Exploring the diversity of exoplanets

Dr Suzanne Aigrain

Cormack lecture 2012 Putting the Earth in context

Earthrise seen by Apollo 8 crew, December 1968 © NASA Planets of the solar system

Sizes are to scale, distances are not

asteroid Gaspra asteroid Gaspra

Halley’s comet (since 240 BCE) asteroid Gaspra

Halley’s comet (since 240 BCE) Nebular theory Swedenborg, Kant, Laplace (1700’s) 100 000 000 000 stars in the Milky Way 100 000 000 000 galaxies in the Universe Star nurseries... planet nurseries Star nurseries... planet nurseries The discovery of Uranus

William & Caroline Herschel (1781) The discovery of Uranus

William & Caroline Herschel (1781) The discovery of Uranus

William & Caroline Herschel (1781) The discovery of Neptune

Orbit of Uranus estimated from a century of observations The discovery of Neptune

Orbit of Uranus estimated from a century of observations

Existence of a new planet predicted by Adams & Le Verrier (1845) The discovery of Neptune

Orbit of Uranus estimated from a century of observations

Existence of a new planet predicted by Adams & Le Verrier (1845) Neptune observed at predicted position the same year by Galle Finding exoplanets I. direct imaging HR 8799 (Marois et al. 2008) Finding exoplanets I. direct imaging HR 8799 (Marois et al. 2008) Finding exoplanets II. radial velocity Finding exoplanets II. radial velocity

dispersing element Finding exoplanets II. radial velocity

absorption lines

dispersing element Finding exoplanets II. radial velocity 51 Peg b (Mayor & Queloz 1995) Finding exoplanets II. radial velocity 51 Peg b (Mayor & Queloz 1995)

hot Jupiters Finding exoplanets III. transits

RV + transits = size + mass = density = buk composition Finding exoplanets III. transits HD 209 458 b (Charbonneau et al. 2000)

RV + transits = size + mass = density = buk composition Finding exoplanets III. transits HD 209 458 b (Charbonneau et al. 2000)

STARE

RV + transits = size + mass = density = buk composition >750 exoplanets known today

Jupiter

Neptune Observational limitations

Earth Mass-Radius relation

Data from www.inscience.ch/transits (maintained by F. Pont & N. Husnoo) Mass-Radius relation

low mass stars

Data from www.inscience.ch/transits (maintained by F. Pont & N. Husnoo)

Mass-radius relations from Fressin et al. (2007), Seager et al. (2007).

brown dwarfs

pure H/He Mass-Radius relation

low mass stars

Data from www.inscience.ch/transits (maintained by F. Pont & N. Husnoo)

Mass-radius relations from Fressin et al. (2007), Irradiation Seager et al. (2007).

brown dwarfs

pure H/He Mass-Radius relation

low mass stars

Data from www.inscience.ch/transits (maintained by F. Pont & N. Husnoo)

Mass-radius relations from Fressin et al. (2007), Irradiation Seager et al. (2007).

solid core brown dwarfs

pure H/He pure H2O

rock Mass-Radius relation

hot Jupiters low mass stars

Data from www.inscience.ch/transits (maintained by F. Pont & N. Husnoo)

Mass-radius relations from Fressin et al. (2007), Irradiation Seager et al. (2007).

solid core brown dwarfs

pure H/He pure H2O

rock Mass-Radius relation

hot Jupiters low mass stars

Data from www.inscience.ch/transits (maintained by F. Pont & N. Husnoo)

Mass-radius relations from Fressin et al. (2007), Irradiation Seager et al. (2007).

solid core brown dwarfs

pure H/He pure H2O

hot Neptunes rock The first transiting hot Neptune: Gliese 436b The first transiting hot Neptune: Gliese 436b

