Pictor A was observed by the Chandra X-ray Observatory on Jan 18 2000 (sequence number 700018, obsid 346) using the Advanced CCD Imaging Spectrometer (ACIS) spectroscopic array, which provides excellent spatial resolution ( 1 ) with medium spectroscopic resolution ( 130 eV FWHM at 1 keV) at the aim point on chip S3 (back illuminated). In order to obtain good spatial resolution on the western hot spot while retaining excellent resolution on the nucleus, the latter was placed about 1 from the aim point in the direction towards the S3/S2 chip boundary (the -Y direction). The nucleus was thus about 1 from the S3/S2 boundary. The western hot spot was 3.'5 from the aim point and also on chip S3. The spatial resolution at the nucleus is barely different from the optimal, while that at the hot spot is 1.''2 1.''9 FWHM. The eastern radio lobe is partly on S3 and partly on S2 (front illuminated). The total integration time (``Good Time Intervals'') was 26.177 ksecs, taken with the default frame time of 3.2 sec. There were no flares in the background count rate, which remained constant at 1.4 counts per second over the S3 chip.
The data were processed using version R4CU4UPD4.2 of the pipeline software, and we have used the updated gain file acisD1999-09-16gainN0003.fits. The data extraction and analysis have been performed using version 1.1 of the CIAO software and version 11.0 of XSPEC.
In order to create an X-ray image covering the whole radio source (shown as Fig. 1), it was necessary to remove CCD artifacts that take the form of broad, linear features along the readout direction in chips S2 and S3. This was accomplished by first copying a spatial subsection of the level 2 events file and rotating it by -37.6, so that the axis of the S array ran horizontally. The image was then smoothed with a Gaussian with = 1.''0 and the average of the rows computed, including rejection of those pixels in each column which were more than 2 from the mean for that column. The average of the rows was then subtracted from each row. Finally, the image was rotated by +37.6, so the rows and columns corresponded to the cardinal directions. This process not only removed most of the linear artifacts, but also the `streak' from the very strong nuclear source that results from its readout and much of the differences in background level between the two chips.
To derive the X-ray spectrum of the hot spot, counts were extracted from a rectangular region in extent which includes all of the X-ray emission from the hot spot. Background counts were taken from an area of identical shape offset to the west. Other extraction regions were tried for the background but no significant changes were seen. We obtain a total of 3413 counts from the hot spot and 125 from the background. Response and auxiliary response files were created using the standard CIAO tools. The source counts were grouped so that each channel has at least 25 counts so that the fitting in XSPEC is meaningful.
We followed a similar procedure to obtain the X-ray spectrum of the jet. Counts were taken from a rectangular region 103 2.''4 that contains the whole jet and minimises background (the 10 gap between the nucleus and the eastern end of the jet was excluded). The background was taken from adjacent regions, and no significant difference was found for the jet's spectrum when using different regions for the background. We obtain a total of 481 counts from the jet and 240 from the background. The X-ray response of the CCDs depends on the location of the source on the chip. To account for this, the back-illuminated S3 chip has 1024 individual calibration measurements (256 per node) arranged in a grid covering the face of the chip. This oversamples the variation in the response of the CCD, and a single calibration measurement may be used for a source extending up to pixels (approximately 32 32 ). The X-ray jet extends 228 pixels () from the nucleus, so we divided it into four sections, each of which has a separate response. We also note that the jet crosses a `node boundary' (the division between sections of the chip that have different read-out amplifiers), and that the response may vary sharply across this boundary. To avoid any problem with this, we also divided the jet at the node boundary so as not to mix events collected by different nodes.
Each of the four spectra were grouped to give each channel at least 20 counts so that fitting is valid. These spectra were fitted simultaneously in XSPEC with each spectrum having its own background, response and auxiliary response file. The spectral properties are assumed to be constant along the length of the jet so the parameters in the fit are tied together with the exception of their normalizations.
One drawback of treating the jet in this way is that the individual spectra will have extremely low count rates, and fitting may only be valid over a limited range in energy. It is also difficult to produce clear figures of the spectral fits since there are four almost coincident data sets. To overcome these problems the entire jet spectrum may be treated as a single data set, using an `averaged' response function that is the photon weighted average of the four individual responses. We use PI spectra, which are gain corrected, to minimize the effect of collecting photons from two different nodes. The results we obtain using this technique are equivalent, within the errors, to those obtained through fitting the spectra individually.