Category Archives: Atmospheres

Transit Spectroscopy, Biosignature Searches, and the Myth of Perfect Stars

Can we detect atmospheric biosignatures in the next two decades? Only if we can meet a major, newly-recognized challenge to our studies of exoplanet atmospheric composition.

Over the past years the Hubble Space Telescope has proven to be our most powerful tool to probe the atmospheres of transiting exoplanets: the comparison of spectra taken before and during the transits can reveal the compositions of the atmospheres. Exciting discoveries included condensate clouds, hazes, extremely efficient scatterers, molecules (water, methane, and carbon-dioxide), and atoms (sodium and potassium). Some of the most ambitious research programs are pushing this technology to levels never envisioned previously as they reach spectacularly precise measurements on increasingly small planets. This technique may even allow the detection of water vapor in the habitable zone earth-sized planets in the TRAPPIST-1 system with HST and additional gases with JWST.

But underpinning these measurements is an approximation that is called in question now – in fact, one that we show is wrong for typical systems.

Astrophysicists focusing on exoplanets often assume that the planet host stars are perfect. As we show in our new paper, led by Steward Observatory graduate student Ben Rackham, this assumption is underpinning high-precision transit spectroscopy. In reality, stellar heterogeneity contaminates exoplanet transmission spectra (HST and JWST) and — unless we figure out how to correct for this effect — it will greatly limit our ability to search for biosignatures in the next two decades.


Because transit measurements are relative measurements stellar spectra cancels out – in the first order approximation:  the difference of a spectrum taken before transit and during transit will provide the transmission spectrum of the exoplanet atmosphere.


It is tempting to think that this approximation holds with infinite precision: in fact, the majority of the transit spectroscopy papers in the literature simply adopt this approximation without further considerations. This may be OK for the typical, less precise measurements, but remains a dangerous assumption for high-precision studies and for all but the least active stars.

In reality, the stellar spectrum does not cancel out in transmission spectroscopy, because the first-order approximation described above confusingly equates two different light sources: the stellar disk (the spectrum of which is observed before the transit) and the actual light source, which is just a very small fraction of the stellar disk — the projection of the transit chord onto the stellar photosphere (see figure).

Mercury’s transit in front of the Sun illustrates that even quiet stars are heterogeneous on the fine length scale of exoplanets.

Stars are not perfect: in reality, no patch of the stellar photosphere has the same exact spectrum as the stellar chord. The difference between the assumed lightsource (stellar disk) and the actual lightsource of unknown spectrum (photosphere under the chord) imprints itself onto the transmission spectrum observed.

The effect itself is not new: Some of the best published studies tried to develop a correction for the suspected stellar contamination. In doing so, almost all groups assumed a linear relation between the photometric variability of the stars and the covering fraction (basically: more variability means more spots).

In our new paper we worked with Mark Giampapa — a solar/stellar astrophysicist —  on the first comprehensive study of this effect and its impact on transit spectra and exoplanet density measurements.

This project brought about important, surprising, and concerning results.

Our team — part of the larger Earths in Other Solar Systems project — has created toy models of a star with starspots and faculae to assess the connection between stellar variability and stellar spot covering fraction; we then used state-of-the-art stellar atmospheric models to predict the stellar contamination in the transit spectra of the planets, which we then compared with atmospheric absorbers (including biosignatures) that could be detected in transmission spectra now and in the near future. Finally we also assessed the impact of the apparent size (and therefore density) of the exoplanets: could starspots lead to an apparently lower planet density (more volatiles/gaseous envelope)?

Photometric variability amplitudes are very poor tracers of stellar spot/facula coverage.

