Entanglement, Nonlocality and Resource Theories

Introduction

Entanglement is one of the defining resources of quantum information theory. It describes correlations between quantum systems that cannot be explained by assigning independent quantum states to the systems. Entanglement is central to quantum teleportation, superdense coding, quantum cryptography, distributed quantum computation, quantum sensing and quantum networks.

Entanglement should not be confused with communication faster than light. Measurements on entangled systems can produce correlations that violate classical limits, but usable information cannot be transmitted instantaneously. The theory of entanglement therefore combines three ideas: the structure of composite quantum states, the operational restrictions on physically allowed parties, and quantitative measures of nonclassical correlation.

This chapter develops bipartite entanglement through Schmidt decomposition, Bell inequalities, reduced density operators, entanglement measures, monogamy, local operations and classical communication (LOCC), and entanglement concentration and dilution. It concludes by treating entanglement as a resource that can be created, transformed, consumed and protected.

Learning objectives

By the end of this chapter, a learner should be able to:

1. Distinguish product, separable and entangled states.
2. Use Schmidt decomposition to identify and quantify pure-state entanglement.
3. Explain Bell inequalities and interpret their experimental violation.
4. Calculate entropy-based entanglement measures for simple states.
5. Explain LOCC and why it defines a resource theory.
6. Describe monogamy of entanglement and its relevance to quantum networks.
7. Distinguish entanglement concentration from dilution.
8. Evaluate practical limitations in the generation and use of entanglement.

1. Composite quantum systems

If systems A and B have Hilbert spaces H_A and H_B, their joint space is the tensor product

A pure joint state is represented by a unit vector |ψ⟩ in H_AB. A product state has the form

where |a⟩ and |b⟩ are states of the individual systems. Such a state contains no quantum correlation between A and B.

A mixed state is represented by a density operator ρ satisfying ρ ≥ 0 and Tr(ρ) = 1. A separable state is any state that can be written as

The states ρ_A^(i) and ρ_B^(i) may be pure or mixed. A state that cannot be written in this form is entangled. Every product state is separable, but a separable mixed state need not be a product state because it may contain classical correlations.

For a joint state ρ_AB, the reduced state of A is obtained by tracing out B:

Similarly, ρ_B = Tr_A(ρ_AB). The reduced state describes all measurements performed only on the corresponding subsystem.

2. Schmidt decomposition

The Schmidt decomposition is the most useful structural theorem for bipartite pure states. For every pure state |ψ⟩_AB, there exist orthonormal sets {|u_i⟩_A} and {|v_i⟩_B} such that

where λ_i ≥ 0, ∑_i λ_i = 1, and r is the Schmidt rank. The λ_i are called Schmidt coefficients in probability form. Some texts call √λ_i the Schmidt coefficients and λ_i their squared values.

The reduced states are

Thus, ρ_A and ρ_B have the same nonzero eigenvalues. This gives several immediate conclusions:

• A pure state is a product state if and only if its Schmidt rank is 1.
• A pure state is entangled if and only if its Schmidt rank is greater than 1.
• The entanglement of a pure state is determined entirely by the distribution {λ_i}.
• The maximum Schmidt rank is min(dim H_A, dim H_B).

For two qubits, a state with Schmidt coefficients λ and 1 − λ is maximally entangled when λ = 1/2. A standard maximally entangled Bell state is

Its reduced state is I/2, where I is the two-dimensional identity operator. Although the joint state is pure, each subsystem individually is maximally mixed.

Worked example: Schmidt decomposition

Consider

This is already in Schmidt form, with λ_1 = 3/4 and λ_2 = 1/4. Since two nonzero Schmidt coefficients exist, the state is entangled. The reduced state of either qubit is

It is not maximally mixed, so the state is entangled but not maximally entangled.

3. Quantifying bipartite entanglement

A useful entanglement measure should be zero for separable states, non-increasing under LOCC on average, and generally invariant under local unitary transformations. Different measures are useful for different operational tasks.

3.1 Entanglement entropy

For a bipartite pure state, the entanglement entropy is the von Neumann entropy of either reduced state:

where

The unit is bits because the logarithm is base 2. If the eigenvalues of ρ are λ_i, then

The convention 0 log₂ 0 = 0 is used. For a d-dimensional maximally mixed state, S = log₂ d. For a two-qubit maximally entangled state, E = 1 ebit, where an ebit is the entanglement equivalent to one Bell pair.

