Hypervelocity Stars II. The Bound Population

0.41 MB·English

by  Warren R. Brown

Accepted to ApJ
PreprinttypesetusingLATEXstyleemulateapjv.6/22/04
HYPERVELOCITY STARS II. THE BOUND POPULATION
Warren R. Brown, Margaret J. Geller, Scott J. Kenyon, Michael J. Kurtz
SmithsonianAstrophysicalObservatory,60GardenSt,Cambridge,MA02138
and
Benjamin C. Bromley
DepartmentofPhysics,UniversityofUtah,115S1400E,Rm201,SaltLakeCity,UT84112
Accepted to ApJ
ABSTRACT
Hypervelocity stars (HVSs) are stars ejected completely out of the Milky Way by three-body in-
7
0 teractions with the massive black hole in the Galactic center. We describe 643 new spectroscopic
0 observations from our targeted survey for HVSs. We find a significant (3.5σ) excess of B-type stars
2 with large velocities +275 < vrf < 450 km s−1 and distances d > 10 kpc that are most plausibly
explainedasanewclassofHVSs: starsejectedfromtheGalacticcenteronboundorbits. IfaGalactic
n
centerejectionoriginiscorrect,thedistributionofHVSsontheskyshouldbeanisotropicforasurvey
a
J completetoafixedlimitingapparentmagnitude. TheunboundHVSsinoursurveyhaveamarginally
anisotropic distribution on the sky, consistent with the Galactic center ejection picture.
1
2 Subject headings: Galaxy: halo — Galaxy: center — Galaxy: stellar content — Galaxy: kinematics
and dynamics — stars: early-type

1
v
0 1. INTRODUCTION the origin of massive stars in the Galactic Center, and
0 HVSs are a natural consequence of a massive black thehistoryofstellarinteractionswiththeMBH.Clearly,
6 a larger sample of HVSs will be a rich source for further
hole (MBH) in a dense stellar environment like that in
1 progress on these issues.
the Galactic center. Hills (1988) first pointed out that a
0 We havedesigneda successful targetedsurvey for new
stellarbinaryencounteringtheMilkyWay’scentralMBH
7 HVSs. We use the 6.5m MMT and the Whipple 1.5m
can eject one member of the binary as a HVS traveling
0 at∼1,000kms−1. HVSs differ fromclassical“runaway” Tillinghast telescopes to obtain radial velocities of faint
h/ stars because 1) HVSs are unbound and 2) the classical B-type stars, stars with lifetimes consistent with travel
p supernovaejection(Blaauw1961)anddynamicalejection timesfromtheGalacticcenterbutwhichshouldnototh-
- (Poveda et al. 1967) mechanisms that explain runaway erwiseexistinthedistanthalo. FourearlierHVSdiscov-
o eries from this survey are published elsewhere (Brown
starscannotproduceejectionvelocitieslargerthan200-
r
t 300kms−1 for B-type stars(Leonard1991,1993;Porte- et al. 2006a,b). Here, we present spectroscopic observa-
s tions of 643 new HVS candidates.
a gies Zwart 2000; Davies et al. 2002; Gualandris et al.
Ourpaperisorganizedasfollows. In§2wediscussour
: 2004; Dray et al. 2005). The first HVS discovered, by
v target selection and spectroscopic identifications. In §3
comparison, has a heliocentric radial velocity of +853
Xi km s−1 and a Galactic rest-frame velocity of at least we present evidence for stars ejected from the Galactic
+709±12kms−1 (Brownetal.2005a),manytimesthat centeronboundorbits. In§4weshowthatthe unbound
r
HVSs have a marginally anisotropic distribution on the
a needed to escape the Milky Way. Photometric follow-up
sky. We conclude in §5. Our new observations are listed
shows that the object is a slowly pulsating B main se-
in Appendix A.
quence star (Fuentes et al. 2006). Only interaction with
aMBHcanplausiblyacceleratethis3M⊙ mainsequence 2. DATA
B star to such an extreme velocity and distance (∼110
2.1. Target Selection
kpc).
HVSs are fascinating because they can be used to un- Brown et al. (2006b) describe our target selection for
derstand the nature and environs of MBHs (see the re- candidate HVSs. Briefly, we use Sloan Digital Sky Sur-
cent theoretical work of Gualandris et al. 2005; Holley- vey (SDSS) photometry to select late B-type stars by
Bockelmann et al. 2005; Ginsburg & Loeb 2006a; Levin color. Figure 1 presents the HVS survey color-color se-
2006; Perets et al. 2006; Baumgardt et al. 2006; Sesana lection. We use the 6.5m MMT telescope to observe ev-
etal.2006;Bromleyetal.2006;Merritt2006;Demarque erystarwith17<g0′ <19.5inthiscolorrange. Wehave
&Virani2007;Ginsburg&Loeb2006b;O’Leary&Loeb observed 136 new HVS candidates and have completed
2006; Gualandris & Portegies Zwart 2007; Kollmeier the high declination region of the SDSS Data Release 4
& Gould 2007). The trajectories of HVSs also pro- (DR4, Adelman-McCarthy et al. 2006). The MMT sur-
vide unique probes of the shape and orientation of the vey is now 80% complete across SDSS DR4 and covers
Galaxy’s dark matter halo (Gnedin et al.2005). Discov- an area of ∼5,000 deg2 or 12% of the entire sky.
eries of additional HVSs (Edelmann et al. 2005; Hirsch We have also carried out a new, complementary HVS
etal.2005;Brownetal.2006a,b)arestartingto provide survey targeting bright stars with 15 < g0′ < 17 and
suggestive limits on the stellar mass function of HVSs, B-type colors in the SDSS. We use the 1.5m Tilling-
hast telescope and the FAST spectrograph (Fabricant
Electronicaddress: [email protected] et al. 1998) to observe these bright HVS candidates.
