1: \documentclass[manuscript]{aastex}
2: \usepackage{emulateapj5,natbib}
3:
4: \citestyle{aa}
5:
6:
7:
8: %%%%%%%%%%%%%%
9: % Title Page %
10: %%%%%%%%%%%%%%
11:
12: \slugcomment{accepted by the Astrophysical Journal}
13:
14: \shortauthors{McCRADY ET AL.}
15: \shorttitle{SUPER--STAR CLUSTER M82-F}
16:
17: \begin{document}
18:
19: \title{ MASS SEGREGATION AND THE INITIAL MASS FUNCTION OF SUPER--STAR
20: CLUSTER M82-F\altaffilmark{1} }
21:
22: \author{
23: Nate McCrady\altaffilmark{2,3}, James R. Graham\altaffilmark{3} and William D. Vacca\altaffilmark{4}
24: }
25: \altaffiltext{1}{
26: Based on observations made at the W.M. Keck Observatory, which is operated
27: as a scientific partnership among the California Institute of Technology,
28: the University of California and the National Aeronautics and Space
29: Administration. The Observatory was made possible by the generous
30: financial support of the W.M. Keck Foundation.}
31:
32: \altaffiltext{2}{
33: nate@astro.berkeley.edu}
34:
35: \altaffiltext{3}{
36: Department of Astronomy, University of California, Berkeley, CA 94720-3411}
37:
38: \altaffiltext{4}{
39: NASA Ames Research Center, Moffett Field, CA, 94035}
40:
41:
42: %%%%%%%%%%%%
43: % Abstract %
44: %%%%%%%%%%%%
45:
46: \begin{abstract}
47:
48: We investigate the Initial Mass Function and mass segregation in
49: super--star cluster M82-F with high resolution Keck/NIRSPEC echelle
50: spectroscopy. Cross-correlation with template supergiant spectra
51: provides the velocity dispersion of the cluster, enabling measurement
52: of the kinematic (virial) mass of the cluster when combined with sizes
53: from NICMOS and ACS images. We find a mass of $6.6 \pm 0.9 \times
54: 10^5 M_{\odot}$ based on near-IR light and $7.0 \pm 1.2 \times 10^5
55: M_{\odot}$ based on optical light. Using PSF-fitting photometry, we
56: derive the cluster's light-to-mass ($L/M$) ratio in both near-IR and
57: optical light, and compare to population synthesis models. The ratios
58: are inconsistent with a normal stellar initial mass function for the
59: adopted age of 40--60 Myr, suggesting a deficiency of low-mass stars
60: within the volume sampled. King model light profile fits to new
61: HST/ACS images of M82-F, in combination with fits to archival near-IR
62: images, indicate mass segregation in the cluster. As a result, the
63: virial mass represents a lower limit on the mass of the cluster.
64: \end{abstract}
65:
66: \keywords{galaxies: individual (M82) --- galaxies: starburst ---
67: galaxies: star clusters --- infrared: galaxies }
68:
69:
70: %%%%%%%%
71: % Body %
72: %%%%%%%%
73:
74: \section{Introduction}
75: \label{intro}
76:
77: Star formation in starburst galaxies can be resolved into young,
78: dense, massive ``super--star clusters'' (SSCs) that represent a
79: substantial fraction of new stars formed in a burst event
80: \citep{meurer95, zepf99}. SSCs identified from the ground two decades
81: ago \citep[e.g.,][]{arp85,melnick85} have been spatially resolved in
82: the nearest cases by the {\it Hubble Space Telescope} (HST) during the
83: last decade using WFPC in visible light
84: \citep[e.g.,][]{o'connell94,whitmore95}.
85:
86: Optical studies of SSCs in dusty star forming regions are hampered by
87: high extinction. The development over the past decade of new high
88: spatial resolution near-infrared imaging and high spectral resolution
89: spectroscopic instruments, including NICMOS aboard HST and the NIRSPEC
90: spectrometer at the W. M. Keck Observatory, has led to a wealth of new
91: data on SSCs. In the optical, the Advanced Camera for Surveys (ACS)
92: imager on HST has substantially improved resolution over the standards
93: set by WFPC2. These new instruments have enabled direct measurement
94: of structural parameters and determination of virial masses for even
95: heavily reddened, optically-invisible clusters.
96:
97: In addition to being a significant mode of star formation in starburst
98: galaxies, SSCs are important for study of the formation of high mass
99: stars, galaxy stellar populations and evolution, and physical
100: conditions analogous to high-redshift star formation. Numerical
101: simulations by \citet{zwart04} further suggest that runaway mergers of
102: high mass stars deep in the potential well of dense SSCs lead to the
103: creation of intermediate-mass black holes, a possible missing link
104: between stellar mass black holes and the supermassive black holes
105: found in the centers of galaxies. Especially important is the stellar
106: initial mass function (IMF), for which SSCs are ideal laboratories:
107: coeval populations with enough stars to extensively sample the IMF.
108: Critical to understanding the IMF is detection of low-mass stars, the
109: light of which is swamped by high luminosity supergiants.
110: Measurement of the kinematic mass of the clusters represents the only
111: means of detecting and quantifying the contribution of low-mass stars.
