Finite-Location Causal Witnesses in Quantum Networks

• The paper introduces finite-location causal witnesses as Hermitian operators that detect causal nonseparability using measurement statistics from a limited set of local operations.
• It details experimental protocols, including the use of quantum switches and interferometry, achieving robust negative witness values (e.g., -0.305 ± 0.001) confirming indefinite causal order.
• The study emphasizes the method's generalizability and robustness, extending traditional Bell tests and do-interventions to certify nonclassical causal structures in quantum networks.

A finite-location causal witness is a Hermitian operator designed to detect causal nonseparability in quantum processes where each party is assigned only a finite number of local operations or spacetime events. Within the process-matrix formalism, these witnesses generalize entanglement witnesses to scenarios involving indefinite or nonclassical causal order, enabling experimental certification in both circuit-based and network quantum architectures. Finite-location causal witnesses are constructed so as to separate the set of causally separable processes—those expressible as convex mixtures of definite global orders—from those that are genuinely causally nonseparable, using only measurement statistics derived from a finite collection of local operations (Araújo et al., 2015, Valibouse et al., 13 Apr 2026, Zamora et al., 9 Jan 2026).

1. Causal Nonseparability and the Process Matrix Framework

A quantum process connecting $N$ separated laboratories (finite locations) is mathematically defined by a process matrix

$W \in \bigotimes_{i=1}^N (A^i_I \otimes A^i_O)$

subject to positivity, normalization ($\operatorname{Tr} W = \prod_i d_{A^i_O}$), and linear constraints excluding forbidden signaling. Each party $A^i$ implements completely positive (CP), trace-nonincreasing maps with Choi matrices. A process is causally separable if it lies in the convex cone generated by all possible total orders of the parties, i.e.,

$$\mathcal{W}^{\rm sep} = \operatorname{conv}\left( \bigcup_\sigma \mathcal{W}^\sigma \right),$$

where each $\mathcal{W}^\sigma$ corresponds to a definite causal order. Any process not of this form is causally nonseparable (Araújo et al., 2015).

2. Definition and Mathematical Properties of Finite-Location Causal Witnesses

A causal witness $S$ is a Hermitian operator acting on the same space as $W$, constructed so that

$\operatorname{Tr}[S W^{\rm sep}] \geq 0$

for all $W^{\rm sep}$, but there exists at least one causally nonseparable process $W \in \bigotimes_{i=1}^N (A^i_I \otimes A^i_O)$0 with $W \in \bigotimes_{i=1}^N (A^i_I \otimes A^i_O)$1 (Araújo et al., 2015, Valibouse et al., 13 Apr 2026). The cone of causal witnesses $W \in \bigotimes_{i=1}^N (A^i_I \otimes A^i_O)$2 is dual to $W \in \bigotimes_{i=1}^N (A^i_I \otimes A^i_O)$3, and characterization theorems (e.g., Theorem 1 in (Araújo et al., 2015)) provide explicit semidefinite (LMI) conditions for valid witnesses. In multipartite (finite-location) cases, the witness cone is characterized by the intersection of the duals of each order-cone.

3. Construction and Measurement Protocols

To implement a finite-location causal witness experimentally:

• Preparation of States and Inputs: In the quantum switch or related ICO processes, ancilla systems are entangled and localized at each finite event, e.g., photons prepared in path and polarization DOFs.
• Time-Delocalized Interactions: Parties such as Bob interact with the system qubit at multiple space-time points (e.g., $W \in \bigotimes_{i=1}^N (A^i_I \otimes A^i_O)$4 and $W \in \bigotimes_{i=1}^N (A^i_I \otimes A^i_O)$5), with the measurement apparatus engineered to yield a single outcome while preserving coherence (quantum eraser effect).
• Measurement and Data Acquisition: Coefficients $W \in \bigotimes_{i=1}^N (A^i_I \otimes A^i_O)$6 for the linear combination of measurement outcomes are found by semidefinite programming for optimal witness sensitivity (Valibouse et al., 13 Apr 2026). Joint outcome probabilities $W \in \bigotimes_{i=1}^N (A^i_I \otimes A^i_O)$7 are empirically measured for a tomographically complete set of local operations, and the witness value $W \in \bigotimes_{i=1}^N (A^i_I \otimes A^i_O)$8 is computed:

$W \in \bigotimes_{i=1}^N (A^i_I \otimes A^i_O)$9

A negative value certifies causal nonseparability.

