Re: [AMBER] diffusion time step

From: Daniel Roe <>
Date: Fri, 2 Jan 2015 12:24:50 -0700


On Fri, Jan 2, 2015 at 9:12 AM, newamber list <> wrote:
>>> It's not clear what you mean by "different temperature controls"
> Sorry I used the term lightly. AMBER one is with Berendsen and CHARMM with
> langevin dynamics.

This will certainly cause differences in diffusion rates. In fact, if
you're looking to compare to experimental values you should use no
thermostat (i.e. NVE ensemble) or an extremely weak thermostat (e.g.
Berendsen with extremely long coupling constant), combined with
tighter SHAKE tolerance and a short (1 fs) time step.

>>> You mention "ions in some constricted space": if there are not very many
> of these, you may have considerably difficulty in getting a converged
> result.
> Yes that is true. For my one case (AMBER) coordinates are saved after 2 ps
> and in this constricted volume I actually take an average over long part of
> simulation and over all the particles. I get quite linear MSD vs time
> relation upto 12-14 ps (linear line with 6 points) from which I get
> diffusion coefficient values. Hope its fine?

Depending on how many frames are in that 12-14 ps and how many
particles you're averaging over it could be - however fitting a line
to only 6 points seems like that's not enough data to me.

> But in CHARMM case, someone can suggest, I have frames saved after 10 ps
> only. So problem is having less samples of particles in constricted space
> (volume) which is not reliable. Other thing is in 10 ps a particle may
> leave and enter the same volume that will be counted as regular path which
> is wrong (single event). As particle can come inside, stay and then leave
> small volume few times in 10 ps those should be counted as different
> events. Also the particle may stay in constricted space for long (~20-30
> ps) then in this case am averaging over only slow diffusing particles hence
> may be am getting three times lower values than AMBER case in which I can
> capture both fast and slow particles as frames are saved after 2ps. I am
> not sure if am thinking correctly.

I feel like you may be over-thinking this a bit actually. Sure, you
have 5x less data if you're saving 1 frame every 10 ps as opposed to 2
ps, but particles aren't going to do all that much more in that
interval at the speeds particles will travel in simulations at typical
temperatures. If you have enough data and all other conditions are the
same the answers should agree well with each other. You can convince
yourself of this by re-doing your diffusion calculation on your 2 ps
data but reading only every 5th frame (e.g. in cpptraj 'trajin 1 last 5') - if you have enough sampling your results should

One important thing to keep in mind - you don't give many details
about your simulations or how you are performing the diffusion
calculation, but it is important that there be no imaging in your
input trajectory. If there is this will certainly throw off your
calculation (since when a molecule is imaged it appears to jump an
entire box length in 1 frame). You can use the cpptraj 'unwrap'
command to remove any imaging.

Hope this helps,


> Indeed it should not matter if we take 2ps or 10 ps time step for
> diffusion calculation if it was done on infinite bulk volume without any
> constrictions/restrictions.
> Thanks for your time and suggestions.
> regards,
> Jiom
> On Fri, Jan 2, 2015 at 1:10 PM, David A Case <>
> wrote:
>> On Thu, Jan 01, 2015, newamber list wrote:
>> >
>> > I have question related to diffusion time step that we define in cpptraj
>> > 'diffusion' calculation. Query divided in two parts:
>> >
>> > 1) I have trajectory coordinates saved after 2ps, thus in cpptraj I will
>> > define time_per_frame as 2. But if same trajectory is used with frames
>> > stripped and saved after 10 ps and using time_per_frame as 10 then
>> should I
>> > expect nearly same diffusion coeff. values?
>> Yes.
>> >
>> > 2) Actually I have some complex situation and I have number of
>> variables. I
>> > have one simulation with AMBER ff and other with CHARMM and time frames
>> > saved after 2 and 10 ps respectively that too with different temperature
>> > controls. I am getting three times difference in the diffusion values for
>> > ions in some constricted space of interest. I am not sure whether I
>> should
>> > expect the differences or not. If yes then such (three times) difference
>> is
>> > acceptable?
>> >
>> It's not clear what you mean by "different temperature controls".
>> Diffusion
>> constants are strongly dependent upon temperature, and also upon the force
>> field. (In particular, TIP3P water is dramatically different from most
>> other
>> popular water models; this may also influence ionic diffusion.
>> Be sure that you are not just looking at anectodal evidence. You mention
>> "ions in some constricted space": if there are not very many of these, you
>> may have considerably difficulty in getting a converged result.
>> ....dac
>> _______________________________________________
>> AMBER mailing list
> _______________________________________________
> AMBER mailing list

Daniel R. Roe, PhD
Department of Medicinal Chemistry
University of Utah
30 South 2000 East, Room 307
Salt Lake City, UT 84112-5820
(801) 587-9652
(801) 585-6208 (Fax)
AMBER mailing list
Received on Fri Jan 02 2015 - 11:30:03 PST
Custom Search