1: \documentclass[12pt,preprint]{aastex}
2: %\documentclass{emulateapj}
3:
4: %\usepackage{float}
5: \usepackage{epsfig}
6: %\usepackage{emulateapj5}
7: %\usepackage{onecolfloat5}
8: %\usepackage{apjfonts}
9: \usepackage{natbib}
10: \bibliographystyle{apj}
11:
12: \def\plotone#1{\centering \leavevmode
13: \includegraphics[clip=, width=.75\columnwidth]{#1}}
14: \def\plottwo#1#2{\centering \leavevmode
15: \includegraphics[width=.45\columnwidth]{#1} \hfil
16: \includegraphics[width=.45\columnwidth]{#2}}
17: \def\plotthree#1#2#3{\centering \leavevmode
18: \includegraphics[width=.3\columnwidth]{#1} \hfil
19: \includegraphics[width=.3\columnwidth]{#2} \hfil
20: \includegraphics[width=.3\columnwidth]{#3}}
21:
22: \newcommand{\cN}[1]{\mathcal{N}}
23: \newcommand{\pn}[1]{\mbox{$(#1)$}}
24: \newcommand{\spa}{\mbox{ }}
25: \def\gsim{\;\rlap{\lower 2.5pt
26: \hbox{$\sim$}}\raise 1.5pt\hbox{$>$}\;}
27: \def\lsim{\;\rlap{\lower 2.5pt
28: \hbox{$\sim$}}\raise 1.5pt\hbox{$<$}\;}
29:
30: \textwidth = 6.5 in
31: \textheight = 9 in
32: \oddsidemargin = 0.0 in
33: \evensidemargin = 0.0 in
34: \topmargin = -0.3 in
35: \headheight = 0.3 in
36: \headsep = 0.5 in
37:
38:
39:
40: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
41: % THE DOCUMENT BEGINS HERE %
42: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
43:
44: \begin{document}
45:
46:
47: %%% Begin front material
48: %\twocolumn[%%% Begin front material
49:
50: \title{%Place-Holder Title\\
51: On Constraining A Transiting Exoplanet's Rotation Rate\\
52: With Its Transit Spectrum}
53:
54: \author{David S. Spiegel\altaffilmark{1}, Zolt\'an Haiman\altaffilmark{1},
55: B. Scott Gaudi\altaffilmark{2}}
56:
57: \affil{$^1$Department of Astronomy, Columbia University, 550 West 120th Street, New York, NY 10027}
58:
59: \affil{$^2$Department of Astronomy, Ohio State University, 140 W. 18th Avenue, Columbus Ohio 43210--1173}
60:
61: \vspace{0.5\baselineskip}
62:
63: \email{dave@astro.columbia.edu, zoltan@astro.columbia.edu,
64: gaudi@astronomy.ohio-state.edu}
65:
66:
67:
68:
69: \begin{abstract}
70: We investigate the effect of planetary rotation on the transit
71: spectrum of an extrasolar giant planet.
72: During ingress and egress, absorption features arising from the planet's
73: atmosphere are Doppler shifted by of order the planet's rotational
74: velocity ($\sim 1-2 {\rm ~km~s^{-1}}$) relative to where they would be
75: if the planet were not rotating. We focus in particular on the case of
76: HD209458b, which ought to be at least as good a target as any other known
77: transiting planet. For HD209458b, this shift should give rise to a small
78: net centroid shift of $\sim 60 {\rm ~cm~s^{-1}}$ on the stellar absorption
79: lines. Using a detailed model of the transmission spectrum due to a
80: rotating star transited by a rotating planet with an isothermal atmosphere,
81: we simulate the effect of the planet's rotation on the shape
82: of the spectral lines, and in particular on the magnitude of their width
83: and centroid shift. We then use this simulation to determine the
84: expected signal--to--noise ratio for distinguishing a rotating from a
85: non--rotating planet, and assess how this S/N scales with
86: various parameters of HD209458b. We find that with a 6~m telescope,
87: an equatorial rotational velocity of $\sim 2 {\rm ~km~s^{-1}}$ could be
88: detected with a S/N $\sim 5$ by accumulating the signal over many transits
89: over the course of several years. With a 30~m telescope, the time required
90: to make such a detection reduces to less than 2 months.
91: \end{abstract}
92:
93: \keywords{astrobiology -- planetary systems -- radiative transfer --
94: stars: atmospheres --
95: stars:individual (HD209458) --
96: astrochemistry}%]%%% End front material
97:
98: \section{Introduction}
99: \label{sec:intro}
100: Since the early 1990s, more than 200 planets have been discovered orbiting
101: other stars.
102: Roughly four fifths of the planets that have been discovered are at least half
103: as massive as Jupiter, and about a quarter of the known planets
104: orbit extremely close to their parent star ($\lsim 0.1 {\rm~AU}$).
105: The Jupiter--mass planets that are in close orbits are highly irradiated by
106: their stars and are therefore called ``Hot Jupiters.''
107: % Make sure you know which 20 (which are the 6 new ones?)
108: % Corot-Exo-1b, no reference...
109: Twenty Hot Jupiters transit along the line of sight between Earth and
110: their star
111: \citep{charbonneau_et_al2000, henry_et_al2000, konacki_et_al2003,
112: bouchy_et_al2004, pont_et_al2004, konacki_et_al2004, konacki_et_al2005,
113: udalski_et_al2002a, udalski_et_al2002b, udalski_et_al2002c, udalski_et_al2003,
114: udalski_et_al2004, alonso_et_al2004, bouchy_et_al2005, mccullough_et_al2006,
115: odonovan_et_al2006b, bakos_et_al2006, collier_et_al2006, burke_et_al2007,
116: gillon_et_al2007, bakos_et_al2007a, bakos_et_al2007bb,
117: johnskrull_et_al2007}\footnote{Corot--Exo--1b is reported at
118: {\it http://www.esa.int/esaCP/SEMCKNU681F\_index\_0.html}.}.
119: Observations of the transiting planets have confirmed
120: their similarity to Jupiter by revealing that these are gas giant planets,
121: with radii comparable to, or somewhat larger than Jupiter's
122: \citep{charbonneau_et_al2000, henry_et_al2000, gaudi2005}.
123:
124: The effects of the tidal torques experienced by an orbiting body have
125: been studied for a long time -- for an early seminal analyis, see
126: \citet{goldreich+peale1966}. Such torques
127: tend to synchronize a satellite's rotation rate to its orbital rate, and
128: if the torque is sufficient this synchronization is achieved and the
129: orbiter is said to be ``tidally locked,'' as the Earth's Moon is. The
130: Hot Jupiter--class extrasolar planets are thought to orbit sufficiently
131: close to their stars that their tidal locking timescales are much shorter
132: than the ages of the planets. The planets, then, are expected to be
133: tidally locked to the stars, with one hemisphere in permanent day and
134: the other in permanent night \citep{harrington_et_al2006}.
135:
136: A tidally locked Hot Jupiter will have a permanent sharp contrast in
137: temperature between the substellar point and the night side, which must
138: have a profound influence on the atmospheric dynamics.
139: \citet{showman+guillot2002} make simple predictions of the day/night
140: temperature difference ($\sim 500 {\rm ~K}$) and the speed of winds (up
141: to $\sim 2 {\rm ~km~s^{-1}}$), and their detailed, three--dimensional
142: simulations agree with their estimates. Shallow--water simulations by
143: \citet{cho_et_al2003} predict longitudinally averaged zonal wind speeds
144: of up to $400 {\rm~m~s^{-1}}$, with local winds approaching
145: $2.7 {\rm~km~s^{-1}}$ (under some assumtions). Simulations by
146: \citet{cooper+showman2005}
147: predict a super--rotational jet (i.e., blowing eastward, where north is
148: defined by the right--hand rule) that blows the hottest part of the planet
149: downstream by about $60\degr$ from the substellar point. Their simulations
150: predict supersonic winds exceeding $9 {\rm~km~s^{-1}}$ at high latitudes,
151: high in the atmosphere (where the optical depth is low) and winds exceeding
152: $4 {\rm~km~s^{-1}}$ at pressures near the photosphere. A Spitzer Space
153: Telescope phase curve for $\upsilon$ Andromedae b rules out a phase--shift
154: as large as $60\degr$ between the substellar point and the hottest
155: spot \citep{harrington_et_al2006}, but a Spitzer phase curve for
156: HD189733b favors a $\sim 30\degr$ shift for that planet
157: \citep{knutson_et_al2007b}, so it remains unclear to what extent available
158: data indicate very strong photospheric winds.
159:
160: Transmission spectroscopy is a way to probe the atmospheres of these planets.
161: \citet{charbonneau_et_al2002} were the first to detect an absorption feature
162: in what is probably the atmosphere of HD209458b, when they found that the
163: effective radius of the planet increases slightly at the wavelength of a
164: strong sodium absorption doublet (the sodium D lines) at $\sim 590 {\rm~nm}$.
165: In addition, \citet{vidal-madjar_et_al2003, vidal-madjar_et_al2004} have
166: reported a number of absorption features in HD209458's transit spectra that
167: are due to various species (including hydrogen Lyman alpha, neutral carbon,
168: oxygen, and sulfur, and some ionization states of carbon, nitrogen, and
169: silicon) in a hot exosphere that is probably only loosely bound to the planet.
170: Intriguingly, through analyzing the red and near--IR portion of HD209458b's
171: transit spectrum \citet{barman2007} found a $10 \sigma$ detection of
172: atmospheric water vapor. Several measurements of the planet's emission
173: spectrum, however, have found results that seem to be inconsistent with
174: high water abundance high in the atmosphere \citep{grillmair_et_al2007,
175: richardson_et_al2007, swain_et_al2007}.
176:
177: Initial work by \citet{seager+sasselov2000} and a comprehensive study by
178: \citet[hereafter B01]{brown2001}
179: have described various other considerations that should affect the details of
180: transit spectra, including the orbital motion of a planet (a few tens of
181: kilometers per second in the radial
182: direction), the rotation of the planet (a few kilometers per second at the
183: equator, according to the hypothesis that the planet is tidally locked), and
184: winds on the planet's surface (in B01's analysis, up to
185: $\sim 1 {\rm~km~s^{-1}}$). These physical effects should tend to broaden or
186: impose Doppler shifts on absorption features due to the planet's atmosphere.
187: B01 constructed an impressively detailed model of radiative transfer
188: through a Hot Jupiter's atmosphere, assuming various models of zonal windflow
189: superimposed on an equatorial bulk rotation speed of
190: $v_{\rm eq} = 2 {\rm~km~s^{-1}}$, which is approximately the value for
191: HD209458b under the assumption that it is tidally locked in its 3.5 day orbit.
192: He finds the height of the cloud deck to be the most important
193: parameter that affects the transmission of light through the planet's
194: atmosphere.
