Mathematical framework for systems with random noise, used in biological and turbulent dynamics.Stochastic differential equations (SDEs) are equations that describe how a quantity changes over time when its evolution is influenced by random fluctuations, or noise, in addition to deterministic forces. Think of a pollen grain jiggling in water (Brownian motion) or a stock price fluctuating under market whims: the underlying trend is there, but randomness constantly perturbs it. SDEs formalize this by combining ordinary differential equations with a term that injects randomness, typically modeled as a Wiener process (a continuous-time random walk). They are essential in fields where noise is not a nuisance but a core feature of the system, such as biology (population dynamics, gene regulation), finance (option pricing), and physics (turbulence, thermal fluctuations).
The mathematical foundations were laid by Kiyosi Itô in the 1940s, who developed the stochastic integral and calculus that bear his name. Earlier, Albert Einstein and Marian Smoluchowski had used heuristic stochastic models for Brownian motion, but Itô provided the rigorous framework. Ruslan Stratonovich introduced an alternative integral in the 1960s, and the theory has since expanded into a rich field with applications across sciences.
An SDE is written like dX_t = a(X_t, t) dt + b(X_t, t) dW_t, where X_t is the state, a is the drift (deterministic trend), b is the diffusion (noise strength), and dW_t is an infinitesimal increment of a Wiener process. The key subtlety is that the noise term is not a smooth function; it has infinite variation, so the integral of b dW_t must be defined carefully. Two main interpretations exist: Itô (where the integrand is evaluated at the left endpoint of each time step) and Stratonovich (evaluated at the midpoint). The choice affects the calculus rules and the physical meaning. Solutions are stochastic processes, and instead of a single trajectory, one obtains a probability distribution over possible paths.
SDEs are used to model option prices in quantitative finance (e.g., the Black-Scholes equation is a partial differential equation derived from an SDE), to simulate neural activity in computational neuroscience, and to describe turbulent flows in physics.
The path of a Wiener process, the fundamental noise in SDEs, is continuous everywhere but differentiable nowhere — it is so jagged that its length over any finite interval is infinite.




