muVT MD Simulations

[Jared O'Leary]

Hi All:

It seems like HOOMD-Blue doesn't have native tools for muVT MD simulations (although they do have such tools for muVT HPMC simulations).

Have any of you written custom code to implement muVT  MD simulations? If not, do you have any suggestions on how I would go about implementing this?

Thanks!

Jared

[Joshua Anderson]

Jared,

I am not aware of anyone performing muVT simulations with HOOMD in MD. You could implement muVT with a custom updater, but the evaluation would cost O(N) for each force evaluation and the overhead from snapshot initialization would be very high. Implementing this efficiently would be a very large development effort with a significant impact on the core codebase.

Have you tried the muVT in HPMC? Why doesn't that work for you?
------
Joshua A. Anderson, Ph.D.
Research Area Specialist, Chemical Engineering, University of Michigan

> --
> You received this message because you are subscribed to the Google Groups "hoomd-users" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to hoomd-users...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/hoomd-users/10b58228-daf6-4882-9818-85b593610453n%40googlegroups.com.

[Michael Howard]

Hi Jared,

To add to Josh's answer, muVT MD is very difficult because particle insertion / deletion introduces discontinuities in the equations of motion. You will find a variety of schemes in the literature (largely based on extended Hamiltonians) for implementing this ensemble, and you would need to think carefully about what one would make sense for your problem.

Regards,

Mike

[Jared O'Leary]

Thank you both for your responses!

Essentially, I am trying to replicate the system used in this paper: https://arxiv.org/pdf/1912.08333.pdf

Here, I need grand-canonical moves and dynamics/cluster moves and HPMC won't do it. 

A possible solution I thought of would be to swap HOOMD-Blue's HPMC for virtual-move MC (people have done this for other systems: http://vmmc.xyz/), but honestly this full implementation is over my head.

Do either of you have any other further suggestions? If not, I think I will just try to work with a different system.

Thanks again for all of your help!

Best,

Jared

[Joshua Anderson]

Jared,

You could implement the model in this paper with:
* Geometric cluster moves: https://hoomd-blue.readthedocs.io/en/v3.0.0/module-hpmc-update.html#hoomd.hpmc.update.Clusters
* A patch potential: https://hoomd-blue.readthedocs.io/en/v3.0.0/module-hpmc-pair-user.html
* And a Spherical particle: https://hoomd-blue.readthedocs.io/en/v3.0.0/module-hpmc-integrate.html#hoomd.hpmc.integrate.Sphere

While the rejection free Geometric cluster moves will not give the same "dynamics" as the virtual move algorithm, it will sample the same equilibrium states.

Implementing virtual moves in HPMC would not require as much code as efficient muVT MD, but would be conceptually more complex. It would also not be possible to parallelize, which negates some (or much?) of the performance advantage.

One more thing - do you really need muVT at all? In the Glotzer group, we regularly study systems that form complex and open structures such as those in https://arxiv.org/pdf/1912.08333.pdf using NVT and NPT ensembles: https://glotzerlab.engin.umich.edu/home/publications/
For example:
* https://doi.org/10.1039/C4SM01612B
* https://doi.org/10.1039/D0SM02073G
* https://doi.org/10.1073/pnas.1905929116

------
Joshua A. Anderson, Ph.D.
Research Area Specialist, Chemical Engineering, University of Michigan

> To view this discussion on the web visit https://groups.google.com/d/msgid/hoomd-users/bdfcc3b2-1c45-46be-b74b-5c3b424fdb94n%40googlegroups.com.