1: \documentclass[12pt,preprint]{aastex}
2: \usepackage{emulateapj5}
3: \usepackage{epsf}
4: \begin{document}
5: \shorttitle{Old-Population High Velocity Stars}
6: \shortauthors{Kollmeier \& Gould}
7: \newcommand{\msun}{M_{\odot}}
8: \newcommand{\kms}{\, {\rm km\, s}^{-1}}
9: \newcommand{\cm}{\, {\rm cm}}
10: \newcommand{\gm}{\, {\rm g}}
11: \newcommand{\erg}{\, {\rm erg}}
12: \newcommand{\kel}{\, {\rm K}}
13: \newcommand{\kpc}{\, {\rm kpc}}
14: \newcommand{\mpc}{\, {\rm Mpc}}
15: \newcommand{\seg}{\, {\rm s}}
16: \newcommand{\kev}{\, {\rm keV}}
17: \newcommand{\hz}{\, {\rm Hz}}
18: \newcommand{\etal}{et al.\ }
19: \newcommand{\yr}{\, {\rm yr}}
20: \newcommand{\mpyr}{{\rm mas}\, {\rm yr}^{-1}}
21:
22: \newcommand{\gyr}{\, {\rm Gyr}}
23: \newcommand{\eq}{eq.\ }
24: \def\arcsec{''\hskip-3pt .}
25: \def\prine{{}}
26: \def\ch{\bf }
27: \title{Where are the Old-Population High Velocity Stars?}
28: \author{Juna A. Kollmeier\altaffilmark{1,2} and Andrew Gould \altaffilmark{3}}
29:
30: \altaffiltext{1}{Observatories of the Carnegie Institution of Washington, 813 Santa Barbara Street, Pasadena, CA 91101}
31: \altaffiltext{2}{Hubble Fellow, Carnegie-Princeton Fellow}
32: \altaffiltext{3}{Dept. of Astronomy, The Ohio State University,
33: 140 W. 18th Ave, Columbus, OH 43210}
34:
35: \begin{abstract}
36:
37: To date, all of the reported high velocity stars (HVSs), which are
38: believed to be ejected from the Galactic center, are blue and
39: therefore almost certainly young. Old-population HVSs could be much
40: more numerous than the young ones that have been discovered, but still
41: have escaped detection because they are hidden in a much denser
42: background of Galactic halo stars. Discovery of these stars would
43: shed light on star formation at the Galactic center, the mechanism by
44: which they are ejected from it, and, if they prove numerous, enable
45: detailed studies of the structure of the dark halo. We analyze the
46: problem of finding these stars and show that the search should be
47: concentrated around the main-sequence turnoff $(0.3<g-i<1.1)$ at
48: relatively faint magnitudes $(19.5<g<21.5)$. If the ratio of turnoff
49: stars to B stars is the same for HVSs as it is in the local disk, such
50: a search would yield about 1 old-population HVS per $10\,{\rm deg}^2$.
51: A telescope similar to the Sloan 2.5m could search about $20\,{\rm
52: deg}^2$ per night, implying that in short order such a population,
53: should it exist, would show up in interesting numbers.
54: \end{abstract}
55:
56: \keywords{Galaxy: halo --- Galaxy: kinematics and dynamics --- Galaxy: stellar content --- Galaxy: center --- Stars: late-type}
57: \section{Introduction
58: \label{sec:intro}}
59:
60: Hypervelocity stars (HVSs---stars with velocities in excess of the
61: Galactic escape speed) have come a long way since \citet{hills88}
62: predicted their existence. It is now appreciated that, beyond being a
63: dynamical curiosity, these stars are useful probes of multi-scale
64: Galactic phenomena. Their frequency, spectral properties, and
65: distribution provide important constraints on the character of star
66: formation in the Galactic Center (GC) as well as the stellar ejection
67: mechanism itself. Furthermore, in sufficient numbers these objects
68: are unique dynamical tracers of the shape of the Milky Way's dark
69: matter halo---a critical quantity in understanding how the Galaxy fits
70: into the overall picture of hierarchical structure formation
71: \citep{gnedin05}.
72:
73:
74: Owing to their rapid Galactic exit times and low predicted ejection
75: rates for plausible dynamical mechanisms (e.g., \citealt{yu03}), these
76: stars are relatively rare. The first HVS discovery was a
77: serendipitous byproduct of a kinematic survey of blue horizontal
78: branch (BHB) stars \citep{brown05}. Through spectroscopic follow-up
79: of 36 faint ($19.75 < g < 20.5$) color selected $(0.8 < u - g <
80: 1.5)\cap(-0.3 < g - r < 0.0 )$ BHB stars from the Sloan Digital Sky
81: Survey (SDSS) First Data Release, these authors discovered a
82: $6\,\sigma$ radial-velocity outlier at $709\,\kms$ with $g = 19.81\pm
83: 0.02$ and dereddened colors $(u - g)_0= 1.04\pm 0.09$ and $(g - r)_0=
84: -0.30\pm 0.03$. This star was subsequently determined to be a
85: pulsating B-type main sequence star \citep{fuentes06}. Shortly after
86: this discovery, two additional HVSs were found, also within surveys
87: designed for the selection of early-type stars. During their survey
88: for subluminous B stars (sdB) \citet{edelman05} discovered an $8\msun$
89: B star with radial velocity of $563\,\kms$ located at $\sim
90: 60\,\kpc$ from the Galactic center, potentially ejected from the LMC.
