1: \documentclass[preprint,12pt]{aastex}
2:
3: \shortauthors{Winn et al.~2008}
4: \shorttitle{Albedo of TrES-3}
5:
6: \begin{document}
7:
8: % ------------------------------------------------------------------------
9: % New commands
10: %
11: \def\ltsima{$\; \buildrel < \over \sim \;$}
12: \def\lsim{\lower.5ex\hbox{\ltsima}}
13: \def\gtsima{$\; \buildrel > \over \sim \;$}
14: \def\gsim{\lower.5ex\hbox{\gtsima}}
15: \def\lam{\lambda=-1\fdg4 \pm 1\fdg1}
16: % -------------------------------------------------------------------------
17: %
18:
19: \bibliographystyle{apj}
20:
21: \title{
22: The Transit Light Curve Project.\\
23: VIII.~Six Occultations of the Exoplanet TrES-3
24: }
25:
26: \author{
27: Joshua N.\ Winn\altaffilmark{1},
28: Matthew J.\ Holman\altaffilmark{2},
29: Avi Shporer\altaffilmark{3},
30: Jos\'e Fern\'andez\altaffilmark{2},\\
31: Tsevi Mazeh\altaffilmark{3},
32: David W.\ Latham\altaffilmark{2},
33: David Charbonneau\altaffilmark{2},
34: Mark E.~Everett\altaffilmark{4}
35: }
36:
37: \altaffiltext{1}{Department of Physics, and Kavli Institute for
38: Astrophysics and Space Research, Massachusetts Institute of
39: Technology, Cambridge, MA 02139, USA}
40:
41: \altaffiltext{2}{Harvard-Smithsonian Center for Astrophysics, 60
42: Garden Street, Cambridge, MA 02138, USA}
43:
44: \altaffiltext{3}{Wise Observatory, Tel Aviv University, Tel Aviv
45: 69978, Israel}
46:
47: \altaffiltext{4}{Planetary Science Institute, 1700 E.~Fort Lowell Rd.,
48: Suite 106, Tucson, AZ 85719}
49:
50: \begin{abstract}
51:
52: We present photometry of the exoplanet host star TrES-3 spanning six
53: occultations (secondary eclipses) of its giant planet. No flux
54: decrements were detected, leading to 99\%-confidence upper limits on
55: the planet-to-star flux ratio of $2.4 \times 10^{-4}$, $5.0\times
56: 10^{-4}$, and $8.6\times 10^{-4}$ in the $i$, $z$, and $R$ bands
57: respectively. The corresponding upper limits on the planet's
58: geometric albedo are 0.30, 0.62, and 1.07. The upper limit in the
59: $i$ band rules out the presence of highly reflective clouds, and is
60: only a factor of 2--3 above the predicted level of thermal radiation
61: from the planet.
62:
63: \end{abstract}
64:
65: \keywords{planetary systems --- stars:~individual (TrES-3,
66: GSC~03089--00929) --- techniques:~photometric}
67:
68: \section{Introduction}
69:
70: The detection of reflected light from close-in giant planets is a
71: difficult but worthwhile goal. Knowledge of the albedo is important
72: because the stellar insolation is a critical factor in the atmospheric
73: structure and the overall thermal balance of these planets (Guillot et
74: al.~1996, Saumon et al.~1996, Seager \& Sasselov~1998). Measurement of
75: the planet's reflection spectrum, or at least the broad
76: wavelength-dependence of the albedo, would provide clues about the
77: dominant scattering mechanisms and constituents in the planetary
78: atmosphere (see, e.g., Seager \& Sasselov 1998, Marley et al.~1999,
79: Sudarsky et al.~2000, Barman et al.~2001). The difficulty is that the
80: reflected light is a very small fraction of the direct starlight,
81: producing a planet-to-star flux ratio at opposition of
82: $\epsilon_\lambda = p_\lambda (R_p/a)^2$, where $R_p$ is the planetary
83: radius, $a$ is the orbital separation, and $p_\lambda$ is the
84: wavelength-dependent geometric albedo.\footnote{The geometric albedo
85: is defined as the flux reflected by the planet when viewed at
86: opposition (full phase), divided by the flux that would be reflected
87: by a flat and perfectly diffusing surface with the same
88: cross-sectional area as the planet.} Even for a planet as large as
89: Jupiter, and an orbital separation as small as 0.05~AU,
90: $\epsilon_\lambda \approx 10^{-4}$~$p_\lambda$. The reflected signal
91: is also expected to vary over the orbital period, in a manner
92: depending on the orbital inclination and the phase function of the
93: planetary atmosphere.
