A Ground-Based Search for Thermal Emission from the Exoplanet TrES-1

a r

X

i

v

:07

5

.

4

2

8

8

v

1

[

a

s

t

r

o

-

p

h

]

2

9

M

a

y

2

7

A Ground-Based Search for Thermal Emission from the Exoplanet TrES-1Heather A.Knutson,David Charbonneau 1Harvard-Smithsonian Center for Astrophysics,60Garden Street,Cambridge,MA 02138hknutson@aa004c86ec3a87c24028c415,dcharbonneau@aa004c86ec3a87c24028c415 Drake Deming Planetary Systems Laboratory,Code 693,Goddard Space Flight Center,Greenbelt,MD 20771ddeming@aa004c86ec3a87c24028c415 and L.Jeremy Richardson Exoplanet and Stellar Astrophysics Laboratory,Code 667,Goddard Space Flight Center,Greenbelt,MD 20771richardsonlj@aa004c86ec3a87c24028c415 ABSTRACT Eclipsing planetary systems give us an important window on extrasolar planet atmospheres.By measuring the depth of the secondary eclipse,when the planet moves behind the star,we can estimate the strength of the thermal emission from

the day side of the planet.Attaining a ground-based detection of one of these eclipses has proven to be a signi?cant challenge,as time-dependent variations in instrument throughput and atmospheric seeing and absorption overwhelm the small signal of the eclipse at infrared wavelengths.We gathered a series of si-multaneous L grism spectra of the transiting planet system TrES-1and a nearby comparison star of comparable brightness,allowing us to correct for these e?ects in aa004c86ec3a87c24028c415bining the data from two eclipses,we demonstrate a detec-tion sensitivity of 0.15%in the eclipse depth relative to the stellar ?ux.This

–2–

approaches the sensitivity required to detect the planetary emission,which the-

oretical models predict should lie between0.05?0.1%of the stellar?ux in our

2.9?4.3μm bandpass.We explore the factors that ultimately limit the precision

of this technique,and discuss potential avenues for future improvements.

Subject headings:infrared:techniques:spectroscopic-eclipses-stars:individual:

TrES-1-extrasolar planets

1.Introduction

Recent observations using the Spitzer Space Telescope have detected the thermal emis-

sion from several transiting planets,including TrES-1(Deming et al.2005,2006;Charbonneau et al. 2005;Grillmair et al.2007;Richardson et al.2007;Swain et al.2007).By measuring the de-crease in?ux when the planet moves behind the star,Charbonneau et al.(2005)estimated

the brightness temperature of the planet at4.5μm and8.0μm.In the shorter wavelength bandpass,they estimated an eclipse depth of0.00066±0.00013in relative?ux for TrES-1, corresponding to a brightness temperature of1010±60K for the planet.This brightness temperature is signi?cantly lower than predicted by models,which indicate that hot Jupiters should have a strong peak in their emission at4μm,created by a gap between absorption bands from CO and H2O.There are a number of possible explanations for the low?ux at

4.5μm,including a higher-than-expected abundance of CO and additional opacity sources in

the atmosphere,but the bandpass-integrated nature of Spitzer photometry make it di?cult

to determine the correct explanation.

In this paper we describe a series of L-band spectroscopic observations of two secondary eclipses of TrES-1(Alonso et al.2004;Sozzetti et al.2004;Laughlin et al.2005;Winn et al. 2007)using the NIRI instrument on Gemini North.Because the ultimate goal is to constrain theoretical model spectra,spectroscopic observations are inherently more useful;although we must still bin our data over a range of wavelengths in order to detect the secondary eclipse,

we could,in principle,vary our bandpasses to measure the amplitude of the4μm peak

and other spectral features directly.For our measurement we observe TrES-1and a close comparison star simultaneously in the same slit and use this second set of spectra to remove

the e?ects of variable atmospheric absorption and slit losses due to telescope pointing jitter

and seeing.

