Two Hornich-Hlawka-type and Gram matrix-based inequalities

Nizar El Idrissi, Hicham Zoubeir · 2026

Read Two Hornich-Hlawka-type and Gram matrix-based inequalities by Nizar El Idrissi, Hicham Zoubeir (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-18768.txt.

Introduction

Nizar El Idrissi and Hicham Zoubeir prove two new claims in the maths of gaps and lengths in spaces with a dot rule. One claim ties the four sides of a four-point form: the long side can't beat the sum set by the rest in a set way. Claims of that kind are old and prized, since they bind how far points can sit from each other. The next claim uses a table of dot scores for a set of points.

That table, called a Gram table, holds the cos-like scores of all pairs. From it, the pair draw fresh bounds on joint gaps. Such bounds feed work in shape maths, in codes, and in the tests that prove a set of points must act in a set way. Pure claims like these are tools that next work can lift whole.

Catalog record

CREATORNizar El Idrissi, Hicham Zoubeir
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE18 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-18768

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Two Hornich-Hlawka-type and Gram matrix-based inequalities Nizar El Idrissi, Hicham Zoubeir Abstract This paper deals with Hornich-Hlawka-type and Gram matrix-based inequalities in inner product spaces (IPS). On the one hand, the Hornich-Hlawka inequality, or Hlawka’s inequality, or quadrilateral inequality, is an inequality concerning three vectors in a quadrilateral. Namely, it states that for all x , y , z x,y,z , ∥ x ∥ + ∥ y ∥ + ∥ z ∥ + ∥ x + y + z ∥ ≥ ∥ x + y ∥ + ∥ x + z ∥ + ∥ y + z ∥ \lVert x\rVert+\lVert y\rVert+\lVert z\rVert+\lVert x+y+z\rVert\geq\lVert x+y\rVert+\lVert x+z\rVert+\lVert y+z\rVert . On the other hand, inequalities based on Gram matrices are very important because they generate generically all inequalities implying three vectors in an IPS. In the first part of our paper, we introduce a stronger version of Hornich-Hlawka inequality, namely, for all vectors x , y , z x,y,z , ‖ x ‖ ​ ‖ y ‖ + ‖ z ‖ | x + y + z | ≥ ‖ x + z ‖ ​ ‖ y + z ‖ \|x\|\|y\|+\|z\|\|x+y+z\|\geq\|x+z\|\|y+z\| . In the second part of our paper, we establish a sharp inequality based on Gram matrices that implies, in particular, the standard Cauchy-Schwarz inequality. 0 0 footnotetext: 2020 Mathematics Subject Classification. 46C05; 26D15 0 0 footnotetext: Key words and phrases. Hornich-Hlawka inequality; Gram matrix; Cauchy-Schwarz inequality; inner product spaces; norm inequalities 1 Introduction On the one hand, the celebrated Hornich-Hlawka inequality [ 3 , 4 , 2 , 5 , 6 , 7 , 8 , 9 , 10 , 12 , 13 ] in normed vector spaces E states \forall x,y,z\in E:\lVert x\rVert+\lVert y\rVert+\lVert z\rVert+\lVert x+y+z\rVert\geq\lVert x+y\rVert+\lVert x+z\rVert+\lVert y+z\rVert While it has been proved that this inequality fails to hold in some normed vector spaces, it does hold, nonetheless, in a large class of normed vector spaces including inner product spaces. On the other hand, important inequalities can be