Saltar al contenido ForoHackers

Computational physics: start here

Física computacional
#745 · Ophuscado · · Denunciar

Computational physics solves the equations of physics numerically when pencil and paper cannot: orbits of many bodies, fluids, fields, quantum systems, statistical ensembles. It trains you to turn a law into an algorithm, to know what the algorithm gets wrong, and to check results against limits you can calculate by hand.

Key concepts

  • Integrating equations of motion: Euler, Runge-Kutta and symplectic methods, which keep energy from drifting in long orbits.
  • Fields and grids: finite differences and finite volumes for waves, diffusion and fluids, and the stability limit that ties time step to grid size.
  • Monte Carlo: random sampling for statistical systems and integrals, such as the Ising model.
  • Conservation laws as tests: energy, momentum and charge should stay constant, so tracking them finds bugs.
  • Scaling and units: nondimensionalising a problem keeps numbers sane and reveals the controlling parameters.

Vocabulary

  • Time step: the interval by which a simulation advances in each iteration.
  • Boundary conditions: what a field does at the edge of the domain.
  • Chaos: sensitive dependence on initial conditions, which limits long-term prediction.
  • Finite-size effect: an error caused by simulating a small piece of a large system.

What to study, in order

  1. Mechanics and calculus, then a language with fast array maths.
  2. Simulate a pendulum with several integrators, and plot energy against time for each.
  3. Solve the diffusion and wave equations on a grid, and find the time step where the scheme goes unstable.
  4. Write a Monte Carlo simulation of the Ising model, and look for the phase transition.
  5. Simulate a many-body problem, and check conservation laws and a known limit.

Related tags

#numerical-analysis #simulation #physics-simulation #computational-chemistry #hpc

Inicia sesión o únete para responder.