scf convergence problems /warnings with hybrid functionals

Dear all,

I am trying to run the Si example using hybrid functionals. However using the same input file as in the examples produces the famous error: "warning(occupy): minimum eigenvalue less than minimum linearisation energy : -2.057302621 -2.000000000 for s.c. lo".

I tried changing the inout file until it became :

! Band structure of silicon. The points listed after plot1d below are the
! vertices joined in the band structure.

tasks
0
5

! use Broyden mixing
mixtype
3

lmaxo
6

gmaxvr
18.0

rgkmax
9

msmooth
4

autolinengy
.true.

xctype
20

hybrid
.true.

maxscl
500

! hybrid mixing parameter
hybridc
0.25

avec
5.13 5.13 0.00
5.13 0.00 5.13
0.00 5.13 5.13

sppath
'/home/eamoujaes/elk-5.2.14/species/'

atoms
1 : nspecies
'Si.in' : spfname
2 : natoms; atposl below
0.0 0.0 0.0
0.25 0.25 0.25

ngridk
3 3 3

! These are the vertices to be joined for the band structure plot
!plot1d
! 7 200 : nvp1d, npp1d
! 0.0 0.0 1.0 : vlvp1d
! 0.5 0.5 1.0
! 0.0 0.0 0.0
! 0.5 0.0 0.0
! 0.5 0.5 0.0
! 0.5 0.25 -0.25
! 0.5 0.0 0.0

However, I am still getting the same error and the energy is fluctuating :

-550.460126129
-550.854497608
-550.460126129
-550.854497608
-550.460126129
-550.854497608
-550.460126129
-550.854497608
-550.460126129
-550.854497608
-550.460126129

with the absoute change in teh total energy 'freezing' at some value :

Absolute change in total energy (target) : 1.930849691 Absolute change in total energy (target) : 0.4274268605 Absolute change in total energy (target) : 0.3901818058 Absolute change in total energy (target) : 0.3950746912 Absolute change in total energy (target) : 0.3942406452 Absolute change in total energy (target) : 0.3943961704 Absolute change in total energy (target) : 0.3943672593 Absolute change in total energy (target) : 0.3943723027 Absolute change in total energy (target) : 0.3943715650 Absolute change in total energy (target) : 0.3943716652 Absolute change in total energy (target) : 0.3943716810 Absolute change in total energy (target) : 0.3943716376 Absolute change in total energy (target) : 0.3943715827 Absolute change in total energy (target) : 0.3943714794 Absolute change in total energy (target) : 0.3943714791 Absolute change in total energy (target) : 0.3943714792 Absolute change in total energy (target) : 0.3943714792 Absolute change in total energy (target) : 0.3943714792 Absolute change in total energy (target) : 0.3943714792 Absolute change in total energy (target) : 0.3943714792 Absolute change in total energy (target) : 0.3943714792 ( 0.1000000000E-03)
( 0.1000000000E-03)
( 0.1000000000E-03)
( 0.1000000000E-03)
( 0.1000000000E-03)
( 0.1000000000E-03)
( 0.1000000000E-03)
( 0.1000000000E-03)
( 0.1000000000E-03)
( 0.1000000000E-03)
( 0.1000000000E-03)
( 0.1000000000E-03)
( 0.1000000000E-03)
( 0.1000000000E-03)
( 0.1000000000E-03)
( 0.1000000000E-03)
( 0.1000000000E-03)
( 0.1000000000E-03)
( 0.1000000000E-03)
( 0.1000000000E-03)
( 0.1000000000E-03)

Can anybody please suugest how to correct this problem. I know this quesion has been asked a lot but I have tried a lot of things as you can see from my input file but in vain.

Thanks

Elie Moujaes
Federal University of Rondonia
Brazil

Andrew Shyichuk - 2019-04-02

Dear Elie,

The "warning(occupy): minimum eigenvalue less than minimum linearisation energy" should rather go away after several cycles.

The "freezing" scf might mean that the basis is too small.

Also, the 3x3x3 grid is way too small for Si, I'd use at least 10x10x10.
Try looking for Si in this forum, there might be more info.

Best regards.
Andrew

Elio - 2019-04-03

Dear Andrew,]

Thanks for your reply. The Si system is intended to be just a test to check how the hybrid functionals work. My system is completely different.

I have managed to change the input and the 'freezing' phenomenon disappeared. I know that 3x3x3 is a small grid and that a hybrid coefficient must alsio be conversed. However the results look weird. The bands got basically shifted and the TDOS curves, although got altered, still look strange to me. i am attaching them . Do these look sensical.

Regards