astro-ph0609513/ms.tex
1: \documentclass[12pt, preprint]{aastex}
2: %% \documentclass[preprint2]{aastex}
3: \newcommand{\ms}{\mbox{m s$^{-1}~$}}
4: \newcommand{\msy}{\mbox{m s$^{-1}$ yr$^{-1}~$}}
5: \newcommand{\kms}{\mbox{km s$^{-1}~$}}
6: \newcommand{\msun}{M$_{\odot}$}
7: \newcommand{\rsun}{R$_{\odot}$}
8: \newcommand{\mjup}{M$_{\rm JUP}$}
9: \newcommand{\mearth}{M$_{\rm EARTH  }$}
10: \newcommand{\msat}{M$_{\rm SAT}$}
11: \newcommand{\rjup}{R$_{\rm JUP}~$}
12: \newcommand{\msini}{$M \sin i~$}
13: \newcommand{\teff}{\mbox{$T_{\rm eff}$}}
14: \newcommand{\teq}{\mbox{$T_{\rm eq}$}}
15: \newcommand{\tint}{\mbox{$T_{\rm int}$}}
16: \newcommand{\etal}{\mbox{\rm et al.~}}
17: \newcommand{\logg}{\mbox{$\log g$}}
18: \newcommand{\fe}{\mbox{[Fe/H]}}
19: \newcommand{\sini}{$sin i~$}
20: \newcommand{\vsini}{V$sin i~$}
21: \newcommand{\ab}{$\sim$}
22: \newcommand{\lessim}{\lower.5ex\hbox{$\; \buildrel < \over \sim \;$}}
23: \newcommand{\Mv}{M$_{\rm v}~$}
24: \newcommand{\simgt}{\lower.5ex\hbox{$\; \buildrel > \over \sim \;$}}
25: \newcommand{\simlt}{\lower.5ex\hbox{$\; \buildrel < \over \sim \;$}}
26: 
27: \received{}
28: \accepted{}
29: 
30: \slugcomment{Submitted to PASP; Revised July 25, 2006}
31: 
32: \shortauthors{Bozorgnia \etal}
33: \shorttitle{Search for an Atmosphere in HD149026}
34: \begin{document}
35: 
36: 
37: \title{Search for an Atmospheric Signature\\
38:      of the Transiting Exoplanet HD 149026b\altaffilmark{1}}
39: \author{Nassim Bozorgnia\altaffilmark{2},
40: Jonathan J. Fortney\altaffilmark{3,4,5}
41: Debra A. Fischer\altaffilmark{2},
42: Chris McCarthy\altaffilmark{2},
43: Geoffrey W. Marcy\altaffilmark{2,6}}
44: 
45: \altaffiltext{1}{Based on observations obtained at the
46: W. M. Keck Observatory, which is operated by the University of  
47: California
48: and the California Institute of Technology. Keck time has been
49: granted by NASA.}
50: \altaffiltext{2}{Department of Physics \& Astronomy, San Francisco State
51: University, San Francisco, CA  94132; fischer@stars.sfsu.edu}
52: \altaffiltext{3}{Space Science and Astrobiology Division, NASA Ames  
53: Research
54: Center, MS 245-3, Moffett Field, CA 94035; jfortney@arc.nasa.gov}
55: \altaffiltext{4}{Spitzer Fellow}
56: \altaffiltext{5}{SETI Institute, 515 North Whisman Road, Mountain  
57: View, CA 94043}
58: \altaffiltext{6}{Department of Astronomy, University of California,
59: Berkeley, CA USA 94720}
60: 
61: \begin{abstract}
62: 
63: HD149026b is a short-period, Saturn-mass planet that transits a metal- 
64: rich
65: star.  The planet radius, determined by photometry, is remarkably small
66: compared to other known transiting planets,
67: with a heavy element core that apparently comprises $\sim 70$\% of  
68: the total
69: planet mass.  Time series spectra were obtained at Keck before and
70: during transit in order to model the Rossiter-McLaughlin effect.
71: Here, we make use of these observations to carry out a differential
72: comparison of spectra obtained in and out-of-transit
73: to search for signatures of neutral atomic lithium and potassium from  
74: the planet atmosphere.  No signal was detected at the 2\% level; we  
75: therefore place upper limits on the column density of these atoms.
76: 
77: %We acknowledge support from NASA, grant number NNG056G164G to DF.
78: 
79: \end{abstract}
80: 
81: \keywords{planetary systems --- stars: individual (HD 149026) ---
82: transit effects}
83: 
84: 
85: \section{Introduction}
86: 
87: Extrasolar planets that transit their host stars offer a
88: unique opportunity to search for constituents in the planet atmosphere.
89: As stellar flux passes through the planet atmosphere, additional
90: absorption features will be imprinted on the (otherwise pure) stellar
91: spectrum (Seager \& Sasselov 2000, Brown 2001, Hubbard \etal 2001).   
92: This offers a possibility
93: for detection of atmospheric transmission features
94: when working differentially with in-transit and out-of-transit spectra. 
95: We have used this fact to search for additional absorption in alkali lines  
96: from the atmosphere of HD 149026b during transit. 
97: 
98: 
99: Transiting planets that reside close
100: to their host stars are strongly irradiated and have atmospheres
101: with temperatures (but not gravities) similar to those of late type
102: L dwarfs.  In the cool atmosphere of sub-stellar objects
103: like brown dwarfs and giant exoplanets, the more refractory metals
104: condense and rain out-of-the atmosphere leaving the less refractory
105: alkali metals in atomic form (Burrows \& Sharp 1999; Lodders 1999).
