Born Rule

The whole apparatus of quantum theory — Hilbert spaces, unit vectors, unitary evolution, interference — would be a closed mathematical game if nothing in it ever touched a laboratory reading. The Born rule is the single bridge that connects the two. It takes the abstract state $|\psi\rangle$ , a unit vector no instrument can display directly, and returns the one thing an experiment actually produces: the relative frequency of each outcome when the same preparation is measured over and over. In its most elementary form, for a measurement in an orthonormal basis $\{|m\rangle\}$ , it says the probability of outcome $m$ is the squared modulus of an amplitude, $p(m)=|\langle m|\psi\rangle|^2$ . Two features make this short statement load-bearing. First, the amplitude $\langle m|\psi\rangle$ is complex, yet a single measurement returns only its squared modulus — so its phase is silent for one outcome but decides everything when amplitudes for different paths are added before squaring. That "add, then square" ordering is the entire origin of quantum interference, and it separates a quantum computer from a classical sampler. Second, the rule scales without seams: projective measurements give $p(m)=\langle\psi|P_m|\psi\rangle$ , mixed states $\mathrm{Tr}(P_m\rho)$ , and general measurements $\mathrm{Tr}(E_m\rho)$ , all collapsing back to $|\langle m|\psi\rangle|^2$ in the simplest case. And the quadratic form is not a free choice — Gleason's theorem shows it is forced by consistency alone in dimension three and above. This article assembles all of that, with every number re-derived.

Key takeaways

The Born rule assigns each measurement outcome a probability equal to the squared modulus of its amplitude: $p(m)=|\langle m|\psi\rangle|^2$ for a normalized state $|\psi\rangle$ .
The amplitude $\langle m|\psi\rangle$ is complex; a single measurement reveals only $|\langle m|\psi\rangle|^2$ , which is why an overall phase $e^{i\theta}$ is physically invisible.
Unit norm does the bookkeeping: $\sum_m p(m)=\sum_m|c_m|^2=\langle\psi|\psi\rangle=1$ , so the probabilities automatically sum to one.
It generalizes to $p(m)=\langle\psi|P_m|\psi\rangle$ (projective), $\mathrm{Tr}(P_m\rho)$ (mixed), and $\mathrm{Tr}(E_m\rho)$ (POVM), with $\langle A\rangle=\langle\psi|A|\psi\rangle$ .
Anchor numbers: a qubit at $\theta=60°$ gives $p(0)=0.75$ , $p(1)=0.25$ ; the same state read in the $\pm$ basis exposes its phase, $p(+)=\tfrac12(1+\sin\theta\cos\varphi)$ .
Gleason's theorem proves the quadratic form is the only additive probability rule for Hilbert-space dimension $\geq 3$ , so the squaring is forced, not stipulated.
Key idea: probability equals amplitude squared, $p(m)=|\langle m|\psi\rangle|^2$ ; a phase is invisible to one measurement but governs interference the instant amplitudes are added before squaring.

The core insight

Probability equals amplitude squared: $p(m)=|\langle m|\psi\rangle|^2$ for a normalized state. The phase of an amplitude is invisible to a single measurement but controls interference the moment two amplitudes are added before squaring.

Where this lives today

The Born rule is both the everyday engine of quantum technology and the target of precision tests. Every quantum computer reads out by sampling from $|\langle x|\psi\rangle|^2$ over bit strings $x$ , and algorithm design is the craft of shaping amplitudes so interference concentrates Born probability on the right answer — the mechanism behind Shor's and Grover's algorithms. Its most direct modern showcase is random-circuit sampling: Google's 2019 Sycamore experiment and its successors ran shallow random circuits and compared observed bit-string frequencies against the ideal Born probabilities through linear cross-entropy benchmarking, whose fidelity score is literally a correlation with $|\langle x|\psi\rangle|^2$ ; the conjectured classical hardness of reproducing that distribution is the whole claim of "quantum advantage." The rule is also under foundational scrutiny: it forbids genuine three-path interference, so a triple-slit experiment tests it by bounding the Sorkin parameter (the residual third-order interference), which Sinha and collaborators found consistent with zero at the $10^{-2}$ level in 2010, with photonic and molecular versions tightening the bound since. And deriving the rule rather than postulating it — from Gleason, from Everettian decision theory, or from Zurek's envariance — remains an open frontier.

