Siesta

mulliken population analysis: problem with DZP basis

Asked by Dongzhe Li on 2018-12-06

Dear Siesta developers and users,

As a new user for Siesta, I got a strange problem with super simple system of benzene molecule in the gas phase. If I use SZ basis set with default parameters, I got reasonable atomic and orbital populations, namely about 0.90 electrons on "H" and 4.1 electrons on "C".

However, if I change to DZP basis defined as follows:

PAO.EnergyShift 100 meV
PAO.SplitNorm 0.15 ## default value
PAO.BasisSize DZP
PAO.BasisType split

I got strange atomic and orbital populations:

Species: H
Atom Qatom Qorb
               1s 1s 1Ppy 1Ppz 1Ppx
   2 1.244 0.842 0.298 0.047 0.037 0.020
   4 1.244 0.841 0.299 0.027 0.037 0.040
   6 1.244 0.841 0.299 0.027 0.037 0.040
   8 1.244 0.842 0.298 0.047 0.037 0.020
  10 1.244 0.841 0.299 0.027 0.037 0.040
  12 1.244 0.841 0.299 0.027 0.037 0.040

Species: C
Atom Qatom Qorb
               2s 2s 2py 2pz 2px 2py 2pz 2px
               2Pdxy 2Pdyz 2Pdz2 2Pdxz 2Pdx2-y2
   1 3.757 0.240 0.646 0.070 0.692 0.462 0.795 0.229 0.492
              0.031 0.008 0.005 0.035 0.053
   3 3.756 0.240 0.646 0.361 0.692 0.168 0.569 0.229 0.719
              0.047 0.028 0.005 0.014 0.036
   5 3.756 0.240 0.646 0.361 0.692 0.168 0.569 0.229 0.719
              0.047 0.028 0.005 0.014 0.036
   7 3.757 0.240 0.646 0.070 0.692 0.462 0.795 0.229 0.492
              0.031 0.008 0.005 0.035 0.053
   9 3.756 0.240 0.646 0.361 0.692 0.168 0.569 0.229 0.719
              0.047 0.028 0.005 0.014 0.036
  11 3.756 0.240 0.646 0.361 0.692 0.168 0.569 0.229 0.719
              0.047 0.028 0.005 0.014 0.036

"H" atom has more than one electron which is completely un-physical...

Then I started to play with "PAO.SplitNorm" and "PAO.SplitNormH" to re-define the radius of second \zeta function as follows:

PAO.EnergyShift 100 meV
PAO.SplitNorm 0.45
PAO.SplitNormH 0.50
PAO.BasisSize DZP
PAO.BasisType split

In this case, I got about 0.93 e on "H" and 4.06 e on "C" which is more reasonable.

So, my question is, what is the strategy to get the most trustable and transferable DZP basis?

One more question, is it enough to use SZ basis to describe well the charge transfer at the molecule-metal interface? Surprisingly, I got a \delta n > 0 (physically it should be <0) in the case of HOMO close to Fermi level...

Thanks a lot in advance for any comments.
Best regards,
Dongzhe Li

Department of Physics
Univ. Konstanz, Germany