THE ASTROPHYSICAL JOURNAL, 454 : L73–L76, 1995 December 1 q 1995. The American Astronomical Society. All rights reserved. Printed in U.S.A.
HUBBLE SPACE TELESCOPE OBSERVATIONS 1 OF GLOBULAR CLUSTERS IN M87 AND AN ESTIMATE OF H 0 BRADLEY C. WHITMORE, 2 WILLIAM B. SPARKS, RAY A. LUCAS, F. DUCCIO MACCHETTO, 3
AND
JOHN A. BIRETTA
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218 Received 1995 July 21; accepted 1995 September 11
ABSTRACT Hubble Space Telescope observations of over 1000 globular clusters in the central region of M87 have been made using the Wide Field and Planetary Camera 2. The limiting magnitude in V is 26 mag, more than 2 mag beyond the turnover of the luminosity function. The distribution is well fitted by a Gaussian profile with a mean of ^m V0 & 5 23.72 H 0.06 mag and a width of 1.40 H 0.06 mag. Assuming a value of ^M V0 & 5 27.4 H 0.25 mag for the globular cluster luminosity function results in a distance modulus of (m 2 M ) 5 31.12 mag to the Virgo Cluster. Using a value of D Coma2Virgo (m 2 M ) 5 3.71 H 0.10 mag and adopting a velocity for the Coma Cluster of 7188 km s 21 result in a value for the Hubble constant of H 0 5 78 H 11 km s 21 Mpc 21 . The V 2 I color distribution is bimodal, with peaks at V 2 I 5 0.95 mag and 1.20 mag. The mean size of the clusters is R eff 5 3 pc, with a scatter of 1 pc. Subject headings: distance scale — galaxies: individual (M87, NGC 4486) — galaxies: star clusters — globular clusters: general sumption, of course, is that the GCLF is indeed ‘‘universal,’’ a hypothesis that needs further verification.
1. INTRODUCTION
The position of M87 (5NGC 4486) in the core of the Virgo Cluster provides an important advantage over other Virgo Cluster galaxies for distance measurements and for the determination of the Hubble constant. For example, a large uncertainty in the Freedman et al. (1994) estimate of H 0 using Cepheid variables is the question of where M100 is relative to the cluster core. Hence, although the scatter in the periodluminosity relation for Cepheid variables is 7 times smaller than the width of the luminosity function for globular clusters, the formal uncertainty in our estimate of H 0 is actually smaller. The incredible number of clusters in M87 also provides an important advantage in allowing statistical uncertainties of less than 0.1 mag in determining the mean of the distribution. Finally, the high spatial resolution of the Hubble Space Telescope (HST) provides a vast improvement in the ability to distinguish globular clusters from both foreground and background objects. The globular cluster luminosity function (hereafter GCLF) has proved to be remarkably similar for a wide range of galaxies. It can be well described by a Gaussian profile with a mean ^M V0 & 5 27.4 H 0.25 mag (see § 3.2) and a width s 5 1.4 H 0.2 mag, where ^M V0 & is the mean of the distribution in the Johnson V band corrected for galactic extinction. However, the turnover in the GCLF has been measured clearly only in the Milky Way and M31. While a few heroic ground-based observations of galaxies in the Virgo Cluster have reached just beyond the turnover (van den Bergh, Pritchet, & Grillmair 1985; Cohen 1988; Harris 1991), observations with HST are now capable of observing roughly 2 mag deeper than ground-based measurements, which makes the measurements of the GCLF potentially one of the most accurate methods of determining distances. The crucial as-
2. OBSERVATIONS AND REDUCTIONS
M87 was observed with HST on 1995 February 21 using the Wide Field and Planetary Camera 2 (hereafter WF 5 Wide Field Camera and PC 5 Planetary Camera). Four 600 s exposures were taken with both the F555W and F814W filters. The four separate exposures provide excellent cosmic-ray rejection, which was performed using the STSDAS task GCOMBINE, followed by the IRAF program COSMICRAYS to remove hot pixels. Object identification was performed on the sum of the F555W and F814W images using DAOFIND from the DAOPHOT package for the first pass, followed by a visual comparison with both the F555W and F814W images. Aperture photometry was performed using a 3 pixel radius and a sky annulus between 5 and 10 pixels. Aperture corrections in F555W were 20.28 mag and 20.17 mag for the PC and WF, respectively, and 20.38 and 20.19 for the F814W filter, based on measurements of isolated stars by Whitmore et al. (1995). Photometric zero points of 22.573 mag, to convert our F555W observations to Johnson V, and 21.709 mag, to convert F814W observations to Cousins I, have been adopted from Whitmore et al. (1995). These are in excellent agreement with the values of Holtzman et al. (1995):22.56 in V and 21.69 in I. Color terms from Table 7 of Holtzman et al. (1995) have been assumed in deriving our zero points. This correction is very small for the M87 globular clusters, with values of about 20.02 in V and 20.04 in I for a typical cluster with V 2 I 5 1.2 mag. Galactic extinction corrections of AV 5 0.067 H 0.04 mag and AI 5 0.032 H 0.02 mag have been adopted based on the value of AB 5 0.09 H 0.06 from Burstein & Heiles (1984) and the reddening curves by Mathis (1990). A comparison with Couture, Harris, & Allwright (1990) shows excellent agreement in our photometry, with a zero-point offset of 0.03 mag for the four brightest objects in common with Couture et al. beyond 500 and a scatter of 0.07 mag. A concentration index was determined using the difference in magnitudes obtained with aperture radii of 0.5 pixel and 3.0
1 Based on observations with the NASA/ESA Hubble Space Telescope, obtained at the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS5-26555. 2
[email protected]. 3 On assignment from the Space Science Department of ESA.