106: Thus, the strongest absorption features are expected for alkali
107: metals such as sodium and potassium
108: (Seager \& Sasselov 2000, Sudarsky, Burrows \& Pinto 2000).
109: 
110: Since the discovery of HD 209458b
111: (Charbonneau et al. 2000; Henry et al. 2000),
112: this star has been the subject of many ground-based and
113: space-based studies.  HD 209458 is a bright star with
114: $V=7.65$, permitting spectroscopy with high resolution and high S/N.
115: The transiting planet has an anomalously large radius of 1.3 \rjup,  
116: making
117: it an excellent candidate to search for atmospheric signatures. In  
118: contrast,
119: searching for atmospheric features in other transiting exoplanets
120: may be more difficult, because the planet to star radius is generally
121: smaller and the host stars have a lower intrinsic brightness.   
122: However, if
123: other exoplanets have tenuous, but extended atmospheres, then it may be
124: possible to identify atmospheric features from the planet.
125: 
126: The most successful searches for exoplanet atmospheres have been carried
127: out from space where telluric contamination is avoided and
128: higher S/N has enabled more sensitive searches.  Based on
129: observations with the \emph{Hubble Space Telescope} (HST),
130: Charbonneau \etal (2002) detected absorption from the sodium resonance
131: doublet at 5893 \AA\ in the dense lower atmosphere of HD 209458b, and
132: Vidal-Madjar \etal (2003) detected an escaping, trailing extended
133: exosphere of hydrogen.  In contrast, it is more challenging
134: to search for exoplanet atmospheres with ground-based telescopes.
135: Bundy \& Marcy (2000) carried out a differential analysis of spectra
136: for HD 209458 obtained in and
137: out-of-transit at the Lick and Keck Observatories.  They also studied  
138: the
139: spectra of 51 Peg, which is not a transiting planet.  Because of
140: contamination from ${\rm H_2O}$ and ${\rm O_2}$ features, they were only
141: able to place weak upper limits on Na I and K I line depth variations.
142: Deming et al. (2005) searched for carbon monoxide
143: absorption features in the transmission spectrum of HD 209458b
144: at $\sim2.3$ $\micron$ using NIRSPEC on Keck II telescope.  Their  
145: analysis
146: required 1077 high-resolution spectra to achieve
147: comparable sensitivity to the HST sodium observations.
148: No changes in the CO features in and out-of-transit were measured,
149: implying the presence of high clouds in the planet atmosphere.
150: 
151: 
152: 
153: \section{HD 149026: A Star with a Transiting Planet}
154: 
155: HD 149026 is a metal-rich G0 IV subgiant with V=8.15 and an absolute  
156: visual
157: magnitude, $M_V=3.66$. The planet's mass is similar to Saturn in our  
158: solar
159: system with a radius of $0.725$ \rjup (Sato \etal 2005)
160: determined from the observed photometric transit depth.
161: Interior models discussed by Sato \etal (2005) imply a
162: planet core of about 70 Earth masses.  HD 149026b orbits with a
163: semimajor axis of only $a=0.045$ AU, and Sato et al. (2005) report an
164: effective temperature in the planet atmosphere of 1540 K.
165: 
166: 
167: \subsection{Expected Atmospheric Features}
168: 
169: The equilibrium temperature of an exoplanet atmosphere at the substellar
170: point is a function of the effective stellar temperature, $T_*$, the
171: stellar radius,
172: $R_*$, and the planet-star separation, $a$. These factors are modulated
173: by the planet Bond albedo, $A_B$, and by a geometrical factor, $f$.
174: If a thermal gradient is established from heating on the tidally  
175: locked day
176: side  of the planet, then a wind will circulate energy to the night side
177: of the planet. Depending on the efficiency of this process, $f$ is 1 for
178: isotropic emission from both sides of the planet, and $f$ scales to 2
179: for emission from the day side only.
180: %
181: \begin{equation}
182: T_{eq}=T_*(R_*/2a)^{1/2}[f(1-A_B)]^{1/4}
183: \end{equation}
184: %
185: Adopting $T_*$ and $R_*$ from Sato \etal (2005), and assuming
186: isotropic emission and a Bond albedo, $A_B = 0.3$, we calculate
187: $T_{eq}=1593$ K.
188: 
189: Fortney \etal (2006) define the planet effective temperature as:
190: 
191: %
192: \begin{equation}
193: T_{eff}^{4}=T_{eq}^{4}+T_{int}^{4}
194: \end{equation}
195: %
196: 
197: \noindent
198: where \teq is the equilibrium temperature of the planet in the presence of stellar radiation,
199: and \tint is the effective intrinsic temperature of the planet in isolation. 
200: These authors note that for strongly
201: irradiated planets, the intrinsic isolation temperature is negligible.
202: They consider atmospheric models for HD 149026b with
203: a range of metallicities: [M/H] = 0., 0.5, 1.0.
204: For a cloud-free model with [M/H]=0.5 and assuming isotropic
205: radiation, Fortney \etal (2006) derive an effective
206: temperature for the planet atmosphere of 1734 K, implying $A_B < 0.1 
207: $.  If radiation is
208: not isotropic, but radiates from the day side only (2$\pi$ steradians),
209: this same atmospheric model yields \teff $ = 2148$ K, and TiO and VO
210: features appear at low pressures (Fortney \etal 2006).
211: 
212: The temperature of HD 149026b is therefore likely to be similar to
213: the temperature of L/T dwarfs (Kirkpatrick 2005).
