PaperlayerView exact revision on arXiv

Source-linked AI summary

The decoupling approach to quantum information theory

Frédéric Dupuis

arXiv:1004.1641v1quant-ph

TL;DR

The thesis frames questions about characterizing memoryless-channel capacities and the asymptotic rate for storing qubits, and develops a locking scheme as a special case of a broader model. Its locking discussion reports a trace-distance criterion, described as strictly stronger than accessible information.

  • Problem

    The thesis notes that memoryless-channel capacities lack a satisfactory characterization and asks for the best asymptotic rate for storing qubits.

    Source: §2.5.2 · p. 35, §4.1 · p. 68

  • Method

    For the locking scheme, the thesis constructs a special case of the model under consideration.

    Source: LOCKING CLASSICAL INFORMATION IN… · p. 103

  • Results

    The main improvement over previous locking schemes is a trace-distance criterion on measurement outputs rather than accessible information, described as strictly stronger.

    Source: §6.3 · p. 113

  • Takeaways & Limitations

    In the discussed quantum-adversary setting, polylog(n) key bits precede reconstruction of the entire key, rather than n−1 bits.

    Source: §6.3 · p. 114

    Whether a converse theorem can be demonstrated without allowing mixed states remains an open problem.

    Source: §5.6 · p. 100

Abstract

from arXiv · show

Quantum information theory studies the fundamental limits that physical laws impose on information processing tasks such as data compression and data transmission on noisy channels. This thesis presents general techniques that allow one to solve many fundamental problems of quantum information theory in a unified framework. The central theorem of this thesis proves the existence of a protocol that transmits quantum data that is partially known to the receiver through a single use of an arbitrary noisy quantum channel. In addition to the intrinsic interest of this problem, this theorem has as immediate corollaries several central theorems of quantum information theory. The following chapters use this theorem to prove the existence of new protocols for two other types of quantum channels, namely quantum broadcast channels and quantum channels with side information at the transmitter. These protocols also involve sending quantum information partially known by the receiver with a single use of the channel, and have as corollaries entanglement-assisted and unassisted asymptotic coding theorems. The entanglement-assisted asymptotic versions can, in both cases, be considered as quantum versions of the best coding theorems known for the classical versions of these problems. The last chapter deals with a purely quantum phenomenon called locking. We demonstrate that it is possible to encode a classical message into a quantum state such that, by removing a subsystem of logarithmic size with respect to its total size, no measurement can have significant correlations with the message. The message is therefore "locked" by a logarithmic-size key. This thesis presents the first locking protocol for which the success criterion is that the trace distance between the joint distribution of the message and the measurement result and the product of their marginals be sufficiently small.

RÉSUMÉ

The thesis develops general techniques for addressing fundamental quantum information problems in a unified framework. It presents a single-use transmission theorem, protocols for additional channel types, and a locking scheme with a trace-distance success criterion.

RÉSUMÉ

  • The thesis aims to address several fundamental quantum information problems using general techniques in a unified framework.Source: RÉSUMÉ · p. 3
  • The central theorem establishes a protocol for transmitting quantum data partially known to the receiver through one use of a noisy quantum channel.The theorem has several central quantum information theorems as immediate corollaries.Source: RÉSUMÉ · p. 3
  • For broadcast channels and channels with transmitter side information, new protocols yield entanglement-assisted and unassisted asymptotic coding theorems.The entanglement-assisted versions are quantum versions of the best known coding theorems for the corresponding classical problems.Source: RÉSUMÉ · p. 3
  • Removing a logarithmic-size subsystem can leave every measurement without significant correlation with the encoded classical message, locking it with a logarithmic-size key.The key size is logarithmic relative to the total state size.Source: RÉSUMÉ · p. 3
  • The thesis presents the first locking protocol whose success criterion is a sufficiently small trace distance between the joint message–measurement distribution and the product of their marginals.Source: RÉSUMÉ · p. 3

CHAPTER 4: QUANTUM CHANNELS WITH SIDE INFORMATION

The excerpted headings identify a chapter on locking classical information and a subsection on security implications for quantum protocols.

CHAPTER 4: QUANTUM CHANNELS WITH SIDE INFORMATION

CHAPTER 7: CONCLUSION

The excerpted conclusion includes open problems and future research directions.

