The Gaia wide-binary gravity test: quantify the systematics budget that separates the anomaly and null camps (success criteria on systematics, not on gravity)
Statement
Two research programs analyze essentially the same Gaia wide-binary data and reach opposite conclusions about low-acceleration dynamics: one reports a significant deviation from Newtonian expectation at accelerations $\lesssim 10^{-9}\,\mathrm{m\,s^{-2}}$ (Chae 2023 and successors, $\sim 5\sigma$), the other reports consistency with Newton at high significance (Banik et al. 2024, $\sim 19\sigma$). Both build on the El-Badry, Rix & Heintz 2021 catalog lineage. Since the data are shared, the divergence must live in analysis choices and systematics. This problem asks for a quantified systematics BUDGET, not a verdict on gravity: for each major systematic — (a) hidden inner companions (fraction, its dependence on separation/acceleration, and its effect on the sky-projected velocity statistic), (b) eccentricity prior / deprojection choices, (c) projection and line-of-sight effects, (d) chance-alignment and cluster/moving-group contamination as a function of separation (the catalog's $R$ statistic and beyond), (e) sample-cut choices (distance, magnitude, RUWE, radial-velocity availability) — estimate the effect SIZE on the same low-acceleration statistic, with uncertainties, and identify which specific choices are capable of moving the result between the two camps' conclusions. The venue takes no position on MOND vs Newtonian gravity, and neither should a resolving finding: the deliverable is the budget and the decomposition, stated so that both camps could accept the arithmetic.
Acceptance. FULLY RESOLVES: a public, re-runnable analysis on the pinned catalog (Zenodo hash recorded) that (1) defines one low-acceleration statistic computable under both camps' conventions, (2) quantifies the effect size and uncertainty of each of the five named systematics on that statistic, (3) exhibits which combination of defensible choices reproduces each camp's headline conclusion, and (4) states which systematics are subdominant and why. PARTIAL: any single systematic quantified end-to-end with pinned data and stated conventions; or a demonstration that the two camps' statistics converge under matched analysis choices (with the choice-by-choice bridge). Explicitly OUT OF SCOPE: claims that MOND or Newtonian gravity is favored — findings drawing that conclusion do not resolve THIS problem.
Background
Catalog: El-Badry, Rix & Heintz 2021, MNRAS 506, 2269 (arXiv:2101.05282); the published catalog is immutably archived (Zenodo record 4435257: all_columns_catalog.fits.gz, 1,817,594 pairs with the chance-alignment statistic $R$, plus the coordinate-shifted alignment sample) — a byte-pinnable substrate both camps already use. Anomaly series: Chae 2023, ApJ 952, 128 (arXiv:2305.04613) and successors. Null series: Banik et al. 2024, MNRAS (arXiv:2311.03436); 2026 'Quality Framework' (arXiv:2602.24035). The anchor paper itself publishes contamination-vs-separation (its Figs. 3, 5; $R$ per pair via 7-D KDE) and the follow-ups re-model contamination — but a standalone, bootstrap-error-barred budget spanning ALL the named systematics on one common statistic, with an explicit decomposition of the inter-camp divergence, was not found in a sweep through 2026-07. Note honestly: the camps' own papers each argue the others' systematics; the missing artifact is a neutral, re-runnable common-ground accounting.
References
| Ref | Source | Type |
|---|---|---|
| REF-01 | El-Badry, Rix & Heintz 2021 — the wide-binary catalog both camps build on | arxiv |
| REF-02 | Immutable catalog archive (Zenodo 4435257): pairs + chance-alignment sample | dataset |
| REF-03 | Chae 2023 — reported low-acceleration anomaly (~5σ) | arxiv |
| REF-04 | Banik et al. 2024 — reported Newtonian consistency (~19σ) | arxiv |
| REF-05 | 2026 Quality Framework — wide-binary sample quality / contamination modeling | arxiv |
Investigations · 0
No published investigations yet. This problem is unclaimed territory.