Geodesic flow as a Hamiltonian flow

Intuition [Beginner]

A geodesic on a curved surface is the path you take when you walk straight without ever turning the wheel. On a sphere it bends back into a great circle; on a saddle it spreads apart from a neighbour; on the flat plane it is an ordinary straight line. The geodesics of a curved space record the shape of the space through the way "going straight" deforms.

There is a second story for the same paths: a particle gliding with no friction and no force, only the constraint of staying on the surface, follows a geodesic. So geodesics are also the free-motion trajectories of a frictionless point mass on the manifold. Two pictures, one set of paths.

The Hamiltonian framing puts both pictures into the same phase-space language used for every other mechanical system. Position is a point on the manifold; momentum is a covector — a way to pair tangent directions with numbers. The energy is half the squared length of the momentum, measured by the metric. Hamilton's equations turn that energy function into an evolution rule, and the resulting motion projects back down to a geodesic on the manifold. The metric becomes an energy function, the symplectic form becomes the rule for converting energy gradients into motion, and geodesic flow falls out as one specific Hamiltonian system.

Visual [Beginner]

A schematic showing a curved surface with a geodesic curve drawn on it, paired with a phase-space trajectory in the cotangent bundle above. An arrow labelled "project" connects a covector path back down to the geodesic on the surface.

The picture marks the two layers — the surface where geodesics live and the phase space where Hamilton's equations run — and the projection that ties them together.

Worked example [Beginner]

Take the flat plane ${\mathbb{R}}^2$ with the ordinary distance. The metric components are simply $g_{ij}={\delta }_{ij}$ and the inverse is the same. The energy function on phase space is

$$H(q,p)=\frac{1}{2}(p_1^2+p_2^2),$$

the standard kinetic energy with mass $1$. Pick a starting point at the origin with momentum pointing in the first coordinate direction: position $(0,0)$ and momentum $(1,0)$.

Hamilton's equations say position changes at the rate of momentum and momentum changes at the rate of minus the position-derivative of the energy. Here the energy does not depend on position, so momentum stays put: $p(t)=(1,0)$ for every time $t$. Position then advances at the constant rate $(1,0)$, giving $q(t)=(t,0)$.

Projecting the phase-space curve back to the plane discards the momentum coordinate and keeps the trajectory $q(t)=(t,0)$. That is a straight line at unit speed — exactly the geodesic of the flat metric starting at the origin and pointing east. The Hamiltonian flow on phase space and the geodesic on the manifold tell the same story.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $(M,g)$ be a smooth Riemannian manifold of dimension $n$. The cotangent bundle $T^{\ast }M$ carries the canonical Liouville one-form $\lambda$ and the canonical symplectic form $\omega =-d\lambda$ 05.02.05. In coordinates $(q^i,p_i)$ adapted to a chart on $M$, $\lambda =p_{i}d{q}^i$ and $\omega =d{q}^i\wedge d{p}_i$.

The metric $g$ on $M$ pulls back to a fibrewise inner product on $T^{\ast }M$ via the inverse metric $g^{ij}(q)$. The kinetic-energy Hamiltonian is the function

$$H:T^{\ast }M\to \mathbb{R},H(q,p)=\frac{1}{2}{g}^{ij}(q)p_i{p}_j.$$

This is half the squared norm of the covector $p$ measured by the metric pulled to the cotangent space. Sign convention: ${\iota }_{X_H}\omega =dH$, matching the conventions of 05.02.01 and 05.02.05.

The Hamiltonian vector field $X_H$ on $T^{\ast }M$ is the unique smooth vector field satisfying ${\iota }_{X_H}\omega =dH$. The cogeodesic flow is the flow of $X_H$ on $T^{\ast }M$. The geodesic flow in the classical sense is the corresponding flow on $TM$ obtained by transporting $X_H$ across the Legendre isomorphism $\mathbb{F}L:TM\to T^{\ast }M$, $(q,\dot{q})\mapsto (q,g_{ij}(q){\dot{q}}^{j}d{q}^i)$, attached to the kinetic Lagrangian $L(q,\dot{q})=\frac{1}{2}{g}_{ij}(q){\dot{q}}^i{\dot{q}}^j$.

Key theorem with proof [Intermediate+]

Theorem (geodesic flow = cogeodesic flow). *Let $(M,g)$ be a Riemannian manifold and $H(q,p)=\frac{1}{2}{g}^{ij}(q)p_i{p}_j$ the kinetic-energy Hamiltonian on $T^{\ast }M$. The integral curves of $X_H$ on $T^{\ast }M$ project under $\pi :T^{\ast }M\to M$ to geodesics of $g$. Conversely, every geodesic of $g$ is the projection of a unique integral curve of $X_H$ specified by its initial position and initial momentum.*

Proof. Hamilton's equations associated with $H$ in canonical coordinates read

$${\dot{q}}^i=\frac{\partial H}{\partial p_i}=g^{ij}(q)p_j,{\dot{p}}_i=-\frac{\partial H}{\partial q^i}=-\frac{1}{2}{\partial }_i{g}^{jk}(q)p_j{p}_k.$$

The first equation is the Legendre relation $p_i=g_{ij}{\dot{q}}^j$ inverted, identifying momentum as the metric-flat of velocity. Differentiating it in $t$,

$${\ddot{q}}^i={\partial }_{\ell }{g}^{ij}{\dot{q}}^{\ell }{p}_j+g^{ij}{\dot{p}}_j.$$

Substitute $p_j=g_{j\ell }{\dot{q}}^{\ell }$ in the first term and ${\dot{p}}_j=-\frac{1}{2}{\partial }_j{g}^{k\ell }{p}_k{p}_{\ell }$ in the second, then convert all $p$'s to $\dot{q}$'s using $p_a=g_{ab}{\dot{q}}^b$:

$${\ddot{q}}^i=({\partial }_{\ell }{g}^{ij})g_{jm}{\dot{q}}^{\ell }{\dot{q}}^m-\frac{1}{2}{g}^{ij}({\partial }_j{g}^{k\ell })g_{ka}{g}_{\ell b}{\dot{q}}^a{\dot{q}}^b.$$

Apply the inverse-metric derivative identity ${\partial }_a{g}^{ij}=-g^{ic}({\partial }_a{g}_{cd})g^{dj}$, valid because $g^{ij}{g}_{jk}={\delta }_k^i$ implies ${\partial }_a{g}^{ij}{g}_{jk}+g^{ij}{\partial }_a{g}_{jk}=0$. After substitution and contraction the right-hand side collapses to

$${\ddot{q}}^i=-{\Gamma }_{ab}^i(q){\dot{q}}^a{\dot{q}}^b,{\Gamma }_{ab}^i=\frac{1}{2}{g}^{ic}({\partial }_a{g}_{cb}+{\partial }_b{g}_{ca}-{\partial }_c{g}_{ab}),$$

which is the geodesic equation in the Levi-Civita connection. The projected curve $q(t)=\pi (\gamma (t))$ therefore satisfies the geodesic equation, and conversely every geodesic admits a horizontal lift to $T^{\ast }M$ by the same Legendre relation, unique once initial momentum is fixed. $\square$

Bridge. The identification here builds toward 05.10.01 (contact manifold), where the level set $\{H=\frac{1}{2}\}$ — the unit cotangent bundle $S(T^{\ast }M)$ — appears again as the canonical contact manifold whose Reeb vector field is exactly the geodesic spray. Putting these together with the foliation picture of 05.02.04, the foundational reason the geodesic flow is integrable on highly symmetric manifolds (round spheres, flat tori, ellipsoids) is exactly that the Hamiltonian framing exposes Killing vector fields as Poisson-commuting integrals; the central insight is that metric symmetry is the same data as Hamiltonian symmetry once one passes to $T^{\ast }M$. Geodesic flow therefore appears again in 05.09.01 (KAM theorem) as the unperturbed integrable system for KAM-style perturbation analysis, and the bridge between the Riemannian-geometry and Hamiltonian-mechanics layers is that single energy function $H=\frac{1}{2}{g}^{ij}{p}_i{p}_j$.

Exercises [Intermediate+]

Lean formalization [Intermediate+]

[object Promise]

The proof in Mathlib would proceed by establishing the Legendre transform as a smooth bundle isomorphism $TM\cong T^{\ast }M$, transporting the Hamiltonian vector field $X_H$ across it, and identifying the resulting vector field on $TM$ as the geodesic spray. Each step is independently a candidate Mathlib contribution, listed in the lean_mathlib_gap field above.

Advanced results [Master]

The Hamiltonian framing of geodesic flow on $T^{\ast }M$ is the natural setting for several deeper structural results.

