1: \documentclass[12pt,preprint]{aastex}
2: \usepackage{epsfig,wrapfig,graphicx}
3:
4: \def\ms{m~s$^{-1}$}
5: \def\ks{km~s$^{-1}$}
6: \def\msini{$M_P\sin{i}$}
7: \def\asini{$a\sin{i}$}
8: \def\vsini{$V_{\rm rot}\sin{i}$}
9: \def\msun{$M_{\odot}$}
10: \def\mjup{$M_{\rm Jup}$}
11: \def\rsun{$R_{\odot}$}
12: \def\lsun{$L_{\odot}$}
13: \def\chisq{$\sqrt{\chi^2_\nu}$}
14: \def\chis{$\chi^2_\nu$}
15: \def\plmn{~$\pm$~}
16: \def\feh{[Fe/H]}
17: \def\rphk{$R^\prime_{HK}$}
18: \def\shk{$S$}
19: \def\trendA{-11}
20: \def\npl{3}
21: \def\npllet{three}
22: \def\nplcap{Three}
23:
24: \include{defs}
25:
26: \def\starA{HD\,210702}
27: \def\starB{HD\,175541}
28: \def\starC{HD\,192699}
29: \def\hippA{HIP\,109577}
30: \def\hippB{HIP\,92895}
31: \def\hippC{HIP\,99894}
32: \def\mstarA{1.85}
33: \def\mstarB{1.65}
34: \def\mstarC{1.68}
35: \def\rstarA{4.72}
36: \def\rstarB{3.85}
37: \def\rstarC{4.25}
38: \def\lstarA{13.1}
39: \def\lstarB{9.56}
40: \def\lstarC{11.5}
41: \def\bvA{0.951}
42: \def\bvB{0.869}
43: \def\bvC{0.867}
44: \def\mvA{2.19}
45: \def\mvB{2.49}
46: \def\mvC{2.30}
47: \def\vmagA{5.93}
48: \def\vmagB{8.03}
49: \def\vmagC{6.44}
50: \def\ageA{1.4}
51: \def\ageB{1.9}
52: \def\ageC{1.8}
53: \def\dA{56}
54: \def\dB{128}
55: \def\dC{67}
56: \def\shkA{0.11}
57: \def\shkB{0.11}
58: \def\shkC{0.12}
59: \def\rhkA{-5.35}
60: \def\rhkB{-5.28}
61: \def\rhkC{-5.29}
62: \def\prA{70}
63: \def\prB{60}
64: \def\prC{60}
65: % SME
66: \def\feA{$+0.12$}
67: \def\feB{$-0.07$}
68: \def\feC{$-0.15$}
69: \def\loggA{3.29}
70: \def\loggB{3.52}
71: \def\loggC{3.44}
72: \def\teffA{5010}
73: \def\teffB{5060}
74: \def\teffC{5220}
75: \def\vsiniA{1.7}
76: \def\vsiniB{2.9}
77: \def\vsiniC{1.9}
78:
79: \begin{document}
80: \title{Retired A Stars and Their Companions: Exoplanets
81: Orbiting \nplcap\ Intermediate--Mass Subgiants$^1$}
82:
83: \author{ John Asher Johnson\altaffilmark{2},
84: Debra A. Fischer\altaffilmark{3},
85: Geoffrey W. Marcy\altaffilmark{2},
86: Jason T. Wright\altaffilmark{2},
87: Peter Driscoll\altaffilmark{4},
88: R. Paul Butler\altaffilmark{5},
89: Saskia Hekker\altaffilmark{6},
90: Sabine Reffert\altaffilmark{7},
91: Steven S. Vogt\altaffilmark{8}
92: }
93:
94: \email{johnjohn@astron.berkeley.edu}
95:
96: \altaffiltext{1}{Based on observations obtained at the Lick
97: Observatory, which is operated by the University of California, and
98: W. M. Keck Observatory, which is operated jointly by the University
99: of California and the California Institute of Technology}
100: \altaffiltext{2}{Department of Astronomy, University of California,
101: Mail Code 3411, Berkeley, CA 94720}
102: \altaffiltext{3}{Department of Physics \& Astronomy, San Francisco
103: State University, San Francisco, CA 94132}
104: \altaffiltext{4}{Department of Earth \& Planetary Sciences, Johns
105: Hopkins University, Baltimore, MD, 21218}
106: \altaffiltext{5}{Department of Terrestrial Magnetism, Carnegie
107: Institution of Washington DC, 5241 Broad Branch Rd. NW, Washington DC,
108: 20015-1305}
109: \altaffiltext{6}{Leiden Observatory, Leiden University, PO Box 9513,
110: 2300 RA Leiden, The Netherlands}
111: \altaffiltext{7}{ZAH-Landessternwarte, K\"onigstuhl 12, 69117
112: Heidelberg, Germany}
113: \altaffiltext{8}{UCO/Lick Observatory, University of California at
114: Santa Cruz, Santa Cruz, CA 95064}
115:
116: \begin{abstract}
117: We report precision Doppler measurements of \npllet\
118: intermediate--mass subgiants obtained at Lick and Keck
119: Observatories. All \npllet\ stars show variability in
120: their radial velocities consistent with planet--mass companions in
121: Keplerian orbits. We find a planet with a minimum mass
122: $M_P\sin{i} = \msiniC$~\mjup\ in a \pC~day orbit around \starC, a planet
123: with a minimum mass of \msiniA~\mjup\ in a \pA~day orbit around
124: \starA, and a planet with a minimum mass of \msiniB~\mjup\ in a
125: \pB~day orbit around \starB. Mass estimates from stellar interior
126: models indicate that all three stars were formerly A--type,
127: main--sequence dwarfs with masses ranging from \mstarB~\msun\ to
128: \mstarA~\msun. These \npllet\ long--period planets would not have been
129: detectable during their stars' main--sequence phases due to the large
130: rotational velocities and stellar jitter exhibited by early--type
131: dwarfs. There are now 9 ``retired'' (evolved) A--type
132: stars ($M_* > 1.6$~\msun) with known planets. All 9 planets orbit at
133: distances $a \geq 0.78$ ~AU, which is significantly different than the
134: semimajor axis
135: distribution of planets around lower--mass stars. We examine the
136: possibility that the observed lack of short--period planets is due to
137: engulfment by their expanding host stars, but we find that this
138: explanation is inadequate given the
139: relatively small stellar radii of K giants ($R_* < 32$~\rsun~$=0.15$~AU) and
140: subgiants ($R_* < 7$~\rsun~$ = 0.03$~AU). Instead, we conclude that planets
141: around intermediate--mass stars reside preferentially beyond
142: $\sim$0.8~AU, which may be a reflection of different formation and
143: migration histories of planets around A--type stars.
