Convex Hull Condition

• Convex hull condition is a criterion that defines when a point, function, or set lies within a convex or generalized convex hull by ensuring every relevant subspace intersects the prescribed set.
• It underpins a range of applications from numerical analysis and PDE theory to variational methods and optimization, guiding algorithms for membership tests and solution invariance.
• The condition facilitates exact relaxations in nonconvex optimization and supports efficient geometric computations, offering actionable insights across theoretical and applied mathematics.

A convex hull condition is a necessary-and-sufficient criterion for determining when a point, set, or function lies within the convex hull (or an appropriate generalization) of a prescribed set. Such conditions encode both geometric and analytic membership, underpinning fundamental theorems in convex geometry, optimization, numerical analysis, and geometric computation. These conditions arise in a diverse set of frameworks, including generalized convexities, non-Euclidean metric spaces, motion planning, variational principles, and algebraic-geometric representations.

1. Generalized Convex Hull and the m-Hull Condition

The classical convex hull $\operatorname{conv}(E)$ of a set $E\subset\mathbb{R}^n$ is the minimal convex set containing $E$. This notion admits a generalization parametrized by an integer $0\leq m\leq n-1$ into the concept of $m$-convexity:

• $E$ is $m$-convex if, for every $x\notin E$, there exists an $m$-dimensional affine subspace $L_x\subset\mathbb{R}^n$ with $x\in L_x$ and $L_x\cap E=\varnothing$.
• The $m$-hull of $E$, denoted $\operatorname{conv}_m E$, is the intersection of all $m$-convex supersets of $E$.

The canonical convex hull is recovered at $m=n$. The primary convex hull condition for point-membership in this context, due to Zelinskii–Vyhovsʹka–Stefanchuk, is:

• For compact sets $K_i\subset\mathbb{R}^n$ ($E = \bigcup_i K_i$) and $x_0\in\mathbb{R}^n$,

$x_0\in \operatorname{conv}_m E\quad\Longleftrightarrow\quad \text{Every $m$-dimensional affine subspace through$x_0$meets at least one$K_i$.}$

This characterization generalizes the classical Carathéodory criterion for convex hull membership and is central to the shadow problem, in which the absence of lines (for $m=1$) avoiding $E$ through a point guarantees inclusion in the $1$-hull (Zelinskii et al., 2015).

2. Convex Hull Conditions in Metric and Weakly Convex Spaces

In metric geometry, convexity and hull conditions can be formulated in spaces admitting geodesic bicombings. Let $(X,d)$ be a metric space with a conical geodesic bicombing $\sigma$. The $\sigma$-convex hull $\operatorname{conv}_\sigma(A)$ of a set $A\subset X$ is the intersection of all closed $\sigma$-convex subsets containing $A$, where $\sigma$-convexity requires all $\sigma$-geodesics between pairs in $A$ to remain in the set.

While in CAT(0) spaces or injective metric spaces, the hull of a compact set remains compact, recent work demonstrates that the generalized convex-hull condition

$\operatorname{conv}_\sigma(A):\text{ all $\sigma$-geodesics between points in$A$ remain inside}$

can fail to ensure compactness in the absence of stronger global curvature assumptions. Specifically, there exist metric spaces with a conical bicombing and finite $A$ such that $\operatorname{conv}_\sigma(A)$ is non-compact, showing that minimal geodesic-convexity is insufficient for compactness of convex hulls (Basso et al., 2023).

3. Convex Hull Property in Discrete, Variational, and PDE Contexts

In variational, PDE, and finite element contexts, the convex hull property supplies a crucial generalization of the maximum principle, guaranteeing that certain solutions remain within the convex hull determined by boundary or initial data. For conforming P$_1$ finite elements on non-obtuse meshes, minimizers $u_h$ of convex energies $E(u)$ satisfy an exact discrete hull condition:

$$u_h(\Omega) \subset \operatorname{conv}\big\{u_h(z)\mid z \in \partial\Omega \big\}$$

provided the energy is strictly convex and monotone in the gradient, and mesh angles are non-obtuse (Diening et al., 2013). A strong form holds for strictly acute triangulations: if an interior nodal value attains an extreme point of the boundary-value convex hull, then $u_h$ is locally constant.

In systems of elliptic or parabolic PDEs, analogous hull properties are established for solutions $u$ to

$-\operatorname{div}(\mathbb{A}(x,u,\nabla u)) = 0$

with suitable matrix decoupling and factorization hypotheses on the coefficients. Under these conditions, solutions remain globally in the convex hull of their boundary (or parabolic boundary) values, with precise algebraic structure conditions on the differential operators (Češík, 2023).

