Home Perturbation theory of asymptotic operators of contact instantons and pseudoholomorphic curves on symplectization
Article Open Access

Perturbation theory of asymptotic operators of contact instantons and pseudoholomorphic curves on symplectization

  • Taesu Kim and Yong-Geun Oh EMAIL logo
Published/Copyright: January 23, 2025
Become an author with De Gruyter Brill

Abstract

In this paper, we first derive the covariant tensorial formula for the asymptotic operators of contact instantons w : Σ̇Q and of pseudoholomorphic curves u : (Σ̇, j) → (Q × ℝ, ) on the symplectization of a contact manifold (Q, λ). The formula is independent of the choice of connection on the nose and exhibits explicit dependence on the compatible pair (λ, J) of a given contact triad (Q, λ, J). Then based on this, we prove that all eigenvalues of the asymptotic operator are simple for a generic choice of compatible CR almost complex structure J for a given contact form λ. This perturbation theory has been missing in the study of pseudoholomorphic curves on symplectization.

1 Introduction and overview

Let (Q, ξ) be a contact manifold and assume that ξ is coorientable. Then we can choose a contact form λ with ker λ = ξ. With λ given, we have the Reeb vector field Rλ uniquely determined by the equations Rλ = 0, Rλλ = 1. Then we have the decomposition TQ = ξ ⊕ ℝ {Rλ}. We denote by π : TQξ the associated projection and by Π = Πλ : TQTQ the associated idempotent whose image is ξ.

A contact triad is a triple (Q, λ, J) where λ is a contact form of ξ, i.e. ker λ = ξ, and J is an endomorphism J : TQTQ that satisfies the following.

Definition 1.1

(CR almost complex structure). Let (Q, ξ = ker λ) be as above. A CR almost complex structure J is an endomorphism J : TQTQ satisfying J2 = − Π, or more explicitly

(J|ξ)2=id|ξ,J(Rλ)=0.

  1. We say J is adapted to λ if (Y, J Y) ≥ 0 for all Yξ with equality only when Y = 0. In this case, we call the pair (λ, J) an adapted pair of the contact manifold (Q, ξ).

  2. We say J is compatible to λ if the bilinear form (X, JY) is symmetric on ξ in addition to (1). In that case, we call the pair (λ, J) a compatible pair of (Q, ξ).

The associated contact triad metric for a λ-compatible pair (λ, J) is given by

gλ:=dλ(,J)+λλ. (1.1)

The symplectization of (Q, λ) is the symplectic manifold (Q × ℝ, d(esλ)) with ℝ-coordinate s also called the (cylindrical) radial coordinate. We equip the symplectization with the s-translation invariant almost complex structure

J~=JJ0

where J0 is the almost complex structure on the plane ℝ { s , Rλ} satisfying J0( s ) = Rλ.

On the other hand, in the joint work [35] by Savelyev and the second-named author, they lifted the theory of contact instantons to the theory of pseudoholomorphic curves on the 𝔩𝔠𝔰-fication (Q × Sρ1 , ωλ), Banyaga’s locally conformal symplectification (which they call the 𝔩𝔠𝔰-fication) of contact manifold (Q, λ), see [4], on

(Q×Sρ1,dλ+dθλ) (1.2)

with the canonical angular form satisfying Sρ1 = 1. According to the terminology adopted in [35], the authors call them the 𝔩𝔠𝔰-fication ‘of nonzero temperature’ on which the theory of pseudoholomorphic curves is developed. Here ‘𝔩𝔠𝔰′ stands for the standard abbreviation of ‘locally conformal symplectic’. The authors of ibid. call the relevant pseudoholomorphic curves lcs instantons. This family can be augmented by including the case of the product Q × ℝ as the ‘zero temperature limit’ with 1/ρ → 0,

(Q×R,ωλ),ωλ:=dπQλ+dsπQλ (1.3)

in physical terms. (We refer interested readers to many physical literature for various physical discussions involving consideration of such a limit. Here, we just take one such example, the discussion in [9, Section 4.1].) We denote by ρ the radius of the circle S1 and by π = πQ (respectively s) the projection to Q (respectively to ℝ) of Q × ℝ. For the simplicity of notation, we will often just write λ for πQλ on Q × ℝ.

The tensorial approach also clarifies the relationship between the background geometries of the contact triad (Q, λ, J), the symplectization

(SQ,d(esλ))=(Q×R,esωλ)

and the lcs manifold (1.2). We write M : = SQ and consider the decomposition

TMTQRsξspanR~λ,sξR2. (1.4)

We denote by λ the unique vector field on [0, ∞) × Q which is invariant under the translation, tangent to the level sets of s and projected to Rλ. When there is no danger of confusion, we will mostly denote it by Rλ. For a given contact triad (Q, λ, J), we have a canonical almost complex structure

J0:Rs,RλRs,Rλ

defined by J0 s = Rλ thanks to the splitting (1.4).

Any smooth map u : Σ̇Q × ℝ has the form u = (w, f) with

f=su,w=πu (1.5)

in the presence of the contact form λ on (Q, ξ). We have the decomposition of the derivative

du=dwdfs

viewed as a TM-valued one-form with respect to the splitting

Hom(TzΣ˙,Tu(z)M)=Hom(TzΣ˙,HTu(z)M)Hom(TzΣ˙,VTu(z)M).

(For the notational simplicity, we often omit ‘⊗’ except in situations that could cause confusion to the readers without it.)

The main purpose of the present paper is to continue the second-named author and his collaborators’ covariant tensorial study of contact instantons and of the pseudoholomorphic curves on symplectization given in [37; 38; 29; 39] and to carry out precise asymptotic analyses near the punctures of finite energy contact instantons and of finite energy pseudoholomorphic curves by developing a generic perturbation theory of asymptotic operators over the change of compatible CR-almost complex structures.

1.1 Pseudoholomorphic curves in symplectization and contact instantons

By definition, we have Q du = dw. It was observed by Hofer [13] that u is -holomorphic if and only if (w, f) satisfies

¯πw=0wλj=df. (1.6)

(We refer to Appendix B for the here unexplained notation π.) A contact instanton is a map w : Σ̇M that satisfies the system of nonlinear partial differential equations

¯πw=0,d(wλj)=0 (1.7)

on a contact triad (M, λ, J). The equation itself had been introduced by Hofer [14, p. 698]. Note that for any map u = (w, f) satisfying (1.6), w satisfies this equation with the additional property that the one-form w*λj is exact, not just closed.

In a series of papers, [37; 38] joined with Wang and in [27], the second-named author systematically developed the analysis of contact instantons (for the closed string case) without taking symplectization by the global covariant tensorial calculations using the notion of contact triad connection introduced by Wang and the second-named author in [36]. A relevant Fredholm theory has been developed by the second-named author in [27], [35] (for the closed string case). More recently the second-named author also studied its open string counterpart of the boundary value problem of (1.7) under the Legendrian boundary condition whose explanation is now in order. We mention that in the early 2000’s Abbas studied the Legendrian chord problem in [1], and more recently Cant [7] developed a detailed Fredholm theory on symplectization in the relative context.

Throughout the paper, we adopt the following notation.

Notation 1.2

We denote by (Σ, j) a closed Riemann surface, by Σ̇ the associated punctured Riemann surface and by Σ the real blow-up of Σ̇ along the punctures.

For the simplicity and for the main purpose of the present paper, we will focus on the genus zero case so that Σ̇ is conformally the unit disc with boundary punctures z0, …, zk∂ D2 ordered counterclockwise, i.e.,

Σ˙D2{z0,,zk}

Then, for a (k + 1)-tuple R⃗ = (R0, R1, …, Rk) of Legendrian submanifolds, which we call an (ordered) Legendrian link, we consider the boundary value problem

¯πw=0,d(wλj)=0,w(zizi+1¯)Ri (1.8)

as an elliptic boundary value problem for a map w : Σ̇M and derive the a priori coercive elliptic estimates. Here zizi+1∂ D2 is the open arc between zi and zi+1.

Remark 1.3

Oh also identified the correct counterpart of Hamiltonian-perturbed contact instantons in [29], [31] and applied them to a systematic quantitative study of contact topology [32], and proved Shelukhin’s conjecture [30].

Let ∇͠ := ∇can be the canonical connection of this almost Hermitian manifold

(Q×R,g~λ,J~),g~λ:=gλ+dsds (1.9)

i.e. the unique Riemannian connection whose torsion T satisfies T(X, J͠ X) = 0 for all XT(Q × ℝ); see [11; 22; 26] for its definition and basic properties. We note that this almost Hermitian structure on Q × ℝ is translational invariant in the radial s-direction. The usage of the canonical connection on the 𝔩𝔠𝔰-fication (Q × ℝ, , ωλ) (or equivalently via the contact triad connection on the triad (Q, λ, J)) plays an important role in the present authors’ analysis of the asymptotic operators of finite energy contact instantons in Section 5, and hence of finite energy pseudoholomorphic curves too. Roughly speaking, our coordinate-free approach enables us to compute the asymptotic operator associated to each isospeed Reeb orbit (γ, T) of a contact instanton w denoted by

A(λ,J,)π:Γ(γξ)Γ(γξ)

simultaneously over all isospeed Reeb orbits in the covariant tensorial way in terms of the pull-back connection under the map w of the given contact triad connection or the Levi–Civita of the triad metric of the given triad (Q, λ, J). (See Definition 4.2 for the precise definition of the operator A(λ,J,)π .)

Definition 1.4

(Isospeed Reeb trajectory). We call a pair (γ, T) an isospeed Reeb trajectory if γ: [0, 1] → Q and T = ∫ γ* λ satisfy

γ˙(t)=TRλ(γ(t)). (1.10)

If γ(1) = γ(0) in addition, we call it an isospeed Reeb orbit.

1.2 Asymptotic operators and their analysis

We first mention a few differences between the way how we study the asymptotic operators and those of [16] and of other literature such as [42, Appendix C], [43; 44], [45], [7].

In [42, Appendix E], [43; 44], [45, Section 3.3], there have been attempts to give a coordinate-free definition of the asymptotic operator along the associated asymptotic Reeb orbit for a pseudoholomorphic curve u = (w, f) on symplectization. However they fall short of a seamless definition of the ‘asymptotic operator’ of the Reeb orbits because the Reeb orbit lives on Q while the pseudoholomorphic curves live on the product Q × ℝ and the asymptotic limit of pseudoholomorphic curves lives at infinity Q × {± ∞}. (Cant studied the asymptotic operator in the relative context in [7, Section 6.3] by adapting Wendl’s.)

What the literature (e.g. [45, Section 3.3], [43; 44]) is describing, however, is actually the asymptotic operator of the contact instanton w but trying to describe it in terms of the pseudoholomorphic curves which prevents them from being able to give a seamless definition: Recall the decomposition

du=dπw(wλRλ)dfs (1.11)

with respect to the splitting (1.4). (Compare their practices with our definition of the asymptotic operator of contact instantons given in Definition 4.2 and compare it therewith. See also [38, Section 11.2 & 11.5] for the precursor of our definition.)

In this regard, we will derive an explicit formula of the asymptotic operator in terms of the contact triad connection, which exhibits the way how the operator explicitly depends on J so it admits a perturbation theory under the change of J.

Proposition 1.5

(Proposition 4.3). Letbe the triad connection, and let (γ, T) be an isospeed Reeb orbit. Then the asymptotic operator A(λ,J,)π is given by

A(λ,J,)π=JtT2JLRλJ.

Combining this with the basic properties of the triad connection and γ̇(t) = T Rλ(γ(t)), we can derive an explicit expression of the asymptotic operator purely in terms of λ, J and (γ, T) independent of the choice of the connection.