144: \end{abstract}
145:
146: \keywords{techniques: radial velocities---planetary systems:
147: formation---stars: individual (HD\,192699, HD\,210702, HD\,175541)}
148:
149: \section{Introduction}
150:
151: Very little is known about the occurrence rate and orbital properties
152: of planets around A--type stars, corresponding to stellar masses
153: ranging from 1.6~\msun\ to 3.0~\msun.
154: Inspection of the Catalog of Nearby Exoplanets (CNE)\footnote{For the
155: updated catalog of extrasolar planet and their parameters see
156: http://exoplanets.org.} reveals that only 6 of the 173 stars with
157: securely detected planetary companions have masses in excess of
158: 1.6~\msun \citep{butler06}. This small number of detections
159: is not a true reflection of the occurrence of planets
160: around A--type stars, but rather the result of a strong selection bias
161: against early--type, main--sequence stars in precision Doppler surveys.
162:
163: Measuring precise Doppler shifts of early--type dwarfs is complicated
164: by their rotationally broadened spectral features, high surface
165: temperatures, and high levels of excess radial velocity noise, or
166: ``jitter'' \citep{saar98, wright05}. \citet{galland05} find that Doppler
167: precision for early--type dwarfs is limited to $\sim 40$~\ms\ at
168: spectral type F5V, and 90--200~\ms\ for A stars, rendering Doppler
169: measurements of these stars sensitive only to planets with large
170: masses and short orbital periods. The lowest mass companion
171: so far detected around an A star is the brown
172: dwarf orbiting HD\,180777 \citep{galland06}. Even though the
173: 28~day orbital solution has a large velocity semiamplitude, $K
174: =1200$~\ms, the signal is only a 3$\sigma$ detection above the stellar
175: jitter and measurement uncertainties.
176:
177: Most of what is known about planet formation around intermediate--mass
178: stars comes from two primary sources: direct imaging of disks around
179: young stars and Doppler detections of planets around evolved stars.
180: While A--type dwarfs are poor Doppler targets, their
181: high intrinsic luminosities facilitate the detection and direct
182: imaging of material in their circumstellar environments.
183: More than a decade before the discoveries of the first extrasolar
184: planets, evidence of planet formation outside of our Solar
185: System came from the infrared detection of collision--generated
186: dust around the A--type, main sequence stars Vega \citep{aumann84} and
187: $\beta$~Pic \citep{smith84}. Since then, advances in high--contrast
188: imaging have resulted in the detection of an optically thick disk around a
189: pre--main--sequence Herbig Ae star \citep{perrin06}, as
190: well as scattered light images of optically thin "debris disks" around
191: 11 main--sequence stars---the majority of which have spectral types F5V
192: or earlier \citep[Table 2 of][and references therein]{kalas06,
193: schneider06, wahhaj07}. Recent observations of the
194: debris disk around the young A star Fomalhaut have revealed a
195: perturbation in the disk structure that may be due to the influence of
196: an orbiting Jovian planet \citep{kalas05}. Studying the relationships
197: between the architectures of disks around young A stars and the
198: distribution of planet properties around their older counterparts will
199: provide key tests of planet formation models.
200:
201: A key to finding planets around A stars using Doppler methods is
202: provided by the effects of
203: stellar evolution. As stars evolve away from the main sequence, they
204: become cooler and rotate slower, which increases the number of narrow
205: absorption lines in their spectra \citep{gray85, schrijver93,
206: donascimento00}.
207: Several Doppler surveys have focused on evolved, intermediate--mass
208: stars on the red giant branch \citep{frink02, mitchell04, hatzes05,
209: lovis05} and
210: clump giant branch \citep{sato03, setiawan03}. These surveys have resulted in
211: the discovery of 6 substellar companions orbiting
212: former A--type stars (Table \ref{massive_table}). That none of these
213: planets would have been detectable during their host stars'
214: main--sequence phases highlights the important role evolved stars play
215: in the study of planets around intermediate--mass stars.
216:
217: Here we present \npllet\ new planet candidates around stars with $M_*
218: > 1.6$~\msun. These detections come from our precision Doppler survey
219: of evolved stars on the subgiant branch of the H--R diagram. We
220: discussed the selection criteria of our target stars in
221: \citet{johnson06b}, along with the discovery of an eccentric hot Jupiter
222: orbiting the 1.28~\msun\ subgiant HD\,185269. We discuss
223: our spectroscopic observations and Doppler measurement technique in
224: \S~\ref{observations}. In \S~\ref{stars}, we present the
225: characteristics of the host stars along with the orbital solutions
226: for their planet candidates. We conclude with a comparison of the
227: semimajor axis distributions of planets around A--type stars
228: and lower--mass stars in \S~\ref{summary}.
229:
230: \section{Observations}
231: \label{observations}
232:
233: We are monitoring a sample of 159 evolved stars at Lick and Keck
234: Observatories \citep{johnson06b}. At Lick Observatory, the Shane 3\,m
235: and 0.6\,m
236: Coude Auxiliary Telescopes (CAT) feed the Hamilton spectrometer
237: \citep{vogt87}, which has a resolution of $R \approx 50,000$ at
238: $\lambda = 5500$~\AA. Spectroscopic observations at Keck Observatory were
239: obtained using the HIRES spectrometer with a resolution of $R \approx
240: 80,000$ at $\lambda = 5500$~\AA\ \citep{vogt94}.
