1: \def\lae{\mathrel{<\kern-1.0em\lower0.9ex\hbox{$\sim$}}}
2: \def\gae{\mathrel{>\kern-1.0em\lower0.9ex\hbox{$\sim$}}}
3: \font\fsmall=cmr8
4: \newcommand{\be}{\begin{equation}}
5: \newcommand{\ee}{\end{equation}}
6:
7: %\documentclass[12pt,preprint]{aastex}
8:
9: \documentclass[apj]{emulateapj}
10: \bibpunct[; ]{(}{)}{;}{a}{}{;}
11: \usepackage{apjfonts}
12:
13: \slugcomment{Accepted for publication in ApJ Letters}
14:
15: \shortauthors{JORD\'AN ET AL.}
16: \shorttitle{GCLF TRENDS}
17:
18: \begin{document}
19:
20: \title{Trends in the Globular Cluster Luminosity Function of Early-Type
21: Galaxies\altaffilmark{1}}
22:
23: \author{
24: Andr\'es Jord\'an\altaffilmark{2},
25: Dean E. McLaughlin\altaffilmark{3},
26: Patrick C\^ot\'e\altaffilmark{4},
27: Laura Ferrarese\altaffilmark{4},
28: Eric W. Peng\altaffilmark{4},
29: John P. Blakeslee\altaffilmark{5},
30: Simona Mei\altaffilmark{6},
31: Daniela Villegas\altaffilmark{2,7},
32: David Merritt\altaffilmark{8},
33: John L. Tonry\altaffilmark{9},
34: Michael J. West\altaffilmark{10,11}
35: }
36:
37: \begin{abstract}
38: We present results from a study of the globular cluster luminosity
39: function (GCLF) in a sample of 89 early-type galaxies observed as part
40: of the ACS Virgo Cluster Survey. Using a Gaussian parametrization of
41: the GCLF, we find a highly significant correlation between the GCLF
42: dispersion, $\sigma$, and the galaxy luminosity, $M_{B,{\rm gal}}$, in
43: the sense that the GC systems in fainter galaxies have narrower
44: luminosity functions. The GCLF dispersions in the Milky Way and M31 are
45: fully consistent with this trend, implying that the correlation between sigma
46: and galaxy luminosity is more fundamental than older suggestions that GCLF
47: shape is a function of galaxy Hubble type.
48: We show that the $\sigma - M_{B,{\rm gal}}$ relation
49: results from a bonafide narrowing of the
50: distribution of (logarithmic) cluster masses in fainter
51: galaxies.
52: We further show that this behavior is mirrored by a
53: steepening of the GC mass function for relatively high masses,
54: ${\cal M} \ga 3\times10^5\,{\cal M}_\odot$, a mass regime in which
55: the shape of the
56: GCLF is not strongly affected by dynamical evolution over a Hubble time. We
57: argue that this trend arises from variations in initial conditions
58: and requires explanation by theories of cluster formation.
59: Finally, we confirm that in bright galaxies, the GCLF ``turns over" at the
60: canonical mass scale of
61: ${\cal M}_{\rm TO}\simeq 2\times 10^5\,{\cal M}_\odot$. However, we find
62: that ${\cal M}_{\rm TO}$ scatters to lower values
63: ($\approx$ 1-2$\times 10^5\,{\cal M}_\odot$) in galaxies fainter than
64: $M_{B,{\rm gal}}\ga -18.5$, an important consideration if
65: the GCLF is to be used as a distance indicator for dwarf ellipticals.
