The study of hypergraphs generalizes ordinary graphs by allowing edges to connect more than two vertices at once. A central question in combinatorics is: if you want to build a very large hypergraph that avoids some forbidden pattern, how dense can it be? This paper addresses a refined version of that question called "uniform Turan density," which asks not just about overall edge density but about how edges are distributed relative to smaller substructures. Specifically, it asks: in a large hypergraph that avoids a forbidden pattern, how densely can edges be packed when measured against every possible subhypergraph of a given intermediate size? The question has a natural "level" parameter that interpolates between previously studied extremes.
Two earlier landmark results had resolved this question at the two boundary levels. At one extreme, Erdos characterized when the answer is zero in the classical sense. At the other extreme, Reiher, Rodl, and Schacht characterized it for the level just below the edge size, and they conjectured that a similar characterization should hold at all intermediate levels too. This paper proves that conjecture. The answer turns out to depend on a simple structural property of the forbidden pattern: the density is zero precisely when you can order the vertices of the forbidden hypergraph so that whenever two edges share a certain number of vertices, those shared vertices always appear in the same positions within both edges. This is a clean combinatorial condition that fully determines whether avoidance of the pattern forces a hypergraph to be sparse in this strong uniform sense.
Beyond resolving the conjecture, the result has a striking quantitative consequence: whenever the uniform Turan density is not zero, it must be at least some explicit positive threshold, even for infinite families of forbidden patterns. The threshold depends only on the edge size and the level, not on the specific forbidden pattern. The proof strategy combines two sophisticated tools: hypergraph regularity, which lets you decompose large hypergraphs into structured pieces, and a probabilistic technique involving random hypergraphs with a special spreadability property, where the authors analyze how swapping adjacent vertices in an ordering affects edge probabilities.