X-ray quasi-periodic oscillations (QPOs) in active galactic nucleus (AGN) allow us to probe and understand the nature of accretion in highly curved space–time, yet the most robust form of detection (i.e. repeat detections over multiple observations) has been limited to a single source to-date, with only tentative claims of single observation detections in several others. The association of those established AGN QPOs with a specific spectral component has motivated us to search the XMM–Newton archive and analyse the energy-resolved light curves of 38 bright AGNs. We apply a conservative false alarm testing routine folding in the uncertainty and covariance of the underlying broad-band noise. We also explore the impact of red-noise leak and the assumption of various different forms (power-law, broken power-law, and Lorentzians) for the underlying broad-band noise. In this initial study, we report QPO candidates in six AGNs (seven including one tentative detection in MRK 766) from our sample of 38, which tend to be found at characteristic energies and, in four cases, at the same frequency across at least two observations, indicating they are highly unlikely to be spurious in nature.

The emissivity of the accretion flow is a key parameter affecting the shape of both the energy and variability power spectrum of AGN. We explore the energy-dependence of the power spectrum for five AGN, across the XMM-Newton bandpass, and across the 0.01-1 mHz frequency range, finding a ubiquitous flattening of the power spectrum towards higher energies. We develop a framework to explore this behaviour and thereby extract the energy dependence of the emissivity assuming a simple disc-like geometry for the inflow. We find that the emissivity ranges from  R−2 at energies around the soft excess and increases to  R−4 or steeper above 4-6 keV. We describe the changing emissivity index with a linear function in energy, finding the best-fitting slopes to vary between AGN. We attempt to correlate the slope of the linear function against key AGN parameters but, as yet, the sample size is too small to confirm hints of a correlation with Eddington ratio.