Skip to content

aspuru-guzik-group/DiffiQult

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

29 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

DiffiQult

DiffiQult is an open source autodifferentiable quantum chemistry package.

docs/h2o_3g_MO_1.gif

Method:

  • RHF

Features:

  • Single point calculations
  • Energy gradients with respect to any parameter of the one-particle basis functions.
  • Energy optimization with respect of any parameter of the Gaussian basis functions.

Getting started with DiffiQult

Requirements

Installation

  • From source:

    git clone https://github.com/ttamayo/DiffiQult.git

    python setup.py install

Usage

Molecular system

We define the parameters of a molecular systems with an System_mol object:

  • molecular geometry in xyz format and atomic units
  • basis sets (data base so far sto_3G
  • number of electrons

For example:

# Basis set is sto_3G
from diffiqult.Basis import basis_set_3G_STO as basis
   # Our molecule H_2
   d = -1.64601435
   mol = [(1,(0.0,0.0,0.20165898)),(1,(0.0,0.0,d))]

   # Number of electrons
   ne = 2
   system = System_mol(mol,                                ## Geometry
                       basis,                              ## Basis set (if shifted it should have the coordinates too)
                       ne,                                 ## Number of electrons
                       shifted=False,                      ## If the basis is going to be on the atoms coordinates
                       mol_name='agua')                    ## Units -> Bohr

Tasks

The jobs in Diffiqult are managed by a Tasks object,

manager = Tasks(system,
             name='h2_sto_3g',      ## Prefix for all optput files
             verbose=True)          ## If there is going to be an output

where we defined the molecular system to optimize with the object system, and output options with verbose.

The class Task contains the method Tasks.runtask, it computes one the following options:

Task Key Description
Single point energies Energy It calculates the RHF energy and updates some attibute in system
Optimization Opt It optimizes a given parameter and updates the basis set in system

Single point calculation

manager.runtask('Energy',
             max_scf=50,                        # Maximum number of SCF cycles
             printcoef=True,                    # This will produce a npy file with the molecular coefficients
             name='Output.molden',              # Name of the output file (Compatible with molden)
             output=True)

Notes:

  • We currently don't have convergence options for the SCF.
  • The molden file also contains an input section that can be used as input for system with the option shifted
  • The geometry and MOs can be vizualized with molden,and the molden file.

Optimization

To optimize one or many input parameters, we use the option Opt. After a succesful optimization or If the optimization reaches the maximum number of steps or convergence, it updates the attributes of the system_mol object.

manager.runtask('Opt',
                 max_scf=50,
                 printcoef=False,
                 argnum=[2],  # Optimization of centers
                 output=True) # We optimized all the steps
print(manager.syste.energy)

where argnum recieves a list with the parameters to optimize with the following convention:

Parameter argnum
Widths 0
Contraction coefficients 1
Gaussian centers 2

for example, we can optimize the atomic centered basis function with respect of their widths and contraction coefficients in the following way.

manager.runtask('Opt',
                    max_scf=50,
                    printcoef=False,
                    argnum=[0,1],  # Optimization of centers
                    output=True)   # We print a molden file of all steps

Additionally, if output is set to True, a molden file of each optimization step is printed.

About

A fully autodifferentiable and variational HF

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages