Slow convergence HSE - surface calculation

[rafael3k]

Dear all,

I am now doing a slab calculation. It is OK, if the LDA, GGA or METAGGA is used. I want to improve the band-structure results, then (HSE) Hybrid functional is probably the best choice. However, I am obtaining a really slow convergence. Please, how could I solve this?.

OSZICAR:
N E dE d eps ncg rms ort
SDA: 1 0.521107446700E+03 0.52111E+03 -0.75608E+03 3600 0.189E+04 0.000E+00
DMP: 2 0.401762863222E+02 -0.48093E+03 -0.28507E+03 3600 0.538E+03 0.458E+03
DMP: 3 -0.114635257422E+03 -0.15481E+03 -0.15027E+03 3600 0.356E+03 0.520E+02
DMP: 4 0.147715338467E+04 0.15918E+04 -0.16387E+05 3600 0.529E+05-0.313E+05
DMP: 5 0.778616708179E+03 -0.69854E+03 -0.88471E+04 3600 0.239E+05-0.466E+04
DMP: 6 0.134921854811E+04 0.57060E+03 -0.39034E+05 3600 0.101E+06-0.100E+05
DMP: 7 0.172890121826E+05 0.15940E+05 -0.14367E+07 3600 0.363E+07-0.998E+05
DMP: 8 0.765845589729E+05 0.59296E+05 -0.10134E+08 3600 0.255E+08-0.330E+06
DMP: 9 0.147922837281E+06 0.71338E+05 -0.15718E+08 3600 0.395E+08-0.568E+06

Thank you,

[alex]

Hi Rafael,

one way to improve speed of convergence is to do an e.g. GGA calculation first, save the wavefunction. Copy this WAVECAR and CHG* into the HSE directory and just start the HSE calc. again. Should work better than without.

Cheers,

alex

[Tarasingh]

jk
<span class='smallblacktext'>[ Edited Fri Nov 15 2013, 12:51PM ]</span>