\begin{table}%t3
\caption{\label{tab:ragtempPBB}Temperature~(keV) and radius (km) evolution with time of $BB_{\rm cold}$  and $BB_{\rm hot}$ in the 2$BB$~model, with uncertainties at 1$\sigma$ confidence level~(68\%).}
%\centerline
 {\small \begin{tabular}{ccccc}
\hline  \hline
Epoch & $kT_{\rm cold}$  & $R_{\rm cold}$  & $kT_{\rm hot}$ & $R_{\rm hot}$\\
    & keV  & km  &   keV &  km     \\
\hline 
Sep. 03 & $0.275\pm0.009$  &$5.8\pm0.4$ &$0.685\pm0.005$ &$1.54\pm0.03$\\ 
Mar. 04 & $0.275\pm0.009$  &$5.0\pm0.4$ &$0.699\pm0.009$ &$1.07\pm0.03$\\
Sep. 04 & $0.251\pm0.005$  &$5.6\pm0.4$ &$0.677\pm0.005$ &$0.85\pm0.02$\\
Mar. 05 & $0.225\pm0.004$  &$6.0\pm0.4$ &$0.597\pm0.006$ &$0.71\pm0.02$\\
Sep. 05 & $0.205\pm0.004$  &$6.6\pm0.4$ &$0.54\pm0.01$   &$0.51\pm0.04$\\
Mar. 06 & $0.192\pm0.004$  &$7.2\pm0.5$ &$0.50\pm0.02$   &$0.42\pm0.05$\\
Sep. 06 & $0.177\pm0.004$  &$8.2\pm0.5$ &$0.36\pm0.02$   &$0.8 \pm0.1$\\
Mar. 07 & $0.170\pm0.005$  &$9.0\pm0.7$ &$0.33\pm0.02$   &$1.0 \pm0.2$\\
Sep. 07 & $0.167\pm0.004$  &$9.3\pm0.7$ &$0.33\pm0.01$   &$0.9 \pm0.1$\\ 
\hline 
\end{tabular}}
\medskip
Given the stability of both the flux and spectrum over the last three observations (less than~8\% variation in flux in the 0.6--10~keV energy band; September 2006--2007), we merged the source photon lists to obtain a higher~S/N spectrum. The calibration of the PN~instrument has proved to be also very stable (Krisch et~al. 2005). The analysis of the merged spectrum significantly improved the determination of the spectral parameters as compared  to each single spectrum (see Table~\ref{tab:3bbtot}). Furthermore, the hotter~$BB$ component remained statistically non-significant also in the merged spectrum. We could thus obtain a more accurate (3$\sigma$) upper limit on its flux: $F_{\rm BBhot}<4.5\times10^{-14}~\ergscm2$ in the 0.6--10~keV energy band.
\end{table}