2 Brown et al.
Fig. 2.— Spectrum of the z = 2.43 quasar located at
15h58m51s.85,+22d21m59s.9(J2000).
and a 2′′ slit. These settings provided a wavelengthcov-
erageof3500˚Ato5400˚Aandaspectralresolutionof2.3
˚A. Like the MMT observations, exposure times ranged
from5to30minutesandwerechosentoyieldS/N =15
inthecontinuumat4000˚A.Comparisonlampexposures
were obtained after every exposure.
Spectra were extracted using IRAF1 in the stan-
Fig. 1.— Color-color diagram showing the target selection re-
dard way. Radial velocities were measured using
gions for the MMT (dashed line) and FAST (solid line) samples,
and the spectroscopic identifications for the combined set of 894 the cross-correlation package RVSAO (Kurtz & Mink
HVS candidates. The brighter FAST sample (15 <g0′ < 17) tar- 1998). Brown et al. (2003) describe in detail the cross-
getsbluerstarsandfindsanumberofstarswithB3andB4spectral correlationtemplatesweuse. Theaverageuncertaintyis
types(solid dots).
±11 km s−1 for the B-type stars.
2.3. Selection Efficiency and Completeness
We are able to survey efficiently much bluer stars with
Our combined observations total 894 HVS candidates
FAST because contamination from white dwarfs is less with 15 < g′ < 19.5, of which 699 (78%) are stars of
0
problematic at bright magnitudes. Figure 1 shows the
B spectral type. The remaining objects include 185 DA
FAST color-color selection: 0.1 < (u′ −g′) < 1.5 and
0 white dwarfs, 3 other (DZ, DB) white dwarfs, 1 com-
(10.67(g′−r′) +3.5)<(u′−g′) <(10.67(g′−r′) +5.07).
0 0 0 pact blue galaxy (see Kewley et al. 2006), 1 quasar, and
This region follows the stellar sequence of B-type stars
5 objects of uncertain identification, possibly broad ab-
in the SDSS photometric system (Fukugita et al. 1996).
sorptionline quasars. The oneobviousquasarisaprevi-
Following Brown et al. (2006a), we impose a color cut ously unidentified g′ =19.67 object at z =2.43, located
−0.5 < (r′ − i′)0 < 0 to reject objects with non- at15h58m51s.85,+22d21m59s.9(J2000). Figure2presents
stellar colors. We also exclude the regionofsky between its spectrum. The quasar has colors (u′−g′) = +0.80
b < −l/5+ 50◦ and b > l/5− 50◦ to avoid excessive and (g′ −r′) = −0.28, unusually red in (u′ −0 g′) and
0 0
contamination from Galactic bulge stars. There are 746 blue in (g′ −r′) compared to other quasars at similar
SDSS DR4 candidate B stars with 15 < g′ < 17 in the 0
0 redshifts (Schneider et al. 2005).
selection region. We have observed 514 stars, 69% of
Figure 1 plots the spectroscopic identifications of the
this total. The averagesurface number density of bright
HVSs, other B-type stars, and the white dwarfs. Inter-
HVS candidates is 0.13 deg−2. Thus we have surveyed
estingly, we find stars with spectral types as early as B3
aneffective areaof ∼4000deg2 for HVS candidates with
inthebrightFASTsample: thesoliddotsinFigure1are
15<g′ <17 with FAST.
0 the61starswithspectraltypesB3andB4. Iftheseearly
B-typestarsaremainsequencestars,theyarelocatedat
2.2. Spectroscopic Observations and Radial Velocities surprisinglylargeheliocentricdistances30.d.60kpc.
Newobservationsatthe6.5mMMTtelescopewereob- We discuss the early B-type stars further in §3.
tainedwiththeBlueChannelspectrographonthenights Our survey is now 100% complete across the high
of 2006 May 23-26 and 2006 June 19-20. The spectro- declination-region of SDSS DR4 located at 25 < l <
graph was operated with the 832 line mm−1 grating in 210◦, b > 0◦ in Galactic coordinates. Curiously, we find
second order and a 1.25′′ slit. These settings provided no new unbound HVSs in the completed region. Our
a wavelength coverage of 3650 ˚A to 4500 ˚A and a spec- previous discovery rate with the MMT was 1 HVS per
tral resolution of 1.2 ˚A. Exposure times ranged from 5 ∼50 B-type stars (Brown et al. 2006b), yet our new set
to 30 minutes and were chosen to yield S/N =15 in the of observations contains 245 (57 from the MMT) B-type
continuum at 4000 ˚A. Comparison lamp exposures were starsinthecompleteregion. Theabsenceofanunbound
obtained after every exposure. HVS in this large region suggests that the distribution
Observations at the 1.5m Tillinghast telescope were
obtainedwith the FAST spectrographoverthe courseof 1 IRAF is distributed by the National Optical Astronomy Ob-
servatories, which are operated by the Association of Universities
19nightsbetween2006January1and2006July21. The
forResearchinAstronomy,Inc.,undercooperativeagreementwith
spectrographwasoperatedwitha600linemm−1 grating theNationalScienceFoundation.
Hypervelocity Stars II. The Bound Population 3
of HVSs on the sky may have structure. We address the
spatial distribution of HVSs in §4.
For the remainderof this paper, we considera “clean”
sample containing 560 stars selected only from the over-
lapping region of color-colorspace: −0.38<(g′−r′) <
0
−0.28 and 2.67(g′−r′) +1.30<(u′−g′) < 2.67(g′−
0 0
r′) +2.0. Main sequence B stars in this selection re-
0
gion have masses of 3 - 5 M⊙. We also consider the full
sample of 699 B-type stars at all colors. The full sample
contains earlier B-type objects from the FAST sample
possibly including main sequence stars with masses up
to 7 M⊙. We emphasize that our photometric selection
is complete andcandetect a B-type staratany velocity.