214: The brown dwarf spectral sequence is a temperature
215: sequence over part of the range, except from mid-/late-L to mid-T  
216: where the
217: behavior of clouds become a dominant effect (Kirkpatrick 2005).
218: The effective temperature estimates for HD 149026b ranging from \teff= 
219: $1540 -2148$ K
220: correspond to the brown dwarf spectral type range from early-L to mid-T.
221: 
222: 
223: Based on the effective temperature range of the planet, observable
224: spectral features in the atmosphere can be predicted.  The spectra
225: of L dwarfs, T dwarfs and irradiated giant planets have
226: dominant absorption lines from neutral alkali metals like Na I, K I,
227: Rb I, Cs I and Li I.  Early L-type dwarfs
228: have prominent molecular lines, TiO and VO bands,
229: hydride bands FeH, CaH, and CrH in their atmospheres.
230: By mid to late L-types, the hydrides and the ground state neutral
231: lines of Na I and K I have grown in strength and are the dominant
232: absorbers, while the TiO and VO have weakened.
233: Another important feature in the spectra of late L-type
234: and mid T-type brown dwarfs is H${_2}$O; the neutral alkali
235: lines are still strong, and the hydrides are less prominent.  In
236: late T-type brown dwarfs, the two major lines of Na I and K I have grown
237: very wide (Kirkpatrick 2005).  Because the rainout of condensates
238: clear the atmosphere of most of its metals, the less
239: refractory neutral alkali metals become important in the
240: atmospheres of strongly irradiated, close-in ($\sim0.05$ AU)
241: exoplanets with \teff\ from 800 to
242: 2000 K (Burrows \& Volobuyev 2002; Sudarsky \etal 2003).
243: Sodium and potassium can be particularly important both in
244: brown dwarfs and in the hotter ($\geq\ 1400$ K) of the giant
245: exoplanet classes.  Other alkali metals such as lithium should
246: also be present in giant exoplanets, but with lower
247: abundances (Sudarsky et al. 2003).
248: Fortney et al. (2006) find that model atmospheres of
249: HD 149026b with [M/H]=0.5 should show strong absorption by Na and K
250: in the optical spectra.  As the \emph{majority} of the planet's mass  
251: must be made up of metals, the metallicity of the atmosphere of HD  
252: 149026b could be quite high.
253: 
254: %In the present search, several atomic lines of the alkali
255: %metals in the wavelength range of 6500-8000 \AA\ are studied,
256: %none of which exhibited variations beyond a 3$\sigma$ level.
257: %We carried out a careful study on the the K line at 7698 \AA\ and
258: %the Li I doublet at 6708 \AA, and discovered no significant
259: %features during transit in those two lines (Fig. 3 and Fig. 5).
260: %Out of transit, both the K line and the Li I doublet exhibit
261: %uncertainties of only $\sim$1.4\% and $\sim$0.4\%.
262: %The part of the spectrum containing the sodium line was
263: %contaminated by iodine lines and we were not able to look
264: %for deviations due to Na.
265: 
266: 
267: \subsection{Observations}
268: 
269: A total of 30 spectra of HD 149026 were obtained at the Keck  
270: telescope, using
271: the HIRES echelle spectrometer (Vogt et al. 1994).  Eighteen of
272: these spectra were obtained while the planet was transiting the host  
273: star.
274: The spectra have a resolution, $R\sim$55000, and span a wavelength
275: range of 3500-8000 \AA\ with typical signal-to-noise ratio of 250.
276: All of these observations were obtained in order to derive velocity
277: information about the star and to model the Rossiter-McLaughlin
278: effect (Wolf \etal 2006). The Rossiter-McLaughlin effect
279: is an apparent deviation in the radial velocity from the Keplerian orbit. 
280: This occurs when an extrasolar planet transits across the face of its 
281: host star, first obscuring the approaching limb of the 
282: rotating star and then the receding limb of the star. The resulting 
283: line profile variations at the blue and red edges of the spectral 
284: lines is spuriously interpreted as a Doppler shift in our analysis. 
285: 
286: In order to carry out Doppler analysis,
287: an iodine cell is positioned in front of the spectrometer slit
288: to imprint narrow reference iodine lines.  These iodine lines
289: span the wavelength range
290: 4800-6200 \AA\, and have a depth of about 20\%.  The iodine lines are
291: used to measure wavelength shifts (velocities) and to derive
292: the instrumental point spread function; however, they
293: interfere with our ability to measure small changes in line
294: absorption strengths.  Unfortunately the Na I doublet is contaminated  
295: by the
296: presence of iodine and could not be included in our analysis.
297: 
298: Redward of 6000 \AA, telluric lines begin to appear in the
299: stellar spectrum.  Differences in barycentric velocities at the times
300: of observations
301: create significant shifts in the stellar spectrum relative to
302: the telluric features. Telluric lines that shift in and out of the
303: spectral lines in our analysis are problematic because they will
304: introduce a spurious signal. To avoid this problem, we
305: cross-correlated a telluric spectrum with each stellar spectrum and
306: masked the regions of the spectra containing telluric lines.
307: For those cases where the telluric lines fall
308: on a stellar line, we masked out the entire spectral line to avoid
309: introducing large errors.  Doppler shifts arising from dynamical orbital
310: motion correspond to subpixel shifts and can be ignored.
311: 
312: 
313: \section{Analysis}
314: 
315: Our goal was to search for constituents of the illuminated planet
316: atmosphere during transit.
