TOPICS
Search

Frobenius Radial Convexity


Frobenius radial convexity of order n is a property of a real function f on a interval J: for every ordered triple A<=C<=B of n×n Hermitian matrices with eigenvalues in J,

 ||f(A)-f(C)||_F<=||f(A)-f(B)||_F, ||f(C)-f(B)||_F<=||f(A)-f(B)||_F.

Here ||·||_F is the Frobenius norm, A<=B means that B-A is positive semidefinite, and the matrix functions are defined by unitary eigen decomposition (Czerwińska et al. 2026). Every increasing or decreasing matrix monotone function of order n has this property.

Kian (2026, Theorem 3.1) proved that for every integer m>=2 there is epsilon_m>0 such that for 0<epsilon<epsilon_m the rational function

 f_(m,epsilon)(t)=sum_(j=1)^m(t+1)/(jt+j+2)-epsilon(t+1)/((m+1)t+m+3), t>0

is matrix monotone of order m but not m+1, and Frobenius radially convex of order 2m-1 but not 2m. Consequently the gap between the two exact orders is unbounded. For every n>=3, Frobenius radial convexity of order n fails to imply matrix monotonicity of that order in either direction.


See also

Frobenius Norm, Hermitian Matrix, Matrix Monotone Function

Explore with Wolfram|Alpha

References

Czerwińska, M. M.; Franco, J. A.; and Merino, A. "Towards a Characterization of Matrix Monotonicity via Radial Convexity." 12 Oct 2026. https://arxiv.org/abs/2610.17393v1.Kian, M. "Unbounded Gaps Between Matrix Monotonicity and Frobenius Radial Convexity." 19 Nov 2026. https://arxiv.org/abs/2611.25596v1.

Cite this as:

Weisstein, Eric W. "Frobenius Radial Convexity." From MathWorld--A Wolfram Resource. https://mathworld.wolfram.com/FrobeniusRadialConvexity.html

Subject classifications