Dirichlet energy

The Dirichlet energy of a function is a measure of its local variation — the integral (or sum) of the squared gradient over its domain. In continuous settings, for a function u on a domain Ω, the Dirichlet energy is E[u] = ∫_Ω |∇u|² dV. On a graph, for a function f on vertices, it is E[f] = Σ_{(i,j)∈E} (f(i) − f(j))². In both cases, the energy quantifies how much the function changes from point to point, or from vertex to vertex.

The Dirichlet energy is the central quantity in the Poincaré inequality and the log-Sobolev inequality, where it controls the global behavior of a function through its local variation. In physics, the Dirichlet energy appears in the Lagrangian of the wave equation and the heat equation; minimizing it subject to boundary conditions yields harmonic functions. In network science, the Dirichlet energy of an opinion vector determines how quickly consensus protocols converge: low energy means opinions are already similar; high energy means disagreement is entrenched.

The concept generalizes to Dirichlet forms — bilinear functionals that abstract the properties of the Dirichlet energy to general state spaces. Dirichlet forms are the analytic foundation of the theory of symmetric Markov processes and provide the bridge between potential theory and probability.

The Dirichlet energy is often treated as a technical quantity in variational calculus. This misses its physical meaning: it is the dissipation rate of a system, the rate at which gradients smooth out and differences equilibrate. A system with high Dirichlet energy is a system far from equilibrium, full of tension and structure. A system with zero Dirichlet energy is a system at rest, uniform and uninteresting. The Dirichlet energy is therefore the measure of a system's capacity to be interesting.