Conditional Quantum Communication

• Conditional quantum communication is a family of quantum-information tasks where outcomes like entanglement generation or secure messaging depend on auxiliary conditions such as side outputs or heralding events.
• The QCMI capacity theorem establishes that the asymptotic entanglement rate is half the maximum conditional mutual information, linking fundamental entropy measures to practical channel coding.
• These protocols extend to cryptographic applications and indefinite causal structures, exemplified by quantum one-time pads, conditional disclosures, and coherently controlled communication orders.

Conditional quantum communication denotes a family of quantum-information tasks in which communication, entanglement generation, secrecy, recovery, or routing is contingent on auxiliary systems or conditions. In the most explicit channel-coding formulation, Alice uses a quantum broadcast channel $\Phi:A\to B\otimes C$, Bob exploits the side output $C$, and the exact asymptotic ebit-generation capacity is

$Q_K(\Phi)=\frac12\max_{\rho_A} I(R:B\mid C),$

thereby giving the quantum conditional mutual information (QCMI) a direct channel-coding meaning (Wang, 24 Jun 2026). Closely related uses of the same conditional structure appear in state redistribution, conditional quantum one-time pads, conditional disclosure of secrets, heralded quantum memories, and protocols in which the signaling order is coherently controlled by a quantum system (Brandao et al., 2014, Sharma et al., 2017, Asadi et al., 2024, Saglamyurek et al., 2011, Feix et al., 2015).

1. Scope of the term in the literature

The literature does not employ a single uniform definition of conditional quantum communication. Instead, the phrase appears across several technically distinct settings in which the recoverability, secrecy, or utility of quantum information depends on side information, a helper system, a public predicate, a heralding event, or a control qubit governing causal order (Wang, 24 Jun 2026, Brandao et al., 2014, Sharma et al., 2017, Asadi et al., 2024, Saglamyurek et al., 2011, Feix et al., 2015).

Setting	Condition	Core statement
Broadcast-channel entanglement generation	Assistant output $C$ helps decoding and must be decoupled	$Q_K(\Phi)=\frac12\max_{\rho_A}I(R:B\mid C)$
State redistribution	Bob’s side information $B$ reduces communication cost	$Q_*=\frac12 I(C;R\mid B)$
Conditional quantum one-time pad	Secure subsystem $B$ versus insecure subsystem $E$	Capacity $I(A;B\mid E)$; quantum-message version $C$0
Quantum conditional disclosure of secrets	Reveal the secret iff $C$1	Correctness and privacy are imposed in diamond norm
Heralded memory readout	Condition on detection of the partner photon	Conditional average storage fidelity $C$2
Superposed communication order	Order is coherently controlled by a qubit	$C$3 in the tripartite task

This suggests a unifying feature: the relevant resource is not merely quantum signaling, but quantum signaling constrained or unlocked by ancillary structure. In some settings the condition is informational, as in QCMI-governed tasks; in others it is cryptographic, optical, or causal.

2. Broadcast-channel conditional quantum communication and the QCMI capacity theorem

In the broadcast-channel model, Alice holds system $C$4. Through a quantum broadcast channel $C$5, Alice’s input emerges partly at Bob as $C$6 and partly at an assistant or “pilot” station as $C$7. No prior entanglement is shared, while classical public communication between the parties is allowed but not counted toward the rate. The goal is to establish noiseless quantum correlation, i.e. ebits, between Alice’s purification reference $C$8 and Bob’s system $C$9, while allowing the assistant $Q_K(\Phi)=\frac12\max_{\rho_A} I(R:B\mid C),$0 to help the decoding; the assistant must be decoupled from the final $Q_K(\Phi)=\frac12\max_{\rho_A} I(R:B\mid C),$1 correlation to prevent leakage (Wang, 24 Jun 2026).

For $Q_K(\Phi)=\frac12\max_{\rho_A} I(R:B\mid C),$2 uses of $Q_K(\Phi)=\frac12\max_{\rho_A} I(R:B\mid C),$3 achieving $Q_K(\Phi)=\frac12\max_{\rho_A} I(R:B\mid C),$4 nearly perfect ebits, the asymptotic rate is

$Q_K(\Phi)=\frac12\max_{\rho_A} I(R:B\mid C),$5

If $Q_K(\Phi)=\frac12\max_{\rho_A} I(R:B\mid C),$6 is a tripartite state, with von Neumann entropy $Q_K(\Phi)=\frac12\max_{\rho_A} I(R:B\mid C),$7, then

$Q_K(\Phi)=\frac12\max_{\rho_A} I(R:B\mid C),$8

and

$Q_K(\Phi)=\frac12\max_{\rho_A} I(R:B\mid C),$9

By strong subadditivity,

$C$0

The central theorem states that for a quantum broadcast channel $C$1,

$C$2

where $C$3 and $C$4 purifies $C$5. Equivalently, if $C$6 denotes the channel environment,

$C$7

The result directly links QCMI to a channel capacity and is presented as a quantum extension of the classical key-generation capacity of Csiszár and Ahlswede. In the purely classical limit, including diagonal states, the expression reduces to the classical $C$8 (Wang, 24 Jun 2026).