94:
95: Early attempts to detect reflected light from close-in giant planets,
96: by Charbonneau et al.~(1999), Collier Cameron et al.~(1999, 2002), and
97: Leigh et al.~(2003a,b), relied on optical spectroscopy. Those
98: investigators sought the reflected copies of the stellar spectral
99: lines by combining suitably Doppler-shifted spectra taken over a range
100: of orbital phases. They did not detect the reflected signal and placed
101: upper limits\footnote{Although Collier Cameron et al.~(1999) reported
102: a detection with greater than 95\% confidence, additional data
103: obtained by the same group did not confirm the detection; see
104: Collier Cameron et al.~(2002).} on the geometric albedo in the
105: visual band of $\sim$0.1--0.5, subject to some ambiguity because of
106: the unknown radius and orbital inclination of the planets they
107: observed. Liu et al.~(2008) recently revisited the spectroscopic
108: technique and applied it to HD~209458b, for which the planetary radius
109: and orbital inclination are known, but the available data provided
110: only a weak constraint on the geometric albedo ($0.8\pm 1.6$ from
111: 554$-$681~nm).
112:
113: Another detection method involves polarimetry. The Stokes parameters
114: are expected to vary over the planetary orbital period because
115: reflected light is preferentially polarized while direct starlight is
116: unpolarized (see, e.g., Hough et al.~2006). Berdyugina et al.~(2007)
117: reported a detection of a time-variable polarized signal from
118: HD~189733. Taking the scattering radius to be the same as the optical
119: radius measured through transit photometry, the implied geometric
120: albedo is larger than $2/3$, which is the geometric albedo of a
121: perfectly diffusing sphere. While this is physically possible, and
122: indeed some Solar system objects have geometric albedos exceeding
123: unity due to strong backscattering, Berdyugina et al.~(2007) preferred
124: an interpretation in which the albedo is smaller and the scattering
125: radius is larger than the optical radius measured during transits.
126:
127: For planets whose orbits are viewed close enough to edge-on that they
128: undergo periodic occultations by the star, a powerful and conceptually
129: simple method is available: when the planet is hidden by the star, the
130: total light should diminish by the fraction
131: $\epsilon_\lambda$. Edge-on systems are also advantageous because the
132: planetary radius and orbital inclination can be determined precisely
133: from transit observations.\footnote{It is possible, however, for a
134: planet on an eccentric orbit to exhibit occultations without
135: transits, or transits without occultations (see, e.g., Irwin et
136: al.~2008).} If the measurement noise were limited only by photon
137: statistics, the photometric technique could be employed with a smaller
138: telescope than would be needed for either the spectroscopic or
139: polarimetric techniques. However, ground-based photometry is generally
140: afflicted by systematic errors that prevent one from achieving the
141: required precision. To date, only spaceborne photometry has provided
142: meaningful upper limits on exoplanetary albedos. In particular, Rowe
143: et al.~(2007) have used the {\it Microvariability and Oscillations of
144: Stars}\, satellite to set a 3$\sigma$ upper limit of 0.17 on the
145: geometric albedo of HD~209458.
146:
147: Discoveries of transiting planets have abounded in the past few years,
148: and there are now several systems known with larger values of
149: $(R_p/a)^2$. One of the most favorable systems is TrES-3, discovered
150: by O'Donovan et al.~(2007), for which $(R_p/a)^2 = 7.5\times
151: 10^{-4}$. In this system, a planet with mass 2~$M_{\rm Jup}$ and
152: radius 1.3~$R_{\rm Jup}$ circles a G dwarf star with an orbital period
153: of 31~hr. This paper relates our attempts to detect reflected light
154: from the TrES-3 planet using ground-based photometry with meter-class
155: telescopes. In \S~2, we describe our observations and data reduction
156: procedures. In \S~3, we present our results. In \S~4, we place these
157: results in the context of other observations and of theoretical
158: expectations for the albedo and the thermal emission spectrum.
159:
160: \section{Observations and Data Reduction}
161:
162: We observed TrES-3 on 6 different nights when planetary occultations
163: were expected, according to the ephemeris of O'Donovan et
164: al.~(2007). Assuming that the planetary orbit is circular, as expected
165: for a planet with such a short orbital period\footnote{The timescale
166: for tidal circularization is $\sim$$2\times 10^7$~yr, using Eqn.~(9)
167: of Rasio et al.~(1996) with the system parameters of O'Donovan et
168: al.~(2007) and a tidal quality factor $Q=10^5$. This is shorter than
169: the estimated age of the star, $5\times 10^8$~yr (Torres et
170: al.~2008). The current radial velocity data are consistent with a
171: circular orbit and give an upper limit on the eccentricity of 0.12
172: with 99\% confidence.}, the uncertainty in the predicted
173: mid-occultation time was smaller than 3 minutes on each night.
174: Observations on other nights were attempted, but only the six nights
175: described below offered clear enough skies for high-precision
176: photometry.
