Abstract
We consider two measures of entanglement of mixed bipartite states of dimension $2 \times 2$: concurrence and negativity. We first prove the conjecture of Eisert and Plenio that concurrence can never be smaller than negativity. We then characterise all states for which concurrence equals negativity and also those states for which the difference between concurrence and negativity is maximal (keeping either the concurrence fixed, or the participation ratio $R = 1/\text{trace}(\rho^2)$).
Introduction
Quantifying entanglement is a central challenge in quantum information theory. For the simplest bipartite system—two qubits ($2 \times 2$)—two primary measures are widely used:
- Concurrence ($C$): Derived from the entanglement of formation, representing the minimum resources required to synthesize a state.
- Negativity ($E_N$): Based on the Peres-Horodecki criterion, measuring the extent to which a state's partial transpose fails to be positive semi-definite.
While these measures coincide for pure states, they diverge for mixed states. This paper formally proves the long-standing Eisert-Plenio conjecture: for any $2 \times 2$ mixed state, concurrence is an upper bound for negativity ($C \geq E_N$).
Methods
The authors utilize a combination of analytical matrix analysis and numerical optimization:
- Weyl’s Inequality: This is the cornerstone of the proof for $C \geq E_N$, relating the eigenvalues of a sum of Hermitian matrices to the sums of their individual eigenvalues.
- Singular Value Decomposition (SVD): SVD is used to extend pure-state results to mixed states by analyzing "optimal ensembles"—the specific decomposition of pure states that minimizes the average entanglement.
- Numerical Search: To identify states with the maximal discrepancy $(C - E_N)$, the authors employed a Nelder-Mead "downhill simplex" optimization in MATLAB, exploring the full parameter space of $2 \times 2$ density matrices.
Key Insights
The discrepancy between concurrence and negativity arises because concurrence is defined via a minimization over all possible pure-state ensembles, whereas negativity is a direct spectral property of the density matrix. Weyl's inequality demonstrates that the "averaging" process in mixed states suppresses negativity more significantly than concurrence.
The authors further show that equality ($C = E_N$) occurs only when the "optimal" pure states in the ensemble have perfectly aligned internal structures (specifically, their Schmidt-vector-related eigenvectors).
Key Results
The primary achievement of this work is the formal proof that $C(\rho) \geq E_N(\rho)$ for all $2 \times 2$ states. Beyond the proof, the paper maps the maximal possible gap between these measures.
For states with a participation ratio $R$ (where $R = 1/\text{Tr}(\rho^2)$) between 2 and 3, the states maximizing the gap are specific "maximally entangled" mixed states. For $R < 2$, numerical results indicate the gap is even larger than simple analytical models suggest, revealing a complex landscape of mixed-state entanglement.
The relationship between the maximal difference and the participation ratio $R$ for rank-2 states is given by $(C - E_N)_{max} = 1 - 1/\sqrt{R}$.
Limitations and Future Work
These results are specific to $2 \times 2$ systems. In higher dimensions (e.g., $3 \times 3$), the relationship is more complex; negativity can be zero for entangled states (bound entanglement), a phenomenon absent in two-qubit systems. Additionally, while the equality conditions are fully characterized, finding analytical solutions for the absolute maximum difference across all levels of mixedness remains a challenge, currently necessitating numerical methods.
Summary updated