Question on the error of the FM

[Cecília Álvares]

Hello,

I understand that in the .force output files that are generated by VOTCA after force-matching the third column is the force error (including the block-averaging procedure, in case more I am using several overdetermined-systems of equations).

I was wondering exactly these errors are (mathematically speaking). I mean, how do they the value of (F_il,model - Fil,ref)^2, which should ideally be 0, at each point?

Also, if it is not much to ask, would someone have ready-to-go the magnitude ( +/- ) of these errors in the force-fields derived via FM for propane in the original VOTCA publication? ( https://doi.org/10.1021/ct900369w) I know I can run this myself and check, so it is just in case someone happens to have this information readily. It would be quite helpful. I am asking just to have a reference on how small these errors should be: I want to be sure that the values in my case are not too large (which would indicate that maybe my system is not well described by pairwise potentials for example). Given that the FM force-fields for hexane look quite beautiful, I wanted to use the underlying error magnitude in the potential as a "reference" of how large they should be.

Thanks a lot!

Cecilia

[Christoph Junghans]

On Fri, Jun 16, 2023 at 4:21 AM Cecília Álvares
<cecilia...@gmail.com> wrote:
>
> Hello,
>
> I understand that in the .force output files that are generated by VOTCA after force-matching the third column is the force error (including the block-averaging procedure, in case more I am using several overdetermined-systems of equations).
>
> I was wondering exactly these errors are (mathematically speaking). I mean, how do they the value of (F_il,model - Fil,ref)^2, which should ideally be 0, at each point?

I guess if you would have an infinitely long trajectory and hence
infinitely long blocks the ends should go to 0.
We implemented this after https://doi.org/10.1007/978-94-010-0173-1_7,
so you can have a look at that as well.

>
> Also, if it is not much to ask, would someone have ready-to-go the magnitude ( +/- ) of these errors in the force-fields derived via FM for propane in the original VOTCA publication? ( https://doi.org/10.1021/ct900369w) I know I can run this myself and check, so it is just in case someone happens to have this information readily. It would be quite helpful. I am asking just to have a reference on how small these errors should be: I want to be sure that the values in my case are not too large (which would indicate that maybe my system is not well described by pairwise potentials for example). Given that the FM force-fields for hexane look quite beautiful, I wanted to use the underlying error magnitude in the potential as a "reference" of how large they should be.

The paper is from 2009, let me see if I can still find it. Please send
me your github username in an email.

Christoph

>
> Thanks a lot!
> Cecilia
>
> --
> Join us on Slack: https://join.slack.com/t/votca/signup
> ---
> You received this message because you are subscribed to the Google Groups "votca" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to votca+un...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/votca/eb34d925-de3c-4468-bb01-c07078abdc75n%40googlegroups.com.



--
Christoph Junghans
Web: http://www.compphys.de

[Cecília Álvares]

Hey Christoph,

"> I guess if you would have an infinitely long trajectory and hence
> infinitely long blocks the ends should go to 0.
> We implemented this after https://doi.org/10.1007/978-94-010-0173-1_7,
> so you can have a look at that as well. "

I agree, but only in a scenario where the potential form of choice (pairwise for example) is indeed such that allows you to reproduce the forces. Maybe some systems require multi-body terms or some things like that.

In any case, sorry, because my sentence " I mean, how do they the value of (F_il,model - Fil,ref)^2, which should ideally be 0, at each point?" was not well written. it should be "I mean, how do these errors relate to the value of (F_il,model - Fil,ref)^2, (...), at each point?"

What I mean is: if you have for example more than one superatom type, you will have minimum 3 pairwise potentials: 1-1, 1-2, 2-2. Each of them will be a separate .force output file and have its own reported "error". But if you look at force-matching's formalism as develop by Voth (which is the one implemented in VOTCA), they work with equations for the global, total force acting on each atom i. So in fact I was curious to know what are these individual errors of 1-1, 1-2 and 2-2 (i.e., how are they calculated by VOTCA?).

"> The paper is from 2009, let me see if I can still find it. Please send
> me your github username in an email."

I sent you an email :)

If I somehow emailed the wrong Christoph Junghans and you didnt receive anything, you can feel free to email me instead: cecilia.sar...@umontpellier.fr