Gillon et al. (2007) The first transiting hot Neptune: Gliese 436b

Gillon et al. (2007) Hot ice going to spaceTerrestrial planets? CoRoT A CoRoT lightcurve

intermediate

stellar variability (>1d)

residuals (smoothed) The first transiting super-Earth: CoRoT-7b

P = 0.8541 d, 0.3 mmag deep The first transiting super-Earth: CoRoT-7b

P = 0.8541 d, 0.3 mmag deep Kepler: taking exoplanet discovery to the next level

Planet Hunters 1 - a circumbinary planet in a quadruple system

Schwamb et al. (2012) A transit closer to home

www.super-collider.com Transit of Venus

First predicted by Johannes Kepler in 1627 (for 1631) First observed by Jeremiah Horrocks in 1639 Measuring the scale of the solar system

First predicted by Johannes Kepler in 1627 (for 1631) First observed by Jeremiah Horrocks in 1639 Observations from widely spaced points on Earth proposed by Edmund Halley (1691, 1716) Measuring the scale of the solar system

First predicted by Johannes Kepler in 1627 (for 1631) First observed by Jeremiah Horrocks in 1639 Observations from widely spaced points on Earth proposed by Edmund Halley (1691, 1716) Successful expeditions 1761, 1769 (Captain Cook) The first extra-terrestrial atmosphere

Mikhail Lomonosov (1761)

2004 transit © D. Kiselman, Royal Swedish Academy of Sciences Using transits to study exoplanet atmospheres

secondary eclipse: emitted or reflected light

transit: planet radius phase curve: temperature distribution and dynamics transmission spectra: absorbers in atmosphere Transmission spectroscopy

to observer

star light

planet Effect of clouds

to observer

star light

planet Effect of gases

to observer

star light

planet How do we measure this? How do we measure this? How do we measure this?

grism How do we measure this?

grism How do we measure this?

grism

4 N. P. Gibson et al.

Figure 3. Extracted 1D spectra of a typical in-transit (green) and out-of-transit (blue) observation of HD 189733 showing the number of electrons collected per pixel channel. The wavelength decreases with increasing x position.

Figure 2. NICMOS image of HD 189733 taken with the G206 grism. The blue line marks the centroids of the first-order spec- trum of HD 189733, with the green lines flanking representing the extent of the extraction region. Immediately left is the second- order spectrum, and below is a companion star to HD 189733. The bright band at the right of the detector is a feature of G206 grism spectra, caused by the warm edge of the aperture mask.

was used to extract the flux for each wavelength channel, after subtraction of a background value for each pixel. A width of 35 pixels was chosen to minimise the RMS in the white light curve of orbits 2, 4 and 5. This was repeated for 90 pixel columns along the dispersion axis. The extraction regions used are marked in Fig. 2. For the background sub- traction, we experimented with various techniques. First, a global background subtraction was used, similarly to S08. Figure 4. Raw ‘white’ light curve of HD 189733 obtained by in- The background was taken to be the average of a large un- tegrating the flux from each spectra over all wavelengths, showing illuminated region above the spectral trace (we tested using the sampling of the transit, alongside its best-fit model. While all dierent regions). This is not ideal, as previously mentioned orbits are seen to suer from systematics, orbit one exhibits by the background varies spatially over the detector. As a first far the largest, commonly attributed to spacecraft ‘settling’, and order correction, we instead calculated the background sep- is excluded from subsequent analysis. arately for each pixel column, as the average value of the un-illuminated region above the spectral trace along the col- umn. Again we note that this does not provide a satisfac- curves. A ‘white’ light curve was constructed by integrating tory correction, as the background varies along both x and y. the flux over the entire wavelength range for each image. Both global and wavelength dependent corrections were used This was extended to include 110 pixel columns, which min- for the subsequent analysis. For the remainder of this paper, imises systematics arising from small changes in the position we will display results from data that is not flat-fielded, and of the spectral trace. This is plotted in Fig. 4, and shows the using separate columns for background subtraction, but our sampling of the light curve over the five orbits. Systemat- conclusions remain the same for each case. ics are evident in each orbit, but particularly for orbit one. Extracted spectra from a typical in-transit and out-of- This is commonly found in similar NICMOS data (see e.g. transit observation are shown in Fig. 3, giving approximately Sects. 3.1 and 4.1). It is attributed to spacecraft ‘settling’ 430000 electrons in the brightest pixel channel, and approx- (e.g. S08; Pont et al. 2009), and orbit one is excluded for the imately 120000 electrons in the faintest channel. The fea- remainder of this work. tures in the spectra do not correspond to real stellar fea- The raw light curves are shown in Fig. 5 for each of the tures, but result in variation of the sensitivity of the grism 18 wavelength channels, after normalising by fitting a line with wavelength. Each wavelength channel in the 1D spec- through orbits 2, 4 and 5, and dividing the light curves by tra is then used to construct a time-series, after binning in this function. The light curves are clearly seen to exhibit 5 pixels along the dispersion direction, resulting in 18 light strong time-correlated noise, which needs to be removed if

