2018MNRAS.480.1999G

Principal component analysis (PCA) is applied to a variety of blazars to examine X-ray spectral variability. Data from nine different objects are analysed in two ways: long-term, which examines variability trends across years or decades, and short-term, which looks at variability within a single observation. The results are then compared to simulated spectra in order to identify the physical components that they correspond to. It is found that long-term variability for all objects is dominated by changes in a single power-law component. The primary component is responsible for more than 84 per cent of the variability in every object, while the second component is responsible for at least 3 per cent. Small differences in the shapes of these components can be used to predict qualities such as the degree to which spectral parameters are varying relative to one another, and correlations between spectral hardness and flux. Short-term variability is less clear-cut, with no obvious physical analogue for some of the PCA results. We discuss the simulation process, and specifically remark on the consequences of the breakdown of the linearity assumption of PCA and how it manifests in the real data. We conclude that PCA is a useful tool for analysing variability, but only if its underlying assumptions and limitations are understood.