Principal component analysis of the Chandra ACIS gain

TL;DR

This paper uses principal component analysis to identify the main spatial components of the Chandra ACIS gain maps, enabling accurate calibration with limited illumination sources as the radioactive calibration source decays.

Contribution

It demonstrates that a few principal components can effectively describe the ACIS gain maps, simplifying calibration with astrophysical sources.

Findings

Gain maps are described by few spatial components.

Calibration accuracy is within 0.6% over most of the chip.

Small-area illumination suffices for gain calibration.

Abstract

Up to 2020, the Chandra ACIS gain has been calibrated using the External Calibration Source (ECS). The ECS consists of an Fe-55 radioactive source and is placed in the ACIS housing such that all chips are fully illuminated. Since the radioactive source decays over time with a half-life of 2.7 years, count rates are becoming too low for gain calibration. Instead, astrophysical calibration sources will be needed, which do not fill and illuminate the entire field of view. Here, we determine the dominant spatial components of the gain maps through principal component analysis (PCA). We find that, given the noise levels observed today, all ACIS gain maps can be sufficiently described by just a few (often only one) spatial components. We conclude that illuminating a small area is sufficient for gain calibration. We apply this to observations of the astrophysical source Cassiopeia A. The resulting calibration is found to be accurate to 0.6% in at least 68% of the chip area, following the same definition for the calibration accuracy that has been used since launch.