Bell States and EPR Pairs

A Bell state, or EPR pair, is a maximally entangled state of two qubits — the smallest system in which quantum correlations become impossible to explain by any story of pre-existing local properties. There are exactly four of them, and together they form an orthonormal basis of the two-qubit space $\mathbb{C}^2\otimes\mathbb{C}^2$ : the Bell basis. Each is built from a product state by a single Hadamard followed by a CNOT, each has a reduced density matrix that is exactly maximally mixed, and each carries precisely one ebit — one unit of bipartite entanglement. They are, in a real operational sense, the currency in which entanglement is counted. Historically they crystallize the 1935 argument of Einstein, Podolsky, and Rosen, recast in spin- $\tfrac12$ form by David Bohm: a pair in the singlet $|\Psi^-\rangle$ is perfectly anticorrelated along every common axis, yet each half, examined alone, is random. EPR read this as evidence that quantum mechanics is incomplete — that the outcomes must be fixed in advance by variables the theory omits. Three decades later John Stewart Bell turned that philosophical unease into an inequality, and the Bell states turned out to be exactly the resource that violates it: they reach Tsirelson's bound $2\sqrt2$ for the CHSH quantity, the maximum any quantum system can achieve. Algebraically the four states are the simultaneous eigenstates of the two commuting stabilizers $X\otimes X$ and $Z\otimes Z$ , whose four $\pm1$ sign patterns label the basis. That single fact ties them at once to error correction (a Bell pair is the simplest stabilizer code state), to Bell measurement (reading two parity bits), and to the protocols — teleportation, superdense coding, entanglement swapping — that make entanglement a usable resource rather than a curiosity. This article defines the four states, derives their structure and correlations, and traces the EPR argument to its experimental resolution.

Key takeaways

The four Bell states $|\Phi^\pm\rangle,|\Psi^\pm\rangle$ form an orthonormal basis of the two-qubit space $\mathbb{C}^2\otimes\mathbb{C}^2$ , each one maximally entangled and each obtained from $|00\rangle$ by a Hadamard then a CNOT.
Every Bell state has Schmidt rank 2 with equal coefficients $1/\sqrt2$ , so the reduced state of either qubit is maximally mixed, $\rho_A=I/2$ , and the entanglement entropy is exactly one ebit.
They are interconverted by local Pauli gates on a single qubit: $I,Z,X,XZ$ carry $|\Phi^+\rangle$ to the other three, which is why two classical bits suffice to select any of the four.
The four states are the simultaneous $\pm1$ eigenstates of the commuting stabilizers $X\otimes X$ and $Z\otimes Z$ ; the four sign patterns are a complete label for the Bell basis.
The singlet obeys $E(\hat a,\hat b)=-\hat a\cdot\hat b$ , a smooth $-\cos\theta$ correlation that reaches the Tsirelson bound $2\sqrt2$ for CHSH — beyond the classical limit of 2 but below the algebraic maximum of 4.
Bell pairs are the canonical resource for teleportation, superdense coding, entanglement swapping, and device-independent cryptography; a Bell measurement is their reading device.
Key idea: The Bell states are the four maximally entangled two-qubit states that diagonalize the commuting operators $X\otimes X$ and $Z\otimes Z$ ; each carries exactly one ebit of entanglement, they are converted into one another by local Pauli gates, and their correlations violate every local hidden-variable bound.

The core insight

The Bell states are the four maximally entangled two-qubit states that diagonalize the commuting operators $X\otimes X$ and $Z\otimes Z$ ; each carries exactly one ebit of entanglement, they are converted into one another by local Pauli gates, and their $-\cos\theta$ correlations violate every local hidden-variable bound.

Where this lives today

Bell pairs stopped being a thought experiment decades ago; they are now generated by the millions and measured to close the last logical loopholes. In 2015 three groups — Hensen and colleagues in Delft using nitrogen-vacancy spins linked by entanglement swapping, and the Giustina and Shalm photonic experiments — reported the first loophole-free Bell tests, closing the locality and detection loopholes simultaneously and ruling out local realism at high confidence. That program earned Aspect, Clauser, and Zeilinger the 2022 Nobel Prize in Physics. Beyond foundations, Bell pairs are the working substrate of quantum technology: teleportation and entanglement swapping chain them into quantum repeaters that fight photon loss over fibre; the Micius satellite distributed entangled pairs to ground stations more than 1200 km apart, enabling intercontinental entanglement-based key distribution; and device-independent QKD and randomness expansion use the degree of CHSH violation itself as a security certificate — the closer to $2\sqrt2$ , the less an eavesdropper can know. In the lab, near-perfect Bell states are heralded routinely in trapped ions, superconducting qubits, and photonic chips, and their fidelity to the ideal is a standard benchmark for any new platform. The open frontier is scaling: building repeater networks and a fault-tolerant "quantum internet" whose elementary link is, still, a single shared ebit.