CHAPTER 7: CONCLUSION

NOTATION

Together, these parts describe the thesis’s use of decoupling as a framework for deriving quantum-channel coding protocols and information-locking results, including a one-shot coding theorem.

General / First names

  • The notation uses log for base-2 logarithms, c∗ for complex conjugation, and E_U[f(U)] for expectation over U; Alice and Bob are sender and receiver.The sender and receiver labels apply across all protocols.Source: General · p. 9, First names · p. 11

Linear Algebra and Quantum Systems

  • A, B, C, . . . label quantum systems or operators and their associated Hilbert spaces; |A| is the dimension of A, and AB has space A ⊗B.A^n denotes a quantum system composed of n copies of A.Source: Linear Algebra and Quantum Systems · p. 9
  • L(A, B) is the space of linear operators from A to B; M† is the adjoint, T is the transpose in canonical bases, and M:A→B indicates membership in L(A, B).The notation also defines Herm(A) and Pos(A) for Hermitian and positive-semidefinite operators.Source: Linear Algebra and Quantum Systems · p. 9

Linear Algebra and Quantum Systems, continued

  • D(A) denotes density operators, N and T denote completely positive linear maps, I_A denotes the identity operator or superoperator, and vec(M) turns an operator into a vector.The vectorization notation refers to Section 2.6.Source: Linear Algebra and Quantum Systems,… · p. 10

Norms and Distance measures / Entropies

  • The notation includes the diamond norm and fidelity, and uses conditional min-entropy, conditional max-entropy, mutual information, and conditional mutual information.The cited definitions are in Sections 2.3.1, 2.10, 2.12, 2.6, and 2.7, respectively.Source: Norms and Distance measures · p. 10, Entropies · p. 11

INTRODUCTION

  • The thesis addresses coding for different quantum channels and information locking, where a small key can unlock information.Quantum information theory studies the physical limits on information processing under quantum mechanics.Source: INTRODUCTION · pp. 13–14

1.1 Decoupling

  • Decoupling means decorrelating the input from everything except the channel output, which can be achieved by destroying correlations.The thesis motivates this condition by noting that a desired correlation between channel input and output is equivalent to decorrelation from the rest of the universe.Source: §1.1 · p. 14
  • The thesis gives a general, more flexible decoupling theorem from which known theorems can be derived, and develops coding theorems for broadcast channels and channels with transmitter side information.The thesis presents decoupling as a unified approach to channel coding problems that had previously been proved separately.Source: §1.1 · pp. 14–15
  • The thesis reports no prior results for the particular tasks involving information at the transmitter and uses the main decoupling theorem to prove an information-locking result.Source: §1.1 · p. 15

1.2 Contributions

  • Chapter 3 states and proves the main decoupling theorem and variants, including a one-shot quantum-channel coding theorem where Bob may know part of the state beforehand.Source: §1.2 · p. 15
  • Chapter 3 rederives the main results of,, and in a more straightforward manner.Source: §1.2 · p. 15

Chapter 4 (Quantum channels with side information at the transmitter):

This chapter develops the quantum-information measures and decoupling tools used to study one-shot quantum transmission over channels with transmitter side information. It derives one-shot coding conditions and connects them to asymptotic and entanglement-assisted capacity results.

Chapter 4 (Quantum channels with side information at the transmitter):

  • The chapter gives a one-shot coding theorem and derives an optimal entanglement-assisted protocol for memoryless channels, with a single-letter capacity formula paralleling the classical solution.The transmitter has side information about the channel noise; the receiver does not.Source: Chapter 4 (Quantum channels with side… · p. 16

PRELIMINARIES

  • The preliminaries introduce notation and concepts needed to understand the thesis.Source: PRELIMINARIES · p. 18

2.1 Notation

  • The notation subsection establishes linear-algebra conventions for quantum information, including Hilbert spaces, operators, and superoperators.Source: §2.1 · pp. 18–19

2.2 Quantum mechanics: an extremely short introduction

  • The quantum mechanics introduction defines states as density operators and describes pure states and composite systems.Source: §2.2 · pp. 19–20
  • Quantum channels are framed as uncontrolled trace-preserving transformations, while measurements can be described by POVMs when only their classical outcomes matter.The text also distinguishes measurement superoperators that preserve a quantum residue from POVMs that describe only measurement results.Source: §2.2 · pp. 20–21

