jackkuo commited on
Commit
2db8c90
·
verified ·
1 Parent(s): 9e3368a

Add files using upload-large-folder tool

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. -dFIT4oBgHgl3EQf9CsF/content/2301.11404v1.pdf +3 -0
  2. -dFIT4oBgHgl3EQf9CsF/vector_store/index.faiss +3 -0
  3. -dFIT4oBgHgl3EQf9CsF/vector_store/index.pkl +3 -0
  4. .gitattributes +61 -0
  5. 0tE2T4oBgHgl3EQfiQfL/content/2301.03957v1.pdf +3 -0
  6. 0tE2T4oBgHgl3EQfiQfL/vector_store/index.pkl +3 -0
  7. 1NAzT4oBgHgl3EQf8v6L/content/tmp_files/2301.01909v1.pdf.txt +1735 -0
  8. 1NAzT4oBgHgl3EQf8v6L/content/tmp_files/load_file.txt +0 -0
  9. 1tE3T4oBgHgl3EQfngpH/content/2301.04625v1.pdf +3 -0
  10. 1tE3T4oBgHgl3EQfngpH/vector_store/index.faiss +3 -0
  11. 1tE3T4oBgHgl3EQfngpH/vector_store/index.pkl +3 -0
  12. 59E4T4oBgHgl3EQfcAxC/vector_store/index.faiss +3 -0
  13. 69E4T4oBgHgl3EQfcgw0/content/tmp_files/2301.05083v1.pdf.txt +1509 -0
  14. 69E4T4oBgHgl3EQfcgw0/content/tmp_files/load_file.txt +0 -0
  15. 7NA0T4oBgHgl3EQfOP_L/content/tmp_files/2301.02159v1.pdf.txt +3207 -0
  16. 7NA0T4oBgHgl3EQfOP_L/content/tmp_files/load_file.txt +0 -0
  17. 7NE1T4oBgHgl3EQfBgKg/content/tmp_files/2301.02853v1.pdf.txt +1502 -0
  18. 7NE1T4oBgHgl3EQfBgKg/content/tmp_files/load_file.txt +0 -0
  19. 7NE2T4oBgHgl3EQfkwf2/vector_store/index.faiss +3 -0
  20. 99E4T4oBgHgl3EQf3g3K/content/tmp_files/2301.05307v1.pdf.txt +0 -0
  21. 99E4T4oBgHgl3EQf3g3K/content/tmp_files/load_file.txt +0 -0
  22. 9dAzT4oBgHgl3EQfSfu4/content/2301.01234v1.pdf +3 -0
  23. 9dAzT4oBgHgl3EQfSfu4/vector_store/index.faiss +3 -0
  24. ENE3T4oBgHgl3EQfVQoB/content/2301.04457v1.pdf +3 -0
  25. ENE3T4oBgHgl3EQfVQoB/vector_store/index.faiss +3 -0
  26. ENE3T4oBgHgl3EQfVQoB/vector_store/index.pkl +3 -0
  27. EdAzT4oBgHgl3EQfUPw6/content/tmp_files/2301.01262v1.pdf.txt +1967 -0
  28. EdAzT4oBgHgl3EQfUPw6/content/tmp_files/load_file.txt +0 -0
  29. EdFRT4oBgHgl3EQfBDfd/vector_store/index.pkl +3 -0
  30. GdFJT4oBgHgl3EQfECxK/content/tmp_files/2301.11436v1.pdf.txt +258 -0
  31. GdFJT4oBgHgl3EQfECxK/content/tmp_files/load_file.txt +205 -0
  32. GtAyT4oBgHgl3EQfrfmY/content/tmp_files/2301.00562v1.pdf.txt +1507 -0
  33. GtAyT4oBgHgl3EQfrfmY/content/tmp_files/load_file.txt +0 -0
  34. H9FJT4oBgHgl3EQfuy1Y/content/tmp_files/2301.11623v1.pdf.txt +1297 -0
  35. H9FJT4oBgHgl3EQfuy1Y/content/tmp_files/load_file.txt +0 -0
  36. HtE1T4oBgHgl3EQfrgVp/content/tmp_files/2301.03355v1.pdf.txt +1489 -0
  37. HtE1T4oBgHgl3EQfrgVp/content/tmp_files/load_file.txt +0 -0
  38. I9AzT4oBgHgl3EQfH_tx/vector_store/index.pkl +3 -0
  39. JdFIT4oBgHgl3EQfZSuO/content/2301.11252v1.pdf +3 -0
  40. JdFIT4oBgHgl3EQfZSuO/vector_store/index.faiss +3 -0
  41. JdFIT4oBgHgl3EQfZSuO/vector_store/index.pkl +3 -0
  42. KNE2T4oBgHgl3EQfpghg/content/tmp_files/2301.04029v1.pdf.txt +0 -0
  43. KNE2T4oBgHgl3EQfpghg/content/tmp_files/load_file.txt +0 -0
  44. KdFQT4oBgHgl3EQfTTZa/content/tmp_files/2301.13292v1.pdf.txt +632 -0
  45. KdFQT4oBgHgl3EQfTTZa/content/tmp_files/load_file.txt +335 -0
  46. KtA0T4oBgHgl3EQfCv_p/content/2301.01995v1.pdf +3 -0
  47. KtAzT4oBgHgl3EQfyf5z/content/tmp_files/2301.01754v1.pdf.txt +1195 -0
  48. KtAzT4oBgHgl3EQfyf5z/content/tmp_files/load_file.txt +0 -0
  49. LNE3T4oBgHgl3EQfvwtG/content/2301.04696v1.pdf +3 -0
  50. LNE3T4oBgHgl3EQfvwtG/vector_store/index.faiss +3 -0
-dFIT4oBgHgl3EQf9CsF/content/2301.11404v1.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7bf91ca401ede7dacee01159a7a46a08c9474f55328254227fbd40bafeb3379b
3
+ size 2507961
-dFIT4oBgHgl3EQf9CsF/vector_store/index.faiss ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1fe09f4336b90997cca6f8475c69b1272b9ef1e9e7056df10e0bf26261543ed1
3
+ size 5111853
-dFIT4oBgHgl3EQf9CsF/vector_store/index.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6ddfdb0b9346b6854453b9ca2ca1be729b769837ceaf359af71e6c2b34a81889
3
+ size 166735
.gitattributes CHANGED
@@ -6031,3 +6031,64 @@ dNFST4oBgHgl3EQfEjgF/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -tex
6031
  bNE2T4oBgHgl3EQfaAeA/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6032
  M9AyT4oBgHgl3EQfUPcx/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6033
  r9E4T4oBgHgl3EQfVQyb/content/2301.05023v1.pdf filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6031
  bNE2T4oBgHgl3EQfaAeA/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6032
  M9AyT4oBgHgl3EQfUPcx/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6033
  r9E4T4oBgHgl3EQfVQyb/content/2301.05023v1.pdf filter=lfs diff=lfs merge=lfs -text
6034
+ ptFKT4oBgHgl3EQfyy7U/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6035
+ htE1T4oBgHgl3EQfzgUS/content/2301.03444v1.pdf filter=lfs diff=lfs merge=lfs -text
6036
+ cdE_T4oBgHgl3EQf0Bzf/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6037
+ rNFAT4oBgHgl3EQffh2A/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6038
+ hdAzT4oBgHgl3EQfa_yH/content/2301.01378v1.pdf filter=lfs diff=lfs merge=lfs -text
6039
+ ENE3T4oBgHgl3EQfVQoB/content/2301.04457v1.pdf filter=lfs diff=lfs merge=lfs -text
6040
+ TdE2T4oBgHgl3EQftAiy/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6041
+ 9dAzT4oBgHgl3EQfSfu4/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6042
+ hdAzT4oBgHgl3EQfa_yH/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6043
+ dtE1T4oBgHgl3EQfeASf/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6044
+ 9dAzT4oBgHgl3EQfSfu4/content/2301.01234v1.pdf filter=lfs diff=lfs merge=lfs -text
6045
+ kdAyT4oBgHgl3EQfkvi4/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6046
+ hNE2T4oBgHgl3EQfHgYi/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6047
+ wdE5T4oBgHgl3EQfMg4L/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6048
+ ENE3T4oBgHgl3EQfVQoB/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6049
+ O9E2T4oBgHgl3EQfrQi-/content/2301.04048v1.pdf filter=lfs diff=lfs merge=lfs -text
6050
+ KtA0T4oBgHgl3EQfCv_p/content/2301.01995v1.pdf filter=lfs diff=lfs merge=lfs -text
6051
+ -dFIT4oBgHgl3EQf9CsF/content/2301.11404v1.pdf filter=lfs diff=lfs merge=lfs -text
6052
+ mdE3T4oBgHgl3EQf6gtE/content/2301.04790v1.pdf filter=lfs diff=lfs merge=lfs -text
6053
+ pdE3T4oBgHgl3EQfLgnu/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6054
+ gtAyT4oBgHgl3EQfxfkk/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6055
+ h9AyT4oBgHgl3EQfxvlP/content/2301.00671v1.pdf filter=lfs diff=lfs merge=lfs -text
6056
+ mdE3T4oBgHgl3EQf6gtE/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6057
+ otE1T4oBgHgl3EQfiAQh/content/2301.03246v1.pdf filter=lfs diff=lfs merge=lfs -text
6058
+ O9E2T4oBgHgl3EQfrQi-/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6059
+ 7NE2T4oBgHgl3EQfkwf2/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6060
+ zdFQT4oBgHgl3EQfzDZw/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6061
+ 1tE3T4oBgHgl3EQfngpH/content/2301.04625v1.pdf filter=lfs diff=lfs merge=lfs -text
6062
+ _9AzT4oBgHgl3EQfvv1V/content/2301.01711v1.pdf filter=lfs diff=lfs merge=lfs -text
6063
+ 1tE3T4oBgHgl3EQfngpH/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6064
+ 59E4T4oBgHgl3EQfcAxC/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6065
+ n9A0T4oBgHgl3EQfJ_86/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6066
+ d9AzT4oBgHgl3EQf3f6y/content/2301.01831v1.pdf filter=lfs diff=lfs merge=lfs -text
6067
+ JdFIT4oBgHgl3EQfZSuO/content/2301.11252v1.pdf filter=lfs diff=lfs merge=lfs -text
6068
+ YNFJT4oBgHgl3EQf6S0s/content/2301.11673v1.pdf filter=lfs diff=lfs merge=lfs -text
6069
+ h9AyT4oBgHgl3EQfxvlP/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6070
+ r9FKT4oBgHgl3EQf1C6f/content/2301.11918v1.pdf filter=lfs diff=lfs merge=lfs -text
6071
+ ytAzT4oBgHgl3EQf7_7g/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6072
+ LNE3T4oBgHgl3EQfvwtG/content/2301.04696v1.pdf filter=lfs diff=lfs merge=lfs -text
6073
+ 0tE2T4oBgHgl3EQfiQfL/content/2301.03957v1.pdf filter=lfs diff=lfs merge=lfs -text
6074
+ Z9E2T4oBgHgl3EQfEwaS/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6075
+ NtFJT4oBgHgl3EQfHCyo/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6076
+ rNFAT4oBgHgl3EQffh2A/content/2301.08582v1.pdf filter=lfs diff=lfs merge=lfs -text
6077
+ wdE5T4oBgHgl3EQfMg4L/content/2301.05481v1.pdf filter=lfs diff=lfs merge=lfs -text
6078
+ rNA0T4oBgHgl3EQfKv_4/content/2301.02110v1.pdf filter=lfs diff=lfs merge=lfs -text
6079
+ qtE3T4oBgHgl3EQfMAny/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6080
+ JdFIT4oBgHgl3EQfZSuO/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6081
+ qtE3T4oBgHgl3EQfMAny/content/2301.04370v1.pdf filter=lfs diff=lfs merge=lfs -text
6082
+ htE1T4oBgHgl3EQfzgUS/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6083
+ VdE4T4oBgHgl3EQfMgyS/content/2301.04948v1.pdf filter=lfs diff=lfs merge=lfs -text
6084
+ S9E4T4oBgHgl3EQfLgwd/content/2301.04938v1.pdf filter=lfs diff=lfs merge=lfs -text
6085
+ QtAzT4oBgHgl3EQfz_6d/content/2301.01778v1.pdf filter=lfs diff=lfs merge=lfs -text
6086
+ -dFIT4oBgHgl3EQf9CsF/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6087
+ atAzT4oBgHgl3EQfZfwz/content/2301.01351v1.pdf filter=lfs diff=lfs merge=lfs -text
6088
+ LNE3T4oBgHgl3EQfvwtG/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6089
+ qNE3T4oBgHgl3EQf8QsZ/content/2301.04805v1.pdf filter=lfs diff=lfs merge=lfs -text
6090
+ fNE0T4oBgHgl3EQfowEk/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6091
+ rNA0T4oBgHgl3EQfKv_4/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6092
+ TNE4T4oBgHgl3EQfmA2x/content/2301.05165v1.pdf filter=lfs diff=lfs merge=lfs -text
6093
+ VdE4T4oBgHgl3EQfMgyS/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text
6094
+ ztAzT4oBgHgl3EQfQ_vk/content/2301.01210v1.pdf filter=lfs diff=lfs merge=lfs -text
0tE2T4oBgHgl3EQfiQfL/content/2301.03957v1.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0976e350c2f52010eae8683d323a0bf3ed459c4de2e42ffce4bd27a40ff4350e
3
+ size 585901
0tE2T4oBgHgl3EQfiQfL/vector_store/index.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:370634a476fc2300313669acd19248c77faeb44b9dc578f84f06c69457ad8d17
3
+ size 90074
1NAzT4oBgHgl3EQf8v6L/content/tmp_files/2301.01909v1.pdf.txt ADDED
@@ -0,0 +1,1735 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ arXiv:2301.01909v1 [math-ph] 5 Jan 2023
2
+ Solid-solid phase transitions in the ‘near-liquid’ limit
3
+ Yury Grabovsky∗
4
+ Lev Truskinovsky†
5
+ January 6, 2023
6
+ Abstract
7
+ In this paper, dedicated to the memory of J. Ericksen, we address the fundamental
8
+ difference between solid-solid and liquid-liquid phase transitions while remaining within
9
+ the Ericksen’s nonlinear elasticity paradigm. To this end we assume that rigidity is
10
+ weak and explore the nature of solid-solid phase transitions in a ‘near-liquid’ limit. In
11
+ the language of calculus of variations we probe limits of quasiconvexity in an ’almost
12
+ liquid’ solid by comparing the thresholds for cooperative (laminate based) and non-
13
+ cooperative (inclusion based) nucleation. We consider a 2D problem and work with a
14
+ prototypical two-phase Hadamard material. Using these two types of nucleation tests
15
+ we obtain for this material surprisingly tight two-sided bounds on the elastic binodal
16
+ without computing the quasi-convex envelope.
17
+ 1
18
+ Introduction
19
+ In 1975 J. Ericksen posed the problem of equilibrium for solids undergoing first order phase
20
+ transitions in the framework of nonlinear elasticity theory. In this way he effectively reformu-
21
+ lated the classical problem of physics into a problem of vectorial calculus of variations. The
22
+ contemporaneous physical theory viewed non-hydrostatically stressed solids as metastable
23
+ and therefore did not distinguish between solid-solid and liquid-liquid phase transitions. J.
24
+ Ericksen realized that at normal conditions the assumption of complete relaxation of non-
25
+ hydrostatic stresses is impractical and his pioneering research program of studying materials
26
+ with non-rank-one convex energies revolutionized elasticity theory. The goal of this paper is
27
+ to elucidate the difference between solid-solid and liquid-liquid phase transitions within the
28
+ Ericksen’s nonlinear elasticity paradigm.
29
+ From the perspective of elasticity theory, the main difference between liquids and solids
30
+ is that liquids do not resist shear [6, 10]. This degeneracy in the elastic constitutive structure
31
+ of liquids is responsible for their peculiar behavior during first order phase transitions vis a
32
+ vis the behavior of solids, characterized by finite rigidity [17]. While in both cases reaching
33
+ phase equilibrium usually leads to the formation of phase mixtures, in the case of solids
34
+ ∗Department of Mathematics, Temple University, Philadelphia, PA 19122, USA
35
+ †PMMH, CNRS – UMR 7636, ESPCI, PSL, 75005 Paris, France
36
+ 1
37
+
38
+ the knowledge of phase fractions carries considerably more information about the geometry
39
+ of the resulting microstructure than in the case of liquids. More specifically, if the phase
40
+ organization in liquid phase transitions is largely controlled by surface tension, in solid phase
41
+ transitions the dominance of elastic long-range interactions leaves to surface tension only a
42
+ minor role of a scale selection.
43
+ First order phase transitions in liquids are well understood at both physical and mathe-
44
+ matical level [32, 9]. The reason is that the scalar problem confronted in the liquid case is
45
+ fully solvable [7]. Instead, despite many dedicated efforts, largely inspired by the pioneering
46
+ contributions of J. Ericksen himself [12, 11, 13, 14, 15], the mathematical understanding of
47
+ elastic phase transitions in solids is still far from being complete as the underlying nonconvex
48
+ vectorial problems of the calculus of variations remain highly challenging.
49
+ To set the stage, we recall that in nonlinear elasticity the energy functional can be written
50
+ in the form E[y] =
51
+
52
+ Ω W(F )dx, where F = ∇y and y : Ω → Rn is the deformation. For
53
+ the energy minimizing configurations the conventional physically informed energy density
54
+ W(F ) can be replaced by a relaxed one QW(F ) = infφ∈C∞
55
+ 0 (D;Rn) |D|−1 �
56
+ D W(F + ∇φ)dx
57
+ which is known as quasiconvexification of W(F ) [8]. To construct the function QW(F )
58
+ one must know the energy minimizing phase microstructures.
59
+ In the case of liquids the
60
+ geometry of such microstructures is irrelevant and the construction of QW(F ) reduces to
61
+ convexification. In solids the task of finding the equilibrium microstructures in a generic
62
+ case is hardly tractable [4, 1].
63
+ With the aim of building a bridge between elastic phase transitions in liquids and solids,
64
+ we consider a special limit of ‘near-liquid’ solids which are characterized by an arbitrarily
65
+ weak resistance to shear. While we pose the general question of how in such a limit the tight
66
+ control on the geometry of optimal microstructures by elastic interactions is progressively
67
+ lost, we address a simpler problem of describing in this limit the boundary of the set of
68
+ stable single-phase configurations. In the case of liquid-liquid phase transitions the incipient
69
+ microstructures do not have any special features. The problem also simplifies in the case
70
+ of ‘strongly-solid’ elastic phase transitions when the equilibrium microstructures are just
71
+ simple laminates [24]. The goal of the present paper is to understand the opposite, ‘weakly-
72
+ solid’ limit, when some of the simplest laminate-based microstructures are proved to be
73
+ suboptimal.
74
+ In the physics of phase transformations, the Maxwell-Gibbs critical/equilibrium condi-
75
+ tions [34, 16], defining the incipient transitions in liquids, are designed to account for the
76
+ possibility of phase nucleation. In other words, their role is to delimit the homogeneous
77
+ configurations that are unstable to perturbations that are small only in extent and the set of
78
+ such configurations is known in physics as the binodal region [36]. From the perspective of
79
+ the mathematical theory of elastic phase transitions the analog of the binodal region would
80
+ incorporate the homogeneous states that fail to be strong minima of the energy functional.
81
+ Therefore, the binodal region is a subset in the configurational space of strain measures
82
+ where the quasi convex envelope lays below the energy density. Locating the boundaries
83
+ of the binodal region (known simply as a binodal) in the ’near-liquid’ limit constitutes the
84
+ main task of the present paper. While remaining nontrivial, this task appears, a priori as
85
+ more tractable than the task of constructing the actual quasiconvex envelope.
86
+ In our prior work we have developed a general method for identifying the subsets of the
87
+ 2
88
+
89
+ d1
90
+ d2
91
+ d
92
+ h(d)
93
+ Figure 1: Double-well structure of the energy density h.
94
+ binodal supporting the laminate type energy minimizing configurations [18, 23, 25]. Behind
95
+ this method is the study of stability of the jump set—a codimension one variety in the phase
96
+ space that has a dual nature. On the one hand it determines the set of pairs F± that could
97
+ be the traces of the deformation gradient at the phase boundary in a stable configuration.
98
+ On the other, the jump set consists of points that are at most marginally stable in the sense
99
+ that their every neighborhood contains points where quasiconvexity fails. Therefore, if one
100
+ can prove quasiconvexity at a point on the jump set, then this point must lie on the binodal.
101
+ In addition, we have also developed tools to constrain the location of the binodal by means
102
+ of addressing nucleation phenomenon directly [19]. As we show in this paper, combined
103
+ together, these two types of approaches can produce in the ’near-liquid’ limit a rather good
104
+ practical understanding of the whole structure of the binodal, and even allow one to obtain
105
+ the exact formulas for the quasiconvex envelope.
106
+ To highlight ideas we focus here only on the simplest family of non-quasiconvex energy
107
+ densities known as Hadamard materials [27, 28]: W(F ) = µ
108
+ 2|F |2 + h(det F ). Specifically,
109
+ we’ll be interested in the case of two space dimensions and assume that the function h(d)
110
+ describes a generic double-well potential modeling isotropic-to-isotropic phase transitions
111
+ (see Fig. 1). The main advantage of this class of elastic materials is that one can identify
112
+ a single parameter µ, scaling the effective rigidity; by varying this parameter we can study
113
+ the entire range of intermediate rigidity responses from ’strong’ (µ ≫ 1) to ’weak’ (µ ≪ 1).
114
+ A notable feature of the Hadamard materials is that the phase with the larger value of det F
115
+ (smaller density) is characterized by a larger tangential (effective) rigidity than the phase
116
+ with the smaller value of det F (larger density). As a result, the latter is more ’liquid-like’
117
+ than the former and therefore the incipient phase transformation induced by compression can
118
+ be expected to be different from the incipient phase transformation induced by stretching.
119
+ As we show in what follows, this asymmetry leads to a coexistence of ’strongly-solid’ and
120
+ ’weakly-solid’ responses inside a single material model as, even in the absence of hysteresis,
121
+ the direct and reverse solid-solid phase transitions proceed according to morphologically
122
+ different transformation mechanisms.
123
+ While for an Hadamard material the double well energy structure is described by the
124
+ simplest scalar potential h(d), the results of relaxation of W(F ) are nontrivial due to the
125
+ inherent incompatibility of the energy wells [3].
126
+ We recall that W(F ) is quasiconvex if
127
+ and only if h(d) is convex [2]. The relaxation of W(F ) with non-convex h(d) is known for
128
+ 3
129
+
130
+ the ‘infinitely-weak’ solids (effectively fluids) with µ = 0, where QW(F ) = h∗∗(det F ) [7].
131
+ Previously we explicitly constructed the quasiconvex envelope for W(F ) in the ‘strongly-
132
+ solid’ limit assuming that the shear modulus µ is sufficiently large and the corresponding
133
+ quadratic term dominates the double-well term. In this case the formula for QW(F ) couples
134
+ |F | and det F and the relaxed energy is sandwiched between W(F ) above and U(F ) =
135
+ µ
136
+ 2|F |2 + h∗∗(det F ) [24].
137
+ In this paper we show that the constraint on µ in [24] was not a technical limitation,
138
+ and that, as µ decreases, our formula for QW(F ) ceases to be valid in the subsets of the
139
+ binodal region close to the ’liquid-like’ phase with smaller rigidity. In the limit of small µ,
140
+ we show that the relaxation of W(F ) goes through a chain of structural transitions with
141
+ simple lamination persisting only in the vicinity of the pure ’solid-like’ phase, being replaced
142
+ by very different phase arrangements close to the ’liquid-like’ phase.
143
+ Our main technical approach is to generate bounds on the binodal surface.
144
+ The simplest bounds is obtained by probing the binodal by means of nucleating first
145
+ rank laminates. Their optimality is proved by establishing their polyconvexity (and therefore
146
+ quasi-convexity). In this setting this is an algebraic problem, because the supporting null-
147
+ Lagrangians can be constructed explicitly, [25]. In contrast with the strongly solid regime of
148
+ large µ analyzed in [24], in the near liquid regime of small µ, not all of the first rank laminate
149
+ bounds are optimal.
150
+ These bounds are then improved by nucleating second rank laminates. However, as shown
151
+ in [26], the second rank laminate bounds are not optimal either, and are further improved
152
+ for hydrostatic strains by means of nucleating a bounded circular inclusion in the infinite
153
+ plane. We conjecture that this bound is optimal. If our conjecture is true, then the values
154
+ of the deformation gradient in the exterior of the circular nucleus would provide a bound
155
+ on the binodal from the outside of the binodal region. Another consequence of the assumed
156
+ optimality of the inclusion-based nucleation bound is the explicit formula for the quasiconvex
157
+ envelope QW(F ) at all hydrostatic strains.
158
+ By juxtaposing the hypothetical bound provided by the study of bounded inclusions and
159
+ unbounded second rank laminates we derive tight two-sided bounds on the binodal. As we
160
+ demonstrate in [20], both bounds remain tight in the full range of parameters for which the
161
+ bounds are meaningful. Moreover, the hypothetical bound being in complete agreement with
162
+ the numerically computed rank-one convex binodal.
163
+ The paper is organized as follows. In Section 2 we recall some general results from the
164
+ calculus of variations for nonconvex vectorial problems, used in the rest of the paper. In
165
+ Section 3 we specialize these results for the Hadamard material and present the numerical
166
+ illustrations of the obtained bounds. Analytical results for the limiting case µ → 0 are pre-
167
+ sented in Section 4 where we also compare them with numerical computations. In Section 5
168
+ we demonstrate the far reaching consequences of the assumed optimality of the nucleation
169
+ bound. The paper ends with a general discussion and conclusions in Section 6.
170
+ 4
171
+
172
+ 2
173
+ Preliminaries
174
+ Binodal region. Hyperelastic materials in a d-dimensional space have the following form of
175
+ the energy stored in the deformed elastic body
176
+ E[y] =
177
+
178
+
179
+ W(∇y(x))dx,
180
+ where Ω ⊂ Rd is the reference configuration, and y : Ω → Rd is the deformation.
181
+ In
182
+ order to understand the stable (i.e. experimentally observable) configurations of the body
183
+ it is often necessary to replace the energy density W(F ) with a relaxed one QW(F ), called
184
+ quasiconvexification. Even though, there is a formula for QW(F ) [8]:
185
+ QW(F ) =
186
+ inf
187
+ φ∈C∞
188
+ 0 (D;Rn)
189
+ 1
190
+ |D|
191
+
192
+ D
193
+ W(F + ∇φ)dx,
194
+ (2.1)
195
+ there is no systematic approaches to compute it. A simpler, but just as useful an object, is
196
+ the elastic binodal.
197
+ Definition 2.1. An elastic binodal is the boundary of the binodal region
198
+ B = {F : W(F ) < QW(F )}.
199
+ (2.2)
200
+ Definition 2.2. We say that the matrix F is stable, if W(F ) = QW(F ).
201
+ Thus, the binodal is the boundary separating the binodal region from the set of stable
202
+ points.
203
+ Jump set. While we acknowledge that there could be rank-one convex, non quasiconvex
204
+ functions, most cases of practical interest in elastic phase transitions feature multiwell ener-
205
+ gies that are not rank-one convex and possess a non-trivial jump set, stable points of which
206
+ form a part of the binodal (or the entire binodal, if one is lucky). The jump set is the set of
207
+ solutions F = F− of the equations
208
+
209
+
210
+
211
+
212
+
213
+
214
+
215
+
216
+
217
+ F+ = F− + a ⊗ n,
218
+ [[P ]]n = 0,
219
+ [[P T ]]a = 0,
220
+ [[W]] − ⟨{{P }}, [[F ]]⟩ = 0,
221
+ (2.3)
222
+ where a ̸= 0 and |n| = 1 are thought to be excluded from the above system resulting in a
223
+ single scalar equation for F . We refer the reader to [26] for a discussion of the geometry of
224
+ the solution set of (2.3). Here we used the standard notations
225
+ P± = WF (F±),
226
+ [[F ]] = F+ − F−,
227
+ {{P }} = P+ + P−
228
+ 2
229
+ ,
230
+ ⟨A, B⟩ = Tr (ABT),
231
+ where WF indicates the matrix of partial derivatives Pij = ∂W/∂Fij.
232
+ 5
233
+
234
+ The points on the jump set belong either to the binodal or to the binodal region B, [18].
235
+ Hence, the jump set always represents a bound on the binodal region from within. One of
236
+ the easy ways to detect the unstable parts of the jump set is to use the Weierstrass condition,
237
+ which is necessary for stability.
238
+ W ◦(F , b ⊗ m) ≥ 0,
239
+ ∀b ∈ Rn, |m| = 1,
240
+ (2.4)
241
+ where
242
+ W ◦(F , H) = W(F + H) − W(F ) − ⟨WF (F ), H⟩.
243
+ We have proved in [22] that the pairs of points F± on the jump set are either both stable
244
+ or both unstable. Hence, a point F+ satisfying (2.4) can be still classified as unstable, if F−
245
+ fails (2.4). While there are other conditions of stability that don’t follow from (2.4) (see [23])
246
+ we will only make use of an easily verifiable corollary of(2.4) that restricts the rank-one test
247
+ fields b ⊗ m to an infinitesimally small neighborhood of [[F ]] = a ⊗ n.
248
+ Currently, the only general tool for establishing stability is polyconvexity, which is suf-
249
+ ficient but rather far from necessary. In two dimensions it reduces to finding a constant
250
+ m ∈ R, such that
251
+ W ◦(F , H) − m det H ≥ 0,
252
+ ∀H ∈ R2×2.
253
+ (2.5)
254
+ If (2.5) holds, then F is stable in the sense of Definition 2.2. For points F± on the jump set,
255
+ however, the only value of m that could possibly work is, as shown in [25],
256
+ m = ⟨[[P ]], cof[[F ]]⟩
257
+ |[[F ]]|2
258
+ .
259
+ (2.6)
260
+ Secondary jump set. An improved bound on the binodal is provided by the secondary
261
+ jump set corresponding to the nucleation of a rank-two laminate in the infinite homoge-
262
+ neously strained space. Thus, the secondary jump set is defined by the system of equations
263
+
264
+
265
+
266
+
267
+
268
+
269
+
270
+
271
+
272
+
273
+
274
+ F = F + b ⊗ m,
275
+ P m = P m,
276
+ P Tb = P
277
+ Tb,
278
+ W(F ) − W = P m · b,
279
+ (2.7)
280
+ where the pair F±, to be determined, is assumed to satisfy the primary jump set equations
281
+ (2.3), while
282
+ W = λW(F+) + (1 − λ)W(F−),
283
+ P = λP+ + (1 − λ)P−,
284
+ (2.8)
285
+ for some λ ∈ [0, 1], which also plays the role of a variable to be solved for in (2.7), along
286
+ with F , b ̸= 0, and |m| = 1. Once again, the secondary jump set represents a bound on the
287
+ binodal region from within.
288
+ Nucleation criterion. Let us now recall another method of probing the binodal: nucleation
289
+ of inclusions either of a prescribed shape [5, 33, 31] or of an optimal inclusion, whose shape
290
+ must be determined [29, 35, 30]. The theory justifying why these tests probe the binodal
291
+ 6
292
+
293
+ was developed in [19]. In the case of “nucleation of a bounded inclusion”, the criterion for
294
+ F0 to be “marginally stable”, i.e. to lie in the closure of B, is the existence of a field
295
+ φ ∈ S = {φ ∈ L2
296
+ loc(Rd) : ∇φ ∈ L2(Rd; Rd)},
297
+ such that
298
+ ∇ · P (F0 + ∇φ) = 0,
299
+ ∇ · P ∗(F0 + ∇φ) = 0
300
+ (2.9)
301
+ in the sense of distribution in Rd, where
302
+ P (F ) = WF (F ),
303
+ P ∗(F ) = W(F )Id − F TP (F ).
304
+ We also need to verify the non-degeneracy of the solution φ:
305
+
306
+ Rd W ◦
307
+ F (F0, ∇φ)dx ̸= 0.
308
+ (2.10)
309
+ In the case of nucleation of an actual inclusion ω with smooth boundary the verification of
310
+ (2.9) consists in verifying that the field φ ∈ S solves ∇ · P (F0 + ∇φ) = 0 both inside and
311
+ outside of ω, together with the condition that the traces F±(x) = F0 + ∇φ±(x) on the two
312
+ sides of ∂ω form a corresponding pair on the jump set for each x ∈ ∂ω. If, in addition, we
313
+ can somehow prove that F + ∇φ(x) is stable in the sense of Definition 2.2, for each x ∈ Rd,
314
+ then F0 must lie on the binodal. Conversely, if it is known that that at some x0 ∈ Rd the
315
+ matrix F0 + ∇φ(x0) is unstable, then F0 must lie in the interior of B.
316
+ 3
317
+ Hadamard material
318
+ In this paper we focus our attention on a particularly simple, yet nontrivial energy
319
+ W(F ) = µ
320
+ 2|F |2 + h(d),
321
+ F ∈ {F ∈ GL(n) : det F > 0},
322
+ d = det F ,
323
+ (3.1)
324
+ where h(d) is a C2(0, +∞) function with a double-well shape. In our explicit computations
325
+ and illustrations we use the quartic double-well energy1
326
+ h(d) = (d − d1)2(d − d2)2,
327
+ (3.2)
328
+ which affords certain simplification of general formulas.
329
+ Jump set. We recall (see [24]) that in two dimensions the jump set of (3.1) consists of
330
+ matrices F±, whose two singular values labelled ε0 and ε± satisfy the equations
331
+ ε0[[h′]] + µ[[ε]] = 0,
332
+ [[h]] − {{h′}}[[d]] = 0,
333
+ d± = det F± = ε0ε±.
334
+ (3.3)
335
+ The notation reflects that for each pair F± on the jump set there is a frame in which
336
+ both matrices are diagonal and share the same singular value ε0 with the same eigenvector.
337
+ 1Formula (3.2) only needs to hold in an arbitrary neighborhoodof [d1, d2]. The potential h(d) can be
338
+ modified outside of that neighborhood arbitrarily, as long as h∗∗(d) = h(d) there. In particular, the singular
339
+ behavior of h(d) as d → 0+, required in nonlinear elasticity, can be easily assured.
340
+ 7
341
+
342
+ Equations (3.3) can be used to derive the semi-explicit parametric equations of the jump set,
343
+ where, say d+ = ε0ε+, can serve as a parameter. Given d+ we can use the second equation in
344
+ (3.3) to compute d− = D(d+). Then, multiplying the first equation in (3.3) by ε0 we obtain
345
+ the parametric equations
346
+
347
+
348
+
349
+ ε0(d+) =
350
+
351
+ −µ[[d]]
352
+ [[h′]] ,
353
+ ε+(d+) =
354
+ d+
355
+ ε0(d+).
356
+ In the case of potential (3.2) we obtain
357
+ [[h]] − {{h′}}[[d]] = [[d]]3(d1 + d2 − d+ − d−).
358
+ Hence, d− = d1 + d2 − d+ = D(d+). It follows that
359
+ {{h′}} = 0,
360
+ ε+ + ε− = d1 + d2
361
+ ε0
362
+ .
363
+ (3.4)
364
+ In particular, we can eliminate h′(d±) from our formulas by means of (3.3) and (3.4):
365
+ h′(d±) = {{h′}} ± 1
366
+ 2[[h′]] = ∓µ
367
+ 2
368
+ [[ε]]
369
+ ε0
370
+ .
371
+ (3.5)
372
+ For quartic energy (3.2) we can also write the equation of the jump set explicitly as
373
+ ε± = ε±(ε0). Indeed, ε± = d±/ε0, while d± solves
374
+ (d± − d1)(d± − d2) = − µ
375
+ 4ε2
376
+ 0
377
+ (3.6)
378
+ The two roots of (3.6) are the values of d±, where, by convention, we denote by d+ the
379
+ larger root. Equation (3.6) has exactly two real roots whenever ε0 > √µ/(d2 − d1). Hence,
380
+ explicitly,
381
+ ε± =
382
+ 1
383
+ 2ε0
384
+
385
+ d1 + d2 ±
386
+
387
+ (d2 − d1)2 − µ
388
+ ε2
389
+ 0
390
+
391
+ .
392
+ (3.7)
393
+ In our calculations we will use equations (3.5) to eliminate all occurrences of h′(d±) and
394
+ equations (3.7) to eliminate ε±, since the pair ε± is uniquely determined by a single parameter
395
+ ε0.
396
+ Numerical illustrations. When µ is large we have shown in [24] that the jump set [18]
397
+ comprises the entire binodal, each point of which corresponds to the nucleation of a simple
398
+ laminate, leading to an explicit formula for the relaxation QW(F ). As the shear modulus
399
+ µ decreases, parts of the jump set will become unstable. The jump set will then undergo
400
+ a topological change at µ = µtop and in the limit µ → 0, which is the main focus of this
401
+ paper, a specific portion of it will remain stable, as we will show using methods from [25].
402
+ Fig. 2 shows the jump sets and indicates their unstable parts for four different values of the
403
+ shear modulus µ. The values of µ in Fig. 2 are chosen to be µ = 0, µtop/3, 0.9µtop, and
404
+ 1.5µtop. Dotted lines indicate “convexification hyperbolas”, i.e., hyperbolas ε2 = d1/ε1 and
405
+ ε2 = d2/ε1, where the interval [d1, d2] is the interval on which h(d) differs from its convex hull.
406
+ 8
407
+
408
+ 1
409
+ 2
410
+ 3
411
+ 1
412
+ 0.5
413
+ 1
414
+ 1.5
415
+ 2
416
+ 2.5
417
+ 3
418
+ 2
419
+ = 0
420
+ 0.5
421
+ 1
422
+ 1.5
423
+ 2
424
+ 2.5
425
+ 1
426
+ 0.5
427
+ 1
428
+ 1.5
429
+ 2
430
+ 2.5
431
+ 2
432
+ = 2.8168
433
+ 0.5
434
+ 1
435
+ 1.5
436
+ 2
437
+ 1
438
+ 0.5
439
+ 1
440
+ 1.5
441
+ 2
442
+ 2
443
+ = 7.6617
444
+ = 12.6757
445
+ 1
446
+ 2
447
+ 3
448
+ 1
449
+ 0.5
450
+ 1
451
+ 1.5
452
+ 2
453
+ 2.5
454
+ 3
455
+ 2
456
+ Figure 2: Jump sets for h(d) given by (3.2) with d1 = 1, d2 = 3, and different values of µ.
457
+ All points outside of the region bounded by the convexification hyperbolas are well-known
458
+ to be stable (see e.g. [9]), since they are obviously polyconvex.
459
+ W-points. In [23] we have shown that the easily computable corollary of the Weierstrass
460
+ condition (2.4) for the energy (3.1) has the form
461
+ ε0 ≥ ε±.
462
+ (3.8)
463
+ In [24] we have shown that this condition is always satisfied for large values of µ as is evident
464
+ from the lower right panel in Fig. 2, while it has an obvious geometric interpretation in the
465
+ two panels in which the part of the jump set failing (3.8) is shown as a dashed line. The
466
+ points marked by red dots in Fig. 2 that delimit the part of the jump set satisfying (3.8)
467
+ will be called the Weierstrass points or W-points, for short. We have shown in [26] that the
468
+ solid portion of the jump set delimited by W-points is polyconvex for all sufficiently small
469
+ µ. As we show below, one can provide an almost explicit characterization of all values of µ
470
+ for which W-points are also points of polyconvexity assuming the quartic nonlinearity (3.2).
471
+ As discussed above, in order to prove the polyconvexity of W-points we need to establish
472
+ (2.5), where m is given by (2.6). This problem has been already analyzed in [24], where we
473
+ showed that (2.5) can be written as Φ(x, y) ≥ Φ(ε±, ε0) for all x, y, where
474
+ Φ(x, y) = µ
475
+ 2(x2 + y2) − αx − βy − h(xy) − mxy,
476
+ 9
477
+
478
+ α = 2√µR{{d}},
479
+ β = µ2 + R4d+d−
480
+ Rõ
481
+ ,
482
+ m = [[h′d]]
483
+ [[d]] ,
484
+ R =
485
+
486
+ −[[h′]]
487
+ [[d]] .
488
+ According to the equations of the jump set (3.3) R = √µ/ε0. Hence, we also have
489
+ α = µ(ε+ + ε−),
490
+ β = µ
491
+
492
+ ε0 + ε+ε−
493
+ ε0
494
+
495
+ ,
496
+ m = {{h′}} − µ{{ε}}
497
+ ε0
498
+ .
499
+ When we minimized Φ(x, y) over all (x, y), such that xy = d we have concluded that the
500
+ minimizer is (d/y, y), where y = y(d) is the largest root of
501
+ y4 − β0y3 + dα0y − d2 = 0,
502
+ α0 = ε+ + ε−,
503
+ β0 = ε0 + ε+ε−
504
+ ε0
505
+ ,
506
+ (3.9)
507
+ while the minimum of Φ(x, y) is achieved at a finite point corresponding to a critical point
508
+ of φ(d) = Φ(d/y(d), y(d)).
509
+ In the special case of W-points we have ε+ = ε0 and therefore α0 = β0 = ε− + ε0. In this
510
+ case equation (3.9) factors
511
+ (y2 − d)(y2 − α0y + d) = 0.
512
+ The largest root is y = 1
513
+ 2(α0 +
514
+
515
+ α2
516
+ 0 − 4d), provided 0 < d ≤ α2
517
+ 0/4. If d > α2
518
+ 0/4, then the
519
+ quartic has only two real roots y = ±
520
+
521
+ d. Thus,
522
+ y(d) =
523
+
524
+ (α0 +
525
+
526
+ α2
527
+ 0 − 4d)/2,
528
+ d ≤ α2
529
+ 0/4,
530
+
531
+ d,
532
+ d > α2
533
+ 0/4.
534
+ In [24] we have also computed
535
+ φ′(d) = µy(d)2 − β0y(d)
536
+ d
537
+ + h′(d) − m.
538
+ In the case of W-points for which β0 = α0 we see that
539
+ y(d)2 − β0y(d)
540
+ d
541
+ = −1
542
+ when d ≤ α2
543
+ 0/4. Hence, any critical points of φ(d) in this regime would have to satisfy
544
+ h′(d) − µ − m = 0.
545
+ One of the solutions is d−, which always satisfies d− ≤ α2
546
+ 0/4. If this equation has 3 solutions,
547
+ the the middle one corresponds to a local maximum of φ(d), while the third d∗ > d+ always
548
+ fails to satisfy d∗ ≤ α2
549
+ 0/4 because d+ = ε2
550
+ 0 > (ε− + ε0)2/4. We conclude that the only critical
551
+ points of φ(d) that need to be checked are the ones that satisfy d > α2
552
+ 0/4, while
553
+ φ′(d) = µ
554
+
555
+ 1 − α0
556
+
557
+ d
558
+
559
+ + h′(d) − m.
560
+ 10
561
+
562
+ Observe that φ′(d) > 0 when d ≥ max(α2
563
+ 0, �d+), where �d+ is the largest root of h′(d) − m.
564
+ Hence we only need to check for critical points in a specific bounded interval. In fact, if h(d)
565
+ is given by (3.2), then it is easy to see that φ′(d) > 0 for all d ≥ α2
566
+ 0. Hence, we only need to
567
+ check for critical points of φ(d) on (α2
568
+ 0/4, α2
569
+ 0). In addition, since {{h′}} = 0 for h(d), given by
570
+ (3.2) we have m = −µ{{ε}}/ε0 = −µα0/(2ε0). Thus, we obtain the following characterization
571
+ of polyconvexity of W-points.
572
+ Theorem 3.1. Let h(d) be given by (3.2), then W-points are polyconvex whenever
573
+ min
574
+ d∈
575
+
576
+ α2
577
+ 0
578
+ 4 ,α2
579
+ 0
580
+
581
+
582
+ h(d) + µ
583
+
584
+ d + α0d
585
+ 2ε0
586
+ − 2α0
587
+
588
+ d
589
+ ��
590
+ = h(ε2
591
+ 0) − µε0
592
+ �ε0
593
+ 2 + 3ε−
594
+ 2
595
+
596
+ .
597
+ (3.10)
598
+ where α0 = ε0 + ε−, with (ε0, ε−), (ε−, ε0), and (ε0, ε0) being the coordinates of W-points.
599
+ The right-hand side in (3.10) is just φ(ε2
600
+ 0), where φ(d) is the function being minimized
601
+ in (3.10). For quartic energy (3.2) we compute the coordinates of W-points by solving
602
+ −4d(d − d1)(d − d2) = µ.
603
+ Then ε2
604
+ 0 is the largest root, and
605
+ ε− = d1 + d2 − ε2
606
+ 0
607
+ ε0
608
+ .
609
+ We can compute the largest value of µ for which (3.10) holds by substituting µ = −4ε2
610
+ 0(ε2
611
+ 0 −
612
+ d1)(ε2
613
+ 0−d2) into (3.10) and regarding ε0 ≤ √d2 as a parameter. When ε0 = √d2, φ(d)−φ(d2)
614
+ is a positive polynomial in x =
615
+
616
+ d. We then seek numerically the largest value of ε0 <
617
+ √d2 for which the polynomial P(x) = (φ(x2) − φ(ε2
618
+ 0))/(x − ε0)2 develops a double root.
619
+ Algebraically this means seeking the largest root ε0 < √d2 of the discriminant (computed in
620
+ Maple). This solution gives the largest value of µ below which the W-points are polyconvex.
621
+ For example, when d1 = 1, d2 = 3, we have polyconvexity of W-points for all µ < 6.35888.
622
+ In this paper we will be interested exclusively in the case when W-points are quasiconvex.
623
+ In Fig. 2 the W-points are polyconvex in the top right panel and unstable in the bottom left
624
+ panel.
625
+ Secondary jump set. The algebraic equations (2.7) describing the secondary jump set can
626
+ generally be solved only numerically. By contrast, when µ is small, the asymptotics of the
627
+ solutions can be computed explicitly, providing an excellent approximation to the computed
628
+ secondary jump set for µ < 3, with d1 = 1, d2 = 3. While the entire secondary jump set is
629
+ unstable [26], we will see that it provides an excellent (inside) bound for the binodal.
630
+ Suppose that F0 lies on the secondary jump set. Then there exists ε±, y and λ ∈ [0, 1],
631
+ such that the pair F0, F , where
632
+ F =
633
+
634
+ ε
635
+ 0
636
+ 0
637
+ ε0
638
+
639
+ ,
640
+ ε = λε+ + (1 − λ)ε−,
641
+ satisfies the jump set equations (2.7). We compute
642
+ P = λP+ + (1 − λ)P− =
643
+
644
+ µε + h′ε0
645
+ 0
646
+ 0
647
+ µε0 + εh′
648
+
649
+ .
650
+ 11
651
+
652
+ We have
653
+ P0 = µF0 + h′(d0)cofF0 = µ
654
+
655
+ ε
656
+ 0
657
+ 0
658
+ ε0
659
+
660
+ + µb ⊗ m + h′(d0)
661
+ ��
662
+ ε0
663
+ 0
664
+ 0
665
+ ε
666
+
667
+ + b⊥ ⊗ m⊥
668
+
669
+ .
670
+ Thus, the second and the third equations in the jump set system (2.7) become
671
+
672
+
673
+
674
+
675
+
676
+
677
+
678
+
679
+
680
+
681
+
682
+
683
+
684
+
685
+
686
+
687
+
688
+
689
+
690
+
691
+  (h′(d0) − h′)ε0
692
+ 0
693
+ 0
694
+ h′(d0)ε − εh′
695
+
696
+  m = −µb,
697
+
698
+  (h′(d0) − h′)ε0
699
+ 0
700
+ 0
701
+ h′(d0)ε − εh′
702
+
703
+  b = −µ|b|2m.
704
+ These equations result in 3 possibilities
705
+ (a) (h′(d0) − h′)ε0 = h′(d0)ε − εh′ = −γ, µb = γm, m ∈ S1
706
+ (b) (h′(d0) − h′)ε0 = −(h′(d0)ε − εh′) = −γ, µb = γI−m, I− =
707
+
708
+ 1
709
+ 0
710
+ 0
711
+ −1
712
+
713
+ , m ∈ S1
714
+ (c) (h′(d0) − h′)ε0 ̸= ±(h′(d0)ε − εh′)
715
+ Possibility (c) implies that F0 must be diagonal, and will be our main focus. In [26] we show
716
+ that possibilities (a) and (b) have no solutions. Let us therefore assume that F± is diagonal
717
+ and has the form
718
+ F± =
719
+
720
+ ε±
721
+ 0
722
+ 0
723
+ ε0
724
+
725
+ .
726
+ This implies that F − F0 = βe2 ⊗ e2. In particular
727
+ F0 =
728
+
729
+ x0
730
+ 0
731
+ 0
732
+ y0
733
+
734
+ ,
735
+ x0 = ε = λε+ + (1 − λ)ε−,
736
+ λ ∈ (0, 1).
737
+ Let us compute the diagonal matrices P± using equations (3.4) and (3.5).
738
+ P 11
739
+ ± = µε± + h′(d±)ε0 = µ{{ε}} = µ(d1 + d2)
740
+ 2ε0
741
+ ,
742
+ P 22
743
+ ± = µε0 + h′(d±)ε± = µ
744
+
745
+ ε0 ∓ [[ε]]ε±
746
+ 2ε0
747
+
748
+ .
749
+ Let us compute the diagonal matrix P0.
750
+ P 11
751
+ 0
752
+ = µx0 + h′(d0)y0 = µε + h′(d0)d0
753
+ ε ,
754
+ P 22
755
+ 0
756
+ = µy0 + h′(d0)x0 = µd0
757
+ ε
758
+ + h′(d0)ε.
759
+ Traction continuity equation (P − P0)e2 = 0 then becomes
760
+ ε0 + [[ε]]
761
+ 2ε0
762
+ (ε− − 2λ{{ε}}) − d0
763
+ ε − h′(d0)
764
+ µ
765
+ ε = 0.
766
+ 12
767
+
768
+ It will be convenient to use ε as a variable in place of λ. Replacing λ above using ε = ε−+λ[[ε]]
769
+ we obtain
770
+ d0
771
+ ε = ε0 + 1
772
+ ε0
773
+ (ε+ε− − {{ε}}ε) − h′(d0)
774
+ µ
775
+ ε.
776
+ (3.11)
777
+ Let us now compute all the terms in the last equation in (2.7).
778
+ W(F0) = µ
779
+ 2 (ε2 + y2
780
+ 0) + h(d0) = µ
781
+ 2
782
+
783
+ ε2 + d2
784
+ 0
785
+ ε2
786
+
787
+ + h(d0).
788
+ Next we compute
789
+ W = W− + λ[[W]] = W− + λµ[[ε]]{{ε}} = W− + µ(ε − ε−){{ε}},
790
+ where [[h]] = −{{h′}}[[d]] = 0 has been used. We compute
791
+ h(d−) = [(d− − d1)(d− ��� d1)]2 =
792
+ µ2
793
+ 16ε4
794
+ 0
795
+ ,
796
+ according to (3.6). Therefore,
797
+ W− = µ
798
+ 2 (ε2
799
+ − + ε2
800
+ 0) + µ2
801
+ 16ε4
802
+ 0
803
+ .
804
+ We then compute F0 − F = (y0 − ε0)e2 ⊗ e2. Therefore
805
+ ⟨P , F0 − F ⟩ = µ
806
+ �d0
807
+ ε − ε0
808
+ � �
809
+ ε0 + 1
810
+ ε0
811
+ (ε+ε− − {{ε}}ε)
812
+
813
+ .
814
+ Finally, the Maxwell equation W(F0) − W = ⟨P , F0 − F ⟩ can be written as
815
+ 1
816
+ 2
817
+
818
+ ε2 + d2
819
+ 0
820
+ ε2
821
+
822
+ +h(d0)
823
+ µ
824
+ −(ε−ε−){{ε}}−1
825
+ 2(ε2
826
+ −+ε2
827
+ 0)− µ
828
+ 16ε4
829
+ 0
830
+ =
831
+ �d0
832
+ ε − ε0
833
+ � �
834
+ ε0 + 1
835
+ ε0
836
+ (ε+ε− − {{ε}}ε)
837
+
838
+ .
839
+ (3.12)
840
+ Next we replace in the above form of the Maxwell relation the expression d0/ε by its expres-
841
+ sion from (3.11) As a result of such a substitution the Maxwell relation will also become a
842
+ quadratic equation in ε. This permits us to eliminate this variable as a rational expression in
843
+ terms of ε0 and d0, while the Maxwell relation will also reduce to a rational relation between
844
+ d0 and ε0. This calculation can only be done with the aid of a computer algebra system,
845
+ since the remaining equation F(ε0, d0) = 0 is very long and complicated. Now for a given
846
+ choice of numerical values of µ, d1 and d2 we can solve F(ε0, d0) = 0 numerically and then
847
+ extract those solutions which satisfy λ ∈ [0, 1]. The result for d1 = 1, d2 = 2 and µ = µtop/3
848
+ is shown as a green curve in Fig. 3. As we can see, it identifies all points between the green
849
+ curve and the dashed lines of the primary jump set as a part of the binodal region—an
850
+ improvement over the primary jump set bound.
851
+ While, it is not apparent from Fig. 3, the secondary jump set consists of two curves
852
+ related by symmetry with respect to the bisector of the first quadrant. Each of the curves
853
+ 13
854
+
855
+ 0.5
856
+ 1
857
+ 1.5
858
+ 2
859
+ 2.5
860
+ 1
861
+ 0.5
862
+ 1
863
+ 1.5
864
+ 2
865
+ 2.5
866
+ 2
867
+ = 2.8168
868
+ Figure 3: Secondary jump set computed by numerically solving equations (3.11) and (3.12).
869
+ are cut-off at their intersection with each other at the bisector. Each curve starts at a W-
870
+ point and ends at a point (not shown) on the dashed part of the jump set. The endpoints
871
+ of the secondary jump set correspond to the extreme values 0 and 1 of the volume fraction
872
+ λ in (2.8). The corresponding points on the secondary jump set must lie on the primary
873
+ jump set. There are two possibilities. Either F ̸= F or F = F at λ = 0 or 1. In the
874
+ former case the limiting position F+ of F is rank-one related to two different points on the
875
+ jump set: F− (layer normal e1) and F (layer normal e2). W-point F+ is the only one with
876
+ this property. All other points F+ on the jump set have a unique counterpart F−. In the
877
+ latter case a detailed asymptotic analysis shows that that the common limit point of F and
878
+ F must achieve equality in the “Legendre-Hadamard for phase boundaries” inequality from
879
+ [23]. This point lies on the dashed part of the jump set and is used in numerical calculations.
880
+ The details of the analysis will be spelled out in a forthcoming paper [20].
881
+ Circular nucleus. In [26] it is shown that the secondary jump set (green curve in Fig. 3)
882
+ is unstable. That means that the corresponding bound on the binodal is not optimal. We
883
+ can improve the bound using another method of probing the binodal: nucleation of equilib-
884
+ rium energy-neutral inclusions. The theory justifying why such nucleation tests probe the
885
+ binodal was developed in [19]. In the case of the isotropic, objective energy (3.1) and a
886
+ hydrostatic loading it is natural that the shape of an optimal precipitate should be circular.
887
+ The deformation gradient inside the circular precipitate must be a constant hydrostatic field
888
+ F0 = εW
889
+ 0 I2, since that field is rank-one connected to an infinite family of fields
890
+ FR = R
891
+
892
+ εW
893
+
894
+ 0
895
+ 0
896
+ εW
897
+ 0
898
+
899
+ RT,
900
+ R ∈ SO(2),
901
+ where (εW
902
+ − , εW
903
+ 0 ) is a coordinate of one of the W-points. The deformation gradient outside of
904
+ the circular inclusion must solve an Euler-Lagrange equation for the energy (3.1)
905
+ µ∆y + (cof∇y)∇h′(det ∇y) = 0,
906
+ x ∈ R2 \ B(0, 1),
907
+ (3.13)
908
+ 14
909
+
910
+ and agree with FR at the point Re1 on the boundary of the circular inclusion:
911
+ ∇y(x) = εW
912
+ − n ⊗ n + εW
913
+ 0 τ ⊗ τ,
914
+ x ∈ ∂B(0, 1).
915
+ (3.14)
916
+ In this case both equations (2.9) will be satisfied for the possibly marginally stable matrix
917
+ F∞ = lim
918
+ |x|→∞ ∇y(x) = ε∞I2.
919
+ We also know that that the values of ∇y(x) inside the circular inclusion and its trace on
920
+ the outside boundary of the inclusion are stable. Our results from [19] then say that either
921
+ F∞ lies on the binodal and all values ∇y(x) in the exterior of the inclusion are stable, or
922
+ F∞ lies in the interior of the binodal region B.
923
+ In our special radially symmetric case we look for a radially symmetric solution of (3.13)
924
+ y = η(r)ˆx,
925
+ |x| > 1.
926
+ The the unknown function η(r) must solve
927
+
928
+ η
929
+ r
930
+ d
931
+ drh′ �
932
+ ηη′
933
+ r
934
+
935
+ + µ
936
+
937
+ η′ + η
938
+ r
939
+ �′ = 0,
940
+ r > 1,
941
+ η′(1) = εW
942
+ − ,
943
+ η(1) = εW
944
+ 0 .
945
+ (3.15)
946
+ The nonlinear second order ODE (3.15) cannot be integrated explicitly, but can be solved
947
+ numerically. In order to do so, we need to convert the infinite range r > 1 into a finite one
948
+ by means of the change of the independent variable x = 1/r2. It will also be convenient to
949
+ change the dependent variable v = η/r, so that v(x) would have a finite limit, when x → 0.
950
+ Then v(x) solves
951
+ v′′ = − (v′)2vh′′(v2 − 2xvv′)
952
+ µ + v2h′′(v2 − 2xvv′),
953
+ x ∈ [0, 1],
954
+ v(1) = εW
955
+ 0 ,
956
+ v′(1) = εW
957
+ 0 − εW
958
+
959
+ 2
960
+ .
961
+ (3.16)
962
+ The value ε∞ = v(0)I2 found numerically is shown as a blue dot in Fig. 4. It provides an
963
+ improved bound on the binodal compared to the secondary jump set (green line in Fig. 4) by
964
+ showing that hydrostatic strains between the blue dot and the green line are unstable. This
965
+ conclusion holds, provided the non-degeneracy condition (2.10) is verified. A calculation
966
+ shows that
967
+
968
+ Rd W ◦
969
+ F (F0, ∇φ)dx = −I2
970
+
971
+ R2 h′′(ε2
972
+ ∞)ε2
973
+
974
+
975
+ η′(r) + η(r)
976
+ r
977
+ − 2ε∞
978
+
979
+ dx.
980
+ Thus,
981
+
982
+ Rd W ◦
983
+ F (F0, ∇φ)dx = −2πh′′(ε2
984
+ ∞)ε2
985
+ ∞I2 lim
986
+ r→∞(rη(r) − ε∞r2).
987
+ To see that the limit above exists and is non-zero, at least for small µ > 0, we simply solve
988
+ (3.15) for µ = 0, for which εW
989
+ 0 = √d2, εW
990
+ − =
991
+ d1
992
+ √d2. The solution is η(r) = √d1r2 + d2 − d1,
993
+ and we easily see that
994
+ lim
995
+ r→∞(rη(r) − ε∞r2) = d2 − d1
996
+ 2√d1
997
+ .
998
+ 15
999
+
1000
+ Hence, the non-degeneracy condition (2.10) will hold, at least for sufficiently small µ > 0.
1001
+ The non-degeneracy will also hold for all µ below the topological transition, because if we
1002
+ write �η(r) = η(r) − ε∞r, then (assuming that �η′(r) → 0, as r → ∞) �η(r) will solve, when r
1003
+ is large, the differential equation
1004
+ ε∞h′′(ε2
1005
+ ∞)
1006
+
1007
+ ε∞
1008
+
1009
+ �η′ + �η
1010
+ r
1011
+
1012
+ + �η′�η
1013
+ r
1014
+
1015
+ + µ
1016
+
1017
+ �η′ + �η
1018
+ r
1019
+
1020
+ = 0.
1021
+ This integrates to
1022
+ ε∞h′′(ε2
1023
+ ∞)(2ε∞r�η + �η2) + 2µr�η = 2c.
1024
+ Since �η, satisfying �η′(r) → 0, as r → ∞, cannot be zero (it is the leading term of η(r)−ε∞r),
1025
+ we conclude that the constant of integration c cannot be zero either. Hence, we obtain that
1026
+ lim
1027
+ r→∞(rη(r) − ε∞r2) = lim
1028
+ r→∞ r�η(r) =
1029
+ c
1030
+ µ + ε2∞h′′(ε2∞) ̸= 0.
1031
+ Polyconvexity limits along εI2. We now turn to the problem of proving polyconvexity at
1032
+ points F = εI2. To succeed we need to find a constant m ∈ R, such that (2.5) holds. For
1033
+ our energy we compute
1034
+ W ◦(F , H) = µ
1035
+ 2|H|2 + h(ε2 + d + εθ) − h(ε2) − εh′(ε2)θ,
1036
+ θ = Tr H, d = det H.
1037
+ We also have
1038
+ |H|2 = 1
1039
+ 2|H − HT|2 − 2d + θ2.
1040
+ Hence we need to find m ∈ R, such that
1041
+ µθ2
1042
+ 2
1043
+ + h(ε2 + d + εθ) − h(ε2) − εh′(ε2)θ ≥ (m + µ)d,
1044
+ ∀d ≤ θ2
1045
+ 4 .
1046
+ (3.17)
1047
+ In particular the inequality must hold for d = θ2/4. In that case we must have
1048
+ m ≤ µ + 4 min
1049
+ θ∈R
1050
+ h(ε2 + θ2/4 + εθ) − h(ε2) − εh′(ε2)θ
1051
+ θ2
1052
+ = m∗.
1053
+ (3.18)
1054
+ The infimum of the smooth function
1055
+ F(d, θ) = µθ2
1056
+ 2
1057
+ + h(ε2 + d + εθ) − h(ε2) − εh′(ε2)θ − (m + µ)d
1058
+ must be attained either at a critical point or at infinity. Obviously, if along the minimizing
1059
+ sequence the quantity δ = ε2 + d + εθ goes to infinity, then the values of F must also go to
1060
+ +∞, which cannot happen along a minimizing sequence. Hence, (d, θ) go to infinity so that
1061
+ δ stays bounded. Hence, we can switch variables and instead of the pair (d, θ) consider the
1062
+ pair (δ, θ). In this case d = δ − εθ − ε2. Hence,
1063
+ F(δ, θ) = µθ2
1064
+ 2
1065
+ + h(δ) − h(ε2) − εh′(ε2)θ − (m + µ)(δ − εθ − ε2),
1066
+ 16
1067
+
1068
+ where δ ≤ (θ/2 + ε)2. Minimizing F(δ, θ) with respect to θ we obtain
1069
+ θ = −ε(m + µ − h′(ε2))
1070
+ µ
1071
+ .
1072
+ Hence we need to minimize
1073
+ f(δ) = h(δ) − h(ε2) − (m + µ)(δ − ε2) − ε2(m + µ − h′(ε2))2
1074
+
1075
+ ,
1076
+ over all δ satisfying
1077
+ δ ≤ ε2(h′(ε2) + µ − m)2
1078
+ 4µ2
1079
+ .
1080
+ (3.19)
1081
+ We remark that taking θ = −4ε in (3.18) we conclude that m∗ ≤ µ + h′(ε2). Thus, the
1082
+ right-hand side of (3.19) is monotone decreasing in m, when m ≤ m∗.
1083
+ Now, the minimum is achieved either at the boundary, where equality in (3.19) holds, or
1084
+ at a critical point. It cannot be “achieved” at infinity, where f(δ) is +∞. If the minimum
1085
+ is achieved at the boundary then f(δ) ≥ 0 for all δ, provided m ≤ m∗. If the minimum is
1086
+ achieved at a critical point
1087
+ h′(δ) = m + µ,
1088
+ then several possibilities need to be considered. Let us first assume that the equation
1089
+ h′(δ) = m∗ + µ
1090
+ (3.20)
1091
+ has a single root δ∗. If that root fails to satisfy (3.19) with m = m∗, then for m = m∗ the
1092
+ function f(δ) has no critical points and polyconvexity holds. If that root satisfies (3.19),
1093
+ then all values δ < δ∗ are admissible. We then substitute m = h′(δ) − µ, δ ≤ δ∗ into f(δ).
1094
+ The resulting function
1095
+ f(δ) = h(δ) − h(ε2) − h′(δ)(δ − ε2) − ε2(h′(δ) − h′(ε2))2
1096
+
1097
+ (3.21)
1098
+ can be plotted on (−∞, δ∗] versus m = h′(δ) − µ to see if there are values of m above which
1099
+ all values of f(δ) are positive.
1100
+ Yet another possibility is when equation (3.20) has 3 real roots. If even the smallest root
1101
+ δ∗ fails to satisfy (3.19) with m = m∗, then there are no critical points and polyconvexity
1102
+ holds. Otherwise, all values of δ ≤ δ∗ are admissible and we can prove failure of polyconvexity
1103
+ by plotting (3.21) versus m(δ) = h′(δ)−µ on δ ≤ δ∗ and checking that it is negative. In fact,
1104
+ we believe that polyconvexity fails in all cases when δ∗ satisfies (3.19). In order to exhibit
1105
+ this failure we only need to produce a single value of admissible δ for which f(δ), given
1106
+ by (3.21), is negative. Hence, if (3.21) is negative for all δ ≤ δ∗, then there is no need to
1107
+ examine other intervals of admissible δ, since for any m ≤ m∗ there is always an admissible
1108
+ δ ≤ δ∗, which makes f(δ) negative. However, if f(δ) has a region where it is positive, then
1109
+ one needs to examine other areas of admissibility and check whether f(δ) is negative for the
1110
+ same values of m. Thus, we obtain an algorithm that can prove polyconvexity or failure of
1111
+ 17
1112
+
1113
+ 1.1
1114
+ 1.15
1115
+ 1.2
1116
+ 1
1117
+ 1.06
1118
+ 1.08
1119
+ 1.1
1120
+ 1.12
1121
+ 1.14
1122
+ 1.16
1123
+ 1.18
1124
+ 1.2
1125
+ 2
1126
+ = 2.8168
1127
+ nucleation bound
1128
+ polyconvexity bound
1129
+ secondary jump set
1130
+ Figure 4: Bounds on the binodal from the inside and the outside of the binodal region along
1131
+ hydrostatic strains.
1132
+ it in many, but not all cases. Polyconvexity holds whenever
1133
+ δ∗ > ε2(h′(ε2) + µ − m∗)2
1134
+ 4µ2
1135
+ .
1136
+ for all solutions δ∗ of h′(δ) = µ + m∗. Polyconvexity fails whenever f(δ) < 0 for all δ < δ∗,
1137
+ provided
1138
+ δ∗ ≤ ε2(h′(ε2) + µ − m∗)2
1139
+ 4µ2
1140
+ .
1141
+ If ε2 = d1, then the minimization problem (3.18) simplifies:
1142
+ min
1143
+ θ∈R
1144
+ h(d1 + θ√d1 + θ2/4)
1145
+ θ2
1146
+ .
1147
+ We first observe that in general θ = 0 is not a minimizer. Then there are 3 minimizers:
1148
+ θ = −4
1149
+
1150
+ d1,
1151
+ θ = ±2
1152
+
1153
+ d2 − 2
1154
+
1155
+ d1.
1156
+ When ε = √d1 + x, then the minimizer θ(x) must be located near one of the above 3
1157
+ minimizers. We can then write θ = θ0 + y for the minimizer, where θ0 denotes one of the 3.
1158
+ If we write the function under the minimum as H(ε, θ), then at the minimum we must have
1159
+ ∂H/∂θ = 0, which gives the equation
1160
+ x ∂2H
1161
+ ∂θ∂ε + y∂2H
1162
+ ∂θ2 = 0.
1163
+ After solving for y and substituting this solution back into H we obtain
1164
+ H = x
1165
+
1166
+
1167
+ ∂H
1168
+ ∂ε − ∂H
1169
+ ∂θ
1170
+ ∂2H
1171
+ ∂θ∂ε
1172
+ ∂2H
1173
+ ∂θ2
1174
+
1175
+
1176
+  ,
1177
+ 18
1178
+
1179
+ where derivatives are evaluated at (√d1, θ0). Maple calculation yields
1180
+ H =
1181
+
1182
+
1183
+
1184
+
1185
+
1186
+ x
1187
+ 2
1188
+ √d1h′′(d1),
1189
+ θ0 = −4√d1,
1190
+ xd1h′′(d1)
1191
+ √d1+√d2 ,
1192
+ θ0 = −2√d2 − 2√d1,
1193
+ xd1h′′(d1)
1194
+ √d1−√d2 ,
1195
+ θ0 = 2√d2 − 2√d1.
1196
+ This shows that θ = 2√d2 − 2√d1 + y is the minimizer, while
1197
+ m∗ = µ −
1198
+ 4xd1
1199
+ √d2 − √d1
1200
+ h′′(d1).
1201
+ In particular, the equation h′(δ) = m∗ + µ will have 3 real roots. The smallest one δ∗ will
1202
+ be near d1:
1203
+ δ∗ = d1 + µ + m∗
1204
+ h′′(d1) .
1205
+ Finally, polyconvexity will hold if (3.19) fails when δ = δ∗ and m = m∗. In other words we
1206
+ must have (asymptotically)
1207
+ ε ≤
1208
+
1209
+ d1 +
1210
+ µ
1211
+ h′′(d1)√d1
1212
+ √d2 − √d1
1213
+ √d2 + √d1
1214
+ .
1215
+ (3.22)
1216
+ Fig. 4 showing the right-hand side of (3.22) as a red dot implies that ε∞I2 fails to be
1217
+ polyconvex, but by a very slim margin. The ordering of the bounds in Fig. 4 persists on
1218
+ the entire range of µ. We see that the gap between established stability (along the bisector
1219
+ below the red dot) and established instability (along the bisector above the blue dot) is very
1220
+ small.
1221
+ 4
1222
+ Limiting case µ → 0
1223
+ In this section we derive explicit asymptotics of the secondary jump set and the nucleation
1224
+ bound.
1225
+ Secondary jump set. Expanding equation (3.7) to first order in µ we obtain
1226
+ ε+ = d2
1227
+ ε0
1228
+
1229
+ µ
1230
+ 4ε3
1231
+ 0(d2 − d1) + O(µ2),
1232
+ ε− = d1
1233
+ ε0
1234
+ +
1235
+ µ
1236
+ 4ε3
1237
+ 0(d2 − d1) + O(µ2).
1238
+ (4.1)
1239
+ When d1 and d2 are fixed we think of ε± as functions of ε0 and µ, even if we suppress this
1240
+ in the notation. Clearly, when µ → 0 we have ε+ → d2/ε0, ε− → d1/ε0.
1241
+ The parametric equations (x0(ε0; µ), y0(ε0; µ)) of secondary jump set converge, when
1242
+ µ → 0, to the hyperbola x0y0 = d1. In particular, d0(ε0, µ) → d1, as µ → 0. The volume
1243
+ fraction λ of the rank-one laminate used in the second rank laminate is also a function of
1244
+ ε0 and µ and must have a limit (at least along a subsequence) λ(ε0; µ) → λ0(ε0), as µ → 0.
1245
+ Equation (3.11) shows that d0 = d1 + µδ + O(µ2), while δ satisfies
1246
+ d
1247
+ ε0
1248
+
1249
+ ε0 + 1
1250
+ ε0
1251
+ �d1
1252
+ d2
1253
+ ε2
1254
+ 0 − d1 + d2
1255
+ 2ε2
1256
+ 0
1257
+ d
1258
+ ��
1259
+ − d1 − 2δ(d2 − d1)2d
1260
+ 2
1261
+ ε2
1262
+ 0
1263
+ = 0,
1264
+ (4.2)
1265
+ 19
1266
+
1267
+ where d = λd2 + (1 − λ)d1. Equation (4.2) was obtained simply by passing to the limit as
1268
+ µ → 0 in equation (3.11).
1269
+ When we pass to the limit as µ → 0 in (3.12) we obtain
1270
+ (d − d1)2(ε4
1271
+ 0 + d
1272
+ 2 − 2d2d)
1273
+ 2ε2
1274
+ 0d
1275
+ 2
1276
+ = 0.
1277
+ (4.3)
1278
+ The dependence of d on the volume fraction λ is essential and should not disappear in the
1279
+ limit µ → 0. Therefore, the solution of (4.3) that we are after is
1280
+ d = d2 −
1281
+
1282
+ d2
1283
+ 2 − ε4
1284
+ 0,
1285
+ (4.4)
1286
+ where the choice of the root was dictated by the requirement that d ≤ d2. Combining this
1287
+ with the requirement that d ≥ d1 shows that
1288
+ 4�
1289
+ d2
1290
+ 2 − (d2 − d1)2 ≤ ε0 ≤
1291
+
1292
+ d2.
1293
+ (4.5)
1294
+ Substituting (4.4) into (4.2) gives the explicit formula for δ:
1295
+ δ = ε4
1296
+ 0(d2 − d1) − 2(d2
1297
+ 2 − ε4
1298
+ 0)(d2 −
1299
+
1300
+ d2
1301
+ 2 − ε4
1302
+ 0)
1303
+ 4ε2
1304
+ 0(d2 − d1)2(d2 −
1305
+
1306
+ d2
1307
+ 2 − ε4
1308
+ 0)2
1309
+ .
1310
+ (4.6)
1311
+ It seem that in order to obtain the correct asymptotics of the secondary jump set we need
1312
+ to obtain the first order asymptotics of ε:
1313
+ ε = d2 −
1314
+
1315
+ d2
1316
+ 2 − ε4
1317
+ 0
1318
+ ε0
1319
+ + �εµ + O(µ2).
1320
+ (4.7)
1321
+ In fact, this is not necessary because the leading order asymptotics of d0 is a constant d1.
1322
+ In that case, as far as the first order asymptotics as µ → 0 is concerned, using (4.7) simply
1323
+ corresponds to reparametrizing the curve
1324
+
1325
+
1326
+
1327
+
1328
+
1329
+
1330
+
1331
+ x0 = d2 −
1332
+
1333
+ d2
1334
+ 2 − ε4
1335
+ 0
1336
+ ε0
1337
+ ,
1338
+ y0 = d1 + µδ(ε0)
1339
+ x0
1340
+ .
1341
+ (4.8)
1342
+ Indeed, if we change the curve parameter ε0 to ε0 + µ�ε/x′
1343
+ 0(ε0), then
1344
+ x0
1345
+
1346
+ ε0 +
1347
+ µ�ε
1348
+ x′
1349
+ 0(ε0)
1350
+
1351
+ = x0(ε0) + µ�ε + O(µ2).
1352
+ At the same time
1353
+ y0
1354
+
1355
+ ε0 +
1356
+ µ�ε
1357
+ x′
1358
+ 0(ε0)
1359
+
1360
+ =
1361
+ d1
1362
+ x0(ε0) −
1363
+ µd1�ε
1364
+ x0(ε0)2 + µδ(ε0)
1365
+ x0(ε0) + O(µ2) = d1 + µδ
1366
+ x0 + µ�ε + O(µ2).
1367
+ 20
1368
+
1369
+ 0
1370
+ 1
1371
+ 2
1372
+ 3
1373
+ 4
1374
+ 5
1375
+ 1
1376
+ 1.05
1377
+ 1.1
1378
+ 1.15
1379
+ asymptotic
1380
+ numerical
1381
+ pcx
1382
+ asymptotic
1383
+ Figure 5: Comparison between the asymptotics (4.12) and ε∞ obtained from the numerical
1384
+ solution of (3.15).
1385
+ We conclude that equation (4.8) correctly describes the asymptotics of the secondary jump
1386
+ set with O(µ2) error, where the parameter ε0 varies according to (4.5). When ε0 = √d2,
1387
+ the secondary jump set enters one of the W-points, while when ε0 =
1388
+ 4�
1389
+ d2
1390
+ 2 − (d2 − d1)2 the
1391
+ secondary jump set enters its other end at the “Legendre-Hadamard for phase boundaries”
1392
+ bound that for small µ lies on the dashed part of the jump set in Fig. 3.
1393
+ The plot of
1394
+ (4.8) is in Fig. 3 and is indistinguishable from the numerically obtained curve using the full
1395
+ (non-asymptotic) versions of secondary jump set equations.
1396
+ Circular nucleus. In the near-liquid limit µ → 0 we can find the asymptotics of the
1397
+ solution explicitly.
1398
+ We know that in the limit µ → 0 the field d(x) = det ∇y(x) must
1399
+ approach d1. Hence,
1400
+ ηη′
1401
+ r
1402
+ = d1 + µδ(r) + O(µ2),
1403
+ r > 1.
1404
+ That implies
1405
+ η(r) =
1406
+
1407
+ d1r2 + c0 + µ�η(r) + O(µ2),
1408
+ (4.9)
1409
+ and therefore,
1410
+ δ(r) = 1
1411
+ r
1412
+
1413
+ �η(r)
1414
+
1415
+ d1r2 + c0
1416
+ �′
1417
+ .
1418
+ Substituting this ansatz into (3.15) we obtain
1419
+ µ
1420
+ √d1r2 + c0
1421
+ r
1422
+ h′′(d1)δ′(r) + µ
1423
+
1424
+ d1r
1425
+ √d1r2 + c0
1426
+ +
1427
+ √d1r2 + c0
1428
+ r
1429
+ �′
1430
+ + O(µ2) = 0.
1431
+ (4.10)
1432
+ Initial conditions from (3.15) imply that
1433
+ c0 = d2 − d1,
1434
+ �η(1) = −
1435
+ d2 − d1
1436
+ 4d3/2
1437
+ 2 h′′(d2)
1438
+ ,
1439
+ �η′(1) = d1(d2 − d1)
1440
+ 2d3/2
1441
+ 2
1442
+
1443
+ 1
1444
+ d1h′′(d1) +
1445
+ 1
1446
+ 2d2h′′(d2)
1447
+
1448
+ .
1449
+ 21
1450
+
1451
+ Equation (4.10) is easy to integrate (observing that √d1r2 + c0/r is decreasing from √d2 to
1452
+ √d1 and is therefore uniformly bounded away from zero and ∞).
1453
+ h′′(d1)�η(r) =
1454
+ c1r2 + c2
1455
+ √d1r2 + c0
1456
+
1457
+ r2
1458
+ 2√d1r2 + c0
1459
+ ln
1460
+ √d1r2 + c0
1461
+ r
1462
+ .
1463
+ (4.11)
1464
+ From initial conditions for �η(r) we obtain
1465
+ c1 = 1
1466
+ 2 ln
1467
+
1468
+ d2,
1469
+ c2 = −(d2 − d1)h′′(d1)
1470
+ 4d2h′′(d2)
1471
+ ,
1472
+ and hence
1473
+ ε∞ =
1474
+ ��
1475
+ d1 +
1476
+ µ
1477
+ 2h′′(d1)√d1
1478
+ ln
1479
+ √d2
1480
+ √d1
1481
+
1482
+ I2 + O(µ2).
1483
+ (4.12)
1484
+ Figure 5 shows the quality of the asymptotics for the entire range of shear moduli µ. The
1485
+ numbers on the y-axis indicate that even for values of µ that are not particularly small the
1486
+ asymptotics (4.12) gives a good approximation of the actual value of ε∞. For example, for
1487
+ µ = 3 the relative discrepancy is only around 0.1%.
1488
+ 5
1489
+ A glimpse into the relaxed energy
1490
+ Hypothetical bounds on the binodal. We have seen in the foregoing discussion that the energy
1491
+ W(F ) is not polyconvex at F = ε∞I2. This is not very surprising, since polyconvexity is
1492
+ usually strictly stronger that quasiconvexity and we expect and conjecture that F = ε∞I2
1493
+ lies on the binodal—at the very edge of quasiconvexity. Here we recall our observation that
1494
+ if someone could prove that F = ε∞I2 is stable, then we would immediately conclude that
1495
+ for every |x| > 1
1496
+ ∇y(x) = η′(r)ˆx ⊗ ˆx + η(r)
1497
+ r (I2 − ˆx ⊗ ˆx)
1498
+ would be stable in the sense of Definition 2.2, providing a bound on the binodal from the
1499
+ outside. For the entire range of µ for which W-points are polyconvex the union of the curves
1500
+
1501
+ ε1 = η(r)
1502
+ r ,
1503
+ ε2 = η′(r),
1504
+ and
1505
+
1506
+ ε1 = η′(r)
1507
+ ε2 = η(r)
1508
+ r ,
1509
+ r > 1
1510
+ (5.1)
1511
+ are indistinguishable from the secondary jump set curves shown in green in Fig. 3. Fig. 6
1512
+ shows the same blown-up part of the strain space as in Fig. 4, where the curves (5.1) shown in
1513
+ magenta are passing through the blue point from Fig. 4. Assuming the conjectured stability
1514
+ of ε∞I2, the magenta curve must lie outside of binodal region, while secondary jump set
1515
+ lies in its interior [26]. Thus, the binodal of the energy (3.1) would have to lie between the
1516
+ green and the magenta curves. We will even go so far as to conjecture that the magenta
1517
+ curve is in fact the actual binodal of the energy (3.1). Regardless, under the assumption of
1518
+ stability of ε∞I2, the magenta line represents a rather tight outside bound on the binodal
1519
+ region. Another byproduct of the assumed stability of ε∞I2 would be the formula for the
1520
+ 22
1521
+
1522
+ 0.5
1523
+ 1
1524
+ 1.5
1525
+ 2
1526
+ 2.5
1527
+ 1
1528
+ 0.5
1529
+ 1
1530
+ 1.5
1531
+ 2
1532
+ 2.5
1533
+ 2
1534
+ = 2.8168
1535
+ hypothetical binodal
1536
+ known binodal region
1537
+ 1.1
1538
+ 1.15
1539
+ 1.2
1540
+ 1
1541
+ 1.06
1542
+ 1.08
1543
+ 1.1
1544
+ 1.12
1545
+ 1.14
1546
+ 1.16
1547
+ 1.18
1548
+ 1.2
1549
+ 2
1550
+ = 2.8168
1551
+ nucleation bound
1552
+ polyconvexity bound
1553
+ secondary jump set
1554
+ outside bound
1555
+ Figure 6: A hypothetical bound on the binodal region from the outside, assuming stability
1556
+ of ε∞I2.
1557
+ quasiconvex envelope QW(F ) for hydrostatic strains F . If F = ε∞I2 is stable, then our
1558
+ radial solution ∇y(x) = η(r)ˆx of (3.15) is also a global minimizer in every finite ball B(0, R),
1559
+ where it satisfies the affine boundary condition y(x) = (η(R)/R)x, x ∈ ∂B(0, R) [21]. The
1560
+ energy of such configurations must necessarily be QW(η(R)I2/R)|B(0, R)|. This permits
1561
+ us to compute QW(εI2) for all ε, as the energy of y(x) = η(r)ˆx in B(0, R). Using the
1562
+ Clapeyron-type formula for the nonlinear elastic energy stored in an equilibrium stationary
1563
+ configuration we obtain for F = η(R)I2/R: [21]
1564
+ |B(0, R)|QW(F ) = 1
1565
+ 2
1566
+
1567
+ ∂B(0,R)
1568
+ {P (∇y)n · y + P ∗(∇y)n · x}dS.
1569
+ (5.2)
1570
+ Substituting n = ˆx, y = η(r)ˆx into (5.2) we obtain
1571
+ QW
1572
+ �η(R)
1573
+ R I2
1574
+
1575
+ = 2(µ − h′(d))d − µη′(R)2 + (2h′(d) + µ)η(R)2
1576
+ R2
1577
+ + 2h(d),
1578
+ (5.3)
1579
+ where
1580
+ d = η′(R)η(R)
1581
+ R
1582
+ .
1583
+ When µ is small we can use the explicit asymptotic formulas (4.9), (4.11) for η(r) to obtain
1584
+ an explicit asymptotics for QW(εI2). The plot of QW(εI2), coming from the numerical
1585
+ solution of (3.15), as well as its explicit asymptotic approximation, superposed on the plot
1586
+ of W(εI2) is shown in Fig. 7.
1587
+ 6
1588
+ Conclusions
1589
+ In this paper our far reaching goal was to solve analytically the relaxation problem for the
1590
+ double well Hadamard energy (3.1) in two space dimensions when the rigidity measure µ
1591
+ 23
1592
+
1593
+ 1
1594
+ 1.2
1595
+ 1.4
1596
+ 1.6
1597
+ 1.8
1598
+ 3
1599
+ 4
1600
+ 5
1601
+ 6
1602
+ 7
1603
+ 8
1604
+ 9
1605
+ Energy
1606
+ W( I2)
1607
+ QW( I2)
1608
+ QWasym( I2)
1609
+ Figure 7: Quasiconvex envelope of W(F ) restricted to hydrostatic strains F = εI2.
1610
+ is sufficiently small. An apparently more attainable target was to locate the corresponding
1611
+ binodal region inside the strain space. The study of the limit µ → 0 was expected to show how
1612
+ the ’cooperative’ , rigidity-controlled microstructures, dominating the quasiconvex envelope
1613
+ at large µ, give rise to more arbitrary and less controlled microstructures characterizing first
1614
+ order phase transitions in zero rigidity liquids.
1615
+ We used some of our previously developed methods to pinpoint a substantial portion
1616
+ of the binodal.
1617
+ While our general methods apply for Hadamard materials in the entire
1618
+ parameter range and are amenable to numerical implementation, here we were able to obtain
1619
+ the explicit asymptotic formulas only in the ’near-liquid’ regime. In particular, we showed
1620
+ that in an ’almost liquid’ limit, a subset of the jump set adjacent to the high strain phase
1621
+ remains stable which ensures that simple lamination delivers the corresponding part of the
1622
+ binodal. This means that even when the reference measure of rigidity µ is small, the high
1623
+ strain phase maintains its tangential rigidity at the level which ensures solid-solid like nature
1624
+ of the incipient phase transition. Instead, our analysis showed that the subset of the jump set
1625
+ adjacent to the low strain and low rigidity phase is unstable in the µ → 0 limit. Moreover, the
1626
+ secondary jump set is also unstable in this limit. This result suggests that laminates of any
1627
+ finite rank are unstable near the corresponding subset of the binodal. As we’ve demonstrated
1628
+ for hydrostatic strains, the reduced rigidity control in this range allows the incipient phase
1629
+ transformation to proceed non-cooperatively through the formation of isolated nuclei of the
1630
+ more rigid phase inside the matrix of the less rigid phase. Such transformation mechanism is
1631
+ already very similar to the one believed to be operating in purely fluid-fluid phase transitions.
1632
+ Whether the revealed asymmetry of the transformation mechanism between the direct
1633
+ and reverse transformation is a peculiarity of the Hadamard material or whether this striking
1634
+ phenomenon has a more general nature, remains to be established. It shows, however, the
1635
+ intricate role of rigidity in structural transformations which, even if weak, can produce rather
1636
+ complex structure of the relaxed energy. This complexity will then reflect a gradual transition
1637
+ from geometrically ordered microstructures, controlled by long range elastic interactions,
1638
+ 24
1639
+
1640
+ to more ’fluid’ microstructures whose spatial organization is mostly affected by molecular
1641
+ interactions operating at short range. In other words, in this limit the direct and reverse
1642
+ solid-solid phase transitions can operate through different transformation mechanisms. The
1643
+ fact that the ensuing complex structure of the relaxed energy at ’almost-liquid’ solid-solid
1644
+ phase transitions is ultimately replaced by a simple energy convexification at fluid-fluid phase
1645
+ transitions points to a singular nature of the limit µ → 0.
1646
+ Acknowledgments.
1647
+ YG was supported by the National Science Foundation under
1648
+ Grant No. DMS-2005538. The work of LT was supported by the French grant ANR-10-
1649
+ IDEX-0001-02 PSL.
1650
+ References
1651
+ [1] J. M. Ball. Progress and puzzles in nonlinear elasticity. In J¨org Schr¨oder and Patrizio
1652
+ Neff, editors, Poly-, Quasi- and Rank-One Convexity in Applied Mechanics, pages 1–15.
1653
+ Springer Vienna, Vienna, 2010.
1654
+ [2] J. M. Ball and F. Murat. W 1,p-quasiconvexity and variational problems for multiple
1655
+ integrals. J. Funct. Anal., 58(3):225–253, 1984.
1656
+ [3] J.M. Ball and R.D. James. Incompatible sets of gradients and metastability. Archive
1657
+ for Rational Mechanics and Analysis, 218(3):1363–1416, 2015.
1658
+ [4] John M. Ball. Some open problems in elasticity. In Geometry, mechanics, and dynamics,
1659
+ pages 3–59. Springer, New York, 2002.
1660
+ [5] D. M. Barnett, J. K. Lee, H. I. Aaronson, and K. C. Russel. The strain energy of a
1661
+ coherent ellipsoidal precipitate. Scripta Metall., 8:1447–1450, 1974.
1662
+ [6] Paul M Chaikin, Tom C Lubensky, and Thomas A Witten. Principles of condensed
1663
+ matter physics, volume 10. Cambridge university press Cambridge, 1995.
1664
+ [7] B. Dacorogna. A relaxation theorem and its application to the equilibrium of gases.
1665
+ Arch. Rational Mech. Anal., 77(4):359–386, 1981.
1666
+ [8] B. Dacorogna. Quasiconvexity and relaxation of nonconvex problems in the calculus of
1667
+ variations. J. Funct. Anal., 46(1):102–118, 1982.
1668
+ [9] B. Dacorogna. Direct methods in the calculus of variations. Springer-Verlag, New York,
1669
+ 2nd edition, 2008.
1670
+ [10] Michelle M Driscoll, Bryan Gin-ge Chen, Thomas H Beuman, Stephan Ulrich, Sid-
1671
+ ney R Nagel, and Vincenzo Vitelli. The role of rigidity in controlling material failure.
1672
+ Proceedings of the National Academy of Sciences, 113(39):10813–10817, 2016.
1673
+ [11] J. Ericksen. Some phase transitions in crystals. Archive for Rational Mechanics and
1674
+ Analysis, 73:99–124, 1980.
1675
+ 25
1676
+
1677
+ [12] J. L. Ericksen. Equilibrium of bars. J. Elasticity, 5(3–4):191–201, 1975.
1678
+ [13] J. L. Ericksen. Twinning of crystals. I. In Metastability and incompletely posed problems
1679
+ (Minneapolis, Minn., 1985), pages 77–93. Springer, New York, 1987.
1680
+ [14] J. L. Ericksen. On kinematic conditions of compatibility. Journal of Elasticity, 26(1):65–
1681
+ 74, 1991.
1682
+ [15] J. L. Ericksen.
1683
+ Bifurcation and martensitic transformations in Bravais lattices.
1684
+ J.
1685
+ Elasticity, 28(1):55–78, 1992.
1686
+ [16] J. Gibbs, Willard. On the equilibrium of heterogeneous substances. Transactions of the
1687
+ Connecticut Academy, III:108–248 and 343–524, 1873 and 1874.
1688
+ [17] Leonardo Golubovi´c and T. C. Lubensky.
1689
+ Nonlinear elasticity of amorphous solids.
1690
+ Physical review letters, 63(10):1082–1085, 1989.
1691
+ [18] Y. Grabovsky and L. Truskinovsky. Roughening instability of broken extremals. Arch.
1692
+ Rat. Mech. Anal., 200(1):183–202, 2011.
1693
+ [19] Y. Grabovsky and L. Truskinovsky. Marginal material stability. Journal of Nonlinear
1694
+ Science, 23(5):891–969, 2013.
1695
+ [20] Yury Grabovsky, M. Oberman, Adam, and Lev Truskinovsky. Rank one convex envelope
1696
+ for hadamard material: numerical and analytical study. In preparation.
1697
+ [21] Yury Grabovsky and Lev Truskinovsky. Delicate regularity and sufficient conditions for
1698
+ lipschitz minimizers of integral functionals. in preparation.
1699
+ [22] Yury Grabovsky and Lev Truskinovsky. Normality condition in elasticity. Journal of
1700
+ Nonlinear Science, 24(6):1125–1146, 2014.
1701
+ [23] Yury Grabovsky and Lev Truskinovsky. Legendre-Hadamard conditions for two-phase
1702
+ configurations. Journal of Elasticity, 123(2):225–243, 2016.
1703
+ [24] Yury Grabovsky and Lev Truskinovsky.
1704
+ Explicit relaxation of a two-well hadamard
1705
+ energy. Journal of Elasticity, 135(1-2):351–373, 2019.
1706
+ [25] Yury Grabovsky and Lev Truskinovsky. When rank-one convexity meets polyconvexity:
1707
+ An algebraic approach to elastic binodal. J. Nonlinear Sci., 28(1):229–253, 2019.
1708
+ [26] Yury Grabovsky and Lev Truskinovsky. Ubiquity of infinite rank laminates. to besub-
1709
+ mitted, In preparation.
1710
+ [27] J. Hadamard. Le¸cons sur la propagation des ondes et les ´equations de l’hydrodynamique.
1711
+ Hermann, Paris, 1903.
1712
+ [28] Fritz John. Plane elastic waves of finite amplitude. hadamard materials and harmonic
1713
+ materials. Communications on Pure and Applied Mathematics, 19(3):309–341, 1966.
1714
+ 26
1715
+
1716
+ [29] V. Kardonski and Roitburd. On the shape of coherent precipitates. Phys. Met. Metal-
1717
+ lurg. USSR, 33:210–212, 1972.
1718
+ [30] A. G. Khachaturyan. Theory of structural transformation in solids. Wiley, New York,
1719
+ 1983.
1720
+ [31] L. B. Kublanov and A. B. Freidin. Nuclei of a solid phase in a deformable material.
1721
+ Prikl. Mat. Mekh., 52(3):493–501, 1988.
1722
+ [32] Lev Davidovich Landau and Evgenii Mikhailovich Lifshitz. Statistical Physics: Volume
1723
+ 5, volume 5. Elsevier, 2013.
1724
+ [33] J. K. Lee, D. M. Barnett, and H. I. Aaronson. The elastic strain energy of coherent
1725
+ ellipsoidal precipitates in anisotropic crystalline solids. Metall. Trans. A, 8A:963–970,
1726
+ 1977.
1727
+ [34] J. C. Maxwell. On the dynamic evidence of the molecular composition of bodies. Nature,
1728
+ 11(279-280):357–359, 374–377, 1875.
1729
+ [35] A. Pineau. Influence of uniaxial stress on the morphology of coherent precipitates during
1730
+ coarsening — elastic energy considerations. Acta Metall., 24:559–564, 1976.
1731
+ [36] J.D. van der Waals. The equilibrium between a solid body and a fluid phase, especially
1732
+ in the neighbourhood of the critical state. In KNAW, Proceedings, volume 6, pages
1733
+ 1903–1904, 1903.
1734
+ 27
1735
+
1NAzT4oBgHgl3EQf8v6L/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
1tE3T4oBgHgl3EQfngpH/content/2301.04625v1.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:37ace31300adbe863e18281b1882f473736d042587a56e9c0904f0b2d7ffe039
3
+ size 420836
1tE3T4oBgHgl3EQfngpH/vector_store/index.faiss ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:37fdcf76d1aa172e0667812fc653d04ad7b1ed05ab14776ed55f120273722d37
3
+ size 3014701
1tE3T4oBgHgl3EQfngpH/vector_store/index.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7e2b97e87cd1449529d0efb80428d7cd5df72bc32fa9b30b81906234c3bf0d57
3
+ size 115523
59E4T4oBgHgl3EQfcAxC/vector_store/index.faiss ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a911caef2b197228e0c78e20e753771fe77c7fc2bc1af3ab8097fc18fab0a907
3
+ size 4784173
69E4T4oBgHgl3EQfcgw0/content/tmp_files/2301.05083v1.pdf.txt ADDED
@@ -0,0 +1,1509 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Gravitational collapse of scalar and vector fields
2
+ Karim Mosani,∗ Koushiki,† Pankaj S. Joshi,‡ and Jay Verma Trivedi§
3
+ International Centre for Space and Cosmology, School of Arts and Sciences,
4
+ Ahmedabad University, Ahmedabad-380009 (Guj), India.
5
+ Tapobroto Bhanja¶
6
+ International Center for Cosmology, & PDPIAS,
7
+ Charotar University of Science and Technology, Anand- 388421 (Guj), India
8
+ (Dated: January 13, 2023)
9
+ We study here the unhindered gravitational collapse of spatially homogeneous (SH) scalar fields φ
10
+ with a potential Vs(φ), as well as vector fields ˜A with a potential Vv(B) where B = g( ˜A, ˜A) and g is
11
+ the metric tensor. We show that in both cases, classes of potentials exist that give rise to black holes
12
+ or naked singularities depending on the choice of the potential. The strength of the naked singular-
13
+ ity is examined, and they are seen to be strong, in the sense of Tipler, for a wide class of respective
14
+ potentials. We match the collapsing scalar/vector field with a generalized Vaidya spacetime outside.
15
+ We highlight that full generality is maintained within the domain of SH scalar or vector field collapse.
16
+ keywords: Gravitational collapse, singularity, scalar field, vector field, causal structure.
17
+ I.
18
+ INTRODUCTION
19
+ The contraction of a matter field under its gravita-
20
+ tional influence is called gravitational collapse. In 1939,
21
+ Oppenheimer and Snyder [1], and independently in 1938,
22
+ Datt [2] developed the first solution of Einstein’s field
23
+ equations (called the OSD model) depicting the gravita-
24
+ tional collapse of a massive star. They considered a very
25
+ specific case of spatially homogeneous (SH) dust collapse
26
+ (By spatial homogeneity, we mean homogeneous on a
27
+ three-dimensional spacelike orbit with a six-dimensional
28
+ isometry group G6 corresponding to the spacetime [3]).
29
+ Such a matter field undergoes gravitational collapse that
30
+ ends up in a singularity. Such a spacetime singularity
31
+ is hidden behind an event horizon, not visible to any ob-
32
+ server, and what we obtain is a black hole as the outcome
33
+ of continual collapse.
34
+ Extending the above special scenario, in 1969, Penrose
35
+ proposed what is now known as the cosmic censorship hy-
36
+ pothesis (CCH) [4]. The weaker version of the hypothesis
37
+ states that all singularities of gravitational collapse are
38
+ hidden within a black hole and hence, cannot be seen
39
+ by a distant observer (a globally naked singularity can-
40
+ not exist). The strong version of the hypothesis states
41
+ that no past inextendible nonspacelike geodesics can ex-
42
+ ist between the singularity and any point in the space-
43
+ time manifold. In other words, a causal geodesic with
44
+ a positive tangent “at” the singularity does not exist (a
45
+ locally naked singularity also cannot exist).
46
+ The sup-
47
+ porting argument for the validity of the strong CCH is
48
+ the desirability of the spacetime manifold to be globally
49
+ ∗ kmosani2014@gmail.com
50
+ † koushiki.malda@gmail.com
51
+ ‡ pankaj.joshi@ahduni.edu.in
52
+ § jay.verma2210@gmail.com
53
+ ¶ tapobroto.bhanja@gmail.com
54
+ hyperbolic. Global hyperbolicity implies the existence of
55
+ Cauchy surfaces embedded in the total manifold, thereby
56
+ making general relativity a deterministic theory [5–7].
57
+ Now singularity theorems of Hawking and Penrose
58
+ [6, 11] do not imply that singularities are hidden from
59
+ an external observer under any possible circumstances.
60
+ In fact, singularity theorems take the causality condition
61
+ as one of the axioms to start with to prove the existence
62
+ of incomplete past (future) directed causal curves. Addi-
63
+ tionally, the OSD model that motivated cosmic censor-
64
+ ship is a special case. Joshi and Malafarina [12] showed
65
+ that any arbitrarily small neighbourhood of the initial
66
+ data giving rise to OSD collapse contains initial data cor-
67
+ responding to collapse evolution giving rise to a singular-
68
+ ity with the following property: one could trace outgoing
69
+ past singular causal geodesics. This means that the end
70
+ state of OSD collapse is unstable under small perturba-
71
+ tions in initial data. Moreover, one can show the forma-
72
+ tion of naked singularities (global and local) as an end
73
+ state of gravitational collapse from suitable, physically
74
+ reasonable initial data for various matter fields [13, 14].
