I’m a PhD Student and a beginner on the use of GULP. I intend to use GULP to fitting potentials to phonon frequencies obtained from DFT.
Is there some video tutorials i can watch in order to know how to use GULP on a correct way? without having to try functionalities without knowing them very well?
Thanks in advance
Sorry - there are no videos, podcasts, Tiktok or YouTube - it’s time to do things the old fashioned way. There’s a manual, help text on the commands, examples, and of course the literature. You should also read previous posts on this topic. Your PhD supervisor is also responsible for providing help, if you want an animated element to the educational process.
Hello, No Worries.
I wrote an gulp input to fit an shell model for SiO2 quartz at som frequencies at gamma point calculated with DFT.
I noticed that at the gulp/Examples folders there are some examples for quartz so I thought it was a good idea to use them as an starting point.
Somehow I keep on having bad fits (with gnorm values being too big). I played with the weights trying to get the best fit possible but I haven’t got anywhere
Could you help me on how to use gulp appropriately? Maybe I’m trying to fit too many frequencies at the same time, Maybe I should use another potential for the shell model, etc.
By the way I’m using gulp 5.0.
Thanks in advance
sio2_buck.gin (2.4 KB)
The first thing to say is that I’d recommend you look through the previous posts on the topic of fitting frequencies, since there has already been discussion of the issue that fitting frequencies can be problematic since how do you know that you are assigning the right frequency to the right mode? How to address this via energy surfaces or use of eigenvectors should be in the earlier posts.
The second issue is that you are probably expecting too much from the model - the shell model has a small number of parameters & so expecting it to reproduce 27 frequencies, plus the structure, is optimistic. Given the errors in the modes, even for a good model, will be at least 10s of cm^-1 which is then squared and multiplied by the weights you’ve input (>1) then it’s not surprising that your sum of squares is large. Remember that it’s arbitrary since it depends on the weights you’ve given so you have to make your own subjective assessment of whether a fit is good or not rather than hoping for the sum of squares to go to zero.
PS You shouldn’t specify relax and simul as keywords since they are mutually exclusive - you should get rid of simul.