The evidence for biosignatures on K2-18b is flimsy, at best

Out there in the Universe, every stellar system is like a lottery ticket. Every time a new star forms, there’s a chance that it’s going to form planets, including possibly rocky, Earth-sized planets. There’s a chance that the right conditions will emerge on that world — a mix of the raw chemical elements, a source of energy, and stable, life-friendly conditions — that allow for the creation of molecules that metabolize resources and can self-replicate. And there’s a chance that those molecules will survive and thrive over time, creating life that can evolve and alter the conditions on the planet itself. While we can be certain that Earth is one of the winning tickets in this cosmic lottery, it’s plausible that there are many other winners, including on worlds that are very different from Earth itself.

Although no one has yet found our second example (after Earth) of a definitively inhabited planet, a recent exciting claim asserts that we’ve now come tantalizingly close: by finding signatures of dimethyl sulfide (DMS) and dimethyl disulfide (DMDS) in the atmosphere of exoplanet K2-18b. According to the study authors — and repeated in outlets such as the New York Times, the BBC, and many others — these molecules:

• exist on a Hycean world,
• with a global ocean beneath an envelope of hydrogen and helium,
• where organic molecules are then lifted up to the upper atmosphere,
• including not just methane but also DMS and/or DMDS,
• which they assert is only produced on Earth by biological processes,
• and is seen in both near-infrared (with JWST’s NIRISS and NIRSpec) and mid-infrared (with JWST’s MIRI) wavelengths,
• at about the 3-sigma significance, meaning there’s only a ~0.3% chance of this being a fluke,
• and therefore, this is perhaps the strongest and most exciting hint of life beyond our Solar System ever discovered.

It’s a wild claim, put forth by one particular group at Cambridge led by exoplanet scientist, Nikku Madhusudhan. Unfortunately, nearly every aspect of this claim is almost certainly wrong. You’ve heard the hype; now learn the science.

The most common “sized” world in the galaxy is a super-Earth, between 2 and 10 Earth masses, such as Kepler 452b, illustrated at right. But the illustration of this world as “Earth-like” in any way may be mistaken, as it’s more likely to either have a large, volatile gas envelope, making it a mini-Neptune, or to be a hot, stripped planetary core: like a scaled-up version of Mercury. K2-18b is not either; it’s unambiguously a cold, full-fledged Neptune-like world.

The story of K2-18b goes all the way back to 2015: when the exoplanet itself was first discovered. A systematic search for transiting planets in the K2 (the follow-on to the Kepler mission) data revealed 36 planetary candidates that transited in front of 31 stars, including the planet K2-18b around the red dwarf star K2-18. K2-18 has about half the mass of the Sun but just 2.3% of the Sun’s intrinsic brightness, while the planet, K2-18b, is about:

• 8.6 times the mass of Earth (or about half the mass of Neptune),
• about 2.6 times the radius of Earth (or about two-thirds the radius of Neptune),
• and orbits its parent star once every ~33 days,
• leading to an average temperature that’s just below the freezing point of water: -8 °C (17 °F).

There were more planets of this variety — with masses and radii in between those of Earth and Neptune — found in the Kepler and K2 data than of any other size or mass. While they were initially all called super-Earths, subsequent research has shown that most of these worlds are more like mini-Neptunes: without rocky surfaces and thin atmospheres, but rather with thick hydrogen-and-helium envelopes. Only if your world is lower than about two Earth masses, and smaller than about 1.4 Earth radii, are they more Earth-like than Neptune-like. K2-18b has a low density: of just 2.7 g/cm³, which is less than half the density of a rocky planet like Earth. Given its large mass and radius, plus its low temperature (being so far from its relatively cool parent star), K2-18b is almost certainly a “cold Neptune” world, rather than a terrestrial, Earth-like one.

When an exoplanet passes in front of its parent star, a portion of that starlight will filter through the exoplanet’s atmosphere, allowing us to break up that light into its constituent wavelengths and to characterize the atomic and molecular composition of the atmosphere. If the planet is inhabited, we may reveal unique biosignatures, but if the planet has either a thick, gas-rich envelope of volatile material around it, or alternatively no atmosphere at all, the prospects for habitability will be very low. Nearly all so-called “super-Earth” worlds that have had their transit spectrum measured have revealed these characteristic volatile envelopes, suggesting that they’re mini-Neptunes instead of super-Earths.

