DC : Never used DMA but have some issues with it. People always say you can’t use point charges from atom in molecules, mainly because those produced by DMA are terrible.
TG: Currently just interested in generating an initial guess of the geometry of the v-sites rather than using the charge value in particular.
DC: Not seen this having been done before with DMA. Most papers try to find minimal number of points to improve agreement with ESP. Other than that most things have been quite AD-HOC.
DM: The big question partly is - is there a way to come up with an initial guess? Use Chris Bayly as the ‘initial guess’? Can we do better though?
DC: For many optimized molecules on QCArchive load into latest cube kit
SB: What constraints on the v-sites? Min distances and max distances? Angles? etc?
DC: We’ve got these built into QUBEKit. For N 0.8 A from parent atom.
Observed for halogen you want sigma hole away from atom but this gives you a massive dipole, but sigma hole is most quadruple, so in this case probably want two v-sites to give the quadruple but balance the dipole.
SB: In the past what’s your go-to assessment property?
DC: Density doesn’t change, H-vap improves a bit, Ghyd improves a bit more.
DM: There was a SAMPYL with challenge with a lot of poly-halogenated atoms that seemed like they need the v-sites.
TG: In general does it almost always add v-sites?
CR: No theres a set of thresh-holds that asks ‘has the ESP improved enough’.
General: Doesn’t seem like there’s much evidence for v-sites on carbonyls
DC: Case of v-sites on C#C bonds
DM: pyridine nitrogen required slightly different BCCs (or LJs?) in order to get the electrostatics to work out.