Temporal nonlocality of a qudit resides in the input state, not the channel, and certifies temporal teleportation up to a fundamental limit

Technical Summary: Temporal Nonlocality of a Qudit

Problem Statement
The paper addresses the nature of quantum correlations across time for a single system (a qudit) passing through a noisy channel. While spatial nonlocality is well-understood as a property of shared entangled states, the source of "temporal nonlocality" (Bell-like violations in time) has been less clear. Specifically, it is unknown whether temporal nonlocality arises from the input state, the channel dynamics, or their interplay. Furthermore, there is a need to establish a rigorous hierarchy of temporal correlations (entanglement, steering, nonlocality) analogous to the spatial hierarchy, and to determine the operational utility of these correlations for tasks like temporal teleportation. A critical open question is whether device-independent certification of temporal correlations provides a faithful guarantee of the channel's ability to transmit quantum information, or if it suffers from "over-certification."

Methodology
The authors employ a unified framework based on semidefinite programming (SDP) to quantify the robustness of temporal correlations against noise. They define three robustness measures for a single qudit sent through a channel $E$ and measured at two time steps ($t_A$ and $t_B$):

1. Temporal Entanglement Robustness ($TER$): Specifically $TER_{sep}$ (based on the separability of the channel Choi state) and $TER_{PPT}$ (based on the Peres-Horodecki PPT criterion).
2. Temporal Steering Robustness ($TSR$): Quantifying the minimal noise required to destroy the assemblage's ability to be explained by a hidden-state model.
3. Temporal Nonlocality Robustness ($TNR$): Quantifying the minimal noise required to render the two-time statistics local (admitting a Local Hidden Variable model).

The study utilizes a two-MUB (Mutually Unbiased Bases) measurement scheme, where Alice's settings include the eigenbasis of the input state $\rho_A$ and its Fourier conjugate. To handle dimensionality $d \geq 3$ and avoid contextuality artifacts associated with standard $SU(d)$ generators, the authors utilize a non-contextual Wigner phase-space realization for the pseudo-density operator (PDO).

The analysis combines:

• Analytic Proofs: Establishing conditions for the vanishing of robustness measures and deriving hierarchy bounds.
• Numerical Verification: Extensive Monte-Carlo sweeps ($10^6$ configurations) for dimensions $d=2$ through $d=5$, testing random states, channels (depolarizing, phase damping, amplitude damping), and times.

Key Contributions and Results

1. State-Boundness of Temporal Nonlocality:
   The central finding is that for standard noise families, temporal nonlocality is asymmetrically state-bound. The robustness $TNR(\rho_A, E)$ vanishes if and only if the input state is maximally mixed ($\rho_A = 1/d$), regardless of the channel (provided the channel is not adversarial in a specific way).

   • Mechanism: The resource is the back-action of the measurement on a non-maximally mixed state. If $\rho_A \neq 1/d$, Alice's choice of basis disturbs the state in a way that Bob's statistics detect, creating a temporal signal. This signal persists even if the channel completely decoheres the state.
   • Implication: The channel itself is dispensable for the existence of temporal nonlocality; the resource resides entirely in the input state's mixedness.

2. Hierarchy of Temporal Robustness:
   The authors establish a strict hierarchy for the robustness measures:

   • Lower Bound: $TSR \geq TNR \geq 0$ holds universally for all inputs and channels.
   • Upper Bound: The inequality $TER_{sep} \geq TSR$ holds only under the No-Signaling-in-Time (NSIT) condition (i.e., when $\rho_A = 1/d$).
   • NSIT Violation: When $\rho_A \neq 1/d$, the input state violates NSIT, and the steering resource ($TSR$) can exceed the channel's entanglement robustness ($TER_{sep}$). The authors derive a universal, NSIT-corrected bound: $TSR \leq TER_{sep} + \frac{1}{2}V_{NSIT}$, where $V_{NSIT}$ is the NSIT violation monitor.

3. Operational Meaning: Temporal Teleportation:
   The paper links $TNR$ to the fidelity of device-independent temporal teleportation (DI-TIT). In a prepare-and-measure protocol using Heisenberg-Weyl encoding:

   • The certified $TNR$ provides a lower bound on the average teleportation fidelity: $F_{DI} \geq 1/d + (d-1)TNR/d$.
   • For $d=3$, the supremal guaranteed fidelity is $7/9$.

4. The "Over-Certification" Trap:
   A significant finding is that $TNR$ can over-certify the channel's performance. Because $TNR$ depends on the input state's back-action rather than the channel's coherence transmission, a "protected" input state (e.g., an eigenstate of a phase-damping channel) can yield maximal $TNR$ even when the channel destroys the coherence necessary for teleporting superpositions.

   • Resolution: Honest certification is guaranteed only if the probe state is sufficiently mixed. The paper provides a mixedness threshold ($\lambda_{min} \geq 1/d - p/(d-1)$) that ensures $TNR$ does not exceed the channel's actual twirl parameter $p$. No choice of probes can make the certificate universally valid for all channels.

5. Numerical Verification:
   The results are verified numerically for $d=2, 3, 4, 5$. The data confirms that $TNR=0$ occurs exclusively at $\rho_A = 1/d$ (under adapted measurements), the lower hierarchy $TSR \geq TNR$ holds to machine precision, and the upper hierarchy breaks exactly in proportion to NSIT violations.

Significance and Claims
The paper claims to resolve the foundational question of where temporal nonlocality resides, identifying it as a property of the input state's mixedness rather than the channel. This contrasts with spatial nonlocality, which is a property of the shared state.

The work provides a unified operational framework for temporal correlations, establishing that:

• Temporal nonlocality is a resource for device-independent temporal teleportation.
• There is a fundamental limit to what can be certified device-independently ($7/9$ for $d=3$) due to the decoupling of the certification metric ($TNR$) from the channel's actual transmission capability.
• The "trap" of over-certification is a general feature of temporal-correlation-based certification, necessitating mixed probes for honest guarantees.

The authors emphasize that their results are derived for standard noise families and adapted measurement schemes, and they explicitly note that the equivalence $TNR=0 \iff \rho_A=1/d$ is not universal over all injective channels (e.g., adversarial unitaries can hide the resource from fixed schemes). The paper does not propose new experimental setups but rather provides the theoretical bounds and certification protocols for existing temporal quantum memory and teleportation scenarios.