M104 R4–7 Eight-sector extraction & QA

Feasibility

Extraction is possible, but four gates come first

The eight sectors are feasible in counts; the real uncertainty is the outer-corner QPB anchor and final fitting constraints of each 45° region on MOS1/MOS2, not the geometric area itself.

Target regions8each 45°, radius 4–7 arcmin

MOS2 candidates8all enter preflight; qualified status not guaranteed in advance

Known MOS1 limitations3Disk W, NW transition, Wind N expected to have no outer anchor

Currently qualified7 / 8Wind N independent review PASS; S1/S2/S4/S5/S6/S7 follow the B006 production-only QPB path

Live progress · 2026-07-11

Phase 2: MOS2 extraction complete; 7/8 qualified (B006 replay unavailable); S8 unavailable

S1–S7 each have region-specific source/QPB/ARF/RMF and a complete no-fit/spatial QA handoff. The Wind N fresh independent review deleg_998c11bd has PASSED; S8 (292.5–337.5°) has only central CCD1, and mos_back per its rules refuses QPB without an outer anchor.

Current status: 7 / 8 qualified MOS2 spectra / 7 / 7 visual QA PASS / 7 / 7 FAST technical pass / S8 unavailable. The REQUEST_CHANGES on Wind N v1 has been formally resolved by the hardened qualifier and a fresh independent review PASS; replay QPB is available. S1/S2/S4/S5/S6/S7 were qualified via the production-only QPB path; B006 documents that their frozen-CCF replay is unavailable, but production logs confirm the pipeline accepted the corresponding CCDs with no surrogate/parent-scaled QPB. All qualified sectors' spectra and the exact DS9 + MOS2 camera/source-mask figures have been inspected. The REQUEST_CHANGES of visual-QA review deleg_7987bec2 has been resolved: all spectra explicitly 0.4–3.2 keV, DS9 parser fail-closed, manifest/WCS/hash tests strengthened, and 375/390/1280 px responsive-width regression passed.

Exact DS9 regions8 / 8CSV+JSON manifest and combined display files generated

Monte Carlo closurePASS2,000,000 points; union = 1.000000000, gap = 0.000000000, overlap = 0.000000000

MOS converter48 / 486 MOS exposure detector images × 8 regions all 45°; independent coordconv centers and 3 interior direction points per region all pass

Qualified spectra7 / 8Wind N replay-verified; S1/S2/S4/S5/S6/S7 production-only QPB (B006); S8 unavailable

MOS2 product sets7 / 8Wind N + S1/S2/S4/S5/S6/S7; all qualified

S8UNAVAILABLEactive CCD=[1]; no outer exposed-corner anchor; no QPB

Remaining batchPIPELINE RC 0wrapper RC1 correctly exposes the S8 missing product; not a pipeline crash

FitFAST TRIAGE DONE5 constrained / 2 boundary-limited; final normal/2D confidence not started

Wind N source counts3,102700–1000 eV = 234 counts; planning≈231 is preflight only

Wind N QPB counts1,674.20700–1000 eV = 65.35 counts; region-specific ESAS QPB

Exposure / sky area80.05 ksBACKSCAL=16,730,648 → 11.62 arcmin²

CCD / responses1, 4, 5outer anchors=4,5; RMF=2,400 channels; ARF max=344.19 cm²

Wind N MOS2 no-fit xsherpa ARF-unfolded surface-brightness spectrum grouped with SNR6 and oversampling5 — **No-fit xsherpa inspect (qualified product QA evidence).** 0.4–3.2 keV, `group_subtracted_oversampling_SNR(5,6)`; data only are shown — no model was constructed or fitted.

Overlay of the Wind N exact DS9 region, MOS2 U005 camera mask, and four source-exclusion circles — **Region + camera-mask QA.** Blue fill/hatch is the included exact Wind N region, cyan is the valid MOS2 camera mask, red dashed are the four excluded source circles.

qualification JSON superseded release status JSON exact DS9 region detector expression inspect manifest overlay manifest

Complete visual QA

Seven qualified sectors: spectra and spatial maps inspected one by one

Per-sector inspection of exact DS9, MOS2 camera/source mask, and three xsherpa no-fit units. All EPIC spectra are y-log, 0.4–3.2 keV, and use group_subtracted_oversampling_SNR(5,6); these QA figures themselves contain no fit — fitting results are in the next section.

