An approximation and alternative-formalism concept defining methods for computing quantum predictions when exact solutions are impractical. It governs controlled expansions, action-based formulations, and phase-space representations that support analytic and numerical work. It does not ensure accuracy outside its regime of validity and requires explicit error assessment or convergence checks. It enables tractable estimates of spectra, transition rates, and dynamical behavior across a wide range of models. The concept is generally stable, though improved algorithms and convergence techniques evolve over time.

Principle

Replace the continuum functional integral over histories by a finite-dimensional sum or integral using time-slicing, lattice discretization, or stochastic sampling, then control the discretization and sampling errors to recover continuum quantum results.

Demonstration

Compute the propagator of a nonrelativistic particle in a potential by dividing the time interval into N slices, approximating the action for each slice, sampling paths with importance sampling (Metropolis or Langevin Monte Carlo), and extrapolating observables to N→∞ to estimate the continuum propagator.

Misapplication

Treating coarse discretization results as exact without convergence checks, using a sampling algorithm that ignores the oscillatory phase (leading to uncontrolled sign/phase problems), or applying Euclidean Monte Carlo methods without justifying analytic continuation back to real time.

Consequence

When properly implemented with controlled extrapolation and error estimation, numerical path integrals provide nonperturbative access to quantum observables, thermal partition functions, and semiclassical instanton contributions that are difficult to obtain otherwise.

Reversal

Instead of sampling paths, one could compute the same amplitudes using operator methods (Schrödinger or Heisenberg picture) or exact diagonalization; the reversal emphasizes algebraic spectral techniques rather than stochastic integration over histories.

Boundary

Applies to systems where the action is known and discretizable; excludes methods that rely purely on analytic resummation, closed-form path integrals for quadratic actions, or classical trajectory integration without quantum interference.

Semantic Tension

Competes with operator-based numerical approaches (matrix diagonalization, tensor-network methods) when finite Hilbert-space truncation is feasible; the tension lies between summing histories versus truncating state space.

Synthesis

Path integral numerical simulation is the practical substitution of the continuum sum-over-histories by controlled discrete sampling or summation procedures that recover quantum amplitudes and nonperturbative effects through convergence and error control.