New geometric constants of isosceles orthogonal type

Qi Liu, Zhijian Yang, Yongjin LiarXiv 2021 · DOI · arXiv

Abstract

This paper introduces a novel geometric constant, $\Omega(X)$, derived from the parallelogram law and the principle of isosceles orthogonality. We investigate the fundamental properties of this constant, specifically its capacity to characterize the structure of Banach spaces. Our findings demonstrate that $\Omega(X) = 1$ if and only if the norm of the space is induced by an inner product. Furthermore, we establish the relationship between $\Omega(X)$ and the property of uniform non-squareness, providing sharp bounds and relating it to established metrics such as the von Neumann-Jordan and James constants.

Introduction

In the study of functional analysis, a recurring theme is the distinction between general Banach spaces and the more structured Hilbert spaces. While all Hilbert spaces are Banach spaces, the converse is not true. The "distance" between a general normed linear space and an ideal inner product space is often measured using geometric constants. This paper proposes a new constant, $\Omega(X)$, which leverages isosceles orthogonality to provide a refined metric for this deviation.

The Concept of Isosceles Orthogonality

Orthogonality in Euclidean space is a straightforward concept. However, in a general Banach space, multiple definitions exist. This research focuses on isosceles orthogonality ($x \perp_I y$), defined by the condition: $$\|x+y\| = \|x-y\|$$ This condition implies that the diagonals of the parallelogram formed by $x$ and $y$ are equal in length. Unlike Birkhoff orthogonality, which can be non-symmetric in general Banach spaces, isosceles orthogonality is inherently symmetric, making it a robust foundation for defining new geometric constants.

Defining the Constant $\Omega(X)$

The authors define the new constant $\Omega(X)$ for a normed space $X$ as: $$\Omega(X) = \sup \left\{ \frac{\|x+2y\|^2 + \|2x+y\|^2}{5\|x+y\|^2} : x, y \in S_X, x \perp_I y \right\}$$ where $S_X$ denotes the unit sphere of the space. This specific formulation is designed to probe the "squareness" and convexity of the space by evaluating specific linear combinations of orthogonal vectors.

To analyze $\Omega(X)$, the study employs several mathematical strategies:

  1. Dvoretzky’s Theorem: Used to establish lower bounds in infinite-dimensional spaces by examining two-dimensional subspaces.
  2. Minkowski Plane Analysis: Simplifies calculations for symmetric 2D normed spaces, allowing for the derivation of equivalent forms of the constant.
  3. Inequality Chaining: Connects $\Omega(X)$ to the von Neumann-Jordan constant ($C_{NJ}(X)$) and the James constant ($J(X)$) using triangle inequalities and norm properties.

Key Insights and Characterizations

The Hilbert Space Identity

A primary result of this work is the "Hilbert Test." The authors prove that for any normed space $X$, $\Omega(X) = 1$ if and only if $X$ is an inner product space. This confirms that $\Omega(X)$ is a sensitive indicator of the underlying geometry of the space, reaching its minimum value only when the parallelogram law is perfectly satisfied.

Uniform Non-Squareness

The constant also serves as a measure of "squareness." A space is considered uniformly non-square if it avoids certain "flat" geometric configurations. The paper establishes that:

  • For any infinite-dimensional Banach space, $1 \le \Omega(X) \le 8/5$.
  • A space fails to be uniformly non-square (meaning it contains square-like regions) if and only if $\Omega(X) = 8/5$.

For example, in the space $\mathbb{R}^2$ with the $L_\infty$ norm (which geometrically represents a square), the constant reaches its maximum value of $8/5$. Conversely, for a specific mixed $\ell_\infty - \ell_1$ norm, the authors calculated a value of $49/40$, demonstrating the constant's sensitivity to different norm structures.

Generalized Constants and Relationships

The authors also introduce a generalized version, $\Omega'(X)$, which relaxes the requirement that $x$ and $y$ reside on the unit sphere. This version is shown to be directly related to the function $\gamma_X(t)$, specifically $\Omega'(X) = \frac{9}{10} \gamma_X(1/3)$. Furthermore, the research establishes the inequality: $$\Omega'(X) \le \frac{2}{5} C_{NJ}(X) + \frac{4}{5}$$ This links the new constant to the well-established von Neumann-Jordan constant, integrating it into the existing landscape of functional analysis metrics.

Conclusion

The introduction of $\Omega(X)$ provides a new lens through which to view the geometry of Banach spaces. By utilizing isosceles orthogonality, the authors have created a constant that not only identifies inner product spaces but also quantifies the degree of non-squareness within a space. While this work is primarily theoretical, it lays the groundwork for future exploration into more complex functional spaces, such as Orlicz or Sobolev spaces, and potential applications in numerical optimization where space geometry influences convergence rates.

Summary updated

Abstract

Based on the parallelogram law and isosceles orthogonality, we define a new orthogonal geometric constant. We first discuss some basic properties of this new constant. Next, we consider the relation between the constant and the uniformly non-square property. Moreover, a generalized constant is also introduced and some basic properties are presented. It is shown that, for a normed space, the constant value is equal to 1 if and only if the norm can be induced by the inner product. Finally, we verify that this constant is closely related to the well-known geometric constants through some inequalities.
BibTeX
@misc{liu2021new,
  author = {Qi Liu and Zhijian Yang and Yongjin Li},
  title = {{New geometric constants of isosceles orthogonal type}},
  year = {2021},
  eprint = {2111.08392},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2111.08392},
}

Loading related content...

Figure