External Methods Keywords in the Jaguar Input File (Machine learning methods and xTB)
- Overview
- Examples
Table 1 lists the keywords that are related to using machine learning methods with Jaguar calculations. Table 2 lists keywords that are related to using xTB with Jaguar calculations.
|
Keyword |
Value |
Description |
|
0 |
Calculate energy and derivatives quantum-mechanically. |
|
|
|
1 |
Calculate energy and derivatives with the Schrodinger-ANI neural-network potential energy surface, derived from fitting to QM calculations for a large set of drug-like molecules [294]. This surface is only available for the elements H, C, N, O, F, P, S, and Cl. As this is essentially a force-field calculation, setting this keyword only permits geometry optimizations and calculation of frequencies. |
|
|
2 |
Calculate energy and derivatives with the ANI-1ccx neural-network potential energy surface, derived from fitting to DFT and CCSD(T) calculations for a large set of drug-like molecules [293]. This surface is only available for the elements H, C, N, O. As this is essentially a force-field calculation, setting this keyword only permits geometry optimizations and calculation of frequencies. |
|
0 |
Calculate energy and derivatives quantum-mechanically. |
|
|
|
1 |
Calculate energy and derivatives with the direct-learned QRNN (charge recursive neural-network) potential energy surface, derived from fitting to DFT calculations for a large set of ionic drug-like molecules and their tautomers [303]. |
|
|
2 |
Calculate energy and derivatives with the delta-learned QRNN (charge recursive neural-network) potential energy surface, derived from fitting to the difference between QM and DFT calculations for a large set of ionic drug-like molecules and their tautomers [303]. |
|
|
3 |
Calculate energy and derivatives with the delta-learned QRNN (charge recursive neural-network) + transfer-learning potential energy surface, derived from fitting to the difference between QM and DFT calculations for a large set of ionic drug-like molecules and their tautomers. |
|
|
4 |
Calculate energy and derivatives with the direct-learned QRNN (charge recursive neural-network) + transfer-learning potential energy surface, derived from fitting to the difference between QM and DFT calculations for a large set of ionic drug-like molecules and their tautomers. |
|
0 |
Do not add implicit solvation to the QRNN calculations. |
|
|
|
1 |
Add implicit solvation to QRNN calculations. This setting only works if qrnn is set to 2 or 3. Select the solvent model with qrnn_solvent_model and the solvent with solvent. |
|
ALPB |
Use the analytic linearized Poisson–Boltzmann (ALPB) solvent model when implicit solvation is enabled [313]. |
|
|
|
GBSA |
Use the Generalized Born and surface area (GBSA) solvent model [325]. |
|
|
COSMO |
Use the Conductor-like Screening Model (COSMO) solvent model[232]. |
|
|
CPCMX |
Use the Extended Conductor-like Polarizable Continuum Model (CPCM-X) solvent model. Note: only single point energy calculations can be performed with this model as first and second derivatives are not available [326]. |
|
Any |
Calculate energy and derivatives with a machine learning force field (MLFF). For MPNICE (Message Passing Network with Iterative Charge Equilibration) machine learning force fields, the keyword value can be any model listed in this table. This method cannot be applied to open-shell systems. As this is essentially a force-field calculation, setting this keyword only permits geometry optimizations and calculation of frequencies. You can also use the UMA model, specified by the |
|
|
0 |
Do not add implicit solvation to the MPNICE calculations. |
|
|
|
1 |
Add implicit solvation to MPNICE calculations. This setting only works if mlff is set to Organic_MPNICE_TB. Select the solvent model with mlff_solvent_model and the solvent with solvent. |
|
ALPB |
Use the analytic linearized Poisson–Boltzmann (ALPB) solvent model when implicit solvation is enabled [313]. |
|
|
|
GBSA |
Use the Generalized Born and surface area (GBSA) solvent model [325]. |
|
|
COSMO |
Use the Conductor-like Screening Model (COSMO) solvent model[232]. |
|
|
CPCMX |
Use the Extended Conductor-like Polarizable Continuum Model (CPCM-X) solvent model. Note: only single point energy calculations can be performed with this model as first and second derivatives are not available [326]. |
|
water |
Solvent name. The allowed values are listed below. |
|
|
0 |
All inputs are treated as a molecular system even if periodic boundary condition (PBC) information is found. |
|
|
|
1 |
Attempt to find periodic boundary condition (PBC) information in the input
|
|
|
2 |
Attempt to find periodic boundary condition (PBC) information in the input
|
|
Keyword |
Value |
Description |
|
0 |
Calculate energy and derivatives quantum-mechanically. |
|
|
|
1 |
Calculate energy and derivatives with the GFN2-xTB semiempirical method [327]. |
|
0 |
Do not add implicit solvation to xTB calculations. |
|
|
|
1 |
Add implicit solvation to xTB calculations. This setting only works if xtb is set to 1. Select the solvent model with xtb_solvent_model and the solvent with solvent. |
|
ALPB |
Use the analytic linearized Poisson–Boltzmann (ALPB) solvent model when implicit solvation is enabled [313]. |
|
|
|
GBSA |
Use the Generalized Born and surface area (GBSA) solvent model [325]. |
|
|
COSMO |
Use the Conductor-like Screening Model (COSMO) solvent model[232]. |
|
|
CPCMX |
Use the Extended Conductor-like Polarizable Continuum Model (CPCM-X) solvent model. Note: only single point energy calculations can be performed with this model as first and second derivatives are not available [326]. |
|
water |
Solvent name. The allowed values are listed below. |
Available Solvents
The solvents available for use with the solvent keyword with the ALPB, COSMO, and CPCMX models are listed below.
|
acetonitrile |
aniline |
benzene |
|
dichloromethane |
chloroform |
carbon_disulfide |
|
N,N-dimethylformamide |
dimethylsulfoxide |
diethyl_ether |
|
ethanol |
ethyl_ethanoate |
n-hexadecane |
|
n-hexane |
nitromethane |
1-octanol |
|
toluene |
tetrahydrofuran |
|