I had a quick glance at how the temperature control works in variable-cell dynamics and couldn't se anything obviously wrong. The temperature is calculated from the kinetic energy of both nuclear and cell degrees of freedom; then both nuclear and cell velocities are rescaled according to the criteria written in the documentation, or at least, it looks so. Note that the temperature reported on output also includes the "kinetic energy" of the cell.
P. On Sat, 2012-12-01 at 10:17 +0100, Davide Ceresoli wrote: > Dear Tian, > I'm experiencing the same problem, also in 5.0.2. I think that the > default choice of the cell mass is not appropriate. The kinetic energy of the > cell d.o.f. is added to the ions kinetic energy, then velocities are > rescaled. If ionic temperature is always smaller than tempw, it might be > that the cell d.o.f. suck a consistent fraction of the ionic temperature. > I'm not 100% sure about that. I will perform some test by reducing wmass, > then I'll report the outcome. > I've also reported this problem to QE developers mailing list, and I'm > waiting for suggestions. > > HTH. > > Best wishes, > Davide > > > On 11/30/2012 09:49 PM, Tian Lan wrote: > > Thanks. I want to use vc-md, because it is NPT enemble. I would like to see > > the > > cell response to temperatures, e.g. . I did not find any explicit comment > > about > > the bug of temperature control. But I did see one recent message saying a > > very > > similar problem. So at this point, I am not sure whether the problem comes > > from > > the setup or from the code itself. > > What I do see is the control commands are not effective at all. The tolp=10 > > and > > tolp=100 almost have the same output, e.g., which is disappointing. When > > tempw=100K, it did converge well, but tempw=300K can only reach 200K. And, a > > smaller cell seems to perform worse in temperature control. So there are a > > lot > > of phenomena. If someone is familar with this, or knows the relationship > > between > > those seemingly messy observations, please let me know. Or it is indeed the > > code > > problem, I will try to use another one. Thanks a lot ! > > Tian > > > _______________________________________________ > Pw_forum mailing list > Pw_forum at pwscf.org > http://pwscf.org/mailman/listinfo/pw_forum -- Paolo Giannozzi, IOM-Democritos and University of Udine, Italy
