Entanglement certification via causal-order interferometry in a quantum switch

Abstract: Entanglement certification is often performed on states that have already undergone noisy transmission or processing. Noise can reduce the surviving entanglement and can also cause a given criterion to fail even when entanglement remains. In this context, the quantum switch, a paradigmatic realization of indefinite causal order (ICO), coherently controls the orders in which two channels act and has been shown to offer advantages across a range of quantum information-processing tasks. Here we ask whether this coherent control enlarges the noise-parameter region in which entanglement remains certifiable. We regard the two order branches as the arms of a causal-order interferometer and insert a local unitary between the channel uses to tune their interference. For stochastic Pauli noise, a postselected ICO output can exhibit greater entanglement negativity than any classical mixture of the two definite orders; in particular, we identify regimes where its negativity remains nonzero while that of every classical mixture vanishes. A suitable local Pauli unitary substantially enlarges this ICO-only region, while an input-dependent path-difference indicator qualitatively links operator noncommutativity to the postselected negativity gain. Numerical examples extend the advantage to local amplitude-damping noise and two-qutrit Weyl noise. At a representative Weyl-noise point for the $3\times 3$ positive-partial-transpose (PPT) Tiles bound-entangled state, a nondecomposable witness detects the postselected ICO output, whereas an analytic bound excludes detection of the definite-order outputs and their mixtures by the entire locally rotated witness family. These results identify causal-order interferometry as a strategy for enhancing entanglement certification across distinct noise models and dimensions.

• The paper develops a separability-preserving quantum-switch framework that uses coherent interference between channel orders and postselection to certify entanglement without creating it.
• The method expands negativity-based certification regions for Pauli, amplitude-damping, and two-qutrit Weyl noise, with selected branches remaining NPT even when all definite-order mixtures become PPT.
• The protocol also detects PPT bound entanglement in a noisy Tiles UPB state, achieving a witness value of approximately −5.93 × 10⁻³ with about 97.01% success probability while the output remains PPT.

The paper develops a framework for certifying bipartite entanglement by preprocessing a noisy state through a quantum switch, treating the two channel orders as the arms of a causal-order interferometer. The central question is whether indefinite causal order (ICO) enlarges the noise-parameter region in which a fixed entanglement criterion—negativity for NPT states or nondecomposable witnesses for PPT states—remains successful, relative to the best classical mixture of the two definite causal orders (DCO). The authors answer affirmatively across Pauli, amplitude-damping, and two-qutrit Weyl noise models, and extend the result to PPT-entangled inputs via an analytically certified witness.

Causal-order interferometry and the certification task

The setup compares two preprocessing strategies applied to a bipartite input $\rho_{AB}$ followed by the same $A|B$ entanglement criterion. In the DCO benchmark, $\rho_{AB}$ passes through channels $\mathcal{E}_1$ and $\mathcal{E}_2$ sequentially, with an optional local unitary $U = U_A \otimes U_B$ inserted between them; the benchmark is the full family of classical mixtures $$\rho_f^{\mathrm{MIX}}(\xi) = \xi\rho_f^{\rightarrow} + (1-\xi)\rho_f^{\leftarrow}$$. In the ICO protocol, a control qubit initialized in $|+\rangle_c$ coherently controls the order; measuring it in the phase basis $|\pm_\phi\rangle$ induces conditional Kraus operators that are coherent sums of the ordered Kraus products, $$M_{ij}^{\pm,\phi} = (H_{ij}^{\rightarrow} \pm e^{-i\phi}H_{ij}^{\leftarrow})/2$$, so postselection retains cross-order interference terms absent from any classical mixture.

A key methodological constraint is that each conditional map must be separability preserving across $A|B$0; otherwise postselection could manufacture the detected entanglement. Lemma 1 establishes that if each trace-nonincreasing branch admits a separable Kraus decomposition, no branch creates entanglement from separable inputs, and if each branch is PPT-preserving, the probability-weighted negativity obeys $A|B$1. The reported gains therefore arise from selection among outcomes rather than entanglement generation.

Exact path selection under Pauli noise

For Pauli Kraus operators and a Pauli-string insertion $A|B$2, Theorem 1 shows that the two ordered products of each path pair differ only by a sign determined by the commutation label $A|B$3, so an $A|B$4-basis control measurement routes each path pair exclusively to one outcome. Because $A|B$5 depends on $A|B$6, the inserted unitary determines which outcome retains which noise components: postselection can discard the Pauli components most harmful to certification.

Applied to a Bell state subject to two global depolarizing channels on $A|B$7, this yields a concrete advantage. With $A|B$8, the negativity gain is positive over a broad region but small and the ICO-only region narrow. With $A|B$9, the ICO-only NPT-certifiable region expands substantially, the peak gain grows by roughly an order of magnitude, and the selected output remains NPT even as one channel approaches complete depolarization—a regime where every DCO output and all their mixtures are PPT. The cost is probabilistic: the optimal success probability never falls below $\rho_{AB}$0 over the scanned region, approaching that minimum precisely in the asymmetric near-depolarizing regime overlapping the ICO-only region, consistent with the average monotonicity bound.

