Controlled fields, rough stochastic calculus, and Itô-Wentzell-Alekseev-Gröbner identities

Jannis R. Dause, Peter K. Friz, Arnulf Jentzen, Jian SongarXiv 2026 · DOI · arXiv

TL;DR

This paper introduces "space-time controlled fields," a novel framework that bridges the gap between rough path theory and classical stochastic calculus. By treating random fields as structured "jets," the authors derive a unified rough stochastic Itô–Wentzell formula, providing a powerful tool for analyzing complex systems like stochastic partial differential equations (SPDEs) and fluid dynamics under minimal regularity assumptions.

What Problem Does This Paper Address?

In stochastic analysis, the Itô formula is the bread and butter for differentiating functions of random processes. However, when the "function" itself is a random field that evolves over time (common in filtering, SPDEs, and finance), you need the more complex Itô–Wentzell formula.

The challenge arises when the underlying noise is "rough"—meaning it is more irregular than standard Brownian motion (e.g., fractional Brownian motion). Classical Itô calculus fails here, and while Rough Path Theory provides a solution for deterministic rough signals, integrating it with probabilistic martingales has historically been clunky. Previous attempts often required restrictive assumptions, like Malliavin smoothness or heavy moment requirements, making them difficult to apply to real-world numerical methods or "backward" stochastic problems.

Methods

The authors develop a two-pronged approach:

  1. Space-Time Controlled Fields: Inspired by Gubinelli’s controlled paths and Hairer’s regularity structures, they define a field not just as a function, but as a "7-tuple" of objects (a jet) that encodes spatial and temporal derivatives simultaneously. They prove that these fields are stable under composition: if you evaluate a controlled field along a controlled path, the result is itself a controlled path.
  2. Strongly Controlled Rough Semimartingales: They introduce a second-order variant of rough semimartingales. The key "aha" moment here is a correspondence principle: "rough stochastic identities" can be derived directly from "rough identities" by replacing deterministic reference paths with a semi-stochastic joint lift. This allows for a "moment-free" calculus that avoids the complex integration spaces usually required in this field.

Datasets

Not discussed in the paper. As a work of theoretical mathematics and stochastic analysis, the paper relies on rigorous proofs and definitions rather than empirical data or benchmarks.

Key Insights and Ideas

  • The "Jet" Perspective: By treating derivatives as intrinsic parts of the field's definition rather than external operations, the authors turn complex calculus problems into more manageable algebraic ones.
  • Algebraic Stability: The main result (Theorem 3.12) shows that the structure of a "controlled field" is preserved under composition. This is the engine that allows them to derive the Itô–Wentzell formula almost as a corollary.
  • Simplifying the IAG Formula: The paper provides a significantly simplified proof of the Itô–Alekseev–Gröbner (IAG) formula. Crucially, they remove the need for Malliavin smoothness assumptions on the coefficients, answering a conjecture from recent literature (Hudde et al., 2024).
  • Unified Framework: The approach handles rough paths, martingales, and drifts in a single equation, making it highly relevant for the "mixed" regimes found in modern Rough Stochastic Differential Equations (RSDEs).

Key Results

  • Theorem 4.21 (The Rough Stochastic Itô–Wentzell Formula): This is the paper's centerpiece. It provides a composition rule for a random field $H_t(x)$ evaluated at a process $Y_t$, where both can have rough and martingale components.
  • Refined IAG Formula: They establish the IAG formula in a "Skorokhod form," which is essential for forward-backward stochastic analysis. Their version requires less integrability and fewer smoothness conditions than previous iterations.
  • Forward-Backward Interpolation: The authors show how their framework naturally derives stochastic interpolation formulas, which are critical for analyzing the error in numerical simulations of SDEs.

Limitations and Future Work

  • Finite Dimensions: The current work focuses on finite-dimensional real Banach spaces. While the authors note that extensions to infinite dimensions (relevant for some SPDEs) are possible, they are not the focus here.
  • $L^0$-Regularity: The authors acknowledge that a full analysis of $L^0$-regularity (convergence in probability without assuming finite moments) for certain rough flows lies beyond the scope of this paper.
  • Applications: While the paper provides the theoretical machinery, applying these identities to specific high-dimensional filtering problems or complex mean-field games remains a promising direction for future research.

Summary updated

Abstract

We develop a calculus of space-time controlled fields for rough stochastic systems. This approach provides a unified composition rule for evaluating random fields along rough semimartingales and yields a rough stochastic Itô-Wentzell formula under natural and verifiable regularity assumptions. Our motivation comes from works of Hudde et al. (2024) and, independently, Del Moral and Singh (2022) where the authors established, respectively, Itô-Alekseev-Gröbner, backward Itô-Wentzell, and diffusion interpolation formulas.
BibTeX
@misc{dause2026controlled,
  author = {Jannis R. Dause and Peter K. Friz and Arnulf Jentzen and Jian Song},
  title = {{Controlled fields, rough stochastic calculus, and Itô-Wentzell-Alekseev-Gröbner identities}},
  year = {2026},
  eprint = {2603.05388},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2603.05388},
}

Loading related content...

Figure