Entanglement by Path Identity

• Entanglement by path identity is a technique that renders photon emissions from multiple SPDC sources indistinguishable, enabling coherent superposition and entangled state formation.
• The method utilizes precise phase stabilization, beam alignment, and mode matching to achieve high-dimensional states with fidelities around 85–90% in experiments.
• Its modular design facilitates scalability to multiphoton and integrated systems, offering promising applications in quantum communication and networked photonic architectures.

Searching arXiv for relevant papers on entanglement by path identity and closely related formulations. Entanglement by path identity is a method of entanglement generation in which multiple emission alternatives are made physically indistinguishable by aligning one or more output paths so completely that no measurement can reveal which source produced the detected quanta. In its canonical photonic form, several coherently pumped spontaneous parametric down-conversion (SPDC) sources emit into path-identical modes; once which-source information is absent even in principle, the pair-generation amplitudes interfere and the detected state becomes a coherent superposition of source-labeled mode components. The entanglement therefore does not originate inside any single source and does not require the photons from different sources to interact (Krenn et al., 2016, Kysela et al., 2019).

1. Historical emergence and conceptual scope

The modern formulation of entanglement by path identity was introduced as a method for creating multipartite and high-dimensional photonic entanglement starting from non-entangled photon pairs. Its two defining ingredients were stated explicitly as the superposition of photon pairs with different origins and the alignment of photons such that their paths are identical (Krenn et al., 2016). The same work emphasized that the technique was identified by analyzing the output of the automated experiment-design algorithm Melvin, and it proposed experimentally feasible realizations for polarization-entangled GHZ and W states as well as arbitrary high-dimensional two-photon states (Krenn et al., 2016).

A complementary formalization was given by Svozil, who described the method as a three-part construction: a universal unitary transformation prepares a coherent superposition of product-state creation branches, SPDC supplies pair creation with well-defined relational structure, and path identification maps outputs from different crystals into identical detection modes so that amplitudes from distinct creation events add coherently (Svozil, 2017). In that formulation, the relevant optical toolbox is the usual network of beam splitters and phase shifters implementing finite-dimensional unitaries.

The scheme was then generalized to many-particle interferometry with two identical sources, where making the paths of one or more particles identical eliminates which-way information and leaves the remaining detected particles in phase-selectable superpositions of Dicke sectors (Lahiri, 2018). In that framework, Bell states and three-particle GHZ-class states appear as special cases, and the visibility of the many-particle interference pattern is directly linked to the amount of entanglement (Lahiri, 2018).

The phrase is also used more broadly in later literature for related but not identical mechanisms, including single-photon optical-path entanglement, path-dependent CPTP evolutions, neutron spin-path entanglement with overlapping branches, and remote entanglement of independent particles. The canonical usage, however, remains the nonlinear-optical source-engineering paradigm in which indistinguishable emission origins are converted into entangled output states.

2. Quantum-mechanical mechanism and state construction

Path identity is implemented by deliberately overlapping the emission paths of multiple, coherently pumped SPDC sources so perfectly that, after the photons leave the apparatus, there is no information about which source generated them. Under that condition, the pair-generation processes interfere quantum mechanically, and the biphoton is a coherent superposition of origin-labeled two-photon mode components (Kysela et al., 2019).

In the simplest two-source construction, two identical SPDC crystals are adjusted to emit into the same fundamental orbital-angular-momentum (OAM) mode, and one branch is tagged by a mode shifter. If the outputs are path-identical, the resulting state is

$$|\psi\rangle = \frac{1}{\sqrt{2}}\left(|0,0\rangle + e^{i\varphi}|1,1\rangle\right),$$

where the ket labels denote the OAM quanta of signal and idler and $\varphi$ is a controllable phase set by a phase shifter (Kysela et al., 2019). The crucial point is that neither crystal alone emits an entangled state; the entanglement is created by coherent superposition after which-source information has been erased.