66: \end{abstract}
67:
68: \keywords{galaxies: elliptical and lenticular, cD ---
69: galaxies: star clusters ---
70: globular clusters: general}
71:
72: \altaffiltext{1}{Based on observations with the NASA/ESA
73: {\it Hubble Space Telescope}
74: obtained at the Space Telescope Science Institute, which is operated
75: by the Association
76: of Universities for Research in Astronomy, Inc., under
77: NASA contract NAS 5-26555}
78: \altaffiltext{2}{European Southern Observatory,
79: Karl-Schwarzschild-Stra{\ss}e 2, 85748 Garching bei M\"unchen, Germany;
80: ajordan@eso.org}
81: \altaffiltext{3}{Department of Physics \& Astronomy, University
82: of Leicester, Leicester, LE1 7RH, UK; dean.mclaughlin@astro.le.ac.uk}
83: \altaffiltext{4}{Herzberg Institute of Astrophysics, Victoria,
84: BC V9E 2E7, Canada}
85: \altaffiltext{5}{Department of Physics and Astronomy,
86: Washington State University, 1245 Webster Hall, Pullman, WA 99163-2814}
87: \altaffiltext{6}{Department of Physics and Astronomy,
88: Johns Hopkins University, Baltimore, MD 21218}
89: \altaffiltext{7}{Departamento de Astronom\'{\i}a y Astrof\'{\i}sica,
90: Pontificia Universidad Cat\'olica de Chile,
91: Avenida Vicu\~na Mackenna 4860, Casilla 306, Santiago 22, Chile}
92: \altaffiltext{8}{Department of Physics, Rochester Institute of Technology,
93: 84 Lomb Memorial Drive, Rochester, NY 14623}
94: \altaffiltext{9}{Institute of Astronomy, University of Hawaii,
95: 2680 Woodlawn Drive, Honolulu, HI 96822}
96: \altaffiltext{10}{Department of Physics \& Astronomy, University of Hawaii,
97: Hilo, HI 96720}
98: \altaffiltext{11}{Gemini Observatory, Casilla 603, La Serena, Chile}
99:
100: \section{Introduction}
101: \label{sec:intro}
102:
103: The luminosity function of globular clusters (GCs) represents
104: one of the most remarkable features of these stellar systems.
105: The distribution of GC magnitudes, commonly referred to as the
106: GC luminosity function (GCLF), shows a turnover, or peak, at
107: $M_V \simeq -7.5$ mag, corresponding to a mass of
108: ${\cal M} \simeq 2 \times 10^5 {\cal M}_\odot$. Observations have shown
109: that this turnover is nearly invariant across and within galaxies, prompting
110: its widespread use as a distance indicator (see, e.g., Harris 2001).
111: %
112: Accounting for this nearly universal mass scale remains
113: an open problem for theories of GC formation and evolution.
114: %
115: It follows that establishing whether or not the GCLF as a whole is
116: universal --- i.e., whether its overall form depends on
117: host galaxy properties --- can help guide and
118: constrain theories for the formation and evolution
119: of galaxies and GC systems.
120:
121: In this {\it Letter}, we present results from a study of
122: the GCLFs of 89 early-type galaxies observed by HST as
123: part of the ACS Virgo Cluster Survey (ACSVCS; C\^ot\'e et al.~2004).
124: We find the clearest evidence to date for a correlation
125: between the width (i.e., Gaussian dispersion) of the GCLF and the
126: luminosity of the host galaxy;
127: we also show that there is some downward scatter in the
128: {\it mass} scale of the GCLF turnover in galaxies fainter
129: than $M_{B,{\rm gal}}\ga -18.5$.
130: Focusing on the observed steepening of the GCLF
131: at the bright (high-mass) end in the faint galaxies,
132: we argue that this behavior was probably imprinted at the time of GC
133: formation. A more detailed discussion of the whole GCLF,
134: including the faint (low-mass) end and the
135: role that long-term dynamical evolution plays in that regime, is deferred to a
136: subsequent paper (Jord\'an et~al.~2006, hereafter J06). That paper
137: presents our data in full and gives details of our analysis techniques,
138: including modeling of the GCLFs with a new, non-Gaussian, physically
139: motivated fitting function.
140:
141: \section{Observations and Analysis}
142: \label{sec:data}
143:
144: One hundred early-type galaxies in the Virgo cluster
145: were observed in the ACSVCS (C\^ot\'e et~al. 2004).
146: Each galaxy was imaged for 750 s in the F475W bandpass
147: ($\simeq$ Sloan~$g$) and for 1210 s in F850LP ($\simeq$ Sloan~$z$).
148: Reductions were performed as described in
149: Jord\'an et~al. (2004). In what follows, we use $g$ and $z$ as
150: shorthand to refer to the F475W and F850LP filters.
151:
152: One of the main scientific objectives of the ACSVCS is the study of
153: GC systems, and thus we have developed methods
154: to: (1) discard foreground stars and background galaxies from the
155: totality of observed sources around each target galaxy in the survey;
156: and (2) estimate the
157: level of residual fore- and background contamination in the remaining sources
158: designated as candidate GCs. These procedures are described and illustrated
159: by Peng et al.~(2006a; their \S2.2 and Figure~1), and discussed in detail in
160: the GCLF context in J06. In the latter paper, we also
161: examine the effects of using alternate selection
162: criteria to define GC samples, and show that the results
163: presented here are fully robust against such subtleties.
