1: 

2: \documentclass[12pt,preprint]{aastex}

3: 

4: %% manuscript produces a one-column, double-spaced document:

5: 

6: %\documentclass[manuscript]{aastex}

7: 

8: %% preprint2 produces a double-column, single-spaced document:

9: 

10: %\documentclass[preprint2]{aastex}

11: 

12: %% \documentclass[preprint2,longabstract]{aastex}

13: 

14: %% If you are submitting to a journal that translates manuscripts

15: %% into SGML, you need to follow certain guidelines when preparing

16: %% your macros. See the AASTeX v5.x Author Guide

17: %% for information.

18: 

19: \newcommand{\vdag}{(v)^\dagger}

20: \newcommand{\myemail}{skywalker@galaxy.far.far.away}

21: 

22: 

23: \begin{document}

24: 

25: %% LaTeX will automatically break titles if they run longer than

26: %% one line. However, you may use \\ to force a line break if

27: %% you desire.

28: 

29: \title{A {\it Spitzer} Spectrum of the Exoplanet HD 189733b}

30: 

31: \author{C. J. Grillmair}

32: \affil{Spitzer Science Center, 1200 E. California Blvd., Pasadena,  CA 91125}

33: \email{carl@ipac.caltech.edu}

34: \author{D. Charbonneau\altaffilmark{1}}

35: \affil{Harvard-Smithsonian Center for Astrophysics, 60 Garden St.,

36:   MS-16, Cambridge, MA 02138}

37: \email{dcharbon@cfa.harvard.edu}

38: \author{A. Burrows}

39: \affil{Steward Observatory, University of Arizona, Tucson, AZ 85721}

40: \email{burrows@.as.arizona.edu}

41: \author{L. Armus}

42: \affil{Spitzer Science Center, 1200 E. California Blvd., Pasadena,  CA 91125}

43: \email{lee@ipac.caltech.edu}

44: \author{J. Stauffer}

45: \affil{Spitzer Science Center, 1200 E. California Blvd., Pasadena,  CA 91125}

46: \email{stauffer@ipac.caltech.edu}

47: \author{V. Meadows}

48: \affil{Spitzer Science Center, 1200 E. California Blvd., Pasadena,  CA 91125}

49: \email{vsm@ipac.caltech.edu}

50: \author{J. Van Cleve}

51: \affil{Ball Aerospace \& Technologies Corp., PO Box 1062, Boulder, CO 80306}

52: \email{jvanclev@ball.com}

53: \and

54: \author{D. Levine}

55: \affil{Spitzer Science Center, 1200 E. California Blvd., Pasadena,  CA 91125}

56: \email{deblev@ipac.caltech.edu}

57: 

58: \altaffiltext{1}{Alfred P. Sloan Research Fellow}

59: 

60: \begin{abstract}

61: 

62: We report on the measurement of the 7.5-14.7 micron spectrum for the

63: transiting extrasolar giant planet HD 189733b using the Infrared

64: Spectrograph on the Spitzer Space Telescope. Though the observations

65: comprise only 12 hours of telescope time, the continuum is well

66: measured and has a flux ranging from 0.6 mJy to 1.8 mJy over the

67: wavelength range, or $0.49 \pm 0.02\%$ of the flux of the parent star.

68: The variation in the measured fractional flux is very nearly flat over

69: the entire wavelength range and shows no indication of significant

70: absorption by water or methane, in contrast with the predictions of

71: most atmospheric models.  Models with strong day/night differences appear

72: to be disfavored by the data, suggesting that heat redistribution to

73: the night side of the planet is highly efficient.

74: 

75: 

76: \end{abstract}

77: 

78: %% Keywords should appear after the \end{abstract} command. The uncommented

79: %% example has been keyed in ApJ style. See the instructions to authors

80: %% for the journal to which you are submitting your paper to determine

81: %% what keyword punctuation is appropriate.