Path difference as an indicator of interference strength

The authors introduce the path-difference operator $\rho_{AB}$1, which decomposes into commutators involving $\rho_{AB}$2, $\rho_{AB}$3, and $\rho_{AB}$4, reducing to $\rho_{AB}$5 when $\rho_{AB}$6. Its aggregate $\rho_{AB}$7 equals four times the unnormalized $\rho_{AB}$8-conditional output and is representation independent. The Hilbert–Schmidt norm $\rho_{AB}$9 serves as an input-dependent scalar indicator. Under general local unitaries—where separability preservation is not guaranteed—the scans show a qualitative correlation between $\mathcal{E}_1$0 and the negativity gain. A spectral analysis in the appendix makes this precise: NPT in the $\mathcal{E}_1$1 branch requires a directional crossing condition on projections of $\mathcal{E}_1$2 onto near-zero eigendirections of $\mathcal{E}_1$3, and a necessary threshold $\mathcal{E}_1$4 holds when the mixed output is PPT. Larger $\mathcal{E}_1$5 permits, but does not guarantee, larger gain.

Non-Pauli noise and high-dimensional extensions

For local amplitude-damping channels with $\mathcal{E}_1$6, exact sign selection fails, yet the composition remains order-independent and every nonvanishing conditional Kraus operator stays a product operator, so Lemma 1 still applies. The negativity gain is positive at every sampled interior grid point but small, and the ICO and DCO NPT regions are largely comparable—showing that exact Pauli path selection is sufficient but not necessary for a gain.

In the two-qutrit setting, Weyl operators replace Pauli strings: ordered products differ by phases rather than signs, and Corollary 1 shows the path weights become phase dependent, $\mathcal{E}_1$7. A qubit control cannot fully sort all Weyl paths; complete phase sorting would require higher-dimensional controls. Nevertheless, for a maximally entangled two-qutrit state under joint depolarizing Weyl channels with $\mathcal{E}_1$8, the fixed $\mathcal{E}_1$9 branch dominates both definite orders throughout the scanned space and certifies NPT entanglement over a strictly broader region.

PPT entanglement detection with the Tiles UPB state

Since negativity cannot detect PPT-entangled states, the paper turns to nondecomposable witnesses using the $\mathcal{E}_2$0 Tiles UPB bound-entangled state. The authors prove, via a Gram-matrix certificate for a sextic polynomial verified in exact arithmetic, that $\mathcal{E}_2$1, validating the witness $\mathcal{E}_2$2 and its entire locally rotated family. At representative Weyl-noise parameters $\mathcal{E}_2$3 with $\mathcal{E}_2$4, the definite-order outputs coincide and admit the analytic lower bound $\mathcal{E}_2$5, attained globally at $\mathcal{E}_2$6. Hence no witness in the continuous family detects either definite-order output or any mixture. The same witness $\mathcal{E}_2$7 gives $\mathcal{E}_2$8 on the ICO conditional output, with success probability $\mathcal{E}_2$9. Notably, the conditional output remains PPT (minimum partial-transpose eigenvalue $U = U_A \otimes U_B$0), so the detection is not obtained by converting the output to an NPT state—the ICO protocol certifies genuinely bound entanglement where the benchmark family fails entirely.

A complementary channel-level result (Proposition 1) gives an analytic advantage region for single-qubit depolarizing channels in the Choi-state picture: whenever $U = U_A \otimes U_B$1 and $U = U_A \otimes U_B$2, all DCO Choi outputs are PPT while the $U = U_A \otimes U_B$3-conditioned output is NPT; in the symmetric case this spans approximately $U = U_A \otimes U_B$4.

Limitations and open questions

Several restrictions are acknowledged explicitly. Separability preservation of the conditional maps is proven only for the Pauli, Weyl, and identity-insertion amplitude-damping constructions; for general local unitaries and general channels, the scans compare output negativities without certifying input entanglement. The analysis is confined to discrete-variable systems with noise acting only on the target, excluding noisy implementations of the switch itself. The correlation between $U = U_A \otimes U_B$5 and negativity gain is established only qualitatively within restricted depolarizing families, and the spectral threshold conditions are necessary rather than sufficient. For Weyl noise, a qubit control cannot achieve complete phase sorting, leaving open how much additional advantage higher-dimensional or multipath controls would provide. Whether the ICO-only regions persist under more general noise structures, and whether the witness-based PPT advantage extends beyond the specific locally rotated Tiles family, remain unresolved.

Conclusion

This work establishes quantum-switch preprocessing as a resource for entanglement certification: by tuning interference between causal orders with an inserted local unitary, postselection can route detrimental noise paths into discarded outcomes, extending the certifiable parameter region for NPT states under Pauli, amplitude-damping, and Weyl noise, and enabling witness-based detection of a PPT bound-entangled output where an analytic bound excludes detection by the entire benchmark witness family. The separability-preservation lemma ensures these gains certify input entanglement rather than artifacts of postselection, grounding the approach as a principled strategy for certification under realistic noise.