Measurement-Induced Entanglement Transitions

• Measurement-induced entanglement transitions are phase changes in quantum systems where unitary dynamics create entanglement and local measurements collapse it.
• Hybrid quantum circuits provide a canonical framework, with a critical measurement rate determining the switch from volume-law to area-law entanglement.
• Key diagnostics such as entanglement entropy, mutual information, and quantum conditional mutual information reveal insights applicable to quantum error correction and algorithm design.

Searching arXiv for recent and foundational papers on measurement-induced entanglement transitions. {"query":"measurement-induced entanglement transitions hybrid quantum circuits conditional mutual information free fermion diffusive dynamics arXiv", "max_results": 10} Measurement-induced entanglement transitions are phase transitions in the structure of quantum entanglement that occur in many-body systems subject to both unitary dynamics and local measurements. In monitored quantum circuits and continuously monitored Hamiltonian systems, unitary evolution scrambles and generates entanglement, while measurements collapse local degrees of freedom and suppress it. The long-time steady state of individual quantum trajectories can therefore exhibit a sharp transition, as a function of measurement rate or strength, between a volume-law entangled phase and an area-law phase. In one dimension, this transition is not a thermal transition; it is purely about the structure of entanglement in quantum trajectories, and it is invisible in the corresponding trajectory-averaged mixed state in several standard settings (Li et al., 2019).

1. Conceptual framework

A standard setting for measurement-induced entanglement transitions is a monitored quantum circuit: an array of qubits evolving in discrete time under random unitary gates and local projective measurements. In this setting, the key control parameter is the measurement rate $p\in[0,1]$, the probability per spacetime location that a measurement is inserted. The physically relevant object is a pure trajectory state $|\psi(t)\rangle$, obtained after conditioning on a specific sequence of measurement outcomes; entanglement is then computed from reduced density matrices of $|\psi(t)\rangle$, not from the unconditional density matrix averaged over outcomes (Li et al., 2019).

The canonical distinction is between a volume-law phase and an area-law phase. For a contiguous subsystem $A$, the volume-law phase has

$S_A \sim s |A|,$

while in one dimension the area-law phase has

$S_A \sim \mathcal{O}(1).$

In hybrid circuit models, low measurement rate stabilizes the volume-law phase and high measurement rate stabilizes the area-law phase, with a critical measurement rate $p_c>0$ in generic chaotic settings (Li et al., 2019). In continuously monitored Hamiltonian systems, the analogous control parameter is often a measurement strength $\gamma$, and the same competition between scrambling and disentangling can produce a steady-state transition (Fuji et al., 2020).

A common misconception is that the transition should be visible in the mixed-state density matrix obtained after averaging over outcomes. For several paradigmatic hybrid circuit models, the average density matrix is essentially infinite temperature for all $p>0$, and the nontrivial phase structure resides only in single-trajectory pure states (Li et al., 2019). Another common misconception is that the transition is simply a thermal or localization transition. In fact, it is a transition in entanglement structure, and it can occur in systems with random circuits, Floquet dynamics, integrable Hamiltonians, nonintegrable Hamiltonians, many-body localized dynamics, and explicitly non-Hermitian effective evolution, depending on the monitoring protocol (Li et al., 2019).

2. Models and realizations

The foundational one-dimensional circuit realization is the hybrid brickwork circuit with two-qubit unitary gates and stochastic local measurements. In the random Clifford workhorse model, local unitaries are random 2-qubit Clifford gates, single-qubit projective Pauli-$Z$ measurements are inserted independently at each spacetime location with probability $|\psi(t)\rangle$0, and stabilizer methods allow simulation up to $|\psi(t)\rangle$1 qubits (Li et al., 2019). Variants include Floquet Clifford circuits with no randomness in the unitaries, random Clifford circuits with quasi-periodic measurement placement, and fully space-time symmetric Clifford circuits in which only the measurement outcomes are stochastic; all exhibit a sharp entanglement transition (Li et al., 2019).

Beyond Clifford, small-system simulations of random Haar circuits and Floquet non-Clifford circuits also display the same qualitative transition. In these cases, mutual information between antipodal sites peaks at a critical measurement rate, and for all Rényi indices $|\psi(t)\rangle$2 the peak location is approximately independent of $|\psi(t)\rangle$3, while the zeroth Rényi entropy $|\psi(t)\rangle$4 can behave differently (Li et al., 2019). This distinction between $|\psi(t)\rangle$5 and $|\psi(t)\rangle$6 recurs across the literature.

