Regarding the use of boson exchange in I-PI

[邓晓]

The problem is this. I found that there seems to be code in I-PI that supports boson exchange (PIMDB). But when I ran PIMDB, in the He-4 system, I found that the difference between the trajectories run using PIMDB and using PIMD was very small, and the cyclic polymer only translated a small distance as a whole. So, I'm not sure if PIMDB's code works. Or, how can I judge whether the PIMDB code is working, by comparing the differences in trajectories or energies between the two? But the difference between the two is very small. I wonder if it is because my system binds the He-4 nucleus very compactly, causing the He-4 wave functions to not overlap each other and thus unable to exchange bosons. I don’t know if my judgment is correct. I look forward to your answers. Thank you.

[Barak Hirshberg]

Hi,

Which version of bosonic PIMD are you using? There is currently an open pull request to merge a new version, which is much faster, and we tested it for helium. 

One reason you might be getting very similar results initially to distinguishable particles is if the temperature is too high, or if the starting configuration has very little overlap between the de Broglie wavelength of the atoms because then only the identity permutation is contributing. This should change though as the atoms thermalize if the temperature is not too low. 

Also, how long are you running for? 

We can try to help you understand this if you share your input with us. It might be only on Monday, after Jewish New Year's tough.

All the best,

Barak

[Barak Hirshberg]

One more comment, you can also start checking you input file by looking into a simple harmonic example, which is available on the i-pi GitHub on:

https://github.com/i-pi/i-pi/tree/master/dev_tests/manual_regression_tests/bosons

[邓晓]

I wish you have a happy Jewish New Year and healthy body.these are my test files.From the line-style trajectory picture,we can see PIMDB only make a little shift with PIMD.
we will have a talk after end of  Jewish New Year .Enjoy your days and Best Wish 

[Yotam Feldman]

Hi,

Thanks for using our code!

Are you able to share with us also the file with the initial positions (conf.xyz), as well as which driver you're using for the forces?
Also, how many steps are depicted in the trajectory visualization you sent?

Thanks,

Yotam

[邓晓]

Hi,

Thank for your reply.I am using deepmd driver for the forces,a machine learning driver,and steps depicted in the  trajectory is equal to 2000000.This machine learning potential file is too large for me to upload.

Looking forward to your reply.

(I have displayed that the post was successfully sent, but I did not see my post after refreshing, so I have sent it several times. I apologize for any inconvenience caused)

Thanks,

Deng

[ipi-users]

Hi Deng,

It seems like you are running three particles in a 2x2x2 A box.

This results in a density that is >x10 larger than typically used for liquid helium.

I think that at these unrealistically high densities, the repulsion between the atoms dominates and you effectively get classical behavior (and no difference between bosons and distinguishable paticles).

We are working on including more tests for helium to the i-pi repo, on top of the ones I sent to you before, so that anyone who wants to run helium simulations could do it easily.

All the best,

Barak

[邓晓]

Hi,Barak

Unfortunately, using PIMDB of He system in lammps driver was still unsuccessful, and I think I need your help. I have already uploaded the files to Yotam.
Thanks for your help,

Deng