317: In order to carry out the strongest possible analysis, we worked
318: differentially, comparing spectra obtained during transit with
319: those obtained out-of-transit.  As a first step, we identified
320: the in-transit and out-of-transit spectra by phase-folding the
321: radial velocity (RV) data with a period of
322: $P=2.875965 +0.000085 -0.000135$ days.  This value was calculated
323: by taking the mean of the values provided by Charbonneau et al. (2005),
324: and Wolf et al. (2006) which agree with each other within the
325: stated uncertainties.  The time at the center of transit,
326: $T_c (HJD)=2453527.87455+0.00085-0.00091$, was adopted from
327: Charbonneau et al. (2005).
328: 
329: Figure 1 represents an update to the phased radial velocity data
330: for HD 149026 presented in Sato et al. (2005).
331: The Keplerian fit is overplotted on the RV data.  Fifteen RV  
332: measurements
333: made at Keck during transit are shown as open diamonds in the plot.  
334: These
335: points were removed when fitting the Keplerian model, because they
336: exhibit the Rossiter-McLaughlin effect and deviate from Keplerian
337: velocities.  We froze the period at $P=2.87596$ days to make this fit,
338: and the RMS RV scatter to the fit is 3.07 \ms consistent with our  
339: velocity
340: precision.  The dashed lines in
341: the plot show the 215 minute transit window.
342: 
343: We constructed a high signal-to-noise, continuum normalized template for
344: HD 149026 by averaging all 12 nontransit spectra.
345: Each individual spectrum (both in and out-of-transit)
346: was normalized and cross-correlated with the template
347: spectrum.  As a first step, we divided each spectrum by the template,  
348: and
349: searched the full
350: wavelength range of 6500-8000 \AA\ by eye for any residuals relative to
351: the template spectrum.   We examined 9 spectral orders this way
352: and found no significant changes in the spectrum greater than a few
353: percent.  Because of contamination from telluric and iodine
354: lines we were not able to make use of other spectral orders.
355: 
356: 
357: We then carried out a more robust analysis by searching individual lines
358: such as potassium and lithium in detail.
359: We considered a wavelength bin centered on the K line at 7698 \AA.  The
360: template spectrum is overplotted on an in-transit spectrum in Figure 2.
361: The difference of the two spectra (offset from zero) and
362: the line core bin for the K line are also shown in the same figure.
363: We calculated the percent deviation at each pixel by taking
364: the difference between each spectrum and the template spectrum, and  
365: dividing
366: by the template spectrum.
367: In order to average over photon noise, we binned the percent  
368: difference in
369: wavelength, choosing bins with widths comparable to the sizes of the
370: absorption line cores (8 pixels or 10.4 \kms).  The percent  
371: deviations are
372: plotted with respect to orbital phase in Figure 3.
373: 
374: The individual spectra have signal-to-noise better than 250, so we 
375: expect the photon noise to be about 0.4\% in the continuum but larger
376: in cores of the spectral lines.  In addition to photon noise, there 
377: are systematic errors that arise from imperfect continuum normalization, from 
378: the uncertainties in vertically registering the individual spectra against
379: the template, from spectrum-to-spectrum variations in weak telluric 
380: lines and perhaps from trace iodine features.
381: 
382: To assess our uncertainties, we computed the standard deviation of
383: the percent differences for 12 {\em out-of-transit} spectra in the 
384: wavelength bins of interest. Since the out-of-transit spectra should 
385: not be contaminated by the planet atmospheres,  
386: this represents the typical uncertainties, including Poisson errors and 
387: systematic errors.  The
388: dashed lines in Figure 3 show the limits where the variations
389: would have exceeded three times the intrinsic fluctuations measured in
390: the out-of-transit spectra. 
391: 
392: A Kolmogorov-Smirnov (K-S)
393: test was applied to the in-transit and out-of-transit residuals
394: and suggests that both sets of spectra were drawn from
395: the same parent population.  The probability that the two samples were
396: drawn from different populations is only 0.36\%, implying that there
397: is no evidence for an excess of potassium in the irradiated planet
398: atmosphere.
399: 
400: 
401: The same analysis was applied to the Li I resonance doublet at  
402: 6707.82 \AA.
403: Figure 4 shows a comparison of the template and an in-transit
404: spectrum; the 10.4 \kms bin used for the lithium line is also plotted
405: in the same figure.
406: The percent deviations for Li are shown in Figure 5, and the dashed
407: lines show the limit of 3 times the intrinsic fluctuations measured in
408: the out-of-transit spectra.  Again,
409: the K-S test shows that the in-transit spectra are statistically
410: indistinguishable from the out-of-transit spectra; no excess lithium
411: absorption is seen during transit.
412: 
413: 
414: We also scrutinized observations taken near ingress and egress times,
415: for evidence of a leading or trailing atmosphere that could enhance
416: absorption when the planet is not actually transiting.
417: To investigate this possibility, we modified our original analysis of
418: the potassium and lithium lines.  In this analysis we constructed a
419: template spectrum from only five out-of-transit spectra
420: which were clearly not near the ingress and egress points.
421: Then we repeated the analysis described above for the K and Li lines.
422: Figure 6 shows the percent deviations for individual observations of  
423: the K
424: line relative to the new template and plotted against orbital phase.
425: Figure 7 shows the percent deviations for the Li I line as a function of
426: orbital phase.  The dashed horizontal lines again show boundaries
427: that represent three times the RMS scatter in the 5 non-transiting
428: spectra.
429: The nontransit observations near ingress or egress are shown as
430: boundary nontransit points (open triangles) in the plot.