Theorem 1.6

(Theorem 5.5). Let (Q, λ, J) be a triad, and let (γ, T) with T ≠ 0 be an isospeed Reeb trajectory. Then we have the formula

A(λ,J,)π=T12LRλJJLRλ12J(LRλJ) (1.12)

when acted upon Γ(γ* TM).

This study enables us to make precise statements on the spectral behavior of asymptotic operators under the perturbation of J’s. The ellipticity of A(λ,J,)π implies that it has a discrete set of eigenvalues, which we enumerate as

<μk<<μ1<0<μ1<<μk<

repeated with finite multiplicities allowed, where μk → ∞ (respectively μk → − ∞) as k → ∞. It also has a uniform spectral gap (by [18, Theorem 6.29]) in that there exists some d > 0 with |μi+1μi| ≥ d > 0 for all i.

Our explicit formula of the asymptotic operator given in Proposition 1.5, which simultaneously applies to all closed Reeb orbits, enables us to provide a generic description of the spectral behavior of asymptotic operators under the change of λ-compatible CR almost complex structures J when λ is fixed. For example, we prove the following natural generic perturbation result of eigenvalues of the asymptotic operator.

Theorem 1.7

(Generic simpleness of eigenvalues, Theorem 5.1). Let (Q, ξ) be a contact manifold. Assume that λ is nondegenerate. For a generic choice of λ-compatible CR almost complex structures J, all eigenvalues μi of the asymptotic operator are simple for all closed Reeb orbits of λ.

See Section 5 for our derivation of the formula of the asymptotic operator.

1.3 Relationship with other literature and discussion

We have no doubt that a parameterized version of the proof of Theorem 5.1 will lead to the following spectral flow description of the index. Consider the nonlinear Fredholm map

Υ(λ,J)cont(w):=(¯πw,d(wλj))

whose domain and codomain are specified as in [27] or in [39] where this map was just written as Υ instead. Since we use the latter for a different purpose later in the present article, we use the different notation Υ(λ,J)cont here for the same operator appearing in [27] or in [39].

Theorem 1.8

Let (Q, λ, J) be a contact triad. Let w : ℝ × S1Q be a contact instanton with its asymptotic limits (γ+, T+) and (γ, T) and write the associated asymptotic operators as

A±=A(λ,J,)π,±:Γ(γ±ξ)Γ(γ±ξ).

Then for J generic in the sense of Theorem 5.1, we have

IndexDΥ(λ,J)cont(w)=μspec(A,A+)

where μspec(A−∞, A+∞) is the spectral flow from A−∞ to A+∞.

The main step of the proof is to establish the contact instanton version of Lemma 3.17 [45] which says that there is a perturbation (inside the space of abstract Fredholm operators) of the given one-parameter family of asymptotic operators such that all eigenvalue families of the perturbed operator are transversal in an explicit sense. In our case, such a transversality result can be obtained inside the smaller natural family of J perturbations by the parameterized version of the proof of Theorem 5.1. We omit the details of its proof leaving them to interested readers or to [45, Appendix C] which handles the case of pseudoholomorphic curves in symplectization the scheme of which can be now easily modified via the perturbations of J utilizing the scheme of our proof of Theorem 5.1.

Because the existing literature on the pseudoholomorphic curves on symplectization lacks an explicit formula of the asymptotic operator as given in Theorem 6.1, it has been the case that the general abstract perturbation theory of linear operators in Kato [18] is just quoted in their study of asymptotic operators which prevents one from making any statement on specific dependence on the compatible almost complex structures. (See [45, Section 3.2] especially see Lemma 3.17, Theorem 3.35 and Appendix C of [45] for example which in turn extends the exposition given in [15] to higher dimensions.)

As far as we can see, the aforementioned status of matter makes the existing description in the literature, at least, of the spectral flow related to the index formula of the linearized operator for the -pseudoholomorphic curves on the fundamental case of ℝ × S1 for the closed string case (or on ℝ × [0, 1] for the open string case) in symplectization or in SFT rather unsatisfying and incomplete. This is because the spectral flow definition over the real line ℝ ≅ (0, 1) starts from the requirement that the asymptotic operators

A±=A(λ,J,)π,±:Γ(γ±ξ)Γ(γ±ξ)

associated to the given have simple eigenvalues at the end points; note, however, that these operators may have eigenvalues with multiplicity before making some perturbation. (Such a requirement is vacuous over the closed circle S1 considered as in the original article [3].) Furthermore one needs this description simultaneously over all asymptotic Reeb orbits for the given almost complex structure . To achieve this requirement the existing studies of such a spectral flow representation in the literature had to use the much bigger perturbations of abstract Fredholm operators in the linearized level but not in the original off-shell nonlinear level. Furthermore the perturbations are made depending on each asymptotic orbit which is not a priori given, especially when even its existence is not known. (See [45, Lemma 3.17 & Appendix C], for example.) Under such current circumstances, it is very cumbersome to develop, for example, a gluing theory leading to the Kuranishi structures compatible over different stata of the compactified moduli space of pseudoholomorphic curves entering in the SFT compactification.

Our Theorem 1.7 cures at least this unsatisfying point on the spectral flow description of the index in the literature by considering the natural family

τAτ:Γ(γτξ)Γ(γτξ);γτ=w(τ,)

which connects the asymptotic operator at τ = ± ∞ for the given J, after one single perturbation thereof in the off-shell level (See (4.3) for the precise expression.) We believe that the kind of perturbation result stated in Theorem 5.1 and Theorem 1.8 will play some role in the construction of Kuranishi structures on the moduli space of finite energy contact instantons so that certain natural functors can be defined in our Fukaya-type category of contact manifolds [21], [20]. (See [5] for a construction of semi-global Kuranishi structures in their definition of contact homology, where the simpleness properties of the eigenvalues of the asymptotic operators is utilized in an essential way.)

We also refer readers to the arXiv version [19] of the present paper for some application to the finer study of asymptotic convergence of contact instantons which simplifies the exposition given in the literature on the pseudoholomorphic curves on symplectization via the systematic tensorial calculus.

Finally we would like to just mention that the same asymptotic study can be made by now in a straightforward way by incorporating the boundary condition as done in [39], [28], [33].

The present work has been first announced in the survey paper [34] submitted for MATRIX Annals for the IBS-CGP and MATRIX workshop on Symplectic Topology held for December 5–16, 2022.

2 Recollection of the analysis of contact instantons

In this section, we recall the basic Weitzenböck-type identities and the various analytical results on contact instantons established via covariant tensorial approach in [36], [37], [38], [27], [28], [29], [35] and [39]. Our study of asymptotic operators of contact instantons and of pseudoholomorphic curves on symplectization also follows this tensorial approach and is based on this analytical foundation on contact instantons.

Denote by (Σ̇, j) a punctured Riemann surface (including the case of closed Riemann surfaces without punctures). The following definition is introduced in [37].

Definition 2.1

(Contact Cauchy–Riemann map). A smooth map w : Σ̇M is called a contact CauchyRiemann map (with respect to the contact triad (M, λ, J)), if w satisfies the following Cauchy–Riemann equation

¯πw:=¯j,Jπw:=12(πdw+Jπdwj)=0.

Recall that for a fixed smooth map w : Σ̇M, the triple

(wξ,wJ,wgξ)

becomes a Hermitian vector bundle over the punctured Riemann surface (Σ̇, j). This introduces a Hermitian bundle structure on Hom(TΣ̇, w* ξ) ≅ T* Σ̇w* ξ over Σ̇, with inner product given by

αζ,βη=h(α,β)gξ(ζ,η),

where α, βΩ1(Σ̇), ζ, ηΓ(w* ξ), and h is the Kähler metric on the punctured Riemann surface (Σ̇, j).

Let ∇π be the contact Hermitian connection. Combining the pulling-back of this connection and the Levi–Civita connection of the Riemann surface, we get a Hermitian connection for the bundle T* Σ̇w* ξΣ̇, which we will still denote by ∇π by a slight abuse of notation. This is the setting where we apply the harmonic theory and Weitzenböck formulae to study the global a priori W1,2-estimate of dπ w: The smooth map w has an associated π-harmonic energy density, the function eπ (w) : Σ̇ → ℝ defined by

eπ(w)(z):=|dπw|2(z).

(Here we use |⋅| to denote the norm from 〈⋅, ⋅〉 which should be clear from the context.)

However the contact Cauchy–Riemann equation itself π w = 0 does not form an elliptic system. By augmenting the closedness condition d(w* λj) = 0 to contact Cauchy–Riemann map equation π w = 0, we arrive at an elliptic system

¯πw=0,d(wλj)=0. (2.1)

2.1 Fundamental equation of contact Cauchy–Riemann maps

The following fundamental identity is derived in [37].

Theorem 2.2

(Fundamental Equation; Theorem 4.2 in [37]). Let w be a contact CauchyRiemann map, i.e. a solution of π w = 0. Then

dπ(dπw)=wλj12(LRλJ)dπw. (2.2)

The following elegant expression of the Fundamental Equation in isothermal coordinates (x, y), i.e. one such that z = x + iy provides a complex coordinate of (Σ̇, j) such that h = dx2 + dy2, will be extremely useful for the study of higher a priori Ck,α Hölder estimates.

Corollary 2.3

(Fundamental Equation in Isothermal Coordinates). Let (x, y) be isothermal coordinates. Write ζ := πwτ for a section of w* ξM. Then

xπζ+Jyπζ12λwx(LRλJ)ζ+12λwy(LRλJ)Jζ=0. (2.3)

The fundamental equation in cylindrical (or strip-like) coordinates is nothing but the linearization equation of the contact Cauchy–Riemann equation in the direction τ . This plays an important role in the derivation of the exponential decay of the derivatives at cylindrical ends; see [38, Part II].

2.2 Generic nondegeneracy of Reeb orbits and of Reeb chords

Nondegeneracy of closed Reeb orbits or of Reeb chords is fundamental in the Fredholm property of the linearized operator of contact instanton equations as well as of pseudoholomorphic curves on symplectization.

2.2.1 The case of closed Reeb orbits

Let γ be a closed Reeb orbit of period T ≠ 0. In other words, γ: ℝ → M is a solution of = Rλ(x) satisfying γ(T) = γ(0). By definition, we can write γ(T) = ϕRλT (γ(0)) for the Reeb flow ϕRλT of the Reeb vector field Rλ. Therefore if γ is a closed orbit, then we have

ϕRλT(γ(0))=γ(0)

i.e. p = γ(0) is a fixed point of the diffeomorphism ϕRλT . Since 𝓛Rλ λ = 0, ϕRλT is a (strict) contact diffeomorphism and so induces an isomorphism

dϕRλT(p)|ξp:ξpξp

which is the linearization restricted to ξp of the Poincaré return map.

Definition 2.5

We say that a T-closed Reeb orbit (T, λ) is nondegenerate if d ϕRλT (p)|ξp : ξpξp with p = γ(0) does not have the eigenvalue 1.

The following generic nondegeneracy result is proved by Albers–Bramham–Wendl in [2].

Theorem 2.5

(Albers–Bramham–Wendl). Let (Q, ξ) be a contact manifold. Then there exists a residual subset 𝓒reg(Q, ξ) ⊂ 𝓒(Q, ξ) such that for any contact form λ ∈ 𝓒reg(Q, ξ) all Reeb orbits are nondegenerate for T ≠ 0.

The case T = 0 can be included as the Morse–Bott nondegenerate case if we allow the action T = 0 by extending the definition of a Reeb trajectory to isospeed Reeb chords of the pairs (γ, T) with γ:[0, 1] → Q with T = ∫ γ* λ as done in [32; 28].

2.2.2 The case of Reeb chords

Let R0, R1 be a pair of Legendrian submanifolds. We first recall the notion of isospeed Reeb trajectories used in [32] and the definition of nondegeneracy of thereof.