164:
165: Of the 100 galaxies in the ACSVCS, we restrict our analysis to those that
166: have more than five GCs, as estimated by subtracting the total number
167: of expected contaminants from the full list of GC candidates for each galaxy.
168: We additionally eliminate two galaxies for which we were
169: unable to obtain useful measurements of the GCLF parameters. This leaves
170: a final sample of 89 galaxies which are studied here, and in J06.
171:
172: Also as part of the ACSVCS, we have measured the distances to 84
173: of our target galaxies using the method of surface
174: brightness fluctuations (SBF; Mei et~al. 2006). We use these SBF distances to
175: transform the observed
176: GC and galaxy magnitudes into absolute ones whenever possible. For those
177: galaxies lacking an SBF distance,
178: we adopt the mean distance modulus to the
179: Virgo cluster: $\langle (m-M)_0 \rangle = 31.09$~mag, or
180: $\langle D \rangle=16.5$~Mpc (see Mei et~al. 2005, 2006).
181:
182: We use an approach similar to that of Secker \& Harris (1993) to
183: characterize the GCLFs: parametric models are fitted to the observed
184: luminosity functions via a maximum-likelihood method that takes into account
185: photometric errors, incompleteness, and the luminosity function of
186: contaminants. Full technical details are given in J06, where we consider
187: two parametric models for the GCLF. The first, on which this paper will focus,
188: is the standard Gaussian distribution,
189: \begin{equation}
190: dN/dz = N_{\rm tot} \, \big( 2\pi\,\sigma_z^2 \big)^{-1/2} \
191: \exp \big[- (z-\mu_z)^2/2\sigma_z^2 \big]\ .
192: \label{eq:gauss}
193: \end{equation}
194: The second is a simple analytical modification of a Schechter (1976) function
195: designed to account for the effects of cluster evaporation (two-body
196: relaxation) on a GC mass function that is assumed to have initially resembled
197: that of the young clusters forming today in local mergers and starbursts. Full
198: details on these two models are given in J06, where we fit each of them
199: to the separate $g$- and $z$-band GCLFs of our 89 program galaxies. In this
200: {\it Letter} we present only the results of Gaussian fits to the $z$-band
201: GCLFs.
202:
203: \section{Results}
204: \label{sec:results}
205:
206: \begin{figure}
207: \epsscale{1.0}
208: \plotone{f1.eps}
209: \caption{Gaussian dispersion, $\sigma_z$, versus galaxy, $M_{B,{\rm gal}}$,
210: for the $z$-band GCLFs of 89 ACSVCS galaxies. The GCLF width varies
211: systematically, being narrower in fainter galaxies. The two anomalously high
212: points at $M_{B, {\rm gal}}=-21.2$ and $-19.9$ correspond to the galaxies
213: VCC 798 and VCC 2095, both of which have large excesses of faint, diffuse
214: clusters (Peng et al.~2006b). The large star is plotted at the
215: spheroid luminosity (de Vaucouleurs \& Pence 1978) and GCLF dispersion (Harris
216: 2001) of the Milky Way. The large triangle marks the bulge luminosity (Kent
217: 1989) and GCLF dispersion (Harris 2001) of M31.
218: \label{fig:sigma_B}}
219: \end{figure}
220:
221:
222:
223: Figure~\ref{fig:sigma_B} shows our main result: GCLFs are narrower in
224: lower-luminosity galaxies. The straight line in this plot of Gaussian
225: dispersion against absolute galaxy magnitude
226: shows the least-squares fit
227: \begin{equation}
228: \sigma_z = (1.12\pm0.01) - (0.093 \pm 0.006)(M_{B, {\rm gal}}+20) \ .
229: \label{eq:sigmaz}
230: \end{equation}
231:
232: It has been reported before that the GCLFs in
233: lower-luminosity galaxies tend to show somewhat lower dispersions
234: (e.g., Kundu \& Whitmore 2001). However, the size and homogeneity of
235: the ACSVCS dataset make this
236: the most convincing demonstration to date of a
237: continuous trend in GCLF shape over a range of $\ga\! 400$ in galaxy
238: luminosity. Monte Carlo simulations and alternate constructions of GCLF
239: samples show that the observed decrease in dispersion is {\it not}
240: an artifact of small-number statistics in the faint galaxies (J06).
