# Notes on Stochastic Differential Equations
*A synthesis of handwritten notes from MATH310 (F21/S22/F24), [video lectures](https://www.youtube.com/playlist?list=PLfzyv9XksyEfntM7sGdNDLp8kw06I5trG) by [Juan MR Parrondo](https://en.wikipedia.org/wiki/J._M._R._Parrondo), and parts of, [An Introduction to Stochastic Differential Equations](https://www.cmor-faculty.rice.edu/~cox/stoch/SDE.course.pdf) Version 1.2, by [Lawrence C. Evans](https://en.wikipedia.org/wiki/Lawrence_C._Evans) (2004)*
---
## Overview and Motivation
These notes trace a path from discrete [random walks](https://en.wikipedia.org/wiki/Random_walk) to continuous [stochastic processes](https://en.wikipedia.org/wiki/Stochastic_process), ultimately arriving at [stochastic differential equations](https://en.wikipedia.org/wiki/Stochastic_differential_equation) and the modified calculus needed to work with them. The journey reveals how noise fundamentally changes the rules of calculus and why the [Itô correction](https://en.wikipedia.org/wiki/It%C3%B4%27s_lemma) is essential for properly modeling real-world phenomena.
### Document Structure
- [[Random-Walks]] - Develops the symmetric random walk from first principles through the binomial distribution to the continuum limit. Shows how the scaling $\Delta y \propto \sqrt{\Delta t}$ naturally emerges and connects to diffusion.
- [[Wiener-Process]] - Constructs the Wiener process as the continuum limit of random walks and explores its mathematical properties. Develops the autocorrelation structure and introduces white noise as the formal derivative, including the L² basis representation.
- [[Ito-Formula]] - Explains why classical calculus fails for Brownian motion through the key example where "y returns the Wiener process itself." Derives Itô's lemma showing how $(dW)^2 = dt$ necessitates the extra second-derivative term.
- [[SDE-Fundamentals]] - Introduces stochastic differential equations starting from noise-perturbed exponential decay. Works through the complete solution of the first-order linear SDE using Itô's lemma and connects to physical systems via the Langevin equation.
---
## 1. From Random Walks to Continuous Processes
*See [[Random-Walks#1. Definition and Basic Structure]] for full development*
### The Discrete Foundation
We begin with the [symmetric random walk](https://en.wikipedia.org/wiki/Random_walk#Symmetric_random_walk) - a particle taking steps of size $\pm\Delta y$ with equal probability. After $n = t/\Delta t$ steps:
- **Mean position**: $\langle y(t) \rangle = 0$
- **Variance**: $\sigma_y^2 = n(\Delta y)^2 = \frac{t(\Delta y)^2}{\Delta t}$
The variance grows linearly with time - a signature of diffusive behavior.
### The Continuum Limit
*See [[Random-Walks#4. The Continuum Limit]] for detailed derivation*
To obtain a meaningful limit as $\Delta t \to 0$, we must scale the step size: $\Delta y = \sqrt{D \Delta t}$, where $D$ is the [diffusion coefficient](https://en.wikipedia.org/wiki/Mass_diffusivity). This yields:
$\sigma_y^2 = Dt$
In this limit, the random walk converges to the **[Wiener process](https://en.wikipedia.org/wiki/Wiener_process)** $W(t)$ - the mathematical model of [Brownian motion](https://en.wikipedia.org/wiki/Brownian_motion).
> [!info] Key Insight
> The scaling $\Delta y \propto \sqrt{\Delta t}$ is crucial. It's why Brownian paths are continuous but nowhere differentiable - the particle moves infinitely fast at infinitesimal scales.
---
## 2. The Wiener Process and White Noise
*Full treatment in [[Wiener-Process#3. Definition of the Wiener Process]]*
### Properties of the Wiener Process
The Wiener process $W(t)$ is characterized by:
1. $W(0) = 0$
2. Independent increments
3. $W(t) - W(s) \sim \mathcal{N}(0, |t-s|)$ for $t > s$
4. Continuous paths (almost surely)
### The Autocorrelation Structure
*Detailed derivation in [[Wiener-Process#Autocorrelation Structure]]*
For the Wiener process:
$E[W(t_1)W(t_2)] = \sigma^2 \min\{t_1, t_2\}$
This leads to a profound result when we consider the formal derivative $\varepsilon(t) = dW/dt$:
$E[\varepsilon(t_1)\varepsilon(t_2)] = \sigma^2 \delta(t_1 - t_2)$
[White noise](https://en.wikipedia.org/wiki/White_noise) is perfectly uncorrelated - a mathematical idealization that captures maximal randomness.
### L² Basis Representation
*See [[Wiener-Process#4. L² Basis Construction]] for complete construction*
We can construct white noise through orthogonal expansions:
$\varepsilon(t) = \sum_{n=1}^{\infty} A_n \phi_n(t)$
where $\{\phi_n\}$ is an orthonormal basis and $A_n$ are independent Gaussian random variables with $E[A_n] = 0$ and $E[A_n^2] = 1$.
---
## 3. Stochastic Differential Equations
*Introduction and examples in [[SDE-Fundamentals#1. Introduction and Motivation]]*
### The Fundamental Problem
Consider adding noise to exponential decay:
$\frac{dy}{dt} = -(\alpha + \xi(t))y$
where $\xi(t)$ is white noise. Since white noise isn't a proper function, we rewrite this using differentials:
$dy = -\alpha y \, dt - y \, dW$
### The Stochastic Integral Challenge
*The key example explored in [[Ito-Formula#1. Why the Classical Chain Rule Fails]] and [[SDE-Fundamentals#3. The Stochastic Integral]]*
When $y$ returns the Wiener process itself, we must compute $\int_0^t W(s) \, dW(s)$. The result depends on how we interpret the integral:
- **[Itô](https://en.wikipedia.org/wiki/It%C3%B4_calculus)** (left endpoint): $E[\int W \, dW] = 0$
- **[Stratonovich](https://en.wikipedia.org/wiki/Stratonovich_integral)** (midpoint): $E[\int W \, dW] = \frac{\sigma^2 t}{2}$
Classical calculus suggests $\frac{W(t)^2}{2}$, which agrees with Stratonovich. But for discrete-time limits, Itô is required.
---
## 4. Itô's Formula: The Modified Chain Rule
*Complete derivation in [[Ito-Formula#1. Why the Classical Chain Rule Fails]]*
### Why Classical Calculus Fails
The key insight: $(dW)^2 = dt$ is not negligible!
For Brownian motion:
$W(t)\Delta W = \frac{1}{2}\Delta(W^2) - \frac{1}{2}(\Delta W)^2$
In the limit:
$W \, dW = \frac{d(W^2)}{2} - \frac{dt}{2}$
The second-order differential of the Wiener process is first-order in time.
### [Itô's Formula](https://en.wikipedia.org/wiki/It%C3%B4%27s_lemma)
*See [[Ito-Formula#3. Itô's Formula: The General Rule]] for full statement and [[Ito-Formula#Heuristic Derivation]] for intuition*
For $Y = u(X,t)$ where $dX = b \, dt + \sigma \, dW$:
$dY = \left(\frac{\partial u}{\partial t} + b\frac{\partial u}{\partial x} + \frac{1}{2}\sigma^2\frac{\partial^2 u}{\partial x^2}\right)dt + \sigma\frac{\partial u}{\partial x}dW$
The extra term $\frac{1}{2}\sigma^2\frac{\partial^2 u}{\partial x^2}dt$ is the **Itô correction**.
