betaFeedback

WorksMathematics

Nearby Lagrangians in the cotangent bundle of the Klein bottle

Abstract

We prove the nearby Lagrangian conjecture for the Klein bottle: every closed connected embedded exact Lagrangian submanifold of its cotangent bundle is Hamiltonian isotopic to the zero section by a compactly supported isotopy. The proof passes to the torus double cover and builds on the theorem of Dimitroglou Rizell, Goodman, and Ivrii for the two-dimensional torus. The main step is to construct an isotopy compatible with the covering involution. We achieve this using holomorphic sphere rulings in S2×S2S^2\times S^2 adapted to the zero section, allowing the construction to descend to the Klein bottle.

Introduction

The nearby Lagrangian conjecture asks whether a closed embedded exact Lagrangian in a cotangent bundle of a closed manifold is Hamiltonian isotopic to its zero section. In this paper, we prove the conjecture in the case of the Klein bottle.

Theorem 1.1. Let KK be the Klein bottle and let (T∗K,ωK=dλK)(T^*K,\omega_K = d\lambda_K) be its cotangent bundle. Every closed connected embedded exact Lagrangian submanifold of (T∗K,dλK)(T^*K,d\lambda_K) is Hamiltonian isotopic to the zero section by a smooth isotopy supported in one compact subset of T∗KT^*K.

Earlier work establishes topological and categorical constraints on closed exact Lagrangians in cotangent bundles. For a simply connected spin base, Fukaya, Seidel, and Smith [10] proved that the cotangent projection of a closed exact spin Lagrangian with vanishing Maslov class induces an isomorphism on complex cohomology. Under the same hypotheses, Nadler [Nad09] identified the Lagrangian with the zero section in the Fukaya category over C\mathbb{C}, up to a grading shift. The work of Abouzaid [1], Kragh [Kra13], and Abouzaid–Kragh [4] shows that the cotangent projection of every closed exact Lagrangian is a simple homotopy equivalence.

For closed surfaces, Hind proved the conjecture for S2S^2 [12], and Li and Wu gave an alternative proof using symplectic cutting [LW12, Theorem 6.6]. In T∗RP2T^*\mathbb{R}P^2, Hind proved Hamiltonian uniqueness for Lagrangian projective planes homotopic to the zero section [13 Corollary 8]. Hind, Pinsonnault, and Wu proved that the space of Lagrangian projective planes in T∗RP2T^*\mathbb{R}P^2 is weakly contractible [HPW16, Theorem 1.3], and Adaloglou gave a compactification proof of Hamiltonian uniqueness [Ada25]. Dimitroglou Rizell, Goodman, and Ivrii established the conjecture for T2T^2 [7 Theorem B], providing the starting point of our argument.

For the Klein bottle, Atallah, Chassé, Leclercq, and Shelukhin proved that every Lagrangian Klein bottle in T∗KT^*K is HH-exact, meaning that symplectic area vanishes on relative two-dimensional homology classes [2 Theorem 44]. This condition is weaker than exactness of the restricted Liouville form, which is assumed in Theorem 1.1. Their argument also uses the torus double cover.

Outline of the proof. The main idea of this paper is to pass the problem of the Klein bottle to its double cover. Let L⊂T∗KL \subset T^*K be as in Theorem 1.1. The Abouzaid–Kragh theorem implies that its full lift E⊂T∗T2E \subset T^*T^2 is a connected exact torus. Write σ:T∗T2→T∗K\sigma: T^*T^2 \to T^*K for the covering involution and ZZ for the zero section of T∗T2T^*T^2. We first use the torus theorem of Dimitroglou Rizell, Goodman, and Ivrii [7 Theorem B] to obtain a compactly supported Hamiltonian diffeomorphism ψ\psi taking E⊂T∗T2E \subset T^*T^2 to the zero section ZZ of T∗T2T^*T^2. Descent to T∗KT^*K requires an isotopy commuting with σ\sigma. As a first step, Lemma 2.2 allows us to choose ψ\psi so that the conjugated involution ρ=ψσψ−1\rho= \psi\sigma\psi^{-1} agrees with σ\sigma near ZZ.

To construct this path, we consider the equal-area compactification Y=S2×S2Y = S^2 \times S^2 of a square cotangent domain in T∗T2T^*T^2, where ZZ becomes the standard Clifford torus. Gromov’s theorem gives two transverse sphere rulings for each compatible almost complex structure on YY [11 Section 2.4.A1]. Our geometric input is an equivariant adaptation of the neck-stretching and smoothing constructions in [7 Sections 4–5]. We construct a single ρ\rho-invariant compatible almost complex structure whose two rulings contain circle families of spheres meeting ZZ in its two families of coordinate circles (Proposition 3.1). Intersections of ruling leaves [8 pp. 11–13], followed by the equivariant annular straightening of Lemma A.1, then give a smooth path of diffeomorphisms ft:Y→Yf_t : Y \to Y satisfying

f0=ψ−1,ftρ=σft,f1(Z)=Z.f_0 = \psi^{-1}, \qquad f_t\rho= \sigma f_t, \qquad f_1(Z) = Z.

Using the convex-form argument of [8 pp. 12–13], we apply a relative Moser deformation to make this path symplectic while preserving its initial value, conjugacy relation, and zero-section endpoint condition. A Hamiltonian correction near the caps then gives an equivariant symplectic isotopy with common compact support, whose descent to T∗KT^*K is Hamiltonian because Hc1(T∗K;R)=0H^1_c(T^*K;\mathbb{R}) = 0.

Structure of the paper. In § 2, we normalize the involution near the zero section, state the central equivariant symplectic isotopy result, and introduce the square compactification together with the extended involution and intersection properties. In § 3, we construct the two rulings with the prescribed intersections with the zero section. In § 4, we use the two rulings to get the desired smooth isotopy and apply Moser deformation to make it symplectic. Finally, we prove the main theorem in § 5.

Acknowledgments. The authors thank Nik Adaloglou, Johannes Hauber, and Jialiang Chen for helpful discussions. The authors formulated the research problem, proposed the double-cover approach, and outlined a program using an equivariant adaptation of the foliation method in [7]. They supplied concrete mathematical directions for constructing and using the foliations in the compactified setting. Under the authors’ direction, a generative large language model produced a preliminary draft of more than 40 pages. The authors guided the mathematical development throughout, examined and revised the generated arguments, and extensively revised the manuscript for clarity and precision. They take full responsibility for the mathematical content of the final manuscript.

The orientation cover and compactification

The orientation cover and normalization

Let ZZ denote the zero section of T∗T2T^*\mathbb{T}^2. We use coordinates

(x,y,px,py)∈(R/Z)2×R2,λ=px dx+py dy,ω=dλ.(x,y,p_x,p_y) \in(\mathbb{R}/\mathbb{Z})^2 \times\mathbb{R}^2,\qquad\lambda= p_x\,\mathrm{d}x + p_y\,\mathrm{d}y,\qquad\omega= \mathrm{d}\lambda.

Write q=(x,y)q=(x,y) and p=(px,py)p=(p_x,p_y) for the base and fiber coordinates. The cotangent lift of the orientation-cover involution is

σ(x,y,px,py)=(x+12,−y,px,−py).(1)\sigma(x,y,p_x,p_y)=\left(x+\frac{1}{2},-y,p_x,-p_y\right). \tag*{(1)}

Therefore, T∗T2/⟨σ⟩=T∗KT^*\mathbb{T}^2/\langle\sigma\rangle=T^*K. Throughout the paper, an isotopy is called compactly supported if all its maps are the identity outside one common fixed compact set. We use the convention ιXHω=−dH\iota_{X_H}\omega=-\mathrm{d}H. We write Ham⁡c(T∗T2)\operatorname{Ham}_{c}(T^*\mathbb{T}^2) for the group of time-one maps of compactly supported Hamiltonian isotopies.

The key technical part of the paper is to construct the following equivariant symplectic isotopy:

Proposition 2.1. Let ψ∈Ham⁡c(T∗T2)\psi\in\operatorname{Ham}_{c}(T^*\mathbb{T}^2), and suppose that ρ=ψσψ−1\rho=\psi\sigma\psi^{-1} agrees with σ\sigma on a neighborhood of ZZ. There is a smooth symplectic isotopy hth_t, supported in one common compact subset of T∗T2T^*\mathbb{T}^2, such that

h0=id⁡,htσ=σht(0≤t≤1),h1(ψ−1(Z))=Z.h_0=\operatorname{id},\qquad h_t\sigma=\sigma h_t\quad(0\leq t\leq1),\qquad h_1(\psi^{-1}(Z))=Z.

We will prove the proposition in Sections 3–4. We first show that any closed exact Lagrangian of T∗KT^*K gives rise to such a ψ\psi that satisfies the assumptions of Proposition 2.1.

Lemma 2.2. If L⊂T∗KL\subset T^*K is a closed connected embedded exact Lagrangian, then its inverse image E⊂T∗T2E\subset T^*\mathbb{T}^2 is a connected exact torus. Moreover, there is a ψ∈Ham⁡c(T∗T2)\psi\in\operatorname{Ham}_{c}(T^*\mathbb{T}^2) with

ψ(E)=Z,ψσψ−1=σnear Z.(2)\psi(E)=Z,\qquad\psi\sigma\psi^{-1}=\sigma\quad\text{near }Z. \tag*{(2)}

Proof. Step 1: Lifting the Lagrangian. The cotangent projection L→KL\to K is a homotopy equivalence by [4 Theorem 1]. Its induced map on fundamental groups therefore pulls the orientation character of KK back to a surjection onto Z/2\mathbb{Z}/2. The associated cover E→LE\to L is connected and has fundamental group Z2\mathbb{Z}^2. The classification of closed surfaces then identifies EE with a torus. Pulling back a primitive for λK∣L\lambda_K|_L proves exactness. By [7 Theorem B], there is a ϕ∈Ham⁡c(T∗T2)\phi\in\operatorname{Ham}_{c}(T^*\mathbb{T}^2) with ϕ(E)=Z\phi(E)=Z.

Step 2: Normalizing the action on the zero section. Let ρ0=ϕσϕ−1\rho_0=\phi\sigma\phi^{-1}. This involution acts freely on ZZ, and

(ρ0∣Z)∗=(100−1)on H1(Z;Z),(\rho_0|_Z)_*= \begin{pmatrix} 1 & 0\\ 0 & -1 \end{pmatrix} \quad\text{on }H_1(Z;\mathbb{Z}),

since Z↪T∗T2Z\hookrightarrow T^*\mathbb{T}^2 is a homotopy equivalence and ϕ\phi is isotopic to the identity. The quotient of ZZ by ρ0\rho_0 is a Klein bottle, and its orientation cover gives a diffeomorphism $c:Z\to Z which conjugates ρ0∣Z\rho_0|_Z to σ∣Z\sigma|_Z. The matrix c∗c_* commutes with diag⁡(1,−1)\operatorname{diag}(1,-1), so it is diagonal with entries ±1\pm1. The corresponding linear torus map dd commutes with σ∣Z\sigma|_Z, including its half-translation. Replacing cc by d−1cd^{-1}c then makes c∗c_* the identity. It follows from the mapping class group calculation for the torus [9 Theorem 2.5] that cc is isotopic to the identity.

Let ctc_t be an isotopy from id⁡Z\operatorname{id}_Z to cc, and let vt=c˙t∘ct−1v_t=\dot{c}_t\circ c_t^{-1} be its generating vector field. The cotangent lift of ctc_t is generated by Ht(q,p)=p(vt(q))H_t(q,p)=p(v_t(q)). Cutting off HtH_t in the fibers gives a compactly supported Hamiltonian isotopy κt\kappa_t preserving ZZ and satisfying κt∣Z=ct\kappa_t|_Z=c_t. Define

ρ1=κ1ρ0κ1−1.\rho_1=\kappa_1\rho_0\kappa_1^{-1}.

Since κ1∣Z=c\kappa_1|_Z=c, we have ρ1∣Z=σ∣Z\rho_1|_Z=\sigma|_Z.

Step 3: Normalizing the germ along the zero section. By the equivariant coisotropic neighborhood theorem [6 Section 30.2, footnote 19], the identity map of ZZ extends to a germ of symplectomorphisms gg of a neighborhood of ZZ, satisfying

g∣Z=id⁡Z,gρ1g−1=σnear Z.g|_Z=\operatorname{id}_Z,\qquad g\rho_1g^{-1}=\sigma\quad\text{near }Z.

We realize gg by a compactly supported Hamiltonian diffeomorphism. Let δs(q,p)=(q,sp)\delta_s(q,p)=(q,sp). Since gg is symplectic and fixes ZZ pointwise,

Dg∣(q,0)=(IB(q)0I),B(q)=B(q)T.Dg|_{(q,0)}= \begin{pmatrix} I & B(q)\\ 0 & I \end{pmatrix}, \qquad B(q)=B(q)^{T}.

Taylor’s formula therefore shows that the germs of symplectomorphisms

gs=δs−1∘g∘δs,0<s≤1,g_s=\delta_s^{-1}\circ g\circ\delta_s,\qquad0<s\leq1,

extend smoothly to g0=id⁡g_0=\operatorname{id}. On a common neighborhood of ZZ, set

Vs=∂sgs∘gs−1.V_s=\partial_sg_s\circ g_s^{-1}.

Since gsg_s is symplectic and fixes ZZ pointwise, the one-form ιVsω\iota_{V_s}\omega is closed and vanishes along ZZ. This implies VsV_s is generated by a family of Hamiltonians −Fs-F_s satisfying Fs∣Z=0F_s|_Z=0. Cutting off the Hamiltonians −Fs-F_s by a fixed cutoff function equal to one near ZZ, we obtain gsg_s on a smaller neighborhood for all ss. Its time-one map G∈Ham⁡c(T∗T2)G\in\operatorname{Ham}_c(T^*T^2) therefore fixes ZZ pointwise and agrees with gg near ZZ.

Finally, set ψ=G∘κ∘ϕ\psi=G\circ\kappa\circ\phi. Then ψ∈Ham⁡c(T∗T2)\psi\in\operatorname{Ham}_c(T^*T^2), ψ(E)=Z\psi(E)=Z, and

ψσψ−1=Gρ1G−1=σnear Z.\psi\sigma\psi^{-1}=G\rho_1G^{-1}=\sigma\quad\text{near }Z.

This proves (2).

The square compactification

Fix ψ\psi and ρ\rho as in Proposition 2.1. Choose R>0R>0 so that supp⁡ψ\operatorname{supp}\psi lies in

U=(R/Z)2×(−R,R)2.U=(\mathbb{R}/\mathbb{Z})^2\times(-R,R)^2.

Collapsing the boundary circles of each annular factor compactifies UU to

(Y,ω)=S2(2R)×S2(2R),(Y,\omega)=S^2(2R)\times S^2(2R),

where each sphere has symplectic area 2R2R. Let a±a_\pm be the poles of the first sphere factor, corresponding to px=±Rp_x=\pm R, and let b±b_\pm be the poles of the second, corresponding to py=±Rp_y=\pm R. The four caps are

D1±={a±}×S2,D2±=S2×{b±},D_1^\pm=\{a_\pm\}\times S^2,\qquad D_2^\pm=S^2\times\{b_\pm\},

and we set D=⋃i,±Di±D=\bigcup_{i,\pm}D_i^\pm. Their four intersection points (a±,b±)(a_\pm,b_\pm) are called the corners. The involution in (1) extends to YY as σ=σ1×σ2\sigma=\sigma_1\times\sigma_2, where in the annular coordinates

σ1(x,px)=(x+12,px),σ2(y,py)=(−y,−py).\sigma_1(x,p_x)=\left(x+\frac{1}{2},p_x\right),\qquad\sigma_2(y,p_y)=(-y,-p_y).

On the compactified spheres, these are rotations by π\pi about the polar axis of the first factor and an equatorial axis of the second factor, respectively. Thus σ1\sigma_1 fixes the poles a±a_{\pm}, whereas σ2\sigma_2 exchanges the poles b±b_{\pm} and fixes exactly two equatorial points, given by (y,py)=(0,0)(y,p_y)=(0,0) and (12,0)(\frac{1}{2},0). The map ψ\psi extends as the identity near DD, so ρ=ψσψ−1\rho=\psi\sigma\psi^{-1} also extends to YY and agrees with σ\sigma near DD. Note that neither ρ\rho nor σ\sigma fixes any corner: both exchange (a±,b+)(a_{\pm},b_+) and (a±,b−)(a_{\pm},b_-).

Let

A=[S2×{pt}],B=[{pt}×S2].A=[S^2\times\{\mathrm{pt}\}], \qquad B=[\{\mathrm{pt}\}\times S^2].

The caps D1±D_1^{\pm} represent BB, and D2±D_2^{\pm} represent AA, and ZZ is the product of the equators. Fix a σ\sigma-invariant ω\omega-compatible product almost complex structure JstdJ_{\mathrm{std}} satisfying Jstd∂x=−∂pxJ_{\mathrm{std}}\partial_x=-\partial_{p_x} and Jstd∂y=−∂pyJ_{\mathrm{std}}\partial_y=-\partial_{p_y} near ZZ. An almost complex structure on YY is called standard near a set if it agrees with JstdJ_{\mathrm{std}} on a neighborhood of that set.

Fix a point p∈Zp\in Z, let Hi±H_i^{\pm} be the four standard symplectic disks with boundary on ZZ, whose classes form a standard basis of H2(Y,Z;Z)H_2(Y,Z;\mathbb{Z}), and Hi±∩Di±=1H_i^{\pm}\cap D_i^{\pm}=1. Each of the disks has area RR and Maslov index 2. Let e1=[x]e_1=[x] and e2=[y]e_2=[y] be the positively oriented coordinate circle classes in H1(Z;Z)H_1(Z;\mathbb{Z}). Then

∂Hi+=−ei,∂Hi−=ei.(3)\partial H_i^+=-e_i,\qquad\partial H_i^-=e_i. \tag*{(3)}

Figure 1 explains the cap and disk labels.

Figure 1. Momentum projection of the square compactification. Edge preimages are the indicated cap spheres, and the center has fiber Z≅T2Z\cong T^2. Each segment is the projection of Hi±H_i^{\pm}, with the complementary base coordinate fixed. The indicated circle collapses along an edge and both circles collapse at a corner.

For every ω\omega-compatible almost complex structure JJ on YY, the JJ-holomorphic spheres in AA and BB form two smooth transverse sphere rulings. Each point of YY lies on exactly one leaf of each ruling, and leaves in different classes intersect once transversely. These rulings also vary smoothly with a smooth family of JJ. If JJ is standard near DD, the four caps are ruling leaves.

Neck stretching argument

The goal of this section is to construct a ρ\rho-invariant almost complex structure for which the two sphere rulings have prescribed intersections with ZZ. This will be used in the reconstruction of equivariant symplectic isotopy in § 4.

Proposition 3.1. There is a smooth ρ\rho-invariant ω\omega-compatible almost complex structure J1J_1 on YY, standard near D∪ZD \cup Z, with the following property. For every c,d∈R/Zc,d \in\mathbb{R}/\mathbb{Z}, the circles

{(x,c,0,0):x∈R/Z},{(d,y,0,0):y∈R/Z}\{(x,c,0,0):x\in\mathbb{R}/\mathbb{Z}\}, \qquad\{(d,y,0,0):y\in\mathbb{R}/\mathbb{Z}\}

are, respectively, the intersections of ZZ with a J1J_1-holomorphic sphere in class AA and a J1J_1-holomorphic sphere in class BB.

We stretch a neck along ZZ and use the area calculation in Subsection 2.2 to determine the limiting configurations of ruling spheres through ZZ. Their exterior planes form four circle families, which we straighten equivariantly and join in opposite pairs across ZZ to symplectic spheres, following [7 Sections 5.3–5.4] We then choose a compatible almost complex structure making both sphere families holomorphic. Their coordinate-circle intersections identify ZZ with a product of circles in the two leaf spaces, which will make the endpoint of the reconstruction in § 4.1 preserve ZZ.

Limiting spheres and the four caps

Equip Z=T2Z=T^2 with the flat metric dx2+dy2dx^2+dy^2, and write r=∣p∣r=|p| in the cotangent coordinates. Choose a ρ\rho-invariant Weinstein neighborhood UU of ZZ, disjoint from DD, on which ρ=σ\rho=\sigma, and a cosphere bundle Σ⊂U\Sigma\subset U. We stretch a neck along Σ\Sigma using the construction of [7 Section 2.5]. Attaching cylindrical ends to the two sides of Σ\Sigma identifies them with T∗T2T^*T^2 and Y∖ZY\setminus Z, respectively, with the fixed normalization of the Weinstein coordinates used in that construction. We first specify the almost complex structures on these two spaces and their cylindrical ends.

On S∗T2S^*T^2, write p=(cos⁡θ,sin⁡θ)p=(\cos\theta,\sin\theta); the contact form and Reeb vector field are

α0=λ∣S∗T2=cos⁡θ dx+sin⁡θ dy,Rα0=cos⁡θ ∂x+sin⁡θ ∂y.\alpha_0=\lambda|_{S^*T^2}=\cos\theta\,dx+\sin\theta\,dy,\qquad R_{\alpha_0}=\cos\theta\,\partial_x+\sin\theta\,\partial_y.

On R×S∗T2\mathbb{R}\times S^*T^2, identified with T∗T2∖ZT^*T^2\setminus Z by ∣p∣=es|p|=e^s, use the cylindrical almost complex structure

Jcyl∂x=−∣p∣∂px,Jcyl∂y=−∣p∣∂py.(4)J_{\mathrm{cyl}}\partial_x=-|p|\partial_{p_x},\qquad J_{\mathrm{cyl}}\partial_y=-|p|\partial_{p_y}. \tag*{(4)}

On T∗T2T^*T^2, fix the almost complex structure of [7 Section 4], denoted here by JinJ_{\mathrm{in}}:

Jin∂x=−f(∣p∣)∂px,Jin∂y=−f(∣p∣)∂py,(5)J_{\mathrm{in}}\partial_x=-f(|p|)\partial_{p_x},\qquad J_{\mathrm{in}}\partial_y=-f(|p|)\partial_{p_y}, \tag*{(5)}

where f>0f>0 is smooth and nondecreasing, equals 11 near zero, and satisfies f(r)=rf(r)=r for r≫0r\gg0. The values on the fiber directions are determined by J2=−idJ^2=-\mathrm{id}. Thus JinJ_{\mathrm{in}} is standard near the zero section and agrees with JcylJ_{\mathrm{cyl}} near fiber infinity. Formula (5) holds throughout the interpolation region as well, so the coordinate planes spanned by ∂x,∂px\partial_x,\partial_{p_x} and by ∂y,∂py\partial_y,\partial_{p_y} remain JinJ_{\mathrm{in}}-invariant. The cylinder classification in [7 Lemma 4.2] applies to this model.

On Y∖ZY\setminus Z, choose a ρ\rho-invariant ω\omega-compatible almost complex structure JoutJ_{\mathrm{out}} which is standard near DD and equals JcylJ_{\mathrm{cyl}} on the end approaching ZZ. Such a choice exists by the standard invariant extension construction: the prescribed local structures are σ\sigma-invariant, and ρ=σ\rho=\sigma near D∪ZD\cup Z.

With JoutJ_{\mathrm{out}} and JinJ_{\mathrm{in}} fixed, the neck-stretching construction gives a family JℓpreJ_\ell^{\mathrm{pre}} of ω\omega-compatible almost complex structures on YY, with cylindrical neck length tending to infinity as ℓ→∞\ell\to\infty. The family can be chosen ρ\rho-invariant because the stretching takes place where ρ=σ\rho=\sigma and commutes with this action. Under the Weinstein coordinates, Jℓpre=JinJ_\ell^{\mathrm{pre}}=J_{\mathrm{in}} on a fixed neighborhood of ZZ for every ℓ\ell. The almost complex structures also equal JstdJ_{\mathrm{std}} on a fixed neighborhood of DD, so the four caps remain holomorphic.

As ℓ→∞\ell\to\infty, SFT compactness [7 Section 2.5 and Theorem 2.2] and [5] gives subsequential limits of JℓpreJ_\ell^{\mathrm{pre}}-holomorphic ruling spheres in either class AA or BB as connected genus-zero holomorphic buildings in the same class. Their components lie in the two completed sides specified above, with zero or finitely many intermediate levels in (R×S∗T2,Jcyl)(\mathbb{R} \times S^{*}T^{2}, J_{\mathrm{cyl}}). We call the components in (Y∖Z,Jout)(Y \setminus Z, J_{\mathrm{out}}) exterior components, and those in (T∗T2,Jin)(T^{*}T^{2}, J_{\mathrm{in}}) bottom components.

A connected genus-zero holomorphic building in class AA or BB in these spaces is called a split sphere. Matched punctures have the same Reeb orbit and multiplicity; ordinary nodes join components in the same target level. An exterior plane is a finite-energy JoutJ_{\mathrm{out}}-holomorphic map from C\mathbb{C} with one negative puncture at infinity, considered up to holomorphic reparametrization. A cylinder has domain C∗\mathbb{C}^{*}; in an intermediate level, a trivial cylinder has image R×γ\mathbb{R} \times\gamma for a Reeb orbit γ\gamma. Exterior punctures approach ∣p∣=0\lvert p\rvert= 0, while bottom punctures approach ∣p∣=∞\lvert p\rvert= \infty in T∗T2T^{*}T^{2}.

Lemma 3.2. A split sphere in AA or BB with a nonconstant bottom component consists of two exterior planes, one primitive inner cylinder, and trivial intermediate cylinders. The relative classes of the two planes have area RR. In class AA, the planes meet D1−D_{1}^{-} and D1+D_{1}^{+} once, respectively, and their asymptotes are the two orientations of the same primitive xx-geodesic. The bottom cylinder has image {y=c,py=b}\{y = c, p_y = b\} for some c∈R/Zc \in\mathbb{R}/\mathbb{Z} and b∈Rb \in\mathbb{R}. For class BB, exchange xx and yy and use the caps D2±D_{2}^{\pm}. For a limit of embedded ruling spheres, the exterior planes are embedded. See Figure 2.

Class-A split sphere showing two exterior caps, two cylinders, and the inner cylinder labeled $\{y=c,p_y=b\}$

Figure 2. The class-AA split sphere in Lemma 3.2. Each exterior plane meets the indicated cap once. The two columns correspond to opposite Reeb directions. The inner cylinder has b=0b=0 in the point-constrained moduli space. The class-BB picture exchanges the coordinate directions.

Proof. We use the relative area formula to obtain the splitting without a genericity assumption on JoutJ_{\mathrm{out}}. Fix 0<ϵ<R/30 < \epsilon< R/3 and a split sphere. Truncate its exterior ends sufficiently far down the cylindrical ends that the annuli completing their boundary curves to ZZ have total area less than ϵ\epsilon. Each nonconstant exterior component represents a relative class (by adding a trivial cylinder with boundary on ZZ) of positive area, hence of area at least RR by the area formula in Subsection 2.2. Its truncated area is therefore at least R−ϵR - \epsilon. The completed relative classes sum to AA or BB, so the total truncated exterior area is at most 2R2R. Since 3(R−ϵ)>2R3(R - \epsilon) > 2R, there are at most two nonconstant exterior components. A closed exterior component would already have area at least 2R2R, leaving no area for a punctured component connecting it to the bottom level.

Due to the genus constraint, the component graph is a tree. Exactness and the noncontractibility of the flat Reeb orbits imply that every nonconstant component in the bottom or intermediate levels has at least two punctures. Every leaf is therefore exterior. There must be at least two leaves, so the tree is a path with exactly two exterior planes and all other components cylinders; in particular, there are no ordinary nodes or constant components. An intermediate cylinder with two negative ends would have dα0d\alpha_0-energy equal to minus the sum of their Reeb actions, contradicting its nonnegativity [7 Section 2.3]. Thus the path contains exactly one bottom cylinder, and each intermediate cylinder has one positive and one negative end.

In class AA, the cap intersections identify the relative classes of the two planes as H1−H_1^{-} and H1+H_1^{+}. Their areas are RR, and their boundary classes in (3) determine the primitive asymptotic directions ±e1\pm e_1. By [7 Lemma 4.2], the bottom cylinder consequently has image {y=c,py=b}\{y = c, p_y = b\}, and [7 Lemma 4.4] identifies the intermediate cylinders as trivial. The class-BB case exchanges the coordinates. Each plane is simply covered because it meets one cap once. For limits of embedded ruling spheres, embeddedness then follows from the positivity-of-intersection argument in [7 proof of Lemma 5.12].

Remark 3.3. The above argument may replace part of the genericity argument in [7 Corollary 5.4], but does not generalize to the non-monotone case, to the best of the authors’ knowledge. This allows us to keep the prescribed ρ\rho-invariant almost complex structure JoutJ_{\mathrm{out}} throughout the splitting argument without going into more complicated transversality issues, albeit achievable.

We now use point constraints to obtain an exterior plane at every coordinate Reeb orbit. Fix z0=(x0,y0,0,0)∈Zz_0=(x_0,y_0,0,0)\in Z and, for each ℓ\ell, take the JℓpreJ^{\mathrm{pre}}_\ell-holomorphic ruling sphere in class AA through z0z_0. SFT compactness with this point constraint gives a limiting building with a nonconstant bottom component through z0z_0, as in [7 Proposition 5.7]. By Lemma 3.2, its bottom cylinder is {y=c,py=b}\{y=c,p_y=b\}. Since it contains z0z_0, we have c=y0c=y_0 and b=0b=0. For class BB, the same argument gives {x=x0,px=0}\{x=x_0,p_x=0\}. The two ends of each cylinder correspond to the two orientations of its coordinate geodesic. As y0y_0 or x0x_0 varies, every signed coordinate Reeb orbit therefore occurs as the asymptote of an embedded exterior plane, paired with an opposite plane in the same limiting sphere.

The four families of planes

For c∈R/Zc\in\mathbb{R}/\mathbb{Z}, denote the unparametrized coordinate Reeb orbits in S∗T2S^*T^2 by

γ1±(c)=(x,c,±1,0)∈R/Z,γ2±(c)=(c,y,0,±1)∈R/Z.(6)\gamma_1^\pm(c)=(x,c,\pm1,0)\in\mathbb{R}/\mathbb{Z},\qquad\gamma_2^\pm(c)=(c,y,0,\pm1)\in\mathbb{R}/\mathbb{Z}. \tag*{(6)}

Their orientations are given by Rα0R_{\alpha_0}.

Lemma 3.4. For every i∈{1,2}i\in\{1,2\}, each choice of sign, and every c∈R/Zc\in\mathbb{R}/\mathbb{Z}, there is a unique exterior plane Pi±(c)P_i^\pm(c) asymptotic to γi±(c)\gamma_i^\pm(c) that meets Di±D_i^\pm once and is disjoint from the other caps. For each ii and each choice of sign, these planes vary smoothly with cc, and their union is an embedded three-dimensional submanifold of Y∖ZY\setminus Z. Any two distinct planes among these four families are disjoint. Moreover,

ρP1±(c)=P1±(−c),ρP2±(c)=P2±(c+12).(7)\rho P_1^\pm(c)=P_1^\pm(-c),\qquad\rho P_2^\pm(c)=P_2^\pm\left(c+\frac{1}{2}\right). \tag*{(7)}

Proof. With D∞=D1+∪D2+D_{\infty}=D_{1}^{+}\cup D_{2}^{+}, our negative and positive cap labels correspond to the small and big planes of [7], respectively. Lemma 3.2 and the preceding point-constrained construction supply the splitting configuration and embedded planes at every coordinate orbit. These planes have Maslov index 2 and Fredholm index 1, so the automatic transversality argument in [7 proof of Proposition 5.11(i)] applies to our fixed JoutJ_{\mathrm{out}}. The arguments of [7 Proposition 5.11 and Lemmas 5.12–5.13] therefore give uniqueness, smooth parametrization by cc, and an embedded three-dimensional union for each family.

Any two distinct planes in the same coordinate direction are disjoint by [7 Lemmas 5.5 and 5.8]. For different coordinate directions, complete the planes to split spheres in classes AA and BB. Their bottom cylinders {y=c,py=0}\{y=c,p_y=0\} and {x=d,px=0}\{x=d,p_x=0\} intersect once. The completion comparison in [7 proof of Lemma 5.8] then accounts for A⋅B=1A\cdot B=1, so positivity excludes any exterior intersection. See Figure 3.

Standard inner cylinders used in Lemma 3.4

Figure 3. The standard inner cylinders used in Lemma 3.4, shown at zero momentum. Cutting off the punctured ends and reconnecting matching boundary circles by annuli gives closed cycles with the displayed homological pairings. In each case the bottom intersections exhaust the pairing, so positivity forces all exterior intersections to vanish.

Finally, ρ\rho preserves JoutJ_{\mathrm{out}} and agrees with σ\sigma near D∪ZD\cup Z. It permutes the cap and orbit data as prescribed in (7), so uniqueness gives that formula.

Each of the four families Pi±P_i^{\pm} therefore fills a hypersurface of Y∖ZY\setminus Z. In the next subsection, we recover the two sphere rulings by joining the opposite families.

Smoothing and the two sphere rulings.

Proof of Proposition 3.1. Apply the end-straightening construction of [7 Proposition 5.16] to the four families in Lemma 3.4. This modifies the planes near ZZ, preserving their symplecticity and the smooth embedded families, so that their ends become standard half-cylinders.

The construction can be performed simultaneously and ρ\rho-equivariantly: the four families of ends lie in disjoint cones around the momentum directions ±e1,±e2\pm e_1,\pm e_2, and we use the same radial cutoff in all four cones. Indeed, ρ=σ\rho=\sigma near ZZ acts by translations and sign changes, permutes these cones, and preserves the radial coordinate, so the cutoff construction commutes with ρ\rho. See Figure 4.

Figure 4. Pairing the ends after simultaneous smoothing in Proposition 3.1. Left: near ZZ, the ends of Pi±(c)P_i^{\pm}(c) lie in four disjoint cones of nonzero momenta. Base coordinates are suppressed. Right: one scalar end-graph component at the fixed (x,c)(x,c), before and after multiplication by χ(s)\chi(s). The modified graphs vanish for s≪0s\ll0 and agree with the original ones near the outer boundary. The same scalar cutoff in all four cones commutes with ρ\rho. The standard opposite ends then close across their common geodesic on ZZ.

For each cc, the straightened planes P1±(c)P_1^{\pm}(c) join smoothly across {y=c,p=0}\{y=c,p=0\} to form a symplectic sphere in class AA, with local model {y=c,py=0}\{y=c,p_y=0\} near ZZ. Similarly, the planes P2±(d)P_2^{\pm}(d) join to a sphere in class BB, with local model {x=d,px=0}\{x=d,p_x=0\}. These give two smooth circle families of spheres. Within each family the spheres are disjoint, while spheres from different families meet only at their coordinate-circle crossing on ZZ. Near ZZ, their tangent planes are the two standard, symplectically orthogonal coordinate planes. Both families are preserved by ρ\rho, and the construction leaves the planes unchanged near DD.

For a single sphere family, the existence of a global tame almost complex structure making all its leaves holomorphic is noted in [7 Proposition 5.16(iii) and Remark 5.17]. Here we need one compatible, ρ\rho-invariant almost complex structure for both families. Each family foliates an embedded three-dimensional submanifold, along which the sphere tangent planes form a smooth symplectic subbundle VV of TYTY. Choose a Riemannian metric gg satisfying V⊥gVωV \mathrel{\perp_g} V^\omega along every leaf. Near ZZ, the standard product metric satisfies this condition for both families. Near DD, we let the metric be associated to Jout=JstdJ_{\mathrm{out}}=J_{\mathrm{std}}. Away from a neighborhood of ZZ, the two family unions are disjoint, so these metric prescriptions extend in separate tubular neighborhoods and then to YY. A partition of unity and averaging over ρ\rho preserve the orthogonality conditions and the prescribed local metrics. The compatible almost complex structure J1J_1 associated to this metric and ω\omega therefore preserves every sphere tangent plane, is ρ\rho-invariant, and is standard near D∪ZD\cup Z. Gromov’s theorem identifies the constructed spheres as leaves of its two sphere rulings, proving the proposition.

Relative reconstruction

We prove Proposition 2.1 using Gromov’s ruling argument, as in [7 Section 5.5], with the reconstruction described in [8 pp. 11–13]. We arrange that the maps conjugate ρ\rho to σ\sigma, preserve each labeled cap, and preserve ZZ at the endpoint.

Leaf intersection maps

Set J0=ψ∗JstdJ_0=\psi^*J_{\mathrm{std}}. Choose a smooth path of ρ\rho-invariant ω\omega-compatible almost complex structures JtJ_t from J0J_0 to J1J_1, equal to JstdJ_{\mathrm{std}} on one fixed neighborhood of DD. Such a path is obtained by interpolating the associated invariant metrics, as in [6 Proposition 12.8].

Let At(p)A_t(p) and Bt(p)B_t(p) denote the two ruling leaves through p∈Yp\in Y. We use the pole labels and factor involutions introduced in Subsection 2.2. Parametrize D1D_1 by v(b)=(a−,b)v(b)=(a_-,b). Choose a σ2\sigma_2-fixed point b0b_0 on the second equator and set P=(a−,b0)P=(a_-,b_0). Since ρ=σ\rho=\sigma near DD, the point PP is fixed by ρ\rho; it lies on the equator of D1−D_1^{-}, away from the corners (a−,b±)(a_{-},b_{\pm}). Uniqueness of the ruling leaf through PP makes At(P)A_t(P) ρ\rho-invariant. Its unique intersection with D1+D_1^{+} is therefore fixed by ρ\rho. This intersection varies continuously in the two-point fixed set of ρ∣D1+\rho|_{D_1^{+}} and starts at (a+,b0)(a_{+},b_0), so it remains there. Thus At(P)A_t(P) meets D1±D_1^{\pm} at (a±,b0)(a_{\pm},b_0) for every tt.

Choose smooth orientation-preserving parametrizations ut:S2→At(P)u_t:S^2\to A_t(P) satisfying

u0(a)=ψ(a,b0),utσ1=ρut.u_0(a)=\psi(a,b_0), \qquad u_t\sigma_1=\rho u_t.

with first coordinate aa near a±a_{\pm}. To obtain them, use the equivariant holomorphic graph parametrizations near the caps, extend the normal velocity of At(P)A_t(P) with these prescribed graph velocities, and average the extension over ρ\rho. Its flow transports u0u_0 to the required utu_t.

Define

Φt:S2×S2→Y,Φt(a,b)=Bt(ut(a))∩At(v(b)).(8)\Phi_t:S^2\times S^2\to Y,\qquad\Phi_t(a,b)=B_t(u_t(a))\cap A_t(v(b)). \tag*{(8)}

The leaf-intersection construction of [8 p. 11] gives a smooth family of diffeomorphisms. The chosen parametrizations, together with vσ2=ρvv\sigma_2=\rho v, give

Φ0=ψ,Φtσ=ρΦt.(9)\Phi_0=\psi,\qquad\Phi_t\sigma=\rho\Phi_t. \tag*{(9)}

The pole labels show that each Φt\Phi_t preserves every labeled cap. Moreover, product spheres sufficiently close to the caps are JtJ_t-holomorphic and hence are ruling leaves, as in [7 proof of Theorem D, Section 5.4]. The chosen parametrizations label these leaves by their product coordinates, so Φt\Phi_t is the identity on one common neighborhood of the four corners. Figure 5 illustrates (8).

Figure 5. Reconstruction from the two sphere rulings. Lines denote sphere leaves. The point Φt(a,b)\Phi_t(a,b) is the intersection of the BB-leaf through ut(a)u_t(a) and the AA-leaf through v(b)v(b).

Set ft=Φt−1f_t=\Phi_t^{-1}. By Proposition 3.1,

f1(Z)=C1×C2,f_1(Z)=C_1\times C_2,

where CiC_i is a σi\sigma_i-invariant embedded circle in the corresponding annular factor. Both circles are essential because Z↪Y∖DZ\hookrightarrow Y\setminus D induces an isomorphism on fundamental groups and f1f_1 preserves DD. Apply Lemma A.1 in each factor and append the resulting product isotopy; see Figure 6. After stationary reparametrizations at the join, the resulting smooth path, still denoted ftf_t, satisfies

Figure 6. Before the appended straightening, the inverse leaf map sends ZZ to a product of invariant essential circles, drawn schematically. The appended equivariant annular isotopies take the circles to the equators and hence their product to ZZ.

f0=ψ−1,ftρ=σft,ft(Di±)=Di±,f1(Z)=Z.(10)f_0=\psi^{-1},\qquad f_t\rho=\sigma f_t,\qquad f_t(D_i^{\pm})=D_i^{\pm},\qquad f_1(Z)=Z. \tag*{(10)}

It remains the identity near the corners. Throughout the original interval, ftf_t takes the two JtJ_t-ruling planes to the product-factor planes and preserves their orientations. This remains true on the appended interval with the almost complex structure fixed at J1J_1.

Relative Moser deformation and cutoff

Consider

ωs,t=(1−s)ω+sft∗ω,0≤s,t≤1.(11)\omega_{s,t}=(1-s)\omega+s f_t^*\omega,\qquad0\le s,t\le1. \tag*{(11)}

These forms are symplectic by the convex-form argument of [8 pp. 12–13]: the two ruling planes are symplectically orthogonal with respect to ft∗ωf_t^*\omega, which is positive on each and hence tames the same JtJ_t as ω\omega. On the appended interval in the previous subsection, use J1J_1. The forms are ρ\rho-invariant by (10) and cohomologous because ftf_t starts at the symplectomorphism ψ−1\psi^{-1}.

We next apply a relative equivariant Moser argument to make the path ftf_t symplectic, while maintaining its initial value and zero-section endpoint condition. A subsequent Hamiltonian cutoff makes the corrected path equal to the identity on a fixed neighborhood of DD.

To keep the zero section at the endpoint, we need

[f1∗ω−ω]=0in H2(Y,Z;R).(12)[f_1^*\omega-\omega]=0\quad\text{in }H^2(Y,Z;\mathbb{R}). \tag*{(12)}

The map f1f_1 preserves ZZ and every labeled cap. It also preserves each cap’s normal orientation, since these signs are constant along the cap-preserving path starting at ψ−1\psi^{-1}. It therefore preserves the cap-intersection coordinates of H2(Y,Z;Z)H_2(Y,Z;\mathbb{Z}). The hemisphere basis in Subsection 2.2 shows that f1f_1 acts trivially on this group, proving (12).

The homotopy formula for ftf_t gives smooth primitives ηt\eta_t of ft∗ω−ωf_t^*\omega-\omega with η0=0\eta_0=0. Average them over ρ\rho, and choose a ρ\rho-invariant primitive βrel\beta^{\mathrm{rel}} at t=1t=1 with βrel∣TZ=0\beta^{\mathrm{rel}}|_{TZ}=0. Since βrel−η1\beta^{\mathrm{rel}}-\eta_1 is closed, setting

βt=ηt+t(βrel−η1)\beta_t=\eta_t+t(\beta^{\mathrm{rel}}-\eta_1)

gives invariant primitives satisfying

β0=0,β1∣TZ=0.(13)\beta_0=0,\qquad\beta_1|_{TZ}=0. \tag*{(13)}

Near the corners, ft=id⁡f_t=\operatorname{id}, so βt\beta_t is closed. Write βt=dbt\beta_t=db_t on disjoint corner balls, with btb_t normalized to vanish at the corners. Subtracting d(χbt)d(\chi b_t) with invariant cutoffs makes βt\beta_t vanish on smaller common corner neighborhoods. This preserves (13), since the balls are disjoint from ZZ and the normalized potentials vanish at t=0t=0.

Each cap CC is symplectic for ωs,t\omega_{s,t}. Choose functions as,tCa^C_{s,t} vanishing on CC whose normal derivatives satisfy

das,tC∣(TC)ωs,t=−βt∣(TC)ωs,t.(14)da^C_{s,t}|_{(TC)^{\omega_{s,t}}}=-\beta_t|_{(TC)^{\omega_{s,t}}}. \tag*{(14)}

A tubular extension and cutoff realize these derivatives. The extensions can be chosen smoothly and linearly in βt\beta_t, with disjoint supports away from the corner neighborhoods and ZZ. Let as,ta_{s,t} be the sum of these functions averaged over ρ\rho, and define β^s,t=βt+das,t\widehat{\beta}_{s,t}=\beta_t+da_{s,t}. Averaging preserves the prescribed derivatives. The corrected primitives vanish at t=0t=0, annihilate TZTZ at t=1t=1, and annihilate (TC)ωs,t(TC)^{\omega_{s,t}} along every cap.

The Moser vector field defined by ιVs,tωs,t=−βs,t\iota_{V_{s,t}}\omega_{s,t}=-\beta_{s,t} is therefore ρ\rho-invariant, tangent to every cap, zero at t=0t=0, and tangent to ZZ at t=1t=1. Its flow from s=0s=0 to s=1s=1, denoted by MtM_t, satisfies Mt∗ft∗ω=ωM_t^*f_t^*\omega=\omega. Thus

St=ftMt(15)S_t=f_tM_t \tag*{(15)}

is a smooth symplectic path with

Stρ=σSt,S0=ψ−1,St(Di±)=Di±,S1(Z)=Z.S_t\rho=\sigma S_t,\qquad S_0=\psi^{-1},\qquad S_t(D_i^\pm)=D_i^\pm,\qquad S_1(Z)=Z.

We now make StS_t equal to the identity near DD by a Hamiltonian cutoff, following the construction at the end of [7 Section 7.2]. Since H1(Y;R)=0H^1(Y;\mathbb{R})=0, the vector field Xt=S˙tSt−1X_t=\dot{S}_tS_t^{-1} has a smooth Hamiltonian HtH_t with ιXtω=−dHt\iota_{X_t}\omega=-dH_t. The conjugacy relation makes XtX_t σ\sigma-invariant, so averaging allows us to take HtH_t invariant as well.

Choose an invariant cutoff χ\chi supported near DD, away from ZZ, and equal to one on a neighborhood OO of DD. Let KtK_t be the Hamiltonian flow of χHt\chi H_t from the identity. It commutes with σ\sigma and fixes a neighborhood of ZZ. Because St(D)=DS_t(D)=D and the parameter interval is compact, there is an invariant neighborhood NN of DD on which ψ=id\psi=\mathrm{id} and such that St(N)⊂OS_t(N)\subset O for all tt. On these trajectories the two Hamiltonian vector fields agree, and their initial maps agree on NN. Uniqueness of the flow gives

Kt∣N=St∣Nfor all t.(16)K_t|_N=S_t|_N\qquad\text{for all }t. \tag*{(16)}

Define

ht=Kt−1Stψ.(17)h_t=K_t^{-1}S_t\psi. \tag*{(17)}

This is a symplectic isotopy from the identity, commuting with σ\sigma and equal to the identity on NN. Its supports lie in the common compact set Y∖NY\setminus N. Restricting to Y∖DY\setminus D and extending by the identity gives the required isotopy on T∗T2T^*T^2. Finally,

h1(ψ−1(Z))=K1−1S1(Z)=Z,h_1\left(\psi^{-1}(Z)\right)=K_1^{-1}S_1(Z)=Z,

because K1K_1 fixes ZZ. This proves Proposition 2.1.

Conclusion of the proof

Proof of Theorem 1.1. Let EE and ψ\psi be as in Lemma 2.2. Proposition 2.1 gives a smooth symplectic isotopy hth_t from the identity with htσ=σhth_t\sigma=\sigma h_t and h1(E)=Zh_1(E)=Z. For the covering projection π:T∗T2→T∗K\pi:T^*T^2\to T^*K, equivariance gives a smooth isotopy hˉt\bar{h}_t defined by

hˉtπ=πht,\bar{h}_t\pi=\pi h_t,

which is symplectic since π\pi is locally symplectic. Its initial map is the identity and hˉ1(L)\bar{h}_1(L) is the zero section. If C⊂T∗T2C\subset T^*T^2 is a common compact support of hth_t, then π(C)\pi(C) is a common compact support for hˉt\bar{h}_t.

Since T∗KT^*K is an oriented four-manifold and retracts onto KK, Poincaré duality with compact supports gives

Hc1(T∗K;R)≅H3(T∗K;R)≅H3(K;R)=0.H^1_c(T^*K;\mathbb{R})\cong H_3(T^*K;\mathbb{R})\cong H_3(K;\mathbb{R})=0.

The descended path is therefore generated by a smooth Hamiltonian with one common compact support, by the Hamiltonian isotopy criterion in [MS17, Proposition 9.3.1 and p. 389]. This proves the theorem.

Equivariant annular straightening

In this appendix we prove the equivariant straightening result used to complete the smooth reconstruction in Subsection 4.1. The inverse leaf-intersection map at the endpoint sends ZZ to a product C1×C2C_1 \times C_2 of essential circles invariant under σ1\sigma_1 and σ2\sigma_2, respectively. The main result is Lemma A.1, which moves each circle to the equator through an isotopy commuting with the corresponding involution and fixed near the annular ends. Appending the product of these isotopies gives f1(Z)=Zf_1(Z)=Z while preserving the conjugacy relation and every labeled cap, providing the endpoint condition needed for the relative Moser deformation in the proof of Proposition 2.1.