112:
113: At a distance of $3.6 \pm 0.3$ Mpc \citep{freedman94}, M82 is the
114: nearest massive starburst galaxy. \citet{lipscy04} find that at least
115: 20\% of the star formation in M82 is occuring in SSCs. The galaxy's
116: high inclination of 81$^\circ$ \citep{achtermann95} and prevalent dust
117: lead to large, patchy extinction; near-IR observations are required to
118: overcome this obstacle in characterizing the SSC population. The
119: nuclear starburst in M82 contains over 20 SSCs that are prominent in
120: the near-infrared, with typical ages of $\sim 10^7$ years
121: \citep{natascha98}. These young, massive clusters offer an excellent
122: opportunity to search for variations in the IMF.
123:
124: Early studies \citep{rieke80, mcleod93} of the IMF in the M82
125: starburst used ground-based observations and were necessarily global
126: in scale; these studies indicated an abnormal IMF for the starburst.
127: An influential work by \citet{rieke93} used population synthesis
128: models to constrain the IMF. They concluded that the large $K$-band
129: luminosity of the M82 starburst relative to the dynamical mass
130: required the IMF to be significantly deficient in low-mass stars ($M <
131: 3M_{\odot}$) relative to the solar neighborhood.
132:
133: Local examination of star formation regions on the scale of individual
134: SSCs provides an important test for the assertion of an abnormal,
135: ``top-heavy'' IMF. \citet{satyapal97} used $1''$-resolution near-IR
136: images to identify pointlike sources and found that at this scale
137: starburst models could match observations without invoking a high-mass
138: biased IMF. \citet{mccrady03} used high-resolution ($0.''2$)
139: HST/NICMOS images and near-IR Keck/NIRSPEC spectroscopy to measure the
140: kinematic (virial) masses of two individual near-IR-bright super--star
141: clusters in the nuclear starburst. Comparison of derived
142: light-to-mass ratios and population synthesis models indicate that
143: cluster MGG-9 has a normal IMF, while MGG-11 may be deficient at the
144: 1-$\sigma$ level in low-mass stars ($M<1$ M$_{\odot}$). This work
145: indicates that the IMF may vary on the scale of individual clusters,
146: and bears on the ongoing debate over the role of physical conditions
147: in a star formation environment in determining the mass distribution
148: function of newly-formed stars.
149:
150: The brightest optically-visible cluster within 500 pc of the center of
151: the galaxy is M82-F. \citet[][hereafter SG01]{smith01} studied the
152: cluster in detail using optical spectra and HST/WFPC2 images to
153: estimate the cluster's virial mass. Based on population synthesis
154: models, they determined that M82-F is too luminous for its assumed
155: age, and must therefore have a top-heavy IMF. With most of its mass
156: in stars of $M > 2 $M$_{\odot}$ and a deficiency of long-lived,
157: low-mass stars, the authors suggested that the cluster is ``doomed''
158: and will begin to dissolve due to stellar evolution over the next 1--2
159: Gyr. We examine some of the limitations of their data in Section
160: \ref{acsdata}. SG01 noted the possibility that their data could
161: instead reflect mass segregation, which would lead to underestimation
162: of the cluster's mass. Using NICMOS images, \citet{mccrady03}
163: measured a smaller radius for M82-F in near-IR images, suggesting that
164: observed light-to-mass ratios may be systematically overestimated due
165: to mass segregation.
166:
167: Mass segregation cannot be observed directly via star counts for
168: clusters in which individual stars are unresolved. However,
169: \citet{sternberg98} notes that a mass-segregated cluster should appear
170: smaller in the UV and IR (light dominated by red supergiant stars)
171: than in the optical (where the light is dominated by intermediate-mass
172: main sequence stars).
173:
174: In this paper, we characterize the IMF of the super star cluster M82-F
175: and investigate indications of mass segregation. We measure the
176: cluster stellar velocity dispersion using new high spatial resolution
177: Keck/NIRSPEC echelle spectra and present analysis of new,
178: high-resolution, multi-waveband optical images of the cluster from the
179: High Resolution Camera on the HST Advanced Camera for Surveys (ACS).
180: As discussed in Section \ref{analysis}, these new images represent a
181: significant improvement in the characterization of the physical
182: structure of M82-F. We fit two-dimensional light profiles to the
183: cluster in the ACS and archival HST/NICMOS images, and derive the
184: kinematic mass and light-to-mass ratio. Section \ref{obs} describes
185: the observations used in this work, Section \ref{analysis} presents
186: analysis of the data and determination of key cluster parameters, and
187: Section \ref{discussion} discusses mass segregation and the form of
188: the IMF in M82-F.
189:
190: \vspace{0.5in}
191:
192: \section{Observations}
193: \label{obs}
194:
195: We observed super star cluster M82-F (MGG-F) with the 10-m Keck II
196: telescope on 2003 January 19, using the facility near-infrared echelle
197: spectrometer NIRSPEC. We obtained high-resolution ($R \sim 22,000$),
198: cross-dispersed spectra in the wavelength range 1.51--1.75 $\mu$m
199: using the NIRSPEC-5 order-sorting filter. The data fall in seven
200: echelle orders, ranging from $m=44$ through 50. The cluster was
201: nodded along the $0.''432$ $\times$ $24''$ slit, with successive nods
202: separated by $\sim 10$ arcsec (seeing was $\sim 0.''8$). Seven
203: separate integrations, each 600 seconds in duration, provided a total
204: time of 4200 seconds on the cluster. The spectra were dark
205: subtracted, flat-fielded and corrected for cosmic rays and bad pixels.