241:
242: Past investigations have pointed to a dependence of the GCLF dispersion on
243: the Hubble type of the GC host galaxies (e.g., Harris
244: 1991). Figure~\ref{fig:sigma_B}
245: includes datapoints at the location of the bulge magnitude and GCLF
246: dispersion of the Milky Way (large star) and M31 (large triangle).
247: Since both systems fall comfortably on the relation defined by
248: our data for early-type galaxies, we conclude that the underlying
249: fundamental correlation is one between $\sigma$ and $M_{B,\rm gal}$,
250: rather than between $\sigma$ and Hubble type.
251:
252: A natural question at this point is whether the observed trend in GCLF
253: dispersion with galaxy magnitude implies a similar trend in the GC
254: {\it mass} function. This is not a foregone conclusion, for the following
255: reason. GC systems are known to have systematically redder
256: and broader (or more strongly bimodal) color distributions in brighter
257: galaxies than in faint ones (see, e.g., Peng et al.~2006a). Equivalently,
258: GCs in giant galaxies are more metal-rich on average, and have larger
259: dispersions in [Fe/H], than those in low-mass dwarfs.
260: Since cluster
261: mass-to-light ratios, $\Upsilon$, are functions of [Fe/H] in
262: general, it is conceivable that the average GC $\Upsilon$ could
263: change systematically in going from bright galaxies to fainter ones,
264: and that the spread of $\Upsilon$ values within a single GC system
265: could also vary systematically as a function of galaxy magnitude.
266: The possibility then exists
267: that narrower GCLFs for faint galaxies might result from these systematics
268: in $\Upsilon$ combined with a more nearly invariant spread in GC
269: masses.
270: %
271: We can show easily, however, that this is not the case.
272:
273: The systematics in $\Upsilon$ vs.~[Fe/H] just mentioned are
274: also a function of wavelength. In bluer filters, such as $B$, $V$, or $g$,
275: mass-to-light ratios of old stellar systems
276: do change significantly (increasing by factors
277: of two or more) in going from cluster metallicities ${\rm [Fe/H]} \la -2$ to
278: ${\rm [Fe/H]}=0$, typical of GCs. But at the much redder wavelengths of our
279: $z$-band data ($\lambda_{\rm pivot} \simeq 9055$~\AA; Sirianni et~al. 2005),
280: this strong metallicity dependence almost completely disappears. We have used
281: the PEGASE population-synthesis
282: model of Fioc \& Rocca-Volmerange (1997) to compute $\Upsilon_z$ as a function
283: of metallicity for clusters with a Kennicutt (1983)
284: stellar IMF and various fixed ages $\tau$. For $\tau=13$~Gyr, we find
285: that $\Upsilon_z\simeq 1.6\ {\cal M}_\odot\,L_\odot^{-1}$ at an extreme
286: ${\rm [Fe/H]}=-2.3$, decreasing to a minimum of
287: $\Upsilon_z\simeq 1.5\ {\cal M}_\odot\,L_\odot^{-1}$ at
288: ${\rm [Fe/H]}\simeq-0.7$, and then increasing slightly to
289: $\Upsilon_z=1.7\ {\cal M}_\odot\,L_\odot^{-1}$ at
290: ${\rm [Fe/H]}=0$. In other words, we always have
291: $\Upsilon_z\approx 1.6\pm0.1$ for any of the globular clusters
292: in any of our sample galaxies --- no matter how red or blue the clusters are,
293: or how broad or narrow the GC color/metallicity distribution.
294: Comparably small ranges of $\Upsilon_z$ result if younger GC ages or different
295: reasonable stellar IMFs are assumed.
296:
297: The effect of variations in mass-to-light ratio on the width of the GCLF at
298: NIR wavelengths is therefore completely negligible. From
299: Figure~\ref{fig:sigma_B}, we have that $\sigma(\log\,L_z)=\sigma_z/2.5\ga0.2$
300: in our galaxies, whereas the discussion above implies that
301: $\Upsilon_z({\rm max})/\Upsilon_z({\rm min})\sim 1.13$ for our GCs, such that
302: the dispersion of mass-to-light ratios in any one system is always
303: $\sigma(\log\,\Upsilon_z)<0.055$ at an absolute maximum.
