Density Operators and Composite Quantum Systems

Introduction

Quantum theory describes physical states, transformations, and measurements. For a system known to be in a particular pure state, a state vector is often sufficient. In practical quantum information, however, complete knowledge is unusual. A preparation device may produce different states with known probabilities, an environment may become correlated with the system, or only part of a larger system may be accessible. In these situations, density operators provide the general language for representing quantum states.

This chapter develops density matrices, also called density operators, and applies them to ensembles and composite systems. The central operation for extracting the state of a subsystem is the partial trace. A particularly important distinction is that a classical probabilistic mixture is not generally the same as a quantum superposition, even when both produce similar measurement probabilities in one chosen basis.

Density operators are essential in quantum communication, quantum cryptography, quantum error correction, quantum sensing, and the simulation of noisy quantum hardware.

Learning objectives

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

1. Define pure states, mixed states, ensembles, and density operators.
2. Construct a density matrix from an ensemble of state vectors.
3. Test whether a matrix is a valid quantum state.
4. Calculate expectation values and measurement probabilities using density operators.
5. Distinguish coherent superpositions from incoherent mixtures.
6. Construct composite-system states using tensor products.
7. Calculate reduced density operators using partial traces.
8. Explain how entanglement produces mixed reduced states.
9. Apply these ideas to a practical quantum-information scenario.

1. Pure states and the need for density operators

A pure state of a finite-dimensional quantum system is represented by a normalized vector |ψ⟩ in a Hilbert space. Normalization requires ⟨ψ|ψ⟩ = 1. The vector is physically equivalent to any nonzero global phase multiple e^{iφ}|ψ⟩, because global phase does not affect measurement probabilities.

For a qubit, a general pure state can be written as

where α and β are complex probability amplitudes. Measurement in the computational basis gives outcome 0 with probability |α|² and outcome 1 with probability |β|².

This description assumes that the state vector is known. Suppose instead that a source prepares |0⟩ with probability 1/2 and |1⟩ with probability 1/2, but the observer does not know which preparation occurred. The observer should not describe the source by the state vector (|0⟩ + |1⟩)/√2. That vector represents a coherent superposition and contains phase-sensitive interference. The appropriate description is a mixed state represented by a density operator.

2. Density operators and ensembles

An ensemble is a collection of possible state preparations {|ψ_i⟩, p_i}, where p_i ≥ 0 and ∑_i p_i = 1. Its density operator is

Here ρ is an operator, |ψ_i⟩⟨ψ_i| is the projector associated with preparation i, and p_i is its classical probability. In a chosen orthonormal basis, ρ is represented by a matrix called a density matrix.

For a pure state |ψ⟩, the density operator is

If |ψ⟩ = α|0⟩ + β|1⟩, then

where * denotes complex conjugation. The diagonal entries describe computational-basis populations, while the off-diagonal entries describe coherence in that basis.

A density operator is physically valid if and only if it satisfies three conditions:

The symbol ρ† denotes the conjugate transpose. Positivity is often checked by verifying that all eigenvalues are non-negative. For a 2 × 2 Hermitian matrix, non-negative eigenvalues are equivalent to non-negative trace and determinant, with trace equal to one for a normalized state.

The purity of a state is

For every valid state, 0 < P ≤ 1. A state is pure exactly when P = 1, equivalently when ρ² = ρ or when ρ has one eigenvalue equal to 1 and all others equal to 0. A mixed state has P < 1. In a d-dimensional system, the maximally mixed state is ρ = I/d and has purity 1/d.

Purity measures state preparation uncertainty, but it should not be interpreted as a complete measure of all forms of quantum complexity. For example, a reduced state can be mixed because of entanglement even when the joint state is pure.

3. Measurement with density operators

For a projective measurement with projectors {Π_m}, the probability of result m is given by the Born rule in density-operator form:

The projectors satisfy Π_m† = Π_m, Π_m² = Π_m, Π_mΠ_n = 0 for m ≠ n, and ∑_m Π_m = I. If the result m is obtained and the measurement is ideal and projective, the post-measurement state is

provided p(m) > 0. The denominator normalizes the conditional state.

