Entanglement Swapping Protocol (ESP)

• ESP is a protocol that converts two entangled pairs into a new entangled connection between non-interacting subsystems using Bell or generalized measurements.
• It encompasses techniques with pure, partially entangled, and mixed-state inputs, highlighting trade-offs between local predictability, coherence, and overall entanglement fidelity.
• The protocol underpins quantum networking by serving as a repeater primitive with applications in fault-tolerant communication, dynamic scheduling, and experimental realizations across various platforms.

Searching arXiv for recent and foundational papers on entanglement swapping and related protocol generalizations. arXiv search query: "entanglement swapping protocol many-body partially entangled qudits quantum networks" Entanglement swapping protocol (ESP) denotes the process by which two entangled pairs are converted into a new entangled pair between subsystems that never interacted directly. In the standard setting, two qubits belonging to an intermediate node are jointly measured in an entangled basis, and the remote qubits become entangled conditionally on that outcome. In contemporary usage, the term also covers a broader family of protocols that replace Bell pairs by partially entangled, mixed, continuous-variable, qudit, or many-body resources, and that reinterpret the intermediate operation as Bell-basis measurement, generalized Bell measurement, inverse-unitary measurement, or other equivalent measurement-induced projections (Maziero et al., 2022, Starke et al., 1 Aug 2025, Abdi et al., 2013, Huhtanen et al., 27 Jun 2025).

1. Canonical two-pair protocol

In the standard four-system formulation, one starts from two bipartite entangled pairs and performs a Bell-state measurement on the two middle systems. For ideal Bell-pair inputs, the basic decomposition is

$$|\Psi\rangle_{1234}=|\Phi^+\rangle_{12}\,|\Phi^+\rangle_{34} =\frac{1}{2}\Big( |\Phi^+\rangle_{14}|\Phi^+\rangle_{23} +|\Phi^-\rangle_{14}|\Phi^-\rangle_{23} +|\Psi^+\rangle_{14}|\Psi^+\rangle_{23} +|\Psi^-\rangle_{14}|\Psi^-\rangle_{23} \Big),$$

so a Bell-state measurement on the middle pair projects the remote pair into the corresponding Bell state (Beccaceci et al., 11 Dec 2025). In the polarization-qubit formulation, the Bell basis is

$$|\Phi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HH\rangle\pm |VV\rangle\bigr),\qquad |\Psi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HV\rangle\pm |VH\rangle\bigr),$$

and an ideal singlet projection on the middle photons yields a singlet on the remote photons (Trávníček et al., 2020).

For arbitrary pure two-qubit inputs,

$|\xi\rangle_{AC},\qquad |\eta\rangle_{C'B},$

Charlie performs a Bell-basis measurement on $C,C'$, and the remote qubits $A,B$ collapse into one of four conditional states. The paper on local predictability and coherence gives the explicit concurrence law

$$E_{con}(\text{output})=\frac{E_{con}(\xi_{AC})E_{con}(\eta_{C'B})}{4Pr(\text{Bell outcome})},$$

together with the average distributed entanglement

$$\langle E_{conc}\rangle =E_{con}(\xi_{AC})E_{con}(\eta_{C'B}) =\sqrt{1-P_C^2-C_C^2}\sqrt{1-P_{C'}^2-C_{C'}^2}.$$

This identifies the Bell measurement as the mechanism that converts local structure in the measured qubits into remote entanglement (Maziero et al., 2022).

Within quantum-repeater language, this canonical protocol is the link-connection primitive: adjacent entangled pairs are generated first, and swapping then converts short-range links into a longer-range link. A hardware realization on IBM ibmqx4 prepared two $|\Phi^+\rangle$ pairs and used an equivalent circuit so that the initial pairing $(A_1,B_1)(A_2,B_2)$ was transformed into $(A_1,A_2)(B_1,B_2)$, with reconstructed fidelities $$|\Phi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HH\rangle\pm |VV\rangle\bigr),\qquad |\Psi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HV\rangle\pm |VH\rangle\bigr),$$0 and $$|\Phi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HH\rangle\pm |VV\rangle\bigr),\qquad |\Psi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HV\rangle\pm |VH\rangle\bigr),$$1 (Behera et al., 2017).

