Source-Target Entanglement Overview

• Source-target entanglement is a framework where entanglement is prepared at a source and then directed towards a target via various quantum protocols.
• The topic spans methods like quantum illumination, heralded state preparation, and asymmetric photonic architectures to achieve effective entanglement transfer and certification.
• It also includes operational approaches such as network capacity analysis and bipartition measures, providing actionable insights into quantum communication and sensing.

Source-target entanglement denotes several distinct constructions rather than a single standardized protocol. In the literature considered here, it can mean entanglement prepared at a source and distributed to remote targets, a source–target subsystem structure used for heralded state preparation, transfer of entanglement from one set of particles to another, or a bipartition in which every node or degree of freedom is assigned “source” and “target” roles. It can also refer to target-space, rather than base-space, entanglement in first-quantized and matrix-model settings. A central negative result is equally important: in quantum illumination and related target-detection models, the operative resource is not literal entanglement between source and target, but entanglement prepared at the source between signal and idler modes, followed by a joint measurement on the noisy return and the stored idler (Zhuang et al., 2017, Sorelli et al., 2020).

1. Entanglement with a target versus entanglement about a target

In quantum illumination, an entanglement source produces $M\gg 1$ iid signal-idler mode pairs in a two-mode squeezed-vacuum state with $N_S\ll 1$. The signal probes a region that may contain a weakly reflecting target, while the idler is retained ideally. The target is modeled only through the hypotheses $h=0$ and $h=1$: under $h=0$, the return is a thermal mode of mean photon number $N_B$; under $h=1$, the return is $$\hat c_{S_m}=\sqrt{\kappa}\,\hat c_{S_{0_m}}+\sqrt{1-\kappa}\,\hat c_{N_m}$$ with $\kappa\ll 1$. The decisive structural difference is the residual phase-sensitive return-idler cross correlation $\sqrt{\kappa N_S(N_S+1)}$, present only under target presence. The paper is explicit that QI target detection is the task of identifying the presence of that phase-sensitive cross correlation, and equally explicit that the environment is entanglement breaking: the sensing advantage survives even though the original signal-idler entanglement is destroyed before the final measurement (Zhuang et al., 2017).

The 2020 review on quantum radar makes the same distinction in broader terms. The transmitter prepares entanglement between a transmitted signal mode and a retained idler mode, while the target is a weak passive reflector embedded in thermal noise, represented by the parameter $N_S\ll 1$0. Under the standard QI operating conditions, the return-idler states under both hypotheses are separable, so the advantage does not rely on surviving entanglement with the target or even on surviving return-idler entanglement. It relies on source-prepared phase-sensitive cross-correlations larger than any classical source can provide at the same signal brightness, together with a joint measurement on return and idler (Sorelli et al., 2020).

A closely related caution appears in entanglement-assisted tomography of a quantum target. There the target is a qubit $N_S\ll 1$1 with $N_S\ll 1$2, probed indirectly by scattering a qubit $N_S\ll 1$3 from it through the Heisenberg interaction $N_S\ll 1$4. In the entanglement-assisted strategy the relevant resource is an initially maximally entangled probe-ancilla state $N_S\ll 1$5; the target is inferred from how the target-dependent scattering channel acts on the probe half of that pair. The paper does not treat pre-existing source-target entanglement as the operative resource (Pasquale et al., 2011).

These results support a precise restriction on the term. In remote sensing and indirect tomography, “source-target entanglement” is generally not the mechanism. What matters is entanglement at the source, target-dependent channels, and measurements that extract the resulting correlation structure.

2. Heralded and asymmetric photonic source–target architectures

A different usage appears in heralded photonic state preparation. In the experimental demonstration of a heralded entanglement source, a single type-II SPDC source is operated on its three-pair emission component,

$N_S\ll 1$6

and the six photons are split into four auxiliary trigger photons and a two-photon target subsystem in output modes $N_S\ll 1$7. After the optical network, the relevant term has the form $N_S\ll 1$8, with

$N_S\ll 1$9

A fourfold trigger projects the output onto the Bell pair. The source/herald subsystem and the target subsystem are therefore initially embedded in a larger six-photon state, but after heralding the trigger photons are measured destructively and function as classical event flags rather than persistent entangled partners. The reported figures are a heralding probability $h=0$0 for a given three-pair input, fidelity better than $h=0$1, and state preparation efficiency up to $h=0$2 (Wagenknecht et al., 2010).

