Gray's stability theorem

Intuition [Beginner]

Gray's theorem says contact structures are rigid under deformation. Take a closed manifold and a smooth one-parameter family of contact structures, ${\xi }_t$ for $t\in [0,1]$. The two endpoint structures ${\xi }_0$ and ${\xi }_1$ might look completely different at first glance, but Gray's theorem produces an isotopy of the manifold — a smooth family of diffeomorphisms starting at the identity — whose time-one map carries ${\xi }_0$ onto ${\xi }_1$.

The conclusion is strong: deformation through contact structures forces diffeomorphism. Two contact structures sitting in the same connected component of the space of contact structures are related by an ambient isotopy of the manifold.

The technique is the contact analogue of the Moser-trick approach to symplectic stability. Build a time-dependent vector field whose flow conjugates the structures along the path, integrate, take the time-one map. The contact condition is what makes the equation for the vector field uniquely solvable.

Visual [Beginner]

A schematic of a closed three-manifold with two contact structures drawn as families of tilted planes, connected by a path of intermediate plane fields, with an arrow labelled ${\psi }_1$ indicating the time-one diffeomorphism produced by the contact-Moser flow.

The key picture is the flow generated by the Gray vector field along the path of contact structures.

Worked example [Beginner]

Take the three-sphere $S^3$ with the standard contact form ${\alpha }_0=\frac{1}{2}(x_{1}d{y}_1-y_{1}d{x}_1+x_{2}d{y}_2-y_{2}d{x}_2)$ restricted from ${\mathbb{R}}^4$. Now perturb to the family ${\alpha }_t=(1+t\cdot f){\alpha }_0$ where $f$ is some smooth positive function on $S^3$ and $t\in [0,1]$.

Each ${\alpha }_t$ defines the same contact distribution ${\xi }_t=\ker {\alpha }_t=\ker {\alpha }_0$, since multiplying a one-form by a positive function does not change its kernel. So in this case the diffeomorphism Gray's theorem produces is the identity. The Gray vector field is zero.

A more interesting case: deform the contact distribution itself by tilting it along a closed loop on $S^3$. The family ${\xi }_t$ now varies through different distributions. Gray's theorem guarantees the existence of a one-parameter family of diffeomorphisms ${\psi }_t$ of $S^3$ with ${\psi }_0=$ identity and $({\psi }_t{)}_{\ast }{\xi }_0={\xi }_t$. You never need to compute ${\psi }_t$ explicitly — the existence comes from the Cartan-formula calculation, the same way Moser stability does in the symplectic setting.

What this tells us: a deformation of a contact structure is the same data as an isotopy of the manifold. The space of contact structures, locally, is just diffeomorphisms acting on a fixed structure.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $M$ be a closed smooth manifold of dimension $2n+1$, and let ${\xi }_t$, $t\in [0,1]$, be a smooth one-parameter family of contact structures on $M$. Smoothness here means: there exist smooth contact forms ${\alpha }_t$ depending smoothly on $t$ with ${\xi }_t=\ker {\alpha }_t$ at each $t$ (in the co-orientable case; in the non-co-orientable case the condition is local and the global path is registered as a section of the appropriate line bundle).

A contact isotopy generating the family ${\xi }_t$ is a smooth family of diffeomorphisms ${\psi }_t:M\to M$ with ${\psi }_0={\mathrm{id}}_M$ and $({\psi }_t{)}_{\ast }{\xi }_0={\xi }_t$ for every $t\in [0,1]$. Equivalently, ${\psi }_t^{\ast }{\alpha }_t={\lambda }_t{\alpha }_0$ for some smooth positive function ${\lambda }_t:M\to {\mathbb{R}}_{>0}$ depending smoothly on $t$ — the conformal factor records the contact-form ambiguity inside a single contact structure.

Gray's theorem asserts the existence of such an isotopy whenever the family ${\xi }_t$ is smooth and the manifold is closed.

Key theorem with proof [Intermediate+]

Theorem (Gray 1959). Let $M$ be a closed smooth manifold of dimension $2n+1$, and let ${\xi }_t$, $t\in [0,1]$, be a smooth one-parameter family of contact structures on $M$. Then there exists a smooth isotopy ${\psi }_t:M\to M$ with ${\psi }_0={\mathrm{id}}_M$ and $(\psi_t)_ \xi_0 = \xi_t$forevery$t \in [0, 1]$.*

Proof. Choose a smooth family of contact forms ${\alpha }_t$ with ${\xi }_t=\ker {\alpha }_t$, smooth in $t$ (locally always available; globally available in the co-oriented case, which we assume — the non-co-oriented case is reduced to the co-oriented case by passing to the orientation double cover).

Look for the desired isotopy ${\psi }_t$ in the form generated by a time-dependent vector field $X_t$, satisfying ${\psi }_t^{\ast }{\alpha }_t={\lambda }_t{\alpha }_0$ for some smooth positive function ${\lambda }_t:M\to {\mathbb{R}}_{>0}$.

Differentiating both sides in $t$ and using the standard identity $\frac{d}{dt}({\psi }_t^{\ast }{\alpha }_t)={\psi }_t^{\ast }({\dot{\alpha }}_t+{\mathcal{L}}_{X_t}{\alpha }_t)$ gives

$${\psi }_t^{\ast }({\dot{\alpha }}_t+{\mathcal{L}}_{X_t}{\alpha }_t)={\dot{\lambda }}_t{\alpha }_0.$$

Apply Cartan's formula ${\mathcal{L}}_{X_t}{\alpha }_t=d{\iota }_{X_t}{\alpha }_t+{\iota }_{X_t}d{\alpha }_t$. Choose the ansatz $X_t\in {\xi }_t$, that is, ${\alpha }_t(X_t)={\iota }_{X_t}{\alpha }_t=0$. Then ${\mathcal{L}}_{X_t}{\alpha }_t={\iota }_{X_t}d{\alpha }_t$ and the differentiated identity becomes

$${\psi }_t^{\ast }({\dot{\alpha }}_t+{\iota }_{X_t}d{\alpha }_t)={\dot{\lambda }}_t{\alpha }_0.$$

Pulling ${\psi }_t^{\ast }$ across, the equation to solve (before applying ${\psi }_t^{\ast }$) is

$${\dot{\alpha }}_t+{\iota }_{X_t}d{\alpha }_t={\mu }_t{\alpha }_t\text{for some smooth function }{\mu }_t,$$

with ${\mu }_t=({\psi }_t^{-1}{)}^{\ast }({\dot{\lambda }}_t/{\lambda }_t)\cdot {\lambda }_t$ — a smooth scalar to be determined. Restricting to ${\xi }_t=\ker {\alpha }_t$, the right-hand side vanishes, leaving

$${\iota }_{X_t}d{\alpha }_t{|}_{{\xi }_t}=-{\dot{\alpha }}_t{|}_{{\xi }_t}.$$

The contact condition states that $d{\alpha }_t$ restricted to ${\xi }_t$ is a non-degenerate two-form (a symplectic form on the rank-$2n$ distribution). The map $X_t\mapsto {\iota }_{X_t}d{\alpha }_t{|}_{{\xi }_t}$ from ${\xi }_t$ to its dual is therefore a vector-bundle isomorphism, and the equation above admits a unique smooth solution $X_t\in {\xi }_t$.

Determining ${\mu }_t$: evaluate the unrestricted equation ${\dot{\alpha }}_t+{\iota }_{X_t}d{\alpha }_t={\mu }_t{\alpha }_t$ on the Reeb field $R_{{\alpha }_t}$ of ${\alpha }_t$. Recall ${\iota }_{R_{{\alpha }_t}}{\alpha }_t=1$ and ${\iota }_{R_{{\alpha }_t}}d{\alpha }_t=0$. So

$${\dot{\alpha }}_t(R_{{\alpha }_t})+d{\alpha }_t(X_t,R_{{\alpha }_t})={\mu }_t,$$

and $d{\alpha }_t(X_t,R_{{\alpha }_t})=-{\iota }_{R_{{\alpha }_t}}{\iota }_{X_t}d{\alpha }_t=0$ because ${\iota }_{R_{{\alpha }_t}}d{\alpha }_t=0$. Therefore ${\mu }_t={\dot{\alpha }}_t(R_{{\alpha }_t})$, a smooth function of $t$ and the point of $M$.

The vector field $X_t$ is now a smooth time-dependent section of $TM$ on the closed manifold $M$. Standard ODE theory gives a smooth flow ${\psi }_t:M\to M$ with ${\psi }_0={\mathrm{id}}_M$, defined for all $t\in [0,1]$. The function ${\lambda }_t$ is recovered by integrating the scalar ODE ${\dot{\lambda }}_t/{\lambda }_t={\psi }_t^{\ast }{\mu }_t$ with ${\lambda }_0\equiv 1$, giving ${\lambda }_t=\exp ({\int }_0^t{\psi }_s^{\ast }{\mu }_{s}ds)$, which is smooth and positive.

By construction $\frac{d}{dt}({\psi }_t^{\ast }{\alpha }_t)-{\dot{\lambda }}_t{\alpha }_0=0$, and at $t=0$ the difference ${\psi }_0^{\ast }{\alpha }_0-{\lambda }_0{\alpha }_0=0$, so ${\psi }_t^{\ast }{\alpha }_t={\lambda }_t{\alpha }_0$ for every $t$. Taking kernels gives $({\psi }_t{)}_{\ast }{\xi }_0=({\psi }_t{)}_{\ast }\ker {\alpha }_0=\ker (({\psi }_t^{-1}{)}^{\ast }{\alpha }_0)=\ker ({\lambda }_t^{-1}{\alpha }_t)={\xi }_t$. $\square$

Bridge. The Gray-stability argument here builds toward 05.10.02 (symplectisation), where the contact-form path ${\alpha }_t$ on $M$ lifts to the symplectic-form path ${\omega }_t=d(e^s{\alpha }_t)$ on $M\times \mathbb{R}$, and Gray's contact-Moser flow appears again in the symplectisation as the lift of a symplectic-Moser flow generated by an $\mathbb{R}$-equivariant vector field — exactly the symplectisation functoriality that organises symplectic field theory and embedded contact homology. The contact-Moser equation ${\iota }_{X_t}d{\alpha }_t{|}_{{\xi }_t}=-{\dot{\alpha }}_t{|}_{{\xi }_t}$ is the symplectic-Moser equation 05.01.05 restricted from $TM$ to the contact distribution and dressed with the conformal factor ${\lambda }_t$ — this bridge between the two-form non-degeneracy on ${\xi }_t$ and the diffeomorphism conclusion is the foundational reason every stability statement for contact structures takes the same template. Putting these together, Gray's theorem says the moduli space of contact structures on a closed manifold has no local invariants beyond the connected component, and the connected component itself is the orbit of the identity-component diffeomorphism group.

Exercises [Intermediate+]

Lean formalization [Intermediate+]

Mathlib does not currently name contact structures, contact forms, the Reeb vector field, or smooth families of any of these as first-class objects. A proposed signature, in Lean 4 / Mathlib syntax, sketching what the formalisation would look like:

[object Promise]

The proof depends on names that do not currently exist in Mathlib (contact-form smooth families, the Reeb field, the time-dependent vector-field existence-uniqueness from non-degeneracy of $d{\alpha }_t$ on ${\xi }_t$, the Cartan-formula identity for one-forms along a flow). Each is a candidate Mathlib contribution; until then this unit ships with lean_status: none.

Advanced results [Master]

Gray's theorem is the foundational rigidity statement of contact topology. Every classification problem for contact structures presupposes it: when a textbook says "two contact structures on a 3-manifold are the same iff they are homotopic through contact structures iff they are diffeomorphic via an ambient isotopy", the equivalence between the second and third clauses is Gray. The theorem is what makes the contact-structure-up-to-diffeomorphism question well-posed as a homotopy problem.

Reeb-flow stability under perturbation. A perturbation $\alpha +ϵ\beta$ of a contact form with $\beta$ a smooth one-form gives a contact form for $ϵ$ small enough (the contact condition is open). The Gray flow conjugating the perturbed structure to the unperturbed one carries the perturbed Reeb field to a vector field on $(M,\xi )$ generating an orbit-equivalent flow. The orbit-equivalence is the right invariance class for the Reeb dynamics: closed Reeb orbits, their homology classes, and their Conley-Zehnder indices are invariants of the contact structure, not of the contact form. This is the structural fact behind the well-definedness of contact homology and embedded contact homology as invariants of $(M,\xi )$.

Foundation for symplectic field theory and embedded contact homology. Counts of pseudoholomorphic curves in the symplectisation $(M\times \mathbb{R},d(e^t\alpha ))$ asymptotic to closed Reeb orbits are well-defined invariants of $(M,\xi )$ because Gray's theorem guarantees that different choices of $\alpha$ within the same contact structure give symplectomorphic-on-the-cylindrical-end symplectisations. In dimension three, Hutchings's embedded contact homology realises this and is now known (Cristofaro-Gardiner, Hutchings, Taubes, 2016) to coincide with monopole Floer homology of the underlying 3-manifold. Each step of the construction uses Gray-stability invariance at the start.

Tight versus overtwisted in dimension three. Eliashberg's 1989 classification of overtwisted contact structures on closed 3-manifolds asserts that two overtwisted structures are isotopic through contact structures (equivalently, by Gray's theorem, ambient-isotopic) iff they are homotopic as plane fields. The h-principle for overtwisted contact structures replaces a contact-topological problem with a homotopy-theoretic one, and Gray's theorem is what converts the homotopy-of-contact-structures conclusion of the h-principle into a diffeomorphism statement.

Higher-dimensional h-principles. Borman, Eliashberg, and Murphy (2015) extended the overtwisted h-principle to all odd dimensions, defining overtwistedness via a model on the standard $D^{2n}$ and proving that overtwisted contact structures in dimension $2n+1$ are classified, up to ambient isotopy, by homotopy classes of almost-contact structures. Gray's theorem is again the bridge from the homotopy conclusion to the diffeomorphism conclusion.

Failure modes.

• Non-compact manifold. Without compactness, the Gray vector field's flow may not exist for all $t\in [0,1]$. The local-on-relatively-compact-sets statement still holds, but global isotopy does not.
• Path of hyperplane fields rather than contact structures. If one structure along the path fails the contact condition, $d{\alpha }_t{|}_{{\xi }_t}$ becomes degenerate and the Gray equation is no longer uniquely solvable. The theorem is genuinely about paths through the contact locus.
• Family with non-smooth dependence on $t$. $C^0$ paths of contact structures need not be Gray-stable. The proof uses the flow of a smooth time-dependent vector field, and the regularity of the flow degrades with the regularity of the path.
• Sign-convention drift in the contact form. When $n$ is odd, the sign of $\alpha \wedge (d\alpha {)}^n$ depends on the orientation of $\alpha$. A path of contact structures must use a consistently oriented family of forms; an unoriented path can require a sign-flip that breaks smoothness.

Synthesis. Gray's theorem stands beside Moser stability and the volume-form theorem of Moser 1965 as one of the three foundational stability theorems of differential geometry, each provable by the same path-method template. The cohomological hypothesis in the symplectic case becomes the contact condition in the contact case: non-degeneracy of $d{\alpha }_t$ on ${\xi }_t$ replaces non-degeneracy of ${\omega }_t$ on $TM$, and the conformal factor ${\lambda }_t$ tracks the multiplicative ambiguity of the contact form. Read in the opposite direction, the Gray construction is dual to the deformation map of the contact-structure space: the tangent vectors at $\xi$ to the contact-structure space are smooth one-forms modulo a kernel, the Gray vector field is the inverse of the bundle isomorphism $\xi \to {\xi }^{\ast }$ given by $d\alpha {|}_{\xi }$, and the existence of the inverse is exactly the contact condition. The bridge between the analytic equation for $X_t$ and the geometric conclusion $({\psi }_t{)}_{\ast }{\xi }_0={\xi }_t$ is the foundational reason the contact-structure moduli space on a closed manifold has flat local geometry — locally, it is just a quotient of the diffeomorphism group. Putting these together, Gray's theorem says contact structures, like symplectic structures, have no local invariants beyond dimension and orientation: the contact analogue of "symplectic geometry has no local invariants" is "contact geometry has no local invariants beyond connected component", and Gray is what makes the connected-component classification the right object of study.

Full proof set [Master]

Lemma (smoothness of the bundle inverse). Let ${\xi }_t\subset TM$ be a smooth one-parameter family of rank-$2n$ distributions on a closed $(2n+1)$-manifold $M$, and let $\omega_t \in \Gamma(\Lambda^2 \xi_t^)$beasmoothfamilyoftwo-formson$\xi_t$,eachnon-degenerate.Thebundleisomorphism$\xi_t \to \xi_t^$givenby$X \mapsto \iota_X \omega_t$hasasmoothinversejointlyin$(t, p) \in [0, 1] \times M$.

Proof. Non-degeneracy at each $(t,p)$ gives an inverse pointwise. Smoothness in $(t,p)$ follows from smoothness of ${\omega }_t$ and the inversion theorem applied to the bundle map: the inverse of a smoothly varying linear isomorphism varies smoothly. The closedness and compactness of $M$ are used to conclude the inverse is uniformly bounded, hence the time-dependent vector field $X_t$ defined by it is smooth and bounded on $[0,1]\times M$. $\square$

Lemma (Cartan-formula identity for one-forms along a flow). For a smooth time-dependent vector field $X_t$ on $M$ with flow ${\psi }_t$ and a smooth one-parameter family of one-forms ${\alpha }_t$,

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

Proof. The chain rule gives $\frac{d}{dt}{\psi }_t^{\ast }{\alpha }_t{|}_{t=t_0}={\psi }_{t_0}^{\ast }{\dot{\alpha }}_{t_0}+\frac{d}{ds}{|}_{s=0}{\psi }_{t_0+s}^{\ast }{\alpha }_{t_0}$. The second term is ${\psi }_{t_0}^{\ast }{\mathcal{L}}_{X_{t_0}}{\alpha }_{t_0}$ by the definition of the Lie derivative along a time-dependent flow. $\square$

Theorem (Gray 1959). Statement and proof as in the Intermediate section.

Proof. The Intermediate-section proof goes through verbatim. The two lemmas above package the technical inputs: the smooth-inversion lemma gives the smooth time-dependent vector field $X_t$ from the contact condition, the Cartan-formula identity gives the differentiated equation for ${\psi }_t^{\ast }{\alpha }_t$. Compactness of $M$ ensures the time-dependent flow exists for all $t\in [0,1]$. $\square$

Theorem (parametric Gray). A smooth $K$-parameter family ${\xi }_{t,s}$ of contact-structure paths on a closed $M$, $s\in K$ a smooth manifold, admits a smooth $K$-parameter family of generating isotopies ${\psi }_{t,s}$.

Proof. Apply the Gray construction with smooth dependence on the parameter $s$. The smooth-inversion lemma extends to smooth dependence on $s$, giving a smooth two-parameter family $X_{t,s}$ of vector fields. The flow construction with parameters gives ${\psi }_{t,s}$, smooth in $(t,s)$ on the compact $M$. $\square$

Corollary (path-component classification). On a closed manifold $M$, two contact structures lie in the same path component of the space of contact structures iff they are related by an isotopy in the identity component ${\mathrm{Diff}}_0(M)$ of the diffeomorphism group.

Proof. If ${\xi }_0$ and ${\xi }_1$ are in the same path component, smooth a continuous path joining them to a smooth path of contact structures (using openness of the contact condition in the smooth topology). Gray gives ${\psi }_t\in {\mathrm{Diff}}_0(M)$ with $({\psi }_1{)}_{\ast }{\xi }_0={\xi }_1$. Conversely, if ${\psi }_1\in {\mathrm{Diff}}_0(M)$ with $({\psi }_1{)}_{\ast }{\xi }_0={\xi }_1$, choose an isotopy ${\psi }_t$ from the identity to ${\psi }_1$ and pull back ${\xi }_1$: the family $({\psi }_t^{-1}{)}_{\ast }{\xi }_1$ is a smooth path of contact structures from ${\xi }_0$ to ${\xi }_1$. $\square$

Connections [Master]

• Contact manifold 05.10.01. Gray's theorem is the global rigidity counterpart to the local Darboux theorem for contact manifolds. Together they say contact structures on a closed manifold have no local invariants and the global classification is up to ambient isotopy on each path component of the contact-structure space.

• Moser's trick 05.01.05. Gray's argument is the contact analogue of Moser's trick. The contact-Moser equation ${\iota }_{X_t}d{\alpha }_t{|}_{{\xi }_t}=-{\dot{\alpha }}_t{|}_{{\xi }_t}$ is the symplectic-Moser equation ${\iota }_{X_t}{\omega }_t=-{\alpha }_t$ restricted from $TM$ to the contact distribution, with the conformal factor ${\lambda }_t$ absorbing the contact-form-rescaling ambiguity that has no symplectic counterpart.

• Symplectisation 05.10.02 (pending). A contact-form path ${\alpha }_t$ on $M$ lifts to a symplectic-form path ${\omega }_t=d(e^s{\alpha }_t)$ on $M\times \mathbb{R}$, and Gray's contact-Moser flow lifts to an $\mathbb{R}$-equivariant symplectic-Moser flow on the symplectisation. This functoriality is the foundation of contact-Floer theories.

• Symplectic manifold 05.01.02. The well-definedness of symplectic-topology invariants for contact-type boundaries — Stein domains, Liouville cobordisms, Weinstein manifolds — uses Gray-stability invariance of the boundary contact structure under deformation.

• Darboux's theorem 05.01.04 and contact Darboux. The contact Darboux theorem proven inside 05.10.01 uses a localised-at-a-point Moser argument; Gray's theorem is the global version. The two results occupy parallel positions: Darboux is local rigidity, Gray is global stability.

• Lagrangian submanifold 05.05.01 and Legendrian submanifolds. Legendrian-isotopy theorems for closed Legendrian submanifolds in a fixed contact manifold are proved by parametric Gray-style arguments applied to families of contact-form paths fixing the Legendrian, parallel to the Hamiltonian-isotopy theory of Lagrangian submanifolds.

Historical & philosophical context [Master]

J. W. Gray proved the stability theorem in 1959 in Some global properties of contact structures (Ann. of Math. (2) 69, 421–450) ^{[Gray 1959]}. The paper introduced the contact-Moser construction in the global setting on a closed manifold, parallel to the path-method that Moser would publish six years later for volume forms. Gray's argument was framed in the language of $G$-structures and equivalence of Pfaffian systems — the contact form was viewed as a $G$-structure on a $(2n+1)$-manifold, and the stability theorem was an equivalence-of-$G$-structures result in the sense of Élie Cartan. The modern framing in terms of contact-form paths and time-dependent vector fields is due to subsequent expositors, particularly Geiges in his Cambridge textbook and McDuff–Salamon in their introduction.

The conceptual position of Gray's theorem alongside Moser's volume-form theorem (1965) and Moser–Weinstein symplectic stability (1965/1971) was clarified in the 1970s and 1980s as contact and symplectic topology developed in parallel. The path-method became a single template applied across three settings: volume forms (the original Moser argument), symplectic forms (Moser–Weinstein), and contact structures (Gray). The unifying observation is that each setting has a non-degeneracy hypothesis that converts a primitive of the time-derivative into a vector field whose flow conjugates the structures.

The implications of Gray's theorem for the classification of contact structures became central in the 1980s with the work of Daniel Bennequin (1983, Entrelacements et équations de Pfaff, Astérisque 107–108) ^{[Bennequin 1983]} on the tight contact structure on ${\mathbb{R}}^3$ and Yakov Eliashberg (1989, Classification of overtwisted contact structures on 3-manifolds, Inventiones Math. 98) ^{[Eliashberg 1989]} on the h-principle for overtwisted contact structures. Eliashberg's result classifies overtwisted contact structures on a closed 3-manifold up to isotopy through contact structures by homotopy classes of plane fields; Gray's theorem converts the conclusion into a classification up to ambient isotopy of the manifold. Borman, Eliashberg, and Murphy extended the overtwisted h-principle to higher dimensions in 2015 (Acta Math. 215), again with Gray's theorem as the bridge from the contact-homotopy conclusion to the diffeomorphism conclusion.

Hans Geiges's An Introduction to Contact Topology (Cambridge Studies in Advanced Mathematics 109, 2008) ^[Geiges] is the standard modern textbook treatment, with the Gray proof presented in Chapter 2 alongside contact Darboux.

Bibliography [Master]

[object Promise]