[CPMD-list] Timestep with BO dynamics

[Axel Kohlmeyer]

On Thu, 17 Jan 2008, Dave Sherman wrote:

DS> How big can a BO dynamics timestep typically be (say for a box with 55
DS> H2O's).   I can just about get away with 5 AU while runnning the system
DS> using CP dynamics (although I'm using 3AUs).

my rule of the thumb number would be that you have to have
5 CP timesteps per BO timestep. so 15 to 25 a.u. (the latter 
would be an agressive choice without replacing hydrogens with
deuterium same as 5 a.u. in CP).

when checking for stability and energy conservation with an NVE 
trajectory, please keep in mind, that for default settings of BO 
dynamics (= use the previous wavefunction as initial guess) you have to 
converge very tightly or else you may get into the situation where a 
drift of the total energy from bad convergence (you will lose energy!) 
would offset the energy drift from too large a time step (where you 
gain). once you have established a good choice of time step, it is
strongly recommended to use wavefunction extrapolation to get an 
improved initial guess (3rd to 4th order), if you have enough memory
on your machine to store the extra copies of the wavefunction.
this will not only provide you with an improved guess, but you should
be able to go to a less strict convergence setting without getting
a drift in the energy and thus giving you an additional speedup.

bala can probably comment on this a little more and give you exact 
numbers on parameters and speedup, since we made some experiments 
recently for a project of his.

cheers,
     axel.


DS> 
DS> Thanks!
DS> 
DS> David M. Sherman
DS> Professor of Geochemistry
DS> Department of Earth Sciences
DS> University of Bristol
DS> Bristol BS8 1RJ UNITED KINGDOM
DS> Phone: 44-(0)117-954-5446
DS> http://mineral.gly.bris.ac.uk
DS> 
DS> 

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.