The physical picture

Picture the state $|\psi\rangle$ as a unit-length arrow in an abstract complex space. A measurement nominates a set of orthogonal axes $\{|m\rangle\}$ and asks which one the system will register on. The Born rule answers geometrically: drop the arrow's shadow onto each axis, and the probability of landing on axis $m$ is the squared length of that shadow, $|\langle m|\psi\rangle|^2$ . Because the arrow has unit length, Pythagoras guarantees the squared shadows add to exactly one — a quantum Pythagorean theorem whose legs are probabilities. The first figure makes this concrete for a qubit, where the two shadows are $\cos^2(\theta/2)$ and $\sin^2(\theta/2)$ .

Figure

State at Bloch angle θ = 60°: projecting the unit state arrow onto the two outcome axes gives squared shadows p(0) = cos²30° = 0.75 and p(1) = sin²30° = 0.25, which sum to exactly 1 — a quantum Pythagorean theorem

The phase of an amplitude is the part of the arrow this squared-shadow picture hides — and hides for a reason. Rotate the entire arrow by a global factor $e^{i\theta}$ and every squared shadow is untouched, so a global phase carries no physical consequence. A relative phase between components is different: when two amplitudes route to the same outcome, the Born rule adds them first and squares afterward, so their relative phase decides whether they reinforce toward certainty or cancel toward impossibility. The slogan is exact — classical probability adds probabilities, quantum mechanics adds amplitudes and squares at the end. The second figure turns that phase loose, plotting the Born probability of a $\pm$ -basis measurement as $\varphi$ is dialed: a full interference fringe.

Figure

Interference fringe: the relative phase φ, invisible in the computational basis, reappears in the ± basis as the Born probability p(+) = cos²(φ/2) = (1 + cosφ)/2 — a full sinusoid running from p(+) = 1 at φ = 0 to p(+) = 0 at φ = π

The squaring is not a modelling convenience that could have been chosen differently. Gleason's theorem establishes that, for any Hilbert space of dimension three or more, the only additive assignment of probabilities to measurement outcomes is the trace form $\mathrm{Tr}(P_m\rho)$ — of which $|\langle m|\psi\rangle|^2$ is the pure-state case. The quadratic law is dictated by the geometry of the state space together with the ordinary axioms of probability, not appended by hand.

The core equations

The elementary Born rule is a statement about one measurement basis. It takes the complex amplitude for an outcome and returns a real, non-negative probability by squaring its modulus — the definition from which every other form below is derived.

p(m)=|\langle m|\psi\rangle|^2 \qquad (1)

where: $p(m)$ — probability of obtaining outcome $m$ ; $|m\rangle$ — orthonormal basis vector labelling outcome $m$ ; $|\psi\rangle$ — normalized state, $\langle\psi|\psi\rangle=1$ ; $\langle m|\psi\rangle$ — the complex amplitude for outcome $m$ . Read it as: complex amplitude in, real probability out. Writing the state in the measurement basis exposes the amplitudes as expansion coefficients, and orthonormality fixes each to the inner product with the corresponding basis vector.

|\psi\rangle=\sum_m c_m|m\rangle,\qquad c_m=\langle m|\psi\rangle \qquad (2)

where: $c_m$ — amplitude (expansion coefficient) of $|\psi\rangle$ along $|m\rangle$ . The identification $c_m=\langle m|\psi\rangle$ makes "coefficient in the sum" and "amplitude in the Born rule" the same object. Because the state is a unit vector, the squared amplitudes already form a normalized distribution — the norm does the work, no extra assumption needed.