206: The curved echelle orders were then rectified onto an orthogonal
207: slit-position/wavelength grid based on a wavelength solution from sky
208: (OH) emission lines. Each pixel in the grid has a width of
209: $\delta\lambda = 0.019$ nm. We sky-subtracted by fitting third-order
210: polynomials to the two dimensional spectra column-by-column.
211:
212: The spectra were extracted using Gaussian weighting functions matched
213: to the wavelength-integrated cluster profile. Atmospheric calibration
214: was performed using the spectrum of HD 173087, a B5V star. The
215: calibration star was observed at an airmass of 1.94, consistent with
216: the airmass range (1.76--2.07) of the M82-F observations. To account
217: for photospheric absorption features (particularly Brackett series and
218: helium lines) and continuum slope, the calibration star spectrum was
219: divided by a spline function fit. The resulting atmospheric
220: absorption spectrum was divided into the cluster spectra to recover
221: absorbed flux. The individual extracted spectra have an average
222: signal-to-noise ratio (S/N) of $\sim 25$ per pixel and S/N $\sim 66$
223: per pixel when all integrations are combined. For comparison, the
224: optical spectra used in SG01 have S/N per pixel of 15--25.
225:
226: For photometric measurements and structural parameters, we obtained
227: new high-resolution optical images of M82-F. The images were taken
228: 2002 June 7 with the ACS High Resolution Camera (HRC) in the F435W,
229: F555W and F814W filters (corresponding approximately to the Johnson
230: $B$, $V$ and $I$ filters). Exposure times for the images are 1320
231: seconds, 400 seconds, and 120 seconds, respectively. Additional
232: observations in the F250W filter were taken on 2003 Jan 20, with
233: exposure time of 10,160 seconds. ACS/HRC has a plate scale of 25 mas
234: per pixel and is thus critically sampled at 477 nm. We relied on
235: pipeline reduction (CALACS) for bias correction, dark subtraction and
236: flatfielding. The resulting set of ``FLT'' images were processed with
237: the Multidrizzle package \citep{koekemoer02} to reject cosmic rays,
238: mask bad pixels and create a ``drizzled'' image for each filter,
239: corrected for geometric distortions.
240:
241: In the near-IR, we used archival HST/NICMOS images in the F160W and
242: F222M filters. For details on these data and their reduction, see
243: \citet{mccrady03}.
244:
245:
246:
247:
248: \section{Analysis}
249: \label{analysis}
250:
251: \subsection{Stellar Velocity Dispersion}
252:
253: To derive the kinematic mass of a cluster using the virial theorem, we
254: need a measurement of the internal velocity dispersion in the cluster
255: and a characteristic radius. Figure \ref{spect} shows a portion of
256: the NIRSPEC echelle spectrum of M82-F, compared to a series of
257: template supergiant spectra. Features found in the spectra of
258: supergiant stars are readily identified in the cluster spectrum,
259: although they appear ``washed out'' due to the stellar velocity
260: dispersion. Especially prominent are the rovibrational CO bandheads
261: and numerous Fe and OH lines. We have assembled an atlas of
262: high-resolution NIRSPEC spectra of supergiant stars for use in
263: cross-correlation analysis \citep{mccrady03}, from which we determine
264: the dominant spectral type and line-of-sight velocity dispersion.
265: Based upon the peak of the cross-correlation function (Figure
266: \ref{ccfplot}), the $H$-band spectrum of M82-F most closely matches
267: spectral types in the range of K4I--M0I. The line-of-sight velocity
268: dispersion based on cross-correlation with templates in this range is
269: $\sigma_r = 13.5 \pm 0.2$ km s$^{-1}$. The value of $\sigma_r$
270: decreases as a function of the similarity to the template spectrum as
271: measured by the peak value of the cross-correlation function. It is
272: therefore possible that the stated value reflects some template
273: mismatch bias due to our limited template spectra atlas, but this
274: effect is small ($<$ 0.5 km s$^{-1}$).
275:
276: \subsection{Cluster Size and Mass}
277:
278: The simplest approach to determine the cluster mass is to measure the
279: half-light radius, assume that light traces mass, the cluster is
280: spherical, and the velocity dispersion is isotropic, then apply the
281: virial theorem. M82-F presents a difficulty for this method: HST
282: images (Figure \ref{fitplot}) clearly show that the cluster is
283: elliptical in projection and cannot, therefore, be spherical. To
284: measure the radius, we fit the cluster using an elliptical version of
285: the empirical \citet{king62} model:
286:
287: \begin{equation}
288: f(x,y) = k_1 \left\{ \left( 1+\frac{x^2}{a^2}+\frac{y^2}{b^2}\right)^{-1/2}
289: - (1+k_2^2)^{-1/2} \right\}^2
290: \end{equation}
291:
292: \noindent where $a$ and $b$ are the characteristic lengths of the
293: minor and major axes, respectively, $k_1$ is a scaling constant and
294: $k_2$ acts as a ``tidal radius,'' truncating the profile beyond a
295: particular scale length. The fit therefore has four free parameters
296: ($k_1$, $k_2$, $a$ and $b$) to describe the light profile and three
297: more to describe the centroid location and position angle. The King
298: model is convolved with a model PSF from Tiny Tim \citep{krist95} and
299: compared to the image. Fit parameters are determined by iterative
300: search over parameter space, using a Levenburg-Marquardt least-squares
301: fit. Figure \ref{fitplot} shows the fit and residuals for the ACS/HRC
302: F814W image of M82-F.