Consider contact triads (Q, λ, J) and the boundary value problem for (γ, T) with γ : [0, 1] → Q

γ˙(t)=TRλ(γ(t)),γ(0)R0,γ(1)R1. (2.4)

Definition 2.6

(Isospeed Reeb trajectory; Definition 2.1 in [28). We call a pair (γ, T) consisting of a smooth curve γ : [0, 1] → Q and T ∈ ℝ an isospeed Reeb trajectory if they satisfy

γ˙(t)=TRλ(γ(t)),γλ=T (2.5)

for all t ∈ [0, 1]. We call (γ, T) an isospeed closed Reeb orbit if γ(0) = γ(1), and an isospeed Reeb chord of (R0, R1) it γ(0) ∈ R0 and γ(1) ∈ R1 from R0 to R1.

With this definition, we state the corresponding notion of nondegeneracy:

Definition 2.7

We say a Reeb chord (γ, T) of (R0, R1) is nondegenerate if the linearization map d ϕRλT (p) : ξpξp satisfies

dϕRλT(p)(Tγ(0)R0)Tγ(1)R1in ξγ(1),or equivalentlydϕRλT(p)(Tγ(0)R0)Tγ(1)ZR1 in Tγ(1)Q.

Here ϕRλT is the flow generated by the Reeb vector field Rλ.

More generally, we consider the following situation. We recall the definition of Reeb trace ZR of a Legendrian submanifold R, which is defined to be

ZR:=tRϕRλt(R).

See [28, Appendix B] for a detailed discussion on its genericity.

Definition 2.8

(Nondegeneracy of Legendrian links). Let R⃗ = (R1, …, Rk) be a chain of Legendrian submanifolds, which we call an (ordered) Legendrian link. We say that the Legendrian link R⃗ is nondegenerate if it satisfies

ZRiRj

for all i, j = 1, …, k.

We now provide the off-shell framework for the proof of nondegeneracy in general. Denote 𝓛(Q) = C(S1, Q) the space of loops z : S1 = ℝ /ℤ → Q. We denote by

C(Q,ξ)

the set of contact forms of (Q, ξ) equipped with C-topology. We denote by 𝓛(Q; R0, R1) the space of paths

γ:([0,1],{0,1})(Q;R0,R1).

We consider the assignment

Φ:(T,γ,λ)γ˙TRλ(γ) (2.6)

as a section of the Banach vector bundle over

(0,)×L1,2(Q;R0,R1)×C(Q,ξ)

where 𝓛1,2(Q; R0, R1) is the W1,2-completion of 𝓛(Q; R0, R1). We have

γ˙TRλ(γ)Γ(γTQ;Tγ(0)R0,Tγ(1)R1).

We define the vector bundle

L2(R0,R1)(0,)×L1,2(Q;R0,R1)×C(Q,ξ)

whose fiber at (T, γ, λ) is L2(γ* TQ). We denote by πi, i = 1, 2, 3 the corresponding projections as before.

We denote ℜ𝔢𝔢𝔟(M, λ; R0, R1) = Φλ1 (0), where

Φλ:=Φ|(0,)×L1,2(Q;R0,R1)×{λ}.

Then we have

Reeb(λ;R0,R1)=Φλ1(0)=Reeb(Q,ξ)π31(λ).

The following relative version of Theorem 2.5 is proved in [28, Appendix B].

Theorem 2.9

(Perturbation of contact forms; Theorem B.3 in [28]). Let (Q, ξ) be a contact manifold. Let (R0, R1) be a pair of Legendrian submanifolds, allowing the case R0 = R1. There exists a residual subset C1reg (Q, ξ) ⊂ 𝓒(Q, ξ) such that for any λ C1reg (Q, ξ) all Reeb chords from R0 to R1 are nondegenerate for T ≠ 0 and BottMorse nondegenerate when T = 0.

The following theorem is also proved in [28].

Theorem 2.10

(Perturbation of boundaries; Theorem B.10 in [28]). Let (Q, ξ) be a contact manifold. Let (R0, R1) be a pair of Legendrian submanifolds, allowing the case R0 = R1. For a given contact form λ and R1, there exists a residual subset

Legreg(Q,ξ)Leg(Q,ξ)

of Legendrian submanifolds such that for all R0Legreg(Q, ξ), all Reeb chords from R0 to R1 are nondegenerate for T ≠ 0 and MorseBott nondegenerate when T = 0.

We refer readers to [28, Appendix B] for the proofs of these results.

2.3 Exponential asymptotic C convergence of contact instantons

In this section, we recall the asymptotic behavior of contact instantons on the Riemann surface (Σ̇, j) associated with a metric h with cylinder-like ends for the closed string context and with strip-like ends for the open string context.

We assume that there exists a compact set KΣΣ̇, such that Σ̇-Int(KΣ) is a disjoint union of interior-punctured disks each of which is isometric to the half cylinder [0, ∞) × S1 or the half strip (−∞, 0] × S1, where the choice of positive or negative strips depends on the choice of analytic coordinates at the punctures. We denote by {pi+}i=1,,l+ the positive punctures, and by {pj}j=1,,l the negative punctures. Here l = l+ + l. The case of boundary-punctured disks is similar. Then we denote by ϕi± such cylinder-like or strip-like coordinates depending on whether they are boundary or interior punctures.

We separately describe the cases of interior punctures and of boundary punctures. We first mainly state our assumptions for the study of the behavior of boundary punctures. (The case of interior punctures is treated in [37, Section 6] and will be briefly mentioned at the end of this section.)

Definition 2.11

Let Σ̇ be a punctured Riemann surface of genus zero with boundary punctures {pi+}i=1,,l+ {pj}j=1,,l equipped with a metric h with strip-like ends outside a compact subset KΣ. Let w : Σ̇M be any smooth map with Legendrian boundary condition. We define the total π-harmonic energy Eπ(w) by

Eπ(w)=E(λ,J;Σ˙,h)π(w)=12Σ˙|dπw|2 (2.7)

where the norm is taken in terms of the given metric h on Σ̇ and the triad metric on M.

Throughout this section, we work locally near one boundary puncture p, i.e., on a punctured semi-disc Dδ(p) ∖ {p}. By taking the associated conformal coordinates ϕ+ = (τ, t) : Dδ(p) ∖ {p} → [0, ∞) × [0, 1] such that h = 2 + dt2, we need only look at a map w defined on the semi-strip [0, ∞) × [0, 1] without loss of generality.

Under the nondegeneracy hypothesis from Definition 2.8 and the transversality hypothesis, the exponential C convergence is derived from the subsequence convergence and the charge-vanishing result in [37], [39]. For readers’ convenience and for the self-containedness of the paper, we recall the subsequence and charge in Appendix C, and the exponential convergence result in this subsection.

Suppose that the tuple R⃗ = (R0, …, Rk) is transversal in the sense all pairwise Reeb chords are nondegenerate. In particular we assume that the entries of the tuple are pairwise disjoint. Firstly, we state the following C0-exponential convergence from [38] for the closed string and from [39] for the open string case.

Proposition 2.12

(Proposition 11.23 in [38], Proposition 6.5 in [39]). There exist some constants C > 0, δ > 0 and τ0 large enough such that for any τ > τ0,

dw(τ,),γ()C0([0,1])Ceδτ

where we have 0 < δ < |μ1| with the first negative eigenvalue μ1 of the asymptotic operator A(λ,J,)π of γ.

Once the above C0-exponential decay is established, the C-exponential convergence w(τ, ⋅) → γ follows by establishing the C-exponential decay of

dwRλ(w)dt.

The proof of the latter decay is carried out in [39] by an alternating bootstrap argument by decomposing

dw=dπw+wλRλ

as follows. Let z = x + i y be any isothermal coordinates on (D2, ∂ D2) ⊂ (Σ̇, ∂ Σ̇) adapted to the boundary, i.e. satisfying that x is tangent to ∂ Σ̇. We set

ζ:=dπw(x),χ:=λwy+1λwx.

Then we show that the fundamental equation (2.2) is transformed into the following system of equations for the pair (ζ, α)

xπζ+Jyπζ+12λ(wy)(LRλJ)ζ12λ(wx)(LRλJ)Jζ=0ζ(z)TRifor zD2 (2.8)

and

¯χ=12|ζ|2χ(z)Rfor zD2 (2.9)

for some i = 0, …, k. With this coupled system of equations for (ζ, χ) at our disposal, the proof follows the alternating bootstrap between ζ and χ similarly as in the proof of higher regularity results carried out by the alternating bootstrap argument in [29; 39].

Combining this and the elliptic Ck,α-estimates given in [37; 39], the proof of the C-convergence of w(τ, ⋅) → γ as τ → ∞ is completed.

So far we have recollected various foundational analytic results on contact instantons both in the closed and in the open string context, which will be used later in our study of asymptotic operators. Since the same arguments can be applied to the open string case with by now straightforward incorporation of the boundary condition exercised in [39], [33], [28] and [40], we will focus on the case of closed strings to streamline our study of asymptotic operators and to highlight the main points of our approach in the rest of the paper.

2.4 Exponential convergence of pseudoholomorphic curves on symplectization

Finally we make a brief mention on how the exponential convergence for pseudoholomorphic curves on symplectization follows from that of contact instantons now. We consider the symplectization

M=Q×R,ω=d(esπλ)=es(dsπλ+dπλ)

of the contact manifold (Q, ξ) equipped with contact form λ.

On Q, the Reeb vector field Rλ associated to the contact form λ is the unique vector field X =: Rλ satisfying

Xλ=1,Xdλ=0. (2.10)

We call (y, s) the cylindrical coordinates. On the cylinder [0, ∞) × Q ⊂ (− ∞, ∞) × Q, we have the natural splitting (1.4) of the tangent bundle TM.

Now we describe a special family of almost complex structure compatible to the given cylindrical structure of M.

Definition 2.13

An almost complex structure J on Q × (0, ∞) is called λ-compatible if it is split into

J=JξJ0:TMξR2TMξR2

where Jξ is compatible to |ξ and J0 : ℝ2 → ℝ2 maps s to Rλ.

For our purpose, we will need to consider a family of symplectic forms to which the given J is compatible and their associated metrics. For any λ-compatible J, the J-compatible metric associated to ω is expressed as

g(ω,J)=ds2+gQ (2.11)

on ℝ × Q. Now we regard the triple (ω, J, g(ω,J)) as an almost Hermitian manifold near the level surface s = 1. We then fix the canonical connection ∇ associated to (ω, J, g(ω,J)); see Appendix A.

The following is a general property of the canonical connection.

Proposition 2.4

Let (W, ω, J) be an almost Hermitian manifold and letbe the canonical connection. Denote by T its torsion tensor. Then

T(JY,Y)=0 (2.12)

for all vector fields Y on W.

Consider the decomposition (1.4) and the canonical connection ∇͠ on Q × ℝ, which in particular is J-linear. Recalling the expression u = (w, f) with f = su and w = πQu for each map u : Σ̇Q × ℝ, we know that if it is J-holomorphic, it satisfies

¯πw=0,wj=df

on Σ̇. In particular w is a contact instanton. Then we have already shown the exponential convergence w to the Reeb orbit w(⋅ , t).

For the convergence of f, we use the equation

wλj=df

which is equivalent to

wλ=dfj

By taking the differential, we obtain

d(dfj)=wdλ=12|dπw|2dτdt.

Firstly, this equation shows that f is a subharmonic function. More explicitly, we derive that

2fτ2+2ft212|dπw|2=0 (2.13)

where we know from Theorem C.5 that the convergence 12 |dπ w|2T2 is exponentially fast. This immediately gives rise to the following exponential convergence of the radial component, which will complete the study of asymptotic convergence property of finite energy pseudoholomorphic planes in symplectization.

