M104 R0–15 EPIC camera comparison

Previous mixed contract vs. traditional 3B masked full region · ObsID 0900170101

Executive conclusion

The traditional 3B full-region result confirms that pn0 still provides the strongest single-camera Fe-L statistics, but the degree of "PN dominating MOS" in the previous comparison was amplified by the shorter GTI and smaller BACKSCAL extraction area of the 4B MOS.

In 0.7–1.05 keV:

• Previous mixed contract(4B MOS + 3B pn0): MOS1+MOS2 net counts are 3,516, pn0 is 8,592, and the pn/MOS net-count ratio is 2.44;
• Traditional 3B masked full region(3B MOS1+MOS2+pn0): MOS1+MOS2 net counts increase to 5,534, pn0 remains 8,592, and the pn/MOS net-count ratio drops to 1.55;
• The MOS-only naive independent-counts S/N increases from 44.7 to 56.4; the naive S/N after adding pn0 increases from 85.5 to 92.2;
• The 3B MOS1/MOS2/pn0 Fe-L high/low values are 0.559 / 0.597 / 0.585, respectively. MOS1–PN0 and MOS2–PN0 differ by only −0.71σ and +0.37σ;
• pn0 is therefore still worth including in the controlled MOS+PN forward-folded validation fit, but neither the 1.63 nor the 1.91 counts-space S/N ratio is a kT precision forecast.

English summary. The all-3B comparison recovers substantially more MOS counts because the traditional MOS products have longer exposures and larger extraction footprints/BACKSCAL sky areas. PN0 remains the strongest single-camera Fe-L dataset, but its net-count advantage over MOS1+MOS2 decreases from 2.44 to 1.55. The response/area-normalized Fe-L colors remain statistically compatible. These are QPB-subtracted quick-look spectra, not fitted source-component significances; the counts-space gain is not a kT precision forecast.

Two data contracts

Detector	Previous mixed	Exposure / PHA area	Traditional 3B	Exposure / PHA area
MOS1	4.background masked full-FOV alias	65.76 ks / 120.3 arcmin²	3B.background_20250124/...exclude_extent_source_mask	78.81 ks / 162.6 arcmin²
MOS2	4.background masked full-FOV alias	65.23 ks / 335.0 arcmin²	3B.background_20250124/...exclude_extent_source_mask	80.03 ks / 396.9 arcmin²
pn0	3B.background_20250124/...pn0...mask	55.37 ks / 369.8 arcmin²	The same pn0 PHA/QPB/ARF/RMF	55.37 ks / 369.8 arcmin²

The PHA, QPB, ARF, and RMF SHA256 of PN0 are identical across the two contracts, so it is a strict control; the old-vs-new curve differences come only from the MOS products. The MOS and PN camera footprints are still not a pixel-identical common aperture.

Direct old–new comparison

0.4–3.2 keV ARF-unfolded surface brightness

Download PDF

The two panels explicitly use the same x/y scales (sb_flux y-limit = 9×10⁻⁸–3×10⁻⁵). The overall shape is stable over 0.4–1.2 keV; the strong structures near 1.49 and 1.74 keV belong to the instrumental-line-sensitive region. In 2–3.2 keV the 3B MOS sb_flux can be higher by ~12–25%, which more likely reflects GTI, residual particle/SP, and response-weighting differences and cannot be directly interpreted as hotter source plasma.

0.6–1.2 keV Fe-L zoom

Download PDF

In the traditional 3B the MOS curves have smaller errors and more grouped bins; the Fe-L continuum slopes of the three instruments remain compatible. The figure has no model or fit and cannot distinguish M104, foreground, CXB, soft proton, and OoT components.

0.7–1.05 keV quantitative comparison

Previous mixed contract

Detector	Source	Scaled QPB	Net	S/N	QPB fraction	Mean ARF
MOS1	1,750	582	1,168	27.5	33.3%	172 cm²
MOS2	4,129	1,781	2,348	35.5	43.1%	164 cm²
pn0	12,447	3,855	8,592	72.9	31.0%	495 cm²

Traditional 3B masked full region

Detector	Source	Scaled QPB	Net	S/N	QPB fraction	Mean ARF
MOS1	2,987	1,029	1,958	35.2	34.4%	173 cm²
MOS2	6,088	2,513	3,575	44.3	41.3%	160 cm²
pn0	12,447	3,855	8,592	72.9	31.0%	495 cm²

MOS change of 3B relative to previous

Detector	Exposure	PHA area	Net counts	S/N	Net rate per area	Approx. sb_flux
MOS1	+19.8%	+35.2%	+67.7%	+28.0%	+3.5%	+2.9%
MOS2	+22.7%	+18.5%	+52.3%	+24.9%	+4.8%	+7.1%
pn0	0.0%	0.0%	0.0%	0.0%	0.0%	0.0%

The conclusion is: most of the increase is counts and S/N brought by the MOS BACKSCAL extraction area × exposure, not an instrumental effective-area (ARF) or surface-brightness rise. The Fe-L mean ARF changes by only +0.6%/−2.2%, and sb_flux by only ~3–7%.

