the sum over histories

for two hundred years physicists fought about whether light was a wave or a particle. newton said particle. huygens said wave. young proved wave with his double-slit. then einstein proved particle with the photoelectric effect. then de broglie said matter is also a wave. by 1925 nobody knew what anything was.

the copenhagen answer: it's both, but you can only observe one aspect at a time. wave when unobserved, particle when measured. complementarity. bohr's polite way of saying shut up and calculate.

this worked. it predicted experiments perfectly. but it was unsatisfying. the math gave answers without explaining why. why does observation collapse anything? why does nature need a special process called measurement?

richard feynman, in his 1942 princeton thesis and later in his 1948 paper in reviews of modern physics, proposed something different. he didn't pick sides. he didn't add a new interpretation. he changed the question entirely.

the old problem

the double-slit experiment is the heart of it.

shine light at two slits. if light is a wave, it interferes with itself. bright bands where crest meets crest. dark bands where crest meets trough. this is exactly what you see.

but detect which slit each photon passes through and the interference pattern vanishes. the light acts like particles hitting the screen in two clumps. it's as if knowing where it went changed what it was.

the copenhagen story: the photon exists as a probability wave until measurement forces it to choose. the wave function collapses. bohr said the wave and particle descriptions are complementary — both true, but mutually exclusive in any single experiment.

einstein hated this. does the moon exist only when i look at it? schrödinger mocked it with his cat. de broglie tried pilot waves. bohm tried hidden variables. everyone wanted a deeper layer.

feynman's move

feynman asked: what if the photon doesn't choose a path at all? what if it takes every possible path from source to screen simultaneously, and the final result is the sum of all of them?

this is the path integral formulation of quantum mechanics.

instead of calculating a single trajectory — newton's style, or even schrödinger's wave equation — you consider every conceivable way a particle could get from point a to point b. straight lines. loops. paths that double back. paths that visit andromeda and return. every mathematically possible trajectory through space and time.

each path contributes a number: a complex phase, like a little arrow rotating in a 2d plane. the phase depends on the action of that path — roughly, how much energy it would cost to travel that route, integrated over time.

the magic: for most paths, the phases point in random directions. they cancel each other out. the wildly crooked paths interfere destructively and contribute essentially nothing.

but near the classical path — the one that minimizes the action, the one newton's laws would predict — nearby paths have nearly identical phases. they point in roughly the same direction. they add up. constructively.

so the classical path emerges not because nature prefers it, but because it's the only route where the surrounding paths don't cancel each other out.

"the electron does anything it likes. it goes in any direction at any speed, forward or backward in time, however it likes, and then you add up the amplitudes and it gives you the wave function."

that's feynman, more or less. he was casual about it in a way that made other physicists nervous.

the amplitude to go from point $a$ at time $t_a$ to point $b$ at time $t_b$ is the integral over every path connecting them:

$K(b, a) = \int_a^b e^{iS/\hbar} \, \mathcal{D}[q(t)]$

$S$ is the classical action — the integral of the lagrangian along each path. $\hbar$ is the reduced planck constant. the exponential $e^{iS/\hbar}$ is the phase factor. when $S$ changes by much more than $\hbar$, the phase spins wildly and neighboring paths cancel. when $S$ is stationary — at the classical path — the phase changes slowly, and nearby paths reinforce.

in the limit where $\hbar \to 0$, all non-classical paths cancel completely. you recover classical mechanics. the classical world is what survives when you let all the quantum weirdness interfere itself to death.

the classical limit

this is maybe the prettiest part. why does the world look classical when $\hbar$ is tiny?

because $\hbar$ is tiny. for macroscopic actions — a baseball, a planet — $S$ is enormous compared to $\hbar$. the phase $e^{iS/\hbar}$ spins billions of times per tiny path variation. destructive interference is total. only paths extremely close to the classical trajectory survive.

for an electron, $S$ is comparable to $\hbar$. the phase changes slowly. many paths contribute. quantum behavior dominates.

the boundary isn't magic. it's a numbers game. small action, many paths matter. large action, only the classical path survives. the transition is continuous.

the payoff

the path integral explains interference naturally. in the double-slit experiment, the photon doesn't go through one slit or the other. it goes through both, and every other path besides. the amplitudes for paths through the left slit and paths through the right slit add, and their phases create the interference pattern. no wave function collapse needed — just addition of complex numbers.

it removes the special status of measurement. in copenhagen, you need a classical observer to collapse the wave function. in the path integral formulation, there's nothing to collapse. you calculate amplitudes for histories and square them to get probabilities. measurement is just a particular boundary condition, not a metaphysical event.

it also makes quantum field theory computable. feynman diagrams — those little stick figures physicists draw — are perturbative approximations of path integrals. each diagram represents a class of paths. the path integral is the generating function from which all feynman diagrams fall out.

and it unifies. the same formalism works for photons, electrons, quarks, fields. you can write path integrals for quantum electrodynamics, for the standard model, for gravity (though the last one breaks down in ways nobody has fixed).

the catch

the path integral over all paths is not always well-defined mathematically. the measure $\mathcal{D}[q(t)]$ — the thing that tells you how to weight each path — is formal. physicists know how to use it, but mathematicians have spent decades figuring out when it rigorously exists.

also, the integral includes paths backward in time, paths that violate energy conservation briefly, paths that do absolutely anything. feynman embraced this. he treated positrons as electrons moving backward in time. he treated virtual particles as real particles on weird paths. this is where the feynman diagram rules come from — they're bookkeeping devices for classes of paths.

dirac had the germ of the idea in 1933, in a paper on the lagrangian in quantum mechanics. but he didn't push it. feynman took dirac's hint, developed the full machinery, and made it a complete alternative formulation equal to heisenberg's matrices and schrödinger's equation.

the deeper thing

the path integral doesn't tell you what the photon really is. it sidesteps the question. wave or particle? the photon is neither. it's an amplitude for a process, calculated by summing over every possible history. the wave nature and particle nature are both emergent — from the interference of phases, from the selection of paths that survive cancellation.

this is why feynman was so irritating to bohr. bohr wanted you to accept complementarity as the final word. feynman just gave you a method that made the whole debate look provincial. why argue about ontology when you can calculate everything from a deeper principle?

"i think i can safely say that nobody understands quantum mechanics."

feynman said that too. but he also gave us the closest thing to understanding that anyone has managed. not by explaining what quantum objects are, but by describing exactly what they do — all of it, at once, across every possible path.

the path integral is now the default language of quantum field theory. the standard model is built on it. string theory is built on it. every calculation in particle physics, from the higgs mass to the muon's $g-2$ anomaly, traces back to some version of this integral.

newton's rings showed that light interferes. young's slits showed it diffracts. einstein showed it arrives in quanta. feynman showed that all of this — waves, particles, interference, quantization — falls out of a single principle: sum over histories.

the photon doesn't choose. nature doesn't choose. everything that can happen, does. and what we see is just what survives the cancellation.

the thought is mine. the words are written by janis, my hermes agent.