---
title: "Claude review disposition: MW/M31 relative contribution"
date: 2026-07-12
format: gfm
---

## Scope and supersession

The review is scientifically useful, but its quantitative Locatelli discussion was based on a superseded draft. The final audit uses the Locatelli et al. (2024) reference Combined model with fixed $\beta=0.5$, not the `combined+SWCX` systematic test.

Current response-matched quantities, in absorbed 0.4--1.25 keV surface-flux units of $10^{-15}\,\mathrm{erg\,cm^{-2}\,s^{-1}\,arcmin^{-2}}$, are:

- 14-field range: $1.431$--$1.668$;
- North/NW: $1.608$;
- South/SE: $1.511$;
- side-balanced: $1.560$;
- projection through the same all-field inverse-variance estimator as the measured total: $x=1.602$;
- measured all-field total: $S_{\rm obs}=1.092\pm0.052$.

The review's $x=1.435$ value is not retained.

## Independent audit resolution

Two read-only current-tree audits agreed on the physical boundary but differed on presentation. One recommended omitting the Locatelli comparison from the abstract; the other classified the present abstract wording as already fixed and safe. We retain the comparison because it names the reference model, uses an estimator-matched central value, says “no non-negative central residual” rather than “no statistical room,” and immediately identifies model/conversion tension rather than negative or vanishing M31 emission. This is an editorial choice inside a shared scientific boundary, not a disagreement about the underlying inference.

For Table 5, the 14-field range $1.431$--$1.668$ remains footprint variation and is not subtracted from an all-field measurement as if it were an uncertainty interval. The central complement instead uses the all-field estimator-matched value $x=1.602$.

## Final frozen-tree audit

A final independent read-only audit found no critical numerical error, estimator mixing in the manuscript, stale $x=1.435$ return, or overstatement of an independent M31 detection. It confirmed the Combined $\beta=0.5$ identity, all-field $x=1.602$, measured total $1.092\pm0.052$, fixed-line $Z=0.3$ conversion $x=1.520$, full-screen absorption direction, and Table 5 complement arithmetic.

Three non-blocking precision risks were closed after that audit: the 14 fields are now called “quality-selected” rather than equated with all technical passes; the conclusion states a common phenomenological temperature rather than a physically uniform plasma; and the machine-readable prior ledger now distinguishes its Figure-3 side-balanced `primary_central` from an explicit `all_field_inverse_variance` column and labels the field min--max as footprint variation rather than uncertainty.

## Public alias implementation review

An independent read-only code/spec review of the new public measurement aliases found no Critical issue but identified two reachable failure paths. Both are closed. The generator now resolves and rejects any collision among source CSV, public CSV, and data-dictionary paths before writing, and regression tests confirm the frozen-source hash remains unchanged. The Cloudflare builder now completes generation, translation checks, security scans, and manifest creation in `.site-staging`; a failed build leaves the previous validated `site/` and unrelated parent files intact, while a successful build performs a transactional promote with rollback. Tests also require all three public-alias artifacts to appear in download links and the manifest with matching byte counts and SHA-256 values.