established from the nonnegativity of Gram matrices. Actually, all inequalities implying three vectors in an IPS can be generically derived from the nonnegativity of the Gram matrices of these vectors. After this introduction, our article is divided into 2 other sections. In section 2 , we prove a stronger form of Hornich-Hlawka inequality in inner product spaces, namely: for all vectors x,y,z , \|x\|\|y\|+\|z\|\|x+y+z\|\geq\|x+z\|\|y+z\| . In section 3 , we prove a sharp inequality based on Gram matrices that strengthens, in particular, the Cauchy-Schwarz inequality. Hereafter, H denotes a real inner product space, with scalar product \langle\cdot,\cdot\rangle and norm \lVert\cdot\rVert . 2 A stronger Hornich-Hlawka inequality Theorem 2.1 (Strong Hornich–Hlawka inequality) . For all x,y,z\in H , we have \lVert x\rVert\lVert y\rVert+\lVert z\rVert\lVert x+y+z\rVert\geq\lVert x+z\rVert\lVert y+z\rVert, Moreover, equality holds if and only if x,y,z are linearly dependent and their dependence coefficients satisfy one of the equivalent conditions listed in proposition 2.3 Remark 2.2 . • When H=\mathbb{R} or \mathbb{C} , the inequality can be easily proved by expanding both sides and using the multiplicativity of the modulus (this property is senseless when we consider vector norms instead of the modulus of complex numbers). • This stronger Hornich-Hlawka inequality implies the classical Hornich-Hlawka inequality. Indeed, since both sides of the classical Hornich-Hlawka inequality are nonnegative, we may replace them with their squares. For the left-hand side, we obtain \bigl(\|x\|+\|y\|+\|z\|+\|x+y+z\|\bigr)^{2}=2\bigl(\|x\|^{2}+\|y\|^{2}+\|z\|^{2}\bigr)+2\bigl(\langle x,y\rangle+\langle y,z\rangle+\langle z,x\rangle\bigr) +2\bigl(\|x\|+\|y\|+\|z\|\bigr)\|x+y+z\|+2\bigl(\lVert x\rVert\lVert y\rVert+\lVert x\rVert\lVert z\rVert+\lVert y\rVert\lVert z\rVert\bigr). For the right-hand side, we have \bigl(\|x+y\|+\|y+z\|+\|z+x\|\bigr)^{2}=2\bigl(\|x\|^{2}+\|y\|^{2}+\|z\|^{2}\bigr)+2\bigl(\langle x,y\rangle+\langle y,z\rangle+\langle z,x\rangle\bigr) +2\bigl(\|x+y\|\cdot\|y+z\|+\|y+z\|\cdot\|z+x\|+\|z+x\|\cdot\|x+y\|\bigr). Thus, the classical Hornich-Hlawka inequalityis is equivalent to \bigl(\|x\|+\|y\|+\|z\|\bigr)\|x+y+z\|+\bigl(\lVert x\rVert\lVert y\rVert+\lVert x\rVert\lVert z\rVert+\lVert y\rVert\lVert z\rVert\bigr). \geq\|x+y\|\cdot\|y+z\|+\|y+z\|\cdot\|z+x\|+\|z+x\|\cdot\|x+y\| which, after regrouping, is equivalent to \bigl(\lVert x\rVert\lVert y\rVert+\|z\|\cdot\|x+y+z\|-\|y+z\|\cdot\|z+x\|\bigr)+ \bigl(\lVert y\rVert\lVert z\rVert+\|x\|\cdot\|x+y+z\|-\|z+x\|\cdot\|x+y\|\bigr)+ \bigl(\lVert x\rVert\lVert z+\|y\|\cdot\|x+y+z\|-\|x+y\|\cdot\|y+z\|\bigr)\geq 0. Therefore, it is clear that the stronger Hornich-Hlawka inequality implies the classical Hornich-Hlawka inequality. Proof. Idea: we reduce the inequality to a quartic form in the Gram matrix entries and show that its minimum over the PSD cone is attained on the boundary. • First step. Reduction. Denote by L and R the left and right hand sides of this inequality, respectively. We have \displaystyle L^{2} \displaystyle=\lVert