Computational Quantum Mechanics

Computational quantum mechanics turns a specified quantum model into spectra, evolving states, correlation functions, and measurable predictions. Its central task is to control the difference between the physical question, the finite representation, and the number returned by an algorithm. This volume develops numerical methods together with the benchmarks, convergence studies, and failure diagnostics needed to interpret their output.

The methods range from single-particle differential equations to interacting many-body systems and stochastic dynamics. Conceptual definitions and physical models remain with their subject volumes; these pages explain how to represent them, solve the resulting problems, and quantify numerical limitations. Every prediction depends on the model and parameter regime used to compute it.

Start with a controlled calculation

A useful first project is the one-dimensional particle in a box. Construct the finite-difference Hamiltonian, normalize its eigenvectors with the correct grid measure, compare low energies with the exact spectrum, and refine the grid. This small problem reveals boundary, units, discretization, and solver errors without hiding them behind an unfamiliar model.

Follow representations as arrays, dimensionless variables, finite differences, matrix diagonalization, and convergence studies. Then propagate a state and compare its phase and observables, as well as its norm, against a known solution.

Readers should be comfortable with linear algebra, quantum states and operators, and ordinary differential equations. Core Formalism, Wave Mechanics, and Mathematical Toolkit supply that background. Programming examples use their stated array and unit conventions; translating a formula into code requires preserving those conventions.

Chapters and their questions

1. Representations and Numerical Reliability — How do basis choices, floating-point arithmetic, symbolic algebra, and error budgets determine what a result means? Includes reproducibility, convergence reports, and parallel computation.
2. Spatial Discretization and Boundaries — How does a differential operator become a grid or basis problem with the right domain and measure? Covers finite differences, Numerov, finite elements, spectral methods, DVR, mappings, absorbers, and relativistic grids.
3. Spectral and Variational Computation — How are eigenpairs and variational approximations obtained and checked? Includes Krylov methods, molecular integrals, mean-field iterations, and correlated wavefunction approximations.
4. Time Propagation — Which algorithms resolve quantum phases and observables over the required time interval? Covers matrix exponentials, splitting, Crank–Nicolson, Magnus and Chebyshev methods, rotating frames, and wavepackets.
5. Scattering and Resonance Computation — How are amplitudes, phase shifts, transmission, and complex resonances extracted without confusing them with finite-box artifacts?
6. Density Matrices and Stochastic Evolution — How are mixed states, generators, trajectories, and noise represented while tracking trace, positivity, sampling error, and physical time scales?
7. Many-Body Simulation — How do occupation bases, symmetry sectors, spin models, and fermionic signs enter exact finite-system calculations and circuit simulation?
8. Tensor Networks — When can entanglement structure make a many-body state compressible, and how are truncation and contraction errors controlled?
9. Quantum Monte Carlo — How do sampling, imaginary-time projection, and path integrals estimate quantum quantities? Covers autocorrelation, the sign problem, and the distinct biases of practical methods.
10. Observables and Quantum Geometry — How are reduced states, entanglement, correlations, response functions, Berry phases, and topological invariants extracted with consistent conventions?

Choose a route by the requested observable

For a few low-lying bound states, build a converged representation, then use a suitable sparse eigensolver or variational method. An accurate energy may coexist with an inaccurate eigenvector near a degeneracy; compare the relevant subspace and observable matrix elements.

For real-time dynamics, choose a propagation method suited to the Hamiltonian structure and separately refine spatial or basis resolution and time step. Norm conservation is a useful check, but a unitary method can still accumulate large phase error. Long-time correlation spectra also depend on the finite observation window.

For an interacting lattice model, begin with exact diagonalization on a small instance. It establishes basis ordering, signs, and symmetry conventions before introducing tensor compression or statistical sampling. A larger approximate calculation should reproduce that overlapping regime before its extrapolations are trusted.

For open systems, distinguish deterministic density-matrix evolution from stochastic unravellings of the same generator. Agreement of trajectory averages requires both controlled time integration and enough independent samples. A numerical absorber used to emulate spatial escape is not, by itself, a complete model of a physical environment.

Separate the sources of uncertainty

A useful calculation makes four comparisons explicit: the model against its physical assumptions; the representation against a continuum or basis limit; the solver against the finite equations; and the estimator against its statistical and finite-time limits. These comparisons are logically different even when one code produces all four outputs.

Use analytic benchmarks to expose signs and normalization errors. Use error budgets to combine controlled approximations without treating systematic biases as independent random noise. Use convergence reports to retain the actual refinement data and settings behind a claimed accuracy.

A small residual, an integer-valued invariant, or a smooth plot can each coexist with an incorrect result. The method pages therefore pair defining equations with examples, diagnostic failures, and solved exercises. Read those checks as part of the algorithm.

References

• Golub, Gene H., and Charles F. Van Loan. Matrix Computations. 4th ed. Johns Hopkins University Press, 2013.
• Higham, Nicholas J. Accuracy and Stability of Numerical Algorithms. 2nd ed. SIAM, 2002.
• Tannor, David J. Introduction to Quantum Mechanics: A Time-Dependent Perspective. University Science Books, 2007.
• Thijssen, Jos. Computational Physics. 2nd ed. Cambridge University Press, 2007.