Measurement-Induced Entanglement

• Measurement-induced entanglement is defined by how quantum measurement backaction can actively generate, recover, or reshape entanglement, employing conditional entanglement entropy as a key metric.
• Experimental protocols in few-body systems and circuit QED demonstrate that partial measurements can restore entanglement, with recovered states even violating the CHSH inequality.
• In hybrid and many-body circuits, the interplay between unitary dynamics and projective measurements governs phase transitions and scaling, highlighting practical noise effects and operator entanglement.

Measurement-induced entanglement denotes entanglement that is created, recovered, reshaped, or operationally revealed by quantum measurement backaction rather than by unitary dynamics alone. Across the literature, the term is used in several closely related senses: conditional entanglement generation in few-body protocols, recovery of entanglement degraded by decoherence, average post-measurement entanglement between unmeasured regions of a many-body state, and entanglement structures arising in hybrid circuits with interspersed unitaries and measurements. In one explicit many-body definition, if a subsystem $C$ is measured with outcome ${\bf m}$ occurring with Born probability $p_{\bf m}$, then the measurement-induced entanglement of region $A$ is $$\mathrm{MIE}(A)=\sum_{\bf m} p_{\bf m} S_{\bf m}(A)$$, with $S_{\bf m}(A)$ the conditional entanglement entropy in the post-measurement state (Khanna et al., 4 Aug 2025).

1. Operational meaning and measurement backaction

A foundational description separates measurement into an entangling interaction between system and meter and a subsequent meter read-out. For an observable $\hat A$ with eigenstates $|a\rangle$, the interaction can be written as $$\hat U_{SM}|a\rangle_S|\Phi_0\rangle_M=|a\rangle_S|\phi(a)\rangle_M$$, so an initial superposition becomes $$\hat U_{SM}|\psi\rangle_S|\Phi_0\rangle_M=\sum_a c_a |a\rangle_S |\phi(a)\rangle_M$$. The conditional meter states ${\bf m}$0 encode both measurement resolution and measurement-induced decoherence. In this formulation, the decoherence matrix is ${\bf m}$1, the resolution is the squared Hellinger distance ${\bf m}$2, and the Hilbert-space algebra gives the bound ${\bf m}$3. For the irreversible part of the disturbance, the paper further states ${\bf m}$4 in the maximally coherent case (Patekar et al., 2019).

This framework clarifies why measurements can either suppress or generate useful entanglement. Read-out bases that maximize distinguishability realize maximal information gain together with maximal irreversible decoherence, while quantum-eraser-type read-outs can set ${\bf m}$5 and ${\bf m}$6. This suggests that measurement-induced entanglement is not a contradiction of wavefunction collapse; rather, it is a controlled consequence of how collapse is conditioned, reversed, or spatially redistributed by the measurement protocol (Patekar et al., 2019).

2. Few-body generation, heralding, and recovery

A direct experimental realization of entanglement recovery was reported for photon pairs in a pure dephasing environment. Starting from the Bell state ${\bf m}$7, one photon passes through quartz plates that induce frequency-dependent phase randomization, producing a density matrix with coherence factor ${\bf m}$8. Without measurement, the concurrence decays as ${\bf m}$9. With a sequence comprising a partial measurement, further dephasing, and a quantum eraser, entanglement can be partially or fully restored; for suitably matched eraser parameters $p_{\bf m}$0, the concurrence recovery can reach up to $p_{\bf m}$1. The recovered states were shown to violate the CHSH inequality, with $p_{\bf m}$2 and $p_{\bf m}$3, and the experiment exhibited entanglement sudden death and rebirth, including revival after $p_{\bf m}$4 (Xu et al., 2010).

