THE ASTRONOMICAL JOURNAL, 121:1136È1146, 2001 February ( 2001. The American Astronomical Society. All rights reserved. Printed in U.S.A.
LIMITS ON LINE BISECTOR VARIABILITY FOR STARS WITH EXTRASOLAR PLANETS M. S. POVICH,1 M. S. GIAMPAPA,2 J. A. VALENTI,3,4 T. TILLEMAN,2 S. BARDEN,3 D. DEMING,5 W. C. LIVINGSTON,2 AND C. PILACHOWSKI3 Received 2000 July 21; accepted 2000 November 3 ABSTRACT We present an analysis of high-resolution synoptic spectra of ten F- and G-type stars, seven of which exhibit periodic radial velocity variations due to the presence of one or more substellar companions. We searched for subtle periodic variations in photospheric line asymmetry, as characterized by line bisectors. In principle, periodic variations in line asymmetry observed at lower spectral resolution could mimic the radial velocity signature of a companion, but we Ðnd no signiÐcant evidence of such behavior in our data. Observations were obtained from 1998 March to 1999 February using the National Solar Observa- tory (NSO) 1.52 m McMath-Pierce Solar Telescope Facility on Kitt Peak in conjunction with the solar- stellar spectrograph, achieving a resolving power of 1.2 ] 105. To characterize line asymmetry, we Ðrst measured line bisectors for the unblended Fe I photospheric line at 625.26 nm. To improve sensitivity to small Ñuctuations, we then combined points in each bisector to form a velocity displacement with respect to the line core. We searched for periodic variations in this displacement, Ðnding no substantial di†er- ence between stars with substellar companions and those without reported companions. We Ðnd no correlation between bisector velocity displacement and the known orbital phase of substellar companions around our target stars. Simulations of a periodic signal with noise levels that mimic our measurement errors suggest that we can exclude bisector variations with amplitudes greater than about 20 m s~1. These results support the conclusion that extrasolar planets best explain the observed periodic variations in radial velocity. Key words: line: proÐles È planetary systems È stars: late-type
1. INTRODUCTION ing a superposition of two proÐles shifted relative to each High-precision radial velocity studies of nearby stars other. The net proÐle is asymmetric because of unequal have resulted in the Ðrst detections of extrasolar planetary contributions from the rising and falling components. In the systems, consisting primarily of ““ massive Jupiters ÏÏ (Mayor Sun, the asymmetry appears as a C-shaped curvature of the & Queloz 1995; Marcy & Butler 1998, and references line bisector, which is the locus of points midway between therein). These detections rely on achieving precisions of the red and blue wings of a spectral line. The bisector is 3È15ms~1. Stellar activity in the form of spots and plages, deÐned and discussed extensively by Gray (1988 and refer- nonradial pulsations, and convective inhomogeneities com- ences therein), especially within the context of convection in bined with rotation can distort stellar photospheric lines late-type stars. beyond the intrinsic asymmetries arising from surface con- Gray (1997) originally suggested that subtle temporal vection, thereby contributing error in the form of ““ stellar variations in (R D 105) line bisectors for 51 Peg might be noise ÏÏ to sensitive radial velocity measurements (Saar & the source of radial velocity variations attributed to the Donahue 1997; Hatzes 1996). Further improvement in gravitational inÑuence of a planetary companion. GrayÏs radial velocity precision beyond this intrinsic variability (1997) challenge of the planetary companion hypothesis was limit will require increased attention to line-proÐle pertur- based on a sparse data-set that appeared to show regular bations caused by stellar atmospheric phenomena. In par- variations in the bisector span of an Fe I photospheric line ticular, changes in line shape arising from stellar with a period equal to the orbital period of the putative atmospheric motion can mimic small radial velocity varia- planet. However, subsequent investigations by Gray (1998) tions at the spectral resolution (R D 5È7 ] 104) typically and Hatzes, Cochran, & Bakker (1998) based on an inten- utilized for planet searches. sive series of observations of 51 Peg did not detect any Convection in the photosphere consists of rising, hotter systematic changes in the line shapes that could be misinter- material (associated with bright granules on the Sun) and preted as radial velocity shifts. Thus, the data are most falling, cooler material (e.g., the intergranular lanes), yield- compatible with the presence of a sub-stellar mass compan- ÈÈÈÈÈÈÈÈÈÈÈÈÈÈÈ ion to 51 Peg. 1 National Solar Observatory, Research Experiences for Under- Following the challenge by Gray (1997), we initiated a graduates, summer 1999. Also at Department of Astronomy, Harvard Uni- long-term program of high spectral resolution observations versity, Cambridge, MA 02138. of seven late-type stars with reported planetary companions 2 National Solar Observatory, National Optical Astronomy Observa- along with three additional late-type stars with no known tory, P.O. Box 26732, Tucson, AZ 85726-6732. The National Solar Obser- vatory and the National Optical Astronomy Observatory are each substellar companions down to the sensitivity limits of operated for the National Science Foundation by the Association of Uni- current radial velocity monitoring programs. We note that versities for Research in Astronomy, Inc. FrancÓois et al. (1999) also conducted an intensive series of 3 National Optical Astronomy Observatory, Tucson, AZ 85726-6732. high spectral resolution observations of 51 Peg and t And 4 Now at Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218. over four nights at the Canada-France-Hawaii Telescope 5 NASA/Goddard Space Flight Center, Code 693, Greenbelt, MD (CFHT). These investigators did not observe any regular 20771. change in line bisector shape that could give rise to spurious 1136 STARS WITH EXTRASOLAR PLANETS 1137
