Multiwfn forum

bernard

Member

Registered: 2020-08-24

Posts: 5

On spin densities of radical cation/anion

Hi Dr. Lu,

I was trying to calculate the Parr functions for an aldehyde at the GFN2-xTB level. From the geometry of the neutral molecule, I did single-point calculations of the cation/anion radical and generated molden files. I submitted these files to Multiwfn > Population Analysis > Mulliken. I did get the spin population for the N+1 below
Population of atoms:
     Atom      Alpha pop.   Beta pop.    Spin pop.     Atomic charge
     1(C )      2.09026      2.00568      0.08457        -0.09594
     2(C )      1.97617      1.89595      0.08023         0.12788
      :                                                                            ....

but obtained a different one for N-1 that looks like the following
Population of atoms:
Atom     1(C )    Population:  3.99823433    Net charge:  0.00176567
Atom     2(C )    Population:  3.78793229    Net charge:  0.21206771
   :                                                                                         .....

which is appears to to be similar in form for the neutral molecule
Population of atoms:
Atom     1(C )    Population:  4.05190053    Net charge: -0.05190053
Atom     2(C )    Population:  3.83075642    Net charge:  0.16924358
   :                                                                                         .....

Am I doing something different?

Thank you.

Bernard

sobereva

Tian Lu (Multiwfn developer)

From: Beijing

Registered: 2017-09-11

Posts: 1,830

Website

Re: On spin densities of radical cation/anion

Dear Bernard,

I didn't encounter this problem. Taking H2CO as example, the .xyz file is

4
sdf
 C                 -0.00000000   -0.00000000   -0.52715724
 H                 -0.00000000    0.93830873   -1.11461622
 O                  0.00000000    0.00000000    0.67402199
 H                  0.00000000   -0.93830873   -1.11461622

I ran the following command for calculating cation (N-1) state:

xtb H2CO.xyz --chrg 1 --uhf 1 --molden --gfn 2

The resulting molden file is attached: molden.zip

The result of Mulliken analysis by Multiwfn 3.7 is

 Population of atoms:
     Atom      Alpha pop.   Beta pop.    Spin pop.     Atomic charge
     1(C )      1.99581      1.80948      0.18633         0.19471
     2(H )      0.41303      0.25619      0.15683         0.33078
     3(O )      3.17814      2.67813      0.50000         0.14373
     4(H )      0.41303      0.25619      0.15683         0.33078
 Total net charge:  1.00000000      Total spin electrons:  1.00000000

The result looks very normal. The version of xtb I am using is version 6.3.2 (2020-07-02). If you are using older version of xtb, please try the latest one.

Best regards,

Tian

bernard

Member

Registered: 2020-08-24

Posts: 5

Re: On spin densities of radical cation/anion

Dear Sir,

Thank you very much for the reply.

Bernard

bernard

Member

Registered: 2020-08-24

Posts: 5

Re: On spin densities of radical cation/anion

Dear Sir,

I tried the suggestion above using the following coordinates for the aldehyde. However, I still cannot obtain the spin densities for the cation. Thank you for the help.

Sincerely,

Bernard

15
aldehyde
C          0.88400       -0.50500        0.00000
C          1.18400        0.87700        0.00000
C         -0.45200       -0.92600        0.00000
C          0.12900        1.79700        0.00000
C         -1.17800        1.36800        0.00000
C         -1.46600        0.00300        0.00000
O          2.43400        1.33000        0.00000
H          0.36900        2.84800        0.00000
C          1.95800       -1.47900        0.00200
H         -0.67400       -1.98200       -0.00000
H         -1.98600        2.08300       -0.00000
Cl        -3.11600       -0.51700       -0.00000
H          3.05100        0.55900        0.00000
O          3.14400       -1.20500        0.00100
H          1.62800       -2.53200       -0.00000

sobereva

Tian Lu (Multiwfn developer)

From: Beijing

Registered: 2017-09-11

Posts: 1,830

Website

Re: On spin densities of radical cation/anion

By default xtb employs Fermi-smearing technique to achieve fractional occupations. If the gap of frontier MOs is not large, in other words, the fronter MOs are nearly degenerate, the orbital occupation numbers will no longer be integer but partial. In this case, the orbital occupation numbers in the exported molden file are also non-integer, and thus Multiwfn is unable to properly deal it as an open-shell wavefunction, but view it as a closed-shell wavefunction represented by natural orbitals, and thus spin population cannot be obtained.

The only workaround is seemingly adding "-etemp 0" argument to xtb command, so that Fermi-smearing is disabled. Unfortunately, I tested your system with this argument, the SCF convergence cannot achieve... Currently I don't know if there is any other way to make SCF converged while maintaining integer occupation number. You may ask developers of xtb code for help.

Alternatively, you can use Fukui function instead. Calculation of Fukui function doesn't suffer from aforementioned problem, since it is only dependent of total density. Multiwfn can easily calculate Fukui function based on molden of xtb.

bernard

Member

Registered: 2020-08-24

Posts: 5

Re: On spin densities of radical cation/anion

Thank you, Sir, for the information.

bernard

Member

Registered: 2020-08-24

Posts: 5

Re: On spin densities of radical cation/anion

Is there a way to get the required wfn files out of the molden file without having to convert them via module 100/2? When I do first convert molden --> wfn, the new wfn files generate the Fukui indices but not the quantities related to IP and EA. I suppose I could use the energies of the radicals minus the energy for the neutral atom for the vertical IP and EA.

Last edited by bernard (2020-08-24 19:45:47)

sobereva

Tian Lu (Multiwfn developer)

From: Beijing

Registered: 2017-09-11

Posts: 1,830

Website

Re: On spin densities of radical cation/anion

Current version of xtb is unable to directly output .wfn file. (You may ask its developer to add this function ;-)

If you use subfunction 16 of main function 100 to get Fukui indices along with IP and EA, .wfn file must be used because Multiwfn not only loads orbital wavefunction, but also loads electronic energy from .wfn files. However, it is not compulsory to record electronic energy in .molden file, and if you use the .wfn converted from .molden file, the .wfn also doesn't record electronic energy. Therefore, if you need IP and EA, please manually calculate them according to their definitions.