PIMD only on some atoms in domain

[Karthik Ganeshan]

Hello everyone, 

Is there any way I can impose PIMD only on a few elements in the system while treating every other species strictly classically? 

I am very new to the Quantum and GLE thermostats, but is it something that can be handled through them instead? I am assuming this as a possibility as the Quantum thermostat is a fit... 

It would be great if the selective PIMD can'be imposed mechanically via the integrators, of course. 

Appreciate all the help! 

Sincerely,
Karthik. 

[Venkat Kapil]

If you want to treat only certain elements quantum mechanically you could simply set very large masses for the atoms you would want to be treated classically (say by multiplying the masses with a large factor), and use physical masses for the rest.  However, if you are simply thinking of it as a means of reducing the computational overhead then treating just a few nuclei quantum mechanically would not reduce the computational cost w.r.t to case the where all the atoms are quantum, as in both cases you would have to compute the energies of replicas that contain all the atoms. I hope this answers the question, otherwise please let us know.

[Michele Ceriotti]

Adding to what Venkat says, there is a "spatially localized contraction" implemented that could help to an extent to achieve something akin to what you mention. 

ref: http://doi.org/10.1063/1.5002537

[Karthik Ganeshan]

Thank you! 

I will take a closer look at the paper. It seems to be quite similar to what I want, except that I am looking for an element type distinction than a spatial one. Is it a module that's already within i-pi? (I might have missed its description in the paper.. only skimmed through it so far). 

But on Venkat's idea: wouldn't increasing the masses of every other element skew the overall dynamics of the system? If I am trying to isolate the H+ and Li+ quantum effects in water, for example, the heavier water would probably not give us the proper (expected) dynamics, right? The velocities would be proportionally lower for the system. 

I am not looking to use this as a way to accelerate PIMD. Just to isolate the effects.

Thanks again, 

Sincerely,
Karthik.

[Venkat Kapil]

Depends on what you want to compute. Are you interested in thermodynamic properties or dynamical ones ? I do not understand how to do "proper dynamics" for a system that has a few classical and a few quantum atoms as in I do not know if there is a rigorous way to do that. While if you want thermodynamic properties then setting the masses the way I recommended makes some sense. 

[Karthik Ganeshan]

Hmm. Yeah, I see your point. I do want some dynamical aspect on the classically-treated atoms, though. I will try to rig up something like the paper Prof. Ceriotti had mentioned. 

Thanks! 

[Venkat Kapil]

So what I really wanted to say in that message is that if you want to compute thermodynamic properties of a system with a few classical particles that can be done by simulating a quantum system with some atoms with physical masses and other (classical) atoms being treated as isotopes with VERY high masses. So I hope it is clear that this is a fictitious system and the results that you would get would not be the real "atomic" quantum effects but the quantum effects of that quasi system. For instance if you have a molecule of water and you set the oxygen atoms to have very high masses that will have an effect on the quantum kinetic energy of the hydrogen atoms, and it would not give you the same results as you would get for the hydrogen atoms if all the particles were quantum. So what I suggested was not a recipe for getting local atomic quantum effects, but just a way to simulate the described fictitious system.

[Michele Ceriotti]

Sorry to be a bit blunt, but I fail to see the point of what you are trying to achieve. If there is no computational advantage in making some atoms classical, there is no reason to do so. 

The "right" degrees of freedom to think about contributions to quantum effects are normal modes that often involve multiple atoms so I don't see what you would learn. 

[Venkat Kapil]

I agree with Michele since what you propose is not a rigorous way of disentangling quantum effects of different elements.

[Karthik Ganeshan]

(sorry for the delayed reply) 

Well.. I suppose in a way it is to accelerate as my plan was to eventually use a more aggressive thermostat on the ones I am treating with PIMD and a regular Nosé-Hoover on the bulk of the system being treated classically. But I understand, this is not a rigorous or particularly a wise thing to do. 

Thank you so much for your responses! 

[Venkat Kapil]

Just to warn you, applying quantum thermostats on a few atoms with classical ones on some others can lead to some zero-point energy leakage.

[Karthik Ganeshan]

Right. I hadn't considered that...