The Godunov Method
Godunov's method advances conservation laws by solving a Riemann problem at each cell interface to compute upwind-consistent fluxes.
Riemann problems at every interface
Godunov's method treats the solution as piecewise constant across cells and recognizes that each cell interface poses a Riemann problem: two constant states separated by a discontinuity. Solving that local Riemann problem gives the exact wave structure at the interface, from which the numerical flux is computed. This builds the correct upwind direction and wave speeds into the scheme automatically, without ad hoc dissipation.
The genius of the approach is that the physics of wave propagation, encoded in the Riemann solution, determines the numerical flux. It is inherently conservative and captures shocks at the right speeds.
First order and its extensions
The original Godunov scheme uses piecewise-constant data and is first-order accurate, robust but diffusive. Higher-order Godunov-type methods (MUSCL, PPM) replace the constant states with limited linear or parabolic reconstructions, so the Riemann problems are posed between more accurate interface states. Slope limiting keeps these reconstructions monotone near discontinuities.
Exact versus approximate solvers
Solving the exact Riemann problem can be expensive, especially for systems like MHD with many wave families. In practice approximate Riemann solvers (Roe, HLL, HLLC, HLLD for MHD) provide the interface flux at a fraction of the cost while retaining the essential upwind character. The choice trades accuracy and robustness against computational effort.
- Piecewise-constant states pose a Riemann problem per interface
- Wave structure sets the upwind numerical flux
- MUSCL and PPM raise the order via limited reconstruction
- Approximate Riemann solvers replace exact solves for efficiency
Godunov-type finite-volume schemes are the standard for compressible and magnetohydrodynamic simulation in fusion research, valued for capturing fast shocks and complex wave interactions conservatively and robustly.