Doppler shifts occurring with the period of the planetary spectra with equal line depths. For 55 Cnc, the background companion. Our investigation extends these studies correction required to match all the slices also yielded line through the addition of a longer temporal baseline of rela- depths in agreement with a spectrum obtained independent- tively frequent observations of several late-type stars ly by G. Marcy using the Lick Hamilton echelle. The excel- reported to have planetary companions. lent agreement between the spectra validated our empirical OBSERVATIONS AND REDUCTION correction scheme. 2. We constructed a reference spectrum for each star by The spectra were obtained with the 1.52 m main telescope reducing and summing spectra from three separate images of the NSO/Kitt Peak McMath-Pierce Solar Telescope (see Fig. 1). The remaining spectra for that star were then Facility in conjunction with the stellar spectrograph. The mapped to the reference spectrum, using a nonlinear least- stellar spectrograph is described by Smith & Giampapa squares Ðtting routine that quadratically mapped each indi- (1987). We utilized the echelle grating with the 180 mm vidual wavelength scale onto the reference wavelength scale, Nikon transfer lens, TI 800 ] 800 UV-enhanced CCD and Ðt a polynomial to match the continua, and applied a back- a10] 200 mm Bowen-Walraven image slicer. This con- ground correction of the form in equation (1) to yield a set Ðguration yielded a resolution of R \ 1.20 ] 105 as mea- of consistent spectra. Finally, we created a high signal-to- sured directly from the full widths at half-maximum noise ratio (S/N [ 1500) template spectrum for each of our (FWHMs) of the ThAr comparison source lines. All spectra target stars by combining all the individual dithered spectra were centered on the Fe I photospheric absorption line at using an Interactive Data Language (IDL) routine which 625.3 nm used by Gray (1998) in his study of 51 Peg. The performs neighborhood averaging on unevenly sampled program stars are presented in Table 1. data. This template spectrum is used to estimate bisector A quartz lamp and a ThAr comparison source were uti- errors (° 2.2). lized for Ñat Ðelding and wavelength calibration, respec- tively. A barycentric correction was applied so that all 2.1. L ine Bisector Analysis spectra for a given program star are on a uniform wave- We construct the bisector for the Fe I 6252.6A line fol- length scale. While not essential for the analysis of proÐle lowing the method described by Gray (1988). SpeciÐcally, shapes, this facilitated the combining of multiple frames, we locate the pixel closest to the line core and then step up including the construction of a very high signal-to-noise the blue side of the proÐle pixel by pixel. From each blue ratio average spectrum that is used as a template spectrum pixel, we extend a horizontal (i.e., same level of residual for the error analysis (° 2.2). intensity) line segment across the proÐle. The wavelength Our initial inspection of the reduced spectra revealed line value for the red end of the segment is determined though depth variations in photospheric features on a night-to- linear interpolation between the neighboring pixels. The night basis. This apparent variability of the line strength is bisector connects the midpoints of these line segments. We an artifact which likely is related to the uneven illumination adopt linear interpolation because error propagation is of the slicer assembly by the stellar image in conditions of more straightforward, bisector points are more independent variable seeing. Extraction of the individual slicer spectra of one another, and outliers do not cause ringing. There is showed that the more faintly illuminated slices yielded shal- likely some curvature across a pixel, especially in the line lower absorption features. To each spectrum, we added an core, but we Ðtted the line core with a parabola and curva- empirical ““ background ÏÏ correction, x, and then renorma- ture is minimal in the steep part of the line proÐles, where lized the continuum, obtaining we use linear interpolation. In addition, the points on one S ] x side of the proÐle are used without interpolation which S \ , (1) further mitigates the e†ect of curvature. Typically, we used corr 1 ] x seven bisector points for Fe I 6252.6A in G-type stars and where S is the original reduced spectrum andScorr is the twelve bisector points for the more rotationally broadened corrected spectrum. This approach yielded individual slicer F-type stars. A typical line bisector is illustrated in Figure 2.
