1. Introduction
Brownian motion, also called a Wiener process, is a cornerstone of modern probability theory and a natural playground for stochastic calculus. It captures the erratic, continuoustime fluctuations observed in physical particles, financial assets, and many other systems. Stochastic calculus provides a formal framework to manipulate and integrate functions of such random paths, yielding tools like Its lemma that have become indispensable in fields ranging from quantitative finance to statistical physics.
2. Basic Properties of Brownian Motion
A standard Brownian motion W(t) on a probability space (,,P) satisfies four key properties:
- Initial condition:
W(0)=0almost surely. - Independent increments: For any
0sW(t)W(s)is independent of the algebra generated by{W(u):us}. - Stationary increments:
W(t)W(s)follows a normal distribution with mean 0 and variancets, i.e.N(0, ts). - Continuity: With probability one, the sample paths
tW(t)are continuous functions of time.
These properties make Brownian motion a Gaussian process with covariance Cov(W(t),W(s))=min(t,s), a fact that will be repeatedly used when deriving stochastic integrals.
3. From Ordinary to Stochastic Integration
In deterministic calculus, the Riemann integral approximates the area under a curve by summing rectangles whose heights are determined by function values at sample points. The same idea can be extended to random functions, but the nondifferentiability of Brownian paths forces a different construction. The It integral of a predictable process X(t) with respect to W(t) is defined as the limit (in mean square) of sums
X(t_i)(W(t_{i+1})W(t_i)) where the partition points t_i become increasingly fine. Crucially, the integrand is evaluated at the left endpoint t_i, which yields an nonanticipating integral suitable for modelling causal systems.
4. Its Isometry
A fundamental result that validates the It integral is Its isometry:
E\Big[ \big(_0^T X(t) dW(t) \big)^2 \Big] = E\Big[ _0^T X(t)^2 dt \Big] This identity mirrors the Pythagorean theorem: the variance of the stochastic integral equals the expected Lnorm of the integrand. It guarantees that the integral is welldefined for any squareintegrable predictable process and provides a convenient tool for estimating moments.
5. Its Lemma
Its lemma is the stochastic analogue of the chain rule. Suppose Y(t)=f(t,W(t)) where fC^{1,2} (once differentiable in time and twice in space). Then
dY(t) = f_t(t,W(t))dt + f_x(t,W(t))dW(t) + f_{xx}(t,W(t))dt The extra term f_{xx} dt arises from the quadratic variation of Brownian motion, which satisfies [W,W]_t = t. This correction term is the source of many counterintuitive results in stochastic calculus, such as the fact that W(t)^2t is a martingale.
6. Stochastic Differential Equations (SDEs)
Many dynamical systems driven by noise are modeled by stochastic differential equations of the form
dX(t) = (t,X(t))dt + (t,X(t))dW(t) where is the drift and the diffusion coefficient. Solutions are interpreted in the It sense unless stated otherwise. The existence and uniqueness of solutions are typically ensured by Lipschitz and linear growth conditions on and .
7. Example: Geometric Brownian Motion
One of the most celebrated SDEs is geometric Brownian motion (GBM), which underlies the BlackScholes model for option pricing. GBM satisfies
dS(t) = S(t)dt + S(t)dW(t) Applying Its lemma to the logarithm yields an explicit solution:
S(t)=S(0)exp\Big[()t + W(t)\Big] The exponential form guarantees positivity of the asset price, a feature that cannot be achieved with a simple linear SDE.
8. Martingales and the Optional Stopping Theorem
A process M(t) is a martingale if E[M(t) | _s] = M(s) for all st. Both W(t) and the compensated square W(t)^2t are martingales. The optional stopping theorem states that, for a bounded stopping time , a martingale evaluated at has the same expectation as at time zero:
E[M()] = M(0) This result is frequently used to prove fairness of gambling strategies and to derive pricing formulas in finance.
9. Numerical Approximation: EulerMaruyama Scheme
Exact solutions of SDEs are rare, so numerical methods are essential. The EulerMaruyama discretisation for a step size t reads
X_{n+1}=X_n + (t_n,X_n)t + (t_n,X_n)W_n where W_n = W(t_{n+1})W(t_n) is a Gaussian random variable with mean zero and variance t. The method converges in the meansquare sense with order , a slower rate than deterministic Euler because of the stochastic increments.
10. Applications Beyond Finance
While financial engineering popularised stochastic calculus, the techniques are equally valuable in other disciplines:
- Physics: Modeling the motion of particles in a fluid (Langevin equations) or the evolution of fields in quantum optics.
- Biology: Describing population dynamics under random environmental fluctuations (stochastic logistic growth).
- Engineering: Analyzing noisedriven control systems and signal processing algorithms.
- Machine learning: Stochastic gradient dynamics can be interpreted as discretised SDEs, providing insight into convergence and escape from local minima.
11. Further Reading
For a deeper dive, the following classics are recommended:
- ksendal, Stochastic Differential Equations: An Introduction with Applications, 6th ed., Springer, 2010.
- Karatzas & Shreve, Brownian Motion and Stochastic Calculus, 2nd ed., Springer, 1991.
- Shreve, Stochastic Calculus for Finance II: ContinuousTime Models, Springer, 2004.
- Protter, Stochastic Integration and Differential Equations, 2nd ed., Springer, 2005.
12. Concluding Remarks
Stochastic calculus on Brownian motion provides a powerful language for describing systems that evolve in the presence of randomness. By extending the ideas of integration and differentiation to nondeterministic paths, it opens a gateway to rigorous analysis of phenomena that would otherwise be intractable. Whether you are pricing exotic derivatives, studying the diffusion of molecules, or designing robust algorithms, the tools presented here form a solid foundation that can be built upon with more advanced concepts such as stochastic partial differential equations, Malliavin calculus, and rough path theory.