304: The intrinsic dispersion of logarithmic GC {\it masses},
305: $\sigma(\log\,{\cal M}) =
306: [ \sigma^2(\log\,L_z) - \sigma^2(\log\,\Upsilon_z) ]^{1/2}$,
307: is thus never more than $\sim\!4\%$ different from the
308: observed $\sigma(\log\,L_z)$.
309: We conclude, unavoidably, that the steady decrease of $\sigma_z$ by more than
310: 50\% from the brightest giants to the faintest dwarfs in
311: Figure~\ref{fig:sigma_B} is an accurate reflection of
312: just such a trend in the intrinsic GC mass distributions.
313:
314: We now turn our attention to the GCLF turnover magnitude.
315: The upper panel of Figure~\ref{fig:mu_B} shows the mean GC absolute
316: magnitude $\mu_z$ from the Gaussian fits to our GCLFs, versus
317: host galaxy $M_{B,{\rm gal}}$.
318: The horizontal line in this plot is drawn at a level typical of
319: galaxies brighter than $M_{B,{\rm gal}}\la -18.5$: $\mu_z=-8.4$. Given
320: a typical $\Upsilon_z\simeq1.5\ {\cal M}_\odot\,L_\odot^{-1}$ in these
321: galaxies (for GC ages 13 Gyr and an average
322: ${\rm [Fe/H]}\approx-1$), this
323: corresponds to a cluster mass scale of ${\cal M}_{\rm TO}\simeq 2.2\times10^5\
324: {\cal M}_\odot$.
325: Estimates of the $z$-band GCLF turnovers in the Milky Way and M31
326: are shown by the large star and large triangle, as in Figure~\ref{fig:sigma_B}.
327: In the lower panel of Figure \ref{fig:mu_B} we plot the turnover
328: masses ${\cal M}_{\rm TO}$ obtained from the fitted $\mu_z$ using
329: PEGASE model mass-to-light ratios.
330: As we have discussed, $z$-band luminosities are very good proxies for total
331: GC masses, so this graph is essentially a mirror image of the one above it.
332:
333: \begin{figure}[!t]
334: \epsscale{1.0}
335: \plotone{f2.eps}
336: \caption{
337: ({\it Top}) GCLF turnover magnitude (absolute mean $\mu_z$) versus galaxy
338: magnitude, $M_{B,{\rm gal}}$, from Gaussian fits to 89 $z$-band GCLFs in the
339: ACSVCS. The outlying points at $M_{B,{\rm gal}}=-21.2$ and $-19.9$ are
340: VCC 798 and VCC 2095, which have large excesses of faint, diffuse clusters
341: (Peng et al.~2006b). The star and triangle show $\mu_z$ values for the Milky
342: Way and M31, estimated from their $V$-band peaks (Harris 2001) by applying an
343: average $(V-z)$ color estimated from the PEGASE population-synthesis code
344: (Fioc \& Rocca-Volmerange 1997).
345: ({\it Bottom}) Turnover {\it mass} ${\cal M}_{\rm TO}$ corresponding to the
346: fitted $\mu_z$, obtained by applying an average GC $\Upsilon_z$ computed for
347: each galaxy using the PEGASE model. Typical errorbars on ${\cal M}_{\rm TO}$
348: as a function of galaxy magnitude are indicated.
349: \label{fig:mu_B}}
350: \end{figure}
351:
352:
353: Figure~\ref{fig:mu_B} shows that there is no strong or systematic
354: variation in $\mu_z$ or
355: ${\cal M}_{\rm TO}$ to match that seen for $\sigma_z$
356: (Figure~\ref{fig:sigma_B}). Nevertheless,
357: there is a clear tendency for the GCLF turnovers of
358: galaxies fainter than $M_{B,{\rm gal}}\ga -18.5$ to scatter to somewhat
359: fainter (less massive) values than is typical of the bright giants.
360: The difference in mass is a factor of $\approx\!1.5$ on average,
361: but it ranges apparently randomly, from a factor of 1 (i.e., no
362: difference) up to factors slightly greater than 2 in some cases.
363: Note that there is a healthy mix of E and S0 or dE and dS0 galaxies
364: at all magnitudes in our ACSVCS sample (see Table 1 of C\^ot\'e et al.~2004).
365: We find no tendency for any particular Hubble type to scatter
366: preferentially away from $\mu_z=-8.4$ or ${\cal M}_{\rm TO} =
367: 2.2 \times 10^5 M_\odot$ in Figure \ref{fig:mu_B}.
368:
369:
370: The lower mean value for ${\cal M}_{\rm TO}$ at faint $M_{B,{\rm gal}}$
371: clearly can impact the use of the GCLF as a standard candle for dwarf
372: galaxies. On the other hand, the effect is wavelength-dependent. Publicly
373: available codes such as PEGASE can be used easily to show that in bluer
374: bandpasses such as $g$ (or the closely related $V$, which is more standard
375: for such studies), the slight decrease we find for the average GC
376: turnover {\it mass} in fainter galaxies is balanced by a comparable
377: decrease in the typical GC mass-to-light ratio (because of the lower
378: cluster metallicities), so that the mean turnover {\it magnitude} does not
379: vary as in the $z$ band. We have also confirmed this directly from our own
380: ACSVCS data. In J06, we obtain plots analogous to Figures \ref{fig:sigma_B}
381: and \ref{fig:mu_B} from fits to the $g$-band GCLFs of our galaxies. The
382: results fully support all of our conclusions here. It is
383: particularly worth noting that we find a relation identical to equation
384: (\ref{eq:sigmaz}) for the dependence of $g$-band GCLF dispersion on parent
385: galaxy luminosity.
386:
387: \section{Discussion}
388: \label{sec:discussion}
389:
390: An obvious question prompted by Figure \ref{fig:sigma_B} is whether the
391: correlation between $\sigma_z$ and $M_{B,{\rm gal}}$ was established
392: at the time of cluster formation or built up afterwards as GCLFs were
393: modified by the dynamical destruction of GCs over a Hubble
394: time. We favor the first interpretation.
395:
396: Star clusters can be destroyed over Gyr timescales as a result of
397: mass loss driven by stellar evolution, dynamical friction, gravitational
398: shocks, and internal two-body
399: relaxation (evaporation) --- processes that have been studied in
400: detail by several
401: groups. Recent discussions, centered specifically on how these affect
402: the GCLF, can be found in Fall \& Zhang (2001) and Vesperini (2000, 2001).
403: Fall \& Zhang in particular show that, while stellar evolution and
404: gravitational shocks certainly deplete the total number of GCs in a
405: galaxy, they do
406: not significantly alter the overall shape of the GCLF. Evaporation, on the
407: other hand, {\it can} change the shape of the GCLF, but significantly so only
408: for cluster masses ${\cal M}\la 2$-$3\times10^5\,{\cal M}_\odot$, i.e.,
409: below the typical GCLF turnover mass.
410:
411: In the theoretical treatments of Fall \& Zhang, Vesperini, and many
412: others, the evaporation rate is independent of cluster mass, which ultimately
413: drives the {\it low-mass} side of the GCLF to a universal
414: shape: a simple exponential $dN/dz \propto 10^{-0.4\,z}$ for the
415: number of GCs per unit magnitude fainter than the turnover (equivalent to
416: a flat distribution for the number of GCs per unit linear luminosity or
417: mass). But fitting a Gaussian model to the GCLF, as we have done
418: here, tacitly assumes that the distribution is symmetric. The
419: results in Figure \ref{fig:sigma_B} might therefore seem to imply that
420: both the bright side {\it and} the faint side of the GCLF become progressively
421: steeper in fainter galaxies.
422: However, various observational uncertainties make it difficult to
423: determine precisely the form of the faintest tail of the GCLF. Thus, in J06
424: we show that good fits to our GCLFs can also be obtained using an alternate
425: model with a universal exponential shape at
426: magnitudes fainter than the turnover --- and that the downward scatter in
427: ${\cal M}_{\rm TO}$ for faint galaxies persists in such a model and so is
428: not an artifact of any assumed Gaussian symmetry. Here we concern
429: ourselves only with the brighter half of the GCLF, which is observationally
430: better defined.