303:
304: The half-light radius in projection along the major axis, $r_{hp}$, is
305: the semimajor axis of the ellipse that encloses half the flux in the
306: fitted King model. We determine $r_{hp}$ numerically by summing the
307: flux in a series of ellipses with the axial ratio defined by $a$ and
308: $b$. Monte Carlo simulations of clusters indicate that $a$, $b$ and
309: $k_2$ are significantly covariant; however, the fitted projected
310: half-light radius along a given axis is accurate to about 2 percent.
311: The half-mass radius may be determined by assuming that light traces
312: mass and deprojecting by dividing $r_{hp}$ by 0.766 \citep{spitzer87}.
313:
314: \citet{mccrady03} fit a spherical King model to M82-F in the NICMOS
315: F160W image and found a projected half-light radius of $89 \pm 11$
316: mas. Our elliptical King function fit to the same image found
317: $r_{hp}=113 \pm 2$ along the major axis and $r_{hp}=62 \pm 1$ along
318: the minor axis. At the adopted distance of M82, the projected
319: half-light radii along the major and minor axes are thus $1.97 \pm
320: 0.16$ pc and $1.1 \pm 0.1$ pc, respectively.
321:
322: To account for the ellipticity of the cluster, we assume it is an
323: oblate spheroid and compute the gravitational potential assuming that
324: the cluster is homoeoidal \citep[][Section 2.3]{binney87}. The
325: numerical constant in the virial mass formula may be separated into
326: factors dependent upon the central concentration and the ellipticity.
327: The compact core plus extended envelope structure of star clusters is
328: similar to an $n=5$ polytrope \citep[][p. 13]{spitzer87}. For an
329: $n=5$ polytropic oblate ellipsoid with an isotropic velocity
330: dispersion, the virial mass is:
331:
332: \begin{equation}
333: M = 10.31 \, (e/\sin^{-1}{e}) \, r_{hp} \, \sigma_r^2 /G
334: \end{equation}
335:
336: \noindent where the eccentricity is $e = \sqrt{1-(Z/R)^2}$, and $R$
337: and $Z$ are the equatorial and polar radii of the oblate spheroid,
338: respectively. Figure \ref{halfradii} shows the fitted cluster
339: half-light radii in projection along the major and minor axes for the
340: NICMOS and ACS images. The observed axial ratio, $a/b$, represents a
341: lower bound on the ratio $Z/R$; an oblate cluster will appear rounder
342: than its intrinsic shape unless viewed directly along the equatorial
343: plane. Averaging over all possible inclinations, we find that the
344: observed axial ratio of $0.55$ corresponds to a most-likely intrinsic
345: axial ratio of $0.35$ with $e=0.77 \pm 0.07$. We represent the
346: uncertainty on the viewing inclination as the difference between the
347: angle-averaged value and the lower bound.
348:
349: We believe the ellipticity of the cluster is intrinsic, rather than
350: the result of differential reddening. Extinction is very low at 2.2
351: $\mu$m, and therefore the observed shape in the F222M images is very
352: likely intrinsic. Moreover, the axial ratio is constant within the
353: uncertainties across all wavebands from 0.4 to 2.2 $\mu$m (Figure
354: \ref{halfradii}). If the ellipticity were due to the distribution of
355: dust around the cluster, the axial ratio should change as a function
356: of wavelength as the extinction is expected to vary significantly
357: between $B$ and $K$. The axial ratio does appear to increase sharply
358: at the shortest HST/ACS waveband (F250W). We interpret this as a
359: result of scattering of cluster light by dust outside the cluster.
360: The ACS images show strong spatial variations in extinction in the
361: immediate vicinity of M82-F. Scattering and absorption in the
362: ultraviolet compromise the quality of the fit to the light profile at
363: very short wavelengths.
364:
365: Combining the NIRSPEC velocity dispersion with the NICMOS radius and
366: estimated eccentricity, we calculate an $H$-band virial mass of $M_H =
367: 6.6 \pm 0.9 \times 10^5$ M$_{\odot}$ for M82-F. The subscript on the
368: mass is used to emphasize that the virial mass is computed from a
369: velocity dispersion and size measured in the $H$ band. In this manner
370: we minimize systematic errors by measuring both variables from the
371: light of the same stars and may seek variations between different
372: wavebands.
373:
374: \subsection{Optical Results \& Photometry}
375:
376: SG01 found a velocity dispersion of $13.4 \pm 0.7$ km s$^{-1}$ for
377: M82-F based on cross-correlation analysis using cluster spectra in the
378: 601--759 nm range (overlapping the ACS F814W bandpass). They
379: concluded that the cluster spectrum most closely matches their
380: template K stars, consistent with the Starburst99 prediction that K II
381: stars dominate the spectrum at the inferred cluster age (discussed in
382: Section \ref{age}). We adopt these results to measure the cluster
383: mass with optical data.