2.3 Distance measures

  • Trace distance characterizes how well measurements distinguish quantum states, while fidelity approaches one as states become closer.Fidelity is also stable under purifications, with the fidelity between states equal to the maximum fidelity between their purifications.Source: §2.3 · pp. 22, 24
  • Fuchs–van de Graaf inequalities relate fidelity distance to trace distance, and the diamond norm characterizes optimal distinguishing probability for two quantum channels.The diamond norm maximizes over mixed states with an ancilla isomorphic to the input; it was introduced to quantum information theory by Kitaev.Source: §2.3 · p. 24, §2.3.1 · p. 25

2.4.1 von Neumann entropy and derived quantities

  • The von Neumann entropy measures information; Schumacher showed that an i.i.d. state ρA⊗n can be compressed into n[H(A)ρ − δ] qubits with error tending to zero as n→∞ for any δ > 0.Source: §2.4.1 · pp. 25–26
  • Quantum mutual information gives the entanglement-assisted classical capacity of memoryless quantum channels, while coherent information gives the best known general unassisted quantum-transmission rate.Conditional mutual information is nonnegative and is used in a technical proof in Chapter 4.Source: §2.4.1 · p. 27

2.4.2 Properties of the von Neumann entropy

  • For mixed states, adding subsystem C cannot decrease the mutual information between A and B, reflecting strong subadditivity.The passage describes this monotonicity as a consequence of strong subadditivity.Source: §2.4.2 · p. 28

2.4.3 One-shot information measures

  • One-shot measures address protocols using a single use of an arbitrary channel; min-entropy is the negative logarithm of the largest eigenvalue.Conditional min-entropy quantifies extractable private randomness and key requirements for encryption; conditional 2-entropy is another quantity used in proofs.Source: §2.4.3 · pp. 28–29
  • One-shot entropies can be sensitive to small state variations, so smooth versions optimize over an ε-ball; these quantities reduce to von Neumann quantities in i.i.d. cases.The i.i.d. reduction is identified with the fully quantum asymptotic equipartition property.Source: §2.4.3 · pp. 31–32

2.5 Quantum channel capacities

  • The thesis considers unassisted and entanglement-assisted quantum capacities, first for one use of an arbitrary channel and then for i.i.d. channels.Source: §2.5 · pp. 32–33

2.5.1 One-shot capacities

  • One-shot quantum transmission asks Alice to send an arbitrary message system through one channel use so Bob can recover it within a tolerated error, including when it is entangled with a reference.Source: §2.5.1 · p. 33
  • A fixed maximally entangled input provides an easier, weaker test than optimizing over every input; slightly reducing the channel-input dimension can upgrade a protocol meeting that test.Source: §2.5.1 · p. 34

2.5.2 Capacities of memoryless channels

  • The coherent-information lower bound is not tight, and channels N and M can each have zero capacity while N ⊗ M has positive capacity [SY08].Source: §2.5.2 · pp. 35–36

2.5.3 Regularization and single-letter converses

  • For any pure input-reference state σ, every rate R < I(A⟩C)ρ, with ρ=N(σ), is achievable for quantum transmission through N.The theorem’s protocol rate is computed from a state produced by one application of N, although the protocol uses the channel repeatedly.Source: §2.5.3 · p. 37
  • The quantum capacity is characterized by a regularized converse over arbitrary positive numbers of channel uses.The passage describes this multiletter characterization as uncomputable and says it does not reveal the structure of capacity-achieving codes.Source: §2.5.3 · p. 38
  • The regularized characterization is uncomputable because no bound is known on how many channel copies approximate capacity, and it reveals nothing about capacity-achieving code structure.Source: §2.5.3 · p. 38
  • Finding easily computable capacity expressions that reveal optimal-code structure is difficult in quantum information theory, and basic unassisted quantum and classical capacities remain central open problems.The passages describe finding such expressions as one of information theory’s main goals.Source: §2.5.3 · p. 39

2.6 The duality between vectors and operators

  • The thesis uses a basis-dependent op transformation to turn multipartite states into operators and its inverse vec transformation to turn operators into vectors.The basis choice is implicit and does not matter in the thesis.Source: §2.6 · pp. 39, 42

