Chips I0, I2, I3, S1, S2, S3 have been used. The idea is to calibrate both BI chips. In many ACIS-S observations, the target covers all of S3, and this setup will allow the observer to model the local sky background component (the excess over the particle background) using chip S1 and then use this model for S3. For this, we had to turn off one ACIS-I chip, either I0 or I1. I show below that I0 and I1 have nearly the same background, so one can be used as a model for the other.
The following two panels show spectra (0.3-3 keV and 3-12 keV bands) of the background from the whole I0 and I1 chips, from the composite 2001 dataset (that includes both sky and particle components). Both spectra are extracted from the same observations, have a total exposure of 500 ks, and were cleaned using the VF mode filter:
The spectra are very similar, except for a 3% difference in the total flux in the 0.3-10 keV band (1% of which is a geometric factor due to the different number of the excluded bad columns). Note that all spectral lines align perfectly. The bin-to-bin ratio of these spectra (using coarser binning) is shown below:
Over the whole band, the deviations are within +-10% of the mean ratio (middle horizontal line at 0.97).
The spatial structure of the I0 and I1 background is also almost the same, when plotted in the coordinates symmetric w.r.t. the aimpoint (the CHIPX axis points toward the aimpoint in I0 but away from it in I1, while CHIPY axes point toward the aimpoint in both). Since the data used for these plots include the vignetted sky component, one expects the maximum brightness at low energies to be near the aimpoint. However, some of this nonuniformity is still seen in the pure particle background in the ACIS-stowed data. The panels below show projections of the background images in chip coordinates (with the CHIPX sign swapped for I1) onto the CHIPX and CHIPY axes, in the 0.5-2 keV band. The I0 curves were scaled up by 3% (see above).
A few deviations >10% in the CHIPX panel are due to the excluded bad columns; other than that, the curves agree to better than 10%. The next figure gives the same comparison in the 5-10 keV band:
It shows the same agreement to better than 10%, except now the sign of CHIPX doesn't matter.
The conclusion is that we can use I0 as a model for I1 with an adequate accuracy, after flipping the CHIPX coordinate and taking care of the different bad columns and the overall background normalization for the two chips.