For the worked example above,

The state therefore contains approximately 0.811 ebits of pure-state entanglement.

3.2 Entanglement of formation

For a mixed state, a decomposition into pure states is generally not unique. The entanglement of formation is defined by the convex-roof construction:

The minimum is over all ensembles satisfying ρ = ∑_i p_i |ψ_i⟩⟨ψ_i|. It represents the least average pure-state entanglement needed to prepare ρ when the preparation information is disregarded. The minimization can be difficult, although closed formulas exist for two-qubit states through concurrence.

3.3 Concurrence for two qubits

For a two-qubit state ρ, define its spin-flipped operator

where σ_y is the Pauli-y matrix and ρ* denotes complex conjugation in the computational basis. Let R = √(√ρ \tilde{ρ}√ρ), with eigenvalues r_1 ≥ r_2 ≥ r_3 ≥ r_4. The concurrence is

For a pure two-qubit state, C ranges from 0 for product states to 1 for maximally entangled states. The entanglement of formation is then

where h(x) = −x log₂ x − (1 − x)log₂(1 − x). This formula applies specifically to two-qubit states.

3.4 Negativity

The partial transpose with respect to B is defined by

The negativity is

where ||·||_1 is the trace norm, equal to the sum of singular values. A negative eigenvalue of ρ^{T_B} proves entanglement; this is the Peres–Horodecki positive-partial-transpose criterion. For systems of dimensions 2 × 2 and 2 × 3, positivity of the partial transpose is also sufficient for separability. In higher dimensions, some entangled states have positive partial transpose and are called bound entangled states.

4. Bell inequalities and quantum nonlocality

Entanglement and nonlocality are related but not identical. Entanglement is a property of a quantum state. Bell nonlocality concerns whether observed correlations can be reproduced by a local hidden-variable model.

In a Bell experiment, two separated observers, Alice and Bob, independently choose measurement settings x and y and obtain outcomes a and b. A local hidden-variable model has the form

Here λ is a shared hidden variable. Locality means that Alice's outcome distribution does not depend on Bob's setting and vice versa, conditioned on λ.

CHSH inequality

The Clauser–Horne–Shimony–Holt expression is

where A_0 and A_1 are Alice's two observables, B_0 and B_1 are Bob's two observables, and E(A_xB_y) is the expectation value of the product of their ±1 outcomes. Every local hidden-variable theory satisfies

Quantum mechanics permits values up to Tsirelson's bound,

The bound is achieved using a maximally entangled two-qubit state and appropriately chosen measurement directions. No-signalling theories can reach the algebraic maximum |S| = 4, but quantum theory does not.

Experimental tests must address loopholes. The detection loophole occurs when inefficient detectors create a biased sample. The locality loophole occurs when measurement events are not space-like separated. The freedom-of-choice loophole concerns possible correlations between measurement settings and hidden variables. Modern experiments have performed increasingly strong loophole-resistant tests, although practical noise and loss remain important in technology.

Bell violation is useful in device-independent quantum cryptography and randomness generation. In such protocols, observed correlations can certify security or randomness with limited trust in the internal operation of devices.

5. Monogamy of entanglement

Entanglement cannot generally be shared arbitrarily among many systems. For three qubits A, B and C, a commonly used monogamy relation is the Coffman–Kundu–Wootters inequality:

Here C_A|BC is the concurrence between A and the joint system BC, while C_AB and C_AC are pairwise concurrences. This relation applies to three-qubit systems and illustrates that strong pairwise entanglement between A and B limits the pairwise entanglement that A can share with C.

For the three-qubit GHZ state,

the pairwise reduced states are separable, although the three-party state has genuine multipartite entanglement. For the W state,

pairwise entanglement is present, and it is distributed differently from GHZ entanglement. These examples show that “total entanglement” is not a single universal quantity for multipartite systems.

Monogamy supports quantum key distribution: an eavesdropper who becomes strongly correlated with a legitimate receiver cannot simultaneously maintain unrestricted correlation with the sender. It also constrains quantum network design, repeater architectures and multipartite sensing.

