Re: Transition State Search BY NAMD

From: Pawel Kedzierski (pawel.kedzierski_at_pwr.edu.pl)
Date: Wed Oct 07 2020 - 02:57:04 CDT

W dniu 07.10.2020 o 09:26, 辛志宏 pisze:
>
> Dear all,
>
> Are there any available tutorial regarding to the  transition State
> Search in a hybrid QM/MM simulation ?
>
I would say NAMD is close to useless for this task, for the following
reasons:

  * There is no Hessian implementation so one can't actually verify
    finding a TS on QM/MM level. Even if you make the QM slave program
    to optimize to TS, NAMD will add MM forces to the QM forces and
    drive the geometry away.
  * I tried 2 years ago to make a regular stationary point optimization
    using NAMD 2.13 git QM/MM and encountered a bug: after some cycles
    of energy lowering NAMD was repeatedly sending the same QM part
    coordinates to the QM slave every few steps (in a periodic cycle),
    thus getting back the same forces and repeating the cycle. This was
    not dependent on the MM force field (tried with AMBER and CHARMM)
    neither on the QM program and method (tried Mopac and ORCA,
    semiempirical, HF and a few DFT functionals). I did submit a bug
    report to the list but to my knowledge it was never fixed.
  * Even if the bug wasn't there, the optimizer implementation in NAMD
    is very limited, you do not even have any option to provide proper
    convergence criteria. And there is no option for TS optimization.

In short, NAMD is (just) a MD engine and while you may use it to sample
dynamics of TS ensemble, to actually locate the TS you need another tool.

Cheers,

Pawel

> Thank you in advance.
>
> Zhihong Xin
> College of Food Science and Technology,
> Nangjing Agricultural Universiy,


This archive was generated by hypermail 2.1.6 : Thu Dec 31 2020 - 23:17:14 CST