Principal component analysis of MCG-06-30-15 with XMM-Newton

We analyse the spectral variability of MCG-06-30-15 with 600 k s of XMM-Newton data, including 300 k s of new data from the joint XMM-Newton and NuSTAR 2013 observational campaign. We use principal component analysis to find high-resolution, model-independent spectra of the different variable components of the spectrum. We find that over 99 per cent of the variability can be described by just three components, which are consistent with variations in the normalization of the power-law continuum (similar to 97 per cent), the photon index (similar to 2 per cent) and the normalization of a relativistically blurred reflection spectrum (similar to 0.5 per cent). We also find a fourth significant component but this is heavily diluted by noise, and we can attribute all the remaining spectral variability to noise. All three components are found to be variable on time-scales from 20 down to 1 k s, which corresponds to a distance from the central black hole of less than 70 gravitational radii. We compare these results with those derived from spectral fitting, and find them to be in very good agreement with our interpretation of the principal components. We conclude that the observed relatively weak variability in the reflected component of the spectrum of MCG-06-30-15 is due to the effects of light-bending close to the event horizon of the black hole, and demonstrate that principal component analysis is an effective tool for analysing spectral variability in this regime.