A second photonic architecture uses asymmetry deliberately. The 810/1550 nm source based on SPDC in MgO:PPLN inside a polarization Sagnac interferometer generates

$h=0$3

with the signal photon at $h=0$4 nm and the idler photon at $h=0$5 nm. The source is simultaneously polarization-entangled and time-energy-entangled, directly fiber coupled, and explicitly designed for heterogeneous links: $h=0$6 nm for low-loss fiber transport and $h=0$7 nm for free-space propagation and efficient Si-APD detection. The reported performance includes heralding efficiencies of $h=0$8 at $h=0$9 nm and $h=1$0 at $h=1$1 nm, polarization fidelity $h=1$2, net polarization visibilities $h=1$3 and $h=1$4, time-energy visibility $h=1$5, spectral brightness $h=1$6, and coincidence rates up to $h=1$7 at $h=1$8. In the authors’ hybrid testbed, a $h=1$9 km free-space arm and a $h=0$0 km fiber arm correspond directly to asymmetric source–target deployment (Dumas et al., 21 May 2026).

Taken together, these works treat the source–target relation as a preparation architecture. In one case the target subsystem is certified by heralding; in the other, different wavelengths assign source-side and remote-target roles to the two photons.

3. Matter–light interfaces and explicit entanglement transfer

A more literal source-to-target entanglement appears in matter–light interfaces. The spinwave-photon source based on a cold $h=0$1 ensemble generates

$h=0$2

where the flying write-out photon is entangled with a stored collective atomic excitation. Here the source subsystem is the emitted photonic polarization, the target subsystem is the spinwave qubit encoded in $h=0$3 and $h=0$4, and the state is heralded by detection of the write-out photon. A polarization-compensated low-finesse ring cavity is engineered to be resonant for both spinwave modes, enabling an intrinsic retrieval efficiency up to $h=0$5, with Bell-CHSH values $h=0$6 at $h=0$7, $h=0$8 at $h=0$9, and $N_B$0 at $N_B$1 (Yang et al., 2015).

In detuned resonance fluorescence, the source-target relation is cast as cascaded quantum driving. A driven two-level system with

$N_B$2

emits spectrally filtered sideband photons whose nonvacuum sector is a Bell state $N_B$3, with the full emitted state described as a superposition of vacuum and that Bell pair. The source then drives a target consisting of coupled bosonic modes modeling exciton-polaritons through a cascaded master equation. When the source sidebands are tuned to the target resonances, the lower and upper polariton branches acquire a postselected state with concurrence about $N_B$4 and fidelity $N_B$5 to $N_B$6. The paper therefore treats the source as a spectrally tunable entangling reservoir for a dissipative target (Carreño et al., 2023).

The paper “Entanglement is better teleported than transmitted” sharpens the contrast between direct transfer and resource-assisted transfer. Alice’s source qubit $N_B$7 is initially entangled with an ancilla $N_B$8, and the goal is to make Bob’s target qubit $N_B$9 entangled with $h=1$0 through a quantum field. For direct transmission, the reduced state $h=1$1 obeys

$h=1$2

so Bob–ancilla negativity does not appear up to second perturbative order. If, instead, Alice and Bob first use the field to generate or harvest entanglement between an auxiliary detector $h=1$3 and $h=1$4, and then consume that resource in teleportation, the transferred negativity appears already at second order. In the maximally entangled source-reference case $h=1$5,

$h=1$6

The paper’s conclusion is therefore that the field is more effective as a source of entanglement than as a direct transmission medium for source-reference entanglement (Yamaguchi et al., 2023).

An older spin-system model studies transfer between two spin pairs. Target particles $h=1$7 interact locally with source particles $h=1$8 via simultaneous Heisenberg couplings $h=1$9 and $$\hat c_{S_m}=\sqrt{\kappa}\,\hat c_{S_{0_m}}+\sqrt{1-\kappa}\,\hat c_{N_m}$$0. For qubit sources and qubit targets,

$$\hat c_{S_m}=\sqrt{\kappa}\,\hat c_{S_{0_m}}+\sqrt{1-\kappa}\,\hat c_{N_m}$$1

so if the source pair is maximally entangled, the target pair becomes maximally entangled at $$\hat c_{S_m}=\sqrt{\kappa}\,\hat c_{S_{0_m}}+\sqrt{1-\kappa}\,\hat c_{N_m}$$2, independently of its initial state. For qutrit sources and qubit targets, the analogous one-shot transfer is state-dependent and generally submaximal, but an iterated operation can drive the target pair close to maximal entanglement (Meng et al., 2010).

