three-dimensional atmospheric circulation of hot …showman/publications/kataria-etal-2013.pdf ·...

19
The Astrophysical Journal, 767:76 (19pp), 2013 April 10 doi:10.1088/0004-637X/767/1/76 C 2013. The American Astronomical Society. All rights reserved. Printed in the U.S.A. THREE-DIMENSIONAL ATMOSPHERIC CIRCULATION OFHOT JUPITERS ON HIGHLY ECCENTRIC ORBITS T. Kataria 1 , A. P. Showman 1 , N. K. Lewis 1 ,2,6 , J. J. Fortney 3 , M. S. Marley 4 , and R. S. Freedman 4 ,5 1 Department of Planetary Sciences and Lunar and Planetary Laboratory, The University of Arizona, Tucson, AZ 85721, USA; [email protected] 2 Department of Earth, Atmospheric and Planetary Sciences, Massachusetts Institute of Technology, Cambridge, MA 02139, USA 3 Department of Astronomy & Astrophysics, University of California, Santa Cruz, CA 95064, USA 4 NASA Ames Research Center 245-3, Moffett Field, CA 94035, USA 5 SETI Institute, 189 Bernardo Ave. 100, Mountain View, CA 94043, USA Received 2012 August 14; accepted 2013 February 28; published 2013 March 26 ABSTRACT Of the over 800 exoplanets detected to date, over half are on non-circular orbits, with eccentricities as high as 0.93. Such orbits lead to time-variable stellar heating, which has major implications for the planet’s atmospheric dynamical regime. However, little is known about the fundamental dynamical regime of such planetary atmospheres, and how it may influence the observations of these planets. Therefore, we present a systematic study of hot Jupiters on highly eccentric orbits using the SPARC/MITgcm, a model which couples a three-dimensional general circulation model (the MITgcm) with a plane-parallel, two-stream, non-gray radiative transfer model. In our study, we vary the eccentricity and orbit-average stellar flux over a wide range. We demonstrate that the eccentric hot Jupiter regime is qualitatively similar to that of planets on circular orbits; the planets possess a superrotating equatorial jet and exhibit large day–night temperature variations. As in Showman & Polvani, we show that the day–night heating variations induce momentum fluxes equatorward to maintain the superrotating jet throughout its orbit. We find that as the eccentricity and/or stellar flux is increased (corresponding to shorter orbital periods), the superrotating jet strengthens and narrows, due to a smaller Rossby deformation radius. For a select number of model integrations, we generate full-orbit light curves and find that the timing of transit and secondary eclipse viewed from Earth with respect to periapse and apoapse can greatly affect what we see in infrared (IR) light curves; the peak in IR flux can lead or lag secondary eclipse depending on the geometry. For those planets that have large temperature differences from dayside to nightside and rapid rotation rates, we find that the light curves can exhibit “ringing” as the planet’s hottest region rotates in and out of view from Earth. These results can be used to explain future observations of eccentric transiting exoplanets. Key words: atmospheric effects – methods: numerical – planets and satellites: atmospheres – planets and satellites: general Online-only material: color figures 1. INTRODUCTION Since the first planetary confirmations in the mid-1990s (Mayor & Queloz 1995; Wolszczan 1994) the detection and characterization of extrasolar planets continue to be major fields in astronomy and planetary science. Over 700 planets have been detected from the ground and space, more than half of which are classified as “hot Jupiters,” Jovian-mass planets that orbit their parent stars at distances less than 0.1 AU. A number of these hot Jupiters transit their host star along our line of sight, allowing us to observe them as they pass in front of and behind their parent star (e.g., Charbonneau et al. 2008; Swain et al. 2008; Pont et al. 2008; Knutson et al. 2007, 2008, 2009). Using these observations, we can infer much about these planets’ atmospheric composition, temperature structure, and circulation. A fifth of these transiting exoplanets have eccentricities greater than 0.1, with values as large as 0.93 (HD80606b; Naef et al. 2001). These eccentric planets are subject to highly time-variable heating which has a significant effect on the planet’s atmospheric dynamics. Among planets amenable to observational follow-up, HAT-P-2b, which has an eccentricity of 0.52, undergoes a factor of nine variation in flux throughout 6 Sagan Postdoctoral Fellow. its orbit. HD 17156b (e = 0.67) experiences a factor of 27 variation in stellar flux, and HD 80606b, with its large eccentricity (e = 0.93), undergoes an impressive factor of 828 variation in flux. Because they are transiting, we can probe their atmospheres as we can for planets on circular orbits. However, while it is clear the strongly variable heating leads to a vastly different forcing regime than for exoplanets on circular orbits, it remains unknown whether this causes a fundamentally different dynamical regime: is the circulation quantitatively similar to that on circular hot Jupiters, or is it a completely new circulation regime? Eccentric transiting exoplanets present unique challenges when one attempts to extract information about their atmo- spheres from observational data. In particular, interpretation of flux maxima and minima in infrared light curves can be com- plicated by the convolution of spatial effects (for example, hot spots on the planet that rotate into and out of view along our line of sight) with temporal effects (planet getting colder/warmer at apoapse/periapse passage). Langton & Laughlin (2008a, 2008b) and Cowan & Agol (2011) address this problem as applied to particular targets, but use only a two-dimensional hydrodynam- ical model and one-dimensional semi-analytic model, respec- tively. To fully capture these effects, a three-dimensional circu- lation model that self-consistently calculates heating and wind velocities is needed. 1

Upload: others

Post on 28-Jun-2020

0 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: THREE-DIMENSIONAL ATMOSPHERIC CIRCULATION OF HOT …showman/publications/kataria-etal-2013.pdf · planets’ atmospheric composition, temperature structure, and circulation. A fifth

The Astrophysical Journal 76776 (19pp) 2013 April 10 doi1010880004-637X767176Ccopy 2013 The American Astronomical Society All rights reserved Printed in the USA

THREE-DIMENSIONAL ATMOSPHERIC CIRCULATION OF HOT JUPITERSON HIGHLY ECCENTRIC ORBITS

T Kataria1 A P Showman1 N K Lewis126 J J Fortney3 M S Marley4 and R S Freedman451 Department of Planetary Sciences and Lunar and Planetary Laboratory The University of Arizona Tucson AZ 85721 USA tkatarialplarizonaedu

2 Department of Earth Atmospheric and Planetary Sciences Massachusetts Institute of Technology Cambridge MA 02139 USA3 Department of Astronomy amp Astrophysics University of California Santa Cruz CA 95064 USA

4 NASA Ames Research Center 245-3 Moffett Field CA 94035 USA5 SETI Institute 189 Bernardo Ave 100 Mountain View CA 94043 USA

Received 2012 August 14 accepted 2013 February 28 published 2013 March 26

ABSTRACT

Of the over 800 exoplanets detected to date over half are on non-circular orbits with eccentricities as high as 093Such orbits lead to time-variable stellar heating which has major implications for the planetrsquos atmospheric dynamicalregime However little is known about the fundamental dynamical regime of such planetary atmospheres and howit may influence the observations of these planets Therefore we present a systematic study of hot Jupiters onhighly eccentric orbits using the SPARCMITgcm a model which couples a three-dimensional general circulationmodel (the MITgcm) with a plane-parallel two-stream non-gray radiative transfer model In our study we vary theeccentricity and orbit-average stellar flux over a wide range We demonstrate that the eccentric hot Jupiter regimeis qualitatively similar to that of planets on circular orbits the planets possess a superrotating equatorial jet andexhibit large dayndashnight temperature variations As in Showman amp Polvani we show that the dayndashnight heatingvariations induce momentum fluxes equatorward to maintain the superrotating jet throughout its orbit We find thatas the eccentricity andor stellar flux is increased (corresponding to shorter orbital periods) the superrotating jetstrengthens and narrows due to a smaller Rossby deformation radius For a select number of model integrationswe generate full-orbit light curves and find that the timing of transit and secondary eclipse viewed from Earth withrespect to periapse and apoapse can greatly affect what we see in infrared (IR) light curves the peak in IR flux canlead or lag secondary eclipse depending on the geometry For those planets that have large temperature differencesfrom dayside to nightside and rapid rotation rates we find that the light curves can exhibit ldquoringingrdquo as the planetrsquoshottest region rotates in and out of view from Earth These results can be used to explain future observations ofeccentric transiting exoplanets

Key words atmospheric effects ndash methods numerical ndash planets and satellites atmospheres ndash planets andsatellites general

Online-only material color figures

1 INTRODUCTION

Since the first planetary confirmations in the mid-1990s(Mayor amp Queloz 1995 Wolszczan 1994) the detection andcharacterization of extrasolar planets continue to be major fieldsin astronomy and planetary science Over 700 planets have beendetected from the ground and space more than half of whichare classified as ldquohot Jupitersrdquo Jovian-mass planets that orbittheir parent stars at distances less than 01 AU A numberof these hot Jupiters transit their host star along our line ofsight allowing us to observe them as they pass in front ofand behind their parent star (eg Charbonneau et al 2008Swain et al 2008 Pont et al 2008 Knutson et al 2007 20082009) Using these observations we can infer much about theseplanetsrsquo atmospheric composition temperature structure andcirculation

A fifth of these transiting exoplanets have eccentricitiesgreater than 01 with values as large as 093 (HD80606bNaef et al 2001) These eccentric planets are subject to highlytime-variable heating which has a significant effect on theplanetrsquos atmospheric dynamics Among planets amenable toobservational follow-up HAT-P-2b which has an eccentricityof 052 undergoes a factor of nine variation in flux throughout

6 Sagan Postdoctoral Fellow

its orbit HD 17156b (e = 067) experiences a factor of27 variation in stellar flux and HD 80606b with its largeeccentricity (e = 093) undergoes an impressive factor of 828variation in flux Because they are transiting we can probe theiratmospheres as we can for planets on circular orbits Howeverwhile it is clear the strongly variable heating leads to a vastlydifferent forcing regime than for exoplanets on circular orbits itremains unknown whether this causes a fundamentally differentdynamical regime is the circulation quantitatively similar tothat on circular hot Jupiters or is it a completely new circulationregime

Eccentric transiting exoplanets present unique challengeswhen one attempts to extract information about their atmo-spheres from observational data In particular interpretation offlux maxima and minima in infrared light curves can be com-plicated by the convolution of spatial effects (for example hotspots on the planet that rotate into and out of view along our lineof sight) with temporal effects (planet getting colderwarmer atapoapseperiapse passage) Langton amp Laughlin (2008a 2008b)and Cowan amp Agol (2011) address this problem as applied toparticular targets but use only a two-dimensional hydrodynam-ical model and one-dimensional semi-analytic model respec-tively To fully capture these effects a three-dimensional circu-lation model that self-consistently calculates heating and windvelocities is needed

1

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

Hence it is crucial to conduct a comprehensive study thatestablishes the dynamical regime temperature structure andobservational implications of eccentric exoplanets We use athree-dimensional atmospheric circulation model coupled toa non-gray radiative transfer scheme to study eccentric hotJupiters as a whole In Section 2 we will describe our modelsetup and integrations Section 3 describes the dynamicalregime and its dependence on eccentricity and mean stellarflux Section 4 presents synthetic light curves and attempts todetermine what can be learned about atmospheric circulationsfrom remote measurements Finally Section 5 concludes andcompares results to known eccentric hot Jupiters

2 MODEL

We adopt the Substellar and Planetary Atmospheric Radiationand Circulation (SPARC) model which couples the MITgcm(Adcroft et al 2004) with a two-stream implementation of themulti-stream non-gray radiative transfer scheme developed byMarley amp McKay (1999) A brief discussion of the model isincluded here for a more complete description see Showmanet al (2009)

The MITgcm solves the primitive equations a simplificationof the NavierndashStokes equations where the horizontal flow lengthscale exceeds vertical length scales these equations are valid atlarge scales in stably stratified atmospheres Hot Jupiters gener-ally satisfy these criteria with horizontal and vertical lengthscales of sim104 km and sim300 km respectively and highlystratified atmospheres While the primitive equations do havelimitations (for example the inability to capture small-scaleweakly stratified features such as fronts and storms) thelarge-scale flow (eg jets and waves) is captured The use ofthe primitive equations also presents many computational ad-vantages over the fully compressible equations The radiativetransfer code solves the two-stream radiative transfer equationsand employs the correlated-k method (eg Mlawer et al 1997Goody et al 1989 Fu amp Liou 1992 Marley amp McKay 1999) tosolve for upwarddownward fluxes and heatingcooling ratesthrough an inhomogeneous atmosphere This method retainsmost of the accuracy of full line-by-line calculations whiledrastically increasing computational efficiency The correlated-kmethod maps the absorption coefficients denoted by k(ν) whereν is frequency from spectral space (k = k(ν)) to a spacewhere k varies with a variable g (k = k(g)) The inversionof this function g(k) is commonly referred to as the cumulativedistribution

gi =int k

0fi(k

prime) dkprime (1)

where fi(k) is the distribution function for the absorptioncoefficient in the ith atmosphere layer in a spectral interval νIn the space of cumulative probability g varies from 0 to 1 andrepresents the probability of the absorption coefficient beingless than or equal to k(g) The total frequency space is splitinto a number of discrete frequency intervals (windows) and theabsorption coefficients at each grid point in each window aresorted by frequency of occurrence From this a g distribution isderived for each window The radiative transfer is then carriedout window by window usually using a g(k) that has beenrepresented by a Gaussian division scheme in order to integratethe distribution over each window

For each window n different monochromatic absorptioncoefficients Ki (where i varies from 1 to n) are computed atn gauss points in order to integrate over the k(g) curve These

values of Ki are used to calculate the transmission T as afunction of column mass u by the equation

T (u) =nsumi

WieminusKiu (2)

where Wi is the gauss weight for each fractional gauss pointIn this way multiple opacity sources can be modeled Thismodel accounts for gaseous (Rayleigh) scattering and pressure-induced absorption by H2 and He Scattering and absorptiondue to clouds aerosols and dust can also be included thoughfor simplicity we do not include these effects here The fluxeswithin each spectral interval are weighted and summed toobtain the upward and downward fluxes for each bin at eachatmospheric layer The sum of all intervals gives the totalwavelength-integrated fluxes for each layer We then calculatethe heating rate by finite-differencing the fluxes and pressuresbetween interfaces over and underlying a given dynamicallevel

q = gpartF

partp (3)

Here q is the heating rate g is the planetrsquos gravity and partFpartpis the gradient of flux with pressure For each simulation weutilize a cubed-sphere grid with a horizontal resolution of C32(approximately equivalent to 64 times 128 in latitude and longitude)and NL = 40 or 76 pressure levels The lowermost NL minus 1levels extend from a mean pressure of 200 bars at the bottom to02 mbar at the top evenly spaced in log pressure The top levelextends from a pressure of 02 mbar to zero

We implement a fourth-order Shapiro filter for temperatureand momentum with a timestep equal to double the dynam-ical timestep Models integrated with this Shapiro filter setup(but excluding any other large-scale drag) conserve total an-gular momentum to better than 01 thereby demonstratingexcellent conservation of angular momentum with our modelconfiguration

21 Updates to SPARCMITgcm

211 Reduced Number of Frequency Bins

While the models of HD 189733b and HD 209458b inShowman et al (2009) and the models of GJ 436b in Lewis et al(2010) used 30 frequency bins to model circulation and heatingon each planet here the opacities are statistically weighted into11 frequency bins to improve computational efficiency For boththe 30- and 11-bin implementations the radiative transfer iscalculated at eight values of k(g) in each of the frequency binsFour of the values of k(g) are sampled at g lt 095 meaning thatfour calculations are done for the weakest 95 of the spectrallines The other four radiative transfer calculations are done forthe strongest 5 of the lines (g gt 095) This allows us tocalculate the radiative transfer for both strong and weak spectrallines

We list the 11 new frequency bins in Table 1 Before imple-menting the updated scheme into our simulations we conductedone-dimensional and three-dimensional tests to reproduce theresults for HD 1897833b from Showman et al (2009) Oncesuccessful we applied the new scheme to the model integra-tions described here (The Appendix describes these validationtests in detail)

2

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

Table 1Bounding Wavelengths of the Frequency Bins Used

in the Radiative Transfer Calculations

Wavelength Wavelength(μm) (μm)

32468 20002000 870870 440440 350350 250250 202202 132132 085085 061061 042042 026

212 Updated Collision-induced Absorption dueto H2ndashH2 and H2ndashHe Collisions

The collision-induced absorption (CIA) by H2 moleculesdue to H2ndashH2 and H2ndashHe collisions has recently been recom-puted from first principles (Saumon et al 2012 and referencestherein) The new formulation uses state-of-the-art quantum me-chanical calculations of the potential energy surface and of theinteraction-induced dipole moment surface for colliding H2ndashH2and H2ndashHe pairs This new calculation is more reliable at highertemperature and higher photon energies than the earlier cal-culations by Borysow and collaborators that radiative transfergroups have been using (httpwwwastrokudksimaborysow)We have included these new tabulations of H2 CIA opacity inour code

22 Grid Simulations

Table 2 lists our model integrations for this study For allmodels we assume the radius and gravity of a generic hotJupiter7 Figure 1 shows the parameter space plotted as afunction of semi-major axis and eccentricity along with detectedexoplanets with semi-major axes less than 05 AU In these

7 HD 189733b RP = 82396 times 107 m g = 214 m sminus2 Note that we couldhave easily chosen any other well-characterized hot Jupiter such as HD209458b However the primary goal of this study is to illustrate the dynamicaland observational effects of eccentricity and average stellar flux The preciseplanetary radius and gravity are of secondary importance

0 01 02 03 04 050

01

02

03

04

05

06

07

08

09

1

semiminusmajor axis

ecce

ntr

icit

y

detected exoplanet

468183 W mminus2

185691 W mminus2

73680 W mminus2

11617 W mminus2

Figure 1 Model integrations from this study plotted against a portion of thecurrent exoplanet population (dots) The green right-pointing triangles reddown-pointing triangles blue left-pointing triangles and magenta up-pointingtriangles correspond to average stellar fluxes of 468183 185691 73680 and11617 W mminus2 respectively See Table 2 and text for more details

(A color version of this figure is available in the online journal)

simulations we systematically vary both the average stellarflux and eccentricity Orbital eccentricities of 00ndash075 andorbit-averaged stellar fluxes 〈F 〉 of 468183 185691 73680and 11617 W mminus2 were explored These values correspond toequilibrium temperatures (Teq) of approximately 1199 951 755and 476 K respectively For these given values of e and 〈F 〉the orbital semi-major axis a is calculated using the equation

〈F 〉 = L

4πa2(1 minus e2)12 (4)

where L is the luminosity of HD 189733 These chosen simula-tions fall well within the current exoplanet population

We assume the atmospheric composition to be 1times solarmetallicity without TiO and VO this composition has hadsuccess in explaining observations of HD 189733b As inShowman et al (2009) opacities are determined assuming localchemical equilibrium (accounting for rainout of condensates)

Table 2List of Model Integrations

〈F 〉 (W mminus2) Teq a e rp ra Porb Prot PorbProt Ω (sminus1)(K) (AU) (AU) (AU) (s) (s)

468183 1199 00313 000 00313 00313 1917 times 105 1917 times 105 10 327 times 10minus5

468183 1199 00318 025 00239 00398 1963 times 105 1421 times 105 14 442 times 10minus5

468183 1199 00385 075 000963 00674 2615 times 105 3010 times 104 87 209 times 10minus4

185691 951 00497 000 00497 00497 3834 times 105 3834 times 105 10 164 times 10minus5

185691 951 00505 025 00379 00631 3927 times 105 2843 times 105 14 221 times 10minus5

185691 951 00534 050 00267 00801 4270 times 105 1522 times 105 28 413 times 10minus5

185691 951 00611 075 00153 01069 5226 times 105 6016 times 104 87 104 times 10minus4

73680 755 00789 000 00789 00789 7668 times 105 7668 times 105 10 819 times 10minus6

73680 755 00802 025 00602 01003 7858 times 105 5689 times 105 14 110 times 10minus5

73680 755 00848 050 00424 01272 8544 times 105 3046 times 105 28 206 times 10minus5

11617 476 01987 000 01987 01987 3067 times 106 3067 times 106 10 205 times 10minus5

11617 476 02019 025 01514 02524 3141 times 106 2274 times 106 14 276 times 10minus5

11617 476 02135 050 01068 03202 3416 times 106 1218 times 106 28 516 times 10minus6

11617 476 02443 075 00611 04275 4181 times 106 4813 times 105 87 131 times 10minus5

3

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

0 500 1000 1500 2000 2500

10minus6

10minus4

10minus2

100

102

104

Temperature (K)

Pre

ssu

re (

bar

s)

11617 W mminus2

73680 W mminus2

185691 W mminus2

468183 W mminus2

Figure 2 Initial 1-D pressurendashtemperature profiles assigned to each column ofthe atmosphere at the start of each simulation at the four different valuesof average stellar flux These profiles were calculated assuming radiativeequilibrium and use the same opacities as the full 3D model

(A color version of this figure is available in the online journal)

at each temperature and pressure over the 3D grid assumingthe elemental abundances of Lodders (2003) We set cp and κto values appropriate for a H2-dominated atmosphere (13 times104 J kgminus1 Kminus1 and 27 respectively)

For eccentric cases we assume the planet is pseudo-synchronously rotating its host starmdashwe assume the planetrsquostidal interactions with the star force the same side of the planetto approximately face the star every periapse passage We cal-

culate this using the Hut (1981) formulation

Prot = Porb

(1 + 3e2 + 3

8e4)

(1 minus e2)32

1 + 152 e2 + 45

8 e4 + 516e6

(5)

Pseudo-synchronization is a commonly used assumption inthe calculation of the rotation rate of eccentric exoplanets (egDeming et al 2007 Irwin et al 2008 Langton amp Laughlin2008a 2008b Lewis et al 2010) Nevertheless the true depen-dence of rotation period on orbital period and eccentricity isvery uncertain (eg Greenberg 2009) and other formulationsbesides Equation (5) have also been suggested in the litera-ture (eg Ivanov amp Papaloizou 2004) We explore the sensitiv-ity of the flow to the assumption of pseudo-synchronization inSection 34

In each integration we assume the winds to be initiallyzero with each vertical atmospheric column assigned the sameglobal-mean radiative equilibrium pressurendashtemperature profilecalculated using the same opacities as the main model (Figure 2)These temperaturendashpressure profiles are calculated for eachvalue of average stellar flux (see above) using the methodsdescribed in Fortney et al (2005 2008)

Most model integrations were run until the winds reacheda statistically steady configuration at upper levels Figure 3shows the rms velocity plotted as a function of pressure andsimulated time for a model integrated for gt6000 Earth daysThis behavior in rms velocity is typical of most model inte-grations Weak drag was applied to the atmosphere with atime constant of 100 days at 200 bars with a drag coeffi-cient that decreases linearly with pressure to zero at 10 barsTests were conducted to explore the sensitivity of drag on themean flow Extending the drag up to 1 bar affects the verti-cal and longitudinal extent of the equatorial jet but not the

0 1000 2000 3000 4000 5000 6000

10minus3

10minus2

10minus1

100

101

102

simulated time (Earth days)

pre

ssu

re (

bar

)

e= 050 a= 00534 AU

00003217611

0000321

200

200 200

400

400 400

600

600 600

800

800 800

1000

1000 1000

1200

1200 1200

1400

1400 1400

1600

16001600

1800

18001800

2000

2000

2000

2200

2200

2200

2400

2400

2600

260028003000

(m sminus1)0

500

1000

1500

2000

2500

3000

Figure 3 rms velocity plotted as a function of pressure and simulation time for one of our model integrations The plot shows a case where Teq = 951 K e = 050and a = 00534 AU rms velocity was calculated at each pressure level from wind speeds output every 100 days The rapid decrease in velocity at the start of the runis merely an artifact at the start of the run

(A color version of this figure is available in the online journal)

4

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus180 minus90 0 90 180True anomaly

minus2 minus15 minus1 minus05 0 05 1 15 2

10minus3

10minus2

10minus1

100

101

102

simulated time (Earth days)

pre

ssu

re (

bar

) 700700

800800

800

900

900

900

1000

10001000

11001100

1100

12001200 1200

1300 1300 1300

1400 1400 1400

1500 1500 1500

T(K)

700

800

900

1000

1100

1200

1300

1400

1500

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00534 AU

minus200

minus200

200

200

600

600

1000

1000

1400

18002200

2600300034003800

(m sminus1)

0

1000

2000

3000

4000

5000

Figure 4 Dynamical response of a model with Teq = 951 K e = 050 anda = 00534 AU Top global horizontally averaged temperature vs pressureand time throughout the full orbit zero is defined as periapse passage The upperaxis gives the true anomaly throughout the orbit Bottom zonally averaged zonal(eastndashwest) winds vs pressure and latitude for the same model time-averagedover a full orbit

(A color version of this figure is available in the online journal)

qualitative nature of the circulation We reiterate that inte-grations without this drag scheme (but including our fourth-order Shapiro filter) conserve angular momentum to betterthan 01

3 RESULTS

31 Circulation Regime

To illustrate the circulation regime of an eccentric exoplanetwe present a nominal model where e = 050 a = 00534 AUand Teq = 951 K representative of the simulations (Figure 4)The top panel shows the horizontally averaged global temper-ature as a function of pressure time (in Earth days) and trueanomaly f where f = 0 at periapse and f = plusmn180 atapoapse Temperatures remain fairly constant at pressures ex-ceeding 1 bar However in the uppermost layers of the atmo-sphere temperatures vary throughout the orbit peaking sim6 hr

after periapse passage The bottom panel shows the zonal-meanzonal wind8 of the same model time-averaged over a full or-bit The flow exhibits a superrotating (eastward) equatorial jetwhich is the dominant feature of the flow with peak winds ofabout 5000 m sminus1 Superrotation is commonly seen in models ofhot Jupiters and hot Neptunes (eg Showman amp Guillot 2002Cooper amp Showman 2005 Showman et al 2008 2009 Dobbs-Dixon amp Lin 2008 Dobbs-Dixon et al 2010 Lewis et al 2010Heng et al 2011 Perna et al 2012 Rauscher amp Menou 20102012) although not by all groups (Cho et al 2008 Thrastarsonamp Cho 2010)

With peak winds exceeding 2ndash5 km sminus1 the equatorial jet issupersonic as the sound speed for this class of planets is ap-proximately 2 km sminus1 As Heng (2012) shows the presence ofthis supersonic flow generates shocks at the dayndashnight termi-nator upstream of the substellar point The SPARCMITgcmdoes not properly account for the removal of kinetic energy as aresult of these shocks The formation of shocks is expected to beshallow (ie at low pressure) in both of the model integrationsthat Heng (2012) presents the Mach number falls below unityat pressures greater than 1 bar and hence shocks are not formedNevertheless the infrared photosphere and the peak jet speedsin our model occur at sufficiently low pressure that shocks couldplay some role particularly when the orbital eccentricity is highMoreover as a possible energy dissipation mechanism shockscould have a non-local effect (ie they could cause an indirecteffect on the flow even at pressures deeper than those where theshocks directly occur) Clearly more work is warranted on therole of shocks in hot Jupiter atmospheres

Snapshots of the wind and temperature profile throughout theorbit at a pressure level of 30 mbar (approximately the levelof the infrared photosphere) show that the planet undergoes aflash heating event as the planet approaches periapse passage(Figure 5) The eight panels show snapshots throughout onecomplete orbit from apoapse (top figure) to hours after periapse(f = 119 follow arrows) and back again to apoapse all withthe same colorscale for temperature Like the circulation ofHD 189733b in its nominally circular orbit (see Showman et al2009) the flow maintains the eastward equatorial superrotatingjet throughout its orbit (Figure 6) As the planet approachesperiapse (Figure 5 f = minus12) temperatures rise to sim1000 Kat the equator and a high-amplitude eddy structure developswith a pattern of temperature and wind vectors that tilt eastwardtoward the equator This so-called ldquochevronrdquo-shaped feature isprominent at periapse (bottom panel) extends over a wide rangeof latitude and longitude and persists hours after periapse Thechevron shape is also seen in our other integrations and indicatesthat eddies transport momentum equatorward (see Section 36)

32 Effect of Eccentricity

Our models show that at constant orbit-mean stellar fluxthe dayside temperatures dayndashnight temperature differencesand wind speeds at periapse passage all increase with increas-ing orbital eccentricity This is illustrated in Figure 7 whichcompares three simulations with an equilibrium temperature of951 K (〈F 〉 = 185691 W mminus2) that vary in eccentricity from00 (top row) to 075 (bottom row) Again we plot each columnon the same colorscale (zonal wind and temperature) for com-parison As eccentricity increases peak temperatures increase

8 Zonal wind is defined as eastndashwest winds with positive (negative) valuesdenoting eastward (westward) winds a zonal average is an average inlongitude

5

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=000deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus1242deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=1242deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus7227deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=7227deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus11988deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=11988deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus18000deg

(K)500

600

700

800

900

1000

1100

1200

Figure 5 Snapshots of the wind and temperature structure at 30 mbar of a planet with Teq = 951 K e = 050 and a = 00534 AU To trace the behavior throughoutone complete orbit the different panels should be followed counterclockwise from the top panel Snapshots are taken from apoapse (true anomaly f = minus1800top panel) to hours after periapse (f = 119 follow arrows) and back again to apoapse Each panel is plotted on the same temperature colorscale Snapshots weretaken between 6103 and 6108 Earth days Note that peak temperatures lag periapse passage Periapse corresponds to f = 0 and apoapse to 180 The vertical linecorresponds to substellar longitude

(A color version of this figure is available in the online journal)

6

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=000deg

minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus1242deg

minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=1242deg

minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus7227deg

minus200

minus200 200

200

200

600

600

1000

1000

1400

1400

18002200260030003400

3800

4200

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=7227deg

minus200

minus200minus200 200

200

200

600

600

1000

1000

1400

180022

002600300034003800

4200

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus11988deg

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=11988deg

minus600minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

1800

2200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus18000deg

minus200

minus200

200

200

600

600

1000

1000

1400

1800

2200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

Figure 6 Snapshots of the zonal-mean zonal wind for the same integration described in Figure 5 (Teq = 951 K e = 050 and a = 00534 AU) Snapshots are againtaken from apoapse (f = minus1800) to hours after periapse (f = 119) and back again to apoapse Each panel is plotted on the same temperature colorscale Thesnapshots correspond to outputs from 6103 to 6108 Earth days Note that the equatorial superrotating jet is persistent throughout the orbit with peak speeds exceeding5000 m sminus1

(A color version of this figure is available in the online journal)

7

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 000

200

200

200600

600

600

1000

1000

10001400

1400

18002200

2600

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 000

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 025

minus600

minus600

minus200 minus200

200

200

600

600

1000

140018002200260030003400

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 025

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 050

minus200

minus200

200

200

600

600

1000

1000

1400

18002200

2600300034003800

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 050

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 075

200

200

200

200

600

600

600

1000

1000

1000

140018

002200260030003400

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 075

(K)

500

600

700

800

900

1000

1100

1200

1300

Figure 7 Comparison of simulations with Teq = 951 K with increasing eccentricity from e = 00 (top row) to e = 075 (bottom row) Shown are plots of thezonal-mean zonal wind averaged over a full orbit (left column) and snapshots of the wind and temperature profiles at 30 mbar (right column) The snapshots of thewind and temperature profiles in the right column were taken at periapse passage The vertical bars in the right column denote the longitude of the substellar pointThe snapshots of wind and temperature correspond to model outputs of (from top to bottom) 4100 5006 6106 and 6904 Earth days

(A color version of this figure is available in the online journal)

8

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

from 1000 K to 1300 K In addition the variation in temper-ature from dayside to nightside increases this strengthens thewinds within the equatorial superrotating jet Peak winds reachsim2500 m sminus1 in the circular case but increase to sim5000 m sminus1

in the high eccentricity caseAs the eccentricity increases the equatorial jet also narrows

in latitude In the case of highest eccentricity the atmospheredevelops westward midlatitude jets and eastward high-latitudejets The narrowing of the equatorial jet arises from the factthat jet width is confined to the Rossby radius of deformation(Showman amp Polvani 2011) the length scale over which theatmosphere adjusts to planetary-scale phenomena The Rossbyradius of deformation (Holton 2004) is defined at the equator as

LR =radic

c

β (6)

where c is the gravity wave speed and β is the derivative ofthe Coriolis parameter f with respect to latitudinal distancey β = df dy = 2Ωrot cos φRp where Ωrot is the planetaryrotation rate φ is the latitude and Rp is the planetary radius Atthe equator β is simply 2ΩrotRp Hence at the equator LR is

LR =radic

cRp

2Ωrot (7)

Therefore LR prop 1radic

Ωrot and an increased rotation rateleads to a narrower deformation radius From the Hut (1981)formulation the combined effect of the larger orbital period andincrease in eccentricity produces a shorter rotational periodThis in turn increases the rotation rate which reduces thedeformation radius For example the ratio between Ωrot forthe e = 00 case and the e = 075 case is 63 hence we wouldexpect that the equatorial jet should be narrower by a factor ofradic

63 sim 25 This is indeed the case as the e = 075 case hasan average jet width of sim40 latitude while the e = 00 casehas an average jet width of sim100 latitude

33 Effect of Varying Average Stellar Flux

An increase in average stellar flux leads to increases intemperature and wind speeds illustrated in Figure 8 Each caseshown has an eccentricity of 025 these maps have independentcolorscales to better show atmospheric structure The planetthat receives the highest average stellar flux throughout its orbitis hottest overall (bottom) with peak temperatures of 1300 KFurthermore because the planet with the highest 〈F 〉 orbits at thesmallest orbital distance it has a faster rotation rate and hencea narrower equatorial jet this comparison makes the effect ofrotation on the jet structure much more apparent

34 Global Average Temperature versus Eccentricity and 〈F 〉Here we illustrate the variation in time of maximum

planet-wide temperature with eccentricity and average stellarflux Figure 9 shows the globally averaged temperature foreach model integration plotted as a function of mean anomaly9

where M = 0 corresponds to periapse passage With increas-ing eccentricity the time at which the global-mean temperaturereaches its peak value decreases This is expected as a planet

9 The mean anomaly M is defined as (2πPorb)(t minus τ ) where τ is the timeof periapse passage

with a larger eccentricity will have a higher global-average tem-perature at periapse and hence shorter radiative time constant(τrad prop T minus3 see Showman et al 2011) Figure 10 plots the timeof peak temperature from periapse versus eccentricity Here theinverse relationship between eccentricity and the time of peaktemperature is clear

Figures 9 and 10 serve as a nice summary of the modelintegrations and their relation to observations as temperatureis one of the planetary properties that shapes light curvesat secondary eclipse Observers can use these predictions toestimate the timing of peak flux (and global temperature) basedon a given stellar flux andeccentricity (see Section 4)

35 Synchronous versus Pseudo-synchronous Rotation

Here we explore the sensitivity of the circulation to rota-tion rate by comparing models performed with the pseudo-synchronous rotation rate to otherwise identical models per-formed with the synchronous rotation rate This is illustratedin Figure 11 Here the rotation rate varies by almost a factorof three Both cases still have an equatorial superrotating jetand the overall atmospheric structure is similar However thereare subtle differences in the synchronous case the equatorialeastward winds and high-latitude westward winds are higher inmagnitude (right column) The synchronous case also exhibitsmore latitudinal and longitudinal temperature variation partic-ularly on the dayside this is more apparent at apoapse (bot-tom row) In the pseudo-synchronous case the faster rotationrate serves to homogenize the temperature in longitude becausethe time for rotation to carry air parcels from day to night isshorter The temperature difference from equator to pole is alsolarger because the faster rotation rate inhibits equator-to-poleheat transport Thus while it is unlikely that an eccentric planetsynchronously rotates its star the distinction can lead to notabledifferences in the mean flow

These results can also be used to compare the Hut (1981)formulation shown by the pseudo-synchronous case to for-mulations by other groups As shown in Figure 11 if anotherprescription for rotation rate differed from Hut (1981) by afactor of 2ndash3 (as illustrated by the synchronous case) the qual-itative picture would not look that different However if theprescription differed by a factor of more than 3ndash4 the picturecould substantially differ For example the rotation rate calcu-lated from Ivanov amp Papaloizou (2004) Equation (63) yields avalue of 14 times 106 s a factor of four larger than the Hut (1981)calculation So results using this prescription would look verydifferent

Still the rotation rate of eccentric exoplanets is a completelyunconstrained problem One must be careful then in choosinga formulation noting its assumptions and limitations

36 Momentum Budget and Equatorial Superrotation

In most of our simulations the planet maintains an equatorialeastward jet throughout its orbit Superrotation is commonin three-dimensional circulation models of hot Jupiters (egShowman amp Guillot 2002 Showman et al 2008 2009 Menouamp Rauscher 2009 Rauscher amp Menou 2010 2012 Cooperamp Showman 2005 2006 Dobbs-Dixon amp Lin 2008 Dobbs-Dixon et al 2010 Heng et al 2011) and is also seen withinour own solar system on Venus Titan Jupiter and SaturnWhile superrotation on solar system planets has been extensivelystudied only recently have the mechanisms for superrotation onsynchronously rotating exoplanets been identified (Showman amp

9

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)a= 00802 AU

minus200

minus200

200

200

200

600

600

1000

100014001800

22002600

(m sminus1)

0

500

1000

1500

2000

2500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00802 AU

(K)

550

600

650

700

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00505 AU

minus600

minus600

minus200 minus200

200

200

600

600

1000

140018002200260030003400

(m sminus1)

minus500

0

500

1000

1500

2000

2500

3000

3500

4000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00505 AU

(K)

700

750

800

850

900

950

1000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00318 AU

minus200

minus200

200

200

200 200

600

600

600

10001400

1800220026003000340038004200

5000

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00318 AU

(K)700

800

900

1000

1100

1200

1300

Figure 8 Comparison of simulations with e = 025 with decreasing average stellar flux from bottom to top panels Shown are plots of the orbit-averaged zonal-meanzonal wind (left column) and the wind and temperature profiles at 30 mbar at periapse (right column) Each plot has an independent colorscale The vertical bars inthe right column denote the substellar longitude The snapshots of wind and temperature correspond to model outputs of (from top to bottom) 3706 5006 and 2201Earth days

(A color version of this figure is available in the online journal)

10

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

0 05 1 15 2 25 3 35620

640

660

680

700

720

740ltFgt = 11617 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050e=075

0 05 1 15 2 25 3 35880

890

900

910

920

930

940

950

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

ltFgt = 73680 Wmminus1

e=025e=050

0 05 1 15 2 25 3 35 4960

980

1000

1020

1040

1060

1080

1100

1120

1140

1160ltFgt = 185691 Wmminus1

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

Mean anomaly

e=025e=050e=075

0 05 1 15 2 25 3 35 4 451100

1150

1200

1250

1300

1350ltFgt = 468183 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050

Figure 9 Plots of global average temperature vs mean anomaly for each modelintegration Each plot contains profiles of a given average stellar flux (468183185691 73680 and 11617 W mminus1) with blue green and red representingprofiles for e = 025 050 and 075 respectively

(A color version of this figure is available in the online journal)

03 04 05 06 070

05

1

15

2

25

3

eccentricity

Pea

k o

ffse

t fr

om

per

iap

se (

Ear

th d

ays)

468183 Wmminus1

185691 Wmminus1

73680 Wmminus1

11617 Wmminus1

Figure 10 Time after periapse passage at which the global-average temperaturepeaks (from Figure 9) vs eccentricity for all model integrations The bluegreen red and teal profiles correspond to average stellar fluxes of 468183185691 73680 and 11617 W mminus1 respectively

(A color version of this figure is available in the online journal)

Polvani 2011 hereafter SP2011) As discussed in their paperthe generation of superrotation requires up-gradient momentumtransport that pumps angular momentum from outside the jet towithin it Only transport by waves and eddies can satisfy thisrequirement (Hide 1969)

SP2011 demonstrate that for synchronous hot Jupiters onnear-circular orbits such as HD 189733b and HD 209458bequatorial superrotation is generated by standing planetaryscale Kelvin and Rossby waves at the equator and midlati-tudes respectively that are themselves the dynamical responseto the planetrsquos large dayndashnight heating contrast Because theKelvin waves propagate eastward and the Rossby waves prop-agate westward a phase tilt of the wind vectors emerges fromnorthwest to southeast in the northern hemisphere and south-west to northeast in the southern hemisphere with an over-all shape resembling a chevron pointing eastward This patterncauses meridional eddy angular momentum fluxes that convergeangular momentum onto the equator generating equatorialsuperrotation

Despite the fact that not all our model integrations aresynchronous our results suggest that similar mechanisms arestill at play The chevron shape that SP2011 describe is exactlywhat we noted earlier in our plots of the wind and temperatureprofiles from Figure 5 suggesting that as in SP2011 our modelsexhibit an equatorward flux of eddy angular momentum Thisis demonstrated quantitatively in Figure 12 which shows thezonally averaged meridional flux of relative zonal eddy angularmomentum uprimevprime cos φ at snapshots throughout the orbit (asin Figures 5 and 6) Here uprime and vprime represent the deviationof the zonal and meridional winds respectively from theirzonal averages Positive values indicate a northward flux ofeastward eddy angular momentum and negative values indicatea southward flux of eastward eddy angular momentum Atapoapse the small dayndashnight forcing causes poleward transportof eddy angular momentum at latitudes greater than 50 butequatorward transport of eddy angular momentum equator of50 latitude the amplitudes are weak due to the low insolationat apoapse As the planet moves closer to the star (middlerow left column) the increase in dayndashnight forcing strengthens

11

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)e = 050 pseudominussynchronous rotation

minus200

200

200

600

600

1000

10001400

180022002600

3000

3400

3800

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

synchronous rotation

minus600

minus600

minus200

minus200 minus200

minus200

minus200

200

200

600

10001400180022002600300034003800

4200

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

Figure 11 Comparison of pseudo-synchronous (Trot = 3046 times 105 s) vs synchronous (Trot = 8544 times 105 s) rotation for a case with Teq = 755 K e = 050 anda = 00848 AU Wind and temperature snapshots are taken at periapse (middle row) and apoapse (bottom row) at 5008 Earth days and 5013 Earth days respectivelyThe vertical bar denotes substellar longitude

(A color version of this figure is available in the online journal)

the equatorward flux of eastward eddy angular momentumat latitudes between 0 and 50 in both hemispheres Themomentum flux grows and persists until hours after periapseAngular momentum thus converges onto the equator andmaintains the equatorial jet against westward accelerationscaused by advection friction (if any) and Coriolis forces

These mechanisms also explain the difference in wind speedsbetween the synchronous and pseudo-synchronous cases inFigure 11 The synchronous case has a higher dayndashnight heatingcontrast hence a higher equatorward flux of eddy momentumleading to a stronger narrower equatorial jet Moreover theequatorial jet in the pseudo-synchronous case extends to deeper

12

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=18000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

Figure 12 Eddy momentum flux uprimevprime cos φ throughout the orbit of a model where Teq = 951 K e = 05 and a = 00534 AU Positive values indicate a northwardflux of eastward eddy angular momentum while negative values indicate a southward flux of eastward eddy angular momentum The maximum momentum fluxoccurs hours after periapse passage (f = 72) The snapshots correspond to model outputs from 6103 to 6108 Earth days

(A color version of this figure is available in the online journal)

13

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

Figure 13 Differing orbital viewing geometries explored in this study ω = 180 ω = 270 and ω = 360

(A color version of this figure is available in the online journal)

pressures than that of the synchronous case this is most likelydue to a change in momentum budget

4 OBSERVATIONAL IMPLICATIONS

Observations of transiting exoplanets in circular orbits areknown to be shaped by the atmospheric circulation As wasfirst predicted for hot Jupiters by Showman amp Guillot (2002)and later observationally confirmed for hot Jupiter HD 189733bby Knutson et al (2007) an equatorial superrotating jet causesan eastward displacement of the hot spot from the substellarpoint under appropriate conditions Hence a light curve showsa peak in infrared flux minutes to hours before secondaryeclipse We expect such a situation for eccentric exoplanetsas well However observations of eccentric planets can becomplicated because the flux received by the planet variesthroughout its orbit Previous studies explored such situationsbut have not used three-dimensional models (eg Langton ampLaughlin 2008a 2008b Cowan amp Agol 2011) We can expandon these studies and the three-dimensional study by Lewis et al(2010) by using our SPARC model results

We choose to generate light curves following the proceduresdescribed in Fortney et al (2006) for each model integrationat three values of the argument of periastron (ω) the anglebetween the radius vector to the ascending node and the periapseof the orbit In particular we choose orientations where transitoccurs before periapse (at f = minus90ω = 180) transit occursat apoapse (ω = 270) and transit occurs after periapse (atf = +90ω = 360) (Figure 13) These values reflect thebroad range of viewing geometries of transiting exoplanets forexample HAT-P-2b (e = 052) HAT-P-17b (e = 035) andHD 97658b (e = 013) have ω near 180 WASP-8b (e = 031)HAT-P-31b (e = 025) and HD 80606b (e = 093) have ω near270 and GJ 436b (e = 015) XO-3b (e = 026) and HAT-P-11b (e = 020) have ω near 360 (see exoplaneteu for a full listof values) Note that when transit occurs the nightside is visibleto Earth while the dayside is visible at the time of secondaryeclipse

41 Observational Effects of Varying ω

Using the nominal case (e = 050 a = 00534 AU Teq =951 K) we illustrate the dependence of full-orbit light curves onthe value of ω Figure 14 shows three light curves of the same

model integration but at the differing values of ω described inthe previous section Hence the differences in light curves resultsolely from the differing viewing geometries The planetstarflux ratio is plotted as a function of Earth days from periapsein the warm Spitzer wavelength bands (36 and 45 μm) andJHK bands (126 165 and 220 μm) The solid and dashedlines correspond to transit and secondary eclipse respectivelyFor the light curve where ω = 180 (top panel) the peak ininfrared (IR) flux occurs approximately 6 hr after periapse andbefore secondary eclipse The timing of the peak is due toa combination of the radiative lag of the planet and the factthat the hottest regions rotate into an Earth-facing orientationhours after periapse passage The dayside which is most visiblenear secondary eclipse is hottest after periapse Additionallytemperatures reach peak values at times close to when thosehottest regions are facing Earth This helps explain the highamplitude of the light curve peak in the ω = 180 case relative tothe other two cases These effects can be visualized by plottingthe temperature near the photosphere along the equator in apolar projection (Figure 15) Snapshots throughout the orbit plotthe temperature in a polar projection as a function of pressureand orbital position at a latitudinal slice near the equatorthe solid and dashed lines denote the substellar longitude andEarth-facing longitude respectively Based on the temperaturesat the Earth-facing longitude it is apparent that the peak in IRflux should occur after periapse and before secondary eclipse(see f = 12 and 72) this is indeed the case

For the same model but instead with ω = 270 the peak inIR flux occurs 1 hr after eclipse and periapse In this case thedayside is visible near periapse and therefore we are seeing theplanet increase in temperature (and IR flux) during its closestapproach to the star However the hottest regions are displacedeastward of the substellar point and therefore the hottest regionsface Earth significantly before periapse passage (Figure 15 atf asymp 72) Thus by the time the peak temperatures are reachednear and after periapse the hottest regions are already rotatingout of view from Earth This explains the lower amplitude ofthe light curve peak as compared to the peak of the ω = 180case The IR peak decreases as the planet moves further awayfrom the star along our line of sight A second ramp-up in fluxoccurs approximately 1 day after periapseeclipse this is due tothe hottest regions rotating back into view as the planet movestoward apoapse

14

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015P

lane

tSta

r F

lux

Rat

io

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

Figure 14 Three light curves of the IR flux (expressed as a planetstar fluxratio) vs time throughout the full orbit for a single model integration whereTeq = 951 K e = 050 and a = 00534 AU The flux is plotted in warm Spitzerbandpasses 36 and 45 μm and JHK bands (126 165 220 μm respectively)These three cases only differ by their value of ω shown in Figure 13 The solidlines denote transit while the dotted lines denote secondary eclipse

(A color version of this figure is available in the online journal)

In the case where ω = 360 two distinct peaks are presentmdashasmaller peak that occurs after secondary eclipse and sim3ndash4 hrbefore periapse and a second higher peak that occurs aftertransit The first peak occurs as some of the dayside is visible asthe planet is heated approaching periapse The peak decreasesas the dayside rotates out of view For this geometry the hottestregions are facing away from Earth when they reach their peaktemperatures (Figure 15) The second peak occurs after transitas the hottest regions again rotate into view At this point in theorbit peak temperatures have already occurred and thereforethis peak is lower in amplitude than the other two light curvesHowever because the planet is hotter than it was after secondary

eclipse this second peak has a higher amplitude than the firstone

42 High-e Model Integrations at Varying 〈F 〉If an eccentric planet has a large enough dayndashnight tempera-

ture difference and a sufficiently rapid rotation rate its resultantlight curve can show a periodic rise and fall in flux through-out its orbit The light curves shown in Figure 16 are twoe = 075 model integrations that differ in 〈F 〉 the top panelshows a case where 〈F 〉 = 468183 W mminus2 (Teq = 1199 K)and a = 00385 AU while the bottom panel is a case where〈F 〉 = 185691 W mminus2 (Teq = 951 K) and a = 00611 AU Forboth cases light curves are shown for the ω = 180 geome-try (see Figure 13) The differences in 〈F 〉 lead to differencesin light curve amplitude in the two models the closer hotterplanet exhibits a higher flux amplitude Not surprisingly thetiming of peak IR flux for both models resembles the ω = 180case in Figure 14 The peak in each case occurs after periapseand before eclipse due to the combined effects of the ther-mal lag of the planet and geometry of the system Additionallyboth light curves exhibit a quasi-periodic rise and fall of IR fluxdays after periapse passage This phenomenon called ldquoringingrdquowas also seen in models by Langton amp Laughlin (2008a) andCowan amp Agol (2011) caused by the hottest point of the planetrotating in and out of view from Earth For ringing to occurthe temperature difference from dayside to nightside must belarge and it must survive for multiple planetary rotation periodsIn the models shown in Figure 16 this is aided by the rapidpseudo-synchronous rotation rate associated with an eccentric-ity of 075 (Table 2) For the top case the dayndashnight tempera-ture difference is over sim1300 K while the bottom case variesby sim750 K (Figure 7) The light curves from Figure 14 do notexhibit ringing because the dayndashnight temperature difference islow (sim300ndash400 K)

Because the radiative and dynamical timescales in the modelsby Cowan amp Agol (2011) are calculated by a simple scaling withtemperature the ringing seen in their simulated light curvesoccurs with a period equal to the planetrsquos assumed solid-bodyrotation period in the inertial frame of the winds In realityan eccentric exoplanet approaching apoapse would becomeincreasingly cold the winds would weaken and decrease andhence the ringing should be non-periodic Indeed for ourthree dimensional models the ringing in each case has varyingperiodsmdashthe top case has a ringing period ranging from 03 to04 days (Trot = 035 days) while the bottom panel has a periodranging 07ndash08 days (Trot = 07 days)

Despite these differences this study and the studies byLangton amp Laughlin (2008a) and Cowan amp Agol (2011) illus-trate the influence of atmospheric dynamics on the observationsof eccentric exoplanets The light curve features described herecould be seen in future follow-up observations of eccentric ex-oplanets and observers can use the conclusions drawn aboveto relate the observations to the planetrsquos atmospheric structureFor example if ringing is present in an IR light curve the planetmust have a strong dayndashnight temperature asymmetry and hencewe would expect the planet to have strong superrotation at theequator (Showman amp Polvani 2011) This is probably the casefor highly eccentric (e gt 05) short-period exoplanets such asHD 17156b and HD 80606b If ringing were observed it wouldalso allow an estimate of the sum of the rotation speed and thesuperrotation speed thereby placing constraints on both windspeeds and rotation rates of eccentric hot Jupiters The timing ofpeak IR flux relative to transit and secondary eclipse and also

15

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

f=000deg

(K)800

850

900

950

1000

f=minus1242deg

(K)800

850

900

950

1000f=1242deg

(K)800

850

900

950

1000

f=minus7227deg

(K)800

850

900

950

1000f=7227deg

(K)800

850

900

950

1000

f=minus11988deg

(K)800

850

900

950

1000f=11988deg

(K)800

850

900

950

1000

f=minus18000deg

(K)800

850

900

950

1000

Figure 15 Plots of temperature vs longitude and pressure in equatorial slices throughout one full orbit for our nominal model with Teq = 951 K e = 050 anda = 00534 AU Each annulus shows temperature vs longitude and log-pressure for a snapshot at a particular orbital position (denoted by the value of f) as viewedfrom the planetrsquos north pole Pressures between 30 and 43 mbar are displayed in each annulus The different annuli mark the time evolution throughout one completeorbit starting from apoapse for the top annulus proceeding to periapse and then back to apoapse (follow arrows moving counterclockwise around the plot) Thesesnapshots correspond to model outputs from 6103 to 6108 Earth days The solid line in each snapshot indicates the longitude of the substellar point while the dashedline denotes the Earth-facing longitude for the particular case of ω = 180 For the cases of ω = 270 and 360 degrees exactly the same plot is valid except that thesub-Earth longitude would be aiming upward for ω = 270 and right for ω = 360

(A color version of this figure is available in the online journal)

16

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

e = 075 a = 00385 AU

-15 -10 -05 00 05 10 15Time after periapse passage (days)

0000

0001

0002

0003

0004

Pla

netS

tar

Flu

x R

atio

e = 075 a = 00611 AU

-3 -2 -1 0 1 2 3Time after periapse passage (days)

00000

00005

00010

00015

00020

00025

Pla

netS

tar

Flu

x R

atio

Figure 16 Light Curves for two different e = 075 model integrationsTeq = 1199 K a = 00385 AU (top) and Teq = 951 K a = 00611 AU(bottom) both with ω = 180 The solid line denotes transit while the dottedline denotes secondary eclipse

(A color version of this figure is available in the online journal)

relative to orbital position can help to constrain the circulationfurther

5 CONCLUSIONS

We present three-dimensional circulation models coupledwith a two-stream non-gray radiative transfer scheme fora number of theoretical eccentric hot Jupiters We haveshown that as in published models with zero eccentricity ourhigh-eccentricity circulation models are dominated by eastwardflow at photospheric levels which cause an eastward displace-ment of the hottest regions from the substellar point The rapidrotation rates associated with pseudo-synchronization at higheccentricity lead to a small Rossby deformation radius andin some cases multiple jets in the atmosphere Global-meantemperatures and dayndashnight temperature differences peak notat periapse but several hours afterward due to finite radiativetimescales in the planetrsquos atmosphere

Furthermore we show that equatorial superrotation is gener-ated and maintained by eddies formed by the strong dayndashnightheating contrast which induce a flux of momentum from mid-latitudes to the equator The eddy magnitudes and momentumfluxes peak just after periapse passage leading to variations inthe zonal-mean flow throughout the orbit

Figure 17 Globally averaged pressurendashtemperature profiles from 1-D radiative-convective radiative transfer models of HD 189733b with 11 frequency bins(red profiles) 30 frequency bins (green profiles) and 196 frequency bins (blackprofiles) From left to right the profiles vary in stellar flux from the nominalvalue as well as 2times 4times and 8times higher flux The 8times higher flux case has atemperature inversion

(A color version of this figure is available in the online journal)

Lastly we have shown that the spatial and temporal vari-ations of the wind and temperature structure as well as theorbital viewing geometry of the system with respect to Earthcan affect the time and amplitude of peak IR flux seen infull-orbit light curves Depending on the viewing geometry ofthe orbit relative to Earth we find that peaks in IR flux thateither lead or lag periapse are possible in all cases a combi-nation of temporal effects (temperatures changing over time)and geometric effects (hot spots rotating into or out of view)are important in controlling the timing and amplitude of theflux peaks In cases where the dayndashnight temperature contrastis large and the rotational period is short the light curve canalso exhibit ldquoringingrdquo in flux as the hottest region of the planetrotates in and out of view This ringing is non-periodic due tothe variation in stellar heating as a function of distance

This work was supported by NASA Origins and PlanetaryAtmospheres grants to APS TK also acknowledges supportfrom the Harriet P Jenkins Pre-Doctoral Fellowship Program(JPFP) Resources supporting this work were provided bythe NASA High-End Computing (HEC) Program through theNASA Advanced Supercomputing (NAS) Division at AmesResearch Center The authors thank the anonymous referee fortheir helpful comments and suggestions

APPENDIX

VALIDATION TESTS FOR UPDATEDRADIATIVE TRANSFER SCHEME

To validate our updated radiative transfer scheme using 11frequency bins we conducted the following tests

1 First we tested the sensitivity of the number of frequencybins in a one-dimensional radiative-convective model ofHD 189733b (a = 00313 AU e = 00 RP = 82396 times107 m g = 214 m sminus2) Figure 17 compares the globallyaveraged pressurendashtemperature profiles of models run with11 frequency bins 30 frequency bins and 196 frequencybins at (from left to right) 1times 2times 4times and 8times theaverage stellar flux of HD 189733b For each group of

17

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)30 opacity bins

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)

11 opacity bins

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

30 opacity bins

minus600

minus200minus200

200

200

600

6001000

1000

14001800

2200

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

11 opacity bins

minus600

minus200

minus200

200

200

600

600

10001000

1400

1800

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

30 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

11 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

Figure 18 Globally averaged pressurendashtemperature profiles (top row) zonal-mean zonal wind (middle row) and the windtemperature profiles at 30 mbar (bottomrow) for the 30-bin (left column) and 11-bin (right column) model integrations of HD 189733b Snapshots were output at 250 Earth days

(A color version of this figure is available in the online journal)

models the 11-bin profile agrees well with the 30- and196-bin cases from 1 μbar to 1 bar the temperature variesby only tens of K Below 10 bars the 30- and 11-binprofiles differ from the 196-bin cases by up to 50 Kdue to the loss of resolution at low wavenumbers Inthe case of the highest average flux the large amount of

heating leads to a temperature inversion that is not fullycaptured by the 11-bin model but can be improved witha better initial guess at the pressure-temperature (P-T)profile

2 After testing the new radiative transfer (RT) scheme forone-dimensional models we ran full three-dimensional

18

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

simulations of HD 189733b at its nominal average stellarflux using the SPARCMITgcm Figure 18 shows the glob-ally averaged P-T profiles (top row) zonal-mean zonal wind(middle row) and windtemperature profiles at 30 mbar(bottom row) for the 30-bin (left column) and 11-bin (rightcolumn) RT setups The globally averaged profiles arenearly identical with any temperature differences withina few K The zonal wind plots also show good agree-ment with similar peak speeds (gt2600 m sminus1) equatorialjet width (sim60) and jet level (lt100 bar) At 30 mbarthe 30- and 11-bin models share similar dayside-nightsidetemperature structure with a hot spot (1100 K) eastward ofthe substellar point and two colder regions on the nightsidein the mid-latitudes

Both tests show that the bulk circulation and temperaturestructure is retained in the transition from 30 to 11 frequencybins hence we proceed to use the SPARCMITgcm with thenew scheme for these and future model simulations

REFERENCES

Adcroft A Campin J-M Hill C amp Marshall J 2004 MWRv 132 2845Charbonneau D Knutson H A Barman T et al 2008 ApJ 686 1341Cho J Y-K Menou K Hansen B M S amp Seager S 2008 ApJ

675 817Cooper C S amp Showman A P 2005 ApJL 629 L45Cooper C S amp Showman A P 2006 ApJ 649 1048Cowan N B amp Agol E 2011 ApJ 726 82Deming D Harrington J Laughlin G et al 2007 ApJL 667 L199Dobbs-Dixon I Cumming A amp Lin D N C 2010 ApJ 710 1395Dobbs-Dixon I amp Lin D N C 2008 ApJ 673 513Fortney J J Cooper C S Showman A P Marley M S amp Freedman R S

2006 ApJ 652 746Fortney J J Lodders K Marley M S amp Freedman R S 2008 ApJ

678 1419Fortney J J Marley M S Lodders K Saumon D amp Freedman R

2005 ApJL 627 L69

Fu Q amp Liou K N 1992 JAtS 49 2140Goody R West R Chen L amp Crisp D 1989 JQSRT 42 539Greenberg R 2009 ApJL 698 L42Heng K 2012 ApJL 761 L1Heng K Frierson D M W amp Phillipps P J 2011 MNRAS 418 2669Hide R 1969 JAtS 26 841Holton J R 2004 An Introduction to Dynamic Meteorology (Waltham MA

Academic)Hut P 1981 AampA 99 126Irwin J Charbonneau D Nutzman P et al 2008 ApJ 681 636Ivanov P B amp Papaloizou J C B 2004 MNRAS 353 1161Knutson H A Charbonneau D Allen L E Burrows A amp Megeath S T

2008 ApJ 673 526Knutson H A Charbonneau D Allen L E et al 2007 Natur 447 183Knutson H A Charbonneau D Cowan N B et al 2009 ApJ 690 822Langton J amp Laughlin G 2008a ApJ 674 1106Langton J amp Laughlin G 2008b AampA 483 25Lewis N K Showman A P Fortney J J et al 2010 ApJ 720 344Lodders K 2003 ApJ 591 1220Marley M S amp McKay C P 1999 Icar 138 268Mayor M amp Queloz D 1995 Natur 378 355Menou K amp Rauscher E 2009 ApJ 700 887Mlawer E J Taubman S J Brown P D Iacono M J amp Clough S A

1997 JGR 102 16663Naef D Latham D W Mayor M et al 2001 AampA 375 L27Perna R Heng K amp Pont F 2012 ApJ 751 59Pont F Knutson H Gilliland R L Moutou C amp Charbonneau D

2008 MNRAS 385 109Rauscher E amp Menou K 2010 ApJ 714 1334Rauscher E amp Menou K 2012 ApJ 750 96Saumon D Marley M S Abel M Frommhold L amp Freedman R S

2012 ApJ 750 74Showman A P Cho J Y-K amp Menou K 2011 in Exoplanets ed S Seager

(Tucson AZ Univ Arizona Press) 471Showman A P Cooper C S Fortney J J amp Marley M S 2008 ApJ

682 559Showman A P Fortney J J Lian Y et al 2009 ApJ 699 564Showman A P amp Guillot T 2002 AampA 385 166Showman A P amp Polvani L M 2011 ApJ 738 71Swain M R Vasisht G amp Tinetti G 2008 Natur 452 329Thrastarson H T amp Cho J Y 2010 ApJ 716 144Wolszczan A 1994 Sci 264 538

19

  • 1 INTRODUCTION
  • 2 MODEL
    • 21 Updates to SPARCMITgcm
    • 22 Grid Simulations
      • 3 RESULTS
        • 31 Circulation Regime
        • 32 Effect of Eccentricity
        • 33 Effect of Varying Average Stellar Flux
        • 34 Global Average Temperature versus Eccentricity and F
        • 35 Synchronous versus Pseudo-synchronous Rotation
        • 36 Momentum Budget and Equatorial Superrotation
          • 4 OBSERVATIONAL IMPLICATIONS
            • 41 Observational Effects of Varying ω
            • 42 High-e Model Integrations at Varying F
              • 5 CONCLUSIONS
              • APPENDIX VALIDATION TESTS FOR UPDATED RADIATIVE TRANSFER SCHEME
              • REFERENCES
Page 2: THREE-DIMENSIONAL ATMOSPHERIC CIRCULATION OF HOT …showman/publications/kataria-etal-2013.pdf · planets’ atmospheric composition, temperature structure, and circulation. A fifth

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

Hence it is crucial to conduct a comprehensive study thatestablishes the dynamical regime temperature structure andobservational implications of eccentric exoplanets We use athree-dimensional atmospheric circulation model coupled toa non-gray radiative transfer scheme to study eccentric hotJupiters as a whole In Section 2 we will describe our modelsetup and integrations Section 3 describes the dynamicalregime and its dependence on eccentricity and mean stellarflux Section 4 presents synthetic light curves and attempts todetermine what can be learned about atmospheric circulationsfrom remote measurements Finally Section 5 concludes andcompares results to known eccentric hot Jupiters

2 MODEL

We adopt the Substellar and Planetary Atmospheric Radiationand Circulation (SPARC) model which couples the MITgcm(Adcroft et al 2004) with a two-stream implementation of themulti-stream non-gray radiative transfer scheme developed byMarley amp McKay (1999) A brief discussion of the model isincluded here for a more complete description see Showmanet al (2009)

The MITgcm solves the primitive equations a simplificationof the NavierndashStokes equations where the horizontal flow lengthscale exceeds vertical length scales these equations are valid atlarge scales in stably stratified atmospheres Hot Jupiters gener-ally satisfy these criteria with horizontal and vertical lengthscales of sim104 km and sim300 km respectively and highlystratified atmospheres While the primitive equations do havelimitations (for example the inability to capture small-scaleweakly stratified features such as fronts and storms) thelarge-scale flow (eg jets and waves) is captured The use ofthe primitive equations also presents many computational ad-vantages over the fully compressible equations The radiativetransfer code solves the two-stream radiative transfer equationsand employs the correlated-k method (eg Mlawer et al 1997Goody et al 1989 Fu amp Liou 1992 Marley amp McKay 1999) tosolve for upwarddownward fluxes and heatingcooling ratesthrough an inhomogeneous atmosphere This method retainsmost of the accuracy of full line-by-line calculations whiledrastically increasing computational efficiency The correlated-kmethod maps the absorption coefficients denoted by k(ν) whereν is frequency from spectral space (k = k(ν)) to a spacewhere k varies with a variable g (k = k(g)) The inversionof this function g(k) is commonly referred to as the cumulativedistribution

gi =int k

0fi(k

prime) dkprime (1)

where fi(k) is the distribution function for the absorptioncoefficient in the ith atmosphere layer in a spectral interval νIn the space of cumulative probability g varies from 0 to 1 andrepresents the probability of the absorption coefficient beingless than or equal to k(g) The total frequency space is splitinto a number of discrete frequency intervals (windows) and theabsorption coefficients at each grid point in each window aresorted by frequency of occurrence From this a g distribution isderived for each window The radiative transfer is then carriedout window by window usually using a g(k) that has beenrepresented by a Gaussian division scheme in order to integratethe distribution over each window

For each window n different monochromatic absorptioncoefficients Ki (where i varies from 1 to n) are computed atn gauss points in order to integrate over the k(g) curve These

values of Ki are used to calculate the transmission T as afunction of column mass u by the equation

T (u) =nsumi

WieminusKiu (2)

where Wi is the gauss weight for each fractional gauss pointIn this way multiple opacity sources can be modeled Thismodel accounts for gaseous (Rayleigh) scattering and pressure-induced absorption by H2 and He Scattering and absorptiondue to clouds aerosols and dust can also be included thoughfor simplicity we do not include these effects here The fluxeswithin each spectral interval are weighted and summed toobtain the upward and downward fluxes for each bin at eachatmospheric layer The sum of all intervals gives the totalwavelength-integrated fluxes for each layer We then calculatethe heating rate by finite-differencing the fluxes and pressuresbetween interfaces over and underlying a given dynamicallevel

q = gpartF

partp (3)

Here q is the heating rate g is the planetrsquos gravity and partFpartpis the gradient of flux with pressure For each simulation weutilize a cubed-sphere grid with a horizontal resolution of C32(approximately equivalent to 64 times 128 in latitude and longitude)and NL = 40 or 76 pressure levels The lowermost NL minus 1levels extend from a mean pressure of 200 bars at the bottom to02 mbar at the top evenly spaced in log pressure The top levelextends from a pressure of 02 mbar to zero

We implement a fourth-order Shapiro filter for temperatureand momentum with a timestep equal to double the dynam-ical timestep Models integrated with this Shapiro filter setup(but excluding any other large-scale drag) conserve total an-gular momentum to better than 01 thereby demonstratingexcellent conservation of angular momentum with our modelconfiguration

21 Updates to SPARCMITgcm

211 Reduced Number of Frequency Bins

While the models of HD 189733b and HD 209458b inShowman et al (2009) and the models of GJ 436b in Lewis et al(2010) used 30 frequency bins to model circulation and heatingon each planet here the opacities are statistically weighted into11 frequency bins to improve computational efficiency For boththe 30- and 11-bin implementations the radiative transfer iscalculated at eight values of k(g) in each of the frequency binsFour of the values of k(g) are sampled at g lt 095 meaning thatfour calculations are done for the weakest 95 of the spectrallines The other four radiative transfer calculations are done forthe strongest 5 of the lines (g gt 095) This allows us tocalculate the radiative transfer for both strong and weak spectrallines

