Density Matrix and Mixed States

The standard quantum mechanics you first learn, built on state vectors $|\psi ⟩$, has a fundamental limitation: it assumes you have complete information about your system. This works perfectly for isolated particles in pristine laboratories, but fails dramatically for real-world scenarios—from a qubit interacting with its noisy environment to a macroscopic object in thermal equilibrium. The density matrix (or density operator) formalism is the essential mathematical framework that extends quantum mechanics to handle systems where we have partial or statistical information. It is the lingua franca for quantum statistical mechanics, decoherence theory, and the study of open quantum systems, providing the tools to quantify entanglement, model thermal states, and track the loss of quantum coherence.

From State Vectors to the Density Operator

We begin with a system in a pure state, described by a normalized state vector $|\psi ⟩$. All its quantum information is contained in this vector. The corresponding density operator $\rho$ for a pure state is defined as the projector onto that state: $$\rho =|\psi ⟩⟨\psi |.$$ This is a Hermitian operator (${\rho }^{†}=\rho$) with key properties: it is positive semidefinite (its eigenvalues are non-negative), it has trace one ($\text{Tr}(\rho )=1$), and for a pure state, it is idempotent (${\rho }^2=\rho$). The trace condition enforces probability conservation.

The power of the density matrix appears when we consider statistical mixtures. A mixed state describes a system that is not in a single, known pure state, but exists with a certain probability distribution over several possible pure states. Imagine an ensemble where a fraction $p_i$ of systems is prepared in state $|{\psi }_i⟩$. The system's state is not $|{\psi }_i⟩$ for any single $i$; it is the statistical mixture itself. The density operator generalizes perfectly to this case: $$\rho =\sum _i{p}_i|{\psi }_i⟩⟨{\psi }_i|,$$ where $0\le p_i\le 1$ and ${\sum }_i{p}_i=1$. This formulation is crucial: the states $|{\psi }_i⟩$ need not be orthogonal, and the decomposition of a given $\rho$ into such a mixture is not unique. The physically meaningful object is $\rho$ itself.

Computing Observables and the Born Rule

How do we compute the expectation value of an observable $\hat{O}$? In the state vector formalism, it's $⟨\psi |\hat{O}|\psi ⟩$. In the density matrix formalism, this generalizes elegantly via the trace: $$⟨\hat{O}⟩=\text{Tr}(\rho \hat{O}).$$ For a pure state $\rho =|\psi ⟩⟨\psi |$, this reduces to the familiar form. For a mixed state $\rho ={\sum }_i{p}_i|{\psi }_i⟩⟨{\psi }_i|$, it gives the statistically weighted average: $⟨\hat{O}⟩={\sum }_i{p}_i⟨{\psi }_i|\hat{O}|{\psi }_i⟩$.

Similarly, the probability of obtaining a specific measurement outcome associated with projector $\hat{P}$ is given by the generalized Born rule: $P=\text{Tr}(\rho \hat{P})$. This trace-based formalism is often more computationally convenient than working directly with state vectors, especially in composite systems.

Entanglement and the Reduced Density Matrix

The density matrix formalism provides the definitive tool for describing subsystems and quantifying entanglement. Consider a composite system AB in a pure state $|{\Psi }_{AB}⟩$. The state of subsystem A alone is not generally a pure state; it is a mixed state described by the reduced density matrix. You obtain it by partial trace: tracing over the degrees of freedom of subsystem B: $${\rho }_A={\text{Tr}}_B(|{\Psi }_{AB}⟩⟨{\Psi }_{AB}|).$$ If $|{\Psi }_{AB}⟩$ is separable (a product state), then ${\rho }_A$ is a pure state. If $|{\Psi }_{AB}⟩$ is entangled, then ${\rho }_A$ is a mixed state. The entanglement between A and B is directly responsible for the "mixedness" of the subsystem.

We can quantify this mixedness using the von Neumann entanglement entropy: $$S_A=-\text{Tr}({\rho }_A\ln {\rho }_A).$$ For a pure composite state, $S_A=S_B$, and it is zero if and only if the subsystems are unentangled. A positive $S_A$ signals entanglement. This entropy is a cornerstone of quantum information theory and quantum statistical mechanics.