384:
385: In the ACS F814W image, we measure a half-light radius of $119 \pm 2$
386: mas. Using this value and the SG01 velocity dispersion, we find an
387: $I$-band virial mass of $M_I = 7.0 \pm 1.2 \times 10^5$ M$_{\odot}$
388: for M82-F. This is significantly lower than the optically-derived
389: SG01 mass estimate of $1.2 \pm 0.1 \times 10^6$ M$_{\odot}$ due to
390: more accurate measure of the cluster's half-light radius (see Section
391: \ref{acsdata}).
392:
393: Photometry was derived by integrating over the fitted King models and
394: applying the appropriate conversion factors for each instrument and
395: filter. Details of the PSF fitting photometry and spectral energy
396: distribution are presented in a companion paper \citep{vacca04}. The
397: observed cluster in-band luminosities as defined in \citet{mccrady03},
398: not corrected for extinction, are $L_{F814} = 3.8 \pm 0.6 \times 10^5$
399: L$_{\odot}$ and $L_{F160} = 4.0 \pm 0.7 \times 10^5$~L$_{\odot}$.
400:
401:
402:
403: \section{Discussion}
404: \label{discussion}
405:
406: By comparing the derived light-to-mass ($L/M$) ratio of M82-F to
407: population synthesis models, we may characterize the IMF of the
408: cluster. Critical to this analysis are the cluster age and
409: line-of-sight extinction.
410:
411: \subsection{Age}
412: \label{age}
413:
414: The spectrum of M82-F immediately places upper and lower bounds on the
415: cluster's age. There is no evidence of nebular emission in the $H$-
416: or $K$-band, which indicates the absence of O stars and a minimum
417: cluster age of 6--7 Myr. \citet{natascha98} finds none of the
418: features expected of AGB stars for the nuclear region of M82, setting
419: an upper limit of $\sim 10^8$ years. For more precise limits on the
420: age, we turn to population synthesis modelling. As noted in Section
421: \ref{analysis}, the F160W light of M82-F most closely matches template
422: spectral types in the range of K4I--M0I. \citet{gilbert02th} used
423: Starburst99 population synthesis models to determine the flux-weighted
424: average spectral type as a function of age for coeval stellar
425: populations. In the $H$-band, the cluster light is dominated by
426: K4--M0 stars for a brief period around 15 Myr and during the ages of
427: $\sim$ 40--60 Myr. SG01 used Starburst99 models to fit the H$\delta$
428: and He I absorption profiles in optical spectra. They found that the
429: best fits to the wings of the Balmer line and depth of the helium line
430: suggest an age of $60 \pm 20$ Myr. This is consistent with the
431: dominant spectral type of the $H$-band light, and we therefore adopt
432: an age range of 40--60 Myr for M82-F.
433:
434: \subsection{Extinction}
435: \label{extinction}
436:
437: SG01 used $BVI$ photometry to determine line-of-sight extinction of
438: $A_V = 2.8$. Applying the \citet{cardelli89} extinction law ($R_V =
439: 3.1$) to the this value gives $A_{F160W} = 0.53$ and $A_{F814W} =
440: 1.63$.
441:
442: As an independent test, we calculate synthetic colors for the near-IR
443: dominant evolved K4--M0 stars based on the \citet{pickles98} stellar
444: spectral library. The synthetic [F160W] $-$ [F222M] color for a K4I
445: star is 0.35, versus 0.52 for M0I. From our photometry, we find
446: [F160W] $-$ [F222M] = $0.36 \pm 0.04$ for M82-F, at the blue end of
447: the expected color range. This implies that $H$-band extinction to
448: M82-F is quite small, essentially negligible.
449:
450: The optical (600-800 nm) light of the cluster is expected to be
451: dominated by KII stars (SG01). Synthetic [F555W] $-$ [F814W] colors
452: for these stars in the Pickles library range from 1.19 to 1.45, versus
453: the measured $1.71 \pm 0.04$ for M82-F. For a standard ($R_V = 3.1$)
454: interstellar extinction curve, this color excess gives $A_{F814W} =
455: 0.34$ to 0.66 mag.
456:
457: To calculate the light-to-mass ratio, we need to deredden the cluster.
458: We adopt extinction of $A_{F160W} = 0.0 ^{+0.1}_{-0.0}$ and $A_{F814W}
459: = 0.5 \pm 0.2$ based on the synthetic photometry results. For sake of
460: comparison, Figure \ref{sb99plot} reflects both these estimates and
461: the extinctions implied by the SG01 estimate of $A_V$.
462:
463: The Galactic dust map of \citet{schlegel98} indicates Galactic
464: extinction along the line of sight to M82-F of $A_V = 0.48$,
465: $A_{F814W} = 0.28$ and $A_{F160W} = 0.09$. This provides a
466: cross-reference for the estimated extinction.
467:
468:
469:
470: \subsection{The IMF}
471:
472: We determine the luminosity and virial mass of the cluster independent
473: of any assumptions about the IMF. This is in contrast to photometric
474: mass determinations, which are based on cluster colors and ages and
475: must therefore assume an IMF. Our method enables us to constrain the
476: cluster IMF by comparing observed light-to-mass ratios in various
477: wavebands to population synthesis models \citep{sternberg98,mccrady03}.
