[QE-users] Best way to choose dt and nstep in AIMD/CPMD simulation using cp.x code in quantum espresso

[Kazume NISHIDATE]

Dear, Jayraj Anadani

> According to the user guide of cp.x we can use a larger time step
> "dt" for electron dynamics = "cg" but for my system of 54 atoms
> (metallic) at 2000 K temperature, it takes a lot of time to take 500
> steps for 36 core, 72 nodes.

The temperature 2000 K is very high and the ions must have high
velocities. It means that the ionic positions at the next time step
will largely different from the current ones. The convergence of the
wave functions will need many 'cg' steps accordingly.

So, you should set short 'dt' to ensure the first convergence of the
wave functions.

> 1) is there any way to balancing between timestep "dt" and number of
>  CP-MD steps "nstep" to achieve fast simulation ?,

It is worth to try the shorter time intervals (dt) with fixing other
parameters such as the total time steps.

> 2) In comparison with VASP most of the simulation like NVT, NPT, NVE
>  took 3 to 5 ps to equilibrate at a certain temperature so is there
>  any strategy to speed up the calculation ?

Generally, 3 to 5 ps is enough to achieve equilibrium state.  But I do
not understand your system and not sure if it is appropriate to your
case.

> 2022/09/15 16:18、Jayraj Anadani <jayrajanadani at spuvvn.edu>のメール:
> 
> Hello QE Community,
> According to the user guide of cp.x we can use a larger time step "dt" for electron dynamics = "cg" but for my system of 54 atoms (metallic) at 2000 K temperature, it takes a lot of time to take 500 steps for 36 core, 72 nodes. 

西館数芽
Kazume NISHIDATE Ph.D