Numerical Methods in Computational Physics
Numerical methods give physics the algorithmic machinery to solve equations that have no closed-form answer, turning differential equations, integrals and matrix problems into finite arithmetic that a computer can execute with controlled error.
Definition
Numerical methods in computational physics are the discretization and approximation algorithms used to convert continuous physical models into finite computations, with attention to truncation error, numerical stability and conservation of physical invariants.
Scope
This area covers the core numerical toolkit on which computational physics is built: integrators for ordinary and partial differential equations, methods for large linear-algebra and eigenvalue problems arising from discretized physics, and root finding and optimization for nonlinear physical conditions. It emphasizes accuracy, stability and the physical interpretation of discretization rather than abstract numerical analysis for its own sake.
Sub-topics
Core questions
- How are continuous differential equations of physics turned into stable, accurate finite-difference or finite-element schemes?
- What controls the trade-off between step size, truncation error and stability in an integrator?
- How are the large sparse linear systems and eigenproblems from discretized physics solved efficiently?
- How do numerical schemes preserve physical invariants such as energy, momentum or symplectic structure?
Key theories
- Discretization and truncation error
- Replacing derivatives and integrals by finite-difference or quadrature approximations introduces truncation error that scales as a power of the step size, setting the order of accuracy of a scheme.
- Numerical stability
- A scheme is stable if errors do not grow without bound as it is iterated; stability conditions such as the Courant-Friedrichs-Lewy criterion constrain the admissible time and space steps for evolution equations.
- Sparse linear algebra and eigenproblems
- Discretized physical operators yield large sparse matrices whose linear systems and eigenvalues are found with iterative Krylov, Lanczos and conjugate-gradient methods rather than dense factorization.
Clinical relevance
These methods underpin essentially all of quantitative physics done on computers: orbital and trajectory integration, electromagnetic and quantum field solvers, fluid and heat-transport simulation, and the solution of the matrix problems behind electronic structure and lattice models.
History
Numerical solution of physical equations dates to hand computation in celestial mechanics and ballistics, was transformed by the electronic computers built for wartime physics in the 1940s, and matured into a standard methodology through reference works such as Numerical Recipes and the rise of computational physics curricula in the late twentieth century.
Key figures
- John von Neumann
- William H. Press
- Cornelius Lanczos
- Rubin H. Landau
Related topics
Seminal works
- press2007
- landau2015
Frequently asked questions
- Why not just use a very small step size to get high accuracy?
- Shrinking the step reduces truncation error but increases the number of steps and the accumulation of round-off error, and for some explicit schemes a too-large step causes instability rather than mere inaccuracy. Good methods balance order of accuracy, stability and cost rather than relying on brute-force small steps.
- How is numerical physics different from numerical analysis?
- Numerical analysis studies algorithms and their error bounds in general, while numerical methods in physics select and adapt those algorithms to physical equations, prioritizing conservation laws, symmetries and physical interpretability of the discretized model.