Lemma A.1. On A=S1×RA=S^1\times\mathbb{R}, let S(q,p)=(q+12,p)S(q,p)=(q+\frac{1}{2},p) and r(q,p)=(−q,−p)r(q,p)=(-q,-p). If CC is a smooth invariant embedded essential circle for either involution τ=S\tau=S or τ=r\tau=r, then there is a smooth isotopy ftf_t with one common compact support such that

f0=id,ftτ=τft,f1(C)=S1×{0}.f_0=\mathrm{id},\qquad f_t\tau=\tau f_t,\qquad f_1(C)=S^1\times\{0\}.

Proof. For SS, the quotient is an annulus and the image of CC is an embedded essential circle. Curve isotopy and smooth isotopy extension [9 Propositions 1.10–1.11] give an ambient isotopy supported in a compact subannulus carrying it to the core. Lifting the vector field gives the required isotopy upstairs.

For rr, choose a large symmetric closed annulus containing CC. Its quotient is a disk with two simple branch points, locally described by w=z2w=z^2. See Figure 7. Since rr acts by −1-1 on the first homology of the annulus, its restriction to CC reverses orientation, hence CC contains the two fixed points and projects to an embedded arc joining the branch points. We first straighten CC near those points by equivariant local isotopies. After rotating its tangent line, write it as y=h(x)y=h(x) with hh odd. The vector field −χ(x,y)h(x)∂y-\chi(x,y)h(x)\partial_y, with an even cutoff equal to one on the local trace, straightens this graph. The two constructions have disjoint supports.

Figure 7. The branched case of Lemma A.1. After the initial local straightening, the invariant circle descends to an arc joining the two branch points. The disk isotopy after the half-twist correction takes this arc to the image of the core.

The coordinate-change principle for surfaces [9 Section 1.3] gives a disk diffeomorphism, fixed near the boundary, taking the resulting arc to the image of the core. Since the arcs have straight endpoint germs, we specify complex linear germs in the centered branch charts. The mapping class group of the disk with two unordered marked points is B2≅ZB_2\cong\mathbb{Z} [9 Sections 9.1–9.2]. A generating half twist preserves the target arc and can also have complex linear branch germs. Postcomposing with a suitable power gives a straightening map GG isotopic to the identity relative to the boundary and the marked points. Choose a smooth isotopy GtG_t from the identity to GG, fixed on one boundary collar; smooth isotopies may be used by [9 discussion after Theorem 1.13]. Figure 7 illustrates the branched quotient and the straightening of the projected arc.

Before lifting, we normalize the branch germs relative to the endpoints of the isotopy. In a centered branch chart, we write

Gt(w)=Atw+Rt(w),At∈GL+(2,R),Rt(w)=O(∣w∣2),G_t(w)=A_t w+R_t(w),\qquad A_t\in\mathrm{GL}^{+}(2,\mathbb{R}),\qquad R_t(w)=O(|w|^2),

uniformly in tt. For sufficiently small fixed ϵ\epsilon, replace this expression by Gt(w)−χ(w/ϵ)Rt(w)G_t(w)-\chi(w/\epsilon)R_t(w), which is exactly AtwA_t w on a smaller disk. The change is O(ϵ)O(\epsilon) in C1C^1, so the maps remain diffeomorphisms. The endpoints are unchanged since their germs are already linear.

We write At=UtPtA_t=U_tP_t in polar form and define

Bt=Pt−1det⁡Pt I.B_t=P_t^{-1}\sqrt{\det P_t}\,I.

Precompose with the time-1 flow of the compactly supported vector field χ0(w)(log⁡Bt)w\chi_0(w)(\log B_t)w. On a sufficiently small common disk this flow has germ BtB_t, making the new germ Utdet⁡Pt IU_t\sqrt{\det P_t}\,I which is complex-linear. The correction is the identity at both endpoints since B0=B1=IB_0=B_1=I. We then perform these constructions in the two disjoint branch charts.

The resulting germs are w↦atww\mapsto a_t w with at≠0a_t\ne0 and a0=1a_0=1. At each branch point, choose the smooth square root along the parameter interval that starts at 1. In the branch chart w=z2w=z^2, the local lift G~t(z)=at z\widetilde G_t(z)=\sqrt{a_t}\,z is smooth in (t,z)(t,z) and satisfies

(G~t(z))2=atz2=Gt(z2).\bigl(\widetilde G_t(z)\bigr)^2=a_tz^2=G_t(z^2).

These local lifts agree with the ordinary isotopy lift away from the branch points by uniqueness and their common initial value. The lift starting at the identity commutes with rr and is fixed near both annular ends. We concatenate it with the initial local straightening to get the required isotopy. All constructions take place in one compact annulus, and stationary reparametrizations make the concatenation smooth.

References

School of Mathematics, Nanjing University, Nanjing, Jiangsu Province 210093, China Email address: [email protected]

School of Mathematical Sciences, Zhejiang University, Hangzhou, Zhejiang Province 310027, China Email address: [email protected]

School of Mathematical Sciences, Eastern Institute of Technology, Ningbo, Zhejiang, 315200, China Email address: [email protected]

References

  1. [1]Mohammed Abouzaid. Nearby Lagrangians with vanishing Maslov class are homotopy equivalent. Invent. Math., 189(2):251–313, 2012.arxiv.org/abs/1005.0358
  2. [2]Marcelo S. Atallah, Jean-Philippe Chassé, Rémi Leclercq, and Egor Shelukhin. Weinstein exactness of nearby Lagrangians and related questions, 2025. Preprint, version 4.arxiv.org/abs/2410.04158
  3. [4]Mohammed Abouzaid and Thomas Kragh. Simple homotopy equivalence of nearby Lagrangians. Acta Math., 220(2):207–237, 2018.arxiv.org/abs/1603.05431
  4. [5]Frédéric Bourgeois, Yakov Eliashberg, Helmut Hofer, Krzysztof Wysocki, and Eduard Zehnder. Compactness results in Symplectic Field Theory. Geom. Topol., 7:799–888, 2003.DOI
  5. [6]Ana Cannas da Silva. Lectures on symplectic geometry, volume 1764 of Lecture Notes in Mathematics. Springer-Verlag, Berlin, 2001.DOI
  6. [7]Georgios Dimitroglou Rizell, Elizabeth Goodman, and Alexander Ivrii. Lagrangian isotopy of tori in S² × S² and CP². Geom. Funct. Anal., 26(5):1297–1358, 2016.
  7. [8]Jonathan Evans. Lecture XIV: Pseudoholomorphic curves II. Lecture notes, https://jde27.uk/symp/lecture14.pdf, 2010.
  8. [9]Benson Farb and Dan Margalit. A primer on mapping class groups, volume 49 of Princeton Mathematical Series. Princeton University Press, Princeton, NJ, 2012.DOI
  9. [10]Kenji Fukaya, Paul Seidel, and Ivan Smith. Exact Lagrangian submanifolds in simply-connected cotangent bundles. Invent. Math., 172(1):1–27, 2008.DOI
  10. [11]Mikhael Gromov. Pseudo holomorphic curves in symplectic manifolds. Invent. Math., 82(2):307–347, 1985.DOI
  11. [12]Richard Hind. Lagrangian spheres in S² × S². Geom. Funct. Anal., 14(2):303–318, 2004.
  12. [13]Richard Hind. Lagrangian unknottedness in Stein surfaces. Asian J. Math., 16(1):1–36, 2012.DOI

Paper details

vlnc:W_t7rt2yqs
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

a generative large language model

View statement
Looking for related papers…
Contents