177:
178: On 2007~April~27, 2007~July~4, 2007~September~14, and 2008~March~12,
179: we used the 1.2~m telescope at the Fred L.\ Whipple Observatory (FLWO)
180: on Mount Hopkins, Arizona. (Here and elsewhere, the quoted date refers
181: to the UT date at the start of the night.) The detector was KeplerCam,
182: a 4096$^2$ CCD with a square field of view $23\farcm 1$ on a side
183: (Szentgyorgi et al.~2005). We binned the images $2\times 2$, giving a
184: scale of $0\farcs68$ per binned pixel. We obtained repeated 60~s
185: exposures over the course of 4-5~hr bracketing the predicted
186: mid-occultation time. The dead time between exposures was
187: 11~s. Autoguiding maintained the pointing to within 2-3 pixels
188: throughout the observations. On the first night, 2007~April~27, we
189: observed through a Sloan $i$ filter ($\approx$0.7--0.85~$\mu$m), as
190: the star rose from an airmass of 1.4 to the meridian. The full-width
191: at half-maximum (FWHM) of stellar images was about 4.5~pixels
192: ($3\arcsec$). On the next two nights, 2007~July~4 and
193: 2007~September~14, we observed through a Sloan $z$ filter
194: ($\approx$0.85--1.0~$\mu$m, with the red cutoff arising from the
195: quantum efficiency of the detector). On both of those nights, the
196: target star began near the meridian and set to an airmass near 2.0
197: throughout the observations, and the FWHM was approximately 3~pixels
198: ($2\arcsec$). On the last night, 2008~March~12, we chose the Sloan $i$
199: filter again and observed through an airmass ranging from 1.1 to 2.0.
200: Conditions were nearly photometric, and the seeing was exceptionally
201: good and stable, with a FWHM of 2.1~pixels ($1\farcs4$).
202:
203: On UT~2007~August~26 and 2007~October~3, we used the 1~m telescope at
204: Wise Observatory in Israel. We used a Princeton Instruments
205: $1340\times 1300$ back-illuminated CCD, giving a field of view of
206: $13\farcm 0 \times 12\farcm 6$ and a pixel scale of $0\farcs 58$. The
207: readout and refresh time was 25~s, and we used exposure times ranging
208: from 60~s to 160~s, depending on the airmass and seeing. The guider
209: generally maintained the pointing to within a few pixels, but there
210: was a glitch on each night that led to a 10-20 pixel offset. We
211: observed through a Bessell $R$ filter ($\approx$0.55--0.7~$\mu$m). On
212: August~26, we observed for 6~hr as the target star set from the
213: meridian to an airmass of 2.8, and on October~3, we observed for 4~hr
214: as the target set from airmass 1.1 to 2.5. In both cases, the data
215: obtained through an airmass greater than 2.0 proved to be much noisier
216: than the rest of the data, and ultimately were not used in our
217: analysis. On both nights the seeing steadily worsened with increasing
218: airmass, from a FWHM of 2.5~pixels to 4.5~pixels ($1\farcs8$ to
219: $3\farcs2$).
220:
221: We used standard IRAF procedures for overscan correction, trimming,
222: bias subtraction, and flat-field division. We performed aperture
223: photometry of TrES-3 and several nearby stars. The sum of the fluxes
224: of the comparison stars was taken to be the comparison signal. The
225: flux timeseries for TrES-3 was divided by the comparison signal, and
226: then by a constant chosen to give a unit mean flux outside of the
227: predicted occultation. We experimented with different choices for the
228: aperture size, and different combinations of the comparison stars,
229: aiming to minimize the standard deviation of the out-of-occultation
230: portion of the TrES-3 light curve. Not surprisingly, the best results
231: were obtained when the aperture diameter was about twice as large as
232: the FWHM of the stellar images, and when as many comparison stars as
233: possible were used with mean fluxes ranging from 50--150\% of the mean
234: flux of TrES-3. For the FLWO data, 10-13 stars were used, and for the
235: Wise data, 6-7 stars were used. The Wise light curve had an abrupt
236: jump at precisely the time when the guiding failed and the pointing
237: changed; to correct for this, we set the mean out-of-occultation flux
238: to be unity for each of the two segments of data. Finally, for all
239: data sets, we performed an airmass correction. Using only the
240: out-of-occultation data, we fitted the light curve to an exponential
241: function of airmass, divided by the best-fitting function, and
242: renormalized the light curve to have unit mean flux outside of the
243: occultation.
244:
245: The final relative photometry is plotted in Figure~1, and are given in
246: numerical form in Table~1. One of the transit light curves of
247: O'Donovan et al.~(2007) is also plotted in Figure~1, because the
248: transit data were used in our model for the occultation data, as
249: described in the next section. The duration of the occultation is
250: expected to be nearly the same as the duration of the transit, since
251: the orbit is presumed to be circular, as discussed further in \S~4.
252:
253: \begin{figure}[p]
254: \epsscale{0.9}
255: \plotone{f1.eps}
256: \caption{Photometry of TrES-3. The top panel shows the $z$-band
257: transit photometry of O'Donovan et al.~(2007). The other 3 panels
258: show our photometry spanning occultations (secondary eclipses).
259: The dotted lines show the expected start and end of the occultation.