We list the 11 new frequency bins in Table 1 Before imple-menting the updated scheme into our simulations we conductedone-dimensional and three-dimensional tests to reproduce theresults for HD 1897833b from Showman et al (2009) Oncesuccessful we applied the new scheme to the model integra-tions described here (The Appendix describes these validationtests in detail)

2

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

Table 1Bounding Wavelengths of the Frequency Bins Used

in the Radiative Transfer Calculations

Wavelength Wavelength(μm) (μm)

32468 20002000 870870 440440 350350 250250 202202 132132 085085 061061 042042 026

212 Updated Collision-induced Absorption dueto H2ndashH2 and H2ndashHe Collisions

The collision-induced absorption (CIA) by H2 moleculesdue to H2ndashH2 and H2ndashHe collisions has recently been recom-puted from first principles (Saumon et al 2012 and referencestherein) The new formulation uses state-of-the-art quantum me-chanical calculations of the potential energy surface and of theinteraction-induced dipole moment surface for colliding H2ndashH2and H2ndashHe pairs This new calculation is more reliable at highertemperature and higher photon energies than the earlier cal-culations by Borysow and collaborators that radiative transfergroups have been using (httpwwwastrokudksimaborysow)We have included these new tabulations of H2 CIA opacity inour code

22 Grid Simulations

Table 2 lists our model integrations for this study For allmodels we assume the radius and gravity of a generic hotJupiter7 Figure 1 shows the parameter space plotted as afunction of semi-major axis and eccentricity along with detectedexoplanets with semi-major axes less than 05 AU In these

7 HD 189733b RP = 82396 times 107 m g = 214 m sminus2 Note that we couldhave easily chosen any other well-characterized hot Jupiter such as HD209458b However the primary goal of this study is to illustrate the dynamicaland observational effects of eccentricity and average stellar flux The preciseplanetary radius and gravity are of secondary importance

0 01 02 03 04 050

01

02

03

04

05

06

07

08

09

1

semiminusmajor axis

ecce

ntr

icit

y

detected exoplanet

468183 W mminus2

185691 W mminus2

73680 W mminus2

11617 W mminus2

Figure 1 Model integrations from this study plotted against a portion of thecurrent exoplanet population (dots) The green right-pointing triangles reddown-pointing triangles blue left-pointing triangles and magenta up-pointingtriangles correspond to average stellar fluxes of 468183 185691 73680 and11617 W mminus2 respectively See Table 2 and text for more details

(A color version of this figure is available in the online journal)

simulations we systematically vary both the average stellarflux and eccentricity Orbital eccentricities of 00ndash075 andorbit-averaged stellar fluxes 〈F 〉 of 468183 185691 73680and 11617 W mminus2 were explored These values correspond toequilibrium temperatures (Teq) of approximately 1199 951 755and 476 K respectively For these given values of e and 〈F 〉the orbital semi-major axis a is calculated using the equation

〈F 〉 = L

4πa2(1 minus e2)12 (4)

where L is the luminosity of HD 189733 These chosen simula-tions fall well within the current exoplanet population

We assume the atmospheric composition to be 1times solarmetallicity without TiO and VO this composition has hadsuccess in explaining observations of HD 189733b As inShowman et al (2009) opacities are determined assuming localchemical equilibrium (accounting for rainout of condensates)

Table 2List of Model Integrations

〈F 〉 (W mminus2) Teq a e rp ra Porb Prot PorbProt Ω (sminus1)(K) (AU) (AU) (AU) (s) (s)

468183 1199 00313 000 00313 00313 1917 times 105 1917 times 105 10 327 times 10minus5

468183 1199 00318 025 00239 00398 1963 times 105 1421 times 105 14 442 times 10minus5

468183 1199 00385 075 000963 00674 2615 times 105 3010 times 104 87 209 times 10minus4

185691 951 00497 000 00497 00497 3834 times 105 3834 times 105 10 164 times 10minus5

185691 951 00505 025 00379 00631 3927 times 105 2843 times 105 14 221 times 10minus5

185691 951 00534 050 00267 00801 4270 times 105 1522 times 105 28 413 times 10minus5

185691 951 00611 075 00153 01069 5226 times 105 6016 times 104 87 104 times 10minus4

73680 755 00789 000 00789 00789 7668 times 105 7668 times 105 10 819 times 10minus6

73680 755 00802 025 00602 01003 7858 times 105 5689 times 105 14 110 times 10minus5

73680 755 00848 050 00424 01272 8544 times 105 3046 times 105 28 206 times 10minus5

11617 476 01987 000 01987 01987 3067 times 106 3067 times 106 10 205 times 10minus5

11617 476 02019 025 01514 02524 3141 times 106 2274 times 106 14 276 times 10minus5

11617 476 02135 050 01068 03202 3416 times 106 1218 times 106 28 516 times 10minus6

11617 476 02443 075 00611 04275 4181 times 106 4813 times 105 87 131 times 10minus5

3

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

0 500 1000 1500 2000 2500

10minus6

10minus4

10minus2

100

102

104

Temperature (K)

Pre

ssu

re (

bar

s)

11617 W mminus2

73680 W mminus2

185691 W mminus2

468183 W mminus2

Figure 2 Initial 1-D pressurendashtemperature profiles assigned to each column ofthe atmosphere at the start of each simulation at the four different valuesof average stellar flux These profiles were calculated assuming radiativeequilibrium and use the same opacities as the full 3D model

(A color version of this figure is available in the online journal)

at each temperature and pressure over the 3D grid assumingthe elemental abundances of Lodders (2003) We set cp and κto values appropriate for a H2-dominated atmosphere (13 times104 J kgminus1 Kminus1 and 27 respectively)

For eccentric cases we assume the planet is pseudo-synchronously rotating its host starmdashwe assume the planetrsquostidal interactions with the star force the same side of the planetto approximately face the star every periapse passage We cal-

culate this using the Hut (1981) formulation

Prot = Porb

(1 + 3e2 + 3

8e4)

(1 minus e2)32

1 + 152 e2 + 45

8 e4 + 516e6

(5)

Pseudo-synchronization is a commonly used assumption inthe calculation of the rotation rate of eccentric exoplanets (egDeming et al 2007 Irwin et al 2008 Langton amp Laughlin2008a 2008b Lewis et al 2010) Nevertheless the true depen-dence of rotation period on orbital period and eccentricity isvery uncertain (eg Greenberg 2009) and other formulationsbesides Equation (5) have also been suggested in the litera-ture (eg Ivanov amp Papaloizou 2004) We explore the sensitiv-ity of the flow to the assumption of pseudo-synchronization inSection 34

In each integration we assume the winds to be initiallyzero with each vertical atmospheric column assigned the sameglobal-mean radiative equilibrium pressurendashtemperature profilecalculated using the same opacities as the main model (Figure 2)These temperaturendashpressure profiles are calculated for eachvalue of average stellar flux (see above) using the methodsdescribed in Fortney et al (2005 2008)

Most model integrations were run until the winds reacheda statistically steady configuration at upper levels Figure 3shows the rms velocity plotted as a function of pressure andsimulated time for a model integrated for gt6000 Earth daysThis behavior in rms velocity is typical of most model inte-grations Weak drag was applied to the atmosphere with atime constant of 100 days at 200 bars with a drag coeffi-cient that decreases linearly with pressure to zero at 10 barsTests were conducted to explore the sensitivity of drag on themean flow Extending the drag up to 1 bar affects the verti-cal and longitudinal extent of the equatorial jet but not the

0 1000 2000 3000 4000 5000 6000

10minus3

10minus2

10minus1

100

101

102

simulated time (Earth days)

pre

ssu

re (

bar

)

e= 050 a= 00534 AU

00003217611

0000321

200

200 200

400

400 400

600

600 600

800

800 800

1000

1000 1000

1200

1200 1200

1400

1400 1400

1600

16001600

1800

18001800

2000

2000

2000

2200

2200

2200

2400

2400

2600

260028003000

(m sminus1)0

500

1000

1500

2000

2500

3000

Figure 3 rms velocity plotted as a function of pressure and simulation time for one of our model integrations The plot shows a case where Teq = 951 K e = 050and a = 00534 AU rms velocity was calculated at each pressure level from wind speeds output every 100 days The rapid decrease in velocity at the start of the runis merely an artifact at the start of the run

(A color version of this figure is available in the online journal)

4

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus180 minus90 0 90 180True anomaly

minus2 minus15 minus1 minus05 0 05 1 15 2

10minus3

10minus2

10minus1

100

101

102

simulated time (Earth days)

pre

ssu

re (

bar

) 700700

800800

800

900

900

900

1000

10001000

11001100

1100

12001200 1200

1300 1300 1300

1400 1400 1400

1500 1500 1500

T(K)

700

800

900

1000

1100

1200

1300

1400

1500

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00534 AU

minus200

minus200

200

200

600

600

1000

1000

1400

18002200

2600300034003800

(m sminus1)

0

1000

2000

3000

4000

5000

Figure 4 Dynamical response of a model with Teq = 951 K e = 050 anda = 00534 AU Top global horizontally averaged temperature vs pressureand time throughout the full orbit zero is defined as periapse passage The upperaxis gives the true anomaly throughout the orbit Bottom zonally averaged zonal(eastndashwest) winds vs pressure and latitude for the same model time-averagedover a full orbit

(A color version of this figure is available in the online journal)

qualitative nature of the circulation We reiterate that inte-grations without this drag scheme (but including our fourth-order Shapiro filter) conserve angular momentum to betterthan 01

3 RESULTS

31 Circulation Regime

To illustrate the circulation regime of an eccentric exoplanetwe present a nominal model where e = 050 a = 00534 AUand Teq = 951 K representative of the simulations (Figure 4)The top panel shows the horizontally averaged global temper-ature as a function of pressure time (in Earth days) and trueanomaly f where f = 0 at periapse and f = plusmn180 atapoapse Temperatures remain fairly constant at pressures ex-ceeding 1 bar However in the uppermost layers of the atmo-sphere temperatures vary throughout the orbit peaking sim6 hr

after periapse passage The bottom panel shows the zonal-meanzonal wind8 of the same model time-averaged over a full or-bit The flow exhibits a superrotating (eastward) equatorial jetwhich is the dominant feature of the flow with peak winds ofabout 5000 m sminus1 Superrotation is commonly seen in models ofhot Jupiters and hot Neptunes (eg Showman amp Guillot 2002Cooper amp Showman 2005 Showman et al 2008 2009 Dobbs-Dixon amp Lin 2008 Dobbs-Dixon et al 2010 Lewis et al 2010Heng et al 2011 Perna et al 2012 Rauscher amp Menou 20102012) although not by all groups (Cho et al 2008 Thrastarsonamp Cho 2010)

With peak winds exceeding 2ndash5 km sminus1 the equatorial jet issupersonic as the sound speed for this class of planets is ap-proximately 2 km sminus1 As Heng (2012) shows the presence ofthis supersonic flow generates shocks at the dayndashnight termi-nator upstream of the substellar point The SPARCMITgcmdoes not properly account for the removal of kinetic energy as aresult of these shocks The formation of shocks is expected to beshallow (ie at low pressure) in both of the model integrationsthat Heng (2012) presents the Mach number falls below unityat pressures greater than 1 bar and hence shocks are not formedNevertheless the infrared photosphere and the peak jet speedsin our model occur at sufficiently low pressure that shocks couldplay some role particularly when the orbital eccentricity is highMoreover as a possible energy dissipation mechanism shockscould have a non-local effect (ie they could cause an indirecteffect on the flow even at pressures deeper than those where theshocks directly occur) Clearly more work is warranted on therole of shocks in hot Jupiter atmospheres

Snapshots of the wind and temperature profile throughout theorbit at a pressure level of 30 mbar (approximately the levelof the infrared photosphere) show that the planet undergoes aflash heating event as the planet approaches periapse passage(Figure 5) The eight panels show snapshots throughout onecomplete orbit from apoapse (top figure) to hours after periapse(f = 119 follow arrows) and back again to apoapse all withthe same colorscale for temperature Like the circulation ofHD 189733b in its nominally circular orbit (see Showman et al2009) the flow maintains the eastward equatorial superrotatingjet throughout its orbit (Figure 6) As the planet approachesperiapse (Figure 5 f = minus12) temperatures rise to sim1000 Kat the equator and a high-amplitude eddy structure developswith a pattern of temperature and wind vectors that tilt eastwardtoward the equator This so-called ldquochevronrdquo-shaped feature isprominent at periapse (bottom panel) extends over a wide rangeof latitude and longitude and persists hours after periapse Thechevron shape is also seen in our other integrations and indicatesthat eddies transport momentum equatorward (see Section 36)

32 Effect of Eccentricity

Our models show that at constant orbit-mean stellar fluxthe dayside temperatures dayndashnight temperature differencesand wind speeds at periapse passage all increase with increas-ing orbital eccentricity This is illustrated in Figure 7 whichcompares three simulations with an equilibrium temperature of951 K (〈F 〉 = 185691 W mminus2) that vary in eccentricity from00 (top row) to 075 (bottom row) Again we plot each columnon the same colorscale (zonal wind and temperature) for com-parison As eccentricity increases peak temperatures increase

8 Zonal wind is defined as eastndashwest winds with positive (negative) valuesdenoting eastward (westward) winds a zonal average is an average inlongitude

5

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=000deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus1242deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=1242deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus7227deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=7227deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus11988deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=11988deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus18000deg

(K)500

600

700

800

900

1000

1100

1200

Figure 5 Snapshots of the wind and temperature structure at 30 mbar of a planet with Teq = 951 K e = 050 and a = 00534 AU To trace the behavior throughoutone complete orbit the different panels should be followed counterclockwise from the top panel Snapshots are taken from apoapse (true anomaly f = minus1800top panel) to hours after periapse (f = 119 follow arrows) and back again to apoapse Each panel is plotted on the same temperature colorscale Snapshots weretaken between 6103 and 6108 Earth days Note that peak temperatures lag periapse passage Periapse corresponds to f = 0 and apoapse to 180 The vertical linecorresponds to substellar longitude

(A color version of this figure is available in the online journal)

6

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=000deg

minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus1242deg

minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=1242deg

minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus7227deg

minus200

minus200 200

200

200

600

600

1000

1000

1400

1400

18002200260030003400

3800

4200

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=7227deg

minus200

minus200minus200 200

200

200

600

600

1000

1000

1400

180022

002600300034003800

4200

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus11988deg

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=11988deg

minus600minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

1800

2200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus18000deg

minus200

minus200

200

200

600

600

1000

1000

1400

1800

2200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

Figure 6 Snapshots of the zonal-mean zonal wind for the same integration described in Figure 5 (Teq = 951 K e = 050 and a = 00534 AU) Snapshots are againtaken from apoapse (f = minus1800) to hours after periapse (f = 119) and back again to apoapse Each panel is plotted on the same temperature colorscale Thesnapshots correspond to outputs from 6103 to 6108 Earth days Note that the equatorial superrotating jet is persistent throughout the orbit with peak speeds exceeding5000 m sminus1

(A color version of this figure is available in the online journal)

7

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 000

200

200

200600

600

600

1000

1000

10001400

1400

18002200

2600

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 000

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 025

minus600

minus600

minus200 minus200

200

200

600

600

1000

140018002200260030003400

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 025

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 050

minus200

minus200

200

200

600

600

1000

1000

1400

18002200

2600300034003800

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 050

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 075

200

200

200

200

600

600

600

1000

1000

1000

140018

002200260030003400

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 075

(K)

500

600

700

800

900

1000

1100

1200

1300

Figure 7 Comparison of simulations with Teq = 951 K with increasing eccentricity from e = 00 (top row) to e = 075 (bottom row) Shown are plots of thezonal-mean zonal wind averaged over a full orbit (left column) and snapshots of the wind and temperature profiles at 30 mbar (right column) The snapshots of thewind and temperature profiles in the right column were taken at periapse passage The vertical bars in the right column denote the longitude of the substellar pointThe snapshots of wind and temperature correspond to model outputs of (from top to bottom) 4100 5006 6106 and 6904 Earth days

(A color version of this figure is available in the online journal)

8

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

from 1000 K to 1300 K In addition the variation in temper-ature from dayside to nightside increases this strengthens thewinds within the equatorial superrotating jet Peak winds reachsim2500 m sminus1 in the circular case but increase to sim5000 m sminus1

in the high eccentricity caseAs the eccentricity increases the equatorial jet also narrows

in latitude In the case of highest eccentricity the atmospheredevelops westward midlatitude jets and eastward high-latitudejets The narrowing of the equatorial jet arises from the factthat jet width is confined to the Rossby radius of deformation(Showman amp Polvani 2011) the length scale over which theatmosphere adjusts to planetary-scale phenomena The Rossbyradius of deformation (Holton 2004) is defined at the equator as

LR =radic

c

β (6)

where c is the gravity wave speed and β is the derivative ofthe Coriolis parameter f with respect to latitudinal distancey β = df dy = 2Ωrot cos φRp where Ωrot is the planetaryrotation rate φ is the latitude and Rp is the planetary radius Atthe equator β is simply 2ΩrotRp Hence at the equator LR is

LR =radic

cRp

2Ωrot (7)

Therefore LR prop 1radic

Ωrot and an increased rotation rateleads to a narrower deformation radius From the Hut (1981)formulation the combined effect of the larger orbital period andincrease in eccentricity produces a shorter rotational periodThis in turn increases the rotation rate which reduces thedeformation radius For example the ratio between Ωrot forthe e = 00 case and the e = 075 case is 63 hence we wouldexpect that the equatorial jet should be narrower by a factor ofradic

63 sim 25 This is indeed the case as the e = 075 case hasan average jet width of sim40 latitude while the e = 00 casehas an average jet width of sim100 latitude

33 Effect of Varying Average Stellar Flux

An increase in average stellar flux leads to increases intemperature and wind speeds illustrated in Figure 8 Each caseshown has an eccentricity of 025 these maps have independentcolorscales to better show atmospheric structure The planetthat receives the highest average stellar flux throughout its orbitis hottest overall (bottom) with peak temperatures of 1300 KFurthermore because the planet with the highest 〈F 〉 orbits at thesmallest orbital distance it has a faster rotation rate and hencea narrower equatorial jet this comparison makes the effect ofrotation on the jet structure much more apparent

34 Global Average Temperature versus Eccentricity and 〈F 〉Here we illustrate the variation in time of maximum

planet-wide temperature with eccentricity and average stellarflux Figure 9 shows the globally averaged temperature foreach model integration plotted as a function of mean anomaly9

where M = 0 corresponds to periapse passage With increas-ing eccentricity the time at which the global-mean temperaturereaches its peak value decreases This is expected as a planet

9 The mean anomaly M is defined as (2πPorb)(t minus τ ) where τ is the timeof periapse passage

with a larger eccentricity will have a higher global-average tem-perature at periapse and hence shorter radiative time constant(τrad prop T minus3 see Showman et al 2011) Figure 10 plots the timeof peak temperature from periapse versus eccentricity Here theinverse relationship between eccentricity and the time of peaktemperature is clear

Figures 9 and 10 serve as a nice summary of the modelintegrations and their relation to observations as temperatureis one of the planetary properties that shapes light curvesat secondary eclipse Observers can use these predictions toestimate the timing of peak flux (and global temperature) basedon a given stellar flux andeccentricity (see Section 4)

35 Synchronous versus Pseudo-synchronous Rotation

Here we explore the sensitivity of the circulation to rota-tion rate by comparing models performed with the pseudo-synchronous rotation rate to otherwise identical models per-formed with the synchronous rotation rate This is illustratedin Figure 11 Here the rotation rate varies by almost a factorof three Both cases still have an equatorial superrotating jetand the overall atmospheric structure is similar However thereare subtle differences in the synchronous case the equatorialeastward winds and high-latitude westward winds are higher inmagnitude (right column) The synchronous case also exhibitsmore latitudinal and longitudinal temperature variation partic-ularly on the dayside this is more apparent at apoapse (bot-tom row) In the pseudo-synchronous case the faster rotationrate serves to homogenize the temperature in longitude becausethe time for rotation to carry air parcels from day to night isshorter The temperature difference from equator to pole is alsolarger because the faster rotation rate inhibits equator-to-poleheat transport Thus while it is unlikely that an eccentric planetsynchronously rotates its star the distinction can lead to notabledifferences in the mean flow

These results can also be used to compare the Hut (1981)formulation shown by the pseudo-synchronous case to for-mulations by other groups As shown in Figure 11 if anotherprescription for rotation rate differed from Hut (1981) by afactor of 2ndash3 (as illustrated by the synchronous case) the qual-itative picture would not look that different However if theprescription differed by a factor of more than 3ndash4 the picturecould substantially differ For example the rotation rate calcu-lated from Ivanov amp Papaloizou (2004) Equation (63) yields avalue of 14 times 106 s a factor of four larger than the Hut (1981)calculation So results using this prescription would look verydifferent

Still the rotation rate of eccentric exoplanets is a completelyunconstrained problem One must be careful then in choosinga formulation noting its assumptions and limitations

36 Momentum Budget and Equatorial Superrotation

In most of our simulations the planet maintains an equatorialeastward jet throughout its orbit Superrotation is commonin three-dimensional circulation models of hot Jupiters (egShowman amp Guillot 2002 Showman et al 2008 2009 Menouamp Rauscher 2009 Rauscher amp Menou 2010 2012 Cooperamp Showman 2005 2006 Dobbs-Dixon amp Lin 2008 Dobbs-Dixon et al 2010 Heng et al 2011) and is also seen withinour own solar system on Venus Titan Jupiter and SaturnWhile superrotation on solar system planets has been extensivelystudied only recently have the mechanisms for superrotation onsynchronously rotating exoplanets been identified (Showman amp

9

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)a= 00802 AU

minus200

minus200

200

200

200

600

600

1000

100014001800

22002600

(m sminus1)

0

500

1000

1500

2000

2500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00802 AU

(K)

550

600

650

700

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00505 AU

minus600

minus600

minus200 minus200

200

200

600

600

1000

140018002200260030003400

(m sminus1)

minus500

0

500

1000

1500

2000

2500

3000

3500

4000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00505 AU

(K)

700

750

800

850

900

950

1000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00318 AU

minus200

minus200

200

200

200 200

600

600

600

10001400

1800220026003000340038004200

5000

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00318 AU

(K)700

800

900

1000

1100

1200

1300

Figure 8 Comparison of simulations with e = 025 with decreasing average stellar flux from bottom to top panels Shown are plots of the orbit-averaged zonal-meanzonal wind (left column) and the wind and temperature profiles at 30 mbar at periapse (right column) Each plot has an independent colorscale The vertical bars inthe right column denote the substellar longitude The snapshots of wind and temperature correspond to model outputs of (from top to bottom) 3706 5006 and 2201Earth days

(A color version of this figure is available in the online journal)

10

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

0 05 1 15 2 25 3 35620

640

660

680

700

720

740ltFgt = 11617 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050e=075

0 05 1 15 2 25 3 35880

890

900

910

920

930

940

950

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

ltFgt = 73680 Wmminus1

e=025e=050

0 05 1 15 2 25 3 35 4960

980

1000

1020

1040

1060

1080

1100

1120

1140

1160ltFgt = 185691 Wmminus1

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

Mean anomaly

e=025e=050e=075

0 05 1 15 2 25 3 35 4 451100

1150

1200

1250

1300

1350ltFgt = 468183 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050

Figure 9 Plots of global average temperature vs mean anomaly for each modelintegration Each plot contains profiles of a given average stellar flux (468183185691 73680 and 11617 W mminus1) with blue green and red representingprofiles for e = 025 050 and 075 respectively

(A color version of this figure is available in the online journal)

03 04 05 06 070

05

1

15

2

25

3

eccentricity

Pea

k o

ffse

t fr

om

per

iap

se (

Ear

th d

ays)

468183 Wmminus1

185691 Wmminus1

73680 Wmminus1

11617 Wmminus1

Figure 10 Time after periapse passage at which the global-average temperaturepeaks (from Figure 9) vs eccentricity for all model integrations The bluegreen red and teal profiles correspond to average stellar fluxes of 468183185691 73680 and 11617 W mminus1 respectively

(A color version of this figure is available in the online journal)

Polvani 2011 hereafter SP2011) As discussed in their paperthe generation of superrotation requires up-gradient momentumtransport that pumps angular momentum from outside the jet towithin it Only transport by waves and eddies can satisfy thisrequirement (Hide 1969)

SP2011 demonstrate that for synchronous hot Jupiters onnear-circular orbits such as HD 189733b and HD 209458bequatorial superrotation is generated by standing planetaryscale Kelvin and Rossby waves at the equator and midlati-tudes respectively that are themselves the dynamical responseto the planetrsquos large dayndashnight heating contrast Because theKelvin waves propagate eastward and the Rossby waves prop-agate westward a phase tilt of the wind vectors emerges fromnorthwest to southeast in the northern hemisphere and south-west to northeast in the southern hemisphere with an over-all shape resembling a chevron pointing eastward This patterncauses meridional eddy angular momentum fluxes that convergeangular momentum onto the equator generating equatorialsuperrotation

Despite the fact that not all our model integrations aresynchronous our results suggest that similar mechanisms arestill at play The chevron shape that SP2011 describe is exactlywhat we noted earlier in our plots of the wind and temperatureprofiles from Figure 5 suggesting that as in SP2011 our modelsexhibit an equatorward flux of eddy angular momentum Thisis demonstrated quantitatively in Figure 12 which shows thezonally averaged meridional flux of relative zonal eddy angularmomentum uprimevprime cos φ at snapshots throughout the orbit (asin Figures 5 and 6) Here uprime and vprime represent the deviationof the zonal and meridional winds respectively from theirzonal averages Positive values indicate a northward flux ofeastward eddy angular momentum and negative values indicatea southward flux of eastward eddy angular momentum Atapoapse the small dayndashnight forcing causes poleward transportof eddy angular momentum at latitudes greater than 50 butequatorward transport of eddy angular momentum equator of50 latitude the amplitudes are weak due to the low insolationat apoapse As the planet moves closer to the star (middlerow left column) the increase in dayndashnight forcing strengthens

11

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)e = 050 pseudominussynchronous rotation

minus200

200

200

600

600

1000

10001400

180022002600

3000

3400

3800

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

synchronous rotation

minus600

minus600

minus200

minus200 minus200

minus200

minus200

200

200

600

10001400180022002600300034003800

4200

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

Figure 11 Comparison of pseudo-synchronous (Trot = 3046 times 105 s) vs synchronous (Trot = 8544 times 105 s) rotation for a case with Teq = 755 K e = 050 anda = 00848 AU Wind and temperature snapshots are taken at periapse (middle row) and apoapse (bottom row) at 5008 Earth days and 5013 Earth days respectivelyThe vertical bar denotes substellar longitude

(A color version of this figure is available in the online journal)

the equatorward flux of eastward eddy angular momentumat latitudes between 0 and 50 in both hemispheres Themomentum flux grows and persists until hours after periapseAngular momentum thus converges onto the equator andmaintains the equatorial jet against westward accelerationscaused by advection friction (if any) and Coriolis forces

These mechanisms also explain the difference in wind speedsbetween the synchronous and pseudo-synchronous cases inFigure 11 The synchronous case has a higher dayndashnight heatingcontrast hence a higher equatorward flux of eddy momentumleading to a stronger narrower equatorial jet Moreover theequatorial jet in the pseudo-synchronous case extends to deeper

12

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=18000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

Figure 12 Eddy momentum flux uprimevprime cos φ throughout the orbit of a model where Teq = 951 K e = 05 and a = 00534 AU Positive values indicate a northwardflux of eastward eddy angular momentum while negative values indicate a southward flux of eastward eddy angular momentum The maximum momentum fluxoccurs hours after periapse passage (f = 72) The snapshots correspond to model outputs from 6103 to 6108 Earth days

(A color version of this figure is available in the online journal)

13

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

Figure 13 Differing orbital viewing geometries explored in this study ω = 180 ω = 270 and ω = 360

(A color version of this figure is available in the online journal)

pressures than that of the synchronous case this is most likelydue to a change in momentum budget

4 OBSERVATIONAL IMPLICATIONS

Observations of transiting exoplanets in circular orbits areknown to be shaped by the atmospheric circulation As wasfirst predicted for hot Jupiters by Showman amp Guillot (2002)and later observationally confirmed for hot Jupiter HD 189733bby Knutson et al (2007) an equatorial superrotating jet causesan eastward displacement of the hot spot from the substellarpoint under appropriate conditions Hence a light curve showsa peak in infrared flux minutes to hours before secondaryeclipse We expect such a situation for eccentric exoplanetsas well However observations of eccentric planets can becomplicated because the flux received by the planet variesthroughout its orbit Previous studies explored such situationsbut have not used three-dimensional models (eg Langton ampLaughlin 2008a 2008b Cowan amp Agol 2011) We can expandon these studies and the three-dimensional study by Lewis et al(2010) by using our SPARC model results

We choose to generate light curves following the proceduresdescribed in Fortney et al (2006) for each model integrationat three values of the argument of periastron (ω) the anglebetween the radius vector to the ascending node and the periapseof the orbit In particular we choose orientations where transitoccurs before periapse (at f = minus90ω = 180) transit occursat apoapse (ω = 270) and transit occurs after periapse (atf = +90ω = 360) (Figure 13) These values reflect thebroad range of viewing geometries of transiting exoplanets forexample HAT-P-2b (e = 052) HAT-P-17b (e = 035) andHD 97658b (e = 013) have ω near 180 WASP-8b (e = 031)HAT-P-31b (e = 025) and HD 80606b (e = 093) have ω near270 and GJ 436b (e = 015) XO-3b (e = 026) and HAT-P-11b (e = 020) have ω near 360 (see exoplaneteu for a full listof values) Note that when transit occurs the nightside is visibleto Earth while the dayside is visible at the time of secondaryeclipse

41 Observational Effects of Varying ω

Using the nominal case (e = 050 a = 00534 AU Teq =951 K) we illustrate the dependence of full-orbit light curves onthe value of ω Figure 14 shows three light curves of the same

model integration but at the differing values of ω described inthe previous section Hence the differences in light curves resultsolely from the differing viewing geometries The planetstarflux ratio is plotted as a function of Earth days from periapsein the warm Spitzer wavelength bands (36 and 45 μm) andJHK bands (126 165 and 220 μm) The solid and dashedlines correspond to transit and secondary eclipse respectivelyFor the light curve where ω = 180 (top panel) the peak ininfrared (IR) flux occurs approximately 6 hr after periapse andbefore secondary eclipse The timing of the peak is due toa combination of the radiative lag of the planet and the factthat the hottest regions rotate into an Earth-facing orientationhours after periapse passage The dayside which is most visiblenear secondary eclipse is hottest after periapse Additionallytemperatures reach peak values at times close to when thosehottest regions are facing Earth This helps explain the highamplitude of the light curve peak in the ω = 180 case relative tothe other two cases These effects can be visualized by plottingthe temperature near the photosphere along the equator in apolar projection (Figure 15) Snapshots throughout the orbit plotthe temperature in a polar projection as a function of pressureand orbital position at a latitudinal slice near the equatorthe solid and dashed lines denote the substellar longitude andEarth-facing longitude respectively Based on the temperaturesat the Earth-facing longitude it is apparent that the peak in IRflux should occur after periapse and before secondary eclipse(see f = 12 and 72) this is indeed the case

For the same model but instead with ω = 270 the peak inIR flux occurs 1 hr after eclipse and periapse In this case thedayside is visible near periapse and therefore we are seeing theplanet increase in temperature (and IR flux) during its closestapproach to the star However the hottest regions are displacedeastward of the substellar point and therefore the hottest regionsface Earth significantly before periapse passage (Figure 15 atf asymp 72) Thus by the time the peak temperatures are reachednear and after periapse the hottest regions are already rotatingout of view from Earth This explains the lower amplitude ofthe light curve peak as compared to the peak of the ω = 180case The IR peak decreases as the planet moves further awayfrom the star along our line of sight A second ramp-up in fluxoccurs approximately 1 day after periapseeclipse this is due tothe hottest regions rotating back into view as the planet movestoward apoapse

14

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015P

lane

tSta

r F

lux

Rat

io

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

Figure 14 Three light curves of the IR flux (expressed as a planetstar fluxratio) vs time throughout the full orbit for a single model integration whereTeq = 951 K e = 050 and a = 00534 AU The flux is plotted in warm Spitzerbandpasses 36 and 45 μm and JHK bands (126 165 220 μm respectively)These three cases only differ by their value of ω shown in Figure 13 The solidlines denote transit while the dotted lines denote secondary eclipse

(A color version of this figure is available in the online journal)

In the case where ω = 360 two distinct peaks are presentmdashasmaller peak that occurs after secondary eclipse and sim3ndash4 hrbefore periapse and a second higher peak that occurs aftertransit The first peak occurs as some of the dayside is visible asthe planet is heated approaching periapse The peak decreasesas the dayside rotates out of view For this geometry the hottestregions are facing away from Earth when they reach their peaktemperatures (Figure 15) The second peak occurs after transitas the hottest regions again rotate into view At this point in theorbit peak temperatures have already occurred and thereforethis peak is lower in amplitude than the other two light curvesHowever because the planet is hotter than it was after secondary

eclipse this second peak has a higher amplitude than the firstone

42 High-e Model Integrations at Varying 〈F 〉If an eccentric planet has a large enough dayndashnight tempera-

ture difference and a sufficiently rapid rotation rate its resultantlight curve can show a periodic rise and fall in flux through-out its orbit The light curves shown in Figure 16 are twoe = 075 model integrations that differ in 〈F 〉 the top panelshows a case where 〈F 〉 = 468183 W mminus2 (Teq = 1199 K)and a = 00385 AU while the bottom panel is a case where〈F 〉 = 185691 W mminus2 (Teq = 951 K) and a = 00611 AU Forboth cases light curves are shown for the ω = 180 geome-try (see Figure 13) The differences in 〈F 〉 lead to differencesin light curve amplitude in the two models the closer hotterplanet exhibits a higher flux amplitude Not surprisingly thetiming of peak IR flux for both models resembles the ω = 180case in Figure 14 The peak in each case occurs after periapseand before eclipse due to the combined effects of the ther-mal lag of the planet and geometry of the system Additionallyboth light curves exhibit a quasi-periodic rise and fall of IR fluxdays after periapse passage This phenomenon called ldquoringingrdquowas also seen in models by Langton amp Laughlin (2008a) andCowan amp Agol (2011) caused by the hottest point of the planetrotating in and out of view from Earth For ringing to occurthe temperature difference from dayside to nightside must belarge and it must survive for multiple planetary rotation periodsIn the models shown in Figure 16 this is aided by the rapidpseudo-synchronous rotation rate associated with an eccentric-ity of 075 (Table 2) For the top case the dayndashnight tempera-ture difference is over sim1300 K while the bottom case variesby sim750 K (Figure 7) The light curves from Figure 14 do notexhibit ringing because the dayndashnight temperature difference islow (sim300ndash400 K)

Because the radiative and dynamical timescales in the modelsby Cowan amp Agol (2011) are calculated by a simple scaling withtemperature the ringing seen in their simulated light curvesoccurs with a period equal to the planetrsquos assumed solid-bodyrotation period in the inertial frame of the winds In realityan eccentric exoplanet approaching apoapse would becomeincreasingly cold the winds would weaken and decrease andhence the ringing should be non-periodic Indeed for ourthree dimensional models the ringing in each case has varyingperiodsmdashthe top case has a ringing period ranging from 03 to04 days (Trot = 035 days) while the bottom panel has a periodranging 07ndash08 days (Trot = 07 days)

Despite these differences this study and the studies byLangton amp Laughlin (2008a) and Cowan amp Agol (2011) illus-trate the influence of atmospheric dynamics on the observationsof eccentric exoplanets The light curve features described herecould be seen in future follow-up observations of eccentric ex-oplanets and observers can use the conclusions drawn aboveto relate the observations to the planetrsquos atmospheric structureFor example if ringing is present in an IR light curve the planetmust have a strong dayndashnight temperature asymmetry and hencewe would expect the planet to have strong superrotation at theequator (Showman amp Polvani 2011) This is probably the casefor highly eccentric (e gt 05) short-period exoplanets such asHD 17156b and HD 80606b If ringing were observed it wouldalso allow an estimate of the sum of the rotation speed and thesuperrotation speed thereby placing constraints on both windspeeds and rotation rates of eccentric hot Jupiters The timing ofpeak IR flux relative to transit and secondary eclipse and also

15

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

f=000deg

(K)800

850

900

950

1000

f=minus1242deg

(K)800

850

900

950

1000f=1242deg

(K)800

850

900

950

1000

f=minus7227deg

(K)800

850

900

950

1000f=7227deg

(K)800

850

900

950

1000

f=minus11988deg

(K)800

850

900

950

1000f=11988deg

(K)800

850

900

950

1000

f=minus18000deg

(K)800

850

900

950

1000

Figure 15 Plots of temperature vs longitude and pressure in equatorial slices throughout one full orbit for our nominal model with Teq = 951 K e = 050 anda = 00534 AU Each annulus shows temperature vs longitude and log-pressure for a snapshot at a particular orbital position (denoted by the value of f) as viewedfrom the planetrsquos north pole Pressures between 30 and 43 mbar are displayed in each annulus The different annuli mark the time evolution throughout one completeorbit starting from apoapse for the top annulus proceeding to periapse and then back to apoapse (follow arrows moving counterclockwise around the plot) Thesesnapshots correspond to model outputs from 6103 to 6108 Earth days The solid line in each snapshot indicates the longitude of the substellar point while the dashedline denotes the Earth-facing longitude for the particular case of ω = 180 For the cases of ω = 270 and 360 degrees exactly the same plot is valid except that thesub-Earth longitude would be aiming upward for ω = 270 and right for ω = 360

(A color version of this figure is available in the online journal)

16

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

e = 075 a = 00385 AU

-15 -10 -05 00 05 10 15Time after periapse passage (days)

0000

0001

0002

0003

0004

Pla

netS

tar

Flu

x R

atio

e = 075 a = 00611 AU

-3 -2 -1 0 1 2 3Time after periapse passage (days)

00000

00005

00010

00015

00020

00025

Pla

netS

tar

Flu

x R

atio

Figure 16 Light Curves for two different e = 075 model integrationsTeq = 1199 K a = 00385 AU (top) and Teq = 951 K a = 00611 AU(bottom) both with ω = 180 The solid line denotes transit while the dottedline denotes secondary eclipse

(A color version of this figure is available in the online journal)

relative to orbital position can help to constrain the circulationfurther

5 CONCLUSIONS

We present three-dimensional circulation models coupledwith a two-stream non-gray radiative transfer scheme fora number of theoretical eccentric hot Jupiters We haveshown that as in published models with zero eccentricity ourhigh-eccentricity circulation models are dominated by eastwardflow at photospheric levels which cause an eastward displace-ment of the hottest regions from the substellar point The rapidrotation rates associated with pseudo-synchronization at higheccentricity lead to a small Rossby deformation radius andin some cases multiple jets in the atmosphere Global-meantemperatures and dayndashnight temperature differences peak notat periapse but several hours afterward due to finite radiativetimescales in the planetrsquos atmosphere

Furthermore we show that equatorial superrotation is gener-ated and maintained by eddies formed by the strong dayndashnightheating contrast which induce a flux of momentum from mid-latitudes to the equator The eddy magnitudes and momentumfluxes peak just after periapse passage leading to variations inthe zonal-mean flow throughout the orbit

Figure 17 Globally averaged pressurendashtemperature profiles from 1-D radiative-convective radiative transfer models of HD 189733b with 11 frequency bins(red profiles) 30 frequency bins (green profiles) and 196 frequency bins (blackprofiles) From left to right the profiles vary in stellar flux from the nominalvalue as well as 2times 4times and 8times higher flux The 8times higher flux case has atemperature inversion

(A color version of this figure is available in the online journal)

Lastly we have shown that the spatial and temporal vari-ations of the wind and temperature structure as well as theorbital viewing geometry of the system with respect to Earthcan affect the time and amplitude of peak IR flux seen infull-orbit light curves Depending on the viewing geometry ofthe orbit relative to Earth we find that peaks in IR flux thateither lead or lag periapse are possible in all cases a combi-nation of temporal effects (temperatures changing over time)and geometric effects (hot spots rotating into or out of view)are important in controlling the timing and amplitude of theflux peaks In cases where the dayndashnight temperature contrastis large and the rotational period is short the light curve canalso exhibit ldquoringingrdquo in flux as the hottest region of the planetrotates in and out of view This ringing is non-periodic due tothe variation in stellar heating as a function of distance

This work was supported by NASA Origins and PlanetaryAtmospheres grants to APS TK also acknowledges supportfrom the Harriet P Jenkins Pre-Doctoral Fellowship Program(JPFP) Resources supporting this work were provided bythe NASA High-End Computing (HEC) Program through theNASA Advanced Supercomputing (NAS) Division at AmesResearch Center The authors thank the anonymous referee fortheir helpful comments and suggestions

APPENDIX

VALIDATION TESTS FOR UPDATEDRADIATIVE TRANSFER SCHEME

To validate our updated radiative transfer scheme using 11frequency bins we conducted the following tests

1 First we tested the sensitivity of the number of frequencybins in a one-dimensional radiative-convective model ofHD 189733b (a = 00313 AU e = 00 RP = 82396 times107 m g = 214 m sminus2) Figure 17 compares the globallyaveraged pressurendashtemperature profiles of models run with11 frequency bins 30 frequency bins and 196 frequencybins at (from left to right) 1times 2times 4times and 8times theaverage stellar flux of HD 189733b For each group of

17

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)30 opacity bins

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)

11 opacity bins

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

30 opacity bins

minus600

minus200minus200

200

200

600

6001000

1000

14001800

2200

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

11 opacity bins

minus600

minus200

minus200

200

200

600

600

10001000

1400

1800

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

30 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

11 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

Figure 18 Globally averaged pressurendashtemperature profiles (top row) zonal-mean zonal wind (middle row) and the windtemperature profiles at 30 mbar (bottomrow) for the 30-bin (left column) and 11-bin (right column) model integrations of HD 189733b Snapshots were output at 250 Earth days

(A color version of this figure is available in the online journal)

models the 11-bin profile agrees well with the 30- and196-bin cases from 1 μbar to 1 bar the temperature variesby only tens of K Below 10 bars the 30- and 11-binprofiles differ from the 196-bin cases by up to 50 Kdue to the loss of resolution at low wavenumbers Inthe case of the highest average flux the large amount of

heating leads to a temperature inversion that is not fullycaptured by the 11-bin model but can be improved witha better initial guess at the pressure-temperature (P-T)profile

2 After testing the new radiative transfer (RT) scheme forone-dimensional models we ran full three-dimensional

18

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

simulations of HD 189733b at its nominal average stellarflux using the SPARCMITgcm Figure 18 shows the glob-ally averaged P-T profiles (top row) zonal-mean zonal wind(middle row) and windtemperature profiles at 30 mbar(bottom row) for the 30-bin (left column) and 11-bin (rightcolumn) RT setups The globally averaged profiles arenearly identical with any temperature differences withina few K The zonal wind plots also show good agree-ment with similar peak speeds (gt2600 m sminus1) equatorialjet width (sim60) and jet level (lt100 bar) At 30 mbarthe 30- and 11-bin models share similar dayside-nightsidetemperature structure with a hot spot (1100 K) eastward ofthe substellar point and two colder regions on the nightsidein the mid-latitudes

Both tests show that the bulk circulation and temperaturestructure is retained in the transition from 30 to 11 frequencybins hence we proceed to use the SPARCMITgcm with thenew scheme for these and future model simulations

REFERENCES

Adcroft A Campin J-M Hill C amp Marshall J 2004 MWRv 132 2845Charbonneau D Knutson H A Barman T et al 2008 ApJ 686 1341Cho J Y-K Menou K Hansen B M S amp Seager S 2008 ApJ

675 817Cooper C S amp Showman A P 2005 ApJL 629 L45Cooper C S amp Showman A P 2006 ApJ 649 1048Cowan N B amp Agol E 2011 ApJ 726 82Deming D Harrington J Laughlin G et al 2007 ApJL 667 L199Dobbs-Dixon I Cumming A amp Lin D N C 2010 ApJ 710 1395Dobbs-Dixon I amp Lin D N C 2008 ApJ 673 513Fortney J J Cooper C S Showman A P Marley M S amp Freedman R S

2006 ApJ 652 746Fortney J J Lodders K Marley M S amp Freedman R S 2008 ApJ

678 1419Fortney J J Marley M S Lodders K Saumon D amp Freedman R

2005 ApJL 627 L69

Fu Q amp Liou K N 1992 JAtS 49 2140Goody R West R Chen L amp Crisp D 1989 JQSRT 42 539Greenberg R 2009 ApJL 698 L42Heng K 2012 ApJL 761 L1Heng K Frierson D M W amp Phillipps P J 2011 MNRAS 418 2669Hide R 1969 JAtS 26 841Holton J R 2004 An Introduction to Dynamic Meteorology (Waltham MA

Academic)Hut P 1981 AampA 99 126Irwin J Charbonneau D Nutzman P et al 2008 ApJ 681 636Ivanov P B amp Papaloizou J C B 2004 MNRAS 353 1161Knutson H A Charbonneau D Allen L E Burrows A amp Megeath S T

2008 ApJ 673 526Knutson H A Charbonneau D Allen L E et al 2007 Natur 447 183Knutson H A Charbonneau D Cowan N B et al 2009 ApJ 690 822Langton J amp Laughlin G 2008a ApJ 674 1106Langton J amp Laughlin G 2008b AampA 483 25Lewis N K Showman A P Fortney J J et al 2010 ApJ 720 344Lodders K 2003 ApJ 591 1220Marley M S amp McKay C P 1999 Icar 138 268Mayor M amp Queloz D 1995 Natur 378 355Menou K amp Rauscher E 2009 ApJ 700 887Mlawer E J Taubman S J Brown P D Iacono M J amp Clough S A

1997 JGR 102 16663Naef D Latham D W Mayor M et al 2001 AampA 375 L27Perna R Heng K amp Pont F 2012 ApJ 751 59Pont F Knutson H Gilliland R L Moutou C amp Charbonneau D

2008 MNRAS 385 109Rauscher E amp Menou K 2010 ApJ 714 1334Rauscher E amp Menou K 2012 ApJ 750 96Saumon D Marley M S Abel M Frommhold L amp Freedman R S

2012 ApJ 750 74Showman A P Cho J Y-K amp Menou K 2011 in Exoplanets ed S Seager

(Tucson AZ Univ Arizona Press) 471Showman A P Cooper C S Fortney J J amp Marley M S 2008 ApJ

682 559Showman A P Fortney J J Lian Y et al 2009 ApJ 699 564Showman A P amp Guillot T 2002 AampA 385 166Showman A P amp Polvani L M 2011 ApJ 738 71Swain M R Vasisht G amp Tinetti G 2008 Natur 452 329Thrastarson H T amp Cho J Y 2010 ApJ 716 144Wolszczan A 1994 Sci 264 538

19

  • 1 INTRODUCTION
  • 2 MODEL
    • 21 Updates to SPARCMITgcm
    • 22 Grid Simulations
      • 3 RESULTS
        • 31 Circulation Regime
        • 32 Effect of Eccentricity
        • 33 Effect of Varying Average Stellar Flux
        • 34 Global Average Temperature versus Eccentricity and F
        • 35 Synchronous versus Pseudo-synchronous Rotation
        • 36 Momentum Budget and Equatorial Superrotation
          • 4 OBSERVATIONAL IMPLICATIONS
            • 41 Observational Effects of Varying ω
            • 42 High-e Model Integrations at Varying F
              • 5 CONCLUSIONS
              • APPENDIX VALIDATION TESTS FOR UPDATED RADIATIVE TRANSFER SCHEME
              • REFERENCES
Page 3: THREE-DIMENSIONAL ATMOSPHERIC CIRCULATION OF HOT …showman/publications/kataria-etal-2013.pdf · planets’ atmospheric composition, temperature structure, and circulation. A fifth

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

Table 1Bounding Wavelengths of the Frequency Bins Used

in the Radiative Transfer Calculations

Wavelength Wavelength(μm) (μm)

32468 20002000 870870 440440 350350 250250 202202 132132 085085 061061 042042 026

212 Updated Collision-induced Absorption dueto H2ndashH2 and H2ndashHe Collisions

The collision-induced absorption (CIA) by H2 moleculesdue to H2ndashH2 and H2ndashHe collisions has recently been recom-puted from first principles (Saumon et al 2012 and referencestherein) The new formulation uses state-of-the-art quantum me-chanical calculations of the potential energy surface and of theinteraction-induced dipole moment surface for colliding H2ndashH2and H2ndashHe pairs This new calculation is more reliable at highertemperature and higher photon energies than the earlier cal-culations by Borysow and collaborators that radiative transfergroups have been using (httpwwwastrokudksimaborysow)We have included these new tabulations of H2 CIA opacity inour code

22 Grid Simulations

Table 2 lists our model integrations for this study For allmodels we assume the radius and gravity of a generic hotJupiter7 Figure 1 shows the parameter space plotted as afunction of semi-major axis and eccentricity along with detectedexoplanets with semi-major axes less than 05 AU In these

7 HD 189733b RP = 82396 times 107 m g = 214 m sminus2 Note that we couldhave easily chosen any other well-characterized hot Jupiter such as HD209458b However the primary goal of this study is to illustrate the dynamicaland observational effects of eccentricity and average stellar flux The preciseplanetary radius and gravity are of secondary importance

0 01 02 03 04 050

01

02

03

04

05

06

07

08

09

1

semiminusmajor axis

ecce

ntr

icit

y

detected exoplanet

468183 W mminus2

185691 W mminus2

73680 W mminus2

11617 W mminus2

Figure 1 Model integrations from this study plotted against a portion of thecurrent exoplanet population (dots) The green right-pointing triangles reddown-pointing triangles blue left-pointing triangles and magenta up-pointingtriangles correspond to average stellar fluxes of 468183 185691 73680 and11617 W mminus2 respectively See Table 2 and text for more details

(A color version of this figure is available in the online journal)

simulations we systematically vary both the average stellarflux and eccentricity Orbital eccentricities of 00ndash075 andorbit-averaged stellar fluxes 〈F 〉 of 468183 185691 73680and 11617 W mminus2 were explored These values correspond toequilibrium temperatures (Teq) of approximately 1199 951 755and 476 K respectively For these given values of e and 〈F 〉the orbital semi-major axis a is calculated using the equation

〈F 〉 = L

4πa2(1 minus e2)12 (4)

where L is the luminosity of HD 189733 These chosen simula-tions fall well within the current exoplanet population

We assume the atmospheric composition to be 1times solarmetallicity without TiO and VO this composition has hadsuccess in explaining observations of HD 189733b As inShowman et al (2009) opacities are determined assuming localchemical equilibrium (accounting for rainout of condensates)

Table 2List of Model Integrations

〈F 〉 (W mminus2) Teq a e rp ra Porb Prot PorbProt Ω (sminus1)(K) (AU) (AU) (AU) (s) (s)

468183 1199 00313 000 00313 00313 1917 times 105 1917 times 105 10 327 times 10minus5

468183 1199 00318 025 00239 00398 1963 times 105 1421 times 105 14 442 times 10minus5

468183 1199 00385 075 000963 00674 2615 times 105 3010 times 104 87 209 times 10minus4

185691 951 00497 000 00497 00497 3834 times 105 3834 times 105 10 164 times 10minus5

185691 951 00505 025 00379 00631 3927 times 105 2843 times 105 14 221 times 10minus5

185691 951 00534 050 00267 00801 4270 times 105 1522 times 105 28 413 times 10minus5

185691 951 00611 075 00153 01069 5226 times 105 6016 times 104 87 104 times 10minus4

73680 755 00789 000 00789 00789 7668 times 105 7668 times 105 10 819 times 10minus6

73680 755 00802 025 00602 01003 7858 times 105 5689 times 105 14 110 times 10minus5

73680 755 00848 050 00424 01272 8544 times 105 3046 times 105 28 206 times 10minus5

11617 476 01987 000 01987 01987 3067 times 106 3067 times 106 10 205 times 10minus5

11617 476 02019 025 01514 02524 3141 times 106 2274 times 106 14 276 times 10minus5

11617 476 02135 050 01068 03202 3416 times 106 1218 times 106 28 516 times 10minus6

11617 476 02443 075 00611 04275 4181 times 106 4813 times 105 87 131 times 10minus5

3

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

0 500 1000 1500 2000 2500

10minus6

10minus4

10minus2

100

102

104

Temperature (K)

Pre

ssu

re (

bar

s)

11617 W mminus2

73680 W mminus2

185691 W mminus2

468183 W mminus2

Figure 2 Initial 1-D pressurendashtemperature profiles assigned to each column ofthe atmosphere at the start of each simulation at the four different valuesof average stellar flux These profiles were calculated assuming radiativeequilibrium and use the same opacities as the full 3D model

(A color version of this figure is available in the online journal)

at each temperature and pressure over the 3D grid assumingthe elemental abundances of Lodders (2003) We set cp and κto values appropriate for a H2-dominated atmosphere (13 times104 J kgminus1 Kminus1 and 27 respectively)

For eccentric cases we assume the planet is pseudo-synchronously rotating its host starmdashwe assume the planetrsquostidal interactions with the star force the same side of the planetto approximately face the star every periapse passage We cal-

culate this using the Hut (1981) formulation

Prot = Porb

(1 + 3e2 + 3

8e4)

(1 minus e2)32

1 + 152 e2 + 45

8 e4 + 516e6

(5)

Pseudo-synchronization is a commonly used assumption inthe calculation of the rotation rate of eccentric exoplanets (egDeming et al 2007 Irwin et al 2008 Langton amp Laughlin2008a 2008b Lewis et al 2010) Nevertheless the true depen-dence of rotation period on orbital period and eccentricity isvery uncertain (eg Greenberg 2009) and other formulationsbesides Equation (5) have also been suggested in the litera-ture (eg Ivanov amp Papaloizou 2004) We explore the sensitiv-ity of the flow to the assumption of pseudo-synchronization inSection 34

In each integration we assume the winds to be initiallyzero with each vertical atmospheric column assigned the sameglobal-mean radiative equilibrium pressurendashtemperature profilecalculated using the same opacities as the main model (Figure 2)These temperaturendashpressure profiles are calculated for eachvalue of average stellar flux (see above) using the methodsdescribed in Fortney et al (2005 2008)

Most model integrations were run until the winds reacheda statistically steady configuration at upper levels Figure 3shows the rms velocity plotted as a function of pressure andsimulated time for a model integrated for gt6000 Earth daysThis behavior in rms velocity is typical of most model inte-grations Weak drag was applied to the atmosphere with atime constant of 100 days at 200 bars with a drag coeffi-cient that decreases linearly with pressure to zero at 10 barsTests were conducted to explore the sensitivity of drag on themean flow Extending the drag up to 1 bar affects the verti-cal and longitudinal extent of the equatorial jet but not the

0 1000 2000 3000 4000 5000 6000

10minus3

10minus2

10minus1

100

101

102

simulated time (Earth days)

pre

ssu

re (

bar

)

e= 050 a= 00534 AU

00003217611

0000321

200

200 200

400

400 400

600

600 600

800

800 800

1000

1000 1000

1200

1200 1200

1400

1400 1400

1600

16001600

1800

18001800

2000

2000

2000

2200

2200

2200

2400

2400

2600

260028003000

(m sminus1)0

500

1000

1500

2000

2500

3000

Figure 3 rms velocity plotted as a function of pressure and simulation time for one of our model integrations The plot shows a case where Teq = 951 K e = 050and a = 00534 AU rms velocity was calculated at each pressure level from wind speeds output every 100 days The rapid decrease in velocity at the start of the runis merely an artifact at the start of the run

(A color version of this figure is available in the online journal)

4

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus180 minus90 0 90 180True anomaly

minus2 minus15 minus1 minus05 0 05 1 15 2

10minus3

10minus2

10minus1

100

101

102

simulated time (Earth days)

pre

ssu

re (

bar

) 700700

800800

800

900

900

900

1000

10001000

11001100

1100

12001200 1200

1300 1300 1300

1400 1400 1400

1500 1500 1500

T(K)

700

800

900

1000

1100

1200

1300

1400

1500

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00534 AU

minus200

minus200

200

200

600

600

1000

1000

1400

18002200

2600300034003800

(m sminus1)

0

1000

2000

3000

4000

5000

Figure 4 Dynamical response of a model with Teq = 951 K e = 050 anda = 00534 AU Top global horizontally averaged temperature vs pressureand time throughout the full orbit zero is defined as periapse passage The upperaxis gives the true anomaly throughout the orbit Bottom zonally averaged zonal(eastndashwest) winds vs pressure and latitude for the same model time-averagedover a full orbit

(A color version of this figure is available in the online journal)

qualitative nature of the circulation We reiterate that inte-grations without this drag scheme (but including our fourth-order Shapiro filter) conserve angular momentum to betterthan 01

3 RESULTS

31 Circulation Regime

To illustrate the circulation regime of an eccentric exoplanetwe present a nominal model where e = 050 a = 00534 AUand Teq = 951 K representative of the simulations (Figure 4)The top panel shows the horizontally averaged global temper-ature as a function of pressure time (in Earth days) and trueanomaly f where f = 0 at periapse and f = plusmn180 atapoapse Temperatures remain fairly constant at pressures ex-ceeding 1 bar However in the uppermost layers of the atmo-sphere temperatures vary throughout the orbit peaking sim6 hr

after periapse passage The bottom panel shows the zonal-meanzonal wind8 of the same model time-averaged over a full or-bit The flow exhibits a superrotating (eastward) equatorial jetwhich is the dominant feature of the flow with peak winds ofabout 5000 m sminus1 Superrotation is commonly seen in models ofhot Jupiters and hot Neptunes (eg Showman amp Guillot 2002Cooper amp Showman 2005 Showman et al 2008 2009 Dobbs-Dixon amp Lin 2008 Dobbs-Dixon et al 2010 Lewis et al 2010Heng et al 2011 Perna et al 2012 Rauscher amp Menou 20102012) although not by all groups (Cho et al 2008 Thrastarsonamp Cho 2010)

With peak winds exceeding 2ndash5 km sminus1 the equatorial jet issupersonic as the sound speed for this class of planets is ap-proximately 2 km sminus1 As Heng (2012) shows the presence ofthis supersonic flow generates shocks at the dayndashnight termi-nator upstream of the substellar point The SPARCMITgcmdoes not properly account for the removal of kinetic energy as aresult of these shocks The formation of shocks is expected to beshallow (ie at low pressure) in both of the model integrationsthat Heng (2012) presents the Mach number falls below unityat pressures greater than 1 bar and hence shocks are not formedNevertheless the infrared photosphere and the peak jet speedsin our model occur at sufficiently low pressure that shocks couldplay some role particularly when the orbital eccentricity is highMoreover as a possible energy dissipation mechanism shockscould have a non-local effect (ie they could cause an indirecteffect on the flow even at pressures deeper than those where theshocks directly occur) Clearly more work is warranted on therole of shocks in hot Jupiter atmospheres

Snapshots of the wind and temperature profile throughout theorbit at a pressure level of 30 mbar (approximately the levelof the infrared photosphere) show that the planet undergoes aflash heating event as the planet approaches periapse passage(Figure 5) The eight panels show snapshots throughout onecomplete orbit from apoapse (top figure) to hours after periapse(f = 119 follow arrows) and back again to apoapse all withthe same colorscale for temperature Like the circulation ofHD 189733b in its nominally circular orbit (see Showman et al2009) the flow maintains the eastward equatorial superrotatingjet throughout its orbit (Figure 6) As the planet approachesperiapse (Figure 5 f = minus12) temperatures rise to sim1000 Kat the equator and a high-amplitude eddy structure developswith a pattern of temperature and wind vectors that tilt eastwardtoward the equator This so-called ldquochevronrdquo-shaped feature isprominent at periapse (bottom panel) extends over a wide rangeof latitude and longitude and persists hours after periapse Thechevron shape is also seen in our other integrations and indicatesthat eddies transport momentum equatorward (see Section 36)

32 Effect of Eccentricity

Our models show that at constant orbit-mean stellar fluxthe dayside temperatures dayndashnight temperature differencesand wind speeds at periapse passage all increase with increas-ing orbital eccentricity This is illustrated in Figure 7 whichcompares three simulations with an equilibrium temperature of951 K (〈F 〉 = 185691 W mminus2) that vary in eccentricity from00 (top row) to 075 (bottom row) Again we plot each columnon the same colorscale (zonal wind and temperature) for com-parison As eccentricity increases peak temperatures increase

8 Zonal wind is defined as eastndashwest winds with positive (negative) valuesdenoting eastward (westward) winds a zonal average is an average inlongitude

5

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=000deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus1242deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=1242deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus7227deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=7227deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus11988deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=11988deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus18000deg

(K)500

600

700

800

900

1000

1100

1200

Figure 5 Snapshots of the wind and temperature structure at 30 mbar of a planet with Teq = 951 K e = 050 and a = 00534 AU To trace the behavior throughoutone complete orbit the different panels should be followed counterclockwise from the top panel Snapshots are taken from apoapse (true anomaly f = minus1800top panel) to hours after periapse (f = 119 follow arrows) and back again to apoapse Each panel is plotted on the same temperature colorscale Snapshots weretaken between 6103 and 6108 Earth days Note that peak temperatures lag periapse passage Periapse corresponds to f = 0 and apoapse to 180 The vertical linecorresponds to substellar longitude

(A color version of this figure is available in the online journal)

6

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=000deg

minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus1242deg

minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=1242deg

minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus7227deg

minus200

minus200 200

200

200

600

600

1000

1000

1400

1400

18002200260030003400

3800

4200

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=7227deg

minus200

minus200minus200 200

200

200

600

600

1000

1000

1400

180022

002600300034003800

4200

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus11988deg

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=11988deg

minus600minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

1800

2200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus18000deg

minus200

minus200

200

200

600

600

1000

1000

1400

1800

2200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

Figure 6 Snapshots of the zonal-mean zonal wind for the same integration described in Figure 5 (Teq = 951 K e = 050 and a = 00534 AU) Snapshots are againtaken from apoapse (f = minus1800) to hours after periapse (f = 119) and back again to apoapse Each panel is plotted on the same temperature colorscale Thesnapshots correspond to outputs from 6103 to 6108 Earth days Note that the equatorial superrotating jet is persistent throughout the orbit with peak speeds exceeding5000 m sminus1

(A color version of this figure is available in the online journal)

7

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 000

200

200

200600

600

600

1000

1000

10001400

1400

18002200

2600

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 000

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 025

minus600

minus600

minus200 minus200

200

200

600

600

1000

140018002200260030003400

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 025

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 050

minus200

minus200

200

200

600

600

1000

1000

1400

18002200

2600300034003800

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 050

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 075

200

200

200

200

600

600

600

1000

1000

1000

140018

002200260030003400

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 075

(K)

500

600

700

800

900

1000

1100

1200

1300

Figure 7 Comparison of simulations with Teq = 951 K with increasing eccentricity from e = 00 (top row) to e = 075 (bottom row) Shown are plots of thezonal-mean zonal wind averaged over a full orbit (left column) and snapshots of the wind and temperature profiles at 30 mbar (right column) The snapshots of thewind and temperature profiles in the right column were taken at periapse passage The vertical bars in the right column denote the longitude of the substellar pointThe snapshots of wind and temperature correspond to model outputs of (from top to bottom) 4100 5006 6106 and 6904 Earth days

(A color version of this figure is available in the online journal)

8

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

from 1000 K to 1300 K In addition the variation in temper-ature from dayside to nightside increases this strengthens thewinds within the equatorial superrotating jet Peak winds reachsim2500 m sminus1 in the circular case but increase to sim5000 m sminus1

in the high eccentricity caseAs the eccentricity increases the equatorial jet also narrows

in latitude In the case of highest eccentricity the atmospheredevelops westward midlatitude jets and eastward high-latitudejets The narrowing of the equatorial jet arises from the factthat jet width is confined to the Rossby radius of deformation(Showman amp Polvani 2011) the length scale over which theatmosphere adjusts to planetary-scale phenomena The Rossbyradius of deformation (Holton 2004) is defined at the equator as

LR =radic

c

β (6)

where c is the gravity wave speed and β is the derivative ofthe Coriolis parameter f with respect to latitudinal distancey β = df dy = 2Ωrot cos φRp where Ωrot is the planetaryrotation rate φ is the latitude and Rp is the planetary radius Atthe equator β is simply 2ΩrotRp Hence at the equator LR is

LR =radic

cRp

2Ωrot (7)

Therefore LR prop 1radic

Ωrot and an increased rotation rateleads to a narrower deformation radius From the Hut (1981)formulation the combined effect of the larger orbital period andincrease in eccentricity produces a shorter rotational periodThis in turn increases the rotation rate which reduces thedeformation radius For example the ratio between Ωrot forthe e = 00 case and the e = 075 case is 63 hence we wouldexpect that the equatorial jet should be narrower by a factor ofradic

63 sim 25 This is indeed the case as the e = 075 case hasan average jet width of sim40 latitude while the e = 00 casehas an average jet width of sim100 latitude

33 Effect of Varying Average Stellar Flux

An increase in average stellar flux leads to increases intemperature and wind speeds illustrated in Figure 8 Each caseshown has an eccentricity of 025 these maps have independentcolorscales to better show atmospheric structure The planetthat receives the highest average stellar flux throughout its orbitis hottest overall (bottom) with peak temperatures of 1300 KFurthermore because the planet with the highest 〈F 〉 orbits at thesmallest orbital distance it has a faster rotation rate and hencea narrower equatorial jet this comparison makes the effect ofrotation on the jet structure much more apparent

34 Global Average Temperature versus Eccentricity and 〈F 〉Here we illustrate the variation in time of maximum

planet-wide temperature with eccentricity and average stellarflux Figure 9 shows the globally averaged temperature foreach model integration plotted as a function of mean anomaly9

where M = 0 corresponds to periapse passage With increas-ing eccentricity the time at which the global-mean temperaturereaches its peak value decreases This is expected as a planet

9 The mean anomaly M is defined as (2πPorb)(t minus τ ) where τ is the timeof periapse passage

with a larger eccentricity will have a higher global-average tem-perature at periapse and hence shorter radiative time constant(τrad prop T minus3 see Showman et al 2011) Figure 10 plots the timeof peak temperature from periapse versus eccentricity Here theinverse relationship between eccentricity and the time of peaktemperature is clear

Figures 9 and 10 serve as a nice summary of the modelintegrations and their relation to observations as temperatureis one of the planetary properties that shapes light curvesat secondary eclipse Observers can use these predictions toestimate the timing of peak flux (and global temperature) basedon a given stellar flux andeccentricity (see Section 4)

35 Synchronous versus Pseudo-synchronous Rotation

Here we explore the sensitivity of the circulation to rota-tion rate by comparing models performed with the pseudo-synchronous rotation rate to otherwise identical models per-formed with the synchronous rotation rate This is illustratedin Figure 11 Here the rotation rate varies by almost a factorof three Both cases still have an equatorial superrotating jetand the overall atmospheric structure is similar However thereare subtle differences in the synchronous case the equatorialeastward winds and high-latitude westward winds are higher inmagnitude (right column) The synchronous case also exhibitsmore latitudinal and longitudinal temperature variation partic-ularly on the dayside this is more apparent at apoapse (bot-tom row) In the pseudo-synchronous case the faster rotationrate serves to homogenize the temperature in longitude becausethe time for rotation to carry air parcels from day to night isshorter The temperature difference from equator to pole is alsolarger because the faster rotation rate inhibits equator-to-poleheat transport Thus while it is unlikely that an eccentric planetsynchronously rotates its star the distinction can lead to notabledifferences in the mean flow

These results can also be used to compare the Hut (1981)formulation shown by the pseudo-synchronous case to for-mulations by other groups As shown in Figure 11 if anotherprescription for rotation rate differed from Hut (1981) by afactor of 2ndash3 (as illustrated by the synchronous case) the qual-itative picture would not look that different However if theprescription differed by a factor of more than 3ndash4 the picturecould substantially differ For example the rotation rate calcu-lated from Ivanov amp Papaloizou (2004) Equation (63) yields avalue of 14 times 106 s a factor of four larger than the Hut (1981)calculation So results using this prescription would look verydifferent

