Source of Acquisition NASA Goddard Space Flight Center 2 The atmospheres of extrasolar planets 2.1 Introduction In this chapter we examine what can be learned about extrasolar planet atmospheres by concentrating on a class of planets that transit their parent stars. As discussed in the previous chapter, one way of detecting an extraso- lar planet is by observing the drop in stellar intensity as the planet passes in front of the star. A transit represents a special case in which the geometry of the planetary system is such that the planet’s orbit is nearly edge-on as seen from Earth. As we will explore, the transiting planets provide opportunities for detailed follow-up observations that allow physical characterization of extrasolar planets, probing their bulk compositions and atmospheres. 2.2 The Primary Eclipse The vast majority of the currently-known extrasolar planets have been de- tected using the radial velocity technique.t As detailed in the previous chapter, the radial velocity method searches for periodic motion of a star caused by the gravitational pull of an orbiting companion. Figure 1.1 shows a sketch of a typical periodic radial velocity signal and the basic geometry of the planetary system. This method is sensitive only to movement of the star towards and away from the observer, that is, along the line of sight from the system to the observer on Earth. Thus, radial velocity observations provide only a determination of the minimum mass M of the planet, and the orbital inclination i of the system remains unknown, as in h4 = M,sini, (2.1) where Mp is the tme mass of the planet. (See Section 1.2 for further details.) t An up-to-date reference and catalog of all known extrasolar planets can be found at http://vo.obspm.fr/exoplaoetes/encyclo/encycl.html 1 Fig. 2.1. Sketch showing a planet crossing the disk of its parent star. aansit light curve shown below. The primary eclipse, or transit, occurs when the planet's orbit happens to be nearly edge-on as seen from Earth. This means that the planet pe- riodically crosses in front of the star as it orbits, and we detect this as a decrease in the light from the star that occurs once per planet revolution, as indicated schematically in Figure 2.1. The dimming is typically a few percent or less for the currently known transiting planets. In this geometry, the orbital inclination is now known to be 90" (and can be determined N precisely from the details of the transit light curve). We can therefore derive the true planetary mass, Mp, from Equation 2.1. A number of other physical parameters of the planet and star can be derived from the shape of the light curve (Seager and Mall&-Ornelas, 2003). The depth of the transit AF (i.e,, the change in flux from outside transit to during transit, as shown in Figure 2.1) is directly proportional to the ratio of the area of the planetary disk to the area of the stellar disk. That is, where F represents the total flux, A is the area of the disk (planet or star), and R is the radius (planet or star). With a stellar mass-radius relation, it is possible to derive both the planetary and stellar radii simultaneously. With the planetary mass Mp and radius Rp, one can immediately calculate 2.3 The Secondary Eclipse 3 the average density of the planet from The discovery of transiting planets allowed a direct measurement of the true mass, radius, and density of planets outside the solar system for the first time. The planetary radius is key to determining the reflection and thermal emission of the planets from flux measurements. The density measurements derived from transit observations indicate that all but one of the transit- ing planets are hydrogen-helium gas giants, similar in bulk composition to Jupiter and Saturn in our own Solar System. When the planet is in front of the star, the planet’s atmosphere appears as an annulus surrounding the planetary disk, and some of the starlight passes through this annulus to the observer. The detection of starlight that has passed through the transiting planet’s atmosphere in this manner is called transmission spectroscopy. By measuring how much starlight is transmitted as a function of wavelength, we can learn about the atomic and molecular species present in the planet’s atmosphere, providing a much greater wealth of information than simply the average density and bulk composition. We introduce the broad study of spectroscopy in Section 2.5 and discuss re- cent observations of transiting planets using transmission spectroscopy in Section 2.7. 