,
The giant elliptical galaxy NGC 4494 was observed in I-band using the Great Ohio Telescope (GOT). Surface brightness profiles were derived and are found to be equally well described by either a Sérsic's or a de Vaucouleurs model with effective radius of ~45 arcsec. The residuals to both fits uncover an excess of light around a radius of 80 arcsec which might be attributed to an underlying disk. The influence of the asymmetric PSF on ellipticity and position angle profiles is carefully investigated. At radii larger than 25 arcsec, these profiles yield approximately constant values of 0.16 for ellipticity and -4° for the position angle. Published B and V photometry is used to determine color gradients in B-I and V-I colors. While the B-I color exhibits no significant gradient (d(B-I)/dlog(r) = 0.008±0.021), V-I shows a quite strong negative gradient of d(V-I)/dlog(r) = -0.365±0.076.
The origin of elliptical galaxies is still a quite controversial debate. The original idea was, that they formed through a monolithic collapse at a rather high redshift, and evolved only passively since then. This scenario resulted in all elliptical galaxies being old, relaxed systems with ages around 12 Gyr. But this formation mechanism fails to explain some recent evidence which has become available with the advance of more detailed CCD photometry. Quite a large fraction of elliptical galaxies are actually now believed to harbor signs of a more active past. Kinematic studies reveal counterrotating cores, surface photometry shows dust in lanes or rings, or shells, which are believed to be features of a recent merger. The existence of a bi-modality in the Globular Cluster (GC) population in some ellipticals also points into the same direction. These findings forced astronomers to reevaluate the formation scenario of elliptical galaxies. Merging and accretion of satellites is now believed to play a key role in the formation and evolution of these early-type galaxies. This hypothesis is supported by N-body simulation of mergers which are able to naturally reproduce the empirical de Vaucouleurs law.
Surface photometry represents one of the most important and most powerful tools to study the properties and history of elliptical galaxies. In this analysis technique, ellipses are fitted to the isophotes of the galaxy. The derived radial ellipticity and position angle profiles provide basic information such as effective radius, isophote twisting, triaxiality or absolute magnitudes. The associated higher order Fourier coefficients from the fits reveal the intrinsic "boxy" or "disky" appearance of the isophotes which can be used to uncover underlying stellar disks. The oldest proposed model to fit the radial intensity profile is the empirical "de Vaucouleurs law" (de Vaucouleurs 1948) that has now been extended with the more flexible Sérsic profile. This model can be written as
,
,
Deep multi-band photometry of elliptical galaxies reveal color gradients in their radial profiles. In general, these gradients are such that the galaxy is redder at its center and gets continuously bluer with increasing radius. This fact is thought to originate from variations in age or metallicity of the stellar population of these objects (e.g. Worthey 1994). Unfortunately, expensive longslit spectroscopy is generally needed to disentangle these two possibilities. Another approach was lately proposed by Wu et. al. (2004) to tackle this so-called age-metallicity degeneracy. They use predictions of stellar population synthesis models to transform the various observed colors into a likelihood distribution in age-metallicity space. Thorough calibration under photometric conditions is needed to convert the instrumental color profiles into meaningful calibrated colors.
NGC 4494 is an intriguing galaxy, which is thought to be a member of the COMA-I group at a distance of 12.8±0.9Mpc (Jacoby, Ciardullo & Harris 1996). This giant E1 galaxy is classified as a LINER. High-resolution HST photometry of the center of this galaxy shows an asymmetric sharp dust ring, with an extinction minimum around only 10-20pc (corresponds to ~1-2arcsec) away from the center (Carollo et al. 1997). Groundbased longslit spectroscopy uncovered a kinematically decoupled core (Bender, Saglia & Gerhard 1994), yet another sign of recent merger activity.
The observations of NGC 4494 were carried out with the 0.25m Great Ohio Telescope (GOT) with a Kodak KAF 1600 Non-ABG CCD mounted on its ST-8 camera. This combination yields a quite large field of view (18x12 arcmin), half of which was filled by the galaxy. NGC 4494's location at a right ascension of 12hrs 31min 24.0sec and a declination of 25° 46'' 29' made it possible to observe the object on May 23rd between 12:30am and 2am, while staying between 1.1 and 1.5 airmasses in all five 600sec exposures. The I filter was chosen for all runs, since no published surface photometry is available in this band. Conditions were not photometric, with a chance of thin clouds affecting the data as could be seen in the morning. Though the 9x9 micrometer pixel scale translates to a angular separation of only 0.71 arcsec/pixel (Steiner 2004, in preparation), poor seeing and large errors in the auto-guiding system of the GOT resulted in a large, asymmetric point spread function (PSF) with its ellipticity climbing up to 0.5 in the center. During the observing run, 22 zero frames, 4 dark frames and 7 flat frames (4 in the morning, 3 in the evening) were taken. Unfortunately, we were unable to obtain calibration data during the observational period, which rendered the original goal to obtain calibrated color profiles unattainable.
The basic data reduction steps were done using the software package IRAF. First, the 22 zero frames were visually inspected and added up to produce a combined zero frame. Dark frames were coadded as well. No dark or zero frames had to be discarded. The combined dark and zero frames were then used to correct the flat fields and object frames. The 3 evening flat fields had to be ignored due to an apparent rotation of the CCD chip, which generated an artificial visible pattern in the reduced data. This rotation was most likely introduced while remounting the CCD to the GOT after polar alignment. One morning flat field had to be discarded as well, due to a "firefly" contamination. The object frames were finally corrected by dividing by the combination of the remaining 3 flat fields. Cosmic rays were removed by the the publicly available program "DCR" by Wojtek Pych (2004), which is very effective in removing the cosmic rays without cutting into the high end of the Poisson noise. The corrected object frames were then registered, trimmed and coadded to increase S/N. After removing a constant background, a slight residual stayed visible in the image. This made it necessary to fit a "tilted plane" through four uncontaminated blank sky fields, spread over the entire ccd. Removal of this residual gradient improved the convergence of the subsequent isophote fitting procedure significantly.
The resulting image was then analyzed with the IRAF package "STSDAS.ISOPHOT". The included "ELLIPSE" task was used to fit ellipses to the image, with point sources masked out "by eye" and ignored during the analysis. The ellipticity and position angle were allowed to vary, whereas the center was kept fixed. The program "BMODEL" was then used to reconstruct a surface brightness profile of the galaxy. Subtraction of this model showed that the procedure worked fine in the outer parts of the galaxy, but left a quite strong residual in the center of the galaxy. Allowing the isophote centers to move by a maximum of 1 pixel for each following step improved the fitting process for this part of the galaxy significantly, leaving almost no residual behind. The center moved a total of less than 2 pixels during this fit. The two surface brightness profiles were then combined, with the variable center for the inner 10arcsec and the fixed center for the outer parts. There are two possible explanations for this rather odd behavior: It might have been simply caused by the asymmetric shape of the PSF. Another possibility is the asymmetric light distribution in the inner 10 parsecs, caused by the excessive amount of light from the dust ring. This highly unresolved feature in our data can be clearly seen in high-resolution HST data (Carollo et al. 1997).
Figure 1a shows the central part final reduced image of NGC 4494. This false color image is 5.91 arcmin across and logarithmically scaled. The smooth model output of the IRAF task "BMODEL" is put on the same scale next to it for comparison. Subtraction of the model from the data resulted in a very clean image (see figure 1c). The residual image shows no signs of prominent asymmetries or shells that could indicate a recent merger. The radial profile of the c4 parameter indicates that the deviations from perfect ellipses are generally negligibly small (see figure 2d). Although the center was modeled quite well, there is some evidence for asymmetry. Since this feature is only visible on the scale of the PSF, this can be completely attributed to the elongated PSF and the misalignment of its position angle (PA) with respect to the PA of the galaxy's major axis (see 4.2 for a more thorough discussion).
  |
  |
Figure 1. Top left: a) Final reduced image of NGC 4494 in I band, observed with the GOT. Top right: b) Model for the galaxy, reconstructed from its I band elliptical surface brightness profile. Bottom left: c) Residual image, the model has been subtracted to bring out possible faint features, which cannot be seen. Bottom right). Figure 1a with ellipses overlaid to bring out the isophotal twist in the inner part of the galaxy, caused by the asymmetric PSF. All images are logarithmically scaled false color images with a size of 5.91 arcmin (500 pixels). |
As already mentioned in the last section, it is crucial to take the two-dimensional shape and profile of the PSF into account for the analysis of the surface brightness profile. Figure 2a shows the radial behavior of the PSF compared to the surface brightness profile of the galaxy. At a radius of about 20 arcsec, the ratio of the PSF contribution drops below a few percent. This suggests a cut-off of at least 15-20 arcsec to gain reliable results.
 |
Figure 2a. Complete I band surface brightness profile of NGC 4494 as observed with the GOT. Green dashed line: PSF profile, scaled to the same central brightness as the galaxy. |
Problems with the auto-guiding system during the observing run, caused by wind and a faint guide star, resulted in a PSF with an elongated shape along the image x-axis. The ellipticity of the PSF rises up to 0.5 in the very center, with a rather constant PA of exactly 90°, as can be seen in figures 2a and b. This quite strong flattening of the PSF, together with the relative twist of the PA with respect to the outer parts of the galaxy, affect strongly the observed galaxy profile. The profile is clearly PSF dominated at small radii (<10arcsec), where the PSF's PA is forced onto the galaxy profile. At a transition radius of about 10 arcsec, the PA starts to flip over to the 'true' and almost constant value of -4° in the outer parts of the galaxy. This also manifests in a circularization of the galaxy image at this transition radius, visible as a prominent drop in ellipticity. The ellipticity stays approximately constant around a value of 0.16 for radii larger than 25 arcsec. The radial profile of the c4 parameter indicates that the deviations from perfect ellipses are generally negligibly small (see figure 2d).
 |
Figure 2b. Ellipticity profile of the galaxy. Green dashed line: PSF profile. |
 |
Figure 2c. Position angle profile of the galaxy. Green dashed line: PA profile of the PSF, elongated along the x-axis. |
 |
Figure 2d. Radial C4 profile of the galaxy. Green dashed line: C4 profile of the PSF. |
A comparison to HST data (Carollo et al., 1997) of the central galaxy core confirms that the isophotal twisting is solely caused by the PSF and has no valuable physical meaning. Their analysis also uncovers a steep cusp in the center, with no additional point source needed for their model. In order to extract meaningful results from this surface brightness profile, I restricted my further analyses to radii larger than 25 arcsec. The application of PSF deconvolution techniques such as the publicly available IDL procedure "MAX_ENTROPY" did result in slightly more circularized images, but introduced negative annuli around a few stars. For this reason, the deconvolved images were not used during the analysis.
The radial surface brightness profiles were fitted with the extremely robust least-squares minimization IDL package "MPFIT", provided by Craig B. Markwardt (NASA/GSFC, available at http://cow.physics.wisc.edu/~craigm/idl/idl.html). Table 1 lists the results for the various fits at different radii.
| fitting range [arcsec] | re | n |
χ2/dof |
re |
χ2/dof |
rcore |
β | χ2/dof |
| 20-40 |
31.82±0.61 |
1.29±0.09 | 0.93 |
56.57±0.96 | 21.26 |
22.86±1.33 | 1.29±0.06 | 1.36 |
| 40-100 |
58.18±1.95 | 4.96±0.91 | 8.32 |
56.47±1.02 | 7.50 |
18.55±2.16 | 1.07±0.02 |
9.38 |
| 100-300 |
65.60±11.00 | 1.56±0.41 | 2.55 |
24.36±2.07 | 2.93 |
105.35±2.31 | 2.31±0.32 |
2.65 |
| 25-250* |
45.89±0.35 | 3.96±0.11 | 13.07 |
45.77±0.33 | 13.45 |
18.44±0.317 | 1.15±0.01 | 26.70 |
Surprisingly, different ranges in radius yield quite different fits to the data. To test the robustness of the fitting process, the starting parameters were altered for several runs, producing almost identical results. An additional free parameter to account for an offset due to inappropriate sky subtraction does not improve the fits. A few interesting things can be noted in this table. Row 4 lists the parameters for the "scientifically valid" range between 25 and 250 arcsec. The Sérsic model seems to be converging toward the empirical de Vaucouleurs law, yielding a value for n that is quite close to 4. Both models give comparable fits, whereas the King model seems to diverge from the data for large radii (see figure 3). A careful investigation of the systematic effects of choosing the fitting range shows that the de Vaucouleurs fit is the most stable fit to the data. The effective radius of 45.77 arcsec (~2.5kpc) describes the data best, with associated systematic errors of &plusm;10 arcsec. In general, the Sérsic profile also yields good fits to the data, yet the scatter in parameters in the various radial bands and signs of parameter coupling render the results unreliable.
Figure 3 shows the surface brightness profile for NGC 4494 in the range between 25 and 250 arcsec. The lower plot shows the residuals from the best fit de Vaucouleurs profile to highlight deviations from the model. A statistically significant excess can be seen at radii around 80 arcsec, clearly visible as a "bump" in the residual image. This excess might represent the contribution of an underlying faint disk with a radius of ~5kpc.
 |
Figure 3. Upper plot: Surface Brightness profile in I-band with radii ranging from 25 to 250 arcsec. The lines correspond to the best fits for this range (see table 1., marked with a *). Red: King, Blue: Sérsic, Green: de Vaucouleurs. Lower plot: Residuals from the best de Vaucouleurs fit, an excess around 80 arcsec can be seen. |
With the help of published surface photometry (Goodfrooij et al., 1994) it was possible to extract color gradients from the data. Figure 4 shows the radial evolution of the B-I and V-I colors and the corresponding linear fits. No B-I color gradient can be identified, and least square minimization yields a color gradient of d(B-I)/dlog(r) = 0.008±0.021. Nevertheless the V-I color plot exhibits a substantial decrease in color with radius, giving a best fit of d(V-I)/dlog(r) = -0.365±0.076. This is consistent with the general trend seen in elliptical galaxies to get redder toward the center.
 |
Figure 4. Color profiles for NGC 4494 together with the best linear fits. The upper plot shows B-I, the lower plot V-I. Only V-I shows a significant gradient. B and V data are from Goodfrooij et al. (1994). Both color plots have an arbitrary offset to make them fit into one plot. |
Trujillo, Graham and Caon (2001) investigated the possibility of parameter coupling in Sérsic profile fits. They came to the conclusion that the observed trend of a larger Sérsic parameter n corresponding to a larger effective radius is real and not caused by fitting issues. It is interesting to note however, that most of our parameter combinations fall right onto their log(n)-log(re) correlation, except those for only large radii.
The color gradient found for NGC 4494 is somewhat stronger than one found in a recent study by Idiart et al. (2002). Unfortunately, no error was given for their value of -0.019, and no plot was shown. This makes it hard to quantitatively compare both results, especially due to the lack of calibration in our case. It might also be important, that their fitting range is between 5 and 80 arcseconds. As a result, this might cause a lower value, since the gradient is more prominent at larger radii (compare figure 4).
It was possible to adequately measure (S/N > 20) the surface brightness profile down to a level below 3% of the sky at a radius of 3 arcmin. With the help of published surface photometry, a color gradient in V-I of d(V-I)/dlog(r) = -0.365±0.076. could be determined. The B-I color profile is consistent with the absence of any gradient. The light distribution is nicely fit by both a Sérsic or a de Vaucouleurs law with effective radius of 45.77 arcsec. The residuals of the fit uncover a possible disk at a radius of 80 arcsec. The parts of the galaxy that are not PSF dominated (> 25 arcsec) yield an approximately constant ellipticity of 0.16 with an equally constant PA of -4°
The original goal of this project was to adequately measure the radial color gradients in V-R and V-I without having to rely on published surface photometry, calibrate the colors and finally convert these gradients into age-metallicity gradients (Wu et al. 2004). Unfortunately, the weather only permitted a limited amount of observing time in one single band. For similar future studies, it is recommended to consider the following general suggestions: It should be possible to choose the target such that calibrated standard stars fall into the field of view of the object or that of a classmate (e.g. check the new Stetson (2000) standard stars, which are located around well-observed objects). Another point that could considerably improve the data quality is the existence of a bright guide star to increase the auto-guiding accuracy. Besides these general recommendations, it would be very beneficial to this particular project to have a B filter at hand, since the radial color profiles often show the strongest gradients in the B-I color.
I would like to thank my classmates that helped during all the nights of observing and data reduction. A special thanks to Mangala Sharma for her help with IRAF and to the course instructor Tom Statler, who was always available for help and open for discussion. I would also like to mention George Eberts whose backyard we were allowed to use for the observations and our system administrator Don Roth, who successfully kept our computers alive.