431: Again, the K-S test shows that the line depths do not vary significantly
432: whether the observation was obtained in-transit, near ingress or  
433: egress, or
434: out-of-transit.  Therefore, we see no line enhancement that suggest a  
435: leading
436: or trailing exosphere with a large optical depth.
437: 
438: 
439: \section{Detectability Simulations}
440: 
441: To determine the increase in optical depth in the planet atmosphere that
442: would be needed to produce a detectable signal, we generated synthetic
443: spectra with injected planetary signals of different strengths.  We  
444: scaled
445: the K line and the Li line in 1\% steps for each of our {\em in-transit}
446: spectra. For each of these pseudo optical depth scaling factors, we
447: applied the two-sided K-S test to find the probability that the  
448: simulated
449: transit data
450: were drawn from the same distribution as our out-of-transit template.
451: For our analysis, we used the template constructed from five
452: observations that were not close to ingress or egress.
453: The pseudo optical depth scaling factors for the K line and the
454: resulting K-S statistic,
455: as well as the detection probabilities are listed in Table 1.
456: Small values of the K-S statistic show that the cumulative distribution
457: function of our scaled up in-transit spectra is significantly different
458: from the out-of-transit template, and therefore yield a high probability
459: detection. Alternatively, the detection probability is one minus the
460: K-S statistic.  A signal with 1.03 pseudo optical depth scaling factor
461: would have been detected at the 95.16\% confidence level.
462: 
463: Figure 8 shows a comparison of in-transit data with a 1.02 pseudo  
464: optical
465: depth scaling for the K line.  This synthetic data set represents a
466: transit by a planet which imprints a
467: spectral signature 2\% stronger than the signal already present in the
468: stellar spectrum.  The top figure shows
469: an overplot of the template and synthetic spectra.  We
470: subtracted the template from the synthetic spectrum and offset the
471: differences from zero for visibility on the same scale.
472: The bottom plot in Figure 8 shows the percent deviations
473: for the out-of-transit and synthetically
474: enhanced in-transit data, relative to the template.
475: The K-S test compares these two populations (out-of-transit data
476: and synthetic in-transit data) to calculate the detection probability
477: listed in Table 1.
478: 
479: Figure 9 shows the synthetic signal corresponding to a 1.05
480: pseudo optical depth scaling factor for the K line.
481: As with Figure 8, the simulated change of 5\% in the transit
482: spectra is overplotted on the template in the top plot.
483: Figure 9 (bottom plot) shows the percent difference
484:  in the 10.4 \kms bin.
485: A K-S test finds a detection probability of 98.8\% for this case.
486: Finally, Figure 10 shows the fake injected signal for a pseudo  
487: optical depth
488: factor of 1.08 at the place of the K line.  Clearly, an enhancement in
489: line depth of 8\% could even be discovered by simple inspection.
490: 
491: We carried out a similar scaling for the Li line at the
492: 10.4 \kms bin at 6707.8 \AA\, and simulated a line with deeper strength.
493: Table 2 shows different values of the pseudo optical depth scaling
494: factor, their corresponding K-S significance level, and the
495: confidence level for detecting the signal.
496: A pseudo optical depth scaling factor of 1.02 results in a 84\%
497: confidence level for detecting lithium, and for a 1.08 scaling
498: factor, the confidence level for detecting the signal is 98.8\%.
499: 
500: 
501: \section{Discussion}
502: The detection of absorption features in the lower atmosphere of a  
503: transiting giant planet requires outstanding precision.  Charbonneau  
504: \etal (2002), who detected neutral atomic sodium, and Deming \etal  
505: (2005), who placed upper limits on CO absorption, obtained a  
506: precision 100 times better for HD209458 than we have in our study.   
507: They were looking for 0.02\% effects.  These studies made use of the  
508: \emph{Hubble Space Telescope} and over 1000 ground-based spectra,  
509: respectively.  However, The detection of species in an extended  
510: exosphere need not necessarily require this same precision, because  
511: the atmosphere's cross-sectional area is significantly larger (Vidal- 
512: Madjar \etal 2003).
513: 
514: If one thinks of our 2\% precision as the smallest cross-sectional  
515: atmospheric area that we could observe, this corresponds to 2.1 $R_ 
516: {\rm J}$ on top of the opaque-at-all-wavelengths 0.725 $R_{\rm J}$  
517: planet.  This large distance makes it immediately clear that we would  
518: only be sensitive to species in a large extended exosphere, such as  
519: that observed by Vidal-Madjar \etal (2003).  Therefore, only if K and  
520: Li had a large enough column abundance to remain optically thick past  
521: 2.1 $R_{\rm J}$ would these atoms have been detected.
522: 
523: Furthermore, these species must remain neutral and atomic, such that  
524: they would absorb stellar flux at our specified wavelengths.  While  
525: these species are likely neutral in the lower atmosphere around $P  
526: \gtrsim$1 mbar, where Charbonneau \etal (2002) detected Na in HD  
527: 209458b, all exospheric models for highly irradiated giant planets  
528: predict temperatures on the order of 10$^4$ K (Lammer \etal 2003,  
529: Yelle 2004, Tian \etal 2005).  Based on calculations by Yelle (2004)  
530: for an HD 2094358b-like planet at 0.04 AU, 2.1 $R_{\rm J}$ likely  
531: corresponds to $P \sim 10^{-10}$ bar and $T \sim 2 \times 10^4$ K.   
532: This high temperature makes it likely that Li and K will be ionized.   
