Computational Quantum Mechanics Roadmap

Computational quantum mechanics is the route for readers who want to turn Hamiltonians, wavefunctions, operators, and density matrices into reproducible calculations. The central habit is not just getting a number; it is knowing what was discretized, what error was introduced, what convergence test was passed, and what physical limit was checked.

This roadmap focuses on durable numerical ideas. Specific software tools can change quickly, but conditioning, convergence, validation, and benchmark problems remain central.

For a named material or material-model claim, Computational Quantum Matter turns these reusable capabilities into a routed approximation, validation, probe-comparison, and evidence workflow; this roadmap retains the capability-building sequence.

Follow a complete calculation

The two paths below cross subject boundaries in the order the calculation needs. Each step links to its existing treatment and gives a checkpoint. Starting a path adds separate Next in this path links below the article; the ordinary Previous/Next links continue to follow that article’s chapter. You can leave the path at any point or return here to choose a different step.

From the harmonic oscillator to a converged spectrum

Compute low-energy oscillator states and separate grid-spacing error from the effect of a finite spatial domain.

Required background. Matrix Diagonalization — Interpret the eigenvalues and normalized eigenvectors of a Hermitian matrix.

1. Quantum Harmonic Oscillator

   Identify the oscillator Hamiltonian, natural units, and exact energy spectrum.

   Checkpoint. Write the expected dimensionless energy levels and explain which physical scale restores the units.

2. Finite Difference Methods

   Translate a second derivative into a finite matrix while specifying the boundary treatment.

   Checkpoint. Construct the three-point second derivative and state the error expected when the grid spacing is halved.

3. Convergence Tests

   Plan independent resolution and domain-size studies before interpreting a computed spectrum.

   Checkpoint. Explain why refining the grid inside a fixed small box cannot eliminate boundary error.

4. Harmonic Oscillator Spectrum

   Run the spectrum experiment, inspect its checks, and extend a convergence study.

   Checkpoint. Compare the lowest energies with their analytic values and report both spacing and domain-size evidence for the accuracy claimed.

Start this path →

From a driven two-level system to a transition probability

Connect the Landau–Zener prediction to finite-duration numerical evolution and distinguish propagation error from the finite observation window.

Required background. Matrix Diagonalization — Find instantaneous eigenstates of a two-by-two Hermitian Hamiltonian.

1. Two-Level Systems

   Relate a two-component state, a Hamiltonian, and probabilities in a specified basis.

   Checkpoint. State which basis defines the initial state and which measurement defines the final transition probability.

2. Time-Dependent Hamiltonians

   Track the difference between instantaneous eigenstates and the state that actually evolves.

   Checkpoint. Explain why diagonalizing the Hamiltonian separately at each instant does not itself solve the time-evolution problem.

3. Landau-Zener Transition

   Identify the avoided crossing, sweep rate, coupling convention, and asymptotic transition prediction.

   Checkpoint. Check the adiabatic and sudden limits using the same state-label convention as the formula.

4. Time-Stepping Methods

   Choose a propagator and distinguish norm conservation from accuracy of amplitudes and phases.

   Checkpoint. Explain why a method that preserves norm can still give the wrong transition probability.

5. Landau–Zener Simulation

   Run two propagators and vary both the integration step and the time window.

   Checkpoint. Report norm conservation, agreement between propagators, and separate step-size and window-size checks before comparing with the asymptotic formula.

Start this path →

Prerequisites

You should know complex vectors and matrices, eigenvalue problems, ordinary and partial differential equations, Fourier transforms, and enough programming to read arrays, loops, tests, and plots. Use Math Needed for Computational QM as the preparation map.

Phase 1: Numerical Linear Algebra

Begin with finite-dimensional approximations:

1. floating-point arithmetic,
2. conditioning and stability,
3. dense matrix diagonalization,
4. sparse matrices,
5. sparse eigensolvers,
6. matrix exponentials,
7. orthogonality and normalization checks.

Use Floating-Point Arithmetic, Conditioning and Stability, Matrix Diagonalization, Sparse Matrices, and Sparse Eigensolvers.

Milestone: you can diagonalize a finite Hamiltonian, check orthonormality, sort eigenpairs, and identify a numerically suspicious result.

Phase 2: Discretizing Wave Mechanics

Learn how continuous wave mechanics becomes a finite computation:

1. grids and basis truncations,
2. finite differences,
3. spectral methods,
4. boundary conditions,
5. numerical quadrature,
6. normalization on grids,
7. convergence under refinement.

Use Discretization, Finite-Difference Methods, Spectral Methods, Numerical Quadrature, and Convergence Tests.

Milestone: you can compute the infinite-square-well or harmonic-oscillator spectrum and demonstrate convergence to the analytic answer.

Phase 3: Time Evolution

Time-dependent calculations require preserving the structure of quantum mechanics:

1. ordinary differential equation solvers,
2. unitary time stepping,
3. matrix exponentials,
4. split-operator methods,
5. Crank-Nicolson-type ideas,
6. time-dependent Hamiltonians,
7. norm and energy checks.

Use ODE Solvers, Time-Stepping Methods, Matrix Exponentials Numerically, and Error Estimates.

Milestone: you can propagate a wave packet and tell whether loss of norm is physical, numerical, or caused by a declared absorbing boundary.

Phase 4: Fourier and PDE Tools

Many quantum simulations move between coordinate and momentum space. Study:

1. Fourier transforms and FFT conventions,
2. spectral differentiation,
3. aliasing and resolution,
4. partial differential equation solvers,
5. boundary-condition artifacts,
6. wave-packet diagnostics.

Use Fast Fourier Transform, Fourier Transform Conventions, PDE Solvers, and Wave Packets.

Milestone: you can explain how grid spacing, box size, and momentum resolution constrain the calculation.

Phase 5: Model Benchmarks

Never trust a new computational method first on a system with no known answer. Build a benchmark ladder:

1. two-level systems,
2. harmonic oscillator,
3. finite and infinite square wells,
4. free-particle wave packets,
5. potential step and barrier scattering,
6. perturbative checks,
7. symmetry checks,
8. limiting cases.

Use Benchmark Problems as the validation map.

Computational AMO and Quantum Chemistry extends this ladder to hydrogen and helium, molecular basis and electronic-structure calculations, optical Bloch and Jaynes–Cummings dynamics, laser cooling, and domain-specific reproducibility records.

Milestone: every new code path has at least one analytic, convergence, or conservation check.

Phase 6: Many-Body and Matrix Growth

The Hilbert space grows quickly. Study:

1. tensor-product basis construction,
2. tensor-product ordering,
3. exact diagonalization,
4. sparse Hamiltonian assembly,
5. symmetries and block diagonalization,
6. Fock-space bases,
7. truncation and basis selection.

The existing Core pages on Tensor Products and Density Operators provide the formal language. Scaling of Hilbert Space owns the basis-size counts and asymptotics, while Tensor Networks Preview explains graph factorizations and cross-family capacity. Matrix Product States Preview develops canonical bonds, transfer spectra, and the physical basis of DMRG in one dimension. Enter Computational Many-Body QM to select the physics-facing numerical branch. The Computational Many-Body Overview connects those structures to detailed method selection, validation, and finite-size inference. Exact Diagonalization Preview owns the finite basis-to-eigensystem workflow; the full Computational QM volume retains reusable implementations.

Milestone: you can estimate Hilbert-space dimension before building a matrix and explain which symmetries reduce the computation.

Phase 7: Reproducibility and Reporting

A computational result should include:

• Hamiltonian and parameter definitions,
• units and constants,
• basis or grid,
• boundary conditions,
• solver method,
• tolerances,
• convergence tests,
• benchmark comparisons,
• code or notebook environment,
• random seeds when relevant,
• known failure modes.

Milestone: another reader can reproduce the result or identify precisely what information is missing.

Common Pitfalls

• Trusting a plot without convergence tests.
• Confusing discretization error with physical effects.
• Ignoring boundary reflections.
• Using dense algorithms on sparse problems without estimating size.
• Forgetting normalization factors on grids.
• Comparing to analytic formulas with different units or Fourier conventions.
• Reporting hardware or software performance without versions.

References

• L. N. Trefethen and D. Bau III, Numerical Linear Algebra, SIAM, 1997.
• Y. Saad, Numerical Methods for Large Eigenvalue Problems, 2nd ed., SIAM, 2011.
• J. W. Thomas, Numerical Partial Differential Equations: Finite Difference Methods, Springer, 1995.
• J. M. Thijssen, Computational Physics, 2nd ed., Cambridge University Press, 2007.
• W. H. Press et al., Numerical Recipes, 3rd ed., Cambridge University Press, 2007.