75
+ This implies that the initial conditions must be fine-tuned
76
+ for the cosmic censorship conjecture to hold.
77
+ In such a context, an important question one can ask
78
+ here is as follows: what will be the end state of an un-
79
+ hindered gravitational collapse of a fundamental matter
80
+ field, such as a scalar field or a vector field, derived from
81
+ an appropriate Lagrangian?
82
+ The answer to this question has been achieved up to
83
+ a certain extent. Scalar fields are fundamental matter
84
+ fields derived from suitable Lagrangian.
85
+ A real scalar
86
+ field is a map defined on a smooth manifold as φ : M →
87
+ R with a suitable continuity condition.
88
+ Christodoulou
89
+ showed that in the case of gravitational collapse of a
90
+ massless scalar field φ (the scalar field Lagrangian is
91
+ Lφ = (1/2)gµν∂µφ∂νφ), the set of initial data giving rise
92
+ to a naked singularity as an end state has positive codi-
93
+ mension in the entire initial data set [15, 16]. This means
94
+ arXiv:2301.05083v1 [gr-qc] 12 Jan 2023
95
+
96
+ 2
97
+ that the initial data set corresponding to naked singular-
98
+ ity has a zero measure in the total initial data set. In
99
+ other words, naked singularity in such cases is unstable
100
+ under arbitrarily small perturbations in the initial data.
101
+ One can have a massless scalar field with a poten-
102
+ tial function Vs(φ) that still be a fundamental matter
103
+ field.
104
+ A massive scalar field will then be a particular
105
+ case of a massless scalar field with a specific potential of
106
+ the form Vs(φ) = (1/2)µ2φ2, where µ is the mass term.
107
+ Goswami and Joshi [17] showed the example of the grav-
108
+ itational collapse of a massless SH scalar field with a cer-
109
+ tain potential Vs(φ) that ends up in a naked singularity.
110
+ Mosani, Dey, Bhattacharya, and Joshi [18] conducted a
111
+ similar investigation for a massless scalar field with a two-
112
+ dimensional analogue of the Mexican hat-shaped Higgs
113
+ field potential and found out that the end state of such
114
+ unhindered scalar field collapse is a naked singularity.
115
+ In addition to scalar fields as fundamental matter
116
+ fields, vector fields are also fundamental matter fields de-
117
+ rived from suitable matter Lagrangian. Geometrically,
118
+ vector fields on a smooth manifold M can be thought of
119
+ as sections on the tangent bundle π : TM → M, where
120
+ π is a continuous surjection. A section is a smooth map
121
+ σ : M → TM such that π ◦ σ is an identity map on
122
+ M. From a particle physics point of view, the funda-
123
+ mental nature of a vector field is different from that of a
124
+ scalar field. There are many aspects, but one of the most
125
+ important ones is that massive or massless vector fields
126
+ mediate most particle physics processes. These represent
127
+ the three fundamental interactions: quantum electrody-
128
+ namics and weak and strong processes. A massless vector
129
+ field with a potential function Vv(B) is again a funda-
130
+ mental matter field. A massive vector field will then be
131
+ a particular case of a massless vector field with a specific
132
+ potential of the form Vv(B) = (1/2)µ2B, where µ is the
133
+ mass term. Garfinkle, Mann, and Vuille [26] have studied
134
+ the collapse of a massive vector field and numerically ob-
135
+ tained the critical initial conditions. To our knowledge,
136
+ much analytical work has not been done in investigating
137
+ the causal structure of the end-state spacetime of the un-
138
+ hindered gravitational collapse of matter fields that are
139
+ vector fields.
140
+ In this paper, in both the massless SH scalar field as
141
+ well as vector field cases, we show that there are broad
142
+ classes of potentials for which the configuration collapses
143
+ and ends up in either a black hole or a naked singu-
144
+ larity depending on the potential function chosen. We
145
+ approach the causality investigation problem of scalar
146
+ field as well as vector field collapse in a unified way, so
147
+ to speak.
148
+ As far as general relativity is concerned, it
149
+ does not discriminate between whether a scalar field or
150
+ a vector field seeds the matter field.
151
+ The matter field
152
+ is entirely identified by a rank two tensor field that we
153
+ call the stress-energy tensor. As far as SH perfect fluid
154
+ is concerned, one can identify a given matter field by
155
+ the functional form of the equation of state parameter
156
+ ω(a), where a is the scale factor of the collapsing cloud.
157
+ We derive relevant equations of collapsing SH scalar field
158
+ φ(a) and vector field ˜A(a) in the sub-sections of section
159
+ II. The main body of section II contains discussions and
160
+ relevant relations regarding the gravitational collapse of
161
+ SH perfect fluids. In section III, we smoothly join the
162
+ interior collapsing perfect fluid with an external gener-
163
+ alized Vaidya spacetime. In section IV, we investigate
164
+ the causal structure of the spacetime (condition of ob-
165
+ taining a naked singularity) at the end of the collapse
166
+ of the interior perfect fluid that is either a scalar field φ
167
+ with potential Vs or a vector field ˜A with potential Vv.
168
+ We also depict a few examples of well-known scalar fields
169
+ and vector fields. In section V, we derive the criteria for
170
+ the singularity, thus obtained in the end, to be strong of
171
+ Tipler’s type. In the last section, we highlight the key
172
+ points of the investigation. Here we use the geometrized
173
+ units 8πG = c = 1 throughout.
174
+ II.
175
+ INTERIOR COLLAPSING MATTER FIELD
176
+ Consider a gravitational collapse of a SH perfect fluid.
177
+ The components of the stress-energy tensor in the coor-
178
+ dinate basis {dxµ � ∂ν|0 ≤ µ, ν ≤ 3} of the comoving
179
+ coordinates (t, x, y, z) are given by
180
+ T µ
181
+ ν = diag (−ρ, p, p, p) .
182
+ (1)
183
+ The spacetime geometry is governed by the flat (k = 0)
184
+ Friedmann–Lemaˆıtre–Robertson–Walker (FLRW) metric
185
+ ds2 = −dt2 + a2dΣ2,
186
+ (2)
187
+ where dΣ2 = dx2 + dy2 + dz2. Here a = a(t) is the scale
188
+ factor such that a(0) = 1 and a(ts) = 0, where ts is the
189
+ time of formation of the singularity. R = R(t, r) is the
190
+ physical radius of the collapsing cloud and can be written
191
+ as
192
+ R(t, r) = ra(t),
193
+ (3)
194
+ where r is the radial spherical coordinate. For a FLRW
195
+ spacetime Eq.(2), we have
196
+ ρ = 3˙a2
197
+ a2 ,
198
+ (4)
199
+ and
200
+ p = −2¨a
201
+ a − ˙a2
202
+ a2 .
203
+ (5)
204
+ The overhead dot denotes the partial time derivative of
205
+ a. Eq.(4) can be rewritten to obtain the dynamics of the
206
+ collapse as
207
+ ˙a = −
208
+
209
+ ρ(a)
210
+ 3 a.
211
+ (6)
212
+ Differentiating the above equation once again gives us
213
+ ¨a = 1
214
+ 3a
215
+ �aρ,a
216
+ 2
217
+ + ρ
218
+
219
+ .
220
+ (7)
221
+
222
+ 3
223
+ Integrating Eq.(6), we obtain the time curve, which is
224
+ t(a) =
225
+ � 1
226
+ a
227
+ �3
228
+ ρ
229
+ da
230
+ a .
231
+ (8)
232
+ The dynamics of the scale factor a(t) is, thus, the inverse
233
+ of the LHS of the above equation. The time of formation
234
+ of the singularity ts = t(0) is
235
+ ts =
236
+ � 1
237
+ 0
238
+ �3
239
+ ρ
240
+ da
241
+ a .
242
+ (9)
243
+ Now, let us consider a particular matter field ˆT from a
244
+ set of all the possible SH perfect fluids. Choosing such
245
+ an element means choosing a specific functional form of
246
+ the equation of state parameter
247
+ ω(a) = p
248
+ ρ.
249
+ (10)
250
+ Using Eq.(4), Eq.(5), and Eq.(10), we can express the
251
+ density of the matter field with the equation of state
252
+ parameter ω as
253
+ ρ = ρ0 exp
254
+ �� 1
255
+ a
256
+ 3 (1 + ω(a))
257
+ a
258
+ da
259
+
260
+ ,
261
+ (11)
262
+ An SH perfect fluid is a fundamental matter field since
263
+ it can be derived by a fundamental matter Lagrangian.
264
+ In the following two subsections, we will describe two
265
+ distinct ways of obtaining such a matter field.
266
+ A.
267
+ Scalar field collapse
268
+ We prove that any SH perfect fluid is equivalent to a
269
+ SH scalar field φ(a) with a suitable potential Vs(a), as
270
+ far as the gravitational collapse is concerned. If φ(a) is
271
+ invertible, then the following statement holds: Any SH
272
+ perfect fluid is gravitationally equivalent to a SH scalar
273
+ field φ with a suitable potential Vs(φ).
274
+ Consider a real scalar field defined on the manifold M
275
+ as
276
+ φ : M → R.
277
+ (12)
278
+ The Lagrangian of a massless scalar field φ with potential
279
+ Vs(a) is given by
280
+ Lφ = 1
281
+ 2gµν∂µφ∂νφ − Vs(φ),
282
+ (13)
283
+ The stress-energy tensor is obtained from the Lagrangian
284
+ Lφ as
285
+ Tµν = −
286
+ 2
287
+ √−g
288
+ δ (√−gLφ)
289
+ δgµν
290
+ .
291
+ (14)
292
+ The density (ρs) and the isotropic pressure (ps) are sub-
293
+ sequently expressed in terms of the time derivative of the
294
+ scalar field and its potential as
295
+ ρs = 1
296
+ 2
297
+ ˙φ2 + Vs
298
+ (15)
299
+ and
300
+ ps = 1
301
+ 2
302
+ ˙φ2 − Vs.
303
+ (16)
304
+ The overhead dot denotes the time derivative of the func-
305
+ tions.
306
+ From Eq.(15) and Eq.(16), and from using the
307
+ chain rule ˙φ = φ,a ˙a, we get
308
+ ρs + ps = φ2
309
+ ,a ˙a2.
310
+ (17)
311
+ We now equate ρs = ρ and ps = p. Using Eq.(5) and
312
+ (17), along with replacing ˙a and ¨a using Eq.(6) and (7),
313
+ one obtains the expression of density as a function of a
314
+ as
315
+ ρs = ρ0 exp
316
+ �� 1
317
+ a
318
+ aφ2
319
+ ,ada
320
+
321
+ .
322
+ (18)
323
+ From Eqs.(15) and (16), we get
324
+ ps = ρs − 2Vs.
325
+ (19)
326
+ Using Eq.(6) in Eq.(17), we get
327
+ ρs
328
+
329
+ 1 − φ2
330
+ ,aa2
331
+ 3
332
+
333
+ + ps = 0.
334
+ (20)
335
+ Using Eqs.(19) and (20), we get
336
+ Vs(φ) = ρs
337
+
338
+ 1 − φ2
339
+ ,aa2
340
+ 6
341
+
342
+ .
343
+ (21)
344
+ Using Eq.(17), Eq.(6) in Eq.(5), one obtains
345
+ ρs,a
346
+ ρs
347
+ = −φ,2
348
+ a
349
+ a .
350
+ (22)
351
+ We have, using Eq.(10), Eq.(15) and Eq.(16),
352
+ Vs = ρs
353
+ 2 (1 − ω) .
354
+ (23)
355
+ Now from Eq.(21) and Eq.(23), we have
356
+ φ(a),a = ±
357
+
358
+ 3 (1 + ω(a))
359
+ a
360
+ .
361
+ (24)
362
+ Integrating the above equation, one obtains
363
+ φ(a) = ±
364
+ � 1
365
+ a
366
+
367
+ 3 (1 + ω(a))
368
+ a
369
+ da + c.
370
+ (25)
371
+ From Eq. (11) and Eq.(21) we have
372
+ Vs(a) = ρ0
373
+ �1 − ω(a)
374
+ 2
375
+
376
+ exp
377
+ �� 1
378
+ a
379
+ 3 (1 + ω(a))
380
+ a
381
+ da
382
+
383
+ .
384
+ (26)
385
+ Hence, we proved that given the functional form of
386
+
387
+ 4
388
+ the equation of state parameter ω(a), one could obtain
389
+ the corresponding scalar field φ(a) given by Eq.
390
+ (25)
391
+ with potential Vs(a) given by Eq. (26). As long as φ(a)
392
+ is invertible (or, in other words, a bijective map from
393
+ (0, 1] → R), we obtain a(φ), at least in principle, using
394
+ which, we get Vs(φ).
395
+ Alternatively, given a scalar field φ(a), one can obtain
396
+ the corresponding perfect fluid ˆT (or the ω(a) by which
397
+ it is identified), using Eq.(24).
398
+ On the other hand, we can also start with a given scalar
399
+ field potential V (φ). One can use Eq.(18) and Eq.(21) to
400
+ obtain the ordinary nonlinear differential equation
401
+ H
402
+
403
+ a, φ, dφ
404
+ da , d2φ
405
+ da2
406
+
407
+ = 0,
408
+ (27)
409
+ that can be solved in principle, to obtain φ(a), and later
410
+ obtain ω(a) using Eq.(24). Hence, given a scalar field po-
411
+ tential Vs(φ), one can obtain the corresponding ˆT (iden-
412
+ tified by ω(a)) in the above manner.
413
+ B.
414
+ Vector field collapse
415
+ We prove that any SH perfect fluid is equivalent to a
416
+ SH vector field ˜A(a) with a suitable potential Vv(a), as far
417
+ as the gravitational collapse is concerned. If B(a) is in-
418
+ vertible, then the following statement holds: Any SH per-
419
+ fect fluid is gravitationally equivalent to a SH vector field
420
+ ˜A with a suitable potential Vv(B) (where B = g( ˜A, ˜A)).
421
+ Consider a vector field
422
+ ˜A : M → TM.
423
+ (28)
424
+ with potential V (B). For a fixed p ∈ M, ˜A(p) = Aµdxµ,
425
+ where Aµ = (A0, Ai), 1 < i < 3 (in the comoving carte-
426
+ sian coordinate basis). Here B = gαβAαAβ. We con-
427
+ sider a SH pure vector field: A0 = 0 and Ai = A ∈ R
428
+ ��i ∈ (1, 2, 3). For such a vector field, B = 3A2/a2.
429
+ The Lagrangian of a massless vector field ˜A with po-
430
+ tential Vv(B) is given by
431
+ L ˜
432
+ A = −1
433
+ 4F µνFµν − Vv(B).
434
+ (29)
435
+ F is a two form called the field strength and can be writ-
436
+ ten in terms of wedge product as F = Fµνdxµ ∧ dxν.
437
+ The field strength is the exterior derivative of the vec-
438
+ tor field ˜A, i.e.F = d ˜A. The components are written as
439
+ Fµν = ∇µAν − ∇νAµ.
440
+ The stress-energy tensor is obtained from the La-
441
+ grangian L ˜
442
+ A as
443
+ Tµν = −
444
+ 2
445
+ √−g
446
+ δ (√−gL ˜
447
+ A)
448
+ δgµν
449
+ .
450
+ (30)
451
+ This gives us
452
+ Tµν = −1
453
+ 4FαβF αβgµν −Vv(B)gµν +FµαF
454
+ α
455
+ ν
456
+ +2V ′
457
+ vAµAν.
458
+ (31)
459
+ The overhead prime denotes the ordinary derivative with
460
+ respect to B. The density and the isotropic pressure are
461
+ subsequently expressed in terms of the time derivative of
462
+ the vector field component and its potential as
463
+ ρv = 3
464
+ 2
465
+ ˙A2
466
+ a2 + Vv(B),
467
+ (32)
468
+ and
469
+ pv = 1
470
+ 2
471
+ ˙A2
472
+ a2 − Vv(B) + 2V ′
473
+ v
474
+ A2
475
+ a2 .
476
+ (33)
477
+ We now equate ρv = ρ and pv = p.
478
+ From Eq.(32)
479
+ and Eq.(4), we obtain
480
+ Vv = ρv
481
+
482
+ 1 − 1
483
+ 2A,2
484
+ a
485
+
486
+ .
487
+ (34)
488
+ Substituting for ρ(a) from Eq.(11), we obtain
489
+ Vv = ρ0 exp
490
+ �� 1
491
+ a
492
+ 3 (1 + ω(a))
493
+ a
494
+ da
495
+ � �
496
+ 1 − 1
497
+ 2A,2
498
+ a
499
+
500
+ (35)
501
+ On differentiating Eq.(34) with respect to B we obtain,
502
+ V ′
503
+ v =
504
+ ρv,a
505
+
506
+ 1 − A,2
507
+ a
508
+ 2
509
+
510
+ − ρvA,a A,aa
511
+ 6A2
512
+ a2
513
+
514
+ A,a
515
+ A − 1
516
+ a
517
+
518
+ (36)
519
+ Using Eq.(33), Eq.(4), and Eq.(5), we obtain
520
+ aρv,a
521
+ 3
522
+ + ρv
523
+
524
+ 1 + 1
525
+ 6A,2
526
+ a
527
+
528
+ = Vv − V ′
529
+ v
530
+ A2
531
+ a2 .
532
+ (37)
533
+ Substituting for Vv and V ′
534
+ v from Eq.(34) and Eq.(36),
535
+ and also substituting for ρ,a (by differentiating Eq.(11))
536
+ in Eq.(37), we obtain a second order nonlinear differential
537
+ equation
538
+ G
539
+
540
+ a, ω, A, dA
541
+ da , d2A
542
+ da2
543
+
544
+ = 0,
545
+ (38)
546
+ where G is
547
+ G =d2A
548
+ da2 − 4
549
+ A
550
+ �dA
551
+ da
552
+ �2
553
+ + 1
554
+ 2a (5 − 3ω) dA
555
+ da + 6
556
+ A (1 + ω)
557
+ − 3 (1 + ω)
558
+ a
559
+ �dA
560
+ da
561
+ �−1
562
+ .
563
+ (39)
564
+ For a fixed ω(a), solving this differential equation with
565
+ two initial conditions gives us A(a), and consequently,
566
+ the vector field ˜A.
567
+ Hence, we proved that given the functional form of the
568
+ equation of state parameter ω(a), one could obtain the
569
+ corresponding vector field ˜A using Eq.(38), and conse-
570
+ quently, the vector field potential Vv(a) using Eq.(35).
571
+
572
+ 5
573
+ SH Perfect Fluid
574
+ Characterised by the equation of state
575
+ parameter 𝜔(a)
576
+ SH Scalar Field
577
+ Characterised by 𝜙(a)
578
+ SH Vector Field
579
+ Characterised by Ã(a)
580
+ FIG. 1: A spatially homogeneous (SH) perfect fluid (governed by a flat FLRW spacetime metric) is completely
581
+ characterized by the equation of state parameter ω(a), Eq.(10) of the matter field. This matter field is obtained
582
+ from fundamental matter Lagrangian. Hence, the same matter field is also characterized by an SH scalar field φ(a),
583
+ Eq.(25) or its potential Vs(a), Eq.(26) (Vs(φ) if φ(a) is invertible). Similarly, it can also be characterized by an SH
584
+ vector field ˜A(a), Eq.(38) or its potential Vv(a), Eq.(35) (Vv(B) if B(a) is invertible). This schematic diagram
585
+ depicts the equivalence between the gravitational collapse of SH Perfect fluid, Scalar field and Vector field. By
586
+ spatial homogeneity, we mean homogeneous on a three-dimensional spacelike orbit with a six-dimensional isometry
587
+ group G6 corresponding to the spacetime [3].
588
+ Now, from the functional form A(a), we obtain B(a). As
589
+ long as B(a) is invertible (or, in other words, a bijec-
590
+ tive map from (0, 1] → R), we obtain a(B), at least in
591
+ principle, using which, we get Vv(B).
592
+ Alternatively, given a vector field ˜A(a), one can obtain
593
+ the corresponding perfect fluid ˆT (or the ω(a) by which
594
+ it is identified), using Eq.(38).
595
+ On the other hand, we can also start with a given vec-
596
+ tor field potential Vv(B). One can differentiate Eq.(35),
597
+ and do some rearrangements to obtain
598
+ ω(a, A, dA
599
+ da , d2A
600
+ da2 )
601
+ as
602
+ ω = 2AV ′
603
+ aV
604
+ �A
605
+ a − dA
606
+ da
607
+
608
+ −a
609
+ 3
610
+ dA
611
+ da
612
+ d2A
613
+ da2
614
+
615
+ 1 − 1
616
+ 2
617
+ �dA
618
+ da
619
+ �2�−1
620
+ −1.
621
+ (40)
622
+ Substituting Eq.(40) in Eq.(38), we obtain
623
+ ˜G
624
+
625
+ a, A, dA
626
+ da , d2A
627
+ da2
628
+
629
+ = 0.
630
+ (41)
631
+ In principle, this differential equation can be solved to
632
+ obtain A(a), which, when substituted in Eq.(40), gives
633
+ us ω(a). Hence, given a vector field potential V (B), one
634
+ can obtain the corresponding ˆT (identified by ω(a)) in
635
+ the above manner.
636
+ III.
637
+ EXTERIOR GENERALIZED VAIDYA
638
+ SPACETIME
639
+ The collapsing vector field spacetime (g−
640
+ µν) can be
641
+ joined smoothly with the exterior generalized Vaidya
642
+ spacetime (g+
643
+ µν) so that their union forms a valid solution
644
+ of the Einstein’s field equations. The interior FLRW and
645
+ the exterior generalized Vaidya spacetime [19] are respec-
646
+ tively given as
647
+ ds2
648
+ − = −dt2 + a(t)2dr2 + r2
649
+ ba(t)2dΩ2,
650
+ (42)
651
+ and
652
+ ds2
653
+ + = −
654
+
655
+ 1 − 2M(R, v)
656
+ R
657
+
658
+ dv2 − 2dvdR + R2dΩ2. (43)
659
+ Here, v is the retarded null coordinate, R is the general-
660
+ ized Vaidya radius, and rb is the value of the radial co-
661
+ ordinate r corresponding to the matching hypersurface,
662
+ or in other words, the radial coordinate of the outermost
663
+ shell of the collapsing scalar/vector field cloud. The mat-
664
+ ter field corresponding to the generalized Vaidya space-
665
+ time is a combination of Type I and type II, such that
666
+ the components of the stress-energy tensor written in the
667
+
668
+ 6
669
+ orthonormal basis appear as
670
+ Tab =
671
+
672
+
673
+
674
+
675
+
676
+
677
+
678
+
679
+
680
+
681
+
682
+
683
+
684
+
685
+
686
+ ¯ϵ
687
+ 2 + ϵ
688
+ ¯ϵ
689
+ 2
690
+ 0
691
+ 0
692
+ ¯ϵ
693
+ 2
694
+ ¯ϵ
695
+ 2 − ϵ 0
696
+ 0
697
+ 0
698
+ 0
699
+ P
700
+ 0
701
+ 0
702
+ 0
703
+ 0 P.
704
+
705
+
706
+
707
+
708
+
709
+
710
+
711
+
712
+
713
+
714
+
715
+
716
+
717
+
718
+
719
+ (44)
720
+ ϵ = P = 0 and ¯ϵ ̸= 0 corresponds the usual Vaidya
721
+ spacetime as a special case. ¯ϵ = 0 and ϵ ̸= 0 corresponds
722
+ to a sub-class of Type I matter field. The generalized
723
+ Vaidya solution encompasses many known Einstein field
724
+ equations solutions. Matching the first and second fun-
725
+ damental forms for the interior and exterior metric on Σ
726
+ gives the following equations:
727
+ R(t) = R(t, rb) (= rba(t)) ,
728
+ (45)
729
+ F(t, rb) = 2M(R, v),
730
+ (46)
731
+ �dv
732
+ dt
733
+
734
+ Σ
735
+ =
736
+ 1 + ˙R
737
+ 1 − F (t,rb)
738
+ R
739
+ ,
740
+ (47)
741
+ and
742
+ M(R, v),R = F(t, rb)
743
+ 2R
744
+ + R ¨R.
745
+ (48)
746
+ Here, F = R ˙R2 is the Misner-Sharp mass function of
747
+ the collapsing spherical SH perfect fluid. Using the rela-
748
+ tion (45), we can relate the generalized Vaidya mass with
749
+ the density of the interior collapsing SH spherical perfect
750
+ fluid cloud as
751
+ M = ρ
752
+ 6R3.
753
+ (49)
754
+ Using Eq.(7), differentiation of Eq.(25) with respect to
755
+ a, and Eq.(49), in Eq.(48) we get
756
+ M,R = 3M
757
+ R
758
+
759
+ 1 + (1 + ω(a))r2
760
+ b
761
+ R2
762
+
763
+ ,
764
+ (50)
765
+ integrating which we obtain
766
+ M(R, v) = M1(v) exp
767
+ ��
768
+ 3
769
+ R
770
+
771
+ 1 + (1 + ˜ω (R))r2
772
+ b
773
+ R2
774
+
775
+ dR
776
+
777
+ .
778
+ (51)
779
+ Here M1(v) is a constant of integration and is a function
780
+ of null coordinate v, and
781
+ ˜ω(R) = ω
782
+ �R
783
+ rb
784
+
785
+ .
786
+ Eq.(51) gives us the expression of the generalized Vaidya
787
+ mass function of the exterior generalized Vaidya space-
788
+ time, in terms of interior collapsing perfect fluid equation
789
+ of state parameter ω, to ensure smooth matching at the
790
+ matching hypersurface.
791
+ For the exterior matter field to satisfy the weak energy
792
+ condition, ¯ϵ and ϵ should be non-negative [19].
793
+ These
794
+ inequalities, in turn, put restrictions on the generalized
795
+ Vaidya mass function as
796
+ M,v ≤ 0,
797
+ and
798
+ M,R ≥ 0.
799
+ (52)
800
+ Using Eq.(50) and Eq.(51) in the above two relations, we
801
+ obtain
802
+ M1,v ≤ 0,
803
+ (53)
804
+ and
805
+ �1 + ω(a)
806
+ a2
807
+
808
+ ≥ 0.
809
+ (54)
810
+ The inequality (54) is always satisfied if the interior col-
811
+ lapsing matter field obeys the weak energy condition.
812
+ Hence, Eq.(53) is the only restriction on the generalized
813
+ Vaidya mass function for the exterior spacetime to obey
814
+ at least the weak energy condition.
815
+ Now,
816
+ we have a complete solution of Einstein’s
817
+ field equations consisting of an interior collapsing SH
818
+ scalar/vector field (with some potential) and the exte-
819
+ rior generalized Vaidya solution, matched smoothly at
820
+ the matching hypersurface. The free functions are cat-
821
+ egorically the potential function (Vs(φ) in case of scalar
822
+ field collapse, and Vv(B) in case of vector field collapse),
823
+ and the component of generalized Vaidya mass function
824
+ M1(v), the latter one restricted by the inequality (53).
825
+ It is evident that the choice of M1(v) does not affect the
826
+ causal structure of the spacetime obtained as an end-
827
+ state of unhindered gravitational collapse.
828
+ Of course,
829
+ instead of considering the potential function Vs(φ) (or
830
+ Vv(B)) as a free function, one could also consider any one
831
+ of the remaining functions: ω(a), ρ(a), φ(a) (or A(a)),
832
+ Vs(a) (or Vv(a)) as a free function, without any trouble.
833
+ In the next section, we study the end state of this class of
834
+ global dynamical spacetime identified by any one of the
835
+ free functions.
836
+ IV.
837
+ CAUSAL STRUCTURE AND STRENGTH
838
+ OF THE SINGULARITY
839
+ Once the singularity is formed as an end state of grav-
840
+ itational collapse of the interior scalar (vector) field with
841
+ potential Vs(φ) (Vv(B)), one can investigate whether or
842
+ not causal geodesics can escape the singularity. Addition-
843
+ ally, one can investigate whether or not such singularity
844
+ is gravitationally strong in the sense of Tipler. The fol-
845
+ lowing two subsections discuss these two properties.
846
+
847
+ 7
848
+ Massless scalar field
849
+ Vs(φ) = 0
850
+ φ(a) = c ±
851
+
852
+ 6 log a
853
+ strong
854
+ BH
855
+ Homogeneous dust (ω = 0)
856
+ Vs(φ) ∝ exp
857
+ �√
858
+
859
+
860
+ φ(a) = c ±
861
+
862
+ 3 log a
863
+ strong
864
+ BH
865
+ Goswami/ Joshi [17] (ω = − 2
866
+ 3) (SF1)
867
+ Vs(φ) ∝ exp φ
868
+ φ(a) = c ± log a
869
+ strong
870
+ NS
871
+ Two dimensional analog of Mexican hat [18]
872
+ (SF2)
873
+ Vs(φ) = 1
874
+ 2µφ2 + λφ4
875
+ φ(a) = ±2
876
+
877
+ 2√c − log a
878
+ weak
879
+ NS
880
+ TABLE I: Four examples of spatially homogeneous scalar fields that collapse to form a singularity that is either
881
+ hidden (blackhole or BH) or (naked singularity or NS). In the fourth example, µ = − 16
882
+ 3 λ. The first three types end
883
+ up in a strong singularity in the sense of Tipler.
884
+ Massless vector field
885
+ Vv(B) = 0
886
+ strong
887
+ BH
888
+ Massive vector field
889
+ Vv(B) = − 1
890
+ 2µ2B
891
+ strong
892
+ BH
893
+ VF1
894
+ Vv(a) as in Fig.(3)
895
+ strong
896
+ NS
897
+ VF2
898
+ Vv(a) as in Fig.(3)
899
+ weak
900
+ NS
901
+ TABLE II: Four examples of spatially homogeneous vector fields that collapse to form a singularity that is either
902
+ hidden within a black hole (BH) or is naked (NS). The ones mentioned in the third and the fourth row are newly
903
+ constructed vector fields from known scalar fields (mentioned in the third [17] and the fourth [18] row of Table 1,
904
+ respectively) by exploiting the gravitational equivalence depicted in Fig.(1). The corresponding vector field
905
+ component A(a) for each case is plotted in Fig.(2-3). The first three types end up in a gravitationally strong
906
+ singularity in the sense of Tipler.
907
+ 0.0
908
+ 0.2
909
+ 0.4
910
+ 0.6
911
+ 0.8
912
+ 1.0
913
+ -0.5
914
+ 0.0
915
+ 0.5
916
+ 1.0
917
+ a
918
+ A
919
+ a
920
+ 0.0
921
+ 0.2
922
+ 0.4
923
+ 0.6
924
+ 0.8
925
+ 1.0
926
+ 0
927
+ 2
928
+ 4
929
+ 6
930
+ a
931
+ V
932
+ FIG. 2: The dynamics of the vector field component A(a) in the case of the massive (µ = 1) vector field ˜A (Left
933
+ panel) and its potential Vv(a) (Right panel). First we obtain ω(a, A, dA
934
+ da , d2A
935
+ da2 ) by substituting Vv(B) = − 1
936
+ 2µ2B in
937
+ Eq.(40). Substituting for ω(a, A, dA
938
+ da , d2A
939
+ da2 ) in Eq.(41) and solving the differential equation with initial conditions
940
+ A(1) = 1 and A′(1) = 2, we obtain A(a). Consequently, substituting Vv(B) = − 1
941
+ 2µ2B, and the obtained A(a) in
942
+ Eq.(40), we obtain ω(a), Further substitution of ω(a) in Eq.(35), we obtain Vv(a).
943
+
944
+ 8
945
+ 0.2
946
+ 0.4
947
+ 0.6
948
+ 0.8
949
+ 1.0
950
+ -3
951
+ -2
952
+ -1
953
+ 0
954
+ 1
955
+ 2
956
+ 3
957
+ a
958
+ Vv
959
+ (a)
960
+ 0.0
961
+ 0.5
962
+ 1.0
963
+ 1.5
964
+ 2.0
965
+ -10
966
+ -8
967
+ -6
968
+ -4
969
+ -2
970
+ 0
971
+ 2
972
+ 4
973
+ B
974
+ Vv
975
+ (b)
976
+ 0.3
977
+ 0.4
978
+ 0.5
979
+ 0.6
980
+ 0.7
981
+ 0.8
982
+ 0.9
983
+ 1.0
984
+ -300
985
+ -200
986
+ -100
987
+ 0
988
+ 100
989
+ 200
990
+ a
991
+ Vv
992
+ (c)
993
+ 0.0
994
+ 0.5
995
+ 1.0
996
+ 1.5
997
+ 2.0
998
+ -800
999
+ -600
1000
+ -400
1001
+ -200
1002
+ 0
1003
+ 200
1004
+ B
1005
+ Vv
1006
+ (d)
1007
+ A1
1008
+ A2
1009
+ 0.3
1010
+ 0.4
1011
+ 0.5
1012
+ 0.6
1013
+ 0.7
1014
+ 0.8
1015
+ 0.9
1016
+ 1.0
1017
+ 0.0
1018
+ 0.2
1019
+ 0.4
1020
+ 0.6
1021
+ 0.8
1022
+ 1.0
1023
+ a
1024
+ A
1025
+ (e)
1026
+ FIG. 3: (a) and (c): Vector field potentials Vv(a) corresponding to newly constructed vector fields VF1 (orange)
1027
+ and VF2 (green), as mentioned in the third and fourth row of the Table (II), respectively. (b) and (d): The same
1028
+ vector field potentials Vv(B) as function of B. (e): The vector field components A(a) in both of these cases. In the
1029
+ latter example, µ = −8/3 and λ = 1. First, we obtain ωi(a) , using Eq.(25) (Here i ∈ 1, 2 corresponds to VF1 and
1030
+ VF2 respectively). Then we obtain the vector field components Ai(a) by solving the differential Eq.(38) with initial
1031
+ conditions Ai(1) = 1 and A′
1032
+ i(1) = 10. Further substitution of ωi(a) and the obtained Ai(a) in Eq.(35), we get
1033
+ Vv(i)(a). Once Vv(i)(a) is obtained, we obtain Vv(i)(B).
1034
+ A.
1035
+ Causal structure of the singularity
1036
+ We say that a singularity formed due to unhindered
1037
+ gravitational collapse is naked if there exists a family of
1038
+ outgoing causal curves whose past endpoint is the singu-
1039
+ larity. In the future, these curves can either reach a far-
1040
+ away observer or fall back to the singularity. The singu-
1041
+ larities are then termed globally naked and locally naked,
1042
+ respectively. Whether or not the singularity is naked es-
1043
+ sentially depends on the geometry of trapped surfaces as
1044
+ the collapse evolves. Trapped surfaces are two-surfaces in
1045
+ the spacetime on which not only the ingoing congruence
1046
+ but also the outgoing congruence necessarily converge.
1047
+ Convergence or otherwise of the outgoing null geodesic
1048
+ congruence is determined by the behaviour of its expan-
1049
+ sion scalar, which we denote here as θl (t, r). It is ex-
1050
+ pressed in terms of the metric coefficients, in comoving
1051
+ spherical coordinates as,
1052
+ θl = 2
1053
+ R
1054
+
1055
+ 1 −
1056
+
1057
+ ρR2
1058
+ 3
1059
+
1060
+ .
1061
+ (55)
1062
+
1063
+ 9
1064
+ EH
1065
+ AH
1066
+ 0.0
1067
+ 0.1
1068
+ 0.2
1069
+ 0.3
1070
+ 0.4
1071
+ R
1072
+ 0.0
1073
+ 0.1
1074
+ 0.2
1075
+ 0.3
1076
+ 0.4
1077
+ 0.5
1078
+ t
1079
+ (a) Massless scalar field (Vs = 0)
1080
+ EH
1081
+ AH
1082
+ 0.0
1083
+ 0.1
1084
+ 0.2
1085
+ 0.3
1086
+ 0.4
1087
+ R
1088
+ 0.00
1089
+ 0.05
1090
+ 0.10
1091
+ 0.15
1092
+ 0.20
1093
+ 0.25
1094
+ 0.30
1095
+ 0.35
1096
+ t
1097
+ (b) Massless vector field (Vv = 0)
1098
+ 0.0
1099
+ 0.1
1100
+ 0.2
1101
+ 0.3
1102
+ 0.4
1103
+ R
1104
+ 0.0
1105
+ 0.5
1106
+ 1.0
1107
+ 1.5
1108
+ 2.0
1109
+ 2.5
1110
+ t
1111
+ (c) SF1/VF1
1112
+ 0.0
1113
+ 0.1
1114
+ 0.2
1115
+ 0.3
1116
+ 0.4
1117
+ R
1118
+ 0.0
1119
+ 0.5
1120
+ 1.0
1121
+ 1.5
1122
+ 2.0
1123
+ 2.5
1124
+ t
1125
+ (d) SF2/VF2
1126
+ FIG. 4: Spacetime diagram of the examples of spatially homogeneous scalar fields and vector fields mentioned in
1127
+ Tables I and II. The solid black curve in each of them represents the boundary of the collapsing cloud. Upper panel:
1128
+ The singularity is not visible in both examples. Lower Panel: In the case of SF1/VF1, the singularity forms in a
1129
+ finite comoving time and is globally visible because of the absence of the apparent and event horizons. In the case of
1130
+ SF2/VF2, the singularity forms in an infinite comoving time. However, an ultra-high density region is obtained in
1131
+ finite comoving time, which can be visible globally because of the absence of the apparent and event horizons.
1132
+ The region in which θl < 0 is called the trapped region.
1133
+ The boundary of the trapped region, given by θl = 0, is
1134
+ called the apparent horizon. If the neighbourhood of the
1135
+ singular center is surrounded by a trapped region since
1136
+ before the time of formation of the singularity ts, then
1137
+ it is covered, and we get a black hole. Hence, the nec-
1138
+ essary condition for singular null geodesic congruence to
1139
+ escape the singularity is the absence of a trapped region,
1140
+ which is ensured by the condition θl(ts, r) > 0 for such
1141
+ congruence. The absence of trapped region in the neigh-
1142
+ bourhood of the singularity (t, r) = (ts, 0) is ensured by
1143
+ the following inequality:
1144
+ lim
1145
+ t→ts
1146
+ ρR2
1147
+ 3
1148
+ ≤ lim
1149
+ a→0
1150
+ ρ(a)r2
1151
+ ba2
1152
+ 3
1153
+ < 1,
1154
+ (56)
1155
+ The inequality (56) is definitely satisfied if
1156
+ lim
1157
+ a→0 ρ(a) < 1
1158
+ a2 .
1159
+ (57)
1160
+ For
1161
+ lim
1162
+ a→0 ρ(a) = k
1163
+ a2 ,
1164
+ for some k ∈ R+, the inequality is satisfied only for
1165
+ rb <
1166
+
1167
+ 3/k. Rewriting the inequality (57) in terms of
1168
+ the equation of state parameter ω(a) using Eq.(11), one
1169
+ obtains
1170
+ lim
1171
+ a→0 ρ0a2 exp
1172
+ �� 1
1173
+ a
1174
+ 3 (1 + ω(a))
1175
+ a
1176
+
1177
+ da < 1.
1178
+ (58)
1179
+ If a collapsing matter field with the equation of state
1180
+ parameter ω(a) satisfies the inequality (58), then it will
1181
+
1182
+ 10
1183
+ up in a naked singularity [17]. In the case of otherwise,
1184
+ the final outcome is a black hole.
1185
+ Hence, as decided by the above inequality, we get a
1186
+ class of SH matter fields that include scalar and vector
1187
+ fields, identified by the functional form ω(a), that goes
1188
+ to either the blackhole or naked singularity final state as
1189
+ an end state of unhindered gravitational collapse.
1190
+ In the case of scalar field collapse, the restriction (58)
1191
+ on ω(a) gives us a restriction on the scalar field φ(a) us-
1192
+ ing Eq.(25), and the scalar field potential function Vs(a)
1193
+ using Eq.(26). Hence, obtaining a class of ω(a) is grav-
1194
+ itationally equivalent to obtaining a class of scalar field
1195
+ potentials Vs(a) that goes to the naked singularity as
1196
+ an end state of unhindered gravitational collapse. More-
1197
+ over, suppose φ(a) is a bijective map from (0, 1] → R.
1198
+ In that case, obtaining a class of ω(a) is gravitationally
1199
+ equivalent to obtaining a class of scalar field potentials
1200
+ Vs(φ) = Vs(a(φ)) that goes to the naked singularity as
1201
+ an end state of gravitational collapse.
1202
+ Similarly, in the case of vector field collapse, the re-
1203
+ striction (58) on ω(a) gives us a restriction on the vector
1204
+ field ˜A (or more specifically, a restriction on the vec-
1205
+ tor field component A(a)) obtained by solving the dif-
1206
+ ferential Eq.(38), and the vector field potential function
1207
+ Vv(a) obtained by substituting A(a) and ρ from Eq.(11),
1208
+ in Eq.(34).
1209
+ Hence, obtaining a class of ω(a) is gravi-
1210
+ tationally equivalent to obtaining a class of vector field
1211
+ potential Vv(a) that goes to the naked singularity as an
1212
+ end state of unhindered gravitational collapse.
1213
+ More-
1214
+ over, suppose A(a) is a bijective map from (0, 1] → R.
1215
+ In that case, obtaining a class of ω(a) is gravitationally
1216
+ equivalent to obtaining a class of vector field potential
1217
+ Vv(A) = Vv(a(A)) that goes to the naked singularity as
1218
+ an end state of unhindered gravitational collapse.
1219
+ In Table (I) and (II), we discuss examples of such
1220
+ scalar field collapse and vector field collapse that ends
1221
+ up in either a black hole or a naked singularity.
1222
+ Ex-
1223
+ ploiting the equivalence between SH perfect fluids, scalar
1224
+ fields with potential Vs(a), and vector fields with poten-
1225
+ tial Vv(a), we construct two examples of collapsing vector
1226
+ fields with potential out-of-known examples of collapsing
1227
+ scalar fields with potentials, giving rise to the naked sin-
1228
+ gularity as an end state.
1229
+ The first example of a collapsing vector field with po-
1230
+ tential Vv(a) is constructed from the scalar field with po-
1231
+ tential mentioned in the third row of Table (I) [17]. The
1232
+ perfect fluid corresponding to such scalar field example
1233
+ has an equation of state parameter ω(a) = − 2
1234
+ 3. The con-
1235
+ structed collapsing vector field ˜A = (0, A, A, A) (in the
1236
+ comoving coordinate basis) has the property (dynamics
1237
+ of A(a) and Vv(a)) as shown in Fig.(3). Refer to the third
1238
+ row of Table (II).
1239
+ The second example of a collapsing vector field with
1240
+ potential Vv(a) is constructed from the scalar field with
1241
+ potential mentioned in the fourth row of Table (I) [18].
1242
+ Such a scalar field has a two-dimensional analogue of
1243
+ Mexican hat-shaped potential. The constructed collaps-
1244
+ ing vector field ˜A = (0, A, A, A) (in the comoving co-
1245
+ ordinate basis) has the property (dynamics of A(a) and
1246
+ Vv(a)) as shown in Fig.(3). Refer to the fourth row of
1247
+ Table (II). The spacetime diagrams of some of the ex-
1248
+ amples in Table (I) and (II) are plotted in Fig.(4).
1249
+ B.
1250
+ Strength of the singularity
1251
+ Generally, a singularity in the spacetime manifold is
1252
+ identified by the existence of at least one past/future in-
1253
+ complete geodesic. However, in the case of singularities
1254
+ forming as the end state of a gravitational collapse, apart
1255
+ from the existence of such incomplete geodesics, one ex-
1256
+ pects an additional physical property as follows: An ob-
1257
+ ject approaching such singularity should be crushed to
1258
+ zero volume. We call such a singularity gravitationally
1259
+ strong in the sense of Tipler [20]. A precise definition of
1260
+ a strong singularity is as follows:
1261
+ Consider a smooth spacetime manifold (M, g) and a
1262
+ causal geodesic γ : [t0, 0) → M. Let λ be an affine pa-
1263
+ rameter along this geodesic. Let ξ(i), (0 ≤ i ≤ 2 in the
1264
+ case of null geodesic, 0 ≤ i ≤ 3 in the case of timelike
1265
+ geodesic) be the independent Jacobi vector fields. The
1266
+ wedge product of these Jacobi fields gives us the volume
1267
+ form V = � ξ(i). We say that a singularity is gravita-
1268
+ tionally strong in the sense of Tipler if this volume form
1269
+ vanishes as λ → 0.
1270
+ Clarke and Krolak [21] related the existence of a Tipler
1271
+ strong singularity with the growth rate of the curvature
1272
+ terms as follows: At least along one null geodesic with
1273
+ affine parameter λ (such that λ → 0 as the singularity is
1274
+ approached), the following inequality
1275
+ lim
1276
+ λ→0 λ2RijKiKj > 0
1277
+ (59)
1278
+ should hold for the singularity to be strong in the sense of
1279
+ Tipler. Here Ki = dxi
1280
+ dλ are the tangents to the chosen null
1281
+ geodesic, and xi is the coordinate system. This condition
1282
+ puts a lower bound on the growth of the curvature scalar.
1283
+ In the spherical coordinate system (t, r, θ, φ), the radial
1284
+ null geodesic equation reads
1285
+ dt
1286
+ dr = a.
1287
+ (60)
1288
+ Hence, we have the relation between the tangents Kt and
1289
+ Kr as
1290
+ Kt = aKr,
1291
+ (61)
1292
+ and subsequently, in terms of the affine parameter,
1293
+ Kt = R
1294
+ λ ,
1295
+ and
1296
+ Kr = r
1297
+ λ.
1298
+ (62)
1299
+ The inequality (59) can then be written in terms of ω as
1300
+ lim
1301
+ a→0
1302
+
1303
+ r2(1 + ω)ρ0 exp
1304
+ �� 1
1305
+ a
1306
+ 3(1 + ω)
1307
+ a
1308
+ da
1309
+ ��
1310
+ > 0
1311
+ (63)
1312
+
1313
+ 11
1314
+ Hence, the singularity formed due to the gravitational
1315
+ collapse of a scalar/vector field is strong in the sense of
1316
+ Tipler if the following inequality holds (assuming that
1317
+ the weak energy condition is respected):
1318
+ lim
1319
+ a→0 exp
1320
+ �� 1
1321
+ a
1322
+ 3(1 + ω)
1323
+ a
1324
+ da
1325
+
1326
+ > 0.
1327
+ (64)
1328
+ Hence, (along with using the condition (58)) one can ob-
1329
+ tain a naked singularity that is strong in the sense of
1330
+ Tipler for that ω that satisfies the following constraint:
1331
+ 0 < lim
1332
+ a→0 exp
1333
+ �� 1
1334
+ a
1335
+ 3(1 + ω)
1336
+ a
1337
+ da
1338
+
1339
+ < O(a−2).
1340
+ (65)
1341
+ This constraint gives us the class of SH collapsing mat-
1342
+ ter fields that we identify by ω(a), which ends up in
1343
+ strong curvature naked singularity. Or in other words,
1344
+ we have a class of scalar/vector field potentials corre-
1345
+ sponding to the given scalar/vector field that collapses
1346
+ to a strong naked singularity.
1347
+ As an example, in Ta-
1348
+ bles (I) and (II), we mention the causal property and
1349
+ the strength of the singularity formed due to the gravi-
1350
+ tational collapse of four different scalar/vector fields.
1351
+ V.
1352
+ CONCLUSIONS AND REMARKS
1353
+ Following are the concluding remarks:
1354
+ 1. Unlike the singularity theorems that provide rig-
1355
+ orous proof of the existence of incomplete causal
1356
+ geodesics under rather generic conditions, one does
1357
+ not currently have proof or disproof of the cosmic
1358
+ censorship hypothesis.
1359
+ In fact, we need a math-
1360
+ ematically rigorous formulation of this conjecture,
1361
+ which is not available currently, before we can prove
1362
+ or disprove it.
1363
+ Under the situation at present, we can only spec-
1364
+ ulate its validity or otherwise.
1365
+ Proposed coun-
1366
+ terexamples, hence have great importance in under-
1367
+ standing whether naked singularities, in fact, exist
1368
+ or not in our universe. Through such analysis of
1369
+ gravitational collapse models only, one could pos-
1370
+ sibly hope to arrive at a suitable formulation of
1371
+ cosmic censorship. The collapse of inhomogeneous
1372
+ dust and the Vaidya null fluids were the first exam-
1373
+ ples proposed to produce naked singularities. How-
1374
+ ever, an important objection could be that, even if
1375
+ astrophysically interesting, they are not fundamen-
1376
+ tal forms of matter [7, 22].
1377
+ One could then ask
1378
+ whether the collapse of matter configuration that
1379
+ is obtained from a fundamental matter Lagrangian
1380
+ ends up in a naked singularity. Scalar fields with
1381
+ potential and vector fields with potential are fun-
1382
+ damental matter fields in this sense. Here we show
1383
+ that not just one particular choice of these fields
1384
+ but an entire class of such types could collapse and
1385
+ form a naked singularity as an end state. This basi-
1386
+ cally divides the allowed class of potential functions
1387
+ into classes that take the unhindered collapse to a
1388
+ black hole or naked singularity.
1389
+ 2. To achieve this, we show equivalence between SH
1390
+ (a) Perfect fluid: characterized by ω(a),
1391
+ (b) Massless scalar field φ: characterized by φ(a)
1392
+ or its potential Vs(a) or Vs(φ) (if φ(a) is in-
1393
+ vertible), and
1394
+ (c) Massless vector field
1395
+ ˜A:
1396
+ characterized by
1397
+ A(a), or its potential Vv(a), or Vv(B) (if B(a)
1398
+ is invertible).
1399
+ as far as the gravitational collapse is concerned.
1400
+ This gravitational equivalence is described in sub-
1401
+ sections of section (II) and depicted in Fig.(1).
1402
+ Now, if the functional form of ω(a) satisfies the
1403
+ inequality (58), then the singular null geodesic con-
1404
+ gruence, if at all there exists, does not get trapped
1405
+ as a → 0.
1406
+ Hence, we have a class of functions
1407
+ ω(a) corresponding to a naked singularity as an end
1408
+ state of gravitational collapse. Now, because of the
1409
+ above equivalence, in the case of an SH scalar field
1410
+ collapse, one then has a class of scalar field func-
1411
+ tion φ(a), or a class of scalar field potential Vs(a),
1412
+ or a class of scalar field potential in terms of φ,
1413
+ i.e. Vs(φ) (provided φ(a) is invertible), that corre-
1414
+ sponds to the naked singularity as an end state.
1415
+ Similarly, in the case of an SH vector field col-
1416
+ lapse, one has a class of vector field component
1417
+ function A(a), or a class of vector field potential
1418
+ Vv(a), or a class of vector field potential in terms
1419
+ of B = g( ˜A, ˜A), i.e. Vs(B) (provided B(a) is in-
1420
+ vertible), that corresponds to the naked singularity
1421
+ as an end state.
1422
+ 3. A naked singularity formed due to gravitational col-
1423
+ lapse may or may not be relevant if they are not
1424
+ gravitationally strong in the sense of Tipler [20].
1425
+ Here, we show a class of ω(a) that satisfies the in-
1426
+ equalities (65) that corresponds to the formation
1427
+ of a strong curvature naked singularity. Using ar-
1428
+ guments similar to point no. 2 of this section, we
1429
+ have equivalently shown a class of scalar field po-
1430
+ tential (in case of scalar field collapse) and a class
1431
+ of vector field potential (in case of vector field col-
1432
+ lapse) that corresponds to a strong curvature naked
1433
+ singularity.
1434
+ 4. For the sake of completion, we study the global
1435
+ spacetime, consisting of the interior collapsing
1436
+ scalar/vector field and the exterior generalized
1437
+ Vaidya spacetime. The smooth matching demands
1438
+ a restriction on the free function, that is, the gener-
1439
+ alized Vaidya mass function, in terms of the prop-
1440
+ erty of the interior collapsing scalar/vector field.
1441
+
1442
+ 12
1443
+ We have fulfilled this demand by deriving the ex-
1444
+ pression of the generalized Vaidya mass in terms
1445
+ of the equation of state parameter of the interior
1446
+ collapsing field in Eq.(51).
1447
+ [1] J. R. Oppenheimer and H. Snyder, Phys. Rev. Journals
1448
+ Archive 56, 455 (1939).
1449
+ [2] S. Datt, Zs. f. Phys. 108 314 (1938).
1450
+ [3] G.F.R. Ellis, S.T.C. Siklos and J. Wainwrighit, in Dy-
1451
+ namical systems in cosmology, Eds. J. Wainwright and
1452
+ G.F.R. Ellis, (Cambridge University Press, Cambridge,
1453
+ England, 1997).
1454
+ [4] R. Penrose, Riv. Nuovo Cimento Soc. Ital. Fis. 1, 252
1455
+ (1969).
1456
+ [5] R. Geroch, Journal of Mathematical Physics, 11, 2, 437-
1457
+ 449 (1970).
1458
+ [6] S. W. Hawking and G. F. R. Ellis, The large scale struc-
1459
+ ture of spacetime, Cambridge University Press (1973).
1460
+ [7] P. S. Joshi, Global Aspects in Gravitation and Cosmology
1461
+ (Clendron Press, Oxford, 1993).
1462
+ [8] R. Geroch and G. Horowitz, ‘Global structure of space-
1463
+ times’, in General Relativity:
1464
+ An Einstein Centenary
1465
+ Survey, eds S. W. Hawking and W. Israel, Cambridge:
1466
+ Cambridge University Press (1979).
1467
+ [9] S. W. Hawking and W.Israel, ‘An introductory survey’, in
1468
+ General Relativity: An Einstein Centenary Survey, eds
1469
+ S. W. Hawking and W. Israel. Cambridge: Cambridge
1470
+ University Press (1979).
1471
+ [10] R. Penrose, ‘Singularities and time asymmetry’, in Gen-
1472
+ eral Relativity: An Einstein Centenary Survey, eds S. W.
1473
+ Hawking and W. Israel. Cambridge: Cambridge Univer-
1474
+ sity Press (1979).
1475
+ [11] R. Penrose, Phys. Rev. Lett. 14, 57 (1965).
1476
+ [12] P. S. Joshi and D. Malafarina, Phys. Rev. D 83, 024009
1477
+ (2011).
1478
+ [13] P. S. Joshi, Gravitational Collapse, and Spacetime Singu-
1479
+ larities, (Cambridge University Press, Cambridge, Eng-
1480
+ land, 2007).
1481
+ [14] K. Mosani, D. Dey, P. S. Joshi, Phys. Rev. D 102,
1482
+ 044037.
1483
+ [15] D. Christodoulou, Annals of Mathematics Annals of
1484
+ Mathematics, 140, 607 (1994).
1485
+ [16] D. Christodoulou, Annals of Mathematics, 149, 183
1486
+ (1999).
1487
+ [17] R. Goswami and P. S. Joshi, Modern Physics Letters A,
1488
+ 22, 01, pp. 65-74 (2007).
1489
+ [18] Karim Mosani, Dipanjan Dey, Kaushik Bhattacharya
1490
+ and Pankaj S. Joshi, Phys. Rev. D 105, 064048 (2022).
1491
+ [19] A. Wang and Y. Wu, Gen. Relativ. Gravit. 31, 107
1492
+ (1999).
1493
+ [20] F. J. Tipler, Phys. Lett. 64A, 8 (1977).
1494
+ [21] C. J. S. Clarke and A. Krolak, J. Geom. Phys. 2, 127
1495
+ (1985).
1496
+ [22] D. M. Eardley, in ’Gravitation in Astrophysics’, ed. B.
1497
+ Carter and J. B. Hartle (Plenum, New York, 1987).
1498
+ [23] Karim Mosani, Dipanjan Dey and Pankaj S. Joshi, Phys.
1499
+ Rev. D 101, 044052 (2020).
1500
+ [24] Demetrios Christodoulou, Commun. Math. Phys. 105,
1501
+ 337-361 (1986).
1502
+ [25] K. S. Virbhadra, S. Jhingan and P. S. Joshi, International
1503
+ Journal of Modern Physics D 06, 357-361 (1997).
1504
+ [26] David Garfinkle, Robert Mann, and Chris Vuille Phys.
1505
+ Rev. D 68, 064015 (2003).
1506
+ [27] E. Poisson, “A Relativist’s Toolkit: The Mathematics
1507
+ of Black-Hole Mechanics,” Cambridge University Press,
1508
+ (2009).
1509
+
69E4T4oBgHgl3EQfcgw0/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
7NA0T4oBgHgl3EQfOP_L/content/tmp_files/2301.02159v1.pdf.txt ADDED
@@ -0,0 +1,3207 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ arXiv:2301.02159v1 [math.NA] 5 Jan 2023
2
+ Finite element approximation of scalar curvature in arbitrary
3
+ dimension
4
+ Evan S. Gawlik∗
5
+ Michael Neunteufel†
6
+ Abstract
7
+ We analyze finite element discretizations of scalar curvature in dimension N ≥ 2.
8
+ Our
9
+ analysis focuses on piecewise polynomial interpolants of a smooth Riemannian metric g on a
10
+ simplicial triangulation of a polyhedral domain Ω ⊂ RN having maximum element diameter h.
11
+ We show that if such an interpolant gh has polynomial degree r ≥ 0 and possesses single-valued
12
+ tangential-tangential components on codimension-1 simplices, then it admits a natural notion
13
+ of (densitized) scalar curvature that converges in the H−2(Ω)-norm to the (densitized) scalar
14
+ curvature of g at a rate of O(hr+1) as h → 0, provided that either N = 2 or r ≥ 1. As a special
15
+ case, our result implies the convergence in H−2(Ω) of the widely used “angle defect” approxima-
16
+ tion of Gaussian curvature on two-dimensional triangulations, without stringent assumptions on
17
+ the interpolated metric gh. We present numerical experiments that indicate that our analytical
18
+ estimates are sharp.
19
+ 1
20
+ Introduction
21
+ Many partial differential equations that arise in mathematical physics and geometric analysis involve
22
+ the Riemann curvature tensor and its contractions.
23
+ The scalar curvature R, which is obtained
24
+ from two contractions of the Riemann curvature tensor, is particularly important; it serves as the
25
+ integrand in the Einstein-Hilbert functional from general relativity, and it appears in the governing
26
+ equation for two-dimensional Ricci flow. To approximate solutions to PDEs involving the scalar
27
+ curvature, it is necessary to discretize the nonlinear differential operator that sends a Riemannian
28
+ metric tensor to its scalar curvature.
29
+ The goal of this paper is to construct and analyze such
30
+ discretizations in arbitrary dimension N ≥ 2.
31
+ We are specifically interested in the setting where a smooth Riemannian metric tensor g on
32
+ a polyhedral domain Ω ⊂ RN is approximated by a piecewise polynomial Regge metric gh on a
33
+ simplicial triangulation T of Ω having maximum element diameter h. Here, a metric is called a
34
+ Regge metric on T if it is piecewise smooth and its tangential-tangential components are single-
35
+ valued on every codimension-1 simplex in T . When such a metric is piecewise polynomial, it belongs
36
+ to a finite element space called the Regge finite element space [11, 12, 21]. Regge metrics are not
37
+ classically differentiable, so our first task will be to assign meaning to the scalar curvature of gh. Our
38
+ definition, which is a natural generalization of one that is now well-established in dimension N = 2,
39
+ treats the scalar curvature of gh as a distribution and regards it as an approximation of the densitized
40
+ scalar curvature of g, i.e. the scalar curvature R times the volume form ω. For piecewise constant
41
+ Regge metrics, our definition reduces to the classical definition of the distributional densitized
42
+ ∗Department of Mathematics, University of Hawai‘i at Manoa, Honolulu, HI, 96822, USA, egawlik@hawaii.edu
43
+ †Institute for Analysis and Scientific Computing, TU Wien, Wiedner Hauptstr. 8-10, 1040 Wien, Austria,
44
+ michael.neunteufel@tuwien.ac.at
45
+ 1
46
+
47
+ curvature on piecewise flat spaces [8, 23]. It is a linear combination of Dirac delta distributions
48
+ supported on (N − 2)-simplices S, weighted by the angle defect at S: 2π minus the sum of the
49
+ dihedral angles incident at S. For piecewise polynomial Regge metrics of higher degree, it includes
50
+ additional contributions involving the scalar curvature in the interior of each N-simplex and the
51
+ jump in the mean curvature across each (N − 1)-simplex.
52
+ We study the convergence of the distributional densitized scalar curvature of gh to the densitized
53
+ scalar curvature of g under refinement of the triangulation. We show in Theorem 4.1 that in the
54
+ H−2(Ω)-norm, this convergence takes place at a rate of O(hr+1) when gh is an optimal-order
55
+ interpolant of g that is piecewise polynomial of degree r ≥ 0, provided that either N = 2 or r ≥ 1.
56
+ Our numerical experiments in Section 5 suggest that these estimates are sharp in general.
57
+ To put this convergence result into context, let us summarize some existing convergence results
58
+ in the literature on finite element approximation of the scalar curvature. We first need to assemble
59
+ some notation.
60
+ Notation.
61
+ In what follows, W s,p(Ω) denotes the Sobolev-Slobodeckij space of differentiability
62
+ index s ∈ [0, ∞) and integrability index p ∈ [1, ∞], and ∥ · ∥W s,p(Ω) and | · |W s,p(Ω) denote the
63
+ associated norm and semi-norm, which we always take with respect to the Euclidean metric. We
64
+ denote Lp(Ω) = W 0,p(Ω) and Hs(Ω) = W s,2(Ω). For k ∈ N, we denote H−k(Ω) = (Hk
65
+ 0 (Ω))′, where
66
+ Hk
67
+ 0 (Ω) denotes the space of functions in Hk(Ω) whose derivatives of order 0 through k − 1 have
68
+ vanishing trace on ∂Ω, and the prime denotes the dual space. Occasionally we use weighted Lp and
69
+ H−k spaces associated with a Riemannian metric g, which we denote by Lp(Ω, g) and H−k(Ω, g);
70
+ see Section 4 and [16, Equation 4.1] for details.
71
+ If g is a smooth Riemannian metric and gh is a Regge metric, then R(g) denotes the scalar
72
+ curvature of g, (Rω)(g) denotes the densitized scalar curvature of g, (Rω)dist(gh) denotes the
73
+ distributional densitized scalar curvature of gh (defined below in Definition 3.1), and R(q)
74
+ h (gh)
75
+ denotes the L2(Ω, gh)-projection of (Rω)dist(gh) onto the Lagrange finite element space of degree q.
76
+ We also use the terms optimal-order interpolant, canonical interpolant, and geodesic interpolant
77
+ below. The first of these is a catch-all term for any piecewise polynomial interpolant gh of g that
78
+ belongs to the Regge finite element space and enjoys error estimates of optimal order in W s,p(T)-
79
+ norms on N-simplices T; see Definition 4.2. The canonical interpolant is a specific interpolant
80
+ (which is optimal-order) detailed in [21, Chapter 2]. The geodesic interpolant of g is the unique
81
+ piecewise constant Regge metric gh with the property that the length of every edge in T , as
82
+ measured by gh, agrees with the geodesic distance between the corresponding vertices in T , as
83
+ measured by g.
84
+ Summary of existing results.
85
+ We can now summarize some existing results about the approx-
86
+ imation of g’s curvature by gh’s distributional curvature. Throughout what follows, the letter r
87
+ denotes the polynomial degree of gh.
88
+ 1. Cheeger, M¨uller, and Schrader [8, Equation (5.7) and Theorem 5.1] proved that if r = 0 and
89
+ gh is the geodesic interpolant of g, then (Rω)dist(gh) converges to (Rω)(g) in the (setwise)
90
+ sense of measures at a rate of O(h) in dimension N = 2 and O(h1/2) in dimension N ≥ 3.
91
+ 2. Gawlik [16, Theorem 4.1] proved that if r ≥ 1, N = 2, and gh is any optimal-order interpolant
92
+ of g, then R(q)
93
+ h (gh) converges to R(g) at a rate of O(hr) in the H−1(Ω, g)-norm and at a rate of
94
+ O(hr−k−1) in the broken Hk(Ω)-norm, k = 0, 1, 2, . . . , r − 2, provided that q ≥ max{1, r − 2}.
95
+ 3. Berchenko-Kogan and Gawlik [4, Corollary 6.2] proved that if r ≥ 1, N = 2, and gh is any
96
+ optimal-order interpolant of g, then (Rω)dist(gh) converges to (Rω)(g) at a rate of O(hr) in
97
+ 2
98
+
99
+ the norm ∥u∥V ′,h = supv∈V,v̸=0⟨u, v⟩V ′,V /∥v∥V,h, where
100
+ V = {v ∈ H1
101
+ 0(Ω) | v|T ∈ H2(T) ∀T ∈ T N}
102
+ (1)
103
+ and ∥v∥V,h = |v|H1(Ω) +
104
+ ��
105
+ T∈T N h2
106
+ T |v|2
107
+ H2(T)
108
+ �1/2
109
+ . Here, hT denotes the diameter of T, and
110
+ T N denotes the set of N-simplices in T .
111
+ 4. Gopalakrishnan, Neunteufel, Sch¨oberl, and Wardetzky [19, Theorem 6.5 and Corollary 6.6]
112
+ proved that if r ≥ 0, N = 2, and gh is the canonical interpolant of g, then R(r+1)
113
+ h
114
+ (gh) converges
115
+ to R(g) at a rate of O(hr+1) in the H−1(Ω, g)-norm and at a rate of O(hr−k) in the broken
116
+ Hk(Ω)-norm, k = 0, 1, 2, . . . , r − 1.
117
+ New results.
118
+ As one can see from above, our analysis in this paper covers two important cases
119
+ that have not yet been addressed in the literature:
120
+ 1. We prove a convergence result in the case where N ≥ 3 and r ≥ 1. This opens the door
121
+ to the use of piecewise polynomial Regge metrics to approximate scalar curvature in high
122
+ dimensions.
123
+ 2. We prove a convergence result in the case where N = 2, r = 0, and gh is an arbitrary
124
+ optimal-order interpolant of g. This has been a longstanding gap in the literature on Gaussian
125
+ curvature approximation. Previous efforts to address the case where N = 2 and r = 0 have
126
+ relied on subtle properties of the geodesic interpolant [8] and the canonical interpolant [19].
127
+ Our results establish the validity of Gaussian curvature approximations involving the angle
128
+ defect without stringent assumptions on the interpolated metric tensor gh.
129
+ Note that our analysis predicts no convergence at all in the H−2(Ω)-norm when N ≥ 3 and r = 0.
130
+ Our numerical experiments suggest that this result is sharp for general optimal-order interpolants.
131
+ However, for the canonical interpolant, numerical experiments suggest that (Rω)dist(gh) converges
132
+ to (Rω)(g) in the H−2(Ω)-norm at a rate of O(h) when N ≥ 3 and r = 0. We intend to study this
133
+ superconvergence phenomenon exhibited by the canonical interpolant in future work.
134
+ Structure of the paper.
135
+ Our strategy for proving convergence of (Rω)dist(gh) to (Rω)(g) con-
136
+ sists of two steps. First, in Sections 2-3, we study the evolution of (Rω)dist(gh) under deformations
137
+ of the metric, leading to an integral formula for the error (Rω)dist(gh) − (Rω)(g) which reads
138
+ ⟨(Rω)dist(gh) − (Rω)(g), v⟩V ′,V =
139
+ � 1
140
+ 0
141
+ bh(�g(t); σ, v) − ah(�g(t); σ, v) dt,
142
+ ∀v ∈ V.
143
+ (2)
144
+ Here, �g(t) = (1 − t)g + tgh, σ =
145
+
146
+ ∂t�g(t) = gh − g, V is the space defined in (1), and bh(�g(t); ·, ·)
147
+ and ah(�g(t); ·, ·) are certain metric-dependent bilinear forms. In Section 4, we use techniques from
148
+ finite element theory to estimate the right-hand side of (2), leading to Theorem 4.1.
149
+ The approach above is similar to the one used in dimension N = 2 in [4, 16, 19], but there are
150
+ a few important differences. First, we work with an integral formula for the error (Rω)dist(gh) −
151
+ (Rω)(g) rather than an integral formula for the curvature itself. Previous analyses in [4, 16, 19]
152
+ hinged on formulas of the latter type. Loosely speaking, in this paper we compute the evolution of
153
+ the error along a one-parameter family of Regge metrics starting at g and ending at gh, whereas
154
+ the papers [4, 16, 19] compute the evolution of the curvature along a pair of one-parameter families
155
+ of metrics: one family that starts at the Euclidean metric δ and ends at gh, and one that starts at
156
+ 3
157
+
158
+ δ and ends at g. The approach based on evolving the error appears to be better suited for proving
159
+ optimal error estimates.
160
+ Another key aspect of our analysis is our use of the H−2(Ω)-norm to measure the error. This
161
+ norm is weaker than the ones used in [4, 16, 19], and it appears to be more natural for measuring
162
+ the error in the curvature. For example, for piecewise constant Regge metrics in dimension N = 2,
163
+ we show that convergence of (Rω)dist(gh) to (Rω)(g) holds in the H−2(Ω)-norm for any optimal-
164
+ order interpolant of g, but numerical experiments suggest that it fails to hold in stronger norms
165
+ when gh is not the canonical interpolant of g. A key tool that we use to prove convergence in
166
+ H−2(Ω) is the near-equivalence of a certain pair of metric-dependent, mesh-dependent norms on
167
+ V ; see Proposition 4.5. This equivalence is similar to one that Walker [27, Theorems 4.1 and 4.3]
168
+ proved for an analogous family of mesh-dependent norms on triangulated surfaces.
169
+ Additional comments.
170
+ The formula (2) is not only useful for the error analysis, but it is also
171
+ interesting in its own right. It has a differential counterpart (see Theorem 3.6) that reads
172
+ d
173
+ dt⟨(Rω)dist(�g(t)), v⟩V ′,V = bh(�g(t); σ, v) − ah(�g(t); σ, v),
174
+ ∀v ∈ V,
175
+ (3)
176
+ which mimics the formula
177
+ d
178
+ dt
179
+
180
+
181
+ Rvω =
182
+
183
+
184
+ (div div Sσ)vω −
185
+
186
+
187
+ ⟨G, σ⟩vω,
188
+ ∀v ∈ V
189
+ (4)
190
+ that holds for a family of smooth Riemannian metrics g(t) with densitized scalar curvature Rω and
191
+ Einstein tensor G = Ric − 1
192
+ 2Rg. Here, Sσ = σ−g Tr σ, and div is the covariant divergence operator;
193
+ see below for more notational details.
194
+ The correspondence between (3) and (4) becomes even more transparent when one inspects the
195
+ formulas for bh and ah (see Theorem 3.6). The bilinear form bh(�g; ·, ·) is (up to the appearance of
196
+ S) a non-Euclidean, N-dimensional generalization of a bilinear form that appears in the Hellan-
197
+ Herrmann-Johnson finite element method [1–3, 5–7, 9, 22].
198
+ It can be regarded as the integral
199
+ of div div Sσ against v, where div div is interpreted in a distributional sense. This link with the
200
+ Hellan-Herrmann-Johnson method has previously been noted and used in dimension N = 2 [4, 16,
201
+ 19].
202
+ The bilinear form ah(�g; ·, ·), which is only nonzero in dimension N ≥ 3, appears to play the role
203
+ of
204
+
205
+ Ω⟨G, σ⟩vω, which is also only nonzero in dimension N ≥ 3. It gives rise to a natural way of
206
+ defining the Einstein tensor in a distributional sense for Regge metrics. We discuss this more in
207
+ Section 3.2. Among other things, we point out that the formula for ah contains a term involving
208
+ the jump in the trace-reversed second fundamental form across codimension-1 simplices; the same
209
+ quantity arises in studies of singular sources in general relativity, where it encodes the well-known
210
+ Israel junction conditions across a hypersurface on which stress-energy is concentrated [20].
211
+ There are a few other connections between our calculations and ones that appear in the physics
212
+ literature. The variation of the Gibbons-Hawking-York boundary term in general relativity [17, 28]
213
+ is one example. It has many parallels to our calculations in Section 2.2, and one can undoubtedly
214
+ find formulas like (6) in the literature after reconciling notations. We still give a full derivation
215
+ of such formulas, not only to familiarize the reader with our notation, but also to provide careful
216
+ derivations that refrain from discarding total derivatives (which integrate to zero on manifolds
217
+ without boundary, but not in general) and minimize the use of local coordinate calculations where
218
+ possible.
219
+ 4
220
+
221
+ 2
222
+ Evolution of geometric quantities
223
+ In this section, we consider an N-dimensional manifold M equipped with a smooth Riemannian
224
+ metric g, and we study the evolution of various geometric quantities under deformations of g.
225
+ We adopt the following notation in this section. The Levi-Civita connection associated with g
226
+ is denoted ∇. If σ is a (p, q)-tensor field, then its covariant derivative is the (p, q + 1)-tensor field
227
+ ∇σ, and its covariant derivative in the direction of a vector field X is the (p, q)-tensor field ∇Xσ.
228
+ Its trace Tr σ is the contraction of σ along the first two indices, using g to raise or lower indices
229
+ as needed. We denote div σ = Tr ∇σ and ∆σ = div ∇σ. The g-inner product of two (p, q)-tensor
230
+ fields σ and ρ is denoted ⟨σ, ρ⟩.
231
+ The volume form associated with g is denoted ω. The Ricci tensor and the scalar curvature of g
232
+ are denoted Ric and R, respectively. When we wish to emphasize their dependence on g, we write
233
+ ω(g), Ric(g), R(g), etc.
234
+ If D is an embedded submanifold of M, then we denote by ωD the induced volume form on D.
235
+ If σ is a tensor field on M, then σ|D denotes the pullback of σ under the inclusion D ֒→ M. Later
236
+ we will introduce some additional notation related to embedded submanifolds of codimension 1,
237
+ like the mean curvature H and second fundamental form II; see Section 2.2.
238
+ We denote the exterior derivative of a differential form α by dα. If α is a one-form, then α♯
239
+ denotes the vector field obtained by raising indices with g. If f is a scalar field, then we sometimes
240
+ interpret the one-form ∇f = df as the vector field (df)♯ without explicitly writing it.
241
+ Later, in Section 4, we will append a subscript g to many quantities like ∇ and ⟨·, ·⟩ to emphasize
242
+ their dependence on g.
243
+ In that section only, an absent subscript will generally signal that the
244
+ quantity in question is computed with respect to the Euclidean metric, which we denote by δ. We
245
+ say more about this notational shift in Section 4.
246
+ 2.1
247
+ Evolution of the densitized scalar curvature
248
+ First we study the evolution of the densitized scalar curvature Rω under deformations of the metric.
249
+ Proposition 2.1. Let g(t) be a family of smooth Riemannian metrics with time derivative ∂
250
+ ∂tg =: σ.
251
+ We have
252
+
253
+ ∂t(Rω) = (div div Sσ)ω − ⟨G, σ⟩ω,
254
+ where G = Ric − 1
255
+ 2Rg denotes the Einstein tensor associated with g and
256
+ Sσ = σ − g Tr σ.
257
+ Proof. We compute
258
+
259
+ ∂t(Rω) = ˙Rω + R ˙ω
260
+ and invoke the well-known formulas [15, Lemma 2]
261
+ ˙R = div div σ − ∆ Tr σ − ⟨Ric, σ⟩
262
+ and [10, Equation 2.4]
263
+ ˙ω = 1
264
+ 2(Tr σ)ω.
265
+ Since ∆ Tr σ = div div(g Tr σ) and Tr σ = ⟨g, σ⟩, the result follows.
266
+ 5
267
+
268
+ 2.2
269
+ Evolution of the mean curvature
270
+ Next we study the evolution of the mean curvature H of a hypersurface F. We assume that the tan-
271
+ gent bundle of F is trivial, so that there exists a smooth, g-orthonormal frame field τ1, τ2, . . . , τN−1
272
+ on F. (If this is not the case, then one can simply fix a point p ∈ F and focus on a neighborhood of p
273
+ on which the tangent bundle is trivial.) We let n be the unit normal to F so that n, τ1, τ2, . . . , τN−1
274
+ forms a right-handed g-orthonormal frame (in the ambient manifold) at each point on F. If the
275
+ metric g varies smoothly in time, then we assume that the vectors n, τ1, τ2, . . . , τN−1 also vary
276
+ smoothly in time and remain g-orthonormal at all times.
277
+ We use the notation
278
+ II(X, Y ) = g(∇Xn, Y ) = −g(n, ∇XY )
279
+ for the second fundamental form on F. Our sign convention is such that Tr II = H, and H is
280
+ positive for a sphere with an outward normal vector. We also let ∇F and divF denote the surface
281
+ gradient and surface divergence operators on F, which have the following meanings. For a scalar
282
+ field v,
283
+ ∇F v = ∇v − n∇nv =
284
+ N−1
285
+
286
+ i=1
287
+ τi∇τiv,
288
+ and for a one-form α,
289
+ divF α = Tr (∇α|F) =
290
+ N−1
291
+
292
+ i=1
293
+ (∇τiα)(τi).
294
+ Note that in the formula ∇F v = ∇v − n∇nv, we have regarded ∇v as a vector field rather than a
295
+ one-form. Recall that the surface divergence operator satisfies the identity
296
+
297
+ F
298
+ (divF α)ωF =
299
+
300
+ ∂F
301
+ α(νF )ω∂F +
302
+
303
+ F
304
+ Hα(n)ωF ,
305
+ (5)
306
+ where νF is the outward unit normal to ∂F and H is the mean curvature of F.
307
+ Proposition 2.2. Let g(t) be a family of smooth Riemannian metrics with time derivative ∂
308
+ ∂tg =: σ.
309
+ Let F be a time-independent hypersurface with mean curvature H and induced volume form ωF.
310
+ Then
311
+
312
+ ∂t(HωF ) = −1
313
+ 2
314
+ ��
315
+ II, σ|F
316
+
317
+ + (div Sσ)(n) + divF (σ(n, ·)) − Hσ(n, n)
318
+
319
+ ωF,
320
+ (6)
321
+ where
322
+ II(X, Y ) = II(X, Y ) − Hg(X, Y )
323
+ is the trace-reversed second fundamental form.
324
+ Remark 2.3. In dimension N = 2, the formula (6) simplifies considerably.
325
+ Letting τ and n
326
+ denote the unit tangent and unit normal to F, we have ∇τn = Hτ, −∇ττ = Hn, and II(τ, τ) =
327
+ g(∇τn, τ) − Hg(τ, τ) = H − H = 0, so II vanishes. In addition,
328
+ divF (σ(n, ·)) − Hσ(n, n) = ∇τ (σ(n, ·)) (τ) − Hσ(n, n)
329
+ = ∇τ (σ(n, τ)) − σ(n, ∇ττ) − Hσ(n, n)
330
+ = ∇τ (σ(n, τ)) .
331
+ Thus, in two dimensions,
332
+
333
+ ∂t(HωF) = −1
334
+ 2 ((div Sσ)(n) + ∇τ (σ(n, τ))) ωF.
335
+ 6
336
+
337
+ To prove Proposition 2.2, we write
338
+ ˙H = −
339
+ N−1
340
+
341
+ i=1
342
+
343
+ ∂tg(n, ∇τiτi)
344
+ (7)
345
+ and use the following lemmas.
346
+ Lemma 2.4. For any time-dependent vector fields X and Y ,
347
+
348
+ ∂t∇Y X = ∇ ˙Y X + ∇Y ˙X + 1
349
+ 2 ((∇Xσ)Y + (∇Y σ)X − (∇σ)(X, Y ))♯ ,
350
+ where (∇σ)(X, Y ) denotes the one-form Z �→ (∇Zσ)(X, Y ), and (∇Xσ)Y denotes the one-form
351
+ Z �→ (∇Xσ)(Y, Z).
352
+ Proof. In coordinates,
353
+ (∇Y X)ℓ = Y j ∂Xℓ
354
+ ∂xj + Γℓ
355
+ ijY jXi,
356
+ where Γℓ
357
+ ij denote the Christoffel symbols of the second kind associated with g. Thus,
358
+
359
+ ∂t(∇Y X)ℓ = ˙Y j ∂Xℓ
360
+ ∂xj + Γℓ
361
+ ij ˙Y jXi + Y j ∂ ˙Xℓ
362
+ ∂xj + Γℓ
363
+ ijY j ˙Xi + ˙Γℓ
364
+ ijY jXi
365
+ = (∇ ˙Y X)ℓ + (∇Y ˙X)ℓ + ˙Γℓ
366
+ ijY jXi.
367
+ Next, we recall the following formula for the rate of change of the Christoffel symbols under a
368
+ metric deformation [10, Equation 2.23]:
369
+ ˙Γℓ
370
+ ij = 1
371
+ 2gℓm ((∇iσ)jm + (∇jσ)im − (∇mσ)ij) .
372
+ It follows that
373
+ ˙Γℓ
374
+ ijY jXi = 1
375
+ 2gℓm �
376
+ (∇Xσ)jmY j + (∇Y σ)imXi − (∇mσ)ijY jXi�
377
+ = 1
378
+ 2 [((∇Xσ)Y + (∇Y σ)X − (∇σ)(X, Y ))]ℓ .
379
+ Hence,
380
+
381
+ ∂t(∇Y X)ℓ = (∇ ˙Y X)ℓ + (∇Y ˙X)ℓ + 1
382
+ 2 ((∇Xσ)Y + (∇Y σ)X − (∇σ)(X, Y ))ℓ .
383
+ Lemma 2.5. For any time-dependent vector field X,
384
+
385
+ ∂tg(n, X) = 1
386
+ 2σ(n, n)g(n, X) + g(n, ˙X).
387
+ Proof. Writing X = ng(n, X) + �N−1
388
+ i=1 τig(τi, X), we compute
389
+
390
+ ∂tg(n, X) = σ(n, X) + g( ˙n, X) + g(n, ˙X)
391
+ = σ(n, n)g(n, X) +
392
+ N−1
393
+
394
+ i=1
395
+ σ(n, τi)g(τi, X) + g( ˙n, n)g(n, X) +
396
+ N−1
397
+
398
+ i=1
399
+ g( ˙n, τi)g(τi, X) + g(n, ˙X)
400
+ = (σ(n, n) + g( ˙n, n)) g(n, X) +
401
+ N−1
402
+
403
+ i=1
404
+ (σ(n, τi) + g( ˙n, τi)) g(τi, X) + g(n, ˙X).
405
+ 7
406
+
407
+ For each i = 1, 2, . . . , N − 1, we have
408
+ 0 = ∂
409
+ ∂tg(n, τi) = σ(n, τi) + g( ˙n, τi) + g(n, ˙τi)
410
+ = σ(n, τi) + g( ˙n, τi)
411
+ since ˙τi is g-orthogonal to n. Likewise,
412
+ 0 = ∂
413
+ ∂tg(n, n) = σ(n, n) + 2g(n, ˙n),
414
+ so the result follows.
415
+ We are now ready to compute the time derivative of the mean curvature H. By Lemma 2.5, we
416
+ have
417
+ ˙H = −
418
+ N−1
419
+
420
+ i=1
421
+
422
+ ∂tg(n, ∇τiτi)
423
+ = −
424
+ N−1
425
+
426
+ i=1
427
+ �1
428
+ 2σ(n, n)g(n, ∇τiτi) + g
429
+
430
+ n, ∂
431
+ ∂t∇τiτi
432
+ ��
433
+ = 1
434
+ 2Hσ(n, n) −
435
+ N−1
436
+
437
+ i=1
438
+ g
439
+
440
+ n, ∂
441
+ ∂t∇τiτi
442
+
443
+ .
444
+ (8)
445
+ Using Lemma 2.4 and the symmetry of the second fundamental form, we can write the second term
446
+ as
447
+ g
448
+
449
+ n, ∂
450
+ ∂t∇τiτi
451
+
452
+ = g(n, ∇ ˙τiτi) + g(n, ∇τi ˙τi) + (∇τiσ)(n, τi) − 1
453
+ 2(∇nσ)(τi, τi)
454
+ = 2g(n, ∇ ˙τiτi) + (∇τiσ)(n, τi) − 1
455
+ 2(∇nσ)(τi, τi).
456
+ The first term above, when summed over i, can be simplified as follows. We write ˙τi = �N−1
457
+ j=1 τjg(τj, ˙τi)
458
+ and use the linearity of ∇XY in X to compute
459
+ 2
460
+ N−1
461
+
462
+ i=1
463
+ g(n, ∇ ˙τiτi) = 2
464
+ N−1
465
+
466
+ i=1
467
+ N−1
468
+
469
+ j=1
470
+ g(n, ∇τjτi)g(τj, ˙τi)
471
+ =
472
+ N−1
473
+
474
+ i=1
475
+ N−1
476
+
477
+ j=1
478
+ g(n, ∇τjτi) (g(τj, ˙τi) + g( ˙τj, τi))
479
+ = −
480
+ N−1
481
+
482
+ i=1
483
+ N−1
484
+
485
+ j=1
486
+ g(n, ∇τjτi)σ(τj, τi)
487
+ = ⟨II, σ|F ⟩.
488
+ Above, we used the symmetry of the second fundamental form to pass from the first line to the
489
+ second, and we used the identity
490
+ 0 = ∂
491
+ ∂tg(τj, τi) = σ(τj, τi) + g(τj, ˙τi) + g( ˙τj, τi)
492
+ 8
493
+
494
+ to pass from the second line to the third. Inserting these results into (8), we get
495
+ ˙H = 1
496
+ 2Hσ(n, n) − ⟨II, σ|F ⟩ +
497
+ N−1
498
+
499
+ i=1
500
+ �1
501
+ 2(∇nσ)(τi, τi) − (∇τiσ)(n, τi)
502
+
503
+ .
504
+ (9)
505
+ Lemma 2.6. We have
506
+ N−1
507
+
508
+ i=1
509
+ �1
510
+ 2(∇nσ)(τi, τi) − (∇τiσ)(n, τi)
511
+
512
+ = 1
513
+ 2 (⟨II, σ|F⟩ − (div Sσ)(n) − divF (σ(n, ·))) .
514
+ (10)
515
+ Proof. The identity 0 = ∇τi (g(n, n)) = 2g(n, ∇τin) shows that ∇τin is in the span of {τj}N−1
516
+ j=1 , so
517
+ the first term on the right-hand side of (10) satisfies
518
+ ⟨II, σ|F ⟩ =
519
+ N−1
520
+
521
+ i=1
522
+ N−1
523
+
524
+ j=1
525
+ σ(τj, τi)g(τj, ∇τin)
526
+ =
527
+ N−1
528
+
529
+ i=1
530
+ σ(∇τin, τi).
531
+ (11)
532
+ The second term on the right-hand side of (10) can be computed as follows. Recalling that Sσ =
533
+ σ − g Tr σ, we have
534
+ (div Sσ)(n) = ∇n(Sσ)(n, n) +
535
+ N−1
536
+
537
+ i=1
538
+ ∇τi(Sσ)(n, τi)
539
+ = (∇nσ)(n, n) − ∇n(g Tr σ)(n, n) +
540
+ N−1
541
+
542
+ i=1
543
+ [(∇τiσ)(n, τi) − ∇τi(g Tr σ)(n, τi)]
544
+ = (∇nσ)(n, n) − g(n, n)∇n Tr σ +
545
+ N−1
546
+
547
+ i=1
548
+ [(∇τiσ)(n, τi) − g(n, τi)∇τi Tr σ]
549
+ = (∇nσ)(n, n) − ∇n Tr σ +
550
+ N−1
551
+
552
+ i=1
553
+ (∇τiσ)(n, τi).
554
+ Since the trace commutes with covariant differentiation,
555
+ ∇n Tr σ = Tr ∇nσ = (∇nσ)(n, n) +
556
+ N−1
557
+
558
+ i=1
559
+ (∇nσ)(τi, τi).
560
+ Thus,
561
+ (div Sσ)(n) =
562
+ N−1
563
+
564
+ i=1
565
+ [(∇τiσ)(n, τi) − (∇nσ)(τi, τi)] .
566
+ (12)
567
+ The third term on the right-hand side of (10) is given by
568
+ divF (σ(n, ·)) =
569
+ N−1
570
+
571
+ i=1
572
+ ∇τi (σ(n, ·)) (τi)
573
+ =
574
+ N−1
575
+
576
+ i=1
577
+ [∇τi (σ(n, τi)) − σ(n, ∇τiτi)] .
578
+ (13)
579
+ 9
580
+
581
+ Combining (11), (12), and (13), we see that
582
+ 1
583
+ 2 (⟨II, σ|F ⟩ − (div Sσ)(n) − divF (σ(n, ·)))
584
+ = 1
585
+ 2
586
+ N−1
587
+
588
+ i=1
589
+ [σ(∇τin, τi) − (∇τiσ)(n, τi) + (∇nσ)(τi, τi) − ∇τi (σ(n, τi)) + σ(n, ∇τiτi)]
590
+ = 1
591
+ 2
592
+ N−1
593
+
594
+ i=1
595
+ [(∇nσ)(τi, τi) − 2(∇τiσ)(n, τi)] .
596
+ Combining Lemma 2.6 with (9), we get
597
+ ˙H = 1
598
+ 2 (−⟨II, σ|F ⟩ − (div Sσ)(n) − divF (σ(n, ·)) + Hσ(n, n)) .
599
+ (14)
600
+ Proposition 2.2 now follows from the identities
601
+
602
+ ∂t(HωF) = ˙HωF + H ˙ωF = ˙HωF + 1
603
+ 2H Tr (σ|F ) ωF
604
+ and
605
+ ⟨II, σ|F ⟩ − H Tr (σ|F) = ⟨II, σ|F⟩.
606
+ 2.3
607
+ Evolution of angles
608
+ Next we study the evolution of angles under deformations of the metric.
609
+ Lemma 2.7. Let g(t) be a family of smooth Riemannian metrics with time derivative
610
+
611
+ ∂tg =: σ.
612
+ Let (¯n(t), ¯τ(t)) be a pair of g(t)-orthonormal vectors, and let (n(t), τ(t)) be another pair of g(t)-
613
+ orthonormal vectors lying in the span of (¯n(t), ¯τ(t)). Let θ(t) be the angle for which
614
+ τ = ¯τ cos θ + ¯n sin θ,
615
+ n = −¯τ sin θ + ¯n cos θ.
616
+ Assume that these vectors vary smoothly in time, and assume that n(t) (respectively, ¯n(t)) is at all
617
+ times g(t)-orthogonal to a time-independent hypersurface F (respectively, ¯F). Then, at all times
618
+ for which θ ∈ (0, π), we have
619
+
620
+ ∂tθ = 1
621
+ 2σ(n, τ) − 1
622
+ 2σ(¯n, ¯τ).
623
+ (15)
624
+ Proof. Differentiating the relation cos θ = g(¯n, n) yields
625
+ − ˙θ sin θ = ∂
626
+ ∂t (g(¯n, n)) .
627
+ In particular, at any time s, we can write
628
+ − ˙θ(s) sin θ(s) = ∂
629
+ ∂t
630
+ ����
631
+ t=s
632
+ (g(t)(¯n(t), n(s))) + ∂
633
+ ∂t
634
+ ����
635
+ t=s
636
+ (g(t)(¯n(s), n(t))) − σ(s)(¯n(s), n(s)).
637
+ 10
638
+
639
+ Using Lemma 2.5 and suppressing the evaluations at t = s, we get
640
+ − ˙θ sin θ = 1
641
+ 2σ(¯n, ¯n)g(¯n, n) + 1
642
+ 2σ(n, n)g(n, ¯n) − σ(¯n, n)
643
+ = 1
644
+ 2σ(¯n, ¯n cos θ − n) + 1
645
+ 2σ(n cos θ − ¯n, n)
646
+ = 1
647
+ 2σ(¯n, ¯τ sin θ) + 1
648
+ 2σ(−τ sin θ, n).
649
+ If θ ∈ (0, π) at time t = s, then we can divide by sin θ to get (15).
650
+ 3
651
+ Distributional densitized scalar curvature
652
+ Let T be a simplicial triangulation of a polyhedral domain Ω ⊂ RN. We use T k to denote the
653
+ set of all k-simplices in T . We also use ˚T k to denote the subset of T k consisting of k-simplices
654
+ that are not contained in the boundary of Ω. We call such simplices interior simplices. We call
655
+ (N − 1)-simplices faces.
656
+ Let g be a Regge metric on T . Recall that this means that g|T is a smooth Riemannian metric
657
+ on each T ∈ T N, and the induced metric g|F is single-valued on each F ∈ ˚
658
+ T N−1 (and consequently
659
+ the induced metric is single-valued on all lower-dimensional simplices in T ).
660
+ On each T ∈ T N, we denote by RT the scalar curvature of g|T . On an interior face F ∈ ˚
661
+ T N−1
662
+ that lies on the boundary of two N-simplices T + and T −, the second fundamental form on F, as
663
+ measured by g|T +, generally differs from that measured by g|T −. We denote by �II�F the jump in
664
+ the second fundamental form across F. More precisely,
665
+ �II�F(X, Y ) = g|T + (∇Xn+, Y ) + g|T − (∇Xn−, Y )
666
+ for any vectors X, Y tangent to F, where n± points outward from T ±, has unit length with respect
667
+ to g|T ±, and is g|T ±-orthogonal to F. We adopt similar notation for the jumps in other quantities
668
+ across F. For instance, �H�F denotes the jump in the mean curvature across F. We sometimes
669
+ drop the subscript F when there is no danger of confusion. If F is contained in ∂Ω, then we define
670
+ the jump in a scalar field v across F to be simply �v�F = v|F .
671
+ On each S ∈ ˚
672
+ T N−2, the angle defect along S is
673
+ ΘS = 2π −
674
+
675
+ T∈T N
676
+ T⊃S
677
+ θST,
678
+ where θST denotes the dihedral angle formed by the two faces of T that contain S, as measured by
679
+ g|T . Generally this angle may vary along S. If F + and F − are the two faces of T that contain S,
680
+ and if n± denotes the unit normal to F ± with respect to g|T pointing outward from T, then
681
+ cos θST = − g|T (n+, n−).
682
+ Let
683
+ V = {v ∈ H1
684
+ 0(Ω) | ∀T ∈ T N, v|T ∈ H2(T)}.
685
+ Note that if v ∈ V , then v admits a single-valued trace on every simplex in T of dimension ≥ N −3.
686
+ Definition 3.1. Let g be a Regge metric. The distributional densitized scalar curvature of g is the
687
+ linear functional (Rω)dist(g) ∈ V ′ defined by
688
+ ⟨(Rω)dist(g), v⟩V ′,V =
689
+
690
+ T∈T N
691
+
692
+ T
693
+ RT vωT + 2
694
+
695
+ F ∈˚
696
+ T N−1
697
+
698
+ F
699
+ �H�FvωF + 2
700
+
701
+ S∈˚
702
+ T N−2
703
+
704
+ S
705
+ ΘSvωS,
706
+ ∀v ∈ V.
707
+ (16)
708
+ 11
709
+
710
+ This definition generalizes Definition 3.1 of [4], where the distributional curvature two-form (i.e.
711
+ the Gaussian curvature times the volume form) is defined for Regge metrics in dimension N = 2.
712
+ Note that the factors of 2 appearing in all but the first term in (16) are consistent with the fact
713
+ that in dimension N = 2, the scalar curvature R is twice the Gaussian curvature.
714
+ One can heuristically motivate Definition 3.1 in much the same way that one motivates its
715
+ two-dimensional counterpart. When g is piecewise constant, Definition 3.1 recovers the classical
716
+ notion [23] that the distributional densitized scalar curvature is a linear combination of Dirac delta
717
+ distributions supported on (N − 2)-simplices, with weights given by angle defects. When g is not
718
+ piecewise constant, additional terms appear which account for the nonzero (classically defined)
719
+ curvature of g in the interior of each N-simplex T and the jump in the mean curvature across each
720
+ interior face F. The jump in the mean curvature across F can be understood by recalling that the
721
+ scalar curvature R at a point p ∈ F can be expressed as (two times) a sum of sectional curvatures
722
+ of N(N − 1)/2 tangent planes that are mutually g-orthogonal at p, (N − 1)(N − 2)/2 of which
723
+ are tangent to F at p and N − 1 of which are g-orthogonal to F at p. The sectional curvatures
724
+ corresponding to planes tangent to F are nonsingular, owing to the tangential-tangential continuity
725
+ of g. The remaining N − 1 sectional curvatures are singular, and by considering an N-dimensional
726
+ region that encloses a portion of F and has small thickness in the direction that is g-orthogonal of F,
727
+ one can use the Gauss-Bonnet theorem (along two-dimensional slices) to approximate the (volume-
728
+ )integrated sum of these sectional curvatures by the (surface-)integrated jump in the mean curvature
729
+ across F.
730
+ (In this calculation, one must bear in mind that sectional curvatures and Gaussian
731
+ curvatures are related via the Gauss-Codazzi equations.) See the discussion after Definition 3.1
732
+ in [4], as well as [26], for more insight in dimension N = 2. See also [13] for a justification of
733
+ Definition 3.1 in the case where g is piecewise constant and N ≥ 2.
734
+ In the sequel, we will consistently use the letters T, F, and S to refer to simplices of dimension
735
+ N, N − 1, and N − 2, respectively. We will therefore write �
736
+ T , �
737
+ F , and �
738
+ S in place of �
739
+ T∈T N ,
740
+
741
+ F ∈T N−1, and �
742
+ S∈T N−2, respectively. When we wish to sum over interior simplices of a given
743
+ dimension, we put a ring on top of the summation symbol. Thus, for example, ˚
744
+
745
+ F is shorthand
746
+ for �
747
+ F ∈˚
748
+ T N−1.
749
+ 3.1
750
+ Evolution of the distributional scalar curvature
751
+ We are interested in how (16) changes under deformations of the metric. To this end, consider a
752
+ one-parameter family of Regge metrics g(t) with time derivative
753
+ σ = ∂
754
+ ∂tg.
755
+ Our goal will be to compute
756
+ d
757
+ dt⟨(Rω)dist(g(t)), v⟩V ′,V
758
+ with v ∈ V arbitrary.
759
+ According to Propositions 2.1 and 2.2, the derivatives of the first two terms on the right-hand
760
+ side of (16) satisfy
761
+ d
762
+ dt
763
+
764
+ T
765
+ RT vωT =
766
+
767
+ T
768
+ (div div Sσ − ⟨G, σ⟩) vωT
769
+ and
770
+ 2 d
771
+ dt
772
+
773
+ F
774
+ �H�FvωF = −
775
+
776
+ F
777
+ ��
778
+ II, σ|F
779
+
780
+ + (div Sσ)(n) + divF (σ(n, ·)) − Hσ(n, n)
781
+
782
+ vωF .
783
+ (17)
784
+ For the third term on the right-hand side of (16), we use the following lemma.
785
+ 12
786
+
787
+ Lemma 3.2. Along any interior (N − 2)-simplex S, we have
788
+
789
+ ∂t(ΘSωS) = 1
790
+ 2
791
+ ��
792
+ F ⊃S
793
+ �σ(n, τ)�F + ΘS Tr(σ|S)
794
+
795
+ ωS,
796
+ where the sum is over all (N −1)-simplices F that contain S, n is the unit normal to F with respect
797
+ to g, and τ is the unit vector with respect to g that points into F from S and is g-orthogonal to
798
+ both S and n. Here, our convention is that if F is shared by two N-simplices T + and T −, then
799
+ �σ(n, τ)�F = σ+(n+, τ) + σ−(n−, τ),
800
+ where σ± = σ|T ± and n± points outward from T ±.
801
+ Remark 3.3. Note that n generally differs on either side of F, whereas τ does not, because g has
802
+ single-valued tangential-tangential components along F.
803
+ Proof. We compute
804
+ ˙ΘS = −
805
+
806
+ T⊃S
807
+ ˙θST
808
+ and use Lemma 2.7 to differentiate each angle θST.
