Loading field map
Geometric measure theory is the branch of analysis that studies the structure of sets and mappings in Euclidean space (and, more generally, in metric spaces) using the tools of measure theory. Its central concern is to give a precise meaning to the notion of "surface area" for highly irregular objects—objects that may have no tangent planes, may be fractal, or may be defined only as limits of smoother approximations. The field asks: what does it mean for a set to have a well-defined lower-dimensional size, and when can such a set be approximated, decomposed, or reconstructed from its infinitesimal pieces?
Ordinary calculus handles surfaces that are smooth, meaning they have tangent planes at every point. But many natural objects—boundaries of domains with corners, soap films spanning wire frames, solutions to variational problems, or the supports of measures arising in physics—fail to be smooth. A first difficulty is that the usual notion of length, area, or volume via parametrization breaks down. A curve that is merely continuous can fill a square (as Peano showed), so its "length" cannot be computed by summing chord lengths over finer and finer partitions; the sum diverges. Similarly, a set can have positive area in the plane yet contain no smooth curve, or can be a fractal with no well-defined tangent anywhere.
The foundational move of geometric measure theory is to replace the geometric object itself with a measure—a function that assigns a size to subsets—and then to study the measure rather than the set directly. The most important such measures are the Hausdorff measures, introduced by Felix Hausdorff in 1918. For any dimension \( s \geq 0 \), the \( s \)-dimensional Hausdorff measure \( \mathcal{H}^s \) assigns to a set a number that generalizes the idea of \( s \)-dimensional volume. For integer \( s \), \( \mathcal{H}^s \) agrees with ordinary length, area, or volume on smooth objects, but it is defined for arbitrary sets, including fractals. The Hausdorff dimension of a set is the critical exponent at which its Hausdorff measure jumps from infinity to zero; this gives a rigorous way to say that a set has "dimension" \( d \) even when it is not a manifold.
Hausdorff measure alone, however, is too crude for many purposes. A set can have infinite \( \mathcal{H}^1 \) measure yet still be a perfectly reasonable curve in a generalized sense. The field therefore developed finer tools: the notion of a rectifiable set, which is a set that can be covered, up to a set of measure zero, by countably many Lipschitz images of Euclidean space. A Lipschitz map is one that does not stretch distances by more than a fixed factor; such maps are allowed to be non-smooth but cannot tear space. Rectifiable sets are the natural generalization of smooth submanifolds: they have tangent planes almost everywhere (in a measure-theoretic sense), and they support integration by parts. Their complements are the purely unrectifiable sets, which meet every Lipschitz image in a set of zero measure; these are the genuinely fractal objects.
The deepest questions in geometric measure theory concern not just static sets but objects that can vary, deform, and serve as solutions to optimization problems. Two families of objects were introduced in the mid-twentieth century to handle these questions, and they remain the field's two main working tools.
Currents were introduced by Georges de Rham in the 1950s as a generalization of surfaces that could be integrated against differential forms. A current is a continuous linear functional on the space of smooth differential forms; it is the measure-theoretic analogue of a distribution (in the sense of Laurent Schwartz) but for forms of arbitrary degree. An oriented smooth surface defines a current by integrating forms over it, and the boundary of a current is defined by the exterior derivative via Stokes's theorem. The crucial advantage is that currents form a complete metric space under a natural norm (the mass norm), so one can take limits of sequences of surfaces and obtain a well-defined limiting current. This compactness property, proved by Herbert Federer and Wendell Fleming in 1960, is the foundation of the modern calculus of variations in higher dimensions: it guarantees that minimizing sequences for area-type functionals have convergent subsequences, provided one works with integral currents—currents with integer multiplicities and finite mass whose boundaries also have finite mass.
The price of this compactness is that the limit of smooth surfaces may be a current with singularities, multiplicities, or cancellations. A current can represent a surface with a "fold" where two sheets coincide with opposite orientations, and these sheets cancel in the limit. This is not a defect but a feature: it allows the theory to handle problems where the optimal object genuinely has such structure, such as soap films that touch themselves.
Varifolds, introduced by Frederick Almgren in the 1960s, take a different approach. A varifold is a measure on the product space of points and Grassmann manifolds of tangent planes; it records not only where a surface lies but also what tangent directions are present, weighted by multiplicity. Varifolds do not require an orientation, so they can represent non-orientable surfaces and surfaces with arbitrary multiplicity without cancellation. They also have a natural notion of first variation—the rate of change of mass under a deformation of space—which allows one to define mean curvature even for very irregular objects. The theory of varifolds is the natural setting for studying minimal surfaces that are not necessarily orientable, and for the mean curvature flow of nonsmooth surfaces.
The relationship between currents and varifolds is complementary. Currents are better suited to problems with a fixed boundary condition and a variational structure, because their boundary operator is exact and their compactness is well understood. Varifolds are better suited to geometric evolution equations and to problems where orientation is irrelevant or impossible. Both can represent the same smooth surface, but they diverge on singular objects: a current sees a folded sheet with cancellation, while a varifold sees it with multiplicity. The choice of object is not a matter of taste but of which properties one needs: exactness of boundary versus flexibility of multiplicity.