3. BOUNDHYPERVELOCITYSTARS
The observed distribution of distances and velocities
of our B-type stars provides ample evidence that they
are not main sequence run-aways. Curiously, the veloc-
ity distribution revealsan asymmetryofstars with large
positive radial velocities. Because the large positive ve-
locityoutliersareveryunlikelymainsequencerun-aways,
we argue that the stars represent a class of bound HVSs
ejected from the Galactic center. Fig. 3.— Galactic rest-frame velocity histogram of the clean
sample of 560 B-type stars (upper panel). The best-fit Gaussian
3.1. Background (thinline)hasdispersion105±5kms−1. Thelowerpanelplotsthe
residualsoftheobservationsfromthebest-fitGaussian,normalized
HighGalacticlatitude B-type starswerefirstreported
by the value of the Gaussian. In addition to the unbound HVSs,
byHumason&Zwicky(1947). Earlyspectroscopicstud- there is a significant asymmetry of 11 positive velocity outliers
ies (Feige 1958; Berger 1963; Greenstein 1966) showed +275<vrf <450kms−1 includingSDSSJ144955.58+310351.4.
that high latitude B-type stars are a mix of Popula-
tion II post-main sequence stars and Population I main
sequence run-aways. Accurate space motions and dis-
that produces a type Ia supernova. In such systems,
tancesarenowknownfornearbyd<10kpcB-typestars
observed by Hipparcos. A detailed analysis by Martin Davies et al. find that 3 M⊙ secondaries have typical
ejection velocities ∼100 km s−1 and maximum ejection
(2004)shows that 2/3ofthe high latitude B stars in the
velocities of 250 km s−1. By comparison, the escape ve-
Hipparcos catalog are main sequence run-aways and 1/3
locity of the Galaxy near the Sun is at least 500 km s−1
are evolved (mostly blue horizontal branch) halo stars.
(Carney et al. 1988).
Martin (2004) finds that all of the main sequence run-
Because main sequence run-awaystravelon bound or-
awayBstarsintheHipparcos cataloghaveorbitsconsis-
bits,theyspendmostoftheirtimeneartheapexoftheir
tent with a disk origin.
trajectoriesabovethediskwithsmallline-of-sightveloci-
Our HVS survey probes much fainter and more dis-
ties. Forexample,astarejectedat240kms−1 vertically
tant B-type stars than observed by Hipparcos. Thus
we cannot rely on proper motions or parallaxes. Brown out of an infinite disk with surface mass density 95 M⊙
pc−2 reaches an apex of z = 11 kpc. This simple calcu-
et al. (2006b) discuss the nature of the late B-type stars
lation illustrates that main sequence run-aways cannot
and conclude that the distribution of radial velocities
simultaneously have large velocities v > +275 km s−1
and metallicities suggests that they are likely a Galac-
and large distances z > 10 kpc. Davies et al. (2002)
tic halo population of post-main sequence stars and/or
use a full Galaxy potential model and compute orbits
blue stragglers. We now re-visit the possibility of main
for1000run-awaystarsrandomlyejectedfromtheirpre-
sequence run-aways in our sample.
dicted velocity distribution. The vast majority of the
Theorists have shown that unbinding a stellar binary
Davies et al. (2002) run-aways are located at |z| < 5
through supernova disruption (Blaauw 1961) or binary-
kpc; at high Galactic latitudes |b| > 30◦ the most dis-
binaryinteractions(Povedaetal.1967)producesamaxi-
mumejectionvelocityof200-300kms−1(Leonard1991, tant run-aways are found ∼25 kpc from the disk.
1993; Portegies Zwart 2000; Davies et al. 2002; Gualan-
3.2. The Asymmetric Velocity Distribution
drisetal.2004;Drayetal.2005). Themaximumvelocity
is setbythe escape velocityfromthe stellarsurface. For Figure 3 plots the distribution of line-of-sight veloci-
example,acontactbinarycontainingtwo3M⊙ starshas ties, corrected to the Galactic rest-frame, for our clean
a Keplerian orbital velocity of 240 km s−1. Extracting sample of 560late B-type stars plus HVS1 (Brownet al.
a large ejection velocity from disrupting such a system 2005a). It is apparent from Figure 3 that the velocities
is difficult. Portegies Zwart (2000) considers supernovae are well described by a Gaussian distribution. We itera-
unbinding binaries and finds that 90% of ejections have tivelyclip3σ outliersandcalculatea105±5kms−1 dis-
velocities between 0 and 100 km s−1; only 1% of 3 M⊙ persionanda +14±4kms−1 meanforthe distribution.
starsareejectedat200kms−1. Daviesetal.(2002)per- We note that the full sample of 699 stars has a statis-
formsimilarcalculationsforbinariesthatundergoacom- tically identical distribution. The lower panel of Figure
mon envelope phase followed by a mass-transfer phase 3 plots the residuals of the observations from this Gaus-
4 Brown et al.
sian distribution, normalized by the value of the Gaus-
sian. Stars with velocities |v |<275 km s−1 show low-
rf
significance deviations from a Gaussian distribution and
probably constitute a halo population of post-main se-
quencestarsand/orbluestragglers(Brownetal.2006b).
Thehighest-significantoutliersarethefirstHVS(HVS1),
thefourHVSspreviouslydiscoveredinthissurvey(HVS4
-HVS7),andanewobject: SDSSJ144955.58+310351.4.