Restriction to the unit cotangent bundle. The level set $S(T^{\ast }M)=\{(q,p):g^{ij}{p}_i{p}_j=1\}=\{H=\frac{1}{2}\}$ is a smooth $(2n-1)$-dimensional submanifold. The Liouville one-form $\lambda$ restricts to a contact form $\alpha$: $\alpha \wedge (d\alpha {)}^{n-1}$ is a volume form on $S(T^{\ast }M)$ because $\omega =-d\lambda$ is non-degenerate and the level set is transverse to $X_H$. The Reeb vector field of $\alpha$ — defined by ${\iota }_R\alpha =1$ and ${\iota }_{R}d\alpha =0$ — coincides on $S(T^{\ast }M)$ with the restricted Hamiltonian vector field $X_H{|}_{S(T^{\ast }M)}$. So the geodesic flow on $S(T^{\ast }M)$ is, intrinsically, a Reeb flow of the canonical contact form on the unit cotangent bundle ^{[Cannas da Silva §1]}.

Conservation laws and Killing vector fields. Energy $H$ is conserved because the Hamiltonian is autonomous: $\dot{H}=\omega (X_H,X_H)=0$. More generally, every Killing vector field $K$ on $(M,g)$ — every infinitesimal isometry — induces a conserved quantity for the geodesic flow. The cotangent lift $\hat{K}$ on $T^{\ast }M$ is the Hamiltonian vector field of the function $f_K(q,p)=p(K(q))$, and $\{f_K,H\}=0$ on the nose because $K$ preserves $g$. On $S^n$ with the round metric, the rotation group $SO(n+1)$ supplies $\left(\genfrac{}{}{0ex}{}{n+1}{2}\right)$ independent Killing fields, far more than the $n$ needed for Liouville-Arnold integrability; the geodesic flow on the round sphere is therefore superintegrable, and every geodesic is a great circle. On a flat torus ${\mathbb{R}}^n/{\mathbb{Z}}^n$ the constant vector fields are Killing and supply $n$ commuting integrals, so the geodesic flow is Liouville integrable with the action-angle coordinates $(p,q\mathrm{mod}1)$ 05.02.04.

Anosov geodesic flow on negative curvature. On a closed Riemannian manifold with strictly negative sectional curvature — most cleanly, a compact hyperbolic surface $\Sigma ={\mathbb{H}}^2/\Gamma$ — the geodesic flow on $S(T^{\ast }\Sigma )$ is Anosov: the tangent bundle of the flow direction admits an invariant splitting into uniformly contracting and uniformly expanding subspaces. The flow is exponentially mixing with respect to the Liouville measure on $S(T^{\ast }\Sigma )$, and orbits are distributed equidistributionally. This is the bridge between symplectic / Riemannian geometry and ergodic theory: the same Hamiltonian flow that integrates explicitly on $S^n$ becomes the canonical example of a chaotic dynamical system on $\Sigma$ ^{[Klingenberg Ch. 3]}.

Liouville-Arnold integrable cases. Beyond constant curvature, the Jacobi-integrable geodesic flow on the triaxial ellipsoid $\{x^2/a^2+y^2/b^2+z^2/c^2=1\}$ in ${\mathbb{R}}^3$ is the classical example of an integrable geodesic flow with no underlying isometry-group transitivity. Jacobi 1838 Vorlesungen über Dynamik showed that elliptic-coordinate separation gives three independent Poisson-commuting integrals on $T^{\ast }$(ellipsoid) — energy plus two extra — and the geodesics close up or wrap densely on Liouville tori according to rationality of the frequency ratios. Generalisations: the Neumann system, Stäckel-separable metrics, and the Manakov integrable cases on $T^{\ast }SO(n)$.

Length spectrum and Laplace spectrum. The set of lengths of closed geodesics on $(M,g)$ — the length spectrum — is the trace of the geodesic flow at periodic orbits. The Selberg trace formula on a compact hyperbolic surface gives an exact identity between the length spectrum and the Laplace spectrum of $-{\Delta }_g$. The marked length spectrum (lengths together with free-homotopy classes) determines the metric on negatively curved closed surfaces (Otal, Croke, 1990). The inverse spectral question popularised by Mark Kac ("Can one hear the shape of a drum?") is the analogous question for the Laplace spectrum and is delicate even on flat tori (Milnor 1964 produced a 16-dimensional counterexample).

