The Bell States and Maximally Entangled Pairs

Two qubits live in a four-dimensional space, and the basis everyone learns first is the product basis $\{|00\rangle,|01\rangle,|10\rangle,|11\rangle\}$ , where every vector factors cleanly into "qubit $A$ is here and qubit $B$ is there." The Bell basis is a different orthonormal basis of the very same space — $|\Phi^\pm\rangle=(|00\rangle\pm|11\rangle)/\sqrt2$ and $|\Psi^\pm\rangle=(|01\rangle\pm|10\rangle)/\sqrt2$ — and its defining feature is that none of its four vectors factors. Each is maximally entangled: trace out either qubit and the other is left in the maximally mixed state $\tfrac12 I$ , with entanglement entropy $\ln 2$ — the largest a two-qubit split allows. Locally, each half looks like pure noise; every drop of structure lives in the correlations. This article proves the three structural facts that make the Bell states the currency of quantum information. First, they form an orthonormal, complete basis — a fact we verify by computing all sixteen inner products and the resolution of the identity. Second, the four are interconverted by a single local Pauli: $(I\otimes\sigma_i)|\Phi^+\rangle$ sweeps out the whole family, so they are "one resource in four flavours." Third — the deepest — the singlet $|\Psi^-\rangle$ is invariant under $U\otimes U$ for every SU(2) rotation, making it the unique spin-zero, basis-free anti-correlator whose perfect correlations launched the EPR argument. We sketch how projecting onto this basis powers dense coding and teleportation, and we are careful throughout: the inequality that turns these correlations into a refutation of local realism — with its Tsirelson value $2\sqrt2$ — is derived next door in Bell's Theorem and the CHSH Inequality, not here.

Key takeaways

The four Bell states $|\Phi^\pm\rangle,|\Psi^\pm\rangle$ are a maximally-entangled orthonormal basis of two qubits; each has reduced state $I/2$ . A single local Pauli maps one into any other, and the singlet $|\Psi^-\rangle$ is $U\otimes U$ -invariant (total spin zero). They are the resource behind teleportation and dense coding.

The core insight

Bell states are the "unit cells" of two-qubit entanglement: maximally correlated, locally random, mutually orthogonal, and all reachable from one another by acting on a single qubit. Measuring in the Bell basis reads out which correlation two qubits share — the operation at the heart of teleportation.

Where this lives today

Bell states are the standard unit of an entangled resource: entanglement is quantified in "ebits," one ebit being one shared Bell pair. Bell-state measurement is the primitive of the quantum internet — entanglement swapping in quantum repeaters is a Bell measurement performed on halves of two independent pairs — and the 2022 Nobel Prize (Aspect, Clauser, Zeilinger) rewarded experiments that first violated Bell inequalities on such pairs (the inequalities themselves are developed in Bell's Theorem and the CHSH Inequality, the experiments in Experimental Tests and Violation of Bell Inequalities). Dense coding and teleportation, both built on the local-Pauli structure above, are the textbook proof that entanglement is a usable resource, not a curiosity.

Origins · 1 minjohn stewart bell and how this was discovered

john stewart bell

1928–1990 · Belfast, Geneva

John Stewart Bell was, by day, an accelerator physicist at CERN — he co-discovered the Adler–Bell–Jackiw anomaly of quantum field theory. Foundations of quantum mechanics were his "hobby," pursued on sabbatical, yet they reshaped the subject. In 1964 he proved that no local hidden-variable theory can reproduce the correlations of an entangled pair like the singlet above — turning the EPR debate from philosophy into an experiment (that inequality is derived in Bell's Theorem and the CHSH Inequality). Two years later, in a review written in 1964 but published in 1966, he dismantled von Neumann's famous 1932 "proof" that hidden variables are impossible, exposing an unphysical assumption — a story told in Hidden Variable Theories. Bell built his argument on exactly the maximally-entangled states of this article; the singlet's axis-independent anti-correlation is the raw material his theorem sharpens into a contradiction with local realism. He died suddenly in 1990, the year he was reportedly to be nominated for the Nobel Prize his work made possible.

