Loading field map
Convex geometry studies convex sets—sets that contain the line segment between any two of their points—and the structures they form. The central tension driving the field is between two ways of understanding these sets: measuring them (their volume, surface area, mixed volumes) and organizing them (their face structure, intersection patterns, combinatorial types). Over the past century, this tension has generated four major frameworks, each emphasizing a different kind of question and a different set of tools. The story of convex geometry is the story of how these frameworks emerged, how they borrowed from and competed with one another, and how they now coexist in a rich, pluralistic discipline.
The first systematic framework for convex geometry was Brunn-Minkowski Theory, which crystallized around 1887 in the work of Hermann Brunn and Hermann Minkowski. Its central object is the convex body—a compact convex set with nonempty interior—and its central operation is Minkowski addition: the sum of two convex bodies is the set of all sums of their points. The Brunn-Minkowski inequality, the framework's signature result, relates the volume of a Minkowski sum to the volumes of its summands, providing a sharp lower bound that encodes deep geometric information.
Brunn-Minkowski Theory gave convex geometry its first powerful analytic engine: mixed volumes. These are symmetric multilinear functionals that arise when one expands the volume of a Minkowski sum of several bodies. Mixed volumes encode not just size but shape: the surface area, mean curvature, and other metric invariants of a convex body all appear as special cases. The framework's original motivation came from the geometry of numbers, where Minkowski used convex bodies to study lattice points and Diophantine approximation. But its influence quickly spread: the Brunn-Minkowski inequality itself became a template for a whole family of isoperimetric and concentration inequalities that later frameworks would inherit and generalize.
Around 1900, a second framework emerged that asked a fundamentally different kind of question. Instead of measuring convex bodies, Combinatorial Convexity asked about the patterns they form when they intersect. The landmark results were Helly's theorem (1913) and Carathéodory's theorem (1907). Helly's theorem says that if a family of convex sets in ℝᵈ has the property that every d+1 of them intersect, then the whole family intersects. Carathéodory's theorem says that any point in the convex hull of a set can be expressed as a convex combination of at most d+1 points from that set.
These theorems are purely combinatorial: they involve no volumes, no measures, only intersection patterns and convex combinations. Combinatorial Convexity did not replace Brunn-Minkowski Theory; it complemented it by providing a qualitative, structural counterpart to the volumetric approach. Where Brunn-Minkowski Theory measured, Combinatorial Convexity classified. The two frameworks coexisted, each addressing a different layer of convex geometry's central tension. Combinatorial Convexity also connected convex geometry to other parts of mathematics—to topology via the Euler characteristic of convex cell complexes, and to linear programming via the facial structure of polytopes.
The third framework, Geometric Functional Analysis, emerged around 1950 and transformed convex geometry by changing its setting. Earlier frameworks worked in finite-dimensional Euclidean space. Geometric Functional Analysis moved the study of convex sets into infinite-dimensional Banach spaces, treating convex bodies as unit balls of norms. This shift was not a mere generalization; it reframed convex geometry as a branch of functional analysis, where the geometry of a space is encoded in the shape of its unit ball.
The key innovation was the Banach-Mazur distance, which measures how far a convex body is from being a Euclidean ball. This invariant allowed convex geometers to ask quantitative questions about the shape of Banach spaces: how round can a convex body be? How flat? The framework also introduced powerful duality arguments: the polar of a convex body corresponds to the dual norm, so geometric properties of a body and its polar are linked by functional-analytic duality. Geometric Functional Analysis absorbed the volumetric tools of Brunn-Minkowski Theory—the Brunn-Minkowski inequality itself holds in infinite dimensions—but added a new layer of analytic structure. It narrowed the field's focus to the geometry of norms, while simultaneously expanding its reach into the theory of Banach spaces.
By 1970, a fourth framework had begun to take shape, driven by a simple observation: the behavior of convex bodies in high dimensions is radically different from their behavior in low dimensions. Asymptotic Convex Geometry studies convex bodies in ℝⁿ as n grows large, asking about typical, universal properties rather than exact, individual ones. Its central phenomenon is concentration of measure: in a high-dimensional convex body, most of the volume lies near the boundary, and most of the mass of a product measure is concentrated near a thin shell. This counterintuitive fact—discovered by Paul Lévy and later developed by Vitali Milman—transformed convex geometry into a probabilistic discipline.
Asymptotic Convex Geometry inherited the metric tools of Brunn-Minkowski Theory and the normed-space perspective of Geometric Functional Analysis, but it added a new kind of question: instead of asking what a particular convex body looks like, it asks what a random projection of a convex body looks like, or what the average width of a convex body is. The framework's signature results include Milman's proof of Dvoretzky's theorem, which shows that every high-dimensional convex body has a nearly Euclidean slice, and the development of the concentration inequality for Lipschitz functions on the sphere. These results revealed a hidden simplicity in high dimensions: despite their complexity, convex bodies in large n exhibit universal, law-like behavior.
Today, all four frameworks remain active, but they have settled into a division of labor. Brunn-Minkowski Theory continues as the volumetric infrastructure for the entire field: its inequalities are used in every other framework, and its mixed volumes remain the primary tool for studying the shape of convex bodies. Combinatorial Convexity has evolved into the study of polytopes and face lattices, with deep connections to algebraic combinatorics and optimization. Geometric Functional Analysis remains the framework of choice for questions about the geometry of Banach spaces and the structure of norms. Asymptotic Convex Geometry is the most active frontier, driving research on concentration, random matrices, and high-dimensional probability.
What the leading frameworks agree on is that convex geometry is fundamentally about inequalities: the Brunn-Minkowski inequality, the isoperimetric inequality, the concentration inequality. They share a common toolbox of volumetric estimates, duality arguments, and combinatorial decompositions. Where they disagree is on what the central questions should be. Geometric Functional Analysis tends to ask about exact, deterministic structure: what is the Banach-Mazur distance between two given bodies? Asymptotic Convex Geometry tends to ask about typical, probabilistic behavior: what does a random section of a convex body look like? This tension between the deterministic and the probabilistic, the exact and the typical, drives much of the current research. Major open problems—the Kannan-Lovász-Simonovits conjecture, the slicing problem, the hyperplane conjecture—sit at the intersection of these frameworks, requiring both the analytic precision of Geometric Functional Analysis and the probabilistic insight of Asymptotic Convex Geometry.
Convex geometry today is not a settled field but a dynamic one, held together by a shared commitment to understanding convex sets through multiple lenses. The four frameworks are not rivals; they are complementary perspectives that together reveal the richness of a single, deceptively simple object: the convex set.