Symplectic capacities of unit codisk bundles. The unit codisk bundle $D(T^{\ast }M)=\{H\le \frac{1}{2}\}$ is a symplectic manifold with boundary $S(T^{\ast }M)$. Its symplectic capacities — Gromov width, Hofer-Zehnder capacity, Ekeland-Hofer capacities — are dynamical invariants of $(M,g)$ defined through the periodic orbits of the geodesic flow on the boundary. Floer-theoretic computations relate them to the volume and the systole. This connects geodesic flow to 05.07.02 symplectic capacity, 05.07.01 non-squeezing, and the symplectic homology of cotangent bundles (Viterbo 1996, Salamon-Weber 2006: $S{H}^{\ast }(T^{\ast }M)\cong H_{\ast }(\Lambda M)$, the homology of the free loop space).

Maupertuis-Jacobi reformulation. For a mechanical Hamiltonian $H=\frac{1}{2}{g}^{ij}{p}_i{p}_j+V(q)$ with potential $V$, on each energy level $\{H=E\}$ where $E-V>0$ the trajectories project to geodesics of the Jacobi metric $\tilde{g}=(E-V)g$, parametrised by $\tilde{g}$-arc length. This converts a mechanical system with potential into a pure geodesic-flow problem at the cost of a metric depending on $E$. The reformulation is due to Maupertuis 1744 in its variational form ("principle of least action") and Jacobi 1837 in its differential-geometric form; it is the bridge from Newtonian mechanics with conservative forces to Riemannian-geometry techniques.

Synthesis. The data of a Riemannian metric and the data of a kinetic-energy Hamiltonian on the cotangent bundle are equivalent: each determines the other via the Legendre transform $\mathbb{F}L:TM\to T^{\ast }M$, and the dynamics agree under projection. Read in one direction, the construction puts geodesic flow into the symplectic toolkit, where Liouville volume preservation, action-angle coordinates, KAM perturbation theory, and Floer homology become available techniques. Read in the opposite direction, every kinetic Hamiltonian on a cotangent bundle is the geodesic flow of some metric — its Hamiltonian comes from a quadratic form on covectors, and any positive-definite quadratic form defines a Riemannian metric. The bridge between the analytic equation $H=\frac{1}{2}{g}^{ij}{p}_i{p}_j$ and the geometric conclusion that integral curves project to geodesics is the foundational reason the symplectic, contact, and Riemannian pictures of free-particle motion coincide. Putting these together, one sees that the moduli space of "kinetic Hamiltonians on $T^{\ast }M$" is the moduli space of Riemannian metrics on $M$, and the central insight is that metric symmetry, Hamiltonian symmetry, and contact symmetry are the same data viewed from three angles.

Full proof set [Master]

Theorem (geodesic flow = cogeodesic flow). Restated: the integral curves of $X_H$ for $H=\frac{1}{2}{g}^{ij}{p}_i{p}_j$ on $T^{\ast }M$ project to geodesics of $g$ on $M$, and this correspondence is bijective on initial conditions specified by point and tangent vector via the Legendre transform $\mathbb{F}L:TM\to T^{\ast }M$.

Proof. Given in the Intermediate Key theorem section above. The substitution chain ${\dot{q}}^i=g^{ij}{p}_j$, ${\dot{p}}_i=-\frac{1}{2}{\partial }_i{g}^{jk}{p}_j{p}_k$, combined with the inverse-metric derivative identity, produces ${\ddot{q}}^i+{\Gamma }_{ab}^i{\dot{q}}^a{\dot{q}}^b=0$, the Levi-Civita geodesic equation. The bijection on initial conditions is the assertion that $\mathbb{F}L$ is a diffeomorphism, which holds because the kinetic Lagrangian is hyper-regular (the Hessian ${\partial }^{2}L/\partial {\dot{q}}^i\partial {\dot{q}}^j=g_{ij}$ is positive-definite). $\square$

Theorem (Reeb = geodesic spray on unit cotangent bundle). *Let $\alpha =\lambda {|}_{S(T^{\ast }M)}$ be the restriction of the canonical Liouville form to the unit cotangent bundle. Then $\alpha$ is a contact form, and its Reeb vector field is the restriction of $X_H$.*