91: \citet{hirsch05} followed up star US 708 as part of a survey of
92: $\sim 100$ subluminous O stars selected from SDSS having colors $(u -
93: g) < 0.2$ and $(g - r) < 0.1$, finding it to be a Helium rich
94: subluminous O star (HesdO) traveling at $\sim 720\,\kms$ at $25\,\kpc$ from
95: the GC. After these initial discoveries, in which HVSs were
96: contaminants in surveys for other blue stars, \citet{brown06}
97: undertook the first {\it targeted} survey for HVSs, in which
98: candidates were selected to be relatively faint, ($17.5 < g < 18.5$),
99: with B-star colors, $[-0.42<(g-r)<-0.27] \cap [1.33 < (u - g) - 2.67(g
100: -r) < 2.0 ]$. This search strategy has two key components: maximizing
101: both the volume covered and the contrast with normal halo stars. The
102: faint magnitudes achieve the first aim, while both the color and
103: magnitude selection contribute to the second. Contrast is improved at
104: faint magnitudes (large distances) because the HVS density should drop
105: off as $R^{-2}$, where $R$ is Galactocentric distance, while the
106: normal halo stars drop off faster than $R^{-3}$. It is improved
107: at blue colors because the only blue halo stars are BHB stars,
108: which have short lifetimes and so low density, and white dwarfs,
109: which are even rarer. Midway through their survey, \citet{brown06}
110: have found 4 probable HVSs out of 192 candidates culled from
111: $3000\,\rm deg^2$, i.e., a density of $(1/750)\,\deg^{-2}$.
112:
113: All reported HVSs are blue, which simply reflects the fact that, after
114: the first three serendipitous discoveries, the searches were conducted
115: among blue stars. One can imagine extending previous work
116: in several possible directions in parameter space, for example
117: by searching for the A stars emerging from the same
118: underlying population as the already discovered B stars. If the HVSs
119: are typical of bulge stars, however, there should be many {\it old}
120: ones. Due to the high background of unevolved halo stars, no one has
121: yet undertaken the daunting (and seemingly hopeless) task of a
122: comprehensive survey for this {\it late-type} population of HVSs.
123:
124: However, determining whether this population exists and in what
125: proportion would be of great interest. In particular the ratio of old
126: to young HVSs would place important constraints on the stellar
127: ejection mechanism itself. It is possible that the distribution of
128: HVS ages reflects the distribution in the stellar cusp near the GC.
129: Models that suggest HVS ejections are due to a short burst of
130: scatterings from an intermediate mass black hole (IMBH) that falls to
131: the GC by dynamical friction and disrupts the stellar cusp there
132: \citep{baumgardt06} would be directly tested by knowledge of the HVS
133: age distribution. With high-precision proper motion measurements of a
134: sufficient number of HVSs, one could measure not only the Galactic
135: potential but also the times of ejection for individual stars
136: \citep{gnedin05}. Regardless of the relative number of late- to
137: early-type HVSs, measurement of this ratio would be interesting. If
138: the ratio proved small, this observational fact could constrain
139: pictures in which the young stars near Sgr A* are brought to the
140: GC in clusters along with an IMBH as suggested by
141: \citet{hansen03}. It is also possible that late-type HVSs vastly
142: outnumber early-type HVSs, but the difficulty in finding this
143: population has prematurely biased our view of it. Should this be the
144: case, such stars would be vital probes of Galactic structure.
145:
146: In brief, the true age distribution of HVSs is simply unknown.
147: We therefore turn to the problem of how to find the old population.
148: Clearly, it is not practical at the current time to attempt spectroscopy
149: of all halo stars. In \S~\ref{sec:needle} we analyze the problem
150: of developing optimal selection criteria to search for them.
151: Then, in \S~\ref{sec:discuss} we comment on the prospects for detecting
152: HVSs in future surveys.
153:
154: \section{Needles in a Haystack
155: \label{sec:needle}}
156:
157: Old-population HVSs must be much less common than even the relatively
158: low-density population of halo stars. Otherwise, they would have been
159: discovered in spectroscopic surveys of high proper-motion stars.
160: Hence, finding such stars against the much more numerous background
161: of halo stars will require a well thought-out search strategy.
162:
163: If HVSs are ejected isotropically from the Galactic center at a rate
164: $\Gamma$, then their density at Galactocentric distance $R$ is
165: \begin{equation}
166: \rho(R) = {\Gamma\langle[v(R)]^{-1}\rangle\over 4\pi R^2},
167: \label{eqn:rho}
168: \end{equation}
169: where $v(R)$ is the velocity of the ejected star as a function of $R$
170: and where the brackets indicate averaging over the inverse velocities.
171: Note that ``$\Gamma$'' could refer to the HVS population as a whole
172: or to any subclass of stars within it.
173:
174: \subsection{Zeroth Order Analysis
175: \label{sec:zero}}
176:
177: To facilitate the exposition, we make a set of simplifying assumptions.
178: Taken together, these lead to a toy model that, while not realistic, does
179: yield a useful starting point for understanding the problem of finding HVSs.
180: In the next section we will sequentially relax these assumptions, allowing
181: the features of the real problem to come into focus.
182:
183: First, we assume that HVSs do not decelerate as they leave the Galaxy.
184: Equation (\ref{eqn:rho}) then simplifies to $\rho(R)\propto R^{-2}$.
185: Second, we assume that the physical density of halo stars also scales
186: $\rho_{\rm halo}\propto R^{-2}$. Then, under this very unrealistic
187: assumption, the ratio of HVSs to halo stars would have the same
188: constant value at any position in the Galaxy. Finally, we assume that
189: the color-magnitude relation of the old-population HVSs is identical
190: to that of halo stars. Together, these assumptions would imply that
191: the ratio of HVSs to halo stars is exactly the same for candidates
192: selected at any color and apparent magnitude and in any direction,
193: provided only that the selection criteria ensured that disk and
194: thick-disk stars were effectively excluded.
195:
196: \subsection{First Order Analysis
197: \label{sec:first}}
198:
199: As each of the three assumptions is relaxed, the fraction of HVSs
200: increases with increasing $R$. First, from equation (\ref{eqn:rho}),
201: progressive deceleration makes the density of HVSs fall more slowly
202: than $R^{-2}$. Second, the density of halo stars falls much
203: more quickly than $R^{-2}$. Locally halo stars fall roughly as
204: $R^{-3.2}$ (e.g.\ \citealt{gould98}), and the relation steepens
205: further out. Third, stars near the Galactic center are generally more
206: metal rich than halo stars, which implies that they are more luminous
207: on the upper main sequence. As we will show below, upper
208: main-sequence and turnoff stars dominate the HVS discovery potential.
209: The fact that HVSs are more luminous implies that they lie at farther
210: distances at fixed magnitude. Hence, they cover a larger range of
211: distance over a fixed magnitude range than the corresponding halo
212: stars. This enhances their density as a function of apparent
213: magnitude.
214:
215: Thus, other factors being equal, one should try to search for
216: HVSs as far from the Galactic center as possible. Within a given
217: field, ``other factors'' are obviously not equal because it is
218: easier to search among bright than faint stars. So this issue will
219: require additional analysis. However, we are at least driven to
220: the conclusion that the search for HVSs will be easiest toward
221: high-latitude, Galactic-anticenter fields: high-latitude to avoid
222: contamination from disk stars, and anticenter to reach the maximum
223: $R$ at fixed apparent magnitude.
224:
225: \subsection{Characteristics of Background
226: \label{sec:background}}
227:
228: Because the selection criteria for HVS candidates can only be color
229: and apparent magnitude, we must begin by analyzing the background
230: in terms of these two variables. We adopt a purely empirical approach,
231: tabulating the density of stars toward a SDSS
232: field centered at approximately RA=4 hr , Dec=$-6^\circ$
233: ($l=196$, $b=-40$). Figure \ref{fig:cmd} shows the stellar density
234: as a function of $g-i$ color for 7 different magnitude bins centered
235: on $g=19.0, 19.5.,\ldots 22.0$.
236: The key point is that in this magnitude range and in the region from the
237: turnoff redward, this density varies by less than a factor 4 altogether
238: and only by a factor 2 in the basic trend with color.
239: Moreover the color profiles at
240: each $g$ magnitude are approximately the same.
241: These characteristics imply that the
242: background does not play a crucial role in devising selection
243: procedures for HVS candidates as it would have were the curves
244: clearly separated: rather color/magnitude selection
245: must be based primarily on maximizing the total number of HVSs
246: and minimizing the amount of observing time required to identify
247: them as HVSs. We return to the issue of turnoff vs. giant and lower MS stars in \S~\ref{sec:color}.
248:
249: \subsection{Color Selection
250: \label{sec:color}}
251:
252: Under certain simplifying assumptions, the color/magnitude selection
253: actually factors into separate selections in color and magnitude.
254: We first introduce and motivate these assumptions and later evaluate
255: how their relaxation would impact our conclusions.