The physical picture

Think of a Bell pair as a single shared coin that has not yet landed. Neither half holds a definite face, yet whenever Alice and Bob inspect the same "side" they always agree — or, for the singlet, always disagree. The agreement is perfectly real and it is present in the correlations at the instant of measurement, but no usable signal travels: each local view, taken by itself, is indistinguishable from pure noise. Everything interesting lives not in either qubit but in the relationship between them. The cleanest way to see a Bell state come into existence is to build one. Start with the plain product state $|00\rangle$ . A Hadamard on the first qubit spreads it into a superposition, $(|0\rangle+|1\rangle)|0\rangle/\sqrt2$ . Then a CNOT — flip the second qubit exactly when the first is $|1\rangle$ — ties their fates together, producing $(|00\rangle+|11\rangle)/\sqrt2=|\Phi^+\rangle$ . Entanglement is born at the CNOT: before it the two qubits factor, after it they do not. The first figure traces this one-Hadamard-one-CNOT circuit and records what the resulting state is stabilized by.

Figure

The entangling circuit — a Hadamard on the top qubit then a CNOT turns the product state |00⟩ into the Bell state |Φ⁺⟩ = (|00⟩ + |11⟩)/√2, the +1 eigenstate of both X⊗X and Z⊗Z, carrying one ebit with reduced state ρ_A = I/2

The four Bell states differ only by which of two commuting parities they carry. Because $X\otimes X$ and $Z\otimes Z$ commute, both can be sharp at once even though neither qubit has a definite $X$ or $Z$ value on its own; the four sign combinations $(\pm1,\pm1)$ index the four states. This is the stabilizer viewpoint, and it is the bridge to quantum error correction. The singlet $|\Psi^-\rangle$ is the exception among the four: it is invariant under $U\otimes U$ for every single-qubit unitary, so its anticorrelation holds along all axes at once. The second figure shows what that rotational invariance buys — the smooth $-\cos\theta$ correlation curve that no local model can match, and the gap that Bell's theorem exploits.

Figure

The singlet correlation E(a,b) = −cos θ as a function of the angle θ between Alice's and Bob's axes: perfect anticorrelation −1 at θ = 0, zero at 90°, perfect correlation +1 at 180°. The best any local hidden-variable model can do is the straight amber line; the gap is where the CHSH quantity climbs from 2 to the Tsirelson bound 2√2 ≈ 2.828

The core equations

The Bell states are usually just written down, but the definition already contains everything: the pattern of signs and swapped labels encodes the entanglement, the orthonormality, and the parity structure all at once. We begin there and unpack the consequences.

\begin{aligned}|\Phi^\pm\rangle&=\tfrac{1}{\sqrt2}(|00\rangle\pm|11\rangle)\\ |\Psi^\pm\rangle&=\tfrac{1}{\sqrt2}(|01\rangle\pm|10\rangle)\end{aligned} \qquad (1)

where: $|0\rangle,|1\rangle$ — computational basis of a single qubit; $|ab\rangle$ — shorthand for $|a\rangle_A\otimes|b\rangle_B$ , qubit A then qubit B; $\tfrac{1}{\sqrt2}$ — normalization giving unit norm and equal Schmidt weights. Reading: the $\Phi$ states pair equal bits, the $\Psi$ states pair opposite bits; the superscript is the relative sign. The four are mutually orthogonal, $\langle B_i|B_j\rangle=\delta_{ij}$ , and span the whole two-qubit space. The most useful fact about the four states is that they are all one Hadamard-and-CNOT away from a product state. This is not just a recipe; it is why any physical platform that can do a controlled gate can prepare entanglement on demand.

|\Phi^+\rangle=\mathrm{CNOT}\,(H\otimes I)\,|00\rangle \qquad (2)

where: $H$ — Hadamard, $H|0\rangle=(|0\rangle+|1\rangle)/\sqrt2$ ; $\mathrm{CNOT}$ — controlled-NOT with qubit 1 as control, qubit 2 as target; $I$ — identity on the target during the Hadamard. Reading: feeding the other three computational inputs $|01\rangle,|10\rangle,|11\rangle$ through the same circuit produces $|\Psi^+\rangle,|\Phi^-\rangle,|\Psi^-\rangle$ respectively, so the circuit is a unitary map from the computational basis onto the Bell basis. Once you hold $|\Phi^+\rangle$ , the other three are a single local gate away. This is the algebraic heart of teleportation and superdense coding: a two-bit classical label $(z,x)$ names a Pauli, and that Pauli names a Bell state.

