Application of Partial Differential Equations (PDEs) in Option Pricing

Visit the Mathema Option Pricing System, supporting FX options and structured product pricing and valuation!

Introduction

Partial Differential Equations (PDEs) are a crucial tool in option pricing theory. The derivation of the Black-Scholes option pricing formula is fundamentally based on a PDE, which models the dynamics of asset price changes and describes the relationship between option prices and variables such as time. Compared to numerical methods (e.g., binomial tree models and Monte Carlo simulations), PDEs provide a foundational framework for establishing option pricing theory and can be used to solve prices for various types of options.

This article introduces the basic principles of PDEs in option pricing, derives the Black-Scholes PDE, and discusses how to solve PDEs using numerical methods (e.g., explicit and implicit finite difference methods). Finally, we demonstrate the application of PDEs in pricing a European call option through a simple example.

1. Basic Principles of PDEs in Option Pricing

1.1 Stochastic Process Modeling for Option Pricing

In financial markets, the price $S$ of the underlying asset is assumed to follow a Geometric Brownian Motion (GBM), which can be expressed as:

dS = \mu S \, dt + \sigma S \, dW_t

where:

$\mu$ : The expected return of the underlying asset;
$\sigma$ : The volatility of the underlying asset;
$W_t$ : Standard Brownian motion.

The option price $V(S, t)$ is a function of the underlying asset price $S$ and time $t$ . Our goal is to describe the evolution of $V(S, t)$ using this stochastic process.

1.2 Risk-Neutral Measure and the Black-Scholes PDE

In an arbitrage-free market, the prices of all assets can be expressed under the risk-neutral measure. In the risk-neutral world, the drift rate $\mu$ of the underlying asset price is replaced by the risk-free interest rate $r$ . Thus, the stochastic process of the underlying asset price becomes:

dS = r S \, dt + \sigma S \, dW_t

According to Itô's Lemma, the total differential of the option price $V(S, t)$ can be expressed as:

dV = \frac{\partial V}{\partial t} dt + \frac{\partial V}{\partial S} dS + \frac{1}{2} \frac{\partial^2 V}{\partial S^2} \sigma^2 S^2 dt

Substituting $dS$ into the above equation and using the no-arbitrage condition of a replicating portfolio, the Black-Scholes PDE can be derived:

\frac{\partial V}{\partial t} + \frac{1}{2} \sigma^2 S^2 \frac{\partial^2 V}{\partial S^2} + r S \frac{\partial V}{\partial S} - r V = 0

1.3 Boundary Conditions

The Black-Scholes PDE is an elliptic partial differential equation, and its solution requires boundary conditions:

Terminal Condition: At the option's maturity, the option's value is determined by its payoff function. For example, for a European call option:
$V(S, T) = \max(S - K, 0)$
where $K$ is the strike price of the option.
Boundary Conditions: For extreme asset prices, boundary values need to be set manually. For example:
- $S \to 0$ : When the underlying asset price approaches zero, the value of a call option approaches zero, i.e., $V(0, t) = 0$ .
- $S \to \infty$ : When the underlying asset price approaches infinity, the value of a call option approaches $S - K$ , i.e., $V(S, t) \approx S - K$ .

2. Methods for Solving PDEs

The Black-Scholes PDE is typically not solvable analytically and requires numerical methods. Below are two commonly used numerical methods:

2.1 Explicit Finite Difference Method

The explicit finite difference method discretizes time $t$ and price $S$ to transform the PDE into a finite difference equation. Assume:

$t$ is divided into $M$ time steps, with each step having a time interval of $\Delta t = \frac{T}{M}$ ;
$S$ is divided into $N$ price points, with each step having a price interval of $\Delta S$ .

The discretized Black-Scholes PDE can be expressed as:

V_{i,j} = a_j V_{i+1,j-1} + b_j V_{i+1,j} + c_j V_{i+1,j+1}

where:

$V_{i,j}$ represents the option value at the $i$ -th time step and $j$ -th price point;
The coefficients $a_j, b_j, c_j$ depend on the asset price, volatility, and risk-free interest rate.

The explicit finite difference method progresses step-by-step from the terminal condition backward. However, due to its poor stability, it is generally only suitable for calculations with small time steps.

2.2 Implicit Finite Difference Method

The implicit finite difference method solves the option value at each time step by constructing a system of linear equations:

a_j V_{i-1,j-1} + b_j V_{i-1,j} + c_j V_{i-1,j+1} = V_{i,j}

Compared to the explicit method, the implicit method is more stable and allows for larger time steps, but each step requires solving a system of linear equations, resulting in higher computational complexity.

