A note on invariants of foliated $3$-sphere bundles

Abstract.

In this note we prove that $H^{*}(\mathrm{B}\mathrm{SO}(4);\mathbb{Q})$ injects into the group cohomology of $\mathrm{Diff}^{+}(\mathbb{S}^{3})$ with rational coefficients. The proof is based on an idea of Nariman who proved that the monomials in the Euler and Pontrjagin classes are nontrivial in $H^{*}(\mathrm{B}\mathrm{Diff}_{+}^{\delta}(\mathbb{S}^{2n-1});\mathbb{Q})$.

1. Introduction

Let $\mathrm{B}\mathrm{S}\Gamma_{d}$ be the classifying space of Haefliger structures for codimension $d$ foliations that are transversely oriented and denote by $\nu:\mathrm{B}\mathrm{S}\Gamma_{d}\rightarrow\mathrm{B}\mathrm{GL}^{+}_{d}(\mathbb{R})$ the map that records the normal bundle of the foliation (see [Hae71] for details). We consider the space of bundle maps $\mathrm{Bun}(TM,\nu^{*}\gamma_{d})$ where $\gamma_{d}$ denotes the universal oriented vector bundle over $\mathrm{B}\mathrm{GL}_{d}^{+}(\mathbb{R})$. It has a $\mathrm{Diff}_{+}(M)$-action by precomposition with the differential of a diffeomorphism, and Nariman proved (cf. [Nar17] and [Nar23, Cor. 2.5]) that the map $\mathrm{B}\mathrm{Diff}_{+}^{\delta}(M)\rightarrow\mathrm{B}\mathrm{Diff}_{+}(M)$ induced by the inclusion of $\mathrm{Diff}_{+}^{\delta}(M)$, the group of diffeomorphisms with the discrete topology, factors through an acyclic map

$$\beta:\mathrm{B}\mathrm{Diff}_{+}^{\delta}(M)\longrightarrow\mathrm{Bun}(TM,\nu^{*}\gamma_{d})/\!\!/\mathrm{Diff}_{+}(M).$$

This means in particular that $\beta$ induces an isomorphism

(1.1) $$H^{*}(\mathrm{B}\mathrm{Diff}^{\delta}_{+}(M);\mathbb{Q})\cong H^{*}(\mathrm{Bun}(TM,\nu^{*}\gamma_{d})/\!\!/\mathrm{Diff}_{+}(M);\mathbb{Q}),$$

and under this isomorphism the map on cohomology induced by $\mathrm{B}\mathrm{Diff}^{\delta}_{+}(M)\to\mathrm{B}\mathrm{Diff}_{+}(M)$ agrees with the map induced by the projection

$$p\colon\mathrm{Bun}(TM,\nu^{*}\gamma_{d})/\!\!/\mathrm{Diff}_{+}(M)\rightarrow\mathrm{B}\mathrm{Diff}_{+}(M).$$

If a Lie group $G$ acts smoothly on $M$, there is a commutative diagram

(1.2)

and Nariman shows in [Nar23, Sect. 3] that if $G$ is a torus acting freely on $M$, then $\mathrm{Bun}(TM,\nu^{*}\gamma_{d})$ has a $G$-fixed point so the left vertical map has a section. Hence, the map on cohomology induced by $\mathrm{B}G\rightarrow\mathrm{B}\mathrm{Diff}_{+}(M)$ factors as

$$H^{*}(\mathrm{B}\mathrm{Diff}_{+}(M))\longrightarrow H^{*}(\mathrm{B}\mathrm{Diff}^{\delta}_{+}(M))\longrightarrow H^{*}(\mathrm{B}G).$$

For an odd sphere one can define classes $e,p_{1},\ldots,p_{n-1}\in H^{*}(\mathrm{B}\mathrm{Diff}_{+}(\mathbb{S}^{2n-1})$ that pull back to the Euler and Pontrjagin classes along the map $\mathrm{B}\mathrm{SO}(2n)\rightarrow\mathrm{B}\mathrm{Diff}_{+}(\mathbb{S}^{2n-1})$ induced by the action of $\mathrm{SO}(2n)$ on $\mathbb{S}^{2n-1}$. If we further restrict to the free $S^{1}$-action, the image of the monomials in the Euler and Pontrjagin classes in the cohomology of $\mathrm{B}S^{1}$ are nontrivial and it follows that they are nontrivial in $H^{*}(\mathrm{B}\mathrm{Diff}^{\delta}_{+}(\mathbb{S}^{2n-1}))$ [Nar23, Thm 1.2]. But this poses the question about injectivity of the map

