LAMMPS WWW Site - LAMMPS Documentation - LAMMPS Mailing List Archives
Re: [lammps-users] [EXTERNAL] how to perform grand canonical Monte Carlo simulation for binary mixture by LAMMPS?
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [lammps-users] [EXTERNAL] how to perform grand canonical Monte Carlo simulation for binary mixture by LAMMPS?


From: "Thompson, Aidan" <athomps@...3...>
Date: Mon, 6 Nov 2017 20:14:24 +0000

Wei,

 

You are right.  Thanks for pointing this out.

 

I ran my own tests and indeed found that there is a systematic bias if the seeds are the same. Using different seeds, the differences in total density, pressure, and excess chemical potential, compared to the 1-component case, were fully eliminated, within statistical noise.  In addition, contrary to my previous assertion, there was a systematic difference in the average number of atoms of type 1 versus type 2, depending on which fix was listed first (ratio of first type to second type is about 40:60). This also went away when both seeds are different:

 

Aidan

 

# Same seeds

# Step  v_n1av v_n2av

    1000       19.77        33.12

    2000        20.9        32.75

    3000       20.71        32.34

    4000       21.36        31.66

    5000       21.18        33.12

    6000       20.27        34.14

    7000       21.24        33.22

    8000       21.38        32.91

    9000       21.71        32.02

   10000       20.65        34.95

 

# Different seeds

# Step  v_n1av v_n2av

    1000       30.18        31.66

    2000       29.37        31.81

    3000      30.73        30.15

    4000      29.53        30.76

    5000       30.06        31.04

    6000      28.99        31.51

    7000      29.15        31.09

    8000      31.12        29.23

    9000      28.65        31.69

   10000      28.99        31.15

 

From: "chenwei@...7134..." <chenwei@...7134...>
Date: Tuesday, October 24, 2017 at 7:05 PM
To: "Thompson, Aidan" <athomps@...3...>
Cc: lammps/lammps LAMMPS Users List <lammps-users@lists.sourceforge.net>
Subject: Re: Re: [EXTERNAL] [lammps-users] how to perform grand canonical Monte Carlo simulation for binary mixture by LAMMPS?

 

Dear Aidan P. Thompson,

 

I fixed this problem by using two different random number seeds in those two commands:

 

fix                1 O gcmc 1 1 1 1 29494 $T 9.0 0.1 group O

fix                2 C gcmc 1 1 1 2 271828 $T 9.0 0.1 group C

 

Best wishes´╝î

 

Wei

-----Original Messages-----
From:"Thompson, Aidan" <athomps@...3...>
Sent Time:2017-10-14 06:03:51 (Saturday)
To: "chenwei@...7134..." <chenwei@...7134...>, "lammps-users@lists.sourceforge.net" <lammps-users@lists.sourceforge.net>
Cc:
Subject: Re: [EXTERNAL] [lammps-users] how to perform grand canonical Monte Carlo simulation for binary mixture by LAMMPS?

There are a lot of complicating factors in your example. Also, it is gigantic. I can't check every possible thing you might be doing wrong. However, I did run my own tests on a modified version of in.gcmc.lj, in which I ran two gcmc fixes with two LAMMPS types with identical interaction coefficients.  The total density and pressure of the system came out slightly lower than for the 1-component case, while the excess chemical potential came out slightly higher. I think is due to the fact the ideal gas contribution to chemical potential is different by shifted by kTln(2) for 2-component case.  In any event, the numbers of types 1 and 2 are the same, within statistical noise.

 

 

--

      Aidan P. Thompson

      01444 Multiscale Science

      Sandia National Laboratories

      PO Box 5800, MS 1322      Phone: 505-844-9702

      Albuquerque, NM 87185     Fax  : 505-845-7442

      E-mail:athomps@...3... Cell : 505-218-1011

 

 

From: "chenwei@...7134..." <chenwei@...7134...>
Date: Tuesday, September 26, 2017 at 3:55 AM
To: lammps/lammps LAMMPS Users List <lammps-users@lists.sourceforge.net>
Subject: [EXTERNAL] [lammps-users] how to perform grand canonical Monte Carlo simulation for binary mixture by LAMMPS?

 

Dear LAMMPS users,

 

I try to perform grand canonical Monte Carlo simulation for binary mixture by LAMMPS.

 

I want a pure GCMC simulator.

 

Firstly, I test the program by the simulation of one component hard sphere fluid. The cutoff distance of LJ 12-6 potential is set equal to sigma, it is not a perfect hard sphere system, but the density of the system from the simulation is very close to the theoretical one.

 

However, there is a problem when I do the simulation for a binary hard-sphere mixture.  The densities predicted by the simulator are always far away from the confirmed theoretical ones.

 

The GCMC exchange is performed by these two commands in my script:

 

fix                1 O gcmc 1 1 1 1 29494 $T 3.982254 0.1 group O

fix                2 C gcmc 1 1 1 2 29494 $T 1.065682 0.1 group C

In addition, the density of the atoms handled by the second fix command is always close to zero. Here,  the density of group C is almost zero, but if I change the order of those two fix commands, then the density of group O is close to zero.

 

The LAMMPS Version is (6 Jul 2017).

 

Please find the full .in script and the .data file from the attachment.

 

Thank you very much,

 

Wei

 





--

State Key Laboratory of Multi-Phase Complex Systems
Institute of Process Engineering (IPE)
Chinese Academy of Sciences (CAS)
P.O. Box 353, Beijing 100190, China
Tel: +86-10-8254 4839

Fax: +86-10-6255 8065
Email: chenwei
@ipe.ac.cn
Http://www.mpcs.cn/emms