533: Examining K specifically, Lodders \& Fegley (2006) review the  
534: chemistry of alkalis in substellar atmospheres and plot a curve where  
535: K I and K II have an equal abundance.  This curve has only weak  
536: pressure dependence.  At the lowest pressure considered, 1 mbar, K I  
537: and K II are equal in abundance at a temperature of 1900 K.  Our  
538: exospheric temperatures are likely 5-10 times hotter, meaning K II is  
539: strongly preferred.  Photodissociation is also likely important and  
540: will further reduce the abundances of Li I and K I.  Fortney \etal  
541: (2003) examined ionization of Na in the atmosphere of HD 209458b and  
542: found that photoionization of this species on the planet's limb could  
543: proceed to pressures as high as $\sim$ 1 mbar.  The atoms Li and K  
544: have even lower ionization potentials.  Due to the high temperatures  
545: and strong photoionizing flux, it appears unlikely that neutral  
546: atomic Li and K will be found in significant abundance in the  
547: extended atmosphere of HD 149026b.  This is consistent with the  
548: observations of Charbonneau \etal (2002), who observed Na I in the  
549: lower atmosphere, but not the exosphere, of HD 209458b.
550: 
551: R.~S.~Freedman has kindly provided us with absorption cross-sections  
552: as a function of wavelength for K and Li across our observation  
553: bands.  These are $\sim 4 \times 10^{-13}$ cm$^2$ for Li and $\sim 1  
554: \times 10^{-13}$ cm$^2$ for K.  Since we find that the exosphere has  
555: an optical depth $\lessim 1$ at 2.1 $R_{\rm J}$, this corresponds to  
556: column densities for these atoms in their neutral atomic ground  
557: state, of $\lesssim 2 \times 10^{12}$ cm$^{-2}$for Li, and $\lessim 9  
558: \times 10^{12}$ cm$^{-2}$ for K, at this radius.
559: 
560: 
561: A more sensitive search from
562: space-based telescopes could result in a detection of the planetary
563: atmosphere. Ionized alkali species would have noble gas electron 
564: configurations and would not 
565: have prominent absorption features even for space based observations.  We
566: advocate a search for H Lyman $\alpha$ from space to give a first  
567: estimate of the extent of an exosphere in HD 149026b.  At the same time, since  
568: there is a
569: clear link between heavy element mass and atmospheric metallicity in the
570: giant planets in our own solar system, the atmosphere and exosphere of
571: HD 149026b could be richer in metals than HD 209458b. The next elements
572: to search for would probably be C, N, and O species.
573: 
574: 
575: We gratefully acknowledge the dedication and support of the Keck
576: Observatory staff,
577: in particular Grant Hill for support with HIRES.  We thank Paul  
578: Butler and
579: Steve Vogt for contributing the observed spectra.  We thank Richard  
580: Freedman for providing us with opacity data.  Thanks to Jessica
581: Lovering for reduction of the CCD images to 1-D spectra.  We thank  
582: the NOAO
583: and NASA Telescope
584: assignment committees for generous allocations of telescope time.
585: Data presented herein were obtained at the W. M. Keck Observatory
586: from telescope time allocated to the National
587: Aeronautics and Space Administration through the agency's scientific
588: partnership with the California Institute of Technology and the
589: University of
590: California.  The Observatory was made possible by the generous financial
591: support of the W. M. Keck Foundation. We thank the Michaelson Science
592: Center for travel
593: support and support through the KDPA program.  DAF is a Cottrell
594: Science Scholar
595: of Research Corporation. We acknowledge support from NASA grant (to
596: DAF); a NASA Postdoctoral Program Fellowship (to JJF);
597: NASA grant NCC5-511; NASA grant NAG5-75005 (to GWM).
598: 
599: 
600: 
601: \clearpage
602: 
603: 
604: \begin{thebibliography}
605: 
606: \bibitem[Brown (2001)]{brown01} Brown, T. M.  2001, \apj, 553, 1006
607: 
608: \bibitem[Bundy \& Marcy (2000)]{bm00}
609: Bundy, K. A., \& Marcy, G. W.  2000, PASP, 112, 1421
610: 
611: \bibitem[Burrows \& Sharp (1999)]{bs99}
612: Burrows, A., \& Sharp, C. M.  1999, \apj, 512, 843
613: 
614: \bibitem[Burrows \& Volobuyev (2002)]{bv02}
615: Burrows, A., \& Volobuyev, M.  2002, \apj, 583, 985
616: 
617: \bibitem[Butler et al. (1998)]{but98} Butler R. P., Marcy, G. W.,
618: Vogt, S. S., \& Apps, K.,  1998, PASP, 110, 1389
619: 
620: \bibitem[Charbonneau et al. (2002)]{char01}
621: Charbonneau, D., Brown, T. M., Noyes, R. W., \& Gilliland, R. L.   
622: 2002, \apj, 586, 377
623: 
624: \bibitem[Charbonneau et al. (2005)]{char05}
625: Charbonneau, D., Winn, J. N., Latham, D. W., Bakos,
626: G., Falco, E. E., Holman, M. J., Noyes, R. W., \& Cs\'ak, B.