241: Doppler shifts are measured from each spectrum
242: using the iodine cell method described by \citet{butler96} \citep[see
243: also][]{marcy92b}. A temperature--controlled Pyrex cell containing
244: gaseous iodine is placed at the entrance slit of the
245: spectrometer. The dense set of narrow molecular lines imprinted on
246: each stellar spectrum from 5000 to 6000~\AA\ provides a
247: robust wavelength scale for each observation, as well as information
248: about the shape of the spectrometer's instrumental response.
249:
250: \begin{figure}[t!]
251: \epsscale{0.8}
252: \plotone{fig1.eps}
253: \caption {\footnotesize{Radial velocity time series for four stable subgiants with
254: $B-V$ colors and absolute visual magnitudes similar to those of the
255: \npllet\ subgiant planet host stars. $\sigma$ represents the
256: standard deviation of the velocities about the mean (dashed
257: lines). Observations
258: from Keck (left panels) and Lick (right panels) show that
259: subgiants in this region of the H--R diagram are typically stable
260: to well within 10~\ms\ over time scales of many years. \label{std_stars}}}
261: \end{figure}
262:
263: Traditionally, the Doppler shift of each stellar observation is made
264: with respect to an observed, iodine--free stellar template spectrum.
265: These template observations require higher signal and resolution than
266: normal radial velocity observations, which leads to increased exposure
267: times. Given our large target list and the small aperture of the CAT,
268: obtaining an observed template for each star would represent a
269: prohibitive cost in observing time. We therefore perform
270: a preliminary analysis of each star's observations using a
271: synthetic, ``morphed'' template spectrum following the method
272: described by \citet{johnson06}. Stars showing conspicuous Doppler
273: variations are reanalyzed using a traditional, observed template to
274: verify the signal and search for a full orbital solution.
275:
276: Doppler measurements from Keck and Lick Observatories for four stable
277: subgiants are shown in Figure~\ref{std_stars}. The error bars
278: represent the internal uncertainties
279: of each measurement, which are approximated by the weighted standard
280: deviation of the mean velocity measured from each of the 700 individual
281: 2~\AA\ wide chunks in each spectrum \citep{butler96}. We
282: typically achieve internal
283: measurement uncertainties of 1-2~\ms\ for Keck observations and
284: 3-5~\ms\ at Lick. Subgiants have an additional 4-6~\ms\ of
285: ``jitter''---velocity scatter in excess of internal errors due to
286: astrophysical sources such as pulsation and rotational modulation of
287: surface features \citep[][]{saar98, wright05}. We therefore adopt a
288: jitter value of 5~\ms\ for our subgiants, which is added in quadrature
289: to the internal uncertainties of the measurements before
290: searching for a best--fit orbital solution.
291:
292: After determining the best--fit Keplerian solution using a
293: Levenberg--Marquardt, least--squares minimization, we estimate the
294: orbital parameter uncertainties using a bootstrap
295: Monte Carlo method. We first subtract the best--fit Keplerian from the
296: measured velocities. The residuals are then scrambled and added back
297: to the original measurements, and a new set of orbital parameters is
298: obtained. This process is repeated for 1000 trials, and the standard
299: deviations of the parameters from all trials are adopted as
300: the formal, 1$\sigma$ uncertainties.
301:
302: \section{Stellar Properties and Orbit Solutions}
303: \label{stars}
304:
305: \subsection{Estimates of Stellar Properties}
306: \label{stellar}
307:
308: We estimated the stellar properties of our target stars using two
309: primary methods: the LTE
310: spectral synthesis method (SME) described by \citet{valenti05}, and
311: the Padova\footnote{See also http://pleiadi.pd.astro.it/} stellar
312: interior models. The spectral synthesis method uses a
313: non--linear least--squares algorithm to vary the parameters of a
314: synthetic spectrum to search for a fit to an iodine--free stellar
315: template spectrum. The free parameters in the fit are the
316: abundances of heavy elements; effective surface temperature, $T_{\rm eff}$;
317: surface gravity, $\log{g}$; and broadening effects due to
318: the star's projected rotation velocity, \vsini. \citet{valenti05}
319: estimate a precision of 0.04 dex in
320: metallicity, 44~K in effective temperature, 0.3 dex in $\log{g}$, and
321: 0.5~\ks\ in rotational velocity.
322:
323: \begin{figure}[t!]
324: \epsscale{0.8}
325: \plotone{fig2.eps}
326: \caption{\footnotesize{This figure illustrates the interpolation method employed to
327: determine accurate stellar masses for the planet host stars. We
328: estimated each star's mass, radius and age by interpolating its
329: Hipparcos $B-V$
330: color and absolute visual magnitude $M_V$ onto grids of four
331: different metallicities: [Fe/H] $= -0.4, 0.0, +0.18, +0.3$ (open
332: diamonds). For each star's measured value of $B-V$ and $M_V$, the
333: three lines show the dependence of stellar mass on the measured
334: [Fe/H], estimated using a cubic spline interpolation between the
335: diamonds. Similar dependencies were determined for stellar radii,
336: luminosities and ages. \label{massint}} }
337: \end{figure}
338:
339: To estimate stellar masses, radii, luminosities and ages, we used the Padova
340: theoretical stellar models, which have been
341: transformed into several photometric systems by
342: \citet{girardi02}. Stellar properties can be inferred by interpolating
343: a star's color, absolute magnitude and metallicity onto these model
344: grids. However, the \citet{girardi02} model grids are defined at
345: widely--spaced metallicity intervals, with [Fe/H]~$ = $~-0.4, 0.0,
346: +0.18 and +0.30. Since the uncertainties in our spectroscopically
347: derived metallicity estimates are much less than the model grid
348: intervals, and because the derived stellar properties do not vary
349: linearly with [Fe/H], we could not simply perform a linear, 3-dimensional
350: interpolation of $M_V$, $B-V$ and [Fe/H]. Instead, we first linearly
351: interpolate the stars' colors and absolute magnitudes onto each of the
352: four metallicity grids. We then use a cubic spline interpolation
353: between the grid points to measure the desired stellar property
354: (e.g. mass) at the star's measured [Fe/H].