Still the rotation rate of eccentric exoplanets is a completelyunconstrained problem One must be careful then in choosinga formulation noting its assumptions and limitations

36 Momentum Budget and Equatorial Superrotation

In most of our simulations the planet maintains an equatorialeastward jet throughout its orbit Superrotation is commonin three-dimensional circulation models of hot Jupiters (egShowman amp Guillot 2002 Showman et al 2008 2009 Menouamp Rauscher 2009 Rauscher amp Menou 2010 2012 Cooperamp Showman 2005 2006 Dobbs-Dixon amp Lin 2008 Dobbs-Dixon et al 2010 Heng et al 2011) and is also seen withinour own solar system on Venus Titan Jupiter and SaturnWhile superrotation on solar system planets has been extensivelystudied only recently have the mechanisms for superrotation onsynchronously rotating exoplanets been identified (Showman amp

9

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)a= 00802 AU

minus200

minus200

200

200

200

600

600

1000

100014001800

22002600

(m sminus1)

0

500

1000

1500

2000

2500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00802 AU

(K)

550

600

650

700

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00505 AU

minus600

minus600

minus200 minus200

200

200

600

600

1000

140018002200260030003400

(m sminus1)

minus500

0

500

1000

1500

2000

2500

3000

3500

4000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00505 AU

(K)

700

750

800

850

900

950

1000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00318 AU

minus200

minus200

200

200

200 200

600

600

600

10001400

1800220026003000340038004200

5000

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00318 AU

(K)700

800

900

1000

1100

1200

1300

Figure 8 Comparison of simulations with e = 025 with decreasing average stellar flux from bottom to top panels Shown are plots of the orbit-averaged zonal-meanzonal wind (left column) and the wind and temperature profiles at 30 mbar at periapse (right column) Each plot has an independent colorscale The vertical bars inthe right column denote the substellar longitude The snapshots of wind and temperature correspond to model outputs of (from top to bottom) 3706 5006 and 2201Earth days

(A color version of this figure is available in the online journal)

10

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

0 05 1 15 2 25 3 35620

640

660

680

700

720

740ltFgt = 11617 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050e=075

0 05 1 15 2 25 3 35880

890

900

910

920

930

940

950

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

ltFgt = 73680 Wmminus1

e=025e=050

0 05 1 15 2 25 3 35 4960

980

1000

1020

1040

1060

1080

1100

1120

1140

1160ltFgt = 185691 Wmminus1

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

Mean anomaly

e=025e=050e=075

0 05 1 15 2 25 3 35 4 451100

1150

1200

1250

1300

1350ltFgt = 468183 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050

Figure 9 Plots of global average temperature vs mean anomaly for each modelintegration Each plot contains profiles of a given average stellar flux (468183185691 73680 and 11617 W mminus1) with blue green and red representingprofiles for e = 025 050 and 075 respectively

(A color version of this figure is available in the online journal)

03 04 05 06 070

05

1

15

2

25

3

eccentricity

Pea

k o

ffse

t fr

om

per

iap

se (

Ear

th d

ays)

468183 Wmminus1

185691 Wmminus1

73680 Wmminus1

11617 Wmminus1

Figure 10 Time after periapse passage at which the global-average temperaturepeaks (from Figure 9) vs eccentricity for all model integrations The bluegreen red and teal profiles correspond to average stellar fluxes of 468183185691 73680 and 11617 W mminus1 respectively

(A color version of this figure is available in the online journal)

Polvani 2011 hereafter SP2011) As discussed in their paperthe generation of superrotation requires up-gradient momentumtransport that pumps angular momentum from outside the jet towithin it Only transport by waves and eddies can satisfy thisrequirement (Hide 1969)

SP2011 demonstrate that for synchronous hot Jupiters onnear-circular orbits such as HD 189733b and HD 209458bequatorial superrotation is generated by standing planetaryscale Kelvin and Rossby waves at the equator and midlati-tudes respectively that are themselves the dynamical responseto the planetrsquos large dayndashnight heating contrast Because theKelvin waves propagate eastward and the Rossby waves prop-agate westward a phase tilt of the wind vectors emerges fromnorthwest to southeast in the northern hemisphere and south-west to northeast in the southern hemisphere with an over-all shape resembling a chevron pointing eastward This patterncauses meridional eddy angular momentum fluxes that convergeangular momentum onto the equator generating equatorialsuperrotation

Despite the fact that not all our model integrations aresynchronous our results suggest that similar mechanisms arestill at play The chevron shape that SP2011 describe is exactlywhat we noted earlier in our plots of the wind and temperatureprofiles from Figure 5 suggesting that as in SP2011 our modelsexhibit an equatorward flux of eddy angular momentum Thisis demonstrated quantitatively in Figure 12 which shows thezonally averaged meridional flux of relative zonal eddy angularmomentum uprimevprime cos φ at snapshots throughout the orbit (asin Figures 5 and 6) Here uprime and vprime represent the deviationof the zonal and meridional winds respectively from theirzonal averages Positive values indicate a northward flux ofeastward eddy angular momentum and negative values indicatea southward flux of eastward eddy angular momentum Atapoapse the small dayndashnight forcing causes poleward transportof eddy angular momentum at latitudes greater than 50 butequatorward transport of eddy angular momentum equator of50 latitude the amplitudes are weak due to the low insolationat apoapse As the planet moves closer to the star (middlerow left column) the increase in dayndashnight forcing strengthens

11

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)e = 050 pseudominussynchronous rotation

minus200

200

200

600

600

1000

10001400

180022002600

3000

3400

3800

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

synchronous rotation

minus600

minus600

minus200

minus200 minus200

minus200

minus200

200

200

600

10001400180022002600300034003800

4200

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

Figure 11 Comparison of pseudo-synchronous (Trot = 3046 times 105 s) vs synchronous (Trot = 8544 times 105 s) rotation for a case with Teq = 755 K e = 050 anda = 00848 AU Wind and temperature snapshots are taken at periapse (middle row) and apoapse (bottom row) at 5008 Earth days and 5013 Earth days respectivelyThe vertical bar denotes substellar longitude

(A color version of this figure is available in the online journal)

the equatorward flux of eastward eddy angular momentumat latitudes between 0 and 50 in both hemispheres Themomentum flux grows and persists until hours after periapseAngular momentum thus converges onto the equator andmaintains the equatorial jet against westward accelerationscaused by advection friction (if any) and Coriolis forces

These mechanisms also explain the difference in wind speedsbetween the synchronous and pseudo-synchronous cases inFigure 11 The synchronous case has a higher dayndashnight heatingcontrast hence a higher equatorward flux of eddy momentumleading to a stronger narrower equatorial jet Moreover theequatorial jet in the pseudo-synchronous case extends to deeper

12

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=18000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

Figure 12 Eddy momentum flux uprimevprime cos φ throughout the orbit of a model where Teq = 951 K e = 05 and a = 00534 AU Positive values indicate a northwardflux of eastward eddy angular momentum while negative values indicate a southward flux of eastward eddy angular momentum The maximum momentum fluxoccurs hours after periapse passage (f = 72) The snapshots correspond to model outputs from 6103 to 6108 Earth days

(A color version of this figure is available in the online journal)

13

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

Figure 13 Differing orbital viewing geometries explored in this study ω = 180 ω = 270 and ω = 360

(A color version of this figure is available in the online journal)

pressures than that of the synchronous case this is most likelydue to a change in momentum budget

4 OBSERVATIONAL IMPLICATIONS

Observations of transiting exoplanets in circular orbits areknown to be shaped by the atmospheric circulation As wasfirst predicted for hot Jupiters by Showman amp Guillot (2002)and later observationally confirmed for hot Jupiter HD 189733bby Knutson et al (2007) an equatorial superrotating jet causesan eastward displacement of the hot spot from the substellarpoint under appropriate conditions Hence a light curve showsa peak in infrared flux minutes to hours before secondaryeclipse We expect such a situation for eccentric exoplanetsas well However observations of eccentric planets can becomplicated because the flux received by the planet variesthroughout its orbit Previous studies explored such situationsbut have not used three-dimensional models (eg Langton ampLaughlin 2008a 2008b Cowan amp Agol 2011) We can expandon these studies and the three-dimensional study by Lewis et al(2010) by using our SPARC model results

We choose to generate light curves following the proceduresdescribed in Fortney et al (2006) for each model integrationat three values of the argument of periastron (ω) the anglebetween the radius vector to the ascending node and the periapseof the orbit In particular we choose orientations where transitoccurs before periapse (at f = minus90ω = 180) transit occursat apoapse (ω = 270) and transit occurs after periapse (atf = +90ω = 360) (Figure 13) These values reflect thebroad range of viewing geometries of transiting exoplanets forexample HAT-P-2b (e = 052) HAT-P-17b (e = 035) andHD 97658b (e = 013) have ω near 180 WASP-8b (e = 031)HAT-P-31b (e = 025) and HD 80606b (e = 093) have ω near270 and GJ 436b (e = 015) XO-3b (e = 026) and HAT-P-11b (e = 020) have ω near 360 (see exoplaneteu for a full listof values) Note that when transit occurs the nightside is visibleto Earth while the dayside is visible at the time of secondaryeclipse

41 Observational Effects of Varying ω

Using the nominal case (e = 050 a = 00534 AU Teq =951 K) we illustrate the dependence of full-orbit light curves onthe value of ω Figure 14 shows three light curves of the same

model integration but at the differing values of ω described inthe previous section Hence the differences in light curves resultsolely from the differing viewing geometries The planetstarflux ratio is plotted as a function of Earth days from periapsein the warm Spitzer wavelength bands (36 and 45 μm) andJHK bands (126 165 and 220 μm) The solid and dashedlines correspond to transit and secondary eclipse respectivelyFor the light curve where ω = 180 (top panel) the peak ininfrared (IR) flux occurs approximately 6 hr after periapse andbefore secondary eclipse The timing of the peak is due toa combination of the radiative lag of the planet and the factthat the hottest regions rotate into an Earth-facing orientationhours after periapse passage The dayside which is most visiblenear secondary eclipse is hottest after periapse Additionallytemperatures reach peak values at times close to when thosehottest regions are facing Earth This helps explain the highamplitude of the light curve peak in the ω = 180 case relative tothe other two cases These effects can be visualized by plottingthe temperature near the photosphere along the equator in apolar projection (Figure 15) Snapshots throughout the orbit plotthe temperature in a polar projection as a function of pressureand orbital position at a latitudinal slice near the equatorthe solid and dashed lines denote the substellar longitude andEarth-facing longitude respectively Based on the temperaturesat the Earth-facing longitude it is apparent that the peak in IRflux should occur after periapse and before secondary eclipse(see f = 12 and 72) this is indeed the case

For the same model but instead with ω = 270 the peak inIR flux occurs 1 hr after eclipse and periapse In this case thedayside is visible near periapse and therefore we are seeing theplanet increase in temperature (and IR flux) during its closestapproach to the star However the hottest regions are displacedeastward of the substellar point and therefore the hottest regionsface Earth significantly before periapse passage (Figure 15 atf asymp 72) Thus by the time the peak temperatures are reachednear and after periapse the hottest regions are already rotatingout of view from Earth This explains the lower amplitude ofthe light curve peak as compared to the peak of the ω = 180case The IR peak decreases as the planet moves further awayfrom the star along our line of sight A second ramp-up in fluxoccurs approximately 1 day after periapseeclipse this is due tothe hottest regions rotating back into view as the planet movestoward apoapse

14

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015P

lane

tSta

r F

lux

Rat

io

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

Figure 14 Three light curves of the IR flux (expressed as a planetstar fluxratio) vs time throughout the full orbit for a single model integration whereTeq = 951 K e = 050 and a = 00534 AU The flux is plotted in warm Spitzerbandpasses 36 and 45 μm and JHK bands (126 165 220 μm respectively)These three cases only differ by their value of ω shown in Figure 13 The solidlines denote transit while the dotted lines denote secondary eclipse

(A color version of this figure is available in the online journal)

In the case where ω = 360 two distinct peaks are presentmdashasmaller peak that occurs after secondary eclipse and sim3ndash4 hrbefore periapse and a second higher peak that occurs aftertransit The first peak occurs as some of the dayside is visible asthe planet is heated approaching periapse The peak decreasesas the dayside rotates out of view For this geometry the hottestregions are facing away from Earth when they reach their peaktemperatures (Figure 15) The second peak occurs after transitas the hottest regions again rotate into view At this point in theorbit peak temperatures have already occurred and thereforethis peak is lower in amplitude than the other two light curvesHowever because the planet is hotter than it was after secondary

eclipse this second peak has a higher amplitude than the firstone

42 High-e Model Integrations at Varying 〈F 〉If an eccentric planet has a large enough dayndashnight tempera-

ture difference and a sufficiently rapid rotation rate its resultantlight curve can show a periodic rise and fall in flux through-out its orbit The light curves shown in Figure 16 are twoe = 075 model integrations that differ in 〈F 〉 the top panelshows a case where 〈F 〉 = 468183 W mminus2 (Teq = 1199 K)and a = 00385 AU while the bottom panel is a case where〈F 〉 = 185691 W mminus2 (Teq = 951 K) and a = 00611 AU Forboth cases light curves are shown for the ω = 180 geome-try (see Figure 13) The differences in 〈F 〉 lead to differencesin light curve amplitude in the two models the closer hotterplanet exhibits a higher flux amplitude Not surprisingly thetiming of peak IR flux for both models resembles the ω = 180case in Figure 14 The peak in each case occurs after periapseand before eclipse due to the combined effects of the ther-mal lag of the planet and geometry of the system Additionallyboth light curves exhibit a quasi-periodic rise and fall of IR fluxdays after periapse passage This phenomenon called ldquoringingrdquowas also seen in models by Langton amp Laughlin (2008a) andCowan amp Agol (2011) caused by the hottest point of the planetrotating in and out of view from Earth For ringing to occurthe temperature difference from dayside to nightside must belarge and it must survive for multiple planetary rotation periodsIn the models shown in Figure 16 this is aided by the rapidpseudo-synchronous rotation rate associated with an eccentric-ity of 075 (Table 2) For the top case the dayndashnight tempera-ture difference is over sim1300 K while the bottom case variesby sim750 K (Figure 7) The light curves from Figure 14 do notexhibit ringing because the dayndashnight temperature difference islow (sim300ndash400 K)

Because the radiative and dynamical timescales in the modelsby Cowan amp Agol (2011) are calculated by a simple scaling withtemperature the ringing seen in their simulated light curvesoccurs with a period equal to the planetrsquos assumed solid-bodyrotation period in the inertial frame of the winds In realityan eccentric exoplanet approaching apoapse would becomeincreasingly cold the winds would weaken and decrease andhence the ringing should be non-periodic Indeed for ourthree dimensional models the ringing in each case has varyingperiodsmdashthe top case has a ringing period ranging from 03 to04 days (Trot = 035 days) while the bottom panel has a periodranging 07ndash08 days (Trot = 07 days)

Despite these differences this study and the studies byLangton amp Laughlin (2008a) and Cowan amp Agol (2011) illus-trate the influence of atmospheric dynamics on the observationsof eccentric exoplanets The light curve features described herecould be seen in future follow-up observations of eccentric ex-oplanets and observers can use the conclusions drawn aboveto relate the observations to the planetrsquos atmospheric structureFor example if ringing is present in an IR light curve the planetmust have a strong dayndashnight temperature asymmetry and hencewe would expect the planet to have strong superrotation at theequator (Showman amp Polvani 2011) This is probably the casefor highly eccentric (e gt 05) short-period exoplanets such asHD 17156b and HD 80606b If ringing were observed it wouldalso allow an estimate of the sum of the rotation speed and thesuperrotation speed thereby placing constraints on both windspeeds and rotation rates of eccentric hot Jupiters The timing ofpeak IR flux relative to transit and secondary eclipse and also

15

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

f=000deg

(K)800

850

900

950

1000

f=minus1242deg

(K)800

850

900

950

1000f=1242deg

(K)800

850

900

950

1000

f=minus7227deg

(K)800

850

900

950

1000f=7227deg

(K)800

850

900

950

1000

f=minus11988deg

(K)800

850

900

950

1000f=11988deg

(K)800

850

900

950

1000

f=minus18000deg

(K)800

850

900

950

1000

Figure 15 Plots of temperature vs longitude and pressure in equatorial slices throughout one full orbit for our nominal model with Teq = 951 K e = 050 anda = 00534 AU Each annulus shows temperature vs longitude and log-pressure for a snapshot at a particular orbital position (denoted by the value of f) as viewedfrom the planetrsquos north pole Pressures between 30 and 43 mbar are displayed in each annulus The different annuli mark the time evolution throughout one completeorbit starting from apoapse for the top annulus proceeding to periapse and then back to apoapse (follow arrows moving counterclockwise around the plot) Thesesnapshots correspond to model outputs from 6103 to 6108 Earth days The solid line in each snapshot indicates the longitude of the substellar point while the dashedline denotes the Earth-facing longitude for the particular case of ω = 180 For the cases of ω = 270 and 360 degrees exactly the same plot is valid except that thesub-Earth longitude would be aiming upward for ω = 270 and right for ω = 360

(A color version of this figure is available in the online journal)

16

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

e = 075 a = 00385 AU

-15 -10 -05 00 05 10 15Time after periapse passage (days)

0000

0001

0002

0003

0004

Pla

netS

tar

Flu

x R

atio

e = 075 a = 00611 AU

-3 -2 -1 0 1 2 3Time after periapse passage (days)

00000

00005

00010

00015

00020

00025

Pla

netS

tar

Flu

x R

atio

Figure 16 Light Curves for two different e = 075 model integrationsTeq = 1199 K a = 00385 AU (top) and Teq = 951 K a = 00611 AU(bottom) both with ω = 180 The solid line denotes transit while the dottedline denotes secondary eclipse

(A color version of this figure is available in the online journal)

relative to orbital position can help to constrain the circulationfurther

5 CONCLUSIONS

We present three-dimensional circulation models coupledwith a two-stream non-gray radiative transfer scheme fora number of theoretical eccentric hot Jupiters We haveshown that as in published models with zero eccentricity ourhigh-eccentricity circulation models are dominated by eastwardflow at photospheric levels which cause an eastward displace-ment of the hottest regions from the substellar point The rapidrotation rates associated with pseudo-synchronization at higheccentricity lead to a small Rossby deformation radius andin some cases multiple jets in the atmosphere Global-meantemperatures and dayndashnight temperature differences peak notat periapse but several hours afterward due to finite radiativetimescales in the planetrsquos atmosphere

Furthermore we show that equatorial superrotation is gener-ated and maintained by eddies formed by the strong dayndashnightheating contrast which induce a flux of momentum from mid-latitudes to the equator The eddy magnitudes and momentumfluxes peak just after periapse passage leading to variations inthe zonal-mean flow throughout the orbit

Figure 17 Globally averaged pressurendashtemperature profiles from 1-D radiative-convective radiative transfer models of HD 189733b with 11 frequency bins(red profiles) 30 frequency bins (green profiles) and 196 frequency bins (blackprofiles) From left to right the profiles vary in stellar flux from the nominalvalue as well as 2times 4times and 8times higher flux The 8times higher flux case has atemperature inversion

(A color version of this figure is available in the online journal)

Lastly we have shown that the spatial and temporal vari-ations of the wind and temperature structure as well as theorbital viewing geometry of the system with respect to Earthcan affect the time and amplitude of peak IR flux seen infull-orbit light curves Depending on the viewing geometry ofthe orbit relative to Earth we find that peaks in IR flux thateither lead or lag periapse are possible in all cases a combi-nation of temporal effects (temperatures changing over time)and geometric effects (hot spots rotating into or out of view)are important in controlling the timing and amplitude of theflux peaks In cases where the dayndashnight temperature contrastis large and the rotational period is short the light curve canalso exhibit ldquoringingrdquo in flux as the hottest region of the planetrotates in and out of view This ringing is non-periodic due tothe variation in stellar heating as a function of distance

This work was supported by NASA Origins and PlanetaryAtmospheres grants to APS TK also acknowledges supportfrom the Harriet P Jenkins Pre-Doctoral Fellowship Program(JPFP) Resources supporting this work were provided bythe NASA High-End Computing (HEC) Program through theNASA Advanced Supercomputing (NAS) Division at AmesResearch Center The authors thank the anonymous referee fortheir helpful comments and suggestions

APPENDIX

VALIDATION TESTS FOR UPDATEDRADIATIVE TRANSFER SCHEME

To validate our updated radiative transfer scheme using 11frequency bins we conducted the following tests

1 First we tested the sensitivity of the number of frequencybins in a one-dimensional radiative-convective model ofHD 189733b (a = 00313 AU e = 00 RP = 82396 times107 m g = 214 m sminus2) Figure 17 compares the globallyaveraged pressurendashtemperature profiles of models run with11 frequency bins 30 frequency bins and 196 frequencybins at (from left to right) 1times 2times 4times and 8times theaverage stellar flux of HD 189733b For each group of

17

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)30 opacity bins

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)

11 opacity bins

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

30 opacity bins

minus600

minus200minus200

200

200

600

6001000

1000

14001800

2200

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

11 opacity bins

minus600

minus200

minus200

200

200

600

600

10001000

1400

1800

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

30 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

11 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

Figure 18 Globally averaged pressurendashtemperature profiles (top row) zonal-mean zonal wind (middle row) and the windtemperature profiles at 30 mbar (bottomrow) for the 30-bin (left column) and 11-bin (right column) model integrations of HD 189733b Snapshots were output at 250 Earth days

(A color version of this figure is available in the online journal)

models the 11-bin profile agrees well with the 30- and196-bin cases from 1 μbar to 1 bar the temperature variesby only tens of K Below 10 bars the 30- and 11-binprofiles differ from the 196-bin cases by up to 50 Kdue to the loss of resolution at low wavenumbers Inthe case of the highest average flux the large amount of

heating leads to a temperature inversion that is not fullycaptured by the 11-bin model but can be improved witha better initial guess at the pressure-temperature (P-T)profile

2 After testing the new radiative transfer (RT) scheme forone-dimensional models we ran full three-dimensional

18

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

simulations of HD 189733b at its nominal average stellarflux using the SPARCMITgcm Figure 18 shows the glob-ally averaged P-T profiles (top row) zonal-mean zonal wind(middle row) and windtemperature profiles at 30 mbar(bottom row) for the 30-bin (left column) and 11-bin (rightcolumn) RT setups The globally averaged profiles arenearly identical with any temperature differences withina few K The zonal wind plots also show good agree-ment with similar peak speeds (gt2600 m sminus1) equatorialjet width (sim60) and jet level (lt100 bar) At 30 mbarthe 30- and 11-bin models share similar dayside-nightsidetemperature structure with a hot spot (1100 K) eastward ofthe substellar point and two colder regions on the nightsidein the mid-latitudes

Both tests show that the bulk circulation and temperaturestructure is retained in the transition from 30 to 11 frequencybins hence we proceed to use the SPARCMITgcm with thenew scheme for these and future model simulations

REFERENCES

Adcroft A Campin J-M Hill C amp Marshall J 2004 MWRv 132 2845Charbonneau D Knutson H A Barman T et al 2008 ApJ 686 1341Cho J Y-K Menou K Hansen B M S amp Seager S 2008 ApJ

675 817Cooper C S amp Showman A P 2005 ApJL 629 L45Cooper C S amp Showman A P 2006 ApJ 649 1048Cowan N B amp Agol E 2011 ApJ 726 82Deming D Harrington J Laughlin G et al 2007 ApJL 667 L199Dobbs-Dixon I Cumming A amp Lin D N C 2010 ApJ 710 1395Dobbs-Dixon I amp Lin D N C 2008 ApJ 673 513Fortney J J Cooper C S Showman A P Marley M S amp Freedman R S

2006 ApJ 652 746Fortney J J Lodders K Marley M S amp Freedman R S 2008 ApJ

678 1419Fortney J J Marley M S Lodders K Saumon D amp Freedman R

2005 ApJL 627 L69

Fu Q amp Liou K N 1992 JAtS 49 2140Goody R West R Chen L amp Crisp D 1989 JQSRT 42 539Greenberg R 2009 ApJL 698 L42Heng K 2012 ApJL 761 L1Heng K Frierson D M W amp Phillipps P J 2011 MNRAS 418 2669Hide R 1969 JAtS 26 841Holton J R 2004 An Introduction to Dynamic Meteorology (Waltham MA

Academic)Hut P 1981 AampA 99 126Irwin J Charbonneau D Nutzman P et al 2008 ApJ 681 636Ivanov P B amp Papaloizou J C B 2004 MNRAS 353 1161Knutson H A Charbonneau D Allen L E Burrows A amp Megeath S T

2008 ApJ 673 526Knutson H A Charbonneau D Allen L E et al 2007 Natur 447 183Knutson H A Charbonneau D Cowan N B et al 2009 ApJ 690 822Langton J amp Laughlin G 2008a ApJ 674 1106Langton J amp Laughlin G 2008b AampA 483 25Lewis N K Showman A P Fortney J J et al 2010 ApJ 720 344Lodders K 2003 ApJ 591 1220Marley M S amp McKay C P 1999 Icar 138 268Mayor M amp Queloz D 1995 Natur 378 355Menou K amp Rauscher E 2009 ApJ 700 887Mlawer E J Taubman S J Brown P D Iacono M J amp Clough S A

1997 JGR 102 16663Naef D Latham D W Mayor M et al 2001 AampA 375 L27Perna R Heng K amp Pont F 2012 ApJ 751 59Pont F Knutson H Gilliland R L Moutou C amp Charbonneau D

2008 MNRAS 385 109Rauscher E amp Menou K 2010 ApJ 714 1334Rauscher E amp Menou K 2012 ApJ 750 96Saumon D Marley M S Abel M Frommhold L amp Freedman R S

2012 ApJ 750 74Showman A P Cho J Y-K amp Menou K 2011 in Exoplanets ed S Seager

(Tucson AZ Univ Arizona Press) 471Showman A P Cooper C S Fortney J J amp Marley M S 2008 ApJ

682 559Showman A P Fortney J J Lian Y et al 2009 ApJ 699 564Showman A P amp Guillot T 2002 AampA 385 166Showman A P amp Polvani L M 2011 ApJ 738 71Swain M R Vasisht G amp Tinetti G 2008 Natur 452 329Thrastarson H T amp Cho J Y 2010 ApJ 716 144Wolszczan A 1994 Sci 264 538

19

  • 1 INTRODUCTION
  • 2 MODEL
    • 21 Updates to SPARCMITgcm
    • 22 Grid Simulations
      • 3 RESULTS
        • 31 Circulation Regime
        • 32 Effect of Eccentricity
        • 33 Effect of Varying Average Stellar Flux
        • 34 Global Average Temperature versus Eccentricity and F
        • 35 Synchronous versus Pseudo-synchronous Rotation
        • 36 Momentum Budget and Equatorial Superrotation
          • 4 OBSERVATIONAL IMPLICATIONS
            • 41 Observational Effects of Varying ω
            • 42 High-e Model Integrations at Varying F
              • 5 CONCLUSIONS
              • APPENDIX VALIDATION TESTS FOR UPDATED RADIATIVE TRANSFER SCHEME
              • REFERENCES
Page 4: THREE-DIMENSIONAL ATMOSPHERIC CIRCULATION OF HOT …showman/publications/kataria-etal-2013.pdf · planets’ atmospheric composition, temperature structure, and circulation. A fifth

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

0 500 1000 1500 2000 2500

10minus6

10minus4

10minus2

100

102

104

Temperature (K)

Pre

ssu

re (

bar

s)

11617 W mminus2

73680 W mminus2

185691 W mminus2

468183 W mminus2

Figure 2 Initial 1-D pressurendashtemperature profiles assigned to each column ofthe atmosphere at the start of each simulation at the four different valuesof average stellar flux These profiles were calculated assuming radiativeequilibrium and use the same opacities as the full 3D model

(A color version of this figure is available in the online journal)

at each temperature and pressure over the 3D grid assumingthe elemental abundances of Lodders (2003) We set cp and κto values appropriate for a H2-dominated atmosphere (13 times104 J kgminus1 Kminus1 and 27 respectively)

For eccentric cases we assume the planet is pseudo-synchronously rotating its host starmdashwe assume the planetrsquostidal interactions with the star force the same side of the planetto approximately face the star every periapse passage We cal-

culate this using the Hut (1981) formulation

Prot = Porb

(1 + 3e2 + 3

8e4)

(1 minus e2)32

1 + 152 e2 + 45

8 e4 + 516e6

(5)

Pseudo-synchronization is a commonly used assumption inthe calculation of the rotation rate of eccentric exoplanets (egDeming et al 2007 Irwin et al 2008 Langton amp Laughlin2008a 2008b Lewis et al 2010) Nevertheless the true depen-dence of rotation period on orbital period and eccentricity isvery uncertain (eg Greenberg 2009) and other formulationsbesides Equation (5) have also been suggested in the litera-ture (eg Ivanov amp Papaloizou 2004) We explore the sensitiv-ity of the flow to the assumption of pseudo-synchronization inSection 34

In each integration we assume the winds to be initiallyzero with each vertical atmospheric column assigned the sameglobal-mean radiative equilibrium pressurendashtemperature profilecalculated using the same opacities as the main model (Figure 2)These temperaturendashpressure profiles are calculated for eachvalue of average stellar flux (see above) using the methodsdescribed in Fortney et al (2005 2008)

Most model integrations were run until the winds reacheda statistically steady configuration at upper levels Figure 3shows the rms velocity plotted as a function of pressure andsimulated time for a model integrated for gt6000 Earth daysThis behavior in rms velocity is typical of most model inte-grations Weak drag was applied to the atmosphere with atime constant of 100 days at 200 bars with a drag coeffi-cient that decreases linearly with pressure to zero at 10 barsTests were conducted to explore the sensitivity of drag on themean flow Extending the drag up to 1 bar affects the verti-cal and longitudinal extent of the equatorial jet but not the

0 1000 2000 3000 4000 5000 6000

10minus3

10minus2

10minus1

100

101

102

simulated time (Earth days)

pre

ssu

re (

bar

)

e= 050 a= 00534 AU

00003217611

0000321

200

200 200

400

400 400

600

600 600

800

800 800

1000

1000 1000

1200

1200 1200

1400

1400 1400

1600

16001600

1800

18001800

2000

2000

2000

2200

2200

2200

2400

2400

2600

260028003000

(m sminus1)0

500

1000

1500

2000

2500

3000

Figure 3 rms velocity plotted as a function of pressure and simulation time for one of our model integrations The plot shows a case where Teq = 951 K e = 050and a = 00534 AU rms velocity was calculated at each pressure level from wind speeds output every 100 days The rapid decrease in velocity at the start of the runis merely an artifact at the start of the run

(A color version of this figure is available in the online journal)

4

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus180 minus90 0 90 180True anomaly

minus2 minus15 minus1 minus05 0 05 1 15 2

10minus3

10minus2

10minus1

100

101

102

simulated time (Earth days)

pre

ssu

re (

bar

) 700700

800800

800

900

900

900

1000

10001000

11001100

1100

12001200 1200

1300 1300 1300

1400 1400 1400

1500 1500 1500

T(K)

700

800

900

1000

1100

1200

1300

1400

1500

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00534 AU

minus200

minus200

200

200

600

600

1000

1000

1400

18002200

2600300034003800

(m sminus1)

0

1000

2000

3000

4000

5000

Figure 4 Dynamical response of a model with Teq = 951 K e = 050 anda = 00534 AU Top global horizontally averaged temperature vs pressureand time throughout the full orbit zero is defined as periapse passage The upperaxis gives the true anomaly throughout the orbit Bottom zonally averaged zonal(eastndashwest) winds vs pressure and latitude for the same model time-averagedover a full orbit

(A color version of this figure is available in the online journal)

qualitative nature of the circulation We reiterate that inte-grations without this drag scheme (but including our fourth-order Shapiro filter) conserve angular momentum to betterthan 01

3 RESULTS

31 Circulation Regime

To illustrate the circulation regime of an eccentric exoplanetwe present a nominal model where e = 050 a = 00534 AUand Teq = 951 K representative of the simulations (Figure 4)The top panel shows the horizontally averaged global temper-ature as a function of pressure time (in Earth days) and trueanomaly f where f = 0 at periapse and f = plusmn180 atapoapse Temperatures remain fairly constant at pressures ex-ceeding 1 bar However in the uppermost layers of the atmo-sphere temperatures vary throughout the orbit peaking sim6 hr

after periapse passage The bottom panel shows the zonal-meanzonal wind8 of the same model time-averaged over a full or-bit The flow exhibits a superrotating (eastward) equatorial jetwhich is the dominant feature of the flow with peak winds ofabout 5000 m sminus1 Superrotation is commonly seen in models ofhot Jupiters and hot Neptunes (eg Showman amp Guillot 2002Cooper amp Showman 2005 Showman et al 2008 2009 Dobbs-Dixon amp Lin 2008 Dobbs-Dixon et al 2010 Lewis et al 2010Heng et al 2011 Perna et al 2012 Rauscher amp Menou 20102012) although not by all groups (Cho et al 2008 Thrastarsonamp Cho 2010)

With peak winds exceeding 2ndash5 km sminus1 the equatorial jet issupersonic as the sound speed for this class of planets is ap-proximately 2 km sminus1 As Heng (2012) shows the presence ofthis supersonic flow generates shocks at the dayndashnight termi-nator upstream of the substellar point The SPARCMITgcmdoes not properly account for the removal of kinetic energy as aresult of these shocks The formation of shocks is expected to beshallow (ie at low pressure) in both of the model integrationsthat Heng (2012) presents the Mach number falls below unityat pressures greater than 1 bar and hence shocks are not formedNevertheless the infrared photosphere and the peak jet speedsin our model occur at sufficiently low pressure that shocks couldplay some role particularly when the orbital eccentricity is highMoreover as a possible energy dissipation mechanism shockscould have a non-local effect (ie they could cause an indirecteffect on the flow even at pressures deeper than those where theshocks directly occur) Clearly more work is warranted on therole of shocks in hot Jupiter atmospheres

Snapshots of the wind and temperature profile throughout theorbit at a pressure level of 30 mbar (approximately the levelof the infrared photosphere) show that the planet undergoes aflash heating event as the planet approaches periapse passage(Figure 5) The eight panels show snapshots throughout onecomplete orbit from apoapse (top figure) to hours after periapse(f = 119 follow arrows) and back again to apoapse all withthe same colorscale for temperature Like the circulation ofHD 189733b in its nominally circular orbit (see Showman et al2009) the flow maintains the eastward equatorial superrotatingjet throughout its orbit (Figure 6) As the planet approachesperiapse (Figure 5 f = minus12) temperatures rise to sim1000 Kat the equator and a high-amplitude eddy structure developswith a pattern of temperature and wind vectors that tilt eastwardtoward the equator This so-called ldquochevronrdquo-shaped feature isprominent at periapse (bottom panel) extends over a wide rangeof latitude and longitude and persists hours after periapse Thechevron shape is also seen in our other integrations and indicatesthat eddies transport momentum equatorward (see Section 36)

32 Effect of Eccentricity

Our models show that at constant orbit-mean stellar fluxthe dayside temperatures dayndashnight temperature differencesand wind speeds at periapse passage all increase with increas-ing orbital eccentricity This is illustrated in Figure 7 whichcompares three simulations with an equilibrium temperature of951 K (〈F 〉 = 185691 W mminus2) that vary in eccentricity from00 (top row) to 075 (bottom row) Again we plot each columnon the same colorscale (zonal wind and temperature) for com-parison As eccentricity increases peak temperatures increase

8 Zonal wind is defined as eastndashwest winds with positive (negative) valuesdenoting eastward (westward) winds a zonal average is an average inlongitude

5

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=000deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus1242deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=1242deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus7227deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=7227deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus11988deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=11988deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus18000deg

(K)500

600

700

800

900

1000

1100

1200

Figure 5 Snapshots of the wind and temperature structure at 30 mbar of a planet with Teq = 951 K e = 050 and a = 00534 AU To trace the behavior throughoutone complete orbit the different panels should be followed counterclockwise from the top panel Snapshots are taken from apoapse (true anomaly f = minus1800top panel) to hours after periapse (f = 119 follow arrows) and back again to apoapse Each panel is plotted on the same temperature colorscale Snapshots weretaken between 6103 and 6108 Earth days Note that peak temperatures lag periapse passage Periapse corresponds to f = 0 and apoapse to 180 The vertical linecorresponds to substellar longitude

(A color version of this figure is available in the online journal)

6

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=000deg

minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus1242deg

minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=1242deg

minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus7227deg

minus200

minus200 200

200

200

600

600

1000

1000

1400

1400

18002200260030003400

3800

4200

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=7227deg

minus200

minus200minus200 200

200

200

600

600

1000

1000

1400

180022

002600300034003800

4200

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus11988deg

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=11988deg

minus600minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

1800

2200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus18000deg

minus200

minus200

200

200

600

600

1000

1000

1400

1800

2200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

Figure 6 Snapshots of the zonal-mean zonal wind for the same integration described in Figure 5 (Teq = 951 K e = 050 and a = 00534 AU) Snapshots are againtaken from apoapse (f = minus1800) to hours after periapse (f = 119) and back again to apoapse Each panel is plotted on the same temperature colorscale Thesnapshots correspond to outputs from 6103 to 6108 Earth days Note that the equatorial superrotating jet is persistent throughout the orbit with peak speeds exceeding5000 m sminus1

(A color version of this figure is available in the online journal)

7

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 000

200

200

200600

600

600

1000

1000

10001400

1400

18002200

2600

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 000

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 025

minus600

minus600

minus200 minus200

200

200

600

600

1000

140018002200260030003400

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 025

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 050

minus200

minus200

200

200

600

600

1000

1000

1400

18002200

2600300034003800

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 050

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 075

200

200

200

200

600

600

600

1000

1000

1000

140018

002200260030003400

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 075

(K)

500

600

700

800

900

1000

1100

1200

1300

Figure 7 Comparison of simulations with Teq = 951 K with increasing eccentricity from e = 00 (top row) to e = 075 (bottom row) Shown are plots of thezonal-mean zonal wind averaged over a full orbit (left column) and snapshots of the wind and temperature profiles at 30 mbar (right column) The snapshots of thewind and temperature profiles in the right column were taken at periapse passage The vertical bars in the right column denote the longitude of the substellar pointThe snapshots of wind and temperature correspond to model outputs of (from top to bottom) 4100 5006 6106 and 6904 Earth days

(A color version of this figure is available in the online journal)

8

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

from 1000 K to 1300 K In addition the variation in temper-ature from dayside to nightside increases this strengthens thewinds within the equatorial superrotating jet Peak winds reachsim2500 m sminus1 in the circular case but increase to sim5000 m sminus1

in the high eccentricity caseAs the eccentricity increases the equatorial jet also narrows

in latitude In the case of highest eccentricity the atmospheredevelops westward midlatitude jets and eastward high-latitudejets The narrowing of the equatorial jet arises from the factthat jet width is confined to the Rossby radius of deformation(Showman amp Polvani 2011) the length scale over which theatmosphere adjusts to planetary-scale phenomena The Rossbyradius of deformation (Holton 2004) is defined at the equator as

LR =radic

c

β (6)

where c is the gravity wave speed and β is the derivative ofthe Coriolis parameter f with respect to latitudinal distancey β = df dy = 2Ωrot cos φRp where Ωrot is the planetaryrotation rate φ is the latitude and Rp is the planetary radius Atthe equator β is simply 2ΩrotRp Hence at the equator LR is

LR =radic

cRp

2Ωrot (7)

Therefore LR prop 1radic

Ωrot and an increased rotation rateleads to a narrower deformation radius From the Hut (1981)formulation the combined effect of the larger orbital period andincrease in eccentricity produces a shorter rotational periodThis in turn increases the rotation rate which reduces thedeformation radius For example the ratio between Ωrot forthe e = 00 case and the e = 075 case is 63 hence we wouldexpect that the equatorial jet should be narrower by a factor ofradic

63 sim 25 This is indeed the case as the e = 075 case hasan average jet width of sim40 latitude while the e = 00 casehas an average jet width of sim100 latitude

33 Effect of Varying Average Stellar Flux

An increase in average stellar flux leads to increases intemperature and wind speeds illustrated in Figure 8 Each caseshown has an eccentricity of 025 these maps have independentcolorscales to better show atmospheric structure The planetthat receives the highest average stellar flux throughout its orbitis hottest overall (bottom) with peak temperatures of 1300 KFurthermore because the planet with the highest 〈F 〉 orbits at thesmallest orbital distance it has a faster rotation rate and hencea narrower equatorial jet this comparison makes the effect ofrotation on the jet structure much more apparent

34 Global Average Temperature versus Eccentricity and 〈F 〉Here we illustrate the variation in time of maximum

planet-wide temperature with eccentricity and average stellarflux Figure 9 shows the globally averaged temperature foreach model integration plotted as a function of mean anomaly9

where M = 0 corresponds to periapse passage With increas-ing eccentricity the time at which the global-mean temperaturereaches its peak value decreases This is expected as a planet

9 The mean anomaly M is defined as (2πPorb)(t minus τ ) where τ is the timeof periapse passage

with a larger eccentricity will have a higher global-average tem-perature at periapse and hence shorter radiative time constant(τrad prop T minus3 see Showman et al 2011) Figure 10 plots the timeof peak temperature from periapse versus eccentricity Here theinverse relationship between eccentricity and the time of peaktemperature is clear

Figures 9 and 10 serve as a nice summary of the modelintegrations and their relation to observations as temperatureis one of the planetary properties that shapes light curvesat secondary eclipse Observers can use these predictions toestimate the timing of peak flux (and global temperature) basedon a given stellar flux andeccentricity (see Section 4)

35 Synchronous versus Pseudo-synchronous Rotation

Here we explore the sensitivity of the circulation to rota-tion rate by comparing models performed with the pseudo-synchronous rotation rate to otherwise identical models per-formed with the synchronous rotation rate This is illustratedin Figure 11 Here the rotation rate varies by almost a factorof three Both cases still have an equatorial superrotating jetand the overall atmospheric structure is similar However thereare subtle differences in the synchronous case the equatorialeastward winds and high-latitude westward winds are higher inmagnitude (right column) The synchronous case also exhibitsmore latitudinal and longitudinal temperature variation partic-ularly on the dayside this is more apparent at apoapse (bot-tom row) In the pseudo-synchronous case the faster rotationrate serves to homogenize the temperature in longitude becausethe time for rotation to carry air parcels from day to night isshorter The temperature difference from equator to pole is alsolarger because the faster rotation rate inhibits equator-to-poleheat transport Thus while it is unlikely that an eccentric planetsynchronously rotates its star the distinction can lead to notabledifferences in the mean flow

These results can also be used to compare the Hut (1981)formulation shown by the pseudo-synchronous case to for-mulations by other groups As shown in Figure 11 if anotherprescription for rotation rate differed from Hut (1981) by afactor of 2ndash3 (as illustrated by the synchronous case) the qual-itative picture would not look that different However if theprescription differed by a factor of more than 3ndash4 the picturecould substantially differ For example the rotation rate calcu-lated from Ivanov amp Papaloizou (2004) Equation (63) yields avalue of 14 times 106 s a factor of four larger than the Hut (1981)calculation So results using this prescription would look verydifferent

Still the rotation rate of eccentric exoplanets is a completelyunconstrained problem One must be careful then in choosinga formulation noting its assumptions and limitations

36 Momentum Budget and Equatorial Superrotation

In most of our simulations the planet maintains an equatorialeastward jet throughout its orbit Superrotation is commonin three-dimensional circulation models of hot Jupiters (egShowman amp Guillot 2002 Showman et al 2008 2009 Menouamp Rauscher 2009 Rauscher amp Menou 2010 2012 Cooperamp Showman 2005 2006 Dobbs-Dixon amp Lin 2008 Dobbs-Dixon et al 2010 Heng et al 2011) and is also seen withinour own solar system on Venus Titan Jupiter and SaturnWhile superrotation on solar system planets has been extensivelystudied only recently have the mechanisms for superrotation onsynchronously rotating exoplanets been identified (Showman amp

9

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)a= 00802 AU

minus200

minus200

200

200

200

600

600

1000

100014001800

22002600

(m sminus1)

0

500

1000

1500

2000

2500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00802 AU

(K)

550

600

650

700

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00505 AU

minus600

minus600

minus200 minus200

200

200

600

600

1000

140018002200260030003400

(m sminus1)

minus500

0

500

1000

1500

2000

2500

3000

3500

4000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00505 AU

(K)

700

750

800

850

900

950

1000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00318 AU

minus200

minus200

200

200

200 200

600

600

600

10001400

1800220026003000340038004200

5000

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00318 AU

(K)700

800

900

1000

1100

1200

1300

Figure 8 Comparison of simulations with e = 025 with decreasing average stellar flux from bottom to top panels Shown are plots of the orbit-averaged zonal-meanzonal wind (left column) and the wind and temperature profiles at 30 mbar at periapse (right column) Each plot has an independent colorscale The vertical bars inthe right column denote the substellar longitude The snapshots of wind and temperature correspond to model outputs of (from top to bottom) 3706 5006 and 2201Earth days

(A color version of this figure is available in the online journal)

10

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

0 05 1 15 2 25 3 35620

640

660

680

700

720

740ltFgt = 11617 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050e=075

0 05 1 15 2 25 3 35880

890

900

910

920

930

940

950

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

ltFgt = 73680 Wmminus1

e=025e=050

0 05 1 15 2 25 3 35 4960

980

1000

1020

1040

1060

1080

1100

1120

1140

1160ltFgt = 185691 Wmminus1

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

Mean anomaly

e=025e=050e=075

0 05 1 15 2 25 3 35 4 451100

1150

1200

1250

1300

1350ltFgt = 468183 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050

Figure 9 Plots of global average temperature vs mean anomaly for each modelintegration Each plot contains profiles of a given average stellar flux (468183185691 73680 and 11617 W mminus1) with blue green and red representingprofiles for e = 025 050 and 075 respectively

(A color version of this figure is available in the online journal)

03 04 05 06 070

05

1

15

2

25

3

eccentricity

Pea

k o

ffse

t fr

om

per

iap

se (

Ear

th d

ays)

468183 Wmminus1

185691 Wmminus1

73680 Wmminus1

11617 Wmminus1

Figure 10 Time after periapse passage at which the global-average temperaturepeaks (from Figure 9) vs eccentricity for all model integrations The bluegreen red and teal profiles correspond to average stellar fluxes of 468183185691 73680 and 11617 W mminus1 respectively

(A color version of this figure is available in the online journal)

Polvani 2011 hereafter SP2011) As discussed in their paperthe generation of superrotation requires up-gradient momentumtransport that pumps angular momentum from outside the jet towithin it Only transport by waves and eddies can satisfy thisrequirement (Hide 1969)

SP2011 demonstrate that for synchronous hot Jupiters onnear-circular orbits such as HD 189733b and HD 209458bequatorial superrotation is generated by standing planetaryscale Kelvin and Rossby waves at the equator and midlati-tudes respectively that are themselves the dynamical responseto the planetrsquos large dayndashnight heating contrast Because theKelvin waves propagate eastward and the Rossby waves prop-agate westward a phase tilt of the wind vectors emerges fromnorthwest to southeast in the northern hemisphere and south-west to northeast in the southern hemisphere with an over-all shape resembling a chevron pointing eastward This patterncauses meridional eddy angular momentum fluxes that convergeangular momentum onto the equator generating equatorialsuperrotation

Despite the fact that not all our model integrations aresynchronous our results suggest that similar mechanisms arestill at play The chevron shape that SP2011 describe is exactlywhat we noted earlier in our plots of the wind and temperatureprofiles from Figure 5 suggesting that as in SP2011 our modelsexhibit an equatorward flux of eddy angular momentum Thisis demonstrated quantitatively in Figure 12 which shows thezonally averaged meridional flux of relative zonal eddy angularmomentum uprimevprime cos φ at snapshots throughout the orbit (asin Figures 5 and 6) Here uprime and vprime represent the deviationof the zonal and meridional winds respectively from theirzonal averages Positive values indicate a northward flux ofeastward eddy angular momentum and negative values indicatea southward flux of eastward eddy angular momentum Atapoapse the small dayndashnight forcing causes poleward transportof eddy angular momentum at latitudes greater than 50 butequatorward transport of eddy angular momentum equator of50 latitude the amplitudes are weak due to the low insolationat apoapse As the planet moves closer to the star (middlerow left column) the increase in dayndashnight forcing strengthens

11

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)e = 050 pseudominussynchronous rotation

minus200

200

200

600

600

1000

10001400

180022002600

3000

3400

3800

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

synchronous rotation

minus600

minus600

minus200

minus200 minus200

minus200

minus200

200

200

600

10001400180022002600300034003800

4200

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

Figure 11 Comparison of pseudo-synchronous (Trot = 3046 times 105 s) vs synchronous (Trot = 8544 times 105 s) rotation for a case with Teq = 755 K e = 050 anda = 00848 AU Wind and temperature snapshots are taken at periapse (middle row) and apoapse (bottom row) at 5008 Earth days and 5013 Earth days respectivelyThe vertical bar denotes substellar longitude

(A color version of this figure is available in the online journal)

the equatorward flux of eastward eddy angular momentumat latitudes between 0 and 50 in both hemispheres Themomentum flux grows and persists until hours after periapseAngular momentum thus converges onto the equator andmaintains the equatorial jet against westward accelerationscaused by advection friction (if any) and Coriolis forces

These mechanisms also explain the difference in wind speedsbetween the synchronous and pseudo-synchronous cases inFigure 11 The synchronous case has a higher dayndashnight heatingcontrast hence a higher equatorward flux of eddy momentumleading to a stronger narrower equatorial jet Moreover theequatorial jet in the pseudo-synchronous case extends to deeper

12

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=18000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

Figure 12 Eddy momentum flux uprimevprime cos φ throughout the orbit of a model where Teq = 951 K e = 05 and a = 00534 AU Positive values indicate a northwardflux of eastward eddy angular momentum while negative values indicate a southward flux of eastward eddy angular momentum The maximum momentum fluxoccurs hours after periapse passage (f = 72) The snapshots correspond to model outputs from 6103 to 6108 Earth days

(A color version of this figure is available in the online journal)

13

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

Figure 13 Differing orbital viewing geometries explored in this study ω = 180 ω = 270 and ω = 360

(A color version of this figure is available in the online journal)

pressures than that of the synchronous case this is most likelydue to a change in momentum budget

4 OBSERVATIONAL IMPLICATIONS

Observations of transiting exoplanets in circular orbits areknown to be shaped by the atmospheric circulation As wasfirst predicted for hot Jupiters by Showman amp Guillot (2002)and later observationally confirmed for hot Jupiter HD 189733bby Knutson et al (2007) an equatorial superrotating jet causesan eastward displacement of the hot spot from the substellarpoint under appropriate conditions Hence a light curve showsa peak in infrared flux minutes to hours before secondaryeclipse We expect such a situation for eccentric exoplanetsas well However observations of eccentric planets can becomplicated because the flux received by the planet variesthroughout its orbit Previous studies explored such situationsbut have not used three-dimensional models (eg Langton ampLaughlin 2008a 2008b Cowan amp Agol 2011) We can expandon these studies and the three-dimensional study by Lewis et al(2010) by using our SPARC model results

We choose to generate light curves following the proceduresdescribed in Fortney et al (2006) for each model integrationat three values of the argument of periastron (ω) the anglebetween the radius vector to the ascending node and the periapseof the orbit In particular we choose orientations where transitoccurs before periapse (at f = minus90ω = 180) transit occursat apoapse (ω = 270) and transit occurs after periapse (atf = +90ω = 360) (Figure 13) These values reflect thebroad range of viewing geometries of transiting exoplanets forexample HAT-P-2b (e = 052) HAT-P-17b (e = 035) andHD 97658b (e = 013) have ω near 180 WASP-8b (e = 031)HAT-P-31b (e = 025) and HD 80606b (e = 093) have ω near270 and GJ 436b (e = 015) XO-3b (e = 026) and HAT-P-11b (e = 020) have ω near 360 (see exoplaneteu for a full listof values) Note that when transit occurs the nightside is visibleto Earth while the dayside is visible at the time of secondaryeclipse

41 Observational Effects of Varying ω

Using the nominal case (e = 050 a = 00534 AU Teq =951 K) we illustrate the dependence of full-orbit light curves onthe value of ω Figure 14 shows three light curves of the same

model integration but at the differing values of ω described inthe previous section Hence the differences in light curves resultsolely from the differing viewing geometries The planetstarflux ratio is plotted as a function of Earth days from periapsein the warm Spitzer wavelength bands (36 and 45 μm) andJHK bands (126 165 and 220 μm) The solid and dashedlines correspond to transit and secondary eclipse respectivelyFor the light curve where ω = 180 (top panel) the peak ininfrared (IR) flux occurs approximately 6 hr after periapse andbefore secondary eclipse The timing of the peak is due toa combination of the radiative lag of the planet and the factthat the hottest regions rotate into an Earth-facing orientationhours after periapse passage The dayside which is most visiblenear secondary eclipse is hottest after periapse Additionallytemperatures reach peak values at times close to when thosehottest regions are facing Earth This helps explain the highamplitude of the light curve peak in the ω = 180 case relative tothe other two cases These effects can be visualized by plottingthe temperature near the photosphere along the equator in apolar projection (Figure 15) Snapshots throughout the orbit plotthe temperature in a polar projection as a function of pressureand orbital position at a latitudinal slice near the equatorthe solid and dashed lines denote the substellar longitude andEarth-facing longitude respectively Based on the temperaturesat the Earth-facing longitude it is apparent that the peak in IRflux should occur after periapse and before secondary eclipse(see f = 12 and 72) this is indeed the case

For the same model but instead with ω = 270 the peak inIR flux occurs 1 hr after eclipse and periapse In this case thedayside is visible near periapse and therefore we are seeing theplanet increase in temperature (and IR flux) during its closestapproach to the star However the hottest regions are displacedeastward of the substellar point and therefore the hottest regionsface Earth significantly before periapse passage (Figure 15 atf asymp 72) Thus by the time the peak temperatures are reachednear and after periapse the hottest regions are already rotatingout of view from Earth This explains the lower amplitude ofthe light curve peak as compared to the peak of the ω = 180case The IR peak decreases as the planet moves further awayfrom the star along our line of sight A second ramp-up in fluxoccurs approximately 1 day after periapseeclipse this is due tothe hottest regions rotating back into view as the planet movestoward apoapse

14

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015P

lane

tSta

r F

lux

Rat

io

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

Figure 14 Three light curves of the IR flux (expressed as a planetstar fluxratio) vs time throughout the full orbit for a single model integration whereTeq = 951 K e = 050 and a = 00534 AU The flux is plotted in warm Spitzerbandpasses 36 and 45 μm and JHK bands (126 165 220 μm respectively)These three cases only differ by their value of ω shown in Figure 13 The solidlines denote transit while the dotted lines denote secondary eclipse

(A color version of this figure is available in the online journal)

In the case where ω = 360 two distinct peaks are presentmdashasmaller peak that occurs after secondary eclipse and sim3ndash4 hrbefore periapse and a second higher peak that occurs aftertransit The first peak occurs as some of the dayside is visible asthe planet is heated approaching periapse The peak decreasesas the dayside rotates out of view For this geometry the hottestregions are facing away from Earth when they reach their peaktemperatures (Figure 15) The second peak occurs after transitas the hottest regions again rotate into view At this point in theorbit peak temperatures have already occurred and thereforethis peak is lower in amplitude than the other two light curvesHowever because the planet is hotter than it was after secondary

eclipse this second peak has a higher amplitude than the firstone

42 High-e Model Integrations at Varying 〈F 〉If an eccentric planet has a large enough dayndashnight tempera-

ture difference and a sufficiently rapid rotation rate its resultantlight curve can show a periodic rise and fall in flux through-out its orbit The light curves shown in Figure 16 are twoe = 075 model integrations that differ in 〈F 〉 the top panelshows a case where 〈F 〉 = 468183 W mminus2 (Teq = 1199 K)and a = 00385 AU while the bottom panel is a case where〈F 〉 = 185691 W mminus2 (Teq = 951 K) and a = 00611 AU Forboth cases light curves are shown for the ω = 180 geome-try (see Figure 13) The differences in 〈F 〉 lead to differencesin light curve amplitude in the two models the closer hotterplanet exhibits a higher flux amplitude Not surprisingly thetiming of peak IR flux for both models resembles the ω = 180case in Figure 14 The peak in each case occurs after periapseand before eclipse due to the combined effects of the ther-mal lag of the planet and geometry of the system Additionallyboth light curves exhibit a quasi-periodic rise and fall of IR fluxdays after periapse passage This phenomenon called ldquoringingrdquowas also seen in models by Langton amp Laughlin (2008a) andCowan amp Agol (2011) caused by the hottest point of the planetrotating in and out of view from Earth For ringing to occurthe temperature difference from dayside to nightside must belarge and it must survive for multiple planetary rotation periodsIn the models shown in Figure 16 this is aided by the rapidpseudo-synchronous rotation rate associated with an eccentric-ity of 075 (Table 2) For the top case the dayndashnight tempera-ture difference is over sim1300 K while the bottom case variesby sim750 K (Figure 7) The light curves from Figure 14 do notexhibit ringing because the dayndashnight temperature difference islow (sim300ndash400 K)

Because the radiative and dynamical timescales in the modelsby Cowan amp Agol (2011) are calculated by a simple scaling withtemperature the ringing seen in their simulated light curvesoccurs with a period equal to the planetrsquos assumed solid-bodyrotation period in the inertial frame of the winds In realityan eccentric exoplanet approaching apoapse would becomeincreasingly cold the winds would weaken and decrease andhence the ringing should be non-periodic Indeed for ourthree dimensional models the ringing in each case has varyingperiodsmdashthe top case has a ringing period ranging from 03 to04 days (Trot = 035 days) while the bottom panel has a periodranging 07ndash08 days (Trot = 07 days)

Despite these differences this study and the studies byLangton amp Laughlin (2008a) and Cowan amp Agol (2011) illus-trate the influence of atmospheric dynamics on the observationsof eccentric exoplanets The light curve features described herecould be seen in future follow-up observations of eccentric ex-oplanets and observers can use the conclusions drawn aboveto relate the observations to the planetrsquos atmospheric structureFor example if ringing is present in an IR light curve the planetmust have a strong dayndashnight temperature asymmetry and hencewe would expect the planet to have strong superrotation at theequator (Showman amp Polvani 2011) This is probably the casefor highly eccentric (e gt 05) short-period exoplanets such asHD 17156b and HD 80606b If ringing were observed it wouldalso allow an estimate of the sum of the rotation speed and thesuperrotation speed thereby placing constraints on both windspeeds and rotation rates of eccentric hot Jupiters The timing ofpeak IR flux relative to transit and secondary eclipse and also

15

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

f=000deg

(K)800

850

900

950

1000

f=minus1242deg

(K)800

850

900

950

1000f=1242deg

(K)800

850

900

950

1000

f=minus7227deg

(K)800

850

900

950

1000f=7227deg

(K)800

850

900

950

1000

f=minus11988deg

(K)800

850

900

950

1000f=11988deg

(K)800

850

900

950

1000

f=minus18000deg

(K)800

850

900

950

1000

Figure 15 Plots of temperature vs longitude and pressure in equatorial slices throughout one full orbit for our nominal model with Teq = 951 K e = 050 anda = 00534 AU Each annulus shows temperature vs longitude and log-pressure for a snapshot at a particular orbital position (denoted by the value of f) as viewedfrom the planetrsquos north pole Pressures between 30 and 43 mbar are displayed in each annulus The different annuli mark the time evolution throughout one completeorbit starting from apoapse for the top annulus proceeding to periapse and then back to apoapse (follow arrows moving counterclockwise around the plot) Thesesnapshots correspond to model outputs from 6103 to 6108 Earth days The solid line in each snapshot indicates the longitude of the substellar point while the dashedline denotes the Earth-facing longitude for the particular case of ω = 180 For the cases of ω = 270 and 360 degrees exactly the same plot is valid except that thesub-Earth longitude would be aiming upward for ω = 270 and right for ω = 360

(A color version of this figure is available in the online journal)

16

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

e = 075 a = 00385 AU

-15 -10 -05 00 05 10 15Time after periapse passage (days)

0000

0001

0002

0003

0004

Pla

netS

tar

Flu

x R

atio

e = 075 a = 00611 AU

-3 -2 -1 0 1 2 3Time after periapse passage (days)

00000

00005

00010

00015

00020

00025

Pla

netS

tar

Flu

x R

atio

Figure 16 Light Curves for two different e = 075 model integrationsTeq = 1199 K a = 00385 AU (top) and Teq = 951 K a = 00611 AU(bottom) both with ω = 180 The solid line denotes transit while the dottedline denotes secondary eclipse

(A color version of this figure is available in the online journal)

relative to orbital position can help to constrain the circulationfurther

5 CONCLUSIONS

We present three-dimensional circulation models coupledwith a two-stream non-gray radiative transfer scheme fora number of theoretical eccentric hot Jupiters We haveshown that as in published models with zero eccentricity ourhigh-eccentricity circulation models are dominated by eastwardflow at photospheric levels which cause an eastward displace-ment of the hottest regions from the substellar point The rapidrotation rates associated with pseudo-synchronization at higheccentricity lead to a small Rossby deformation radius andin some cases multiple jets in the atmosphere Global-meantemperatures and dayndashnight temperature differences peak notat periapse but several hours afterward due to finite radiativetimescales in the planetrsquos atmosphere

Furthermore we show that equatorial superrotation is gener-ated and maintained by eddies formed by the strong dayndashnightheating contrast which induce a flux of momentum from mid-latitudes to the equator The eddy magnitudes and momentumfluxes peak just after periapse passage leading to variations inthe zonal-mean flow throughout the orbit

Figure 17 Globally averaged pressurendashtemperature profiles from 1-D radiative-convective radiative transfer models of HD 189733b with 11 frequency bins(red profiles) 30 frequency bins (green profiles) and 196 frequency bins (blackprofiles) From left to right the profiles vary in stellar flux from the nominalvalue as well as 2times 4times and 8times higher flux The 8times higher flux case has atemperature inversion

(A color version of this figure is available in the online journal)

Lastly we have shown that the spatial and temporal vari-ations of the wind and temperature structure as well as theorbital viewing geometry of the system with respect to Earthcan affect the time and amplitude of peak IR flux seen infull-orbit light curves Depending on the viewing geometry ofthe orbit relative to Earth we find that peaks in IR flux thateither lead or lag periapse are possible in all cases a combi-nation of temporal effects (temperatures changing over time)and geometric effects (hot spots rotating into or out of view)are important in controlling the timing and amplitude of theflux peaks In cases where the dayndashnight temperature contrastis large and the rotational period is short the light curve canalso exhibit ldquoringingrdquo in flux as the hottest region of the planetrotates in and out of view This ringing is non-periodic due tothe variation in stellar heating as a function of distance

This work was supported by NASA Origins and PlanetaryAtmospheres grants to APS TK also acknowledges supportfrom the Harriet P Jenkins Pre-Doctoral Fellowship Program(JPFP) Resources supporting this work were provided bythe NASA High-End Computing (HEC) Program through theNASA Advanced Supercomputing (NAS) Division at AmesResearch Center The authors thank the anonymous referee fortheir helpful comments and suggestions

APPENDIX

VALIDATION TESTS FOR UPDATEDRADIATIVE TRANSFER SCHEME

To validate our updated radiative transfer scheme using 11frequency bins we conducted the following tests

1 First we tested the sensitivity of the number of frequencybins in a one-dimensional radiative-convective model ofHD 189733b (a = 00313 AU e = 00 RP = 82396 times107 m g = 214 m sminus2) Figure 17 compares the globallyaveraged pressurendashtemperature profiles of models run with11 frequency bins 30 frequency bins and 196 frequencybins at (from left to right) 1times 2times 4times and 8times theaverage stellar flux of HD 189733b For each group of

17

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)30 opacity bins

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)

11 opacity bins

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

30 opacity bins

minus600

minus200minus200

200

200

600

6001000

1000

14001800

2200

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

11 opacity bins

minus600

minus200

minus200

200

200

600

600

10001000

1400

1800

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

30 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

11 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

Figure 18 Globally averaged pressurendashtemperature profiles (top row) zonal-mean zonal wind (middle row) and the windtemperature profiles at 30 mbar (bottomrow) for the 30-bin (left column) and 11-bin (right column) model integrations of HD 189733b Snapshots were output at 250 Earth days

(A color version of this figure is available in the online journal)

models the 11-bin profile agrees well with the 30- and196-bin cases from 1 μbar to 1 bar the temperature variesby only tens of K Below 10 bars the 30- and 11-binprofiles differ from the 196-bin cases by up to 50 Kdue to the loss of resolution at low wavenumbers Inthe case of the highest average flux the large amount of

heating leads to a temperature inversion that is not fullycaptured by the 11-bin model but can be improved witha better initial guess at the pressure-temperature (P-T)profile

2 After testing the new radiative transfer (RT) scheme forone-dimensional models we ran full three-dimensional

18

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

simulations of HD 189733b at its nominal average stellarflux using the SPARCMITgcm Figure 18 shows the glob-ally averaged P-T profiles (top row) zonal-mean zonal wind(middle row) and windtemperature profiles at 30 mbar(bottom row) for the 30-bin (left column) and 11-bin (rightcolumn) RT setups The globally averaged profiles arenearly identical with any temperature differences withina few K The zonal wind plots also show good agree-ment with similar peak speeds (gt2600 m sminus1) equatorialjet width (sim60) and jet level (lt100 bar) At 30 mbarthe 30- and 11-bin models share similar dayside-nightsidetemperature structure with a hot spot (1100 K) eastward ofthe substellar point and two colder regions on the nightsidein the mid-latitudes

Both tests show that the bulk circulation and temperaturestructure is retained in the transition from 30 to 11 frequencybins hence we proceed to use the SPARCMITgcm with thenew scheme for these and future model simulations

REFERENCES

Adcroft A Campin J-M Hill C amp Marshall J 2004 MWRv 132 2845Charbonneau D Knutson H A Barman T et al 2008 ApJ 686 1341Cho J Y-K Menou K Hansen B M S amp Seager S 2008 ApJ

675 817Cooper C S amp Showman A P 2005 ApJL 629 L45Cooper C S amp Showman A P 2006 ApJ 649 1048Cowan N B amp Agol E 2011 ApJ 726 82Deming D Harrington J Laughlin G et al 2007 ApJL 667 L199Dobbs-Dixon I Cumming A amp Lin D N C 2010 ApJ 710 1395Dobbs-Dixon I amp Lin D N C 2008 ApJ 673 513Fortney J J Cooper C S Showman A P Marley M S amp Freedman R S

2006 ApJ 652 746Fortney J J Lodders K Marley M S amp Freedman R S 2008 ApJ

678 1419Fortney J J Marley M S Lodders K Saumon D amp Freedman R

2005 ApJL 627 L69

Fu Q amp Liou K N 1992 JAtS 49 2140Goody R West R Chen L amp Crisp D 1989 JQSRT 42 539Greenberg R 2009 ApJL 698 L42Heng K 2012 ApJL 761 L1Heng K Frierson D M W amp Phillipps P J 2011 MNRAS 418 2669Hide R 1969 JAtS 26 841Holton J R 2004 An Introduction to Dynamic Meteorology (Waltham MA

Academic)Hut P 1981 AampA 99 126Irwin J Charbonneau D Nutzman P et al 2008 ApJ 681 636Ivanov P B amp Papaloizou J C B 2004 MNRAS 353 1161Knutson H A Charbonneau D Allen L E Burrows A amp Megeath S T

2008 ApJ 673 526Knutson H A Charbonneau D Allen L E et al 2007 Natur 447 183Knutson H A Charbonneau D Cowan N B et al 2009 ApJ 690 822Langton J amp Laughlin G 2008a ApJ 674 1106Langton J amp Laughlin G 2008b AampA 483 25Lewis N K Showman A P Fortney J J et al 2010 ApJ 720 344Lodders K 2003 ApJ 591 1220Marley M S amp McKay C P 1999 Icar 138 268Mayor M amp Queloz D 1995 Natur 378 355Menou K amp Rauscher E 2009 ApJ 700 887Mlawer E J Taubman S J Brown P D Iacono M J amp Clough S A

