Astron. Astrophys. 326, 528-536 (1997)

3. Analysis

3.1. Photometry of the resolved stars

Since the crowding is not severe, stellar fluxes can be measured using the standard aperture photometry package apphot of IRAF, by following the method of Paresce et al. (1995). The main difficulty here comes from the unresolved galaxy light which increases by a factor of more than 20 from the edge of our field to the inner bulge and does not allow to maintain simultaneously the same significance of detection and the same limiting magnitude over the entire field of view. Subtracting an image obtained by smoothing the light distribution can help the visual identification of point source objects but does not change the fact that the magnitude of the fluctuations increases toward the center, as the average flux level increases. In order to keep a relatively homogeneous significance of star detection (at the expense of an uniform limiting magnitude), we have run the daofind star finding routine with different threshold values. To this end we have divided the image in four domains in which it is reasonable to have a constant threshold value. We have set this threshold at approximately 6 above the average sky level. These domains are defined by elliptical contours with semi-major axis 2:003, 1:004, 0:0057 and axis ratio 0.7, 0.58, 0.5 respectively. The procedure was run on the F175W image and the list of detected stars that we obtained was used for aperture photometry in both images.

The magnitudes m₁₇₅ and m₃₄₂ of each star were calculated as

where I is the inverse sensitivity of the modes used (7.65 10^-17 and 3.50 10^-18 ergs cm^-2 s^-1 Å^-1 per count/s respectively for the F175W and F342W frames), c is the total number of counts attributable to the star, t the exposure time in s and the energy fraction associated with the aperture photometry. The latter factor accounts for the amount of light missed by the aperture as well as unduly subtracted out in the background reference annulus. We have adopted an aperture radius of 3 pixels and a reference annulus with inner and outer radii of 4 and 7 pixels, respectively. We have measured = 0.48 and = 0.59 with available PSF images. A total of 130 stars, measured in both images (we have discarded a few measurements as discussed later), are displayed in the color-magnitude diagram (CMD) of Fig. 2 obtained by plotting the observed m₁₇₅ magnitude versus the m m₃₄₂ color.

Fig. 2. Color-magnitude diagram obtained from the magnitudes measured on the summed F175W and F342W images. The dashed line shows the detection limit in the F342W image. The solid lines show the isochrones of Bertelli et al. for the ages log t = 6.6, 7.0, 7.2, 7.4, 7.6, 7.8 years (from top to bottom).

The statistical and systematic uncertainties expected in the FOC images analysed with the aperture photometry technique have been discussed by Paresce et al. (1995). In our case the uncertainty due to the absolute calibration should be slightly increased over the quoted value of 15% because the energy fraction factor could not be verified very accurately in our image and a nominal value was adopted. It should be noted that the color term is less affected by this additional source of error. The statistical uncertainty associated with the aperture photometry measurement and returned by the phot package is and at 22 mag with the F175W and F342W filters, respectively. This uncertainty increases at fainter magnitudes and in areas of large background from the galaxy light.

The completeness in our sample is a severe issue because it is not only a function of magnitude but also a function of the surface brightness of the background galaxy light. This is especially true for studying the distribution of the individual stars. As a first approximation, we have evaluated the incompleteness in each domain defined above where the limit of detection is expected to be reasonably uniform. Practically, we have added artificial stars at random places in our image and repeated our finding operation for several trials at different magnitudes. We find in this way that our photometry is 95% complete at m , 28% at m in the outer domain where the detection is the deepest. In the second domain (the outer domain is the first and we move to the center) the completeness fractions decrease to 66% and 7% (for respectively the same magnitudes as before), then we obtain 78% at m , 30% at m in the third domain. No realistic evaluation was made in our central domain.

The F175W image has been used to set the detection limit of the sample because it allows the selection of the hottest stars and it was less deep than the F342W image. As a consequence of the latter, faint stars are seen in the F342W image that have no counterpart in the F175W frame. These stars are red and have no impact on the completeness of our sample. On the blue side of the CMD (in Fig. 2) we have plotted the (dashed) line resulting from the detection limit in the F342W frame (scaled down according to the sensitivity difference from the detection limit in the F175W frame). The F342W image is deep enough that all stars above the detection limit in the F175W image with realistic blue colors are detected. In order to check this, we have visually examined three stars that were not detected in the F342W frame and two objects with abnormally blue color. We have found that their detection was already marginal in the F175W frame; the two latter objects have been discarded.

3.2. Galaxy light distribution

As noted above, individual stars are resolved in both images against the diffuse light of the integrated stellar population of the bulge. The contribution of the resolved stars to the total light is minimal (of the order of 2 to 3 percent), implying that we are only resolving the tip of the blue stellar population. For all practical purposes the study of the unresolved stellar population is equivalent to that of the total light. This allows direct comparison with previous photometric data based on integrated light and avoids painful completeness corrections for removing the contribution of resolved stars down to a given magnitude.

We have first calculated the total number of counts due to galaxy light in the field of view. In order to avoid the warped edges of the images, the calculation was made in a circular aperture of radius 220 pixels, i.e. 3:0016 and an area of 31.3 square arcsec. For the F175W co-added images we get a total count rate of 280 count/s (after subtracting a detector background of 7 10^-4 count/pixel/s). Only a fraction of these are due to true UV light because of the F175W filter red leak. In order to determine this fraction and to make comparison with previous IUE measurements (Burstein et al. 1988), we need to use a spectrum of the galaxy. For a reasonable approximation we extend the IUE spectrum into the visible with a model spectral energy distribution (see Sect. 4.2) that we have normalized to a magnitude V = 11.94 in the equivalent IUE aperture of 154 square arcsec (Burstein et al. 1988). Assuming that the spectral shape is the same in the area of 31.3 square arcsec, we scale down the spectral energy distribution by a factor of 1.84 according to the aperture photometry of van den Bergh (1976) in V light. By synthetic photometry using the synphot package we predict 160 count/s whereas we get 280 count/s. This contradiction cannot be explained alone by a possible underestimation of the detector background (9 10^-4 count/pixel/s is possible instead of 7 10^-4 count/pixel/s). It probably tells us that the scaling factor based on the V light profile is not appropriate for the UV and suggests a color gradient between the UV and the visible in the domain covered by the two apertures we refer to (radius of 3:001 used in the FOC image and equivalent radius of 7" for the IUE diaphragm), in the sense of the central region getting bluer. A bluer nucleus has also been reported by van den Bergh (1976) and Pritchet (1979). Our conclusion is however affected by the uncertainties accompanying the approximation of the spectral energy distribution and the comparison of aperture photometry with different instruments (orientation and shape of apertures). Assuming that our approximation of the spectrum is valid for the purpose of red leak calculation, synthetic photometry shows that only 75% of the counts are due to true UV light, i.e. below 2400 Å.

Fig. 3. Comparison of the average radial light profiles resulting from isophote fitting: F342W, solid lines (there are two curves because the image with the neutral density filters has been used for the central region); F175W, dashed line (in the outer part the profile is very depending on the correction for detector background). The dot-dashed line is the B-luminosity profile f rom Pritchet (1979). At the distance of NGC 5102 1" is 15 pc.

Fig. 4. The average (175-342) color profile resulting from isophotal fitting as in Fig. 3. The log scale for the distance emphasizes the very central region. At the distance of NGC 5102 1" is 15 pc and a pixel is 0.2 pc. The wrinkles are due to the noise accompanying isophotal fitting with steps as small as 1 pixel in the central regions. The F342W image with the neutral density filter is used in the central region. The dashed line is the color variation without the correction for non-linearity effects in the F175W image. The error bars show the uncertainty at the edge of the field resulting from dark current subtraction (upper bound 0.9 10-3 count/pixel/s; lower bound 0.6 10-3 count/pixel/s).

In order to reduce some of the above uncertainties and to take advantage of the angular resolution of the HST, we have concentrated on the study of a possible color gradient within the images themselves. Average light profiles have been calculated by fitting elliptical isophotes with the IRAF/STSDAS task ellipse and have been plotted in Fig. 3 (1" corresponds to 15 pc and 1 pixel to 0.2 pc in NGC 5102). The profile in F342W light is obtained with the co-added images using radii from 30 pixels to 240 pixels and with the image obtained with neutral density filters for radii from 5 to 45 pixels; the agreement looks reasonable in the regime of overlap. The profile in F175W light has been plotted assuming the nominal detector dark current of 7 10^-4 count/pixel/s; it would be of course steeper in the outer part if more of the count rate is ascribed to detector background. In Fig. 4 we directly display the m₁₇₅--m₃₄₂ color variation while emphasizing the very central region. From 1:004 outward, the observed profiles are comparable to the profile in B-light from Pritchet (1979) and reveal a shallow color gradient (getting redder outward) consistent with the gradient inferred from our comparison above with the IUE measurements and the UB gradient reported by van den Bergh (1976). A quantitative comparison is not possible given the uncertainties of our dark current subtraction and the fact that ground-based photometry was made through large diaphragms. From 1:004 inward, the m₁₇₅--m₃₄₂ blue color gradient gets much steeper (from m₁₇₅--m₃₄₂ ~ 0.0 at 1:004 to at 0:004) but does not extend into the very center where the color gets back to 0.2, a value comparable to that typically found at about 3" from the center. We have performed a correction for non-linearity effects in the F175W image due to the high count rates in the central regions of the galaxy and Fig. 4 shows the impact of this correction on the color. Although this correction is uncertain, it cannot explain the upturn of the color gradient in the very center: at 0:002 from the center the count rate is already less than 0.1 count s^-1. From the brightness profiles in the very center (Fig. 5) there is no evidence for a core and NGC 5102 falls into the category of galaxies that have profiles that continue into the resolution limit as steep power laws (Lauer et al. 1995).

Fig. 5. Same as Fig. 3 with only the F342W (plus neutral filter) average radial light profile and the log scale emphasizing the central region.

3.3. The spatial distribution of resolved stars

Examination of Fig. 1 shows that the individual stars are preferentially found in the direction of the galaxy major axis (almost the vertical in Fig. 1). This visual impression has been quantitatively confirmed by counting the stars as a function of position angle, using 10^o wide angular bins with the center of the galaxy as the summit. With stars in our two external domains only, we find that the direction favoured is at a position angle of 10^o 10 from vertical, i.e. 55^o from North and close to the major axis of the galaxy (optical and HI axis, respectively 48^o and 43^o from North, as reported by van Woerden et al. 1993). This approach is not severely affected by the photometric incompleteness; if these effects are further reduced by considering only stars brighter than 22 mag, the result is affected by small number statistics but not basically changed.

Whether the resolved blue stars follow the overall galaxy light distribution is more difficult to answer, especially since the detection limit and the completeness are depending on this background light. In a first approach we have tried to stay as consistent as possible with our analysis of the completeness; we have counted stars in our three external domains, corrected these numbers for incompleteness according to the magnitude range of the stars and compared the resulting numbers with the integrated light in the same domains. We find that the number of stars is in direct proportion to the integrated light in the two external domains but we find slightly more stars closer to the center than one would expect from the integrated light. As this approach is limited to a comparison between two series of only three numbers, we have attempted a comparison of the surface density of resolved stars with the radial light profile obtained as a function of the equivalent semi-major axis (see previous section). Because the incompleteness is more difficult to account for in this approach, we limit ourselves to stars brighter than 21.75 mag. The comparison can be made either with a limited number of stars close to the major axis ( ^o) or with all the stars assuming that they are in a disk and their surface density calculated therein adopting an inclination angle of 30^o. Both comparisons are displayed in Fig. 6. Within the limits of completeness, small number statistics and the lack of investigation in the inner 40 pixels (0:006), Fig. 6 shows that the spatial distribution of resolved stars follows the galaxy light distribution with the stars being slightly more concentrated. This concentration seems even more dramatic with the brightest stars in the sense that no stars with m 21 are observed in the outer parts but an effect from small-number statistics cannot be excluded. A similar result would be obtained with an angle of 70^o, perhaps indicating that a disk structure is not required by the data. Another feature of the distribution of resolved stars is the absence of clustering as usually found in spiral arms of late-type galaxies and star-forming galaxies. The poor angular resolution of the HI distribution has prevented any detailed comparison with our data but we note that, although the HI distribution has a central depression (van Woerden et al. 1993), the average gas surface density rises again in the region under study to a value of 0.9 M pc^-2.

Fig. 6. Comparison of the surface density of resolved stars (m 21.75) with the UV light radial profile. The surface density is in arbitrary units. The light profile results from ellipse fitting and is expressed in count/pixel. The distance along x-axis is the semi-major axis distance for the light profile, the distance from the center of the galaxy for stars within 15o of the major axis (circle), the radial distance of the stars calculated in a disk of inclination angle 30o (triangle).

3.4. The dust patches

The images reveal several dust patches close to the center. The presence of dust is not unusual in early-type galaxies and S0 galaxies (Kormendy and Djorgovski 1989) and has been specifically reported in the outer parts of NGC 5102 (Danks et al. 1979). The HST angular resolution is now known to enhance the detection capability of dust lanes close to the center of early-type galaxies (Van Dokkum and Franx 1995). The dust patches are best seen in the F342W image where the larger number of counts allows a better contrast. The three most prominent are found at position angles of 45^o, 88^o and 195^o, and at distances of 2:0075, 1:002 and 2:002 respectively from the center of the galaxy. They do not form a particular pattern and have themselves irregular shapes with typical sizes of 0:002. The first one noted above is found to extend in a sort of fainter ring shape. The third one merges into a larger and fainter structure at the edge of the FOC field that is possibly the inner part of the dust lane reported at 3" and position angle 200^o by Danks et al. (1979). In this third dust patch we evaluate an average extinction (for the F342W filter) of 1 mag (i.e. a color excess E(BV) of 0.2) by comparison of light profiles on and off the absorbed area. The same type of measurement with the F175W image shows that the count rate right in the patch is not significantly different from the detector dark current, preventing even a crude determination of the extinction in the UV and therefore any constraint on the extinction law at UV wavelength.

© European Southern Observatory (ESO) 1997

Online publication: October 15, 1997
helpdesk.link@springer.de