[CPMD-list] Re: two questions about MD

[Alexandr Isayev]

Dear Zhang,

If you have one variable your CONSTRAINT looks like:

  30001     1         2.2333309887E-03    2.5728999289E+00
  30002     1         1.5494648330E-03    2.5728999290E+00
  30003     1         9.1925447001E-04    2.5728999290E+00
  30004     1         3.8264338162E-04    2.5728999291E+00
  30005     1        -1.2575423713E-04    2.5728999291E+00
  30006     1        -6.8543558750E-04    2.5728999292E+00

where is the first column - step number, second - constraint number,
third - value of the Lagrange multiplier and the last one is actual
value of your constraint.

Sincerely,
Alexandr

Wednesday, February 28, 2007, 6:55:07 AM, you wrote:

q> Hi,professor!
q>    Thanks for the reply!
q>    I learned from some tutorial that the average force can be calculated using d
q> F/dq=<Z-1/2[-lamda+kTG]>/<Z-1/2>, and what's the quantity in the CONSTRSINT file
q> ? Is it lagrange multiplier lamda?

q>    best wishes
q>                                                   Zhang Qianfan
q>                                                    2007-02-28
q>    

q> Axel Kohlmeyer д:

>> On Sun, 25 Feb 2007, qfzhang wrote:
>> 
>> dear zhang,
>> 
>> ZQ> Hi,professor
>> ZQ>   Thanks for the reply! I still have some question about the constraint fo
q> rce.
>> ZQ>   Do you mean that the constrain force of one particular distance differen
q> ce is 
>> ZQ> just the average of the trajectory, and the barrier can be calculated just
q>  by in
>> ZQ> tegrating such \'averages\'?
>> 
>> that is correct, provided the your system is in equilibrium.
>> please check out the literature about umbrella sampling and
>> statistical mechanis in general. there are some references 
>> in the manual and most modern textbooks on MD cover it decently.
>> there are also a number of alternatives to umbrella sampling 
>> (which is considered to be rather accurate - if done correctly, 
>> but quite expensive in terms of cpu time requirements).
>> 
>> ZQ>   Another problem is that the constraint forces in the CONSTRAINT file var
q> y heav
>> ZQ> ily along the trajectory. Is there some method to solve such a problem? Ge
q> nerall
>> ZQ> y speaking,how many timestep should be done to get the converged constrain
q> t forc
>> ZQ> e? 
>> 
>> this is to be expected. the instantaneous forces should fluctuate in a 
>> more-or-less gaussian distribution around the \'true\' values and then
>> you can use conventional statistical measures to judge whether your 
>> average is converged or not. you also have to balance the accuracy 
>> of each sampling point against the accuracy of integrating the mean 
>> forces with a limited number of sampling points. of course the number 
>> of sampling points depends on the shape of your PMF.
>> 
>> ZQ>   From the example given in cpmd tutorial, when specifing CONSTRAINTS, the
q> re is 
>> ZQ> no NOSE keyword, should that be needed in the constant temperature simulat
q> ion?
>> ZQ>    best wishes 
>> 
>> can you please be more specific to which example of what tutorial you 
>> are referring to. 
>> 
>> cheers,
>>    axel.
>> 
>> 


q> _______________________________________________
q> CPMD-list mailing list
q> CPMD-list at cpmd.org
q> http://cpmd.org/mailman/listinfo/cpmd-list



-------------------------------------------------------
Alexandr Isayev,
Graduate Research Assistant, and System Administrator
@ Computational Center for Molecular Structure
and Interactions (CCMSI),
Jackson State University,
Jackson, MS USA
   Tel:  +(601) 979-1134
e-mail:  alex(at)ccmsi.us
   Web:  http://www.ccmsi.us
--------------------------------------------------------