4. Convex Hull Conditions in Optimization and Nonlinear Algebraic Settings

The convex hull condition is pivotal in providing exact relaxations for non-convex optimization problems. For example, in quadratically constrained quadratic programs (QCQPs) with epigraph set

$$S = \{(x,t)\in\mathbb{R}^n\times\mathbb{R} : q_0(x)\leq 2t,\, q_i(x)\leq 0\}$$

one seeks conditions under which $\operatorname{conv}(S)$ is precisely represented by the projected feasible set of the standard semidefinite program (SDP) relaxation.

• The convex hull is exact ($\operatorname{conv}(S)=\text{SDP}$) iff, for every boundary point $(x,t)$ of the SDP relaxation outside $S$, there exists a nonzero direction in which $(x,t)$ can be "rounded" into $S$ while remaining in the feasible SDP region. This is formalized via a tangent cone $R'(x,t)$ constructed using the facial structure of the cone of convex Lagrange multipliers, and its nontriviality is both necessary and sufficient under a geometric exposedness assumption (Wang et al., 2024).
• Sufficient conditions involving quadratic eigenvalue multiplicity or block-structured matrices yield explicit and verifiable convex-hull exactness in several practical cases (2002.01566).

In polynomial, algebraic, and moment curve settings, the convex hull condition attains a combinatorial algebraic character. Every point in the convex hull of the parametric moment curve $x(t) = (t, t^2, \ldots, t^N)$ can be written as a convex combination of at most $M = \lceil (N+1)/2\rceil$ curve points, with the evaluation mapping modulo coefficient-merge equivalences forming a homeomorphism onto the convex hull (Mazur, 2017).

Convex hull representations for sets defined by bounded bilinear (product) terms, such as $z=x y$, are determined by known polyhedral constraints (the McCormick inequalities) supplemented, in the case of additional bounds on $z$, by one or more second-order cone (SOC) constraints. These characterize the convex hull exactly, and the optimality of the associated representations is established via explicit volume formulas (Anstreicher et al., 2020).

5. Algorithmic Convex Hull Membership and Data Structures

Efficient algorithmic determination of convex hull membership is essential in computational geometry. For $P\subset\mathbb{R}^2$, a query point $q$ is in $\operatorname{conv}(P)$ if and only if it is below every oriented edge of the upper hull and above every edge of the lower hull covering its $x$-coordinate. For dynamic data structures, efficient membership checking and convex hull maintenance rely on bridge-finding routines that, using only three slope/intersection cases, reduce the complexity of queries to $O(\log n)$. This arises from the observation that every candidate for the "bridge" in a decomposable convex hull must satisfy one of three mutually exclusive slope relations, circumventing the need for cumbersome case analyses previously used (Gæde et al., 2023).

Application	Convex Hull Condition	Reference
Generalized hulls	All $m$-subspaces through $x$ intersect $E$	(Zelinskii et al., 2015)
Metric spaces	$\sigma$-hull equals completion under midpoint iteration	(Basso et al., 2023)
PDEs/Variational	Solution lies in convex hull of boundary/initial values	(Diening et al., 2013, Češík, 2023)
QCQP Optimization	Nontrivial tangent to SDP hull at all relaxed, non-epigraph points	(Wang et al., 2024, 2002.01566)
Algorithmic geom.	Point under all hull edges; dynamic slope/intersection case algorithm	(Gæde et al., 2023)

6. Logical, Semantic, and Geometric Interpretations

Convex hull conditions also manifest as semantic evaluators in logic and abstract convex geometry. In preferential conditional logic, a conditional $A\Rightarrow B$ holds in a finite point model iff all extreme points of the antecedent set $A$ satisfy $B$, equivalently if $A$ is contained in the convex hull of $A\cap B$. Completeness results establish that the axiomatization precisely captures this geometric hull-based semantic (Marti, 2020).

In the context of generalized geometric or algebraic structures, important representational theorems (e.g., Richter–Rogers representation) guarantee that finite convex geometries can always be realized as points or unions of convex polygons in $\mathbb{R}^2$, with the convex-hull condition interpreting abstract closure operators as geometric hulls.

---

In summary, the convex hull condition constitutes a unifying principle across a broad spectrum of pure and applied mathematics, serving as a test for inclusion not only in classical convex sets but in generalized, lifted, discrete, and metric settings. It is intimately connected to separation properties, solution invariance, and optimality criteria and is central to the structural analysis of convexity in analytic, algorithmic, and logical frameworks.