Spectrum files42 / 42 spectrum PNG7 sectors × rate/sb/sb_flux × 1col/2col; plus 42 paired PDFs

Spatial overlays14 / 14 spatial overlay PNG7 sectors × 1col/2col; exact image/mask WCS equality PASS

Visual verdict7 / 7 visual QA PASSall sector cards and four atlases checked by human and machine

InterpretationNO FITspectra are product QA only; spatial maps labeled spatial QA only

Spatial QA atlas of S1-S7 exact DS9 regions, MOS2 camera masks, and source exclusions — **Spatial QA atlas.** Each sector's included annular sector is blue fill+hatch, the valid camera mask is cyan, source exclusions are red dashed; RA/Dec, center markers, and the 5′ scale bar are all shown.

S1-S7 xsherpa ARF-unfolded no-fit surface-brightness spectra — **ARF-unfolded sb_flux atlas.** All 7 regions have binned data/error bars; no model curve. Checked together with the rate/sb atlases for unit-transformation consistency.

S1-S7 xsherpa no-fit rate spectra — **Rate atlas.** counts s⁻¹ keV⁻¹; for directly checking grouping, error bars, and count morphology.

S1-S7 xsherpa no-fit area-normalized surface-brightness spectra — **Surface-brightness atlas.** counts s⁻¹ keV⁻¹ arcmin⁻²; normalized by each region's BACKSCAL.

S1 Disk W MOS2 spatial and three-unit spectrum QA card — **S1 · Disk W · PASS.** PA 247.5–292.5°; CCDs 1,6; 4,291 source counts; production-only QPB (B006).

S2 NW transition MOS2 spatial and three-unit spectrum QA card — **S2 · NW transition · PASS.** PA 292.5–337.5°; CCDs 1,5,6; 3,576 source counts; production-only QPB (B006).

S3 Wind N MOS2 spatial and three-unit spectrum QA card — **S3 · Wind N · PASS.** PA 337.5–22.5°; CCDs 1,4,5; 3,102 source counts; replay-verified QPB + independent review PASS.

S4 NE transition MOS2 spatial and three-unit spectrum QA card — **S4 · NE transition · PASS.** PA 22.5–67.5°; CCDs 1,4; 3,165 source counts; production-only QPB (B006).

S5 Disk E MOS2 spatial and three-unit spectrum QA card — **S5 · Disk E · PASS.** PA 67.5–112.5°; CCDs 1,3,4; 3,190 source counts; production-only QPB (B006).

S6 SE transition MOS2 spatial and three-unit spectrum QA card — **S6 · SE transition · PASS.** PA 112.5–157.5°; CCDs 1,3; 3,305 source counts; production-only QPB (B006).

S7 Wind S MOS2 spatial and three-unit spectrum QA card — **S7 · Wind S · PASS.** PA 157.5–202.5°; CCDs 1,3; 4,217 source counts; production-only QPB (B006).

visual QA manifest S1/S2/S4/S5/S6/S7 qualification JSON Wind N qualification JSON public asset SHA-256

v22 temperature triage · 2026-07-12

Can the seven sectors
be fitted for a temperature?

This section preserves the completed historical mosIgn140-200 triage: source SP shape+norm frozen to R0–15, CXB norm free, MOS Line3=0; the effective band is 0.4–3.2 keV excluding MOS 1.4–2.0 keV. It is a FAST triage + source-only 1σ profile, not the final normal/2D confidence. The later controlled 4-vs-7 refit uses the unified 1.15–1.45 keV gap; the two must not be conflated as the same result.

Conclusion: 7 / 7 FAST technical pass; 5 constrained / 2 boundary-limited. S1/S2/S3/S5/S6 obtained two-sided finite 1σ intervals that do not touch the soft kT boundary. The 1σ upper bounds of S4 and S7 cross the current kT.max=1.0 keV, so they can only be called boundary-limited; final normal/2D confidence not started. The first trial run produced non-positive dof because groups straddled the 0.4/1.4/2.0/3.2 keV boundaries with in-gap line parameters still free; it was archived, fixed, and fully re-run.

