[QE-users] CPMD Electron Relaxation takes too long.

Giuliana Materzanini giuliana.materzanini at epfl.ch
Fri Jan 11 07:52:28 CET 2019


Dear Oier,


here are my comments/advices:


(*) I would recommend stricter convergence criteria: ekin_conv_thr = 
1.0d-11 (or 1.0d-9 at least), etot_conv_thr = 1.0d-6.


(**) dt is not the only parameter to set in CP: you should tune emass 
and emass_cutoff, for which you are now using the default values. Beside 
choosing a lower value of dt, also choosing a lower value of 
emass_cutoff should help you to avoid that "ortho goes bananas". In any 
case, you solved this. But remember that tuning these three parameters, 
emass, emass_cutoff and dt, will be crucial for your MD simulation (you 
should tune them in such a way that ekin_electrons is much lower than 
ekin_ions and that econt is conserved - see file .evp for these quantities).


(***) for electron_dynamics, why don't you use damped dynamics? Put 
electron_dynamics='damp', electron_damping=0.2, and if needed tune 
electron_damping (see the user's guide). In my experience damped 
dynamics is faster than sd, and this could be the reason why you are stuck.


(****) Just to clarify: step 1) is reaching the GS, that means relaxing 
the electrons; step 2) is relaxing the system, i.e. relaxing the ions.



Hope this helps!


Best regards,

    Giuliana


-- 
****************************************************
Giuliana Materzanini
Theory and Simulation of Materials
École Polytechnique Fédérale de Lausanne
****************************************************



On 10/01/19 17:29, Oier Arcelus wrote:
> Dear All,
>
> I am trying to run the cp.x to analyze Na ion dynamics into a rigid 
> layered host structure. The first test is to try to run the simulation 
> on a small unit cell of layered NaFeO2. As I read in the user guide 
> https://www.quantum-espresso.org/Doc/cp_user_guide/ there are few 
> steps. 1) Reach the GS 2) Relax the electrons 3) Run the dynamics. For 
> the first step I am using the following input file:
> User's Guide for The Quantum ESPRESSO Car-Parrinello Molecular 
> Dynamics <https://www.quantum-espresso.org/Doc/cp_user_guide/>
> Pietro Delugas 2018-03-16
> www.quantum-espresso.org
>
>  &control
>    calculation='cp'
>    restart_mode='from_scratch',
>    iprint = 10
>    isave = 100
>    ndr = 51
>    ndw = 51
>    nstep = 100
>    tstress = .TRUE.
>    tprnfor = .TRUE.
>    dt = 2.5d0
>    etot_conv_thr = 1.0d-5
>    ekin_conv_thr = 1.0d-4
>    pseudo_dir='/home/enxararo/NaFeO2/pseudos/',
>    outdir='./'
>    prefix='nafeo'
>  /
>  &system
>    ibrav = 0
>    nat= 12
>    ntyp= 3
>    ecutwfc = 90
>    ecutrho = 900
>    nr1b = 20, nr2b = 20, nr3b = 65
>    nspin = 2
>    tot_magnetization = 5.0
>    lda_plus_u = .TRUE.
>    Hubbard_U(2) = 3.9
>    vdw_corr = TS
>  /
>  &electrons
>    electron_dynamics = 'sd'
>  /
>  &ions
>    ion_dynamics = 'none'
>  /
>  &cell
>    cell_dynamics = 'none'
>    press = 0.0d0
>  /
>  ATOMIC_SPECIES
>  Na 51 Na.pbe-spnl-rrkjus_psl.1.0.0.UPF
>  Fe 72 Fe.pbe-n-rrkjus_psl.1.0.0.UPF
>  O 16.0 O.pbe-n-rrkjus_psl.1.0.0.UPF
>  ATOMIC_POSITIONS (crystal)
>   Na   0.333332980         0.666666973         0.166666995
>   Na   0.000000000         0.000000000         0.500000000
>   Na   0.666666991         0.333332969         0.833333020
>   Fe   0.000000000         0.000000000         0.000000000
>   Fe   0.666666991         0.333332969         0.333332991
>   Fe   0.333332980         0.666666973         0.666666980
>   O   0.666666991         0.333332969         0.067304000
>   O   0.000000000         0.000000000         0.266029012
>   O   0.333332980         0.666666973         0.400636998
>   O   0.666666991         0.333332969         0.599363031
>   O   0.000000000         0.000000000         0.733970988
>   O   0.333332980         0.666666973         0.932695992
>  CELL_PARAMETERS {angstrom}
>         3.0578720570         0.0000000000 0.0000000000
>        -1.5289360285         2.6481948829 0.0000000000
>         0.0000000000         0.0000000000  16.2363090515
>
> I've noticed a few things.
> 1) Trying to use the estimation of nr1b, nr2b, and nr3b that appears 
> in the user guide point 4.4.4 resulted in a SIGSEV:segmentation fault. 
> Reducing it to the values of the input I attached solved the problem.
> 2) Using a value of dt = 5.0 resulted in an error from routine 'ortho: 
> orthogonalization went bananas' or something like that. Setting it to 
> dt = 2.5 gets rid of this problem.
> 3) Last, and most importantly, I've been running this 12 atom system 
> for about 6 hours now and it is stuck on the following:
>
> .
> .
> .
>    Wave Initialization: random initial wave-functions
>    Occupation number from init
>    spin =   1 nbnd =    46
>     1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
>     1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
>     1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
>     1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
>     1.00 1.00 1.00 1.00 1.00 1.00
>    spin =   2 nbnd =    41
>     1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
>     1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
>     1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
>     1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
>     1.00
>
>    formf: eself=   519.42285
>    formf:     vps(g=0)=  -0.0016489     rhops(g=0)=  -0.0091918
>    formf: sum_g vps(g)=  -0.7698402 sum_g rhops(g)=  -6.4702082
>    formf:     vps(g=0)=   0.0136252     rhops(g=0)=  -0.0081705
>    formf: sum_g vps(g)=   4.1620329 sum_g rhops(g)=  -5.7512961
>    formf:     vps(g=0)=  -0.0001694     rhops(g=0)=  -0.0061279
>    formf: sum_g vps(g)=   0.3628532 sum_g rhops(g)=  -4.3134721
>    Delta V(G=0):   0.077012Ry,    2.095591eV
>
>    from rhoofr: total integrated electronic density
>    spin up
>    in g-space =     46.000000   in r-space =    46.000000
>    spin down
>    in g-space =     41.000000   in r-space =    41.000000
>
>      TS-vdW Pressure (GPa)        2.30153      0
>      Total Electronic Pressure (GPa)     7162.86920      0
>
> Is this normal? The scf calculations with pw.x does not take more than 
> a minute to solve this system. Any help is very much appreciated!
>
> Thanks in advance,
>
> Best regards,
>
> Oier.
>
> _______________________________________________
> users mailing list
> users at lists.quantum-espresso.org
> https://lists.quantum-espresso.org/mailman/listinfo/users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20190111/2d2b4708/attachment.html>


More information about the users mailing list