6
$\begingroup$

Suppose $\kappa$ is a no-where vanishing 1-form, then its kernel is integrable is equivalent to condition $d\kappa \wedge \kappa = 0$.

My question is, can such foliation smoothly deformed such that actually $d\kappa = 0$?

EDIT: Smooth means real $C^\infty$, and I want the $\kappa$ to be nowhere-vanishing along the deformation. Thanks Loïc Teyssier pointing out the lack of context.

Thank you.

$\endgroup$
2
  • 1
    $\begingroup$ It is still not clear on which manifold you consider your $1$-form, and whether you wish to deform the foliation by preserving the transverse structure or not, is it a deformation that can be placed within a bigger foliation...? I maintain that you cannot rule out trivial examples without additional hypothesis. $\endgroup$ Sep 28, 2013 at 15:05
  • $\begingroup$ You are right that my questions does not specify enough information. But the actual (physics; I am physics majored) problem I am dealing with that leads to my post does not see these details (which probably will in the future). Anyhow, I know it is rather elementary, but could you educate me/provide some references on the meaning of "transverse structure" and "deformation that preserve transverse structure"? I feel that they'll be relevant in the future. $\endgroup$
    – Lelouch
    Sep 29, 2013 at 1:57

3 Answers 3

7
$\begingroup$

If $\mathcal F$ is a smooth codimension one foliation on $S^3$ then $\mathcal F$ cannot be defined by a closed $1$-form. This follows from Novikov's compact leaf theorem which says that one of the leaves of $\mathcal F$ is a torus $T$ bounding a Reeb component. If $\mathcal F$ were defined by a closed $1$-form then every leaf near $T$ would also be compact. Indeed, at a neighborhood of $T$, any closed $1$-form defining $\mathcal F$ would admit a primitive as we can normalize local primitives by setting them to be zero at $T$. Level sets of this primitive would be other compact leaves for $\mathcal F$. Reeb global stability Theorem implies that every leaf of $\mathcal F$ is compact, and that $Sˆ3$ admits a map to $S^1$ with connected fibers. But this is impossible since from it we get a surjective map from $\pi_1(S^3)=0$ to $\pi_1(S^1)=\mathbb Z$.

The same argument applies to codimension one foliations on compact $3$-folds with finite fundamental group.

$\endgroup$
0
16
$\begingroup$

Although jvp's argument is correct, there's a more elementary argument that a foliation on $S^3$, or any closed integral homology sphere $M$ ($b_1(M)=0$), cannot admit a foliation defined by a closed (nowhere zero) 1-form. The point is that if $d\kappa=0$, then since $H_1(M;\mathbb{R})=0$, there exists a smooth function $f:M\to \mathbb{R}$ so that $\kappa=df$. Let $p\in M$ be a point at which the maximum of $f$ is achieved, then $\kappa_p=df_p = 0$, contradicting that $\kappa$ is nowhere zero.

In fact, there is a complete characterization of foliations defined by closed nowhere-zero 1-forms. Suppose $\kappa$ is a closed, nowhere zero 1-form. Then $ker(\kappa)$ defines an integrable plane field, which integrates to a foliation $\mathcal{F}$ with a transverse measure given by integration agains $\kappa$. A closed manifold with a measured foliation must fiber over $S^1$, and $\kappa$ may be deformed to a closed nowhere zero 1-form $\alpha$ with integral periods, which defines a submersion to $S^1$, and therefore a fibration $\Sigma \to M \to S^1$, where $\Sigma$ is a closed surface tangent to $ker(\alpha)$.

The deformation argument is simple. Take $k=b_1(M)$ loops $c_1,\ldots,c_k$ which generate $H_1(M;\mathbb{Z})/Torsion$. For any closed 1-form $\beta$, we obtain a $k$-tuple of periods $(\beta(c_1),\ldots,\beta(c_k))\in \mathbb{R}^k$. Choose $k$ closed 1-forms $\alpha_1,\ldots,\alpha_k$ generating $H^1(M;\mathbb{R})$. Then for small $\epsilon$, we have $\kappa+\sum t_i \alpha_i$ is a nowhere zero closed 1-form for $|t_i|< \epsilon$. Moreover, the period map takes these forms to a small open set in $\mathbb{R}^k$. This set must contain a point $\kappa+\sum t_i\alpha_i$ with rational coordinates, so we choose $\alpha=m(\kappa +\sum t_i \alpha_i )$, where $m$ is the $lcm$ of the denominators of the coordinates. Clearly $\alpha$ smoothly deforms through closed nonzero 1-forms to $\kappa$, and $\alpha$ has integral periods.

