[QE-users] can i use this trick to do faster MD calculation using pw.x ?

[Kazume NISHIDATE]

Dear Jayraj;

> the temperature fluctuations are too high compared to the other
>  literature.

Because you are using "thousands of atoms" in LAMMPS while you are using
only the 128 atoms in QE.

> MD simulation of thousands of atoms in LAMMPS
…….
> On my metallic system, which has 128 atoms,


You must understand the concept of ensemble average.
A good text was introduced by Nicola:

> 3) not sure where to start - you need to learn and understand MD - a
>  good start is if you google "ercolessi md primer"


In addition to his elemental comments, I want to add some trivial
things.

> "conv_thr" (i.e. ecutwfc=5 or 10) below a minimum cutoff of
>  pseudopotential? So that I can get the next ionic position very

Suppose you have a potential curvature having a shape like a cup and 
a ball is falling along its curvature from its edge. The ball will
perform oscillating behavior continually if there is no energy loss by
a friction (energy conservation).

In a primitive molecular dynamics calculation (MD), total energy
(kinetic+potential) must be preserved in the long time scale (say
10000 time steps) when the system does not have a restriction on its
kinetics (like temperature regulation). 

The "ecutwfc=5" apparently insufficient to evaluate the potential
curvature and hence your MD calculation will collapse. It is 
equivalent to a MD in LAMMPS with random interaction potentials.

Use the default cutoff energy and monitor the fluctuation of total
energies in the MD runs with changing the dt (time increment value) to
find the appropriate (largest) dt with preserving the total energies.

And you can also consider a gamma-point only calculation if your 
system is large enough.



西館数芽
Kazume NISHIDATE Ph.D

Department of Systems Innovation Engineering,
Graduate School of Science and Engineering, Iwate University
4-3-5 Ueda, Morioka, Iwate, 020-8551 JAPAN
Phone:+81-19-621-6391
kazume.nishidate at gmail.com, nisidate at iwate-u.ac.jp
https://sites.google.com/site/nisidatelab/