Long-term observations have shown that black hole X-ray binaries exhibit strong, aperiodic variability on time-scales of a few milliseconds to seconds. The observed light curves display various characteristic features like a log-normal distribution of flux and a linear rms-flux relation, which indicate that the underlying variability process is stochastic in nature. It is also thought to be intrinsic to accretion. This variability has been modelled as inward propagating fluctuations of mass accretion rate, although the physical process driving the fluctuations remains puzzling. In this work, we analyse five exceptionally long duration general relativistic magnetohydrodynamic (GRMHD) simulations of optically thin, geometrically thick, black hole accretion flows to look for hints of propagating fluctuations in the simulation data. We find that the accretion profiles from these simulations do show evidence for inward propagating fluctuations below the viscous frequency by featuring strong radial coherence and positive time lags when comparing smaller to larger radii, although these time lags are generally shorter than the viscous time-scale and frequency independent. Our simulations also support the notion that the fluctuations in $\dot{M}$ build up in a multiplicative manner, as the simulations exhibit linear rms-mass flux relations, as well as log-normal distributions of their mass fluxes. When combining the mass fluxes from the simulations with an assumed emissivity profile, we additionally find broad agreement with observed power spectra and time lags, including a recovery of the frequency dependency of the time lags.

The nature and geometry of the hard state in black hole binaries is controversial. The broadband continuum spectrum and fast variability properties can be explained in a model where the inner disc evaporates into a geometrically thick, hot flow. However these models are challenged by the persistent detection of an extremely broad iron line, which requires that the disc extends down to the last stable orbit of a high spin black hole. This line width can be considerably reduced if the Comptonisation continuum is multi-component rather than single temperature, but such models are highly degenerate. Here we show a specific model of a radially stratified continuum coupled to a model of propagating fluctuations, fit to some of the best hard state data from GX 339-4. This full spectral-timing model can fit the time averaged spectrum, the power spectra in different energy bands, and the frequency dependent lags between these bands. For the first time we also include disc reverberation and show that this same spectral-timing model successfully \textit{predicts} the lag-energy spectra on all timescales. This gives a more robust method to determine the inner radius of the disc, which is of order $20~R_g$, i.e. significantly truncated. This opens up the way to use the fast variability spectral-timing data to trace the source geometry of black hole binaries in all states.

Recent monitoring campaigns designed to map the accretion regime in AGN show major discrepancies with models where the optical/ultraviolet (UV) is produced by X-ray-illuminated, optically thick disc material within a few hundred gravitational radii. However, these campaigns only monitored X-rays below $10$ keV, whereas the bolometric luminosity for most of these AGN peaks above $50$ keV. We use data from the recent multiwavelength campaign by \cite{E17} on NGC 4151 - the only AGN bright enough to be monitored at higher energies with \textit{Swift} BAT. We develop a spectral-timing model with a hot corona, warm Comptonisation, and outer standard disc. This fits the time-averaged spectrum well, but completely fails to match the UV variability predicted from the X-ray lightcurve. However, it reveals that NGC 4151 had a bolometric luminosity around $1.4\%$ of the Eddington luminosity during this campaign, close to the luminosity at which AGN show a `changing-state' transition, where the broad optical lines disappear. Stellar mass black holes show a state transition at a similarly low Eddington fraction, which is broadly interpreted as the inner disc being replaced by an optically thin flow. We find that the UV lightcurve can instead be matched by reprocessing of the X-ray flux on size scales of the broad line region (BLR; $1.5-20$ light-days) and rule out there being optically thick material inwards of this, as expected if the thin disc is replaced by the flow below the inner radius of the BLR. These results emphasise the need for even longer-timescale, multiwavelength monitoring campaigns on variable AGN.

We develop new techniques to deconvolve the radial structure of the X-ray emission region in the bright low/hard state of the black hole Cygnus X-1 using both spectral and timing data in the 3-35~keV range. The spectrum at these energies is dominated by Comptonisation rather than the disc, but there is a complex pattern in the time lags between different energy bands and differences in the normalisation and shape in the power spectra of these bands, which clearly shows that the Comptonisation is not produced from a single, homogeneous region. We use a physically based model of density fluctuations propagating through a spectrally inhomogeneous flow, setting the spectral components by jointly fitting to the time-averaged and Fourier resolved spectra. The predicted variability in any band is modelled analytically in Fourier space so it can be fit directly to the observed power spectra and lags. We find that the best fit model picks out three distinct radii in the flow, each with a distinct Compton spectrum. The variability and luminosity produced at these radii is enhanced, while propagation of fluctuations from larger radii is suppressed. We associate these radii with the disc truncation, the inner edge of the flow, and (more speculatively) the jet launch radius. These distinct radii are most evident where the source is close to a transition between the low/hard and high/soft states. We suggest that the smoother power spectra seen at lower luminosities imply that the source structure is simpler away from the transition.

We build a full spectral-timing model for the low/hard state of black hole binaries assuming that the spectrum of the X-ray hot flow can be produced by two Comptonisation zones. Slow fluctuations generated at the largest radii/softest spectral region of the flow propagate down to modulate the faster fluctuations produced in the spectrally harder region close to the black hole. The observed spectrum and variability are produced by summing over all regions in the flow, including its emission reflected from the truncated disc. This produces energy-dependent Fourier lags qualitatively similar to those in the data. Given a viscous frequency prescription, the model predicts Fourier power spectral densities and lags for any energy bands. We apply this model to archival RXTE data from Cyg X-1, using the time-averaged energy spectrum together with an assumed emissivity to set the radial bounds of the soft and hard Comptonisation regions. We find that the power spectra cannot be described by any smooth model of generating fluctuations, instead requiring that there are specific radii in the flow where noise is preferentially produced. We also find fluctuation damping between spectrally distinct regions is required to prevent all the variability power generated at large radii being propagated into the inner regions. Even with these additions, we can fit either the power spectra at each energy, or the lags between energy bands, but not both. We conclude that either the spectra are more complex than two zone models, or that other processes are important in forming the variability.