History

1935

— Schrödinger identifies the singlet's non-classical correlations

1951

— Bohm recasts EPR using the spin singlet, the modern Bell state

1964

— Bell proves entangled correlations violate local hidden variables

1992

— Bennett and Wiesner: superdense coding from Bell states

1993

— Bennett and collaborators: quantum teleportation via Bell measurement

The physical picture

Start from $|\Phi^+\rangle=(|00\rangle+|11\rangle)/\sqrt2$ . If you measure qubit $A$ in the computational basis you get $0$ or $1$ with equal probability, and instantly qubit $B$ is forced to match. The two outcomes are perfectly correlated — yet before measurement neither qubit has a value. That is the signature of maximal entanglement: no local property, only a shared rule. The other three Bell states encode different shared rules: $|\Phi^-\rangle$ is the same correlation with a relative sign, $|\Psi^+\rangle$ is anti-correlation with a plus sign, and $|\Psi^-\rangle$ — the singlet — is anti-correlation along every axis at once. The four are an orthonormal basis, so any two-qubit state is a superposition of them, and any complete two-qubit measurement can be a Bell measurement: a single joint measurement that asks "which of the four correlation patterns are you in?" This is a genuinely non-local measurement — it cannot be done by measuring $A$ and $B$ separately — and it is the primitive operation of quantum communication. Teleportation is, at bottom, "Alice does a Bell measurement, tells Bob two classical bits, Bob applies the matching Pauli." The reason a single Pauli shuffles the whole family is that the Pauli matrices are a basis for two-qubit operations on one side, and each flips exactly one of the two "rules" (the sign, or the correlation-vs-anticorrelation). And the reason the singlet is special is symmetry: it is the total-spin-zero state, and a spin-zero object looks the same after any rotation — so it carries no directional information whatsoever, only the bare instruction "opposite, always."

Figure

The Bell basis — four maximally entangled states, all reachable from |Φ⁺⟩ by one local Pauli

Figure

Locally random, globally correlated — each half of a Bell pair is pure noise

Derivation AThe four Bell states form a maximally-entangled orthonormal basis

We show the four vectors are orthonormal, complete, and that each leaves either qubit maximally mixed. These three facts together are what "maximally entangled orthonormal basis" means, and they are the foundation everything else rests on. The tools are the inner product on $\mathbb{C}^2\otimes\mathbb{C}^2$ and the partial trace defined in The Reduced Density Matrix and Partial Trace.

Step 1. Orthonormality. Compute inner products directly: $\langle\Phi^+|\Phi^-\rangle=\tfrac12(\langle00|+\langle11|)(|00\rangle-|11\rangle)=\tfrac12(1-1)=0$ , and $\langle\Phi^\pm|\Psi^\pm\rangle=0$ because $|\Phi\rangle$ lives in the $\{|00\rangle,|11\rangle\}$ sector and $|\Psi\rangle$ in the $\{|01\rangle,|10\rangle\}$ sector. Each is normalized: $\langle\Phi^+|\Phi^+\rangle=\tfrac12(1+1)=1$ . Why: the plus/minus signs make orthogonal partners inside each sector, and the two sectors are already orthogonal — all sixteen inner products give the identity matrix, confirmed in the verify script.

Step 2. Completeness. Sum the projectors: $\sum_B|B\rangle\langle B|=I_4$ . Why: four orthonormal vectors in a four-dimensional space are automatically a basis, so their projectors resolve the identity — the Bell basis is a legitimate complete measurement.

Step 3. Each is maximally entangled. Take $|\Phi^+\rangle$ and trace out $B$ : $\rho_A=\mathrm{Tr}_B|\Phi^+\rangle\langle\Phi^+|=\tfrac12(|0\rangle\langle0|+|1\rangle\langle1|)=\tfrac12 I$ . Why: the two Schmidt terms $|00\rangle,|11\rangle$ have equal weight $\tfrac12$ , so the reduced state is the flat mixture — the Schmidt coefficients $\{1/\sqrt2,1/\sqrt2\}$ from The Schmidt Decomposition are as spread as possible.

