NEB update https://docs.google.com/presentation/d/1EgIMxyKO6GK0Dz1e067uP3Rw4MbK_rbeRP558kkn1Q8/edit#slide=id.g1e4c09bfa0e_0_72 | Trevor | Slides will be uploaded DM (slide 5) – if you compared differences between torsion conformer geometries, they would be higher for TD than NEB, right? The NEB would be smoother TG – correct. Points close to each other NEB are close to each other in conformation space, not necessarily true in TD DLM: could also argue that’s a downside
Slide 6: conformation is the first point that’s not a minimum, i.e. second point from left. Should be 15 degrees BS (slide 10): the discontinuity of the “TD” conformations at 90 degrees is probably from the out of plane pucker at the ring carbon that’s discontinuous depending on which side you approach it from. Somehow the other methods are smoother there TG (slide 11): all three NEBs unbent the out-of-plane pucker, i.e. made it flat again TG (slide 12): got essentially same result with 1 degree TD LPI completely underlaid under NEB_from_LPI (slide 13)
PB: is it a higher force because of the constrained minimisation in TorsionDrives? TG: they’re single point so they don’t see TD forces or NEB forces LPW: I think NEB doesn’t impose constraints literally, but it’s doing something similar, because the structures along the pathway are subject to a force parallel to the path. So I wouldn’t necessarily expect TD to give a higher force every time.
LPW: given that the PES might be pretty well represented in two directions – torsion angle and pucker – I wonder if it’s possible to draw out the 2D potential surface, where we can see what TD and NEB are doing differently. TG: should be straightforward for TD, but for NEB, one dimension might be zero. LPW: Was thinking you might be able to draw a diagram that compares the two
TG: wondering what I should expect / look for in fitting Forces are smaller in NEB, how would that show up in FF fitting? We haven’t really looked at this before PB: we didn’t get very far with looking at fitting to gradients
CB: are the magnitudes of the forces on Slide 6 very small, i.e. residual forces? TG: I’m showing the result of removing the constraint around the central bond. Magnitudes are normalized here, but not on Slide 11. CB: I like LPW’s rationale that the non-planar ring could represent a minima, maybe a 1-4 interaction between the ring C and the Cl? Could there be a minimum when it’s planar, and one where it’s deformed for the interaction? BS: I’m wondering if these gradients are really comparable, given that TD does constrained optimization. There’s a force exerted by the constraints in TD that might be included there, but not including it in the NEB gradients. TG: I’m doing single points after the fact, so it’s a post-hoc gradient. I’m not using either of the TD or NEB forces to calculate these single-points
LPW: if you pick a geometry where the torsion is, e.g., 45 degrees and plot the forces from the TD, I’d expect the forces to be aligned with the torsion degree of freedom. (Not necessarily 45 degrees, just anywhere the magnitude of the force isn’t small).
TG: LPW, could I set up a constrained single point? TG: if I picked 4 other atoms to define the torsions, should I expect similar results?
|