809
+ The resulting expression for ˙ΘS involves
810
+ differences between σ(n, τ) evaluated on consecutive pairs of faces F emanating from S. This sum
811
+ can be rearranged to give
812
+ ˙ΘS = 1
813
+ 2
814
+
815
+ F ⊃S
816
+ �σ(n, τ)�F.
817
+ (18)
818
+ We thus get
819
+
820
+ ∂t(ΘSωS) = ˙ΘSωS + ΘS ˙ωS
821
+ = 1
822
+ 2
823
+
824
+ F ⊃S
825
+ �σ(n, τ)�F ωS + 1
826
+ 2ΘS Tr (σ|S) ωS.
827
+ It follows from the above lemma that
828
+ 2 d
829
+ dt
830
+
831
+ S
832
+ ΘSvωS =
833
+
834
+ S
835
+
836
+ F ⊃S
837
+ �σ(n, τ)�F vωS +
838
+
839
+ S
840
+ ΘS Tr(σ|S)vωS
841
+ =
842
+
843
+ S
844
+
845
+ F ⊃S
846
+ �σ(n, τ)�F vωS +
847
+
848
+ S
849
+ ⟨ΘSg|S, σ|S⟩ vωS.
850
+ Collecting our results, we obtain
851
+ d
852
+ dt⟨(Rω)dist(g(t)), v⟩V ′,V =
853
+
854
+ T
855
+
856
+ T
857
+ (div div Sσ)vωT
858
+ − ˚
859
+
860
+ F
861
+
862
+ F
863
+ �(div Sσ)(n) + divF (σ(n, ·)) − Hσ(n, n)�F vωF + ˚
864
+
865
+ S
866
+
867
+ S
868
+
869
+ F ⊃S
870
+ �σ(n, τ)�F vωS
871
+ (19)
872
+
873
+
874
+ T
875
+
876
+ T
877
+ ⟨G, σ⟩vωT − ˚
878
+
879
+ F
880
+
881
+ F
882
+
883
+ �II�F, σ|F
884
+
885
+ vωF + ˚
886
+
887
+ S
888
+
889
+ S
890
+ ⟨ΘSg|S, σ|S⟩ vωS.
891
+ We will now use integration by parts to rewrite the first three terms in a way that involves no
892
+ derivatives of σ.
893
+ 13
894
+
895
+ Lemma 3.4. For any v ∈ V , we have
896
+
897
+ T
898
+
899
+ T
900
+ (div div Sσ)vωT − ˚
901
+
902
+ F
903
+
904
+ F
905
+ �(div Sσ)(n) + divF (σ(n, ·)) − Hσ(n, n)�F vωF
906
+ + ˚
907
+
908
+ S
909
+
910
+ S
911
+
912
+ F ⊃S
913
+ �σ(n, τ)�FvωS =
914
+
915
+ T
916
+
917
+ T
918
+ ⟨Sσ, ∇∇v⟩ω −
919
+
920
+ F
921
+
922
+ F
923
+ Sσ(n, n)�∇nv�ωF.
924
+ Proof. We have
925
+
926
+ T
927
+
928
+ T
929
+ ⟨Sσ, ∇∇v⟩ω −
930
+
931
+ F
932
+
933
+ F
934
+ Sσ(n, n)�∇nv�ωF
935
+ (20)
936
+ =
937
+
938
+ T
939
+ � �
940
+ T
941
+ ⟨Sσ, ∇∇v⟩ω −
942
+
943
+ ∂T
944
+ Sσ(n, n)∇nv ω∂T
945
+
946
+ =
947
+
948
+ T
949
+ � �
950
+ ∂T
951
+ Sσ(n, ∇v)ω∂T −
952
+
953
+ T
954
+ (div Sσ)(∇v)ω −
955
+
956
+ ∂T
957
+ Sσ(n, n)∇nv ω∂T
958
+
959
+ =
960
+
961
+ T
962
+ � �
963
+ ∂T
964
+ Sσ(n, ∇v)ω∂T −
965
+
966
+ ∂T
967
+ (div Sσ)(n)vω∂T +
968
+
969
+ T
970
+ (div div Sσ)vω
971
+
972
+
973
+ ∂T
974
+ Sσ(n, n)∇nv ω∂T
975
+
976
+ .
977
+ (21)
978
+ Note that here we are regarding ∇v as a vector field rather than a one-form. On each N-simplex
979
+ T, we can write
980
+
981
+ ∂T Sσ(n, ∇v)ω∂T −
982
+
983
+ ∂T Sσ(n, n)∇nv ω∂T as a sum of integrals over faces F ⊂ ∂T:
984
+
985
+ ∂T
986
+ Sσ(n, ∇v)ω∂T −
987
+
988
+ ∂T
989
+ Sσ(n, n)∇nv ω∂T =
990
+
991
+ F ⊂∂T
992
+
993
+ F
994
+ Sσ(n, ∇v − n∇nv)ωF
995
+ =
996
+
997
+ F ⊂∂T
998
+
999
+ F
1000
+ Sσ(n, ∇F v)ωF
1001
+ =
1002
+
1003
+ F ⊂∂T
1004
+
1005
+ F
1006
+ σ(n, ∇F v)ωF .
1007
+ In the last line above, we used the fact that ∇F v is g-orthogonal to n, so
1008
+ Sσ(n, ∇Fv) = σ(n, ∇F v) − g(n, ∇F v) Tr σ = σ(n, ∇F v).
1009
+ Each integral over F can be integrated by parts as follows. We have
1010
+ σ(n, ∇F v) = divF (σ(n, ·)v) − divF (σ(n, ·)) v,
1011
+ so the identity (5) applied to α = σ(n, ·)v implies that
1012
+
1013
+ F
1014
+ σ(n, ∇F v)ωF =
1015
+
1016
+ ∂F
1017
+ σ(n, νF )vω∂F −
1018
+
1019
+ F
1020
+ (divF (σ(n, ·)) − Hσ(n, n)) vωF.
1021
+ Now we insert this result into (21) to get
1022
+
1023
+ T
1024
+
1025
+ T
1026
+ ⟨Sσ, ∇∇v⟩ω −
1027
+
1028
+ F
1029
+
1030
+ F
1031
+ Sσ(n, n)�∇nv�ωF
1032
+ =
1033
+
1034
+ T
1035
+ � �
1036
+ F ⊂∂T
1037
+
1038
+ ∂F
1039
+ σ(n, νF )vω∂F −
1040
+
1041
+ F ⊂∂T
1042
+
1043
+ F
1044
+ (divF (σ(n, ·)) − Hσ(n, n)) vωF
1045
+
1046
+
1047
+ ∂T
1048
+ (div Sσ)(n)vω∂T +
1049
+
1050
+ T
1051
+ (div div Sσ)vω
1052
+
1053
+ .
1054
+ 14
1055
+
1056
+ The first term can be re-expressed as a sum over interior (N − 2)-simplices S using our notation
1057
+ from Lemma 3.2, and the next two terms can be re-expressed in terms of jumps across interior
1058
+ faces F. (Integrals over (N − 2)-simplices S ⊂ ∂Ω and (N − 1)-simplices F ⊂ ∂Ω vanish because
1059
+ v = 0 on ∂Ω.) The result is
1060
+
1061
+ T
1062
+
1063
+ T
1064
+ ⟨Sσ, ∇∇v⟩ω −
1065
+
1066
+ F
1067
+
1068
+ F
1069
+ Sσ(n, n)�∇nv�ωF = ˚
1070
+
1071
+ S
1072
+
1073
+ S
1074
+
1075
+ F ⊃S
1076
+ �σ(n, τ)�FvωS
1077
+ − ˚
1078
+
1079
+ F
1080
+
1081
+ F
1082
+ �divF (σ(n, ·)) − Hσ(n, n) + (div Sσ)(n)� vωF +
1083
+
1084
+ T
1085
+
1086
+ T
1087
+ (div div Sσ)vω.
1088
+ Remark 3.5. Many of the above calculations are similar to the ones in [4, Proposition 4.2], except
1089
+ that here we are in dimension N rather than 2.
1090
+ We can now state the main result of this subsection.
1091
+ Theorem 3.6. Let g(t) be a family of Regge metrics with time derivative
1092
+
1093
+ ∂tg =: σ. For every
1094
+ v ∈ V , we have
1095
+ d
1096
+ dt⟨(Rω)dist(g(t)), v⟩V ′,V = bh(g; σ, v) − ah(g; σ, v),
1097
+ (22)
1098
+ where
1099
+ bh(g; σ, v) =
1100
+
1101
+ T
1102
+
1103
+ T
1104
+ ⟨Sσ, ∇∇v⟩ω −
1105
+
1106
+ F
1107
+
1108
+ F
1109
+ Sσ(n, n)�∇nv�FωF,
1110
+ ah(g; σ, v) =
1111
+
1112
+ T
1113
+
1114
+ T
1115
+ ⟨G, σ⟩vωT + ˚
1116
+
1117
+ F
1118
+
1119
+ F
1120
+
1121
+ �II�F, σ|F
1122
+
1123
+ vωF − ˚
1124
+
1125
+ S
1126
+
1127
+ S
1128
+ ⟨ΘSg|S, σ|S⟩ vωS.
1129
+ Proof. Combine (19) with Lemma 3.4.
1130
+ 3.2
1131
+ Distributional densitized Einstein tensor
1132
+ We now pause to make a few remarks about the bilinear forms ah(g; ·, ·) and bh(g; ·, ·) appearing
1133
+ in Theorem 3.6. These remarks will play no role in our analysis, but they help to elucidate the
1134
+ content of Theorem 3.6. The reader can safely skip ahead to Section 4 if desired.
1135
+ Numerical analysts will likely recognize the bilinear form bh(g; ·, ·) appearing in Theorem 3.6. As
1136
+ we mentioned in Section 1, it is (up to the appearance of S) a non-Euclidean, N-dimensional gener-
1137
+ alization of a bilinear form that appears in the Hellan-Herrmann-Johnson finite element method [1–
1138
+ 3, 5–7, 9, 22]. It can be regarded as the integral of div div Sσ against v, where div div is interpreted
1139
+ in a distributional sense.
1140
+ The bilinear form ah(g; ·, ·) can be understood by comparing Theorem 3.6 with Proposition 2.1,
1141
+ which, when integrated against a continuous function v, states that for a family of smooth Rieman-
1142
+ nian metrics g(t) with scalar curvature R,
1143
+ d
1144
+ dt
1145
+
1146
+
1147
+ Rvω =
1148
+
1149
+
1150
+ (div div Sσ)vω −
1151
+
1152
+
1153
+ ⟨G, σ⟩vω,
1154
+ (23)
1155
+ where σ = ∂
1156
+ ∂tg and G = Ric − 1
1157
+ 2Rg is the Einstein tensor associated with g. A comparison of (23)
1158
+ with (22) suggests that for a Regge metric g, the bilinear form ah(g; σ, v) should be regarded as a
1159
+ distributional counterpart of
1160
+
1161
+ Ω⟨G, σ⟩vω.
1162
+ 15
1163
+
1164
+ This motivates the following definition. Fix a number s > 1, and let Σ denote the space of
1165
+ square-integrable symmetric (0, 2)-tensor fields σ with the following properties: the restriction of
1166
+ σ to each T ∈ T N belongs to Hs(T), and the tangential-tangential components of σ along any
1167
+ face F ∈ ˚
1168
+ T N−1 are single-valued. Note that these conditions imply that the tangential-tangential
1169
+ components of σ along any S ∈ ˚
1170
+ T N−2 are well-defined and single-valued as well.
1171
+ Definition 3.7. Let g be a Regge metric. The distributional densitized Einstein tensor associated
1172
+ with g is the linear functional (Gω)dist(g) ∈ Σ′ defined by
1173
+ ⟨(Gω)dist(g), σ⟩Σ′,Σ =
1174
+
1175
+ T
1176
+
1177
+ T
1178
+ ⟨G, σ⟩ωT + ˚
1179
+
1180
+ F
1181
+
1182
+ F
1183
+
1184
+ �II�F, σ|F
1185
+
1186
+ ωF − ˚
1187
+
1188
+ S
1189
+
1190
+ S
1191
+ ⟨ΘSg|S, σ|S⟩ ωS,
1192
+ ∀σ ∈ Σ.
1193
+ Remark 3.8. In dimension N = 2, we have (Gω)dist(g) = 0 for any Regge metric g, because
1194
+ G vanishes within each triangle, ¯II vanishes on each edge, and the restriction of σ to each vertex
1195
+ vanishes.
1196
+ Remark 3.9. The appearance of the trace-reversed second fundamental form II in Definition 3.7
1197
+ is quite natural. The same quantity arises in studies of singular sources in general relativity, with
1198
+ the jump in II encoding the well-known Israel junction conditions across a hypersurface on which
1199
+ stress-energy is concentrated [20].
1200
+ Remark 3.10. If we define a map (div div S)dist : Σ → V ′ by
1201
+ ⟨(div div S)distσ, v⟩V ′,V = bh(g; σ, v),
1202
+ ∀v ∈ V,
1203
+ then, by construction, we have
1204
+ d
1205
+ dt
1206
+ ����
1207
+ t=0
1208
+ ⟨(Rω)dist(g + tσ), v⟩V ′,V = ⟨(div div S)distσ, v⟩V ′,V − ⟨(Gω)dist(g), vσ⟩Σ′,Σ
1209
+ for every piecewise smooth σ ∈ Σ and every smooth function v with compact support in Ω. In
1210
+ particular, suppose that Ω has no boundary (e.g., suppose that Ω is an N-dimensional cube and
1211
+ we identify its opposing faces). Then bh(g; σ, 1) = 0 and
1212
+ d
1213
+ dt
1214
+ ����
1215
+ t=0
1216
+ ⟨(Rω)dist(g + tσ), 1⟩V ′,V = −⟨(Gω)dist(g), σ⟩Σ′,Σ
1217
+ for every piecewise smooth σ ∈ Σ. This implies that a Regge metric g is a stationary point of
1218
+ ⟨(Rω)dist(g), 1⟩Σ′,Σ if its distributional densitized Einstein tensor vanishes: (Gω)dist(g) = 0.
1219
+ The functional ⟨(Rω)dist(g), 1⟩Σ′,Σ is a counterpart of the Einstein-Hilbert functional
1220
+
1221
+ Ω Rω from
1222
+ general relativity, whose stationary points are solutions to the (vacuum) Einstein field equations
1223
+ G = 0. It reduces to the Regge action from Regge calculus when g is piecewise constant. That is,
1224
+ ⟨(Rω)dist(g), 1⟩Σ′,Σ = 2 ˚
1225
+
1226
+ S
1227
+ ΘSVS,
1228
+ if g is piecewise constant,
1229
+ where VS =
1230
+
1231
+ S ωS denotes the volume of S. If g varies with t and remains piecewise constant for
1232
+ all t, then
1233
+ d
1234
+ dt2 ˚
1235
+
1236
+ S
1237
+ ΘSVS = 2 ˚
1238
+
1239
+ S
1240
+ ˙ΘSVS + 2 ˚
1241
+
1242
+ S
1243
+ ΘS ˙VS,
1244
+ 16
1245
+
1246
+ and one checks that (on a domain without boundary)
1247
+ 2 ˚
1248
+
1249
+ S
1250
+ ˙ΘSVS = bh(g; σ, 1) = 0
1251
+ and
1252
+ 2 ˚
1253
+
1254
+ S
1255
+ ΘS ˙VS = −ah(g; σ, 1) = −⟨(Gω)dist(g), σ⟩Σ′,Σ,
1256
+ where σ =
1257
+
1258
+ ∂tg. The fact that ˚
1259
+
1260
+ S ˙ΘSVS = 0 for any piecewise constant Regge metric g (on a
1261
+ domain without boundary) was proved in Regge’s classic paper [23] using very different techniques.
1262
+ Remark 3.11. If g is a Regge metric and σ = gv for some smooth function v with compact support
1263
+ in Ω, then:
1264
+ 1. On each N-simplex T, we have
1265
+ ⟨G, σ⟩ = ⟨G, g⟩v = (Tr G)v = −
1266
+ �N − 2
1267
+ 2
1268
+
1269
+ Rv.
1270
+ 2. On either side of each interior (N − 1)-simplex F, we have:
1271
+
1272
+ II, σ|F
1273
+
1274
+ = ⟨II, g|F ⟩ v − ⟨g|F , g|F⟩ Hv
1275
+ = Hv − (N − 1)Hv
1276
+ = −(N − 2)Hv.
1277
+ 3. On each interior (N − 2)-simplex S, we have
1278
+ ⟨ΘSg|S, σ|S⟩ = ΘSv Tr(g|S) = (N − 2)ΘSv.
1279
+ This shows that
1280
+ ⟨(Gω)dist(g), gv⟩Σ′,Σ = −
1281
+ �N − 2
1282
+ 2
1283
+ � ��
1284
+ T
1285
+
1286
+ T
1287
+ RT vωT + 2 ˚
1288
+
1289
+ F
1290
+
1291
+ F
1292
+ �H�FvωF + 2 ˚
1293
+
1294
+ S
1295
+
1296
+ S
1297
+ ΘSvωS
1298
+
1299
+ = −
1300
+ �N − 2
1301
+ 2
1302
+
1303
+ ⟨(Rω)dist(g), v⟩V ′,V
1304
+ for every smooth function v with compact support in Ω. One can interpret this as saying that the
1305
+ trace of (Gω)dist(g) is −
1306
+ � N−2
1307
+ 2
1308
+
1309
+ (Rω)dist(g).
1310
+ Remark 3.12. If g is a piecewise constant Regge metric and σ ∈ Σ is piecewise constant, then
1311
+ ⟨(Gω)dist(g), σ⟩Σ′,Σ = − ˚
1312
+
1313
+ S
1314
+
1315
+ S
1316
+ ΘS Tr(σ|S)ωS.
1317
+ If we linearize around the Euclidean metric g = δ, then we see from (18) that
1318
+ d
1319
+ dt
1320
+ ����
1321
+ t=0
1322
+ ⟨(Gω)dist(δ + tρ), σ⟩Σ′,Σ = − ˚
1323
+
1324
+ S
1325
+
1326
+ S
1327
+ ˙ΘS Tr(σ|S)ωS
1328
+ = −1
1329
+ 2
1330
+ ˚
1331
+
1332
+ S
1333
+
1334
+ S
1335
+
1336
+ F ⊃S
1337
+ �ρ(n, τ)�F Tr(σ|S)ωS
1338
+ 17
1339
+
1340
+ for every piecewise constant ρ, σ ∈ Σ. (Note that there are no additional terms on the right-hand
1341
+ side because ΘS = 0 at t = 0.) Hence, if Ω has no boundary, then
1342
+ d2
1343
+ dt2
1344
+ ����
1345
+ t=0
1346
+ ⟨(Rω)dist(δ + tσ), 1⟩V ′,V = − d
1347
+ dt
1348
+ ����
1349
+ t=0
1350
+ ⟨(Gω)dist(δ + tσ), σ⟩Σ′,Σ
1351
+ = 1
1352
+ 2
1353
+ ˚
1354
+
1355
+ S
1356
+
1357
+ S
1358
+
1359
+ F ⊃S
1360
+ �σ(n, τ)�F Tr(σ|S)ωS
1361
+ for every piecewise constant σ ∈ Σ. This is equivalent to Christiansen’s formula [12, Theorem 2
1362
+ and Equations (25-26)] for the second variation of the Regge action around the Euclidean metric
1363
+ in dimension N = 3. (There, the Regge action is taken to be 1
1364
+ 2⟨(Rω)dist(g), 1⟩V ′,V rather than
1365
+ ⟨(Rω)dist(g), 1⟩V ′,V .)
1366
+ 4
1367
+ Convergence
1368
+ In this section, we prove a convergence result for the distributional densitized scalar curvature in
1369
+ the norm
1370
+ ∥u∥H−2(Ω) =
1371
+ sup
1372
+ v∈H2
1373
+ 0(Ω),
1374
+ v̸=0
1375
+ ⟨u, v⟩H−2(Ω),H2
1376
+ 0(Ω)
1377
+ ∥v∥H2(Ω)
1378
+ .
1379
+ (24)
1380
+ Our convergence result will be applicable to a family {gh}h>0 of Regge metrics defined on a shape-
1381
+ regular family {Th}h>0 of triangulations of Ω parametrized by h = maxT∈T N
1382
+ h hT , where hT =
1383
+ diam(T). Shape-regularity means that there exists a constant C0 independent of h such that
1384
+ max
1385
+ T∈T N
1386
+ h
1387
+ hT
1388
+ ρT
1389
+ ≤ C0
1390
+ for all h > 0, where ρT denotes the inradius of T.
1391
+ Theorem 4.1. Let Ω ⊂ RN be a polyhedral domain equipped with a smooth Riemannian metric g.
1392
+ Let {gh}h>0 be a family of Regge metrics defined on a shape-regular family {Th}h>0 of triangulations
1393
+ of Ω. Assume that limh→0 ∥gh − g∥L∞(Ω) = 0 and C1 := suph>0 maxT∈T N
1394
+ h ∥gh∥W 1,∞(T) < ∞. The
1395
+ following statements hold:
1396
+ (i) If N = 2, then there exist positive constants C and h0 such that
1397
+ ∥(Rω)dist(gh) − (Rω)(g)∥H−2(Ω) ≤ C
1398
+
1399
+ 1 + max
1400
+ T
1401
+ h−1
1402
+ T ∥gh − g∥L∞(T) + max
1403
+ T
1404
+ |gh − g|W 1,∞(T)
1405
+
1406
+ ×
1407
+
1408
+ ∥gh − g∥2
1409
+ L2(Ω) +
1410
+
1411
+ T
1412
+ h2
1413
+ T |gh − g|2
1414
+ H1(T)
1415
+ �1/2
1416
+ (25)
1417
+ for all h ≤ h0. The constants C and h0 depend on ∥g∥W 1,∞(Ω), ∥g−1∥L∞(Ω), C0, and C1.
1418
+ 18
1419
+
1420
+ (ii) If N ≥ 3, assume additionally that C2 := suph>0 maxT∈T N
1421
+ h |gh|W 2,∞(T) < ∞. Then there exist
1422
+ positive constants C and h0 such that
1423
+ ∥(Rω)dist(gh) − (Rω)(g)∥H−2(Ω) ≤ C
1424
+
1425
+ 1 + max
1426
+ T
1427
+ h−2
1428
+ T ∥gh − g∥L∞(T) + max
1429
+ T
1430
+ h−1
1431
+ T |gh − g|W 1,∞(T)
1432
+
1433
+ ×
1434
+
1435
+ ∥gh − g∥2
1436
+ L2(Ω) +
1437
+
1438
+ T
1439
+ h2
1440
+ T |gh − g|2
1441
+ H1(T) +
1442
+
1443
+ T
1444
+ h4
1445
+ T |gh − g|2
1446
+ H2(T)
1447
+ �1/2
1448
+ (26)
1449
+ for all h ≤ h0. The constants C and h0 depend on N, ∥g∥W 1,∞(Ω), ∥g−1∥L∞(Ω), C0, C1, and
1450
+ C2.
1451
+ The above theorem leads immediately to error estimates of optimal order for piecewise poly-
1452
+ nomial interpolants of g having degree r ≥ 0, provided that either N = 2 or r ≥ 1. To make this
1453
+ statement precise, we introduce a definition. Recall that the Regge finite element space of degree
1454
+ r ≥ 0 consists of symmetric (0, 2)-tensor fields on Ω that are piecewise polynomial of degree at
1455
+ most r and possess single-valued tangential-tangential components on interior (N − 1)-simplices.
1456
+ Definition 4.2. Let Ih be a map that sends smooth symmetric (0, 2)-tensor fields on Ω to the Regge
1457
+ finite element space of degree r ≥ 0. We say that Ih is an optimal-order interpolation operator of
1458
+ degree r if there exists a number m ∈ {0, 1, . . . , N} and a constant C3 = C3(N, r, hT /ρT , t, s) such
1459
+ that for every p ∈ [1, ∞], every s ∈ (m/p, r + 1], every t ∈ [0, s], and every symmetric (0, 2)-tensor
1460
+ field g possessing W s,p(Ω)-regularity, Ihg exists (upon continuously extending Ih) and satisfies
1461
+ |Ihg − g|W t,p(T) ≤ C3hs−t
1462
+ T
1463
+ |g|W s,p(T)
1464
+ (27)
1465
+ for every T ∈ T N
1466
+ h .
1467
+ We call the number m the codimension index of Ih.
1468
+ A Regge metric gh
1469
+ is called an optimal-order interpolant of g having degree r and codimension index m if it is the
1470
+ image of a Riemannian metric g under an optimal-order interpolation operator having degree r and
1471
+ codimension index m.
1472
+ An example of an optimal-order interpolation operator is the canonical interpolation operator
1473
+ onto the degree-r Regge finite element space introduced in [21, Chapter 2]. Its degrees of freedom
1474
+ involve integrals over simplices of codimension at most N − 1, so its action on a tensor field g is
1475
+ well-defined so long as g admits traces on simplices of codimension at most N − 1, i.e. g possesses
1476
+ W s,p(Ω)-regularity with s > (N − 1)/p. Correspondingly, its codimension index is m = N − 1.
1477
+ Corollary 4.3. Let Ω, g, and {Th}h>0 be as in Theorem 4.1. Let {gh}h>0 be a family of optimal-
1478
+ order interpolants of g having degree r ≥ 0 and codimension index m. If N ≥ 3, assume that r ≥ 1.
1479
+ Then there exist positive constants C and h0 such that
1480
+ ∥(Rω)dist(gh) − (Rω)(g)∥H−2(Ω) ≤ C
1481
+ ��
1482
+ T
1483
+ hp(r+1)
1484
+ T
1485
+ |g|p
1486
+ W r+1,p(T)
1487
+ �1/p
1488
+ for all h ≤ h0 and all p ∈ [2, ∞] satisfying p >
1489
+ m
1490
+ r+1.
1491
+ (We interpret the right-hand side as
1492
+ C maxT hr+1
1493
+ T
1494
+ |g|W r+1,∞(T) if p = ∞.) The constants C and h0 depend on the same quantities listed
1495
+ in (i) (if N = 2) and (ii) (if N ≥ 3), as well as on Ω, r, and (if N ≥ 3) |g|W 2,∞(Ω).
1496
+ 19
1497
+
1498
+ Remark 4.4. The corollary above continues to hold if we allow slightly more general interpolants
1499
+ in Definition 4.2. For example, it holds if (27) is replaced by
1500
+ |Ihg − g|W t,p(T) ≤ C3hs−t
1501
+ T
1502
+
1503
+ T ′:T ′∩T̸=∅
1504
+ |g|W s,p(T ′),
1505
+ (28)
1506
+ where the sum is over all T ′ ∈ T N
1507
+ h
1508
+ that share a subsimplex with T.
1509
+ In what follows, we reuse the letter C to denote a positive constant that may change at each
1510
+ occurrence and may depend on N, ∥g∥W 1,∞(Ω), ∥g−1∥L∞(Ω), C0, and C1. Beginning in Lemma 4.8,
1511
+ we allow C to also depend on C2.
1512
+ Our strategy for proving Theorem 4.1 will be to consider an evolving metric
1513
+ �g(t) = (1 − t)g + tgh
1514
+ with time derivative
1515
+ σ = ∂
1516
+ ∂t�g(t) = gh − g.
1517
+ Note that �g(t), being piecewise smooth and tangential-tangential continuous, is a Regge metric for
1518
+ all t ∈ [0, 1], and it happens to be a (globally) smooth Riemannian metric at t = 0. Since �g(0) = g
1519
+ and �g(1) = gh, Theorem 3.6 implies that
1520
+ ⟨(Rω)dist(gh) − (Rω)(g), v⟩V ′,V =
1521
+ � 1
1522
+ 0
1523
+ bh(�g(t); σ, v) − ah(�g(t); σ, v) dt,
1524
+ ∀v ∈ V.
1525
+ Thus, we can estimate (Rω)dist(gh) − (Rω)(g) by estimating the bilinear forms bh(�g(t); ·, ·) and
1526
+ ah(�g(t); ·, ·).
1527
+ To do this, we introduce some notation. Given any Regge metric g, we let ∇g and ∇ denote
1528
+ the covariant derivatives with respect to g and δ, respectively. Similarly, we append a subscript
1529
+ g to other operators like Tr, S, and div when they are taken with respect to g, and we omit the
1530
+ subscript when they are taken with respect to δ. On the boundary of any N-simplex T, we let ng
1531
+ and n denote the outward unit normal vectors with respect to g|T and δ, respectively. These two
1532
+ vectors are related to one another in coordinates via
1533
+ ng =
1534
+ 1
1535
+
1536
+ nT g−1n
1537
+ g−1n,
1538
+ (29)
1539
+ where we are thinking of g as a matrix and n and ng as column vectors. We write ⟨·, ·⟩g for the
1540
+ g-inner product of two tensor fields. If D is a submanifold of Ω on which the induced metric g|D
1541
+ is well-defined, and if ρ is a tensor field on D, then we denote
1542
+ ∥ρ∥Lp(D,g) =
1543
+ ���
1544
+ D |ρ|p
1545
+ g ωD(g)
1546
+ �1/p ,
1547
+ if 1 ≤ p < ∞,
1548
+ supD |ρ|g,
1549
+ if p = ∞,
1550
+ where ωD(g) is the induced volume form on D and |ρ|g = ⟨ρ, ρ⟩1/2
1551
+ g
1552
+ . We abbreviate ∥ · ∥Lp(D) =
1553
+ ∥ · ∥Lp(D,δ) and | · | = | · |δ.
1554
+ We introduce two metric-dependent, mesh-dependent norms. For v ∈ V , we set
1555
+ ∥v∥2
1556
+ 2,h,g =
1557
+
1558
+ T
1559
+ ∥∇g∇gv∥2
1560
+ L2(T,g) +
1561
+
1562
+ F
1563
+ h−1
1564
+ F ∥�dv(ng)�∥2
1565
+ L2(F,g) .
1566
+ 20
1567
+
1568
+ If σ is a symmetric (0, 2)-tensor field with the property that σ(ng, ng) is well-defined and single-
1569
+ valued on every F ∈ T N−1
1570
+ h
1571
+ , then we set
1572
+ ∥σ∥2
1573
+ 0,h,g =
1574
+
1575
+ T
1576
+ ∥σ∥2
1577
+ L2(T,g) +
1578
+
1579
+ F
1580
+ hF ∥σ(ng, ng)∥2
1581
+ L2(F,g),
1582
+ where hF is the Euclidean diameter of F. Note that the image under Sg of any symmetric (0, 2)-
1583
+ tensor field possessing single-valued tangential-tangential components along faces automatically
1584
+ possesses single-valued normal-normal components along faces, because
1585
+ Sgσ(ng, ng) = σ(ng, ng) − g(ng, ng) Trg σ = − Trg (σ|F ) .
1586
+ Now we return to the setting of Theorem 4.1 and the discussion thereafter: g is a smooth
1587
+ Riemannian metric, gh is a Regge metric, �g(t) = (1 − t)g + tgh, and σ = gh − g. We assume
1588
+ throughout what follows that limh→0 ∥gh − g∥L∞(Ω) = 0 and suph>0 maxT∈T N
1589
+ h ∥gh∥W 1,∞(T) < ∞.
1590
+ These assumptions have some elementary consequences that we record here for reference (see [16]
1591
+ for a derivation). For every h sufficiently small, every t ∈ [0, 1], and every vector w with unit
1592
+ Euclidean length,
1593
+ ∥�g∥L∞(Ω) + ∥�g−1∥L∞(Ω) ≤ C,
1594
+ (30)
1595
+ max
1596
+ T
1597
+ |�g|W 1,∞(T) ≤ C,
1598
+ (31)
1599
+ C−1 ≤ inf
1600
+ Ω (wT �gw) ≤ sup
1601
+
1602
+ (wT �gw) ≤ C,
1603
+ (32)
1604
+ where we are thinking of �g as a matrix and w as a column vector in the last line. Note that the
1605
+ last line implies the existence of positive lower and upper bounds on wT �g−1w as well:
1606
+ C−1 ≤ inf
1607
+ Ω (wT �g−1w) ≤ sup
1608
+
1609
+ (wT �g−1w) ≤ C.
1610
+ (33)
1611
+ In addition, the inequalities ∥�g∥L∞(Ω) ≤ C and ∥�g−1∥L∞(Ω) ≤ C imply that
1612
+ C−1∥ρ∥Lp(D,�g(t2)) ≤ ∥ρ∥Lp(D,�g(t1)) ≤ C∥ρ∥Lp(D,�g(t2))
1613
+ (34)
1614
+ and
1615
+ C−1∥ρ∥Lp(D) ≤ ∥ρ∥Lp(D,�g(t1)) ≤ C∥ρ∥Lp(D)
1616
+ (35)
1617
+ for every t1, t2 ∈ [0, 1], every admissible submanifold D, every p ∈ [1, ∞], every tensor field ρ having
1618
+ finite Lp(D)-norm, and every h sufficiently small. We select h0 > 0 so that (30-35) hold for all
1619
+ h ≤ h0, and we tacitly use these inequalities throughout our analysis.
1620
+ We will show the following near-equivalence of the norms ∥ · ∥2,h,�g and ∥ · ∥2,h,g.
1621
+ Proposition 4.5. For every v ∈ V , every h ≤ h0, and every t ∈ [0, 1],
1622
+ ∥v∥2
1623
+ 2,h,�g ≤ C
1624
+
1625
+ ∥v∥2
1626
+ 2,h,g +
1627
+
1628
+ max
1629
+ T
1630
+ h−2
1631
+ T ∥gh − g∥2
1632
+ L∞(T) + max
1633
+ T
1634
+ |gh − g|2
1635
+ W 1,∞(T)
1636
+
1637
+ ×
1638
+
1639
+ T
1640
+
1641
+ ∥dv∥2
1642
+ L2(T) + h2
1643
+ T |dv|2
1644
+ H1(T)
1645
+ � �
1646
+ .
1647
+ The proof of Proposition 4.5 relies on the following lemma.
1648
+ 21
1649
+
1650
+ Lemma 4.6. Let g1 and g2 be two symmetric positive definite matrices, and let n be a unit vector.
1651
+ Let
1652
+ ngi =
1653
+ 1
1654
+
1655
+ nTg−1
1656
+ i
1657
+ n
1658
+ g−1
1659
+ i
1660
+ n,
1661
+ i = 1, 2.
1662
+ Then there exists a constant c depending on |g1|, |g2|, |g−1
1663
+ 1 |, |g−1
1664
+ 2 | such that
1665
+ |ng1 − ng2| ≤ c|g1 − g2|.
1666
+ Proof. Using the identity
1667
+ 1
1668
+
1669
+ nT g−1
1670
+ 1 n
1671
+
1672
+ 1
1673
+
1674
+ nTg−1
1675
+ 2 n
1676
+ =
1677
+ nT(g−1
1678
+ 2
1679
+ − g−1
1680
+ 1 )n
1681
+ nTg−1
1682
+ 1 n
1683
+
1684
+ nTg−1
1685
+ 2 n + nT g−1
1686
+ 2 n
1687
+
1688
+ nT g−1
1689
+ 1 n
1690
+ ,
1691
+ (36)
1692
+ we can write
1693
+ ng1 − ng2 =
1694
+ nT (g−1
1695
+ 2
1696
+ − g−1
1697
+ 1 )n
1698
+ nT g−1
1699
+ 1 n
1700
+
1701
+ nT g−1
1702
+ 2 n + nTg−1
1703
+ 2 n
1704
+
1705
+ nTg−1
1706
+ 1 n
1707
+ g−1
1708
+ 1 n +
1709
+ 1
1710
+
1711
+ nT g−1
1712
+ 2 n
1713
+ (g−1
1714
+ 1
1715
+ − g−1
1716
+ 2 )n.
1717
+ Since g−1
1718
+ 1
1719
+ − g−1
1720
+ 2
1721
+ = g−1
1722
+ 1 (g2 − g1)g−1
1723
+ 2 , the bound follows easily.
1724
+ Notice that in view of (29), Lemma 4.6 implies that
1725
+ ∥n�g − ng∥L∞(F ) ≤ C∥�g − g∥L∞(F )
1726
+ (37)
1727
+ on either side of any face F.
1728
+ Now we are ready to begin proving Proposition 4.5. Consider the term �
1729
+ F h−1
1730
+ F
1731
+ ���dv(n�g)�
1732
+ ��2
1733
+ L2(F,�g)
1734
+ that appears in the definition of ∥v∥2
1735
+ 2,h,�g. Notice that
1736
+ dv(n�g) = dv(ng) + dv(n�g − ng),
1737
+ and we can use the bound (37) to estimate
1738
+ ∥dv(n�g − ng)∥L2(F,�g) ≤ C∥dv(n�g − ng)∥L2(F )
1739
+ ≤ C∥dv∥L2(F )∥n�g − ng∥L∞(F )
1740
+ ≤ C∥dv∥L2(F )∥�g − g∥L∞(F )
1741
+ ≤ C∥dv∥L2(F )∥gh − g∥L∞(F )
1742
+ on either side of F. Using the trace inequality
1743
+ ∥dv∥2
1744
+ L2(F ) ≤ C
1745
+
1746
+ h−1
1747
+ T ∥dv∥2
1748
+ L2(T) + hT |dv|2
1749
+ H1(T)
1750
+
1751
+ ,
1752
+ F ⊂ T ∈ T N
1753
+ h ,
1754
+ (38)
1755
+ it follows that
1756
+
1757
+ F
1758
+ h−1
1759
+ F ∥�dv(n�g)�∥2
1760
+ L2(F,�g)
1761
+ ≤ C
1762
+ ��
1763
+ F
1764
+ h−1
1765
+ F ∥�dv(ng)�∥2
1766
+ L2(F,g) +
1767
+
1768
+ T
1769
+ h−1
1770
+ T
1771
+
1772
+ h−1
1773
+ T ∥dv∥2
1774
+ L2(T) + hT |dv|2
1775
+ H1(T)
1776
+
1777
+ ∥gh − g∥2
1778
+ L∞(T)
1779
+
1780
+ = C
1781
+ ��
1782
+ F
1783
+ h−1
1784
+ F ∥�dv(ng)�∥2
1785
+ L2(F,g) +
1786
+
1787
+ T
1788
+
1789
+ h−2
1790
+ T ∥gh − g∥2
1791
+ L∞(T)∥dv∥2
1792
+ L2(T) + ∥gh − g∥2
1793
+ L∞(T)|dv|2
1794
+ H1(T)
1795
+ ��
1796
+ ,
1797
+ 22
1798
+
1799
+ where we have used (34), (38), and the bound hT ≤ ChF, which follows from the shape-regularity
1800
+ of Th.
1801
+ Next, consider the term �
1802
+ T ∥∇�g∇�gv∥2
1803
+ L2(T,�g) that appears in the definition of ∥v∥2
1804
+ 2,h,�g. Notice
1805
+ that
1806
+
1807
+ ∇�g∇�gv
1808
+
1809
+ ij = (∇g∇gv)ij + (Γk
1810
+ ij − �Γk
1811
+ ij) ∂v
1812
+ ∂xk ,
1813
+ where Γk
1814
+ ij and �Γk
1815
+ ij are the Christoffel symbols of the second kind associated with g and �g, respec-
1816
+ tively. We have
1817
+ ∥Γk
1818
+ ij − �Γk
1819
+ ij∥L∞(T) ≤ C∥�g − g∥W 1,∞(T) ≤ C∥gh − g∥W 1,∞(T),
1820
+ so
1821
+ ∥∇�g∇�gv∥L2(T,�g) ≤ C∥∇�g∇�gv∥L2(T)
1822
+ ≤ C
1823
+
1824
+ ∥∇g∇gv∥L2(T) + ∥gh − g∥W 1,∞(T)∥dv∥L2(T)
1825
+
1826
+ ≤ C
1827
+
1828
+ ∥∇g∇gv∥L2(T,g) + ∥gh − g∥W 1,∞(T)∥dv∥L2(T)
1829
+
1830
+ .
1831
+ It follows that
1832
+ ∥v∥2
1833
+ 2,h,�g ≤ C
1834
+
1835
+ ∥v∥2
1836
+ 2,h,g +
1837
+
1838
+ max
1839
+ T
1840
+ h−2
1841
+ T ∥gh − g∥2
1842
+ L∞(T) + max
1843
+ T
1844
+ |gh − g|2
1845
+ W 1,∞(T)
1846
+
1847
+ ×
1848
+
1849
+ T
1850
+
1851
+ ∥dv∥2
1852
+ L2(T) + h2
1853
+ T |dv|2
1854
+ H1(T)
1855
+ � �
1856
+ .
1857
+ This completes the proof of Proposition 4.5.
1858
+ Our next step will be to estimate the bilinear form bh(�g; ·, ·).
1859
+ Proposition 4.7. For every h ≤ h0, every t ∈ [0, 1], and every v ∈ H2
1860
+ 0(Ω), we have (with
1861
+ σ = gh − g)
1862
+ |bh(�g; σ, v)| ≤ C
1863
+
1864
+ ∥gh − g∥2
1865
+ L2(Ω) +
1866
+
1867
+ T
1868
+ h2
1869
+ T |gh − g|2
1870
+ H1(T)
1871
+ �1/2
1872
+ ×
1873
+
1874
+ 1 + max
1875
+ T
1876
+ h−1
1877
+ T ∥gh − g∥L∞(T) + max
1878
+ T
1879
+ |gh − g|W 1,∞(T)
1880
+
1881
+ ∥v∥H2(Ω).
1882
+ Proof. In view of the definitions of ∥ · ∥0,h,�g and ∥ · ∥2,h,�g, we have
1883
+ |bh(�g; σ, v)| ≤ ∥S�gσ∥0,h,�g∥v∥2,h,�g.
1884
+ (39)
1885
+ Recalling that
1886
+ ∥S�gσ∥2
1887
+ 0,h,�g =
1888
+
1889
+ T
1890
+ ∥S�gσ∥2
1891
+ L2(T,�g) +
1892
+
1893
+ F
1894
+ hF ∥S�gσ(n�g, n�g)∥2
1895
+ L2(F,�g),
1896
+ we compute
1897
+ ⟨S�gσ, S�gσ⟩�g =
1898
+
1899
+ σ − �g⟨�g, σ⟩�g, σ − �g⟨�g, σ⟩�g
1900
+
1901
+ �g
1902
+ = ⟨σ, σ⟩�g − 2⟨�g, σ⟩2
1903
+ �g + ⟨�g, �g⟩�g⟨�g, σ⟩2
1904
+ �g
1905
+ = ⟨σ, σ⟩�g + (N − 2)⟨�g, σ⟩2
1906
+ �g,
1907
+ 23
1908
+
1909
+ which leads to the bound
1910
+ ∥S�gσ∥L2(T,�g) ≤ C∥σ∥L2(T,�g) ≤ C∥σ∥L2(T).
1911
+ Also, by the trace inequality,
1912
+ ∥S�gσ(n�g, n�g)∥2
1913
+ L2(∂T,�g) ≤ C∥S�gσ∥2
1914
+ L2(∂T,�g)
1915
+ ≤ C∥σ∥2
1916
+ L2(∂T)
1917
+ ≤ C
1918
+
1919
+ h−1
1920
+ T ∥σ∥2
1921
+ L2(T) + hT |σ|2
1922
+ H1(T)
1923
+
1924
+ .
1925
+ (Here we are measuring the L2(∂T, �g)-norm of the full tensor S�gσ rather than its restriction to the
1926
+ tangent bundle of ∂T.) Thus,
1927
+ ∥S�gσ∥2
1928
+ 0,h,�g ≤ C
1929
+
1930
+ ∥σ∥2
1931
+ L2(Ω) +
1932
+
1933
+ T
1934
+ h2
1935
+ T |σ|2
1936
+ H1(T)
1937
+
1938
+ = C
1939
+
1940
+ ∥gh − g∥2
1941
+ L2(Ω) +
1942
+
1943
+ T
1944
+ h2
1945
+ T |gh − g|2
1946
+ H1(T)
1947
+
1948
+ .
1949
+ (40)
1950
+ Consider now the term ∥v∥2,h,�g in (39). Proposition 4.5 implies that
1951
+ ∥v∥2,h,�g ≤ C
1952
+
1953
+ ∥v∥2,h,g +
1954
+
1955
+ max
1956
+ T
1957
+ h−1
1958
+ T ∥gh − g∥L∞(T) + max
1959
+ T
1960
+ |gh − g|W 1,∞(T)
1961
+
1962
+ ∥v∥H2(Ω)
1963
+
1964
+ since v ∈ H2
1965
+ 0(Ω).
1966
+ Furthermore, since g is smooth and v ∈ H2
1967
+ 0(Ω), we have �dv(ng)� = 0 on
1968
+ every interior face F and �dv(ng)� = dv(ng) = 0 on every face F ⊂ ∂Ω.
1969
+ Thus, ∥v∥2
1970
+ 2,h,g =
1971
+
1972
+ T ∥∇g∇gv∥2
1973
+ L2(T,g) = ∥∇g∇gv∥2
1974
+ L2(Ω,g). Since
1975
+ (∇g∇gv)ij = (∇∇v)ij − Γk
1976
+ ij
1977
+ ∂v
1978
+ ∂xk ,
1979
+ we see that
1980
+ ∥v∥2,h,g = ∥∇g∇gv∥L2(Ω) ≤ C(|v|H2(Ω) + |v|H1(Ω)) ≤ C∥v∥H2(Ω).
1981
+ Thus,
1982
+ ∥v∥2,h,�g ≤ C
1983
+
1984
+ 1 + max
1985
+ T
1986
+ h−1
1987
+ T ∥gh − g∥L∞(T) + max
1988
+ T
1989
+ |gh − g|W 1,∞(T)
1990
+
1991
+ ∥v∥H2(Ω).
1992
+ (41)
1993
+ Combining (39), (40), and (41) completes the proof.
1994
+ At this point, we have finished proving part (i) of Theorem 4.1. Indeed, in dimension N = 2,
1995
+ ah vanishes, so we can write
1996
+ ��⟨(Rω)dist(gh) − (Rω)(g), v⟩V ′,V
1997
+ �� ≤
1998
+ � 1
1999
+ 0
2000
+ |bh(�g(t); σ, v)| dt
2001
+ and apply Proposition 4.7 to deduce (25).
2002
+ To prove part (ii) of Theorem 4.1, we suppose that N ≥ 3 and that suph>0 maxT∈T N
2003
+ h |gh|W 2,∞(T) <
2004
+ ∞, and we proceed as follows. Recall that
2005
+ ah(�g; σ, v) =
2006
+
2007
+ T
2008
+
2009
+ T
2010
+ ⟨G(�g), σ⟩�gvωT (�g)+ ˚
2011
+
2012
+ F
2013
+
2014
+ F
2015
+
2016
+ �II(�g)�F, σ|F
2017
+
2018
+ �g vωF(�g)− ˚
2019
+
2020
+ S
2021
+
2022
+ S
2023
+ ⟨ΘS(�g)�g|S, σ|S⟩�gvωS(�g),
2024
+ (42)
2025
+ 24
2026
+
2027
+ where have made all dependencies on the metric explicit in the notation. We will bound each of
2028
+ the three terms above, beginning with the first. Throughout what follows, we continue to denote
2029
+ σ = gh − g, and we let v be an arbitrary member of V .
2030
+ Lemma 4.8. We have
2031
+ �����
2032
+
2033
+ T
2034
+
2035
+ T
2036
+ ⟨G(�g), σ⟩�g vωT (�g)
2037
+ ����� ≤ C∥gh − g∥L2(Ω)∥v∥L2(Ω).
2038
+ Proof. Since we are now assuming that suph>0 maxT∈T N
2039
+ h ∥gh∥W 2,∞(T) < ∞, the Einstein tensor
2040
+ associated with �g satisfies
2041
+ ∥G(�g)∥L∞(T) ≤ C
2042
+ for every h ≤ h0, every t ∈ [0, 1], and every T ∈ T N
2043
+ h . It follows that
2044
+ ����
2045
+
2046
+ T
2047
+ ⟨G(�g), σ⟩�g vωT (�g)
2048
+ ���� ≤ ∥G(�g)∥L∞(T,�g)∥σ∥L2(T,�g)∥v∥L2(T,�g)
2049
+ ≤ C∥G(�g)∥L∞(T)∥σ∥L2(T)∥v∥L2(T)
2050
+ ≤ C∥σ∥L2(T)∥v∥L2(T)
2051
+ = C∥gh − g∥L2(T)∥v∥L2(T).
2052
+ Summing over all T ∈ T N
2053
+ h
2054
+ completes the proof.
2055
+ Lemma 4.9. We have
2056
+ �����
2057
+ ˚
2058
+
2059
+ F
2060
+
2061
+ F
2062
+
2063
+ �II(�g)�F, σ|F
2064
+
2065
+ �g vωF(�g)
2066
+ ����� ≤ C max
2067
+ T
2068
+
2069
+ h−1
2070
+ T ∥gh − g∥W 1,∞(T)
2071
+
2072
+ ×
2073
+ ��
2074
+ T
2075
+ ∥gh − g∥2
2076
+ L2(T) + h2
2077
+ T |gh − g|2
2078
+ H1(T)
2079
+ �1/2 ��
2080
+ T
2081
+ ∥v∥2
2082
+ L2(T) + h2
2083
+ T |v|2
2084
+ H1(T)
2085
+ �1/2
2086
+ .
2087
+ Proof. Consider an interior (N − 1)-simplex F. By applying a Euclidean rotation and translation
2088
+ to the coordinates, we may assume without loss of generality that F lies in the plane xN = 0. In
2089
+ these coordinates, the second fundamental form associated with �g is given by
2090
+ IIij(�g) = −�g(n�g, ∇�g,eiej)
2091
+ = −�g(n�g, �Γk
2092
+ ijek)
2093
+ = −nℓ
2094
+ �g�gℓk�Γk
2095
+ ij,
2096
+ i, j = 1, 2, . . . , N − 1,
2097
+ where e1, e2, . . . , eN are the Euclidean coordinate basis vectors. Since n�g = �g−1n/
2098
+
2099
+ nT �g−1n and n
2100
+ points in the xN direction, we get
2101
+ IIij(�g) = −
2102
+ 1
2103
+
2104
+ nT �g−1n
2105
+ �ΓN
2106
+ ij .
2107
+ The jump in this quantity across F can be computed using the identity �ab� = �a�{b} + {a}�b�,
2108
+ where {·} denotes the average across F, giving
2109
+ −�IIij(�g)� =
2110
+
2111
+ 1
2112
+
2113
+ nT �g−1n
2114
+ � �
2115
+ �ΓN
2116
+ ij
2117
+
2118
+ +
2119
+
2120
+ 1
2121
+
2122
+ nT �g−1n
2123
+ � �
2124
+ �ΓN
2125
+ ij
2126
+
2127
+ .
2128
+ 25
2129
+
2130
+ In view of (36), we have
2131
+ �����
2132
+
2133
+ 1
2134
+
2135
+ nT �g−1n
2136
+ ������
2137
+ L∞(F )
2138
+ ≤ C ∥��g�∥L∞(F )
2139
+ ≤ C ∥�gh − g�∥L∞(F )
2140
+ ≤ C
2141
+
2142
+ ∥gh − g∥L∞(T1) + ∥gh − g∥L∞(T2)
2143
+
2144
+ ,
2145
+ where T1 and T2 are the two N-simplices that share the face F.
2146
+ Here, we used the fact that
2147
+ �g = g + t(gh − g) and g is smooth. Similarly, we have
2148
+ ���
2149
+
2150
+ �ΓN
2151
+ ij
2152
+ ����
2153
+ L∞(F ) ≤ C∥��g�∥W 1,∞(F )
2154
+ ≤ C∥�gh − g�∥W 1,∞(F )
2155
+ ≤ C
2156
+
2157
+ ∥gh − g∥W 1,∞(T1) + ∥gh − g∥W 1,∞(T2)
2158
+
2159
+ .
2160
+ (43)
2161
+ Thus,
2162
+ ∥�II(�g)�∥L∞(F ) ≤ C
2163
+
2164
+ ∥gh − g∥W 1,∞(T1) + ∥gh − g∥W 1,∞(T2)
2165
+
2166
+ .
2167
+ From this it follows easily that the same bound holds, possibly with a larger constant C, for the
2168
+ trace-reversed tensor II(�g) = II(�g) − H(�g)�g:
2169
+ ∥�II(�g)�∥L∞(F ) ≤ C
2170
+
2171
+ ∥gh − g∥W 1,∞(T1) + ∥gh − g∥W 1,∞(T2)
2172
+
2173
+ .
2174
+ It follows that
2175
+ ����
2176
+
2177
+ F
2178
+
2179
+ �II(�g)�F , σ|F
2180
+
2181
+ �g vωF (�g)
2182
+ ����
2183
+ ≤ ∥�II(�g)�∥L∞(F,�g)∥σ|F ∥L2(F,�g)∥v∥L2(F,�g)
2184
+ ≤ C∥�II(�g)�∥L∞(F )∥σ|F ∥L2(F )∥v∥L2(F )
2185
+ ≤ C
2186
+ � 2
2187
+
2188
+ i=1
2189
+ ∥gh − g∥W 1,∞(Ti)
2190
+ � �
2191
+ h−1
2192
+ T1 ∥σ∥2
2193
+ L2(T1) + hT1|σ|2
2194
+ H1(T1)
2195
+ �1/2 �
2196
+ h−1
2197
+ T1 ∥v∥2
2198
+ L2(T1) + hT1|v|2
2199
+ H1(T1)
2200
+ �1/2
2201
+ .
2202
+ By the shape-regularity of Th, we have C−1 ≤ hT1/hT2 ≤ C for some constant C independent of h
2203
+ and F, so
2204
+ �����
2205
+ ˚
2206
+
2207
+ F
2208
+
2209
+ F
2210
+
2211
+ �II(�g)�F, σ|F
2212
+
2213
+ �g vωF(�g)
2214
+ ����� ≤ C max
2215
+ T
2216
+
2217
+ h−1
2218
+ T ∥gh − g∥W 1,∞(T)
2219
+
2220
+ ×
2221
+ ��
2222
+ T
2223
+ ∥gh − g∥2
2224
+ L2(T) + h2
2225
+ T |gh − g|2
2226
+ H1(T)
2227
+ �1/2 ��
2228
+ T
2229
+ ∥v∥2
2230
+ L2(T) + h2
2231
+ T |v|2
2232
+ H1(T)
2233
+ �1/2
2234
+ .
2235
+ Remark 4.10. If gh is piecewise constant, then in (43) we have the sharper bound
2236
+ ∥�gh − g�∥W 1,∞(F ) = ∥�gh − g�∥L∞(F ) ≤ C
2237
+
2238
+ ∥gh − g∥L∞(T1) + ∥gh − g∥L∞(T2)
2239
+
2240
+ 26
2241
+
2242
+ because ∂gh
2243
+ ∂xi = 0 and
2244
+ ∂g
2245
+ ∂xi is continuous for each i. This implies that for piecewise constant gh, we
2246
+ can replace ∥gh − g∥W 1,∞(T) by ∥gh − g∥L∞(T) in Lemma 4.9, yielding
2247
+ �����
2248
+ ˚
2249
+
2250
+ F
2251
+
2252
+ F
2253
+
2254
+ �II(�g)�F, σ|F
2255
+
2256
+ �g vωF(�g)
2257
+ ����� ≤ C max
2258
+ T
2259
+
2260
+ h−1
2261
+ T ∥gh − g∥L∞(T)
2262
+
2263
+ ×
2264
+ ��
2265
+ T
2266
+ ∥gh − g∥2
2267
+ L2(T) + h2
2268
+ T |gh − g|2
2269
+ H1(T)
2270
+ �1/2 ��
2271
+ T
2272
+ ∥v∥2
2273
+ L2(T) + h2
2274
+ T |v|2
2275
+ H1(T)
2276
+ �1/2
2277
+ .
2278
+ Now we turn our attention toward the third integral in (42). In preparation for this, we will
2279
+ first use the shape-regularity assumption to show that the dihedral angles of every N-simplex in
2280
+ Th (measured in the Euclidean metric) are uniformly bounded above and below.
2281
+ Lemma 4.11. There exist constants θmin, θmax ∈ (0, π) such that for every h > 0 and every
2282
+ T ∈ T N
2283
+ h , the dihedral angles in T (measured in the Euclidean metric) all lie between θmin and θmax.
2284
+ Proof. This fact is proved in dimension N = 3 in [18, Lemma 3.6].
2285
+ We generalize their proof
2286
+ to dimension N ≥ 3 as follows. Given N + 1 points x1, x2, . . . , xN+1 in general position in RN,
2287
+ let T = [x1, x2, . . . , xN+1] denote the N-simplex with vertices x1, x2, . . . , xN+1.
2288
+ Consider two
2289
+ faces F1 = [x1, x3, x4, . . . , xN+1] and F2 = [x2, x3, x4, . . . , xN+1] that intersect along the (N − 2)-
2290
+ dimensional subsimplex S = [x3, x4, . . . , xN+1]. Throughout what follows, we work in the Euclidean
2291
+ metric. Let A be the orthogonal projection of x1 onto the (N−1)-dimensional hyperplane containing
2292
+ F2, and let B be the orthogonal projection of x1 onto the (N −2)-dimensional hyperplane containing
2293
+ S. Observe that both [x1, A] and [x1, B] are orthogonal to S, since S ⊂ F2. Thus, the triangle
2294
+ [x1, A, B] is orthogonal to S. This triangle is a right triangle with hypotenuse [x1, B], so the dihedral
2295
+ angle θST along S satisfies
2296
+ sin θST = |[x1, A]|
2297
+ |[x1, B]|,
2298
+ where | · | denotes the Euclidean volume (i.e. length in this case). Obviously, |[x1, B]| is bounded
2299
+ above by hT , the diameter of T. In addition, |[x1, A]| is bounded from below by 2 times ρT , the
2300
+ inradius of T. To see why, we generalize the argument in [18, Proposition 2.3], bearing in mind
2301
+ that our definition of ρT differs from theirs by a factor of 2. Consider the inscribed (N − 1)-sphere
2302
+ in T, whose center C lies at a distance ρT from F2. Let D be the point where this inscribed sphere
2303
+ touches F2, and let E be the point diametrically opposite to D on this sphere. The line segment
2304
+ [D, E] is orthogonal to F2, so the volume of the N-simplex T ′ = [E, x2, x3, x4, . . . , xN+1] satisfies
2305
+ |T ′| = 1
2306
+ N |[D, E]||F2| = 2ρT
2307
+ N |F2|.
2308
+ Since T ′ ⊂ T, we have
2309
+ |T ′| ≤ |T| = 1
2310
+ N |[x1, A]||F2|,
2311
+ so
2312
+ 2ρT ≤ |[x1, A]|.
2313
+ Thus,
2314
+ sin θST ≥ 2ρT
2315
+ hT
2316
+ .
2317
+ The result follows from this bound and the shape-regularity of Th.
2318
+ 27
2319
+
2320
+ Next we show that Lemma 4.11 remains valid when one measures angles with g rather than the
2321
+ Euclidean metric δ.
2322
+ Lemma 4.12. Upon reducing the value of h0 if necessary, there exist constants θmin,g, θmax,g ∈ (0, π)
2323
+ such that for every h ≤ h0, every T ∈ T N
2324
+ h , every (N − 2)-simplex S ⊂ ∂T, and every point p ∈ S,
2325
+ the dihedral angle in T at p (measured by g) lies between θmin,g and θmax,g.
2326
+ Proof. If there were no such lower bound θmin,g > 0, then there would exist a sequence of N-
2327
+ simplices T1 ∈ Th1, T2 ∈ Th2, . . . with faces F (1)
2328
+ 1
2329
+ , F (2)
2330
+ 1
2331
+ ⊂ T1, F (1)
2332
+ 2 , F (2)
2333
+ 2
2334
+ ⊂ T2, . . . and points
2335
+ p1 ∈ F (1)
2336
+ 1
2337
+ ∩ F (2)
2338
+ 1 , p2 ∈ F (1)
2339
+ 2
2340
+ ∩ F (2)
2341
+ 2
2342
+ , . . . such that
2343
+ ∠ g|Ti(pi)(F (1)
2344
+ i
2345
+ , F (2)
2346
+ i
2347
+ ) → 0
2348
+ as i → ∞, where ∠g(X, Y ) denotes the angle between X and Y as measured by g. Using the
2349
+ compactness of the Grassmannian, this implies that, after extracting a subsequence which we do
2350
+ not relabel,
2351
+ ∠δ(F (1)
2352
+ i
2353
+ , F (2)
2354
+ i
2355
+ ) → 0,
2356
+ where ∠δ(X, Y ) denotes the angle between X and Y as measured by the Euclidean metric δ. This
2357
+ contradicts the assumed positive lower bound on the Euclidean dihedral angles. The existence of
2358
+ an upper bound θmax,g < π is proved similarly.
2359
+ Now we are ready to estimate the third integral in (42).
2360
+ Lemma 4.13. We have
2361
+ �����
2362
+ ˚
2363
+
2364
+ S
2365
+
2366
+ S
2367
+ ⟨ΘS(�g) �g|S , σ|S⟩�g vωS(�g)
2368
+ �����
2369
+ ≤ C
2370
+
2371
+ max
2372
+ T
2373
+ h−2
2374
+ T ∥gh − g∥L∞(T)
2375
+ � ��
2376
+ T
2377
+ ∥gh − g∥2
2378
+ L2(T) + h2
2379
+ T |gh − g|2
2380
+ H1(T) + h4
2381
+ T |gh − g|2
2382
+ H2(T)
2383
+ �1/2
2384
+ ×
2385
+ ��
2386
+ T
2387
+ ∥v∥2
2388
+ L2(T) + h2
2389
+ T |v|2
2390
+ H1(T) + h4
2391
+ T |v|2
2392
+ H2(T)
2393
+ �1/2
2394
+ .
2395
+ Proof. Fix an interior (N − 2)-simplex S and an N-simplex T containing S. At any point p along
2396
+ S, we have
2397
+ cos θST(g) − cos θST (�g) = �g(n(1)
2398
+ �g , n(2)
2399
+ �g ) − g(n(1)
2400
+ g , n(2)
2401
+ g )
2402
+ = �g(n(1)
2403
+ �g
2404
+ − n(1)
2405
+ g , n(2)
2406
+ �g
2407
+ − n(2)
2408
+ g ) + �g(n(1)
2409
+ �g
2410
+ − n(1)
2411
+ g , n(2)
2412
+ g ) + �g(n(1)
2413
+ g , n(2)
2414
+ �g
2415
+ − n(2)
2416
+ g )
2417
+ + �g(n(1)
2418
+ g , n(2)
2419
+ g ) − g(n(1)
2420
+ g , n(2)
2421
+ g ),
2422
+ where n(1)
2423
+ g
2424
+ and n(2)
2425
+ g
2426
+ are suitably oriented unit normal vectors (with respect to g|T ) to the two faces
2427
+ of T containing S, and similarly for n(1)
2428
+ �g
2429
+ and n(2)
2430
+ �g . Using Lemma 4.6, we see that at the point p,
2431
+ | cos θST(�g) − cos θST(g)| ≤ C|�g − g| ≤ C|gh − g|
2432
+ for all h sufficiently small.
2433
+ Since there are constants θmin,g, θmax,g ∈ (0, π) such that θmin,g ≤
2434
+ θST(g) ≤ θmax,g, we get
2435
+ |θST (�g) − θST(g)| ≤ C|gh − g| ≤ C∥gh − g∥L∞(T).
2436
+ 28
2437
+
2438
+ Summing over T ⊃ S and noting that �
2439
+ T⊃S θST (g) = 2π, we get
2440
+ |ΘS(�g)| = |ΘS(�g) − ΘS(g)| ≤
2441
+
2442
+ T⊃S
2443
+ |θST(�g) − θST(g)| ≤ C
2444
+
2445
+ T⊃S
2446
+ ∥gh − g∥L∞(T).
2447
+ (44)
2448
+ Now we are almost ready to estimate the integral
2449
+
2450
+ S ⟨ΘS(�g) �g|S , σ|S⟩�g vωS(�g). We first note that
2451
+ ∥v∥2
2452
+ L2(S) ≤ C
2453
+
2454
+ h−2
2455
+ T ∥v∥2
2456
+ L2(T) + |v|2
2457
+ H1(T) + h2
2458
+ T |v|2
2459
+ H2(T)
2460
+
2461
+ ,
2462
+ which can be proved using a codimension-2 trace inequality and a scaling argument, or by applying
2463
+ the codimension-1 trace inequality (38) twice (to v rather than dv).
2464
+ If T1, T2, . . . , Tm are the
2465
+ N-simplices that share the (N − 2)-simplex S, then we have
2466
+ ����
2467
+
2468
+ S
2469
+ ⟨ΘS(�g) �g|S , σ|S⟩�g vωS(�g)
2470
+ ����
2471
+ ≤ C∥ΘS(�g)∥L∞(S,�g)∥σ|S∥L2(S,�g)∥v∥L2(S,�g)
2472
+ ≤ C∥ΘS(�g)∥L∞(S)∥σ|S∥L2(S)∥v∥L2(S)
2473
+ ≤ C
2474
+ � m
2475
+
2476
+ i=1
2477
+ ∥gh − g∥L∞(Ti)
2478
+ � �
2479
+ h−2
2480
+ T1 ∥σ∥2
2481
+ L2(T1) + |σ|2
2482
+ H1(T1) + h2
2483
+ T1|σ|2
2484
+ H2(T1)
2485
+ �1/2
2486
+ ×
2487
+
2488
+ h−2
2489
+ T1 ∥v∥2
2490
+ L2(T1) + |v|2
2491
+ H1(T1) + h2
2492
+ T1|v|2
2493
+ H2(T1)
2494
+ �1/2
2495
+ .
2496
+ The proof is completed by summing over all interior (N − 2)-simplices S and substituting σ =
2497
+ gh − g.
2498
+ Collecting our results, we can state a bound on the bilinear form ah(�g; ·, ·).
2499
+ Proposition 4.14. For every h ≤ h0, every t ∈ [0, 1], and every v ∈ V , we have (with σ = gh −g),
2500
+ |ah(�g; σ, v)| ≤ C
2501
+
2502
+ 1 + max
2503
+ T
2504
+ h−2
2505
+ T ∥gh − g∥L∞(T) + max
2506
+ T
2507
+ h−1
2508
+ T |gh − g|W 1,∞(T)
2509
+
2510
+ ×
2511
+ ��
2512
+ T
2513
+ ∥gh − g∥2
2514
+ L2(T) + h2
2515
+ T |gh − g|2
2516
+ H1(T) + h4
2517
+ T |gh − g|2
2518
+ H2(T)
2519
+ �1/2
2520
+ ×
2521
+ ��
2522
+ T
2523
+ ∥v∥2
2524
+ L2(T) + h2
2525
+ T |v|2
2526
+ H1(T) + h4
2527
+ T |v|2
2528
+ H2(T)
2529
+ �1/2
2530
+ .
2531
+ Proof. Combine Lemmas 4.8, 4.9, and 4.13.
2532
+ Upon combining Proposition 4.7 with Proposition 4.14, we see that
2533
+ ∥(Rω)dist(gh) − (Rω)(g)∥H−2(Ω) ≤ C
2534
+
2535
+ 1 + max
2536
+ T
2537
+ h−2
2538
+ T ∥gh − g∥L∞(T) + max
2539
+ T
2540
+ h−1
2541
+ T |gh − g|W 1,∞(T)
2542
+
2543
+ ×
2544
+
2545
+ ∥gh − g∥2
2546
+ L2(Ω) +
2547
+
2548
+ T
2549
+ h2
2550
+ T |gh − g|2
2551
+ H1(T) +
2552
+
2553
+ T
2554
+ h4
2555
+ T |gh − g|2
2556
+ H2(T)
2557
+ �1/2
2558
+ .
2559
+ 29
2560
+
2561
+ This completes the proof of Theorem 4.1. Corollary 4.3 then follows from (27) and the bounds
2562
+ ∥gh − g∥L2(Ω) ≤ |Ω|1/2−1/p∥gh − g∥Lp(Ω),
2563
+ ��
2564
+ T
2565
+ h2
2566
+ T |gh − g|2
2567
+ H1(T)
2568
+ �1/2
2569
+ ≤ |Ω|1/2−1/p
2570
+ ��
2571
+ T
2572
+ hp
2573
+ T |gh − g|p
2574
+ W 1,p(T)
2575
+ �1/p
2576
+ ,
2577
+ ��
2578
+ T
2579
+ h4
2580
+ T |gh − g|2
2581
+ H2(T)
2582
+ �1/2
2583
+ ≤ |Ω|1/2−1/p
2584
+ ��
2585
+ T
2586
+ h2p
2587
+ T |gh − g|p
2588
+ W 2,p(T)
2589
+ �1/p
2590
+ ,
2591
+ which hold for all p ∈ [2, ∞] (with the obvious modifications for p = ∞).
2592
+ Remark 4.15. Notice that the analysis above yields
2593
+ |bh(�g; σ, v)| = O(hr+1),
2594
+ (by Proposition 4.7),
2595
+ (45)
2596
+ �����
2597
+
2598
+ T
2599
+
2600
+ T
2601
+ ⟨G(�g), σ⟩�g vωT (�g)
2602
+ ����� = O(hr+1),
2603
+ (by Lemma 4.8),
2604
+ (46)
2605
+ �����
2606
+ ˚
2607
+
2608
+ F
2609
+
2610
+ F
2611
+
2612
+ �II(�g)�F , σ|F
2613
+
2614
+ �g vωF (�g)
2615
+ ����� =
2616
+
2617
+ O(h),
2618
+ if r = 0,
2619
+ O(h2r),
2620
+ if r ≥ 1,
2621
+ (by Remark 4.10),
2622
+ (by Lemma 4.9),
2623
+ (47)
2624
+ �����
2625
+ ˚
2626
+
2627
+ S
2628
+
2629
+ S
2630
+ ⟨ΘS(�g) �g|S , σ|S⟩�g vωS(�g)
2631
+ ����� = O(h2r),
2632
+ (by Lemma 4.13)
2633
+ (48)
2634
+ for any optimal-order interpolant gh of g having degree r ≥ 0. Bearing in mind that (46-48) vanish
2635
+ when N = 2, we see that the above estimates lead to an optimal error estimate ∥(Rω)dist(gh) −
2636
+ (Rω)(g)∥H−2(Ω) = O(hr+1) in all cases except when N ≥ 3 and r = 0, where we obtain ∥(Rω)dist(gh)−
2637
+ (Rω)(g)∥H−2(Ω) = O(1) because of (48). Numerical experiments suggest that these analytical re-
2638
+ sults are sharp for a general optimal-order interpolant, whereas for the canonical interpolant the
2639
+ estimate (48) improves to O(h2(r+1)), yielding ∥(Rω)dist(gh) − (Rω)(g)∥H−2(Ω) = O(h) when r = 0;
2640
+ cf. Figure 2.
2641
+ 5
2642
+ Numerical examples
2643
+ In this section we present numerical experiments in dimension N = 2, 3 to illustrate the predicted
2644
+ convergence rates. The examples were performed in the open source finite element library NGSolve1
2645
+ [24, 25], where the Regge finite elements are available for arbitrary polynomial order. We construct
2646
+ an optimal-order interpolant gh of a given metric tensor g as follows. On each element T, the local
2647
+ L2 best-approximation ¯gh|T of g|T is computed. Then the tangential-tangential degrees of freedom
2648
+ shared by two or more neighboring elements are averaged to obtain a globally tangential-tangential
2649
+ continuous interpolant gh.
2650
+ We verify in Appendix A that this interpolant is an optimal-order
2651
+ interpolant in the sense of Remark 4.4 on shape-regular, quasi-uniform triangulations.
2652
+ To compute the H−2(Ω)-norm of the error f := (Rω)dist(gh) − (Rω)(g) we make use of the fact
2653
+ that ∥f∥H−2(Ω) is equivalent to ∥u∥H2(Ω), where u ∈ H2
2654
+ 0(Ω) solves the biharmonic equation ∆2u = f.
2655
+ This equation will be solved numerically using the (Euclidean) Hellan–Herrmann–Johnson method.
2656
+ To prevent the discretization error from spoiling the real error, we use for uh two polynomial orders
2657
+ more than for gh.
2658
+ 1www.ngsolve.org
2659
+ 30
2660
+
2661
+ We consider in dimension N = 2 the numerical example proposed in [16], where on the square
2662
+ Ω = (−1, 1)2 the smooth Riemannian metric tensor
2663
+ g(x, y) :=
2664
+
2665
+ 1 + (∂f
2666
+ ∂x)2
2667
+ ∂f
2668
+ ∂x
2669
+ ∂f
2670
+ ∂y
2671
+ ∂f
2672
+ ∂x
2673
+ ∂f
2674
+ ∂y
2675
+ 1 + (∂f
2676
+ ∂y )2
2677
+
2678
+ with f(x, y) := 1
2679
+ 2(x2 + y2) − 1
2680
+ 12(x4 + y4) is defined. This metric corresponds to the surface induced
2681
+ by the embedding
2682
+
2683
+ x, y
2684
+
2685
+ �→
2686
+
2687
+ x, y, f(x, y)
2688
+
2689
+ , and its exact scalar curvature is given by
2690
+ R(g)(x, y) =
2691
+ 162(1 − x2)(1 − y2)
2692
+ (9 + x2(x2 − 3)2 + y2(y2 − 3)2)2 .
2693
+ For a three-dimensional example we consider the cube Ω = (−1, 1)3 and the Riemannian metric
2694
+ tensor induced by the embedding
2695
+
2696
+ x, y, z
2697
+
2698
+ �→
2699
+
2700
+ x, y, z, f(x, y, z)
2701
+
2702
+ , where f(x, y, z) := 1
2703
+ 2(x2 + y2 +
2704
+ z2) − 1
2705
+ 12(x4 + y4 + z4). The scalar curvature is
2706
+ R(g)(x, y, z) = 18
2707
+
2708
+ (1 − x2)(1 − y2)(9 + q(z)) + (1 − y2)(1 − z2)(9 + q(x)) + (1 − z2)(1 − x2)(9 + q(y))
2709
+
2710
+ (9 + q(x) + q(y) + q(z))2
2711
+ ,
2712
+ where q(x) = x2(x2 − 3)2.
2713
+ We start with a structured mesh consisting of 2·22k triangles and 6·23k tetrahedra, respectively,
2714
+ in two and three dimensions with ˜h = maxT hT =
2715
+
2716
+ N 21−k (and minimal edge length 21−k)
2717
+ for k = 0, 1, . . . . To avoid possible superconvergence due to mesh symmetries, we perturb each
2718
+ component of the inner mesh vertices by a random number drawn from a uniform distribution in
2719
+ the range [−˜h 2−(2N+1)/2, ˜h 2−(2N+1)/2]. As depicted in Figure 1 (left) and listed in Table 1, linear
2720
+ convergence is observed when N = 2 and gh has polynomial degree r = 0. This is consistent with
2721
+ Theorem 4.1(i). For r = 1 and r = 2, higher convergence rates are obtained as expected.
2722
+ In the three-dimensional case, the same convergence rates as for N = 2 are obtained, cf. Figure 1
2723
+ (right) and Table 2. This indicates that Theorem 4.1(ii) is sharp for r ≥ 1. For r = 0 we observe
2724
+ numerically linear convergence, which is better than predicted by Theorem 4.1(ii). However, further
2725
+ investigation suggests that the observed linear convergence for r = 0 is pre-asymptotic. Indeed, to
2726
+ test if (48) is sharp, we compute the H−2(Ω)-norm of the linear functional
2727
+ v �→
2728
+ � 1
2729
+ 0
2730
+ ˚
2731
+
2732
+ S
2733
+
2734
+ S
2735
+ ⟨ΘS(�g(t)) �g(t)|S , σ|S⟩�g(t) vωS(�g(t)) dt,
2736
+ (49)
2737
+ where we approximate the parameter integral by a Gauss quadrature of order seven. As depicted
2738
+ in Figure 2, the norm of this functional for the optimal-order interpolant gh with r = 0 stagnates at
2739
+ about 4·10−4, which is below the overall error of 4.296·10−3 for the finest grid; cf. Table 2. There-
2740
+ fore, the lack of convergence predicted by Theorem 4.1(ii) is not yet visible in Figure 1. For r = 1, 2
2741
+ the proven rate of O(h2r) for (49) (see (48)) is clearly obtained. Interestingly, using the canonical
2742
+ interpolant appears to increase the convergence rate of (49) to O(h2(r+1)) (i.e. an increase of two
2743
+ orders), as observed in Figure 2. Thus, it appears that the canonical interpolant achieves conver-
2744
+ gence in the lowest-order case. We intend to study this superconvergence phenomenon exhibited
2745
+ by the canonical interpolant in future work.
2746
+ Acknowledgments
2747
+ We thank Yasha Berchenko-Kogan for many helpful discussions, especially about the mean curva-
2748
+ ture term in Definition 3.1. We also thank Snorre Christiansen for pointing out the link with the
2749
+ 31
2750
+
2751
+ 101
2752
+ 102
2753
+ 103
2754
+ 104
2755
+ 105
2756
+ 106
2757
+ 10−8
2758
+ 10−6
2759
+ 10−4
2760
+ 10−2
2761
+ 100
2762
+ ndof
2763
+ error
2764
+ r = 0
2765
+ r = 1
2766
+ r = 2
2767
+ O(h)
2768
+ O(h2)
2769
+ O(h3)
2770
+ 101
2771
+ 102
2772
+ 103
2773
+ 104
2774
+ 105
2775
+ 106
2776
+ 107
2777
+ 108
2778
+ 10−6
2779
+ 10−5
2780
+ 10−4
2781
+ 10−3
2782
+ 10−2
2783
+ 10−1
2784
+ 100
2785
+ ndof
2786
+ error
2787
+ r = 0
2788
+ r = 1
2789
+ r = 2
2790
+ O(h)
2791
+ O(h2)
2792
+ O(h3)
2793
+ Figure 1: Convergence of the distributional scalar curvature in the H−2(Ω)-norm for N = 2 (left)
2794
+ and N = 3 (right) with respect to the number of degrees of freedom (ndof) of gh for r = 0, 1, 2.
2795
+ 101
2796
+ 102
2797
+ 103
2798
+ 104
2799
+ 105
2800
+ 106
2801
+ 107
2802
+ 10−12
2803
+ 10−10
2804
+ 10−8
2805
+ 10−6
2806
+ 10−4
2807
+ 10−2
2808
+ ndof
2809
+ H−2(Ω)-norm of (49)
2810
+ r = 0
2811
+ r = 1
2812
+ r = 2
2813
+ r = 0 c.i.
2814
+ r = 1 c.i.
2815
+ r = 2 c.i.
2816
+ O(h2)
2817
+ O(h4)
2818
+ O(h6)
2819
+ Figure 2: Convergence of (49) in the H−2(Ω)-norm with respect to number of degrees of freedom
2820
+ (ndof) for an optimal-order interpolant and the canonical interpolant (c.i.)
2821
+ for r = 0, 1, 2 in
2822
+ dimension N = 3.
2823
+ r = 0
2824
+ r = 1
2825
+ r = 2
2826
+ h
2827
+ Error
2828
+ Order
2829
+ Error
2830
+ Order
2831
+ Error
2832
+ Order
2833
+ 2.828 · 10−0
2834
+ 1.534 · 10−0
2835
+ 8.584 · 10−1
2836
+ 4.609 · 10−1
2837
+ 2.417 · 10−1
2838
+ 1.251 · 10−1
2839
+ 6.260 · 10−2
2840
+ 3.198 · 10−2
2841
+ 2.237 · 10−1
2842
+ 1.945 · 10−1
2843
+ 0.23
2844
+ 6.220 · 10−2
2845
+ 1.96
2846
+ 2.336 · 10−2
2847
+ 1.57
2848
+ 9.434 · 10−3
2849
+ 1.41
2850
+ 4.457 · 10−3
2851
+ 1.14
2852
+ 2.181 · 10−3
2853
+ 1.03
2854
+ 1.067 · 10−3
2855
+ 1.06
2856
+ 8.613 · 10−2
2857
+ 8.448 · 10−2
2858
+ 0.03
2859
+ 4.565 · 10−2
2860
+ 1.06
2861
+ 1.335 · 10−2
2862
+ 1.98
2863
+ 3.689 · 10−3
2864
+ 1.99
2865
+ 9.205 · 10−4
2866
+ 2.11
2867
+ 2.280 · 10−4
2868
+ 2.02
2869
+ 5.777 · 10−5
2870
+ 2.04
2871
+ 2.720 · 10−2
2872
+ 1.364 · 10−2
2873
+ 1.13
2874
+ 2.213 · 10−3
2875
+ 3.13
2876
+ 3.615 · 10−4
2877
+ 2.91
2878
+ 4.189 · 10−5
2879
+ 3.34
2880
+ 5.504 · 10−6
2881
+ 3.08
2882
+ 7.028 · 10−7
2883
+ 2.97
2884
+ 8.784 · 10−8
2885
+ 3.1
2886
+ Table 1: Same as Figure 1 (left), but in tabular form.
2887
+ 32
2888
+
2889
+ r = 0
2890
+ r = 1
2891
+ r = 2
2892
+ h
2893
+ Error
2894
+ Order
2895
+ Error
2896
+ Order
2897
+ Error
2898
+ Order
2899
+ 3.464 · 10−0
2900
+ 1.850 · 10−0
2901
+ 9.709 · 10−1
2902
+ 4.999 · 10−1
2903
+ 2.753 · 10−1
2904
+ 1.358 · 10−1
2905
+ 6.878 · 10−2
2906
+ 7.869 · 10−2
2907
+ 3.215 · 10−1
2908
+ -2.24
2909
+ 1.132 · 10−1
2910
+ 1.62
2911
+ 4.152 · 10−2
2912
+ 1.51
2913
+ 1.838 · 10−2
2914
+ 1.37
2915
+ 8.733 · 10−3
2916
+ 1.05
2917
+ 4.296 · 10−3
2918
+ 1.04
2919
+ 1.359 · 10−1
2920
+ 6.613 · 10−2
2921
+ 1.15
2922
+ 2.912 · 10−2
2923
+ 1.27
2924
+ 8.633 · 10−3
2925
+ 1.83
2926
+ 2.391 · 10−3
2927
+ 2.15
2928
+ 6.194 · 10−4
2929
+ 1.91
2930
+ 1.579 · 10−4
2931
+ 2.01
2932
+ 1.871 · 10−2
2933
+ 4.133 · 10−2
2934
+ -1.26
2935
+ 5.286 · 10−3
2936
+ 3.19
2937
+ 7.342 · 10−4
2938
+ 2.97
2939
+ 9.753 · 10−5
2940
+ 3.38
2941
+ 1.261 · 10−5
2942
+ 2.89
2943
+ 1.604 · 10−6
2944
+ 3.03
2945
+ Table 2: Same as Figure 1 (right), but in tabular form.
2946
+ Israel formalism mentioned in Remark 3.9. EG was supported by NSF grant DMS-2012427. MN
2947
+ acknowledges support by the Austrian Science Fund (FWF) project F 65.
2948
+ A
2949
+ Optimal-order interpolation via averaging
2950
+ Below we verify that the interpolant described in Section 5 is an optimal-order interpolant in the
2951
+ sense of Remark 4.4, assuming that {Th}h>0 is shape-regular and quasi-uniform. Recall that quasi-
2952
+ uniformity means that maxT∈T N
2953
+ h h/hT is bounded above by a constant independent of h. In what
2954
+ follows, the letter C may depend on this constant as well as on the parameters N, hT /ρT , r, s, and
2955
+ t appearing below.
2956
+ Let ℓ(1), ℓ(2), . . . , ℓ(M) denote the canonical degrees of freedom for the Regge finite element space
2957
+ of degree r ≥ 0 on Th [21, Equation (2.4b)]. Each linear functional ℓ(i) is associated with a simplex
2958
+ D ∈ T k
2959
+ h of dimension k ≥ 1 in the following sense: ℓ(i) sends a symmetric (0, 2)-tensor field g to
2960
+ the integral of g|D against a (symmetric tensor-valued) polynomial of degree ≤ r − k + 1 over D.
2961
+ We enumerate these degrees of freedom with a local numbering system as follows.
2962
+ On a
2963
+ given N-simplex T ∈ T N
2964
+ h , the degrees of freedom associated with subsimplices of T are denoted
2965
+ ℓT
2966
+ 1 , ℓT
2967
+ 2 , . . . , ℓT
2968
+ MT . If T, T ′ ∈ T N
2969
+ h
2970
+ are two N-simplices with nonempty intersection, then it may happen
2971
+ that ℓT
2972
+ i and ℓT ′
2973
+ j
2974
+ coincide for some and i and j. We let S(i, T) denote the set of all pairs (j, T ′) for
2975
+ which ℓT
2976
+ i and ℓT ′
2977
+ j
2978
+ coincide.
2979
+ With the above local numbering system, let ψT
2980
+ 1 , ψT
2981
+ 2 , . . . , ψT
2982
+ MT denote the basis for the degree-r
2983
+ Regge finite element space that is dual to the above degrees of freedom. That is,
2984
+ ℓT
2985
+ i (ψT ′
2986
+ j ) =
2987
+
2988
+ 1,
2989
+ if (j, T ′) ∈ S(i, T),
2990
+ 0,
2991
+ otherwise.
2992
+ Let us assume that the degrees of freedom and basis functions above are first defined on a reference
2993
+ simplex and then transported to T via an affine transformation. A scaling argument shows that [21,
2994
+ Lemma 2.11]
2995
+ ∥ψT
2996
+ i ∥Lp(T) ≤ ChN/p−2
2997
+ T
2998
+ (50)
2999
+ and
3000
+ |ℓT
3001
+ i (g)| ≤ Ch−N/p+2
3002
+ T
3003
+ ∥g∥Lp(T)
3004
+ (51)
3005
+ for all g in the domain of ℓT
3006
+ i . Note that the −2 and the +2 appearing in the exponents above
3007
+ arise because of the way that pullbacks of (0, 2)-tensor fields behave under affine transformations;
3008
+ see [21, Lemma 2.11].
3009
+ 33
3010
+
3011
+ Let g be a symmetric (0, 2)-tensor field possessing W s,p(Ω)-regularity for every p ∈ [1, ∞] and
3012
+ every s > (N − 1)/p. The canonical interpolation operator Jh onto the Regge finite element space
3013
+ is defined elementwise by
3014
+ Jhg|T = J T
3015
+ h (g|T ) =
3016
+ MT
3017
+
3018
+ i=1
3019
+ ℓT
3020
+ i (g)ψT
3021
+ i .
3022
+ Let ¯gh denote the elementwise L2-projection of g onto the space of discontinuous piecewise
3023
+ polynomial symmetric (0, 2)-tensor fields of degree at most r. Since Jh is a projector, we have
3024
+ ¯gh|T = J T
3025
+ h ( ¯gh|T ) =
3026
+ MT
3027
+
3028
+ i=1
3029
+ ℓT
3030
+ i (¯gh)ψT
3031
+ i .
3032
+ The interpolant discussed in Section 5 is defined by
3033
+ gh|T =
3034
+ MT
3035
+
3036
+ i=1
3037
+
3038
+
3039
+ 1
3040
+ |S(i, T)|
3041
+
3042
+ (j,T ′)∈S(i,T)
3043
+ ℓT ′
3044
+ j (¯gh)
3045
+
3046
+  ψT
3047
+ i ,
3048
+ where |S(i, T)| denotes the cardinality of S(i, T).
3049
+ To analyze the error gh − g, let p ∈ [1, ∞], s ∈ ((N − 1)/p, r + 1], and t ∈ [0, s]. We have
3050
+ |gh − g|W t,p(T) ≤ |gh − Jhg|W t,p(T) + |Jhg − g|W t,p(T).
3051
+ The second term satisfies [21, Theorem 2.5]
3052
+ |Jhg − g|W t,p(T) ≤ Chs−t
3053
+ T
3054
+ |g|W s,p(T).
3055
+ (52)
3056
+ To bound the first term, we use the fact that
3057
+ ℓT
3058
+ i (g) =
3059
+ 1
3060
+ |S(i, T)|
3061
+
3062
+ (j,T ′)∈S(i,T)
3063
+ ℓT ′
3064
+ j (g)
3065
+ to write
3066
+ (gh − Jhg)|T =
3067
+ MT
3068
+
3069
+ i=1
3070
+ 1
3071
+ |S(i, T)|
3072
+
3073
+ (j,T ′)∈S(i,T)
3074
+ ℓT ′
3075
+ j (¯gh − g)ψT
3076
+ i .
3077
+ Using an inverse estimate, (50), (51), and a standard error estimate [14, Proposition 1.135] for the
3078
+ elementwise L2-projector, we obtain
3079
+ |gh − Jhg|W t,p(T) ≤ Ch−t
3080
+ T ∥gh − Jhg∥Lp(T)
3081
+ ≤ Ch−t
3082
+ T
3083
+
3084
+ T ′:T ′∩T̸=∅
3085
+ h−N/p+2
3086
+ T ′
3087
+ ∥¯gh − g∥Lp(T ′)hN/p−2
3088
+ T
3089
+ ≤ Ch−t
3090
+ T
3091
+
3092
+ T ′:T ′∩T̸=∅
3093
+ ∥¯gh − g∥Lp(T ′)
3094
+ ≤ Ch−t
3095
+ T
3096
+
3097
+ T ′:T ′∩T̸=∅
3098
+ hs
3099
+ T ′|g|W s,p(T ′)
3100
+ ≤ Chs−t
3101
+ T
3102
+
3103
+ T ′:T ′∩T̸=∅
3104
+ |g|W s,p(T ′).
3105
+ (53)
3106
+ Here, we have repeatedly used the fact that the ratio hT /hT ′ is bounded uniformly above and below
3107
+ by positive constants. Combining (52) and (53) shows that the error gh − g satisfies (28).
3108
+ 34
3109
+
3110
+ References
3111
+ [1]
3112
+ D. N. Arnold and F. Brezzi. “Mixed and nonconforming finite element methods: implementa-
3113
+ tion, postprocessing and error estimates”. In: ESAIM: Mathematical Modelling and Numerical
3114
+ Analysis 19.1 (1985), pp. 7–32.
3115
+ [2]
3116
+ D. N. Arnold and S. W. Walker. “The Hellan–Herrmann–Johnson method with curved ele-
3117
+ ments”. In: SIAM Journal on Numerical Analysis 58.5 (2020), pp. 2829–2855.
3118
+ [3]
3119
+ I. Babuˇska, J Osborn, and J. Pitk¨aranta. “Analysis of mixed methods using mesh dependent
3120
+ norms”. In: Mathematics of Computation 35.152 (1980), pp. 1039–1062.
3121
+ [4]
3122
+ Y. Berchenko-Kogan and E. S. Gawlik. “Finite element approximation of the Levi-Civita con-
3123
+ nection and its curvature in two dimensions”. In: Foundations of Computational Mathematics,
3124
+ to appear (2022).
3125
+ [5]
3126
+ D. Braess, R. H. Hoppe, and C. Linsenmann. “A two-energies principle for the biharmonic
3127
+ equation and an a posteriori error estimator for an interior penalty discontinuous Galerkin
3128
+ approximation”. In: ESAIM: Mathematical Modelling and Numerical Analysis 52.6 (2018),
3129
+ pp. 2479–2504.
3130
+ [6]
3131
+ D. Braess, A. S. Pechstein, and J. Sch¨oberl. “An equilibration-based a posteriori error bound
3132
+ for the biharmonic equation and two finite element methods”. In: IMA Journal of Numerical
3133
+ Analysis (2019).
3134
+ [7]
3135
+ F. Brezzi and P.-A. Raviart. “Mixed finite element methods for 4th order elliptic equations”.
3136
+ In: Topics in Numerical Analysis, III. London: Academic Press, 1977, pp. 33–56.
3137
+ [8]
3138
+ J. Cheeger, W. M¨uller, and R. Schrader. “On the curvature of piecewise flat spaces”. In:
3139
+ Communications in Mathematical Physics 92.3 (1984), pp. 405–454.
3140
+ [9]
3141
+ L. Chen, J. Hu, and X. Huang. “Multigrid methods for Hellan–Herrmann–Johnson mixed
3142
+ method of Kirchhoff plate bending problems”. In: Journal of Scientific Computing 76.2 (2018),
3143
+ pp. 673–696.
3144
+ [10]
3145
+ B. Chow, P. Lu, and L. Ni. Hamilton’s Ricci Flow. Vol. 77. American Mathematical Society,
3146
+ 2006.
3147
+ [11]
3148
+ S. H. Christiansen. “A characterization of second-order differential operators on finite element
3149
+ spaces”. In: Mathematical Models and Methods in Applied Sciences 14.12 (2004), pp. 1881–
3150
+ 1892.
3151
+ [12]
3152
+ S. H. Christiansen. “On the linearization of Regge calculus”. In: Numerische Mathematik
3153
+ 119.4 (2011), pp. 613–640.
3154
+ [13]
3155
+ S. H. Christiansen. “Exact formulas for the approximation of connections and curvature”. In:
3156
+ arXiv preprint arXiv:1307.3376 (2013).
3157
+ [14]
3158
+ A. Ern and J.-L. Guermond. Theory and Practice of Finite Elements. Vol. 159. Springer
3159
+ Science & Business Media, 2004.
3160
+ [15]
3161
+ A. E. Fischer and J. E. Marsden. “Deformations of the scalar curvature”. In: Duke Mathe-
3162
+ matical Journal 42.3 (1975), pp. 519–547.
3163
+ [16]
3164
+ E. S. Gawlik. “High-order approximation of Gaussian curvature with Regge finite elements”.
3165
+ In: SIAM Journal on Numerical Analysis 58.3 (2020), pp. 1801–1821.
3166
+ [17]
3167
+ G. W. Gibbons and S. W. Hawking. “Action integrals and partition functions in quantum
3168
+ gravity”. In: Euclidean Quantum Gravity. World Scientific, 1993, pp. 233–237.
3169
+ 35
3170
+
3171
+ [18]
3172
+ S. Gong, J. Guzm´an, and M. Neilan. “A note on the shape regularity of Worsey-Farin splits”.
3173
+ In: arXiv preprint arXiv:2205.05059 (2022).
3174
+ [19]
3175
+ J. Gopalakrishnan, M. Neunteufel, J. Sch¨oberl, and M. Wardetzky. “Analysis of curvature
3176
+ approximations via covariant curl and incompatibility for Regge metrics”. In: arXiv preprint
3177
+ arXiv:2206.09343 (2022).
3178
+ [20]
3179
+ W. Israel. “Singular hypersurfaces and thin shells in general relativity”. In: Il Nuovo Cimento
3180
+ B (1965-1970) 44.1 (1966), pp. 1–14.
3181
+ [21]
3182
+ L. Li. “Regge finite elements with applications in solid mechanics and relativity”. PhD thesis.
3183
+ University of Minnesota, May 2018.
3184
+ [22]
3185
+ A. S. Pechstein and J. Sch¨oberl. “The TDNNS method for Reissner–Mindlin plates”. In:
3186
+ Numerische Mathematik 137.3 (2017), pp. 713–740.
3187
+ [23]
3188
+ T. Regge. “General relativity without coordinates”. In: Il Nuovo Cimento (1955-1965) 19.3
3189
+ (1961), pp. 558–571.
3190
+ [24]
3191
+ J. Sch¨oberl. “C++ 11 implementation of finite elements in NGSolve”. In: Institute for Analysis
3192
+ and Scientific Computing, Vienna University of Technology (2014).
3193
+ [25]
3194
+ J. Sch¨oberl. “NETGEN An advancing front 2D/3D-mesh generator based on abstract rules”.
3195
+ In: Computing and Visualization in Science 1.1 (1997), pp. 41–52.
3196
+ [26]
3197
+ R. S. Strichartz. “Defining Curvature as a Measure via Gauss–Bonnet on Certain Singular
3198
+ Surfaces”. In: The Journal of Geometric Analysis 30.1 (2020), pp. 153–160.
3199
+ [27]
3200
+ S. W. Walker. “Poincar´e inequality for a mesh-dependent 2-norm on piecewise linear surfaces
3201
+ with boundary”. In: Computational Methods in Applied Mathematics 22.1 (2022), pp. 227–
3202
+ 243.
3203
+ [28]
3204
+ J. W. York Jr. “Role of conformal three-geometry in the dynamics of gravitation”. In: Physical
3205
+ Review Letters 28.16 (1972), pp. 1082–1085.
3206
+ 36
3207
+
7NA0T4oBgHgl3EQfOP_L/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
7NE1T4oBgHgl3EQfBgKg/content/tmp_files/2301.02853v1.pdf.txt ADDED
@@ -0,0 +1,1502 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Using a Penalized Likelihood to Detect Mortality
2
+ Deceleration
3
+ Silvio C. Patricio*1 and Trifon I. Missov1
4
+ 1The Interdisciplinary Centre on Population Dynamics, University of Southern Denmark
5
+ Abstract
6
+ We propose a novel method to detect deceleration in mortality patterns. For a gamma-
7
+ Gompertz frailty model, we suggest maximizing a penalized likelihood in a Bayesian setting
8
+ as an alternative to traditional likelihood inference and hypothesis testing. We compare the
9
+ performance of the two methods on simulated and real mortality data.
10
+ Keywords: Gompertz model; gamma-Gompertz model, mortality deceleration; penalized
11
+ likelihood function; maximum a posteriori probability.
12
+ 1
13
+ Introduction
14
+ Human death-rate patterns are astoundingly log-linear over a wide range of adult ages. The
15
+ Gompertz distribution (Gompertz, 1825) with an exponentially increasing hazard function cap-
16
+ tures this accurately. The theory of unobserved heterogeneity and the associated frailty model
17
+ (Vaupel et al., 1979) predicts a downward deviation at the oldest ages, to which only the most
18
+ robust individuals in the population survive. Detecting such a deceleration in real data is not
19
+ always successful (Gavrilova and Gavrilov, 2015; Newman, 2018), even though the vast major-
20
+ ity of studies indicate that death rates at older ages increase at lower rates and can even level
21
+ off (Curtsinger et al., 1992; Fukui et al., 1993, 1996; Carey et al., 1995; Khazaeli et al., 1998;
22
+ Gampe, 2010, 2021; Rootz´en and Zholud, 2017; Alvarez et al., 2021; Camarda, 2022; Belzile
23
+ et al., 2022). In a frailty model setting, testing for mortality deceleration is equivalent to testing
24
+ whether the non-negative frailty parameter is strictly positive.
25
+ Formally, denote by X a non-negative continuous random variable that describes individual
26
+ human lifespans (complete or after a given adult age). If X ∼ Gompertz(a, b), where a is the
27
+ mortality level at the initial age and b is the rate of aging, the associated hazard function (force
28
+ of mortality) at time x
29
+ µ(x) = lim
30
+ ε↓0 P(x ≤ X < x + ε|X ≥ x).
31
+ *silca@sam.sdu.dk
32
+ 1
33
+ arXiv:2301.02853v1 [stat.ME] 7 Jan 2023
34
+
35
+ is µ(x) = aebx . Vaupel et al. (1979) introduce a positive continuous random variable Z, called
36
+ frailty, that acts multiplicatively on µ(x) and captures one’s unobserved susceptibility to death.
37
+ The force of mortality for an individual with frailty Z = z is
38
+ µ(x | Z = z) = z µ(x) .
39
+ For a gamma-distributed frailty with E(Z) = 1 and VAR(Z) = σ2, the force of mortality of the
40
+ population, i.e., the marginal hazard is
41
+ ¯µ(x) =
42
+ aebx
43
+ 1 + σ2 a
44
+ b (ebx − 1)
45
+ (1)
46
+ (see Vaupel et al. (1979) and Vaupel and Missov (2014) for all technicalities). Note that the
47
+ variance of Z is often denoted by γ (e.g., in Vaupel and Missov, 2014) because it is also equal
48
+ to the squared coefficient of variation of the distribution of frailty among survivors to any age x.
49
+ If σ2 > 0, the force of mortality for the population ¯µ(x) starts deviating from the exponential
50
+ pattern with increasing x and reaches an asymptote b/σ2. When σ2 = 0, i.e., when there is
51
+ no unobserved heterogeneity, the model for the population reduces to the (Gompertz) model for
52
+ individuals with an exponentially increasing hazard function µ(x) = aebx.