627: 2005, \apj, in press
628: 
629: \bibitem[Deming \& Brown (2005)]{db05}
630: Deming, D., Brown, T. M., Charbonneau, D., Harrington, J., \&  
631: Richardson,
632: L. J. 2005, \apj, 622, 1149
633: 
634: \bibitem[Deming et al. (2005)]{dem05}
635: Deming, D., Seager, S., Richardson,
636: L. J., \& Harrington, J.  2005, Nature, 434, 740
637: 
638: \bibitem[Fischer et al. (2004)]{fisch04}
639: Fischer, D. A., Laughlin, G., Butler, P., Marcy, G.,
640: Johnson, J., Henry, G., Valenti, J., Vogt, S., Ammons, M.,
641: Robinson, S., Spear, G., Strader, J., Driscoll, P., Fuller, A.,
642: Johnson, T., Manrao, E., McCarthy, C., Munoz, M.,
643: Tah, K. L., Wright, J., Ida, S., Sato, B., Toyota, E.,
644: \& Minniti, D.  2004, \apj, 620, 481
645: 
646: \bibitem[Fischer et al. (2006)]{fisch06}
647: Fischer, D. A., Laughlin, G., Marcy, G. W., Butler, R. P.,
648: Vogt, S. S., Johnson, J. A.,
649: Henry, G. W., McCarthy, C., Ammons, M., Robinson, S.,
650: Strader, J., Valenti, J. A., McCullough, P. R.,
651: Charbonneau, D., Haislip, J., Knutson, H. A.,
652: Reichart, D. E., McGee, P., Monard, B., Wright, J. T.,
653: Ida, S., Sato, B., \& Minniti, D.  2006, \apj, 637, 1094
654: 
655: \bibitem[Fortney (2005)]{fort05}
656: Fortney, J. J. 2005, \mnras, 364, 649
657: 
658: \bibitem[Fortney (2003)]{fort03} Fortney, J. J., Sudarsky, D.,  
659: Hubeny, I., Cooper, C. S., Hubbard, W. B., Burrows, A., \& Lunine, J.  
660: I.  2003, \apj, 589, 615
661: 
662: \bibitem[Fortney et al. (2006)]{fortn06}
663: Fortney, J. J., Saumon, D., Marley, M. S., Lodders, K.,  \&
664: Freedman, R. S. 2005, \apj, 642, 495
665: 
666: \bibitem[Henry et al. (2000)]{henry00}
667: Henry, G. W., Marcy, G. W., Butler, R. P.,  \&
668: Vogt, S. S. 2000, \apj, 529, L41
669: 
670: \bibitem[Hubbard (2001)]{hubb01}  Hubbard, W. B., Fortney, J. J.,  
671: Lunine, J. I., Burrows, A., Sudarsky, D., \& Pinto, P.  2001, \apj,  
672: 560, 413
673: 
674: \bibitem[Johnson (2006)]{john06}  Johnson, et al. 2006, in press
675: 
676: \bibitem[Kirkpatrick (2005)]{kirk05}  Kirkpatrick, J. D.  2005, ARA 
677: \&A, 43, 195
678: 
679: \bibitem[Lammer (2003)]{lamm03}  Lammer, H., Selsis, F., Ribas, I.,  
680: Guinan, E. F., Bauer, S. J., \& Weiss, W. W.  2003, \apj, 598, L121
681: 
682: \bibitem[Lodders (1999)]{lod99}  Lodders, K. 1999, \apj, 519, 793
683: 
684: \bibitem[Lodders and Fegley (2006)]{lod06}  Lodders, K., Fegley, B.,  
685: Astrophysics Update v. 2, in press, astro-ph/0601381
686: 
687: \bibitem[Sato et al.(2005)]{sat05} Sato, B., Fischer, D. A., Henry,
688: G. W., Laughlin, G., Butler, R. P., Marcy, G. W.,
689: Vogt, S. S., Bodenheimer, P., Ida, S., Toyota, E., Wolf, A.,
690: Valenti, J. A., Boyd, L. J., Johnson, J. A., Wright, J. T.,
691: Ammons, M., Robinson, S., Strader, J., McCarthy, C.,
692: Tah, K. L., \& Minniti, D.  2005, \apj, 633, 465
693: 
694: \bibitem[Seager \& Sasselov(2000)]{ss00}
695: Seager, S., \& Sasselov, D. D.  2000, \apj, 537, 916
696: 
697: \bibitem[Sudarsky et al. (2003)]{sud03} Sudarsky, D.,
698: Burrows, A., \& Hubeny, I.  2003, \apj, 588, 1121
699: 
700: \bibitem[Sudarsky et al. (2000)]{sud00} Sudarsky, D.,
701: Burrows, A., \& Pinto, P.  2000, \apj, 538, 885
702: 
703: \bibitem[Tian et al.(2005)]{tian05} Tian, F., Toon, O. B., Pavlov, A.  
704: A., \& De Sterck, H.  2005, \apj, 621, 1049
705: 
706: \bibitem[Vidal-Madjar et al.(2003)]{vm03}
707: Vidal-Madjar, A., Lecavelier des Etangs, A., D\'esert, J. M.,
708:   Ballester, G. E., Ferlet, R, H\'ebrard, G., Mayor, M.  2003, \nat,  
709: 422, 143
710: 
711: \bibitem[Vogt(1994)]{vogt94} Vogt, S. S., et al.  1994, Proc.SPIE,  
712: 2198, 362
713: 
714: \bibitem[Wolf(2006)]{wolf06} Wolf, A., et al.  2006, in press
715: 
716: \bibitem[Yelle(2004)]{yelle04} Yelle, R. V.  2004, Icarus, 170, 167
717: 
718: \end{thebibliography}
719: 
720: \clearpage
721: 
722: 
723: \begin{figure}
724: \begin{center}
725: \leavevmode
726: \scalebox{1}{\includegraphics{f1.eps}}
727: \end{center}
728: \caption{Updated phased radial velocities for HD 149026. The velocities
729: measured at Keck during transit are shown as open diamonds in the plot.
