Hot Rocks Survey: MIRI 15µm Exoplanet Atmospheres

Updated 9 July 2026

Hot Rocks Survey is a JWST General Observer program that uses 15µm secondary-eclipse photometry to assess if rocky exoplanets orbiting M dwarfs retain secondary atmospheres or are bare rocks.
The program employs repeated eclipse measurements with advanced pixel-level modeling, including Gaussian process detrending and Bayesian model averaging, to robustly extract dayside brightness temperatures and flux ratios.
Its comparative design across varied irradiation environments helps establish the atmospheric retention boundary and calibrates the limits of single-band thermal photometry.

Hot Rocks Survey is a JWST General Observer program (GO 3730) that uses MIRI secondary-eclipse photometry at $15\,\mu\mathrm{m}$ to test whether nine rocky exoplanets orbiting M dwarfs retain secondary atmospheres or instead are consistent with low-albedo bare rocks (August et al., 2024, Allen et al., 19 Aug 2025). The survey is motivated by the observational uncertainty of terrestrial atmospheres around M dwarfs: transmission studies of rocky planets have often yielded flat or ambiguous spectra, with degeneracies among high-mean-molecular-weight atmospheres, clouds, or no atmosphere at all, and M-dwarf stellar heterogeneity can further confuse interpretation. The survey therefore emphasizes thermal emission at $15\,\mu\mathrm{m}$ , where an atmosphere should generally reduce the observed eclipse depth by redistributing heat to the nightside, and where CO $_2$ has strong opacity (Allen et al., 19 Aug 2025).

1. Scientific rationale and survey design

The Hot Rocks Survey is framed around a specific atmospheric-evolution problem: close-in rocky planets around M dwarfs are expected to lose primordial H/He envelopes, but the survivability of heavier secondary atmospheres remains uncertain (Valdés et al., 25 Mar 2025, August et al., 2024). M dwarfs are simultaneously attractive and difficult hosts for this test. They are small, cool, and common, so the relative planet-to-star signal is enhanced, but they also have long and active pre-main-sequence evolution, extended X-ray saturation, frequent flaring, intense XUV irradiation, and strong stellar winds, all of which can drive atmospheric escape (August et al., 2024).

Within that framework, the survey is not a single-target program but a comparative one. It observes targets across a range of irradiation environments in order to identify where atmospheres begin to appear and which parameters correlate with atmospheric onset (Allen et al., 19 Aug 2025). One paper describes the sample as spanning equilibrium temperatures of roughly $420$–$910$ K (Fortune et al., 28 May 2025). Another emphasizes that LTT 3780 b anchors the extreme hot end, while LHS 1140c anchors the cool end (Allen et al., 19 Aug 2025, Fortune et al., 28 May 2025).

The program’s reliance on secondary eclipses in MIRI/F1500W is methodologically deliberate. A deep $15\,\mu\mathrm{m}$ eclipse is expected from a dark, inefficiently redistributing bare surface, whereas a shallower eclipse can arise from atmospheric heat transport and/or CO $_2$ absorption in the bandpass (Allen et al., 19 Aug 2025, Holmberg et al., 2 Apr 2026). This observing strategy also made the survey a “relevant primer” for the later Director’s Discretionary Time Rocky Worlds program (August et al., 2024).

2. Observational architecture and inference framework

All published Hot Rocks Survey installments use JWST/MIRI imaging in F1500W and rely on repeated occultation measurements rather than single visits. The program extracts eclipse depths, converts them to a dayside brightness temperature or dayside flux ratio, and then compares the result to forward models for bare-rock surfaces and idealized atmospheres (August et al., 2024, Valdés et al., 25 Mar 2025, Allen et al., 19 Aug 2025).

A recurring element is the conversion between eclipse depth and dayside temperature. One survey paper writes the band-integrated planet/star flux ratio as

$\frac{F_p}{F_s} = \left(\frac{R_p}{R_s}\right)^2 \frac{\int \frac{\pi B_p(T_{day}, \lambda) M(\lambda)}{hc/\lambda} d\lambda}{\int \frac{F_\star(T_\star, \lambda) M(\lambda)}{hc/\lambda} d\lambda},$

with $M(\lambda)$ the MIRI/F1500W throughput and $F_\star$ an interpolated stellar spectrum (Allen et al., 19 Aug 2025). Bare-rock comparisons then use a dayside temperature scaling in which $15\,\mu\mathrm{m}$ 0 corresponds to a no-atmosphere, no-redistribution flux-weighted dayside average (Allen et al., 19 Aug 2025).

