<div dir="ltr">Dear Community, while performing some tests, I encountered a problem in performing realtime propagation with CP2k. A given input file for such a simulation works fine with version 2.5.0 or earlier, but does no longer converge with version 2.6.0 (and the current trunk). In more detail: The attached input file performs a realtime propagation of some organic molecule (non-periodic system) with an electric field applied in the first step, starting from a SCF wavefunction which is optimized in the beginning of the run (to very high accuracy). In order to make sure that I did not miscompile CP2k, I used the statically linked precompiled executables ("sopt") from the homepage for this test. Attached you find the log files from version 2.5.0 and version 2.6.0. In 2.5.0 (and prior), the realtime propagation runs fine. Starting from version 2.6.0, already the fifth propagation step does not converge, and the simulation stops. The suggested approaches to handle this issue (reduce the propagation time step, increase the max. number of cycles, change EPS_ITER) all do not cure it. The simulation runs slightly longer then, but finally always reaches a point where no convergence can be achieved. This means the new versions cannot be used at all to perform these calculations, not even with a tiny propagation time step and so on... I have several (self-compiled) CP2k versions on my machine here, and I checked them all. The last version which is working for me is r13642, which is almost identical to 2.5.0 (appeared in the end of Feb 2014). Other versions I tried are r14009, r14394, and r15164, and they all show the problem. So this issue was introduced between February and May 2014. For the moment I have to use the old 2.5.0 to perform my calculations. However, it would be very nice if this could be investigated and fixed in the trunk. Or is there a (somehow hidden) parameter which I missed to modify, which could make these problems go away? Best regards, Martin </div>