\sum_m p(m)=\sum_m|c_m|^2=\langle\psi|\psi\rangle=1 \qquad (3)

where: $\langle\psi|\psi\rangle$ — squared norm of the state, equal to one for any physical state; if unnormalized, divide by it to recover genuine probabilities. When outcomes correspond to subspaces rather than single vectors — a degenerate observable — the amplitude form is replaced by a projector sandwiched between the state and its dual.

p(m)=\langle\psi|P_m|\psi\rangle,\qquad P_m=P_m^\dagger=P_m^2,\quad \sum_m P_m=I \qquad (4)

where: $P_m$ — orthogonal projector onto the subspace of outcome $m$ ; $P_m^\dagger$ — its adjoint, and Hermiticity guarantees a real probability; $I$ — identity, so completeness $\sum_m P_m=I$ enforces normalization. When the preparation is a statistical mixture rather than a pure state, the state is a density operator and the Born rule becomes a trace — the single formula covering every case in this article.

p(m)=\mathrm{Tr}(P_m\rho),\qquad \rho\geq 0,\;\mathrm{Tr}\,\rho=1 \qquad (5)

where: $\rho$ — density operator, positive semidefinite with unit trace; $\mathrm{Tr}$ — the trace, a sum of diagonal matrix elements. For a pure state $\rho=|\psi\rangle\langle\psi|$ this collapses back to equation (4). Averaging each eigenvalue against its Born probability gives the expectation value of an observable as a quadratic form in the state — the quantity most experiments report.

\langle A\rangle=\sum_m a_m\,p(m)=\langle\psi|A|\psi\rangle \qquad (6)

where: $A$ — observable with spectral decomposition $A=\sum_m a_m P_m$ ; $a_m$ — the eigenvalue recorded for outcome $m$ . The middle expression is the lab definition (weighted average of readings), the right is the operator shortcut. Two consequences worth isolating: a global phase cancels identically, and an equal superposition read in the diagonal basis converts a relative phase into an interference term.

|\psi\rangle\to e^{i\theta}|\psi\rangle\;\Rightarrow\;p(m)=|e^{i\theta}\langle m|\psi\rangle|^2=|\langle m|\psi\rangle|^2 \qquad (7)

p(\pm)=\tfrac{1}{2}\,|c_0\pm c_1|^2 \qquad (8)

where: $e^{i\theta}$ — arbitrary global phase, $|e^{i\theta}|=1$ ; $c_0,c_1$ — amplitudes for $|0\rangle,|1\rangle$ ; $|\pm\rangle=\tfrac{1}{\sqrt2}(|0\rangle\pm|1\rangle)$ . Equation (7) is why physical states are rays; equation (8) is interference in one line.

Derivation AAmplitudes squared and why the probabilities sum to one

The goal is to establish the elementary Born rule as a self-consistent probability law rather than a bare assertion: to show that the expansion coefficient $c_m$ is exactly the inner product $\langle m|\psi\rangle$ , that the resulting squared moduli are automatically real and non-negative, and that they sum to precisely one because — and only because — the state is a unit vector. We project the state onto the orthonormal basis to pin down the coefficients, then close with the unit-norm condition so that total probability need not be assumed separately.

Step 1. Expand the state in the measurement basis, $|\psi\rangle=\sum_m c_m|m\rangle$ , and act with the bra $\langle k|$ . Linearity of the inner product gives $\langle k|\psi\rangle=\sum_m c_m\langle k|m\rangle$ . Why: the inner product is the tool that reaches inside a superposition and extracts one component, linearly.

Step 2. Impose orthonormality, $\langle k|m\rangle=\delta_{km}$ , collapsing the sum to its single surviving term: $\langle k|\psi\rangle=c_k$ . So the coefficient is the amplitude, $c_m=\langle m|\psi\rangle$ . Why: this fuses the algebraic "coefficient" with the physical "amplitude" — without orthonormality they would differ.