355: Our procedure is illustrated in Figure~\ref{massint}, which
356: shows stellar mass as a function of [Fe/H] for each star's
357: absolute magnitude and color. The same procedure was used for stellar
358: radii, luminosities and ages.
359:
360: We compared our interpolated stellar properties to the \citet{takeda06}
361: theoretical interior models of the stars in the Spectroscopic
362: Properties of Cool Stars catalog
363: \citep[SPOCS][]{valenti05}. We found a subset of 11 evolved stars
364: in the catalog with $2.0 < M_V < 3.0$ and $0.7 < B-V < 1.1$.
365: Differences between our inferred values and those from \citet{takeda06}
366: had an rms scatter of 7\% in mass, 12\%
367: in radius, with a median offset of -2\% and -4\% in each parameter,
368: respectively. Ages of this subset of evolved stars estimated by
369: the two methods have a difference
370: of -0.4~Gyr with and rms scatter of 1.1~Gyr. We therefore adopt
371: fractional uncertainties of 7\% for our derived masses, 12\% for
372: radii and 1~Gyr for ages. We list the full set of derived stellar
373: properties of the three candidate planet host stars in
374: Table~\ref{stellartable}. We summarize each star's properties and
375: orbital solution in the following subsections.
376:
377: \subsection{\starC}
378:
379: \starC\ (\hippC) is listed with a G5 spectral type in the \emph{Hipparcos}
380: Catalog, with $V = \vmagC$, $B - V = \bvC$ and a parallax--based
381: distance of \dC~pc \citep{hipp}. However, no luminosity class is
382: given. Based on its distance, we calculate $M_V = \mvC$, which at its
383: $B-V$ color places the star 3.7 mag above the mean main--sequence of stars
384: in the Solar neighborhood, as defined by \citet{wright04}. Based on
385: its color and absolute magnitude, we find that \starC\ is
386: likely a G8\,IV subgiant near the base of the red giant
387: branch. Commensurate with its evolved status,
388: \starC\ is chromospherically inactive, with \shk~$=$~\shkC\ and
389: \rphk~$=$~\rhkC\ as measured from the CaII H\&K line core and averaged
390: over all observations \citep{wright04b}.
391:
392: Based on our LTE spectral analysis, we find that \starC\ is
393: metal--poor, with $\rm [Fe/H] = -0.15$, and slowly rotating, with
394: \vsini~$ = \vsiniC$~\ks. The other stellar parameters derived
395: from our spectral analysis are listed in Table~\ref{stellartable}. We
396: interpolated the star's color, absolute magnitude and metallicity
397: onto the \citet{girardi02} theoretical stellar model grids using the
398: method described
399: in \S~\ref{stellar}. Our interpolation yields a stellar mass $M_* =
400: \mstarC$~\msun, radius $R_* = \rstarC$~\rsun, and an age of \ageC~Gyr.
401:
402: We began observing \starC\ in 2004 May at Lick Observatory using the
403: 3~m Shane Telescope and 0.6~m CAT.
404: Table~\ref{vel192699} lists our \nobsC\ velocity measurements, along with
405: their times of observation and internal measurement uncertainties
406: (without jitter). Our
407: first 7 observations, initially analyzed using a synthetic stellar template
408: spectrum \citep{johnson06}, showed correlated variations spanning
409: two observing seasons. We obtained a high--quality observed template
410: using the Shane~3m telescope and initiated intensive follow--up
411: observations during the Fall 2006 observing season. The Keplerian signal
412: is visible to the eye (Figure \ref{orbitC}), obviating a periodogram
413: analysis.
414:
415: \begin{figure}[t!]
416: \epsscale{1}
417: \plotone{fig3.eps}
418: \caption {\footnotesize{Radial velocity time series for \starC\
419: measured at Lick
420: Observatory. The error bars reflect the quadrature sum of the
421: internal measurement uncertainties and 5~\ms\ of jitter. The dashed
422: line shows the best--fit orbital solution, which has \chisq$ =
423: \chiC$. \label{orbitC}} }
424: \end{figure}
425:
426: The best--fit
427: Keplerian orbit has a period of $P = \pC$~d, velocity amplitude $K =
428: \kC$~\ms, and eccentricity $e = \eC \pm \eeC$. With an assumed stellar
429: mass of \mstarC~\msun, we estimate a minimum planet mass \msini~$ =
430: \msiniC$~\mjup\ and orbital separation $a = \arelC$~AU. The fit has
431: $\rm rms = \rmsC$~\ms\ and a reduced \chisq~$= \chiC$, consistent with
432: the measurement errors and jitter. The full set of orbital parameters
433: and uncertainties is listed in Table~\ref{orbittable}.
434:
435: \subsection{\starA}
436: \label{starA}
437:
438: \starA\ (\hippA, HR\,8461) is listed in the Hipparcos catalog as a K1
439: star (no luminosity class given) with $V = \vmagA$, $B-V = \bvA$, and a
440: parallax--based distance of \dA~pc. Given its distance and apparent
441: magnitude, we calculate an absolute magnitude $M_V = \mvA$, which
442: places it 4.2~mag above the average main sequence of stars in the
443: Solar neighborhood \citep{wright04}. We therefore estimate that
444: \starA\ is a class K1\,IV subgiant near the base of the red giant branch.
445:
446: Based on our LTE spectral analysis, we find that \starA\ is somewhat
447: metal--rich, with \feh~$=\feA \pm 0.04$, and slowly rotating, with \vsini~$ = \vsiniC$~\ks. Our
448: interpolation of the star's color, absolute magnitude and metallicity
449: onto the \citet{girardi02} stellar model grids yields a stellar mass $M_* =
450: \mstarA$~\msun, stellar radius $R_* = \rstarA$~\rsun, and an age of
451: \ageA~Gyr. Consistent with its post--main--sequence evolutionary
452: status, \starA\ is chromospherically inactive with \shk~$=$~\shkA\ and
453: \rphk~$=$~\rhkA, as measured from its CaII~H\&K emission
454: \citep{wright04b}. The other stellar parameters derived
455: from our spectral analysis and stellar model interpolation are listed
456: in Table~\ref{stellartable}.
