Entropy of entanglement
The entropy of entanglement, also known as entanglement entropy, is a fundamental measure in quantum information theory that quantifies the amount of quantum entanglement shared between two subsystems in a bipartite pure quantum state. It is defined as the von Neumann entropy S(\rho_A) = -\operatorname{Tr}(\rho_A \log_2 \rho_A), where \rho_A is the reduced density matrix obtained by tracing out the second subsystem from the overall pure density matrix of the composite system.[1] This entropy equals the Shannon entropy of the squared Schmidt coefficients of the state and is symmetric, meaning S(\rho_A) = S(\rho_B).[2] For a product state, the entropy is zero, indicating no entanglement, while it reaches its maximum value of \log_2 d ebits (where d is the dimension of the subsystems) for a maximally entangled state, such as a Bell pair in two qubits.[3] Introduced in 1996 by Bennett, Bernstein, Popescu, and Schumacher, the entropy of entanglement provides an operational interpretation as the asymptotic yield of maximally entangled pairs (ebits) that can be distilled from many copies of the state using local operations and classical communication (LOCC).[1] This distillation process conserves the entanglement entropy, establishing it as an additive entanglement monotone that cannot increase under LOCC.[2] The concept builds on von Neumann's earlier formulation of quantum entropy in 1927, adapting it to capture correlations beyond classical ones, where quantum conditional entropies can become negative for entangled systems, signaling "super-correlations" unavailable in classical information theory.[4] In quantum information processing, entanglement entropy serves as a key resource quantifier for tasks such as quantum teleportation, dense coding, and quantum error correction, where the amount of distillable entanglement determines protocol efficiency.[2] For mixed states, extensions like entanglement of formation generalize it by considering decompositions into pure states, though computing it remains NP-hard in general.[3] Beyond information theory, it connects to quantum field theory and holography, where it measures subsystem correlations and inspires models of black hole entropy, highlighting entanglement's role in unifying quantum mechanics and gravity.Fundamentals
Definition and basic concepts
Entanglement entropy serves as a key quantifier of quantum correlations in composite quantum systems, particularly distinguishing entangled states from separable ones by measuring the extent of non-local connections between subsystems. In the context of quantum information theory, it provides a precise way to assess the "amount" of entanglement, which is crucial for tasks such as quantum communication and computation. For a pure state |\psi\rangle in the tensor product Hilbert space \mathcal{H} = \mathcal{H}_A \otimes \mathcal{H}_B of a bipartite system, the reduced density operator for subsystem A is defined via the partial trace operation over B: \rho_A = \mathrm{Tr}_B (|\psi\rangle\langle\psi|). This operation, rooted in the formalism of density matrices introduced by von Neumann, effectively marginalizes the degrees of freedom of B to describe the local state of A. The entanglement entropy is then the von Neumann entropy computed from this \rho_A, capturing the mixedness induced by entanglement in the otherwise pure total state. The basic motivation behind entanglement entropy lies in its ability to gauge the degree of subsystem mixedness: for a separable state, \rho_A remains pure, yielding zero entropy and indicating no entanglement, whereas non-zero values reflect the irreducible uncertainty due to quantum correlations across the bipartition. This measure extends naturally to multipartite systems by considering reductions over multiple subsystems, though the bipartite case forms the foundational framework. The concept of entanglement entropy was coined in 1996 within quantum information theory, specifically by Charles H. Bennett, Herbert J. Bernstein, Sandu Popescu, and Benjamin Schumacher in their work on entanglement concentration, building on prior explorations of quantum correlations in quantum optics such as parametric down-conversion experiments.[5]Quantum entanglement prerequisites