THE DECOUPLING THEOREM

  • A general decoupling framework connects coding to making the reference system and environment nearly product, enabling decoder existence and flexible generalizations of FQSW and state merging.The chapter presents a generalization of both protocols for use in more diversified contexts.Source: THE DECOUPLING THEOREM · pp. 44–45

3.1 Uhlmann’s theorem

  • Uhlmann’s theorem provides a decoder when the environment is nearly uncorrelated with the reference system purifying the transmitted qubits.The theorem applies when the relevant states are purifications, even if their purifying systems are not isomorphic.Source: §3.1 · pp. 45–46

3.2 The decoupling theorem

  • A Haar-random unitary followed by an arbitrary completely positive superoperator yields good decoupling on average, generalizing known decoupling theorems.The superoperator may implement a complex operation, beyond tracing out part of the system or the measurement used in state merging.Source: §3.2 · p. 47
  • The decoupling theorem directly shows that the environment is decoupled from systems of interest.The theorem uses smooth conditional 2-entropies rather than purities and dimension bounds.Source: §3.2 · pp. 47, 50
  • The smoothed-entropy version permits applying the theorem directly to i.i.d. states and channels.It replaces the H2 upper bound with smoothed versions and enables use of the fully quantum AEP.Source: §3.2 · p. 54
  • With very high probability, the trace-distance quantity in Theorem 3.7 is close to its expected value.The probability is over the choice of unitary U.Source: §3.2 · pp. 55–56
  • A unitary 2-design can replace Haar-random unitaries in the results above, except for Theorem 3.9.The text notes that Haar-random choices may be impractical to implement efficiently as quantum circuits.Source: §3.2 · pp. 56–57
  • For all theorems in this section except Theorem 3.9, Haar measure in integrals of this type can be replaced by a unitary 2-design without affecting the rest of the theorem; for Theorem 3.9, this replacement is not yet known.Source: §3.2 · p. 57

3.3 Corollaries of the decoupling theorem

  • The decoupling theorem yields the Fully Quantum Slepian-Wolf theorem [ADHW06], state merging [HOW07], and a random-subspaces corollary.The passage also introduces a protocol blending FQSW and random subspaces, though its precise guarantee is not visible in the supplied text.Source: §3.3 · pp. 57–59

3.4 Quantum coding theorems via decoupling

  • Theorem 3.14 guarantees an encoding and decoding protocol for sending Alice’s share through one use of any channel when Bob already holds a share.The proof strategy makes the complementary channel break correlations with the purifying reference, enabling Bob to reconstruct the message.Source: §3.4 · pp. 59–60
  • Theorem 3.15 establishes achievable rate pairs when Q + E < H(A)_σ and Q − E < I(A⟩C)_N(σ).These conditions apply to quantum transmission through any channel N, for a pure state σ_AA′ with A′ isomorphic to A.Source: §3.4 · p. 64
  • The unassisted rate is I(A⟩C) qubits per transmission, boosted by one ebit per qubit until Q+E reaches H(A)σ; saturating that constraint yields the fully entanglement-assisted rate 2Q < I(A; C)N(σ).Source: §3.4 · pp. 64–65
  • For any Q and E satisfying Q+E < H(A)σ and Q−E < I(A⟩C)N(σ), a protocol exists whose error vanishes as n →∞.Source: §3.4 · p. 65

3.5 Destroying correlations by adding classical randomness

  • Using Theorem 3.14, the chapter recovers ’s decoupling result and establishes a one-shot version for arbitrary bipartite states using a set of 2^k unitaries.Source: §3.5 · pp. 66–67

QUANTUM CHANNELS WITH SIDE INFORMATION AT THE TRANSMITTER

This section develops coding results for quantum channels with transmitter-side channel-state information, including one-shot and memoryless transmission theorems and entanglement-assisted rate and capacity characterizations.