2.3 The Secondary Eclipse A planet that crosses in front of its parent star will disappear behind the star later in its orbit. This disappearance is called the secondary eclipse. For a circular orbit, the secondary eclipse occurs exactly one-half of an or- bital period after the primary eclipse. However, for a non-circular orbit, the secondary eclipse can occur earlier or later (depending on the eccentricity and the orientation of the orbit), and its duration can differ from that of the primary eclipse (Charbonneau, 2003). In addition to clues about the eccen- tricity of the planet’s orbit from the secondary eclipse timing and duration, the secondary eclipse yields information about the nature of the planet’s atmosphere. For example, in visible light, the secondary eclipse probes the amount of starlight reflected by the planet’s atmosphere (called the albedo). In the infrared, however, it measures the direct thermal emission (or intrinsic heat output) of the planet. In neither case does this imply imaging the planet; 4 The atmospheres of extrasolar planets 1. OM) .s E 0.995 -e! 3 ;L 0.990 U - 0.985 - - - - 0.5micron 20 microns .I -0.6 -0.4 -0.2 0.0 0.2 Orbital Phase Fig. 2.2. Comparison of primary and secondary eclipses in the visible and infrared for the thermal emission of the planet HD 209458 b. These curves were calculated from a simple model that asumes the star and planet emit blackbody radiation only. + rather, the idea is to observe the total energy output of the system (star planet) and attempt to detect a decrease as the planet is hidden from view. Figure 2.2 illustrates this decrease in the total energy output of the system during secondary eclipse and shows that the thermal emission of the planet may be detectable at infrared wavelengths using this technique. The basic situation is that the incident starlight (which peaks in the visible for a Sun- like star) is absorbed and reprocessed by the planet’s atmosphere, and some of that radiation is later emitted at infrared wavelengths. The figure shows the thermal emission of the planet HD 209458 b relative to its parent star. This calculation assumes that both the star and planet emit only blackbody radiation (Equation 2.6), and it assumes that the planet emits uniformly in both hemispheres. In the visible region (solid curve), the secondary eclipse is undetectable, both because the planet has virtually no emission at these wavelengths and because the reflected light from the planet is << 0.01% of the stellar output. However, as the figure shows, the situation is quite different at 20 pm. The total intensity relative to the star is higher outside of the eclipse, because the planet has a small but measurable intrinsic energy output at this wavelength. The secondary eclipse appears as a dip of ~ 0 . 3 % in the total intensity as the planet is hidden by the star. The eclipse depths at visible and infrared wavelengths can be estimated 5 4.5 ' - ' " ' ' ' " ' ' ' ' . ' 4.0 x o - r '1116 h 3.5 209458b. 3.0 149O26r +=-THf;ES-l Very Hot Jupiters 2.5 a" 2.0 Hot Jupiters 189733r +--+--i 1.5 I 1136, : 1.0' , . , , , ,-.'0.35 ' 56' b ' * ;e*' 0.6' ' -- Fig. 2.3. The ten currently-known transiting planets, as a function of planetary mass. The upper panel shows the period vs. mass, and the lower panel shows radius vs. mass. Dashed curves indicate curves of constant density in g/cm3. For comparison, Jupiter and Saturn are shown, marked by 'J' and 'S',r espectively. with the following flux ratios. For reflected light, where A, is the geometric albedo (the fraction of incident radiation scattered back into space when the planet is in full phase), Rp is the planetary radius, and a is the orbital semi-major axis. For thermal emission, where Tp and T, are the planet and star effective temperatures (see equa- tion 2.19 for an estimate of Tp).H ere we have used the approximation for the Wien tail of the blackbody flux whereby the flux ratio translates into a temperature ratio. 6 The atmospheres of extrasolar planets Table 2.1. Physical properties of transiting extrasolar planets. Planet Mas T* Teq (MJ) (K) (.K.) OGLETR-56b2t3 1.212 1.25 f 0.08 1.24 f 0.13 6119 1929 OGLE-TR-113b2i4 1.432 1.09 42 0.10 1.08 f 0.28 4804 1234 OGLETR-132b5>6 1.690 1.13 f 0.08 1.19 rt 0.13 6411 1933 HD 189733b 7y8 2.219 1.154 f 0.032 1.15 rt 0.04 5050 1096 HD 149026 bg>l0 2.876 0.726 ? :!!:! 0.33 f 0.023 6147 1593 nES-1 2,11,12 3.030 1.08 f. 0.05 0.729 f 0.036 5226 1059 OGLETR-10b2~13~14 3.101 1.14 f 0.09 0.63 f 0.14 6075 1402 HD 209458 b2715*16 3.525 1.320 f 0.025 0.657 f 0.006 6117 1363 XO-ib17 3.942 1.30 f 0.11 0.90 rt 0.07 5750 1148 OGLE-TR-11 lb2>18 4.016 0.97 f 0.06 0.52 =t0 .13 5044 935 ‘Calculated from Equation 2.19 with f = 1 and AB = 0.3; (Santos et al., 2006); ’‘ (Torres et al., 2004); (Konacki et al., 2004); (Bouchy et al., 2004); (Moutou et al., 2004); (Bouchy et al., 2005); (Bakos et al., 2006); (Sato et al., 2005); lo Charbonneau et al., 2005); l1 (Laughlin et al., 2005); l2 (Alonso et al., 2004); (Konacki et al., 2005); l4 (Holman et al., 2005); l5 (Knutson et al., 2006); (Winn et al., 2005); l7 (McCullough et al., 2006); l8 (Pont et al., 2004); 2.4 Characteristics of Known Transiting Planets A total of ten transiting extrasolar planets have been discovered as of May 2006. Their physical characteristics are given in Table 2.1, and they are plotted in Figure 2.3. The upper panel (period