Manifolds modeled on locally convex spaces

1 Introduction

This blueprint describes the mathematics formalized in the Lean library lc-manifolds: manifolds modeled on locally convex spaces, their tangent bundles, vector fields and flows, differential forms, submanifolds and immersions, vector bundles, connections and geodesics, metrics, manifolds of mappings, analytic, complex and symplectic manifolds, and a collection of examples from functional analysis and from groups of diffeomorphisms. Every definition and theorem below is linked to its Lean counterpart, and the dependency graph shows how the results build on each other. All the statements are proved in Lean, with no sorry, and the main theorems depend only on the standard axioms of Lean.

The setting

The differential calculus is the one of Michal and Bastiani, formalized in the companion library lc-analysis [ LCA ] : a map between open subsets of locally convex spaces is \(C^{1}\) if its directional derivatives exist and depend continuously on the point and on the direction, jointly. This is the calculus of infinite-dimensional Lie theory, as presented by Neeb [ Nee06 ] and by Glöckner and Neeb [ GN26 ] , which we follow for the foundations. For the geometry of Banach manifolds (sprays, connections, Riemannian metrics) we follow Lang [ Lan99 ] and the book [ Lar26 ] , and for metric geometry [ Lar19 ] .

A manifold modeled on \(E\) is a charted space whose transition maps are smooth diffeomorphisms between open subsets of \(E\). Mathlib [ mC20 ] accepts an arbitrary topological model space in its notion of charted space, so we do not need a separate theory of manifolds: we only add the groupoid of Michal–Bastiani diffeomorphisms.

Conventions

Three conventions are used throughout, and we state them once and for all.

  • Model spaces need not be complete. The model space \(E\) is a real locally convex Hausdorff space, with no completeness assumption, following Glöckner and Neeb [ GN26 ] . We call a manifold Banach, Hilbert or Fréchet only when its model is complete; a manifold modeled on a normed space that is not complete is not a Banach manifold. Whenever a result needs completeness, it is part of the statement.

  • No inverse function theorem beyond Banach spaces. Outside Banach spaces there is no inverse function theorem, and there are no existence theorems for ordinary differential equations. On general models we avoid both: submanifolds are given by slice charts, examples are built with explicit charts, and geodesics are characterized, not constructed. Flows, geodesics, the exponential map and the Frobenius and Darboux theorems are proved for Banach models only.

  • Infinite-dimensional subtleties are flagged. Notions that coincide in finite dimensions often split apart here: immersions and weak immersions, embedded and initial submanifolds, strong and weak metrics, continuity on the tangent bundle and continuity along curves. We keep a separate name for each notion, and each splitting is illustrated by a formalized counterexample whenever we have one.

Metrics: two questions, two words

Chapter 13 uses two pairs of words that should not be confused. For a field of Finsler seminorms \(\mu \), the question is how \(\mu _p\) depends on \(p\). A Finsler metric is continuous on the tangent bundle; this is what the term means without further qualification. A Finsler length structure is only continuous along \(C^{1}\) curves: it is what one needs to measure lengths, and quotient metrics on homogeneous spaces are of this kind. Every Finsler metric is a Finsler length structure, and the converse fails already on \(\mathbb {R}^2\). For a field of bilinear forms \(g\), the question is which topology \(g_x\) induces on the tangent space: a strong metric induces the topology of the model, a weak one may induce a coarser topology. Weak Riemannian metrics are Finsler metrics; Lorentzian and pseudo-Riemannian metrics, of any index, are not.

Organization

The chapters follow the folders of the library.

  • Foundations. Manifolds and smooth maps (Chapter 2), tangent bundles (Chapter 3), vector fields and the Frobenius theorem (Chapter 4), flows (Chapter 5), manifolds with rough boundary (Chapter 6), and submanifolds and immersions (Chapter 7).

  • Forms and bundles. Differential forms (Chapter 8), vector bundles (Chapter 9), connections (Chapter 10), and sprays and geodesics (Chapter 11).

  • Metrics. Metric geometry of length spaces (Chapter 12), and Finsler, Riemannian, Lorentzian and pseudo-Riemannian metrics on manifolds (Chapter 13).

  • Further structures. Local additions and manifolds of mappings (Chapter 14), analytic manifolds (Chapter 15), complex manifolds (Chapter 16) and symplectic manifolds (Chapter 17).

  • Examples (Chapter 18) and counterexamples (Chapter 19).

How the library was written

@glaroton set the mathematical program, chose the framework and the references, fixed the design constraints and reviewed the results. The Lean proofs were written by the AI assistant Claude (Anthropic), working under his direction. Correctness does not rest on trusting either of them: Lean’s kernel [ dMU21 ] checks every proof, and the statements are kept close to the literature so that they can be compared with it directly.