The possible HVS, SDSS J144955.58+310351.4, is a
bright g′ = 15.70±0.03 star with an A1±1.4 spectral
0
type and solar metallicity [Fe/H] = −0.3 ± 0.6. A
Wk
main sequence star of this type has heliocentric distance
d ≃ 17 kpc (Schaller et al. 1992); an evolved blue hori- Fig. 4.— Predicted velocity distributions of 3-5 M⊙ HVSs
zontal branch (BHB) star has d ≃ 7 kpc (Clewley et al. withdistance fromtheGalactic center (Bromleyetal.2006). We
2005). Located at (l,b)=(48.7◦,63.9◦), the star is high calculate distances for our B-type stars assuming they are main
sequence stars: the four HVSs discovered by this survey (solid
above the Galactic plane at z ≃ 15 kpc and a Galacto-
centric distance of R ≃ 17 kpc (assuming d = 17 kpc). skqmuasr−e1s)(oapnedntshqeua1r1esp)oinsictliuvdeinvgelSoDciStySoJu14tl4ie9r5s5.+582+7531<03v5r1f.4<(op4e5n0
The star’s +363±10 km s−1 heliocentric radial velocity squarewithplussign).
corresponds to a minimum velocity in the Galactic rest
frameof+447kms−1 (seeBrownetal.2006b). Thestar
is listed with no proper motion in the USNOB1 (Monet
et al. 2003) catalog, and thus is probably bound to the tainlycome fromthe GalacticCenter. The presenceofa
Milky Way. 3.6×106 M⊙ MBH (Ghez et al. 2005; Eisenhauer et al.
Curiously, we find a set of more distant stars with 2005)inthecrowdedGalacticcenterinevitablyproduces
equally large, but bound, velocities. Figure 3 shows a a“fountain”ofHVSsejections. TheactualHVSejection
significant asymmetry in the wings of the observed ve- rate depends on the number of stars with radial orbits
locity distribution: 11 stars (excluding the 5 unbound in the MBH’s so-called “loss cone.” If the MBH’s loss
HVSs) have large positive velocities +275<vrf <+450 cone is replenished with stars scattered in by massive
km s−1; only 1 star has a comparable negative velocity star clusters, molecular clouds, or IMBHs (Perets et al.
at v = −286±12 km s−1. This asymmetry was first 2006),then the ejection rate can be ordersof magnitude
rf
pointed out by Brown et al. (2006b) but is more signifi- larger than the original Yu & Tremaine (2003) predic-
cantinthislargersample. Thereisa0.00022probability tions. Regardless of the exact rate, this fountain picture
of drawing 11 stars with +275< v < 450 km s−1 and results in a broad spectrum of HVS ejection velocities
rf
1 star with −450<v <−275kms−1 froma Gaussian whichincludeboundandunboundorbits(Hills1991;Yu
rf
distributions with the observed parameters. Integrating & Tremaine 2003; Gualandris et al. 2005; Levin 2006;
the wings of a Gaussian with the observed parameters, Baumgardt et al. 2006; Perets et al. 2006; Sesana et al.
wewouldexpecttofind3.6starswith+275<v <450 2006;Bromleyetal.2006;O’Leary&Loeb2006). Brom-
rf
km s−1 and 1.6 stars −450<v <−275 km s−1. Thus leyetal.(2006)recentlycalculatedejectionvelocitiesfor
rf
there is an excess of ∼7 stars with large positive veloc- binariescontaining3-4M⊙primariesmatchedtooursur-
ity and no apparent excess of stars with large negative vey of B-type stars.
velocity, significant at the ∼3.5σ level. Overthevolumesampledbyoursurvey,Bromleyetal.
(2006)predictcomparablenumbersofHVSsejectedinto
Ifthepositivevelocityoutliersaremainsequencestars,
they are located at large distances 20 - 80 kpc (see Ta- bound and unbound orbits. Our survey has discovered
ble 1). Yet the outliers cannot be run-away stars, be- 4 HVSs on unbound orbits suggesting that 4±2 of our
cause the run-away mechanisms cannot produce simul- excess ∼7 positive velocity outliers are plausibly HVSs
taneously large velocities and large distances. If, on the on bound orbits. Although the predicted and observed
other hand, the positive velocity outliers are halo stars numbers of bound HVSs are statistically consistent, we
on radial orbits, we would expect to find equal numbers note that additional HVS mechanisms may be at work.
For example, O’Leary & Loeb (2006) show that single
of stars moving toward and away from us, contrary to
observation. Compact binary systems may also produce starscanbeejectedbyencounterswithstellarmassblack
outliersinthevelocitydistribution. Butvelocityoutliers holesorbitingthecentralMBH.SuchHVSstendtohave
resulting from compact binaries should be distributed lower ejection velocities than HVSs ejected by the Hills
symmetrically, again contrary to observation. We note (1988) mechanism, and thus may account for additional
that neutron stars are observedtraveling at velocities of HVSs on bound orbits.
1000 km s−1, but such neutron stars are the remnants Figure 4 plots the Bromley et al. (2006) predicted ve-
locity distributions for main sequence HVSs ejected by
of asymmetric supernova explosions, and are not B-type
stars. the central MBH as a function of Galactocentric radius.
HVSvelocitiesincreasewithdepthbecauseonlythemost
rapidly movingHVSs survive long enoughto reachlarge
3.3. Bound HVSs
Galactocentric distances. The large squares in Figure
One explanation for the significant excess of B-type 4 are our 4 HVSs (solid squares) and the 11 stars with
stars traveling +275 < v < 450 km s−1 is the HVS +275 < v < 450 km s−1 (open squares). We assume
rf rf
mechanism. HVSs are ejected by three-body interac- that the stars are main-sequence stars, and use the ob-
tions involving a MBH (Hills 1988). HVSs almost cer- served colors and spectral types of the stars to estimate
Hypervelocity Stars II. The Bound Population 5
stellar luminosity, distance, and stellar mass (see Table
1). The solid lines in Figure 4 indicate our survey limits TABLE 1
POSSIBLEBOUND HVSs
for main sequence stars of a given stellar mass.
We evaluate the probability of drawing our observed
sample from the Bromley et al. (2006) predicted dis- CatalogID delg dbeg kmvrsf−1 type kRpac kRpbc
tributions. We use the two-dimensional, two-sample
Kolmogorov-Smirnovtest (Presset al.1992)to compare SDSSJ074950.24+243841.2 196.1 23.2 +293 B8 66 27
the observed and predicted distributions of HVS veloc- SDSSJ075055.24+472822.9 171.5 29.4 +307 B6 42 17
SDSSJ075712.93+512938.0 167.0 30.9 +329 B7 32 13
ity and distance. First, we generate 105 realizations of
SDSSJ081828.07+570922.1 160.4 34.2 +283 B9 42 24
the observations–4 HVSs plus 7boundHVSs randomly SDSSJ090710.08+365957.5 186.3 42.2 +280 B9 57 29
drawn from the 11 high velocity outliers – and calculate SDSSJ110224.37+025002.8 251.2 54.4 +326 A1 51 23
SDSSJ115245.91-021116.2 274.9 57.5 +305 A1 53 18
themaximumdifferenceinintegratedprobabilityagainst
SDSSJ140432.38+352258.4 65.3 72.4 +293 B8 51 18
the Bromley et al. (2006)distributions. We use only the SDSSJ141723.34+101245.7 357.2 63.6 +289 A1 48 19
portion of the Bromley et al. (2006) distributions that SDSSJ142001.94+124404.8 2.5 64.8 +300 B9 27 12
fall within our observational survey limits (see Figure SDSSJ144955.58+310351.4 48.7 63.9 +447 A1 16 9
4). Second, we randomly sample 11 objects from the
Bromley et al. (2006) distributions 105 times and calcu- aGalactocentric distance for a main sequence star of the observed
late the maximum difference in integrated probabilities spectraltype.
b
GalactocentricdistanceforaBHBstaroftheobservedcolor.
against the Bromley et al. (2006) distributions. Again,
we use only the portionof the Bromley et al. (2006)dis-
tributions that fall within our observational survey lim-
its. Finally,thelikelihoodisthefractionoftimethatthe starswiththeirtraveltimesfromtheGalacticcenter? It
probabilities of the synthetic data sets exceed the prob- ispossiblethattheearlyB-typestarsarebluestragglers,
abilities of the real data sets. The mean likelihood of main sequence stars that have undergone mass transfer
drawingourobservedsetsofstarsfromthepredicteddis- or mergers. Blue stragglers are usually associated with
tributionsisapproximately0.128,whichneitherstrongly globular clusters, yet surveys of field BHB stars in the
supports nor strongly rejects the Bromley et al. (2006) halo find that half of the stars are in fact high surface-
model. Weconcludethatthemostplausibleexplanation gravity blue stragglers or A dwarfs (Norris & Hawkins
forthe observedexcessofpositive velocityoutliersis the 1991; Preston et al. 1994; Wilhelm et al. 1999; Brown
HVS mechanism ejecting stars from the Galactic center et al. 2003; Clewley et al. 2004; Brown et al. 2005b).
on bound orbits. Because HVSs likely originate in tight binary systems,
We note that other stars in the literature may also be it is possible that some HVSs experience mass-transfer
explained as bound HVSs. For example, the 5 M⊙ main fromtheir former companionpriorto being ejectedfrom
sequence B star HIP 60350has a heliocentric velocity of the Galactic center. A blue straggleroriginis one expla-
+230 km s−1 and a full space velocity of 417 km s−1 nation for HVS2, an 8 M⊙ HVS with a main sequence
basedonHipparcos measurements (Maitzen etal. 1998). lifetime otherwise inconsistent with traveltime from the
The star’s large radial motion U =−352 km s−1 means Galactic center (Edelmann et al. 2005).
that it originated from well inside the solar circle, per- It is also possible that the early B-type stars are main
haps from the Galactic center. sequencestarsejectedelsewhereintheGalaxybydynam-
ical interactions with compact objects like intermediate
3.4. Mystery of the Early B-type Stars
mass black holes (IMBHs). However, as the mass of a
Our full sample of B-type stars contains many early- blackholedecreases,thecross-sectionofHVSinteraction
type stars at bright magnitudes. Four of the early B decreases (e.g. Hills 1988). A stellar binary must come
stars in the full sample are possibly bound HVSs with much closer to an IMBH before the gravitational tidal
+275 < v < 400 km s−1. If they are main sequence energy exceeds the binding energy of the binary. Thus
rf
stars, the four early B stars have masses of 5 - 7 M⊙ anIMBHproducesfewerHVSsthanaMBHinthesame
and they are located at distances ranging 30 - 60 kpc environment. UnlessIMBHsareubiquitousindensestel-
(boundedbytheFASTsurveylimitingmagnitudes). We larenvironmentssuchasthecentersofglobularclusters,
now ask whether the number of early and late B-type weexpectthatthe overallHVSrateisdominatedby the
bound HVSs are consistent with a standard initial mass MBH in the Galactic center.
function. Assuming the stars are main sequence stars, A final possibility is that the early B-type stars are
there are nine 3 - 5 M⊙ stars in our sample located in post main-sequence stars. Demarque & Virani (2007)
the same volume as the four 5 - 7 M⊙ stars: a ratio of argue that the Galactic center contains many more old
2.25. A Salpeter initial mass function predicts a ratio of dwarf stars than young B stars, and thus the MBH
2.7, similar to what we observe. However, the lifetimes should eject many more BHB stars than main sequence
of 5 - 7 M⊙ main sequence stars are only 40 - 90 Myr B stars. Davies & King (2005) and Dray et al. (2006)
(Schaller et al. 1992), a factor of 2 - 3 shorter than 3 - argue that the Galactic center contains tidally-stripped
5 M⊙ stars. The travel times of the four early B stars, evolved stars (in effect, BHB stars with small hydrogen
assumingv istheirfullspacevelocity,rangefrom∼100 envelopes) that mimic B stars. B-type BHB stars with
rf
to ∼150Myr from the Galactic center. Thus if the early hoteffective temperaturesandsmallhydrogenenvelopes
BstarsaremainsequencestarsejectedfromtheGalactic spend most of their helium-burning lifetimes at intrinsi-
center,theydonothavehighenoughvelocitiestosurvive callyfaintluminosities (Yi etal.1997;Drayet al.2006).