260: \label{fig:1}}
261: \end{figure}
262:
263: \begin{deluxetable}{lccc}
264: \tabletypesize{\normalsize}
265: \tablecaption{Photometry of TrES-3\label{tbl:photometry}}
266: \tablewidth{0pt}
267:
268: \tablehead{
269: \colhead{Telescope} &
270: \colhead{Filter} &
271: \colhead{Heliocentric Julian Date} &
272: \colhead{Relative flux}
273: }
274:
275: \startdata
276: FLWO & i & 2454217.806989 & 0.99905 \\
277: FLWO & i & 2454217.807834 & 1.00095 \\
278: FLWO & i & 2454217.808667 & 1.00006 \\
279: FLWO & i & 2454217.809524 & 1.00156 \\
280: FLWO & i & 2454217.810369 & 1.00144 \\
281: FLWO & i & 2454217.811214 & 0.99904
282: %\input table1.tex
283: \enddata
284:
285: \tablecomments{The time stamps represent the Heliocentric Julian Date
286: at the time of mid-exposure. We intend for this Table to appear in
287: entirety in the electronic version of the journal. An excerpt is
288: shown here to illustrate its format. The data are also available
289: from the authors upon request.}
290:
291: \end{deluxetable}
292:
293: Since no flux decrement is obvious in any of the light curves, the
294: results are upper limits on the planet-to-star flux ratio in each
295: bandpass. This makes it especially important to understand the
296: characteristics of the noise. The out-of-occultation FLWO $i$ light
297: curves from 2007~April~26 and 2008~March~12 have standard deviations
298: of $9.1\times 10^{-4}$ and $9.0\times 10^{-4}$, respectively. The
299: expected level of photon-counting (Poisson) noise is $8\times
300: 10^{-4}$, and the expected level of scintillation noise is $3\times
301: 10^{-5}$, according to Eqn.~1 of Young~(1967). The quadrature sum of
302: these terms is $8.5\times 10^{-4}$. The FLWO $z$ light curves from
303: 2007~July~14 and 2007~September~14 have standard deviations of
304: $1.8\times 10^{-3}$ and $1.4\times 10^{-3}$, as compared to the
305: expected level of about $1.2\times 10^{-3}$ from Poisson and
306: scintillation noise. Thus, most of the observed noise can be accounted
307: for. The noise in the $i$-band light curves, in particular, is only
308: 6-7\% larger than the noise from these two expected sources.
309:
310: The out-of-occultation Wise data from 2007~August~26 and
311: 2007~October~3 have standard deviations of $2.2\times 10^{-3}$ and
312: $1.9\times 10^{-3}$. In both cases the expected level of
313: photon-counting noise is about $10^{-3}$ and the expected level of
314: scintillation noise is about $3\times 10^{-5}$. Thus, the noise in the
315: Wise data is about twice as large as one would expect from these two
316: sources. Because of the jump that was observed in each light curve at
317: the time of a large pixel shift, we suspect that at least some (and
318: perhaps most) of the excess noise is due to an imperfect flat
319: field. On other nights when the telescope lost tracking more
320: frequently the noise level was indeed even higher, although those
321: other nights were also characterized by poorer sky conditions, making
322: it impossible to isolate the source of excess noise.
323:
324: Next we tested for any correlations with time or other external
325: variables, and for significant non-Gaussianity. We found no
326: significant correlation (a correlation coefficient $<$0.2) between the
327: final relative flux data and the airmass, the pixel position, or the
328: measured shape parameters (FWHM, ellipticity, position angle) of
329: stellar images. Of course, we had already decorrelated against the
330: airmass, and for the Wise data we had already removed the effects of
331: the two pointing glitches, so the lack of correlations with those
332: variables was expected. For each light curve, we also examined the
333: histogram of flux values, the autocorrelation function, and the factor
334: by which the noise level is reduced when the data are binned. Because
335: the histograms are approximately Gaussian, the autocorrelations are
336: generally as small as would be expected for uncorrelated noise, and
337: the noise falls approximately as $1/\sqrt{N}$, in what follows we
338: treat the noise as Gaussian and uncorrelated.
339:
340: \section{Determination of Upper Limits on the Albedo}
341:
342: One might imagine measuring the planet-to-star flux ratio,
343: $\epsilon_\lambda$, by finding the ratio of the mean flux during the
344: occultation to the mean out-of-occultation flux. However, because the
345: TrES-3 planet is on a nearly-grazing trajectory, the durations of the
346: ingress and egress cannot be ignored. Furthermore, the uncertainty in
347: the orbital inclination $i$ introduces some uncertainty in the
348: expected shape of the light curve, and the conversion from
349: $\epsilon_\lambda$ into the geometric albedo $p_\lambda$ requires
350: knowledge of the ratio $R_p/a$ and its uncertainty.
