A note on some discrepancies in convolution models in X-ray spectral analysis

TL;DR

This paper investigates discrepancies in convolution models used in X-ray spectral analysis, revealing that they can deviate significantly from non-convolution models depending on various factors, which impacts spectral feature interpretation.

Contribution

It identifies and characterizes the deviations in convolution models in X-ray spectral analysis, highlighting their dependence on energy grid and kernel shape, and advises caution in their application.

Findings

Discrepancies can reach up to 10% of flux in certain energy bins.

Deviations depend on energy grid, kernel shape, and underlying spectrum.

Convolution models may produce significantly different profiles from non-convolution models.

Abstract

Convolution models are powerful tools in many fields of spectral and image analysis owing to their wide applicability, and X-ray astrophysical spectral analysis is no exception. We found that relativistically broadened Fe K${\alpha}$ line profiles obtained through many convolution models both within and without Xspec show deviations from the profiles produced by their non-convolution counterparts. These discrepancies depend on the energy grid considered and on the shape of both the kernel and the underlying spectrum, but can reach as high as 10% of the flux in certain energy bins. We believe that this effect should be taken into consideration, considering how often these models are used to study spectral features of lower relative intensity, and advise great discretion in using them.