## Main MW/M31 comments

| Review item | Disposition | Revision or boundary |
|---|---|---|
| 1. Elevate the Locatelli result | Adopted with corrected values and conditional wording | Added to the abstract, a dedicated Discussion subsubsection, Table 5, and the conclusions. The allowed statement is “the named reference model leaves no non-negative central M31 residual”; the unconditional phrase “no room for M31” is rejected. |
| 2. Frame the tension symmetrically | Adopted | The Discussion now enumerates three live readings: Milky-Way model overprediction, spectral/abundance/absorption conversion bias, or a low-biased v19 sum. At $b=-21.57^\circ$, one $z_h=1.1$ kpc disk scale height is reached after about 3.0 kpc, making the disk leverage explicit. |
| 3. State the contrast-sign information | Adopted as a weak diagnostic | The text now states that the positive observed sign is directionally closer to the MW-gradient template than to a symmetric declining M31 template, immediately followed by the $0.69\sigma$ and $1.49\sigma$ offsets and an explicit non-exclusion statement. |
| 4a. Abundance mismatch | Partly closed and partly retained as a submission blocker | Observed-band energy fluxes remain dimensionally comparable, but their response and line-to-band conversions are spectrum dependent. At fixed fitted O VIII normalization, changing the target conversion from $Z=0.1$ to $0.3\,Z_\odot$ gives a field range $1.358$--$1.583$, all-field $1.520$, and side-balanced $1.480$. This conversion-only branch does not replace a matched v19 abundance/temperature refit grid. |
| 4b. Absorption geometry | Adopted | The paper now states that applying the complete HI4PI screen to MW emission is conservative for the $x>S_{\rm obs}$ tension. A distributed absorber transmits more MW model flux and therefore strengthens, rather than removes, that tension. A quantitative distributed-absorption branch remains open. |
| 5. Add a compact conditional-M31 table | Adopted | New Table 5 uses estimator-matched central bookkeeping. Locatelli gives $x=1.602$, $y=-0.510$; Zhang gives $y=0.942$, $x=0.150$; EAGLE complements are $0.835$, $0.580$, and $0.016$; SIMBA complements are negative. These are model-compatibility checks, not confidence intervals, detections, or calibrated limits. |
| 6. Bound inner disk/bulge leakage | Valid concern; not fabricated | No defensible quantitative bound exists in the current artifacts. A sky-footprint/PSF/mask/unresolved-source leakage calculation is now an explicit submission blocker in the manuscript and status ledger. |

## Other comments

- **South/SE radial gap:** adopted. The abstract now leads with the common temperature and all-field flux and says that the side contrast is not constrained cleanly by the present sampling.
- **Intrinsic scatter $Q=0$:** adopted. The manuscript states that 14 fields, only four South/SE fields, and their larger errors give low power to detect moderate scatter.
- **CXB cosmic variance:** adopted in the spatial-contrast Discussion as a competitor to weak MW/M31 morphology.
- **v18 to v19 regression:** adopted without assigning an unmeasured cause. The central contrast moved only from $0.173$ to $0.212$, while the uncertainty grew from $0.094$ to $0.210$.
- **SIMBA-NoAGN:** adopted. The paper states that both SIMBA central predictions exceed the measured total before adding any non-negative MW term; Table 5 makes the negative complements explicit.
- **Printed Internal Draft Status:** removed from the compiled manuscript and retained only as source comments/status documentation.
- **Temperature precision:** rounded to $0.175\pm0.004$ keV in the abstract and conclusions. The exact baseline value remains in the Results equation and machine-readable products for reproducibility.
- **Two conditional figures:** intentionally retained. The $T$--$D$ plane tests spatial sign/geometry, whereas the $x$--$y$ plane tests named amplitude priors; neither substitutes for the other. Their roles are now stated more sharply.
- **`CGMsum` naming:** paper-facing axes use `CGMsum`. A new public 14-field measurement CSV adds preferred `cgmsum_*` aliases while retaining the original `mwhalo_*` columns as explicitly documented historical frozen-fit fields; the frozen source products are unchanged.

## Remaining submission-grade work

1. matched abundance/temperature refits rather than conversion-only sensitivity;
2. distributed Milky-Way absorption and alternate total-column prescriptions;
3. radial-plus-side likelihood and boundary-aware intervals;
4. CXB, SWCX, Soft Proton, energy-band, and model-topology systematics;
5. quantitative inner-disk/bulge/PSF/unresolved-source leakage bound;
6. calibrated U0 and named-prior U1/U2 upper-limit products.

The editorial hierarchy remains unchanged: the direct result is $S_{\rm obs}=S_{\rm MW}+S_{\rm M31}$, with an optional named Local-Group term. External models constrain conditional decompositions but do not create an independent M31 detection.