The paper studies a surprising connection between two areas of mathematics that might seem unrelated: algebraic geometry and knot theory. On the geometry side, it looks at certain algebraic spaces called braid varieties, which include important families known as Richardson varieties and positroid varieties. These spaces carry a natural symmetry group called a torus, and a central question is how "large" the orbits of this torus action are. On the knot theory side, it investigates links, which are collections of closed loops tangled together in three-dimensional space, and specifically asks how easily a link can be "filled in" by a surface in four dimensions, a property called being smoothly slice.
The main result is that a torus has a dense orbit in one of these braid varieties if and only if the associated link is smoothly slice, meaning it bounds a smooth disk in four-dimensional space. More broadly, the paper shows that when the torus orbit is not dense, the gap in dimension between the orbit and the full space is directly related to how complicated the link is from a four-dimensional perspective, a quantity called the slice genus. For knots specifically, this codimension is exactly twice the slice genus, giving a clean algebraic way to read off a topological quantity.
The paper also addresses a more combinatorial question: which of these varieties are actually algebraic tori themselves, meaning they have the simplest possible structure? The answer is governed by a pattern-avoidance condition in a combinatorial object called a Bruhat interval. Finally, for the special family of positroid links, the authors show that two different ways of measuring link complexity, the Seifert genus and the slice genus, always agree. This means a positroid link can be sliced cleanly only if it was never truly knotted to begin with, that is, only if it is an unlink.