Independent-review remediation (deleg_a559b0b0; follow-ups deleg_26850b86/deleg_eb1a6716/deleg_c8f1c9f8): the final re-run uses the schema-v2 strict manifest; the loader enforces exact 38 runtime data inputs, 7 code dependencies, 7 sectors, role/cardinality/path uniqueness, and a one-to-one sector↔top-level correspondence. All MOS PHA are forced to match the expected OBS_ID and contiguous channels; ARF/RMF are validated for finiteness, positive widths, monotonic ENERG/EBOUNDS, ARF↔MATRIX grid consistency, strictly increasing non-overlapping RMF groups, and zero-groups carrying no positive payload. The fit and confidence runners both reject MAX_JOBS outside 1–3 before creating directories/preflight. Each sector is mutex-protected by flock; session and CSV are published via staging, with the sidecar commit marker completing an atomic commit last. The restore validator item-by-item checks the exact source/QPB/ARF/RMF/RASS inventory, that SP/CXB are indeed in the current active model, that the current sole thawed set is CXB norm/source kT/source norm, and stat/dof/rstat/qval. The release gate compares per-region summary, FAST CSV, the FAST values embedded in constraints, confidence kT, session/sidecar hashes, source PHA and filter, rejecting legal-but-cross-generation mixed artifacts. The deep-semantic-hardened full re-run agrees item-by-item with the previous numbers.

Technical fits7 / 7 PASSdof=6–17; source norm all >0; fit succeeded

Temperature intervals5 constrainedS1/S2/S3/S5/S6; 1σ fractional half-widths roughly 10–20%

Boundary cases2 limitedS4/S7 near the kT=1.0 keV soft upper bound; no precise temperature conclusion

Science statusTRIAGEanswers "most sectors are measurable"; not yet the final publication fit

kT and one-sigma intervals of the seven axis-centered MOS2 sectors of M104 R4-7, with S4 and S7 using hollow boundary-limited markers — **kT profile.** Sky PA from North through East; orange=Disk, blue=Wind, gray=transition; solid points constrained, hollow boundary-limited. S7/S4 are hotter but their upper bounds are limited by the 1 keV soft boundary; N–S wind and E–W disk do not show simple symmetry.

MOS2 data, best-fit total model, and shared residual diagnostics for the seven sectors — **Fitted spectrum + residual atlas.** Gray marks the strictly ignored 1.4–2.0 keV band. S6 has a roughly −3σ point at low energies with rstat=1.95; S2's rstat=0.27 suggests few bins/conservative errors — both need final-fit re-checks.

7-sector vs historical 4-sector. Blue is the final 7×45° source-only 1σ profile, purple diamonds are the historical 4×90° FAST best-fits (no kT intervals), and the gray SW area is the unavailable S8. The kT(4-sector)−kT(7-sector) values along the four principal axes are N +0.074, E −0.131, S −0.221, W +0.008 keV: W agrees essentially, E and S differ more. Not apples-to-apples: the final 7-sector strictly excludes MOS 1.4–2.0 keV, the historical 4-sector excluded only 1.15–1.45 keV, and the angular widths and error contracts differ; this is a diagnostic display only — the differences must not be directly interpreted as true azimuthal temperature differences.

mobile 1col PNG 1col PDF 2col PNG 2col PDF comparison CSV

Sector	Role / PA center	kT (keV)	1σ interval (keV)	rstat / dof / Q	Constraint
S1 disk W	270°	0.698	0.626–0.759	0.82 / 17 / 0.675	constrained
S2 NW transition	315°	0.807	0.718–0.892	0.27 / 9 / 0.982	constrained
S3 wind N	0°	0.707	0.565–0.829	0.45 / 6 / 0.844	constrained
S4 NE transition	45°	0.911	0.746–1.095	1.39 / 6 / 0.213	boundary-limited
S5 disk E	90°	0.808	0.701–0.914	0.80 / 7 / 0.588	constrained
S6 SE transition	135°	0.771	0.679–0.856	1.95 / 6 / 0.0685	constrained
S7 wind S	180°	0.981	0.891–1.078	1.16 / 15 / 0.292	boundary-limited

S1 disk W fitted spectrum and residual — **S1 disk W.** kT=0.698 [0.626,0.759] keV.

S2 NW transition fitted spectrum and residual — **S2 NW transition.** kT=0.807 [0.718,0.892] keV; the low rstat warrants caution.

S3 wind N fitted spectrum and residual — **S3 wind N.** kT=0.707 [0.565,0.829] keV; the interval is wide.