53
+ Testing for mortality deceleration in this setting reduces to statistical testing whether σ2 = 0
54
+ given the alternative σ2 > 0. The frailty parameter σ2 can take a value on the boundary of
55
+ the parameter space (σ2 = 0). This violates the standard underlying assumptions about the
56
+ asymptotic properties of likelihood-based inference and statistical hypothesis testing (see, for
57
+ example, B¨ohnstedt and Gampe, 2019). As a result, the asymptotic distribution of the maximum
58
+ likelihood estimator may not be Gaussian.
59
+ In this paper, we treat the problem of identifying whether σ2 > 0 or σ2 = 0 as a model
60
+ misspecification problem, i.e., we consider the gamma-Gompertz model when it is the Gompertz
61
+ model that actually holds. In this setting, we suggest subtracting a penalty from the log-likelihood
62
+ function. This penalty will be responsible for shrinking σ2 to zero when there is no heterogeneity,
63
+ as well as for adding a small bias to the Maximum Likelihood Estimator (MLE) when the effect
64
+ of unobserved heterogeneity is non-negligible. We carry out Monte Carlo simulation experiments
65
+ to evaluate the accuracy and precision of the estimates obtained by maximizing the likelihood
66
+ function, on the one hand, and the penalized likelihood function, on the other.
67
+ In Section 2, we formulate the model misspecification problem and introduce inference
68
+ methodology taking advantage of the maximum a posteriori probability (MAP). Then we carry
69
+ out a Monte Carlo simulation study to compare the performance of maximizing a standard and a
70
+ penalized likelihood. In Section 3, we compare the latter on mortality data for France, Japan and
71
+ the USA. Section 4 discusses the advantages and drawbacks of applying our method to detect
72
+ heterogeneity (deceleration) in mortality patterns.
73
+ 2
74
+ Methodology
75
+ Suppose X is a random sample with a cumulative distribution function G(x), and we fit the
76
+ incorrect family of densities {f(x; θ), θ ∈ Θ} to the data using MLE. The misspecified log-
77
+ likelihood is
78
+ 2
79
+
80
+ ℓ(θ; X) =
81
+ n
82
+
83
+ i=1
84
+ log f(Xi; θ).
85
+ Applying the law of large numbers, we get in the limit what the misspecified log-likelihood
86
+ function ℓ(θ; X) looks like for each θ ∈ Θ (see the right-hand side below):
87
+ 1
88
+ nℓ(θ; X) = 1
89
+ n
90
+ n
91
+
92
+ i=1
93
+ log f(Xi; θ)
94
+ a.s
95
+ −→ Eg (log f(X1; θ)) =
96
+
97
+ Im(X1)
98
+ log f(x; θ)dG(x) .
99
+ (2)
100
+ Assume there is no heterogeneity in the data (σ2 = 0), and we fit a gamma-Gompertz model.
101
+ In other words, we observe an exponential death-rate increase in the data, but we estimate a
102
+ model that implies a downward deviation from the exponential at the oldest ages. As shown in
103
+ (2), we will estimate σ2 close to but never equal to zero.
104
+ In this model setting, the standard technique is to estimate both the Gompertz and the gamma-
105
+ Gompertz models and compare their goodness of fit. However, minor changes in the data can
106
+ result in different models being selected, which can reduce prediction accuracy and lead to mis-
107
+ interpretations about the mortality deceleration and the mortality plateau. B¨ohnstedt and Gampe
108
+ (2019) derive the asymptotic distribution of the likelihood ratio test statistic to detect heterogene-
109
+ ity. Here, we would like to suggest an alternative that does not involve hypothesis testing. Using
110
+ the latter has been widely discussed and rethought in the Statistics community (Berk et al., 2010;
111
+ Head et al., 2015; Vidgen and Yasseri, 2016; Bruns and Ioannidis, 2016), especially in relation
112
+ to the arbitrary choice of the α-level (most often 0.1, 0.05, or 0.01) and sample size issues.
113
+ Maximum likelihood estimators, obtained by maximizing the log-likelihood function, often
114
+ have low bias and large variance. Estimation accuracy can sometimes be improved by shrinking
115
+ some parameters to zero (Tibshirani, 1996). The associated shrinkage estimator improves the
116
+ overall prediction accuracy at the expense of introducing a small bias to reduce the variance of
117
+ the parameters. This class of estimators is implicit in Bayesian inference and penalized likelihood
118
+ inference. Using shrinkage estimators is applied as an alternative to hypothesis testing. Lasso,
119
+ Ridge and Stein-type estimators are the most widely used examples of penalizing methods (see,
120
+ for example, Hastie et al., 2009).
121
+ 2.1
122
+ Inference
123
+ Let Dx be the number of deaths in a given age interval [x, x + 1) for x = 0, . . . , m, and Ex
124
+ denote the number of person-years lived in the same interval (see, for example Brillinger, 1986;
125
+ Macdonald et al., 2018). Define D = (D0, D1, . . . , Dm)⊤ and E = (E0, E1, . . . , Em)⊤. In
126
+ addition, let θ = (a, b, σ2)⊤ ∈ Θ be the parameter vector that characterizes the force of mortality
127
+ at age x of the gamma-Gompertz model given by (1).
128
+ Assume Dx are Poisson-distributed with E(Dx) = VAR(Dx) = µ(x; θ)Ex for x = 0, . . . , m
129
+ (Brillinger, 1986). Under this assumption, the log-likelihood function for θ = (a, b, σ2)⊤ is
130
+ given by
131
+ ℓ(θ) = ℓ(θ|D, E) =
132
+ m
133
+
134
+ x=0
135
+ [Dx ln µ(x; θ)) − Ex µ(x; θ)] .
136
+ (3)
137
+ 3
138
+
139
+ Maximizing ℓ(θ) with respect to θ = (a, b, σ2)⊤ yields the maximum-likelihood (ML) estimate
140
+ ˆθ.
141
+ Let us now define a penalized log-likelihood function as
142
+ ℓp(θ) = ℓ(θ) − p(σ2) ,
143
+ (4)
144
+ where ℓ(θ) is the standard log-likelihood (3), while p(σ2) is a penalty function. The penalized
145
+ maximum-likelihood estimate is obtained by maximizing ℓp(θ) with respect to θ = (a, b, σ2)⊤.
146
+ For the problem addressed in this paper, the penalty function p(σ2) must be a non-decreasing
147
+ monotonic continuous function and lim
148
+ σ2↓0 p(σ2) > p(σ2) for all σ2 > 0.
149
+ In a Bayesian framework, maximizing (4) is equivalent to maximizing a posterior distribution
150
+ in a setting, in which e−p(σ2)/Cp, Cp :=
151
+
152
+ Θ e−p(σ2)∇θ < ∞, is taken as a prior distribution of
153
+ θ. This procedure yields the maximum a posteriori probability (MAP) estimator. MAP is the
154
+ only Bayesian estimator that minimizes the expected canonical loss (Pereyra, 2019) and is widely
155
+ used in image and video processing (Greig et al., 1989; Afonso et al., 2010; Belekos et al., 2010).
156
+ As σ2 describes the variance of frailty at the starting age of analysis, the standard approach
157
+ would be to specify an inverse gamma prior distribution for it (Gelman et al., 1995). The inverse
158
+ gamma distribution is heavy-tailed and keeps probability mass further from zero than the gamma
159
+ distribution. In addition, while the inverse-gamma mode is always positive, the gamma mode
160
+ can also be zero (Llera and Beckmann, 2016). As we aim to test whether σ2 = 0 or σ2 > 0, we
161
+ will use the log-kernel of the gamma distribution to define the penalty function as
162
+ p(σ2) = λ
163
+
164
+ σ2 + ln σ2�
165
+ (5)
166
+ for some non-negative λ. When λ < 1, using (5) is equivalent to specifying a gamma prior
167
+ distribution for σ2 with parameters α = 1 − λ and β = λ.
168
+ When m → ∞, the effect of the penalty diminishes regardless of the size of λ. For human
169
+ life table data m is finite, thus λ ≥ 0 is a constant that controls the relative impact of the penalty
170
+ function on the estimates. When λ = 0, the penalty term has no effect, and maximizing the
171
+ penalized likelihood will produce the standard maximum likelihood estimates (MLE). However,
172
+ as λ → ∞, the impact of the penalty grows, and the maximum penalized likelihood estimates
173
+ for σ2 will approach zero, providing high precision, but low accuracy.
174
+ Choosing λ is sensible in a wide range of applications (Li et al., 2009; Bhattacharya and
175
+ McNicholas, 2014). Therefore, in accordance with the recommendations in Li et al. (2009),
176
+ we carry out a pilot simulation study, in which we find that choosing λ = 1
177
+ 2 provides similar
178
+ precision to the one by MLE when σ2 > 0, but better accuracy and precision when σ2 = 0
179
+ (simulation results are presented in the next subsection). As a result, the final expression for the
180
+ penalized log-likelihood we propose is
181
+ ℓp(θ) =
182
+ m
183
+
184
+ x=0
185
+ [Dx ln µ(x; θ) − Ex µ(x; θ)] − 1
186
+ 2
187
+
188
+ ln σ2 + σ2�
189
+ .
190
+ (6)
191
+ From a Bayesian perspective, choosing λ = 1
192
+ 2 provides an informative prior distribution for
193
+ σ2. As for human populations we are likely to estimate σ2 < 1 (Missov, 2013), the specified
194
+ prior will provide for σ2 a distribution with a mode equal to zero, a median equal to 0.4549, and
195
+ 4
196
+
197
+ 0.00000
198
+ 0.00010
199
+ 0.00020
200
+ 0.00030
201
+ −600
202
+ −400
203
+ −200
204
+ 0
205
+ Parameter a
206
+ a
207
+ log−likelihood
208
+ MLE
209
+ MAP
210
+ a = 1e−04
211
+ b = 0.1
212
+ σ2 = 0.1
213
+ 0.00
214
+ 0.05
215
+ 0.10
216
+ 0.15
217
+ 0.20
218
+ 0.25
219
+ 0.30
220
+ −8000
221
+ −4000
222
+ 0
223
+ Parameter b
224
+ b
225
+ log−likelihood
226
+ MLE
227
+ MAP
228
+ a = 1e−04
229
+ b = 0.1
230
+ σ2 = 0.1
231
+ 0.00
232
+ 0.05
233
+ 0.10
234
+ 0.15
235
+ 0.20
236
+ 0.25
237
+ 0.30
238
+ −5.5
239
+ −4.5
240
+ −3.5
241
+ Parameter σ2
242
+ σ2
243
+ log−likelihood
244
+ MLE
245
+ MAP
246
+ a = 1e−04
247
+ b = 0.1
248
+ σ2 = 0.1
249
+ 0.00000
250
+ 0.00010
251
+ 0.00020
252
+ 0.00030
253
+ −800
254
+ −600
255
+ −400
256
+ −200
257
+ 0
258
+ Parameter a
259
+ a
260
+ log−likelihood
261
+ MLE
262
+ MAP
263
+ a = 1e−04
264
+ b = 0.1
265
+ σ2 = 0
266
+ 0.00
267
+ 0.05
268
+ 0.10
269
+ 0.15
270
+ 0.20
271
+ 0.25
272
+ 0.30
273
+ −8000
274
+ −4000
275
+ 0
276
+ Parameter b
277
+ b
278
+ log−likelihood
279
+ MLE
280
+ MAP
281
+ a = 1e−04
282
+ b = 0.1
283
+ σ2 = 0
284
+ 0.000
285
+ 0.005
286
+ 0.010
287
+ 0.015
288
+ 0.020
289
+ 0.025
290
+ 0.030
291
+ −3.55
292
+ −3.45
293
+ −3.35
294
+ Parameter σ2
295
+ σ2
296
+ log−likelihood
297
+ MLE
298
+ MAP
299
+ a = 1e−04
300
+ b = 0.1
301
+ σ2 = 0
302
+ Figure 1: Plots of the profile log-likelihood and penalized log-likelihood functions of the param-
303
+ eters. In the first row we used synthetic data from a gamma-Gompertz model with parameters
304
+ a = 0.0001, b = 0.1 and σ2 = 0.1, in the second row we from a Gompertz model with parameters
305
+ a = 0.0001 and b = 0.1.
306
+ mean equal to 1. Furthermore, the prior provides a probability mass of 0.6826 in the interval
307
+ (0, 1].
308
+ Figure 1 shows the log-likelihood and penalized log-likelihood functions for all parameters
309
+ when σ2 > 0 (first row) and σ2 = 0 (second row). When σ2 > 0, the penalty function affects
310
+ neither the shape of the log-likelihood, nor the location of its maximum. However, when σ2 = 0,
311
+ adding a penalty yields a higher maximum at 0. Moreover, when σ2 = 0, the first and second
312
+ derivatives of the penalized log-likelihood are higher than their respective counterparts of the
313
+ log-likelihood. As a result, derivative-based optimization methods may reach the maximum
314
+ point faster, and the estimator ˆσ2 may have a smaller variance.
315
+ 2.2
316
+ Monte Carlo simulations
317
+ We carry out Monte Carlo simulations to explore the performance of the MAP and ML methods
318
+ in estimating the gamma-Gompertz model parameters. We use the R software (Team et al.,
319
+ 2022) to maximize the log-likelihood and the penalized log-likelihood functions via the optim
320
+ function applying as a pre-step differential evolution (Storn and Price, 1997; Ardia et al., 2011).
321
+ The performance of the ML and MAP estimators are evaluated by calculating two measures: the
322
+ bias and the standard deviation.
323
+ We generate 10,000 random samples from this model for some parameter values (scenarios
324
+ with sample sizes of 2,000 and 5,000 were also considered, and are presented in the appendix).
325
+ From these samples, we generate life tables and use them to estimate model parameters via the
326
+ 5
327
+
328
+ MAP and MLE methods. This process was repeated 2,000 times. In the presence of unobserved
329
+ heterogeneity, the true parameter values are a1 = 0.0001 and a2 = 0.00001 for a, b1 = 0.1 and
330
+ b2 = 0.15 for b, and σ2
331
+ 1 = 0.2 and σ2
332
+ 2 = 0.8 for σ2. When there is no heterogeneity (σ2 = 0),
333
+ the true parameter values are a1 = 0.0001, a2 = 0.0003 and a3 = 0.0005 for a, and b1 = 0.09,
334
+ b2 = 0.10 and b3 = 0.11 for b.
335
+ Table 1: Simulation results: gamma-Gompertz model and sample size 10,000.
336
+ There is heterogeneity
337
+ MLE estimator
338
+ MAP estimator
339
+ Bias
340
+ Standard deviation
341
+ Bias
342
+ Standard deviation
343
+ Parameter
344
+ a
345
+ b
346
+ σ2
347
+ a
348
+ b
349
+ σ2
350
+ a
351
+ b
352
+ σ2
353
+ a
354
+ b
355
+ σ2
356
+ (a1, b1, σ2
357
+ 1)
358
+ 0.000053
359
+ -0.000051
360
+ -0.000223
361
+ 0.000051
362
+ 0.001499
363
+ 0.020721
364
+ 0.000055
365
+ -0.000134
366
+ -0.001626
367
+ 0.000052
368
+ 0.001502
369
+ 0.020791
370
+ (a1, b1, σ2
371
+ 2)
372
+ 0.000060
373
+ -0.000292
374
+ -0.007787
375
+ 0.000056
376
+ 0.001784
377
+ 0.035822
378
+ 0.000061
379
+ -0.000354
380
+ -0.009229
381
+ 0.000056
382
+ 0.001783
383
+ 0.035795
384
+ (a1, b2, σ2
385
+ 1)
386
+ 0.000077
387
+ 0.000131
388
+ 0.004431
389
+ 0.000056
390
+ 0.002181
391
+ 0.020569
392
+ 0.000080
393
+ 0.000015
394
+ 0.003096
395
+ 0.000057
396
+ 0.002186
397
+ 0.020631
398
+ (a1, b2, σ2
399
+ 2)
400
+ 0.000085
401
+ -0.000262
402
+ -0.003547
403
+ 0.000061
404
+ 0.002557
405
+ 0.034714
406
+ 0.000087
407
+ -0.000348
408
+ -0.004920
409
+ 0.000061
410
+ 0.002556
411
+ 0.034687
412
+ (a2, b1, σ2
413
+ 1)
414
+ 0.000007
415
+ -0.000349
416
+ -0.003349
417
+ 0.000008
418
+ 0.001315
419
+ 0.019464
420
+ 0.000008
421
+ -0.000417
422
+ -0.004597
423
+ 0.000008
424
+ 0.001318
425
+ 0.019523
426
+ (a2, b1, σ2
427
+ 2)
428
+ 0.000009
429
+ -0.000635
430
+ -0.013592
431
+ 0.000008
432
+ 0.001515
433
+ 0.032551
434
+ 0.000009
435
+ -0.000683
436
+ -0.014810
437
+ 0.000008
438
+ 0.001514
439
+ 0.032528
440
+ (a2, b2, σ2
441
+ 1)
442
+ 0.000009
443
+ -0.000170
444
+ 0.002295
445
+ 0.000008
446
+ 0.001963
447
+ 0.019377
448
+ 0.000009
449
+ -0.000268
450
+ 0.001078
451
+ 0.000008
452
+ 0.001966
453
+ 0.019430
454
+ (a2, b2, σ2
455
+ 2)
456
+ 0.000011
457
+ -0.000650
458
+ -0.007809
459
+ 0.000009
460
+ 0.002273
461
+ 0.032534
462
+ 0.000011
463
+ -0.000721
464
+ -0.009014
465
+ 0.000009
466
+ 0.002272
467
+ 0.032510
468
+ There is no heterogeneity
469
+ MLE estimator
470
+ MAP estimator
471
+ Bias
472
+ Standard deviation
473
+ Bias
474
+ Standard deviation
475
+ Parameter
476
+ a
477
+ b
478
+ σ2
479
+ a
480
+ b
481
+ σ2
482
+ a
483
+ b
484
+ σ2(10−16)
485
+ a
486
+ b
487
+ σ2(10−15)
488
+ (a1, b1, σ2)
489
+ 0.000005
490
+ -0.000055
491
+ 0.000181
492
+ 0.000006
493
+ 0.000875
494
+ 0.008460
495
+ 0.000007
496
+ -0.000239
497
+ 0.125942
498
+ 0.000006
499
+ 0.000791
500
+ 0.6308937
501
+ (a1, b2, σ2)
502
+ 0.000006
503
+ -0.000050
504
+ 0.000222
505
+ 0.000006
506
+ 0.000968
507
+ 0.008907
508
+ 0.000007
509
+ -0.000407
510
+ 0.060433
511
+ 0.000006
512
+ 0.000870
513
+ 0.1460452
514
+ (a1, b3, σ2)
515
+ 0.000006
516
+ -0.000057
517
+ 0.000428
518
+ 0.000006
519
+ 0.001082
520
+ 0.009726
521
+ 0.000008
522
+ -0.000305
523
+ 0.089035
524
+ 0.000006
525
+ 0.000978
526
+ 8.212895
527
+ (a2, b1, σ2)
528
+ 0.000011
529
+ 0.000138
530
+ 0.001864
531
+ 0.000016
532
+ 0.000913
533
+ 0.009010
534
+ 0.000016
535
+ -0.000198
536
+ 0.004470
537
+ 0.000015
538
+ 0.000811
539
+ 0.294978
540
+ (a2, b2, σ2)
541
+ 0.000014
542
+ 0.000117
543
+ 0.001094
544
+ 0.000016
545
+ 0.001042
546
+ 0.009873
547
+ 0.000019
548
+ -0.000264
549
+ 0.003661
550
+ 0.000015
551
+ 0.000859
552
+ 4.922260
553
+ (a2, b3, σ2)
554
+ 0.000015
555
+ 0.000164
556
+ 0.002204
557
+ 0.000016
558
+ 0.001150
559
+ 0.010369
560
+ 0.000022
561
+ -0.000216
562
+ 0.007167
563
+ 0.000016
564
+ 0.001017
565
+ 3.551245
566
+ (a3, b1, σ2)
567
+ 0.000018
568
+ 0.000143
569
+ 0.001062
570
+ 0.000025
571
+ 0.000979
572
+ 0.009712
573
+ 0.000027
574
+ -0.000138
575
+ 0.001124
576
+ 0.000025
577
+ 0.000876
578
+ 1.898573
579
+ (a3, b2, σ2)
580
+ 0.000024
581
+ 0.000076
582
+ 0.000505
583
+ 0.000025
584
+ 0.001057
585
+ 0.009721
586
+ 0.000030
587
+ -0.000112
588
+ 0.001177
589
+ 0.000025
590
+ 0.000942
591
+ 8.114498
592
+ (a3, b3, σ2)
593
+ 0.000025
594
+ 0.000117
595
+ 0.001427
596
+ 0.000025
597
+ 0.001178
598
+ 0.009999
599
+ 0.000033
600
+ -0.000171
601
+ 0.001415
602
+ 0.000024
603
+ 0.000990
604
+ 0.000025
605
+ The simulation results are presented in Table 1. In the presence of unobserved heterogeneity,
606
+ both methods underestimate b and σ2. They also introduce a small positive bias to a, the one pro-
607
+ vided by ML estimator being slightly smaller. However, in general the ML and MAP estimators
608
+ perform equally well, with a similar bias and standard deviation.
609
+ In the absence of unobserved heterogeneity, the ML estimator provides again a smaller bias
610
+ for a and b than the MAP estimator. However, in this case, the MAP method estimates more
611
+ precisely the frailty parameter σ2, with a bias and a standard deviation close to zero (∝ 10−15).
612
+ The MAP estimator also provides a slight reduction in the standard deviation of parameter b.
613
+ By the Monte Carlo simulation we also calculate the proportion of trials in which MAP
614
+ estimates σ2 > 0 when the true values is σ2 = 0 (error type I), as well as the proportion of trials
615
+ in which MAP estimates σ2 = 0 when the true values is σ2 > 0 (error type II). Based on our
616
+ simulations, the type I errro equals 0.001502, while the type II error is 0.001126.
617
+ The Monte Carlo simulations show that using a penalizing likelihood function (6) is an alter-
618
+ native to hypothesis testing, the latter being dependent on the asymptotic distribution of the ML
619
+ estimator, sample size and the arbitrary choice of the α-level (B¨ohnstedt and Gampe, 2019).
620
+ 3
621
+ Performance of MAP and ML estimators on HMD data
622
+ In this section, we estimate the gamma-Gompertz model via ML and MAP using mortality data
623
+ from the Human Mortality Database (HMD, 2022). We take exposures and raw death counts for
624
+ 6
625
+
626
+ the female population of France, Japan and the USA in the years 1960, 1980, 2000, and 2020,
627
+ after age 70. We apply again R (Team et al., 2022) to compute the ML and MAP estimates of
628
+ θ = (a, b, σ2)′ by using differential evolution. We use the mean squared error given by
629
+ MSE = 1
630
+ n
631
+ m
632
+
633
+ x=0
634
+
635
+ ln mx − ln ¯µ(x; ˆθ)
636
+ �2
637
+ ,
638
+ to assess the goodness of fit.
639
+ Table 2: Life expectancy: gamma Gompertz–Makeham model and ML estimates.
640
+ ML Estimates
641
+ MAP Estimates
642
+ Country
643
+ Year
644
+ a
645
+ b
646
+ σ2
647
+ MSE
648
+ a
649
+ b
650
+ σ2
651
+ MSE
652
+ France
653
+ 1960
654
+ 0.003582
655
+ 0.107599
656
+ 0.016726
657
+ 0.100588
658
+ 0.003593
659
+ 0.107483
660
+ 0.015902
661
+ 0.100540
662
+ 1980
663
+ 0.002210
664
+ 0.112032
665
+ 0.003393
666
+ 0.045776
667
+ 0.002220
668
+ 0.111729
669
+ 0.003117
670
+ 0.046747
671
+ 2000
672
+ 0.001247
673
+ 0.117749
674
+ 0.000001
675
+ 0.073648
676
+ 0.001250
677
+ 0.117689
678
+ 0
679
+ 0.073262
680
+ 2020
681
+ 0.000957
682
+ 0.119494
683
+ 0.000002
684
+ 0.094243
685
+ 0.000960
686
+ 0.119416
687
+ 0
688
+ 0.093697
689
+ Japan
690
+ 1960
691
+ 0.004782
692
+ 0.105858
693
+ 0.039989
694
+ 0.011366
695
+ 0.004782
696
+ 0.105845
697
+ 0.039593
698
+ 0.011493
699
+ 1980
700
+ 0.002009
701
+ 0.117886
702
+ 0.015251
703
+ 0.053944
704
+ 0.002009
705
+ 0.117941
706
+ 0.015942
707
+ 0.053328
708
+ 2000
709
+ 0.001140
710
+ 0.115268
711
+ 0.000015
712
+ 0.064604
713
+ 0.001142
714
+ 0.115118
715
+ 0
716
+ 0.063728
717
+ 2020
718
+ 0.000575
719
+ 0.125814
720
+ 0.000233
721
+ 0.104944
722
+ 0.000574
723
+ 0.125870
724
+ 0.000225
725
+ 0.105597
726
+ USA
727
+ 1960
728
+ 0.004701
729
+ 0.095797
730
+ 0.032146
731
+ 0.111711
732
+ 0.004699
733
+ 0.095814
734
+ 0.032802
735
+ 0.110740
736
+ 1980
737
+ 0.003612
738
+ 0.093688
739
+ 0.000001
740
+ 0.054483
741
+ 0.003609
742
+ 0.093720
743
+ 0
744
+ 0.054652
745
+ 2000
746
+ 0.002712
747
+ 0.100566
748
+ 0.000003
749
+ 0.030967
750
+ 0.002714
751
+ 0.100540
752
+ 0
753
+ 0.030873
754
+ 2020
755
+ 0.002473
756
+ 0.101652
757
+ 0.000001
758
+ 0.022735
759
+ 0.002476
760
+ 0.101612
761
+ 0
762
+ 0.022618
763
+ Table 2 shows the results of applying ML and MAP methods to the datasets described above.
764
+ The MAP estimator provides lower MSEs in 8 of the 12 datasets. When the standard ML method
765
+ estimates σ2 < 10−4, our novel method estimates σ2 = 0 and provides a smaller MSE. This
766
+ suggests that the MAP provides a slightly better fit to the data. Overall, MAP performs better
767
+ than ML when unobserved heterogeneity is not detected, and while for estimates of ˆσ2 > 0 ML
768
+ has a slight advantage.
769
+ The results from the real-data application back up the results from the Monte Carlo simula-
770
+ tions in Section 2. In the presence of unobserved heterogeneity, the MLE method provides the
771
+ most precise and accurate estimates. The MAP method, though, has just slightly lower precision.
772
+ On the other hand, in the absence of unobserved heterogeneity, the MAP provides smaller bias
773
+ and variance in its estimates compared to MLE.
774
+ 3.1
775
+ Examples when MAP and ML estimators yield different outcomes
776
+ Using MAP and ML estimators does not always lead to the same statistical inference. One of
777
+ them can detect heterogeneity in cases when the other does not. We will illustrate this on HMD
778
+ data for the Japanese female population in 2009 and the French female population born in 1848,
779
+ ages 70+. To assess the goodness of fit, we will use again MSE.
780
+ For Japanese females in 2009, ML yields estimates ˆθMLE = (0.006359, 0.133805, 0.070513)′
781
+ with standard errors SE(a) = 0.000188, SE(b) = 0.002263 and SE(σ2) = 0.021156. The 95%
782
+ confidence interval for σ2 is (0.029047, 0.111978) indicating stasitically significant unobserved
783
+ 7
784
+
785
+ heterogeneity, i.e., the existence of mortality deceleration. On the other hand, the MAP method
786
+ estimates ˆθMAP = (0.006966, 0.125440, 0)′, indicating the absence of unobserved heterogeneity.
787
+ Comparing the goodness of fit of both methods speaks in favor of the MAP outcome: MAP’s
788
+ MSE is by 37% lower than ML’s LSE (0.018691 for MAP vs 0.029958 for ML). It indicates that
789
+ unobserved heterogeneity is negligible and that the gamma-Gompertz model is misspecified.
790
+ 70
791
+ 80
792
+ 90
793
+ 100
794
+ 110
795
+ −5
796
+ −4
797
+ −3
798
+ −2
799
+ −1
800
+ 0
801
+ Age
802
+ log−force of mortality
803
+ Mortality rate
804
+ MLE
805
+ MAP
806
+ Japan
807
+ 70
808
+ 75
809
+ 80
810
+ 85
811
+ 90
812
+ 95
813
+ 100
814
+ −3.0
815
+ −2.0
816
+ −1.0
817
+ 0.0
818
+ Age
819
+ log−force of mortality
820
+ Mortality rate
821
+ MLE
822
+ MAP
823
+ France
824
+ Figure 2: MAP and MLE estimates of the force of mortality for the Japanese population in 2009
825
+ and the Swedish population born in 1881, after age 70
826
+ The left panel of Figure 2 shows that both methods estimate a similar logarithmic force of
827
+ mortality at most ages. However, after age 100, the MLE deviates downward from the observed
828
+ logarithmic death rates.
829
+ The MAP also provides a better fit and different conclusion for the cohort of French females
830
+ born in 1848. While ML estimates ˆθMLE = (0.053748, 0.090552, 0.008604)′ with SE(a) =
831
+ 0.000317, SE(b) = 0.001273, SE(σ2) = 0.007562 and provides an MSE equal 0.046222,
832
+ MAP estimates ˆθMAP = (0.053113, 0.094921, 0.036466)′ and provides MSE = 0.034226, i.e.,
833
+ MAP’s MSE is by 26% smaller than ML’s MSE.
834
+ Furthermore, while the MAP estimate of σ2 suggests that there is non-negligible unobserved
835
+ heterogeneity, the ML estimate and standard error for σ2 indicates the opposite: the amount
836
+ of unobserved heterogeneity is not statistically significant. The right panel of Figure 2 shows
837
+ the difference between these estimates. MAP’s estimate shows a leveling-off in the force of
838
+ mortality, while the MLE shows a log-linear increase in the hazard function.
839
+ 4
840
+ Concluding remarks
841
+ B¨ohnstedt and Gampe (2019) introduced a formal procedure to identify whether σ2 > 0 or
842
+ σ2 = 0 in a hypothesis testing setting: they studied the asymptotic properties of the maximum
843
+ likelihood estimator and the likelihood ratio test (LRT) for H0 : σ2 = 0 vs. H1 : σ2 = 0 for
844
+ 8
845
+
846
+ the gamma-Gompertz model. However, LRTs are based on the asymptotic distribution of the
847
+ maximum likelihood estimator, hence its convergence depends on the sample size. Moreover,
848
+ conclusions drawn from hypothesis tests are dependent on the arbitrary choice of the significance
849
+ level or p-value.
850
+ We suggest an alternative method by considering the problem as model misspecification. We
851
+ add a penalty function to the likelihood so that we make sure that ˆσ2 = 0 when there is no het-
852
+ erogeneity. We also present a Bayesian interpretation (MAP) to our method. We take advantage
853
+ of robust Monte Carlo simulations to measure the bias and standard deviation of the ML and
854
+ MAP methods in scenarios with and without unobserved heterogeneity. We also compare the
855
+ performance of both methods for estimating the gamma-Gompertz model parameters using ac-
856
+ tual mortality data from the Human Mortality Database. The two methods work almost equally
857
+ well, the ML having a slight advantage, in the presence of unobserved heterogeneity. However,
858
+ in the absence of the latter, the MAP method provides an estimate closer to 0 (ˆσ2 ≈ 10−20) and a
859
+ better fit to the model in comparison to ML. As a result, the method we propose here can be used
860
+ as an alternative to likelihood ratio testing for the gamma-Gompertz model with H0 : σ2 = 0 vs.
861
+ H1 : σ2 > 0. On the one hand, the MAP method does not depend on any asymptomatic distribu-
862
+ tion, its performance is not strongly affected by sample size, and it also does not depend on the
863
+ arbitrary choice of the significance level. On the other hand, MAP provides similar estimates to
864
+ the ones by ML when σ2 > 0 and more accurate estimates when σ2 = 0.
865
+ Acknowledgments
866
+ The research leading to this publication is a part of a project that has received funding from
867
+ the European Research Council (ERC) under the European Union’s Horizon 2020 research and
868
+ innovation programme (Grant agreement No. 884328 – Unequal Lifespans). Silvio C. Patricio
869
+ gratefully acknowledges the support provided from AXA Research Fund, through the funding
870
+ for the “AXA Chair in Longevity Research”.
871
+ Bibliography
872
+ Afonso, M. V., Bioucas-Dias, J. M., and Figueiredo, M. A. (2010). An augmented lagrangian
873
+ approach to the constrained optimization formulation of imaging inverse problems. IEEE
874
+ transactions on image processing, 20(3):681–695.
875
+ Alvarez, J.-A., Villavicencio, F., Strozza, C., and Camarda, C. G. (2021). Regularities in human
876
+ mortality after age 105. PloS one, 16(7):e0253940.
877
+ Ardia, D., Boudt, K., Carl, P., Mullen, K., and Peterson, B. G. (2011). Differential evolution with
878
+ deoptim: an application to non-convex portfolio optimization. The R Journal, 3(1):27–34.
879
+ Belekos, S. P., Galatsanos, N. P., and Katsaggelos, A. K. (2010). Maximum a posteriori video
880
+ super-resolution using a new multichannel image prior. IEEE Transactions on Image Process-
881
+ ing, 19(6):1451–1464.
882
+ 9
883
+
884
+ Belzile, L., Davison, A., Gampe, J., Rootzen, H., and Zholud, D. (2022). Is there a cap on
885
+ longevity? a statistical review. Annual Review of Statistics and Its Application, 9:21–45.
886
+ Berk, R., Brown, L., and Zhao, L. (2010). Statistical inference after model selection. Journal of
887
+ Quantitative Criminology, 26(2):217–236.
888
+ Bhattacharya, S. and McNicholas, P. D. (2014). A lasso-penalized bic for mixture model selec-
889
+ tion. Advances in Data Analysis and Classification, 8(1):45–61.
890
+ B¨ohnstedt, M. and Gampe, J. (2019). Detecting mortality deceleration: Likelihood inference and
891
+ model selection in the gamma-gompertz model. Statistics & Probability Letters, 150:68–73.
892
+ Brillinger, D. R. (1986). A biometrics invited paper with discussion: the natural variability of
893
+ vital rates and associated statistics. Biometrics, pages 693–734.
894
+ Bruns, S. B. and Ioannidis, J. P. (2016). P-curve and p-hacking in observational research. PloS
895
+ one, 11(2):e0149144.
896
+ Camarda, C. G. (2022). The curse of the plateau. measuring confidence in human mortality
897
+ estimates at extreme ages. Theoretical Population Biology, 144:24–36.
898
+ Carey, J., Liedo, P., and Vaupel, J. (1995). Mortality dynamics of density in the mediterranean
899
+ fruit fly. Experimental Gerontology, 30(6):605–629.
900
+ Curtsinger, J., Fukui, H., Townsend, D., and Vaupel, J. (1992).
901
+ Demography of genotypes:
902
+ Failure of the limited life-span paradigm in drosophila melanogaster. Science, 258:461–463.
903
+ Fukui, H., Ackert, L., and Curtsinger, J. (1996). Deceleration of age-specific mortality rates
904
+ in chromosomal homozygotes and heterozygotes of drosophila melanogaster. Experimental
905
+ Gerontology, 36(4):517–531.
906
+ Fukui, H., Curtsinger, J., and Xiu, L. (1993). Slowing of age-specific mortality rates in drosophila
907
+ melanogaster. Experimental Gerontology, 28:585–599.
908
+ Gampe, J. (2010). Human mortality beyond age 110. In Supercentenarians, pages 219–230.
909
+ Springer.
910
+ Gampe, J. (2021). Mortality of supercentenarians: Estimates from the updated idl. In Exceptional
911
+ Lifespans, pages 29–35. Springer, Cham.
912
+ Gavrilova, N. S. and Gavrilov, L. A. (2015). Biodemography of old-age mortality in humans
913
+ and rodents. Journals of Gerontology Series A: Biomedical Sciences and Medical Sciences,
914
+ 70(1):1–9.
915
+ Gelman, A., Carlin, J. B., Stern, H. S., and Rubin, D. B. (1995). Bayesian data analysis. Chap-
916
+ man and Hall/CRC.
917
+ Gompertz, B. (1825). Xxiv. on the nature of the function expressive of the law of human mor-
918
+ tality, and on a new mode of determining the value of life contingencies. in a letter to francis
919
+ baily, esq. frs &c. Philosophical transactions of the Royal Society of London, (115):513–583.
920
+ 10
921
+
922
+ Greig, D. M., Porteous, B. T., and Seheult, A. H. (1989). Exact maximum a posteriori estima-
923
+ tion for binary images. Journal of the Royal Statistical Society: Series B (Methodological),
924
+ 51(2):271–279.
925
+ Hastie, T., Tibshirani, R., Friedman, J. H., and Friedman, J. H. (2009). The elements of statistical
926
+ learning: data mining, inference, and prediction, volume 2. Springer.
927
+ Head, M. L., Holman, L., Lanfear, R., Kahn, A. T., and Jennions, M. D. (2015). The extent and
928
+ consequences of p-hacking in science. PLoS biology, 13(3):e1002106.
929
+ HMD (2022). The human mortality database. http://www.mortality.org/.
930
+ Khazaeli, A., Pletcher, S., and Curtsinger, J. (1998). The fractionation experiment: reducing
931
+ heterogeneity to investigate age-specific mortality in drosophila. Mechanisms of Aging and
932
+ Development, 105:301–317.
933
+ Li, P., Chen, J., and Marriott, P. (2009). Non-finite fisher information and homogeneity: an em
934
+ approach. Biometrika, 96(2):411–426.
935
+ Llera, A. and Beckmann, C. (2016). Estimating an inverse gamma distribution. arXiv preprint
936
+ arXiv:1605.01019.
937
+ Macdonald, A. S., Richards, S. J., and Currie, I. D. (2018). Modelling mortality with actuarial
938
+ applications. Cambridge University Press.
939
+ Missov, T. I. (2013). Gamma-gompertz life expectancy at birth. Demographic Research, 28:259–
940
+ 270.
941
+ Newman, S. J. (2018). Errors as a primary cause of late-life mortality deceleration and plateaus.
942
+ PLoS biology, 16(12):e2006776.
943
+ Pereyra, M. (2019). Revisiting maximum-a-posteriori estimation in log-concave models. SIAM
944
+ Journal on Imaging Sciences, 12(1):650–670.
945
+ Rootz´en, H. and Zholud, D. (2017). Human life is unlimited–but short. Extremes, 20(4):713–
946
+ 728.
947
+ Storn, R. and Price, K. (1997). Differential evolution – a simple and efficient heuristic for global
948
+ optimization over continuous spaces. Journal of Global Optimization, 11:341–359.
949
+ Team, R. C. et al. (2022). R: A language and environment for statistical computing.
950
+ Tibshirani, R. (1996). Regression shrinkage and selection via the lasso. Journal of the Royal
951
+ Statistical Society: Series B (Methodological), 58(1):267–288.
952
+ Vaupel, J. and Missov, T. (2014). Unobserved population heterogeneity: A review of formal
953
+ relationships. Demographic Research, 31(22):659–686.
954
+ Vaupel, J. W., Manton, K. G., and Stallard, E. (1979). The impact of heterogeneity in individual
955
+ frailty on the dynamics of mortality. Demography, 16(3):439–454.
956
+ 11
957
+
958
+ Vidgen, B. and Yasseri, T. (2016). P-values: misunderstood and misused. Frontiers in Physics,
959
+ 4:6.
960
+ Appendices
961
+ Table 3: Simulation results: gamma-Gompertz model and sample size 2,000.
962
+ There is heterogeneity
963
+ MLE estimator
964
+ MAP estimator
965
+ Bias
966
+ Standard deviation
967
+ Bias
968
+ Standard deviation
969
+ Parameter
970
+ a
971
+ b
972
+ σ2
973
+ a
974
+ b
975
+ σ2
976
+ a
977
+ b
978
+ σ2
979
+ a
980
+ b
981
+ σ2
982
+ (a1, b1, σ2
983
+ 1)
984
+ 0.000089
985
+ -0.001172
986
+ -0.020361
987
+ 0.000111
988
+ 0.003286
989
+ 0.047684
990
+ 0.000104
991
+ -0.001709
992
+ -0.029261
993
+ 0.000117
994
+ 0.003465
995
+ 0.051082
996
+ (a1, b1, σ2
997
+ 2)
998
+ 0.000110
999
+ -0.002075
1000
+ -0.048634
1001
+ 0.000128
1002
+ 0.003972
1003
+ 0.078621
1004
+ 0.000118
1005
+ -0.002397
1006
+ -0.055984
1007
+ 0.000129
1008
+ 0.003970
1009
+ 0.078467
1010
+ (a1, b2, σ2
1011
+ 1)
1012
+ 0.000104
1013
+ -0.000906
1014
+ -0.008883
1015
+ 0.000124
1016
+ 0.004826
1017
+ 0.047186
1018
+ 0.000120
1019
+ -0.001603
1020
+ -0.016820
1021
+ 0.000130
1022
+ 0.005032
1023
+ 0.049798
1024
+ (a1, b2, σ2
1025
+ 2)
1026
+ 0.000124
1027
+ -0.001978
1028
+ -0.029154
1029
+ 0.000139
1030
+ 0.005738
1031
+ 0.077069
1032
+ 0.000133
1033
+ -0.002420
1034
+ -0.036104
1035
+ 0.000140
1036
+ 0.005733
1037
+ 0.076901
1038
+ (a2, b1, σ2
1039
+ 1)
1040
+ 0.000030
1041
+ -0.003810
1042
+ -0.046694
1043
+ 0.000021
1044
+ 0.003024
1045
+ 0.045092
1046
+ 0.000034
1047
+ -0.004407
1048
+ -0.057147
1049
+ 0.000024
1050
+ 0.003388
1051
+ 0.052275
1052
+ (a2, b1, σ2
1053
+ 2)
1054
+ 0.000038
1055
+ -0.005056
1056
+ -0.097144
1057
+ 0.000023
1058
+ 0.003338
1059
+ 0.070146
1060
+ 0.000039
1061
+ -0.005316
1062
+ -0.103429
1063
+ 0.000024
1064
+ 0.003333
1065
+ 0.069973
1066
+ (a2, b2, σ2
1067
+ 1)
1068
+ 0.000027
1069
+ -0.003998
1070
+ -0.030152
1071
+ 0.000021
1072
+ 0.004259
1073
+ 0.043531
1074
+ 0.000030
1075
+ -0.004673
1076
+ -0.038167
1077
+ 0.000022
1078
+ 0.004481
1079
+ 0.046813
1080
+ (a2, b2, σ2
1081
+ 2)
1082
+ 0.000032
1083
+ -0.005191
1084
+ -0.065676
1085
+ 0.000025
1086
+ 0.005158
1087
+ 0.075375
1088
+ 0.000034
1089
+ -0.005565
1090
+ -0.071836
1091
+ 0.000025
1092
+ 0.005146
1093
+ 0.075141
1094
+ There is no heterogeneity
1095
+ MLE estimator
1096
+ MAP estimator
1097
+ Bias
1098
+ Standard deviation
1099
+ Bias
1100
+ Standard deviation
1101
+ Parameter
1102
+ a
1103
+ b
1104
+ σ2
1105
+ a
1106
+ b
1107
+ σ2
1108
+ a
1109
+ b
1110
+ σ2(10−12)
1111
+ a
1112
+ b
1113
+ σ2(10−12)
1114
+ (a1, b1, σ2)
1115
+ 0.000015
1116
+ -0.001225
1117
+ 0.000002
1118
+ 0.000013
1119
+ 0.001615
1120
+ 0.006546
1121
+ 0.000016
1122
+ -0.001341
1123
+ 0.021800
1124
+ 0.000014
1125
+ 0.001693
1126
+ 0.000287
1127
+ (a1, b2, σ2)
1128
+ 0.000014
1129
+ -0.001224
1130
+ 0.000002
1131
+ 0.000013
1132
+ 0.001833
1133
+ 0.007389
1134
+ 0.000016
1135
+ -0.001312
1136
+ 0.018541
1137
+ 0.000014
1138
+ 0.001858
1139
+ 0.004010
1140
+ (a1, b3, σ2)
1141
+ 0.000014
1142
+ -0.001323
1143
+ 0.000002
1144
+ 0.000013
1145
+ 0.001998
1146
+ 0.008857
1147
+ 0.000015
1148
+ -0.001324
1149
+ 0.024195
1150
+ 0.000015
1151
+ 0.002098
1152
+ 0.001130
1153
+ (a2, b1, σ2)
1154
+ 0.000028
1155
+ -0.000591
1156
+ 0.000003
1157
+ 0.000031
1158
+ 0.001703
1159
+ 0.009242
1160
+ 0.000031
1161
+ -0.000983
1162
+ 0.000923
1163
+ 0.000033
1164
+ 0.001666
1165
+ 0.000003
1166
+ (a2, b2, σ2)
1167
+ 0.000029
1168
+ -0.000837
1169
+ 0.000003
1170
+ 0.000031
1171
+ 0.001857
1172
+ 0.010440
1173
+ 0.000031
1174
+ -0.000842
1175
+ 0.000731
1176
+ 0.000031
1177
+ 0.001804
1178
+ 0.000009
1179
+ (a2, b3, σ2)
1180
+ 0.000030
1181
+ -0.000810
1182
+ 0.000004
1183
+ 0.000031
1184
+ 0.002067
1185
+ 0.011604
1186
+ 0.000035
1187
+ -0.001196
1188
+ 0.000701
1189
+ 0.000034
1190
+ 0.002086
1191
+ 0.000008
1192
+ (a3, b1, σ2)
1193
+ 0.000034
1194
+ -0.000398
1195
+ 0.000003
1196
+ 0.000046
1197
+ 0.001739
1198
+ 0.011618
1199
+ 0.000037
1200
+ -0.000606
1201
+ 0.000220
1202
+ 0.000048
1203
+ 0.001681
1204
+ 0.000001
1205
+ (a3, b2, σ2)
1206
+ 0.000036
1207
+ -0.000503
1208
+ 0.000005
1209
+ 0.000049
1210
+ 0.002062
1211
+ 0.013431
1212
+ 0.000040
1213
+ -0.000670
1214
+ 0.000151
1215
+ 0.000048
1216
+ 0.001901
1217
+ 0.000002
1218
+ (a3, b3, σ2)
1219
+ 0.000040
1220
+ -0.000238
1221
+ 0.000008
1222
+ 0.000050
1223
+ 0.002247
1224
+ 0.014168
1225
+ 0.000045
1226
+ -0.000722
1227
+ 0.000108
1228
+ 0.000051
1229
+ 0.002061
1230
+ 0.000008
1231
+ Table 4: Simulation results: gamma-Gompertz model and sample size 5,000.
1232
+ There is heterogeneity
1233
+ MLE estimator
1234
+ MAP estimator
1235
+ Bias
1236
+ Standard deviation
1237
+ Bias
1238
+ Standard deviation
1239
+ Parameter
1240
+ a
1241
+ b
1242
+ σ2
1243
+ a
1244
+ b
1245
+ σ2
1246
+ a
1247
+ b
1248
+ σ2
1249
+ a
1250
+ b
1251
+ σ2
1252
+ (a1, b1, σ2
1253
+ 1)
1254
+ 0.000067
1255
+ -0.000379
1256
+ -0.004470
1257
+ 0.000071
1258
+ 0.002078
1259
+ 0.029046
1260
+ 0.000071
1261
+ -0.000552
1262
+ -0.007370
1263
+ 0.000072
1264
+ 0.002090
1265
+ 0.029281
1266
+ (a1, b1, σ2
1267
+ 2)
1268
+ 0.000070
1269
+ -0.000582
1270
+ -0.015178
1271
+ 0.000077
1272
+ 0.002506
1273
+ 0.049888
1274
+ 0.000074
1275
+ -0.000708
1276
+ -0.018078
1277
+ 0.000077
1278
+ 0.002504
1279
+ 0.049808
1280
+ (a1, b2, σ2
1281
+ 1)
1282
+ 0.000090
1283
+ -0.000273
1284
+ 0.001254
1285
+ 0.000077
1286
+ 0.002969
1287
+ 0.028061
1288
+ 0.000095
1289
+ -0.000509
1290
+ -0.001483
1291
+ 0.000078
1292
+ 0.002982
1293
+ 0.028253
1294
+ (a1, b2, σ2
1295
+ 2)
1296
+ 0.000094
1297
+ -0.000514
1298
+ -0.007569
1299
+ 0.000083
1300
+ 0.003583
1301
+ 0.048230
1302
+ 0.000097
1303
+ -0.000687
1304
+ -0.010324
1305
+ 0.000084
1306
+ 0.003579
1307
+ 0.048152
1308
+ (a2, b1, σ2
1309
+ 1)
1310
+ 0.000014
1311
+ -0.001320
1312
+ -0.014568
1313
+ 0.000011
1314
+ 0.001786
1315
+ 0.026808
1316
+ 0.000014
1317
+ -0.001465
1318
+ -0.017208
1319
+ 0.000011
1320
+ 0.001795
1321
+ 0.027016
1322
+ (a2, b1, σ2
1323
+ 2)
1324
+ 0.000015
1325
+ -0.001650
1326
+ -0.033574
1327
+ 0.000013
1328
+ 0.002344
1329
+ 0.049087
1330
+ 0.000016
1331
+ -0.001748
1332
+ -0.036030
1333
+ 0.000013
1334
+ 0.002342
1335
+ 0.049016
1336
+ (a2, b2, σ2
1337
+ 1)
1338
+ 0.000014
1339
+ -0.001299
1340
+ -0.006177
1341
+ 0.000012
1342
+ 0.002706
1343
+ 0.026681
1344
+ 0.000015
1345
+ -0.001505
1346
+ -0.008713
1347
+ 0.000012
1348
+ 0.002717
1349
+ 0.026863
1350
+ (a2, b2, σ2
1351
+ 2)
1352
+ 0.000015
1353
+ -0.001571
1354
+ -0.019479
1355
+ 0.000014
1356
+ 0.003352
1357
+ 0.046735
1358
+ 0.000016
1359
+ -0.001714
1360
+ -0.021904
1361
+ 0.000014
1362
+ 0.003349
1363
+ 0.046666
1364
+ There is no heterogeneity
1365
+ MLE estimator
1366
+ MAP estimator
1367
+ Bias
1368
+ Standard deviation
1369
+ Bias
1370
+ Standard deviation
1371
+ Parameter
1372
+ a
1373
+ b
1374
+ σ2
1375
+ a
1376
+ b
1377
+ σ2
1378
+ a
1379
+ b
1380
+ σ2(10−12)
1381
+ a
1382
+ b
1383
+ σ2(10−12)
1384
+ (a1, b1, σ2)
1385
+ 0.000008
1386
+ -0.000312
1387
+ 0.000009
1388
+ 0.000008
1389
+ 0.001139
1390
+ 0.006564
1391
+ 0.000008
1392
+ -0.000441
1393
+ 0.024410
1394
+ 0.000009
1395
+ 0.001142
1396
+ 0.000058
1397
+ (a1, b2, σ2)
1398
+ 0.000008
1399
+ -0.000339
1400
+ 0.000011
1401
+ 0.000009
1402
+ 0.001290
1403
+ 0.007315
1404
+ 0.000009
1405
+ -0.000497
1406
+ 0.024494
1407
+ 0.000009
1408
+ 0.001259
1409
+ 0.000179
1410
+ (a1, b3, σ2)
1411
+ 0.000008
1412
+ -0.000237
1413
+ 0.000011
1414
+ 0.000009
1415
+ 0.001398
1416
+ 0.007896
1417
+ 0.000009
1418
+ -0.000483
1419
+ 0.037635
1420
+ 0.000009
1421
+ 0.001417
1422
+ 0.000985
1423
+ (a2, b1, σ2)
1424
+ 0.000015
1425
+ 0.000055
1426
+ 0.000823
1427
+ 0.000021
1428
+ 0.001212
1429
+ 0.008106
1430
+ 0.000016
1431
+ -0.000102
1432
+ 0.001137
1433
+ 0.000022
1434
+ 0.001184
1435
+ 0.000012
1436
+ (a2, b2, σ2)
1437
+ 0.000017
1438
+ 0.000028
1439
+ 0.000732
1440
+ 0.000022
1441
+ 0.001363
1442
+ 0.009211
1443
+ 0.000020
1444
+ -0.000228
1445
+ 0.001954
1446
+ 0.000023
1447
+ 0.001319
1448
+ 0.001225
1449
+ (a2, b3, σ2)
1450
+ 0.000019
1451
+ 0.000122
1452
+ 0.000182
1453
+ 0.000022
1454
+ 0.001473
1455
+ 0.009371
1456
+ 0.000021
1457
+ -0.000024
1458
+ 0.002634
1459
+ 0.000022
1460
+ 0.001387
1461
+ 0.000014
1462
+ (a3, b1, σ2)
1463
+ 0.000023
1464
+ 0.000212
1465
+ 0.001357
1466
+ 0.000033
1467
+ 0.001265
1468
+ 0.009088
1469
+ 0.000023
1470
+ -0.000068
1471
+ 0.000266
1472
+ 0.000033
1473
+ 0.001200
1474
+ 0.000001
1475
+ (a3, b2, σ2)
1476
+ 0.000025
1477
+ 0.000198
1478
+ 0.000845
1479
+ 0.000034
1480
+ 0.001391
1481
+ 0.009982
1482
+ 0.000030
1483
+ -0.000154
1484
+ 0.000558
1485
+ 0.000035
1486
+ 0.001332
1487
+ 0.000048
1488
+ (a3, b3, σ2)
1489
+ 0.000026
1490
+ 0.000127
1491
+ 0.000722
1492
+ 0.000035
1493
+ 0.001590
1494
+ 0.011029
1495
+ 0.000034
1496
+ -0.000151
1497
+ 0.001303
1498
+ 0.000035
1499
+ 0.001464
1500
+ 0.003727
1501
+ 12
1502
+
7NE1T4oBgHgl3EQfBgKg/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
7NE2T4oBgHgl3EQfkwf2/vector_store/index.faiss ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6c66897110c596dbb01220f402f2a2683864f4c77e796fe3ba76890f8f929733
3
+ size 1572909
99E4T4oBgHgl3EQf3g3K/content/tmp_files/2301.05307v1.pdf.txt ADDED
The diff for this file is too large to render. See raw diff
 
99E4T4oBgHgl3EQf3g3K/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
9dAzT4oBgHgl3EQfSfu4/content/2301.01234v1.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2a45c6480b577e65f59dd0ea5cebd771f7afc7bf77a26b2330369a9043cbeec3
3
+ size 906636
9dAzT4oBgHgl3EQfSfu4/vector_store/index.faiss ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4ddcea0ec1ad5c27881b9090e8687cce8975971d9c72a944d747ec6755db523e
3
+ size 2097197
ENE3T4oBgHgl3EQfVQoB/content/2301.04457v1.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aba5a5bd4e62a9056c9cb6dff2f72a90f1115c2f32c33df319cad293b860c703
3
+ size 396354
ENE3T4oBgHgl3EQfVQoB/vector_store/index.faiss ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:687897775cdc4323f465d2296a0486f07bd494380f1a09139be05ba16ff436c4
3
+ size 6160429
ENE3T4oBgHgl3EQfVQoB/vector_store/index.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c4f38b43080b878303fb56f3b5e650cb7ade61ba043eaed8a3877dcc0ffdd9ed
3
+ size 210938
EdAzT4oBgHgl3EQfUPw6/content/tmp_files/2301.01262v1.pdf.txt ADDED
@@ -0,0 +1,1967 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Reconfigurable microresonators induced in side-coupled optical fibers
2
+ V. VASSILIEV AND M. SUMETSKY*
3
+ Aston Institute of Photonic Technologies, Aston University, Birmingham B4 7ET, UK
4
+ *Email: m.sumetsky@aston.ac.uk
5
+
6
+
7
+ We experimentally demonstrate that side-coupling of coplanar bent optical fibers can induce a high Q-factor whispering
8
+ gallery mode (WGM) optical microresonator. To explain the effect, we consider WGMs with wavelengths close to the cutoff
9
+ wavelengths (CWs) of these fibers which slowly propagate along the fiber axes. In the vicinity of the touching region, WGMs
10
+ of adjacent fibers are coupled to each other, and CWs experience sub-nanoscale axial variation proportional to the coupling
11
+ strength. We show that in certain cases the CW variation leads to full localization of the WGMs and the creation of an optical
12
+ microresonator. By varying the characteristic curvature fiber radius from the centimeter order to millimeter order, we
13
+ demonstrate fully mechanically reconfigurable high Q-factor optical microresonators with dimensions varying from the
14
+ millimeter order to 100-micron order and free spectral range varying from a picometer to hundreds of picometers. The new
15
+ microresonators may find applications in cavity QED, microresonator optomechanics, frequency comb generation with
16
+ tunable repetition rate, tunable lasing, and tunable processing and delay of optical pulses.
17
+
18
+
19
+ 1. Introduction
20
+ Microphotonic devices and circuits commonly consist of one or
21
+ multiple connected basic elements, such as waveguides, couplers,
22
+ and ring resonators [1, 2]. In addition to the requirements of high
23
+ fabrication precision and low losses [2, 3], the tunability of these
24
+ circuits and devices is of critical importance for a variety of
25
+ applications [4, 5]. While more complex tunable microphotonics
26
+ circuits are targeted at tunability enabling quite arbitrary
27
+ predetermined signal processing (see e.g., [1]), simple microdevices,
28
+ such as standing along tunable three-dimensional microresonators,
29
+ allow for unique functionalities not possible to achieve by other
30
+ means. For a variety of applications, the tunability of spherical,
31
+ toroidal, and bottle microresonators has been demonstrated using
32
+ mechanical stretching, heating, and nonlinear light effects including
33
+ those in monolithic and specially coated microresonators [6-10]. In
34
+ most of these approaches, it is only possible to tune series of
35
+ wavelength eigenvalues simultaneously without noticeable change
36
+ in their separation.
37
+ However, for several applications, which include cavity QED [8,
38
+ 11, 12], optomechanics [13, 14], frequency microcomb generation
39
+ [15, 16], optical signal processing and delay [4, 5, 17], and lasing
40
+ [18-21], it is critical to have microresonators with tunable
41
+ eigenwavelength separation. For example, the latter allows the
42
+ creation of optical frequency microcomb generators and microlasers
43
+ with continuously tunable repetition rate and wavelength and to tune
44
+ the microresonator eigenfrequency separation in resonance with the
45
+ frequency of its mechanical oscillations. Considerable variation of
46
+ the eigenwavelength separation commonly requires the variation of
47
+ microresonator dimension and/or its refractive index parameters by
48
+ the quantity comparable with their original values. One approach to
49
+ solve this problem consists in using Fabry-Perot microresonators
50
+ with tunable mirror separation which contain the optical materials
51
+ under interest [12, 21, 22]. Additional flexibility of tuning can be
52
+ achieved by employing Fabry-Perot microresonators with a liquid
53
+ material inside [21] or translating a wedge-shaped solid optical
54
+ material to vary its dimensions inside the Fabry-Perot
55
+ microresonator [23].
56
+
57
+ Fig. 1. (a) Coplanar bent optical fibers touching each other. The fiber
58
+ profile is manipulated by bending and translation of the fiber tails
59
+ indicated by curved and straight arrows. (b) Illustration of coupling
60
+ between the input-output microfiber and WGMs in Fiber 1 and Fiber 2
61
+ near cutoff wavelengths.
62
+ Alternatively, of special interest is attaining the eigenwavelength
63
+ separation tunability in three-dimensional monolithic high Q-factor
64
+ microresonators, e.g., those with spherical, toroidal, and bottle
65
+
66
+ (a)
67
+ Fiber 2
68
+ Fiber 1
69
+ (b)
70
+ Coupled
71
+ WGMs
72
+ (12)
73
+ Fiber 2
74
+ (z)
75
+ 2n2
76
+ nin?
77
+ Direct
78
+ Fiber 1
79
+ l1n1
80
+ contact
81
+ x
82
+ Microfiber
83
+ Kn1
84
+ Zshapes. This will allow us to add tunability to the emerging
85
+ applications of these microresonators in QED, optomechanics,
86
+ lasing, and frequency comb generation noted above. However, the
87
+ deformation of most of these monolithic microresonators to achieve
88
+ significant change of their eigenwavelength separation is unfeasible.
89
+ A unique exception, though, is exhibited by SNAP (Surface
90
+ Nanoscale Axial Photonics) microresonators [24]. These
91
+ microresonators are introduced at the surface of an optical fiber by
92
+ its nanoscale deformation, which causes the nanoscale variation of
93
+ the cutoff wavelengths (CWs) controlling the slow propagation of
94
+ whispering gallery modes (WGMs) along the fiber axis (see [24, 25]
95
+ and references therein). In Ref. [26], a SNAP microresonator
96
+ induced and fully reconfigurable by local heating of an optical fiber
97
+ was demonstrated. In Ref. [27], it was shown that it is possible to
98
+ create a SNAP microresonator and control its dimensions by local
99
+ bending of an optical fiber. Both approaches allow for tuning of
100
+ eigenwavelength separation of microresonators by the quantity
101
+ comparable to or larger than its original value. However, in both
102
+ approaches, the induced microresonator shapes had limited
103
+ flexibility and their characteristic axial dimensions could not be
104
+ reduced below several millimeters. In the first case, this restriction
105
+ was caused by the imposed length of the characteristic heat
106
+ distribution along the fiber. In the second case, the reduction of
107
+ microresonator size was limited by the smallest curvature radius
108
+ corresponding to the fiber breakage threshold.
109
+ In this paper we report on our discovery of a new type of WGM
110
+ optical microresonators which belongs to the group of SNAP
111
+ microresonators. We show that side coupled coplanar bent fibers
112
+ (Fig. 1) can induce a high Q-factor SNAP microresonator localized
113
+ in the region of fiber coupling. The configuration of fibers shown in
114
+ Fig. 1 allows us to flexibly tune the shape of the induced SNAP
115
+ microresonators and their axial dimensions from several tens of
116
+ microns to several millimeters and, respectively, tune their
117
+ eigenwavelength separation from hundreds of picometers to a
118
+ picometer.
119
+ 2. Cutoff wavelengths of uncoupled and side-
120
+ coupled straight fibers
121
+ First, it is instructive to consider the behavior of CWs for
122
+ uncoupled and side-coupled straight optical fibers. For this
123
+ purpose, we cleave a 125-micron diameter uncoated
124
+ commercial optical fiber into two pieces (Fiber 1 and Fiber 2),
125
+ which are then coaxially aligned and put into contact along
126
+ 3.5 mm of their length as shown in Fig. 2(a). Light is launched
127
+ into Fiber 1 by a transversely oriented taper with the
128
+ micrometer
129
+ diameter
130
+ waist
131
+ (input-output
132
+ microfiber)
133
+ connected to the Optical Spectrum Analyzer (OSA). After
134
+ coupling into Fiber 1, light forms WGMs propagating along
135
+ the fiber surface. In the region of direct contact of fibers (Fig.
136
+ 2(a)), WGMs in Fiber 1 and Fiber 2 are coupled to each other.
137
+ To characterize the effect of interfiber coupling, we measured the
138
+ spectrograms of the configured fiber system. For this purpose, the
139
+ input-output microfiber was translated along Fiber 1 (Figs. 1(b) and
140
+ 2(a)) touching it periodically with the spatial resolution of 2 µm. At
141
+ the cut end of Fiber 1, the microfiber was moved towards Fiber 2 and
142
+ continued scanning Fiber 2. The spectrograms of transmission
143
+ power 𝑃�𝜆, 𝑧� were measured as a function of wavelength  and
144
+ microfiber position z along the axis of Fiber 1.
145
+
146
+ Fig. 2. (a) Illustration of side-coupled straight optical fiber configuration.
147
+ (b) Spectrogram of this configuration. (c) Magnified section outlined in
148
+ the spectrogram (b).
149
+ The measured spectrogram of our fiber system is shown in Fig.
150
+ 2(b). The left- and right-hand sides of this spectrogram show the
151
+ spectrograms of uncoupled Fiber 1 and Fiber 2, respectively. Lines
152
+ in spectrogram shown in Fig. 2(b) indicate the CWs of uncoupled
153
+ and coupled fibers. These CWs correspond to WGMs with different
154
+ azimuthal and radial quantum numbers. The magnified copy of the
155
+ section outlined in Fig. 2(b) is shown in Fig. 2(c). It is seen that the
156
+ CWs appear as straight lines slightly tilted with respect to the
157
+ horizontal direction. From the measured magnitude of tilt, ε� �
158
+ 0.015 nm/mm, we determine the linear variation of the fiber radius
159
+ ∆𝑟� � 𝑟�ε�/𝜆� � 0.6 nm/mm [28]. In the latter rescaling relation,
160
+ we used 𝑟� � 62.5 µm and 𝜆� � 1.55 nm. By linear extrapolation
161
+ of CWs of Fiber 1 and Fiber 2 (dashed white lines), we confirm that,
162
+ as expected, their positions (horizontal black dashed line) coincide at
163
+ the cut ends of these fibers.
164
+ At the 3.5 mm long region of fiber touching, WGMs in Fiber 1
165
+ couple to WGMs in Fiber 2 and the corresponding CWs split. The
166
+ structure and positions of CWs in the touching region depend on the
167
+ magnitude of coupling and will be further discussed in Section 4.
168
+ Here we note that the value of CW splitting found, e.g., from Fig.
169
+ 2(c) is ~ 0.1 nm, which coincides with characteristic values of CW
170
+ variation in SNAP microresonators [24, 25]. In particular, the
171
+ positive CW shift in the coupling region leads to the WGM
172
+ localization and creation of a microresonator which can be tuned by
173
+ changing the length of the side-coupled fiber segment. In our current
174
+ experiment, the Q-factor of the induced SNAP resonator was poor
175
+ due to the scattering of light at the imperfectly cleaved fiber ends,
176
+
177
+ (a)
178
+ Nottoscale
179
+ 3.5 mm
180
+ Fiber2
181
+ Fiber1
182
+ 125μm
183
+ Microfiber
184
+ 1549.00
185
+ (b)
186
+ 1548.50
187
+ -5
188
+ TransmissionPower (dB)
189
+ (wu)
190
+ 1548.00
191
+ length
192
+ 1547.50
193
+ 10
194
+ Wavel
195
+ 1547.00
196
+ 15
197
+ 1546.50
198
+ 1546.00
199
+ -20
200
+ 1545.50
201
+ 0
202
+ 1
203
+ 2
204
+ 3
205
+ 4
206
+ 5
207
+ 6
208
+ Distance along fiber (mm)
209
+ 0
210
+ 1548.50
211
+ (c)
212
+ 1548.40
213
+ -5
214
+ length
215
+ 1548.20
216
+ -10
217
+ 1548.10
218
+ Javel
219
+ 15
220
+ 1548.00
221
+ 1547.90
222
+ -20
223
+ 1547.80
224
+ 0
225
+ 1
226
+ 2
227
+ 3
228
+ 4
229
+ 5
230
+ 6
231
+ Distancealongfiber(mm)which, typically, ensure around 70% WGM reflectivity [29].
232
+ Nevertheless, we suggest that the demonstrated resonator can be
233
+ directly used to create miniature broadly tunable optical delay lines
234
+ generalizing our previous results based on the SNAP
235
+ microresonators with fixed dimensions [30, 31]. Indeed, in these
236
+ devices the WGM pulses complete only a single round trip along the
237
+ fiber axis and therefore their attenuation at the fiber facets may
238
+ reduce the output light power by around 50% only. We also suggest
239
+ that, after feasible improvement, the Q-factor of these
240
+ microresonators can be significantly improved as further discussed
241
+ in Section 6.
242
+ 3. Basic experiment
243
+ In our proof-of-concept experiments, we used 125-micron diameter
244
+ uncoated commercial silica optical fibers touching each other as
245
+ shown in Fig. 1(a). The ends of Fiber 1 and Fiber 2 were bent and
246
+ translated to arrive at the required profile of these fibers near their
247
+ coupling region illustrated in Fig. 1(b). The fibers used were either
248
+ originally straight or preliminary softened in a flame and bent
249
+ permanently. As described in the previous section, WGMs were
250
+ launched into Fiber 1 by a transversely oriented microfiber
251
+ connected to the OSA. If the separation between Fiber 1 and Fiber 2
252
+ is small enough, WGMs penetrate from Fiber 1 into Fiber 2.
253
+ In the simplest configuration considered in this Section, Fiber 1
254
+ was straight, and coplanar Fiber 2 was bent. The fibers were put in
255
+ contact and then slightly pushed towards each other to increase the
256
+ coupling region. The photograph of the fiber configuration used in
257
+ this experiment is shown in Fig. 3(a). From this picture, we estimated
258
+ the curvature radius of the bent fiber as 𝑅~30 mm (see further
259
+ discussion of the fiber profile in Section 4). Fig. 3(b) shows the
260
+ spectrogram of the configured structure measured along the 3.5 nm
261
+ bandwidth within the 700 µm axial length of Fiber 1. At the edges of
262
+ the scanned region, the interfiber coupling is negligible. In these
263
+ regions, CWs do not noticeably change with distance 𝑧 and, thus,
264
+ correspond to Fiber 1 only. The arrangement of CWs in these regions
265
+ is similar to that in Fig. 2(b).
266
+
267
+ Fig. 3. (a) Photograph of the side-coupled fibers used in the experiment. The upper fiber is bent with the curvature radius 𝑅~30 mm and the lower
268
+ fiber has the curvature radius greater than 1 m. (b) The spectrogram measured in the vicinity of the coupling region of these fibers. (b1) and (b2)
269
+ Spectrograms showing the magnified sections outlined in the spectrogram (b). (c1) and (c2) Spectrograms of the microresonators numerically
270
+ calculated in the two-mode approximation detailed in the text, which replicate the experimental spectrograms in Figs. (b1) and (b2), respectively.
271
+
272
+ (a)
273
+ R~30 mm
274
+ Experiment
275
+ 3-10
276
+ nsertion
277
+ 1mm
278
+ -15
279
+ (b1)
280
+ 1549.65
281
+ 1549.46
282
+ 1549.48
283
+ 1547.45
284
+ (b)
285
+ (b2)
286
+ Wavelength (um)
287
+ 1549.60
288
+ 1550.00
289
+ -2
290
+ 1547.40
291
+ 1549.55
292
+ (b2)
293
+ -4
294
+ 1549.50
295
+ 1549.50
296
+ -6
297
+ Transmission Power (dB)
298
+ 1549.45
299
+ Wavelength (nm)
300
+ 1549.00
301
+ -8
302
+ 1549.40
303
+ 1547.20
304
+ 1549.35
305
+ 1548.50
306
+ -10
307
+ 1547.15
308
+ 1549.30
309
+ -12
310
+ 1548.00
311
+ 200
312
+ 400
313
+ 600
314
+ 200
315
+ 400
316
+ 600
317
+ -14
318
+ Distancealongfiber(um)
319
+ Distance along fiber (μm)
320
+ 1547.50
321
+ (b1)
322
+ -16
323
+ -18
324
+ 1547.00
325
+ Theory
326
+ -20
327
+ 0
328
+ 200
329
+ 400
330
+ 600
331
+ Distance along fiber (μm)
332
+ (c1)
333
+ (c2)0.4
334
+ 0.4
335
+ -5
336
+ Wavelengthvariation(nm)
337
+ Wavelength variation (nm)
338
+ 0.3
339
+ 0.3
340
+ 0.2
341
+ 0.2
342
+ -15
343
+ 0.1
344
+ 0.1
345
+ 0.0
346
+ 0.0
347
+ -20
348
+ 0
349
+ 200
350
+ 400
351
+ 600
352
+ 0
353
+ 200
354
+ 400
355
+ 600
356
+ Distancealongfiber(um)
357
+ Distance along fiber (μm) The effect of coupling shows up in the central region of the
358
+ spectrogram in Fig. 3(b). In this region, different CWs exhibit
359
+ different positive and negative variations along the axial length 𝑧.
360
+ The exemplary regions of this spectrogram named (b1) and (b2) are
361
+ magnified in Figs. 3(b1) and 3(b2), respectively. It is seen that, as
362
+ expected, in contrast to negative variations, positive CW variations
363
+ lead to the WGM confinement and the creation of microresonators.
364
+ Our estimates illustrated in the inset of Fig. 3(b2) show that the Q-
365
+ factor of the created microresonator (which measurement was
366
+ limited by the 1.3 pm resolution of the OSA used) exceeds 10�. The
367
+ observed CW variations in Figs. 3(b1) and (b2) can be explained by
368
+ the theory described below.
369
+ 4. Basic theory
370
+ We assume that the fiber bending is small enough so that the
371
+ propagation of light along the axial direction of side-coupled
372
+ fibers (Fig. 1(b)) can be considered as propagation along a
373
+ single waveguide with asymmetric cross-section including
374
+ both fibers. The wavelengths of slow WGMs are close to the
375
+ CWs 𝜆��𝑧� of this compound waveguide. To determine the
376
+ complex-valued CWs 𝜆��𝑧�, we introduce the original CWs
377
+ 𝜆��� � �
378
+ �𝛾��� and 𝜆��� � �
379
+ �𝛾��� of unbent Fiber 1 and Fiber 2
380
+ with the imaginary parts determined primarily by material
381
+ losses and scattering of light at the fiber surface. We assume
382
+ that there are 𝑁� and 𝑁� cutoff wavelengths in Fibers 1 and
383
+ Fiber 2, respectively, which contribute to the resonant
384
+ transmission, so that 𝑛� � 1,2, … , 𝑁�, 𝑗 � 1,2. We refer to the
385
+ integers 𝑛, 𝑛� and 𝑛� as to the transverse quantum numbers.
386
+ Variation of 𝜆��𝑧� is caused by bending of fibers [27] and, in
387
+ our case, primarily by their coupling. In the absence of the
388
+ input-output fiber, the CWs of our system, 𝜆 � 𝜆��𝑧�, 𝑛 �
389
+ 1,2, … , 𝑁� � 𝑁�, are determined as the roots of the
390
+ determinant:
391
+
392
+
393
+
394
+ det
395
+ ( )
396
+ 0
397
+ z
398
+  
399
+
400
+ I
401
+ Ξ
402
+ (1)
403
+
404
+ Here 𝐈 is the unitary �𝑁� � 𝑁�� � �𝑁� � 𝑁�� matrix and
405
+ matrix
406
+
407
+ 1
408
+ 1
409
+ 12
410
+
411
+ 12
412
+ 2
413
+ 2
414
+ ( )
415
+ ( )
416
+ ( )
417
+ ( )
418
+ ( )
419
+ z
420
+ z
421
+ z
422
+ z
423
+ z
424
+
425
+
426
+
427
+  
428
+
429
+
430
+
431
+
432
+ Λ
433
+ Δ
434
+ Δ
435
+ Ξ
436
+ Δ
437
+ Λ
438
+ Δ
439
+ . (2)
440
+
441
+ The submatrices in Eq. (2) determine the original CWs of
442
+ Fiber 1 and Fiber 2, 𝚲� � �𝜆��� � �
443
+ �𝛾����, couplings inside
444
+ each of the fiber caused by bending, 𝚫��𝑧� � �δ����
445
+ ���
446
+ �𝑧��, and
447
+ interfiber
448
+ couplings
449
+ 𝚫���𝑧� � �δ����
450
+ ���� �𝑧��
451
+ ,
452
+ 𝑚�, 𝑛� �
453
+ 1,2, … 𝑁�.
454
+ As in SNAP [24], dramatically small nanometer and sub-
455
+ nanometer scale variations of CWs 𝜆��𝑧� along the compound
456
+ fiber waveguide can localize WGMs and induce an optical
457
+ microresonator having eigenwavelengths 𝜆�� with axial
458
+ quantum numbers 𝑞. Due to the smooth and small CW variation
459
+ and proximity of the localized WGM wavelengths 𝜆�� to
460
+ 𝜆��𝑧�, the corresponding eigenmode can be presented as
461
+ 𝐸���𝑥, 𝑦, 𝑧� � Ψ���𝑧�Ω��𝑥, 𝑦, 𝑧� where the transverse WGM
462
+ distribution Ω��𝑥, 𝑦, 𝑧� is calculated at the CW 𝜆��𝑧� and
463
+ depends on 𝑧 parametrically slow [32], and function Ψ���𝑧�
464
+ determines the axial dependence of the microresonator
465
+ eigenmode amplitude and satisfies the one-dimensional wave
466
+ equation [24]
467
+
468
+ 2
469
+ 3/2
470
+ 2
471
+ 2
472
+ 3/ 2
473
+ 2
474
+ ( , )
475
+ 0,
476
+ ( , )
477
+ ( )
478
+ .
479
+ n
480
+ r
481
+ n
482
+ n
483
+ n
484
+ n
485
+ n
486
+ d
487
+ n
488
+ z
489
+ z
490
+ z
491
+ dz
492
+
493
+
494
+
495
+
496
+
497
+
498
+
499
+
500
+  
501
+  
502
+
503
+
504
+ (3)
505
+
506
+ where 𝑛� is the refractive index of the fibers.
507
+ The coupling parameters 𝜅���𝑧�between WGM 𝐸���𝑥, 𝑦, 𝑧�
508
+ and the input-output wave in the microfiber is determined by
509
+ their overlap integral. Commonly, the microfiber diameter is
510
+ much smaller than the characteristic axial variation length of
511
+ 𝐸���𝑥, 𝑦, 𝑧�. For this reason, similar to the analogous
512
+ approximation in the SNAP platform [24, 33], the coupling
513
+ parameters 𝜅���𝑧� are proportional to the values of
514
+ 𝐸���𝑥, 𝑦, 𝑧� at the axial coordinate 𝑧 of the input-output
515
+ microfiber. Then, calculations based on the Mahaux-
516
+ Weidenmüller theory [34-36] presented in Supplementary
517
+ Material allowed us to express the transmission power 𝑃�𝜆, 𝑧�
518
+ through the input-output microfiber coupled to the considered
519
+ fiber configuration (Fig. 1(b)) as
520
+
521
+ 1
522
+ 2
523
+ 1
524
+ 2
525
+ 2
526
+ *
527
+ 1
528
+ 1
529
+ 1
530
+ ( )
531
+ ( , , )
532
+ ( , )
533
+ 1
534
+ ( )
535
+ ( , , )
536
+ N
537
+ N
538
+ n
539
+ n
540
+ n
541
+ N
542
+ N
543
+ n
544
+ n
545
+ n
546
+ D z G z z
547
+ P z
548
+ D z G z z
549
+
550
+
551
+
552
+
553
+
554
+
555
+
556
+
557
+
558
+
559
+
560
+
561
+ . (4)
562
+
563
+ Here 𝐺��𝑧�, 𝑧�, 𝜆) is the Green’s function of Eq. (3). Eq. (4)
564
+ generalizes the expression for the transmission power
565
+ previously derived in Ref. [24]. As shown below, functions
566
+ 𝐷��𝑧� can be expressed through and have characteristic values
567
+ similar
568
+ to
569
+ the
570
+ coupling
571
+ D-parameters
572
+ which
573
+ were
574
+ experimentally measured previously and typically have the real
575
+ and imaginary parts ~ 0.01 µm-1 [24, 33]. Close to the resonance
576
+ condition, 𝜆 � 𝜆��, for sufficiently small losses and coupling, and
577
+ separated CWs 𝜆��𝑧�, only one Green’s function with number 𝑛
578
+ contributes to the sums in Eq. (4). Then, Eq. (4) coincides with that
579
+ previously derived in Ref. [24]. However, generally, the
580
+ contribution of more than one term to the sums in Eq. (4) may be
581
+ significant.
582
+ Before the detailed description of the spectrograms in Figs.
583
+ 2(b) and 3(b), we note that the transmission power plots in
584
+ these figures characterize the CWs of the coupled fiber system
585
+ determined by Eq. (1) viewed by the input-output microfiber
586
+ and, subsequently, OSA. Therefore, the CWs of Fiber 2, which are
587
+ the solutions of Eq. (1) but uncoupled from Fiber 1 cannot be
588
+ seen by the OSA. On the other hand, the number of CWs which
589
+ can show up in the coupling region can be as many as 𝑁� � 𝑁�,
590
+ i.e., significantly greater than the number ��� of visible
591
+ uncoupled CWs of Fiber 1 (see Fig. 2(b) as an example).
592
+ To clarify the effect of coupling between WGMs in adjacent
593
+ fibers, we consider the two-mode approximation, 𝑁� � 𝑁� � 1,
594
+ assuming that the wavelength 𝜆 of the input light is close to an
595
+
596
+ unperturbed single WGM CW 𝜆�� � �
597
+ �𝛾 of Fiber 1 and a single CW
598
+ 𝜆�� � �
599
+ �𝛾 of Fiber 2 having the same imaginary part. Consequently,
600
+ in Fig. 1(b) we now set 𝑛� � 𝑛� � 1. We neglect the effect of the
601
+ CW variation due to the fiber bending [27], which is usually smaller
602
+ than the effect of fiber coupling, setting 𝛿��
603
+ ��� � 0. Then, the CWs
604
+ 𝜆��𝑧� and 𝜆��𝑧� of the compound fiber are found from Eq. (1) as
605
+
606
+
607
+
608
+
609
+
610
+
611
+
612
+ 2
613
+ 2
614
+ (12)
615
+ 1,2
616
+ 11
617
+ 21
618
+ 11
619
+ 21
620
+ 11
621
+ 1
622
+ 1
623
+ ( )
624
+ ( )
625
+ 2
626
+ 4
627
+ z
628
+ i
629
+ z
630
+
631
+
632
+
633
+
634
+
635
+
636
+
637
+
638
+
639
+
640
+
641
+
642
+
643
+ (5)
644
+
645
+ The dependence on the transverse coordinates 𝑥 and 𝑦 (Fig. 1(b))
646
+ of the compound WGM corresponding to CWs 𝜆��𝑧� can be
647
+ calculated as follows. We introduce the unperturbed WGMs in Fiber
648
+ 1 and 2 (considered unbent and uncoupled) calculated at their CWs
649
+ 𝜆�� and 𝜆�� as Ω�
650
+ ����𝑥, 𝑦� and Ω�
651
+ ����𝑥, 𝑦�. Then, in the two-mode
652
+ approximation, the compound modes generated by weak coupling
653
+ of modes Ω�
654
+ ����𝑥, 𝑦� and Ω�
655
+ ����𝑥, 𝑦� are determined as [37]
656
+
657
+ (1)
658
+ (2)
659
+ 1
660
+ 1
661
+ 1
662
+ 2
663
+ 2
664
+ (1)
665
+ (2)
666
+ 2
667
+ 1
668
+ 1
669
+ 2
670
+ 2
671
+ (12)
672
+ 11
673
+ 11
674
+ 21
675
+ 1
676
+ ( )
677
+ ( , , )
678
+ ( , )
679
+ ( , ),
680
+ 1
681
+ ( )
682
+ 1
683
+ ( )
684
+ ( )
685
+ 1
686
+ ( , , )
687
+ ( , )
688
+ ( , ),
689
+ 1
690
+ ( )
691
+ 1
692
+ ( )
693
+ ( )
694
+ ( )
695
+ .
696
+ z
697
+ x y z
698
+ x y
699
+ x y
700
+ z
701
+ z
702
+ z
703
+ x y z
704
+ x y
705
+ x y
706
+ z
707
+ z
708
+ z
709
+ z
710
+
711
+
712
+
713
+
714
+
715
+
716
+
717
+
718
+
719
+
720
+
721
+
722
+
723
+
724
+
725
+
726
+
727
+
728
+  
729
+
730
+
731
+
732
+
733
+
734
+
735
+
736
+
737
+
738
+
739
+
740
+
741
+
742
+
743
+ (6)
744
+
745
+ Consequently, the coupling parameters to the microfiber entering
746
+ Eq. (4) at coordinate 𝑧 are
747
+
748
+ 1
749
+ 2
750
+ 2
751
+ 2
752
+ ( )
753
+ ( )
754
+ ,
755
+ ( )
756
+ ,
757
+ 1
758
+ ( )
759
+ 1
760
+ ( )
761
+ D
762
+ D
763
+ z
764
+ D z
765
+ D z
766
+ z
767
+ z
768
+
769
+
770
+
771
+
772
+  
773
+
774
+
775
+
776
+
777
+
778
+ (7)
779
+
780
+ where 𝐷 is the z-independent coupling parameter between the input-
781
+ output microfiber and Fiber 1 [24, 33].
782
+ To map the bent fiber axial profile ℎ�𝑧� to the CW envelope
783
+ profiles of the induced microresonators, we have to determine the
784
+ relation between ℎ�𝑧� and coupling coefficient 𝛿��
785
+ �����𝑧�. Similar
786
+ to calculations in Refs. [38, 39], for the smooth and small ℎ�𝑧�
787
+ considered here, we find
788
+
789
+
790
+
791
+ 1/2
792
+ (12)
793
+ 2
794
+ 11
795
+ 0
796
+ 2
797
+ ( )
798
+ exp
799
+ 1
800
+ ( )
801
+ r
802
+ z
803
+ n
804
+ h z
805
+
806
+
807
+
808
+
809
+
810
+
811
+
812
+
813
+
814
+
815
+
816
+
817
+
818
+ , (8)
819
+
820
+ where 𝛿� is 𝑧-independent. Assuming the simplest profile of the
821
+ bent fiber having the curvature radius 𝑅 as
822
+
823
+ ℎ�𝑧� � 𝑧�/2𝑅 (9)
824
+
825
+ for silica fibers with 𝑛� �1.44, we estimate the FWHM of 𝛿��
826
+ �����𝑧�
827
+ as 𝑧����~0.5�𝜆𝑅��/�. At 𝜆~1.55 µm and 𝑅~30 mm of our
828
+ experiment, we have 𝑧����~100 µm. From Eqs. (5) and (8), we
829
+ find that the FWHM of the CW, depending on the value of 𝜆�� �
830
+ 𝜆��, is between 𝑧���� and 2𝑧���� which is only in qualitative
831
+ agreement with the microresonator FWHM 𝑧����~ 250 µm
832
+ found from experimental data in Figs. 3(b1) and (b2).
833
+ The results of our numerical modeling in the two-mode
834
+ approximation considered based on Eqs. (3)-(9) are shown in Figs.
835
+ 3(c1) and 3(c2). To fit the experimental data, we set the average CW
836
+ 0.5(𝜆�� � 𝜆��) = 1.55 µm, the CW difference 𝜆�� � 𝜆�� � 0.05
837
+ nm in Fig. 3(c1) and 𝜆�� � 𝜆�� � �0.05 nm in Fig. 3(c2),
838
+ coupling parameter 𝐷 � �0.01 � 0.01𝑖 µm-1 [24, 33], Q-factor
839
+ 𝑄 � 10�, the microresonator FWHM 𝑧����~ 250 µm and its
840
+ spectral height ~ 0.15 nm, similar to these values found from Figs.
841
+ 2(b1) and (b2).
842
+ The experimental spectrograms in Fig. 3(b1) and (b2) and
843
+ theoretical spectrograms in Figs. 3(c1) and (c2) look nicely similar.
844
+ However, important differences between them should be noted.
845
+ From Eqs. (8) and (9), the FWHM value 𝑧����~ 250 µm
846
+ corresponds to the Fiber 2 curvature radius 𝑅~66 mm, which is
847
+ twice as large as that measured from the fiber image shown in Fig.
848
+ 3(a). We suggest that the difference is caused by the deviation of the
849
+ shape of Fiber 2 from parabolic in the coupling region as well as by
850
+ the fiber misalignment. The additional deformation of fibers may be
851
+ induced by their electrostatic attraction and pressuring, which are not
852
+ visible in Fig. 3(a). Our suggestion is confirmed by the experimental
853
+ profiles of the induced microresonator envelopes and CW shapes in
854
+ Figs. 3(b1) and (b2) which, as compared to those in the theoretical
855
+ spectrograms in Figs. 3(c1) and (c2), have larger side slopes and are
856
+ flatter in the middle. Next, we notice that, in the theoretical
857
+ spectrograms, the CW wavelength profiles are more mirror-
858
+ symmetric to the microresonator envelopes with respect to the
859
+ horizontal line (following Eq. (5)), while, in the experimental
860
+ spectrograms, the lower CW profiles are shallower than the
861
+ microresonator envelopes. We suggest that this deviation can be
862
+ eliminated by taking into account the coupling with other WGMs
863
+ ignored in the two-mode approximation considered.
864
+ 5. Tunability
865
+ Bending and translating the tails of Fiber 1 and Fiber 2 side-
866
+ coupled to each other as illustrated in Fig. 1 allowed us to tune
867
+ the dimensions of the fiber coupling region and thereby tune the
868
+ dimensions of created microresonators. As in the previous
869
+ sections, in our experiments we used 125 µm optical fibers. We
870
+ investigated the cases of the smallest microresonators
871
+ containing a few wavelength eigenvalues and having the
872
+ characteristic axial dimensions of hundred microns (Figs. 5(a1)-
873
+ (a4)), as well as larger microresonators with dimensions of
874
+ several hundred microns (Figs. 5(b1)-(b4) and (c1)-(c3)) and
875
+ the largest microresonator having the axial length of 5
876
+ millimeters (Fig. 4(d)).
877
+ Considering the smallest microresonators, we monitored the
878
+ process of their creation. Side-coupling of a straight Fiber 1 and
879
+ Fiber 2 bent with a sufficiently small curvature radius of ~ 1 mm
880
+ introduced small perturbation in CWs shown in the
881
+ spectrogram in Fig. 4(a1). Increasing the fiber radius further, we
882
+ arrived at the microresonator with a single eigenwavelength
883
+ (Fig. 4(a2)). The inset inside the Fig. 4(a2) spectrogram, which
884
+ magnifies the region near this eigenwavelength, shows that the
885
+ axial dimension of the corresponding eigenmode is ~ 200 µm.
886
+ Remarkably, except for the axial dimension of localized WGMs
887
+ with
888
+ uniform
889
+ magnitude
890
+ in
891
+ specially
892
+ designed
893
+ bat
894
+
895
+ microresonators [39, 40], this dimension (which expansion is
896
+ critical, e.g., for QED applications [41]) is the record large
897
+ characteristic
898
+ WGM
899
+ dimension
900
+ demonstrated
901
+ in
902
+ microresonators to date. The measured Q-factor of this
903
+ microresonator (limited by the 1.3 pm resolution of the OSA
904
+ used) was slightly greater than 10�.
905
+ Larger bending radii of Fiber 2 having the order of 10 mm led to
906
+ the creation of microresonators with millimeter-order axial
907
+ dimensions having the spectrograms shown in Figs. 5(b1)-(b4) and
908
+ (c1)-(c3). The close to parabolic shape of these microresonators
909
+ suggests that they can be used, e.g., as tunable optical frequency
910
+ comb generators [42]. We note that the behavior of the CWs and
911
+ microresonators envelopes in most of these spectrograms cannot be
912
+ accurately described by the two-mode approximations of Section 4.
913
+ Of particular interest is the spectrogram shown in Fig. 4(c2). At first
914
+ sight, the envelop of the microresonator in this spectrogram is the
915
+ continuation of the CW of Fiber 1 (compare with Figs. 3(b1) and
916
+ (c1)). Unexpectedly, the axial WGM localization in this
917
+ microresonator (caused by the WGM reflection from the CW-
918
+ generated turning points [24]) sharply dissolves inside the
919
+ microresonator area.
920
+
921
+ Fig. 4. Tunability of microresonators. (a1)-(a4) Spectrograms of induced microresonators for small curvature radius of Fiber 2 ~ 1 mm. (b1)-(b4) and
922
+ (c1)-(c3) spectrograms of induced microresonators for a lager radius of Fiber 2 ~ 10 mm. (d) Spectrogram of a 5 mm long microresonator induced by
923
+ touching straight Fiber 1 and Fiber 2 which was preliminary permanently bent at the ends as shown in the inset.
924
+
925
+ (a1)
926
+ (a2)
927
+ (a3)
928
+ (a4)
929
+ R=1.2mm
930
+ R=1.5mm
931
+ R=1.6mm
932
+ 1.7mm
933
+ 1546.80
934
+ 1546.80
935
+ 1546.80
936
+ 1546.80
937
+ ap
938
+ (w
939
+ (wu
940
+ 1546.6
941
+ (wu
942
+ (wu
943
+ 1546.70
944
+ 1546.70
945
+ -5
946
+ -45
947
+ -5
948
+ 1546.60
949
+ 1546.60
950
+ -6
951
+ 1546.60
952
+ 1546.60
953
+ 0200400600
954
+ 0200400600
955
+ 0200400600
956
+ 0200400600
957
+ Distancealongfiber(μm)
958
+ Distancealongfiber(um)
959
+ Distancealongfiber(μm)
960
+ Distancealongfiber(um)
961
+ (b1)
962
+ (b2)
963
+ (b3)
964
+ (b4)
965
+ R=6.1mm
966
+ R=7.7mm
967
+ R=8.1mm
968
+ R=16.3mm
969
+ 1551.90
970
+ 1551.90
971
+ 1551.90
972
+ 1551.90
973
+ 10
974
+ 10
975
+ 1551.80
976
+ -20
977
+ 1551.80
978
+ 32
979
+ 1551.70
980
+ 1551.70
981
+ 1551.70
982
+ 1551.70
983
+ lavele
984
+ 4
985
+ -5
986
+ 1551.60
987
+ 1551.60
988
+ ≤1551.60
989
+ 1551.60
990
+ -5
991
+ 1551.50
992
+ 1551.50
993
+ 1551.50
994
+ 200400600800
995
+ 200400600800
996
+ 200400600800
997
+ 1551.50
998
+ 0
999
+ 200400600800
1000
+ Distancealongfiber(um)
1001
+ Distancealongfibor(μm)
1002
+ Distancealongfiber(um)
1003
+ Distancealong fiber(um)
1004
+ (c1)
1005
+ (c2)
1006
+ (c3)
1007
+ R=18mm
1008
+ R~30mm
1009
+ R~30mm
1010
+ 1551.80
1011
+ 1551.80
1012
+ Transmission Power (dB)
1013
+ 1551.80
1014
+ 2
1015
+ (dB)
1016
+ -2
1017
+ 1551.70
1018
+ 4
1019
+ -4
1020
+ nbu
1021
+ .6
1022
+ 6
1023
+ 1551.60
1024
+ 1551.60
1025
+ AeM
1026
+ -8
1027
+ 8
1028
+ 1551.50
1029
+ 1551.50
1030
+ 1551.50
1031
+ -10
1032
+ -10
1033
+ -10
1034
+ 0
1035
+ 400
1036
+ 800
1037
+ 1200
1038
+ 1600
1039
+ 400
1040
+ 800
1041
+ 1200
1042
+ 1600
1043
+ 0
1044
+ 400
1045
+ 800
1046
+ 1200
1047
+ 1600
1048
+ Distancealong fiber(um)
1049
+ Distancealongfiber(um)
1050
+ Distancealongfiber (um)
1051
+ (d)
1052
+ 1548.60
1053
+ length
1054
+ 4
1055
+ 6
1056
+ -10
1057
+ 1548.30
1058
+ 0
1059
+ 1000
1060
+ 2000
1061
+ 3000
1062
+ 4000
1063
+ 5000
1064
+ 6000
1065
+ Distancealong fiber(um)To create longer microresonators, we, first, permanently bent
1066
+ the tails of Fiber 2 as illustrated in the inset of Fig. 4(d). This
1067
+ allowed us to arrive at an arbitrarily large curvature radius of
1068
+ this fiber including its straight shape between the bent tails. As
1069
+ an example, Fig. 4(d) shows the spectrogram of a 5 mm long
1070
+ microresonator. Though the eigenwavelength width of this
1071
+ microresonator is greater than its free spectral range, we
1072
+ suggest that, in contrast to the lossy microresonators induced by
1073
+ side-coupled cleaved straight fibers demonstrated in Section 2,
1074
+ its Q-factor is similar to that of the smaller microresonators
1075
+ considered in this section and Section 3.
1076
+ 6. Discussion
1077
+ The effect of induction of high Q-factor WGM tunable optical
1078
+ microresonators in side-coupled optical fibers discovered in this
1079
+ paper enables a range of exciting generalizations and applications.
1080
+ Further extension of tuning flexibility can be achieved by enabling
1081
+ different boundary conditions at the fiber tails (Fig. 1(a)), different
1082
+ interfiber touching stresses, and different preliminary permanent
1083
+ fiber bending.
1084
+ Configurations of fibers, which are potentially attractive for future
1085
+ research and applications, are illustrated in Fig. 5. Fig. 5(a) shows a
1086
+ way to create long microresonators alternative to the method
1087
+ utilizing fibers with permanently bent tails illustrated in Fig. 4(d). In
1088
+ the configuration of Fig. 5(a), the length of the induced
1089
+ microresonator increases as the curvature radii of touching fibers
1090
+ approach each other. Provided that the variation of the fiber radii can
1091
+ be performed so that the parabolicity of the induced microresonators
1092
+ was maintained, the configuration of Fig. 5(a) can serve for the
1093
+ generation of the optical frequency combs with a tunable repetition
1094
+ rate.
1095
+
1096
+ Fig. 5. (a) Bent fibers with increased coupling region. (b) Bent fibers
1097
+ with increased coupling region and abrupt side of the induced
1098
+ microresonator. (c) A bottle microresonator side-coupled to a fiber. (d)
1099
+ Side coupled straight fibers with tapered facets forming a rectangular
1100
+ microresonator. (e) Two straight fibers with tapered facets coupled to
1101
+ the third straight fiber forming a rectangular microresonator. (f)
1102
+ Twisted side-coupled fibers. (g) A microcapillary fiber filled with liquid
1103
+ and side coupled to a bent fiber. (h) Three straight coupled fibers.
1104
+ In Fig. 5(b), the lower fiber is terminated with a short taper, which
1105
+ can be introduced using, e.g., a CO2 laser. Simple estimates show
1106
+ that a taper with a characteristic length of 100 µm at the end of a 125
1107
+ µm diameter optical fiber creates an abrupt CW barrier with a slope
1108
+ of ~ 100 nm/µm at 1.5 µm wavelength. The steepness of the slope
1109
+ of this barrier (critical for impedance matching of light from the
1110
+ input-output microfiber [43]) is 100 times greater than that
1111
+ demonstrated in Ref. [44] with the femtosecond laser inscription.
1112
+ The configuration shown in Fig. 5(b) can be used for the creation of
1113
+ miniature dispersionless tunable optical delay lines provided that the
1114
+ shape of the induced microresonator is kept semi-parabolic in the
1115
+ process of tuning [43].
1116
+ Experimental investigation and development of the theory of
1117
+ WGMs in a microresonator side-coupled to an optical fiber is of
1118
+ particular interest. Fig. 5(c) illustrates the side coupling of a fiber and
1119
+ a bottle microresonator. While the fiber is open-ended, coupling of
1120
+ the bottle microresonator to the straight fiber can cause the
1121
+ localization of light in the fiber, similar to the coupling between bent
1122
+ optical fibers considered above. The configuration shown in Fig. 5(c)
1123
+ suggests a way of tuning the microresonator eigenwavelengths.
1124
+ The fiber configuration shown in Fig. 5(d) is similar to two
1125
+ straight side-coupled fibers considered in Section 2. To improve the
1126
+ Q-factor of the microresonator induced along the coupling region,
1127
+ the cleaved ends of fibers shown in Fig. 2(a) are modified by the
1128
+ tapered ends. The configuration of fibers shown in Fig. 5(e)
1129
+ illustrates an alternative way to create tunable microresonators when
1130
+ the position of both their sides can be tuned. The rectangular
1131
+ microresonators induced in both configurations can be used for the
1132
+ creation of tunable delay lines which, as shown in Ref. [31], can be
1133
+ dispersionless with a good accuracy.