The construction scales directly. For $d$ crystals with $d-1$ phase- and mode-shifters, the target state takes the form

$$|\psi\rangle = \sum_{\ell=0}^{d-1} c_\ell |\ell,\ell\rangle,$$

with complex amplitudes $c_\ell$ determined by pump powers and relative phases set by inserted phase shifters (Kysela et al., 2019). In the equal-amplitude case,

$$|\psi\rangle = \frac{1}{\sqrt{d}}\left(|0,0\rangle + e^{i\bar{\varphi}_1}|1,1\rangle + \cdots + e^{i\bar{\varphi}_{d-1}}|d-1,d-1\rangle\right),$$

with

$\bar{\varphi}_i = \sum_{j=d-i}^{d-1}\varphi_j.$

This architecture is modular: each added source contributes one new basis component and one additional dimension of entanglement (Kysela et al., 2019).

The coherence requirements are stringent. All sources must be indistinguishable spectrally and in polarization; the down-conversion paths must be overlapped spatially; the pump must remain coherent across all source arms; and the single-pair generation probability per source must be low enough that multi-pair and simultaneous multi-source emissions are negligible (Kysela et al., 2019). If any residual distinguishing information remains—spectral, temporal, or spatial—the coherent superposition is replaced by a statistical mixture (Kysela et al., 2019).

3. High-dimensional OAM implementation and verified performance

The first detailed experimental demonstration of high-dimensional entanglement by path identity used three periodically poled KTP crystals, labeled A, B, and C, each operating as an independent type-II, frequency-degenerate SPDC source and pumped coherently at $405\,\mathrm{nm}$ in a Mach–Zehnder interferometric configuration (Kysela et al., 2019). Pump splitting ratios controlled amplitudes, while trombone phase shifters controlled relative phases. The assigned OAM content was engineered through the pump: crystal A used pump OAM $0$, crystal B used a spiral phase plate producing pump OAM $\varphi$0, and crystal C used the same $\varphi$1 element followed by a mirror that flipped the OAM sign, producing $\varphi$2 in the down-converted pair (Kysela et al., 2019).

The experimentally implemented three-dimensional state was

$\varphi$3

with $\varphi$4 set by pump splitting and $\varphi$5 set by trombone systems (Kysela et al., 2019). Path identity was enforced with interferometers and $\varphi$6 imaging; dichroic mirrors and a band-pass filter separated pump and down-conversion beams; higher-order OAM contributions were suppressed; and deterministic signal-idler separation was achieved with a PBS. Projective OAM measurements used SLM holograms, single-mode fibers, and single-photon counters, followed by maximum-likelihood state tomography with fidelity

$\varphi$7

The experiment realized both two-dimensional and three-dimensional target states, including three mutually orthogonal maximally entangled qutrit states formed with $\varphi$8 (Kysela et al., 2019).

Target state	Form	Fidelity
$\varphi$9	$d$0	$d$1
$d$2	$d$3	$d$4
$d$5	$d$6	$d$7
$d$8	$d$9	$d-1$0
$d-1$1	$d-1$2	$d-1$3
$d-1$4	$d-1$5	$d-1$6
$d-1$7	$d-1$8	$d-1$9

The average fidelity of the maximally entangled three-dimensional set $$|\psi\rangle = \sum_{\ell=0}^{d-1} c_\ell |\ell,\ell\rangle,$$0 was $$|\psi\rangle = \sum_{\ell=0}^{d-1} c_\ell |\ell,\ell\rangle,$$1, while the two-dimensional average was $$|\psi\rangle = \sum_{\ell=0}^{d-1} c_\ell |\ell,\ell\rangle,$$2, leading to the conclusion that the quality did not significantly degrade in scaling from $$|\psi\rangle = \sum_{\ell=0}^{d-1} c_\ell |\ell,\ell\rangle,$$3 to $$|\psi\rangle = \sum_{\ell=0}^{d-1} c_\ell |\ell,\ell\rangle,$$4 (Kysela et al., 2019). The paper did not report a high-dimensional Bell test or a Schmidt-number witness in the main text; entanglement was inferred from complete tomography and high fidelities to target states (Kysela et al., 2019). The stated dominant imperfections were slight source distinguishability, imperfect OAM transformations caused by slight non-collinearity of about $$|\psi\rangle = \sum_{\ell=0}^{d-1} c_\ell |\ell,\ell\rangle,$$5, phase drift in the interferometers, and residual multi-pair events suppressed by low pump power (Kysela et al., 2019).