1997 JGR 102 16663Naef D Latham D W Mayor M et al 2001 AampA 375 L27Perna R Heng K amp Pont F 2012 ApJ 751 59Pont F Knutson H Gilliland R L Moutou C amp Charbonneau D

2008 MNRAS 385 109Rauscher E amp Menou K 2010 ApJ 714 1334Rauscher E amp Menou K 2012 ApJ 750 96Saumon D Marley M S Abel M Frommhold L amp Freedman R S

2012 ApJ 750 74Showman A P Cho J Y-K amp Menou K 2011 in Exoplanets ed S Seager

(Tucson AZ Univ Arizona Press) 471Showman A P Cooper C S Fortney J J amp Marley M S 2008 ApJ

682 559Showman A P Fortney J J Lian Y et al 2009 ApJ 699 564Showman A P amp Guillot T 2002 AampA 385 166Showman A P amp Polvani L M 2011 ApJ 738 71Swain M R Vasisht G amp Tinetti G 2008 Natur 452 329Thrastarson H T amp Cho J Y 2010 ApJ 716 144Wolszczan A 1994 Sci 264 538

19

  • 1 INTRODUCTION
  • 2 MODEL
    • 21 Updates to SPARCMITgcm
    • 22 Grid Simulations
      • 3 RESULTS
        • 31 Circulation Regime
        • 32 Effect of Eccentricity
        • 33 Effect of Varying Average Stellar Flux
        • 34 Global Average Temperature versus Eccentricity and F
        • 35 Synchronous versus Pseudo-synchronous Rotation
        • 36 Momentum Budget and Equatorial Superrotation
          • 4 OBSERVATIONAL IMPLICATIONS
            • 41 Observational Effects of Varying ω
            • 42 High-e Model Integrations at Varying F
              • 5 CONCLUSIONS
              • APPENDIX VALIDATION TESTS FOR UPDATED RADIATIVE TRANSFER SCHEME
              • REFERENCES
Page 5: THREE-DIMENSIONAL ATMOSPHERIC CIRCULATION OF HOT …showman/publications/kataria-etal-2013.pdf · planets’ atmospheric composition, temperature structure, and circulation. A fifth

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus180 minus90 0 90 180True anomaly

minus2 minus15 minus1 minus05 0 05 1 15 2

10minus3

10minus2

10minus1

100

101

102

simulated time (Earth days)

pre

ssu

re (

bar

) 700700

800800

800

900

900

900

1000

10001000

11001100

1100

12001200 1200

1300 1300 1300

1400 1400 1400

1500 1500 1500

T(K)

700

800

900

1000

1100

1200

1300

1400

1500

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00534 AU

minus200

minus200

200

200

600

600

1000

1000

1400

18002200

2600300034003800

(m sminus1)

0

1000

2000

3000

4000

5000

Figure 4 Dynamical response of a model with Teq = 951 K e = 050 anda = 00534 AU Top global horizontally averaged temperature vs pressureand time throughout the full orbit zero is defined as periapse passage The upperaxis gives the true anomaly throughout the orbit Bottom zonally averaged zonal(eastndashwest) winds vs pressure and latitude for the same model time-averagedover a full orbit

(A color version of this figure is available in the online journal)

qualitative nature of the circulation We reiterate that inte-grations without this drag scheme (but including our fourth-order Shapiro filter) conserve angular momentum to betterthan 01

3 RESULTS

31 Circulation Regime

To illustrate the circulation regime of an eccentric exoplanetwe present a nominal model where e = 050 a = 00534 AUand Teq = 951 K representative of the simulations (Figure 4)The top panel shows the horizontally averaged global temper-ature as a function of pressure time (in Earth days) and trueanomaly f where f = 0 at periapse and f = plusmn180 atapoapse Temperatures remain fairly constant at pressures ex-ceeding 1 bar However in the uppermost layers of the atmo-sphere temperatures vary throughout the orbit peaking sim6 hr

after periapse passage The bottom panel shows the zonal-meanzonal wind8 of the same model time-averaged over a full or-bit The flow exhibits a superrotating (eastward) equatorial jetwhich is the dominant feature of the flow with peak winds ofabout 5000 m sminus1 Superrotation is commonly seen in models ofhot Jupiters and hot Neptunes (eg Showman amp Guillot 2002Cooper amp Showman 2005 Showman et al 2008 2009 Dobbs-Dixon amp Lin 2008 Dobbs-Dixon et al 2010 Lewis et al 2010Heng et al 2011 Perna et al 2012 Rauscher amp Menou 20102012) although not by all groups (Cho et al 2008 Thrastarsonamp Cho 2010)

With peak winds exceeding 2ndash5 km sminus1 the equatorial jet issupersonic as the sound speed for this class of planets is ap-proximately 2 km sminus1 As Heng (2012) shows the presence ofthis supersonic flow generates shocks at the dayndashnight termi-nator upstream of the substellar point The SPARCMITgcmdoes not properly account for the removal of kinetic energy as aresult of these shocks The formation of shocks is expected to beshallow (ie at low pressure) in both of the model integrationsthat Heng (2012) presents the Mach number falls below unityat pressures greater than 1 bar and hence shocks are not formedNevertheless the infrared photosphere and the peak jet speedsin our model occur at sufficiently low pressure that shocks couldplay some role particularly when the orbital eccentricity is highMoreover as a possible energy dissipation mechanism shockscould have a non-local effect (ie they could cause an indirecteffect on the flow even at pressures deeper than those where theshocks directly occur) Clearly more work is warranted on therole of shocks in hot Jupiter atmospheres

Snapshots of the wind and temperature profile throughout theorbit at a pressure level of 30 mbar (approximately the levelof the infrared photosphere) show that the planet undergoes aflash heating event as the planet approaches periapse passage(Figure 5) The eight panels show snapshots throughout onecomplete orbit from apoapse (top figure) to hours after periapse(f = 119 follow arrows) and back again to apoapse all withthe same colorscale for temperature Like the circulation ofHD 189733b in its nominally circular orbit (see Showman et al2009) the flow maintains the eastward equatorial superrotatingjet throughout its orbit (Figure 6) As the planet approachesperiapse (Figure 5 f = minus12) temperatures rise to sim1000 Kat the equator and a high-amplitude eddy structure developswith a pattern of temperature and wind vectors that tilt eastwardtoward the equator This so-called ldquochevronrdquo-shaped feature isprominent at periapse (bottom panel) extends over a wide rangeof latitude and longitude and persists hours after periapse Thechevron shape is also seen in our other integrations and indicatesthat eddies transport momentum equatorward (see Section 36)

32 Effect of Eccentricity

Our models show that at constant orbit-mean stellar fluxthe dayside temperatures dayndashnight temperature differencesand wind speeds at periapse passage all increase with increas-ing orbital eccentricity This is illustrated in Figure 7 whichcompares three simulations with an equilibrium temperature of951 K (〈F 〉 = 185691 W mminus2) that vary in eccentricity from00 (top row) to 075 (bottom row) Again we plot each columnon the same colorscale (zonal wind and temperature) for com-parison As eccentricity increases peak temperatures increase

8 Zonal wind is defined as eastndashwest winds with positive (negative) valuesdenoting eastward (westward) winds a zonal average is an average inlongitude

5

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=000deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus1242deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=1242deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus7227deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=7227deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus11988deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=11988deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus18000deg

(K)500

600

700

800

900

1000

1100

1200

Figure 5 Snapshots of the wind and temperature structure at 30 mbar of a planet with Teq = 951 K e = 050 and a = 00534 AU To trace the behavior throughoutone complete orbit the different panels should be followed counterclockwise from the top panel Snapshots are taken from apoapse (true anomaly f = minus1800top panel) to hours after periapse (f = 119 follow arrows) and back again to apoapse Each panel is plotted on the same temperature colorscale Snapshots weretaken between 6103 and 6108 Earth days Note that peak temperatures lag periapse passage Periapse corresponds to f = 0 and apoapse to 180 The vertical linecorresponds to substellar longitude

(A color version of this figure is available in the online journal)

6

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=000deg

minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus1242deg

minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=1242deg

minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus7227deg

minus200

minus200 200

200

200

600

600

1000

1000

1400

1400

18002200260030003400

3800

4200

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=7227deg

minus200

minus200minus200 200

200

200

600

600

1000

1000

1400

180022

002600300034003800

4200

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus11988deg

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=11988deg

minus600minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

1800

2200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus18000deg

minus200

minus200

200

200

600

600

1000

1000

1400

1800

2200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

Figure 6 Snapshots of the zonal-mean zonal wind for the same integration described in Figure 5 (Teq = 951 K e = 050 and a = 00534 AU) Snapshots are againtaken from apoapse (f = minus1800) to hours after periapse (f = 119) and back again to apoapse Each panel is plotted on the same temperature colorscale Thesnapshots correspond to outputs from 6103 to 6108 Earth days Note that the equatorial superrotating jet is persistent throughout the orbit with peak speeds exceeding5000 m sminus1

(A color version of this figure is available in the online journal)

7

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 000

200

200

200600

600

600

1000

1000

10001400

1400

18002200

2600

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 000

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 025

minus600

minus600

minus200 minus200

200

200

600

600

1000

140018002200260030003400

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 025

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 050

minus200

minus200

200

200

600

600

1000

1000

1400

18002200

2600300034003800

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 050

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 075

200

200

200

200

600

600

600

1000

1000

1000

140018

002200260030003400

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 075

(K)

500

600

700

800

900

1000

1100

1200

1300

Figure 7 Comparison of simulations with Teq = 951 K with increasing eccentricity from e = 00 (top row) to e = 075 (bottom row) Shown are plots of thezonal-mean zonal wind averaged over a full orbit (left column) and snapshots of the wind and temperature profiles at 30 mbar (right column) The snapshots of thewind and temperature profiles in the right column were taken at periapse passage The vertical bars in the right column denote the longitude of the substellar pointThe snapshots of wind and temperature correspond to model outputs of (from top to bottom) 4100 5006 6106 and 6904 Earth days

(A color version of this figure is available in the online journal)

8

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

from 1000 K to 1300 K In addition the variation in temper-ature from dayside to nightside increases this strengthens thewinds within the equatorial superrotating jet Peak winds reachsim2500 m sminus1 in the circular case but increase to sim5000 m sminus1

in the high eccentricity caseAs the eccentricity increases the equatorial jet also narrows

in latitude In the case of highest eccentricity the atmospheredevelops westward midlatitude jets and eastward high-latitudejets The narrowing of the equatorial jet arises from the factthat jet width is confined to the Rossby radius of deformation(Showman amp Polvani 2011) the length scale over which theatmosphere adjusts to planetary-scale phenomena The Rossbyradius of deformation (Holton 2004) is defined at the equator as

LR =radic

c

β (6)

where c is the gravity wave speed and β is the derivative ofthe Coriolis parameter f with respect to latitudinal distancey β = df dy = 2Ωrot cos φRp where Ωrot is the planetaryrotation rate φ is the latitude and Rp is the planetary radius Atthe equator β is simply 2ΩrotRp Hence at the equator LR is

LR =radic

cRp

2Ωrot (7)

Therefore LR prop 1radic

Ωrot and an increased rotation rateleads to a narrower deformation radius From the Hut (1981)formulation the combined effect of the larger orbital period andincrease in eccentricity produces a shorter rotational periodThis in turn increases the rotation rate which reduces thedeformation radius For example the ratio between Ωrot forthe e = 00 case and the e = 075 case is 63 hence we wouldexpect that the equatorial jet should be narrower by a factor ofradic

63 sim 25 This is indeed the case as the e = 075 case hasan average jet width of sim40 latitude while the e = 00 casehas an average jet width of sim100 latitude

33 Effect of Varying Average Stellar Flux

An increase in average stellar flux leads to increases intemperature and wind speeds illustrated in Figure 8 Each caseshown has an eccentricity of 025 these maps have independentcolorscales to better show atmospheric structure The planetthat receives the highest average stellar flux throughout its orbitis hottest overall (bottom) with peak temperatures of 1300 KFurthermore because the planet with the highest 〈F 〉 orbits at thesmallest orbital distance it has a faster rotation rate and hencea narrower equatorial jet this comparison makes the effect ofrotation on the jet structure much more apparent

34 Global Average Temperature versus Eccentricity and 〈F 〉Here we illustrate the variation in time of maximum

planet-wide temperature with eccentricity and average stellarflux Figure 9 shows the globally averaged temperature foreach model integration plotted as a function of mean anomaly9

where M = 0 corresponds to periapse passage With increas-ing eccentricity the time at which the global-mean temperaturereaches its peak value decreases This is expected as a planet

9 The mean anomaly M is defined as (2πPorb)(t minus τ ) where τ is the timeof periapse passage

with a larger eccentricity will have a higher global-average tem-perature at periapse and hence shorter radiative time constant(τrad prop T minus3 see Showman et al 2011) Figure 10 plots the timeof peak temperature from periapse versus eccentricity Here theinverse relationship between eccentricity and the time of peaktemperature is clear

Figures 9 and 10 serve as a nice summary of the modelintegrations and their relation to observations as temperatureis one of the planetary properties that shapes light curvesat secondary eclipse Observers can use these predictions toestimate the timing of peak flux (and global temperature) basedon a given stellar flux andeccentricity (see Section 4)

35 Synchronous versus Pseudo-synchronous Rotation

Here we explore the sensitivity of the circulation to rota-tion rate by comparing models performed with the pseudo-synchronous rotation rate to otherwise identical models per-formed with the synchronous rotation rate This is illustratedin Figure 11 Here the rotation rate varies by almost a factorof three Both cases still have an equatorial superrotating jetand the overall atmospheric structure is similar However thereare subtle differences in the synchronous case the equatorialeastward winds and high-latitude westward winds are higher inmagnitude (right column) The synchronous case also exhibitsmore latitudinal and longitudinal temperature variation partic-ularly on the dayside this is more apparent at apoapse (bot-tom row) In the pseudo-synchronous case the faster rotationrate serves to homogenize the temperature in longitude becausethe time for rotation to carry air parcels from day to night isshorter The temperature difference from equator to pole is alsolarger because the faster rotation rate inhibits equator-to-poleheat transport Thus while it is unlikely that an eccentric planetsynchronously rotates its star the distinction can lead to notabledifferences in the mean flow

These results can also be used to compare the Hut (1981)formulation shown by the pseudo-synchronous case to for-mulations by other groups As shown in Figure 11 if anotherprescription for rotation rate differed from Hut (1981) by afactor of 2ndash3 (as illustrated by the synchronous case) the qual-itative picture would not look that different However if theprescription differed by a factor of more than 3ndash4 the picturecould substantially differ For example the rotation rate calcu-lated from Ivanov amp Papaloizou (2004) Equation (63) yields avalue of 14 times 106 s a factor of four larger than the Hut (1981)calculation So results using this prescription would look verydifferent

Still the rotation rate of eccentric exoplanets is a completelyunconstrained problem One must be careful then in choosinga formulation noting its assumptions and limitations

36 Momentum Budget and Equatorial Superrotation

In most of our simulations the planet maintains an equatorialeastward jet throughout its orbit Superrotation is commonin three-dimensional circulation models of hot Jupiters (egShowman amp Guillot 2002 Showman et al 2008 2009 Menouamp Rauscher 2009 Rauscher amp Menou 2010 2012 Cooperamp Showman 2005 2006 Dobbs-Dixon amp Lin 2008 Dobbs-Dixon et al 2010 Heng et al 2011) and is also seen withinour own solar system on Venus Titan Jupiter and SaturnWhile superrotation on solar system planets has been extensivelystudied only recently have the mechanisms for superrotation onsynchronously rotating exoplanets been identified (Showman amp

9

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)a= 00802 AU

minus200

minus200

200

200

200

600

600

1000

100014001800

22002600

(m sminus1)

0

500

1000

1500

2000

2500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00802 AU

(K)

550

600

650

700

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00505 AU

minus600

minus600

minus200 minus200

200

200

600

600

1000

140018002200260030003400

(m sminus1)

minus500

0

500

1000

1500

2000

2500

3000

3500

4000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00505 AU

(K)

700

750

800

850

900

950

1000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00318 AU

minus200

minus200

200

200

200 200

600

600

600

10001400

1800220026003000340038004200

5000

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00318 AU

(K)700

800

900

1000

1100

1200

1300

Figure 8 Comparison of simulations with e = 025 with decreasing average stellar flux from bottom to top panels Shown are plots of the orbit-averaged zonal-meanzonal wind (left column) and the wind and temperature profiles at 30 mbar at periapse (right column) Each plot has an independent colorscale The vertical bars inthe right column denote the substellar longitude The snapshots of wind and temperature correspond to model outputs of (from top to bottom) 3706 5006 and 2201Earth days

(A color version of this figure is available in the online journal)

10

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

0 05 1 15 2 25 3 35620

640

660

680

700

720

740ltFgt = 11617 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050e=075

0 05 1 15 2 25 3 35880

890

900

910

920

930

940

950

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

ltFgt = 73680 Wmminus1

e=025e=050

0 05 1 15 2 25 3 35 4960

980

1000

1020

1040

1060

1080

1100

1120

1140

1160ltFgt = 185691 Wmminus1

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

Mean anomaly

e=025e=050e=075

0 05 1 15 2 25 3 35 4 451100

1150

1200

1250

1300

1350ltFgt = 468183 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050

Figure 9 Plots of global average temperature vs mean anomaly for each modelintegration Each plot contains profiles of a given average stellar flux (468183185691 73680 and 11617 W mminus1) with blue green and red representingprofiles for e = 025 050 and 075 respectively

(A color version of this figure is available in the online journal)

03 04 05 06 070

05

1

15

2

25

3

eccentricity

Pea

k o

ffse

t fr

om

per

iap

se (

Ear

th d

ays)

468183 Wmminus1

185691 Wmminus1

73680 Wmminus1

11617 Wmminus1

Figure 10 Time after periapse passage at which the global-average temperaturepeaks (from Figure 9) vs eccentricity for all model integrations The bluegreen red and teal profiles correspond to average stellar fluxes of 468183185691 73680 and 11617 W mminus1 respectively

(A color version of this figure is available in the online journal)

Polvani 2011 hereafter SP2011) As discussed in their paperthe generation of superrotation requires up-gradient momentumtransport that pumps angular momentum from outside the jet towithin it Only transport by waves and eddies can satisfy thisrequirement (Hide 1969)

SP2011 demonstrate that for synchronous hot Jupiters onnear-circular orbits such as HD 189733b and HD 209458bequatorial superrotation is generated by standing planetaryscale Kelvin and Rossby waves at the equator and midlati-tudes respectively that are themselves the dynamical responseto the planetrsquos large dayndashnight heating contrast Because theKelvin waves propagate eastward and the Rossby waves prop-agate westward a phase tilt of the wind vectors emerges fromnorthwest to southeast in the northern hemisphere and south-west to northeast in the southern hemisphere with an over-all shape resembling a chevron pointing eastward This patterncauses meridional eddy angular momentum fluxes that convergeangular momentum onto the equator generating equatorialsuperrotation

Despite the fact that not all our model integrations aresynchronous our results suggest that similar mechanisms arestill at play The chevron shape that SP2011 describe is exactlywhat we noted earlier in our plots of the wind and temperatureprofiles from Figure 5 suggesting that as in SP2011 our modelsexhibit an equatorward flux of eddy angular momentum Thisis demonstrated quantitatively in Figure 12 which shows thezonally averaged meridional flux of relative zonal eddy angularmomentum uprimevprime cos φ at snapshots throughout the orbit (asin Figures 5 and 6) Here uprime and vprime represent the deviationof the zonal and meridional winds respectively from theirzonal averages Positive values indicate a northward flux ofeastward eddy angular momentum and negative values indicatea southward flux of eastward eddy angular momentum Atapoapse the small dayndashnight forcing causes poleward transportof eddy angular momentum at latitudes greater than 50 butequatorward transport of eddy angular momentum equator of50 latitude the amplitudes are weak due to the low insolationat apoapse As the planet moves closer to the star (middlerow left column) the increase in dayndashnight forcing strengthens

11

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)e = 050 pseudominussynchronous rotation

minus200

200

200

600

600

1000

10001400

180022002600

3000

3400

3800

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

synchronous rotation

minus600

minus600

minus200

minus200 minus200

minus200

minus200

200

200

600

10001400180022002600300034003800

4200

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

Figure 11 Comparison of pseudo-synchronous (Trot = 3046 times 105 s) vs synchronous (Trot = 8544 times 105 s) rotation for a case with Teq = 755 K e = 050 anda = 00848 AU Wind and temperature snapshots are taken at periapse (middle row) and apoapse (bottom row) at 5008 Earth days and 5013 Earth days respectivelyThe vertical bar denotes substellar longitude

(A color version of this figure is available in the online journal)

the equatorward flux of eastward eddy angular momentumat latitudes between 0 and 50 in both hemispheres Themomentum flux grows and persists until hours after periapseAngular momentum thus converges onto the equator andmaintains the equatorial jet against westward accelerationscaused by advection friction (if any) and Coriolis forces

These mechanisms also explain the difference in wind speedsbetween the synchronous and pseudo-synchronous cases inFigure 11 The synchronous case has a higher dayndashnight heatingcontrast hence a higher equatorward flux of eddy momentumleading to a stronger narrower equatorial jet Moreover theequatorial jet in the pseudo-synchronous case extends to deeper

12

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=18000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

Figure 12 Eddy momentum flux uprimevprime cos φ throughout the orbit of a model where Teq = 951 K e = 05 and a = 00534 AU Positive values indicate a northwardflux of eastward eddy angular momentum while negative values indicate a southward flux of eastward eddy angular momentum The maximum momentum fluxoccurs hours after periapse passage (f = 72) The snapshots correspond to model outputs from 6103 to 6108 Earth days

(A color version of this figure is available in the online journal)

13

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

Figure 13 Differing orbital viewing geometries explored in this study ω = 180 ω = 270 and ω = 360

(A color version of this figure is available in the online journal)

pressures than that of the synchronous case this is most likelydue to a change in momentum budget

4 OBSERVATIONAL IMPLICATIONS

Observations of transiting exoplanets in circular orbits areknown to be shaped by the atmospheric circulation As wasfirst predicted for hot Jupiters by Showman amp Guillot (2002)and later observationally confirmed for hot Jupiter HD 189733bby Knutson et al (2007) an equatorial superrotating jet causesan eastward displacement of the hot spot from the substellarpoint under appropriate conditions Hence a light curve showsa peak in infrared flux minutes to hours before secondaryeclipse We expect such a situation for eccentric exoplanetsas well However observations of eccentric planets can becomplicated because the flux received by the planet variesthroughout its orbit Previous studies explored such situationsbut have not used three-dimensional models (eg Langton ampLaughlin 2008a 2008b Cowan amp Agol 2011) We can expandon these studies and the three-dimensional study by Lewis et al(2010) by using our SPARC model results

We choose to generate light curves following the proceduresdescribed in Fortney et al (2006) for each model integrationat three values of the argument of periastron (ω) the anglebetween the radius vector to the ascending node and the periapseof the orbit In particular we choose orientations where transitoccurs before periapse (at f = minus90ω = 180) transit occursat apoapse (ω = 270) and transit occurs after periapse (atf = +90ω = 360) (Figure 13) These values reflect thebroad range of viewing geometries of transiting exoplanets forexample HAT-P-2b (e = 052) HAT-P-17b (e = 035) andHD 97658b (e = 013) have ω near 180 WASP-8b (e = 031)HAT-P-31b (e = 025) and HD 80606b (e = 093) have ω near270 and GJ 436b (e = 015) XO-3b (e = 026) and HAT-P-11b (e = 020) have ω near 360 (see exoplaneteu for a full listof values) Note that when transit occurs the nightside is visibleto Earth while the dayside is visible at the time of secondaryeclipse

41 Observational Effects of Varying ω

Using the nominal case (e = 050 a = 00534 AU Teq =951 K) we illustrate the dependence of full-orbit light curves onthe value of ω Figure 14 shows three light curves of the same

model integration but at the differing values of ω described inthe previous section Hence the differences in light curves resultsolely from the differing viewing geometries The planetstarflux ratio is plotted as a function of Earth days from periapsein the warm Spitzer wavelength bands (36 and 45 μm) andJHK bands (126 165 and 220 μm) The solid and dashedlines correspond to transit and secondary eclipse respectivelyFor the light curve where ω = 180 (top panel) the peak ininfrared (IR) flux occurs approximately 6 hr after periapse andbefore secondary eclipse The timing of the peak is due toa combination of the radiative lag of the planet and the factthat the hottest regions rotate into an Earth-facing orientationhours after periapse passage The dayside which is most visiblenear secondary eclipse is hottest after periapse Additionallytemperatures reach peak values at times close to when thosehottest regions are facing Earth This helps explain the highamplitude of the light curve peak in the ω = 180 case relative tothe other two cases These effects can be visualized by plottingthe temperature near the photosphere along the equator in apolar projection (Figure 15) Snapshots throughout the orbit plotthe temperature in a polar projection as a function of pressureand orbital position at a latitudinal slice near the equatorthe solid and dashed lines denote the substellar longitude andEarth-facing longitude respectively Based on the temperaturesat the Earth-facing longitude it is apparent that the peak in IRflux should occur after periapse and before secondary eclipse(see f = 12 and 72) this is indeed the case

For the same model but instead with ω = 270 the peak inIR flux occurs 1 hr after eclipse and periapse In this case thedayside is visible near periapse and therefore we are seeing theplanet increase in temperature (and IR flux) during its closestapproach to the star However the hottest regions are displacedeastward of the substellar point and therefore the hottest regionsface Earth significantly before periapse passage (Figure 15 atf asymp 72) Thus by the time the peak temperatures are reachednear and after periapse the hottest regions are already rotatingout of view from Earth This explains the lower amplitude ofthe light curve peak as compared to the peak of the ω = 180case The IR peak decreases as the planet moves further awayfrom the star along our line of sight A second ramp-up in fluxoccurs approximately 1 day after periapseeclipse this is due tothe hottest regions rotating back into view as the planet movestoward apoapse

14

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015P

lane

tSta

r F

lux

Rat

io

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

Figure 14 Three light curves of the IR flux (expressed as a planetstar fluxratio) vs time throughout the full orbit for a single model integration whereTeq = 951 K e = 050 and a = 00534 AU The flux is plotted in warm Spitzerbandpasses 36 and 45 μm and JHK bands (126 165 220 μm respectively)These three cases only differ by their value of ω shown in Figure 13 The solidlines denote transit while the dotted lines denote secondary eclipse

(A color version of this figure is available in the online journal)

In the case where ω = 360 two distinct peaks are presentmdashasmaller peak that occurs after secondary eclipse and sim3ndash4 hrbefore periapse and a second higher peak that occurs aftertransit The first peak occurs as some of the dayside is visible asthe planet is heated approaching periapse The peak decreasesas the dayside rotates out of view For this geometry the hottestregions are facing away from Earth when they reach their peaktemperatures (Figure 15) The second peak occurs after transitas the hottest regions again rotate into view At this point in theorbit peak temperatures have already occurred and thereforethis peak is lower in amplitude than the other two light curvesHowever because the planet is hotter than it was after secondary

eclipse this second peak has a higher amplitude than the firstone

42 High-e Model Integrations at Varying 〈F 〉If an eccentric planet has a large enough dayndashnight tempera-

ture difference and a sufficiently rapid rotation rate its resultantlight curve can show a periodic rise and fall in flux through-out its orbit The light curves shown in Figure 16 are twoe = 075 model integrations that differ in 〈F 〉 the top panelshows a case where 〈F 〉 = 468183 W mminus2 (Teq = 1199 K)and a = 00385 AU while the bottom panel is a case where〈F 〉 = 185691 W mminus2 (Teq = 951 K) and a = 00611 AU Forboth cases light curves are shown for the ω = 180 geome-try (see Figure 13) The differences in 〈F 〉 lead to differencesin light curve amplitude in the two models the closer hotterplanet exhibits a higher flux amplitude Not surprisingly thetiming of peak IR flux for both models resembles the ω = 180case in Figure 14 The peak in each case occurs after periapseand before eclipse due to the combined effects of the ther-mal lag of the planet and geometry of the system Additionallyboth light curves exhibit a quasi-periodic rise and fall of IR fluxdays after periapse passage This phenomenon called ldquoringingrdquowas also seen in models by Langton amp Laughlin (2008a) andCowan amp Agol (2011) caused by the hottest point of the planetrotating in and out of view from Earth For ringing to occurthe temperature difference from dayside to nightside must belarge and it must survive for multiple planetary rotation periodsIn the models shown in Figure 16 this is aided by the rapidpseudo-synchronous rotation rate associated with an eccentric-ity of 075 (Table 2) For the top case the dayndashnight tempera-ture difference is over sim1300 K while the bottom case variesby sim750 K (Figure 7) The light curves from Figure 14 do notexhibit ringing because the dayndashnight temperature difference islow (sim300ndash400 K)

Because the radiative and dynamical timescales in the modelsby Cowan amp Agol (2011) are calculated by a simple scaling withtemperature the ringing seen in their simulated light curvesoccurs with a period equal to the planetrsquos assumed solid-bodyrotation period in the inertial frame of the winds In realityan eccentric exoplanet approaching apoapse would becomeincreasingly cold the winds would weaken and decrease andhence the ringing should be non-periodic Indeed for ourthree dimensional models the ringing in each case has varyingperiodsmdashthe top case has a ringing period ranging from 03 to04 days (Trot = 035 days) while the bottom panel has a periodranging 07ndash08 days (Trot = 07 days)

Despite these differences this study and the studies byLangton amp Laughlin (2008a) and Cowan amp Agol (2011) illus-trate the influence of atmospheric dynamics on the observationsof eccentric exoplanets The light curve features described herecould be seen in future follow-up observations of eccentric ex-oplanets and observers can use the conclusions drawn aboveto relate the observations to the planetrsquos atmospheric structureFor example if ringing is present in an IR light curve the planetmust have a strong dayndashnight temperature asymmetry and hencewe would expect the planet to have strong superrotation at theequator (Showman amp Polvani 2011) This is probably the casefor highly eccentric (e gt 05) short-period exoplanets such asHD 17156b and HD 80606b If ringing were observed it wouldalso allow an estimate of the sum of the rotation speed and thesuperrotation speed thereby placing constraints on both windspeeds and rotation rates of eccentric hot Jupiters The timing ofpeak IR flux relative to transit and secondary eclipse and also

15

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

f=000deg

(K)800

850

900

950

1000

f=minus1242deg

(K)800

850

900

950

1000f=1242deg

(K)800

850

900

950

1000

f=minus7227deg

(K)800

850

900

950

1000f=7227deg

(K)800

850

900

950

1000

f=minus11988deg

(K)800

850

900

950

1000f=11988deg

(K)800

850

900

950

1000

f=minus18000deg

(K)800

850

900

950

1000

Figure 15 Plots of temperature vs longitude and pressure in equatorial slices throughout one full orbit for our nominal model with Teq = 951 K e = 050 anda = 00534 AU Each annulus shows temperature vs longitude and log-pressure for a snapshot at a particular orbital position (denoted by the value of f) as viewedfrom the planetrsquos north pole Pressures between 30 and 43 mbar are displayed in each annulus The different annuli mark the time evolution throughout one completeorbit starting from apoapse for the top annulus proceeding to periapse and then back to apoapse (follow arrows moving counterclockwise around the plot) Thesesnapshots correspond to model outputs from 6103 to 6108 Earth days The solid line in each snapshot indicates the longitude of the substellar point while the dashedline denotes the Earth-facing longitude for the particular case of ω = 180 For the cases of ω = 270 and 360 degrees exactly the same plot is valid except that thesub-Earth longitude would be aiming upward for ω = 270 and right for ω = 360

(A color version of this figure is available in the online journal)

16

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

e = 075 a = 00385 AU

-15 -10 -05 00 05 10 15Time after periapse passage (days)

0000

0001

0002

0003

0004

Pla

netS

tar

Flu

x R

atio

e = 075 a = 00611 AU

-3 -2 -1 0 1 2 3Time after periapse passage (days)

00000

00005

00010

00015

00020

00025

Pla

netS

tar

Flu

x R

atio

Figure 16 Light Curves for two different e = 075 model integrationsTeq = 1199 K a = 00385 AU (top) and Teq = 951 K a = 00611 AU(bottom) both with ω = 180 The solid line denotes transit while the dottedline denotes secondary eclipse

(A color version of this figure is available in the online journal)

relative to orbital position can help to constrain the circulationfurther

5 CONCLUSIONS

We present three-dimensional circulation models coupledwith a two-stream non-gray radiative transfer scheme fora number of theoretical eccentric hot Jupiters We haveshown that as in published models with zero eccentricity ourhigh-eccentricity circulation models are dominated by eastwardflow at photospheric levels which cause an eastward displace-ment of the hottest regions from the substellar point The rapidrotation rates associated with pseudo-synchronization at higheccentricity lead to a small Rossby deformation radius andin some cases multiple jets in the atmosphere Global-meantemperatures and dayndashnight temperature differences peak notat periapse but several hours afterward due to finite radiativetimescales in the planetrsquos atmosphere

Furthermore we show that equatorial superrotation is gener-ated and maintained by eddies formed by the strong dayndashnightheating contrast which induce a flux of momentum from mid-latitudes to the equator The eddy magnitudes and momentumfluxes peak just after periapse passage leading to variations inthe zonal-mean flow throughout the orbit

Figure 17 Globally averaged pressurendashtemperature profiles from 1-D radiative-convective radiative transfer models of HD 189733b with 11 frequency bins(red profiles) 30 frequency bins (green profiles) and 196 frequency bins (blackprofiles) From left to right the profiles vary in stellar flux from the nominalvalue as well as 2times 4times and 8times higher flux The 8times higher flux case has atemperature inversion

(A color version of this figure is available in the online journal)

Lastly we have shown that the spatial and temporal vari-ations of the wind and temperature structure as well as theorbital viewing geometry of the system with respect to Earthcan affect the time and amplitude of peak IR flux seen infull-orbit light curves Depending on the viewing geometry ofthe orbit relative to Earth we find that peaks in IR flux thateither lead or lag periapse are possible in all cases a combi-nation of temporal effects (temperatures changing over time)and geometric effects (hot spots rotating into or out of view)are important in controlling the timing and amplitude of theflux peaks In cases where the dayndashnight temperature contrastis large and the rotational period is short the light curve canalso exhibit ldquoringingrdquo in flux as the hottest region of the planetrotates in and out of view This ringing is non-periodic due tothe variation in stellar heating as a function of distance

This work was supported by NASA Origins and PlanetaryAtmospheres grants to APS TK also acknowledges supportfrom the Harriet P Jenkins Pre-Doctoral Fellowship Program(JPFP) Resources supporting this work were provided bythe NASA High-End Computing (HEC) Program through theNASA Advanced Supercomputing (NAS) Division at AmesResearch Center The authors thank the anonymous referee fortheir helpful comments and suggestions

APPENDIX

VALIDATION TESTS FOR UPDATEDRADIATIVE TRANSFER SCHEME

To validate our updated radiative transfer scheme using 11frequency bins we conducted the following tests

1 First we tested the sensitivity of the number of frequencybins in a one-dimensional radiative-convective model ofHD 189733b (a = 00313 AU e = 00 RP = 82396 times107 m g = 214 m sminus2) Figure 17 compares the globallyaveraged pressurendashtemperature profiles of models run with11 frequency bins 30 frequency bins and 196 frequencybins at (from left to right) 1times 2times 4times and 8times theaverage stellar flux of HD 189733b For each group of

17

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)30 opacity bins

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)

11 opacity bins

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

30 opacity bins

minus600

minus200minus200

200

200

600

6001000

1000

14001800

2200

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

11 opacity bins

minus600

minus200

minus200

200

200

600

600

10001000

1400

1800

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

30 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

11 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

Figure 18 Globally averaged pressurendashtemperature profiles (top row) zonal-mean zonal wind (middle row) and the windtemperature profiles at 30 mbar (bottomrow) for the 30-bin (left column) and 11-bin (right column) model integrations of HD 189733b Snapshots were output at 250 Earth days

(A color version of this figure is available in the online journal)

models the 11-bin profile agrees well with the 30- and196-bin cases from 1 μbar to 1 bar the temperature variesby only tens of K Below 10 bars the 30- and 11-binprofiles differ from the 196-bin cases by up to 50 Kdue to the loss of resolution at low wavenumbers Inthe case of the highest average flux the large amount of

heating leads to a temperature inversion that is not fullycaptured by the 11-bin model but can be improved witha better initial guess at the pressure-temperature (P-T)profile

2 After testing the new radiative transfer (RT) scheme forone-dimensional models we ran full three-dimensional

18

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

simulations of HD 189733b at its nominal average stellarflux using the SPARCMITgcm Figure 18 shows the glob-ally averaged P-T profiles (top row) zonal-mean zonal wind(middle row) and windtemperature profiles at 30 mbar(bottom row) for the 30-bin (left column) and 11-bin (rightcolumn) RT setups The globally averaged profiles arenearly identical with any temperature differences withina few K The zonal wind plots also show good agree-ment with similar peak speeds (gt2600 m sminus1) equatorialjet width (sim60) and jet level (lt100 bar) At 30 mbarthe 30- and 11-bin models share similar dayside-nightsidetemperature structure with a hot spot (1100 K) eastward ofthe substellar point and two colder regions on the nightsidein the mid-latitudes

Both tests show that the bulk circulation and temperaturestructure is retained in the transition from 30 to 11 frequencybins hence we proceed to use the SPARCMITgcm with thenew scheme for these and future model simulations

REFERENCES

Adcroft A Campin J-M Hill C amp Marshall J 2004 MWRv 132 2845Charbonneau D Knutson H A Barman T et al 2008 ApJ 686 1341Cho J Y-K Menou K Hansen B M S amp Seager S 2008 ApJ

675 817Cooper C S amp Showman A P 2005 ApJL 629 L45Cooper C S amp Showman A P 2006 ApJ 649 1048Cowan N B amp Agol E 2011 ApJ 726 82Deming D Harrington J Laughlin G et al 2007 ApJL 667 L199Dobbs-Dixon I Cumming A amp Lin D N C 2010 ApJ 710 1395Dobbs-Dixon I amp Lin D N C 2008 ApJ 673 513Fortney J J Cooper C S Showman A P Marley M S amp Freedman R S

2006 ApJ 652 746Fortney J J Lodders K Marley M S amp Freedman R S 2008 ApJ

678 1419Fortney J J Marley M S Lodders K Saumon D amp Freedman R

2005 ApJL 627 L69

Fu Q amp Liou K N 1992 JAtS 49 2140Goody R West R Chen L amp Crisp D 1989 JQSRT 42 539Greenberg R 2009 ApJL 698 L42Heng K 2012 ApJL 761 L1Heng K Frierson D M W amp Phillipps P J 2011 MNRAS 418 2669Hide R 1969 JAtS 26 841Holton J R 2004 An Introduction to Dynamic Meteorology (Waltham MA

Academic)Hut P 1981 AampA 99 126Irwin J Charbonneau D Nutzman P et al 2008 ApJ 681 636Ivanov P B amp Papaloizou J C B 2004 MNRAS 353 1161Knutson H A Charbonneau D Allen L E Burrows A amp Megeath S T

2008 ApJ 673 526Knutson H A Charbonneau D Allen L E et al 2007 Natur 447 183Knutson H A Charbonneau D Cowan N B et al 2009 ApJ 690 822Langton J amp Laughlin G 2008a ApJ 674 1106Langton J amp Laughlin G 2008b AampA 483 25Lewis N K Showman A P Fortney J J et al 2010 ApJ 720 344Lodders K 2003 ApJ 591 1220Marley M S amp McKay C P 1999 Icar 138 268Mayor M amp Queloz D 1995 Natur 378 355Menou K amp Rauscher E 2009 ApJ 700 887Mlawer E J Taubman S J Brown P D Iacono M J amp Clough S A

1997 JGR 102 16663Naef D Latham D W Mayor M et al 2001 AampA 375 L27Perna R Heng K amp Pont F 2012 ApJ 751 59Pont F Knutson H Gilliland R L Moutou C amp Charbonneau D

2008 MNRAS 385 109Rauscher E amp Menou K 2010 ApJ 714 1334Rauscher E amp Menou K 2012 ApJ 750 96Saumon D Marley M S Abel M Frommhold L amp Freedman R S

2012 ApJ 750 74Showman A P Cho J Y-K amp Menou K 2011 in Exoplanets ed S Seager

(Tucson AZ Univ Arizona Press) 471Showman A P Cooper C S Fortney J J amp Marley M S 2008 ApJ

682 559Showman A P Fortney J J Lian Y et al 2009 ApJ 699 564Showman A P amp Guillot T 2002 AampA 385 166Showman A P amp Polvani L M 2011 ApJ 738 71Swain M R Vasisht G amp Tinetti G 2008 Natur 452 329Thrastarson H T amp Cho J Y 2010 ApJ 716 144Wolszczan A 1994 Sci 264 538

19

  • 1 INTRODUCTION
  • 2 MODEL
    • 21 Updates to SPARCMITgcm
    • 22 Grid Simulations
      • 3 RESULTS
        • 31 Circulation Regime
        • 32 Effect of Eccentricity
        • 33 Effect of Varying Average Stellar Flux
        • 34 Global Average Temperature versus Eccentricity and F
        • 35 Synchronous versus Pseudo-synchronous Rotation
        • 36 Momentum Budget and Equatorial Superrotation
          • 4 OBSERVATIONAL IMPLICATIONS
            • 41 Observational Effects of Varying ω
            • 42 High-e Model Integrations at Varying F
              • 5 CONCLUSIONS
              • APPENDIX VALIDATION TESTS FOR UPDATED RADIATIVE TRANSFER SCHEME
              • REFERENCES
Page 6: THREE-DIMENSIONAL ATMOSPHERIC CIRCULATION OF HOT …showman/publications/kataria-etal-2013.pdf · planets’ atmospheric composition, temperature structure, and circulation. A fifth

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=000deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus1242deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=1242deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus7227deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=7227deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus11988deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=11988deg

(K)500

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

3037 mbar f=minus18000deg

(K)500

600

700

800

900

1000

1100

1200

Figure 5 Snapshots of the wind and temperature structure at 30 mbar of a planet with Teq = 951 K e = 050 and a = 00534 AU To trace the behavior throughoutone complete orbit the different panels should be followed counterclockwise from the top panel Snapshots are taken from apoapse (true anomaly f = minus1800top panel) to hours after periapse (f = 119 follow arrows) and back again to apoapse Each panel is plotted on the same temperature colorscale Snapshots weretaken between 6103 and 6108 Earth days Note that peak temperatures lag periapse passage Periapse corresponds to f = 0 and apoapse to 180 The vertical linecorresponds to substellar longitude

(A color version of this figure is available in the online journal)

6

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=000deg

minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus1242deg

minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=1242deg

minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus7227deg

minus200

minus200 200

200

200

600

600

1000

1000

1400

1400

18002200260030003400

3800

4200

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=7227deg

minus200

minus200minus200 200

200

200

600

600

1000

1000

1400

180022

002600300034003800

4200

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus11988deg

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=11988deg

minus600minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

1800

2200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus18000deg

minus200

minus200

200

200

600

600

1000

1000

1400

1800

2200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

Figure 6 Snapshots of the zonal-mean zonal wind for the same integration described in Figure 5 (Teq = 951 K e = 050 and a = 00534 AU) Snapshots are againtaken from apoapse (f = minus1800) to hours after periapse (f = 119) and back again to apoapse Each panel is plotted on the same temperature colorscale Thesnapshots correspond to outputs from 6103 to 6108 Earth days Note that the equatorial superrotating jet is persistent throughout the orbit with peak speeds exceeding5000 m sminus1

(A color version of this figure is available in the online journal)

7

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 000

200

200

200600

600

600

1000

1000

10001400

1400

18002200

2600

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 000

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 025

minus600

minus600

minus200 minus200

200

200

600

600

1000

140018002200260030003400

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 025

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 050

minus200

minus200

200

200

600

600

1000

1000

1400

18002200

2600300034003800

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 050

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 075

200

200

200

200

600

600

600

1000

1000

1000

140018

002200260030003400

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 075

(K)

500

600

700

800

900

1000

1100

1200

1300

Figure 7 Comparison of simulations with Teq = 951 K with increasing eccentricity from e = 00 (top row) to e = 075 (bottom row) Shown are plots of thezonal-mean zonal wind averaged over a full orbit (left column) and snapshots of the wind and temperature profiles at 30 mbar (right column) The snapshots of thewind and temperature profiles in the right column were taken at periapse passage The vertical bars in the right column denote the longitude of the substellar pointThe snapshots of wind and temperature correspond to model outputs of (from top to bottom) 4100 5006 6106 and 6904 Earth days

(A color version of this figure is available in the online journal)

8

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

from 1000 K to 1300 K In addition the variation in temper-ature from dayside to nightside increases this strengthens thewinds within the equatorial superrotating jet Peak winds reachsim2500 m sminus1 in the circular case but increase to sim5000 m sminus1

in the high eccentricity caseAs the eccentricity increases the equatorial jet also narrows

in latitude In the case of highest eccentricity the atmospheredevelops westward midlatitude jets and eastward high-latitudejets The narrowing of the equatorial jet arises from the factthat jet width is confined to the Rossby radius of deformation(Showman amp Polvani 2011) the length scale over which theatmosphere adjusts to planetary-scale phenomena The Rossbyradius of deformation (Holton 2004) is defined at the equator as

LR =radic

c

β (6)

where c is the gravity wave speed and β is the derivative ofthe Coriolis parameter f with respect to latitudinal distancey β = df dy = 2Ωrot cos φRp where Ωrot is the planetaryrotation rate φ is the latitude and Rp is the planetary radius Atthe equator β is simply 2ΩrotRp Hence at the equator LR is

LR =radic

cRp

2Ωrot (7)

Therefore LR prop 1radic

Ωrot and an increased rotation rateleads to a narrower deformation radius From the Hut (1981)formulation the combined effect of the larger orbital period andincrease in eccentricity produces a shorter rotational periodThis in turn increases the rotation rate which reduces thedeformation radius For example the ratio between Ωrot forthe e = 00 case and the e = 075 case is 63 hence we wouldexpect that the equatorial jet should be narrower by a factor ofradic

63 sim 25 This is indeed the case as the e = 075 case hasan average jet width of sim40 latitude while the e = 00 casehas an average jet width of sim100 latitude

33 Effect of Varying Average Stellar Flux

An increase in average stellar flux leads to increases intemperature and wind speeds illustrated in Figure 8 Each caseshown has an eccentricity of 025 these maps have independentcolorscales to better show atmospheric structure The planetthat receives the highest average stellar flux throughout its orbitis hottest overall (bottom) with peak temperatures of 1300 KFurthermore because the planet with the highest 〈F 〉 orbits at thesmallest orbital distance it has a faster rotation rate and hencea narrower equatorial jet this comparison makes the effect ofrotation on the jet structure much more apparent

34 Global Average Temperature versus Eccentricity and 〈F 〉Here we illustrate the variation in time of maximum

planet-wide temperature with eccentricity and average stellarflux Figure 9 shows the globally averaged temperature foreach model integration plotted as a function of mean anomaly9

where M = 0 corresponds to periapse passage With increas-ing eccentricity the time at which the global-mean temperaturereaches its peak value decreases This is expected as a planet

9 The mean anomaly M is defined as (2πPorb)(t minus τ ) where τ is the timeof periapse passage

with a larger eccentricity will have a higher global-average tem-perature at periapse and hence shorter radiative time constant(τrad prop T minus3 see Showman et al 2011) Figure 10 plots the timeof peak temperature from periapse versus eccentricity Here theinverse relationship between eccentricity and the time of peaktemperature is clear

Figures 9 and 10 serve as a nice summary of the modelintegrations and their relation to observations as temperatureis one of the planetary properties that shapes light curvesat secondary eclipse Observers can use these predictions toestimate the timing of peak flux (and global temperature) basedon a given stellar flux andeccentricity (see Section 4)

35 Synchronous versus Pseudo-synchronous Rotation

Here we explore the sensitivity of the circulation to rota-tion rate by comparing models performed with the pseudo-synchronous rotation rate to otherwise identical models per-formed with the synchronous rotation rate This is illustratedin Figure 11 Here the rotation rate varies by almost a factorof three Both cases still have an equatorial superrotating jetand the overall atmospheric structure is similar However thereare subtle differences in the synchronous case the equatorialeastward winds and high-latitude westward winds are higher inmagnitude (right column) The synchronous case also exhibitsmore latitudinal and longitudinal temperature variation partic-ularly on the dayside this is more apparent at apoapse (bot-tom row) In the pseudo-synchronous case the faster rotationrate serves to homogenize the temperature in longitude becausethe time for rotation to carry air parcels from day to night isshorter The temperature difference from equator to pole is alsolarger because the faster rotation rate inhibits equator-to-poleheat transport Thus while it is unlikely that an eccentric planetsynchronously rotates its star the distinction can lead to notabledifferences in the mean flow

These results can also be used to compare the Hut (1981)formulation shown by the pseudo-synchronous case to for-mulations by other groups As shown in Figure 11 if anotherprescription for rotation rate differed from Hut (1981) by afactor of 2ndash3 (as illustrated by the synchronous case) the qual-itative picture would not look that different However if theprescription differed by a factor of more than 3ndash4 the picturecould substantially differ For example the rotation rate calcu-lated from Ivanov amp Papaloizou (2004) Equation (63) yields avalue of 14 times 106 s a factor of four larger than the Hut (1981)calculation So results using this prescription would look verydifferent

Still the rotation rate of eccentric exoplanets is a completelyunconstrained problem One must be careful then in choosinga formulation noting its assumptions and limitations

36 Momentum Budget and Equatorial Superrotation

In most of our simulations the planet maintains an equatorialeastward jet throughout its orbit Superrotation is commonin three-dimensional circulation models of hot Jupiters (egShowman amp Guillot 2002 Showman et al 2008 2009 Menouamp Rauscher 2009 Rauscher amp Menou 2010 2012 Cooperamp Showman 2005 2006 Dobbs-Dixon amp Lin 2008 Dobbs-Dixon et al 2010 Heng et al 2011) and is also seen withinour own solar system on Venus Titan Jupiter and SaturnWhile superrotation on solar system planets has been extensivelystudied only recently have the mechanisms for superrotation onsynchronously rotating exoplanets been identified (Showman amp

9

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)a= 00802 AU

minus200

minus200

200

200

200

600

600

1000

100014001800

22002600

(m sminus1)

0

500

1000

1500

2000

2500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00802 AU

(K)

550

600

650

700

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00505 AU

minus600

minus600

minus200 minus200

200

200

600

600

1000

140018002200260030003400

(m sminus1)

minus500

0

500

1000

1500

2000

2500

3000

3500

4000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00505 AU

(K)

700

750

800

850

900

950

1000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00318 AU

minus200

minus200

200

200

200 200

600

600

600

10001400

1800220026003000340038004200

5000

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00318 AU

(K)700

800

900

1000

1100

1200

1300

Figure 8 Comparison of simulations with e = 025 with decreasing average stellar flux from bottom to top panels Shown are plots of the orbit-averaged zonal-meanzonal wind (left column) and the wind and temperature profiles at 30 mbar at periapse (right column) Each plot has an independent colorscale The vertical bars inthe right column denote the substellar longitude The snapshots of wind and temperature correspond to model outputs of (from top to bottom) 3706 5006 and 2201Earth days

(A color version of this figure is available in the online journal)

10

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

0 05 1 15 2 25 3 35620

640

660

680

700

720

740ltFgt = 11617 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050e=075

0 05 1 15 2 25 3 35880

890

900

910

920

930

940

950

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

ltFgt = 73680 Wmminus1

e=025e=050

0 05 1 15 2 25 3 35 4960

980

1000

1020

1040

1060

1080

1100

1120

1140

1160ltFgt = 185691 Wmminus1

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

Mean anomaly

e=025e=050e=075

0 05 1 15 2 25 3 35 4 451100

1150

1200

1250

1300

1350ltFgt = 468183 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050

Figure 9 Plots of global average temperature vs mean anomaly for each modelintegration Each plot contains profiles of a given average stellar flux (468183185691 73680 and 11617 W mminus1) with blue green and red representingprofiles for e = 025 050 and 075 respectively

(A color version of this figure is available in the online journal)

03 04 05 06 070

05

1

15

2

25

3

eccentricity

Pea

k o

ffse

t fr

om

per

iap

se (

Ear

th d

ays)

468183 Wmminus1

185691 Wmminus1

73680 Wmminus1

11617 Wmminus1

Figure 10 Time after periapse passage at which the global-average temperaturepeaks (from Figure 9) vs eccentricity for all model integrations The bluegreen red and teal profiles correspond to average stellar fluxes of 468183185691 73680 and 11617 W mminus1 respectively

(A color version of this figure is available in the online journal)

Polvani 2011 hereafter SP2011) As discussed in their paperthe generation of superrotation requires up-gradient momentumtransport that pumps angular momentum from outside the jet towithin it Only transport by waves and eddies can satisfy thisrequirement (Hide 1969)

SP2011 demonstrate that for synchronous hot Jupiters onnear-circular orbits such as HD 189733b and HD 209458bequatorial superrotation is generated by standing planetaryscale Kelvin and Rossby waves at the equator and midlati-tudes respectively that are themselves the dynamical responseto the planetrsquos large dayndashnight heating contrast Because theKelvin waves propagate eastward and the Rossby waves prop-agate westward a phase tilt of the wind vectors emerges fromnorthwest to southeast in the northern hemisphere and south-west to northeast in the southern hemisphere with an over-all shape resembling a chevron pointing eastward This patterncauses meridional eddy angular momentum fluxes that convergeangular momentum onto the equator generating equatorialsuperrotation

Despite the fact that not all our model integrations aresynchronous our results suggest that similar mechanisms arestill at play The chevron shape that SP2011 describe is exactlywhat we noted earlier in our plots of the wind and temperatureprofiles from Figure 5 suggesting that as in SP2011 our modelsexhibit an equatorward flux of eddy angular momentum Thisis demonstrated quantitatively in Figure 12 which shows thezonally averaged meridional flux of relative zonal eddy angularmomentum uprimevprime cos φ at snapshots throughout the orbit (asin Figures 5 and 6) Here uprime and vprime represent the deviationof the zonal and meridional winds respectively from theirzonal averages Positive values indicate a northward flux ofeastward eddy angular momentum and negative values indicatea southward flux of eastward eddy angular momentum Atapoapse the small dayndashnight forcing causes poleward transportof eddy angular momentum at latitudes greater than 50 butequatorward transport of eddy angular momentum equator of50 latitude the amplitudes are weak due to the low insolationat apoapse As the planet moves closer to the star (middlerow left column) the increase in dayndashnight forcing strengthens

11

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)e = 050 pseudominussynchronous rotation

minus200

200

200

600

600

1000

10001400

180022002600

3000

3400

3800

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

synchronous rotation

minus600

minus600

minus200

minus200 minus200

minus200

minus200

200

200

600

10001400180022002600300034003800

4200

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

Figure 11 Comparison of pseudo-synchronous (Trot = 3046 times 105 s) vs synchronous (Trot = 8544 times 105 s) rotation for a case with Teq = 755 K e = 050 anda = 00848 AU Wind and temperature snapshots are taken at periapse (middle row) and apoapse (bottom row) at 5008 Earth days and 5013 Earth days respectivelyThe vertical bar denotes substellar longitude

(A color version of this figure is available in the online journal)

the equatorward flux of eastward eddy angular momentumat latitudes between 0 and 50 in both hemispheres Themomentum flux grows and persists until hours after periapseAngular momentum thus converges onto the equator andmaintains the equatorial jet against westward accelerationscaused by advection friction (if any) and Coriolis forces

These mechanisms also explain the difference in wind speedsbetween the synchronous and pseudo-synchronous cases inFigure 11 The synchronous case has a higher dayndashnight heatingcontrast hence a higher equatorward flux of eddy momentumleading to a stronger narrower equatorial jet Moreover theequatorial jet in the pseudo-synchronous case extends to deeper

12

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=18000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

Figure 12 Eddy momentum flux uprimevprime cos φ throughout the orbit of a model where Teq = 951 K e = 05 and a = 00534 AU Positive values indicate a northwardflux of eastward eddy angular momentum while negative values indicate a southward flux of eastward eddy angular momentum The maximum momentum fluxoccurs hours after periapse passage (f = 72) The snapshots correspond to model outputs from 6103 to 6108 Earth days

(A color version of this figure is available in the online journal)

13

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

Figure 13 Differing orbital viewing geometries explored in this study ω = 180 ω = 270 and ω = 360

(A color version of this figure is available in the online journal)

pressures than that of the synchronous case this is most likelydue to a change in momentum budget

4 OBSERVATIONAL IMPLICATIONS

Observations of transiting exoplanets in circular orbits areknown to be shaped by the atmospheric circulation As wasfirst predicted for hot Jupiters by Showman amp Guillot (2002)and later observationally confirmed for hot Jupiter HD 189733bby Knutson et al (2007) an equatorial superrotating jet causesan eastward displacement of the hot spot from the substellarpoint under appropriate conditions Hence a light curve showsa peak in infrared flux minutes to hours before secondaryeclipse We expect such a situation for eccentric exoplanetsas well However observations of eccentric planets can becomplicated because the flux received by the planet variesthroughout its orbit Previous studies explored such situationsbut have not used three-dimensional models (eg Langton ampLaughlin 2008a 2008b Cowan amp Agol 2011) We can expandon these studies and the three-dimensional study by Lewis et al(2010) by using our SPARC model results

We choose to generate light curves following the proceduresdescribed in Fortney et al (2006) for each model integrationat three values of the argument of periastron (ω) the anglebetween the radius vector to the ascending node and the periapseof the orbit In particular we choose orientations where transitoccurs before periapse (at f = minus90ω = 180) transit occursat apoapse (ω = 270) and transit occurs after periapse (atf = +90ω = 360) (Figure 13) These values reflect thebroad range of viewing geometries of transiting exoplanets forexample HAT-P-2b (e = 052) HAT-P-17b (e = 035) andHD 97658b (e = 013) have ω near 180 WASP-8b (e = 031)HAT-P-31b (e = 025) and HD 80606b (e = 093) have ω near270 and GJ 436b (e = 015) XO-3b (e = 026) and HAT-P-11b (e = 020) have ω near 360 (see exoplaneteu for a full listof values) Note that when transit occurs the nightside is visibleto Earth while the dayside is visible at the time of secondaryeclipse

41 Observational Effects of Varying ω

Using the nominal case (e = 050 a = 00534 AU Teq =951 K) we illustrate the dependence of full-orbit light curves onthe value of ω Figure 14 shows three light curves of the same

model integration but at the differing values of ω described inthe previous section Hence the differences in light curves resultsolely from the differing viewing geometries The planetstarflux ratio is plotted as a function of Earth days from periapsein the warm Spitzer wavelength bands (36 and 45 μm) andJHK bands (126 165 and 220 μm) The solid and dashedlines correspond to transit and secondary eclipse respectivelyFor the light curve where ω = 180 (top panel) the peak ininfrared (IR) flux occurs approximately 6 hr after periapse andbefore secondary eclipse The timing of the peak is due toa combination of the radiative lag of the planet and the factthat the hottest regions rotate into an Earth-facing orientationhours after periapse passage The dayside which is most visiblenear secondary eclipse is hottest after periapse Additionallytemperatures reach peak values at times close to when thosehottest regions are facing Earth This helps explain the highamplitude of the light curve peak in the ω = 180 case relative tothe other two cases These effects can be visualized by plottingthe temperature near the photosphere along the equator in apolar projection (Figure 15) Snapshots throughout the orbit plotthe temperature in a polar projection as a function of pressureand orbital position at a latitudinal slice near the equatorthe solid and dashed lines denote the substellar longitude andEarth-facing longitude respectively Based on the temperaturesat the Earth-facing longitude it is apparent that the peak in IRflux should occur after periapse and before secondary eclipse(see f = 12 and 72) this is indeed the case

For the same model but instead with ω = 270 the peak inIR flux occurs 1 hr after eclipse and periapse In this case thedayside is visible near periapse and therefore we are seeing theplanet increase in temperature (and IR flux) during its closestapproach to the star However the hottest regions are displacedeastward of the substellar point and therefore the hottest regionsface Earth significantly before periapse passage (Figure 15 atf asymp 72) Thus by the time the peak temperatures are reachednear and after periapse the hottest regions are already rotatingout of view from Earth This explains the lower amplitude ofthe light curve peak as compared to the peak of the ω = 180case The IR peak decreases as the planet moves further awayfrom the star along our line of sight A second ramp-up in fluxoccurs approximately 1 day after periapseeclipse this is due tothe hottest regions rotating back into view as the planet movestoward apoapse

14

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015P

lane

tSta

r F

lux

Rat

io

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

Figure 14 Three light curves of the IR flux (expressed as a planetstar fluxratio) vs time throughout the full orbit for a single model integration whereTeq = 951 K e = 050 and a = 00534 AU The flux is plotted in warm Spitzerbandpasses 36 and 45 μm and JHK bands (126 165 220 μm respectively)These three cases only differ by their value of ω shown in Figure 13 The solidlines denote transit while the dotted lines denote secondary eclipse

(A color version of this figure is available in the online journal)

In the case where ω = 360 two distinct peaks are presentmdashasmaller peak that occurs after secondary eclipse and sim3ndash4 hrbefore periapse and a second higher peak that occurs aftertransit The first peak occurs as some of the dayside is visible asthe planet is heated approaching periapse The peak decreasesas the dayside rotates out of view For this geometry the hottestregions are facing away from Earth when they reach their peaktemperatures (Figure 15) The second peak occurs after transitas the hottest regions again rotate into view At this point in theorbit peak temperatures have already occurred and thereforethis peak is lower in amplitude than the other two light curvesHowever because the planet is hotter than it was after secondary

eclipse this second peak has a higher amplitude than the firstone

42 High-e Model Integrations at Varying 〈F 〉If an eccentric planet has a large enough dayndashnight tempera-

ture difference and a sufficiently rapid rotation rate its resultantlight curve can show a periodic rise and fall in flux through-out its orbit The light curves shown in Figure 16 are twoe = 075 model integrations that differ in 〈F 〉 the top panelshows a case where 〈F 〉 = 468183 W mminus2 (Teq = 1199 K)and a = 00385 AU while the bottom panel is a case where〈F 〉 = 185691 W mminus2 (Teq = 951 K) and a = 00611 AU Forboth cases light curves are shown for the ω = 180 geome-try (see Figure 13) The differences in 〈F 〉 lead to differencesin light curve amplitude in the two models the closer hotterplanet exhibits a higher flux amplitude Not surprisingly thetiming of peak IR flux for both models resembles the ω = 180case in Figure 14 The peak in each case occurs after periapseand before eclipse due to the combined effects of the ther-mal lag of the planet and geometry of the system Additionallyboth light curves exhibit a quasi-periodic rise and fall of IR fluxdays after periapse passage This phenomenon called ldquoringingrdquowas also seen in models by Langton amp Laughlin (2008a) andCowan amp Agol (2011) caused by the hottest point of the planetrotating in and out of view from Earth For ringing to occurthe temperature difference from dayside to nightside must belarge and it must survive for multiple planetary rotation periodsIn the models shown in Figure 16 this is aided by the rapidpseudo-synchronous rotation rate associated with an eccentric-ity of 075 (Table 2) For the top case the dayndashnight tempera-ture difference is over sim1300 K while the bottom case variesby sim750 K (Figure 7) The light curves from Figure 14 do notexhibit ringing because the dayndashnight temperature difference islow (sim300ndash400 K)

Because the radiative and dynamical timescales in the modelsby Cowan amp Agol (2011) are calculated by a simple scaling withtemperature the ringing seen in their simulated light curvesoccurs with a period equal to the planetrsquos assumed solid-bodyrotation period in the inertial frame of the winds In realityan eccentric exoplanet approaching apoapse would becomeincreasingly cold the winds would weaken and decrease andhence the ringing should be non-periodic Indeed for ourthree dimensional models the ringing in each case has varyingperiodsmdashthe top case has a ringing period ranging from 03 to04 days (Trot = 035 days) while the bottom panel has a periodranging 07ndash08 days (Trot = 07 days)

Despite these differences this study and the studies byLangton amp Laughlin (2008a) and Cowan amp Agol (2011) illus-trate the influence of atmospheric dynamics on the observationsof eccentric exoplanets The light curve features described herecould be seen in future follow-up observations of eccentric ex-oplanets and observers can use the conclusions drawn aboveto relate the observations to the planetrsquos atmospheric structureFor example if ringing is present in an IR light curve the planetmust have a strong dayndashnight temperature asymmetry and hencewe would expect the planet to have strong superrotation at theequator (Showman amp Polvani 2011) This is probably the casefor highly eccentric (e gt 05) short-period exoplanets such asHD 17156b and HD 80606b If ringing were observed it wouldalso allow an estimate of the sum of the rotation speed and thesuperrotation speed thereby placing constraints on both windspeeds and rotation rates of eccentric hot Jupiters The timing ofpeak IR flux relative to transit and secondary eclipse and also

15

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

f=000deg

(K)800

850

900

950

1000

f=minus1242deg

(K)800

850

900

950

1000f=1242deg

(K)800

850

900

950

1000

f=minus7227deg

(K)800

850

900

950

1000f=7227deg

(K)800

850

900

950

1000

f=minus11988deg

(K)800

850

900

950

1000f=11988deg

(K)800

850

900

950

1000

f=minus18000deg

(K)800

850

900

950

1000

Figure 15 Plots of temperature vs longitude and pressure in equatorial slices throughout one full orbit for our nominal model with Teq = 951 K e = 050 anda = 00534 AU Each annulus shows temperature vs longitude and log-pressure for a snapshot at a particular orbital position (denoted by the value of f) as viewedfrom the planetrsquos north pole Pressures between 30 and 43 mbar are displayed in each annulus The different annuli mark the time evolution throughout one completeorbit starting from apoapse for the top annulus proceeding to periapse and then back to apoapse (follow arrows moving counterclockwise around the plot) Thesesnapshots correspond to model outputs from 6103 to 6108 Earth days The solid line in each snapshot indicates the longitude of the substellar point while the dashedline denotes the Earth-facing longitude for the particular case of ω = 180 For the cases of ω = 270 and 360 degrees exactly the same plot is valid except that thesub-Earth longitude would be aiming upward for ω = 270 and right for ω = 360

(A color version of this figure is available in the online journal)

16

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

e = 075 a = 00385 AU

-15 -10 -05 00 05 10 15Time after periapse passage (days)

0000

0001

0002

0003

0004

Pla

netS

tar

Flu

x R

atio

e = 075 a = 00611 AU

-3 -2 -1 0 1 2 3Time after periapse passage (days)

00000

00005

00010

00015

00020

00025

Pla

netS

tar

Flu

x R

atio

Figure 16 Light Curves for two different e = 075 model integrationsTeq = 1199 K a = 00385 AU (top) and Teq = 951 K a = 00611 AU(bottom) both with ω = 180 The solid line denotes transit while the dottedline denotes secondary eclipse

(A color version of this figure is available in the online journal)

relative to orbital position can help to constrain the circulationfurther

5 CONCLUSIONS

We present three-dimensional circulation models coupledwith a two-stream non-gray radiative transfer scheme fora number of theoretical eccentric hot Jupiters We haveshown that as in published models with zero eccentricity ourhigh-eccentricity circulation models are dominated by eastwardflow at photospheric levels which cause an eastward displace-ment of the hottest regions from the substellar point The rapidrotation rates associated with pseudo-synchronization at higheccentricity lead to a small Rossby deformation radius andin some cases multiple jets in the atmosphere Global-meantemperatures and dayndashnight temperature differences peak notat periapse but several hours afterward due to finite radiativetimescales in the planetrsquos atmosphere

Furthermore we show that equatorial superrotation is gener-ated and maintained by eddies formed by the strong dayndashnightheating contrast which induce a flux of momentum from mid-latitudes to the equator The eddy magnitudes and momentumfluxes peak just after periapse passage leading to variations inthe zonal-mean flow throughout the orbit

