Boundary effect of deterministic dense coding

Zhengfeng Ji, Yuan Feng, Runyao Duan, Mingsheng YingPhysical 2006 · DOI · arXiv

Abstract

In quantum communication, deterministic dense coding allows for the transmission of classical information using shared entanglement and a single quantum carrier. While a maximally entangled pair of $d$-level systems (qudits) can transmit an alphabet of size $d^2$, it has been numerically observed that even a slight deviation from maximal entanglement reduces this capacity significantly. This paper provides a rigorous mathematical proof of this "boundary effect," demonstrating that $d^2 - 1$ is an impossible alphabet size for any isometric deterministic dense coding scheme utilizing $d$-level partial entanglement.

Introduction

Quantum dense coding is a fundamental protocol in quantum information theory, demonstrating how entanglement can be leveraged to exceed classical communication limits. In the standard scenario, Alice and Bob share a maximally entangled state of two $d$-level systems. By performing one of $d^2$ local unitary operations on her qudit and sending it to Bob, Alice can communicate $\log_2(d^2)$ bits of information—twice the capacity of a classical channel using the same particle.

However, the robustness of this protocol under non-ideal conditions—specifically, partial entanglement—presents a more complex picture. Previous research by Mozes, Oppenheim, and Reznik (2005) utilized numerical methods to suggest a strange "gap" in the capacity of these channels. They observed that as soon as the shared state $|\psi\rangle$ deviates from maximal entanglement, the maximum number of reliably transmittable messages ($N_{max}$) does not merely decrease to $d^2 - 1$, but drops immediately to $d^2 - 2$ or lower. This paper formalizes this observation, proving that $d^2 - 1$ is a forbidden alphabet size for isometric encoding.

The Geometry of Isometric Encoding

The protocol assumes Alice and Bob share a pure state $|\psi\rangle \in \mathcal{H}_A \otimes \mathcal{H}_B$, where $\text{dim}(\mathcal{H}_A) = \text{dim}(\mathcal{H}_B) = d$. Alice encodes a message $i \in \{1, \dots, N\}$ by applying a local unitary (or more generally, an isometry) $U_i$ to her qudit. The resulting global states are:

$$|\psi_i\rangle = (U_i \otimes I) |\psi\rangle$$

For the coding to be deterministic and error-free, Bob must be able to distinguish these states perfectly, which requires them to be mutually orthogonal:

$$\langle \psi_i | \psi_j \rangle = \delta_{ij}$$

Proof Methodology

The proof relies on analyzing the constraints that the orthogonality condition imposes on the set of operators $\{U_i\}$ when the shared state is not maximally entangled.

1. Schmidt Decomposition and Reduced Density Matrices

The shared state $|\psi\rangle$ is represented via Schmidt decomposition to analyze the distribution of entanglement coefficients ($\lambda_k$). The reduced density matrix for Alice's system is $\rho_A = \text{Tr}_B(|\psi\rangle\langle\psi|)$. For a maximally entangled state, $\rho_A$ is proportional to the identity matrix. For partially entangled states, the eigenvalues are not all equal, breaking the system's symmetry.

2. Orthogonality and Trace Constraints

The condition $\langle \psi_i | \psi_j \rangle = \delta_{ij}$ translates to a requirement on the operators: $\text{Tr}(U_j^\dagger U_i \rho_A) = \delta_{ij}$. This implies that the operators $U_i$ must be orthonormal with respect to the Hilbert-Schmidt inner product weighted by the density matrix $\rho_A$.

3. The Dimensionality Contradiction

The core of the proof involves examining the subspace of operators that can satisfy these trace conditions. By utilizing the concavity properties of von Neumann entropy, the authors show that if $d^2 - 1$ messages were possible, the encoding operators would have to satisfy extremely restrictive commutation relations with $\rho_A$. The proof demonstrates that the number of required orthogonal operators ($d^2 - 1$) exceeds the maximum possible dimension of the subspace allowed by those relations, leading to a logical contradiction for any state that is not maximally entangled.

Key Insights

The All-or-Nothing Gap

Unlike many physical phenomena where performance degrades linearly with the quality of resources, dense coding exhibits a discrete "jump." The transition from $d^2$ to $d^2 - 2$ suggests that the perfect symmetry of maximal entanglement is a prerequisite for high-density encoding; once that symmetry is lost, the capacity cost is higher than one might intuitively expect.

Symmetry and Rigidity

Maximal entanglement provides a high degree of unitary symmetry. Partial entanglement breaks this symmetry, imposing a rigid geometric structure on the Hilbert space. The proof reveals that this "rigidity" is so restrictive that it is impossible to "fit" $d^2 - 1$ orthogonal states into the space once the perfect symmetry of the identity matrix is lost.

Conclusion

This work provides a definitive theoretical boundary for quantum communication. It confirms that for any $d$-level system that is not maximally entangled, the maximum alphabet size for isometric deterministic dense coding is at most $d^2 - 2$.

While this proof covers isometric (unitary) encoding, it remains an open question whether more general quantum operations—such as non-unitary channels or POVMs—could potentially reach the $d^2 - 1$ threshold. Additionally, understanding how this discrete gap manifests in real-world, noisy quantum hardware remains a vital area for future exploration.

Summary updated

Abstract

We present a rigorous proof of an interesting boundary effect of deterministic dense coding first observed by S. Mozes, J. Oppenheim, and B. Reznik [Phys. Rev. A 71, 012311 (2005)]. Namely, it is shown that ${d}^{2}\ensuremath{-}1$ cannot be the maximal alphabet size of any isometric deterministic dense coding schemes utilizing $d$-level partial entanglement.
BibTeX
@inproceedings{ji2006boundary,
  author = {Zhengfeng Ji and Yuan Feng and Runyao Duan and Mingsheng Ying},
  title = {{Boundary effect of deterministic dense coding}},
  booktitle = {Physical},
  year = {2006},
  doi = {10.1103/physreva.73.034307},
}

Loading related content...

Figure