A typical experimental realization for the quantum switch achieved $\operatorname{Tr} W = \prod_i d_{A^i_O}$0 with a theoretical minimum of $\operatorname{Tr} W = \prod_i d_{A^i_O}$1, confirming robust detection of indefinite causal order under realistic noise (Valibouse et al., 13 Apr 2026).

4. Generalization to Network Scenarios and the Latent Splitting Technique

Finite-location witnesses extend naturally to complex quantum networks where classical interventions fail due to space-like separation. The latent splitting procedure generalizes interventions by severing a specific latent edge and replacing it with a controlled local quantum state, resulting in new interventional distributions. In the triangle network, this allows construction of causal witnesses via explicit inequalities—for the RGB4 scenario, a polynomial witness:

$\operatorname{Tr} W = \prod_i d_{A^i_O}$2

and in the minimal binary-outcome case, a nonlinear "interventional CHSH" inequality:

$\operatorname{Tr} W = \prod_i d_{A^i_O}$3

Quantum mechanical strategies achieve $\operatorname{Tr} W = \prod_i d_{A^i_O}$4 for suitable parameters, violating classical bounds and witnessing nonclassicality (Zamora et al., 9 Jan 2026). Latent splitting thus provides a systematic, finite-location-compatible strategy for detecting nonclassical causal relations in scenarios inaccessible to standard node interventions.

5. Optimization, Completeness, and Robustness

Optimal causal witnesses are obtained via semidefinite programming duality (Araújo et al., 2015). The primal-dual pair ensures that for any causally nonseparable process, there exists a detectable $\operatorname{Tr} W = \prod_i d_{A^i_O}$5, and that the value $\operatorname{Tr} W = \prod_i d_{A^i_O}$6 quantifies the generalised robustness of nonseparability. The procedure is complete and computationally efficient for any finite set of local laboratories, including the multipartite setting.

Witness construction is robust against experimental imperfections. For example, the RGB4 polynomial witness in the triangle network admits a symmetric noise threshold $\operatorname{Tr} W = \prod_i d_{A^i_O}$7, with varying robustness depending on source symmetry; the minimal-binary nonlinear witness is similarly tolerant of practical imperfections (Zamora et al., 9 Jan 2026, Valibouse et al., 13 Apr 2026).

6. Comparison and Relation to Standard Interventions and Bell Tests

Finite-location causal witnesses extend standard techniques from causality and Bell scenarios. Unlike classical "do"-interventions, which become uninformative under space-like separation because they only yield marginals of observational distributions, latent splitting and process-matrix-based witnesses reveal nonclassical causal structure. The mathematical structure is analogous to Bell inequalities and entanglement witnesses but is specifically tailored to test for causal, rather than mere correlation, nonclassicality.

7. Application Spectrum and Outlook

Finite-location causal witnesses have broad applicability across quantum information architectures:

• Interferometric quantum switches: Allowing in-situ certification of ICO without destroying path coherence (Valibouse et al., 13 Apr 2026).
• General network nonclassicality: Detecting causal structures in scenarios such as the triangle or instrumental network, with witnesses that combine interventional and observational data (Zamora et al., 9 Jan 2026).
• Robust and tomographically complete characterization: Applying to any finite set of spacetime-localized laboratories, including multipartite and time-delocalized protocols.

A plausible implication is that finite-location causal witnesses will underpin future device-dependent certification of complex quantum network resources and inform the design of noise-tolerant protocols for fundamental and applied quantum processing.