195:
196: The original discovery of the roughly Jupiter--mass planet in a
197: close, $\sim 4$ day orbit around 51 Pegasi \citep{mayor+queloz1995} prompted
198: interest in the dynamics and structure that must govern a highly insolated
199: gas giant planet \citep{guillot_et_al1996}. Observations of the transiting
200: Hot Jupiters heightened this interest when they revealed a puzzling feature of
201: these planets: at least several of them are a bit puffier than Jupiter, with
202: diameters ranging from slightly larger than Jupiter's to as much as $\sim 80\%$
203: larger. It is not clear what allows some planets to maintain such large radii.
204: It has been suggested that if a
205: Jovian planet migrates very quickly, from its presumed formation location at
206: least several AU from its star, to its eventual several day orbit, then it
207: might reach its final home before it has cooled enough to shrink to Jupiter's
208: radius. Accordingly, some authors have investigated the migration processes
209: that lead gas giant planets to such close orbits as have been found
210: \citep[e.g.][]{trilling_et_al2002}. Others have investigated various ways
211: in which a gas giant could either be heated once it ends up near its star, or
212: otherwise maintain sufficient internal energy to sustain its inflated size
213: \citep{guillot+showman2002, burrows_et_al2003, laughlin_et_al2005,
214: bodenheimer_et_al2003, guillot2005, burrows_et_al2007, chabrier+baraffe2007}.
215: Although various physical mechanisms have been suggested as the apparently
216: missing energy source that allows the unexpectedly large radii sometimes seen,
217: the lesson of these investigations {\it in toto} is that it is not easy to
218: explain the inflated sizes, either in terms of the greater stellar flux that
219: these planets experience by virtue of being so close to their stars, or in
220: terms of their evolutionary migratory histories. A recent paper by
221: \citet{winn+holman2005} propose that, contrary to
222: the commonly accepted paradigm, Hot Jupiters might be trapped in a
223: Cassini state with large obliquity, in which the spin--axis precesses in
224: resonance with the orbit, but lies nearly in the orbital plane. Such a state
225: might be stable against perturbation, and yet able to generate sufficient
226: internal energy to increase a gas giant planet's radius to the observed
227: values. In light of an even more recent analysis by
228: \citet{levrard_et_al2007}, however, it appears that the probability of
229: capture into a Cassini state 2 resonance is quite small for a planet with
230: semi--major axis $a < 0.1 {\rm ~AU}$. Furthermore, \citet{fabrycky_et_al2007}
231: argue that even if a planet is captured into Cassini state 2, it is likely to
232: remain there for a time that is short relative to the age of the system.
233:
234: High--resolution transit spectra that have high signal--to--noise ratios will
235: allow us to distinguish between various models of orbit, rotation, and
236: weather, as discussed by B01. Because the orbit is known to high
237: accuracy, and the predictions of the effects of weather (or climate) are
238: highly uncertain, as described above, we will focus in this paper on the
239: much more easily predicted effect of a planet's rotation on a
240: transit--spectrum. If we neglect winds, then the large--obliquity Cassini
241: state described by \citet{winn+holman2005} should have a spectral signature
242: that is very similar to that of a non--rotating model. In contrast, the
243: rotation of a tidally locked planet should impose a Doppler distortion
244: on spectral lines arising from the planet's atmosphere that
245: is roughly an overall redshift during ingress, as the planet is just
246: entering the stellar disk, and a similar distortion that is roughly an
247: overall blueshift during egress, as the planet is just exiting the disk.
248: During mid--transit, the spectral distortion is more similar to rotational
249: broadening. In the present investigation, we address whether there is any
250: hope that these spectral distortions from tidally--locked rotation can be
251: observed. In our study, we focus only on the sodium doublet detected
252: by \citet{charbonneau_et_al2002}. As we will show below, the sensitivity
253: of a measurement of rotation scales with the square root of the number of lines
254: under consideration. Model spectra from, e.g., \citet{sudarsky_et_al2003} and
255: \citet{barman2007} predict a strong potassium doublet at
256: $\sim 770 {\rm~nm}$, strong water absorption features in the near--infrared,
257: and a handful of near--UV lines. If some of these are confirmed in the
258: atmosphere of a transiting planet, they will provide a modest
259: increase in S/N. Since the sodium lines are expected to be the strongest,
260: however, it seems unlikely that observing multiple lines will yield a
261: boost in S/N by more than a factor of a few.
262:
263: We emphasize that it may not be at all justified to neglect winds. It is
264: quite likely that there are super--rotational winds on Hot Jupiters, which
265: are probably necessary to heat the ``night'' side. As indicated above, some
266: models predict, and the observed phase curve for HD189733b suggests, that at
267: the photosphere these winds might be significantly (100\% or more) greater
268: than the equatorial rotation rate, and therefore might contribute
269: importantly to the Doppler distortion induced by the motion of the planet's
270: atmosphere. Nevertheless, in order to isolate the contribution of rotation,
271: we do neglect winds in this study. The Doppler distortions that we predict
272: can therefore probably be taken as a lower bound on the distortions that would
273: be observed for a tidally--locked transiting Hot Jupiter.
274:
275: We find that the spectral shifts induced by rotation will be difficult to
276: detect with current technology, but perhaps not insurmountably so, at least
277: with technology that might be available in the not--to--distant future.
278: The measurements we will describe are limited by a paucity of photons.
279: As such, their signal--to--noise ratio will be enhanced by a bright star and
280: a puffy planet (i.e., a planet with a large scale--height). HD209458 is
281: at least a magnitude brighter than any other star with a known transiting
282: planet except HD189733, and its planet is larger than HD189733b; so HD209458b
283: should be a better target than any other known transiting planet except
284: possibly HD189733b. In this paper, we model the HD209458b system because it is
285: the best--studied system, and it is unlikely that any currently--known
286: planets would be significantly better targets.
287: In a single transit, observations of HD209458 with a 6~m
288: telescope that has a high--resolution ($>50,000$) optical spectrograph
289: with good throughput ($\sim 18\%$) could only show the influence of tidally
290: locked rotation at the $\sim 0.2\sigma$ level. With ultrahigh--resolution
291: ($\gsim 700,000$) and good throughput ($\sim 4\%$) this effect would still
292: only show up at the $\sim 0.6 \sigma$ level. In less than a year, the signal
293: of rotation could be present at five times the noise ($S/N = 5$). Of
294: course, a telescope with larger collecting area, higher spectral
295: resolution, or better throughput would cause the signal to be apparent
296: at that significance level in less time.
297:
298: Other studies have approached the problem of determining the rotation
299: rate from a different angle. \citet{seager+hui2002} and
300: \citet{barnes+fortney2003} suggest that an oblate spheroid will have a
301: different transit light curve from a perfect sphere, and so measuring
302: the oblateness from transit photometry will provide a handle on the
303: rotation rate. The oblateness is somewhat degenerate with several
304: other parameters that are not perfectly known, however, so they
305: conclude that it would be difficult to actually determine that
306: rotation rate in this manner. The method we describe here could
307: eventually prove to be an important complement to other observations
308: to constrain the rotation rate.
309:
310: In the remainder of this paper, we address this idea in detail.
311: One complication that we discuss below is that the technique of this paper
312: is not immune from several near--degeneracies among the many
313: attributes of transiting extrasolar planets that influence light
314: curves or spectra. Although it is likely that current or near--future
315: instruments will be sensitive enough that the spectral distortion
316: imposed by HD209458b's rotation (if it is tidally locked) is
317: visible, it might still be very challenging to discern the fingerprint
318: of rotation from other attributes that affect the spectra at a similar
319: level. In this paper, we tackle the forward problem of calculating the
320: amount of distortion that is caused by rotation. The inverse
321: problem -- determining from observations whether a planet is tidally
322: locked -- is more difficult and should be the topic of a future study.
323:
324: The structure of the rest of this paper is as follows:
325: In \S~\ref{sec:overview}, we describe qualitatively what happens to
326: the starlight received on Earth when a planet transits its star; we
327: give a rough order of magnitude estimate of the the magnitude and
328: detectability of the spectral distortions caused by tidally locked
329: rotation; and we briefly describe some technological progress and
330: remaining challenges relevant to our task of deducing bulk motions
331: in a planet's atmosphere from transit spectra. In \S~\ref{sec:model},
332: we describe our computational model of a transit spectrum.
333: In \S~\ref{sec:results}, we describe the results of our model
334: according to various assumed input parameters.
335: In \S~\ref{sec:disc}, we discuss the scaling of S/N on various model
336: parameters and we address the prospects of actually observationally
337: determining whether a transitting planet is tidally locked. In
338: \S~\ref{sec:conc}, we conclude
339: by describing various ways to boost our predicted S/N to a more
340: optimistic value.
341:
342:
343:
344: \section{Overview of the Problem}
345: \label{sec:overview}
346:
347: The practical feasibility of the investigation we undertake depends on a
348: few factors: understanding the various detailed processes that affect
349: the starlight that reaches Earth when a planet transits its star; the
350: magnitude of the distortion that tidally locked rotation induces; and the
351: technology available to measure such distortions. In this section, we
352: give an overview of these three factors -- in particular, in
353: \S~\ref{ssec:preview}, we give a simple estimate of the results that
354: we will later (in \S~\ref{sec:results}) calculate in detail.
355:
356: \subsection{Relevant Processes}
357: \label{ssec:processes}
358: A planet transiting in front of its star affects the starlight that ultimately
359: reaches Earth in many ways. The motion of the planet's atmosphere
360: (rotation and winds) is a small perturbation on top of several more
361: dominant effects. We therefore summarize below the physical processes
362: that are at least as significant as the effect of tidally locked rotation.
363: Figure~\ref{fig:schematic} schematically represents this situation, and
364: captures nearly all of the processes described below:
365: a rotating planet (of exaggerated relative size) transits in front of a
366: rotating star. The figure depicts a snapshot partway through ingress, when
367: half of the planet is in front of the stellar disk\footnote{The planet is
368: above the star's equator to represent a slight inclination in its orbit.}.
369: The white circle indicates a hypothetical sharp demarcation between the opaque
370: part of the planet (in black) and the optically thin part, labeled
371: ``Atmosphere'' (described further below).
372: \begin{figure}[h!]
373: \plotone
374: {fig1.eps}
375: \caption{Rotating planet beginning to transit in front of rotating star.
376: The vertical black arrows represent the rotation axes of the planet and the
377: star, and the curved arrows indicate the direction of rotation for each.
378: The X's on the right--sides of both the planet and the star indicate regions
379: that are receding from the observer and are therefore redshifted;
380: the unmarked left sides of the planet and the star are moving toward the
381: observer and are therefore blueshifted.