4. Relation to other entanglement mechanisms and recurring misconceptions

Entanglement by path identity is distinct from the standard single-SPDC route to OAM entanglement. In a conventional single-crystal SPDC source, OAM conservation usually gives anti-correlated states of the form $$|\psi\rangle = \sum_{\ell=0}^{d-1} c_\ell |\ell,\ell\rangle,$$6 with a non-uniform spiral spectrum, so maximally entangled high-dimensional states often require pump engineering or Procrustean filtering that discards photons (Kysela et al., 2019). By contrast, path identity can produce arbitrary superpositions of desired pair modes, here of the form $$|\psi\rangle = \sum_{\ell=0}^{d-1} c_\ell |\ell,\ell\rangle,$$7, with tunable amplitudes and phases and with photons created directly in the target modes rather than post-selected into them (Kysela et al., 2019).

It is also distinct from entanglement swapping. Swapping requires interference and measurement of intermediate photons in order to project remote photons into an entangled state. Path-identity-induced entanglement requires no Bell-state measurement and no direct interaction between the photons that end up entangled; the entanglement arises entirely from the indistinguishability of the generation pathways (Kysela et al., 2019, Wang et al., 2024).

A common terminological confusion concerns single-photon optical-path entanglement. In that regime, one photon is coherently delocalized across several paths, yielding states such as

$$|\psi\rangle = \sum_{\ell=0}^{d-1} c_\ell |\ell,\ell\rangle,$$8

and the problem is to certify entanglement in a subspace with at most one photon per path (Monteiro et al., 2015). Path identity can be used to prepare such states by erasing which-path information, but the physical resource is different from the biphoton, source-superposition mechanism used for high-dimensional pair-state engineering. The distinction is operationally significant because the certification tools are different: the optical-path-entanglement witness of local small displacements and on/off detection was designed precisely for the single-photon regime and does not rely on post-selection (Monteiro et al., 2015).

A related distinction appears in induced-coherence-based proposals for remote sensing. There, quantum-induced coherence by path identity can generate first-order coherence in one arm when the partner arm is made path-identical. The literature stresses that this single-photon interference effect does not by itself imply genuine entanglement; in the remote-sensing proposal using quantum frequency combs, the induced coherence transfers target information nonlocally, while entanglement originates from the two-mode squeezed-vacuum structure and the coherent phase-locked combination of modes (Dalvit et al., 2024).

5. Degrees of freedom, integrated architectures, and network implementations

The path-identity mechanism is not restricted to OAM. The experimental high-dimensional study explicitly described it as agnostic to the mode family and stated that the same architecture can be applied to polarization, path, time-bin, and frequency-bin encodings; in that context, the familiar cross-crystal polarization source was identified as the simplest example (Kysela et al., 2019). The same work also noted that each additional source increases the entanglement dimension linearly, with amplitudes and phases independently adjustable, and argued that photons are not discarded because they are generated directly in the desired modes (Kysela et al., 2019).

This modularity has motivated integrated implementations. An on-chip proposal transferred the path-identity idea to silicon photonics, using spontaneous four-wave mixing in micro-ring resonators or nanowires, reversed Hong–Ou–Mandel separation, multimode interferometers, Mach–Zehnder interferometers, and two-dimensional grating couplers that map path to polarization (Feng et al., 2020). That work proposed on-chip GHZ states, W states, and a generalized graph formalism for odd-photon W-state generation, with the central claim that monolithic integration mitigates the stability and scalability limitations of bulk-optical multi-source arrangements (Feng et al., 2020).

