Skip to content

Entanglement in Quantum Information

Entanglement becomes a resource when a task specifies who can act on which systems, what communication is available, and what must be produced. A Bell pair enables teleportation together with a classical message; it cannot transmit a controllable message by itself. The same distinction between joint states and local information connects entanglement to coding, computation, and networks.

Helpful background. Bell states supplies the basic shared pair, partial trace predicts local observations, and LOCC identifies operations available to separated laboratories. Information-theoretic foundations develops the operational definitions used in full protocols.

A qubit is a two-dimensional quantum system with Hilbert space C2\mathbb C^2. An nn-qubit register has

Hn=(C2)⊗n.\mathcal H_n = (\mathbb C^2)^{\otimes n}.

The computational basis is

{∣b1b2⋯bn⟩:bj∈{0,1}}.\{ \lvert b_1b_2\cdots b_n\rangle : b_j\in\{0,1\} \}.

This tensor-product structure is what makes entanglement possible. A two-qubit pure state

∣ψ⟩=∑i,j=01Cij∣ij⟩\lvert\psi\rangle = \sum_{i,j=0}^{1} C_{ij}\lvert ij\rangle

is a product state exactly when the coefficient matrix has rank one. Otherwise it is entangled.

Quantum information often packages this same structure into circuits. A one-qubit gate is local relative to a chosen register split. A two-qubit gate can be local or nonlocal depending on which subsystems it couples. Entangling gates are valuable because local operations alone cannot create entanglement from product inputs.

The standard Bell pair

∣Φ+⟩=12(∣00⟩+∣11⟩)\lvert\Phi^+\rangle = \frac{1}{\sqrt2} \left( \lvert00\rangle+\lvert11\rangle \right)

has maximally mixed one-qubit reductions:

ρA=ρB=12I.\rho_A = \rho_B = \frac12 I.

Its entanglement entropy is one bit:

S(ρA)=1bit.S(\rho_A) = 1 \quad \text{bit}.

One maximally entangled two-qubit pair is called one ebit. The ebit is a unit of bipartite pure-state entanglement, much as a bit is a unit of classical information. It is not a new particle or a new observable; it is a resource unit defined relative to a bipartite tensor-product split.

Bell pairs are the cleanest examples, but quantum information uses many other entangled states: partially entangled pairs, GHZ states, graph states, stabilizer states, cluster states, continuous-variable squeezed states, and noisy mixed states. The right state depends on the protocol and noise model.

Quantum Teleportation uses shared entanglement plus classical communication to simulate the transfer of an unknown qubit state. In resource shorthand,

1 ebit+2 classical bits⟶1 qubit transmission.1\ \text{ebit} + 2\ \text{classical bits} \longrightarrow 1\ \text{qubit transmission}.

This notation is not an equation of physical particles. It is resource accounting: if Alice and Bob already share one Bell pair, then Alice can perform a Bell-basis measurement, send two classical bits, and Bob can apply a Pauli correction to recover the input state.

Before Alice’s message, Bob’s state is the average over the four equally likely corrections:

14∑P∈{I,X,Y,Z}PρP†=I2.\frac14\sum_{P\in\{I,X,Y,Z\}}P\rho P^\dagger=\frac I2.

It is independent of the input ρ\rho, not merely an unknown version Bob could learn by measuring. The classical message lets him apply the outcome-dependent correction. It therefore carries an essential causal part of the protocol.

The teleportation derivation proves the Bell-basis identity and checks that the protocol also preserves correlations with an external reference.

Superdense Coding reverses the resource viewpoint. If Alice and Bob already share one Bell pair, then Alice can encode two classical bits by applying one of four local Pauli operations to her qubit and sending that qubit to Bob:

1 ebit+1 qubit transmission⟶2 classical bits.1\ \text{ebit} + 1\ \text{qubit transmission} \longrightarrow 2\ \text{classical bits}.

The algebra is the local-Pauli labeling of Bell states:

Alice’s local operation Resulting Bell state, up to global phase
II ∣Φ+⟩\lvert\Phi^+\rangle
ZZ ∣Φ−⟩\lvert\Phi^-\rangle
XX ∣Ψ+⟩\lvert\Psi^+\rangle
XZXZ ∣Ψ−⟩\lvert\Psi^-\rangle

Bob can distinguish the four Bell states by a Bell-basis measurement after receiving Alice’s qubit.

The resource accounting includes the shared Bell pair prepared in advance. Dense coding does not say that an isolated qubit always carries two classical bits. The protocol derivation explains how the pre-shared correlations and the transmitted system work together.