Heralded generation protocols exploit the same logic in circuit QED. In one proposal, a single photon traverses a Mach-Zehnder interferometer with a transmon qubit in each arm. Dispersive cavity scattering produces a qubit-state-dependent phase shift, and when the relative phase satisfies $p_{\bf m}$5, the output port implements a parity measurement: detector $p_{\bf m}$6 projects onto $p_{\bf m}$7, detector $p_{\bf m}$8 onto $p_{\bf m}$9. Because entanglement is heralded by photon detection, photon loss lowers the success probability but does not lower the fidelity of successful events (Ohm et al., 2015).

A related bad-cavity protocol continuously measures the amplitude of the field transmitted through a cavity containing two qubits. In the bad-cavity limit, adiabatic elimination yields an effective collective decay $A$0, and the spin-singlet acts as a dark state of the cavity-mediated dynamics. Homodyne monitoring probabilistically selects trajectories consistent with the singlet, with infidelity scaling linearly with the qubit-decoherence rate, $A$1, while more selective post-processing can give $A$2 (Julsgaard et al., 2012).

Measurement can also increase spatial entanglement in discrete-time quantum walks. After a two-dimensional walk, projective measurement of the coin degree of freedom yields an average $A$3-$A$4 entanglement $A$5 that can exceed the value obtained by tracing out the coin, and for optimal measurement bases the alternate quantum walk and the Grover walk were reported to produce exactly the same maximal spatial entanglement for the same number of steps (Franco et al., 2013).

3. Hybrid circuits and measurement-induced phase transitions

In monitored many-body circuits, measurement-induced entanglement is governed by the competition between entangling unitary dynamics and disentangling projective measurements. The canonical transition separates a volume-law phase from an area-law phase at a critical measurement rate. For circuits with a conserved quantity, a central practical result is that fluctuations of a conserved subsystem observable mirror the scaling of entanglement entropy. In a $A$6-conserving chain, the subsystem variance $A$7 and the mutual fluctuation $A$8 display the same volume-law, area-law, and critical scaling structure as entanglement entropy and mutual information. The paper emphasizes that this gives an exponential shortcut: for a subsystem of size $A$9, fluctuation measurements need only $$\mathrm{MIE}(A)=\sum_{\bf m} p_{\bf m} S_{\bf m}(A)$$0 settings rather than $$\mathrm{MIE}(A)=\sum_{\bf m} p_{\bf m} S_{\bf m}(A)$$1, and the phase transition can be revealed by measuring fluctuations of only a handful of qubits (Moghaddam et al., 2023).

The same competition appears in structured variational circuits. For the Hamiltonian Variational Ansatz for the XXZ model and the Hardware Efficient Ansatz, intermediate local projective measurements induce a transition from volume-law to area-law entanglement. Finite-size scaling was reported in the form $$\mathrm{MIE}(A)=\sum_{\bf m} p_{\bf m} S_{\bf m}(A)$$2, with $$\mathrm{MIE}(A)=\sum_{\bf m} p_{\bf m} S_{\bf m}(A)$$3 for XXZ-HVA and $$\mathrm{MIE}(A)=\sum_{\bf m} p_{\bf m} S_{\bf m}(A)$$4 for HEA, and critical exponents $$\mathrm{MIE}(A)=\sum_{\bf m} p_{\bf m} S_{\bf m}(A)$$5 and $$\mathrm{MIE}(A)=\sum_{\bf m} p_{\bf m} S_{\bf m}(A)$$6, respectively. The same work connects the transition to a landscape transition in trainability: above $$\mathrm{MIE}(A)=\sum_{\bf m} p_{\bf m} S_{\bf m}(A)$$7, gradient variance stops decaying exponentially with $$\mathrm{MIE}(A)=\sum_{\bf m} p_{\bf m} S_{\bf m}(A)$$8, mitigating barren plateaus (Wiersema et al., 2021).