x\rVert^{2}\lVert y\rVert^{2}+\lVert z\rVert^{2}\left(\lVert x\rVert^{2}+\lVert y\rVert^{2}+\lVert z\rVert^{2}+2\langle x,y\rangle+2\langle x,z\rangle+2\langle y,z\rangle\right)+2\lVert x\rVert\lVert y\rVert\lVert z\rVert\lVert x+y+z\rVert \displaystyle=\lVert x\rVert^{2}\lVert y\rVert^{2}+\lVert z\rVert^{2}\lVert x\rVert^{2}+\lVert z\rVert^{2}\lVert y\rVert^{2}+\lVert z\rVert^{4}+2\lVert z\rVert^{2}\langle x,y\rangle+2\lVert z\rVert^{2}\langle x,z\rangle+2\lVert z\rVert^{2}\langle y,z\rangle \displaystyle\quad\quad+2\lVert x\rVert\lVert y\rVert\lVert z\rVert\lVert x+y+z\rVert and \displaystyle R^{2} \displaystyle=\left(\lVert x\rVert^{2}+\lVert z\rVert^{2}+2\langle x,z\rangle\right)\left(\lVert y\rVert^{2}+\lVert z\rVert^{2}+2\langle y,z\rangle\right) \displaystyle=\lVert x\rVert^{2}\lVert y\rVert^{2}+\lVert z\rVert^{2}\lVert x\rVert^{2}+\lVert z\rVert^{2}\lVert y\rVert^{2}+\lVert z\rVert^{4}+2\langle x,z\rangle\lVert y\rVert^{2}+2\langle x,z\rangle\lVert z\rVert^{2} \displaystyle\quad\quad+2\langle y,z\rangle\lVert x\rVert^{2}+2\langle y,z\rangle\lVert z\rVert^{2}+4\langle x,z\rangle\langle y,z\rangle We have \displaystyle L\geq R \displaystyle\Leftrightarrow L^{2}\geq R^{2} \displaystyle\Leftrightarrow\frac{L^{2}-\left(\lVert x\rVert^{2}\lVert y\rVert^{2}+\lVert z\rVert^{2}\lVert x\rVert^{2}+\lVert z\rVert^{2}\lVert y\rVert^{2}+\lVert z\rVert^{4}\right)}{2}\geq\frac{R^{2}-\left(\lVert x\rVert^{2}\lVert y\rVert^{2}+\lVert z\rVert^{2}\lVert x\rVert^{2}+\lVert z\rVert^{2}\lVert y\rVert^{2}+\lVert z\rVert^{4}\right)}{2} \displaystyle\Leftrightarrow\lVert z\rVert^{2}\langle x,y\rangle+\lVert z\rVert^{2}\langle x,z\rangle+\lVert z\rVert^{2}\langle y,z\rangle+\lVert x\rVert\lVert y\rVert\lVert z\rVert\lVert x+y+z\rVert\geq \displaystyle\quad\quad\quad\langle x,z\rangle\lVert y\rVert^{2}+\langle x,z\rangle\lVert z\rVert^{2}+\langle y,z\rangle\lVert x\rVert^{2}+\langle y,z\rangle\lVert z\rVert^{2}+2\langle x,z\rangle\langle y,z\rangle \displaystyle\Leftrightarrow\lVert z\rVert^{2}\langle x,y\rangle+\lVert x\rVert\lVert y\rVert\lVert z\rVert\lVert x+y+z\rVert\geq\lVert x\rVert^{2}\langle y,z\rangle+\lVert y\rVert^{2}\langle x,z\rangle+2\langle x,z\rangle\langle y,z\rangle This inequality is equivalent to \Leftrightarrow\lVert x\rVert\lVert y\rVert\lVert z\rVert\lVert x+y+z\rVert\geq\lVert x\rVert^{2}\langle y,z\rangle+\lVert y\rVert^{2}\langle x,z\rangle-\lVert z\rVert^{2}\langle x,y\rangle+2\langle x,z\rangle\langle y,z\rangle Denote by L and R the left and right hand sides of this last inequality, respectively. If \textbf{R}\leq 0 , the inequality is trivial, so we can suppose that \textbf{R}\geq 0 . Set a=\lVert x\rVert,\quad b=\lVert y\rVert,\quad c=\lVert z\rVert,\quad p=\langle x,y\rangle,\quad q=\langle x,z\rangle,\quad r=\langle y,z\rangle. We have \displaystyle\textbf{L}^{2} \displaystyle=\lVert x\rVert^{2}\lVert y\rVert^{2}\lVert z\rVert^{2}\left(\lVert x\rVert^{2}+\lVert y\rVert^{2}+\lVert z\rVert^{2}+2\langle x,y\rangle+2\langle x,z\rangle+2\langle y,z\rangle\right) \displaystyle=a^{2}b^{2}c^{2}(a^{2}+b^{2}+c^{2}+2p+2q+2r). and \displaystyle\textbf{R}^{2} \displaystyle=\lVert