All Figures

$\begin{figure} \par\includegraphics[width=12cm,clip]{9148f01.ps}\end{figure}$ Figure 1: Performance of the photometric redshifts used for the spectroscopic target selection (not the final ones from Pelló et al., in prep.). Plotted are the members of the cluster(s) at the targeted redshift(s) for the given field (21 clusters in 19 fields). For example, if the target selection was $z_{\rm phot}= 0.58\pm 0.2$ for the given field (cf. Table 1), then the z = 0.58 cluster in that field is plotted, but not the z = 0.43 cluster. Four galaxies out of 568 are outside the plotted y-range. Abbreviated cluster names are given on the plot. The 5 clusters not plotted are cl1037a (z = 0.43), cl1103a (z = 0.63), cl1103 (z = 0.96), cl1138a (z = 0.45) and cl1227a (z = 0.58).
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=17cm,clip]{9148f02.ps}\end{figure}$ Figure 2: Flowchart for the science frames. The upper branch is for traditional sky subtraction: a) ``raw'' frame; b) after removal of spatial curvature; c) after flat fielding; d) after cutting-out the individual spectra; e) after application of the 2D wavelength calibration; f) after sky subtraction. The lower branch is for the improved sky subtraction: a) ``raw'' frame; g) after flat fielding; h) after sky subtraction; i) after removal of spatial curvature; j) after cutting-out the individual spectra; k) after application of the 2D wavelength calibration. See text for details. We note that the figure is schematic: it is not to scale, and only 3 spectra per mask (instead of $\sim$ 30) and 1 skyline per spectrum (instead of >100) are shown.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=17cm,clip]{9148f03.ps}\end{figure}$ Figure 3: Illustration of the irregular grid. Panels a) and b) show regular grids (the type of images usually used in astronomy), i.e. with the data points located at integer-valued coordinate positions. Panel c) shows an irregular grid, i.e. with the data points located at real-valued coordinate positions. The different panels show: a) Part of a ``raw'' frame centered on a bright skyline (no galaxy continuum is present in the shown section), i.e. data that have not been interpolated (rebinned), cf. Fig. 2a. The image is pixelised in (x,y). b) The same data after interpolation in y to remove spatial curvature and in x to apply the wavelength calibration, cf. Fig. 2e. The image is pixelised in $(x_{\rm r},y_{\rm t})$ . c) The exact data values from panel a), i.e. uninterpolated values, but shown as an irregular grid in $(x_{\rm r},y_{\rm t})$ space. d) Just a different visualisation of the irregular grid in panel c), corresponding to painting the spectrum in panel a) on a piece of rubber and stretching it to have axes corresponding to $(x_{\rm r},y_{\rm t})$ space. e) The values from panel b) plotted versus $x_{\rm r}$ . We note the scatter caused by the interpolation of the sharp edges of the skyline. f) The values from panel c) plotted versus $x_{\rm r}$ . We note the very low scatter.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=15cm,clip]{milvang_Fig4.eps}\end{figure}$ Figure 4: Examples of the results from the improved and the traditional sky subtraction. For each example, we show 4 panels, e.g. a)- d). The panels show: a) ``Raw'' data (i.e. combined but uninterpolated frame). b) Sky-subtracted data (improved method), still uninterpolated. c) Sky-subtracted data (improved method), interpolated (rectified). d) Sky-subtracted data (traditional method), interpolated (rectified). An identical greyscale is used in panels b)- d). We note that all the panels show regular grids. The first two panels (e.g. a) and b)) are pixelised in (x,y), and the last two panels (e.g. c) and d)) are pixelised in $(x_{\rm r},y_{\rm t})$ and are thus rectified: wavelength is on the abscissa and spatial position is on the ordinate. The greyscale varies from example to example. Examples 1-3 show spectra from tilted slits (slit angles $26.6^\circ$ , $-3.4^\circ$ and $-40.0^\circ$ , resp.). Examples 4-5 show spectra from untilted slits, but in example 5 the skyline is nevertheless slightly tilted. For spectra with tilted skylines the improved sky subtraction method gives better results than the traditional one. We note that the tilt of the emission lines, seen in the rectified frames in examples 1-3, is due to the rotation of the galaxies. Example 2 shows the [OII] emission line of a z=0.7 galaxy, and the corresponding 1D spectrum is shown in Figs. 5a, b.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=8.8cm,clip]{9148f05.ps}\end{figure}$ Figure 5: Comparison of the traditional and the improved sky subtraction method for two typical, one-dimensional cluster spectra extracted from tilted slits (slit angles of $-3.4^\circ$ and $-40.0^\circ$ , resp.). The spectra are flux-calibrated and telluric-absorption corrected (see Sect. 3.4). Fluxes are given in units of $10^{-18}~{\rm erg}~{\rm cm}^{-2}~{\rm s}^{-1}~{\rm \AA}^{-1}$ . For objects 1 and 2 the significant improvement of [OII] and the G-band, resp., is indicated. The [OII] doublet of object 1 is also shown in Fig. 4 as example 2. The spectra are shown at their native pixel size of 1.6 Å (spectral resolution is 6 Å FWHM) without any smoothing. In order to mark the positions of strong skylines, panel e) shows a representative sky spectrum.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=8.5cm,clip]{9148f06.eps}\end{figure}$ Figure 6: Illustration of the method used to create a flux calibration which is valid for all slit positions. Panel a) shows 3 spectra of the same star (LTT7379), obtained through slits placed at the extreme right (blue), the centre (green) and the extreme left (red). Panel b) shows screen flat spectra taken at the same 3 slit positions. The spectral shape has been left untouched, and only the level has been normalised by dividing all 3 spectra by the same constant (cf. the dotted lines; see also the text). Panel c) shows the star spectra (from panel a)) divided by the screen flats (from panel b)), i.e. the panel shows the star SED divided by the screen flat lamp SED, a ratio that is used as a tool - the screen flat lamp SED cancels out at the end of the flux calibration procedure. Panel d) shows the 3 ``spectra'' from panel c) merged.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=8.5cm,clip]{9148f07.ps}\end{figure}$ Figure 7: Illustration of 2nd order contamination or lack thereof. a) Part of a raw MXU arc frame. Most of the lines are from the 1st order, but the 2 lines displaced upwards by 4 px are from the 2nd order. b)- e) 2D spectra of four objects. f)- i) The corresponding spatial profiles. The figure illustrates that while a modest 2nd order contamination is present (redwards of 8000 Å) in the spectra of the standard star used to establish the flux calibration (LTT7379, panel e)), no 2nd order contamination is seen in the galaxy spectra (cf. panel c)) due to their redder observed-frame SEDs.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=8.4cm,clip]{9148f08a.eps}\hspace*{4mm} \includegraphics[width=8.4cm,clip]{9148f08c.eps}\end{figure}$ Figure 8: Comparison of colours synthesised from the spectra with colours from the photometry. Panels a) and b) concern (V-R), while panels c) and d) concern (R-I). To generate this figure, we only used spectra that spanned the wavelength range 4800-7500 Å for panels a) and b), and 5700-8700 Å for panels c) and d). The different width of these minimum wavelength ranges explains the different number of plotted points. Only galaxies at z = 0.15-1.05 are shown. Only spectra from fields with VRI photometry are used. Most of these are from run 3; the few spectra from run 4 (from 2 masks) are shown with crosses. All colours and magnitudes have been corrected for Galactic extinction and are on the AB system (Oke & Gunn 1983). The photometric colours and magnitudes have been measured within a circular aperture of radius 1'' in images corrected to the same fiducial seeing (cf. White et al. 2005). The used transformations to AB are $V_{\rm AB} = V_{\rm Vega} + 0.036$ , $R_{\rm AB} = R_{\rm Vega} + 0.216$ , and $I_{\rm AB} = I_{\rm Vega} + 0.438$ .
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=15.2cm,clip]{9148f09.ps}\end{figure}$ Figure 9: Redshift histograms for the 10 mid-z fields. The clusters for which we have measured a redshift and a velocity dispersion (Halliday et al. 2004, or this paper) are indicated with the $\pm$ $3\sigma _{\rm cl}$ range shown. The labels are ``M'' for the main cluster and ``a'' or ``b'' for secondary clusters. Both 0-colon redshifts (``secure'') and 1-colon redshifts (``secure but with larger uncertainties'') have been used in the plot. The binsize in z, $\Delta z$ , varies with z in such a way that the binsize in rest-frame velocity, $\Delta v_{\rm rest} = c \Delta z/(1+z)$ , is kept constant at 1000 km s^-1. This is achieved binning in $\log~(1+z)$ space with a constant binsize of $\log~(\Delta v_{\rm rest}/c+1)$ .
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=15.45cm,clip]{9148f10.ps}\end{figure}$ Figure 10: Redshift histograms for the 10 high-z fields. Otherwise this figure is similar to Fig. 9.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=17cm,clip]{9148f11.ps}\end{figure}$ Figure 11: The target selection process as a function of magnitude for the EDisCS fields with long spectroscopic exposures plus the 1238 field. For each panel, the figure shows: solid histogram: objects in the photometric catalogue, dashed histogram: targets, dotted histogram: observed targets, long-dashed histogram: targets with a secure redshift, light red filled histogram: galaxies that are members of any cluster in the field (cf. Table 5), dark red filled histogram: galaxies that are members of the cluster(s) at the redshift(s) targeted by the $z_{\rm phot}$ -based target selection. The dotted and the long-dashed histograms often coincide, indicating that a secure redshift was obtained for all the observed targets. All numbers have been computed within the region on the sky spanned by the spectroscopic observations, with the range in x being given on the panels, and the range in y being 100-1950 px, cf. the xy plots (Fig. 16). The cluster redshifts are given on the figure. If more than one value is given, the order is: main cluster, secondary cluster ``a'', secondary cluster ``b''. Values in parentheses are for clusters that are not at the redshift(s) targeted by the $z_{\rm phot}$ -based target selection.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=17.5cm,clip]{9148f12.ps}\end{figure}$ Figure 12: The completeness as function of magnitude for the EDisCS fields with long spectroscopic exposures plus the 1238 field. The completeness is defined as $N_{{\rm secure}~z}/N_{\rm targets}$ , with $N_{{\rm secure}~z}$ being the number of targets for which a secure redshift was obtained (long-dashed line in Fig. 11), and $N_{\rm targets}$ being the number of targets (dashed line in Fig. 11). The shown error bars are too large, since they were calculated assuming that the errors on $N_{{\rm secure}~z}$ and $N_{\rm targets}$ are uncorrelated.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=17.5cm,clip]{9148f13.ps}\end{figure}$ Figure 13: Comparison of the cluster velocity dispersions determined using the 5 methods that we have tested. From left to right for each cluster, the methods are: method 1, method 2₂₀₀, method 2₃₀₀, method 2₅₀₀ and method 2₁₀₀₀(see Sect. 6). For most clusters, the results from all methods agree, but for 6 clusters marked with ``(*)'' this is not the case. For these clusters, the 5 estimates of $\sigma _{\rm cl}$ are illustrated in Fig. 14.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=17.2cm,clip]{9148f14.ps}\end{figure}$ Figure 14: Histograms of peculiar velocities in the cluster rest-frame, $v_{\rm pec}^{\rm rest}= c (z - z_{\rm cl}) / (1 + z_{\rm cl})$ , for the 6 clusters for which the 5 methods of measuring the cluster redshifts and velocity dispersions do not all agree (cf. Fig. 13). Each row shows a given cluster. We note that all panels in a given row are based on the same redshift data, but since the histograms show $v_{\rm pec}^{\rm rest}$ which depends on the cluster redshift $z_{\rm cl}$ , the histograms do not always look completely identical. Each column shows a given method, from left to right: method 1, method 2₂₀₀, method 2₃₀₀, method 2₅₀₀ and method 2₁₀₀₀. For the analysis we adopt method 2₃₀₀. The remaining features of the plots are described in the caption of Fig. 15.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=16.7cm,clip]{9148f15.ps}\end{figure}$ Figure 15: Histograms of peculiar velocities in the cluster rest-frame, $v_{\rm pec}^{\rm rest}= c (z - z_{\rm cl}) / (1 + z_{\rm cl})$ , for the 26 EDisCS clusters. The solid histograms are for galaxies having redshifts without colons (indicating ``secure'' redshifts). The dashed histograms include the 1-colon redshifts (``secure but with larger uncertainties''). The dotted histograms include the 2-colon redshifts (``not secure''). The binsize is 250 km s^-1and the plotted range is $\pm$ 4000 km s^-1. The overplotted Gaussians illustrate the measured velocity dispersion $\sigma _{\rm cl}$ of the given cluster. The vertical dot-dashed lines indicate $\pm$ $3\sigma _{\rm cl}$ , the limits used to define cluster membership. The number of cluster members having redshifts without colons is given as $N_{\rm mem,~0}$ , and the area underneath the Gaussian corresponds to this number. The red skeletal arrows are located at $v_{\rm pec}^{\rm rest}= 0$ km s^-1 and thus indicate the adopted cluster redshifts. The blue filled arrows indicate the adopted BCGs (except where no redshift is available).
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=16.2cm,clip]{9148f16.ps}\end{figure}$ Figure 16: xy plots for the 26 EDisCS clusters. North is up and east is to the left. The units on the axes are pixels = 0.2''. Only galaxies with no colons on their redshifts are shown. The small dots are the non-members. The large symbols are the cluster members. Depending on in which bin $v_{\rm pec}^{\rm rest}$ falls, the symbols are: blue triangles: $[-3\sigma _{\rm cl},-1\sigma _{\rm cl}[$ , green circles: $[-1\sigma _{\rm cl},+1\sigma _{\rm cl}]$ , red squares: ] $+1\sigma _{\rm cl},+3\sigma _{\rm cl}]$ . The cross indicates the adopted BCG, which in the case of cl1059.2-1253 and cl1037.9-1243 does not have any spectroscopy, cf. White et al. (2005). 1 Mpc bars are shown for the assumed cosmology ( $\Omega _{\rm m} = 0.3$ , $\Omega _\Lambda = 0.7$ and $H_0 = 70~{\rm km}~{\rm s}^{-1}~{\rm Mpc}^{-1}$ ).
Open with DEXTER

In the text

$\begin{figure} \makebox[\textwidth]{ \makebox[\fourthwidth]{ \includegraphics[... ... } \makebox[\fourthwidth]{ ~ } \makebox[\fourthwidth]{ ~ } } \end{figure}$ Figure 17: Dressler-Shectman (DS) plots. The DS analysis has only been performed on clusters with at least 20 members. The plots show the x,y location of the cluster members. The radii of the plotted circles are equal to $e^{{\delta }/{2}}$ where $\delta$ is the DS measurement of local deviation from the global velocity dispersion and mean recession velocity (cf. Eq. (3)). The blue/green/red circles (also shown as dotted/solid/hashed) indicate velocity in the same way as in the xy plots (Fig. 16). The probability P given on the figure is the probability of there being no substructure in the dataset; thus, a small value (e.g. less than 0.05) indicates that substructure has been detected. The number of members is also given on the figure; only redshifts without colons have been used. The 9 clusters in this figure are shown in the same order as in Table 6. We note that DS plots for 5 more clusters are found in Halliday et al. (2004).
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=8.5cm,clip]{9148f18.ps}\end{figure}$ Figure 18: Comparison of the velocity dispersions obtained from the weak lensing analysis (Clowe et al. 2006), $\sigma _{\rm lens}$ , with the velocity dispersions obtained from the spectroscopy (Halliday et al. 2004 and this paper), $\sigma _{\rm spec}$ . The figure shows the 19 main EDisCS clusters ( z = 0.42-0.96). The blue triangles are the clusters with other structures near enough to possibly affect the lensing measurements (Clowe et al. 2006), and the red squares are the rest of the clusters. The 3 circled clusters are those for which the Dressler-Shectman test gives a significant detection of substructure (Halliday et al. 2004 and this paper). The dotted line shows the one to one correspondence. Abbreviated cluster names are given on the figure. The 2 major outliers of the blue triangles are cl1059.2-1253 and cl1103.7-1245, both of which were identified in Clowe et al. (2006) as having extremely high mass-to-light ratios. The 3 major outliers of the red squares are cl1420.3-1236, cl1054.7-1245 and cl1054.4-1146.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=8.5cm,clip]{9148f19.ps}\end{figure}$ Figure 19: The distribution of velocity dispersion $\sigma$ vs. lookback time for EDisCS and for two other well-studied cluster samples at similar redshifts, as well as for a well-studied local sample. The figure shows: SDSS (blue histogram) at z < 0.1, MORPHS (red circles) at 0.37 < z < 0.56, EDisCS (black triangles) at 0.40 < z < 0.96, and ACS GTO (blue squares) at 0.8 < z < 1.3. The EDisCS clusters fill the gap in lookback time between the MORPHS and the ACS GTO clusters and have a large range in $\sigma$ . The dashed lines show how $\sigma$ is expected to evolve with redshift from z=1 to z=0 (Poggianti et al. 2006). From these curves it is apparent that EDisCS is a high redshift cluster sample for which a majority of the clusters can be progenitors of ``typical'' low redshift clusters. References for the plotted velocity dispersions: SDSS: von der Linden et al. (2007); MORPHS: Girardi & Mezzetti (2001); EDisCS: Halliday et al. (2004) and this paper; ACS GTO: Gioia et al. (2004); Demarco et al. (2004,2005); Gal & Lubin (2004); see also Postman et al. (2005). The shown SDSS sample contains 488 clusters selected to have $\sigma > 200~{\rm km}~{\rm s}^{-1}$ and $\sigma /{\rm uncertainty}(\sigma ) > 4$ as measured by von der Linden et al. (2007) for a subset of the C4 cluster sample originally compiled by Miller et al. (2005).
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=17.8cm,clip]{9148fa1.ps}\end{figure}$ Figure A.1: Illustration of the performance of the two sky subtraction methods for the spectra from all the masks and for all wavelengths. The figure shows histograms of $I_{\rm skysubtr}/\sigma _{\rm CCD}$ , where $I_{\rm skysubtr}$ are the pixel values in the background-subtracted 2D galaxy spectra, and $\sigma _{\rm CCD}$ are the corresponding 2D spectra giving the standard deviation expected from the CCD noise model (photon noise and read-out noise; Eqs. (A.1) and (A.4)). Only pixels in the manually-determined background regions were used to make the histograms, thus excluding practically all signal from known objects in the spectra (galaxies, stars). The histograms therefore illustrate the scatter caused by both natural noise sources (photon noise and read-out noise) and by possible imperfections in the sky subtraction. Solid/blue histograms: improved sky subtraction (i.e. sky subtraction performed on the unrebinned data); dash-dotted/red histograms: traditional sky subtraction (i.e. sky subtraction performed on the rebinned data). Dotted/green curves: Gaussians with $\sigma = 1$ , for reference. Panels a)- c) show unrebinned data (where sky-subtracted frames are only available for the improved sky subtraction) while panels d)- f) show rebinned data (where sky-subtracted frames are available for both types of sky subtraction). Panels a)+ d) show data from all slits, while the data have been split in tilted and untilted slits in panels b)+ e) and c)+ f), respectively. The main conclusions from this figure are: (i) The result from the improved sky subtraction is close to the CCD noise limit, since the data in panel a) [solid/blue curve] agree so well with a $\sigma = 1$ Gaussian [dotted/green curve]; (ii) The improved sky subtraction is better than the traditional one, with a large improvement for tilted slits (panel e)) and a smaller improvement for untilted slits (panel f)).
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=17.5cm,clip]{9148fa2.ps}\end{figure}$ Figure A.2: Illustration of the performance of the two sky subtraction methods for the spectra from all the masks as function of wavelength. Thedata are plotted against wavelength in bins of 1.6 Å. Panel a) shows a sky spectrum for reference. Panel b) shows $\sigma (I_{\rm skysubtr}/\sigma _{\rm CCD})$ , which is the standard deviation of the $I_{\rm skysubtr}/\sigma _{\rm CCD}$ values in the given bin. Results from both sky subtraction methods are plotted. A value of 1 representsthe noise floor set by photon noise and read-out noise. Panel c) shows the ratio of $\sigma (I_{\rm skysubtr}/\sigma _{\rm CCD})$ for the two methods, illustrating that the traditional sky subtraction has several times more noise than the improved sky subtraction at the locations of skylines. This plot is based on all masks; had we only plotted the masks with the longest exposure times (and hence the largest number of collected sky counts) the difference between the two methods would have been larger. Panels b) and c) are for spectra from tilted slits, whereas panels d) and e) are for spectra from untilted slits.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=8.5cm,clip]{9148fa3.ps}\end{figure}$ Figure A.3: Illustration of the fact that the traditional sky subtraction method has the highest extra noise at the edges of the skylines, i.e. where the gradient in the sky background is the largest (cf. Kelson 2003). This figure is akin to a zoom of Fig. A.2 centered at the strong 6300 Å skyline, but only data from a single mask have been used (and we note that the y-axis range for panel b) has been increased). If this figure had been made using all 51 masks it would have looked rather similar, but the peaks in panel b) and c) would not have been so sharp due to the small wavelength shifts that exist between the masks due to instrument flexure. The dotted lines indicate the wavelength region used for the statistics shown in Fig. A.4.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=8.5cm,clip]{9148fa4.ps} \end{figure}$ Figure A.4: Illustration of how the noise in the sky-subtracted spectra depends on the collected sky counts. Only data in the narrow wavelength range $6300\pm8~$ Å have been used (cf. Fig. A.3). Data from all 51 masks have been used. The x-axis shows the mean collected sky counts over the total exposure time for the given spectrum, with the mean being taken over the used wavelength range. The quantity on the x-axis thus depends linearly on the total exposure time and on the sky brightness at 6300 Å. The y-axis shows $\sigma (I_{\rm skysubtr}/\sigma _{\rm CCD})$ which was also used in Figs. A.2 and A.3, just here computed in the single wavelength bin of $6300\pm8~$ Å instead of in multiple bins of 1.6 Å. The quantity on the y-axis is the noise relative to $\sigma _{\rm CCD}$ (the noise expected from photon noise and read-out noise). The horizontal dot-dashed line at 1 represents thenoise floor set by photon noise and read-out noise. The points for the traditional sky subtraction show a square root like behaviour. Since the quantity on the y-axis has already been divided by $\sigma _{\rm CCD}$ (which essentially is proportional to the square root of the quantity on the x-axis) the plot indicates that the extra noise in the traditional sky subtraction goes linearly with the number of collected sky counts.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=15.45cm,clip]{9148f10.ps}\end{figure}$	Figure 10: Redshift histograms for the 10 high-z fields. Otherwise this figure is similar to Fig. 9.
Open with DEXTER