to their inferred distances. Thus, if they are BHB stars, the early B-type stars are
How can we reconcile the lifetimes of the early B-type located 10 - 20 kpc from the Galactic center and have
6 Brown et al.
Fig. 5.— Aitoff sky map, in Galactic coordinates, showing the clean sample of 560 B-type HVS candidates. Radial velocities, in the
Galactic rest frame, are indicated by the color of the solid squares: purple is -300, green is 0, and red is +300 km s−1. The HVSs are
completelyoffthiscolorscaleandaremarkedby: HVS1(plus),thefourHVSsdiscoveredbythissurvey(solidsquares),andthe11positive
velocity outliers +275 < vrf < 400 km s−1 (open squares) including SDSS J144955.58+310351.4 (open square with plus sign). Dotted
linesarelinesofconstant unboundHVSfractioncalculated fromtheBromleyetal.(2006) models.
travel times ranging from 25 - 50 Myr. The BHB pro- tic rest frame, of the stars. The four unbound HVSs
genitor lifetimes easily exceed the inferred travel times, (solid black squares) discovered by this survey are com-
reconcilinganydifficultieswiththe MBHejectionorigin. pletely off the color scale; HVS1 is marked by a plus
Totestthepost-mainsequenceexplanationrequireshigh sign. The11possibleboundHVSsareindicatedbyopen
dispersion spectroscopy to determine the stellar nature squares; SDSS J144955.58+310351.4 is plotted with an
of the bound HVSs. open square with a plus sign. The dotted lines in Fig-
ure 5 are lines of constant HVS fraction calculated from
4. HVSSPATIALANISOTROPY the Bromleyet al.(2006)models. We note that our sur-
veyiscurrentlyincompleteintheSDSSequatorialslices,
The distribution of HVSs on the sky is interesting located at l >210◦ and b<0◦.
because the spatial distribution of HVSs is linked to
It is striking that all of the unbound HVSs in our sur-
their origin. Bromley et al. (2006) show that fraction
vey are located towards the anti-center. We randomize
ofunboundHVSsincreaseswithradialdistancefromthe
the observed velocities among the positions of all 560
Galacticcenter(seeFigure4). Thefractionchangeswith
stars, and find a 0.064 probability of randomly draw-
depth because only the most rapidly moving HVSs sur-
ing the 4 unbound HVSs from our survey in the hemi-
vivetoreachlargeGalactocentricdistances. Becausethe sphere 90 < l < 270◦. Thus the unbound HVSs have a
Sunislocated8kpcfromtheGalacticcenter,oursurvey
marginallyanisotropic spatialdistribution, being prefer-
reaches approximately 16 kpc deeper towards the anti-
entially located in the anti-center hemisphere with ∼2σ
center than towards the Galactic center. Thus, if the
confidence. We note that the 3 unbound HVSs not from
fountainmodelofHVSsejectedfromtheGalacticcenter
this survey, HVS1 (Brown et al. 2005a), HVS2 (Edel-
iscorrect,weexpecttofindmoreunboundHVSstowards
mann et al. 2005), and HVS3 (Hirsch et al. 2005), are
the anti-center. alsolocatedintheanti-centerhemisphere90<l<270◦.
We use the Bromleyet al.(2006)models (Figure 4)to
The 11 possible bound HVSs, on the other hand, ap-
calculate the expected fraction of unbound HVSs in our
pear more evenly distributed across the sky: 6 bound
survey volume as a function of position on the sky. The HVSs are located 90 < l < 270◦ and 5 bound HVSs
predicted fraction of unbound HVSs ranges from 53% are located |l| < 90◦. To calculate the variation in un-
towards the Galactic center to 63% towards the anti-
bound HVS fraction, we consider only 7 bound HVSs
center. The fraction does not vary much because it is
(the excess number of high velocity outliers in Figure 3)
dominatedby the surveyvolume atlargedistances. The
andsplitthem4/3betweentheanti-centerandGalactic-
HVSs found in this survey, however, are not located at
center hemispheres, respectively. The fraction of un-
distances beyond ∼80 kpc. If we limit the survey vol-
bound HVSs is then 50±30% in the anti-center hemi-
ume to the observed range of HVS distances, then the
sphere and 0±40% in the Galactic-center hemisphere.
predictedfractionof unbound HVSs rangesfrom31%to
Although small number statistics overwhelm our mea-
46%. We now consider the observations.
surement, the unbound HVS fraction appears consistent
Figure 5 plots the distribution of our clean sample
with the predictions of Bromley et al. (2006) and the
of 560 B-type stars across the sky in Galactic coordi-
Galactic center ejection picture.
nates. Color indicates the radial velocity, in the Galac-
Hypervelocity Stars II. The Bound Population 7
There is another possible explanation for spatial Stellar rotation is a useful discriminant between rapidly
anisotropy: HVSs ejected via single star encounters rotating main sequence B stars (Abt et al. 2002; Mar-
with a MBH binary are preferentially ejected in the tin 2004)and slowly rotating BHB stars (Petersonet al.
orbital plane of the MBH binary (Gualandris et al. 1995;Behr2003). Metallicityisalsoagooddiscriminant
2005;Holley-Bockelmannetal.2005;Levin2006;Sesana betweenyoung mainsequence B stars andpost-main se-
et al. 2006; Merritt 2006; Baumgardt et al. 2006). Re- quence halo stars. Echelle observations of the brightest
cent N-body simulations for the case of an IMBH inspi- HVSs, underway now, will reveal the stars’ true nature.
rallingintoaMBH,however,showthatHVSsareejected A Galactic center origin predicts that unbound HVSs
nearlyisotropicallybecause the angularmomentum vec- shouldbe foundpreferentiallytowardsthe anti-centeras
tor of the IMBH rapidly changes on a ∼1 Myr timescale aresultofsurveyselection. Weobservethattheunbound
(Baumgardt et al. 2006). Moreover, our observed set of HVSs have a marginally anisotropic distribution at 2-
HVSsdonotsharecommontraveltimesfromtheGalac- σ confidence. Our estimates of unbound HVS fraction
tic center. Different IMBH inspiral events would likely are consistent with the Galactic center ejection picture.