457:
458: We began monitoring \starA\ in 2004 August at Lick Observatory. The
459: first 9 observations were Doppler--analyzed using a synthetic
460: template, and showed excessive variability with $\rm
461: rms=19$~\ms. We then obtained a traditional, observed template
462: to confirm the variations with higher Doppler precision. The full set
463: of velocities is listed in Table~\ref{vel210702} (without jitter) and
464: plotted in Figure \ref{orbitA}. The error bars in Figure~\ref{orbitA}
465: have been augmented by adding 5~\ms\ of
466: jitter in quadrature to the internal measurement uncertainties.
467:
468: \begin{figure}[h!]
469: \epsscale{1}
470: \plotone{fig4.eps}
471: \caption {\footnotesize{Radial velocity time series for \starA\
472: measured at Lick Observatory. The dashed line shows the
473: best--fit orbital solution, which has \chisq$=\chiA$.\label{orbitA}} }
474: \end{figure}
475:
476: The best--fit Keplerian orbital solution is shown in
477: Figure~\ref{orbitA} overplotted on the velocities. The solution has a
478: \pA~day period, an eccentricity $e = \eA \pm \eeA$, and a
479: semiamplitude $K = \kA$~\ms. The fit residuals have $\rm rms =
480: \rmsA$~\ms\ and reduced \chisq~$=$~\chiA, consistent with the
481: internal measurement uncertainties and jitter. Assuming a stellar mass
482: $M_* = \mstarA$~\msun, the best--fit solution yields a relative separation $a
483: = \arelA$~AU.
484:
485: We find that the inclusion of a linear trend in the orbital solution
486: yields a slight improvement in the quality of fit, decreasing the rms
487: scatter of the residuals from \rmsA~\ms\ to 6.7~\ms, and the reduced
488: \chisq\ from \chiA\ to 1.00 after accounting for the
489: extra free parameter in the
490: Keplerian--plus--trend model. We tested the validity of the trend
491: using the prescription of \citet{wright07}, and found a false--alarm
492: probability of 49\%. The large FAP indicates that the apparent
493: linear trend is likely due to noise rather than and additional orbital
494: companion. Indeed, the trend appears to be driven primarily by the
495: three outliers near JD~$=$ 100, 400 and 800 (Figure~\ref{orbitA}). We
496: therefore favor the single--planet Keplerian model summarized in
497: Table~\ref{orbittable}.
498:
499: \subsection{\starB}
500: \label{sectionB}
501:
502: \starB\ (\hippB) is listed in the
503: \emph{Hipparcos} Catalog as a G8V star with $V = \vmagB$, $B-V = \bvB$
504: and a parallax--based distance of \dB~pc \citep{hipp}. Given its
505: distance, the star
506: has $M_V = \mvB$, placing it 3.5 mag above the mean main--sequence of
507: stars in the Solar neighborhood \citep{wright04}. Like most evolved stars,
508: \starB\ is chromospherically quiet with \shk~$=$~\shkB\ and
509: \rphk~$=$~\rhkB\ \citep{wright04b}.
510: Its low chromospheric activity and location in the H--R
511: diagram indicate that \starB\ is most likely a luminosity
512: class IV star on the subgiant branch, rather than a class V dwarf.
513:
514: \starB\ is listed in the SPOCS Catalog
515: \citep{valenti05} with a metal abundance slightly below Solar
516: (\feh~$ = \feB \pm 0.04$) and projected rotational velocity
517: \vsini~$= \vsiniB$~\ks. Interpolation of the
518: star's $B-V$ color, absolute magnitude and metallicity onto the
519: \citet{girardi02} stellar model grids yields a stellar mass $M_* =
520: \mstarB$~\msun, radius $R_* = \rstarB$~\rsun, and an age of
521: \ageB~Gyr. The interior models of \citet{takeda06} yield
522: $M_* = 1.52$~\msun and $R_* = 3.72$~\rsun. The
523: SPOCS Catalog lists $M_* =
524: 1.74$~\msun, and $R_* = 4.11$~\rsun\ \citep{valenti05}. The
525: variances of these different mass and
526: radius estimates are 0.1~\msun\ and 0.2~\rsun,
527: respectively, which are consistent with our estimate of uncertainties
528: in \S \ref{stellar}. The other stellar properties are listed in
529: Table~\ref{stellartable}.
530:
531: \begin{figure}[t!]
532: \epsscale{1}
533: \plotone{fig5.eps}
534: \caption{\footnotesize{Radial velocities for \starB\ measured at Keck
535: Observatory. The error bars represent the internal uncertainty of
536: each measurement without accounting for stellar jitter. \label{velplotB}} }
537: \end{figure}
538:
539: \begin{figure}[h!]
540: \epsscale{1}
541: \plotone{fig6.eps}
542: \caption{\footnotesize{\emph{Top:} Periodogram analysis of the RV time series of
543: \starB. A strong peak is visible near $P = 300$~d with an analytic
544: false--alarm probability FAP$< 0.1$\%. \emph{Bottom:} Empirical assessment of
545: the FAP of the best--fit Keplerian model. The original, unscrambled
546: velocities yield an orbital solution with \chisq$ = \chiB$ (arrow).
547: The histogram shows the distribution of \chisq\ obtained from the
548: best--fit orbital solution for each of the scrambled--velocities trials.
549: None of the $10^4$ trials
550: produced a value of \chisq\ lower than the value obtained from the
551: original time series, resulting in FAP$< 0.01$\% (cf
552: \S~\ref{sectionB}). \label{pgramB}} }
553: \end{figure}
554: \clearpage
555:
556: \begin{figure}[h!]
