Higher Order Variability Properties of Accreting Black Holes

To better constrain the emission mechanism underlying the hard state of galactic black hole candidates, we use high-time resolution RXTE lightcurves for Cyg X-1 and GX 399-4 to compute two higher order variability statistics for these objects, the skewness and the Fourier bispectrum. Similar analyses, in particular using the skewness measure, have been attempted before, but the photon collection area of RXTE allows us to present results of much greater statistical significance. The results for the two objects are qualitatively similar, reinforcing the idea that the same basic mechanisms are at work in both. We find a significantly positive skewness for variability timescales less than $\sim 1 $ second, and a {\it negative} skewness for timescale $1- 5 $ sec. Such a skewness pattern cannot be reproduced by the simplest shot variability models where individual shots have a fixed profile and intensity and are uncorrelated in time. Further evidence against simple shot models comes from the significant detection of a non-zero bicoherence for Fourier periods $\sim 0.1-10$ sec, implying that significant coupling does exist between variations on these timescales. We discuss how current popular models for variability in black hole systems can be modified to match these observations. Using simulated light curves, we suggest that the most likely way to reproduce this observed behavior is to have the variability come in groups of many shots, with the number of shots per unit time fitting an envelope function which has a rapid rise and slow decay, while the individual shots have a slow rise and a rapid decay. Invoking a finite energy reservoir that is depleted by each shot is a natural way of producing the required shot correlations.