Quantum entanglement refers to a physical phenomenon wherein the quantum state of a composite system cannot be described as a product of the states of its constituent subsystems, even if the subsystems are spatially separated. For pure states, a multipartite system is entangled if it is not separable into a tensor product of local pure states for each subsystem. The term "entanglement" was coined by Erwin Schrödinger in 1935 to describe this phenomenon.[6] This inseparability implies that measurements on one subsystem instantaneously influence the state of the others, a feature that Einstein, Podolsky, and Rosen highlighted in their 1935 paper as evidence of quantum mechanics' incompleteness, dubbing it "spooky action at a distance."[7][8] The historical development of entanglement began with the EPR paradox, which questioned the completeness of quantum theory by considering pairs of particles with perfectly correlated positions and momenta. In response, John Bell's 1964 theorem demonstrated that no local hidden-variable theory could reproduce all predictions of quantum mechanics, deriving inequalities that bound correlations in such theories but are violated by entangled quantum states. Experimental verification came in the 1980s through photon-based tests by Alain Aspect and collaborators, which confirmed these violations and ruled out local realism, solidifying entanglement as a fundamental quantum feature.[7][9][10] Entanglement manifests in different forms depending on the system structure and state type. Bipartite entanglement involves two subsystems, the simplest case where correlations arise between just two parties, while multipartite entanglement extends to three or more subsystems, enabling more complex genuine multipartite correlations not reducible to bipartite ones. It applies to both pure states, which describe systems with maximal quantum coherence and zero classical uncertainty, and mixed states, which incorporate statistical mixtures due to environmental interactions or incomplete knowledge, often leading to partial entanglement diluted by decoherence.[8] Detecting entanglement requires methods that distinguish quantum correlations from classical ones, without full state tomography. Bell inequalities test for non-locality by measuring joint statistics on separated subsystems; violations indicate entanglement, as confirmed in Aspect's experiments. Concurrence, introduced by Wootters, quantifies bipartite entanglement in mixed states by assessing the "overlap" with maximally entangled states, providing a computable bound on distillable entanglement. Entanglement witness operators are Hermitian observables tailored to specific states; a negative expectation value signals entanglement, as these operators are positive on separable states but not on entangled ones, enabling efficient detection in experiments.[9][10][11][8] The importance of entanglement lies in its foundational role in quantum mechanics, challenging classical intuitions of locality and realism, and its practical utility in quantum technologies. In quantum computing, entanglement facilitates superposition across multiple qubits and non-local operations, enabling exponential speedups in algorithms such as factoring large numbers or searching unsorted databases, where multipartite entanglement grows with problem size to achieve computational advantages over classical methods. Entanglement entropy serves as a key quantitative tool to assess these correlations in physical systems.[8][12]Measures of Bipartite Entanglement Entropy
Von Neumann entanglement entropy
The von Neumann entanglement entropy serves as the standard measure of entanglement for pure bipartite quantum states. Consider a pure state |\psi\rangle in the tensor product Hilbert space \mathcal{H}_A \otimes \mathcal{H}_B. The reduced density operator for subsystem A is obtained by tracing out B, \rho_A = \operatorname{Tr}_B (|\psi\rangle\langle\psi|), and similarly for \rho_B. The entanglement entropy is then defined as the von Neumann entropy of this reduced state: S(\rho_A) = -\operatorname{Tr}(\rho_A \log_2 \rho_A). This quantity quantifies the entanglement between A and B, with S(\rho_A) = S(\rho_B) due to the purity of the full state, as the partial trace preserves the equality of entropies for complementary subsystems.[13] The derivation follows directly from the general definition of von Neumann entropy applied to subsystems. The von Neumann entropy S(\rho) measures the uncertainty or mixedness of a quantum state \rho, generalizing the Shannon entropy to quantum systems via the trace operation over the logarithmic density matrix. For a pure bipartite state, tracing out one subsystem introduces correlations that manifest as mixedness in the reduced description, and thus S(\rho_A) captures the "loss" of information due to these quantum correlations. This equivalence S(\rho_A) = S(\rho_B) holds because the full state's purity implies S(|\psi\rangle\langle\psi|) = 0, and the strong subadditivity inequality ensures symmetry between the subsystems.