JWST Exoplanet Atmospheres, Weekly Edition (28 Sep to 5 Oct 2026): Stellar Contamination and Retrieval Reliability Recur

Executive Summary

A recurring problem in this week's JWST exoplanet-atmosphere preprints is separating a planet's signal from its star or from analysis artifacts. Of the seven papers I reviewed, three address stellar contamination or retrieval reliability, three present new spectra (sub-Neptunes, an ultrahot Neptune, a rocky planet's surface), and one is a population-level interpretation. The headline astrophysical result is a press-reported atmosphere on the lava world HD 3167 b, the coolest lava world yet found with one. Every arXiv item below is a preprint, not yet peer reviewed, and several are explicitly under review or newly submitted. Coverage is limited to what I could open: an arXiv API search of astro-ph.EP and astro-ph.SR for papers with JWST in the abstract, submitted from 26 September, the abstract pages of seven papers, and one news report. I did not find or open a NASA/STScI press release on exoplanet atmospheres dated inside the window, and I did not query MAST for data releases, so those two parts of the brief are not covered.

Key discoveries and new spectra

Date (2026)	Target	Instrument	Finding	Status	Source
4 Oct (news)	HD 3167 b	JWST mid-infrared eclipse	Day side cooler than the maximum possible, read as strong evidence of an atmosphere	Reported in ApJ Letters (per news report; paper not opened)	SkyNightly
28 Sep	LTT 9779 b	NIRISS/SOSS + NIRSpec G395H	Muted features, clouds, cool terminator	Under review at AAS Journals	arXiv:2609.36206
28 Sep	TOI-125 b, c	NIRSpec G395H	Compatible with high metallicity or high clouds	Resubmitted to AJ after review	arXiv:2609.35607
29 Sep	LHS 3844 b	MIRI/LRS	Highest-SNR rocky-planet emission spectrum; tentative 10 to 12 micron feature	Revised version submitted to ApJL after first referee report	arXiv:2609.37982

HD 3167 b (lava world, news report): A University of Chicago-led team reported in The Astrophysical Journal Letters that the rocky super-Earth HD 3167 b, 154 light-years away, is the coldest lava world found so far with evidence of an atmosphere. The evidence is indirect: in secondary eclipse the day side is noticeably cooler than its expected maximum, which heat redistribution or reflective clouds would explain. The team says the atmosphere's composition is still undetermined. It is the first result from a program observing 10 ultra-hot lava worlds to look for a critical temperature above which atmospheres appear. I only read a secondary news report, not the paper, so treat the details as press-level. Classification: new detection (indirect), single source.

LTT 9779 b (ultrahot Neptune): The authors combine a re-reduction of NIRISS/SOSS data with new NIRSpec/G395H data to give a 0.63 to 5.165 micron spectrum. Retrievals find muted features consistent with clouds and a low terminator temperature of about 560 K for free chemistry and about 910 to 950 K for equilibrium chemistry. A cloud-metallicity degeneracy remains, and the authors say mid-infrared silicate features are needed to confirm the cloud species. Classification: new constraint on a planet whose atmosphere was already reported as confirmed (the authors note earlier results on its content differ).

TOI-125 b and c (sub-Neptunes): Both planets have NIRSpec G395H spectra with median error bars of 50 ppm (b) and 85 ppm (c). Retrievals allow either about 150x solar metallicity or clouds above about 3 mbar for b, and about 100x solar or clouds above about 10 mbar for c. The authors say this may support metal enrichment in sub-Neptunes but further complicates the proposed trend of feature size with temperature, and that shorter-wavelength follow-up is needed. Planet c's transit was only 70% covered because of larger-than-predicted transit timing variations.

LHS 3844 b (rocky, surface constraint): Eight new and three archival eclipses give the highest-SNR emission spectrum for any rocky exoplanet so far, median SNR 29 across 12 bins. The spectrum resembles a blackbody shortward of 10 microns, disfavoring a transparency feature, with a tentative feature at 1.9 to 3.5 sigma between 10 and 12 microns that could match fayalite but might be instrumental. This is a surface constraint, not an atmosphere detection.

Methods: the star is now the main error source

TRAPPIST-1 c: Rathcke and Buchhave fit four NIRSpec PRISM transits with a Gaussian process per epoch for stellar contamination, calibrated on 17 archival transits of the airless TRAPPIST-1 b. Detection thresholds are 146 ppm (moderate evidence) and 171 ppm (strong), versus 96 ppm without contamination. Planet c is consistent with a featureless, airless baseline, with a 95% upper limit of 90 ppm for CO2-shaped signals that rules out clear pure-CO2 atmospheres of at least 0.1 bar. Water shows a persistent stellar floor of +55 +/- 34 ppm. Classification: methodological advance and a non-detection limit; submission to MNRAS was still planned.
GJ 1214 b: Coordinated multi-band photometry and high-resolution spectroscopy around the JWST visits indicates the observations occurred at a favourable rotation phase with low contamination. A neural-network method (UnSPOTTER) constrains corrections to about 15 ppm. The authors warn that near activity maximum the contamination would be strongly chromatic and affect retrievals.
Retrieval reliability: Chakravertty and Raj report that amortized simulation-based inference collapsed on real WASP-39b spectra (effective sample size of 1). They attribute this to a missing planet-radius latent variable rather than missing physics, and their fix lowers chi-squared per point from 301 to 0.06. This is a six-page preprint, so independent replication is needed.

Population-level context

A single-author preprint compiles eight M-dwarf sub-Neptunes with published JWST spectra and a constrained mean molecular weight. It finds four of the eight satisfy neither the gas-dwarf nor the water-world class, with envelopes of 3.8 < mu <= 18 amu, and calls these "hybrid worlds". Only one is a strict water world. The sample is small and the author notes selection effects, so this is an interpretive argument, not a settled result. TOI-125 adds two more sub-Neptunes whose spectra are compatible with either high metallicity or clouds, and the authors say shorter-wavelength data are needed to tell whether they are truly muted. No paper this week settles that.

What would change this picture

A peer-reviewed HD 3167 b paper with explicit retrieved day-side temperatures would test how strong the "atmosphere" inference is; I did not read it.
Mid-infrared silicate features for LTT 9779 b, and shorter-wavelength data for TOI-125 b and c, are the follow-ups the authors themselves request.
More TRAPPIST-1 c transits would lower the CO2 threshold; the authors' scaling suggests 80 ppm at ten transits.