Bottinelli et al. (1991) and Tammann (1993) argued that PNLF distances to Virgo ellipticals were underestimated because the luminosity depth of the PN surveys was not adequate to sample beyond the brightest (0.5 mag) edge of the PNLF. Since that edge is nearly linear in the logarithmic PNLF, the method becomes insensitive to distance modulus. In addition, those authors challenged the PNLF distances on the basis that a shallow survey of a large galaxy will suffer from a sample size bias. The sense of this argument is that N objects are more likely to be drawn from the low probability bright tail of the large elliptical galaxy PN sample than are N objects from the smaller sample in M31's bulge.
While it is true that PN surveys must extend deep enough to sense the curvature of the PNLF reliably with statistical methods, the required depth is only 0.8 mag. With the exception of NGC 4649 which was observed under poor conditions, Jacoby, Ciardullo, & Ford (1990) estimated the depth of their surveys for 6 Virgo galaxies to be ~ 1.0 mag. Thus, it seemed unlikely that a serious systematic error was contaminating those distances.
An independent assessment of the likelihood of a serious systematic error is provided by recent Cepheid distances to Virgo galaxies. The average PNLF distance to Virgo, based on 6 galaxies, is 15.3 Mpc (using the modern M31 distance and extinction for the zero-point). This result agrees very well with the Ferrarese et al. (1996) distance of 15.8 Mpc to M100 based on HST Cepheids, and the Pierce et al. (1994) distance of 14.7 Mpc to NGC 4571 based on CFHT Cepheids. In addition, Sandage et al. (1996) reports HST Cepheid distances to three near-Virgo galaxies: NGC 4496 at 16.6 Mpc, NGC 4536 at 16.6 Mpc, and NGC 4639 at 25.1 Mpc. Thus, four galaxies are reported in the range 14.7 to 16.6 Mpc, and these are very comparable to the PNLF range of distances (14.3 to 16.2 Mpc). One galaxy, though, is behind all of these. It is unclear which, if any, of these spirals represents the distances to the ellipticals, but it is evident that most of the spirals (four out of five) have distances that support the PNLF distances.
A direct resolution of the challenges to the PNLF distances lies in a short observing project. Deep PN observations in M87 can push well into the plateau region of the PNLF. Data were obtained in April 1995 with the KPNO 4-m telescope to examine the claim that the earlier Virgo data were not deep enough. A total of 7 hours of on-line integration were devoted to detecting fainter PN. This survey also extends to large radial (10 arcmin) distances from M87's nucleus. The survey results are described below.
7.1. A deep PNLF distance contradicts the challenges
Figure 3 shows the new PNLF for M87. A total of 320 PN were identified, but many are fainter than the completeness limit. A total of 201 PN are in the complete sample which extends ~ 1.2 mag down the PNLF.
|
Figure 3. The 1995 PNLF for M87 based on 7 hours at the KPNO 4-m. This PNLF extends 1.2 mag down the PNLF and clearly reaches beyond the linear portion of the bright edge (25.9 < m5007 < 26.3) of the PNLF. The scaled and shifted PNLF from M31 is superposed to illustrate the agreement with the reference galaxy's PNLF and demonstrates that the M87 PNLF is not a power law over this regime. Note the single very luminous object at m5007 = 25.6 which was found first in the 1990 survey. See Jacoby et al. (1996) for a discussion of what this luminous object may be. |
Figure 4 shows a curious effect, though, which has not been fully evaluated at the time of this conference. That is, when the sample of PN is divided in half such that the PN drawn from M87's halo are separated, the PNLF for the inner half (out to 4 arcmin in radius) are systematically fainter by 0.3 mag than the halo sample. I return to this point below, but note here that the more reliable distance is derived from the inner sample because it is less likely to be contaminated by intracluster PN. That distance, 14.4 ± 1.3 Mpc (on the modern M31 distance scale), is nearly identical to the Jacoby, Ciardullo, & Ford (1990) value of 14.9 ± 1.2 Mpc.
|
Figure 4. The 1995 PNLF for M87 where solid points represent those PN found in the inner 4 arcmin and open circles show those PN found beyond 4 arcmin and out to 10 arcmin. |
Thus, a deeper PNLF argues against the contentions of Bottinelli et al. (1991) and Tammann (1993) that the PNLF distance to Virgo is underestimated as a consequence of inadequate survey depth.
7.2. Sample size effects are small
With the new larger sample of PN, it is possible to investigate the effects that different sample sizes have on the final PNLF distance. Subsets of PN were drawn randomly from the sample of 201 PN. Distances for these subsamples of 20, 30, 40, 50, and 100 PN were derived following our standard procedures and compared to the distance based on 201 PN. Figure 5 shows the magnitude of the effect of sample size differences.
|
Figure 5. Results of a Monte Carlo experiment to estimate the effect of deriving distances with different sample sizes. The error bars represent the scatter in the multiple attempts to derive distances with a given number of PN in the sample. |
In the worst case, for a sample of 20 PN, there is a slight tendency to overestimate the distance to a galaxy by up to 3%. Again, this contradicts the challenges of Bottinelli et al. (1991) and Tammann (1993) who claim that our distances would be underestimated. The reason that the effect is small is that the statistical process described by Ciardullo et al. (1989) is cognizant of the sample size and adjusts the derived distances to the most likely one for a given sample. That is, a statistical correction for sample size has always been applied to the PNLF results.
As noted above, M87's halo PNLF is ~ 0.3 mag brighter than the central PNLF. We can all agree that the halo of M87 is not 15% closer than its core! Thus, something must be artificially enhancing the brightness of the PN in the outer regions. Since we have not seen this effect in the two other samples that permit a similar radial test (Cen A, Hui et al. 1993; NGC 4494, Jacoby, Ciardullo, & Harris (1996)), we consider what could cause such an effect here. Five possibilities come to mind:
The color gradient in ellipticals is such that the outer halos are bluer than the inner regions. The first 3 possibilities above have been suggested as possible causes of gradients.
The first, a metallicity decrease, reduces the luminosity of PN, in contradiction to the observed effect. The second, the presence of a young population can enhance the PNLF luminosity if the ages of the halo stars are < 0.5 Gyrs, provided the central population is > 3 Gyrs. We can neither dismiss nor confirm this possibility.
The third option, dust in the central regions, has been suggested by Wise & Silva (1996), Goudfrooij & De Jong (1995), and Witt et al. (1992) to explain the color gradients. To explain the PNLF enhancement, central reddening of E(B - V) ~ 0.06 is needed. Again, this is plausible.
Although the effect has not been seen before, a brightness enhancement in the halo could be caused by instrumental effects. Thus far, flat-fielding errors and filter transmission variations have been dismissed.
The fifth option initially seems highly speculative. The key point is that a large population of intracluster stars exists and can produce PN having a range of distances representing the full depth extent of the Virgo cluster. Thus, some PN may be foreground to M87 and appear brighter than M87's own PN, while other PN will be in the background and be lost in the faint end of the PNLF. Since the number of intracluster PN found is proportional to the area surveyed, there is a survey bias against finding foreground PN in the smaller central region, while simultaneously, there is a bias in favor of finding true M87 PN at the center where the stellar density is high.
What is the likelihood of finding intracluster PN? Arnabaldi et al. (1996) found that 3 of 19 PN in the Jacoby, Ciardullo, & Ford (1990) NGC 4406 sample are intracluster. Since that sample is biased against intracluster PN due to velocity rejection in the survey filter (NGC 4406 was sampled at its systemic velocity of -220 km/s, or about 1500 km/s from the Virgo systemic velocity), there are likely to be many intracluster PN. Arnabaldi et al. (1996) discuss the intracluster population in detail. To zeroth order, the number density of intracluster PN is not a problem.
A definitive source of the enhanced halo PNLF in M87 is not possible at this time. Kinematics can be used to investigate the likelihood that intracluster PN are contaminating the halo sample. The other likely causes, extinction and very young populations, seem less secure at this time because we don't know that they exist while we do know that intracluster PN do exist.