Exoplanets - transmission spectroscopy - atmospheric chemistry

How We Smell an Exoplanet Without Going There

Transmission spectroscopy is the strange art of letting an alien planet pass in front of its star, catching the filtered starlight, and measuring molecule-sized bites in a spectrum.

We are not literally smelling an exoplanet. No telescope is inhaling WASP-39 b through a cosmic straw. But chemically, the analogy works: molecules identify themselves by the wavelengths they absorb. The quantitative game is to measure transit-depth changes of tens to hundreds of parts per million and prove they are not detector behavior, stellar contamination, or wishful thinking with error bars.

A planet blocks light. Its atmosphere edits light. The spectrograph reads the edits.
28 May 2026 - Hertford, UK Transmission spectroscopy Interactive atmosphere lab JWST - clouds - biosignatures
Original concept artwork for transmission spectroscopy: starlight filtered through an exoplanet atmosphere with wavelength fingerprints
Original concept direction: Biswajit Jana. The image shows an exoplanet atmosphere as a thin backlit chemical filter, with molecular absorption features translated into wavelength-dependent transit depth.
Interactive lab - separate flight window

Tighten your seatbelt. Open the atmosphere simulator.

The full lab now lives in its own page so the article can breathe. It includes JWST-style wavelength axes from 0.6 to 5.3 um, molecular bands for Na, K, H2O, CH4, CO2, CO, and SO2, instrument modes, noise, repeated transits, scale height, cloud decks, and live ppm-level readouts.

Launch Full Simulator
Atmospheres - spectroscopy - indirect evidence

The planet is hidden, but its atmosphere can still touch the light.

Exoplanet science is full of slightly unfair measurements. We cannot fly to most of these worlds, cannot scoop up their air, and usually cannot even image the planet directly. Yet we can still ask what the atmosphere contains because light is chemically vulnerable. Put gas in front of a light source, and some wavelengths survive better than others.

Transmission spectroscopy uses that vulnerability. When a planet transits, the opaque body blocks starlight at every wavelength. But the thin ring of atmosphere around the planet is selective. Molecules such as water vapor, carbon dioxide, methane, carbon monoxide, sodium, and potassium absorb particular wavelength bands. The atmosphere makes the planet look slightly larger where it is more opaque.

0.6-5.3 umJWST/NIRSpec PRISM-like coverage used in many public transmission-spectrum diagrams.
2.32e-4Extra sodium-band absorption reported for HD 209458 b by Charbonneau et al. (2002), about 232 ppm.
4.3 umA major CO2 band that made WASP-39 b a landmark Webb target.
sqrt(N)The friendly part of repeat observations: independent random noise averages down roughly as the square root of the number of transits.
Chapter 01 - The metaphor

First, no, we are not literally smelling the planet.

Smell is chemistry interpreted by biology. A molecule reaches your nose, binds to receptors, and your brain turns that molecular contact into perception. Exoplanet astronomy obviously does not work like that. No carbon dioxide molecule from WASP-39 b is drifting into a telescope and politely introducing itself.

But the metaphor is useful because the scientific act is similar in spirit: identify chemical species from the signatures they leave. A nose uses receptors. A spectrograph uses wavelength. Molecules do not need to arrive at Earth; their absorption fingerprints can travel inside the starlight.

We do not smell the planet by collecting its air. We smell it by noticing which colors of starlight its air refused to let through.
Chapter 02 - The transit window

A transit turns a star into a backlight.

The ordinary transit method measures the total dip in brightness when a planet crosses the stellar disk. Transmission spectroscopy adds one question: does the dip have exactly the same depth at every wavelength?

If the planet had no atmosphere and the star were perfectly simple, the depth would mainly be the area ratio between planet and star. With an atmosphere, some starlight skims through the limb of the planet before reaching us. At wavelengths where atmospheric gases absorb strongly, that limb becomes more opaque, so the effective planet radius increases slightly.

Transit depth with wavelength dependence
\[ \Delta F(\lambda) \simeq \left(\frac{R_p(\lambda)}{R_\star}\right)^2 \]

Here \(R_p(\lambda)\) is not the solid surface radius. It is the effective radius of the planet plus the part of the atmosphere that blocks light at wavelength \(\lambda\).

Opaque bodyBlocks starlight almost equally across wavelength, giving the basic transit depth.
Atmospheric limbBlocks extra light at molecular absorption bands, creating peaks in the transmission spectrum.
Out-of-transit baselineThe unfiltered stellar spectrum gives the reference for what the star looked like without the planet in front.
In-transit spectrumThe filtered spectrum contains the small atmospheric imprint, if the signal survives noise and systematics.
Chapter 03 - The shape of the clue

A transmission spectrum is a graph of missing light.

