No Persistent Helium Absorption in LHS 1140 b: Four JWST/NIRISS SOSS Transits and Multi-epoch Stellar He I Variability

Abstract: The search for atmospheres on temperate terrestrial planets is important for understanding how these objects form and evolve, and their potential habitability. Recent transit observations of LHS 1140 b, a temperate ($T_{\rm eq}=226$ K) planet straddling the radius valley ($R_{\rm p}\approx1.7 R_\oplus$), with the WINERED high-resolution spectrograph yielded a detection of planetary atmospheric escape through the measurement of excess absorption ($1.24\pm0.23\%$) in the metastable helium triplet, although a later second visit resulted in a non-detection. We analyze four transits of LHS 1140 b and two of LHS 1140 c observed with JWST/NIRISS SOSS. We find no evidence of helium absorption in either planet, with all amplitudes consistent with zero within $1σ$. For LHS 1140 b, we derive $3σ$ upper limits of $0.72$--$1.21\%$. Individual visits disfavor the reported WINERED absorption at $2.6$--$3.7σ$, while their joint constraint disfavors a persistent signal at $4.5σ$. We find that the stellar He I line of LHS 1140 varies substantially between NIRPS and WINERED epochs, with both its strength and fractional variability consistent with the behavior of other M dwarfs. We also identify a moderate correlation between the 2024 He I depth and seeing, suggesting a possible seeing-dependent instrumental contribution. If the high-resolution detection is indeed planetary, the rate of such atmospheric loss events must be relatively low ($f=22_{-12}<sup>{+17}\%$). Alternatively, stellar He I variability may contribute to the reported excess absorption. Additional high-resolution observations, both in and out of transit, are required to distinguish between these scenarios.

• The paper finds no metastable helium absorption in four JWST/NIRISS SOSS transits, with a combined depth of 0.05 ± 0.13% that rules out a persistent prior signal at 4.5σ.
• The analysis also detects no helium absorption or extended transit tails from LHS 1140 c and constrains strong escape events on planet b to an estimated occurrence rate of 22^{+17}_{−12}%.
• Multi-epoch spectroscopy reveals stellar He I variability spanning roughly 20–50 mÅ, while a moderate correlation with seeing suggests stellar activity and instrumental effects may have contributed to the earlier detection.

Context and motivation

LHS 1140 b is a temperate ($T_\mathrm{eq} = 226$ K) super-Earth with $R_\mathrm{p} \approx 1.7\,R_\oplus$ and $M_\mathrm{p} = 5.6\,M_\oplus$, orbiting an M4.5V star every 24.7 days. Its density is slightly below a purely rocky composition, consistent with either a water world or a mini-Neptune, and prior NIRISS/SOSS and NIRSpec/G395H transmission spectra showed no atmospheric features (Cadieux et al., 2024, Damiano et al., 2024). Against this backdrop, Cherubim et al. (2026) reported excess absorption of $1.24 \pm 0.23\%$ in the metastable He I triplet at 10833 Å during one WINERED transit of LHS 1140 b on the 6.5 m Magellan Clay telescope — a detection that would constitute atmospheric escape from a habitable-zone rocky planet and support a helium-world scenario driven by escape-induced mass fractionation. A second WINERED visit yielded a non-detection (upper limit of 0.6%), attributed by those authors to temporal variability of the escape.

This Letter tests the persistence of that signal using four NIRISS/SOSS transits of LHS 1140 b (two from DDT 6543 in December 2023, two from GO 7073 in August 2025 and July 2026), two transits of the inner planet LHS 1140 c, and multi-epoch high-resolution stellar spectroscopy from NIRPS and WINERED.

Observations and data reduction

The five SOSS visits used SUBSTRIP256 with NISRAPID readouts, covering orders 1 and 2 (0.6–2.8 μm). The first DDT visit suffered a target-acquisition failure that displaced the traces by $-157$ pixels along dispersion and $-12$ pixels cross-dispersion; the authors recovered the wavelength solution for this visit by cross-correlating its median spectrum against the correctly positioned second visit and remapping the PASTASOSS solution. Reductions used the transitspectroscopy pipeline with time-series-optimized jump detection, 15-pixel box apertures, and $5\sigma$ outlier replacement. Spectroscopic light curves at native pixel resolution were fitted with juliet/batman/dynesty under two treatments: a conservative fit with free limb-darkening coefficients and a Matérn-3/2 Gaussian process, and a fiducial fit with limb darkening fixed from the lowest-scatter visit and a linear temporal slope. Orbital parameters were fixed to a joint JWST/RV analysis.

Helium analysis methodology

Two complementary searches were performed on 21–22 pixel channels spanning 1.073–1.093 μm. The low-resolution search modeled the triplet as a Gaussian at SOSS resolution ($R=600$) and fitted its peak amplitude $A_\mathrm{LR}$ directly to the transmission spectrum. The WINERED-informed fit instead assumed the intrinsic line profile reported by Cherubim et al. (FWHM = 0.86 Å), convolved it with the SOSS line-spread function, and scaled its bin-integrated template to infer an intrinsic depth $A_\mathrm{He}$. Negative amplitudes were permitted to avoid positive-boundary bias. Injection–recovery tests confirmed that a W24-like profile would be recoverable above $R_\mathrm{p} \approx 1.7\,R_\oplus$0 in each visit (recovery probability 63.8% for the noisiest visit, >95% for the other three).

Non-detection of planetary helium absorption

No visit shows evidence of He I absorption: all inferred amplitudes are consistent with zero within $R_\mathrm{p} \approx 1.7\,R_\oplus$1. Under the fiducial treatment, intrinsic amplitudes range from $R_\mathrm{p} \approx 1.7\,R_\oplus$2 to $R_\mathrm{p} \approx 1.7\,R_\oplus$3, with $R_\mathrm{p} \approx 1.7\,R_\oplus$4 upper limits of 0.72–1.21%; the conservative treatment gives consistent results. Individual visits disfavor the W24 amplitude at $R_\mathrm{p} \approx 1.7\,R_\oplus$5–$R_\mathrm{p} \approx 1.7\,R_\oplus$6, with three of four exceeding $R_\mathrm{p} \approx 1.7\,R_\oplus$7. Multiplying the independent posteriors under the assumption of a constant intrinsic depth yields $R_\mathrm{p} \approx 1.7\,R_\oplus$8, disfavoring a persistent W24-level signal at $R_\mathrm{p} \approx 1.7\,R_\oplus$9. This is the central quantitative result: if the 2024 WINERED detection is real and planetary, it cannot be a steady feature of the planet's transmission spectrum.

The two transits of LHS 1140 c likewise show no helium absorption, with $M_\mathrm{p} = 5.6\,M_\oplus$0 upper limits of 1.02% and 0.58%. Inspection of one- and two-pixel light curves centered on the triplet reveals no coherent flux decrease during the extended pre-transit (0.84 hr) or post-transit (0.75 hr) intervals interpreted as escaping-helix tails by Cherubim et al., removing support for temporally extended absorption as well. An Hα flare search in order 2 across four visits found no enhancement both localized to Hα and coherent in time, though the authors caution that order 2 has lower SNR and field-star contamination in some visits.

Under a binomial model with one W24-like event among six transits, the occurrence rate of such strong escape events is constrained to $M_\mathrm{p} = 5.6\,M_\oplus$1. If the WINERED detection is planetary, significant atmospheric loss from LHS 1140 b must therefore be episodic rather than persistent.

Stellar He I variability

The authors reanalyzed the out-of-transit WINERED stellar templates from 2024 September 23 and 2025 September 29, obtaining mutually consistent He I equivalent widths of $M_\mathrm{p} = 5.6\,M_\oplus$2 and $M_\mathrm{p} = 5.6\,M_\oplus$3 mÅ. Twenty-nine archival NIRPS commissioning spectra from late 2022 yield a mean EW of 23.7 mÅ with night-to-night dispersion of 3.5 mÅ; the WINERED measurements exceed this by roughly $M_\mathrm{p} = 5.6\,M_\oplus$4–$M_\mathrm{p} = 5.6\,M_\oplus$5, providing tentative evidence for epoch-to-epoch variability. Adding the W24 planetary excess ($M_\mathrm{p} = 5.6\,M_\oplus$6 mÅ) gives an in-transit EW of $M_\mathrm{p} = 5.6\,M_\oplus$7 mÅ, $M_\mathrm{p} = 5.6\,M_\oplus$8 above the NIRPS average. Across all available epochs, the stellar He I EW of LHS 1140 spans approximately 20–50 mÅ.

This variability is placed in context using the CARMENES M-dwarf sample of Fuhrmeister et al.: mean He I strength decreases toward cooler effective temperatures while fractional variability increases, and LHS 1140 follows both trends. High-cadence Sun-as-a-star NIRPS observations show solar He I variations of ~10% peak-to-peak on several-hour timescales comparable to a transit duration, whereas the W24 event corresponds to a ~30% fractional increase. The authors note plainly that whether fully convective mid-M dwarfs can exhibit such short-timescale variability remains unknown, so a stellar origin for the W24 signal is plausible but not demonstrated.

Possible instrumental contribution

Because Cherubim et al. adopted a fixed instrumental resolution, seeing-dependent illumination of the 0.3″ WINERED slit could modulate the measured depth of the narrow He I line. Using spatial FWHM measurements of the 55 usable integrations from the 2024 sequence (seeing range ~0.57–0.97″), the authors find a moderate positive correlation between seeing and He I excess ($M_\mathrm{p} = 5.6\,M_\oplus$9), with a slope differing from zero at $1.24 \pm 0.23\%$0. Correcting to the mean seeing reduces the inferred pre/in/post-transit excess relative to the out-of-transit baseline by ~20%, from $1.24 \pm 0.23\%$1 to $1.24 \pm 0.23\%$2. The large reduced chi-square ($1.24 \pm 0.23\%$3) indicates underestimated error bars by a factor of ~3.7 and that this simple relation does not capture the full time-series behavior. The authors are careful to state that this dependence does not establish an instrumental origin, only that seeing-dependent effects may contribute to the measured contrast. They also note that the sensitivity of the inferred contrast to the choice of out-of-transit baseline — which excluded the pre- and post-transit intervals interpreted as extended planetary absorption — further affects the reported amplitude.

Limitations and open questions

Several caveats bear directly on the interpretation. The joint $1.24 \pm 0.23\%$4 constraint assumes the intrinsic helium depth was constant across the four SOSS epochs; genuinely intermittent escape at the W24 level would evade this test, and the binomial occurrence-rate estimate rests on treating six heterogeneous observations (two instruments, different precision) as independent draws. The CARMENES comparison sample contains no stars as cool as LHS 1140 ($1.24 \pm 0.23\%$5 K versus the sample limit of ~3300 K), so the stellar-variability argument relies on extrapolation rather than direct calibration at the relevant temperature. The Hα diagnostic is limited by order-2 SNR and contamination. Finally, quantifying the instrumental contribution would require characterizing the time-dependent spectral resolution of both WINERED sequences, which is beyond the scope of this work; an equivalent-width-based measurement, less sensitive to resolution changes, is proposed as a complementary test. The paper leaves open whether fully convective M dwarfs vary in He I on transit-comparable timescales, and whether new high-resolution observations in and out of transit can separate intermittent planetary escape from stellar or instrumental contributions.

Conclusion

Four JWST/NIRISS SOSS transits of LHS 1140 b show no metastable helium absorption, jointly disfavoring a persistent W24-level signal at $1.24 \pm 0.23\%$6; the same holds for LHS 1140 c. Multi-epoch NIRPS and WINERED spectroscopy reveals substantial stellar He I variability in LHS 1140 itself, consistent with trends in warmer M dwarfs, and a moderate seeing correlation in the 2024 WINERED data suggests a possible instrumental contribution to the originally reported excess. Three interpretations remain viable — intermittent planetary escape at a rate of $1.24 \pm 0.23\%$7, stellar He I variability, or instrumental systematics — and distinguishing among them requires additional high-resolution observations spanning both in- and out-of-transit phases together with deeper reanalysis of the WINERED data.