Atmospheric modeling across the survey has been centered on CO $15\,\mu\mathrm{m}$ 1 and H $15\,\mu\mathrm{m}$ 2O, typically using HELIOS radiative-convective calculations with cloud-free assumptions and parameterized heat redistribution (Allen et al., 19 Aug 2025, Valdés et al., 25 Mar 2025, Holmberg et al., 2 Apr 2026). This is not incidental: at $15\,\mu\mathrm{m}$ 3, CO $15\,\mu\mathrm{m}$ 4-rich atmospheres are expected to produce especially strong suppression of dayside flux (Allen et al., 19 Aug 2025). The survey’s interpretation is therefore not limited to “atmosphere versus no atmosphere,” but often specifically “CO $15\,\mu\mathrm{m}$ 5-rich secondary atmosphere versus bare rock.”

A major technical theme is that the inferential bottleneck is not only photon noise. The published papers repeatedly show that MIRI detector settling, extraction choices, visit-to-visit systematics, and correlated noise modeling can materially affect the ease of interpretation, even when the eclipse depth itself is comparatively robust (August et al., 2024, Fortune et al., 28 May 2025, Allen et al., 19 Aug 2025, Holmberg et al., 2 Apr 2026).

3. Published targets and principal observational results

The first five survey installments cover LHS 1478 b, TOI-1468 b, LHS 1140c, LTT 3780 b, and GJ 3473 b (August et al., 2024, Valdés et al., 25 Mar 2025, Fortune et al., 28 May 2025, Allen et al., 19 Aug 2025, Holmberg et al., 2 Apr 2026). Together they establish the survey’s current empirical basis.

Planet	F1500W eclipse result	Main interpretation
LHS 1478 b	$15\,\mu\mathrm{m}$ 6 ppm	Possible shallow eclipse; majority of atmospheric scenarios considered are consistent, but two visits are inconsistent (August et al., 2024)
TOI-1468 b	$15\,\mu\mathrm{m}$ 7 ppm	Mostly consistent with no atmosphere and zero Bond albedo at 1.65- $15\,\mu\mathrm{m}$ 8; pure CO $15\,\mu\mathrm{m}$ 9 or H $_2$ 0O atmospheres above 1 bar ruled out over 3- $_2$ 1 (Valdés et al., 25 Mar 2025)
LHS 1140c	$_2$ 2 ppm	$_2$ 3 eclipse; highly consistent with a low-albedo bare rock (Fortune et al., 28 May 2025)
LTT 3780 b	$_2$ 4 ppm	Consistent with thermal emission from a bare rock surface; CO $_2$ 5-based atmospheres down to 0.01 bar ruled out at $_2$ 6 (Allen et al., 19 Aug 2025)
GJ 3473 b	$_2$ 7 ppm	Atmospheric and bare-rock interpretations both remain consistent; thick CO $_2$ 8 atmospheres excluded (Holmberg et al., 2 Apr 2026)

The individual papers sharpen this comparison. LHS 1478 b, the first published target, produced a possible shallow eclipse at the expected time for a circular orbit, but one of the two visits was a non-detection because of significantly larger unexplained systematics (August et al., 2024). That paper therefore treated the result as tentative rather than definitive.

TOI-1468 b produced three eclipses with individual depths of $_2$ 9 ppm, $420$0 ppm, and $420$1 ppm, and a joint depth of $420$2 ppm. The measured thermal emission was mostly consistent with no atmosphere and zero Bond albedo at 1.65-$420$3, while pure CO$420$4 or H$420$5O atmospheres with a surface pressure above 1 bar were ruled out over 3-$420$6 (Valdés et al., 25 Mar 2025).

LHS 1140c, the coolest and least irradiated survey target, yielded a deep eclipse detected at $420$7, with a recovered dayside brightness temperature of $420$8 K, close to the theoretical maximum of $420$9 K. The paper ruled out pure CO$910$0 atmospheres with surface pressure $910$1 mbar and pure H$910$2O atmospheres with surface pressure $910$3 bar at $910$4 (Fortune et al., 28 May 2025).