TABLE 1 TARGET STARS
Perioda K a,b ~1 HR Name Type (days) T0 (m s ) 458...... t Andc F8 V 4.6163 50542.736 60.1 1101 ...... 10Tau F9IVÈV ...... 3522 ...... 55Cnc G8V 14.6597 50484.188 72.8 4277 ...... 47UMa G1V 1026.58 50405.303 47.1 5072 ...... 70Vir G4V 116.666 50539.287 317.0 5185 ...... q Boo F6 IV 3.31253 50537.569 468.2 5933 ...... c Ser F6 IVÈV ...... 5968 ...... o CrB G0 Va 39.8412 50975.357 60.7 7503 ...... 16CygA G1.5 Vb ...... 8729 ...... 51Peg G2.5 IVa 4.23110 50509.695 56.1
a Calculated from data courtesy G. Marcy. b JD [2,400,000. c Multiple planet systemÈDoppler parameters are given for the innermost companion only (t And b). 1.2 1.2 υ Andromedae 10 Tauri
1.0 1.0 V I 6251.8 V I 6251.8
0.8 0.8
0.6 0.6 Fe I 6252.6
0.4 0.4 Fe I 6252.6
0.2 0.2
0.0 0.0 6240 6245 6250 6255 6260 6240 6245 6250 6255 6260 Wavelength (Å) Wavelength (Å)
1.2 1.2 55 Cancri 47 Ursae Majoris
1.0 1.0 V I 6251.8 0.8 0.8
0.6 0.6
0.4 V I 6251.8 0.4 Fe I 6252.6
0.2 0.2 Fe I 6252.6
0.0 0.0 6240 6245 6250 6255 6260 6240 6245 6250 6255 6260 Wavelength (Å) Wavelength (Å)
1.2 1.2 70 Virginis τ Bootis
1.0 1.0 V I 6251.8
0.8 0.8 V I 6251.8 Fe I 6252.6
0.6 0.6
0.4 0.4 Fe I 6252.6 0.2 0.2
0.0 0.0 6240 6245 6250 6255 6260 6240 6245 6250 6255 6260 6265 Wavelength (Å) Wavelength (Å)
FIG. 1.ÈTemplate spectra of all the stars in our data set. Spectral resolution (j/dj)is^1.2 ] 105 with S/N [ 1500. The F-type stars (see Table 1) exhibit shallower, rotationally broadened absorption lines. The spectrum of 55 Cnc contains many more lines than are seen in our other target stars, most likely due to an enhanced CN content in the photosphere. STARS WITH EXTRASOLAR PLANETS 1139
1.2 1.2 γ Serpentis ρ Coronae Borealis
1.0 1.0 V I 6251.8 V I 6251.8 0.8 0.8 Fe I 6252.6
0.6 0.6
0.4 0.4 Fe I 6252.6
0.2 0.2
0.0 0.0 6240 6245 6250 6255 6260 6265 6240 6245 6250 6255 6260 Wavelength (Å) Wavelength (Å)
1.2 1.2 16 Cygni A 51 Pegasi
1.0 1.0 V I 6251.8
0.8 0.8 V I 6251.8
0.6 0.6
0.4 0.4 Fe I 6252.6 Fe I 6252.6 0.2 0.2
0.0 0.0 6240 6245 6250 6255 6260 6240 6245 6250 6255 6260 6265 Wavelength (Å) Wavelength (Å)
FIG. 1.ÈContinued
Once we have derived the bisector, we measure velocity mization. This method is usually not feasible for stellar o†setsv1, v2, andv3, corresponding to residual intensities observations which are typically characterized by lower y1, y2, andy3 (see Fig. 2). Velocity o†sets are measured with spectral resolution and S/N. In the context of this investiga- respect to the center of a parabolic Ðt to the line core. We tion, we deÐne the velocity span as determinev for a giveny by linear interpolation of adja- i i v \ v [ v cent bisector points. Following Gray (1998) and Gray & s 3 1 . (2) \ Hatzes (1997), we chose(y1, y 2, y3) (0.83, 0.7, 0.5) for the The other principal measure of line asymmetry is bisector Fe I line, which for 51 Peg has a residual intensity of ^0.3 in curvature, which has been described as the di†erence the core. For other stars, the measurement depths, y, were between bisector spans computed for the top and bottom scaled according to observed line depth. halves of a line (Gray & Hatzes 1997; Gray 1998). In our Line asymmetries are typically characterized by velocity terminology, this is equivalent to span (or amplitude) or bisector curvature. For example, the \ [ [ [ former approach is used by Livingston et al. (1999) to cb (v3 v2) (v2 v1) . (3) measure line-proÐle asymmetry in solar spectra. Livingston Since the curvature,c , is a second derivative (whereas \ [ b et al. (1999) deÐne the bisector amplitude as ab jc xmin, span is a Ðrst derivative), the signal-to-noise requirements wherejc is the observed central wavelength of the absorp- are more demanding in order to obtain signiÐcant