Step 3. Form the total probability $\sum_m p(m)=\sum_m|c_m|^2$ . Each term $|c_m|^2=c_m^*c_m$ is a squared modulus, hence real and $\geq 0$ , so every $p(m)$ is legitimate before normalization is even checked. Why: reality and non-negativity are free — built into "squared modulus", which is why the rule never produces a negative or complex probability.

Go deeper · Derivation B · 1 minPhase made visible — interference in a rotated basis

Derivation BPhase made visible — interference in a rotated basis

The goal is to demonstrate that the phase of an amplitude, provably invisible to a single measurement in one basis, becomes fully observable in another. Starting from an equal-weight superposition carrying a tunable relative phase $\varphi$ , we measure in the diagonal $\pm$ basis, add the two path amplitudes before squaring exactly as the Born rule demands, and recover $p(+)=\cos^2(\varphi/2)$ — a probability that sweeps from certainty to impossibility as $\varphi$ runs from $0$ to $\pi$ , the mathematical content of every interference experiment.

Step 1. Prepare $|\psi\rangle=\tfrac{1}{\sqrt2}\big(|0\rangle+e^{i\varphi}|1\rangle\big)$ . In the computational basis the Born rule gives $p(0)=p(1)=\tfrac12$ for every $\varphi$ : the phase is completely hidden. Why: $|\tfrac{1}{\sqrt2}|^2$ and $|\tfrac{1}{\sqrt2}e^{i\varphi}|^2$ both equal $\tfrac12$ , so no computational-basis statistic sees $\varphi$ .

Step 2. Compute the amplitude for $|+\rangle=\tfrac{1}{\sqrt2}(|0\rangle+|1\rangle)$ : $\langle +|\psi\rangle=\tfrac12(1+e^{i\varphi})$ — the two paths' amplitudes have added. Why: this addition-before-squaring is the whole mechanism of interference; the relative phase now sits inside a sum, where it cannot cancel.

Step 3. Square the modulus: $p(+)=\tfrac14|1+e^{i\varphi}|^2=\tfrac14(2+2\cos\varphi)=\cos^2\tfrac{\varphi}{2}$ , using $|1+e^{i\varphi}|^2=2+2\cos\varphi$ . Why: the hidden phase re-emerges as a real, measurable modulation — the fringe of the second figure.

Step 4. The complementary outcome is $p(-)=\sin^2\tfrac{\varphi}{2}$ . At $\varphi=0$ the state is exactly $|+\rangle$ and $p(+)=1$ ; at $\varphi=\pi$ it is $|-\rangle$ and $p(+)=0$ — the phase is now fully determined by the statistics. Why: what was unobservable in one basis is fully observable in another; basis choice is an experimental degree of freedom.

Go deeper · Derivation C · 1 minFrom state vectors to operators, and why the form is forced

Derivation CFrom state vectors to operators, and why the form is forced

The goal is to climb from the vector form of the Born rule to its operator forms and then explain why the quadratic law is not an arbitrary modelling choice. We write a pure state as the rank-one projector $\rho=|\psi\rangle\langle\psi|$ , use cyclicity of the trace to prove $\mathrm{Tr}(P_m\rho)=\langle\psi|P_m|\psi\rangle$ , assemble expectation values into $\langle A\rangle=\langle\psi|A|\psi\rangle$ , and recall Gleason's theorem, which shows this trace expression is the unique additive probability assignment once the Hilbert space has dimension three or more.

Step 1. Represent a pure state by its rank-one projector $\rho=|\psi\rangle\langle\psi|$ and insert it: $p(m)=\mathrm{Tr}(P_m\rho)=\mathrm{Tr}(P_m|\psi\rangle\langle\psi|)$ . Why: promoting the state to an operator lets pure and mixed states share one formula; the pure case must be recovered.

Step 2. Apply cyclicity $\mathrm{Tr}(ABC)=\mathrm{Tr}(CAB)$ to move $\langle\psi|$ to the front: the argument becomes the scalar $\langle\psi|P_m|\psi\rangle$ , whose trace is itself, so $p(m)=\langle\psi|P_m|\psi\rangle$ — equation (4), and for $P_m=|m\rangle\langle m|$ this is $|\langle m|\psi\rangle|^2$ . Why: cyclicity is exact and basis-independent, so the reduction holds for any projector and normalized state.