730: These measurements
731: exhibit the Rossiter-McLaughlin effect and were removed when fitting
732: a Keplerian model.\label{fig1}}
733: \end{figure}
734: 
735: \clearpage
736: 
737: 
738: 
739: \begin{figure}
740: %\centering
741: \begin{center}
742: \leavevmode
743: \scalebox{0.75}{\includegraphics{f2.eps}}
744: \end{center}
745: \caption{The template nontransit spectrum for the K line overplotted  
746: on a
747: transit spectrum. The difference of the two (offset from zero) and the
748: 10.4 \kms wavelength bin (dashed lines) are also shown in
749: this plot.\label{fig2}}
750: \end{figure}
751: 
752: \clearpage
753: 
754: 
755: \begin{figure}
756: %\centering
757: \begin{center}
758: %\leavevmode
759: \scalebox{1}{\includegraphics{f3.eps}}
760: \end{center}
761: \caption{Deviations of the K line. The percent deviation is averaged  
762: over
763: the 10.4 \kms bin. This percent deviation was calculated by taking the
764: difference between each spectrum and the template nontransit spectrum
765: (which was constructed by averaging 12 nontransit spectra) and  
766: dividing by
767: the value of the template spectrum. The dashed lines indicate the
768: 3$\sigma$ detection limits, where $\sigma$ is the RMS scatter in the
769: out-of-transit spectra.\label{fig3}}
770: \end{figure}
771: 
772: \clearpage
773: 
774: 
775: \begin{figure}
776: %\centering
777: \begin{center}
778: %\leavevmode
779: \scalebox{0.75}{\includegraphics{f4.eps}}
780: \end{center}
781: \caption{The template nontransit spectrum for the Li line overplotted  
782: on a
783: transit spectrum. The difference of the two (offset from zero) and the
784: 10.4 \kms wavelength bin (dashed lines) are also shown in this
785: plot.\label{fig4}}
786: \end{figure}
787: 
788: \clearpage
789: 
790: 
791: \begin{figure}
792: %\centering
793: \begin{center}
794: %\leavevmode
795: \scalebox{1}{\includegraphics{f5.eps}}
796: \end{center}
797: \caption{Deviations of the Li line. The percent deviation is averaged
798: over the 10.4 \kms bin. This percent deviation is calculated by  
799: taking the
800: difference between each spectrum and the template nontransit spectrum
801: (which is constructed by averaging 12 nontransit spectra) and  
802: dividing by
803: the value of the template spectrum. The 3$\sigma$ detection limits are
804: plotted as dashed lines.\label{fig5}}
805: \end{figure}
806: 
807: \clearpage
808: 
809: 
810: \begin{figure}
811: \begin{center}
812: \scalebox{1}{\includegraphics{f6.eps}}
813: \end{center}
814: \caption{Deviations of the K line. These deviations were computed
815: using a template spectrum
816: that was constructed from 5 nontransit spectra. The triangles show
817: the boundary nontransit observations which were close to the ingress
818: and were not used in the template nontransit spectrum.\label{fig6}}
819: \end{figure}
820: \clearpage
821: 
822: \begin{figure}
823: \begin{center}
824: \scalebox{1}{\includegraphics{f7.eps}}
825: \end{center}
826: \caption{Deviations of the Li line. These deviations were computed
827: using a template
828: spectrum that was constructed from 5 nontransit spectra. The  
829: triangles show
830: the boundary nontransit observations which were close to the ingress
831: and were not used in the template nontransit spectrum.\label{fig7}}
832: \end{figure}
833: 
834: \clearpage
835: 
836: 
837: 
838: 
839: 
840: \begin{figure}
841: \centering
842: \scalebox{0.6}{\includegraphics{f8a.eps}}
843: \scalebox{0.6}{\includegraphics{f8b.eps}}
844: \caption{Synthetic data. A 2\% signal is introduced at 7698 \AA.
845: (a) The synthetic transit spectrum overplotted on
846: the template nontransit spectrum. The difference (offset from zero)
847: is also plotted. (b) The percent deviations
848: for such a signal. There is an 84\%\ confidence level to detect such a
849: signal.\label{fig8}}
850: \end{figure}
851: 
852: \clearpage
853: 
854: 
855: \begin{figure}
856: \centering
857: \scalebox{0.6}{\includegraphics{f9a.eps}}
858: \scalebox{0.6}{\includegraphics{f9b.eps}}
859: \caption{Synthetic data. A 5\% signal is introduced at 7698 \AA.
860: (a) The synthetic transit spectrum overplotted on
861: the template nontransit spectrum. The difference (offset from zero)
862: is also plotted. (b) The percent deviations
863: for such a signal. There is a 98.8 \%\ confidence level to detect such
864: a signal.\label{fig9}}
865: \end{figure}
866: 
867: \clearpage
868: 
869: 
870: \begin{figure}
871: \centering
872: \scalebox{0.6}{\includegraphics{f10a.eps}}
873: \scalebox{0.6}{\includegraphics{f10b.eps}}
874: \caption{Synthetic data. An 8\% signal is introduced at 7698 \AA.
875: (a) The difference (offset from zero) of the synthetic transit
876: spectrum and the template nontransit spectrum. (b) The percent  
877: deviations
878: for such a signal. Such a signal would clearly be detected.\label 
879: {fig10}}
880: \end{figure}
881: 
882: \clearpage
883: 
884: \include{tab1.tex}
885: \include{tab2.tex}
886: 
887: 
888: 
889: 
890: \end{document}
891: 
892: 
893: 
894: 
895: