Skip to content

Driver to combine dftd4 and minpack/nlopt for damping parameter optimization

License

LGPL-3.0, GPL-3.0 licenses found

Licenses found

LGPL-3.0
COPYING.LESSER
GPL-3.0
COPYING
Notifications You must be signed in to change notification settings

dftd4/dftd4-fit

Repository files navigation

DFT-D4 damping parameter optimization

This project provides a driver for optimization of damping parameters in the DFT-D4 method.

Installation

To use this project the following dependencies are required

  • dftd4 version 3.0.0 or newer, for evaluation of the DFT-D4 dispersion correction
  • nlopt version 2.0.0 to 2.5.0, for the optimization of the damping parameters
  • nlopt-f, to provide Fortran bindings for nlopt
  • minpack, for the optimization of the damping parameters
  • mctc-lib, for reading geometry inputs when creating the data set

Meson

Create a new build with

meson setup _build --prefix ~/.local

The Fortran compiler can be adjusted by setting the FC environment variable. Meson will find installed dependencies automatically or fetch them automatically if they are not available.

To build the project use

meson compiler -C _build

Finally, you install the binary with

meson install -C _build

Fortran package manager

This project is built with the Fortran package manager (fpm). Fpm will handle the dependencies for dftd4, minpack, nlopt-f and mctc-lib, only nlopt has to be provided on the system.

To create the binaray use

fpm install --profile release --flag -fopenmp

This will install dftd4-fit to ~/.local/bin on Unix and %APPDATA%\local\bin on Windows. Make sure the respective directories are in your PATH or adjust the installation prefix with the --prefix option.

Usage

git clone https://github.com/dftd4/dftd4-fitset
dftd4-fit -C dftd4-fitset ./data.csv

The data set must contain the missing dispersion energies for each (association) reaction in Hartree. The missing dispersion energy is calculated as follows:

$$\left(E^{\text{complex}}_{\text{reference}} - \sum_i^\text{monomers} E^{i}_{\text{reference}}\right) - \left(E^{\text{complex}}_{\text{DFT}} - \sum_i^\text{monomers} E^{i}_{\text{DFT}} \right) = \sum_i^\text{reactants} \vartheta^i E^{i}_{\text{reference}} - \sum_i^\text{reactants} \vartheta^i E^{i}_{\text{DFT}}$$

Correspondingly, this energy is negative for most functionals, because the reference method binds the complex more strongly. The lack of dispersion in DFT reduces the interaction energy.

Data set format 1: No stoichiometry coefficients

This format is the default (--format 1).

Note that the first entry must be the product (stoichiometry factor $\vartheta = 1$). All other entries are assumed to be educts (stoichiometry factor $\vartheta = -1$). An example is given below:

S22x5/01-0.9, S22x5/01-A, S22x5/01-B, -1.0007611865e-03
S22x5/01-1.0, S22x5/01-A, S22x5/01-B, -1.5228237266e-03
S22x5/01-1.2, S22x5/01-A, S22x5/01-B, -1.6586059147e-03
S22x5/01-1.5, S22x5/01-A, S22x5/01-B, -1.2297590834e-03
S22x5/01-2.0, S22x5/01-A, S22x5/01-B, -6.2420992500e-04
...

Data set format 2: Explicit stoichiometry coeffiecients

This option can be requested with --format 2.

Here, the stoichiometry coefficients must be given after the directory name, i.e., directory name and stoichiometry coefficient are given in an alternating fashion. The example from above would now look as follows:

S22x5/01-0.9, 1, S22x5/01-A, -1, S22x5/01-B, -1, -1.0007611865e-03
S22x5/01-1.0, 1, S22x5/01-A, -1, S22x5/01-B, -1, -1.5228237266e-03
S22x5/01-1.2, 1, S22x5/01-A, -1, S22x5/01-B, -1, -1.6586059147e-03
S22x5/01-1.5, 1, S22x5/01-A, -1, S22x5/01-B, -1, -1.2297590834e-03
S22x5/01-2.0, 1, S22x5/01-A, -1, S22x5/01-B, -1, -6.2420992500e-04
...

Data set format: Legacy

This option can be requested with --format 0.

In the first versions, the missing dispersion energy (target energy) is calculated the other way round:

$$E_{\text{target}} = \left(E^{\text{complex}}_{\text{DFT}} - \sum_i^\text{monomers} E^{i}_{\text{DFT}}\right) - \left(E^{\text{complex}}_{\text{reference}} - \sum_i^\text{monomers} E^{i}_{\text{reference}} \right) = \sum_i^\text{reactants} \vartheta^i E^{i}_{\text{DFT}} - \sum_i^\text{reactants} \vartheta^i E^{i}_{\text{reference}}$$

Hence, the energies are mostly positive.

In the fit, however, the error is defined as

$$E_{\text{target}} - E_{\text{D4}} = (E_{\text{DFT}} - E_{\text{reference}}) - E_{\text{D4}} = (E_{\text{DFT}} - E_{\text{D4}}) - E_{\text{reference}} \text{,}$$

which introduces a wrong sign ($- E_{\text{D4}}$ is positive, making the dispersion-corrected energy even more repulsive). Therefore, the reactions in the data set are taken as dissociations, i.e., the first entry must be the educt (stoichiometry factor $\vartheta = -1$) and all other entries are assumed to be products (stoichiometry factor $\vartheta = 1$). An example is given below:

S22x5/01-0.9, S22x5/01-A, S22x5/01-B, 1.0007611865e-03
S22x5/01-1.0, S22x5/01-A, S22x5/01-B, 1.5228237266e-03
S22x5/01-1.2, S22x5/01-A, S22x5/01-B, 1.6586059147e-03
S22x5/01-1.5, S22x5/01-A, S22x5/01-B, 1.2297590834e-03
S22x5/01-2.0, S22x5/01-A, S22x5/01-B, 6.2420992500e-04
...

More Information

For more information checkout the project help page

fpm run -- --help

For an overview over all command line arguments use the --help argument or checkout the dftd4-fit(1) manpage.

License

This project is free software: you can redistribute it and/or modify it under the terms of the Lesser GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This project is distributed in the hope that it will be useful, but without any warranty; without even the implied warranty of merchantability or fitness for a particular purpose. See the Lesser GNU General Public License for more details.

Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in this project by you, as defined in the Lesser GNU General Public license, shall be licensed as above, without any additional terms or conditions.