478:
479: By applying the adopted extinction corrections for M82-F, we find
480: de-reddened light-to-mass ratios of $L/M = 0.6 \pm 0.1$
481: (L$_\odot$/M$_\odot$) at 1.6$\mu$m and $L/M = 0.9 \pm 0.2$ at
482: 0.8$\mu$m. Figure \ref{sb99plot} compares the measured $L/M$ ratio to
483: population synthesis model predictions for two fiducial IMF forms. We
484: used Starburst99 version 4.0, with an instantaneous burst, solar
485: metallicity \citep{mcleod93} and the Hillier \& Pauldrach atmosphere
486: models. The derived $L/M$ ratio of M82-F is too high for the standard
487: \citet{kroupa01} IMF over the full range of stellar masses (0.1--100
488: M$_{\odot}$). Rather it appears that the cluster's IMF is deficient
489: in low-mass stars. For example, the $L/M$ ratio is roughly consistent
490: with a \citet{salpeter55} IMF truncated at a lower mass of about 2
491: M$_{\odot}$. If the cluster age were approximately 15 Myr, the
492: derived $L/M$ ratios would be consistent with the Kroupa IMF; an
493: independent determination of the cluster's age could rule out this
494: possibility.
495:
496: In our analysis we assume a particular form for the IMF (i.e., Kroupa
497: or Salpeter power laws) and modify the lower-mass cutoff to generate
498: the observed $L/M$ ratio at the adopted age. This method allows us to
499: test the top-heavy IMF hypothesis --- if the observed $L/M$ ratio is
500: fit with an IMF extending down to 0.1 M$_{\odot}$, as in cluster MGG-9
501: in \citet{mccrady03}, there is no need to invoke an abnormal IMF. In
502: the present case, a normal (Kroupa) IMF cannot explain the relatively
503: large $L/M$ ratio of the cluster. An elevated lower-mass cutoff is
504: not required, however. The observed $L/M$ ratios could be caused at
505: the cluster's age by flattening the slope of the IMF, which would
506: change the relative proportions of stellar masses. Although our data
507: do not distinguish between changes to the lower mass cutoff or IMF
508: slope, it is evident that the IMF is different than the Kroupa form,
509: and different from other nearby SSCs in the M82 nuclear starburst.
510:
511: A similar study has been conducted by \citet{mengel02} in the Antennae
512: (NGC 4038/4039). They determined virial masses and light-to-mass
513: ratios for a sample of five bright SSCs, and compared the results to
514: various IMF forms using population synthesis models. The clusters
515: studied by Mengel et al. appear to exhibit a range of IMFs, with some
516: evidence of dependence on location within the merger environment. In
517: contrast, \citet{larsen04} measured virial masses for five clusters in
518: nearby galaxies and found light-to-mass ratios consistent with a
519: Kroupa or Salpeter IMF. They found no evidence for a deficiency of
520: low-mass stars in these clusters.
521:
522:
523:
524: \subsection{Optical Images}
525: \label{acsdata}
526:
527: The ACS observations, designed specifically to study the properties of
528: M82-F, represent a significant advance in quality over all earlier
529: optical images. \citet{o'connell95} imaged the cluster with the
530: pre-repair mission Planetary Camera in 1992. The cluster fell at the
531: edge of or between chips on the detector in their $V$- and $I$-band
532: equivalent images, and they deemed their $V$-band deconvolution ``not
533: highly reliable.'' SG01 used archival WFPC2 images in the F439W,
534: F555W and F814W filters \citep{degrijs01}. These images were designed
535: for a separate study of M82, and were not ideal for M82-F. The
536: cluster fell on the WF4 CCD, and were significantly undersampled due
537: to the 100-mas pixels. The two longer-wavelength images were both
538: saturated in the cluster core. The ACS/HRC images used exposure times
539: specifically designed for study of M82-F, and with 25 mas pixels, are
540: critically sampled above 477 nm. SG01 did not deconvolve the PSF from
541: the WFPC2 F439W image, instead estimating the broadening
542: heuristically. By contrast, we treated the PSF more rigorously using
543: a model PSF as described in Section \ref{obs}. Both the
544: \citet{o'connell95} and SG01 studies found a projected half-light
545: radius of $160 \pm 20$ mas for M82-F, using images at 555 nm and 439
546: nm, respectively. Our fits to the major axis of the cluster in
547: ACS/HRC images yield significantly smaller projected half-light radii
548: of $120 \pm 2$ mas and $123 \pm 2$ mas at 555 nm and 435 nm,
549: respectively. These values are more precise and more accurate than
550: the radii determined from the lower-resolution optical data in earlier
551: works.
552:
553:
554: \subsection{Mass Segregation}
555:
556: The axial ratio of M82-F is $\sim 0.55$ in all but the shortest
557: wavelength image (Figure \ref{halfradii}). Fits to the images show a
558: distinct trend of decreasing cluster size with increasing wavelength.
559: Light at shorter wavelengths is increasingly dominated by hotter stars
560: still on the main sequence. In a coeval population, these stars will
561: be intermediate mass stars. Longer wavelength light is dominated by
562: cooler evolved stars, stars originally more massive than those at the
563: cluster's main sequence turnoff point. The negative correlation
564: between cluster size and observed wavelength suggests that the massive
565: red evolved stars that dominate the near-IR light are more centrally
566: concentrated than the intermediate-mass main sequence stars which
567: dominate the optical light \citep{sternberg98}.