1134
+ The coupling of twisted optical fibers illustrated in Fig. 5(f) is
1135
+ interesting to investigate both theoretically and experimentally. In
1136
+ the cylindrical coordinates �𝑧, 𝜌, 𝜑� of one of the fibers, the curve
1137
+ along which the fibers touch each other corresponds to the azimuthal
1138
+ angle 𝜑 � 𝜑� � 𝛼𝑧 , where 𝛼 is the twisting coefficient. The
1139
+ corresponding value of the WGM field is proportional to
1140
+ exp �𝑖𝛽𝑧 � 𝑖𝑚�𝜑� � 𝛼𝑧�� where 𝛽 is the propagation constant.
1141
+ From this expression, a WGM at CW corresponding to 𝛽 � 0 is
1142
+ seen by another fiber as a mode with nonzero propagation constant.
1143
+ Thus, in contrast to the untwisted fibers, coupling between the side-
1144
+ coupled twisted fibers is essentially three dimensional.
1145
+ Fig. 5(g) shows a microcapillary fiber filled with liquid and side-
1146
+ coupled to a bent fiber. For the microcapillary with sufficiently thin
1147
+ walls, a microresonator induced inside it by the side-coupled fiber
1148
+ performs nonlocal sensing of liquid [45]. In Refs. [46] and [47], such
1149
+ microresonators were introduced with the CO2 laser and slow
1150
+ cooking methods. Fig. 5(g) suggests the simplest approach for the
1151
+ realization of nonlocal microfluidic sensing.
1152
+ Fig. 5(h) illustrates three straight side-coupled fibers. In contrast
1153
+ to two coupled fibers, WGMs launched into this configuration will
1154
+ propagate into both azimuthal direction and, in particular, into the
1155
+ positive and negative directions of the input-output microfiber with
1156
+ approximately the same amplitudes. The channel formed between
1157
+ these fibers can be used for gas and microfluidic sensing. Unlike the
1158
+ microcapillary illustrated in Fig. 5(g), no ultrathin wall enabling the
1159
+ WGM sensing of the internal channel is required in this case.
1160
+ While the model of two coupled CWs developed here
1161
+ qualitatively explains some characteristic features of the
1162
+
1163
+ (a)
1164
+ (f)
1165
+ (b)
1166
+ (g)
1167
+ (c)
1168
+ (d)
1169
+ (h)
1170
+ (e)experimentally measured spectrograms, the complete explanation
1171
+ and quantitative fitting of the experimental data should include the
1172
+ effect of several CWs and be based on the further development of
1173
+ the coupled wave theory. The future theory should also allow us
1174
+ to express the fiber profiles and deformation in the region of
1175
+ coupling through the values of forces and moments applied to
1176
+ the fiber tails (Fig.1(a)) including the effect of electrostatic fiber
1177
+ attraction.
1178
+ We suggest that the fixed submicron-wide gaps between
1179
+ coupled fibers and input-output microfiber, rather than their
1180
+ direct contact considered here, will allow us to demonstrate the
1181
+ proposed microresonators with the Q-factor exceeding 108 [8].
1182
+ While such large Q-factors are not required for the realization of
1183
+ tunable delay lines [43], signal processors [25], and microlasers
1184
+ [19-21], they may be important for the realization of frequency
1185
+ comb generators with tunable repetition rate [15, 16, 42], as
1186
+ well as for the cavity QED [8, 11,12] and optomechanical
1187
+ applications [13, 14].
1188
+
1189
+
1190
+
1191
+ Supplementary material
1192
+ Expression for the transmission power
1193
+ We introduce the discrete eigenwavelengths of the microresonator
1194
+ in the compound fiber system, 𝜆� � �
1195
+ �𝛾� , 𝑚 � 1,2, … , 𝑀 and
1196
+ coupling
1197
+ coefficients 𝜅��𝑧� between
1198
+ the
1199
+ corresponding
1200
+ eigenmodes and the input-output microfiber positioned at axial
1201
+ coordinate 𝑧. We calculate the transmission power 𝑃�𝜆, 𝑧� of our
1202
+ system by applying the Mahaux-Weidenmüller formula [34-36]:
1203
+
1204
+
1205
+
1206
+ 2
1207
+ 1
1208
+
1209
+
1210
+ 2
1211
+ ( , )
1212
+ 1
1213
+ ( , ) ,
1214
+ ( , )
1215
+ ( )
1216
+ ( )
1217
+ ( )
1218
+ ( )
1219
+ ( )
1220
+ i
1221
+ P
1222
+ z
1223
+ iT
1224
+ z
1225
+ T
1226
+ z
1227
+ z
1228
+ z
1229
+ z
1230
+ z
1231
+
1232
+
1233
+
1234
+
1235
+
1236
+
1237
+
1238
+
1239
+
1240
+ Κ
1241
+ Δ
1242
+ Κ
1243
+ Κ
1244
+ Κ
1245
+ , (S1)
1246
+
1247
+ where
1248
+
1249
+ 1
1250
+ 2
1251
+
1252
+ 2
1253
+ 1
1254
+ 1
1255
+ 2
1256
+ 1
1257
+ 1
1258
+ 2
1259
+ 2
1260
+ ( )
1261
+ ( )
1262
+ ( )
1263
+ ( )
1264
+ ( )
1265
+ ( ) ,
1266
+ ( )
1267
+ ,
1268
+ ...
1269
+ ( )
1270
+ 0
1271
+ ...
1272
+ 0
1273
+ 0
1274
+ ...
1275
+ 0
1276
+ ( )
1277
+ .
1278
+ ...
1279
+ ...
1280
+ ...
1281
+ ...
1282
+ 0
1283
+ 0
1284
+ ...
1285
+ i
1286
+ M
1287
+ i
1288
+ i
1289
+ i
1290
+ M
1291
+ M
1292
+ z
1293
+ z
1294
+ z
1295
+ z
1296
+ z
1297
+ z
1298
+ z
1299
+
1300
+
1301
+
1302
+
1303
+
1304
+
1305
+
1306
+
1307
+
1308
+
1309
+
1310
+
1311
+
1312
+
1313
+
1314
+
1315
+
1316
+
1317
+
1318
+
1319
+
1320
+
1321
+  
1322
+
1323
+
1324
+
1325
+
1326
+
1327
+
1328
+
1329
+
1330
+
1331
+
1332
+
1333
+
1334
+
1335
+
1336
+
1337
+
1338
+
1339
+  
1340
+
1341
+
1342
+
1343
+
1344
+
1345
+
1346
+
1347
+
1348
+
1349
+ Θ
1350
+ Δ
1351
+ Κ
1352
+ Κ
1353
+ Κ
1354
+ Δ
1355
+ (S2)
1356
+
1357
+ It is assumed in Eq. (S1) that the coupling to the input-output
1358
+ waveguide
1359
+ does
1360
+ not
1361
+ introduce
1362
+ the
1363
+ shifts
1364
+ of
1365
+ the
1366
+ eigenwavelengths [35] which will be added later. We simplify
1367
+ the expression for the transmission power by expanding the
1368
+ inverse matrix in Eq. (S1) as follows:
1369
+
1370
+
1371
+
1372
+   
1373
+
1374
+  
1375
+  
1376
+
1377
+
1378
+  
1379
+ 1
1380
+
1381
+ 2
1382
+ 1
1383
+
1384
+ 1
1385
+ 2
1386
+ 0
1387
+ 1
1388
+
1389
+ 2
1390
+ 2
1391
+ 1
1392
+
1393
+ 1
1394
+
1395
+ 2
1396
+ 1
1397
+ 1
1398
+
1399
+ 1
1400
+
1401
+ 1
1402
+ 2
1403
+ 2
1404
+ 1
1405
+
1406
+ 2
1407
+ 2
1408
+ 2
1409
+ ( )
1410
+ ( )
1411
+ ( )
1412
+ ( )
1413
+ ( )
1414
+ ( )
1415
+ ( )
1416
+ 1
1417
+ ( )
1418
+ ( )
1419
+ ( )
1420
+ ( )
1421
+ ( )
1422
+ ( ) ( )
1423
+ ( )
1424
+ ( )
1425
+ ...
1426
+ ( )
1427
+ ( )
1428
+ ( ) ( )
1429
+ ( )
1430
+ ( )
1431
+ ...
1432
+ ( )
1433
+ ( )
1434
+ 1
1435
+ ( )
1436
+ ( )
1437
+ ( )
1438
+ i
1439
+ n
1440
+ n
1441
+ i
1442
+ n
1443
+ i
1444
+ i
1445
+ n
1446
+ n
1447
+ i
1448
+ n
1449
+ m
1450
+ i
1451
+ i
1452
+ i
1453
+ m
1454
+ m
1455
+ m
1456
+ z
1457
+ z
1458
+ z
1459
+ z
1460
+ z
1461
+ z
1462
+ z
1463
+ z
1464
+ z
1465
+ z
1466
+ z
1467
+ z
1468
+ z
1469
+ z
1470
+ z
1471
+ z
1472
+ z
1473
+
1474
+
1475
+
1476
+
1477
+
1478
+
1479
+
1480
+
1481
+
1482
+
1483
+
1484
+
1485
+
1486
+
1487
+
1488
+
1489
+
1490
+
1491
+
1492
+
1493
+
1494
+
1495
+
1496
+
1497
+
1498
+
1499
+
1500
+
1501
+
1502
+
1503
+
1504
+
1505
+
1506
+
1507
+
1508
+
1509
+
1510
+
1511
+
1512
+
1513
+ 
1514
+
1515
+
1516
+
1517
+
1518
+
1519
+ Δ
1520
+ Κ
1521
+ Κ
1522
+ Δ
1523
+ Κ
1524
+ Κ
1525
+ Δ
1526
+ Δ
1527
+ Κ
1528
+ Κ
1529
+ Δ
1530
+ Κ
1531
+ Κ
1532
+ Δ
1533
+ Κ
1534
+ Κ
1535
+ Δ
1536
+ Κ
1537
+ Κ
1538
+ Δ
1539
+ Κ
1540
+ Κ
1541
+ Δ
1542
+ Δ
1543
+ Κ
1544
+ Κ
1545
+ 1
1546
+ 0
1547
+ 1
1548
+ 1
1549
+
1550
+ 1
1551
+ 2
1552
+ 2
1553
+ 2
1554
+ 1
1555
+ 2
1556
+ ( )
1557
+ ( )
1558
+ ( )
1559
+ ( )
1560
+ 1
1561
+ ( )
1562
+ ( )
1563
+ 1
1564
+ n
1565
+ M
1566
+ n
1567
+ i
1568
+ M
1569
+ m
1570
+ i
1571
+ i
1572
+ m
1573
+ m
1574
+ m
1575
+ z
1576
+ z
1577
+ z
1578
+
1579
+
1580
+
1581
+
1582
+
1583
+
1584
+
1585
+
1586
+
1587
+
1588
+
1589
+
1590
+
1591
+
1592
+
1593
+
1594
+
1595
+
1596
+
1597
+
1598
+
1599
+
1600
+
1601
+
1602
+
1603
+
1604
+
1605
+
1606
+
1607
+
1608
+
1609
+
1610
+
1611
+
1612
+
1613
+
1614
+
1615
+
1616
+
1617
+
1618
+
1619
+
1620
+
1621
+
1622
+
1623
+
1624
+
1625
+
1626
+
1627
+
1628
+
1629
+ Δ
1630
+ Δ
1631
+ Κ
1632
+ Κ
1633
+ Δ
1634
+
1635
+
1636
+ Substituting this expression into Eq. (S1), we find:
1637
+
1638
+ 2
1639
+ 2
1640
+ 2
1641
+ 1
1642
+ 2
1643
+ 2
1644
+ 2
1645
+ 1
1646
+ 2
1647
+ ( )
1648
+ 1
1649
+ ( , )
1650
+ ( )
1651
+ 1
1652
+ M
1653
+ m
1654
+ i
1655
+ i
1656
+ m
1657
+ m
1658
+ m
1659
+ M
1660
+ m
1661
+ i
1662
+ i
1663
+ m
1664
+ m
1665
+ m
1666
+ z
1667
+ P
1668
+ z
1669
+ z
1670
+
1671
+
1672
+
1673
+
1674
+
1675
+
1676
+
1677
+
1678
+
1679
+
1680
+
1681
+
1682
+
1683
+
1684
+
1685
+
1686
+
1687
+
1688
+
1689
+
1690
+ (S4)
1691
+
1692
+ We separate the series of eigenwavelengths 𝜆� � �
1693
+ �𝛾� and
1694
+ coupling coefficients 𝜅��𝑧� by their correspondence to CWs 𝜆��𝑧�
1695
+ entering Eq. (3) of the main text. For this purpose, we rewrite these
1696
+ parameters as 𝜆�� � �
1697
+ �𝛾� and 𝜅���𝑧�, where 𝑞 is the axial quantum
1698
+ number of the eigenmode 𝐸���𝑥, 𝑦, 𝑧� � Ψ���𝑧�Ω��𝑥, 𝑦, 𝑧�.
1699
+ Here Ψ���𝑧� satisfies Eq. (3) and Ω��𝑥, 𝑦, 𝑧� is a parametrically
1700
+ slow function of the axial coordinate 𝑧. Substituting 𝛾� → 𝛾� we
1701
+ assume that the material losses do not depend on the axial quantum
1702
+ number 𝑞. Then, similar to the arguments of Ref. [24] (see Eq. (13)
1703
+ in this reference), the coupling coefficients can be factorized as
1704
+ �𝜅���𝑧��
1705
+ � � 2𝑖𝐷��𝑧��Ψ���𝑧��
1706
+ � . Using the expression for the
1707
+ Green’s function of Eq. (3),
1708
+
1709
+ 2
1710
+ 2
1711
+ ( )
1712
+ ( , , )
1713
+ qn
1714
+ n
1715
+ i
1716
+ q
1717
+ qn
1718
+ n
1719
+ z
1720
+ G z z 
1721
+
1722
+
1723
+
1724
+
1725
+
1726
+
1727
+
1728
+
1729
+ , (S5)
1730
+
1731
+ we rewrite Eq. (S4) as
1732
+
1733
+ 2
1734
+ *
1735
+ 1
1736
+ 1
1737
+ 1
1738
+ ( )
1739
+ ( , , )
1740
+ ( , )
1741
+ 1
1742
+ ( )
1743
+ ( , , )
1744
+ N
1745
+ n
1746
+ n
1747
+ n
1748
+ N
1749
+ n
1750
+ n
1751
+ n
1752
+ D z G z z
1753
+ P
1754
+ z
1755
+ D z G z z
1756
+
1757
+
1758
+
1759
+
1760
+
1761
+
1762
+
1763
+
1764
+
1765
+
1766
+ . (S6)
1767
+
1768
+ To identify the physical meaning of parameters 𝐷��𝑧�, we recall the
1769
+ expression for the transmission power of a SNAP microresonator
1770
+ under the assumption of a single CW contribution (𝑁 � 1) and
1771
+
1772
+ lossless coupling to the input-output microfiber [24]:
1773
+
1774
+
1775
+ 2
1776
+ *
1777
+ 1
1778
+ 1
1779
+ 1
1780
+ 1
1781
+ 1
1782
+ 1
1783
+ ( , , )
1784
+ ( , )
1785
+ 1
1786
+ ( , , )
1787
+ D G z z
1788
+ P
1789
+ z
1790
+ D G z z
1791
+
1792
+
1793
+
1794
+
1795
+
1796
+
1797
+ (S7)
1798
+
1799
+ Here complex parameter 𝐷�, which was experimentally
1800
+ measured and analyzed previously [24, 33], determines the
1801
+ coupling to the input-output microfiber as well as the WGM
1802
+ phase shift due to this coupling. Importantly, while the
1803
+ imaginary part of 𝐷��𝑧� contributes to the widths of the
1804
+ resonances, its real part (not taken into account in the original Eq.
1805
+ (S1)) determines the WGM phase shifts caused by the coupling to
1806
+ the input-output microfiber.
1807
+
1808
+ Funding. The Engineering and Physical Sciences Research Council
1809
+ (EPSRC), grants EP/P006183/1 and EP/W002868/1. Horizon 2020
1810
+ MSCA-ITN-EID grant 814147.
1811
+ Disclosures. The authors declare no conflicts of interest.
1812
+ Data availability. Data underlying the results presented in
1813
+ this paper are not publicly available at this time but may be
1814
+ obtained from the authors upon reasonable request.
1815
+ References
1816
+ 1. W. Bogaerts, D. Pérez, J. Capmany, D. A. B. Miller, J. Poon, D. Englund, F.
1817
+ Morichetti, and A. Melloni, “Programmable photonic circuits,” Nature
1818
+ 586, 207 (2020).
1819
+ 2. S. Y. Siew, B. Li, F. Gao, H. Y. Zheng, W. Zhang, P. Guo, S. W. Xie, A. Song,
1820
+ B. Dong, L. W. Luo, C. Li, X. Luo, and G.-Q. Lo, “Review of Silicon Photonics
1821
+ Technology and Platform Development,” J. Lightwave Technol. 39, 4374
1822
+ (2021).
1823
+ 3. Z. Lu, J. Jhoja, J. Klein, X. Wang, A. Liu, J. Flueckiger, J. Pond, and L.
1824
+ Chrostowski, "Performance prediction for silicon photonics integrated
1825
+ circuits with layout-dependent correlated manufacturing variability,"
1826
+ Opt. Express 25, 9712 (2017).
1827
+ 4. C. Lian, C. Vagionas, T. Alexoudi, N. Pleros, N. Youngblood, and C. Ríos,
1828
+ “Photonic (computational) memories: tunable nanophotonics for data
1829
+ storage and computing,” Nanophotonics 11, 3823 (2022).
1830
+ 5. J. H. Ko, Y. J. Yoo, Y. Lee, H.-H. Jeong, and Y. M. Song, “A review of tunable
1831
+ photonics: Optically active materials and applications from visible to
1832
+ terahertz,” iScience 25, 104727 (2022).
1833
+ 6. A. A. Savchenkov, V. S. Ilchenko, A. B. Matsko and L. Maleki, “Tunable filter
1834
+ based on whispering gallery modes,” Electron. Lett. 39, 389-390 (2003).
1835
+ 7. D. Armani, B. Min, A. Martin, and K. J. Vahala, “Electrical thermo-optic
1836
+ tuning of ultrahigh-Q microtoroid resonators,” Appl. Phys. Lett. 85, 5439-
1837
+ 5441 (2004).
1838
+ 8. M. Pöllinger, D. O'Shea, F. Warken, and A. Rauschenbeutel, “Ultra-high-Q
1839
+ tunable whispering-gallery-mode microresonator,” Phys. Rev. Lett. 103,
1840
+ Art. 053901 (2009).
1841
+ 9. M. Sumetsky, Y. Dulashko, and R. S. Windeler, “Super free spectral range
1842
+ tunable optical microbubble resonator,” Opt. Lett. 35, 1866 (2010).
1843
+ 10. A. Kovach, J. He, P. J. G. Saris, D. Chen, and A. M. Armani, “Optically
1844
+ tunable microresonator using an azobenzene monolayer,” AIP Advances
1845
+ 10, 045117 (2020).
1846
+ 11. J. R. Buck and H. J. Kimble, “Optimal sizes of dielectric microspheres for
1847
+ cavity QED with strong coupling,” Phys. Rev. A 67, Art. 033806 (2003).
1848
+ 12. H. Pfeifer, L. Ratschbacher, J. Gallego, C. Saavedra, A. Faßbender, A. von
1849
+ Haaren, W. Alt, S. Hofferberth, M. Köhl, S. Linden, and D. Meschede,
1850
+ “Achievements and perspectives of optical fiber Fabry–Perot cavities,”
1851
+ Appl. Phys. B 128, 29 (2022).
1852
+ 13. G. Bahl, J. Zehnpfennig, M. Tomes, and T. Carmon, “Stimulated
1853
+ optomechanical excitation of surface acoustic waves in a microdevice,”
1854
+ Nat. Commun. 2, 403 (2011).
1855
+ 14. N. J. Lambert, A. Rueda, F. Sedlmeir, and H. G. L. Schwefel, “Coherent
1856
+ Conversion Between Microwave and Optical Photons—An Overview of
1857
+ Physical Implementations,” Adv. Quantum Technol. 3, 1900077 (2020).
1858
+ 15. H. Bao, A. Cooper, M. Rowley, L. Di Lauro, J. S. T. Gongora, S. T. Chu, B. E.
1859
+ Little, G.-L. Oppo, R. Morandotti, D. J. Moss, B. Wetzel, M. Peccianti, and
1860
+ A. Pasquazi, “Laser cavity-soliton microcombs,” Nat. Photonics 13, 384
1861
+ (2019).
1862
+ 16. L. Chang, S. Liu, and J. E. Bowers, “Integrated optical frequency comb
1863
+ technologies,” Nat. Photon. 16, 95 (2022).
1864
+ 17. X. Wang, L. Zhou, R. Li, J. Xie, L. Lu, K. Wu, and J. Chen, “Continuously
1865
+ tunable ultra-thin silicon waveguide optical delay line,” Optica 4, 507
1866
+ (2017).
1867
+ 18. W. Zhang, J. Yao, and Y. S. Zhao, “Organic Micro/Nanoscale Lasers,” Acc.
1868
+ Chem. Res. 49, 1691 (2016).
1869
+ 19. S. Zhu, L. Shi, B. Xiao, X. Zhang, and X. Fan, “All-Optical Tunable
1870
+ Microlaser Based on an Ultrahigh-Q Erbium-Doped Hybrid Microbottle
1871
+ Cavity,” ACS Photonics 5, 3794 (2018).
1872
+ 20. S. Zhu, B. Xiao, B. Jiang, L. Shi, and X. Zhang, “Tunable Brillouin and
1873
+ Raman
1874
+ microlasers
1875
+ using
1876
+ hybrid
1877
+ microbottle
1878
+ resonators,”
1879
+ Nanophotonics 8, 931 (2019).
1880
+ 21. X. Yang, C. Gong, C. Zhang, Y. Wang, G.-F. Yan, L. Wei, Y.-C. Chen, Y.-J.
1881
+ Rao, and Y. Gong, “Fiber Optofluidic Microlasers: Structures,
1882
+ Characteristics, and Applications,” Laser Photonics Rev. 16, 2100171
1883
+ (2022).
1884
+ 22. L. Greuter, S. Starosielec, D. Najer, A. Ludwig, L. Duempelmann, D.
1885
+ Rohner, and R. J. Warburton, “A small mode volume tunable microcavity:
1886
+ development and characterization,” Appl. Phys. Lett. 105, 121105 (2014).
1887
+ 23. S. Flågan, P. Maletinsky, R. J. Warburton, and D. Riedel, “Microcavity
1888
+ platform for widely tunable optical double resonance,” Optica 9, 1197
1889
+ (2022).
1890
+ 24. M. Sumetsky, “Theory of SNAP devices: basic equations and comparison
1891
+ with the experiment,” Opt. Express 20, 22537 (2012).
1892
+ 25. M. Sumetsky, “Optical bottle microresonators,” Prog. Quantum
1893
+ Electron. 64, 1 (2019).
1894
+ 26. A. Dmitriev, N. Toropov and M. Sumetsky, “Transient reconfigurable
1895
+ subangstrom-precise photonic circuits at the optical fiber surface,” 2015
1896
+ IEEE Photonics Conference (IPC), 2015, pp. 1-2 (post-deadline paper).
1897
+ 27. D. Bochek, N. Toropov, I. Vatnik, D. Churkin, and M. Sumetsky, “SNAP
1898
+ microresonators introduced by strong bending of optical fibers,” Opt.
1899
+ Lett. 44, 3218 (2019).
1900
+ 28. M. Sumetsky and Y. Dulashko, "Radius variation of optical fibers with
1901
+ angstrom accuracy," Opt. Lett. 35, 4006 (2010).
1902
+ 29. D. V. Kudashkin, A. A. Deriskiba, I. D. Vatnik, S. V. Suchkov, and D. V.
1903
+ Churkin, “Reflection of whispering gallery modes propagating on a
1904
+ surface of an optical fiber from its cleave,” Opt. Express 28, 34530 (2020).
1905
+ 30. M. Sumetsky, “Delay of Light in an Optical Bottle Resonator with
1906
+ Nanoscale Radius Variation: Dispersionless, Broadband, and Low Loss,”
1907
+ Phys. Rev. Lett. 111, 163901 (2013).
1908
+ 31. N. Toropov, S. Zaki, T. Vartanyan, and M. Sumetsky, “Microresonator
1909
+ devices lithographically introduced at the optical fiber surface,” Opt. Lett.
1910
+ 46, 1784 (2021).
1911
+ 32. A. W. Snyder and J. D. Love, Optical Waveguide Theory, (New York,
1912
+ Chapman & Hall, 1983).
1913
+ 33. D. L. P. Vitullo, S. Zaki, D. E. Jones, M. Sumetsky, and M. Brodsky,
1914
+ “Coupling between waveguides and microresonators: the local
1915
+ approach,” Opt. Express 28, 25908 (2020).
1916
+ 34. C. Mahaux and H. A. Weidenmüller, Shell-Model Approach to
1917
+ Nuclear Reactions (North-Holland, Amsterdam, 1969).
1918
+ 35. F.-M. Dittes, “The decay of quantum systems with a small number
1919
+ of open channels,” Phys. Rep. 339, 215 (2000).
1920
+
1921
+ 36. M. Sumetsky, “Mahaux-Weidenmüller approach to cavity quantum
1922
+ electrodynamics and complete resonant down-conversion of the single-
1923
+ photon frequency,” Phys. Rev. A 100, 013801 (2019).
1924
+ 37. C. Cohen-Tannoudji, B. Diu, and F. Laloe, Quantum mechanics, vol.
1925
+ 1 (Wiley-Interscience, New York, 1977).
1926
+ 38. B. E. Little, J.-P. Laine, and H. A. Haus, “Analytic theory of coupling
1927
+ from tapered fibers and half-blocks into microsphere resonators,”
1928
+ J. Lightwave Technol. 17, 704 (1999).
1929
+ 39. M. Sumetsky, “Fundamental limit of microresonator field uniformity and
1930
+ slow light enabled ultraprecise displacement metrology,” Opt. Lett. 46,
1931
+ 1656 (2021).
1932
+ 40. Y. Yang, M. Crespo-Ballesteros and M. Sumetsky, “Experimental
1933
+ demonstration of a bat microresonator,” 2021 Conference on
1934
+ Lasers and Electro-Optics Europe & European Quantum
1935
+ Electronics
1936
+ Conference
1937
+ (CLEO/Europe-EQEC),
1938
+ 2021.
1939
+ doi:
1940
+ 10.1109/CLEO/Europe-EQEC52157.2021.9542687.
1941
+ 41. D. E. Chang, J. S. Douglas, A. González-Tudela, C.-L. Hung, and H. J.
1942
+ Kimble, “Colloquium: Quantum matter built from nanoscopic
1943
+ lattices of atoms and photons,” Rev. Mod. Phys. 90, 031002
1944
+ (2018).
1945
+ 42. S. V. Suchkov, M. Sumetsky, and A. A. Sukhorukov, "Frequency
1946
+ comb generation in SNAP bottle resonators," Opt. Lett. 42, 2149
1947
+ (2017).
1948
+ 43. M. Sumetsky, “Delay of light in an optical bottle resonator with
1949
+ nanoscale radius variation: dispersionless, broadband, and low
1950
+ loss,” Phys. Rev. Lett. 111, 163901 (2013).
1951
+ 44. Q. Yu, S. Zaki, Y. Yang, N. Toropov, X. Shu, and M. Sumetsky,
1952
+ “Rectangular SNAP microresonator fabricated with a femtosecond
1953
+ laser,” Opt. Lett. 44, 5606 (2019).
1954
+ 45. M. Sumetsky, “Slow light optofluidics: a proposal,” Opt. Lett. 39,
1955
+ 5578 (2014).
1956
+ 46. T. Hamidfar, A. Dmitriev, B. Mangan, P. Bianucci, and M.
1957
+ Sumetsky, "Surface nanoscale axial photonics at a capillary fiber,"
1958
+ Opt. Lett. 42, 3060 (2017).
1959
+ 47. G. Gardosi, B. J. Mangan, G. S. Puc, and M. Sumetsky, “Photonic
1960
+ Microresonators Created by Slow Optical Cooking,” ACS Photonics
1961
+ 8, 436 (2021).
1962
+
1963
+
1964
+
1965
+
1966
+
1967
+
EdAzT4oBgHgl3EQfUPw6/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
EdFRT4oBgHgl3EQfBDfd/vector_store/index.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f16fbd8b3694ed9a59615a9fabc7449260037188cbe6edd047a2b53b8e4c5ed7
3
+ size 276041
GdFJT4oBgHgl3EQfECxK/content/tmp_files/2301.11436v1.pdf.txt ADDED
@@ -0,0 +1,258 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Synesthetic Dice: Sensors, Actuators, And Mappings
2
+ Albrecht Kurze
3
+ Chemnitz University of Technology, Albrecht.Kurze@informatik.tu-chemnitz.de
4
+ How bright can you cry? How loud does the sun shine? We developed a multisensory and multimodal tool, the Loaded Dice, for use in
5
+ co-design workshops to research the design space of IoT usage scenarios. The Loaded Dice incorporate the principle of a technical
6
+ synesthesia, being able to map any of the included sensors to any of the included actuators. With just a turn of one of the cubical
7
+ devices it is possible to create a new combination. We discuss the core principles of the Loaded Dice, what sensors and actuators are
8
+ included, how they relate to human senses, and how we realized a meaningful mapping between sensors and actuators. We further
9
+ discuss where we see additional potential in the Loaded Dice to support synesthetic exploration – as Synesthetic Dice – so that you can
10
+ eventually find out who cries brighter.
11
+ CCS CONCEPTS • Human-centered computing~Human computer interaction (HCI)
12
+ Additional Keywords and Phrases: multisensory, multimodal, synesthesia, design, ideation, tools, methods, IoT,
13
+ Internet of Things, haptic technology, cubic shape, tangible interactive devices, input and output devices, tangibles
14
+ ACM Reference Format:
15
+ Albrecht Kurze. 2022. Synesthetic Dice: Sensors, Actuators, And Mappings. In Workshop Sensory Sketching (CHI’22).
16
+ April 22, 2022. 4 pages.
17
+ 1
18
+ INTRODUCTION
19
+ Some years ago we designed and developed the Loaded Dice [8,9], a multisensory and multimodal hybrid toolkit to
20
+ ideate Internet of Things (IoT) devices and scenarios, e.g. for the ‘smart’ home, and with different groups of co-
21
+ designers [3,7,8]. The Loaded Dice filled a gap between analog, non-functional tools, often card-based, e.g. KnowCards
22
+ [1], and functional but tinkering based tools, e.g. littleBits [2], for multisensory und multimodal exploration, ideation
23
+ and prototyping.
24
+ We introduce the Loaded Dice, the core concepts that they are built on, the used sensors and actuators, and how
25
+ they map to different human senses. We will then continue to discuss how we realized mappings between sensed raw
26
+ value, normalized intermediate values, and actuated values. While the mappings that we currently use are sufficiently
27
+ good enough for current purposes, we see big potential in some extended uses as ‘Synesthetic Dice’.
28
+ This brings us to our core question: How can the Loaded Dice be used for exploration and research of synesthetic
29
+ mappings between sensors and actuators, e.g. for innovative interactions and non-verbal communication?
30
+
31
+ SENSORDIE
32
+ TemperatureSensor
33
+ Light Sensor
34
+ Microphone
35
+ MovementSensor
36
+ Potentiometer
37
+ Distance Sensor
38
+ ACTUATORDIE
39
+ Vibration
40
+ Heating Surface
41
+ LED-Bargraph
42
+ Loudspeaker
43
+ Power-LEDs
44
+ FanFigure 1: The Loaded Dice; left: example of devices in use, turning heat into light (sensor die with temperature sensor active and
45
+ actuator die with power LED active) [9]; right faces and functions – sensors and actuators [8]
46
+ 2
47
+ THE LOADED DICE - SENSES, SENSORS, ACTUATORS
48
+ The Loaded Dice1 are a set of two cubical devices wirelessly connected (fig. 2a). Each cube has six sides, offering in one
49
+ cube six sensors and in the other cube six actuators, one on each side, suitable for multisensory and multimodal
50
+ environmental and user interactions. The sensor cube normalizes a raw sensor value meaningfully, transmits it, and
51
+ then the other cube actuates it mapped on an output. The cubical shape communicates the intuitive reading that the
52
+ top side is active, like a die, offering an easy and spontaneous way to re-combine sensors and actuators. Every sensor-
53
+ in and actuator-out combination is possible resulting in 36 combinations in total. [5]
54
+ The “traditional five” human senses are sight, hearing, taste, smell and touch. Secondary senses are temperature,
55
+ pain, proprioception and balance. Due to the constraints of the technical platform we could not address all human
56
+ senses with sensors and actuators. Overall the Loaded Dice holds sensors and actuators equivalent to some human
57
+ senses directly (see fig. 1 and table 1 for details). It is also possible to think about effects to address other senses using
58
+ the given sensors and actuators, e.g. to inflict pain via the Peltier element through excessive heat or cold (not intended
59
+ nor recommended). It is also possible (but currently not implemented) to use the internal inertial measurement unit
60
+ (IMU), consisting of an accelerometer and gyrometer, not only for interaction controls but also as a sense, as an
61
+ equivalent to proprioception and balance (movement and position).
62
+ New multisensory interaction modalities are possible but not yet implemented, e.g. olfactory / smell. They have the
63
+ potential to broaden interaction qualities even further and especially in an emotional way [6].
64
+ Table 1. Human senses vs. sensors and actuators in the Loaded Dice
65
+ Human Sense
66
+ Sensor
67
+ Actuator
68
+ sight
69
+ (visual stimuli)
70
+ luxmeter (visible light luminosity/ brightness)
71
+ passive infrared detector (PIR movement)
72
+ ultrasonic transceiver (distance)
73
+ power LED (brightness)
74
+ LED ring-graph (count, overall brightness, color)
75
+ hearing
76
+ (auditive stimuli)
77
+ microphone (amplitude)
78
+ sound (modulated note for instrument)
79
+ (vibration motor, rattling noise)
80
+ (fan, air flow noise)
81
+ touch
82
+ (tactile stimuli)
83
+ potentiometer (manual angular dial of 270°)
84
+ vibration motor (vibration)
85
+ fan (mechanical stimulation on hairs)
86
+ temperature
87
+ (thermal stimuli)
88
+ infrared thermometer (thermopile / thermal
89
+ radiation)
90
+ Peltier element (cooling and heating plate)
91
+ fan (cooling by chill effect on skin)
92
+ 1 video demonstrating the Loaded Dice: https://www.youtube.com/watch?v=-E5aUiktCic
93
+ 2
94
+
95
+ SENSORDIE
96
+ TemperatureSensor
97
+ Light Sensor
98
+ Microphone
99
+ MovementSensor
100
+ Potentiometer
101
+ Distance Sensor
102
+ ACTUATORDIE
103
+ Vibration
104
+ Heating Surface
105
+ LED-Bargraph
106
+ Loudspeaker
107
+ Power-LEDs
108
+ Fan3
109
+ SYNESTHESIA - MAPPING SENSES AND MODALITIES
110
+ Synesthesia describes the phenomenon of an event being experienced by another, separate sensory modality [4]. While
111
+ medical not exact, in principle, this means a sound might not only be heard but also be seen as a color (as an example).
112
+ Most existing tools, i.e. for IoT ideation, do not employ synesthesia effects as a design opportunity in order to break
113
+ with existing sensing stereotypes for framing design spaces. Such a stereotype could be e.g. that making noise should
114
+ always be connected with hearing noise. While most related digital (IoT) ideation tools do allow for flexible
115
+ combinations of different sensors and actuators in principle, this is not ad hoc possible. Instead they require necessary
116
+ steps in combining parts or mapping sensor values to actuator values. Thus, they demand an initial idea of how the
117
+ combination should play out. Our tool allows users to explore such synesthetic effects ad hoc.
118
+ We implemented a meaningful mapping between every sensor and actuator that is used in the Loaded Dice. This
119
+ includes reasonably chosen sampling rates, ranges and steppings for raw input values, their normalization on internal
120
+ values and the conversion back to meaningful output values. All this is done internally in hard- and software, without
121
+ the need of user intervention. Selecting a new sensor-actuator combination just requires bringing another side to the
122
+ top. Based on the presented design rationale, a co-designer can transport heat over a distance by choosing the infrared
123
+ thermometer and Peltier element sides of both cubes. Rotating the actuator cube to the power-LEDs would transform
124
+ the temperature into light, thus mimicking synesthesia-like perception.
125
+ The possibilities of the Loaded Dice can be used in a framed scenario-driven co-design approach, in open
126
+ exploration or even just for ‘sensory sketching’, even for ‘weird’ synesthetic combinations, e.g:
127
+
128
+ to try out how bright sunlight sounds or feels as vibration
129
+
130
+ what temperature a loud cry has
131
+
132
+ how much air-flow half a meter distance is
133
+
134
+ whether you can feel the flickering of light …
135
+ We use meaningful but simple functions for preprocessing of raw sensor values and normalization to an
136
+ intermediate data value as well back to actuations (table 2). Overall, the mappings are done in a predefined ‘static’
137
+ way. However, static does not mean one fits all. It is necessary to consider non-linearities and dynamics, e.g. for light
138
+ and sound, as these senses are not perceived in a linear or static manner by humans. However, we applied ‘just good
139
+ enough’ assumptions for meaningfulness without the claim of physical or psychometric correctness, sometimes even a
140
+ bit off to make effects clearer. Currently also the sensor as well as the selected actuator are considered for the mapping
141
+ in addition to the normalized value. We do this mainly for technical reasons as the different modalities operate at
142
+ different speeds. Currently, only the LED ring graphic signals which sensor has sampled the data by changing color.
143
+ Table 2: Current mapping from sensed values to intermediate values and then to actuated values
144
+ Sensor
145
+ Sensor Mapping
146
+ Value
147
+ Actuator Mapping
148
+ Actuator
149
+ potentiometer
150
+ 0..270° AD sampling 0..1023  linear  0..24
151
+ 0..24
152
+  Neopixels count 0..24, color coded by sensor,
153
+ brightness per pixel static
154
+ ring-graph
155
+ thermometer
156
+ digital read-out 0..50 °C  linear  0..24
157
+  sqr  0..576  0..255 RGB brightness
158
+ power LED
159
+ microphone
160
+ 50ms window AD sampling 0..1023  max-min
161
+ difference  0..1023  linear  0..24
162
+ 0  0; 1..24  MIDI noteOn(value+50)
163
+ sound
164
+ distance
165
+ 0  0; 1..72 cm  linear  1..24
166
+ 0..12  -255..0 (cooling) 12..24  0..255 (heating) PWM
167
+ or 0..24  0..255 (from neutral to heating only) PWM
168
+ Peltier thermo
169
+ PIR movement
170
+ binary 0  0; 1  24
171
+ 0  0; 1..24  64..255 PWM
172
+ vibration
173
+ light
174
+ digital read-out 0..65535 lx  sqrt  0..48  0..24
175
+ 0  0; 1..24  160..255 PWM
176
+ fan
177
+ Every combination is possible, alignment in lines just as examples. AD: analogdigital conversion, PWM: pulse width modulation
178
+ 3
179
+
180
+ SENSORDIE
181
+ TemperatureSensor
182
+ Light Sensor
183
+ Microphone
184
+ MovementSensor
185
+ Potentiometer
186
+ Distance Sensor
187
+ ACTUATORDIE
188
+ Vibration
189
+ Heating Surface
190
+ LED-Bargraph
191
+ Loudspeaker
192
+ Power-LEDs
193
+ FanWhile we are quite satisfied what the Loaded Dice can already do there are some new possibilities at hand:
194
+
195
+ more use of colors: for power LED element and LED ring-graph (NeoPixels are colorful…)
196
+
197
+ other use of sound: other (music/midi) instruments, modulation of velocity and pitch, other sounds
198
+ (artificial or sampled in nature)
199
+
200
+ use of spatial component: position of the LEDs of the ring-graph, color fades, patterns
201
+
202
+ use of temporal components: from time static value to dynamic patterns for sound, vibration, light, air
203
+ flow etc.
204
+ A flexible “sketching” of a new mapping function would allow to bring in completely new synesthesia effects, also not
205
+ necessarily only limited to one input sensor and one output actuator at one time.
206
+ 4
207
+ CONCLUSION
208
+ While the Loaded Dice can already be used meaningfully for activities associated with synesthesia, e.g. for ideation, we
209
+ see a lot of potential in more flexible mappings and even other creative uses of what the sensors and actuators might
210
+ do. We are open for inspirations and ideas.
211
+ ACKNOWLEDGMENTS
212
+ This research is funded by the German Ministry of Education and Research (BMBF), grant FKZ 16SV7116.
213
+ References
214
+ [1]
215
+ Tina Aspiala and Alexandra Deschamps-Sonsino. 2016. Know Cards: Learn. Play. Collect. Know Cards. Retrieved December 6, 2016 from
216
+ http://know-cards.myshopify.com/
217
+ [2]
218
+ Ayah Bdeir. 2009. Electronics As Material: LittleBits. In Proceedings of the 3rd International Conference on Tangible and Embedded Interaction (TEI
219
+ ’09), 397–400. https://doi.org/10.1145/1517664.1517743
220
+ [3]
221
+ Arne Berger, William Odom, Michael Storz, Andreas Bischof, Albrecht Kurze, and Eva Hornecker. 2019. The Inflatable Cat: Idiosyncratic Ideation
222
+ Of Smart Objects For The Home. In CHI Conference on Human Factors in Computing Systems Proceedings. https://doi.org/10.1145/3290605.3300631
223
+ [4]
224
+ Peter G. Grossenbacher and Christopher T. Lovelace. 2001. Mechanisms of synesthesia: cognitive and physiological constraints. Trends in cognitive
225
+ sciences 5, 1: 36–41. Retrieved December 15, 2016 from http://www.sciencedirect.com/science/article/pii/S1364661300015710
226
+ [5]
227
+ Albrecht Kurze. 2021. Interaction Qualities For Interactions With, Between, And Through IoT Devices. In 11th International Conference on the
228
+ Internet of Things (IoT ‘21), November 08-12, 2021, St.Gallen, Switzerland. https://doi.org/10.1145/3494322.3494348
229
+ [6]
230
+ Albrecht Kurze. 2021. Scented Dice: New interaction qualities for ideating connected devices. In Workshop Smell, Taste, and Temperature Interfaces
231
+ at Conference on Human Factors in Computing Systems (CHI ’21). Retrieved from https://arxiv.org/abs/2201.10484
232
+ [7]
233
+ Albrecht Kurze, Kevin Lefeuvre, Michael Storz, Andreas Bischof, Sören Totzauer, and Arne Berger. 2016. Explorative Co-Design-Werkzeuge zum
234
+ Entwerfen von Smart Connected Things am Beispiel eines Workshops mit Blinden und Sehbehinderten. In Technische Unterstützungssysteme, die
235
+ die Menschen wirklich wollen, 395–400. Retrieved January 19, 2017 from http://tinyurl.com/janya26
236
+ [8]
237
+ Kevin Lefeuvre, Sören Totzauer, Andreas Bischof, Albrecht Kurze, Michael Storz, Lisa Ullmann, and Arne Berger. 2016. Loaded Dice: Exploring the
238
+ Design Space of Connected Devices with Blind and Visually Impaired People. In Proceedings of the 9th Nordic Conference on Human-Computer
239
+ Interaction (NordiCHI ’16), 31:1-31:10. https://doi.org/10.1145/2971485.2971524
240
+ [9]
241
+ Kevin Lefeuvre, Sören Totzauer, Andreas Bischof, Michael Storz, Albrecht Kurze, and Arne Berger. 2017. Loaded Dice: How to cheat your way to
242
+ creativity. In Proceedings of the 3rd Biennial Research Through Design Conference. https://doi.org/10.6084/m9.figshare.4746976.v1
243
+ 4
244
+
245
+ SENSORDIE
246
+ TemperatureSensor
247
+ Light Sensor
248
+ Microphone
249
+ MovementSensor
250
+ Potentiometer
251
+ Distance Sensor
252
+ ACTUATORDIE
253
+ Vibration
254
+ Heating Surface
255
+ LED-Bargraph
256
+ Loudspeaker
257
+ Power-LEDs
258
+ Fan
GdFJT4oBgHgl3EQfECxK/content/tmp_files/load_file.txt ADDED
@@ -0,0 +1,205 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf,len=204
2
+ page_content='Synesthetic Dice: Sensors, Actuators, And Mappings Albrecht Kurze Chemnitz University of Technology, Albrecht.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
3
+ page_content='Kurze@informatik.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
4
+ page_content='tu-chemnitz.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
5
+ page_content='de How bright can you cry?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
6
+ page_content=' How loud does the sun shine?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
7
+ page_content=' We developed a multisensory and multimodal tool, the Loaded Dice, for use in co-design workshops to research the design space of IoT usage scenarios.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
8
+ page_content=' The Loaded Dice incorporate the principle of a technical synesthesia, being able to map any of the included sensors to any of the included actuators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
9
+ page_content=' With just a turn of one of the cubical devices it is possible to create a new combination.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
10
+ page_content=' We discuss the core principles of the Loaded Dice, what sensors and actuators are included, how they relate to human senses, and how we realized a meaningful mapping between sensors and actuators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
11
+ page_content=' We further discuss where we see additional potential in the Loaded Dice to support synesthetic exploration – as Synesthetic Dice – so that you can eventually find out who cries brighter.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
12
+ page_content=' CCS CONCEPTS • Human-centered computing~Human computer interaction (HCI) Additional Keywords and Phrases: multisensory, multimodal, synesthesia, design, ideation, tools, methods, IoT, Internet of Things, haptic technology, cubic shape, tangible interactive devices, input and output devices, tangibles ACM Reference Format: Albrecht Kurze.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
13
+ page_content=' 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
14
+ page_content=' Synesthetic Dice: Sensors, Actuators, And Mappings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
15
+ page_content=' In Workshop Sensory Sketching (CHI’22).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
16
+ page_content=' April 22, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
17
+ page_content=' 4 pages.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
18
+ page_content=' 1 INTRODUCTION Some years ago we designed and developed the Loaded Dice [8,9], a multisensory and multimodal hybrid toolkit to ideate Internet of Things (IoT) devices and scenarios, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
19
+ page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
20
+ page_content=' for the ‘smart’ home, and with different groups of co- designers [3,7,8].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
21
+ page_content=' The Loaded Dice filled a gap between analog, non-functional tools, often card-based, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
22
+ page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
23
+ page_content=' KnowCards [1], and functional but tinkering based tools, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
24
+ page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
25
+ page_content=' littleBits [2], for multisensory und multimodal exploration, ideation and prototyping.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
26
+ page_content=' We introduce the Loaded Dice, the core concepts that they are built on, the used sensors and actuators, and how they map to different human senses.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
27
+ page_content=' We will then continue to discuss how we realized mappings between sensed raw value, normalized intermediate values, and actuated values.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
28
+ page_content=' While the mappings that we currently use are sufficiently good enough for current purposes, we see big potential in some extended uses as ‘Synesthetic Dice’.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
29
+ page_content=' This brings us to our core question: How can the Loaded Dice be used for exploration and research of synesthetic mappings between sensors and actuators, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
30
+ page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
31
+ page_content=' for innovative interactions and non-verbal communication?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
32
+ page_content=' SENSORDIE TemperatureSensor Light Sensor Microphone MovementSensor Potentiometer Distance Sensor ACTUATORDIE Vibration Heating Surface LED-Bargraph Loudspeaker Power-LEDs FanFigure 1: The Loaded Dice;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
33
+ page_content=' left: example of devices in use, turning heat into light (sensor die with temperature sensor active and actuator die with power LED active) [9];' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
34
+ page_content=' right faces and functions – sensors and actuators [8] 2 THE LOADED DICE - SENSES, SENSORS, ACTUATORS The Loaded Dice1 are a set of two cubical devices wirelessly connected (fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
35
+ page_content=' 2a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
36
+ page_content=' Each cube has six sides, offering in one cube six sensors and in the other cube six actuators, one on each side, suitable for multisensory and multimodal environmental and user interactions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
37
+ page_content=' The sensor cube normalizes a raw sensor value meaningfully, transmits it, and then the other cube actuates it mapped on an output.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
38
+ page_content=' The cubical shape communicates the intuitive reading that the top side is active, like a die, offering an easy and spontaneous way to re-combine sensors and actuators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
39
+ page_content=' Every sensor- in and actuator-out combination is possible resulting in 36 combinations in total.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
40
+ page_content=' [5] The “traditional five” human senses are sight, hearing, taste, smell and touch.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
41
+ page_content=' Secondary senses are temperature, pain, proprioception and balance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
42
+ page_content=' Due to the constraints of the technical platform we could not address all human senses with sensors and actuators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
43
+ page_content=' Overall the Loaded Dice holds sensors and actuators equivalent to some human senses directly (see fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
44
+ page_content=' 1 and table 1 for details).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
45
+ page_content=' It is also possible to think about effects to address other senses using the given sensors and actuators, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
46
+ page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
47
+ page_content=' to inflict pain via the Peltier element through excessive heat or cold (not intended nor recommended).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
48
+ page_content=' It is also possible (but currently not implemented) to use the internal inertial measurement unit (IMU), consisting of an accelerometer and gyrometer, not only for interaction controls but also as a sense, as an equivalent to proprioception and balance (movement and position).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
49
+ page_content=' New multisensory interaction modalities are possible but not yet implemented, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
50
+ page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
51
+ page_content=' olfactory / smell.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
52
+ page_content=' They have the potential to broaden interaction qualities even further and especially in an emotional way [6].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
53
+ page_content=' Table 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
54
+ page_content=' Human senses vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
55
+ page_content=' sensors and actuators in the Loaded Dice Human Sense Sensor Actuator sight (visual stimuli) luxmeter (visible light luminosity/ brightness) passive infrared detector (PIR movement) ultrasonic transceiver (distance) power LED (brightness) LED ring-graph (count,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
56
+ page_content=' overall brightness,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
57
+ page_content=' color) hearing (auditive stimuli) microphone (amplitude) sound (modulated note for instrument) (vibration motor,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
58
+ page_content=' rattling noise) (fan,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
59
+ page_content=' air flow noise) touch (tactile stimuli) potentiometer (manual angular dial of 270°) vibration motor (vibration) fan (mechanical stimulation on hairs) temperature (thermal stimuli) infrared thermometer (thermopile / thermal radiation) Peltier element (cooling and heating plate) fan (cooling by chill effect on skin) 1 video demonstrating the Loaded Dice: https://www.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
60
+ page_content='youtube.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
61
+ page_content='com/watch?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
62
+ page_content='v=-E5aUiktCic 2 SENSORDIE TemperatureSensor Light Sensor Microphone MovementSensor Potentiometer Distance Sensor ACTUATORDIE Vibration Heating Surface LED-Bargraph Loudspeaker Power-LEDs Fan3 SYNESTHESIA - MAPPING SENSES AND MODALITIES Synesthesia describes the phenomenon of an event being experienced by another, separate sensory modality [4].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
63
+ page_content=' While medical not exact, in principle, this means a sound might not only be heard but also be seen as a color (as an example).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
64
+ page_content=' Most existing tools, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
65
+ page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
66
+ page_content=' for IoT ideation, do not employ synesthesia effects as a design opportunity in order to break with existing sensing stereotypes for framing design spaces.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
67
+ page_content=' Such a stereotype could be e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
68
+ page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
69
+ page_content=' that making noise should always be connected with hearing noise.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
70
+ page_content=' While most related digital (IoT) ideation tools do allow for flexible combinations of different sensors and actuators in principle, this is not ad hoc possible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
71
+ page_content=' Instead they require necessary steps in combining parts or mapping sensor values to actuator values.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
72
+ page_content=' Thus, they demand an initial idea of how the combination should play out.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
73
+ page_content=' Our tool allows users to explore such synesthetic effects ad hoc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
74
+ page_content=' We implemented a meaningful mapping between every sensor and actuator that is used in the Loaded Dice.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
75
+ page_content=' This includes reasonably chosen sampling rates, ranges and steppings for raw input values, their normalization on internal values and the conversion back to meaningful output values.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
76
+ page_content=' All this is done internally in hard- and software, without the need of user intervention.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
77
+ page_content=' Selecting a new sensor-actuator combination just requires bringing another side to the top.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
78
+ page_content=' Based on the presented design rationale, a co-designer can transport heat over a distance by choosing the infrared thermometer and Peltier element sides of both cubes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
79
+ page_content=' Rotating the actuator cube to the power-LEDs would transform the temperature into light, thus mimicking synesthesia-like perception.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
80
+ page_content=' The possibilities of the Loaded Dice can be used in a framed scenario-driven co-design approach, in open exploration or even just for ‘sensory sketching’, even for ‘weird’ synesthetic combinations, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
81
+ page_content='g: to try out how bright sunlight sounds or feels as vibration what temperature a loud cry has how much air-flow half a meter distance is whether you can feel the flickering of light … We use meaningful but simple functions for preprocessing of raw sensor values and normalization to an intermediate data value as well back to actuations (table 2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
82
+ page_content=' Overall, the mappings are done in a predefined ‘static’ way.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
83
+ page_content=' However, static does not mean one fits all.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
84
+ page_content=' It is necessary to consider non-linearities and dynamics, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
85
+ page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
86
+ page_content=' for light and sound, as these senses are not perceived in a linear or static manner by humans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
87
+ page_content=' However, we applied ‘just good enough’ assumptions for meaningfulness without the claim of physical or psychometric correctness, sometimes even a bit off to make effects clearer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
88
+ page_content=' Currently also the sensor as well as the selected actuator are considered for the mapping in addition to the normalized value.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
89
+ page_content=' We do this mainly for technical reasons as the different modalities operate at different speeds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
90
+ page_content=' Currently, only the LED ring graphic signals which sensor has sampled the data by changing color.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
91
+ page_content=' Table 2: Current mapping from sensed values to intermediate values and then to actuated values Sensor Sensor Mapping Value Actuator Mapping Actuator potentiometer 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
92
+ page_content='.270° AD sampling 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
93
+ page_content='.1023 \uf0e0 linear \uf0e0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
94
+ page_content='.24 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
95
+ page_content='.24 \uf0e0 Neopixels count 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
96
+ page_content='.24, color coded by sensor, brightness per pixel static ring-graph thermometer digital read-out 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
97
+ page_content='.50 °C \uf0e0 linear \uf0e0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
98
+ page_content='.24 \uf0e0 sqr \uf0e0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
99
+ page_content='.576 \uf0e0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
100
+ page_content='.255 RGB brightness power LED microphone 50ms window AD sampling 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
101
+ page_content='.1023 \uf0e0 max-min difference \uf0e0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
102
+ page_content='.1023 \uf0e0 linear \uf0e0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
103
+ page_content='.24 0 \uf0e0 0;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
104
+ page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
105
+ page_content='.24 \uf0e0 MIDI noteOn(value+50) sound distance 0 \uf0e0 0;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
106
+ page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
107
+ page_content='.72 cm \uf0e0 linear \uf0e0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
108
+ page_content='.24 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
109
+ page_content='.12 \uf0e0 -255.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
110
+ page_content='.0 (cooling) 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
111
+ page_content='.24 \uf0e0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
112
+ page_content='.255 (heating) PWM or 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
113
+ page_content='.24 \uf0e0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
114
+ page_content='.255 (from neutral to heating only) PWM Peltier thermo PIR movement binary 0 \uf0e0 0;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
115
+ page_content=' 1 \uf0e0 24 0 \uf0e0 0;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
116
+ page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
117
+ page_content='.24 \uf0e0 64.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
118
+ page_content='.255 PWM vibration light digital read-out 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
119
+ page_content='.65535 lx \uf0e0 sqrt \uf0e0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
120
+ page_content='.48 \uf0e0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
121
+ page_content='.24 0 \uf0e0 0;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
122
+ page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
123
+ page_content='.24 \uf0e0 160.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
124
+ page_content='.255 PWM fan Every combination is possible, alignment in lines just as examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
125
+ page_content=' AD: analog\uf0e0digital conversion,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
126
+ page_content=' PWM: pulse width modulation 3 SENSORDIE TemperatureSensor Light Sensor Microphone MovementSensor Potentiometer Distance Sensor ACTUATORDIE Vibration Heating Surface LED-Bargraph Loudspeaker Power-LEDs FanWhile we are quite satisfied what the Loaded Dice can already do there are some new possibilities at hand: more use of colors: for power LED element and LED ring-graph (NeoPixels are colorful…) other use of sound: other (music/midi) instruments,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
127
+ page_content=' modulation of velocity and pitch,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
128
+ page_content=' other sounds (artificial or sampled in nature) use of spatial component: position of the LEDs of the ring-graph,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
129
+ page_content=' color fades,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
130
+ page_content=' patterns use of temporal components: from time static value to dynamic patterns for sound,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
131
+ page_content=' vibration,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
132
+ page_content=' light,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
133
+ page_content=' air flow etc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
134
+ page_content=' A flexible “sketching” of a new mapping function would allow to bring in completely new synesthesia effects, also not necessarily only limited to one input sensor and one output actuator at one time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
135
+ page_content=' 4 CONCLUSION While the Loaded Dice can already be used meaningfully for activities associated with synesthesia, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
136
+ page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
137
+ page_content=' for ideation, we see a lot of potential in more flexible mappings and even other creative uses of what the sensors and actuators might do.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
138
+ page_content=' We are open for inspirations and ideas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
139
+ page_content=' ACKNOWLEDGMENTS This research is funded by the German Ministry of Education and Research (BMBF), grant FKZ 16SV7116.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
140
+ page_content=' References [1] Tina Aspiala and Alexandra Deschamps-Sonsino.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
141
+ page_content=' 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
142
+ page_content=' Know Cards: Learn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
143
+ page_content=' Play.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
144
+ page_content=' Collect.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
145
+ page_content=' Know Cards.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
146
+ page_content=' Retrieved December 6, 2016 from http://know-cards.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
147
+ page_content='myshopify.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
148
+ page_content='com/ [2] Ayah Bdeir.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
149
+ page_content=' 2009.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
150
+ page_content=' Electronics As Material: LittleBits.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
151
+ page_content=' In Proceedings of the 3rd International Conference on Tangible and Embedded Interaction (TEI ’09), 397–400.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
152
+ page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
153
+ page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
154
+ page_content='1145/1517664.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
155
+ page_content='1517743 [3] Arne Berger, William Odom, Michael Storz, Andreas Bischof, Albrecht Kurze, and Eva Hornecker.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
156
+ page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
157
+ page_content=' The Inflatable Cat: Idiosyncratic Ideation Of Smart Objects For The Home.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
158
+ page_content=' In CHI Conference on Human Factors in Computing Systems Proceedings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
159
+ page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
160
+ page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
161
+ page_content='1145/3290605.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
162
+ page_content='3300631 [4] Peter G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
163
+ page_content=' Grossenbacher and Christopher T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
164
+ page_content=' Lovelace.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
165
+ page_content=' 2001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
166
+ page_content=' Mechanisms of synesthesia: cognitive and physiological constraints.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
167
+ page_content=' Trends in cognitive sciences 5, 1: 36–41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
168
+ page_content=' Retrieved December 15, 2016 from http://www.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
169
+ page_content='sciencedirect.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
170
+ page_content='com/science/article/pii/S1364661300015710 [5] Albrecht Kurze.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
171
+ page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
172
+ page_content=' Interaction Qualities For Interactions With, Between, And Through IoT Devices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
173
+ page_content=' In 11th International Conference on the Internet of Things (IoT ‘21), November 08-12, 2021, St.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
174
+ page_content='Gallen, Switzerland.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
175
+ page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
176
+ page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
177
+ page_content='1145/3494322.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
178
+ page_content='3494348 [6] Albrecht Kurze.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
179
+ page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
180
+ page_content=' Scented Dice: New interaction qualities for ideating connected devices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
181
+ page_content=' In Workshop Smell, Taste, and Temperature Interfaces at Conference on Human Factors in Computing Systems (CHI ’21).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
182
+ page_content=' Retrieved from https://arxiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
183
+ page_content='org/abs/2201.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
184
+ page_content='10484 [7] Albrecht Kurze, Kevin Lefeuvre, Michael Storz, Andreas Bischof, Sören Totzauer, and Arne Berger.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
185
+ page_content=' 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
186
+ page_content=' Explorative Co-Design-Werkzeuge zum Entwerfen von Smart Connected Things am Beispiel eines Workshops mit Blinden und Sehbehinderten.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
187
+ page_content=' In Technische Unterstützungssysteme, die die Menschen wirklich wollen, 395–400.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
188
+ page_content=' Retrieved January 19, 2017 from http://tinyurl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
189
+ page_content='com/janya26 [8] Kevin Lefeuvre, Sören Totzauer, Andreas Bischof, Albrecht Kurze, Michael Storz, Lisa Ullmann, and Arne Berger.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
190
+ page_content=' 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
191
+ page_content=' Loaded Dice: Exploring the Design Space of Connected Devices with Blind and Visually Impaired People.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
192
+ page_content=' In Proceedings of the 9th Nordic Conference on Human-Computer Interaction (NordiCHI ’16), 31:1-31:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
193
+ page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
194
+ page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
195
+ page_content='1145/2971485.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
196
+ page_content='2971524 [9] Kevin Lefeuvre, Sören Totzauer, Andreas Bischof, Michael Storz, Albrecht Kurze, and Arne Berger.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
197
+ page_content=' 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
198
+ page_content=' Loaded Dice: How to cheat your way to creativity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
199
+ page_content=' In Proceedings of the 3rd Biennial Research Through Design Conference.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
200
+ page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
201
+ page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
202
+ page_content='6084/m9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
203
+ page_content='figshare.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
204
+ page_content='4746976.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
205
+ page_content='v1 4 SENSORDIE TemperatureSensor Light Sensor Microphone MovementSensor Potentiometer Distance Sensor ACTUATORDIE Vibration Heating Surface LED-Bargraph Loudspeaker Power-LEDs Fan' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/GdFJT4oBgHgl3EQfECxK/content/2301.11436v1.pdf'}
GtAyT4oBgHgl3EQfrfmY/content/tmp_files/2301.00562v1.pdf.txt ADDED
@@ -0,0 +1,1507 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Age-Optimal Multi-Channel-Scheduling under
2
+ Energy and Tolerance Constraints
3
+ Xujin Zhou, Irem Koprulu, Atilla Eryilmaz
4
+ Electrical and Computer Engineering
5
+ The Ohio State University
6
+ Columbus, US
7
+ {zhou.2400@osu.edu, irem.koprulu@gmail.com, eryilmaz.2@osu.edu}
8
+ Abstract—We study the optimal scheduling problem where n
9
+ source nodes attempt to transmit updates over L shared wireless
10
+ on/off fading channels to optimize their age performance under
11
+ energy and age-violation tolerance constraints. Specifically, we
12
+ provide a generic formulation of age-optimization in the form of
13
+ a constrained Markov Decision Processes (CMDP), and obtain
14
+ the optimal scheduler as the solution of an associated Linear
15
+ Programming problem. We investigate the characteristics of the
16
+ optimal single-user multi-channel scheduler for the important
17
+ special cases of average-age and violation-rate minimization.
18
+ This leads to several key insights on the nature of the optimal
19
+ allocation of the limited energy, where a usual threshold-based
20
+ policy does not apply and will be useful in guiding scheduler
21
+ designers. We then investigate the stability region of the optimal
22
+ scheduler for the multi-user case. We also develop an online
23
+ scheduler using Lyapunov-drift-minimization methods that do
24
+ not require the knowledge of channel statistics. Our numerical
25
+ studies compare the stability region of our online scheduler to the
26
+ optimal scheduler to reveal that it performs closely with unknown
27
+ channel statistics.
28
+ I. INTRODUCTION
29
+ In recent years, the Internet of Things (IoT) has become
30
+ one of the most important frameworks of the next-generation
31
+ wireless networks, whereby a large number of mobile devices
32
+ need to be supported over an ultra-wide frequency spectrum
33
+ (see, for example, [1]). In particular, for many real-time IoT
34
+ applications, it is necessary for the devices to send fresh
35
+ updates over the shared spectrum. To measure the freshness
36
+ of data, the concept of Age of Information (AoI) has been
37
+ introduced over the last decade (see, for example, [2]–[4]),
38
+ which is defined concisely as the elapsed time since the
39
+ generation time of the last received status update. Since
40
+ the introduction of the AoI metric, numerous related studies
41
+ emerged in various networking scenarios, including wireless
42
+ random access networks (e.g., [5], [6]), content distribution
43
+ networks (e.g., [7], [8]), scheduling (e.g., [9]–[13]), queuing
44
+ networks (e.g., [14], [15]), and vehicular networks (e.g., [16]).
45
+ Recently, other AoI related metrics have been developed in
46
+ order to address more generalized or different forms of ageing,
47
+ such as: non-linear AoI (e.g., [4], [17]), peak AoI (e.g., [18]),
48
+ time-since-last-service (e.g., [19]), age upon decisions (e.g.,
49
+ [20]), to name a few. Among them, the metric, called the age-
50
+ violation-rate (see [15], [21], [22]) is of particular interest for
51
+ real-time IoT services that have hard age-deadline constraints
52
+ and a limited tolerance to violating this deadline (see [23],
53
+ [24] for further motivation of this metric).
54
+ In view of its significance for next generation IoT networks,
55
+ in this paper, we study the general optimal multi-channel
56
+ scheduling problem to optimize varying forms of age perfor-
57
+ mances under energy and age-violation tolerance constraints.
58
+ Our contributions can be listed as:
59
+ • We provide a generic formulation of age-optimization
60
+ as a Constrained Markov Decision Problem (CMDP)
61
+ (see [25]–[27]) and obtain the age-optimal multi-channel
62
+ scheduler as the solution of an associated Linear Pro-
63
+ gramming problem, first for the single-source (in Sec-
64
+ tion III) and then for general the multi-source (in Sec-
65
+ tion IV) scenarios.
66
+ • For the single-source multi-channel scenario, we in-
67
+ vestigate the characteristics of the optimal schedulers
68
+ under energy constraints for two age metrics that are
69
+ important for IoT applications: (i) average-age mini-
70
+ mization; and (ii) age-violation-rate minimization, a non-
71
+ convex/concave metric (in Section III-C). Our investiga-
72
+ tions reveal various insights on different energy allocation
73
+ structures, as well as the common monotonicity proper-
74
+ ties of the optimal schedulers for minimizing these two
75
+ metrics, which is useful for guiding scheduler designers.
76
+ • For the multi-source age optimal scheduling problem,
77
+ we also study the feasibility region of the average-
78
+ age-optimal scheduler under age-violation-rate tolerance
79
+ constraints to contrast its results with those of related
80
+ earlier works that are developed for the single-channel
81
+ multi-user scenario (see Section IV-C and Section VI).
82
+ • Moreover, we develop (in Section V) an online scheduler
83
+ using Lyapunov-drift-minimization methods (e.g., [28])
84
+ that does not require the knowledge of channel statistics,
85
+ and compare its performance to the optimal and earlier
86
+ designs to reveal how much the knowledge of channel
87
+ statistics affects the feasibility region (see Section VI).
88
+ Our work relates to, but also differs from several other
89
+ related works in this domain. Many early works (e.g., [9],
90
+ [12], [29]) aim to minimize AoI under power constraints
91
+ but with the assumption of reliable channels as opposed to
92
+ the fading channels that we consider. More recent works
93
+ arXiv:2301.00562v1 [cs.IT] 2 Jan 2023
94
+
95
+ (e.g., [10], [30]) aim to minimize AoI-related costs based
96
+ on max-age matching, while other works (e.g., [29], [31])
97
+ proposed AoI minimization schedulers based on Whittle Index
98
+ approach. However, to the best of our knowledge, prior works
99
+ predominantly assume that one source can choose at most
100
+ one channel, which is an important factor in proving the
101
+ Whittle Indexability of the corresponding problems they solve.
102
+ In contrast, one of the key features our setting is the possibility
103
+ of each user to transmit over multiple channels as enabled by
104
+ new wireless technologies. Furthermore, most of the above
105
+ mentioned works have average or peak AoI as the objective
106
+ function, while we consider more general age-based objective
107
+ functions, which for example allows the objective function to
108
+ be a non-convex metric such as the age-violation-rate. In this
109
+ multi-channel setting with general objectives, we observe (cf.
110
+ Section III-C) that the optimal solution can in fact possess non-
111
+ monotone characteristics, which make the Whittle Indexability
112
+ approach infeasible in general. The work in [21] has con-
113
+ sidered the multi-source single-channel scheduling problem
114
+ under tolerance constraints, which is a special case of our
115
+ setting. We would like to note that this interesting work
116
+ [21] has been a primary motivation for our current work in
117
+ exploring a different approach based on the CMDP framework
118
+ that guarantees optimality and applies to more general multi-
119
+ channel scenarios with additional energy constraints. There
120
+ are also works (e.g., [32], [33]) that focus on learning-based
121
+ approaches which can be considered as complementary to the
122
+ focus of this work.
123
+ II. SYSTEM MODEL
124
+ We consider the operation of a discrete-time wireless access
125
+ system, whereby n source nodes share L on/off fading wireless
126
+ channels to update their ageing status at a receiver (such as a
127
+ base station) under energy and violation tolerance constraints
128
+ (see Figure 1).
129
+ Figure 1. n sources share L on-off fading channels to update their status to
130
+ a receiver under energy and tolerance constraints in order to keep their age
131
+ levels low.
132
+ Our goal is to develop generic solution strategies to find
133
+ optimal schedulers that can optimize diverse age-based metrics
134
+ while meeting certain requirements on energy consumption
135
+ and tolerance levels. We describe the key terminology and the
136
+ essential system dynamics in the rest of this section. Then,
137
+ in the following sections we formulate and solve classes of
138
+ age-optimization problems for single and multi-source cases,
139
+ subsequently.
140
+ Scheduling policy and age-violation-tolerance: We assume
141
+ that each source node i ∈ {1, · · · , n} refreshes its status
142
+ and creates a new packet at the beginning of every time
143
+ slot t ∈ {1, 2, 3, · · · }. Source nodes attempt to transmit their
144
+ freshest packet to the receiver, for example a base station
145
+ (BS), whenever they get a chance to transmit. Every time the
146
+ BS successfully receives a new status from source node i,
147
+ it saves the current status and discards all previous packets
148
+ received from that node. As such, the BS keeps only one
149
+ packet from each source node, namely the freshest one. We
150
+ use Xi[t] to denote the generation time of the packet stored
151
+ at the BS from source i at time t. We define the age Ai[t]
152
+ of source node i at time t as the time that has elapsed since
153
+ the generation of its last received packet1: Ai[t] ≜ t − Xi[t].
154
+ We use2 A[t] ≜ (A1[t], · · · , An[t]) to denote the ages of all
155
+ sources at time slot t.
156
+ At the beginning of each time slot, the centralized scheduler
157
+ decides which channels each of the source nodes will use to
158
+ transmit to the base station based on the ages A[t] of all source
159
+ nodes. Let ui(A[t]) be the number of channels source node
160
+ i uses to transmit at time t. Each transmission attempt can
161
+ resolve in success or failure which we will describe below
162
+ as part of the channel success model. If the base station
163
+ successfully receives the packet from source i at time t, then
164
+ its age at time t + 1 will reset to 1, otherwise its age will
165
+ increase by one, i.e.,
166
+ Ai[t + 1] =
167
+
168
+ 1, if transmission of source i succeeds
169
+ Ai[t] + 1,
170
+ otherwise.
171
+ We allow each source i to have a desired age thresh-
172
+ old/deadline τi. The information of source i is up-to-date if
173
+ its age is less than or equal to this threshold τi. Otherwise, we
174
+ speak of an age violation in that slot. In particular, we define
175
+ the age-violation-rate of source i as the long-term average
176
+ fraction of time slots when the source’s age Ai[t] exceeds
177
+ its threshold τi, i.e., lim
178
+ T →∞
179
+ 1
180
+ T
181
+ T
182
+
183
+ t=1
184
+ 1 {Ai[t] > τi}. We use ϵi ∈
185
+ [0, 1] to indicate the tolerance of source i that measures the
186
+ maximum allowed age-violation-rate for its updates. (ϵi = 1
187
+ indicates that there is no violation rate constraint, and ϵi = 0
188
+ indicates that we do not allow any deadline violation.) When
189
+ the age violation rate is no greater than the tolerance rate, the
190
+ age violation tolerance constraint is satisfied.
191
+ Channel success model and energy constraints: The n
192
+ source nodes share L wireless on/off fading channels, each
193
+ of which can accommodate at most one packet transmission.
194
+ However, even when there is a single transmission over
195
+ a channel, a successful transmission is not guaranteed. In
196
+ 1This metric is also referred to as Age-of-Information (AoI) and Time-Since-
197
+ Last-Service (TSLS) in different contexts. In the rest of the paper, we will refer
198
+ to it as AoI or simple as age, interchangeably.
199
+ 2We will consistently use bold symbols to represent vectors.
200
+
201
+ particular, source node i has a channel success probability of
202
+ µi when transmitting over each of its assigned channels3.
203
+ We call the update of source i in a slot to be a success
204
+ if any one of its transmissions over its assigned channels is
205
+ successful. Since the channel is a collision channel, for an
206
+ optimal scheduler we always have
207
+ n
208
+
209
+ i=1
210
+ ui(A[t]) ≤ L. Once
211
+ the value of ui(A[t]) is decided for all i, the scheduler will
212
+ assign different channels to different sources, so that no two
213
+ sources transmit over the same channel. Also, note that under
214
+ the described channel success model, the probability for the
215
+ BS to successfully receive an update from source node i when
216
+ the node uses l channels is 1 − (1 − µi)l.
217
+ We assume that each transmission over a channel comes
218
+ with an energy cost of 1 unit4. We require that the aggregate
219
+ time-average energy cost for source i is not greater than a
220
+ given constraint bi channels per slot, i.e., we require
221
+ lim
222
+ T →∞
223
+ 1
224
+ T
225
+ T
226
+
227
+ t=1
228
+ ui (A[t]) ≤ bi,
229
+ bi ∈ R+.
230
+ It is obvious that transmitting over more channels will
231
+ increase the success probability of a source, but increase
232
+ energy consumption. We are interested in finding the number
233
+ of channels that when allocated to sources optimize the desired
234
+ age performance given the current age state, as well as energy
235
+ and and tolerance constraints discussed above. In the next
236
+ section, we attack this problem within the constrained Markov
237
+ Decision Process (MDP) framework first for a single user, and
238
+ then extend our approach to cover the multi-user setting.
239
+ III. AGE-OPTIMAL MULTI-CHANNEL SCHEDULING FOR A
240
+ SINGLE USER
241
+ In this section, we first consider the single-user age-optimal
242
+ multi-channel scheduling problem. This not only allows us to
243
+ simplify the notation by omitting the subscripts, but also is of
244
+ particular interest for the next generation ultra-wideband wire-
245
+ less communication technologies that are expected to support
246
+ low-delay access over multiple fading channels. We formulate
247
+ a general age-optimal optimization problem which can be
248
+ used in different scenarios in Section III-A and following the
249
+ analysis of the performance in Section III-B. To that end,
250
+ in Section III-C, we study the characterization and insights
251
+ of the optimal schedulers for two important special cases of
252
+ minimizing the average-age and the age-violation-rate, which
253
+ will be useful in guiding scheduler designers.
254
+ A. Problem formulation
255
+ The problem of minimizing time-averaged age-based ob-
256
+ jectives under average energy and tolerance constraints can
257
+ 3All our development can be generalized to the case when the success
258
+ probability between source i and channel j is allowed to be different as µij.
259
+ However, this is omitted here as it increases the complexity of the exposition
260
+ without adding to the substance.
261
+ 4This can also be generalized to non-uniform energy costs over different
262
+ channels, but omitted to avoid cumbersome notation.
263
+ be generally formulated as the following constrained Markov
264
+ decision problem [25]:
265
+ min
266
+ u(A)
267
+ lim
268
+ T →∞
269
+ 1
270
+ T
271
+ T
272
+
273
+ t=1
274
+ E [ω0(A[t])]
275
+ (1)
276
+ s.t :
277
+ lim
278
+ T →∞
279
+ 1
280
+ T
281
+ T
282
+
283
+ t=1
284
+ E [u (A[t])] ≤ b,
285
+ (2)
286
+ lim
287
+ T →∞
288
+ 1
289
+ T
290
+ T
291
+
292
+ t=1
293
+ E [ωk (A[t])] ≤ ck, k = 1, · · · , K,
294
+ u(A[t]) ∈ {0, 1, · · · , L}.
295
+ The optimization is performed over Markovian policies
296
+ described by a function u(·) that maps age levels to number
297
+ of channels. It is known that such Markovian policies are
298
+ sufficient for optimal operation [25].
299
+ The first constraint on the time-averaged u(·) captures the
300
+ average energy constraint discussed in the system model. The
301
+ functions ωk(·) serve as general functions that map the current
302
+ state A[t] to a value that measures the cost of that age with
303
+ respect to various measures5 By setting different mappings for
304
+ the weight function ω0(A[t]), the objective can be changed
305
+ into different commonly used age-related objectives: letting
306
+ ω0(a) = −1{a = 1} transform the objective to maximizing
307
+ the average throughput; letting ω0(a) = a makes the objective
308
+ minimize the average AoI; letting ω0(a) = 1{a ≥ d} make the
309
+ objective minimize the average age-violation rate. Note that
310
+ this allows the objective function to be a non-convex/concave
311
+ function.
312
+ B. Performance analysis
313
+ Next, we will analyze the generic constrained optimization
314
+ problem under energy constraint by showing that the problem
315
+ is equivalent to a Linear Programming (LP) problem and thus
316
+ describe the optimal policy.
317
+ Theorem 1: The solution of the generic age-optimization
318
+ problem (1) can be obtained by solving the following linear
319
+ programming problem:
320
+ min
321
+ yla
322
+ D
323
+
324
+ a=1
325
+ L
326
+
327
+ l=0
328
+ yl
329
+ aω0(a)
330
+ s.t:
331
+ D
332
+
333
+ a=1
334
+ L
335
+
336
+ l=0
337
+ yl
338
+ a · l ≤ b,
339
+ D
340
+
341
+ a=1
342
+ L
343
+
344
+ l=0
345
+ yl
346
+ aωk(a) ≤ ck, k = 1, · · · , K,
347
+ 0 ≤ yl
348
+ a ≤ 1
349
+ ∀1 ≤ a ≤ D, 0 ≤ l ≤ L,
350
+ D
351
+
352
+ a=1
353
+ L
354
+
355
+ l=0
356
+ yl
357
+ a = 1,
358
+ Qy = 0,
359
+ where y is a column vector of size DL with y
360
+ =
361
+ (y1
362
+ 1, · · · , yL
363
+ 1 , · · · , y1
364
+ D, · · · , yL
365
+ D)T as its components; D is an
366
+ 5We note that the problem can also solved with the same approach
367
+ (but heavier notation) by more generally defining ωk(A[t], u(A[t])) to be
368
+ functions of both the age and the action.
369
+
370
+ upper bound on the age state in the system which can be
371
+ set sufficiently large so that the probability of reaching D
372
+ is vanishing.6 Qy = 0 is the matrix representation of the
373
+ following (global balance) equations:
374
+ L
375
+
376
+ l=0
377
+ yl
378
+ a+1 −
379
+ L
380
+
381
+ l=0
382
+ yl
383
+ a(1 − µ)l = 0
384
+ ∀a = 1, · · · , D − 2,
385
+ L
386
+
387
+ l=0
388
+
389
+ 1 − (1 − µ)l�
390
+ yl
391
+ D −
392
+ L
393
+
394
+ l=0
395
+ yl
396
+ D−1(1 − µ)l = 0,
397
+
398
+ L
399
+
400
+ l=0
401
+ yl
402
+ 1(1 − µ)l +
403
+ D
404
+
405
+ a=2
406
+ L
407
+
408
+ l=0
409
+ yl
410
+ a
411
+
412
+ 1 − (1 − µ)l�
413
+ = 0.
414
+ If this LP is feasible, and y is an optimal solution, then the
415
+ optimal policy u∗(a) is a probabilistic policy, whereby the
416
+ probability f l
417
+ a of choosing l channels when the age is at state
418
+ a equals:
419
+ f l
420
+ a =
421
+
422
+
423
+
424
+
425
+
426
+
427
+
428
+
429
+
430
+
431
+
432
+
433
+
434
+
435
+
436
+ yl
437
+ a
438
+ L
439
+
440
+ l=0
441
+ yl
442
+ a
443
+ ,
444
+ if
445
+ L
446
+
447
+ l=0
448
+ yl
449
+ a ̸= 0
450
+ 1
451
+ L,
452
+ if
453
+
454
+ l
455
+ yl
456
+ a = 0
457
+ (3)
458
+ for l = 0, 1, · · · , L and a = 1, 2, · · · , D.
459
+ Proof:
460
+ As shown in [25], it is enough for us to optimize
461
+ over the Markovian policies for Problem 1. Since the process
462
+ is not affected by a shift in time, we can define the probabilistic
463
+ scheduling policy where f l
464
+ a denotes the probability of choosing
465
+ l channels when the AoI of single source is at state a. The
466
+ normalization constraint of the probabilistic scheduling policy
467
+ requires
468
+ L
469
+
470
+ l=0
471
+ f l
472
+ a = 1 and f l
473
+ a ⩾ 0 for all a.
474
+ Notice that the system state can be fully characterized by a
475
+ one-dimensional Markov chain with age A[t] as state. Given
476
+ the current state information A[t], the system state at the next
477
+ time slot A[t+1] depends only on the current state A[t] (with
478
+ no dependence on earlier states) and the current action u[t].
479
+ In addition, the objective and constraints only depend on the
480
+ current state and action. So an equivalent MDP problem can
481
+ be formulated. Let λa2
482
+ a1 denote the transition probability from
483
+ state a1 to a2, and define ¯µ ≜ 1 − µ as the probability of
484
+ channel failure. Then based on the channel success model,
485
+ λa2
486
+ a1 =
487
+
488
+
489
+
490
+
491
+
492
+
493
+
494
+
495
+
496
+
497
+
498
+
499
+
500
+
501
+
502
+
503
+
504
+
505
+
506
+
507
+
508
+
509
+
510
+
511
+
512
+
513
+
514
+ L
515
+
516
+ l=1
517
+ f l
518
+ a1 ¯µl,
519
+ 1 ≤ a1 ≤ D − 1, a2 = a1 + 1
520
+ L
521
+
522
+ l=1
523
+ f l
524
+ a(1 − ¯µl),
525
+ a1 = 1, · · · , D, a2 = 1
526
+ L
527
+
528
+ l=1
529
+ f l
530
+ D(1 − ¯µl),
531
+ a1 = D, a2 = D
532
+ 0,
533
+ otherwise.
534
+ (4)
535
+ Since there are finitely many states, there exists a stationary
536
+ distribution π(a) for every a. Let C be the set of all recurrent
537
+ 6In practice, moderate level of D is enough so that the dimension of LP
538
+ won’t be large. Also, when there is only age violation related objective and
539
+ constraints, it’s enough to set D = d + 1. See III-C and IV-C for references.
540
+ states, then C is irreducible and closed, thus C is positive
541
+ recurrent. When a ∈ C the stationary distribution π(a) is equal
542
+ to the long term average lim
543
+ T →∞
544
+ 1
545
+ T
546
+ T
547
+
548
+ t=1
549
+ 1{A[t] = a} independent
550
+ of the starting point. When state a /∈ C, then both the stationary
551
+ distribution and the long term average are equal to zero. So the
552
+ optimization problem is equivalent to the following constraint
553
+ MDP problem:
554
+ min
555
+ f la
556
+ D
557
+
558
+ a=1
559
+ π(a)ω0(a)
560
+ s.t:
561
+ D
562
+
563
+ a=1
564
+ L
565
+
566
+ l=0
567
+ π(a)f l
568
+ al ≤ b
569
+ D
570
+
571
+ a=1
572
+ π(a)ωk(a) ≤ ck, k = 1, · · · , K
573
+ (5)
574
+ L
575
+
576
+ l=0
577
+ f l
578
+ a = 1, f l
579
+ a ⩾ 0
580
+ ∀a ≤ D, l ≤ L
581
+ (6)
582
+ H · Π = Π,
583
+ 1 · Π = 1
584
+ (7)
585
+ where Π = [π(1), · · · , π(D)]T is the stationary distribution of
586
+ the Markov Chain and H is the D × D transition matrix with
587
+ hij = λi
588
+ j. Let us define yl
589
+ a = π(a)f l
590
+ a, then π(a) =
591
+ L
592
+
593
+ l=0
594
+ yl
595
+ a for
596
+ a ≤ D. Then the constraint 5 becomes:
597
+ D
598
+
599
+ a=1
600
+ L
601
+
602
+ l=0
603
+ yl
604
+ aωk(a) ≤ ck, k = 1, · · · , K.
605
+ The
606
+ normalization
607
+ constraint
608
+ in
609
+ Equation
610
+ 7
611
+ requires
612
+ D
613
+
614
+ a=1
615
+ L
616
+
617
+ l=0
618
+ yl
619
+ a = 1. Substituting yl
620
+ a into the CMDP problem and
621
+ after simplifying, we establish the equivalency of the Linear
622
+ Programming problem. After obtaining the solution y, we
623
+ let f l
624
+ a = yl
625
+ a/π(a) for π(a) ̸= 0.States a with π(a) = 0, are
626
+ transient states, and the actions at these states do not affect
627
+ the average results. For those states we adopt a simple policy
628
+ as in Equation 3, then the constraint 7 is also satisfied.
629
+ C. Characterization and Insights on Age-Optimal Schedulers
630
+ Our general framework encompasses a wide range of objec-
631
+ tives and constraints for different choices of ωk(·) functions
632
+ using different age and age-violation metrics. In this section,
633
+ we focus on two important problems that can be expressed
634
+ within our framework: average age minimization and age-
635
+ violation-rate minimization. This effort will enable us to
636
+ characterize their optimal schedulers and gain insights into
637
+ their nature.
638
+ Optimal scheduler minimizing average age: When we set
639
+ ω0(a) = a in (1), the objective of the optimization problem
640
+ becomes to minimize the average age
641
+ lim
642
+ T →∞
643
+ 1
644
+ T
645
+ T
646
+
647
+ t=1
648
+ E{A[t]} =
649
+ D
650
+
651
+ a=1
652
+ a π(a).
653
+
654
+ 0
655
+ 5
656
+ 10
657
+ 15
658
+ AoI
659
+ 0
660
+ 2
661
+ 4
662
+ 6
663
+ 8
664
+ Average number of activated channels
665
+ =0.12
666
+ =0.1
667
+ =0.08
668
+ =0.04
669
+ Figure 2. Optimal number of channels to choose to minimize average AoI
670
+ when b = 2.
671
+ For this problem formulation, we retain the energy constraint
672
+ lim
673
+ T →∞
674
+ 1
675
+ T
676
+ T
677
+
678
+ t=1
679
+ E [u (A[t])] ≤ b; but do not need additional age
680
+ constraints. Hence, ωk(a) = 0 and ck = 0, for all k and a.
681
+ Figure 2 depicts the average number of activated channels
682
+ of the average-age optimal scheduler as a function of the age
683
+ states under different channel success probabilities µ for the
684
+ energy constraint b = 2. We will further discuss these results
685
+ at the end of this section in comparison with the next scheduler
686
+ of interest.
687
+ Optimal scheduler minimizing age-violation-rate: Setting
688
+ ω0(a) = 1{a > τ} n (1), the objective becomes minimizing
689
+ the average age-violation-rate
690
+ lim
691
+ T →∞
692
+ 1
693
+ T
694
+ T
695
+
696
+ t=1
697
+ E{1{A[t] > τ}} =
698
+ D
699
+
700
+ a=τ+1
701
+ π(a).
702
+ As before, we keep the energy constraint, but do not need
703
+ additional age constraints. Hence, ωk(a) = 0 and ck = 0, for
704
+ all k and a.
705
+ With this, the problem becomes minimizing the age-
706
+ violation-rate under an energy constraint. Unlike in the previ-
707
+ ous problem, our goal is not to minimize the average age but
708
+ to avoid age-violation events. In this scenario, we can view
709
+ all the states with a > τ as state τ + 1, so it’s enough to set
710
+ D = τ + 1.
711
+ Figure 3 depicts the average number of activated channels
712
+ of the violation-rate optimal scheduler as a function of the age
713
+ states under different channel success probabilities µ for age
714
+ threshold τ = 8 and the same energy constraint b = 2. Next,
715
+ we compare the optimal policies of these two schedulers and
716
+ discuss the insights that can be gained from their study.
717
+ Insights on the two optimal schedulers: We start by noting
718
+ the similarities of the optimal policy under both scenarios:
719
+ (i) Each optimal policy is a probabilistic combination of at
720
+ most two deterministic policies, which matches the result
721
+ that the number of randomization is at most the number
722
+ of constraints, as shown in [25].
723
+ (ii) For each scenario, as the channel success probability
724
+ increases, the corresponding optimal policy starts trans-
725
+ 2
726
+ 4
727
+ 6
728
+ 8
729
+ AoI
730
+ 0
731
+ 5
732
+ 10
733
+ 15
734
+ 20
735
+ 25
736
+ Average number of activated channels
737
+ =0.12
738
+ =0.1
739
+ =0.08
740
+ =0.04
741
+ Figure 3. Optimal number of channels to choose to minimize AoI violation
742
+ rate when b = 2 and τ = 8.
743
+ mitting at lower age levels, and also tends to choose
744
+ more channels at the same age level. This is a somewhat
745
+ counter-intuitive characteristic that indicates that the opti-
746
+ mal policy should be more active and active earlier when
747
+ the channels are more reliable.
748
+ (iii) The optimal policy in each scenario is idle when AoI is
749
+ relatively small. This is meaningful once we observe that,
750
+ when the age is relatively small, a successful transmission
751
+ will not benefit the objective as much as when the age
752
+ is large. Hence, the optimal scheduler saves energy for
753
+ larger age states.
754
+ However, we also notice differences between the two sets
755
+ of schedulers:
756
+ (i) The optimal policy in the average age minimization prob-
757
+ lem has an activation function u∗(·) that is monotone non-
758
+ decreasing with increasing age state. On the other hand,
759
+ the monotonicity does not hold in the age violation rate
760
+ minimization problem. This difference comes from the
761
+ non-convex nature of the the age violation rate function
762
+ in the latter case. In [25] and many related works (e.g.,
763
+ [9], [34]), the authors exploit the monotone structure and
764
+ threshold nature of the optimal scheduling policy for solv-
765
+ ing the CMDP, revealing insights as well as simplifying
766
+ the algorithm by using the convexity or concavity of the
767
+ objective functions. However, in our general treatment,
768
+ the objective functions, such as age violation rate, are not
769
+ necessarily convex or concave, which prevents us from
770
+ using the same approach. Hence, to obtain the optimal
771
+ policy, we use the generally applicable LP method despite
772
+ the higher computational complexity that it may require
773
+ in order to develop insights about the optimal solution.
774
+ (ii) In the average age minimization problem, the number
775
+ of activated channels of the optimal policy experiences
776
+ a sub-linear/concave like increase with respect to ages
777
+ after the age level that the number of activated channels
778
+ starts to be above zero. In contrast, the age violation rate
779
+ minimizing schedulers experience a super-linear/convex
780
+ like increasing with respect to age until the deadline level
781
+ τ. This difference can be interpreted as follows: in the age
782
+
783
+ 1
784
+ 2
785
+ 3
786
+ 4
787
+ 5
788
+ 6
789
+ AoI
790
+ 0
791
+ 5
792
+ 10
793
+ 15
794
+ 20
795
+ 25
796
+ Average number of activated channels
797
+ =0.1
798
+ =0.001
799
+ =0.0001
800
+ Figure 4.
801
+ Optimal number of channels to choose to minimize average age
802
+ under violation rate constraint when τ = 5, b = 3, µ = 0.2
803
+ violation rate minimization problem, the penalty happens
804
+ only when the age is beyond the age deadline, and hence
805
+ the optimal scheduler will be more aggressive as the
806
+ threshold level is approached from below. In contrast,
807
+ for the average age minimization problem, the number
808
+ of activated channels increases more gradually to balance
809
+ the tradeoff between consuming energy unnecessarily at
810
+ very low age levels and waiting too long to consume the
811
+ available energy, which yields an indefinitely increasing
812
+ cost.
813
+ These insights on the structure of the allocation functions of
814
+ the optimal schedulers can guide designers in restricting their
815
+ search to classes of functions with sufficiently flexible but also
816
+ tractable forms whenever the solution through the LP strategy
817
+ is not possible due to lack of prior statistical information as
818
+ well as computational resources.
819
+ To demonstrate how the age violation rate constraint effects
820
+ the shape of the scheduler more clearly, in Figure 4 we set
821
+ the objective function to be ω0(a) = a, the energy constraint
822
+ to be b = 3, and the channel success probability to be µ =
823
+ 0.2. In addition, we set ω1(a) = 1{a > τ}, where the age
824
+ deadline τ = 5. We set c1 = ϵ and show how the number of
825
+ activated channels changes over age states under different ϵ
826
+ levels. By adding and tightening the tolerance constraint, we
827
+ can see the transition from concave (or sublinear) to convex
828
+ (or superlinear) form. As such, the optimal scheduler becomes
829
+ more aggressive when the age increases. This reveals a trade-
830
+ off between the average age and the age-violation-rate, namely
831
+ that reducing the age violation rate calls for an increasingly
832
+ more aggressive allocation function.
833
+ IV. AGE-OPTIMAL MULTI-CHANNEL SCHEDULING FOR
834
+ MULTIPLE USERS
835
+ In this section, we extend our framework to the general
836
+ multi-user multi-channel age-optimal scheduling problem. As
837
+ before, this formulation allows us to cover a range of scenarios
838
+ depending on the choice for objective function and constraints.
839
+ To that end, we investigate the feasibility and stability region
840
+ of the optimal policy along with alternatives from related
841
+ literature associated with multi-user settings.
842
+ A. Problem Formulation
843
+ The formulation of the optimization problem for the multi-
844
+ user case is similar to single user case (1):
845
+ min
846
+ u(A)
847
+ lim
848
+ T →∞
849
+ 1
850
+ T
851
+ T
852
+
853
+ t=1
854
+ E [ω0(A[t])]
855
+ (8)
856
+ s.t :
857
+ lim
858
+ T →∞
859
+ 1
860
+ T
861
+ T
862
+
863
+ t=1
864
+ E [ui (A[t])] ≤ bi, i = 1, · · · , n,
865
+ lim
866
+ T →∞
867
+ 1
868
+ T
869
+ T
870
+
871
+ t=1
872
+ E [ωk (A[t])] ≤ ck, k = 1, · · · , K,
873
+ ui(A[t]) ∈ {0, 1, · · · , L}, i = 1, · · · , n,
874
+ n
875
+
876
+ i=1
877
+ ui(A[t]) ≤ L
878
+ where = (u1(A), · · · , un(A)) denotes the scheduling policy
879
+ at state A with ui(A) as the number of channels allocated to
880
+ source i. The weight functions ωk(·), k = 0, 1, · · · , K, map
881
+ the age states to cost values that capture age-related objectives
882
+ and constraints. Source nodes can have heterogeneous energy
883
+ constraints bi, which means node i can transmit over at most
884
+ bi channels per slot on average.
885
+ B. Performance analysis
886
+ Next, we establish the equivalence of the multi-user problem
887
+ formulation to a linear programming (LP) problem, as we
888
+ did for the single user case in Section III-B. To enable a
889
+ more compact notation, we will use a ≜ (a1, a2, · · · , an)
890
+ and l ≜ (l1, l2, · · · , ln) to denote values of A[t] and u(A),
891
+ respectively. We further define sets A ≜ {1, · · · , D}n, L ≜
892
+ {1, · · · , L}n, and L1 ≜ {l : lΣ ≤ L} where lΣ ≜
893
+ n
894
+
895
+ i=1
896
+ li.
897
+ Theorem 2: The solution of the multi-user age-optimization
898
+ problem (8) can be obtained by solving the following linear
899
+ programming problem:
900
+ min
901
+ yl
902
+ a
903
+
904
+ a∈A
905
+
906
+ l∈L1
907
+ yl
908
+ aω0(a)
909
+ s.t:
910
+
911
+ a∈A
912
+
913
+ l∈L1
914
+ yl
915
+ ali ≤ bi, i = 1, 2, · · · , n
916
+ 0 ≤ yl
917
+ a ≤ 1
918
+ ∀l ∈ L, a ∈ A
919
+ yl
920
+ a = 0
921
+ ∀l ∈ L/L1
922
+
923
+ a∈A
924
+
925
+ l∈L1
926
+ yl
927
+ a = 1
928
+
929
+ a∈A
930
+
931
+ l∈L1
932
+ yl
933
+ aωk(a) ≤ ck, k = 1, · · · , K
934
+ (9)
935
+ Qy = 0
936
+ where y is a column vector with yl
937
+ a as components and
938
+ Q represents the transition matrix associated with the age
939
+
940
+ dynamics, exactly in the same form as in the single-user case
941
+ (cf. Theorem 1).
942
+ If this LP is feasible and y is an optimal solution, then
943
+ the optimal policy u∗
944
+ i (a) is a probabilistic policy, whereby
945
+ the probability f l
946
+ a of choosing l channels for source nodes
947
+ i = 1, · · · , n when the AoI is at state a equals:
948
+ f l
949
+ a =
950
+
951
+
952
+
953
+
954
+
955
+
956
+
957
+
958
+
959
+ yl
960
+ a
961
+
962
+ l∈L
963
+ yl
964
+ a
965
+ ,
966
+ if
967
+
968
+ l∈L
969
+ yl
970
+ a ̸= 0
971
+ 1
972
+ |L|,
973
+ if
974
+
975
+ l∈L
976
+ yl
977
+ a = 0
978
+ for l ∈ L and a ∈ A.
979
+ Proof:
980
+ We will use f l
981
+ a to denote the probability of choosing
982
+ l = (l1, · · · , ln) channels for source nodes (1, · · · , n) when
983
+ the AoI is at state a. Thus �
984
+ l∈L
985
+ f l
986
+ a = 1, and f l
987
+ a ≥ 0 for all a.
988
+ Similarly as in Theorem 1, the constraint MDP problem with
989
+ n−dimensional Markov Chains for multi-user scheduling can
990
+ be generally formulated as:
991
+ min
992
+
993
+ a
994
+ π(a)ω0(a)
995
+ s.t:
996
+
997
+ a
998
+
999
+ l
1000
+ π(a)f l
1001
+ ali ≤ bi, i = 1, 2, · · · , n
1002
+ (10)
1003
+ f l
1004
+ a = 0
1005
+ ∀l ∈ L/L1
1006
+
1007
+ a
1008
+ π(a)ωk(a) ≤ ck
1009
+ k = 1, · · · , K
1010
+ H · Π = Π,
1011
+ 1 · Π = 1,
1012
+ (11)
1013
+ where the indices range over a ∈ A and l ∈ L; π(a) is the
1014
+ stationary distribution of state a; and ωk(a), k = 0, 1, · · · , K,
1015
+ are age related objective and cost functions. The constraints 10
1016
+ bound the average energy of nodes i by bi for i = 1, · · · , n.
1017
+ In the constraint 11, Π is a Dn × 1 stationary distribution
1018
+ vector with π(a), a
1019
+
1020
+ A as entries.7 H represents the
1021
+ Dn×Dn transaction matrix with hi,j equals the probability of
1022
+ transaction from the jth state in Π to the ith state in Π, which
1023
+ can be detailed by using the age evolution and channel success
1024
+ probability equations similarly as in Equation 4. Similarly, we
1025
+ will define
1026
+ yl
1027
+ a ≜ yl1,l2,··· ,ln
1028
+ a1,a2,··· ,an = π(a)f l
1029
+ a.
1030
+ By changing the value of the weight functions, we can get
1031
+ different AoI related metrics, but all are linear with respect to
1032
+ yl
1033
+ a. Then,
1034
+ π(a) =
1035
+
1036
+ l
1037
+ yl
1038
+ a,
1039
+ and the normalization constraint requires:
1040
+
1041
+ a
1042
+
1043
+ l
1044
+ yl
1045
+ a = 1.