In many diagrams, the vertical axis is the amount of light blocked and the horizontal axis is wavelength. A flat line would mean the apparent planet size is nearly the same across the observed range. Peaks mean more light is blocked at those wavelengths, often because a molecule absorbs there.

This is slightly counterintuitive at first. In a normal stellar absorption spectrum, absorption appears as dips in flux. In a transmission spectrum plotted as transit depth, absorption often appears as a peak because the planet looks bigger at that wavelength. Same physics, different bookkeeping.

Approximate atmospheric scale
\[ H = \frac{k_B T}{\mu m_H g} \]

The atmospheric scale height \(H\) grows with temperature and shrinks with stronger gravity or heavier mean molecular weight \(\mu\). Here \(\mu\) is in proton-mass units and \(m_H\) is the hydrogen mass. Puffy hot atmospheres produce larger transmission features than compact heavy atmospheres.

Signal directionAbsorption appears upward when the graph is transit depth, because the atmosphere makes the planet's apparent radius larger at selected wavelengths.
Useful wavelength spanJWST/NIRISS, NIRSpec, NIRCam, and MIRI together turn the trick from optical sodium lines into near- and mid-infrared chemistry.
Feature widthReal bands are not single needles. Instrument resolution, pressure broadening, temperature structure, and clouds smear molecular opacity across wavelength.
Model warningA labelled peak is a hypothesis, not a trophy. Retrievals compare families of atmospheres against the data and the instrument systematics.
Transmission spectroscopy diagram showing transit geometry, an atmospheric chemical filter, and a JWST-style spectrum with molecular absorption features
Original visual concept: Biswajit Jana. The image compresses the measurement chain into three steps: transit geometry, atmospheric filtering, and the final wavelength-dependent chemical fingerprint.
Chapter 04 - Order-of-magnitude physics

How loud is an atmosphere in ppm?

The most useful intuition is that transmission spectroscopy measures an annulus, not a disk. The opaque planet blocks an area roughly \(\pi R_p^2\). The atmosphere adds a thin wavelength-dependent ring. If a strong molecular band samples \(N_H\) scale heights, the extra transit depth is approximately:

Atmospheric annulus signal
\[ \Delta D_\mathrm{atm} \approx \frac{2N_H R_p H}{R_\star^2} \]

This is not a retrieval model. It is a scaling law. It explains why hot Jupiters around small stars are generous, and Earth-like atmospheres around Sun-like stars are brutal.

Hot Jupiter around Sun-like star\(T=1200\) K, \(\mu=2.3\), \(g=10\) m/s^2 gives \(H\approx430\) km. Five scale heights around a 1.2 \(R_J\) planet can be roughly 760 ppm.
Warm Neptune around M dwarf\(T=700\) K, \(\mu=2.8\), \(g=11\) m/s^2, \(R_\star=0.45R_\odot\): five scale heights can land near 480 ppm.
Earth-like heavy atmosphere\(T=250\) K, \(\mu=28\), \(g=9.8\) m/s^2 gives \(H\approx8\) km. Around a 0.12 \(R_\odot\) star, five scale heights are still only about 70 ppm.
Hydrogen-rich super-EarthA lower-\(\mu\) atmosphere can produce hundreds of ppm, but that may mean the planet is not Earth-like in the way people emotionally want it to be.
Opacity as the hidden variable
\[ R_p^2(\lambda)=R_0^2+2\int_{R_0}^{\infty}\left[1-e^{-\tau_\lambda(b)}\right]b\,db \]

The effective radius comes from integrating over atmospheric chords. The optical depth \(\tau_\lambda\) depends on abundance, pressure, temperature, clouds, and molecular cross-sections.

Chapter 05 - Molecular fingerprints

Molecules are picky about wavelength, which is convenient for us.

Molecules absorb light when photons match allowed transitions in their rotational and vibrational energy structure. The result is not a single neat barcode line, but a pattern of bands. Water vapor has strong near-infrared bands. Carbon dioxide has a prominent feature around 4.3 microns. Methane absorbs in several infrared bands. Sodium and potassium can leave visible and near-infrared atomic signatures.

The exact spectrum depends on temperature, pressure, abundance, clouds, haze, stellar contamination, and instrument response. That is why the job is not simply spotting a bump and shouting a molecule's name. Atmospheric retrieval is model comparison under uncertainty.

