negative homo - lumo gap

[Claudio]

Dear all,

sometimes after I converge a calculation to an accuracy of say 1.0E-5
I get things like

HOMO - LUMO gap [eV] : -0.980594

This is because the Homo Energy is 0.22417642 and the Lumo Energy is
0.18814027 in AU.

I have requested the same accuracy for homo and lumo and have added
enough iterations to the lumo calculation loop to make sure the
program claims it is converged. I am using the OT method. How do I
prevent this from happening? I mean how do I ensure that the
calculation yields Lumo energies that are higher than the homo energy?


Sorry to bother you and thanks for your help!

Claudio

[marci]

Hi Claudio,

typically this is an indication that your system is metallic and that
the
selected optimization procedure (OT) is not suitable. You may try to
use diagonalization
as optimization scheme, adding more MOS and smearing the occupation
numbers around the Fermi energy.
Regards
Marcella

[Ian Hamilton]

Hi Marcella,

One of my students has encountered this problem. He tried changing the OT and increasing the number of MOs (we don't know what you mean by "smearing the occupation numbers around the Fermi energy") but it didn't correct the problem. I’m posting this on his behalf because he was unable to- he got a message that he had “limited access to public groups in his domain.”

Do you (or anyone else) have another suggestion? In his pdos files all the levels look fine except for the HOMO and LUMO.

From his output:

 Fermi Energy [eV] :   -5.966796

  Lowest Eigenvalues of the unoccupied subspace spin            1

 -----------------------------------------------------

  Reached convergence in          149  iterations 

      -0.23489989     -0.21167052     -0.18538586     -0.17943768

      -0.13796253     -0.13007839

 HOMO - LUMO gap [eV] :   -0.425155

 ENERGY| Total FORCE_EVAL ( QS ) energy (a.u.):            -1076.972424168399129

Best regards,

Ian Haminton

[marcella Iannuzzi]

Hi Ian,

without more information about the calculation you are performing it is hard to guess what is going on.

By default, the  unoccupied states are calculated only a posteriori. Here the problem seems to be that the ground state is not properly optimised.

Either the system has no gap, i.e., it is metallic, or there is some problem with the DFT settings. 

If the system is metallic, one should use a different scheme for the optimisation that is able to apply the partial occupation of the KS states, by assigning the occupation number as a function of the KS energy and of the Fermi energy. 

This is not possible using OT. From your message it is not clear what did Claudio actually try to correct the problem.

Best

Marcella

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.

[Ian Hamilton]

Hi Marcella,

Thank you for your reply. Let me tell you a little bit about the system and why, although a negative HOMO-LUMO gap doesn't make sense, I'm no longer

worried about the situation.

The student is doing calculations on graphene quantum dots which (unlike graphene) are well-known to gave a gap. He is looking at the effect on the gap 

of replacing hydrogens with either electron donating or electron withdrawing groups. He is also looking at the effect on the gap of replacing hydrogens with

bulkier groups at either armchair or zigzag positions since it has been suggested that a gap can form in graphene if it is strained and that the effect is different

for strain in the armchair versus zigzag directions. 

The calculation for which he got the negative HOMO-LUMO gap, was one in which, due to too much strain or congestion, the "optimized" geometry was completely 

nonphysical. 

Best,

Ian