LTT 3780 b, the survey’s highest-irradiation target at $910$5 Earth’s instellation, delivered a joint eclipse depth of $910$6 ppm. The inferred dayside temperature was $910$7 K, corresponding to $910$8 of the maximum no-redistribution temperature. That paper confidently ruled out CO$910$9-dominated atmospheres down to 0.01 bar surface pressure to greater than $15\,\mu\mathrm{m}$ 0 (Allen et al., 19 Aug 2025).

GJ 3473 b produced a secure average eclipse detection of $15\,\mu\mathrm{m}$ 1 ppm from four visits, but unlike the clearer bare-rock cases, both atmospheric and bare-rock interpretations remained consistent with the data. The paper nevertheless excluded thick CO $15\,\mu\mathrm{m}$ 2 atmospheres, placing a 95% credible upper limit of 1.2–6.5 bar on the surface pressure (Holmberg et al., 2 Apr 2026).

Taken together, the published installments suggest that the survey is finding multiple planets on the “bare rock” side of the atmospheric-retention boundary, but with at least one ambiguous case and one tentative shallow-eclipse case that remains unresolved.

4. Comparative interpretation across the sample

The survey papers place strong weight on comparative structure, not isolated detections. LTT 3780 b is explicitly treated as a “key anchor” at the hot extreme because it is the most highly irradiated target in the 9-planet sample and therefore a stringent test of whether highly heated rocky M-dwarf planets can sustain volatile envelopes at all (Allen et al., 19 Aug 2025). Its dayside temperature, at essentially the maximum expected for a synchronously rotating low-albedo airless planet, situates it clearly on the “bare rock” side of the emerging boundary (Allen et al., 19 Aug 2025).

At the other end of the published sample, LHS 1140c matters because it is the coolest and least irradiated target in the survey, yet it also appears highly consistent with a low-albedo bare rock (Fortune et al., 28 May 2025). This does not by itself define the retention boundary, but it pushes the bare-rock regime to lower irradiation than the hotter archetypes.

TOI-1468 b reinforces that pattern. Its eclipse is again consistent with a hot bare rock and inconsistent with substantial cloud-free CO $15\,\mu\mathrm{m}$ 3- or H $15\,\mu\mathrm{m}$ 4O-rich atmospheres at and above the 1 bar level, although the paper also notes that the planet appears marginally hotter than expected and therefore discusses thermal inversion, induction heating, and instrumental artifact as possible explanations (Valdés et al., 25 Mar 2025).

The main deviation from a simple bare-rock narrative is LHS 1478 b, where the first observation favored a shallow eclipse and the majority of atmospheric scenarios considered were consistent with the data, spanning CO $15\,\mu\mathrm{m}$ 5-rich atmospheres with surface pressures from 0.1 to 10 bar (August et al., 2024). However, because the two observations did not yield consistent results, that paper explicitly stressed that more observations were needed (August et al., 2024).

GJ 3473 b is the clearest published demonstration that a secure eclipse does not automatically imply a secure atmospheric classification. Its average depth is somewhat lower than expected for a blackbody, but the overlap between reflective/weathered bare-rock surfaces and thin or moderate atmospheres leaves the atmosphere question unresolved (Holmberg et al., 2 Apr 2026). A plausible implication is that the survey is not only mapping a physical boundary, but also identifying where single-band $15\,\mu\mathrm{m}$ 6 eclipse photometry becomes intrinsically degenerate.

5. Methodological advances and MIRI time-series systematics

One of the survey’s most substantial contributions is methodological. The papers show that MIRI time-series analysis is highly sensitive to the initial settling ramp, aperture choice, extraction weighting, and pixel-level behavior, but they also show that robust eclipse recovery is possible when these effects are handled carefully (Allen et al., 19 Aug 2025, Fortune et al., 28 May 2025, Holmberg et al., 2 Apr 2026).

Hot Rocks Survey III introduced a “novel approach” that joint-fits individual pixel light curves with a shared eclipse model and a flexible multi-dimensional Gaussian process that models changes in the PSF over time (Fortune et al., 28 May 2025). In simulated data, the method can weight away from particular pixels that show increased systematics, allowing eclipse depths to be recovered more robustly and precisely (Fortune et al., 28 May 2025). That paper also identified a potential trend between detector settling and the previous filter used by MIRI (Fortune et al., 28 May 2025).

