nstep
Posted: Thu Jun 27, 2013 9:54 pm
Hello:
I have been running a geometry optimization (ionmov 3 and optcell 2) on two systems which only differ by the identity of one element. In both cases, it is very difficult to converge the electronic scf of the initial geometry, so that the nstep limit is reached before the stopping tolerance (tolvrs 1.0d-18). This is true even if I increase nstep and read in the wavefunction from a previous calculation.
My question is has to do with what happens after the first step. One of the systems continues on and takes a new geometry step despite not reaching electronic convergence. This is fine with me, because the calculation converges well after this. The second system, however, quits after not reaching convergence in the first geometry iteration. I am using the same input file for both.
Is there a reason that one calculation quits and one goes on? Is there a way to force the other system to go on to the second geometry iteration?
Thank you,
Kale
I have been running a geometry optimization (ionmov 3 and optcell 2) on two systems which only differ by the identity of one element. In both cases, it is very difficult to converge the electronic scf of the initial geometry, so that the nstep limit is reached before the stopping tolerance (tolvrs 1.0d-18). This is true even if I increase nstep and read in the wavefunction from a previous calculation.
My question is has to do with what happens after the first step. One of the systems continues on and takes a new geometry step despite not reaching electronic convergence. This is fine with me, because the calculation converges well after this. The second system, however, quits after not reaching convergence in the first geometry iteration. I am using the same input file for both.
Is there a reason that one calculation quits and one goes on? Is there a way to force the other system to go on to the second geometry iteration?
Thank you,
Kale