351:
352: To take all of these factors into account, we simultaneously fitted a
353: parameterized model to our $R$, $i$, and $z$-band occultation
354: photometry and the high-precision $B$ and $z$-band transit photometry
355: of O'Donovan et al.~(2007). The model and the fitting method are
356: similar to those we have employed in previous papers in this series
357: (see, e.g., Winn et al.~2007, Holman et al.~2007). The model posits a
358: circular orbit of a planet with radius $R_p$ and a star with radius
359: $R_\star$, with orbital inclination $i$, orbital separation $a$, and
360: period $P$. The flux ratio between the full disk of the star and
361: fully-illuminated disk of the planet is $\epsilon_k$, where $k$ refers
362: to the bandpass. When the planet is projected in front of the star, or
363: vice versa, we use the analytic formulae of Mandel \& Agol (2002) to
364: compute the appropriate flux decrement. We assumed that the flux
365: received from the planet does not vary appreciably over the 5~hr span
366: of our occultation data, i.e., that the planet's phase function is
367: constant within $30\arcdeg$ of opposition. For the star, we assumed
368: the limb-darkening law to be quadratic, with
369: coefficients\footnote{Specifically, we used $a=0.2508$,~$b=0.3019$ for
370: the $z$-band data; and $a=0.7250$,~$b=0.0967$ for the $B$-band
371: data. These are based on interpolations for a star with $T_{\rm
372: eff}=5720$~K, $\log g=4.6$, and solar
373: metallicity. Southworth~(2008) has found that the uncertainties in
374: the transit parameters (the scaled radii of the star and planet, and
375: the orbital inclination) are underestimated when the limb-darkening
376: coefficients are fixed at theoretical values. However, our results
377: for the geometric albedo are not susceptible to this problem, as we
378: confirmed by repeating the fit with a linear limb-darkening law and
379: allowing the coefficients to be free parameters.} taken from
380: Claret~(2000, 2004). The model parameters were $\epsilon_i$,
381: $\epsilon_z$, $\epsilon_R$, $R_p/a$, $R_\star/a$, $i$, $P$, and $T_c$
382: (a particular time of midtransit), along with the airmass correction
383: parameters for each light curve.
384:
385: For the transit light curves, we adopted the same flux uncertainties
386: as O'Donovan et al.~(2007). For the occultation light curves, based on
387: the investigation of the photometric errors described in the previous
388: section, we took the standard deviation of the out-of-occultation flux
389: to be the uncertainty in each flux measurement. To determine the
390: best-fitting values of the parameters and their uncertainties, we used
391: a Markov Chain Monte Carlo (MCMC) algorithm. This algorithm creates a
392: random sequence of points in parameter space whose density is an
393: approximation of the {\it a posteriori}\, probability distribution of
394: the parameter values. The sequence is generated from an initial point
395: by iterating a ``jump function,'' which in our case was the addition
396: of a Gaussian random number to a randomly-chosen parameter. If the new
397: point has a lower $\chi^2$ than the previous point, the jump is
398: executed; if not, the jump is executed with probability
399: $\exp(-\Delta\chi^2/2)$. We used an isotropic prior for the orbital
400: inclination (uniform in $\cos i$), and a Gaussian prior in the orbital
401: period to enforce consistency with the value $P=1.30619 \pm
402: 0.00001$~d, which was determined by O'Donovan et al.~(2007) using the
403: yearlong baseline between the TrES survey data and the high-precision
404: light curves. We used uniform priors for all other parameters. The
405: planet-to-star flux ratios were required to be positive. After
406: creating several independent chains to verify that they all converged
407: to the same region of parameter space, we created one long chain of
408: $5\times 10^6$ points for our final results.
409:
410: In Table~2, we provide the results for the 99\%-confidence upper limit
411: on the planet-to-star flux ratio and the geometric albedo for each
412: bandpass. The geometric albedo was computed as $p_\lambda =
413: \epsilon_\lambda(a/R_p)^2$. The upper limit was defined as the value
414: for which the cumulative {\it a posteriori}\, probability distribution
415: took the value 0.99. Figure~2 shows the {\it a posteriori}\,
416: probability distribution for the geometric albedo based on the data
417: from each bandpass, and for all of the data in combination. The
418: results for the transit-based parameters $R_p/R_\star$, $R_\star/a$,
419: $i$, and $T_c$ were in agreement with the published values of
420: O'Donovan et al.~(2007).
421:
422: \begin{figure}[ht]
423: \epsscale{0.5}
424: \plotone{f2.eps}
425: \caption{Constraints on the geometric albedo of exoplanet TrES-3, averaged
426: over various bandpasses. Each curve shows the {\it a posteriori}\,
427: probability distribution determined with the MCMC algorithm.
428: \label{fig:2}}
429: \end{figure}
430:
431: \begin{deluxetable}{ccc}
432: \tabletypesize{\normalsize}
433: \tablecaption{Results of Occultation Photometry of TrES-3\label{tbl:params}}
434: \tablewidth{0pt}
435: %\tablewidth{4.5in}
436:
437: \tablehead{
438: \colhead{Bandpass} &
439: \multicolumn{2}{c}{Upper Limit (99\% confidence)} \\
440: \colhead{} &
441: \colhead{Planet-to-star flux ratio} &
442: \colhead{Geometric albedo}
443: }
444:
445: \startdata
446: $i$ & $0.00024$ & $0.30$ \\
447: $z$ & $0.00050$ & $0.62$ \\
448: $R$ & $0.00086$ & $1.07$
449: \enddata
450:
451: \end{deluxetable}
452:
453: \section{Discussion and Summary}
454:
455: This study has resulted in empirical upper limits on the
456: planet-to-star flux ratio, and the geometric albedo, in the wavelength
457: range 0.6$-$1.0~$\mu$m. The most stringent upper limit was obtained in
458: the Sloan $i$ band, where $p_i < 0.30$ with 99\% confidence. To put
459: this result in perspective, we remind the reader that a
460: perfectly-diffusing sphere has a geometric albedo of $2/3$. The TrES-3
461: planet is less reflective than such a sphere, at least for
462: observations in the $i$ and $z$ bands. The bandpass-averaged geometric
463: albedos of Jupiter in the $R$, $i$, and $z$ bands are approximately
464: 0.50, 0.45, and 0.25, respectively (Karkoschka 1994). Our results show
465: that TrES-3 is less reflective than Jupiter in the $i$ band, although
466: they do not enable a meaningful comparison in the other two
467: bands. Firm comparisons to other exoplanets are also difficult at this
468: stage because almost all published reports are upper limits. Using the
469: spectroscopic method, Charbonneau et al.~(1999), Collier Cameron et
470: al.~(2002), and Leigh et al.~(2003a,b) have placed upper limits of
471: 0.1--0.4 on various systems, although in those cases the results are
472: subject to extra uncertainty because the planetary radius and orbital
473: inclinations are not known. Rowe et al.~(2007) used several months of
474: spaceborne photometry of HD~209458 to conclude $p_{\rm opt} <0.17$ for
475: that system with 3$\sigma$ (99.73\%) confidence, over an optical
476: bandpass ranging from 0.4--0.7~$~\mu$m. At the same confidence level
477: of 99.73\%, our upper limit in the Sloan $i$ band is $p_i < 0.35$. As
478: mentioned in \S~1, Berdyugina et al.~(2007) found a surprisingly large
479: planet-to-star flux ratio for HD~189733, indeed large enough that
480: those authors suggested the scattering radius is larger than the
481: transit radius. Our results do not permit such an interpretation for
482: TrES-3.
483:
484: Theoretical reflection spectra for strongly irradiated giant planets
485: such as TrES-3 depend upon many factors and are especially sensitive
486: to the presence or absence of clouds. Without highly reflective
487: clouds, they are generally expected to have very low optical albedos.
488: The underlying cause of the low albedo in cloud-free models is strong
489: absorption in the pressure-broadened resonance doublet transitions of
490: sodium and potassium (see, e.g., Seager et al.~2000, Marley et
491: al.~1999, Sudarsky et al.~2000) or, for the hottest planets, the
492: plentiful electronic transitions of gaseous TiO and VO (Hubeny et
493: al.~2003, Sharp \& Burrows 2007, Fortney et al.~2007). Condensates may
494: form reflective clouds that can drastically raise the albedo. The
495: particular condensates that are available will depend on the
496: temperature and pressure profile of the atmosphere.
497:
498: As a specific example, Sudarsky et al.~(2000, 2003) predicted that gas
499: giants with effective temperatures between about 900~K and 1500~K
500: (``Class IV'' in their terminology) have optical albedos $\lsim 0.2$
501: in the bandpasses we employed, while for temperatures $\gsim 1500$~K
502: (``Class V''), the albedo is as large as 0.6. The difference is that
503: the hotter planets have a high layer of opaque silicate and iron
504: clouds in their upper atmospheres. The case of TrES-3 is borderline,
505: with an effective temperature of 1643~$f^{1/4} (1-A)^{1/4}$~K
506: (O'Donovan et al.~2007), where $A$ is the Bond albedo (the ratio of
507: reflected to incident bolometric power) and $f$ is a phenomenological
508: factor accounting for atmospheric circulation (with $f=1$ for
509: isotropic reradiation and $f=2$ for radiation from the dayside
510: only). For TrES-3 it would be possible to obtain a self-consistent
511: Class~V solution with highly-reflective silicate and iron clouds,
512: e.g., $f=1$ and $A\approx 0.4$, giving $T_{\rm eff}=1500$~K. Our upper
513: limit on $p_i$ rules this out, and hence may be regarded as evidence
514: against the presence of highly reflective clouds in the planet's upper
515: atmosphere.
