Integration of nonsmooth 2-forms: from Young to Itô and Stratonovich (Q6184586)

This paper provides an interesting argument to show that geometric integrals of the type \(\int _{\Omega}fdg_1\wedge dg_2\) can be defined over a two-dimensional domain \(\Omega\) when the functions \(f, g_1, g_2 : \mathbb{R}^2 \rightarrow \mathbb{R}\) are just Hölder continuous with sufficiently large Hölder exponents and the boundary of \({\Omega}\) has sufficiently small dimension, by summing over a refining sequence of partitions the discrete Stratonovich or Itô type terms. Hence, it shows an equivalence of Young, Itô and Stratonovich integration for the two-dimensional extension of the classical Young integral that coincides with the integral introduced recently by \textit{R. Züst}, [Calc. Var. Partial Differ. Equ. 40, No. 1--2, 99--124 (2011; Zbl 1219.49036)]. The respective integral may be called both Itô and Stratonovich, and in fact generalizes the one-dimensional Young integral. In a broad sense the authors construct explicitly, via the appropriate discrete approximations, the extension of the classical notion of the integral of the differential \(2\)-form \(fdg_1\wedge dg_2\) over any sufficiently nice oriented planar domain \({\Omega} \subset \mathbb{R}^2\) to the case when the maps \(f: \mathbb{R}^2 \rightarrow \mathbb{R}, g := (g_1, g_2): \mathbb{R}^2 \rightarrow \mathbb{R}^2\) are only Hölder continuous. The so-defined integral can be extended to a large class of bounded open sets \({\Omega} \subset \mathbb{R}^2\) having sufficiently small box-counting dimension of the topological boundary. These results give a partial answer to the curious and important question, along what kind of ``surfaces'' (or, more generally, against which de Rham currents) one can integrate the ``rough forms'' of the type \(fdg_1\wedge dg_2\) with just Hölder \(f, g_1, g_2\). Additionally, if \(f\) is represented in the particular form \(f(x) = F(x, g(x))\), then the conditions for the existence of the integral extending the classical one for smooth forms, i.e., the requirements on the Hölder exponents of \(g^i\), may be significantly relaxed at the price of requiring \(F : \mathbb{R}^2\times \mathbb{R}^2 \rightarrow \mathbb{R}\) to be sufficiently regular. This relaxation is achieved by employing Stratonovich germs. This is a very particular feature of Stratonovich summation and does not hold for Itô summation. Most of the derivations refer to rough paths theory and rough path forms. One could legitimately ask if the analogy used throughout the article with stochastic integrals could be extended even further for a broader type of stochastic calculus (say, of Malliavin nature). Additional extensions to probability theory as well as differential geometry and functional analysis could be interesting to explore.