Legendre transformation

Legendre transformation
Name	Legendre transformation
Field	Mathematical analysis; Theoretical physics
Introduced	19th century
Notable	Adrien-Marie Legendre; Joseph Fourier; Henri Poincaré

Legendre transformation is a mathematical operation that converts one real-valued function into another by exchanging dependent and independent variables via a convexity-related conjugation. It plays a central role in classical mechanics, thermodynamics, convex analysis, and partial differential equations, linking formulations such as the Hamiltonian and Lagrangian descriptions and relating thermodynamic potentials like internal energy and free energy.

Definition

Given a smooth convex function f on a convex domain, the Legendre transformation produces a new function g defined on the space of slopes (or momenta) by taking the supremum of linear functionals minus f. In analytic terms, for x in the domain of f and p representing the gradient variable, g(p) = sup_x (⟨p, x⟩ − f(x)). This construction is coordinate-free and relies on convexity to ensure the supremum is attained and unique; the gradient map x ↦ ∇f(x) establishes a diffeomorphism under strict convexity. The operation is involutive under suitable regularity: applying the transform twice recovers the convex closure of the original function.

Mathematical properties

The transform converts pointwise addition and scalar multiplication into operations dual under convex conjugation and exchanges minima with maxima in variational problems. For C^2 strictly convex f, the map ∇f implements a Legendre diffeomorphism between primal and dual spaces, and the second derivative matrices are inverse to one another: D^2g(p) = [D^2f(x)]^{-1} with p = ∇f(x). The transformation respects lower semicontinuity and convexity: the Legendre transform of a proper closed convex function is proper closed convex. It intertwines subdifferentials: the subgradient of f at x corresponds to the inverse image under the subgradient of the transform. In finite dimensions, Fenchel–Moreau theorem characterizes biconjugation and identifies the double transform with the convex lower semicontinuous envelope. Under smoothness, the Legendre transform preserves analytic regularity and converts Hamilton–Jacobi type PDEs into dual PDE forms; symplectic geometry interprets the transform as a change of Lagrangian submanifold coordinates in cotangent bundles.

Examples and applications

Classical mechanics employs the transform to pass between Lagrangian and Hamiltonian formalisms: given a kinetic-plus-potential Lagrangian depending on generalized velocities, the Legendre map sends velocities to canonical momenta and yields the Hamiltonian function generating Hamilton's equations. Thermodynamics uses the construction to relate thermodynamic potentials: internal energy as a function of entropy and volume is Legendre-transformed to obtain Helmholtz free energy with temperature as the conjugate variable, and similarly Gibbs free energy arises when transforming with respect to pressure and chemical potential. In optics, Fermat-type variational principles and geometrical optics connect to Hamiltonian optics via Legendre conjugation of optical path length. In large deviations theory and statistical mechanics, rate functions and cumulant generating functions are linked by Legendre transforms, underpinning the equivalence of ensembles. Calculus of variations and classical field theory utilize the transform to derive canonical Hamiltonian densities from Lagrangian densities. Examples include quadratic forms where the transform produces the dual quadratic with inverse coefficient matrix, and the entropy function on probability simplices whose transform yields log-partition functions central to exponential families in statistics.

Generalizations and Legendre–Fenchel transform

The Legendre–Fenchel transform (also called convex conjugate) extends the construction to non-smooth, extended-real-valued functions on topological vector spaces. It replaces gradients with subdifferentials and supremum operations in infinite-dimensional Banach or Hilbert spaces, enabling duality theory in optimization and variational calculus. Fenchel duality provides conditions for strong duality in convex optimization problems encountered in control theory and signal processing. In symplectic and contact geometry, generating functions of Lagrangian submanifolds generalize classical Legendre transforms to produce Maslov indices and relate to Floer homology in the study of Hamiltonian diffeomorphisms. Other extensions include the convex conjugate on manifolds, the use in Young measures for relaxation of variational problems, and tropical or idempotent analogues appearing in optimization over semirings.

Computational methods and inversion

Computing Legendre transforms analytically is straightforward for smooth strictly convex functions but requires optimization methods in general. For differentiable f, inversion uses the gradient map: solve p = ∇f(x) for x and substitute to evaluate g(p) = ⟨p, x⟩ − f(x). For non-differentiable or high-dimensional settings, convex optimization algorithms such as interior-point methods, projected gradient descent, and proximal algorithms compute the supremum defining the conjugate. Discrete transforms appear in numerical convex analysis where Fenchel conjugates are approximated via convex hull computations or epi-convergence techniques; in signal processing, fast convex conjugation exploits problem structure for efficient implementation. Regularization and smoothing (Moreau envelope) aid numerical stability and provide differentiable approximations whose transforms are easier to invert. In many applied problems, dual optimization yields computational advantages by converting constrained primal problems into unconstrained dual formulations amenable to parallelization.

Historical context

The transform traces roots to 19th-century analysis and variational calculus, with early contributions by Adrien-Marie Legendre in the context of elliptic integrals and convexity ideas later formalized by Joseph Fourier and Henri Poincaré. The systematic convex-conjugate viewpoint was developed in the 20th century through the work of Ernst Fischer, Werner Fenchel, John von Neumann, and Léonard Moreau, establishing Fenchel duality and proximal calculus. The adoption of the transform in physics accelerated with the formalization of analytical mechanics by William Rowan Hamilton and Carl Gustav Jacob Jacobi, and its central role in thermodynamics crystallized with Ludwig Boltzmann and Josiah Willard Gibbs. Contemporary research connects the transform to optimization theory, symplectic topology, and statistical mechanics, reflecting its interdisciplinary influence across mathematics and physics.

Category:Mathematical transforms