---
title: Cherry Vector Fields
url: https://www.emergentmind.com/topics/cherry-vector-fields
type: topic
---

# Cherry Vector Fields

Searching arXiv for recent and foundational papers on Cherry vector fields, Cherry flows, and Cherry maps.
Cherry vector fields, also called Cherry fields, are \(C^\infty\) vector fields on the torus \(T^2\) without closed trajectories and with exactly two singularities, a sink and a saddle, both hyperbolic [1203.6301]. The flow generated by such a field is a Cherry flow. In the modern literature, Cherry vector fields are studied primarily through the structure of their trajectories on \(T^2\), the existence of global Poincaré sections, and the induced one-dimensional circle maps with a flat interval known as Cherry maps [1203.6301, 2508.12350].

## 1. Definition and basic dynamical model

The standard setting is the two-dimensional torus,
\[
T^2 \simeq S^1 \times [0,1]/\sim,\qquad (s,0)\sim (s,1).
\]
Within this setting, a Cherry field is characterized by three phase-portrait properties: the ambient manifold is \(T^2\), there are no periodic orbits, and the singular set consists of exactly one hyperbolic sink and one hyperbolic saddle [1203.6301]. The 2025 treatment denotes the set of such vector fields by
\[
\mathfrak X_c(\mathbb T^2)
\]
and uses the same toral definition [2508.12350].

A standard structural fact recalled in the literature is that every Cherry flow admits a Poincaré global section, defined as a transversal, simple, closed \(C^\infty\) curve \(\Sigma\) intersecting every one-dimensional trajectory of the flow [1203.6301]. Another recalled fact is that every Cherry field has a quasi-minimal set, namely the closure of a non-trivial recurrent trajectory, and that this set is locally homeomorphic to the Cartesian product of a Cantor set and a segment [1203.6301]. This makes Cherry dynamics a canonical example of singular torus dynamics in which recurrence persists despite the absence of periodic orbits.

The singular configuration is essential. In the cited treatments, the sink and saddle are not incidental defects of an otherwise regular foliation; they are the mechanism by which the torus flow acquires a nontrivial return map with a flat interval and irrational rotation behavior [1203.6301]. A plausible implication is that Cherry vector fields occupy a boundary regime between smooth circle-like dynamics and singular surface flows with recurrent invariant sets.

## 2. Global sections and induced Cherry maps

The first-return construction is the basic reduction from a Cherry vector field to one-dimensional dynamics. Let \(X\) be a Cherry field and let \(\Sigma\) be a global Poincaré section. After identifying the complement of \(\Sigma\) with an annulus, one considers points on a transversal circle whose positive orbit reaches the opposite boundary circle. If \(W\subset S^1\times\{0\}\) denotes the set of such points and \(t(x)\) the first positive return time, the induced map is
\[
f(x)=X_{t(x)}(x)\in S^1,
\]
or, in equivalent flow notation,
\[
f(x)=\Phi_X^{t(x)}(x)
\]
on the domain of definition [1203.6301, 2508.12350]. On the complementary interval \(U\), the map is extended by collapsing \(U\) to a single point. Thus the return map is continuous on the circle, \(C^\infty\) outside the boundary points of \(U\), and constant on \(U\) [1203.6301].

This places the return map in the class denoted \(\mathcal L\) or \(\mathscr L\), consisting of order-preserving degree-one circle maps with a flat interval. The defining local model is an interval \(U=(a,b)\) such that \(f(U)\) is one point, \(f\) restricts to a diffeomorphism off \(U\), and near the endpoints there are power-law forms
\[
f(x)=h_l\big((x-a)^{l_1}\big),\qquad
f(x)=h_r\big((x-b)^{l_2}\big),
\]
with \(h_l,h_r\) local diffeomorphisms [2205.12915, 2508.12350]. In the Cherry-flow setting, the first return map is symmetric, and the critical exponent is determined by the saddle eigenvalues \(\lambda_1>0>\lambda_2\):
\[
\alpha=\frac{|\lambda_2|}{\lambda_1}.
\]
More precisely, on a right-sided neighborhood of \(b\),
\[
f(x)=h_r\big((x-b)^\alpha\big),
\]
and analogously on a left-sided neighborhood of \(a\) [1203.6301].

The rotation number of a lift \(F\) of the induced map is
\[
\rho(f):=\lim_{n\to\infty}\frac{F^n(0)}{n}\pmod 1,
\]
and in the Cherry-flow case it is irrational because the underlying flow has no periodic trajectories [1203.6301, 2205.12915]. This semiconjugates Cherry dynamics to irrational rotation combinatorics while retaining a singular flat interval.