measures tion feature andxmin is the most extreme blue point in the of variability in the bisector curvature. We therefore bisector. This approach is especially appropriate for solar adopted an alternative characterization which is a variant observations where the combination of very high spectral of the span method designed to increase sensitivity. We resolutions(R Z 3 ] 105)( and S/NZ5000) yields well- deÐne this velocity displacement as deÐned bisectors with a large number of independent ] ] points. Consequently, derivatives of the bisector itself can \ v1 v2 v3 [ vb jc , (4) be determined andxmin can be deduced through mini- 3 1140 POVICH ET AL. Vol. 121
v (m s−1) the template is between 0.95 and 1.02. We assign uncer- −300 −200 −100 0 100 200 300 1.0 tainties(ps) to every point in the spectrum by multiplying the continuum uncertainty by the square root of the
y1 v residual intensity. In other words, we assume that uncer- 1 tainties are controlled by Poisson statistics in the stellar 0.8 signal, which is approximately true for these bright stars.
y2 v Uncertainties in residual intensity Ðrst propagate into 2 our bisector determination when interpolating the red wing of the proÐle as illustrated in Figure 2. Uncertainty in the 0.6 wavelength corresponding to a particular residual intensity is given by
y3 Residual Intensity v \ 3 pw ps/m , (5) 0.4 where m is the slope of the spectrum (Gray 1988), which is measured using the essentially noiseless template. We then propagate errors through successive steps in the bisector λ c construction and interpolation onto reference depths (y)to 0.2 obtainp , p , andp , which are the uncertainties in v , v , 6252.3 6252.4 6252.5 6252.6 6252.7 1 2 3 1 2 Wavelength λ (Å) andv3, respectively. Uncertainty in the wavelength of the line core,pjc, is derived from estimated uncertainties in the FIG. 2.ÈSchematic diagram of a spectral line bisector, illustrating the residual intensity points being Ðtted. The Ðnal error invb is C shape and evaluation pointsv1, v2, andv3 (plotted as times cross). The bisector points (small Ðlled circles) are the midpoints of the horizontal, p2]p2]p2 dashed-dotted line segments, plotted on the wavelength scale of the spec- 2\ 1 2 3 ] 2 pb pjc . (6) trum. Filled squares are data points in the observed spectrum, while open 9 squares indicate where linear interpolation has been used to determine the red endpoints of the bisector segments. The solid line with large, open The observed scatter in bisector velocity displacement is circles is the entire bisector, plotted on the velocity (v) scale, which is consistent with the formal uncertainties calculated in this expanded by a factor of 40 relative to the wavelength scale. Zero velocity is manner, as discussed below. deÐned asjc, determined from a parabola Ðt to the line core (dashed curve). We also examined the uncertainties associated with other methods of characterizing bisectors, namely span and cur- vature. In particular, the uncertainty in span from equation wherevb is the average displacement of the evaluation points from the line core. (2) is 2\ 2] 2 A comparison of the formulations in equations (2)È(4) ps p1 p3 , (7) suggests that for a C-shaped bisector,v andv will be nega- b s while the uncertainty in curvature from equation (3) is tive while the curvature,cb, will be positive. The principal advantage ofv as deÐned by equation (4) is that by utilizing 2\ 2] 2] 2 b pc p1 4p2 p3 . (8) the average of all three evaluation points, the impact of the uncertainties in individual evaluation points is mitigated. In Table 2, we compare various characterizations of line We also note that our method of bisector measurement bisectors for 51 Peg, namely, velocity displacement (vb), relies on an accurate determination of the location of the span(vs), and curvature (cb). For each technique, we give the weighted mean of all measured values for 51 Peg (col. [2]), line core(jc), which we ascertain by Ðtting a parabola to the line core. the standard deviation about the mean (col. [3]), and the average formal uncertainty (col. [4]). The last column con- In equation (4), averagingv1, v2, andv3 reduces uncer- tainties relative to the separate use of these points in a Ðrst tains the observed standard deviation divided by the or second derivative. Uncertainties in the velocity shift of average formal error. Values near unity indicate that our the line core are reduced by Ðtting a parabola to a few estimates of formal uncertainty provide a realistic indicator points in the line core. In essence, we are utilizing two aver- of actual errors. aged points for the determination of the bisector amplitude. Table 2 shows that for our data, velocity displacement as In contrast, the curvature formulation in equation (3) uses deÐned in equation (4) yields smaller formal errors than span or curvature. Moreover, the mean velocity displace- three independent points with one point(v2) used twice. We Ðnd from our data that this latter approach signiÐcantly ment is as large as other characterizations, makingvb the increases measurement uncertainties (see ° 2.2). most sensitive diagnostic of variability. Since our purpose Finally, we do not include any additional bisector points here is to search for periodicity in bisector behavior, we in the construction ofv since shallower points are intrinsi- b TABLE 2 cally noisy and only further increase the uncertainty in v . b ISECTOR MPLITUDES FOR EG Deeper points impinge on the line core, which would com- B A 51 P promise the independence of our two bisector span refer- Measurea Mean (m s~1) Scatter (m s~1) Error (m s~1) ence points. (1) (2) (3) (4) Ratio [ 2.2. Error Analysis vb ...... 15 42 29 1.45 To estimate noise in the continuum, we divide each spec- vs ...... 2 83 61 1.36 trum by a summed template spectrum, which is essentially cb ...... 14 53 88 0.60 noiseless. We then calculate the standard deviation of this a Bisector amplitudes:vb is deÐned by eq. (4);vs is given by eq. (2);cb is ratio in continuum windows where the residual intensity of calculated according to eq. (3). No. 2, 2001 STARS WITH EXTRASOLAR PLANETS 1141
TABLE 3 SUMMARY OF RESULTS
~1 ~1 ~1 HR Name Type SvbT (m s ) SpT (m s ) pk (m s ) Nobs (1) (2) (3) (4) (5) (6) (7)
458...... t And F8 V 12 81 16 64 1101 ...... 10Tau F9IVÈV3136680 3522 ...... 55Cnc G8V [41 36 4 225 4277 ...... 47UMa G1V [3 43 6 121 5072 ...... 70Vir G4V [17 32 4 135 5185 ...... q Boo F6 IV [45 200 22 118 5933 ...... c Ser F6 IVÈV 46 129 15 120 5968 ...... o CrB G0 Va 5 33 4 105 7503 ...... 16CygA G1.5 Vb [15 42 5 82 8729 ...... 51Peg G2.5 IVa [15 29 4 115 adopt the formulation that provides the strongest con- to positive. Inspection of Figure 5 reveals no obvious straints on bisector amplitude variations, namely velocity periodic variability in bisector velocity displacement, displacement. For illustrative purposes only, we present in though excursions for individual observations may be mar- Figure 3 bisector curvature and span versus orbital phase ginally signiÐcant in some cases. for 51 Peg. In Table 4 we present the ratio of the observed scatter, pv, RESULTS AND DISCUSSION to the mean of the formal errors, SpT. This quantity is 3. analogous to the last entry in the Ðrst row of Table 1, but 3.1. SigniÐcance of Variability for all stars. For ratios greater than unity, bisector varia- In Table 3 we summarize our measurements and formal tions may be signiÐcant, even if not