Passive Environment-Assisted Quantum Communication

• Passive environment-assisted quantum communication is a model where a helper pre-sets the environment’s state, transforming the induced channel and altering capacity characteristics.
• It employs techniques like matched GKP ancillas, non-Gaussian ancillas, and trigger-signal steering to overcome the standard pure-loss limitations of bosonic channels.
• The framework extends to capacity theory and practical QKD implementations, addressing challenges such as energy constraints, memory effects, and experimental imperfections.

Passive environment-assisted quantum communication is a communication model in which a quantum channel is realized as a unitary interaction between an information-carrying system and an environment, while a helper controls only the environment’s initial state and does not measure the environment, apply feedforward, or use classical side information during transmission. In this setting, the induced channel is changed from the standard “vacuum environment” or “thermal environment” form by replacing the environment input with a tailored state, and the resulting communication properties can differ sharply from those of the unassisted channel. The subject spans an abstract capacity theory for unitary dilations, explicit bosonic constructions for beam splitters and attenuators, non-Gaussian ancilla design, memory-based passive steering of environments, and experimentally motivated passive architectures in continuous-variable and time-bin quantum communication (Karumanchi et al., 2014, Wang et al., 2024).

1. Formal model and operational meaning

In the general passive-helper model, a single use of the physical link is represented by an isometry or unitary

$U^{AE \to BE'}: A \otimes E \to B \otimes E',$

where $A$ is the sender’s input, $E$ is the environment input, $B$ is the receiver’s output, and $E'$ is the environment output. Fixing an environment state $\sigma_E$ induces the completely positive trace-preserving map

$$\mathcal{N}_\sigma(\rho_A)=\operatorname{Tr}_{E'}\!\left[U(\rho_A\otimes \sigma_E)U^\dagger\right].$$

The helper may prepare product environment states across channel uses or arbitrary correlated states across uses, but remains passive after preparation. In an extended model, the helper and receiver may also share prior entanglement, changing the induced channel to $A\to BH$ (Karumanchi et al., 2014).

This notion is distinct from active environment assistance. In active schemes, the environment output is measured and the outcome is used for feedforward correction, possibly with two-way classical communication. In passive assistance, none of those operations are allowed: the only resource is the initial choice of the environment state. A common misconception is that “passive” means “no control.” In the formal usage of the subject, passive assistance can involve highly structured environment preparation, including non-Gaussian states, correlated helper inputs across uses, or helper–receiver entanglement prepared before communication begins (Karumanchi et al., 2014, Wang et al., 2024).

For bosonic channels, the same idea is expressed through a beam splitter or other Gaussian unitary. If a system mode $a$ and an environment mode $b$ interact through a beam splitter of transmissivity $A$0, then passive assistance corresponds to choosing a fixed ancilla state $A$1 at the environment input and using the induced channel

$A$2

The choice $A$3 reproduces the standard pure-loss channel, but a nontrivial $A$4 can fundamentally change the channel seen by the receiver (Wang et al., 2024).

2. Bosonic loss, anti-degradability, and the threshold problem

For the unassisted bosonic pure-loss channel, the environment is vacuum and the channel is the canonical attenuator of transmissivity $A$5. In the unlimited-energy limit, its quantum capacity per mode is

$A$6

so $A$7 for $A$8. This is the well-known 50% loss threshold. The beam splitter realization is specified by the Heisenberg relations

$A$9

or, in quadratures,

$E$0

With vacuum at the environment input, the reduced map on the system is pure loss; with a structured environment state, the channel is no longer the standard pure-loss map (Wang et al., 2024).

The same threshold appears in the Gaussian-helper analysis of two-mode Gaussian unitaries. Under a Gaussian helper restriction, beam splitters with $E$1 are Gaussian universally degradable and beam splitters with $E$2 are Gaussian universally anti-degradable. In that restricted setting, the single-letter quantum capacity is

$E$3

and it vanishes for $E$4. The same work also shows that no nontrivial two-mode Gaussian unitary is universally degradable or universally anti-degradable for all environment states, Gaussian or not, unless $E$5 (Oskouei et al., 2021).

A central point of passive assistance is therefore not that it “improves” the standard pure-loss channel while keeping the same channel model, but that it replaces the vacuum-environment channel by a different induced channel. In characteristic-function language, for a beam splitter one has

$E$6

$E$7

so zeros or lattice structure in the ancilla characteristic function can suppress logical signatures at the environment output while preserving them at the receiver output (Voss et al., 25 Feb 2026).