havedifferentorbitalhistories,thusitseemsunlikelythat Measuring the spatial distribution of a larger sample of
theobservedHVSspatialdistributioncanpointbacktoa HVSs may provide a strong test of the stars’ origin.
binary MBH origin. A set of HVSs with common travel We note that full space motions for the HVSs will be
times must be found in order to test the binary MBH knowninafewyears. AHubbleSpaceTelescopeprogram
origin. is measuring the stars’ positions with 0.5 mas accura-
cies by registering backgroundgalaxies across Advanced
5. CONCLUSIONS Camera for Surveys images. The HVSs have predicted
We discuss our targeted survey for HVSs, a spectro- proper motions of 0.5 - 1 mas yr−1, thus 3σ proper mo-
scopic survey of stars with B-type colors. We introduce tionmeasurementsarepossibleoverathreeyearbaseline
anewcomponentofthesurveythatsamplesbrighterand (Gnedin et al. 2005).
bluerstarsusingthe FAST spectrograph. Ourcombined We are continuing our targeted HVS survey of faint
set of observations now contains 699 stars of B spectral B-type stars. SDSS Data Release 5 provides additional
type, 188 white dwarfs, and 1 previously unidentified targetsthatwearenowfollowing-upwiththe MMTand
z = 2.43 quasar located at 15h58m51s.85,+22d21m59s.9 Whipple 1.5m telescopes. Given our current discovery
(J2000). rate, we expect to find another few unbound HVSs in
We find no new unbound HVS, but we find evidence the coming months. Identifying HVSs throughout the
foranew classofboundHVSs. The velocitydistribution Galaxy will allow us to measure the mass function of
ofthe B-typestarsshowsa significantexcessof∼7stars stars in the Galactic center and the history of stellar
with large positive velocities +275 < v < 450 km s−1 interactions with the central MBH.
rf
andnoapparentexcessofstarswithlargenegativeveloc-
ity, significant at the ∼3.5σ level. Neither halo stars on
radial orbits, compact binary stars, nor main sequence
run-away stars ejected from the disk can explain simul- We thank M. Alegria, J. McAfee, and A. Milone for
taneouslythelargevelocitiesanddistancesofthesestars. their assistance with observations obtained at the MMT
The most plausible explanation is that the positive ve- Observatory, a joint facility of the Smithsonian Insti-
locityoutliersareHVSs ejectedfromthe Galacticcenter tution and the University of Arizona. We thank H.
on bound orbits. Perets for helpful correspondence. This project makes
Our bright FAST survey contains a large number of use of data products from the Sloan Digital Sky Sur-
early B-type stars, four of which have velocities +275< vey, which is managed by the Astrophysical Research
v <450kms−1. ThesefourpossibleboundHVSshave Consortium for the Participating Institutions. This re-
rf
main sequence lifetimes too short to survive to their in- search has made use of NASA’s Astrophysics Data Sys-
ferred distances, however. One explanation is that the temBibliographicServices. This workwassupportedby
stars are post-main sequence stars. Establishing HVSs W. Brown’s Clay Fellowship and the Smithsonian Insti-
as main sequence or post-main sequence stars is impor- tution.
tant for measuring the stellar mass function of HVSs Facilities: MMT (Blue Channel Spectrograph),
and probingthe types of stars that orbit near the MBH. FLWO:1.5m (FAST Spectrograph)
REFERENCES
Abt,H.A.,Levato, H.,&Grosso,M.2002,ApJ,573,359 Brown,W.R.,Geller,M.J.,Kenyon, S.J.,Kurtz,M.J.,Allende
Adelman-McCarthy,J.K.etal.2006,ApJS,162,38 Prieto,C.,Beers,T.C.,&Wilhelm,R.2005b, AJ,130,1097
Baumgardt, H., Gualandris, A., & Portegies Zwart, S. 2006, Carney,B.W.,Laird,J.B.,&Latham,D.W.1988,AJ,96,560
MNRAS,372,174 Clewley, L., Warren, S. J., Hewett, P. C., Norris,J. E., & Evans,
Behr,B.B.2003,ApJS,149,67 N.W.2004,MNRAS,352,285
Berger,J.1963, PASP,75,393 Clewley,L.,Warren,S.J.,Hewett,P.C.,Norris,J.E.,Wilkinson,
Blaauw,A.1961, Bull.Astron.Inst.Netherlands,15,265 M.I.,&Evans,N.W.2005,MNRAS,362,349
Bromley, B. C., Kenyon, S. J., Geller, M. J., Barcikowski, E., Davies,M.B.&King,A.2005,ApJ,624,L25
Brown,W.R.,&Kurtz,M.J.2006,ApJ,653,1194 Davies,M.B.,King,A.,&Ritter,H.2002,MNRAS,333,463
Brown,W.R.,AllendePrieto,C.,Beers,T.C.,Wilhelm,R.,Geller, Demarque,P.&Virani,S.2007,A&A,461,651
M.J.,Kenyon,S.J.,&Kurtz,M.J.2003,AJ,126,1362 Dray, L. M., Dale, J. E., Beer, M. E., Napiwotzki, R., & King,
