[Paper Review] Comparison of XMM-Newton EPIC, Chandra ACIS-S3, ASCA SIS and GIS, and ROSAT PSPC Results for G21.5-0.9, 1E0102.2-7219, and MS1-54.4-0321

S. L. Snowden|arXiv (Cornell University)|Mar 19, 2002

Calibration and Measurement Techniques3 citations

TL;DR

This paper cross-calibrates X-ray spectral responses across XMM-Newton EPIC, Chandra ACIS-S3, ASCA SIS/GIS, and ROSAT PSPC using three astrophysical sources: the heavily absorbed SNR G21.5-0.9, the line-rich SNR 1E0102.2-7219, and the high-redshift cluster MS1054.4-0321. It finds flux normalizations agree within ±10% across instruments, with XMM-Newton EPIC MOS and Chandra ACIS-S3 showing the best consistency, and identifies a systematic ~7% flux deficit in EPIC PN relative to MOS.

ABSTRACT

This paper presents a ``man on the street'' view of the current status of the spectral cross calibration between the XMM-Newton EPIC, Chandra ACIS-S3, ASCA SIS and GIS, and ROSAT PSPC instruments. Using publicly released software for the extraction of spectra and the production of spectral redistribution response matrices and effective areas, the spectral fits of data from three astronomical objects are compared. The three sources are G21.5-0.9 (a heavily absorbed Galactic SNR with a power law spectrum), 1E0102.2-7219 (a SNR in the SMC with a line-dominated spectrum), and MS1054.4-0321 (a high redshift cluster with a thermal spectrum). The agreement between the measured fluxes of the various instruments is within the 10% range, and is better when just XMM-Newton and Chandra are compared. Fitted spectral parameters are also in relatively good agreement although the results are more limited.

Motivation & Objective

• To assess the relative accuracy of spectral calibration across multiple X-ray observatories: XMM-Newton, Chandra, ASCA, and ROSAT.
• To evaluate spectral flux and parameter consistency among instruments using three astrophysical calibration sources with varying spectral complexity.
• To identify systematic discrepancies in flux normalization and spectral parameters across different instruments and detector modes.
• To provide a benchmark for cross-calibration reliability using publicly available software and calibration data.
• To support ongoing calibration refinement by highlighting discrepancies, such as the EPIC PN flux deficit.

Proposed method

• Used publicly released software (SAS, CIAO, HEASoft) to extract source and background spectra, generate response matrices (RMFs), and create ancillary response files (ARFs).
• Fitted spectral models (power law, thermal) to data from G21.5-0.9, 1E0102.2-7219, and MS1054.4-0321 using Xspec, with data grouped via grppha for statistical reliability.
• Performed simultaneous spectral fits across instruments to improve statistical precision, particularly for G21.5-0.9 and MS1054.4-0321.
• Applied energy band-specific fits: 2–10 keV for G21.5-0.9, 0.5–2.0 keV for 1E0102.2-7219, and 1.0–5.0 keV for MS1054.4-0321.
• Accounted for known calibration effects, including a carbon K-edge absorption edge (τ=1.0) in ACIS-S3 fits to correct for systematic area calibration discrepancies.
• Used confidence contours to assess parameter uncertainties and consistency across instruments, especially for temperature and column density in MS1054.4-0321.

Experimental results

Research questions

• RQ1 How consistent are flux measurements across XMM-Newton EPIC, Chandra ACIS-S3, ASCA SIS/GIS, and ROSAT PSPC for the same astrophysical sources?
• RQ2 To what extent do spectral parameters such as photon index, temperature, and absorption column density agree between instruments?
• RQ3 What systematic differences exist between instruments, particularly between XMM-Newton EPIC MOS and PN, or between Chandra ACIS-S3 and other instruments?
• RQ4 How do data reduction choices, background selection, and spectral model assumptions affect cross-calibration results?
• RQ5 Can independent observations of spectrally stable sources like SNRs and high-redshift clusters serve as reliable cross-calibration references despite their inherent complexities?

Key findings

• Flux normalizations across all instruments agree within ±10% for the 2–10 keV band in G21.5-0.9, with XMM-Newton and Chandra showing the tightest agreement.
• The EPIC PN instrument measures fluxes approximately 7% lower than EPIC MOS across all sources, a consistent systematic discrepancy noted in other studies.
• For G21.5-0.9, the power-law photon index is consistent to within 0.05 (≈3%) across EPIC, ACIS-S3, and SIS, and within 0.1 when GIS is included.
• For MS1054.4-0321, the temperature and absorption column density derived from EPIC and ACIS-S3 data are in complete agreement within confidence contours, despite limited photon statistics.
• The ROSAT PSPC flux for G21.5-0.9 is significantly lower (0.89× relative to MOS1) due to high absorption and limited energy band (0.5–2.5 keV), reducing its utility for calibration.
• SIS and GIS data show reasonable flux agreement with other instruments, though SIS data are less effective for constraining spectral parameters due to lower signal-to-noise.