GFN-FF - A general force-field for elements Z=1-86

This repository contains a standalone implementation of the GFN-FF method by S.Spicher and S.Grimme (https://doi.org/10.1002/anie.202004239), and was adapted from the xtb code.

The default CMake build compiles a statically linked library (libgfnff.a) that can be linked in other Fortran, C and C++ projects.

---

Instructions (building from source)

Make sure you have the following dependencies installed:

• CMake and make
• Fortran and C compilers (e.g., gfortran/gcc), including LAPACK and OpenMP libraries (e.g. openblas)

Follow these steps to build the project assuming you are currently within the cloned repository:

1. Create a build directory and navigate to it

   mkdir _build
   cd _build

2. Run CMake to set up the build:

   • either directly:

     cmake ..
   • or if you wish to build the minimal example app:

     cmake .. -Dbuild_exe=true

3. Build the project and run the testsuite:

   make
   make test
   

libgfnff.a (and a gfnff app, when -Dbuild_exe=true was used in the CMake setup) can now be found in this (_build) directory.

---

In-code instructions

main.f90 in app/ demonstrates the (Fortran) in-code usage. main.c and main.cpp in test/ demonstrate the C and C++ in-code usage.

Two steps are required for using the GFN-FF library: 1. Initializing the calculator class (which stores topology and parametrization) and 2. calling the energy+gradient routine from it. A minimal Fortran example would look something like this:

use gfnff_interface
type(gfnff_data) :: calculator   !> model storage
integer :: nat                   !> number of atoms in the system 
integer :: at(nat)               !> atomic number for each atom
real(kind=real64) :: xyz(3,nat)  !> atomic coordinates in BOHR 
integer :: ichrg                 !> molecular charge
real(kind=real64) :: energy      !> energy in Hartree
real(kind=real64) :: grad(3,nat) !> gradient in Hartree/Bohr
integer :: io                    !> output status   

!> Read-in/define the system 
[...]

!> Initialize the calculator
call calculator%init(nat,at,xyz,ichrg=ichrg)

!> Calculate the energy and gradient routine
call calculator%singlepoint(nat,at,xyz,energy,gradient,iostatus=io)

!> Destroy calculator (once you are done)
call calculator%deallocate()

The calculator initialization is often more expensive than the actual energy evaluation due to the topology setup. However, once the calculator has been initialized, singlepoint energies can be called repeatedly.