Figure 17 Globally averaged pressurendashtemperature profiles from 1-D radiative-convective radiative transfer models of HD 189733b with 11 frequency bins(red profiles) 30 frequency bins (green profiles) and 196 frequency bins (blackprofiles) From left to right the profiles vary in stellar flux from the nominalvalue as well as 2times 4times and 8times higher flux The 8times higher flux case has atemperature inversion

(A color version of this figure is available in the online journal)

Lastly we have shown that the spatial and temporal vari-ations of the wind and temperature structure as well as theorbital viewing geometry of the system with respect to Earthcan affect the time and amplitude of peak IR flux seen infull-orbit light curves Depending on the viewing geometry ofthe orbit relative to Earth we find that peaks in IR flux thateither lead or lag periapse are possible in all cases a combi-nation of temporal effects (temperatures changing over time)and geometric effects (hot spots rotating into or out of view)are important in controlling the timing and amplitude of theflux peaks In cases where the dayndashnight temperature contrastis large and the rotational period is short the light curve canalso exhibit ldquoringingrdquo in flux as the hottest region of the planetrotates in and out of view This ringing is non-periodic due tothe variation in stellar heating as a function of distance

This work was supported by NASA Origins and PlanetaryAtmospheres grants to APS TK also acknowledges supportfrom the Harriet P Jenkins Pre-Doctoral Fellowship Program(JPFP) Resources supporting this work were provided bythe NASA High-End Computing (HEC) Program through theNASA Advanced Supercomputing (NAS) Division at AmesResearch Center The authors thank the anonymous referee fortheir helpful comments and suggestions

APPENDIX

VALIDATION TESTS FOR UPDATEDRADIATIVE TRANSFER SCHEME

To validate our updated radiative transfer scheme using 11frequency bins we conducted the following tests

1 First we tested the sensitivity of the number of frequencybins in a one-dimensional radiative-convective model ofHD 189733b (a = 00313 AU e = 00 RP = 82396 times107 m g = 214 m sminus2) Figure 17 compares the globallyaveraged pressurendashtemperature profiles of models run with11 frequency bins 30 frequency bins and 196 frequencybins at (from left to right) 1times 2times 4times and 8times theaverage stellar flux of HD 189733b For each group of

17

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)30 opacity bins

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)

11 opacity bins

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

30 opacity bins

minus600

minus200minus200

200

200

600

6001000

1000

14001800

2200

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

11 opacity bins

minus600

minus200

minus200

200

200

600

600

10001000

1400

1800

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

30 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

11 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

Figure 18 Globally averaged pressurendashtemperature profiles (top row) zonal-mean zonal wind (middle row) and the windtemperature profiles at 30 mbar (bottomrow) for the 30-bin (left column) and 11-bin (right column) model integrations of HD 189733b Snapshots were output at 250 Earth days

(A color version of this figure is available in the online journal)

models the 11-bin profile agrees well with the 30- and196-bin cases from 1 μbar to 1 bar the temperature variesby only tens of K Below 10 bars the 30- and 11-binprofiles differ from the 196-bin cases by up to 50 Kdue to the loss of resolution at low wavenumbers Inthe case of the highest average flux the large amount of

heating leads to a temperature inversion that is not fullycaptured by the 11-bin model but can be improved witha better initial guess at the pressure-temperature (P-T)profile

2 After testing the new radiative transfer (RT) scheme forone-dimensional models we ran full three-dimensional

18

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

simulations of HD 189733b at its nominal average stellarflux using the SPARCMITgcm Figure 18 shows the glob-ally averaged P-T profiles (top row) zonal-mean zonal wind(middle row) and windtemperature profiles at 30 mbar(bottom row) for the 30-bin (left column) and 11-bin (rightcolumn) RT setups The globally averaged profiles arenearly identical with any temperature differences withina few K The zonal wind plots also show good agree-ment with similar peak speeds (gt2600 m sminus1) equatorialjet width (sim60) and jet level (lt100 bar) At 30 mbarthe 30- and 11-bin models share similar dayside-nightsidetemperature structure with a hot spot (1100 K) eastward ofthe substellar point and two colder regions on the nightsidein the mid-latitudes

Both tests show that the bulk circulation and temperaturestructure is retained in the transition from 30 to 11 frequencybins hence we proceed to use the SPARCMITgcm with thenew scheme for these and future model simulations

REFERENCES

Adcroft A Campin J-M Hill C amp Marshall J 2004 MWRv 132 2845Charbonneau D Knutson H A Barman T et al 2008 ApJ 686 1341Cho J Y-K Menou K Hansen B M S amp Seager S 2008 ApJ

675 817Cooper C S amp Showman A P 2005 ApJL 629 L45Cooper C S amp Showman A P 2006 ApJ 649 1048Cowan N B amp Agol E 2011 ApJ 726 82Deming D Harrington J Laughlin G et al 2007 ApJL 667 L199Dobbs-Dixon I Cumming A amp Lin D N C 2010 ApJ 710 1395Dobbs-Dixon I amp Lin D N C 2008 ApJ 673 513Fortney J J Cooper C S Showman A P Marley M S amp Freedman R S

2006 ApJ 652 746Fortney J J Lodders K Marley M S amp Freedman R S 2008 ApJ

678 1419Fortney J J Marley M S Lodders K Saumon D amp Freedman R

2005 ApJL 627 L69

Fu Q amp Liou K N 1992 JAtS 49 2140Goody R West R Chen L amp Crisp D 1989 JQSRT 42 539Greenberg R 2009 ApJL 698 L42Heng K 2012 ApJL 761 L1Heng K Frierson D M W amp Phillipps P J 2011 MNRAS 418 2669Hide R 1969 JAtS 26 841Holton J R 2004 An Introduction to Dynamic Meteorology (Waltham MA

Academic)Hut P 1981 AampA 99 126Irwin J Charbonneau D Nutzman P et al 2008 ApJ 681 636Ivanov P B amp Papaloizou J C B 2004 MNRAS 353 1161Knutson H A Charbonneau D Allen L E Burrows A amp Megeath S T

2008 ApJ 673 526Knutson H A Charbonneau D Allen L E et al 2007 Natur 447 183Knutson H A Charbonneau D Cowan N B et al 2009 ApJ 690 822Langton J amp Laughlin G 2008a ApJ 674 1106Langton J amp Laughlin G 2008b AampA 483 25Lewis N K Showman A P Fortney J J et al 2010 ApJ 720 344Lodders K 2003 ApJ 591 1220Marley M S amp McKay C P 1999 Icar 138 268Mayor M amp Queloz D 1995 Natur 378 355Menou K amp Rauscher E 2009 ApJ 700 887Mlawer E J Taubman S J Brown P D Iacono M J amp Clough S A

1997 JGR 102 16663Naef D Latham D W Mayor M et al 2001 AampA 375 L27Perna R Heng K amp Pont F 2012 ApJ 751 59Pont F Knutson H Gilliland R L Moutou C amp Charbonneau D

2008 MNRAS 385 109Rauscher E amp Menou K 2010 ApJ 714 1334Rauscher E amp Menou K 2012 ApJ 750 96Saumon D Marley M S Abel M Frommhold L amp Freedman R S

2012 ApJ 750 74Showman A P Cho J Y-K amp Menou K 2011 in Exoplanets ed S Seager

(Tucson AZ Univ Arizona Press) 471Showman A P Cooper C S Fortney J J amp Marley M S 2008 ApJ

682 559Showman A P Fortney J J Lian Y et al 2009 ApJ 699 564Showman A P amp Guillot T 2002 AampA 385 166Showman A P amp Polvani L M 2011 ApJ 738 71Swain M R Vasisht G amp Tinetti G 2008 Natur 452 329Thrastarson H T amp Cho J Y 2010 ApJ 716 144Wolszczan A 1994 Sci 264 538

19

  • 1 INTRODUCTION
  • 2 MODEL
    • 21 Updates to SPARCMITgcm
    • 22 Grid Simulations
      • 3 RESULTS
        • 31 Circulation Regime
        • 32 Effect of Eccentricity
        • 33 Effect of Varying Average Stellar Flux
        • 34 Global Average Temperature versus Eccentricity and F
        • 35 Synchronous versus Pseudo-synchronous Rotation
        • 36 Momentum Budget and Equatorial Superrotation
          • 4 OBSERVATIONAL IMPLICATIONS
            • 41 Observational Effects of Varying ω
            • 42 High-e Model Integrations at Varying F
              • 5 CONCLUSIONS
              • APPENDIX VALIDATION TESTS FOR UPDATED RADIATIVE TRANSFER SCHEME
              • REFERENCES
Page 7: THREE-DIMENSIONAL ATMOSPHERIC CIRCULATION OF HOT …showman/publications/kataria-etal-2013.pdf · planets’ atmospheric composition, temperature structure, and circulation. A fifth

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=000deg

minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus1242deg

minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=1242deg

minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus7227deg

minus200

minus200 200

200

200

600

600

1000

1000

1400

1400

18002200260030003400

3800

4200

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=7227deg

minus200

minus200minus200 200

200

200

600

600

1000

1000

1400

180022

002600300034003800

4200

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus11988deg

minus200

minus200

200

200

200

600

600

1000

1000

1400

18002200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=11988deg

minus600minus200

minus200

minus200

200

200

200

600

600

1000

1000

1400

1800

2200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus18000deg

minus200

minus200

200

200

600

600

1000

1000

1400

1800

2200260030003400

3800

(m sminus1)minus1000

0

1000

2000

3000

4000

5000

Figure 6 Snapshots of the zonal-mean zonal wind for the same integration described in Figure 5 (Teq = 951 K e = 050 and a = 00534 AU) Snapshots are againtaken from apoapse (f = minus1800) to hours after periapse (f = 119) and back again to apoapse Each panel is plotted on the same temperature colorscale Thesnapshots correspond to outputs from 6103 to 6108 Earth days Note that the equatorial superrotating jet is persistent throughout the orbit with peak speeds exceeding5000 m sminus1

(A color version of this figure is available in the online journal)

7

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 000

200

200

200600

600

600

1000

1000

10001400

1400

18002200

2600

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 000

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 025

minus600

minus600

minus200 minus200

200

200

600

600

1000

140018002200260030003400

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 025

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 050

minus200

minus200

200

200

600

600

1000

1000

1400

18002200

2600300034003800

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 050

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 075

200

200

200

200

600

600

600

1000

1000

1000

140018

002200260030003400

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 075

(K)

500

600

700

800

900

1000

1100

1200

1300

Figure 7 Comparison of simulations with Teq = 951 K with increasing eccentricity from e = 00 (top row) to e = 075 (bottom row) Shown are plots of thezonal-mean zonal wind averaged over a full orbit (left column) and snapshots of the wind and temperature profiles at 30 mbar (right column) The snapshots of thewind and temperature profiles in the right column were taken at periapse passage The vertical bars in the right column denote the longitude of the substellar pointThe snapshots of wind and temperature correspond to model outputs of (from top to bottom) 4100 5006 6106 and 6904 Earth days

(A color version of this figure is available in the online journal)

8

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

from 1000 K to 1300 K In addition the variation in temper-ature from dayside to nightside increases this strengthens thewinds within the equatorial superrotating jet Peak winds reachsim2500 m sminus1 in the circular case but increase to sim5000 m sminus1

in the high eccentricity caseAs the eccentricity increases the equatorial jet also narrows

in latitude In the case of highest eccentricity the atmospheredevelops westward midlatitude jets and eastward high-latitudejets The narrowing of the equatorial jet arises from the factthat jet width is confined to the Rossby radius of deformation(Showman amp Polvani 2011) the length scale over which theatmosphere adjusts to planetary-scale phenomena The Rossbyradius of deformation (Holton 2004) is defined at the equator as

LR =radic

c

β (6)

where c is the gravity wave speed and β is the derivative ofthe Coriolis parameter f with respect to latitudinal distancey β = df dy = 2Ωrot cos φRp where Ωrot is the planetaryrotation rate φ is the latitude and Rp is the planetary radius Atthe equator β is simply 2ΩrotRp Hence at the equator LR is

LR =radic

cRp

2Ωrot (7)

Therefore LR prop 1radic

Ωrot and an increased rotation rateleads to a narrower deformation radius From the Hut (1981)formulation the combined effect of the larger orbital period andincrease in eccentricity produces a shorter rotational periodThis in turn increases the rotation rate which reduces thedeformation radius For example the ratio between Ωrot forthe e = 00 case and the e = 075 case is 63 hence we wouldexpect that the equatorial jet should be narrower by a factor ofradic

63 sim 25 This is indeed the case as the e = 075 case hasan average jet width of sim40 latitude while the e = 00 casehas an average jet width of sim100 latitude

33 Effect of Varying Average Stellar Flux

An increase in average stellar flux leads to increases intemperature and wind speeds illustrated in Figure 8 Each caseshown has an eccentricity of 025 these maps have independentcolorscales to better show atmospheric structure The planetthat receives the highest average stellar flux throughout its orbitis hottest overall (bottom) with peak temperatures of 1300 KFurthermore because the planet with the highest 〈F 〉 orbits at thesmallest orbital distance it has a faster rotation rate and hencea narrower equatorial jet this comparison makes the effect ofrotation on the jet structure much more apparent

34 Global Average Temperature versus Eccentricity and 〈F 〉Here we illustrate the variation in time of maximum

planet-wide temperature with eccentricity and average stellarflux Figure 9 shows the globally averaged temperature foreach model integration plotted as a function of mean anomaly9

where M = 0 corresponds to periapse passage With increas-ing eccentricity the time at which the global-mean temperaturereaches its peak value decreases This is expected as a planet

9 The mean anomaly M is defined as (2πPorb)(t minus τ ) where τ is the timeof periapse passage

with a larger eccentricity will have a higher global-average tem-perature at periapse and hence shorter radiative time constant(τrad prop T minus3 see Showman et al 2011) Figure 10 plots the timeof peak temperature from periapse versus eccentricity Here theinverse relationship between eccentricity and the time of peaktemperature is clear

Figures 9 and 10 serve as a nice summary of the modelintegrations and their relation to observations as temperatureis one of the planetary properties that shapes light curvesat secondary eclipse Observers can use these predictions toestimate the timing of peak flux (and global temperature) basedon a given stellar flux andeccentricity (see Section 4)

35 Synchronous versus Pseudo-synchronous Rotation

Here we explore the sensitivity of the circulation to rota-tion rate by comparing models performed with the pseudo-synchronous rotation rate to otherwise identical models per-formed with the synchronous rotation rate This is illustratedin Figure 11 Here the rotation rate varies by almost a factorof three Both cases still have an equatorial superrotating jetand the overall atmospheric structure is similar However thereare subtle differences in the synchronous case the equatorialeastward winds and high-latitude westward winds are higher inmagnitude (right column) The synchronous case also exhibitsmore latitudinal and longitudinal temperature variation partic-ularly on the dayside this is more apparent at apoapse (bot-tom row) In the pseudo-synchronous case the faster rotationrate serves to homogenize the temperature in longitude becausethe time for rotation to carry air parcels from day to night isshorter The temperature difference from equator to pole is alsolarger because the faster rotation rate inhibits equator-to-poleheat transport Thus while it is unlikely that an eccentric planetsynchronously rotates its star the distinction can lead to notabledifferences in the mean flow

These results can also be used to compare the Hut (1981)formulation shown by the pseudo-synchronous case to for-mulations by other groups As shown in Figure 11 if anotherprescription for rotation rate differed from Hut (1981) by afactor of 2ndash3 (as illustrated by the synchronous case) the qual-itative picture would not look that different However if theprescription differed by a factor of more than 3ndash4 the picturecould substantially differ For example the rotation rate calcu-lated from Ivanov amp Papaloizou (2004) Equation (63) yields avalue of 14 times 106 s a factor of four larger than the Hut (1981)calculation So results using this prescription would look verydifferent

Still the rotation rate of eccentric exoplanets is a completelyunconstrained problem One must be careful then in choosinga formulation noting its assumptions and limitations

36 Momentum Budget and Equatorial Superrotation

In most of our simulations the planet maintains an equatorialeastward jet throughout its orbit Superrotation is commonin three-dimensional circulation models of hot Jupiters (egShowman amp Guillot 2002 Showman et al 2008 2009 Menouamp Rauscher 2009 Rauscher amp Menou 2010 2012 Cooperamp Showman 2005 2006 Dobbs-Dixon amp Lin 2008 Dobbs-Dixon et al 2010 Heng et al 2011) and is also seen withinour own solar system on Venus Titan Jupiter and SaturnWhile superrotation on solar system planets has been extensivelystudied only recently have the mechanisms for superrotation onsynchronously rotating exoplanets been identified (Showman amp

9

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)a= 00802 AU

minus200

minus200

200

200

200

600

600

1000

100014001800

22002600

(m sminus1)

0

500

1000

1500

2000

2500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00802 AU

(K)

550

600

650

700

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00505 AU

minus600

minus600

minus200 minus200

200

200

600

600

1000

140018002200260030003400

(m sminus1)

minus500

0

500

1000

1500

2000

2500

3000

3500

4000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00505 AU

(K)

700

750

800

850

900

950

1000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00318 AU

minus200

minus200

200

200

200 200

600

600

600

10001400

1800220026003000340038004200

5000

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00318 AU

(K)700

800

900

1000

1100

1200

1300

Figure 8 Comparison of simulations with e = 025 with decreasing average stellar flux from bottom to top panels Shown are plots of the orbit-averaged zonal-meanzonal wind (left column) and the wind and temperature profiles at 30 mbar at periapse (right column) Each plot has an independent colorscale The vertical bars inthe right column denote the substellar longitude The snapshots of wind and temperature correspond to model outputs of (from top to bottom) 3706 5006 and 2201Earth days

(A color version of this figure is available in the online journal)

10

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

0 05 1 15 2 25 3 35620

640

660

680

700

720

740ltFgt = 11617 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050e=075

0 05 1 15 2 25 3 35880

890

900

910

920

930

940

950

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

ltFgt = 73680 Wmminus1

e=025e=050

0 05 1 15 2 25 3 35 4960

980

1000

1020

1040

1060

1080

1100

1120

1140

1160ltFgt = 185691 Wmminus1

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

Mean anomaly

e=025e=050e=075

0 05 1 15 2 25 3 35 4 451100

1150

1200

1250

1300

1350ltFgt = 468183 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050

Figure 9 Plots of global average temperature vs mean anomaly for each modelintegration Each plot contains profiles of a given average stellar flux (468183185691 73680 and 11617 W mminus1) with blue green and red representingprofiles for e = 025 050 and 075 respectively

(A color version of this figure is available in the online journal)

03 04 05 06 070

05

1

15

2

25

3

eccentricity

Pea

k o

ffse

t fr

om

per

iap

se (

Ear

th d

ays)

468183 Wmminus1

185691 Wmminus1

73680 Wmminus1

11617 Wmminus1

Figure 10 Time after periapse passage at which the global-average temperaturepeaks (from Figure 9) vs eccentricity for all model integrations The bluegreen red and teal profiles correspond to average stellar fluxes of 468183185691 73680 and 11617 W mminus1 respectively

(A color version of this figure is available in the online journal)

Polvani 2011 hereafter SP2011) As discussed in their paperthe generation of superrotation requires up-gradient momentumtransport that pumps angular momentum from outside the jet towithin it Only transport by waves and eddies can satisfy thisrequirement (Hide 1969)

SP2011 demonstrate that for synchronous hot Jupiters onnear-circular orbits such as HD 189733b and HD 209458bequatorial superrotation is generated by standing planetaryscale Kelvin and Rossby waves at the equator and midlati-tudes respectively that are themselves the dynamical responseto the planetrsquos large dayndashnight heating contrast Because theKelvin waves propagate eastward and the Rossby waves prop-agate westward a phase tilt of the wind vectors emerges fromnorthwest to southeast in the northern hemisphere and south-west to northeast in the southern hemisphere with an over-all shape resembling a chevron pointing eastward This patterncauses meridional eddy angular momentum fluxes that convergeangular momentum onto the equator generating equatorialsuperrotation

Despite the fact that not all our model integrations aresynchronous our results suggest that similar mechanisms arestill at play The chevron shape that SP2011 describe is exactlywhat we noted earlier in our plots of the wind and temperatureprofiles from Figure 5 suggesting that as in SP2011 our modelsexhibit an equatorward flux of eddy angular momentum Thisis demonstrated quantitatively in Figure 12 which shows thezonally averaged meridional flux of relative zonal eddy angularmomentum uprimevprime cos φ at snapshots throughout the orbit (asin Figures 5 and 6) Here uprime and vprime represent the deviationof the zonal and meridional winds respectively from theirzonal averages Positive values indicate a northward flux ofeastward eddy angular momentum and negative values indicatea southward flux of eastward eddy angular momentum Atapoapse the small dayndashnight forcing causes poleward transportof eddy angular momentum at latitudes greater than 50 butequatorward transport of eddy angular momentum equator of50 latitude the amplitudes are weak due to the low insolationat apoapse As the planet moves closer to the star (middlerow left column) the increase in dayndashnight forcing strengthens

11

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)e = 050 pseudominussynchronous rotation

minus200

200

200

600

600

1000

10001400

180022002600

3000

3400

3800

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

synchronous rotation

minus600

minus600

minus200

minus200 minus200

minus200

minus200

200

200

600

10001400180022002600300034003800

4200

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

Figure 11 Comparison of pseudo-synchronous (Trot = 3046 times 105 s) vs synchronous (Trot = 8544 times 105 s) rotation for a case with Teq = 755 K e = 050 anda = 00848 AU Wind and temperature snapshots are taken at periapse (middle row) and apoapse (bottom row) at 5008 Earth days and 5013 Earth days respectivelyThe vertical bar denotes substellar longitude

(A color version of this figure is available in the online journal)

the equatorward flux of eastward eddy angular momentumat latitudes between 0 and 50 in both hemispheres Themomentum flux grows and persists until hours after periapseAngular momentum thus converges onto the equator andmaintains the equatorial jet against westward accelerationscaused by advection friction (if any) and Coriolis forces

These mechanisms also explain the difference in wind speedsbetween the synchronous and pseudo-synchronous cases inFigure 11 The synchronous case has a higher dayndashnight heatingcontrast hence a higher equatorward flux of eddy momentumleading to a stronger narrower equatorial jet Moreover theequatorial jet in the pseudo-synchronous case extends to deeper

12

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=18000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

Figure 12 Eddy momentum flux uprimevprime cos φ throughout the orbit of a model where Teq = 951 K e = 05 and a = 00534 AU Positive values indicate a northwardflux of eastward eddy angular momentum while negative values indicate a southward flux of eastward eddy angular momentum The maximum momentum fluxoccurs hours after periapse passage (f = 72) The snapshots correspond to model outputs from 6103 to 6108 Earth days

(A color version of this figure is available in the online journal)

13

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

Figure 13 Differing orbital viewing geometries explored in this study ω = 180 ω = 270 and ω = 360

(A color version of this figure is available in the online journal)

pressures than that of the synchronous case this is most likelydue to a change in momentum budget

4 OBSERVATIONAL IMPLICATIONS

Observations of transiting exoplanets in circular orbits areknown to be shaped by the atmospheric circulation As wasfirst predicted for hot Jupiters by Showman amp Guillot (2002)and later observationally confirmed for hot Jupiter HD 189733bby Knutson et al (2007) an equatorial superrotating jet causesan eastward displacement of the hot spot from the substellarpoint under appropriate conditions Hence a light curve showsa peak in infrared flux minutes to hours before secondaryeclipse We expect such a situation for eccentric exoplanetsas well However observations of eccentric planets can becomplicated because the flux received by the planet variesthroughout its orbit Previous studies explored such situationsbut have not used three-dimensional models (eg Langton ampLaughlin 2008a 2008b Cowan amp Agol 2011) We can expandon these studies and the three-dimensional study by Lewis et al(2010) by using our SPARC model results

We choose to generate light curves following the proceduresdescribed in Fortney et al (2006) for each model integrationat three values of the argument of periastron (ω) the anglebetween the radius vector to the ascending node and the periapseof the orbit In particular we choose orientations where transitoccurs before periapse (at f = minus90ω = 180) transit occursat apoapse (ω = 270) and transit occurs after periapse (atf = +90ω = 360) (Figure 13) These values reflect thebroad range of viewing geometries of transiting exoplanets forexample HAT-P-2b (e = 052) HAT-P-17b (e = 035) andHD 97658b (e = 013) have ω near 180 WASP-8b (e = 031)HAT-P-31b (e = 025) and HD 80606b (e = 093) have ω near270 and GJ 436b (e = 015) XO-3b (e = 026) and HAT-P-11b (e = 020) have ω near 360 (see exoplaneteu for a full listof values) Note that when transit occurs the nightside is visibleto Earth while the dayside is visible at the time of secondaryeclipse

41 Observational Effects of Varying ω

Using the nominal case (e = 050 a = 00534 AU Teq =951 K) we illustrate the dependence of full-orbit light curves onthe value of ω Figure 14 shows three light curves of the same

model integration but at the differing values of ω described inthe previous section Hence the differences in light curves resultsolely from the differing viewing geometries The planetstarflux ratio is plotted as a function of Earth days from periapsein the warm Spitzer wavelength bands (36 and 45 μm) andJHK bands (126 165 and 220 μm) The solid and dashedlines correspond to transit and secondary eclipse respectivelyFor the light curve where ω = 180 (top panel) the peak ininfrared (IR) flux occurs approximately 6 hr after periapse andbefore secondary eclipse The timing of the peak is due toa combination of the radiative lag of the planet and the factthat the hottest regions rotate into an Earth-facing orientationhours after periapse passage The dayside which is most visiblenear secondary eclipse is hottest after periapse Additionallytemperatures reach peak values at times close to when thosehottest regions are facing Earth This helps explain the highamplitude of the light curve peak in the ω = 180 case relative tothe other two cases These effects can be visualized by plottingthe temperature near the photosphere along the equator in apolar projection (Figure 15) Snapshots throughout the orbit plotthe temperature in a polar projection as a function of pressureand orbital position at a latitudinal slice near the equatorthe solid and dashed lines denote the substellar longitude andEarth-facing longitude respectively Based on the temperaturesat the Earth-facing longitude it is apparent that the peak in IRflux should occur after periapse and before secondary eclipse(see f = 12 and 72) this is indeed the case

For the same model but instead with ω = 270 the peak inIR flux occurs 1 hr after eclipse and periapse In this case thedayside is visible near periapse and therefore we are seeing theplanet increase in temperature (and IR flux) during its closestapproach to the star However the hottest regions are displacedeastward of the substellar point and therefore the hottest regionsface Earth significantly before periapse passage (Figure 15 atf asymp 72) Thus by the time the peak temperatures are reachednear and after periapse the hottest regions are already rotatingout of view from Earth This explains the lower amplitude ofthe light curve peak as compared to the peak of the ω = 180case The IR peak decreases as the planet moves further awayfrom the star along our line of sight A second ramp-up in fluxoccurs approximately 1 day after periapseeclipse this is due tothe hottest regions rotating back into view as the planet movestoward apoapse

14

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015P

lane

tSta

r F

lux

Rat

io

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

Figure 14 Three light curves of the IR flux (expressed as a planetstar fluxratio) vs time throughout the full orbit for a single model integration whereTeq = 951 K e = 050 and a = 00534 AU The flux is plotted in warm Spitzerbandpasses 36 and 45 μm and JHK bands (126 165 220 μm respectively)These three cases only differ by their value of ω shown in Figure 13 The solidlines denote transit while the dotted lines denote secondary eclipse

(A color version of this figure is available in the online journal)

In the case where ω = 360 two distinct peaks are presentmdashasmaller peak that occurs after secondary eclipse and sim3ndash4 hrbefore periapse and a second higher peak that occurs aftertransit The first peak occurs as some of the dayside is visible asthe planet is heated approaching periapse The peak decreasesas the dayside rotates out of view For this geometry the hottestregions are facing away from Earth when they reach their peaktemperatures (Figure 15) The second peak occurs after transitas the hottest regions again rotate into view At this point in theorbit peak temperatures have already occurred and thereforethis peak is lower in amplitude than the other two light curvesHowever because the planet is hotter than it was after secondary

eclipse this second peak has a higher amplitude than the firstone

42 High-e Model Integrations at Varying 〈F 〉If an eccentric planet has a large enough dayndashnight tempera-

ture difference and a sufficiently rapid rotation rate its resultantlight curve can show a periodic rise and fall in flux through-out its orbit The light curves shown in Figure 16 are twoe = 075 model integrations that differ in 〈F 〉 the top panelshows a case where 〈F 〉 = 468183 W mminus2 (Teq = 1199 K)and a = 00385 AU while the bottom panel is a case where〈F 〉 = 185691 W mminus2 (Teq = 951 K) and a = 00611 AU Forboth cases light curves are shown for the ω = 180 geome-try (see Figure 13) The differences in 〈F 〉 lead to differencesin light curve amplitude in the two models the closer hotterplanet exhibits a higher flux amplitude Not surprisingly thetiming of peak IR flux for both models resembles the ω = 180case in Figure 14 The peak in each case occurs after periapseand before eclipse due to the combined effects of the ther-mal lag of the planet and geometry of the system Additionallyboth light curves exhibit a quasi-periodic rise and fall of IR fluxdays after periapse passage This phenomenon called ldquoringingrdquowas also seen in models by Langton amp Laughlin (2008a) andCowan amp Agol (2011) caused by the hottest point of the planetrotating in and out of view from Earth For ringing to occurthe temperature difference from dayside to nightside must belarge and it must survive for multiple planetary rotation periodsIn the models shown in Figure 16 this is aided by the rapidpseudo-synchronous rotation rate associated with an eccentric-ity of 075 (Table 2) For the top case the dayndashnight tempera-ture difference is over sim1300 K while the bottom case variesby sim750 K (Figure 7) The light curves from Figure 14 do notexhibit ringing because the dayndashnight temperature difference islow (sim300ndash400 K)

Because the radiative and dynamical timescales in the modelsby Cowan amp Agol (2011) are calculated by a simple scaling withtemperature the ringing seen in their simulated light curvesoccurs with a period equal to the planetrsquos assumed solid-bodyrotation period in the inertial frame of the winds In realityan eccentric exoplanet approaching apoapse would becomeincreasingly cold the winds would weaken and decrease andhence the ringing should be non-periodic Indeed for ourthree dimensional models the ringing in each case has varyingperiodsmdashthe top case has a ringing period ranging from 03 to04 days (Trot = 035 days) while the bottom panel has a periodranging 07ndash08 days (Trot = 07 days)

Despite these differences this study and the studies byLangton amp Laughlin (2008a) and Cowan amp Agol (2011) illus-trate the influence of atmospheric dynamics on the observationsof eccentric exoplanets The light curve features described herecould be seen in future follow-up observations of eccentric ex-oplanets and observers can use the conclusions drawn aboveto relate the observations to the planetrsquos atmospheric structureFor example if ringing is present in an IR light curve the planetmust have a strong dayndashnight temperature asymmetry and hencewe would expect the planet to have strong superrotation at theequator (Showman amp Polvani 2011) This is probably the casefor highly eccentric (e gt 05) short-period exoplanets such asHD 17156b and HD 80606b If ringing were observed it wouldalso allow an estimate of the sum of the rotation speed and thesuperrotation speed thereby placing constraints on both windspeeds and rotation rates of eccentric hot Jupiters The timing ofpeak IR flux relative to transit and secondary eclipse and also

15

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

f=000deg

(K)800

850

900

950

1000

f=minus1242deg

(K)800

850

900

950

1000f=1242deg

(K)800

850

900

950

1000

f=minus7227deg

(K)800

850

900

950

1000f=7227deg

(K)800

850

900

950

1000

f=minus11988deg

(K)800

850

900

950

1000f=11988deg

(K)800

850

900

950

1000

f=minus18000deg

(K)800

850

900

950

1000

Figure 15 Plots of temperature vs longitude and pressure in equatorial slices throughout one full orbit for our nominal model with Teq = 951 K e = 050 anda = 00534 AU Each annulus shows temperature vs longitude and log-pressure for a snapshot at a particular orbital position (denoted by the value of f) as viewedfrom the planetrsquos north pole Pressures between 30 and 43 mbar are displayed in each annulus The different annuli mark the time evolution throughout one completeorbit starting from apoapse for the top annulus proceeding to periapse and then back to apoapse (follow arrows moving counterclockwise around the plot) Thesesnapshots correspond to model outputs from 6103 to 6108 Earth days The solid line in each snapshot indicates the longitude of the substellar point while the dashedline denotes the Earth-facing longitude for the particular case of ω = 180 For the cases of ω = 270 and 360 degrees exactly the same plot is valid except that thesub-Earth longitude would be aiming upward for ω = 270 and right for ω = 360

(A color version of this figure is available in the online journal)

16

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

e = 075 a = 00385 AU

-15 -10 -05 00 05 10 15Time after periapse passage (days)

0000

0001

0002

0003

0004

Pla

netS

tar

Flu

x R

atio

e = 075 a = 00611 AU

-3 -2 -1 0 1 2 3Time after periapse passage (days)

00000

00005

00010

00015

00020

00025

Pla

netS

tar

Flu

x R

atio

Figure 16 Light Curves for two different e = 075 model integrationsTeq = 1199 K a = 00385 AU (top) and Teq = 951 K a = 00611 AU(bottom) both with ω = 180 The solid line denotes transit while the dottedline denotes secondary eclipse

(A color version of this figure is available in the online journal)

relative to orbital position can help to constrain the circulationfurther

5 CONCLUSIONS

We present three-dimensional circulation models coupledwith a two-stream non-gray radiative transfer scheme fora number of theoretical eccentric hot Jupiters We haveshown that as in published models with zero eccentricity ourhigh-eccentricity circulation models are dominated by eastwardflow at photospheric levels which cause an eastward displace-ment of the hottest regions from the substellar point The rapidrotation rates associated with pseudo-synchronization at higheccentricity lead to a small Rossby deformation radius andin some cases multiple jets in the atmosphere Global-meantemperatures and dayndashnight temperature differences peak notat periapse but several hours afterward due to finite radiativetimescales in the planetrsquos atmosphere

Furthermore we show that equatorial superrotation is gener-ated and maintained by eddies formed by the strong dayndashnightheating contrast which induce a flux of momentum from mid-latitudes to the equator The eddy magnitudes and momentumfluxes peak just after periapse passage leading to variations inthe zonal-mean flow throughout the orbit

Figure 17 Globally averaged pressurendashtemperature profiles from 1-D radiative-convective radiative transfer models of HD 189733b with 11 frequency bins(red profiles) 30 frequency bins (green profiles) and 196 frequency bins (blackprofiles) From left to right the profiles vary in stellar flux from the nominalvalue as well as 2times 4times and 8times higher flux The 8times higher flux case has atemperature inversion

(A color version of this figure is available in the online journal)

Lastly we have shown that the spatial and temporal vari-ations of the wind and temperature structure as well as theorbital viewing geometry of the system with respect to Earthcan affect the time and amplitude of peak IR flux seen infull-orbit light curves Depending on the viewing geometry ofthe orbit relative to Earth we find that peaks in IR flux thateither lead or lag periapse are possible in all cases a combi-nation of temporal effects (temperatures changing over time)and geometric effects (hot spots rotating into or out of view)are important in controlling the timing and amplitude of theflux peaks In cases where the dayndashnight temperature contrastis large and the rotational period is short the light curve canalso exhibit ldquoringingrdquo in flux as the hottest region of the planetrotates in and out of view This ringing is non-periodic due tothe variation in stellar heating as a function of distance

This work was supported by NASA Origins and PlanetaryAtmospheres grants to APS TK also acknowledges supportfrom the Harriet P Jenkins Pre-Doctoral Fellowship Program(JPFP) Resources supporting this work were provided bythe NASA High-End Computing (HEC) Program through theNASA Advanced Supercomputing (NAS) Division at AmesResearch Center The authors thank the anonymous referee fortheir helpful comments and suggestions

APPENDIX

VALIDATION TESTS FOR UPDATEDRADIATIVE TRANSFER SCHEME

To validate our updated radiative transfer scheme using 11frequency bins we conducted the following tests

1 First we tested the sensitivity of the number of frequencybins in a one-dimensional radiative-convective model ofHD 189733b (a = 00313 AU e = 00 RP = 82396 times107 m g = 214 m sminus2) Figure 17 compares the globallyaveraged pressurendashtemperature profiles of models run with11 frequency bins 30 frequency bins and 196 frequencybins at (from left to right) 1times 2times 4times and 8times theaverage stellar flux of HD 189733b For each group of

17

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)30 opacity bins

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)

11 opacity bins

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

30 opacity bins

minus600

minus200minus200

200

200

600

6001000

1000

14001800

2200

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

11 opacity bins

minus600

minus200

minus200

200

200

600

600

10001000

1400

1800

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

30 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

11 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

Figure 18 Globally averaged pressurendashtemperature profiles (top row) zonal-mean zonal wind (middle row) and the windtemperature profiles at 30 mbar (bottomrow) for the 30-bin (left column) and 11-bin (right column) model integrations of HD 189733b Snapshots were output at 250 Earth days

(A color version of this figure is available in the online journal)

models the 11-bin profile agrees well with the 30- and196-bin cases from 1 μbar to 1 bar the temperature variesby only tens of K Below 10 bars the 30- and 11-binprofiles differ from the 196-bin cases by up to 50 Kdue to the loss of resolution at low wavenumbers Inthe case of the highest average flux the large amount of

heating leads to a temperature inversion that is not fullycaptured by the 11-bin model but can be improved witha better initial guess at the pressure-temperature (P-T)profile

2 After testing the new radiative transfer (RT) scheme forone-dimensional models we ran full three-dimensional

18

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

simulations of HD 189733b at its nominal average stellarflux using the SPARCMITgcm Figure 18 shows the glob-ally averaged P-T profiles (top row) zonal-mean zonal wind(middle row) and windtemperature profiles at 30 mbar(bottom row) for the 30-bin (left column) and 11-bin (rightcolumn) RT setups The globally averaged profiles arenearly identical with any temperature differences withina few K The zonal wind plots also show good agree-ment with similar peak speeds (gt2600 m sminus1) equatorialjet width (sim60) and jet level (lt100 bar) At 30 mbarthe 30- and 11-bin models share similar dayside-nightsidetemperature structure with a hot spot (1100 K) eastward ofthe substellar point and two colder regions on the nightsidein the mid-latitudes

Both tests show that the bulk circulation and temperaturestructure is retained in the transition from 30 to 11 frequencybins hence we proceed to use the SPARCMITgcm with thenew scheme for these and future model simulations

REFERENCES

Adcroft A Campin J-M Hill C amp Marshall J 2004 MWRv 132 2845Charbonneau D Knutson H A Barman T et al 2008 ApJ 686 1341Cho J Y-K Menou K Hansen B M S amp Seager S 2008 ApJ

675 817Cooper C S amp Showman A P 2005 ApJL 629 L45Cooper C S amp Showman A P 2006 ApJ 649 1048Cowan N B amp Agol E 2011 ApJ 726 82Deming D Harrington J Laughlin G et al 2007 ApJL 667 L199Dobbs-Dixon I Cumming A amp Lin D N C 2010 ApJ 710 1395Dobbs-Dixon I amp Lin D N C 2008 ApJ 673 513Fortney J J Cooper C S Showman A P Marley M S amp Freedman R S

2006 ApJ 652 746Fortney J J Lodders K Marley M S amp Freedman R S 2008 ApJ

678 1419Fortney J J Marley M S Lodders K Saumon D amp Freedman R

2005 ApJL 627 L69

Fu Q amp Liou K N 1992 JAtS 49 2140Goody R West R Chen L amp Crisp D 1989 JQSRT 42 539Greenberg R 2009 ApJL 698 L42Heng K 2012 ApJL 761 L1Heng K Frierson D M W amp Phillipps P J 2011 MNRAS 418 2669Hide R 1969 JAtS 26 841Holton J R 2004 An Introduction to Dynamic Meteorology (Waltham MA

Academic)Hut P 1981 AampA 99 126Irwin J Charbonneau D Nutzman P et al 2008 ApJ 681 636Ivanov P B amp Papaloizou J C B 2004 MNRAS 353 1161Knutson H A Charbonneau D Allen L E Burrows A amp Megeath S T

2008 ApJ 673 526Knutson H A Charbonneau D Allen L E et al 2007 Natur 447 183Knutson H A Charbonneau D Cowan N B et al 2009 ApJ 690 822Langton J amp Laughlin G 2008a ApJ 674 1106Langton J amp Laughlin G 2008b AampA 483 25Lewis N K Showman A P Fortney J J et al 2010 ApJ 720 344Lodders K 2003 ApJ 591 1220Marley M S amp McKay C P 1999 Icar 138 268Mayor M amp Queloz D 1995 Natur 378 355Menou K amp Rauscher E 2009 ApJ 700 887Mlawer E J Taubman S J Brown P D Iacono M J amp Clough S A

1997 JGR 102 16663Naef D Latham D W Mayor M et al 2001 AampA 375 L27Perna R Heng K amp Pont F 2012 ApJ 751 59Pont F Knutson H Gilliland R L Moutou C amp Charbonneau D

2008 MNRAS 385 109Rauscher E amp Menou K 2010 ApJ 714 1334Rauscher E amp Menou K 2012 ApJ 750 96Saumon D Marley M S Abel M Frommhold L amp Freedman R S

2012 ApJ 750 74Showman A P Cho J Y-K amp Menou K 2011 in Exoplanets ed S Seager

(Tucson AZ Univ Arizona Press) 471Showman A P Cooper C S Fortney J J amp Marley M S 2008 ApJ

682 559Showman A P Fortney J J Lian Y et al 2009 ApJ 699 564Showman A P amp Guillot T 2002 AampA 385 166Showman A P amp Polvani L M 2011 ApJ 738 71Swain M R Vasisht G amp Tinetti G 2008 Natur 452 329Thrastarson H T amp Cho J Y 2010 ApJ 716 144Wolszczan A 1994 Sci 264 538

19

  • 1 INTRODUCTION
  • 2 MODEL
    • 21 Updates to SPARCMITgcm
    • 22 Grid Simulations
      • 3 RESULTS
        • 31 Circulation Regime
        • 32 Effect of Eccentricity
        • 33 Effect of Varying Average Stellar Flux
        • 34 Global Average Temperature versus Eccentricity and F
        • 35 Synchronous versus Pseudo-synchronous Rotation
        • 36 Momentum Budget and Equatorial Superrotation
          • 4 OBSERVATIONAL IMPLICATIONS
            • 41 Observational Effects of Varying ω
            • 42 High-e Model Integrations at Varying F
              • 5 CONCLUSIONS
              • APPENDIX VALIDATION TESTS FOR UPDATED RADIATIVE TRANSFER SCHEME
              • REFERENCES
Page 8: THREE-DIMENSIONAL ATMOSPHERIC CIRCULATION OF HOT …showman/publications/kataria-etal-2013.pdf · planets’ atmospheric composition, temperature structure, and circulation. A fifth

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 000

200

200

200600

600

600

1000

1000

10001400

1400

18002200

2600

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 000

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 025

minus600

minus600

minus200 minus200

200

200

600

600

1000

140018002200260030003400

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 025

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 050

minus200

minus200

200

200

600

600

1000

1000

1400

18002200

2600300034003800

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 050

(K)

500

600

700

800

900

1000

1100

1200

1300

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

e= 075

200

200

200

200

600

600

600

1000

1000

1000

140018

002200260030003400

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e= 075

(K)

500

600

700

800

900

1000

1100

1200

1300

Figure 7 Comparison of simulations with Teq = 951 K with increasing eccentricity from e = 00 (top row) to e = 075 (bottom row) Shown are plots of thezonal-mean zonal wind averaged over a full orbit (left column) and snapshots of the wind and temperature profiles at 30 mbar (right column) The snapshots of thewind and temperature profiles in the right column were taken at periapse passage The vertical bars in the right column denote the longitude of the substellar pointThe snapshots of wind and temperature correspond to model outputs of (from top to bottom) 4100 5006 6106 and 6904 Earth days

(A color version of this figure is available in the online journal)

8

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

from 1000 K to 1300 K In addition the variation in temper-ature from dayside to nightside increases this strengthens thewinds within the equatorial superrotating jet Peak winds reachsim2500 m sminus1 in the circular case but increase to sim5000 m sminus1

in the high eccentricity caseAs the eccentricity increases the equatorial jet also narrows

in latitude In the case of highest eccentricity the atmospheredevelops westward midlatitude jets and eastward high-latitudejets The narrowing of the equatorial jet arises from the factthat jet width is confined to the Rossby radius of deformation(Showman amp Polvani 2011) the length scale over which theatmosphere adjusts to planetary-scale phenomena The Rossbyradius of deformation (Holton 2004) is defined at the equator as

LR =radic

c

β (6)

where c is the gravity wave speed and β is the derivative ofthe Coriolis parameter f with respect to latitudinal distancey β = df dy = 2Ωrot cos φRp where Ωrot is the planetaryrotation rate φ is the latitude and Rp is the planetary radius Atthe equator β is simply 2ΩrotRp Hence at the equator LR is

LR =radic

cRp

2Ωrot (7)

Therefore LR prop 1radic

Ωrot and an increased rotation rateleads to a narrower deformation radius From the Hut (1981)formulation the combined effect of the larger orbital period andincrease in eccentricity produces a shorter rotational periodThis in turn increases the rotation rate which reduces thedeformation radius For example the ratio between Ωrot forthe e = 00 case and the e = 075 case is 63 hence we wouldexpect that the equatorial jet should be narrower by a factor ofradic

63 sim 25 This is indeed the case as the e = 075 case hasan average jet width of sim40 latitude while the e = 00 casehas an average jet width of sim100 latitude

33 Effect of Varying Average Stellar Flux

An increase in average stellar flux leads to increases intemperature and wind speeds illustrated in Figure 8 Each caseshown has an eccentricity of 025 these maps have independentcolorscales to better show atmospheric structure The planetthat receives the highest average stellar flux throughout its orbitis hottest overall (bottom) with peak temperatures of 1300 KFurthermore because the planet with the highest 〈F 〉 orbits at thesmallest orbital distance it has a faster rotation rate and hencea narrower equatorial jet this comparison makes the effect ofrotation on the jet structure much more apparent

34 Global Average Temperature versus Eccentricity and 〈F 〉Here we illustrate the variation in time of maximum

planet-wide temperature with eccentricity and average stellarflux Figure 9 shows the globally averaged temperature foreach model integration plotted as a function of mean anomaly9

where M = 0 corresponds to periapse passage With increas-ing eccentricity the time at which the global-mean temperaturereaches its peak value decreases This is expected as a planet

9 The mean anomaly M is defined as (2πPorb)(t minus τ ) where τ is the timeof periapse passage

with a larger eccentricity will have a higher global-average tem-perature at periapse and hence shorter radiative time constant(τrad prop T minus3 see Showman et al 2011) Figure 10 plots the timeof peak temperature from periapse versus eccentricity Here theinverse relationship between eccentricity and the time of peaktemperature is clear

Figures 9 and 10 serve as a nice summary of the modelintegrations and their relation to observations as temperatureis one of the planetary properties that shapes light curvesat secondary eclipse Observers can use these predictions toestimate the timing of peak flux (and global temperature) basedon a given stellar flux andeccentricity (see Section 4)

35 Synchronous versus Pseudo-synchronous Rotation

Here we explore the sensitivity of the circulation to rota-tion rate by comparing models performed with the pseudo-synchronous rotation rate to otherwise identical models per-formed with the synchronous rotation rate This is illustratedin Figure 11 Here the rotation rate varies by almost a factorof three Both cases still have an equatorial superrotating jetand the overall atmospheric structure is similar However thereare subtle differences in the synchronous case the equatorialeastward winds and high-latitude westward winds are higher inmagnitude (right column) The synchronous case also exhibitsmore latitudinal and longitudinal temperature variation partic-ularly on the dayside this is more apparent at apoapse (bot-tom row) In the pseudo-synchronous case the faster rotationrate serves to homogenize the temperature in longitude becausethe time for rotation to carry air parcels from day to night isshorter The temperature difference from equator to pole is alsolarger because the faster rotation rate inhibits equator-to-poleheat transport Thus while it is unlikely that an eccentric planetsynchronously rotates its star the distinction can lead to notabledifferences in the mean flow

These results can also be used to compare the Hut (1981)formulation shown by the pseudo-synchronous case to for-mulations by other groups As shown in Figure 11 if anotherprescription for rotation rate differed from Hut (1981) by afactor of 2ndash3 (as illustrated by the synchronous case) the qual-itative picture would not look that different However if theprescription differed by a factor of more than 3ndash4 the picturecould substantially differ For example the rotation rate calcu-lated from Ivanov amp Papaloizou (2004) Equation (63) yields avalue of 14 times 106 s a factor of four larger than the Hut (1981)calculation So results using this prescription would look verydifferent

Still the rotation rate of eccentric exoplanets is a completelyunconstrained problem One must be careful then in choosinga formulation noting its assumptions and limitations

36 Momentum Budget and Equatorial Superrotation

In most of our simulations the planet maintains an equatorialeastward jet throughout its orbit Superrotation is commonin three-dimensional circulation models of hot Jupiters (egShowman amp Guillot 2002 Showman et al 2008 2009 Menouamp Rauscher 2009 Rauscher amp Menou 2010 2012 Cooperamp Showman 2005 2006 Dobbs-Dixon amp Lin 2008 Dobbs-Dixon et al 2010 Heng et al 2011) and is also seen withinour own solar system on Venus Titan Jupiter and SaturnWhile superrotation on solar system planets has been extensivelystudied only recently have the mechanisms for superrotation onsynchronously rotating exoplanets been identified (Showman amp

9

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)a= 00802 AU

minus200

minus200

200

200

200

600

600

1000

100014001800

22002600

(m sminus1)

0

500

1000

1500

2000

2500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00802 AU

(K)

550

600

650

700

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00505 AU

minus600

minus600

minus200 minus200

200

200

600

600

1000

140018002200260030003400

(m sminus1)

minus500

0

500

1000

1500

2000

2500

3000

3500

4000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00505 AU

(K)

700

750

800

850

900

950

1000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00318 AU

minus200

minus200

200

200

200 200

600

600

600

10001400

1800220026003000340038004200

5000

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00318 AU

(K)700

800

900

1000

1100

1200

1300

Figure 8 Comparison of simulations with e = 025 with decreasing average stellar flux from bottom to top panels Shown are plots of the orbit-averaged zonal-meanzonal wind (left column) and the wind and temperature profiles at 30 mbar at periapse (right column) Each plot has an independent colorscale The vertical bars inthe right column denote the substellar longitude The snapshots of wind and temperature correspond to model outputs of (from top to bottom) 3706 5006 and 2201Earth days

(A color version of this figure is available in the online journal)

10

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

0 05 1 15 2 25 3 35620

640

660

680

700

720

740ltFgt = 11617 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050e=075

0 05 1 15 2 25 3 35880

890

900

910

920

930

940

950

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

ltFgt = 73680 Wmminus1

e=025e=050

0 05 1 15 2 25 3 35 4960

980

1000

1020

1040

1060

1080

1100

1120

1140

1160ltFgt = 185691 Wmminus1

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

Mean anomaly

e=025e=050e=075

0 05 1 15 2 25 3 35 4 451100

1150

1200

1250

1300

1350ltFgt = 468183 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050

Figure 9 Plots of global average temperature vs mean anomaly for each modelintegration Each plot contains profiles of a given average stellar flux (468183185691 73680 and 11617 W mminus1) with blue green and red representingprofiles for e = 025 050 and 075 respectively

(A color version of this figure is available in the online journal)

03 04 05 06 070

05

1

15

2

25

3

eccentricity

Pea

k o

ffse

t fr

om

per

iap

se (

Ear

th d

ays)

468183 Wmminus1

185691 Wmminus1

73680 Wmminus1

11617 Wmminus1

Figure 10 Time after periapse passage at which the global-average temperaturepeaks (from Figure 9) vs eccentricity for all model integrations The bluegreen red and teal profiles correspond to average stellar fluxes of 468183185691 73680 and 11617 W mminus1 respectively

(A color version of this figure is available in the online journal)

Polvani 2011 hereafter SP2011) As discussed in their paperthe generation of superrotation requires up-gradient momentumtransport that pumps angular momentum from outside the jet towithin it Only transport by waves and eddies can satisfy thisrequirement (Hide 1969)

SP2011 demonstrate that for synchronous hot Jupiters onnear-circular orbits such as HD 189733b and HD 209458bequatorial superrotation is generated by standing planetaryscale Kelvin and Rossby waves at the equator and midlati-tudes respectively that are themselves the dynamical responseto the planetrsquos large dayndashnight heating contrast Because theKelvin waves propagate eastward and the Rossby waves prop-agate westward a phase tilt of the wind vectors emerges fromnorthwest to southeast in the northern hemisphere and south-west to northeast in the southern hemisphere with an over-all shape resembling a chevron pointing eastward This patterncauses meridional eddy angular momentum fluxes that convergeangular momentum onto the equator generating equatorialsuperrotation

Despite the fact that not all our model integrations aresynchronous our results suggest that similar mechanisms arestill at play The chevron shape that SP2011 describe is exactlywhat we noted earlier in our plots of the wind and temperatureprofiles from Figure 5 suggesting that as in SP2011 our modelsexhibit an equatorward flux of eddy angular momentum Thisis demonstrated quantitatively in Figure 12 which shows thezonally averaged meridional flux of relative zonal eddy angularmomentum uprimevprime cos φ at snapshots throughout the orbit (asin Figures 5 and 6) Here uprime and vprime represent the deviationof the zonal and meridional winds respectively from theirzonal averages Positive values indicate a northward flux ofeastward eddy angular momentum and negative values indicatea southward flux of eastward eddy angular momentum Atapoapse the small dayndashnight forcing causes poleward transportof eddy angular momentum at latitudes greater than 50 butequatorward transport of eddy angular momentum equator of50 latitude the amplitudes are weak due to the low insolationat apoapse As the planet moves closer to the star (middlerow left column) the increase in dayndashnight forcing strengthens

11

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)e = 050 pseudominussynchronous rotation

minus200

200

200

600

600

1000

10001400

180022002600

3000

3400

3800

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

synchronous rotation

minus600

minus600

minus200

minus200 minus200

minus200

minus200

200

200

600

10001400180022002600300034003800

4200

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

Figure 11 Comparison of pseudo-synchronous (Trot = 3046 times 105 s) vs synchronous (Trot = 8544 times 105 s) rotation for a case with Teq = 755 K e = 050 anda = 00848 AU Wind and temperature snapshots are taken at periapse (middle row) and apoapse (bottom row) at 5008 Earth days and 5013 Earth days respectivelyThe vertical bar denotes substellar longitude

(A color version of this figure is available in the online journal)

the equatorward flux of eastward eddy angular momentumat latitudes between 0 and 50 in both hemispheres Themomentum flux grows and persists until hours after periapseAngular momentum thus converges onto the equator andmaintains the equatorial jet against westward accelerationscaused by advection friction (if any) and Coriolis forces

These mechanisms also explain the difference in wind speedsbetween the synchronous and pseudo-synchronous cases inFigure 11 The synchronous case has a higher dayndashnight heatingcontrast hence a higher equatorward flux of eddy momentumleading to a stronger narrower equatorial jet Moreover theequatorial jet in the pseudo-synchronous case extends to deeper

12

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=18000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

Figure 12 Eddy momentum flux uprimevprime cos φ throughout the orbit of a model where Teq = 951 K e = 05 and a = 00534 AU Positive values indicate a northwardflux of eastward eddy angular momentum while negative values indicate a southward flux of eastward eddy angular momentum The maximum momentum fluxoccurs hours after periapse passage (f = 72) The snapshots correspond to model outputs from 6103 to 6108 Earth days

(A color version of this figure is available in the online journal)

13

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

Figure 13 Differing orbital viewing geometries explored in this study ω = 180 ω = 270 and ω = 360

(A color version of this figure is available in the online journal)

pressures than that of the synchronous case this is most likelydue to a change in momentum budget

4 OBSERVATIONAL IMPLICATIONS

Observations of transiting exoplanets in circular orbits areknown to be shaped by the atmospheric circulation As wasfirst predicted for hot Jupiters by Showman amp Guillot (2002)and later observationally confirmed for hot Jupiter HD 189733bby Knutson et al (2007) an equatorial superrotating jet causesan eastward displacement of the hot spot from the substellarpoint under appropriate conditions Hence a light curve showsa peak in infrared flux minutes to hours before secondaryeclipse We expect such a situation for eccentric exoplanetsas well However observations of eccentric planets can becomplicated because the flux received by the planet variesthroughout its orbit Previous studies explored such situationsbut have not used three-dimensional models (eg Langton ampLaughlin 2008a 2008b Cowan amp Agol 2011) We can expandon these studies and the three-dimensional study by Lewis et al(2010) by using our SPARC model results

We choose to generate light curves following the proceduresdescribed in Fortney et al (2006) for each model integrationat three values of the argument of periastron (ω) the anglebetween the radius vector to the ascending node and the periapseof the orbit In particular we choose orientations where transitoccurs before periapse (at f = minus90ω = 180) transit occursat apoapse (ω = 270) and transit occurs after periapse (atf = +90ω = 360) (Figure 13) These values reflect thebroad range of viewing geometries of transiting exoplanets forexample HAT-P-2b (e = 052) HAT-P-17b (e = 035) andHD 97658b (e = 013) have ω near 180 WASP-8b (e = 031)HAT-P-31b (e = 025) and HD 80606b (e = 093) have ω near270 and GJ 436b (e = 015) XO-3b (e = 026) and HAT-P-11b (e = 020) have ω near 360 (see exoplaneteu for a full listof values) Note that when transit occurs the nightside is visibleto Earth while the dayside is visible at the time of secondaryeclipse

41 Observational Effects of Varying ω

Using the nominal case (e = 050 a = 00534 AU Teq =951 K) we illustrate the dependence of full-orbit light curves onthe value of ω Figure 14 shows three light curves of the same

model integration but at the differing values of ω described inthe previous section Hence the differences in light curves resultsolely from the differing viewing geometries The planetstarflux ratio is plotted as a function of Earth days from periapsein the warm Spitzer wavelength bands (36 and 45 μm) andJHK bands (126 165 and 220 μm) The solid and dashedlines correspond to transit and secondary eclipse respectivelyFor the light curve where ω = 180 (top panel) the peak ininfrared (IR) flux occurs approximately 6 hr after periapse andbefore secondary eclipse The timing of the peak is due toa combination of the radiative lag of the planet and the factthat the hottest regions rotate into an Earth-facing orientationhours after periapse passage The dayside which is most visiblenear secondary eclipse is hottest after periapse Additionallytemperatures reach peak values at times close to when thosehottest regions are facing Earth This helps explain the highamplitude of the light curve peak in the ω = 180 case relative tothe other two cases These effects can be visualized by plottingthe temperature near the photosphere along the equator in apolar projection (Figure 15) Snapshots throughout the orbit plotthe temperature in a polar projection as a function of pressureand orbital position at a latitudinal slice near the equatorthe solid and dashed lines denote the substellar longitude andEarth-facing longitude respectively Based on the temperaturesat the Earth-facing longitude it is apparent that the peak in IRflux should occur after periapse and before secondary eclipse(see f = 12 and 72) this is indeed the case

For the same model but instead with ω = 270 the peak inIR flux occurs 1 hr after eclipse and periapse In this case thedayside is visible near periapse and therefore we are seeing theplanet increase in temperature (and IR flux) during its closestapproach to the star However the hottest regions are displacedeastward of the substellar point and therefore the hottest regionsface Earth significantly before periapse passage (Figure 15 atf asymp 72) Thus by the time the peak temperatures are reachednear and after periapse the hottest regions are already rotatingout of view from Earth This explains the lower amplitude ofthe light curve peak as compared to the peak of the ω = 180case The IR peak decreases as the planet moves further awayfrom the star along our line of sight A second ramp-up in fluxoccurs approximately 1 day after periapseeclipse this is due tothe hottest regions rotating back into view as the planet movestoward apoapse

14

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015P

lane

tSta

r F

lux

Rat

io

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

Figure 14 Three light curves of the IR flux (expressed as a planetstar fluxratio) vs time throughout the full orbit for a single model integration whereTeq = 951 K e = 050 and a = 00534 AU The flux is plotted in warm Spitzerbandpasses 36 and 45 μm and JHK bands (126 165 220 μm respectively)These three cases only differ by their value of ω shown in Figure 13 The solidlines denote transit while the dotted lines denote secondary eclipse

(A color version of this figure is available in the online journal)

In the case where ω = 360 two distinct peaks are presentmdashasmaller peak that occurs after secondary eclipse and sim3ndash4 hrbefore periapse and a second higher peak that occurs aftertransit The first peak occurs as some of the dayside is visible asthe planet is heated approaching periapse The peak decreasesas the dayside rotates out of view For this geometry the hottestregions are facing away from Earth when they reach their peaktemperatures (Figure 15) The second peak occurs after transitas the hottest regions again rotate into view At this point in theorbit peak temperatures have already occurred and thereforethis peak is lower in amplitude than the other two light curvesHowever because the planet is hotter than it was after secondary

eclipse this second peak has a higher amplitude than the firstone

42 High-e Model Integrations at Varying 〈F 〉If an eccentric planet has a large enough dayndashnight tempera-

ture difference and a sufficiently rapid rotation rate its resultantlight curve can show a periodic rise and fall in flux through-out its orbit The light curves shown in Figure 16 are twoe = 075 model integrations that differ in 〈F 〉 the top panelshows a case where 〈F 〉 = 468183 W mminus2 (Teq = 1199 K)and a = 00385 AU while the bottom panel is a case where〈F 〉 = 185691 W mminus2 (Teq = 951 K) and a = 00611 AU Forboth cases light curves are shown for the ω = 180 geome-try (see Figure 13) The differences in 〈F 〉 lead to differencesin light curve amplitude in the two models the closer hotterplanet exhibits a higher flux amplitude Not surprisingly thetiming of peak IR flux for both models resembles the ω = 180case in Figure 14 The peak in each case occurs after periapseand before eclipse due to the combined effects of the ther-mal lag of the planet and geometry of the system Additionallyboth light curves exhibit a quasi-periodic rise and fall of IR fluxdays after periapse passage This phenomenon called ldquoringingrdquowas also seen in models by Langton amp Laughlin (2008a) andCowan amp Agol (2011) caused by the hottest point of the planetrotating in and out of view from Earth For ringing to occurthe temperature difference from dayside to nightside must belarge and it must survive for multiple planetary rotation periodsIn the models shown in Figure 16 this is aided by the rapidpseudo-synchronous rotation rate associated with an eccentric-ity of 075 (Table 2) For the top case the dayndashnight tempera-ture difference is over sim1300 K while the bottom case variesby sim750 K (Figure 7) The light curves from Figure 14 do notexhibit ringing because the dayndashnight temperature difference islow (sim300ndash400 K)

Because the radiative and dynamical timescales in the modelsby Cowan amp Agol (2011) are calculated by a simple scaling withtemperature the ringing seen in their simulated light curvesoccurs with a period equal to the planetrsquos assumed solid-bodyrotation period in the inertial frame of the winds In realityan eccentric exoplanet approaching apoapse would becomeincreasingly cold the winds would weaken and decrease andhence the ringing should be non-periodic Indeed for ourthree dimensional models the ringing in each case has varyingperiodsmdashthe top case has a ringing period ranging from 03 to04 days (Trot = 035 days) while the bottom panel has a periodranging 07ndash08 days (Trot = 07 days)

Despite these differences this study and the studies byLangton amp Laughlin (2008a) and Cowan amp Agol (2011) illus-trate the influence of atmospheric dynamics on the observationsof eccentric exoplanets The light curve features described herecould be seen in future follow-up observations of eccentric ex-oplanets and observers can use the conclusions drawn aboveto relate the observations to the planetrsquos atmospheric structureFor example if ringing is present in an IR light curve the planetmust have a strong dayndashnight temperature asymmetry and hencewe would expect the planet to have strong superrotation at theequator (Showman amp Polvani 2011) This is probably the casefor highly eccentric (e gt 05) short-period exoplanets such asHD 17156b and HD 80606b If ringing were observed it wouldalso allow an estimate of the sum of the rotation speed and thesuperrotation speed thereby placing constraints on both windspeeds and rotation rates of eccentric hot Jupiters The timing ofpeak IR flux relative to transit and secondary eclipse and also

15

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

f=000deg

(K)800

850

900

950

1000

f=minus1242deg

(K)800

850

900

950

1000f=1242deg

(K)800

850

900

950

1000

f=minus7227deg

(K)800

850

900

950

1000f=7227deg

(K)800

850

900

950

1000

f=minus11988deg

(K)800

850

900

950

1000f=11988deg

(K)800

850

900

950

1000

f=minus18000deg

(K)800

850

900

950

1000

Figure 15 Plots of temperature vs longitude and pressure in equatorial slices throughout one full orbit for our nominal model with Teq = 951 K e = 050 anda = 00534 AU Each annulus shows temperature vs longitude and log-pressure for a snapshot at a particular orbital position (denoted by the value of f) as viewedfrom the planetrsquos north pole Pressures between 30 and 43 mbar are displayed in each annulus The different annuli mark the time evolution throughout one completeorbit starting from apoapse for the top annulus proceeding to periapse and then back to apoapse (follow arrows moving counterclockwise around the plot) Thesesnapshots correspond to model outputs from 6103 to 6108 Earth days The solid line in each snapshot indicates the longitude of the substellar point while the dashedline denotes the Earth-facing longitude for the particular case of ω = 180 For the cases of ω = 270 and 360 degrees exactly the same plot is valid except that thesub-Earth longitude would be aiming upward for ω = 270 and right for ω = 360

(A color version of this figure is available in the online journal)

16

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

e = 075 a = 00385 AU

-15 -10 -05 00 05 10 15Time after periapse passage (days)

0000

0001

0002

0003

0004

Pla

netS

tar

Flu

x R

atio

e = 075 a = 00611 AU

-3 -2 -1 0 1 2 3Time after periapse passage (days)

00000

00005

00010

00015

00020

00025

Pla

netS

tar

Flu

x R

atio

Figure 16 Light Curves for two different e = 075 model integrationsTeq = 1199 K a = 00385 AU (top) and Teq = 951 K a = 00611 AU(bottom) both with ω = 180 The solid line denotes transit while the dottedline denotes secondary eclipse

(A color version of this figure is available in the online journal)

relative to orbital position can help to constrain the circulationfurther

5 CONCLUSIONS

We present three-dimensional circulation models coupledwith a two-stream non-gray radiative transfer scheme fora number of theoretical eccentric hot Jupiters We haveshown that as in published models with zero eccentricity ourhigh-eccentricity circulation models are dominated by eastwardflow at photospheric levels which cause an eastward displace-ment of the hottest regions from the substellar point The rapidrotation rates associated with pseudo-synchronization at higheccentricity lead to a small Rossby deformation radius andin some cases multiple jets in the atmosphere Global-meantemperatures and dayndashnight temperature differences peak notat periapse but several hours afterward due to finite radiativetimescales in the planetrsquos atmosphere

Furthermore we show that equatorial superrotation is gener-ated and maintained by eddies formed by the strong dayndashnightheating contrast which induce a flux of momentum from mid-latitudes to the equator The eddy magnitudes and momentumfluxes peak just after periapse passage leading to variations inthe zonal-mean flow throughout the orbit

Figure 17 Globally averaged pressurendashtemperature profiles from 1-D radiative-convective radiative transfer models of HD 189733b with 11 frequency bins(red profiles) 30 frequency bins (green profiles) and 196 frequency bins (blackprofiles) From left to right the profiles vary in stellar flux from the nominalvalue as well as 2times 4times and 8times higher flux The 8times higher flux case has atemperature inversion

(A color version of this figure is available in the online journal)

Lastly we have shown that the spatial and temporal vari-ations of the wind and temperature structure as well as theorbital viewing geometry of the system with respect to Earthcan affect the time and amplitude of peak IR flux seen infull-orbit light curves Depending on the viewing geometry ofthe orbit relative to Earth we find that peaks in IR flux thateither lead or lag periapse are possible in all cases a combi-nation of temporal effects (temperatures changing over time)and geometric effects (hot spots rotating into or out of view)are important in controlling the timing and amplitude of theflux peaks In cases where the dayndashnight temperature contrastis large and the rotational period is short the light curve canalso exhibit ldquoringingrdquo in flux as the hottest region of the planetrotates in and out of view This ringing is non-periodic due tothe variation in stellar heating as a function of distance