However, an astounding (and very dubious) claim came out in 2019: using data from the Hubble Space Telescope, there appeared to be signatures consistent with water vapor in the atmosphere of K2-18b. It was a tricky signature; because K2-18b wasn’t Earth-sized, but rather was less dense, it had to be filled with something far more rarified (less compact) than Earth-like material. There were two main possibilities that quickly emerged:

• a large, deep, global ocean, which is lightweight and low in density,
• or a large, deep, thick global atmosphere composed largely of hydrogen and helium: the lightest and most diffuse gases of all.

There were definitely clouds found in this exo-atmosphere, but Hubble’s observations, which focused on a very narrow window of the infrared spectrum of K2-18b (from 1.1 to 1.7 microns), allowed us to perform transmission spectroscopy: where the light from the parent star filters through the atmosphere of the exoplanet, revealing its contents. There were indeed absorption signatures present, which appeared to be consistent with water vapor.

However, over very narrow wavelength ranges, many molecular signatures — sometimes thought of as chemical fingerprints — look very similar. In particular, from 1.1 to 1.7 microns, the atmospheric signatures of water and methane look extremely similar.

This plot shows the spectrum, from 0.8-5.0 microns, of exoplanet K2-18b as taken with the NIRSpec and NIRISS instruments aboard JWST. The signal is shown with data points with error bars; the interpretation of the signal by the discovering group is shown alongside it. The earlier Hubble data only ran from 1.1 to 1.7 microns: where signatures of methane and water vapor would have been virtually indistinguishable.

Then, in 2023, JWST also took transmission spectroscopy data of this world using its NIRSpec and NIRISS instruments, including at much longer wavelengths: from 0.8 microns all the way up to 5.0 microns. What they found was definitive: the molecule doing the absorbing wasn’t water vapor (H₂O), but rather was methane (CH₄), teaching us that the Hycean world hypothesis, with a global liquid water ocean, was completely unsupported by the data.

That should’ve been “case closed” by any reasonable standard. This wasn’t a waterworld, but rather a “cold Neptune” world, like, you know, Neptune, which is blue largely because of the abundant presence of methane in its atmosphere.

Furthermore, there were also signatures of carbon dioxide seen in the atmosphere, but not water vapor, not carbon monoxide, and not ammonia. This led researchers to explore a number of conceptual structures for this world, K2-18b:

• a thin atmosphere world with a rocky surface,
• a thin atmosphere world with a global water ocean,
• a thick atmosphere world with no surface at all, just supercritical layers of gases,
• a thick atmosphere world with a liquid, silicate magma surface (favored by many, especially at pressures several times what’s found at the bottom of Earth’s oceans),
• and a deep atmosphere with a supercritical water ocean, seemingly resurrecting the Hycean world hypothesis.

While the rocky world scenario was ruled out, the others all remained at least marginally consistent with the data.

Five possible models for the atmosphere and surface properties of exoplanet K2-18b are shown here, as determined as of 2024: after the JWST near-infrared data was acquired. Only the first (thin atmosphere, rocky surface) scenario was able to be ruled out; the other four all remain viable in some capacity.

Then the hype train came in. A team of researchers — led by Nikku Madhusudhan at Cambridge — claimed that at the long-wavelength portion of the JWST spectrum, there was a detected signature that appeared to be dimethyl sulfide (DMS): a molecule that they claimed was only produced on Earth by biological processes. It’s true that most of Earth’s dimethyl sulfide is produced by phytoplankton and bacteria, and DMS represents the dominant form of organic sulfur found in Earth’s oceans.

But there were two problems with this assertion.

1. The “detection” of DMS in the near-infrared portion of the spectrum was very, very flimsy: at only 1-sigma significance, which means there’s greater than a ~30% chance that the detection is a fluke. In all sub-field of physics and astronomy, a “1-sigma detection” is known as a non-detection; there is no meaningful signal seen here.
2. While DMS on Earth may primarily arise from biological production mechanisms, it is found all throughout the Universe and, in those environments, is produced entirely by non-biological means. It’s been found in the interstellar medium, including in the galactic center’s large molecular cloud G+0.693-0.027. It’s been produced abiotically in the laboratory right here on Earth (including since 1995) by simple processes, and can be made photochemically elsewhere in the Universe. And it’s been found on comets as well, including on the famous comet 67P/Churyumov-Gerasimenko, which was the target of ESA’s Rosetta mission.