S4 NE transition fitted spectrum and residual — **S4 NE transition.** kT=0.911 [0.746,1.095] keV; upper boundary-limited.

S5 disk E fitted spectrum and residual — **S5 disk E.** kT=0.808 [0.701,0.914] keV.

S6 SE transition fitted spectrum and residual — **S6 SE transition.** kT=0.771 [0.679,0.856] keV; rstat=1.95.

S7 wind S fitted spectrum and residual — **S7 wind S.** kT=0.981 [0.891,1.078] keV; upper boundary-limited.

mobile 1col kT profile temperature CSV FAST summary CSV hash-bound fit release JSON

Response-folded color diagnostic · 2026-07-12

Which two bands best distinguish
0.7 from 0.8 keV?

Folding phabs*apec one by one through the seven real MOS2 ARF+RMF sets of S1–S7, with NH=3×10²⁰ cm⁻², Z=0.3, z=0.003416 fixed. What is optimized here is the detector count ratio, not the unfolded theoretical line flux.

Fe-L ratio signal+42.9%0.7→0.8 keV; d ln R / d ln kT = 2.67

Source-only efficiency4.88σper 1,000 source counts at 0.50–1.15 keV

Current-sector precision1.03σmedian single-sector discrimination; a statistical floor, not yet including background covariance

Proxy vs full fitρ = 0.86median |ΔkT|=0.046 keV; shares the fit background, not an independent validation

Count spectra of 0.7 and 0.8 keV APEC through the real MOS2 response, and the ratio changes of the O VII, O VIII, and Fe-L candidates — **Theoretical sensitivity.** As 0.7→0.8 keV, the Fe-L blend centroid shifts visibly toward higher energies; internal Fe-L hardness is about +42.9%, while O VIII/O VII is only −7.8%.

Calibration curves of the Fe-L high over low and O VIII over O VII count ratios as a function of APEC temperature, folded through the real MOS2 response — **Calibration curve.** Fe-L high/low rises monotonically and rapidly over 0.55–1.0 keV; O VIII/O VII declines slowly. The spread across the seven sector responses is far smaller than the line width.

Azimuthal comparison of the S1-to-S7 Fe-L ratio temperature proxy against the full spectral-fit kT, with S8 unavailable — **Current S1–S7 trial computation.** QPB and the same session's non-source background model are subtracted first, then the Fe-L ratio is mapped to kT. The ordering still agrees well with the full fit (Spearman ρ=0.86), but with larger errors and compressed high-temperature amplitude; it suits cross-checks or high-count merged regions and cannot replace the full spectral fit.

Why not just use the O VII, O VIII, Fe-L "three lines"? In the current sectors the M104 source contributes only a median 8.5% of the O VII window and 18.5% of the O VIII window; the 0.7-vs-0.8 discrimination of O VIII/O VII has a median of only 0.049σ. Testing Z=0.2–0.5 and NH=(2–4)×10²⁰ cm⁻², this ratio changes by up to 12.3%, already exceeding its 7.8% temperature signal. The recommended three-band scheme: O control 0.50–0.70, Fe-L low 0.70–0.875, Fe-L high 0.875–1.05 keV; the main temperature color uses the latter two, with O VII/O VIII only monitoring foreground / absorption consistency. If computed from grouped PHA, grouping breaks must be forced at 0.70/0.875/1.05 keV; the more robust route is to forward-fold in ungrouped PI/counts space.

sensitivity 1col PNG calibration 1col PNG sector proxy 1col PNG summary JSON candidate ratios CSV sector realism CSV NH/Z systematics CSV full band scan CSV

Exact half-open component imaging · 2026-07-13

Fe-L color map with
point sources retained

Per the user's requirement the main map does not subtract point sources; the discrete PI strictly uses 700≤PI<875 and 875≤PI<1050; data, QPB, SP, and sky background all enter the same spatial grid in counts space, then share a 60″ Gaussian kernel for low/high. It is a morphology cross-check, not a diffuse-gas 2-D temperature map.