568:
569: A key assumption of our method is that light traces mass within the
570: cluster. Mass segregation renders the interpretation of the
571: light-to-mass ratio problematic. The near-IR measurements presented
572: here trace the light of supergiant stars --- the most massive stars
573: currently present in the cluster. If there is mass segregation, the
574: near-IR virial measurement probes only the core of the cluster, i.e.,
575: mass contained within the volume populated by these highest-mass
576: stars. As such, the derived mass would be a lower limit and the IMF
577: of the entire cluster may follow the Kroupa form. In this case, the
578: IMF measured in the core would appear ``top-heavy'' because of mass
579: segregation. A nearby example of this effect is the young, massive
580: cluster R136 in the 30 Doradus nebula in the Large Magellanic Cloud
581: (LMC). \citet{brandl96} found that the mass function in R136 steepens
582: with increasing distance from the cluster center, indicating strong
583: mass segregation.
584:
585: At the adopted cluster age, stars more massive than $\sim 8$~M$_\odot$
586: have exploded as supernovae, and stars in the 6--8 M$_\odot$ range
587: have evolved off the main sequence \citep{schaller92}. If we assume
588: that the core consists solely of stars that are (and remnants of
589: progenitors which were) larger than 2 M$_\odot$ as indicated by the
590: $L/M$ ratio, integration of the Kroupa IMF implies that the measured
591: core mass represents only one-third of the current cluster mass. The
592: remaining stars (with masses smaller than 2 M$_\odot$) would be
593: distributed outside the core. This distribution of stars would thus
594: require a total cluster mass of $\sim 2 \times 10^6$ M$_{\odot}$ to be
595: consistent with a Kroupa IMF for the population of the entire cluster.
596:
597: Mass segregation is generally associated with the gradual
598: equipartition of energy via stellar encounters in old globular
599: clusters \citep{spitzer87}. High mass stars sink to the center of a
600: cluster through dynamic interactions over the course of the relaxation
601: time, typically of order $10^8$ years for a globular cluster. We do
602: not expect M82-F to exhibit mass segregation over its full extent at
603: its adopted age. If we assume an average stellar mass of 0.87~M$_{\odot}$
604: for a full Kroupa IMF, the half-mass relaxation time \citep{meylan87}
605: for M82-F is $t_{rh} = 4 \times 10^8$ years --- roughly an order of
606: magnitude longer than the cluster's current age.
607:
608: Recent studies, however, have found evidence of mass segregation in
609: significantly younger clusters. \citet{lynne98} found that the
610: highest mass stars in the 0.8 Myr-old Orion Nebula Cluster are
611: preferentially located in the cluster center, and that stars down to
612: 0.3 $M_{\odot}$ are less centrally concentrated than more massive
613: stars within the inner 1.0 pc. The massive cluster R136 is mass
614: segregated at an age of only 3 Myr \citep{brandl96}. Less massive LMC
615: clusters NGC 1805 \citep{degrijs02b}, NGC 1818, NGC 2004 and NCG 2100
616: \citep{gouliermis04} as well as NGC 330 \citep{sirianni02} in the
617: Small Magellanic Cloud all display mass segregation at ages of 10--50
618: Myr.
619:
620: Numerical simulations demonstrate that segregation of a cluster's most
621: massive stars occurs much more rapidly than the half-mass relaxation
622: time \citep{gerhard00}. Indeed, if high mass stars are somehow
623: concentrated at the center of the cluster at the time of formation,
624: the relaxation timescale there will be shorter. Dynamical mass
625: segregation will thus proceed more rapidly at the core, on the order
626: of a few crossing times \citep{degrijs02b}. Based on the measured
627: velocity dispersion and half-light radius, the crossing time for M82-F
628: is $t_c = 1.4 \times 10^5$ years. This is significantly less than the
629: age of the cluster, which implies that the core has had time to
630: undergo mass segregation. \citet{degrijs02a} showed that the cores of
631: SSCs may undergo significant dynamical evolution in as little as 25
632: Myr. The young age of M82-F relative to its half-mass relaxation time
633: may therefore imply rapid mass segregation for the massive stars or
634: support aggregation \citep{bonnell02} as a mechanism for high mass
635: star formation.
636:
637: \section{Summary}
638:
639: We present new, high resolution Keck/NIRSPEC echelle spectrometry
640: and new HST/ACS imaging of the super--star cluster M82-F. The main
641: results of this study are:
642:
643: \begin{enumerate}
644:
645: \item Measuring the stellar velocity dispersion and half-light radius
646: of the cluster in the $H$-band, we find a virial mass of $6.6 \pm 0.9
647: \times 10^5$ M$_{\odot}$ within the half-light radius. In the
648: $I$-band we find a mass of $7.0 \pm 1.2 \times 10^5$ M$_{\odot}$,
649: which supersedes the larger value quoted by SG01.
650:
651: \item The cluster's light-to-mass ratio in the $I$- and $H$-bands are
652: inconsistent with a Kroupa IMF over the full range of stellar masses
653: from 0.1 to 100 M$_{\odot}$. This may be due to a top-heavy IMF, with
654: a lower mass cutoff of about 2 M$_{\odot}$.
655:
656: \item The cluster is successively larger at each shorter wavelength
657: imaged. We interpret this result as due to mass segregation, even at
658: cluster's young age of about 50 Myr.
659:
660: \item Mass segregation tends to overstate the $L/M$ ratio, as our
661: observations of $H$-band light are insensitive to the mass of stars at
662: radii larger than the volume occupied by red supergiant stars. As
663: such, the measured mass represents a lower limit on the mass of the
664: entire cluster.
665:
666: \item The IMF of M82-F differs from the IMF of nearby SSC MGG-9. The
667: IMF in the nuclear starburst of M82 thus varies from cluster to
668: cluster. This may be due to either IMF variation that is inconsistent
669: with a single, ``universal'' form or the degree of mass segregation
670: within individual clusters.
671:
672:
673: \end{enumerate}
674:
675: %%%%%%%%%%%%%%%%%%%
676: % Acknowledgments %
677: %%%%%%%%%%%%%%%%%%%
678:
679: \acknowledgments
680:
681: We would like to thank the staff of the Keck Observatory, and
682: observing assistant Gary Puniwai specifically. We also thank
683: L. J. Smith for providing us with optical spectroscopy of M82-F. NM
684: thanks I. R. King for helpful discussions on elliptical clusters,
685: A. Cotton-Clay for mathematical assistance, and J. Terrell for
686: generous computing support. The authors wish to recognize and
687: acknowledge the very significant cultural role and reverence that the
688: summit of Mauna Kea has always had within the indigenous Hawaiian
689: community. We are most fortunate to have the opportunity to conduct
690: observations from this mountain. This work has been supported by NSF
691: Grant AST--0205999.
692:
693: %%%%%%%%%%%%%%
694: % References %
695: %%%%%%%%%%%%%%
696:
697: \bibliographystyle{apj}
698: \bibliography{apj-jour,astro_refs}
699:
700:
701: %%%%%%%%%%%
702: % Figures %
703: %%%%%%%%%%%
704:
705: \clearpage
706:
707: %------------%
708: % IR Spectra %
709: %------------%
710:
711: \begin{figure}
712: \centering \epsscale{0.95}
713: \plotone{f1.eps}
714: \caption{Near-IR rest-frame spectra of echelle orders 49 and 47 for M82-F
715: and a range of template supergiants. The spectra are identically
716: normalized, but offset vertically for clarity (zero points are $-$0.7,
717: 0.3, 1.3, 2.3 and 3.3, from bottom to top). The cluster spectrum
718: closely resembles the supergiant spectra, with the features ``washed
719: out'' by the stellar velocity dispersion (e.g., the CO bandheads at
720: $1.5582 \mu$m and $1.6189 \mu$m).}
721: \label{spect}
722: \end{figure}
723:
724: %----------%
725: % CCF plot %
726: %----------%
727:
728: \begin{figure}
729: \centering
730: \epsscale{0.4}
731: \plotone{f2.eps}
732: \caption{Cross-correlation function (CCF) for the cluster and the
733: template M0Iab star, averaged over all echelle orders. The peak of
734: the template star auto-correlation function has been scaled to the
735: peak of the CCF to emphasize the difference in width (due to the
736: convolution with the velocity distribution function of the cluster).
737: The solid line is the antisymmetric part of the CCF, which is seen to
738: be highly symmetric. }
739: \label{ccfplot}
740: \end{figure}
741:
742: %-----------------%
743: % data, fit plots %
744: %-----------------%
745:
746: \begin{figure}
747: \centering \epsscale{0.5} \plotone{f3.eps}
748: \caption{Comparison of data and residuals from the fitted
749: two-dimensional King function. Plot (a) shows the ACS/F814W image of
750: M82-F, logarithmically scaled to accentuate faint features. Plot (b)
751: shows the data image minus the fitted King function, with the same
752: scaling as plot (a). Contours show the fit to the cluster, at
753: 1,2,4,8,16,32 and 64 percent of the peak value. Scale bars in the two
754: images represent 5 pc ($0.''29$) at the adopted distance of M82. }
755: \label{fitplot}
756: \end{figure}
757:
758: %--------------%
759: % radius vs wl %
760: %--------------%
761:
762: \begin{figure}
763: \centering
764: \epsscale{0.4}
765: \plotone{f4.eps}
766: \caption{Measured half-light radius and observed axial ratio as a
767: function of wavelength. Lines are presented as a guide to the eye.
768: The general decrease in size with wavelength is consistent with mass
769: segregation. Note that the axial ratio remains constant within the
770: error for all wavelengths other than 250 nm.}
771: \label{halfradii}
772: \end{figure}
773:
774:
775: %-----------%
776: % L/M plots %
777: %-----------%
778:
779: \begin{figure}
780: \centering
781: \epsscale{0.4}
782: \plotone{f5.eps}
783: \caption{Comparison of the derived light-to-mass ratios with
784: predictions from Starburst99 models for two IMFs. The L/M ratio for
785: the cluster is clearly inconsistent with a normal Kroupa IMF.
786: $H$-band is shown in the top plot and $I$-band is shown in the bottom
787: plot. Circles mark L/M based on extinction estimates derived from
788: synthetic photometry (Section \ref{extinction}) while triangles mark
789: values based on the SG01 extinction estimate.}
790: \label{sb99plot}
791: \end{figure}
792:
793:
794:
795: \end{document}
796:
797:
798: