cathedralpkg
commented
3 years ago Dear Kr0n0ss
First of all, sorry for the delay. We did not realize that users could send messages over here. In relation to your doubt:
The points along the MEP to be calculated depend on the type of reaction. For instance, if in your reaction a hydrogen atom is being transferred, quantum effects (tunneling) will be of importance and you need a well converged MEP. A stepsize of 0.1 is too large for a gradient calculation and the points along the MEP will end up with several imaginary frequencies.
On the other hand, if you suspect that variational effects are of importance (a low value of the imaginary frequency at the transition state) you only need to calculate the MEP only at points close to the transition state to evaluate the variational effects (probably between -0.30 and 0.30 bohr). Finally, if quantum effects and variational effects on your reaction are expected to be negligible, you can probably go with transition state theory, for which you only need information on the transition state and the reactants. If you face a situation in which quantum effects are important and you cannot effort to make a lot of Hessians you should evaluate the MEP at an affordable electronic structure level (maybe, a semiempirical method), but that provides a similar barrier than UB3LYP. This will be your low-level calculation. Once you have done that, you can correct the MEP using the ISPE option of Pilgrim, i.e., by performing UB3LYP calculations at selected points along the low-level MEP. In any case, I do not recommend stepsizes larger than 0.01 bohr for gradient calculations and 0.09 or 0.10 bohr (every 9 or 10 gradient steps) for Hessian calculations. A different thing is to evaluate the MEP to know the reactants and products that lead to your transition state. In that case you can use a large stepsize, but this is not desirable if you want to extract properties from your MEP calculation.
Best regards and we apologize again for the delay in answering
Kr0n0ss
Hi, Very nice code! I am a beginner at running CVT and your code really makes things easier. I would like to compute the CVT recrossing transmission coefficient in reactions that involve transition metal complexes of around 70-100 heavy atoms. I was wondering up to which point in --path would be a good idea to reduce the default step size (from 0.01 to 0.1) and also decrease the Hessian update frequency so I can make things feasible for my systems (since I am tide up to use UB3LYP/def2-SVP ). What should I check to be sure I am not compromising too much by setting their defaults to looser values.
Thanks.