Our findings are detailed here, but the key points are:

  • The amplitude of stellar variability (photometrically determined brightness variations) that is sometimes used as a basis to argue for low spot covering fractions is an extremely poor measure of the stellar heterogeneity: the linear correlation many published papers assume is wrong in most cases.
  • Considering the actual spot covering fraction range that typical stellar variability amplitude really translates to, a much broader range of stellar contamination is possible than previously considered.
  • The contamination is not only limited to changing the slope of the spectrum (which is the effect most are aware of), but it will also introduce spectral features – especially for red dwarf host stars, whose spectra are rich in molecular features. The results are concerning: without additional information, it can be extremely difficult or downright impossible to distinguish water absorption in the star from water absorption in the planet. We find that even molecules not present in the stellar photosphere (O2 and O3) are difficult to identify given the large contamination from the star.
  • For TRAPPIST-1, currently the most exciting planetary system for transit spectroscopy studies, we predict that stellar contamination could be 4-7 times greater than the intrinsic planetary features.
  • Finally, we also show that for host stars with a larger number of spots, the planet density calculated will be too low – errors as larger as 15-25% are to be expected. The error is huge if one’s goal is to understand the possible composition of a small, mostly rocky planet; but less of a problem for hot jupiters.

Expected range of stellar contamination for M-type stars. The contamination can suppress or mimic several of the key absorbers expected from planetary atmospheres, including water and oxygen.

The stellar contamination in transit spectra can be very significant: in fact, our predictions are that the contamination levels are high enough the be present in numerous Hubble Space Telescope studies already published.

How can we recognize and distinguish stellar contamination from genuine planetary atmospheric features?

We are working on this and testing multiple ideas; multi-epoch data, high-resolution spectra, and better understanding of the spot and facula properties are likely to be part of the solution. Of course, the larger fraction of the exoplanet community thinks about this challenge, the more likely it is that we can solve the problem to inform upcoming Hubble and James Webb Space Telescope observations, which may then lead us to sampling habitable zone exo-earth atmospheres within the next five years.

Further reading: Rackham, Apai, Giampapa 2017 Astrophysical Journal, in press (arXiv)

Extrasolar Storms: Belts, Spots, and Waves in Brown Dwarfs

Our new paper came out today in Science, presenting evidence for bands, zones, spots, and waves in brown dwarfs and a model that explains well several until-now mysterious changes in the brightnesses of brown dwarfs.

Podcast: Learn more about our project from the Science Magazine’s podcast!


I am excited about our results because they open a new window on very fundamental processes in brown dwarfs (atmospheric circulation, heat exchanges, and cloud formation) and, at the same time, they also explain a number of past observations that puzzled brown dwarf experts. As always with brown dwarfs, the results are much more far-reaching than people often realize: brown dwarfs are excellent proxies for giant exoplanets: often what we cannot learn from giant exoplanets  we learn from brown dwarfs.

Brown dwarfs and Exoplanets: This is the decade of exoplanets, so one may wonder why are brown dwarfs important. News often describes brown dwarfs as “failed stars”, a label I find misleading: in fact, most brown dwarfs are much more similar to giant planets than to stars. What’s more, it is almost certain that the brown dwarf population contains a large number of ejected giant planets — bona fide exoplanets that were booted from their natal systems by more massive siblings. Known brown dwarfs have temperatures between 250 K to about 2,500K — completely overlapping with the temperatures of giant exoplanets; the compositions of many brown dwarfs are likely very similar or identical to many of the giant exoplanets. But most excitingly, the physical and chemical processes in brown dwarf and exoplanet atmospheres are the same; the identical processes, combined with the fact that brown dwarfs are much easier to study is the reason why we learn so much about exoplanets from brown dwarfs.

A great video summary of our results by JPL:

So, what’s new? Our study shows atmospheric circulation in brown dwarfs for the first time: it shows that brown dwarfs have bands and zones, spots, and that cloud thickness in the zones is continuously changed by atmospheric waves. We found that brown dwarfs are similar to the gas giants in the Solar System (in that they have zonal circulation) , but that they are more like Neptune and less like Jupiter (their brightness variations are driven by large-scale waves in zones rather than Great Red Spot-like storms as in Jupiter). The waves are an interesting piece of the puzzle: we see large-scale waves in the solar system planets (including Earth), but we have not yet seen waves with wavelengths similar to the entire planet — like the ones we now found in brown dwarfs.

Why is atmospheric circulation important? Atmospheric circulation — large-scale flows of air in atmospheres — is very important as it sets how heat and particles/droplets/gas are distributed in a planet. For example, in Earth atmospheric circulation (such as Hadley cells) transport heat between the warmer equatorial regions to the cool polar regions and this circulation pattern not only determines the temperature distribution, but also sets which regions on Earth are dry or rainy and how clouds form over the planet.