An entangling gate is a unitary operation that can turn at least one product input into an entangled output. A standard example is Hadamard followed by CNOT:

(H⊗I)∣00⟩=12(∣00⟩+∣10⟩),(H\otimes I)\lvert00\rangle = \frac{1}{\sqrt2} \left( \lvert00\rangle+\lvert10\rangle \right),

and then

CNOT⁡(H⊗I)∣00⟩=12(∣00⟩+∣11⟩)=∣Φ+⟩.\operatorname{CNOT} (H\otimes I)\lvert00\rangle = \frac{1}{\sqrt2} \left( \lvert00\rangle+\lvert11\rangle \right) = \lvert\Phi^+\rangle.

An entangling gate need not entangle every input: CNOT leaves ∣00⟩\lvert00\rangle unchanged, and also leaves ∣+⟩∣+⟩\lvert+\rangle\lvert+\rangle product because X∣+⟩=∣+⟩X\lvert+\rangle=\lvert+\rangle. Conversely, SWAP is not a product of one-qubit gates but maps every product vector to a product vector. “Acts on two qubits” and “creates entanglement” are different conditions.

Local one-qubit gates can rotate, rephase, or relabel entanglement, but they cannot create it from a product state. Nonlocal gates, interactions, or measurements with appropriate conditioning are needed to generate entanglement between registers.

This distinction is central in circuit models, measurement-based computation, and hardware design. Entanglement is not the only ingredient in quantum computational advantage, but it is one of the structural resources that separates genuinely multiparty quantum evolution from independent one-qubit dynamics.

Quantum error correction can be understood through entanglement with a reference system. Imagine a logical system LL entangled with an inaccessible reference RR. A good code protects the information in LL by preserving the joint correlations between RR and the encoded system despite noise.

This viewpoint avoids a common misconception: a code is not just preserving a list of possible basis states. It must preserve arbitrary superpositions, and therefore it must preserve entanglement with external systems. If a noise process destroys the entanglement between the encoded system and a reference, it has destroyed quantum information even if some classical labels remain recoverable.

For example, complete dephasing preserves both basis states ∣0⟩\lvert0\rangle and ∣1⟩\lvert1\rangle, but acting on half of ∣Φ+⟩RL\lvert\Phi^+\rangle_{RL} gives

ρRL′=12(∣00⟩⟨00∣+∣11⟩⟨11∣).\rho_{RL}'=\frac12(\lvert00\rangle\langle00\rvert+\lvert11\rangle\langle11\rvert).

It is separable and has Bell-state fidelity ⟨Φ+∣ρRL′∣Φ+⟩=1/2\langle\Phi^+|\rho_{RL}'|\Phi^+\rangle=1/2. Protecting classical basis labels was not enough. An exact recovery channel must restore every encoded state, including its correlations with an untouched reference.

Realistic shared states are noisy. A mixed state may be entangled but not close to a Bell pair. Entanglement distillation asks whether many noisy shared copies can be transformed by LOCC into fewer high-fidelity Bell pairs.

Schematically,

ρAB⊗n→ LOCC⁡ ∣Φ+⟩⊗mapproximately.\rho_{AB}^{\otimes n} \xrightarrow{\ \operatorname{LOCC}\ } \lvert\Phi^+\rangle^{\otimes m} \quad \text{approximately}.

The rate m/nm/n depends on the state and on the allowed protocols. Some entangled mixed states are not distillable under standard assumptions; these are called bound entangled states. This is one reason mixed-state entanglement is much richer than pure-state Schmidt entropy.

LOCC cannot create entanglement from separable inputs, but it can concentrate pre-existing entanglement into fewer pairs. A successful conditional branch must be considered together with its success probability and resource consumption. Entanglement distillation develops recurrence and hashing protocols with those costs explicit.

Quantum networks distribute entanglement between spatially separated nodes. The basic operations include:

  • preparing entangled links;
  • storing quantum states in memories;
  • performing local measurements at intermediate nodes;
  • sending classical heralding signals;
  • applying local corrections conditioned on those signals.

Entanglement Swapping is the simplest network primitive. Suppose AA is entangled with BB, and CC is entangled with DD:

∣Φ+⟩AB⊗∣Φ+⟩CD.\lvert\Phi^+\rangle_{AB} \otimes \lvert\Phi^+\rangle_{CD}.

A Bell-basis measurement on B,CB,C can conditionally prepare an entangled state of A,DA,D, even though AA and DD did not directly interact. The conditioning information is classical, so again there is no faster-than-light signaling.

If the Bell result is discarded in this example, the endpoint state is IA/2⊗ID/2I_A/2\otimes I_D/2. Conditioning is what identifies the entangled endpoint branch. Entanglement swapping develops the correction rules and success probabilities, including limitations of particular measurement devices.

  • C. H. Bennett, G. Brassard, C. Crépeau, R. Jozsa, A. Peres, and W. K. Wootters, “Teleporting an unknown quantum state via dual classical and Einstein–Podolsky–Rosen channels,” Physical Review Letters 70, 1895–1899 (1993), doi:10.1103/PhysRevLett.70.1895.
  • C. H. Bennett, D. P. DiVincenzo, J. A. Smolin, and W. K. Wootters, “Mixed-state entanglement and quantum error correction,” Physical Review A 54, 3824–3851 (1996), doi:10.1103/PhysRevA.54.3824.
  • C. H. Bennett and S. J. Wiesner, “Communication via one- and two-particle operators on Einstein–Podolsky–Rosen states,” Physical Review Letters 69, 2881–2884 (1992), doi:10.1103/PhysRevLett.69.2881.
  • R. Horodecki, P. Horodecki, M. Horodecki, and K. Horodecki, “Quantum entanglement,” Reviews of Modern Physics 81, 865–942 (2009), doi:10.1103/RevModPhys.81.865.
  • J. Watrous, The Theory of Quantum Information, Cambridge University Press (2018), author’s text.
  1. Bell-pair generation. Show that applying HH to the first qubit of ∣00⟩\lvert00\rangle and then CNOT gives ∣Φ+⟩\lvert\Phi^+\rangle.
Solution

First,

(H⊗I)∣00⟩=12(∣00⟩+∣10⟩).(H\otimes I)\lvert00\rangle = \frac{1}{\sqrt2} \left( \lvert00\rangle+\lvert10\rangle \right).

CNOT flips the second qubit when the first qubit is 11, so

CNOT⁡12(∣00⟩+∣10⟩)=12(∣00⟩+∣11⟩)=∣Φ+⟩.\operatorname{CNOT} \frac{1}{\sqrt2} \left( \lvert00\rangle+\lvert10\rangle \right) = \frac{1}{\sqrt2} \left( \lvert00\rangle+\lvert11\rangle \right) = \lvert\Phi^+\rangle.
  1. One ebit. Compute the reduced state of either qubit of ∣Φ+⟩\lvert\Phi^+\rangle and its entropy in bits.
Solution

Tracing out either qubit gives

ρA=12∣0⟩⟨0∣+12∣1⟩⟨1∣=12I.\rho_A = \frac12 \lvert0\rangle\langle0\rvert + \frac12 \lvert1\rangle\langle1\rvert = \frac12 I.

The eigenvalues are 1/21/2 and 1/21/2, so

S(ρA)=−2(12log⁡212)=1.S(\rho_A) = -2\left( \frac12\log_2\frac12 \right) = 1.

The Bell pair has one ebit of entanglement.

  1. Why teleportation needs a classical message. In one sentence, explain why teleportation does not allow faster-than-light signaling.
Solution

Before the message, Bob’s averaged density operator is I/2I/2 for every input, so no local measurement can reveal the input; Alice’s classical bits are needed to recover its dependence through a conditional correction.

  1. Local gates. Can local one-qubit unitaries turn ∣00⟩\lvert00\rangle into a Bell state? Explain.
Solution

No. Local unitaries have the form UA⊗UBU_A\otimes U_B, so

(UA⊗UB)∣00⟩=(UA∣0⟩)⊗(UB∣0⟩),(U_A\otimes U_B)\lvert00\rangle = (U_A\lvert0\rangle)\otimes(U_B\lvert0\rangle),

which is still a product state. Creating a Bell state from ∣00⟩\lvert00\rangle requires a nonlocal entangling operation such as CNOT after a local superposition has been prepared.

  1. Entanglement swapping. Why can a Bell-basis measurement on B,CB,C entangle A,DA,D in the state ∣Φ+⟩AB⊗∣Φ+⟩CD\lvert\Phi^+\rangle_{AB}\otimes\lvert\Phi^+\rangle_{CD} without signaling from B,CB,C to A,DA,D?
Solution

The Bell-basis measurement creates outcome-dependent conditional states. Once the classical outcome is known, parties can assign the corresponding entangled state to A,DA,D and apply any needed local correction. When the outcome is ignored, their joint state is I4/4=IA/2⊗ID/2I_4/4=I_A/2\otimes I_D/2, exactly as before the measurement. The protocol uses entanglement plus classical conditioning, not faster-than-light communication.