L73
L74
WHITMORE ET AL.
Vol. 454
FIG. 2a
FIG. 2b
FIG. 2.—(a) The GCLF for the full sample in the Johnson V band. The dashed line shows the data after the background and completeness corrections have been made. (b) The corrected GCLF for the full sample in the Johnson V band. The dotted line shows a Gaussian fit to the corrected data.
pixel (i.e., D 0.523.0 ). This index was used to weed out diffuse objects from the sample (i.e., background/foreground galaxies). Values of D 0.523.0 , 2.7 mag for the PC and D 0.523.0 , 2.2 mag for the WF provided a reliable discriminant. Measurements of D 0.523.0 were also used to determine the effective radii of the clusters, as will be discussed in § 3.3. Numerical experiments were used to determine completeness corrections in six annular regions around the nucleus of M87. Objects with the point-spread function (PSF) of an average cluster were artificially added to our fields and were then identified using our normal object identification technique. We define the completeness threshold as the magnitude at which 50% of the objects are missed. This occurs at a magnitude of V 5 24.8 mag within 160 of the center of M87 and at 26.2 mag for the outermost region between 790 and 1140. A correction for the effects of contamination by foreground stars was also made, based on observations in a nearby field observed by the Medium Deep Survey (Griffiths et al. 1994). The high density of globular clusters near the center of M87 makes this correction nearly negligible (e.g., about 5% at V 5 25 mag). 3. RESULTS
3.1. The Globular Cluster Luminosit y Function Figure 1 (Plate L7) shows a portion of the V image from WF chip No. 2. The squares show the clusters in the range 23.4 mag , m V0 , 23.9 mag, near the peak of the luminosity function. The circles show the clusters in the range 25.0 mag , m V0 , 25.5 mag, well beyond the turnover in the luminosity function. The major result of this Letter is immediately obvious from even this small sample, with 18 clusters in the bright range but only nine in the faint range. The photograph also shows that it is relatively easy to identify 25th magnitude clusters. The incompleteness is only about 20% in this range. The objects labeled ‘‘C’’ are from the Couture et al. (1990) ground-based observations. Figure 2a shows the GCLF in the Johnson V passband for 1032 globular clusters in M87. The dotted line shows the distribution after the background/foreground and complete-
ness corrections have been made, down to our mean completeness threshold for the total sample at V 5 25.5 mag. The mean of the corrected distribution is ^m V0 & 5 23.74 with a width s 5 1.44 mag, as determined by fitting a Gaussian profile in the range from 20.8 mag to 25.55 mag using the NGAUSSFIT task in STSDAS. Figure 2b shows a Gaussian fit to the total corrected distribution. Table 1 lists the values of ^m V0 & and s for five different distributions: (1) the total sample; (2) clusters beyond 540, where the completeness corrections are smaller; (3) clusters bluer than V 2 I 5 1.12 mag, the mean color of the clusters in M87; (4) clusters redder than V 2 I 5 1.12 mag; and (5) the I measurements with a constant offset of V 2 I 5 1.12 mag added. The scatter in these five measurements provides an estimate of the statistical uncertainty in our measurement of ^m V0 &. We adopt as our best estimate the mean of these values, with samples 1, 2, and 5 double weighted. This yields ^m V0 & 5 23.72 H 0.06 mag and a mean width s 5 1.40 H 0.06 mag. 3.2. An Estimate of the Hubble Constant The main result of our Letter is the finding that the GCLF in M87 is well described by a Gaussian profile with ^m V0 & 5 23.72 mag and s 5 1.40 mag. The waters become somewhat murky as we proceed from this clear-cut result to an TABLE 1 GAUSSIAN FITS Sample 1. 2. 3. 4. 5.
m V0
Total sample. . . . . . . . . . . . . . . . R . 540. . . . . . . . . . . . . . . . . . . . . . . . . . V 2 I , 1.12 mag . . . . . . . . . . . . . . . V 2 I . 1.12 mag . . . . . . . . . . . . . . . Total m I0 sample. . . . . . . . . . . . . . . . Weighted mean b . . . . . . . . . . . . . . . .