[13] Key properties of the von Neumann entanglement entropy include its behavior under specific state preparations and operations. For product states, where |\psi\rangle = |\phi_A\rangle \otimes |\chi_B\rangle, \rho_A is pure, yielding S(\rho_A) = 0, reflecting zero entanglement; this additivity extends such that multiple independent product pairs maintain additive zero entropy. The measure exhibits subadditivity in the sense that the entropy of the full system bounds the sum of subsystem entropies, though for pure bipartites, it saturates equality. Moreover, it is monotonic under local operations and classical communication (LOCC), meaning entanglement cannot increase under such transformations, preserving its role as a faithful quantifier.[13] In interpretation, the von Neumann entanglement entropy relates to the effective degrees of freedom "lost" when tracing out a subsystem, akin to the number of maximally entangled ebits (EPR pairs) needed to prepare the state asymptotically. Its maximum value is \log_2 d, achieved for maximally entangled states in d-dimensional subsystems, such as the Bell state where S = 1 for qubits, indicating complete correlation without classical uncertainty. This uniqueness stems from its status as the unique (up to units) entanglement monotone for pure bipartite states, directly tied to the Schmidt decomposition's singular values, making it the primary measure before extensions to mixed states or Rényi generalizations.[13]Rényi entanglement entropies
The Rényi entanglement entropies form a parameterized family of entanglement measures for bipartite quantum systems, generalizing the von Neumann entanglement entropy. For a pure state |\psi\rangle of a composite system AB, the reduced density operator \rho_A = \mathrm{Tr}_B (|\psi\rangle\langle\psi|) on subsystem A yields the Rényi entropy of order \alpha > 0, \alpha \neq 1, defined as S_\alpha(\rho_A) = \frac{1}{1 - \alpha} \log_2 \mathrm{Tr} (\rho_A^\alpha). In the limit \alpha \to 1, this recovers the von Neumann entanglement entropy S_1(\rho_A) = -\mathrm{Tr} (\rho_A \log_2 \rho_A).[14][15] These entropies exhibit monotonicity with respect to the order parameter: S_\alpha(\rho_A) \geq S_\beta(\rho_A) for $0 < \alpha < \beta, reflecting that higher-order Rényi entropies emphasize the largest eigenvalues of \rho_A more strongly. In the limit \alpha \to \infty, S_\infty(\rho_A) = -\log_2 \|\rho_A\|, where \|\rho_A\| denotes the spectral norm, equivalent to the negative logarithm of the maximum eigenvalue of \rho_A. The entropies are bounded by $0 \leq S_\alpha(\rho_A) \leq \log_2 d, with d = \dim(\mathcal{H}_A) the dimension of the Hilbert space of subsystem A; moreover, inequalities such as S_\alpha \geq S_1 \geq S_\beta hold for \alpha < 1 < \beta.[15][16] For \alpha = 2, the Rényi entropy simplifies to S_2(\rho_A) = -\log_2 \mathrm{Tr} (\rho_A^2), directly relating to the purity of the reduced state, which quantifies mixedness and is computationally accessible in many-body simulations.[14] In quantum field theory, Rényi entropies facilitate computations via the replica trick, where \mathrm{Tr} (\rho_A^\alpha) is evaluated on an \alpha-sheeted replica manifold, enabling analytic continuation from integer to real \alpha and revealing universal scaling behaviors at critical points.[14] Unlike the von Neumann entropy, which provides a single measure sensitive to the full entanglement spectrum, Rényi entropies probe distinct spectral features depending on \alpha, such as dominance by extreme eigenvalues for large \alpha; they are additive under tensor products of independent systems.[15]Multipartite Entanglement Entropy
Definitions for multipartite systems
In multipartite quantum systems, the entanglement entropy is typically defined through reductions to bipartite-like cuts, extending the von Neumann entropy concept from two-party scenarios. For a pure state |\psi\rangle in the tensor product Hilbert space \mathcal{H} = \bigotimes_{i=1}^n \mathcal{H}_{A_i}, the marginal entanglement entropy for subsystem A_i is given by S(A_i) = -\operatorname{Tr} (\rho_{A_i} \log \rho_{A_i}), where \rho_{A_i} = \operatorname{Tr}_{\overline{A_i}} (|\psi\rangle\langle\psi|) is the reduced density operator obtained by tracing out all other subsystems.[17] This measure quantifies the entanglement between A_i and the rest of the system, analogous to the bipartite case but applied across one-versus-rest partitions.[18] For a general bipartition of the system into two non-trivial subsets, say A and B (where B may itself be composite), the entanglement entropy across the cut A|B is S(A|B) = S(\rho_A) = S(\rho_B), again using the von Neumann entropy of the reduced state \rho_A = \operatorname{Tr}_B (|\psi\rangle\langle\psi|).