This work was supported by NASA Origins and PlanetaryAtmospheres grants to APS TK also acknowledges supportfrom the Harriet P Jenkins Pre-Doctoral Fellowship Program(JPFP) Resources supporting this work were provided bythe NASA High-End Computing (HEC) Program through theNASA Advanced Supercomputing (NAS) Division at AmesResearch Center The authors thank the anonymous referee fortheir helpful comments and suggestions

APPENDIX

VALIDATION TESTS FOR UPDATEDRADIATIVE TRANSFER SCHEME

To validate our updated radiative transfer scheme using 11frequency bins we conducted the following tests

1 First we tested the sensitivity of the number of frequencybins in a one-dimensional radiative-convective model ofHD 189733b (a = 00313 AU e = 00 RP = 82396 times107 m g = 214 m sminus2) Figure 17 compares the globallyaveraged pressurendashtemperature profiles of models run with11 frequency bins 30 frequency bins and 196 frequencybins at (from left to right) 1times 2times 4times and 8times theaverage stellar flux of HD 189733b For each group of

17

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)30 opacity bins

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)

11 opacity bins

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

30 opacity bins

minus600

minus200minus200

200

200

600

6001000

1000

14001800

2200

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

11 opacity bins

minus600

minus200

minus200

200

200

600

600

10001000

1400

1800

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

30 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

11 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

Figure 18 Globally averaged pressurendashtemperature profiles (top row) zonal-mean zonal wind (middle row) and the windtemperature profiles at 30 mbar (bottomrow) for the 30-bin (left column) and 11-bin (right column) model integrations of HD 189733b Snapshots were output at 250 Earth days

(A color version of this figure is available in the online journal)

models the 11-bin profile agrees well with the 30- and196-bin cases from 1 μbar to 1 bar the temperature variesby only tens of K Below 10 bars the 30- and 11-binprofiles differ from the 196-bin cases by up to 50 Kdue to the loss of resolution at low wavenumbers Inthe case of the highest average flux the large amount of

heating leads to a temperature inversion that is not fullycaptured by the 11-bin model but can be improved witha better initial guess at the pressure-temperature (P-T)profile

2 After testing the new radiative transfer (RT) scheme forone-dimensional models we ran full three-dimensional

18

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

simulations of HD 189733b at its nominal average stellarflux using the SPARCMITgcm Figure 18 shows the glob-ally averaged P-T profiles (top row) zonal-mean zonal wind(middle row) and windtemperature profiles at 30 mbar(bottom row) for the 30-bin (left column) and 11-bin (rightcolumn) RT setups The globally averaged profiles arenearly identical with any temperature differences withina few K The zonal wind plots also show good agree-ment with similar peak speeds (gt2600 m sminus1) equatorialjet width (sim60) and jet level (lt100 bar) At 30 mbarthe 30- and 11-bin models share similar dayside-nightsidetemperature structure with a hot spot (1100 K) eastward ofthe substellar point and two colder regions on the nightsidein the mid-latitudes

Both tests show that the bulk circulation and temperaturestructure is retained in the transition from 30 to 11 frequencybins hence we proceed to use the SPARCMITgcm with thenew scheme for these and future model simulations

REFERENCES

Adcroft A Campin J-M Hill C amp Marshall J 2004 MWRv 132 2845Charbonneau D Knutson H A Barman T et al 2008 ApJ 686 1341Cho J Y-K Menou K Hansen B M S amp Seager S 2008 ApJ

675 817Cooper C S amp Showman A P 2005 ApJL 629 L45Cooper C S amp Showman A P 2006 ApJ 649 1048Cowan N B amp Agol E 2011 ApJ 726 82Deming D Harrington J Laughlin G et al 2007 ApJL 667 L199Dobbs-Dixon I Cumming A amp Lin D N C 2010 ApJ 710 1395Dobbs-Dixon I amp Lin D N C 2008 ApJ 673 513Fortney J J Cooper C S Showman A P Marley M S amp Freedman R S

2006 ApJ 652 746Fortney J J Lodders K Marley M S amp Freedman R S 2008 ApJ

678 1419Fortney J J Marley M S Lodders K Saumon D amp Freedman R

2005 ApJL 627 L69

Fu Q amp Liou K N 1992 JAtS 49 2140Goody R West R Chen L amp Crisp D 1989 JQSRT 42 539Greenberg R 2009 ApJL 698 L42Heng K 2012 ApJL 761 L1Heng K Frierson D M W amp Phillipps P J 2011 MNRAS 418 2669Hide R 1969 JAtS 26 841Holton J R 2004 An Introduction to Dynamic Meteorology (Waltham MA

Academic)Hut P 1981 AampA 99 126Irwin J Charbonneau D Nutzman P et al 2008 ApJ 681 636Ivanov P B amp Papaloizou J C B 2004 MNRAS 353 1161Knutson H A Charbonneau D Allen L E Burrows A amp Megeath S T

2008 ApJ 673 526Knutson H A Charbonneau D Allen L E et al 2007 Natur 447 183Knutson H A Charbonneau D Cowan N B et al 2009 ApJ 690 822Langton J amp Laughlin G 2008a ApJ 674 1106Langton J amp Laughlin G 2008b AampA 483 25Lewis N K Showman A P Fortney J J et al 2010 ApJ 720 344Lodders K 2003 ApJ 591 1220Marley M S amp McKay C P 1999 Icar 138 268Mayor M amp Queloz D 1995 Natur 378 355Menou K amp Rauscher E 2009 ApJ 700 887Mlawer E J Taubman S J Brown P D Iacono M J amp Clough S A

1997 JGR 102 16663Naef D Latham D W Mayor M et al 2001 AampA 375 L27Perna R Heng K amp Pont F 2012 ApJ 751 59Pont F Knutson H Gilliland R L Moutou C amp Charbonneau D

2008 MNRAS 385 109Rauscher E amp Menou K 2010 ApJ 714 1334Rauscher E amp Menou K 2012 ApJ 750 96Saumon D Marley M S Abel M Frommhold L amp Freedman R S

2012 ApJ 750 74Showman A P Cho J Y-K amp Menou K 2011 in Exoplanets ed S Seager

(Tucson AZ Univ Arizona Press) 471Showman A P Cooper C S Fortney J J amp Marley M S 2008 ApJ

682 559Showman A P Fortney J J Lian Y et al 2009 ApJ 699 564Showman A P amp Guillot T 2002 AampA 385 166Showman A P amp Polvani L M 2011 ApJ 738 71Swain M R Vasisht G amp Tinetti G 2008 Natur 452 329Thrastarson H T amp Cho J Y 2010 ApJ 716 144Wolszczan A 1994 Sci 264 538

19

  • 1 INTRODUCTION
  • 2 MODEL
    • 21 Updates to SPARCMITgcm
    • 22 Grid Simulations
      • 3 RESULTS
        • 31 Circulation Regime
        • 32 Effect of Eccentricity
        • 33 Effect of Varying Average Stellar Flux
        • 34 Global Average Temperature versus Eccentricity and F
        • 35 Synchronous versus Pseudo-synchronous Rotation
        • 36 Momentum Budget and Equatorial Superrotation
          • 4 OBSERVATIONAL IMPLICATIONS
            • 41 Observational Effects of Varying ω
            • 42 High-e Model Integrations at Varying F
              • 5 CONCLUSIONS
              • APPENDIX VALIDATION TESTS FOR UPDATED RADIATIVE TRANSFER SCHEME
              • REFERENCES
Page 9: THREE-DIMENSIONAL ATMOSPHERIC CIRCULATION OF HOT …showman/publications/kataria-etal-2013.pdf · planets’ atmospheric composition, temperature structure, and circulation. A fifth

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

from 1000 K to 1300 K In addition the variation in temper-ature from dayside to nightside increases this strengthens thewinds within the equatorial superrotating jet Peak winds reachsim2500 m sminus1 in the circular case but increase to sim5000 m sminus1

in the high eccentricity caseAs the eccentricity increases the equatorial jet also narrows

in latitude In the case of highest eccentricity the atmospheredevelops westward midlatitude jets and eastward high-latitudejets The narrowing of the equatorial jet arises from the factthat jet width is confined to the Rossby radius of deformation(Showman amp Polvani 2011) the length scale over which theatmosphere adjusts to planetary-scale phenomena The Rossbyradius of deformation (Holton 2004) is defined at the equator as

LR =radic

c

β (6)

where c is the gravity wave speed and β is the derivative ofthe Coriolis parameter f with respect to latitudinal distancey β = df dy = 2Ωrot cos φRp where Ωrot is the planetaryrotation rate φ is the latitude and Rp is the planetary radius Atthe equator β is simply 2ΩrotRp Hence at the equator LR is

LR =radic

cRp

2Ωrot (7)

Therefore LR prop 1radic

Ωrot and an increased rotation rateleads to a narrower deformation radius From the Hut (1981)formulation the combined effect of the larger orbital period andincrease in eccentricity produces a shorter rotational periodThis in turn increases the rotation rate which reduces thedeformation radius For example the ratio between Ωrot forthe e = 00 case and the e = 075 case is 63 hence we wouldexpect that the equatorial jet should be narrower by a factor ofradic

63 sim 25 This is indeed the case as the e = 075 case hasan average jet width of sim40 latitude while the e = 00 casehas an average jet width of sim100 latitude

33 Effect of Varying Average Stellar Flux

An increase in average stellar flux leads to increases intemperature and wind speeds illustrated in Figure 8 Each caseshown has an eccentricity of 025 these maps have independentcolorscales to better show atmospheric structure The planetthat receives the highest average stellar flux throughout its orbitis hottest overall (bottom) with peak temperatures of 1300 KFurthermore because the planet with the highest 〈F 〉 orbits at thesmallest orbital distance it has a faster rotation rate and hencea narrower equatorial jet this comparison makes the effect ofrotation on the jet structure much more apparent

34 Global Average Temperature versus Eccentricity and 〈F 〉Here we illustrate the variation in time of maximum

planet-wide temperature with eccentricity and average stellarflux Figure 9 shows the globally averaged temperature foreach model integration plotted as a function of mean anomaly9

where M = 0 corresponds to periapse passage With increas-ing eccentricity the time at which the global-mean temperaturereaches its peak value decreases This is expected as a planet

9 The mean anomaly M is defined as (2πPorb)(t minus τ ) where τ is the timeof periapse passage

with a larger eccentricity will have a higher global-average tem-perature at periapse and hence shorter radiative time constant(τrad prop T minus3 see Showman et al 2011) Figure 10 plots the timeof peak temperature from periapse versus eccentricity Here theinverse relationship between eccentricity and the time of peaktemperature is clear

Figures 9 and 10 serve as a nice summary of the modelintegrations and their relation to observations as temperatureis one of the planetary properties that shapes light curvesat secondary eclipse Observers can use these predictions toestimate the timing of peak flux (and global temperature) basedon a given stellar flux andeccentricity (see Section 4)

35 Synchronous versus Pseudo-synchronous Rotation

Here we explore the sensitivity of the circulation to rota-tion rate by comparing models performed with the pseudo-synchronous rotation rate to otherwise identical models per-formed with the synchronous rotation rate This is illustratedin Figure 11 Here the rotation rate varies by almost a factorof three Both cases still have an equatorial superrotating jetand the overall atmospheric structure is similar However thereare subtle differences in the synchronous case the equatorialeastward winds and high-latitude westward winds are higher inmagnitude (right column) The synchronous case also exhibitsmore latitudinal and longitudinal temperature variation partic-ularly on the dayside this is more apparent at apoapse (bot-tom row) In the pseudo-synchronous case the faster rotationrate serves to homogenize the temperature in longitude becausethe time for rotation to carry air parcels from day to night isshorter The temperature difference from equator to pole is alsolarger because the faster rotation rate inhibits equator-to-poleheat transport Thus while it is unlikely that an eccentric planetsynchronously rotates its star the distinction can lead to notabledifferences in the mean flow

These results can also be used to compare the Hut (1981)formulation shown by the pseudo-synchronous case to for-mulations by other groups As shown in Figure 11 if anotherprescription for rotation rate differed from Hut (1981) by afactor of 2ndash3 (as illustrated by the synchronous case) the qual-itative picture would not look that different However if theprescription differed by a factor of more than 3ndash4 the picturecould substantially differ For example the rotation rate calcu-lated from Ivanov amp Papaloizou (2004) Equation (63) yields avalue of 14 times 106 s a factor of four larger than the Hut (1981)calculation So results using this prescription would look verydifferent

Still the rotation rate of eccentric exoplanets is a completelyunconstrained problem One must be careful then in choosinga formulation noting its assumptions and limitations

36 Momentum Budget and Equatorial Superrotation

In most of our simulations the planet maintains an equatorialeastward jet throughout its orbit Superrotation is commonin three-dimensional circulation models of hot Jupiters (egShowman amp Guillot 2002 Showman et al 2008 2009 Menouamp Rauscher 2009 Rauscher amp Menou 2010 2012 Cooperamp Showman 2005 2006 Dobbs-Dixon amp Lin 2008 Dobbs-Dixon et al 2010 Heng et al 2011) and is also seen withinour own solar system on Venus Titan Jupiter and SaturnWhile superrotation on solar system planets has been extensivelystudied only recently have the mechanisms for superrotation onsynchronously rotating exoplanets been identified (Showman amp

9

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)a= 00802 AU

minus200

minus200

200

200

200

600

600

1000

100014001800

22002600

(m sminus1)

0

500

1000

1500

2000

2500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00802 AU

(K)

550

600

650

700

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00505 AU

minus600

minus600

minus200 minus200

200

200

600

600

1000

140018002200260030003400

(m sminus1)

minus500

0

500

1000

1500

2000

2500

3000

3500

4000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00505 AU

(K)

700

750

800

850

900

950

1000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00318 AU

minus200

minus200

200

200

200 200

600

600

600

10001400

1800220026003000340038004200

5000

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00318 AU

(K)700

800

900

1000

1100

1200

1300

Figure 8 Comparison of simulations with e = 025 with decreasing average stellar flux from bottom to top panels Shown are plots of the orbit-averaged zonal-meanzonal wind (left column) and the wind and temperature profiles at 30 mbar at periapse (right column) Each plot has an independent colorscale The vertical bars inthe right column denote the substellar longitude The snapshots of wind and temperature correspond to model outputs of (from top to bottom) 3706 5006 and 2201Earth days

(A color version of this figure is available in the online journal)

10

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

0 05 1 15 2 25 3 35620

640

660

680

700

720

740ltFgt = 11617 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050e=075

0 05 1 15 2 25 3 35880

890

900

910

920

930

940

950

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

ltFgt = 73680 Wmminus1

e=025e=050

0 05 1 15 2 25 3 35 4960

980

1000

1020

1040

1060

1080

1100

1120

1140

1160ltFgt = 185691 Wmminus1

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

Mean anomaly

e=025e=050e=075

0 05 1 15 2 25 3 35 4 451100

1150

1200

1250

1300

1350ltFgt = 468183 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050

Figure 9 Plots of global average temperature vs mean anomaly for each modelintegration Each plot contains profiles of a given average stellar flux (468183185691 73680 and 11617 W mminus1) with blue green and red representingprofiles for e = 025 050 and 075 respectively

(A color version of this figure is available in the online journal)

03 04 05 06 070

05

1

15

2

25

3

eccentricity

Pea

k o

ffse

t fr

om

per

iap

se (

Ear

th d

ays)

468183 Wmminus1

185691 Wmminus1

73680 Wmminus1

11617 Wmminus1

Figure 10 Time after periapse passage at which the global-average temperaturepeaks (from Figure 9) vs eccentricity for all model integrations The bluegreen red and teal profiles correspond to average stellar fluxes of 468183185691 73680 and 11617 W mminus1 respectively

(A color version of this figure is available in the online journal)

Polvani 2011 hereafter SP2011) As discussed in their paperthe generation of superrotation requires up-gradient momentumtransport that pumps angular momentum from outside the jet towithin it Only transport by waves and eddies can satisfy thisrequirement (Hide 1969)

SP2011 demonstrate that for synchronous hot Jupiters onnear-circular orbits such as HD 189733b and HD 209458bequatorial superrotation is generated by standing planetaryscale Kelvin and Rossby waves at the equator and midlati-tudes respectively that are themselves the dynamical responseto the planetrsquos large dayndashnight heating contrast Because theKelvin waves propagate eastward and the Rossby waves prop-agate westward a phase tilt of the wind vectors emerges fromnorthwest to southeast in the northern hemisphere and south-west to northeast in the southern hemisphere with an over-all shape resembling a chevron pointing eastward This patterncauses meridional eddy angular momentum fluxes that convergeangular momentum onto the equator generating equatorialsuperrotation

Despite the fact that not all our model integrations aresynchronous our results suggest that similar mechanisms arestill at play The chevron shape that SP2011 describe is exactlywhat we noted earlier in our plots of the wind and temperatureprofiles from Figure 5 suggesting that as in SP2011 our modelsexhibit an equatorward flux of eddy angular momentum Thisis demonstrated quantitatively in Figure 12 which shows thezonally averaged meridional flux of relative zonal eddy angularmomentum uprimevprime cos φ at snapshots throughout the orbit (asin Figures 5 and 6) Here uprime and vprime represent the deviationof the zonal and meridional winds respectively from theirzonal averages Positive values indicate a northward flux ofeastward eddy angular momentum and negative values indicatea southward flux of eastward eddy angular momentum Atapoapse the small dayndashnight forcing causes poleward transportof eddy angular momentum at latitudes greater than 50 butequatorward transport of eddy angular momentum equator of50 latitude the amplitudes are weak due to the low insolationat apoapse As the planet moves closer to the star (middlerow left column) the increase in dayndashnight forcing strengthens

11

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)e = 050 pseudominussynchronous rotation

minus200

200

200

600

600

1000

10001400

180022002600

3000

3400

3800

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

synchronous rotation

minus600

minus600

minus200

minus200 minus200

minus200

minus200

200

200

600

10001400180022002600300034003800

4200

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

Figure 11 Comparison of pseudo-synchronous (Trot = 3046 times 105 s) vs synchronous (Trot = 8544 times 105 s) rotation for a case with Teq = 755 K e = 050 anda = 00848 AU Wind and temperature snapshots are taken at periapse (middle row) and apoapse (bottom row) at 5008 Earth days and 5013 Earth days respectivelyThe vertical bar denotes substellar longitude

(A color version of this figure is available in the online journal)

the equatorward flux of eastward eddy angular momentumat latitudes between 0 and 50 in both hemispheres Themomentum flux grows and persists until hours after periapseAngular momentum thus converges onto the equator andmaintains the equatorial jet against westward accelerationscaused by advection friction (if any) and Coriolis forces

These mechanisms also explain the difference in wind speedsbetween the synchronous and pseudo-synchronous cases inFigure 11 The synchronous case has a higher dayndashnight heatingcontrast hence a higher equatorward flux of eddy momentumleading to a stronger narrower equatorial jet Moreover theequatorial jet in the pseudo-synchronous case extends to deeper

12

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=18000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

Figure 12 Eddy momentum flux uprimevprime cos φ throughout the orbit of a model where Teq = 951 K e = 05 and a = 00534 AU Positive values indicate a northwardflux of eastward eddy angular momentum while negative values indicate a southward flux of eastward eddy angular momentum The maximum momentum fluxoccurs hours after periapse passage (f = 72) The snapshots correspond to model outputs from 6103 to 6108 Earth days

(A color version of this figure is available in the online journal)

13

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

Figure 13 Differing orbital viewing geometries explored in this study ω = 180 ω = 270 and ω = 360

(A color version of this figure is available in the online journal)

pressures than that of the synchronous case this is most likelydue to a change in momentum budget

4 OBSERVATIONAL IMPLICATIONS

Observations of transiting exoplanets in circular orbits areknown to be shaped by the atmospheric circulation As wasfirst predicted for hot Jupiters by Showman amp Guillot (2002)and later observationally confirmed for hot Jupiter HD 189733bby Knutson et al (2007) an equatorial superrotating jet causesan eastward displacement of the hot spot from the substellarpoint under appropriate conditions Hence a light curve showsa peak in infrared flux minutes to hours before secondaryeclipse We expect such a situation for eccentric exoplanetsas well However observations of eccentric planets can becomplicated because the flux received by the planet variesthroughout its orbit Previous studies explored such situationsbut have not used three-dimensional models (eg Langton ampLaughlin 2008a 2008b Cowan amp Agol 2011) We can expandon these studies and the three-dimensional study by Lewis et al(2010) by using our SPARC model results

We choose to generate light curves following the proceduresdescribed in Fortney et al (2006) for each model integrationat three values of the argument of periastron (ω) the anglebetween the radius vector to the ascending node and the periapseof the orbit In particular we choose orientations where transitoccurs before periapse (at f = minus90ω = 180) transit occursat apoapse (ω = 270) and transit occurs after periapse (atf = +90ω = 360) (Figure 13) These values reflect thebroad range of viewing geometries of transiting exoplanets forexample HAT-P-2b (e = 052) HAT-P-17b (e = 035) andHD 97658b (e = 013) have ω near 180 WASP-8b (e = 031)HAT-P-31b (e = 025) and HD 80606b (e = 093) have ω near270 and GJ 436b (e = 015) XO-3b (e = 026) and HAT-P-11b (e = 020) have ω near 360 (see exoplaneteu for a full listof values) Note that when transit occurs the nightside is visibleto Earth while the dayside is visible at the time of secondaryeclipse

41 Observational Effects of Varying ω

Using the nominal case (e = 050 a = 00534 AU Teq =951 K) we illustrate the dependence of full-orbit light curves onthe value of ω Figure 14 shows three light curves of the same

model integration but at the differing values of ω described inthe previous section Hence the differences in light curves resultsolely from the differing viewing geometries The planetstarflux ratio is plotted as a function of Earth days from periapsein the warm Spitzer wavelength bands (36 and 45 μm) andJHK bands (126 165 and 220 μm) The solid and dashedlines correspond to transit and secondary eclipse respectivelyFor the light curve where ω = 180 (top panel) the peak ininfrared (IR) flux occurs approximately 6 hr after periapse andbefore secondary eclipse The timing of the peak is due toa combination of the radiative lag of the planet and the factthat the hottest regions rotate into an Earth-facing orientationhours after periapse passage The dayside which is most visiblenear secondary eclipse is hottest after periapse Additionallytemperatures reach peak values at times close to when thosehottest regions are facing Earth This helps explain the highamplitude of the light curve peak in the ω = 180 case relative tothe other two cases These effects can be visualized by plottingthe temperature near the photosphere along the equator in apolar projection (Figure 15) Snapshots throughout the orbit plotthe temperature in a polar projection as a function of pressureand orbital position at a latitudinal slice near the equatorthe solid and dashed lines denote the substellar longitude andEarth-facing longitude respectively Based on the temperaturesat the Earth-facing longitude it is apparent that the peak in IRflux should occur after periapse and before secondary eclipse(see f = 12 and 72) this is indeed the case

For the same model but instead with ω = 270 the peak inIR flux occurs 1 hr after eclipse and periapse In this case thedayside is visible near periapse and therefore we are seeing theplanet increase in temperature (and IR flux) during its closestapproach to the star However the hottest regions are displacedeastward of the substellar point and therefore the hottest regionsface Earth significantly before periapse passage (Figure 15 atf asymp 72) Thus by the time the peak temperatures are reachednear and after periapse the hottest regions are already rotatingout of view from Earth This explains the lower amplitude ofthe light curve peak as compared to the peak of the ω = 180case The IR peak decreases as the planet moves further awayfrom the star along our line of sight A second ramp-up in fluxoccurs approximately 1 day after periapseeclipse this is due tothe hottest regions rotating back into view as the planet movestoward apoapse

14

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015P

lane

tSta

r F

lux

Rat

io

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

Figure 14 Three light curves of the IR flux (expressed as a planetstar fluxratio) vs time throughout the full orbit for a single model integration whereTeq = 951 K e = 050 and a = 00534 AU The flux is plotted in warm Spitzerbandpasses 36 and 45 μm and JHK bands (126 165 220 μm respectively)These three cases only differ by their value of ω shown in Figure 13 The solidlines denote transit while the dotted lines denote secondary eclipse

(A color version of this figure is available in the online journal)

In the case where ω = 360 two distinct peaks are presentmdashasmaller peak that occurs after secondary eclipse and sim3ndash4 hrbefore periapse and a second higher peak that occurs aftertransit The first peak occurs as some of the dayside is visible asthe planet is heated approaching periapse The peak decreasesas the dayside rotates out of view For this geometry the hottestregions are facing away from Earth when they reach their peaktemperatures (Figure 15) The second peak occurs after transitas the hottest regions again rotate into view At this point in theorbit peak temperatures have already occurred and thereforethis peak is lower in amplitude than the other two light curvesHowever because the planet is hotter than it was after secondary

eclipse this second peak has a higher amplitude than the firstone

42 High-e Model Integrations at Varying 〈F 〉If an eccentric planet has a large enough dayndashnight tempera-

ture difference and a sufficiently rapid rotation rate its resultantlight curve can show a periodic rise and fall in flux through-out its orbit The light curves shown in Figure 16 are twoe = 075 model integrations that differ in 〈F 〉 the top panelshows a case where 〈F 〉 = 468183 W mminus2 (Teq = 1199 K)and a = 00385 AU while the bottom panel is a case where〈F 〉 = 185691 W mminus2 (Teq = 951 K) and a = 00611 AU Forboth cases light curves are shown for the ω = 180 geome-try (see Figure 13) The differences in 〈F 〉 lead to differencesin light curve amplitude in the two models the closer hotterplanet exhibits a higher flux amplitude Not surprisingly thetiming of peak IR flux for both models resembles the ω = 180case in Figure 14 The peak in each case occurs after periapseand before eclipse due to the combined effects of the ther-mal lag of the planet and geometry of the system Additionallyboth light curves exhibit a quasi-periodic rise and fall of IR fluxdays after periapse passage This phenomenon called ldquoringingrdquowas also seen in models by Langton amp Laughlin (2008a) andCowan amp Agol (2011) caused by the hottest point of the planetrotating in and out of view from Earth For ringing to occurthe temperature difference from dayside to nightside must belarge and it must survive for multiple planetary rotation periodsIn the models shown in Figure 16 this is aided by the rapidpseudo-synchronous rotation rate associated with an eccentric-ity of 075 (Table 2) For the top case the dayndashnight tempera-ture difference is over sim1300 K while the bottom case variesby sim750 K (Figure 7) The light curves from Figure 14 do notexhibit ringing because the dayndashnight temperature difference islow (sim300ndash400 K)

Because the radiative and dynamical timescales in the modelsby Cowan amp Agol (2011) are calculated by a simple scaling withtemperature the ringing seen in their simulated light curvesoccurs with a period equal to the planetrsquos assumed solid-bodyrotation period in the inertial frame of the winds In realityan eccentric exoplanet approaching apoapse would becomeincreasingly cold the winds would weaken and decrease andhence the ringing should be non-periodic Indeed for ourthree dimensional models the ringing in each case has varyingperiodsmdashthe top case has a ringing period ranging from 03 to04 days (Trot = 035 days) while the bottom panel has a periodranging 07ndash08 days (Trot = 07 days)

Despite these differences this study and the studies byLangton amp Laughlin (2008a) and Cowan amp Agol (2011) illus-trate the influence of atmospheric dynamics on the observationsof eccentric exoplanets The light curve features described herecould be seen in future follow-up observations of eccentric ex-oplanets and observers can use the conclusions drawn aboveto relate the observations to the planetrsquos atmospheric structureFor example if ringing is present in an IR light curve the planetmust have a strong dayndashnight temperature asymmetry and hencewe would expect the planet to have strong superrotation at theequator (Showman amp Polvani 2011) This is probably the casefor highly eccentric (e gt 05) short-period exoplanets such asHD 17156b and HD 80606b If ringing were observed it wouldalso allow an estimate of the sum of the rotation speed and thesuperrotation speed thereby placing constraints on both windspeeds and rotation rates of eccentric hot Jupiters The timing ofpeak IR flux relative to transit and secondary eclipse and also

15

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

f=000deg

(K)800

850

900

950

1000

f=minus1242deg

(K)800

850

900

950

1000f=1242deg

(K)800

850

900

950

1000

f=minus7227deg

(K)800

850

900

950

1000f=7227deg

(K)800

850

900

950

1000

f=minus11988deg

(K)800

850

900

950

1000f=11988deg

(K)800

850

900

950

1000

f=minus18000deg

(K)800

850

900

950

1000

Figure 15 Plots of temperature vs longitude and pressure in equatorial slices throughout one full orbit for our nominal model with Teq = 951 K e = 050 anda = 00534 AU Each annulus shows temperature vs longitude and log-pressure for a snapshot at a particular orbital position (denoted by the value of f) as viewedfrom the planetrsquos north pole Pressures between 30 and 43 mbar are displayed in each annulus The different annuli mark the time evolution throughout one completeorbit starting from apoapse for the top annulus proceeding to periapse and then back to apoapse (follow arrows moving counterclockwise around the plot) Thesesnapshots correspond to model outputs from 6103 to 6108 Earth days The solid line in each snapshot indicates the longitude of the substellar point while the dashedline denotes the Earth-facing longitude for the particular case of ω = 180 For the cases of ω = 270 and 360 degrees exactly the same plot is valid except that thesub-Earth longitude would be aiming upward for ω = 270 and right for ω = 360

(A color version of this figure is available in the online journal)

16

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

e = 075 a = 00385 AU

-15 -10 -05 00 05 10 15Time after periapse passage (days)

0000

0001

0002

0003

0004

Pla

netS

tar

Flu

x R

atio

e = 075 a = 00611 AU

-3 -2 -1 0 1 2 3Time after periapse passage (days)

00000

00005

00010

00015

00020

00025

Pla

netS

tar

Flu

x R

atio

Figure 16 Light Curves for two different e = 075 model integrationsTeq = 1199 K a = 00385 AU (top) and Teq = 951 K a = 00611 AU(bottom) both with ω = 180 The solid line denotes transit while the dottedline denotes secondary eclipse

(A color version of this figure is available in the online journal)

relative to orbital position can help to constrain the circulationfurther

5 CONCLUSIONS

We present three-dimensional circulation models coupledwith a two-stream non-gray radiative transfer scheme fora number of theoretical eccentric hot Jupiters We haveshown that as in published models with zero eccentricity ourhigh-eccentricity circulation models are dominated by eastwardflow at photospheric levels which cause an eastward displace-ment of the hottest regions from the substellar point The rapidrotation rates associated with pseudo-synchronization at higheccentricity lead to a small Rossby deformation radius andin some cases multiple jets in the atmosphere Global-meantemperatures and dayndashnight temperature differences peak notat periapse but several hours afterward due to finite radiativetimescales in the planetrsquos atmosphere

Furthermore we show that equatorial superrotation is gener-ated and maintained by eddies formed by the strong dayndashnightheating contrast which induce a flux of momentum from mid-latitudes to the equator The eddy magnitudes and momentumfluxes peak just after periapse passage leading to variations inthe zonal-mean flow throughout the orbit

Figure 17 Globally averaged pressurendashtemperature profiles from 1-D radiative-convective radiative transfer models of HD 189733b with 11 frequency bins(red profiles) 30 frequency bins (green profiles) and 196 frequency bins (blackprofiles) From left to right the profiles vary in stellar flux from the nominalvalue as well as 2times 4times and 8times higher flux The 8times higher flux case has atemperature inversion

(A color version of this figure is available in the online journal)

Lastly we have shown that the spatial and temporal vari-ations of the wind and temperature structure as well as theorbital viewing geometry of the system with respect to Earthcan affect the time and amplitude of peak IR flux seen infull-orbit light curves Depending on the viewing geometry ofthe orbit relative to Earth we find that peaks in IR flux thateither lead or lag periapse are possible in all cases a combi-nation of temporal effects (temperatures changing over time)and geometric effects (hot spots rotating into or out of view)are important in controlling the timing and amplitude of theflux peaks In cases where the dayndashnight temperature contrastis large and the rotational period is short the light curve canalso exhibit ldquoringingrdquo in flux as the hottest region of the planetrotates in and out of view This ringing is non-periodic due tothe variation in stellar heating as a function of distance

This work was supported by NASA Origins and PlanetaryAtmospheres grants to APS TK also acknowledges supportfrom the Harriet P Jenkins Pre-Doctoral Fellowship Program(JPFP) Resources supporting this work were provided bythe NASA High-End Computing (HEC) Program through theNASA Advanced Supercomputing (NAS) Division at AmesResearch Center The authors thank the anonymous referee fortheir helpful comments and suggestions

APPENDIX

VALIDATION TESTS FOR UPDATEDRADIATIVE TRANSFER SCHEME

To validate our updated radiative transfer scheme using 11frequency bins we conducted the following tests

1 First we tested the sensitivity of the number of frequencybins in a one-dimensional radiative-convective model ofHD 189733b (a = 00313 AU e = 00 RP = 82396 times107 m g = 214 m sminus2) Figure 17 compares the globallyaveraged pressurendashtemperature profiles of models run with11 frequency bins 30 frequency bins and 196 frequencybins at (from left to right) 1times 2times 4times and 8times theaverage stellar flux of HD 189733b For each group of

17

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)30 opacity bins

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)

11 opacity bins

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

30 opacity bins

minus600

minus200minus200

200

200

600

6001000

1000

14001800

2200

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

11 opacity bins

minus600

minus200

minus200

200

200

600

600

10001000

1400

1800

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

30 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

11 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

Figure 18 Globally averaged pressurendashtemperature profiles (top row) zonal-mean zonal wind (middle row) and the windtemperature profiles at 30 mbar (bottomrow) for the 30-bin (left column) and 11-bin (right column) model integrations of HD 189733b Snapshots were output at 250 Earth days

(A color version of this figure is available in the online journal)

models the 11-bin profile agrees well with the 30- and196-bin cases from 1 μbar to 1 bar the temperature variesby only tens of K Below 10 bars the 30- and 11-binprofiles differ from the 196-bin cases by up to 50 Kdue to the loss of resolution at low wavenumbers Inthe case of the highest average flux the large amount of

heating leads to a temperature inversion that is not fullycaptured by the 11-bin model but can be improved witha better initial guess at the pressure-temperature (P-T)profile

2 After testing the new radiative transfer (RT) scheme forone-dimensional models we ran full three-dimensional

18

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

simulations of HD 189733b at its nominal average stellarflux using the SPARCMITgcm Figure 18 shows the glob-ally averaged P-T profiles (top row) zonal-mean zonal wind(middle row) and windtemperature profiles at 30 mbar(bottom row) for the 30-bin (left column) and 11-bin (rightcolumn) RT setups The globally averaged profiles arenearly identical with any temperature differences withina few K The zonal wind plots also show good agree-ment with similar peak speeds (gt2600 m sminus1) equatorialjet width (sim60) and jet level (lt100 bar) At 30 mbarthe 30- and 11-bin models share similar dayside-nightsidetemperature structure with a hot spot (1100 K) eastward ofthe substellar point and two colder regions on the nightsidein the mid-latitudes

Both tests show that the bulk circulation and temperaturestructure is retained in the transition from 30 to 11 frequencybins hence we proceed to use the SPARCMITgcm with thenew scheme for these and future model simulations

REFERENCES

Adcroft A Campin J-M Hill C amp Marshall J 2004 MWRv 132 2845Charbonneau D Knutson H A Barman T et al 2008 ApJ 686 1341Cho J Y-K Menou K Hansen B M S amp Seager S 2008 ApJ

675 817Cooper C S amp Showman A P 2005 ApJL 629 L45Cooper C S amp Showman A P 2006 ApJ 649 1048Cowan N B amp Agol E 2011 ApJ 726 82Deming D Harrington J Laughlin G et al 2007 ApJL 667 L199Dobbs-Dixon I Cumming A amp Lin D N C 2010 ApJ 710 1395Dobbs-Dixon I amp Lin D N C 2008 ApJ 673 513Fortney J J Cooper C S Showman A P Marley M S amp Freedman R S

2006 ApJ 652 746Fortney J J Lodders K Marley M S amp Freedman R S 2008 ApJ

678 1419Fortney J J Marley M S Lodders K Saumon D amp Freedman R

2005 ApJL 627 L69

Fu Q amp Liou K N 1992 JAtS 49 2140Goody R West R Chen L amp Crisp D 1989 JQSRT 42 539Greenberg R 2009 ApJL 698 L42Heng K 2012 ApJL 761 L1Heng K Frierson D M W amp Phillipps P J 2011 MNRAS 418 2669Hide R 1969 JAtS 26 841Holton J R 2004 An Introduction to Dynamic Meteorology (Waltham MA

Academic)Hut P 1981 AampA 99 126Irwin J Charbonneau D Nutzman P et al 2008 ApJ 681 636Ivanov P B amp Papaloizou J C B 2004 MNRAS 353 1161Knutson H A Charbonneau D Allen L E Burrows A amp Megeath S T

2008 ApJ 673 526Knutson H A Charbonneau D Allen L E et al 2007 Natur 447 183Knutson H A Charbonneau D Cowan N B et al 2009 ApJ 690 822Langton J amp Laughlin G 2008a ApJ 674 1106Langton J amp Laughlin G 2008b AampA 483 25Lewis N K Showman A P Fortney J J et al 2010 ApJ 720 344Lodders K 2003 ApJ 591 1220Marley M S amp McKay C P 1999 Icar 138 268Mayor M amp Queloz D 1995 Natur 378 355Menou K amp Rauscher E 2009 ApJ 700 887Mlawer E J Taubman S J Brown P D Iacono M J amp Clough S A

1997 JGR 102 16663Naef D Latham D W Mayor M et al 2001 AampA 375 L27Perna R Heng K amp Pont F 2012 ApJ 751 59Pont F Knutson H Gilliland R L Moutou C amp Charbonneau D

2008 MNRAS 385 109Rauscher E amp Menou K 2010 ApJ 714 1334Rauscher E amp Menou K 2012 ApJ 750 96Saumon D Marley M S Abel M Frommhold L amp Freedman R S

2012 ApJ 750 74Showman A P Cho J Y-K amp Menou K 2011 in Exoplanets ed S Seager

(Tucson AZ Univ Arizona Press) 471Showman A P Cooper C S Fortney J J amp Marley M S 2008 ApJ

682 559Showman A P Fortney J J Lian Y et al 2009 ApJ 699 564Showman A P amp Guillot T 2002 AampA 385 166Showman A P amp Polvani L M 2011 ApJ 738 71Swain M R Vasisht G amp Tinetti G 2008 Natur 452 329Thrastarson H T amp Cho J Y 2010 ApJ 716 144Wolszczan A 1994 Sci 264 538

19

  • 1 INTRODUCTION
  • 2 MODEL
    • 21 Updates to SPARCMITgcm
    • 22 Grid Simulations
      • 3 RESULTS
        • 31 Circulation Regime
        • 32 Effect of Eccentricity
        • 33 Effect of Varying Average Stellar Flux
        • 34 Global Average Temperature versus Eccentricity and F
        • 35 Synchronous versus Pseudo-synchronous Rotation
        • 36 Momentum Budget and Equatorial Superrotation
          • 4 OBSERVATIONAL IMPLICATIONS
            • 41 Observational Effects of Varying ω
            • 42 High-e Model Integrations at Varying F
              • 5 CONCLUSIONS
              • APPENDIX VALIDATION TESTS FOR UPDATED RADIATIVE TRANSFER SCHEME
              • REFERENCES
Page 10: THREE-DIMENSIONAL ATMOSPHERIC CIRCULATION OF HOT …showman/publications/kataria-etal-2013.pdf · planets’ atmospheric composition, temperature structure, and circulation. A fifth

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)a= 00802 AU

minus200

minus200

200

200

200

600

600

1000

100014001800

22002600

(m sminus1)

0

500

1000

1500

2000

2500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00802 AU

(K)

550

600

650

700

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00505 AU

minus600

minus600

minus200 minus200

200

200

600

600

1000

140018002200260030003400

(m sminus1)

minus500

0

500

1000

1500

2000

2500

3000

3500

4000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00505 AU

(K)

700

750

800

850

900

950

1000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

a= 00318 AU

minus200

minus200

200

200

200 200

600

600

600

10001400

1800220026003000340038004200

5000

(m sminus1)

0

1000

2000

3000

4000

5000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e ( deg

)

a= 00318 AU

(K)700

800

900

1000

1100

1200

1300

Figure 8 Comparison of simulations with e = 025 with decreasing average stellar flux from bottom to top panels Shown are plots of the orbit-averaged zonal-meanzonal wind (left column) and the wind and temperature profiles at 30 mbar at periapse (right column) Each plot has an independent colorscale The vertical bars inthe right column denote the substellar longitude The snapshots of wind and temperature correspond to model outputs of (from top to bottom) 3706 5006 and 2201Earth days

(A color version of this figure is available in the online journal)

10

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

0 05 1 15 2 25 3 35620

640

660

680

700

720

740ltFgt = 11617 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050e=075

0 05 1 15 2 25 3 35880

890

900

910

920

930

940

950

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

ltFgt = 73680 Wmminus1

e=025e=050

0 05 1 15 2 25 3 35 4960

980

1000

1020

1040

1060

1080

1100

1120

1140

1160ltFgt = 185691 Wmminus1

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

Mean anomaly

e=025e=050e=075

0 05 1 15 2 25 3 35 4 451100

1150

1200

1250

1300

1350ltFgt = 468183 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050

Figure 9 Plots of global average temperature vs mean anomaly for each modelintegration Each plot contains profiles of a given average stellar flux (468183185691 73680 and 11617 W mminus1) with blue green and red representingprofiles for e = 025 050 and 075 respectively

(A color version of this figure is available in the online journal)

03 04 05 06 070

05

1

15

2

25

3

eccentricity

Pea

k o

ffse

t fr

om

per

iap

se (

Ear

th d

ays)

468183 Wmminus1

185691 Wmminus1

73680 Wmminus1

11617 Wmminus1

Figure 10 Time after periapse passage at which the global-average temperaturepeaks (from Figure 9) vs eccentricity for all model integrations The bluegreen red and teal profiles correspond to average stellar fluxes of 468183185691 73680 and 11617 W mminus1 respectively

(A color version of this figure is available in the online journal)

Polvani 2011 hereafter SP2011) As discussed in their paperthe generation of superrotation requires up-gradient momentumtransport that pumps angular momentum from outside the jet towithin it Only transport by waves and eddies can satisfy thisrequirement (Hide 1969)

SP2011 demonstrate that for synchronous hot Jupiters onnear-circular orbits such as HD 189733b and HD 209458bequatorial superrotation is generated by standing planetaryscale Kelvin and Rossby waves at the equator and midlati-tudes respectively that are themselves the dynamical responseto the planetrsquos large dayndashnight heating contrast Because theKelvin waves propagate eastward and the Rossby waves prop-agate westward a phase tilt of the wind vectors emerges fromnorthwest to southeast in the northern hemisphere and south-west to northeast in the southern hemisphere with an over-all shape resembling a chevron pointing eastward This patterncauses meridional eddy angular momentum fluxes that convergeangular momentum onto the equator generating equatorialsuperrotation

Despite the fact that not all our model integrations aresynchronous our results suggest that similar mechanisms arestill at play The chevron shape that SP2011 describe is exactlywhat we noted earlier in our plots of the wind and temperatureprofiles from Figure 5 suggesting that as in SP2011 our modelsexhibit an equatorward flux of eddy angular momentum Thisis demonstrated quantitatively in Figure 12 which shows thezonally averaged meridional flux of relative zonal eddy angularmomentum uprimevprime cos φ at snapshots throughout the orbit (asin Figures 5 and 6) Here uprime and vprime represent the deviationof the zonal and meridional winds respectively from theirzonal averages Positive values indicate a northward flux ofeastward eddy angular momentum and negative values indicatea southward flux of eastward eddy angular momentum Atapoapse the small dayndashnight forcing causes poleward transportof eddy angular momentum at latitudes greater than 50 butequatorward transport of eddy angular momentum equator of50 latitude the amplitudes are weak due to the low insolationat apoapse As the planet moves closer to the star (middlerow left column) the increase in dayndashnight forcing strengthens

11

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)e = 050 pseudominussynchronous rotation

minus200

200

200

600

600

1000

10001400

180022002600

3000

3400

3800

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

synchronous rotation

minus600

minus600

minus200

minus200 minus200

minus200

minus200

200

200

600

10001400180022002600300034003800

4200

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

Figure 11 Comparison of pseudo-synchronous (Trot = 3046 times 105 s) vs synchronous (Trot = 8544 times 105 s) rotation for a case with Teq = 755 K e = 050 anda = 00848 AU Wind and temperature snapshots are taken at periapse (middle row) and apoapse (bottom row) at 5008 Earth days and 5013 Earth days respectivelyThe vertical bar denotes substellar longitude

(A color version of this figure is available in the online journal)

the equatorward flux of eastward eddy angular momentumat latitudes between 0 and 50 in both hemispheres Themomentum flux grows and persists until hours after periapseAngular momentum thus converges onto the equator andmaintains the equatorial jet against westward accelerationscaused by advection friction (if any) and Coriolis forces

These mechanisms also explain the difference in wind speedsbetween the synchronous and pseudo-synchronous cases inFigure 11 The synchronous case has a higher dayndashnight heatingcontrast hence a higher equatorward flux of eddy momentumleading to a stronger narrower equatorial jet Moreover theequatorial jet in the pseudo-synchronous case extends to deeper

12

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=18000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

Figure 12 Eddy momentum flux uprimevprime cos φ throughout the orbit of a model where Teq = 951 K e = 05 and a = 00534 AU Positive values indicate a northwardflux of eastward eddy angular momentum while negative values indicate a southward flux of eastward eddy angular momentum The maximum momentum fluxoccurs hours after periapse passage (f = 72) The snapshots correspond to model outputs from 6103 to 6108 Earth days

(A color version of this figure is available in the online journal)

13

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

Figure 13 Differing orbital viewing geometries explored in this study ω = 180 ω = 270 and ω = 360

(A color version of this figure is available in the online journal)

pressures than that of the synchronous case this is most likelydue to a change in momentum budget

4 OBSERVATIONAL IMPLICATIONS

Observations of transiting exoplanets in circular orbits areknown to be shaped by the atmospheric circulation As wasfirst predicted for hot Jupiters by Showman amp Guillot (2002)and later observationally confirmed for hot Jupiter HD 189733bby Knutson et al (2007) an equatorial superrotating jet causesan eastward displacement of the hot spot from the substellarpoint under appropriate conditions Hence a light curve showsa peak in infrared flux minutes to hours before secondaryeclipse We expect such a situation for eccentric exoplanetsas well However observations of eccentric planets can becomplicated because the flux received by the planet variesthroughout its orbit Previous studies explored such situationsbut have not used three-dimensional models (eg Langton ampLaughlin 2008a 2008b Cowan amp Agol 2011) We can expandon these studies and the three-dimensional study by Lewis et al(2010) by using our SPARC model results

We choose to generate light curves following the proceduresdescribed in Fortney et al (2006) for each model integrationat three values of the argument of periastron (ω) the anglebetween the radius vector to the ascending node and the periapseof the orbit In particular we choose orientations where transitoccurs before periapse (at f = minus90ω = 180) transit occursat apoapse (ω = 270) and transit occurs after periapse (atf = +90ω = 360) (Figure 13) These values reflect thebroad range of viewing geometries of transiting exoplanets forexample HAT-P-2b (e = 052) HAT-P-17b (e = 035) andHD 97658b (e = 013) have ω near 180 WASP-8b (e = 031)HAT-P-31b (e = 025) and HD 80606b (e = 093) have ω near270 and GJ 436b (e = 015) XO-3b (e = 026) and HAT-P-11b (e = 020) have ω near 360 (see exoplaneteu for a full listof values) Note that when transit occurs the nightside is visibleto Earth while the dayside is visible at the time of secondaryeclipse

41 Observational Effects of Varying ω

Using the nominal case (e = 050 a = 00534 AU Teq =951 K) we illustrate the dependence of full-orbit light curves onthe value of ω Figure 14 shows three light curves of the same

model integration but at the differing values of ω described inthe previous section Hence the differences in light curves resultsolely from the differing viewing geometries The planetstarflux ratio is plotted as a function of Earth days from periapsein the warm Spitzer wavelength bands (36 and 45 μm) andJHK bands (126 165 and 220 μm) The solid and dashedlines correspond to transit and secondary eclipse respectivelyFor the light curve where ω = 180 (top panel) the peak ininfrared (IR) flux occurs approximately 6 hr after periapse andbefore secondary eclipse The timing of the peak is due toa combination of the radiative lag of the planet and the factthat the hottest regions rotate into an Earth-facing orientationhours after periapse passage The dayside which is most visiblenear secondary eclipse is hottest after periapse Additionallytemperatures reach peak values at times close to when thosehottest regions are facing Earth This helps explain the highamplitude of the light curve peak in the ω = 180 case relative tothe other two cases These effects can be visualized by plottingthe temperature near the photosphere along the equator in apolar projection (Figure 15) Snapshots throughout the orbit plotthe temperature in a polar projection as a function of pressureand orbital position at a latitudinal slice near the equatorthe solid and dashed lines denote the substellar longitude andEarth-facing longitude respectively Based on the temperaturesat the Earth-facing longitude it is apparent that the peak in IRflux should occur after periapse and before secondary eclipse(see f = 12 and 72) this is indeed the case

For the same model but instead with ω = 270 the peak inIR flux occurs 1 hr after eclipse and periapse In this case thedayside is visible near periapse and therefore we are seeing theplanet increase in temperature (and IR flux) during its closestapproach to the star However the hottest regions are displacedeastward of the substellar point and therefore the hottest regionsface Earth significantly before periapse passage (Figure 15 atf asymp 72) Thus by the time the peak temperatures are reachednear and after periapse the hottest regions are already rotatingout of view from Earth This explains the lower amplitude ofthe light curve peak as compared to the peak of the ω = 180case The IR peak decreases as the planet moves further awayfrom the star along our line of sight A second ramp-up in fluxoccurs approximately 1 day after periapseeclipse this is due tothe hottest regions rotating back into view as the planet movestoward apoapse

14

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015P

lane

tSta

r F

lux

Rat

io

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

Figure 14 Three light curves of the IR flux (expressed as a planetstar fluxratio) vs time throughout the full orbit for a single model integration whereTeq = 951 K e = 050 and a = 00534 AU The flux is plotted in warm Spitzerbandpasses 36 and 45 μm and JHK bands (126 165 220 μm respectively)These three cases only differ by their value of ω shown in Figure 13 The solidlines denote transit while the dotted lines denote secondary eclipse

(A color version of this figure is available in the online journal)

In the case where ω = 360 two distinct peaks are presentmdashasmaller peak that occurs after secondary eclipse and sim3ndash4 hrbefore periapse and a second higher peak that occurs aftertransit The first peak occurs as some of the dayside is visible asthe planet is heated approaching periapse The peak decreasesas the dayside rotates out of view For this geometry the hottestregions are facing away from Earth when they reach their peaktemperatures (Figure 15) The second peak occurs after transitas the hottest regions again rotate into view At this point in theorbit peak temperatures have already occurred and thereforethis peak is lower in amplitude than the other two light curvesHowever because the planet is hotter than it was after secondary

eclipse this second peak has a higher amplitude than the firstone

42 High-e Model Integrations at Varying 〈F 〉If an eccentric planet has a large enough dayndashnight tempera-

ture difference and a sufficiently rapid rotation rate its resultantlight curve can show a periodic rise and fall in flux through-out its orbit The light curves shown in Figure 16 are twoe = 075 model integrations that differ in 〈F 〉 the top panelshows a case where 〈F 〉 = 468183 W mminus2 (Teq = 1199 K)and a = 00385 AU while the bottom panel is a case where〈F 〉 = 185691 W mminus2 (Teq = 951 K) and a = 00611 AU Forboth cases light curves are shown for the ω = 180 geome-try (see Figure 13) The differences in 〈F 〉 lead to differencesin light curve amplitude in the two models the closer hotterplanet exhibits a higher flux amplitude Not surprisingly thetiming of peak IR flux for both models resembles the ω = 180case in Figure 14 The peak in each case occurs after periapseand before eclipse due to the combined effects of the ther-mal lag of the planet and geometry of the system Additionallyboth light curves exhibit a quasi-periodic rise and fall of IR fluxdays after periapse passage This phenomenon called ldquoringingrdquowas also seen in models by Langton amp Laughlin (2008a) andCowan amp Agol (2011) caused by the hottest point of the planetrotating in and out of view from Earth For ringing to occurthe temperature difference from dayside to nightside must belarge and it must survive for multiple planetary rotation periodsIn the models shown in Figure 16 this is aided by the rapidpseudo-synchronous rotation rate associated with an eccentric-ity of 075 (Table 2) For the top case the dayndashnight tempera-ture difference is over sim1300 K while the bottom case variesby sim750 K (Figure 7) The light curves from Figure 14 do notexhibit ringing because the dayndashnight temperature difference islow (sim300ndash400 K)

Because the radiative and dynamical timescales in the modelsby Cowan amp Agol (2011) are calculated by a simple scaling withtemperature the ringing seen in their simulated light curvesoccurs with a period equal to the planetrsquos assumed solid-bodyrotation period in the inertial frame of the winds In realityan eccentric exoplanet approaching apoapse would becomeincreasingly cold the winds would weaken and decrease andhence the ringing should be non-periodic Indeed for ourthree dimensional models the ringing in each case has varyingperiodsmdashthe top case has a ringing period ranging from 03 to04 days (Trot = 035 days) while the bottom panel has a periodranging 07ndash08 days (Trot = 07 days)

Despite these differences this study and the studies byLangton amp Laughlin (2008a) and Cowan amp Agol (2011) illus-trate the influence of atmospheric dynamics on the observationsof eccentric exoplanets The light curve features described herecould be seen in future follow-up observations of eccentric ex-oplanets and observers can use the conclusions drawn aboveto relate the observations to the planetrsquos atmospheric structureFor example if ringing is present in an IR light curve the planetmust have a strong dayndashnight temperature asymmetry and hencewe would expect the planet to have strong superrotation at theequator (Showman amp Polvani 2011) This is probably the casefor highly eccentric (e gt 05) short-period exoplanets such asHD 17156b and HD 80606b If ringing were observed it wouldalso allow an estimate of the sum of the rotation speed and thesuperrotation speed thereby placing constraints on both windspeeds and rotation rates of eccentric hot Jupiters The timing ofpeak IR flux relative to transit and secondary eclipse and also

15

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

f=000deg

(K)800

850

900

950

1000

f=minus1242deg

(K)800

850

900

950

1000f=1242deg

(K)800

850

900

950

1000

f=minus7227deg

(K)800

850

900

950

1000f=7227deg

(K)800

850

900

950

1000

f=minus11988deg

(K)800

850

900

950

1000f=11988deg

(K)800

850

900

950

1000

f=minus18000deg

(K)800

850

900

950

1000

Figure 15 Plots of temperature vs longitude and pressure in equatorial slices throughout one full orbit for our nominal model with Teq = 951 K e = 050 anda = 00534 AU Each annulus shows temperature vs longitude and log-pressure for a snapshot at a particular orbital position (denoted by the value of f) as viewedfrom the planetrsquos north pole Pressures between 30 and 43 mbar are displayed in each annulus The different annuli mark the time evolution throughout one completeorbit starting from apoapse for the top annulus proceeding to periapse and then back to apoapse (follow arrows moving counterclockwise around the plot) Thesesnapshots correspond to model outputs from 6103 to 6108 Earth days The solid line in each snapshot indicates the longitude of the substellar point while the dashedline denotes the Earth-facing longitude for the particular case of ω = 180 For the cases of ω = 270 and 360 degrees exactly the same plot is valid except that thesub-Earth longitude would be aiming upward for ω = 270 and right for ω = 360

(A color version of this figure is available in the online journal)

16

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

e = 075 a = 00385 AU

-15 -10 -05 00 05 10 15Time after periapse passage (days)

0000

0001

0002

0003

0004

Pla

netS

tar

Flu

x R

atio

e = 075 a = 00611 AU

-3 -2 -1 0 1 2 3Time after periapse passage (days)

00000

00005

00010

00015

00020

00025

Pla

netS

tar

Flu

x R

atio

Figure 16 Light Curves for two different e = 075 model integrationsTeq = 1199 K a = 00385 AU (top) and Teq = 951 K a = 00611 AU(bottom) both with ω = 180 The solid line denotes transit while the dottedline denotes secondary eclipse

(A color version of this figure is available in the online journal)

relative to orbital position can help to constrain the circulationfurther

5 CONCLUSIONS

We present three-dimensional circulation models coupledwith a two-stream non-gray radiative transfer scheme fora number of theoretical eccentric hot Jupiters We haveshown that as in published models with zero eccentricity ourhigh-eccentricity circulation models are dominated by eastwardflow at photospheric levels which cause an eastward displace-ment of the hottest regions from the substellar point The rapidrotation rates associated with pseudo-synchronization at higheccentricity lead to a small Rossby deformation radius andin some cases multiple jets in the atmosphere Global-meantemperatures and dayndashnight temperature differences peak notat periapse but several hours afterward due to finite radiativetimescales in the planetrsquos atmosphere

Furthermore we show that equatorial superrotation is gener-ated and maintained by eddies formed by the strong dayndashnightheating contrast which induce a flux of momentum from mid-latitudes to the equator The eddy magnitudes and momentumfluxes peak just after periapse passage leading to variations inthe zonal-mean flow throughout the orbit

Figure 17 Globally averaged pressurendashtemperature profiles from 1-D radiative-convective radiative transfer models of HD 189733b with 11 frequency bins(red profiles) 30 frequency bins (green profiles) and 196 frequency bins (blackprofiles) From left to right the profiles vary in stellar flux from the nominalvalue as well as 2times 4times and 8times higher flux The 8times higher flux case has atemperature inversion

(A color version of this figure is available in the online journal)

Lastly we have shown that the spatial and temporal vari-ations of the wind and temperature structure as well as theorbital viewing geometry of the system with respect to Earthcan affect the time and amplitude of peak IR flux seen infull-orbit light curves Depending on the viewing geometry ofthe orbit relative to Earth we find that peaks in IR flux thateither lead or lag periapse are possible in all cases a combi-nation of temporal effects (temperatures changing over time)and geometric effects (hot spots rotating into or out of view)are important in controlling the timing and amplitude of theflux peaks In cases where the dayndashnight temperature contrastis large and the rotational period is short the light curve canalso exhibit ldquoringingrdquo in flux as the hottest region of the planetrotates in and out of view This ringing is non-periodic due tothe variation in stellar heating as a function of distance

This work was supported by NASA Origins and PlanetaryAtmospheres grants to APS TK also acknowledges supportfrom the Harriet P Jenkins Pre-Doctoral Fellowship Program(JPFP) Resources supporting this work were provided bythe NASA High-End Computing (HEC) Program through theNASA Advanced Supercomputing (NAS) Division at AmesResearch Center The authors thank the anonymous referee fortheir helpful comments and suggestions

APPENDIX

VALIDATION TESTS FOR UPDATEDRADIATIVE TRANSFER SCHEME

To validate our updated radiative transfer scheme using 11frequency bins we conducted the following tests

1 First we tested the sensitivity of the number of frequencybins in a one-dimensional radiative-convective model ofHD 189733b (a = 00313 AU e = 00 RP = 82396 times107 m g = 214 m sminus2) Figure 17 compares the globallyaveraged pressurendashtemperature profiles of models run with11 frequency bins 30 frequency bins and 196 frequencybins at (from left to right) 1times 2times 4times and 8times theaverage stellar flux of HD 189733b For each group of

17

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)30 opacity bins

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)

11 opacity bins

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

30 opacity bins

minus600

minus200minus200

200

200

600

6001000

1000

14001800

2200

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

11 opacity bins

minus600

minus200

minus200

200

200

600

600

10001000

1400

1800

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

30 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

11 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

Figure 18 Globally averaged pressurendashtemperature profiles (top row) zonal-mean zonal wind (middle row) and the windtemperature profiles at 30 mbar (bottomrow) for the 30-bin (left column) and 11-bin (right column) model integrations of HD 189733b Snapshots were output at 250 Earth days

(A color version of this figure is available in the online journal)

models the 11-bin profile agrees well with the 30- and196-bin cases from 1 μbar to 1 bar the temperature variesby only tens of K Below 10 bars the 30- and 11-binprofiles differ from the 196-bin cases by up to 50 Kdue to the loss of resolution at low wavenumbers Inthe case of the highest average flux the large amount of

heating leads to a temperature inversion that is not fullycaptured by the 11-bin model but can be improved witha better initial guess at the pressure-temperature (P-T)profile

2 After testing the new radiative transfer (RT) scheme forone-dimensional models we ran full three-dimensional

18

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

simulations of HD 189733b at its nominal average stellarflux using the SPARCMITgcm Figure 18 shows the glob-ally averaged P-T profiles (top row) zonal-mean zonal wind(middle row) and windtemperature profiles at 30 mbar(bottom row) for the 30-bin (left column) and 11-bin (rightcolumn) RT setups The globally averaged profiles arenearly identical with any temperature differences withina few K The zonal wind plots also show good agree-ment with similar peak speeds (gt2600 m sminus1) equatorialjet width (sim60) and jet level (lt100 bar) At 30 mbarthe 30- and 11-bin models share similar dayside-nightsidetemperature structure with a hot spot (1100 K) eastward ofthe substellar point and two colder regions on the nightsidein the mid-latitudes

Both tests show that the bulk circulation and temperaturestructure is retained in the transition from 30 to 11 frequencybins hence we proceed to use the SPARCMITgcm with thenew scheme for these and future model simulations

REFERENCES

Adcroft A Campin J-M Hill C amp Marshall J 2004 MWRv 132 2845Charbonneau D Knutson H A Barman T et al 2008 ApJ 686 1341Cho J Y-K Menou K Hansen B M S amp Seager S 2008 ApJ

675 817Cooper C S amp Showman A P 2005 ApJL 629 L45Cooper C S amp Showman A P 2006 ApJ 649 1048Cowan N B amp Agol E 2011 ApJ 726 82Deming D Harrington J Laughlin G et al 2007 ApJL 667 L199Dobbs-Dixon I Cumming A amp Lin D N C 2010 ApJ 710 1395Dobbs-Dixon I amp Lin D N C 2008 ApJ 673 513Fortney J J Cooper C S Showman A P Marley M S amp Freedman R S

2006 ApJ 652 746Fortney J J Lodders K Marley M S amp Freedman R S 2008 ApJ

678 1419Fortney J J Marley M S Lodders K Saumon D amp Freedman R

2005 ApJL 627 L69

Fu Q amp Liou K N 1992 JAtS 49 2140Goody R West R Chen L amp Crisp D 1989 JQSRT 42 539Greenberg R 2009 ApJL 698 L42Heng K 2012 ApJL 761 L1Heng K Frierson D M W amp Phillipps P J 2011 MNRAS 418 2669Hide R 1969 JAtS 26 841Holton J R 2004 An Introduction to Dynamic Meteorology (Waltham MA

Academic)Hut P 1981 AampA 99 126Irwin J Charbonneau D Nutzman P et al 2008 ApJ 681 636Ivanov P B amp Papaloizou J C B 2004 MNRAS 353 1161Knutson H A Charbonneau D Allen L E Burrows A amp Megeath S T

2008 ApJ 673 526Knutson H A Charbonneau D Allen L E et al 2007 Natur 447 183Knutson H A Charbonneau D Cowan N B et al 2009 ApJ 690 822Langton J amp Laughlin G 2008a ApJ 674 1106Langton J amp Laughlin G 2008b AampA 483 25Lewis N K Showman A P Fortney J J et al 2010 ApJ 720 344Lodders K 2003 ApJ 591 1220Marley M S amp McKay C P 1999 Icar 138 268Mayor M amp Queloz D 1995 Natur 378 355Menou K amp Rauscher E 2009 ApJ 700 887Mlawer E J Taubman S J Brown P D Iacono M J amp Clough S A

1997 JGR 102 16663Naef D Latham D W Mayor M et al 2001 AampA 375 L27Perna R Heng K amp Pont F 2012 ApJ 751 59Pont F Knutson H Gilliland R L Moutou C amp Charbonneau D

2008 MNRAS 385 109Rauscher E amp Menou K 2010 ApJ 714 1334Rauscher E amp Menou K 2012 ApJ 750 96Saumon D Marley M S Abel M Frommhold L amp Freedman R S

2012 ApJ 750 74Showman A P Cho J Y-K amp Menou K 2011 in Exoplanets ed S Seager

(Tucson AZ Univ Arizona Press) 471Showman A P Cooper C S Fortney J J amp Marley M S 2008 ApJ

682 559Showman A P Fortney J J Lian Y et al 2009 ApJ 699 564Showman A P amp Guillot T 2002 AampA 385 166Showman A P amp Polvani L M 2011 ApJ 738 71Swain M R Vasisht G amp Tinetti G 2008 Natur 452 329Thrastarson H T amp Cho J Y 2010 ApJ 716 144Wolszczan A 1994 Sci 264 538

19

  • 1 INTRODUCTION
  • 2 MODEL
    • 21 Updates to SPARCMITgcm
    • 22 Grid Simulations
      • 3 RESULTS
        • 31 Circulation Regime
        • 32 Effect of Eccentricity
        • 33 Effect of Varying Average Stellar Flux
        • 34 Global Average Temperature versus Eccentricity and F
        • 35 Synchronous versus Pseudo-synchronous Rotation
        • 36 Momentum Budget and Equatorial Superrotation
          • 4 OBSERVATIONAL IMPLICATIONS
            • 41 Observational Effects of Varying ω
            • 42 High-e Model Integrations at Varying F
              • 5 CONCLUSIONS
              • APPENDIX VALIDATION TESTS FOR UPDATED RADIATIVE TRANSFER SCHEME
              • REFERENCES
Page 11: THREE-DIMENSIONAL ATMOSPHERIC CIRCULATION OF HOT …showman/publications/kataria-etal-2013.pdf · planets’ atmospheric composition, temperature structure, and circulation. A fifth

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

0 05 1 15 2 25 3 35620

640

660

680

700

720

740ltFgt = 11617 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050e=075

0 05 1 15 2 25 3 35880

890

900

910

920

930

940

950

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

ltFgt = 73680 Wmminus1

e=025e=050

0 05 1 15 2 25 3 35 4960

980

1000

1020

1040

1060

1080

1100

1120

1140

1160ltFgt = 185691 Wmminus1

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

Mean anomaly

e=025e=050e=075

0 05 1 15 2 25 3 35 4 451100

1150

1200

1250

1300

1350ltFgt = 468183 Wmminus1

Mean anomaly

Glo

bal

lyminusa

vera

ged

tem

per

atu

re (

K)

e=025e=050

Figure 9 Plots of global average temperature vs mean anomaly for each modelintegration Each plot contains profiles of a given average stellar flux (468183185691 73680 and 11617 W mminus1) with blue green and red representingprofiles for e = 025 050 and 075 respectively

(A color version of this figure is available in the online journal)

03 04 05 06 070

05

1

15

2

25

3

eccentricity

Pea

k o

ffse

t fr

om

per

iap

se (

Ear

th d

ays)

468183 Wmminus1

185691 Wmminus1

73680 Wmminus1

11617 Wmminus1

Figure 10 Time after periapse passage at which the global-average temperaturepeaks (from Figure 9) vs eccentricity for all model integrations The bluegreen red and teal profiles correspond to average stellar fluxes of 468183185691 73680 and 11617 W mminus1 respectively

(A color version of this figure is available in the online journal)

Polvani 2011 hereafter SP2011) As discussed in their paperthe generation of superrotation requires up-gradient momentumtransport that pumps angular momentum from outside the jet towithin it Only transport by waves and eddies can satisfy thisrequirement (Hide 1969)

SP2011 demonstrate that for synchronous hot Jupiters onnear-circular orbits such as HD 189733b and HD 209458bequatorial superrotation is generated by standing planetaryscale Kelvin and Rossby waves at the equator and midlati-tudes respectively that are themselves the dynamical responseto the planetrsquos large dayndashnight heating contrast Because theKelvin waves propagate eastward and the Rossby waves prop-agate westward a phase tilt of the wind vectors emerges fromnorthwest to southeast in the northern hemisphere and south-west to northeast in the southern hemisphere with an over-all shape resembling a chevron pointing eastward This patterncauses meridional eddy angular momentum fluxes that convergeangular momentum onto the equator generating equatorialsuperrotation