More general measurements are described by positive operator-valued measures (POVMs), a set of positive operators {E_m} satisfying ∑_m E_m = I. Their outcome probabilities are also

A POVM element need not be a projector. This formalism models noisy, indirect, or generalized measurements.

For an observable A with spectral decomposition A = ∑_a aΠ_a, the expectation value is

The formula is valid when A is an observable represented by a Hermitian operator. It reduces to ⟨ψ|A|ψ⟩ for a pure state.

Worked example 1: a mixed qubit

Consider

The computational-basis probabilities are p(0) = 0.7 and p(1) = 0.3. The purity is

Thus the state is mixed. Measuring in the computational basis reveals the preparation probabilities, but another measurement basis can reveal whether off-diagonal coherence exists. Here there is no coherence between |0⟩ and |1⟩ in this basis.

4. Mixture versus superposition

Consider the two states

and

The pure superposition has density matrix

The mixture has density matrix

Both states give probabilities 1/2 and 1/2 in the computational basis. They differ in their off-diagonal entries and therefore in interference experiments. For example, measurement in the {|+⟩, |−⟩} basis gives outcome + with probability 1 for ρ_+, but probability 1/2 for ρ_mix.

The distinction is physical, not merely a difference in notation. A superposition has a definite phase relationship between components. A mixture represents classical uncertainty about which state was prepared and has no corresponding relative phase between alternatives.

A useful operational rule is: if a phase-sensitive measurement can distinguish two descriptions, they are different quantum states. Merely matching probabilities in one basis does not establish equivalence.

Different ensembles can sometimes produce the same density operator. For example, I/2 can be prepared by an equal mixture of |0⟩ and |1⟩, or by an equal mixture of |+⟩ and |−⟩. The density operator, rather than the preparation story, determines all measurement statistics available on that system alone.

5. Composite quantum systems

If systems A and B have Hilbert spaces of dimensions d_A and d_B, their joint system has the tensor-product space

with dimension d_A d_B. If |a⟩ is a state of A and |b⟩ is a state of B, the product state is |a⟩⊗|b⟩, usually written |a⟩|b⟩ or |ab⟩.

A general product pure state has the form |ψ⟩_A⊗|φ⟩_B. A joint state is separable if it can be written as a product state for pure states, or as a probabilistic mixture of product states for mixed states:

A state that cannot be written in this form is entangled.

For product density operators,

The tensor product preserves normalization because Tr(ρ_A ⊗ ρ_B) = Tr(ρ_A)Tr(ρ_B) = 1.

The Bell state

is not a product state. Its joint density operator is |Φ+⟩⟨Φ+|, which is pure. Nevertheless, the individual qubits are maximally mixed. This apparently surprising result is explained by the partial trace.

6. Partial trace and reduced states

When only subsystem A is accessible, the correct state assigned to it is the reduced density operator

The partial trace over B is defined by choosing any orthonormal basis {|j⟩_B} for B and calculating

Although a basis appears in the calculation, the final operator does not depend on which orthonormal basis is used.

Similarly,

For a product state, partial tracing gives the expected local state:

The reduced state reproduces every measurement statistic for measurements performed only on A. If M_A is an observable on A, represented on the joint system by M_A ⊗ I_B, then

This identity is the operational reason reduced density operators are used.

Component rule for two qubits

Let the joint matrix elements be written in the computational basis as ρ_{ij,kl}, where i,k refer to A and j,l refer to B. Then

The indices belonging to B are set equal and summed. This is a practical method for hand calculations.

Worked example 2: reducing a Bell state

For |Φ+⟩ = (|00⟩ + |11⟩)/√2,

Tracing over B gives

The same calculation gives ρ_B = I/2. Thus each qubit separately is maximally mixed, even though the pair is in a pure state. If the first qubit is measured in the computational basis, the second qubit is then found with the same result. The correlations are contained in ρ_AB and cannot be reconstructed from either reduced state alone.

7. Entanglement, correlations, and ignorance

There are several distinct sources of mixedness:

1. Classical preparation uncertainty: an ensemble contains unknown alternatives.
2. Interaction with an environment: information about a system becomes distributed into inaccessible degrees of freedom.
3. Entanglement: a subsystem of a pure entangled state has a mixed reduced state.
4. Operational noise: state preparation, transmission, or measurement is imperfect.

A reduced mixed state does not by itself prove that the global state is entangled. For example, a classically correlated state

has the same local reductions I/2 as |Φ+⟩, but it is separable and has no off-diagonal terms linking |00⟩ and |11⟩. Joint measurements are required to distinguish these cases.

For a pure bipartite state, the eigenvalues of either reduced state are the squared Schmidt coefficients. The entropy of entanglement is therefore

For a maximally entangled pair, ρ_A = I/2 and S(ρ_A) = 1 bit. The logarithm base 2 measures information in bits. Entropy is zero for a pure reduced state.

8. Density operators and quantum channels

Physical processes can be represented by completely positive, trace-preserving (CPTP) maps, commonly called quantum channels. A channel E transforms a state according to

A channel is trace-preserving when Tr[E(ρ)] = Tr(ρ), and completely positive when positivity is preserved even after extending the channel by an identity operation on an arbitrary ancilla.

A useful representation is the operator-sum or Kraus form:

The second condition ensures trace preservation. For a qubit bit-flip channel with error probability p,

where X is the Pauli-X operator. This describes a state that remains unchanged with probability 1 − p and undergoes a bit flip with probability p. At p = 1/2, computational-basis information is strongly degraded, although the channel's full action depends on the input coherences as well.

The environment provides a physical explanation for many channels. If a system S interacts unitarily with an environment E initially in |e₀⟩, then

The global evolution is reversible and unitary, while the reduced evolution can appear irreversible because environmental information is not collected.

9. Bloch-vector representation of a qubit

Every qubit density operator can be expressed as

where X, Y, and Z are Pauli matrices, σ = (X,Y,Z), and r = (r_x,r_y,r_z) is the Bloch vector. Physical states satisfy |r| ≤ 1. Pure states lie on the unit sphere; mixed states lie inside it; the maximally mixed state is at the origin.

The components can be obtained from

The purity is

Thus a longer Bloch vector corresponds to greater purity. The representation is especially useful for visualizing single-qubit noise, rotations, and measurement statistics, although it does not scale efficiently as a complete visualization for many qubits.

10. Applied case: diagnosing a remote qubit source

Background

A laboratory receives pairs of photons from a source claimed to produce the Bell state |Φ+⟩. One photon is sent to station A and the other to station B. Each station can measure in the computational basis or in the diagonal basis {|+⟩, |−⟩}.

Problem

Local measurements alone show approximately equal frequencies of 0 and 1 at each station. The technicians must determine whether the source produces entangled pairs or merely the classical mixture

Both hypotheses have local reduced state I/2 and therefore cannot be distinguished by measurements at A or B independently.

Analysis and approach

The team records joint outcomes. In the computational basis, both hypotheses predict correlated outcomes: 00 and 11 occur, ideally, with equal probability. The diagonal basis is more informative. The Bell state can be written

so its diagonal-basis results are also perfectly correlated. The classical mixture has no coherence between |00⟩ and |11⟩. In the diagonal basis it produces all four joint outcomes with probabilities 1/4, although its local outcomes remain individually random.

The experiment therefore estimates joint probabilities and correlation functions rather than relying on marginal frequencies. In a real laboratory, detector inefficiency, dark counts, accidental coincidences, and imperfect alignment must be included in uncertainty analysis. A stronger certification may use several measurement settings and a Bell-inequality test, but the basic density-operator analysis already demonstrates why local reduced states are insufficient.

Outcome

If diagonal-basis correlations are close to perfect, the data support the entangled-state model over the classical mixture. If they are absent, the source may be classically correlated or noisy. Statistical confidence requires repeated trials and an explicit model of experimental error; a small sample should not be treated as proof.

Learning points

• Identical reduced states can correspond to different joint states.
• Coherence is revealed by measurements in suitable bases.
• Correlations require joint data.
• Experimental claims must account for noise and finite-sample uncertainty.

