[QE-users] NVE simulation in cp.x - large fluctuation in temperature

[Ari P Seitsonen]

Dear Jie,

   I do not know about the other options, but

  1) simulating a periodic system of only three atoms with only the Gamma 
point cannot lead to an accurate description of the electronic structure, 
and thus the forces on the ionic cores,

  2) the standard deviation in the fluctuations is supposed to be 
~sqrt[2/(3*N_atoms)) * T_average ~= 350 K in your case, and the 
instantaneous fluctuations are of course larger

   Just my quick thoughts...

     Greetings from Sunny Zurich,

        apsi

-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-
   Ari Paavo Seitsonen / Ari.P.Seitsonen at iki.fi / http://www.iki.fi/~apsi/
     Ecole Normale Supérieure (ENS), Département de Chimie, Paris
     Mobile (F) : +33 789 37 24 25    (CH) : +41 79 71 90 935


On Fri, 13 Apr 2018, Jie Peng wrote:

> Dear all
> I have been running MD simulations on HfS2 using cp.x code in Quantum espresso. I start from initial
> configuration obtained from pwscf vc-relax, and relax the system using cp.x by consecutive steps of:
> electron relaxation->ionic relaxation->cell relaxation. Then, I just directly start a NVE simulation
> starting from the equilibrium configuration. I expect the system to almost stay stationary or the
> temperature should be very small since I am allowing dynamics in a system that is already in
> equilibrium. However, what I see is a huge fluctuation in the tmpp output of cp.x, as I attach a figure
> showing variation of tmpp (Ionic temperature) with simulation time
> [IMAGE]
> 
> I did this because it is suggested in the user guide you should apply an initial displacement to the
> atoms in your system after the relaxation since otherwise there will not be any dynamics. But what I
> see here is a large fluctuation of the system temperature.
> 
> The thinking or questions here are
> 
> 1.Does the tmpp represents the physical temperature of the system here? I think it should be since it
> is the temperature corresponding to kinetic energy of the ions.
> 
> 2.It above point is true, why is the temperature varying so fiercely? Am I setting incorrect
> parameters, for instance the timestep or the fictitious mass? But I took those from previous simulation
> steps where I did the relaxation, and they all worked well since they successfully drived my system to
> equilibrium, satisfying the convergence threshold on total energy, forces acting on atoms, and the
> fictitious electron kinetic energy. I am confused at this point.
> 
> The input file for NVE simulation is attached here:
> 
> &control
>     calculation='cp',
>     title='Halfnium disulfide'
>     restart_mode='restart',
>     ndr=53,
>     ndw=54,
>     nstep=50000,
>     iprint=10
>     isave=100,
>     tstress = .true.
>     tprnfor = .true.
>     dt=10,
>     wf_collect=.true.
>     etot_conv_thr=1e-6
>     forc_conv_thr=1e-3
>     ekin_conv_thr=1e-5
>     prefix='HfS2',
>     pseudo_dir='/home/jpeng/HfS2/potential'
>     outdir='./tmp/',
>  /
>  &system
>     ibrav= 4,
>     a=3.6529
>    c=5.6544
>     nat=  3, ntyp= 2,
>     ecutwfc =50
>     vdw_corr='DFT-D',
>  !   lspinorb=.true.
>  !   noncolin=.true.
>  !   ecutrho=300
>  !   nbnd=14
> !    occupations='smearing'
> !    smearing='gaussian'
> !   degauss=0.01
>  !  nspin=2
>  !   starting_magnetization(1)=0.1
> ! Hf  95.94  Hf.pbe-mt_fhi.UPF
> ! S  32.065  S.pbe-mt_fhi.UPF
> /
>  &electrons
>     electron_dynamics='verlet'
>     electron_velocities='zero'
>     emass=400
>     emass_cutoff=1
> /
>  &ions
>     ion_dynamics = 'verlet'
>     ion_damping=0.1
> !    ion_nstepe=10
>  /
>  &cell
>     cell_dynamics = 'none'
>  
> /
> ATOMIC_SPECIES
>  Hf  95.94  Hf.pbe-mt_fhi.UPF
>  S  32.065  S.pbe-mt_fhi.UPF
> ATOMIC_POSITIONS (crystal)
> Hf      -0.000000000  -0.000000000  -0.000000000
> S        0.666666667   0.333333333   0.257234636
> S        0.333333333   0.666666667  -0.257234636
> 
> Anyone could help me on it? Thank you very much.
> 
> Best
> Jie
> --
> ------------------------------------------------------------------------------------------------------
> ------------------
> Jie Peng
> PhD student
> 2134 Glenn Martin Hall, Mechanical Engineering, University of Maryland
> College Park, Maryland, USA
> Phone:(+1) 240-495-9445
> Email: jiepeng at umd.edu
> 
> 
>