3. GKP-assisted passive transmission through beam splitters

The most explicit passive construction known in the bosonic setting uses Gottesman–Kitaev–Preskill states. GKP codes encode finite-dimensional logical systems in a single bosonic mode by stabilizing against small displacements in phase space. For a square-lattice qubit code, the stabilizers are

$E$8

and logical operators can be chosen as

$E$9

More generally, a symplectic lattice basis $B$0 with $B$1 defines a $B$2-dimensional GKP code $B$3 (Wang et al., 2024).

The passive construction matches the system GKP code and the environment GKP code to the beam splitter transmissivity. If the environment mode is prepared in a $B$4-dimensional GKP code with lattice basis vectors $B$5, then for the system mode one defines

$B$6

with

$B$7

Because $B$8 and $B$9 are environment stabilizers, the evolved logicals can be represented so that the receiver output contains the system logical information while the environment output carries no logical information about it. The physical pictures given for this mechanism are position-space interference and characteristic-function filtering: the environment GKP characteristic function acts like a periodic filter that removes the logical harmonics from the wrong output while preserving them in the right one (Wang et al., 2024).

For rational transmissivities, the construction becomes exact. If

$E'$0

and the system and environment lattices satisfy

$E'$1

with $E'$2 and $E'$3, then the beam splitter perfectly and simultaneously transmits the code in mode 1 to its output port and the code in mode 2 to its output port, with no measurement or feedforward. The output marginal states lie in larger GKP codes with

$E'$4

so the beam splitter embeds each input GKP code into a larger output code. The two outputs are entangled, but each input logical subsystem is mapped to a local logical subsystem at its corresponding output (Wang et al., 2024).

This result removes the unassisted 50% threshold at the logical level. The threshold applies to the standard pure-loss channel with vacuum environment; by choosing a matched GKP environment state, the induced channel becomes an isometry on the encoded GKP subspace for any $E'$5 with $E'$6. The construction also extends beyond the ideal infinite-energy setting. Finite-energy GKP states are taken as

$E'$7

with mean photon number approximately

$E'$8

For equal $E'$9 on the two inputs, $\sigma_E$0 commutes with the beam splitter, and the entanglement infidelity scales approximately as

$\sigma_E$1

At $\sigma_E$2 and $\sigma_E$3, tri-convex optimization yields $\sigma_E$4 under an average energy constraint $\sigma_E$5 per mode, with coherent information $\sigma_E$6 qubits per mode; the optimal environment and encoder are well fit by GKP states, with fidelity $\sigma_E$7 to finite-energy GKP on a hexagonal lattice (Wang et al., 2024).

4. Non-Gaussian ancillas beyond ideal GKP and passive steering by memory

A major practical issue is that ideal GKP ancillas are difficult to realize experimentally. A later bosonic study therefore considers more experimentally accessible non-Gaussian ancillas, including Fock states, cat states, and squeezed cat states, together with optimized encoders and decoders. The framework maximizes entanglement fidelity

$\sigma_E$8

and also reports the single-letter coherent information

$\sigma_E$9

Numerically, the optimization is performed by alternating semidefinite programming over encoder and decoder, typically converging in about 150 rounds from random initializations (Voss et al., 25 Feb 2026).

