Orion Bar H2: NIRCam filter maps checked against NIRSpec spectroscopy (JWST ERS 1288) — v1.0

This dataset tests how well simple JWST NIRCam filter maps of the H2 1–0 S(1) 2.12 µm and 0–0 S(9) 4.69 µm lines in the Orion Bar reproduce the line intensity measured by NIRSpec IFU spectroscopy (PDRs4All, JWST ERS 1288). The compared quantity is the H2 line's contribution to the band-averaged surface brightness in F212N or F470N (MJy/sr, proportional to line intensity), on NIRSpec spaxels (0.1″ grid, 0.25″ FWHM) inside the NIRSpec stripe with line S/N > 10 (n = 1609 at 2.12 µm, n = 1359 at 4.69 µm). Ratios are NIRCam-derived / NIRSpec-measured per spaxel, given as median and 16th–84th percentile across spaxels; the percentile range is spread, not an uncertainty on the median, and spaxels are spatially correlated. Method: NIRCam images were put on a common grid and PSF-matched. NIRSpec cubes were converted to synthetic filter fluxes, split into H2 line, other lines and continuum. The two were cross-calibrated with L = (F_N − c·F_M − h·B)/(1 − c·ε), where an H I term B is traced by Br-alpha (F405N). Result: the naive NIRCam filter-map estimate at 2.12 µm (F212N − F210M) recovers a median of 0.374 of the NIRSpec-measured line intensity (16th–84th percentile −0.049 to 0.572; 18.8 % of spaxels negative). The corrected map gives a median ratio of 1.065 (0.849–1.339), with no negative spaxels; split-half medians 1.071 and 1.176. At 4.69 µm the ratio is 0.906 (naive) and 1.129 (corrected). No uncertainty on the median was computed. Filter-artefact test (defined in scripts/step5_morphology.py and step5b_summary.py): a structure passes if it keeps local-contrast S/N ≥ 3 after correction and with c at its 16th/84th percentile, is seen at S/N ≥ 3 in the other H2 line, and has no opposite-sign ≥ 3σ counterpart in the medium-band filter or Br-alpha. In the signal region, 52 of 123 bright 2.12 µm ridges and 14 of 81 dark lanes pass this test. Passing it does not by itself establish what the ridges physically are. A five-fold spatial cross-validation (pooled out-of-fold, same ratio definition per the summary file; original CV script not found) gives median 1.058 (0.835–1.353) at 2.12 µm and 1.135 (0.999–1.257) at 4.69 µm, i.e. the 4.69 µm correction gives about 13.5 % higher line intensity than NIRSpec. Per-fold medians range 0.94–1.244 and 0.89–1.207. An independent re-implementation of this cross-validation (made 2026-10-26; NOT the original script; definitions fixed before running, no tuning) reproduces the full-sample calibration exactly (all 1262 values in krydskalibrering.json) and gives pooled out-of-fold medians of 1.045 (0.814–1.335) at 2.12 µm and 1.136 (0.985–1.257) at 4.69 µm. Folds 2–5 agree with the summary (identical coefficients; at 4.69 µm fold medians 0.0008–0.0011 higher); fold 1 differs (2.12 µm: 1.026 vs 1.109; 4.69 µm: 0.851 vs 0.890), a possible explanation is a different fold-1 edge definition in the original, but the cause is unresolved and cannot be determined without the original script. Scope: validation applies only inside the NIRSpec stripe (~25″ × 3″). Only line intensity is compared; no mass or column density is derived. The correction is linear with constant coefficients; an unexplained 15–17 % calibration offset between the narrow- and medium-band filters remains. No new physical PDR claims are made. The deposit contains (1) Orion_H2_analyse_public_subset.zip: the author's own files (scripts, results, derived FITS maps, figures, report) copied byte-identically from a frozen original zip (SHA-256 3131d2356cbd14840717b0e7955397d0cea3554f9c7fe6285ae696638756e231; not uploaded because it also contains third-party files), with PUBLIC_SUBSET_MANIFEST.txt; (2) the unmodified spatial-CV summary zip; (3) Orion_H2_CV_independent_rerun_2026-10-26.zip: the independent CV re-implementation (code, pre-declared definitions, results, log; labelled as not the original script); (4) Orion_H2_helfelt_corr_LW_maps.zip: the corrected 2.12 µm and 4.69 µm H2 maps over the whole NIRCam field (LW grid, 4 FITS halves, MJy/sr band-averaged line contribution; validated only inside the NIRSpec stripe, extrapolated elsewhere); (5) Orion_H2_CV_intermediate_inputs.zip: the author's intermediate files nirspec_synth_212/469/bra.fits (synthetic filter fluxes from the NIRSpec cubes) and crosscal_maps.fits (cross-calibration maps), which remove the need for NIRSpec raw data when rerunning the CV; (6) Orion_H2_NIRCam_LW_grids.zip: the author's NIRCam LW filter grids from step1 (F210M, F212N, F405N, F470N, F480M on the F480M grid, PSF-matched, in two halves each, plus the mask) with a script that rebuilds the full grids byte-identically. With (5) and (6) the cross-validation re-implementation can be rerun from this deposit alone, without raw data; tested 2026-10-26, it reproduced the shipped outputs byte-identically; (7) a wrapper. Third-party PDRs4All template spectra and SVO filter curves are not included; sources and SHA-256 are listed in THIRD_PARTY_MATERIAL.md. JWST data are not included and must be retrieved from MAST; file names and SHA-256 hashes are listed. The analysis was carried out with the help of AI assistants; the author directed it and is responsible for the content.