JWST MIRI Medium Resolution Spectrometer Point Fixed Pattern Corrections: Cleaner and Higher Signal-to-Noise Spectra of Point Sources

Abstract: The JWST Mid-Infrared Instrument Medium Resolution Spectrometer provides the capability to obtain spectra from 5-28 micron. The JWST data reduction pipeline removes the majority but not all of the instrument artifacts and the signal-to-noise (S/N) of the resulting spectra are limited by fixed pattern noise. Building on previous work, Point Fixed Pattern Corrections (PFPCs) are constructed using observations of O, A, and G dwarf flux calibration stars and asteroids taken using the default four point dither pattern. The PFPCs can be applied to spectra of point sources taken with target acquisition and the same dither pattern. They can be used alone or with the pipeline residual fringe correction depending on the sources spectral properties. Both narrow and broad artifacts are removed by the PFPCs improving the spectra regardless of their S/N. For higher S/N observations, the PFPCs significantly improve the S/N by up to factors of a few and S/N values of 1000 or more. The MRS-PFPC python package is provided to allow anyone to utilize the PFPCs for their own data.

• The paper introduces 48 dither- and segment-specific Point Fixed Pattern Corrections built from O, A, and G dwarf stars plus featureless asteroids across the 5–28 μm MIRI MRS range.
• The corrections improve point-source spectral signal-to-noise by up to 8× when residual fringe correction is unsuitable, raise short-wavelength S/N above 1000, and reduce segment-overlap corrections to near unity.
• The paper demonstrates that the reported 5.8 μm carbonyl absorption feature is instrumental, while noting that the corrections remain tied to pipeline version 2.0.0 and the standard four-point dither pattern.

Motivation and context

The MIRI Medium Resolution Spectrometer (MRS) delivers $R \sim 3000$ integral field spectroscopy from 5–28 $\mu$m across four nested IFU channels, each with three grating settings. Although the jwst pipeline removes most instrumental signatures, the achievable signal-to-noise ratio (S/N) of point source spectra remains limited by fixed pattern noise: residual fringes of 1–2% amplitude after the static fringe flat (which is derived from extended sources and mismatches the point-source fringe response), $\sim$1.5% flat field uncertainties in the 5–18 $\mu$m range, sampling artifacts, and broad percent-level absorption artifacts such as the feature near 5.8 $\mu$m. The pipeline's residual fringe correction, which fits Fourier components within known fringe frequency ranges, can remove genuine astrophysical signal when a source exhibits regularly spaced emission or absorption lines at similar frequencies—for example molecular bands—making it unusable for many science targets.

Gordon & Law build on prior dither-dependent corrections derived from single A stars (Gasman et al., 2022) or asteroids plus one A star (Pontoppidan et al., 2023), with the explicit goal of combining multiple calibration sources to push MRS point source calibration into the sub-1% regime. Unlike the pointing-based defringing work of Gasman et al. [(Gasman et al., 2024); 2025A&A...697A..58G], which accounts for delivered sub-pixel positions but yields only modest gains in dither-averaged spectra, this work derives a single wavelength-dependent correction per dither position per channel/grating combination, relying on the excellent repeatability of JWST target acquisition.

Calibration data and construction

The Point Fixed Pattern Corrections (PFPCs) are constructed from observations of two O dwarfs (10 Lac, $\mu$ Col), three A dwarfs ($\delta$ UMi, HR 5467, HD 2811), four G dwarfs (HR 6538, HD 37962, 16 Cyg B, HD 167060), and two Themis-family asteroids (515 Athalia, 526 Jena), all obtained with the default 4-POINT, NEGATIVE, POINT-SOURCE, ALL_MRS dither pattern from cycles 1–3 flux calibration programs and the JDISC asteroid program. Data were reduced with jwst pipeline version 2.0.0 using bad-pixel self-calibration, mingrad pixel replacement, band-level IFU-aligned cube building, autocentroided extraction, and correction of the 3A spectral leak.

For each of the 12 channel/grating segments and each of the 4 dither positions—48 corrections in total—the PFPC is measured as the ratio of the extracted spectrum to its model: CALSPEC models for the stars and quadratic continuum fits for the asteroids. This complementary source selection is central to the method's robustness. Stars are excluded from channel 4 (low throughput combined with rapidly declining SEDs) and asteroids from channel 1 (SEDs peaked longward); regions around strong stellar lines are masked, with O stars filling in hydrogen-line regions masked in A dwarfs at short wavelengths, G stars filling in segment 1C, and asteroids and G stars filling in masked regions for segments 2A–3C. Individual measurements agree well and are combined via sigma clipping, with residual bad-pixel spikes removed by inspection. The authors note that the quadratic asteroid model is validated by the excellent agreement between asteroid- and star-derived PFPCs—an assumption that holds only because these Themis asteroids have effectively featureless mid-IR continua.

Properties of the fixed pattern noise

The PFPCs exhibit strong structure at the expected 1–2% level, varying markedly with channel, grating, and dither position. Channel 1 shows the largest variation between dither positions; channel 2 and segment 4C show the largest amplitudes. After applying the pipeline residual fringe correction to the PFPCs themselves, significant structure remains, attributable to sampling artifacts (notably pixel-phase effects in channel 1, dither 1 where part of the PSF falls outside the IFU field of view), flat field S/N limits, interpolation artifacts over strong lines inherited from the NGC 7027-derived flat field (e.g., near bright S IV emission at 10.5 $\mu$m), and large-scale ripples at wavelengths including 5.8, 6.8, 12.3, and 20.2 $\mu$m whose origins remain unidentified.

