[gmx-users] Re: surface tension

André Farias de Moura moura at ufscar.br
Tue Mar 15 05:09:01 CET 2011

Dear Elisabeth,

actually, it is the other way around, you need increase the box length in
one direction, thus keeping periodic boundary conditions in the other two
directions while a (infinitely periodic) surface is created. and notice that
using genconf with -nbox 3 3 1 will increase your system but will not make
it a surface system, unless you increase the box length in one direction.

as regards the size, the larger the model system, the smaller should be
the fluctuations. but mind that you should increase the size by a few
order of magnitude for any noticeable decrease on the (huge) RMSD
values you are getting. if you want to check the results for convergence
maybe you could try either a block averaging or a running average (grace
can do that for you).

best regards


On Mon, Mar 14, 2011 at 7:57 PM, Elisabeth <katesedate at gmail.com> wrote:
> Hello,
> Thank you for your answer.
> 1- If I am right I have to increase the length in two directions rather than
> one, to create a plane parallel to XY for example?
> 2- Can you please give me an idea on how many molecules I need to have in
> the box and also what should be the thickness of layer? I have now 3nm X 9 X
> 9 dimensions. That is thickness of 3nm. What I did was replicating a 3nm box
> using genconf -nbox 3 3 1. I dont know what is the correct way of creating a
> layer for surface tension calculation.
> I appreciate any comments about number of molecules, box dimensions for such
> a study.
> 3- my last question is how can I make sure surface tension reported by
> g_energy is the equilibrated one. RMSD is very big compared to surf. ten. !
> Thanks for your time.
> Elisabeth
> ******************************************
> if you are interested in the surface tension of a pure liquid, which I
> assume is
> true from your message, then you need to create at least one surface, since
> periodic boundary conditions make the model system infinite, i.e., without a
> surface whatsoever.
> the easiest way to make that happen is to increase the length of the box in
> one direction, say the z direction. that way you will end up with a system
> that
> resemble a (thin) liquid film with vacuum below and above, meaning that you
> now have two surfaces. run a regular simulation (NVT) e use g_energy to get
> the surface tension.
> btw: as any other pressure related property, fluctuations are huge.
> best
> Andre
> On Wed, Mar 9, 2011 at 12:25 PM, Elisabeth <katesedate at gmail.com> wrote:
>> Dear gmx users,
>> Since I am new to surface tension topic I need to ask very trivial
>> questions. Please help me out with these simple questions.
>> As a starting point I am going to calculate surface tension of a pure
>> alkane
>> in a cubic box and compare with experimental values.
>> 1- g_energy is giving #Surf*SurfTen by default. On the other hand surface
>> tension can be obtained by gamma = (Pzz - (Pxx+Pyy)/2) / Lz. i.e
>> Pres-XX-(bar),  Pres-YY(bar),  Pres--(bar)
>> Can anyone tell me what the difference between these two is?
>> 2- In pressure coupling settings there is surface_tension option which I
>> guess is applicable where surface tension needs to be kept fixed. If one
>> want to calculate surface tension I dont think this option make sense. Am
>> I
>> right?
>> 3- I am using the following setting: I calculate the average for a 2ns run
>> and different start times as shown below. Although T, P and other
>> quantities
>> are equilibrated after 200ps, surface tension is not giving a constant
>> value. Is that because I am not using berenden P coupling? (As mentioned
>> in
>> the manual surface tension works with berendsen)
>> Pcoupl              =  Parrinello-Rahman
>> Pcoupltype          =  isotropic
>> tau_p               =  1      1
>> compressibility     =  4.5e-5
>> ref_p               =  40
>> time period for which average is calculated   Average       RMSD
>> Fluct.      Drift  Tot-Drift
>> -------------------------------------------------------------------------------
>> 1-2000 ps run: #Surf*SurfTen                    6.43844    3588.74
>> 3588.35   0.091406    182.721
>> 500-2000 ps    #Surf*SurfTen                   12.8518    3605.72
>> 3605.26   0.132126    198.189
>> 1000-2000ps   #Surf*SurfTen                    18.8821    3610.97
>> 3610.8    0.11819    118.191
>> 1500-2000ps   #Surf*SurfTen                     23.0072    3585.51
>> 3584.93  -0.444037   -222.019
>> 4- Assuming I am getting surface tension for a cubic box, to compare this
>> with reported values in literature I need to divide by 6 (no. of
>> surfaces)?
>> 5- Does box six affect the results? (mine is 3.3 nm ).
>> Thank you,

More information about the gromacs.org_gmx-users mailing list