The network setting has also been realized experimentally. A telecom experiment distributed heralded single-photon path entanglement over a total of $$|\psi\rangle = \sum_{\ell=0}^{d-1} c_\ell |\ell,\ell\rangle,$$9 of optical fiber in a repeater-like architecture, with independent SPDC sources at Alice and Bob, idlers combined on a 50/50 beam splitter at a central station, and path identity enforced by spectral filtering, temporal alignment, polarization control, and active phase stabilization (Caspar et al., 2020). The heralding rate was $c_\ell$0 at $c_\ell$1, and the displacement-based entanglement witness yielded

$c_\ell$2

corresponding to a $c_\ell$3 violation without relying on post-selection in the witness evaluation (Caspar et al., 2020). That result addressed a different state class from the OAM biphoton source, but it showed that path identity is compatible with telecom wavelengths, kilometer-scale distribution, and loss-tolerant entanglement certification.

A compact summary of representative implementations is useful.

System	Physical platform	Reported result
High-dimensional OAM source	Three coherently pumped ppKTP SPDC crystals	Two- and three-dimensional entangled states with fidelities around $c_\ell$4–$c_\ell$5 (Kysela et al., 2019)
Heralded telecom path entanglement	Two independent SPDC sources over $c_\ell$6 fiber	Heralding rate $c_\ell$7, witness violation $c_\ell$8 (Caspar et al., 2020)
On-chip multiphoton proposal	Silicon photonics with SFWM sources	GHZ and W architectures by path identity (Feng et al., 2020)

These results support the general claim that path identity is simultaneously a state-engineering principle and a network primitive. The literature repeatedly associates it with high brightness, modular scaling, and compatibility with integrated photonics, though the same papers also emphasize the continuing need for precise phase stabilization and stringent mode matching (Kysela et al., 2019, Feng et al., 2020).

6. Theoretical limits, broader formulations, and current directions

Later theory has clarified both the power and the limits of the method. A 2025 study of OAM-state engineering showed that path identity with multiple crystals is equivalent, at the level of OAM spectra and fidelities, to a single-crystal SPDC source pumped by a spatially engineered superposition of Laguerre–Gaussian modes (Bernecker et al., 19 Aug 2025). The same work identified fixed amplitude ratios among $c_\ell$9 modes on an OAM-conservation line as a fundamental obstacle to perfect full-space maximally entangled states when $$|\psi\rangle = \frac{1}{\sqrt{d}}\left(|0,0\rangle + e^{i\bar{\varphi}_1}|1,1\rangle + \cdots + e^{i\bar{\varphi}_{d-1}}|d-1,d-1\rangle\right),$$0, and concluded that the high-fidelity building blocks are restricted to pump OAM $$|\psi\rangle = \frac{1}{\sqrt{d}}\left(|0,0\rangle + e^{i\bar{\varphi}_1}|1,1\rangle + \cdots + e^{i\bar{\varphi}_{d-1}}|d-1,d-1\rangle\right),$$1 (Bernecker et al., 19 Aug 2025). Within that model, the optimal total dimensionality obtainable from $$|\psi\rangle = \frac{1}{\sqrt{d}}\left(|0,0\rangle + e^{i\bar{\varphi}_1}|1,1\rangle + \cdots + e^{i\bar{\varphi}_{d-1}}|d-1,d-1\rangle\right),$$2 crystals is $$|\psi\rangle = \frac{1}{\sqrt{d}}\left(|0,0\rangle + e^{i\bar{\varphi}_1}|1,1\rangle + \cdots + e^{i\bar{\varphi}_{d-1}}|d-1,d-1\rangle\right),$$3 when each crystal contributes a high-fidelity Bell pair, and a three-crystal Gaussian-pumped scheme plus OAM shifts was reported to reach full-space fidelity $$|\psi\rangle = \frac{1}{\sqrt{d}}\left(|0,0\rangle + e^{i\bar{\varphi}_1}|1,1\rangle + \cdots + e^{i\bar{\varphi}_{d-1}}|d-1,d-1\rangle\right),$$4 for the target $$|\psi\rangle = \frac{1}{\sqrt{d}}\left(|0,0\rangle + e^{i\bar{\varphi}_1}|1,1\rangle + \cdots + e^{i\bar{\varphi}_{d-1}}|d-1,d-1\rangle\right),$$5 (Bernecker et al., 19 Aug 2025).