TO
VARIOUS SAMPLES
Number
^m V0 & (mag)
s (mag)
1032 473 531 499 1032
23.74 23.71 23.58 23.71 23.79 a
1.44 1.35 1.32 1.35 1.48
23.72 H 0.06
1.40 H 0.06
NOTE.—The fits were made using the NGAUSSFIT task in STSDAS allowing the amplitude, mean, and width of the Gaussian to vary. a Using ^m I0 & 5 22.67 mag plus a mean color of V 2 I 5 1.12 mag. b Double weighting samples 1, 2, and 5.
No. 2, 1995
GLOBULAR CLUSTERS IN M87 AND ESTIMATE OF H 0
estimate for a value of the Hubble constant, and different readers may have different prescriptions for making this determination. The first step is a determination of ^M V0 &, the intrinsic value of the mean absolute V magnitude of a Gaussian describing the GCLF. Using the Milky Way and M31, Secker (1992) finds a value of ^M V0 & 5 27.40 H 0.10 mag, while Sandage & Tammann (1995) find a value of ^M V0 & 5 27.62 H 0.08 mag. The primary reason for the difference is the use of a new RR Lyrae calibration scale by Sandage & Tammann. A potential problem to consider is the use of data on spiral galaxies to calibrate our results for an elliptical galaxy. Ashman, Conti, & Zepf (1995) suggest that metallicity differences between globular clusters in spirals and ellipticals should result in values of ^M V0 & that are about 0.2 mag fainter for ellipticals. Van den Bergh et al. (1985) find ^B 2 V& 5 0.80 mag in M87, roughly 0.10 mag redder (more metal rich) than Sandage & Tammann (1995) find for the Milky Way and M31, indicating that such a correction may be relevant. In addition, we see some evidence for a small shift in the GCLF as a function of color in Table 1, where the red sample has a value of ^m V0 & that is 0.13 mag fainter than the blue sample, although the difference is only marginally significant. We shall adopt a value of ^M V0 & 5 27.4 H 0.25 mag in the present Letter. This covers the range from ^M V0 & 5 27.62 mag (i.e., Sandage & Tammann 1995, assuming no corrections between spirals and ellipticals) to ^M V0 & 5 27.20 mag (i.e., Secker 1992, with a 0.2 mag correction between spirals and ellipticals). Combining our value of ^m V0 & 5 23.72 mag with ^M V0 & 5 27.4 mag yields a distance modulus of (m 2 M) 5 31.12 to the Virgo Cluster, corresponding to a distance of 16.75 Mpc. The most straightforward approach at this point would be simply to divide the velocity of the Virgo Cluster by this distance. However, a variety of problems makes the determination of the Virgo velocity problematic (Huchra 1988). We therefore adopt the technique of using the well-determined value of D Coma2Virgo (m 2 M ) 5 3.71 H 0.10 mag (e.g., van den Bergh 1992; de Vaucouleurs 1993; Jerjen & Tammann 1993; Freedman et al. 1994) to estimate the distance modulus to the Coma Cluster. This cluster is roughly 6 times further than the Virgo Cluster; hence, uncertainties introduced by peculiar local velocities have a much smaller effect on our estimate of the Hubble constant. The resulting distance modulus of the Coma Cluster is (m 2 M ) 5 34.83 mag, corresponding to a distance of 92.5 Mpc. Adopting a value of 7188 km s 21 for the velocity of the Coma Cluster (Jerjen & Tammann 1993) results in a value for the Hubble constant of H 0 5 78 H 11 km s 21 Mpc 21 . Table 2 shows an estimate of our error budget. It is interesting to note that using this value of the Hubble constant and a distance of 16.75 Mpc for Virgo predicts a velocity for the Virgo Cluster of 1307 km s 21 , intermediate between the values of 1404 km s 21 (Huchra 1988) and 1179 km s 21 (Jerjen & Tammann 1993). The main question that remains is whether the population of globular clusters in M87 is similar to those in other galaxies. After all, M87 has a specific globular cluster frequency that is 3–10 times higher than a typical galaxy (Harris 1991). In addition, it is in the core of a rich cluster of galaxies where conditions may be quite different. Perhaps the massive black hole in the center of M87 has influenced the GCLF for the inner clusters, or perhaps the X-ray corona influences the GCLF properties. Holtzman et al. (1992) discovered a popu-
L75
TABLE 2 ERROR BUDGET
FOR
ESTIMATE
OF
H0
Source of Uncertainty
Estimated Error (mag)
Photometric zero point. . . . . . . . . . . . . Extinction correction . . . . . . . . . . . . . . . Aperture correction . . . . . . . . . . . . . . . . Statistical uncertainties in ^m V0 & . . . . Intrinsic spread in ^M V0 & . . . . . . . . . . . . D Coma2Virgo (m 2 M). . . . . . . . . . . . . . . . . Velocity of Coma Cluster . . . . . . . . . .