1046
+ Substituting yl
1047
+ a into the CMDP problem, we obtain the equiv-
1048
+ alence of the LP problem.
1049
+ 7The existence of the stationary distribution follows by the same proof as
1050
+ in Theorem 1.
1051
+ C. Characterization and insights on multi-user scheduling
1052
+ problem with violation tolerance Constraints
1053
+ Since there is no closed form solution to the general age-
1054
+ optimal problem, we will study the multi-user single-channel
1055
+ scheduling feasibility problem with age-violation tolerance
1056
+ constraint as a common setting to investigate its performance
1057
+ and characteristics.
1058
+ In particular, we will compare the stability region of the
1059
+ optimal scheduler with a previously developed algorithm that
1060
+ was developed for the special case of multi-user single-channel
1061
+ setting [21]. To that end, we set L = 1 and bi > 1. Thus, all
1062
+ the energy constraints will be inactive, and we can focus on the
1063
+ tolerance constraint, as in [21]. Since we are only interested
1064
+ in feasibility, we set ω0(a) = 1 for all a. To express the
1065
+ age-violation rate constraints we define the weight functions
1066
+ ωk(a) =
1067
+
1068
+ 0,
1069
+ if ak ≤ τk
1070
+ 1,
1071
+ if ak ≥ τk + 1,
1072
+ and set ck = ϵk for k = 1, 2, · · · , K = n, to represent the
1073
+ heterogeneous age-violation tolerance level for the kth source.
1074
+ Then the constraint
1075
+
1076
+ a
1077
+ π(a)ωk(a) ≤ ck becomes
1078
+ πk(τk + 1) ≤ ϵk
1079
+ ∀k = 1, · · · , K = n,
1080
+ where πk(τk + 1) denotes the total probability (under the
1081
+ stationary distribution) that source k violates its age threshold
1082
+ τk. Since
1083
+ πk(τk+1) =
1084
+
1085
+ j1,...,jk−1,jk+1,...,jn
1086
+ π(j1, ...jk−1, τk+1, jk+1..., jn),
1087
+ the constraint (9) in the linear programming problem becomes
1088
+
1089
+ j1,...,jk−1,jk+1,...,jn
1090
+
1091
+ l
1092
+ yl
1093
+ j1,...,jk−1,τk+1,jk+1,...,jn ≤ ϵk.
1094
+ For the sake of easy visualization, we study the case with
1095
+ n = 2 users. In this case, the LP problem is formulated as:
1096
+ min
1097
+ 1
1098
+ s.t:
1099
+ 0 ≤ yl1,l2
1100
+ a1,a2 ≤ 1
1101
+ ∀l1, l2 = 0, 1
1102
+ yl1,l2
1103
+ a1,a2 = 0
1104
+ ∀l1 + l2 > 1
1105
+
1106
+ j
1107
+
1108
+ l1,l2
1109
+ yl1,l2
1110
+ τ1+1,j ≤ ϵ1
1111
+
1112
+ j
1113
+
1114
+ l1,l2
1115
+ yl1,l2
1116
+ j,τ2+1 ≤ ϵ2
1117
+ The numerical results can be seen in Figures5 and 6 for
1118
+ different parameters where the upper right area of the solid
1119
+ blue line is the stability region of the optimal scheduler.
1120
+ These typical examples reveal the non-negligible gap between
1121
+ the performance of the optimal scheduler and the previously
1122
+ proposed design, even for a small two user setting.
1123
+ This motivates the search for new algorithms that can
1124
+ perform closer to the optimal scheduler, even when the channel
1125
+ statistics are unknown a priori. This is performed in the next
1126
+ section along with further discussion about these numerical
1127
+ results after we discuss our online scheduling algorithm.
1128
+
1129
+ Before we proceed, we note even the above numerical
1130
+ results are for two-user single-channel scheduling problem
1131
+ under tolerance constraints for visualization purposes, our
1132
+ methods apply to the more general multi-user multi-channel
1133
+ scheduling problem under violation tolerance and energy
1134
+ constraints. Although the computational complexity may be
1135
+ relatively high for the LP solution compared to other solutions
1136
+ that exploit the special structure of particular problems, as we
1137
+ mentioned above, due to the non-convexity and non-concavity
1138
+ of the tolerance constraints, the monotone and threshold
1139
+ structure of the optimal policy does not hold. The Whittle
1140
+ Index approach (used, for example, in [29], [31]) which have
1141
+ relatively low complexity also does not apply to our multi-
1142
+ channel scheduling problems since each user in our setting
1143
+ is allowed to transmit over multiple channels simultaneously,
1144
+ whereby the Whittle’s Indexability condition does not hold.
1145
+ Using the generally applicable LP-based approach reveals key
1146
+ insights that can guide the designers in developing efficient
1147
+ schedulers for future multi-channel wireless technologies.
1148
+ V. ONLINE SCHEDULING UNDER UNKNOWN CHANNEL
1149
+ STATISTICS
1150
+ Until this point, we have assumed that the channel success
1151
+ probabilities are known when solving the optimization prob-
1152
+ lems. In this section, we use a Lyapunov-drift-plus-penalty
1153
+ approach(see [28]) to solve the multi-user online age related
1154
+ optimization problem in the scenario when only the current
1155
+ channel states are known, but the channel statistics are un-
1156
+ known.
1157
+ We will transfer all the energy and age-related constraints
1158
+ into the virtual queues and view the objective as a penalty term
1159
+ with parameter M. For the energy constraint of the source i,
1160
+ let us define the corresponding virtual queue as Q1,i[t], whose
1161
+ initial value is Q1,i[0] = 0 and update equation is:
1162
+ Q1,i[t + 1] = (Q1,i[t] + ui (A[t]) − bi)+ .
1163
+ Similarly, we define the virtual queue Q2,k[t] for the kth age-
1164
+ related constraint, whose initial value is Q2,k[0] = 0 and
1165
+ update equation is:
1166
+ Q2,k[t + 1] = (Q2,k[t] + ωk (A[t]) − ck)+ .
1167
+ Generically, if the virtual queue Q1,i[t] is stable, then its input
1168
+ rate lim
1169
+ T →∞
1170
+ 1
1171
+ T
1172
+ T
1173
+
1174
+ t=1
1175
+ E [ui (A[t])] will be less than its output rate
1176
+ bi [28], so that the corresponding constraint can be satisfied.
1177
+ Define the state of both virtual queues and age at time t
1178
+ as Q[t] = (Q1,1[t], · · · , Q1,n[t], Q2,1[t], · · · , Q2,K[t], A[t]).
1179
+ Based on the virtual queues, we will define the quadratic
1180
+ Lyapunov function as:
1181
+ V [t] = 1
1182
+ 2(
1183
+ n
1184
+
1185
+ i=1
1186
+ Q2
1187
+ 1,i[t] +
1188
+ K
1189
+
1190
+ k=1
1191
+ Q2
1192
+ 2,k[t]),
1193
+ and develop an online algorithm to greedily minimize
1194
+ the upper bound of the Lyapunov-drift-plus-penalty func-
1195
+ tion ∆V (q) + ME[ω0(a)] given the current state q
1196
+ =
1197
+ (q1,1, · · · , q1,n, q2,1, · · · , q2,K, a), where:
1198
+ ∆V (q) = E[V [t] − V [t − 1]|Q[t] = q].
1199
+ We consider the multi-user single-channel scheduling prob-
1200
+ lem under tolerance constraints as a specific example to
1201
+ present the design. Since there are no energy constraints,
1202
+ we do not need the set of virtual queues {Q1,i[t]}i. In
1203
+ order to express the kth violation rate constraint for source
1204
+ k = 1, · · · , n, we let ωk (A[t]) = 1 (Ak[t + 1] > τk) and
1205
+ ck = ϵk. Then the virtual queue Q2,k[t], whose initial value
1206
+ is Q2,k[t] = 0, updates as follows:
1207
+ Q2,k[t + 1] = (Q2,k[t] + 1 (Ak[t + 1] > τk) − ϵk)+ ,
1208
+ where Ak[t + 1] = 1 + Ak[t](1 − Sk[t]Uk[t]); Sk[t] represents
1209
+ the channel success; Uk[t] represents whether the source is
1210
+ scheduled to transmit or not. If virtual queue Q2,k[t] is stable,
1211
+ its input rate, the threshold violation rate πk(τk + 1) =
1212
+ limT →∞ 1
1213
+ T
1214
+ �T
1215
+ t=1 1 (Ak[t + 1] > τk) , will be less than its
1216
+ output rate ϵk.
1217
+ The conditional Lyapunov drift can be bounded as follows:
1218
+ ∆V (q)
1219
+
1220
+ n
1221
+
1222
+ k=1
1223
+ q2,kE [Rk − ϵk|q2,k] +
1224
+ n
1225
+
1226
+ k=1
1227
+ E
1228
+
1229
+ (Rk − ϵk)2
1230
+ 2
1231
+ |q2,k
1232
+
1233
+ ,
1234
+ where Rk
1235
+ ∆= 1{1 + Ak (1 − SkCk) > τk}. At every time slot
1236
+ t, we can develop an online algorithm as summarized below
1237
+ to greedily minimize the upper bound of the Lyapunov drift
1238
+ given the queue lengths Q[t − 1] and A[t − 1] since there is
1239
+ no objective or penalty term in this case.
1240
+ Algorithm 1 A Heuristic Scheduling Policy
1241
+ 1: Input current system state: Ai[t],Qi[t].
1242
+ 2: Define available transmission decision set: only one Ui[t]
1243
+ can be 1.
1244
+ 3: Choose U[t] to minimize the upper bound of Lyapunov
1245
+ drift function in the above inequality.
1246
+ 4: Update queue lengths for next time slot.
1247
+ Again, for the sake of easy visualization, we will only
1248
+ present the simulation results for the two-user online schedul-
1249
+ ing problem under age tolerance constraints, but the online
1250
+ algorithm can be simply applied to any number of sources.
1251
+ The simulation results are illustrated in Fig 5 and Fig 6 for
1252
+ different parameters where the upper right area of the dash-dot
1253
+ purple line is the stability region of the online scheduler when
1254
+ the channel condition µi. The comparison will be in the next
1255
+ section.
1256
+ VI. COMPARISON OF STABILITY REGIONS UNDER AGE
1257
+ VIOLATION CONSTRAINTS
1258
+ In this section, we compare the performance of three dif-
1259
+ ferent algorithms for the two-user single channel scheduling
1260
+ feasibility problem under age violation tolerance constraints.
1261
+ These are: the optimal scheduler from Section IV; the prior
1262
+
1263
+ design from [21] developed for a single-channel multi-user
1264
+ setting; and our online scheduler from Section V that does
1265
+ not require channel statistics.
1266
+ We first focus on the case when the two source nodes are
1267
+ symmetric. In Figure 5, there are two source nodes with the
1268
+ same age thresholds of τ1 = τ2 = 2 and the same channel
1269
+ success probabilities of µ1 = µ2 = 0.85. The upper right
1270
+ area of the blue line is the stability region for the optimal
1271
+ scheduling algorithm in Section IV-C. The yellow and orange
1272
+ lines correspond to the algorithm in [21] and capture the two
1273
+ cases when the rate vector does or does not possess a special
1274
+ property (called step-down rate vector). The purple line marks
1275
+ the stability region for the online algorithm when the channel
1276
+ conditions µ1, µ2 are unknown. Several observations are in
1277
+ order from these simulation results:
1278
+ (i) The stability regions are all symmetric, as can be expected
1279
+ due to the homogeneous deadline thresholds and channel
1280
+ conditions.
1281
+ (ii) The optimum policy (blue line) outperforms other poli-
1282
+ cies, with markedly better performance in cases where
1283
+ the tolerance levels are greatly different from each other.
1284
+ (iii) The online algorithm (purple line) performs very closely
1285
+ to the optimal policy, experiencing a small performance
1286
+ loss only at some extreme range of tolerance levels.
1287
+ (iv) When compared with the algorithms from [21](yellow
1288
+ and red lines), the online algorithm performs particularly
1289
+ better when one of the tolerance rates is smaller than the
1290
+ corresponding channel loss probability, as observed by
1291
+ the vertical gap between purple and yellow lines.
1292
+ (v) The online and optimal policies are continuous with
1293
+ respect to the tolerance level, which eliminates the need
1294
+ to check if the tolerance rate vector satisfies certain
1295
+ properties, such as the step-down rate condition in [21].
1296
+ To compare the advantages and disadvantages of the al-
1297
+ gorithms under non-homogeneous scenarios, in Figure 6, we
1298
+ consider two source nodes with asymmetric age thresholds of
1299
+ τ1 = 2, τ2 = 4 and a common channel success probability of
1300
+ µ1 = µ2 = 0.85. Since the violation rate depends on both
1301
+ the age thresholds and the channel success probabilities, this
1302
+ is a non-homogeneous scenario even though µ1 = µ2. In this
1303
+ figure, in contrast to the previous figure, we can further see
1304
+ that the optimal policy outperforms others when one of the
1305
+ 0
1306
+ 0.2
1307
+ 0.4
1308
+ 0.6
1309
+ 0.8
1310
+ 1
1311
+ 1
1312
+ 0
1313
+ 0.2
1314
+ 0.4
1315
+ 0.6
1316
+ 0.8
1317
+ 1
1318
+ 2
1319
+ Optimal scheduling algrithm
1320
+ Algorithm proposed in [21]
1321
+ under general rate vector
1322
+ Algorithm proposed in [21]
1323
+ under step-down rate vector
1324
+ Online scheduling algorithm
1325
+ Figure 5. Stability region (upper-righter) comparison for symmetric case.
1326
+ 0
1327
+ 0.2
1328
+ 0.4
1329
+ 0.6
1330
+ 0.8
1331
+ 1
1332
+ 1
1333
+ 0
1334
+ 0.2
1335
+ 0.4
1336
+ 0.6
1337
+ 0.8
1338
+ 1
1339
+ 2
1340
+ Optimal scheduling algrithm
1341
+ Algorithm proposed in [21]
1342
+ under general rate vector
1343
+ Algorithm proposed in [21]
1344
+ under step-down rate vector
1345
+ Online scheduling algorithm
1346
+ Figure 6. Stability region (upper-righter) comparison for asymmetric case.
1347
+ tolerance constraints is very strict, namely when ϵ1 approaches
1348
+ 1. In this regime, the feasible tolerance level ϵ2 of user 2
1349
+ other algorithms is bounded away from zero while the optimal
1350
+ algorithm decreases towards zero.
1351
+ These simulation results are typical of other circumstances,
1352
+ with the common observation that our online scheduler per-
1353
+ forms close to the optimal scheduler and typically non-
1354
+ negligibly better than the most closely related state-of-art
1355
+ algorithm from [21], despite the fact that it operates without
1356
+ the knowledge of channel statistics that is assumed in the other
1357
+ designs.
1358
+ VII. CONCLUSIONS
1359
+ In this paper, we considered a general class of age-optimal
1360
+ scheduling problems for multi-source multi-channel commu-
1361
+ nication. We formulated the generic age-optimization problem
1362
+ with flexible weight functions ωk under energy and tolerance
1363
+ constraints in the form of a CMDP. We solved this generic
1364
+ problem, which a usual threshold-based structure policy does
1365
+ not apply, by relating it to the solution an associated linear
1366
+ programming problem using the powerful theory of CMDPs.
1367
+ Then, we focused on the special case of single-source multi-
1368
+ channel scenario to investigate the characteristics of optimal
1369
+ scheduler for the important special cases of average-age and
1370
+ violation-rate minimization.
1371
+ Our investigations revealed several interesting insights, in-
1372
+ cluding the observation that age-violation-rate minimizing
1373
+ scheduler employs a super-linearly like growing energy al-
1374
+ location strategy with increasing age, as opposed to the
1375
+ sub-linearly like growing allocation for the average-age-
1376
+ minimizing scheduler. These insights may provide useful
1377
+ guidelines for IoT network designers in developing effective
1378
+ update strategies based on different sensitivities of applications
1379
+ to age performance.
1380
+ We also studied the special case of multi-source single-
1381
+ channel scheduling problem with age violation rate constraints
1382
+ to investigate the feasibility region of the optimal scheduler
1383
+ together with that of most closely related prior works. Finally,
1384
+ we have developed an online scheduler that does not require
1385
+ the knowledge of channel statistics, and compared its perfor-
1386
+ mance to the optimal scheduler through simulations to observe
1387
+ that it performs closely to the optimal scheduler despite its lack
1388
+ of information on channel statistics.
1389
+
1390
+ REFERENCES
1391
+ [1] F. C. Commission et al., “Fcc adopts rules to facilitate next generation
1392
+ wireless technologies,” FCC, July, vol. 14, 2016.
1393
+ [2] S. Kaul, M. Gruteser, V. Rai, and J. Kenney, “Minimizing age of
1394
+ information in vehicular networks,” in 2011 8th Annual IEEE communi-
1395
+ cations society conference on sensor, mesh and ad hoc communications
1396
+ and networks.
1397
+ IEEE, 2011, pp. 350–358.
1398
+ [3] S. Kaul, R. Yates, and M. Gruteser, “Real-time status: How often should
1399
+ one update?” in 2012 Proceedings IEEE INFOCOM.
1400
+ IEEE, 2012.
1401
+ [4] A. Kosta, N. Pappas, V. Angelakis et al., “Age of information: A new
1402
+ concept, metric, and tool,” Foundations and Trends® in Networking,
1403
+ vol. 12, no. 3, pp. 162–259, 2017.
1404
+ [5] H. Chen, Y. Gu, and S.-C. Liew, “Age-of-information dependent random
1405
+ access for massive iot networks,” in INFOCOM WKSHPS. IEEE, 2020,
1406
+ pp. 930–935.
1407
+ [6] X. Zhou, I. Koprulu, A. Eryilmaz, and M. J. Neely, “Efficient dis-
1408
+ tributed mac for dynamic demands: Congestion and age based designs,”
1409
+ IEEE/ACM Transactions on Networking, pp. 1–14, 2022.
1410
+ [7] B. Abolhassani, J. Tadrous, A. Eryilmaz, and E. Yeh, “Fresh caching of
1411
+ dynamic content over the wireless edge,” IEEE/ACM Transactions on
1412
+ Networking, 2022.
1413
+ [8] R. Liu, E. Yeh, and A. Eryilmaz, “Proactive caching for low access-
1414
+ delay services under uncertain predictions,” Proceedings of the ACM on
1415
+ Measurement and Analysis of Computing Systems, vol. 3, no. 1, 2019.
1416
+ [9] H. Tang, J. Wang, L. Song, and J. Song, “Scheduling to minimize age
1417
+ of information in multi-state time-varying networks with power con-
1418
+ straints,” in 2019 57th Annual Allerton Conference on Communication,
1419
+ Control, and Computing (Allerton).
1420
+ IEEE, 2019, pp. 1198–1205.
1421
+ [10] B. Sombabu and S. Moharir, “Age-of-information aware scheduling for
1422
+ heterogeneous sources,” in Proceedings of the 24th Annual International
1423
+ Conference on Mobile Computing and Networking, 2018, pp. 696–698.
1424
+ [11] M. Li, C. Chen, H. Wu, X. Guan, and X. Shen, “Age-of-information
1425
+ aware scheduling for edge-assisted industrial wireless networks,” IEEE
1426
+ Transactions on Industrial Informatics, vol. 17, no. 8, pp. 5562–5571,
1427
+ 2020.
1428
+ [12] P. R. Jhunjhunwala and S. Moharir, “Age-of-information aware schedul-
1429
+ ing,” in SPCOM.
1430
+ IEEE, 2018.
1431
+ [13] D. Han, W. Chen, and Y. Fang, “Joint channel and queue aware
1432
+ scheduling for latency sensitive mobile edge computing with power
1433
+ constraints,” IEEE Transactions on Wireless Communications, vol. 19,
1434
+ no. 6, pp. 3938–3951, 2020.
1435
+ [14] L. Hu, Z. Chen, Y. Dong, Y. Jia, L. Liang, and M. Wang, “Status update
1436
+ in iot networks: Age-of-information violation probability and optimal
1437
+ update rate,” IEEE Internet of Things Journal, vol. 8, no. 14, 2021.
1438
+ [15] N. Pappas and M. Kountouris, “Delay violation probability and age of
1439
+ information interplay in the two-user multiple access channel,” in 20th
1440
+ International Workshop on SPAWC.
1441
+ IEEE, 2019, pp. 1–5.
1442
+ [16] M. Chen, Y. Xiao, Q. Li, and K.-c. Chen, “Minimizing age-of-
1443
+ information for fog computing-supported vehicular networks with deep
1444
+ q-learning,” in ICC 2020-2020 IEEE International Conference on Com-
1445
+ munications (ICC).
1446
+ IEEE, 2020, pp. 1–6.
1447
+ [17] Y. Sun and B. Cyr, “Sampling for data freshness optimization: Non-
1448
+ linear age functions,” Journal of Communications and Networks, vol. 21,
1449
+ no. 3, pp. 204–219, 2019.
1450
+ [18] M. Emara, H. Elsawy, and G. Bauch, “A spatiotemporal model for peak
1451
+ aoi in uplink iot networks: Time versus event-triggered traffic,” IEEE
1452
+ internet of things journal, vol. 7, no. 8, pp. 6762–6777, 2020.
1453
+ [19] B. Li, A. Eryilmaz, and R. Srikant, “Emulating round-robin in wireless
1454
+ networks,” in Proceedings of the 18th ACM International Symposium
1455
+ on Mobile Ad Hoc Networking and Computing, 2017, pp. 1–10.
1456
+ [20] Y. Dong, Z. Chen, S. Liu, P. Fan, and K. B. Letaief, “Age-upon-decisions
1457
+ minimizing scheduling in internet of things: To be random or to be
1458
+ deterministic?” IEEE Internet of Things Journal, vol. 7, no. 2, 2019.
1459
+ [21] C. Li, Q. Liu, S. Li, Y. Chen, Y. T. Hou, and W. Lou, “On scheduling
1460
+ with aoi violation tolerance,” in IEEE INFOCOM, 2021, pp. 1–9.
1461
+ [22] M. Song, H. H. Yang, H. Shan, J. Lee, H. Lin, and T. Q. Quek, “Analysis
1462
+ of aoi violation probability in wireless networks,” in 17th ISWCS. IEEE,
1463
+ 2021.
1464
+ [23] R. Devassy, G. Durisi, G. C. Ferrante, O. Simeone, and E. Uysal-
1465
+ Biyikoglu, “Delay and peak-age violation probability in short-packet
1466
+ transmissions,” in ISIT.
1467
+ IEEE, 2018, pp. 2471–2475.
1468
+ [24] L. Hu, Z. Chen, Y. Dong, Y. Jia, M. Wang, L. Liang, and C. Chen,
1469
+ “Optimal status update in iot systems: An age of information violation
1470
+ probability perspective,” in VTC2020-Fall.
1471
+ IEEE, 2020, pp. 1–5.
1472
+ [25] E. Altman, Constrained Markov decision processes: stochastic modeling.
1473
+ Routledge, 1999.
1474
+ [26] D. A. Dolgov and E. H. Durfee, “Stationary deterministic policies for
1475
+ constrained mdps with multiple rewards, costs, and discount factors,” in
1476
+ IJCAI, vol. 19.
1477
+ Citeseer, 2005, pp. 1326–1331.
1478
+ [27] H. Satija, P. Amortila, and J. Pineau, “Constrained markov decision
1479
+ processes via backward value functions,” in International Conference
1480
+ on Machine Learning.
1481
+ PMLR, 2020, pp. 8502–8511.
1482
+ [28] M. J. Neely, “Stochastic network optimization with application to
1483
+ communication and queueing systems,” Synthesis Lectures on Commu-
1484
+ nication Networks, vol. 3, no. 1, pp. 1–211, 2010.
1485
+ [29] Y.-P. Hsu, “Age of information: Whittle index for scheduling stochastic
1486
+ arrivals,” in ISIT.
1487
+ IEEE, 2018, pp. 2634–2638.
1488
+ [30] V. Tripathi and S. Moharir, “Age of information in multi-source sys-
1489
+ tems,” in GLOBECOM.
1490
+ IEEE, 2017, pp. 1–6.
1491
+ [31] Y. Zou, K. T. Kim, X. Lin, and M. Chiang, “Minimizing age-of-
1492
+ information in heterogeneous multi-channel systems: A new partial-
1493
+ index approach,” in Proceedings of the Twenty-second International
1494
+ Symposium on Theory, Algorithmic Foundations, and Protocol Design
1495
+ for Mobile Networks and Mobile Computing, 2021, pp. 11–20.
1496
+ [32] A. Elgabli, H. Khan, M. Krouka, and M. Bennis, “Reinforcement
1497
+ learning based scheduling algorithm for optimizing age of information
1498
+ in ultra reliable low latency networks,” in ISCC.
1499
+ IEEE, 2019, pp. 1–6.
1500
+ [33] M. Li, C. Chen, C. Hua, and X. Guan, “Learning-based autonomous
1501
+ scheduling for aoi-aware industrial wireless networks,” IEEE Internet of
1502
+ Things Journal, vol. 7, no. 9, pp. 9175–9188, 2020.
1503
+ [34] H. Tang, J. Wang, L. Song, and J. Song, “Minimizing age of infor-
1504
+ mation with power constraints: Multi-user opportunistic scheduling in
1505
+ multi-state time-varying channels,” IEEE Journal on Selected Areas in
1506
+ Communications, vol. 38, no. 5, pp. 854–868, 2020.
1507
+
GtAyT4oBgHgl3EQfrfmY/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
H9FJT4oBgHgl3EQfuy1Y/content/tmp_files/2301.11623v1.pdf.txt ADDED
@@ -0,0 +1,1297 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Higher-Order Patterns Reveal Causal Timescales
2
+ of Complex Systems
3
+ Luka V. Petrovi´c1, Anatol Wegner2, and Ingo Scholtes1,2
4
+ 1Data Analytics Group, University of Zurich, Z¨urich, Switzerland
5
+ 2Center for Artificial Intelligence and Data Science (CAIDAS),
6
+ Julius-Maximilians-Universit¨at W¨urzburg, W¨urzburg, Germany
7
+ January 30, 2023
8
+ Abstract
9
+ The analysis of temporal networks heavily depends on the analysis of time-respecting paths.
10
+ However, before being able to model and analyze the time-respecting paths, we have to infer
11
+ the timescales at which the temporal edges influence each other. In this work we introduce
12
+ temporal path entropy, an information theoretic measure of temporal networks, with the aim to
13
+ detect the timescales at which the causal influences occur in temporal networks. The measure
14
+ can be used on temporal networks as a whole, or separately for each node. We find that the
15
+ temporal path entropy has a non-trivial dependency on the causal timescales of synthetic and
16
+ empirical temporal networks.
17
+ Furthermore, we notice in both synthetic and empirical data
18
+ that the temporal path entropy tends to decrease at timescales that correspond to the causal
19
+ interactions. Our results imply that timescales relevant for the dynamics of complex systems
20
+ can be detected in the temporal networks themselves, by measuring temporal path entropy.
21
+ This is crucial for the analysis of temporal networks where inherent timescales are unavailable
22
+ and hard to measure.
23
+ 1
24
+ Introduction
25
+ The research of dynamic complex systems has in recent years advanced beyond static graph repre-
26
+ sentations [Lambiotte et al., 2019, Battiston et al., 2020]. The focus has shifted to various general-
27
+ izations of diadic interactions in graphs: multiple types of interactions in multilayer network [Kivel¨a
28
+ et al., 2014], multibody interactions in the form of simplicial complexes and hypergraphs [Petri and
29
+ Barrat, 2018] and models that incorporate concepts of memory [Scholtes et al., 2014, Lambiotte
30
+ et al., 2015, Williams et al., 2022]. Such generalized relationships allow us to better model complex
31
+ systems, because they can represent richer data.
32
+ Temporal networks are one kind of such rich data; they record not only who interacted with
33
+ whom, but also when each interaction happened. They bring us closer to understanding the dynamics
34
+ of complex systems, but require us to perform analysis beyond the static networks approach [Holme
35
+ and Saram¨aki, 2012, Holme, 2015]. The time information can yield valuable insights on its own [Goh
36
+ and Barab´asi, 2008], and, although initially the temporal and topological aspects of temporal net-
37
+ works were mostly studied independently, even richer insights are hidden in the coupling of the
38
+ temporal and topological patterns [Ceria et al., 2022]. Such coupling can affect the statistics of
39
+ 1
40
+ arXiv:2301.11623v1 [physics.soc-ph] 27 Jan 2023
41
+
42
+ time-respecting paths [Holme and Saram¨aki, 2012] in temporal networks, which impacts e.g., anal-
43
+ ysis of accessibility [Lentz et al., 2013], reachability [Badie-Modiri et al., 2020], spreading [Masuda
44
+ et al., 2013, Scholtes et al., 2014, Lambiotte et al., 2015, Badie-Modiri et al., 2022], clustering [Ros-
45
+ vall et al., 2014], centralities [Scholtes et al., 2016], and visualization [Perri and Scholtes, 2020].
46
+ Although there are many possible ways in which temporal and topological patterns can couple in
47
+ complex systems, one of the most basic cases is when an incoming temporal edge to a node causes
48
+ a change of frequencies of the edges emanating from the node within a given time-window. For
49
+ instance, in a communication network we expect an incoming message to induce a outgoing message
50
+ on the same topic, e.g. in the form of a reply, within a certain time window reflecting the minimal
51
+ reaction time and memory of the recipient. However, information on the timescales relevant for the
52
+ temporal network dynamics is rarely available in real world settings.
53
+ In this work, we define an information theoretic measure to detect timescales at which interactions
54
+ in a complex system cause each other. We demonstrate its effectiveness in both synthetic and real
55
+ world data.
56
+ 2
57
+ Background
58
+ Let Γ = (V, E) be a temporal network consisting of a set of nodes V and a set of time-stamped edges
59
+ E ⊆ V × V × R. We denote the set of unique edges with E ⊂ V × V . A temporal edge (v, w, t) ∈ E
60
+ represents a direct link from node v to node w at time t. For simplicity, we assume that the temporal
61
+ edges are instantaneous, however the method and algorithms can be modified in a straightforward
62
+ fashion to the case where edges have finite duration. Formally, we call a sequence of time-stamped
63
+ edges (v1, w1, t1), . . . , (vk, wk, tk) a time-respecting path iff for all i ∈ {2, . . . , k} they satisfy the
64
+ following conditions [Pan and Saram¨aki, 2011, Holme and Saram¨aki, 2012, Casteigts et al., 2021]:
65
+ wi−1 = vi
66
+ (1)
67
+ τmin < ti − ti−1 < τmax.
68
+ (2)
69
+ The parameters τmin and τmax naturally introduce a timescale that affects all analyses of temporal
70
+ networks that are based on time-respecting paths.
71
+ The timescale has to be defined differently for processes on the temporal network or the processes
72
+ of the temporal network [Holme and Saram¨aki, 2012]. In the former case, the timescale is defined by
73
+ the process running on the temporal network, e.g. in the case of an epidemic that is spreading over
74
+ a temporal network of contacts, the timescale is a property of a disease, related to the time interval
75
+ in which a person is contagious and not related to the timescales at which contacts occur 1. In the
76
+ latter case, the timescale is part of the process of edge activation, and thus shapes the temporal
77
+ network itself. For example, information that is spreading between individuals is also affecting the
78
+ individuals’ choice with whom to share the information: a person would be more likely to share the
79
+ family-related information with a family member and work-related information with a colleague. In
80
+ this letter, we investigate the latter case, more specifically, we investigate whether interactions in a
81
+ complex system induce one another at a given timescale τ = [τmin, τmax].
82
+ In the literature, there exist a variety of definitions of timescales in temporal networks, as well as
83
+ a variety of methods aimed at detecting them. The various definitions of timescales are based on the
84
+ different structural features of temporal networks. One popular definition of timescales in temporal
85
+ networks is the approach based on splitting the network into time-slices and aggregating the edges
86
+ 1We note that the processes on and of the temporal network may interact [Gross and Sayama, 2009], and thus blur
87
+ the distinction.
88
+ 2
89
+
90
+ inside the time-interval [Caceres and Berger-Wolf, 2013, Darst et al., 2016]. In the same framework,
91
+ Ghasemian et al. [2016] and Taylor et al. [2016] investigate the limitations of detectability of cluster
92
+ structures dependent on the timescales of aggregation. Since this framework is based on aggregating
93
+ the temporal network into a sequence of static time-aggregated networks, it loses information of
94
+ the time-respecting paths and is therefore not in line with our aims. Other lines of research often
95
+ related to timescale detection are change point detection [Peixoto and Gauvin, 2018], and analysis
96
+ of large-scale structures. Gauvin et al. [2014] detects clusters and their temporal activations in a
97
+ temporal network using tensor decomposition. Similarly, Peixoto [2015] proposed a method to detect
98
+ the change points of cluster structure in a temporal network. Peixoto and Rosvall [2017] proposed
99
+ a method to simultaneously detect the clusters and timescales in temporal network, however, they
100
+ model the temporal network as a single sequence of tokens (similar to [Peixoto and Gauvin, 2018])
101
+ that represent temporal edges, and their timescale inference refers to the number of tokens in the
102
+ memory of a Markov chain that models such a sequence. In our view, these works focus on mesoscale
103
+ structures, and take a coarse grained view of temporal networks, while in this work, we propose a
104
+ complementary approach by focusing on local correlations between temporal edges incident on a
105
+ node and subsequent temporal edges emanating from it. Among the works that took a fine-grained
106
+ view, Williams et al. [2022] investigated pairwise correlations between the temporal edges. Different
107
+ from the approach that we took, they aggregate the network in time-slices as a preprocessing step,
108
+ and the timescale is defined as a maximum number of time slices back in time at which correlations
109
+ are detectable. Scholtes et al. [2016] found that correlations between edges on time-respecting paths
110
+ affect centralities; they modeled the time-respecting paths with higher-order models and found that
111
+ this approach improves the centrality rankings. They identified the issue of timescale detection in
112
+ the context of time-respecting paths, which our work addresses. Our work also complements the
113
+ work of Pfitzner et al. [2013] which introduces betweenness preference that can be used to study
114
+ over- and under-represented time-respected paths in temporal networks, but does not address the
115
+ problem of detecting the timescales at which these paths occur. To the best of our knowledge, our
116
+ work is the first to address the issue of timescale detection for time-respecting paths in temporal
117
+ networks.
118
+ 3
119
+ Temporal Path Entropy
120
+ We address the issue of timescale detection by analysing the statistics of time-respecting paths
121
+ Pk
122
+ τ (Γ) of length k at timescales τ = [τmin, τmax] in a temporal network Γ. We define “temporal path
123
+ entropy” H for paths (v0, v1, . . . , vk) of length k as the entropy of the last node vk conditional on
124
+ the sub-path (v0, v1, . . . , vk−1):
125
+ H = H(vk|v0, . . . , vk−1)
126
+ (3)
127
+ = H(v0, . . . , vk) − H(v0, . . . , vk−1),
128
+ (4)
129
+ where H(P) = − �
130
+ i pi ln(pi) is the Shannon entropy.
131
+ The identity in Eq. 4 can be obtained
132
+ by applying the chain rule (see Appendix for derivation). By definition, temporal path entropy
133
+ H measures uncertainty in the last step of time-respecting paths given the k − 1 previous steps.
134
+ A lower value of the entropy indicates a high correlation between the memory of time-respecting
135
+ paths and subsequent steps. Hence the τ for which the entropy reaches its minimum gives us the
136
+ timescale for which time-respecting paths become most predictable, i.e.
137
+ where the correlations
138
+ between subsequent temporal edges are the most pronounced. The entropy can also be defined for
139
+ a single node v, by simply fixing vk−1 = v, allowing for a more fine grained analysis that could be
140
+ important if nodes differ significantly with respect to the timescales they operate on. The intuition
141
+ 3
142
+
143
+ behind the temporal path entropy is to measure how much the target vk of an edge emanating
144
+ form the node vk−1 depends on the incoming paths that influenced it in the past. Testing those
145
+ dependencies at different timescales would thus point to the timescales at which the dependencies
146
+ are most pronounced. When we compute temporal path entropy for the whole temporal network,
147
+ we use all time-respecting paths of length k in the temporal network, while when we compute it for
148
+ a node v, we select only the paths where vk−1 = v.
149
+ 1.75
150
+ 2.00
151
+ 2.25
152
+ H[ nat ]
153
+ Synthetic-2
154
+ 0
155
+ 1
156
+ 2
157
+ WB-DE
158
+ 2
159
+ 3
160
+ HC-email
161
+ 50
162
+ 100
163
+ 150
164
+ 200
165
+ 250
166
+ 0.0
167
+ 0.5
168
+ 1.0
169
+ counts [103]
170
+ 100
171
+ 102
172
+ 104
173
+ 106
174
+ time [s]
175
+ 0
176
+ 20
177
+ 103
178
+ 105
179
+ 0.00
180
+ 0.05
181
+ 0.10
182
+ m
183
+ m
184
+ h
185
+ d
186
+ w
187
+ s
188
+ m
189
+ h
190
+ d
191
+ w
192
+ original
193
+ shuffled
194
+ inter-event times
195
+ Figure 1:
196
+ Top: temporal path entropy as a function of causal temporal scales in datasets Synthetic-
197
+ 2, WB-DE, and HC-email (transparent red) and in the temporal networks with shuffled timestamps
198
+ (transparent blue).
199
+ The height of a bar represents temporal path entropy (error bars represent
200
+ the estimation error) and the x-limits of a bar represent the interval τ = [τmin, τmax] on which the
201
+ temporal path entropy was measured. We indicate on x-axis the timescales of one minute (m),
202
+ hour (h), day (d), week (w), and year (y). We observe that the temporal path entropy differs more
203
+ between the original and the shuffled network at causal timescales. Bottom: histogram of causal
204
+ inter-event times.
205
+ In practice, the temporal path entropy can be estimated from the counts of time-respecting
206
+ paths Pk
207
+ τ (Γ) by assuming multinomial distributions with respective probabilities p(v0, . . . , vk−1) and
208
+ p(v0, . . . , vk). The counts of time-respecting paths can be computed e.g. using the methods from
209
+ [Kivel¨a et al., 2018, Petrovi´c and Scholtes, 2021]. The estimation of the entropy can be challenging
210
+ especially for small ranges of timescales, since the temporal network can get temporally disconnected,
211
+ resulting in very few paths of order k being observed. As a result we require an efficient method for
212
+ estimating the entropy that performs well even in such under-sampled regimes. The simplest estima-
213
+ tor of a multinomial distribution, called the plug-in estimator, is based on the maximum likelihood
214
+ estimation, which, however, is known to severely underestimate the entropy in the undersampled
215
+ regime and has various corrections [e.g. Miller, 1955, Grassberger, 2003]). An alternative to the
216
+ plug-in estimator is to follow a Bayesian approach which results in entropy estimators that strongly
217
+ depend on the choice of prior. To counteract this dependency, the NSB estimator [Nemenman et al.,
218
+ 2001] directly infers the entropy from the counts by averaging over different priors for the transition
219
+ probabilities, rather than inferring the transition probabilities themselves. Being a Bayesian method,
220
+ the NSB estimator can also be used to quantify the uncertainty of the estimates. Assuming that
221
+ the estimates of H(v0, . . . , vk) and H(v0, . . . , vk−1) have independent errors σk and σk−1, we can
222
+ approximate the total error of the estimate as σ = (σ2
223
+ k + σ2
224
+ k−1)1/2. As the NSB estimator requires
225
+ the size of the alphabet to be known, it is most suitable for cases where the number of nodes is fixed
226
+ and improves further if the set of edges that can occur are known a priory as this further restricts
227
+ the number of potential paths. In cases when the number of nodes in the system is unknown, the
228
+ 4
229
+
230
+ 0
231
+ 2
232
+ EU-email-A
233
+ 100
234
+ 101
235
+ 102
236
+ 103
237
+ 104
238
+ 105
239
+ 106
240
+ 0.0
241
+ 2.5
242
+ DNC-16
243
+ s
244
+ m
245
+ h
246
+ d
247
+ w
248
+ 0.0
249
+ 0.2
250
+ 0.4
251
+ 0.6
252
+ 0.8
253
+ 1.0
254
+ time [s]
255
+ 0.0
256
+ 0.2
257
+ 0.4
258
+ 0.6
259
+ 0.8
260
+ 1.0
261
+ H[ nat ]
262
+ original
263
+ shuffled
264
+ Figure 2: Temporal path entropy as a function of the timescale τ in EU-email-A and DNC-16 and in
265
+ timestamp shuffled networks. Timescale τ is represented with the x-limits of the bar, and temporal
266
+ path entropy is represented as the height of the bar. Error bars indicate the error of the temporal
267
+ path entropy estimates.
268
+ Pitman-Yor Mixture entropy estimator [Archer et al., 2014] could be used instead.
269
+ Finally, we address testing whether an interval τ is a causal timescale of a temporal network Γ.
270
+ To do so, we need to assume the null hypothesis that there are no temporal correlations between
271
+ temporal edges, but the main issue is to obtain a sample of temporal networks under this assump-
272
+ tion. To resolve this issue, we repeatedly randomize the observed temporal network Γ by randomly
273
+ permuting timestamps between its temporal edges [Holme and Saram¨aki, 2012]. These samples of
274
+ temporal networks would preserve both the edge frequencies and timestamp distribution while de-
275
+ stroying the correlations between temporal edges. We can use the samples to determine whether
276
+ temporal path entropy of the observed network has an unexpected value under the null assumption.
277
+ 4
278
+ Experiments
279
+ In the following part, we first show the behavior of temporal path entropy in synthetically generated
280
+ temporal networks with known causal timescales (the description of the generation process can
281
+ be found in the Appendix); we then present how it behaves in two real-world networks with the
282
+ information about the ground truth timescales, and two real world networks without the information
283
+ about the ground truth timescales.
284
+ In the top left panel of Fig. 1, we present the temporal path entropy H (y-axis) for various
285
+ timescales (x-axis): the left and right x limit of a bar represents τmin and τmax, and the height of
286
+ the bar represents H. The results are shown both for the synthetic network and its shuffled network.
287
+ In the bottom left panel, we show the histogram of inter-event times on causal paths. We observe
288
+ that the temporal path entropy behaves as expected and decreases in accordance with the planted
289
+ timescale at which the interactions cause one another. Moreover, this pattern disappears when the
290
+ timestamps of edges are shuffled, demonstrating that temporal path entropy captures the interplay
291
+ of temporal and topological patterns.
292
+ We consider here two empirical temporal networks where we have information about the ground
293
+ truth causal structure and two empirical temporal networks where we have no information about
294
+ 5
295
+
296
+ the ground truth causal structure. As a first data set, we consider the bipartite temporal network of
297
+ German Wikibooks co-editing patterns (WB-DE) [Wikimedia Foundation, Peixoto, 2020]. This data
298
+ contains information about edits on the Wikibooks website: for each edit, we know the editor, the
299
+ article that was edited, and the time at which the edit occurred. We preprocess this data to obtain
300
+ a temporal network of editors: if editor v edited an article prior to editor w who edited the same
301
+ article at time t, we assume that a link (v, w, t) occurred in the temporal network of editors. We
302
+ define causal inter-event times based on the articles: we extract the time intervals between successive
303
+ edits of each article. In WB-DE data, we analyze the timescales of the whole temporal network.
304
+ As a second data set, we consider public data set of Hillary Clinton’s emails (HC-email) [Kaggle,
305
+ 2022], which contains the sender, the receiver, the timestamp, and the subject of each email. In this
306
+ data set we analyze the timescales of node representing Hillary Clinton. While sender, receiver and
307
+ the timestamp constitute a temporal network, email subjects allow us to obtain causal inter-event
308
+ times: for each incoming email, we extract the time duration until an email with the same subject
309
+ was sent. We use the inter-event times between emails with the same subject and the inter-event
310
+ times of articles for evaluation; the temporal networks contain only the temporal edges and not any
311
+ additional information about the ground truth timescales. We also use two email data sets without a
312
+ ground truth timescales: EU-email-A [Paranjape et al., 2017], which contains email correspondence
313
+ between researchers of an EU institution from four deparments, and DNC-16 [Rossi and Ahmed,
314
+ 2015], which contains emails of the US Democratic National Committee. Results on other datasets
315
+ as well as details of all datasets are in the Appendix. Reproducibility package is available at [Petrovi´c
316
+ et al., 2023].
317
+ Results of the WB-DE and HC-email data are in Fig. 1 (middle and right, respectively). When
318
+ we compare the histogram of causal inter-event times with the temporal path entropy at different
319
+ timescales of the temporal network, we see that increased number of causal interactions increases the
320
+ difference in temporal path entropy between the original and the shuffled network. The temporal
321
+ path entropy converges for large timescales because the interval sizes increase, the density of causal
322
+ interactions decreases, and the noise increases. In Fig. 2, although we do not have the ground truth,
323
+ we see that the largest difference between the original and the shuffled datasets are at timescales
324
+ between a minute a day, which is what we would expect from email correspondence.
325
+ We identify four limitations of our approach. First, our base assumption is that the interactions,
326
+ represented by edges, cause one another, and our measure can not separate that case that from
327
+ the case when edges are generated by some common factor. Second, being based on directed paths
328
+ the current method is restricted in the types of causal interactions it considers namely interactions
329
+ where a incoming link into a vertex effects the subsequent links emanating from the vertex. The
330
+ method could potentially be generalized to other types of interactions by considering other patterns
331
+ to alleviate this shortcoming. Third, our method cannot detect timescales at which the incoming
332
+ edges to a node change the overall activity of the node without changing the relative frequencies of
333
+ the outgoing edges. Detecting timescales of such causal influences is thus an open problem. Fourth,
334
+ real data can contain time-varying timescales, e.g. during day or night, which would probably require
335
+ an application of time warping techniques.
336
+ 5
337
+ Conclusion
338
+ To summarize, the analysis of temporal networks heavily depends on the analysis of time-respecting
339
+ paths [Holme and Saram¨aki, 2012, Holme, 2015, Pan and Saram¨aki, 2011, Masuda et al., 2013,
340
+ Scholtes et al., 2016, Kivel¨a et al., 2018]. However, in order to model and analyze the time-respecting
341
+ paths, we first need to identify the correct timescale.
342
+ In this work we address this problem by
343
+ 6
344
+
345
+ introducing an information theoretic measure, the temporal path entropy, that is able to can identify
346
+ timescales at which the influences are highly correlated. Using real world data we demonstrated that
347
+ the measure can be applied to temporal networks as a whole as well as to a single node. We showed
348
+ that the temporal path entropy can capture the causal timescales in both synthetic and empirical
349
+ temporal networks. We further support our findings by observing that the differences in the temporal
350
+ path entropy between the original and shuffled networks coincide with increases in the number of
351
+ causal paths. The temporal path entropy allows system-relevant timescales to be inferred from the
352
+ temporal networks themselves which is crucial for the analysis of temporal networks where inherent
353
+ timescales are unavailable and hard to measure.
354
+ Acknowledgments
355
+ The authors would like to thank Christopher Bl¨ocker, Chester Tan, and Franziska Heeg for valu-
356
+ able comments on the manuscript. LP and IS acknowledge support by the Swiss National Science
357
+ Foundation, grant 176938.
358
+ References
359
+ E. Archer, I. M. Park, and J. W. Pillow. Bayesian entropy estimation for countable discrete distri-
360
+ butions. The Journal of Machine Learning Research, 15(1):2833–2868, 2014.
361
+ A. Badie-Modiri, M. Karsai, and M. Kivel¨a. Efficient limited-time reachability estimation in temporal
362
+ networks. Physical Review E, 101(5):052303, 2020.
363
+ A. Badie-Modiri, A. K. Rizi, M. Karsai, and M. Kivel¨a. Directed percolation in temporal networks.
364
+ Physical Review Research, 4(2):L022047, 2022.
365
+ F. Battiston, G. Cencetti, I. Iacopini, V. Latora, M. Lucas, A. Patania, J.-G. Young, and G. Petri.
366
+ Networks beyond pairwise interactions: structure and dynamics. Physics Reports, 2020.
367
+ B. Blonder and A. Dornhaus. Time-ordered networks reveal limitations to information flow in ant
368
+ colonies. PloS one, 6(5):e20298, 2011.
369
+ R. S. Caceres and T. Berger-Wolf. Temporal scale of dynamic networks. In Temporal networks,
370
+ pages 65–94. Springer, 2013.
371
+ A. Casteigts, A.-S. Himmel, H. Molter, and P. Zschoche. Finding temporal paths under waiting
372
+ time constraints. Algorithmica, 83(9):2754–2802, 2021.
373
+ A. Ceria, S. Havlin, A. Hanjalic, and H. Wang. Topological–temporal properties of evolving networks.
374
+ Journal of Complex Networks, 10(5):cnac041, 2022.
375
+ R. K. Darst, C. Granell, A. Arenas, S. G´omez, J. Saram¨aki, and S. Fortunato. Detection of timescales
376
+ in evolving complex systems. Scientific reports, 6(1):1–8, 2016.
377
+ L. Gauvin, A. Panisson, and C. Cattuto. Detecting the community structure and activity patterns
378
+ of temporal networks: a non-negative tensor factorization approach. PloS one, 9(1):e86028, 2014.
379
+ V. Gemmetto, A. Barrat, and C. Cattuto. Mitigation of infectious disease at school: targeted class
380
+ closure vs school closure. BMC infectious diseases, 14(1):1–10, 2014.
381
+ 7
382
+
383
+ M. G´enois, C. L. Vestergaard, J. Fournet, A. Panisson, I. Bonmarin, and A. Barrat. Data on face-
384
+ to-face contacts in an office building suggest a low-cost vaccination strategy based on community
385
+ linkers. Network Science, 3(3):326–347, 2015.
386
+ A. Ghasemian, P. Zhang, A. Clauset, C. Moore, and L. Peel. Detectability thresholds and optimal
387
+ algorithms for community structure in dynamic networks. Physical Review X, 6(3):031005, 2016.
388
+ K.-I. Goh and A.-L. Barab´asi.
389
+ Burstiness and memory in complex systems.
390
+ EPL (Europhysics
391
+ Letters), 81(4):48002, 2008.
392
+ P. Grassberger. Entropy estimates from insufficient samplings. arXiv preprint physics/0307138,
393
+ 2003.
394
+ T. Gross and H. Sayama. Adaptive networks. In Adaptive networks, pages 1–8. Springer, 2009.
395
+ P. Holme. Modern temporal network theory: a colloquium. The European Physical Journal B, 88
396
+ (9):234, 2015.
397
+ P. Holme and J. Saram¨aki. Temporal networks. Physics reports, 519(3):97–125, 2012.
398
+ L. Isella, J. Stehl´e, A. Barrat, C. Cattuto, J.-F. Pinton, and W. Van den Broeck. What’s in a crowd?
399
+ analysis of face-to-face behavioral networks. Journal of theoretical biology, 271(1):166–180, 2011.
400
+ Kaggle.
401
+ Hillary clinton emails., 2022.
402
+ URL https://www.kaggle.com/datasets/kaggle/
403
+ hillary-clinton-emails.
404
+ M. Kivel¨a, A. Arenas, M. Barthelemy, J. P. Gleeson, Y. Moreno, and M. A. Porter. Multilayer
405
+ networks. Journal of complex networks, 2(3):203–271, 2014.
406
+ M. Kivel¨a, J. Cambe, J. Saram¨aki, and M. Karsai.
407
+ Mapping temporal-network percolation to
408
+ weighted, static event graphs. Scientific reports, 8(1):1–9, 2018.
409
+ R. Lambiotte, V. Salnikov, and M. Rosvall. Effect of memory on the dynamics of random walks on
410
+ networks. Journal of Complex Networks, 3(2):177–188, 2015.
411
+ R. Lambiotte, M. Rosvall, and I. Scholtes. From networks to optimal higher-order models of complex
412
+ systems. Nature physics, page 1, 2019.
413
+ H. H. Lentz, T. Selhorst, and I. M. Sokolov. Unfolding accessibility provides a macroscopic approach
414
+ to temporal networks. Physical review letters, 110(11):118701, 2013. doi: 10.1103/PhysRevLett.
415
+ 110.118701. URL http://link.aps.org/doi/10.1103/PhysRevLett.110.118701.
416
+ R. Mastrandrea, J. Fournet, and A. Barrat.
417
+ Contact patterns in a high school: a comparison
418
+ between data collected using wearable sensors, contact diaries and friendship surveys. PloS one,
419
+ 10(9):e0136497, 2015.
420
+ N. Masuda, K. Klemm, and V. M. Egu´ıluz. Temporal networks: slowing down diffusion by long
421
+ lasting interactions. Physical Review Letters, 111(18):188701, 2013.
422
+ G. Miller. Note on the bias of information estimates. information theory in psychology: Problems
423
+ and methods. Quastler H, pages 95–100, 1955.
424
+ I. Nemenman, F. Shafee, and W. Bialek.
425
+ Entropy and inference, revisited. Advances in neural
426
+ information processing systems, 14, 2001.
427
+ 8
428
+
429
+ R. K. Pan and J. Saram¨aki. Path lengths, correlations, and centrality in temporal networks. Physical
430
+ Review E, 84(1):016105, 2011.
431
+ A. Paranjape, A. R. Benson, and J. Leskovec. Motifs in temporal networks. In Proceedings of the
432
+ tenth ACM international conference on web search and data mining, pages 601–610, 2017.
433
+ T. P. Peixoto. Inferring the mesoscale structure of layered, edge-valued, and time-varying networks.
434
+ Physical Review E, 92(4):042807, 2015.
435
+ T. P. Peixoto. The netzschleuder network catalogue and repository, 2020. URL https://networks.
436
+ skewed.de/.
437
+ T. P. Peixoto and L. Gauvin. Change points, memory and epidemic spreading in temporal networks.
438
+ Scientific reports, 8(1):1–10, 2018.
439
+ T. P. Peixoto and M. Rosvall. Modelling sequences and temporal networks with dynamic community
440
+ structures. Nature communications, 8(1):1–12, 2017.
441
+ V. Perri and I. Scholtes. Hotvis: Higher-order time-aware visualisation of dynamic graphs. In Graph
442
+ Drawing and Network Visualization - 28th International Symposium, -, 2020.
443
+ G. Petri and A. Barrat. Simplicial activity driven model. Physical review letters, 121(22):228301,
444
+ 2018.
445
+ L. V. Petrovi´c and I. Scholtes. Paco: Fast counting of causal paths in temporal network data. In
446
+ Companion Proceedings of the Web Conference 2021, pages 521–526, 2021.
447
+ L. V. Petrovi´c, A. Wegner, and I. Scholtes. Reproducibility Package for the Paper ”Higher- Order
448
+ Patterns Reveal Causal Timescales of Complex Systems”, Jan. 2023. URL https://doi.org/10.
449
+ 5281/zenodo.7575182.
450
+ R. Pfitzner, I. Scholtes, A. Garas, C. J. Tessone, and F. Schweitzer. Betweenness preference: Quan-
451
+ tifying correlations in the topological dynamics of temporal networks. Physical review letters, 110
452
+ (19):198701, 2013. doi: 10.1103/PhysRevLett.110.198701. URL http://link.aps.org/doi/10.
453
+ 1103/PhysRevLett.110.198701.
454
+ R. A. Rossi and N. K. Ahmed. The network data repository with interactive graph analytics and
455
+ visualization. In AAAI, 2015. URL http://networkrepository.com.
456
+ M. Rosvall, A. V. Esquivel, A. Lancichinetti, J. D. West, and R. Lambiotte. Memory in network
457
+ flows and its effects on spreading dynamics and community detection. Nature communications, 5:
458
+ 4630, 2014.
459
+ I. Scholtes, N. Wider, R. Pfitzner, A. Garas, C. J. Tessone, and F. Schweitzer. Causality-driven slow-
460
+ down and speed-up of diffusion in non-markovian temporal networks. Nature communications, 5:
461
+ 5024, 2014. doi: 10.1038/ncomms6024. URL https://doi.org/10.1038/ncomms6024.
462
+ I. Scholtes, N. Wider, and A. Garas. Higher-order aggregate networks in the analysis of temporal
463
+ networks: path structures and centralities. The European Physical Journal B, 89(3):61, 2016.
464
+ J. Stehl´e, N. Voirin, A. Barrat, C. Cattuto, L. Isella, J.-F. Pinton, M. Quaggiotto, W. Van den
465
+ Broeck, C. R´egis, B. Lina, et al. High-resolution measurements of face-to-face contact patterns in
466
+ a primary school. PloS one, 6(8):e23176, 2011.
467
+ 9
468
+
469
+ D. Taylor, S. Shai, N. Stanley, and P. J. Mucha. Enhanced detectability of community structure in
470
+ multilayer networks through layer aggregation. Physical review letters, 116(22):228301, 2016.
471
+ P. Vanhems, A. Barrat, C. Cattuto, J.-F. Pinton, N. Khanafer, C. R´egis, B.-a. Kim, B. Comte, and
472
+ N. Voirin. Estimating potential infection transmission routes in hospital wards using wearable
473
+ proximity sensors. PloS one, 8(9):e73970, 2013.
474
+ Wikimedia Foundation. Wikimedia downloads. URL http://dumps.wikimedia.org/.
475
+ O. E. Williams, L. Lacasa, A. P. Mill´an, and V. Latora. The shape of memory in temporal networks.
476
+ Nature communications, 13(1):1–8, 2022.
477
+ M. S. Zanetti, I. Scholtes, C. J. Tessone, and F. Schweitzer. Categorizing bugs with social networks:
478
+ a case study on four open source software communities. In 2013 35th International Conference
479
+ on Software Engineering (ICSE), pages 1032–1041. IEEE, 2013.
480
+ 6
481
+ Datasets
482
+ In this work we considered synthetic and empirical temporal networks.
483
+ To generate synthetic temporal networks Synthetic-1, Synthetic-2 and Synthetic-3 with a ground
484
+ truth timescale ¯τ = [¯τmin, ¯τmax], we start from a static Erd˝os-R´enyi random graph with 50 nodes and
485
+ 500 directed edges. We sample a random subset Pcausal of nu.p. = 500 unique paths of length k = 2
486
+ in the static network which correspond to causal influences in the system. We sample with repetition
487
+ np = 5000 paths from Pcausal to generate dataset Synthetic-1, np = 10000 paths to generate dataset
488
+ Synthetic-2 and np = 20000 paths to generate dataset Synthetic-3. To add each path (v0, v1, v2)
489
+ to the temporal network, we sample a random starting time t uniformly from [0, Ttotal − ¯τmax] and
490
+ create a temporal edge (v0, v1, t); we then sample temporal distance δ between edges on the path
491
+ (inter-event time) uniformly from ¯τ and create the temporal edge (v1, v2, t + δ). We choose ¯τ with
492
+ ¯τmin = 100 and ¯τmax = 200. To add some noise to the system, we uniformly sample 20000 edges
493
+ from the static graph, and sample their timestamps uniformly from [0, Ttotal]. The temporal network
494
+ Synthetic-4 contains two time-scales relevant for the dynamics. To do so, we generated two different
495
+ temporal networks based on two random graphs of 50 nodes (with the same node names) and 500
496
+ edges and based on the different timescales τ 1 = [50, 100] and τ 2 = [150, 200]. We used the same
497
+ procedure as above with parameters nu.p. = 500; np = 5000; Ttotal = 105; nr.e. = 10000. We merged
498
+ the two temporal networks into one; the details of the resulting network are in Table 1. The dataset
499
+ Synthetic-5 contains paths of length three. Again, there are 50 nodes and 500 edges in the static
500
+ Erd˝os R´enyi graph.
501
+ We sample nu.p. = 20 unique paths, we sample np = 20000 of them, and
502
+ spread them across Ttotal = 105 using the same procedure and timescale τ = [100, 200]. We add
503
+ nr.e = 10000 random edges to the network as noise.
504
+ We also use empirical dataset where can get access to the ground truth causal path structure. We
505
+ consider the bipartite temporal network of Wikibooks co-edits in Arabic (WB-AR), French (WB-
506
+ FR) and German (WB-DE) [Wikimedia Foundation, Peixoto, 2020]. This data contains information
507
+ about edits on the Wikibooks website: for each edit, we know the editor, the article that was edited,
508
+ and the time at which the edit occurred. We preprocess this data to obtain a temporal network
509
+ of editors: if editor v edited an article prior to editor w who edited the same article at time t, we
510
+ assume that a link (v, w, t) occurred in the temporal network of editors. We define causal inter-event
511
+ times based on the articles: we extract the time intervals between successive edits of each article.
512
+ In these data, we analyze the timescales of the whole temporal network. Another dataset where we
513
+ 10
514
+
515
+ dataset
516
+ |V |
517
+ |E|
518
+ |E|
519
+ Ttotal [s]
520
+ Ants-1-1
521
+ 89
522
+ 947
523
+ 1911
524
+ 1.44e+03
525
+ Ants-1-2
526
+ 72
527
+ 862
528
+ 1820
529
+ 1.75e+03
530
+ Ants-2-1
531
+ 71
532
+ 636
533
+ 975
534
+ 1.44e+03
535
+ Ants-2-2
536
+ 69
537
+ 769
538
+ 1917
539
+ 1.8e+03
540
+ Ants-3-1
541
+ 11
542
+ 37
543
+ 78
544
+ 1.13e+03
545
+ Ants-3-2
546
+ 6
547
+ 21
548
+ 104
549
+ 1.42e+03
550
+ DNC-16
551
+ 1891
552
+ 5598
553
+ 39264
554
+ 8.49e+07
555
+ EU-email-1
556
+ 309
557
+ 3031
558
+ 61046
559
+ 6.94e+07
560
+ EU-email-2
561
+ 162
562
+ 1772
563
+ 46772
564
+ 6.94e+07
565
+ EU-email-3
566
+ 89
567
+ 1506
568
+ 12216
569
+ 6.93e+07
570
+ EU-email-4
571
+ 142
572
+ 1375
573
+ 48141
574
+ 6.94e+07
575
+ EU-email-A
576
+ 986
577
+ 24929
578
+ 332334
579
+ 6.95e+07
580
+ Gallery
581
+ 10972
582
+ 89034
583
+ 831824
584
+ 6.95e+06
585
+ HC-email
586
+ 326
587
+ 385
588
+ 8313
589
+ 1.19e+08
590
+ Hospital
591
+ 75
592
+ 2278
593
+ 64848
594
+ 3.48e+05
595
+ Hypertext
596
+ 113
597
+ 4392
598
+ 41636
599
+ 2.12e+05
600
+ OSS
601
+ 5789
602
+ 6888
603
+ 12583
604
+ 3.54e+08
605
+ Primary
606
+ 242
607
+ 16634
608
+ 251546
609
+ 1.17e+05
610
+ School-13
611
+ 327
612
+ 11636
613
+ 377016
614
+ 3.64e+05
615
+ Synthetic-1
616
+ 50
617
+ 500
618
+ 30000
619
+ 1e+05
620
+ Synthetic-2
621
+ 50
622
+ 500
623
+ 40000
624
+ 1e+05
625
+ Synthetic-3
626
+ 50
627
+ 500
628
+ 60000
629
+ 1e+05
630
+ Synthetic-4
631
+ 50
632
+ 898
633
+ 40000
634
+ 1e+05
635
+ Synthetic-5
636
+ 50
637
+ 500
638
+ 50000
639
+ 1e+05
640
+ WB-AR
641
+ 1124
642
+ 3334
643
+ 27166
644
+ 3.89e+08
645
+ WB-DE
646
+ 10999
647
+ 54700
648
+ 464089
649
+ 4.87e+08
650
+ WB-FR
651
+ 9735
652
+ 53606
653
+ 362094
654
+ 4.88e+08
655
+ Work-13
656
+ 92
657
+ 1510
658
+ 19654
659
+ 9.88e+05
660
+ Table 1: The sizes of the sets of nodes V , unique edges E, and temporal edges E of temporal
661
+ networks that we analyzed in the experiments. Datasets synth-2, HC email and WB DE are in the
662
+ main paper. The other datasets are shown in the Appendix.
663
+ can get access to the ground truth causal structure is the public data set of Hillary Clinton’s emails
664
+ (HC-email) [Kaggle, 2022], which contains the sender, the receiver, the timestamp, and the subject
665
+ of each email. In this data set we analyze the timescales of node representing Hillary Clinton. While
666
+ sender, receiver and the timestamp form a temporal network, email subjects allow us to obtain
667
+ causal inter-event times: for each incoming email, we extract the time duration until an email with
668
+ the same subject was sent. We use the inter-event times between emails with the same subject and
669
+ the inter-event times of articles for evaluation; the temporal networks contain only the temporal
670
+ edges and not any additional information about the ground truth timescales. The details of each
671
+ data-set are in Table 1.
672
+ Finally, we also use empirical temporal networks where we do not know the ground truth causal
673
+ path structure. Datasets Ants-1-1, Ants-1-2, Ants-2-1, Ants-2-2, Ants-3-1, and Ants-3-3 [Blonder
674
+ and Dornhaus, 2011] contain antenna contacts in ant colonies. Dataset DNC-16 [Rossi and Ahmed,
675
+ 2015] contains emails of the US Democratic National Committee leaked in 2016.
676
+ Datasets EU-
677
+ 11
678
+
679
+ email-1, EU-email-2, EU-email-3, EU-email-4, and EU-email-A [Paranjape et al., 2017] contain
680
+ email correspondence between researchers of an EU institution from first, second, third, fourth and
681
+ all departments, respectively. Datasets Gallery [Isella et al., 2011], Hospital [Vanhems et al., 2013],
682
+ Hypertext [Isella et al., 2011], Primary [Gemmetto et al., 2014, Stehl´e et al., 2011], Work-13 [G´enois
683
+ et al., 2015] and School-13 [Mastrandrea et al., 2015] contain human face-to-face interactions in
684
+ different settings measured by the SocioPatterns collaborations. Dataset OSS [Zanetti et al., 2013]
685
+ contains ASSIGN relationships between members of the Open Source Software community Apache.
686
+ 7
687
+ Results: Synthetic Data
688
+ We present results for datasets Synthetic-1, Synthetic-3, Synthetic-4, Synthetic-5.
689
+ 1.5
690
+ 2.0
691
+ H[ nat ]
692
+ Synthetic-1
693
+ 50
694
+ 100
695
+ 150
696
+ 200
697
+ 250
698
+ time [s]
699
+ 0.0
700
+ 0.2
701
+ 0.4
702
+ counts [103]
703
+ original
704
+ shuffled
705
+ inter-event times
706
+ Figure 3: Top: temporal path entropy as a function of the timescale τ in temporal network Synthetic-
707
+ 1 and in Synthetic-1 with shuffled timestamps. Timescale τ is represented with the x-limits of the
708
+ bar, and temporal path entropy is represented as the height of the bar. Error bars indicate the error
709
+ of the temporal path entropy estimates. Bottom: histogram of inter-event times of synthetic causal
710
+ interactions.
711
+ 8
712
+ Results: Empirical Data with Ground Truth
713
+ In this section we show results on other Wikibooks datasets that we used to test the method. In
714
+ Fig. 7, we test temporal path entropy on the WB-AR dataset, and in Fig. 8, we test it on the WB-FR
715
+ dataset. Similar to the WB-DE in the main paper, the bottom panel shows the yellow histogram of
716
+ inter-event times of edits per article for all articles.
717
+ 9
718
+ Empirical data without the ground truth
719
+ In this section, we show multiple datasets in which we do not have access to the ground truth
720
+ temporal scales. Although the lack of ground truth in these datasets makes objective evaluation
721
+ 12
722
+
723
+ 1.5
724
+ 2.0
725
+ H[ nat ]
726
+ Synthetic-3
727
+ 50
728
+ 100
729
+ 150
730
+ 200
731
+ 250
732
+ time [s]
733
+ 0
734
+ 1
735
+ 2
736
+ counts [103]
737
+ original
738
+ shuffled
739
+ inter-event times
740
+ Figure 4: Equivalent of Fig. 3, for Synthetic-3.
741
+ 2
742
+ 3
743
+ H[ nat ]
744
+ Synthetic-4
745
+ 50
746
+ 100
747
+ 150
748
+ 200
749
+ 250
750
+ time [s]
751
+ 0.0
752
+ 0.5
753
+ 1.0
754
+ counts [103]
755
+ original
756
+ shuffled
757
+ inter-event times
758
+ Figure 5: Equivalent of Fig. 3, for Synthetic-4.
759
+ of the method difficult, the results across datasets are consistent and in accordance with what one
760
+ would expect: e.g. in the email datasets, temporal path entropy is different between the original
761
+ and the shuffled network for timescales between one minute and a few days, which corresponds to
762
+ what we would expect to be the interval in which emails are responded to.
763
+ 13
764
+
765
+ 1.0
766
+ 1.2
767
+ H[ nat ]
768
+ k = 2
769
+ Synthetic-5
770
+ 0.75
771
+ 1.00
772
+ 1.25
773
+ H[ nat ]
774
+ k = 3
775
+ 50
776
+ 75
777
+ 100
778
+ 125
779
+ 150
780
+ 175
781
+ 200
782
+ 225
783
+ 250
784
+ time [s]
785
+ 0.0
786
+ 2.5
787
+ counts [103]
788
+ original
789
+ shuffled
790
+ inter-event times
791
+ Figure 6: Temporal path entropy as a function of the timescale τ in temporal network Synthetic-5
792
+ and in Synthetic-5 with shuffled timestamps for orders k = 2 (top) and k = 3 (middle). Timescale τ
793
+ is represented with the x-limits of the bar, and temporal path entropy is represented as the height of
794
+ the bar. Error bars indicate the error of the temporal path entropy estimates. Bottom: histogram
795
+ of inter-event times of synthetic causal interactions.
796
+ 0.0
797
+ 0.5
798
+ 1.0
799
+ H[ nat ]
800
+ WB-AR
801
+ 100
802
+ 102
803
+ 104
804
+ 106
805
+ time [s]
806
+ 0
807
+ 2
808
+ counts [103]
809
+ s
810
+ m
811
+ h
812
+ d
813
+ w
814
+ original
815
+ shuffled
816
+ inter-event times
817
+ Figure 7:
818
+ Top: temporal path entropy as a function of the timescale τ in WB-AR temporal network
819
+ and of WB-AR temporal network with shuffled timestamps. Timescale τ is represented with the
820
+ x-limits of the bar, and temporal path entropy is represented as the height of the bar. Error bars
821
+ indicate the error of the temporal path entropy estimates. Bottom: histogram of inter-event times
822
+ for all articles of edits of the same article.
823
+ 14
824
+
825
+ 0.0
826
+ 0.5
827
+ 1.0
828
+ H[ nat ]
829
+ WB-FR
830
+ 100
831
+ 102
832
+ 104
833
+ 106
834
+ time [s]
835
+ 0
836
+ 20
837
+ counts [103]
838
+ s
839
+ m
840
+ h
841
+ d
842
+ w
843
+ original
844
+ shuffled
845
+ inter-event times
846
+ Figure 8:
847
+ Equivalent of Fig. 7 for WB-FR.
848
+ 15
849
+
850
+ 100
851
+ 101
852
+ 102
853
+ 103
854
+ time [s]
855
+ 0
856
+ 1
857
+ 2
858
+ 3
859
+ 4
860
+ 5
861
+ H[ nat ]
862
+ Ants-1-1
863
+ original
864
+ shuffled
865
+ s
866
+ m
867
+ 100
868
+ 101
869
+ 102
870
+ 103
871
+ time [s]
872
+ 0
873
+ 1
874
+ 2
875
+ 3
876
+ 4
877
+ H[ nat ]
878
+ Ants-1-2
879
+ original
880
+ shuffled
881
+ s
882
+ m
883
+ 100
884
+ 101
885
+ 102
886
+ 103
887
+ time [s]
888
+ 0
889
+ 1
890
+ 2
891
+ 3
892
+ 4
893
+ H[ nat ]
894
+ Ants-2-1
895
+ original
896
+ shuffled
897
+ s
898
+ m
899
+ 100
900
+ 101
901
+ 102
902
+ 103
903
+ time [s]
904
+ 0
905
+ 1
906
+ 2
907
+ 3
908
+ H[ nat ]
909
+ Ants-2-2
910
+ original
911
+ shuffled
912
+ s
913
+ m
914
+ 100
915
+ 101
916
+ 102
917
+ 103
918
+ time [s]
919
+ 0.0
920
+ 0.5
921
+ 1.0
922
+ 1.5
923
+ 2.0
924
+ 2.5
925
+ 3.0
926
+ H[ nat ]
927
+ Ants-3-1
928
+ original
929
+ shuffled
930
+ s
931
+ m
932
+ 100
933
+ 101
934
+ 102
935
+ 103
936
+ time [s]
937
+ 0.0
938
+ 0.5
939
+ 1.0
940
+ 1.5
941
+ 2.0
942
+ H[ nat ]
943
+ Ants-3-2
944
+ original
945
+ shuffled
946
+ s
947
+ m
948
+ Figure 9: Temporal path entropy as a function of the timescale τ in temporal networks of antenna
949
+ contacts in ant collonies. For each temporal network, we show the temporal path entropy of the
950
+ original and of a shuffled network. Timescale τ is represented with the x-limits of the bar, and
951
+ temporal path entropy is represented as the height of the bar. Error bars indicate the error of the
952
+ temporal path entropy estimates.
953
+ 16
954
+
955
+ 100
956
+ 101
957
+ 102
958
+ 103
959
+ 104
960
+ 105
961
+ 106
962
+ time [s]
963
+ 0
964
+ 1
965
+ 2
966
+ 3
967
+ 4
968
+ H[ nat ]
969
+ DNC-16
970
+ original
971
+ shuffled
972
+ s
973
+ m
974
+ h
975
+ d
976
+ w
977
+ 100
978
+ 101
979
+ 102
980
+ 103
981
+ 104
982
+ 105
983
+ 106
984
+ time [s]
985
+ 0.0
986
+ 0.5
987
+ 1.0
988
+ 1.5
989
+ 2.0
990
+ 2.5
991
+ 3.0
992
+ H[ nat ]
993
+ EU-email-A
994
+ original
995
+ shuffled
996
+ s
997
+ m
998
+ h
999
+ d
1000
+ w
1001
+ 100
1002
+ 101
1003
+ 102
1004
+ 103
1005
+ 104
1006
+ 105
1007
+ 106
1008
+ time [s]
1009
+ 0
1010
+ 1
1011
+ 2
1012
+ 3
1013
+ H[ nat ]
1014
+ EU-email-1
1015
+ original
1016
+ shuffled
1017
+ s
1018
+ m
1019
+ h
1020
+ d
1021
+ w
1022
+ 100
1023
+ 101
1024
+ 102
1025
+ 103
1026
+ 104
1027
+ 105
1028
+ 106
1029
+ time [s]
1030
+ 0
1031
+ 1
1032
+ 2
1033
+ 3
1034
+ 4
1035
+ 5
1036
+ H[ nat ]
1037
+ EU-email-2
1038
+ original
1039
+ shuffled
1040
+ s
1041
+ m
1042
+ h
1043
+ d
1044
+ w
1045
+ 100
1046
+ 101
1047
+ 102
1048
+ 103
1049
+ 104
1050
+ 105
1051
+ 106
1052
+ time [s]
1053
+ 0
1054
+ 1
1055
+ 2
1056
+ 3
1057
+ 4
1058
+ 5
1059
+ H[ nat ]
1060
+ EU-email-3
1061
+ original
1062
+ shuffled
1063
+ s
1064
+ m
1065
+ h
1066
+ d
1067
+ w
1068
+ 100
1069
+ 101
1070
+ 102
1071
+ 103
1072
+ 104
1073
+ 105
1074
+ 106
1075
+ time [s]
1076
+ 0
1077
+ 1
1078
+ 2
1079
+ 3
1080
+ 4
1081
+ 5
1082
+ 6
1083
+ H[ nat ]
1084
+ EU-email-4
1085
+ original
1086
+ shuffled
1087
+ s
1088
+ m
1089
+ h
1090
+ d
1091
+ w
1092
+ Figure 10: Temporal path entropy as a function of the timescale τ in temporal networks of email
1093
+ correspondence. For each temporal network, we show the temporal path entropy of the original and
1094
+ of a shuffled network. Timescale τ is represented with the x-limits of the bar, and temporal path
1095
+ entropy is represented as the height of the bar. Error bars indicate the error of the temporal path
1096
+ entropy estimates.
1097
+ 17
1098
+
1099
+ 102
1100
+ 103
1101
+ 104
1102
+ time [s]
1103
+ 0
1104
+ 2
1105
+ 4
1106
+ 6
1107
+ H[ nat ]
1108
+ Gallery
1109
+ original
1110
+ shuffled
1111
+ m
1112
+ h
1113
+ 102
1114
+ 103
1115
+ 104
1116
+ 105
1117
+ time [s]
1118
+ 0
1119
+ 1
1120
+ 2
1121
+ 3
1122
+ 4
1123
+ 5
1124
+ 6
1125
+ H[ nat ]
1126
+ School-13
1127
+ original
1128
+ shuffled
1129
+ m
1130
+ h
1131
+ d
1132
+ 102
1133
+ 103
1134
+ 104
1135
+ 105
1136
+ time [s]
1137
+ 0
1138
+ 1
1139
+ 2
1140
+ 3
1141
+ 4
1142
+ 5
1143
+ 6
1144
+ H[ nat ]
1145
+ Hospital
1146
+ original
1147
+ shuffled
1148
+ m
1149
+ h
1150
+ d
1151
+ 102
1152
+ 103
1153
+ 104
1154
+ 105
1155
+ time [s]
1156
+ 0
1157
+ 1
1158
+ 2
1159
+ 3
1160
+ 4
1161
+ 5
1162
+ 6
1163
+ H[ nat ]
1164
+ Hypertext
1165
+ original
1166
+ shuffled
1167
+ m
1168
+ h
1169
+ d
1170
+ 102
1171
+ 103
1172
+ 104
1173
+ 105
1174
+ time [s]
1175
+ 0
1176
+ 2
1177
+ 4
1178
+ 6
1179
+ H[ nat ]
1180
+ Primary
1181
+ original
1182
+ shuffled
1183
+ m
1184
+ h
1185
+ d
1186
+ 102
1187
+ 103
1188
+ 104
1189
+ 105
1190
+ time [s]
1191
+ 0
1192
+ 1
1193
+ 2
1194
+ 3
1195
+ 4
1196
+ 5
1197
+ H[ nat ]
1198
+ Work-13
1199
+ original
1200
+ shuffled
1201
+ m
1202
+ h
1203
+ d
1204
+ w
1205
+ Figure 11: Temporal path entropy as a function of the timescale τ in temporal networks of human
1206
+ face-to-face interactions measured by the SocioPatterns collaboration. For each temporal network,
1207
+ we show the temporal path entropy of the original and of a shuffled network.
1208
+ Timescale τ is
1209
+ represented with the x-limits of the bar, and temporal path entropy is represented as the height
1210
+ of the bar. Error bars indicate the error of the temporal path entropy estimates.
1211
+ 18
1212
+
1213
+ 100
1214
+ 101
1215
+ 102
1216
+ 103
1217
+ 104
1218
+ 105
1219
+ 106
1220
+ time [s]
1221
+ 0
1222
+ 1
1223
+ 2
1224
+ 3
1225
+ 4
1226
+ H[ nat ]
1227
+ OSS
1228
+ original
1229
+ shuffled
1230
+ s
1231
+ m
1232
+ h
1233
+ d
1234
+ w
1235
+ Figure 12: Temporal path entropy as a function of the timescale τ in temporal networks ASSIGN
1236
+ relationships between members of the Open Source Software community Apache.
1237
+ We show the
1238
+ temporal path entropy of the original and of a shuffled network. Timescale τ is represented with the
1239
+ x-limits of the bar, and temporal path entropy is represented as the height of the bar. Error bars
1240
+ indicate the error of the temporal path entropy estimates.
1241
+ 19
1242
+
1243
+ 10
1244
+ Conditional entropy: The chain rule
1245
+ For discrete random variables X and Y , the definition of the entropy (in nats) is
1246
+ H(X) = −
1247
+
1248
+ x
1249
+ p(X = x) ln p(X = x)
1250
+ and the definition of conditional entropy (in nats) H(Y |X) is:
1251
+ H(Y |X) = −
1252
+
1253
+ x,y
1254
+ p(X = x, Y = y) ln p(X = x, Y = y)
1255
+ p(X = x)
1256
+ In the following, we use the above definitions to derive the chain rule of conditional entropy:
1257
+ H(Y |X) = −
1258
+
1259
+ x,y
1260
+ p(X = x, Y = y) (ln p(X = x, Y = y) − ln p(X = x)) =
1261
+ = −
1262
+
1263
+ x,y
1264
+ p(X = x, Y = y) ln p(X = x, Y = y) −
1265
+
1266
+
1267
+
1268
+ x,y
1269
+ p(X = x, Y = y) ln(p(X = x)))
1270
+
1271
+ =
1272
+ = H(X, Y ) −
1273
+
1274
+
1275
+
1276
+ x,y
1277
+ p(Y = y|X = x)p(X = x) ln(p(X = x)))
1278
+
1279
+ =
1280
+ = H(X, Y ) −
1281
+
1282
+ �−
1283
+
1284
+ x
1285
+ p(X = x) ln(p(X = x)))
1286
+ 
1287
+ :1
1288
+ ��
1289
+ y
1290
+ p(Y = y|X = x)
1291
+
1292
+
1293
+ � =
1294
+ = H(X, Y ) − H(X).
1295
+ (5)
1296
+ 20
1297
+
H9FJT4oBgHgl3EQfuy1Y/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
HtE1T4oBgHgl3EQfrgVp/content/tmp_files/2301.03355v1.pdf.txt ADDED
@@ -0,0 +1,1489 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Charge transfer mediated giant photo-amplification in air-stable α-CsPbI3
2
+ nanocrystals decorated 2D-WS2 photo-FET with asymmetric contacts
3
+ Shreyasi Das1, Arup Ghorai1,2, Sourabh Pal3, Somnath Mahato1, Soumen Das4, Samit K. Ray5 *
4
+ 1School of Nano Science and Technology, IIT Kharagpur, Kharagpur, West Bengal, India, 721302
5
+ 2Department of Materials Science and Engineering, Pohang University of Science and Technology,
6
+ Pohang 790-784, Korea
7
+ 3Advanced Technology Development Centre, IIT Kharagpur, Kharagpur, West Bengal, India, 721302
8
+ 4School of Medical Science and Technology, IIT Kharagpur, Kharagpur, West Bengal, India, 721302
9
+ 5Department of Physics, IIT Kharagpur, Kharagpur, West Bengal, India, 721302
10
+ Email : physkr@phy.iitkgp.ac.in
11
+
12
+ Abstract
13
+ Hybrid heterostructure based phototransistors are attractive owing to their high gain induced
14
+ by photogating effect. However, the absence of an in-plane built-in electric field in the single
15
+ channel layer transistor results in a relatively higher dark current and require a large operating
16
+ gate voltage of the device. Here, we report novel air-stable cesium lead iodide/tungsten di-
17
+ sulfide (CsPbI3/WS2) mixed dimensional heterostructure based photo-field-effect-transistors
18
+ (photo-FETs) with asymmetric metal electrodes (Cr/WS2/Au), exhibiting extremely low dark
19
+ current (~10-12 A) with a responsivity of ~ 102 A/W at zero gate bias. The Schottky barrier
20
+ (WS2/Au interface) induced rectification characteristics in the channel accompanied by the
21
+ excellent photogating effect from solution-processed α-phase CsPbI3 NCs sensitizers, resulting
22
+ in gate-tunable broadband photodetection with a very high responsivity (~104 A/W) and
23
+ excellent sensitivity (~106). Most interestingly, the device shows superior performance even
24
+ under high humidity (50-65%) conditions owing to the formation of cubic α-phase CsPbI3
25
+ nanocrystals with a relatively smaller lattice constant (a = 6.2315 Å) and filling of surface
26
+ vacancies (Pb2+ centres) with the sulfur atoms from WS2 layer, thus protecting it from
27
+ environmental degradation. These results emphasise a novel strategy for developing mixed
28
+ dimensional hybrid heterostructure based phototransistors for futuristic integrated nano-
29
+ optoelectronic systems.
30
+ Keywords: Two dimensional TMDs, Inorganic perovskites, Sensitizers, Asymmetric
31
+ electrodes, Mixed dimensional phototransistors
32
+
33
+
34
+
35
+ Introduction
36
+ Fabrication of high performance phototransistors demands superior channel material, which
37
+ should be of high carrier mobility for high gain bandwidth product, a direct bandgap for
38
+ efficient optical absorption, a thinner layer for full depletion leading to ultralow dark current
39
+ and very low trap state density for low subthreshold swings. Layered semiconducting two
40
+ dimensional (2D) TMDs fulfil most of these requirements [1–4] making them a potential
41
+ candidate for photo-field-effect transistor (photo-FET). However, some trade-offs are found in
42
+ single semiconductor channel based phototransistors, due to the simultaneous occurrence of
43
+ both absorption and amplification processes within the same layer, compromising the device
44
+ performance [5–7]. To overcome this shortcoming as well as for fabrication simplicity,
45
+ attention has been paid on few-layer TMDs over monolayer with a wider spectral
46
+ photoresponse and relatively higher absorbance [8–10], which in turn increases the dark current
47
+ values requiring a higher gate voltage to operate the device in full depletion mode [11]. This
48
+ issue can be addressed via incorporating a Schottky barrier by judiciously selecting the metal
49
+ contacts and utilizing the developed depletion region to facilitate an unidirectional current
50
+ transport, which leads to the significant lowering of dark current in few layered TMD based
51
+ photo-FETs at zero gate bias [12–14]. The built-in electric field of the Schottky interface
52
+ further helps in the separation of photogenerated charge carriers across the channel layer [15]
53
+ leading to zero gate bias driven enhanced photosensitivity, although the responsivity of these
54
+ devices are limited owing to the lower absorption coefficient of 2D TMD material. For the
55
+ further improvement of photoresponsivity, recent works are focused on sensitizing the thin
56
+ channel material with semiconductor nanocrystals (NCs), referred to as sensitizers, having
57
+ excellent absorption characteristics and opposite doping polarity to create a vertical junction
58
+ for subsequent charge separation [16–19].
59
+
60
+ Recently, all-inorganic cesium lead halide (CsPbX3: X = I, Br, and Cl) perovskite NCs have
61
+ drawn tremendous interests in the field of optoelectronics [20], featuring superior emissive
62
+ (approaching photoluminescence quantum yield ~ 100%) [21] characteristics, extremely high
63
+ absorption coefficient [22], large carrier diffusion lengths (>3 mm), fast radiative
64
+ recombination rates and most importantly their improved stability [23] over organic-inorganic
65
+ hybrid perovskites [24,25]. Especially colloidal synthesised zero-dimensional (0D) α-CsPbI3
66
+ NCs exhibit extended photoabsorption band covering the whole visible spectrum with high
67
+ absorption coefficient (~ 3 × 104 cm-1 at 680 nm) [26] and better photostability, making them
68
+ attractive as a photoactive material for high performance optoelectronic devices under ambient
69
+ condition. Weerd et al. have recently reported colloidal CsPbI3 NCs with high quantum yield
70
+ of ~ 98% prepared by hot-injection method which reveals highly efficient carrier multiplication
71
+ as well as longer build-up of free carrier concentration [27]. Among four existing phases (α-
72
+ cubic, β-tetragonal, γ-orthorhombic, and δ-orthorhombic) of CsPbI3 NCs, cubic α-phase has
73
+ high stability due to its low surface-to-volume ratio (lattice constant a = 6.23 Å) without any
74
+ octahedral inclination or lattice distortion in the [PbI6]4− octahedra as well as in the unit
75
+ cell [28–30]. These extraordinary properties of α-phase CsPbI3 NCs provoke to utilize them as
76
+ an effective sensitizer in 2D channel based hybrid phototransistor devices. However, owing to
77
+ the vulnerability towards moisture for halide perovskite NCs, recently, the surface passivation
78
+ of CsPbI3 via coordination of Pb2+ centres with sulphur donors has been explored to protect
79
+ them from environmental degradation [31] to improve the device stability under ambient
80
+ conditions.
81
+ In this work, we present a proof-of-concept for 0D/2D CsPbI3/WS2 based mixed dimensional
82
+ van der Waals heterostructure (MvWH) photo-FETs utilizing the superior photoabsorption
83
+ attributes of α-phase CsPbI3 NCs as sensitizers, along with a sub 5-nm thick 2D WS2 channel
84
+ which acts as an expressway for carrier transport. In our device configuration, the developed
85
+
86
+ built-in electrical field at the heterostructure interface facilitates efficient transfer of
87
+ photogenerated carriers from CsPbI3 NCs into the WS2 layer resulting in an excellent
88
+ broadband visible photoabsorption in the hybrid system. Moreover, asymmetric metal contacts
89
+ (Au-Cr) as source and drain electrodes are purposefully chosen to utilize the large built-in
90
+ potential along the WS2/Au Schottky junction for a diode-like unidirectional current flow and
91
+ effective separation of photogenerated electron-hole pairs, leading to an excellent rectifying
92
+ ratio of ~ 104 with a very low dark current of the order of ~ pA under a low source-to-drain
93
+ bias (VDS) without any applied back gate voltage. Fabricated hybrid phototransistor devices
94
+ exhibit a very high photoresponsivity of ~ 104 A/W at ~ 40 V gate bias upon visible light
95
+ illumination (~ 0.1 µW) due to the photogating effect and a broad spectral bandwidth across
96
+ the entire visible range. In addition, the coordination of sulphur atoms of WS2 layer with Pb2+
97
+ centres of CsPbI3 NCs reveals an excellent device stability under 50-65% humidity condition.
98
+ Our results not only open up new avenues for studying fundamental carrier transport and
99
+ relaxation pathways in hybrid van der Waals heterojunctions, but also pave the way for
100
+ constructing high-performance optoelectronic devices using mixed-dimensional 0D/2D hybrid
101
+ building blocks, rather than using purely 2D layered materials.
102
+ Experimental section:
103
+ Synthesis of α-phase CsPbI3 NCs:
104
+ This is a two-step reaction, where, in the first step Caesium oleate (Cs-OA) was prepared
105
+ followed by the synthesis of α-phase CsPbI3 NCs in the second step.
106
+ i) Firstly, caesium carbonate (Cs2CO3) of 814 mg and 40 ml Olive oil were added in a round
107
+ bottom two-neck flask, which was heated at 120⁰C for 1 hr under vacuum condition.
108
+ Followed by the rise in temperature to 150⁰C under N2 atmosphere for 10-15 mins, the
109
+ desired transparent solution of Cs-OA was stored for further use to synthesize CsPbI3.
110
+ ii) Next, 870 mg of PbI2 and 5 ml of olive oil (instead of 1-octadecene (ODE)) were mixed
111
+ which was heated to 120⁰C under vacuum for 1hr. Thereafter, we swiftly injected 1 ml of
112
+
113
+ Oleyl amine (OLAm) in the reaction mixture under N2 atmosphere to get a transparent
114
+ solution. Finally, preheated (100⁰C) Cs-OA was injected to the reaction mixture and cooled
115
+ immediately in an ice bath to quench the reaction to obtain desired α-phase CsPbI3 NCs.
116
+ Finally, as-synthesized CsPbI3 was purified through centrifugation using excess hexane. The
117
+ centrifugation process (20000 rpm) was repeated for several times to remove excess
118
+ OLAm/olive oil from the product. Finally, the sedimentation was collected and re-dispersed in
119
+ hexane. The collected dispersion was stored in a sealed vial for further characterisations and
120
+ device fabrications.
121
+ CsPbI3/WS2 MvWH photo-FET device fabrication:
122
+ For the fabrication of the CsPbI3/WS2 MvWH photo-FET device, WS2 flakes were
123
+ mechanically exfoliated from the bulk WS2 crystal (2D semiconductor Inc. Scottsdale, AZ,
124
+ USA) using a Scotch tape (3M Inc. USA) on polydimethylsiloxane (PDMS) gel film (Gel-Pak
125
+ Inc. Hayward, CA, USA) and the interested few layer flakes were identified under optical
126
+ microscope, followed by the layer confirmation via Raman characteristics and AFM height
127
+ profile. Note that, mechanically exfoliated flakes were chosen for their high quality and clean
128
+ interface promising greater mobility and high performance device fabrication. Further, for the
129
+ Au contact with WS2, electrodes were patterned in advance on a pre-patterned SiO2/Si (285 nm
130
+ oxide thickness) substrate using e- beam lithography technique and then Cr (5 nm)/Au (30 nm)
131
+ were deposited via e-beam evaporation followed by lift-off with acetone. Then, the selected
132
+ few layer flakes were deterministically transferred on the targeted Au electrodes from PDMS
133
+ gel film following the dry transfer technique to make sure residue free clean transfer. After the
134
+ successful transfer of the flakes on Au electrodes, again electrodes were patterned using second
135
+ step e-beam lithography followed by metal deposition Cr (5 nm)/Au (30 nm) for Cr contacts
136
+ on WS2. To remove the resist residue and improve the contact conductance, the fabricated
137
+ devices were annealed at 150°C for 2 hrs in a high vacuum of ~ 10−3 Torr. Finally, the
138
+ synthesized diluted solution (0.1 mg mL−1) of perovskite NCs was uniformly spin-coated
139
+ several times (varying from one to four) onto WS2 layer with a speed of 2000 rpm.
140
+
141
+ Characterisations and measurements:
142
+ X-ray diffraction (XRD, Philips MRD X-ray diffractometer) patterns were recorded using
143
+ characteristic Cu-Kα (λ = 1.5418 Å) radiation with 2.0° grazing incidence angle. For the
144
+ Transmission electron microscopy (TEM) sample preparation, CsPbI3 NCs solution was
145
+ dissolved in Hexane and then placed into a TEM grid and dried it for few minutes and did the
146
+ measurement. The TEM images were carried out using TECNAI G2 TF20-ST and JEM-2100F
147
+ Field Emission Electron Microscope operating at 200 kV equipped with Gaytan’s latest CMOS
148
+ camera. All the images were proceeding by Digital Micrograph Software for the estimation of
149
+ d-Spacing & Indexing. UV–vis–NIR absorption spectra of as synthesised CsPbI3 samples were
150
+ recorded using a fiber probe-based UV–vis–NIR spectrophotometer (Model: U-2910
151
+ Spectrophotometer, HITACHI) and a broadband light source. Raman and PL spectra were
152
+ recorded using a semiconductor laser of excitation wavelength 532 nm, equipped with a CCD
153
+ detector, an optical microscope of 100x objective lens and a spectrometer (WITec alpha-300R).
154
+ The photogenerated carrier lifetime was measured by exciting the material with a pulsed diode
155
+ laser of wavelength 372 nm and detecting the signal using Edinburgh LifeSpec-II fluorescence
156
+ lifetime spectrometer fitted with a PMT detector. Room-temperature current−voltage
157
+ characteristics were recorded using a Keithly semiconductor parameter analyzer (4200 SCS)
158
+ in the presence of an Argon laser (514 nm) and a broadband solar simulator (AM 1.5, 100
159
+ mW/cm2) as a visible light source.
160
+ Results and discussion
161
+
162
+
163
+ FIG. 1. (a) Rietveld refinements (α-phase fitting) of the XRD pattern of a film of cubic CsPbI3
164
+ NCs. (b) 1×1 3D VESTA visualization image of α-CsPbI3 cubic crystal structure. (c) Typical
165
+ HRTEM image of CsPbI3 NCs with an energy 200 keV revealing cubic morphology. (d) FFT
166
+ patterns from a region marked by the red dotted square on the micrograph (c). (e) A magnified
167
+ view of the corresponding HRTEM image in the selected yellow square region on micrograph
168
+ (c). (f) SAED pattern of α-CsPbI3 NCs showing well defined diffraction spots indexed as (200),
169
+ (220) and (020) planes viewed along [004] zone axis.
170
+
171
+ To study the crystal structure of as-synthesized CsPbI3 NCs, we have recorded X-ray
172
+ diffraction pattern, followed by their fitting with Rietveld refinement full proof software, which
173
+ are presented in Fig. 1(a). An excellent agreement with fitted results indicates the growth of
174
+ single-phase (α-phase) cubic CsPbI3. The crystal structure of CsPbI3 NCs visualized using
175
+ VESTA 3D software through Rietveld fitting is shown in Fig. 1(b). The VESTA 3D (1x1)
176
+ structure shows the absence of any octahedral inclination in the perovskite NCs, which is
177
+ known to be beneficial for achieving higher stability under laboratory ambient (45-50%
178
+ humidity). Typical high resolution transmission electron microscopy (HRTEM) image reveals
179
+ almost cubic shape of the synthesised NCs (15.05 nm × 18.04 nm), as shown in Fig. 1(c).
180
+ Corresponding first Fourier transform (FFT) pattern presented in Fig. 1(d) of the red squared
181
+
182
+ (a)
183
+ C
184
+ (200)
185
+ &
186
+ Observed
187
+ (100)
188
+ 米0
189
+ Calculated
190
+ 15.05nm
191
+ Intensity (arb. units)
192
+ (210)
193
+ Difference
194
+ d= 0.62 nm
195
+ Braggposition
196
+ 11001
197
+ d= 0.62 nm
198
+ [100]
199
+ 10nm
200
+ 10
201
+ 20
202
+ 20 (degree)
203
+ 30
204
+ 40
205
+ 50
206
+ (b)
207
+ (d)
208
+ ZA[002]
209
+ a-CsPbl3
210
+ f)
211
+ ZA [004]
212
+ a-CsPbl
213
+ (220)
214
+ (210)
215
+ (200)
216
+ (200)
217
+ (200)
218
+ (210)
219
+ (220)
220
+ (220) ( ()
221
+ 2 nm
222
+ 2 1/nm
223
+ (020)
224
+ Cs
225
+ Pbportion of the Fig. 1(c), shows pure cubic α-phase pattern along the zone axis [002]. Whereas,
226
+ the high-resolution fringe pattern from the yellow squared region of Fig. 1(c) shows a d-spacing
227
+ of 0.62 nm, which is in well matched with the cubic α-phase of CsPbI3 [30], as shown in Fig.
228
+ 1(e). The result indicates (100) directional growth of cubic phase CsPbI3 NCs, which is in well
229
+ agreement with our previously reported results [30]. Corresponding selected area electron
230
+ diffraction (SAED) patterns shown in the Fig. 1(f), with indexed (200), (220) and (020) planes
231
+ along the zone axis [004], also corroborate the pure cubic structure of synthesized CsPbI3.
232
+ Figure 2(a) presents the optical absorption and emission properties of the as-synthesised α-
233
+ phase CsPbI3 NCs in the visible wavelength range with an absorption maxima at ~ 680 nm and
234
+ the corresponding bandgap value is ~ 1.814 eV [30], extracted from the Tauc plot shown in
235
+ Fig. S1 within the Supplimental Material. Further, the photoluminescence (PL) maxima at ~
236
+ 687 nm confirms the formation of excitons (Xα) across the direct bandgap (∼1.80 eV) of α-
237
+ phase CsPbI3 NCs represented via blue curve in Fig. 2(a). The Gaussian line shape of the PL
238
+ spectrum and the absence of any other PL peaks clearly dictate the synthesis of pure α-phase
239
+ CsPbI3 without presence of any mixed phase. To examine the charge transfer mechanism at the
240
+ CsPbI3 NCs/WS2 interface, Raman spectroscopy and micro-PL (µ-PL) measurements have
241
+ been carried out by spin-coating of a dilute solution of CsPbI3 NCs uniformly on the exfoliated
242
+ WS2 surface. For the room temperature µ-Raman-PL measurements, samples have been
243
+ excited with a CW laser having a wavelength of 532 nm with the laser power being kept at a
244
+ very low value to avoid any local heating induced sample degradation. Figure 2(b) represents
245
+ comparative Raman spectra of WS2 and mixed dimensional van der Waals heterostructure
246
+ (MvWH) samples, showing intense in-plane 2LA+E12g Raman modes at ∼ 351 cm–1 and out-
247
+ of-plane vibrational A1g peaks at ∼ 420 cm–1 [32]. The A1g vibrational Raman mode, which
248
+ preserves the symmetry of the lattice, is clearly red-shifted by ∼ 4.7 cm-1 in case of CsPbI3
249
+ decorated WS2 layer [inset of Fig. 2(b)], revealing the interfacial charge transfer phenomena.
