Re: [ESPResSo] : tcl/Espresso sources : Good news

[Limbach, Hans Joerg, LAUSANNE, Food Materials]

 Hello Michael,

First, please use the Espresso mailing list for your questions (I will see and 
treat them the same way as if you write me personally). The advantage is that 
other users are able to profit from what we discuss.

About the units: Espresso does not contain a fixed set of units (read chapter 
1.4 in the user guide!). Better, consult a standard textbook on MD as 
Frenkel/Smith or Allen/Tildesley to get the basics of MD simulations. If you do 
not have these basics whatever I'll advice you here will not be sufficient to 
be able to safely run a MD simulation (I mean one, where you can trust the 
results). Espresso runs in MD units which are defined by sigma, epsilon and tau 
for length, energy and time. For atomistic simulations you can for example set 
1 sigma = 1 Angstrom, 1 epsilon = 1 KJ/mol and 1 tau = 1 ps. You have then to 
convert the units from your force field into these units and feed them with the 
inter command to espresso. But everybody needs to work this out personally once 
(No way around this one!).

If you have done this, all energies calculated in Espresso will have the same 
energy (also the kinetic one).

You need to choose your time step such that you sample the potential. This 
means the time step must be at least 20 times larger then the fastest 
oscillation in your system (usually the vibrations of a bond). If not you're 
results will have nothing to do with the potential yoou use and if you are 
lucky the simulation blows up (broken bonds, NaN values and so on). Lucky 
because you see that something is wrong. Also this part you have to work out 
yourself once at least. 

And start with something simple where you can find results in the literature to 
compare with. The g(r) function in Espresso is correct. This means if this 
gives strange results there are errors in the input parameters.

But don't give up. It took me months before I could run a proper simulation.

Best regards,
Hanjo

P.S. given that I will not go into details of the files you have sent me.

> -----Original Message-----
> From: Michael Winokur [mailto:address@hidden 
> Sent: mardi, 1. septembre 2009 21:38
> To: Limbach,Hans Joerg,LAUSANNE,Food Materials
> Subject: Re: : tcl/Espresso sources : Good news
> 
> Hi,
> 
> So I've continued to plug away at this and made slow but 
> steady progress.  At issue appears to be my understanding of 
> the units associated with the energies.  The empirical force 
> field coefficient in the tables are given in chemists units 
> of "kcal/mole" where as the kinetic energy referred to 
> something completely different.  Once I "rescaled" the mass 
> upwards by a factor of ~200000 or so the atoms started to 
> behave in a more or less 
> reasonable manner.    Can you give me some guidance on how I 
> should view the 
> kinetic energy?  In the "off the self" package (i.e., Tinker) 
> the relatively scaling has been worked out so it seems 
> sensible (I can't prove it is quantitatively correct).  My 
> initial thought is to observed how the energies vary with 
> temperature and then apply the equipartition theorem.  I'm 
> sure there are better/more correct methods. 
> 
> Espresso still is quirky because for large values of the time 
> step the program (over 0.5) the program eventually stops 
> because either bonds become "broken" 
> or the energy suddenly becomes "nan" (not a number).  It may 
> be that at this step some atom atom distances become so short 
> that the LJ potential blows up.
> 
> I've also tried the g(r) function an so far I'm not sure if I 
> believe what I see.  I've some tested home codes which do 
> this for systems with periodic boundary conditions and so 
> I'll see if I can make a one to one comparison.  
> As I said, slow but steady.
> 
> Is there anything else I should be warned of?
> Thanks,
> 
> Michael
> 
> 
> On Thursday 27 August 2009 06:05:43 am you wrote:
> > Hi,
> >
> > Sorry, but this week is getting a bit crazy. Remind me if 
> you do not 
> > hear anything from my side next week. Best regards, Hanjo
> >
> > > -----Original Message-----
> > > From: Michael Winokur [mailto:address@hidden
> > > Sent: mardi, 25. août 2009 21:12
> > > To: Limbach,Hans Joerg,LAUSANNE,Food Materials
> > > Subject: tcl/Espresso sources
> > >
> > > Hello Hanjo,
> > >
> > > Thanks for the offer of aid.  (nestle.com.....the Swiss food 
> > > conglomerate....so they have an interest in MD simulations?!)
> > >
> > > With the thermostat off things are even stranger.  To be 
> fair I am 
> > > much more familiar with home grown Monte Carlo 
> simulations and some 
> > > commercial modeling software (i.e.,
> > > Macromodel) rather than codes like Espresso.  The thermosetting 
> > > approaches are still very much a black box to me.
> > >
> > > As per your request I include the unwieldy tcl base program 
> > > (trial2.tcl), input data file (otp_mm3.dat) for connectivity, 
> > > customizations of the source code (my C programming skills are 
> > > minimal and so I did this by brute force.....fortran and perl are 
> > > the ones I usually use).
> > >
> > > trial2.tcl is configured to build a single orthoterphenyl 
> molecule 
> > > and then run a simulation.  I can send these to the 
> espresso e-mail 
> > > address if you think this is better.
> > >
> > > Please ask if you've a suggestion of anything else I should do.
> > >
> > > Michael
> > >
> > > --
> > >
> > > //////////////////////////////////////////////////////////////
> > > /////////////////////////////////////////
> > > / Michael Winokur, Phone:(608)263-7475
> > > / Department of Physics    FAX:  (608)265-2334
> > > / University of Wisconsin  Email:address@hidden
> > > / Madison, WI 53706         also: address@hidden
> > > / Home Page: http://romano.physics.wisc.edu 
> > > //////////////////////////////////////////////////////////////
> > > //////////////////////////////////////////
> 
> -- 
> 
> //////////////////////////////////////////////////////////////
> /////////////////////////////////////////
> / Michael Winokur, Professor  Phone:(608)263-7475          
> / Department of Physics    FAX:  (608)265-2334                  
> / University of Wisconsin  Email:address@hidden  
> / Madison, WI 53706         also: address@hidden  
> / Home Page: http://romano.physics.wisc.edu 
> //////////////////////////////////////////////////////////////
> //////////////////////////////////////////
> 
> 
>