Proposition 2.15

(Exponential convergence of radial component). Let u = (w, f) be a finite energy J͠-holomorphic plane in Q × ℝ. Then we have the convergence

dfTdτ

exponentially fast. More precisely, there exists some c ∈ ℝ such that

|f(τ,t)(Tτ+c)|0

as τ → ∞ exponentially fast.

Proof

We have already established that w* λTdt before in Corollary C.6. By composing it with j, the statement follows. □

3 Covariant linearization operator and its Fredholm theory

In this section, we work out the Fredholm theories of pseudoholomorphic curves on symplectization by recalling the exposition given in [27] for the case of contact instantons and the one given in [35] for the case of 𝔩𝔠𝔰-instantons. The zero-temperature limit of the latter also provides the relevant Fredholm theory for pseudoholomorphic curves just by incorporating the presence of the ℝ-factor in the product M = Q2n−1 × ℝ into that of contact instantons. Explanation of this point is now in order.

Let Σ be a closed Riemann surface and let Σ̇ be its associated punctured Riemann surface. We allow the set of punctures to be empty, i.e., Σ̇ = Σ. We recall the splitting TM = ξ ⊕ 𝓥 from (1.4). Using this splitting, we would like to regard the assignment uJ u which can be decomposed into

u=(w,f)¯πw,wλjfds=:Υ(u)

for a map w : Σ̇Q as a section of the (infinite dimensional) vector bundle over the space of maps of w. In this section, we lay out the precise relevant off-shell framework of functional analysis. We recall the definition of the (0, 1)-projection of the covariant differential dπ

¯π:=12π+Jπj.

(We recall the definition of general covariant differential in Appendix B, which in particular applies to dπ.) We decompose

Υ(u)=(Υ1(u),Υ2(u))

into the ξ and the Reeb components respectively. Then we have the formulae

Υ1(u)=¯πu,Υ2(u)=wλjfds. (3.1)

Theorem 3.1

(Theorem 10.1 in [35]). We decompose dπ = dπ w + w* λRλ and Y = Yπ + λ(Y) Rλ, and X = (Y, v) ∈ Ω0(w* T(Q × ℝ)). Denote κ = λ(Y) and b = ds(v). Then we have

DΥ1(u)(Y,v)=¯πYπ+B(0,1)(Yπ)+Tdwπ,(0,1)(Yπ)+12κLRλJ)J(πw (3.2)
DΥ2(u)(Y,v)=w(LYλ)jLvds=dκjdb+w(Ydλ)j (3.3)

where B(0,1) and Tdwπ,(0,1) are the (0, 1)-components of B and Tdwπ and B, Tdwπ : Ω0(w* TQ) → Ω1(w* ξ) are zero-order differential operators given by

B(Y)=12wλ(LRλJ)JYandTdwπ(Y)=πT(Y,dw).

Now we consider a punctured Riemann surface Σ̇. We consider some choice of weighted Sobolev spaces

Wδ;ηk,pΣ˙,Q×R;γ+,γ

as the off-shell function space and linearize the map

(w,f~)¯πw,df~.

This linearization operator then becomes cylindrical in cylindrical coordinates near the punctures. The Fredholm property of the linearization map

DΥ(λ,J)(u):Ωk,p;δ0(uT(Q×R);J;γ+,γ)Ωk1,p;δ(0,1)(wξ)Ωk1,p(0,1)(uV)

and its index are computed in [27], [28] and in [39] respectively.

We briefly recall the aforementioned Fredholm property here. We have the decomposition

Ωk,p;δ0(wT(Q×R);J;γ+,γ)=Ωk,p;δ0(wξ)Ωk,p;δ0(uV) (3.4)

and again the operator

DΥ(λ,J)(u):Ωk,p;δ0(wT(Q×R);J;γ+,γ)Ωk1,p;δ(0,1)(wξ)Ωk1,p;δ(0,1)(uV) (3.5)

which is decomposed into

DΥ1(u)(Y,v)DΥ2(u)(Y,v)

where the summands are given as in (3.2) and (3.3) respectively.

In terms of the decomposition (3.4), the linearized operator (λ,J)(u) can be written as in the following succinct matrix form

¯π+B(0,1)+Tdwπ,(0,1),12()(LRλJ)J(πw)()πdλj,¯. (3.6)

For the calculation of the index of the linearized operator, we would like to homotope to the block-diagonal form, i.e. into the direct sum operator

¯π+B(0,1)+Tdwπ,(0,1)¯=Dw¯π¯

via a continuous path of Fredholm operators. The Fredholm property of all elements in the path used in [27] and [35] relies on the following asymptotic property of the off-diagonal terms. This was implicitly used in the calculation of the index in [27] and [35], and its proof is given in [34, Proposition 13.6]. For readers’ convenience, we reproduce its proof here to show how the asymptotic exponential convergence and the explicit formula of the linearized operator enter in the asymptotic property.

Proposition 3.2

The off-diagonal terms decay exponentially fast as |τ| → ∞. In particular, the off-diagonal term is a compact operator relatively to the diagonal term.

Proof

For the (1, 2)-term of the matrix (3.6), we derive

wτπ+Jwtπ=0

from the equation π w = 0 by evaluating it against τ . Therefore we have

πwτ=12wτπJwtπ=Jwtπ.

By the exponential convergence wt T Rλ(γ(t)), we derive

Jπwτ=wtπ0

since wt T Rλ.

For the (2, 1)-term, we evaluate

(Yπdλ)jτ=dλY,wt(Yπdλ)jt=dλY,wτ.

Therefore we have derived

(Yπdλ)j=dλY,wtdτdλY,wτdt.

This proves that

()πdλjdλ,wtdτdλ,wτdt

as |τ| → ∞. Since wt T Rλ, the first term converges to zero, and the second term converges to

dλ(,JTRλ)=Tdλ,s=0,

all exponentially fast.

Combining the two, we infer that the off-diagonal term converges to the zero operator exponentially fast. The last statement about the relatively compactness is an immediate consequence of this exponential decay. □

Because of this asymptotic vanishing, the path

s[0,1]¯π+B(0,1)+Tdwπ,(0,1),1s2()(LRλJ)J(πw)(1s)()πdλj,¯=:Ls (3.7)

carries the same asymptotic operator and hence is a continuous path of Fredholm operators

Ls:Ωk,p;δ0(wT(Q×R);J;γ+,γ)Ωk1,p;δ(0,1)(wξ)Ωk1,p;δ(0,1)(uV)

such that L0 = (λ,J)(u) and

L1=¯π+B(0,1)+Tdwπ,(0,1)¯=Dw¯π¯.

Therefore we have only to compute the index of the diagonal operator L1 which was given in [27] and [35].

4 Definition of the asymptotic operator and its covariant formulae

Recall that for any -holomorphic curve (w, f) on the symplectization, w is a contact instanton for J on Q. Furthermore we have

(w,f)T(Q×R)=wTQfTR=wξspanRs,Rλ.

The last splitting is respected by the canonical connection of the almost Hermitian manifold

(Q×R,dλ+dsλ,J~).

Indeed, we have

can=π0

where ∇π = π ∇|ξ and ∇0 is the trivial connection on span { s , Rλ}.

Remark 4.1

Here again we would like to emphasize the usage of the canonical connection of the almost Hermitian manifold, not the Levi–Civita connection, admits this splitting.

Now we study a finer analysis of the asymptotic behavior along the Reeb orbit. Our discussion thereof is close to the one given in [38, Sections 11.2 and 11.5] where the more general Morse–Bott case is studied.

For this purpose, we evaluate the linearization operator against τ . We have already checked that the off-diagonal terms of the matrix representation of (w) decay exponentially fast in the direction τ in the previous section, and so we have only to examine the diagonal terms 1(w) and 2(w). First we consider 2 and rewrite

DΥ2(w)=¯=12(τ+it).

Therefore we have the asymptotic operator

A(λ,J,):=it (4.1)

which does not depend on the choice of J ∈ 𝓙λ(Q, ξ). The eigenfunction expansions for this operator is nothing but the standard Fourier series for the functions in L2(S1, ℂ).

This being said, we now focus on the Q-component 1(w) of the asymptotic operator, and compute

D¯π(w)τ=12(τπ+Jtπ)+Tdwπ,(0,1)τ+B(0,1)τ. (4.2)

In fact, this is nothing but the left hand side of (2.3) by the calculation of the torsion term, which we omit since we have already have the formula (2.3). We write

D¯π(w)τ=12τπA(λ,J,)τ.

and define the family of operators

Aτ=A(λ,J,)τ:Γ(wτξ)Γ(wτξ)

given by the formula

Aτ:=JtπTdwπ,(0,1)τ+B(0,1)τ. (4.3)

Thanks to the exponential convergence of wτ = w(τ, ⋅) → γ± as τ → ± ∞, we can take the limit of the conjugate operators

ΠτA(λ,J,)τ(Πτ)1:Γ(γ±ξ)Γ(γ±ξ) (4.4)

as τ → ± ∞ respectively, where Πτ is the parallel transport along the short geodesics from w(τ, t) to w(∞, t). This conjugate is defined for all sufficiently large |τ|.

Since the discussion at τ = − ∞ will be the same, we will focus our discussion on the case at τ = + ∞ from now on.

Definition 4.2

(Asymptotic operator). Let (τ, t) be the cylindrical (or strip-like) coordinate, and let ∇π be the almost Hermitian connection on w* ξ induced by the contact triad connection ∇ of (Q, λ, J). We define the asymptotic operator of a contact instanton w to be the limit operator

A(λ,J,)π:=limτ+ΠτA(λ,J,)τ(Πτ)1. (4.5)

Obviously we can define the asymptotic operator at negative punctures in the similar way.

4.1 Asymptotic operator in contact triad connection

Now we find the formula for the above limit operator with respect to the contact triad connection As T(Rλ, ⋅ ) = 0, wτπ=Jwtπ andwt(τ,)TRλ exponentially fast, we obtain

Tdwπ,(0,1)(τ)=Tπwτ,0.

On the other hand, we have

2B(0,1)τ=12λwτLRλJ12λwtJLRλJ

This converges to T2JRλ since w* λT dt as τ → ∞.

This immediately gives rise to the following simple explicit formula for the asymptotic operator.

Proposition 4.3

Letbe the contact triad connection associated to any compatible pair (λ, J). Then the asymptotic operator A(λ,J,)π is given by

A(λ,J,)π=Jt+T2LRλJJ

In particular, it induces a (real) self-adjoint operator on (ξ, g|ξ) with respect to the triad metric g.

Proof

We have already shown the formula above. The symmetry of the opeartor −Jt can be directly checked from the J-linearity of ∇π by the integration by parts.

Now we recall from [6] that that the operator 𝓛Rλ J J defines a symmetric operator on Γ(γ* ξ) by the following general identity and hence so is A(λ,J,)π .

Lemma 4.4

(Lemma 6.2 in [6]). The linear maps 𝓛Rλ J J and 𝓛Rλ J are (pointwise) symmetric with respect to the triad metric.

This finishes the proof of the proposition. □

Proposition 4.3 enables us to derive the following.

Proposition 4.5

We have

RλY=[Rλ,Y]+12(LRλJ)JY=LRλY+12(LRλJ)JY (4.6)

for all Yξ.

Proof

We recall the formula

YRλ=12(LRλJ)JY

from (A.2). By the definition of the torsion, we also have

YRλ=RλY+[Y,Rλ]+T(Y,Rλ).

