Gaussian Steering-Annihilating Channels

• The topic defines Gaussian steering-annihilating channels as a subclass of Gaussian quantum channels that guarantee unsteerable outputs under Gaussian measurements.
• It establishes both sufficient and necessary vector criteria, using matrix inequalities to link channel parameters with the complete annihilation of EPR steering.
• The analysis outlines operational implications for continuous-variable quantum information tasks, highlighting noise thresholds, directional steering, and error-mitigation strategies.

Gaussian steering-annihilating channels constitute a rigorously defined subclass of Gaussian quantum channels for continuous-variable (CV) systems that enforce the complete destruction of EPR steering (from one party to another) under Gaussian measurements. This property is operationally significant, distinguishing channels that render all outputs unsteerable regardless of the input Gaussian resource. In recent theoretical developments, sharp definitions, matrix inequalities, and critical parameter thresholds for steering-annihilation have been established and linked to physical channel models, including noise and loss, as well as to the structure of free transformations in quantum information tasks involving steering.

1. Formal Definition and Characterization

Let $H_A \otimes H_B$ be an $(m+n)$-mode CV system. A general Gaussian channel acts as

$$\phi = \phi(K, M, \mathbf{d}) : \rho(\Gamma, \mathbf{d}_\rho) \longmapsto \rho\bigl(K\Gamma K^T + M,\, K\mathbf{d}_\rho + \mathbf{d} \bigr)$$

where $K$ is a real matrix, $M$ a real, symmetric, positive matrix, and $\mathbf{d}$ a real vector. The channel is completely positive iff $M + i\Omega_{m+n} - iK\Omega_{m+n}K^T \ge 0$.

The channel $\phi$ is Gaussian steering-annihilating (GSA) from $A$ to $B$ if, for every input Gaussian state $\rho$, the output is unsteerable $A \to B$ using Gaussian measurements. At the covariance-matrix (CM) level, with block-diagonal symplectic form $\Omega_m \oplus \Omega_n$, the formal requirement is

$$\Gamma_{\phi(\rho)} + 0_{2m} \oplus i \Omega_n \ge 0 \quad \forall\, \rho \in \mathcal{GS}(H_A \otimes H_B)$$

where $\Gamma_{\phi(\rho)}$ is the output CM.

Sufficient and Necessary Criteria

• Sufficient Criterion:

If

$$M + \bigl( 0_{2m} \oplus i\Omega_n \bigr) - i K (\Omega_m \oplus \Omega_n) K^T \ge 0 \tag{1}$$

then $\phi$ is GSA from $A$ to $B$.

• Necessary and Sufficient Criterion:

$\phi$ is steering-annihilating if and only if, for all $\mathbf{w} \in \mathbb{C}^{2(m+n)}$,

$$\mathbf{w}^\dagger M\, \mathbf{w} + \big| \mathbf{w}^\dagger K (\Omega_m \oplus \Omega_n)K^T \mathbf{w}\big| \ge \big| \mathbf{w}^\dagger (0_{2m} \oplus \Omega_n) \mathbf{w}\big| \tag{2}$$

These conditions directly link the channel's structural parameters $(K, M)$ with the requirement for output unsteerability via Gaussian measurements.

2. Mathematical Derivation and Main Results

The derivation proceeds from the general action of a Gaussian channel on first and second moments: $$\mathbf{d}_\rho \mapsto K\mathbf{d}_\rho + \mathbf{d}, \quad \Gamma_\rho \mapsto K\Gamma_\rho K^T + M$$ A Gaussian state is unsteerable $A\to B$ iff

$\Gamma + 0_{2m} \oplus i\Omega_n \ge 0$

Thus, $\phi$ is GSA iff, for all $\Gamma \ge \pm i(\Omega_m \oplus \Omega_n)$,

$K\Gamma K^T + M + 0_{2m} \oplus i\Omega_n \ge 0$

or, equivalently,

$K\Gamma K^T + M \ge \pm i (0_{2m} \oplus \Omega_n)$

The minimal value of $\mathbf{w}^\dagger K \Gamma K^T \mathbf{w}$ for $\Gamma$ in the set is $$\,\big| \mathbf{w}^\dagger K (\Omega_m \oplus \Omega_n) K^T \mathbf{w} \big|$$, reducing the full criterion to the scalar inequality (2).

Example: Channel Satisfying the Full GSA Criterion

Example 3.1 presents a $(1+1)$-mode channel $\phi_1 = \phi(K_1, M_1, 0)$: $$K_1 = \mathrm{diag}(1.03, 1.03, 0.1, 0.1), \quad M_1 = I_4$$ This channel fails the sufficient matrix test (1) but satisfies the full necessary and sufficient vector criterion (2) for all $\mathbf{w} \in \mathbb{C}^4$, verified numerically for the entire parameter space. Thus, $\phi_1$ is a steering-annihilating Gaussian channel despite not satisfying the simpler matrix criterion.

3. Gaussian Channel Models and Physical Interpretation

The operational relevance of GSA channels is contextualized by channel models that act on two-mode squeezed states (TMSS), which underpin continuous-variable EPR-steering experiments and protocols.

Lossy Channel (Beam-Splitter Model)

A lossy channel with transmissivity $\eta_L \in [0,1]$ transforms the mode $B$ as

$$\hat{B}_L = \sqrt{\eta_L}\, \hat{B} + \sqrt{1 - \eta_L}\, \hat{v}$$

and the corresponding covariance matrix parameters become

$$\alpha_L = \alpha,\quad \beta_L = \eta_L \beta + (1-\eta_L),\quad \gamma_L = \sqrt{\eta_L} \gamma$$

where $\beta,\gamma$ refer to the TMSS parameters.

