Online Material

\begin{figure} \par\includegraphics[width=16cm,clip]{4626fig10.eps} \vspace{5mm}\end{figure} Figure 2:

The 250 $\mu $m SPIRE map around the $\beta $ Pic disk. The $10 \times 10$$^\prime $ region delimited by the white square shows more than 50 background sources comparable to the cold blob seen in the southwest of the disk.

Open with DEXTER

\begin{figure} \par\mbox{\includegraphics[width=58mm,clip]{4626fig07.eps} \inclu... ...} \includegraphics[width=58mm,clip]{4626fig09.eps} } \vspace{-3mm} \end{figure} Figure 6:

The 250, 350 and 500 $\mu $m SPIRE PSFs, rotated to match the position angle at the time of the $\beta $ Pic observations. The PSF images are scaled linearly, contour lines are in steps of 10% of the peak flux. The white circle shows the beam FWHM.

Open with DEXTER

Table 1:   Observation log.

Appendix A: Data reduction

The PACS data were processed in the Herschel interactive analysis environment HIPE (v3.0), applying the standard pipeline steps. The flux conversion was done using version 5 of the response calibration. Signal glitches due to cosmic ray impacts were masked out in two steps. First the PACS photMMTDeglitching task in HIPE was applied on the detector timeline. Then a first coarse map was projected, which is then used as a reference for the second level deglitching HIPE task IIndLevelDeglitch. In the detector time series we masked the region around the source prior to applying a high-pass filter to remove the low frequency drifts. The scale of the high pass filter was taken to be half the length of an individual scan leg on the sky, i.e. 3.7$^\prime $. The detector time series signals were then summed up into a map using the PACS photProject task. The pixel scale for the 70 and 100 $\mu $m maps was set to 1 $^{\prime \prime }$, while the scale for the 160 $\mu $m map was 2 $^{\prime \prime }$. For the deep map in the 70 and 160 $\mu $m filter we combined the two detector time series and projected these together.

The SPIRE data were also reduced using HIPE and maps were obtained via the default naiveMapper task. The SPIRE observation consists of several repetitions of a map observation of the same area. As a result it was possible to project the data with a pixel size of 4, 6, and 9 $^{\prime \prime }$ while still maintaining complete sampling across the source.