Despite the fact that not all our model integrations aresynchronous our results suggest that similar mechanisms arestill at play The chevron shape that SP2011 describe is exactlywhat we noted earlier in our plots of the wind and temperatureprofiles from Figure 5 suggesting that as in SP2011 our modelsexhibit an equatorward flux of eddy angular momentum Thisis demonstrated quantitatively in Figure 12 which shows thezonally averaged meridional flux of relative zonal eddy angularmomentum uprimevprime cos φ at snapshots throughout the orbit (asin Figures 5 and 6) Here uprime and vprime represent the deviationof the zonal and meridional winds respectively from theirzonal averages Positive values indicate a northward flux ofeastward eddy angular momentum and negative values indicatea southward flux of eastward eddy angular momentum Atapoapse the small dayndashnight forcing causes poleward transportof eddy angular momentum at latitudes greater than 50 butequatorward transport of eddy angular momentum equator of50 latitude the amplitudes are weak due to the low insolationat apoapse As the planet moves closer to the star (middlerow left column) the increase in dayndashnight forcing strengthens

11

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)e = 050 pseudominussynchronous rotation

minus200

200

200

600

600

1000

10001400

180022002600

3000

3400

3800

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

synchronous rotation

minus600

minus600

minus200

minus200 minus200

minus200

minus200

200

200

600

10001400180022002600300034003800

4200

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

Figure 11 Comparison of pseudo-synchronous (Trot = 3046 times 105 s) vs synchronous (Trot = 8544 times 105 s) rotation for a case with Teq = 755 K e = 050 anda = 00848 AU Wind and temperature snapshots are taken at periapse (middle row) and apoapse (bottom row) at 5008 Earth days and 5013 Earth days respectivelyThe vertical bar denotes substellar longitude

(A color version of this figure is available in the online journal)

the equatorward flux of eastward eddy angular momentumat latitudes between 0 and 50 in both hemispheres Themomentum flux grows and persists until hours after periapseAngular momentum thus converges onto the equator andmaintains the equatorial jet against westward accelerationscaused by advection friction (if any) and Coriolis forces

These mechanisms also explain the difference in wind speedsbetween the synchronous and pseudo-synchronous cases inFigure 11 The synchronous case has a higher dayndashnight heatingcontrast hence a higher equatorward flux of eddy momentumleading to a stronger narrower equatorial jet Moreover theequatorial jet in the pseudo-synchronous case extends to deeper

12

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=18000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

Figure 12 Eddy momentum flux uprimevprime cos φ throughout the orbit of a model where Teq = 951 K e = 05 and a = 00534 AU Positive values indicate a northwardflux of eastward eddy angular momentum while negative values indicate a southward flux of eastward eddy angular momentum The maximum momentum fluxoccurs hours after periapse passage (f = 72) The snapshots correspond to model outputs from 6103 to 6108 Earth days

(A color version of this figure is available in the online journal)

13

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

Figure 13 Differing orbital viewing geometries explored in this study ω = 180 ω = 270 and ω = 360

(A color version of this figure is available in the online journal)

pressures than that of the synchronous case this is most likelydue to a change in momentum budget

4 OBSERVATIONAL IMPLICATIONS

Observations of transiting exoplanets in circular orbits areknown to be shaped by the atmospheric circulation As wasfirst predicted for hot Jupiters by Showman amp Guillot (2002)and later observationally confirmed for hot Jupiter HD 189733bby Knutson et al (2007) an equatorial superrotating jet causesan eastward displacement of the hot spot from the substellarpoint under appropriate conditions Hence a light curve showsa peak in infrared flux minutes to hours before secondaryeclipse We expect such a situation for eccentric exoplanetsas well However observations of eccentric planets can becomplicated because the flux received by the planet variesthroughout its orbit Previous studies explored such situationsbut have not used three-dimensional models (eg Langton ampLaughlin 2008a 2008b Cowan amp Agol 2011) We can expandon these studies and the three-dimensional study by Lewis et al(2010) by using our SPARC model results

We choose to generate light curves following the proceduresdescribed in Fortney et al (2006) for each model integrationat three values of the argument of periastron (ω) the anglebetween the radius vector to the ascending node and the periapseof the orbit In particular we choose orientations where transitoccurs before periapse (at f = minus90ω = 180) transit occursat apoapse (ω = 270) and transit occurs after periapse (atf = +90ω = 360) (Figure 13) These values reflect thebroad range of viewing geometries of transiting exoplanets forexample HAT-P-2b (e = 052) HAT-P-17b (e = 035) andHD 97658b (e = 013) have ω near 180 WASP-8b (e = 031)HAT-P-31b (e = 025) and HD 80606b (e = 093) have ω near270 and GJ 436b (e = 015) XO-3b (e = 026) and HAT-P-11b (e = 020) have ω near 360 (see exoplaneteu for a full listof values) Note that when transit occurs the nightside is visibleto Earth while the dayside is visible at the time of secondaryeclipse

41 Observational Effects of Varying ω

Using the nominal case (e = 050 a = 00534 AU Teq =951 K) we illustrate the dependence of full-orbit light curves onthe value of ω Figure 14 shows three light curves of the same

model integration but at the differing values of ω described inthe previous section Hence the differences in light curves resultsolely from the differing viewing geometries The planetstarflux ratio is plotted as a function of Earth days from periapsein the warm Spitzer wavelength bands (36 and 45 μm) andJHK bands (126 165 and 220 μm) The solid and dashedlines correspond to transit and secondary eclipse respectivelyFor the light curve where ω = 180 (top panel) the peak ininfrared (IR) flux occurs approximately 6 hr after periapse andbefore secondary eclipse The timing of the peak is due toa combination of the radiative lag of the planet and the factthat the hottest regions rotate into an Earth-facing orientationhours after periapse passage The dayside which is most visiblenear secondary eclipse is hottest after periapse Additionallytemperatures reach peak values at times close to when thosehottest regions are facing Earth This helps explain the highamplitude of the light curve peak in the ω = 180 case relative tothe other two cases These effects can be visualized by plottingthe temperature near the photosphere along the equator in apolar projection (Figure 15) Snapshots throughout the orbit plotthe temperature in a polar projection as a function of pressureand orbital position at a latitudinal slice near the equatorthe solid and dashed lines denote the substellar longitude andEarth-facing longitude respectively Based on the temperaturesat the Earth-facing longitude it is apparent that the peak in IRflux should occur after periapse and before secondary eclipse(see f = 12 and 72) this is indeed the case

For the same model but instead with ω = 270 the peak inIR flux occurs 1 hr after eclipse and periapse In this case thedayside is visible near periapse and therefore we are seeing theplanet increase in temperature (and IR flux) during its closestapproach to the star However the hottest regions are displacedeastward of the substellar point and therefore the hottest regionsface Earth significantly before periapse passage (Figure 15 atf asymp 72) Thus by the time the peak temperatures are reachednear and after periapse the hottest regions are already rotatingout of view from Earth This explains the lower amplitude ofthe light curve peak as compared to the peak of the ω = 180case The IR peak decreases as the planet moves further awayfrom the star along our line of sight A second ramp-up in fluxoccurs approximately 1 day after periapseeclipse this is due tothe hottest regions rotating back into view as the planet movestoward apoapse

14

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015P

lane

tSta

r F

lux

Rat

io

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

Figure 14 Three light curves of the IR flux (expressed as a planetstar fluxratio) vs time throughout the full orbit for a single model integration whereTeq = 951 K e = 050 and a = 00534 AU The flux is plotted in warm Spitzerbandpasses 36 and 45 μm and JHK bands (126 165 220 μm respectively)These three cases only differ by their value of ω shown in Figure 13 The solidlines denote transit while the dotted lines denote secondary eclipse

(A color version of this figure is available in the online journal)

In the case where ω = 360 two distinct peaks are presentmdashasmaller peak that occurs after secondary eclipse and sim3ndash4 hrbefore periapse and a second higher peak that occurs aftertransit The first peak occurs as some of the dayside is visible asthe planet is heated approaching periapse The peak decreasesas the dayside rotates out of view For this geometry the hottestregions are facing away from Earth when they reach their peaktemperatures (Figure 15) The second peak occurs after transitas the hottest regions again rotate into view At this point in theorbit peak temperatures have already occurred and thereforethis peak is lower in amplitude than the other two light curvesHowever because the planet is hotter than it was after secondary

eclipse this second peak has a higher amplitude than the firstone

42 High-e Model Integrations at Varying 〈F 〉If an eccentric planet has a large enough dayndashnight tempera-

ture difference and a sufficiently rapid rotation rate its resultantlight curve can show a periodic rise and fall in flux through-out its orbit The light curves shown in Figure 16 are twoe = 075 model integrations that differ in 〈F 〉 the top panelshows a case where 〈F 〉 = 468183 W mminus2 (Teq = 1199 K)and a = 00385 AU while the bottom panel is a case where〈F 〉 = 185691 W mminus2 (Teq = 951 K) and a = 00611 AU Forboth cases light curves are shown for the ω = 180 geome-try (see Figure 13) The differences in 〈F 〉 lead to differencesin light curve amplitude in the two models the closer hotterplanet exhibits a higher flux amplitude Not surprisingly thetiming of peak IR flux for both models resembles the ω = 180case in Figure 14 The peak in each case occurs after periapseand before eclipse due to the combined effects of the ther-mal lag of the planet and geometry of the system Additionallyboth light curves exhibit a quasi-periodic rise and fall of IR fluxdays after periapse passage This phenomenon called ldquoringingrdquowas also seen in models by Langton amp Laughlin (2008a) andCowan amp Agol (2011) caused by the hottest point of the planetrotating in and out of view from Earth For ringing to occurthe temperature difference from dayside to nightside must belarge and it must survive for multiple planetary rotation periodsIn the models shown in Figure 16 this is aided by the rapidpseudo-synchronous rotation rate associated with an eccentric-ity of 075 (Table 2) For the top case the dayndashnight tempera-ture difference is over sim1300 K while the bottom case variesby sim750 K (Figure 7) The light curves from Figure 14 do notexhibit ringing because the dayndashnight temperature difference islow (sim300ndash400 K)

Because the radiative and dynamical timescales in the modelsby Cowan amp Agol (2011) are calculated by a simple scaling withtemperature the ringing seen in their simulated light curvesoccurs with a period equal to the planetrsquos assumed solid-bodyrotation period in the inertial frame of the winds In realityan eccentric exoplanet approaching apoapse would becomeincreasingly cold the winds would weaken and decrease andhence the ringing should be non-periodic Indeed for ourthree dimensional models the ringing in each case has varyingperiodsmdashthe top case has a ringing period ranging from 03 to04 days (Trot = 035 days) while the bottom panel has a periodranging 07ndash08 days (Trot = 07 days)

Despite these differences this study and the studies byLangton amp Laughlin (2008a) and Cowan amp Agol (2011) illus-trate the influence of atmospheric dynamics on the observationsof eccentric exoplanets The light curve features described herecould be seen in future follow-up observations of eccentric ex-oplanets and observers can use the conclusions drawn aboveto relate the observations to the planetrsquos atmospheric structureFor example if ringing is present in an IR light curve the planetmust have a strong dayndashnight temperature asymmetry and hencewe would expect the planet to have strong superrotation at theequator (Showman amp Polvani 2011) This is probably the casefor highly eccentric (e gt 05) short-period exoplanets such asHD 17156b and HD 80606b If ringing were observed it wouldalso allow an estimate of the sum of the rotation speed and thesuperrotation speed thereby placing constraints on both windspeeds and rotation rates of eccentric hot Jupiters The timing ofpeak IR flux relative to transit and secondary eclipse and also

15

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

f=000deg

(K)800

850

900

950

1000

f=minus1242deg

(K)800

850

900

950

1000f=1242deg

(K)800

850

900

950

1000

f=minus7227deg

(K)800

850

900

950

1000f=7227deg

(K)800

850

900

950

1000

f=minus11988deg

(K)800

850

900

950

1000f=11988deg

(K)800

850

900

950

1000

f=minus18000deg

(K)800

850

900

950

1000

Figure 15 Plots of temperature vs longitude and pressure in equatorial slices throughout one full orbit for our nominal model with Teq = 951 K e = 050 anda = 00534 AU Each annulus shows temperature vs longitude and log-pressure for a snapshot at a particular orbital position (denoted by the value of f) as viewedfrom the planetrsquos north pole Pressures between 30 and 43 mbar are displayed in each annulus The different annuli mark the time evolution throughout one completeorbit starting from apoapse for the top annulus proceeding to periapse and then back to apoapse (follow arrows moving counterclockwise around the plot) Thesesnapshots correspond to model outputs from 6103 to 6108 Earth days The solid line in each snapshot indicates the longitude of the substellar point while the dashedline denotes the Earth-facing longitude for the particular case of ω = 180 For the cases of ω = 270 and 360 degrees exactly the same plot is valid except that thesub-Earth longitude would be aiming upward for ω = 270 and right for ω = 360

(A color version of this figure is available in the online journal)

16

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

e = 075 a = 00385 AU

-15 -10 -05 00 05 10 15Time after periapse passage (days)

0000

0001

0002

0003

0004

Pla

netS

tar

Flu

x R

atio

e = 075 a = 00611 AU

-3 -2 -1 0 1 2 3Time after periapse passage (days)

00000

00005

00010

00015

00020

00025

Pla

netS

tar

Flu

x R

atio

Figure 16 Light Curves for two different e = 075 model integrationsTeq = 1199 K a = 00385 AU (top) and Teq = 951 K a = 00611 AU(bottom) both with ω = 180 The solid line denotes transit while the dottedline denotes secondary eclipse

(A color version of this figure is available in the online journal)

relative to orbital position can help to constrain the circulationfurther

5 CONCLUSIONS

We present three-dimensional circulation models coupledwith a two-stream non-gray radiative transfer scheme fora number of theoretical eccentric hot Jupiters We haveshown that as in published models with zero eccentricity ourhigh-eccentricity circulation models are dominated by eastwardflow at photospheric levels which cause an eastward displace-ment of the hottest regions from the substellar point The rapidrotation rates associated with pseudo-synchronization at higheccentricity lead to a small Rossby deformation radius andin some cases multiple jets in the atmosphere Global-meantemperatures and dayndashnight temperature differences peak notat periapse but several hours afterward due to finite radiativetimescales in the planetrsquos atmosphere

Furthermore we show that equatorial superrotation is gener-ated and maintained by eddies formed by the strong dayndashnightheating contrast which induce a flux of momentum from mid-latitudes to the equator The eddy magnitudes and momentumfluxes peak just after periapse passage leading to variations inthe zonal-mean flow throughout the orbit

Figure 17 Globally averaged pressurendashtemperature profiles from 1-D radiative-convective radiative transfer models of HD 189733b with 11 frequency bins(red profiles) 30 frequency bins (green profiles) and 196 frequency bins (blackprofiles) From left to right the profiles vary in stellar flux from the nominalvalue as well as 2times 4times and 8times higher flux The 8times higher flux case has atemperature inversion

(A color version of this figure is available in the online journal)

Lastly we have shown that the spatial and temporal vari-ations of the wind and temperature structure as well as theorbital viewing geometry of the system with respect to Earthcan affect the time and amplitude of peak IR flux seen infull-orbit light curves Depending on the viewing geometry ofthe orbit relative to Earth we find that peaks in IR flux thateither lead or lag periapse are possible in all cases a combi-nation of temporal effects (temperatures changing over time)and geometric effects (hot spots rotating into or out of view)are important in controlling the timing and amplitude of theflux peaks In cases where the dayndashnight temperature contrastis large and the rotational period is short the light curve canalso exhibit ldquoringingrdquo in flux as the hottest region of the planetrotates in and out of view This ringing is non-periodic due tothe variation in stellar heating as a function of distance

This work was supported by NASA Origins and PlanetaryAtmospheres grants to APS TK also acknowledges supportfrom the Harriet P Jenkins Pre-Doctoral Fellowship Program(JPFP) Resources supporting this work were provided bythe NASA High-End Computing (HEC) Program through theNASA Advanced Supercomputing (NAS) Division at AmesResearch Center The authors thank the anonymous referee fortheir helpful comments and suggestions

APPENDIX

VALIDATION TESTS FOR UPDATEDRADIATIVE TRANSFER SCHEME

To validate our updated radiative transfer scheme using 11frequency bins we conducted the following tests

1 First we tested the sensitivity of the number of frequencybins in a one-dimensional radiative-convective model ofHD 189733b (a = 00313 AU e = 00 RP = 82396 times107 m g = 214 m sminus2) Figure 17 compares the globallyaveraged pressurendashtemperature profiles of models run with11 frequency bins 30 frequency bins and 196 frequencybins at (from left to right) 1times 2times 4times and 8times theaverage stellar flux of HD 189733b For each group of

17

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)30 opacity bins

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)

11 opacity bins

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

30 opacity bins

minus600

minus200minus200

200

200

600

6001000

1000

14001800

2200

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

11 opacity bins

minus600

minus200

minus200

200

200

600

600

10001000

1400

1800

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

30 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

11 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

Figure 18 Globally averaged pressurendashtemperature profiles (top row) zonal-mean zonal wind (middle row) and the windtemperature profiles at 30 mbar (bottomrow) for the 30-bin (left column) and 11-bin (right column) model integrations of HD 189733b Snapshots were output at 250 Earth days

(A color version of this figure is available in the online journal)

models the 11-bin profile agrees well with the 30- and196-bin cases from 1 μbar to 1 bar the temperature variesby only tens of K Below 10 bars the 30- and 11-binprofiles differ from the 196-bin cases by up to 50 Kdue to the loss of resolution at low wavenumbers Inthe case of the highest average flux the large amount of

heating leads to a temperature inversion that is not fullycaptured by the 11-bin model but can be improved witha better initial guess at the pressure-temperature (P-T)profile

2 After testing the new radiative transfer (RT) scheme forone-dimensional models we ran full three-dimensional

18

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

simulations of HD 189733b at its nominal average stellarflux using the SPARCMITgcm Figure 18 shows the glob-ally averaged P-T profiles (top row) zonal-mean zonal wind(middle row) and windtemperature profiles at 30 mbar(bottom row) for the 30-bin (left column) and 11-bin (rightcolumn) RT setups The globally averaged profiles arenearly identical with any temperature differences withina few K The zonal wind plots also show good agree-ment with similar peak speeds (gt2600 m sminus1) equatorialjet width (sim60) and jet level (lt100 bar) At 30 mbarthe 30- and 11-bin models share similar dayside-nightsidetemperature structure with a hot spot (1100 K) eastward ofthe substellar point and two colder regions on the nightsidein the mid-latitudes

Both tests show that the bulk circulation and temperaturestructure is retained in the transition from 30 to 11 frequencybins hence we proceed to use the SPARCMITgcm with thenew scheme for these and future model simulations

REFERENCES

Adcroft A Campin J-M Hill C amp Marshall J 2004 MWRv 132 2845Charbonneau D Knutson H A Barman T et al 2008 ApJ 686 1341Cho J Y-K Menou K Hansen B M S amp Seager S 2008 ApJ

675 817Cooper C S amp Showman A P 2005 ApJL 629 L45Cooper C S amp Showman A P 2006 ApJ 649 1048Cowan N B amp Agol E 2011 ApJ 726 82Deming D Harrington J Laughlin G et al 2007 ApJL 667 L199Dobbs-Dixon I Cumming A amp Lin D N C 2010 ApJ 710 1395Dobbs-Dixon I amp Lin D N C 2008 ApJ 673 513Fortney J J Cooper C S Showman A P Marley M S amp Freedman R S

2006 ApJ 652 746Fortney J J Lodders K Marley M S amp Freedman R S 2008 ApJ

678 1419Fortney J J Marley M S Lodders K Saumon D amp Freedman R

2005 ApJL 627 L69

Fu Q amp Liou K N 1992 JAtS 49 2140Goody R West R Chen L amp Crisp D 1989 JQSRT 42 539Greenberg R 2009 ApJL 698 L42Heng K 2012 ApJL 761 L1Heng K Frierson D M W amp Phillipps P J 2011 MNRAS 418 2669Hide R 1969 JAtS 26 841Holton J R 2004 An Introduction to Dynamic Meteorology (Waltham MA

Academic)Hut P 1981 AampA 99 126Irwin J Charbonneau D Nutzman P et al 2008 ApJ 681 636Ivanov P B amp Papaloizou J C B 2004 MNRAS 353 1161Knutson H A Charbonneau D Allen L E Burrows A amp Megeath S T

2008 ApJ 673 526Knutson H A Charbonneau D Allen L E et al 2007 Natur 447 183Knutson H A Charbonneau D Cowan N B et al 2009 ApJ 690 822Langton J amp Laughlin G 2008a ApJ 674 1106Langton J amp Laughlin G 2008b AampA 483 25Lewis N K Showman A P Fortney J J et al 2010 ApJ 720 344Lodders K 2003 ApJ 591 1220Marley M S amp McKay C P 1999 Icar 138 268Mayor M amp Queloz D 1995 Natur 378 355Menou K amp Rauscher E 2009 ApJ 700 887Mlawer E J Taubman S J Brown P D Iacono M J amp Clough S A

1997 JGR 102 16663Naef D Latham D W Mayor M et al 2001 AampA 375 L27Perna R Heng K amp Pont F 2012 ApJ 751 59Pont F Knutson H Gilliland R L Moutou C amp Charbonneau D

2008 MNRAS 385 109Rauscher E amp Menou K 2010 ApJ 714 1334Rauscher E amp Menou K 2012 ApJ 750 96Saumon D Marley M S Abel M Frommhold L amp Freedman R S

2012 ApJ 750 74Showman A P Cho J Y-K amp Menou K 2011 in Exoplanets ed S Seager

(Tucson AZ Univ Arizona Press) 471Showman A P Cooper C S Fortney J J amp Marley M S 2008 ApJ

682 559Showman A P Fortney J J Lian Y et al 2009 ApJ 699 564Showman A P amp Guillot T 2002 AampA 385 166Showman A P amp Polvani L M 2011 ApJ 738 71Swain M R Vasisht G amp Tinetti G 2008 Natur 452 329Thrastarson H T amp Cho J Y 2010 ApJ 716 144Wolszczan A 1994 Sci 264 538

19

  • 1 INTRODUCTION
  • 2 MODEL
    • 21 Updates to SPARCMITgcm
    • 22 Grid Simulations
      • 3 RESULTS
        • 31 Circulation Regime
        • 32 Effect of Eccentricity
        • 33 Effect of Varying Average Stellar Flux
        • 34 Global Average Temperature versus Eccentricity and F
        • 35 Synchronous versus Pseudo-synchronous Rotation
        • 36 Momentum Budget and Equatorial Superrotation
          • 4 OBSERVATIONAL IMPLICATIONS
            • 41 Observational Effects of Varying ω
            • 42 High-e Model Integrations at Varying F
              • 5 CONCLUSIONS
              • APPENDIX VALIDATION TESTS FOR UPDATED RADIATIVE TRANSFER SCHEME
              • REFERENCES
Page 12: THREE-DIMENSIONAL ATMOSPHERIC CIRCULATION OF HOT …showman/publications/kataria-etal-2013.pdf · planets’ atmospheric composition, temperature structure, and circulation. A fifth

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)e = 050 pseudominussynchronous rotation

minus200

200

200

600

600

1000

10001400

180022002600

3000

3400

3800

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

synchronous rotation

minus600

minus600

minus200

minus200 minus200

minus200

minus200

200

200

600

10001400180022002600300034003800

4200

(m sminus1)minus500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=000deg

(K)500

550

600

650

700

750

800

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 pseudominussynchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

e = 050 synchronous rotation f=18000deg

(K)

520

540

560

580

600

620

640

660

680

Figure 11 Comparison of pseudo-synchronous (Trot = 3046 times 105 s) vs synchronous (Trot = 8544 times 105 s) rotation for a case with Teq = 755 K e = 050 anda = 00848 AU Wind and temperature snapshots are taken at periapse (middle row) and apoapse (bottom row) at 5008 Earth days and 5013 Earth days respectivelyThe vertical bar denotes substellar longitude

(A color version of this figure is available in the online journal)

the equatorward flux of eastward eddy angular momentumat latitudes between 0 and 50 in both hemispheres Themomentum flux grows and persists until hours after periapseAngular momentum thus converges onto the equator andmaintains the equatorial jet against westward accelerationscaused by advection friction (if any) and Coriolis forces

These mechanisms also explain the difference in wind speedsbetween the synchronous and pseudo-synchronous cases inFigure 11 The synchronous case has a higher dayndashnight heatingcontrast hence a higher equatorward flux of eddy momentumleading to a stronger narrower equatorial jet Moreover theequatorial jet in the pseudo-synchronous case extends to deeper

12

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=18000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

Figure 12 Eddy momentum flux uprimevprime cos φ throughout the orbit of a model where Teq = 951 K e = 05 and a = 00534 AU Positive values indicate a northwardflux of eastward eddy angular momentum while negative values indicate a southward flux of eastward eddy angular momentum The maximum momentum fluxoccurs hours after periapse passage (f = 72) The snapshots correspond to model outputs from 6103 to 6108 Earth days

(A color version of this figure is available in the online journal)

13

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

Figure 13 Differing orbital viewing geometries explored in this study ω = 180 ω = 270 and ω = 360

(A color version of this figure is available in the online journal)

pressures than that of the synchronous case this is most likelydue to a change in momentum budget

4 OBSERVATIONAL IMPLICATIONS

Observations of transiting exoplanets in circular orbits areknown to be shaped by the atmospheric circulation As wasfirst predicted for hot Jupiters by Showman amp Guillot (2002)and later observationally confirmed for hot Jupiter HD 189733bby Knutson et al (2007) an equatorial superrotating jet causesan eastward displacement of the hot spot from the substellarpoint under appropriate conditions Hence a light curve showsa peak in infrared flux minutes to hours before secondaryeclipse We expect such a situation for eccentric exoplanetsas well However observations of eccentric planets can becomplicated because the flux received by the planet variesthroughout its orbit Previous studies explored such situationsbut have not used three-dimensional models (eg Langton ampLaughlin 2008a 2008b Cowan amp Agol 2011) We can expandon these studies and the three-dimensional study by Lewis et al(2010) by using our SPARC model results

We choose to generate light curves following the proceduresdescribed in Fortney et al (2006) for each model integrationat three values of the argument of periastron (ω) the anglebetween the radius vector to the ascending node and the periapseof the orbit In particular we choose orientations where transitoccurs before periapse (at f = minus90ω = 180) transit occursat apoapse (ω = 270) and transit occurs after periapse (atf = +90ω = 360) (Figure 13) These values reflect thebroad range of viewing geometries of transiting exoplanets forexample HAT-P-2b (e = 052) HAT-P-17b (e = 035) andHD 97658b (e = 013) have ω near 180 WASP-8b (e = 031)HAT-P-31b (e = 025) and HD 80606b (e = 093) have ω near270 and GJ 436b (e = 015) XO-3b (e = 026) and HAT-P-11b (e = 020) have ω near 360 (see exoplaneteu for a full listof values) Note that when transit occurs the nightside is visibleto Earth while the dayside is visible at the time of secondaryeclipse

41 Observational Effects of Varying ω

Using the nominal case (e = 050 a = 00534 AU Teq =951 K) we illustrate the dependence of full-orbit light curves onthe value of ω Figure 14 shows three light curves of the same

model integration but at the differing values of ω described inthe previous section Hence the differences in light curves resultsolely from the differing viewing geometries The planetstarflux ratio is plotted as a function of Earth days from periapsein the warm Spitzer wavelength bands (36 and 45 μm) andJHK bands (126 165 and 220 μm) The solid and dashedlines correspond to transit and secondary eclipse respectivelyFor the light curve where ω = 180 (top panel) the peak ininfrared (IR) flux occurs approximately 6 hr after periapse andbefore secondary eclipse The timing of the peak is due toa combination of the radiative lag of the planet and the factthat the hottest regions rotate into an Earth-facing orientationhours after periapse passage The dayside which is most visiblenear secondary eclipse is hottest after periapse Additionallytemperatures reach peak values at times close to when thosehottest regions are facing Earth This helps explain the highamplitude of the light curve peak in the ω = 180 case relative tothe other two cases These effects can be visualized by plottingthe temperature near the photosphere along the equator in apolar projection (Figure 15) Snapshots throughout the orbit plotthe temperature in a polar projection as a function of pressureand orbital position at a latitudinal slice near the equatorthe solid and dashed lines denote the substellar longitude andEarth-facing longitude respectively Based on the temperaturesat the Earth-facing longitude it is apparent that the peak in IRflux should occur after periapse and before secondary eclipse(see f = 12 and 72) this is indeed the case

For the same model but instead with ω = 270 the peak inIR flux occurs 1 hr after eclipse and periapse In this case thedayside is visible near periapse and therefore we are seeing theplanet increase in temperature (and IR flux) during its closestapproach to the star However the hottest regions are displacedeastward of the substellar point and therefore the hottest regionsface Earth significantly before periapse passage (Figure 15 atf asymp 72) Thus by the time the peak temperatures are reachednear and after periapse the hottest regions are already rotatingout of view from Earth This explains the lower amplitude ofthe light curve peak as compared to the peak of the ω = 180case The IR peak decreases as the planet moves further awayfrom the star along our line of sight A second ramp-up in fluxoccurs approximately 1 day after periapseeclipse this is due tothe hottest regions rotating back into view as the planet movestoward apoapse

14

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015P

lane

tSta

r F

lux

Rat

io

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

Figure 14 Three light curves of the IR flux (expressed as a planetstar fluxratio) vs time throughout the full orbit for a single model integration whereTeq = 951 K e = 050 and a = 00534 AU The flux is plotted in warm Spitzerbandpasses 36 and 45 μm and JHK bands (126 165 220 μm respectively)These three cases only differ by their value of ω shown in Figure 13 The solidlines denote transit while the dotted lines denote secondary eclipse

(A color version of this figure is available in the online journal)

In the case where ω = 360 two distinct peaks are presentmdashasmaller peak that occurs after secondary eclipse and sim3ndash4 hrbefore periapse and a second higher peak that occurs aftertransit The first peak occurs as some of the dayside is visible asthe planet is heated approaching periapse The peak decreasesas the dayside rotates out of view For this geometry the hottestregions are facing away from Earth when they reach their peaktemperatures (Figure 15) The second peak occurs after transitas the hottest regions again rotate into view At this point in theorbit peak temperatures have already occurred and thereforethis peak is lower in amplitude than the other two light curvesHowever because the planet is hotter than it was after secondary

eclipse this second peak has a higher amplitude than the firstone

42 High-e Model Integrations at Varying 〈F 〉If an eccentric planet has a large enough dayndashnight tempera-

ture difference and a sufficiently rapid rotation rate its resultantlight curve can show a periodic rise and fall in flux through-out its orbit The light curves shown in Figure 16 are twoe = 075 model integrations that differ in 〈F 〉 the top panelshows a case where 〈F 〉 = 468183 W mminus2 (Teq = 1199 K)and a = 00385 AU while the bottom panel is a case where〈F 〉 = 185691 W mminus2 (Teq = 951 K) and a = 00611 AU Forboth cases light curves are shown for the ω = 180 geome-try (see Figure 13) The differences in 〈F 〉 lead to differencesin light curve amplitude in the two models the closer hotterplanet exhibits a higher flux amplitude Not surprisingly thetiming of peak IR flux for both models resembles the ω = 180case in Figure 14 The peak in each case occurs after periapseand before eclipse due to the combined effects of the ther-mal lag of the planet and geometry of the system Additionallyboth light curves exhibit a quasi-periodic rise and fall of IR fluxdays after periapse passage This phenomenon called ldquoringingrdquowas also seen in models by Langton amp Laughlin (2008a) andCowan amp Agol (2011) caused by the hottest point of the planetrotating in and out of view from Earth For ringing to occurthe temperature difference from dayside to nightside must belarge and it must survive for multiple planetary rotation periodsIn the models shown in Figure 16 this is aided by the rapidpseudo-synchronous rotation rate associated with an eccentric-ity of 075 (Table 2) For the top case the dayndashnight tempera-ture difference is over sim1300 K while the bottom case variesby sim750 K (Figure 7) The light curves from Figure 14 do notexhibit ringing because the dayndashnight temperature difference islow (sim300ndash400 K)

Because the radiative and dynamical timescales in the modelsby Cowan amp Agol (2011) are calculated by a simple scaling withtemperature the ringing seen in their simulated light curvesoccurs with a period equal to the planetrsquos assumed solid-bodyrotation period in the inertial frame of the winds In realityan eccentric exoplanet approaching apoapse would becomeincreasingly cold the winds would weaken and decrease andhence the ringing should be non-periodic Indeed for ourthree dimensional models the ringing in each case has varyingperiodsmdashthe top case has a ringing period ranging from 03 to04 days (Trot = 035 days) while the bottom panel has a periodranging 07ndash08 days (Trot = 07 days)

Despite these differences this study and the studies byLangton amp Laughlin (2008a) and Cowan amp Agol (2011) illus-trate the influence of atmospheric dynamics on the observationsof eccentric exoplanets The light curve features described herecould be seen in future follow-up observations of eccentric ex-oplanets and observers can use the conclusions drawn aboveto relate the observations to the planetrsquos atmospheric structureFor example if ringing is present in an IR light curve the planetmust have a strong dayndashnight temperature asymmetry and hencewe would expect the planet to have strong superrotation at theequator (Showman amp Polvani 2011) This is probably the casefor highly eccentric (e gt 05) short-period exoplanets such asHD 17156b and HD 80606b If ringing were observed it wouldalso allow an estimate of the sum of the rotation speed and thesuperrotation speed thereby placing constraints on both windspeeds and rotation rates of eccentric hot Jupiters The timing ofpeak IR flux relative to transit and secondary eclipse and also

15

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

f=000deg

(K)800

850

900

950

1000

f=minus1242deg

(K)800

850

900

950

1000f=1242deg

(K)800

850

900

950

1000

f=minus7227deg

(K)800

850

900

950

1000f=7227deg

(K)800

850

900

950

1000

f=minus11988deg

(K)800

850

900

950

1000f=11988deg

(K)800

850

900

950

1000

f=minus18000deg

(K)800

850

900

950

1000

Figure 15 Plots of temperature vs longitude and pressure in equatorial slices throughout one full orbit for our nominal model with Teq = 951 K e = 050 anda = 00534 AU Each annulus shows temperature vs longitude and log-pressure for a snapshot at a particular orbital position (denoted by the value of f) as viewedfrom the planetrsquos north pole Pressures between 30 and 43 mbar are displayed in each annulus The different annuli mark the time evolution throughout one completeorbit starting from apoapse for the top annulus proceeding to periapse and then back to apoapse (follow arrows moving counterclockwise around the plot) Thesesnapshots correspond to model outputs from 6103 to 6108 Earth days The solid line in each snapshot indicates the longitude of the substellar point while the dashedline denotes the Earth-facing longitude for the particular case of ω = 180 For the cases of ω = 270 and 360 degrees exactly the same plot is valid except that thesub-Earth longitude would be aiming upward for ω = 270 and right for ω = 360

(A color version of this figure is available in the online journal)

16

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

e = 075 a = 00385 AU

-15 -10 -05 00 05 10 15Time after periapse passage (days)

0000

0001

0002

0003

0004

Pla

netS

tar

Flu

x R

atio

e = 075 a = 00611 AU

-3 -2 -1 0 1 2 3Time after periapse passage (days)

00000

00005

00010

00015

00020

00025

Pla

netS

tar

Flu

x R

atio

Figure 16 Light Curves for two different e = 075 model integrationsTeq = 1199 K a = 00385 AU (top) and Teq = 951 K a = 00611 AU(bottom) both with ω = 180 The solid line denotes transit while the dottedline denotes secondary eclipse

(A color version of this figure is available in the online journal)

relative to orbital position can help to constrain the circulationfurther

5 CONCLUSIONS

We present three-dimensional circulation models coupledwith a two-stream non-gray radiative transfer scheme fora number of theoretical eccentric hot Jupiters We haveshown that as in published models with zero eccentricity ourhigh-eccentricity circulation models are dominated by eastwardflow at photospheric levels which cause an eastward displace-ment of the hottest regions from the substellar point The rapidrotation rates associated with pseudo-synchronization at higheccentricity lead to a small Rossby deformation radius andin some cases multiple jets in the atmosphere Global-meantemperatures and dayndashnight temperature differences peak notat periapse but several hours afterward due to finite radiativetimescales in the planetrsquos atmosphere

Furthermore we show that equatorial superrotation is gener-ated and maintained by eddies formed by the strong dayndashnightheating contrast which induce a flux of momentum from mid-latitudes to the equator The eddy magnitudes and momentumfluxes peak just after periapse passage leading to variations inthe zonal-mean flow throughout the orbit

Figure 17 Globally averaged pressurendashtemperature profiles from 1-D radiative-convective radiative transfer models of HD 189733b with 11 frequency bins(red profiles) 30 frequency bins (green profiles) and 196 frequency bins (blackprofiles) From left to right the profiles vary in stellar flux from the nominalvalue as well as 2times 4times and 8times higher flux The 8times higher flux case has atemperature inversion

(A color version of this figure is available in the online journal)

Lastly we have shown that the spatial and temporal vari-ations of the wind and temperature structure as well as theorbital viewing geometry of the system with respect to Earthcan affect the time and amplitude of peak IR flux seen infull-orbit light curves Depending on the viewing geometry ofthe orbit relative to Earth we find that peaks in IR flux thateither lead or lag periapse are possible in all cases a combi-nation of temporal effects (temperatures changing over time)and geometric effects (hot spots rotating into or out of view)are important in controlling the timing and amplitude of theflux peaks In cases where the dayndashnight temperature contrastis large and the rotational period is short the light curve canalso exhibit ldquoringingrdquo in flux as the hottest region of the planetrotates in and out of view This ringing is non-periodic due tothe variation in stellar heating as a function of distance

This work was supported by NASA Origins and PlanetaryAtmospheres grants to APS TK also acknowledges supportfrom the Harriet P Jenkins Pre-Doctoral Fellowship Program(JPFP) Resources supporting this work were provided bythe NASA High-End Computing (HEC) Program through theNASA Advanced Supercomputing (NAS) Division at AmesResearch Center The authors thank the anonymous referee fortheir helpful comments and suggestions

APPENDIX

VALIDATION TESTS FOR UPDATEDRADIATIVE TRANSFER SCHEME

To validate our updated radiative transfer scheme using 11frequency bins we conducted the following tests

1 First we tested the sensitivity of the number of frequencybins in a one-dimensional radiative-convective model ofHD 189733b (a = 00313 AU e = 00 RP = 82396 times107 m g = 214 m sminus2) Figure 17 compares the globallyaveraged pressurendashtemperature profiles of models run with11 frequency bins 30 frequency bins and 196 frequencybins at (from left to right) 1times 2times 4times and 8times theaverage stellar flux of HD 189733b For each group of

17

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)30 opacity bins

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)

11 opacity bins

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

30 opacity bins

minus600

minus200minus200

200

200

600

6001000

1000

14001800

2200

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

11 opacity bins

minus600

minus200

minus200

200

200

600

600

10001000

1400

1800

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

30 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

11 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

Figure 18 Globally averaged pressurendashtemperature profiles (top row) zonal-mean zonal wind (middle row) and the windtemperature profiles at 30 mbar (bottomrow) for the 30-bin (left column) and 11-bin (right column) model integrations of HD 189733b Snapshots were output at 250 Earth days

(A color version of this figure is available in the online journal)

models the 11-bin profile agrees well with the 30- and196-bin cases from 1 μbar to 1 bar the temperature variesby only tens of K Below 10 bars the 30- and 11-binprofiles differ from the 196-bin cases by up to 50 Kdue to the loss of resolution at low wavenumbers Inthe case of the highest average flux the large amount of

heating leads to a temperature inversion that is not fullycaptured by the 11-bin model but can be improved witha better initial guess at the pressure-temperature (P-T)profile

2 After testing the new radiative transfer (RT) scheme forone-dimensional models we ran full three-dimensional

18

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

simulations of HD 189733b at its nominal average stellarflux using the SPARCMITgcm Figure 18 shows the glob-ally averaged P-T profiles (top row) zonal-mean zonal wind(middle row) and windtemperature profiles at 30 mbar(bottom row) for the 30-bin (left column) and 11-bin (rightcolumn) RT setups The globally averaged profiles arenearly identical with any temperature differences withina few K The zonal wind plots also show good agree-ment with similar peak speeds (gt2600 m sminus1) equatorialjet width (sim60) and jet level (lt100 bar) At 30 mbarthe 30- and 11-bin models share similar dayside-nightsidetemperature structure with a hot spot (1100 K) eastward ofthe substellar point and two colder regions on the nightsidein the mid-latitudes

Both tests show that the bulk circulation and temperaturestructure is retained in the transition from 30 to 11 frequencybins hence we proceed to use the SPARCMITgcm with thenew scheme for these and future model simulations

REFERENCES

Adcroft A Campin J-M Hill C amp Marshall J 2004 MWRv 132 2845Charbonneau D Knutson H A Barman T et al 2008 ApJ 686 1341Cho J Y-K Menou K Hansen B M S amp Seager S 2008 ApJ

675 817Cooper C S amp Showman A P 2005 ApJL 629 L45Cooper C S amp Showman A P 2006 ApJ 649 1048Cowan N B amp Agol E 2011 ApJ 726 82Deming D Harrington J Laughlin G et al 2007 ApJL 667 L199Dobbs-Dixon I Cumming A amp Lin D N C 2010 ApJ 710 1395Dobbs-Dixon I amp Lin D N C 2008 ApJ 673 513Fortney J J Cooper C S Showman A P Marley M S amp Freedman R S

2006 ApJ 652 746Fortney J J Lodders K Marley M S amp Freedman R S 2008 ApJ

678 1419Fortney J J Marley M S Lodders K Saumon D amp Freedman R

2005 ApJL 627 L69

Fu Q amp Liou K N 1992 JAtS 49 2140Goody R West R Chen L amp Crisp D 1989 JQSRT 42 539Greenberg R 2009 ApJL 698 L42Heng K 2012 ApJL 761 L1Heng K Frierson D M W amp Phillipps P J 2011 MNRAS 418 2669Hide R 1969 JAtS 26 841Holton J R 2004 An Introduction to Dynamic Meteorology (Waltham MA

Academic)Hut P 1981 AampA 99 126Irwin J Charbonneau D Nutzman P et al 2008 ApJ 681 636Ivanov P B amp Papaloizou J C B 2004 MNRAS 353 1161Knutson H A Charbonneau D Allen L E Burrows A amp Megeath S T

2008 ApJ 673 526Knutson H A Charbonneau D Allen L E et al 2007 Natur 447 183Knutson H A Charbonneau D Cowan N B et al 2009 ApJ 690 822Langton J amp Laughlin G 2008a ApJ 674 1106Langton J amp Laughlin G 2008b AampA 483 25Lewis N K Showman A P Fortney J J et al 2010 ApJ 720 344Lodders K 2003 ApJ 591 1220Marley M S amp McKay C P 1999 Icar 138 268Mayor M amp Queloz D 1995 Natur 378 355Menou K amp Rauscher E 2009 ApJ 700 887Mlawer E J Taubman S J Brown P D Iacono M J amp Clough S A

1997 JGR 102 16663Naef D Latham D W Mayor M et al 2001 AampA 375 L27Perna R Heng K amp Pont F 2012 ApJ 751 59Pont F Knutson H Gilliland R L Moutou C amp Charbonneau D

2008 MNRAS 385 109Rauscher E amp Menou K 2010 ApJ 714 1334Rauscher E amp Menou K 2012 ApJ 750 96Saumon D Marley M S Abel M Frommhold L amp Freedman R S

2012 ApJ 750 74Showman A P Cho J Y-K amp Menou K 2011 in Exoplanets ed S Seager

(Tucson AZ Univ Arizona Press) 471Showman A P Cooper C S Fortney J J amp Marley M S 2008 ApJ

682 559Showman A P Fortney J J Lian Y et al 2009 ApJ 699 564Showman A P amp Guillot T 2002 AampA 385 166Showman A P amp Polvani L M 2011 ApJ 738 71Swain M R Vasisht G amp Tinetti G 2008 Natur 452 329Thrastarson H T amp Cho J Y 2010 ApJ 716 144Wolszczan A 1994 Sci 264 538

19

  • 1 INTRODUCTION
  • 2 MODEL
    • 21 Updates to SPARCMITgcm
    • 22 Grid Simulations
      • 3 RESULTS
        • 31 Circulation Regime
        • 32 Effect of Eccentricity
        • 33 Effect of Varying Average Stellar Flux
        • 34 Global Average Temperature versus Eccentricity and F
        • 35 Synchronous versus Pseudo-synchronous Rotation
        • 36 Momentum Budget and Equatorial Superrotation
          • 4 OBSERVATIONAL IMPLICATIONS
            • 41 Observational Effects of Varying ω
            • 42 High-e Model Integrations at Varying F
              • 5 CONCLUSIONS
              • APPENDIX VALIDATION TESTS FOR UPDATED RADIATIVE TRANSFER SCHEME
              • REFERENCES
Page 13: THREE-DIMENSIONAL ATMOSPHERIC CIRCULATION OF HOT …showman/publications/kataria-etal-2013.pdf · planets’ atmospheric composition, temperature structure, and circulation. A fifth

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=1242deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=7227deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=minus11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=11988deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

minus50 0 50

10minus2

100

102

latitude (deg)

pre

ssu

re (

bar

)

f=18000deg

(m2 sminus2)minus1

minus08

minus06

minus04

minus02

0

02

04

06

08

1x 10

6

Figure 12 Eddy momentum flux uprimevprime cos φ throughout the orbit of a model where Teq = 951 K e = 05 and a = 00534 AU Positive values indicate a northwardflux of eastward eddy angular momentum while negative values indicate a southward flux of eastward eddy angular momentum The maximum momentum fluxoccurs hours after periapse passage (f = 72) The snapshots correspond to model outputs from 6103 to 6108 Earth days

(A color version of this figure is available in the online journal)

13

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

Figure 13 Differing orbital viewing geometries explored in this study ω = 180 ω = 270 and ω = 360

(A color version of this figure is available in the online journal)

pressures than that of the synchronous case this is most likelydue to a change in momentum budget

4 OBSERVATIONAL IMPLICATIONS

Observations of transiting exoplanets in circular orbits areknown to be shaped by the atmospheric circulation As wasfirst predicted for hot Jupiters by Showman amp Guillot (2002)and later observationally confirmed for hot Jupiter HD 189733bby Knutson et al (2007) an equatorial superrotating jet causesan eastward displacement of the hot spot from the substellarpoint under appropriate conditions Hence a light curve showsa peak in infrared flux minutes to hours before secondaryeclipse We expect such a situation for eccentric exoplanetsas well However observations of eccentric planets can becomplicated because the flux received by the planet variesthroughout its orbit Previous studies explored such situationsbut have not used three-dimensional models (eg Langton ampLaughlin 2008a 2008b Cowan amp Agol 2011) We can expandon these studies and the three-dimensional study by Lewis et al(2010) by using our SPARC model results

We choose to generate light curves following the proceduresdescribed in Fortney et al (2006) for each model integrationat three values of the argument of periastron (ω) the anglebetween the radius vector to the ascending node and the periapseof the orbit In particular we choose orientations where transitoccurs before periapse (at f = minus90ω = 180) transit occursat apoapse (ω = 270) and transit occurs after periapse (atf = +90ω = 360) (Figure 13) These values reflect thebroad range of viewing geometries of transiting exoplanets forexample HAT-P-2b (e = 052) HAT-P-17b (e = 035) andHD 97658b (e = 013) have ω near 180 WASP-8b (e = 031)HAT-P-31b (e = 025) and HD 80606b (e = 093) have ω near270 and GJ 436b (e = 015) XO-3b (e = 026) and HAT-P-11b (e = 020) have ω near 360 (see exoplaneteu for a full listof values) Note that when transit occurs the nightside is visibleto Earth while the dayside is visible at the time of secondaryeclipse

41 Observational Effects of Varying ω

Using the nominal case (e = 050 a = 00534 AU Teq =951 K) we illustrate the dependence of full-orbit light curves onthe value of ω Figure 14 shows three light curves of the same

model integration but at the differing values of ω described inthe previous section Hence the differences in light curves resultsolely from the differing viewing geometries The planetstarflux ratio is plotted as a function of Earth days from periapsein the warm Spitzer wavelength bands (36 and 45 μm) andJHK bands (126 165 and 220 μm) The solid and dashedlines correspond to transit and secondary eclipse respectivelyFor the light curve where ω = 180 (top panel) the peak ininfrared (IR) flux occurs approximately 6 hr after periapse andbefore secondary eclipse The timing of the peak is due toa combination of the radiative lag of the planet and the factthat the hottest regions rotate into an Earth-facing orientationhours after periapse passage The dayside which is most visiblenear secondary eclipse is hottest after periapse Additionallytemperatures reach peak values at times close to when thosehottest regions are facing Earth This helps explain the highamplitude of the light curve peak in the ω = 180 case relative tothe other two cases These effects can be visualized by plottingthe temperature near the photosphere along the equator in apolar projection (Figure 15) Snapshots throughout the orbit plotthe temperature in a polar projection as a function of pressureand orbital position at a latitudinal slice near the equatorthe solid and dashed lines denote the substellar longitude andEarth-facing longitude respectively Based on the temperaturesat the Earth-facing longitude it is apparent that the peak in IRflux should occur after periapse and before secondary eclipse(see f = 12 and 72) this is indeed the case

For the same model but instead with ω = 270 the peak inIR flux occurs 1 hr after eclipse and periapse In this case thedayside is visible near periapse and therefore we are seeing theplanet increase in temperature (and IR flux) during its closestapproach to the star However the hottest regions are displacedeastward of the substellar point and therefore the hottest regionsface Earth significantly before periapse passage (Figure 15 atf asymp 72) Thus by the time the peak temperatures are reachednear and after periapse the hottest regions are already rotatingout of view from Earth This explains the lower amplitude ofthe light curve peak as compared to the peak of the ω = 180case The IR peak decreases as the planet moves further awayfrom the star along our line of sight A second ramp-up in fluxoccurs approximately 1 day after periapseeclipse this is due tothe hottest regions rotating back into view as the planet movestoward apoapse

14

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015P

lane

tSta

r F

lux

Rat

io

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

Figure 14 Three light curves of the IR flux (expressed as a planetstar fluxratio) vs time throughout the full orbit for a single model integration whereTeq = 951 K e = 050 and a = 00534 AU The flux is plotted in warm Spitzerbandpasses 36 and 45 μm and JHK bands (126 165 220 μm respectively)These three cases only differ by their value of ω shown in Figure 13 The solidlines denote transit while the dotted lines denote secondary eclipse

(A color version of this figure is available in the online journal)

In the case where ω = 360 two distinct peaks are presentmdashasmaller peak that occurs after secondary eclipse and sim3ndash4 hrbefore periapse and a second higher peak that occurs aftertransit The first peak occurs as some of the dayside is visible asthe planet is heated approaching periapse The peak decreasesas the dayside rotates out of view For this geometry the hottestregions are facing away from Earth when they reach their peaktemperatures (Figure 15) The second peak occurs after transitas the hottest regions again rotate into view At this point in theorbit peak temperatures have already occurred and thereforethis peak is lower in amplitude than the other two light curvesHowever because the planet is hotter than it was after secondary

eclipse this second peak has a higher amplitude than the firstone

42 High-e Model Integrations at Varying 〈F 〉If an eccentric planet has a large enough dayndashnight tempera-

ture difference and a sufficiently rapid rotation rate its resultantlight curve can show a periodic rise and fall in flux through-out its orbit The light curves shown in Figure 16 are twoe = 075 model integrations that differ in 〈F 〉 the top panelshows a case where 〈F 〉 = 468183 W mminus2 (Teq = 1199 K)and a = 00385 AU while the bottom panel is a case where〈F 〉 = 185691 W mminus2 (Teq = 951 K) and a = 00611 AU Forboth cases light curves are shown for the ω = 180 geome-try (see Figure 13) The differences in 〈F 〉 lead to differencesin light curve amplitude in the two models the closer hotterplanet exhibits a higher flux amplitude Not surprisingly thetiming of peak IR flux for both models resembles the ω = 180case in Figure 14 The peak in each case occurs after periapseand before eclipse due to the combined effects of the ther-mal lag of the planet and geometry of the system Additionallyboth light curves exhibit a quasi-periodic rise and fall of IR fluxdays after periapse passage This phenomenon called ldquoringingrdquowas also seen in models by Langton amp Laughlin (2008a) andCowan amp Agol (2011) caused by the hottest point of the planetrotating in and out of view from Earth For ringing to occurthe temperature difference from dayside to nightside must belarge and it must survive for multiple planetary rotation periodsIn the models shown in Figure 16 this is aided by the rapidpseudo-synchronous rotation rate associated with an eccentric-ity of 075 (Table 2) For the top case the dayndashnight tempera-ture difference is over sim1300 K while the bottom case variesby sim750 K (Figure 7) The light curves from Figure 14 do notexhibit ringing because the dayndashnight temperature difference islow (sim300ndash400 K)

Because the radiative and dynamical timescales in the modelsby Cowan amp Agol (2011) are calculated by a simple scaling withtemperature the ringing seen in their simulated light curvesoccurs with a period equal to the planetrsquos assumed solid-bodyrotation period in the inertial frame of the winds In realityan eccentric exoplanet approaching apoapse would becomeincreasingly cold the winds would weaken and decrease andhence the ringing should be non-periodic Indeed for ourthree dimensional models the ringing in each case has varyingperiodsmdashthe top case has a ringing period ranging from 03 to04 days (Trot = 035 days) while the bottom panel has a periodranging 07ndash08 days (Trot = 07 days)

Despite these differences this study and the studies byLangton amp Laughlin (2008a) and Cowan amp Agol (2011) illus-trate the influence of atmospheric dynamics on the observationsof eccentric exoplanets The light curve features described herecould be seen in future follow-up observations of eccentric ex-oplanets and observers can use the conclusions drawn aboveto relate the observations to the planetrsquos atmospheric structureFor example if ringing is present in an IR light curve the planetmust have a strong dayndashnight temperature asymmetry and hencewe would expect the planet to have strong superrotation at theequator (Showman amp Polvani 2011) This is probably the casefor highly eccentric (e gt 05) short-period exoplanets such asHD 17156b and HD 80606b If ringing were observed it wouldalso allow an estimate of the sum of the rotation speed and thesuperrotation speed thereby placing constraints on both windspeeds and rotation rates of eccentric hot Jupiters The timing ofpeak IR flux relative to transit and secondary eclipse and also

15

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

f=000deg

(K)800

850

900

950

1000

f=minus1242deg

(K)800

850

900

950

1000f=1242deg

(K)800

850

900

950

1000

f=minus7227deg

(K)800

850

900

950

1000f=7227deg

(K)800

850

900

950

1000

f=minus11988deg

(K)800

850

900

950

1000f=11988deg

(K)800

850

900

950

1000

f=minus18000deg

(K)800

850

900

950

1000

Figure 15 Plots of temperature vs longitude and pressure in equatorial slices throughout one full orbit for our nominal model with Teq = 951 K e = 050 anda = 00534 AU Each annulus shows temperature vs longitude and log-pressure for a snapshot at a particular orbital position (denoted by the value of f) as viewedfrom the planetrsquos north pole Pressures between 30 and 43 mbar are displayed in each annulus The different annuli mark the time evolution throughout one completeorbit starting from apoapse for the top annulus proceeding to periapse and then back to apoapse (follow arrows moving counterclockwise around the plot) Thesesnapshots correspond to model outputs from 6103 to 6108 Earth days The solid line in each snapshot indicates the longitude of the substellar point while the dashedline denotes the Earth-facing longitude for the particular case of ω = 180 For the cases of ω = 270 and 360 degrees exactly the same plot is valid except that thesub-Earth longitude would be aiming upward for ω = 270 and right for ω = 360

(A color version of this figure is available in the online journal)

16

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

e = 075 a = 00385 AU

-15 -10 -05 00 05 10 15Time after periapse passage (days)

0000

0001

0002

0003

0004

Pla

netS

tar

Flu

x R

atio

e = 075 a = 00611 AU

-3 -2 -1 0 1 2 3Time after periapse passage (days)

00000

00005

00010

00015

00020

00025

Pla

netS

tar

Flu

x R

atio

Figure 16 Light Curves for two different e = 075 model integrationsTeq = 1199 K a = 00385 AU (top) and Teq = 951 K a = 00611 AU(bottom) both with ω = 180 The solid line denotes transit while the dottedline denotes secondary eclipse

(A color version of this figure is available in the online journal)

relative to orbital position can help to constrain the circulationfurther

5 CONCLUSIONS

We present three-dimensional circulation models coupledwith a two-stream non-gray radiative transfer scheme fora number of theoretical eccentric hot Jupiters We haveshown that as in published models with zero eccentricity ourhigh-eccentricity circulation models are dominated by eastwardflow at photospheric levels which cause an eastward displace-ment of the hottest regions from the substellar point The rapidrotation rates associated with pseudo-synchronization at higheccentricity lead to a small Rossby deformation radius andin some cases multiple jets in the atmosphere Global-meantemperatures and dayndashnight temperature differences peak notat periapse but several hours afterward due to finite radiativetimescales in the planetrsquos atmosphere

Furthermore we show that equatorial superrotation is gener-ated and maintained by eddies formed by the strong dayndashnightheating contrast which induce a flux of momentum from mid-latitudes to the equator The eddy magnitudes and momentumfluxes peak just after periapse passage leading to variations inthe zonal-mean flow throughout the orbit

Figure 17 Globally averaged pressurendashtemperature profiles from 1-D radiative-convective radiative transfer models of HD 189733b with 11 frequency bins(red profiles) 30 frequency bins (green profiles) and 196 frequency bins (blackprofiles) From left to right the profiles vary in stellar flux from the nominalvalue as well as 2times 4times and 8times higher flux The 8times higher flux case has atemperature inversion

(A color version of this figure is available in the online journal)

Lastly we have shown that the spatial and temporal vari-ations of the wind and temperature structure as well as theorbital viewing geometry of the system with respect to Earthcan affect the time and amplitude of peak IR flux seen infull-orbit light curves Depending on the viewing geometry ofthe orbit relative to Earth we find that peaks in IR flux thateither lead or lag periapse are possible in all cases a combi-nation of temporal effects (temperatures changing over time)and geometric effects (hot spots rotating into or out of view)are important in controlling the timing and amplitude of theflux peaks In cases where the dayndashnight temperature contrastis large and the rotational period is short the light curve canalso exhibit ldquoringingrdquo in flux as the hottest region of the planetrotates in and out of view This ringing is non-periodic due tothe variation in stellar heating as a function of distance

This work was supported by NASA Origins and PlanetaryAtmospheres grants to APS TK also acknowledges supportfrom the Harriet P Jenkins Pre-Doctoral Fellowship Program(JPFP) Resources supporting this work were provided bythe NASA High-End Computing (HEC) Program through theNASA Advanced Supercomputing (NAS) Division at AmesResearch Center The authors thank the anonymous referee fortheir helpful comments and suggestions

APPENDIX

VALIDATION TESTS FOR UPDATEDRADIATIVE TRANSFER SCHEME

To validate our updated radiative transfer scheme using 11frequency bins we conducted the following tests

1 First we tested the sensitivity of the number of frequencybins in a one-dimensional radiative-convective model ofHD 189733b (a = 00313 AU e = 00 RP = 82396 times107 m g = 214 m sminus2) Figure 17 compares the globallyaveraged pressurendashtemperature profiles of models run with11 frequency bins 30 frequency bins and 196 frequencybins at (from left to right) 1times 2times 4times and 8times theaverage stellar flux of HD 189733b For each group of

17

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)30 opacity bins

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)

11 opacity bins

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

30 opacity bins

minus600

minus200minus200

200

200

600

6001000

1000

14001800

2200

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

11 opacity bins

minus600

minus200

minus200

200

200

600

600

10001000

1400

1800

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

30 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

11 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

Figure 18 Globally averaged pressurendashtemperature profiles (top row) zonal-mean zonal wind (middle row) and the windtemperature profiles at 30 mbar (bottomrow) for the 30-bin (left column) and 11-bin (right column) model integrations of HD 189733b Snapshots were output at 250 Earth days

(A color version of this figure is available in the online journal)

models the 11-bin profile agrees well with the 30- and196-bin cases from 1 μbar to 1 bar the temperature variesby only tens of K Below 10 bars the 30- and 11-binprofiles differ from the 196-bin cases by up to 50 Kdue to the loss of resolution at low wavenumbers Inthe case of the highest average flux the large amount of

heating leads to a temperature inversion that is not fullycaptured by the 11-bin model but can be improved witha better initial guess at the pressure-temperature (P-T)profile

2 After testing the new radiative transfer (RT) scheme forone-dimensional models we ran full three-dimensional

18

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

simulations of HD 189733b at its nominal average stellarflux using the SPARCMITgcm Figure 18 shows the glob-ally averaged P-T profiles (top row) zonal-mean zonal wind(middle row) and windtemperature profiles at 30 mbar(bottom row) for the 30-bin (left column) and 11-bin (rightcolumn) RT setups The globally averaged profiles arenearly identical with any temperature differences withina few K The zonal wind plots also show good agree-ment with similar peak speeds (gt2600 m sminus1) equatorialjet width (sim60) and jet level (lt100 bar) At 30 mbarthe 30- and 11-bin models share similar dayside-nightsidetemperature structure with a hot spot (1100 K) eastward ofthe substellar point and two colder regions on the nightsidein the mid-latitudes

Both tests show that the bulk circulation and temperaturestructure is retained in the transition from 30 to 11 frequencybins hence we proceed to use the SPARCMITgcm with thenew scheme for these and future model simulations

REFERENCES

Adcroft A Campin J-M Hill C amp Marshall J 2004 MWRv 132 2845Charbonneau D Knutson H A Barman T et al 2008 ApJ 686 1341Cho J Y-K Menou K Hansen B M S amp Seager S 2008 ApJ

675 817Cooper C S amp Showman A P 2005 ApJL 629 L45Cooper C S amp Showman A P 2006 ApJ 649 1048Cowan N B amp Agol E 2011 ApJ 726 82Deming D Harrington J Laughlin G et al 2007 ApJL 667 L199Dobbs-Dixon I Cumming A amp Lin D N C 2010 ApJ 710 1395Dobbs-Dixon I amp Lin D N C 2008 ApJ 673 513Fortney J J Cooper C S Showman A P Marley M S amp Freedman R S

2006 ApJ 652 746Fortney J J Lodders K Marley M S amp Freedman R S 2008 ApJ

678 1419Fortney J J Marley M S Lodders K Saumon D amp Freedman R

2005 ApJL 627 L69

Fu Q amp Liou K N 1992 JAtS 49 2140Goody R West R Chen L amp Crisp D 1989 JQSRT 42 539Greenberg R 2009 ApJL 698 L42Heng K 2012 ApJL 761 L1Heng K Frierson D M W amp Phillipps P J 2011 MNRAS 418 2669Hide R 1969 JAtS 26 841Holton J R 2004 An Introduction to Dynamic Meteorology (Waltham MA

Academic)Hut P 1981 AampA 99 126Irwin J Charbonneau D Nutzman P et al 2008 ApJ 681 636Ivanov P B amp Papaloizou J C B 2004 MNRAS 353 1161Knutson H A Charbonneau D Allen L E Burrows A amp Megeath S T

2008 ApJ 673 526Knutson H A Charbonneau D Allen L E et al 2007 Natur 447 183Knutson H A Charbonneau D Cowan N B et al 2009 ApJ 690 822Langton J amp Laughlin G 2008a ApJ 674 1106Langton J amp Laughlin G 2008b AampA 483 25Lewis N K Showman A P Fortney J J et al 2010 ApJ 720 344Lodders K 2003 ApJ 591 1220Marley M S amp McKay C P 1999 Icar 138 268Mayor M amp Queloz D 1995 Natur 378 355Menou K amp Rauscher E 2009 ApJ 700 887Mlawer E J Taubman S J Brown P D Iacono M J amp Clough S A

1997 JGR 102 16663Naef D Latham D W Mayor M et al 2001 AampA 375 L27Perna R Heng K amp Pont F 2012 ApJ 751 59Pont F Knutson H Gilliland R L Moutou C amp Charbonneau D

2008 MNRAS 385 109Rauscher E amp Menou K 2010 ApJ 714 1334Rauscher E amp Menou K 2012 ApJ 750 96Saumon D Marley M S Abel M Frommhold L amp Freedman R S

2012 ApJ 750 74Showman A P Cho J Y-K amp Menou K 2011 in Exoplanets ed S Seager

(Tucson AZ Univ Arizona Press) 471Showman A P Cooper C S Fortney J J amp Marley M S 2008 ApJ

682 559Showman A P Fortney J J Lian Y et al 2009 ApJ 699 564Showman A P amp Guillot T 2002 AampA 385 166Showman A P amp Polvani L M 2011 ApJ 738 71Swain M R Vasisht G amp Tinetti G 2008 Natur 452 329Thrastarson H T amp Cho J Y 2010 ApJ 716 144Wolszczan A 1994 Sci 264 538

19

  • 1 INTRODUCTION
  • 2 MODEL
    • 21 Updates to SPARCMITgcm
    • 22 Grid Simulations
      • 3 RESULTS
        • 31 Circulation Regime
        • 32 Effect of Eccentricity
        • 33 Effect of Varying Average Stellar Flux
        • 34 Global Average Temperature versus Eccentricity and F
        • 35 Synchronous versus Pseudo-synchronous Rotation
        • 36 Momentum Budget and Equatorial Superrotation
          • 4 OBSERVATIONAL IMPLICATIONS
            • 41 Observational Effects of Varying ω
            • 42 High-e Model Integrations at Varying F
              • 5 CONCLUSIONS
              • APPENDIX VALIDATION TESTS FOR UPDATED RADIATIVE TRANSFER SCHEME
              • REFERENCES
Page 14: THREE-DIMENSIONAL ATMOSPHERIC CIRCULATION OF HOT …showman/publications/kataria-etal-2013.pdf · planets’ atmospheric composition, temperature structure, and circulation. A fifth

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

Figure 13 Differing orbital viewing geometries explored in this study ω = 180 ω = 270 and ω = 360

(A color version of this figure is available in the online journal)

pressures than that of the synchronous case this is most likelydue to a change in momentum budget

4 OBSERVATIONAL IMPLICATIONS

Observations of transiting exoplanets in circular orbits areknown to be shaped by the atmospheric circulation As wasfirst predicted for hot Jupiters by Showman amp Guillot (2002)and later observationally confirmed for hot Jupiter HD 189733bby Knutson et al (2007) an equatorial superrotating jet causesan eastward displacement of the hot spot from the substellarpoint under appropriate conditions Hence a light curve showsa peak in infrared flux minutes to hours before secondaryeclipse We expect such a situation for eccentric exoplanetsas well However observations of eccentric planets can becomplicated because the flux received by the planet variesthroughout its orbit Previous studies explored such situationsbut have not used three-dimensional models (eg Langton ampLaughlin 2008a 2008b Cowan amp Agol 2011) We can expandon these studies and the three-dimensional study by Lewis et al(2010) by using our SPARC model results

We choose to generate light curves following the proceduresdescribed in Fortney et al (2006) for each model integrationat three values of the argument of periastron (ω) the anglebetween the radius vector to the ascending node and the periapseof the orbit In particular we choose orientations where transitoccurs before periapse (at f = minus90ω = 180) transit occursat apoapse (ω = 270) and transit occurs after periapse (atf = +90ω = 360) (Figure 13) These values reflect thebroad range of viewing geometries of transiting exoplanets forexample HAT-P-2b (e = 052) HAT-P-17b (e = 035) andHD 97658b (e = 013) have ω near 180 WASP-8b (e = 031)HAT-P-31b (e = 025) and HD 80606b (e = 093) have ω near270 and GJ 436b (e = 015) XO-3b (e = 026) and HAT-P-11b (e = 020) have ω near 360 (see exoplaneteu for a full listof values) Note that when transit occurs the nightside is visibleto Earth while the dayside is visible at the time of secondaryeclipse

