Eva Miranda, a geometer and professor at the University of Barcelona, presents a lecture on geometric quantization — a mathematical framework that bridges classical and quantum mechanics using the geometry of symplectic and Poisson manifolds. The talk moves from the historical origins of quantization, through the geometric structures underlying classical mechanics, to the modern machinery of geometric quantization and its open problems.
Classical vs. Quantum Mechanics
Classical mechanics is governed by Newton’s laws and reformulated through Hamiltonian dynamics, where the state of a system lives on a cotangent bundle (position-momentum pairs) and evolves via Hamilton’s equations, driven by conservation of energy (the Hamiltonian).
The double-slit experiment revealed wave-particle duality — even electrons produce interference patterns — showing that classical mechanics is incomplete and that a quantum description is needed.
Quantum mechanics emerged from work by Planck, Einstein, Bohr, de Broglie, Heisenberg, and Schrödinger in the early 20th century, introducing discrete energy quanta and wave functions.
Feynman observed that “nature isn’t classical” — suggesting quantum mechanics is more fundamental — yet in practice, physicists often start with a classical Lagrangian and then quantize it, a procedure some criticize as backwards.
Quantization as a Bridge
Quantization is the process of assigning a quantum system (operators on a Hilbert space, commutators) to a classical system (functions on phase space, Poisson brackets).
Dirac’s dream: a quantization map where the Poisson bracket of two classical observables maps to the commutator of their corresponding quantum operators, satisfying {f,g} → (1/iℏ)[f̂,ĝ].
This works perfectly for linear functions (position and momentum), yielding the Heisenberg uncertainty principle from the Darboux theorem — a coincidence that initially suggested a general rule.
However, Dirac’s rule fails for general functions (e.g., quadratic functions like x² and p²), because classical multiplication is commutative while operator multiplication is not — an issue of operator ordering.
Groenwald’s no-go theorem (1946) proves it is impossible to construct a quantization map satisfying Dirac’s correspondence for all classical observables — quantization is not a functor.
Symplectic Geometry: The Language of Classical Mechanics
Classical mechanics is naturally formulated on symplectic manifolds — even-dimensional spaces equipped with a closed, non-degenerate 2-form ω (the symplectic form).
Hamilton’s equations are equivalent to the geometric equation: the contraction of the Hamiltonian vector field X_H with ω equals -dH.
The symplectic form provides a canonical isomorphism between the tangent and cotangent bundles — this is why position and momentum are naturally paired.
Darboux’s theorem: locally, all symplectic manifolds look the same — there are no local invariants (unlike Riemannian geometry, which has curvature). Locally, every symplectic manifold is a cotangent bundle with the standard form ω = Σ dpᵢ ∧ dqᵢ.
The Poisson bracket {f,g} = ω(X_f, X_g) encodes the dynamics and satisfies antisymmetry and the Jacobi identity.
Poisson Manifolds: Beyond Symplectic
Poisson manifolds generalize symplectic manifolds — the Poisson bracket can be defined even when the symplectic form is degenerate.
Example: the dual of the Lie algebra so(3) (infinitesimal rotations in 3D) carries a natural Poisson structure that cannot be symplectic (odd-dimensional), but its level sets (spheres) inherit symplectic forms — this is the Kirillov-Kostant-Souriau structure.
Casimir functions (like x₁² + x₂² + x₃² for so(3)) are conserved quantities that Poisson-commute with everything.
Allen-Weinstein splitting theorem (1983): any Poisson manifold is locally a product of a symplectic manifold and a transverse Poisson manifold vanishing at a point — a deep structural result proved surprisingly late.
Integrable Systems and Action-Angle Coordinates
An integrable system on a 2n-dimensional symplectic manifold has n independent functions (including the Hamiltonian) that Poisson-commute — half the dimension.
Examples: the harmonic oscillator, the Kepler problem (two-body), and the real and imaginary parts of holomorphic functions (via Cauchy-Riemann equations).