2. Partially entangled and mixed-state inputs

A central development in the ESP literature is the move away from maximally entangled inputs. For pure two-qubit states with arbitrary local predictability and local coherence, the complete complementarity relation

$$|\Phi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HH\rangle\pm |VV\rangle\bigr),\qquad |\Psi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HV\rangle\pm |VH\rangle\bigr),$$2

makes the resource tradeoff explicit: higher local predictability or local coherence reduces the entanglement available on average for swapping, because those quantities consume the unit budget in the complementarity relation (Maziero et al., 2022).

For partially entangled pure states of arbitrary Schmidt rank, deterministic entanglement swapping can be reformulated as a state-discrimination problem. The standard deterministic protocol is equivalent to optimal minimum-error discrimination, while a probabilistic enhancement is obtained by optimal maximum-confidence discrimination, including sequential maximum-confidence measurements. In that formulation, deterministic swapping is always available, but the amount of entanglement in the swapped state is, in general, lower than the maximum achievable for the quantum channels involved; by contrast, successful maximum-confidence branches yield higher entanglement and fidelity than the deterministic minimum-error branch, at the price of nondeterminism (Solís-Prosser et al., 2014).

A related result is that Bell-state discrimination is not always required for attaining the EPR projection. In the probabilistic framework with partially entangled initial pairs and a generally nonmaximal measurement basis, the maximal probability of accomplishing an EPR projection becomes a constant for concurrences higher than or equal to a threshold entanglement value, so maximal entanglement in the measuring-basis is not required for attaining the EPR projection (Oppliger et al., 2021). This directly qualifies the common assumption that Bell measurement is the unique optimal intermediate resource.

The mixed-state case is analytically tractable for arbitrary two-qubit inputs. The general solution for swapping of arbitrary two-qubit states shows that the concurrence of a partially mixed state is conserved when this state is swapped with a Bell state; it also establishes upper and lower bounds on the concurrence of the state resulting from entanglement swapping for various classes of input states, and a general relationship between the ranks of the initial states and the rank of the final state after swapping (Kirby et al., 2015). This extends ESP beyond the pure-state setting that dominates many elementary treatments.

A network-level variant with partially entangled inputs introduces the family

$$|\Phi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HH\rangle\pm |VV\rangle\bigr),\qquad |\Psi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HV\rangle\pm |VH\rangle\bigr),$$3

and shows that a single swap maps

$$|\Phi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HH\rangle\pm |VV\rangle\bigr),\qquad |\Psi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HV\rangle\pm |VH\rangle\bigr),$$4

with explicit probabilities. Repeated swapping through $$|\Phi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HH\rangle\pm |VV\rangle\bigr),\qquad |\Psi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HV\rangle\pm |VH\rangle\bigr),$$5 identical elementary links yields the binomial law

$$|\Phi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HH\rangle\pm |VV\rangle\bigr),\qquad |\Psi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HV\rangle\pm |VH\rangle\bigr),$$6

so ESP acts not only as a distribution mechanism but also as a probabilistic concentration mechanism for partially entangled resources (Guerra et al., 6 Aug 2025).

3. Higher-dimensional, continuous-variable, and many-body extensions

The qudit generalization replaces the Bell basis by the generalized Bell basis

$$|\Phi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HH\rangle\pm |VV\rangle\bigr),\qquad |\Psi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HV\rangle\pm |VH\rangle\bigr),$$7

and specializes the input states, without loss of generality, to Schmidt-diagonal pure states

$$|\Phi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HH\rangle\pm |VV\rangle\bigr),\qquad |\Psi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HV\rangle\pm |VH\rangle\bigr),$$8

The conditional output state on $$|\Phi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HH\rangle\pm |VV\rangle\bigr),\qquad |\Psi^\pm\rangle=\frac{1}{\sqrt{2}}\bigl(|HV\rangle\pm |VH\rangle\bigr),$$9 is

$|\xi\rangle_{AC},\qquad |\eta\rangle_{C'B},$0

with outcome probability

$|\xi\rangle_{AC},\qquad |\eta\rangle_{C'B},$1

Using the $|\xi\rangle_{AC},\qquad |\eta\rangle_{C'B},$2-norm entanglement monotone, the average distributed entanglement obeys the bounds

$|\xi\rangle_{AC},\qquad |\eta\rangle_{C'B},$3

and

$|\xi\rangle_{AC},\qquad |\eta\rangle_{C'B},$4

while for qubits the product bound is exact and for qutrits the average is exactly one half of the product (Starke et al., 1 Aug 2025).

