Python for Quantum Chemistry

A Full Stack Programming Guide

1st Edition, Volume 23 - March 28, 2025
Author: Qiming Sun
Language: English
Paperback ISBN:
9 7 8 - 0 - 4 4 3 - 2 3 8 3 7 - 6
eBook ISBN:
9 7 8 - 0 - 4 4 3 - 2 3 8 3 6 - 9

Quantum chemistry requires ever higher computational performance, with more and more sophisticated and dedicated Python scripts being required to solve challenging problems. Al… Read more

Purchase options

World Book Day Celebration!

Save up to 30% on books and eBooks!

Shop now

Institutional subscription on ScienceDirect

Request a sales quote

Resources

Textbook support for instructors(opens in new tab/window)

Quantum chemistry requires ever higher computational performance, with more and more sophisticated and dedicated Python scripts being required to solve challenging problems. Although resources for basic use of Python are widely (and often freely) available online and in literature, truly cohesive materials for advanced Python programming skills are lacking.

Qiming Sun, a developer of the popular Python package PySCF, provides a comprehensive, end-to-end practical resource for researchers and engineers who have basic Python programming experiences chiefly in computational chemistry but want to take their use of the software forwards to the next level, the book provides an insightful exploration of Numpy, Pandas, and other data analysis tools. Readers will learn how to manage their Python computational projects in a professional way, with various tools and protocols for computational chemistry research and general scientific computing tasks exhibited and analysed from a technical perspective. Multiple programming paradigms including object-oriented, functional, meta-programming, dynamic, concurrent, and vector-oriented are illustrated in various technology scenarios allowing readers to properly use them to enhance their program projects. Readers will also learn how to use the presented optimization technologies to speed up their Python applications, even to the level as fast as a native C++ implementation. The applications of these technologies are then demonstrated using quantum chemistry Python applications.

Python for Quantum Chemistry: A Full Stack Programming Guide is written primarily for graduate students, researchers and software engineers working primarily in the fields of theoretical chemistry, computational chemistry, condensed matter physics, material modelling, molecular simulations, and quantum computing.

Title of Book
Cover image
Title page
Table of Contents
Copyright
Preface
List of acronyms
Chapter 1: Python programming environment
1.1. Python package system
1.1.1. Python runtime
1.1.2. Package, sub-package, and module
1.1.3. Python program as CLI tools
1.1.4. Virtual environment
1.1.5. Conda
1.2. Python programs in Docker
1.3. Managing a Python project
1.3.1. Version control
1.3.2. Testing a Python program
1.3.3. Coding style and static checks
1.3.4. Releasing a Python package
1.3.5. CI and DevOps workflow
1.4. Modular design and object-oriented programming
1.4.1. Method resolution order in multiple inheritances
1.4.2. Mixins
1.5. Working with IPython and Jupyter
1.5.1. Startup configurations
1.5.2. Extensions
1.5.3. Magics
1.5.4. Remote execution
Summary
Chapter 2: Data processing
2.1. Vectorized data processing with NumPy
2.1.1. The basics of NumPy
2.1.2. Universal functions (ufunc)
2.1.3. Inplace operations
2.1.4. Broadcasting
2.1.5. Fancy indexing
2.1.6. Mask array
2.1.7. Data structure of NumPy ndarray
2.1.8. Array views
2.1.9. The reshape function
2.2. Data types in NumPy
2.2.1. Type casting
2.2.2. Scalar type and zero-dimensional array
2.2.3. Infinity (inf) and not-a-number (nan)
2.2.4. Data with high precision
2.2.5. Structured array
2.3. Data with labels: Pandas
2.3.1. Pandas data objects
2.3.2. Broadcasting
2.3.3. Indexing
2.3.4. query and eval methods
2.3.5. Altering structure of DataFrame
2.3.6. Data types
2.3.7. Missing data
2.3.8. Grouping and aggregation
2.3.9. View and copy
Summary
Chapter 3: Visualization
3.1. Matplotlib
3.2. Pandas visualization
3.3. Mayavi for 3D plotting
3.4. Quantum chemistry visualization
3.4.1. Jinja template
3.4.2. Molden format
3.4.3. Cube format
Summary
Chapter 4: Scientific computing tools
4.1. Linear algebra
4.2. Sparse matrices
4.2.1. Storage formats
4.2.2. Linear algebra for sparse matrix
4.2.3. Linear operator
4.3. Tensor contractions
4.4. Discrete Fourier transforms
4.4.1. SciPy FFT
4.4.2. Intel FFT
4.4.3. PyFFTW
4.4.4. FFT performance benchmark
4.4.5. How to choose FFT libraries?
Summary
Chapter 5: Meta-programming and non-numerical computation
5.1. Code generation
5.1.1. eval and exec
5.1.2. Composing code programmatically
5.1.3. Manipulating classes and functions at runtime
5.1.4. Python AST
5.2. Symbolic computation
5.3. Automatic differentiation
5.3.1. PyTorch
5.3.2. JAX
Summary
Chapter 6: Input and output
6.1. Serialization
6.1.1. Pickle
6.1.2. JSON
6.1.3. YAML
6.2. File I/O
6.2.1. The npy format
6.2.2. HDF5 storage
6.2.3. Memory mapping
6.3. I/O buffering
6.4. In-memory I/O
6.5. Network I/O
6.5.1. Network requests over HTTP
6.5.2. REST API
6.5.3. RPC over HTTP
6.5.4. gPRC
6.5.5. Apache arrow
Summary
Chapter 7: Working with cloud
7.1. Utilizing cloud computing
7.1.1. Comparison between cloud and supercomputers
7.1.2. Designing a workflow
7.1.3. Computing resources
7.1.4. Communications among cloud services
7.1.5. Function-as-a-Service
7.2. Distributed job executors
7.2.1. Celery
7.2.2. Dask
7.2.3. Ray
Summary
Chapter 8: Foreign language interfaces
8.1. Inter-process communication interfaces
8.2. C/C++ interfaces
8.2.1. Data types and type conversion
8.2.2. Cython
8.2.3. pybind11
8.2.4. Compiling C++ code into a Python module
8.3. Foreign function interfaces
8.3.1. Ctypes
8.3.2. CFFI
8.3.3. Memory leaks
8.3.4. Duplicated function names in extensions
8.4. Fortran interfaces
8.4.1. ctypes for Fortran
8.4.2. f2py compiler
8.5. Rust interfaces
Summary
Chapter 9: Program performance optimization
9.1. Principles for performance optimization
9.1.1. Cost comparison between Python and C/C++ operations
9.1.2. Hardware and operating system overhead
9.1.3. Latency and throughput
9.1.4. Strategies for optimizing latency and throughput
9.1.5. Computation bound and I/O bound
9.1.6. Instruction level parallelism
9.2. Profiling
9.2.1. Benchmark tests
9.2.2. Python built-in profiling tools
9.2.3. Line profiler
9.2.4. Sampling profiler
9.2.5. Perf
9.3. Python level optimization
9.3.1. The dis module
9.3.2. Performance-friendly Python code
9.3.3. Utilizing tensor operations
9.3.4. Optimizing tensor indexing efficiency
9.4. Compiling Python code
9.4.1. Numba
9.4.2. Cython
9.4.3. Pythran
9.4.4. Comparison of Cython, Pythran, and Numba
9.5. Optimization with compiled languages
9.5.1. GCC compiler
9.6. Optimization for I/O
9.6.1. Storage layout
9.6.2. Compressing data
9.6.3. Overlapping computation and I/O
9.7. Precomputation and memoization
9.7.1. LRU cache
9.7.2. Functional programming
9.7.3. Dynamic programming
9.8. Optimization with lazy evaluation
Summary
Chapter 10: Parallel computation
10.1. Multithreading
10.1.1. The threading module
10.1.2. ThreadPoolExecutor
10.2. Lock and thread synchronization
10.2.1. Mutex
10.2.2. Event and condition variable
10.3. Producer-consumer model
10.4. Pipeline executor
10.5. Asynchronous program
10.5.1. Similarity between asynchronous programming and lazy evaluation
10.5.2. Asynchronous program with coroutines and asyncio
10.6. Multiprocessing
10.6.1. Data communication in multiprocessing.Process
10.6.2. Data communication in ProcessPoolExecutor
10.6.3. Locks in the multiprocessing program
10.7. Inter-process communication
10.8. OpenMP
10.9. MPI
10.9.1. Naming conventions in MPI4Py
10.9.2. SPMD MPI programs in Python
10.9.3. Integrating MPI into serial program
10.9.4. Shared memory
Summary
Chapter 11: GPU programming
11.1. Configuring GPU runtime environment
11.2. Architecture-independent optimization
11.2.1. CuPy
11.2.2. PyTorch
11.2.3. JAX
11.3. Architecture-aware optimization
11.3.1. Pinned memory
11.3.2. CUDA stream
11.4. Custom GPU kernels in Python
11.4.1. Kernel in CUDA code
11.4.2. Numba CUDA JIT
11.4.3. CuPy custom kernel
11.4.4. Integrating CUDA kernels using ctypes
Summary
Chapter 12: Integral evaluation
12.1. Analytical integral evaluation for Gaussian type orbitals
12.1.1. Data structure for GTO basis
12.1.2. Basic types for analytical GTO integrals
12.1.3. Recurrence relations and the dynamic programming implementation
12.1.4. Converting recursion to iteration
12.1.5. Eliminating branches and adjusting iterations
12.1.6. Optimization with Numba JIT compilation
12.1.7. Optimization with Cython compilation
12.1.8. Optimization with meta-programming techniques
12.1.9. Performance benchmark
12.1.10. Other performance factors
12.2. Numerical integration
12.2.1. Gaussian quadrature
12.2.2. Generating quadrature roots and weights
12.2.3. Approximating quadratures
12.2.4. Numerical integration with Fourier transform
12.3. Integral transformation
Summary
Chapter 13: Mean-field methods
13.1. The self-consistency iteration program
13.2. Coulomb and exchange matrices
13.2.1. Integral screening for Coulomb and exchange matrices
13.2.2. J-engine
13.3. Integrals for exchange-correlation functionals
13.4. DIIS
13.5. Design of Python classes for mean-field methods
13.5.1. New mean-field classes via class inheritances
13.5.2. Mean-field classes via the visitor pattern
13.5.3. Dynamic patches to mean-field objects
13.5.4. Dynamic mean-field classes
13.6. Speeding up mean-field calculations
13.6.1. Cholesky decomposition and resolution of identity methods
13.6.2. Improving convergence using DIIS
13.6.3. Improving initial guess
Summary
Chapter 14: Post Hartree-Fock I: full configuration interaction
14.1. Theory of full configuration interaction
14.2. The string representation
14.3. Davidson diagonalization
14.4. Direct CI algorithm
14.5. Optimizing tensor contractions
14.6. Optimization for memory efficiency
14.7. Parallel computation
Summary
Chapter 15: Post Hartree-Fock II: coupled cluster
15.1. Coupled cluster theory
15.2. CCD program
15.3. Optimizing data transferring
15.4. Just-in-time compilation for I/O readahead
15.5. Symbolic programming for coupled cluster theory
15.5.1. Defining fundamental symbolic elements
15.5.2. Implementing computation rules
15.5.3. Evaluating expressions symbolically
15.5.4. Generating CCD equations symbolically
15.5.5. Applying permutation symmetry to CCD tensors
Summary
Chapter 16: Molecular properties
16.1. Molecular properties for single-particle operators
16.2. Analytical nuclear gradients
16.2.1. Basic equations of Hartree-Fock derivatives
16.2.2. Derivatives of integrals
16.2.3. Nuclear gradients for RHF energy
16.2.4. Nuclear gradients with finite difference
16.2.5. First order molecular orbitals
16.3. Krylov subspace linear equation solver
16.4. Nuclear gradients with automatic differentiation
16.4.1. Differentiable Python class
16.4.2. Implicit differentiation
16.4.3. Higher-order derivatives
16.4.4. Conversion between JVP and VJP
Summary
Index

Life Sciences

Physical Sciences & Engineering

Social Sciences & Humanities

Health