6. LOCC and the resource theory of entanglement

LOCC consists of operations that separated parties can perform using:

• Local quantum operations on their own systems.
• Local measurements and preparation.
• Classical communication of measurement results.
• Shared randomness and local ancillary systems.

Parties cannot send quantum systems to one another under a strict LOCC model. Local unitaries cannot change entanglement, while measurements and classical communication cannot increase the average entanglement of an ensemble. This asymmetry defines entanglement as a resource.

A resource theory specifies four elements:

1. Objects: quantum states, channels or correlations.
2. Free states: states that require no resource, such as separable states.
3. Free operations: operations allowed without consuming the resource, such as LOCC.
4. Resource measures: quantities that cannot increase under free operations.

A valid monotone E should satisfy, at minimum, E(ρ) = 0 for free states and

for deterministic LOCC operation Λ_LOCC. For a measurement with outcomes i, a stronger average condition is

where p_i is the probability of outcome i and ρ_i is the corresponding post-measurement state. Some measures may increase conditionally for a particular outcome, but not on average when the outcome statistics are included.

LOCC is more restrictive than separable operations and differs from operational models such as PPT-preserving operations, which are useful in mathematical analyses. The chosen free-operation class affects which transformations are possible.

7. Entanglement concentration and dilution

Suppose Alice and Bob share many identical copies of a partially entangled pure state |ψ⟩. Entanglement concentration converts these copies, using LOCC, into fewer nearly maximally entangled Bell pairs. In the asymptotic limit, the optimal yield is the entanglement entropy:

where n is the number of input copies and m is the number of output ebits as n becomes large. Concentration is therefore lossy in copy number but can produce high-quality standard entanglement.

Entanglement dilution performs the reverse task: it uses Bell pairs to prepare copies of a partially entangled state. For pure states in the asymptotic limit, the number of ebits required per output copy is also E(|ψ⟩). Thus concentration and dilution are reversible in the ideal asymptotic pure-state setting, although finite-size effects, classical communication, noise and imperfect operations make real processes irreversible.

For a single copy, LOCC transformations are characterized by majorization. Let λ and μ be the ordered Schmidt probability vectors of two pure states. The transformation |ψ⟩ → |φ⟩ by deterministic LOCC is possible if λ is majorized by μ, written λ ≺ μ. A maximally entangled state has the most uniform Schmidt vector and can be converted into any state of the same or lower local dimension. The reverse conversion is generally not deterministic.

8. Entanglement distribution and practical systems

Entanglement is generated in several ways:

• Parametric down-conversion produces correlated photon pairs.
• Trapped ions and superconducting circuits generate entanglement through controlled interactions.
• Quantum dots and color centers can produce matter–photon entanglement.
• Quantum repeaters aim to distribute entanglement over long distances using elementary links and entanglement swapping.

Entanglement swapping illustrates that two particles that have never directly interacted can become entangled. If Alice–Bob and Charlie–David share entangled pairs, a Bell-state measurement on Bob and Charlie projects Alice and David into an entangled state, subject to known Pauli corrections communicated classically.

Practical performance is limited by loss, decoherence, imperfect state preparation, detector inefficiency, mode mismatch and finite memory coherence. An entanglement measure calculated from an ideal state may therefore overestimate the usable resource. Engineers must distinguish state fidelity, entanglement measures, heralding probability, throughput and end-to-end secret-key rate.

9. Illustrative application: a metropolitan quantum network

Background

Consider an illustrative three-node network connecting a university laboratory, a hospital and a data centre. The nodes are linked by optical fibre. The objective is to establish entanglement between distant nodes for device-independent key generation and distributed quantum sensing.

Problem

Direct transmission of photons over fibre suffers exponential loss with distance. The network must determine whether generated correlations are sufficiently strong for the intended application and whether intermediate nodes should be trusted.

Analysis

The network first characterizes each elementary link using state tomography or, where trust is limited, Bell-inequality tests. A measured CHSH value above 2 indicates nonlocal correlations under the relevant experimental assumptions. However, a value below 2 does not prove that the state is separable; noise, measurement choices or insufficient statistics may explain the result.