60″ vs 90″ morphologyρ = 0.951R4–R7 common-valid pixels; median |ΔR|=0.039

Point-included vs maskedρ = 0.857R4–R7; median |ΔR|=0.054

Image ratio vs full-fit kTρ = 0.750point sources retained; p=0.052; 7 qualified sectors

Masked ratio vs full-fit kTρ = 0.786p=0.036; same observation, not an independent discovery significance

MOS2 point-source-included Fe-L low/high net rate, background-corrected ratio, and statistical error maps — **Primary product.** Exact-band QPB + native MOS2 SP + response-folded sky background subtracted; the orange ellipses only mark the source masks — the main map does not apply them.

Point-source-included and masked Fe-L color, masked uncertainty, and the ESAS source mask — **Point-source control.** After masking, small-scale hotspots and central structure visibly decrease, but the high Wind S ratio and lower Disk W ratio persist; the primary product remains the included version at top left.

Stability comparison of the MOS2 Fe-L ratio under 30, 60, and 90 arcsec shared Gaussian kernels — **Scale control.** The 30″ effective area is small and details are unstable; the 60 and 90″ large-scale morphology agrees. Running `adapt` separately per band and then dividing is forbidden.

S1-to-S8 integrated Fe-L ratios, background and point-source controls, and the full-fit kT comparison — **Sector cross-check.** S7 Wind S is highest; S3's masked fractional error reaches 77%, uninterpretable. S4/S7 full fits remain boundary-limited.

Component contract: full-map QPB low/high=1,602.69/1,567.99 counts; sky=1,314.96/984.25; native SP only 21.21/21.67. The R0–15 native MOS2 model closes to 0.9% and 0.4% in the two bands respectively. Applying the MOS1 common-SP seed directly to MOS2 would be 17–22× higher than native SP; it is used only as an over-subtraction stress test.Open the full bilingual methods, equations, sector table, caveats, and downloads page.

Claude Code Opus4.8 high-effort independent audit: the band-ratio color-morphology release after the half-open reconstruction is PASS; the 8-sector spectral publication is REQUEST_CHANGES. S4/S7 remain limited by the 1.0 keV kT ceiling, and the Wind N/S ordering flips between the 4-sector and 8-sector binning — "Wind-S-hot" cannot be stated as a binning-robust quantitative conclusion.Download the full independent audit report.

Full HTML report primary FITS.gz masked-control FITS.gz sector CSV input/hash inventory method contract

The eight axis-centered 45-degree exact DS9 regions of M104 R4-7 overlaid on the actual XMM MOS2 sky image — **Actual geometry gate figure.** Orange are the Disk W/E axis bins, blue the Wind N/S axis bins, hatched diagonals the four transition bins; the colored radial lines are the W/N/E/S sector central axes. The background is the actual MOS2 sky image. The source-exclusion mask is a separate product, not shown here.

2col PNG 2col PDF DS9 manifest CSV DS9 manifest JSON verification summary MOS conversion CSV wrap negative-control CSV replayable release JSON

Review-hardened checks: the Monte Carlo now derives PA from the actual DS9 endpoints; a deterministic circular-adjacency and total-span=360° check is added. The converter expression requires exactly one DETX/DETY sector+annulus, radii 4800/8400, finite values; canonical angles are normalized to midpoint∈[-180°,180°). The positive wrap 337.5→22.5 = 315°, rejected; the negative-angle wrap -22.5→22.5 = 45° is accepted. All 48 conversions across the six MOS exposure detector images and the 6 wrap-control sets pass. MOS2 WCS planning counts assigned=2,547, unassigned=0, multiply-assigned=0; the eight centroids are 0.82–4.20° from their sector centers. Release m104-r47-eight-axis-centered-phase1-v1 records absolute paths, sizes, and SHA-256 of 6 code files, 11 FITS inputs, and 20 outputs; the tests recompute each item to reject stale mixed products.

Exact geometry

Eight 45° regions,
axis-centered

All boundaries are shifted 22.5° relative to the old scheme, placing the disk / wind axes at sector centers. File names record the DS9 pie angle; physical directions continue to use sky PA = DS9 angle − 90°.