Water vaporOften searched for in near-infrared transmission spectra, especially around broad bands near 1.4 microns and beyond.
Carbon dioxideImportant near 4.3 microns and useful for formation and atmospheric composition studies.
MethaneChemically interesting, especially for cooler planets, but can be absent or depleted in hot atmospheres.
Sodium and potassiumAtomic features that can appear in optical transmission spectra, sometimes shaped by haze and pressure broadening.
Chapter 06 - Clouds and hazes

Clouds can erase the confession.

An atmosphere may contain the molecules we are looking for and still refuse to show them clearly. Clouds and hazes can mute spectral features by blocking the deeper atmospheric layers where stronger absorption would otherwise be visible. In a transmission spectrum, this can flatten the line into something frustratingly quiet.

This is not failure. A flat spectrum is also information. It may indicate a high-altitude cloud deck, haze scattering, a compact high-metallicity atmosphere, low scale height, or simply data that are not precise enough. The universe has many ways to say "try again, but with better error bars."

The simulation above includes a cloud-opacity control because it is one of the fastest ways to understand why atmospheric retrieval is hard: the same molecule can look obvious in a clear atmosphere and nearly invisible behind haze.
Chapter 07 - Precision problem

The signal is tiny, which is where instrumentation enters the room.

Transmission features are often measured in parts per million or small fractions of a percent. The star is bright. The atmospheric annulus is thin. The instrument is imperfect. The star may have spots, faculae, limb darkening, and time-variable behavior. The detector may have ramps, drifts, persistence, flat-field issues, and wavelength calibration quirks.

This is why transmission spectroscopy is not only an astrophysics problem. It is also an instrumentation problem, a detector problem, a calibration problem, and a statistics problem. The planet's atmosphere is not just hidden behind distance. It is hidden behind every small imperfection in the measurement chain.

Cartoon signal-to-noise idea
\[ \mathrm{SNR} \propto \frac{\mathrm{feature\ amplitude}}{\mathrm{noise}/\sqrt{N_\mathrm{transits}}} \]

More repeated transits can improve precision when noise behaves nicely. Real systematics are less generous, because they do not always average down like independent random noise.

Spectral resolving power
\[ R = \frac{\lambda}{\Delta\lambda} \]

Higher \(R\) separates narrower features, but often with fewer photons per bin. In practice, observers trade spectral resolution against signal-to-noise and the width of the molecular band they want to test.

Chapter 08 - Real telescope examples

JWST turned this from promising to spectacularly useful.

The James Webb Space Telescope has made transmission spectroscopy feel less like a distant ambition and more like a working atmospheric laboratory. WASP-39 b is a famous example because Webb observations revealed a rich transmission spectrum with species such as water, carbon dioxide, carbon monoxide, sulfur dioxide, sodium, and potassium discussed in NASA's public release material.

The same idea appears across other targets. WASP-96 b showed water vapor signatures with evidence that clouds and possible haze affected the shape. WASP-107 b combined Hubble and Webb measurements across a broad wavelength range, showing how multiple instruments can build a wider atmospheric picture. WASP-17 b even brought mineral cloud chemistry into the conversation through mid-infrared spectroscopy.

WASP-39 bNASA lists Na, K, H2O, CO, CO2, H2S, and SO2 in Webb public material. The 4.3 um CO2 feature became one of the headline results.
WASP-96 bNIRISS covered 0.6-2.8 um in a single observation, showing water vapor plus clouds/haze inferred from muted feature heights.
WASP-107 bHubble WFC3, Webb NIRCam, and Webb MIRI together produced a broad spectrum with H2O, CO2, CO, CH4, SO2, and NH3 discussed by NASA.
WASP-17 bWebb/MIRI low-resolution spectroscopy from about 5-12 um pointed to quartz/silica cloud particles.
Chapter 09 - Retrievals

A spectrum is not the atmosphere. It is a puzzle with priors.

Atmospheric retrieval means proposing many possible atmospheres, generating spectra from them, comparing those spectra with the data, and asking which parameter combinations survive. The output is usually not a single atmosphere but a probability distribution over temperature, radius reference pressure, abundances, cloud properties, and sometimes chemistry assumptions.

This is where degeneracy enters. A weaker water band could mean less water. It could also mean higher clouds, higher mean molecular weight, lower temperature, smaller scale height, or uncorrected stellar contamination. Retrieval is the art of being honest about those overlapping explanations.

Forward modelChoose temperature-pressure structure, chemistry, opacity tables, cloud prescriptions, and instrument response; compute the expected spectrum.
LikelihoodCompare model and data with uncertainties, including correlated noise when the detector or reduction pipeline demands it.
PosteriorUse Bayesian sampling or nested sampling to map which atmospheres are plausible rather than pretending one curve is the truth.
Chapter 10 - Stellar contamination

The star can forge atmospheric features.