The coding theorem places QCMI in a direct operational role that had remained elusive in channel coding. Earlier operational meanings of QCMI involved redistribution, deconstruction, erasure, or secrecy tasks rather than a broadcast-channel capacity.

3. Proof architecture and the role of QCMI beyond channel coding

The achievability proof proceeds by selecting an input ensemble $C$9 that approximately maximizes $Q_K(\Phi)=\frac12\max_{\rho_A}I(R:B\mid C)$0. A packing lemma is then used to code classical labels into typical subspaces of $Q_K(\Phi)=\frac12\max_{\rho_A}I(R:B\mid C)$1 so that Bob, with access to $Q_K(\Phi)=\frac12\max_{\rho_A}I(R:B\mid C)$2, can reliably identify the label whenever the label rate is below $Q_K(\Phi)=\frac12\max_{\rho_A}I(R:B\mid C)$3. A covering lemma randomizes over the ensemble to decouple $Q_K(\Phi)=\frac12\max_{\rho_A}I(R:B\mid C)$4 from the final $Q_K(\Phi)=\frac12\max_{\rho_A}I(R:B\mid C)$5 system; this requires a randomness rate up to $Q_K(\Phi)=\frac12\max_{\rho_A}I(R:B\mid C)$6. Converting classical communication cost into quantum cost introduces at most a factor two through entanglement-assisted and superdense-coding arguments, yielding

$Q_K(\Phi)=\frac12\max_{\rho_A}I(R:B\mid C)$7

ebits per channel use. The converse uses Fannes-Alicki continuity and the data-processing inequality to show

$Q_K(\Phi)=\frac12\max_{\rho_A}I(R:B\mid C)$8

so any larger rate would either defeat reliable decoding or leave the assistant correlated with $Q_K(\Phi)=\frac12\max_{\rho_A}I(R:B\mid C)$9 (Wang, 24 Jun 2026).

QCMI already had a canonical operational interpretation in state redistribution. For a pure state $B$0, with Alice holding $B$1, Bob holding $B$2, and $B$3 purifying $B$4, the optimal forward quantum communication cost for transferring subsystem $B$5 to Bob is

$B$6

while the entanglement rate is

$B$7

This establishes QCMI as the exact rate for a conditional transfer task and furnishes the main ingredient in reconstruction-based lower bounds on QCMI (Brandao et al., 2014).

The reconstruction perspective asks how well $B$8 can be recovered from $B$9 or $Q_*=\frac12 I(C;R\mid B)$0 by a CPTP reconstruction map. The Fawzi-Renner inequality gives

$Q_*=\frac12 I(C;R\mid B)$1

and stronger bounds upper-limit the regularized relative-entropy distance and the measured relative-entropy distance to reconstructed states. In this framework, $Q_*=\frac12 I(C;R\mid B)$2 if and only if the state is a quantum Markov chain $Q_*=\frac12 I(C;R\mid B)$3, while small QCMI implies closeness to a reconstructed state in fidelity, measured relative entropy, or regularized quantum relative entropy (Brandao et al., 2014).

Taken together, these results show that QCMI governs at least three distinct operational regimes: redistribution cost, distinguishability from reconstructed states, and broadcast-channel entanglement generation. A plausible implication is that the 2026 capacity theorem converts a previously structural entropy quantity into a directly code-theoretic one.

4. Example channels and implications for code design

Two example classes illustrate how the side output $Q_*=\frac12 I(C;R\mid B)$4 can unlock correlations that are inaccessible if one looks only at $Q_*=\frac12 I(C;R\mid B)$5 (Wang, 24 Jun 2026).

For correlated bit-flip and phase-flip noise, consider a two-qubit map $Q_*=\frac12 I(C;R\mid B)$6 with Kraus operators

$Q_*=\frac12 I(C;R\mid B)$7

With maximally mixed input $Q_*=\frac12 I(C;R\mid B)$8,

$Q_*=\frac12 I(C;R\mid B)$9

Here the side channel $B$0 perfectly flags the error type—bit-flip, phase-flip, or none—and unlocks the full two bits of correlation.

A second class starts from a single-user channel $B$1 with isometry $B$2, followed by an isometric split $B$3 via $B$4. The resulting broadcast channel $B$5 obeys

$B$6

For the rank-2 qubit channel with Kraus matrices

$B$7

the gain

$B$8

was computed for four splitting options: AD + AD, Deph + Deph, Deph + AD, and AD + Deph. In the AD+AD case with parameters $B$9 and $E$0,

$E$1

and numerically $E$2 for $E$3.

