Are there Spectral Features in the MIRI/LRS Transmission Spectrum of K2-18b?

Jake Taylor · 2025

Read Are there Spectral Features in the MIRI/LRS Transmission Spectrum of K2-18b? by Jake Taylor (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-15916.txt.

Introduction

Jake Taylor re-examines, in a 2025 solo-authored paper, claims of biosignature gases in the atmosphere of the exoplanet K2-18b. A recent study reported a 3.4-sigma hint of spectral features, possibly dimethyl sulfide, in telescope data from the MIRI instrument.

Taylor fitted Gaussian bumps of the expected shape in six different ways. Five of the six tests preferred a plain flat line over any feature. Centering the bump exactly where the two candidate gases absorb flipped the verdict mildly, but only at about two sigma. His verdict: no strong statistical evidence for spectral features in this spectrum, and the extraordinary claim stays unproven.

Catalog record

CREATORJake Taylor
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE11 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-15916

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Are there Spectral Features in the MIRI/LRS Transmission Spectrum of K2-18b? Jake Taylor Abstract Determining the composition of an exoplanet atmosphere relies on the presence of detectable spectral features. The strongest spectral features, including DMS, look approximately Gaussian. Here, I perform a suite of Gaussian feature analyses to find any statistically significant spectral features in the recently published MIRI/LRS spectrum of K2-18b ( N. Madhusudhan et al. 2025 ) . In N. Madhusudhan et al. 2025 , they claim a 3.4- σ \sigma detection of spectral features compared to a flat line. In 5 out of 6 tests, I find the data preferred a flat line over a Gaussian model, with a χ ν 2 \chi^{2}_{\nu} of 1.06. When centering the Gaussian where the absorptions for DMS and DMDS peak, I find ln(B) = 1.21 in favour of the Gaussian model, with a χ ν 2 \chi^{2}_{\nu} of 0.99. With only ∼ \sim 2- σ \sigma in favour of Gaussian features, I conclude no strong statistical evidence for spectral features. show][email protected] I Introduction The search for signs of life is a key goal of the exoplanet community. The James Webb Space Telescope (JWST) has improved our ability to study the atmospheres of temperate worlds, and many of the first attempts to study atmospheres with the telescope being of those around rocky planets. However, spectra of rocky planets are difficult to analyse, given their small features in transmission spectra, with at present no confirmed atmospheric detection ( E. M. May et al. 2023 , e.g.) . On the other hand, H 2 -rich sub-Neptunes, with their low mean molecular weight atmospheres, are readily available to study, with recent theoretical and observational work determining these could be our current best opportunity to find life elsewhere in the universe ( S. Seager et al. 2021 ; S. Seager et al. 2025 ) . One hypothesised class of sub-Neptune is “Hycean” worlds ( N. Madhusudhan et al. 2021 ) , defined by a H 2 rich atmosphere above a global liquid water ocean. The planet K2-18b ( T_{\rm{eq}} = 254.9 \pm 3.9 K, R p = 2.610 \pm 0.087 R ⊕ , M p = 8.63 \pm 1.35 M ⊕ ( R. Cloutier et al. 2019 ; B. Benneke et al. 2019 ) ) has been extensively studied, with competing hypotheses on whether it is a Hycean world ( N. Madhusudhan et al. 2023 ) , mini-Neptune ( N. F. Wogan et al. 2024 ) , or a gas dwarf with a surface magma ocean ( O. Shorttle et al. 2024 ) . K2-18b has now been observed in transmission using 3 instrument modes of JWST: NIRISS/SOSS, NIRSpec/G395H, and MIRI/LRS. The initial near-infrared observations suggested the presence of dimethyl sulphide (DMS) ( N. Madhusudhan et al. 2023 ) , but this molecule was not confirmed in a re-analysis of the same data ( S. P. Schmidt et al. 2025 ) . Recently, mid-infrared observations of K2-18b suggested the presence of DMS and/or dimethyl disulphide (DMDS) at 3- \sigma significance ( N. Madhusudhan et al. 2025 ) . The potential presence of DMS or DMDS has been raised as evidence for a biosphere on K2-18b ( N. Madhusudhan et al. 2023 ; N. Madhusudhan et al. 2025 ) , given that these molecules are produced by phytoplankton in Earth’s oceans ( U. Alcolombri et al. 2015 ) . N. Madhusudhan et al. 2025 claims to reject a flat line at 3.4- \sigma significance. They compare a 16-parameter “canonical” model to a 1-parameter flat line model. However, this is not a nested analysis, as their flat line model is not a subset of their “canonical” model. It is possible to compare the Bayesian evidence of these models to determine whether one is favoured over the other, however, careful consideration of the prior ranges need to be taken, as the prior space can greatly impact the Bayesian evidence ( R. Trotta 2008 ) . The authors do observe this behaviour; they comment that when increasing the prior range for their flat line model, they achieve a stronger significance. When using their 32-parameter ‘‘maximal’’ model 1 1 1 These models are being used to fit 29 data points. One should generally avoid fitting data with more free parameters than data points to avoid overfitting the data. , they find they can only reject a flat line at 3- \sigma , the significance has gone down because the prior space between the “maximal” and “canonical” models are different. The 3- \sigma detection of DMS/DMDS claimed by N. Madhusudhan et al. 2025 was established using a curated “canonical” model as opposed to their “maximal” model. However, the “canonical” model neglects opacity contributions from other potential molecules, which consequently boosts the detection significance by neglecting alternative species in the Bayesian evidence. They performed the same tests with their ”maximal” model and find they can only reject a flat line model compared to their atmosphere models without DMS/DMDS at less than the 2- \sigma level. Furthermore, the authors postulate that the abundance of DMS/DMDS they obtain are consistent with \sim 20 \times Earth’s biogenic flux, as predicted by S.-M. Tsai et al. 2024 . However, in S.-M. Tsai et al. 2024 they show that ethane (C 2 H 6 ) would be more readily detectable with this level of biogenic flux compared to DMS, yet no ethane is seen in the MIRI/LRS observations. These concerns motivate a careful reassessment of the reliability of the claimed detection of DMS and/or DMDS in K2-18b’s atmosphere. In this research note, I explore whether a flat line can be rejected by the K2-18b MIRI transmission spectrum, and therefore determine if an atmospheric signal can be claimed. II Methods When using the Bayesian evidence to compute whether one model is favoured by a set of data compared to another model (e.g. an atmospheric model compared to a flat line), one model should ideally be a subset (‘nested’) of the other model with the same priors on the common parameters. In this section, I perform a nested approach to determine if a Gaussian model is statistically preferred relative to a flat line. Following previous work ( JWST Transiting Exoplanet Community Early Release Science Team et al. 2023 ; E. M. May et al. 2023 , e.g.) , I search for spectral features, while remaining agnostic to the particular gas causing the feature, with a Gaussian function: \delta_{\lambda}(A,\mu,\sigma_{m},c)=A\,\text{exp}\Bigg(-\frac{(\lambda-\mu)^{2})}{2\sigma^{2}_{m}}\Bigg)+c (1) where \delta_{\lambda} is the model transit depth, and \mu , \sigma_{m} , and A are the centre, standard deviation, and amplitude of the Gaussian, respectively, and c is a vertical offset. Given that the spectral shapes of DMS and DMDS are Gaussian-like (see Figure 3 of N. Madhusudhan et al. 2025 ), this is a reasonable approximation. I fit for the flat line using a model of the form: \delta_{\lambda}(c)=c\text{.} (2) For the analysis, I only consider the observations obtained with the JexoRes pipeline, which is the same data shown in Figure 2 of N. Madhusudhan et al. 2025 . I did perform the same analysis using the observations using the data obtained from JexoPipe , however, the conclusions remain similar. 2 2 2 The data was obtained here https://osf.io/gmhw3/

Read the complete text