(1.3) $$\mathbb{Q}[e,p_{1},\ldots,p_{n-1}]\subset H^{*}(\mathrm{B}\mathrm{Diff}_{+}(\mathbb{S}^{2n-1});\mathbb{Q})\rightarrow H^{*}(\mathrm{B}\mathrm{Diff}^{\delta}_{+}(\mathbb{S}^{2n-1});\mathbb{Q}).$$

The obvious obstacle to making a statement about injectivity of (1.3) is that the Krull dimension of $H^{*}(\mathrm{B}S^{1};\mathbb{Q})$ is too small. Instead, we should use the action of the maximal torus $T^{n}\subset\mathrm{SO}(2n)$ on $\mathbb{S}^{2n-1}$. However, $T^{n}$ is not acting freely on $\mathbb{S}^{2n-1}$ so that one does not know whether $\mathrm{Bun}(T\mathbb{S}^{2n-1},\nu^{*}\gamma_{2n-1})$ has a $T^{n}$-fixed point from Nariman’s construction. The main result of this note circumvents the need to construct actual fixed points for $n=2$.

Theorem 1.1.

The induced map on cohomology

(1.4) $$H^{*}(\mathrm{B}T^{2};\mathbb{Q})\longrightarrow H^{*}(\mathrm{Bun}(T\mathbb{S}^{3},\nu^{*}\gamma_{3})/\!\!/T^{2};\mathbb{Q})$$

is an injection.

Corollary 1.2.

The subring $\mathbb{Q}[p_{1},e]\subset H^{*}(\mathrm{B}\mathrm{Diff}_{+}(\mathbb{S}^{3});\mathbb{Q})$ defined by the Euler and Pontrjagin class injects into $H^{*}(\mathrm{B}\mathrm{Diff}^{\delta}_{+}(\mathbb{S}^{3});\mathbb{Q})$

Proof.

We know that $\mathbb{Q}[p_{1},\ldots,p_{n-1},e]\subset H^{*}(\mathrm{B}\mathrm{Diff}_{+}(\mathbb{S}^{2n-1};\mathbb{Q})$ injects into $H^{*}(\mathrm{B}T^{n};\mathbb{Q})$ under the map $\mathrm{B}T^{n}\rightarrow\mathrm{B}\mathrm{Diff}_{+}(\mathbb{S}^{2n-1})$. It then follows from commutativity of (1.2) and Theorem 1.1 that the composite

$$\mathbb{Q}[p_{1},e]\rightarrow H^{*}(\mathrm{B}\mathrm{Diff}^{\delta}_{+}(\mathbb{S}^{3}))\rightarrow H^{*}(\mathrm{Bun}(T\mathbb{S}^{3},\nu^{*}\gamma_{3})/\!\!/\mathrm{Diff}_{+}(\mathbb{S}^{3})).$$

is a injection, so that the first map is injective as well. ∎

Remark 1.3.

In a forthcoming paper [Pri] that previously appeared as an appendix to [Nar23] we proved that $H^{*}(\mathrm{B}\mathrm{SO}(4);\mathbb{R})$ injects into the smooth cohomology $H^{*}_{\mathrm{sm}}(\mathrm{Diff}_{+}(\mathbb{S}^{3});\mathbb{R})$ using the method developed by Haefliger [Hae78]. This statement follows from Corollary 1.2, but the real homotopy theory computation is still interesting as it showcases an interesting relation between $p_{1}^{2}$ and a continuously varying cohomology class as pointed out by Morita (cf. [Nar23]).

