Multiwfn forum

wham09

Member

Registered: Yesterday

Posts: 7

Question on charge decomposition analysis

Dear all,

I am having a problem doing charge decomposition analysis. The analysis program itself seems to be working, but there are multiple odd things.

First, my complex and the two components are all not open-shell systems, but the programs says otherwise. See below:

Loading basic information of complex... Please wait
Alpha electrons:    193     Beta electrons:    193     Multiplicity:  1
The number of atoms in complex:       109
The number of basis functions in complex:      1460
....
Input .mwfn/.fch/.molden/.gms file of fragment   1
....
Loading basic information of this fragment... Please wait
Alpha electrons:    144     Beta electrons:    144     Multiplicity:  1
The number of basis functions in this fragment:      1108
The number of atoms in this fragment:        85
....
Input .mwfn/.fch/.molden/.gms file of fragment   2
Loading basic information of this fragment... Please wait
Alpha electrons:     49     Beta electrons:     49     Multiplicity:  1
The number of basis functions in this fragment:       352
The number of atoms in this fragment:        24

Loading orbitals information for complex...
Note: The complex is an open-shell system
Flip electron spin of complex? (y/n)
y
Loading orbitals information for fragment   1...
Note: This fragment is an open-shell system
Flip electron spin of fragment   1? (y/n)
y
Loading orbitals information for fragment   2...
Note: This fragment is an open-shell system
Flip electron spin of fragment   2? (y/n)
Y

So I tried all 8 combinations of yes and no's to try analysis anyways, but the following problems were not solved by this.

1. The numbers of d, b, r are often very large and seems wrong. Below is a part of my data:
....
     178    1.000000  -25.809571  -22.818702   -2.990869  -43.363119
     179    1.000000  366.313880 -135.214450  501.528330 -202.616715
     180    1.000000  -50.070482  -10.590196  -39.480286  185.599172
     181    1.000000  272.544276  -13.180447  285.724723  -46.430657
     182    1.000000 -130.499120   -2.286675 -128.212445  -59.984506
....
      Sum:     193.000000  200.675262  294.833313  -94.158051 -270.692556

2. The fragment orbital contributions to molecular orbitals look really wrong. See below:
Alpha orbital   127 of fragment  1, Occ: 1.00000    Contribution:11725.51 %
Alpha orbital   128 of fragment  1, Occ: 1.00000    Contribution:38893.07 %
Alpha orbital   129 of fragment  1, Occ: 1.00000    Contribution:******** %
Alpha orbital   130 of fragment  1, Occ: 1.00000    Contribution: 5021.00 %
Alpha orbital   131 of fragment  1, Occ: 1.00000    Contribution:45494.56 %
Alpha orbital   132 of fragment  1, Occ: 1.00000    Contribution:62854.03 %

My Gaussian input commands look like this:

%nprocshared=8
%mem=16GB
%chk=....chk
#p
scf=(maxcycle=300)
sp
pop=(full)
iop(3/33=1)
uM06/gen
pseudo=read
nosym
int=ultrafine

[filename]

[charge/mult]

[coordinates]

P C H F 0
6-311+G**
****
Pd 0
SDD
****

Pd 0
SDD

After calculations were done, I generated .fch files by formchk and used them as the input files.

I'd be very grateful if anyone can tell what is the problem.

Thanks for your help in advance!

sobereva

Tian Lu (Multiwfn developer)

From: Beijing

Registered: 2017-09-11

Posts: 2,076

Website

Re: Question on charge decomposition analysis

You can find relevant description in Section 3.19 of Multiwfn manual.

1 If the whole system or ANY fragment is open-shell, then CDA will be conducted in the open-shell way, so you will need to choose whether or not to flip spin for each fragment.

Please do not arbitrarily try combinations. Please keep in mind that you need to properly set spin flip, so that sum of alpha (beta) electrons of all fragments equals to alpha (beta) electrons of the whole system, otherwise the analysis cannot be performed.

2 Please NEVER employ diffuse functions, which usually make CDA result very terrible. def2-TZVP commonly is a nice choice.

By the way, for Gaussian users, I suggest directly using .fch file as input file, and you do not need to add "pop=(full) iop(3/33=1)" in the input file.

wham09

Member

Registered: Yesterday

Posts: 7

Re: Question on charge decomposition analysis

Thank you for the quick reply!

1. Would M06 / def2-TZVP / SDD be a fine choice then?

2. The whole system and both fragments are not open-shell (you can see above that multiplicity of all three are 1. So I still don't understand why this happens.

wham09

Member

Registered: Yesterday

Posts: 7

Re: Question on charge decomposition analysis

I have one more question. Should I NOT employ diffuse functions also when I want to do orbital composition analysis or NCI analysis?

sobereva

Tian Lu (Multiwfn developer)

From: Beijing

Registered: 2017-09-11

Posts: 2,076

Website

Re: Question on charge decomposition analysis

Thank you for the quick reply!

1. Would M06 / def2-TZVP / SDD be a fine choice then?

2. The whole system and both fragments are not open-shell (you can see above that multiplicity of all three are 1. So I still don't understand why this happens.

1 It is fine

2 Because you have set UM06, therefore the system is calculated in the unrestricted formalism even if it is essentially a singlet closed-shell system. For a close-shell system, there is no reason to add "U" prefix to the functional name.

sobereva

Tian Lu (Multiwfn developer)

From: Beijing

Registered: 2017-09-11

Posts: 2,076

Website

Re: Question on charge decomposition analysis

I have one more question. Should I NOT employ diffuse functions also when I want to do orbital composition analysis or NCI analysis?

Different orbital composition analysis methods have different compatibility with diffuse functions, they have been documented in the manual. For example, if you use Mulliken or SCPA method to calculate orbital compositions, then diffuse functions should not be used, while methods like Hirshfeld and Becke can be used in any case.

All real space function analysis, such as IGMH, IRI, valence electron density, NCI, all kinds of AIM analysis, etc., are fully compatible with diffuse functions.

wham09

Member

Registered: Yesterday

Posts: 7

Re: Question on charge decomposition analysis

I have one more question. Should I NOT employ diffuse functions also when I want to do orbital composition analysis or NCI analysis?

Different orbital composition analysis methods have different compatibility with diffuse functions, they have been documented in the manual. For example, if you use Mulliken or SCPA method to calculate orbital compositions, then diffuse functions should not be used, while methods like Hirshfeld and Becke can be used in any case.

All real space function analysis, such as IGMH, IRI, valence electron density, NCI, all kinds of AIM analysis, etc., are fully compatible with diffuse functions.

What about NBO method for orbital compositions?

Thanks for the explanations!

sobereva

Tian Lu (Multiwfn developer)

From: Beijing

Registered: 2017-09-11

Posts: 2,076

Website

Re: Question on charge decomposition analysis

What about NBO method for orbital compositions?

Thanks for the explanations!

All NBO related analyses are fully compatible with diffuse functions, because these analyses are not directly performed on the top of original basis functions, diffuse functions do not notably affect the NBO results.

wham09

Member

Registered: Yesterday

Posts: 7

Re: Question on charge decomposition analysis

Thank you very much!