Partial Differential Equations
Partial differential equations (PDEs) are equations that relate a function of multiple variables to its partial derivatives. They are the native language of systems whose state depends on both time and space — heat diffusing through a conductor, waves propagating through a medium, fluids flowing around an obstacle, patterns forming on a chemical substrate. Where ordinary differential equations describe how a single quantity evolves, PDEs describe how a field evolves: how the temperature at every point changes, how the velocity at every point changes, how the concentration at every point changes. The shift from ODEs to PDEs is the shift from particle thinking to field thinking, and it is one of the most consequential conceptual moves in the history of science.
The defining feature of a PDE is that its unknown is not a number or a finite vector but a function — an infinite-dimensional object. The equation constrains how this function changes in different directions. A heat equation relates the time derivative of temperature to the sum of its second spatial derivatives, encoding the physical law that heat flows from hot to cold at a rate proportional to the curvature of the temperature field. A wave equation relates the second time derivative of displacement to the second spatial derivative, encoding the law that acceleration is proportional to local curvature. These are not merely analogies to physical laws. They are the laws, expressed in the only language precise enough to handle spatial variation.
Classification and the Nature of Information Flow
PDEs are classified by their highest-order derivatives into three families, and this classification is not bureaucratic — it determines how information moves through the system.
Elliptic equations (the Laplace equation, Poisson equation) describe equilibrium states. They have no time variable. Information at any point depends on information at all other points, instantaneously. A charge distribution determines the electric potential everywhere, and the potential at any point is the average of the potential on any surrounding sphere. This global-to-local coupling is the mathematical signature of systems that have settled into balance.
Parabolic equations (the heat equation, diffusion equations) describe smoothing processes. Information flows forward in time but diffuses backward in space — the future temperature at a point depends on the present temperature in a neighborhood, but not on the precise far-field details. The smoothing is irreversible: diffusion destroys fine structure. The Fourier transform reveals this directly — high-frequency spatial modes decay exponentially fast, leaving only the slowly varying components.
Hyperbolic equations (the wave equation, equations of fluid dynamics) describe propagating disturbances. Information travels along characteristic curves — trajectories in spacetime along which the PDE reduces to an ordinary differential equation. A sound wave reaches your ear because the acoustic equation has characteristics that propagate the initial disturbance forward at finite speed. This is not a detail of the physics. It is a structural property of hyperbolic PDEs: finite propagation speed, preservation of discontinuities, and the possibility of shock formation.
The classification maps onto the broader systems distinction between equilibrium, dissipation, and propagation. It is the same trichotomy that appears in network dynamics (consensus, diffusion, and epidemic spreading), in control theory (steady-state regulation, damping, and feedforward response), and in thermodynamics (equilibrium, irreversible processes, and reversible transformations). The mathematics of PDEs is not a specialty. It is a general theory of how local rules generate global patterns.
PDEs as Emergence Engines
The deepest reason PDEs matter for systems theory is that they are engines of emergence. A PDE does not merely describe an already-existing pattern. It specifies the local rules from which the pattern self-organizes.
Consider the reaction-diffusion systems that produce Turing patterns. Each point in space contains chemicals that react with each other and diffuse to neighboring points. The PDE specifies nothing about stripes, spots, or labyrinthine structures. It specifies only reaction rates and diffusion coefficients. Yet under certain conditions — when an activator diffuses slowly and an inhibitor diffuses quickly — the uniform state becomes unstable, and the system spontaneously organizes into stable spatial patterns. The PDE is the genotype. The pattern is the phenotype. The environment is the initial condition and the boundary constraints.
The Navier-Stokes equations exhibit a different kind of emergence: turbulence. The equations are deterministic and local. A smooth initial condition evolves, and at sufficient Reynolds number, the solution develops fine-scale vortices, energy cascades across scales, and effectively unpredictable behavior. The unpredictability is not injected from outside. It is generated by the nonlinearity of the PDE itself. Local interactions produce a global state that cannot be reconstructed from local measurements — a form of emergence in which the whole is not merely greater than the sum of the parts but qualitatively different from any finite sample of the parts.
The Analytic-Synthetic Tension
PDEs sit at a methodological crossroads that mirrors a broader tension in science. The analytic tradition — Fourier, Hilbert, Sobolev — treats PDEs as problems to be solved, emphasizing existence, uniqueness, and regularity of solutions. The synthetic tradition — Turing, Prigogine, reaction-diffusion biologists — treats PDEs as models to be simulated, emphasizing pattern formation, bifurcation, and dynamical behavior.
These traditions have different standards of understanding. An analyst knows a PDE is understood when a theorem guarantees a unique solution in a Sobolev space. A synthesist knows a PDE is understood when a simulation reproduces the striped shell of a mollusk or the spiral waves of a Belousov-Zhabotinsky reaction. Neither standard is complete. Theorems without physical interpretation are formal exercises. Simulations without analytical backing are numerology.
The productive convergence occurs in weak solutions — solutions defined not by pointwise satisfaction of the equation but by satisfaction of an integral identity. Weak solutions capture shock waves, vortex sheets, and other singular structures that classical solutions cannot describe. They are the mathematical formalization of the physical intuition that what matters is not the value of the solution at every point but its integrated effect on test functions. This is the same intuition that underlies quantum mechanics (observables as operators on states) and spectral graph theory (eigenvectors as generalized coordinates).
The belief that PDEs are a branch of analysis rather than a general theory of how local interactions produce global structure is a disciplinary accident. Every field that studies extended systems — physics, biology, ecology, economics, sociology — eventually arrives at PDEs, because extended systems are what PDEs describe. The mathematicians did not invent partial differential equations. They found them already operating in the world.