c 2002 RAS, MNRAS 000,1–17 How do we measure this?

grism

4 N. P. Gibson4 N. P. et Gibson al. et al.

Figure 3. Extracted 1D spectra of a typical in-transit (green) and out-of-transit (blue) observation of HD 189733 showing the number of electrons collected per pixel channel. The wavelength decreases with increasing x position.

Figure 2. NICMOS image of HD 189733 taken with the G206 Figure 3. Extracted 1D spectra of a typical in-transit (green) grism. The blue line marks the centroids of the first-order spec- and out-of-transit (blue) observation of HD 189733 showing the trum of HD 189733, with the green lines flanking representing the extent of the extraction region. Immediately left is the second- number of electrons collected per pixel channel. The wavelength order spectrum, and below is a companion star to HD 189733. decreases with increasing x position. The bright band at the right of the detector is a feature of G206 grism spectra, caused by the warm edge of the aperture mask.

Figure 2. NICMOS image of HD 189733 taken with the G206 was used to extract the flux for each wavelength channel, grism. The blueafter line marks subtraction the of centroids a background of the value first-order for each pixel. spec- A trum of HD 189733,width with of 35 the pixels green was lines chosen flanking to minimise representing the RMS in the the extent of the extractionwhite light region. curve of Immediately orbits 2, 4 and 5. left This is was the repeated second- for order spectrum,90 and pixel below columns is a along companion the dispersion star axis. to HD The 189733. extraction regions used are marked in Fig. 2. For the background sub- The bright bandtraction, at the right we experimented of the detector with various is a feature techniques. of G206First, a grism spectra, causedglobal by background the warm subtraction edge of was the used, aperture similarly mask. to S08. Figure 4. Raw ‘white’ light curve of HD 189733 obtained by in- The background was taken to be the average of a large un- tegrating the flux from each spectra over all wavelengths, showing illuminated region above the spectral trace (we tested using the sampling of the transit, alongside its best-fit model. While all dierent regions). This is not ideal, as previously mentioned orbits are seen to suer from systematics, orbit one exhibits by was used to extractthe background the flux varies for spatially each wavelength over the detector. channel, As a first far the largest, commonly attributed to spacecraft ‘settling’, and order correction, we instead calculated the background sep- is excluded from subsequent analysis. after subtractionarately of a for background each pixel column, value as the for average each value pixel. of A the width of 35 pixelsun-illuminated was chosen region to above minimise the spectral the trace RMS along in the the col- white light curveumn. of Againorbits we 2, note 4 and that 5. this This does was not provide repeated a satisfac- for curves. A ‘white’ light curve was constructed by integrating 90 pixel columnstory along correction, the as dispersion the background axis. varies The along extraction both x and y. the flux over the entire wavelength range for each image. Both global and wavelength dependent corrections were used This was extended to include 110 pixel columns, which min- regions used arefor marked the subsequent in Fig. analysis. 2. For For the the remainder background of this sub- paper, imises systematics arising from small changes in the position traction, we experimentedwe will display resultswith variousfrom data techniques.that is not flat-fielded, First, and a of the spectral trace. This is plotted in Fig. 4, and shows the global backgroundusing subtraction separate columns was for background used, similarly subtraction, to but S08. our sampling of the light curve over the five orbits. Systemat- conclusions remain the same for each case. icsFigure are evident 4. Raw in each ‘white’ orbit, but light particularly curve of for HD orbit 189733 one. obtained by in- The background was taken to be the average of a large un- Extracted spectra from a typical in-transit and out-of- Thistegrating is commonly the flux found from in similar each spectra NICMOS over data all (see wavelengths, e.g. showing transit observation are shown in Fig. 3, giving approximately Sects. 