4.1 Introduction

  • The entanglement-assisted quantum capacity has the same form as the classical side-information expression (4.1) and admits a single-letter converse.Source: §4.1 · pp. 68–69

4.2 Definition of quantum channels with side information at the transmitter

  • The model gives the transmitter access to side information about the channel state, representing both quantum and classical side information in a unified framework.Source: §4.2 · p. 69
  • For entanglement-assisted coding over i.i.d. channels, the preceding theorem’s i.i.d. input distribution is optimal.Source: §4.2 · p. 70, §4.4 · p. 76

4.3 Direct coding theorem

  • The one-shot theorem gives a coding guarantee for arbitrary channels with transmitter side information, with its error bound controlled by a smooth conditional-entropy difference and an additive 24ε term.Source: §4.3 · p. 71
  • For memoryless channels, rate points meeting the theorem’s conditions are achievable with rate-limited entanglement, and ebits trade for qubits one-for-one until the first condition binds.Source: §4.3 · pp. 74–75
  • Achievable rates satisfy Q + E < H(A|S)_σ − Δ1 and Q − E < I(A⟩C)_N(σ) − Δ2, with both deltas vanishing as n grows.The first condition limits the message qubits and Alice’s entanglement share to fit in the input; the second limits the transmission rate minus entanglement by coherent information.Source: §4.3 · pp. 75–76
  • Adding the rate conditions to eliminate E gives the entanglement-assisted bound 2Q < H(A|S)_σ + I(A⟩C)_N(σ).The passage states that a simple calculation shows this bound is equivalent to a further expression, but that expression is incomplete in the supplied text.Source: §4.3 · p. 76

4.4 Optimality for entanglement-assisted coding

The entanglement-assisted quantum capacity with transmitter side information is characterized by a single-letter expression, and the section proves that the achievable rate is optimal. Its converse adapts the argument of to quantum states.

4.4 Optimality for entanglement-assisted coding

  • For any achievable rate Q, the converse bounds Q by a single-letter term I(A;C)_N(σ) − I(A;S)_σ for some state σ.The proof considers arbitrary block-size-n encoders and decoders, then uses a direct coding theorem to show the largest term in its sum is achievable.Source: §4.4 · pp. 76–78
  • The capacity is the supremum, over mixed σ_AA′S with σ_S = φ_S, of 1/2 I(A;C)_N(σ) − 1/2 I(A;S)_σ.The section states that this expression is achievable and proves no higher rate is possible.Source: §4.4 · p. 76
  • The converse follows the approach of, with quantum adaptations that track which state defines each mutual information.The proof considers states in which only a fraction of the n channel instances have been applied.Source: §4.4 · pp. 76, 78
  • Summing the inequalities across channel uses yields the converse bound; the key comparison has nonnegative difference I(C_i−1;C_i)_ω(i).This difference follows from mutual-information identities and nonnegativity.Source: §4.4 · pp. 78–79

4.5 Discussion

The discussion highlights unresolved bounds needed to compute a capacity and shows that optimal inputs need not be pure. The broadcast-channel chapter develops one-shot and i.i.d. protocols for two receivers, with distinct rate-region behavior depending on entanglement assistance.

4.5 Discussion

  • The discussion draws a parallel between entanglement-assisted quantum capacities and classical coding theory, and identifies systematic conversion of classical coding theorems as an open interest.The capacity formula is described as having the same form as its classical version.Source: §4.5 · p. 79
  • The capacity formula lacks an upper bound on the dimension of A, so the single-letter converse does not provide a way to compute capacity.The discussion leaves this dimension-bound problem open.Source: §4.5 · p. 80
  • A qubit-to-qubit channel applying one of four Pauli operations equally has capacity one qubit per transmission, yet its optimal input cannot be pure when the transmitter’s side information identifies the operation.If σ were pure, purifying the four-dimensional S would require that dimension in the two-dimensional A′, which is impossible.Source: §4.5 · p. 80

5.1 Introduction

  • The broadcast-channel task is to transfer Alice’s separate quantum shares to Bob 1 and Bob 2; the chapter gives a one-shot theorem and specializes it to i.i.d. channels.The i.i.d. setting uses repeated channel N and entangled pairs between Alice and a reference and each Bob.Source: §5.1 · pp. 81–82
  • With maximum entanglement assistance, the broadcast-channel rate region recovers a quantum version of Marton’s region; without assistance, the rates appear to have no independent sum-rate constraint.The unassisted information to the two Bobs appears to “talk past each other.”Source: §5.1 · p. 83
  • The chapter also proves a regularized converse for the fully entanglement-assisted broadcast-channel case.Source: §5.1 · p. 83

5.2 Definitions

This section defines quantum broadcast coding setups and presents one-shot and asymptotic transmission guarantees. It also describes coding regions obtained through receiver-role switching and time-sharing, including a fully entanglement-assisted quantum version of Marton’s region with the constraint Q1 + Q2 < 1.

