Monotone Riemannian Metrics

• Monotone Riemannian metrics are a class of contractive metrics on positive-definite matrices and quantum states defined via operator monotone and convex functions.
• They are constructed using left/right multiplication operators and characterized by complete positivity under quantum stochastic maps.
• These metrics find applications in quantum information theory, statistical mechanics, and condensed matter physics for quantifying statistical distinguishability.

Monotone Riemannian metrics are a distinguished class of contractive Riemannian metrics on spaces of positive-definite matrices or quantum states, defined through operator means and fully classified via operator monotone and operator convex functions. These metrics arise naturally in quantum information geometry, statistical mechanics, and condensed matter physics as they provide a geometric quantification of statistical distinguishability that is compatible with quantum stochastic maps (completely positive trace-preserving maps). The theory is built on key results by Petz, Morozova, Chencov, and collaborators, and connects analytic, spectral, and information-theoretic approaches.

1. Definition and Characterizing Properties

Let $M_d$ denote the space of $d \times d$ matrices, $P_d\subset M_d$ the strictly positive-definite matrices, and consider the manifold of normalized density matrices $\{D\in P_d : \Tr D = 1\}$. A monotone Riemannian metric is a smoothly varying positive-definite quadratic form $g^k_D$ on the tangent spaces to this manifold such that, for every completely positive trace-preserving map $\Phi$ and tangent vector $A$ at $D$,

$g^k_{\Phi(D)}(\Phi(A), \Phi(A)) \leq g^k_D(A, A)$

This contractivity condition ensures that the metric decreases under stochastic (quantum) coarse-graining.

For each $k$ in the set

$d \times d$0

the associated metric is defined via the linear operator

$d \times d$1

where $d \times d$2 and $d \times d$3 are left and right multiplication. Setting $d \times d$4, one has

$d \times d$5

The metric is then

$d \times d$6

restricted to traceless Hermitian $d \times d$7.

2. Classification via Operator Convex and Monotone Functions

Every monotone metric corresponds bijectively to a function $d \times d$8. Moreover, $d \times d$9 is a Bauer simplex with explicit integral (Choquet) representation:

$P_d\subset M_d$0

for a probability measure $P_d\subset M_d$1 on $P_d\subset M_d$2. The extreme points are

$P_d\subset M_d$3

with $P_d\subset M_d$4 the minimal and $P_d\subset M_d$5 the maximal metric in the pointwise order (Hiai et al., 2012).

Canonical families within $P_d\subset M_d$6 include the Heinz, binomial (power), power-difference (A–L–G), Wigner–Yanase–Dyson (WYD), and Stolarsky means, each with parameterized domains where contractivity holds. For instance:

• Heinz: $P_d\subset M_d$7
• WYD: $P_d\subset M_d$8

Explicit inclusion/exclusion ranges for $P_d\subset M_d$9 (see below) are established for each family (Hiai et al., 2012).

3. Spectral, Commutator, and Dynamic Structure Representations

Monotone metrics admit several analytic forms. For a one-parameter family of Gibbs states $\{D\in P_d : \Tr D = 1\}$0 with $\{D\in P_d : \Tr D = 1\}$1:

• Spectral representation: For $\{D\in P_d : \Tr D = 1\}$2, the quadratic form is

$\{D\in P_d : \Tr D = 1\}$3

where $\{D\in P_d : \Tr D = 1\}$4 (Tonchev, 2015, Tonchev, 2021).

• Commutator expansion: Metrics can be expanded as a Kubo-type sum rule in terms of nested commutators of $\{D\in P_d : \Tr D = 1\}$5 with $\{D\in P_d : \Tr D = 1\}$6,

$\{D\in P_d : \Tr D = 1\}$7

with $\{D\in P_d : \Tr D = 1\}$8 involving $\{D\in P_d : \Tr D = 1\}$9 and $g^k_D$0 arising from Taylor expansion of deviation functions (Tonchev, 2015, Tonchev, 2021).

