Encarni and I are currently working on an reverberation mapping project and we have three semesters of SALT observations of three different AGN. One data set has been fully reduced while our collaborators are currently working on the other data set.

Right now, we are trying to catch up on the literature and some of the state of the art methods for measuring lags. One method (SPEARS or it’s python implementation called JAVELIN) by Zu, Kochanek, and Peterson (2011) assumes that the continuum variation is a damped random walk and that the line flux should vary in a corresponding way. If I think I am summarising it correctly, a statistical model of the of continuum is built up based on assuming that the continuum will vary as a damped random walk process, ie. the covariance between the fluxes at a time t1 and t2 will be based on some random fluctuation dampened by a parameter τd. Then, the time lag can be determined by looking at the average covariance between the continuum and the line flux or the covariance of the line flux with itself. I’m going to have to look further into the code and paper to work through the exact details of how the time lag determination actually works.

However, Grier et al. (2013) is a very cool demonstration of how this process works. They use JAVELIN and MEMECHO (Horne et al. 1991) to measure the delay for five systems they have regularly monitored. By splitting up the Hβ line by velocity, they can look for different lags in different parts of the line and create velocity-delay maps. The shape produced in different maps is highly dependent on the shape of the gas around the black hole. Depending on whether (and how) the gas is infalling, in a disk, or in a shell, the velocity maps will have different shapes. As it turns out, each of their AGN did have different shapes and also different shapes were seen for different transitions even.