Spectra of K2-18 b, obtained with Webb’s NIRISS (Near-Infrared Imager and Slitless Spectrograph) and NIRSpec (Near-Infrared Spectrograph), display an abundance of methane and carbon dioxide in the exoplanet’s atmosphere, as well as a possible detection of a molecule called dimethyl sulfide (DMS). However, the significance of the data could potentially indicate many other outcomes besides the one pushed by the study’s authors.

This is important. You can’t simply say, “We’ve found this gas, and therefore there’s biological activity on this world that produced it.” Back in 2021, the astrobiology community held a workshop on standards of evidence for biosignatures, which led to a community report discussing responsible standards of evidence and quantifying seven different levels of confidence for announcing biosignatures, with the lowest-confidence levels going to detections that cannot discriminate between different scenarios and mid-confidence levels going to definitive detections that cannot rule out abiotic pathways. To announce “we’ve found a biosignature” would require multiple, independent, unambiguous signatures of molecules that cannot be produced abiotically at all.

As a new paper in the Proceedings of the National Academies of Sciences correctly reminded us just a few days ago:

“Characterizing rocky or sub-Neptune-size exoplanets with JWST is an intricate task, and moves us away from the notion of finding a definitive “silver bullet” biosignature gas. Indeed, JWST results necessitate us to allow “parallel interpretations” that will perhaps not be resolved until the next generation of observatories.”

After all, there are three now-obvious reasons to be skeptical of the claim of finding a biosignature on an exoplanet like K2-18b.

1. The molecule in question that is supposed to be a biosignature, DMS in this case, is not produced solely by organic processes, but by abiotic (non-life) processes as well. Chemistry happens all over the Universe, including under very foreign conditions compared to what we find on Earth, and small molecules are particularly easy to make abiotically, especially under exotic conditions.
2. There are many viable scenarios for the planetary interior of this exoplanet, and each one has different chemical conditions with a unique set of potential atmospheric signatures. You cannot simply assume that one model (like the Hycean world model) is correct and interpret your data accordingly; you must discriminate between different scenarios and acquire data that’s sufficient to tell them apart.
3. And finally, even if we had found a molecule like DMS at very high significance, which we did not, we don’t understand its spectral signature well enough to be certain we weren’t confusing it with something else. We need better significance and a more comprehensive suite of models to know whether DMS is even present or not, and whether the Hycean world hypothesis is the best (or the only) fit to the data that works.

While examining the transmission spectrum of exoplanet TOI-270d with JWST, methane, carbon dioxide, and water vapor were detected on this cold Neptune-like exoplanet. Yet several viable models for its interior persist, and all must be considered when modeling the atmosphere for the presence of any additional species.

Importantly, we’ve found other exoplanets with similar signatures and properties to K2-18b previously, including the exoplanet TOI-270d, which is also a mini-Neptune exoplanet found at near-Earth-like temperatures. It has methane and carbon dioxide in its atmosphere, but unlike K2-18b, it also shows signs of water vapor in its atmosphere, too, with a few sulfur-containing compounds detected at the edge of statistical significance. As you can see, above, there are several scenarios that remain viable for its planetary interior, highlighting the difficulty with drawing definitive conclusions about a planet’s overall properties, and even the “chemical fingerprint” of a particular species of gas from a low-confidence detection, from even a high-quality spectrum of its atmosphere.

So, clearly, we learned our lesson from this saga, and no one in the community would ever make such an irresponsible claim again, right?

What wishful thinking that would be. Instead, the same team of researchers that originally claimed to find DMS in the atmosphere of K2-18b went back and acquired JWST MIRI (or mid-infrared) spectral data of this exoplanet, which is good, and then put out a wildly irresponsible press release that doubled down on and oversold claims of DMS, now joined (possibly) by DMDS, in the spectrum of K2-18b. While there’s a fairly responsible scientific paper out discussing these claims, the press release — and Madhusudhan himself in several interviews — are now making a series of scientifically indefensible statements that the data itself simply cannot support.

This shows the data of the transmission spectrum acquired by the mid-infrared instrument (MIRI) aboard JWST while examining the atmosphere of transiting exoplanet K2-18b. The error bars are shown on each data point in red, while the best-fit overall spectrum is shown with the black line.

They do successfully acquire a MIRI spectrum for this transiting Neptune-like exoplanet at a distance of 124 light-years, which is a remarkable achievement and an exciting proof-of-concept for what JWST can do for such planets even at such great distances. The data are good enough to rule out a “featureless spectrum” for the exoplanet, indicating that there are spectral features and signatures of gases in the atmosphere. And they do explore one of the possible not-yet-ruled-out scenarios for the interior of the exoplanet K2-18b, that of the Hycean world, and conclude that — for that scenario — the data is slightly better fit by an atmosphere that includes carbon dioxide, methane, and either DMS or DMDS than by carbon dioxide, methane, and no DMS or DMDS.