Fe-L color consistency

Physical half-open bands: low [0.700,0.875) keV, high [0.875,1.050) keV. Statistics use the RMF-channel midpoint lo <= E_mid < hi to avoid double-counting the nonlinear pn EBOUNDS boundary channel.

The errors and z-scores in the table below are count/QPB-statistical only: they propagate only source Poisson and QPB STAT_ERR, assume the two bands are independent, and exclude ARF/RMF calibration, soft-proton, sky/OoT, and common systematic errors.

Contract	MOS1 high/low	MOS2 high/low	pn0 high/low	MOS1–PN0	MOS2–PN0
Previous mixed	0.517 ± 0.039	0.589 ± 0.033	0.585 ± 0.017	−1.61σ	+0.11σ
Traditional 3B	0.559 ± 0.032	0.597 ± 0.027	0.585 ± 0.017	−0.71σ	+0.37σ

After 3B processing the MOS1–pn0 difference shrinks; MOS2 agrees with pn0 in both contracts. This change is consistent with sensitivity to the compound processing/footprint contract, but since PHA, QPB, ARF, and RMF all change at once, the specific cause cannot be isolated, and this does not constitute a precision cross-calibration measurement.

Traditional 3B-only inspect

Broad sb_flux

PDF

Fe-L zoom sb_flux

PDF

All inspections use xsherpa.inspect.SpectrumInspector; energy = 0.4–3.2 keV; group_subtracted_oversampling_SNR(sampling_rate=5, SNR=6); QPB subtraction; y-log; no model; no fit. The mandatory rate, sb, sb_flux, and 1col/2col PNG/PDF are all kept in the local science output.

Scientific limits

1. QPB-subtracted is not pure M104. The spectra still contain sky foreground/background, CXB, residual soft protons, instrumental residuals, and possible pn OoT.
2. PATTERN==0 is not an OoT correction。 pn0 is a single-pixel/high-resolution selection; the Extended Full Frame OoT must be handled separately by the matched OoT product.
3. The mask is not a common aperture. The chip gaps, bad pixels, and camera coverage of MOS1/MOS2/pn differ; PHA BACKSCAL records only the respective extraction area, not the ARF instrumental effective area.
4. ARF-unfold is for quick-look only. Quantitative inference for extended sources must use forward-folded detector-specific models.
5. 3B does not automatically become the production baseline because of this. It provides more MOS statistics, but the previous project had its GTI/QPB screening reasons for adopting 4B. A decision requires MOS-only and MOS+pn forward-folded validation fits under the same source/background/GTI/OoT contract.

Reproducibility and downloads

Independent audit status

• Claude Sonnet medium reached its audit turn limit (error_max_turns,15 turns reported against a14-turn cap) without a verdict; one internal tool attempt also emitted a Conda/plugin error. No Claude verdict was used.

• Network-enabled Codex gpt-5.6-sol max independently reopened the FITS products, recomputed all key numbers without importing the project statistics functions, checked current ESA/SAS documentation, and returned PASS WITH CAVEATS with no Critical/High finding.

• Both Medium findings were corrected: BACKSCAL extraction area is now distinct from ARF instrumental effective area, and the MOS1 interpretation is limited to compound-contract sensitivity with count/QPB-statistical-only errors.

• Explicit common sb_flux y-limits, exact 24/24 manifest cardinality, Git-head/SHA binding, removal of the four-byte public audit artifact, and10 focused tests were verified. Final current-tree re-review: PASS.

• Combined statistics CSV

• Combined summary JSON

• Input paths and SHA256 inventory

• Previous statistics CSV

• Previous summary JSON

• Previous independent Codex audit

• Traditional 3B independent Codex audit

• Previous analysis driver

• Traditional 3B comparison driver

• Public asset manifest

All published science assets are byte-bound to their local source artifacts by SHA256 in asset_manifest.json. The manifest also records the Git HEAD; because the working tree is uncommitted, the formal release identity is the per-asset SHA256 set rather than a clean commit.