region basename	DS9 pie geometry	Actual sky PA	Physical role	Sector center	MOS2 700–1000 eV planning counts	detector expectation
`Angle337.5-22.5`	`pie(-22.5,22.5)`	247.5–292.5°	Disk W	W · 270°	≈430	MOS2 candidate; MOS1 unavailable expected
`Angle22.5-67.5`	`pie(22.5,67.5)`	292.5–337.5°	NW transition	NW · 315°	≈265	MOS2 candidate; MOS1 unavailable expected
`Angle67.5-112.5`	`pie(67.5,112.5)`	337.5–22.5°	Wind N	N · 0°	≈231	MOS2 candidate; MOS1 unavailable expected
`Angle112.5-157.5`	`pie(112.5,157.5)`	22.5–67.5°	NE transition	NE · 45°	≈239	MOS1+MOS2 candidate; per-region check
`Angle157.5-202.5`	`pie(157.5,202.5)`	67.5–112.5°	Disk E	E · 90°	≈279	MOS1+MOS2 candidate; per-region check
`Angle202.5-247.5`	`pie(202.5,247.5)`	112.5–157.5°	SE transition	SE · 135°	≈261	MOS1+MOS2 candidate; per-region check
`Angle247.5-292.5`	`pie(247.5,292.5)`	157.5–202.5°	Wind S	S · 180°	≈349	MOS1+MOS2 candidate; per-region check
`Angle292.5-337.5`	`pie(292.5,337.5)`	202.5–247.5°	SW transition	SW · 225°	≈493	MOS1+MOS2 candidate; per-region check

Counting note: these values come from the MOS2 700–1000 eV ESAS source-image mosaic of the four qualified parents, split read-only along the corrected sky-PA boundaries; 2,547 counts total with no leakage across the eight regions. They serve feasibility only — formal counts must come from the exact detector expressions converted from the new DS9 files; this table must not be treated as PHA counts or fit results.

Execution plan

A four-stage pipeline from geometry to temperature

Every stage is a fail-closed gate. If the previous step did not pass, the next is not started; any partial directory is quarantined first and never mixed with production products.

Generate and verify regions · PASS

An independent directory R4-7_eight_axis_centered_45deg_sectors was established, without overwriting the existing four-sector products or the deprecated boundary-aligned plan.
8 DS9 FK5 files, a combined display file, and a manifest with DS9 angle, sky PA, central direction, and axis/transition roles were output.
The 2,000,000-point Monte Carlo passed: full-annulus union=1, gap=0, overlap=0; the centers of the four axis bins fall on W/N/E/S respectively.
This is an independent partition rotated 22.5° from the old 90° parents; no requirement exists that any two new sectors reconstruct an old parent.
All 48 ds9_to_xmmexpr conversions over the 6 MOS exposure detector images yield 45° detector sectors.
The actual DS9 overlay passed machine/WCS/human confirmation; the Phase 2 gate was lifted.

Production ESAS extraction · COMPLETE / 7 QUALIFIED

S1–S7 each have source/QPB/ARF/RMF and passed qualification; S8 unavailable.
S8 has only CCD1; mos_back skipped it per the fail-closed rule; no parent-scaled/surrogate QPB was constructed.
Wind N replay-verified; the other 6 regions accepted via the B006 production-only QPB path.
All qualified products come from region-specific mos-spectra + mos_back source / QPB / ARF / RMF.
Per-region command CCDs, EXPOSURE, BACKSCAL, counts, RMF/ARF, and PN exclusions were verified.

Per-spectrum QA handoff · PASS

The 7 regions' exact DS9 FK5/RA-Dec regions and converted detector expressions are all retained.
42 spectrum PNGs, 42 paired PDFs, and 14 spatial overlay PNGs complete.
All spectra are no-fit xsherpa.inspect; all spatial maps overlay the exact region + actual camera/source mask.
7 four-panel sector QA cards and 4 overview atlases were visually inspected.

Homogeneous refit and axis control

The approved four-sector values and the existing mosIgn140-200 seven-sector triage are retained as non-overwritten historical baselines; new results use independent, long, explicit output suffixes.
Following the v22 source/background topology: R0–15 SP shape+surface norm frozen, CXB norm free.
The user-selected unified contract is MOS2-only, excluding 1.15–1.45 keV, MOS Line3 (≈1.3 keV) norm=0; the energy gap and the Gaussian cancellation are two independent operations. Al/Si instrumental lines are not a universal forbidden band for MOS analysis and continue to be handled explicitly per branch.
Re-run the four 90° controls and the seven 45° sectors under identical grouping, background, free/frozen parameters, and source-only confidence definitions, without overwriting old sessions.
FAST_MODE triage first; normal/confidence only for sectors passing technical and science QC.
The final report will state the four axis-centered 45° temperatures, four transition temperatures, the corresponding 90° axis controls, detector composition, and confidence status; it will no longer claim that new sectors can pairwise reconstruct the old parents.

