I have found that if your initial configuration is distributed well enough,
then a classical MD thermalization isn't necessary....but you'll need a good
program that does that for you. If you already have a classical MD code
you're familiar with, then use that. Otherwise, the skew start method of
MOLDY (http://www.ccp5.ac.uk/librar.shtml#MOLDY) works really well. MOLDY
also has MD capability, but like I said the skew start method for generating
an initial configuration has been sufficient for me in SIESTA (although I
haven't used it for water).


On Mon, Dec 7, 2009 at 8:27 AM, N H <[email protected]> wrote:

> Another thing you cann try is equilibrating the system with a classical MD
> run at first ... It should work.
>
>
> On Mon, Dec 7, 2009 at 5:10 PM, Szyja, B.M. <[email protected]> wrote:
>
>> Hi Sufian,
>>
>> In general, if the random distribution of molecules gives us the system
>> close enough to it's optimal geometry (not necessarily in its minimum), you
>> could start MD without prior optimization, and you only need the
>> equilibration run to "take care" of the proper geometry. But if you observe
>> such jumps in temperature, it usually means too much or too close contacts.
>> I would suggest you to do at least a couple of optimization steps and then
>> try again. From my experience I can say that water is quite tricky solvent,
>> it is not only due to the density of the simulated cell, but also the
>> relative orientation of molecules (presence of hydrogen bonds network) can
>> play an important role.
>>
>> Success!
>> Bartek
>>
>> ________________________________________
>> From: [email protected] [[email protected]]
>> Sent: Monday, December 07, 2009 11:32 AM
>> To: [email protected]
>> Cc: [email protected]
>> Subject: RE: [SIESTA-L] Energy flauctuation in MD
>>
>> Hi Bartek
>> i am planning to optimize the net structure (16H2O + Formaamide)  after i
>> "shake" the unit cell with those water molecules only
>> random distribution of water molecules were chosen alone firstly,
>> I really do not know if i need to optimize H2O molecules before MD run?
>>
>> yours,
>> sufian
>> >
>> > Hi Sufian,
>> >
>> > My guess is the geometry being too far from optimal. Have you optimized
>> > the system (atomic coordinates and/or cell size) before the MD run?
>> >
>> > Cheers,
>> > Bartek
>> >
>> > ________________________________________
>> > From: [email protected] [[email protected]]
>> > Sent: Monday, December 07, 2009 10:58 AM
>> > To: [email protected]
>> > Cc: [email protected]
>> > Subject: Re: [SIESTA-L] Energy flauctuation in MD
>> >
>> >  Hi Andrey
>> >
>> >  I have 16-H2O molecule in my unit cell, i am trying to cool the system
>> >  down from 1000K to 300K to be more realistic. The problem still there,
>> > the
>> >  temperature goes very high and won't stop at the target temperature, i
>> >  tried to use annealing, Nose,... all those options give the same
>> problem,
>> >  i am wondering if there is any unseen option that can be used to fix
>> this
>> >  problem?
>> >
>> >  yours,
>> >  sufain
>> >>
>> >>
>> >>> Hi Sufian,
>> >>>
>> >>> How many molecules does your "bunch" contain? It's normal to have
>> >>> temperature fluctuations of ~T/sqrt(N), where N is the number of
>> >>> particles. Also I encountered non-physical temperature fluctuations
>> >>> when initial temperature of run was not equal to final temperature.
>> >>> Seems that Nose thermostat in Siesta is not capable to implement
>> >>> temperature gradients. And the final remark - if you want to model
>> >>> gas-liquid phase transition, Siesta might not be the right tool to do
>> >>> it, since it needs a big supercell and large simulation time.
>> >>>
>> >>> Cheers,
>> >>> Andrey.
>> >>>
>> >>>> Dear Siesta Users
>> >>>>
>> >>>> I am trying to run a molecular dynamics simulation for a bunch of H2O
>> >>>> molecules,i chose the Tini =800 K and Tfin= 300 K  my run starts
>> >>>> normally for the first one or two steps. but after that a huge
>> >>>> fluctuations in temperature appear, it reached a couple of thousands
>> >>>> of kelvins i,tried many options but i still has the same problem,
>> >>>> can any body assist me in this problem?
>> >>>> Thank you all,
>> >>>>
>> >>>> sufian
>> >>>>
>> >>>
>> >>
>> >>
>> >
>> >
>> >
>>
>>
>>
>

Responder a