Feynman Path Integral: Sum Over Histories

Every possible path counts, not just the classical one

Richard Feynman's path-integral formulation, developed in the 1940s, reframes quantum mechanics around a radical idea: to find the probability amplitude for a particle to travel from point A to point B, sum a small complex number -- a phasor -- over literally every conceivable path connecting A to B, not just the one smooth trajectory Newtonian mechanics would predict. Each path contributes a phasor of equal length but a phase set by its classical action S, the time-integral of the Lagrangian along that path:

amplitude(A to B) = sum over all paths of  exp(i * S[path] / hbar)

Squaring the magnitude of that total summed amplitude gives the actual observable probability of the particle arriving at B. Wildly impractical-looking paths -- ones that zigzag, loop backward in space, even briefly move faster than light in the non-relativistic toy version -- all get included in the sum with exactly equal weight; what varies from path to path is only the phase angle each one contributes.

Why the classical path still wins, most of the time

For everyday, macroscopic objects, the action S along nearby paths changes enormously from one path to a slightly different neighbouring one, because S is measured in units of hbar times a huge number for anything macroscopic. That means the phasors from neighbouring non-classical paths point in wildly different, essentially random directions and cancel each other out almost completely when summed -- destructive interference. The one region where this cancellation fails is right around the path that makes the action stationary (a minimum, maximum, or saddle point under small variations) -- the classical trajectory predicted by Newton's laws or, more generally, the principle of least action. Neighbouring paths near that stationary path have action values that change only slowly, so their phases stay nearly aligned and add up constructively, which is precisely why the classical path dominates the sum and macroscopic objects appear to follow one single, definite trajectory.

Where quantum weirdness survives: small action, small mass

For a particle light enough or a distance short enough that competing paths differ in action by only a fraction of hbar, the destructive-interference argument above breaks down -- neighbouring paths keep roughly the same phase over a much wider spread, so many distinct paths contribute constructively rather than just one. This is exactly the regime where genuinely quantum behaviour -- interference, diffraction, tunnelling -- shows up, and the path-integral picture explains it as literally more of the sum surviving cancellation, rather than needing a separate rule bolted onto classical mechanics.

The double slit, reframed as two path bundles

The path-integral view gives an unusually direct explanation for double-slit interference: the sum over paths naturally splits into a bundle of paths passing through slit one and a bundle passing through slit two, and the interference pattern on the screen is just the ordinary result of adding those two bundles' amplitudes together, phase and all, before squaring. Block one slit and you remove an entire bundle of paths from the sum, which is exactly why the interference pattern disappears and a plain single-slit diffraction pattern remains -- no separate wave-particle-duality postulate is needed, it falls straight out of summing over histories.

From an elegant idea to a working numerical method

Beyond its conceptual clarity, the path integral became a genuine computational tool. In quantum field theory it is the natural language for computing scattering amplitudes via Feynman diagrams (each diagram is a organized way of enumerating a class of contributing paths, now including particle creation and annihilation). In condensed matter and statistical mechanics, path-integral Monte Carlo methods sample the dominant paths numerically -- since summing every conceivable path exactly is impossible, but a computer can sample the ones that matter most, in a Wick-rotated version of the sum where the phases become ordinary probabilities -- to compute properties of interacting many-body quantum systems that have no closed-form solution.

Frequently asked questions

Try it live

Everything above runs in your browser — open Feynman Path Integral and change the parameters while it is running. Nothing is installed, nothing is uploaded, the whole model lives in one tab.