5.2 Definitions

  • The definitions model a quantum broadcast channel as a CPTP map with multiple output subsystems and specify separate receiver data, entanglement, and transmission rates.The one-shot setup assigns outputs to two receivers; the i.i.d. setup considers repeated channel use for arbitrary quantum data.Source: §5.2 · pp. 83–84

5.3 Direct coding theorem

  • Theorem 5.2 establishes a one-use protocol for sending two quantum inputs to separate broadcast-channel receivers using an encoder and parallel decoders.The purified protocol’s decoder outputs are close to the initial state.Source: §5.3 · pp. 85–86, Figure 5.1 · p. 86
  • Theorem 5.3 gives achievable rate conditions for quantum transmission through a broadcast channel with rate-limited entanglement assistance.In the fully entanglement-assisted case, it yields a quantum version of Marton’s region.Source: §5.3 · pp. 91–92
  • In the i.i.d. coding theorem, the condition Q2 + E2 < H(A2) is included among the stated achievable-rate constraints.The theorem uses the fully quantum AEP to obtain vanishing error as the number of channel uses grows.Source: §5.3 · pp. 91–92
  • Switching Bob 1 and Bob 2 yields the other rate region, and convex combinations of the regions correspond to time-sharing between protocols.Source: §5.3 · p. 93
  • The fully entanglement-assisted region has the constraint Q1 + Q2 < 1.Source: §5.3 · p. 93

5.4 Regularized converse

For rate-limited quantum transmission over a broadcast channel, the capacity region is characterized by regularizing the achievable rate points over many channel uses. The converse proof places any achievable rate tuple inside the stated region, while the authors caution that this is a weak characterization.

5.4 Regularized converse

  • The characterization is only weak: distinct formulas for rate regions may agree after regularization.Source: §5.4 · p. 94
  • The capacity region for rate-limited quantum transmission over a broadcast channel is the convex hull of the union of rate points satisfying the theorem’s conditions.The characterization regularizes over many uses of the channel.Source: §5.4 · pp. 93–94
  • The converse argues that any achievable rate tuple lies inside the region in Equation (5.14), establishing the stated capacity characterization.The proof uses bounds from recoverability, continuity, and subadditivity.Source: §5.4 · pp. 94–95
  • The single-letter conjecture is false even for unassisted quantum transmission through a channel with one receiver.Source: §5.4 · p. 95

5.5 Single-letter example

For classical deterministic broadcast channels, the entanglement-assisted capacity region matches the achievable region of Theorem 5.3. This provides an example where the region has a single-letter converse and is optimal.

5.5 Single-letter example

  • The example addresses the unusual availability of a single-letter converse for a quantum information-theory rate region.The authors note that entanglement is unlikely to help classical transmission through a classical channel, but this still gives an example where their theorem is optimal.Source: §5.5 · p. 96
  • Every achievable rate point for the example channel is a convex combination of rates achieved by the coding method.The channel has deterministic outputs specified by functions of its input.Source: §5.5 · p. 96
  • For classical deterministic broadcast channels, the entanglement-assisted capacity region equals the achievable rate region of Theorem 5.3.The proof bounds the individual and sum rates and concludes that the region is optimal for these channels.Source: §5.5 · pp. 96–97

5.6 Discussion

The discussion presents a quantum broadcast communication protocol and relates its entanglement-assisted rate region to classical coding results. It also identifies open questions about converses, optimality, and the unassisted region.

