[lammps-users] question about fix heat

LAMMPS users,

I’m trying to calculate the thermal conductivity of Cu.EAM using fix heat but I’m getting some weird results. The domain has 72 nm in length and cross section of 52 nm2 with PBC on the xy axis and a surface on the z axis (direction of heat flow). I’m setting a heat flow of 10 eV / ps (with positive value on one face and negative in the other) and applying the fix every step for 200000 steps (with tstep of 1 fs). At the end of the run the temperature profile looks linear, except that for that length of material, the temperature gradient should’ve been ~ 6 K, but I’m measuring 150 K, which gives a very low conductivity (Cu has a conductivity of 386 W/mK).

any comments / suggestions?

thanks
Jaime

Paul - any comments on this Q?

Steve

Jaime,

Wouldn't most of the thermal conductivity be via the electrons? If so, that might be the part missing in your model. Perhaps you could try using a TTM+MD approach. If you'd like to go this route, let me know and I can point you to papers and/or source code.

Paul


Paul,
I also have an interest in simulating thermal transport in metals ( I have not done anything yet). Can you please clarify what TTM is? I also am interested in the papers/source code you mentioned.

Thanks,
Sreekant

— On Mon, 2/23/09, Crozier, Paul S <[email protected]…3…> wrote:



> From: Crozier, Paul S [email protected]
> Subject: Re: [lammps-users] question about fix heat
> To: “lammps-users” [email protected]
> Date: Monday, February 23, 2009, 1:02 PM
>
> <br>> Jaime,<br>> <br>> Wouldn't most of the thermal conductivity be via the electrons? If so, that<br>> might be the part missing in your model. Perhaps you could try using a TTM+MD<br>> approach. If you'd like to go this route, let me know and I can point you to<br>> papers and/or source code.<br>> <br>> Paul<br>> <br>> > -----Original Message-----<br>> > From: Steve Plimpton [mailto:[email protected]] <br>> > Sent: Monday, February 23, 2009 8:14 AM<br>> > To: Crozier, Paul S; lammps-users<br>> > Subject: Fwd: question about fix heat<br>> > <br>> > Paul - any comments on this Q?<br>> > <br>> > Steve<br>> > <br>> > <br>> > ---------- Forwarded message ----------<br>> > From: Jaime Sanchez <[email protected]><br>> > Date: Fri, Feb 20, 2009 at 4:25 PM<br>> > Subject: question about fix heat<br>> > To: LAMMPS user list <[email protected]>, <br>> > Steve Plimpton <[email protected]><br>> > <br>> > <br>> > LAMMPS users,<br>> > <br>> > I'm trying to calculate the thermal conductivity of Cu.EAM <br>> > using fix heat but I'm getting some weird results. The domain <br>> > has 72 nm in length and cross section of 52 nm2 with PBC on <br>> > the xy axis and a surface on the z axis (direction of heat <br>> > flow). I'm setting a heat flow of 10 eV / ps (with positive <br>> > value on one face and negative in the other) and applying the <br>> > fix every step for 200000 steps (with tstep of 1 fs). At the <br>> > end of the run the temperature profile looks linear, except <br>> > that for that length of material, the temperature gradient <br>> > should've been ~ 6 K, but I'm measuring 150 K, which gives a <br>> > very low conductivity (Cu has a conductivity of 386 W/mK).<br>> > <br>> > any comments / suggestions?<br>> > <br>> > thanks<br>> > Jaime<br>> > <br>> > <br>> <br>> ------------------------------------------------------------------------------<br>> Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA<br>> -OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise<br>> -Strategies to boost innovation and cut costs with open source participation<br>> -Receive a $600 discount off the registration fee with the source code: SFAD<br>> http://p.sf.net/sfu/XcvMzF8H<br>> _______________________________________________<br>> lammps-users mailing list<br>> [email protected]<br>> https://lists.sourceforge.net/lists/listinfo/lammps-users<br>> <br>>

|

TTM = “two temperature model”. I’ll attach some papers and source code that provide more details. I have no idea whether or not this is a good idea to try for simulating thermal transport in metals. My comment to Jaime was aimed at identifying a piece of physics that might be missing from his model, and suggesting a possible remedy. I’m offering no guarantee that the proposed remedy is any good. The way I’ve implemented this assumes 3D periodic, which would have to be changed in Jaime’s case.

Paul

modelling electron ion interactions using a Langevinthermostat.pdf (225 KB)

Langevin thermostat to couple electrons and ions.pdf (227 KB)

fix_inhomogeneous_langevin.h (1.7 KB)

fix_inhomogeneous_langevin.cpp (21.2 KB)

Hi Paul,

I have implemented the TTM (two temperature model) coupled with MD before in LAMMPS, but the implementation requires that you “know” what the electron component of the thermal conductivity is. since in metals the thermal conductivity is dominated by electron interactions then the lattice equation can be directly replaced with MD.

what I wanted to do was validate a numerical approach that uses fix heat (by calculating a known material’s thermal conductivity), and then use the same method to characterize an unknown material

any comments / suggestions?

thanks
Jaime

Hi Jaime. I think that the problem is with the model that you’re using rather than a problem with LAMMPS. I don’t know of a better model to use, and have no further suggestions to offer.

Paul