The phrase has also been extended into a more abstract quantum-information setting. For population-preserving CPTP maps on path-superposed particles, a path-entangling evolution can be written as a map that multiplies off-diagonal density-matrix elements by path-dependent phase factors and possible dephasing terms. In that framework, inseparability is equivalent to entangling power, and for a diagonal phase operation on $$|\psi\rangle = \frac{1}{\sqrt{d}}\left(|0,0\rangle + e^{i\bar{\varphi}_1}|1,1\rangle + \cdots + e^{i\bar{\varphi}_{d-1}}|d-1,d-1\rangle\right),$$6 the relevant non-additivity witness is

$$|\psi\rangle = \frac{1}{\sqrt{d}}\left(|0,0\rangle + e^{i\bar{\varphi}_1}|1,1\rangle + \cdots + e^{i\bar{\varphi}_{d-1}}|d-1,d-1\rangle\right),$$7

with output negativity

$$|\psi\rangle = \frac{1}{\sqrt{d}}\left(|0,0\rangle + e^{i\bar{\varphi}_1}|1,1\rangle + \cdots + e^{i\bar{\varphi}_{d-1}}|d-1,d-1\rangle\right),$$8

in the unitary case (Matsumura, 2021). This usage is not identical to the original SPDC source-engineering scheme, but it preserves the central idea that path labels alone can mediate nonlocal coherence and entanglement without changing path populations.

Experimentally, the path-identity principle has continued to move beyond the original “common-source” picture. A 2024 four-source SPDC experiment showed that photons that did not interact and did not share a common past could still be entangled by making the origins of their partner photons indistinguishable (Wang et al., 2024). In the four-photon heralded configuration, the Bell parameter was

$$|\psi\rangle = \frac{1}{\sqrt{d}}\left(|0,0\rangle + e^{i\bar{\varphi}_1}|1,1\rangle + \cdots + e^{i\bar{\varphi}_{d-1}}|d-1,d-1\rangle\right),$$9

the fidelity to $\bar{\varphi}_i = \sum_{j=d-i}^{d-1}\varphi_j.$0 was $\bar{\varphi}_i = \sum_{j=d-i}^{d-1}\varphi_j.$1, and the concurrence was $\bar{\varphi}_i = \sum_{j=d-i}^{d-1}\varphi_j.$2; in a one-ancilla configuration the fidelity was $\bar{\varphi}_i = \sum_{j=d-i}^{d-1}\varphi_j.$3 and the concurrence $\bar{\varphi}_i = \sum_{j=d-i}^{d-1}\varphi_j.$4 (Wang et al., 2024). The paper explicitly framed this as entanglement generation without direct interaction, prior entanglement, or Bell-state measurement.

Matter-wave analogues broaden the concept further. In neutron interferometry, spin-path entanglement prepared with magnetic Wollaston prisms and RF flippers remained robust even when the path branches overlapped, with CHSH witnesses saturating the Tsirelson bound scaled by polarization across large variations in entanglement length, coherence length, and RF-induced energy differences (Kuhn et al., 2020). That result does not implement multi-source photonic path identity, but it reinforces the broader thesis that the relevant resource is not large path separation but the absence of usable which-path information.

A plausible implication of the current literature is that entanglement by path identity has matured from a specific nonlinear-optical trick into a general design principle for indistinguishability-based entanglement engineering. The common structure across these variants is the same: coherent alternatives are created, source or path labels are rendered operationally inaccessible, and the surviving amplitudes define the entangled state. The remaining challenges are likewise common: phase stability, mode overlap, control of higher-order emissions, and—in the high-dimensional OAM setting—fundamental spectral and mode-structure constraints that limit full-space maximally entangled targets (Kysela et al., 2019, Bernecker et al., 19 Aug 2025).