DihedralParametrizer GUI

WaNo and Settings

Input files and parameters: (DihedralParametrizer2 tab)

  1. Molecule input:

    • Deposit PDB: Import PDB file of the molecule, either from your hard drive or from the Parametrizer WaNo preceding the DihedralParametrizer WaNo in the Workflow.
    • Deposit SPF: Import SPF file containing the nonbonded force field (electrostatic and Lennard Jones) of the molecule, either from your hard drive or from the Parametrizer WaNo preceding the DihedralParametrizer WaNo in the Workflow. Note: Dihedral energy profiles and force-fields are only computed if the spf file contains dihedral information (automatically generated by the Parametrizer).
  2. Settings The standard settings are depicted in the screenshot of the DihedralParametrizer WaNo above.

    • Number of Steps: Number of data points for the spline fit of the resulting dihedral force field. Recommended value: 32 or larger. Explanation: To compute the potential of a dihedral angle, each dihedral angle is rotated around 360° in equal sized steps. At each step the dihedral angle is kept fixed at the current value and the rest of the geometry is relaxed using Metropolis Monte-Carlo with a classical force field. DFT single point calculation is then performed on the resulting configuration and the energies of all configurations are used to compute a dihedral force field in form of a spline fit on the data points.
    • Engine: Like QuantumPatch, the DihedralParametrizer employs Quantum Chemistry Engines to compute molecular energies. For the DihedralParametrizer, DFTBplus (DFTB as implemented in DFTB+), XTB and Turbomole (DFT) are available. If Turbomole (requires a Turbomole license and installation not included in the software package) is chosen, additional settings for basis-set and functional are enabled. For reasons of efficiency we recommend to use DFTB, unless your molecule contains atom types not supported in DFTB+ (see list of supported atom types). The XTB option is relatively new, so please use with precaution. You can find a description of QM engines here.
    • Advanced settings: As of now, we recommend to use standard settings.
      • The "optimize forcefield" option provides better forcefields, but is still in an experimental phase.
        • Algorithm settings: *scf iterations: The number of iterations of the above mentioned algorithm. In each iteration, the dihedral force-field of the previous step is applied for the relaxation of the molecule, leading to an increased accuracy of the force-field energy. For standard runs, 2 scf iterations are sufficient. If the "optimize forcefield" is used (see below), do at least 3 scf iterations.
          • MC step multiplier: For each evaluation point of the dihedral potential, the remaining molecular configuration is relaxed to avoid clashes. For a very complex molecule, the standard number of MC relaxation steps can be scaled up using this value.
          • evaluation points: Number of random dihedral configurations used to assses quality of the dihedral forcefield (see CorrelationPlots.png in output files) validated against single point DFT/DFTB values.
        • intra forcefield settings:
          • There are various options to generate internal forcefields. The standard approach uses an internal LJ potential (with standard LJ parameters) and a spline dihedral energy fitted to DFT energies. To use this standard forcefield, leave everything as is.
          • To generate forcefields with better correlation to quantumchemistry eneriges (experimental phase) proceed as follows:
            • check the "optimize forcefield" option
            • increase scf_iterations (see above) to 3 or more.
            • Leave Trainigsset as "Set of Vac Mols". The dihedral parametrizer module will then generate random molecular configurations to fit the forcefield parameters.
            • Number of evaluation points needs to be increased to four times the number of atoms in your molecule.
            • Leave train set acc temp at 1000K
          • Remarks:
            • Only molecules with intramolecular forcefields generated by the same dihedral parametrization protocol can be run simultaneously in Deposit, e.g. for mixed morphologies.
            • The optimization option increases the runtime of the dihedral parametrizer. If molecules do not converge, please let us know at info.nanomatch.com.
            • You can see the impact of the forcefield optimization in this pdf.
      • Estimate configurational disorder: At the end of the DHP run, a set of random molecular configurations is sampled using the internal DH-FF along with internal interactions applied later in Deposit. HOMOs and LUMOs of these samples are computed to estimate configurational disorder of this material. Note: Runtime is increased significantly when this option is set.

Parallelization and Performance: (Resources tab)

DihedralParametrizer is parallelized and scales almost linarly with the number of processors up to the total number of performed QC computations. This number increases with the number of dihedral angles in the molecule and "Number of steps" (see above). For a molecule with only a single dihedral angle, e.g. the number of cores should be "Number of steps" plus one or lower. The required memory depends on the size of your compound. For average sized small organic compounds (~100 Atoms) 1000MB are sufficient. Check the manual of your DFT engine for details. For larger molecules with multiple dihedral angles, allocate a full node (e.g. 32 processors with 64GB RAM).

For every option not checked in the „Resources“ tab, the server defaults that depend on your HPC machine will be used. Check individual options to modify specific values.

Recommended resources for DihedralParametrizer:

Resource Recommendation
CPUs per Node: The calculation of each data point (see „Number of Steps“ above) can be performed on a CPU each. For larger molecules with multiple dihedral angles, allocate a full node.
Number of Nodes: 1 (No parallelization on more than a single node available)
Memory: 2GB per core recommended.
Walltime: Depending on the size of the compounds and the number of data points and the number of dihedrals of your molecule, the parametrization takes between 1h and 24h.
Queue: Default

Output Files

Overview

Filename File description Interface to module...
molecule.pdb PDB-file containing atom positions. Deposit
dihedral_forcefield.spf Force-field file for Deposit including the nonbonded force field (partial charges and Lennard-Jones parameters) and the dihedral force field computed for this module. Deposit
iter_X/snap_files_LJ/*.png Plots of dihedral forcefield (dihedral_.png), VdW contribution to the forcefield (vdw_only_.png) and DFT(B) contribution to the forcefield (dft_only_*.png). None
dihedral_X/homos_dihedral_X.png HOMO profile of dihedral with id X when rotating around 2pi. An example is given below. None
dihedral_X/lumos_dihedral_X.png LUMO profile of dihedral with id X when rotating around 2pi. An example is given below. None
disorder_estimate.yml Configurational disorder estimated by computing HOMO/LUMO for 500 random configurations sampled at 300K (temperature can be adapted via command line) using final dihedral forcefield and internal interactions. None

Example for HOMO/LUMO profiles

HOMO profile for Triphenylamine LUMO profile for Triphenylamine

 

 

The results of the search are