Comparisons with the Methodology of Nichol et al. (1997).

Although Nichol et al. (1997) have reanalyzed the EMSS sample using ROSAT PSPC data we feel that a further reanalysis is necessary. This is because we feel that further improvements can be made to the reanalysis.

Our reanalysis differs from that of Nichol et al. (1997) in three important aspects. Firstly, our samples of clusters at $z>0.3$ are different. We adopted the methodology of using the original sample of Henry et al. (1992) as closely as possible but supplemented with newer measurements and identifications wherever appropriate. Accordingly we have used the revised sample of Gioia & Luppino (1994) throughout, except where recent measurements have shown certain objects (viz. MS1333.3+1725, MS1209.0+3917, MS1610.4+6616; see §2) not to be clusters, and we have used new redshift measurements where appropriate. The methodology of Nichol et al. (1997) is somewhat different in that a measurement of the extent of the X-ray emission of each cluster is made to decide whether the object is a cluster or not. Sources which are unresolved in ROSAT PSPC data are flagged as uncertain. Some of these sources were then examined in the High Resolution Imager (HRI) to check their extent, and for those clusters for which there were no high-quality HRI data Monte Carlo simulations were performed to assess the probability that such compact sources would be observed by the PSPC. Two clusters (at $z>0.3$) were excluded by Nichol et al. (1997) following this method, and two more were not excluded but were classified as ambiguous. This method of identifying clusters is not satisfactory since MS1208.7+3928, which was excluded, has been found to be a cluster (Stocke et al. 1999); it is also included in the survey of Luppino et al. (1999) and is found independently in the survey of Vikhlinin et al. (1998a). Also two clusters (MS1333.3+1725 and MS1610.4+6616, see §2) which are included by Nichol et al. (1997) have since been found not to be clusters. The two clusters (MS2137.3-2353 and MS1512.4+3647) which were classed as ambiguous in Nichol et al. (1997) are also found to be clusters in HRI observations by Molikawa et al. (1999), which adds further doubt to the robustness of the classification scheme of Nichol et al. (1997).

The second way in which our reanalysis differs with that of Nichol et al. (1997) is in the data reduction method. We have used a spectral cube when performing our background subtraction. This allows the photon vignetting to be calculated more accurately as a function of energy than when using an image in a single passband, as in Nichol et al. (1997), when a mean photon energy must be assumed.

Finally, the third way in which our reanalysis is different to that of Nichol et al. (1997) is in the data analysis. Nichol et al. (1997) measure count rates within a fixed metric aperture that is calculated (as a function of redshift and PSF) to encompass 85 % of the total flux. However this calculation is made assuming a King profile of $\beta = \frac{2}{3}$ and a core radius of $r_{\textrm{\scriptsize {c}}}=250$ kpc. We improve upon this by measuring the core radius of each cluster and correcting from detected flux to total flux using the King profile with the measured core radius (see §2.2). The count rates were converted to fluxes assuming a 6 keV Raymond-Smith spectrum in Nichol et al. (1997), whereas we had the advantage of being able to use measured temperatures (for all but one cluster) from ASCA measurements.