[Paper Review] Some observations on the simplex

Igor Rivin|ArXiv.org|Aug 26, 2003

Point processes and geometric inequalities3 references3 citations

TL;DR

This paper investigates the geometric and algebraic structure of simplices in Euclidean space using linear algebra, particularly the Gram matrix and its dual. It proves that the set of Gram matrices of non-degenerate simplices forms a convex cone, and establishes a duality between face areas and the null space of the dual Gram matrix, providing a complete characterization of simplex configurations via symmetric positive-definite matrices and their adjugates.

ABSTRACT

We investigate the space of simplices in Euclidean Space

Motivation & Objective

• To understand the conditions under which a set of edge lengths can form a non-degenerate simplex in $\mathbb{E}^n$, extending the triangle inequality to higher dimensions.
• To analyze the geometric constraints on simplices beyond triangle inequalities, showing that these are insufficient for higher-dimensional non-degeneracy.
• To characterize the space of simplices using the Gram matrix and its dual, establishing a duality between face areas and the null space of the dual Gram matrix.
• To demonstrate that the set of valid Gram matrices forms a convex cone, providing a linear-algebraic framework for studying simplex configurations.

Proposed method

• Uses the Gram matrix $G = V^tV$, where $V$ is the matrix of vertex vectors from the origin, to represent the inner products of edge vectors.
• Applies the Rayleigh-Ritz characterization to show that the smallest eigenvalue of a symmetric matrix is a concave function, leading to convexity of the positive-definite cone.
• Employs Cramer’s rule and the adjugate matrix to relate the null space of a singular matrix to the outer product of its left and right null vectors.
• Introduces the dual Gram matrix $G^*$, whose entries are the cosines of exterior dihedral angles between faces, and proves it is symmetric and positive semi-definite with a one-dimensional null space.
• Uses the divergence theorem to show that the vector of face areas lies in the null space of the dual Gram matrix.
• Applies the adjugate theorem to derive a ratio formula for squared face areas in terms of principal minors of the dual Gram matrix.

Experimental results

Research questions

• RQ1 What conditions on edge lengths are necessary and sufficient for the existence of a non-degenerate simplex in $\mathbb{E}^n$?
• RQ2 Why do triangle inequalities fail to guarantee the existence of a simplex in dimensions $n > 2$?
• RQ3 How can the space of all simplices in $\mathbb{E}^n$ be parametrized algebraically, and is this space convex?
• RQ4 What is the geometric meaning of the null space of the dual Gram matrix, and how is it related to face areas?
• RQ5 Can the duality between the Gram matrix and its adjugate be used to derive identities involving face areas and dihedral angles?

Key findings

• The set of Gram matrices of non-degenerate simplices in $\mathbb{E}^n$ forms a convex cone, as the set of symmetric positive-definite matrices is convex.
• The dual Gram matrix $G^*$ of a simplex is symmetric and positive semi-definite with exactly one zero eigenvalue, and its null space is spanned by the vector of face areas.
• The face area vector $\mathbf{A} = (A_0, \dots, A_n)$ satisfies $G^* \mathbf{A} = 0$, which follows from the divergence theorem applied to the polyhedral simplex.
• The ratio of squared face areas is given by $\frac{A_i^2}{A_j^2} = \frac{\widehat{M}_{ii}}{\widehat{M}_{jj}}$, where $M$ is the dual Gram matrix and $\widehat{M}$ is its adjugate.
• For a singular matrix $M$ of nullity 1 with left and right null vectors $w$ and $v$, the adjugate satisfies $\widehat{M} = c w \otimes v$, where $c$ is the product of non-zero eigenvalues divided by $\langle v, w \rangle$.
• The paper establishes that the space of simplices is not convex when parametrized by edge lengths, as shown by a counterexample with two simplices whose average edge lengths do not form a valid simplex.