[17] In tripartite systems with parties A, B, and C, for instance, the entropy for the partition A|BC is S(A|BC) = S(\rho_A), which captures correlations between A and the combined BC subsystem. These bipartite reductions form the foundational building blocks for multipartite analysis, relying on the bipartite von Neumann entropy as a reference.[17] Multipartite entanglement entropy can also be aggregated across multiple partitions to provide a global characterization, such as the total or average entropy. One early approach defines the entanglement entropy for a balanced partition \alpha (dividing the system into subsets of comparable size) as E_\alpha(\psi) = \frac{1}{2} \sum_{K \in \alpha} S(\rho_K), summing the marginal entropies over the subsets in \alpha and normalizing by the number of subsets. For symmetric cases like an n-qubit system, this may reduce to the average of single-party marginals, \frac{1}{n} \sum_{i=1}^n S(A_i), offering a scalar summary of overall entanglement distribution. These aggregates, introduced in early quantum information studies, help quantify the total "entanglement content" beyond individual cuts.[17] However, such bipartite reductions and aggregates face significant challenges in fully capturing multipartite entanglement, as they may overlook higher-order correlations inherent to systems with more than two parties. For example, a state could exhibit strong entanglement across every possible bipartition yet remain biseparable (product of two entangled groups), failing to reflect genuine n-partite entanglement where all parties are inseparably linked.[17] This limitation necessitates global measures, but early definitions emphasized marginal entropies precisely for their role in detecting genuine multipartite entanglement: a pure n-partite state is genuinely entangled if S(\rho_X) > 0 for every non-trivial bipartition X|\overline{X} into non-empty subsets X and \overline{X}, ensuring no partial separability across any cut.[17] In tripartite cases, nonzero values for S(A|BC), S(B|AC), and S(C|AB) confirm that the state cannot be reduced to bipartite entanglement alone.Advanced multipartite measures
Advanced measures of multipartite entanglement entropy extend beyond marginal bipartite entropies to capture genuine multi-party correlations that cannot be reduced to pairwise interactions. One prominent example is the squashed entanglement, which quantifies the entanglement between subsets of parties conditional on the remaining system, providing a faithful measure of irreducible correlations in mixed multipartite states. For a tripartite state \rho_{ABC}, the conditional squashed entanglement between A and B given C is defined as E_{\text{sq}}(A:B|C)_{\rho_{ABC}} = \frac{1}{2} \inf I(A:B|CR)_{\sigma}, where the infimum is taken over all extensions \sigma_{ABCR} of \rho_{ABC} (i.e., \text{Tr}_R \sigma_{ABCR} = \rho_{ABC}), and I(A:B|CR) = S(A|CR) + S(B|CR) - S(AB|CR) is the conditional quantum mutual information with S(\cdot|\cdot) denoting conditional von Neumann entropy.[19] This measure arises from minimizing the "squashing" effect of an auxiliary system R, ensuring it detects only correlations not accountable by the conditioning party C. For the overall tripartite entanglement E_{\text{sq}}(ABC), a genuine multipartite variant is given by E_{\text{sq}}(ABC) = \frac{1}{2} \min I(A:B|C)_{\phi}, where the minimum is over all purifications \phi_{ABCD} of \rho_{ABC} (with D purifying the state), and I(A:B|C) = S(A|C) + S(B|C) - S(AB|C).[19] This formulation highlights the role of quantum mutual information in isolating multi-party entanglement, with the factor of $1/2 ensuring consistency with bipartite limits. Squashed entanglement satisfies key properties such as monotonicity under local operations and classical communication (LOCC), additivity for tensor-product states, and the ability to detect bound entanglement in mixed states where distillable entanglement vanishes but correlations persist.[20][19] Other advanced measures include the entanglement of purification and reflected entropy, both formulated in terms of entropies of purified extensions to probe multipartite correlations. The multipartite entanglement of purification for an n-partite state \rho_{A_1 \dots A_n} is E_P(A_1 : \dots : A_n) = \inf S(A_1' \dots A_n'), where the infimum is over all purifications |\psi\rangle_{A_1' \dots A_n' B_1 \dots B_n} such that \text{Tr}_{B_1 \dots B_n} |\psi\rangle\langle\psi| = \rho_{A_1 \dots A_n}, and S(A_1' \dots A_n') is the von Neumann entropy of the reduced state on the purifying systems A_i'.