---
## 5. Solving the Linear SDE
*Complete solution in [[SDE-Fundamentals#6. Solving the First-Order Linear SDE]]*
### The Problem
$\frac{dy}{dt} = -(\alpha + \xi(t))y, \quad y(0) = y_0$
### Solution via [Itô's Lemma](https://en.wikipedia.org/wiki/It%C3%B4%27s_lemma)
*Step-by-step derivation in [[SDE-Fundamentals#6. Solving the First-Order Linear SDE]] and [[Ito-Formula#4. Key Examples]]*
Let $u = \ln(y)$. Applying Itô's formula:
$du = -\alpha dt - dW - \frac{\sigma^2}{2}dt = -\left(\alpha + \frac{\sigma^2}{2}\right)dt - dW$
Integrating:
$y(t) = y_0 \exp\left[-\left(\alpha + \frac{\sigma^2}{2}\right)t - W(t)\right]$
### Key Insights
1. The Itô correction appears as $\sigma^2/2$ in the drift
2. Despite the noise, $E[y(t)] = y_0 e^{-\alpha t}$ - decay is preserved on average
3. In finance, this becomes geometric Brownian motion: $S(t) = S_0 \exp[\sigma W(t) + (\mu - \sigma^2/2)t]$
---
## 6. Physical Context
*Historical development in [[Wiener-Process#1. Historical Development]] and [[SDE-Fundamentals#5. Physical Interpretations]]*
### Einstein's Brownian Motion
The [Langevin equation](https://en.wikipedia.org/wiki/Langevin_equation) describes a particle subject to friction and thermal noise:
$m\ddot{x} = -\gamma \dot{x} - \nabla V(x) + \sqrt{2\gamma k_B T} \xi(t)$
The fluctuation-dissipation theorem relates noise intensity to temperature:
$\langle \eta(t)\eta(t') \rangle = 2\gamma k_B T \delta(t-t')$
### Connection to PDEs
*See [[SDE-Fundamentals#7. Connection to Partial Differential Equations]] for derivation*
The [Fokker-Planck equation](https://en.wikipedia.org/wiki/Fokker%E2%80%93Planck_equation) describes the evolution of the probability density:
$\frac{\partial p}{\partial t} = -\frac{\partial}{\partial x}[f(x)p] + \frac{1}{2}\frac{\partial^2}{\partial x^2}[g(x)^2 p]$
This bridges between individual stochastic trajectories and ensemble behavior.
---
## Key Takeaways
1. **Random walks converge to Brownian motion** with the crucial scaling $\Delta y \propto \sqrt{\Delta t}$
2. **White noise** is the formal derivative of the Wiener process - infinitely rough but mathematically tractable
3. **Stochastic integrals** depend on interpretation (Itô vs Stratonovich)
4. **Itô's formula** modifies the chain rule: $(dW)^2 = dt$ changes everything
5. **The Itô correction** $\sigma^2/2$ appears whenever we transform stochastic processes
6. **Physical systems** naturally lead to SDEs through thermal fluctuations
---
## Cross-References
- [[Random-Walks]] - Detailed development of the discrete foundation
- [[Wiener-Process]] - Mathematical properties and construction
- [[Ito-Formula]] - The modified chain rule and applications
- [[SDE-Fundamentals]] - Introduction and first-order linear SDE
---
## Video Lectures
### Parrondo's Six-Part Series
- [Part 1: Overview and Motivation](https://youtu.be/vcGpD7nZ3UM)
- [Part 2: The Wiener Process](https://youtu.be/vRsfjcXBGuI)
- [Part 3: Stochastic Integrals](https://youtu.be/9zfw_CoPYNE)
- [Part 4: Itô Calculus](https://youtu.be/h1eNpKDOa2c)
- [Part 5: Applications](https://youtu.be/7J82tcLynaU)
---
## References
### Course Materials
- MATH310 F21 Handwritten Notes: Stochastic Differential Equations
- MATH310 F24: Random Walks lecture notes
- Evans, L.C. *An Introduction to Stochastic Differential Equations*
### Historical Sources
- [Einstein, A.](https://en.wikipedia.org/wiki/Albert_Einstein) (1905). "Über die von der molekularkinetischen Theorie der Wärme geforderte Bewegung von in ruhenden Flüssigkeiten suspendierten Teilchen"
- [Itô, Kiyosi](https://en.wikipedia.org/wiki/Kiyosi_It%C3%B4) (1944). "Stochastic Integral"
- [Bachelier, L.](https://en.wikipedia.org/wiki/Louis_Bachelier) (1900). "Théorie de la spéculation"