516:
517: For cloud-free atmospheres, the thermal radiation from the planet is
518: expected to produce a larger signal than the reflected light (see,
519: e.g., L\'opez-Morales \& Seager 2007, Fortney et al.~2007). One might
520: wonder whether our limits on the planet-to-star flux ratio also
521: constrain the thermal emission spectrum of the planet. The thermally
522: emitted flux can be estimated from the effective temperature of the
523: planet. For TrES-3, even for the hottest possible temperature of
524: 2100~K (for $f=2$ and $A=0$) we estimate that the maximum
525: planet-to-star flux ratio due solely to thermal emission is $2\times
526: 10^{-4}$ at a wavelength of 0.9~$\mu$m. This rough estimate agrees
527: within a factor of 2 with a more detailed atmospheric model of TrES-3
528: by Fortney et al.~(2007). Those investigators classify TrES-3 as a pM
529: planet\footnote{The distinction between pM and pL planets, in the
530: nomenclature proposed by Fortney et al.~(2007), is analogous to the
531: M/L transition in low-mass stars and brown dwarfs. The pM planets
532: are hot enough for TiO and VO to exist in gaseous form, leading to
533: strong absorption of the stellar flux at low pressure, a temperature
534: inversion in the planet's upper atmosphere, and a brightness
535: temperature in excess of the equilibrium temperature in the optical
536: and infrared bands.}, for which they predict a very low albedo
537: (because conditions are too hot for any condensates) and an
538: ``anomalously hot'' brightness temperature (because of a high-altitude
539: temperature inversion). For our observing bandpass the calculated
540: planet-to-star flux ratio is $\sim$$10^{-4}$. This is smaller than our
541: most constraining upper limit of $2.4\times 10^{-4}$ in the $i$-band,
542: and hence we conclude that the expected level of thermal emission is
543: beneath our detection limit by a factor of 2--3.
544:
545: In summary, through high-precision photometry of an especially
546: favorable target, we have placed meaningful upper limits on the albedo
547: of a transiting exoplanet. This is the first time this has been done
548: with ground-based occultation photometry. Our results seem to be
549: limited by random noise, suggesting that the acquisition of more data
550: will lead to increased sensitivity. For example, with 8 more light
551: curves comparable in quality to the two $i$-band light curves
552: presented here, the photon-limited $1\sigma$ error in the albedo would
553: be approximately 0.03. Even if the albedo proves to be very small, as
554: expected in cloud-free atmospheric models, such a data set would
555: enable a $\sim$4$\sigma$ detection of the expected level of thermal
556: emission. Of course, it is possible that sources of systematic noise
557: will become limiting factors, such as flat-fielding errors, or
558: time-variable differential extinction beyond a gradual airmass
559: dependence. Nevertheless, our results give reason to hope for an
560: unambiguous detection of reflected light or thermal emission from
561: exoplanets based on ground-based photometry using meter-class
562: telescopes.
563:
564: \acknowledgments We thank Jonathan Fortney, Willie Torres, and Alex
565: Sozzetti for helpful conversations and correspondence; John
566: Southworth, for his code for computing limb-darkening coefficients;
567: and an anonymous referee for a timely and detailed review. This
568: research was supported by Grant No.~2006234 from the United
569: States-Israel Binational Science Foundation (BSF), Jerusalem,
570: Israel. KeplerCam was developed with partial support from the {\it
571: Kepler}\, mission under NASA Cooperative Agreement NCC2-1390
572: (PI:~D.~Latham).
573:
574: \begin{thebibliography}{}
575:
576: \bibitem[Barman et al.(2001)]{2001ApJ...556..885B} Barman, T.~S.,
577: Hauschildt, P.~H., \& Allard, F.\ 2001, \apj, 556, 885
578:
579: \bibitem[Berdyugina et al.(2008)]{2008ApJ...673L..83B} Berdyugina,
580: S.~V., Berdyugin, A.~V., Fluri, D.~M., \& Piirola, V.\ 2008, \apjl,
581: 673, L83
582:
583: \bibitem[Charbonneau et al.(1999)]{1999ApJ...522L.145C} Charbonneau,
584: D., Noyes, R.~W., Korzennik, S.~G., Nisenson, P., Jha, S., Vogt,
585: S.~S., \& Kibrick, R.~I.\ 1999, \apjl, 522, L145