256:
257: First, we assume that deceleration is negligible so that, as mentioned
258: in \S~\ref{sec:zero}, $\rho\propto R^{-2}$. In fact, for an isothermal
259: sphere of circular velocity $v_{\rm circ}=220\,\kms$, the velocity-squared
260: falls by $\Delta v^2 = 2v_{\rm circ}^2\ln(R_2/R_1)$ between $R_1$
261: and $R_2$. For example, a star traveling at $800\,\kms$ at $15\, \kpc$
262: will slow by 15\% to $675\,\kms$ at $100\, \kpc$. This is not completely
263: negligible but it is modest compared to other factors in the problem.
264: Under this assumption, the number of HVSs of a fixed absolute magnitude (and
265: so by assumption fixed color) and a narrow range of
266: apparent magnitudes $\Delta g$, is
267: \begin{equation}
268: N = {\ln 10\over 5}\,{\Gamma\langle v^{-1}\rangle\over 4 \pi}\,
269: \Omega\Delta g
270: {r^3\over R^2}
271: \label{eqn:ntrue}
272: \end{equation}
273: where $\Omega$ is the angular size of the field and $r$ is the distance
274: from the observer to a star at the center of the magnitude bin.
275:
276: Second, we assume that $R=r$, which reduces the last term in equation
277: (\ref{eqn:ntrue}) from $r^3/R^2\rightarrow r$. That is,
278: $N_{\rm naive}\propto r$.
279: The ratio of this naive estimate to the true number is
280: \begin{equation}
281: {N_{\rm naive}\over N} = 1 - 2\cos l\cos b{R_0\over r} + {R_0^2\over r^2}
282: \rightarrow 1 + 1.47{R_0\over r} + {R_0^2\over r^2},
283: \label{eqn:nnaive}
284: \end{equation}
285: where $R_0=8\,\kpc$ is the solar Galactocentric radius.
286: Clearly this correction can be fairly large, so we will have to
287: carefully assess its impact after the selection criteria are derived.
288:
289: Because radial velocity (RV) measurements are most efficiently
290: carried out in the $g$-band part of the spectrum, RV precision for
291: a fixed exposure time is basically a function of $g$ magnitude.
292: This is not exactly true because the metal lines, from which these
293: determinations are primarily derived for FGK stars, are stronger at
294: lower temperatures. However, this is a modest correction, which we
295: will ignore for the moment but to which we will return below.
296:
297:
298: Consider now an ensemble of HVSs, drawn randomly from a common
299: old-star isochrone and ejected isotropically and stochastically
300: from the Galactic center. We now select stars at a fixed $g$-magnitude
301: (or rather in a narrow interval $\Delta g$ centered at fixed $g$),
302: which have a variety of absolute magnitudes $M_g$ and so (through
303: the color-magnitude relation of the isochrone), a variety of $g-i$ colors
304: (which are what we actually observe). The stars at $M_g$ will
305: be seen over a range of distance
306: $\Delta r = (\ln 10/5)10^{0.2(g-M_g+5)}\Delta g\,$pc, i.e.
307: $\Delta r \propto 10^{-0.2\,M_g}$. Under the above two assumptions,
308: the relative number of such stars in the sample will be
309: \begin{equation}
310: N_{det}(M_g) \propto 10^{-0.2\,M_g} \Phi(M_g),
311: \label{eqn:lumdep}
312: \end{equation}
313: where $\Phi(M_g)$ is the
314: fraction of stars from the isochrone in the $M_g$ bin. Note, in
315: particular that this relative number does not depend on $g$,
316: the apparent magnitude at which they are selected.
317:
318: Figure \ref{fig:cumulative} shows the cumulative distribution of these
319: relative numbers as a function of color (the observed quantity) for an
320: isochrone of solar metallicity and age of 10 Gyr \citep{yale}. Since
321: these are given in the Johnson/Cousins system, we convert to SDSS
322: bands using the transformations given on the SDSS web site
323: \citep{lupton05}. When the isochrone is viewed as a ``function'' of
324: color, it is double-valued. To illustrate the role of the two
325: branches, we plot their cumulative distributions separately, although
326: of course these could not be distinguished from color/magnitude data
327: alone. The solid curves illustrate the result under the assumption
328: that $r=R$, i.e., effectively that $g=\infty$. The dashed and
329: dot-dashed curves are for the more realistic cases of $g=22$ and
330: $g=21$ in the direction $(l,b)=(196^\circ ,-40^\circ)$.
331:
332: There are several important features of this diagram. First, and by
333: far the most important, the great majority of potential sensitivity to
334: old-population HVSs comes from stars with $g-i$ colors within 0.5
335: magnitudes of the turnoff, i.e., with $g-i<1.1$. This is already
336: basically true for the naive ``$g=\infty$'' case, but strictly applies
337: in realistic cases, $g\la 22$. By contrast, both the M and late-K dwarfs
338: on the lower branch and the M and late-K giants on the upper branch,
339: contribute very little, the former because they are so close and the
340: latter because they are so rare. Note that the dominance of turnoff
341: stars is a specific result of the $L^{1/2}$ luminosity-dependence in
342: equation~(\ref{eqn:lumdep}). If $N\propto L^{3/2}$ (as in a
343: magnitude-limited sample of uniform-density population) then giants
344: would dominate. If $N\propto L^0$ (as in a magnitude-limited sample
345: of an $r^{-3}$ halo-star-like population) then dwarfs would dominate.