periodic. A comparison of the mean value of this ratio for stars with planets com- errors. Column (4) gives the mean ofvb for individual obser- vations, weighted by their corresponding uncertainties. Column (6) gives the formal uncertainty in the weighted TABLE 4 ATIOS OF CATTER TO EAN NCERTAINTIES mean,pk, calculated from uncertainties invb for each obser- R S M U vation. See equation (4.31) of Bevington & Robinson (1992) for details. Column (5) gives the mean value of the formal HR Name Type pv/SpT errors, SpT. Measurements for each night are depicted 458...... t And F8 V 0.7 graphically as time series in Figure 4. For convenient refer- 1101 ...... 10Tau F9IVÈV 1.1 ence,SvbT, pv, and SpT are given in the Ðgure. 3522 ...... 55Cnc G8V 1.5 A key goal of this investigation is to test whether line 4277 ...... 47UMa G1V 1.1 asymmetries vary periodically on time scales consistent 5072 ...... 70Vir G4V 1.2 with precisely measured radial velocity variations. For each 5185 ...... q Boo F6 IV 0.7 star, we plot in Figure 5 bisector velocities as a function of 5933 ...... c Ser F6 IVÈV 0.6 Julian date and orbital phase, as computed from planetary 5968 ...... o CrB G0 Va 1.4 orbital periods and reference times(T ) in Table 1. At the 7503 ...... 16CygA G1.5 Vb 0.9 0 8729 ...... 51Peg G2.5 IVa 1.4 reference time, the radial velocity transitions from negative
200 200 51 Peg −1
100
0 ) ) 50 −1 −1 (m s (m s c s v 0 v −100
−50
−200 −100 51 Peg −1
FIG. 3.ÈTime series of bisector amplitudes for 51 Peg according to (a) the curvature method and (b) the span deÐnition 150 150
100
100
50 ) ) −1 −1 0 50 (m s (m s b b v v
−50
0
−100 υ And 10 Tau −1 −1
200 100
100 50
0 0 ) ) −1 −1 (m s (m s b b v v −100 −50
−200 −100 55 Cnc 47 UMa −1 −1
100 400
200 50
0 ) ) −1 −1 0 (m s (m s b b v v −200
−50 −400 70 Vir τ Boo −1 −1
300 100
200 50
100 0 ) ) −1 −1 (m s (m s b b v v 0 −50
−100 −100 γ Ser ρ CrB −1 −1
100 100
50 50
0 ) ) −1 −1 0 (m s (m s b b v v −50
−50 −100 16 Cyg A 51 Peg −1 −1
150 200
100 100
50 0 ) ) −1 −1 0 (m s (m s b b v v −100 −50
−200 −100 υ And 55 Cnc −1 −1
100 100
50 50
0 ) ) −1 −1 0 (m s (m s b b v v −50
−50 −100 47 UMa 70 Vir −1 −1
IG F . 5.ÈPlots ofvb vs. phase / for those stars exhibiting periodic Doppler velocity variations due to the presence of a substellar companion. Orbital phase of the companion was computed from the parameters listed in Table 1. permit detection of a periodic signal with this amplitude. 3.4. Trend with Spectral Type The high v sin i and the relatively low amplitude of radial velocity variation, along with our errors, would mask any Finally, we display in Figure 7 mean bisector velocity displacementSv T as a function of B[V color. Inspection regular variations invb that might be present. In all other b program stars with planets, the periodogram readily of Figure 7 suggests that theSvbT is correlated with stellar revealed the simulated periodic signal with false alarm spectral type, a result that is consistent with Gray (1988). In probabilities that ranged from about several times 10~2 to particular, Gray (1988) notes that the redward turn of the D10~11. We conclude that for these stars, changes in line bisector in the line core (the bottom part of the C) is absent asymmetry are not the major cause of observed radial in early-type stars, only manifesting itself near spectral type velocity variations. G0. We would expect stars showing only the top of the C to For illustrative purposes, we present in Figure 6 the have positivevb, and this is indeed what we observe for most periodogram for the 51 Peg time series of observations of our F stars. For cooler stars, the velocity displacement along with a simulation that includes a periodic signal at becomes more