vs. mass) illustrates the two groups of transiting planets. The ‘hot Jupiters’ (to the upper left of the plot) have masses smaller than that of Jupiter and orbital periods greater than -2.5 days. This name is something of a misnomer, since the so-called hot Jupiters are quite different from our own Jupiter-because of the fact that they orbit at such small orbital distances, they are much hotter and therefore have different chemical species present in their atmospheres. The other group, often called the ‘very hot Jupiters,’ is characterized by plan- ets that orbit much closer to their parent stars (with orbital periods less than 2.5 days) and are more massive than Jupiter. These two dynamically distinct groups of planets may have different evolutionary histories, possi- bly resulting from different migration mechanisms (Gaudi et al., 2005), and thus could potentially have very different atmospheric properties. We now have one bright planet from each group-HD 209458 b and HD 198733 b- 2.5 Spectroscopy 7 ’ allowing us to compare observations of the two planets and gain insights into their atmospheric structure and evolutionary history. The lower panel of Figure 2.3 shows the radius of each planet vs. mass. The dashed curves indicating constant density provide context for under- standing the bulk composition of the planets. For example, most of the transiting planets are similar in average density to Jupiter (p = 1.33 g/cm2) and Saturn (p = 0.69 g/cm2). However, these ‘close-in’ extrasolar planets are likely to be quite different from our own Jupiter, due to the fact that they are much closer in to their stars. At orbital distances of a < 0.1 AU (by comparison, Mercury is at a 0.38 AU), these planets are bombarded N by radiation from their parent stars and are therefore expected to be hot (2’ > 1000 K). Jupiter, at 5 AU from the Sun, has a blackbody temperature of only 110 K. Because of the large temperature difference, we expect the at- mospheric composition of the hot Jupiters to be significantly different from that of Jupiter. For example, at low temperatures (2’ < 1000 K), chemical equilibrium calculations show that carbon is mostly present in the form of CH4, while at higher temperatures it appears as CO (Burrows and Sharp, 1999). Finally, from Figure 2.3, we note that not all of the known transiting planets have densities similar to Jupiter and Saturn. The most extreme example is HD 209458 b, with an average density of -0.35 g/cm2. This planet was the first one found to exhibit a transit (Charbonneau et al., 2000; Henry et al., 2000), although it was originally detected using the radial velocity method (Mazeh et al., 2000). Since then, it has been observed extensively from the ground and from space, as we shall discuss further in Section 2.7. From the beginning, it was unclear why the planet appears to have a larger radius than is predicted by theory (Burrows et al., 2000; Burrows et al., 2003), and this remains one of the unanswered questions in the field today (Winn and Holman, 2005). 2.5 Spectroscopy We now turn from a general discussion of the transiting planets to the spe- cific topic of spectroscopy and the radiative transfer equation. By studying the spectroscopy of extrasolar planets, we can gain key insights into the atmospheric composition, temperature, and structure of these planets. We begin this section by introducing the Planck blackbody law, which describes the thermal emission of an object in the absence of scattering or absorbing particles, and move to the radiative transfer equation, which does account for the effects of scattering and absorption. The radiative transfer equation 8 The atmospheres of extrasolar planets governs the interaction of energy (in the form of emitted or absorbed radi- ation) with matter (in this case the particles that make up the planetary atmosphere). In Section 2.6 we give an overview of how models of planetary atmospheres are computed. Spectra derived from such models help interpret observational results and facilitate the design of future planet atmosphere detection instruments. At the most basic level we can approximate the star and planet as black- bodies. In that case, where we ignore the details of the atmosphere, the emission is given by Planck’s blackbody law: where X is wavelength, c is the speed of light in vacuum, IC is Boltzmann’s constant, T is the temperature of the blackbody, and h is Planck’s constant. For a stellar or planetary atmosphere which contains a variety of different species that absorb, emit, and scatter radiation, however, the blackbody law is not sufficient to describe the resulting spectrum, and it becomes necessary to understand how matter interacts with the radiation. Following convention, we begin by considering a pencil of radiation trav- eling through a medium. The energy in the beam is given by dE, = I, cos 8 dAdQ dud t, (2-7) where u is wavenumber (l/X), I, is the monochromatic (spectral) radiance (sometimes