2025-12-16 FF Fitting Meeting

Barbara Morales: Benchmarking of Water Mixtures

CC: What is sigma_repl?

• BM: That is the average standard deviation of Hmix between replicates for each data point

MS: It’s interesting that the nitrogen aromatic cyclic molecules seem to be the only class of molecules that are systematically off for density. We’ve tried fixing that with virtual sites but it didn’t work

• BM: Given that we are avoiding a virtual site in water do we want to use them for these nitrogenated aromatic rings? What is the difficult part?

• DM: If someone proved that a four point water model is significantly better then we would consider

• JW: From the infrastructure side, Amber is the only issue, but if I remember correctly we previously decided that it’s not a blocker.

• CC: It’s more difficult to fit more parameters

• MS: It is clear in the literature that four point water models are needed for pure water properties, that being said OPC3 is closer to OPC than TIP3P

CI: Sorry I’m late, could we go back to the comparison of Hmix. I’m trying to refit to other water models like TIP3P-ML. In that model the epsilon in oxygen is much stronger

• MS: How did they optimize it because OPC3 and TIP3P-FB are pretty optimized?

• CI: I don’t remember off hand, but I’ll be able to present more on it in Jan. with refitting Sage 2.2.1 with different water models.

• BM: Looking forward to it!