[CPMD-list] how to do constraint optimization?

[Lan-Feng Yuan (lyuan@Princeton.EDU)]

----- Original Message -----
From: Axel Kohlmeyer <akohlmey at vitae.cmm.upenn.edu>
Date: Wednesday, December 7, 2005 10:32 am
Subject: Re: [CPMD-list] how to do constraint optimization?
To: "Lan-Feng Yuan (lyuan at Princeton.EDU)" <lyuan at Princeton.EDU>
Cc: cpmd-list at cpmd.org

> On Tue, 6 Dec 2005, Lan-Feng Yuan (lyuan at Princeton.EDU) wrote:
> 
> LY>   Dear CPMD users,
> 
> dear lan-feng,
> 
> LY>   I want to optimize a structure under a constraint, e.g., the
> LY> distance between atoms 1 and 2 must be 3 A. According to 
> manual, the
> LY> way is to add CONSTRAINTS ... END CONSTRAINTS in &ATOMS. Emm 
> ... but
> LY> how about the initial geometry? It seems that I don't need to make
> LY> the input geometry satisfy the constraint, because in the 
> output it
> LY> prints the current value and its difference with the desired 
> value.LY> However, the program updates the structure by thousands 
> of times,
> LY> but never gets convergence. And the constraint is not reached
> LY> neither. What's the algorithm to do constraint optimization? 
> Is this
> LY> feature kind of broken, or do I miss something? Thank you very 
> much!
> geometry optimization with constraints is somewhat tricky, compared
> to MD. how about, you start with an simulated annealing and then
> switch over to geometry optimization when you are close to the 
> minimum?
> axel.
> 
  Dear Axel,

  Do you propose to run a simulated annealing MD with constraint to find some nearly optimized structure, then run constraint optimization? First I didn't know simulated annealing MD can run with constraint, thank you ... But I doubt it's much helpful, because I already have the fully optimized structure (i.e., without constraint), then change the atom 1-2 distance a bit, so the start geometry should not be too far from constraint optimized structure. But it still runs for thousands of steps without a conclusion.

  Thanks! Regards,

  Lan-Feng