The central technical challenge of geometric measure theory is the regularity problem: given a set or current that minimizes area (or some related functional), how singular can it be? The field's most celebrated results concern this question.
For area-minimizing hypersurfaces (codimension one), the regularity theory is essentially complete. Ennio De Giorgi in 1961 and, independently, Frederick Almgren and Ernst Reifenberg proved that an area-minimizing current of dimension \( n-1 \) in \( \mathbb{R}^n \) is a smooth manifold except on a singular set of Hausdorff dimension at most \( n-8 \). This means that in dimensions up to 7, area-minimizing hypersurfaces are completely smooth; in dimension 8 and above, singularities can occur but are rare and well understood. The prototype is Simons's cone in \( \mathbb{R}^8 \), which is area-minimizing but singular at the origin. This result is sharp and explains why the regularity theory of minimal surfaces in higher dimensions is fundamentally different from the classical theory in three dimensions.
For area-minimizing currents of higher codimension (surfaces of dimension \( k \) in \( \mathbb{R}^n \) with \( n-k \geq 2 \)), the situation is more complicated. Almgren's big regularity theorem, proved in the 1980s and published posthumously in 2000, showed that the singular set has Hausdorff dimension at most \( k-2 \), but the proof was extraordinarily long and complex. A substantially simplified proof was given by Camillo De Lellis and Emanuele Spadaro in the 2010s, using a different method based on "center manifold" techniques. The key difference from the hypersurface case is that in higher codimension, singularities can form along lower-dimensional sets rather than just at isolated points, and the structure of these singularities is more varied.
For varifolds, the regularity theory is less complete. A stationary varifold—one whose first variation vanishes—need not be smooth; there are examples of stationary varifolds with large singular sets. The best general result, due to Allard in 1972, is a regularity theorem that says: if a varifold has bounded mean curvature and its mass density is sufficiently close to that of a flat disk at a point, then it is a smooth manifold in a neighborhood of that point. This is a epsilon-regularity theorem: it gives smoothness only where the varifold is already nearly flat. The global structure of singularities of stationary varifolds remains an open problem.
A closely related thread, often grouped under geometric measure theory, concerns functions of bounded variation (BV functions). A function \( f \) on a domain in \( \mathbb{R}^n \) is of bounded variation if its distributional derivative is a vector-valued measure with finite total variation. Such functions can have jump discontinuities along hypersurfaces, and their level sets are sets of finite perimeter. The theory of sets of finite perimeter, developed by Renato Caccioppoli and De Giorgi in the 1950s, is the study of sets whose characteristic functions are BV. These sets have a well-defined measure-theoretic boundary, and the perimeter is the total variation of the characteristic function.
This theory is not a separate subject but an integral part of geometric measure theory: it provides the natural framework for the isoperimetric problem, for the study of minimal surfaces with free boundaries, and for image processing and other applied fields where one seeks to recover sharp edges from noisy data. The BV framework also connects geometric measure theory to the calculus of variations and to partial differential equations, since many variational problems with nonsmooth solutions are naturally posed in BV spaces.
Contemporary geometric measure theory is organized around several active research fronts. One is the study of quantitative versions of the classical theorems: instead of asking whether a set is rectifiable, one asks how close it is to being rectifiable, measured by the size of the set where it fails. This has led to the theory of uniform rectifiability, developed by Guy David and Stephen Semmes, which characterizes sets that are rectifiable in a scale-invariant way and has deep connections to harmonic analysis and singular integral operators.
Another front is the extension of the theory from Euclidean space to metric spaces. The notions of Hausdorff measure and rectifiability make sense in any metric space, but many tools—such as the compactness theorems for currents—do not. The development of a geometric measure theory on metric spaces, particularly on Heisenberg groups and on spaces with Ricci curvature bounded below, is an active area. The Heisenberg group is a model for sub-Riemannian geometry, where the usual Euclidean metric is replaced by a metric that only allows motion along certain directions; rectifiability in this setting behaves differently and is not yet fully understood.
A third front is the use of geometric measure theory in the study of partial differential equations, particularly in the regularity theory of free boundary problems and of harmonic maps. The methods of Almgren, De Giorgi, and Allard—blow-up analysis, monotonicity formulas, and dimension reduction—have become standard tools in these areas. The field also interacts with geometric analysis through the study of minimal submanifolds in Riemannian manifolds, where the regularity theory of currents and varifolds provides the existence theory for minimal surfaces in arbitrary ambient spaces.
The field's foundational texts—Federer's Geometric Measure Theory (1969) and the later monographs by Simon, Mattila, and others—remain in use, but the subject has moved well beyond them. The modern practitioner is as likely to work on quantitative rectifiability, on the structure of measures with fractional dimension, or on the regularity of solutions to nonlinear PDEs as on the classical Plateau problem. What unifies the field is not a single technique but a shared commitment to the idea that geometric objects can be studied rigorously even when they are too irregular for classical differential geometry, and that the right tools for this study come from measure theory and analysis rather than from topology or algebra alone.