Representative results show that low-energy non-Gaussian ancillas can already lift the below-threshold regime. With a Fock ancilla $$\mathcal{N}_\sigma(\rho_A)=\operatorname{Tr}_{E'}\!\left[U(\rho_A\otimes \sigma_E)U^\dagger\right].$$0 and optimized encoding at $$\mathcal{N}_\sigma(\rho_A)=\operatorname{Tr}_{E'}\!\left[U(\rho_A\otimes \sigma_E)U^\dagger\right].$$1, the reported entanglement fidelity is approximately $$\mathcal{N}_\sigma(\rho_A)=\operatorname{Tr}_{E'}\!\left[U(\rho_A\otimes \sigma_E)U^\dagger\right].$$2 and the coherent information is approximately $$\mathcal{N}_\sigma(\rho_A)=\operatorname{Tr}_{E'}\!\left[U(\rho_A\otimes \sigma_E)U^\dagger\right].$$3. Cat ancillas with $$\mathcal{N}_\sigma(\rho_A)=\operatorname{Tr}_{E'}\!\left[U(\rho_A\otimes \sigma_E)U^\dagger\right].$$4 and squeezed cat ancillas with $$\mathcal{N}_\sigma(\rho_A)=\operatorname{Tr}_{E'}\!\left[U(\rho_A\otimes \sigma_E)U^\dagger\right].$$5, $$\mathcal{N}_\sigma(\rho_A)=\operatorname{Tr}_{E'}\!\left[U(\rho_A\otimes \sigma_E)U^\dagger\right].$$6 also yield positive coherent information for some $$\mathcal{N}_\sigma(\rho_A)=\operatorname{Tr}_{E'}\!\left[U(\rho_A\otimes \sigma_E)U^\dagger\right].$$7. For Fock $$\mathcal{N}_\sigma(\rho_A)=\operatorname{Tr}_{E'}\!\left[U(\rho_A\otimes \sigma_E)U^\dagger\right].$$8 ancillas and $$\mathcal{N}_\sigma(\rho_A)=\operatorname{Tr}_{E'}\!\left[U(\rho_A\otimes \sigma_E)U^\dagger\right].$$9–$A\to BH$0, the numerically optimized encodings resemble hexagonal GKP states; evaluated against $A\to BH$1 environments with $A\to BH$2, coherent information around $A\to BH$3 is observed near optimal $A\to BH$4 (Voss et al., 25 Feb 2026).

Several analytical schemes complement the numerics. In a cat-ancilla model, a special operating point yields a degradable channel with

$A\to BH$5

while the decoded channel obtained from the Petz map satisfies

$A\to BH$6

and the optimal channel fidelity under Petz equals $A\to BH$7. In comb-ancilla constructions, for $A\to BH$8 one obtains

$A\to BH$9

and

$a$0

as the number of comb teeth $a$1 increases. In a high-Fock scheme with environment $a$2 and a $a$3 code, optimizing over $a$4 gives $a$5 near $a$6, and an explicit decoder at $a$7 yields $a$8 and $a$9 (Voss et al., 25 Feb 2026).

Passive assistance need not rely on direct laboratory access to the environment input. A memory-based protocol for bosonic attenuators realizes passive assistance by sending “trigger signals” before the data-carrying signal. If successive uses are separated by times much shorter than the environment relaxation time, the environment does not reset, and the trigger train steers it into a favorable non-thermal state. For a collective trigger mode

$b$0

preparing the trigger state $b$1 leads to an environment state $b$2 obeying

$b$3

With only two triggers, one has

$b$4

and for sufficiently small $b$5 the protocol yields

$b$6

This operationalizes the die-hard effect $b$7 for every transmissivity $b$8 without directly accessing the environment state (Mele et al., 2022).

Scheme	Passive environment strategy	Representative result
Matched GKP ancilla	Tailored GKP state at the beam-splitter environment port	Perfect transmission for rational $b$9 with ideal GKP; $A$00 at $A$01, $A$02 (Wang et al., 2024)
Accessible non-Gaussian ancilla	Fock, cat, or squeezed cat state at the dark port	At $A$03, Fock $A$04 gives $A$05 and $A$06 (Voss et al., 25 Feb 2026)
Trigger-signal steering	Passive preconditioning of a memoryful environment	Two triggers yield $A$07 for small $A$08 (Mele et al., 2022)

5. Capacity theory, helper correlations, and structural order results

The abstract capacity theory of passive environment assistance is more general than the bosonic beam-splitter setting. For a unitary interaction $A$09, the unrestricted-helper quantum capacity is

$A$10

where the helper may choose a correlated environment state $A$11 across uses. The product-helper version restricts $A$12 to product states, and a further extension allows prior entanglement between helper and receiver, yielding $A$13 (Karumanchi et al., 2014).

A basic structural fact is that helper correlations matter. The product-helper capacity can differ from the unrestricted capacity, and prior shared entanglement between helper and receiver can make a further difference. For two-qubit unitaries, a single-letter expression exists for the product-helper capacity because the induced qubit channel is either degradable or anti-degradable for every pure environment input:

$A$14

By contrast, the unrestricted-helper setting can exhibit superactivation and self-superactivation: there are explicit families for which $A$15 but $A$16, and there are pairs $A$17 with $A$18 but $A$19 when the helper entangles the environments across uses (Karumanchi et al., 2014).