3.1 and 4.1). It is attributed to spacecraft ‘settling’ illuminated region above the spectral trace (we tested using the sampling of the transit, alongside its best-fit model. While all dierent regions).430000 This electrons is not in ideal, the brightest as previously pixel channel, mentioned and approx- (e.g. S08; Pont et al. 2009), and orbit one is excluded for the imately 120000 electrons in the faintest channel. The fea- remainderorbits are of this seen work. to suer from systematics, orbit one exhibits by the backgroundtures varies in the spatially spectra do over not the correspond detector. to real As stellar a first fea- farThe the raw largest, light curves commonly are shown attributed in Fig. 5 for to each spacecraft of the ‘settling’, and order correction,tures, we but instead result incalculated variation of the the sensitivity background of the sep- grism 18is wavelength excluded from channels, subsequent after normalising analysis. by fitting a line arately for eachwith pixel wavelength. column, Each as wavelength the average channel value in the of 1D the spec- through orbits 2, 4 and 5, and dividing the light curves by tra is then used to construct a time-series, after binning in this function. The light curves are clearly seen to exhibit un-illuminated region5 pixels along above the the dispersion spectral direction, trace resulting along the in 18 col- light strong time-correlated noise, which needs to be removed if umn. Again we note that this does not provide a satisfac- curves. A ‘white’ light curve was constructed by integrating c 2002 RAS, MNRAS 000,1–17 tory correction, as the background varies along both x and y. the flux over the entire wavelength range for each image. Both global and wavelength dependent corrections were used This was extended to include 110 pixel columns, which min- for the subsequent analysis. For the remainder of this paper, imises systematics arising from small changes in the position we will display results from data that is not flat-fielded, and of the spectral trace. This is plotted in Fig. 4, and shows the using separate columns for background subtraction, but our sampling of the light curve over the five orbits. Systemat- conclusions remain the same for each case. ics are evident in each orbit, but particularly for orbit one. Extracted spectra from a typical in-transit and out-of- This is commonly found in similar NICMOS data (see e.g. transit observation are shown in Fig. 3, giving approximately Sects. 3.1 and 4.1). It is attributed to spacecraft ‘settling’ 430000 electrons in the brightest pixel channel, and approx- (e.g. S08; Pont et al. 2009), and orbit one is excluded for the imately 120000 electrons in the faintest channel. The fea- remainder of this work. tures in the spectra do not correspond to real stellar fea- The raw light curves are shown in Fig. 5 for each of the tures, but result in variation of the sensitivity of the grism 18 wavelength channels, after normalising by fitting a line with wavelength. Each wavelength channel in the 1D spec- through orbits 2, 4 and 5, and dividing the light curves by tra is then used to construct a time-series, after binning in this function. The light curves are clearly seen to exhibit 5 pixels along the dispersion direction, resulting in 18 light strong time-correlated noise, which needs to be removed if

c 2002 RAS, MNRAS 000,1–17 How do we measure this?

grism

4 N. P. Gibson4 N. P. et Gibson al. et al. University of Exeter — Astrophysics Group / CGAFD Rolling Grant Application — June 2008

Figure 3: Hubble Space Telescope time series for the transiting planet system HD 189733 in ten wavelength intervals between 0.6 and 1.05 mi- crons (Pont et al. 2008). The transmission spectrum of the planetary at- mosphere is imprinted in these data as minute depth dierences in the transit signals. Figure 3. Extracted 1D spectra of a typical in-transit (green) and out-of-transit (blue) observation of HD 189733 showing the number of electrons collected per pixel channel. The wavelength decreases with increasing x position.