Our data represent the?rst ground-based attempt to detect the secondary eclipse of TrES-1;previous ground-based studies(Richardson et al.2003a,b;Snellen2005;Deming et al. 2007)focused on HD209458b,which lacks a nearby comparison star of similar infrared

–3–

brightness.As a result,Richardson et al.(2003a,b),Deming et al.(2007)and Snellen(2005) were forced to nod between the target and the calibration star,which limited their ability to adequately remove these time-dependent variations.Deming et al.(2007)note that they achieved the most accurate correction using the nearer of their two comparison stars,at a separation of1.7?,but the lack of coherence between the observed?uxes for the two stars indicated that absorption from telluric water vapor was still incoherent on this angular scale. Snellen(2007)obtained relative photometry spanning secondary eclipse for the transiting planet system OGLE-TR-113,which is located in a crowded?eld.The1.3′square?eld of view contained several nearby stars of comparable brightness for calibration.They report a tentative detection of the secondary eclipse,indicating that the use of nearby comparison stars can signi?cantly reduce errors.Similarly,we selected TrES-1for our experiment be-cause it has a nearby comparison star;by placing both stars in the slit simultaneously,we planned to simply ratio the two observed spectra to correct for time-dependent instrumental and telluric variations.

2.Observations and Analysis

We used the Near InfraRed Imager(NIRI)on Gemini North(Hodapp et al.2003)to observe TrES-1and a nearby star(2MASS19041058+3638409,44′′distant)in the same slit during secondary eclipses on UT2006May11and July26.We obtained a series of758L spectra of each star spanning the wavelength range from2.9?4.3μm.We chose the widest slit available for L grism spectroscopy,with dimensions of0.75′′×110′′,to maximize the total ?ux and minimize slit losses due to guiding errors.We used3s exposures,with15coadds each,for our images.

2.1.Image Reduction

In order to remove the sky background we nod the telescope in an ABBA pattern by by±5′′for the May eclipse and±2.5′′for the July eclipse,then di?erence the resulting pairs of images.Immediately after di?erencing the two images we?t the data with a function designed to remove periodic detector noise(see the end of§2.3for a full description of this step).Because the?ats were taken with the shutter closed they are dominated by the thermal blackbody spectrum of the warm shutter.We remove this e?ect by taking the central200rows in the?at-?eld image and binning them to make a spectrum,then divide all rows in our?at-?eld image by this composite spectrum.After?at-?elding the data, we calculate the errors associated with each pixel in our images using the equation derived

–4–

by Vacca,Cushing,&Rayner(2004).The variance in electron counts associated with an individual pixel in the images before?at-?elding is given by:

V I=gI+2n cσ2read(1) where I is the?ux in ADU,g is the gain(g=12.3e?ADU?1for these images),n c is the number of co-adds,andσread is the rms read noise(σread=70e?pix?1).Once the raw images are di?erenced and?at-?elded,the per pixel variance is given by:

V A+V B

V D=

–5–

by dividing the timeseries by the median out-of-transit value.The resulting timeseries are shown in Fig.2and3.

2.3.Removing Systematic Trends

Variations at the level of10?20%are apparent in the timeseries.These variations are well-correlated with the o?sets of the spectra in the x(dispersion)direction,which serves as a proxy for the position of the star along the short axis of the slit.We estimate these o?sets by cross-correlating the strong telluric feature around3.3μm in individual spectra with a composite spectrum made by combining ten individual spectra gathered at low airmass(this ensures our template spectrum has a higher signal to noise than the individual spectra). We attempt to remove these trends by?tting the out-of-eclipse data at each nod position separately with a quadratic function of the x shift.Additional degrees of freedom did not produce signi?cant improvements.This decorrelation reduces the level of variation to1?2% (see Figure4).We searched for additional correlations with variables including airmass, detector temperature,y o?set(cross-dispersion direction),various proxies for changes in focus including the width of the dispersion pro?le of the2D spectra and the width of the cross-correlation function of the extracted1D spectra,and the total?ux from the calibration star.None of these variables correlate with the residuals,although the small size of these trends relative to the scatter in the data makes it di?cult to rule out subtle correlations with a high degree of con?dence.

