astro-legacy-archive/white-mountain

White Mountain absolute CMB spectrum

This dataset contains all four scientific tables published on LAMBDA's White Mountain Collaboration data-product page: radiometer properties, cold-load components, representative measurements, and final CMB temperatures. The source publishes the tables directly in HTML; it links no separate measurement file.

The White Mountain Collaboration measured the absolute difference between the antenna temperature of the zenith sky and a liquid-helium-temperature blackbody at 2.5, 4.75, 10, 33, and 90 GHz. Observations were made in 1982 and 1983 at the University of California White Mountain Research Station. A separate 9.4 GHz radiometer measured atmospheric emission.

How to use

Install datasets, huggingface_hub, and pyarrow, then load a configuration by its name:

python -m venv .venv && .venv/bin/pip install datasets huggingface_hub pyarrow

from datasets import load_dataset
dataset = load_dataset("astro-legacy-archive/white-mountain", "final_results", split="train[:2]", columns=["Frequency (GHz)", "Combined 1982 and 1983 results (K)"])
for row in dataset:
    print(row["Frequency (GHz)"], row["Combined 1982 and 1983 results (K)"])

2.5 2.79±0.15 Ref.1
4.75 2.70±0.07 Ref.2

The flattened source values, exact caption text and parser-normalized cell-inner HTML, field semantics, units, marker notes, bibliography, and provenance are embedded in every Parquet schema:

import json
from huggingface_hub import hf_hub_download
import pyarrow.parquet as pq
path = hf_hub_download("astro-legacy-archive/white-mountain", "final_results.parquet", repo_type="dataset")
print(json.loads(pq.read_metadata(path).metadata[b"statistical_product"])["schema"])

astro-legacy-archive/statistical-product

Configurations and fields

All columns have Arrow string type, and every row and column remains in source order. Field identifiers below are the exact text obtained by joining the source HTML cell's stripped text nodes with spaces; HTML character references are decoded by the parser.

radiometer_properties contains six rows and five columns:

• Frequency (GHz) — observing frequency, in GHz.
• System bandwidth (GHz) — radiometer system bandwidth, in GHz.
• Antenna Half-power beam width (deg) — antenna half-power beam width, in degrees.
• System noise (K/Hz 1/2 ) — radiometer system noise per square-root bandwidth, in K/Hz^1/2.
• Laboratory (Ref) — laboratory and source bibliography marker for the radiometer; the source header supplies no scalar unit.

cold_load_components contains five rows and six columns:

• Frequency (GHz) — observing frequency, in GHz.
• Eccosorb (o) (mK) — Eccosorb cold-load antenna-temperature component, in mK.
• Windows + gas (mK) — cold-load antenna-temperature contribution from windows and gas, in mK.
• Dewar walls (mK) — cold-load antenna-temperature contribution from Dewar walls, in mK.
• Reflected signals (mK) — cold-load antenna-temperature contribution from reflected signals, in mK.
• Total (mK) — total cold-load antenna temperature, in mK.

representative_measurements contains six rows and eight columns:

• Frequency (GHz) — observing frequency, in GHz.
• T A,load (mK) — cold-load antenna temperature, in mK.
• ΔT (sky-load) (mK) — antenna-temperature difference between sky and cold load, in mK.
• ΔT offset (mK) — published antenna-temperature offset, in mK.
• T ground (mK) — ground-emission antenna-temperature contribution, in mK.
• T atmosphere (mK) — atmospheric-emission antenna-temperature contribution, in mK.
• T Galaxy (mK) — Galactic-emission antenna-temperature contribution, in mK.
• Year — observation year corresponding by position to the paired source values; the source header supplies no scalar unit.

final_results contains five rows and five columns:

• Frequency (GHz) — observing frequency, in GHz.
• Number of observations — numbers of observations for 1982 and 1983 in source order; the source header supplies no scalar unit.
• CMB antenna temperature (K) — CMB antenna temperature for 1982 and 1983 in source order, in K.
• CMB thermodynamic temperature (K) — CMB thermodynamic temperature for 1982 and 1983 in source order, in K.
• Combined 1982 and 1983 results (K) — combined 1982 and 1983 CMB thermodynamic-temperature result and bibliography marker, in K.

Decoding the source text

Composite cells are retained rather than decomposed. In the cold-load table, (a) means the 1982 value and (b) the 1983 value; (o) records the note (o) Eccosorb/Liquid Helium physical temperature 3776±4 mK,. In the representative measurements, paired entries correspond by position to Year, whose flattened cell is 1982 1983; published_cell_html retains its <br> boundary in parser-normalized inner HTML. The source note is (a)included in T A,load. The final-results rule is recorded verbatim as For Columns 2-4, the first row is for 1982 data and the second row is for 1983 data. The last column is the combined result followed by its Ref. marker.

