geometric measure theory

Suppose you want to prove that among all surfaces spanning a given wire, one of least area actually exists. The natural plan — take a minimizing sequence of surfaces and pass to a limit — runs into a wall: the limit of nice smooth surfaces can be a crinkly, possibly self-overlapping object that is no longer a manifold. Geometric measure theory is the toolkit that lets you work with such limit objects anyway. It builds a flexible class of 'generalized surfaces' that is closed under taking limits, proves that minimizers exist within that class, and then shows that the minimizers are actually smooth except on a small singular set. It is the analytic foundation under minimal surfaces, the Plateau problem, and much of geometric analysis.

The central objects are measure-theoretic stand-ins for submanifolds. A k-rectifiable set is a set that is, up to a measure-zero error, a countable union of pieces of C^1 k-dimensional surfaces — it has a well-defined approximate tangent plane almost everywhere even if it looks rough. To carry orientation and boundary you use integral currents: roughly, a k-current is a 'surface' you can integrate k-forms over, defined as a continuous linear functional on k-forms, and an integral current is one supported on a rectifiable set with integer multiplicity and finite mass. The boundary operator is dual to the exterior derivative (Stokes' theorem becomes the definition of boundary), and the key analytic engine is the compactness theorem of Federer-Fleming: a sequence of integral currents with uniformly bounded mass and boundary mass has a subsequence converging to an integral current. That single theorem is what makes the direct method work — minimizing sequences now have limits in the right space. Varifolds are a parallel unoriented version that also tracks tangent-plane information and is the natural setting for stationary (not necessarily minimizing) surfaces and for mean-curvature flow.

Why it matters: GMT proves existence of area-minimizers (solving Plateau in great generality) and then a deep regularity theory shows the minimizer is a smooth manifold away from a singular set of low dimension — for area-minimizing hypersurfaces the singular set has codimension at least 7, which is exactly why the Simons cone in R^8 is the first singular example. The honest cautions. First, existence is cheap (compactness), regularity is expensive — the smoothness of the limit is a separate, hard theorem and is genuinely false past the critical dimension, so 'a minimizer exists' must not be read as 'a smooth minimizer exists'. Second, currents and varifolds are different tools for different jobs: currents carry orientation and boundary and are right for minimizing; varifolds are unoriented and right for stationary and flowing surfaces — using the wrong one loses essential information (a varifold cannot see cancellation of opposite orientations, a current can).