5.6 Discussion

  • The protocol gives a rate region for quantum communication through broadcast channels with rate-limited entanglement assistance.The region is achieved for every mixed input state.Source: §5.6 · pp. 98–99
  • The entanglement-assisted broadcast region is formally similar to Marton’s classical broadcast region, with factors of 1/2.For classical channels, teleportation can directly achieve the entanglement-assisted rates found here.Source: §5.6 · p. 99
  • The lack of a single-letter converse for Marton’s region and the quantum region leaves open whether a better broadcast region could break their analogy.A better characterization of the quantum regions is suggested as a step toward resolving this uncertainty.Source: §5.6 · p. 99
  • The regularized converse has not been proved without allowing mixed states, leaving open whether a converse can avoid that requirement.The discarded system D in Theorem 5.3 is equivalent to optimizing over mixed rather than only pure states.Source: §5.6 · p. 100
  • The unassisted region has no independent sum-rate constraint, yet the region is already suboptimal for channels with a single receiver.Whether this also holds for the true capacity region remains open.Source: §5.6 · p. 100

LOCKING CLASSICAL INFORMATION IN QUANTUM STATES

This section develops a quantum information-locking scheme, using quasi-measurements to establish a unitary family that limits what arbitrary measurements can reveal about the message under a trace-distance secrecy criterion.

LOCKING CLASSICAL INFORMATION IN QUANTUM STATES

  • Locking is measured by the trace distance between the message–measurement-result joint distribution and the product of their marginals, rather than by accessible information.The definition also requires the encoded states to be perfectly distinguishable when both ciphertext and key are available, while measurements on the ciphertext alone reveal practically no information.Source: LOCKING CLASSICAL INFORMATION IN… · p. 102
  • Classically, a key must be nearly as long as the message to keep the message distribution close to uniform given the ciphertext.The passage explains that if only log K additional bits are needed to reconstruct the message, the conditional distribution is supported on at most K messages and remains far from uniform unless K is nearly the total number of messages.Source: LOCKING CLASSICAL INFORMATION IN… · p. 103

6.1 The locking scheme

  • The scheme embeds equiprobable classical messages into a joint ciphertext–key system of dimension at least N, with the key itself a quantum state.The proof seeks a unitary that makes the resulting states a good locking scheme for every measurement on the ciphertext.Source: §6.1 · pp. 103–104
  • Theorem 6.2 establishes that a unitary exists for which every ciphertext measurement satisfies the locking criterion.The proof considers Haar-random unitaries, uses a concentration bound for quasi-measurements, and applies a union bound to obtain one unitary that works across the required measurements.Source: §6.1 · pp. 104, 106
  • The operator Chernoff bound ensures that independent measurements on n copies yield an (n, k)-quasi-measurement with high probability.Lemma 6.5 bounds the probability that the copies fail to form such a quasi-measurement; the supplied passage omits the full bound.Source: §6.1 · p. 107
  • Theorem 6.6 establishes a unitary U for which the stated property holds for all measurement operators.The supplied theorem statement is truncated, so its exact bound is not recoverable here.Source: §6.1 · p. 108
  • The proof combines the Lemma 6.4 bound with Lemma 6.5’s failure-probability estimate, then chooses constants to satisfy both constraints.It sets k = 2 and r = ε before deriving the stated parameter conditions.Source: §6.1 · pp. 108–109

6.2 Implications for the security of quantum protocols against quantum adversaries

The section explains how locking can undermine a quantum key-distribution security definition based on accessible information. It describes a protocol that preserves low accessible information while allowing an adversary to recover the key after learning a small part of it.

6.2 Implications for the security of quantum protocols against quantum adversaries

  • A security definition based on Eve’s accessible information can be inadequate when she retains quantum data and later measures it using partial knowledge of the secret.The cited work gives an example where learning the first n−1 bits lets Eve recover the final bit by measuring her quantum register.Source: §6.2 · p. 110
  • The locking scheme yields a QKD protocol with negligible accessible information about the final key, yet Eve can recover the entire key after learning a very small fraction.The section describes this as a failure of the security definition based on accessible information.Source: §6.2 · p. 110

6.2.1 Description of the protocol

  • The construction starts from a truly secure protocol, splits its n-bit key into a log-size key and the remaining string, then sends a locked version of the latter to Eve.The small key is used to create the quantum state containing the locked string.Source: §6.2.1 · pp. 110–111
  • If Eve learns the small key, she can recover the remaining string, even though immediately after the protocol no measurement reveals information about the key.The theorem considers measurements on Eve’s quantum registers, including the register containing the locked string.Source: §6.2.1 · p. 111
  • The construction shows that low accessible information alone is not an adequate security definition for quantum key distribution.The protocol combines low accessible information with insecurity due to locking effects.Source: §6.2.1 · p. 113

6.3 Discussion

This discussion presents a stronger operational definition of locking and explains why locked quantum information matters for cryptography. A quantum adversary may lack classical information from measurements while holding quantum information that a small amount of additional information can unlock.