557: \epsscale{1}
558: \plotone{fig7.eps}
559: \caption{\footnotesize{Radial velocity observations of \starB\ phased at
560: $P = \pB$~d. The gray points lie outside of phases 0.0 and 1.0 and
561: are included to guide the eye. The dashed line shows the best-fit
562: Keplerian orbital solution, which has \chisq$=\chiB$. \label{orbitB}} }
563: \end{figure}
564:
565: \starB\ was one of the original stars added to the CCPS Keck program in
566: 1996, and was subsequently added to our list of intermediate--mass
567: stars in 2004. Table \ref{vel175541} lists our \nobsB\ Doppler
568: measurements along with their observation dates and internal
569: uncertainties (without jitter). Figure \ref{velplotB} shows that the rms scatter of
570: the velocity measurements is a factor of 6 greater than the mean internal
571: uncertainty ($\bar{\sigma_v} \approx 2$~\ms), and 2--3 times larger
572: than the rms scatter of stable Keck subgiants (Figure \ref{std_stars}).
573: A Lombe--Scargle periodogram analysis of the velocities reveals a
574: pronounced peak near $P=300$~d, with an analytical false--alarm
575: probability $\rm < 0.1\%$ (Figure \ref{pgramB}).
576:
577:
578: To search for the best--fit orbital solution, we added 5~\ms\ of
579: jitter in quadrature to the internal measurement uncertainties. We find
580: that a Keplerian with $P=\pB$~d, $K = \kB$~\ms\
581: and $e = \eB$ provides the best fit to the data, resulting in
582: $\rm rms = \rmsB$~\ms\ and \chisq~$ = $~\chiB. Figure \ref{orbitB}
583: shows the the radial velocities
584: phased at $P = \pB$~day, along with the best--fit orbital solution
585: (the gray points show the measurements at phases outside of phases 0.0
586: and 1.0, in order to guide the eye). Assuming a stellar
587: mass of \mstarB~\msun, we estimate a minimum planet mass \msini~$=
588: \msiniB$~\mjup\ and orbital separation $a = \arelB$~AU.
589:
590:
591: While the strong periodogram peak and low \chisq\ are
592: indicative of a correlated signal resulting from an orbiting planet,
593: it is still possible that random variability could
594: conspire to produce a false periodicity in our sparse series of
595: measurements. To test the null hypothesis, we used the
596: ``scrambled'' velocity false--alarm test described by
597: \citet{marcy05b}. For $10^4$ separate trials, we held the observation
598: times constant and scrambled the order of the measurements using a
599: pseudo random number generator. This has the effect of keeping the
600: sampling constant while removing any true temporal coherence, if such
601: a signal exists. For each
602: of the scrambled trials, we perform a full search for the
603: best--fit Keplerian orbital solution---with jitter---and record the
604: \chisq\ from the fit.
605:
606: The distribution of \chisq\ generated from the scrambled--velocity
607: trials is then compared to the fit obtained from the original time
608: series, as shown in the lower panel of Figure \ref{pgramB}. None of the
609: $10^4$ scrambled trials produced a \chisq\ equal to or lower than the
610: best--fit solution to the original time series, resulting in a
611: false--alarm probability of $< 0.01$\%. From this test, we
612: conclude that the temporally correlated signal seen in the velocity
613: time series is likely real, rather than an artifact of random
614: noise. We find that the best explanation of the periodic signal is the
615: presence of an unseen planetary companion orbiting \starB.
616:
617: \section{Summary and Discussion}
618: \label{summary}
619:
620: We present precision Doppler measurements of \npllet\
621: intermediate--mass subgiants that show periodic variations
622: in their radial velocities consistent with planet--mass orbital
623: companions. Interpolation of the stars' absolute magnitudes, colors
624: and metallicities onto the \citet{girardi02} stellar interior models
625: shows that all \npllet\ stars have
626: masses ranging from 1.65~\msun\ to 1.85~\msun. Figure \ref{sg_hr} shows these
627: massive host stars on an H--R diagram, along with their theoretical
628: evolution tracks. Following the tracks back to the zero--age
629: main sequence reveals that these present--day subgiants were originally
630: early--type dwarfs with $B-V \lesssim 0.2$ and spectral types
631: ranging from A2V to A5V. The \npllet\ long--period planets presented here
632: would not have been detectable during their stars'
633: main--sequence phases due to the jitter and rotational line broadening
634: typical for intermediate--mass dwarfs. These planets orbiting
635: ``retired'' A stars illustrate how evolved stars provide a unique
636: window into stellar mass and planetary domains otherwise
637: inaccessible to Doppler--based planet searches.
638:
639: \begin{figure}[t!]
640: \epsscale{1}
641: \plotone{fig8.eps}
642: \caption {\footnotesize{H--R diagram illustrating the properties of the
643: \npllet\ subgiant planet host stars (pentagrams) compared to their
644: main--sequence progenitors (filled circles). The connecting lines
645: represent each star's \citet{girardi02} theoretical mass track,
646: interpolated for that star's metallicity. The thick, diagonal line
647: is the theoretical zero--age main sequence for [Fe/H]=0.0. The hashed
648: region shows the approximate range of colors and magnitudes of stars with
649: spectral types A2V--A8V. \label{sg_hr}} }
650: \end{figure}
651:
652: There are now 9 former
653: A--type stars ($1.6 \lesssim M_* < 3.0$~\msun) with planetary
654: companions. We list some of the properties of these massive host
655: stars and their planets in
656: Table~\ref{massive_table}. All 9 planets orbit beyond
657: $\sim$0.78~AU from their stars. This paucity of planets with semimajor
658: axes $a < 0.78$~AU is unlikely to be due to a detection bias. For
659: a given planet mass and stellar mass, the velocity
660: semiamplitude of a star scales as $K \sim a^{-1/2}$, making planets in
661: smaller orbits easier to detect. The detectability
662: of close--in planets is also facilitated by the increased number of
663: orbital cycles that are observable over a given time span.
664:
665: We consider two possible explanations for the observed lack of
666: close--in planets around intermediate--mass stars. The first
667: possibility is that planets around A--type stars have the same
668: semimajor axis distribution as planets orbiting lower--mass stars, but
669: the close--in planets were destroyed by the expanding atmospheres of
670: their giant host
671: stars. Alternatively, planets orbiting A--type stars may have a
672: different semimajor axis distribution than lower--mass stars, with
673: planets residing preferentially in long--period orbits beyond
674: $\sim0.8$~AU.