Noisy (Thermal-Noise) Channel

A thermal-noise channel with transmissivity $\eta_N$ and added noise $N$ (shot-noise units): $$\hat{B}_N = \sqrt{\eta_N}\, \hat{B} + \sqrt{1-\eta_N}\, \hat{v} + \sqrt{1-\eta_N}\, \hat{n}$$ produces

$$\alpha_N = \alpha,\quad \beta_N = \eta_N\beta + (1-\eta_N)(1+N),\quad \gamma_N = \sqrt{\eta_N}\gamma$$

Correlated-Noise (Non-Markovian) Channel

Injection of correlated noise allows the engineering of revival protocols: $c = \frac{(1-\eta_N)N}{1-T}$ with $T$ the transmissivity of a secondary beam splitter.

A notable property is that pure-loss channels ($N = 0$) never induce sudden death of steering—steering decays smoothly with loss and vanishes asymptotically in the limit $\eta\to0$. In contrast, thermal-noise channels yield a finite critical excess noise for annihilation.

4. Thresholds and Directionality of Steering Annihilation

EPR steering in the CV regime is directionally asymmetric, with distinct thresholds for steering annihilation in the $A\to B$ and $B\to A$ directions. Steering (e.g., $A\to B$) disappears exactly when

$\gamma^2 \le \alpha(\beta - 1)$

with the channel-affected output parameters substituted accordingly.

The critical excess-noise thresholds for steering annihilation follow: $$N_{\rm crit}^{A\to B} = \frac{ \eta_N \left(\gamma^2 - \alpha(\beta-1)\right) }{ \alpha(1-\eta_N) }$$

$$N_{\rm crit}^{B\to A} = \frac{ \eta_N \left(\gamma^2 - \beta(\alpha-1)\right) }{ \beta(1-\eta_N) }$$

One-way steering is obtained for $N_{\rm crit}^{A\to B} < N < N_{\rm crit}^{B\to A}$. Both directions are completely annihilated when $$N \ge \max\{N_{\rm crit}^{A\to B}, N_{\rm crit}^{B\to A}\}$$.

The sudden death and revival protocols in non-Markovian correlated-noise channels allow recovery of steering lost to excess noise, provided the noise-correlation parameter is tuned as above, at the expense of additional effective loss.

5. Relationship to Related Channel Classes

The hierarchy of Gaussian channel classes is elucidated:

Symbol	Definition	Contains / Contained In
$\mathcal{GC}_{\rm SA}$	Steering-annihilating: every output unsteerable ($A \to B$)	$\subset \mathcal{GC}_{\rm MUS}$
$\mathcal{GC}_{\rm SB}$	Steering-breaking: $(\psi \otimes \mathrm{id})(\rho)$ unsteerable for every Gaussian $\rho$	Not subset/superset of GSA
$\mathcal{GC}_{\rm MUS}$	Maximal unsteerable: sends unsteerable states to unsteerable states	--

Neither $\mathcal{GC}_{\rm SA}$ nor $\mathcal{GC}_{\rm SB}$ strictly contains the other. Channels in $\mathcal{GC}_{\rm SB}$ (e.g., constant channels) may not be steering-annihilating, and vice versa. GSA channels form a strict subset of maximal unsteerable channels.

6. Operational Implication and Applications

A Gaussian steering-annihilating channel enforces the loss of all Gaussian steerability in either direction when applied to a bipartite resource. Such channels model practical Gaussian noise and loss mechanisms that render a previously EPR-steerable state ineffective for protocols requiring one-sided device-independent security, such as quantum key distribution, subchannel discrimination, and related CV quantum information tasks.

The predictive necessary–sufficient vector criterion enables systematic diagnosis and engineering of CV systems robust against steering-annihilating environments. In quantum networks, this insight governs noise budgeting, informs security thresholds, and underlies the feasibility of active noise-mitigation through correlated channel engineering. The distinction between smooth and finite-time decoherence of steering, along with revival strategies, provides a foundation for exploring resilience and error correction in CV quantum information.

7. Design Guidelines and Future Directions

The structural results yield concise design rules:

• Pure-loss Gaussian channels ($N = 0$): steering vanishes only at zero transmissivity.
• Thermal-noise channels ($N > 0$): finite critical noise thresholds for annihilation in each direction, computable from the above formulas.
• Non-Markovian correlated-noise channels can remove channel-induced noise via engineered correlations given $c = \frac{(1-\eta)N}{1-T}$, at the cost of increased effective loss.
• Directionality is ubiquitous except in special symmetric cases; steering annihilation must be assessed separately for both $A \to B$ and $B \to A$.

A plausible implication is that GSA channels delineate a practical boundary for one-sided device-independent quantum systems based on continuous variables. Their full mathematical characterization informs both the limits of quantum resource transmission in the presence of noise and the strategies for channel engineering in large-scale CV quantum networks.

For foundational protocols and a taxonomy of channel types, see "Several kinds of Gaussian quantum channels" (Ma et al., 7 Nov 2025) and for experimental/analytical studies of sudden death and revival, "Sudden death and revival of Gaussian Einstein-Podolsky-Rosen steering in noisy channels" (Deng et al., 2021).