6.3 Discussion

  • A quantum adversary may hold much quantum information about a string despite having no classical information from measurements.The discussion connects this possibility to the need for cryptographic security definitions that account for quantum information.Source: §6.3 · p. 113
  • The locking phenomenon allows a very small amount of additional information to unlock quantum information about a classical message.Source: §6.3 · p. 113
  • The new locking definition uses trace distance between measurement outputs rather than accessible information, a strictly stronger criterion.Source: §6.3 · p. 113
  • Under this criterion, measurements on locked messages are negligibly distinguishable from outcomes generated independently of the message.Source: §6.3 · p. 113
  • Compared with, the adversary here needs only polylog(n) key bits to reconstruct the key, rather than n −1.Source: §6.3 · pp. 113–114

CONCLUSION

The conclusion presents decoupling as a general tool for quantum information problems, then applies it to channel coding and information locking. It also identifies open questions about constructivity, optimality, and extensions of the theory.

CONCLUSION

  • A general decoupling theorem applies a unitary to part of a state before a channel, with decorrelation quality governed by two parameters.The thesis presents versions of the theorem adapted to different settings.Source: CONCLUSION · p. 115
  • The framework yields achievable quantum transmission rates and entanglement-assisted capacities, plus new coding theorems for broadcast and transmitter-side-information channels.The conclusion also identifies the entanglement-assisted capacities for classical and quantum transmission.Source: CONCLUSION · p. 115
  • The coding proofs first establish single-use results depending on min- and max-entropies, then specialize them to repeated uses of a channel.For memoryless channels, the fully quantum asymptotic equipartition property bounds smooth min-entropies and yields von Neumann entropies.Source: CONCLUSION · pp. 115–116
  • In locking, a large ciphertext and small key encode a classical message so ciphertext-only measurements reveal no significant information, while both systems together provide full information.The conclusion describes the key as very small relative to the ciphertext.Source: CONCLUSION · p. 116
  • The locking criterion bounds trace distance between measurement outcomes with and without message dependence, making a measured locked message virtually indistinguishable from a random state.The conclusion calls this condition stronger than previous locking definitions.Source: CONCLUSION · p. 116

7.1 Open problems and future research directions

  • Haar-random proofs establish that protocols exist but do not directly construct them; the thesis identifies a constructive locking scheme as an open problem.The conclusion notes that most results can use unitary 2-designs, but identifies the locking scheme as an exception.Source: §7.1 · pp. 116–117

3.9 (or perhaps something slightly more general) would most likely achieve this

  • The one-shot coding theorems lack converses, although the thesis considers them likely optimal for some input distributions.Some special cases, including state merging, had already been shown optimal.Source: §3.9 · p. 117
  • A general classical–quantum principle remains unsettled, and broader smooth min-entropy bounds could extend the results to channels with memory.The conclusion notes similar capacity expressions in classical and quantum cases, while saying the principle’s scope and definitions remain unclear.Source: §3.9 · pp. 117–118

Appendix I

Appendix I gathers technical lemmas and inequalities used throughout the thesis, including operator bounds, probabilistic estimates, continuity inequalities, and finite nets.

Appendix I

  • Appendix I collects technical lemmas and inequalities used at multiple points throughout the thesis.Its contents include operator inequalities and tools such as Markov’s inequality, the operator Chernoff bound, Fannes-type inequalities, and ε-nets.Source: Appendix I · pp. 128–129
  • A lemma constructs a positive operator on B bounded by the identity whose partial expectation against a bipartite pure state recovers a specified operator on A.The stated condition is that the operator on A is bounded above by the pure state's marginal on A.Source: Appendix I · p. 129
  • The appendix uses Markov’s inequality to establish a unitary satisfying multiple conditions at once.It also introduces an operator Chernoff bound and Fannes-type inequalities among the technical tools.Source: Appendix I · pp. 131–132
Loading 1004.1641v1…