Real analysis, quantitative topology, and geometric complexity (Q1348591)

This expository paper reviews several classes of problems in which studying properties of mappings in terms of behavior of local averages versus pointwise behavior allows one to quantify useful features of the mappings. The properties in question roughly fall into the categories of parameterization and distortion. The author discusses three particular settings. The first involves a combinatorial problem in topology: when can a \(d\)-dimensional polyhedron be identified locally with \(\mathbb R^d\) via a piecewise linear map in a neighborhood of any point? This problem leads quickly to issues of algorithmic complexity in topology. In the second setting, imposing a metric structure enables one to control geometric complexity by putting bounds on distortion. Questions then arise how different conditions on magnitudes of derivatives can be imposed to provide senses in which a mapping is close to a rigid motion, at least in the case of Euclidean space. The author argues -- through specific observations -- that, when behavior on average is at stake, BMO is often a nice measure of distortion to use. For instance, any second order derivative of \(f\) is controlled, in BMO norm, by the uniform norm of the Laplacian of \(f\). While BMO proves to be a useful norm in Euclidean settings, it is not always useful in the case of manifolds, particularly of rough ones. For example, one might ask when the value of a function \(f\) defined on a manifold \(M\) with metric \(d\) is controlled by the average value of its gradient in the sense that \[ |f(x)|\leq C\int_M \frac{1}{d(x,y)^{n-1}}|\nabla f(y)|dV(y)\tag{*} \] in which \(dV\) denotes the volume element. Here \(M\) is locally equivalent to \(\mathbb R^n\) in the sense of local biLipschitz mappings. Such an estimate holds when the volume element satisfies a certain doubling condition and \(M\) satisfies a certain local contractivity condition. The doubling condition is somewhat restrictive. A less restrictive condition on a locally \(d\)-dimensional manifold \(M\) embedded in \(\mathbb R^{n}\) is that of Ahlfors regularity -- essentially the existence of a measure \(\mu\) supported on \(M\) such that \(\mu(B(x;r))\approx r^d\). In this third setting one replaces the condition (*) by a notion of ``uniform rectifiability'' which says, roughly, that for each \(x\in M\) and \(r>0\) a large subset of \(M\cap \overline{B}(x;r)\) is locally biLipschitz equivalent to \(\mathbb R^d\) with a Lipschitz bound independent of \(x,r\). A theorem of David and Semmes then says that if \(M\) is Ahlfors regular and locally linearly contractible then \(M\) is uniformly rectifiable. Consequences in the calculus of variations are also outlined.