82: 

83: \keywords{Stars: --- Planetary Systems, Stars: Binaries: Eclipsing,

84: Stars: Individual (HD 189733)}

85: 

86: \section{Introduction}

87: 

88: HD 189733 is a nearby K0V star with a magnitude of K = 5.5 and an

89: expected flux of 250 mJy at 10 microns.  HD 189733b was initially

90: discovered in the ELODIE search for transiting exoplanets

91: \citep{bouchy2005} and subsequently detected in existing Hipparcos

92: data \citep{hebrard2006}. In addition to HD 189733b, HD 189733 is

93: orbited by an M dwarf at a distance of 216 AU \citep{bakos2006a}. HD

94: 189733b orbits its parent star in 2.2 days and has an estimated radius

95: of $R_p = 1.154 \pm 0.033 R_J$ \citep{bakos2006b}. From first to last

96: contact, the duration of the secondary eclipse is $\approx 1.9$ hours.

97: 

98: The planet-to-star surface area ratio for HD 189733b is $\approx

99: 0.024$ \citep{bakos2006b}.  Owing to an orbital radius of 0.031 AU,

100: the upper atmospheric temperature of the planet is $\geq 1000$ K

101: \citep{fortney2006a}. Together these factors yield an expected

102: planet-to-star flux ratio that is almost a factor of two greater at 10

103: microns than for either TrES-1 or HD 209458b, two other exoplanets

104: whose thermal emission has recently been detected by the Spitzer Space

105: Telescope \citep{charbonneau2005, deming2005}.  This expectation has

106: been borne out by \citet{deming2006}, who used the Peak-up Imaging

107: array portion of {\it Spitzer's} Infrared Spectrograph (IRS) to

108: measure a secondary eclipse depth at $16 \micron$ of $0.551 \pm

109: 0.030\%$.  This makes HD 189733b a prime candidate for IRS

110: spectroscopy and a more detailed characterization of its physical

111: properties.

112: 

113: We briefly describe the observations in Section \ref{observations}

114: and the analysis in Section \ref{analysis}. We discuss the

115: spectrum of HD 189733b in Section \ref{discussion}. 

116: 

117: 

118: \section{Observations \label{observations}}

119: 

120: HD 189733 was observed with the Infrared Spectrograph (IRS) on October

121: 21st and November 21st, 2006 as part of General Observer program

122: \#30473. In each case the star was observed for a six-hour span

123: centered on the predicted time of secondary eclipse.  All observations

124: were made in first order ($\lambda = 7.4-14.5 \micron$) using the

125: Short-Low module, giving a spectral resolution of $\approx 80$.  The

126: performance and capabilities of the Short-Low module, which has the

127: cleanest and most sensitive of the IRS detectors, is discussed in

128: detail by \citet{houck2004}.

129: 

130: The Oct 21st observations were carried out in ``Staring'' mode,

131: wherein the telescope was periodically repointed to place the target

132: alternately at two dither positions along the slit.  Alternating

133: between two slit positions provided some insurance against the

134: possible large buildup of latent charge (see below). The observations

135: were preceded by a high-accuracy peak-up approximately 3 hours before

136: secondary eclipse and were taken continuously through the

137: out-of-eclipse, ingress, secondary eclipse, egress, and out-of-eclipse

138: portions of HD 189733b's orbit. Thirty successive 14-second exposures

139: were taken at each slit position before moving to the next dither

140: position. A total of 30 telescope moves were commanded, yielding 450

141: independent integrations at each dither position over the course of

142: six hours.

143: 

144: A preliminary analysis of the Oct 21st observations showed that charge

145: latents were considerably more benign than initial estimates.  An

146: observing sequence scheduled for Nov 21st was consequently modified to

147: place the target at the center of the Short-Low slit for the entire

148: observing sequence. Small pointing errors associated with any

149: commanded telescope movement (which contribute significantly to the

150: noise budget) were thus avoided. Having thereby also eliminated

151: telescope slew and settle times, we were able to fit a total 950

152: 14-second exposures into the six hours available.