4. Distribution from a central source and communication-theoretic capacities

The most explicit network formulation of source-to-target entanglement distribution considers $$\hat c_{S_m}=\sqrt{\kappa}\,\hat c_{S_{0_m}}+\sqrt{1-\kappa}\,\hat c_{N_m}$$3 noisy quantum channels $$\hat c_{S_m}=\sqrt{\kappa}\,\hat c_{S_{0_m}}+\sqrt{1-\kappa}\,\hat c_{N_m}$$4. A central source prepares an $$\hat c_{S_m}=\sqrt{\kappa}\,\hat c_{S_{0_m}}+\sqrt{1-\kappa}\,\hat c_{N_m}$$5-partite input state, sends subsystem $$\hat c_{S_m}=\sqrt{\kappa}\,\hat c_{S_{0_m}}+\sqrt{1-\kappa}\,\hat c_{N_m}$$6 through $$\hat c_{S_m}=\sqrt{\kappa}\,\hat c_{S_{0_m}}+\sqrt{1-\kappa}\,\hat c_{N_m}$$7 uses of channel $$\hat c_{S_m}=\sqrt{\kappa}\,\hat c_{S_{0_m}}+\sqrt{1-\kappa}\,\hat c_{N_m}$$8, and the remote receivers $$\hat c_{S_m}=\sqrt{\kappa}\,\hat c_{S_{0_m}}+\sqrt{1-\kappa}\,\hat c_{N_m}$$9 then use LOCC to distill a target entangled state. In this model the source retains no noiseless subsystem after transmission. The paper defines the EPR distribution capacity $\kappa\ll 1$0 and the GHZ distribution capacity $\kappa\ll 1$1, proves lower and upper bounds, and obtains exact characterizations in special cases. For two erasure channels,

$\kappa\ll 1$2

For multipartite GHZ distribution, if the most noisy channel is dephasing and $\kappa\ll 1$3 for all other links, then

$\kappa\ll 1$4

The framework distinguishes pairwise target entanglement from genuinely multipartite target entanglement and treats both as capacities of the source-to-target star architecture (Chen et al., 2024).

A related but more abstract operational usage appears in entanglement-assisted zero-error source-channel coding. Alice and Bob receive correlated classical inputs from a source, communicate through a noisy classical one-way channel, and are assisted only by shared entanglement. The graph-theoretic parameters are the entangled chromatic number $\kappa\ll 1$5, the entangled independence number $\kappa\ll 1$6, the entangled Witsenhausen rate $\kappa\ll 1$7, the entangled Shannon capacity $\kappa\ll 1$8, and the entangled source-channel cost rate $\kappa\ll 1$9. A central bound is

$\sqrt{\kappa N_S(N_S+1)}$0

On quarter-orthogonality graphs $\sqrt{\kappa N_S(N_S+1)}$1, the paper proves

$\sqrt{\kappa N_S(N_S+1)}$2

and, for the source-channel setting,

$\sqrt{\kappa N_S(N_S+1)}$3

Here the “source-target” language is operational rather than subsystem-based: shared entanglement changes how source symbols can be reconstructed at the receiving end with zero error (Briët et al., 2013).

These two lines of work suggest a broad operational interpretation of source-target entanglement: a source distributes, assists, or steers entanglement subject to channel constraints, and the relevant figures of merit are capacities or exact zero-error rates.

5. Intrinsic source–target bipartitions and interferometric generation

One paper uses the term literally as a graph-native entanglement measure for discrete-time quantum walks on irregular networks. For a graph $\sqrt{\kappa N_S(N_S+1)}$4, each node $\sqrt{\kappa N_S(N_S+1)}$5 is duplicated into a source copy $\sqrt{\kappa N_S(N_S+1)}$6 and a target copy $\sqrt{\kappa N_S(N_S+1)}$7, turning an arc state $\sqrt{\kappa N_S(N_S+1)}$8 into $\sqrt{\kappa N_S(N_S+1)}$9. A general walk state

$N_S\ll 1$00

is embedded as

$N_S\ll 1$01

and the source-target entanglement is the von Neumann entropy

$N_S\ll 1$02

of the Schmidt spectrum of $N_S\ll 1$03. The paper proves that if $N_S\ll 1$04 is the size of the largest matching of the bipartite double cover contained in the support of $N_S\ll 1$05, then

$N_S\ll 1$06

It further identifies maximum matchings, not degree alone, as the structural quantity governing the entanglement capacity of the network (Prerana et al., 1 May 2026).

