next previous
Up: Spatially resolved spectroscopy of


   
4 Spectroscopic observations and data reduction

Long-slit spectra centered on the 5170 Å Mg triplet of the galaxies were taken along their major axes. Table 2 describes the setups used during the 6 runs of observations at the 2.4 m telescope of the Michigan-Dartmouth-M.I.T. (MDM) observatory at Kitt Peak, the 2.7 m telescope at Mc Donald observatory (McD) and the 3.5 m telescope of the German-Spanish Astronomical Center on Calar Alto (CA). Table 3 gives the log of the observations. Depending on the telescope the seeing was either measured by a seeing monitor or by measuring the point spread function (PSF) of a star image in the beginning of the night and controlling it via the guiding probe. The seeing listed in Table 3 is the worst value measured for the individual exposures. As we also needed "velocity standard'' stars to determine the kinematical parameters of the galaxies, we obtained spectra for several G9 to K1 - giant template stars for all three setups. To make sure that the illumination across the slit was uniform and like in the galaxies we trailed and wiggled the stars across the slit. This ensured that the instrumental broadening was always the same in the stars as in the galaxies. For the kinematic analysis itself we finally only used spectra of HR 6817 (K1III) and HD 172401 (KOIII) both from Gonzales (1993, G93). Additionally we observed the flux standard star G191B2B or HD 19228 to calibrate the flux of the spectra before line strength indices were derived. The standard reduction (bias & dark subtraction, flat fielding, wavelength calibration, cleaning for cosmics, sky subtraction, correction for CCD misalignment) was carried out with our own programs running under the image processing package MIDAS provided by ESO and which are described in details in Bender et al. (1994, BSG94). Finally multiple exposures for one galaxy were centered and summed up to increase the final signal-to-noise.

  \begin{figure}
\psfig{figure=1702_f09.eps,width=12cm}\end{figure} Figure 2: The distribution of galaxies as a function of the projected distance D from the cluster center. a) all galaxies; b) Es; c) S0s; d) E/S0. Galaxies with D > 0.71$^{\circ }$(dashed line) belong to the outer Sample


  \begin{figure}
{
\psfig{figure=1702_f10.eps,width=10cm} }
\end{figure} Figure 3: The total galaxy magnitude $R_{\rm KC_{\rm T}}$ as a function of the projected distance D from the galaxy center. Filled triangles indicate cDs, open squares Es, open triangles S0s, open stars E/S0s. Galaxies with D > 0.71$^{\circ }$ (vertical dotted line) belong to the outer sample. The long-dashed line indicates the completeness limit of the inner sample ( $R_{\rm KC_{\rm T}}=12.63$; horizontal dashed line)

The pixel-to-pixel noise in the normalized flat field is < 0.2%. A sky subtraction better than 1% was achieved, while the errors in the wavelength calibration were $\le 0.15$ Å. The variation of the spectral resolution with $\lambda$ was also $\le 0.15$ Å and stable during each observing run ( $\delta \lambda \le 0.05$ Å). The quality of the final spectrum - and hence the resulting S/N - depends on the spectral resolution, the efficiency of the spectrograph as well as on the seeing conditions. Figure 4 shows examples of central spectra covering the 3 quality classes listed in Table 3 for the different instrumental resolutions $\sigma _{\rm inst}$. The spectra are averaged within the radius $r_{\rm L}$, which corresponds to a circular standard aperture having a diameter of $2 r_{\rm A} = 3.4''$ (see Sect. 5). For $\sigma _{\rm inst}$ = 67.9 km s-1 (run 2, 3, 5, 6), the quality parameter is either 1 or 2, while for $\sigma _{\rm inst}$ = 129.4 km s-1 (run 1 & 4) it is only 2 or 3. Because of the low instrumental resolution we excluded the latter ones from having quality 1 by default.

  \begin{figure}
{
\psfig{figure=1702_f11.eps,width=16cm} }
\end{figure} Figure 4: Example spectra averaged within a circular standard aperture (see text) covering the range of quality classes for different instrumental resolutions: a) and b) show spectra with high instrumental resolution ($\sigma $ = 67.9 km s-1) having very good (1) and good (2) quality. There are no spectra of quality worse than two for this instrumental resolutions. c) and d) show spectra with low instrumental resolution ($\sigma $ = 129.4 km s-1) having good (2) and fair (3) quality. Because of the low instrumental resolution we excluded the latter ones to have quality 1 by default. Note that the panels show relative flux and have a false zero point for viewing convenience

In the next step the spectra were rebinned along the slit (spatial direction) to achieve a minimum signal-to-noise (S/N) $\geq$ 40/Å at all radii (for details of the procedure see BSG94). Of course the central pixels have higher S/N (up to a factor of 5). Monte Carlo simulations described by BSG94 showed that a minimum on S/N = 30 - 40 is necessary to derive meaningful kinematical parameters with negligible systematic errors. As a final step of the pre-processing, the galaxy continuum was removed by fitting a forth to sixth order polynomial. We then determined the line-of-sight-velocity-distributions (LOSVDs) by using the Fourier Correlation Quotient (FCQ) method (Bender 1990), which provides the stellar rotational velocities $v_{\rm rot}$, velocity dispersions $\sigma $ and first orders of asymmetric (H3) and symmetric (H4) deviations of the LOSVDs from real Gaussian profiles (van der Marel & Franx 1993; Gerhard 1993; BSG94). As expected, we find that the FCQ method is little influenced by template mismatching (Bender 1990). Following BSG94, Monte Carlo simulations were made to find the best fit-order for the continuum and to check for systematic effects. It turns out that for runs 2, 3, 5, and 6, which have high spectral resolution there are no systematic errors for all kinematic parameters. Even for runs 1 & 4, that have rather low spectral resolution (129.4 km s-1), $v_{\rm rot}$, $\sigma $ and H3 do not show any systematic errors. Only the derived H4 amplitude for this setup turns out to be systematically negative for S/N $\le$ 80. At S/N = 40 the systematic error is -0.04 and hence is of the order of the error bars (see below). We corrected for this systematic error using the dependence of H4 on S/N delivered by the Monte Carlo simulations. The error bars were derived from photon statistics and CCD read-out noise and were calibrated via Monte Carlo simulations described by Gerhard et al. (1998): Noise is added to template stars, broadened following the observed values of H4 and $\sigma $, matching the power spectrum noise to peak ratio of the galaxy. The accuracy of the estimated error bars is about 20%.

Additionally we tested whether our estimated kinematic errors indeed reproduce the RMS scatter between different exposures of the same galaxy. For all setups the expected errors - derived via our Monte Carlo simulations and shown in Fig. 6 - are of the order of the RMS. In fact for most galaxies the errors were slightly larger than the RMS by a factor of $\approx$ 1.2. In a few cases the estimated errors were overestimated up to a factor of 2.

Mg, Fe and H$\beta $ line strength indices were derived following Faber et al. (1985) and Worthey (1992) from flux calibrated spectra, rebinned radially as before. For the 11 galaxies observed with the MDM setup (run 1 and 4), that covers a larger $\lambda$ range, we could in addition derive the NaD line index profiles. We corrected all measured indices for velocity dispersion broadening and calibrated our measurements to the Lick system using stars from Faber et al. (1985). The errors are derived from photon statistics and CCD read-out noise. As for the kinematic parameters we tested whether our estimated errors indeed reproduce the RMS scatter between different exposures of the same galaxy. Again, for all setups the estimated errors - derived via our Monte Carlo simulations and shown in Fig. 6 - are of the order of the RMS. For most galaxies the errors were slightly larger than the RMS by a factor of $\approx$ 1.2 and in a few cases the estimated errors were overestimated up to a factor of 2.

In the following we indicate the average Iron index with $\langle$Fe$\rangle=$ (Fe5270+Fe 5335)/2 (Gorgas et al. 1990) and the usual combined Magnesium-Iron index with [MgFe $]=\sqrt{\hbox{\rm Mg}b \langle \hbox{\rm Fe}\rangle} $ (G93).

Some of the line index profiles needed further treatment. At small radii some of them show unreal features caused by the varying focus of the spectrograph in the dispersion direction. Following G93, this effect is only detectable if the variation of the focus' point spread function (PSF) is dominant compared to the atmospherical seeing (FWHM). For some of the galaxies the focus variation was the dominant effect and we followed the procedure described in Mehlert et al. (1998, MSBW98) to correct it. The Mg1and Mg2 indices defined by Faber et al. (1985) are affected most, because their pseudo continua are $\approx$ 200 Å apart from the line windows. For all the other line indices the continuum and line windows are close to each other. H$_\beta$ and the Fe indices are in the red and blue part of the spectra and thus slightly affected, while the Mgb index at the central wavelength is almost unaffected.

Recent investigations showed evidence for small amount of ionized gas and dust in the interstellar medium (ISM) of many elliptical galaxies (e.g. Bregman et al. 1992; Goudfrooij et al. 1994). Consequently, the detection of some emission (e.g. H$\beta $, N, and O) in their spectra is expected. Since the velocities of the stars and gas in the galaxies may differ by up to 100 km s-1 (Bertola et al. 1995) asymmetric contamination of line indices is possible. In addition the dust and ionized gas show a wide range of distribution: Smoothly along their major/minor axis or in rather patchy or filamentary features (Goudfrooij et al. 1994). If one wants to investigate the stellar population of elliptical galaxies (as we are going to do in Paper II) the contamination of H$\beta $ by emission is especially crucial. Emission will weaken the age indicating H$\beta $ absorption index and hence lead to an overestimate of the age of the dominant stellar population (see Worthey 1994, W94). Simulations with the kinematical template stars we used showed that we can detect emission in H$\beta $ for equivalent width larger than about 0.3 Å. Figure 5 shows examples of the strong (a) and weak (b) H$\beta $ emission seen in the our spectra. We will refer to these two classes in the comments on the galaxies given in Sect. 5. In particular Fig. 5a demonstrates how asymmetric H$\beta $ emission can be. Since no OIII emission is detected in this galaxy (GMP 4315) at all, these data illustrate that it is impossible to correct for H$\beta $emission via OIII as suggested by G93.

Additionally Goudfrooij & Emsellem (1996) showed that NI emission may influence the measured Mg indices ($\Delta $Mgb up to 0.5 Å and $\Delta $Mg2 up to 0.03 mag). The influence of asymmetric NI emission can be seen in the Mgb profile of GMP 2390 (galaxy 6 in Fig. 6).


   
Table 2: Setups of the spectroscopic runs. The spectral resolutions in km s-1 were derived at the 5170 Å Mg triplet
Run Date Telescope Detector $\lambda$ - range Scale Slit- spectral
      Spectrograph [Å] [''/pix] width resolution $[\sigma]$
1 3/95 MDM TI: 1024 $\times$ 1024 4300-6540 0.777 1.7'' 2.23 Å
4 3/96 2.4 m Mark III       129.4 km s-1

2

4/95 McD TI: 800 $\times$ 800 4850-5560 0.635 2.5'' 1.17 Å
5 4/96 2.7 m LCS       67.9 km s-1
3 5/95 CA TI: 1024 $\times$ 1024 4730-5700 0.896 3.6'' 1.17 Å
6 5/96 3.5 m TWIN/R       67.9 km s-1


   
Table 3: Log of the spectroscopic observations. Galaxy numbers as in Table 1. For galaxy No. 3 the spectra from run 3 and 6 are summed up; for galaxy Nos. 9 and 12 the spectra from run 1 and 4 are summed up. The seeing has been measured for each single exposure. The maximum value detected in the most broadend spectrum for one galaxy is listed. The parameter Q estimates the quality of the spectra: 1, very good; 2, good; 3, medium (see Fig. 4)
No. GMP No. NGC/IC No. DN Run Single exp. times Total exp. times PA Seeing Q
          [s] [s] (N $\geq$ E) (FWHM)  
1 3329 NGC 4874 129 2 $ 3 \times 3600 $ 10800 45$^{\circ }$ 1.8'' 2
2 2921 NGC 4889 148 2 $ 2 \times 3600 + 1800 $ 9000 81$^{\circ }$ 2.2'' 2
3 4928 NGC 4839 31 1 $2 \times 3600$ 7200 63$^{\circ }$ 1.8'' 3
        2 $2 \times 2700 + 3350$ 8750 63$^{\circ }$ 2.0'' 3
        3 1800   63$^{\circ }$   1
        6 1800 3600 63$^{\circ }$ 2.8'' 1
4 4822 NGC 4841A 240 1 $2 \times 3600 + 2400$ 9600 105$^{\circ }$ 1.5'' 2
5 1750 NGC 4926 49 1 3600 + 4500 + 2700 10800 60$^{\circ }$ 1.8'' 2
6 2390 IC 4051 143 1 $2 \times 3600 + 2000 + 1600 $ 10800 106$^{\circ }$ 1.8'' 2
7 2795 NGC 4895 206 1 $ 3 \times 3600 $ 10800 156$^{\circ }$ 1.5'' 2
8 3792 NGC 4860 194 1 3600 + 1520 + 3000 + 2400 10520 127$^{\circ }$ 1.4'' 2
9 2629 NGC 4896 232 1 $ 2 \times 3600 +2010 $   7$^{\circ }$   2
        4 3000 + 2850 15060 7$^{\circ }$ 1.8'' 2
10 3561 NGC 4865 179 1 $ 3 \times 3000 + 1800 $ 10800 115$^{\circ }$ 1.8'' 2
11 2000 NGC 4923 78 3 $2 \times 3600$ 7200 77$^{\circ }$ 2.2'' 1
12 2413 - 230 1 2700 + 2050 + 3000   25$^{\circ }$   2
        4 $ 2 \times 3000 $ 13750 25$^{\circ }$ 1.8'' 2
13 4829 NGC 4840 46 1 $ 3 \times 3300 $ 9900 106$^{\circ }$ 1.5'' 2
14 3510 NGC 4869 105 3 $2 \times 3600$ 7200 169$^{\circ }$ 2.5'' 1
15 2417 NGC 4908 167 3 $2 \times 3600$ 7200 55$^{\circ }$ 2.5'' 1
16 2440 IC 4045 168 3 2300 + 2500 + 2000 6800 108$^{\circ }$ 2.2'' 1
17 3414 NGC 4871 131 3 3120 + 3200 + 2400 8700 178$^{\circ }$ 2.8'' 1
18 4315 NGC 4850 137 3 $2 \times 3600$ 7200 153$^{\circ }$ 2.2'' 1
19 3073 NGC 4883 175 3 $2 \times 3600$ 7200 106$^{\circ }$ 2.2'' 1
20 1853 - 190 4 $ 3 \times 3600 $ 10800 88$^{\circ }$ 1.6'' 2
21 3201 NGC 4876 124 4 $ 2 \times 4500 + 3600 $ 12600 24$^{\circ }$ 1.7'' 2
22 3661 - 13 5 $ 4 \times 3600 + 2400 $ 16800 139$^{\circ }$ 1.4'' 1
23 4679 - 75 5 $ 2 \times 4800 + 5700 $ 15300 114$^{\circ }$ 1.6'' 1
24 3352 NGC 4872 130 5 $ 2 \times 5400 + 3600 $ 14400 111$^{\circ }$ 1.5'' 1
25 2535 IC 4041 145 5 $ 3 \times 4800 $ 14400 48$^{\circ }$ 1.3'' 1
26 3958 IC 3947 72 5 $ 3 \times 4800 $ 14400 102$^{\circ }$ 1.5'' 1
27 2776 - 39 5 $ 3 \times 4800 $ 14400 77$^{\circ }$ 2.3'' 1
28 0144 NGC 4957 - 6 4300 + 4500 8800 91$^{\circ }$ 1.5'' 1
29 0282 NGC 4952 - 6 4100 + 5000 9100 135$^{\circ }$ 2.5'' 1
30 0756 NGC 4944 - 6 $ 2 \times 4500 $ 9000 88$^{\circ }$ 2.0'' 1
31 1176 NGC 4931 - 6 $ 2 \times 4500 $ 9000 78$^{\circ }$ 2.0'' 1
32 1990 - - 6 5100 + 4500 9600 135$^{\circ }$ 2.2'' 1
33 5279 NGC 4827 - 6 4500 + 5330 9830 56$^{\circ }$ 1.5'' 1
34 5568 NGC 4816 - 6 3600 + 3400 7000 78$^{\circ }$ 1.2'' 1
35 5975 NGC 4807 - 6 4500 + 3000 7500 23$^{\circ }$ 2.0'' 1


  \begin{figure}
{
\psfig{figure=1702_f12.eps,width=10cm} }
\end{figure} Figure 5: Examples of the two classes of H$\beta $ emission strength detected in our spectra (solid line): a) strong and b) weak emission. For comparison we also show the H$\beta $ absorption line at the opposite sides of the galaxies, where no emission is detected (dashed-dotted line). The vertical dashed lines indicate the redshifted position of the H$\beta $ absorption window. The arrow indicates the position of the detected emission. In particular panel a) also shows the strong asymmetry of the H$\beta $ absorption in GMP 4315. While strong H$\beta $ emission exists at  r =+2.81'', there is none at all at the opposite side of the galaxy (r=-2.57''; dashed-dotted line). Since no OIII emission is detected in this galaxy at all, these data also demonstrate, that it is not possible to correct for H$\beta $ emission via OIII. Note that the panels show relative flux and have a false zero point for viewing convenience


next previous
Up: Spatially resolved spectroscopy of

Copyright The European Southern Observatory (ESO)