|\Phi^-\rangle=(Z\otimes I)|\Phi^+\rangle,\quad |\Psi^+\rangle=(X\otimes I)|\Phi^+\rangle,\quad |\Psi^-\rangle=(XZ\otimes I)|\Phi^+\rangle \qquad (3)

where: $X,Z$ — Pauli operators on qubit A; $XZ=-iY$ up to global phase. Reading: $Z$ flips the relative sign, $X$ swaps the paired bits, and their product does both, taking $|\Phi^+\rangle$ to the singlet. All four Bell states lie in one local-Pauli orbit — no entangling gate is needed to move between them. The signature of maximal entanglement is what you find when you throw one qubit away. Tracing out either party leaves no information at all: the local state is the maximally mixed $I/2$ .

\rho_A=\mathrm{Tr}_B\,|\Phi^+\rangle\langle\Phi^+|=\tfrac12 I,\qquad S(\rho_A)=1\ \text{ebit} \qquad (4)

where: $\mathrm{Tr}_B$ — partial trace over qubit B; $I$ — the $2\times2$ identity, so $\rho_A$ has eigenvalues $\tfrac12,\tfrac12$ ; $S(\rho_A)=-\mathrm{Tr}(\rho_A\log_2\rho_A)$ — the von Neumann entropy, equal to one ebit. Reading: a local measurement on one qubit is a fair coin, carrying zero information about the global state — entanglement lives in the correlations, never in the marginals. The four states are told apart by two commuting observables measured together. This parity picture is what makes a Bell measurement possible and what connects Bell pairs to stabilizer codes.

X\!\otimes\!X\,|\Phi^+\rangle=+|\Phi^+\rangle,\qquad Z\!\otimes\!Z\,|\Phi^+\rangle=+|\Phi^+\rangle,\qquad [X\!\otimes\!X,\,Z\!\otimes\!Z]=0 \qquad (5)

where: $X\otimes X$ — parity of the two qubits in the $X$ basis; $Z\otimes Z$ — parity in the $Z$ basis; the commutator vanishes because each pair of single-qubit Paulis anticommutes and the two sign flips cancel. Reading: the four eigenvalue patterns are $|\Phi^+\rangle\to(+,+)$ , $|\Phi^-\rangle\to(-,+)$ , $|\Psi^+\rangle\to(+,-)$ , $|\Psi^-\rangle\to(-,-)$ — a complete, non-redundant label. Finally, the correlation that started it all. For the singlet, the joint outcome of measuring spin along any two directions is fixed entirely by the angle between them.

E(\hat a,\hat b)=\langle\Psi^-|(\hat a\!\cdot\!\vec\sigma)\otimes(\hat b\!\cdot\!\vec\sigma)|\Psi^-\rangle=-\,\hat a\cdot\hat b=-\cos\theta_{ab} \qquad (6)

where: $\hat a,\hat b$ — Alice's and Bob's measurement axes; $\vec\sigma=(\sigma_x,\sigma_y,\sigma_z)$ ; $E$ — the average product of the two $\pm1$ outcomes; $\theta_{ab}$ — the angle between axes. Reading: at $\theta=0$ the outcomes are perfectly anticorrelated ( $E=-1$ ); at $\theta=180^\circ$ perfectly correlated; the smooth cosine in between is what no local hidden-variable model can reproduce and what the CHSH inequality measures.

Derivation AMaximal entanglement and one ebit of entropy

The goal is to prove rigorously that a Bell state is maximally entangled in the precise, quantitative sense used throughout quantum information: its Schmidt decomposition has two equal coefficients, its reduced density matrix is the maximally mixed state $I/2$ , and its entanglement entropy equals exactly one ebit — the largest value two qubits allow. We start from the definition, read off the Schmidt form by inspection, take the partial trace, and compute the von Neumann entropy, so that the single number "one ebit" is anchored to an explicit calculation rather than asserted.

Step 1. Write $|\Phi^+\rangle=\tfrac{1}{\sqrt2}|0\rangle_A|0\rangle_B+\tfrac{1}{\sqrt2}|1\rangle_A|1\rangle_B$ . This is already a Schmidt decomposition: $\{|0\rangle,|1\rangle\}$ is orthonormal on each side, and the coefficients are $\lambda_1=\lambda_2=1/\sqrt2$ . Why: a state is maximally entangled precisely when its Schmidt coefficients are all equal, so recognizing the equal $1/\sqrt2$ weights settles the question before any density matrix is formed.