The strings preserve ±, reference markers, the missing-value text --, the atmospheric-radiometer marker (*), and the source's exponent typography. The source HTML carries <30±10 and <10±5 as bare, unescaped inequality strings. The declared BeautifulSoup html.parser retains both as cell text; other HTML parsers may treat their leading less-than signs as markup. A consumer that needs numeric arrays can split these strings using the positional rules above and the embedded self_description; no year-wise row expansion, numeric decomposition, unit stripping, or reference parsing has been applied here.

The source marks only the 9.4(*) row in radiometer properties with the note (*) entirely dedicated to study the atmospheric emission. The 9.4 GHz row in representative measurements is bare 9.4; the note is absent from that table, the cold-load table, and the final-results table exactly as published. The product page's outer layout table is page structure, not a statistical product; the four inner scientific tables are recorded in metadata with their source-table positions, captions, flattened row matrices, and parser-normalized cell-inner HTML.

Provenance

The authoritative White Mountain product page was checked on 2026-08-31. Its pinned HTML is 25,088 bytes with SHA-256 99a3a55a56c0880ec6ae788b44c954d16b564a77150004bb45a51f186b63de83. The experiment page is pinned at 15,761 bytes with SHA-256 6ca45c4cc56de017f9d828f50f837644be3bbf449c608a7b9b2163ff58595703. The White Mountain publications page was checked on 2026-09-05 and is pinned at 15,596 bytes with SHA-256 1ef15ce0b4923fc6d2079bd183d516107d691066ebe2ff4c2ae01d18ba60f86f.

Run python tools/make_white_mountain.py --out data/white-mountain --source data/white-mountain --provenance-commit eef444757ee26e323e34e524b62868ab4d3534f5 from the repository root to regenerate the four Parquet files. The build pins the product-page byte count and checksum, requires exactly five HTML tables, checks each scientific table's complete shape and unique published headers, and verifies equality between every extracted source string and the built Parquet table. Each file carries the current supported statistical-product schema version 1.3.0, its source-table position, complete published rows, field descriptions, structured unit provenance, marker notes, bibliography, rights, credit, and the regeneration command.

Considerations for use

These are compact published summary tables, not time-ordered observations or a likelihood. The reported uncertainties and combined results should be interpreted through the cited experiment papers. Source text can combine two years, an uncertainty, and a bibliography marker in one cell, so treating every field as a scalar without applying the documented positional rules will lose meaning.

The mirrored LAMBDA tables contain three source-level discrepancies that are retained rather than corrected. At 90 GHz, the representative-measurement cold load is 2083±17 mK, while the product page's cold-load component table gives 2083±37 mK. Witebsky et al. (1986) gives a cold-load total of 2083 mK with ±37 mK uncertainty in Section VII(a) and Tables 4–5 on pages 152–153. Product-page reference 6 prints De Amici et al. at ApJ 298, 210, while LAMBDA's publications page identifies ApJ 298, 710 (1985ApJ...298..710D). Finally, the LAMBDA summary retains only its published 1982 and 1983 rows. Witebsky et al. reports observations in 1982, 1983, and 1984 in the abstract on page 145; the note to Table 6 on page 156 identifies its three rows at each wavelength as those years in order.

Licensing and credit

HEASARC states: “HEASARC materials are all available freely for your use.” See the HEASARC data policy, pinned at 23,756 bytes with SHA-256 0bea39b0bdccafdebf3cb3e32c3b5b81f310d606b14185b3ba377ab93cf2970a.

LAMBDA requests:

We acknowledge the use of the Legacy Archive for Microwave Background Data Analysis (LAMBDA), part of the High Energy Astrophysics Science Archive Center (HEASARC). HEASARC/LAMBDA is a service of the Astrophysics Science Division at the NASA Goddard Space Flight Center.

The LAMBDA contact page carrying that acknowledgement is pinned at 13,730 bytes with SHA-256 1b3c2496e3036f4be3b39bcc23b4e0d7e97d8198ec27dddea5e1f6047c95d1d0.

Citation

The LAMBDA pages cite Smoot, G.F. et al. 1983, Phys.Rev.Lett. 51, 1099 (ADS), and Smoot, G.F. et al. 1985, Ap.J., 291, L23 (ADS). The table's reference markers use this published bibliography (each entry is reproduced verbatim):

• (1) Sironi G. and Bonelli G., 1986, ApJ, 311, 418
• (2) Mandolesi N. et al 1986, Ap.J. 310, 561
• (3) Partridge R.B. et al., 1984, Phys.Rev. D, 29, 2683
• (4) Danese L. and Partridge R.B., 1989, ApJ 342, 604
• (5) Kogut A. et al., 1988, ApJ, 325, 1
• (6) De Amici G. et al., 1985, ApJ , 298, 210
• (7) Witebsky C. et al., 1986, ApJ, 310, 145

Cite the publication relevant to the frequencies or values used and acknowledge LAMBDA with the requested wording above.

Errata and requests

Report corrections or dataset requests in the dataset's Community tab.