250
+
251
+ The external electron doping in 2D WS2 leads to the filling-up of antibonding states of the
252
+ conduction band, mostly made up of d z2 orbitals of transition metal atoms [33]. This makes the
253
+ bonds weaker and the A1g peak of pristine WS2 is shifted towards a lower wavenumber on
254
+ significant electron doping from CsPbI3 NCs [34].
255
+
256
+ FIG. 2. (a) Absorption (Green) and photoluminescence (Blue) spectra of as-synthesised cubic
257
+ phase CsPbI3 NCs. (b) Comparative Raman spectra of WS2 flakes before and after CsPbI3 NCs
258
+ decoration showing characteristic E2g and A1g peaks of layered WS2. Inset shows the magnified
259
+ image of the out-of-plane A1g mode revealing a clear peak shift to lower wavenumbers due to
260
+ electron doping in WS2 flakes from CsPbI3 NCs. (c) Deconvoluted PL spectra of (i) a bare ML
261
+ WS2 flake and (ii-v) the heterostructure samples with varying number of spin coated layers of
262
+ CsPbI3 NCs on the WS2 flake. The spectra (ii), (iii), (iv) and (v) represent the PL emission from
263
+ the first, second, third and fourth spin coated layers of CsPbI3 NCs, respectively. The green
264
+ (red) peak represents the A excitonic (A- trionic) emission from ML WS2 flakes and the blue
265
+ peak represents the emission from band to band transition of cubic α-phase CsPbI3 NCs. (d)
266
+ Energy band diagram of CsPbI3/WS2 hybrid heterostructures showing effective electron doping
267
+ in WS2 from CsPbI3 sensitizers and hole trapping in the NCs giving rise to a strong photogating
268
+ effect. (e) Schematic representation of the charge transfer mechanism in 0D/2D CsPbI3/WS2
269
+ hybrid heterostructures giving rise to trion formation in ML WS2. (f) Normalised time resolved
270
+ PL decay curves of CsPbI3 NCs (Blue curve) and CsPbI3/WS2 hybrids (Red curve), measured
271
+ using an excitation wavelength of 372 nm.
272
+
273
+ On the other hand, the monolayer (ML) WS2 PL emission characteristics [Fig. 2c(i)] consist of
274
+ a strong A-excitonic emission at ∼ 1.995 eV, corresponding to the direct band-to-band
275
+ 550
276
+ 600
277
+ 650
278
+ 700
279
+ 750
280
+
281
+
282
+
283
+
284
+ PL intensity (arb. units)
285
+ Wavelength (nm)
286
+
287
+
288
+
289
+
290
+ (v)
291
+ (iv)
292
+ (iii)
293
+ (ii)
294
+
295
+
296
+ (i)
297
+ WS2
298
+ A0
299
+ 160
300
+ 180
301
+ 200
302
+ 220
303
+ 240
304
+ tav = 33.7 ns
305
+ PL (arb. units)
306
+ Time (ns)
307
+ CsPbI3
308
+ CsPbI3/WS2
309
+ tav = 40.2 ns
310
+ 300
311
+ 400
312
+ 500
313
+ 410 420 430
314
+
315
+
316
+
317
+ 4.7 cm-1
318
+ 400
319
+ 410
320
+ 420
321
+ 430
322
+ 440
323
+
324
+
325
+
326
+ 4.7 cm-1
327
+ CsPbI3/WS2
328
+
329
+
330
+ Intensity (arb. units)
331
+ Raman shift (cm-1)
332
+ WS2
333
+
334
+ A-
335
+ CsPbI3/WS2
336
+ 500
337
+ 600
338
+ 700
339
+ 800
340
+ Wavelength (nm)
341
+ Nor. Abs./PL (arb. units)
342
+ Absorption
343
+ Photoluminescence
344
+ WS2
345
+ CsPbI3
346
+
347
+ Electron
348
+ transfer
349
+ Hole
350
+ trapping
351
+ Charge
352
+ transfer
353
+ WS2
354
+ Exciton
355
+ Trion
356
+ CsPbI3
357
+
358
+ Electrons
359
+ Holes
360
+ (d)
361
+ (c)
362
+ (a)
363
+ (f)
364
+ (b)
365
+ (e)
366
+ Increasing CsPbI3 coating layer
367
+
368
+ transitions at the K (and/or K’) point of the Brillouin zone, and a weaker trionic A– emission at
369
+ ∼ 1.965 eV with a binding energy of ∼ 30 meV, which are in good agreement with the
370
+ previously reported results [34]. It is to be noted that, ML WS2 is purposefully chosen for the
371
+ charge transfer study via PL measurements due to its extraordinary luminescence property at
372
+ room temperature owing to the direct bandgap transition. The existence of trions in the room
373
+ temperature emission spectrum indicates the unintentional doping in the un-passivated WS2
374
+ flake from the substrate as well as the surrounding environment [35,36]. A systematic PL study
375
+ of the hybrid structure with increasing layer numbers of CsPbI3 NCs spin-coated over ML WS2
376
+ shows a pronounced excitonic-PL quenching in MvWH as compared to both the pristine
377
+ materials (Fig. S2 within the Supplimental Material). Further, the CsPbI3/WS2 MvWHs show
378
+ relatively broad PL spectra with combined contributions from CsPbI3 NCs as well as ML WS2
379
+ and the spectral shape changes with increasing CsPbI3 spin-coated layer numbers [Figs. 2c(ii-
380
+ v)]. To explore the effect of CsPbI3 coating over WS2, we have fitted each spectrum with three
381
+ Gaussian peaks containing the characteristics excitonic features of both the materials and
382
+ analysed their intensity variation with increasing concentration of CsPbI3 treated on WS2
383
+ flakes, as shown in Fig. 2(c). It is noticed that the distinctive trion peak (A–) of WS2 becomes
384
+ prominent with increasing density (coating number) of CsPbI3 NCs and finally excitonic to
385
+ trionic (integrated intensity) crossover is observed above a critical concentration of CsPbI3 NCs
386
+ (Fig. S3 within the Supplimental Material). The possible explanation behind these observations
387
+ is as follows: upon illumination, photoexcited electron-hole pairs are generated in both WS2
388
+ and CsPbI3, however, due to the type-II energy band alignment of the heterostructures,
389
+ electrons are easily transferred from CsPbI3 NCs to ML WS2, as illustrated in Fig. 2(d). On the
390
+ other hand, photogenerated holes remain trapped in the NCs, resulting in a reduced
391
+ recombination rate and giving rise to a quenched excitonic PL intensity for CsPbI3 NCs and
392
+ increased trionic emission in ML WS2. The enhanced generation rate of trions results in the
393
+
394
+ reduced density of excitons inside the system, leading to the suppression of excitonic peak
395
+ intensity and the dominance of the trion peak in the PL spectra of MvWH, as schematically
396
+ depicted in Fig. 2(e). [34] To further confirm the charge transfer phenomena, time-resolved
397
+ photoluminescence (Tr-PL) spectra have been measured. Fig. 2(f) shows the Tr-PL decay
398
+ curves of CsPbI3 NCs (blue curve) and CsPbI3/WS2 MvWHs (red curve). The PL decay curves
399
+ have been fitted using a bi-exponential function to extract the average excitonic life time (τav).
400
+ The τav of CsPbI3 NCs decreases from 40.2 ns to 33.7 ns after hybridization with WS2,
401
+ corroborating the successful charge transfer mechanism from CsPbI3 NCs to WS2 flakes. As a
402
+ conclusion, the type-II band alignment in CsPbI3/WS2 heterostructure facilitates an efficient
403
+ electron–hole pair separation and strong electron doping into WS2 channel, making the hybrid
404
+ system ideal for fabrication of superior performance phototransistor devices [37].
405
+
406
+ FIG. 3. (a) Schematic 3D view of the fabricated back gated phototransistor comprising of
407
+ CsPbI3 sensitized WS2 channel with asymmetric electrodes (Au and Cr) acting as a source and
408
+ drain. An optical micrograph of the device is shown in the inset. (b) Linear IDS-VDS
409
+ characteristics of three fabricated devices with different source-drain contacts (i) Cr-Cr
410
+ (Yellow curve), (ii) Au-Au (Orange curve) and (iii) Au-Cr (Brown curve) without any back
411
+ gate bias. Inset: the corresponding semi-logarithmic IDS-VDS characteristics plots. (c) Transfer
412
+
413
+ (a)
414
+ (b)
415
+ 6
416
+ 10-
417
+ 4
418
+ 10-11
419
+ (vu)
420
+ 2
421
+ 10-13
422
+ T
423
+ Vps (M)
424
+ U
425
+ DS
426
+ -2
427
+ Cr/Cr
428
+ Au
429
+ Au/Au
430
+ 4
431
+ Au/Cr
432
+ Watoms
433
+ Si02
434
+ .6
435
+ CsPbl, NCs
436
+ Satoms
437
+ -2
438
+ -1
439
+ 0
440
+ 1
441
+ 2
442
+ V
443
+ Ds (V)
444
+ (c)
445
+ (d)
446
+ 10°
447
+ 60
448
+ 20
449
+ A0
450
+ 5V
451
+ 10V
452
+ 15V
453
+ (vu)
454
+ (vu)
455
+ 15
456
+ 20V
457
+ 'DS
458
+ 10-11
459
+ DS
460
+ 10
461
+ -20
462
+ 20
463
+ 40
464
+ 20
465
+ WS
466
+ 5
467
+ CsPbI,/WS
468
+ 4-0
469
+ 0
470
+ -20
471
+ 0
472
+ 20
473
+ 40
474
+ 0.0
475
+ 0.1
476
+ 0.2
477
+ 0.3
478
+ 0.4
479
+ 0.5
480
+ VGs (V)
481
+ V
482
+ (V)
483
+ DS(IDS-VGS) characteristics of the CsPbI3/WS2 hybrid transistor (Green curve) and control WS2
484
+ transistor (Blue curve) devices in linear scale and logarithmic scale (Inset). The current value
485
+ in the accumulation region decreases and the threshold voltage is shifted to a higher positive
486
+ voltage in hybrid device due to charge transfer through the CsPbI3/WS2 junction. (d) Output
487
+ characteristics of the hybrid device with varying gate voltage.
488
+
489
+ The CsPbI3 NCs/WS2 MvWH photo-FET with asymmetric metal contacts has been
490
+ demonstrated by exploiting the Schottky barrier induced dark current suppression for the zero
491
+ gate bias driven photosensitivity of the device with a simpler fabrication technique. Figure
492
+ 3(a) schematically demonstrates the as-fabricated phototransistor device structure consisting
493
+ of a few layer WS2 channel and asymmetric Cr and Au electrodes as source and drain terminals,
494
+ respectively. The inset shows the optical micrograph of the connected few layer WS2 flake
495
+ (5×2 µm2) having a thickness of ~ 4.5 nm corresponding to 4-5 atomic layers of WS2, further
496
+ corroborated by the AFM analysis (Fig. S4 within the Supplimental Material). The deposition
497
+ of a lower work function Cr (ΦCr = 4.5 eV) and higher work function Au (ΦAu = 5.1 eV) on n-
498
+ type WS2 as asymmetric contacts reveals room temperature rectifying diode characteristics
499
+ with rectification ratio up to 5.2 × 102 even at zero applied back gate bias, as depicted via the
500
+ current-voltage (IDS-VDS) characteristics in Fig. 3(b). Under an applied reverse drain voltage,
501
+ the potential barrier height between Au and WS2 becomes higher to suppress the current flow
502
+ through the junction compared to Cr, and thus exhibits the IDS-VDS characteristics of an ideal
503
+ diode [13]. The current through the Au–WS2-Cr device follows the diode behaviour and the
504
+ Schottky barrier (Φb) at Au-WS2 interface is capable of reducing the dark current significantly
505
+ making the device architecture an ideal prototype for operating in full depletion mode even at
506
+ zero gate bias, leading to a very high ON-OFF ratio of the device. On the other hand, a linear
507
+ IDS-VDS characteristics with a comparatively larger current value (100 nA) confirms the
508
+ formation of an Ohmic-like junction with a very low contact resistance for Cr-WS2-Cr
509
+ device [38]. Further, Au-WS2-Au system reveals a rectifying output characteristics with
510
+ relatively lower current than Cr contacts, confirming a typical high resistive back-to-back
511
+
512
+ Schottky diode [15]. The energy band alignment with different metal contacts is schematically
513
+ depicted in Fig. S5 within the Supplimental Material, revealing an easy current flow through
514
+ the Ohmic Cr junction and restricted flow via built-in potential barrier in the Au Schottky
515
+ junction. Fig. 3(c) shows the transfer (IDS-VGS) characteristics of the WS2 phototransistor with
516
+ asymmetric Cr–Au contacts at a reverse drain voltage of -2V revealing excellent n-type channel
517
+ properties at room temperature with off-currents of the order of 10 pA and the transistor ON-
518
+ OFF ratio ~104. Further, to understand the effect of CsPbI3 treatment on the device
519
+ performance, the CsPbI3/WS2 hybrid transistor characteristics is compared with the pristine
520
+ WS2 one, referred to as the control device. The incorporation of sensitizing perovskite NCs on
521
+ 2D-WS2 layer results in a junction formation via Fermi level alignment in equilibrium under
522
+ dark condition. In this process, the draining of electrons from the WS2 channel towards CsPbI3
523
+ NCs results in the depletion of majority carriers in WS2 leading to the lowering of the current
524
+ flow in the channel under dark condition. Further, we have studied the output characteristics
525
+ of the hybrid phototransistor on application of gate voltage varying from 0 to +20 V, as depicted
526
+ in Fig. 3(d). For a higher positive gate voltage, more electrons are induced in the WS2 channel
527
+ and the transistor has a higher current in the saturation state. On the other hand, under a negative
528
+ gate bias a small amount of current flows through the channel owing to the depletion of carriers,
529
+ leading to the OFF state of the transistor [Fig. 3(c)].
530
+ The performance of the CsPbI3/WS2 MvWH photo-FET has been analysed by recording the
531
+ room temperature IDS-VDS characteristics for zero gate bias under dark as well as visible
532
+ illumination using a Newport solar simulator having broadband emission with irradiance of
533
+ 100 mW/cm2 under air mass (AM) 1.5G condition, as shown in Fig. 4(a). For comparison, the
534
+ characteristics of the pristine WS2 control device is also presented. The suppression of dark
535
+ current to the order of tens of pA even without any gate bias along with the significant reduction
536
+ of noise currents are attributed to the built-in electric field at the Au/WS2 Schottky junction,
537
+
538
+ FIG. 4. (a) Comparative IDS-VDS characteristics of pristine WS2 and CsPbI3/WS2 MvWH
539
+ photo-FET under dark and illumination via broadband light source for zero gate bias. (b)
540
+ Spectral responsivity curves of MvWH photo-FET at VGS = 0V with increasing reverse VDS
541
+ from 0V to -2V, as shown via yellow, green and blue curves. The blue curve represents the
542
+ responsivity of the photo-FET at a maximum VDS of -2V, while the spectral responsivity of the
543
+ control device (orange curve) showing an order of magnitude lower device response at same
544
+ VDS. (c) COMSOL Multiphysics simulated E-field distribution at the vicinity of the hybrid
545
+ system upon excitation with an excitation wavelength of 680 nm. (d) The transfer
546
+ characteristics (IDS-VGS) of MvWH photo-FET for a range of incident powers (from 0.1 to 35
547
+ μW) with an illumination of wavelength 514 nm at VDS = -2 V. (e) The variation of responsivity
548
+ of the device with incident illumination power for VGS varying from 0 to 40 V. (f) The shift in
549
+ threshold voltage (ΔVTh) with increasing illumination power (Pin) fitted with a power law. Blue
550
+ dots represent the extracted data points from panel (a) and green line represents the fitted curve.
551
+ Inset: The power law fit of the variation of photocurrent with incident power at VDS = -2V and
552
+ VGS = 40V showing a sublinear photocurrent dependency with incident optical power.
553
+
554
+ which further helps in effective separation of photogenerated carriers created in WS2 channel.
555
+ On illuminating the heterojunction device, the reverse current tends to increase due to the
556
+ collection of photogenerated minority carriers (holes) at the electrodes. The photo-to-dark
557
+ current ratio of MvWH photo-FETs by illuminating with a broadband light source is estimated
558
+ to be much higher compared to the pristine WS2 one (~1000 times) under the applied reverse
559
+ bias condition, which reaches to a value of ~1.08×106 at VDS of -2V, as shown in Fig. S6 within
560
+ the Supplimental Material. The decoration of WS2 channel with superior light absorbing
561
+
562
+ (a)
563
+ (b)
564
+ (c)
565
+ 120
566
+ (AW)
567
+ ×102
568
+ E2/E.?
569
+ 6
570
+ 80
571
+ 10-8
572
+ 4
573
+ MvWHlight
574
+ 40
575
+ MvWH dark
576
+ WS,light
577
+ 2
578
+ WS,dark
579
+ WS,1
580
+ 0
581
+ -2
582
+ -1
583
+ 0
584
+ 1
585
+ 2
586
+ 300400500600700800900
587
+ Vps (V)
588
+ Wavelength(nm)
589
+ (d)
590
+ (e)
591
+ (f)
592
+ 0.3
593
+ -10
594
+ 35μW
595
+ (A/W)
596
+ - 40V
597
+ 20V
598
+ 200
599
+ 10-7
600
+ 104
601
+ -15
602
+ (nA)
603
+ 40V
604
+ 30V
605
+ -10V
606
+ 150
607
+ (μA)
608
+ 0.2
609
+ +
610
+ -OV
611
+ 20
612
+ DS
613
+ 100
614
+ 10-1
615
+ 0μw
616
+ Tocp0.48
617
+ 10
618
+ 50
619
+ 10-13
620
+ AV
621
+ -25
622
+ -20
623
+ Vcs (M)
624
+ 40
625
+ 0
626
+ 10
627
+ 20
628
+ 30
629
+ 30
630
+ Pin (μW)
631
+ Resi
632
+ 10
633
+ -35
634
+ ocp0.17
635
+ 0.0
636
+ 40
637
+ -20
638
+ 20
639
+ 40
640
+ 0.1
641
+ 1
642
+ 10
643
+ 0
644
+ 7
645
+ 14
646
+ 0
647
+ 21
648
+ 28
649
+ 35
650
+ (V)perovskite CsPbI3 NCs facilitates enhancement in the photocurrent by elevating the
651
+ photogenerated carriers in the channel via efficient charge transfer from CsPbI3 to WS2 due to
652
+ type-II energy band alignment [7,19]. This explains the significant enhancement (~103 times)
653
+ of response of MvWH transistor over the control one, revealing the role of photoabsorbing
654
+ CsPbI3 NCs in boosting the performance of the phototransistor. Further, the spectral
655
+ responsivity of the fabricated MvWH photo-FET, the most important figure of merit to evaluate
656
+ a detector performance, has been studied displaying a broadband spectral photoresponse
657
+ covering the entire visible wavelength range, as shown in Fig. 4(b). It may be noted that a peak
658
+ responsivity of ~1.05×102 A/W at ~ 460 nm at an applied bias (VDS) of -2 V is achieved, which
659
+ is close to the C-exciton absorption edge of WS2. Two other peaks at ~ 620 nm (R ~ 0.97×102
660
+ A/W) and ~ 720 nm (R ~ 0.77×102 A/W) correspond to the direct bandgap absorption of few
661
+ layer WS2 and α-phase CsPbI3 NCs, respectively. Further, the spectral responsivity increases
662
+ with increasing reverse VDS that assists in the efficient extraction of photogenerated carriers.
663
+ On the other hand, the control WS2 based device also exhibits a similar trend with increasing
664
+ bias showing a maximum peak responsivity of ~ 10 A/W at ~ 460 nm at -2 V applied VDS (Fig.
665
+ S7 within the Supplimental Material). It is to be noted that the decoration of CsPbI3 NCs on
666
+ WS2 flakes not only improves the detector responsivity by more than 10-fold but also extends
667
+ the spectral responsivity window up to 800 nm, as shown comparatively in Fig. 4(b). Hence,
668
+ the decoration of WS2 active channel layer with excellent photoabsorbing CsPbI3 NCs appear
669
+ to be a promising approach for next generation high performance optoelectronic applications.
670
+ To further investigate the role of CsPbI3 in photocarrier generation and efficient charge
671
+ transfer, the electromagnetic simulations have been performed using the COMSOL
672
+ Multiphysics software. Figure 4(c) shows the electric field distribution of the hybrid
673
+ CsPbI3/WS2 device, illuminated with an electromagnetic plane wave of wavelength λ = 680
674
+ nm from top, which propagates through air and the nanostructure. The distribution clearly
675
+
676
+ depicts that the electric field is trapped along the edges of the nano-cubes of CsPbI3 with the
677
+ maximum confinement occurring near the base (as demonstrated by the colour index profile),
678
+ resulting in strong charge transport in CsPbI3/WS2 hybrid heterostructure.
679
+ Further to explore the impact of gate bias on transistor performance, the photo-induced transfer
680
+ characteristics (IDS−VGS) of the WS2/CsPbI3 MvWH photo-FET is recorded under the dark
681
+ (black markers) and 514 nm illumination with a range of optical powers (from 0.1 μW to 35
682
+ μW) at a constant VDS of -2V, as illustrated in Fig. 4(d). The corresponding logarithmic current
683
+ representation is depicted in the inset. Under illumination of a fixed power of 35 μW, the drain
684
+ current of the MvWH photo-FET is enhanced by ~ 13 times (from ~ 20 nA to ~ 0.26 μA) at a
685
+ constant gate voltage of ~ 40V, manifested by the strong photoabsorption in CsPbI3 and
686
+ subsequent transfer of photoexcited electrons to the WS2 channel. Further, with the increase of
687
+ laser power, the photocurrent significantly increases in the accumulation region (i.e. VGS>VTh)
688
+ and the transfer curves are gradually shifted to a negative gate voltage. As illustrated in Fig.
689
+ 2(d), the favourable energy band alignment rules out the possibility of hole injection from
690
+ CsPbI3 into WS2, leading to the trapped holes induced strong photogating effect in the hybrid
691
+ system. This leads to significant photocurrent increment in the accumulation region and
692
+ negative threshold voltage shift (ΔVTh) with increasing incident power density of
693
+ illumination [39]. To investigate in greater detail, the calculated responsivity as a function of
694
+ illumination power has been plotted for different gate bias voltages in Fig. 4(e). Here, the
695
+ responsivity value increases with increasing positive gate bias in case of MvWH photo-FETs
696
+ and reaches to a high value of ~ 1.1 × 104 A W−1 at a back gate voltage of ~ 40 V under an
697
+ illumination power of 0.1 μW, which is quite remarkable compared to those previously
698
+ reported 0D/2D hybrid phototransistors [37,39]. Note that, for all the gate voltages, the
699
+ measured responsivity dropped with increasing power because of the saturation of sensitizing
700
+ traps in CsPbI3 NCs, which is a characteristic footprint of trap-dominated photoresponse [40–
701
+
702
+ 42] . Further, we have extracted the threshold voltage via extrapolating the linear region of
703
+ each transfer curve under different incident laser powers and the shift in threshold voltage is
704
+ plotted as a function of incident power. The variation is fitted with the power law function
705
+ 𝑉𝑇ℎ ∝ 𝑃𝑏, as depicted in Fig. 4(f), to understand the possible photoconduction mechanism. The
706
+ extracted fitting exponent, b ~ 0.17 clearly indicates a sublinear dependency on laser power
707
+ confirming the existence of photogating dominant carrier conduction in MvWH photo-
708
+ FETs [43]. Further, it is also observed that the change in VTh is large in the lower power region
709
+ and starts to saturate gradually at a higher power owing to the saturated trap states present in
710
+ sensitizer interface leading to the saturation of the photogating effect. The photocurrent IPh =
711
+ IPhoto − IDark versus gate voltage for different illumination intensity [Fig. 5(a)] shows a strong
712
+ modulation with VGS, and a clear maximum in response can be identified around +35 V. The
713
+ strongest response of the FET device corresponds to the region with highest transconductance,
714
+ due to the favourable Fermi level alignment, for low-contact resistance operation leading to
715
+ many cycles of electron circulation to produce maximum gain. Hitherto, in this region the FET
716
+ device operates at a relatively higher dark current, compromising the signal-to-noise ratio
717
+ (SNR) of the device, which is also a very important figure of merit of photo-FETs. The SNR
718
+ defined as IPhoto/IDark is illustrated in the same panel, Fig. 5(a), which reveals the potential of
719
+ 0D/2D hybrid phototransistors for highest sensitivity detection in its depletion regime with VGS
720
+ from 0 to 5V. In this region, a lowest dark current and a maximum sensitivity are achieved,
721
+ despite the devices’ concurrent drop in the photocurrent. So the maximum sensitivity of the
722
+ device can be achieved via contact engineering where the transistor is operated in the depletion
723
+ region, even without applying any gate bias, hitherto unreported for photo-FET devices. While
724
+ the peak responsivity of our device is comparable or superior to the reported 2D materials based
725
+ hybrid phototransistor devices with perovskite sensitizers, the sensitivity is found to be
726
+ significantly higher without application of any external gate bias (see Table 1). These results
727
+
728
+ illustrate the superior performance of broadband phototransistor, with ultrahigh sensitivity and
729
+ responsivity, using CsPbI3 NCs sensitized 2D WS2 layer.
730
+
731
+ FIG. 5. (a) Back-gate bias dependent photocurrent (right axis) and photo-to-dark current ratio,
732
+ (left axis) of the phototransistor device under five different illumination intensities (from 0.1
733
+ µW to 10 µW) for 514 nm. Despite the strongest photoresponse at higher gate bias (VGS ≈ 40
734
+ V), highest sensitivity of the device is achieved in the depletion regime (VGS ≈ 0V). The
735
+ schematic representation of channel current transport mechanism and energy band diagram of
736
+ the asymmetric contact hybrid phototransistor under reverse drain-source voltage with (b) zero
737
+ and (c-d) different gate bias conditions.
738
+
739
+ On the other hand, a remarkable photoresponse of CsPbI3/WS2 MvWH photo-FET is explained
740
+ by considering the influence of positive gate voltage on energy band alignment at the contact
741
+ interfaces and heterostuctures leading to efficient charge injection into n-type WS2 channel
742
+ Table 1. Comparison of device performances with reported 2D material based hybrid photo-
743
+ FETs with perovskite sensitizers
744
+
745
+ (a)
746
+ (b)
747
+ hy
748
+ High sen sitivity
749
+ High photoresponse
750
+ 104
751
+ 90
752
+ 0.1 μW
753
+ 0.5 μW
754
+ (vu)
755
+ 5 μW
756
+ 10μW
757
+ CsPbI3
758
+ 103
759
+ 0.00
760
+ 60
761
+ Photo
762
+ 10
763
+ Au
764
+ WS2
765
+ Cr
766
+ 30
767
+ 10
768
+ -ve
769
+ +ve
770
+ 100
771
+ 00
772
+ 20
773
+ 0
774
+ 20
775
+ 40
776
+ Vcs (V)
777
+ Underillumination
778
+ (c)
779
+ (p)
780
+ hy
781
+ CsPbl
782
+ CsPbI3
783
+ Au
784
+ Au
785
+ Cr
786
+ WS2
787
+ -ve
788
+ -ve
789
+ Cr
790
+ WS2
791
+ +ve
792
+ +ve
793
+ Depletion region
794
+ Accumulation regionDevice
795
+ structure
796
+ Sensitizer
797
+ Operational
798
+ spectral
799
+ range
800
+ Idark
801
+ w/o
802
+ applied
803
+ VGS
804
+ Iphoto/Idark
805
+ @
806
+ VGS=0V
807
+ Responsivity
808
+ for different
809
+ values of VGs
810
+ Ref.
811
+ Au / ML
812
+ WS2 / Au
813
+ CH3NH3PbI3
814
+ 450-700 nm
815
+ 5nA
816
+ 104
817
+ 2.5 A/W @ 0V [44]
818
+ Au / ML
819
+ MoS2 / Au
820
+ CsPbBr3
821
+ 350-550nm
822
+ 0.2 nA
823
+ 103
824
+ 4.4 A/W @ 0V [45]
825
+ Au / Ti / ML
826
+ MoS2 / Ti /
827
+ Au
828
+ Ch3NH3PbBr3
829
+ / CsPbI3-xBrx
830
+ 532 and 355
831
+ nm
832
+ 4 nA
833
+ 104
834
+ 7 × 104 A/W
835
+ @ 60V
836
+ [46]
837
+ Au / Ti / FL
838
+ MoS2 / Ti /
839
+ Au
840
+ CsPbBr3
841
+ 405 nm
842
+ 10 nA
843
+ 10
844
+ 4.7 × 104 A/W
845
+ @ 20V
846
+ [47]
847
+ Au / FL BP /
848
+ Au
849
+ CsPbBr3
850
+ 405 nm
851
+ 2 nA
852
+ 102
853
+ 357.2 mA/W
854
+ @ 0V
855
+ [48]
856
+ Au / ML
857
+ MoS2 / Au
858
+ CsPbI3-xBrx
859
+ 532 nm
860
+ 0.2 µA
861
+ 103
862
+ 1.13 × 105
863
+ A/W @ 60V
864
+ [49]
865
+ Au / FL BP /
866
+ Al / Au
867
+ MAPbI3−xClx
868
+ 400-900 nm
869
+ 0.1 µA
870
+ /µm
871
+ 102
872
+ 4 × 106 A/W
873
+ @ 40V
874
+ [50]
875
+ Au / Ti / FL
876
+ MoSe2(WSe2
877
+ ) / Ti / Au
878
+ CsPb(Cl/Br)3
879
+ 455 nm
880
+ 0.8 nA
881
+ 10
882
+ 102 A/W @
883
+ 50V
884
+ [51]
885
+ Au / Cr / FL
886
+ Ta2NiSe5 / Cr
887
+ / Au
888
+ CH3NH3PbI3
889
+ 800 nm
890
+ 3.5 µA
891
+ 10
892
+ 2.4 × 102 A/W
893
+ @ 0V
894
+ [37]
895
+ Au / Cr / FL
896
+ WS2 / Au /
897
+ Cr
898
+ α-phase
899
+ CsPbI3
900
+ 400-800 nm
901
+ 2 pA
902
+ 106
903
+ 104 A/W @
904
+ 40V
905
+ Our
906
+ work
907
+
908
+ layer from photoabsorbing CsPbI3 NCs. As illustrated in Fig. 5(b), the Schottky barrier at the
909
+ Au/WS2 interface is high enough to inhibit the charge conduction mechanism across the WS2
910
+ channel layer at reverse drain bias without any gate electric field under dark condition. Hence,
911
+ the transistor immediately goes to the OFF state with very low dark current in the order of pA.
912
+ At this condition, when the visible light is illuminated on the 0D/2D heterostructure, the
913
+ photogeneration takes place in both CsPbI3 NCs as well as WS2 channel layer, as depicted in
914
+ Fig. 5(b). The effective photogenerated carrier separation takes place by the built-in electric
915
+ field at the Schottky junction (WS2/Au) as well as at CsPbI3/WS2 interfaces. The subsequent
916
+ transition of photoexcited electrons from CsPbI3 to WS2 starts to populate the active channel
917
+
918
+ layer which are collected by the external electrodes under an applied reverse VDS, leading to
919
+ the photoresponsivity of ~ 102 A/W at VDS = ‒2V and VGS = 0V. Further, the application of a
920
+ back gate voltage (VGS) to the device modulates the Schottky barrier height at Au/WS2 interface
921
+ as shown in Figs. 5(c)-(d) [52,53]. An application of negative gate bias (VGS < VTh) increases
922
+ the barrier height leading to the transistor operation in the depletion region, where the
923
+ photosensitivity (IPhoto/IDark) of the device is maximum. On the other hand, on increasing the
924
+ VGS beyond VTh initiates the lowering of the Schottky barrier at Au/WS2 interface, resulting in
925
+ a higher magnitude of charge carrier injection from the Au electrode to WS2 channel layer
926
+ through thermoionic as well as tunnelling mechanisms, as illustrated in Fig. 5(d). Thus, the
927
+ cumulative effects of CsPbI3 NCs decoration mediated strong photogating phenomena as well
928
+ as the gate voltage induced Schottky barrier lowering result in a drastic enhancement of the
929
+ photocurrent (IPhoto − IDark) through the transistor channel at ON state (VGS > VTh). This leads
930
+ to an ultrahigh photoresponsivity of the order of ~ 104 A/W at VGS = 40 V. Such gate modulated
931
+ responsivity and sensitivity of MvWH photo-FET devices via interface engineering offers a
932
+ novel pathway for next generation high performance and low power integrated photonic
933
+ technology.
934
+ Temporal photoresponse is also an important parameter for the phototransistors performance
935
+ in terms of switching speed and device stability. The transient photoresponse of the as-
936
+ fabricated CsPbI3/WS2 MvWH photo-FET upon visible illumination (𝜆 = 514 nm) at VGS = 0V
937
+ with varying reverse VDS is demonstrated in Fig. 6(a). Upon illumination of four periodic
938
+ pulses of the Argon-ion laser, relatively fast and consistent photocurrent modulation
939
+ characteristics of the device reveals the stability and reproducibility of the as-fabricated MvWH
940
+ photo-FET. The device exhibits a much stronger photoresponse characteristics revealing ratio
941
+ of ~106 as compared to the control device with pristine WS2 with the value ~103 (Fig. S8 within
942
+ the Supplimental Material), which is attributed to the injection of high density
943
+
944
+
945
+ FIG. 6. (a) Transient response of the MvWH photo-FET device under illumination of a 514 nm
946
+ laser at different applied VDS. (b) Temporal photocurrent response of the MvWH device for a
947
+ wavelength of 514 nm with and without any applied gate bias. The temporal response indicates
948
+ a significant decrease in rise time (from 43.8 to 34 ms) as well as fall time (from 32.7 to 24
949
+ ms), measured at a relatively higher power of 35 µW. (c) Operational stability of the fabricated
950
+ MvWH photo-FET device under visible illumination for more than half an hour. (d) The
951
+ transient photocurrent response of the fabricated transistor over a span of seven days from the
952
+ beginning and end of the stability test. (e) Stability of the device tested under extreme humid
953
+ conditions (varying from 50 to 65% RH). The last four cycle is the response under 65% of
954
+ humidity showing around 5% decay in the photoresponse.
955
+
956
+ photogenerated charge carriers into the WS2 channel from strong light absorbing CsPbI3 NCs.
957
+ With the increment of reverse VDS, a consistent photocurrent enhancement is distinctly noticed
958
+ from the switching characteristics owing to the increase of depletion region width at the
959
+ Schottky barrier interface and subsequent separation of photogenerated charge carriers.
960
+ Further, the rise and fall times of the fabricated device in the absence of gate bias have been
961
+ estimated using an enlarged single cycle response [Fig. 6(b)] and are found to be around ∼
962
+ 43.8 ms and ∼ 32.7 ms, respectively, which are further reduced to 34 ms and 24 ms,
963
+ respectively on applying a gate voltage of 40 V. The response speed of these devices are found
964
+ to be relatively slower, which is attributed to the trapping of charge carriers in various structural
965
+
966
+ (a)
967
+ (c)
968
+ 3
969
+ 21
970
+ -1.5V
971
+ -1V
972
+ 2.5
973
+ 2.0
974
+ HA
975
+ 1.5
976
+ DS
977
+ DS
978
+ 1.0
979
+ 0.5
980
+ ON
981
+ ON
982
+ 0
983
+ OFF
984
+ 0.0
985
+ 0
986
+ 20
987
+ 40
988
+ 60
989
+ 80
990
+ 0
991
+ 20
992
+ 40
993
+ 0
994
+ 602
995
+ 1205
996
+ 1807
997
+ 18901920
998
+ Time (sec)
999
+ Time (sec)
1000
+ Time (sec)
1001
+ Time (sec)
1002
+ (b)
1003
+ (d)
1004
+ (e)
1005
+ 1.2
1006
+ 1.2
1007
+ oV
1008
+ 3
1009
+ VDs = -2V
1010
+ 2=514nm
1011
+ 50% humidity
1012
+ 65% humidity
1013
+ GS
1014
+ Dav 1
1015
+ Day3
1016
+ Day 5
1017
+ Dav 7
1018
+ Normalized
1019
+ 0.8
1020
+ Normalized
1021
+ 0.8
1022
+ (vn)
1023
+ 2
1024
+ 0.4
1025
+ DS
1026
+ 0.4
1027
+ 0.0
1028
+ 0.0
1029
+ 5%
1030
+ decay
1031
+ 6.75
1032
+ 6.80
1033
+ 6.85
1034
+ 6.90
1035
+ 6.95
1036
+ 7.00
1037
+ Time (sec)
1038
+ Time (sec)
1039
+ Time (sec)and surface defect states present in WS2 as well as CsPbI3 NCs and their local junction
1040
+ interfaces. These interface traps present in WS2 layer are mostly empty when biased under
1041
+ depletion condition, i.e. VGS < VTh owing to lack of enough mobile carriers in the channel. This
1042
+ allows a large number of photogenerated electrons to get trapped by the defect states while
1043
+ some of the gate induced electrons, although small in number, can be trapped as well. This
1044
+ results in a relatively slow rise of current, as depicted in the photocurrent dynamic response.
1045
+ On the other hand, the interface traps are nearly filled up with gate-induced electrons in
1046
+ accumulation condition, when VGS > VTh, as shown in Fig. 6(b). Hence, the trapping
1047
+ probability of photogenerated carriers is lower and a relatively faster response (~34 ms) is
1048
+ observed in MvWH photo-FET devices. Further, owing to the fact that the perovskite materials
1049
+ are prone to environmental degradation via oxygen diffusion through iodide vacancies upon
1050
+ illumination, the long term operation stability of the fabricated devices have been tested in this
1051
+ study upon visible light illumination at zero gate bias for prolonged duration (more than 60
1052
+ min). From the I–t curves for the first 100 s [Fig. 6(c), left] and the last 100 s [Fig. 6(c), right],
1053
+ it is observed that the photocurrent has almost no attenuation, indicating that these devices
1054
+ show an excellent light stability under ambient condition, even without the use of a glovebox
1055
+ or encapsulation. The device stability has also been tested via recording the photocurrent under
1056
+ illumination over a period of one week, as illustrated in Fig. 6(d). Here, the phototransistor
1057
+ sustains under laboratory ambient conditions (relative humidity (RH) ~ 45-50%, temperature
1058
+ ~ 22oC) for one week with negligible change in the photocurrent via degradation after storing.
1059
+ Further, as CsPbI3 NCs are vulnerable to environmental humidity, to explore the device
1060
+ performance in the extreme humid condition, we have performed the temporal response under
1061
+ 65% RH showing an insignificant degradation (5% decay) in terms of device response [Fig.
1062
+ 6(e)]. This superior performance stability is due to the surface defect passivation of CsPbI3
1063
+ through the interaction with the sulphur of WS2 ensuring the outstanding environmental
1064
+
1065
+ stability of as-fabricated CsPbI3/WS2 MvWH photo-FETs. The sulfur atoms present on the top
1066
+ layer of WS2 may have stronger coordination to the Pb2+ centers of CsPbI3 NCs leads to reduced
1067
+ defect states in perovskites enabling higher reluctance to the degradation [31]. It may be noted
1068
+ that the performance of the devices could be further improved by process optimization, device
1069
+ encapsulation and incorporation of buffer layers. This work reveals the significant potential of
1070
+ colloidal synthesized air-stable α-CsPbI3 NCs on 2D materials in fabricating 0D/2D mixed-
1071
+ dimensional heterostructure photo-FETs for applications in next generation optoelectronic
1072
+ devices.
1073
+ Conclusion:
1074
+ To summarize, significant improvements in performance have been realized in CsPbI3/WS2
1075
+ 0D/2D mixed-dimensional phototransistors with asymmetric metal electrodes leading to
1076
+ combinatorial effect of Schottky barrier induced suppression of dark current and efficient
1077
+ charge transfer from photoabsorbing CsPbI3 nanocrystals, resulting in enhanced
1078
+ photosensitivity and spectral responsivity. The WS2 channel with asymmetric contacts
1079
+ (Cr/WS2/Au) shows a rectifying I-V characteristics under an applied VDS with the dark current
1080
+ in the order of pA. Further, by combining the channel sensitization via decorating the WS2 with
1081
+ photosensitive air-stable α-phase CsPbI3 NCs, a responsivity of ~102 A/W has been achieved
1082
+ at low VDS (~ -2V) for an incident optical power of 0.1 µW even without any external gate
1083
+ bias. The device exhibits a broad spectral photoresponsivity between 400 and 800 nm due to
1084
+ the extended visible light absorption features of CsPbI3 NCs. Using gate-controlled carrier
1085
+ modulation in the transistor channel, a peak responsivity ~104 A/W (VGS = +40 V) has been
1086
+ achieved owing to the photogating effect mediated charge conduction whereas the maximum
1087
+ sensitivity (~ 106 at ~ VDS = -2 V) in terms of signal-to-noise ratio is observed by depleting the
1088
+ channel carries (VGS = 0 to 5 V). These devices show superior performance in terms of
1089
+ environment stability, owing to the filling of surface trap states present in CsPbI3 NCs via
1090
+
1091
+ conjugation with sulfur atoms of 2D WS2 layer. The fabricated hybrid heterostructure devices
1092
+ combining 2D TMDs and superior light absorbing 0D perovskite nanocrsytals, through proper
1093
+ interface engineering, would open up new pathways for novel optoelectronic functionalities
1094
+ and energy-harvesting applications.
1095
+ Acknowledgement:
1096
+ SKR acknowledges the support of Chair Professor Fellowship of the Indian National Academy
1097
+ of Engineering (INAE).
1098
+ Conflicts of interest
1099
+ There are no conflicts to declare.
1100
+ References:
1101
+ [1]
1102
+ K. S. Novoselov, D. Jiang, F. Schedin, T. J. Booth, V. V Khotkevich, S. V Morozov,
1103
+ and A. K. Geim, Two-Dimensional Atomic Crystals, Proc. Natl. Acad. Sci. 102, 10451
1104
+ (2005).
1105
+ [2]
1106
+ J. Pu, Y. Yomogida, K.-K. Liu, L.-J. Li, Y. Iwasa, and T. Takenobu, Highly Flexible
1107
+ MoS 2 Thin-Film Transistors with Ion Gel Dielectrics, Nano Lett. 12, 4013 (2012).
1108
+ [3]
1109
+ B. Radisavljevic, A. Radenovic, J. Brivio, V. Giacometti, and A. Kis, Single-Layer
1110
+ MoS2 Transistors, Nat. Nanotechnol. 6, 147 (2011).
1111
+ [4]
1112
+ K. F. Mak, C. Lee, J. Hone, J. Shan, and T. F. Heinz, Atomically Thin MoS2 : A New
1113
+ Direct-Gap Semiconductor, Phys. Rev. Lett. 105, 136805 (2010).
1114
+ [5]
1115
+ D. Kufer and G. Konstantatos, Photo-FETs: Phototransistors Enabled by 2D and 0D
1116
+ Nanomaterials, ACS Photonics 3, 2197 (2016).
1117
+ [6]
1118
+ N. Huo, S. Gupta, and G. Konstantatos, MoS 2 -HgTe Quantum Dot Hybrid
1119
+ Photodetectors beyond 2 Μm, Adv. Mater. 29, 1606576 (2017).
1120
+ [7]
1121
+ G. Konstantatos, M. Badioli, L. Gaudreau, J. Osmond, M. Bernechea, F. P. G. de
1122
+ Arquer, F. Gatti, and F. H. L. Koppens, Hybrid Graphene–Quantum Dot
1123
+ Phototransistors with Ultrahigh Gain, Nat. Nanotechnol. 7, 363 (2012).
1124
+ [8]
1125
+ W. Choi, M. Y. Cho, A. Konar, J. H. Lee, G.-B. Cha, S. C. Hong, S. Kim, J. Kim, D.
1126
+ Jena, J. Joo, and S. Kim, High-Detectivity Multilayer MoS 2 Phototransistors with
1127
+ Spectral Response from Ultraviolet to Infrared, Adv. Mater. 24, 5832 (2012).
1128
+ [9]
1129
+ J. Kwon, Y. K. Hong, G. Han, I. Omkaram, W. Choi, S. Kim, and Y. Yoon, Giant
1130
+ Photoamplification in Indirect-Bandgap Multilayer MoS 2 Phototransistors with Local
1131
+
1132
+ Bottom-Gate Structures, Adv. Mater. 27, 2224 (2015).
1133
+ [10] N. Perea-López, A. L. Elías, A. Berkdemir, A. Castro-Beltran, H. R. Gutiérrez, S.
1134
+ Feng, R. Lv, T. Hayashi, F. López-Urías, S. Ghosh, B. Muchharla, S. Talapatra, H.
1135
+ Terrones, and M. Terrones, Photosensor Device Based on Few-Layered WS 2 Films,
1136
+ Adv. Funct. Mater. 23, 5511 (2013).
1137
+ [11] K. Natori, Ballistic Metal‐oxide‐semiconductor Field Effect Transistor, J. Appl. Phys.
1138
+ 76, 4879 (1994).
1139
+ [12] D. Li, X. Sun, H. Song, Z. Li, H. Jiang, Y. Chen, G. Miao, and B. Shen, Effect of
1140
+ Asymmetric Schottky Barrier on GaN-Based Metal-Semiconductor-Metal Ultraviolet
1141
+ Detector, Appl. Phys. Lett. 99, 261102 (2011).
1142
+ [13] J. Kim, A. Venkatesan, N. A. N. Phan, Y. Kim, H. Kim, D. Whang, and G. Kim,
1143
+ Schottky Diode with Asymmetric Metal Contacts on WS 2, Adv. Electron. Mater. 8,
1144
+ 2100941 (2022).
1145
+ [14] M. Fontana, T. Deppe, A. K. Boyd, M. Rinzan, A. Y. Liu, M. Paranjape, and P.
1146
+ Barbara, Electron-Hole Transport and Photovoltaic Effect in Gated MoS2 Schottky
1147
+ Junctions, Sci. Rep. 3, 1634 (2013).
1148
+ [15] J.-Y. Wu, Y. T. Chun, S. Li, T. Zhang, and D. Chu, Electrical Rectifying and
1149
+ Photosensing Property of Schottky Diode Based on MoS 2, ACS Appl. Mater.
1150
+ Interfaces 10, 24613 (2018).
1151
+ [16] S. Zhang, X. Wang, Y. Chen, G. Wu, Y. Tang, L. Zhu, H. Wang, W. Jiang, L. Sun, T.
1152
+ Lin, H. Shen, W. Hu, J. Ge, J. Wang, X. Meng, and J. Chu, Ultrasensitive Hybrid MoS
1153
+ 2 –ZnCdSe Quantum Dot Photodetectors with High Gain, ACS Appl. Mater. Interfaces
1154
+ 11, 23667 (2019).
1155
+ [17] D.-H. Kwak, P. Ramasamy, Y.-S. Lee, M.-H. Jeong, and J.-S. Lee, High-Performance
1156
+ Hybrid InP QDs/Black Phosphorus Photodetector, ACS Appl. Mater. Interfaces 11,
1157
+ 29041 (2019).
1158
+ [18] Y. Lee, J. Kwon, E. Hwang, C.-H. Ra, W. J. Yoo, J.-H. Ahn, J. H. Park, and J. H. Cho,
1159
+ High-Performance Perovskite-Graphene Hybrid Photodetector, Adv. Mater. 27, 41
1160
+ (2015).
1161
+ [19] Z. Sun, Z. Liu, J. Li, G. Tai, S.-P. Lau, and F. Yan, Infrared Photodetectors Based on
1162
+ CVD-Grown Graphene and PbS Quantum Dots with Ultrahigh Responsivity, Adv.
1163
+ Mater. 24, 5878 (2012).
1164
+ [20] S. Pal, A. Ghorai, D. K. Goswami, and S. K. Ray, Strain Mediated Light Emission
1165
+ Using Heterojunctions of All-Inorganic Mixed-Halide Perovskite Nanocrystals via
1166
+ Piezo-Phototronic Effect, Nano Energy 87, 106200 (2021).
1167
+ [21] F. Liu, Y. Zhang, C. Ding, S. Kobayashi, T. Izuishi, N. Nakazawa, T. Toyoda, T. Ohta,
1168
+ S. Hayase, T. Minemoto, K. Yoshino, S. Dai, and Q. Shen, Highly Luminescent Phase-
1169
+ Stable CsPbI3 Perovskite Quantum Dots Achieving Near 100% Absolute
1170
+ Photoluminescence Quantum Yield, ACS Nano 11, 10373 (2017).
1171
+ [22] A. Ghorai, S. Mahato, S. K. Srivastava, and S. K. Ray, Atomic Insights of Stable,
1172
+ Monodispersed CsPbI 3− x Br x ( x = 0, 1, 2, 3) Nanocrystals Synthesized by Modified
1173
+ Ligand Cell, Adv. Funct. Mater. 32, 2202087 (2022).
1174
+
1175
+ [23] A. Ghorai, A. Midya, and S. K. Ray, Surfactant-Induced Anion Exchange and
1176
+ Morphological Evolution for Composition-Controlled Caesium Lead Halide
1177
+ Perovskites with Tunable Optical Properties, ACS Omega 4, 12948 (2019).
1178
+ [24] M. A. Becker, R. Vaxenburg, G. Nedelcu, P. C. Sercel, A. Shabaev, M. J. Mehl, J. G.
1179
+ Michopoulos, S. G. Lambrakos, N. Bernstein, J. L. Lyons, T. Stöferle, R. F. Mahrt, M.
1180
+ V. Kovalenko, D. J. Norris, G. Rainò, and A. L. Efros, Bright Triplet Excitons in
1181
+ Caesium Lead Halide Perovskites, Nature 553, 189 (2018).
1182
+ [25] Q. A. Akkerman, G. Rainò, M. V. Kovalenko, and L. Manna, Genesis, Challenges and
1183
+ Opportunities for Colloidal Lead Halide Perovskite Nanocrystals, Nat. Mater. 17, 394
1184
+ (2018).
1185
+ [26] K. M. Sim, A. Swarnkar, A. Nag, and D. S. Chung, Phase Stabilized α-CsPbI3
1186
+ Perovskite Nanocrystals for Photodiode Applications, Laser Photonics Rev. 12, 1
1187
+ (2018).
1188
+ [27] C. de Weerd, L. Gomez, A. Capretti, D. M. Lebrun, E. Matsubara, J. Lin, M. Ashida,
1189
+ F. C. M. Spoor, L. D. A. Siebbeles, A. J. Houtepen, K. Suenaga, Y. Fujiwara, and T.
1190
+ Gregorkiewicz, Efficient Carrier Multiplication in CsPbI3 Perovskite Nanocrystals,
1191
+ Nat. Commun. 9, 4199 (2018).
1192
+ [28] T. Wu, Y. Wang, Z. Dai, D. Cui, T. Wang, X. Meng, E. Bi, X. Yang, and L. Han,
1193
+ Efficient and Stable CsPbI 3 Solar Cells via Regulating Lattice Distortion with Surface
1194
+ Organic Terminal Groups, Adv. Mater. 31, 1900605 (2019).
1195
+ [29] J.-K. Sun, S. Huang, X.-Z. Liu, Q. Xu, Q.-H. Zhang, W.-J. Jiang, D.-J. Xue, J.-C. Xu,
1196
+ J.-Y. Ma, J. Ding, Q.-Q. Ge, L. Gu, X.-H. Fang, H.-Z. Zhong, J.-S. Hu, and L.-J. Wan,
1197
+ Polar Solvent Induced Lattice Distortion of Cubic CsPbI 3 Nanocubes and
1198
+ Hierarchical Self-Assembly into Orthorhombic Single-Crystalline Nanowires, J. Am.
1199
+ Chem. Soc. 140, 11705 (2018).
1200
+ [30] S. Mahato, A. Ghorai, A. Mondal, S. K. Srivastava, M. Modak, S. Das, and S. K. Ray,
1201
+ Atomic-Scale Imaging and Nano-Scale Mapping of Cubic α-CsPbI 3 Perovskite
1202
+ Nanocrystals for Inverted Perovskite Solar Cells, ACS Appl. Mater. Interfaces 14,
1203
+ 9711 (2022).
1204
+ [31] J. C. Hamill, O. Romiluyi, S. A. Thomas, J. Cetola, J. Schwartz, M. F. Toney, P.
1205
+ Clancy, and Y.-L. Loo, Sulfur-Donor Solvents Strongly Coordinate Pb 2+ in Hybrid
1206
+ Organic–Inorganic Perovskite Precursor Solutions, J. Phys. Chem. C 124, 14496
1207
+ (2020).
1208
+ [32] A. Berkdemir, H. R. Gutiérrez, A. R. Botello-Méndez, N. Perea-López, A. L. Elías, C.-
1209
+ I. Chia, B. Wang, V. H. Crespi, F. López-Urías, J. Charlier, H. Terrones, and M.
1210
+ Terrones, Identification of Individual and Few Layers of WS2 Using Raman
1211
+ Spectroscopy, Sci. Rep. 3, 1755 (2013).
1212
+ [33] B. Chakraborty, A. Bera, D. V. S. Muthu, S. Bhowmick, U. V. Waghmare, and A. K.
1213
+ Sood, Symmetry-Dependent Phonon Renormalization in Monolayer MoS 2 Transistor,
1214
+ Phys. Rev. B 85, 161403(R) (2012).
1215
+ [34] S. Das, R. K. Chowdhury, D. Karmakar, S. Das, and S. K. Ray, Substrate-Dependent
1216
+ Synergistic Many-Body Effects in Atomically Thin Two-Dimensional W S2, Phys. Rev.
1217
+ Mater. 5, 124001 (2021).
1218
+
1219
+ [35] N. Peimyoo, W. Yang, J. Shang, X. Shen, Y. Wang, and T. Yu, Chemically Driven
1220
+ Tunable Light Emission of Charged and Neutral Excitons in Monolayer WS2, ACS
1221
+ Nano 8, 11320 (2014).
1222
+ [36] B. Liu, W. Zhao, Z. Ding, I. Verzhbitskiy, L. Li, J. Lu, J. Chen, G. Eda, and K. P. Loh,
1223
+ Engineering Bandgaps of Monolayer MoS 2 and WS 2 on Fluoropolymer Substrates by
1224
+ Electrostatically Tuned Many-Body Effects, 6457 (2016).
1225
+ [37] J. Qiao, F. Feng, S. Song, T. Wang, M. Shen, G. Zhang, X. Yuan, and M. G. Somekh,
1226
+ Perovskite Quantum Dot‐Ta 2 NiSe 5 Mixed‐Dimensional Van Der Waals
1227
+ Heterostructures for High‐Performance Near‐Infrared Photodetection, Adv. Funct.
1228
+ Mater. 32, 2110706 (2022).
1229
+ [38] M. Hussain, S. Aftab, S. H. A. Jaffery, A. Ali, S. Hussain, D. N. Cong, R. Akhtar, Y.
1230
+ Seo, J. Eom, P. Gautam, H. Noh, and J. Jung, Asymmetric Electrode Incorporated 2D
1231
+ GeSe for Self-Biased and Efficient Photodetection, Sci. Rep. 10, 1 (2020).
1232
+ [39] H. Wu, H. Si, Z. Zhang, Z. Kang, P. Wu, L. Zhou, S. Zhang, Z. Zhang, Q. Liao, and Y.
1233
+ Zhang, All-Inorganic Perovskite Quantum Dot-Monolayer MoS 2 Mixed-Dimensional
1234
+ van Der Waals Heterostructure for Ultrasensitive Photodetector, Adv. Sci. 5, 1801219
1235
+ (2018).
1236
+ [40] S. Cao, Y. Xing, J. Han, X. Luo, W. Lv, W. Lv, B. Zhang, and Z. Zeng, Ultrahigh-
1237
+ Photoresponsive UV Photodetector Based on a BP/ReS 2 Heterostructure p–n Diode,
1238
+ Nanoscale 10, 16805 (2018).
1239
+ [41] Y. Yu, Y. Zhang, X. Song, H. Zhang, M. Cao, Y. Che, H. Dai, J. Yang, H. Zhang, and
1240
+ J. Yao, PbS-Decorated WS 2 Phototransistors with Fast Response, ACS Photonics 4,
1241
+ 950 (2017).
1242
+ [42] C. Hu, D. Dong, X. Yang, K. Qiao, D. Yang, H. Deng, S. Yuan, J. Khan, Y. Lan, H.
1243
+ Song, and J. Tang, Synergistic Effect of Hybrid PbS Quantum Dots/2D-WSe 2 Toward
1244
+ High Performance and Broadband Phototransistors, Adv. Funct. Mater. 27, 1603605
1245
+ (2017).
1246
+ [43] A.-Y. Lee, H.-S. Ra, D.-H. Kwak, M.-H. Jeong, J.-H. Park, Y.-S. Kang, W.-S. Chae,
1247
+ and J.-S. Lee, Hybrid Black Phosphorus/Zero-Dimensional Quantum Dot
1248
+ Phototransistors: Tunable Photodoping and Enhanced Photoresponsivity, ACS Appl.
1249
+ Mater. Interfaces 10, 16033 (2018).
1250
+ [44] C. Ma, Y. Shi, W. Hu, M. Chiu, Z. Liu, A. Bera, F. Li, H. Wang, L. Li, and T. Wu,
1251
+ Heterostructured WS 2 /CH 3 NH 3 PbI 3 Photoconductors with Suppressed Dark
1252
+ Current and Enhanced Photodetectivity, Adv. Mater. 28, 3683 (2016).
1253
+ [45] X. Song, X. Liu, D. Yu, C. Huo, J. Ji, X. Li, S. Zhang, Y. Zou, G. Zhu, Y. Wang, M.
1254
+ Wu, A. Xie, and H. Zeng, Boosting Two-Dimensional MoS 2 /CsPbBr 3
1255
+ Photodetectors via Enhanced Light Absorbance and Interfacial Carrier Separation,
1256
+ ACS Appl. Mater. Interfaces 10, 2801 (2018).
1257
+ [46] H. Wu, Z. Kang, Z. Zhang, Z. Zhang, H. Si, Q. Liao, S. Zhang, J. Wu, X. Zhang, and
1258
+ Y. Zhang, Interfacial Charge Behavior Modulation in Perovskite Quantum Dot-
1259
+ Monolayer MoS 2 0D-2D Mixed-Dimensional van Der Waals Heterostructures, Adv.
1260
+ Funct. Mater. 28, 1802015 (2018).
1261
+ [47] R. Lin, X. Li, W. Zheng, and F. Huang, Balanced Photodetection in Mixed-
1262
+
1263
+ Dimensional Phototransistors Consisting of CsPbBr3 Quantum Dots and Few-Layer
1264
+ MoS2, ACS Appl. Nano Mater. 2, 2599 (2019).
1265
+ [48] K. Chen, Y. Wang, J. Liu, J. Kang, Y. Ge, W. Huang, Z. Lin, Z. Guo, Y. Zhang, and
1266
+ H. Zhang, In Situ Preparation of a CsPbBr 3 /Black Phosphorus Heterostructure with
1267
+ an Optimized Interface and Photodetector Application, Nanoscale 11, 16852 (2019).
1268
+ [49] H. Wu, Z. Kang, Z. Zhang, H. Si, S. Zhang, Z. Zhang, Q. Liao, and Y. Zhang, Ligand
1269
+ Engineering for Improved All‐Inorganic Perovskite Quantum Dot‐MoS 2 Monolayer
1270
+ Mixed Dimensional van Der Waals Phototransistor, Small Methods 3, 1900117
1271
+ (2019).
1272
+ [50] X. Zou, Y. Li, G. Tang, P. You, and F. Yan, Schottky Barrier‐Controlled Black
1273
+ Phosphorus/Perovskite Phototransistors with Ultrahigh Sensitivity and Fast Response,
1274
+ Small 15, 1901004 (2019).
1275
+ [51] S. Lee, J. Y. Kim, S. Choi, Y. Lee, K.-S. Lee, J. Kim, and J. Joo, Photosensitive N-
1276
+ Type Doping Using Perovskite CsPbX 3 Quantum Dots for Two-Dimensional MSe 2
1277
+ (M = Mo and W) Field-Effect Transistors, ACS Appl. Mater. Interfaces 12, 25159
1278
+ (2020).
1279
+ [52] S. Aftab, M. W. Iqbal, A. M. Afzal, M. F. Khan, G. Hussain, H. S. Waheed, and M. A.
1280
+ Kamran, Formation of an MoTe 2 Based Schottky Junction Employing Ultra-Low and
1281
+ High Resistive Metal Contacts, RSC Adv. 9, 10017 (2019).
1282
+ [53] M. H. Yang, K. B. K. Teo, W. I. Milne, and D. G. Hasko, Carbon Nanotube Schottky
1283
+ Diode and Directionally Dependent Field-Effect Transistor Using Asymmetrical
1284
+ Contacts, Appl. Phys. Lett. 87, 253116 (2005).
1285
+
1286
+
1287
+
1288
+
1289
+ Supplemental Material
1290
+
1291
+ Charge transfer mediated giant photo-amplification in air-stable α-CsPbI3
1292
+ nanocrystals decorated 2D-WS2 photo-FET with asymmetric contacts
1293
+ Shreyasi Das1, Arup Ghorai1,2, Sourabh Pal3, Somnath Mahato1, Soumen Das4, Samit K. Ray5 *
1294
+ 1School of Nano Science and Technology, IIT Kharagpur, Kharagpur, West Bengal, India, 721302
1295
+ 2Department of Materials Science and Engineering, Pohang University of Science and Technology,
1296
+ Pohang 790-784, Korea
1297
+ 3Advanced Technology Development Centre, IIT Kharagpur, Kharagpur, West Bengal, India, 721302
1298
+ 4School of Medical Science and Technology, IIT Kharagpur, Kharagpur, West Bengal, India, 721302
1299
+ 5Department of Physics, IIT Kharagpur, Kharagpur, West Bengal, India, 721302
1300
+ Email : physkr@phy.iitkgp.ac.in
1301
+
1302
+
1303
+
1304
+
1305
+ S1: Tauc plot of CsPbI3 NCs
1306
+
1307
+ Fig. S1. Tauc plot of as synthesised α-phase CsPbI3 NCs
1308
+
1309
+ S2: Photoluminescence spectra of WS2, CsPbI3 and CsPbI3/WS2 hybrid
1310
+
1311
+ Fig. S2. Comparative PL spectrum of bare ML WS2 flake, CsPbI3 NCs and their
1312
+ heterostructures where three different concentrations of CsPbI3 NCs (different steps of spin
1313
+ coating) incorporated on WS2 flakes. After formation of heterostructures, the PL intensity of
1314
+ both bare ML WS2 as well as CsPbI3 NCs get reduced.
1315
+
1316
+ (αhv)? (a.u.)
1317
+ E. = 1.814 eV
1318
+ 1.6
1319
+ 1.8
1320
+ 2.0
1321
+ 2.2
1322
+ 2.4
1323
+ 2.6
1324
+ Energy (eV)PL Intensity (a.u.)
1325
+ Ws?
1326
+ CsPbI3
1327
+ Step 1
1328
+ Step 2
1329
+ Step 3
1330
+ 550
1331
+ 600
1332
+ 650
1333
+ 700
1334
+ 750
1335
+ Wavelength (nm)S3: PL integrated intensity ratio vs coating step
1336
+
1337
+ Fig. S3. Variation of PL integrated intensity ratio of WS2 trion peak (A-) to excitonic peak
1338
+ (A) with increasing concentration of CsPbI3 decoration (increasing spin coating step) on WS2
1339
+ flakes.
1340
+
1341
+ S4: Thickness of the exfoliated flake analysis using AFM
1342
+
1343
+ Fig. S4. Atomic force microscopy image of the few layer WS2 flakes. Inset shows the height
1344
+ profile along the yellow dashed line confirming the thickness of the flakes around 4.5 nm.
1345
+
1346
+ 7
1347
+ Increasing CsPbI,
1348
+ 6
1349
+ concentration
1350
+ 5
1351
+ 2
1352
+ 1
1353
+ 0
1354
+ Step 3
1355
+ Step 2
1356
+ Step 1
1357
+ Ws.0
1358
+ 5
1359
+ 10
1360
+ 15
1361
+ 20
1362
+ 25
1363
+ 30
1364
+ 35
1365
+ 40
1366
+ 45μm
1367
+ nm
1368
+ 30
1369
+ Height (nm)
1370
+ 5
1371
+ 27.5
1372
+ 25
1373
+ 10-
1374
+ 22.5
1375
+ 15
1376
+ -20
1377
+ 20
1378
+ 0
1379
+ 17.5
1380
+ 25
1381
+ 0.0
1382
+ 0.4
1383
+ 0.8
1384
+ 1.2
1385
+ -15
1386
+ Distance (um)
1387
+ 30-
1388
+ 12.5
1389
+ -10
1390
+ 35
1391
+ 7.5
1392
+ 40
1393
+ -5
1394
+ 45
1395
+ 2.5
1396
+ 50
1397
+ umS5: Energy band structures of WS2 at the contacts
1398
+
1399
+ Fig. S5. The corresponding energy band structures with different combination of metal
1400
+ electrodes before contact and after contact condition under applied reverse bias.
1401
+
1402
+ S6: Photo to dark current ratio
1403
+
1404
+ Fig. S6. Photo to dark current ratio for control device (WS2 FET) and MvWH photo-FET
1405
+ with varying drain to source voltage.
1406
+ Cr
1407
+ Cr
1408
+ WS2
1409
+ Ohmic contact
1410
+ +ve
1411
+ -ve
1412
+ Au
1413
+ Au
1414
+ WS2
1415
+ Symmetric contact
1416
+ +ve
1417
+ -ve
1418
+ Au
1419
+ -ve
1420
+ Cr
1421
+ +ve
1422
+ WS2
1423
+ Asymmetric contact
1424
+ Schottky diode
1425
+ Au
1426
+ Cr
1427
+ WS2
1428
+ 4.6 eV
1429
+ 4.5 eV
1430
+ 5.1 eV
1431
+ Evac
1432
+ EF
1433
+ Before contact
1434
+
1435
+ 1.2x10
1436
+ Ws
1437
+ 9.0x1(
1438
+ sPbI.
1439
+ rk
1440
+ Dal
1441
+ 6.0x10
1442
+ 3.0x10
1443
+ 0.0
1444
+ 2.0
1445
+ -1.5
1446
+ -1.0
1447
+ -0.5
1448
+ 0.0
1449
+ Vps (V)S7: Spectral responsivity of the control WS2 FET device
1450
+
1451
+ Fig. S7. Spectral responsivity curves of control WS2 FET device at VGS = 0V with increasing
1452
+ reverse VDS from 0V to -2V
1453
+
1454
+ S8: Transient response of the control WS2 FET device
1455
+
1456
+
1457
+ Fig. S8. Transient response of the control WS2 FET device under illumination of a 514 nm
1458
+ laser at different applied VDS.
1459
+
1460
+ 300 400 500 600 700 800 900
1461
+ 0
1462
+ 3
1463
+ 5
1464
+ 8
1465
+ 10
1466
+ Responsivity (A/W)
1467
+ WS2
1468
+ Wavelength (nm)
1469
+ -2V
1470
+ -1V
1471
+ 0V
1472
+
1473
+ 8
1474
+ -1.5V
1475
+ -1V
1476
+ 40
1477
+ 6
1478
+ (nA)
1479
+ 4
1480
+ 2
1481
+ ON
1482
+ OFF ON
1483
+ 0
1484
+ 0
1485
+ 20
1486
+ 40
1487
+ 60
1488
+ 80
1489
+ Time (sec)
HtE1T4oBgHgl3EQfrgVp/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
I9AzT4oBgHgl3EQfH_tx/vector_store/index.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2d7eb3c625c4ef6ff9ebec8701c21fe01a6afac6f49df8e4cd400cf07c550834
3
+ size 192025
JdFIT4oBgHgl3EQfZSuO/content/2301.11252v1.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:07352ed61f9bcbe177049399fde20f721566559ce4efa730dcf761650198e02f
3
+ size 356424
JdFIT4oBgHgl3EQfZSuO/vector_store/index.faiss ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:54c6defce0f1c8bd7a59a30bc3bfa49dcd1e1a398116243f1462f980d699764b
3
+ size 1703981
JdFIT4oBgHgl3EQfZSuO/vector_store/index.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e87bd6fb92435a83169ff399bdabb5399107821bff7213643b468b8a0a53654
3
+ size 69172
KNE2T4oBgHgl3EQfpghg/content/tmp_files/2301.04029v1.pdf.txt ADDED
The diff for this file is too large to render. See raw diff
 
KNE2T4oBgHgl3EQfpghg/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
KdFQT4oBgHgl3EQfTTZa/content/tmp_files/2301.13292v1.pdf.txt ADDED
@@ -0,0 +1,632 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ On the structure of entropy solutions to the Riemann
2
+ problem for a degenerate nonlinear parabolic equation
3
+ Evgeny Yu. Panov
4
+ Yaroslav-the-Wise Novgorod State University, Veliky Novgorod, Russian Federation,
5
+ Research and Development Center, Veliky Novgorod, Russian Federation.
6
+ Abstract
7
+ We find an explicit form of entropy solutions to a Riemann problem for a de-
8
+ generate nonlinear parabolic equation with piecewise constant velocity and diffusion
9
+ coefficients. It is demonstrated that this solution corresponds to the minimum point
10
+ of some strictly convex function of a finite number of variables.
11
+ 1
12
+ Introduction
13
+ In a half-plane Π = {(t, x) | t > 0, x ∈ R}, we consider a nonlinear parabolic equation
14
+ ut + v(u)ux − t(a2(u)ux)x = 0,
15
+ (1)
16
+ where v(u), a(u) ∈ L∞(R), a(u) ≥ 0. Since the diffusion coefficient a(u) may take zero
17
+ value, equation (1) is degenerate. In the case when a(u) ≡ 0 it reduces to a first order
18
+ conservation law
19
+ ut + ϕ(u)x = 0,
20
+ (2)
21
+ where ϕ′(u) = v(u). Similarly, a general equation (1) can be written in the conservative
22
+ form
23
+ ut + ϕ(u)x − tA(u)xx = 0
24
+ with A′(u) = a2(u), which allows to define weak solutions of this equation. Unfortunately,
25
+ weak solutions to a Cauchy problem for equation (1) are not unique in general, and some
26
+ additional entropy conditions are required. We consider the Cauchy problem with initial
27
+ data
28
+ u(0, x) = u0(x),
29
+ (3)
30
+ where u0(x) ∈ L∞(R). Recall the notion of entropy solution (e.s. for short) in the sense of
31
+ Carrillo [1].
32
+ Definition 1. A function u = u(t, x) ∈ L∞(Π) is called an e.s. of (1), (3) if
33
+ (i) the distribution A(u) ∈ L2
34
+ loc(Π);
35
+ (ii) for all k ∈ R
36
+ |u − k|t + (sign(u − k)(ϕ(u) − ϕ(k)))x − (t sign(u − k)(A(u) − A(k)))xx ≤ 0
37
+ (4)
38
+ in the sense of distributions (in D′(Π));
39
+ (iii) ess lim
40
+ t→0
41
+ u(t, ·) = u0 in L1
42
+ loc(R).
43
+ 1
44
+ arXiv:2301.13292v1 [math.AP] 30 Jan 2023
45
+
46
+ Entropy condition (4) means that for each nonnegative test function f = f(t, x) ∈
47
+ C2
48
+ 0(Π)
49
+
50
+ Π
51
+ [|u − k|ft + sign(u − k)((ϕ(u) − ϕ(k))fx + t(A(u) − A(k))fxx)]dtdx ≥ 0.
52
+ (5)
53
+ In the case of conservation laws (2) the notion of e.s. reduces to the notion of generalized
54
+ e.s. in the sense of Kruzhkov [2]. Taking in (4) k = ±M, M ≥ ∥u∥∞, we derive that
55
+ ut + ϕ(u)x − tA(u)xx = 0 in D′(Π),
56
+ that is, an e.s. u of (1), (3) is a weak solution of this problem. It is known that an e.s. of (1),
57
+ (3) always exists and is unique. In general multidimensional setting this was demonstrated
58
+ in [2] for conservation laws and in [1] for the general case. If to be precise, in [1] the case of
59
+ usual diffusion term A(u)xx was studied but the proofs can be readily adapted to the case
60
+ of the self-similar diffusion tA(u)xx.
61
+ If u = u(t, x) is a piecewise C2-smooth e.s. of equation (1) then it must satisfy this
62
+ equation in classic sense in each smoothness domain.
63
+ Applying relation (1) to a test
64
+ function f = f(t, x) ∈ C2
65
+ 0(Π) supported in a neighborhood of a discontinuity line x = x(t)
66
+ and integrating by parts, we then obtain the identity
67
+ (−x′(t)[u] + [ϕ(u)] − t[A(u)x])f + t[A(u)]fx = 0
68
+ (6)
69
+ a.e. on the line x = x(t). Here we denote by [w] the jump of a function w = w(t, x) on the
70
+ line x = x(t) so that
71
+ [w] = w(t, x(t)+) − w(t, x(t)−),
72
+ where w(t, x(t)±) =
73
+ lim
74
+ y→x(t)± w(t, y).
75
+ Since the functions f, fx are arbitrary and independent on the line x = x(t), identity (6)
76
+ implies the following two relations of Rankine-Hugoniot type
77
+ [A(u)] = 0,
78
+ (7)
79
+ −x′(t)[u] + [ϕ(u)] − t[A(u)x] = 0.
80
+ (8)
81
+ Similarly, it follows from entropy relation (5), after integration by parts, that
82
+ (−x′(t)[|u − k|] + [sign(u − k)(ϕ(u) − ϕ(k))] − t[sign(u − k)A(u)x])f+
83
+ t[sign(u − k)(A(u) − A(k))]fx ≤ 0.
84
+ (9)
85
+ Since the function A(u) increases, it follows from (7) that A(u) = const when u lies between
86
+ the values u(t, x(t)−) and u(t, x(t)+). This implies that [sign(u − k)(A(u) − A(k))] = 0
87
+ and in view of arbitrariness of f ≥ 0 it follows from (9) that
88
+ − x′(t)[|u − k|] + [sign(u − k)(ϕ(u) − ϕ(k))] − t[sign(u − k)A(u)x] ≤ 0.
89
+ (10)
90
+ In the case when k lies out of the interval with the endpoints u± .= u(t, x(t)±) relation (10)
91
+ follows from (8) and fulfils with equality sign. When u− < k < u+ this relation reads
92
+ −x′(t)(u+ + u− − 2k) + ϕ(u+) + ϕ(u−) − 2ϕ(k) − t(A(u)+
93
+ x + A(u)−
94
+ x ) ≤ 0,
95
+ 2
96
+
97
+ where A(u)±
98
+ x = A(u)x(t, x(t)±). Adding (8) to this relation and dividing the result by 2,
99
+ we arrive at the following analogue of the famous Oleinik condition (see [3]) known for
100
+ conservation laws.
101
+ − x′(t)(u+ − k) + ϕ(u+) − tA(u)+
102
+ x − ϕ(k) ≤ 0
103
+ ∀k ∈ [u−, u+].
104
+ (11)
105
+ In the case u+ < u− this condition has the form
106
+ − x′(t)(u+ − k) + ϕ(u+) − tA(u)+
107
+ x − ϕ(k) ≥ 0
108
+ ∀k ∈ [u+, u−]
109
+ (12)
110
+ and can be derived similarly.
111
+ Geometric interpretation of these conditions is that the
112
+ graph of the flux function ϕ(u) lies not below (not above) of the segment connecting the
113
+ points (u−, ϕ(u−) − tA(u)−
114
+ x ), (u+, ϕ(u+) − tA(u)+
115
+ x ) when u− ≤ u ≤ u+ (respectively, when
116
+ u+ ≤ u ≤ u−), see Figure 1. We take here into account that in view of condition (8)
117
+ the vector (−x′(t), 1) is a normal to the indicated segment. We also notice that it follows
118
+ from relations (11), (12) with k = u± and from the Rankine-Hugoniot condition (8) that
119
+ A(u)±
120
+ x ≥ 0 (A(u)±
121
+ x ≤ 0) whenever u+ > u− (u+ < u−).
122
+ u
123
+ u-
124
+ u+
125
+ y=φ(u)
126
+ y
127
+ tA(u)x
128
+ -
129
+ tA(u)x
130
+ +
131
+ (-x',1)
132
+ k
133
+ Figure 1: Oleinik condition.
134
+ 2
135
+ The case of piecewise constant coefficients.
136
+ Below we will assume that the functions v(u), a(u) are piecewise constant, v(u) = vk,
137
+ a(u) = ak when uk < u < uk+1, k = 0, . . . , n − 1, where
138
+ α = u0 < u1 < · · · < un−1 < un = β.
139
+ We will study problem (1), (3) with the Riemann data u0(x) =
140
+ � α,
141
+ x < 0,
142
+ β,
143
+ x > 0.
144
+ Since this
145
+ problem is invariant under the scaling transformations t → λt, x → λx, λ > 0 then, by
146
+ the uniqueness, the e.s. u = u(t, x) is self-similar: u(t, x) = u(λt, λx). This implies that
147
+ u = u(x/t). Suppose that ak > 0. Then in a domain where uk < u(ξ) < uk+1 with ξ = x/t
148
+ equation (1) reduces to the second order ODE
149
+ (vk − ξ)u′ − a2
150
+ ku′′ = 0,
151
+ 3
152
+
153
+ the general solution of which is u = C1F((ξ − vk)/ak) + C2; C1, C2 = const, where
154
+ F(z) =
155
+ 1
156
+
157
+
158
+ � z
159
+ −∞
160
+ e−s2/2ds
161
+ is the error function. Therefore, it is natural to seek the e.s. of our problem in the following
162
+ form
163
+ u(ξ) =
164
+ � uk +
165
+ uk+1−uk
166
+ F((ξk+1−vk)/ak)−F((ξk−vk)/ak)(F((ξ − vk)/ak) − F((ξk − vk)/ak))
167
+ ,
168
+ ak > 0,
169
+ uk
170
+ ,
171
+ ak = 0,
172
+ (13)
173
+ ξk < ξ < ξk+1, k = 0, . . . , d,
174
+ where
175
+ d =
176
+ � n − 1
177
+ ,
178
+ an−1 > 0,
179
+ n
180
+ ,
181
+ an−1 = 0,
182
+ − ∞ = ξ0 < ξ1 ≤ · · · ≤ ξd < ξd+1 = +∞,
183
+ and we agree that an = 0, F(−∞) = 0, F(+∞) = 1. We also assume that ξk+1 > ξk
184
+ whenever ak > 0. The rays x = ξkt for finite ξk are (weak or strong) discontinuity lines of
185
+ u, they correspond to discontinuity points ξk of the function u(ξ). Observe that conditions
186
+ (7), (8) turns into the following relations at points ξk
187
+ [A(u)] = A(u(ξk+)) − A(u(ξk−)) = 0,
188
+ (14)
189
+ −ξk[u] + [ϕ(u)] − [A(u)′] = −ξk(u(ξk+) − u(ξk−)) + ϕ(u(ξk+))−
190
+ ϕ(u(ξk−)) − A(u)′(ξk+) + A(u)′(ξk−) = 0.
191
+ (15)
192
+ Here w(ξk±) denotes unilateral limits of a function w(ξ) at the point ξk. Similarly, the
193
+ Oleinik condition (11) reads
194
+ − ξk(u(ξk+) − k) + ϕ(u(ξk+)) − A(u)′(ξk+) − ϕ(k) ≤ 0
195
+ ∀k ∈ [u(ξk−), u(ξk+)].
196
+ (16)
197
+ Notice that our solution (13) is a nonstrictly increasing function of the self-similar variable
198
+ ξ and, therefore, u(ξk−) ≤ u(ξk+).
199
+ Let us firstly analyze the solution (13) in the case ξk−1 < ξk < ξk+1. If ak−1, ak > 0
200
+ then u(ξk−) = u(ξk+) = uk so that condition (14) fulfils while (15) reduces to the equality
201
+ [A(u)′] = 0, which is revealed as
202
+ ak(uk+1 − uk)
203
+ F((ξk+1 − vk)/ak) − F((ξk − vk)/ak)F ′((ξk − vk)/ak) =
204
+ ak−1(uk − uk−1)
205
+ F((ξk − vk−1)/ak−1) − F((ξk−1 − vk−1)/ak−1)F ′((ξk − vk−1)/ak−1).
206
+ (17)
207
+ In this situation ξ = ξk is a weak discontinuity point, the function u(ξ) itself is continuous,
208
+ only its derivative u′(ξ) may be discontinuous. Moreover, it follows from (17) that both
209
+ functions u(ξ) and u′(ξ) are continuous at point ξk if ak = ak−1 > 0.
210
+ If ak−1 > ak = 0 then again u(ξ) is continuous at uk and (15) reduces to the relation
211
+ ak−1(uk − uk−1)
212
+ F((ξk − vk−1)/ak−1) − F((ξk−1 − vk−1)/ak−1)F ′((ξk − vk−1)/ak−1) = 0,
213
+ (18)
214
+ 4
215
+
216
+ which is impossible. If ak > ak−1 = 0 then u(ξk−) = uk−1 < uk = u(ξk+), that is, ξk is
217
+ a strong discontinuity point. Condition (14) holds because A(u) is constant on [uk−1, uk]
218
+ (A′(u) = a2
219
+ k−1 = 0) while (15) turns into
220
+ (vk−1 − ξk)(uk − uk−1) −
221
+ ak(uk+1 − uk)
222
+ F((ξk+1 − vk)/ak) − F((ξk − vk)/ak)F ′((ξk − vk)/ak) = 0, (19)
223
+ where we use the fact that ϕ(uk) − ϕ(uk−1) = vk−1(uk − uk−1). It remains to analyze the
224
+ situation when ak = ak−1 = 0. In this case again u(ξk−) = uk−1 < uk = u(ξk+) and
225
+ A(uk−1) = A(uk) while condition (15) turns into the simple relation
226
+ ξk = vk−1.
227
+ (20)
228
+ Finally, since the function ϕ(u) is affine on the segment [uk−1, uk] and A(u)′(ξk±) ≥ 0, then
229
+ entropy relation (16) is always satisfied.
230
+ Now we consider the case when there exists a nontrivial family of mutually equaled
231
+ values ξi, ξi = ξk for i = k, . . . , l, where l > k. We can assume that this family is maximal,
232
+ that is,
233
+ −∞ ≤ ξk−1 < ξk = · · · = ξl < ξl+1 ≤ +∞.
234
+ Then ai = 0 for i = k, . . . , l − 1, and the point ξ = c .= ξk is a discontinuity point of u(ξ)
235
+ with the unilateral limits
236
+ u(c+) = ul, u(c−) = uk′,
237
+ where k′ =
238
+ � k
239
+ ,
240
+ ak−1 > 0,
241
+ k − 1
242
+ ,
243
+ ak−1 = 0 .
244
+ Since a(u) = 0 for u(c−) < u < u(c+), we find that A(u(c−)) = A(u(c+)) and condition
245
+ (14) is satisfied. Further, we notice that
246
+ l−1
247
+
248
+ i=k′
249
+ (−ξi+1(ui+1 − ui)) = −c
250
+ l−1
251
+
252
+ i=k′
253
+ (ui+1 − ui) = −c(ul − uk���),
254
+ l−1
255
+
256
+ i=k′
257
+ vi(ui+1 − ui) =
258
+ l−1
259
+
260
+ i=k′
261
+ (ϕ(ui+1) − ϕ(ui)) = ϕ(ul) − ϕ(uk′).
262
+ Therefore, condition (15) can be written in the form
263
+ l−1
264
+
265
+ i=k′
266
+ (vi − ξi+1)(ui+1 − ui) − (A(u)′(c+) − A(u)′(c−)) = 0,
267
+ (21)
268
+ where, as is easy to verify,
269
+ A(u)′(c−) =
270
+
271
+ 0
272
+ ,
273
+ ak−1 = 0,
274
+ ak−1(uk−uk−1)
275
+ F((ξk−vk−1)/ak−1)−F((ξk−1−vk−1)/ak−1)F ′((ξk − vk−1)/ak−1)
276
+ ,
277
+ ak−1 > 0,
278
+ (22)
279
+ A(u)′(c+) =
280
+
281
+ 0
282
+ ,
283
+ al = 0,
284
+ al(ul+1−ul)
285
+ F((ξl+1−vl)/al)−F((ξl−vl)/al)F ′((ξl − vl)/al)
286
+ ,
287
+ al > 0.
288
+ (23)
289
+ 5
290
+
291
+ In the similar way we can write the Oleinik condition (16) as follows
292
+ l−1
293
+
294
+ i=j
295
+ (vi − ξi+1)(ui+1 − ui) − A(u)′(c+) ≤ 0
296
+ k′ < j < l.
297
+ (24)
298
+ We use here the fact the function ϕ(u) is piecewise affine and, therefore, it is enough to
299
+ verify the Oleinik condition (16) only at the nodal points k = uj.
300
+ The above reasoning remains valid also in the case when l = k. In this case, relation
301
+ (21) reduces to one of conditions (17), (18), (19), (20) while (24) is trivial.
302
+ 3
303
+ The entropy function
304
+ We introduce the convex cone Ω ⊂ Rd consisting of points ¯ξ = (ξ1, . . . , ξd) with increasing
305
+ coordinates, ξ1 ≤ ξ2 ≤ · · · ≤ ξd such that ξk+1 > ξk whenever ak > 0, k = 1, . . . , d − 1.
306
+ Each point ¯ξ ∈ Ω determines a function u(ξ) in correspondence with formula (13). Assume
307
+ firstly that ¯ξ ∈ Int Ω, that is, the values ξk are strictly increasing. Then conditions (17),
308
+ (18), (19), (20) coincides with the equality
309
+
310
+ ∂ξk E(¯ξ) = 0, where
311
+ E(¯ξ) = −
312
+
313
+ k=0,...,n−1,ak>0
314
+ (ak)2(uk+1 − uk) ln(F((ξk+1 − vk)/ak) − F((ξk − vk)/ak))+
315
+ 1
316
+ 2
317
+
318
+ k=0,...,n−1,ak=0
319
+ (uk+1 − uk)(ξk+1 − vk)2,
320
+ ¯ξ = (ξ1, . . . , ξd) ∈ Ω.
321
+ (25)
322
+ We will call this function the entropy because it depends only on the discontinuities of a
323
+ solution. Thus, for ¯ξ ∈ Int Ω the e.s. (13) corresponds to a critical point of the entropy. We
324
+ are going to demonstrate that the entropy is strictly convex and coercive in Ω. Therefore,
325
+ it has a unique global minimum point in Ω. In the case when this minimum point lies in
326
+ Int Ω it is a unique critical point.
327
+ Obviously, E(¯ξ) ∈ C∞(Ω). Notice that for all k = 0, . . . , n − 1, such that ak > 0
328
+ ln(F((ξk+1 − vk)/ak) − F((ξk − vk)/ak)) < 0.
329
+ Therefore, all terms in expression (25) are nonnegative and, in particular, E(¯ξ) ≥ 0.
330
+ Proposition 1 (coercivity). The sets E(¯ξ) ≤ c are compact for each constant c ≥ 0.
331
+ Proof. If E(¯ξ) ≤ c then it follows from nonnegativity of all terms in (25) that for all
332
+ k = 0, . . . , n − 1
333
+ −(ak)2(uk+1 − uk) ln(F((ξk+1 − vk)/ak) − F((ξk − vk)/ak)) ≤ E(¯ξ) ≤ c if ak > 0,
334
+ (26)
335
+ (uk+1 − uk)(ξk+1 − vk)2/2 ≤ c if ak = 0.
336
+ (27)
337
+ Relation (26) implies the estimate
338
+ F((ξk+1 − vk)/ak) − F((ξk − vk)/ak) ≥ δ .= exp(−c/m) > 0,
339
+ (28)
340
+ where m =
341
+ min
342
+ k=0,...,n−1,ak>0(ak)2(uk+1 − uk) > 0. If a0 > 0 relation (28) with k = 0 reads
343
+ F((ξ1 − v0)/a0) > δ (notice that F((ξ0 − v0)/a0) = F(−∞) = 0), which implies that
344
+ ξ1 ≥ v0 + a0F −1(δ).
345
+ 6
346
+
347
+ On the other hand, if a0 = 0 then (u1 − u0)(ξ1 − v0)2 ≤ 2c, in view of (27) with k = 0, and
348
+ ξ1 ≥ v0 − (2c/(u1 − u0))1/2.
349
+ In any case,
350
+ ξ1 ≥ r1 .= v0 + min(a0F −1(δ), −(2c/(u1 − u0))1/2).
351
+ (29)
352
+ To get an upper bound, we remark that in the case an−1 > 0 it follows from (28) with
353
+ k = d = n − 1 that F(−(ξn−1 − vn−1)/an−1) = 1 − F((ξn−1 − vn−1)/an−1) ≥ δ (observe that
354
+ F((ξn − vn−1)/an−1) = F(+∞) = 1), which implies the estimate
355
+ ξd ≤ vn−1 − an−1F −1(δ).
356
+ If an−1 = 0 then d = n and in view of inequality (27) with k = n − 1 we find (un −
357
+ un−1)(ξn − vn−1)2/2 ≤ c, that is,
358
+ ξd ≤ vn−1 + (2c/(un − un−1))1/2.
359
+ In both cases
360
+ ξd ≤ r2 .= vn−1 + max(−an−1F −1(δ), (2c/(un − un−1))1/2).
361
+ (30)
362
+ Since all coordinates of ¯ξ lie between ξ1 and ξd, estimates (29), (30) imply the bound
363
+ |¯ξ|∞ = max
364
+ k=1,...,d |ξk| ≤ r .= max(|r1|, |r2|).
365
+ Further, since F ′(x) =
366
+ 1
367
+
368
+ 2πe−x2/2 < 1, the function F(x) is Lipschitz with constant 1 and
369
+ it follows from (28) that
370
+ (ξk+1 − ξk)/ak ≥ F((ξk+1 − vk)/ak) − F((ξk − vk)/ak) ≥ δ,
371
+ k = 1, . . . , d − 1, ak > 0.
372
+ We find that
373
+ ξk+1 − ξk ≥ akδ
374
+ (this also includes the case ak = 0). We conclude tat the set E(¯ξ) ≤ c lies in the compact
375
+ set
376
+ K = { ¯ξ = (ξ1, . . . , ξd) ∈ Rd | |¯ξ|∞ ≤ r, ξk+1 − ξk ≥ akδ ∀k = 1, . . . , d − 1 }.
377
+ By the continuity of E(¯ξ) the set E(¯ξ) ≤ c is a closed subset of K and therefore is
378
+ compact.
379
+ We take c > N .= inf E(¯ξ). Then the set E(¯ξ) ≤ c is not empty. By Proposition 1
380
+ this set is compact and therefore the continuous function E(¯ξ) reaches the minimal value
381
+ on it, which is evidently equal to N. We proved the existence of global minimum E(¯ξ0) =
382
+ min E(¯ξ). The uniqueness of the minimum point is a consequence of strict convexity of the
383
+ entropy, which is stated in Proposition 2 below. The following lemma plays a key role.
384
+ Lemma 1. The function P(x, y) = − ln(F(x) − F(y)) is strictly convex in the half-plane
385
+ x > y.
386
+ 7
387
+
388
+ Proof. The function P(x, y) is infinitely differentiable in the domain x > y. To prove the
389
+ lemma, we need to establish that the Hessian D2P is positive definite at every point. By
390
+ the direct computation we find
391
+ ∂2
392
+ ∂x2P(x, y) = (F ′(x))2 − F ′′(x)(F(x) − F(y))
393
+ (F(x) − F(y))2
394
+ ,
395
+ ∂2
396
+ ∂y2P(x, y) = (F ′(y))2 − F ′′(y)(F(y) − F(x))
397
+ (F(x) − F(y))2
398
+ ,
399
+ ∂2
400
+ ∂x∂yP(x, y) = −
401
+ F ′(x)F ′(y)
402
+ (F(x) − F(y))2.
403
+ We have to prove positive definiteness of the matrix Q = (F(x) − F(y))2D2P(x, y) with
404
+ the components
405
+ Q11 = (F ′(x))2 − F ′′(x)(F(x) − F(y)),
406
+ Q22 = (F ′(y))2 − F ′′(y)(F(y) − F(x)), Q12 = Q21 = −F ′(x)F ′(y).
407
+ Since F ′(x) = e−x2/2, then F ′′(x) = −xF ′(x) and the diagonal elements of this matrix can
408
+ be written in the form
409
+ Q11 = F ′(x)(x(F(x) − F(y)) + F ′(x)) =
410
+ F ′(x)(x(F(x) − F(y)) + (F ′(x) − F ′(y))) + F ′(x)F ′(y),
411
+ Q22 = F ′(y)(y(F(y) − F(x)) + (F ′(y) − F ′(x))) + F ′(x)F ′(y).
412
+ By Cauchy mean value theorem there exists such a value z ∈ (y, x) that
413
+ F ′(x) − F ′(y)
414
+ F(x) − F(y) = F ′′(z)
415
+ F ′(z) = −z.
416
+ Therefore,
417
+ Q11 = F ′(x)(F(x) − F(y))(x − z) + F ′(x)F ′(y),
418
+ Q22 = F ′(y)(F(x) − F(y))(z − y) + F ′(x)F ′(y),
419
+ and it follows that Q = R1 +F ′(x)F ′(y)R2, where R1 is a diagonal matrix with the positive
420
+ diagonal elements F ′(x)(F(x)−F(y))(x−z), F ′(y)(F(x)−F(y))(z−y) while R2 =
421
+ � 1
422
+ −1
423
+ −1
424
+ 1
425
+
426
+ .
427
+ Since R1 > 0, R2 ≥ 0, then the matrix Q > 0, as was to be proved.
428
+ Corollary 1. The functions P(x, −∞) = − ln F(x), P(+∞, x) = − ln(1 − F(x)) of single
429
+ variable are strictly convex.
430
+ Proof. Since 1 − F(x) = F(−x), we see that P(+∞, x) = P(−x, −∞), and it is sufficient
431
+ to prove the strict convexity of the function P(x, −∞) = − ln F(x). By Lemma 1 in the
432
+ limit as y → −∞ we obtain that this function is convex, moreover,
433
+ 0 ≤ (F(x))2 d2
434
+ dx2P(x, −∞) = lim
435
+ y→−∞ Q11 = F ′(x)(xF(x) + F ′(x)).
436
+ If
437
+ d2
438
+ dx2P(x, −∞) = 0 at some point x = x0 then 0 = x0F(x0)+F ′(x0) is the minimum of the
439
+ nonnegative function xF(x) + F ′(x). Therefore, its derivative (xF + F ′)′(x0) = 0. Since
440
+ F ′′(x) = −xF ′(x), this derivative
441
+ (xF + F ′)′(x0) = F(x0) + x0F ′(x0) + F ′′(x0) = F(x0) > 0.
442
+ But this contradicts our assumption. We conclude that
443
+ d2
444
+ dx2P(x, −∞) > 0 and the function
445
+ P(x, −∞) is strictly convex.
446
+ 8
447
+
448
+ Proposition 2 (convexity). The entropy function E(¯ξ) is strictly convex on Ω.
449
+ Proof. For k = 0, . . . , n − 1 we denote Pk(¯ξ) = − ln(F((ξk+1 − vk)/ak) − F((ξk − vk)/ak))
450
+ if ak > 0, and Pk(¯ξ) = (ξk+1 − vk)2 if ak = 0. In view of (25) the entropy E(¯ξ) is a linear
451
+ combination of the functions Pk with positive coefficients, and convexity of the entropy
452
+ readily follows from the statements of Lemma 1 and Corollary 1. To establish the strict
453
+ convexity, we have to demonstrate that the Hessian matrix D2E(¯ξ) is strictly positive.
454
+ Assume that for some ζ = (ζ1, . . . , ζd) ∈ Rd
455
+ D2E(¯ξ)ζ · ζ =
456
+ d
457
+
458
+ i,j=1
459
+ ∂2E(¯ξ)
460
+ ∂ξi∂ξj
461
+ ζiζj = 0.
462
+ (31)
463
+ Since E(¯ξ) is a linear combination of convex functions Pk(¯ξ) with positive coefficients, we
464
+ find that
465
+ D2Pk(¯ξ)ζ · ζ = 0
466
+ ∀k = 0, . . . , n − 1.
467
+ This can be written in the form
468
+
469
+ i,j=k,k+1
470
+ ∂2Pk(¯ξ)
471
+ ∂ξi∂ξj
472
+ ζiζj = 0 if 0 < k < n − 1, ak > 0;
473
+ ∂2Pk(¯ξ)
474
+ ∂ξ2
475
+ k+1
476
+ ζ2
477
+ k+1 if k = 0 or ak = 0.
478
+ In view of Lemma 1 and Corollary 1 the functions Pk in above equalities are strictly convex
479
+ as functions of either two variables (ξk, ξk+1) or single variable ξk+1.
480
+ Therefore, these
481
+ equalities imply that in any case ζk+1 = 0, k = 0, . . . , n − 2, and ζn = 0 if an−1 = 0 (when
482
+ d = n). We conclude that all coordinates ζi = 0, i = 1, . . . , d. Hence, equality (31) can
483
+ hold only for ζ = 0 and the matrix D2P(¯ξ) > 0 for all ¯ξ ∈ Ω. This completes the proof.
484
+ 4
485
+ The variational formulation
486
+ Let ¯ξ0 = (ξ1, . . . , ξd) ∈ Ω be the unique minimum point of E(¯ξ).
487
+ The necessary and
488
+ sufficient condition for ¯ξ0 to be a minimum point is the following one
489
+ ∇E(¯ξ0) · p ≥ 0
490
+ ∀p ∈ T(¯ξ0) = { p ∈ Rd | ∃h > 0 ¯ξ0 + hp ∈ Ω },
491
+ (32)
492
+ so that T(¯ξ0) is the tangent cone to Ω at the point ¯ξ0. If ¯ξ0 ∈ Int Ω then T(¯ξ0) = Rd
493
+ and (32) reduces to the requirement ∇E(¯ξ0) = 0. As we have already demonstrated, this
494
+ requirement coincides with jump conditions (17), (18), (19), (20) for all k = 1, . . . , d. But
495
+ these conditions are equivalent to the statement that the function (13) is an e.s. of (1),
496
+ (3). In the general situation when ¯ξ0 can belong to the boundary of Ω, the coordinates of
497
+ ¯ξ0 may coincides. Let ξk = · · · = ξl = c be a maximal family of coinciding coordinates,
498
+ that is, ξk−1 < ξk = ξl < ξl+1 (it is possible here that k = l). Then, as is easy to realize,
499
+ the vector p = (p1, . . . , pd), with arbitrary increasing coordinates pk ≤ · · · ≤ pl and with
500
+ zero remaining coordinates, belong to the tangent cone T(¯ξ0). In view of (32)
501
+ l
502
+
503
+ i=k
504
+
505
+ ∂ξi
506
+ E(¯ξ0)pi ≥ 0
507
+ 9
508
+
509
+ for any such a vector. Using the summation by parts formula, we realize that the above
510
+ condition is equivalent to the following requirements
511
+ l
512
+
513
+ i=k
514
+
515
+ ∂ξi
516
+ E(¯ξ0) = 0,
517
+ (33)
518
+ l
519
+
520
+ i=j
521
+
522
+ ∂ξi
523
+ E(¯ξ0) ≥ 0, k < j ≤ l.
