Page 1 of 1

Error in the discription of DYNAMICAL MATRIX?

Posted: Mon Jan 25, 2010 6:39 pm
by jianmin
Dear Prof. Graeme,

In the section "CALCULATING THE FIRST DYNAMICAL MATRIX" of the following webpage,
http://theory.cm.utexas.edu/vtsttools/dynmat/
step 4 is "Set up the INCAR file. You need to have the following variables set: ICHAIN=1, POTIM=0.0, IMAGES=NN (only if running in multiple geometries at once mode from step 2), NSW=(Degrees of freedom)/N+1." It says that NSW=(Degrees of freedom)/N+1.

But in the paragraph below it, you write "For instance, in the calculation of ethane modes, the NSW tag would be 25 (3 degrees of freedom for each of 8 atoms, plus 1)." It seems like NSW=(Degrees of freedom)+1, not =(Degrees of freedom)/N+1.

One of them must be wrong. So which one should I use?

Re: Error in the discription of DYNAMICAL MATRIX?

Posted: Mon Jan 25, 2010 10:18 pm
by graeme
The second description is most appropriate for getting started. This is when you are calculating your finite difference forces one after the other. It is also possible to run vasp with multiple images (NN>1) so that the finite difference forces are calculated in parallel, but this is a more advanced mode. Start with N=1 so that NSW=DOF+1.

Re: Error in the discription of DYNAMICAL MATRIX?

Posted: Mon Jan 25, 2010 10:21 pm
by jianmin
Prof. Graeme, thank you very much for your prompt reply.

Now I understand that:
To do a serial calculation, use NSW=DOF+1 and do not set the IMAGES parameter;
To do a parallel calculation on N cores, use NSW=DOF/N+1 and set IMAGES=N.