Step 3. Build the expectation value by averaging eigenvalues against Born probabilities: $\langle A\rangle=\sum_m a_m\langle\psi|P_m|\psi\rangle=\langle\psi|\big(\sum_m a_m P_m\big)|\psi\rangle=\langle\psi|A|\psi\rangle$ , using $A=\sum_m a_m P_m$ . Why: this turns a laboratory average of discrete readings into a single quadratic form, the quantity most experiments report and algorithms optimize.

Step 4. Why must the form be quadratic? Gleason's theorem: any probability assignment on projectors that is additive over orthogonal ones and sums to one on a complete set must equal $\mathrm{Tr}(P_m\rho)$ for some $\rho$ , provided $\dim\mathcal H\geq 3$ . Why: the squaring is a theorem, not a postulate — additivity plus subspace geometry leaves no other consistent rule.

Limiting cases

Case	Result	Meaning
Basis measurement of a pure state	$p(m)=\vert\langle m\vert\psi\rangle\vert^2$	The elementary form: $P_m=\vert m\rangle\langle m\vert$ is rank-one and $\langle\psi\vert P_m\vert\psi\rangle=\vert\langle m\vert\psi\rangle\vert^2$ .
Eigenstate of the measurement	$\vert\psi\rangle=\vert k\rangle\Rightarrow p(m)=\delta_{mk}$	A state already aligned with an outcome gives that outcome with certainty and all others probability zero.
Mixed state	$p(m)=\mathrm{Tr}(P_m\rho)$	For $\rho=\vert\psi\rangle\langle\psi\vert$ this collapses to $\langle\psi\vert P_m\vert\psi\rangle$ , recovering the pure-state rule.
Degenerate outcome (subspace)	$p(m)=\langle\psi\vert P_m\vert\psi\rangle$ , $\mathrm{rank}\,P_m>1$	The amplitude form is insufficient; sum the squared amplitudes over an orthonormal basis of the subspace.
POVM (most general measurement)	$p(m)=\mathrm{Tr}(E_m\rho)$ , $\sum_m E_m=I$	With positive operators $E_m\geq 0$ ; projective measurement is the special case $E_m=P_m$ with $P_m^2=P_m$ .
Global phase change	$\vert\psi\rangle\to e^{i\theta}\vert\psi\rangle\Rightarrow p(m)$ unchanged	Overall phase is unobservable; physical states are rays, not vectors.

Common trapProbability is the amplitude squared — not the amplitude, and not its real part

The most common error is to treat the amplitude itself, or its real part, as a probability. Amplitudes are complex, can be negative, and interfere — they add before squaring; probabilities are non-negative reals that never interfere. So $\langle m|\psi\rangle=-i/\sqrt2$ is a fine amplitude, but the probability it carries is $|-i/\sqrt2|^2=\tfrac12$ , not $-i/\sqrt2$ and not $0$ (its real part). A quieter trap: $|\langle m|\psi\rangle|^2$ is a probability only when $|\psi\rangle$ is normalized — otherwise divide by $\langle\psi|\psi\rangle$ , or the "probabilities" will not sum to one. And a relative phase that vanishes from one basis has not vanished from the physics; measure in another basis and it returns as a full interference fringe.

Worked examples

1 · One state, two bases (fully worked)

A qubit is prepared at Bloch angles $\theta=60°$ , $\varphi=40°$ : $|\psi\rangle=\cos\tfrac{\theta}{2}|0\rangle+e^{i\varphi}\sin\tfrac{\theta}{2}|1\rangle$ . Find its probabilities in the computational and $\pm$ bases, and $\langle Z\rangle$ .

Step 1. Read the amplitudes with $\theta/2=30°$ : $\langle 0|\psi\rangle=\cos30°=0.8660$ and $\langle 1|\psi\rangle=0.5\,e^{i40°}$ . The phase lives entirely on the $|1\rangle$ amplitude.

Step 2. Computational-basis probabilities: $p(0)=|0.8660|^2=\mathbf{0.75}$ and $p(1)=|0.5\,e^{i40°}|^2=\mathbf{0.25}$ . The phase $e^{i40°}$ vanishes under the modulus, so this basis is blind to $\varphi$ ; check $0.75+0.25=1$ .

Step 3. Expectation of $Z$ (eigenvalues $\pm1$ ): $\langle Z\rangle=(0.75)-(0.25)=\mathbf{0.5}=\cos\theta$ — the Bloch $z$ -coordinate, exactly as $\langle\psi|Z|\psi\rangle$ predicts.

Step 4. Now the $\pm$ basis: $p(+)=\tfrac12|\cos\tfrac{\theta}{2}+e^{i\varphi}\sin\tfrac{\theta}{2}|^2=\tfrac12(1+\sin\theta\cos\varphi)$ . With $\sin60°=0.8660$ , $\cos40°=0.7660$ : $p(+)=\tfrac12(1+0.6634)=\mathbf{0.832}$ , so $p(-)=\mathbf{0.168}$ .

Step 5. The lesson: the same state gives $(0.75,0.25)$ in one basis and $(0.832,0.168)$ in another. The phase $\varphi=40°$ was invisible in Step 2 and decisive in Step 4 — measurement basis is a physical choice, with $\langle Z\rangle=0.5$ but $\langle X\rangle=\sin\theta\cos\varphi=0.663$ .

2 · How many shots? The Born rule as a frequency law

The qubit above has $p(1)=0.25$ . How many counts of outcome $1$ do you expect in $N=1000$ shots, with what spread, and how many shots pin $p(1)$ to $\pm 0.01$ ?

Step 1. Each shot is a Bernoulli trial with success probability $p=0.25$ , so the expected count is $Np=\mathbf{250}$ .

Step 2. Its standard deviation is $\sqrt{Np(1-p)}=\sqrt{187.5}=\mathbf{13.7}$ , so a run of 1000 shots typically returns $250\pm14$ counts — the irreducible shot noise of the Born distribution.

Step 3. The estimate $\hat p=(\text{count})/N$ has standard error $\sqrt{p(1-p)/N}=\sqrt{0.1875/1000}=\mathbf{0.0137}$ . To reach $0.01$ , invert: $N=0.1875/0.0001=\mathbf{1875}$ shots.

Step 4. The point: the Born rule predicts a distribution, so extracting even one probability costs $O(1/\varepsilon^2)$ repetitions for precision $\varepsilon$ , and full tomography multiplies this by the number of bases — why characterizing a device is so measurement-hungry. 250 ± 14 counts; ~1875 shots for ±0.01.

3 · A degenerate outcome on two qubits

Two qubits are in $|\psi\rangle=\tfrac{1}{\sqrt3}(|00\rangle+|01\rangle+|10\rangle)$ . Measure only whether the first qubit is $0$ , with projector $P_0=|0\rangle\langle 0|\otimes I$ (rank 2). Find $p_0$ and the post-measurement state.

Step 1. $P_0$ selects the terms whose first qubit is $0$ — $|00\rangle$ and $|01\rangle$ , each with amplitude $1/\sqrt3$ . This is a degenerate outcome, so sum squared amplitudes over the subspace: $p_0=\tfrac13+\tfrac13=\mathbf{2/3}$ .

Step 2. The complementary outcome (first qubit $1$ ) keeps only $|10\rangle$ : $p_1=\tfrac13$ , and $p_0+p_1=1$ .

Step 3. The post-measurement state renormalizes the projection: $P_0|\psi\rangle/\sqrt{p_0}=\tfrac{1}{\sqrt2}(|00\rangle+|01\rangle)$ . p₀ = 2/3, collapsing to (|00⟩+|01⟩)/√2. This is why equation (4), not the bare amplitude form, is the tool for coarse-grained measurements.

Physics of the portrait

The Born rule is where the wavefunction stopped being a puzzle and became a probability — added almost as an afterthought, in a footnote corrected in proof, yet it reshaped what physics claims to describe.

Origins · 1 minMax Born and how this was discovered

Max Born

1882–1970 · Göttingen, Edinburgh · the statistical interpretation of the wavefunction

Max Born was the mathematician-physicist at the center of the Göttingen group that, with his assistants Heisenberg and Jordan, built matrix mechanics in 1925. But his most quoted contribution came the following year. In the June 1926 paper Zur Quantenmechanik der Stoßvorgänge ("On the quantum mechanics of collision processes"), analyzing how a wavefunction scatters, Born proposed that $\psi$ does not describe a density of stuff but a probability: the quantity governing where a particle is found is the squared modulus of $\psi$ . Famously he first wrote the probability as proportional to $\psi$ itself in the body of the paper, then corrected it to $|\psi|^2$ in a footnote added in proof — one of the most consequential footnotes in physics. Einstein never accepted the resulting indeterminism ("God does not play dice"), and Born defended it for the rest of his life. The interpretation dissolved the paradox of a spreading wave that still registers as a single localized click: the wave sets the odds, the click is the draw. It reframed physics from predicting trajectories to predicting distributions, and made quantum mechanics a statistical theory at its foundation, not merely in practice. Born received the Nobel Prize in 1954 for exactly "his statistical interpretation of the wavefunction" — nearly three decades after the footnote that earned it.

History

1926

— Born interprets |ψ|² as a probability, corrected from |ψ| in a footnote added in proof

1927

— Dirac and Jordan's transformation theory recasts amplitudes as probability transition kernels

1932

— von Neumann axiomatizes measurement and the projection postulate in his Grundlagen

1954

— Born awarded the Nobel Prize for the statistical interpretation of the wavefunction

1957

— Gleason proves the Born trace form is forced for Hilbert-space dimension three or more

Going further

Texts. Nielsen & Chuang, Quantum Computation and Quantum Information, §2.2.3 (measurement postulate and Born rule) · Preskill, Quantum Information lecture notes, Ch. 2 · Watrous, The Theory of Quantum Information, §2.3 (Gleason and measurements) · Wilde, Quantum Information Theory, Ch. 3 · Kitaev, Shen & Vyalyi, Classical and Quantum Computation, Ch. 8 · Mermin, Quantum Computer Science, Ch. 1. Key results to keep in mind: $c_m=\langle m|\psi\rangle$ follows from orthonormality plus completeness; the qubit at $\theta=60°$ gives $p(0)=0.75$ , $p(1)=0.25$ , $\langle Z\rangle=\cos\theta=0.5$ ; the interference result $p(+)=\cos^2(\varphi/2)=\tfrac12(1+\cos\varphi)$ follows from $|1+e^{i\varphi}|^2=2+2\cos\varphi$ , and the two-basis value at $\varphi=40°$ is $p(+)=\tfrac12(1+\sin\theta\cos\varphi)=0.832$ ; the trace identity $\mathrm{Tr}(P_m|\psi\rangle\langle\psi|)=\langle\psi|P_m|\psi\rangle$ holds by cyclicity; the shot-noise numbers are $Np=250$ , $\sqrt{Np(1-p)}=13.7$ , $N=1875$ ; and the degenerate-outcome probability is $p_0=2/3$ . Gleason's theorem (dimension $\geq 3$ ) justifies the quadratic form.

Builds on: The Qubit and Its State Space · Dirac Bra-Ket Notation · Inner Product, Norm, and Normalization · Superposition Principle. Leads to: The Measurement Postulate · Projective Measurements · Expectation Values and Variance · Density Operators · Global and Relative Phase.

Ask the tutor: Want to see how a chosen measurement basis turns a hidden relative phase into a visible interference fringe — or how many shots your target precision really costs? Ask, and we will work it through with numbers.

Why the squared modulus, and not the amplitude or its bare square?

Three constraints leave no alternative. A probability must be real and non-negative, but amplitudes are complex — only $|c|^2=c^*c$ is guaranteed real and $\geq 0$ for every complex $c$ . The probabilities must sum to one, and $\sum_m|c_m|^2=\langle\psi|\psi\rangle$ does so exactly when the state is normalized. And the rule must be additive over orthogonal outcomes consistently across all measurements — which Gleason's theorem shows forces the trace form $\mathrm{Tr}(P_m\rho)$ , whose pure-state case is $|\langle m|\psi\rangle|^2$ . The plain amplitude fails reality; the bare square $c^2$ is complex. Only $|c|^2$ survives all three.

If a global phase is unobservable, why track phase at all?

Because a relative phase is very observable, and you cannot tell "the phase that matters" from "the phase that does not" until you fix a basis. A global phase multiplies every amplitude equally and cancels in every Born probability, so $|\psi\rangle$ and $e^{i\theta}|\psi\rangle$ are the same physical state. But the relative phase between components — the $\varphi$ in $\tfrac{1}{\sqrt2}(|0\rangle+e^{i\varphi}|1\rangle)$ — controls interference and appears directly in the right basis as the fringe $p(+)=\cos^2(\varphi/2)$ . You track phases throughout a calculation because relative phases accumulate under gates and only become discardable at the very end.

Does the Born rule apply to a single measurement, or only to many?

It assigns a probability to the single next outcome, but a probability is only testable against a frequency, which needs repetition. One shot on a qubit with $p(1)=0.25$ yields a definite bit — $0$ or $1$ — and no way to check the number $0.25$ . Repeat the identical preparation a thousand times and the frequency of $1$ converges to $0.25$ with a spread of about $\sqrt{p(1-p)/N}$ . So the rule is a statement about an ensemble of identically prepared systems, or equivalently your rational betting odds on the one system in front of you; both readings give the same arithmetic.

What is the difference between the projective form and a POVM?

A projective measurement uses orthogonal projectors $P_m$ with $P_mP_{m'}=\delta_{mm'}P_m$ and $\sum_m P_m=I$ ; its outcomes are mutually exclusive subspaces and, repeated immediately, it gives the same result (it is "sharp"). A POVM replaces the projectors by any positive operators $E_m\geq 0$ with $\sum_m E_m=I$ , dropping $E_m^2=E_m$ and orthogonality. The Born rule keeps its form — $p(m)=\mathrm{Tr}(E_m\rho)$ — but POVMs can have more outcomes than the dimension and need not be repeatable. They arise naturally when a system is measured jointly with an ancilla, and are essential for tasks like unambiguous state discrimination.

How does the Born rule connect to expectation values and variances?

An observable $A=\sum_m a_m P_m$ has expectation value equal to its eigenvalues weighted by their Born probabilities: $\langle A\rangle=\sum_m a_m p(m)=\langle\psi|A|\psi\rangle$ . That quadratic form is what most experiments and variational algorithms estimate, since a mean needs no full distribution. The variance follows the same way, $\mathrm{Var}(A)=\langle\psi|A^2|\psi\rangle-\langle\psi|A|\psi\rangle^2$ , quantifying the spread of Born outcomes — and vanishing exactly when $|\psi\rangle$ is an eigenstate of $A$ , i.e. when the outcome is certain.

Can the Born rule be derived, or must it be assumed?

Both positions are defensible, which is why it stays interesting. In the standard axioms it is a postulate. But Gleason's theorem shows that if you accept the state-space structure and demand additive probabilities over orthogonal projectors, the trace form is forced for dimension $\geq 3$ — a strong derivation, though it presupposes that probabilities attach to projectors at all. Other programs go further: decision-theoretic arguments in the Everettian framework (Deutsch, Wallace) and Zurek's envariance argument from the symmetry of entangled states both aim to derive the $|c|^2$ weights from more primitive assumptions. None is universally accepted, so "why $|\psi|^2$ ?" remains a live foundational question even as the rule is confirmed to high precision.