Dear all,

it took me some time to figure this out. Therefore, I will share the how to: the trick is the modifiction of the first line of each skf file AND the counting which differs from the file format as published on dftb.org.

http://www.dftb.org/fileadmin/DFTB/public/misc/slakoformat.pdf

So, this is what you have to do:

1) Homonuclear Skf files,

a. line 1: 2.000000000000E-02 650 6 6

one has to add the atom numbers for the elements concerned as values 3 and 4

b. line 1: value 2: the value defines the Hamiltonian and overlap matrix values on the grid has be set so that with (value2 + 3) you reach the line with the word Spline

2) Heteronuclear Skf files,

a. line 1: .000000000000E-02 651 6 7

one has to add the atom numbers for the elements concerned as values 3 and 4

b. line 1: value 2: the values define the Hamiltonian and overlap matrix on the grid

has be set so that with (value2 + 2) you reach the line with the word Spline

c. for each heteronuclear interaction there have to be 2 files, e.g. CN.skf and NC.skf

3) Gaussian input file, keyword line:

a. # dftb OPT

b. frequencies have to be calculated numerically according to the gaussian manual

4) AtFiles: ( the skf files have to be loadedwhich is indicated by a @ )

a. Prevent the contents of the skf files to be printed in the output by adding a /N to each line

_______________________

#p dftb OPT

cyano

-1 1

C 0.00000 0.00000 0.00000

N 1.20000 0.00000 0.00000

@/path/nn.skf/N

@/ path /cn.skf/N

@/ path /nc.skf/N

@/ path /cc.skf/N

__________________________

Dear Linda, you can to visit the web site of Gaussian software and make all your quastions and you may have the answers.

good luke.

Nice, I was trying to figure out how to make this work for months.

The manual was not clear.

I have an additional question : suppose you need to do a TS search

using opt=(ts,calcfc) or calchf, you get :

Unimplemented Fock Drv or Freq. in FoFSE

I think there is a problem with the evaluation of the hessian. The only solution I can imagine is to store a freq job in checkpoint and use it to start the TS search. Is there a more simple way?

Dear Linda,

thank you very much for the offer to help. Where you able to read my last descriptions of what I did before they where erased? I do not have access to my office PC and therefore can only try to describe what I did from my memory.

I tried to add an empty line, the "1", another empty line and then the @path/file.skf

That did not help either. I experiemtned with reducing the number of grid points to 500 by removing lines, I tried to remove the word Spline, the spline block, the line itself etc.

Do you have a running example that you would be willing to share including the skf file. I have the files myself in the framework of a collaboration project, so I am allowed to use them anyhow.

If so, I would provide you with my email address.

Absolutely...I sent you an e-mail. I can't do that just at this moment, but I certainly will.

What parameter file are you working with (pbs, mio, etc)?

Dear all,

it took me some time to figure this out. Therefore, I will share the how to: the trick is the modifiction of the first line of each skf file AND the counting which differs from the file format as published on dftb.org.

http://www.dftb.org/fileadmin/DFTB/public/misc/slakoformat.pdf

So, this is what you have to do:

1) Homonuclear Skf files,

a. line 1: 2.000000000000E-02 650 6 6

one has to add the atom numbers for the elements concerned as values 3 and 4

b. line 1: value 2: the value defines the Hamiltonian and overlap matrix values on the grid has be set so that with (value2 + 3) you reach the line with the word Spline

2) Heteronuclear Skf files,

a. line 1: .000000000000E-02 651 6 7

one has to add the atom numbers for the elements concerned as values 3 and 4

b. line 1: value 2: the values define the Hamiltonian and overlap matrix on the grid

has be set so that with (value2 + 2) you reach the line with the word Spline

c. for each heteronuclear interaction there have to be 2 files, e.g. CN.skf and NC.skf

3) Gaussian input file, keyword line:

a. # dftb OPT

b. frequencies have to be calculated numerically according to the gaussian manual

4) AtFiles: ( the skf files have to be loadedwhich is indicated by a @ )

a. Prevent the contents of the skf files to be printed in the output by adding a /N to each line

_______________________

#p dftb OPT

cyano

-1 1

C 0.00000 0.00000 0.00000

N 1.20000 0.00000 0.00000

@/path/nn.skf/N

@/ path /cn.skf/N

@/ path /nc.skf/N

@/ path /cc.skf/N

__________________________

Thanks to Linda and Andreas, and I have solved the problem for runing DFTB with Gaussian 09 (D.01). One thing to supplement to Andreas's answer is that, the format of the skf files are quite free, but the input for Gaussian is strick.  the blank line between 'Spline' string and the "integral table" data should be deleted. Then change the line 1 as Andreas has said.

Hi, I was wondering if I could get some help with running dftb with Gaussian. I've read through these previous posts and have adjusted my .skf files. I keep on running into the same error message which says it is expecting a floating point number, but receiving a string. I know that string is Spline and I have adjusted line 1: value 2 many times and keep on running into the same error. Has anyone had the same problem or can help me?

Michael,

When I ran into that particular error, the culprit ended up being not counting the number of points correctly.  Good news is, this is an easy fix.  If you would like, send me your input file and I can take a quick look.

Best,

Linda

Hi Linda. 

I have the same problem as Michael.

Can you share your solution please?

Thanks!

Luiz,

First, check the number of points you have.  I use excel for this.  If you are confident in the number of points, check your input file.  Here is a sample:

# opt geom=connectivity dftb

Title Card Required

0 1

C -1.92291219 0.18518518 -0.02517231 

H -1.56623935 0.68958337 -0.89882382

H -2.99291219 0.18519837 -0.02517231

1 2 1.0 3 1.0  #atom connectivity, atom1 is connected to atom 2 via a single bond..

2

3

@C:/G09W/C-H_pbc.skf  # all atom pair interactions must be included

@C:/G09W/H-C_pbc.skf

@C:/G09W/H-H_pbc.skf

Connectivity and atomic interactions are two of the more common sources of error.  Make sure you have the line spacing as Gauss expects it to be.

Pair interactions need to account for ALL atom pairs, bonded or non bonded.  So, COOH would look like:

C-O

O-C

O-O

O-H

H-O

C-H

H-C

If none of the above are the source for your errors, feel free to send me your skf and input files, and I will take a look.

Best,

Linda

Denis, did you find a solution to  opt=(ts,calcfc)?

Maria,

Has someone sent you the file yet?

Linda

Cristina,

I tried running the dispersion correction on G09 (C.01) with the IOP but didn't have any luck.  I was told Gauss was possibly going to include the ability for proper dispersion correction in later versions but have not followed up on it to see if it came to fruition.

I use DFTB+ (linux platform but is free for academics) for dftb dispersion correction calculations.  DFTB+ runs the parameter files as they are without any alterations.

Best,

Linda

Thanks to Linda's help I was able to run DFTB with external slako files in Gaussian. However, there are important Gaussian features that don't work with DFTB. For example, opt=(ts,calcfc) as mentioned by Denis. It seems that those features weren't improved in G16. An advice, don't rely on Gaussian as far as DFTB is concerned. There are other codes e.g. DFTB+ ( recommend by Linda) CP2K and ADF. Maybe they don't have all Gaussian buttons (features) but I believe they are worth a try.