The operators compare the measured density matrix with the target Bell state using fidelity,

for density operators ρ and σ. Fidelity close to one indicates similarity but is not itself identical to an entanglement measure. Partial-transpose analysis and an appropriate entanglement measure are also applied.

Approach

The network uses entanglement swapping at the intermediate node. Memory systems store successful elementary links until neighbouring links are ready. Classical messages announce Bell-measurement outcomes, allowing endpoint Pauli-frame corrections. Link purification may be used if multiple noisy pairs are available, but purification consumes pairs and requires sufficiently high initial fidelity.

Outcome

The network can deliver fewer but higher-quality entangled pairs than the raw source rate. The final performance is reported using several metrics: delivered-pair rate, fidelity, measured Bell value, memory lifetime, failure probability and classical communication delay.

Learning points

1. Entanglement is a usable resource only when quality and delivery rate are considered together.
2. Bell violation can provide device-independent evidence, but experimental loopholes and finite statistics must be analysed.
3. Entanglement swapping extends range but does not remove loss; quantum memories and repeaters are needed for scalable systems.
4. A resource-theoretic analysis prevents the incorrect assumption that every correlated pair is equally useful.

10. Limitations, governance and best practice

Entanglement does not permit faster-than-light signalling. Any operational claim must respect the no-signalling condition, which requires that Alice's marginal outcome distribution be independent of Bob's measurement choice:

for all choices of y, assuming properly normalized quantum measurements.

Experimental claims should report uncertainties, calibration procedures, detection efficiency, sampling assumptions and loophole analysis. Security claims must specify the threat model, device assumptions, finite-key corrections and side-channel controls. Quantum hardware should be evaluated not only by entanglement fidelity but also by reproducibility, scalability, energy use, maintainability and interoperability.

Current research includes high-dimensional entanglement, continuous-variable systems, network nonlocality, multipartite resource theories, entanglement-assisted error correction, quantum error-correcting codes and device-independent protocols. A significant challenge is that entanglement is often fragile while useful quantum applications require reliable distribution across noisy systems.

Chapter Summary

Bipartite entanglement is a property of composite quantum states that cannot be represented as mixtures of product states. Schmidt decomposition provides a complete description of pure-state bipartite entanglement: the state is separable exactly when it has Schmidt rank one, and its entanglement entropy is the von Neumann entropy of either reduced state. Mixed-state entanglement requires more advanced measures, including entanglement of formation, concurrence and negativity.

Bell inequalities distinguish local hidden-variable correlations from quantum correlations. The CHSH bound is 2 for local models and 2√2 for quantum theory. Bell nonlocality is stronger than entanglement in the operational sense: some entangled states do not violate a particular Bell inequality. Monogamy relations limit how entanglement can be distributed among multiple systems.

LOCC defines the principal resource theory of entanglement. Separable states are free states, while entangled states are resources that cannot be increased on average by LOCC. Entanglement concentration converts partially entangled states into fewer Bell pairs, while dilution uses Bell pairs to prepare partially entangled states. In the asymptotic pure-state regime, both rates are governed by entanglement entropy. Real systems face loss, decoherence, imperfect detection and finite-size effects, so practical evaluation must combine theoretical measures with engineering metrics.

Key Terms

Entanglement: A quantum correlation that cannot be represented as a mixture of product states.

Product state: A state of the form |a⟩⊗|b⟩ with no quantum correlation between subsystems.

Separable state: A state expressible as a probabilistic mixture of product states.

Density operator: A positive, unit-trace operator representing a quantum state.

Partial trace: The operation used to obtain a subsystem's reduced state.

Schmidt decomposition: Canonical expansion of a bipartite pure state in paired orthonormal bases.

Schmidt rank: Number of nonzero Schmidt coefficients.

Entanglement entropy: Von Neumann entropy of a reduced state for a pure bipartite state.

Ebit: Standard unit of bipartite entanglement represented by one Bell pair.

Bell inequality: Constraint satisfied by local hidden-variable theories.

CHSH inequality: A two-party, two-setting Bell inequality with local bound 2.

Bell nonlocality: Correlations incompatible with every local hidden-variable model.

Monogamy: Restriction on how strongly entanglement can be shared among systems.

