HOMO-LUMO gap in Aun

Clicked A Few Times

3:35:30 PM PDT - Wed, Mar 19th 2014

Hi all,

I am a casual user of NWChem, excuse me for my questions

In output of nwchem I get

     Non-variational initial energy

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

Total energy =  -18741.839624

1-e energy   =  -27215.363774

2-e energy   =    7860.022803

HOMO         =       0.048963

LUMO         =       0.048963

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

Warning - the HOMO and LUMO are degenerate and you are using symmetry.

This can lead to non-variational energies and poor convergence.

Modify the initial guess, or use an open-shell wavefunction, or turn

off symmetry.

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

My questions are:

1. this value of homo-lumo is really the final value (I'm confused by the initial word)

2. In table below first row from gaussian (lanl2dz) second row from nwchem (zora/sarc-zora) e final row from dirac12 (dyall.v2z).
Au 0,33 0,32 0,36
Au2 0,33 0,25 0,28
Au3 0,32 0,19 0,30
Au4 0,29 0,18 0,22
Au5 0,30 0,12 0,24

what happens that the values ??differ so much with increasing number of carbon atoms,
gaussian and dirac are close...

3. I worry about the same Warnning above the total energy being similar to other simulations?

thanks
Prof. Phd. André Luiz Fassone Canova

Forum Regular

6:27:45 PM PDT - Tue, Apr 8th 2014
Hi Andre, Ad 1. If the HOMO and LUMO orbitals are degenerate than any rotation between them will generate another wavefunction with supposedly the same energy. However, when you are using symmetry these orbitals might belong to different irreps and rotations among them would break the symmetry of the wavefunction. When that happens the rules that the code uses for eliminating integrals that are supposed to be zero by symmetry no longer hold and trouble might result. There are a number of ways to deal with this issue: use a different initial guess that lifts the degeneracy (obviously will not help if the degeneracy is supposed to exist in the final solution); use an open-shell wavefunction to replace a 2 0 occupation with a 1 1 occupation for example; or turn the use of symmetry off so that the code makes no assumption about the symmetry of the orbitals and the associated integrals. I guess the latter is actually the easiest option. Ad 2. I am not sure what the numbers you are listing are. I am assuming they are orbital energies. A problem I have with these numbers is that they are obtained using different basis sets and different methods. Whereas one might think the differences are a bit big it is impossible to tell this way whether this is right or wrong. NWChem should have the LanL2DZ basis set. So running the same NWChem as you did for Gaussian and then comparing the orbital energies wil allow you to check the agreement between Gaussian and NWChem. I am not sure whether the LanL2DZ basis sets were optimized with or without relativistic effects. Depending on that you may want to subsequently try SARC-ZORA with or without ZORA and see what that does to the orbital energies. Alternatively, it might be possible to the Dyall.v2z basis in NWChem with ZORA. Ad 3. Running the calculation without symmetry can always be used as a baseline check. However, we are not aware of any issues at present. Huub

Forum >> NWChem's corner >> General Topics