By combining these with the torsion property T(⋅ , Rλ) = 0 of the triad connection, we obtain the first equality. The second is just the definition of the Lie derivative 𝓛Rλ acting upon vector fields. □

We note that the right hand side formula in (4.6) is canonically defined depending only on λ and J, independent of the choice of the connection. In other words, we can derive the first variation of the operator ∇Rλ|ξ associated to the triad connection of (Q, λ, J) with respect to the compatible pair (λ, J) or with respect to J when λ is fixed.

Now in the context of pseudoholomorphic curves on symplectization, we proceed by decomposing the full asymptotic operator A(λ,J,)π of the pseudoholomorphic curves

A(λ,J,)π:γξCγξC

into

A(λ,J,)=A(λ,J,)πA(λ,J,).

Here ℂ stands for the pull-back bundle

(πuτ)Rs,Rλ=wτRs,RλRs,Rλ

which is canonically trivialized, and hence may be regarded as a vector bundle over the curves wτ on Q. (Compare this with [41, Definition 2.28].)

4.2 Asymptotic operator in Levi–Civita connection

Up until now, we have emphasized the usage of the triad connection which gives rise to an optimal form of tensorial expressions. We recall that ∇Y J = 0 for all Yξ for the contact triad connection ∇ by one of the defining axioms of the contact triad connection. We have ∇Rλ J ≠ 0 for the connection (in fact, ∇Rλ J = 0 if and only if Rλ is a Killing vector field, i.e. ∇ Rλ = 0 with respect to ∇; see [37, Remark 2.4]). On the other hand, while YLCJ ≠ 0 for the Levi–Civita connection ∇LC in general, the Levi–Civita connection has the following surprising property

RλLCJ=0,

see Lemma 6.1 in [6], Proposition 4 in [36]. Although it will not be used in the present paper, we also convert the formula of the asymptotic operator A(λ,J,∇) into one written in terms of the Levi–Civita connection for a possible future purpose.

For this purpose, the following lemma is crucial.

Lemma 4.6

For any YΓ(ξ), we have

RλY=RλLCY12JY. (4.7)

Proof

Applying the torsion property T(Rλ, Y) = 0 of the triad connection ∇ and the torsionfreeness of the Levi–Civita connection, (4.7) is equivalent to

YRλ+[Rλ,Y]=YLCRλ+[Rλ,Y]12JY.

By cancelling the bracket terms, it is enough to show that

YRλ=YLCRλ12JY.

This follows from the formula ∇Y Rλ = 12 (𝓛Rλ J) J, see (A.2), and

YLCRλ=12JY+12(LRλJ)J

for all Yξ, see Lemma 6.2 in [6], Lemma 9 in [36]. □

Therefore we have derived the simple relationship

RλLC=12J+Rλ

between RλLC and ∇Rλ on Γ(ξ). Then taking the pull-back of this identity along the loop γ and using Proposition 4.3, we can rewrite the asymptotic operator in terms of the Levi–Civita connection as follows.

Corollary 4.7

Let (λ, J) be any compatible pair and letLC be the LeviCivita connection of the triad metric of (Q, λ, J). Let w be a contact instanton with its asymptotic limit γ = w(± ∞, ⋅) in cylindrical coordinates (τ, t) near a puncture. Assume that

γλ=T0

and let A(λ,J,∇) be the asymptotic operator of w. Then we have

  1. [tLC,J](=tLCJ)=0 and

  2. A(λ,J,)π=JtLCT2Id+T2(LRλJ)J. (4.8)

It follows from Corollary 4.7 that the self-adjoint operator

A(λ,J,)π:L2(γξ)L2(γξ)

can be decomposed into A = A′ + A″ where A′ and A″ are the J-linear and the anti-J-linear parts: We have

A=JtLCT2Id,A=T2LRλJJ (4.9)

where A′ defines a compact operator on Dom( A(λ,J,)π ) on a self-adjoint extension of A(λ,J,)π which we regard as a linear map W1,2(γ* ξ) → L2(γ* ξ).

The same kind of property also holds for the open string case for the Legendrian pair (R0, R1). The explicit formula for the asymptotic operator given above enables us to study a series of perturbation results on the eigenfunctions and eigenvalues of the asymptotic operators under the perturbation of J’s inside the set 𝓙λ of λ-compatible CR almost complex structures J. Discussion on this perturbation theory is now in order.

5 Spectral analysis of asymptotic operators

In this section, we derive consequences of Corollary 4.7 assuming T ≠ 0.

The ellipticity of A(λ,J,)π implies that it has a discrete set of eigenvalues, which we enumerate as

<μk<<μ1<0<μ1<<μk<

with repeated finite multiplicities allowed, where μk → ∞ (respectively μk → −∞) as k → ∞. It also has a uniform spectral gap (by [18, Theorem 6.29]) in that there exists some d = d( A(λ,J,)π ) > 0 such that

|μi+1μi|d>0for all i. (5.1)

The explicit form of the asymptotic operator on J and λ given in the previous section enables us to study the perturbation theory of the asymptotic operator in terms of the change of J, which we do in the next section. In the rest of this section, we prove the following generic simpleness statement using the perturbation theory of self-adjoint operators.

Theorem 5.1

(Simpleness of eigenvalues). Let (Q, ξ) be a contact manifold. Assume that λ is nondegenerate. For a generic choice of a compatible pair (λ, J), all eigenvalues μi of the asymptotic operator are simple for all closed Reeb orbits.

We first outline the scheme of the proof as follows:

  1. For each k, we take the spectral decomposition

    W2,2(γξ)=MkMk

    where Mk := ker( A(λ,J,)π μk Id) is the eigenspace with eigenvlue μi. We denote by Πk : L2(γ* ξ) → L2(γ* ξ) the idempotent associated to the orthogonal projection πk : L2(γ* ξ) → Mk. We know that Mk is a finite-dimensional subspace and that A(λ,J,)π restricts to a diagonalizable linear map denoted by Ak;γ(J) : MkMk for each k ≥ 0. We also define the associated map

    Ak,γ:JλEnd(Mk):JAk;γ(J). (5.2)
  2. For each k, we apply the perturbation theory of linear maps in finite-dimensional vector spaces, see [18, Chapter 1], and determine the set of all J for which the linear map Ak;γ(J) has simple eigenvalues.

5.1 Spectral perturbation theory under J’s

Let k ≥ 0 be given and let

W2,2(γξ)=MkMk

be the decomposition mentioned as above. We consider the diagonalizable linear map denoted by

Ak;γ(J):MkMk

given by restricting the operator A(λ,J,)π to Mk, i.e.

Ak;γ(J):=πkA(λ,J,)π|Mk

where πk is the L2-projection to Mk.

By choosing an L2-orthonormal basis of Mk, say

{e1,e2,,emk}=:Bk,

we define a map

JλEnd(Rmk):J[Ak;γ(J)]Bk

where [Ak;γ(J)]𝓑k is the matrix of Ak;γ(J) with respect to the basis 𝓑k. This matrix is a symmetric matrix.

Definition 5.2

Define Symsimp(ℝmk) to be the set of all symmetric matrices that have simple eigenvalues. We call the complement

Sym(Rmk)Symsimp(Rmk) (5.3)

the discriminantal variety.

Here the following remark provides that the complement (5.3) is an algebraic variety for which we can apply the (stratawise) transversality theorem.

Remark 5.3

Consider the characteristic polynomial of Ak;γ(J) given by

pk,γ(J)(μ):=det(μIdAk;γ(J))

i.e.

pk,γ(J)=CharAk;γ(J)

where Char(A) := det((⋅) IA) is the characteristic polynomial of the matrix A. We denote its discriminant by Δ(pk,γ(J)); recall that the discriminant Δ(f) = Δn(f) in general is a certain homogeneous polynomial with degree 2n − 2 of the coefficients of a degree n polynomial

f=a0+a1x++anxn

such that f has a multiple root if and only if Δ(f) = 0. Therefore the zero set of Δ(f) defines a codimension 1 algebraic variety of ℝ P2n−2 (see [12, Chapter 12] for a summary of the properties of discriminants). Then (5.3) is the preimage

pk,γ1(Δ1(0))=(Δpk,γ)1(0).

Recall that the tangent space TJ 𝓙λ can be written as

TJJλ={BΓ(End(ξ))BJ+JB=0,dλ(B(),J())+dλ(J(),B())=0} (5.4)

where the second equation means nothing but that B is a symmetric endomorphism of the metric (⋅ , J ⋅)|ξ on ξ (see [10] and [26, p. 339] for a similar description in the symplectic case). We note that by definition TJ 𝓙λ can be expressed as a fiber bundle

SλQ (5.5)

whose fiber is given by Sλ,x = TJ 𝓙λ|x which is isomorphic to

Sλ,x={BEnd(R2n)BJ0+J0B=0,dλ(B(),J0())+dλ(J0(),B())=0}

where J0 is the standard complex structure of complex multiplication by 1 with the identification ℝ2n ≅ ℂn.

Now we have the following proposition the proof of which we postpone till the next subsection:

Proposition 5.4

Let λ be any nondegenerate contact form (Q, ξ), and let (γ, T) be an isospeed closed Reeb orbit with the loop γ : S1Q andγ* λ = T. Consider the pull-back fiber bundle γ* SλS1. Then the map

Ak;γ:JλEnd(Mk)Mmk×mk(R)

is stratawise transverse to (Δmk)−1(0) for all γ ∈ ℜ𝔢𝔢𝔟(λ).

Postponing the proof of this proposition till the next subsection, we proceed with the proof of Theorem 5.1.

This proposition proves that the subset Ak;γ1(Δmk1(0)) is a stratified smooth submanifold of real codimension 1 for each given k ≥ 1. In particular the complement

JλAk;γ1(Δmk1(0))

is a residual subset of 𝓙λ. Therefore their countable intersection

k1JλAk;γ1(Δmk1(0))=:Jλsm(γ)

is still a residual subset of 𝓙λ. Since the set ℜ𝔢𝔢𝔟(λ) is a countable set, the intersection

γReeb(λ)Jλsm(γ)

is still a residual subset of 𝓙λ. By definition, this last set consists precisely of those J’s for which the associated asymptotic operator carries simple eigenvalues.

This will complete the proof of Theorem 5.1, except for the proof of Proposition 5.4 which is now in order.

5.2 Proof of generic simpleness of eigenvalues

Let γ ∈ ℜ𝔢𝔢𝔟(λ) be given and k be fixed. We consider the assignment JA(λ,J,∇) as a map

F:JλFredW2,p(γξ),W1,p(γξ),

obtained by assigning the operator

A(λ,J,)π=Jt+T2(LRλJ)J

to J in (4.8) where Fred(W2,2(γ* ξ), W1,2(γ* ξ)) is the Banach space of Fredholm operators from W2,2(γ* ξ) to W1,2(γ* ξ). To compute the variation of the assignment

F:JA(λ,J,)π

we use (4.6) to convert the formula for A(λ,J,)π into the following one which explicitly shows that the operator does not depend on the choice of connection.

Theorem 5.5

Let (Q, λ, J) be a triad. Let (γ, T) be an isospeed Reeb trajectory. Then we have the formula

A(λ,J,)π=T12LRλJJLRλ12J(LRλJ) (5.6)

when acted upon Γ(γ* TM).

Proof

Recalling that γ̇(t) = TRλ(γ(t)) and the definition of the pull-back connection in general, we have

(tη)(t)=TRλ(γ(t))Y|γ(t)=TRλ(γ(t))Y|γ(t)

where ηΓ(γ* ξ) and Y is a vector field Y such that η(t) = Y(γ(t)) is locally defined near the point γ(t) for given tS1.

Therefore with a slight abuse of notation, utilizing (4.6), we rewrite

A(λ,J,)π=JTRλ+T2(LRλJ)J=T2JLRλJJTJLRλT2J(LRλJ)=T2LRλJTJLRλT2J(LRλJ)=T12LRλJJLRλ12J(LRλJ).

This finishes the proof. □

Therefore its variation under δ J = B is given by

dJF(B)=T12LRλBBLRλ12BLRλJ12JLRλB=T2Id+JLRλBTBLRλ+12LRλJ (5.7)

for B satisfying JB + BJ = 0.

In the following paragraph, we adapt the exposition of [17, p. 73] in our study of deformation of the asymptotic operators, which is used in the context of Gromov–Witten theory.

We fix the Fredholm index ι and denote the set of such Fredholm operators (with index ι) of dim ker = k by Fredk(J, γ; μ) and their union by Fredkι (γ; μ), i.e.

Fredkι(γ;μ)=JJλFredkι(J,γ;μ) (5.8)

as an infinite-dimensional fiber bundle over 𝓙λ. In general, by a theorem of Koschorke [24] we have

Fredι=k,γFredkι

where Fredkι of index ι is a submanifold of Fred with real codimension k(kι) : The normal bundle of Fredkι in Fred at an operator D ∈ Fredk is

HomRker(DμId),coker(DμId).

Considering the L2-adjoint, we may identify this with

HomR(ker(DμId),ker(DμId)).

To show that a similar statement holds for the union of subsets thereof

Fredι(γ;μ)=kFredkι(J,γ;μ), (5.9)

we need to verify that the restricted variation arising from the change of J is big enough.

Proposition 5.6

The map

dJAk;γ:TJJλTAk;γ(J)Mmk×mk(R)Mmk×mk(R)

is an epimorphism at every J ∈ (Δk;γpk;γ)−1(0).

Proof

To prove the proposition, we need to prove that for any element

0κ,cker(A(λ,J,)πμkId)Γ(γξ)

we can find a variation BTJ 𝓙λ such that

c,BκL20.

The following lemma is a fundamental ingredient, the counterpart of [45, Lemma C.1]. However our variation is restricted to those arising from J-variation of the assignment

JA(λ,J,)π

based on the explicit J-dependent formula of A(λ,J,)π , while Wendl used an abstract variation of Fredholm operators. Because of this our proof is much more nontrivial than that of [45, Lemma C.1] which strongly relies on the special form of J-dependence of the linearization A(λ,J,)π .

Lemma 5.7

Let i ≥ 0 and L ∈ End(ker A(λ,J,)π μi Id) that is symmetric with respect to the L2 inner product. Then there exists a smooth section

BTJJλ=Γ(γSλ)Γ(γEnd(ξ))

such that

c,BκL2=c,LκL2

for all κ, c ∈ ker ( A(λ,J,)π μi I).

Proof

We fix a basis 𝓑 := {e1, …, emi} of ker( A(λ,J,)π μi Id). Recall that each eigenfunction is a solution to a first-order linear ODE and hence it is a nowhere varnishing section of γ* ξ. Furthermore

{e1(t),,emi(t)}ξγ(t)=B(t)

is linearly independent at each tS1 and so spans an mi-dimensional subspace of ξγ(t). In particular, we must have mi ≤ 2n with dim Q = 2n + 1.

We extend the set 𝓑(t) to an orthogonal basis 𝓑͠(t) of ξγ(t) and define a linear map (t) on ξγ(t) so that

L~(t)(ek(t))=(Lek)(t) (5.10)

for all k = 1, …, mi. This (t) may not be symmetric. Since it satisfies 〈κ, L͠cL21 = 〈κ, L cL2 for all κ, c ∈ ker( A(λ,J,)π μi Id), we can replace by its symmetrization ( + ()T)/2 if needed.

Writing A = [dF(J)(B)]𝓑, we have

Aij=ei,dJF(B)ejL2=S1ei(t),(dJF(B)ej)(t)dt.

Lemma 5.8

For any eigenfunction η of A(λ,J,∇), we have

LRλη+12(LRλJ)η=μTη12J(LRλJ)η.

Proof

This is a direct consequence of (5.6). □

Therefore substituting this into (5.7), we have

dJF(B)ej=T2Id+JLRλBejTBLRλej+12(LRλJ)ej=T2Id+J(LRλB)+TBμTId+12J(LRλJ)ej.

We write the last operator in the big parentheses as

MLRλB+BN

where

M=T2(Id+J),N=TμTId+12JLRλJ

and note that M is an invertible map pointwise: the inverse of M is given by

M1=1T(IdJ). (5.11)

(Recall that the nondegeneracy of Reeb orbits implies that T ≠ 0.)

After pulling back the operator by γ, we can write the pull-back operator as

MtϕB+BN

and define the map tϕ by

tϕB:=LRλB~|γ(t)

where is a smooth extension of the map

γ(t)Bt

to a neighborhood of Image γ. By the definition of the Lie derivative, this definition does not depend on the choice of extension . It is straightforward to check that this is the induced connection of the one on the vector bundle γ* ξS1 defined, which we still denote by tϕ as follows. The following lemma is standard and so its proof is omitted.

Lemma 5.9

Let (γ, T) be the given isospeed Reeb orbit. Consider the pull-back vector bundle γ* ξS1. Define the operator tϕ : Γ(γ* ξ) → Γ(γ* ξ) by

tϕ(η):=TLRλY|γ(t)

for any (locally defined) vector field Y tangent to ξ defined near the point γ(t) such that Y(γ(t)) = η(t). Then the map is well-defined and defines a connection on the vector bundle γ* ξS1.

Therefore we can express it as

tϕ=t+Cϕ

where ∇t is the pull-back connection of the Hermitian connection on ξQ induced by the contact triad connection on the triad (Q, λ, J), and Cϕ is a zero order operator along γ. Then we can rewrite dJF(B) as

MtB+MCϕB+BN.

Furthermore we also have π ∇|ξ = πLC|ξ. (See [36, Section 6], more precisely see its arXiv version 1212.4817(v2).)

The following lemma is a crucial ingredient in our proof that enables us to use a priori much smaller set of variations arising from J-variations of A(λ,J,)π instead of the abstract variations used in [45, Section 3.2] especially in the proof of [45, Lemma C.1].

Lemma 5.10

The following initial valued problem can be uniquely solved:

MtLCB+MCϕB+BN=LB(0)=B0

for any given smooth section L of End(γ* ξ).

Proof

We first recall from (5.11) that the coefficient matrix M is invertible. Therefore the equation can be rewritten as

tLCB+CϕB+M1BN=M1L

which is an inhomogeneous linear first-order ODE for the variable B. By conjugating B by the parallel transport map Πγt : Tγ(0) QTγ(t) Q, we consider

B¯(t):=ΠγtB(t)(Πγt)1:Tγ(0)QTγ(0)Q.

Then the equation is further transformed to

dB¯dt+C¯ϕB¯+M¯1B¯N¯=M¯1L¯

where all ‘bar’ operators are conjugates by the parallel transport as for . This is the genuine first-order linear system of ODE valued in the vector space Tγ(0) Q.

Now we apply the standard argument of ‘variation of constants’ for solving the initial valued problem of the inhomogeneous first-order linear ODE for . By writing back B(t)=(Πγt)1B¯(t)Πγt, we have uniquely solved the required equation, which finishes the proof. □

Applying this lemma by setting L = , the operator given in (5.10), we have finished the proof of Lemma 5.7. □

This proves that the map J A(λ,J,)π is a submersion from 𝓙λ to the stratum Δk;γ1 (0) along

(Δk;γAk,γ)1(0)

for all k, which finishes the proof of Proposition 5.6. □

Now we have finished the proof of Proposition 5.4.

Funding statement: This work is supported by the IBS project # IBS-R003-D1. Both authors would like to also acknowledge MATRIX and the Simons Foundation for their support and funding through the MATRIX-Simons Collaborative Fund of the IBS-CGP and MATRIX workshop on Symplectic Topology.

Acknowledgements

We thank MATRIX for providing an excellent research environment where the present research was initiated. We also thank Hutchings for attracting our attention to Siefring’s paper [43] after we posted our survey paper [34] on the arXiv.

  1. Communicated by: K. Ono

References

[1] C. Abbas, The chord problem and a new method of filling by pseudoholomorphic curves. Int. Math. Res. Not. no. 18 (2004), 913–927. MR2037757 Zbl 1105.5306510.1155/S1073792804132236Search in Google Scholar

[2] P. Albers, B. Bramham, C. Wendl, On nonseparating contact hypersurfaces in symplectic 4-manifolds. Algebr. Geom. Topol. 10 (2010), 697–737. MR2606798 Zbl 1207.3201910.2140/agt.2010.10.697Search in Google Scholar

[3] M. F. Atiyah, V. K. Patodi, I. M. Singer, Spectral asymmetry and Riemannian geometry. III. Math. Proc. Cambridge Philos. Soc. 79 (1976), 71–99. MR397799 Zbl 0325.5801510.1017/S0305004100052105Search in Google Scholar

[4] A. Banyaga, Some properties of locally conformal symplectic structures. Comment. Math. Helv. 77 (2002), 383–398. MR1915047 Zbl 1020.5305010.1007/s00014-002-8345-zSearch in Google Scholar

[5] E. Bao, K. Honda, Semi-global Kuranishi charts and the definition of contact homology. Adv. Math. 414 (2023), Paper No. 108864, 148 pages. MR4539062 Zbl 1510.5309110.1016/j.aim.2023.108864Search in Google Scholar

[6] D. E. Blair, Riemannian geometry of contact and symplectic manifolds. Birkhäuser, Boston 2010. MR2682326 Zbl 1246.5300110.1007/978-0-8176-4959-3Search in Google Scholar

[7] D. J. Cant, A dimension formula for relative symplectic field theory. PhD Thesis, Stanford University, 2022.Search in Google Scholar

[8] S. S. Chern, Complex manifolds without potential theory. Van Nostrand, Princeton, %N.J.-Toronto, Ont.-London 1967. MR225346 Zbl 0158.33002Search in Google Scholar

[9] P. Fendley, M. P. A. Fisher, C. Nayak, Topological entanglement entropy from the holographic partition function. J. Stat. Phys. 126 (2007), 1111–1144. MR2312943 Zbl 1119.8201110.1007/s10955-006-9275-8Search in Google Scholar

[10] A. Floer, The unregularized gradient flow of the symplectic action. Comm. Pure Appl. Math. 41 (1988), 775–813. MR948771 Zbl 0633.5305810.1002/cpa.3160410603Search in Google Scholar

[11] P. Gauduchon, Hermitian connections and Dirac operators. Boll. Un. Mat. Ital. B (7) 11 (1997), 257–288. MR1456265 Zbl 0876.53015Search in Google Scholar

[12] I. M. Gel’fand, M. M. Kapranov, A. V. Zelevinsky, Discriminants, resultants, and multidimensional determinants. Birkhäuser 1994. MR1264417 Zbl 0827.1403610.1007/978-0-8176-4771-1Search in Google Scholar

[13] H. Hofer, Pseudoholomorphic curves in symplectizations with applications to the Weinstein conjecture in dimension three. Invent. Math. 114 (1993), 515–563. MR1244912 Zbl 0797.5802310.1007/BF01232679Search in Google Scholar

[14] H. Hofer, Holomorphic curves and real three-dimensional dynamics. Geom. Funct. Anal. (2000), Special Volume, Part II, 674–704. MR1826268 Zbl 1161.5336210.1007/978-3-0346-0425-3_5Search in Google Scholar

[15] H. Hofer, K. Wysocki, E. Zehnder, Properties of pseudo-holomorphic curves in symplectisations. II. Embedding controls and algebraic invariants. Geom. Funct. Anal. 5 (1995), 270–328. MR1334869 Zbl 0845.5702710.1007/BF01895669Search in Google Scholar