11. Practical calculation and implementation guidance

When working with density matrices, use the following procedure:

1. Specify the physical system and basis ordering, such as |00⟩, |01⟩, |10⟩, |11⟩.
2. Normalize all state vectors and ensemble probabilities.
3. Construct each projector |ψ_i⟩⟨ψ_i| before taking the weighted sum.
4. Check Hermiticity, trace, and eigenvalue positivity.
5. Use Tr(Πρ) or Tr(Eρ) for measurement probabilities.
6. For a subsystem, identify the tensor-factor indices carefully before tracing.
7. Compare purity or eigenvalues when classifying pure and mixed states.
8. Keep joint and local states separate; a reduced state does not contain all correlations.

Numerical software should use tolerances because floating-point calculations can produce tiny negative eigenvalues, such as −10⁻¹⁴, for a theoretically positive matrix. Such values may be numerical round-off rather than physical negativity, but substantial negative eigenvalues indicate an invalid calculation or an invalid estimated state. In tomography, enforcing physicality may require constrained estimation rather than simply setting negative entries to zero.

12. Limitations and contemporary relevance

A density matrix contains all operational information about a quantum state, but it can become exponentially large. An n-qubit density matrix has 2ⁿ × 2ⁿ complex entries, although Hermiticity, trace, and positivity reduce the number of independent real parameters. This scaling limits direct classical simulation of large quantum systems.

Quantum state tomography is also measurement-intensive. Estimating a general state requires many measurement settings and careful statistical treatment. For large systems, researchers use compressed sensing, tensor-network models, shadow tomography, randomized measurements, and problem-specific models.

In quantum computing, density operators are indispensable for modelling decoherence, thermal initialization, readout errors, noisy gates, and error-correction syndromes. In quantum communication, they describe mixed signals, eavesdropper knowledge, and entangled resources. In quantum machine learning and sensing, they support statistical inference from imperfect quantum data.

Good governance requires clear separation between an estimated density matrix and a directly observed fact. Tomographic reconstructions depend on calibration, measurement assumptions, and finite data. Claims about entanglement or secure communication should therefore report uncertainties, test procedures, and possible loopholes.

Chapter Summary

Density operators generalize state vectors to include pure states, classical ensembles, and reduced states of larger systems. An ensemble {|ψ_i⟩, p_i} is represented by ρ = ∑_i p_i|ψ_i⟩⟨ψ_i|. A valid density operator is Hermitian, positive semidefinite, and has unit trace. Measurement probabilities are calculated using Tr(Eρ), and observable averages using Tr(ρA).

A coherent superposition contains off-diagonal phase information, whereas a classical mixture represents probabilistic uncertainty and generally lacks those coherence terms. Composite systems are described in tensor-product spaces. The partial trace produces the state of an accessible subsystem and preserves all statistics of measurements performed only on that subsystem. Entangled joint states can have mixed reduced states, so local mixedness alone does not prove classical uncertainty or entanglement.

Density operators also provide the natural language for quantum channels, environmental noise, quantum tomography, and practical quantum-device analysis. Reliable work requires careful basis ordering, positivity checks, explicit treatment of noise, and distinction between local data and joint correlations.

Key Terms

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

Density matrix: The matrix representation of a density operator in a chosen basis.

Pure state: A state represented by one state vector, with purity equal to one.

Mixed state: A statistical state represented by a nontrivial convex combination of pure states.

Ensemble: A set of possible preparations together with their probabilities.

Projector: An operator P satisfying P² = P; |ψ⟩⟨ψ| is the projector onto |ψ⟩.

Hermitian operator: An operator equal to its conjugate transpose.

Positive semidefinite: An operator whose expectation value is non-negative for every vector.

Purity: The quantity Tr(ρ²), indicating whether a state is pure or mixed.

Coherence: Phase-sensitive off-diagonal structure in a density matrix.

Superposition: A coherent linear combination of basis states.

Classical mixture: A probabilistic ensemble without coherent phase relations between alternatives.

Tensor product: The mathematical construction describing a composite quantum system.

Composite system: A system formed from two or more subsystems.

Partial trace: The operation that removes an inaccessible subsystem from a joint density operator.

Reduced state: The density operator of a subsystem obtained by a partial trace.

Entanglement: A nonclassical correlation that prevents a joint state from being expressed as a mixture of product states.