431:
432: We have performed maximum-likelihood fits of
433: exponential models $dN/dz \propto 10^{0.4 (\beta_z-1)\,z}$ (corresponding to
434: power-law {\it mass} distributions, $dN/d{\cal M}\propto {\cal M}^{-\beta_z}$)
435: to the GCLFs at absolute magnitudes $-8.7 \ga z \ga -10.8$
436: (cluster masses $\simeq\! 3\times10^5$--$2\times10^6\ {\cal M}_\odot$)
437: in 66 of our galaxies. Such distributions accurately describe the
438: bright sides of giant-galaxy GCLFs (Harris \& Pudritz 1994;
439: Larsen et al. 2001), and with $\beta_z\simeq2$ they also give good
440: matches to the mass functions of young star clusters in nearby mergers
441: and starbursts (Zhang \& Fall 1999).
442:
443: Figure~\ref{fig:power} shows the results of this exercise.
444: There is a clear steepening in the
445: power-law exponent, from $\beta_z\simeq 1.8$ in bright galaxies to
446: $\beta_z\simeq3$ in the faintest systems. However the faint side of
447: the GCLF behaves in detail, the bright side alone suggests that smaller
448: galaxies were unable to form very massive clusters in the same {\it relative}
449: proportions as giant galaxies.
450:
451: \begin{figure}
452: \epsscale{1.0}
453: \plotone{f3.eps}
454: \caption{Slope of the power law that best fits our $z$-band GCLF data,
455: $\beta_z$, for masses $3\times 10^5 \lae ({\cal M}/{\cal M}_\odot) \lae 2\times
456: 10^6$, plotted against host galaxy absolute magnitude, $M_{B,{\rm gal}}$.
457: %
458: The star and triangle show $\beta$ values for the Milky Way and M31 respectively,
459: measured in the same mass regime using the data from Harris (1996) and Reed et~al. (1994)
460: assuming a $V$-band mass-to-light ratio $M/L_V = 2$.
461: %
462: The bright
463: side of the GCLF is steeper in fainter galaxies.
464: \label{fig:power}}
465: \end{figure}
466:
467:
468:
469: A potential complication here is dynamical friction. A
470: cluster of mass ${\cal M}$ on an orbit of radius $r$ in a galaxy with
471: circular speed $V_c$ will spiral in to the center of the galaxy on a
472: timescale $\tau_{\rm df} \propto {\cal M}^{-1} r^2 V_c$ (Binney \& Tremaine
473: 1987). In the Milky Way and larger galaxies, $\tau_{\rm df} > 13$~Gyr
474: for all but the very most massive clusters at small radii, and thus
475: dynamical friction does not significantly affect
476: their GCLFs (e.g., Fall \& Zhang 2001). In
477: dwarfs with low $V_c$, however, $\tau_{\rm df}$ can be interestingly
478: short for smaller GCs at larger $r$ --- suggesting, perhaps,
479: that the process might significantly deplete the bright side of the GCLF
480: in small galaxies and contribute to the type of
481: trend seen in Figure~\ref{fig:power}. However, Vesperini (2000, 2001) has
482: modeled the GCLF evolution over a Hubble time in galaxies with a wide range
483: of mass, and his results strongly suggest that dynamical
484: friction does {\it not} suffice to explain our observations. In
485: particular, the widths of the Gaussian GCLFs in his models do not decrease,
486: even in dwarf galaxies, to anywhere near the extent seen in the data.
487: Thus, any significant galaxy-to-galaxy variations in the shape of the GCLF
488: above the turnover mass probably reflect initial conditions
489: (see J06 for further discussion).
490:
491: In summary, the gradual narrowing of the GCLF as a function of galaxy
492: luminosity --- or the steepening of the
493: mass distribution above the classic turnover point --- presents
494: a new constraint for theories of GC formation and evolution.
495: In our view, it is the cluster formation process in
496: particular that is likely to be most relevant to the observed behavior at
497: the high-mass end of the GCLF. Exactly what factors might lead to more
498: massive galaxies forming massive clusters in greater relative numbers,
499: is an open question of some interest.
500:
501: \acknowledgements
502:
503: Support for program GO-9401 was provided
504: through a grant from the Space
505: Telescope Science Institute, which is operated by the Association of
506: Universities for Research in Astronomy, Inc., under NASA contract NAS5-26555.
