Why Complex Numbers in Quantum Theory

The puzzle classical probability cannot solve

Imagine a beam of light hitting a half-silvered mirror. Classical probability says: each photon either reflects (probability $1/2$ ) or transmits (probability $1/2$ ). Add a second mirror to reunite the two paths — a Mach–Zehnder interferometer — and classical probability predicts each detector fires half the time. Experiment says otherwise: with equal path lengths, one detector fires with certainty and the other never fires.

The difference is interference. Classical probabilities add; quantum amplitudes add, and interference can make a total amplitude of zero even when each partial amplitude is nonzero. To capture cancellation you need quantities that can subtract — quantities that can be negative, or more generally, complex.

What an amplitude is

Quantum mechanics replaces the classical probability $p(x)$ of an outcome $x$ with a probability amplitude $\psi(x)$ , a complex number. The actual probability is recovered by the Born rule:

P(x) = |\psi(x)|^2.

Because $|\psi|^2 \geq 0$ always, probabilities are non-negative. But $\psi$ itself can be negative, imaginary, or anywhere in the complex plane. Before squaring, amplitudes from different paths can interfere.

Consider two paths to the same outcome, with amplitudes $\psi_1$ and $\psi_2$ . The combined amplitude is $\psi_1 + \psi_2$ , so the probability is

|\psi_1 + \psi_2|^2 = |\psi_1|^2 + |\psi_2|^2 + 2\operatorname{Re}(\psi_1^* \psi_2).

The last term is the interference term. If $\psi_1 = A$ and $\psi_2 = -A$ (equal magnitudes, opposite signs), the interference term equals $-2A^2$ , exactly cancelling the first two terms and giving probability zero. Real numbers are sufficient for this sign flip, but once path lengths or phases are involved you need full complex arithmetic — angles in the complex plane, not just plus and minus.

Why real numbers alone are not enough

Could we restrict amplitudes to real numbers? Real numbers do allow interference (two paths can cancel), but they cannot describe the continuous, smooth phase evolution that quantum systems exhibit. The time-dependent Schrödinger equation,

i\hbar\frac{\partial\psi}{\partial t} = \hat{H}\psi,

has an explicit $i$ on the left-hand side. For a real Hamiltonian acting on a real-valued $\psi$ , the right-hand side $\hat{H}\psi$ is real, so the equation forces $\partial\psi/\partial t$ to be purely imaginary: $\partial\psi/\partial t = -\tfrac{i}{\hbar}\hat{H}\psi$ . The state therefore cannot stay real — it is immediately driven into the complex plane (the only exception is $\hat{H}\psi = 0$ , where it does not change at all). Complex numbers are therefore not a cosmetic choice — they are forced on us by the need to describe time evolution.

Amplitudes in a two-outcome measurement

A qubit (or any two-level system) has a state

|\psi\rangle = \alpha|0\rangle + \beta|1\rangle,

where $\alpha, \beta \in \mathbb{C}$ . The Born rule gives

P(\text{measure } 0) = |\alpha|^2, \quad P(\text{measure } 1) = |\beta|^2.

Normalization requires $|\alpha|^2 + |\beta|^2 = 1$ . The values $\alpha$ and $\beta$ are not probabilities; they are amplitudes that encode both magnitude and phase. The phase difference between $\alpha$ and $\beta$ affects what happens when the state is acted on by a quantum gate, and it shows up as interference in subsequent measurements.

The takeaway

Complex numbers appear in quantum mechanics because quantum systems exhibit interference, and capturing interference requires quantities that can cancel by phase — not just by sign. The Born rule $P = |\psi|^2$ then converts these complex amplitudes into real, non-negative probabilities at the moment of measurement. Everything else in quantum theory — unitaries, inner products, expectation values — is built from this one fact.