The idea for the proof of Theorem 1.1 is quite simple and we learned it from [AP93, Example 3.1.16]. Given a space $X$ with a torus action so that infinitely many distinct subtori $K_{i}\subset T$ of codimension $1$ occur as stabilizers, the projection $p\colon X/\!\!/T\rightarrow\mathrm{B}T$ induces an injection on cohomology. This is because the map induced by the projection $p:X/\!\!/K_{i}\rightarrow\mathrm{B}K_{i}$ is an injection on cohomology (as there is a $K_{i}$-fixed point) and from the commutativity of the diagram

it follows that

$$\ker(H^{*}(\mathrm{B}T)\rightarrow H^{*}(X/\!\!/T))\subset\ker(H^{*}(\mathrm{B}T)\rightarrow H^{*}(\mathrm{B}K_{i})),$$

where $\ker(H^{*}(\mathrm{B}T)\rightarrow H^{*}(\mathrm{B}K_{i}))$ is an ideal generated by a linear polynomial (with integral coefficients) $f_{i}\in H^{2}(\mathrm{B}T)$. Hence, an element $x\in\ker(H^{*}(\mathrm{B}T)\rightarrow H^{*}(X/\!\!/T))$ is divisible by infinitely may distinct linear polynomials $f_{i}$ as the tori $K_{i}$ are distinct and therefore $x=0$.

So the idea of the proof is to construct bundle maps $\tau\in\mathrm{Bun}(T\mathbb{S}^{2n-1},\nu^{*}\gamma_{2n-1})$ with prescribed stabilizers, which as it turns out is still difficult to do. However, for $n=2$ we are able to show that there exists infinitely many tori $K\subset T^{2}$ with the property that

$$H^{*}(\mathrm{B}K)\hookrightarrow H^{*}(\mathrm{Bun}(T\mathbb{S}^{3},\nu^{*}\gamma_{3})/\!\!/K).$$

without using actual fixed points of the action, which proves Theorem 1.1.

2. Main result

For integers $m,n\in\mathbb{Z}$ we denote by $c_{m,n}\colon S^{1}\rightarrow T^{2}$ the group homomorphism defined by $c_{m,n}(\lambda)=(\lambda^{m},\lambda^{n})$. Codimension 1 tori are indexed by integers $(m,n)\in\mathbb{Z}^{2}/((m,n)\sim-(m,n))$ with $\mathrm{gcd}(m,n)=1$ as the image $K_{m,n}=\mathrm{im}(c_{m,n})$. Theorem 1.1 follows from the following statement which we prove in this section.

Proposition 2.1.

Let $m\in\mathbb{N}$ be odd and $n=m+2$, then the map

$$H^{*}(\mathrm{B}K_{m,n};\mathbb{Q})\longrightarrow H^{*}(\mathrm{Bun}(T\mathbb{S}^{3},\nu^{*}\gamma_{3})/\!\!/K_{m,n};\mathbb{Q})$$

is injective.

We consider the $T^{2}$-action on $\mathbb{S}^{3}$ given by $(\lambda_{1},\lambda_{2})\cdot(z_{1},z_{2})=(\lambda_{1}\cdot z_{1},\lambda_{2}\cdot z_{2})$, and we denote by $(S^{3},K_{m,n})$ the $3$-sphere with the restricted $K_{m,n}$-action. For $m=n=1$ this corresponds to the usual free $S^{1}$-action on $\mathbb{S}^{3}$ which we denote by $(\mathbb{S}^{3},S^{1})$.

Lemma 2.2.

The map

$$\pi_{m,n}\colon(\mathbb{S}^{3},S^{1})\rightarrow(\mathbb{S}^{3},K_{m,n}),\qquad\pi_{m,n}(z_{1},z_{2})=\frac{(z_{1}^{m},z_{2}^{n})}{||(z_{1}^{m},z_{2}^{n})||}$$

is equivariant with respect to $c_{m,n}$.

The pullback $\pi_{m,n}^{*}T\mathbb{S}^{3}$ is an $S^{1}$-equivariant vector bundle over $(\mathbb{S}^{3},S^{1})$ and there is a canonical bundle map $\bar{\pi}_{m,n}:\pi_{m,n}^{*}T\mathbb{S}^{3}\rightarrow T\mathbb{S}^{3}$ covering $\pi_{m,n}$ which is equivariant with respect to $c_{m,n}$. Precomposition with $\bar{\pi}_{m,n}$ defines a map

(2.1) $$\psi\colon\mathrm{Bun}(T\mathbb{S}^{3},\nu^{*}\gamma_{3})\longrightarrow\mathrm{Bun}(\pi^{*}_{m,n}T\mathbb{S}^{3},\nu^{*}\gamma_{3})$$

which is equivariant by construction with respect to $c_{m,n}^{-1}\colon K_{m,n}\rightarrow S^{1}$ if the inverse exists (i.e. if $\gcd(m,n)=1$).

Lemma 2.3.

The map $\psi$ in (2.1) is a rational equivalence.

Proof.

The tangent bundle $T\mathbb{S}^{3}$ is (non-equivariantly) trivial and hence so is the pullback $\pi_{m,n}^{*}T\mathbb{S}^{3}$. This implies that the space of bundle maps is (non-equivariantly) homotopy equivalent to $\mathrm{map}(\mathbb{S}^{3},\overline{\mathrm{B}\Gamma}_{3})$, where $\overline{\mathrm{B}\Gamma}_{d}$ denotes the homotopy fibre of $\nu:\mathrm{B}\mathrm{S}\Gamma_{d}\rightarrow\mathrm{B}\mathrm{GL}_{d}^{+}(\mathbb{R})$. Since $\overline{\mathrm{B}\Gamma}_{3}$ is $4$-connected [Thu74], these mapping spaces are connected and $\psi$ corresponds to the map of mapping spaces induced by precomposition with $\pi_{m,n}$, which is a rational equivalence since $\pi_{m,n}$ is (this follows directly by inspection of the rational models of mapping spaces, for example cf. [Ber15]). ∎

In the following, we denote by $V_{k}=(\mathbb{C},\rho_{k})$ for $k\in\mathbb{Z}$ the complex $S^{1}$-representation given by $\rho_{k}(\lambda)(z)=\lambda^{k}\cdot z$. We obtain $T^{2}$-representations by pulling back along the group homomorphism $\Delta\colon T^{2}\rightarrow S^{1}$ given by $\Delta(\lambda_{1},\lambda_{2})=\lambda_{2}/\lambda_{1}$. In the next section we show that $T\mathbb{S}^{3}$ is $T^{2}$-equivariantly $V$-trivial for $V=\Delta^{*}V_{1}\oplus\mathbb{R}$, where $\mathbb{R}$ denotes the trivial representation (see Lemma 3.1). This implies that $\pi_{m,n}^{*}T\mathbb{S}^{3}$ is $c_{m,n}^{*}V$-trivial and we use the following simple criterion to show that it is also $\mathbb{R}^{3}$-trivial.

Lemma 2.4.

Let $G$ be a Lie group acting on $X$ and $V,W$ be two $G$-representations. There is an isomorphism of equivariant $G$-vector bundles $X\times V\cong_{G}X\times W$ if and only if there is a $G$-equivariant map $X\rightarrow\mathrm{Iso}(V,W)$, where $\mathrm{Iso}(V,W)$ denotes the space of isomorphisms with respect to the conjugation $G$-action.

Proposition 2.5.

Let $n=m+2$, then the pullback $\pi_{m,n}^{*}T\mathbb{S}^{3}$ is isomorphic to the trivial $S^{1}$-bundle $\mathbb{S}^{3}\times\mathbb{R}^{3}$, i.e. $S^{1}$ acts diagonally on $\mathbb{S}^{3}\times\mathbb{R}^{3}$ and trivially on $\mathbb{R}^{3}$.

Proof.

By Lemma 3.1 the tangent bundle $T\mathbb{S}^{3}$ is $T^{2}$-equivariantly $V$-trivial for $V=\mathbb{R}\oplus\Delta^{*}V_{1}$, and hence $\pi_{m,n}^{*}T\mathbb{S}^{3}\cong_{S^{1}}\mathbb{S}^{3}\times(\mathbb{R}\oplus c_{m,n}^{*}\Delta^{*}V_{1})$. The statement follows from Lemma 2.4 if we can construct an $S^{1}$-equivariant map

(2.2) $$\mathbb{S}^{3}\longrightarrow\text{Iso}(\mathbb{R}^{3},\mathbb{R}\oplus c_{m,n}^{*}\Delta^{*}V_{1}).$$

We identify $\text{Iso}(\mathbb{R}^{3},V)\cong\mathrm{GL}_{3}(\mathbb{R})$ and denote by $D(\lambda)\in\mathrm{SO}(2)$ the rotation corresponding to $\lambda\in S^{1}$. Then under this identification the $S^{1}$-action is given left multiplication with

$$\left(\begin{array}[]{c c}1&0\\ 0&D(\lambda^{n-m})\end{array}\right)\in\mathrm{SO}(3).$$

If $n-m=2$ then the double cover $f\colon\mathbb{S}^{3}\rightarrow\mathrm{SO}(3)\subset\mathrm{GL}_{3}(\mathbb{R})$ is $S^{1}$-equivariant with respect to this action by Lemma 3.2 which concludes the proof. ∎

Proof of Prop. 2.1.

It follows from Proposition 2.5 there is an equivariant homeomorphism $\mathrm{Bun}(\pi_{m,n}^{*}T\mathbb{S}^{3},\nu^{*}\gamma_{3})\approx_{S^{1}}\mathrm{Bun}(\mathbb{S}^{3}\times\mathbb{R}^{3},\nu^{*}\gamma_{3})$. The latter has a fixed point by the same argument as in [Nar23, Sect. 3] so that

$$H^{*}(\mathrm{B}S^{1})\rightarrow H^{*}(\mathrm{Bun}(\pi_{m,n}^{*}T\mathbb{S}^{3},\nu^{*}\gamma_{3})/\!\!/S^{1})$$

is injective. If $m\in\mathbb{N}$ is odd and $n=m+2$ then $\gcd(m,n)=1$ and $c_{m,n}\colon S^{1}\rightarrow K_{m,n}$ is an isomorphism, and by Proposition 2.3 we have a commutative diagram

(2.3)

which proves the statement. ∎

3. Two facts about $\mathbb{S}^{3}$

We prove two elementary statements about the $3$-sphere that were used in Proposition 2.5, both relying on the fact that $\mathbb{S}^{3}$ can be identified with the group of unit quaternions.

Lemma 3.1.

The tangent bundle $T\mathbb{S}^{3}$ is $T^{2}$-equivariantly $V$ trivial for $V=\Delta^{*}V_{1}\oplus\mathbb{R}$, i.e. there is an isomorphism of $T^{2}$-equivariant vector bundles $T\mathbb{S}^{3}\cong_{T^{2}}\mathbb{S}^{3}\times V$.

Proof.

Choose an inner product on $\mathbb{H}$ with orthonormal basis given by $1,i,j,k$ and consider $\mathbb{S}^{3}$ as the set of unit quaternions $\{z_{1}+z_{2}j\,|\,z_{i}\in\mathbb{C},|z_{1}|^{2}+|z_{2}|^{2}=1\}\subset\mathbb{H}$ and $T\mathbb{S}^{3}=\{(w_{1},w_{2})\in\mathbb{S}^{3}\times\mathbb{H}\,|\,w_{2}\bot w_{1}\}\subset\mathbb{S}^{3}\times\mathbb{H}$. We choose a basis of $T_{1}\mathbb{S}^{3}$ given by $e_{1}=(1,i)$, $e_{2}=(1,j)$ and $e_{3}=(1,k)$ and define a trivialization of $T\mathbb{S}^{3}$ by

$$\tau\colon\mathbb{S}^{3}\times T_{1}\mathbb{S}^{3}\rightarrow T\mathbb{S}^{3},\quad(p,X_{1})\mapsto Dl_{p}(X_{1})$$

where $l_{w}\colon\mathbb{S}^{3}\rightarrow\mathbb{S}^{3}$ denotes left multiplication with $w=z_{1}+z_{2}j\in\mathbb{S}^{3}$ so that

$$\tau(w,e_{1}) =(w,z_{1}i-z_{2}k),$$

$$\tau(w,e_{2}) =(w,-z_{2}+z_{1}j),$$

$$\tau(w,e_{3}) =(w,z_{2}i+z_{1}k).$$

The tangent bundle $T\mathbb{S}^{3}$ has a left $T^{2}$-action via the differential and one computes that

$$D\mu_{(\lambda_{1},\lambda_{2})}\tau(w,e_{1}) =((\lambda_{1},\lambda_{2})\cdot w,\lambda_{1}z_{1}i-\lambda_{2}z_{2}k),$$

$$D\mu_{(\lambda_{1},\lambda_{2})}\tau(w,e_{2}) =((\lambda_{1},\lambda_{2})\cdot w,-\lambda_{1}z_{2}+\lambda_{2}z_{1}j),$$

$$D\mu_{(\lambda_{1},\lambda_{2})}\tau(w,e_{3}) =((\lambda_{1},\lambda_{2})\cdot w,\lambda_{1}z_{2}i+\lambda_{2}z_{1}k),$$

where $\mu_{\lambda_{1},\lambda_{2}}\colon\mathbb{S}^{3}\rightarrow\mathbb{S}^{3}$ denotes action of $(\lambda_{1},\lambda_{2})\in T^{2}$.