Step 2. The reduced state follows from the Schmidt form directly, $\rho_A=\sum_i\lambda_i^2\,|i\rangle_A\langle i|=\tfrac12|0\rangle\langle0|+\tfrac12|1\rangle\langle1|=\tfrac12 I$ . Equivalently, expand $|\Phi^+\rangle\langle\Phi^+|$ into four terms and trace over B; the cross terms $|0\rangle\langle1|\,\langle1|0\rangle$ vanish because $\langle1|0\rangle=0$ , leaving the two diagonal terms. Why: equal squared Schmidt coefficients force the marginal to be proportional to the identity, the operational meaning of "one qubit alone tells you nothing".

Step 3. Compute the entropy of the eigenvalues $\tfrac12,\tfrac12$ : $S(\rho_A)=-\tfrac12\log_2\tfrac12-\tfrac12\log_2\tfrac12=1$ . This is the Shannon entropy of a fair coin, the maximum for a single qubit. Why: the entanglement entropy of a pure bipartite state is the entropy of its reduced state, so this one bit is the entanglement, by definition, of the Bell pair — one ebit.

Step 4. The same argument applies to all four Bell states, since each is a local unitary image of $|\Phi^+\rangle$ and local unitaries leave the Schmidt coefficients unchanged. Why: entanglement is invariant under local operations, so all four states sit at the identical maximum — there is no "more entangled" two-qubit pure state. *Verified: eigenvalues of $\rho_A$ are $1/2,1/2$ ; $-\tfrac12\log_2\tfrac12-\tfrac12\log_2\tfrac12=1$ recomputed numerically (Nielsen & Chuang §11.3, §12.1).*

Go deeper · Derivation B · 1 minEPR perfect anticorrelation along every axis

Derivation BEPR perfect anticorrelation along every axis

The goal is to establish the property at the very center of the Einstein–Podolsky–Rosen argument: the singlet $|\Psi^-\rangle$ is perfectly anticorrelated not just along the computational axis but along every common measurement direction $\hat n$ , so that whatever Alice finds, Bob is certain to find the opposite. We prove it twice over — first by direct action of $\sigma_z\otimes\sigma_z$ , then by exploiting the singlet's $SU(2)$ rotational invariance to lift the result to an arbitrary axis — which is exactly the symmetry that makes the correlation seem to demand pre-existing values.

Step 1. Act with $\sigma_z\otimes\sigma_z$ on the singlet: $(\sigma_z\otimes\sigma_z)\tfrac{1}{\sqrt2}(|01\rangle-|10\rangle)=\tfrac{1}{\sqrt2}(-|01\rangle+|10\rangle)=-|\Psi^-\rangle$ , using $\sigma_z|0\rangle=|0\rangle$ , $\sigma_z|1\rangle=-|1\rangle$ . So $|\Psi^-\rangle$ is a $-1$ eigenstate and $E(\hat z,\hat z)=-1$ . Why: a single explicit eigenvalue computation pins the anticorrelation along one concrete axis, giving a foothold before invoking symmetry.

Step 2. Show the singlet is rotationally invariant: for any $U\in SU(2)$ , $(U\otimes U)|\Psi^-\rangle=(\det U)\,|\Psi^-\rangle=|\Psi^-\rangle$ , because $\det U=1$ on $SU(2)$ and the antisymmetric combination transforms by the determinant. Why: this invariance is the hallmark of a total-spin-zero state, and it means no direction is special — the physics along $\hat z$ must hold along every axis.

Step 3. Rotate the observable: any $\hat n\cdot\vec\sigma=U^\dagger\sigma_z U$ for the $U$ that carries $\hat z$ into $\hat n$ . Then $E(\hat n,\hat n)=\langle\Psi^-|(U^\dagger\sigma_zU)\otimes(U^\dagger\sigma_zU)|\Psi^-\rangle$ . Slide the $U$ 's onto the state using invariance: $(U\otimes U)|\Psi^-\rangle=|\Psi^-\rangle$ , so the expression collapses to $\langle\Psi^-|\sigma_z\otimes\sigma_z|\Psi^-\rangle=-1$ . Why: the invariance converts a hard "arbitrary axis" calculation into the trivial $\hat z$ case, proving $E(\hat n,\hat n)=-1$ for all $\hat n$ at once.

Step 4. For unequal axes the same manipulation gives $E(\hat a,\hat b)=-\hat a\cdot\hat b=-\cos\theta_{ab}$ , the full correlation function. It is this smooth cosine — not merely the $\theta=0$ anticorrelation — that no local model can reproduce, as Bell's inequality makes quantitative. Why: EPR saw only the perfect $\theta=0$ correlations and concluded incompleteness; Bell saw the whole curve and concluded nonlocality. *Verified: $\sigma_z\otimes\sigma_z\,|\Psi^-\rangle=-|\Psi^-\rangle$ by direct action; $SU(2)$ invariance from $\det U=1$ ; $E(\hat a,\hat b)=-\hat a\cdot\hat b$ reproduced for sampled angles (Preskill Ch. 4; Sakurai, angular momentum).*

Go deeper · Derivation C · 1 minFrom the singlet cosine to the Tsirelson bound

Derivation CFrom the singlet cosine to the Tsirelson bound

The goal is to carry the correlation function $E(\hat a,\hat b)=-\cos\theta$ into the CHSH combination and show that a Bell state reaches the value $2\sqrt2$ — Tsirelson's bound — decisively above the local-realist limit of 2 and below the algebraic maximum of 4. This is the quantitative statement of "Bell states are the optimal nonlocal resource": we assemble the CHSH quantity from four correlators, choose the standard optimal angles, and evaluate, so the famous $2\sqrt2$ appears as a plain arithmetic result.

Step 1. Define the CHSH quantity for two settings per party, $S=E(\hat a,\hat b)-E(\hat a,\hat b')+E(\hat a',\hat b)+E(\hat a',\hat b')$ . Any local hidden-variable model assigns each measurement a predetermined $\pm1$ ; a short case check on these values gives $|S|\le2$ . Why: this classical bound is the target the quantum state must beat, and it is the inequality Bell's theorem is built on.

Step 2. Use the state $|\Phi^+\rangle$ with axes in the $x$ – $z$ plane, for which $E(\hat a,\hat b)=\cos(\alpha-\beta)$ where $\alpha,\beta$ are the axis angles. Pick the optimal settings $\alpha=0^\circ,\ \alpha'=90^\circ$ for Alice and $\beta=45^\circ,\ \beta'=135^\circ$ for Bob. Why: these angles space the four correlators so that three add constructively and the subtracted one flips sign — the geometry that saturates the quantum bound.

Step 3. Evaluate the four correlators: $E(\hat a,\hat b)=\cos(-45^\circ)=\tfrac{1}{\sqrt2}$ , $E(\hat a,\hat b')=\cos(-135^\circ)=-\tfrac{1}{\sqrt2}$ , $E(\hat a',\hat b)=\cos(45^\circ)=\tfrac{1}{\sqrt2}$ , $E(\hat a',\hat b')=\cos(-45^\circ)=\tfrac{1}{\sqrt2}$ . The minus sign in front of the second term turns it positive, so $S=4\cdot\tfrac{1}{\sqrt2}=2\sqrt2\approx2.828$ . Why: all four contributions now have the same sign and magnitude, which is precisely what pushes $S$ to its quantum maximum.

Step 4. No quantum state does better: Tsirelson's theorem caps $S$ at $2\sqrt2$ for any quantum system, while a hypothetical super-quantum "PR box" would reach the algebraic $S=4$ . Bell states sit exactly at the quantum ceiling. Why: this places Bell states at the boundary of quantum theory itself — maximally nonlocal within what nature allows, yet still short of signalling. *Verified: $S=E(a,b)-E(a,b')+E(a',b)+E(a',b')=2.8284$ for angles $(0,90,45,135)$ , matching $2\sqrt2$ (Tsirelson 1980; Wilde Ch. 3).*

Limiting cases

The Bell states are defined by sitting at the extremes of every entanglement and correlation measure at once; the table reads off where each quantity lands.

Case	Result	Meaning
Maximal entropy	$S(\rho_A)=-\mathrm{Tr}(\rho_A\log_2\rho_A)=1\ \text{ebit}$	Reduced state $\rho_A=I/2$ is maximally mixed; one ebit is the ceiling for two qubits.
Maximal concurrence	$C(\|\Phi^+\rangle)=1$	Wootters' concurrence saturates its upper bound of 1 for every Bell state.
Perfect correlation	$E(\hat a,\hat a)=\mp1$	Matched axes give a certain outcome; $\|\Psi^-\rangle$ anticorrelates, $\|\Phi^+\rangle$ correlates.
Tsirelson bound	$\langle\mathcal{B}_{\mathrm{CHSH}}\rangle=2\sqrt2\approx2.828$	Optimal angles hit the maximal quantum CHSH value, above the classical 2, below the PR-box 4.
Schmidt rank	$\lambda_1=\lambda_2=\tfrac{1}{\sqrt2}$	Two equal Schmidt coefficients define maximal entanglement; a product state has rank 1.
Product-state limit	$\|\Phi^+\rangle\to\|00\rangle$ if one weight $\to0$	Unequal weights $\alpha\|00\rangle+\beta\|11\rangle$ are only partially entangled; equal weights are the special maximal point.

Common trapMaximal entanglement is not a box of pre-set answers

The perfect correlations of a Bell state tempt a fatal mental picture: that each qubit secretly "already knows" what it will give for every measurement, and the pair merely shares a matching list drawn in advance. That is exactly the local hidden-variable model Bell's theorem rules out. Two facts break the picture. First, each qubit alone is completely random, $\rho_A=I/2$ — there is no local record to read. Second, the correlation is the smooth curve $E(\hat a,\hat b)=-\cos\theta$ , and no assignment of pre-existing $\pm1$ values to every axis can reproduce a cosine; the best any such list achieves is the straight-line bound of the second figure, and the CHSH quantity it yields never exceeds 2, whereas the Bell state reaches $2\sqrt2$ . Note too the basis subtlety: only the singlet $|\Psi^-\rangle$ is rotationally invariant, so its anticorrelation holds on every axis; $|\Phi^\pm\rangle$ and $|\Psi^+\rangle$ correlate perfectly only in specific bases. Maximal entanglement is a statement about correlations that have no local explanation, not about synchronized hidden lists.

Worked examples

1 · A Bell state violates CHSH (fully worked)

Alice and Bob share $|\Phi^+\rangle$ and measure spin along axes in the $x$ – $z$ plane, for which $E(\hat a,\hat b)=\cos(\alpha-\beta)$ . Alice uses $\alpha=0^\circ$ or $\alpha'=90^\circ$ ; Bob uses $\beta=45^\circ$ or $\beta'=135^\circ$ . Compute the CHSH quantity and compare to the classical and Tsirelson bounds.

Step 1. Evaluate the four correlators from the angle differences. $E(\hat a,\hat b)=\cos(0-45^\circ)=\cos(-45^\circ)=\mathbf{+0.7071}$ . $E(\hat a,\hat b')=\cos(0-135^\circ)=\cos(-135^\circ)=\mathbf{-0.7071}$ . $E(\hat a',\hat b)=\cos(90-45^\circ)=\cos(45^\circ)=\mathbf{+0.7071}$ . $E(\hat a',\hat b')=\cos(90-135^\circ)=\cos(-45^\circ)=\mathbf{+0.7071}$ .

Step 2. Assemble $S=E(\hat a,\hat b)-E(\hat a,\hat b')+E(\hat a',\hat b)+E(\hat a',\hat b')$ . The subtracted term carries a minus sign, so the $-0.7071$ becomes $+0.7071$ and all four contributions align: $S=0.7071-(-0.7071)+0.7071+0.7071=4\times0.7071$ .

Step 3. Therefore $S=\mathbf{2.828}=\mathbf{2\sqrt2}$ . This exceeds the local-realist bound $|S|\le2$ by a wide margin — the correlations cannot come from any pre-shared classical list — and it lands exactly on Tsirelson's quantum ceiling, so no measurement choice or state could push it higher. A Bell state with these four angles gives S = 2√2 ≈ 2.828, the maximal quantum violation.

2 · Superdense coding: two classical bits down one qubit

Alice and Bob pre-share $|\Phi^+\rangle$ . Alice wants to send the two-bit message $z x = 10$ . Show how one local gate and one transmitted qubit deliver both bits, and how Bob decodes.

Step 1. Encode. The message $(z,x)=(1,0)$ selects the Pauli $Z^zX^x=Z^1X^0=Z$ . Alice applies $Z$ to her qubit only: $(Z\otimes I)|\Phi^+\rangle=|\Phi^-\rangle=\tfrac{1}{\sqrt2}(|00\rangle-|11\rangle)$ by equation (3). No entangling operation and no communication has happened yet.

Step 2. Transmit. Alice sends her single physical qubit to Bob. He now holds both halves and performs a Bell measurement — a CNOT then a Hadamard on the control, then two $Z$ readouts — which maps the Bell basis onto the computational basis: $|\Phi^+\rangle\to00,\ |\Phi^-\rangle\to10,\ |\Psi^+\rangle\to01,\ |\Psi^-\rangle\to11$ .

Step 3. Decode. Bob's readout is $\mathbf{10}$ , exactly Alice's message. One qubit carried two classical bits, using one pre-shared ebit — Holevo's single-qubit limit of one bit is beaten precisely because the entanglement was distributed in advance. Running the same map backwards is teleportation: there Alice measures in the Bell basis and sends the two bits, and Bob applies the matching Pauli $I,X,Z,XZ$ to reconstruct an unknown state.

3 · The marginal is a fair coin

Confirm by explicit partial trace that measuring one qubit of $|\Psi^-\rangle$ in the computational basis gives 0 and 1 with equal probability, carrying no information about the message encoded globally.

Step 2. So $\rho_A=\tfrac12|0\rangle\langle0|+\tfrac12|1\rangle\langle1|=\tfrac12 I$ , with $\Pr(0)=\Pr(1)=\tfrac12$ . Alice's local outcome is a fair coin regardless of which of the four Bell states she shares — which is exactly why superdense coding transmits nothing until Alice's qubit physically arrives, and why entanglement alone cannot signal faster than light.

Physics of the portrait

The Bell states carry Bell's name for a reason: they are the states that make his theorem bite. The portrait belongs to the man who turned a philosophical dispute about EPR into a quantitative, testable statement — and lost, gloriously, to the experiments.

Origins · 1 minJohn Stewart Bell and how this was discovered

John Stewart Bell

1928–1990 · quantum foundations and the impossibility of local hidden variables

John Stewart Bell was a Northern Irish physicist who spent his career at CERN designing particle accelerators, and did his revolutionary work on quantum foundations as, in his own words, a "hobby". Trained in Belfast and Birmingham, he was troubled from his student days by the measurement problem and by the way the Copenhagen orthodoxy discouraged asking what quantum states really are. In 1964, sparked by David Bohm's hidden-variable reconstruction of quantum mechanics, Bell asked whether any theory of local pre-existing values could reproduce the singlet's correlations. His answer — Bell's theorem — was a simple inequality that every local hidden-variable model must obey but that quantum mechanics, using precisely the entangled states now named after him, violates. He had converted Einstein, Podolsky, and Rosen's 1935 metaphysical complaint into an experiment. Clauser, Horne, Shimony, and Holt recast it into the testable CHSH form in 1969; Clauser and Freedman, then Aspect in the early 1980s, and finally the loophole-free experiments of 2015 all came down decisively against local realism. Bell died suddenly in 1990, the year he was rumoured to be a Nobel candidate; the 2022 physics Nobel to Aspect, Clauser, and Zeilinger honoured the program he began. He was also famously careful about what his theorem did not show — it does not permit signalling — and he coined "beable" for the elements of reality a sensible theory should posit.

History

1935

— Einstein, Podolsky, and Rosen argue quantum mechanics is incomplete

1935

— Schrödinger names the phenomenon entanglement and analyzes it

1951

— Bohm recasts the EPR argument in spin-half (qubit) form

1964

— Bell derives his inequality, making local hidden variables testable

1969

— Clauser, Horne, Shimony, and Holt give the experimentally usable CHSH form

Going further

Texts. Nielsen & Chuang, Quantum Computation and Quantum Information, §1.3.6, §2.3, §12.1 · Preskill, Lecture Notes on Quantum Computation, Ch. 4 · Watrous, The Theory of Quantum Information, Ch. 6 · Wilde, Quantum Information Theory, Ch. 3 · Mermin, Quantum Computer Science, Ch. 6 · Einstein, Podolsky & Rosen, Phys. Rev. 47, 777 (1935); Bell, Physics 1, 195 (1964); Bohm, Quantum Theory (1951).

Provenance. Bell-state definitions, orthonormality, and the entangling circuit are from Nielsen & Chuang §1.3.6 and §2.3; the Schmidt/entropy result ( $\rho_A=I/2$ , $S=1$ ebit) from §11.3 and §12.1. The singlet's $SU(2)$ invariance and the correlation $E(\hat a,\hat b)=-\hat a\cdot\hat b$ follow Preskill Ch. 4 and Bohm's spin formulation. The stabilizer eigenvalue table is verifiable by direct two-qubit matrix action; the Tsirelson bound $2\sqrt2$ is standard (Tsirelson 1980; Wilde Ch. 3). A verifier should recompute: the CNOT $(H\otimes I)|00\rangle$ circuit; the partial trace giving $I/2$ ; the $\pm1$ eigenvalues of $X\otimes X$ and $Z\otimes Z$ ; the $\sigma_z\otimes\sigma_z$ action on $|\Psi^-\rangle$ ; and the CHSH sum $S=2\sqrt2$ for angles $(0,90,45,135)$ — all confirmed numerically here.

Prerequisites. Separable versus Entangled States · Schmidt Decomposition. Leads to. CHSH Inequality and Its Quantum Violation · Bell's Theorem and Local Hidden Variables · Entanglement Entropy of Pure States · Concurrence and Wootters' Formula.

Ask the tutor: Ask me to walk through a full teleportation or superdense-coding protocol step by step, or to show exactly why a straight-line local model can never match the singlet's $-\cos\theta$ curve.

Why are there exactly four Bell states, not more or fewer?

Because the two-qubit space is four-dimensional, and the Bell states are an orthonormal basis for it. Any basis of $\mathbb{C}^2\otimes\mathbb{C}^2$ has four elements; what makes the Bell basis special is that all four are maximally entangled at once, which is unusual — most bases mix entangled and product states. The count four is also visible in the stabilizer picture: two commuting parities $X\otimes X$ and $Z\otimes Z$ each have two eigenvalues $\pm1$ , giving $2\times2=4$ sign patterns, one per Bell state. And it is the count that matters operationally: four outcomes carry exactly two classical bits, which is why teleportation needs two bits of feedforward and superdense coding sends two bits per qubit.

What is the difference between an EPR pair and a Bell state?

In modern usage they are the same object under two names. "EPR pair" is the historical term, from the 1935 Einstein–Podolsky–Rosen paper, and traditionally refers to the singlet $|\Psi^-\rangle$ — the spin-zero state Bohm used, with its perfect anticorrelation along every axis. "Bell state" is the information-theoretic term for any of the four maximally entangled two-qubit states $|\Phi^\pm\rangle,|\Psi^\pm\rangle$ . Since all four are related by local Pauli gates and carry identical entanglement, the distinction is cosmetic for most purposes; people say "an EPR pair" to evoke the foundations story and "a Bell state" to evoke the protocol or basis. When precision matters, name the state: $|\Phi^+\rangle$ is the usual textbook default.

Does the instantaneous correlation let Alice send Bob a message faster than light?

No, and the reason is sharp: Bob's qubit is described by the maximally mixed state $\rho_B=I/2$ whether or not Alice has measured, and whatever she chose to measure. His local statistics are completely independent of her actions, so he learns nothing until a classical message arrives through an ordinary sub-light channel. This is the "no-signalling" content of quantum mechanics, and it holds for every Bell state and every measurement. The correlations are real and, in the quantum description, established at measurement — but they are only visible once the two parties compare notes, and that comparison travels no faster than light. Entanglement is a correlation resource, not a communication channel.

Why is the singlet Ψ⁻ singled out as the special one of the four?

Because it is the only one that is rotationally invariant: $(U\otimes U)|\Psi^-\rangle=|\Psi^-\rangle$ for every single-qubit unitary $U$ , since it is the total-spin-zero state and $\det U=1$ on $SU(2)$ . The practical consequence is that its perfect anticorrelation holds along every common measurement axis, not just the computational one — measure both spins along any shared direction and they always disagree. The other three Bell states correlate perfectly only in specific bases: $|\Phi^+\rangle$ correlates in $Z$ but its behaviour in $X$ differs. This invariance is also why the singlet is the natural state for the original EPR/Bohm argument and for many Bell tests, where you want axis-independent correlations.

How do you actually measure which Bell state you have?

You cannot read the four states apart with local measurements alone — the Bell projectors are entangled operators, so a Bell measurement is inherently non-local. The standard trick is to rotate the states rather than the detector: apply a CNOT (control on qubit A) followed by a Hadamard on qubit A, which is exactly the inverse of the preparation circuit and maps the Bell basis onto the computational basis. Then two ordinary single-qubit $Z$ -measurements read out the label: $|\Phi^+\rangle\to00$ , $|\Phi^-\rangle\to10$ , $|\Psi^+\rangle\to01$ , $|\Psi^-\rangle\to11$ . The two readout bits are the $Z\otimes Z$ and $X\otimes X$ parities. With photons and passive linear optics, though, you can distinguish at most two of the four without ancillas — a hard $50\%$ ceiling that shapes real quantum-network design.

Is a Bell state the most entangled state that exists?

It is the most entangled state of two qubits — no two-qubit pure state carries more than one ebit, and every Bell state saturates that bound. For larger systems there are more entangled states in the sense of carrying more total entanglement: two qutrits can hold $\log_2 3$ ebits, and $n$ ebits live in $2n$ qubits arranged as $n$ parallel Bell pairs. There are also multipartite states, like GHZ and W states on three or more qubits, whose entanglement is genuinely many-body and not just a collection of pairs — these are not simply "more Bell states". But as the unit of bipartite entanglement, the Bell pair is the reference against which all of these are measured: entanglement is quite literally counted in ebits, and one Bell pair is one ebit.