ramess101 / mie-swf Goto Github PK

From Mostafa:

Richard,

Thank you for forwarding this email. This special issue sounds like a good outlet for our research. Any ideas what could the title of our submission be?

Generally, I can imagine a few routes for our research:

1- A comparison between different potential models: shifted-force, force-switch, extended LJ, Mie, Yukawa, etc.
2- Developing force fields for additional compounds using MBAR/ITIC
3- Building an online framework for customized force field development similar to SpeadMD where people can develop their own force fields by taking advantage of MBAR
4- Developing a transport-properties optimized force field: participating in IFPSC challenge (See http://fluidproperties.org/10th) could further motivate us for focusing on transport properties

Obviously, these are brainstorming ideas and their feasibility and priority can be discussed. It would be nice if we complete such a list and prioritize them.

4/11/2018

Richard as scribe

Mohammad: There may be a problem with MBAR for polar systems because the basis functions don't work.
Rich: As long as you keep the charges constant and only vary eps and sigma, there shouldn't be a problem.
Jeff: If we use reaction field, then we should be able to get a basis function for the short-range contribution. In the long term, we just need to split the short range from long range then vary the partial charges.
Rich: It might be more straightforward to calculate the repulsive and attractive contributions. It is also possible to just return the overall non-bonded for different values of epsilon and sigma. The second approach results in more precise basis functions because the overall magnitude is always about the same, whereas the repulsive and attractive contributions can have very large magnitudes.
Rich: Currently working on GCMC+MBAR. This is like histogram reweighting but also varying epsilon and sigma.
Rich: For Special Issue of FPE, we could show how to convert tc potential to swf potential using MBAR.
Jeff: That would be a good fit.
Rich: We are planning a submission for IFPSC viscosity, in addition to the FPE issue. We were considering testing the UA Mie potentials for viscosity.
Jeff: Testing the Mie potential for viscosity just hasn't been done, so we would really like to see how it does. Peter Gordon tried something like that but he used a poorly optimized force field.
Rich: All the force fields are poor for density at high pressures. LJ underestimates but Mie overestimates. An exponent like 13 is optimal at high pressures but gives poor VLE. Only the TraPPE-AUA model gives good density at high pressure. TAMie, AUA4, and Exp6 AUA are each too high.
Jeff: Have you run any all-atom models for viscosity of ethane?
Rich: No, but have used some AA models for other molecules. In general, OPLS-AA or other AA models don't work well for engineering properties because they have been parameterized based on completely different data.
Mostafa: Now that I hear Jeff's perspective on Mie viscosities, I'm more interested in "Option C," which involves Mie viscosity predictions. Would that be a good fit for FPE too?
Jeff: Yes, really good fit. There's something fundamental to learn about how these potential parameters impact viscosity predictions.
Rich: A key weakness of MBAR is that it can't predict viscosity. We would need a way
Me: We should also consider Ungerer AUA.
Jeff: What about the >C< of AUA? Does that diameter get smaller like CH2 and CH?
Rich: I looked it up in some code I have and … yes it does. 0.3607 for CH3, 0.3431 for CH2, 0.3363 for CH, and 0.244 for C is the order.
Me: Zhang et al ~2014 give a good description of how to compute viscosities. We need to get more experience applying that in the near future, right Mostafa?
Mostafa: Yes, we will compute some viscosities before two weeks is up.
Mohammad: It would be helpful if Mostafa could wrap up his efforts implementing swf and sf in GOMC.
Mostafa: I think that would be easy, so I put that on my todo list.
JRE: I thought Martini was already in GOMC and Martini includes swf. Can we leverage that?
Mohammad: Yes, it would need to be advisable to ensure that Mostafa's swf be well-coordinated with Martini swf. The Martini swf is coded as a separate function, so it should be fairly easy. Just need to be careful about header files.
Jeff: Life would be easier if you make these changes sooner than later because there is a large batch of upgrades coming and having swf was part of the upgrade it would be easier to use the latest version as you apply the swf to your calculations.

@msoroush @jpotoff

When determining the VLE properties for the reference force field, MBAR is mathematically equivalent to histogram reweighting.

To demonstrate this, I have created a GitHub repo at https://github.com/ramess101/MBAR_GCMC. The MBAR_GCMC_class.py code results in this figure for n-hexane:

Note that the simulation snapshots of U and N are the n-hexane data that you provided me in Issue #5 at GOMC-WSU/GOMC_Examples#5.

Although it might be advantageous to use MBAR instead of histogram reweighting with increased histogram dimensions, e.g. mixtures, the main reason for this exercise was to demonstrate that we obtain the same results using MBAR as with histogram reweighting. We can now work on extending the MBAR-GCMC class so that it can predict VLE for different force field parameters and so that we can combine simulation data from multiple reference force fields.