x\rVert^{4}\langle y,z\rangle^{2}+\lVert y\rVert^{4}\langle x,z\rangle^{2}+\lVert z\rVert^{4}\langle x,y\rangle^{2}+4\langle x,z\rangle^{2}\langle y,z\rangle^{2} \displaystyle\quad+2\lVert x\rVert^{2}\lVert y\rVert^{2}\langle x,z\rangle\langle y,z\rangle-2\lVert x\rVert^{2}\lVert z\rVert^{2}\langle x,y\rangle\langle y,z\rangle-2\lVert y\rVert^{2}\lVert z\rVert^{2}\langle x,y\rangle\langle x,z\rangle \displaystyle\quad+4\lVert x\rVert^{2}\langle x,z\rangle\langle y,z\rangle^{2}+4\lVert y\rVert^{2}\langle y,z\rangle\langle x,z\rangle^{2}-4\lVert z\rVert^{2}\langle x,y\rangle\langle x,z\rangle\langle y,z\rangle \displaystyle=a^{4}r^{2}+b^{4}q^{2}+c^{4}p^{2}+4q^{2}r^{2}+2a^{2}b^{2}qr-2a^{2}c^{2}pr-2b^{2}c^{2}pq+4a^{2}qr^{2}+4b^{2}rq^{2}-4c^{2}pqr Consider \displaystyle\xi \displaystyle=\textbf{L}^{2}-\textbf{R}^{2} \displaystyle=a^{2}b^{2}c^{2}(a^{2}+b^{2}+c^{2}+2p+2q+2r) \displaystyle\quad-\left(a^{4}r^{2}+b^{4}q^{2}+c^{4}p^{2}+4q^{2}r^{2}+2a^{2}b^{2}qr-2a^{2}c^{2}pr-2b^{2}c^{2}pq+4a^{2}qr^{2}+4b^{2}rq^{2}-4c^{2}pqr\right) • Second step. The minimum of \xi over the PSD cone is attained on the boundary. Consider \xi as a quadratic function in one variable, say p , for fixed a,b,c,q,r (satisfying the relevant conditions, see next). The coefficient of p^{2} is -c^{4}\leq 0 , so it is a downward-opening parabola. The set of allowed values a,b,c\geq 0 , p,q,r are such that the Gram matrix G=\begin{pmatrix}a^{2}&p&q\\ p&b^{2}&r\\ q&r&c^{2}\end{pmatrix} is positive semidefinite. This requires all principal minors to be nonnegative: |p|\leq ab , |q|\leq ac , |r|\leq bc , and \det G\geq 0 . Since the allowed range for p is the interval where \det G\geq 0 and |p|\leq ab , which is closed and bounded, the minimum of \xi over this interval occurs at one of the endpoints. The endpoints correspond to boundaries where either \det G=0 or p=\pm ab (but note that p=\pm ab implies the 2x2 minor for x,y is zero, which implies x,y are linearly dependent, which forces \det G=0 ). Thus, the minima occur on the boundary where G is singular ( \det G=0 ). To show \xi\geq 0 on this boundary, assume linear dependence: – If z=\lambda x+\mu y , then substitute \displaystyle c^{2} \displaystyle=\lambda^{2}a^{2}+\mu^{2}b^{2}+2\lambda\mu p \displaystyle q \displaystyle=\lambda a^{2}+\mu p \displaystyle r \displaystyle=\lambda p+\mu b^{2} After substitution and simplification, \xi factors as \xi=(a^{2}b^{2}-p^{2})\left(a^{2}\lambda(\lambda+1)-b^{2}\mu(\mu+1)\right)^{2}. The first factor is nonnegative by the Cauchy-Schwarz inequality, and the second factor is a square, hence nonnegative. Therefore, \xi\geq 0 . – If x=\lambda y+\mu z , then substitute \displaystyle a^{2} \displaystyle=\lambda^{2}b^{2}+\mu^{2}c^{2}+2\lambda\mu r, \displaystyle p \displaystyle=\lambda b^{2}+\mu r, \displaystyle q \displaystyle=\lambda r+\mu c^{2}. After substitution and simplification, \xi factors as \xi=\Bigl(b^{2}c^{2}-r^{2}\Bigr)\Bigl(b^{2}\lambda(\lambda+1)+c^{2}\mu(\mu+1)+2\lambda(\mu+1)r\Bigr)^{2}. The first factor is nonnegative by the Cauchy-Schwarz inequality, and the second factor is

Read the complete text