Hamiltonian realizations replace random circuit layers by deterministic many-body evolution under monitored quantum dynamics. In a one-dimensional hard-core boson chain with continuous position measurement, the stochastic Schrödinger equation with jump operators $|\psi(t)\rangle$7 produces a transition in the steady-state trajectory entanglement as the monitoring strength $|\psi(t)\rangle$8 is tuned (Fuji et al., 2020). In the quantum Ising chain, two different unravellings—stochastic quantum-state diffusion and the no-click limit described by a non-Hermitian Hamiltonian—both exhibit a transition from a critical phase with logarithmic scaling of the entanglement to an area-law phase as $|\psi(t)\rangle$9 is increased, and both occur at the same value of the measurement rate $|\psi(t)\rangle$0 (Turkeshi et al., 2021).

Many-body localized systems provide a distinctive Hamiltonian setting because emergent l-bits make the measurement basis relevant. In a disordered Heisenberg chain deep in the MBL regime, projective measurements in the $|\psi(t)\rangle$1 basis have $|\psi(t)\rangle$2, while $|\psi(t)\rangle$3-basis measurements exhibit a finite critical point $|\psi(t)\rangle$4 (Lunt et al., 2020). This basis dependence is absent in Haar-random circuits, where all local operators are scrambled in time.

Long-range and nonlocal architectures further broaden the class of realizations. Monitored variable-range Clifford circuits use a distance-dependent probability $|\psi(t)\rangle$5 for choosing gate pairs on a ring. For $|\psi(t)\rangle$6, these circuits show a transition between volume-law and area-law phases, whereas for $|\psi(t)\rangle$7 no area-law phase is found for any $|\psi(t)\rangle$8 (Otsuka et al., 4 Apr 2025). In a Brownian all-to-all circuit with local measurements, the large-$|\psi(t)\rangle$9 problem maps to a one-dimensional quantum chain in the semiclassical limit, and the transition is controlled by the shape of an effective potential as a function of $A$0 (Yu et al., 2022).

3. Phases, scaling laws, and critical diagnostics

In one-dimensional hybrid circuits, the low-measurement phase is a stable volume-law phase, but its entanglement structure is not simply the Page law of a purely unitary chaotic circuit. For $A$1, the steady-state entanglement entropy of a contiguous region obeys

$A$2

so the volume-law piece is accompanied by a robust subleading logarithmic correction whenever $A$3 (Li et al., 2019). At the critical point, the extensive coefficient vanishes and one finds a purely logarithmic law,

$A$4

with $A$5 in bits for the Clifford models studied there (Li et al., 2019).

The area-law phase has finite entanglement at large subsystem size. In the Clifford stabilizer description, this behavior is tied to the stabilizer-length distribution. In the clipped gauge, the normalized length distribution has a power-law $A$6 tail throughout the volume-law phase, a pure $A$7 form at criticality, and an exponentially cut off form in the area-law phase. This directly explains why logarithmic entanglement persists at criticality and why the volume-law phase carries a logarithmic correction (Li et al., 2019).

Finite-size scaling yields a correlation-length exponent. In the original hybrid circuit study, the scaling form

$A$8

led to $A$9 in the random Clifford circuit and essentially the same value in all Clifford variants (Li et al., 2019). In monitored variable-range Clifford circuits, QCMI-based finite-size scaling gives $S_A \sim s |A|,$0 for $S_A \sim s |A|,$1, in excellent agreement with the nearest-neighbor Clifford circuit result $S_A \sim s |A|,$2, while also providing a thermodynamic-limit estimate $S_A \sim s |A|,$3 for the coefficient of the logarithmic critical entropy in the shortest-range case (Otsuka et al., 4 Apr 2025).

Mutual information is a central diagnostic. For antipodal intervals, it develops a peak that sharpens with increasing $S_A \sim s |A|,$4, and away from criticality it decays exponentially with system size,

$S_A \sim s |A|,$5

with $S_A \sim s |A|,$6 (Li et al., 2019). At criticality, mutual information in one dimension obeys conformal-cross-ratio scaling,

$S_A \sim s |A|,$7

and for small $S_A \sim s |A|,$8 one finds

$S_A \sim s |A|,$9

so two distant single sites satisfy

$S_A \sim \mathcal{O}(1).$0

These are among the strongest pieces of evidence for emergent conformal invariance at the transition in short-range monitored circuits (Li et al., 2019).