2.3 Crank-Nicolson Method

The Crank-Nicolson method combines the advantages of explicit and implicit methods by averaging the time steps to improve accuracy and stability:

\frac{1}{2} \left[ \text{Explicit Finite Difference Formula} + \text{Implicit Finite Difference Formula} \right]

It is a standard method for solving the Black-Scholes PDE, offering both high accuracy and stability.

3. Example: Pricing a European Call Option Using PDEs

3.1 Problem Description

Suppose we want to price a European call option with the following parameters:

Current underlying asset price $S_0 = 50$ ;
Strike price $K = 52$ ;
Volatility $\sigma = 0.2$ ;
Risk-free interest rate $r = 5\%$ ;
Time to maturity $T = 1$ year.

3.2 Using the Explicit Finite Difference Method

Discretization Parameters:
- Number of time steps $M = 100$ , number of price steps $N = 100$ ;
- Time interval $\Delta t = \frac{T}{M} = 0.01$ ;
- Price interval $\Delta S = \frac{S_{\text{max}}}{N} = 1$ , assuming $S_{\text{max}} = 100$ .
Initialize Terminal Condition:
- $V(S, T) = \max(S - K, 0)$ .
Recursively Calculate Option Value:
- Use the explicit finite difference formula to step backward from $t = T$ to $t = 0$ .

3.3 Results

Using the explicit finite difference method, the theoretical price of the European call option is 4.48 (consistent with the Black-Scholes closed-form solution).

4. Further Applications and In-Depth Research on PDE Models

The application of Partial Differential Equation (PDE) models in financial derivative pricing and risk management extends far beyond the classical Black-Scholes framework. With the development of financial markets, researchers continue to explore the use of PDE models in pricing more complex derivatives, risk management, multi-asset modeling, and other areas, proposing many improved methods and extended models. The following sections will provide a detailed introduction to the further applications of PDE models and related cutting-edge research.

4.1 Further Applications of PDE Models

4.1.1 Pricing American Options

American options allow exercise at any time before maturity, so their pricing must account for the possibility of early exercise. This makes American option pricing a free-boundary problem, requiring the simultaneous determination of the option value and the optimal exercise boundary.

PDE Pricing Methods:

Free Boundary Condition:
At each time step, the optimal exercise boundary satisfies:
$V(S_b(t), t) = \max(S_b(t) - K, 0)$
where $S_b(t)$ is the optimal exercise boundary.
Numerical Techniques:
- Penalty Method: Introduces a penalty term into the PDE to transform the free-boundary problem into a fixed-boundary problem.
- Variational Inequality Method: Reformulates the problem as a linear complementarity problem (LCP).
- Front-Fixing Method: Directly tracks the position of the optimal exercise boundary.

Applications:

Stock option pricing (American options are more popular in high-volatility markets).
Fixed-income instruments (e.g., callable bonds).

4.1.2 Pricing Barrier Options

Barrier options are path-dependent options whose value depends on whether the underlying asset price reaches a preset barrier level. Pricing barrier options requires introducing additional boundary conditions.

PDE Pricing Methods:

Barrier Conditions:
- Knock-In Options: The option value is zero before the barrier is hit; it becomes a standard option after being hit.
- Knock-Out Options: The option expires immediately if the barrier is hit.
Boundary Treatment:
Reflective or absorbing boundary conditions are set at the barrier level. For example, for a knock-out call option at the barrier price $S_b$ :
$V(S_b, t) = 0$

Applications:

Foreign exchange options (FX Barrier Options).
Structured products (e.g., yield-enhancing products with barrier triggers).

4.1.3 Pricing Asian Options

Asian options are path-dependent options whose value depends on the average price of the underlying asset. Due to the dynamic nature of the cumulative average, pricing Asian options requires introducing an additional state variable.

PDE Pricing Methods:

State Extension:
Introduce the cumulative average $A$ as an additional dimension, constructing a two-dimensional PDE:
$\frac{\partial V}{\partial t} + r S \frac{\partial V}{\partial S} + \frac{1}{2} \sigma^2 S^2 \frac{\partial^2 V}{\partial S^2} + \frac{\partial V}{\partial A} = r V$
Numerical Methods:
- Use finite difference methods (FDM) or Monte Carlo simulations combined with PDE solving.
- Special handling of the cumulative variable is required for large time steps.