Current evidence

The plan is based on real four-sector products

Below is the current actual ESAS source-image overview. The corrected 45° partition is rotated 22.5° relative to these 90° hatches: the four axis bins sit at hatch centers, the four transition bins straddle adjacent hatch boundaries. This figure does not replace the Phase 1 exact-region overlay.

The current M104 R4-7 four-sector actual ESAS source images with the R4-6 parent comparison — **Current evidence.** All four MOS2 90° sectors qualified; MOS1 has only Disk E and Wind S qualified. After the eight-sector split, every 45° sub-region must redo its camera-specific CCD/QPB gate.

Actual extraction audit

Verifying each
mos-spectra companion image

Not just region file names: the audit binds directly to the PHA used in the fit, mos*-obj-im-700-1000.fits, MASK, the copied/standard DS9 regions, and rProfile. The historical four sectors have 6 camera-region products, the final S1–S7 have 7, 13 in total.

The historical four sectors' actual mos-spectra 700-to-1000 eV companion images overlaid with the standard R4-7 regions, camera masks, and point-source exclusions — **Historical 4-sector: 6/6 PASS.** The actual counts of MOS2 W/N and MOS1+MOS2 E/S all lie in the correct R=4–7 arcmin, 90° sky sectors; for E/S the MOS1 and MOS2 are in the same sky region.

The final S1-to-S7 actual MOS2 mos-spectra 700-to-1000 eV companion images overlaid with the standard R4-7 regions, camera masks, and point-source exclusions — **Final 7/8-sector: 7/7 PASS.** S1–S7 fall in their own 45° sectors in W/NW/N/NE/E/SE/S order; S8 remains missing because its QPB condition is not met.

Quantitative conclusion: all 13/13 PHA/image identities, detector annulus/sector widths, active DS9 geometry, image/MASK WCS, and image/rProfile total counts agree; relative to the M104 center, within a 2.5-arcsec pixel tolerance, the positive pixels crossing the R4/R7 or PA boundaries are all zero. Wrong-region extraction can therefore be excluded from the explanation of the 4-vs-7 temperature differences. Two provenance caveats: the historical four product-directory region annotations suffered an old label misalignment, but the active pie(...) and the actual images are correct; the surviving rProfile radial axis is centered on the FITS RA_OBJ/DEC_OBJ, 2.134′ from the M104 center — it can only be used for its spatial panels/total counts in verification, not to judge the M104 R4–7 boundaries from its radial axis.

4-sector 1col PNG 4-sector 2col PDF 7-sector 1col PNG 7-sector 2col PDF 13-record CSV summary JSON

Go / no-go

Stop, merge, and publication gates

"Extraction succeeded" does not mean "the eight-point temperature map is credible". Technical availability, fit constraints, and physical interpretation must be recorded separately.

Stop immediately

Geometry gap/overlap, 45° converter width errors, physical labels inconsistent with WCS, missing QPB, CCD1-only without outer anchor, any missing source/QPB/response.

Keep but mark Questioned

kT touching a boundary, confidence failure, obvious camera residual conflicts, insufficient source/background constraints, anomalous fit statistic. Questioned points cannot be used to claim 45° temperature structure.

Fall back when necessary

If an axis-centered sector is unavailable, fall back to the corresponding verified 90° axis control; if a transition sector is unavailable, record the explicit gap or design a separate symmetric rebin. No arbitrary adjacent new sectors may be spliced into an old parent, and no surrogate fills the 8 points.

Delivery contract

Final deliverables

After each stage the same Cloudflare Pages report and the M104 Discord thread are updated, keeping plan, regions, spectra, and fit results in one place.

Geometry package

8 DS9 FK5 regions, combined file, CSV/JSON manifests, MOS1/MOS2 detector expressions, 1col+2col overlay PNG/PDF, and the WCS regression test.

Spectrum QA package

Qualification manifest, source/QPB/ARF/RMF paths and hashes for every qualified camera/sector, no-fit inspect figures, region+camera-mask real-XMM overlays.

Temperature package

8-sector kT polar/linear plots (1col+2col), spectrum+residual atlas, parent-control comparison, QC classification, and a mobile-readable Cloudflare HTML.