524
+ (34)
525
+ Recall that ai = 0 for k ≤ i < l. By the direct computation we find
526
+
527
+ ∂ξi
528
+ E(¯ξ0) = (ui − ui−1)(ξi − vi−1),
529
+ k < i < l,
530
+
531
+ ∂ξk
532
+ E(¯ξ0) =
533
+ � (uk − uk−1)(ξk − vk−1)
534
+ ,
535
+ ak−1 = 0,
536
+ −A(u)′(c−)
537
+ ,
538
+ ak−1 > 0;
539
+
540
+ ∂ξl
541
+ E(¯ξ0) = A(u)′(c+),
542
+ where A(u)′(c±) are given by (22), (23). Putting these expressions into (33), (34), we
543
+ obtain exactly the jump conditions (21), (24). Therefore, the function (13) corresponding
544
+ to the point ¯ξ0 is an e.s. of (1), (3). Conversely, if (13) is an e.s. then relations (33), (34)
545
+ holds for all groups of coinciding coordinates. As is easy to verify, this is equivalent to the
546
+ criterion (32). We have proved our main result.
547
+ Theorem 1. The function (13) is an e.s. of (1), (3) if and only if ¯ξ0 = (ξ1, . . . , ξd) is the
548
+ minimum point of the entropy E(¯ξ).
549
+ Remark 1. Adding to the entropy (25) the constant
550
+
551
+ k=0,...,n−1,ak>0
552
+ (ak)2(uk+1 − uk) ln((uk+1 − uk)/ak),
553
+ we obtain the alternative variant of the entropy
554
+ E1(¯ξ) = −
555
+
556
+ k=0,...,n−1,ak>0
557
+ (ak)2(uk+1 − uk) ln
558
+ �F((ξk+1 − vk)/ak) − F((ξk − vk)/ak)
559
+ (uk+1 − uk)/ak
560
+
561
+ +1
562
+ 2
563
+
564
+ k=0,...,n−1,ak=0
565
+ (uk+1 − uk)(ξk+1 − vk)2. (35)
566
+ If we consider the values vk, ak as a piecewise constant approximation of an arbitrary
567
+ velocity function v(u) and, respectively, a diffusion function a(u) ≥ 0 then, passing in
568
+ (35) to the limit as max(uk+1 − uk) → 0, we find that the entropy E1(¯ξ) turns into the
569
+ variational functional
570
+ J(ξ) = −
571
+
572
+ {u∈[α,β],a(u)>0}
573
+ (a(u))2 ln(F ′((ξ(u) − v(u))/a(u))ξ′(u))du+
574
+ 1
575
+ 2
576
+
577
+ {u∈[α,β],a(u)=0}
578
+ (ξ(u) − v(u))2du,
579
+ 10
580
+
581
+ where ξ(u) is an increasing function on [α, β], which is expected to be the inverse function
582
+ to a self-similar solution u = u(ξ) of the problem (1), (3). Taking into account that
583
+ ln(F ′((ξ(u) − v(u))/a(u))ξ′(u)) = ln F ′((ξ(u) − v(u))/a(u)) + ln ξ′(u) =
584
+ −(ξ(u) − v(u))2
585
+ 2a2(u)
586
+ + ln ξ′(u),
587
+ we may simplify the expression for the functional J(ξ)
588
+ J(ξ) =
589
+ � β
590
+ α
591
+ [(ξ(u) − v(u))2/2 − (a(u))2 ln(ξ′(u))]du.
592
+ (36)
593
+ We see that this functional is strictly convex. The corresponding Euler-Lagrange equation
594
+ has the form
595
+ ξ(u) − v(u) + ((a(u))2/ξ′(u))′ = 0.
596
+ (37)
597
+ Since u′(ξ) = 1/ξ′(u), u = u(ξ), we can transform (37) as follows
598
+ ξ(u) − v(u) + ((a(u))2u′(ξ))′
599
+ u = 0.
600
+ Multiplying this equation by u′(ξ), we obtain the equation
601
+ (a2u′)′ = (v − ξ)u′,
602
+ u = u(ξ),
603
+ which is exactly our equation (1) written in the self-similar variable.
604
+ Remark 2. In the case of conservation laws (2) the e.s. u = u(ξ) of (2), (3) is piecewise
605
+ constant, and, by expression (13),
606
+ u(ξ) = uk,
607
+ ξk < ξ < ξk+1, k = 0, . . . , n,
608
+ where −∞ = ξ0 < ξ1 ≤ · · · ≤ ξn < ξn+1 = +∞. In this case the entropy function is
609
+ particularly simple, it is the quadratic function
610
+ E(¯ξ) = 1
611
+ 2
612
+ n
613
+
614
+ k=1
615
+ (uk − uk−1)(ξk − vk−1)2,
616
+ defined on the closed polyhedral cone
617
+ Ω = { ¯ξ = (ξ1, . . . , ξn) ∈ Rn | ξk+1 ≥ ξk ∀k = 1, . . . , n − 1 }.
618
+ Existence and uniqueness of a minimal point in this case is trivial. By Theorem 1 and
619
+ Remark 1 we obtain new, variational formulation of the entropy solution.
620
+ Acknowledgments
621
+ The research was supported by the Russian Science Foundation, grant 22-21-00344.
622
+ 11
623
+
624
+ References
625
+ [1] J. Carrillo, Entropy solutions for nonlinear degenerate problems, Arch. Ration. Mech.
626
+ Anal., 147 (1999), 269–361.
627
+ [2] S. N. Kruzhkov, First order quasilinear equations in several independent variables, Mat.
628
+ Sb. (N.S.), 81 (1970), 228–255.
629
+ [3] O. A. Oleinik, Uniqueness and stability of the generalized solution of the Cauchy prob-
630
+ lem for a quasi-linear equation, Uspekhi Mat. Nauk, 14:2(86) (1959), 165–170.
631
+ 12
632
+
KdFQT4oBgHgl3EQfTTZa/content/tmp_files/load_file.txt ADDED
@@ -0,0 +1,335 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf,len=334
2
+ page_content='On the structure of entropy solutions to the Riemann problem for a degenerate nonlinear parabolic equation Evgeny Yu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
3
+ page_content=' Panov Yaroslav-the-Wise Novgorod State University, Veliky Novgorod, Russian Federation, Research and Development Center, Veliky Novgorod, Russian Federation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
4
+ page_content=' Abstract We find an explicit form of entropy solutions to a Riemann problem for a de- generate nonlinear parabolic equation with piecewise constant velocity and diffusion coefficients.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
5
+ page_content=' It is demonstrated that this solution corresponds to the minimum point of some strictly convex function of a finite number of variables.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
6
+ page_content=' 1 Introduction In a half-plane Π = {(t, x) | t > 0, x ∈ R}, we consider a nonlinear parabolic equation ut + v(u)ux − t(a2(u)ux)x = 0, (1) where v(u), a(u) ∈ L∞(R), a(u) ≥ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
7
+ page_content=' Since the diffusion coefficient a(u) may take zero value, equation (1) is degenerate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
8
+ page_content=' In the case when a(u) ≡ 0 it reduces to a first order conservation law ut + ϕ(u)x = 0, (2) where ϕ′(u) = v(u).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
9
+ page_content=' Similarly, a general equation (1) can be written in the conservative form ut + ϕ(u)x − tA(u)xx = 0 with A′(u) = a2(u), which allows to define weak solutions of this equation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
10
+ page_content=' Unfortunately, weak solutions to a Cauchy problem for equation (1) are not unique in general, and some additional entropy conditions are required.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
11
+ page_content=' We consider the Cauchy problem with initial data u(0, x) = u0(x), (3) where u0(x) ∈ L∞(R).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
12
+ page_content=' Recall the notion of entropy solution (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
13
+ page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
14
+ page_content=' for short) in the sense of Carrillo [1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
15
+ page_content=' Definition 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
16
+ page_content=' A function u = u(t, x) ∈ L∞(Π) is called an e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
17
+ page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
18
+ page_content=' of (1), (3) if (i) the distribution A(u) ∈ L2 loc(Π);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
19
+ page_content=' (ii) for all k ∈ R |u − k|t + (sign(u − k)(ϕ(u) − ϕ(k)))x − (t sign(u − k)(A(u) − A(k)))xx ≤ 0 (4) in the sense of distributions (in D′(Π));' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
20
+ page_content=' (iii) ess lim t→0 u(t, ·) = u0 in L1 loc(R).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
21
+ page_content=' 1 arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
22
+ page_content='13292v1 [math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
23
+ page_content='AP] 30 Jan 2023 Entropy condition (4) means that for each nonnegative test function f = f(t, x) ∈ C2 0(Π) � Π [|u − k|ft + sign(u − k)((ϕ(u) − ϕ(k))fx + t(A(u) − A(k))fxx)]dtdx ≥ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
24
+ page_content=' (5) In the case of conservation laws (2) the notion of e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
25
+ page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
26
+ page_content=' reduces to the notion of generalized e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
27
+ page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
28
+ page_content=' in the sense of Kruzhkov [2].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
29
+ page_content=' Taking in (4) k = ±M, M ≥ ∥u∥∞, we derive that ut + ϕ(u)x − tA(u)xx = 0 in D′(Π), that is, an e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
30
+ page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
31
+ page_content=' u of (1), (3) is a weak solution of this problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
32
+ page_content=' It is known that an e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
33
+ page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
34
+ page_content=' of (1), (3) always exists and is unique.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
35
+ page_content=' In general multidimensional setting this was demonstrated in [2] for conservation laws and in [1] for the general case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
36
+ page_content=' If to be precise, in [1] the case of usual diffusion term A(u)xx was studied but the proofs can be readily adapted to the case of the self-similar diffusion tA(u)xx.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
37
+ page_content=' If u = u(t, x) is a piecewise C2-smooth e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
38
+ page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
39
+ page_content=' of equation (1) then it must satisfy this equation in classic sense in each smoothness domain.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
40
+ page_content=' Applying relation (1) to a test function f = f(t, x) ∈ C2 0(Π) supported in a neighborhood of a discontinuity line x = x(t) and integrating by parts, we then obtain the identity (−x′(t)[u] + [ϕ(u)] − t[A(u)x])f + t[A(u)]fx = 0 (6) a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
41
+ page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
42
+ page_content=' on the line x = x(t).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
43
+ page_content=' Here we denote by [w] the jump of a function w = w(t, x) on the line x = x(t) so that [w] = w(t, x(t)+) − w(t, x(t)−), where w(t, x(t)±) = lim y→x(t)± w(t, y).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
44
+ page_content=' Since the functions f, fx are arbitrary and independent on the line x = x(t), identity (6) implies the following two relations of Rankine-Hugoniot type [A(u)] = 0, (7) −x′(t)[u] + [ϕ(u)] − t[A(u)x] = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
45
+ page_content=' (8) Similarly, it follows from entropy relation (5), after integration by parts, that (−x′(t)[|u − k|] + [sign(u − k)(ϕ(u) − ϕ(k))] − t[sign(u − k)A(u)x])f+ t[sign(u − k)(A(u) − A(k))]fx ≤ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
46
+ page_content=' (9) Since the function A(u) increases, it follows from (7) that A(u) = const when u lies between the values u(t, x(t)−) and u(t, x(t)+).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
47
+ page_content=' This implies that [sign(u − k)(A(u) − A(k))] = 0 and in view of arbitrariness of f ≥ 0 it follows from (9) that − x′(t)[|u − k|] + [sign(u − k)(ϕ(u) − ϕ(k))] − t[sign(u − k)A(u)x] ≤ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
48
+ page_content=' (10) In the case when k lies out of the interval with the endpoints u± .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
49
+ page_content='= u(t, x(t)±) relation (10) follows from (8) and fulfils with equality sign.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
50
+ page_content=' When u− < k < u+ this relation reads −x′(t)(u+ + u− − 2k) + ϕ(u+) + ϕ(u−) − 2ϕ(k) − t(A(u)+ x + A(u)− x ) ≤ 0, 2 where A(u)± x = A(u)x(t, x(t)±).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
51
+ page_content=' Adding (8) to this relation and dividing the result by 2, we arrive at the following analogue of the famous Oleinik condition (see [3]) known for conservation laws.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
52
+ page_content=' − x′(t)(u+ − k) + ϕ(u+) − tA(u)+ x − ϕ(k) ≤ 0 ∀k ∈ [u−, u+].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
53
+ page_content=' (11) In the case u+ < u− this condition has the form − x′(t)(u+ − k) + ϕ(u+) − tA(u)+ x − ϕ(k) ≥ 0 ∀k ∈ [u+, u−] (12) and can be derived similarly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
54
+ page_content=' Geometric interpretation of these conditions is that the graph of the flux function ϕ(u) lies not below (not above) of the segment connecting the points (u−, ϕ(u−) − tA(u)− x ), (u+, ϕ(u+) − tA(u)+ x ) when u− ≤ u ≤ u+ (respectively, when u+ ≤ u ≤ u−), see Figure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
55
+ page_content=' We take here into account that in view of condition (8) the vector (−x′(t), 1) is a normal to the indicated segment.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
56
+ page_content=' We also notice that it follows from relations (11), (12) with k = u± and from the Rankine-Hugoniot condition (8) that A(u)± x ≥ 0 (A(u)± x ≤ 0) whenever u+ > u− (u+ < u−).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
57
+ page_content=" u u- u+ y=φ(u) y tA(u)x tA(u)x + (-x',1) k Figure 1: Oleinik condition." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
58
+ page_content=' 2 The case of piecewise constant coefficients.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
59
+ page_content=' Below we will assume that the functions v(u), a(u) are piecewise constant, v(u) = vk, a(u) = ak when uk < u < uk+1, k = 0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
60
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
61
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
62
+ page_content=' , n − 1, where α = u0 < u1 < · · · < un−1 < un = β.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
63
+ page_content=' We will study problem (1), (3) with the Riemann data u0(x) = � α, x < 0, β, x > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
64
+ page_content=' Since this problem is invariant under the scaling transformations t → λt, x → λx, λ > 0 then, by the uniqueness, the e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
65
+ page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
66
+ page_content=' u = u(t, x) is self-similar: u(t, x) = u(λt, λx).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
67
+ page_content=' This implies that u = u(x/t).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
68
+ page_content=' Suppose that ak > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
69
+ page_content=' Then in a domain where uk < u(ξ) < uk+1 with ξ = x/t equation (1) reduces to the second order ODE (vk − ξ)u′ − a2 ku′′ = 0, 3 the general solution of which is u = C1F((ξ − vk)/ak) + C2;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
70
+ page_content=' C1, C2 = const, where F(z) = 1 √ 2π � z −∞ e−s2/2ds is the error function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
71
+ page_content=' Therefore, it is natural to seek the e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
72
+ page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
73
+ page_content=' of our problem in the following form u(ξ) = � uk + uk+1−uk F((ξk+1−vk)/ak)−F((ξk−vk)/ak)(F((ξ − vk)/ak) − F((ξk − vk)/ak)) , ak > 0, uk , ak = 0, (13) ξk < ξ < ξk+1, k = 0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
74
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
75
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
76
+ page_content=' , d, where d = � n − 1 , an−1 > 0, n , an−1 = 0, − ∞ = ξ0 < ξ1 ≤ · · · ≤ ξd < ξd+1 = +∞, and we agree that an = 0, F(−∞) = 0, F(+∞) = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
77
+ page_content=' We also assume that ξk+1 > ξk whenever ak > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
78
+ page_content=' The rays x = ξkt for finite ξk are (weak or strong) discontinuity lines of u, they correspond to discontinuity points ξk of the function u(ξ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
79
+ page_content=' Observe that conditions (7), (8) turns into the following relations at points ξk [A(u)] = A(u(ξk+)) − A(u(ξk−)) = 0, (14) −ξk[u] + [ϕ(u)] − [A(u)′] = −ξk(u(ξk+) − u(ξk−)) + ϕ(u(ξk+))− ϕ(u(ξk−)) − A(u)′(ξk+) + A(u)′(ξk−) = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
80
+ page_content=' (15) Here w(ξk±) denotes unilateral limits of a function w(ξ) at the point ξk.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
81
+ page_content=' Similarly, the Oleinik condition (11) reads − ξk(u(ξk+) − k) + ϕ(u(ξk+)) − A(u)′(ξk+) − ϕ(k) ≤ 0 ∀k ∈ [u(ξk−), u(ξk+)].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
82
+ page_content=' (16) Notice that our solution (13) is a nonstrictly increasing function of the self-similar variable ξ and, therefore, u(ξk−) ≤ u(ξk+).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
83
+ page_content=' Let us firstly analyze the solution (13) in the case ξk−1 < ξk < ξk+1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
84
+ page_content=' If ak−1, ak > 0 then u(ξk−) = u(ξk+) = uk so that condition (14) fulfils while (15) reduces to the equality [A(u)′] = 0, which is revealed as ak(uk+1 − uk) F((ξk+1 − vk)/ak) − F((ξk − vk)/ak)F ′((ξk − vk)/ak) = ak−1(uk − uk−1) F((ξk − vk−1)/ak−1) − F((ξk−1 − vk−1)/ak−1)F ′((ξk − vk−1)/ak−1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
85
+ page_content=' (17) In this situation ξ = ξk is a weak discontinuity point, the function u(ξ) itself is continuous, only its derivative u′(ξ) may be discontinuous.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
86
+ page_content=' Moreover, it follows from (17) that both functions u(ξ) and u′(ξ) are continuous at point ξk if ak = ak−1 > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
87
+ page_content=' If ak−1 > ak = 0 then again u(ξ) is continuous at uk and (15) reduces to the relation ak−1(uk − uk−1) F((ξk − vk−1)/ak−1) − F((ξk−1 − vk−1)/ak−1)F ′((ξk − vk−1)/ak−1) = 0, (18) 4 which is impossible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
88
+ page_content=' If ak > ak−1 = 0 then u(ξk−) = uk−1 < uk = u(ξk+), that is, ξk is a strong discontinuity point.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
89
+ page_content=' Condition (14) holds because A(u) is constant on [uk−1, uk] (A′(u) = a2 k−1 = 0) while (15) turns into (vk−1 − ξk)(uk − uk−1) − ak(uk+1 − uk) F((ξk+1 − vk)/ak) − F((ξk − vk)/ak)F ′((ξk − vk)/ak) = 0, (19) where we use the fact that ϕ(uk) − ϕ(uk−1) = vk−1(uk − uk−1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
90
+ page_content=' It remains to analyze the situation when ak = ak−1 = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
91
+ page_content=' In this case again u(ξk−) = uk−1 < uk = u(ξk+) and A(uk−1) = A(uk) while condition (15) turns into the simple relation ξk = vk−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
92
+ page_content=' (20) Finally, since the function ϕ(u) is affine on the segment [uk−1, uk] and A(u)′(ξk±) ≥ 0, then entropy relation (16) is always satisfied.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
93
+ page_content=' Now we consider the case when there exists a nontrivial family of mutually equaled values ξi, ξi = ξk for i = k, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
94
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
95
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
96
+ page_content=' , l, where l > k.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
97
+ page_content=' We can assume that this family is maximal, that is, −∞ ≤ ξk−1 < ξk = · · · = ξl < ξl+1 ≤ +∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
98
+ page_content=' Then ai = 0 for i = k, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
99
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
100
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
101
+ page_content=' , l − 1, and the point ξ = c .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
102
+ page_content='= ξk is a discontinuity point of u(ξ) with the unilateral limits u(c+) = ul, u(c−) = uk′, where k′ = � k , ak−1 > 0, k − 1 , ak−1 = 0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
103
+ page_content=' Since a(u) = 0 for u(c−) < u < u(c+), we find that A(u(c−)) = A(u(c+)) and condition (14) is satisfied.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
104
+ page_content=' Further, we notice that l−1 � i=k′ (−ξi+1(ui+1 − ui)) = −c l−1 � i=k′ (ui+1 − ui) = −c(ul − uk′), l−1 � i=k′ vi(ui+1 − ui) = l−1 � i=k′ (ϕ(ui+1) − ϕ(ui)) = ϕ(ul) − ϕ(uk′).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
105
+ page_content=' Therefore, condition (15) can be written in the form l−1 � i=k′ (vi − ξi+1)(ui+1 − ui) − (A(u)′(c+) − A(u)′(c−)) = 0, (21) where, as is easy to verify, A(u)′(c−) = � 0 , ak−1 = 0, ak−1(uk−uk−1) F((ξk−vk−1)/ak−1)−F((ξk−1−vk−1)/ak−1)F ′((ξk − vk−1)/ak−1) , ak−1 > 0, (22) A(u)′(c+) = � 0 , al = 0, al(ul+1−ul) F((ξl+1−vl)/al)−F((ξl−vl)/al)F ′((ξl − vl)/al) , al > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
106
+ page_content=' (23) 5 In the similar way we can write the Oleinik condition (16) as follows l−1 � i=j (vi − ξi+1)(ui+1 − ui) − A(u)′(c+) ≤ 0 k′ < j < l.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
107
+ page_content=' (24) We use here the fact the function ϕ(u) is piecewise affine and, therefore, it is enough to verify the Oleinik condition (16) only at the nodal points k = uj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
108
+ page_content=' The above reasoning remains valid also in the case when l = k.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
109
+ page_content=' In this case, relation (21) reduces to one of conditions (17), (18), (19), (20) while (24) is trivial.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
110
+ page_content=' 3 The entropy function We introduce the convex cone Ω ⊂ Rd consisting of points ¯ξ = (ξ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
111
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
112
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
113
+ page_content=' , ξd) with increasing coordinates, ξ1 ≤ ξ2 ≤ · · · ≤ ξd such that ξk+1 > ξk whenever ak > 0, k = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
114
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
115
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
116
+ page_content=' , d − 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
117
+ page_content=' Each point ¯ξ ∈ Ω determines a function u(ξ) in correspondence with formula (13).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
118
+ page_content=' Assume firstly that ¯ξ ∈ Int Ω, that is, the values ξk are strictly increasing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
119
+ page_content=' Then conditions (17), (18), (19), (20) coincides with the equality ∂ ∂ξk E(¯ξ) = 0, where E(¯ξ) = − � k=0,.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
120
+ page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
121
+ page_content=',n−1,ak>0 (ak)2(uk+1 − uk) ln(F((ξk+1 − vk)/ak) − F((ξk − vk)/ak))+ 1 2 � k=0,.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
122
+ page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
123
+ page_content=',n−1,ak=0 (uk+1 − uk)(ξk+1 − vk)2, ¯ξ = (ξ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
124
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
125
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
126
+ page_content=' , ξd) ∈ Ω.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
127
+ page_content=' (25) We will call this function the entropy because it depends only on the discontinuities of a solution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
128
+ page_content=' Thus, for ¯ξ ∈ Int Ω the e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
129
+ page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
130
+ page_content=' (13) corresponds to a critical point of the entropy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
131
+ page_content=' We are going to demonstrate that the entropy is strictly convex and coercive in Ω.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
132
+ page_content=' Therefore, it has a unique global minimum point in Ω.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
133
+ page_content=' In the case when this minimum point lies in Int Ω it is a unique critical point.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
134
+ page_content=' Obviously, E(¯ξ) ∈ C∞(Ω).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
135
+ page_content=' Notice that for all k = 0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
136
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
137
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
138
+ page_content=' , n − 1, such that ak > 0 ln(F((ξk+1 − vk)/ak) − F((ξk − vk)/ak)) < 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
139
+ page_content=' Therefore, all terms in expression (25) are nonnegative and, in particular, E(¯ξ) ≥ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
140
+ page_content=' Proposition 1 (coercivity).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
141
+ page_content=' The sets E(¯ξ) ≤ c are compact for each constant c ≥ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
142
+ page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
143
+ page_content=' If E(¯ξ) ≤ c then it follows from nonnegativity of all terms in (25) that for all k = 0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
144
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
145
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
146
+ page_content=' , n − 1 −(ak)2(uk+1 − uk) ln(F((ξk+1 − vk)/ak) − F((ξk − vk)/ak)) ≤ E(¯ξ) ≤ c if ak > 0, (26) (uk+1 − uk)(ξk+1 − vk)2/2 ≤ c if ak = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
147
+ page_content=' (27) Relation (26) implies the estimate F((ξk+1 − vk)/ak) − F((ξk − vk)/ak) ≥ δ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
148
+ page_content='= exp(−c/m) > 0, (28) where m = min k=0,.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
149
+ page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
150
+ page_content=',n−1,ak>0(ak)2(uk+1 − uk) > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
151
+ page_content=' If a0 > 0 relation (28) with k = 0 reads F((ξ1 − v0)/a0) > δ (notice that F((ξ0 − v0)/a0) = F(−∞) = 0), which implies that ξ1 ≥ v0 + a0F −1(δ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
152
+ page_content=' 6 On the other hand, if a0 = 0 then (u1 − u0)(ξ1 − v0)2 ≤ 2c, in view of (27) with k = 0, and ξ1 ≥ v0 − (2c/(u1 − u0))1/2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
153
+ page_content=' In any case, ξ1 ≥ r1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
154
+ page_content='= v0 + min(a0F −1(δ), −(2c/(u1 − u0))1/2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
155
+ page_content=' (29) To get an upper bound, we remark that in the case an−1 > 0 it follows from (28) with k = d = n − 1 that F(−(ξn−1 − vn−1)/an−1) = 1 − F((ξn−1 − vn−1)/an−1) ≥ δ (observe that F((ξn − vn−1)/an−1) = F(+∞) = 1), which implies the estimate ξd ≤ vn−1 − an−1F −1(δ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
156
+ page_content=' If an−1 = 0 then d = n and in view of inequality (27) with k = n − 1 we find (un − un−1)(ξn − vn−1)2/2 ≤ c, that is, ξd ≤ vn−1 + (2c/(un − un−1))1/2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
157
+ page_content=' In both cases ξd ≤ r2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
158
+ page_content='= vn−1 + max(−an−1F −1(δ), (2c/(un − un−1))1/2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
159
+ page_content=' (30) Since all coordinates of ¯ξ lie between ξ1 and ξd, estimates (29), (30) imply the bound |¯ξ|∞ = max k=1,.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
160
+ page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
161
+ page_content=',d |ξk| ≤ r .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
162
+ page_content='= max(|r1|, |r2|).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
163
+ page_content=' Further, since F ′(x) = 1 √ 2πe−x2/2 < 1, the function F(x) is Lipschitz with constant 1 and it follows from (28) that (ξk+1 − ξk)/ak ≥ F((ξk+1 − vk)/ak) − F((ξk − vk)/ak) ≥ δ, k = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
164
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
165
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
166
+ page_content=' , d − 1, ak > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
167
+ page_content=' We find that ξk+1 − ξk ≥ akδ (this also includes the case ak = 0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
168
+ page_content=' We conclude tat the set E(¯ξ) ≤ c lies in the compact set K = { ¯ξ = (ξ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
169
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
170
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
171
+ page_content=' , ξd) ∈ Rd | |¯ξ|∞ ≤ r, ξk+1 − ξk ≥ akδ ∀k = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
172
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
173
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
174
+ page_content=' , d − 1 }.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
175
+ page_content=' By the continuity of E(¯ξ) the set E(¯ξ) ≤ c is a closed subset of K and therefore is compact.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
176
+ page_content=' We take c > N .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
177
+ page_content='= inf E(¯ξ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
178
+ page_content=' Then the set E(¯ξ) ≤ c is not empty.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
179
+ page_content=' By Proposition 1 this set is compact and therefore the continuous function E(¯ξ) reaches the minimal value on it, which is evidently equal to N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
180
+ page_content=' We proved the existence of global minimum E(¯ξ0) = min E(¯ξ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
181
+ page_content=' The uniqueness of the minimum point is a consequence of strict convexity of the entropy, which is stated in Proposition 2 below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
182
+ page_content=' The following lemma plays a key role.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
183
+ page_content=' Lemma 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
184
+ page_content=' The function P(x, y) = − ln(F(x) − F(y)) is strictly convex in the half-plane x > y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
185
+ page_content=' 7 Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
186
+ page_content=' The function P(x, y) is infinitely differentiable in the domain x > y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
187
+ page_content=' To prove the lemma, we need to establish that the Hessian D2P is positive definite at every point.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
188
+ page_content=' By the direct computation we find ∂2 ∂x2P(x, y) = (F ′(x))2 − F ′′(x)(F(x) − F(y)) (F(x) − F(y))2 , ∂2 ∂y2P(x, y) = (F ′(y))2 − F ′′(y)(F(y) − F(x)) (F(x) − F(y))2 , ∂2 ∂x∂yP(x, y) = − F ′(x)F ′(y) (F(x) − F(y))2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
189
+ page_content=' We have to prove positive definiteness of the matrix Q = (F(x) − F(y))2D2P(x, y) with the components Q11 = (F ′(x))2 − F ′′(x)(F(x) − F(y)), Q22 = (F ′(y))2 − F ′′(y)(F(y) − F(x)), Q12 = Q21 = −F ′(x)F ′(y).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
190
+ page_content=' Since F ′(x) = e−x2/2, then F ′′(x) = −xF ′(x) and the diagonal elements of this matrix can be written in the form Q11 = F ′(x)(x(F(x) − F(y)) + F ′(x)) = F ′(x)(x(F(x) − F(y)) + (F ′(x) − F ′(y))) + F ′(x)F ′(y), Q22 = F ′(y)(y(F(y) − F(x)) + (F ′(y) − F ′(x))) + F ′(x)F ′(y).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
191
+ page_content=' By Cauchy mean value theorem there exists such a value z ∈ (y, x) that F ′(x) − F ′(y) F(x) − F(y) = F ′′(z) F ′(z) = −z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
192
+ page_content=' Therefore, Q11 = F ′(x)(F(x) − F(y))(x − z) + F ′(x)F ′(y), Q22 = F ′(y)(F(x) − F(y))(z − y) + F ′(x)F ′(y), and it follows that Q = R1 +F ′(x)F ′(y)R2, where R1 is a diagonal matrix with the positive diagonal elements F ′(x)(F(x)−F(y))(x−z), F ′(y)(F(x)−F(y))(z−y) while R2 = � 1 −1 −1 1 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
193
+ page_content=' Since R1 > 0, R2 ≥ 0, then the matrix Q > 0, as was to be proved.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
194
+ page_content=' Corollary 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
195
+ page_content=' The functions P(x, −∞) = − ln F(x), P(+∞, x) = − ln(1 − F(x)) of single variable are strictly convex.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
196
+ page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
197
+ page_content=' Since 1 − F(x) = F(−x), we see that P(+∞, x) = P(−x, −∞), and it is sufficient to prove the strict convexity of the function P(x, −∞) = − ln F(x).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
198
+ page_content=' By Lemma 1 in the limit as y → −∞ we obtain that this function is convex, moreover, 0 ≤ (F(x))2 d2 dx2P(x, −∞) = lim y→−∞ Q11 = F ′(x)(xF(x) + F ′(x)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
199
+ page_content=' If d2 dx2P(x, −∞) = 0 at some point x = x0 then 0 = x0F(x0)+F ′(x0) is the minimum of the nonnegative function xF(x) + F ′(x).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
200
+ page_content=' Therefore, its derivative (xF + F ′)′(x0) = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
201
+ page_content=' Since F ′′(x) = −xF ′(x), this derivative (xF + F ′)′(x0) = F(x0) + x0F ′(x0) + F ′′(x0) = F(x0) > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
202
+ page_content=' But this contradicts our assumption.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
203
+ page_content=' We conclude that d2 dx2P(x, −∞) > 0 and the function P(x, −∞) is strictly convex.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
204
+ page_content=' 8 Proposition 2 (convexity).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
205
+ page_content=' The entropy function E(¯ξ) is strictly convex on Ω.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
206
+ page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
207
+ page_content=' For k = 0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
208
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
209
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
210
+ page_content=' , n − 1 we denote Pk(¯ξ) = − ln(F((ξk+1 − vk)/ak) − F((ξk − vk)/ak)) if ak > 0, and Pk(¯ξ) = (ξk+1 − vk)2 if ak = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
211
+ page_content=' In view of (25) the entropy E(¯ξ) is a linear combination of the functions Pk with positive coefficients, and convexity of the entropy readily follows from the statements of Lemma 1 and Corollary 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
212
+ page_content=' To establish the strict convexity, we have to demonstrate that the Hessian matrix D2E(¯ξ) is strictly positive.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
213
+ page_content=' Assume that for some ζ = (ζ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
214
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
215
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
216
+ page_content=' , ζd) ∈ Rd D2E(¯ξ)ζ · ζ = d � i,j=1 ∂2E(¯ξ) ∂ξi∂ξj ζiζj = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
217
+ page_content=' (31) Since E(¯ξ) is a linear combination of convex functions Pk(¯ξ) with positive coefficients, we find that D2Pk(¯ξ)ζ · ζ = 0 ∀k = 0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
218
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
219
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
220
+ page_content=' , n − 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
221
+ page_content=' This can be written in the form � i,j=k,k+1 ∂2Pk(¯ξ) ∂ξi∂ξj ζiζj = 0 if 0 < k < n − 1, ak > 0;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
222
+ page_content=' ∂2Pk(¯ξ) ∂ξ2 k+1 ζ2 k+1 if k = 0 or ak = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
223
+ page_content=' In view of Lemma 1 and Corollary 1 the functions Pk in above equalities are strictly convex as functions of either two variables (ξk, ξk+1) or single variable ξk+1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
224
+ page_content=' Therefore, these equalities imply that in any case ζk+1 = 0, k = 0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
225
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
226
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
227
+ page_content=' , n − 2, and ζn = 0 if an−1 = 0 (when d = n).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
228
+ page_content=' We conclude that all coordinates ζi = 0, i = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
229
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
230
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
231
+ page_content=' , d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
232
+ page_content=' Hence, equality (31) can hold only for ζ = 0 and the matrix D2P(¯ξ) > 0 for all ¯ξ ∈ Ω.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
233
+ page_content=' This completes the proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
234
+ page_content=' 4 The variational formulation Let ¯ξ0 = (ξ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
235
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
236
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
237
+ page_content=' , ξd) ∈ Ω be the unique minimum point of E(¯ξ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
238
+ page_content=' The necessary and sufficient condition for ¯ξ0 to be a minimum point is the following one ∇E(¯ξ0) · p ≥ 0 ∀p ∈ T(¯ξ0) = { p ∈ Rd | ∃h > 0 ¯ξ0 + hp ∈ Ω }, (32) so that T(¯ξ0) is the tangent cone to Ω at the point ¯ξ0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
239
+ page_content=' If ¯ξ0 ∈ Int Ω then T(¯ξ0) = Rd and (32) reduces to the requirement ∇E(¯ξ0) = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
240
+ page_content=' As we have already demonstrated, this requirement coincides with jump conditions (17), (18), (19), (20) for all k = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
241
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
242
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
243
+ page_content=' , d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
244
+ page_content=' But these conditions are equivalent to the statement that the function (13) is an e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
245
+ page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
246
+ page_content=' of (1), (3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
247
+ page_content=' In the general situation when ¯ξ0 can belong to the boundary of Ω, the coordinates of ¯ξ0 may coincides.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
248
+ page_content=' Let ξk = · · · = ξl = c be a maximal family of coinciding coordinates, that is, ξk−1 < ξk = ξl < ξl+1 (it is possible here that k = l).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
249
+ page_content=' Then, as is easy to realize, the vector p = (p1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
250
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
251
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
252
+ page_content=' , pd), with arbitrary increasing coordinates pk ≤ · · · ≤ pl and with zero remaining coordinates, belong to the tangent cone T(¯ξ0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
253
+ page_content=' In view of (32) l � i=k ∂ ∂ξi E(¯ξ0)pi ≥ 0 9 for any such a vector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
254
+ page_content=' Using the summation by parts formula, we realize that the above condition is equivalent to the following requirements l � i=k ∂ ∂ξi E(¯ξ0) = 0, (33) l � i=j ∂ ∂ξi E(¯ξ0) ≥ 0, k < j ≤ l.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
255
+ page_content=' (34) Recall that ai = 0 for k ≤ i < l.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
256
+ page_content=' By the direct computation we find ∂ ∂ξi E(¯ξ0) = (ui − ui−1)(ξi − vi−1), k < i < l, ∂ ∂ξk E(¯ξ0) = � (uk − uk−1)(ξk − vk−1) , ak−1 = 0, −A(u)′(c−) , ak−1 > 0;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
257
+ page_content=' ∂ ∂ξl E(¯ξ0) = A(u)′(c+), where A(u)′(c±) are given by (22), (23).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
258
+ page_content=' Putting these expressions into (33), (34), we obtain exactly the jump conditions (21), (24).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
259
+ page_content=' Therefore, the function (13) corresponding to the point ¯ξ0 is an e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
260
+ page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
261
+ page_content=' of (1), (3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
262
+ page_content=' Conversely, if (13) is an e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
263
+ page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
264
+ page_content=' then relations (33), (34) holds for all groups of coinciding coordinates.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
265
+ page_content=' As is easy to verify, this is equivalent to the criterion (32).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
266
+ page_content=' We have proved our main result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
267
+ page_content=' Theorem 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
268
+ page_content=' The function (13) is an e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
269
+ page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
270
+ page_content=' of (1), (3) if and only if ¯ξ0 = (ξ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
271
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
272
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
273
+ page_content=' , ξd) is the minimum point of the entropy E(¯ξ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
274
+ page_content=' Remark 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
275
+ page_content=' Adding to the entropy (25) the constant � k=0,.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
276
+ page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
277
+ page_content=',n−1,ak>0 (ak)2(uk+1 − uk) ln((uk+1 − uk)/ak), we obtain the alternative variant of the entropy E1(¯ξ) = − � k=0,.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
278
+ page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
279
+ page_content=',n−1,ak>0 (ak)2(uk+1 − uk) ln �F((ξk+1 − vk)/ak) − F((ξk − vk)/ak) (uk+1 − uk)/ak � +1 2 � k=0,.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
280
+ page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
281
+ page_content=',n−1,ak=0 (uk+1 − uk)(ξk+1 − vk)2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
282
+ page_content=' (35) If we consider the values vk,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
283
+ page_content=' ak as a piecewise constant approximation of an arbitrary velocity function v(u) and,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
284
+ page_content=' respectively,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
285
+ page_content=' a diffusion function a(u) ≥ 0 then,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
286
+ page_content=' passing in (35) to the limit as max(uk+1 − uk) → 0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
287
+ page_content=' we find that the entropy E1(¯ξ) turns into the variational functional J(ξ) = − � {u∈[α,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
288
+ page_content='β],' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
289
+ page_content='a(u)>0} (a(u))2 ln(F ′((ξ(u) − v(u))/a(u))ξ′(u))du+ 1 2 � {u∈[α,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
290
+ page_content='β],' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
291
+ page_content='a(u)=0} (ξ(u) − v(u))2du,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
292
+ page_content=' 10 where ξ(u) is an increasing function on [α,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
293
+ page_content=' β],' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
294
+ page_content=' which is expected to be the inverse function to a self-similar solution u = u(ξ) of the problem (1),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
295
+ page_content=' (3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
296
+ page_content=' Taking into account that ln(F ′((ξ(u) − v(u))/a(u))ξ′(u)) = ln F ′((ξ(u) − v(u))/a(u)) + ln ξ′(u) = −(ξ(u) − v(u))2 2a2(u) + ln ξ′(u), we may simplify the expression for the functional J(ξ) J(ξ) = � β α [(ξ(u) − v(u))2/2 − (a(u))2 ln(ξ′(u))]du.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
297
+ page_content=' (36) We see that this functional is strictly convex.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
298
+ page_content=' The corresponding Euler-Lagrange equation has the form ξ(u) − v(u) + ((a(u))2/ξ′(u))′ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
299
+ page_content=' (37) Since u′(ξ) = 1/ξ′(u), u = u(ξ), we can transform (37) as follows ξ(u) − v(u) + ((a(u))2u′(ξ))′ u = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
300
+ page_content=' Multiplying this equation by u′(ξ), we obtain the equation (a2u′)′ = (v − ξ)u′, u = u(ξ), which is exactly our equation (1) written in the self-similar variable.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
301
+ page_content=' Remark 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
302
+ page_content=' In the case of conservation laws (2) the e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
303
+ page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
304
+ page_content=' u = u(ξ) of (2), (3) is piecewise constant, and, by expression (13), u(ξ) = uk, ξk < ξ < ξk+1, k = 0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
305
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
306
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
307
+ page_content=' , n, where −∞ = ξ0 < ξ1 ≤ · · · ≤ ξn < ξn+1 = +∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
308
+ page_content=' In this case the entropy function is particularly simple, it is the quadratic function E(¯ξ) = 1 2 n � k=1 (uk − uk−1)(ξk − vk−1)2, defined on the closed polyhedral cone Ω = { ¯ξ = (ξ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
309
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
310
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
311
+ page_content=' , ξn) ∈ Rn | ξk+1 ≥ ξk ∀k = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
312
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
313
+ page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
314
+ page_content=' , n − 1 }.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
315
+ page_content=' Existence and uniqueness of a minimal point in this case is trivial.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
316
+ page_content=' By Theorem 1 and Remark 1 we obtain new, variational formulation of the entropy solution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
317
+ page_content=' Acknowledgments The research was supported by the Russian Science Foundation, grant 22-21-00344.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
318
+ page_content=' 11 References [1] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
319
+ page_content=' Carrillo, Entropy solutions for nonlinear degenerate problems, Arch.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
320
+ page_content=' Ration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
321
+ page_content=' Mech.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
322
+ page_content=' Anal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
323
+ page_content=', 147 (1999), 269–361.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
324
+ page_content=' [2] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
325
+ page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
326
+ page_content=' Kruzhkov, First order quasilinear equations in several independent variables, Mat.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
327
+ page_content=' Sb.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
328
+ page_content=' (N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
329
+ page_content='S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
330
+ page_content=' ), 81 (1970), 228–255.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
331
+ page_content=' [3] O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
332
+ page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
333
+ page_content=' Oleinik, Uniqueness and stability of the generalized solution of the Cauchy prob- lem for a quasi-linear equation, Uspekhi Mat.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
334
+ page_content=' Nauk, 14:2(86) (1959), 165–170.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
335
+ page_content=' 12' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/KdFQT4oBgHgl3EQfTTZa/content/2301.13292v1.pdf'}
KtA0T4oBgHgl3EQfCv_p/content/2301.01995v1.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5fc369ae1d7c8a45f049a5f204a0ef73ee652ff0084a2d9ad6e15198f8620215
3
+ size 17207322
KtAzT4oBgHgl3EQfyf5z/content/tmp_files/2301.01754v1.pdf.txt ADDED
@@ -0,0 +1,1195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Status of leptoquark models after LHC Run-2 and discovery
2
+ prospects at future colliders
3
+ Nishita Desai∗
4
+ Department of Theoretical Physics,
5
+ Tata Institute of Fundamental Research,
6
+ Mumbai, India 400005
7
+ Amartya Sengupta
8
+ Meghnad Saha Pally, Burdwan, India, 713104 †
9
+ We study limits from dilepton searches on leptoquark completions to the Standard
10
+ Model in the parameter space motivated by anomalies in the b → s sector. After a full
11
+ Run-2 analysis by LHCb, the disparity in lepton flavour violation has disappeared.
12
+ However, the mismatch in angular distributions as well as in Bs → µ+µ− partial
13
+ width is still unresolved and still implies a possible new physics contribution. We
14
+ probe three models of leptoquarks — scalar models S3 and R2 as well as vector
15
+ leptoquark model U1 using non-resonant dilepton searches to place limit on both the
16
+ mass and couplings to SM fermions. The exclusions of leptoquarks coupling either
17
+ non-uniformly to different lepton flavours or uniformly is examined. Interestingly, if
18
+ leptoquark couplings to electrons and muons are indeed universal, then the U1 model
19
+ parameter space that corresponds to the anomalous contribution should already
20
+ accessible with Run-2 data in the non-resonant eµ channel. In the non-universal
21
+ case, there is a significant exclusion in couplings, but not enough to reach regions
22
+ that explain observed anomalies. We, therefore, examine the prospective sensitivity
23
+ at the HL-LHC as well as of a 3 TeV future muon collider. For the vector leptoquark
24
+ model, we find that a muon collider can probe all of the relevant parameter space at
25
+ 95% confidence with just 1 fb−1 data whereas R2 and S3 models can be excluded at
26
+ 95% with 5 fb−1 and 6.5 fb−1 luminosity respectively.
27
+ ∗ nishita.desai@tifr.res.in
28
+ † amartya.sengupta@studenti.unipd.it
29
+ arXiv:2301.01754v1 [hep-ph] 4 Jan 2023
30
+
31
+ 2
32
+ I.
33
+ INTRODUCTION
34
+ An exciting development in recent years has been the measurement of ratios of de-
35
+ cay widths in the semileptonic rare decays of B-mesons [1–4], hinting at lepton flavour-
36
+ universality violation (LFV). The latest of these [1, 2] showed a measurement consistent
37
+ with the SM for certain lepton universality, however, there remains a mismatch with the
38
+ measured branching fraction of Bs → µ+µ− [3, 4] and in the angular distribution in the
39
+ decay B → K∗µ+µ− [5, 6]. Unsurprisingly, this has led to a spirited effort to understand
40
+ the source of the mismatch with the predictions of the Standard Model (SM) and to provide
41
+ new physics explanations for it. In particular, there have been several dedicated studies that
42
+ determine global fits to data in terms of effective field theoretic operators (see e.g. [7–11]).
43
+ There has also been some effort to explain the anomalies in terms of new particles, notably
44
+ with new vector bosons or leptoquarks [12–15]. The effects of the presence of such new
45
+ particles can generally be seen in other observables besides the LFV ratios, and in partic-
46
+ ular, in the high energy tails of certain distributions observable at the LHC. In this paper,
47
+ we examine the expected effects of leptoquarks with minimally required properties to cause
48
+ observed anomalies in the B-sector and report on current constraints and future prospects
49
+ of their detection.
50
+ We start by providing a bare-bones introduction to how the Effective Field Theoretic
51
+ (EFT) framework is used and translated to the measurement of the high-energy observables
52
+ that we examine in this paper. EFT provides a useful method to describe the low-energy
53
+ physics processes in which the short-distance (i.e. high-energy or UV) physics is encapsulated
54
+ in the Wilson coefficients whilst the rest of the long-distance physics is expressed in terms of
55
+ effective operators with those having dimensions higher than four being suppressed by powers
56
+ of an energy scale to maintain the mass dimension of each term in the Lagrangian. The
57
+ analytic form of the Wilson coefficient can then be calculated by “matching” the expressions
58
+ calculated from the EFT with the expressions from the full UV theory. We can use the
59
+ published value by one of the multiple groups to translate the B-meson observations into
60
+ best-fit values of the appropriate Wilson coefficients [7–11, 16]. We then match these values
61
+ to the expressions derived from the leptoquark model under study and study the consequence
62
+ of what that means on other production mechanisms at the LHC.
63
+ The anomalies seen in the data fall into two categories — (1) in the neutral current sector
64
+
65
+ 3
66
+ with b → s transitions, and (2) in the charged current sector with b → c transitions. In
67
+ this work, we concentrate mainly on models that explain the first of these [17], however, it
68
+ is known that one of the models we study viz. the U1 vector leptoquark can explain both
69
+ simultaneously(see e.g. table 2. of [12])
70
+ The relevant observations that motivate this work based on the full Run 1 and 2 dataset
71
+ are shown in table II in the appendix. For completeness, we show both the pre-December
72
+ 2022 LHCb announcement [1, 2] numbers, as well as the latest measurements.
73
+ The low-energy effective theory for the b → s flavour changing neutral current sector is
74
+ described in terms of an effective Hamiltonian which can be written as
75
+ Heff = −4GF
76
+
77
+ 2 VtbV ∗
78
+ ts
79
+ � �
80
+ Ci(µ)Oi(µ)
81
+
82
+ where Ci(µ) are the Wilson coefficients. The effective operators relevant to our study are
83
+ Ol1l2
84
+ 9
85
+ =
86
+ e2
87
+ (4π)2(¯sγµPLb)(¯l1γµl2),
88
+ Ol1l2
89
+ 10 =
90
+ e2
91
+ (4π)2(¯sγµPLb)(¯l1γµγ5l2)
92
+ (I.1)
93
+ Multiple fitting studies have found that the operator whose Wilson coefficient shows sig-
94
+ nificant deviation from the predicted SM value is the C9 and that the most likely discrepancy
95
+ seems to be in the Cµ+µ−
96
+ 9
97
+ coefficient. To stay consistent with the latest data, we use the
98
+ Author (Year)
99
+ Model Dependent Data Driven
100
+ Ciuchini et al (2022) [7]
101
+ [−1.25, −0.72]
102
+ [−1.10, 1.05]
103
+ Ciuchini et al (2019) [8]
104
+ [−1.37, −1.05]
105
+ [−1.47, −0.93]
106
+ Alguer´o et al (2019) [9]
107
+ [−1.15, −0.81]
108
+ Alok et al (2019) [10]
109
+ [−1.27, −0.91]
110
+ Mahmoudi et al (2021) [11]
111
+ [−1.07, −0.83]
112
+ TABLE I. Best Fit values for the new physics contribution to the operator C9. The first of these
113
+ contains the updated 2022 results. The fits taking into account angular distributions still favour
114
+ a similar range as before the 2022 LHCb data release even though the overall best fit 1σ range is
115
+ now consistent with the SM value of zero.
116
+
117
+ 4
118
+ most recent best-fit results as reported in [7]. We shall use the best fit values that correctly
119
+ give the angular correlations as well (the so called “model-dependent” fit). However, later
120
+ in the paper when we examine future prospects, we also show the overlap with the fully
121
+ agnostic data-driven fits. For an overview of the best-fit C9 values see table I. Currently, we
122
+ proceed by using the value
123
+ Cµ+µ−
124
+ 9
125
+ = −0.98 ± 0.27,
126
+ Multiple studies have also examined the leptoquark UV completion and calculated explicit
127
+ expressions for Cµ+µ−
128
+ 9
129
+ from each model. In this work, we use these expressions to investigate
130
+ the LHC constraints on the couplings and mass of the leptoquarks.
131
+ We make only the
132
+ minimal assumptions, i.e. only the couplings that are necessary to give a contribution to
133
+ the b → s anomalies is assumed to be non-zero. As we shall see, in each leptoquark model,
134
+ the Wilson coefficients C9(10) depend on three parameters roughly as
135
+ C9 ∼
136
+ �y22 y32
137
+ M
138
+ �2
139
+ where y22 is the sµ coupling, y32 is the bµ coupling and M is the mass of the leptoquark. We
140
+ start by constraining (y22, y32, M) in other production modes without any further assump-
141
+ tions on other leptoquark couplings. This results in the most conservative limits. In the case
142
+ where there is no LFV, one would expect identical couplings of the leptoquark to electrons,
143
+ i.e. y22 = y21 and y32 = y31. This would also lead to signatures with different flavored
144
+ dileptons which often have much stronger constraints. These constraints are examined in
145
+ section III. In the flavour universal case, the strongest limits on leptoquark masses will come
146
+ from µ → e processes including µ → eγ [18] and µ → 3e[19] measurements. However, it
147
+ might be possible that the effects of leptoquarks could be cancelled in loop-induced processes
148
+ by the presence of other new particles. Studying direct leptoquark production at the LHC
149
+ allows us to directly probe the lepton-universal case because the observed number of events
150
+ in µµ, ee and µe channels will be correlated.
151
+ Our paper is structured as follows: we start by listing out the model Lagrangian and
152
+ the resulting Wilson coefficients for C9 in section II. We then examine the current LHC
153
+ constraints in various search channels in section III and expected detection prospects of
154
+ future colliders are calculated in section IV.
155
+
156
+ 5
157
+ II.
158
+ LEPTOQUARK MODELS
159
+ Leptoquarks are bosons which carry both SU(2)L and colour SU(3) charges and therefore
160
+ couple to both leptons and quarks. Given that we need to get the right contribution to
161
+ Cµ+µ−
162
+ 9
163
+ , this corresponds to a leptoquark that at a minimum couples to muons and to b and
164
+ s quarks. There are three known leptoquark models that give the right kind of contribu-
165
+ tion [12–14, 20], which we describe below. We use the standard names for the fields, viz. S3,
166
+ R2 and U1 and the numbers in brackets that follow correspond to (n-plet of SU(3), n-plet of
167
+ SU(2), U(1)Y hypercharge). Of these, S3 and R2 are scalar fields and U1 is a vector field.
168
+ A.
169
+ Scalar Leptoquark S3 (¯3, 3, 1/3)
170
+ The first leptoquark model we consider is S3(¯3, 3, 1/3) which is a SU(2)L triplet of scalar
171
+ leptoquark states with hypercharge 1/3. S3 is the only scalar leptoquark model that can
172
+ simultaneously predict Rexp
173
+ K∗ < RSM
174
+ K∗ and Rexp
175
+ K∗ < RSM
176
+ K∗ at tree level [21–24]. The Lagrangian
177
+ for the S3 model is
178
+ LS3 = yij
179
+ L ¯QC
180
+ i iτ2(τkSk
181
+ 3)Lj + h.c.,
182
+ (II.1)
183
+ where Qi and Lj are SU(2)L doublet fermion fields corresponding to quarks and leptons of
184
+ the ith( jth) generation respectively, τk are the generators of SU(2)L, and yij
185
+ L stands for a
186
+ Yukawa matrix for the left-handed fermions. The three triplet component states of S3 carry
187
+ charges Q = −2/3, 1/3 and 4/3 respectively. Expanding out the SU(2)L components and
188
+ referring to the leptoquarks as SQ
189
+ 3 , we get
190
+ LS3 = −yij
191
+ L ¯dC
192
+ LiνLjS1/3
193
+ 3
194
+
195
+
196
+ 2yij
197
+ L ¯dC
198
+ LiℓLjS4/3
199
+ 3
200
+ +
201
+
202
+ 2(V ∗yL)ij¯uC
203
+ LiνLjS−2/3
204
+ 3
205
+ − (V ∗yL)ij¯uC
206
+ LiℓLjS1/3
207
+ 3
208
+ + h.c.,
209
+ (II.2)
210
+ of which only the ¯dC
211
+ LiℓLjS4/3
212
+ 3
213
+ term contributes to O9. One can extract the Wilson coefficients
214
+ for the b → sl−l+ decay [12–14, 20],
215
+ Cℓ1ℓ2
216
+ 9
217
+ = −Cℓ1ℓ2
218
+ 10
219
+ =
220
+ πv2
221
+ VtbV ∗
222
+ tsαem
223
+ ybℓ1
224
+ L (ysℓ2
225
+ L )∗
226
+ m2
227
+ S3
228
+ ,
229
+ (II.3)
230
+
231
+ 6
232
+ B.
233
+ Scalar Leptoquark R2 (3, 2, 7/6)
234
+ The second case we consider is a weak doublet of scalar leptoquarks with hypercharge Y =
235
+ 7/6, i.e. R2 (3, 2, 7/6).[25] The most general Lagrangian describing the Yukawa interactions
236
+ with R2 can be written as,
237
+ LR2 = yij
238
+ R ¯QilRjR2 − yij
239
+ L ¯uRiR2iτ2Lj + h.c.,
240
+ (II.4)
241
+ where yL and yR are the Yukawa matrices corresponding to left- and right-handed lepton
242
+ fields respectively. In terms of the components with RQ
243
+ 2 denoting each leptoquark state with
244
+ charge Q, the Lagrangian can be written as
245
+ LR2 = (V yR)ij¯uLiℓRjR5/3
246
+ 2
247
+ + (yR)ij ¯dLiℓRjR2/3
248
+ 2
249
+ + (yL)ij¯uRiνLjR2/3
250
+ 2
251
+ − (yL)ij¯uRiℓLjR5/3
252
+ 2
253
+ + h.c.
254
+ (II.5)
255
+ The tree-level contribution to the Wilson coefficients C9 through the term (yR)ij ¯dLiℓRjR2/3
256
+ 2
257
+ amounts to
258
+ Cℓ1ℓ2
259
+ 9
260
+ = Cℓ1ℓ2
261
+ 10
262
+ = −
263
+ πv2
264
+ 2VtbV ∗
265
+ tsαem
266
+ ysℓ1
267
+ R (ybℓ2
268
+ R )∗
269
+ m2
270
+ R2
271
+ ,
272
+ (II.6)
273
+ C.
274
+ Vector Leptoquark U1 (3, 1, 2/3)
275
+ Finally, we describe the only vector leptoquark model considered in this paper, mainly
276
+ because it has been the only model that could simultaneously explain both charged current
277
+ and neutral current anomalies [12]. We consider the U1 (3, 1, 2/3) model which gives a single
278
+ leptoquark state with charge 2/3. The most general Lagrangian consistent with the SM
279
+ gauge symmetry allows couplings to both left-handed and right-handed fermions, namely
280
+ LU1 = βij
281
+ L ¯QiγµLjU µ
282
+ 1 + βij
283
+ R ¯dRiγµℓRjU µ
284
+ 1 + h.c.,
285
+ (II.7)
286
+ with couplings βij
287
+ L and βij
288
+ R. The contributions to the left-handed couplings to the effective
289
+ Lagrangian amount to
290
+ Cℓ1ℓ2
291
+ 9
292
+ = −Cℓ1ℓ2
293
+ 10
294
+ = −
295
+ πv2
296
+ VtbV ∗
297
+ tsαem
298
+ βsℓ1
299
+ L (βbℓ2
300
+ L )∗
301
+ m2
302
+ U1
303
+ ,
304
+ (II.8)
305
+
306
+ 7
307
+ III.
308
+ LHC LIMITS
309
+ Our goal is to use published LHC data to simultaneously constrain the mass and Yukawa
310
+ couplings of the leptoquarks. The Wilson coefficient C9 depends on three parameters roughly
311
+ as
312
+ Cℓ,ℓ
313
+ 9
314
+
315
+ �y2ℓ y3ℓ
316
+ M
317
+ �2
318
+ where yij refers to the leptoquark coupling between the ith generation of quark and jth
319
+ generation lepton. This corresponds to Yukawa couplings for S3 and R2 models and the gauge
320
+ coupling for the U1 model. Therefore, its possible to find a surface in the 3D parameter space
321
+ that gives the required value of C9. However, most LHC search constraints are in principle
322
+ only 2D — one coupling that determines the cross section of the final state and one mass.
323
+ We, therefore, have several options in which to view the full constraints.
324
+ Let us start with ℓ = 2 (i.e. µ) which contributes to Cµµ
325
+ 9 . To be able to independently
326
+ constrain the two Yukawa couplings y22 and y32, we study three different cases — first
327
+ setting only y22 non-zero (see figure 1, second setting only y32 non-zero (see figure 4) and
328
+ third, setting both equal (see figure 5). Using the upper limits from the non-resonant dimuon
329
+ search gives us an upper limit on y22 at each mass value. It is possible to also determine the
330
+ minimal allowed value of y22 that is consistent with C9 by requiring y32 ≤ 1.
331
+ Since the latest LHCb data seem to indicate that electrons and muons have identical
332
+ behaviour, we can indeed also do a similar exercise with y21 and y31 which would contribute
333
+ to Cee
334
+ 9 . Besides these, non-zero values of all four couplings (or even a single electron and
335
+ a single muon coupling) — y21, y31, y22 and y32 can give signatures that have differently
336
+ flavoured leptons in the final state, but without missing energy and therefore with no SM
337
+ background.
338
+ It should be noted that in the case where a single leptoquark state can couple to both
339
+ electrons and muons, the strongest constraints on couplings and mass of course come from
340
+ low energy processes in the µ → e sector [18, 19, 26]. However, it can still be an interesting
341
+ exercise to directly probe the case where both yk1 and yk2 are non-zero. As we see in figure 2,
342
+ this case is strongly constrained by the LHC, with the U1 model likely to be ruled out already
343
+ with full run-2 data of 139 fb−1.
344
+ Since multiple leptoquark states come from the same multiplet, they have identical mass
345
+ and switching on a single coupling allows the production of multiple states. For calculating
346
+
347
+ 8
348
+ the LHC limits, we allow the production of all leptoquark states and select only that fraction
349
+ that decays into the final state selected for by the analysis being reinterpreted. For example,
350
+ in the S3 case, if we look for pair production of leptoquark followed by decay of each into
351
+ a muon and a jet by turning on y22 ̸= 0 alone, we allow both the production of pairs of
352
+ S4/3
353
+ 3
354
+ → ¯sµ+ as well as pairs of S1/3
355
+ 3
356
+ → ¯cµ+. Our limits, therefore, are not identical to the
357
+ simplified model limits that the experimental analysis publishes by producing only one state
358
+ at a time, with 100% branching fraction into a certain channel. Similarly, when looking at
359
+ dilepton distributions, we take into account, with interference, all leptoquark states in the
360
+ t-channel that are allowed by non-zero couplings.
361
+ A.
362
+ Computational setup
363
+ Since we examine the limits from dilepton searches which have been presented in the form
364
+ of upper limits on generator-level cross sections with fiducial cuts, our computational setup
365
+ is much simplified. We generate events using
366
+ Madgraph5 amc@NLO [27] with the required
367
+ fiducial cuts and do not need to perform further detector simulation. This approach has
368
+ been proven to work well [28] and reproduces expected limits. For the UV models, we use
369
+ the scalar leptoquark models for S3 and R2 described in [29] and for the vector leptoquark
370
+ model for the U1 case, we use the model described in [30–32]. When more complicated
371
+ functionality is required, we use Pythia8 [33] to shower, hadronize and apply the required
372
+ kinematic cuts on events.
373
+ B.
374
+ Limits from resonant and non-resonant dilepton searches
375
+ We re-interpreted both the dilepton resonance search with 139 fb−1 [34] and the non-
376
+ resonant dilepton search at 139 fb−1 [35] from ATLAS. We find that the non-resonant search
377
+ results in much stronger limits and we continue with this search for the rest of our study.
378
+ The exclusive dilepton state can only be seen with a t-channel leptoquark exchange. It is
379
+ possible to have a dilepton plus two jets from strong production of leptoquarks, however,
380
+ this process does not depend on the leptoquark-fermion couplings and results in only a mass
381
+ limit which we deal with in the next subsection. With the interference of SM Drell-Yan
382
+ production of leptons with the t-channel leptoquark mediated production, one expects to
383
+
384
+ 9
385
+ 1000
386
+ 2000
387
+ 3000
388
+ 4000
389
+ 5000
390
+ 6000
391
+ 7000
392
+ 8000
393
+ 0.001
394
+ 0.005
395
+ 0.010
396
+ 0.050
397
+ 0.100
398
+ 0.500
399
+ 1
400
+ MLQ [GeV]
401
+ y2k
402
+ Allowed
403
+ S3
404
+ C9 ⇒ y3 k > 1
405
+ 1000
406
+ 2000
407
+ 3000
408
+ 4000
409
+ 5000
410
+ 6000
411
+ 7000
412
+ 8000
413
+ 0.001
414
+ 0.005
415
+ 0.010
416
+ 0.050
417
+ 0.100
418
+ 0.500
419
+ 1
420
+ MLQ [GeV]
421
+ y2k
422
+ Allowed
423
+ R2
424
+ C9 ⇒ y3 k > 1
425
+ 1000
426
+ 2000
427
+ 3000
428
+ 4000
429
+ 5000
430
+ 6000
431
+ 7000
432
+ 8000
433
+ 0.001
434
+ 0.005
435
+ 0.010
436
+ 0.050
437
+ 0.100
438
+ 0.500
439
+ 1
440
+ MLQ [GeV]
441
+ β2k
442
+ Allowed
443
+ U1
444
+ C9 ⇒ β3 k > 1
445
+ FIG. 1.
446
+ Exclusion plots y2ℓ versus Mass of leptoquark for the S3 (top-left), R2 (top-right) and
447
+ U1 models (bottom). The bright red regions at the top are disallowed from dimuon searches. The
448
+ corresponding di-electron limit is the lighter line inside the red region. The solid regions at the
449
+ bottom are from requiring perturbative couplings consistent with allowed C9. The vertical lines
450
+ are mass limits from direct leptoquark pair production with the solid line corresponding to second
451
+ generation leptons and the dotted corresponding to first generation. The limits correspond to 139
452
+ fb−1 data.
453
+ see a change in the shape of the dilepton invariant mass distribution mℓℓ where ℓ = µ or e.
454
+ We apply the limits from the ATLAS non-resonant dilepton search by generating events
455
+ using
456
+ Madgraph5 amc@NLO according to fiducial cuts listed in [35] and using the 95% upper
457
+ limits for the most conservative signal region called the “µ+µ− constructive signal region” (or
458
+
459
+ 10
460
+ analogously the e+e− constructive signal region). The constructive signal region corresponds
461
+ to the case where you expect signal events above the EW expectation, which is similar to
462
+ our case. The experimental analysis uses LO signal shape to model the expected number of
463
+ events and we therefore also do not use any NLO corrections. The upper limits are provided
464
+ on the additional cross section above the expected SM Electro-Weak (EW) prediction in the
465
+ cumulative signal region where mµ+µ− ≥ 2070 GeV (or me+e− ≥ 2200 GeV).
466
+ As expected, the effect of having heavy new leptoquarks in t-channel dies down when
467
+ either the leptoquark mass is too high or the Yukawa coupling is too small. To account for
468
+ the interference correctly, we use the difference of the cross-section pp → ℓ+ℓ− with both
469
+ leptoquark and EW bosons, and with only EW gauge bosons as our new physics contribution.
470
+ The result is an excluded region near high Yukawa coupling values, with a larger range ruled
471
+ out for smaller leptoquark masses. This is shown as a bright red region in figure 1. The
472
+ highest allowed value of y2k is referred to as y2k max and can be used to further restrict what
473
+ values of y3k are consistent with C9.
474
+ Currently, there is one different flavour dilepton search [36] performed at 13 TeV, but
475
+ with only 3.2 fb data analysed. Aside from cuts on pT of 65 and 50 GeV on electrons and
476
+ muons respectively, there are requirements that missing energy be less than 25 GeV and
477
+ mT < 50 GeV to remove contamination from W-boson production which we apply using
478
+ Pythia 8.3 [33]. The expected background for meµ > 2 TeV is 0.02 ± 0.02. They see one
479
+ event and interpreting it as a statistical fluctuation, set a limit on new physics cross section.
480
+ We extrapolate the expected limits from this search at 139 fb−1. The limits on the eµ case
481
+ for the U1 model can be seen in figure 2. The expected background at 139 fb−1 is 2.78
482
+ events, resulting in an expected 95% upper limit of 0.0185 fb on production cross section
483
+ times branching. As can be seen, the U1 model should be completely ruled out with 139
484
+ fb−1 data. For results in the eµ channel for S3 and R2 models, refer to appendix C.
485
+ C.
486
+ Limits from leptoquark-pair production
487
+ Direct limits on the mass of the leptoquark based on strong pair-production mode followed
488
+ by the decay of each leptoquark into a lepton and a jet are presented in [37]. The limits are
489
+ also presented on generator-level cross-section times branching fraction and can be applied
490
+ directly to our model. The resulting limit is shown as a solid black vertical line. Since there
491
+
492
+ 11
493
+ is no significant improvement in the limit from b-tagging, we use the general lepton+jet
494
+ limits in all cases. When only yk2 is non-zero, i.e. the leptoquark decays to a muon and a
495
+ jet, we obtain a mass limit for S3 leptoquark at 1774 GeV, for the R2 leptoquark at 1720
496
+ GeV and the U1 leptoquark at 2309 GeV. For the case where the leptoquark decays into
497
+ electron alone, we get a mass limit for S3 leptoquark at 1828 GeV, for the R2 leptoquark at
498
+ 1773 GeV and the U1 leptoquark at 2419 GeV.
499
+ There is no direct limit on the case with an eµ final state in the published search, which
500
+ if it existed, would give a far better exclusion simply because there is no irreducible SM
501
+ background and the dominant background would be from mis-identification of leptons.
502
+ D.
503
+ Missing search: top FCNC decay
504
+ Given the need for non-zero leptoquark coupling to the third generation of quarks, this
505
+ also implies a coupling between the top quark and second generation leptons for both the
506
+ S3 and U1 models. In the R2 case, the coupling is either CKM suppressed (in the case of
507
+ left-handed) or entirely independent and therefore set to zero (in the right-handed case). It
508
+ would therefore be possible to search directly for FCNC top decay via t → cµµ.
509
+ Currently, there are no searches for t → cµ+µ− except for a t → cZ search which requires
510
+ the dimuon mass to be within 15 GeV of the Z mass [38] and therefore is not directly
511
+ applicable to our model. A similar measurement from CMS [39] is available from the 8 TeV
512
+ run.
513
+ The main background for a t → cµ+µ− search is from the SM production of t¯tµ+µ−
514
+ via an off-shell Z or γ produced in association with t¯t. To remove contamination from on-
515
+ shell Z, we apply a cut instead Mℓℓ > 105 which is outside the Z-mass window selected for
516
+ by the t → cZ searches. Assuming the identification acceptances do not change, we can
517
+ estimate the background for our proposed search using the data driven estimate presented
518
+ in [38] (denoted by σBG,ATLAS).
519
+ Since we have identical SM production modes for t¯tZ
520
+ and t¯tµ+µ−, we assume that the generator level transfer factor between these processes is
521
+ transmitted all the way to the final selection. The kinematic effect of changing the mℓℓ cut
522
+ from |Mℓℓ − MZ| < 15 to Mℓℓ > 105 can be estimated at generator level and is encapsulated
523
+ in a single number fℓℓ Also, we assume that the enhancement in production of t¯tZ in going
524
+
525
+ 12
526
+ 1000
527
+ 2000
528
+ 3000
529
+ 4000
530
+ 5000
531
+ 6000
532
+ 7000
533
+ 8000
534
+ 0.001
535
+ 0.005
536
+ 0.010
537
+ 0.050
538
+ 0.100
539
+ 0.500
540
+ MLQ [GeV]
541
+ β2k
542
+ Allowed
543
+ U1
544
+ C9 ⇒ β3 K > 1
545
+ 1000
546
+ 2000
547
+ 3000
548
+ 4000
549
+ 5000
550
+ 6000
551
+ 7000
552
+ 8000
553
+ 0.001
554
+ 0.005
555
+ 0.010
556
+ 0.050
557
+ 0.100
558
+ 0.500
559
+ MLQ [GeV]
560
+ β3k
561
+ C9 ⇒ β2 k > 1
562
+ U1
563
+ C9 ⇒ β2 k > β2 k max
564
+ 1000
565
+ 2000
566
+ 3000
567
+ 4000
568
+ 5000
569
+ 6000
570
+ 7000
571
+ 8000
572
+ 0.001
573
+ 0.005
574
+ 0.010
575
+ 0.050
576
+ 0.100
577
+ 0.500
578
+ MLQ [GeV]
579
+ β2k,3k
580
+ C9 best fit
581
+ U1
582
+ FIG. 2.
583
+ Limits for the Leptoquark Couplings versus mass for the U1 Model. The dilepton process,
584
+ in this case, is pp → µe which does not exist in the SM. We, therefore, have strong limits even
585
+ with 3.2 fb−1 data as published in [36]. The top-left panel shows limits on the coupling to second
586
+ generation quarks with y22 = y21, the top-right panel on the coupling to third generation quarks
587
+ with y32 = y31 and the bottom panel shows the case where all four couplings are equal. The green
588
+ band shows the values corresponding to the best fit values of C9 The dotted line in this figure
589
+ shows the expected limit after analysing full 139 fb−1 of run-2 data by ATLAS (only partial result
590
+ is published so far). We see clearly that the universal scenario is likely already ruled out by run-2
591
+ data.
592
+
593
+ 13
594
+ from 13 TeV to 13.6 TeV (fE = σ13.6
595
+ σ13 ) remains the same also for t¯tµ+µ−. Thus we have
596
+ σBG(√s = 13.6) =
597
+ σBG,ATLAS
598
+ × fE × fℓℓ
599
+ × σ(pp → t¯tµ+µ−; √s = 13)
600
+ σ(pp → t¯tZ; √s = 13)
601
+ (III.1)
602
+ Using this, and the expected background cross section from ATLAS, we calculate an
603
+ expected background of 7±2 events. Given that with the Z-window, the background is esti-
604
+ mated at 119±10 events, this would correspond to over an order of magnitude improvement
605
+ in the sensitivity to FCNC branching fraction of the top quark.
606
+ IV.
607
+ FUTURE PROSPECTS
608
+ The best-fit value of the Wilson coefficients for operators that explain the b → s anomalies
609
+ suggests a high suppression scale. Using equations (II.6), (II.3) and (II.8), we find that the
610
+ required scale for both couplings set to one is 16183 GeV for the R2 case and 22887 GeV
611
+ for the S3 and U1 cases. Naturally, resonantly producing a leptoquark of this mass scale is
612
+ out of the question at the LHC. We, therefore, investigate both the expected reach of the
613
+ LHC after the planned high-luminosity run and estimate a conservative reach for a muon
614
+ collider with CM energy of 3 TeV [40–43]. To illustrate the highest sensitivity case, we
615
+ choose y22 = y32 for this calculation. This also allows us to make a comment on the ability
616
+ of the collider to explore the entire parameter space of interest. A summary of the expected
617
+ reach of future colliders can be seen in figure 3
618
+ A.
619
+ LHC High-Lumi expected limits
620
+ Projections for the HL-LHC are made with the luminosity of 3000 fb−1. From previous
621
+ experience, we know that the improvements in limits scale with about the square root of
622
+ luminosity. Using the expected number of signal and background events for the non-resonant
623
+ dilepton search, we can probe effects of leptoquarks up to mass 5 TeV for the S3, 3 TeV for
624
+ the R2 and 9.5 TeV for the U1 model. Conversely, we can probe coupling values as small as
625
+ 0.4 for S3, 0.55 for R2 and 0.15 for U1 models respectively at 1 TeV leptoquark mass. For
626
+
627
+ 14
628
+ 1000
629
+ 2000
630
+ 3000
631
+ 4000
632
+ 5000
633
+ 6000
634
+ 7000
635
+ 8000
636
+ 0.0
637
+ 0.2
638
+ 0.4
639
+ 0.6
640
+ 0.8
641
+ 1.0
642
+ MLQ [GeV]
643
+ y22,32
644
+ LHC13
645
+ MuonC, 1/fb
646
+
647
+
648
+ C9
649
+ fit
650
+ S3
651
+ LHC13-Mass
652
+ 1000
653
+ 2000
654
+ 3000
655
+ 4000
656
+ 5000
657
+ 6000
658
+ 7000
659
+ 8000
660
+ 0.0
661
+ 0.2
662
+ 0.4
663
+ 0.6
664
+ 0.8
665
+ 1.0
666
+ MLQ [GeV]
667
+ y22,32
668
+ LHC13
669
+ MuonC, 1/fb
670
+
671
+
672
+ C9
673
+ fit
674
+ R2
675
+ LHC13-Mass
676
+ 1000
677
+ 2000
678
+ 3000
679
+ 4000
680
+ 5000
681
+ 6000
682
+ 7000
683
+ 8000
684
+ 0.0
685
+ 0.2
686
+ 0.4
687
+ 0.6
688
+ 0.8
689
+ 1.0
690
+ MLQ [GeV]
691
+ β22,32
692
+ LHC13
693
+ MuonC, 1/fb
694
+
695
+
696
+ C9
697
+ fit
698
+ U1
699
+ LHC13-Mass
700
+ FIG. 3. Current and future reach in leptoquark coupling to muons with leptoquark mass for the S3
701
+ Model (top-left), R2 Model (top-right) and U1 Model (bottom). The green region corresponds to
702
+ the 1σ region given by global fit C9 values in the model-dependent case whereas the yellow is the
703
+ data-driven 1σ region ([7], also see table I). The solid red region is the current 139 fb−1 limits with
704
+ the dotted red line the expected reach after 3 ab−1 at the HL-LHC. The solid and dotted vertical
705
+ lines correspond to mass limits from pair production again corresponding to the 139 fb−1 and 3
706
+ ab−1 luminosity respectively. The blue region corresponds to the parameter space that can be
707
+ discovered with a 5σ significance at a 3 TeV muon collider with 1 fb−1 whereas the orange region
708
+ corresponds to the further region that can be probed at 95% confidence at the same collider. The
709
+ U1 model can be fully excluded with just 1 fb−1 data. The S3 and R2 models can also be fully
710
+ probed with 6.5fb−1 and 5fb−1 respectively.
711
+ comparison, C9 best fit predicts a minimum value of coupling at 0.04, 0.06 and 0.04 for the
712
+ three models when we set both couplings equal.
713
+ The direct search limits from strong production are calculated in a similar way using the
714
+
715
+ 15
716
+ published upper limits at 139/fb. We find that the HL-LHC can exclude leptoquark masses
717
+ of 2.2 TeV for both the S3 and R2 case and 2.8 TeV for the U1 case for the leptoquark
718
+ decaying into a muon and a jet and 2.3 TeV for both the S3 and R2 case and 2.9 TeV for
719
+ the U1 case for the leptoquark decaying into an electron and a jet.
720
+ B.
721
+ Reach of a Future Muon Collider
722
+ Estimating the reach of a future muon collider is more difficult since we do not currently
723
+ have a detector configuration to be able to simulate a realistic analysis. However, taking
724
+ lessons from the dilepton and dijet searches at the LHC, we know that a single-bin analysis
725
+ with a high enough cut on the invariant mass provides a very reliable estimate of reach. We
726
+ look at µ+µ− → jj as our signal. Obviously using b-tagging will be a further improvement
727
+ that can pinpoint the underlying scenario. However, for this estimate, we just use untagged
728
+ jets. Given that acceptance efficiencies of jets are expected to be similar for both signal and
729
+ background events for a simple dijet search, we proceed with using just generator-level cross
730
+ sections. A further advantage is the much reduced probability of extra initial state radiation
731
+ jets from initial state muons (in sharp contrast to a pp machine).
732
+ The main background from the SM comes from s-channel photon or Z exchange. In the
733
+ presence of the leptoquark, another Feynman diagram with a t-channel leptoquark exchange
734
+ needs to be taken into account. We look only at events with Mjj > 500 GeV. The SM-only
735
+ cross section at LO is 5.96×10−2 pb which corresponds to a statistical error of about 8 events
736
+ at a luminosity of 1 fb−1. Using this, we can calculate the parameter space corresponding to
737
+ a 5σ discovery as well as regions that can be excluded at 2σ. They are shown in figure 3 as
738
+ blue and orange regions respectively. In the U1 case, we see that a muon collider is capable
739
+ of excluding the entire viable parameter space with 1 fb−1. To exclude the R2 and S3 models
740
+ would need a luminosity of 6.5 fb−1 for S3 and 5 fb−1 for R2.
741
+ V.
742
+ SUMMARY AND CONCLUSIONS
743
+ We examine the limits from direct collider searches on leptoquark models that are capable
744
+ of explaining the anomalous measurements in the decays of B-mesons. We focus on three
745
+ specific models — two scalar leptoquark models S3 and R2 and one vector leptoquark model
746
+
747
+ 16
748
+ U1. Aside from limits on the mass of the leptoquarks (which can be pair-produced by strong
749
+ interactions), it is possible to also constrain the couplings to fermions by looking at changes
750
+ to the shape of the dilepton mass spectrum. Reinterpreting full Run-2 limits from the pair
751
+ production and non-resonant dilepton searches by ATLAS experiment, we find that current
752
+ mass limits are 1.77 TeV, 1.72 TeV and 2.3 TeV respectively for the three models. We can
753
+ expect to reach up to 2.2 TeV for S3 and R2 and 2.8 TeV for the U1 respectively with the
754
+ High-Luminosity LHC run.
755
+ Effects of leptoquarks with couplings to muons can potentially be probed in a muon
756
+ collider. Since there has been considerable interest in a future muon collider recently, we
757
+ also estimate what the reach of the proposed 3 TeV muon collider would be for the three
758
+ models in question. We find that with very minimal assumptions, S3, R2 and U1 models
759
+ show significant deviation in dijet distributions that can be observable for the entire range
760
+ of interest with less than 6 fb−1 data for all three models.
761
+ ACKNOWLEDGEMENTS
762
+ ND is supported by the Ramanujan Fellowship grant SB/S2/RJN-070 from the Department
763
+ of Science and Technology of the Government of India.
764
+ [1] LHCb collaboration, Test of lepton universality in b → sℓ+ℓ− decays, 2212.09152.
765
+ [2] LHCb collaboration, Measurement of lepton universality parameters in B+ → K+ℓ+ℓ− and
766
+ B0 → K∗0ℓ+ℓ− decays, 2212.09153.
767
+ [3] ATLAS collaboration, Study of the rare decays of B0
768
+ s and B0 mesons into muon pairs using
769
+ data collected during 2015 and 2016 with the ATLAS detector, JHEP 04 (2019) 098
770
+ [1812.03017].
771
+ [4] CMS collaboration, Measurement of properties of B0
772
+ s → µ+µ− decays and search for
773
+ B0 → µ+µ− with the CMS experiment, JHEP 04 (2020) 188 [1910.12127].
774
+ [5] LHCb collaboration, Angular Analysis of the B+ → K∗+µ+µ− Decay, Phys. Rev. Lett. 126
775
+ (2021) 161802 [2012.13241].
776
+
777
+ 17
778
+ [6] N. Gubernari, M. Reboud, D. van Dyk and J. Virto, Improved theory predictions and global
779
+ analysis of exclusive b → sµ+µ− processes, JHEP 09 (2022) 133 [2206.03797].
780
+ [7] M. Ciuchini, M. Fedele, E. Franco, A. Paul, L. Silvestrini and M. Valli, Constraints on
781
+ Lepton Universality Violation from Rare B Decays, 2212.10516.
782
+ [8] M. Ciuchini, A. M. Coutinho, M. Fedele, E. Franco, A. Paul, L. Silvestrini et al., New
783
+ Physics in b → sℓ+ℓ− confronts new data on Lepton Universality, Eur. Phys. J. C 79 (2019)
784
+ 719 [1903.09632].
785
+ [9] M. Alguer´o, B. Capdevila, A. Crivellin, S. Descotes-Genon, P. Masjuan, J. Matias et al.,
786
+ Emerging patterns of New Physics with and without Lepton Flavour Universal contributions,
787
+ Eur. Phys. J. C 79 (2019) 714 [1903.09578].
788
+ [10] A. K. Alok, A. Dighe, S. Gangal and D. Kumar, Continuing search for new physics in
789
+ b → sµµ decays: two operators at a time, JHEP 06 (2019) 089 [1903.09617].
790
+ [11] T. Hurth, F. Mahmoudi, D. M. Santos and S. Neshatpour, More Indications for Lepton
791
+ Nonuniversality in b → sℓ+ℓ−, Phys. Lett. B 824 (2022) 136838 [2104.10058].
792
+ [12] A. Angelescu, D. Beˇcirevi´c, D. A. Faroughy and O. Sumensari, Closing the window on single
793
+ leptoquark solutions to the B-physics anomalies, JHEP 10 (2018) 183 [1808.08179].
794
+ [13] D. Beˇcirevi´c, N. Koˇsnik, O. Sumensari and R. Zukanovich Funchal, Palatable Leptoquark
795
+ Scenarios for Lepton Flavor Violation in Exclusive b → sℓ1ℓ2 modes, JHEP 11 (2016) 035
796
+ [1608.07583].
797
+ [14] S. Descotes-Genon, L. Hofer, J. Matias and J. Virto, Global analysis of b → sℓℓ anomalies,
798
+ JHEP 06 (2016) 092 [1510.04239].
799
+ [15] C. Cornella, J. Fuentes-Martin and G. Isidori, Revisiting the vector leptoquark explanation of
800
+ the B-physics anomalies, JHEP 07 (2019) 168 [1903.11517].
801
+ [16] J. Aebischer, W. Altmannshofer, D. Guadagnoli, M. Reboud, P. Stangl and D. M. Straub,
802
+ B-decay discrepancies after Moriond 2019, Eur. Phys. J. C 80 (2020) 252 [1903.10434].
803
+ [17]
804
+ [18] MEG collaboration, New constraint on the existence of the µ+ → e+γ decay, Phys. Rev.
805
+ Lett. 110 (2013) 201801 [1303.0754].
806
+ [19] SINDRUM collaboration, Search for the Decay mu+ —> e+ e+ e-, Nucl. Phys. B 299
807
+ (1988) 1.
808
+
809
+ 18
810
+ [20] A. Angelescu, D. Beˇcirevi´c, D. A. Faroughy, F. Jaffredo and O. Sumensari, Single leptoquark
811
+ solutions to the B-physics anomalies, Phys. Rev. D 104 (2021) 055017 [2103.12504].
812
+ [21] I. Dorˇsner, S. Fajfer, D. A. Faroughy and N. Koˇsnik, The role of the S3 GUT leptoquark in
813
+ flavor universality and collider searches, JHEP 10 (2017) 188 [1706.07779].
814
+ [22] G. Hiller and M. Schmaltz, RK and future b → sℓℓ physics beyond the standard model
815
+ opportunities, Phys. Rev. D 90 (2014) 054014 [1408.1627].
816
+ [23] G. Hiller and I. Nisandzic, RK and RK∗ beyond the standard model, Phys. Rev. D 96 (2017)
817
+ 035003 [1704.05444].
818
+ [24] C. Hati, G. Kumar, J. Orloff and A. M. Teixeira, Reconciling B-meson decay anomalies with
819
+ neutrino masses, dark matter and constraints from flavour violation, JHEP 11 (2018) 011
820
+ [1806.10146].
821
+ [25]
822
+ [26] SINDRUM II collaboration, Test of lepton flavor conservation in mu —> e conversion on
823
+ titanium, Phys. Lett. B 317 (1993) 631.
824
+ [27] J. Alwall, M. Herquet, F. Maltoni, O. Mattelaer and T. Stelzer, MadGraph 5 : Going
825
+ Beyond, JHEP 06 (2011) 128 [1106.0522].
826
+ [28] D. Bhatia, N. Desai and A. Dighe, Frugal U(1)X models with non-minimal flavor violation
827
+ for b → sℓℓ anomalies and neutrino mixing, JHEP 04 (2022) 163 [2109.07093].
828
+ [29] I. Dorˇsner and A. Greljo, Leptoquark toolbox for precision collider studies, JHEP 05 (2018)
829
+ 126 [1801.07641].
830
+ [30] M. J. Baker, J. Fuentes-Mart´ın, G. Isidori and M. K¨onig, High- pT signatures in
831
+ vector–leptoquark models, Eur. Phys. J. C 79 (2019) 334 [1901.10480].
832
+ [31] L. Di Luzio, J. Fuentes-Martin, A. Greljo, M. Nardecchia and S. Renner, Maximal Flavour
833
+ Violation: a Cabibbo mechanism for leptoquarks, JHEP 11 (2018) 081 [1808.00942].
834
+ [32] C. Cornella, D. A. Faroughy, J. Fuentes-Martin, G. Isidori and M. Neubert, Reading the
835
+ footprints of the B-meson flavor anomalies, JHEP 08 (2021) 050 [2103.16558].
836
+ [33] C. Bierlich et al., A comprehensive guide to the physics and usage of PYTHIA 8.3,
837
+ 2203.11601.
838
+ [34] ATLAS collaboration, Search for high-mass dilepton resonances using 139 fb−1 of pp
839
+ collision data collected at √s =13 TeV with the ATLAS detector, Phys. Lett. B 796 (2019)
840
+ 68 [1903.06248].
841
+
842
+ 19
843
+ [35] ATLAS collaboration, Search for new non-resonant phenomena in high-mass dilepton final
844
+ states with the ATLAS detector, JHEP 11 (2020) 005 [2006.12946].
845
+ [36] ATLAS collaboration, Search for new phenomena in different-flavour high-mass dilepton
846
+ final states in pp collisions at √s = 13 Tev with the ATLAS detector, Eur. Phys. J. C 76
847
+ (2016) 541 [1607.08079].
848
+ [37] ATLAS collaboration, Search for pairs of scalar leptoquarks decaying into quarks and
849
+ electrons or muons in √s = 13 TeV pp collisions with the ATLAS detector, JHEP 10 (2020)
850
+ 112 [2006.05872].
851
+ [38] ATLAS collaboration, Search for flavour-changing neutral current top-quark decays t → qZ
852
+ in proton-proton collisions at √s = 13 TeV with the ATLAS detector, JHEP 07 (2018) 176
853
+ [1803.09923].
854
+ [39] CMS collaboration, Search for associated production of a Z boson with a single top quark
855
+ and for tZ flavour-changing interactions in pp collisions at √s = 8 TeV, JHEP 07 (2017)
856
+ 003 [1702.01404].
857
+ [40] R. Palmer et al., Muon collider design, Nucl. Phys. B Proc. Suppl. 51 (1996) 61
858
+ [acc-phys/9604001].
859
+ [41] R. B. Palmer, Muon Colliders, Rev. Accel. Sci. Tech. 7 (2014) 137.
860
+ [42] C. M. Ankenbrandt et al., Status of muon collider research and development and future
861
+ plans, Phys. Rev. ST Accel. Beams 2 (1999) 081001 [physics/9901022].
862
+ [43] J. C. Gallardo et al., µ+µ− Collider: Feasibility Study, eConf C960625 (1996) R4.
863
+ [44] M. Bordone, G. Isidori and A. Pattori, On the Standard Model predictions for RK and RK∗,
864
+ Eur. Phys. J. C 76 (2016) 440 [1605.07633].
865
+ [45] LHCb collaboration, Test of lepton universality with B0 → K∗0ℓ+ℓ− decays, JHEP 08
866
+ (2017) 055 [1705.05802].
867
+ [46] LHCb collaboration, Test of lepton universality in beauty-quark decays, Nature Phys. 18
868
+ (2022) 277 [2103.11769].
869
+ [47] M. Beneke, C. Bobeth and R. Szafron, Power-enhanced leading-logarithmic QED corrections
870
+ to Bq → µ+µ−, JHEP 10 (2019) 232 [1908.07011].
871
+ [48] LHCb collaboration, Measurement of Form-Factor-Independent Observables in the Decay
872
+ B0 → K∗0µ+µ−, Phys. Rev. Lett. 111 (2013) 191801 [1308.1707].
873
+
874
+ 20
875
+ [49] LHCb collaboration, Angular analysis of the B0 → K∗0µ+µ− decay using 3 fb−1 of
876
+ integrated luminosity, JHEP 02 (2016) 104 [1512.04442].
877
+ [50] S. Descotes-Genon, T. Hurth, J. Matias and J. Virto, Optimizing the basis of B → K∗ll
878
+ observables in the full kinematic range, JHEP 05 (2013) 137 [1303.5794].
879
+
880
+ 21
881
+ Appendix A: Relevant observables in the b → s sector
882
+ Observable
883
+ Experiment
884
+ Theory (SM)
885
+ RK[0.1,1.1]
886
+ 0.994 +0.090
887
+ −0.082 (stat) +0.029
888
+ −0.027 (syst) [2022] [1, 2]
889
+ 1.00 ± 0.01 [44])
890
+ RK∗[0.1,1.1]
891
+ 0.927 +0.093
892
+ −0.087 (stat) +0.036
893
+ −0.035 (syst) [2022] [1, 2]
894
+ 1.00 ± 0.01 [44])
895
+ RK[1.1,6]
896
+ 0.949 +0.042
897
+ −0.041 (stat) +0.022
898
+ −0.022 (syst) [2022] [1, 2]
899
+ 1.00 ± 0.01 [44])
900
+ RK∗[1.1,6]
901
+ 1.027 +0.072
902
+ −0.068 (stat) +0.027
903
+ −0.026 (syst) [2022] [1, 2]
904
+ 1.00 ± 0.01 [44])
905
+ R[0.045,1.1]
906
+ K∗
907
+ 0.66+0.11
908
+ −0.07 ± 0.03 [2021] [45]
909
+ 0.906 ± 0.028 [44]
910
+ R[1.1,6.0]
911
+ K∗
912
+ 0.69+0.11
913
+ −0.07 ± 0.05 [2021] [45]
914
+ 1.00 ± 0.01 [44]
915
+ R[1.1,6.0]
916
+ K
917
+ 0.846+0.042+0.013
918
+ −0.039−0.012 [2021] [46]
919
+ 1.00 ± 0.01 [44]
920
+ B(Bs → µ+µ−)
921
+ (2.85+0.32
922
+ −0.31) × 10−9 [3, 4]
923
+ (3.66 ± 0.14) × 10−9[47])
924
+ P ′
925
+ 5 in B → K(∗) l+ l−
926
+ [5, 48, 49]
927
+ [6, 50]
928
+ TABLE II. A summary of the most relevant experimental results and SM predictions for the
929
+ observables in b → s sector.
930
+
931
+ 22
932
+ Appendix B: Limits on leptoquark couplings to third generation quarks y3k.
933
+ 1000
934
+ 2000
935
+ 3000
936
+ 4000
937
+ 5000
938
+ 6000
939
+ 7000
940
+ 8000
941
+ 0.005
942
+ 0.010
943
+ 0.050
944
+ 0.100
945
+ 0.500
946
+ 1
947
+ MLQ [GeV]
948
+ y3k
949
+ C9 ⇒ y2 k > 1
950
+ S3
951
+ C9 ⇒ y2 k > y2 k max
952
+ 1000
953
+ 2000
954
+ 3000
955
+ 4000
956
+ 5000
957
+ 6000
958
+ 7000
959
+ 8000
960
+ 0.001
961
+ 0.005
962
+ 0.010
963
+ 0.050
964
+ 0.100
965
+ 0.500
966
+ 1
967
+ MLQ [GeV]
968
+ y3k
969
+ C9 ⇒ y2 K > 1
970
+ R2
971
+ C9 ⇒ y2 K > y2 K max
972
+ 1000
973
+ 2000
974
+ 3000
975
+ 4000
976
+ 5000
977
+ 6000
978
+ 7000
979
+ 8000
980
+ 0.001
981
+ 0.005
982
+ 0.010
983
+ 0.050
984
+ 0.100
985
+ 0.500
986
+ 1
987
+ MLQ [GeV]
988
+ β3k
989
+ C9 ⇒ β2 k > β2 k max
990
+ U1
991
+ FIG. 4.
992
+ Exclusion plots y3ℓ versus Mass of leptoquark for the S3 (top-left), R2 (top-right) and
993
+ U1 models (bottom). The solid regions at the bottom are from requiring perturbative couplings
994
+ consistent with allowed C9. The darker region is inconsistent with the observed upper limits on y2k
995
+ in figure 1. The vertical lines are mass limits from direct leptoquark pair production with the solid
996
+ line corresponding to second generation leptons and the dotted corresponding to first generation.
997
+ The limits correspond to 139 fb−1 data.
998
+
999
+ 23
1000
+ 1000
1001
+ 2000
1002
+ 3000
1003
+ 4000
1004
+ 5000
1005
+ 6000
1006
+ 7000
1007
+ 8000
1008
+ 0.001
1009
+ 0.005
1010
+ 0.010
1011
+ 0.050
1012
+ 0.100
1013
+ 0.500
1014
+ 1
1015
+ MLQ [GeV]
1016
+ y2k,3k
1017
+ C9 best fit
1018
+ S3
1019
+ 1000
1020
+ 2000
1021
+ 3000
1022
+ 4000
1023
+ 5000
1024
+ 6000
1025
+ 7000
1026
+ 8000
1027
+ 0.001
1028
+ 0.005
1029
+ 0.010
1030
+ 0.050
1031
+ 0.100
1032
+ 0.500
1033
+ 1
1034
+ MLQ [GeV]
1035
+ y2k,3k
1036
+ C9 best fit
1037
+ R2
1038
+ 1000
1039
+ 2000
1040
+ 3000
1041
+ 4000
1042
+ 5000
1043
+ 6000
1044
+ 7000
1045
+ 8000
1046
+ 0.001
1047
+ 0.005
1048
+ 0.010
1049
+ 0.050
1050
+ 0.100
1051
+ 0.500
1052
+ 1
1053
+ MLQ [GeV]
1054
+ β2k,3k
1055
+ C9 best fit
1056
+ U1
1057
+ FIG. 5.
1058
+ Exclusion plots in the limited case of y2ℓ = y3ℓ versus Mass of leptoquark for the S3
1059
+ (top-left), R2 (top-right) and U1 models (bottom).
1060
+ The solid red region at the top are limits
1061
+ from non-resonant dilepton searches in µ+µ−. The lighter lines inside this region correspond to
1062
+ subleading limits from the similar e+e− search.
1063
+ The vertical lines are mass limits from direct
1064
+ leptoquark pair production with the solid line corresponding to second generation leptons and the
1065
+ dotted corresponding to first generation. The limits correspond to 139 fb−1 data. The green band
1066
+ is the region that corresponds to the coefficient C9 within one sigma of best fit to data.
1067
+
1068
+
1069
+ 24
1070
+ Appendix C: Limits on S3 and R2 model parameters in the Lepton Flavour Universal
1071
+ case
1072
+ 1000
1073
+ 2000
1074
+ 3000
1075
+ 4000
1076
+ 5000
1077
+ 6000
1078
+ 7000
1079
+ 8000
1080
+ 0.001
1081
+ 0.005
1082
+ 0.010
1083
+ 0.050
1084
+ 0.100
1085
+ 0.500
1086
+ 1
1087
+ MLQ [GeV]
1088
+ y2k
1089
+ Allowed
1090
+ S3
1091
+ C9 ⇒ y3 K > 1
1092
+ 1000
1093
+ 2000
1094
+ 3000
1095
+ 4000
1096
+ 5000
1097
+ 6000
1098
+ 7000
1099
+ 8000
1100
+ 0.005
1101
+ 0.010
1102
+ 0.050
1103
+ 0.100
1104
+ 0.500
1105
+ 1
1106
+ MLQ [GeV]
1107
+ y3k
1108
+ C9 ⇒ y2 k > 1
1109
+ S3
1110
+ C9 ⇒ y2 k > y2 k max
1111
+ 1000
1112
+ 2000
1113
+ 3000
1114
+ 4000
1115
+ 5000
1116
+ 6000
1117
+ 7000
1118
+ 8000
1119
+ 0.001
1120
+ 0.005
1121
+ 0.010
1122
+ 0.050
1123
+ 0.100
1124
+ 0.500
1125
+ 1
1126
+ MLQ [GeV]
1127
+ y2k,3k
1128
+ C9 best fit
1129
+ S3
1130
+ FIG. 6. Limits on the leptoquark couplings via the process p p → µ e in the case of flavour universal
1131
+ couplings to electrons and muons for the S3 Model.
1132
+
1133
+ 25
1134
+ 1000
1135
+ 2000
1136
+ 3000
1137
+ 4000
1138
+ 5000
1139
+ 6000
1140
+ 7000
1141
+ 8000
1142
+ 0.001
1143
+ 0.005
1144
+ 0.010
1145
+ 0.050
1146
+ 0.100
1147
+ 0.500
1148
+ 1
1149
+ MLQ [GeV]
1150
+ y2k
1151
+ Allowed
1152
+ R2
1153
+ C9 ⇒ y3 K > 1
1154
+ 1000
1155
+ 2000
1156
+ 3000
1157
+ 4000
1158
+ 5000
1159
+ 6000
1160
+ 7000
1161
+ 8000
1162
+ 0.001
1163
+ 0.005
1164
+ 0.010
1165
+ 0.050
1166
+ 0.100
1167
+ 0.500
1168
+ 1
1169
+ MLQ [GeV]
1170
+ y3k
1171
+ C9 ⇒ y2 k > 1
1172
+ R2
1173
+ C9 ⇒ y2 k > y2 k max
1174
+ 1000
1175
+ 2000
1176
+ 3000
1177
+ 4000
1178
+ 5000
1179
+ 6000
1180
+ 7000
1181
+ 8000
1182
+ 0.001
1183
+ 0.005
1184
+ 0.010
1185
+ 0.050
1186
+ 0.100
1187
+ 0.500
1188
+ 1
1189
+ MLQ [GeV]
1190
+ y2k,3k
1191
+ C9 best fit
1192
+ R2
1193
+ FIG. 7. Limits on the leptoquark couplings via the process p p → µ e in the case of flavour universal
1194
+ couplings to electrons and muons for the R2 Model.
1195
+
KtAzT4oBgHgl3EQfyf5z/content/tmp_files/load_file.txt ADDED
The diff for this file is too large to render. See raw diff
 
LNE3T4oBgHgl3EQfvwtG/content/2301.04696v1.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a536b246e30d62009922c1d5a97dc783071189fc9b3adcd1811dcecd9b927173
3
+ size 823235
LNE3T4oBgHgl3EQfvwtG/vector_store/index.faiss ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:50a6fbcc18d3b0aafb87b54f070914d406b481df91750fa5dab737dc0e6078fb
3
+ size 2228269