(1) 6 eigenvalues represent rotation and translation. For (very) small
molecules, these can be quite substantial, see Carlsson and Aqvist, /J.
Phys. Chem. B,/ *109* (13), 6448 -6456, 2005. By fitting you remove the
rotation and translation. You can search the literature for papers that
discuss the QH approximation as well. The motions are not really
harmonic - this is why it's an approximation.
I've received very similar results to Carlsson and Aqvist for benzene
and palmatic acid with GMX.
(2) The values you get depend on the sampling and the conversion of the
simulations. To improve sampling, you have to store the coordinates
frequently enough (so you get more samples). In addition, the simulation
should be long enough to give you meaningful results - and both depend
on the system which you study. Checking for convergence can be done by
repeating the calculations on different time windows, as you suggested.
(3) If you want to study the entropy, you have to sure it convergence.
Otherwise, it depends on what you want to calculate.
(4) Try to see if you really simulate the same system, and maybe try
other systems like the ones in the above mentioned paper. There can be
dozens of things that can change and since I didn't run these
simulations I can't be of much help here. As for the units, I don't
really remember what I used for frequencies, only that the final result
should be in J/(mol K) or Cal/(mol K). By following the script, with the
remarks, you should get the right units.
Ran.
Ran Friedman wrote:
Dear Li Yang,
I forward your email to the GMX mailing list, which may be better for
you since other users can contribute as well. I'll reply there - I hope
you've subscribed to the list.
Ran.
Li Yang wrote:
Dear Ran Friedman
I'm sorry to disturb you, my name is Li Yang, I'm a chinese student.
I've read some paper about the calculation of QH entropy:
(a)Jurgen Schlitter_ChemPhysLett1993_215_617,
(b)Ioan Andricioaei_JChemPhys2001_115_6289(quasiharmonic approximation).
There is still something confuse me.
(1) Why the number of the eigenvalues is 3n-6 (the last 6 values are close
to zero) but not 3n ? I practice some small examples by gromacs. (256 argon
atoms in a box of 2.3nm^3, you can find this example in paper(b)). For
g_covar, when -nofit, the number of eigenvalues is 3n; when -fit the number
is 3n-6. It seems there are some freedoms constrainted in the fit process.
The paper' conclusion is based on a assumption that hwkT, say, the high
frequencies vibrate (both rotation and translations, right?) can be omitted
for the entropy calculations, as it were, the contributions of them is too
small to be omitted. Are the freedoms mentioned above represent the freedoms
of rotations and translations of the molecules. I don't know. Maybe the
answer is in those papers, but I cannot catch it.
While how to use the nofit result and fit result? The eigenv.agr in the
attachment includes fit and nofit results for the example: 256 argon
atoms in a box of 2.3nm^3. Why there is a big difference between them?
(2) I split some time-segment to obtain the entropies in each stage to get
the convergence variation of the entropy. But I doubt the feasibility of
this method. If wrong, how to do?
eg, time points: 0, 1, 2, 3, 4, 5. and time stage:0-1, 0-2, 0-3, 0-4, 0-5,
right?
why not 0-1, 1-2, 2-3, 3-4, 4-5.
In the maillist of gmx, the latter is not wrong because of undersampling, I
don't know this meaning. Could you please offer me some suggestions or refs?
(3) In entropy calculations, a system need to run a long time for entropy
convergence, the time seems to be longer than the one which needed for
energy convergence. While, for equilibrium thermodynamics simulations, how
to justify whether or not the system has achieving a equilibrium stage,
based on energy convergence or entropy confvergence?
(4) For the example mentioned in the paper Ioan
Andricioaei_JChemPhys2001_115_6289. I use your perl script for entropy
calculation. But I don't reproduce the result. The needed time of entropy
convergence is longer than the time mentioned in the paper, and so larger of
my entropy.
I don't know why, perphas the simulation conditions is not right. My
simulation files are included in the attachment. Could you give some
suggestions?
BTW, in line 77 of your script: $w=$ev*$u*10**(-18), Does 10^-18 mean
nm^-2?
I apologize in advance if I disturb you.
Thank you very much. Waiting for you reply.
Best
Li Yang
Li Yang
[EMAIL PROTECTED]
2008-09-17
--
--
Ran Friedman
Postdoctoral Fellow
Computational Structural Biology Group (A. Caflisch)
Department of Biochemistry
University of Zurich
Winterthurerstrasse 190
CH-8057 Zurich, Switzerland
Tel. +41-44-6355593
Email: [EMAIL PROTECTED]
Skype: ran.friedman