numerical simulation

History

The first numerical experiments were done by hand back in the 17th century, but it really began with the advent of computers in the 1940s for calculating nuclear explosions. In astrophysics, star cluster modeling began in the 1960s.

How it works

We take a complex equation describing, for example, how planets move, and break it down into simple time steps: where will the planet be in a second? And in another second? Repeating this millions of times gives the path. The smaller the step, the more accurate, but more computing power is needed.

History

The first numerical experiments were done by hand back in the 17th century, but it really began with the advent of computers in the 1940s for calculating nuclear explosions. In astrophysics, star cluster modeling began in the 1960s.

How it works

We take a complex equation describing, for example, how planets move, and break it down into simple time steps: where will the planet be in a second? And in another second? Repeating this millions of times gives the path. The smaller the step, the more accurate, but more computing power is needed.

History

The origins go back to manual numerical orbit calculations performed by 19th-century astronomers. Decisive impetus came from research in turbulence and nuclear physics in the mid-20th century (work by von Neumann and Ulam on the Monte Carlo method). In astrophysics, pioneering were the works of Hénon and Aarseth on modeling star clusters in the 1960s. Henri Poincaré laid the foundations of dynamical systems theory, crucial for understanding chaos in numerical calculations. With the growth of supercomputers in the 1980s–2000s, full-scale cosmological simulations emerged (Millennium, EAGLE, Illustris).

How it works

It is based on discretizing partial differential equations. For example, in hydrodynamics, the conservation laws for mass, momentum, and energy are written as ∂U/∂t + ∇·F = S, where U is the vector of conservative variables, F are fluxes, and S are sources. The heart of the algorithm is the Courant–Friedrichs–Lewy (CFL) condition ensuring stability: Courant number C = c·Δt/Δx ≤ 1, where c is the wave speed, Δt and Δx are time and space steps. Modern codes use adaptive mesh refinement and parallelization over thousands of cores. For gravity problems, fast Fourier transforms or hierarchical tree methods are used.

Open problems & fun facts

Insufficient spatial resolution to describe turbulence and star formation; Parameterization of subgrid processes (star formation, feedback); Beam instabilities and accumulation of round-off errors; Convergence and code verification in the strongly nonlinear regime

💡 The "GigaEris" simulation used over 250 billion particles for a single galaxy cluster, requiring about 1 petabyte of disk space.