In continuous-variable form, ESP is implemented by mixing the two Bell modes on a balanced beam splitter and homodyning the output quadratures $|\xi\rangle_{AC},\qquad |\eta\rangle_{C'B},$5 and $|\xi\rangle_{AC},\qquad |\eta\rangle_{C'B},$6. A modification with local certification upgrades each local resource from bipartite to tripartite, with remote mode $|\xi\rangle_{AC},\qquad |\eta\rangle_{C'B},$7, Bell mode $|\xi\rangle_{AC},\qquad |\eta\rangle_{C'B},$8, and certification mode $|\xi\rangle_{AC},\qquad |\eta\rangle_{C'B},$9. In the symmetric Gaussian case, the minimum partially-transposed symplectic eigenvalues after swapping satisfy

$C,C'$0

so the purity condition

$C,C'$1

guarantees

$C,C'$2

This permits remote entanglement to be certified using only local optical measurements at the intermediary station, an especially useful feature when the remote modes are mechanical resonators (Abdi et al., 2013). The same construction was earlier framed as “entanglement swapping with local certification,” with the same certifying condition and the same optomechanical application to remote micromechanical resonators (Abdi et al., 2012).

The most radical extension is the general many-body entanglement swapping protocol. Here two distant, non-signaling parties each prepare the same $C,C'$3-qubit target state

$C,C'$4

keep complementary subsets according to an arbitrary bipartition, and send the remaining qubits to an intermediary Eve. Eve applies the inverse preparation circuit in the reordered basis, $C,C'$5, and postselects the computational-basis outcome $C,C'$6. If

$C,C'$7

then the swapped state is

$C,C'$8

The protocol therefore preserves the Schmidt vectors exactly and cubes and renormalizes the Schmidt coefficients. Its fidelity with the target is

$C,C'$9

and the success probability is

$A,B$0

The cost is thus controlled by the 3rd Rényi entanglement entropy of the chosen partition, not directly by the number of qubits (Huhtanen et al., 27 Jun 2025).

4. Certification, diagnostics, and fault tolerance

Verification of ESP need not rely on full tomography. One line of work uses a collective entanglement witness, collectibility, to diagnose the protocol through four conditioned probabilities $A,B$1 and the local probability $A,B$2. For ideal swapping from two $A,B$3 pairs through singlet projection, the expected pattern is

$A,B$4

and the measured witness was

$A,B$5

close to the ideal theoretical value $A,B$6. The same probability set distinguishes depolarization, phase damping, amplitude damping, and imperfect Bell-state measurement, so the diagnostic is fault-sensitive rather than merely entanglement-sensitive (Trávníček et al., 2020).

Certification in the continuous-variable and optomechanical setting takes a different form. There the Bell measurement creates both remote entanglement and local certification-mode entanglement, and the criterion

$A,B$7

ensures that entanglement of the certification modes implies entanglement of the remote modes. This replaces direct nonlocal verification of hard-to-access subsystems by local optical verification at the Bell station (Abdi et al., 2012).

Fault tolerance is discussed explicitly in the many-body setting. Because the derivation is agnostic to the elementary building blocks, a logical qubit encoded in many physical qubits can be treated exactly as a physical qubit in the protocol. A GHZ state is interpreted as a Bell state of logical qubits encoded by the repetition code

$A,B$8

This makes error detection more natural at Eve’s side: a single readout error in one of Eve’s measured qubits can be recognized from the code structure, unlike in ordinary pair swapping where a single measurement error can directly corrupt the swapped entanglement. The fault-tolerant extension remains conceptual, but the proposed route is to combine the protocol with stabilizer codes and, for arbitrary target states, with magic-state distillation (Huhtanen et al., 27 Jun 2025).

An unusual variant replaces pairwise inter-particle entanglement by intra-photon path–polarization entanglement. In that linear-optical scheme, two outer photons are internally entangled between path and polarization, a third indistinguishable photon mediates Hong–Ou–Mandel bunching in chained Mach–Zehnder interferometers, and a heralding event yields the inter-photon state

$A,B$9

Bell-basis discrimination is not required in that construction, which is a direct counterexample to the idea that Bell-state analysis is structurally indispensable for every form of swapping (Bera et al., 2018).

5. ESP as a networking and control primitive

In quantum-network optimization, ESP is not only a state-transformation primitive but also a scheduling problem. One influential reformulation is the shift from path selection to entanglement-swapping trees. Given a path $$E_{con}(\text{output})=\frac{E_{con}(\xi_{AC})E_{con}(\eta_{C'B})}{4Pr(\text{Bell outcome})},$$0 from source $$E_{con}(\text{output})=\frac{E_{con}(\xi_{AC})E_{con}(\eta_{C'B})}{4Pr(\text{Bell outcome})},$$1 to destination $$E_{con}(\text{output})=\frac{E_{con}(\xi_{AC})E_{con}(\eta_{C'B})}{4Pr(\text{Bell outcome})},$$2, any complete binary tree over the ordered links in $$E_{con}(\text{output})=\frac{E_{con}(\xi_{AC})E_{con}(\eta_{C'B})}{4Pr(\text{Bell outcome})},$$3 gives a way to generate an entangled pair over $$E_{con}(\text{output})=\frac{E_{con}(\xi_{AC})E_{con}(\eta_{C'B})}{4Pr(\text{Bell outcome})},$$4. Leaves represent adjacent physical links, internal nodes are intermediate subpath entangled pairs produced by swapping, and the root is the target end-to-end entangled pair. Under the waiting protocol, one entangled pair can be stored while its sibling pair is being generated, so tree shape affects both expected latency and qubit age. Dynamic programming and balanced-tree heuristics developed in this setting outperform path-based approaches by roughly an order of magnitude in entangled-pair generation rate in many settings (Ghaderibaneh et al., 2021).

A later development replaces the precomputed, static swapping plan by a runtime-adaptive swapping-order policy. In that model, the decision at each trigger event is which currently possible swap to perform next, or whether to wait, based on available entangled pairs, their ages, active links, and elapsed generation times. The proposed greedy strategy uses a DP-with-ages cost-to-go estimate, and in larger networks it achieves up to 40% lower generation latency than the optimal static approach and up to 45% lower latency than Swap-ASAP; the multipath version achieves about 10–20% lower latency than the static LP baseline in most cases (Sundaram et al., 2024).

ESP has also been recast in protocol-stack terms for Ethernet-assisted quantum networks. In that setting, a Quantum Protocol carried in Ethernet frames establishes a virtual circuit, generates adjacent entangled links, and then performs swapping sequentially from both ends toward the center using tokens and levels. The total delay for $$E_{con}(\text{output})=\frac{E_{con}(\xi_{AC})E_{con}(\eta_{C'B})}{4Pr(\text{Bell outcome})},$$5 switches is expressed in terms of classical request and acknowledgment delays, quantum transmission delay, and the average number of swap trials $$E_{con}(\text{output})=\frac{E_{con}(\xi_{AC})E_{con}(\eta_{C'B})}{4Pr(\text{Bell outcome})},$$6, and the design goal is to keep total time delay within the decoherence threshold of qubits (Chen-Hu et al., 26 Feb 2025). A plausible implication is that, in networked deployments, ESP performance depends as much on control-plane design and retry policy as on the physical Bell-state measurement itself.

6. Experimental realizations, performance regimes, and open limitations

Proof-of-concept realizations span superconducting processors, linear optics, optomechanics, and deterministic solid-state emitters. On IBM superconducting hardware, entanglement swapping between two Bell pairs was implemented on ibmqx4 with output fidelities $$E_{con}(\text{output})=\frac{E_{con}(\xi_{AC})E_{con}(\eta_{C'B})}{4Pr(\text{Bell outcome})},$$7 and $$E_{con}(\text{output})=\frac{E_{con}(\xi_{AC})E_{con}(\eta_{C'B})}{4Pr(\text{Bell outcome})},$$8, and a separate purification routine improved Bell-pair fidelity from $$E_{con}(\text{output})=\frac{E_{con}(\xi_{AC})E_{con}(\eta_{C'B})}{4Pr(\text{Bell outcome})},$$9 to $$\langle E_{conc}\rangle =E_{con}(\xi_{AC})E_{con}(\eta_{C'B}) =\sqrt{1-P_C^2-C_C^2}\sqrt{1-P_{C'}^2-C_{C'}^2}.$$0 (Behera et al., 2017). A later many-body proof of concept used IBM’s 127-qubit processor ibm_brisbane to share GHZ states, executing the protocol for systems up to 12 qubits total and allowing GHZ states of up to 6 shared qubits; in that GHZ example, postselection is unnecessary because the nontrivial outcomes differ only by local bit flips (Huhtanen et al., 27 Jun 2025).