507:
508: {\it Facility:} \facility{HST (ACS/WFC)}
509:
510:
511:
512: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
513: %%%%%%%%%%%%%%%%%%%%%%%%%%%%% BIBLIOGRAPHY %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
514: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
515:
516: \begin{thebibliography}{00}
517: \bibitem[Binney \& Tremaine 1987]{bt87} Binney, J.,
518: \& Tremaine, S. 1987, Galactic Dynamics (Princeton:
519: Princeton Univ. Press)
520: \bibitem[C\^ot\'e et~al. 2004]{acsvcs_i} C\^ot\'e, P., et~al.\
521: 2004, \apjs, 153, 223
522: \bibitem[de Vaucouleurs \& Pence 1978]{devp78} de Vaucouleurs, G., \& Pence,
523: W. D. 1978, \aj, 83, 1163
524: \bibitem[Fall \& Zhang 2001]{fz01} Fall, S.M., \& Zhang, Q. 2001,
525: \apj, 561, 751
526: \bibitem[Fioc \& Rocca-Volmerange 1997]{pegase} Fioc, M. \& Rocca-Volmerange,
527: B. 1997, \aap, 326, 950
528: \bibitem[Harris 1991]{harris91} Harris, W.E. 1991, \araa, 29, 543
529: \bibitem[Harris 1996]{harris96} Harris, W.E. 1996, \aj, 112, 1487
530: \bibitem[Harris 2001]{harris01} Harris, W.E. 2001, in Star Clusters
531: (28th Saas-Fee Advanced Course) ed. L. Labhardt \& B. Binggeli (Berlin:
532: Springer), 223
533: \bibitem[Harris \& Pudritz(1994)]{harris94} Harris, W. E., \& Pudritz,
534: R. E. 1994, \apj, 429, 177
535: \bibitem[Jord\'an et~al. 2004]{acsvcs_ii} Jord\'an, A. et~al.\
536: 2004, \apjs, 154, 509
537: \bibitem[Jord\'an et~al. 2006]{acsvcs_xii} Jord\'an, A. et~al.\
538: 2006, submitted to \apjs\ (J06)
539: \bibitem[Kennicutt 1983]{kennicutt83} Kennicutt, R. C. 1983, \apj, 272, 54
540: \bibitem[Kent 1989]{kent89} Kent, S. M. 1989, \aj, 97, 1614
541: \bibitem[Kundu \& Whitmore 2001]{kunduwhi01a} Kundu, A., \& Whitmore, B.C.
542: 2001, \aj, 121, 2950
543: \bibitem[Larsen et al.(2001)]{larsen01} Larsen, S. S., Brodie, J. P., Huchra,
544: J. P., Forbes, D. A., \& Grillmair, C. J. 2001, \aj, 121, 2974
545: \bibitem[Mei et~al. 2005]{mei05} Mei, S., et~al.\ 2005, \apj, 625, 121
546: \bibitem[Mei et~al. 2006]{mei06} Mei, S., et~al.\ 2006, submitted to \apj
547: \bibitem[Peng et~al.\ 2006a]{peng06a} Peng, E.W. et~al.\ 2006a, \apj, 639, 95
548: \bibitem[Peng et~al.\ 2006b]{peng06b} Peng, E.W. et~al.\ 2006b, \apj, 639, 838
549: \bibitem[Reed et~al.\ 1994]{reed+94} Reed, L.G., Harris, G.L.H., \& Harris,
550: W.E. 1994, \aj, 107, 555
551: \bibitem[Schechter 1976]{schech76} Schechter, P.L. 1976, \apj, 203, 297
552: \bibitem[Secker \& Harris 1993]{secker+harris93} Secker, J., \& Harris, W.E.
553: 1993, \aj, 105, 1358
554: \bibitem[Sirianni et~al. 2005]{sirianni+05} Sirianni, M., et~al. 2005, \pasp,
555: 117, 1049
556: \bibitem[Vesperini 2000]{vesperini00} Vesperini, E. 2000, \mnras, 318, 841
557: \bibitem[Vesperini 2001]{vesperini01} Vesperini, E. 2001, \mnras, 322, 247
558: \bibitem[Zhang \& Fall(1999)]{zhang99} Zhang, Q., \& Fall, S. M. 1999, \apj,
559: 527, L81
560: \end{thebibliography}
561:
562: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
563: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% FIGURES %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
564: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
565:
566:
567: \end{document}
568: