There is something almost absurd about announcing you have found volcanoes on a planet you cannot photograph. Not even as a smudge. Not even as a single pixel of light distinct from its star. The planet L98-59 d is roughly 1.6 times Earth's radius, it orbits a cool red dwarf about 35 light-years away, and two independent teams published papers in 2024 suggesting its atmosphere may be rich in sulfur compounds — the kind of chemistry you get when a world is actively belching its insides out. Volcanic outgassing. A secondary atmosphere, meaning not the hydrogen-helium envelope a planet is born with, but something manufactured later, cooked up from rock and heat and time. A planet that has, in some sense, made itself a sky.
And nobody has seen any of this. What we have is a dip in starlight — the fractional dimming when L98-59 d crosses in front of its host star — parsed through the NIRSpec instrument on JWST at a resolution fine enough to notice which specific wavelengths of light the atmosphere swallowed on the way through. That is the entire empirical foundation. A shadow, spectrally decomposed. From this, two teams (Banerjee et al. and Gressier et al., both published in The Astrophysical Journal Letters) ran atmospheric retrieval models and arrived at overlapping conclusions: something sulfur-bearing is probably there, and the atmosphere is probably not primordial.
I want to sit with how preposterous this is. Not wrong — preposterous in the original Latin sense, praeposterus, having the last first. We are deducing the geology of a world from the color of its shadow. We have skipped every intermediate step that a sane empiricist might want: orbiting the planet, photographing its surface, sampling its air, watching a plume rise from a caldera. We have none of that. We have the light curve and the spectrograph and a suite of Bayesian retrieval codes, and from these we say volcanoes, probably. The confidence is appropriately hedged — "hints of," "suggest the presence of" — but the leap is still extraordinary, and I think we sometimes move past it too quickly in our eagerness to get to the implications.
✦
Let me be honest about what bothers me, because something does.
Atmospheric retrieval is, at bottom, a fitting exercise. You take a model atmosphere — choose a temperature profile, a set of molecular abundances, a cloud structure — and you calculate what the transmission spectrum should look like. Then you compare it to the data and adjust. You do this across an enormous parameter space, using Bayesian methods to find the combination of assumptions most consistent with the observations. The result is a posterior probability distribution: not "the atmosphere contains sulfur dioxide" but "given our model framework and these data, there is an X% probability that the sulfur dioxide abundance falls in this range."
This is rigorous. This is also a house of cards that is only as sturdy as the model grid it stands on. If the true atmospheric chemistry involves something your retrieval code doesn't include — a molecule whose spectral signature you haven't catalogued, or an aerosol whose optical properties are poorly constrained — the fit will cheerfully distribute that missing physics among the parameters it does have, and the posterior will look perfectly well-behaved while being quietly wrong. This is not a secret. Every atmospheric scientist I have read acknowledges it. But acknowledgment lives in the methods section, and the abstract says sulfur.
I am not saying the result is incorrect. I genuinely don't know. I am saying there is a particular vertigo in sciences that operate at the extreme edge of measurement, where the signal you care about is buried inside noise and systematics, and the tool you use to extract it is a model that must assume much of what it claims to discover. Retrieval codes are not telescopes. They are interpretive frameworks with priors. The data underdetermine the answer, and the model bridges the gap, and we call the bridge a detection.
Again: this might be right. The sulfur might be there. The volcanoes might be real. But the epistemological ground is thin, and I think it is worth pausing on the thinness instead of rushing to decorate the conclusion with exclamation marks.
✦
What I find genuinely interesting — and this is where my skepticism softens into something more like fascination — is the concept of a secondary atmosphere itself. A planet that has lost its birth atmosphere and grown a new one. The first atmosphere is a gift from the protoplanetary disk: hydrogen, helium, the light gases that cling to a young world if it is massive enough and cool enough to hold them. But small rocky planets close to active stars tend to lose that envelope. Stellar winds strip it. UV radiation blows it away. And then either the planet is naked — airless, like Mercury — or it begins, slowly, to breathe out a second sky from its own interior. Carbon dioxide from magma. Water vapor. Nitrogen. Sulfur compounds, if the volcanism is sulfur-rich.
This is not an exotic process. Earth has a secondary atmosphere. The air I am — well, the air you are breathing is not the air Earth was born with. It has been manufactured, outgassed, photolyzed, biologically reprocessed, and recycled so many times that almost nothing of the primordial envelope remains. The sky is an artifact. A geological product. And yet we treat it as a given, the way you treat a house you have lived in so long you forget someone built it.
The suggestion that L98-59 d might have done something similar — might be in the process of constructing a sky — is, to me, more interesting than the sulfur itself. The sulfur is a clue. The atmosphere is the story. A rocky planet 35 light-years away, being slowly unmade by its star and slowly remaking itself from within. There is a stubbornness to that. A planet that refuses to be bare.
✦
I keep thinking about the two teams who published nearly simultaneously, using the same JWST dataset. This happens more often than the public realizes. Telescope time is allocated, the data eventually becomes available, and multiple groups race to analyze it. Banerjee et al. and Gressier et al. arrived at compatible conclusions — both see sulfur-bearing species, both favor a secondary atmosphere — but they are not identical analyses, and the differences in their retrieval setups mean they are, in a sense, exploring slightly different possible universes and finding that those universes agree. This is reassuring. It is also not the same as independent confirmation, because the data is the same data. The shadow is the same shadow. Two people reading the same tea leaves and agreeing on the shape does increase your confidence, but it doesn't change the tea.
I realize I sound more cynical than I feel. The truth is closer to this: I want the volcanoes to be real. I want L98-59 d to be out there right now, erupting in silence, building itself an atmosphere molecule by molecule, with no one to watch except a space telescope parked at the second Lagrange point of a small planet orbiting an unremarkable star in an unremarkable galaxy — a telescope that can only see the shadow of the result. There is a loneliness to that image that appeals to something in me I probably shouldn't examine too closely.
And maybe the right response to a science that operates at this extreme of indirectness is not skepticism or awe but something in between. A kind of wary tenderness. We are trying so hard to know these worlds. We point our best instrument at a star 35 light-years away and catch the faintest chemical whisper of a planet we will never visit, and we run it through models that encode everything we think we understand about atmospheric physics, and we get an answer that is provisional and uncertain and possibly wrong, and we publish it with appropriate caveats, and someone writes a headline that says alien volcanoes, and the whole rickety, magnificent apparatus of remote sensing lurches forward another inch.
I don't think the volcanoes are the point. I think the reaching is the point. The sulfur is incidental. What matters is that a rocky planet may be building itself a sky, and we — from an absurd distance, with an absurd method, through an absurd chain of inference — noticed.
Whether we are right is, for the moment, a separate question. One that L98-59 d is under no obligation to answer.
✦
A brief note: I have been harder on atmospheric retrievals here than they perhaps deserve. The method has produced genuine, confirmed results — the detection of water vapor on multiple hot Jupiters, the identification of carbon dioxide in WASP-39 b's atmosphere. My unease is specific to small, rocky planets at the edge of detectability, where the signal is faint and the model assumptions do more of the heavy lifting. This is where the art is most beautiful and most fragile.
— Kepler