153: 

154: 

155: \section{Analysis \label{analysis}}

156: 

157: Our analysis makes use of the two-dimensional Basic Calibrated Data

158: products from the S14 version of the IRS pipeline. Using Version 1.4

159: of the SPICE spectral extraction tool, we apply the optimal

160: extraction feature to extract one-dimensional spectra for each of the

161: 1850 exposures separately. We fix the center of each extraction window

162: to take advantage of the pointing performance of the telescope and

163: avoid adding noise associated with a measurement of the

164: sub-pixel position of the star for each exposure.

165: 

166: For reasons that will become clear, we treat the integrations at

167: the two dither positions on Oct 21st and the single-pointing

168: integrations taken on Nov 21st as three separate sets of

169: observations. In principal, it should be possible to simply coadd all

170: integrations in each observation set taken during eclipse (starlight

171: only) and subtract them from a coaddition of all spectra taken out of

172: eclipse (star plus planet). In practice, due to the limited

173: integration time available when the planet is in eclipse, several

174: systematics can manifest themselves as either increased noise or

175: spectral offsets.

176: 

177: The Spitzer Space Telescope's stationary pointing performance is not

178: perfect. Due to a small, cyclic heat source within the spacecraft

179: structure, there is a periodic variation in alignment between the Star

180: Tracker assembly and the telescope boresight. This leads to a pointing

181: oscillation with a period of about an hour and a magnitude of $\sim

182: 0.05\arcsec$ \citep{morales2006}. Though small, this oscillation has

183: significant consequences for observations that require precision on

184: the order of 0.1\%. In addition to variations in the total

185: amount of light entering the slit, we are subject to significant signal

186: modulation due to drifts across pixels with different (and poorly

187: calibrated) sensitivities.

188: 

189: We analyze each wavelength bin as an independent time series and then

190: combine the results to produce a final ratio spectrum. As an example,

191: we show in Figure 1 the measured flux as a function of time at dither

192: position 1 for the wavelength bin centered on $7.93 \micron$ . Several

193: characteristics that affect all wavelength bins to a greater or

194: lesser degree are in evidence: 1) There is a rapid rise in the

195: counts during the first 10 minutes of the sequence that is clearly

196: related to latent charge buildup. 2) There is a longer term variation

197: that is probably related to both the buildup and decay of latent

198: charge, as well as to possible drifts in and out of the slit. 3) There

199: is a cyclic behavior due to the pointing oscillations.

200: 

201: We have found that the buildup and decay of latent charge can vary

202: greatly from one pixel to the next. Rather than trying to model the

203: ensemble, short-term latent behavior for each wavelength bin, we

204: simply exclude the first 10 minutes of data from our analysis. Since

205: our final signal-to-noise ratio is determined primarily by

206: data taken during secondary eclipse, the loss of the first 10 minutes

207: of data has negligible consequences for the final result.  We fit the

208: longer-term variation with a 3rd-order polynomial and then divide it

209: out to flatten the time series. We find this to be the minimum order

210: necessary to reproduce the observed drifts while leaving the periodic

211: variations largely unaffected.

212: 

213: Observations with the InfraRed Array Camera (IRAC) indicate that the

214: Spitzer Space Telescope's pointing oscillation behaves rather more

215: like a sawtooth than a sinusoid \citep{morales2006}.  We indeed find

216: that an asymmetric sawtooth best fits the cyclic time behavior in the

217: data, and that the shape of the sawtooth is fairly consistent from one

218: slit position to the next. We find that the phase and asymmetry of the

219: flux modulation (which depend on both the telescope motion and the

220: distribution of imperfectly calibrated pixels) are essentially

221: constant for all wavelength bins (modulo a $180\arcdeg$ phase flip),

222: and we adopt a fixed modulation pattern, with only the amplitude of

223: the oscillation being a free parameter. The sawtooth that best fits

224: the data for dither position 1 is shown over plotted on the

225: drift-corrected data in the middle panel of Figure 1.