All-photonic swapping with remote deterministic quantum-dot sources marks a different regime. Two separate GaAs quantum dots were tuned into resonance so that a partial linear-optics Bell-state measurement on one photon from each pair heralded entanglement between the remaining photons. The measured input-pair fidelities were $$\langle E_{conc}\rangle =E_{con}(\xi_{AC})E_{con}(\eta_{C'B}) =\sqrt{1-P_C^2-C_C^2}\sqrt{1-P_{C'}^2-C_{C'}^2}.$$1 and $$\langle E_{conc}\rangle =E_{con}(\xi_{AC})E_{con}(\eta_{C'B}) =\sqrt{1-P_C^2-C_C^2}\sqrt{1-P_{C'}^2-C_{C'}^2}.$$2, the raw Hong–Ou–Mandel visibilities were $$\langle E_{conc}\rangle =E_{con}(\xi_{AC})E_{con}(\eta_{C'B}) =\sqrt{1-P_C^2-C_C^2}\sqrt{1-P_{C'}^2-C_{C'}^2}.$$3 and $$\langle E_{conc}\rangle =E_{con}(\xi_{AC})E_{con}(\eta_{C'B}) =\sqrt{1-P_C^2-C_C^2}\sqrt{1-P_{C'}^2-C_{C'}^2}.$$4, and the maximum reported swapped fidelities reached $$\langle E_{conc}\rangle =E_{con}(\xi_{AC})E_{con}(\eta_{C'B}) =\sqrt{1-P_C^2-C_C^2}\sqrt{1-P_{C'}^2-C_{C'}^2}.$$5 and $$\langle E_{conc}\rangle =E_{con}(\xi_{AC})E_{con}(\eta_{C'B}) =\sqrt{1-P_C^2-C_C^2}\sqrt{1-P_{C'}^2-C_{C'}^2}.$$6, more than 10 standard deviations above the classical limit (Beccaceci et al., 11 Dec 2025). The dominant limitations were imperfect photon indistinguishability, finite fine-structure splitting, energy-time entanglement of cascade photons, residual charge noise, detector timing jitter, and the inherent incompleteness of linear-optics Bell-state measurement.

Several recurrent limitations appear across the ESP literature. First, postselection overhead can dominate performance. In the many-body protocol, the success probability

$$\langle E_{conc}\rangle =E_{con}(\xi_{AC})E_{con}(\eta_{C'B}) =\sqrt{1-P_C^2-C_C^2}\sqrt{1-P_{C'}^2-C_{C'}^2}.$$7

can become exponentially small for highly entangled states, and for near-maximally entangled states of $$\langle E_{conc}\rangle =E_{con}(\xi_{AC})E_{con}(\eta_{C'B}) =\sqrt{1-P_C^2-C_C^2}\sqrt{1-P_{C'}^2-C_{C'}^2}.$$8 qubits evenly split the paper states $$\langle E_{conc}\rangle =E_{con}(\xi_{AC})E_{con}(\eta_{C'B}) =\sqrt{1-P_C^2-C_C^2}\sqrt{1-P_{C'}^2-C_{C'}^2}.$$9 (Huhtanen et al., 27 Jun 2025). Second, many derivations are restricted to pure-state inputs, with mixed-state extensions either absent or handled only in special two-qubit settings (Starke et al., 1 Aug 2025, Kirby et al., 2015). Third, exact reproduction of the target state is generally exceptional rather than generic: exactness holds for uniform Schmidt spectra in the many-body case, for maximally entangled inputs in the canonical case, or for locally correctable special cases such as GHZ (Huhtanen et al., 27 Jun 2025). Fourth, protocol optimality depends on the performance criterion. Deterministic minimum-error strategies maximize unconditional reliability, whereas maximum-confidence and threshold-based schemes maximize entanglement or fidelity on successful branches only (Solís-Prosser et al., 2014, Oppliger et al., 2021).

Taken together, these developments establish ESP as more than a single Bell-measurement trick. It is a family of measurement-induced entanglement-distribution procedures whose precise form depends on the resource class—Bell pairs, partially entangled pure states, mixed states, qudits, Gaussian tripartite states, or many-body targets—and whose practical performance is set by entanglement structure, distinguishability, postselection cost, and control architecture. The modern literature therefore treats ESP simultaneously as a nonlocality demonstration, a repeater primitive, a state-engineering tool, and a network-scheduling problem (Guerra et al., 6 Aug 2025, Ghaderibaneh et al., 2021).