Morning Glory clouds in Australia are atmospheric waves rendered visible by cloud formation.

What are these waves? On a fundamental level waves are changes that propagate through a medium. For example, dropping a pebble in a lake will force the water to move away from its equilibrium  — and that change will propagate across the surface of the lake. Atmospheres have many different types of waves: for example, (gravity) waves are common and they often propagate on the interface of warm air sitting on top of cold air — these waves are invisible to us (as air is mostly transparent), but they can lead to spectacular sights when clouds highlight them. (This Berkeley meteorology class’s page gives a couple of cool examples). The two examples shown here are small waves — atmospheric circulation is driven by large-scale waves, with wavelengths that are hundreds or thousands of kilometers.

Atmospheric waves at the interface of rapidly flowing air (above) and near-stationary air (below), leading to mixing and heat transport. Photo by Benjamin Foster.

What does this tell us about exoplanets? Whatever we find in brown dwarfs should be pretty much be the same for most giant exoplanets in the galaxy — only the rare hot jupiters (very heavily irradiated by their host stars) should look different (but even for those, the underlying processes that shape their atmospheres should be the same). So, based on our results we would expect that most giant exoplanets will have zonal circulation; we should expect that their atmospheres are not homogeneous, structureless, but in fact should display large brightness variations in the infrared. We should also expect that giant waves will propagate in their atmospheres (parallel with their equators) and that these waves will change the thickness of the clouds. Our next steps will be to figure out what processes drive these waves (probably some combination of heat transport, winds, and rotation) and to improve the cloud models — the same cloud models that are used to interpret exoplanet atmospheres, too. Importantly, we also learned that the atmospheres of gaseous exoplanets should have regions with very different appearance: where the clouds are thinner or lower, we can see into the deeper, hotter atmosphere. This should be an important warning for most current studies that use one-dimensional atmospheric models: in other worlds they assume that every bit of the planet is like the rest.

I am also excited about our results because they demonstrate how much we can learn from unresolved data — basically, from a single pixel. This is crucial as this is all we are going to get for exo-earths: we will not be able to build large enough telescopes to take detailed images of the surfaces of exoplanets — but we will still be able to learn about their atmospheres (and surfaces) from time-resolved observations!

Near-infrared images show bands on Jupiter. Brighter regions have thinner cloud cover, allowing radiation to escape more easily from its hotter interior.

How did we do it? We used the Spitzer Space Telescope and watched the brown dwarfs rotate. As they rotate their brightness changes: when a brighter spot rotates in the visible hemisphere, Spitzer will see the brightening. The brown dwarfs we observe take between 1.5 and 13 hours to turn around fully: we used Spitzer to observe 32 complete rotations of each brown dwarf. This allowed us not only to map the cloud distribution, but also how it changes from rotation to rotation and also over longer timescales: our observations were following the brown dwarfs for more than a year. We then used a novel computer algorithm developed by my colleague Theodora Karalidi to figure out how the brown dwarfs look and how the clouds change. Another team member, Mark Marley from NASA Ames, used a different set of models (cloud structure and light propagating through the clouds) to help figure out how high in the atmosphere are the clouds we see. Initially, we expected that the changes we see are driven by Great Red Spot-like stable features (the GRS has been seen in Jupiter for more than 300 years) — but the brightnesses of the brown dwarfs changed way too much to be explained by spots, Waves, however, worked extremely well. We then realized that the waves and bands not only explain our own data, but a humber of other puzzling findings reported by other teams. We had an excellent team of experts who all contributed different pieces to solving this puzzle.

Very large and very rapid changes in the lightcurve of 2M1324. From Apai et al. (2017, Science).

What’s next? One of our next steps is to expand this study to directly imaged giant exoplanets, which will allow us to explore how cloud properties and dynamics change with the mass of the objects — this cannot be done well with the sensitive, but low-resolution Spitzer Space Telescope. We are using the Hubble Space Telescope in our program Cloud Atlas, to prepare coarse cloud maps for about a dozen or so cool brown dwarfs and exoplanets.