Quantitatively, the PFPC S/N sets a hard ceiling on achievable MRS point source S/N. Without any residual fringe correction—applicable to targets whose line-rich spectra preclude the pipeline step—the limiting S/N ranges from 100–300 per segment. Where the residual fringe correction can be applied, the limiting S/N is 180–430. These are segment averages; local values vary within segments.

Application and validation

Application requires non-standard processing: extraction of individual-dither spectra, division by the dither-specific PFPC, multiplicative leveling to the four-dither mean, sigma-clipped averaging, and optionally a final residual fringe correction on the average. All steps are implemented in the public MRS-PFPC package (pfpc_proc, pfpc_cor, pfpc_plot).

Validation uses HD 163466, an A6V flux monitor excluded from the PFPC construction. Its individual dither spectra reproduce the PFPC structure closely, and the corrected spectra show dramatically less fixed pattern noise than default pipeline products. Coadding 25 epochs from cycles 1–4 reveals further differences: pipeline spectra require segment overlap corrections of 0.91–1.03 and show continuum "wiggles," whereas PFPC spectra require overlap corrections of only 0.96–1.00 and match the CALSPEC model shape closely. Notably, the HI 7–6 line at 13.37 $\mu$m shows core emission absent from the LTE-based CALSPEC model, seen in both pipeline and PFPC reductions and in other A dwarfs but not G dwarfs—possible evidence of non-LTE effects in A dwarf atmospheres, relevant since CALSPEC models assume LTE.

A particularly consequential result concerns the 5.8 $\mu$0m feature previously reported as a candidate carbonyl dust absorption band. In the PFPC-corrected spectrum of this low-extinction sightline the feature is entirely absent, and the dither-resolved PFPCs show it appearing in dithers 3 and 4, weakly in dither 1, and not in dither 2. The paper concludes this feature is fully instrumental. This directly undermines any astrophysical interpretation of the 5.8 $\mu$1m absorption in MRS spectra, with implications for dust composition studies using these data.

Signal-to-noise performance

Measured in visually clean, line-free regions, the S/N improvements depend on whether the residual fringe correction is applicable:

Regime	Improvement	Notes
No residual fringe correction possible	up to ~8$\mu$2	Largest in channel 2 and segment 1C (factors 2–6 for single epochs)
Residual fringe correction applicable	up to ~3$\mu$3	Generally 1–2$\mu$4; largest in channel 2
Residual fringe correction applied after PFPC	up to ~2$\mu$5	Channel 1 only; negligible at longer wavelengths

The delivered S/N reaches 1000 or more at shorter wavelengths where the residual fringe correction can also be used. For coadds, the improvement saturates below photon-noise expectations: channel 1A improves by a factor of two from single epoch (~600) to the 25-epoch coadd (~1200) rather than the factor of five expected from photon statistics alone, indicating the coadded S/N is limited by the PFPC's own S/N. A controlled test supports this interpretation—adding cycles 1–2 HD 163466 observations to the PFPC inputs raised the coadded S/N of the independent cycles 3–4 epochs above the original 25-epoch coadd. More calibration observations therefore translate directly into higher achievable S/N, though the final PFPCs deliberately exclude HD 163466 to avoid bias toward one star. No significant time dependence is found across cycles 1–4 in channels 1–3, suggesting the time-dependent throughput correction is adequately handled and that epoch-to-epoch scatter arises from uncorrected effects such as variable cosmic ray rates.

Limitations and open questions

Several constraints bound the applicability of these results. The PFPCs are tied to jwst pipeline version 2.0.0 and its specific reference files and algorithms; updates to flats or calibration steps will require regenerated corrections, which the MRS-PFPC package is designed to version-track. The corrections apply only to the default four-point dither pattern with target acquisition; extension to other patterns awaits suitable calibration observations. Sub-pixel offsets from nominal dither positions—driven largely by MRS grating wheel non-repeatability—are not accounted for, and the residual fringe behavior in channels 1 and 2 suggests such offsets matter there specifically; incorporating mini-grid observations could address this. The origin of the large-scale ripples at several wavelengths remains unexplained, as does the anomalously low S/N of the cycle 2 seventh epoch of HD 163466, which was excluded pending investigation. Whether and how PFPCs could be incorporated into the standard pipeline is left open.

Conclusion

This work provides empirically derived, dither-specific fixed pattern corrections for MRS point source spectroscopy covering the full 5–28 $\mu$6m range, built from a deliberately heterogeneous set of O, A, and G dwarfs and featureless asteroids so that stellar line regions are filled by complementary calibrators. The corrections raise the practical S/N ceiling of MRS point source spectra from roughly 100–430 to above 1000 at short wavelengths, deliver improvements up to nearly 8$\mu$7 for line-rich sources where the pipeline residual fringe correction cannot be used, and demonstrably remove broad instrumental artifacts—including establishing the instrumental origin of the 5.8 $\mu$8m feature. The main caveats are the pipeline-version dependence of the corrections, restriction to the default dither pattern, and unmodeled sub-pixel pointing sensitivity, particularly in channels 1 and 2.