VS1 Approach III: constrained charge fit

 

Constrained fit to initial values.

Overview

Summary

In Approach II, the virtual site charge ~0.17e on the pyridine was regarded as somewhat low. This fit adds a weak charge restraint to initial values for all virtual sites, which primarily affected the pyridine virtual site. The resulting charge around 0.3 e did not improve performance on condensed phase properties. In addition, we took a closer look at Br properties, which also showed worse performance. This may be due to a lack of multi-site Br fitting data. In Approach IV we attempted a re-fit with additional data.

GitHub repo/branch

 

Status

Won't progress

 Milestones and metrics

Stage

Milestone/Benchmark

Contributors

Deadline

Status

Stage

Milestone/Benchmark

Contributors

Deadline

Status

Train virtual sites and BCCs to existing HF/6-31G* data, with a weak constraint to initial charge

Re-fit 2.1.0 BCCs and virtual sites to ESP data

@Lily Wang

Oct 2023

Completed

Re-fit valence and vdW parameters to condensed phase properties

Re-fit FF terms for vsites-candidate

 

Dec 2023

Completed

Benchmark

Improved or equivalent performance for molecules with virtual sites added (Cl, Br, pyridines) on training data

 

Dec 2023

*Note: this was delayed due to an unexpected outage with lilac access.

Failed

Presentation: https://docs.google.com/presentation/d/1fQjJn2OfApAmxfSCdxxwPuGveWoS0gPRYZ9MVL32OVk/edit#slide=id.g2a39d07bb62_0_21

 

Progress and findings

Re-fitting virtual sites with charge restraints

The pyridine virtual site stabilized around 0.3 e:

Pyridine distance/charge over training

No real effect was observed on the halogens with the restraint:

Cl distance/charge over training

 

Fitting to condensed phase properties

This change in pyridine charges didn’t appear to substantially affect the condensed phase properties.