question about energy minimization, quickmin & cg


I have a very similar problem as the one mention by Xin Yan in : .
Without getting into my problem in detail, Steve you mentioned something which would actually solve it. You said : "you could just fix the positions of the atoms in the first and last replica." That's exactly what I would need. Can you tell me how to do it?

For more detail about my problem :
I have 3 local minimums A B and C.
I want to get the minimum energy path between A and B and between B and C. Although I relaxed the B configuration before running the NEB, it further "relax" the B configuration during the NEB (I said relax but it actually does not relax since the energy increases by few meV). Finally at the end of the NEBs, the B configuration has changed and is not the same for the two NEB calculations. I actually want to compare the potential energy surface of the empirical potential I use with DFT. The small change in atomic position during the NEB gives a small change of energy with the empirical potential but a big one with DFT (more than 1eV)...

Thank you in advance


If you want to freeze the entire config of the endpoints of a NEB calculation
you could use the partition command with fix setforce to zero the forces

on all atoms in those 2 partitions ( assuming the velocities are also 0.0).

This should normally not be necessary. If the endpoints are already in a minimum,
then they won’t move during NEB, since they don’t feel spring forces from
intermediate states.


Thank you Steve

I am running the neb on 12 proc (12 replica), if I got it right to freeze the initial and final conf I should add in my input file :
partition no 2*11 all setforce 0.0 0.0 0.0

but it does not work .

I thought from your previous reply (in that because of the velocity you could get different convergence for the end points of the replica... But you are now saying that if the end points are well minimized before the neb calculation this should not have any influence. Right?
Another issue with my potential is that I have some charge equilibration which are not taken into account in the force calculation so that the forces are not exact. It might be the reason why I get difference. I will try to turn of the charge equilibration to check that...

thanks again


not sure why it isn’t working with fix setforce 0.0

you are correct that if the end points are well-minimized
they should not move much regardless