346: Second, the lower branch contributes a bit more than double the upper
347: branch for realistic cases. That is, the sample is dominated by stars
348: just below the turnoff, with a significant, though clearly secondary,
349: contribution from stars just above the turnoff. This implies that the
350: validity of our approximations is basically determined by how well
351: they hold up at the turnoff. Third, the small contribution from
352: late-type giants obviates another potential complication. Depending
353: on the precise form of the ejection mechanisms, it is possible that
354: giant-star ejection is suppressed relative to smaller stars. For
355: example, some or all of the ejections might take place from disruption
356: of relatively tight binaries that are too close to permit giant-star
357: survival. Had Figure~\ref{fig:cumulative} implied giants dominated
358: the HVS distribution, this would lead to significant uncertainty.
359: However, the small contribution of giants, particularly late-type
360: giants, implies that any such suppression would also have small
361: impact. The one exception to this is the clump giants, which are not
362: included in the Yale Isochrones and hence are not represented in this
363: figure. They would contribute a small ``bump'' in the ``above
364: turnoff'' curve, similar in amplitude to the bump at $g-i\sim 1$ that
365: is actually seen in this curve, which is due to first-ascent giants.
366: With radii ten times solar, these stars are themselves relatively
367: small. However, with ages of only $100\,$Myr, they are younger than
368: the transport time to their current location, roughly $200\,$Myr at
369: $g=21$ and velocity $v=700\,\kms$. Hence, the progenitors of these
370: stars would have had to have been ejected when they were very
371: distended. In any event, they are not included in the figure.
372: Finally, the slightly greater RV precision (at fixed $g$ and fixed
373: exposure time) of cooler stars also has negligible impact, again
374: because of the small contribution of these stars.
375:
376: From this analysis, we conclude that at fixed magnitude, selection
377: should stretch from the turnoff
378: and proceed redward to $g-i=1.1$. In practice, the old-population
379: HVSs will not come from a single isochrone, but from a superposition
380: of many isochrones with a variety of ages and metallicities. However,
381: all of these are qualitatively similar, with just slightly varying
382: turnoff colors. Indeed, we investigated a 5 Gyr isochrone and
383: found results qualitatively similar to those shown in
384: Figure \ref{fig:cumulative}.
385: The important practical point is just to sample
386: the field stars beginning blue enough to cover all such turnoffs.
387: The cost of moving the blue boundary further blueward by $\Delta (g-i)=0.3$
388: is quite small, since this region of the observed field-star color-magnitude
389: diagram has few stars. We therefore advocate color selection $0.3<g-i<1.1$.
390:
391: \subsection{Magnitude Selection
392: \label{sec:magnitude}}
393:
394: Observations of a fixed exposure time can potentially measure RVs
395: to a given precision down to a certain apparent-magnitude limit $g$.
396: Once this is established, one could in principle measure RVs for all
397: stars within this limit, or (if fibers/slits were scarce) only those
398: within 2 magnitudes of the limit, which would contain $>60\%$ of all the
399: HVSs within the magnitude limit.
400: The following arguments apply equally to either strategy.
401:
402: Let us first suppose that ``downtime'' (for slewing, readout, and
403: changing slit masks or fiber positions) is negligible compared to the
404: exposure time. Let us compare two observation strategies, the first
405: with a single field exposed for time $\Delta t$ and the second with
406: two fields each exposed for $\Delta t/2$. Let us initially assume
407: that the magnitude limit is above sky in both cases. Then the flux
408: limit will be a factor 2 larger in the second case to maintain the
409: same signal-to-noise ratio (S/N). The maximum observable distance
410: will therefore be reduced by $2^{-1/2}$, which will decrease the
411: number of HVSs detected in each field by the same factor. However,
412: since there are twice as many fields, the total number of detected
413: HVSs will increase by $2^{1/2}$. Hence it is always better to go to
414: shorter exposures of more fields. If both limits are below sky, then
415: the flux limit increases by $2^{1/2}$, so the distance limit decreases
416: by $2^{-1/4}$, and the number of HVSs from both fields increases by
417: $2^{3/4}$. That is, the same argument applies even more strongly.
418:
419: Now consider the opposite limit, in which the exposure time is
420: negligible compared to the downtime. Shortening the exposure time
421: then still increases the flux limit and so reduces the number of HVSs
422: detected by $2^{-1/2}$ but in this case there is no compensating
423: increase in the number of fields covered. Comparing the two cases, it
424: is clear that the exposure times should be set approximately equal to
425: the downtime. It can be shown that the optimal exposure time is
426: exactly equal to the downtime if the magnitude limit is above sky and
427: equal to 1/3 of the downtime if it is below the sky.
428:
429: This argument somewhat overstates the case: it would be strictly valid
430: if the cumulative distributions illustrated in
431: Figure~\ref{fig:cumulative} were identical for the limiting magnitudes
432: corresponding to the two different exposure times. These curves are
433: nearly identical for the upper branch, but less so for the lower
434: branch. However, the argument remains qualitatively valid, the
435: correction being toward exposures that are somewhat longer than the
436: downtime.
437:
438: \subsection{Observing Strategy: General Considerations
439: \label{sec:generalities}}
440:
441: Before analyzing the characteristics of specific spectrographs, there
442: are two general points to consider. First, from Figure \ref{fig:cmd},
443: the stellar density in our recommended color range, $0.3<g-i<1.1$, is
444: about $150\,\rm mag^{-1}\,deg^{-2}$. Hence, if one is to cover 2 or 3
445: magnitudes in $g$, this requires monitoring 300--500 stars per square
446: degree. Note that in another direction, $(l,b)=(275^\circ,62^\circ)$,
447: we find stellar densities in this color-mag range that are about 2.5
448: times higher, confirming that it is substantially easier (in terms of
449: the sheer number of background contaminants) to search for
450: old-population HVS in high-latitude anticenter fields.
451:
452: Second, the RV precision requirements to distinguish HVSs from halo stars
453: are not very severe: $\sigma=50\,\kms$ would be quite adequate. Typical
454: HVSs have RVs $v_r\sim 700\,\kms$, well separated from halo stars at
455: $|v_r|\sim O(170\,\kms)$. There are, of course, halo stars moving
456: closer to the escape velocity, but these are relatively rare, and the
457: chance that one of these would further upscatter by $\sim 5\,\sigma$
458: into the HVS range is quite small. This precision is not difficult
459: to achieve even in very noisy spectra, particularly on FGK stars,
460: which have many spectral features.
461:
462: \subsection{Specific Evaluations
463: \label{sec:specifics}}
464:
465: From this point forward, concrete development of an observing strategy
466: obviously depends
467: on the detailed characteristics of the multi-object spectrograph, which
468: cannot be treated completely generally.
469: However, to give some broad guidance and to help understand the sensitivity
470: of the search under realistic conditions, we consider two specific
471: multi-object spectrographs with radically different characteristics.
472:
473: First, we consider the resolution ${\cal R}= 2000$ SDSS spectrograph, which has
474: 640 $3''$ diameter fibers spanning a $7\,\rm deg^2$ field on a 2.5 m telescope.
475: The fiber plug-plates require about 10 minutes to change. In principle,
476: one should consider the time required for repointing the telescope,
477: but this will be relatively infrequent because even in the anticenter fields,
478: there are about 2000 viable targets (for a 2-magnitude interval) and only
479: 640 fibers, so there should be about three plug-plates per pointing.
480: If we strictly applied the ``exposure time equals downtime''
481: rule, the resulting 10 minute
482: exposure would yield per pixel S/N=10
483: at about $g=19.5$. Given the $3''$ fibers, this
484: is about 1 mag below dark sky. In practice, it is probably impractical
485: to change plug plates so frequently, so 45 minute exposures (in keeping
486: with current SDSS practice) would appear more realistic. Taking account
487: of sky, this yields a per pixel S/N=8 at $g=20.5$ and S/N=4 at $g=21.5$.
488: This latter is probably adequate for $\sigma=50\,\kms$ measurements, but this
489: should be tested directly. Hence, in one 9-hour night, the SDSS
490: telescope could cover a total of $21\,\rm deg^2$ over $19.5<g<21.5$ and
491: $0.3<g-i<1.1$, i.e., 3 45-minute exposures (1 for each of 3 plug plates)
492: on each of 3 fields. One additional wrinkle should be noted. The
493: 640 fibers are divided into 320 red and 320 blue channels, with
494: the dividing line at 6000$\,$\AA . The blue-channel fibers are
495: more sensitive to RV because of the greater density of lines.
496: However, this can easily be compensated by applying the red-channel
497: fibers to the brighter stars. Being systematically 1 magnitude
498: brighter (and recalling that the target stars are mostly below sky),
499: these would have about 2.5 times higher S/N per pixel.
500:
501: Second, we consider the ${\cal R}=20000$ IMACS F/4 spectrograph, which
502: accommodates up to 1000 slits spanning a $0.067\,\rm deg^2$ field on
503: the Magellan 6.5 m telescope. Changing slit masks requires about 15
504: minutes, but in fact this is not the relevant scale of ``downtime''
505: because each field has only of order 20--30 available targets, far
506: fewer than the available slits. Rather, each mask could be cut to
507: serve of order 10 fields (with a total of 200-300 targets). Hence,
508: the downtime is primarily set by the time required to acquire a new
509: field (without changing the mask). This is roughly 5 minutes, which
510: by the guideline derived in \S~\ref{sec:magnitude} would indicate an
511: exposure time also of 5 minutes. Taking account of sky noise and
512: assuming $4\,e^{-}$ read noise, this leads to an estimate of per
513: pixel S/N=1.2 at $g=21.5$. To evaluate the utility of such
514: signal levels, we construct synthetic spectra, add noise, and then
515: fit the results to a (4-parameter) quadratic polynomial plus a template
516: spectrum, offset by various velocities from the constructed spectrum.