In Gaussian bosonic communication, the capacity theory persists under energy constraints. For a Gaussian isometry $A$20, the energy-constrained passive-helper quantum capacity is

$A$21

with $A$22 satisfying the helper energy constraint. The same work gives a multi-letter formula for the classical capacity and an uncertainty-type relation for classical product-state capacities,

$A$23

This lower bound implies a nontrivial conferencing-encoder rate even when neither assisted capacity is characterized exactly (Oskouei et al., 2021).

Another structural line concerns order preservation rather than explicit rates. Passive-environment bosonic channels, defined by energy-preserving linear coupling to an environment in a passive state followed by tracing out the environment, preserve Fock majorization:

$A$24

On passive states, standard majorization is also preserved. The underlying channel action on Fock-diagonal inputs is represented by lower-triangular, column-stochastic transition matrices, reflecting the fact that a passive environment cannot supply work to the signal mode. These results constrain output entropies, energy monotones, and majorization orderings under passive-environment noise, even though the work does not provide explicit capacity formulas (Jabbour et al., 2018).

6. Implementations, broader passive architectures, and open directions

Beam-splitter-based passive assistance is particularly natural for quantum transduction. A two-mode transducer implements an effective beam splitter between bosonic modes such as microwave–optical, optical–optical, or microwave–microwave modes, and the idle input port of the other mode is precisely the environment port that can be prepared passively. In the GKP construction, the required resources are finite-energy GKP preparation in at least one mode, a stable beam splitter interaction with known $A$25, and standard GKP syndrome extraction or decoding at the receiver. Platforms explicitly mentioned for these tasks are superconducting microwave cavities, trapped-ion motion, and rapidly advancing optical implementations (Wang et al., 2024).

Finite-energy and non-Gaussian imperfections determine the near-term feasibility of passive assistance. In the GKP beam-splitter setting, simulations show that high entanglement fidelity and large coherent information persist for a few-percent intrinsic loss; at $A$26 and $A$27, tri-convex optimization yields positive coherent information up to approximately $A$28 intrinsic loss probability. Thermal noise, mode mismatch, phase drifts, finite detector efficiency, and approximate GKP preparation degrade the periodic filtering mechanism, but finite-energy tolerance helps. In the non-Gaussian ancilla setting, deterministic generation of low-$A$29 Fock states is standard in cavities and circuits, optical cats with $A$30–$A$31 and squeezed cats with $A$32 are within reach, and optical GKP states are maturing, although bright multimode GKP remains difficult (Wang et al., 2024, Voss et al., 25 Feb 2026).

A broader protocol-level use of passivity appears in quantum key distribution. In continuous-variable QKD, passive state preparation uses the intrinsic Gaussian fluctuations of an amplified spontaneous emission source rather than active amplitude and phase modulators. The emitted ensemble is identical to the Gaussian-modulated coherent-state ensemble from Eve’s perspective, and the main technical parameter is the passive-preparation excess noise

$A$33

Experimentally, the excess noise is effectively suppressed by optical attenuation, and secure key generation over metro-area distances is reported, with simulations yielding distances beyond $A$34 km at $A$35 dB/km (Qi et al., 2020).

Time-bin QKD over multimode channels provides a different passive design philosophy. A field-widened, imaging 4$A$36 multimode interferometer and a reference-frame-independent protocol remove the need for active mode filtering, adaptive optics, active basis selection, and active phase stabilization. Over a $A$37 m graded-index multimode fiber channel, the experiment reports visibilities of approximately $A$38–$A$39, $A$40 or $A$41, and a sustained asymptotic secure key rate greater than $A$42 bits/coincidence; the reference-frame-independent parameter

$A$43

remains essentially constant even under an imposed phase drift of $A$44 rad/s (Tannous et al., 2023).

Open questions remain at several levels. In the capacity theory, exact energy-constrained environment-assisted quantum capacities are open, as are conditions for single-letterization beyond special degradable cases. In bosonic implementations, energy–rate tradeoffs for ancilla preparation, decoder complexity outside Petz-optimal points, and extensions to thermal-loss, phase-noise, multimode, and memory channels remain unsettled. For GKP-based passive assistance, exceptional transmissivities of the form $A$45 support only trivial $A$46 under the rational-matching construction, and full classifications of Gaussian unitaries and matched lattices are still of interest. These limitations coexist with a clear general conclusion: passive control of the environment can qualitatively alter quantum communication thresholds, capacities, and implementation strategies without requiring measurements or feedforward during transmission (Wang et al., 2024, Voss et al., 25 Feb 2026, Oskouei et al., 2021).