Online material
Appendix A: Details on the fitting procedure
The fitting procedure was performed as follows: for each pointing, we performed a 2 × 2 binning of all the spaxels to improve the signaltonoise ratio of the resulting spectra. Then, for each spaxel, we fitted the spectral range [6620, 6780] Å to a set of Gaussians corresponding to the emission lines [Nii]λ6548, 6583 Å and Hα, under the following conditions:

The intensities of the four emission lines must always be positiveor zero (no underlying absorption in Hα is assumed). For Hα we only allowed an intensity higher than zero.

The recession velocity of the lines is restricted to the interval [5400, 6900] km s^{1} (this seems reasonable since the recession velocities of all the galaxies of SQ are contained within this range).

The theoretical relation [Nii]λ6548 / 6583 = 0.333 is preserved.

The velocity dispersion (σ) of the three emission lines is forced to be the same, > 1.35 Å (which is the nominal dispersion of the spectra), and < 4.5 Å. This last restriction is imposed to avoid nonphysical solutions that lead to arbitrarily broad lines for some spectra.
The continuum was estimated before the fit, it was assumed to be the same for the three emission lines and to be equal to the median flux in the spectral interval [6600, 6660] Å ∪ [6750, 6800] Å. The background uncertainty (Σ_{bkg}) was taken to be equal to the standard deviation of the flux along this interval. After several tries, this was found to be a proper estimation given the almost null slope of the continuum along this spectral range. The fit was performed with the IDLbased routine MPFITEXPR (Markwardt 2009). This code requires a set of initial input parameters (I_{Hα}, I_{6548}, v_{rec} and σ) and iteratively finds the solution that matches the spectrum best.
As mentioned in a previous section, Figs. 2 to 4 suggest the possibility that in some spaxels the emission comes from more than a single kinematical component. To decide the number of components, we first tried a fit with a single component. Figure 12 shows the stacked spectra of the residuals after a
onecomponent fit for our three pointings. In the three cases the stacked residuals show an emission feature around 6710 Å. This feature is broad for pointings S and M, and narrow and welldefined in pointing N. These residual features could imply that we are missing another component, so we tried the fitting procedure again with two components. Figure 13 shows stacked residuals after this twocomponent fit and in this case we do not see any emission feature, the result being reasonably flat. For this reason, we decided to fix the number of kinematical components at two. Figures 14 to 16 show the residuals in the [6650, 6750] Å spectral interval, after the subtraction of the twocomponent Hα + [Nii] fitting. The residuals are quite flat, implying that our choice is reasonable.
Once the fit of these three lines was made, we fitted the [Oi]λ6300 Å line in the red spectra and the [Oiii]λ4959, 5007 Å and Hβ lines in the blue spectra to Gaussians with the redshifts and velocity dispersions resulting from the fit to the corresponding red spectra.
To illustrate the quality of the fits, we show in Figs. 17 to 19 three examples of typical fits of each pointing. Spaxel S[2, 2] was fitted to a single component, corresponds to one of the Hii regions and shows a typical Hiilike spectrum. Spaxels M[3, 4] and N[4, 4] were fitted to two components and correspond to the shocked region. The red spectra show broad lines unlike the blue spectra where few emission lines were detected. Figure 17 shows a conpicuous asymmetric residual at ≈ 6685 Å, coincident with the strong Hα line detected in spaxel S[2, 2]. This residual is unexplained and because we do not see it for other bright Hα lines or arcs, it is unlikely be caused by an instrumental effect.
To avoid spurious results due to low signaltonoise, we kept throughout this paper only those components for which the intensity peak of the Hα line is above 5 × Σ_{bkg}. This means that the maximum number of components detected for individual spaxels is 2, but could be 1 or even zero depending on the signaltonoise ratio of the Hα line. For the remaining lines, we fitted a Gaussian with σ and redshift equal to those of the corresponding Hα line, and we measured the fluxes of all lines for which the intensity peak is above 1 × Σ_{bkg}.
Log of the observations.
Properties of the emission lines for the spaxels of pointing S.
Oxygen abundances for the brightest spaxels showing Hiilike spectra in pointings S and M, estimated following the Pettini & Pagel (2004) method, using the ([Oiii]λ5007 Å/Hβ)/([Nii]λ6583 Å/Hα) (O3N2) indicator.
Fig. 12
Stacked residuals of the 64 spaxels of the three pointings after a onecomponent fit. The dashed line corresponds to the median along 15 pixels in the Xaxis. 

Open with DEXTER 
Fig. 13
Stacked residuals of the 64 spaxels of the three pointings after a twocomponent fit. The dashed line corresponds to the median along 15 pixels in the Xaxis. 

Open with DEXTER 
Fig. 14
Spatial arrangement of the residuals after the fitting procedure of pointing S. The horizontal axis of all spectra ranges from 6650 Å to 6750 Å. The vertical axis scale is the same as in Fig. 2. The number at the top left corner of each panel indicates the number of components resulting from the fitting procedure. A spaxel labeled with “0” means that 0 components were assigned to this spaxel. 

Open with DEXTER 
Fig. 15
Spatial arrangement of the residuals after the fitting procedure of pointing S. The horizontal axis of all spectra ranges from 6650 Å to 6750 Å. The vertical axis scale is the same as in Fig. 3. The number at the top left corner of each panel indicates the number of components resulting from the fitting procedure. A spaxel labeled with “0” means that 0 components were assigned to this spaxel. 

Open with DEXTER 
Fig. 16
Spatial arrangement of the residuals after the fitting procedure of pointing S. The horizontal axis of all spectra ranges from 6650 Å to 6750 Å. The vertical axis scale is the same as in Fig. 4. The number at the top left corner of each panel indicates the number of components resulting from the fitting procedure. A spaxel labeled with “0” means that 0 components were assigned to this spaxel. 

Open with DEXTER 
Fig. 17
Top: blue spectrum of spaxel S[2, 2]. Middle: red spectrum of spaxel S[2,2]. Bottom: residuals of the bestfit to the spectrum of spaxel S[2, 2] in the red wavelength range. The bestfits to the emission lines of components A and B for which the intensity peak is above Σ_{bkg} are shown in blue and red, respectively. 

Open with DEXTER 
Fig. 18
Same as Fig. 17 for spaxel M[3, 4]. 

Open with DEXTER 
Fig. 19
Same as Fig. 17 for spaxel N[4, 4]. 

Open with DEXTER 
© ESO, 2012