The paper identifies several code-design implications. In architectures with multi-qubit gates with crosstalk, correlated noise in quantum memories, or photonic networks with environment probes, the auxiliary degrees of freedom $E$4 may carry partial information about errors on $E$5. The conditional coding framework shows how such information can be exploited to boost the logical rate relative to discarding $E$6. Code construction is described in terms of combining standard quantum error-correcting codes for the main channel with a “sensing” code for the side output $E$7, while typical-subspace methods plus classical randomness enforce decoupling of $E$8. The same capacity formula is proposed as a figure of merit for end-to-end entanglement distribution in repeater chains or multi-node networks when intermediate measurement outcomes or syndrome data are reused. Future directions explicitly mentioned include threshold analyses in hardware-informed noise models, explicit code families achieving half the QCMI, and investigations of additivity and superadditivity in this channel class (Wang, 24 Jun 2026).

5. Cryptographic conditional communication

A cryptographic realization of conditional quantum communication appears in the conditional quantum one-time pad. Alice and Bob share many copies of a tripartite state $E$9, Alice holds $I(A;B\mid E)$0, and Bob holds $I(A;B\mid E)$1, but $I(A;B\mid E)$2 is an insecure sector controlled by Eve. Alice communicates over an ideal quantum channel that Eve may intercept. The task is to encode a classical message so that Bob, with access to $I(A;B\mid E)$3, can decode it, while Eve, with access to $I(A;B\mid E)$4, learns nothing. The exact secret-communication capacity is

$I(A;B\mid E)$5

The achievability proof again uses packing and covering: $I(A;B\mid E)$6 for reliable recovery and $I(A;B\mid E)$7 for secrecy, so the net rate is

$I(A;B\mid E)$8

The same work states that the quantum-message version has rate $I(A;B\mid E)$9, and it relates the minimized quantity $C$00 to squashed entanglement (Sharma et al., 2017).

A second cryptographic branch is conditional disclosure of secrets with quantum resources. Here Alice and Bob do not communicate with one another; Alice receives $C$01, Bob receives $C$02, they may share entanglement, and both send quantum messages to a referee. Correctness requires that whenever $C$03, there exists a decoder $C$04 with

$C$05

while privacy requires that whenever $C$06, there exists a simulator $C$07 independent of the secret such that

$C$08

This model is equivalent, up to parameter losses, to $C$09-routing, where the quantum secret is routed to one party or the other depending on $C$10. The theory includes closure under negation, amplification from a single-qubit secret to $C$11 qubits with exponentially small error at linear overhead, and lower bounds from one-way quantum communication complexity, $C$12, two-message quantum interactive proofs, and honest-verifier QSZK (Asadi et al., 2024).

The comparison between classical and quantum conditional disclosure sharpens the complexity-theoretic picture. For perfectly correct CDS, a promise version of not-equals admits a quantum upper bound of $C$13 and a classical lower bound of $C$14. More generally, quantum CDS satisfies a lower bound

$C$15

and there is a poly$C$16 quantum CDS protocol for forrelation. The same line also gives an exponential separation between classical and quantum private simultaneous message passing for a partial function (Girish et al., 5 May 2025).

These cryptographic models differ from the broadcast-channel capacity problem in their objective: the central issue is not entanglement generation assisted by a helper, but controlled revelation of a secret under correctness and privacy constraints. Nevertheless, the common entropy-theoretic motif is the conditional accessibility of quantum information.

6. Heralding, conditional detection, and superposed communication order

An experimental use of conditional quantum communication appears in heralded detection after storage in a quantum memory. In the waveguide experiment on time-bin qubits, each 795 nm photon is encoded as

$C$17

and stored in a Ti:Tm:LiNbO$C$18 single-mode channel waveguide implementing an atomic-frequency-comb memory. Detection of the paired 1532 nm photon serves as a herald, so that one analyzes only those trials for which the partner photon has been detected. The conditional retrieval probability is written as $C$19, with intrinsic AFC recall efficiency $C$20. The conditional average storage fidelity is

$C$21

compared with the unconditioned

$C$22

The heralded value exceeds both the measure-and-prepare benchmark $C$23 and the universal-cloning bound $C$24, while heralding rejects $C$25 of detector dark counts (Saglamyurek et al., 2011).

A conceptually different usage identifies conditional quantum communication with coherent control of communication order. In this formulation, one replaces a definite signaling order such as $C$26 or $C$27 by a quantum superposition controlled by an auxiliary qubit. In the tripartite Hamming game with $C$28 and total communication $C$29, the best average success probabilities obey

$C$30

whereas the quantum switch achieves

$C$31

The construction is expressed in process-matrix language by a switch process $C$32 containing off-diagonal coherence terms, and operationally by routing the target through Alice then Bob or Bob then Alice depending coherently on the control qubit. Because any causally separable process is bounded by $C$33, observing $C$34 under the qubit-dimension assumption acts as a semi-device-independent witness of causal nonseparability (Feix et al., 2015).

These examples show that the adjective “conditional” ranges from heralding and side information to secrecy predicates and coherent causal control. The 2026 broadcast-channel theorem isolates the QCMI-governed variant with the sharpest channel-coding statement, while the surrounding literature demonstrates that conditional structure has become a recurring design principle across quantum Shannon theory, quantum cryptography, quantum memories, and indefinite-causal-order protocols.