[QE-users] Relax calculation with meta-gga r2SCAN+rVV10

[Michele Re Fiorentin]

Dear QE developers and users,

I’m running GPU-accelerated QE 7.3.1 (on the CINECA LEONARDO cluster) to perform the relaxation of an MoTe2 bilayer using the r2scan meta-gga functional + rVV10 through Libxc:

input_dft='XC-000i-000i-000i-000i-497L-498L-rvv10'

SCF calculations do not pose any problem and convergence can be achieved without issues. However, relax calculations do not converge and energy and forces start oscillating. Moreover, several bfgs steps give energy_new > energy_old. I’ve increased the cutoff value to ensure force convergence and also reduced the trust_radius to decrease the ionic displacements, unfortunately to no avail.
I know that meta-gga functionals can be tricky, I was wondering if you have any suggestions to solve this problem.
I’ve attached the input/output of the relax calculation.

Thank you very much for your attention and help.
Best,



Michele




--
Michele Re Fiorentin, PhD

Department of Applied Science and Technology (DISAT)
Politecnico di Torino
corso Duca degli Abruzzi 24, 10129 Torino (Italy)
tel: +39 0110904333

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20241206/bb7bdc84/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: input_output.zip
Type: application/zip
Size: 57041 bytes
Desc: input_output.zip
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20241206/bb7bdc84/attachment.zip>