The total binned?uxes from the two stars varied by a factor of three during our obser-vations as the stellar centroids moved in and out of the slit center.Although most of this variation is removed when we take the ratio of the TrES-1spectrum to the companion star’s spectrum,it is possible that a nonlinear detector response might account for some of the residual trends.Our raw images,which are dominated by sky emission,have a typical?ux ranging from1×104e?at the shortest wavelengths to8×104e?at the longest wavelengths. Although this is well below the saturation limit of3×106e?,the instrument is still nonlin-ear at a level of approximately1%for these illumination levels(Andrew Stephens,personal correspondence,Sept.2006).To quantify the e?ect that this nonlinearity might have on our ?nal binned timeseries,we apply the following nonlinearity correction to our raw images. We assumed the nonlinearity scaled with?ux,according to:

f1=f 1+K f

–6–

created our?nal timeseries using these corrected images,individual values in the normalized timeseries changed by only0.03%?0.07%on average and the overall variance in the data was e?ectively unchanged.

We note that the RMS variation of the decorrelated timeseries exceeds the expectations from Poisson and read noise by a factor of4.It is possible that some of this noise might be related to the properties of the detector;infrared detectors often have sign?cant power close to the Nyquist frequency.Indeed,when we examined our images in Fourier space we noticed there was signi?cant row-to-row power at a period of two pixels.Although most of this noise is removed when the images are pair-subtracted,a small but signi?cant peak remains in the power spectrum at a frequency of0.25pix?1.

To remove these residuals,we treat each quadrant in the pair-subtracted images sepa-rately.We?t each column with a sine function with a?xed period of four pixels and solve for the phase and the amplitude,which was permitted to vary quadratically as a function of position.We then subtract the best-?t function from the column,and repeat for all the columns in the quadrant.This method removed98%of the power at four-pixel frequencies. When we created our timeseries using images with four-pixel frequencies removed,the?ux values of individual points changed by0.15%on average and the overall variance in the data was unchanged.

3.Discussion

For each of the four timeseries,we estimate the depth of the eclipse(Table1).We?x the system parameters to the best-?t values published by Winn et al.(2007),and allowed the depth of the eclipse to vary over both positive and negative values.We calculated our errors using a bootstrap Monte Carlo analysis.We note that the only time series to give a positive eclipse depth(July26Nod B)also contains the largest correlated residuals (Fig4).When we take the weighted average of these estimates,we?nd an eclipse depth of?0.0010±0.0015in relative?ux,which is consistent with no variation.We also note that the eclipse depth may vary over time due to changing emission patterns on the planet (Rauscher et al.2006),although this variation is predicted to be much smaller than the sensitivity of our measurements.Published models of the emitted dayside spectrum for hot Jupiters(Fortney et al.2005;Seager et al.2005;Barman et al.2005;Burrows et al.2006) predict that the depth of the eclipse at the wavelengths of our observations will depend on the strength of various molecular absorption bands,including CO and H2O,as well as the e?ciency with which energy is circulated between the permanent day and night sides of the planet.Fortney et al.(2005)predict eclipse depths for TrES-1of?0.0010to?0.0005for

–7–

this bandpass,with the smaller eclipse depths corresponding to a model in which energy is e?ciently recirculated from the dayside to the nightside in the planetary atmosphere, and the larger eclipse depths corresponding to ine?cient circulation and a large day-night temperature di?erential.

We also considered a restricted bandpass from3.8?4.1μm,centered on the4μm peak in the planet’s emission predicted by theoretical models.Fortney et al.(2005)predict that TrES-1should have an eclipse depth between?0.0015and?0.0009for this wavelength range.We estimate a value of+0.0061±0.0026(i.e.an increase in?ux during the eclipse) for this wavelength range.The trends that are visible in Figure4are even larger for these smaller wavelength ranges,thus we conclude that the measurement is not signi?cant.We also binned the spectrum into a red and blue timeseries and took the ratio of these two timeseries to search for color-dependent variations in the depth of the eclipse,but this increased the overall noise level and failed to remove any of the systematic trends,which are di?erent in these two regions of the spectrum.