Brown,W.R.,Geller,M.J.,Kenyon,S.J.,&Kurtz,M.J.2005a, A.R.2005,MNRAS,364,59
ApJ,622,L33 Dray,L.M.,King,A.R.,&Davies,M.B.2006, MNRAS,910
—.2006a,ApJ,640,L35 Edelmann, H., Napiwotzki, R., Heber, U., Christlieb, N., &
—.2006b,ApJ,647,303 Reimers,D.2005,ApJ,634,L181
8 Brown et al.
Eisenhauer,F.etal.2005,ApJ,628,246 Leonard, P. J. T. 1993, in ASP Conf. Ser. 45, Luminous High-
Fabricant,D.,Cheimets,P.,Caldwell,N.,&Geary,J.1998,PASP, LatitudeStars,ed.D.D.Sasselov,360
110,79 Levin,Y.2006,ApJ,653,1203
Feige,J.1958,ApJ,128,267 Maitzen, H. M., Paunzen, E., Pressberger, R., Slettebak, A., &
Fuentes, C. I., Stanek, K. Z., Gaudi, B. S., McLeod, B. A., Wagner, R.M.1998, A&A,339,782
Bogdanov, S.,Hartman, J.D.,Hickox, R.C.,& Holman,M.J. Martin,J.C.2004,AJ,128,2474
2006, ApJ,636,L37 Merritt,D.2006,ApJ,648,976
Fukugita, M.,Ichikawa, T., Gunn, J. E.,Doi, M.,Shimasaku, K., Monet,D.G.etal.2003, AJ,125,984
&Schneider,D.P.1996, AJ,111,1748 Norris,J.E.&Hawkins,M.R.S.1991, ApJ,380,104
Ghez, A. M., Salim, S., Hornstein, S. D., Tanner, A., Lu, J. R., O’Connell,W.O.R.1973,AJ,78,1074
Morris,M.,Becklin,E.E.,&Duchene, G.2005,ApJ,620,744 O’Leary,R.M.&Loeb,A.2006,preprintastro-ph/0609046
Ginsburg,I.&Loeb,A.2006a, MNRAS,368,221 Perets,H. B.,Hopman, C.,&Alexander, T. 2006, preprintastro-
—.2006b,preprintastro-ph/0609440 ph/0606443
Gnedin, O. Y., Gould, A., Miralda-Escud´e, J., & Zentner, A. R. Peterson, R. C., Rood, R. T., & Crocker, D. A. 1995, ApJ, 453,
2005, ApJ,634,344 214
Greenstein,J.L.1966,ApJ,144,496 PortegiesZwart,S.F.2000,ApJ,544,437
Gualandris,A.&Portegies Zwart,S.2007,MNRAS,accepted Poveda, A., Ruiz, J., & Allen, C. 1967, Bol. Obs Tonantzintla
Gualandris, A., Portegies Zwart, S., & Eggleton, P. P. 2004, Tacubaya, 4,860
MNRAS,350,615 Press,W.H.,Teukolsky,S.A.,Vetterling,W.T.,&Flannery,B.P.
Gualandris, A., Portegies Zwart, S. P., & Sipior, M. S. 2005, 1992, Numerical recipes in C. The art of scientific computing
MNRAS,363,223 (Cambridge:UniversityPress,2nded.)
Hills,J.G.1988, Nature,331,687 Preston, G. W., Beers, T. C., & Shectman, S. A. 1994, AJ, 108,
—.1991,AJ,102,704 538
Hirsch,H.A.,Heber,U.,O’Toole,S.J.,&Bresolin,F.2005,A&A, Schaller,G.,Schaerer,D.,Meynet,G.,&Maeder,A.1992,A&AS,
444,L61 96,269
Holley-Bockelmann, K., Sigurdsson, S., Mihos, C. J., Feldmeier, Schneider,D.P.etal.2005, AJ,130,367
J. J., Ciardullo, R., & McBride, C. 2005, preprint astro- Sesana,A.,Haardt,F.,&Madau,P.2006, ApJ,inpress
ph/0512344 Wilhelm,R.,Beers,T.C.,Sommer-Larsen,J.,Pier,J.R.,Layden,
Humason,M.L.&Zwicky,F.1947,ContributionsfromtheMount A.C.,Flynn,C.,Rossi,S.,&Christensen,P.R.1999, AJ,117,
WilsonObservatory /Carnegie Institution of Washington, 724, 2329
1 Worthey, G., Faber, S. M., Gonzalez, J. J., & Burstein, D. 1994,
Kewley,L.J.,Brown,W.R.,Geller,M.J.,Kenyon,S.J.,&Kurtz, ApJS,94,687
M.J.2006,AJ,accepted Yi,S.,Demarque,P.,&Kim,Y.-C.1997,ApJ,482,677
Kollmeier,J.A.&Gould,A.2007, preprintastro-ph/0701350 Yu,Q.&Tremaine,S.2003,ApJ,599,1129
Kurtz,M.J.&Mink,D.J.1998,PASP,110,934
Leonard,P.J.T.1991,AJ,101,562
APPENDIX
DATA TABLE
Table 2 [on-line data table, attached as tab2.datin the source]lists the 643 new observations from our HVS survey,
excluding the extra-galactic objects. 134 HVS candidates were observed with the MMT and 509 were observed with
FAST (see §2). Table 2 includes columns of RA and Dec coordinates (J2000), g′ apparent magnitude, (u′−g′) and
0
(g′ −r′) color, and our heliocentric velocity v . The column WD indicates whether the object is a B-type star
0 helio
(WD=0) or a white dwarf (WD=1).
We note that (u′−g′) color correlates strongly with the observed spectral type of our B-type stars. Our spectral
0
types are completely independent of the SDSS photometry. We classify spectraltypes basedon O’Connell (1973)and
Worthey et al. (1994)line indices as described in Brownet al. (2003). Typical uncertainty is ±1.6 spectral sub-types.
FigureA6 plots our spectraltypes andthe (u′−g′) colorforour fullsample of699B-type stars. The solidline is the
0
best-fit relation: spectral type = 9.5(u′−g′) +11.5, where spectral type 10=B0, 15=B5, 20=A0, etc. The spectral
0
types of the stars in our HVS survey range B3 - A2, with a dispersion around the best-fit relation of ±1 spectral
sub-types or ±0.1 mag in (u′−g′) .
0
Fig. A6.— Spectraltypes determinedfromourspectroscopy
andSDSS(u′−g′)0 colorforthefullsampleof699B-typestars.