The Thermal Equilibrium State

A quintessential application is describing a quantum system in thermal equilibrium with a heat bath at temperature $T$. Its state is not a pure state but a statistical mixture—the canonical ensemble. The density operator is $${\rho }_{\text{th}}=\frac{e^{-\beta \hat{H}}}{\mathcal{Z}},$$ where $\beta =1/(k_{B}T)$, $\hat{H}$ is the system's Hamiltonian, and $\mathcal{Z}=\text{Tr}(e^{-\beta \hat{H}})$ is the partition function. This is a mixed state where the probability $p_n$ of being in energy eigenstate $|n⟩$ with energy $E_n$ is given by the Boltzmann factor: $p_n=e^{-\beta E_n}/\mathcal{Z}$. Expectation values become thermal averages: $⟨\hat{O}{⟩}_{\text{th}}=\text{Tr}({\rho }_{\text{th}}\hat{O})$. This formalism bridges quantum mechanics and statistical mechanics.

Decoherence and Open Quantum Systems

For a system interacting with an environment (an open quantum system), the composite system+environment may be in a pure state, but the system alone is described by a reduced density matrix ${\rho }_S(t)={\text{Tr}}_E(|{\Psi }_{SE}(t)⟩⟨{\Psi }_{SE}(t)|)$. The evolution of ${\rho }_S$ is generally not unitary. A common master equation that approximates this evolution is the Lindblad equation: $$\frac{d{\rho }_S}{dt}=-\frac{i}{\hslash }[\hat{H},{\rho }_S]+\sum _k\left({\hat{L}}_k{\rho }_S{\hat{L}}_k^{†}-\frac{1}{2}\{{\hat{L}}_k^{†}{\hat{L}}_k,{\rho }_S\}\right).$$ The first term gives unitary evolution. The Lindblad operators ${\hat{L}}_k$ model the non-unitary, dissipative effects of the environment, leading to decoherence—the rapid loss of quantum superposition (off-diagonal elements in ${\rho }_S$ in a preferred basis) and the emergence of classical probability distributions. This explains why large objects don't exhibit quantum interference: their density matrix diagonals rapidly in the position basis.

Common Pitfalls

1. Confusing mixtures with superpositions. A superposition $|\psi ⟩={\sum }_i{c}_i|i⟩$ is a single pure state. A mixture $\rho ={\sum }_i{p}_i|i⟩⟨i|$ represents a classical probability distribution over states. The coherence (phase information) between terms in a superposition is present in the off-diagonal elements of $\rho$ for the pure state but absent in the diagonal mixed state. Adding the probabilities $|c_i{|}^2$ does not create a mixture.
2. Assuming a unique ensemble decomposition. A given density matrix $\rho$ can be prepared from infinitely many different statistical mixtures of pure states. For example, the maximally mixed state for a qubit, $\rho =I/2$, can be described as an equal mix of $|0⟩$ and $|1⟩$, or an equal mix of $|+⟩$ and $|-⟩$. All physical predictions depend only on $\rho$, not on which ensemble you imagine.
3. Misinterpreting the reduced density matrix. If subsystem A is described by a mixed state ${\rho }_A$, it does not mean someone simply "doesn't know" the pure state. It means A is fundamentally entangled with B. No local description of A can be a pure state. This mixedness is a signature of quantum correlations.
4. Applying unitary evolution to mixed states incorrectly. For a mixed state $\rho ={\sum }_i{p}_i|{\psi }_i⟩⟨{\psi }_i|$, if each pure state evolves unitarily as $|{\psi }_i(t)⟩=U(t)|{\psi }_i(0)⟩$, then the density matrix evolves as $\rho (t)=U(t)\rho (0)U^{†}(t)$. You cannot simply put time dependence into the probabilities $p_i$; the evolution acts on the projectors themselves.

Summary

• The density operator $\rho$ is the fundamental object for describing quantum systems when information is incomplete, generalizing the pure state vector to both pure (${\rho }^2=\rho$) and mixed states.
• Expectation values and measurement probabilities are computed using the trace: $⟨\hat{O}⟩=\text{Tr}(\rho \hat{O})$.
• The state of a subsystem is given by the reduced density matrix, obtained via a partial trace. The von Neumann entropy $S=-\text{Tr}(\rho \ln \rho )$ of this reduced state quantifies entanglement for a pure composite system.
• A system in thermal equilibrium is described by the canonical density matrix ${\rho }_{\text{th}}=e^{-\beta \hat{H}}/\mathcal{Z}$, forming the basis of quantum statistical mechanics.
• For open quantum systems, interactions with an environment lead to non-unitary evolution (e.g., via a Lindblad master equation), causing decoherence—the loss of quantum coherence and the emergence of classical behavior in the system's reduced density matrix.