In mesoscopic transport, chiral Majorana interferometry realizes a source-to-target entanglement distribution architecture in a different sense. Contacts $N_S\ll 1$07 and $N_S\ll 1$08 are biased source terminals, contacts $N_S\ll 1$09 and $N_S\ll 1$10 are grounded target/output terminals, and a four-terminal Majorana interferometer converts incoming electrons into entangled outgoing electron-hole superpositions. The entanglement witness extracted from current cross-correlations is

$N_S\ll 1$11

For

$N_S\ll 1$12

the paper obtains

$N_S\ll 1$13

The source and target are not entangled as macroscopic systems; rather, source-injected particles become entangled with one another and are delivered to target/output leads (Chirolli et al., 2018).

A photonic interferometric variant appears in entanglement by path identity. Four independent SPDC sources generate product-polarization pairs, but the optical arrangement makes the origin of a successful fourfold event fundamentally indistinguishable. After postselecting one photon in each path $N_S\ll 1$14, the surviving amplitudes yield

$N_S\ll 1$15

so the target photons in paths $N_S\ll 1$16 and $N_S\ll 1$17 are entangled even though there is no prior entangled pair, no Bell-state measurement, and no direct interaction between them. The reported evidence includes $N_S\ll 1$18 in a Bell-CHSH test, fidelity $N_S\ll 1$19 with $N_S\ll 1$20, and concurrence $N_S\ll 1$21 (Wang et al., 2024).

These constructions broaden the meaning of the term. Source-target entanglement can be a canonical bipartition on a network, an output-entanglement architecture in transport, or entanglement generated by indistinguishable source alternatives.

6. Target-space entanglement and adjacent terminological usages

In first-quantized quantum mechanics, string-inspired settings, and matrix models, the relevant distinction is not between source and target subsystems but between base space and target space. Target-space entanglement entropy is defined by choosing a subalgebra $N_S\ll 1$22 of observables associated with a target-space subregion $N_S\ll 1$23, restricting the state to $N_S\ll 1$24, and computing

$N_S\ll 1$25

For fermionic Slater determinants this becomes an exact overlap-matrix formula. If

$N_S\ll 1$26

then

$N_S\ll 1$27

where $N_S\ll 1$28 are the eigenvalues of $N_S\ll 1$29. In the free one-matrix or free-fermion ground state on a circle, the entropy of a single interval of length $N_S\ll 1$30 scales as

$N_S\ll 1$31

and the mutual information of two intervals remains finite at large $N_S\ll 1$32 (Mazenc et al., 2019, Sugishita, 2021).

A nearby but distinct usage appears in LOCC convertibility theory. “Source entanglement” and “accessible entanglement” are defined from the sets $N_S\ll 1$33 and $N_S\ll 1$34 of states that can reach $N_S\ll 1$35 or be reached from $N_S\ll 1$36 by deterministic LOCC. Their volume-based definitions are

$N_S\ll 1$37

For bipartite pure states, majorization gives exact source and accessible polytopes. In the two-qubit case with Schmidt vector $N_S\ll 1$38,

$N_S\ll 1$39

In this context, “source” and “accessible” refer to LOCC order structure rather than physical source and target subsystems (Sauerwein et al., 2015).

These neighboring usages clarify an important point. Across several subfields, “source” and “target” may refer to channel endpoints, heralding subsystems, configuration space, or LOCC preorder. The phrase is therefore stable only within a given formalism.

Source-target entanglement is best understood as a family of constructions organized around a directed relation: generation at a source, certification by a source-side subsystem, transfer to a target, or entropic partition into source-like and target-like roles. The literature surveyed here separates three especially important cases. First, some papers explicitly deny that entanglement with the target is the operative resource and replace it by source-prepared correlations exploited at detection. Second, several architectures do realize a genuine source-to-target flow of entanglement, either by heralding, cascaded driving, field-assisted teleportation, or central-source distribution. Third, other works use “source” and “target” to define a bipartition or an operational order rather than a physical entangled pair. This suggests that the term is most precise when accompanied by the underlying mechanism: source-generated return-idler correlation, heralded preparation, asymmetric distribution, entanglement transfer, central-source capacity, graph bipartition, or target-space algebra.