Hot Rocks Survey V pushed the detector analysis further by identifying three distinct settling behaviors across the PSF: low-illumination edge pixels with little or no settling, highly illuminated core pixels with strong exponential-like ramps, and an intermediate annulus with exponential behavior of opposite sign to the core (Holmberg et al., 2 Apr 2026). Because different apertures weight these regions differently, the apparent light-curve ramp changes with aperture size (Holmberg et al., 2 Apr 2026). The fourth GJ 3473 b visit, which had different illumination history because the target had previously been observed in F560W, was highlighted as an especially instructive case (Holmberg et al., 2 Apr 2026).

Hot Rocks Survey IV on LTT 3780 b is notable for its exhaustive reduction-dependence analysis. The authors performed two independent reductions, explored linear, exponential, and Gaussian-process detrending models, varied the number of integrations trimmed from the start of the visit from 100 to 600, and used Bayesian model averaging over the best-supported models to incorporate systematic-model uncertainty (Allen et al., 19 Aug 2025). They emphasized that nearly all tested fits agreed within $15\,\mu\mathrm{m}$ 7, and almost all within $15\,\mu\mathrm{m}$ 8, with discrepant fits strongly disfavored by Bayesian evidence (Allen et al., 19 Aug 2025).

Hot Rocks Survey I provides the contrasting cautionary example: a plausible eclipse candidate can remain scientifically inconclusive when one visit is degraded by a systematic of the same order as the astrophysical signal (August et al., 2024). The combined lesson is that survey-level conclusions depend not only on nominal sensitivity, but also on visit repeatability, illumination history, pixel-level modeling, and multi-pipeline cross-checking.

6. Scientific significance, limits of single-band photometry, and future directions

The published Hot Rocks Survey papers collectively strengthen the case that many highly irradiated rocky planets around M dwarfs appear airless or nearly so in thermal emission (Valdés et al., 25 Mar 2025, Fortune et al., 28 May 2025, Allen et al., 19 Aug 2025). LTT 3780 b is explicitly described as a hot-boundary anchor on the “bare rock” side of the retention threshold (Allen et al., 19 Aug 2025). TOI-1468 b and LHS 1140c extend that picture to other parts of the survey parameter space (Valdés et al., 25 Mar 2025, Fortune et al., 28 May 2025).

At the same time, the survey has exposed the limits of what single-band F1500W eclipse photometry can accomplish. GJ 3473 b shows that even a confident eclipse detection can remain degenerate between atmospheric and airless interpretations, because reflective, textured, or weathered surfaces can overlap the same $15\,\mu\mathrm{m}$ 9 depth range as thin or moderate atmospheres (Holmberg et al., 2 Apr 2026). LHS 1478 b shows that shallow-eclipse interpretations can be undermined by visit-to-visit inconsistency (August et al., 2024). This suggests that the survey is not merely producing yes-or-no atmospheric classifications; it is also empirically delimiting the regimes in which single-band thermal photometry is decisive and the regimes in which it is not.

The survey papers therefore converge on a common follow-up strategy. LTT 3780 b is described as an excellent target for follow-up JWST observations because its short period makes repeated eclipse or phase-curve observations efficient, and simulated NIRSpec/G395H and MIRI/LRS follow-up could distinguish some surface compositions and probe nightside emission directly (Allen et al., 19 Aug 2025). GJ 3473 b is identified as requiring future spectroscopic or phase-curve observations to determine whether it hosts a substantial atmosphere (Holmberg et al., 2 Apr 2026). TOI-1468 b likewise identifies a full phase curve as the most direct route to determining whether an atmosphere is present (Valdés et al., 25 Mar 2025).

A broader implication is that the Hot Rocks Survey has become both a scientific survey and a calibration program for JWST rocky-world eclipse work. It has already shown that $_2$ 0 eclipse photometry can decisively rule out wide classes of CO $_2$ 1-rich and H $_2$ 2O-rich atmospheres on some targets, but it has also shown that the same observable cannot always uniquely distinguish a bare rock from a more weakly absorbing or more compositionally complex atmosphere (Fortune et al., 28 May 2025, Allen et al., 19 Aug 2025, Holmberg et al., 2 Apr 2026). In that sense, the survey’s lasting significance lies in both outcomes: it is mapping the atmospheric-retention boundary around M dwarfs, and it is defining the methodological boundary of what MIRI/F1500W eclipse photometry alone can establish.