517: We find that this S/N is sufficient for an
518: accurate RV measurement, provided that at least $80\,$\AA\ are sampled
519: centered on $\lambda=5175\,$\AA . In fact, the formal error in the
520: measurement is less than $10\,\kms$, so that it would appear that even
521: lower S/N would be tolerable. However, we find if the S/N is further
522: reduced, that while the width of the correlation peak does not
523: increase dramatically, multiple minima (each quite narrow) begin to
524: appear, undermining the measurement. Similarly, if the wavelength
525: coverage is reduced at constant S/N, then multiple minima also appear.
526: In any event, if appropriate blocking filters permit this $80\,$\AA\
527: (800 pixel) window or larger,
528: then the field can be reliably probed to $g=21.5$.
529: Of order 50 fields could be searched during a 9 hour night, covering
530: about $3.3\,\rm deg^2$.
531:
532: Thus, while these two spectrographs differ in aperture, resolution,
533: and field size by factors of 7, 10, and 100 respectively, they
534: are capable of broadly similar searches for HVSs. We conclude
535: that it is feasible to conduct the search over tens of square degrees
536: on a variety of telescopes without exorbitant effort.
537:
538: \section{Discussion
539: \label{sec:discuss}}
540:
541: How likely is it that such a survey of several tens of square degrees
542: will detect old-population HVSs? At one level, as emphasized in
543: \S~\ref{sec:intro}, we have no idea: based on what we know now,
544: old-population HVSs could equally well be very common or non-existent.
545: However, in the absence of any hard information, we might guess
546: that the ratio of old-population to B-type HVSs might be similar
547: to the ratio of the underlying populations near the GC. This itself
548: is not known, but as proxy we evaluate the same ratio in the
549: solar neighborhood. In fact, what is required is the ratio
550: of turnoff stars to B stars, since our proposed survey is
551: most sensitive to turnoff stars, while B stars form the only
552: population of HVSs that have been reliably tabulated. The
553: distance ranges are similar: the B stars have typically been
554: detected at 60 kpc while turnoff stars at $g=21.5$ lie at about 25 kpc
555: (a factor 2.5 advantage for the B stars) but the B stars were searched
556: over 1 mag in $g$ while the turnoff stars would be searched over 2 mags
557: (which basically compensates).
558:
559: We estimate the ratio of $B$ stars to turnoff stars in the solar
560: neighborhood as follows. We analyze samples of each stellar class drawn
561: from the Hipparcos catalog \citep{hip}, restricted to $V<7.3$
562: (Hipparcos completeness limit), distance $r<300\,$pc (to ensure
563: good parallaxes and low extinction), and distance from the Galactic plane
564: less than 50 pc. We define turnoff stars as having absolute magnitudes
565: $3.5<M_V<4.5$ and near-turnoff colors ($0.3<B_T -V_T<0.8$). For
566: B stars, we probe a 2-magnitude interval $-2.0<M_V<0.0$, which approximately
567: corresponds to the \citet{brown06} $g-r$ selection criterion, and we enforce
568: ($-0.5<B_T -V_T<0.2$) to distinguish these from giants. For each
569: class of star, we tabulate $\sum_i (V_{{\rm eff},i})^{-1}$,
570: where $V_{{\rm eff},i}$ is the ``effective volume'' over which that
571: star could have been found. That is $V_{{\rm eff}}= 4\pi r_{\rm max}^3$
572: where $r_{\rm max}$ is the maximum distance that star could have been
573: detected given its observed absolute magnitude, its known direction,
574: and the two constraints on distance given above. We find effective
575: densities of $0.91\times 10^{-5}\,\rm pc^{-3}$ and
576: $60.7\times 10^{-5}\,\rm pc^{-3}$ for the two classes, indicating
577: that turnoff stars are about 67 times more common that B stars.
578:
579: Applying this rather crudely derived multiplier to the B-star HVS density
580: found by \citet{brown06}, we estimate that there could be 1 turnoff HVS
581: per $10\,\rm deg^2$. Hence a survey of a few tens of square degrees
582: could probe the existence of this putative population.
583:
584: \acknowledgments
585: We thank Alan Dressler and Mike Gladders for discussions
586: at OCIW morning tea that stimulated the genesis of this paper.
587: We are also grateful to Scott Gaudi, Jennifer Johnson,
588: Steve Shectman, and Ian Thompson, for helpful discussions.
589: J.A.K. was supported by NASA through a Hubble Fellowship.