382: The white circle surrounding the opaque black part of the planet denotes
383: the cloud deck, or the boundary between the partially transparent and the
384: fully opaque portions of the planet's disk. The planet is orbiting in the
385: same sense as both it and the star are rotating. The planet is shown above
386: the star's midplane to represent the inclination of the orbit relative to
387: the line--of--sight.}
388: \label{fig:schematic}
389: \end{figure}
390:
391: \begin{enumerate}
392: \label{list:transiteffects}
393: \item {\it Geometric Occultation}:\\
394: The largest effect is an overall dimming by a factor of roughly the
395: ratio of the area of the planet to that of the star: $(R_p/R_*)^2$. Since
396: stars are not perfectly uniform disks, but instead tend to darken toward
397: the limb at most visible wavelengths, the fractional dimming due to being
398: in the planet's shadow tends to be slightly less than the ratio of the
399: areas when the planet is near the edge of the stellar disk and slightly
400: more than this ratio when the planet is near the center.
401: \item {\it Stellar Wobble}:\\
402: The primary spectral effect of the planet orbiting the star is the radial
403: velocity wobble induced by the planet's gravity. This periodic spectral
404: shift is of course in effect during the transit, when, for a close--in planet
405: like HD209458b, it has an influence on the order of
406: $\sim \pm 10 {\rm ~m~s^{-1}}$.
407: This effect is a redshift as the planet begins to transit across the disk
408: (during ingress) and a blueshift during egress.
409: \item {\it Rossiter--McLaughlin Effect}:\\
410: A more subtle effect arises because, during the transit, the planet moves
411: across -- and therefore blocks -- parts of the star that have different
412: recessional velocities. If (as is expected) the planet's orbit is aligned
413: with the star's spin, then during ingress the planet is blocking a part of
414: the star that is slightly blueshifted, and during egress it is blocking a part
415: of the star that is slightly redshifted. Figure~\ref{fig:schematic}
416: illustrates the planet blocking some of the bluest parts of the star
417: during ingress.
418:
419: The parts of the star that are occluded during ingress/egress have spectra
420: that are blue/redshifted by a velocity that is approximately the equatorial
421: rotational speed of the star, or about $\sim 1$-$2 {\rm ~km~s^{-1}}$ for a
422: Sun--like star.
423: As the figure indicates, during ingress/egress, the integrated
424: spectrum of the remaining (unblocked) parts of the star is on average slightly
425: redder/bluer than it would be if the planet were entirely transparent.
426: Therefore, during ingress, the centroids of stellar lines are shifted
427: slightly to the red, and during egress the centroids are correspondingly
428: shifted to the blue.
429:
430: This so--called Rossiter--McLaughlin effect (RME), described originally by
431: \citet{rossiter1924} and \citet{mclaughlin1924} in the case of eclipsing binary
432: stars, adds to the shifts already caused by the radial velocity induced by
433: the planet's gravity, described in (2.) above. The RME has been described in
434: depth more recently in the context of extrasolar planets by
435: \citet{ohta_et_al2005}, \citet{gimenez_et_al2006}, and \citet{gaudi+winn2007}.
436: These centroid--shifts are expected to be comparable in magnitude to the radial
437: velocity wobble from the planet's gravity, and can be roughly estimated as
438: \[
439: \left| \delta v_{\rm R-M} \right| \sim 1 {\rm ~km~s^{-1}} \times (R_p/R_*)^2 \sim 10 {\rm ~m~s^{-1}},
440: \]
441: In fact, the amount of the shift can be predicted precisely for a given
442: orientation of the planet's orbit, and so measuring the shift is
443: tantamount to measuring the alignment between the star's spin and the planet's
444: orbit. Three years ago, \citet{winn_et_al2005} first found that the spin of
445: HD209458 and the orbital plane of its planet are nearly aligned. The degree
446: of alignment has been measured for two other systems -- \citet{winn_et_al2006}
447: found that the spin of HD189733 and its planet are also nearly aligned,
448: and \citet{narita_et_al2007} measured a mis--alignment between these two
449: vectors by $\sim (30 \pm 20)\degr$ in the TrES--1 system.
450: \item {\it Planet's Atmospheric Opacity}:\\
451: Furthermore, a gas--giant planet's opacity surely does not have a
452: perfectly sharp discontinuity at an outer boundary, as a billiard ball does.
453: Instead, it has an extended atmosphere in which the opacity must vary more or
454: less smoothly. There may be a cloud layer, below which the planet is
455: entirely opaque to tangential rays and above which the opacity varies smoothly.
456: Most critical to our investigation, at a given radius, the planet's opacity
457: to tangential lines of sight must vary with wavelength, depending on the
458: contents of its atmosphere. At wavelengths of strong atomic or molecular
459: transitions, the planet's atmosphere will be more opaque than at other
460: wavelengths. As a result, the effective radius of the planet, or the
461: radius at which the optical depth along a tangential ray is of order unity,
462: is greater at some wavelengths than at others. These effects have been
463: described in detail by B01.
464: \item {\it Planet's Orbital Motion}:\\
465: The motion of the planet's atmosphere must influence the transit spectrum in
466: several delicate ways. As B01 points out, there are three main mechanisms by
467: which the motion of a planet's atmosphere relative to its star can affect
468: the spectrum: the planet's orbital velocity along the line--of--sight, the
469: planet's (possibly tidally locked) rotation, and winds in its atmosphere.
470: The largest effect of these three is the orbital velocity, which imposes a
471: bulk blue/redshift during ingress/egress of $\sim 15 {\rm ~km~s^{-1}}$ to
472: spectral lines arising from the planet's atmosphere. These shifts are of
473: opposite sign to the radial velocity wobble and to the shifts from the RME, and
474: therefore tend to lessen the apparent RME slightly.
475: \item {\it Planet's Atmospheric Motion}:\\
476: The most dynamically interesting (and subtlest) effects are those caused by
477: the planetary rotational velocity and atmospheric winds. Since a tidally
478: locked planet rotates in the same sense as it orbits, the rotational velocity
479: of its outside edge has the same sign as its orbital velocity, and the
480: rotational velocity of its inside edge has the opposite sign. As a result,
481: during the beginning of ingress and the end of egress, when only the inside
482: edge of the planet is over the star, tidally locked rotation will impose a
483: spectral distortion that is in the opposite sense of that caused by the bulk
484: orbital velocity described in (5.) above, and that is in the same sense
485: as the RME: the distortions are roughly equivalent to a relative redshift
486: during ingress (graphically represented in Figure~\ref{fig:schematic})
487: and a relative blueshift during egress.
488: During mid--transit, with some parts of the rotating planet's atmosphere
489: moving toward and other parts away from the star relative to an otherwise
490: identical but non--rotating planet, the overall influence of the planet's
491: rotation is approximately equivalent to rotational broadening.
492:
493: Winds complicate the picture even further. It is likely that winds tend
494: to rush from the substellar hot spot to the colder night side of the planet.
495: With the substellar point on the opposite side of the planet from Earth
496: during a transit, this corresponds to winds rushing toward us at several
497: hundred to several thousand meters per second. This would tend to
498: blueshift the spectrum throughout the transit. Zonal wind bands, somewhat
499: similar to those on Jupiter but with much higher speeds, or other more detailed
500: winds, can have an even more intricate effect.
501: \item {\it Additional Effects}:\\
502: If a transiting planet were to have nonzero orbital eccentricity, or rings,
503: these could complicate a measurement of rotation rate. Nonzero eccentricity
504: would break the symmetry between ingress and egress. Still, if the orbit were
505: well-known, this could be modeled and taken into account. It seems unlikely
506: that a Hot Jupiter could maintain rings: Icy rings would sublimate,
507: and, if not continuously replenished, dusty/rocky rings would quickly
508: succumb to the Poynting--Robertson effect \citep{poynting1903,robertson1937}.
509: But if, somehow, a ring were to find a way to persevere around a transiting
510: Hot Jupiter, it could confound -- perhaps hopelessly -- a measurement of
511: rotation. The consequences of rings for the Rossiter--McLaughlin effect is
512: addressed in \citet{ohta_et_al2006}. Saturn's rings are nearly four--times the
513: area of the planet, so for a planet (with equatorial rings that are as
514: relatively large as Saturn's) whose orbit is tilted an angle 0.1 (in radians)
515: from edge-on, the rings would be $\sim 40\%$ the area of the planet, which
516: would increase the RME by $\sim 40\%$. Uncertainty about the presence and
517: size of a ring introduces an uncertainty in the size of the RME effect that is
518: probably larger than the size of the rotation effect. Furthermore, a ring
519: would occlude a (small) part of the planet's atmosphere, which would (slightly)
520: reduce the strength of the rotation signal.
521: \end{enumerate}
522:
523: Other interesting phenomena that primarily affect a transit light curve,
524: rather than the spectrum, include star--spots \citep{silva2003}, atmospheric
525: lensing \citep{hui+seager2002}, and finite--speed--of--light effects
526: \citep{loeb2005}. Although \citet{winn+holman2005} describe a possible
527: configuration (Cassini state 2) that would produce a spectral signature
528: that is nearly identical to what would be expected from a non--rotating
529: planet, the likelihood that any Hot Jupiters are in such a configuration
530: might be low, and it seems quite likely that {\it some} transiting
531: planets are not in this state. Nonetheless, the motion of a transiting
532: planet's atmosphere -- rotational, wind, or other -- is clearly interesting,
533: and the basic technique that we describe below is applicable to any model of
534: atmospheric motion.
535:
536: \subsection{Preview of Results}
537: \label{ssec:preview}
538: A rough estimate of the velocity--shift that is imposed during ingress to the
539: centroids of the stellar Na D--lines by the planet's tidally locked rotation
540: (on top of the RME and the shift from the planet's orbital velocity, both of
541: which would be present even if the planet were not rotating) is the following:
542: \begin{eqnarray}
543: \nonumber \delta v & \sim & \left(\left< \cos[\phi] \right>_{-\pi/2}^{~\pi/2}\right) \times \left( \frac{1}{2} \times \frac{{R_p}^2}{{R_*}^2}\right) \times \left( \frac{2 \pi R_p \Pi_{\rm atm}}{\pi {R_p}^2} \right) \times v_{\rm eq} \\
544: \nonumber & \sim & 0.64 \times 1\% \times 15\% \times 2000 {\rm ~m~s^{-1}} \\
545: \label{eq:Delta v estimated} & = & 1.9 {\rm ~m~s^{-1}}.
546: \end{eqnarray}
547: In this equation, $\phi$ is a planet--centered azimuthal angle,
548: $R_p$ and $R_*$ are the planet's and star's radius, respectively,
549: $\Pi_{\rm atm}$ is the height of the planet's atmosphere, and $v_{\rm eq}$
550: is the equatorial rotation speed. The rotation speed at angle $\phi$ is
551: $v_{\rm eq} \cos[\phi]$. We take the average of $\cos[\phi]$ from
552: $-\pi/2$ to $\pi/2$ to get the average planetary rotation speed.