Figure 2. NICMOS image of HD 189733 taken with the G206 Figure 3. Extracted 1D spectra of a typical in-transit (green) grism. The blue line marks the centroids of the first-order spec- and out-of-transit (blue) observation of HD 189733 showing the Transmission spectroscopy with the HST presents specific trum of HD 189733, with the green lines flanking representing the Analysis of Space Telescope spectroscopic time series extent of the extraction region. Immediately left is the second- number of electrons collected per pixel channel. The wavelength data analysis challenges. The global signal-to-noise of the data is enormous, because the target stars are very bright order spectrum, and below is a companion star to HD 189733. decreases with increasing x position. The bright band at the right of the detector is a feature of G206 for the 2.4 m HST mirror. The spectrum typically saturates the whole detector in seconds. But the signal to measure grism spectra, caused by the warm edge of the aperture mask. 5 is extremely tiny, the wavelength dependence of the transit depth being a few 10 of the total flux. Several sources Figure 2. NICMOS image of HD 189733 taken with the G206 of noise are higher than this level, such as instrument/telescope systematics, uncertainties due to the flux distribution was used to extract the flux for each wavelength channel, grism. The blueafter line marks subtraction the of centroids a background of the value first-order for each pixel. spec- A on the surface of the host star (limb darkening and star spots). To exploit the potential extreme signal-to-noise ratio trum of HD 189733,width with of 35 the pixels green was lines chosen flanking to minimise representing the RMS in the the extent of the extractionwhite light region. curve of Immediately orbits 2, 4 and 5. left This is was the repeated second- for of the data series, exquisitely refined decorrelation of the systematics must be perfected. order spectrum,90 and pixel below columns is a along companion the dispersion star axis. to HD The 189733. extraction regions used are marked in Fig. 2. For the background sub- We have done this for the ACS data on HD 189733, reaching the highest accuracy ever for a planetary transmission The bright bandtraction, at the right we experimented of the detector with various is a feature techniques. of G206First, a grism spectra, causedglobal by background the warm subtraction edge of was the used, aperture similarly mask. to S08. spectrum (equivalent to 50 km height on the planet, Pont et al. 2008), and are working on NICMOS data for the Figure 4. Raw ‘white’ light curve of HD 189733 obtained by in- The background was taken to be the average of a large un- tegrating the flux from each spectra over all wavelengths, showing hot Neptune GJ 436b, and the WASP-1, WASP-2 and WASP-3 transiting systems. We have applied for time on the illuminated region above the spectral trace (we tested using the sampling of the transit, alongside its best-fit model. While all dierent regions). This is not ideal, as previously mentioned orbits are seen to suer from systematics, orbit one exhibits by refurbished HST to obtain a comprehensive transmission spectrum of HD 189733b with STIS (or COS in case of was used to extractthe background the flux varies for spatially each wavelength over the detector. channel, As a first far the largest, commonly attributed to spacecraft ‘settling’, and failure to repair STIS) and NICMOS. With the GJ 436 data, we have the first spectrum of a Neptune-mass extrasolar after subtractionorder of correction, a background we instead value calculated for the each background pixel. sep- A is excluded from subsequent analysis. arately for each pixel column, as the average value of the planet, a significant step towards the ultimate goal of biosignature detection on Earth-like planets. width of 35 pixelsun-illuminated was chosen region to above minimise the spectral the trace RMS along in the the col- white light curveumn. of Againorbits we 2, note 4 and that 5. this This does was not provide repeated a satisfac- for curves. A ‘white’ light curve was constructed by integrating tory correction, as the background varies along both x and y. the flux over the entire wavelength range for each image. Advanced decorrelation As a key part of this project, we will perfect the decorrelation algorithm for space-grade 90 pixel columns along the dispersion axis. The extraction Both global and wavelength dependent corrections were used This was extended to include 110 pixel columns, which min- regions used arefor marked the subsequent in Fig. analysis. 