We have not been able to determine the precise source of the noise that ultimately limits our measurements,but there are several conclusions that we draw from the analysis. First,there do not appear to be any signi?cant correlations between relative?ux variations and airmass,humidity,temperature,and other properties of the local observing conditions. This is an improvement over previous studies(Richardson et al.2003a,b;Snellen2005,2007; Deming et al.2007),which concluded that the correction for atmospheric absorption was the limiting factor when the comparison star was located at distances of2′or more.

A central limitation to our experimental design is the slit,which when coupled to point-ing jitter and variations in the focus and seeing introduces large changes in the apparent ?ux.In particular,we see a strong correlation between the centroid of the stars on the slit and the measured?ux levels,even after dividing the measured?ux for TrES-1by the companion’s?ux.Although we are able to remove most of these variations by decorrelating with the x position of the spectra,there are still some trends remaining after our decorre-lation.We note that most of the issues listed above would be mitigated with the use of a wider slit;although we selected the widest slit available on NIRI,this slit width was still comparable to the point spread functions of the two stars and required us to make multiple adjustments of the telescope pointing over the course of the night to counteract the drift of stars out of the slit.Because we must bin the data over a relatively wide range in wave-length to reduce the photon noise to the level required to seek the secondary eclipse,the loss in wavelength resolution that would result from a wider slit would not be signi?cant. Although spectroscopic observations of the secondary eclipse are scienti?cally more valuable, we note that photometric observations would also avoid many of the problems encountered

–8–

here.Previous ground-based attempts to detect the secondary eclipse of HD209458were limited by the lack of nearby bright stars in the?eld of view;it is entirely possible that these observations could be successful if they were repeated with targets such as TrES-1,which have such nearby companions.

4.Conclusions

We estimate a depth of?0.0010±0.0015for the eclipse from2.9?4.3μm,which is con-sistent with zero.Despite employing a method designed to remove time-dependent variations in atmospheric absorption and seeing,we were unable to reduce the noise to the level required to address theoretical models of the planetary emission at these wavelengths.The noise in our data is signi?cantly higher than that predicted from Poisson and detector read noise. This may be related to the interplay between seeing and slit losses,although we were unable to verify this directly.Based on our experience with these data,it is possible to accurately correct for time-dependent variations in atmospheric absorption for stars with nearby(less than1′)bright companions,and remaining issues with slit losses may be addressed with either a wider slit or purely photometric(as opposed to spectroscopic)observations.Although our proposed modi?cations would reduce our ability to resolve features in the emission spectra of these planets,such compromises may be necessary in order to achieve a successful detec-tion of a secondary eclipse from the ground.This will become increasingly important in the next several years,as Spitzer is predicted to run out of cryogen in early2009.Additionally, ground-based transit surveys(Alonso et al.2004;McCullough et al.2006;O’Donovan et al. 2006;Collier Cameron et al.2007;Bakos et al.2007)are expected to double the number of known transiting planet systems by the end of2007.These surveys tend to target bright stars in relatively crowded?elds,and it is likely that many of these newly discovered systems will have bright nearby comparison stars that make them suitable for the kinds of observations we have described above.

This work is based on observations obtained as part of program GN-2006A-Q-3at the Gemini Observatory,which is operated by the Association of Universities for Research in Astronomy,Inc.,under a cooperative agreement with the NSF on behalf of the Gemini partnership:the National Science Foundation(United States),the Particle Physics and Astronomy Research Council(United Kingdom),the National Research Council(Canada), CONICYT(Chile),the Australian Research Council(Australia),CNPq(Brazil)and CON-ICET(Argentina).We are grateful to Chad Trujillo and the entire Gemini team for their assistance throughout this process.HAK was supported by a National Science Foundation Graduate Research Fellowship.LJR was supported by a NASA Postdoctoral Fellowship at

–9–

NASA Goddard.