The Arnold-Liouville-Mineur theorem: the level sets of an integrable system are tori (Liouville tori), and in a neighborhood of a regular torus, there exist action-angle coordinates (I, θ) where the symplectic form simplifies to ω = Σ dIᵢ ∧ dθᵢ — a global version of Darboux’s theorem.
Mineur (1935) first gave this formula; he was also a member of the French Resistance who died at 55.
Toric Manifolds and Polytopes
Toric manifolds are symplectic manifolds with a Hamiltonian action of a torus of half the dimension.
The moment map of a toric action maps the manifold to a convex polytope (Delzant’s theorem).
Atiyah-Guillemin-Sternberg convexity theorem: the image of the moment map for any torus action is a convex polytope.
Examples: the 2-sphere maps to an interval; complex projective space CP² maps to a triangle.
This connects complicated symplectic geometry to linear algebra and combinatorics — polytopes encode the geometry.
Geometric Quantization: The Machinery
Geometric quantization constructs a quantum Hilbert space from a classical symplectic manifold in two steps:
Pre-quantization: requires a complex line bundle with connection whose curvature is ω — this is possible only if the cohomology class of ω is integral (the symplectic area of any surface is an integer). This is the pre-quantum line bundle.
Polarization: a Lagrangian foliation — a partition of the manifold into n-dimensional submanifolds on which ω vanishes. This reduces the 2n variables to n, cutting down the pre-quantum Hilbert space (which is too big, L²(ℝ²ⁿ)) to the correct size (L²(ℝⁿ)).
The quantum Hilbert space consists of flat sections of the line bundle — sections that are constant along the leaves of the polarization.
Bohr-Sommerfeld Quantization
A leaf of the polarization is a Bohr-Sommerfeld leaf if it admits globally defined flat sections.
For the cotangent bundle of a circle, this condition forces the momentum to be a multiple of 2πℏ — reproducing the Bohr-Sommerfeld quantization condition from early quantum theory (quantized electron orbits in the hydrogen atom).
For toric manifolds, Bohr-Sommerfeld leaves correspond to integer lattice points inside the moment polytope — quantization becomes counting lattice points.
This connects to shift cohomology (an algebraic framework) and can be corrected with half-form corrections (moving lattice points by 1/2).
Singularities and Open Problems
Integrable systems generically have singularities (e.g., the simple pendulum has elliptic singularities; the spherical pendulum has focus-focus singularities).
At singularities, the torus fibration breaks down, and naive quantization can give infinite-dimensional Hilbert spaces — unphysical results.
With collaborator Mark Hamilton, Miranda developed corrections for these singularities, showing that for reasonable singularities (elliptic, hyperbolic, focus-focus), one can recover finite, physically meaningful quantization.
B-symplectic (log-symplectic) manifolds: Poisson manifolds where the symplectic form has controlled singularities (e.g., on a sphere, the area form blows up at the equator). Here, Bohr-Sommerfeld leaves on north and south hemispheres have opposite orientations, causing infinite contributions to cancel — yielding finite quantization that matches physical expectations.
Current work with Richard Nest and Jonathan Weitsman extends geometric quantization to general Poisson manifolds.
Topological Quantum Field Theory (TQFT)
First quantization (geometric quantization) is not a functor — it depends on choices (polarization) and doesn’t always work.
Second quantization (TQFT) is a functor — it assigns vector spaces to manifolds and linear maps to cobordisms, providing a more robust mathematical framework.
TQFT connects to Miranda’s other work on Navier-Stokes equations and Turing computability (to be discussed in a future episode).
Key Takeaways
Quantization is not a unique or canonical procedure — it is “an art” requiring choices (polarization, operator ordering) and is subject to no-go theorems.
Symplectic and Poisson geometry provide the natural language for classical mechanics and the starting point for quantization.
Integrable systems are the “friendly” cases where quantization can be carried out explicitly, with action-angle coordinates and Bohr-Sommerfeld conditions.
The geometry of polytopes encodes the quantization of toric manifolds — a stunning bridge between combinatorics and quantum physics.
Open problems include: quantizing general Poisson manifolds, handling singularities robustly, and understanding whether physical systems are Turing computable (can a fluid flow simulate any computation?).
