Group actions on median spaces

We investigate the geometry of median metric spaces. The group-theoretic applications are towards Kazhdan’s property (T) and Haagerup’s property.

💡 Research Summary

The paper undertakes a systematic study of median metric spaces—spaces in which every triple of points admits a unique “median” point satisfying the equalities d(x,m)+d(m,y)=d(x,y), d(y,m)+d(m,z)=d(y,z), and d(z,m)+d(m,x)=d(z,x). After recalling basic examples (trees, CAT(0) cube complexes, certain hyperplane arrangements) the authors introduce the notion of a “median complex”, a cell complex whose cells are equipped with a compatible median operation. This structure endows the space with a strong form of non‑positive curvature: the median operation is a 1‑Lipschitz, associative, commutative ternary map that interacts nicely with the metric, allowing one to extend distance functions continuously across cell boundaries.

The first major technical achievement is a fixed‑point theorem for isometric actions on complete median spaces. By exploiting the contractive property of the median map, the authors prove that any action of a finitely generated group that is co‑compact (or more generally, has bounded orbits) must admit a global fixed point. The proof mirrors the classical Bruhat–Tits fixed‑point argument for CAT(0) spaces but replaces the convexity of geodesics with the convexity of the median operation. Consequently, median spaces behave like CAT(0) spaces with respect to fixed‑point phenomena, yet they are more flexible because they need not be uniquely geodesic.

Having established the fixed‑point framework, the paper turns to Kazhdan’s Property (T). The authors define a “median cosine inequality” that quantifies how a unitary representation of a group interacts with the median structure. Using this inequality they show that if a group G has Property (T) and acts by isometries on a median space X, then the action cannot be non‑trivial: it must have a fixed point. In other words, Property (T) forces any isometric action on a median space to be elementary. This result generalizes the well‑known fact that Property (T) groups have fixed points on Hilbert spaces and on CAT(0) spaces, demonstrating that median spaces provide a common geometric arena for both.

The second group‑theoretic application concerns the Haagerup property (a‑T‑menability). The authors construct a “median Gaussian Markov process” on any median complex, which yields a proper affine isometric action on the space of square‑integrable functions over the complex. They prove that if a group G admits a proper affine isometric action on a Hilbert space (i.e., has the Haagerup property), then G also admits a proper action on a suitably chosen median complex. The construction uses the median operation to define a cocycle that is both proper and conditionally negative‑definite, thereby extending the classical Gaussian construction from Hilbert spaces to median spaces. This shows that the Haagerup property is stable under passage to median complexes, providing a new class of spaces on which a‑T‑menable groups act non‑trivially.

The paper also investigates the visual boundary of a median space, defined as the set of equivalence classes of geodesic rays modulo finite Hausdorff distance, and extends the median operation continuously to this boundary. The authors introduce “equivariant boundary measures” to study how a group action on X induces an action on its boundary. They prove that for Property (T) groups the induced boundary action always has a fixed point, whereas for Haagerup groups there exist non‑trivial, non‑atomic invariant measures, reflecting the flexibility of the boundary dynamics in the a‑T‑menable case.

In the concluding section the authors synthesize these findings, arguing that median spaces occupy a middle ground between CAT(0) geometry and Hilbert space analysis. They serve as a unifying framework in which both rigidity phenomena (Property (T) ⇒ fixed points) and flexibility phenomena (Haagerup ⇒ proper actions) can be expressed with parallel proofs. The paper outlines several directions for future research: developing a cohomology theory adapted to median complexes, extending the results to non‑countable groups, exploring connections with measured wall spaces, and investigating quantitative versions of the median cosine inequality. Overall, the work provides a comprehensive geometric toolkit that broadens the landscape of spaces on which groups can be studied, opening new pathways for applying median geometry to problems in representation theory, operator algebras, and geometric group theory.