553: We have used $\Pi_{\rm atm} = 7500 {\rm ~km}$, or 15 times the
554: presumed scale height of $500 {\rm ~km}$, because the sodium lines
555: are so heavily saturated that at the assumed abundance and cloud deck
556: height in our model the line cores do not become optically thin until
557: that height. \citet{burrows_et_al2004} and \citet{fortney2005} describe
558: how the optical depth along tangential rays is greater than the optical depth
559: on normal rays. The product
560: \[
561: \delta_{\rm atm} \approx \left( \frac{1}{2} \times \frac{{R_p}^2}{{R_*}^2}\right) \left( \frac{2 \pi R_p \Pi_{\rm atm}}{\pi {R_p}^2} \right) = \left(\frac{R_p}{R_*}\right)^2 \left( \frac{\Pi_{\rm atm}}{R_p} \right)
562: \]
563: is the ratio of the area of the portion of the planet's atmosphere that is in
564: front of the star halfway through ingress to the total area of the disk of the
565: star. Based on this estimate, we expect a maximum velocity shift of
566: $\delta v \sim 190 {\rm ~cm~s^{-1}}$. If we take into account that HD209458b's
567: orbit is actually slightly inclined relative to the line of sight, the cosine
568: average decreases to $\sim 0.45$, and the total estimate decreases to
569: $\sim 140 {\rm~cm~s^{-1}}$. This estimate is in reasonably good agreement with
570: the centroid--shifts predicted by the full model calculation below
571: ($\sim 60 {\rm ~cm~s^{-1}}$); the
572: difference between the estimates is most likely due to the difference between
573: the shapes of the stellar and planetary lines.
574:
575: We now estimate the signal--to--noise ratio for the detectability of this
576: effect in an observation of duration $\Delta t$, with a telescope that has
577: diameter $D$ and throughput efficiency $\eta$. The signal is the distortion
578: of the spectrum relative to a non--rotating planet, and for now we will assume
579: that the noise is
580: dominated by photon noise. If a spectrum $F[\lambda]$ with a symmetric
581: absorption feature of depth $\Delta F$ centered at $\lambda_0$ is redshifted
582: by an amount $\Delta \lambda$ to
583: $\widehat{F}[\lambda] \equiv F[\lambda - \Delta \lambda]$, what is the
584: integrated absolute difference $|F-\widehat{F}|$ over some wavelength range
585: $2L$ centered on $\lambda_0$? If the absorption feature is wide compared
586: with $\Delta \lambda$, then, by symmetry,
587: \begin{eqnarray}
588: \nonumber S & = & \int_{\lambda_0 - L}^{\lambda_0 + L} \left| F[\lambda] - F[\lambda - \Delta \lambda] \right| d\lambda \\
589: \label{eq:signal integral} & \approx & 2 \int_{\lambda_0}^{\lambda_0 + L} \left( F[\lambda] - F[\lambda - \Delta \lambda] \right) d\lambda;
590: \end{eqnarray}
591: and if $\Delta \lambda$ is small then
592: \begin{eqnarray}
593: \nonumber S & \approx & 2 \Delta \lambda \int_{\lambda_0}^{\lambda_0 + L} F'[\lambda] d\lambda \\
594: \label{eq:integrated diff} & \approx & 2 (\Delta \lambda) (\Delta F).
595: \end{eqnarray}
596:
597: We may now estimate the S/N of our effect (for a single absorption line) using
598: the lesson of equation~(\ref{eq:integrated diff}), provided we know the
599: absolute normalization of the stellar spectrum (the number of photons per unit
600: wavelength). A spherical blackbody of radius $R_*$ and temperature $T_*$,
601: at distance $d$ from the telescope, has a photon number flux at wavelength
602: $\lambda$ of
603: \begin{eqnarray}
604: \nonumber \frac{d \dot{N}_\gamma}{d \lambda} & \sim & B_\lambda[\lambda,T] \left( \frac{1}{h c/\lambda} \right) \left( \frac{\pi {R_*}^2}{d^2} \right) \times \eta \pi (D/2)^2\\
605: \label{eq:spectrum} & = & \frac{\pi^2 c}{2 \lambda^4 \left( \exp[(h c)/(\lambda k T_*)] -1 \right)} \times \eta \left(\frac{{R_* D}}{d}\right)^2,
606: \end{eqnarray}
607: where $B_\lambda$ is the Planck function.
608: Since the fractional decrease in the spectrum at the line--center is
609: approximately $\delta_{\rm atm}$, we may express the parameter $\Delta F$
610: from equation~(\ref{eq:integrated diff}) as
611: $\Delta F \approx \delta_{\rm atm} (d \dot{N}_\gamma / d \lambda)$.
612: Similarly, since the root--mean--square velocity shift during ingress is
613: $\left< v^2 \right>^{1/2} \sim (1/2) \times( 2000 {\rm~m~s^{-1}}) = 1000 {\rm ~m~s^{-1}}$,\footnote{We write $(1/2) \times (2000 {\rm~m~s^{-1}})$ because the
614: mean value of $\cos^2$ from $-\pi/2$ to $\pi/2$ is $1/2$}
615: we may express the parameter $\Delta \lambda$ as
616: $\Delta \lambda \sim (\left< v^2 \right>^{1/2} / c) \times \lambda_0$.
617: The distortion (the signal) from a single line can therefore be estimated as
618: \begin{eqnarray}
619: \nonumber S & = & \delta N_\gamma \sim 2 (\Delta \lambda) \left( \delta_{\rm atm} \frac{d \dot{N}_\gamma}{d \lambda}\right) \Delta t\\
620: \label{eq:signal estimate} & = & \frac{\pi^2 c \left(\delta_{\rm atm}\right) \Delta \lambda}{\lambda^4 \left( \exp[(h c)/(\lambda k T_*)] -1 \right)} \times \eta \left(\frac{{R_* D}}{d}\right)^2 (\Delta t).
621: \end{eqnarray}
622: The shot--noise is the square root of the number of photons in a wavelength
623: range $2L$ roughly equal to the FWHM of the line, or about $7 {\rm ~km~s^{-1}}$
624: for a heavily saturated line such as the Na D lines under consideration:
625: \begin{eqnarray}
626: \nonumber N & \sim & \sqrt{\frac{d \dot{N}_\gamma}{d\lambda} (2L) (\Delta t)} \\
627: \label{eq:noise estimate} & \sim & \sqrt{\frac{\pi^2 L c (\Delta t)}{\lambda^4 \left( \exp[(h c)/(\lambda k T_*)] -1 \right)}} \times \sqrt{\eta} \left( \frac{R_* D}{d} \right).
628: \end{eqnarray}
629: We estimate the total signal--to--noise ratio arising from a single
630: absorption line, during an ingress integration of duration $\Delta t$, to be
631: roughly
632: \begin{eqnarray}
633: \nonumber S/N & \sim & \frac{\pi(\delta_{\rm atm})}{\sqrt{\exp[(h c)/(\lambda k T_*)] -1}} \left( \frac{ \Delta \lambda}{\lambda} \right) \left( \sqrt{\frac{c \Delta t}{ L}} \right) \left( \frac{R_* D}{d \lambda} \right) \sqrt{\eta} \\
634: \nonumber & \sim & \left( 6.6\times 10^{-4} \right) \left( 3.3\times 10^{-6} \right) \left( 2.1\times 10^{11} \right) \left( 5.0\times 10^{-3} \right) \sqrt{\eta} \\
635: \label{eq:SN estimate} & \sim & 2.3 \sqrt{\eta}.
636: \end{eqnarray}
637: The above calculation uses parameters for HD209458 and its planet, a
638: sodium D line, and a 6~m telescope:
639: $\lambda = 600 {\rm~nm}$; $\Delta t = 1000 {\rm~s}$;
640: $R_* = 7.3\times 10^{10} {\rm~cm}$;
641: $T_* = 6100 {\rm ~K}$, $d = 47 {\rm~pc}$;
642: and $D = 600 {\rm~cm}$. For two identical
643: absorption lines, we gain a factor of $2^{1/2}$ in S/N, and for egress we gain
644: another factor of $2^{1/2}$, giving a total one--transit S/N of roughly
645: $4.6 \eta^{1/2}$, not counting the additional signal available during
646: mid--transit (see further discussion below). This S/N ratio is in principle
647: independent of the spectral resolution of the spectrograph, for sufficiently
648: high spectral resolution. For low spectral resolution, however, the S/N could
649: be lower than this estimate (below, we conclude that the S/N loses its
650: dependence on resolving power for spectral resolution $\gsim$~500,000).
651:
652: There were several optimistic assumptions that went into this estimate.
653: Still, this rough estimate of the degree to which a planet's rotation
654: influences its transit spectrum indicates that the more in--depth study
655: that we perform below is warranted.
656:
657: \subsection{Available Technology}
658: \label{ssec:tech}
659: Detecting the centroid--shifts caused by tidally locked rotation
660: ($\lsim 1 {\rm ~m~s^{-1}}$) will require very precise measurements of
661: stellar transit spectra. Obtaining such high precision spectra will be
662: quite challenging, for a number of reasons, several of which were
663: described in the groundbreaking paper by \citet{butler_et_al1996}
664: that analyzes the limits of Doppler precision. Of particular concern,
665: stellar pulsations and turbulent motions in stellar photospheres can cause
666: small regions of the stellar disk to move at up to $300 {\rm ~m~s^{-1}}$
667: \citep{dravins1985, ulrich1991}. These motions tend to average out to produce
668: stellar spectra that are largely stable; but it is likely that at least some
669: giant convection cells are not small relative to the size of a planet, and
670: these could introduce a contaminating source of noise when they are located
671: behind the planet or its atmosphere.
672: \citet{butler_et_al1996} reviewed what was then known about the
673: variability of stellar line--profiles; the upshot is that line--widths may
674: vary by up to several meters per second over several years, but it is not
675: clear to what extent spurious apparent velocity shifts may be induced by
676: convection, and such stellar jitters may prove to be a significant source of
677: noise that would render it difficult to measure sub meter--per--second
678: velocity--shifts. More recently, \citet{bouchy_et_al2005b} have actually
679: achieved sub meter--per--second accuracy with the HARPS instrument (spectral
680: resolution of 115,000), and they have found a dispersion in night--averaged
681: radial velocity measurements for a particular star (HD160691) of
682: $\sim 0.4 {\rm ~cm~s^{-1}}$ for nights when they took many ($\gsim 200$)
683: observations. Since in our situation (taking spectra during ingress, say) we
684: have minutes, not hours, available, the rms scatter in ingress--averaged radial
685: velocity measurements is likely to be larger than what they found. In addition
686: to the difficulties posed by several systematic sources of noise, achieving
687: sufficient photon statistics will be difficult for two reasons: for a given
688: throughput efficiency $\eta$, higher spectral resolution means fewer photons
689: per bin; and $\eta$ tends to decrease with increasing spectral resolution
690: $R_S$.