We need to compute $\tau^{-1}\D\mu_{\lambda_{1},\lambda_{2}}\tau$ in order to understand the induced $T^{2}$-action on $\mathbb{S}^{3}\times T_{1}S^{1}$. We see directly that

(3.1) $$D\mu_{(\lambda_{1},\lambda_{2})}\tau(w,e_{1})=\tau((\lambda_{1},\lambda_{2})\cdot w,e_{1}),$$

and for the other two cases let $\psi\in[0,2\pi)$ such that $\lambda_{1}/\lambda_{2}=\cos(\psi)+i\cdot\sin(\psi)$, then a straight forward calculation shows that

(3.2) $$\begin{split}D\mu_{(\lambda_{1},\lambda_{2})}\tau(w,e_{2})&=\cos(\psi)\tau((\lambda_{1},\lambda_{2})\cdot w,e_{2})-\sin(\psi)\tau((\lambda_{1},\lambda_{2})\cdot w,e_{3})\\ D\mu_{(\lambda_{1},\lambda_{2})}\tau(w,e_{3})&=\sin(\psi)\tau((\lambda_{1},\lambda_{2})\cdot w,e_{2})+\cos(\psi)\tau((\lambda_{1},\lambda_{2})\cdot w,e_{3}).\end{split}$$

Finally, observe that the coefficients in (3.2) only depend on $(\lambda_{1},\lambda_{2})$ and not on $w$, so the induced action on $\mathbb{S}^{3}\times T_{1}S^{1}$ is a product action. Hence, $T_{1}S^{1}$ is a $T^{2}$-representation and $f\colon T_{1}S^{1}\rightarrow\mathbb{R}\oplus\Delta^{*}V_{1}$ defined by $f(e_{1})=(1,0)$, $f(e_{2})=(0,1)$ and $f(e_{3})=(0,i)$ is an isomorphism of $T^{2}$-representations. ∎

Lemma 3.2.

The double cover group homomorphism $f\colon\mathbb{S}^{3}\rightarrow\mathrm{SO}(3)$ is $S^{1}$-equivariant with respect to the $S^{1}$-action on $\mathrm{SO}(3)$ given by left multiplication with

$$\left(\begin{array}[]{c c}1&0\\ 0&D(\lambda^{2})\end{array}\right)\in\mathrm{SO}(3).$$

for $\lambda\in S^{1}$.

Proof.

Identify $\mathbb{R}^{3}$ with the subspace of pure quaternions $\{xi+yj+zk\,|x,y,z\in\mathbb{R}\}\subset\mathbb{H}$ and $\mathbb{S}^{3}\subset\mathbb{H}$ with the unit quaternions, i.e. $z=z_{1}+z_{2}j\in\mathbb{S}^{3}$ and $w=ix+z_{3}j\in\mathbb{R}^{3}$ for $x\in\mathbb{R}$ and $z_{1},z_{2},z_{3}\in\mathbb{C}$. We then define

$$f(z)(ix+jy+kz)\colon=z\cdot w\cdot\bar{z},$$

which is a group homomorphism and preserves the inner product on $\mathbb{R}^{3}$. Observe that $zj=j\bar{z}$ for $z\in\mathbb{C}$ and hence for any pure quaternion $w=ix+zj$ and $\lambda\in S^{1}\subset\mathbb{C}$ we have that

$$\lambda w\bar{\lambda}=ix+\lambda zj\bar{\lambda}=ix+\lambda z\lambda j=ix+\lambda^{2}zj,$$

which corresponds to the rotation by

(3.5) $$\left(\begin{array}[]{c c}1&0\\ 0&D(\lambda^{2})\end{array}\right)\in\mathrm{SO}(3).$$

Finally, since

$$f(\lambda z)(w) =\lambda z\cdot w\cdot\bar{z}\bar{\lambda}=\lambda f(z)(w)\bar{\lambda},$$

this corresponds to the left multiplication of $f(z)$ by (3.5). ∎

Acknowledgements

I would like to thank Sam Nariman for introducing me to this aspect of foliation theory and posing many interesting questions in our conversations. I also thank Ronno Das for numerous helpful discussions and reading this note. This research was supported by the Knut and Alice Wallenberg foundation through grant no. 2019.0519.