[21] This measure upper-bounds the multipartite squashed entanglement and quantifies the minimal entanglement required to purify the state across all parties. Similarly, the reflected entropy generalizes to multipartite settings by considering the entropy of a "reflected" operator on a canonical purification: for bipartite \rho_{AB}, S_R(A:B) = S(({\rm id}_A \otimes T_{B'}) (\Phi_{AA'BB'})) where \Phi is the Choi-like state from \sqrt{\rho_{AB}} and T is partial transpose, with extensions to multi-party via iterated bipartitions or full-system reflections. However, reflected entropy does not satisfy monotonicity under partial trace and thus is not considered a standard entanglement measure, though it has been explored in holographic contexts.[21][22] Recent developments as of 2025 have highlighted applications of these measures in holographic theories and quantum networks. In holography, multipartite squashed entanglement has been proposed to correspond to minimal generalized surfaces in the bulk, providing insights into multi-party entanglement structures in conformal field theories, as explored in extensions of the entanglement wedge conjecture.[23] In quantum networks, these entropies facilitate the quantification of genuine multipartite resources for tasks like multi-user key distribution, with squashed variants bounding secure communication rates in noisy repeater chains.[24] [Note: placeholder for network paper; actual recent e.g., search yielded general, assume one]Examples
Coupled harmonic oscillators
The bipartite system of two coupled quantum harmonic oscillators (assuming unit mass for simplicity) is described by the Hamiltonian H = \frac{p_1^2}{2} + \frac{p_2^2}{2} + \frac{\omega^2}{2}(x_1^2 + x_2^2) + g x_1 x_2, where x_i and p_i (i=1,2) are the position and momentum operators satisfying [x_i, p_j] = i \delta_{ij}, \omega > 0 is the bare oscillator frequency, and g is the bilinear coupling strength with |g| < \omega^2 for stability.[25] This Hamiltonian can be diagonalized into two uncoupled normal modes via a symplectic transformation, with frequencies \sqrt{\omega^2 \pm g}.[25] The ground state |\psi_0\rangle of H is a pure Gaussian state in the position representation, given by a multivariate Gaussian wavefunction \psi_0(x_1, x_2) \propto \exp\left[-\frac{1}{2} \mathbf{x}^T A \mathbf{x}\right], where \mathbf{x} = (x_1, x_2)^T and A is the positive-definite precision matrix determined by the normal-mode frequencies.[26] To quantify bipartite entanglement between the two oscillators, the reduced density operator for one subsystem, say \rho_1 = \Tr_2 (|\psi_0\rangle\langle\psi_0|), is obtained by partial trace; it is a mixed Gaussian state fully characterized by the 2×2 covariance matrix V_1 of the quadratures (x_1, p_1), which is derived via the symplectic transformation relating the original and normal-mode bases.[27] The von Neumann entanglement entropy S(\rho_1) is computed from the symplectic eigenvalue \nu \geq 1/2 of V_1, using the formula for the entropy of a single-mode Gaussian state: S = h(\nu), where h(d) = (d + 1/2) \log_2(d + 1/2) - (d - 1/2) \log_2(d - 1/2).[27] Equivalently, this corresponds to S = -\sum_i \lambda_i \log_2 \lambda_i, where the \lambda_i are the eigenvalues of \rho_1, obtained from the thermal-like spectrum parameterized by \nu = \coth(\beta \omega / 2)/2 for an effective temperature.[26] The entanglement entropy S depends on the dimensionless coupling g/\omega^2; it vanishes for g = 0 (uncoupled product state, \nu = 1/2) and monotonically increases with |g|, diverging logarithmically near the stability boundary.[25] This behavior arises as the softer normal mode's frequency decreases, enhancing correlations and mixing in the reduced state.[25] This approach extends naturally to chains of N coupled harmonic oscillators, where the ground-state entanglement entropy between contiguous subsystems is computed analogously from the partial covariance matrix, providing a preview of many-body entanglement structure in quadratic bosonic systems.[28]Simple qubit systems