226: 

227: After dividing the data by the best fitting periodic function, we fit

228: the data using the ephemerides and the secondary eclipse profile of

229: \citet{bakos2006b}. The uncertainty in the predicted time of the

230: center of the secondary eclipse is 5.4 minutes assuming a circular

231: orbit. The ephemerides we have adopted are consistent with updated

232: values recently published by \citet{winn2007}. The timing

233: uncertainty has been reduced to 0.5 minutes and is not a significant source

234: of uncertainty in our analysis. In fitting the light curve, the two

235: free parameters are the mean flux level outside of eclipse and the

236: scaling of Bakos et al.'s light curve required to match the in-eclipse

237: data. This allows us to use data taken during ingress and egress as

238: well as during totality. We optimize the fit by varying the amplitude

239: of the periodic function and minimizing $\chi^2$ during the in-eclipse

240: portion of HD 189733b's orbit. Due to its greater extent, the mean

241: level of the time series outside of eclipse is largely unaffected by

242: changes in the periodic function, and can be independently measured to

243: high precision.

244: 

245: To constrain the non-periodic behavior of pixels in each wavelength

246: bin over the course of six hours, we first divide the pre-correction,

247: in-eclipse data with the light curve of \citet{bakos2006b}.  We

248: find that simply fitting to all data with $t < -2500$ s and $t > 2500$

249: s is not sufficient to constrain long term drifts during eclipse and

250: adds noticeably to the noise in the final spectrum.  Bakos et al.'s

251: (2006) light curve is initially scaled so that the upward correction

252: at eclipse center is 0.5\%.  We then iterate on the final solution for

253: each wavelength bin by rescaling the correction based on the previous

254: flux estimate. The solution converges rapidly in all cases, and we

255: halt the calculation when the difference in computed eclipse depth

256: between successive iterations becomes less than 0.002\%.  Changing the

257: initial scaling of the light curve by a factor of two up or down has

258: no effect on the final spectrum.

259: 

260: Given a perfectly stable detector and platform, we would normally

261: expect to be dominated by photon noise over the entire wavelength

262: range for a star as bright as HD 189733.  After making the corrections

263: described above, the RMS spread of the individual measurements about

264: the best fit light curve are indeed within 20\% of the expected photon

265: noise limit over most of the wavelength range. The RMS spread in the

266: individual measurements exceeds what one would expect just from photon

267: noise by up to 50\% in the regions $9-10 \micron, 11.5-12.5 \micron$,

268: and $13.7-14.5 \micron$ for both the Oct 21st and Nov 21st observations.

269: Conversely, the regions $8-9 \micron, 10-11 \micron$, and $12.5-13.5

270: \micron$ show an RMS spread that is consistent with pure photon

271: statistics. This suggests that there are remaining, unmodeled temporal

272: variations on the order of 0.5\%, due perhaps to fringe motions and

273: subpixel variations in responsivity.

274: 

275: In Figure 2 we show the HD 189733b/HD 189733 flux ratios for each of

276: the three separately analyzed data sets.  The plotted uncertainties

277: reflect only the light curve fitting errors and are smallest for the

278: Nov 21st observations by virtue of their greater number and continuous

279: coverage through secondary eclipse. 

280: 

281: 

282: \section{Discussion \label{discussion}}

283: 

284: 

285: 

286: HD 189733b's continuum is well measured and, multiplying the relative

287: fluxes in Figure 2 by the observed spectrum of HD 189733, we obtain

288: absolute fluxes ranging from 1.8 mJy at $7.8 \micron$ to 0.5 mJy at

289: $14.7 \micron$. Our measurement is consistent with the 660 $\mu$Jy

290: integrated flux measurement at $16 \micron$ by \citet{deming2006}.

291: Comparing the three spectra to one another, the signal-to-noise ratio

292: per resolution element ranges from about 15 at 10 microns to about 6

293: at 14.5 microns.

