Hlawka–Popoviciu inequalities on positive definite tensors

Wolfgang Berndt, Suvrit SraLinear 2015 · DOI · arXiv

Abstract

This paper establishes a rigorous mathematical bridge between classical scalar inequalities and operator theory. The authors extend the Hlawka and Popoviciu inequalities—originally formulated for vector norms and convex functions—to the domain of positive definite operators. By leveraging tensor products and the Löwner partial order, the study proves a new suite of multivariable inequalities that resolve several existing conjectures regarding determinants and permanents.

Introduction

In classical geometry and analysis, Hlawka’s inequality relates the norms of sums of vectors, while Popoviciu’s inequality characterizes the behavior of convex functions. While these relationships are well-established for real numbers and vectors, extending them to operators (matrices) is a non-trivial challenge.

Unlike real numbers, matrix multiplication is non-commutative ($AB \neq BA$), and matrices do not follow a total order. To address this, the authors utilize the Löwner partial order, where $A \succeq B$ signifies that the matrix $A - B$ is positive semi-definite. This paper investigates whether these foundational inequalities remain valid when scalar values are replaced by Hermitian positive definite operators.

Methodological Framework

The research employs several sophisticated techniques from multilinear algebra to navigate the complexities of operator theory:

  • Tensor Power Notation: To simplify complex proofs, the authors treat $p$-fold tensor products ($A \otimes A \otimes \dots \otimes A$) as powers ($A^p$). This notation allows for the manipulation of tensor expressions using algebraic intuition analogous to scalar mathematics.
  • Löwner Partial Ordering: All inequalities are established within the framework of the Löwner order, the standard metric for comparing operator magnitudes in quantum mechanics and optimization.
  • Double Induction: For multivariable cases involving $n$ distinct matrices, a robust double induction strategy is used to generalize from 3-operator cases to $n$-operator sums.
  • Symmetry Classes: The authors utilize representation theory and the properties of the Grassmann exterior product to derive results for specific functions, such as determinants and permanents.

Key Insights and Theoretical Contributions

The Role of Superadditivity

A fundamental driver of the proofs is the property of "strong superadditivity" in tensor products: $(A+B)^p \succeq A^p + B^p$. The authors demonstrate that this property is the primary mechanism behind Hlawka-type behavior in higher-dimensional operator spaces.

Structural Symmetry

The paper identifies a deep structural link between Hlawka’s and Popoviciu’s inequalities. By proving the operator version of one, the authors are frequently able to derive the other as a direct corollary, suggesting a unified underlying logic for these operator relations.

Resolving Determinantal Conjectures

The research formally proves a conjecture by Berndt regarding elementary symmetric determinantal polynomials ($s_n + s_{n-2} + \dots \succeq s_{n-1} + s_{n-3} + \dots$). This confirms that such patterns are inherent to the structure of positive definite tensors.

Divergence from Scalar Analogues

An important finding is that while the operator version of the multivariable inequality holds, a direct "convex function" analogue for scalars actually fails for $n=4$. This highlights that operator inequalities are not merely "scaled up" versions of scalar ones; they possess unique properties that do not always translate back to the scalar domain.

Principal Results

  1. Theorem 2.1 (The 3-Operator Hlawka): The authors prove that for positive definite operators $A, B, C$, the inequality $(A+B+C)^p + A^p + B^p + C^p \succeq (A+B)^p + (A+C)^p + (B+C)^p$ holds for all integers $p \ge 1$.
  2. Theorem 3.3 (Multivariable Generalization): This result extends the 3-operator case to $n$ operators using symmetric tensor sums, providing a generalized framework for matrix analysis.
  3. Matrix Function Corollaries: The authors successfully recover several recently published inequalities for determinants and "G-immanants" (generalized matrix functions) as specific special cases of their broader tensor theorems.

Limitations and Future Research

While the paper provides a significant advancement in operator theory, several avenues for future work remain:

  • Real-Valued Powers: The primary results are currently limited to integer powers $p$. Determining if these hold for real-valued $p > 1$ (similar to the Loewner-Heinz theorem) remains an open question.
  • The $n=4$ Gap: The failure of the scalar convex analogue for $n=4$ suggests a "missing link" in our understanding of when scalar convex inequalities can be perfectly translated to the operator setting.
  • Block Matrices: Preliminary findings suggest a "completely positive" version for block matrices, which may serve as a fruitful area for future research in matrix analysis.

Summary updated

BibTeX
@inproceedings{berndt2015hlawkapopoviciu,
  author = {Wolfgang Berndt and Suvrit Sra},
  title = {{Hlawka–Popoviciu inequalities on positive definite tensors}},
  booktitle = {Linear},
  year = {2015},
  doi = {10.1016/j.laa.2015.08.028},
}

Loading related content...

Figure