## 3. Phase transition in return-map geometry

The most detailed quantitative results presently cited for Cherry vector fields pass through the associated circle map with a flat interval. In the symmetric setting, the critical threshold is the exponent
\[
l=2.
\]
The geometric scaling sequence
\[
\tau_n:=\frac{|(\underline{0},\underline{q_n})|}{|(\underline{0},\underline{q_{n-2}})|}
\]
measures the geometry near the flat interval. When \(\tau_n\to 0\), the geometry is called degenerate; when \(\tau_n\) is bounded away from zero, the geometry is called bounded [1203.6301].

The sharp transition is summarized by the theorem that if the critical exponent \(l\le 2\), then the scalings \(\tau_n\) tend to zero at least exponentially fast, whereas for maps with rotation number of bounded type and critical exponent \(l>2\), the sequence \(\tau_n\) is bounded away from zero [1203.6301]. The corresponding invariant set for the map is the non-wandering set
\[
\Omega = S^1\setminus \bigcup_{i=0}^\infty f^{-i}(U).
\]
For \(l\in(1,2]\), its Hausdorff dimension is \(0\); for \(l>2\), its Hausdorff dimension is strictly greater than \(0\) [1203.6301].

Translated back to a Cherry flow, the threshold becomes
\[
\frac{|\lambda_2|}{\lambda_1}=2
\quad\Longleftrightarrow\quad
|\lambda_2|=2\lambda_1.
\]
If \(X\) is a Cherry vector field with saddle eigenvalues \(\lambda_1>0>\lambda_2\), and if
\[
|\lambda_2|>2\lambda_1
\]
and the first return map has rotation number of bounded type, then the quasi-minimal set has Hausdorff dimension strictly greater than \(1\) [1203.6301]. The mechanism is the local product structure recalled in the same source: near a global section, the quasi-minimal set is \(C^2\)-equivalent to
\[
I\times \Omega.
\]
This makes the return-map phase transition into a geometric phase transition for the flow itself.

## 4. Pairs of Cherry vector fields and bi-Lagrangian structures

A distinct line of work places Cherry vector fields in a symplectic and foliation-theoretic framework. On a symplectic surface, a pair of transversal one-dimensional foliations defines a bi-Lagrangian structure
\[
(M,\omega,\mathcal F_1,\mathcal F_2),
\]
where \(\omega\) is closed and nondegenerate and \(\mathcal F_1,\mathcal F_2\) are transversal Lagrangian foliations [2205.12915]. On \(\mathbb T^2\), a pair of transversal vector fields without singularity therefore defines a bi-Lagrangian structure [2205.12915].

Cherry vector fields introduce an important complication: they necessarily have singularities. The cited torus papers therefore separate two stories. One story concerns nonsingular transverse fields and bi-Lagrangian geometry. The other concerns Cherry vector fields and Cherry maps. The distinction is explicit: the bi-Lagrangian setup uses nonsingular transverse fields, whereas Cherry vector fields necessarily have singularities [2205.12915].

Even so, pairs of Cherry vector fields enter the theory through return-map constructions. A single Cherry vector field generates a symmetric Cherry map. A pair of Cherry vector fields can generate a broader class of maps with asymmetric critical exponents [2205.12915, 2508.12350]. Concretely, if \(X_1,X_2\) generate \(f_1,f_2\in\mathcal L\) or \(\mathscr L\), with flat pieces
\[
U_1=(a_1,b_1),\qquad U_2=(a_2,b_2),
\]
symmetric critical exponents \((l_1,l_1)\) and \((l_2,l_2)\), and if
\[
f_1(U_1)=f_2(U_2),\qquad a_1<a_2\le b_1<b_2,
\]
then the patched map has flat interval
\[
U=(a_1,b_2)
\]
and critical exponents
\[
(l_1,l_2)
\]
or, in the 2025 notation,
\[
(\ell_1,\ell_2)
\]
[2205.12915, 2508.12350]. The proved statement is not that every Cherry map arises this way, but that some maps with different left and right exponents do.

## 5. Hess connections, prolongations, and induced conjugacy

The bi-Lagrangian viewpoint leads to a canonical connection. For a bi-Lagrangian structure, the Hess connection is the unique torsion-free connection preserving both foliations and parallelizing the symplectic form:
\[
\nabla\omega=0.
\]
In the 2025 torus paper, if \((X,Y)\) is a pair of transversal Cherry vector fields with the same singularities, then after removing the singular set
\[
\mathring{\mathbb T^2}=\mathbb T^2\setminus \mathrm{Sing}(X,Y),
\]
the pair determines foliations \(\mathcal F_X,\mathcal F_Y\) and hence a bi-Lagrangian structure on the punctured torus, with Hess connection \({}^{X,Y}\nabla\) [2508.12350]. Under an additional compatibility condition on all pairs generating the same Cherry map, the restriction \({}^{X,Y}\nabla_{|\mathbb S^1}\) is defined as a linear connection associated to the map [2508.12350].

The same paper proves that the push-forward action on pairs of Cherry vector fields induces conjugation on the associated subclass of Cherry maps. Writing
\[
\mathfrak X_c^2(\mathbb T^2)=\mathfrak X_c(\mathbb T^2)\times \mathfrak X_c(\mathbb T^2),
\]
the action
\[
(\psi,(X_1,X_2))\mapsto (\psi_*X_1,\psi_*X_2)
\]
induces
\[
(\varphi,f)\mapsto \varphi\circ f\circ \varphi^{-1}
\]
on the subset \(\mathscr L_c\subset \mathscr L\) of Cherry maps generated by pairs of Cherry vector fields [2508.12350]. The proof uses the standard flow identity
\[
\left.\frac{d}{dt}\right|_{t=0}\bigl(\psi\circ \Phi_X^t\circ \psi^{-1}(y)\bigr)=(\psi_*X)_y,
\]
so the first-hit map transforms by conjugacy [2508.12350].

These papers also develop prolongations of bi-Lagrangian structures to \(TM\) and \(T^*M\), including
\[
(T^*M,d\theta,N^*\mathcal F_1,N^*\mathcal F_2),\qquad
(T^*M,\pi^*\omega+d\theta,N^*\mathcal F_1,N^*\mathcal F_2),\qquad
(TM,\omega^c,\mathcal F_1^c,\mathcal F_2^c),
\]
but their relevance to Cherry vector fields is indirect: the Cherry-map results are formulated mainly through push-forward dynamics and return maps rather than through the tangent- and cotangent-bundle liftings themselves [2205.12915, 2508.12350].

## 6. Scope, limitations, and adjacent frameworks

Several misconceptions are addressed explicitly in the cited literature. First, Cherry vector fields should be distinguished from Cherry maps. A Cherry vector field is a singular torus vector field with no closed trajectories; a Cherry map is a degree-one circle map with a flat interval and endpoint critical exponents [1203.6301, 2508.12350]. The return map construction gives a one-way link from flow to map, but the cited papers do not prove that every Cherry map arises from a Cherry flow or from a pair of vector fields on \(\mathbb T^2\) [2205.12915, 2508.12350].

Second, bi-Lagrangian structures on \(\mathbb T^2\) come from pairs of transversal nonsingular vector fields, whereas Cherry vector fields necessarily possess singularities. The relation between the two theories is therefore motivational and partial rather than a single unified theorem [2205.12915]. The punctured-torus construction in [2508.12350] is one way of reconciling this tension.

Third, not every analytic framework for vector fields is directly applicable to Cherry dynamics. The inverse-problem paper "Ray Transforms and Vector Fields" studies reconstruction of a function from integrals over trajectories of planar vector fields by complexifying
\[
X_\lambda=a(z,\lambda)\partial_z+b(z,\lambda)\partial_{\bar z}
\]
and imposing type-\(H\) or \(H_{k,l}\) conditions [1102.1736]. Cherry vector fields are not explicitly mentioned there, and the stated assessment is that direct applicability is generally no; partial or local applicability is possible only on a simply connected planar subdomain, away from singularities and recurrent obstructions, and only if the required complexification and foliation hypotheses can be verified [1102.1736].

Finally, the phrase “vector field” may refer to entirely different frameworks. The scheme-theoretic paper "Vector fields and differential schemes" defines vector fields as derivations of the structure sheaf and develops leaves and trajectories for schemes [1011.1806]. That theory is not about Cherry vector fields in the dynamical-systems sense. This distinction matters because Cherry dynamics is tied to \(C^\infty\) torus flows, Poincaré sections, return maps, quasi-minimal sets, and saddle-source singular geometry, none of which is the object of [1011.1806].

Taken together, the cited works present Cherry vector fields as a toral singular-flow class defined by a hyperbolic sink, a hyperbolic saddle, and absence of periodic orbits; reduced analytically to monotone circle maps with a flat interval; organized geometrically by a sharp threshold at \(|\lambda_2|=2\lambda_1\); and, in recent work, embedded into a bi-Lagrangian and Hess-connection framework for pairs of Cherry vector fields [1203.6301, 2205.12915, 2508.12350].

Source: https://www.emergentmind.com/topics/cherry-vector-fields