Cédric Villani — Fields Medal 2010

1. Optimal Transport & the Wasserstein Geometry

Monge’s 1781 problem — how to transport a pile of soil from a source distribution to a target distribution at minimum total cost — was reformulated by Kantorovich in the 1940s as a linear-programming relaxation. Villani’s contribution was to develop the geometric structure of the resulting space of probability measures equipped with the Wasserstein distance\(W_p\):

\[ W_p(\mu, \nu) \;=\; \left(\inf_{\pi \in \Pi(\mu, \nu)} \int |x - y|^p \, d\pi(x, y)\right)^{1/p} \]

where \(\Pi(\mu,\nu)\) is the set of joint distributions with prescribed marginals. Villani, with Otto, Sturm, McCann and Lott, established the Lott–Sturm–Villani synthetic Ricci-curvature lower bound: a metric measure space has “Ricci curvature \(\geq K\)” in the synthetic sense iff certain entropy functionals are \(K\)-convex along Wasserstein geodesics. This breakthrough lets Riemannian-geometric tools transfer to non-smooth and singular spaces — metric trees, sub-Riemannian manifolds, and probabilistic limits of graphs.

The downstream impact has been enormous: optimal-transport-based loss functions in machine learning (Wasserstein GANs, Sinkhorn divergences), shape registration in computational anatomy, regularised entropy methods in mathematical economics, and the Otto–Villani interpretation of PDE flows as Wasserstein gradient flows.

2. Boltzmann’s H-theorem & the Cercignani Conjecture

Boltzmann’s 1872 H-theorem proves that for a dilute gas, the entropy functional

\[ H(f) \;=\; \int f \log f \, d\mathbf{v}\,d\mathbf{x}, \qquad \frac{dH}{dt} \leq 0 \]

decreases monotonically toward equilibrium. The qualitative result was clear; the quantitative question — how fast? — was open for over a century. Cercignani’s conjecture proposed that entropy production satisfies a linear lower bound:\(D(f) \geq C\,(H(f) - H_{\mathrm{eq}})\), which would imply exponential decay to equilibrium.

Villani’s 2003 paper with Desvillettes (Inventiones Math.) and the follow-up 2009 work proved the conjecture in dimension 3 for soft potentials, with quantitative entropy-production bounds. The technique combined hypocoercivity theory — a generalisation of standard coercivity to non-self-adjoint evolution operators where dissipation comes from the interaction between transport and collisions — with sharp functional inequalities (logarithmic Sobolev, modified Poincaré).

The proof established a programme — hypocoercivity — that has since been applied to Fokker–Planck equations, kinetic Vlasov flows, and stochastic-Liouville equations far beyond the original Boltzmann setting.

3. Nonlinear Landau Damping

Lev Landau in 1946 showed that small perturbations of a homogeneous plasma equilibrium decay exponentially — even in the collisionless Vlasov–Poisson regime where one would expect only wave persistence. The mechanism, “Landau damping,” relies on phase mixing: high-velocity-gradient eigenmodes of the linearised Vlasov operator interfere out of constructive coherence.

The linear case was settled in the 1940s. The nonlinear case — whether the full Vlasov–Poisson system decays the same way for sufficiently smooth perturbations — remained open for sixty years. Mouhot & Villani’s 2010 paper (the work cited in Villani’s Fields Medal) gave the proof: Gevrey-class smooth perturbations of a stable homogeneous equilibrium decay exponentially in time.

The proof technique was an enormous Newton iteration (the “Nash–Moser with losses” scheme) that controlled echoes — resonant interactions that would naïvely cause the perturbation series to diverge. The estimate is:

\[ \|f(t,\cdot,\cdot) - f^\infty\|_{\mathcal{G}^\nu} \;\leq\; C\,e^{-\lambda t} \quad \text{for some } \lambda > 0 \]

with \(\mathcal{G}^\nu\) a Gevrey class of regularity. Subsequent work (Bedrossian, Masmoudi, Faou et al.) extended the result to lower regularity, larger amplitudes, and 2D Euler / SQG analogues.

4. Logarithmic Sobolev & Functional Inequalities

A continuous theme through Villani’s work is the use of functional inequalities — logarithmic Sobolev (Gross 1975), modified Poincaré, Talagrand inequality — as quantitative tools for proving relaxation rates. The Otto–Villani 2000 theorem proved that log-Sobolev implies Talagrand’s\(T_2\) inequality, structurally tying entropy contraction to Wasserstein geometry. The same toolbox underwrites both the Boltzmann hypocoercivity proof and the modern probabilistic analysis of Markov chain mixing times.

5. Books, Lectures & Public Mathematics

Villani is unusually prolific as an expositor:

• Topics in Optimal Transportation(2003, AMS) — the textbook that introduced a generation of mathematicians to the field.
• Optimal Transport: Old and New(2009, Springer Grundlehren) — the 1000-page reference, including the synthetic Ricci curvature programme, gradient-flow interpretation, and infinite-dimensional generalisations.
• Hypocoercivity (2009, AMS Memoirs) — the monograph laying out the technical framework that won the Fields Medal.
• Théorème Vivant(2012, English: Birth of a Theorem) — a memoir of the Mouhot–Villani Landau-damping proof, written for a general audience.

From 2017 to 2022 he served as a member of the French National Assembly (Essonne's 5th constituency, La République En Marche!) — an unusual political turn for a Fields medallist that gave him a public platform for advocating science funding and AI ethics. After leaving politics he resumed his research and outreach activity.

6. Influence Across Adjacent Fields

• Machine learning: Wasserstein GANs (Arjovsky 2017), Sinkhorn divergences (Cuturi 2013), entropic optimal transport, neural-network OT solvers.
• Probability: Wasserstein-distance lower bounds for Markov chain mixing times, geometric ergodicity proofs, displacement-convex stochastic differential equations.
• Geometry: Lott–Sturm–Villani synthetic Ricci curvature programme, RCD spaces, applications to Alexandrov geometry and Ricci flow.
• PDE: hypocoercivity as a unifying framework for kinetic relaxation, applied to Boltzmann, Vlasov, Fokker–Planck, and BGK models.