41 Observational Effects of Varying ω

Using the nominal case (e = 050 a = 00534 AU Teq =951 K) we illustrate the dependence of full-orbit light curves onthe value of ω Figure 14 shows three light curves of the same

model integration but at the differing values of ω described inthe previous section Hence the differences in light curves resultsolely from the differing viewing geometries The planetstarflux ratio is plotted as a function of Earth days from periapsein the warm Spitzer wavelength bands (36 and 45 μm) andJHK bands (126 165 and 220 μm) The solid and dashedlines correspond to transit and secondary eclipse respectivelyFor the light curve where ω = 180 (top panel) the peak ininfrared (IR) flux occurs approximately 6 hr after periapse andbefore secondary eclipse The timing of the peak is due toa combination of the radiative lag of the planet and the factthat the hottest regions rotate into an Earth-facing orientationhours after periapse passage The dayside which is most visiblenear secondary eclipse is hottest after periapse Additionallytemperatures reach peak values at times close to when thosehottest regions are facing Earth This helps explain the highamplitude of the light curve peak in the ω = 180 case relative tothe other two cases These effects can be visualized by plottingthe temperature near the photosphere along the equator in apolar projection (Figure 15) Snapshots throughout the orbit plotthe temperature in a polar projection as a function of pressureand orbital position at a latitudinal slice near the equatorthe solid and dashed lines denote the substellar longitude andEarth-facing longitude respectively Based on the temperaturesat the Earth-facing longitude it is apparent that the peak in IRflux should occur after periapse and before secondary eclipse(see f = 12 and 72) this is indeed the case

For the same model but instead with ω = 270 the peak inIR flux occurs 1 hr after eclipse and periapse In this case thedayside is visible near periapse and therefore we are seeing theplanet increase in temperature (and IR flux) during its closestapproach to the star However the hottest regions are displacedeastward of the substellar point and therefore the hottest regionsface Earth significantly before periapse passage (Figure 15 atf asymp 72) Thus by the time the peak temperatures are reachednear and after periapse the hottest regions are already rotatingout of view from Earth This explains the lower amplitude ofthe light curve peak as compared to the peak of the ω = 180case The IR peak decreases as the planet moves further awayfrom the star along our line of sight A second ramp-up in fluxoccurs approximately 1 day after periapseeclipse this is due tothe hottest regions rotating back into view as the planet movestoward apoapse

14

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015P

lane

tSta

r F

lux

Rat

io

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

Figure 14 Three light curves of the IR flux (expressed as a planetstar fluxratio) vs time throughout the full orbit for a single model integration whereTeq = 951 K e = 050 and a = 00534 AU The flux is plotted in warm Spitzerbandpasses 36 and 45 μm and JHK bands (126 165 220 μm respectively)These three cases only differ by their value of ω shown in Figure 13 The solidlines denote transit while the dotted lines denote secondary eclipse

(A color version of this figure is available in the online journal)

In the case where ω = 360 two distinct peaks are presentmdashasmaller peak that occurs after secondary eclipse and sim3ndash4 hrbefore periapse and a second higher peak that occurs aftertransit The first peak occurs as some of the dayside is visible asthe planet is heated approaching periapse The peak decreasesas the dayside rotates out of view For this geometry the hottestregions are facing away from Earth when they reach their peaktemperatures (Figure 15) The second peak occurs after transitas the hottest regions again rotate into view At this point in theorbit peak temperatures have already occurred and thereforethis peak is lower in amplitude than the other two light curvesHowever because the planet is hotter than it was after secondary

eclipse this second peak has a higher amplitude than the firstone

42 High-e Model Integrations at Varying 〈F 〉If an eccentric planet has a large enough dayndashnight tempera-

ture difference and a sufficiently rapid rotation rate its resultantlight curve can show a periodic rise and fall in flux through-out its orbit The light curves shown in Figure 16 are twoe = 075 model integrations that differ in 〈F 〉 the top panelshows a case where 〈F 〉 = 468183 W mminus2 (Teq = 1199 K)and a = 00385 AU while the bottom panel is a case where〈F 〉 = 185691 W mminus2 (Teq = 951 K) and a = 00611 AU Forboth cases light curves are shown for the ω = 180 geome-try (see Figure 13) The differences in 〈F 〉 lead to differencesin light curve amplitude in the two models the closer hotterplanet exhibits a higher flux amplitude Not surprisingly thetiming of peak IR flux for both models resembles the ω = 180case in Figure 14 The peak in each case occurs after periapseand before eclipse due to the combined effects of the ther-mal lag of the planet and geometry of the system Additionallyboth light curves exhibit a quasi-periodic rise and fall of IR fluxdays after periapse passage This phenomenon called ldquoringingrdquowas also seen in models by Langton amp Laughlin (2008a) andCowan amp Agol (2011) caused by the hottest point of the planetrotating in and out of view from Earth For ringing to occurthe temperature difference from dayside to nightside must belarge and it must survive for multiple planetary rotation periodsIn the models shown in Figure 16 this is aided by the rapidpseudo-synchronous rotation rate associated with an eccentric-ity of 075 (Table 2) For the top case the dayndashnight tempera-ture difference is over sim1300 K while the bottom case variesby sim750 K (Figure 7) The light curves from Figure 14 do notexhibit ringing because the dayndashnight temperature difference islow (sim300ndash400 K)

Because the radiative and dynamical timescales in the modelsby Cowan amp Agol (2011) are calculated by a simple scaling withtemperature the ringing seen in their simulated light curvesoccurs with a period equal to the planetrsquos assumed solid-bodyrotation period in the inertial frame of the winds In realityan eccentric exoplanet approaching apoapse would becomeincreasingly cold the winds would weaken and decrease andhence the ringing should be non-periodic Indeed for ourthree dimensional models the ringing in each case has varyingperiodsmdashthe top case has a ringing period ranging from 03 to04 days (Trot = 035 days) while the bottom panel has a periodranging 07ndash08 days (Trot = 07 days)

Despite these differences this study and the studies byLangton amp Laughlin (2008a) and Cowan amp Agol (2011) illus-trate the influence of atmospheric dynamics on the observationsof eccentric exoplanets The light curve features described herecould be seen in future follow-up observations of eccentric ex-oplanets and observers can use the conclusions drawn aboveto relate the observations to the planetrsquos atmospheric structureFor example if ringing is present in an IR light curve the planetmust have a strong dayndashnight temperature asymmetry and hencewe would expect the planet to have strong superrotation at theequator (Showman amp Polvani 2011) This is probably the casefor highly eccentric (e gt 05) short-period exoplanets such asHD 17156b and HD 80606b If ringing were observed it wouldalso allow an estimate of the sum of the rotation speed and thesuperrotation speed thereby placing constraints on both windspeeds and rotation rates of eccentric hot Jupiters The timing ofpeak IR flux relative to transit and secondary eclipse and also

15

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

f=000deg

(K)800

850

900

950

1000

f=minus1242deg

(K)800

850

900

950

1000f=1242deg

(K)800

850

900

950

1000

f=minus7227deg

(K)800

850

900

950

1000f=7227deg

(K)800

850

900

950

1000

f=minus11988deg

(K)800

850

900

950

1000f=11988deg

(K)800

850

900

950

1000

f=minus18000deg

(K)800

850

900

950

1000

Figure 15 Plots of temperature vs longitude and pressure in equatorial slices throughout one full orbit for our nominal model with Teq = 951 K e = 050 anda = 00534 AU Each annulus shows temperature vs longitude and log-pressure for a snapshot at a particular orbital position (denoted by the value of f) as viewedfrom the planetrsquos north pole Pressures between 30 and 43 mbar are displayed in each annulus The different annuli mark the time evolution throughout one completeorbit starting from apoapse for the top annulus proceeding to periapse and then back to apoapse (follow arrows moving counterclockwise around the plot) Thesesnapshots correspond to model outputs from 6103 to 6108 Earth days The solid line in each snapshot indicates the longitude of the substellar point while the dashedline denotes the Earth-facing longitude for the particular case of ω = 180 For the cases of ω = 270 and 360 degrees exactly the same plot is valid except that thesub-Earth longitude would be aiming upward for ω = 270 and right for ω = 360

(A color version of this figure is available in the online journal)

16

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

e = 075 a = 00385 AU

-15 -10 -05 00 05 10 15Time after periapse passage (days)

0000

0001

0002

0003

0004

Pla

netS

tar

Flu

x R

atio

e = 075 a = 00611 AU

-3 -2 -1 0 1 2 3Time after periapse passage (days)

00000

00005

00010

00015

00020

00025

Pla

netS

tar

Flu

x R

atio

Figure 16 Light Curves for two different e = 075 model integrationsTeq = 1199 K a = 00385 AU (top) and Teq = 951 K a = 00611 AU(bottom) both with ω = 180 The solid line denotes transit while the dottedline denotes secondary eclipse

(A color version of this figure is available in the online journal)

relative to orbital position can help to constrain the circulationfurther

5 CONCLUSIONS

We present three-dimensional circulation models coupledwith a two-stream non-gray radiative transfer scheme fora number of theoretical eccentric hot Jupiters We haveshown that as in published models with zero eccentricity ourhigh-eccentricity circulation models are dominated by eastwardflow at photospheric levels which cause an eastward displace-ment of the hottest regions from the substellar point The rapidrotation rates associated with pseudo-synchronization at higheccentricity lead to a small Rossby deformation radius andin some cases multiple jets in the atmosphere Global-meantemperatures and dayndashnight temperature differences peak notat periapse but several hours afterward due to finite radiativetimescales in the planetrsquos atmosphere

Furthermore we show that equatorial superrotation is gener-ated and maintained by eddies formed by the strong dayndashnightheating contrast which induce a flux of momentum from mid-latitudes to the equator The eddy magnitudes and momentumfluxes peak just after periapse passage leading to variations inthe zonal-mean flow throughout the orbit

Figure 17 Globally averaged pressurendashtemperature profiles from 1-D radiative-convective radiative transfer models of HD 189733b with 11 frequency bins(red profiles) 30 frequency bins (green profiles) and 196 frequency bins (blackprofiles) From left to right the profiles vary in stellar flux from the nominalvalue as well as 2times 4times and 8times higher flux The 8times higher flux case has atemperature inversion

(A color version of this figure is available in the online journal)

Lastly we have shown that the spatial and temporal vari-ations of the wind and temperature structure as well as theorbital viewing geometry of the system with respect to Earthcan affect the time and amplitude of peak IR flux seen infull-orbit light curves Depending on the viewing geometry ofthe orbit relative to Earth we find that peaks in IR flux thateither lead or lag periapse are possible in all cases a combi-nation of temporal effects (temperatures changing over time)and geometric effects (hot spots rotating into or out of view)are important in controlling the timing and amplitude of theflux peaks In cases where the dayndashnight temperature contrastis large and the rotational period is short the light curve canalso exhibit ldquoringingrdquo in flux as the hottest region of the planetrotates in and out of view This ringing is non-periodic due tothe variation in stellar heating as a function of distance

This work was supported by NASA Origins and PlanetaryAtmospheres grants to APS TK also acknowledges supportfrom the Harriet P Jenkins Pre-Doctoral Fellowship Program(JPFP) Resources supporting this work were provided bythe NASA High-End Computing (HEC) Program through theNASA Advanced Supercomputing (NAS) Division at AmesResearch Center The authors thank the anonymous referee fortheir helpful comments and suggestions

APPENDIX

VALIDATION TESTS FOR UPDATEDRADIATIVE TRANSFER SCHEME

To validate our updated radiative transfer scheme using 11frequency bins we conducted the following tests

1 First we tested the sensitivity of the number of frequencybins in a one-dimensional radiative-convective model ofHD 189733b (a = 00313 AU e = 00 RP = 82396 times107 m g = 214 m sminus2) Figure 17 compares the globallyaveraged pressurendashtemperature profiles of models run with11 frequency bins 30 frequency bins and 196 frequencybins at (from left to right) 1times 2times 4times and 8times theaverage stellar flux of HD 189733b For each group of

17

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)30 opacity bins

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)

11 opacity bins

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

30 opacity bins

minus600

minus200minus200

200

200

600

6001000

1000

14001800

2200

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

11 opacity bins

minus600

minus200

minus200

200

200

600

600

10001000

1400

1800

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

30 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

11 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

Figure 18 Globally averaged pressurendashtemperature profiles (top row) zonal-mean zonal wind (middle row) and the windtemperature profiles at 30 mbar (bottomrow) for the 30-bin (left column) and 11-bin (right column) model integrations of HD 189733b Snapshots were output at 250 Earth days

(A color version of this figure is available in the online journal)

models the 11-bin profile agrees well with the 30- and196-bin cases from 1 μbar to 1 bar the temperature variesby only tens of K Below 10 bars the 30- and 11-binprofiles differ from the 196-bin cases by up to 50 Kdue to the loss of resolution at low wavenumbers Inthe case of the highest average flux the large amount of

heating leads to a temperature inversion that is not fullycaptured by the 11-bin model but can be improved witha better initial guess at the pressure-temperature (P-T)profile

2 After testing the new radiative transfer (RT) scheme forone-dimensional models we ran full three-dimensional

18

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

simulations of HD 189733b at its nominal average stellarflux using the SPARCMITgcm Figure 18 shows the glob-ally averaged P-T profiles (top row) zonal-mean zonal wind(middle row) and windtemperature profiles at 30 mbar(bottom row) for the 30-bin (left column) and 11-bin (rightcolumn) RT setups The globally averaged profiles arenearly identical with any temperature differences withina few K The zonal wind plots also show good agree-ment with similar peak speeds (gt2600 m sminus1) equatorialjet width (sim60) and jet level (lt100 bar) At 30 mbarthe 30- and 11-bin models share similar dayside-nightsidetemperature structure with a hot spot (1100 K) eastward ofthe substellar point and two colder regions on the nightsidein the mid-latitudes

Both tests show that the bulk circulation and temperaturestructure is retained in the transition from 30 to 11 frequencybins hence we proceed to use the SPARCMITgcm with thenew scheme for these and future model simulations

REFERENCES

Adcroft A Campin J-M Hill C amp Marshall J 2004 MWRv 132 2845Charbonneau D Knutson H A Barman T et al 2008 ApJ 686 1341Cho J Y-K Menou K Hansen B M S amp Seager S 2008 ApJ

675 817Cooper C S amp Showman A P 2005 ApJL 629 L45Cooper C S amp Showman A P 2006 ApJ 649 1048Cowan N B amp Agol E 2011 ApJ 726 82Deming D Harrington J Laughlin G et al 2007 ApJL 667 L199Dobbs-Dixon I Cumming A amp Lin D N C 2010 ApJ 710 1395Dobbs-Dixon I amp Lin D N C 2008 ApJ 673 513Fortney J J Cooper C S Showman A P Marley M S amp Freedman R S

2006 ApJ 652 746Fortney J J Lodders K Marley M S amp Freedman R S 2008 ApJ

678 1419Fortney J J Marley M S Lodders K Saumon D amp Freedman R

2005 ApJL 627 L69

Fu Q amp Liou K N 1992 JAtS 49 2140Goody R West R Chen L amp Crisp D 1989 JQSRT 42 539Greenberg R 2009 ApJL 698 L42Heng K 2012 ApJL 761 L1Heng K Frierson D M W amp Phillipps P J 2011 MNRAS 418 2669Hide R 1969 JAtS 26 841Holton J R 2004 An Introduction to Dynamic Meteorology (Waltham MA

Academic)Hut P 1981 AampA 99 126Irwin J Charbonneau D Nutzman P et al 2008 ApJ 681 636Ivanov P B amp Papaloizou J C B 2004 MNRAS 353 1161Knutson H A Charbonneau D Allen L E Burrows A amp Megeath S T

2008 ApJ 673 526Knutson H A Charbonneau D Allen L E et al 2007 Natur 447 183Knutson H A Charbonneau D Cowan N B et al 2009 ApJ 690 822Langton J amp Laughlin G 2008a ApJ 674 1106Langton J amp Laughlin G 2008b AampA 483 25Lewis N K Showman A P Fortney J J et al 2010 ApJ 720 344Lodders K 2003 ApJ 591 1220Marley M S amp McKay C P 1999 Icar 138 268Mayor M amp Queloz D 1995 Natur 378 355Menou K amp Rauscher E 2009 ApJ 700 887Mlawer E J Taubman S J Brown P D Iacono M J amp Clough S A

1997 JGR 102 16663Naef D Latham D W Mayor M et al 2001 AampA 375 L27Perna R Heng K amp Pont F 2012 ApJ 751 59Pont F Knutson H Gilliland R L Moutou C amp Charbonneau D

2008 MNRAS 385 109Rauscher E amp Menou K 2010 ApJ 714 1334Rauscher E amp Menou K 2012 ApJ 750 96Saumon D Marley M S Abel M Frommhold L amp Freedman R S

2012 ApJ 750 74Showman A P Cho J Y-K amp Menou K 2011 in Exoplanets ed S Seager

(Tucson AZ Univ Arizona Press) 471Showman A P Cooper C S Fortney J J amp Marley M S 2008 ApJ

682 559Showman A P Fortney J J Lian Y et al 2009 ApJ 699 564Showman A P amp Guillot T 2002 AampA 385 166Showman A P amp Polvani L M 2011 ApJ 738 71Swain M R Vasisht G amp Tinetti G 2008 Natur 452 329Thrastarson H T amp Cho J Y 2010 ApJ 716 144Wolszczan A 1994 Sci 264 538

19

  • 1 INTRODUCTION
  • 2 MODEL
    • 21 Updates to SPARCMITgcm
    • 22 Grid Simulations
      • 3 RESULTS
        • 31 Circulation Regime
        • 32 Effect of Eccentricity
        • 33 Effect of Varying Average Stellar Flux
        • 34 Global Average Temperature versus Eccentricity and F
        • 35 Synchronous versus Pseudo-synchronous Rotation
        • 36 Momentum Budget and Equatorial Superrotation
          • 4 OBSERVATIONAL IMPLICATIONS
            • 41 Observational Effects of Varying ω
            • 42 High-e Model Integrations at Varying F
              • 5 CONCLUSIONS
              • APPENDIX VALIDATION TESTS FOR UPDATED RADIATIVE TRANSFER SCHEME
              • REFERENCES
Page 15: THREE-DIMENSIONAL ATMOSPHERIC CIRCULATION OF HOT …showman/publications/kataria-etal-2013.pdf · planets’ atmospheric composition, temperature structure, and circulation. A fifth

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015P

lane

tSta

r F

lux

Rat

io

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

-2 -1 0 1 2Time after periapse passage (days)

00000

00005

00010

00015

Pla

netS

tar

Flu

x R

atio

Figure 14 Three light curves of the IR flux (expressed as a planetstar fluxratio) vs time throughout the full orbit for a single model integration whereTeq = 951 K e = 050 and a = 00534 AU The flux is plotted in warm Spitzerbandpasses 36 and 45 μm and JHK bands (126 165 220 μm respectively)These three cases only differ by their value of ω shown in Figure 13 The solidlines denote transit while the dotted lines denote secondary eclipse

(A color version of this figure is available in the online journal)

In the case where ω = 360 two distinct peaks are presentmdashasmaller peak that occurs after secondary eclipse and sim3ndash4 hrbefore periapse and a second higher peak that occurs aftertransit The first peak occurs as some of the dayside is visible asthe planet is heated approaching periapse The peak decreasesas the dayside rotates out of view For this geometry the hottestregions are facing away from Earth when they reach their peaktemperatures (Figure 15) The second peak occurs after transitas the hottest regions again rotate into view At this point in theorbit peak temperatures have already occurred and thereforethis peak is lower in amplitude than the other two light curvesHowever because the planet is hotter than it was after secondary

eclipse this second peak has a higher amplitude than the firstone

42 High-e Model Integrations at Varying 〈F 〉If an eccentric planet has a large enough dayndashnight tempera-

ture difference and a sufficiently rapid rotation rate its resultantlight curve can show a periodic rise and fall in flux through-out its orbit The light curves shown in Figure 16 are twoe = 075 model integrations that differ in 〈F 〉 the top panelshows a case where 〈F 〉 = 468183 W mminus2 (Teq = 1199 K)and a = 00385 AU while the bottom panel is a case where〈F 〉 = 185691 W mminus2 (Teq = 951 K) and a = 00611 AU Forboth cases light curves are shown for the ω = 180 geome-try (see Figure 13) The differences in 〈F 〉 lead to differencesin light curve amplitude in the two models the closer hotterplanet exhibits a higher flux amplitude Not surprisingly thetiming of peak IR flux for both models resembles the ω = 180case in Figure 14 The peak in each case occurs after periapseand before eclipse due to the combined effects of the ther-mal lag of the planet and geometry of the system Additionallyboth light curves exhibit a quasi-periodic rise and fall of IR fluxdays after periapse passage This phenomenon called ldquoringingrdquowas also seen in models by Langton amp Laughlin (2008a) andCowan amp Agol (2011) caused by the hottest point of the planetrotating in and out of view from Earth For ringing to occurthe temperature difference from dayside to nightside must belarge and it must survive for multiple planetary rotation periodsIn the models shown in Figure 16 this is aided by the rapidpseudo-synchronous rotation rate associated with an eccentric-ity of 075 (Table 2) For the top case the dayndashnight tempera-ture difference is over sim1300 K while the bottom case variesby sim750 K (Figure 7) The light curves from Figure 14 do notexhibit ringing because the dayndashnight temperature difference islow (sim300ndash400 K)

Because the radiative and dynamical timescales in the modelsby Cowan amp Agol (2011) are calculated by a simple scaling withtemperature the ringing seen in their simulated light curvesoccurs with a period equal to the planetrsquos assumed solid-bodyrotation period in the inertial frame of the winds In realityan eccentric exoplanet approaching apoapse would becomeincreasingly cold the winds would weaken and decrease andhence the ringing should be non-periodic Indeed for ourthree dimensional models the ringing in each case has varyingperiodsmdashthe top case has a ringing period ranging from 03 to04 days (Trot = 035 days) while the bottom panel has a periodranging 07ndash08 days (Trot = 07 days)

Despite these differences this study and the studies byLangton amp Laughlin (2008a) and Cowan amp Agol (2011) illus-trate the influence of atmospheric dynamics on the observationsof eccentric exoplanets The light curve features described herecould be seen in future follow-up observations of eccentric ex-oplanets and observers can use the conclusions drawn aboveto relate the observations to the planetrsquos atmospheric structureFor example if ringing is present in an IR light curve the planetmust have a strong dayndashnight temperature asymmetry and hencewe would expect the planet to have strong superrotation at theequator (Showman amp Polvani 2011) This is probably the casefor highly eccentric (e gt 05) short-period exoplanets such asHD 17156b and HD 80606b If ringing were observed it wouldalso allow an estimate of the sum of the rotation speed and thesuperrotation speed thereby placing constraints on both windspeeds and rotation rates of eccentric hot Jupiters The timing ofpeak IR flux relative to transit and secondary eclipse and also

15

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

f=000deg

(K)800

850

900

950

1000

f=minus1242deg

(K)800

850

900

950

1000f=1242deg

(K)800

850

900

950

1000

f=minus7227deg

(K)800

850

900

950

1000f=7227deg

(K)800

850

900

950

1000

f=minus11988deg

(K)800

850

900

950

1000f=11988deg

(K)800

850

900

950

1000

f=minus18000deg

(K)800

850

900

950

1000

Figure 15 Plots of temperature vs longitude and pressure in equatorial slices throughout one full orbit for our nominal model with Teq = 951 K e = 050 anda = 00534 AU Each annulus shows temperature vs longitude and log-pressure for a snapshot at a particular orbital position (denoted by the value of f) as viewedfrom the planetrsquos north pole Pressures between 30 and 43 mbar are displayed in each annulus The different annuli mark the time evolution throughout one completeorbit starting from apoapse for the top annulus proceeding to periapse and then back to apoapse (follow arrows moving counterclockwise around the plot) Thesesnapshots correspond to model outputs from 6103 to 6108 Earth days The solid line in each snapshot indicates the longitude of the substellar point while the dashedline denotes the Earth-facing longitude for the particular case of ω = 180 For the cases of ω = 270 and 360 degrees exactly the same plot is valid except that thesub-Earth longitude would be aiming upward for ω = 270 and right for ω = 360

(A color version of this figure is available in the online journal)

16

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

e = 075 a = 00385 AU

-15 -10 -05 00 05 10 15Time after periapse passage (days)

0000

0001

0002

0003

0004

Pla

netS

tar

Flu

x R

atio

e = 075 a = 00611 AU

-3 -2 -1 0 1 2 3Time after periapse passage (days)

00000

00005

00010

00015

00020

00025

Pla

netS

tar

Flu

x R

atio

Figure 16 Light Curves for two different e = 075 model integrationsTeq = 1199 K a = 00385 AU (top) and Teq = 951 K a = 00611 AU(bottom) both with ω = 180 The solid line denotes transit while the dottedline denotes secondary eclipse

(A color version of this figure is available in the online journal)

relative to orbital position can help to constrain the circulationfurther

5 CONCLUSIONS

We present three-dimensional circulation models coupledwith a two-stream non-gray radiative transfer scheme fora number of theoretical eccentric hot Jupiters We haveshown that as in published models with zero eccentricity ourhigh-eccentricity circulation models are dominated by eastwardflow at photospheric levels which cause an eastward displace-ment of the hottest regions from the substellar point The rapidrotation rates associated with pseudo-synchronization at higheccentricity lead to a small Rossby deformation radius andin some cases multiple jets in the atmosphere Global-meantemperatures and dayndashnight temperature differences peak notat periapse but several hours afterward due to finite radiativetimescales in the planetrsquos atmosphere

Furthermore we show that equatorial superrotation is gener-ated and maintained by eddies formed by the strong dayndashnightheating contrast which induce a flux of momentum from mid-latitudes to the equator The eddy magnitudes and momentumfluxes peak just after periapse passage leading to variations inthe zonal-mean flow throughout the orbit

Figure 17 Globally averaged pressurendashtemperature profiles from 1-D radiative-convective radiative transfer models of HD 189733b with 11 frequency bins(red profiles) 30 frequency bins (green profiles) and 196 frequency bins (blackprofiles) From left to right the profiles vary in stellar flux from the nominalvalue as well as 2times 4times and 8times higher flux The 8times higher flux case has atemperature inversion

(A color version of this figure is available in the online journal)

Lastly we have shown that the spatial and temporal vari-ations of the wind and temperature structure as well as theorbital viewing geometry of the system with respect to Earthcan affect the time and amplitude of peak IR flux seen infull-orbit light curves Depending on the viewing geometry ofthe orbit relative to Earth we find that peaks in IR flux thateither lead or lag periapse are possible in all cases a combi-nation of temporal effects (temperatures changing over time)and geometric effects (hot spots rotating into or out of view)are important in controlling the timing and amplitude of theflux peaks In cases where the dayndashnight temperature contrastis large and the rotational period is short the light curve canalso exhibit ldquoringingrdquo in flux as the hottest region of the planetrotates in and out of view This ringing is non-periodic due tothe variation in stellar heating as a function of distance

This work was supported by NASA Origins and PlanetaryAtmospheres grants to APS TK also acknowledges supportfrom the Harriet P Jenkins Pre-Doctoral Fellowship Program(JPFP) Resources supporting this work were provided bythe NASA High-End Computing (HEC) Program through theNASA Advanced Supercomputing (NAS) Division at AmesResearch Center The authors thank the anonymous referee fortheir helpful comments and suggestions

APPENDIX

VALIDATION TESTS FOR UPDATEDRADIATIVE TRANSFER SCHEME

To validate our updated radiative transfer scheme using 11frequency bins we conducted the following tests

1 First we tested the sensitivity of the number of frequencybins in a one-dimensional radiative-convective model ofHD 189733b (a = 00313 AU e = 00 RP = 82396 times107 m g = 214 m sminus2) Figure 17 compares the globallyaveraged pressurendashtemperature profiles of models run with11 frequency bins 30 frequency bins and 196 frequencybins at (from left to right) 1times 2times 4times and 8times theaverage stellar flux of HD 189733b For each group of

17

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)30 opacity bins

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)

11 opacity bins

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

30 opacity bins

minus600

minus200minus200

200

200

600

6001000

1000

14001800

2200

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

11 opacity bins

minus600

minus200

minus200

200

200

600

600

10001000

1400

1800

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

30 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

11 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

Figure 18 Globally averaged pressurendashtemperature profiles (top row) zonal-mean zonal wind (middle row) and the windtemperature profiles at 30 mbar (bottomrow) for the 30-bin (left column) and 11-bin (right column) model integrations of HD 189733b Snapshots were output at 250 Earth days

(A color version of this figure is available in the online journal)

models the 11-bin profile agrees well with the 30- and196-bin cases from 1 μbar to 1 bar the temperature variesby only tens of K Below 10 bars the 30- and 11-binprofiles differ from the 196-bin cases by up to 50 Kdue to the loss of resolution at low wavenumbers Inthe case of the highest average flux the large amount of

heating leads to a temperature inversion that is not fullycaptured by the 11-bin model but can be improved witha better initial guess at the pressure-temperature (P-T)profile

2 After testing the new radiative transfer (RT) scheme forone-dimensional models we ran full three-dimensional

18

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

simulations of HD 189733b at its nominal average stellarflux using the SPARCMITgcm Figure 18 shows the glob-ally averaged P-T profiles (top row) zonal-mean zonal wind(middle row) and windtemperature profiles at 30 mbar(bottom row) for the 30-bin (left column) and 11-bin (rightcolumn) RT setups The globally averaged profiles arenearly identical with any temperature differences withina few K The zonal wind plots also show good agree-ment with similar peak speeds (gt2600 m sminus1) equatorialjet width (sim60) and jet level (lt100 bar) At 30 mbarthe 30- and 11-bin models share similar dayside-nightsidetemperature structure with a hot spot (1100 K) eastward ofthe substellar point and two colder regions on the nightsidein the mid-latitudes

Both tests show that the bulk circulation and temperaturestructure is retained in the transition from 30 to 11 frequencybins hence we proceed to use the SPARCMITgcm with thenew scheme for these and future model simulations

REFERENCES

Adcroft A Campin J-M Hill C amp Marshall J 2004 MWRv 132 2845Charbonneau D Knutson H A Barman T et al 2008 ApJ 686 1341Cho J Y-K Menou K Hansen B M S amp Seager S 2008 ApJ

675 817Cooper C S amp Showman A P 2005 ApJL 629 L45Cooper C S amp Showman A P 2006 ApJ 649 1048Cowan N B amp Agol E 2011 ApJ 726 82Deming D Harrington J Laughlin G et al 2007 ApJL 667 L199Dobbs-Dixon I Cumming A amp Lin D N C 2010 ApJ 710 1395Dobbs-Dixon I amp Lin D N C 2008 ApJ 673 513Fortney J J Cooper C S Showman A P Marley M S amp Freedman R S

2006 ApJ 652 746Fortney J J Lodders K Marley M S amp Freedman R S 2008 ApJ

678 1419Fortney J J Marley M S Lodders K Saumon D amp Freedman R

2005 ApJL 627 L69

Fu Q amp Liou K N 1992 JAtS 49 2140Goody R West R Chen L amp Crisp D 1989 JQSRT 42 539Greenberg R 2009 ApJL 698 L42Heng K 2012 ApJL 761 L1Heng K Frierson D M W amp Phillipps P J 2011 MNRAS 418 2669Hide R 1969 JAtS 26 841Holton J R 2004 An Introduction to Dynamic Meteorology (Waltham MA

Academic)Hut P 1981 AampA 99 126Irwin J Charbonneau D Nutzman P et al 2008 ApJ 681 636Ivanov P B amp Papaloizou J C B 2004 MNRAS 353 1161Knutson H A Charbonneau D Allen L E Burrows A amp Megeath S T

2008 ApJ 673 526Knutson H A Charbonneau D Allen L E et al 2007 Natur 447 183Knutson H A Charbonneau D Cowan N B et al 2009 ApJ 690 822Langton J amp Laughlin G 2008a ApJ 674 1106Langton J amp Laughlin G 2008b AampA 483 25Lewis N K Showman A P Fortney J J et al 2010 ApJ 720 344Lodders K 2003 ApJ 591 1220Marley M S amp McKay C P 1999 Icar 138 268Mayor M amp Queloz D 1995 Natur 378 355Menou K amp Rauscher E 2009 ApJ 700 887Mlawer E J Taubman S J Brown P D Iacono M J amp Clough S A

1997 JGR 102 16663Naef D Latham D W Mayor M et al 2001 AampA 375 L27Perna R Heng K amp Pont F 2012 ApJ 751 59Pont F Knutson H Gilliland R L Moutou C amp Charbonneau D

2008 MNRAS 385 109Rauscher E amp Menou K 2010 ApJ 714 1334Rauscher E amp Menou K 2012 ApJ 750 96Saumon D Marley M S Abel M Frommhold L amp Freedman R S

2012 ApJ 750 74Showman A P Cho J Y-K amp Menou K 2011 in Exoplanets ed S Seager

(Tucson AZ Univ Arizona Press) 471Showman A P Cooper C S Fortney J J amp Marley M S 2008 ApJ

682 559Showman A P Fortney J J Lian Y et al 2009 ApJ 699 564Showman A P amp Guillot T 2002 AampA 385 166Showman A P amp Polvani L M 2011 ApJ 738 71Swain M R Vasisht G amp Tinetti G 2008 Natur 452 329Thrastarson H T amp Cho J Y 2010 ApJ 716 144Wolszczan A 1994 Sci 264 538

19

  • 1 INTRODUCTION
  • 2 MODEL
    • 21 Updates to SPARCMITgcm
    • 22 Grid Simulations
      • 3 RESULTS
        • 31 Circulation Regime
        • 32 Effect of Eccentricity
        • 33 Effect of Varying Average Stellar Flux
        • 34 Global Average Temperature versus Eccentricity and F
        • 35 Synchronous versus Pseudo-synchronous Rotation
        • 36 Momentum Budget and Equatorial Superrotation
          • 4 OBSERVATIONAL IMPLICATIONS
            • 41 Observational Effects of Varying ω
            • 42 High-e Model Integrations at Varying F
              • 5 CONCLUSIONS
              • APPENDIX VALIDATION TESTS FOR UPDATED RADIATIVE TRANSFER SCHEME
              • REFERENCES
Page 16: THREE-DIMENSIONAL ATMOSPHERIC CIRCULATION OF HOT …showman/publications/kataria-etal-2013.pdf · planets’ atmospheric composition, temperature structure, and circulation. A fifth

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

f=000deg

(K)800

850

900

950

1000

f=minus1242deg

(K)800

850

900

950

1000f=1242deg

(K)800

850

900

950

1000

f=minus7227deg

(K)800

850

900

950

1000f=7227deg

(K)800

850

900

950

1000

f=minus11988deg

(K)800

850

900

950

1000f=11988deg

(K)800

850

900

950

1000

f=minus18000deg

(K)800

850

900

950

1000

Figure 15 Plots of temperature vs longitude and pressure in equatorial slices throughout one full orbit for our nominal model with Teq = 951 K e = 050 anda = 00534 AU Each annulus shows temperature vs longitude and log-pressure for a snapshot at a particular orbital position (denoted by the value of f) as viewedfrom the planetrsquos north pole Pressures between 30 and 43 mbar are displayed in each annulus The different annuli mark the time evolution throughout one completeorbit starting from apoapse for the top annulus proceeding to periapse and then back to apoapse (follow arrows moving counterclockwise around the plot) Thesesnapshots correspond to model outputs from 6103 to 6108 Earth days The solid line in each snapshot indicates the longitude of the substellar point while the dashedline denotes the Earth-facing longitude for the particular case of ω = 180 For the cases of ω = 270 and 360 degrees exactly the same plot is valid except that thesub-Earth longitude would be aiming upward for ω = 270 and right for ω = 360

(A color version of this figure is available in the online journal)

16

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

e = 075 a = 00385 AU

-15 -10 -05 00 05 10 15Time after periapse passage (days)

0000

0001

0002

0003

0004

Pla

netS

tar

Flu

x R

atio

e = 075 a = 00611 AU

-3 -2 -1 0 1 2 3Time after periapse passage (days)

00000

00005

00010

00015

00020

00025

Pla

netS

tar

Flu

x R

atio

Figure 16 Light Curves for two different e = 075 model integrationsTeq = 1199 K a = 00385 AU (top) and Teq = 951 K a = 00611 AU(bottom) both with ω = 180 The solid line denotes transit while the dottedline denotes secondary eclipse

(A color version of this figure is available in the online journal)

relative to orbital position can help to constrain the circulationfurther

5 CONCLUSIONS

We present three-dimensional circulation models coupledwith a two-stream non-gray radiative transfer scheme fora number of theoretical eccentric hot Jupiters We haveshown that as in published models with zero eccentricity ourhigh-eccentricity circulation models are dominated by eastwardflow at photospheric levels which cause an eastward displace-ment of the hottest regions from the substellar point The rapidrotation rates associated with pseudo-synchronization at higheccentricity lead to a small Rossby deformation radius andin some cases multiple jets in the atmosphere Global-meantemperatures and dayndashnight temperature differences peak notat periapse but several hours afterward due to finite radiativetimescales in the planetrsquos atmosphere

Furthermore we show that equatorial superrotation is gener-ated and maintained by eddies formed by the strong dayndashnightheating contrast which induce a flux of momentum from mid-latitudes to the equator The eddy magnitudes and momentumfluxes peak just after periapse passage leading to variations inthe zonal-mean flow throughout the orbit

Figure 17 Globally averaged pressurendashtemperature profiles from 1-D radiative-convective radiative transfer models of HD 189733b with 11 frequency bins(red profiles) 30 frequency bins (green profiles) and 196 frequency bins (blackprofiles) From left to right the profiles vary in stellar flux from the nominalvalue as well as 2times 4times and 8times higher flux The 8times higher flux case has atemperature inversion

(A color version of this figure is available in the online journal)

Lastly we have shown that the spatial and temporal vari-ations of the wind and temperature structure as well as theorbital viewing geometry of the system with respect to Earthcan affect the time and amplitude of peak IR flux seen infull-orbit light curves Depending on the viewing geometry ofthe orbit relative to Earth we find that peaks in IR flux thateither lead or lag periapse are possible in all cases a combi-nation of temporal effects (temperatures changing over time)and geometric effects (hot spots rotating into or out of view)are important in controlling the timing and amplitude of theflux peaks In cases where the dayndashnight temperature contrastis large and the rotational period is short the light curve canalso exhibit ldquoringingrdquo in flux as the hottest region of the planetrotates in and out of view This ringing is non-periodic due tothe variation in stellar heating as a function of distance

This work was supported by NASA Origins and PlanetaryAtmospheres grants to APS TK also acknowledges supportfrom the Harriet P Jenkins Pre-Doctoral Fellowship Program(JPFP) Resources supporting this work were provided bythe NASA High-End Computing (HEC) Program through theNASA Advanced Supercomputing (NAS) Division at AmesResearch Center The authors thank the anonymous referee fortheir helpful comments and suggestions

APPENDIX

VALIDATION TESTS FOR UPDATEDRADIATIVE TRANSFER SCHEME

To validate our updated radiative transfer scheme using 11frequency bins we conducted the following tests

1 First we tested the sensitivity of the number of frequencybins in a one-dimensional radiative-convective model ofHD 189733b (a = 00313 AU e = 00 RP = 82396 times107 m g = 214 m sminus2) Figure 17 compares the globallyaveraged pressurendashtemperature profiles of models run with11 frequency bins 30 frequency bins and 196 frequencybins at (from left to right) 1times 2times 4times and 8times theaverage stellar flux of HD 189733b For each group of

17

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)30 opacity bins

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)

11 opacity bins

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

30 opacity bins

minus600

minus200minus200

200

200

600

6001000

1000

14001800

2200

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

11 opacity bins

minus600

minus200

minus200

200

200

600

600

10001000

1400

1800

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

30 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

11 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

Figure 18 Globally averaged pressurendashtemperature profiles (top row) zonal-mean zonal wind (middle row) and the windtemperature profiles at 30 mbar (bottomrow) for the 30-bin (left column) and 11-bin (right column) model integrations of HD 189733b Snapshots were output at 250 Earth days

(A color version of this figure is available in the online journal)

models the 11-bin profile agrees well with the 30- and196-bin cases from 1 μbar to 1 bar the temperature variesby only tens of K Below 10 bars the 30- and 11-binprofiles differ from the 196-bin cases by up to 50 Kdue to the loss of resolution at low wavenumbers Inthe case of the highest average flux the large amount of

heating leads to a temperature inversion that is not fullycaptured by the 11-bin model but can be improved witha better initial guess at the pressure-temperature (P-T)profile

2 After testing the new radiative transfer (RT) scheme forone-dimensional models we ran full three-dimensional

18

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

simulations of HD 189733b at its nominal average stellarflux using the SPARCMITgcm Figure 18 shows the glob-ally averaged P-T profiles (top row) zonal-mean zonal wind(middle row) and windtemperature profiles at 30 mbar(bottom row) for the 30-bin (left column) and 11-bin (rightcolumn) RT setups The globally averaged profiles arenearly identical with any temperature differences withina few K The zonal wind plots also show good agree-ment with similar peak speeds (gt2600 m sminus1) equatorialjet width (sim60) and jet level (lt100 bar) At 30 mbarthe 30- and 11-bin models share similar dayside-nightsidetemperature structure with a hot spot (1100 K) eastward ofthe substellar point and two colder regions on the nightsidein the mid-latitudes

Both tests show that the bulk circulation and temperaturestructure is retained in the transition from 30 to 11 frequencybins hence we proceed to use the SPARCMITgcm with thenew scheme for these and future model simulations

REFERENCES

Adcroft A Campin J-M Hill C amp Marshall J 2004 MWRv 132 2845Charbonneau D Knutson H A Barman T et al 2008 ApJ 686 1341Cho J Y-K Menou K Hansen B M S amp Seager S 2008 ApJ

675 817Cooper C S amp Showman A P 2005 ApJL 629 L45Cooper C S amp Showman A P 2006 ApJ 649 1048Cowan N B amp Agol E 2011 ApJ 726 82Deming D Harrington J Laughlin G et al 2007 ApJL 667 L199Dobbs-Dixon I Cumming A amp Lin D N C 2010 ApJ 710 1395Dobbs-Dixon I amp Lin D N C 2008 ApJ 673 513Fortney J J Cooper C S Showman A P Marley M S amp Freedman R S

2006 ApJ 652 746Fortney J J Lodders K Marley M S amp Freedman R S 2008 ApJ

678 1419Fortney J J Marley M S Lodders K Saumon D amp Freedman R

2005 ApJL 627 L69

Fu Q amp Liou K N 1992 JAtS 49 2140Goody R West R Chen L amp Crisp D 1989 JQSRT 42 539Greenberg R 2009 ApJL 698 L42Heng K 2012 ApJL 761 L1Heng K Frierson D M W amp Phillipps P J 2011 MNRAS 418 2669Hide R 1969 JAtS 26 841Holton J R 2004 An Introduction to Dynamic Meteorology (Waltham MA

Academic)Hut P 1981 AampA 99 126Irwin J Charbonneau D Nutzman P et al 2008 ApJ 681 636Ivanov P B amp Papaloizou J C B 2004 MNRAS 353 1161Knutson H A Charbonneau D Allen L E Burrows A amp Megeath S T

2008 ApJ 673 526Knutson H A Charbonneau D Allen L E et al 2007 Natur 447 183Knutson H A Charbonneau D Cowan N B et al 2009 ApJ 690 822Langton J amp Laughlin G 2008a ApJ 674 1106Langton J amp Laughlin G 2008b AampA 483 25Lewis N K Showman A P Fortney J J et al 2010 ApJ 720 344Lodders K 2003 ApJ 591 1220Marley M S amp McKay C P 1999 Icar 138 268Mayor M amp Queloz D 1995 Natur 378 355Menou K amp Rauscher E 2009 ApJ 700 887Mlawer E J Taubman S J Brown P D Iacono M J amp Clough S A

1997 JGR 102 16663Naef D Latham D W Mayor M et al 2001 AampA 375 L27Perna R Heng K amp Pont F 2012 ApJ 751 59Pont F Knutson H Gilliland R L Moutou C amp Charbonneau D

2008 MNRAS 385 109Rauscher E amp Menou K 2010 ApJ 714 1334Rauscher E amp Menou K 2012 ApJ 750 96Saumon D Marley M S Abel M Frommhold L amp Freedman R S

2012 ApJ 750 74Showman A P Cho J Y-K amp Menou K 2011 in Exoplanets ed S Seager

(Tucson AZ Univ Arizona Press) 471Showman A P Cooper C S Fortney J J amp Marley M S 2008 ApJ

682 559Showman A P Fortney J J Lian Y et al 2009 ApJ 699 564Showman A P amp Guillot T 2002 AampA 385 166Showman A P amp Polvani L M 2011 ApJ 738 71Swain M R Vasisht G amp Tinetti G 2008 Natur 452 329Thrastarson H T amp Cho J Y 2010 ApJ 716 144Wolszczan A 1994 Sci 264 538

19

  • 1 INTRODUCTION
  • 2 MODEL
    • 21 Updates to SPARCMITgcm
    • 22 Grid Simulations
      • 3 RESULTS
        • 31 Circulation Regime
        • 32 Effect of Eccentricity
        • 33 Effect of Varying Average Stellar Flux
        • 34 Global Average Temperature versus Eccentricity and F
        • 35 Synchronous versus Pseudo-synchronous Rotation
        • 36 Momentum Budget and Equatorial Superrotation
          • 4 OBSERVATIONAL IMPLICATIONS
            • 41 Observational Effects of Varying ω
            • 42 High-e Model Integrations at Varying F
              • 5 CONCLUSIONS
              • APPENDIX VALIDATION TESTS FOR UPDATED RADIATIVE TRANSFER SCHEME
              • REFERENCES
Page 17: THREE-DIMENSIONAL ATMOSPHERIC CIRCULATION OF HOT …showman/publications/kataria-etal-2013.pdf · planets’ atmospheric composition, temperature structure, and circulation. A fifth

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

e = 075 a = 00385 AU

-15 -10 -05 00 05 10 15Time after periapse passage (days)

0000

0001

0002

0003

0004

Pla

netS

tar

Flu

x R

atio

e = 075 a = 00611 AU

-3 -2 -1 0 1 2 3Time after periapse passage (days)

00000

00005

00010

00015

00020

00025

Pla

netS

tar

Flu

x R

atio

Figure 16 Light Curves for two different e = 075 model integrationsTeq = 1199 K a = 00385 AU (top) and Teq = 951 K a = 00611 AU(bottom) both with ω = 180 The solid line denotes transit while the dottedline denotes secondary eclipse

(A color version of this figure is available in the online journal)

relative to orbital position can help to constrain the circulationfurther

5 CONCLUSIONS

We present three-dimensional circulation models coupledwith a two-stream non-gray radiative transfer scheme fora number of theoretical eccentric hot Jupiters We haveshown that as in published models with zero eccentricity ourhigh-eccentricity circulation models are dominated by eastwardflow at photospheric levels which cause an eastward displace-ment of the hottest regions from the substellar point The rapidrotation rates associated with pseudo-synchronization at higheccentricity lead to a small Rossby deformation radius andin some cases multiple jets in the atmosphere Global-meantemperatures and dayndashnight temperature differences peak notat periapse but several hours afterward due to finite radiativetimescales in the planetrsquos atmosphere

Furthermore we show that equatorial superrotation is gener-ated and maintained by eddies formed by the strong dayndashnightheating contrast which induce a flux of momentum from mid-latitudes to the equator The eddy magnitudes and momentumfluxes peak just after periapse passage leading to variations inthe zonal-mean flow throughout the orbit

Figure 17 Globally averaged pressurendashtemperature profiles from 1-D radiative-convective radiative transfer models of HD 189733b with 11 frequency bins(red profiles) 30 frequency bins (green profiles) and 196 frequency bins (blackprofiles) From left to right the profiles vary in stellar flux from the nominalvalue as well as 2times 4times and 8times higher flux The 8times higher flux case has atemperature inversion

(A color version of this figure is available in the online journal)

Lastly we have shown that the spatial and temporal vari-ations of the wind and temperature structure as well as theorbital viewing geometry of the system with respect to Earthcan affect the time and amplitude of peak IR flux seen infull-orbit light curves Depending on the viewing geometry ofthe orbit relative to Earth we find that peaks in IR flux thateither lead or lag periapse are possible in all cases a combi-nation of temporal effects (temperatures changing over time)and geometric effects (hot spots rotating into or out of view)are important in controlling the timing and amplitude of theflux peaks In cases where the dayndashnight temperature contrastis large and the rotational period is short the light curve canalso exhibit ldquoringingrdquo in flux as the hottest region of the planetrotates in and out of view This ringing is non-periodic due tothe variation in stellar heating as a function of distance

This work was supported by NASA Origins and PlanetaryAtmospheres grants to APS TK also acknowledges supportfrom the Harriet P Jenkins Pre-Doctoral Fellowship Program(JPFP) Resources supporting this work were provided bythe NASA High-End Computing (HEC) Program through theNASA Advanced Supercomputing (NAS) Division at AmesResearch Center The authors thank the anonymous referee fortheir helpful comments and suggestions

APPENDIX

VALIDATION TESTS FOR UPDATEDRADIATIVE TRANSFER SCHEME

To validate our updated radiative transfer scheme using 11frequency bins we conducted the following tests

1 First we tested the sensitivity of the number of frequencybins in a one-dimensional radiative-convective model ofHD 189733b (a = 00313 AU e = 00 RP = 82396 times107 m g = 214 m sminus2) Figure 17 compares the globallyaveraged pressurendashtemperature profiles of models run with11 frequency bins 30 frequency bins and 196 frequencybins at (from left to right) 1times 2times 4times and 8times theaverage stellar flux of HD 189733b For each group of

17

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)30 opacity bins

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)

11 opacity bins

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

30 opacity bins

minus600

minus200minus200

200

200

600

6001000

1000

14001800

2200

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

11 opacity bins

minus600

minus200

minus200

200

200

600

600

10001000

1400

1800

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

30 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

11 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

Figure 18 Globally averaged pressurendashtemperature profiles (top row) zonal-mean zonal wind (middle row) and the windtemperature profiles at 30 mbar (bottomrow) for the 30-bin (left column) and 11-bin (right column) model integrations of HD 189733b Snapshots were output at 250 Earth days

(A color version of this figure is available in the online journal)

models the 11-bin profile agrees well with the 30- and196-bin cases from 1 μbar to 1 bar the temperature variesby only tens of K Below 10 bars the 30- and 11-binprofiles differ from the 196-bin cases by up to 50 Kdue to the loss of resolution at low wavenumbers Inthe case of the highest average flux the large amount of

heating leads to a temperature inversion that is not fullycaptured by the 11-bin model but can be improved witha better initial guess at the pressure-temperature (P-T)profile

2 After testing the new radiative transfer (RT) scheme forone-dimensional models we ran full three-dimensional

18

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

simulations of HD 189733b at its nominal average stellarflux using the SPARCMITgcm Figure 18 shows the glob-ally averaged P-T profiles (top row) zonal-mean zonal wind(middle row) and windtemperature profiles at 30 mbar(bottom row) for the 30-bin (left column) and 11-bin (rightcolumn) RT setups The globally averaged profiles arenearly identical with any temperature differences withina few K The zonal wind plots also show good agree-ment with similar peak speeds (gt2600 m sminus1) equatorialjet width (sim60) and jet level (lt100 bar) At 30 mbarthe 30- and 11-bin models share similar dayside-nightsidetemperature structure with a hot spot (1100 K) eastward ofthe substellar point and two colder regions on the nightsidein the mid-latitudes

Both tests show that the bulk circulation and temperaturestructure is retained in the transition from 30 to 11 frequencybins hence we proceed to use the SPARCMITgcm with thenew scheme for these and future model simulations

REFERENCES

Adcroft A Campin J-M Hill C amp Marshall J 2004 MWRv 132 2845Charbonneau D Knutson H A Barman T et al 2008 ApJ 686 1341Cho J Y-K Menou K Hansen B M S amp Seager S 2008 ApJ

675 817Cooper C S amp Showman A P 2005 ApJL 629 L45Cooper C S amp Showman A P 2006 ApJ 649 1048Cowan N B amp Agol E 2011 ApJ 726 82Deming D Harrington J Laughlin G et al 2007 ApJL 667 L199Dobbs-Dixon I Cumming A amp Lin D N C 2010 ApJ 710 1395Dobbs-Dixon I amp Lin D N C 2008 ApJ 673 513Fortney J J Cooper C S Showman A P Marley M S amp Freedman R S

2006 ApJ 652 746Fortney J J Lodders K Marley M S amp Freedman R S 2008 ApJ

678 1419Fortney J J Marley M S Lodders K Saumon D amp Freedman R

2005 ApJL 627 L69

Fu Q amp Liou K N 1992 JAtS 49 2140Goody R West R Chen L amp Crisp D 1989 JQSRT 42 539Greenberg R 2009 ApJL 698 L42Heng K 2012 ApJL 761 L1Heng K Frierson D M W amp Phillipps P J 2011 MNRAS 418 2669Hide R 1969 JAtS 26 841Holton J R 2004 An Introduction to Dynamic Meteorology (Waltham MA

Academic)Hut P 1981 AampA 99 126Irwin J Charbonneau D Nutzman P et al 2008 ApJ 681 636Ivanov P B amp Papaloizou J C B 2004 MNRAS 353 1161Knutson H A Charbonneau D Allen L E Burrows A amp Megeath S T

2008 ApJ 673 526Knutson H A Charbonneau D Allen L E et al 2007 Natur 447 183Knutson H A Charbonneau D Cowan N B et al 2009 ApJ 690 822Langton J amp Laughlin G 2008a ApJ 674 1106Langton J amp Laughlin G 2008b AampA 483 25Lewis N K Showman A P Fortney J J et al 2010 ApJ 720 344Lodders K 2003 ApJ 591 1220Marley M S amp McKay C P 1999 Icar 138 268Mayor M amp Queloz D 1995 Natur 378 355Menou K amp Rauscher E 2009 ApJ 700 887Mlawer E J Taubman S J Brown P D Iacono M J amp Clough S A

1997 JGR 102 16663Naef D Latham D W Mayor M et al 2001 AampA 375 L27Perna R Heng K amp Pont F 2012 ApJ 751 59Pont F Knutson H Gilliland R L Moutou C amp Charbonneau D

2008 MNRAS 385 109Rauscher E amp Menou K 2010 ApJ 714 1334Rauscher E amp Menou K 2012 ApJ 750 96Saumon D Marley M S Abel M Frommhold L amp Freedman R S

2012 ApJ 750 74Showman A P Cho J Y-K amp Menou K 2011 in Exoplanets ed S Seager

(Tucson AZ Univ Arizona Press) 471Showman A P Cooper C S Fortney J J amp Marley M S 2008 ApJ

682 559Showman A P Fortney J J Lian Y et al 2009 ApJ 699 564Showman A P amp Guillot T 2002 AampA 385 166Showman A P amp Polvani L M 2011 ApJ 738 71Swain M R Vasisht G amp Tinetti G 2008 Natur 452 329Thrastarson H T amp Cho J Y 2010 ApJ 716 144Wolszczan A 1994 Sci 264 538

19

  • 1 INTRODUCTION
  • 2 MODEL
    • 21 Updates to SPARCMITgcm
    • 22 Grid Simulations
      • 3 RESULTS
        • 31 Circulation Regime
        • 32 Effect of Eccentricity
        • 33 Effect of Varying Average Stellar Flux
        • 34 Global Average Temperature versus Eccentricity and F
        • 35 Synchronous versus Pseudo-synchronous Rotation
        • 36 Momentum Budget and Equatorial Superrotation
          • 4 OBSERVATIONAL IMPLICATIONS
            • 41 Observational Effects of Varying ω
            • 42 High-e Model Integrations at Varying F
              • 5 CONCLUSIONS
              • APPENDIX VALIDATION TESTS FOR UPDATED RADIATIVE TRANSFER SCHEME
              • REFERENCES
Page 18: THREE-DIMENSIONAL ATMOSPHERIC CIRCULATION OF HOT …showman/publications/kataria-etal-2013.pdf · planets’ atmospheric composition, temperature structure, and circulation. A fifth

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)30 opacity bins

800 1000 1200 1400 1600

10minus4

10minus2

100

102

104

temperature (K)

pre

ssu

re (

bar

)

11 opacity bins

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

30 opacity bins

minus600

minus200minus200

200

200

600

6001000

1000

14001800

2200

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus80 minus60 minus40 minus20 0 20 40 60 80

10minus3

10minus2

10minus1

100

101

102

latitude (deg)

pre

ssu

re (

bar

)

11 opacity bins

minus600

minus200

minus200

200

200

600

600

10001000

1400

1800

2200

2600

(m sminus1)minus1000

minus500

0

500

1000

1500

2000

2500

3000

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

30 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

minus150 minus100 minus50 0 50 100 150

minus80

minus60

minus40

minus20

0

20

40

60

80

longitude (deg)

lati

tud

e (deg

)

11 opacity bins 3036 mbar

(K)

600

700

800

900

1000

1100

1200

Figure 18 Globally averaged pressurendashtemperature profiles (top row) zonal-mean zonal wind (middle row) and the windtemperature profiles at 30 mbar (bottomrow) for the 30-bin (left column) and 11-bin (right column) model integrations of HD 189733b Snapshots were output at 250 Earth days

(A color version of this figure is available in the online journal)

models the 11-bin profile agrees well with the 30- and196-bin cases from 1 μbar to 1 bar the temperature variesby only tens of K Below 10 bars the 30- and 11-binprofiles differ from the 196-bin cases by up to 50 Kdue to the loss of resolution at low wavenumbers Inthe case of the highest average flux the large amount of

heating leads to a temperature inversion that is not fullycaptured by the 11-bin model but can be improved witha better initial guess at the pressure-temperature (P-T)profile

2 After testing the new radiative transfer (RT) scheme forone-dimensional models we ran full three-dimensional

18

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

simulations of HD 189733b at its nominal average stellarflux using the SPARCMITgcm Figure 18 shows the glob-ally averaged P-T profiles (top row) zonal-mean zonal wind(middle row) and windtemperature profiles at 30 mbar(bottom row) for the 30-bin (left column) and 11-bin (rightcolumn) RT setups The globally averaged profiles arenearly identical with any temperature differences withina few K The zonal wind plots also show good agree-ment with similar peak speeds (gt2600 m sminus1) equatorialjet width (sim60) and jet level (lt100 bar) At 30 mbarthe 30- and 11-bin models share similar dayside-nightsidetemperature structure with a hot spot (1100 K) eastward ofthe substellar point and two colder regions on the nightsidein the mid-latitudes

Both tests show that the bulk circulation and temperaturestructure is retained in the transition from 30 to 11 frequencybins hence we proceed to use the SPARCMITgcm with thenew scheme for these and future model simulations

REFERENCES

Adcroft A Campin J-M Hill C amp Marshall J 2004 MWRv 132 2845Charbonneau D Knutson H A Barman T et al 2008 ApJ 686 1341Cho J Y-K Menou K Hansen B M S amp Seager S 2008 ApJ

675 817Cooper C S amp Showman A P 2005 ApJL 629 L45Cooper C S amp Showman A P 2006 ApJ 649 1048Cowan N B amp Agol E 2011 ApJ 726 82Deming D Harrington J Laughlin G et al 2007 ApJL 667 L199Dobbs-Dixon I Cumming A amp Lin D N C 2010 ApJ 710 1395Dobbs-Dixon I amp Lin D N C 2008 ApJ 673 513Fortney J J Cooper C S Showman A P Marley M S amp Freedman R S

2006 ApJ 652 746Fortney J J Lodders K Marley M S amp Freedman R S 2008 ApJ

678 1419Fortney J J Marley M S Lodders K Saumon D amp Freedman R

2005 ApJL 627 L69

Fu Q amp Liou K N 1992 JAtS 49 2140Goody R West R Chen L amp Crisp D 1989 JQSRT 42 539Greenberg R 2009 ApJL 698 L42Heng K 2012 ApJL 761 L1Heng K Frierson D M W amp Phillipps P J 2011 MNRAS 418 2669Hide R 1969 JAtS 26 841Holton J R 2004 An Introduction to Dynamic Meteorology (Waltham MA

Academic)Hut P 1981 AampA 99 126Irwin J Charbonneau D Nutzman P et al 2008 ApJ 681 636Ivanov P B amp Papaloizou J C B 2004 MNRAS 353 1161Knutson H A Charbonneau D Allen L E Burrows A amp Megeath S T

2008 ApJ 673 526Knutson H A Charbonneau D Allen L E et al 2007 Natur 447 183Knutson H A Charbonneau D Cowan N B et al 2009 ApJ 690 822Langton J amp Laughlin G 2008a ApJ 674 1106Langton J amp Laughlin G 2008b AampA 483 25Lewis N K Showman A P Fortney J J et al 2010 ApJ 720 344Lodders K 2003 ApJ 591 1220Marley M S amp McKay C P 1999 Icar 138 268Mayor M amp Queloz D 1995 Natur 378 355Menou K amp Rauscher E 2009 ApJ 700 887Mlawer E J Taubman S J Brown P D Iacono M J amp Clough S A

1997 JGR 102 16663Naef D Latham D W Mayor M et al 2001 AampA 375 L27Perna R Heng K amp Pont F 2012 ApJ 751 59Pont F Knutson H Gilliland R L Moutou C amp Charbonneau D

2008 MNRAS 385 109Rauscher E amp Menou K 2010 ApJ 714 1334Rauscher E amp Menou K 2012 ApJ 750 96Saumon D Marley M S Abel M Frommhold L amp Freedman R S

2012 ApJ 750 74Showman A P Cho J Y-K amp Menou K 2011 in Exoplanets ed S Seager

(Tucson AZ Univ Arizona Press) 471Showman A P Cooper C S Fortney J J amp Marley M S 2008 ApJ

682 559Showman A P Fortney J J Lian Y et al 2009 ApJ 699 564Showman A P amp Guillot T 2002 AampA 385 166Showman A P amp Polvani L M 2011 ApJ 738 71Swain M R Vasisht G amp Tinetti G 2008 Natur 452 329Thrastarson H T amp Cho J Y 2010 ApJ 716 144Wolszczan A 1994 Sci 264 538

19

  • 1 INTRODUCTION
  • 2 MODEL
    • 21 Updates to SPARCMITgcm
    • 22 Grid Simulations
      • 3 RESULTS
        • 31 Circulation Regime
        • 32 Effect of Eccentricity
        • 33 Effect of Varying Average Stellar Flux
        • 34 Global Average Temperature versus Eccentricity and F
        • 35 Synchronous versus Pseudo-synchronous Rotation
        • 36 Momentum Budget and Equatorial Superrotation
          • 4 OBSERVATIONAL IMPLICATIONS
            • 41 Observational Effects of Varying ω
            • 42 High-e Model Integrations at Varying F
              • 5 CONCLUSIONS
              • APPENDIX VALIDATION TESTS FOR UPDATED RADIATIVE TRANSFER SCHEME
              • REFERENCES
Page 19: THREE-DIMENSIONAL ATMOSPHERIC CIRCULATION OF HOT …showman/publications/kataria-etal-2013.pdf · planets’ atmospheric composition, temperature structure, and circulation. A fifth

The Astrophysical Journal 76776 (19pp) 2013 April 10 Kataria et al

simulations of HD 189733b at its nominal average stellarflux using the SPARCMITgcm Figure 18 shows the glob-ally averaged P-T profiles (top row) zonal-mean zonal wind(middle row) and windtemperature profiles at 30 mbar(bottom row) for the 30-bin (left column) and 11-bin (rightcolumn) RT setups The globally averaged profiles arenearly identical with any temperature differences withina few K The zonal wind plots also show good agree-ment with similar peak speeds (gt2600 m sminus1) equatorialjet width (sim60) and jet level (lt100 bar) At 30 mbarthe 30- and 11-bin models share similar dayside-nightsidetemperature structure with a hot spot (1100 K) eastward ofthe substellar point and two colder regions on the nightsidein the mid-latitudes

Both tests show that the bulk circulation and temperaturestructure is retained in the transition from 30 to 11 frequencybins hence we proceed to use the SPARCMITgcm with thenew scheme for these and future model simulations

REFERENCES

Adcroft A Campin J-M Hill C amp Marshall J 2004 MWRv 132 2845Charbonneau D Knutson H A Barman T et al 2008 ApJ 686 1341Cho J Y-K Menou K Hansen B M S amp Seager S 2008 ApJ

675 817Cooper C S amp Showman A P 2005 ApJL 629 L45Cooper C S amp Showman A P 2006 ApJ 649 1048Cowan N B amp Agol E 2011 ApJ 726 82Deming D Harrington J Laughlin G et al 2007 ApJL 667 L199Dobbs-Dixon I Cumming A amp Lin D N C 2010 ApJ 710 1395Dobbs-Dixon I amp Lin D N C 2008 ApJ 673 513Fortney J J Cooper C S Showman A P Marley M S amp Freedman R S

2006 ApJ 652 746Fortney J J Lodders K Marley M S amp Freedman R S 2008 ApJ

678 1419Fortney J J Marley M S Lodders K Saumon D amp Freedman R

2005 ApJL 627 L69

Fu Q amp Liou K N 1992 JAtS 49 2140Goody R West R Chen L amp Crisp D 1989 JQSRT 42 539Greenberg R 2009 ApJL 698 L42Heng K 2012 ApJL 761 L1Heng K Frierson D M W amp Phillipps P J 2011 MNRAS 418 2669Hide R 1969 JAtS 26 841Holton J R 2004 An Introduction to Dynamic Meteorology (Waltham MA

Academic)Hut P 1981 AampA 99 126Irwin J Charbonneau D Nutzman P et al 2008 ApJ 681 636Ivanov P B amp Papaloizou J C B 2004 MNRAS 353 1161Knutson H A Charbonneau D Allen L E Burrows A amp Megeath S T

2008 ApJ 673 526Knutson H A Charbonneau D Allen L E et al 2007 Natur 447 183Knutson H A Charbonneau D Cowan N B et al 2009 ApJ 690 822Langton J amp Laughlin G 2008a ApJ 674 1106Langton J amp Laughlin G 2008b AampA 483 25Lewis N K Showman A P Fortney J J et al 2010 ApJ 720 344Lodders K 2003 ApJ 591 1220Marley M S amp McKay C P 1999 Icar 138 268Mayor M amp Queloz D 1995 Natur 378 355Menou K amp Rauscher E 2009 ApJ 700 887Mlawer E J Taubman S J Brown P D Iacono M J amp Clough S A

1997 JGR 102 16663Naef D Latham D W Mayor M et al 2001 AampA 375 L27Perna R Heng K amp Pont F 2012 ApJ 751 59Pont F Knutson H Gilliland R L Moutou C amp Charbonneau D

2008 MNRAS 385 109Rauscher E amp Menou K 2010 ApJ 714 1334Rauscher E amp Menou K 2012 ApJ 750 96Saumon D Marley M S Abel M Frommhold L amp Freedman R S

2012 ApJ 750 74Showman A P Cho J Y-K amp Menou K 2011 in Exoplanets ed S Seager

(Tucson AZ Univ Arizona Press) 471Showman A P Cooper C S Fortney J J amp Marley M S 2008 ApJ

682 559Showman A P Fortney J J Lian Y et al 2009 ApJ 699 564Showman A P amp Guillot T 2002 AampA 385 166Showman A P amp Polvani L M 2011 ApJ 738 71Swain M R Vasisht G amp Tinetti G 2008 Natur 452 329Thrastarson H T amp Cho J Y 2010 ApJ 716 144Wolszczan A 1994 Sci 264 538

19

  • 1 INTRODUCTION
  • 2 MODEL
    • 21 Updates to SPARCMITgcm
    • 22 Grid Simulations
      • 3 RESULTS
        • 31 Circulation Regime
        • 32 Effect of Eccentricity
        • 33 Effect of Varying Average Stellar Flux
        • 34 Global Average Temperature versus Eccentricity and F
        • 35 Synchronous versus Pseudo-synchronous Rotation
        • 36 Momentum Budget and Equatorial Superrotation
          • 4 OBSERVATIONAL IMPLICATIONS
            • 41 Observational Effects of Varying ω
            • 42 High-e Model Integrations at Varying F
              • 5 CONCLUSIONS
              • APPENDIX VALIDATION TESTS FOR UPDATED RADIATIVE TRANSFER SCHEME
              • REFERENCES