Negative bias in Well tempered metadynamics.

[galmas]

Hi everybody,
I'm doing the Adaptive variables I tutorial (https://plumed.github.io/doc-v2.3/user-doc/html/belfast-2.html) and I was wondering why when performing metadynamocs one needs to calculate the negative bias instead of the free energy? Also, can anyone explain what exactly is the correction being calculated?

I would really appreciate any help.

Thanks.

[Giovanni Bussi]

Hi!

With metadynamics you usually have a relationship between F and V (F=-V in non-well-tempered, F=-gamma/(gamma-1)*V in well tempered).

When using adaptive hills the relationship between F and V this relationship is lost and the only way to get F is through reweighting. As it is shown in Branduardi, Bussi, and Parrinello, JCTC (2012) this reweighting procedure can be done using a umbrella-sampling-like anaysis using the final bias potential.

Giovanni

--
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+unsubscribe@googlegroups.com.
To post to this group, send email to plumed...@googlegroups.com.
Visit this group at https://groups.google.com/group/plumed-users.
To view this discussion on the web visit https://groups.google.com/d/msgid/plumed-users/024d4f8b-79c9-4cf1-8637-99fe0d0707f2%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[galmas]

I see, thank you Giovanni I'll take a look at this article.

Cheers

On Monday, October 24, 2016 at 6:33:46 PM UTC+3, Giovanni Bussi wrote:

Hi!

With metadynamics you usually have a relationship between F and V (F=-V in non-well-tempered, F=-gamma/(gamma-1)*V in well tempered).

When using adaptive hills the relationship between F and V this relationship is lost and the only way to get F is through reweighting. As it is shown in Branduardi, Bussi, and Parrinello, JCTC (2012) this reweighting procedure can be done using a umbrella-sampling-like anaysis using the final bias potential.

Giovanni

On Mon, Oct 24, 2016 at 5:28 PM, galmas <galma...@gmail.com> wrote:

Hi everybody,
I'm doing the Adaptive variables I tutorial (https://plumed.github.io/doc-v2.3/user-doc/html/belfast-2.html) and I was wondering why when performing metadynamocs one needs to calculate the negative bias instead of the free energy? Also, can anyone explain what exactly is the correction being calculated?

I would really appreciate any help.

Thanks.

--
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.

[galmas]

Hi, sorry for reposting this, but I have a related question concerning free-energy calculation.

If I performed a well-tempered metadynamics simulation and I want to calculate the free energy surface of my reaction then the correct way do to so will be through the HISTOGRAM and CONVERT_TO_FES tools in plumed-2.3 and not using sum_hills?

Cheers,
Gal.

[Gareth Tribello]

Hello 

The answer is it depends.  There are two scenarios:

(1)  You want the free energy as a function of the CV that you used in the metadynamics simulation.  To be clear what I mean lets suppose that your PLUMED input looked like this:

d1: DISTANCE ATOMS=1,2

METAD ARG=d1 <some other keywords> 

If you want the free energy as a function of the distance between atom 1 and atom 2 you can use either sum hills or HISTOGRAM and CONVERT_TO_FES.  They should give pretty much the same result.

(2)  You wan the free energy as a function of some other CV.  For the example above you want the free energy as a function of some CV that is not the distance between atom 1 and atom 2.  In this case you can only use HISTOGRAM and CONVERT_TO_FES.  You cannot use sum hills as it will give you the free energy as a function of the distance between atoms 1 and 2 only.

Gareth

To view this discussion on the web visit https://groups.google.com/d/msgid/plumed-users/f119895f-d7c7-4a54-b584-1af0ca418511%40googlegroups.com.

[galmas]

So in principle, if I'm working with one CV (an angle CV in my case) the following should give me similar results:

Using sum_hills:

plumed sum_hills --hills HILLS --negbias --outfile negative_bias.dat --spacing XXX --min YYY --max ZZZ

plumed sum_hills --histo HILLS --spacing XXX --min YYY --max ZZZ --kt 2.577483 --sigma AAA --outhisto correction.dat

Using HISTOGRAM and CONVERT_TO_FES:

angle: READ FILE=COLVAR VALUES=angle IGNORE_FORCES IGNORE_TIME

metad: METAD ARG=ang PACE=1 HEIGHT=1.5 SIGMA=0.03 FILE=HILLS GRID_MIN=YYY GRID_MAX=ZZZ GRID_SPACING=XXX BIASFACTOR=BBB TEMP=310.0 REWEIGHTING_NGRID=CCC REWEIGHTING_NHILLS=50

bias: REWEIGHT_METAD ARG=metad.rbias TEMP=310

histo: HISTOGRAM ARG=angle GRID_MIN=YYY GRID_MAX=ZZZ GRID_SPACING=XXX BANDWIDTH=DDD LOGWEIGHTS=bias

fes: CONVERT_TO_FES GRID=histo TEMP=310.0

Now, in practice these two approaches give me slightly different profiles.

Also the free energy scale is different - for example the first method gives a minimum free energy of -350 kJ/mol while the second one -20 kJ/mol.

Is there something wrong with my usage of sum_hills and HISTOGRAM/CONVERT_TO_FES?

Could the reason be that my simulation is not converged yet?

Cheers,
Gal. 

[Gareth Tribello]

Hello

You need to remember that any method for determining the free energy profile gives the free energy modulo an additive constant. The absolute values of the free energy that you get out are thus not particularly meaningful.  The only thing that matters is the difference in free energy in moving from state A to state B.   If these differences are the same with the two different method that would be a reasonable indication that your simulation had converged. 

Gareth

To view this discussion on the web visit https://groups.google.com/d/msgid/plumed-users/80816243-ec72-4e33-bbbd-425d68a63d89%40googlegroups.com.

[galmas]

Ok,
Thank you Gareth, helpful as always!

Cheers.