Applications:

Energy and commodity options (average price hedging).
Foreign exchange derivatives (hedging volatility risk).

4.1.4 Stochastic Volatility Models

The classical Black-Scholes model assumes that the underlying asset's volatility $\sigma$ is constant, but in real markets, volatility is often stochastic. Stochastic volatility models (e.g., the Heston model) introduce a stochastic volatility process, constructing higher-dimensional PDEs.

PDE Pricing Methods:

Heston Model:
Assume volatility follows a mean-reverting process:
$dv = \kappa (\theta - v) dt + \xi \sqrt{v} dW_2$
where $v = \sigma^2$ is the variance.
The corresponding PDE is two-dimensional:
$\frac{\partial V}{\partial t} + \frac{1}{2} S^2 v \frac{\partial^2 V}{\partial S^2} + \rho \xi S v \frac{\partial^2 V}{\partial S \partial v} + \frac{1}{2} \xi^2 v \frac{\partial^2 V}{\partial v^2} + r S \frac{\partial V}{\partial S} + \kappa (\theta - v) \frac{\partial V}{\partial v} - r V = 0$
Numerical Methods:
- Use ADI (Alternating Direction Implicit) methods to solve multi-dimensional PDEs.
- Combine Monte Carlo simulations with PDEs to determine implied volatility.

Applications:

Volatility arbitrage strategies (e.g., VIX options).
Volatility swaps and related derivatives.

4.1.5 Multi-Asset Option Pricing

Multi-asset options (e.g., basket options and correlation options) involve multiple underlying assets, requiring the introduction of multiple price dimensions and the construction of higher-dimensional PDEs.

PDE Pricing Methods:

Multi-Dimensional PDEs:
For two underlying assets $S_1$ and $S_2$ , their prices follow a joint geometric Brownian motion: $dS_1 = \mu_1 S_1 dt + \sigma_1 S_1 dW_1, \quad dS_2 = \mu_2 S_2 dt + \sigma_2 S_2 dW_2$ The correlation between them is described by $\rho$ . The corresponding PDE is: $\frac{\partial V}{\partial t} + \frac{1}{2} \sigma_1^2 S_1^2 \frac{\partial^2 V}{\partial S_1^2} + \frac{1}{2} \sigma_2^2 S_2^2 \frac{\partial^2 V}{\partial S_2^2} + \rho \sigma_1 \sigma_2 S_1 S_2 \frac{\partial^2 V}{\partial S_1 \partial S_2} + r S_1 \frac{\partial V}{\partial S_1} + r S_2 \frac{\partial V}{\partial S_2} - r V = 0$

Numerical Methods:

Use finite difference methods or finite element methods (FEM) to solve high-dimensional PDEs.
For high-dimensional problems, dimensionality reduction techniques (e.g., principal component analysis) can be applied.

Applications:

Basket options (e.g., index options).
Correlation asset derivatives (e.g., cross-currency options).

4.2 In-Depth Research on PDE Models

4.2.1 Efficient Numerical Methods

As the dimensionality of PDEs and the complexity of problems increase, researchers continue to develop efficient numerical methods for solving PDEs:

Fast Fourier Transform (FFT): Used to solve integral problems in pricing formulas.
Sparse Grid Methods: Reduce computational complexity in high-dimensional problems.
Machine Learning and PDEs: Use deep learning (e.g., PINNs, Physics-Informed Neural Networks) to solve complex PDEs.

4.2.2 Nonlinear PDEs

In some cases, option pricing problems involve nonlinear PDEs. For example:

Transaction Cost Models: Introduce nonlinear terms into the PDE when transaction costs are considered.
Liquidity Risk Pricing: Illiquidity may lead to nonlinear pricing models.

4.2.3 Stochastic Partial Differential Equations (SPDEs)

When stochastic parameters (e.g., volatility, interest rates) exist in the market, classical PDEs are extended to Stochastic Partial Differential Equations (SPDEs). Solving SPDEs typically requires combining Monte Carlo methods with numerical PDE techniques.

4.2.4 Non-Uniform Time Grids

In certain scenarios (e.g., short-term high-volatility options or barrier options), standard time grids may fail to capture rapid price changes, prompting researchers to develop non-uniform time grid methods.

5. Finite Difference Methods (FDM) and Partial Differential Equations (PDEs)

The relationship between Finite Difference Methods (FDM) and Partial Differential Equations (PDEs) can be summarized as: FDM is a numerical method for solving PDEs. PDEs are mathematical models describing dynamic relationships between continuous variables, while FDM discretizes PDEs, transforming continuous problems into discrete problems that can be solved on a computer.

