Unlike ordinary differential equations, which involve functions of a single independent variable, partial differential equations (PDEs) describe functions of multiple variables. This shift opens up a richer world of phenomena to model, from the flow of heat in a metal plate to the vibration of a guitar string.

We'll focus on the most common and foundational type: second-order linear PDEs. For a function $u(x, t)$, the general form of such an equation is:

The character of a PDE, including the type of physical phenomena it can describe and the methods used to solve it, is determined primarily by the coefficients of its highest-order derivatives: $A$, $B$, and $C$. This leads to a powerful classification system.

Just as you can classify conic sections (ellipses, parabolas, hyperbolas) using a discriminant, we can classify second-order linear PDEs. The expression $B^2 - 4AC$ sorts these equations into three fundamental families. This isn't just a mathematical convenience; it separates the equations based on how they propagate information.

Let's look at the canonical examples:

Hyperbolic: The wave equation, $u_{tt} - c^2 u_{xx} = 0$. Here, $A=-c^2$, $B=0$, and $C=1$. The discriminant is $0^2 - 4(-c^2)(1) = 4c^2 > 0$. Hyperbolic equations describe phenomena with a finite speed of propagation, like waves traveling along a string. A disturbance at one point is only felt later at another.

Parabolic: The heat equation, $u_t - \alpha u_{xx} = 0$. Here, $A=-\alpha$, $B=0$, and $C=0$. The discriminant is $0^2 - 4(-\alpha)(0) = 0$. Parabolic equations model diffusion processes. A disturbance, like a point source of heat, is felt instantly everywhere, but its magnitude decays with distance.

Elliptic: The Laplace equations, $u_{xx} + u_{yy} = 0$. Here, $A=1$, $B=0$, and $C=1$. The discriminant is $0^2 - 4(1)(1) = -4 < 0$. Elliptic equations describe steady-state or equilibrium systems. The solution at any point depends on the values along the entire boundary of the domain, implying that information is transmitted "infinitely fast" to reach a balance.

Not every PDE you can write down has a meaningful solution. A problem is considered well-posed if it meets three criteria defined by the mathematician Jacques Hadamard:

1. Existence: A solution must exist.
2. Uniqueness: There must be only one solution for a given set of conditions.
3. Stability: The solution must depend continuously on the initial and boundary conditions. A small change in the input data should only lead to a small change in the solution.

This stability criterion is crucial for physical modeling. If a tiny fluctuation in initial measurements could cause a wildly different outcome, our model would be useless for making predictions.

To ensure well-posedness, we must provide appropriate initial and boundary conditions. The type of PDE dictates the conditions needed. For example, a hyperbolic equation like the wave equation typically requires two initial conditions (initial position and initial velocity) and boundary conditions. An elliptic equation, describing equilibrium, only needs boundary conditions.

A key feature of linear PDEs is the principle of superpositions. If $u_1$ and $u_2$ are both solutions to a linear homogeneous PDE, then any linear combination $c_1 u_1 + c_2 u_2$ is also a solution. This principle is the backbone of many solution techniques, like Fourier series, where we build up a complex solution by summing simpler ones.

Understanding this classification and the requirements for a well-posed problem provides the essential framework for tackling PDEs. It tells you what kind of behavior to expect from a solution and guides your choice of analytical and numerical methods.