294: 

295: The flux ratios in Figure 2 are essentially flat; the emission

296: spectrum of HD 189733b over the wavelength range 7.5 to $14.7 \micron$

297: is evidently consistent with that of a blackbody. The integrated flux

298: ratios are 0.50\%, 0.51\%, and 0.46\%, respectively, yielding a mean

299: of $0.49 \pm 0.02\%$. Though the mean relative flux is in good agreement,

300: the spectrum taken at dither position 1 appears somewhat problematic,

301: with a number of peaks and troughs that are not seen at the other

302: slit positions. Of the three, the spectrum taken at slit center is

303: clearly the highest quality measurement and demonstrates the utility

304: of keeping the telescope as motionless as possible for observations of

305: this type. While there are interesting similarities between the

306: spectra taken at dither position 2 and at slit center, the differences

307: are sufficiently large and numerous that we cannot identify individual

308: absorption or emission features with any confidence.

309: 

310: In Figure 3 we compare the relative flux spectrum measured at slit

311: center with a model of HD 189733b from \citet{burrows2006}. The model

312: shown is for superior conjunction, with no clouds and 50\% energy

313: redistribution to the night side of the planet.  The mean levels for

314: the model and the observations are quite similar. Redward of $10

315: \micron$, where the flux ratio is predicted to be fairly flat, the

316: model predicts a mean relative flux which is about 20\% higher than

317: that of the observations.  Near superior conjunction, less atmospheric

318: energy redistribution would push the relative planet/star flux still

319: higher.  In contrast to the findings of \citet{harrington2006} for

320: $\upsilon$ And b, it appears that strong day/night differences on HD

321: 189733b are disfavored by the data.

322: 

323: Blueward of 8.2 microns, the model predicts a quite significant drop

324: in the relative flux due to water absorption. In contrast, the

325: observed flux ratios continue unabated to the blue end of the

326: spectrum. This difference is significant; scaling

327: \citet{burrows2006}'s model to match the data beyond $9.5 \micron$ and

328: summing over all bins blueward of 8.2 microns, the model lies $\approx

329: 9\sigma$ below the data. The measured flux at the blue end is not the

330: product of a few bad measurements, but is reflected separately in each

331: of the three individual spectra, and two independent analyses. It is

332: also seen in simple subtractions of combined out-of-eclipse and

333: in-eclipse spectra.

334: 

335: A reduced planet-to-star ratio blueward of $10\micron$ due to water

336: opacity is a near universal prediction in published models of hot

337: Jupiters to date (e.g. \citet{fortney2005, barman2005, seager2005,

338: burrows2006}), and hence the absence of this signature from our data

339: is intriguing. Simply reducing the amount of water in the models may

340: not, by itself, bring them into agreement with the data.  For example,

341: our observations also appear to disagree with the low H$_2$O, high

342: carbon to oxygen ratio spectrum computed for HD 209458b by

343: \citet{seager2005}. In this case a similar drop in the flux blueward

344: of 9 microns is expected due to methane absorption. We note that

345: \citet{fortney2006b} have recently considered dynamical models of hot

346: Jupiter atmospheres \citep{cooper2006} to predict emergent spectra as

347: a function of orbital phase, and have shown that spectral features in

348: day-side spectra may be suppressed due to an isothermal

349: pressure-temperature profile. A more thorough analysis of the

350: consequences of this spectrum for models of giant planet atmospheres

351: is forthcoming.

352: 

353: 

354: 

355: The detection of HD 189733b's spectral continuum, accomplished with

356: only 12 hours of Spitzer time, is a remarkable demonstration of the

357: capability and utility of the Spitzer Space Telescope and the Infrared

358: Spectrograph. Integrating over many eclipses, we can expect

359: substantial increases in spectral signal-to-noise ratio, broader

360: wavelength coverage, and the detection of atmospheric constituents for

361: both HD 189733b and other transiting extrasolar giant planets. While

362: Spitzer's cryogen lasts, we can look forward to an exciting era of

363: quantitative, spectral characterization of extrasolar planets, and

364: ultimately a deeper understanding of planetary atmospheres under

365: strong irradiation.