• Dynamical structure factor (DSF) moments: The metric admits re-expression as an infinite series in moments of the dynamic structure factor $g^k_D$1,

$g^k_D$2

where $g^k_D$3 (Tonchev, 2015, Tonchev, 2021).

The expansions clarify that each monotone metric is determined by a reference metric (BKM, MC) plus system-dependent corrections captured by DSF moments or commutator expectation values.

4. Complete Positivity, Order Structure, and Geometric Bridges

Mapping properties are central for the classification. For $g^k_D$4, the operator $g^k_D$5 is completely positive (CP) for all $g^k_D$6 if and only if $g^k_D$7 is positive-definite on $g^k_D$8. The sets $g^k_D$9 and $\Phi$0 partition $\Phi$1. The only function in $\Phi$2 for which both $\Phi$3 and its inverse are CP is $\Phi$4.

An order relation is defined by

$\Phi$5

and all classical one-parameter families in $\Phi$6 are monotone in their defining parameter with respect to this order (Hiai et al., 2012).

Interpolation and convexification are encoded in geometric bridges:

$\Phi$7

If $\Phi$8 and $\Phi$9 lie in $A$0, then so does $A$1. When their quotient is infinitely divisible, bridges preserve CP properties, providing a mechanism to interpolate between monotone metrics.

Table: Parameter domains for complete positivity in key families (Hiai et al., 2012) | Family | $A$2 parameter range | Endpoint metric | |-----------------|----------------------|---------------------| | Heinz | $A$3 | $A$4 | | Binomial | $A$5 | $A$6 | | Power-difference| $A$7 | $A$8 | | WYD | $A$9 | $D$0 | | Stolarsky | $D$1 | $D$2, $D$3 |

5. Analytical and Fourier Techniques

Verification of complete positivity and infinite divisibility employs Fourier analysis. By Bochner’s theorem, $D$4 is positive-definite if and only if it is the Fourier transform of a positive measure. For $D$5, explicit criteria for CP can be checked by considering $D$6. For more delicate assessments (e.g., boundaries of geometric bridge positivity), positivity or signedness of Fourier transforms, including asymptotics of closed-form integrals, play a crucial role (Hiai et al., 2012).

Elementary positive-definite functions (arising from hyperbolic/trigonometric quotients or combinations) underpin constructive proofs. Infinite divisibility links bridge interpolation to Lévy-Khinchin decompositions of these Fourier transforms.

6. Physical and Information-Theoretic Applications

Monotone Riemannian metrics have significant roles in quantum information theory (quantum Fisher informations), statistical mechanics, and condensed matter. Through Petz’s theorem, there is a bijection between monotone metrics and operator monotone functions, linking geometry and data-processing inequalities (Tonchev, 2015, Tonchev, 2021). In quantum statistical mechanics, expansions in DSF moments and susceptibility express the metrics in terms of linear response observables, with closed-form expressions available for physically relevant models like spins in a uniform field and single-mode bosonic oscillators.

Comparison inequalities between key metrics (Bures, BKM, MC, Haro/SLD, etc.) establish ordering relations valuable for bounding quantum information and thermodynamic fluctuations. The analytical structure allows for calculations in Lie-algebraic models and shows the direct relevance of monotone geometry in both information-theoretic and thermodynamic fluctuation analysis (Tonchev, 2015, Tonchev, 2021).

7. Broader Geometric Monotonicity and Comparison Principles

A distinct yet structurally analogous notion of monotonicity arises for Riemannian volumes with respect to boundary distance functions. In the classical, finite-dimensional Riemannian geometry context, the local volume monotonicity theorem states that—among simple Riemannian metrics—volume is locally monotone with respect to the boundary distance function, with holonomic consequences for geodesic ray transform injectivity and minimal filling problems (Ivanov, 2011). While this setting is distinct from the noncommutative geometry of quantum states, both reveal the fundamental role of monotonicity in the geometric quantification of information and physical observables.