590: A.G. was supported by grant AST 042758 from the NSF.
591: Any opinions, findings, and conclusions or recommendations expressed in
592: this material are those of the authors and do not necessarily reflect the
593: views of the NSF.
594:
595:
596:
597:
598: \begin{thebibliography}{}
599:
600: \bibitem[Baumgardt et al.(2006)]{baumgardt06}
601: Baumgardt, H., Gualandris, A., \& Portegies Zwart, S.\ 2006, Journal
602: of Physics Conference Series, 54, 301
603:
604: \bibitem[Brown et al.(2005)]{brown05}
605: Brown, W.~R., Geller, M.~J., Kenyon, S.~J., \& Kurtz, M.~J.\ 2005,
606: \apjl, 622, L33
607:
608: \bibitem[Brown et al.(2006)]{brown06}
609: Brown, W.,R.~et al.\ 2006, astro-ph/0604111
610:
611: \bibitem[ESA(1997)]{hip} European Space Agency (ESA). 1997, The Hipparcos and
612: Tycho Catalogues (SP-1200; Noordwijk: ESA)
613:
614: \bibitem[Edelmann et al.(2005)]{edelman05}
615: Edelmann, H., Napiwotzki, R., Heber, U., Christlieb, N., \& Reimers,
616: D.\ 2005, \apjl, 634, L181
617:
618: \bibitem[Fuentes et al.(2006)]{fuentes06}
619: Fuentes, C.~I., Stanek, K.~Z., Gaudi, B.~S., McLeod, B.~A., Bogdanov,
620: S., Hartman, J.~D., Hickox, R.~C., \& Holman, M.~J.\ 2006, \apjl, 636,
621: L37
622:
623: \bibitem[Gould et al.(1998)]{gould98}Gould, A., Flynn, C., \& Bahcall, J.N.
624: 1998, \apj, 503, 798
625:
626: \bibitem[Gnedin et al.(2005)]{gnedin05}
627: Gnedin, O.~Y., Gould, A., Miralda-Escud{\'e}, J., \& Zentner, A.~R.\
628: 2005, \apj, 634, 344
629:
630:
631: \bibitem[Hansen \& Milosavljevi{\'c}(2003)]{hansen03}
632: Hansen, B.~M.~S., \& Milosavljevi{\'c}, M.\ 2003, \apjl, 593, L77
633:
634:
635: \bibitem[Hills(1988)]{hills88}
636: Hills, J.~G.\ 1988, \nat, 331, 687
637:
638: \bibitem[Hirsch et al.(2005)]{hirsch05} Hirsch, H.~A., Heber,
639: U., O'Toole, S.~J., \& Bresolin, F.\ 2005, \aap, 444, L61
640:
641: \bibitem[Lupton(2005)]{lupton05} Lupton, R. 2005,
642: http://www.sdss.org/dr5/algorithms/sdssUBVRITransform.html.
643:
644:
645: \bibitem[Demarque et al.(2004)]{yale} Demarque, P., Woo,
646: > J.-H., Kim, Y.-C., \& Yi, S.K.\ 2004, \apjs, 155, 667
647:
648: \bibitem[Yu \& Tremaine(2003)]{yu03} Yu, Q., \& Tremaine,
649: S.\ 2003, \apj, 599, 1129
650: \end{thebibliography}
651:
652: \begin{figure}
653: \plotone{f1.ps}
654: \caption{Density ($\rm mag^{-2}\,deg^{-2}$) of stars in the color
655: magnitude diagram of the SDSS high-latitude anti-center field toward
656: $(l,b)=(196,-40)$. Curves correspond
657: to constant-magnitude bins at half-magnitude intervals from $g=19$ to
658: $g=22$. To a first approximation, all the curves are the same.
659: Moreover, they are roughly flat redward of the turnoff (at $g-i\sim
660: 0.65$) and even on the turnoff they are only about double the plateau
661: value. }
662: \label{fig:cmd}
663: \end{figure}
664:
665: \begin{figure}
666: \plotone{f2.ps}
667: \caption{Cumulative distribution of expected high-velocity-star detections
668: of apparent magnitude $g$
669: as a function of $g-i$ color, under the assumption that all stars
670: in the underlying old-population are equally likely to be ejected from
671: the Galactic center. The red curves represent stars below the main-sequence
672: turnoff while the blue curves represent stars above the turnoff.
673: At very faint magnitudes, the result is independent of magnitude and
674: is shown by the two {\it solid curves}, whose total is arbitrarily
675: normalized to unity. At realistic magnitudes, $g=22$ ({\it dot-dashed}) or
676: $g=21$ ({\it dashed}), the detections are
677: somewhat suppressed by eq.~(\ref{eqn:nnaive}), particularly for the
678: below-turnoff branch. For these curves, the great majority of expected
679: detections are at $g-i<1.1$.
680: }
681: \label{fig:cumulative}
682: \end{figure}
683:
684:
685:
686:
687: \end{document}
688: