We use the Cappellari & Copin (2003; hereafter CC03) Voronoi tesselation algorithm (hereafter VT) to build the XMM EPIC temperature map of A3266. This is the first time that this technique has been applied to XMM EPIC data. We first extract source and background images in the 0.5-7.5 keV energy band, optimizing the contrast between cluster signal and particle background over the temperature range of the cluster. These two images are used to estimate the Signal-to-Noise (S/N) of each pixel.

As X-ray events follow a Poisson distribution, many of the pixels have low S/N and it is therefore inappropriate to apply the CC03 algorithm as written directly to our data. Our implementation of the algorithm thus involves two steps. First we select all the pixels with sufficiently high S/N (i.e. (S–N)/N ≥ 1.05) and use the CC03 algorithm to bin them into meta-pixel groups with S/N~130 (our final goal for the temperature map). Since these meta-pixels are obtained from the high S/N subset, they are not generated from a continuous region. In the second step, we assign each of the unbinned pixels to its closest meta-pixel. While including the lower S/N pixels clearly increases the scatter of the meta-pixel S/N distribution, the resulting set of convex meta-pixels (cells) covers the whole image without significantly degrading the overall S/N.

In applying this technique to the A3266 mosaic event list, we obtain a total of 138 cells. We fit an absorbed mekal model to the spectrum of each cell using XSPEC v11.2. The absorption is fixed to the Galactic value (N_H = 1.6×10²⁰ cm^-2; Dickey & Lockman 1990), and the abundance is fixed to Z/Z_sol=0.2. This value is obtained by fitting a two-temperature mekal model with a single abundance to the global spectrum, extracted within a circle of 10 arcmin and excluding point sources. This best fitting global abundance is in reasonable agreement with the value found by Henriksen & Tittley (2002) in the central region mapped by Chandra (see also De Grandi & Molendi, 1999). We note that a single temperature model provides an acceptable fit to each of the 138 cell spectra.