Helium escaping from the atmosphere of a nearby rocky exoplanet orbiting in a habitable zone

AbstractObservations of highly irradiated gas giant exoplanets have shown helium escaping from their atmospheres. There is limited evidence for atmospheres on rocky exoplanets, perhaps because they have already escaped. We report near-infrared spectroscopic observations of LHS 1140b, a rocky exoplanet that orbits in the habitable zone of a nearby low-mass star. The transit spectra show absorption by helium escaping from the planet’s atmosphere. Helium absorption is detected in 2024 but not in 2025, indicating time-variable atmospheric escape. We interpret these results as indicating an upper atmosphere dominated by helium and depleted in hydrogen, with other volatile species trapped at lower altitudes, consistent with atmospheric fractionation models. No helium absorption is detected for LHS 1140c, a smaller and more strongly irradiated exoplanet in the same system. SIGN UP FOR THE AWARD-WINNING SCIENCEADVISER NEWSLETTER The latest news, commentary, and research, free to your inbox daily Theoretical models predict that the atmospheres of rocky exoplanets can regulate the climate, shield the surface from ionizing radiation, and enable the presence of liquid water (1, 2). Atmospheres have been observed on large, gas-rich, highly irradiated exoplanets (3, 4). Observing atmospheres on smaller, cooler, rocky exoplanets is technically challenging because they are dwarfed (in size and brightness) by the stars that they orbit. Those challenges can be reduced by studying planets that orbit red dwarf stars (M dwarfs), whose small sizes and low brightness reduce the contrast between the star and any orbiting planet. However, M dwarfs are more active than Sun-like stars, emitting high-energy radiation that can drive atmospheric escape from closely orbiting planets. It is therefore unclear whether such planets can retain their atmospheres for billions of years (5, 6).

Observations of small, rocky exoplanets have mostly revealed airless worlds or atmospheres too tenuous to detect (7, 8), with some debated evidence for atmospheres (9–11).The LHS 1140 systemThe transiting rocky exoplanet LHS 1140b has a mass of 5.60 ± 0.19 Earth masses (M⊕) and a radius of 1.730 ± 0.025 Earth radii (R⊕). These values are consistent with an Earth-like bulk composition with an additional a low-density component, such as an atmosphere or a high abundance of water. It has an orbital period of 24.7 days and receives 42% of the stellar irradiation received by Earth, giving it an equilibrium temperature Teq = 226 ± 4 K (assuming zero albedo), placing it in the liquid-water habitable zone (12, 13). There is another transiting rocky planet in the same system, LHS 1140c (1.91 ± 0.06M⊕ and 1.272 ± 0.026 R⊕) with an orbital period of 3.78 days, which receives about five times the irradiation received by Earth (14). The host star LHS 1140 (also cataloged as GJ 3053) is an old [> 3 Gyr (12, 13, 15,)] inactive (12, 16) M dwarf located 14.96 ± 0.01 parsecs from the Sun (13).Spectroscopic observations of LHS 1140We observed the LHS 1140 system using the Warm Infrared Echelle Spectrograph to Realize Extreme Dispersion (WINERED) mounted on the Magellan Clay telescope at Las Campanas Observatory, as part of the WINERED Helium Consortium project. On 2024 September 23, we observed the system for 6.5 hours, covering one transit of each planet, separated by 39 min.

A total of 70 spectra were collected in total: 35 out-of-transit, 12 during the transit of LHS 1140c, and 23 during the transit of LHS 1140b. We used the WINERED Automatic Reduction Pipeline [WARP (17)] for the initial data reduction (18). To construct time series spectra and search for excess absorption (wavelength-dependent absorption of stellar radiation by the planet’s atmosphere), we divided each spectrum by a stellar template Fout, the mean stellar flux of all out-of-transit spectra (Fig. 1A). The resulting time series (Fig. 1B) shows an absorption feature near 10,833 Å during the transit, pre-ingress, and post-egress of LHS 1140b. This feature is consistent with the presence of metastable helium.Fig. 1. Time series spectra for LHS 1140b observed in 2024.(A) The average out-of-transit stellar template spectrum, constructed from exposures with no helium absorption apparent in the time series (see Fig. 3). Arrows indicate absorption features from Earth’s atmosphere that are masked in (B). (B) Time series spectra of LHS 1140b in the stellar rest frame. Colors indicate the percentage difference from the stellar template in (A). Horizontal lines enclose the expected transits of LHS 1140b (black dashed) and LHS 1140c (white dotted). The vertical dashed white lines indicate the expected positions of helium absorption lines, moving with the same velocity as LHS 1140b. Cross hatching indicates data that were excluded due to Earth’s atmospheric features (A) and a single noisy exposure.We produced a planetary transmission spectrum of LHS 1140b (Fig. 2) by computing the mean of all in-transit excess absorption spectra in the planetary rest frame (18). This transmission spectrum contains correlated noise, which we modeled using a Gaussian process [GP; (18)].

Metastable helium is expected to produce a triplet of closely spaced absorption lines, which we modeled with three Gaussian profiles at 10,832.057, 10,833.217 and 10,833.306 Å [rest wavelengths in vacuum (19)]; the latter two lines are blended at the resolution of the WINERED spectra. We used a Markov chain Monte Carlo (MCMC) analysis to fit the model to the data and determine the uncertainties in our measurements. The MCMC process followed a Bayesian retrieval framework with five free parameters: the three peak amplitudes, a shared peak width, and a shared Doppler shift (18).Fig. 2. Raw transmission spectrum for LHS 1140b in 2024.Excess absorption (black dots) is plotted as the mean of the in-transit spectra shown in Fig. 1. Pre-ingress and post-egress spectra have been excluded, even when they contain evidence for helium absorption. The vertical blue lines indicate the rest wavelengths of the helium absorption lines. The horizontal gray line indicates zero absorption. The gap is due to the data excluded in Fig. 1B. Error bars show 1σ uncertainties.We report the median values and 16 to 84% confidence intervals of the MCMC posterior probability distributions. The excess absorption depth is 1.24−0.23+0.22% at the position of the two blended long-wavelength peaks and 0.25−0.12+0.14% for the single short-wavelength peak, with a Doppler shift of 0.072−0.073+0.080 Å relative to the planet velocity, equivalent to 2.0−2.2+2.0 km s−1. This corresponds to an equivalent opaque radius (the planetary radius including an opaque atmospheric layer that would produce the observed absorption feature) of 1.52 times the radius of LHS 1140b. The full-width at half-maximum (FWHM) of the blended helium absorption lines is 0.86−0.27+0.15 Å, corresponding to 23.9−7.5+4.2 km s−1.

The measured ratio of the blended red peak amplitude to the blue peak amplitude is 6.7−3.1+12.7. This is consistent with the ratio of 8 we expect due to the fine structure of the helium triplet (20, 21); this assumes negligible optical depth (τ≪1) in the thin upper atmosphere where metastable helium can persist.We also fitted the nine pre-ingress spectra with apparent helium absorption in the time series (Fig. 1B) using the same model and MCMC process. The resulting absorption depth is 1.01−0.34+0.31% and the Doppler shift is 0.3−7.5+3.5 km s−1. This is consistent with escaping helium ahead of the planet in its orbit (a leading tail). Leading tails have been previously observed for some gas giant exoplanets with escaping atmospheres (22–24) and interpreted as resulting from stellar winds or interactions between the magnetic fields of the star and planet (25, 26). We also performed the same analysis for the eight post-egress spectra, finding an absorption depth of ≤0.76% (1σ upper limit) and a Doppler shift of 0.53−0.42+0.33 Å, equivalent to 14.7−9.0+11.8 km s−1, which we regard as tentative evidence of a trailing tail. We calculated the mean helium absorption between 10,833 and 10,834 Å to construct a transit light curve of the blended red absorption lines (Fig. 3A). The transmission spectra of the LHS 1140b transit, leading tail, and possible trailing tail are shown in Fig. 3, B to D.Fig. 3. Helium absorption as a function of time and line profiles observed in 2024.(A) Mean helium absorption, calculated from 10,833 to 10,834 Å, as a function of time, measured relative to the mid-transit time (t0) of LHS 1140b. Gray points are all observed spectra, and black points have been binned by a factor of three (17 min).

The pink shaded regions indicate the data used to construct the stellar template (Fig. 1A). The blue shaded regions show the pre-ingress and post-egress data used to construct the transmission spectra shown in (C) and (D), which were chosen by eye from Fig. 1B. Vertical lines show the expected transit times of LHS 1140b (gray) and LHS 1140c (green). (B) The transmission spectrum constructed from the in-transit data, (C) pre-ingress data, and (D) post-egress data. In each panel, black dots show the data after subtraction of the Gaussian process model. Thick blue lines show the best-fitting absorption line models, which have the labeled values of the absorption depth δ and Doppler shift Δλ. Thin purple lines show 100 random samples drawn from the MCMC fitting process. All error bars show 1σ uncertainties.We repeated the same analysis for the transit of LHS 1140c in 2024 (see supplementary text), but found no evidence of helium absorption (fig. S1). The transit of LHS 1140b in 2025 also shows no evidence of helium absorption (Fig. 4 and fig. S2). To investigate whether these differences between transits could be explained by choices in the data reduction process, the 2024 and 2025 datasets for LHS 1140b were independently re-reduced using a different reduction code; we found no difference in the results [fig. S3; (18)].Fig. 4. Comparison between the 2024 and 2025 observations of LHS 1140b.Data points are the raw transmission spectrum in 2025 (gray points) and the GP-subtracted transmission spectrum in 2024 (black points, as in Fig. 2). Error bars show 1σ uncertainties. (A) Colored lines are p-winds models with different XUV flux (see legend) and escape rates (assumed to be proportional to the XUV flux), compared to the best-fitting model of the 2024 data (dark blue line).