675:
676: These possibilities can be explored by comparing the properties of
677: planets in Table~\ref{massive_table} to planets orbiting lower--mass
678: stars listed in the CNE. We exclude extremely low--mass planets with
679: $K < 15$~\ms\ that would not be easily detectible around higher--mass
680: subgiants and giants. We use a one--sided Kolmogorov--Smirnov (K--S)
681: test to compare the semimajor axis distributions of planets around
682: intermediate--mass and lower--mass stars \citep{press}. We find the
683: probability that the two distributions are identical is only
684: 0.06\%. Under the assumption that the semimajor axis distribution of
685: planets is independent of stellar mass, short--period planets orbiting
686: evolved A--type stars must be efficiently destroyed by the expanding
687: atmospheres of their giant host stars. The validity of this
688: hypothesis depends on whether the radii of $\sim2$~\msun\
689: giants are large enough to engulf planets out to $\sim0.8$~AU.
690:
691: Figure~\ref{rstar_hj} shows the evolution of the radius of a
692: 2.0~\msun\ star according to the \citet{girardi02} stellar evolution
693: models. As the star crosses the Hertzsprung Gap during its subgiant
694: phase, its radius remains nearly constant at $a = 5$~\rsun~$ =
695: 0.023$~AU, which is within the orbit of a $P = 3$~day hot
696: Jupiter. Not until the star begins to ascend the RGB
697: does its outer atmosphere begin to encroach on the orbits of
698: short--period planets. But even at the tip of the RGB (near the helium
699: flash), the radius of a
700: 2~\msun\ star is only at the distance of a 10~day hot Jupiter at $a
701: \approx 26$~\rsun~$=0.12$~AU (the radius of a 2.5~\msun\ red giant is
702: not much larger at $a \approx 32$~\rsun~$=0.15$~AU). Thus, engulfment
703: cannot be solely responsible for the lack of close--in
704: planets around subgiants and K giants. Indeed, engulfment can only be
705: important for 4 of the
706: stars in Table~\ref{massive_table}: the post--helium--flash clump
707: giants HD\,104985, HD\,11977 and $\epsilon$~Tau, and HD\,13189 which
708: has a poorly constrained radius due to its highly uncertain parallax.
709:
710: \begin{figure}
711: \epsscale{1}
712: \plotone{fig9.eps}
713: \caption {Radius of a 1.9~\msun\ star as it evolves off of the main
714: sequence (becoming redder). The horizontal dashed line s
715: depict the semimajor axes of planets with periods of 3 and 10
716: days. The vertical
717: dot--dashed line shows the approximate $B-V$ color of the star as it
718: begins to ascend the red giant branch. Because the radii of
719: subgiants are small enough to
720: avoid interference with close--in planets, our Doppler survey
721: is sensitive to the same range of orbital separations as
722: surveys of main--sequence stars. Planets with $a \lesssim 30$~\rsun\ may
723: be destroyed by the expansion of their host stars on the red giant
724: branch. \label{rstar_hj}}
725: \end{figure}
726:
727: The evolution of planetary orbits from 0.05--0.15~AU in the presense of
728: an expanding stellar atmosphere has not been examined in detail. The
729: effects of planet engulfment on its host star have been studied by
730: \citet{siess99}, but a key assumption in their model is that the
731: substellar companion is destroyed. Since it is unclear what
732: happens to a planet when it interacts with the atmosphere of its
733: expanding host star, we simply assume that planets orbiting within the radius
734: of a giant star are destroyed\footnote{\citet{maxted06} discovered
735: a short--period substellar companion that apparently survived
736: engulfment as its parent star evolved into a white dwarf. However, no
737: Jovian planet has yet been detected around a white dwarf.}.
738: Under this assumption, we would expect a deficiency of hot Jupiters
739: around clump giants out to $\sim$0.15~AU, but no corresponding
740: deficiency around subgiants and K giants.
741:
742: We now analyze the lack of close--in planets around the sample in
743: Table~\ref{massive_table} accounting the possible destruction of hot
744: Jupiters around clump giants. For subgiants and giants we can use the
745: K--S test as before, which yields a probability of 0.7\% that the
746: semimajor axis distribution is the same as lower--mass stars in the
747: CNE. For clump giants we exclude planets from the CNE with $a <
748: 0.15$~AU, and the corresponding probability from the K--S test is
749: 1.7\%. Thus, the distribution of close--in
750: planets around former A--type stars remains inconsistent with the
751: distribution of planets in the CNE. Since engulfment does not provide
752: an adequate explanation for the lack of close--in planets in
753: Table~\ref{massive_table}, we are left with the possibility that the
754: semimajor axis distribution of planet around A--type stars is
755: significantly different than the distribution around lower--mass stars
756: ($M_* < 1.6$~\msun).
757:
758: Differences between the semimajor axes of planets around stars of
759: various masses has previously been investigated by \citet{burkert06}.
760: From their study of the orbital properties of known exoplanets, they
761: find evidence of a gap in the semimajor axis distribution
762: around stars with masses $M_* \geq 1.2$~\msun, with fewer planets
763: between 0.08 AU and 0.6 AU compared to lower--mass stars. They were
764: able to reproduce this gap in their Monte Carlo simulations of planet
765: migration, and they attribute the gap to the shorter depletion
766: timescales of disks around intermediate--mass stars.
767:
768: The semimajor axis distribution of planets as a function of stellar
769: mass can be investigated further with the inclusion of a larger sample
770: of intermediate--mass subgiants in Doppler-based planet searches. As
771: Figure~\ref{rstar_hj} shows, Doppler surveys of subgiants can probe
772: occurrence of Jovian planets at orbital distances ranging from many AU
773: down to as close as 0.05~AU, the realm of hot Jupiters. The smaller radii of
774: subgiants also result in higher surface gravities compared to giants,
775: which leads to lower levels of pulsation--induced
776: jitter. \citet{hekker06} show that giants with $B-V > 1.2$ typically
777: have jitter values greater than 20~\ms, ostensibly due to radial and
778: non--radial pulsation modes. Only giants blueward of this limit are
779: stable to within 20~\ms, compared to the 4--6~\ms\ of jitter seen in
780: subgiants ($B-V < 1.0, M_V \lesssim 2.0$). This increased velocity
781: stability, coupled with their relatively small radii, therefore make
782: subgiants ideal proxies for A--type dwarfs in Doppler--based planet
783: searches.