691:
692: By the mid--1990s, the timeless quest for high--resolution spectrographs
693: reached a milestone at the Anglo--Australian Telescope with the development
694: of UHRF and its resolving power of up to $1,000,000$ \citep{diego_et_al1995}.
695: Despite impressive throughput relative to previous endeavors, however, its
696: efficiency was insufficient to obtain the sub decameter--per--second Doppler
697: precision on a $V \ge 7$ star that would be required for planet
698: searches. With a $R_S = 600,000$ spectrograph built at Steward Observatory,
699: \citet{ge_et_al2002} obtained stellar spectra with $R_S \sim 250,000$ and
700: throughput of 0.8\%. Furthermore, they predicted that by optimizing their
701: technology they could increase the throughput to 4\%. More recently,
702: \citet{ge_et_al2006} detected a new planet, around HD 102195, with the
703: Exoplanet Tracker instrument at Kitt Peak. This instrument has resolution
704: of $R_S \sim 60,000$ and total throughput of 18\%.
705: Plans for a spectrograph that has resolving power of 120,000 on a
706: thirty meter telescope \citep{tokunaga_et_al2006} give cause for optimism
707: that increased aperture area and efficiency feeding high and
708: ultrahigh--resolution spectrographs will, in coming years, provide accurate
709: enough spectra that tidally locked rotation of HD209458b has a detectable
710: influence.
711:
712:
713:
714: \section{A Model of a Planetary Transit}
715: \label{sec:model}
716: We consider the spectrum of a star whose companion planet transits across the
717: face of the stellar disk from Earth's perspective. The primary effect of the
718: planet is to reduce the stellar flux at all wavelengths, but the planet's
719: chemical composition, internal structure, and rotation rate influence
720: the spectrum in wavelength--dependent ways. Since each
721: of these factors -- and others too, such as the star's rotation --
722: influences the observed spectrum, we built a model that incorporates
723: the many parameters related to each process.
724: The star and the planet are both assumed to rotate as solid bodies, with no
725: other (nonthermal) motion in their atmospheres. Since deviations from
726: pure solid body rotation are likely to be no more than 25\% over the disk of
727: the star -- e.g., the Sun's equator--to--pole variation in rotation rate
728: is about 21\%, as per \citet{howard_et_al1984}, this is probably a reasonable
729: assumption for the star. For the planet, this assumption might fail,
730: because wind--speeds in excess of the equatorial rotation speed of
731: $v_{\rm eq} \approx 2 {\rm ~km~s^{-1}}$ are predicted by many models, as
732: described in \S~\ref{sec:intro} above. Still, when making this initial study
733: of the spectral effect of the motion of a transiting planet's atmosphere,
734: separating rotation from other processes makes the problem more tractable.
735: We set parameter values to match measured values from the HD209458b system
736: where possible.
737:
738: The planet is modeled as an inner component that is entirely opaque and an
739: outer component that is isothermal and drops off exponentially. We compute
740: the wavelength--dependent optical depth due to the sodium D-doublet at
741: $\approx 590$~nm in the planet's atmosphere;
742: important parameters include the temperature and density of the planet's
743: atmosphere and its Na--content. We use the Voigt profile -- as described
744: by, e.g., \citet{press+rybicki1993} -- to calculate $\tau[\lambda]$, the
745: optical depth to absorption along the line of sight.
746:
747: As the planet transits the star, there are four points of
748: ``contact'' between the planet and the star (really between their projections
749: on the sky): when the disk of the planet first touches the disk of the
750: star; when the planet is first entirely over the stellar disk; when the planet
751: is last entirely over the stellar disk; and when the planet last
752: touches the stellar disk. We will additionally sometimes refer to
753: ``$1.5^{\rm th}$'' contact (half--way between first and second contact),
754: and analogously to ``$2.5^{\rm th}$'' and ``$3.5^{\rm th}$'' contact.
755:
756: As described in \S~\ref{sec:overview} above, the type of distortion that a
757: planet's rotation imposes relative to a non--rotating planet changes
758: depending on when during the transit the observation is made. During
759: ingress or egress, the rotation of a tidally locked planet's atmosphere will
760: impose a distortion similar to an overall shift relative to a non--rotating planet: redshift during ingress; blueshift during egress.
761: When the planet is in mid--transit, in the middle of the stellar disk,
762: the overall distortion to the spectrum imposed by its rotation is
763: akin to a star's rotational broadening. Since the line--centers of the lines
764: we are considering are heavily saturated and therefore flat at their
765: cores, rotational broadening has the somewhat counterintuitive effect
766: of steepening the cores of the profiles while broadening the wings.
767: We will discuss this in greater detail in the next section.
768: Although the type of distortion is different during ingress and egress from
769: during mid--transit, it turns out that the amount of distortion, in terms of
770: S/N ratio, is nearly constant throughout transit. This, too, we will discuss
771: in \S~\ref{sec:results} below.
772:
773: We simulate the HD209458b system, with a $1.32 R_J$ planet in a 3.5 day orbit,
774: orbiting a G0 star at with radius $1.05 R_\sun$ that is 47~pc away. Our
775: model star has the limb darkening profile that \citet{knutson_et_al2007a}
776: measured for HD209458. In order to approximate the fits to the data in
777: \citet{charbonneau_et_al2002}, we assign our model planet's atmosphere a
778: sodium--content and cloud deck height (1\% solar, and 0.01~bars) that are
779: comparable to the parameter--combinations that result in the best fits in
780: that paper. Finally, we present results at our simulation's spectral resolution
781: ($R_S = 700,000$), and we simulate transit events observed using two different
782: lower resolution spectrographs, one with spectral resolution
783: $R^\prime_S = 50,000$ and one with $R^\prime_S = 150,000$.
784: All spectrographs (and associated optical paths) in our simulations have
785: 100\% throughput efficiency. In the remainder of this section, we provide
786: a detailed description of our parameterization of the problem.
787:
788:
789: \subsection{Parameters of the Star}
790: \label{ssec:pstar}
791: The parameters related to the star are listed in Table~\ref{ta:pstar}.
792: They are set to match measured parameters for HD209458, and we use the
793: limb--darkening profile from \citet{knutson_et_al2007a}. We normalize
794: the flux to that of a blackbody of temperature $T_*$ of the size and at
795: the distance of HD209458.
796:
797: \begin{deluxetable}{llr}
798: \tablewidth{400pt}
799: \tablecolumns{3}
800: \small
801: \tablecaption{Model Transit Parameters: Star}
802: \tablehead{\colhead{Parameter} & \colhead{Description} & \colhead{Value}}
803: \startdata
804: $M_*$ & Star Mass & 1.05 $M_\sun = 2.09\times 10^{33} {\rm~g}$\\
805: $R_*$ & Star Radius & 1.05 $R_\sun = 7.35\times 10^{10} {\rm~cm}$\\
806: $T_*$ & Star Temperature & 6100 K\\
807: $d_*$ & Distance to star & 47 pc\\
808: $\tau_*$ & Stellar Rotation Period & 1 month\\
809: \enddata
810: \label{ta:pstar}
811: \vspace{-0.4cm}
812: \end{deluxetable}
813:
814:
815: \subsection{Parameters of the Planet}
816: \label{ssec:pplanet}
817: The parameters related to the planet are in Table~\ref{ta:pplanet}.
818: We model the planet as an inner component that is essentially a billiard ball
819: (completely opaque at all wavelengths) and an outer component that is an
820: isothermal atmosphere with scale height $H = R_{\rm gas}T_p/\mu g$, where
821: $R_{\rm gas}$ is the gas constant, $\mu$ is the molar mass, and $g$ is the
822: acceleration of gravity. The density of our model planet's atmosphere varies
823: as $\rho = \rho_0 \exp[(r-{R_p}_0)/H]$, where ${R_p}_0$ is the radius of
824: the optically thick part (some authors have called this radius the
825: ``cloud--deck'' \citep{charbonneau_et_al2002}). This hypothetical
826: cloud deck could cause the planet to be optically thick at a higher
827: altitude than would otherwise be expected, as discussed in,
828: e.g., \citet{richardson_et_al2003} and \citet{sudarsky_et_al2000}. The cloud
829: deck causes the optical depth as a function of radius in our model to have
830: a singular discontinuity at radius ${R_p}_0$.
831:
832: \begin{deluxetable}{llr}
833: %\tablewidth{300pt}
834: \tablecolumns{3}
835: \small
836: \tablecaption{Model Transit Parameters: Planet}
837: \tablehead{\colhead{Parameter} & \colhead{Description} & \colhead{Value}}
838: \startdata
839: $M_p$ & Planet Mass & $0.69 M_J = 1.31\times 10^{30} {\rm~g}$\\
840: ${R_p}_0$ & Optically Thick Planet Radius & $1.32 R_J = 9.44\times 10^5 {\rm~km}$\\
841: $P_0$ & Planet Pressure at ${R_p}_0$ & 0.01 bars\\
842: $H$ & Planet Atmosphere Scale Height & 500 km\\
843: $T_p$ & Planet Atmosphere Temperature & 1300 K\\
844: $f_{\rm TL}$ & Frac. Tidal Locked Rot. Rate & \hspace{1in} 0 or 1 ($v_{\rm eq} = 0$ or $2 {\rm~km~s^{-1}}$)\\
845: $a$ & Semi--Major Axis & 0.046 AU\\
846: $\#_H$ & Number of Scale Heights in Atm. & 15\\
847: \enddata
848: \label{ta:pplanet}
849: \vspace{-0.4cm}
850: \tablenotetext{a}{Parameter values are set to match measured values from
851: the HD209458b system where possible.
852: }
853: \end{deluxetable}
854:
855:
856: \subsection{Spectral Parameters}
857: \label{ssec:pspec}
858: The parameters pertaining to the shape of the observed spectrum are in
859: Table~\ref{ta:pspec}. In addition to the location of the planet within the
860: stellar disk, the shape of the stellar spectrum and the wavelength--dependent
861: opacity of the planet's atmosphere together influence the transmission
862: spectrum.
863:
864: {\tt Spec\_Shape} is a parameter that can take on the values
865: ``Flat'', ``Blackbody'', or ``Solar'', and determines the rest--frame spectrum
866: of the model stellar photosphere. (The integrated stellar spectrum is the
867: convolution of the rest--frame spectrum with the stellar rotation profile.)
