O1NumHess_QC

This is a Python library related to quantum chemistry calculations. This module interfaces with chemical software such as BDF, ORCA, etc., as well as the O1NumHess library, implementing the functionality to calculate the Hessian matrix by only needing to calculate gradients O(1) times when calling chemical software like BDF.

Users can calculate the Hessian matrix of molecules by specifying the .xyz molecular coordinate file and the corresponding input file for gradient calculations.

O1NumHess is unrelated to quantum chemistry and calculates the Hessian matrix by accepting a vector x and a user-provided gradient function g. O1NumHess perturbs each component of the input vector x separately, calls function g to calculate the gradient after each perturbation, and finally uses the gradients calculated from multiple perturbations to derive the Hessian.

Details of the O1NumHess algorithm, as well as preliminary benchmark results, can be found in our preprint paper: https://arxiv.org/abs/2508.07544

requirement

• Linux System
• python >= 3.6
• numpy
• O1NumHess

install

python3 setup.py install

this will create config files in ~/.O1NumHess_QC, copy the files xxx_config_example.py to xxx_config.py and modify it with your own condition.

Note that the above command installs the package for all users, and requires root privileges. If the user does not have access to root privileges, or if it is not desired to install the package for all users, then one should use the following command instead:

python3 setup.py install --prefix ~/.local

example

Detailed concept and instruction can be found in the documentation.

Here is an example shows how to use O1NumHess_QC and BDF to calculate Hessian.

test files:

• benzene.xyz:

  12
  
  C         -0.26396        0.29101        0.87006
  C         -0.72047       -0.95180        0.41795
  C          0.12718       -2.06420        0.45530
  C          1.43129       -1.93382        0.94486
  C          1.88770       -0.69106        1.39714
  C          1.04011        0.42134        1.35968
  H         -0.91966        1.15163        0.84107
  H          1.39331        1.38276        1.70952
  H          2.89652       -0.59018        1.77601
  H          2.08713       -2.79433        0.97377
  H         -1.72939       -1.05266        0.03928
  H         -0.22596       -3.02564        0.10546
  

• benzene.inp:

  $COMPASS
  Title
  benzene grad
  Basis
  cc-pvdz
  Geometry
  file=benzene.xyz
  End geometry
  $END
  
  $xuanyuan
  $end
  
  $scf
  RKS
  dft
  B3LYP
  spinmulti
  1
  $end
  
  $resp
  geom
  $end
  

• testQC.py:

  import numpy as np
  from O1NumHess_QC import O1NumHess_QC
  
  # read XYZ from file
  qc = O1NumHess_QC("../benzene.xyz")
  # qc = O1NumHess_QC("../benzene.xyz", unit="angstrom")  # specify the unit of xyz file manually
  # print(qc.xyz_angstrom)
  
  # parallel calculate Hessian
  hessian: np.ndarray = qc.calcHessian_BDF(
      method = "single",
      delta = 1e-3,
      core = 4,
      mem = "4G",
      inp = "../benzene.inp",
      encoding = "utf-8",
      tempdir = "~/tmp",      # BDF tempdir, each calculate will generate a subfolder at there
      task_name = "abc",      # all the output file will start with this string
      config_name = "",       # config name in your ~/.O1NumHess_QC folder config file
  )
  print(hessian)

it's recommended to run like this:

.
├── BDF_RUN/         # run in this folder
├── benzene.inp
├── benzene.xyz
└── testQC.py

run in BDF_RUN folder with python3 ../testQC.py, so that all the output file will in BDF_RUN.

unit test

run this command under project root folder:

python3 -m unittest

or simply run files like test_BDF.py in the test folder.