366: 

367: 

368: \acknowledgments

369: 

370: 

371: We are grateful to an anonymous referee for several suggestions which

372: significantly improved the analysis and presentation of this

373: work. This work is based on observations made with the {\it Spitzer

374: Space Telescope}, which is operated by the Jet Propulsion Laboratory,

375: California Institute of Technology under a contract with NASA. Support

376: for this work was provided by NASA through an award issued by

377: JPL/Caltech. This study was supported in part by NASA grant NNGO4GL22G

378: and through the NASA Astrobiology Institute under Cooperative

379: Agreement No. CAN-02-OSS-02 issued through the Office of Space

380: Science.

381: 

382: 

383: 

384: \begin{thebibliography}{}

385: 

386: \bibitem[Bakos et al. (2006a)]{bakos2006a} Bakos, G. A., Pal, A.,

387:   Latham, D. W., Noyes, R. W., \& Stefanik, R. P. 2006a, \apjl, 641, 57

388: 

389: \bibitem[Bakos et al. (2006b)]{bakos2006b} Bakos, G. A., Knutson, H.,

390:   Pont, F., Moutou, C., Charbonneau, D., Shporer, A., Bouchy, F.,

391:   Everett, M., Hergenrother, C., Latham, D. W., Mayor, M., Mazeh, T.,

392:   Noyes, R. W., Queloz, D., Pal, A., \& Udry, S. 2006b, \apj, 650, 1160

393: 

394: \bibitem[Barman et al.(2005)]{barman2005} Barman, T.~S., Hauschildt,

395: P.~H., \& Allard, F.\ 2005, \apj, 632, 1132

396: 

397: \bibitem[Bouchy et al. (2005)]{bouchy2005} Bouchy, F., Udry, S.,

398:   Mayor, M., Moutou, C., Pont, F., Iribarne, N., Da Silva, R.,

399:   Ilovaisky, S., Queloz, D., Santos, N. C., Segransan, D., \& Zucker,

400:   S. 2005, \aap, 444, L15

401: 

402: \bibitem[Burrows, Sudarsky, \& Hubeny (2006)]{burrows2006} Burrows,

403:   A., Sudarsky, D., \& Hubeny, I. 2006, \apj, 650, 1140

404: 

405: \bibitem[Charbonneau et al. (2005)]{charbonneau2005} Charbonneau, D.,

406:   Allen, L. E., Megeath, S. T., Torres, G., Alonso, R., Brown, T. M.,

407:   Gilliland, R. L., Latham, D. W., Mandushev, G., O'Donovan, F. T.,

408:   \& Sozzetti, A. 2005, \apj, 626, 523

409: 

410: \bibitem[Cooper \& Showman (2006)]{cooper2006} Cooper, C.~S., \&

411: Showman, A.~P.\ 2006, \apj, 649, 1048

412: 

413: \bibitem[Deming et al. (2005)]{deming2005} Deming, D., Seager, S., Richardson,

414:   L. J., \& Harrington, J. 2005, \nat, 434, 740

415: 

416: \bibitem[Deming et al. (2006)]{deming2006} Deming, D., Harrington, J.,

417:   Seager, S., \& Richardson, L. J. 2006, \apj, 644, 560

418: 

419: \bibitem[Fortney et al.(2005)]{fortney2005} Fortney, J.~J., Marley,

420: M.~S., Lodders, K., Saumon, D., \& Freedman, R.\ 2005, \apjl, 627, L69

421: 

422: \bibitem[Fortney et al. (2006a)]{fortney2006a} Fortney, J.~J., Saumon,

423:   D., Marley, M.~S., Lodders, K., Freedman, R.~S. 2006, \apj, 642, 495

424: 

425: \bibitem[Fortney et al.(2006b)]{fortney2006b} Fortney, J.~J.,

426: Cooper, C.~S., Showman, A.~P., Marley, M.~S., \& Freedman, R.~S.\ 2006,

427: \apj, 652, 746

428: 

429: \bibitem[Harrington et al. (2006)]{harrington2006} Harrington, J.,

430:   Hansen, B. M., Luszcz, S. H., Seager, S., Deming, D., Menou, K.,

431:   Cho, J. Y.-K., \& Richardson, L. J. 2006, {\it Science}, 314, 623

432: 

433: \bibitem[Hebrard \& Lecavelier Des Etangs (2006)]{hebrard2006}

434:   Hebrard, G., \& Lecavelier Des Etangs, A. 2006, \aap, 445, 341

435: 

436: \bibitem[Houck et al. (2004)]{houck2004} Houck, J. R. et al. 2004,

437:   \apjs, 154, 103

438: 

439: \bibitem[Morales-Calderon et al. (2006)]{morales2006}

440:   Morales-Calderon, M., Stauffer, J. R., Kirkpatrick, J. D., Carey,

441:   S., Gelino, C. R., Barrado y Navascues, D., Rebull, L., Lowrance,

442:   P., Marley, M.S., Charbonneau, D., Patten, B. M., Megeath, S. T., \&

443:   Buzasi, D. 2006, \apj, 653, 1454

444: 

445: \bibitem[Seager et al. (2005)]{seager2005} Seager, S., Richardson,

446:   L. J., Hansen, B. M. S., Menou, K., Cho, J. Y.-K., \& Deming,

447:   D. 2005, \apj, 632, 1122

448: 

449: \bibitem[Winn et al. (2007)]{winn2007} Winn, J.~N., et al. 2007, \aj,

450: in press (astro-ph/061224)

451: 

452: 

453: \end{thebibliography}

454: 

455: 

456: \clearpage

457: 

458: \begin{figure}

459: \epsscale{0.8}

460: \plotone{f1.eps}

461: \caption{The best-fit solution for the planetary flux in the

462:   wavelength bin centered on 7.93 microns, showing from bottom to top

463:   the corrections applied. The flux measurements are shown as a function

464:   of time from secondary eclipse. The upper and lower data

465:   sets have been vertically offset for clarity. The filled circles in

466:   the bottom plot show the uncorrected flux as extracted using

467:   SPICE. The asterisks highlight data points actually used to (1)

468:   constrain the long term drift, (2) fit the amplitude of the periodic

469:   function, and (3) fit the eclipse depth. To better constrain (1) and

470:   (2), the asterisked data have been corrected upwards during

471:   secondary eclipse using the predicted light curve of

472:   \citet{bakos2006b}. The first 10 minutes of data are not used as

473:   they are generally affected by rapid latent charge buildup. The

474:   middle set of data points shows the effect of dividing by a

475:   3rd-order polynomial, and over plotted is the sawtooth function that

476:   best fits the pointing-induced oscillations. The upper set of data

477:   points have been corrected for long term and periodic behavior, and

478:   the best fitting light curve is shown over plotted. The filled

479:   squares are each a mean of 30 successive data points and serve

480:   simply as an aid in visualizing the quality of the

481:   fit. \label{fig1}}

482: \end{figure}

483: 

484: \clearpage

485: 

486: \begin{figure}

487: \epsscale{1.0}

488: \plotone{f2.eps}

489: \caption{The flux of HD 189733b relative to that of HD 189733, as

490: measured at three different slit positions.  The spectra are

491: unsmoothed, and the plotted uncertainties correspond to the formal

492: fitting errors.}

493: \end{figure}

494: 

495: \clearpage

496: 

497: \begin{figure}

498: 

499: \plotone{f3.eps}

500: \caption{Comparison of the flux ratios measured at slit center (filled

501: circles) with a model of HD 189733b from \citet{burrows2006}. The

502: model shown is for superior conjunction, with no clouds and 50\%

503: energy redistribution to the night side of the planet. The upper panel

504: shows the model as published by \citet{burrows2006}, while in the

505: lower panel the model has been scaled to match the data at $\lambda >

506: 9.5 \micron$. Intriguingly, the observed spectrum does not show the

507: expected decrease in relative flux at short wavelengths due to

508: increasing water opacity.}

509: \end{figure}

510: 

511: 

512: \end{document}

513: 

514: %%

515: 

516: