Contact manifold

Intuition [Beginner]

A contact manifold is the odd-dimensional cousin of a symplectic manifold. Where a symplectic manifold pairs every direction with a unique partner direction in even dimension, a contact manifold sits in odd dimension and carries a hyperplane at every point — a slice of the tangent space one dimension below the whole — that twists from point to point in the most violent way the geometry of forms allows. The hyperplanes never line up to give a foliation. They refuse to integrate.

A useful image: think of stacking pancakes. Most ways of laying a hyperplane field onto a manifold give pancake stacks — the planes fit together into surfaces. The contact condition says the hyperplane field is the opposite of pancake-stacked. No matter how small a region you look at, the planes spiral.

Contact geometry is where the geometry of first-order differential equations lives, where the Reeb flow that builds toward modern Floer theory lives, and where the boundary of every symplectic manifold-with-boundary appears again in the next chapter as a contact manifold by restriction.

Visual [Beginner]

A schematic of ${\mathbb{R}}^3$ with the standard contact distribution: a planar field that rotates around the $z$-axis, twisting clockwise as you walk along $z$ and tilting as you walk away from the axis. The picture shows the family of tangent planes pinned to lattice points on a coarse grid, each tilted by a different amount.

The key picture is the twist: the hyperplane field never settles into a foliation, no matter how local you go.

Worked example [Beginner]

Take the manifold ${\mathbb{R}}^3$ with coordinates $(x,y,z)$ and the one-form $\alpha =dz-ydx$. At each point this one-form picks out a plane: the plane on which $\alpha$ vanishes. At the origin, the plane is $\{dz=0\}$, so it is the $(x,y)$-plane. At the point $(0,1,0)$, the plane is $\{dz=dx\}$, tilted by 45 degrees.

Walk along the $y$-axis and watch the plane tilt: at $y=0$ the plane is horizontal, at $y=1$ it has tilted 45 degrees, at $y=2$ it has tilted further. The planes form a spiralling family. They never form a stack of surfaces because they tilt at every step.

The contact condition is the statement that this tilting is as strong as it can be: the three-form $\alpha \wedge d\alpha =dz\wedge dy\wedge dx$ is nowhere zero. This three-form measures the failure of the planes to stack, and the contact condition demands that the failure is total.

What this tells us: contact geometry is geometry where hyperplane fields are forced to be maximally twisted, and one explicit one-form on ${\mathbb{R}}^3$ already shows the picture in full.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $M$ be a smooth manifold of dimension $2n+1$. A contact structure on $M$ is a smooth hyperplane distribution $\xi \subset TM$ — a smooth field of codimension-one subspaces — such that, locally around every point, $\xi =\ker \alpha$ for a smooth one-form $\alpha$ satisfying

$$\alpha \wedge (d\alpha {)}^n\ne 0$$

at every point of the local domain. The form $\alpha$ is called a (local) contact form for $\xi$. A pair $(M,\xi )$ is called a contact manifold.

The contact structure is co-orientable if there exists a single smooth $\alpha$ defined on all of $M$ with $\xi =\ker \alpha$. Two such global forms differ by multiplication by a nowhere-vanishing smooth function: if $\xi =\ker \alpha =\ker {\alpha }^{\prime }$, then ${\alpha }^{\prime }=f\alpha$ for some $f:M\to \mathbb{R}\setminus \{0\}$. The contact-condition top-form transforms by ${\alpha }^{\prime }\wedge (d{\alpha }^{\prime }{)}^n=f^{n+1}\alpha \wedge (d\alpha {)}^n$, so the sign of $\alpha \wedge (d\alpha {)}^n$ depends on the orientation choice when $n$ is odd.

When $\xi$ is non-co-orientable, the choice of $\alpha$ exists only up to local sign; the global object is a section of the line bundle $TM/\xi \to M$, and the contact condition is the requirement that this section be nowhere zero in the appropriate jet sense.

For a co-oriented $(M,\xi =\ker \alpha )$ the Reeb vector field $R_{\alpha }$ is defined by the two equations

$${\iota }_{R_{\alpha }}\alpha =1,{\iota }_{R_{\alpha }}d\alpha =0.$$

Uniqueness and existence of $R_{\alpha }$ follow from the fact that $d\alpha$ restricted to $\xi$ is a symplectic form on $\xi$ (a vector space of dimension $2n$), so the kernel of $d\alpha$ in $TM$ is exactly one-dimensional; the normalisation $\alpha (R_{\alpha })=1$ then pins $R_{\alpha }$ down.

Standard examples.

• ${\mathbb{R}}^{2n+1}$ with coordinates $(x_1,y_1,\dots ,x_n,y_n,z)$ and contact form $\alpha =dz-{\sum }_i{y}_{i}d{x}_i$. The Reeb field is $R_{\alpha }={\partial }_z$.
• The unit sphere $S^{2n+1}\subset {\mathbb{C}}^{n+1}$ with $\alpha$ the restriction of $\frac{i}{2}{\sum }_j(z_{j}d{\bar{z}}_j-{\bar{z}}_{j}d{z}_j)$. The Reeb flow is the Hopf flow.
• The unit cotangent bundle $S(T^{\ast }N)$ of a Riemannian manifold $N$, with the canonical contact structure inherited from the Liouville one-form on $T^{\ast }N$. The Reeb flow is the geodesic flow.
• The first jet bundle $J^1(N,\mathbb{R})=T^{\ast }N\times \mathbb{R}$ with $\alpha =dz-{\theta }_N$, where ${\theta }_N$ is the Liouville one-form pulled up from $T^{\ast }N$.

A submanifold $L\subset M$ of dimension $n$ is Legendrian if $T_{p}L\subset {\xi }_p$ for every $p\in L$; this is the maximum dimension a submanifold tangent to $\xi$ can have, and the Legendrian condition is the contact analogue of the Lagrangian condition in symplectic geometry.

Key theorem with proof [Intermediate+]

Theorem (contact Darboux). Let $(M,\xi )$ be a contact manifold of dimension $2n+1$ with co-oriented contact form $\alpha$. Around any point $p\in M$ there exist local coordinates $(x_1,y_1,\dots ,x_n,y_n,z)$ on a neighbourhood $U$ of $p$ in which

$$\alpha {|}_U=dz-\sum _{i=1}^n{y}_{i}d{x}_i.$$

Proof. Pick a chart at $p$ identifying a neighbourhood with ${\mathbb{R}}^{2n+1}$ in such a way that $\alpha (p)$ has the linear standard form: this is the Darboux normal form for a contact one-form on a single tangent space, available because $d\alpha {|}_{{\xi }_p}$ is a symplectic form on the $2n$-dimensional vector space ${\xi }_p$ and the linear theory of symplectic forms gives a basis. Call the resulting coordinates $(x_i,y_i,z)$ and write ${\alpha }_1$ for the given contact form in these coordinates and ${\alpha }_0=dz-\sum y_{i}d{x}_i$ for the standard model.

Both ${\alpha }_0$ and ${\alpha }_1$ agree at $p$ to first order: ${\alpha }_0(p)={\alpha }_1(p)$ and $d{\alpha }_0(p)=d{\alpha }_1(p)$ on the chosen basis. The path

$${\alpha }_t=(1-t){\alpha }_0+t{\alpha }_1,t\in [0,1],$$

is therefore a path of one-forms on a neighbourhood of $p$, all of which satisfy the contact condition ${\alpha }_t\wedge (d{\alpha }_t{)}^n\ne 0$ on a possibly smaller neighbourhood (the contact condition is open, and it holds at $p$ for every $t$).

Differentiating gives ${\dot{\alpha }}_t={\alpha }_1-{\alpha }_0$, a closed-by-construction one-form vanishing at $p$ to first order. By the relative Poincaré lemma applied to the vanishing-at-$p$ contact correction, write ${\dot{\alpha }}_t={\beta }_t$ with ${\beta }_t(p)=0$ and ${\beta }_t$ smooth in $t$.

Define the time-dependent vector field $X_t$ on the neighbourhood by the contact-Moser equation

$${\iota }_{X_t}{\alpha }_t=0,{\iota }_{X_t}d{\alpha }_t=-{\beta }_t+{\beta }_t(R_t){\alpha }_t,$$

where $R_t$ is the Reeb field of ${\alpha }_t$. The first equation forces $X_t\in {\xi }_t=\ker {\alpha }_t$; the restriction of $d{\alpha }_t$ to ${\xi }_t$ is non-degenerate, so the second equation has a unique solution $X_t$ in ${\xi }_t$. Both equations hold smoothly in $t$.

A direct calculation using the Cartan formula gives

$$\frac{d}{dt}({\psi }_t^{\ast }{\alpha }_t)={\psi }_t^{\ast }({\mathcal{L}}_{X_t}{\alpha }_t+{\dot{\alpha }}_t)={\psi }_t^{\ast }(d{\iota }_{X_t}{\alpha }_t+{\iota }_{X_t}d{\alpha }_t+{\beta }_t)={\psi }_t^{\ast }({\beta }_t(R_t){\alpha }_t),$$

so ${\psi }_t^{\ast }{\alpha }_t={\lambda }_t{\alpha }_0$ for some smooth positive function ${\lambda }_t$ (the conformal factor that records the multiplicative ambiguity of the contact form). Multiplying ${\psi }_1$ by a final point-by-point rescaling absorbs ${\lambda }_1$ and produces a diffeomorphism $\Psi$ of a neighbourhood of $p$ with ${\Psi }^{\ast }{\alpha }_1={\alpha }_0$. Pulling back the standard coordinates gives the asserted Darboux chart. $\square$

Bridge. The contact Darboux argument here builds toward 05.10.03 (Gray's theorem), where the same path-of-contact-forms construction appears again in the global setting: a compact manifold's path of contact structures, rather than a local pair, is shown to be generated by an isotopy of the manifold. The contact Moser equation ${\iota }_{X_t}{\alpha }_t=0$, ${\iota }_{X_t}d{\alpha }_t=-{\beta }_t\mathrm{mod}{\alpha }_t$ identifies primitives of ${\dot{\alpha }}_t$ along ${\xi }_t$ with diffeomorphisms — this is exactly the symplectic-Moser principle 05.01.05 specialised to the contact setting, and the bridge between the analytic equation for $X_t$ and the geometric conclusion ${\Psi }^{\ast }{\alpha }_1={\alpha }_0$ is the foundational reason every local-form result in contact geometry follows the same template. Putting these together, contact Darboux is the contact analogue of the symplectic Darboux theorem 05.01.04, and its symplectisation 05.10.02 recovers the symplectic version on $M\times \mathbb{R}$.

Exercises [Intermediate+]

Lean formalization [Intermediate+]

Mathlib does not yet name contact manifolds, contact forms, or Reeb vector fields as first-class objects. A proposed signature, in Lean 4 / Mathlib syntax, sketching what the formalisation would look like:

[object Promise]

The construction depends on names that do not currently exist in Mathlib (CotangentSpace, wedge product on ManifoldForms, the contact-condition predicate as a smooth section, the implicit-function-style existence of $R_{\alpha }$). Each is a candidate Mathlib contribution; until then this unit ships with lean_status: none.

Advanced results [Master]

The contact category is structurally parallel to the symplectic category but with its own rigidity phenomena, its own classification problems, and its own Floer-theoretic apparatus.

Equivalence of contact forms and contact structures. Two contact forms $\alpha ,{\alpha }^{\prime }$ on a manifold define the same contact structure iff ${\alpha }^{\prime }=f\alpha$ for a nowhere-vanishing $f$. The contact structure is the equivalence class; the contact form is a co-orientation. The Reeb field depends on the choice of form, not on the structure: rescaling $\alpha \mapsto f\alpha$ replaces $R_{\alpha }$ by a vector field whose flow lines (the Reeb orbits) are the same as those of $R_{\alpha }$ up to reparametrisation only when $f$ is constant. The non-uniqueness of the Reeb dynamics within a contact structure is one source of the difficulty of the Weinstein conjecture.

Symplectisation. Given a co-oriented contact manifold $(M,\alpha )$, the symplectisation is $(M\times \mathbb{R},\omega =d(e^t\alpha ))$, which is an exact symplectic manifold with the $\mathbb{R}$-action $(p,t)\mapsto (p,t+s)$ acting by symplectic conformal scaling: the action multiplies $\omega$ by $e^s$. Every contact-geometric question can be lifted to a symplectic question on the symplectisation, and many cylindrical Floer theories — symplectic field theory, contact homology — work in the symplectisation rather than the contact manifold itself. The symplectisation is the foundational bridge between contact and symplectic geometry, and the construction extends naturally to the Liouville cobordisms between contact manifolds that organise modern symplectic topology.

Legendrian submanifolds. A Legendrian $L^n\subset M^{2n+1}$ is the contact analogue of a Lagrangian. The space of Legendrian embeddings carries an isotopy class invariant called the Thurston-Bennequin invariant in dimension three, and Legendrian knot theory in ${\mathbb{R}}^3$ with the standard contact structure is a rich area with its own combinatorics (Chekanov-Eliashberg differential graded algebra, Legendrian contact homology).

Tight versus overtwisted. In dimension three, contact structures fall into two disjoint classes. A contact structure on a 3-manifold is overtwisted if it admits an embedded disk on which the contact planes are tangent along the boundary and twist by a full turn going inwards; tight otherwise. Bennequin (1983) proved the standard contact structure on ${\mathbb{R}}^3$ is tight, and Eliashberg (1989) classified overtwisted contact structures up to isotopy: on a closed 3-manifold, overtwisted contact structures up to isotopy are in bijection with homotopy classes of plane fields. Tight contact structures, by contrast, satisfy strong rigidity: the classification on $S^3$ is a single tight structure, and on torus bundles and lens spaces the classification is a deep finite-list theorem.

Gray's theorem. Isotopic contact structures on a compact manifold are diffeomorphic via an ambient isotopy. The proof is the Moser-trick adapted to the contact setting — the contact-Moser equation produces a vector field whose flow conjugates one structure to the other. Gray's theorem is the rigidity statement that contact structures are stable under deformation, and it is the contact analogue of Moser stability for symplectic forms.

The Weinstein conjecture. On every closed contact $(2n+1)$-manifold, every contact form $\alpha$ has at least one closed Reeb orbit. Proved in dimension three by Taubes (2007) using Seiberg-Witten theory; open in higher dimensions in full generality. The conjecture is the contact-geometric source of Floer-theoretic invariants: contact homology and embedded contact homology are designed precisely to count Reeb orbits.

Failure modes.

• Frobenius foliation. If $\alpha \wedge d\alpha \equiv 0$ then $\ker \alpha$ integrates by Frobenius. Such structures are foliations, not contact structures. The contact condition is the strongest possible obstruction to integrability.
• Confusion with co-symplectic. A co-symplectic structure is a pair $(\alpha ,\omega )$ on a $(2n+1)$-manifold with $\alpha$ closed and $\omega$ closed and $\alpha \wedge {\omega }^n\ne 0$. Co-symplectic structures are foliated, not contact: the closed Reeb-like field gives a foliation of $M$ by even-dimensional leaves.
• Sign convention drift. Some texts define $\alpha \wedge (d\alpha {)}^n>0$ (oriented contact); others allow either sign. The unoriented version is the natural notion for non-co-orientable structures; the oriented version is needed when $n$ is odd to fix the orientation of $M$.

Synthesis. Contact geometry generalises the geometry of first-order differential equations $z=z(x,y)$ on a domain $D\subset {\mathbb{R}}^2$: the graph $z=z(x,y)$ in $J^1(D,\mathbb{R})$ is Legendrian iff $z$ is a smooth function and $\partial z/\partial x_i=y_i$ holds, so contact geometry is the geometry where one-jets of functions are first-class objects. The bridge between the contact one-form $\alpha =dz-\sum y_{i}d{x}_i$ and the calculus of one-jets is exactly that $\alpha$ vanishes on a section iff the section is the one-jet of a function — so contact geometry identifies first-order differential equations with submanifolds of jet space. Read in the opposite direction, contact geometry specialises symplectic geometry: a contact manifold is the cone-base of its symplectisation, and the symplectisation is exactly the cylindrical exact symplectic manifold whose $\mathbb{R}$-action recovers the contact structure on every level.

Putting these together, the central insight is that contact geometry is the geometry one dimension below symplectic geometry, and the bridge between odd-dimensional contact manifolds and even-dimensional symplectic manifolds is the symplectisation functor — this is exactly the same correspondence that recurs as the Liouville-cobordism organisation of symplectic field theory and as the boundary correspondence for symplectic manifolds with contact-type boundary. The foundational reason contact and symplectic geometry are sister subjects rather than independent ones is that the contact-Moser equation ${\iota }_{X_t}{\alpha }_t=0$, ${\iota }_{X_t}d{\alpha }_t\equiv -{\beta }_t(\mathrm{mod}{\alpha }_t)$ is the symplectic-Moser equation restricted to the contact distribution and reduced by the conformal action — the bridge is the symplectisation, and on the symplectisation the contact and symplectic Moser arguments coincide.

Full proof set [Master]

Lemma (uniqueness of the Reeb field). On a co-oriented contact manifold $(M^{2n+1},\alpha )$ the Reeb vector field $R_{\alpha }$ defined by ${\iota }_{R_{\alpha }}\alpha =1$, ${\iota }_{R_{\alpha }}d\alpha =0$ exists and is unique.

Proof. The two-form $d\alpha$ is closed, and on the hyperplane distribution $\xi =\ker \alpha$ it restricts to a non-degenerate two-form (because $\alpha \wedge (d\alpha {)}^n\ne 0$ forces $(d\alpha {)}^n{|}_{\xi }$ to be a volume form on $\xi$, and a closed two-form whose top wedge is a volume form is symplectic). On the full tangent space $TM$, $d\alpha$ has a one-dimensional kernel: any vector $V$ with ${\iota }_{V}d\alpha =0$ must lie in the radical of $d\alpha$ on $TM$, and the radical has codimension exactly $2n$ inside $TM$, so dimension one. Pick any nonzero $V$ in this radical; then $\alpha (V)\ne 0$ (otherwise $V\in \xi$ and $V$ is in the kernel of $d\alpha {|}_{\xi }$, contradicting non-degeneracy). Set $R_{\alpha }=V/\alpha (V)$. Uniqueness follows from the one-dimensionality of the radical and the normalisation. $\square$

Lemma (relative Poincaré at a point, contact version). Let $\beta$ be a closed one-form on a neighbourhood of $0\in {\mathbb{R}}^{2n+1}$ vanishing at $0$. Then $\beta$ has a primitive $f$ vanishing to second order at $0$.

Proof sketch. The standard Poincaré-lemma cone construction $f(x)={\int }_0^1\beta (tx)(x)dt$ produces a primitive on a star-shaped neighbourhood. Vanishing of $\beta$ at zero gives an extra factor of $t$ in the integrand, hence vanishing of $f$ to second order at $0$. $\square$

Theorem (contact Darboux, restated). Local form result of the Intermediate section.

Proof. As in the Intermediate section: relative Moser for contact forms with ${\alpha }_0$ the standard model and ${\alpha }_1$ the given form, the path ${\alpha }_t$ remains contact on a small neighbourhood, the contact-Moser equation defines $X_t$ uniquely in ${\xi }_t$, and the Cartan-formula calculation closes the loop up to a conformal factor that is absorbed in the final point-by-point rescaling. $\square$

Theorem (Gray stability, statement). On a compact manifold, a smooth path ${\xi }_t$ of contact structures, $t\in [0,1]$, is generated by an isotopy: there exists a smooth path of diffeomorphisms ${\varphi }_t$ of $M$ with ${\varphi }_0=\mathrm{id}$ and $(\phi_t)_ \xi_0 = \xi_t$.*

Proof. Stated here without proof; the full argument is the contact-Moser construction in the global setting and forms the content of unit 05.10.03. $\square$

Connections [Master]

• Symplectic manifold 05.01.02. Contact geometry is the odd-dimensional sibling of symplectic geometry — a contact manifold's symplectisation is an exact symplectic manifold, and the boundary of a symplectic manifold of contact type is a contact manifold by restriction.

• Moser's trick 05.01.05. The contact-Darboux theorem above is a direct application of the Moser-trick template adapted to one-forms with a conformal-rescaling correction. The contact-Moser equation ${\iota }_{X_t}{\alpha }_t=0$, ${\iota }_{X_t}d{\alpha }_t\equiv -{\beta }_t(\mathrm{mod}{\alpha }_t)$ is the symplectic-Moser equation reduced by the contact-distribution constraint.

• Darboux's theorem 05.01.04. The contact-Darboux theorem proven above is the contact analogue: every contact manifold is locally the standard model $({\mathbb{R}}^{2n+1},dz-\sum y_{i}d{x}_i)$, just as every symplectic manifold is locally the standard model $({\mathbb{R}}^{2n},\sum d{q}_i\wedge d{p}_i)$.

• Symplectisation 05.10.02 (pending). The symplectisation $(M\times \mathbb{R},d(e^t\alpha ))$ converts contact into symplectic; the next unit develops the construction in detail and works out the canonical examples ($S^{2n+1}$, $S(T^{\ast }N)$, jet bundles).

• Gray's theorem 05.10.03 (pending). The contact rigidity theorem — isotopic contact structures are diffeomorphic — uses the contact-Moser argument from this unit in the global setting.

• Lagrangian submanifold 05.05.01. Legendrians are the contact analogue of Lagrangians. The dimension count, the integrability characterisation, and the role in Floer theory all parallel the symplectic case.

Historical & philosophical context [Master]

Sophus Lie introduced contact transformations in his 1872 work on the geometry of differential equations ^{[Lie 1872 unsourced primary, Geiges Ch. 2 secondary]}, in the form of point transformations of the space of one-jets of functions. The contact one-form $dz-\sum y_{i}d{x}_i$ on the jet space $J^1(D,\mathbb{R})$ is Lie's original contact form, and the contact condition is the integrability requirement for the relation $\partial z/\partial x_i=y_i$ to be preserved.

Élie Cartan's early-twentieth-century work on Pfaffian systems, exterior differential systems, and the equivalence problem placed contact geometry inside the manifold-theoretic framework. Cartan recognised the contact form as a $G$-structure on a $(2n+1)$-manifold and developed the equivalence theorems that are the precursors of the contact Darboux theorem ^{[Geiges Ch. 2]}.

Georges Reeb's 1952 Sur certaines propriétés topologiques des trajectoires des systèmes dynamiques introduced what is now called the Reeb vector field and initiated the dynamical-systems study of contact manifolds.

The modern contact-topology programme is a mid-twentieth-century-onwards development. J.W. Gray's 1959 Some global properties of contact structures (Ann. Math. 69) ^{[Gray 1959]} proved the stability theorem for contact structures and established that contact geometry has rigidity in the same sense that symplectic geometry does. Daniel Bennequin's 1983 Entrelacements et équations de Pfaff (Astérisque 107-108) ^{[Bennequin 1983]} introduced the tight/overtwisted dichotomy in dimension three, proving that the standard contact structure on ${\mathbb{R}}^3$ is tight by constructing the Bennequin inequality on Legendrian knots. Yakov Eliashberg's 1989 Classification of overtwisted contact structures on 3-manifolds (Inventiones Math. 98) ^{[Eliashberg 1989]} showed that overtwisted contact structures on a closed 3-manifold are classified, up to isotopy, by homotopy classes of plane fields — a homotopy-theoretic classification that contrasts sharply with the rigid finite-list classification of tight contact structures.

The transition from contact geometry as a curiosity of differential equations to contact geometry as a central pillar of low-dimensional topology and symplectic topology was driven by Eliashberg's work in the 1980s and 1990s and by the Weinstein conjecture, posed by Alan Weinstein in 1979, on the existence of closed Reeb orbits. Hans Geiges's An Introduction to Contact Topology (Cambridge, 2008) ^[Geiges] is the standard modern textbook treatment.

Bibliography [Master]

[object Promise]