586:
587: \bibitem[Claret(2000)]{2000A&A...363.1081C} Claret, A.\ 2000, \aap,
588: 363, 1081
589:
590: \bibitem[Claret(2004)]{2004A&A...428.1001C} Claret, A.\ 2004, \aap,
591: 428, 1001
592:
593: \bibitem[Collier Cameron et al.(1999)]{1999Natur.402..751C} Cameron,
594: A.~C., Horne, K., Penny, A., \& James, D.\ 1999, \nat, 402, 751
595:
596: \bibitem[Collier Cameron et al.(2002)]{2002MNRAS.330..187C} Collier
597: Cameron, A., Horne, K., Penny, A., \& Leigh, C.\ 2002, \mnras, 330,
598: 187
599:
600: \bibitem[Fortney et al.(2007)]{2007arXiv0710.2558F} Fortney, J.~J.,
601: Lodders, K., Marley, M.~S., \& Freedman, R.~S.\ 2007, ArXiv
602: e-prints, 710, arXiv:0710.2558
603:
604: \bibitem[Guillot et al.(1996)]{1996ApJ...459L..35G} Guillot, T.,
605: Burrows, A., Hubbard, W.~B., Lunine, J.~I., \& Saumon, D.\ 1996,
606: \apjl, 459, L35
607:
608: \bibitem[Holman et al.(2007)]{2007ApJ...664.1185H} Holman, M.~J., et
609: al.\ 2007, \apj, 664, 1185
610:
611: \bibitem[Hough et al.(2006)]{2006SPIE.6269E..25H} Hough, J.~H., Lucas,
612: P.~W., Bailey, J.~A., Tamura, M., \& Hirst, E.\ 2006, \procspie,
613: 6269, 62690S-2
614:
615: \bibitem[Hubeny et al.(2003)]{2003ApJ...594.1011H} Hubeny, I.,
616: Burrows, A., \& Sudarsky, D.\ 2003, \apj, 594, 1011
617:
618: \bibitem[Irwin et al.(2008)]{2008arXiv0801.1496I} Irwin, J., et al.\
619: 2008, ArXiv e-prints, 801, arXiv:0801.1496
620:
621: \bibitem[Karkoschka(1994)]{1994Icar..111..174K} Karkoschka, E.\ 1994,
622: Icarus, 111, 174
623:
624: \bibitem[Leigh et al.(2003a)]{2003MNRAS.346L..16L} Leigh, C., Collier
625: Cameron, A., Udry, S., Donati, J.-F., Horne, K., James, D., \&
626: Penny, A.\ 2003a, \mnras, 346, L16
627:
628: \bibitem[Leigh et al.(2003b)]{2003MNRAS.344.1271L} Leigh, C., Cameron,
629: A.~C., Horne, K., Penny, A., \& James, D.\ 2003b, \mnras, 344, 1271
630:
631: \bibitem[Liu et al.(2008)]{2007arXiv0711.2304L} Liu, X., et al.\ 2008,
632: ArXiv e-prints, 711, arXiv:0711.2304
633:
634: \bibitem[L{\'o}pez-Morales \& Seager(2007)]{2007ApJ...667L.191L}
635: L{\'o}pez-Morales, M., \& Seager, S.\ 2007, \apjl, 667, L191
636:
637: \bibitem[Mandel \& Agol(2002)]{2002ApJ...580L.171M} Mandel, K., \&
638: Agol, E.\ 2002, \apjl, 580, L171
639:
640: \bibitem[Marley et al.(1999)]{1999ApJ...513..879M} Marley, M.~S.,
641: Gelino, C., Stephens, D., Lunine, J.~I., \& Freedman, R.\ 1999,
642: \apj, 513, 879
643:
644: \bibitem[O'Donovan et al.(2007)]{2007ApJ...663L..37O} O'Donovan,
645: F.~T., et al.\ 2007, \apjl, 663, L37
646:
647: \bibitem[Rasio et al.(1996)]{1996ApJ...470.1187R} Rasio, F.~A., Tout,
648: C.~A., Lubow, S.~H., \& Livio, M.\ 1996, \apj, 470, 1187
649:
650: \bibitem[Rowe et al.(2007)]{2007arXiv0711.4111R} Rowe, J.~F., et al.\
651: 2007, ArXiv e-prints, 711, arXiv:0711.4111
652:
653: \bibitem[Saumon et al.(1996)]{1996ApJ...460..993S} Saumon, D.,
654: Hubbard, W.~B., Burrows, A., Guillot, T., Lunine, J.~I., \&
655: Chabrier, G.\ 1996, \apj, 460, 993
656:
657: \bibitem[Seager \& Sasselov(1998)]{1998ApJ...502L.157S} Seager, S., \&
658: Sasselov, D.~D.\ 1998, \apjl, 502, L157
659:
660: \bibitem[Seager et al.(2000)]{2000ApJ...540..504S} Seager, S.,
661: Whitney, B.~A., \& Sasselov, D.~D.\ 2000, \apj, 540, 504
662:
663: \bibitem[Sharp \& Burrows(2007)]{2007ApJS..168..140S} Sharp, C.~M., \&
664: Burrows, A.\ 2007, \apjs, 168, 140
665:
666: \bibitem[Southworth(2008)]{2008arXiv0802.3764S} Southworth, J.\ 2008,
667: ArXiv e-prints, 802, arXiv:0802.3764
668:
669: \bibitem[Sudarsky et al.(2000)]{2000ApJ...538..885S} Sudarsky, D.,
670: Burrows, A., \& Pinto, P.\ 2000, \apj, 538, 885
671:
672: \bibitem[Sudarsky et al.(2003)]{2003ApJ...588.1121S} Sudarsky, D.,
673: Burrows, A., \& Hubeny, I.\ 2003, \apj, 588, 1121
674:
675: \bibitem[Szentgyorgyi et al.(2005)]{2005AAS...20711010S} Szentgyorgyi,
676: A.~H., et al.\ 2005, Bulletin of the American Astronomical Society,
677: 37, 1339
678:
679: \bibitem[Torres et al.(2008)]{2008arXiv0801.1841T} Torres, G., Winn,
680: J.~N., \& Holman, M.~J.\ 2008, ArXiv e-prints, 801, arXiv:0801.1841
681:
682: \bibitem[Winn et al.(2007)]{2007AJ....134.1707W} Winn, J.~N., et al.\
683: 2007, \aj, 134, 1707
684:
685: \bibitem[Young(1967)]{1967AJ.....72..747Y} Young, A.~T.\ 1967, \aj,
686: 72, 747
687:
688: \end{thebibliography}
689:
690: \end{document}
691: