Post date: Oct 31, 2011 8:46:26 AM

When trying to get times (both implied timescales and mean first passage time) one has to balance two things:

    1. Time scales are reliable only when lag >> λ where λ is the time scale which guarantees memoryless-ness. Until that regime, timescales (typically) increase with lag.
    2. Only timescales >> lag can be resolved. This is due to the fact that all of the motions which occur faster than lag are averaged out.

The regime for getting quantitative timescales (TS) is therefore λ << lag << TS. See eg. the same MFPT isosurface picture, computed with a long and a small lag. In the latter case, the binding region (small timescales) looks qualitatively ok, because lag << TS, but values are not converged (times are underestimated).