called intensity), 8 is the angle from the normal to the surface, dA is a differential area element intercepting the beam, s2 is the solid angle in steradians, and t is time. Next, we explicitly describe how the radiation beam changes as it interacts with matter [of density p), traveling through a distance ds: dl, = -k,pI,ds+ j,pds. (2.8) The first term on the right hand side represents the amount of radiation removed from the beam (extinction cross section A,) and the second term represents the amount of radiation added to the beam (emission cross section j,). Defining the source function S, as the ratio of the emission cross section to the extinction cross section, we have dI, + s,. ---- - -I, P k, ds This is the radiative transfer equation and it governs the fundamental physics at work in the atmosphere. The simplistic form of the radiative transfer 2.5 Spectroscopy 9 equation hides its true complexity. The main problem lies in the nonlinearity of the equation. The solution of I, depends on j,, but if there is scattering in the atmosphere j, also depends on I,. A second problem lies in the definition of the source function S, = j,/k,. The opacities that make up kv and j, can be composed of millions of lines for molecular species, and in the m e o f cloud opacities can involve a number of free parameters. Finally, we can write Equation 2.9 in a more conventional form by making a few more definitions. If we consider a plane-parallel atmosphere, we are interested only in radiation flowing in the vertical direction z. We can define e, = cos (2.10) where 0 is again the angle measured from the vertical, or the zenith angle. The distance ds can be projected along the vertical axis, as in ds = cosedz = pdz. (2.11) We can now define the opticaE depth r as The minus sign appears because the optical depth is by convention measured from the top of the atmosphere increasing downward. Using the definition of the optical depth, we can rewrite Equation 2.9 as (2.13) The detailed solution of this equation is beyond the scope of this work, and we refer the interested reader to more comprehensive works that describe the solution and application of the radiative transfer equation (e.g. Mihalas, 1970; Liou, 2002; Salby, 1996). Fortunately, under specific assumptions, the solution to Equation 2.13 becomes simple. As discussed in Section 2.2, during transit the planet passes in front of the star, and some starlight passes through the annulus of the planetary atmosphere before reaching the observer. At visible wavelengths (where the thermal emission is negligible) the starlight is attenuated by the absorbing gases in the planet atmosphere. In this case, we take the emission, and thus the source function S,, to be zero, and Equation 2.13 reduces to dLJ P- = I,, (2.14) dT which can easily be integrated to obtain I, = Iv(o)e-T~’~. (2.15) The atmospheres of extrasolar planets 10. This equation is known as Beer's Lay or Lambert's Law (Liou, 2002). It describes the dissipation of radiation as it travels through a medium. Be- cause atoms and molecules absorb at specific wavelengths, the amount of starlight that is trasmitted through the planetary atmosphere changes with wavelength. Another physical situation with a simple solution to the radiative transfer equation is the case of thermal emission and no scattering. This situation would hold at infrared wavelengths if clbuds &e. scattering particles) were not present. In this case of thermal emission, the source function is simply the blackbody function: S, = B, (2.16) The radiative transfer equation (2.13) then reduces to a linear form: (2.17) The solution is (2.18) With a given vertical temperature and pressure profile, the opacities and hence B, can be computed, and the right hand side of the above equation is straightforward to integrate. 2.6 Model Atmospheres A full model atmosphere computation is needed to understand the details of the planetary spectrum. Usually the models assume that the planetary atmosphere is one-dimensional and plane-parallel (no curvature). The mod- els produce the temperature and pressure as a function of altitude and the radiation field (that is, the emergent flux from the atmosphere) as a function of altitude and wavelength (see Seager et al. (2005) and references therein). To derive these three quantities, three equations are solved: the radiative transfer equation (Equation 2.13), the equation of hydrostatic equilibrium, and the radiative and convective equilibrium. The boundary conditions are the incident stellar radiation at the top of the atmosphere and the inte- rior energy (assumed) at the bottom of the atmosphere. With this type of calculation, only the planetary surface gravity and the incident stellar radi- ation are known with certainty. Although the physics governing the model is relatively simple, a number of assumptions are necessary in order for the calculation to proceed, including (Seager et al., 2005; Marley et al., 2006):