Quantum conditional mutual information provides a more specialized but especially clean diagnostic. With carefully chosen partitions, it is extensive in the volume-law phase, vanishes in the area-law phase, and becomes size-independent at criticality because the logarithmic critical terms cancel. In the monitored variable-range Clifford circuit, this makes QCMI suitable for locating $S_A \sim \mathcal{O}(1).$1 and extracting the universal coefficient of the logarithmic entropy by crossing-point analysis (Otsuka et al., 4 Apr 2025).

4. Universality classes, extensions, and controversies

A central issue in the field is universality. Short-range one-dimensional Clifford and non-Clifford hybrid circuits exhibit a robust universality class with logarithmic critical entanglement, $S_A \sim \mathcal{O}(1).$2, and $S_A \sim \mathcal{O}(1).$3, apparently insensitive to randomness versus periodicity in the unitary gates and randomness versus periodicity in measurement locations (Li et al., 2019). Variational quantum circuits show the same qualitative transition, and the extracted exponents $S_A \sim \mathcal{O}(1).$4 for an XXZ Hamiltonian variational ansatz and $S_A \sim \mathcal{O}(1).$5 for a hardware-efficient ansatz are consistent with the random-circuit universality class (Wiersema et al., 2021).

Percolation occupies a special but limited role. Earlier work mapped the zeroth Rényi entropy $S_A \sim \mathcal{O}(1).$6 to a bond percolation or first-passage percolation problem, but this mapping does not apply to Rényi entropies with $S_A \sim \mathcal{O}(1).$7. The distinction is explicit in several models: in generalized-measurement settings one can observe a transition in $S_A \sim \mathcal{O}(1).$8 while $S_A \sim \mathcal{O}(1).$9 shows no transition, and the critical logarithmic coefficient in Clifford circuits differs from the percolation prediction (Li et al., 2019). This establishes that percolation is not the correct universality class for the physically relevant entanglement measures $p_c>0$0.

Free-fermion systems remain the most prominent controversy. One line of work argues that monitored free fermions with standard Gaussian dynamics do not support a stable volume-law phase in one dimension, and that apparent transitions are finite-size crossovers. Another line finds BKT-type or logarithmic critical phases under continuous monitoring, long-range hopping, disorder, or non-Hermitian deformations. A recent review emphasizes this controversy, including the role of non-Hermitian skin effects and the sensitivity to protocol and observable (Li et al., 27 Mar 2025). A distinct resolution is proposed in work on non-Gaussian monitoring: a stable MIPT with a genuine volume-law phase requires non-Gaussianity of quantum trajectories, and non-Gaussianity from measurements alone is sufficient to restore the transition, even in a free-fermion system (Lumia et al., 2023).

Long-range interactions and correlated measurement patterns can change the universality class. In monitored variable-range Clifford circuits, $p_c>0$1 is compatible with short-range conformal-like criticality, but for $p_c>0$2 the QCMI-extracted coefficient $p_c>0$3 begins to depend on the partition, suggesting that emergent conformal invariance and simple $p_c>0$4 scaling start to break down (Otsuka et al., 4 Apr 2025). In a random Clifford circuit monitored by classically diffusing particles, diffusive correlations in the measurement density are a relevant perturbation to the usual space-time random MIPT critical point and produce a new universality class with altered dynamical scaling and Griffiths-like effects (Ha et al., 2024).

The transition can also intertwine with topology and symmetry. In a $p_c>0$5D symmetric random circuit with $p_c>0$6 symmetry, competing stabilizer measurements and trivializing measurements produce symmetry-protected topological, trivial, and volume-law phases. In the absence of unitary dynamics there is a purely measurement-induced critical point at $p_c>0$7 with logarithmic entanglement scaling, exactly mapped to two copies of a classical 2D percolation problem; once unitaries are added, this point becomes a tricritical point separating two critical lines with an intervening volume-law phase (Fuji et al., 2020).

Measurement protocol itself can alter universality. In the monitored quantum Ising chain, stochastic quantum-state diffusion and the no-click limit share the same critical measurement strength $p_c>0$8, but the effective central charges extracted from logarithmic entanglement scaling do not match near the transition, suggesting different universality classes for the two protocols (Turkeshi et al., 2021).

5. Diagnostics, experiments, and applications

Direct measurement of entanglement entropy is difficult, and much of the recent literature develops alternatives. One approach exploits conserved quantities. In a monitored $p_c>0$9-symmetric random circuit, the variance of subsystem charge,

$\gamma$0

and the mutual fluctuation

$\gamma$1

follow the same finite-size scaling as bipartite entanglement entropy and mutual information. This provides an “exponential shortcut” because tomography requires exponentially many measurement settings in subsystem size, whereas fluctuation measurements require only computational-basis readout and repeated sampling (Moghaddam et al., 2023). Remarkably, the transition can be revealed by measuring fluctuations of only a handful of qubits (Moghaddam et al., 2023).