Transmission spectroscopy assumes we know the stellar light source. Real stellar disks have spots, faculae, granulation, limb darkening, and wavelength-dependent surface structure. If the planet crosses one stellar region while the out-of-transit disk contains a different mixture of surface temperatures, the inferred transmission spectrum can be biased.

This is especially important for small planets around active cool stars. A planet may appear to have a spectral slope or molecular-looking feature because the star is heterogeneous. Rackham, Apai, and Giampapa call this the transit light source effect, and it is one reason biosignature claims will need extreme caution.

Contamination idea
\[ D_\mathrm{obs}(\lambda) \neq D_\mathrm{planet}(\lambda) \]

The measured depth can include the planet plus wavelength-dependent stellar surface effects. The inequality is the whole headache.

Chapter 11 - The life question

This is also why biosignatures are hard.

Once we can read atmospheric chemistry, the tempting next question is life. Could oxygen, methane, carbon dioxide, water vapor, ozone, or other combinations reveal biology? Maybe, but the word "maybe" is doing real work. A molecule is not automatically a biosignature just because biology can produce it.

Context matters. Stellar type, planetary temperature, surface pressure, atmospheric escape, photochemistry, volcanic activity, ocean chemistry, clouds, and false-positive pathways all matter. A convincing biosignature will likely require a pattern of gases in environmental context, not a single heroic molecule carrying the whole claim on its back.

Transmission spectroscopy may one day help us search for life, but it will not let us skip chemistry, geology, stellar physics, or caution.
Chapter 12 - Instrumentation bridge

This is where exoplanets meet the lab bench.

The reason I like this topic is that it sits exactly between astrophysics and instrumentation. The scientific question is cosmic: what are distant worlds made of? The practical question is brutally local: how stable is the spectrograph, how well do we calibrate the detector, how repeatable is the optical path, how much systematic noise is being smuggled into the result?

That is the bridge from transmission spectroscopy to precision spectrographs, feedback control, detector stability, and environmental monitoring. An atmospheric feature in an exoplanet spectrum is only as convincing as the chain of optics, electronics, software, calibration, and statistical modelling that survived between star and conclusion.

In other words: to smell an exoplanet, first build an instrument that knows how not to smell itself.
Takeaway

The planet never visits us. Its filtered starlight does.

Transmission spectroscopy is one of the most elegant tricks in modern astronomy. It turns a transit into chemistry, a spectrum into an atmosphere, and a distant unresolved planet into a physical world with gases, clouds, temperature structure, and history.

It is not easy. The signal is tiny, the models are conditional, the instruments must behave, and clouds enjoy ruining simple explanations. But when it works, the result is astonishing: a planet hundreds of light-years away leaves enough of itself in starlight for us to begin reading its air.

References and source notes

  1. Seager, S. and Sasselov, D. D. (2000). Theoretical transmission spectra during extrasolar giant planet transits.
  2. Brown, T. M. (2001). Transmission spectra as diagnostics of extrasolar giant planet atmospheres.
  3. Charbonneau, D. et al. (2002). Detection of an extrasolar planet atmosphere through sodium absorption.
  4. Lecavelier des Etangs, A. et al. (2008). Rayleigh scattering in the transit spectrum of HD 189733 b.
  5. Sing, D. K. et al. (2016). A continuum from clear to cloudy hot-Jupiter exoplanets without primordial water depletion.
  6. Madhusudhan, N. (2019). Exoplanetary atmospheres: key insights, challenges, and prospects.
  7. Rackham, B. V., Apai, D. and Giampapa, M. S. (2018). The transit light source effect.
  8. Kempton, E. M.-R. et al. (2018). A framework for prioritizing the TESS planetary candidates most amenable to atmospheric characterization.
  9. NASA Science. Exoplanet WASP-39 b transmission spectra.
  10. ESA/Webb. Hot Gas Giant Exoplanet WASP-39 b NIRSpec transmission spectrum.
  11. NASA Science. Webb detects carbon dioxide in an exoplanet atmosphere.
  12. NASA Science. Exoplanet WASP-96 b NIRISS transmission spectrum.
  13. NASA Science. WASP-107 b Hubble, NIRCam, and MIRI transmission spectrum.
  14. NASA Science. WASP-17 b MIRI transmission spectrum.
  15. JWST Transiting Exoplanet Community Early Release Science Program papers on WASP-39 b, including NIRISS, NIRSpec, NIRCam, and photochemical SO2 analyses.

Comments and reactions

Comments are powered by GitHub Discussions through giscus. Leave a question, correction, or reaction below.