[Pw_forum] Binding Energy for H2O with PAW

[Ari P Seitsonen]

Dear Felix,

   First of all, the "hydrogen becomes metallic" probably means that you 
should apply spin polarisation as you have only one atom. Please remember 
that the absolute energies do not make sense, the total energy of the atom 
should be calculated by removing the energy difference between a pseudo 
atom and the all-electron atom, but this is prone to numerical 
(convergence) errors.

   Second, dissociating a water molecule by pulling one hydrogen apart 
would in principle lead to a double radical; well, in practise it would 
lead to two non-neutral constituents (in one calculation the electrons can 
move freely, in "metallic language" the Fermi level is global and same 
everywhere, even if the atoms would be 1 km apart). And further, most of 
the approximations, in particular GGAs, have a problem when pulling apart 
atoms, they give wrong asymptotics (famous examples are dissociating H_2).

   As I do not know what your target goal is, I would once more look at the 
literature on DFT and these asymptotic failures, whether the DFT (and the 
XC functional that you plan to use) is able to capture the correct 
asymptotics.

     Greetings from the Austrian Alps,

        apsi

PS Indeed, the quantity you would be calculating is rather dissociating 
energy, or?

-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-
   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, 26 Feb 2016, Felix Brockherde wrote:

> Dear all,
> 
> We would like to calculate the binding energy for non-equilibrium H2O geometries with PAW pseudo potentials.
> 
> The initial thought was to calculate the single-atom energies. However, for H the system becomes metallic. We used smearing but the results seem too far from the experimental
> value for us to use them (-0.91784682 Ry).
> 
> Now, we would like to calculate the binding energy by separating the H2O molecule as far as necessary. We would thus avoid inaccuracies caused by smearing because it is not a
> metallic system. However, when pulling H2O apart, the DFT calculations do not converge.
> Can anybody help with either how to get the separated H2O converged or a different approach to get the binding energy with PAW pseudo potentials?
> 
> We used PWSCF v.5.1.1 and I attached an input file where QE does not converge.
> 
> Best regards,
> Felix Brockherde
> TU Berlin
> 
>  &CONTROL
>     calculation = 'scf',
>     wfcdir = 'tmp',
>     pseudo_dir = '/home/user/pseudo_potentials',
>     tprnfor = .true.,
>     prefix = 'prefix',
>     tstress = .false.,
>     outdir = './',
>     wf_collect = .true.,
>  /
>  &SYSTEM
>     nat = 3,
>     ntyp = 2,
>     ecutwfc = 90,
>     ecutrho = 360,
>     celldm(1) = 30,
>     ibrav = 1,
>  /
>  &ELECTRONS
>  /
> ATOMIC_SPECIES
>     H 1.00794 H_pbe-kjpaw_psl.0.1.UPF
>     O 15.9994 O_pbe-kjpaw_psl.0.1.UPF
> ATOMIC_POSITIONS alat
>     O 0.5 0.3 0.5
>     H 0.5 0.7 0.3
>     H 0.5 0.7 0.7
> K_POINTS automatic
>    1 1 1   0 0 0
> 
>