The foliations described by closed 1-forms may be characterized up to isotopy: there is a unique such foliation for each element in the projective cone over the fibered faces of the Thurston norm unit ball. If two nowhere zero closed 1-forms are cohomologous, one can show that the corresponding foliations are isotopic, and the same if one is a multiple of the other (by rescaling).

Associated to each face of the Thurston norm unit ball, Thurston showed that there is a canonical element of $H^2(M)$ which realizes the Thurston norm when evaluated on norm-minimizing surfaces in that face. For a nowhere zero integrable 1-form $\alpha$ defining a foliation (so $\alpha \wedge d\alpha=0$), the Euler class of $ker(\alpha)$ has the property that it gives the Euler characteristic of closed leaves. Deformation of integrable 1-forms preserves the Euler class. Thus, if an integrable 1-form is deformable to a closed 1-form, then it must induce the same Euler class as the class associated to a fibered face of the Thurston norm unit ball. For example, associated to each homology class on the boundary of a fibered face, there is a depth one foliation realizing the Euler class of the fibered face, but which is not itself cohomologous to a fibration. To completely answer your question, one would therefore like to know if integrable 1-forms which define the same Euler class are deformations of each other. This question is partially resolved. Suppose that two foliations of a 3-manifold are deformable, then their tangent bundles are homotopic. Eynard-Bontemps proved a converse, that if two oriented foliations have homotopic tangent bundles, then they are deformable, but only through $C^1$ foliations. If the two foliations are taut, then the deformation may be made smoothly. In your case, at one end the foliation is taut (when defined by a closed 1-form), but the initial foliation defined by $\kappa$ might not be taut.

$\endgroup$
2
  • $\begingroup$ This is defintely a much better answer than mine. In retrospect it is difficult to understand why I favoured such convoluted argument. $\endgroup$ Sep 29, 2013 at 0:04
  • $\begingroup$ Great explanation. Thx! $\endgroup$
    – Lelouch
    Sep 29, 2013 at 4:58
4
$\begingroup$

The answer is "no" in some rather natural sense of "deformation". There is a construction due to G. Hector of an everywhere regular smooth (=real $C^\infty$) foliation of $\mathbb R^3=\{(x,y,z)\}$ of codimension $1$, everywhere transverse to the fibers of the projection $(x,y,z)\mapsto(x,y)$, with a distinguished leaf $\{z=0\}$, such that the holonomy computed on some transverse $\Sigma:=\{(x,y)=cst\}$ by lifting the unit circle of $\{z=0\}$, coincides outside a neighborhood of $\Sigma\cap \{z=0\}$ with some beforehand-given smooth diffeomorphism $h : \mathbb R\to \mathbb R$. Now deforming this foliation in the space of Frobenius-integrable foliations could mean that the holonomy is preserved up to $C^\infty $ conjugacy (what would you want to deform a foliation for if it means breaking the transverse structure?). But only the identity can be obtained as the monodromy of a regular closed $1$-form on $\mathbb R^3$ since $d\kappa=0$ implies the foliation has a smooth first-integral $f$, i.e. $\kappa=df$.

$\endgroup$
3
  • $\begingroup$ But then you only get $d\kappa_{\varepsilon}=0$ at $\varepsilon=0$, and then you don't have a foliation. $\endgroup$
    – Ben McKay
    Sep 27, 2013 at 19:25
  • $\begingroup$ The real question is whether you can smoothly deform any smooth foliation into a smooth foliation which is globally the kernel of a closed nowhere-vanishing 1-form, I think. $\endgroup$
    – Ben McKay
    Sep 27, 2013 at 19:26
  • 1
    $\begingroup$ @BenMcKay second comment: does "smooth" mean real $C^\infty $ or real- (complex-)analytic? What is the geometric/topological nature of the manifold on which $\kappa$ is given? Is $\kappa$ regular? singular? With isolated singularities? This question clearly lacks context... $\endgroup$ Sep 27, 2013 at 19:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.