868: When ``Flat'' is chosen, the rest--frame model stellar spectrum intensity is
869: set to the mean value of the blackbody intensity in the specified wavelength
870: range $[\lambda_{\rm min},\lambda_{\rm max}]$, which, in our simulation, is
871: set to $[580 {\rm~nm}, 600 {\rm~nm}]$. When ``Solar'' is chosen, the
872: model stellar spectrum instensity is set to a high--resolution solar spectrum
873: that is normalized to the flux from
874: HD209458\footnote{From {\it ftp://solarch.tuc.noao.edu/}.}; but the Na D
875: lines in this high--resolution spectrum have been replaced by
876: Gaussian fits to the solar lines.
877:
878: The planet's atmosphere has $N_{\rm abs}$ absorption features, each of which
879: is due to an element with a given fraction of the solar abundance. In
880: the models presented in this paper, $N_{\rm abs} = 2$: we consider the Na
881: doublet at 588.9950 nm and 589.5924 nm, with sodium at fractional abundance
882: $f_\sun \equiv {X_{\rm Na~}}_p / {X_{\rm Na~}}_\sun = 0.01$ of
883: the solar abundance. Each line is modeled as a Voigt profile, as described
884: in, e.g., Press \& Rybicki (1993).
885:
886: \begin{deluxetable}{llr}
887: %\tablewidth{300pt}
888: \tablecolumns{3}
889: \small
890: \tablecaption{Model Transit Parameters: Spectral Features}
891: \tablehead{\colhead{Parameter} & \colhead{Description} & \colhead{Value}}
892: \startdata
893: {\tt Spec\_Shape} & Shape of Star Spectrum & Flat, Blackbody, or Solar\\
894: $\lambda_{\rm min}$ & Min. Wavelength in Sim. & 580~nm\\
895: $\lambda_{\rm max}$ & Min. Wavelength in Sim. & 600~nm\\
896: $N_{\rm abs}$ & \# Abs. Features in P. Atm. & 2\\
897: ${f_\sun}_1$ & Frac. Solar Abund., First Line & $0.01$\\
898: ${\lambda_0}_1$ & First Line--Center & 588.9950 nm\\
899: ${A_{ki}}_1$ & Transition Prob. First Line & $6.16\times 10^{-9} {\rm~s^{-1}}$\\
900: ${g_i}_1$ & Stat. Wt. Lower Level First Line & 2\\
901: ${g_k}_1$ & Stat. Wt. Upper Level First Line & 4\\
902: ${f_\sun}_2$ & Frac. Solar Abund., Second Line & $0.01$\\
903: ${\lambda_0}_2$ & Second Line--Center & 589.5924 nm\\
904: ${A_{ki}}_2$ & Transition Prob. Second Line & $6.14\times 10^{-9} {\rm~s^{-1}}$\\
905: ${g_i}_2$ & Stat. Wt. Lower Level Second Line & 2\\
906: ${g_k}_2$ & Stat. Wt. Upper Level Second Line & 2\\
907: \enddata
908: \label{ta:pspec}
909: \vspace{-0.4cm}
910: \tablenotetext{a}{In parameters that have $i$ and $k$ subscripts,
911: $i$ indicates the lower level ($3 s_{1/2}$ for both lines) and $k$
912: indicates the upper level ($3 p_{3/2}$ for the bluer line and
913: $3 p_{1/2}$ for the redder line). The fractional solar abundance is set
914: to 0.01 in order to achieve modest agreement with data observed for the Na D
915: doublet in HD209458b's atmosphere.
916: }
917: \end{deluxetable}
918:
919:
920: \subsection{Parameters of Observing and Computing}
921: \label{ssec:pobscomp}
922: The final set of parameters, listed in Table~\ref{ta:pobscomp}, includes
923: those that specify the observer and those that determine how the
924: observation is discretized for the purpose of numerical computation.
925: The model observational setup is determined by three parameters:
926: the telescope's diameter $D$ (6~m in our simulations) and efficiency
927: $\eta$ (100\%), and the spectrograph's spectral resolution $R_S$ (we set
928: $R_S$ to 700,000 for the purpose of computing the model, and we
929: re--bin to lower, more easily achieved resolutions -- 150,000 and
930: 50,000 -- after computing a model). These three parameters prescribe
931: the sizes of the spectral bins and the rate at which those bins
932: are capable of collecting light.
933:
934: In order to compute the flux at Earth as a function of wavelength, we
935: begin by dividing the stellar disk into ${n_b}_*$ concentric annuli,
936: and we divide each annulus into ${n_\phi}_*$ azimuthal sections. In
937: each section, the redshifted spectrum and the normalization must
938: both be computed.
939:
940: Knowing the stellar rotation rate and axis, we may calculate the recessional
941: velocity of any point on the star's surface as a function of its location on
942: the projected stellar disk, and we redshift the spectrum from each part
943: of the star accordingly.
944: When the planet is in--transit, we separate the stellar disk into an annulus
945: that contains the planet and the rest of the disk that we treat as described
946: above. The annulus that contains the planet is treated almost as above --
947: divided into ${n_b}_{*A}$ sub--annuli, each of which has ${n_\phi}_{*A}$
948: azimuthal sections -- but the sub--annuli are incomplete, interrupted by
949: the planet.
950:
951: In order to sample the planet's atmosphere, we divide the region that overlaps
952: the star into ${n_b}_p$ concentric annuli around the planet's center,
953: each of which is divided into ${n_\phi}_p$ azimuthal sections. In each
954: section, we must determine the optical depth and multiply by
955: $\exp{\left(-\tau\right)}$. In calculating the optical depth, we note
956: that in the case that the planet's rotation axis is entirely normal to the
957: line--of--sight, if the planet rotates as a solid body then the radial
958: component of its recessional velocity is constant along a ray:
959: \begin{equation}
960: \tau[b_p, \phi, \lambda] = N[b_p] \times \sigma\left[\frac{\lambda}{1 + (v_p[b_p, \phi_p] /c)}\right],
961: \label{eq:tau}
962: \end{equation}
963: where the column density is calculated in terms of a function $G$ that
964: is specified below: $N[b_p] = n_0 G[b_p, {R_p}_0, H]$. In
965: equation~(\ref{eq:tau}), $v_p[b_p, \phi_p]$ is the recessional velocity of
966: the planet, as a function of radius and azimuth, which depends upon
967: the orbit and the rotation. Note that there is a single $v_p$ along a given
968: line--of--sight defined by a $(b_p,\phi_p)$ pair only under the assumption of
969: solid body rotation. The rest--frame cross--section $\sigma[\lambda]$ is
970: computed according to the Voigt profile. The function $G$ is defined as
971: the following
972: integral:
973: \begin{equation}
974: G[b_p, {R_p}_0, H] = \left\{ \begin{array}{ll}
975: \int_{-\infty}^{\infty} \exp\left[-\frac{\sqrt{b^2 + l^2} - {R_p}_0}{H}\right] d l & b_p > R_p \\
976: \infty & b_p \leq R_p
977: \end{array} \right. .
978: \label{eq:funcG}
979: \end{equation}
980:
981: \begin{deluxetable}{llr}
982: %\tablewidth{300pt}
983: \tablecolumns{3}
984: \small
985: \tablecaption{Model Transit Parameters: Observation and Computing}
986: \tablehead{\colhead{Parameter} & \colhead{Description} & \colhead{Value}}
987: \startdata
988: $D$ & Telescope Diameter & $2.4 {\rm~m} - 30 {\rm~m}$\\
989: $\eta$ & Spectroscope Efficiency & $1.00$ \\
990: $R^\prime_S$ & Obs. Spec. Resolution & 50,000 - 700,000\\
991: $\mathcal{T}_{\rm int}$ & Integration Time & $932.088 {\rm~s}$\\
992: --------------&--------------------------------&---------------------------\\
993: $R_S$ & Comp. Spec. Resolution & 700,000\\
994: $\Delta t$ & Time--Step in Integration & $50 {\rm~s}$\\
995: ${n_b}_*$ & \# of Star Annuli & 10\\
996: ${n_\phi}_*$ & \# of Star Azimuthal Sections & 16\\
997: ${n_b}_{*A}$ & \# of S. Annuli in P. Annulus & 10\\
998: ${n_\phi}_{*A}$& \# of S. Azim. Sec.'s in P. Ann. & 10\\
999: ${n_b}_p$ & \# of Planet Atm. Annuli & 20\\
1000: ${n_\phi}_p$ & \# of Planet Atm. Azim. Sections & 20\\
1001: \enddata
1002: \label{ta:pobscomp}
1003: \vspace{-0.4cm}
1004: \tablenotetext{a}{Parameter values are set to match measured values from
1005: the HD209458b system where possible.
1006: }
1007: \end{deluxetable}
1008:
1009:
1010: \section{Model Transit Spectra}
1011: \label{sec:results}
1012: As described in \S~\ref{ssec:preview}, we seek the expected signal--to--noise
1013: ratio for distinguishing between the spectrum that would be observed due
1014: to a non--rotating planet (or one that is is in a Cassini state with its
1015: rotation axis nearly in the plane of orbit) and the spectrum that would be
1016: observed due to a tidally locked planet.
1017: The computed model spectrum $\mathcal{N}[\lambda]$ is the time integral
1018: of the instantaneous spectrum $\dot{\mathcal{N}}[\lambda]$ and consists
1019: of the number of photons detected per wavelength bin:
1020: \[
1021: \mathcal{N}[\lambda] \approx \dot{\mathcal{N}}[\lambda] \Delta t_{\rm obs},
1022: \]
1023: for some small exposure time $\Delta t_{\rm obs}$.
1024:
1025: The model signal (of rotation) per bin that we are looking for is the
1026: difference between the rotating model spectrum $\mathcal{N}_{\rm rot}$
1027: and the non--rotating model spectrum $\mathcal{N}_{\rm no~rot}$:
1028: \begin{equation}
1029: S_b = \left(\dot{\mathcal{N}}_{\rm rot}[\lambda] - \dot{\mathcal{N}}_{\rm no~rot}[\lambda]\right) \Delta t_{\rm obs}.
1030: \label{eq:signal}
1031: \end{equation}
1032: We make the optimistic approximation that the noise per bin is just the
1033: photon--noise:
1034: \begin{equation}
1035: N_b = \sqrt{\dot{\mathcal{N}}_{\rm no~rot}[\lambda] \Delta t_{\rm obs}}.
1036: \label{eq:noise}
1037: \end{equation}
1038: The total signal--to--noise ratio in a single exposure, then, is the sum in
1039: quadrature of $S_b/N_b$ for all wavelength bins $\lambda_i$:
1040: \begin{equation}
1041: S/N = \sqrt{\sum_{i = 1}^{\# \rm bins} \left( \frac{\dot{\mathcal{N}}_{\rm rot}[\lambda_i] - \dot{\mathcal{N}}_{\rm no~rot}[\lambda_i]}{\sqrt{\dot{\mathcal{N}}_{\rm no~rot}[\lambda_i]}} \right)^2 } \times \sqrt{\Delta t_{\rm obs}}.