The universality of the transition depends on the scrambling properties of the unitary dynamics. In many-body localized systems, measurements in a scrambled basis such as $$\mathrm{MIE}(A)=\sum_{\bf m} p_{\bf m} S_{\bf m}(A)$$9 yield a finite critical probability $S_{\bf m}(A)$0, while measurements in a basis aligned with the l-bits yield $S_{\bf m}(A)$1. At the finite-$S_{\bf m}(A)$2 transition, the reported critical exponents are $S_{\bf m}(A)$3 and $S_{\bf m}(A)$4, with Rényi entropies scaling as $S_{\bf m}(A)$5. In an all-to-all Brownian bilocal circuit, the transition instead occurs at $S_{\bf m}(A)$6, separating a “cusp phase” with a Page-curve-like first-derivative discontinuity from a “smooth phase” with a smooth entropy curve. For free fermions, by contrast, the review literature highlights a controversy: some field-theory and numerical results support a measurement-induced transition in one dimension, while other analytical and large-scale studies indicate that any finite measurement rate may ultimately produce area-law scaling in the thermodynamic limit; the same review emphasizes the role of non-Hermitian skin effects and higher-dimensional extensions (Lunt et al., 2020, Yu et al., 2022, Li et al., 27 Mar 2025).

4. Basis dependence, Born averaging, and sign structure

In quantum critical one-dimensional systems described by Tomonaga-Luttinger liquids, MIE admits an analytic conformal-field-theory treatment. For measurements of the local charge operator in two disjoint regions, the resulting MIE is described as universal and conformally invariant, depending on the geometry through the conformal cross ratio and on microscopic data only through the Luttinger parameter $S_{\bf m}(A)$7. The calculation uses a double replica trick to average the nonlinear entropy over Born-distributed outcomes, and the resulting exact expression for Rényi-MIE shows that physical MIE is fundamentally different from the entanglement obtained by forcing a specific measurement outcome. The field-theory interpretation is a Born-weighted average over conformally invariant boundary conditions rather than a single fixed boundary condition (Khanna et al., 4 Aug 2025).

The same framework extends from the mean to the full statistics. Closed-form expressions have been derived for all cumulants of MIE in Tomonaga-Luttinger liquids, with Born-averaging over microscopic charge-basis outcomes becoming equivalent at low energy to averaging over conformal boundary conditions weighted by their partition functions. In the maximally separated-interval regime, the cumulants obey a universal scaling $S_{\bf m}(A)$8 for sufficiently large $S_{\bf m}(A)$9. The full distribution of post-measurement entanglement entropy is reported to be generically bimodal and to exhibit fat-tails (Khanna et al., 15 Dec 2025).

A separate line of work uses MIE to diagnose sign structure. For sign-free stabilizer states, the bound $\hat A$0 was proved, and for sign-free qubit wavefunctions a two-qubit bound $\hat A$1 was established. These bounds can be violated in sign-structured states: the cluster state and gapless symmetry-protected topological states can yield non-decaying MIE after measuring the complement, and at critical points of Haar and Clifford hybrid circuits the reported scaling dimensions satisfy $\hat A$2, in explicit violation of the sign-free inequality $\hat A$3 (Lin et al., 2022).

Protocol dependence is equally pronounced in ground-state projection problems. In the quantum Ising chain, stochastic local measurements define different projected ensembles depending on whether outcomes are sampled by the Born rule or forced. Large-scale simulations reported that forced on-site measurements can enhance both bipartite and multipartite entanglement, especially in the paramagnetic regime for the forced-down protocol, whereas Born-rule and forced-up protocols generally reduce entanglement monotonically with the measurement density $\hat A$4. This directly contradicts the common simplification that local projective measurement merely disentangles degrees of freedom (Paviglianiti et al., 2023).

5. Multipartite and long-range structure

The multipartite content of measurement-induced entanglement is sharper than its bipartite manifestations. In one-dimensional hybrid Haar circuits, a graphical representation based on spanning graphs was introduced to track genuine multipartite entanglement among distant subregions. The reported numerics found genuine three-party entanglement at all separations, and at criticality the decay is consistent with a power law whose tripartite exponent is strictly larger than the corresponding exponent for bipartite logarithmic negativity. The same framework was extended to four-party diagnostics, emphasizing that monitored dynamics can sustain long-range genuine multipartite entanglement beyond what is typical in non-monitored systems (Avakian et al., 2024).

A more systematic hierarchy was developed in terms of entanglement clusters and measure-weighted graphs. Near measurement-induced phase transitions, the proposal is that $\hat A$5-party genuine multiparty entanglement decays as $\hat A$6, with an infinite hierarchy of entanglement exponents. Three general relations were conjectured: classical dominance, $\hat A$7; monotonicity, $\hat A$8; and subadditivity, $\hat A$9. In a one-dimensional measurement-only circuit that maps to percolation, non-unitary conformal field theory gives exact exponents $|a\rangle$0; in a two-dimensional measurement-only circuit mapping to classical three-dimensional percolation, the first numerically extracted exponents are consistent with $|a\rangle$1 (Allen et al., 15 Sep 2025).

Collective monitored systems realize a distinct multipartite scenario because both the unitary dynamics and the measurement are collectively entangling. For an ensemble of spin-$|a\rangle$2 particles undergoing scrambling collective dynamics and collective Gaussian measurements of $|a\rangle$3, the average Quantum Fisher Information displays three regimes as a function of the monitoring strength. Weak and strong measurements both give extensive QFI density with Heisenberg scaling, while an intermediate regime yields sub-Heisenberg scaling and more classical-like states. Here the appropriate diagnostic is $|a\rangle$4, rather than bipartite entropy (Poggi et al., 2023).

6. Experimental access, noise, and simulability

Large-scale experimental access to measurement-induced entanglement in superconducting hardware has relied on dual descriptions of monitored dynamics. On noisy quantum processors with up to $|a\rangle$5 superconducting qubits, a space-time duality mapping was used to avoid mid-circuit measurements while accessing entanglement scaling and measurement-induced teleportation in a unified framework. The measurements included second Rényi entropies, $|a\rangle$6, and a decoding-based order parameter $|a\rangle$7 with proxy entropy $|a\rangle$8. The experiment reported finite-size signatures of a phase transition and showed that the entangling and disentangling phases have sharply different sensitivity to noise; the same difference in noise sensitivity was used as a diagnostic of the phase structure (Hoke et al., 2023).

Noise also strongly constrains measurement-induced entanglement in two-dimensional random Clifford circuits. For column-by-column sampling, the operator entanglement of the boundary state, $|a\rangle$9, exhibits a finite-depth area-to-volume-law transition in the noiseless limit. With on-site probabilistic trace noise at any constant rate $$\hat U_{SM}|a\rangle_S|\Phi_0\rangle_M=|a\rangle_S|\phi(a)\rangle_M$$0, however, the maximal $$\hat U_{SM}|a\rangle_S|\Phi_0\rangle_M=|a\rangle_S|\phi(a)\rangle_M$$1 obeys an area law in the boundary length and scales approximately as $$\hat U_{SM}|a\rangle_S|\Phi_0\rangle_M=|a\rangle_S|\phi(a)\rangle_M$$2, while stabilizer generators become exponentially localized and conditional mutual information decays exponentially across buffered tripartitions. The paper concludes that constant local noise destroys long-range, volume-law measurement-induced entanglement in these 2D circuits and conjectures efficient tensor-network sampling in the corresponding noisy regimes (Wei et al., 14 Oct 2025).

Experimental feasibility is therefore highly protocol dependent. Some diagnostics reduce the measurement overhead from exponential to linear in subsystem size, some monitored phases can be accessed without mid-circuit readout by duality mappings, and some noisy regimes remain classically tractable because measurement-induced operator entanglement is forced into an area law. A distinct application even treats bipartite measurement-induced entanglement as a possible signature of non-classicality in coincident gravitational-wave detections, with a normalized entanglement entropy on the order of a few percent of the mean number of gravitons interacting with the detectors (Moghaddam et al., 2023, Hoke et al., 2023, Wei et al., 14 Oct 2025, Jones et al., 2024).