Step 4. Read off the entropy. With $\rho_A=I/2$ the von Neumann entropy is $S(\rho_A)=-\tfrac12\ln\tfrac12-\tfrac12\ln\tfrac12=\ln2$ , the maximum for one qubit — hence "maximally entangled." Why: entanglement entropy (built in Entanglement Entropy and Measures of Entanglement) equals the local mixedness, and $I/2$ is the most mixed a qubit can be. All four Bell states give the same $\ln2$ . *Verified in `verify_aq-qm-bell-states.py`: all sixteen inner products form $I$ , $\sum_B|B\rangle\langle B|=I_4$ , every $\rho_A=I/2$ , every $S(\rho_A)=\ln2$ .*

Go deeper · Derivation B · 1 minLocal-Pauli interconversion and the singlet's rotational invariance

Derivation BLocal-Pauli interconversion and the singlet's rotational invariance

Now the two algebraic gems: any Bell state is one local Pauli away from any other, and the singlet is the unique member invariant under identical rotations of both qubits. The first fact powers dense coding; the second is why the singlet is the natural "reference frame–free" resource and the seed of the EPR argument.

Step 1. One Pauli walks the family. Act on qubit $B$ of $|\Phi^+\rangle$ : $(I\otimes\sigma_x)|\Phi^+\rangle=(|01\rangle+|10\rangle)/\sqrt2=|\Psi^+\rangle$ ; $(I\otimes\sigma_z)|\Phi^+\rangle=(|00\rangle-|11\rangle)/\sqrt2=|\Phi^-\rangle$ ; $(I\otimes i\sigma_y)|\Phi^+\rangle=|\Psi^-\rangle$ . Why: $\sigma_x$ flips $B$ 's bit (correlation $\to$ anti-correlation), $\sigma_z$ inserts a relative sign, and their product does both — the three non-trivial Paulis reach the three other Bell states, so two classical bits' worth of choice is encoded in one local operation. This is exactly dense coding.

Step 2. The singlet is a total-spin-zero state. Write $|\Psi^-\rangle=(|\uparrow\downarrow\rangle-|\downarrow\uparrow\rangle)/\sqrt2$ — the $j=0$ combination from Addition of Angular Momenta. Apply the total spin operator: $\hat{\mathbf{S}}_{\text{tot}}^2|\Psi^-\rangle=0$ . Why: the antisymmetric combination of two spin- $\tfrac12$ 's carries zero total angular momentum, and a spin-zero object is a scalar — it points nowhere.

Step 3. Hence $U\otimes U$ -invariance. For any single-qubit rotation $U\in$ SU(2), $(U\otimes U)|\Psi^-\rangle=|\Psi^-\rangle$ up to an irrelevant global phase. Why: rotating a total-spin-zero state by the same $U$ on both halves is a rotation of a scalar, which does nothing. The triplet states are not invariant — they transform as a vector — so the singlet is singled out as the unique basis-free anti-correlator.

Step 4. Consequence: axis-independent anti-correlation. Because $|\Psi^-\rangle$ is the same in every rotated basis, measuring both spins along any common axis $\hat n$ gives $\langle\sigma_{\hat n}^A\sigma_{\hat n}^B\rangle=-1$ — perfect anti-correlation, every direction. Why: invariance under $U\otimes U$ means the correlation cannot depend on the chosen axis; this axis-independence is precisely the "element of reality" engine of The EPR Paradox and Local Realism. Verified in the script: $(I\otimes\sigma_i)$ maps generate the family; $(U\otimes U)|\Psi^-\rangle=|\Psi^-\rangle$ for six random SU(2) matrices while $|\Phi^+\rangle$ is not invariant; $\hat{\mathbf S}_{\text{tot}}^2|\Psi^-\rangle=0$ ; and $\langle\sigma_{\hat n}\sigma_{\hat n}\rangle=-1$ on five different axes.

Limiting cases

Bell state	Correlation $\langle\sigma_z\sigma_z\rangle$	$\langle\sigma_x\sigma_x\rangle$	$\langle\sigma_y\sigma_y\rangle$	Symmetry
$\vert\Phi^+\rangle$	$+1$	$+1$	$-1$	symmetric
$\vert\Phi^-\rangle$	$+1$	$-1$	$+1$	symmetric
$\vert\Psi^+\rangle$	$-1$	$+1$	$+1$	symmetric
$\vert\Psi^-\rangle$ (singlet)	$-1$	$-1$	$-1$	antisymmetric, $U\otimes U$ -invariant

Only the singlet is anti-correlated on all three axes — the fingerprint of a rotational scalar. The three triplets each fail on exactly one axis, which is how a Bell measurement distinguishes them.

Common trap“A Bell measurement is just measuring both qubits.”

It is not, and this is the operational heart of the subject. Measuring $A$ and then $B$ in the computational basis destroys the very superposition that defines a Bell state — you would collapse $|\Phi^+\rangle$ to $|00\rangle$ or $|11\rangle$ and never learn the relative sign or which of the four states you had. A Bell measurement is a single, joint, entangling measurement that projects onto the four-dimensional Bell basis at once. In hardware it is realized by first disentangling — apply a CNOT then a Hadamard, which maps the Bell basis onto the computational basis — and only then measuring locally. Skip the CNOT and you are not doing a Bell measurement; you are doing a product measurement that throws the correlation information away.

Worked examples

1 · Building and checking a Bell state (confidence builder)

Enumerate the standard preparation circuit: start from $|00\rangle$ , apply a Hadamard to qubit $A$ , then a CNOT with $A$ controlling $B$ . Apply it: the Hadamard gives $(|0\rangle+|1\rangle)/\sqrt2\otimes|0\rangle=(|00\rangle+|10\rangle)/\sqrt2$ ; the CNOT flips $B$ whenever $A=1$ , sending $|10\rangle\to|11\rangle$ , so $(|00\rangle+|11\rangle)/\sqrt2=|\Phi^+\rangle$ . Boundary number: verify orthonormality — the sixteen inner products among $|\Phi^\pm\rangle,|\Psi^\pm\rangle$ form the $4\times4$ identity exactly, and $\sum_B|B\rangle\langle B|=I_4$ . So the two-gate circuit $(\text{CNOT})(H\otimes I)$ turns a trivial product state into a maximally entangled one, and the four outputs (one per computational input) are a complete basis. This is the standard entangler at the front of every teleportation and dense-coding protocol.

2 · Singlet anti-correlation in every basis (exam level)

Take the singlet $|\Psi^-\rangle$ and measure both spins along an axis $\hat n(\theta,\phi)$ tilted away from $z$ . Enumerate the joint observable $\sigma_{\hat n}\otimes\sigma_{\hat n}$ with $\sigma_{\hat n}=\sin\theta\cos\phi\,\sigma_x+\sin\theta\sin\phi\,\sigma_y+\cos\theta\,\sigma_z$ . Apply: the expectation $\langle\Psi^-|\sigma_{\hat n}\otimes\sigma_{\hat n}|\Psi^-\rangle$ comes out to $-1$ for every choice of $\theta,\phi$ — we checked $\hat z$ , $\hat x$ , $\hat y$ , and two skew axes $(0.7,1.2)$ and $(2.0,3.0)$ radians, all giving exactly $-1$ . Boundary number: contrast $|\Phi^+\rangle$ , which gives $\langle\sigma_z\sigma_z\rangle=+1$ , $\langle\sigma_x\sigma_x\rangle=+1$ , but $\langle\sigma_y\sigma_y\rangle=-1$ — it fails to be anti-correlated on the $x$ and $z$ axes and flips sign on $y$ . Only the singlet is perfectly anti-correlated everywhere, precisely because it is $U\otimes U$ -invariant. This axis-independence is the property EPR seized on: whatever axis Alice picks, Bob's outcome is fixed with certainty, so it "must be real."

Going further

The Bell states are the two-qubit maximally-entangled reference points; their local-Pauli structure is the algebra behind superdense coding (two classical bits per qubit) and teleportation (Bennett et al. 1992–93), whose full protocols live in the quantum-information track. Their perfect correlations feed straight into The EPR Paradox and Local Realism (the singlet's axis-independent anti-correlation as "elements of reality"), and the inequality that turns those correlations into a refutation of local realism — together with its maximal quantum value $2\sqrt2$ (Tsirelson) — is proved in Bell's Theorem and the CHSH Inequality, with the experiments in Experimental Tests and Violation of Bell Inequalities. The three-party analogue, where a single measurement run refutes local realism, is The GHZ State and All-or-Nothing Nonlocality. The reduced-state calculation ( $\rho_A=I/2$ ) is from The Reduced Density Matrix and Partial Trace, and the entropy $\ln2$ from Entanglement Entropy and Measures of Entanglement.

Are the Bell states the only maximally entangled two-qubit states?

No — any state local-unitarily equivalent to a Bell state is also maximally entangled. The Bell basis is a convenient orthonormal choice of maximally entangled states, but $(U_A\otimes U_B)|\Phi^+\rangle$ is equally maximal for any local unitaries. What is special about the Bell basis is that its four members are mutually orthogonal, which is what lets them serve as a measurement basis and encode two bits.

Why does the singlet get all three ⟨σσ⟩ negative while the triplets don't?

Because the singlet is a rotational scalar and the triplets are the three components of a vector. A scalar looks identical along every axis, so its (anti)correlation is axis-independent — all $-1$ . Each triplet, being a vector component, singles out a direction and therefore correlates differently on different axes. Concretely $|\Psi^+\rangle$ gives $-1$ on $z$ but $+1$ on $x$ and $y$ ; only $|\Psi^-\rangle$ gives $-1$ everywhere.

How do Bell states enable teleportation in one sentence of physics?

Alice performs a Bell measurement jointly on the unknown qubit and her half of a shared Bell pair; this projects Bob's half into the unknown state up to one of four Pauli corrections, and the two classical bits of her outcome tell Bob which Pauli to undo. The entanglement supplies the correlated "channel," and the local-Pauli structure of the Bell basis is exactly why two classical bits suffice — the full protocol is developed in the quantum-information track.

Is a Bell state the same as an EPR pair?

Yes, colloquially. "EPR pair" is the physicist's older name for a maximally entangled two-qubit state, after the Einstein–Podolsky–Rosen argument (see The EPR Paradox and Local Realism), which used the singlet's perfect correlations. In quantum-information language "Bell pair" and "one ebit" are the standard terms for the same resource.

Do Bell states let Alice signal Bob faster than light?

No. Bob's reduced state is $I/2$ no matter what Alice does to her qubit — the no-signalling theorem proved in The Reduced Density Matrix and Partial Trace. Alice's measurement steers the conditional state Bob would find given her outcome, but she cannot control which outcome she gets, so the unconditioned state Bob holds is unchanged. Teleportation still needs the two classical bits sent at light speed; the entanglement alone transmits nothing.

Can more than two qubits share a single Bell-like state?

Not in the Bell sense — Bell states are strictly bipartite. Three or more qubits have genuinely multipartite entangled states (GHZ, W) that are not related by local unitaries and have no single "maximally entangled" representative; this is exactly the Schmidt-rank obstruction noted in The Schmidt Decomposition. The three-qubit GHZ state and its all-or-nothing nonlocality are treated in The GHZ State and All-or-Nothing Nonlocality.