Python Simulations of Chemistry Framework

Credit/Thank You

Board of Directors

  • Timothy Berkelbach (Columbia and Flatiron Institute)
  • Garnet Chan (Caltech)
  • Sandeep Sharma (CU Boulder)
  • Alexander Sokolov (Ohio State)
  • Qiming Sun (Axiomquant LLC)

Maintainers

  • Zhihao Cui (Caltech)
  • Yang Gao (Caltech)
  • James Smith (Flatiron Institute)
  • Verena Neufeld (Columbia)
  • Hong-Zhou Ye (Columbia)
  • Xiao Wang (UC Santa Cruz)
  • Susi Lehtola (MolSSI)

Outline

  1. What is PySCF?
  2. Feature Highlights
  3. Installation
  4. Open-source community and documentation

What does PySCF do?

Mostly generic two-body Hamiltonians: \[ \hat{\text{H}} = \sum_{pq} h_{pq} \hat{a}_p^\dagger \hat{a}_q + \sum_{pqrs} V_{pqrs} \hat{a}_p^\dagger \hat{a}_q \hat{a}_r^\dagger \hat{a}_s \]

Special forms (e.g. sparsity of V) are rarely exploited.

PySCF calculates \(h_{pq} \) and \( V_{pqrs} \) or users can provide them

Several basis options: Gaussian-type orbitals (GTOs) for molecules and solids and plane-wave

What does PySCF do (pt. 2)?

  • Energies (ground and excited state)
  • Forces and geometry optimization
  • Response properties (e.g. dipole moments)
  • Growing support for spectroscopic properties
  • Relativstic effects and non-collinear spins
  • Users can study systems at many levels of theory

Design Philosophy

A sandbox, but a performant one Python user interface with a C back-end 
-----------------------------------------------|---------
Language          files      blank    comment  |     code
-----------------------------------------------|---------
Python             1145      38662      52046  |   192523
C                    66       3006       2779  |    43508
reStructuredText     48       1414        541  |     4074
Lisp                  5         78        145  |      886
C/C++ Header         18        176        253  |      640
YAML                  7         25         61  |      302
CMake                11         63        189  |      247
...
-----------------------------------------------|---------
SUM:               1328      43546      56085  |   242730
-----------------------------------------------|---------
Since inception, our goal has been to treat molecules and solids on equal footing

Molecules and Solids!

Molecular 
from pyscf import gto, scf, cc

# Setup molecule
mol = gto.Mole()




mol.atom = """
C 0.00 0.00 0.00
C 1.68 1.68 1.68"""
mol.basis = "cc-pvdz"

mol.build()

# Run HF

my_mf = scf.HF(mol)
hf_energy = my_mf.kernel()

# Run Coupled Cluster
my_cc = cc.CCSD(my_mf)
cc_energy = cc.kernel()
Solid 
from pyscf.pbc import gto, scf, cc
    
# Setup molecule
cell = gto.Mole()
cell.a = """
0.00 3.37 3.37
3.37 0.00 3.37
3.37 3.37 0.00"""
cell.atom = """
C 0.00 0.00 0.00
C 1.68 1.68 1.68"""
cell.basis  = "cc-pvdz"
#cell.pseudo = "gth-pade"
cell.build()

# Run HF
cell.make_kpts([2,2,2])
my_mf = scf.HF(cell, kpts)
hf_energy = my_mf.kernel()

# Run Coupled Cluster
my_cc = cc.KCCSD(my_mf)
cc_energy = cc.kernel()

Correlated Methods

We have a lot of correlated methods that work well in molecules and solids

Xiao Wang

And now... plane waves

Hongzhou Ye and Verena Neufeld

Custom Hamiltonians

h1 = numpy.zeros((n,n))
for i in range(n-1):
    h1[i,i+1] = h1[i+1,i] = -1.0
    h1[n-1,0] = h1[0,n-1] = -1.0
eri = numpy.zeros((n,n,n,n))
for i in range(n):
    eri[i,i,i,i] = 2.0

mf = scf.RHF(mol)
mf.get_hcore = lambda *args: h1
mf.get_ovlp = lambda *args: numpy.eye(n)
mf.kernel()

mymp = mp.MP2(mf)
mymp.kernel()

mycc = cc.CCSD(mf)
mycc.kernel()

mycas = mcscf.CASSCF(mf, 4, 4)
mycas.kernel()

Interfaces with PySCF

DMRG
Block (Chan)
CheMPS2 (Wouters)
Selected CI
Arrow (Umrigar)
Dice (Sharma)

QMC
AFQMC (Morales/Zhang)
QMCPACK (LLNL)
NECI (Alavi)
Hande (Thom)
NN VMC
FermiNet (Deepmind)
DeepQMC/PauliNet (Noe)
NetKet (Carleo)
Quantum Computing
OpenFermion
Qiskit (IBM)
OpenQemist (1Qbit)

Somewhere in the future?
Maybe your research code!

How can I get PySCF?

Option 1: Using pip


								pip install pyscf

Option 2: Using pip to installing the latest version

pip install -e git+https://github.com/pyscf/pyscf.git

Option 3: Build PySCF from sources

Requirements:

  • C and Fortran Compiler
  • BLAS
  • CMake
  • Python 3, NumPy, SciPy, h5py
git clone https://github.com/pyscf/pyscf.git
cd pyscf/pyscf/lib
cmake -B build
cmake --build build --parallel

Documentation

Documentation

Examples

Over 350 examples!!

HF
MP2
CCSD
CASSCF
DFT
GW
TDDFT
EOM-CCSD
MRPT
ADC
FCI
X2C
MUCH MORE!
  • pip-install-able
  • Multiple levels of theory to calculate properties
  • If you see something you want and it's not in PySCF already, open an issue on Github or talk to me!
  • Let us know if you want to get involved!

Let's checkout the demos!

https://github.com/jamesETsmith/2022_simons_collab_pyscf_workshop

EXTRA SLIDES

Capabilities in Detail