[16] H. Hofer, K. Wysocki, E. Zehnder, Properties of pseudoholomorphic curves in symplectisations. I. Asymptotics. Ann. Inst. H. Poincaré C Anal. Non Linéaire 13 (1996), 337–379. MR1395676 Zbl 0861.5801810.1016/s0294-1449(16)30108-1Search in Google Scholar

[17] E.-N. Ionel, T. H. Parker, Relative Gromov–Witten invariants. Ann. of Math. (2) 157 (2003), 45–96. MR1954264 Zbl 1039.5310110.4007/annals.2003.157.45Search in Google Scholar

[18] T. Kato, Perturbation theory for linear operators. Springer 1995. MR1335452 Zbl 0836.4700910.1007/978-3-642-66282-9Search in Google Scholar

[19] T. Kim, Y.-G. Oh, Perturbation theory of asymptotic operators of contact instantons and of pseudoholomorphic curves on symplectization. Preprint 2014, arXiv:2303.01011Search in Google Scholar

[20] J. Kim, Y.-G. Oh, Rational contact instantons and Legendrian Fukaya category. In preparation.Search in Google Scholar

[21] T. Kim, Y.-G. Oh, Kuranishi structures on contact instanton moduli spaces. In preparation.Search in Google Scholar

[22] S. Kobayashi, Natural connections in almost complex manifolds. In: Explorations in complex and Riemannian geometry, volume 332 of Contemp. Math., 153–169, Amer. Math. Soc. 2003. MR2018338 Zbl 1052.53032)10.1090/conm/332/05935Search in Google Scholar

[23] S. Kobayashi, K. Nomizu, Foundations of differential geometry. Vol. II. Wiley-Interscience 1996. MR1393941 Zbl 0175.48504Search in Google Scholar

[24] U. Koschorke, Infinite dimensional K-theory and characteristic classes of Fredholm bundle maps. In: Global Analysis (Proc. Sympos. Pure Math., Vol. XV, Berkeley, Calif., 1968), 95–133, Amer. Math. Soc. 1970. MR279838 Zbl 0207.5360210.1090/pspum/015/0279838Search in Google Scholar

[25] R. B. Lockhart, R. C. McOwen, Elliptic differential operators on noncompact manifolds. Ann. Scuola Norm. Sup. Pisa Cl. Sci. (4) 12 (1985), 409–447. MR837256 Zbl 0615.58048Search in Google Scholar

[26] Y.-G. Oh, Symplectic topology and Floer homology. Vol. 1. New Mathematical Monographs, vol. 28. Cambridge Univ. Press 2015. MR3443239 Zbl 1338.53004Search in Google Scholar

[27] Y.-G. Oh, Analysis of contact Cauchy–Riemann maps III: Energy, bubbling and Fredholm theory. Bull. Math. Sci. 13 (2023), Paper No. 2250011, 61 pages. MR4581187 Zbl 1518.53066Search in Google Scholar

[28] Y.-G. Oh, Bordered contact instantons and their Fredholm theory and generic transversalities. Preprint 2023 (to appear in the proceedings of Bumsig Kim’s Memorial Conference, October 2021, KIAS), arXiv:2209.03548Search in Google Scholar

[29] Y.-G. Oh, Contact Hamiltonian dynamics and perturbed contact instantons with Legendrian boundary condition. Preprint 2021, arXiv:2103.15390Search in Google Scholar

[30] Y.-G. Oh, Contact instantons, anti-contact involution and proof of Shelukhin’s conjecture. Preprint 2024, arXiv:2212.03557Search in Google Scholar

[31] Y.-G. Oh, Geometric analysis of perturbed contact instantons with Legendrian boundary conditions. Preprint 2022, arXiv:2205.12351Search in Google Scholar

[32] Y.-G. Oh, Geometry and analysis of contact instantons and entangement of Legendrian links I. Preprint 2024, arXiv:2111.02597Search in Google Scholar

[33] Y.-G. Oh, Gluing theories of contact instantons and of pseudoholomorphic curves in SFT. Preprint 2022, arXiv:2205.00370Search in Google Scholar

[34] Y.-G. Oh, T. Kim, Analysis of Pseudoholomorphic Curves on Symplectization: Revisit via Contact Instantons. In: 2021–2022 MATRIX annals, volume 5 of MATRIX Book Ser., 635–732, Springer 2024. MR4807297 Zbl 0793663410.1007/978-3-031-47417-0_32Search in Google Scholar

[35] Y.-G. Oh, Y. Savelyev, Pseudoholomorphic curves on the 𝔏ℭ𝔖-fication of contact manifolds. Adv. Geom. 23 (2023), 153–190. MR4596225 Zbl 1520.5306510.1515/advgeom-2023-0004Search in Google Scholar

[36] Y.-G. Oh, R. Wang, Canonical connection on contact manifolds. In: Real and complex submanifolds, volume 106 of Springer Proc. Math. Stat., 43–63, Springer 2014. MR3333367 Zbl 1323.5308610.1007/978-4-431-55215-4_5Search in Google Scholar

[37] Y.-G. Oh, R. Wang, Analysis of contact Cauchy–Riemann maps I: a priori Ck estimates and asymptotic convergence. Osaka J. Math. 55 (2018), 647–679. MR3862779 Zbl 1454.53074Search in Google Scholar

[38] Y.-G. Oh, R. Wang, Analysis of contact Cauchy–Riemann maps II: Canonical neighborhoods and exponential convergence for the Morse–Bott case. Nagoya Math. J. 231 (2018), 128–223. MR3845592 Zbl 1410.5308010.1017/nmj.2017.17Search in Google Scholar

[39] Y.-G. Oh, S. Yu, Contact instantons with Legendrian boundary condition: a priori estimates, asymptotic convergence and index formula. Internat. J. Math. 35 (2024), Paper No. 2450019, 60 pages. MR4756719 Zbl 1543.53081Search in Google Scholar

[40] Y.-G. Oh, S. Yu, Legendrian spectral invariants on the one jet bundle via perturbed contact instantons. Preprint 2023, arXiv:2301.06704Search in Google Scholar

[41] J. Pardon, Contact homology and virtual fundamental cycles. J. Amer. Math. Soc. 32 (2019), 825–919. MR3981989 Zbl 1422.5307110.1090/jams/924Search in Google Scholar

[42] J. W. Robbin, D. A. Salamon, Asymptotic behaviour of holomorphic strips. Ann. Inst. H. Poincaré C Anal. Non Linéaire 18 (2001), 573–612. MR1849689 Zbl 0999.5304810.1016/s0294-1449(00)00066-4Search in Google Scholar

[43] R. Siefring, Relative asymptotic behavior of pseudoholomorphic half-cylinders. Comm. Pure Appl. Math. 61 (2008), 1631–1684. MR2456182 Zbl 1158.5306810.1002/cpa.20224Search in Google Scholar

[44] R. Siefring, Intersection theory of punctured pseudoholomorphic curves. Geom. Topol. 15 (2011), 2351–2457. MR2862160 Zbl 1246.3202810.2140/gt.2011.15.2351Search in Google Scholar

[45] C. Wendl, Lectures on Symplectic Field Theory. Book manuscript. Preprint 2016, arXiv:1612.01009Search in Google Scholar

A Appendix: Contact triad connection and canonical connection

Let (Q, ξ) be a given contact manifold. When a contact form λ is given, we have the projection π = πλ from TQ to ξ associated to the decomposition

TQ=ξRRλ.

We denote by Π = Πλ : TQTQ the corresponding idempotent, i.e. the endomorphism of TQ satisfying Π2 = Π, Im Π = ξ and ker Π = ℝ〈Rλ〉.

A.1 Contact triads and triad connections

Let (Q, λ, J) be a contact triad of dimension 2n + 1 for the contact manifold (Q, ξ), and equip it with the contact triad metric g = gξ + λλ. In [36], Wang and the second-named author introduced the contact triad connection associated to every contact triad (Q, λ, J) with the contact triad metric and proved its existence, uniqueness and naturality.

Theorem A.1

(Contact Triad Connection [36]). For every contact triad (Q, λ, J), there exists a unique affine connection ∇, called the contact triad connection, satisfying the following properties:

  1. The connectionis metric with respect to the contact triad metric, i.e. ∇ g = 0;

  2. The torsion tensor T ofsatisfies T(Rλ, ⋅) = 0;

  3. The covariant derivatives satisfyRλ Rλ = 0, andY Rλξ for any Yξ;

  4. The projectionπ := π ∇|ξ defines a Hermitian connection of the vector bundle ξM with Hermitian structure (|ξ, J);

  5. The ξ-projection of the torsion T, denoted by Tπ := π T, satisfies the following property:

    Tπ(JY,Y)=0 (A.1)

    for all Y tangent to ξ;

  6. For Yξ we have the following

    YRλ:=12(YRλJJYRλ)=0.

From this theorem, we see that the contact triad connection ∇ canonically induces a Hermitian connection ∇π for the Hermitian vector bundle (ξ, J, gξ), and we call it the contact Hermitian connection. This connection is used for the study of various a priori estimates of contact instantons starting from [37].

Corollary A.2

Letbe the contact triad connection. Then

  1. For any vector field Y on Q,

    YRλ=12(LRλJ)JY; (A.2)
  2. λ(T) = dλ as a two-form on Q.

We refer readers to [36] for more discussion on the contact triad connection and its relation with other related canonical type connections.

A.2 Almost Hermitian manifolds and canonical connections

Next we give the definition of canonical connection on general almost Hermitian manifolds and apply it to the case of 𝔩𝔠𝔰-fication of contact triad (Q, λ, J). See [26, Chapter 7] for the exposition of canonical connection for almost Hermitian manifolds and its relationship with the Levi–Civita connection, and in relation to the study of pseudoholomorphic curves on general symplectic manifolds emphasizing the Weitzenböck formulae in the study of elliptic regularity in the same spirit of the present survey.

Let (M, J) be any almost complex manifold.

Definition A.3

A metric g on (M, J) is called Hermitian, if g satisfies

g(Ju,Jv)=g(u,v),u,vTxM,xM.

We call the triple (M, J, g) an almost Hermitian manifold.

For any given almost Hermitian manifold (M, J, g), the bilinear form

Φ:=g(J,)

is called the fundamental two-form in [23], which is nondegenerate.

Definition A.4

An almost Hermitian manifold (M, J, g) is an almost Kähler manifold if the two-form Φ above is closed.

Definition A.5

A (almost) Hermitian connection∇ is an affine connection satisfying

g=0=J.

Existence of such a connection is easy to check. In general the torsion T = T of the almost Hermitian connection ∇ is not zero, even when J is integrable. The following is the almost complex version of the Chern connection in complex geometry.

Theorem A.6

([11], [22]). On any almost Hermitian manifold (M, J, g), there exists a unique Hermitian connectionon TM satisfying

T(X,JX)=0 (A.3)

for all XTM.

In complex geometry [8] where J is integrable, a Hermitian connection satisfying (A.3) is called the Chern connection.

Definition A.7

A canonical connection of an almost Hermitian connection is defined to be one that has the torsion property (A.3).

The triple

(Q×R,J~,g~λ)

is a natural example of an almost Hermitian manifold associated to the contact triad (Q, λ, J).

Let ∇͠ be the canonical connection thereof. Then we have the following which also provides a natural relationship between the contact triad connection and the canonical connection.

Proposition A.8

(Canonical connection versus contact triad connection). Let g͠ = λ be the almost Hermitian metric given above. Let ∇͠ be the canonical connection of the almost Hermitian manifold (1.9), and letbe the contact triad connection for the triad (Q, λ, J). Then ∇͠ preserves the splitting (1.4) and satisfies ∇͠|ξ = ∇|ξ.