A distinct route is algorithmic rather than observable-based. In monitored random circuits, a hybrid quantum-classical “unitary mirror” constructs a matrix product state approximation $\gamma$2 of a trajectory state $\gamma$3, builds a unitary mirror circuit $\gamma$4 such that $\gamma$5, and then measures the overlap

$\gamma$6

Polynomial-sized tensor networks can represent area-law states but fail exponentially in the volume-law phase, so the breakdown of the mirror can locate the transition and provide an upper bound on half-chain entropy in terms of $\gamma$7 and bond dimension $\gamma$8 (Yanay et al., 2024).

The first direct experimental realization on superconducting hardware implemented a non-Clifford hybrid random circuit with true mid-circuit measurements. By varying the rate of projective measurements, the experiment directly observed extensive and sub-extensive scaling of entanglement entropy in the volume-law and area-law phases, respectively, and performed a finite-size scaling collapse for different system sizes (Koh et al., 2022). Weak measurements implemented through ancilla-assisted null-type measurement channels produced a ridge of enhanced entropy variance in the $\gamma$9 plane, extending the projective-measurement phase diagram (Koh et al., 2022).

Measurement-induced entanglement transitions also intersect with quantum algorithms. In variational quantum circuits, intermediate projective measurements induce a transition from volume-law to area-law entanglement as the measurement rate increases. The same transition coincides with a landscape transition in classical optimization: for $p>0$0, the gradient variance decays exponentially with system size, signaling severe barren plateaus, whereas for $p>0$1 it becomes essentially independent of system size (Wiersema et al., 2021). This suggests that controlled intermediate measurements can act as an entanglement-management tool in variational quantum algorithms.

A broader generalization replaces projective monitoring by quantum-data collection. In a 1D random brickwork circuit, after each layer a fraction $p>0$2 of sites undergo noisy quantum transduction that transfers quantum information from the system to a quantum computer while introducing noise from an environment. Under the condition that the environment obtains the same amount of information as gained by the computer, the circuit shows a transition from volume-law to area-law entanglement as $p>0$3 is increased above a critical threshold (Kelly et al., 2023). This reframes standard MIPTs as one instance of a more general information-exchange–driven entanglement transition.

6. Broader significance and open directions

Measurement-induced entanglement transitions establish that scrambling and thermalization are not inevitable in monitored many-body quantum systems. Even strongly chaotic unitary dynamics can be held in an area-law phase by sufficiently frequent monitoring, while weak measurements can preserve a stable volume-law phase (Li et al., 2019). This has immediate conceptual consequences for open-system dynamics, non-Hermitian many-body physics, quantum error correction, and hybrid quantum information processing.

Several open questions remain central. One is the precise field theory of the short-range critical point. The logarithmic entanglement, conformal-cross-ratio scaling of mutual information, and critical exponents suggest an effective two-dimensional statistical mechanics description, but the microscopic mapping for $p>0$4 remains incomplete (Li et al., 2019). A second is dimensionality: most precise results are in one dimension, while higher-dimensional transitions remain less understood (Li et al., 2019). A third is the dynamical exponent and transient scaling near criticality, especially in settings with long-range couplings, diffusive monitoring fields, or non-Hermitian constraints (Ha et al., 2024).

Free-fermion systems remain unresolved territory. The literature documents an explicit controversy over whether one-dimensional monitored free fermions support a genuine thermodynamic transition, how this depends on Gaussianity, and when non-Hermitian skin effects or measurement-induced non-Gaussianity alter the outcome (Li et al., 27 Mar 2025). This suggests that “measurement-induced entanglement transition” is not a single rigid phenomenon but a family of transitions whose existence and universality class depend on the scrambling structure of the unitary dynamics, the monitoring channel, and the symmetry content of the model.

A final recurring theme is that the most accessible diagnostics are often not entanglement entropies themselves. QCMI, mutual information, subsystem fluctuations, ancilla probes, and tensor-network mirror fidelities all exploit cancellations, symmetry, or classical simulability to expose the same transition in more practical ways (Otsuka et al., 4 Apr 2025). This suggests that the long-term development of the subject will likely proceed through a combination of field-theoretic characterization, trajectory-level diagnostics, and experimentally scalable probes rather than through direct entropy tomography alone.