784:
785: The primary limitation of subgiants is their relative scarcity, which
786: restricts the number of bright targets suitable for high--resolution
787: spectroscopic observations. The
788: time it takes stars to cross the Hertzsprung Gap is small compared to
789: the star's lifetime---of order 100~Myr---rendering Hertzsprung Gap
790: stars within 200~pc rare compared to main--sequence stars and
791: giants. Additional targets can be
792: found further from the Sun, with fainter apparent magnitudes ($V
793: \gtrsim 7.5$). In the near future, we plan to expand our sample of
794: subgiants using the Keck telescope and HIRES spectrometer in order to
795: further investigate the orbital properties, planet masses and
796: occurrence rate of planets orbiting intermediate--mass stars.
797: As the number of subgiants included in
798: Doppler surveys increases, it will become apparent whether the lack
799: of short--period planets around intermediate--mass stars is a result of
800: different formation and migration mechanisms in the disks of A--type
801: stars, or simply a consequence of the small number of massive subgiants
802: currently surveyed.
803:
804: \acknowledgements
805:
806: We extend our gratitude to the many CAT observers who have helped
807: with this project, including Howard Isaacson, Julia Kregenow, Karin
808: Sandstrom, Bernie
809: Walp, Peter Williams, Katie Peek and Shannon Patel. Special thanks
810: to Herv\'e Buoy and Francisco Ramos-Stierle for lending a portion of
811: their 3m observing time to observe HD\,192699 before it set in 2006.
812: We thank Michael Fitzgerald and Marshall Perrin for their useful
813: discussions, and Tim Robishaw for sharing his data display expertise
814: and IDL plotting routines. We
815: also gratefully acknowledge the efforts and dedication of the Lick
816: Observatory and Keck Observatory staff, and the time assignment
817: committees of NASA, NOAO and University of California for their
818: generous allocations of observing time. We appreciate funding from
819: NASA grant NNG05GK92G (to GWM), and the NSF for its grant AST-0307493
820: (to SSV) for supporting this research. DAF is a Cottrell
821: Science Scholar of Research Corporation and acknowledges support from
822: NASA Grant NNG05G164G that made this work possible. This research has
823: made use of the Simbad database operated at CDS, Strasbourg France,
824: and the NASA ADS database. The authors wish to
825: extend special thanks to those of Hawaiian ancestry on whose sacred
826: mountain of Mauna Kea we are privileged to be guests. Without their
827: generous hospitality, the Keck observations presented herein would not
828: have been possible.
829:
830: \include{bbl}
831:
832: \include{table1}
833:
834: \clearpage
835: \begin{deluxetable}{lllll}
836: \tablecaption{Stellar Parameters\label{stellartable}}
837: \tablewidth{0pt}
838: \tablehead{\colhead{Parameter} &
839: \colhead{\starC} &
840: \colhead{\starA} &
841: \colhead{\starB} \\
842: }
843: \startdata
844: V & \vmagC & \vmagA & \vmagB \\
845: $M_V$ & \mvC & \mvA & \mvB \\
846: B-V & \bvC & \bvA & \bvB \\
847: Distance (pc) & \dC & \dA & \dB \\
848: ${\rm [Fe/H]}$ & \feC~(0.04) & \feA~(0.04) & \feB~(0.04) \\
849: $T_{eff}$~(K) & \teffC~(44) & \teffA~(44) & \teffB~(44) \\
850: \vsini~(\ks) & \vsiniC~(0.5) & \vsiniA~(0.5) & \vsiniB~(0.5) \\
851: $\log{g}$ & \loggC~(0.3) & \loggA~(0.3) & \loggB~(0.3) \\
852: $M_{*}$~(\msun) & \mstarC~(0.12) & \mstarA~(0.13) & \mstarB~(0.12) \\
853: $R_{*}$~(\rsun) & \rstarC~(0.51) & \rstarA~(0.57) & \rstarB~(0.46) \\
854: $L_{*}$~(\lsun) & \lstarC~(0.3) & \lstarA~(0.3) & \lstarB~(0.3) \\
855: Age~(Gyr) & \ageC~(1.0) & \ageA~(1.0) & \ageB~(1.0) \\
856: $S_{HK}$ & \shkC & \shkA & \shkB \\
857: $\log R'_{HK}$ & \rhkC & \rhkA & \rhkB \\
858: \enddata
859: \end{deluxetable}
860:
861: \include{table2}
862:
863: \clearpage
864: \begin{deluxetable}{lllll}
865: \tablecaption{Orbital Parameters\label{orbittable}}
866: \tablewidth{0pt}
867: \tablehead{\colhead{Parameter} &
868: \colhead{\starC\,b} &
869: \colhead{\starA\,b} &
870: \colhead{\starB\,b} \\
871: }
872: \startdata
873: P (d) & \pC~(\peC) & \pA~(\peA) & \pB~(\peB) \\
874: T$_p$\tablenotemark{a}~(JD) & \tpC~(\tpeC) & \tpA~(\tpeA) & \tpB~(\tpeB) \\
875: e & \eC~(\eeC) & \eA~(\eeA) & \eB~(\eeB) \\
876: K$_1$~(\ms) & \kC~(\keC) & \kA~(\keA) & \kB~(\keB) \\
877: $\omega$~(deg) & \omC~(\omeC) & \omA~(\omeA) & \omB~(\omeB) \\
878: \msini~(\mjup) & \msiniC & \msiniA & \msiniB \\
879: $a$~(AU) & \arelC & \arelA & \arelB \\
880: Fit RMS~(\ms) & \rmsC & \rmsA & \rmsB \\
881: \chisq & \chiC & \chiA & \chiB \\
882: N$_{\rm obs}$ & \nobsC & \nobsA & \nobsB \\
883: \enddata
884: \tablenotetext{a}{Time of periastron passage.}
885: \end{deluxetable}
886:
887: \include{table3}
888: \include{table4}
889:
890: \end{document}