1042: \label{eq:sn}
1043: \end{equation}
1044: A similar summation in quadrature applies over all exposures.
1045: Note that, in principle, the expression in equation~(\ref{eq:sn}) is
1046: insensitive to the sizes of bins and hence to the spectral resolution $R_S$,
1047: as long as the bins are small relative to the Gaussian width of the
1048: absorption feature under consideration. Our simulations indicate that the
1049: spectral resolution must be $\gsim 500,000$ in order for S/N to be nearly
1050: independent of $R_S$.
1051:
1052: The effect of rotation, both during ingress and during mid--transit,
1053: is illustrated in Figure~\ref{fig:spec. and diff.}.
1054: For illustrative purposes, in this figure we assume a uniform star (flat
1055: spectrum, non--rotating, no limb--darkening). On the left panels of
1056: Figure~\ref{fig:spec. and diff.}, we show a snap--shot during ingress
1057: (at $1.5^{\rm th}$ contact) and on the right, we show a snap--shot during the
1058: middle of a transit ($2.5^{\rm th}$ contact). The quantity plotted is
1059: $\mathcal{R}' = \mathcal{R}-1$ from B01, where
1060: \begin{equation}
1061: \mathcal{R}[\lambda,t] = \frac{\dot{N}_{\rm in~transit}[\lambda,t]}{\dot{N}_{\rm out~of~transit}[\lambda,t]}
1062: \label{eq:Brown R}
1063: \end{equation}
1064: The bottom panels of Figure~\ref{fig:spec. and diff.} show the
1065: difference spectra between the models with a tidally locked planet and the
1066: models with a non--rotating planet ($\mathcal{R}_{\rm rot} - \mathcal{R}_{\rm no~rot}$).
1067: \begin{figure}[h!]
1068: \plotone
1069: {fig2.eps}
1070: \caption{Upper panels show snap--shot spectra for one of the the Na D lines
1071: for two different model planets (tidally locked and non--rotating); lower
1072: panels show the difference between the two model spectra.The quantities
1073: plotted are $\mathcal{R}' = \mathcal{R}-1$ (upper panels) and
1074: $\Delta \mathcal{R}'$ (lower panels), where
1075: $\mathcal{R}[\lambda,t] = \dot{N}_{\rm in~transit}[\lambda,t]/\dot{N}_{\rm out~of~transit}[\lambda,t]$.
1076: In the upper panels, the blue curve is the tidally locked planet's transit
1077: spectrum, and the green curve is the non--rotating planet's transit spectrum.
1078: In the lower panels, the difference between the rotating and non--rotating
1079: planet's spectra.
1080: {\it Left}: Halfway through ingress (at $1.5^{\rm th}$ contact).
1081: {\it Right}: Halfway through the whole transit ($2.5^{\rm th}$ contact).}
1082: \label{fig:spec. and diff.}
1083: \end{figure}
1084:
1085: As described in \S~\ref{sec:overview} above, a planet's rotation causes
1086: the centroids of stellar absorption features to shift relative to a
1087: non--rotating planet. In Figure~\ref{fig:delta v}, centroid shifts
1088: (in velocity units) are plotted as a function of position in transit, for
1089: a planet transiting in front of a realistic star model with a Sun--like
1090: spectrum. The recessional velocity increases roughly sinusoidally during
1091: ingress, reaching a peak of about $60 {\rm ~cm~s^{-1}}$ at $1.5^{\rm th}$
1092: contact. During mid--transit, between $2^{\rm nd}$ and $3^{\rm rd}$
1093: contacts, the net velocity shift is much smaller. Egress is nearly
1094: perfectly symmetrical with ingress, though the velocity shifts have the
1095: opposite sign.
1096: \begin{figure}[h!]
1097: \plotone
1098: {fig3.eps}
1099: \caption{Centroid--shift of Na D lines from tidally locked rotation from the
1100: beginning to the end of a transit, relative to an identical but non--rotating
1101: planet; Sun--like stellar spectrum.
1102: The vertical lines denote $1.5^{\rm th}$ and $3.5^{\rm th}$ contact (cyan)
1103: and second and third contact (magenta).
1104: Between first and second contact, the
1105: spectrum with the rotating planet is redshifted relative to the non--rotating
1106: planet by up to about $60 {\rm ~cm~s^{-1}}$; between third and fourth contact,
1107: it is blueshifted by the same amount. This plot samples the transit at 60
1108: regularly--spaced points. Parameters were chosen to represent the HD209458
1109: system.}
1110: \label{fig:delta v}
1111: \end{figure}
1112:
1113: The cumulative and incremental signal--to--noise ratio across the transit
1114: are shown in the top and bottom panels, respectively, of
1115: Figure~\ref{fig:inc. and cum.}. As a planet proceeds in its transit from
1116: first contact to fourth contact, the S/N builds up steadily throughout
1117: the transit, as can be seen in the top panel of
1118: this figure. This cumulative increase reflects the steady
1119: incremental S/N per second of observation, which, for a planet
1120: crossing in front of a uniform star, is shown in the bottom panel.
1121: The three curves in the bottom panel represent, in decreasing
1122: order of S/N, the incremental S/N curves that are expected for a spectrograph
1123: with ultrahigh--resolution (the simulation's resolution $R_S = 700,000$),
1124: for a spectrograph with as high--resolution as bHROS on Gemini
1125: ($R^\prime_S = 150,000$), and for a spectrograph with more standard
1126: high--resolution
1127: ($R^\prime_S = 50,000$). In the top panel, we show the cumulative S/N (assuming
1128: the simulation's resolution) for a planet in front of a uniform star, and for
1129: a planet in front of a more realistic sun--like star that rotates once per
1130: month, has a limb--darkening profile, and has a solar spectrum. It
1131: is apparent in the top panel of Figure~\ref{fig:inc. and cum.} that a
1132: spectrograph with our simulation's resolution would see the effect of rotation
1133: at $S/N = 7.1 \eta^{1/2}$ in a single transit in the case of the simplified
1134: uniform star, and at $S/N = 3.2 \eta^{1/2}$ in one transit of the realistic star.
1135: The effect of including in the simulation the realistic features of stellar rotation,
1136: limb--darkening, and a solar spectrum is therefore to depress the S/N of the
1137: effect of tidally locked rotation by a factor of slightly more than 2. The bottom
1138: panel of the figure indicates that our predicted S/N for one transit of a realistic
1139: star ($3.2 \eta^{1/2}$) might be adjusted downward by $\sim 50\%$, depending on the
1140: spectral resolution, indicating a total one--transit S/N for the case of the
1141: realistic star of about $\sim 1.7 \eta^{1/2}$. Finally, we do not show the
1142: incremental S/N for the realistic star model, but, as the top panel of the figure
1143: indicates, a larger proportion of the total signal comes from ingress and egress
1144: in the realistic star model than in the uniform star model.
1145: \begin{figure}[h!]
1146: \plotone
1147: {fig4.eps}
1148: \caption{{\it Top}: S/N for distinguishing between a rotating and a
1149: non--rotating planet, accumulated during a single transit. Upper curve
1150: displays the cumulative S/N for the distinction of a tidally locked planet
1151: from a non-rotating model, assuming a completely uniform star: the model star
1152: has a flat spectrum, is not rotating, and there is no limb--darkening.
1153: Lower curve displays the same, but for a more realistic star: the model
1154: star rotates once per month, has the limb--darkening profile determined by
1155: \citet{knutson_et_al2007a}, and has a solar spectrum.
1156: These curves assume the spectral resolution used in the simulation
1157: ($R_S = 700,000$); for more easily achievable spectral resolution, these
1158: curves are adjusted downward by 30\% - 80\%, as displayed in the bottom panel.
1159: The vertical lines denote $1.5^{\rm th}$ and $3.5^{\rm th}$ contact (cyan)
1160: and second and third contact (magenta).
1161: Also, both top and bottom panels show model results assuming 100\% throughput
1162: efficiency ($\eta = 100\%$). In practice, these curves would
1163: be adjusted downward by a factor $\eta^{1/2}$, or $\sim 0.04^{1/2} = 0.2$ for
1164: an ultrahigh resolution spectrograph.
1165: {\it Bottom}: Incremental S/N for sixty 1--second observations across the
1166: transit, assuming a uniform, non--rotating star with a flat spectrum. Top
1167: curve shows the incremental S/N, assuming a spectrograph with the spectral
1168: resolution of the simulation ($R_S = 700,000$). Middle and bottom curves
1169: show the same, assuming spectral resolutions $R^\prime_S = 150,000$ and
1170: $R^\prime_S = 50,000$, respectively. Parameters of the system (except for
1171: the stellar spectrum in the bottom panel) were chosen to represent the
1172: HD209458 system.}
1173: \label{fig:inc. and cum.}
1174: \end{figure}
1175:
1176:
1177:
1178: \section{Discussion}
1179: \label{sec:disc}
1180: The $S/N$ scales with the square root of the number of absorption lines under
1181: consideration, with the square root of the number of transits, and with
1182: several other parameters, as follows:
1183: \begin{eqnarray}
1184: \nonumber S/N & \sim &
1185: \left( \sqrt{\frac{N_{\rm abs}}{2} } \right)
1186: \left( \sqrt{ \#~\rm transits} \right)
1187: \left( \frac{D \sqrt{\eta}}{\rm 6~m} \right) \times \\
1188: \nonumber & &
1189: 10^{-0.2(V_*-V_0)} \times
1190: \left(\frac{R_*}{R_0}\right)^{-2}
1191: \left(\frac{R_p H}{{R_p}_0 H_0}\right)
1192: \left(\frac{v}{2000 \rm~m~s^{-1}}\right) \times \\
1193: \label{eq:snscale} & &
1194: \left(\psi\left[R^\prime_S\right]\right)
1195: \left( S/N \right)_0
1196: \end{eqnarray}
1197: In this equation, $N_{\rm abs}$ is the number of absorption lines,
1198: $D$ is the telescope's diameter, $\eta$ is the efficiency of the instrumental
1199: setup, $V_*$ is the star's magnitude, $V_0$ is HD209458's magnitude
1200: (about 7.6), $R_*$ is the stellar radius, $R_0$ is HD209458's radius,
1201: $R_p$ and $H$ are the planet's radius and scale height, and ${R_p}_0$ and
1202: $H_0$ are HD209458b's radius and scale height.
1203: $\psi[R^\prime_S]$ is a function that characterizes the (nonlinear)
1204: response of S/N to the spectral resolution of the spectrograph; for instance,
1205: $\psi[700,000] \approx 1$, $\psi[150,000] \sim 0.6$ and
1206: $\psi[50,000] \sim 0.2$. Finally, $(S/N)_0 = 7.1$ for a uniform star and
1207: ${X_{\rm Na~}}_p/{X_{\rm Na~}}_\sun = 0.01$; and $(S/N)_0 = 3.2$ for a more
1208: realistic (rotating, limb--darkened) star with a solar spectrum.
1209:
1210: Culling the signal from mid--transit, however, is much more difficult
1211: than from ingress and egress, because the shape of the distortion depends
1212: more sensitively on the structure of the planet. For the realistic
1213: star, our model predicts $S/N \sim 1.3 \eta^{1/2}$ (with the 6~m
1214: telescope, with Poisson--dominated photon noise) for ingress alone, and
1215: therefore a factor $2^{1/2}$ greater, or $S/N \sim 1.8 \eta^{1/2}$ for
1216: ingress and egress.
1217:
1218: In Table~\ref{ta:norbits}, we present the number of orbits that must be
1219: observed in order to make a $5\sigma$ detection of rotation, for various
1220: combinations of parameters. In all cases, we assume a realistically achievable
1221: optical setup, with spectral resolution $R^\prime_S = 150,000$ and
1222: throughput $\eta = 4\%$. Improvement in $R^\prime_S$ will yield a modest
1223: improvement in S/N, and improvement in $\eta$ could be quite significant.
1224: In the third column of Table~\ref{ta:norbits} (A), we present the total
1225: available S/N for the whole transit, according to our model, while in the
1226: fourth column (B), we present the available S/N from just ingress and egress.
1227: The fifth and sixth columns are based on the S/N in the fourth column.
1228: \begin{deluxetable}{crcccr}
1229: %\tablewidth{240pt}
1230: \tablecolumns{6}
1231: \small
1232: \tablecaption{Required Number of Transits For $5\sigma$ Detection}
1233: \tablehead{\colhead{{\tt Spec\_Shape}} & \colhead{$D$} & \colhead{(S/N)/Tr. (A)} & \colhead{(S/N)/Tr. (B)} & \colhead{Req. \# Transits (A)} & \colhead{Duration (A)}}
1234: \startdata
1235: Flat (uniform star) & 6~m & 1.1 & 0.52 & $\sim 25$ & $\sim 3$ months \\
1236: Solar & 6~m & 0.48 & 0.27 & $\sim 110$ & $\sim 1.25$ years \\
1237: Solar & 10~m & 0.79 & 0.45 & $\sim 40$ & $\sim 5$ months \\
1238: Solar & 30~m & 2.4 & 1.3 & $\sim 5$ & $\sim 3$ weeks \\
1239: \enddata
1240: \label{ta:norbits}
1241: \vspace{-0.4cm}
1242: \tablenotetext{a}{A table showing the number of orbits required to achieve a
1243: $S/N =5$ detection of tidally locked rotation of HD209458b, assuming a
1244: planetary Na--content that is 1\% solar. Assumptions: we are monitoring two
1245: spectral lines (the Na doublet) with a spectrograph with resolving power
1246: $R^\prime_S = 150,000$, and an efficiency in the optical setup of $0.04$.
1247: The third column ``(S/N)/Tr. (A)'' is the S/N per transit, as shown in
1248: Figure~\ref{fig:inc. and cum.}, and the fourth column is the S/N per
1249: transit, considering only ingress and egress. The ``(A)'' in the fifth and
1250: sixth columns indicates that the numbers of transits and the required
1251: duration are tabulated for S/N values from column three.
1252: }
1253: \end{deluxetable}
1254:
1255: Finally, as a sanity check, our model can be tested by comparing
1256: it to the analysis of
1257: \citet{charbonneau_et_al2002}. That analysis suggests that
1258: the sodium content of a planet's atmosphere can be determined
1259: by comparing the flux in a narrow band centered on the sodium
1260: resonance lines with the flux in a wider band surrounding but
1261: excluding the lines. In that paper, the decrement in the
1262: narrow band containing the sodium features is named $n_{\rm Na}$,
1263: and was measured to be $-2.32\times 10^{-4}$ for HD209458 during
1264: the middle of transit. They presented several models, all of
1265: which over--predicted the sodium decrement. The model that
1266: predicted the smallest magnitude of the decrement had 1\% solar
1267: metalicity and cloud deck at $0.0368 {\rm ~bar}$; this model
1268: predicts $n_{\rm Na} \sim -3.4 \times 10^{-4}$ in mid--transit.
1269: Our model (1\% solar metalicity, cloud deck at $0.01 {\rm ~bar}$)
1270: predicts $\sim -4.1 \times 10^{-4}$ in mid--transit, as shown
1271: in Figure~\ref{fig:charbonneau}. We conclude that our model is in
1272: reasonable agreement with both Charbonneau et al.'s model ($\sim 20\%$)
1273: and with the actual data ($\sim 40\%$).
1274: \begin{figure}[h!]
1275: \plotone
1276: {fig5.eps}
1277: \caption{The flux decrement due to sodium in the planet's atmosphere,
1278: as shown by the quantity $n_{\rm Na}$ defined by \citet{charbonneau_et_al2002}.
1279: This decrement reaches a maximum in magnitude of $-4.1 \times 10^{-4}$,
1280: halfway through transit, in our model.}
1281: \label{fig:charbonneau}
1282: \end{figure}
1283:
1284:
1285:
1286: \section{Conclusion}
1287: \label{sec:conc}
1288: Our investigation indicates that, with currently available instruments,
1289: it will be difficult to obtain the sensitivity needed to achieve a
1290: minimal $S/N \gsim 5$ detection of tidally locked rotation of the
1291: planet HD209458b. Nevertheless, it appears that the effect of rotation
1292: will have significant (though small) influence on transit--spectra taken with
1293: current or near--future instruments.
1294: Because this influence is so small, it is worth considering ways that the
1295: $S/N$ could be improved, according to the scalings summarized in
1296: equation~(\ref{eq:snscale}).
1297:
1298: \begin{itemize}
1299: \item The most obvious way to boost S/N is to increase the collecting area of
1300: the telescope, the effect of which is shown in Table~\ref{ta:norbits}. With an
1301: extremely large (30~m--class) telescope, the required discrimination power
1302: could be achieved in $\sim 5-15$ transits, if our optimistic assumptions about
1303: the noise are not too far off.
1304: \item As the abundances of other elements or molecules in the planet's
1305: atmosphere are identified, the number of absorption lines $N_{\rm abs}$ that
1306: can be utilized can increase. Observing
1307: four absorption features instead of two will boost the $S/N$ per orbit by
1308: $2^{1/2}$ and will therefore cut in half the required observing time to
1309: achieve a fixed target $S/N$.
1310: \item The HD189733 system is only 19.3~pc away \citep{bouchy_et_al2005},
1311: but the star is smaller and dimmer than HD209458, and has the same apparent
1312: magnitude. We find that, overall, it has no relative advantage in terms of
1313: $S/N$. It may be unlikely that a star significantly brighter than HD209458
1314: will be found to have a transiting giant planet companion, but if this should
1315: happen the signature of rotation will be more readily apparent in the system
1316: with the brighter star.
1317: \item Finally, since S/N is proportional to the size of the atmosphere,
1318: represented by $\delta_{\rm atm}$ in equation~(\ref{eq:SN estimate}), a planet
1319: with a more extended atmosphere would show the distortion due to rotation more
1320: readily. The size of the atmosphere depends on the product of $R_p$ and the
1321: height of the atmosphere $\Pi_{\rm atm} \sim 15 H$, so the S/N could be
1322: increased in the case of a larger planet, a hotter one, or one with lower
1323: surface gravity. Unfortunately, although some other planets are somewhat
1324: larger, none of the puffier planets that have been discovered so far happen to
1325: transit a star as bright as HD209458. If a more bloated planet is found to
1326: transit a star as bright as HD209458, such a planet would probably be a more
1327: promising target for the type of study described in this paper.
1328: \end{itemize}
1329:
1330: Although the optical transit spectrum of an extrasolar giant planet contains
1331: many precious clues regarding the nature of rotation and climate or weather, we
1332: emphasize that we have so far addressed only the forward problem: deep,
1333: high spectral resolution transit spectra {\it will} be distorted by
1334: HD209458b's tidally locked rotation, if indeed it is tidally locked.
1335: But seeing the effect of rotation in a spectrum and {\it knowing} that we are
1336: seeing the effect of rotation are two very different things.
1337: The inverse problem -- identifying the rotation rate -- will be much more
1338: challenging, and is beyond the scope of this paper's investigation.
1339: In brief, it is impossible to decode the clues that the spectrum holds
1340: without a Rosetta stone -- namely, without already possessing accurate
1341: knowledge of abundances in the planet's atmosphere.
1342: We note that when it comes to solving the inverse problem, it may be possible
1343: to make do without a perfect Rosetta stone by comparing the star's spectrum
1344: at two different times of transit instead of comparing it to an accurate
1345: input model. Specifically, comparing the ingress spectrum to the egress
1346: spectrum should eliminate various systematic uncertainties that may effect both
1347: ingress and egress spectra. This strategy involves foregoing spectra between
1348: second and third contacts, at a cost of a factor of $\sim 2$ in $S/N$, but has
1349: the advantage of being more robust with respect to errors in modeling the
1350: planetary atmosphere. Furthermore, if detailed analysis along the lines of
1351: \citet{charbonneau_et_al2002} should reveal many more planetary absorption
1352: features, a large current--generation telescope could detect the motion of an
1353: extrasolar planet's atmosphere in of order a month.
1354:
1355: A final point to keep in mind is that the technique presented in this paper
1356: is not limited to discerning the difference between a tidally locked planet
1357: and a non--rotating planet; this technique can be used to investigate the
1358: applicability of any model of the motion of a transiting planet's atmosphere.
1359: Some atmospheric circulation models of Hot Jupiters predict extremely fast
1360: winds, up to $\sim 9 {\rm ~km~s^{-1}}$, or more then four times the
1361: predicted equatorial rotation rate. It might therefore turn out to be easier
1362: to detect winds than to detect rotation. Moreover, if an analysis similar to
1363: the one described here were to find a deviation from the hypothesis of
1364: tidally locked rotation, this would mean that {\it something} interesting
1365: is going on -- the planet could be rotating roughly as a solid body but at
1366: a different rate; there could be significant non--rotational motion in
1367: the atmosphere; or there could be some combination of these phenomena.
1368:
1369:
1370:
1371:
1372: % Acknowledgments
1373: \vspace{\baselineskip}
1374: \vspace{\baselineskip}
1375: We thank Caleb Scharf, Frits Paerels, and Greg Bryan for extremely helpful
1376: discussions. We gratefully acknowledge George Rybicki for use of a Voigt
1377: profile routine. We acknowledge many helpful comments by our anonymous
1378: referee.
1379:
1380:
1381:
1382:
1383: \bibliography{biblio.bib}
1384:
1385:
1386: \end{document}
1387: