All Projects Senior Thesis: Finite-Difference Quantum Eigenvalue Problems
Computational Physics

Senior Thesis: Finite-Difference Quantum Eigenvalue Problems

Numerical solutions for 1D quantum mechanical eigenvalue problems using finite-difference methods, solving the time-independent Schrödinger equation with sparse Hamiltonian matrices.

PythonNumPySciPyMatplotlibJupyterLaTeXGit
GitHub Repo

Problem

Solving quantum eigenvalue problems analytically is only possible for a few idealized potentials. For realistic systems, efficient and accurate numerical methods are needed — but the trade-offs between accuracy, grid resolution, and computational cost aren't always clear from theory alone.

Solution

Implemented 2nd and 4th-order finite-difference discretizations of the 1D Schrödinger equation, assembled sparse Hamiltonian matrices, and solved for eigenvalues/eigenstates using ARPACK. Ran convergence studies comparing stencil orders, validated against analytical solutions, and produced publication-ready figures.


What I Did

  • Built core numerical modules for grid construction and Hamiltonian assembly using 2nd and 4th-order finite-difference stencils, leveraging sparse matrices for efficiency.

  • Implemented eigenvalue analysis tools with analytical reference solutions, numerical solvers (ARPACK via scipy.sparse.linalg), error analysis, and convergence helpers.

  • Ran convergence studies comparing 2nd vs 4th-order accuracy on infinite square well (ISW) and quantum simple harmonic oscillator (QSHO) systems, verifying expected convergence rates.

  • Validated against analytical solutions for the infinite square well and harmonic oscillator, confirming eigenvalue accuracy and eigenfunction node counts.

  • Generated publication-ready figures including convergence curves, eigenstate plots, and analytical-vs-numerical comparisons for the written thesis.

  • Wrote a full senior thesis document presenting the mathematical framework, numerical methods, results, and analysis.

Key Results

  • 4th-order stencils achieved significantly better accuracy per grid point compared to 2nd-order, with clear O(h4)\mathcal{O}(h^4) convergence confirmed on log-log plots.
  • Eigenvalues matched analytical solutions to high precision at moderate grid sizes.
  • Sparse matrix methods kept computation efficient even at fine grid resolutions.

Thesis Document

The full written thesis is available in the GitHub repository.