0.05 0.04 0.02 0.06 0.25 0.10 0.09
Total. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
0.30
lation of ultraluminous blue star clusters in NGC 1275, the cD galaxy in the core of the Perseus Cluster. They believe these might be protoglobulars that have recently formed during an accretion event. If a similar burst has occurred in M87 within the last few billion years, this might skew the distribution and affect our estimate of ^m V0 &. However, there is no empirical evidence that the GCLF in M87 is not t ypical of an y normal elliptical galax y. Van den Bergh et al. (1985) find that it is not possible to reject the hypothesis that M87 and Local Group globular clusters are drawn from the same distribution. Lauer & Kormendy (1986) find the GCLF of the clusters in the central region of M87 is indistinguishable from the GCLF of clusters outside the central 600. We find that the width of the distribution is s 5 1.40 H 0.06 mag, which is normal for a GCLF (Harris 1991). As we shall discuss below, there is no strong correlation of the cluster sizes with the distance from the center of M87, the V 2 I distribution is similar to other galaxies (although it does show evidence of bimodality), and there is no trend for the brighter clusters to be bluer, and hence no evidence of a very recent burst of cluster formation. 3.3. Colors and Sizes Figure 3 shows the V 2 I versus V diagram for the clusters with measurement errors less than 0.3 mag in V 2 I. There is no clear trend for the brighter clusters to be bluer, which indicates that there has not been a recent burst of cluster formation similar to the cD galaxy NGC 1275 (Holtzman et al.
FIG. 3.—V 2 I vs. V for the clusters with measurement uncertainties V 2 I , 0.3 mag.
L76
WHITMORE ET AL. clusters observed using the Faint Object Camera, which provides better sampling of the PSF (i.e., pixels are one-third the size of the PC pixels), give similar results. 4. SUMMARY
FIG. 4.—V 2 I histogram for clusters with measurement uncertainties V 2 I , 0.3 mag. Note the bimodal distribution with peaks at V 2 I 5 0.95 mag and V 2 I 5 1.20 mag.
1992) or the merger remnants NGC 7252 (Whitmore et al. 1993) and NGC 4038/4039 (Whitmore & Schweizer 1995). In particular, the average value for the 10 brightest objects is ^V 2 I & 5 1.13 H 0.03 mag, essentially identical to the total population, which has ^V 2 I & 5 1.12 H 0.01 mag. The distribution appears to be slightly bimodal in color, with denser concentrations along the edges of the distribution rather than in the center. Figure 4 shows the corresponding color histogram. The distribution again appears to be bimodal, with peaks at about V 2 I 5 0.95 and V 2 I 5 1.20 mag. This is in good agreement with the result of Elson & Santiago (1995) obtained from a smaller sample farther out in M87, once a correction of 10.15 mag has been made to their V 2 I measurements to bring them into agreement with the measurements of Couture et al. (1990) and our own measurements. Such bimodality is relatively common in galaxies, as discussed by Zepf & Ashman (1993), but only appears to affect the value of ^M V0 & at about the H0.1 mag level. The clusters on the Planetary Camera of the WFPC2 are clearly resolved. Our values of the effective radius, R eff , defined as the radius encompassing half the light from the cluster, are based on our measurements of D 0.523.0 (see § 2) and numerical experiments similar to those described in Whitmore & Schweizer (1995). The average effective radius is R eff 5 3.1 pc, with a scatter of only 0.9 pc and no obvious correlation with the distance from the center of M87. This is similar to effective radii found in Milky Way clusters that have an average value of about R eff 5 3 pc (van den Bergh 1995). Observations of five
HST offers several important advantages for measuring the globular cluster luminosity function for distant galaxies. These include a detection limit which is 2 or 3 mag deeper than ground-based observations, the high resolution necessary to distinguish pointlike clusters from the foreground and background galaxies, and the ability to use very small apertures, which reduces the effect of the background galaxy and allows measurements to be made much nearer the center of the galaxy where the density is highest. Similarly, M87 has several advantages over other galaxies in the measurement of the GCLF and the determination of the Hubble constant. These include an incredibly dense population of clusters, which provides high statistical accuracy and minimizes the effects of background contamination, and a position in the core of the Virgo Cluster, which removes the uncertainty of whether the galaxy is in front of or behind the cluster. We have combined these attributes by using HST to measure over 1000 clusters in M87, with a limiting magnitude which is more than 2 mag beyond the turnover in the luminosity function. We find the GCLF is well fitted by a Gaussian profile with a value of ^m V0 & 5 23.72 H 0.06 mag and a width of 1.40 H 0.06 mag. Adopting a value of ^M V0 & 5 27.4 H 0.25 mag results in a distance modulus of (m 2 M) 5 31.12 to the Virgo Cluster. Bootstrapping our way to the Coma Cluster leads to an estimate of H 0 5 78 H 11 km s 21 Mpc 21 for the Hubble constant (see § 3.2 for details). At present, the largest uncertainties are introduced by the local calibrators and the question of the universality of the GCLF. We look forward to the time when these questions have been resolved, since this promises to reduce the already small uncertainty even further. We would like to thank Stefano Casertano, Harry Ferguson, Wendy Freedman, Ken Freeman, Dan Golombek, William Harris, Vera Rubin, and Marc Postman for helpful discussions and correspondence; Bryan Miller for allowing us to use his software for performing the artificial star experiments; and especially Allan Sandage, Gustav Tammann, and Steve Zepf for a careful reading of this Letter and several useful suggestions. Support for this work was provided by NASA through grant GO-5477.01-93A from the Space Telescope Science Institute, which is operated by AURA, Inc., under NASA contract NAS5-26555.
REFERENCES Ashman, K. M., Conti, A., & Zepf, S. E. 1995, AJ, in press Burstein, D., & Heiles, C. 1984, ApJS, 54, 33 Cohen, J. 1988, AJ, 95, 682 Couture, J., Harris, W. E., & Allwright, J. W. B. 1990, ApJS, 73, 671 de Vaucouleurs, G. 1993, ApJ, 415, 10 Elson, R. A., & Santiago, B. X. 1995, MNRAS, in press Freedman, W. L., et al. 1994, Nature, 371, 27 Griffiths, R., et al. 1994, ApJ, 437, 67 Harris, W. 1991, ARA&A, 29, 543 Holtzman, J., et al. (The WFPC2 Team) 1992, AJ, 103, 691 Holtzman, J., Burrows, C. J., Casertano, S., Hester, J. J., Trauger, J. T., Watson, A. M., & Worthey, G. 1995, PASP, in press Huchra, J. 1988, in ASP Conf. Ser. 4, Extragalactic Distance Scale, ed. S. van den Bergh & C. J. Pritchet (San Francisco: ASP), 257
Jerjen, H., & Tammann, G. A. 1993, A&A, 273, 354 Lauer, T. R., & Kormendy, J. 1986, ApJ, 303, L1 Mathis, J. S. 1990, ARA&A, 28, 37 Sandage, A., & Tammann, G. A. 1995, ApJ, 446, 1 Secker, J. 1992, AJ, 104, 1472 van den Bergh, S. 1992, PASP, 104, 861 ———. 1995, Nature, 374, 215 van den Bergh, S., Pritchet, C., & Grillmair, C. 1985, AJ, 90, 595 Whitmore, B. C., Miller, B., Schweizer, & Fall, S. M. 1995, in preparation Whitmore, B. C., & Schweizer, F. 1995, AJ, 109, 960 Whitmore, B. C., Schweizer, F., Leitherer, C., Borne, K., & Robert, C. 1993, AJ, 106, 1354 Zepf, S. E., & Ashman, K. M. 1993, MNRAS, 264, 611
PLATE L7
FIG. 1.—A portion of the V image from WF chip No. 2. The squares show the clusters in the range 23.4 mag , m V0 , 23.9 mag, near the peak of the GCLF. The circles show the clusters in the range 25.0 mag , m V0 , 25.5 mag, beyond the turnover in the GCLF. The objects labeled ‘‘C’’ are from the ground-based observations of Couture et al. (1990). The arrow points north and is 20 long. The vast majority of the other objects are clusters that fall outside the specified magnitude ranges. Note that there are twice as many clusters in the brighter range, which clearly shows that the turnover in the luminosity function has been reached, and that 25th magnitude clusters are easy to identify. WHITMORE et al. (see 454, L74)