LOCC: Local operations and classical communication between separated parties.

Entanglement monotone: A quantity that does not increase under allowed free operations.

Concurrence: A two-qubit entanglement measure ranging from 0 to 1.

Negativity: Entanglement measure based on the negative eigenvalues of a partial transpose.

Entanglement concentration: Conversion of many partially entangled states into fewer maximally entangled states.

Entanglement dilution: Preparation of partially entangled states using maximally entangled resources.

Entanglement swapping: Creation of remote entanglement through a joint measurement on intermediate systems.

Review Questions

Short Answer Questions

1. What distinguishes a product state from an entangled state?
2. What information is provided by the Schmidt rank?
3. Define entanglement entropy for a pure bipartite state.
4. What is the local hidden-variable bound for the CHSH expression?
5. Why is entanglement described as a resource under LOCC?

Long Answer Questions

1. Explain Schmidt decomposition and use it to characterize pure-state bipartite entanglement.
2. Compare entanglement entropy, concurrence and negativity, including their domains of usefulness.
3. Explain the CHSH inequality, its quantum violation and the main experimental loopholes.
4. Describe monogamy of entanglement and discuss its relevance to quantum cryptography.
5. Explain entanglement concentration and dilution in the asymptotic pure-state regime.

Analytical / Application Questions

1. For the state |ψ⟩ = √(3/4)|00⟩ + √(1/4)|11⟩, calculate the reduced state and entanglement entropy. Interpret the result.
2. A two-qubit experiment measures a CHSH value of 2.35. Explain what can and cannot be concluded about locality, entanglement and experimental reliability.
3. Design a high-level LOCC strategy for distributing entanglement between two distant nodes using entanglement swapping. Identify at least three practical sources of performance loss.

Multiple Choice Questions with Answers

Question 1

Which condition characterizes a product pure state?

A. Its Schmidt rank is 1.

B. Its reduced states are maximally mixed.

C. It violates the CHSH inequality.

D. Its von Neumann entropy is always 1 bit.

Correct Answer: A. Its Schmidt rank is 1.

Question 2

For a bipartite pure state, entanglement entropy is calculated from which state?

A. The global state only

B. Either reduced state

C. The classical probability distribution of measurement settings

D. The tensor-product dimension only

Correct Answer: B. Either reduced state

Question 3

What is the CHSH bound for local hidden-variable theories?

A. 1

B. 2

C. 2√2

D. 4

Correct Answer: B. 2

Question 4

What is the maximum quantum value of the CHSH expression?

A. 2

B. √2

C. 2√2

D. 4

Correct Answer: C. 2√2

Question 5

Which operation is not part of strict LOCC?

A. A local measurement

B. Classical communication

C. A local unitary operation

D. Sending a quantum system between parties

Correct Answer: D. Sending a quantum system between parties

Question 6

For the state √(3/4)|00⟩ + √(1/4)|11⟩, what is its Schmidt rank?

A. 1

B. 2

C. 3

D. 4

Correct Answer: B. 2

Question 7

What does a negative eigenvalue of a partial transpose certify?

A. The state is maximally mixed.

B. The state is entangled.

C. The state is classical.

D. The state violates every Bell inequality.

Correct Answer: B. The state is entangled.

Question 8

In the asymptotic pure-state limit, the optimal concentration rate is governed by which quantity?

A. The trace of the global density matrix before normalization

B. The entanglement entropy

C. The detector dark-count rate

D. The CHSH algebraic maximum

Correct Answer: B. The entanglement entropy

Question 9

What does monogamy of entanglement imply?

A. Entanglement can be freely copied.

B. Strong pairwise entanglement limits simultaneous sharing with other systems.

C. All multipartite states have identical pairwise entanglement.

D. Classical communication creates unlimited entanglement.

Correct Answer: B. Strong pairwise entanglement limits simultaneous sharing with other systems.

Question 10

A measured CHSH value of 2.35 is observed. Under suitable loophole-resistant conditions, what does this indicate?

A. A violation of the local hidden-variable bound

B. A proof that the state is separable

C. A value impossible in quantum mechanics

D. Faster-than-light signalling

Correct Answer: A. A violation of the local hidden-variable bound