REFERENCES

Alonso,R.,et al.2004,ApJ,613,L153

Bakos,G.′A.et al.2007,ApJ,656,552

Barman,T.S.,Hauschildt,P.H.,&Allard,F.2005,ApJ,632,1132

Burrows,A.,Sudarsky,D.,&Hubeny,I.ApJ,650,1140

Charbonneau,D.,et al.2005,ApJ,626,523

Collier Cameron,A.et al.2007,MNRAS,375,951

Cushing,M.,Vacca,W.,&Rayner,J.2004,PASP,116,362

Deming,D.,Seager,S.,&Richardson,L.J.2005,Nature,434,740

Deming,D.,Harrington,J.,Seager,S.,&Richardson,L.J.ApJ,644,560

Deming,D.,Richardson,L.J.,&Harrington,J.2007,MNRAS,in press

Fortney,J.J.,Marley,M.S.,Lodders,K.,Saumon,D.,&Freedman,R.2005,ApJ,627,L69 Grillmair,C.J.et al.2007,ApJ,658,L115

Hodapp,K.W.et al.2003,PASP,115,1388

Horne,K.1986,PASP,98,609

Laughlin,G.et al.2005,ApJ,621,1072

McCullough,P.R.,et al.2006,ApJ,648,1228

O’Donovan,F.T.et al.2006,ApJ,651,L61

Rauscher,E.,Menou,K.Cho,J.Y.-K.,Seager,S.,&Hansen,B.2006,ApJ,submitted, astro-ph/0612413

Richardson,L.J.,Deming,D.,Wiedemann,G.,Goukenleuque,C.,Steyert,D.,Harrington, J.,Esposito,L.W.2003,ApJ,584,1053

Richardson,L.J.,Deming,D.,&Seager,S.2003,ApJ,597,581

–10–

Richardson,L.J.,Deming,D.,Horning,K.,Seager,S.,&Harrington,J.2007,Nature,445, 892

Seager,S.et al.2005,ApJ,632,1122

Snellen,I.2005,MNRAS,363,211

Snellen,I.&Covino,E.2007,MNRAS,375,307

Sozzetti,A.et al.2004,ApJ,616,L167

Swain,M.R.,Bouwman,J.,Akeson,R.,Lawler,S.,Beichman,C.2007,ApJ,submitted, astro-ph/0702593

Vacca,W.,Cushing,M.,&Rayner,T.2004,PASP,116,352

Winn,J.J.,Holman,M.J.,&Roussanova,A.2007,ApJ,657,1098

–11–

Table1.Estimates of the eclipse depth(2.9?4.3μm) Date Depth at Position A Depth at Position B

–12–

3.0 3.2 3.4 3.6 3.8

4.0 4.2

Wavelength (microns)1

2

3

4

5

F l u x (104 e )TrES-1Comparison Fig.1.—These are the averaged spectra for TrES-1(bottom)and the companion star (top)from UT 2006July 26.The dominant spectral features are from telluric absorption.

–13–

Time from Center of Transit (d)R e l a t i v e F l u x

Shift in X Position (pixels)R e l a t i v e F l u

x

Time from Center of Transit (d)

R e l a t i v e F l u x

Shift in X Position (pixels)

R e l a t i v e F l u x

Fig.2.—The panels on the left show the relative ?ux values as a function of time for the UT 2006May 11eclipse,for nod positions A (top)and B (bottom),and the panels on the right show the correlation between these values and shift in the x (dispersion)direction.

–14–

Time from Center of Transit (d)R e l a t i v e F l u x

Shift in X Position (pixels)R e l a t i v e F l u

x Time from Center of Transit (d)

R e l a t i v e F l u x

Shift in X Position (pixels)

R e l a t i v e F l u x

Fig.3.—The panels on the left show the relative ?ux values as a function of time for the UT 2006July 26eclipse,for nod positions A (top)and B (bottom),and the panels on the right show the correlation between these values and the shift in the x (dispersion)direction.

–15–

Time from Center of Transit [d]R e l a t i v e F l u x + C o n s t a n t Fig.4.—The relative change in ?ux for both eclipses and nod positions,along with best-?t transit curves for each timeseries.We treat each of the two nod positions separately in our analysis,as they have di?erent dependencies on the shift in the x (dispersion)direction and airmass.

本文来源：https://www.bwwdw.com/article/v1pq.html