Problems with metadynamics postprocessing

[Jing Hu]

Dear cp2k users and developers

I'm learning the metadynamic recently and met the following problems about postprocessing.

The first one is that in the postprocessing of constructing the free energy surface, I used 'graph.popt -ndim 2 -ndw 1 2 -file *.restart -find-minima -cp2k' to get the minimum point on the free energy surface. I got  three sets of data which include two CV values ​​and corresponding free energy values, respectively. But awkwardly, I cannot find  the corresponding CV value in the original *-COLVAR.metadynLog file and thus fail to find the specific structural configuration . So, I want to know how to find the structural configuration  corresponding to the minimum point on the free energy surface.

The second question is that I tried to use 'graph.popt -ndim 2 -ndw 1 2 -file *.restart -find path -point-a CVs -point-b CVs -cp2k' to get the path between the minimums, but the command lines went wrong as follows:

forrtl: severe (59): list-directed I/O syntax error, unit -5, file Internal List-Directed Read
Image              PC                Routine            Line        Source            
graph.popt         00000000004F55FB  Unknown               Unknown  Unknown
graph.popt         000000000051C5B1  Unknown               Unknown  Unknown
graph.popt         000000000051B2D5  Unknown               Unknown  Unknown
graph.popt         000000000040E935  MAIN__                    330  graph.F
graph.popt         000000000040CA62  Unknown               Unknown  Unknown
libc-2.17.so       00002B42666F4555  __libc_start_main     Unknown  Unknown
graph.popt         000000000040C969  Unknown               Unknown  Unknow

I have no idea about this problem. If someone can offer some help on any question between these two, I would be really appreciate!

Regards!

JHu

[Marcella Iannuzzi]

Dear JHu

It is not surprising that you do not find exactly the same CV values in the *-COLVAR.metadynLog.

The FES generated by graph is reconstructed from the metadynamics potential, made of Gaussian contributions centred on the points where the trajectory is found when the Gaussian is spawn. 

The second issue is related to an error in reading the input.

Regards

Marcella

[Jing Hu]

Dear Marcella

Really thanks for your reply. So, as for the first problem, do you know how to locate the exact minimum of FE and find the corresponding structure? Sorry that I have no idea about it.

As for the second, since the FES is reconstructed, the data of "-point-a" should belong to the *.restart file rather than the values I got from the first step 'graph.popt -ndim 2 -ndw 1 2 -file *.restart -find-minima -cp2k' , right?

Regards

JHu

[Marcella Iannuzzi]

Dear JHu

Each minimum on the FES corresponds to specific values of the CVs, as you defined them. Hence, any structure for which the CVs take those values  belongs to the minimum. There is no one specific set of coordinates. 

I expect that the points among which the path is calculated should be taken from the generated FES and not from the restart file. 

Kind regards

Marcella

[Jing Hu]

Dear Marcella

Thanks and sorry to bother you again. I understand the fact that the FES generated by graph is reconstructed but the minima I got from graph which contain the CVs' values cannot be found in *-COLVAR.metadynLog. Is this situation normal?

And I tried another command line "graph.popt -ndim 2 -ndw 1 2   -opt-path   *HILLS.metadynLog  -orac".  And it still went wrong as follows:

 ABORT in metadyn_tools/graph.F:198 CPASSERT failed
forrtl: error (76): Abort trap signal

Image              PC                Routine            Line        Source             

graph.popt         0000000000500274  Unknown               Unknown  Unknown
libpthread-2.17.s  00002B6FF116F630  Unknown               Unknown  Unknown
libc-2.17.so       00002B6FF5641387  gsignal               Unknown  Unknown
libc-2.17.so       00002B6FF5642A78  abort                 Unknown  Unknown
graph.popt         000000000044E345  base_hooks_mp_cp_          78  base_hooks.F
graph.popt         000000000040D97E  MAIN__                    198  graph.F

graph.popt         000000000040CA62  Unknown               Unknown  Unknown

libc-2.17.so       00002B6FF562D555  __libc_start_main     Unknown  Unknown
graph.popt         000000000040C969  Unknown               Unknown  Unknown
Aborted (core dumped)

Maybe you can give me some advice about the graph usage. Thanks in advance!

Regards

JHu

[Marcella Iannuzzi]

Dear  JHu

if you understand how the reconstruction  of the FES works, you know that the Gaussian contributions to the metadynamics potential have a certain extension, hence they also contribute to parts of the CVs meatspace that do not correspond to points visited by the metatrajectory, which is the one reported in *-COLVAR.metadynLog. Moreover these Gaussian contributions sum up, and if the matadynamics settings are not properly chosen they might even produce misleading FES topographies. I cannot judge whether this is your case, but for sure it is to be expected that  the minima on the FES are not exact points of the metatrajectory in *-COLVAR.metadynLog. If they are in completely unexpected regions, far from those most frequently visited by the trajectory, the metadynamics settings are most probably not properly chosen. 

The -opt-path option of graph requires as input a path file, not the Hills file. 

Regards

Marcella

[Jing Hu]

Dear Marcella

I really really  appreciate your explanation for me.  After read your explanation, I realize I may need to complete my metadynamics settings since the CVs as the minima I got from the graph is not corresponding to my desired structure.

And I'm sorry that I even don't know what the path file refers to. Probably I chose the print level of "LOW", so the path file may not produced.

For the last thing, I was new to this area so there are so many fundamental things I may not understand. And respectively speaking, the resources for cp2k learning is not as many as others. So I'm really appreciated for your help and thanks for your time.

Kind regards

JHu