G A Matzeu

NGC 7582 (z = 0.005264; D = 22.5 Mpc) is a highly variable, changing-look AGN. In this work, we explore the X-ray properties of this source using XMM–Newton and NuSTAR archival observations in the 3 – 40 keV range, from 2001 to 2016. NGC 7582 exhibits a long-term variability between observations but also a short-term variability in two observations that has not been studied before. To study the variability, we perform a time-resolved spectral analysis using a phenomenological model and a physically motivated model (uxclumpy). The spectral fitting is achieved using a nested sampling Monte Carlo method. uxclumpy enables testing various geometries of the absorber that may fit AGN spectra. We find that the best model is composed of a fully covering clumpy absorber. From this geometry, we estimate the velocity, size, and distance of the clumps. The column density of the absorber in the line of sight varies from Compton-thin to Compton-thick between observations. Variability over the time-scale of a few tens of kiloseconds is also observed within two observations. The obscuring clouds are consistent with being located at a distance not larger than 0.6 pc, moving with a transverse velocity exceeding ∼700 km s−1. We could put only a lower limit on the size of the obscuring cloud being larger than 1013 cm. Given the sparsity of the observations, and the limited exposure time per observation available, we cannot determine the exact structure of the obscuring clouds. The results are broadly consistent with comet-like obscuring clouds or spherical clouds with a non-uniform density profile.

We explore the possibility of using machine learning to estimate physical parameters directly from active galactic nucleus (AGN) X-ray spectra without needing computationally expensive spectral fitting. Specifically, we consider survey quality data, rather than long pointed observations, to ensure that this approach works in the regime where it is most likely to be applied. We simulate Athena Wide Field Imager spectra of AGN with warm absorbers, and train simple neural networks to estimate the ionization and column density of the absorbers. We find that this approach can give comparable accuracy to spectral fitting, without the risk of outliers caused by the fit sticking in a false minimum, and with an improvement of around three orders of magnitude in speed. We also demonstrate that using principal component analysis to reduce the dimensionality of the data prior to inputting it into the neural net can significantly increase the accuracy of the parameter estimation for negligible computational cost, while also allowing a simpler network architecture to be used.