Spectral geometric mean and trace characterizations

Abstract: We use nearly parallel pure states to characterize positive linear functionals $φ$ on $\mathbb{M}_n$ as positive multiples of the trace if and only if $φ(A \natural B) \leq \sqrt{φ(A) φ(B)}$ for all positive definite matrices $A$ and $B$. Here $A \natural B = (A^{-1} # B)^{1/2} A (A^{-1} # B)^{1/2}$ represents the spectral geometric mean. For further clarification, we establish novel characterizations through the inequality $φ(A \natural B) \leq φ((A+B)/2)$ for all positive definite matrices $A$ and $B$. We also present a trace inequality related to quantum fidelity that applies to all positive definite matrices, and demonstrate that it does not characterize the trace.

• The paper shows that spectral geometric mean inequalities uniquely characterize the trace functional if and only if the associated operator is a scalar multiple of the identity.
• It employs rank-one projection tests and perturbative estimates to reveal that non-scalar positive linear functionals fail to satisfy the derived inequalities.
• The study bridges operator algebras and quantum information by linking spectral means with quantum fidelity metrics and the Bures–Wasserstein distance.

Spectral Geometric Mean and Trace Characterizations: Technical Summary

Introduction and Background

This paper addresses the characterization of trace functionals on matrix algebras via inequalities involving the spectral geometric mean, with significant connections to operator theory and quantum information theory. The trace functional on $\mathbb{M}_n$—a normalized, positive linear functional invariant under cyclic permutations and unitary conjugation—provides a central example in both classical and quantum frameworks, particularly as the maximally mixed state in quantum mechanics.

Prior work characterizes trace via various operator inequalities involving the metric geometric mean $(A^{1/2}BA^{1/2})^{1/2}$, with connections to quantum divergences such as the Bures--Wasserstein distance. The present investigation extends this direction by introducing new characterizations that hinge on the spectral geometric mean and associated trace inequalities.

Spectral Geometric Mean

The spectral geometric mean $A \natural B$ (Fiedler--Pták) for $A, B \in \mathbb{P}_n$ is defined by

$$A \natural B = \left(A^{-1}\# B\right)^{1/2} A \left(A^{-1}\# B\right)^{1/2},$$

where $\#$ denotes the Kubo--Ando metric geometric mean. This mean coincides with the metric mean if and only if $A$ and $B$ commute, providing a genuinely noncommutative construction essential for operator inequalities beyond the scalar or commuting matrix case.

Trace Characterization via Inequalities

Characterization via Spectral Geometric Mean

The paper establishes that, for a positive linear functional $\phi(X) = \operatorname{Tr}(SX)$ with $S \in \mathbb{P}_n$, the following inequality: $(A^{1/2}BA^{1/2})^{1/2}$0 for all $(A^{1/2}BA^{1/2})^{1/2}$1, holds if and only if $(A^{1/2}BA^{1/2})^{1/2}$2 is a scalar multiple of the identity, i.e., $(A^{1/2}BA^{1/2})^{1/2}$3 is proportional to the trace. The proof involves rank-one projection tests (reducing to pure state scenarios) and asymptotic arguments where non-scalar $(A^{1/2}BA^{1/2})^{1/2}$4 leads to constructive violations using nearly parallel pure states (quantum counterparts of nearly indistinguishable states), exploiting continuity and regularization techniques.

An analogous, slightly weaker characterization is given by the inequality: $(A^{1/2}BA^{1/2})^{1/2}$5 which also characterizes the trace among positive linear functionals. The technical analysis uses two-variable perturbations and reductions to $(A^{1/2}BA^{1/2})^{1/2}$6 subspaces, ultimately relying on spectral properties precluding non-scalar $(A^{1/2}BA^{1/2})^{1/2}$7 from satisfying these inequalities for all positive definite matrices.

Further Two-Point Strengthenings

A functional version of the Kadison inequality,

$(A^{1/2}BA^{1/2})^{1/2}$8

for all $(A^{1/2}BA^{1/2})^{1/2}$9, also characterizes scalar multiples of the trace. This is shown by detailed spectral expansions, contradiction arguments for non-scalar $A \natural B$0 in low-dimensional settings, and perturbative estimates, demonstrating that the condition is essentially a two-variable strengthening of Kadison's famous convexity inequality.

Failure of Quantum Fidelity-Based Characterizations

The paper also considers inequalities inspired by quantum fidelity—namely,

$A \natural B$1

for all $A \natural B$2. It is shown via explicit construction and limiting cases that such an inequality, when $A \natural B$3 ranges over positive linear functionals normalized to be a state, cannot characterize the trace: it is trivially violated except in trivial or degenerate cases (e.g., rank-one projections), and thus does not provide a characterization akin to those for the spectral geometric mean.

Conversely, if the normalization is set as $A \natural B$4 (un-normalized but with full trace), then the fidelity-inspired inequality does characterize the trace, as any $A \natural B$5 not equal to $A \natural B$6 will fail the condition on suitable test vectors.

Implications and Outlook

These results consolidate the spectral geometric mean as a robust noncommutative mean for characterizing the trace functional via operator inequalities, thus grounding an abstract algebraic property (traciality) in concrete analytic inequalities. The necessity of the trace as the unique functional satisfying these inequalities among all positive functionals aligns with the role of the maximally mixed state and unitary invariance in quantum theory.

Numerical implications include the observation that rank-one and nearly rank-one test projections are sufficient to test failure of these inequalities for non-tracial functionals, meaning that quantum information theoretic witness states (noisy or nearly pure) are mathematically optimal probes in these contexts.

Theoretical significance arises for operator algebras, matrix analysis, and quantum information, suggesting that spectral geometric means and their trace inequalities encode deep structural symmetries not enforceable by more elementary or one-variable inequalities.

Future directions may include:

• Generalizing these characterizations to broader classes of operator algebras (e.g., infinite-dimensional),
• Exploring other operator means for similar characterizations,
• Investigating operational interpretations in quantum hypothesis testing and quantum channel discrimination, leveraging the fidelity and Bures--Wasserstein connections,
• Extending to monotone or convex functional frameworks beyond positive definite matrix cones.

Conclusion

The paper rigorously demonstrates that certain spectral geometric mean inequalities tightly characterize the canonical trace among positive linear functionals on matrix algebras, with comprehensive counterexamples showing the failure of related quantum fidelity inequalities as characterizations in the abstract. The bridge built between geometric means, trace functionals, and quantum states provides a foundation for further theoretical developments in matrix analysis and quantum information.