Proof. Non-degeneracy of $\omega =-d\lambda$ on $T^{\ast }M$ together with transversality of $X_H$ to $S(T^{\ast }M)$ — which holds because $X_H(H)=0$ but $X_H\ne 0$ off the zero section — implies $\alpha \wedge (d\alpha {)}^{n-1}$ is a volume form on $S(T^{\ast }M)$. So $\alpha$ is contact. For the Reeb identification, on $S(T^{\ast }M)$ one has ${\iota }_{X_H}\lambda =p_i\cdot d{q}^i(X_H)=p_i{g}^{ij}{p}_j=2H=1$, and ${\iota }_{X_H}d\lambda =-{\iota }_{X_H}\omega =-dH$. Restrict to $S(T^{\ast }M)$ and observe $TS(T^{\ast }M)=\ker dH$, so ${\iota }_{X_H}d\alpha {|}_{S(T^{\ast }M)}=0$. The pair $({\iota }_R\alpha =1,{\iota }_{R}d\alpha =0)$ uniquely determines the Reeb vector field, and $X_H{|}_{S(T^{\ast }M)}$ satisfies both. $\square$

Theorem (Killing → conserved quantity). Let $K$ be a Killing vector field on $(M,g)$, with cotangent lift function $f_K(q,p)=p(K(q))$. Then $\{f_K,H\}=0$, and $f_K$ is conserved along the cogeodesic flow.

Proof. In coordinates $f_K=K^i(q)p_i$. Compute the Poisson bracket:

$$\{f_K,H\}=\frac{\partial f_K}{\partial q^j}\frac{\partial H}{\partial p_j}-\frac{\partial f_K}{\partial p_j}\frac{\partial H}{\partial q^j}={\partial }_j{K}^i{p}_i{g}^{jk}{p}_k-K^j\cdot \frac{1}{2}{\partial }_j{g}^{ik}{p}_i{p}_k.$$

Symmetrise the first term in $(i,k)$: the coefficient of $p_i{p}_k$ is $\frac{1}{2}(g^{jk}{\partial }_j{K}^i+g^{ji}{\partial }_j{K}^k)-\frac{1}{2}{K}^j{\partial }_j{g}^{ik}$. The Killing equation ${\nabla }_i{K}_j+{\nabla }_j{K}_i=0$ (equivalently ${\mathcal{L}}_{K}g=0$) translates in the inverse metric to ${\mathcal{L}}_K{g}^{-1}=0$, which in coordinates is exactly $g^{jk}{\partial }_j{K}^i+g^{ji}{\partial }_j{K}^k-K^j{\partial }_j{g}^{ik}=0$. So the bracket vanishes. By Hamilton's equations ${\dot{f}}_K=\{f_K,H\}=0$. $\square$

Theorem (Maupertuis-Jacobi). On the energy level $\{H=E\}$ of the mechanical Hamiltonian $H=\frac{1}{2}{g}^{ij}{p}_i{p}_j+V(q)$, the trajectories of $X_H$ coincide as unparametrised curves with the trajectories of the cogeodesic flow of the Jacobi metric $\tilde{g}=(E-V)g$, restricted to $\{\tilde{H}=1\}$.

Proof. Solution given in Exercise 7. The crucial identity is that on $\{H=E\}$ one has $g^{ij}{p}_i{p}_j=2(E-V)$, and ${\tilde{g}}^{ij}=(E-V{)}^{-1}{g}^{ij}$, so ${\tilde{g}}^{ij}{p}_i{p}_j=2$ on this set, identifying it with $\{\tilde{H}=1\}$. The vector fields $X_H$ and $X_{\tilde{H}}$ are parallel on the common level set, so trajectories coincide as point sets. Reparametrisation by $\tilde{g}$-arc length identifies $\tilde{g}$-geodesics with the original mechanical trajectories. $\square$

Connections [Master]

• Cotangent bundle 05.02.05. The Hamiltonian framing of geodesic flow is built directly on the canonical symplectic structure of $T^{\ast }M$. The Liouville one-form $\lambda$, the symplectic form $\omega =-d\lambda$, and the projection $\pi :T^{\ast }M\to M$ all play load-bearing roles in the construction, and the kinetic-energy Hamiltonian is the simplest non-zero function on $T^{\ast }M$ that requires the full Riemannian data.

• Hamiltonian vector field 05.02.01. Geodesic flow is the canonical example of a Hamiltonian vector field on a cotangent bundle. The general theory of $X_H$ — energy conservation, Liouville volume preservation, Cartan-formula identities — specialises to give the standard properties of geodesic flow without separate calculation.

• Action-angle coordinates 05.02.04. On a manifold whose isometry group acts transitively (round spheres, flat tori, ellipsoids treated by Jacobi), the geodesic flow is Liouville integrable, and action-angle coordinates linearise the flow on each Liouville torus. This is the direct route from Riemannian symmetry to Hamiltonian integrability.

• Contact manifold 05.10.01. The unit cotangent bundle $S(T^{\ast }M)$ is the canonical contact manifold attached to $(M,g)$, and the geodesic flow is its Reeb flow. The contact-geometry / Riemannian-geometry bridge runs through this identification.

• KAM theorem 05.09.01. A small Hamiltonian perturbation of an integrable geodesic flow — for example, a small symmetric perturbation of the round metric on $S^n$ — produces a Hamiltonian system to which KAM applies: most invariant Liouville tori survive, and the perturbed geodesic flow remains quasi-periodic on a positive-measure set. Geodesic flow on highly symmetric metrics is the canonical "unperturbed system" for KAM-style analysis.

• Symplectic reduction 05.04.02. When a Lie group $G$ acts on $(M,g)$ by isometries, the cotangent lift acts on $T^{\ast }M$ by symplectomorphisms preserving $H$. The Marsden-Weinstein reduced spaces are then phase spaces for the reduced geodesic flow on the quotient orbifold. Examples: geodesic flow on $S^2\cong SO(3)/SO(2)$ as reduction of the free-particle flow on $T^{\ast }SO(3)$.

• Symplectic capacity 05.07.02 and non-squeezing 05.07.01. Symplectic capacities of unit codisk bundles of $(M,g)$ are quantitative invariants computed from periodic orbits of the geodesic flow. Floer-theoretic computations of capacities of $D(T^{\ast }M)$ recover information about the Riemannian metric — volume, systole, length spectrum.

Historical & philosophical context [Master]

Carl Gustav Jacob Jacobi gave the first identification of geodesic flow with Hamiltonian flow in 1837, in Note sur l'intégration des équations différentielles de la dynamique (Comptes Rendus Acad. Sci. Paris) ^{[Jacobi 1837]}. The 1837 note announced the principle: trajectories of a conservative mechanical system on a configuration manifold reduce, after Maupertuis reparametrisation, to geodesics of a metric depending on the energy level. The full proofs appeared in Jacobi's Vorlesungen über Dynamik (1842-43 lectures, published posthumously 1866), where elliptic-coordinate separation of variables produced the integration of the geodesic flow on the triaxial ellipsoid — the first integrable geodesic flow established by techniques other than direct symmetry reduction.

The modern coordinate-free framing — Hamiltonian as a function on $T^{\ast }M$, Liouville form $\lambda$, symplectic form $\omega =-d\lambda$, kinetic energy $H=\frac{1}{2}{g}^{ij}{p}_i{p}_j$ — was crystallised in the symplectic-geometry programme of Souriau, Smale, Abraham, and Marsden in the 1960s and 1970s. Abraham-Marsden Foundations of Mechanics (1978) ^{[Abraham-Marsden]} §3.7 gives the canonical exposition of the cogeodesic flow on $T^{\ast }M$ and its Legendre relation to the geodesic spray on $TM$, in the language used by every subsequent textbook.

Klingenberg's Riemannian Geometry (1982) ^{[Klingenberg]} develops the geodesic spray and the symplectic structure of $TM$ in the Riemannian-geometry tradition, providing the translation between the symplectic-mechanics and the Riemannian-geometry literatures. Cannas da Silva's Lectures on Symplectic Geometry ^{[Cannas da Silva §1]} places the cogeodesic flow as the headline example of a cotangent-bundle Hamiltonian system in the modern symplectic-geometry pedagogical canon.

The connection between geodesic flow and ergodic theory was developed by Hadamard 1898 (negative curvature on surfaces of finite type produces dense geodesics), Hopf 1939 (mixing of the geodesic flow on closed surfaces of negative curvature), and Anosov 1967 (uniform hyperbolicity in arbitrary dimension), establishing geodesic flow as the canonical example of a chaotic dynamical system. The Selberg trace formula (1956) and the Otal-Croke length-spectrum rigidity theorems (1990) place the closed-geodesic spectrum at the centre of inverse spectral geometry.

Bibliography [Master]

[object Promise]