POVM: A collection of positive measurement operators summing to the identity.

Kraus operator: An operator used in the operator-sum representation of a quantum channel.

Bloch vector: A three-dimensional vector representation of a qubit density operator.

Review Questions

Short Answer Questions

1. What three mathematical conditions must a density operator satisfy?
2. Why can a mixture of |0⟩ and |1⟩ not generally be replaced by the superposition (|0⟩ + |1⟩)/√2?
3. What physical information does the partial trace provide?
4. Define purity and state its value for a pure state.
5. Why can an entangled two-qubit state have a mixed one-qubit reduced state?

Long Answer Questions

1. Derive the density matrix of a general pure qubit and explain the meaning of its diagonal and off-diagonal entries.
2. Explain the validity conditions for density operators and describe practical tests for each condition.
3. Compare the Bell state |Φ+⟩ with the classical correlated mixture 1/2|00⟩⟨00| + 1/2|11⟩⟨11|.
4. Explain the partial trace and derive the reduced state of one qubit from |Φ+⟩.
5. Describe how Kraus operators model noisy quantum processes and explain the trace-preserving condition.

Analytical / Application Questions

1. A source produces |0⟩ with probability 0.6 and |+⟩ with probability 0.4. Construct its density matrix in the computational basis and calculate its purity.
2. Design a measurement strategy that distinguishes the state |+⟩ from the equal mixture of |0⟩ and |1⟩. Explain why computational-basis measurements alone are insufficient.
3. A two-qubit experiment reports maximally mixed local states. Explain why this result does not determine whether the joint state is entangled, and propose joint measurements that could distinguish an entangled Bell state from a classical correlated mixture.

Multiple Choice Questions with Answers

Question 1

Which condition is required for a valid density operator?

A. Its trace must be zero

B. It must be positive semidefinite

C. All its entries must be real

D. Its determinant must equal one

Correct Answer: B. It must be positive semidefinite

Question 2

What is the density operator for an equal incoherent mixture of |0⟩ and |1⟩?

A. |+⟩⟨+|

B. |0⟩⟨1|

C. I/2

D. 2I

Correct Answer: C. I/2

Question 3

For a measurement described by POVM element E_m, the outcome probability is

A. Tr(E_m + ρ)

B. Tr(E_mρ)

C. Tr(E_m)Tr(ρ) only

D. det(E_mρ)

Correct Answer: B. Tr(E_mρ)

Question 4

What is the purity of the state ρ = diag(0.7, 0.3)?

A. 0.30

B. 0.49

C. 0.58

D. 1.00

Correct Answer: C. 0.58

Question 5

The reduced state of subsystem A is obtained from a joint state ρ_AB by

A. Taking the determinant over B

B. Transposing subsystem B

C. Applying the partial trace over B

D. Setting ρ_AB equal to ρ_A ⊗ ρ_B

Correct Answer: C. Applying the partial trace over B

Question 6

For the Bell state |Φ+⟩ = (|00⟩ + |11⟩)/√2, the reduced state of either qubit is

A. |0⟩⟨0|

B. |1⟩⟨1|

C. |+⟩⟨+|

D. I/2

Correct Answer: D. I/2

Question 7

Which feature distinguishes |+⟩ from the equal mixture of |0⟩ and |1⟩?

A. Their computational-basis probabilities

B. Their dimensions

C. Their off-diagonal coherence terms

D. Their trace

Correct Answer: C. Their off-diagonal coherence terms

Question 8

If systems A and B have dimensions 2 and 3, respectively, the composite Hilbert space has dimension

A. 1

B. 5

C. 6

D. 9

Correct Answer: C. 6

Question 9

In a Kraus representation of a trace-preserving channel, which condition must hold?

A. ∑_k K_kK_k† = 0

B. ∑_k K_k†K_k = I

C. Each K_k must be unitary

D. Each K_k must have trace one

Correct Answer: B. ∑_k K_k†K_k = I

Question 10

What is the Bloch-vector condition for a physical qubit state?

A. |r| < 0

B. |r| = 2

C. |r| ≤ 1

D. |r| ≥ 1

Correct Answer: C. |r| ≤ 1