[Pw_forum] Dispersion correction issue

[Michael Fischer]

Hello everyone,

I am trying to do some single point calculations for a porous system 
including a guest molecule using QuantumEspresso 5.0.2, employing the 
PBE functional and the Grimme-type dispersion correction (London=.true., 
other related parameters on default settings).

When comparing my results to earlier calculations using CASTEP with 
comparable settings, I found rather large differences, and it became 
clear quite quickly that the majority of these differences are due to 
the dispersion interactions. Below, I have extracted the dispersion 
contribution to the total energy (in Ry) obtained with the two different 
codes (sorry for awkward formatting):

                                CASTEP QuantumEspresso
Zeo                          -0.11569 -0.23136
Guest                      -0.00010                        -0.00020
Zeo + Guest            -0.12646                        -0.25290

It turns out that the dispersion energy from QuantumEspresso is 
practically exactly twice as large as the energy calculated by CASTEP. 
Looking at previous literature studies, I have good reason to "trust" 
the CASTEP energies (they are at least in the right ball park), while I 
must assume that the QE results are too large. Is this problem known to 
anyone?

I should possibly add that I successfully ran the test that involves the 
Grimme-type correction (the dispersion contributions in vdw.out and 
vdw.ref are the same).

Any help will be greatly appreciated.
Kind regards,
Michael