B Appendix: Covariant differential of vector-valued forms

In this appendix, we recall the standard exterior calculus of vector-valued forms borrowing from the exposition in [37, Appendix].

Assume that (P, h) is a Riemannian manifold of dimension n with metric h, and that D is the Levi–Civita connection. Let EP be any vector bundle with inner product 〈 ⋅ , ⋅ 〉, and assume that ∇ is a connection on E which is compatible with 〈 ⋅ , ⋅ 〉.

Denote by Ωk(E) the space of E-valued k-forms on P. The connection ∇ induces an exterior derivative by

d:Ωk(E)Ωk+1(E)d(αζ)=dαζ+(1)kαζ.

It is not hard to check that for any one-form β, equivalently one can write

dβ(v1,v2)=(v1β)(v2)(v2β)(v1),

where v1, v2TP.

For the purpose of the present paper, we mostly apply the last formula to the w* TQ or w* ξ one-forms. Some illustrations thereof are now in order.

For any given map w : Σ̇Q, not necessarily arising from the symplectization, we can decompose its derivative dw, regarded as a w* TQ-valued one-form on Σ̇, into

dw=dπw+wλRλ (B.1)

where dπ w := π dw. Furthermore dπ w is decomposed into

dπw=¯πw+πw (B.2)

where π w := (dwπ)J(0,1) (respectively π w := (dwπ)J(1,0) ) is the anti-complex linear part (respectively the complex linear part) of dπ w : (TΣ̇, j) → (ξ, J|ξ). (For the simplicity of notation, we will abuse our notation by often denoting J|ξ by J. We also simply write (()π)J(0,1)=()π(0,1)and(()π)J(1,0)=()π(1,0) in general.)

Here are some examples of vector-valued forms that appear in the main text of the present paper:

  1. A vector field Y (respectively Yπ) along the map w is a w* TQ-valued (respectively a w* ξ-valued) zero-form.

  2. The covariant derivative ∇ Y is a w* TQ-valued one-form.

  3. π Yπ is a w* ξ-valued one-form.

  4. We regard dw (respectively dπ w) as a w* TQ-valued (respectively w* ξ-valued) one-form.

C Appendix: Subsequence convergence and charge vanishing

In this section, we recall subsequence and charge vanishing results on contact instantons from [37] and [39]. We put the following hypotheses in our asymptotic study of the finite energy contact instanton maps w as in [37]:

Hypothesis C.1

Let h be the metric on Σ̇ given above. Assume that w : Σ̇M satisfies the contact instanton equations (1.8), and

  1. E(λ,J;Σ˙,h)π(w) < ∞ (finite π-energy);

  2. dwC0(Σ̇) < ∞.

  3. Image w ⊂ K ⊂ M for some compact subset K.

The above finite π-energy and C0 bound hypotheses imply that

[0,)×[0,1]|dπw|2dτdt<,dwC0([0,)×[0,1])< (C.1)

in these coordinates.

Definition C.2

(Asymptotic action and charge). Assume that the limit of w(τ, ⋅) as τ → ∞ exists. Then we can associate two natural asymptotic invariants at each puncture defined as

T:=limr[0,1]wrλ (C.2)
Q:=limr[0,1]wr(λj). (C.3)

where wr : [0, 1] → Q is the map defined by wr(t) := w(r, t). (Here we only look at positive punctures. The case of negative punctures is similar.) We call T the asymptotic contact action and Q the asymptotic contact charge of the contact instanton w at the given puncture.

The following open string version of subsequence convergence and charge vanishing result is proved in [29] and [39].

Theorem C.3

Let w : [0, ∞) × [0, 1] → M satisfy the contact instanton equations (1.8) and converge as |τ| → ∞. Then for any sequence sk → ∞, there exists a subsequence, still denoted by sk, and a massless instanton w(τ, t) (i.e. Eπ(w) = 0) on the strip ℝ × [0, 1] that satisfies the following:

  1. π w = 0 and

    limkw(sk+τ,t)=w(τ,t)

    in the Cl(K × [0, 1], M) sense for any l, where K ⊂ [0, ∞) is an arbitrary compact set.

  2. w has vanishing asymptotic charge Q = 0 and satisfies w(τ, t) = γ(t) for some Reeb chord γ joining R0 and R1 with period T at each puncture.

  3. T ≠ 0 at each puncture with the associated pair (R, R′) of boundary conditions with RR′ = ∅.

Corollary C.4

(Corollary 5.11 in [39]). Assume that the pair (λ, R⃗) is nondegenerate in the sense of Definition 2.8. Let w : Σ̇M satisfy the contact instanton equation (1.8) and Hypothesis (C.1). Then on each strip-like end with strip-like coordinates (τ, t) ∈ [0, ∞) × [0, 1] near a puncture

limsπwτ(s+τ,t)=0,limsπwt(s+τ,t)=0limsλ(wτ)(s+τ,t)=0,limsλ(wt)(s+τ,t)=T

and

lims|ldw(s+τ,t)|=0foranyl1.

All the limits are uniform for (τ, t) in K × [0, 1] with compact K ⊂ ℝ.

We also state that the same holds for the closed string case for which the charge Q may not vanish. The proof of the following subsequence convergence result is proved in [37, Theorem 6.4].

Theorem C.5

(Subsequence Convergence, Theorem 6.4 in [37]). Let w : [0, ∞) × S1M satisfy the contact instanton equation (1.8) and Hypothesis (C.1). Then for any sequence sk → ∞, there exists a subsequence, still denoted by sk, and a massless instanton w(τ, t) (i.e. Eπ(w) = 0) on the cylinder ℝ × [0, 1] that satisfies the following:

  1. π w = 0 and

    limkw(sk+τ,t)=w(τ,t)

    in the Cl(K × [0, 1], M) sense for any l, where K ⊂ [0, ∞) is an arbitrary compact set.

  2. wλ=Qdτ+Tdt.

We have the same exponential convergence as Corollary C.4 for the closed strong case, provided Q = 0.

Corollary C.6

Assume that λ is nondegenerate. Suppose that wτ converges as |τ| → ∞ and its massless limit instanton has Q = 0 but T ≠ 0, then the wτ converges to a Reeb orbit of period |T| exponentially fast.

D Appendix: Off-shell setting of the linearization operator

We now provide some details of the Fredholm theory and the index calculation referring readers to the original articles [27] and [35] for complete details.

We fix an elongation function ρ : ℝ → [0, 1] so that 0 ≤ ρ′(τ) ≤ 2 and

ρ(τ)=1τ10τ0.

Then we consider sections of w* TQ by

Y¯i=ρ(τR0)Rλ(γi+(t)),Y_j=ρ(τ+R0)Rλ(γj+(t)) (D.1)

and denote by Γs+,sΓ(w* TQ) the subspace defined by

Γs+,s=i=1s+R{Y¯i}j=1sR{Y_j}.

Let k ≥ 2 and p > 2. The local model of the tangent space of Wδk,p (Σ̇, Q; J; γ+, γ) at

wCδ(Σ˙,Q)Wδk,p(Σ˙,Q)

is given by

Γs+,sWδk,p(wTQ) (D.2)

where Wδk,p (w* TQ) is the Banach space

{Y=(Yπ,λ(Y)Rλ)eδp|τ|YπWk,p(Σ˙,wξ),λ(Y)Wk,p(Σ˙,R)}Wk,p(Σ˙,R)Rλ(w)Wk,p(Σ˙,wξ).

Here we measure the various norms in terms of the triad metric of the triad (Q, λ, J).

We choose δ > 0 so that 0 < δ/p < 1 is smaller than the spectral gap

gap(γ+,γ):=mini,j{dH(specA(Ti,zi),0),dH(specA(Tj,zj),0)}. (D.3)

We denote by

Wδk,p(Σ˙,Q;J;γ+,γ),k2

the Banach manifold such that

limτw((τ,t)i)=γi+(Ti(t+ti)),limτw((τ,t)j)=γj(Tj(ttj))

for some elements ti, tjS1, where (τ, t)i and (τ, t)j are the cylindrical coordinates at the punctures i, j respectively and

Ti=S1(γi+)λ,Tj=S1(γj)λ.

Here ti (respectively tj) depends on the given analytic coordinate (τ, t)i (respectively (τ, t)j) and the parametrization of the Reeb orbits.

Now for each given w Wδk,p := Wδk,p (Σ̇, Q; J; γ+, γ), we consider the Banach space

Ωk1,p;δ(0,1)(wξ)

the Wδk1,p -completion of Ω(0,1)(w* ξ) and form the bundle

Hk1,p;δ(0,1)(ξ)=wWδk,pΩk1,p;δ(0,1)(wξ)

over Wδk,p . Then we can regard the assignment

Υ1:(w,f)¯πw

as a smooth section of the bundle Hk1,p;δ(0,1)(ξ)Wδk,p.

Similarly we consider

Ωk1,p;δ(0,1)(uV)

the Wδk1,p -completion of Ω(0,1)(u* 𝓥) and form the bundle

Hk1,p;δ(0,1)(V)=wWδk,pΩk1,p;δ(0,1)(uV)

over Wδk,p . Then the assignment

Υ2:u=(w,f)wλjfds

defines a smooth section of the bundle

Hk1,p;δ(0,1)(V)Wδk,p.

We have already computed the linearization of each of these maps in the previous section.

With these preparations, the following is a corollary of exponential estimates established in [37].

Proposition D.1

(Corollary 6.5 in [37]). Assume that λ is nondegenerate. Let w : Σ̇Q be a contact instanton and let w* λ = a1 + a2 dt. Suppose that

limτa1,i=Q(pi),limτa2,i=T(pi)limτa1,j=Q(pj),limτa2,j=T(pj) (D.4)

at each puncture pi and qj. Then w Wδk,p (Σ̇, Q; J; γ+, γ).

Now we are ready to describe the moduli space of pseudoholomorphic curves on symplectization with prescribed asymptotic condition as the zero set

M(Σ˙,Q;J;γ+,γ)=Wδk,p(Σ˙,Q;J;γ+,γ)Wδk,p(Σ˙,R)Υ1(0) (D.5)

whose definition does not depend on the choice of k, p or δ as long as k ≥ 2, p > 2 and δ > 0 is sufficiently small. One can also vary λ and J and define the universal moduli space whose detailed discussion is postponed. We see therefrom that (λ,J) is the first-order differential operator whose first-order part is given by the direct sum operator

(Yπ,(κ,b))¯πYπ(dκjdb)

where we write (Y, v) = (Yπ + κ Rλ, b s ) for κ = λ(Y), b = ds(v). Here we have

¯π:Ωk,p;δ0(wξ;J;γ+,γ)Ωk1,p;δ(0,1)(wξ)

and the second summand can be written as the standard Cauchy–Riemann operator

¯:Wk,p(Σ˙;C)Ωk1,p(0,1)(Σ˙,C);b+iκ=:φ¯φ.

The following proposition can be derived from the arguments used by Lockhart and McOwen [25].

Proposition D.2

Suppose that δ > 0 satisfies the inequality

0<δ<mingap(γ+,γ)p,2p

where gap(γ+, γ) is the spectral gap, given in (D.3), of the asymptotic operators A(Tj,zj) or A(Ti,zi) associated to the corresponding punctures. Then the operator (3.5) is Fredholm.

Received: 2023-08-21
Revised: 2024-07-09
Published Online: 2025-01-23
Published in Print: 2025-01-29

© 2025 Walter de Gruyter GmbH, Berlin/Boston

This work is licensed under the Creative Commons Attribution 4.0 International License.

Downloaded on 8.10.2025 from https://www.degruyterbrill.com/document/doi/10.1515/advgeom-2024-0033/html
Scroll to top button