2. For For the the remainder background of this sub- paper, imises systematics arising from small changes in the position spectroscopic time series. Our work on correlated noise in time series (Pont, Zucker & Queloz 2006, MNRAS 373, traction, we experimentedwe will display resultswith variousfrom data techniques.that is not flat-fielded, First, and a of the spectral trace. This is plotted in Fig. 4, and shows the 231), has been parallel with the development of the widely-used decorrelation algorithm Sysrem (Tamuz, Mazeh global backgroundusing subtraction separate columns was for background used, similarly subtraction, to but S08. our sampling of the light curve over the five orbits. Systemat- conclusions remain the same for each case. icsFigure are evident 4. Raw in each ‘white’ orbit, but light particularly curve of for HD orbit 189733 one. obtained by in- & Zucker 2005, MNRAS 356, 1466), and we have a close on-going collaboration with the Tel-Aviv group (S. The background was taken to be the average of a large un- Extracted spectra from a typical in-transit and out-of- Thistegrating is commonly the flux found from in similar each spectra NICMOS over data all (see wavelengths, e.g. showing transit observation are shown in Fig. 3, giving approximately Sects. 3.1 and 4.1). It is attributed to spacecraft ‘settling’ Zucker, T. Mazeh, O. Tamuz). We plan to use the same matrix approach for the HST data. The basic idea is to illuminated region above the spectral trace (we tested using the sampling of the transit, alongside its best-fit model. While all dierent regions).430000 This electrons is not in ideal, the brightest as previously pixel channel, mentioned and approx- (e.g. S08; Pont et al. 2009), and orbit one is excluded for the consider spectrum time series as single matrices, with the two dimensions being wavelength and time, and to isolate imately 120000 electrons in the faintest channel. The fea- remainderorbits are of this seen work. to suer from systematics, orbit one exhibits by the backgroundtures varies in the spatially spectra do over not the correspond detector. to real As stellar a first fea- farThe the raw largest, light curves commonly are shown attributed in Fig. 5 for to each spacecraft of the ‘settling’, and the planetary signal, instrumental systematics and wavelength-dependent variations of the host star by treating the order correction,tures, we but instead result incalculated variation of the the sensitivity background of the sep- grism 18is wavelength excluded from channels, subsequent after normalising analysis. by fitting a line matrix as a whole. Sysrem simply finds the eigenvectors of the object versus time matrix in multi-object photometry. arately for eachwith pixel wavelength. column, Each as wavelength the average channel value in the of 1D the spec- through orbits 2, 4 and 5, and dividing the light curves by tra is then used to construct a time-series, after binning in this function. The light curves are clearly seen to exhibit But in the time of HST spectral time series, the wavelength and time residuals are highly correlated, and this un-illuminated region5 pixels along above the the dispersion spectral direction, trace resulting along the in 18 col- light strong time-correlated noise, which needs to be removed if umn. Again we note that this does not provide a satisfac- curves. A ‘white’ light curve was constructed by integrating must be taken into account. We hope that the eort of perfecting such a matrix-based decorrelation algorithm for c 2002 RAS, MNRAS 000,1–17 tory correction, as the background varies along both x and y. the flux over the entire wavelength range for each image. spectral time series will be repaid by more robust and higher signal-to-noise transmission spectroscopy of planetary Both global and wavelength dependent corrections were used This was extended to include 110 pixel columns, which min- for the subsequent analysis. For the remainder of this paper, imises systematics arising from small changes in the position we will display results from data that is not flat-fielded, and of the spectral trace. This is plotted in Fig. 4, and shows the using separate columns for background subtraction, but our sampling of the light curve over the five orbits. Systemat- conclusions remain the same for each case. ics are evident in each orbit, but particularly for orbit one. Extracted spectra from a typical in-transit and out-of- This is commonly found in similar NICMOS data (see e.g. transit observation are shown in Fig. 3, giving approximately Sects. 3.1 and 4.1). It is attributed to spacecraft ‘settling’ Figure 4: HST time series of the water 1.4-micron absorption band in the 430000 electrons in the brightest pixel channel, and approx- (e.g. S08; Pont et al. 2009), and orbit one is excluded for the imately 120000 electrons in the faintest channel. The fea- remainder of this work. atmosphere of the transiting ice planet around GJ 436 (HST programme tures in the spectra do not correspond to real stellar fea- The raw light curves are shown in Fig. 5 for each of the GO-11306, PI Pont). The accuracy is su⇥cient to detect the presence of water at the 1 level, showing the way ahead for the characterisation tures, but result in variation of the sensitivity of the grism 18 wavelength channels, after normalising by fitting a line ⇥ with wavelength. Each wavelength channel in the 1D spec- through orbits 2, 4 and 5, and dividing the light curves by of the atmosphere of smaller planets. tra is then used to construct a time-series, after binning in this function. The light curves are clearly seen to exhibit 5 pixels along the dispersion direction, resulting in 18 light strong time-correlated noise, which needs to be removed if

c 2002 RAS, MNRAS 000,1–17

Page 11 The first complete transmission spectrum of an exoplanet HD 189 733 b

Data: Sing et al. (2012) Pont et al (2008) Sing et al. (2009) Huitson et al. (2012) Gibson et al. (2012b) Sing et al. (2010)

] Ehrenreich et al. (2010)

Jupiter Radius [R Radius

300 nm 500 nm 1 μ 2 μ 8 μ Wavelength The first complete transmission spectrum of an exoplanet HD 189 733 b

UV visible near-infrared mid-infrared

Data: Sing et al. (2012) Pont et al (2008) Sing et al. (2009) Huitson et al. (2012) Gibson et al. (2012b) Sing et al. (2010)

] Ehrenreich et al. (2010)

Jupiter Radius [R Radius

300 nm 500 nm 1 μ 2 μ 8 μ Wavelength The first complete transmission spectrum of an exoplanet HD 189 733 b

UV visible near-infrared mid-infrared

Data: Sing et al. (2012) Pont et al (2008) sodium & Sing et al. (2009) potassium Huitson et al. (2012) Gibson et al. (2012b) Sing et al. (2010)

] Ehrenreich et al. (2010)

Jupiter Radius [R Radius

300 nm 500 nm 1 μ 2 μ 8 μ Wavelength The first complete transmission spectrum of an exoplanet HD 189 733 b

UV visible near-infrared mid-infrared

Data: Sing et al. (2012) Pont et al (2008) haze sodium & Sing et al. (2009) potassium Huitson et al. (2012) Gibson et al. (2012b) Sing et al. (2010)

] Ehrenreich et al. (2010)

Jupiter Radius [R Radius

300 nm 500 nm 1 μ 2 μ 8 μ Wavelength Atmospheric haze

to observer

star light

haze

Titan Find out more about exoplanet atmospheres Future perspectives

• ongoing

• Kepler: measure incidence of habitable planets (and many other kinds)

• Hubble space telescope: characterise atmospheres of handful of hot Jupiters

• very soon

• Direct imaging on large ground-based telescopes: young solar systems

• next decade

• TESS/PLATO: find transiting planets around bright stars

• EChO/JWST: characterise atmospheres of hot Jupiters, Neptunes and Super-Earths

• E-ELT: direct detection and atmospheric characterisation of smaller, cooler planets Signs of life? Signs of life?

Look for life in atmospheres out of chemical and thermodynamical equilibrium. Alpha Centauri Bb...(?)

Dumusque et al. (2012) © Richard Long