negative. Note that 55 Cnc (G8V) is the only the orbital period and the velocity amplitude of the report- star in our sample with molecular CN lines visible in our ed planet. In the speciÐc case of 51 Peg, we performed simu- summed templates. The blue wing of the Fe I 6252.6A line is lations for a range of amplitudes extending from5ms~1 to blended with a line of CN, which may a†ect our bisector 50ms~1. The results reveal that variations at the orbital characterization for 55 Cnc. period with an amplitude of only 25 m s~1 are readily The signiÐcant departure of q Boo from the trend shown detectable within our errors at a false alarm probability in Figure 7 may be related to its relatively rapid rotation D10~3. But at amplitudes as small as 10 m s~1, any such period of 3.3 ^ 0.5 days (Baliunas et al. 1997). As discussed regular variations that may be present would not be dis- by Gray (1988), and conÐrmed in the case of the Sun by cernible in our data. Smith, Huang, & Livingston (1987), line asymmetries as 400 100
200 50
0 0 ) ) −1 −1 (m s (m s b b v v −200 −50
−400 −100 τ Boo ρ CrB −1 −1
100
50
0 ) −1 (m s b v −50
−100 51 Peg −1
FIG. 5.ÈContinued
10 10
51 Peg (observed) 51 Peg (simulation)
8 8
6 6
4 4 Normalized Power Normalized Power
2 2
0 0 0.05 0.10 0.15 0.20 0.25 0.05 0.10 0.15 0.20 0.25 Frequency (day−1) Frequency (day−1)
FIG. 6.ÈPeriodograms computed for the observed time series for 51 Peg (left) and a simulation assuming a periodic signal at the planetary orbital period (right). The simulation incorporates the observing window and errors that are present in the left panel. No signiÐcant peaks are seen in the observational data. A signiÐcant feature is evident in the simulation at a frequency of 0.2373 day~1, corresponding to a period of 4.214 days, near the planetary orbital period of 4.231 days. 1146 POVICH ET AL.
60 from theoretical and empirical considerations that extra- solar giant planet systems with small separations such as q γ Ser Boo should be characterized by both strong magnetic and 40 tidal interactions. In fact, q Boo exhibited the strongest 10 Tau magnetic interaction among the stars considered in their
20 theoretical investigation. In view of the well-known inter- υ And action between magnetic Ðelds and convection in the Sun ) −1 47 UMa and late-type stars, this speculation merits investigation. 0 ρ CrB > (m s b 4. CONCLUSIONS
REFERENCES Baliunas, S. L., Henry, G. W., Donahue, R. A., Fekel, F. C., & Soon, W. H. Marcy, G. W., & Butler, R. P. 1998, ARA&A, 36, 57 1997, ApJ, 474, L119 Mayor, M., & Queloz, D. 1995, Nature, 378, 355 Bevington, P. R., & Robinson, D. K. 1992, Data Reduction and Error Livingston, W., Wallace, L., Huang, Y., & Moise, E. 1999, in ASP Conf. Analysis for the Physical Sciences (2d ed.; New York: McGraw-Hill), 64 Ser. 183, High Resolution Solar Physics: Theory, Observations, and Cuntz, M., Saar, S. H., & Musielak, Z. E. 2000, ApJ, 533, L151 Techniques, ed. T. R. Rimmele, K. S. Balasubramaniam, & R. R. Radick FrancÓois, R., Briot, D., Spite, F., & Schneider, J. 1999, A&A, 349, 220 (San Francisco: ASP), 494 Gray, D. F. 1988, Lectures on Spectral-Line Analysis: F, G, and K Stars Press, W. H., & Rybicki, G. B. 1989, ApJ, 338, 277 (Arva, Ontario: Publisher) Saar, S. H., & Donahue, R. A. 1997, ApJ, 485, 319 ÈÈÈ. 1997, Nature, 385, 795 Smith, M. A., & Giampapa, M. S. 1987, in Lecture Notes in Physics 291, ÈÈÈ. 1998, Nature, 391, 153 Cool Stars, Stellar Systems, and the Sun, ed. J. Linsky & R. Stencel (New Gray, D. F., & Hatzes, A. P. 1997, ApJ, 490, 412 York: Springer), 477 Hatzes, A. P. 1996, PASP, 108, 839 Smith, M. A., Huang, Y.-R., & Livingston, W. 1987, PASP, 99, 297 Hatzes, A. P., Cochran, W. D., & Bakker, E. J. 1998, Nature, 391, 154