To clarify the relationship between the two, the following sections will elaborate on the connection and differences between FDM and PDEs from theoretical, methodological, and application perspectives.

5.1 Theoretical Relationship

5.1.1 The Core Role of PDEs

PDEs are mathematical models that describe dynamic processes and are widely used in physics, engineering, and finance. In option pricing, the Black-Scholes equation is a classic two-dimensional PDE that describes the evolution of the option price $V(S, t)$ with respect to time $t$ and asset price $S$ :

\frac{\partial V}{\partial t} + \frac{1}{2} \sigma^2 S^2 \frac{\partial^2 V}{\partial S^2} + r S \frac{\partial V}{\partial S} - r V = 0

Such PDEs typically lack analytical solutions (especially for complex derivatives), necessitating numerical methods for approximate solutions.

5.1.2 FDM as a Discretization Tool for PDEs

FDM is a numerical method for solving PDEs. Its core idea is:

Discretization Grid: Divide continuous time $t$ and asset price $S$ into discrete grid points.
Finite Difference Approximation: Replace partial derivatives in the PDE with finite differences. For example:
- Time derivative: $\frac{\partial V}{\partial t} \approx \frac{V(t_{i+1}, S) - V(t_i, S)}{\Delta t}$
- First-order spatial derivative: $\frac{\partial V}{\partial S} \approx \frac{V(S_{j+1}, t) - V(S_{j-1}, t)}{2 \Delta S}$
- Second-order spatial derivative: $\frac{\partial^2 V}{\partial S^2} \approx \frac{V(S_{j+1}, t) - 2V(S_j, t) + V(S_{j-1}, t)}{\Delta S^2}$
Discretized Equations: Transform the PDE into a system of algebraic equations and solve it on the grid.

Thus, FDM is a tool that uses discretization to convert PDEs from continuous descriptions to discrete representations.

5.2 Methodological Comparison

5.2.1 Solving PDEs

PDE solving methods can be divided into two categories:

Analytical Methods:
- Suitable for simple PDEs and boundary conditions.
- For example, the closed-form solution of the Black-Scholes equation can be derived using transformations and integration.
- Advantages: Exact solutions; disadvantages: Only applicable to a few simple problems.
Numerical Methods:
- Suitable for complex PDEs or boundary conditions.
- Main methods include:
  - Finite Difference Methods (FDM)
  - Finite Element Methods (FEM)
  - Spectral Methods
  - Monte Carlo Simulations

FDM is a classical numerical method for solving PDEs, particularly suitable for problems on regular grids.

5.2.2 Specific Applications of FDM

FDM is a numerical method primarily used to discretize PDEs and solve the resulting algebraic equations. Based on the time-stepping approach, FDM includes three core methods:

Explicit Methods:
- Use current time-step values to compute the next time-step values.
- Simple to implement but less stable.
Implicit Methods:
- Require solving a system of linear equations, where current time-step values depend on the entire time step.
- More stable but computationally intensive.
Crank-Nicolson Method:
- A compromise between explicit and implicit methods, offering second-order accuracy and good stability.
- Suitable for financial derivative pricing.

5.3 Application Relationship

5.3.1 Solving the Black-Scholes PDE with FDM

In option pricing, FDM is one of the primary numerical methods for solving PDEs. The specific steps for applying FDM to the Black-Scholes equation are as follows:

Construct the Grid:
- Discretize time $T$ into $M$ time steps with interval $\Delta t = \frac{T}{M}$ .
- Discretize asset price $S$ into $N$ price steps with interval $\Delta S$ .
Terminal Condition:
- At $t = T$ , initialize the grid based on the option's payoff function. For example, for a European call option: $V(S, T) = \max(S - K, 0)$
Boundary Conditions:
- $S \to 0$ : Option value approaches zero.
- $S \to \infty$ : Call option value approaches $S - K e^{-r(T-t)}$ .
Discretize the PDE:
- Use FDM to transform the Black-Scholes equation into a system of algebraic equations.
- For example, the explicit method's recursive formula is: $V_i^j = a_j V_{i-1}^{j-1} + b_j V_{i-1}^j + c_j V_{i-1}^{j+1}$ where $a_j, b_j, c_j$ are discretization coefficients.
Step-by-Step Backward Calculation:
- Starting from $t = T$ , step backward to $t = 0$ to obtain the option price.

