DFTB3 D3(BJ) GAMMA

[Dobromir A Kalchevski]

Hello valued developers and users,

I am trying to setup DFTB3 with D3 correction and GAMMA damping and I can't find the answers to all of my questions in the forum threads. Also some answers are not completely clear to me.

1) If I am to use the D3 correction - is it enough to include 

          DISPERSION_TYPE D3(BJ)
          DISPERSION_PARAMETER_FILE ${LIBPATH}/dftd3.dat

or should I also include the &VDW_POTENTIAL section and where if I am not using DFT (&XC section)

2) does D3(BJ) work with DFTB3 or does only D3 work

3) are both the D3 and the D3(BJ) parameters in the dft3d.dat file. If not - for D3(BJ) I can find only a1, a2 and s8, but I can't find anywhere the fourth one. Why are the default values set to 0

4) is the keyword DFTB3_PARAM in KIND the Hubbard parameter and is it necessary. The 3ob .skf files technically include it, basically does CP2K read it from the .skfs.

5) is using 3ob enough for DFTB3 or do I have to include the keyword DIAGONAL_DFTB3 in &DFTB

6) if HB_SR_GAMMA is the GAMMA damping for hydrogen bonds, is its use recommended for general organics at the moment or it would rather worsen the results, as before

I am attaching an example input file for QM/MM Metadynamics.

Best Regards,

Dobromir

  

[Jürg Hutter]

Hi

1) and 2) Have a look in "tests/DFTB/regtest-vdw" . There you find some examples on how to use D3 with DFTB.
3) The D3(BJ) parameters have to be given in the input file. The data file only contains the atom specific parameters,
not the method parameters.
4) Yes, this is the DFTB3 parameter. I don't know if CP2K can read the parameter from the DFTB parameter file. I would have to check.
5) You need the keyword. According to the manual the default is not to use it.
6) I cannot answer this question.

regards
JH

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Dobromir A Kalchevski <dobro...@gmail.com>
Sent: Saturday, January 31, 2026 11:21 AM
To: cp2k
Subject: [CP2K:22073] DFTB3 D3(BJ) GAMMA

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com<mailto:cp2k+uns...@googlegroups.com>.
To view this discussion visit https://groups.google.com/d/msgid/cp2k/e18f9019-605d-408b-a268-100e577c52f8n%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/e18f9019-605d-408b-a268-100e577c52f8n%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Dobromir A Kalchevski]

Hello,

many thanx for the answers.

Yes, I found the DFTB3 examples in the test files and I tried those inputs, but to no avail. I played around with some of the settings and I think I found something that works, but I'm no sure it gives the right barrier.

With the settings below (the rest of the input is the same, as the uploaded in the first post) I get broken calculations - either the reagent breaks or different H atoms (one from the MM region, one from the QM) get way too close and the calculation stops:

DFTB3 + D3(BJ) with default D3BJ_SCALING from the test files, no HB_SR_GAMMA (as per the test files) and no EWALD (as per the test files) - eventually the reagent (QM region) breaks in pieces:

      &DFTB
        DIAGONAL_DFTB3 T
        !ORTHOGONAL_BASIS F
        SELF_CONSISTENT T
        DO_EWALD F
        DISPERSION T
        !HB_SR_GAMMA T
        &PARAMETER
          PARAM_FILE_PATH ${LIBPATH}/DFTB/3ob-3-1
          PARAM_FILE_NAME param_file

          DISPERSION_TYPE D3(BJ)
          DISPERSION_PARAMETER_FILE ${LIBPATH}/dftd3.dat

          D3BJ_SCALING 1.0 0.4 1.5 4.0
          !HB_SR_PARAM 4.0
        &END PARAMETER
      &END DFTB

----

DFTB3 + D3(BJ) with default D3BJ_SCALING from the test files + HB_SR_GAMMA (= 4.2, as per the original paper) and no EWALD (as per the test files) - H atoms get way too close (< 1 A) and the calculations stops:

      &DFTB
        DIAGONAL_DFTB3 T
        !ORTHOGONAL_BASIS F
        SELF_CONSISTENT T
        DO_EWALD T
        DISPERSION T
        HB_SR_GAMMA T
        &PARAMETER
          PARAM_FILE_PATH ${LIBPATH}/DFTB/3ob-3-1
          PARAM_FILE_NAME param_file

          DISPERSION_TYPE D3(BJ)
          DISPERSION_PARAMETER_FILE ${LIBPATH}/dftd3.dat

          D3BJ_SCALING 1.0 0.4 1.5 4.0
          HB_SR_PARAM 4.2
        &END PARAMETER
      &END DFTB

----

DFTB3 + D3(BJ) with original D3BJ_SCALING from the original paper + HB_SR_GAMMA (= 4.2, as per the original paper), no EWALD (as per the test files) - H atoms get way too close (< 1 A) and the calculations stops:

      &DFTB
        DIAGONAL_DFTB3 T
        !ORTHOGONAL_BASIS F
        SELF_CONSISTENT T
        DO_EWALD F
        DISPERSION T
        HB_SR_GAMMA T
        &PARAMETER
          PARAM_FILE_PATH ${LIBPATH}/DFTB/3ob-3-1
          PARAM_FILE_NAME param_file

          DISPERSION_TYPE D3(BJ)
          DISPERSION_PARAMETER_FILE ${LIBPATH}/dftd3.dat

          D3BJ_SCALING 1.0 .5719 .5883 3.6017
          HB_SR_PARAM 4.2
        &END PARAMETER
      &END DFTB

---

ORTHOGONAL_BASIS T always breaks the calculation from the beginning (not a single geometry in the output), so I guess 3ob-3-1 does not have an orthogonal basis or DFTB3 simply does not use one. As I can see - the default in CP2K is FALSE, anyway.

DO_EWALD F breaks even DFTB2 calculations - the reagent (QM region) breaks in pieces

---

However, the following setup seems to work, although I'm not sure I can trust the barrier (it is very sensitive to intermolecular interactions and it seems kind of high, and there is no available experimental value):

DFTB3 + D3(BJ), no D3BJ_SCALING keyword (I let cp2k determine s6, sr6, s8 automatically, as suggested in a recent new thread in this google group) + HB_SR_GAMMA (=4 as usual default), with EWALD - it seems to work:

      &DFTB
        DIAGONAL_DFTB3 T
        SELF_CONSISTENT  T
        DO_EWALD  T
        DISPERSION  T
        HB_SR_GAMMA T
        &PARAMETER
          PARAM_FILE_PATH ${LIBPATH}/DFTB/3ob-3-1
          PARAM_FILE_NAME param_file
        !  UFF_FORCE_FIELD uff_table

          DISPERSION_TYPE D3(BJ)
          DISPERSION_PARAMETER_FILE ${LIBPATH}/dftd3.dat

          HB_SR_PARAM 4
        &END PARAMETER
      &END DFTB

---

Next I will try to explicitly provide the Hubbard parameters with in the input with DFTB3_PARAM.

Thank you again for the help.

If you can think of anything else - please - tell me.

Best Regards,

Dobromir

[Jürg Hutter]

Hi

I would strongly suggest to check that CP2K can correctly read your SK files.
When DFTB was included in CP2K, only the files distributed with CP2K were
available. Since then the DFTB community has added new functionality and
parameters that are missing in CP2K.

regards
JH

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Dobromir A Kalchevski <dobro...@gmail.com>

Sent: Tuesday, February 3, 2026 1:12 AM
To: cp2k
Subject: Re: [CP2K:22078] DFTB3 D3(BJ) GAMMA

To view this discussion visit https://groups.google.com/d/msgid/cp2k/17f23652-327a-4472-a00e-62ffbce76edfn%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/17f23652-327a-4472-a00e-62ffbce76edfn%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Dobromir A Kalchevski]

Thank you again,  Jürg.

Best Regards,

Dobromir