Impact of ngridq on Ground-State Indirect Band Gap with Spin-Orbit Coupling in GW Calculations

exciting
1

Hello everyone,
I am seeking clarification regarding the convergence of the indirect band gap with spin-orbit coupling (SOC) using different ngridq values. Specifically, I observed that the ground-state indirect band gap (KS + SOC) also changes as reflected in the GW_INFO.OUT files when ngridq is varied.

Interestingly, the values remained consistent when ngridq was set to 6 6 6 or lower, but they began to differ after increasing ngridq to 8 8 8.

Despite performing numerous GW calculations to converge parameters such as lmaxmb, gmb, etc., using the same ground-state parameters, only increasing ngridq led to changes in the KS + SOC indirect band gap.

Here are the details of my input files:

<groundstate
   do="fromscratch"
   rgkmax="8.0"
   epsengy="0.0000005"
   epschg="0.000001"
   ngridk="10 10 10"
   xctype="GGA_PBE"
   nempty="100">
   <spin spinorb="true"/>
</groundstate>

GW input:

<groundstate
   do="skip"
   rgkmax="8.0"
   epsengy="0.0000005"
   epschg="0.000001"
   ngridk="10 10 10"
   xctype="GGA_PBE"
   nempty="100">
   <spin spinorb="true"/>
</groundstate>
<gw
   taskname="g0w0"
   ngridq="8 8 8"
   nempty="100"
   ibgw="1"
   nbgw="20"
   coreflag="xal">
   <mixbasis
      lmaxmb="4"
      epsmb="1.d-4"
      gmb="1.0">
   </mixbasis>
   <freqgrid
      nomeg="12">
   </freqgrid>
</gw>

Best regards,
Sareh

2

Also, I should mention that the band energy in BAND-QP.OUT in GW+SOC wasn’t shifted with the Fermi energy, unlike in GW, KS or KS+SOC.