But then there’s what they didn’t do.

They didn’t explore the other still-viable scenarios for the interior of exoplanet K2-18b, including scenarios with thick atmospheres and either no surface at all or a liquid magma ocean beneath the gas-phase atmosphere. They didn’t test for the possible signatures one would expect if you had a magma ocean in contact with a hydrogen-and-helium rich atmosphere: signatures that include elements like silicon, magnesium, and aluminum, which are heavily prevalent on magma-rich exoplanets and which have been shown to have extraordinarily different chemical fingerprints from compounds that only contain light elements like carbon, nitrogen, oxygen, and hydrogen. Instead, they only tested for compounds that contained those elements, along with sulfur, and nothing else.

Although the JWST MIRI spectrum of exoplanet K2-18b is consistent with a series of light molecules like methane and carbon dioxide along with DMS and/or DMDS, the “significance” of 3-sigma was only obtained because all other possible gas species that could exhibit a strong absorption feature beginning at 9 microns were excluded from the analysis. There are other strongly viable scenarios that must be considered as well.

The big problem, as you can see from the graph above (lifted directly from their paper), is what they did and didn’t look for. Sure, in the paper, they determined that the data favors the presence of DMS or DMDS at the 3-sigma significance level over a scenario with the absence of DMS or DMDS, but this was practically guaranteed from how they designed their (very flimsy) test.

• They assumed a Hycean world model: where the allowable species they considered didn’t include any chemical species that would be expected exclusively in the alternative interior scenarios.
• Then they performed an analysis of the data by performing a likelihood analysis with only those allowed species for the model they selected.
• Next, they created what they called a simpler model, where they removed all other molecules that have the strong absorption features at long mid-infrared wavelengths (greater than about 9 microns).
• And finally, they compared, within those classes of simpler models, “possible spectra with DMS or DMDS” with “possible spectra with no DMS or DMDS,” and they found that the data favors spectra with DMS or DMDS at the 3-sigma level.

In the world of science, or anywhere else really, this is what we call a rigged test. They basically built a model where the only allowable candidate molecules to explain half of the spectrum (the longer wavelength half) were DMS or DMDS, and then concluded that “therefore, DMS or DMDS must exist” in the spectrum of this exoplanet. As Dr. Ryan MacDonald has pointed out, if you perform the proper statistical analysis correctly, the probability of there not existing DMS or DMDS within this spectrum is about 28%, which translates to a 1-sigma detection confidence.

Although they are approximately the same physical size, Neptune is about 30% denser than Uranus. This is despite the fact that Uranus appears to be more centrally condensed than Neptune. An orbiter mission to Uranus and Neptune, possibly containing probes that can enter their upper atmospheres, would be the ideal way to determine why they have these physical properties.

In summary, we are still searching for our first high-quality evidence of a biosignature on a world other than Earth, and we haven’t found it yet. K2-18b, if we’re being honest about what we can say about it, is a cold Neptune-like exoplanet with methane and carbon dioxide definitively found in its atmosphere, and that there are several scenarios for its interior that are still in play. We can also say that, using JWST, we’ve taken spectral data of this exoplanet’s atmosphere as it transited in front of its parent star, and acquired high-quality data from 0.8 microns up through about 12 microns in wavelength: an impressive wavelength range for sure. What we can’t say is that:

• we’ve found significant evidence for DMS or DMDS within the atmosphere, as we have not,
• we’ve determined that it is likely a Hycean world, as other scenarios (including and perhaps especially the magma ocean scenario) are at least just as good,
• and that if DMS or DMDS exists in the atmosphere, that it supports a biological origin rather than an abiotic origin for the molecular species.

While we will no doubt continue to acquire spectra of transiting exoplanets with JWST and beyond, it’s important to remember not to oversell our interpretations of the data and to also remember that we can investigate “cold Neptune” worlds in our own Solar System just by visiting them. We haven’t sent a mission to Uranus or Neptune since Voyager 2 flew past them in the late 1980s, and our window for sending a mission to both will open in 2034. If we want to discover whether life is out there, it’s important to look as comprehensively and scrupulously as possible, and not to fool ourselves by listening to those who loudly but prematurely make positive claims when the data does not yet support such strong conclusions.