OPES_METAD_EXPLORE does not accept initial sigma?

[Qinghua Liao]

Hello,

In a simulation with PLUMED 2.8, I want to try a smaller sigma with OPES_METAD_EXPLORE, but it seems that it does not accept the sigma I provided.

However, I tried it with OPES_METAD, it worked well. I am wondering it is this case,

or there is some problem? Thanks!

All the best,

Qinghua

[Michele Invernizzi]

Dear Qinghua,

What do you mean with "does not accept"? Is there an error message or the simulation fails? 

Could you share the plumed log and/or input file?

Best,

Michele

--
You received this message because you are subscribed to the Google Groups "PLUMED users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to plumed-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/plumed-users/7fa820c5-98b2-4ba7-8331-6e27a53599e2n%40googlegroups.com.

[Qinghua Liao]

Dear Michele,

Thanks for your reply! I am running QM/MM simulation with CP2K/PLUMED, but it should not be related to CP2K. With OPES_METAD,

opes: OPES_METAD ARG=diff1,diff2 PACE=100 sigma=0.2,0.2 BARRIER=25 TEMP=300 STATE_WFILE=opes.restart STATE_WSTRIDE=100 STORE_STATES

I got the following notes from PLUMED:

PLUMED:   temperature = 300
PLUMED:   beta = 1.6774
PLUMED:   depositing new kernels with PACE = 100
PLUMED:   expected BARRIER is 25
PLUMED:   using target distribution with BIASFACTOR gamma = 41.935
PLUMED:   kernels have initial SIGMA =  0.2 0.2
PLUMED:   kernels are truncated with KERNELS_CUTOFF = 9.26924

However, if I just switch to OPES_METAD_EXPLORE:

opes: OPES_METAD_EXPLORE ARG=diff1,diff2 PACE=100 sigma=0.2,0.2 BARRIER=25 TEMP=300 STATE_WFILE=opes.restart STATE_WSTRIDE=100 STORE_STATES

Then I got the following notes:

PLUMED:   temperature = 300
PLUMED:   beta = 1.6774
PLUMED:   depositing new kernels with PACE = 100
PLUMED:   expected BARRIER is 25
PLUMED:   using target distribution with BIASFACTOR gamma = 41.935
PLUMED:   kernels have initial SIGMA =  1.29514 1.29514
PLUMED:   kernels are truncated with KERNELS_CUTOFF = 9.15805

I am wondering why OPES_METAD_EXPLORE does not take the initial sigma I gave. Thanks!

All the best,

Qinghua

[Michele Invernizzi]

Dear Qinghua,

This is the correct behavior, as briefly mentioned in the supporting information of the paper. The idea is that since OPES_META_EXPLORE has to estimate the sampled distribution (thus a well-tempered version of the original Boltzmann distribution), one should start with an initial sigma that is `sigma0 * sqrt(gamma)`, where sigma0 is typically the standard deviation of the CV obtained from a short unbiased run. 

To make things easier for the user I do this multiplication in the code (see here) instead of asking to do it explicitly. This also makes it more straightforward to switch between OPES_METAD and OPES_METAD_EXPLORE, since one can keep the same value for SIGMA, and the code internally makes the needed adjustment.

Hope this makes it clearer.

Best,
Michele

To view this discussion on the web visit https://groups.google.com/d/msgid/plumed-users/0a07c248-4853-4c82-ac6a-1ed80264cfa6n%40googlegroups.com.

[Qinghua Liao]

Dear Michele,

Thanks very much for your clarification. That was also my initial guess, as my understanding of OPES_METAD_EXPLORE.

I tried to provide smaller sigma as I want the "explore" not to be too wild, but slowly at the beginning of the simulation.

All the best,

Qinghua