5.3.2 Pricing American Options

Pricing American options is a free-boundary problem, requiring the comparison of holding value and exercise value at each time step. The steps for applying FDM are as follows:

Discretize the Black-Scholes PDE:
- Use implicit or Crank-Nicolson methods to discretize the PDE.
Exercise Condition:
- Update grid points at each time step: $V_i^j = \max(V_i^j, S_j - K)$
Solve the Linear Complementarity Problem:
- Combine the stability of implicit methods to solve step-by-step.

5.3.3 Extensions to High-Dimensional Problems

FDM can also be used to solve high-dimensional PDEs, such as:

Two-Asset Options:
$\frac{\partial V}{\partial t} + \frac{1}{2} \sigma_1^2 S_1^2 \frac{\partial^2 V}{\partial S_1^2} + \frac{1}{2} \sigma_2^2 S_2^2 \frac{\partial^2 V}{\partial S_2^2} + \rho \sigma_1 \sigma_2 S_1 S_2 \frac{\partial^2 V}{\partial S_1 \partial S_2} + r S_1 \frac{\partial V}{\partial S_1} + r S_2 \frac{\partial V}{\partial S_2} - r V = 0$
- FDM uses a two-dimensional grid for discretization.
- Cross-term $\frac{\partial^2 V}{\partial S_1 \partial S_2}$ requires special handling.
Stochastic Volatility Models:
- Introduce volatility as an additional state variable, constructing three-dimensional PDEs.

5.4 Comparison of FDM with Other Numerical Methods

Method	Applicable Scenarios	Advantages	Disadvantages
FDM	Solving PDEs on regular grids	Simple to implement, suitable for low-dimensional problems	Less efficient for high-dimensional problems
FEM	Solving PDEs on irregular grids	Suitable for complex boundary conditions and geometries	Complex implementation, high computational cost
Monte Carlo	Path-dependent options, stochastic models	Suitable for high-dimensional problems, easy to parallelize	Slow convergence, high estimation variance
Spectral Methods	Smooth problems	High accuracy	Only suitable for smooth problems, difficult to handle complex boundaries

Spectral Methods are a relatively niche but efficient numerical approach for solving partial differential equations (PDEs) in computational fields. Although they are more commonly used in computational fluid dynamics (CFD), weather modeling, and physics, they are less frequently mentioned in financial engineering. However, with the continuous development of financial mathematics, spectral methods are gradually being applied to financial derivative pricing, characteristic function methods, and numerical solutions for high-dimensional problems due to their high accuracy and fast convergence.

5.5 Summary

Relationship Between FDM and PDEs:

PDEs are mathematical models of problems, while FDM is a numerical method for solving PDEs.
FDM discretizes PDEs, transforming continuous problems into discrete problems and solving them step-by-step on a grid.

Applications of FDM:

FDM is widely used in option pricing, particularly for European options, American options, and path-dependent options.
As problem complexity increases, FDM is also extended to high-dimensional problems and stochastic volatility models.

Although FDM is versatile and flexible for solving PDEs, its computational complexity increases with higher dimensions. In the future, combining FDM with other numerical methods (e.g., FEM and Monte Carlo simulations) and machine learning techniques will further expand its applications and efficiency.

6. References

Black, F., & Scholes, M. (1973). The Pricing of Options and Corporate Liabilities. Journal of Political Economy.
- Introduces the classical Black-Scholes model.
Wilmott, P., Howison, S., & Dewynne, J. (1995). The Mathematics of Financial Derivatives. Cambridge University Press.
- Provides detailed derivations of PDEs and numerical methods.
Heston, S. L. (1993). A Closed-Form Solution for Options with Stochastic Volatility with Applications to Bond and Currency Options. Review of Financial Studies.
- Introduces the closed-form solution and PDE for stochastic volatility models.
Tavella, D., & Randall, C. (2000). Pricing Financial Instruments: The Finite Difference Method. Wiley.
- Detailed applications of finite difference methods in pricing financial instruments.
Han, J., Jentzen, A., & E, W. (2018). Solving High-Dimensional Partial Differential Equations Using Deep Learning. Proceedings of the National Academy of Sciences (PNAS).
- Explores the application of deep learning in solving high-dimensional PDEs.

Conclusion

PDE models provide a powerful mathematical framework for pricing financial derivatives, extending beyond the classical Black-Scholes European options to include American options, path-dependent options, multi-asset options, and stochastic volatility models. With advancements in numerical methods and the integration of machine learning techniques, PDE models will find broader applications in high-dimensional and nonlinear problems, further unlocking their potential in financial engineering.