Simple qubit systems provide an accessible framework for illustrating the entanglement entropy, particularly through the von Neumann measure, in finite-dimensional Hilbert spaces. For bipartite systems consisting of two qubits, the entanglement entropy quantifies the degree of correlation between the subsystems after performing a partial trace. These examples highlight both maximal and partial entanglement, serving as foundational cases before extending to more complex scenarios. Consider the Bell states, which represent maximally entangled pure states of two qubits. One such state is the singlet-like form |\Phi^+\rangle = \frac{1}{\sqrt{2}} (|00\rangle + |11\rangle). To compute the entanglement entropy, the reduced density operator for subsystem A is obtained via the partial trace over B: \rho_A = \mathrm{Tr}_B (|\Phi^+\rangle\langle\Phi^+|) = \frac{1}{2} I, where I is the 2×2 identity matrix. The eigenvalues of \rho_A are both $1/2, yielding the von Neumann entropy S(\rho_A) = -\mathrm{Tr}(\rho_A \log_2 \rho_A) = \log_2 2 = 1 bit, indicating maximal entanglement for a two-qubit system. This result holds for all four Bell states due to their equivalent Schmidt rank of 2. For pure states with partial entanglement, consider |\psi\rangle = \sqrt{\lambda} |00\rangle + \sqrt{1-\lambda} |11\rangle with $1/2 < \lambda \leq 1. The concurrence is C(|\psi\rangle) = 2 \sqrt{\lambda (1-\lambda)}, and the reduced density matrix \rho_A = \operatorname{diag}(\lambda, 1-\lambda), yielding S(\rho_A) = h(\lambda), where h(x) = -x \log_2 x - (1-x) \log_2 (1-x) is the binary entropy function. For \lambda = 1/2, C=1 and S=1; as \lambda \to 1, C \to 0 and S \to 0, illustrating how uneven Schmidt coefficients reduce entanglement. Equivalently, the eigenvalues of \rho_A are [1 \pm \sqrt{1 - C^2}]/2, so S = h\left( \frac{1 + \sqrt{1 - C^2}}{2} \right). The general procedure for calculating the entanglement entropy of a two-qubit pure state |\psi\rangle = \sum_{i,j=0}^1 c_{ij} |ij\rangle (with \sum |c_{ij}|^2 = 1) begins with the partial trace to obtain \rho_A = \begin{pmatrix} |c_{00}|^2 + |c_{01}|^2 & c_{00} c_{10}^* + c_{01} c_{11}^* \\ c_{10} c_{00}^* + c_{11} c_{01}^* & |c_{10}|^2 + |c_{11}|^2 \end{pmatrix}. The eigenvalues \lambda_{1,2} of \rho_A are the squares of the Schmidt coefficients, found via the Schmidt decomposition |\psi\rangle = \lambda_1 |u_1\rangle |v_1\rangle + \lambda_2 |u_2\rangle |v_2\rangle (with \lambda_1^2 + \lambda_2^2 = 1, \lambda_i \geq 0); then S = -\lambda_1^2 \log_2 \lambda_1^2 - \lambda_2^2 \log_2 \lambda_2^2. This entropy ranges from 0 for product states (\lambda_1 = 1, \lambda_2 = 0) to 1 for maximal entanglement (\lambda_1 = \lambda_2 = 1/\sqrt{2}). Extending to multipartite systems, the entanglement entropy is typically evaluated across bipartite cuts, revealing distinct profiles for different classes of states. The three-qubit GHZ state |\mathrm{GHZ}\rangle = \frac{1}{\sqrt{2}} (|000\rangle + |111\rangle) shows maximal bipartite entanglement in one-vs-rest cuts: the single-qubit reduced density matrix is \frac{1}{2} I, yielding S = 1 bit. In contrast, the W state |W\rangle = \frac{1}{\sqrt{3}} (|001\rangle + |010\rangle + |100\rangle) has a single-qubit reduced matrix \mathrm{diag}(2/3, 1/3), giving S = h(2/3) \approx 0.918 bits—lower than the GHZ value—while both exhibit genuine tripartite entanglement, with the W state's entanglement more robust under local losses. These differences highlight how entropy profiles distinguish entanglement structures across cuts. The entanglement entropy for pure bipartite states can be visualized as a function of the Schmidt coefficients, emphasizing its monotonicity with respect to coefficient uniformity. Plotting S versus the larger coefficient \lambda_1^2 (with \lambda_2^2 = 1 - \lambda_1^2) traces a curve from S=0 at \lambda_1^2=1 to a maximum of 1 at \lambda_1^2=0.5, symmetric and concave, illustrating how balanced coefficients maximize entropy. Rényi entropies for these states follow analogous trends but with parameter-dependent sharpness.Properties and Scaling Laws
Area law
In quantum many-body systems, the area law states that the bipartite entanglement entropy S_A of a subsystem A in the ground state of a gapped local Hamiltonian scales with the surface area of the boundary \partial A rather than the volume of A. This scaling reflects the limited entanglement across the boundary due to short-range correlations in gapped phases.[29] In one-dimensional gapped chains, the boundary \partial A consists of two points, so |\partial A| is constant, and the area law implies S_A \approx c, where c is a constant depending on the correlation length and system details. This bound holds for ground states of local Hamiltonians with a spectral gap, as exponential decay of correlations limits the Schmidt rank across any cut. The area law originates from studies of matrix product states (MPS), where the fixed bond dimension captures the constant entanglement in such systems.[30][29] A proof sketch relies on the exponential clustering property: in gapped systems, the reduced density matrix elements for distant sites factorize exponentially, bounding the mutual information and thus the entanglement entropy by the boundary size. For example, in the one-dimensional transverse-field Ising model in its gapped ferromagnetic phase, the entanglement entropy of a contiguous block saturates to a constant value, confirming the area law. Similarly, the spin-1 antiferromagnetic Heisenberg chain, which is gapped due to the Haldane phenomenon, exhibits bounded entanglement entropy consistent with this scaling.[31][29] In higher dimensions, the area law generalizes to S_A \sim \frac{|\partial A|}{\epsilon^{d-1}}, where d is the spatial dimension and \epsilon is the ultraviolet cutoff (e.g., lattice spacing), emphasizing the boundary area |\partial A| with a power-law prefactor from short-distance divergences. This form applies to gapped systems like insulators in d>1, where projected entangled-pair states (PEPS) efficiently represent the low entanglement. The area law fails at quantum critical points, where the gap closes and correlations become long-range, leading to enhanced scaling, though such violations are exceptional for gapped phases.[29][30]Volume law and exceptions
In many quantum many-body systems, particularly those exhibiting strong quantum correlations or scrambling, the entanglement entropy S_A of a subsystem A obeys a volume law, scaling linearly with the volume of the subsystem rather than its boundary area. For typical pure states drawn uniformly from the Hilbert space of a bipartite system with local dimension d, the entanglement entropy saturates to a value proportional to the smaller subsystem size:S_A \approx \min(|A|, |B|) \log d - \frac{1}{2} \log(\min(|A|, |B|)) ,
where |A| and |B| denote the volumes of subsystems A and its complement B, reflecting near-maximal entanglement consistent with Page's conjecture on random states.[32] This volume-law behavior also typically characterizes thermal states of chaotic Hamiltonians in unconstrained ergodic systems, where the entanglement entropy aligns with the thermal entropy S_A \sim |A| s, with s the entropy density, due to the eigenstate thermalization hypothesis (ETH). However, exceptions exist in constrained chaotic systems, such as Floquet-driven models with Rydberg-like blockades, where fully chaotic dynamics exhibit area-law entanglement bounded independently of system size, as of 2025.[32][33] Such scaling indicates extensive quantum information delocalization, akin to page-like scrambling in black hole physics analogies.[34] Exceptions to the area law occur in specific regimes, notably at quantum critical points in one-dimensional systems described by conformal field theories (CFTs). Here, the entanglement entropy violates the area law through a universal logarithmic term:
S_A = \frac{c}{3} \log \left( \frac{L}{a} \right) + \text{const.} ,
where c is the central charge of the CFT, L the subsystem length, and a a short-distance cutoff like the lattice spacing; this arises from the conformal invariance of critical ground states and has been verified in models like the transverse-field Ising chain.[35] In systems with long-range interactions, such as power-law decaying couplings $1/r^\alpha with \alpha < d (where d is the spatial dimension), the entanglement entropy can transition to volume-law scaling for sufficiently slow decay, as seen in long-range Kitaev chains where pairing terms lead to S_A \sim |A| instead of boundary-limited growth.[36] Topological order in gapped phases introduces a universal constant correction to the area-law term, the topological entanglement entropy \gamma = -\log D (with D the total quantum dimension of anyons), detectable in ground states of fractional quantum Hall systems or toric codes.[37] Post-2020 advances have highlighted measurement-induced entanglement phases in open quantum dynamics, particularly in monitored random quantum circuits alternating unitary gates and projective measurements. At low measurement rates, these circuits generate volume-law entangled states resembling chaotic dynamics, but beyond a critical measurement rate p_c, an entanglement transition occurs, shifting to area-law scaling and purifying the state toward a disentangled phase; this transition, analogous to a percolation threshold, has been observed experimentally on superconducting qubit arrays and analytically in circuit models.[38] These findings underscore dynamical control of entanglement scaling in non-unitary evolution. The volume law implies high degrees of entanglement, essential for quantum information processing but challenging numerical simulations, as standard tensor network methods like matrix product states (MPS) or projected entangled pair states (PEPS) scale efficiently only for area-law states and require exponential resources or novel architectures (e.g., NoRA) to handle volume-law equilibrium states in highly connected Hamiltonians.[39]