This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

A Lower Bound of the Hofer-Zehnder Capacity via Delzant Polytopes

Yichen Liu Mathematics Department
University of Illinois at Urbana-Champaign
Champaign, Illinois 61801
yichen23@illinois.edu
Abstract.

Given a symplectic toric manifold, the moment maps of sub-circle actions can be modified to be admissible functions in the sense of Hofer-Zehnder. By exploiting the relationship between the period of Hamiltonian sub-circle actions of a symplectic toric manifold and its Delzant polytope, we develop an invariant of Delzant polytopes which gives a lower bound of the Hofer-Zehnder capacity.

2020 Mathematics Subject Classification:
Primary 53D20, 53D35

1. Introduction

In 1980s, Gromov [8] proved the celebrated non-squeezing theorem which states that a ball B2n(r)B^{2n}(r) of radius rr can be symplectically embedded into a cylinder B2(R)×2n2B^{2}(R)\times{\mathbb{R}}^{2n-2} of radius RR if and only if rRr\leq R. This led to the notion of the Gromov width of a symplectic manifold (M,ω)(M,\omega), which is the radius of the largest ball of the same dimension that can be symplectically embedded into MM. The Gromov width is the first example of a symplectic capacity. In [6], Ekeland and Hofer defined symplectic capacities in an axiomatic way and constructed many concrete examples of symplectic capacities. Since then, symplectic capacities play an important role in the study of the symplectic geometry.

Among symplectic capacities, the Hofer-Zehnder capacity, introduced in [10], is based on properties of periodic orbits of Hamiltonian flows on (M,ω)(M,\omega). For a smooth function HH on MM, let XHX_{H} denote its Hamiltonian vector field. For a compact symplectic manifold (M,ω)(M,\omega), we say H:MH:M\to{\mathbb{R}} is admissible if

  • there exist open sets U,VU,V such that H|U=HmaxH|_{U}=H_{\max} and H|V=HminH|_{V}=H_{\min}, and

  • XHX_{H} has no non-constant periodic orbits of period less than 11.

We denote the set of admissible functions by ad(M,ω)={HC(M)\mathcal{H}_{ad}(M,\omega)=\{H\in C^{\infty}(M) HH is admissible}\}. The Hofer-Zehnder capacity of (M,ω)(M,\omega) is defined by

cHZ(M,ω)=sup{HmaxHmin|Had(M,ω)}.c_{\textrm{HZ}}(M,\omega)=\textrm{sup}\{H_{\max}-H_{\min}\big{|}\,H\in\mathcal{H}_{ad}(M,\omega)\}.

Recall that we say a circle S1=/S^{1}={\mathbb{R}}/{\mathbb{Z}} acts on a symplectic manifold (M,ω)(M,\omega) in a Hamiltonian fashion if there exists a map H:MH:M\to{\mathbb{R}} such that dH=ω(X,)dH=\omega(X,\cdot), where XX is the fundamental vector field of the S1S^{1}-action. Such an HH is called a moment map. Moment maps can be modified to admissible functions (see Lemma 3.4). Hence, by studying S1S^{1}-actions on a symplectic manifold, we can estimate the Hofer-Zehnder capacity using the information from the moment maps. Moreover, with the presence of group actions, one can also study equivariant capacities introduced in [7]. The equivariant technique used in this paper is similar to those in [5], [14], where the authors obtained bounds for equivariant capacities. In fact, the authors computed packing densities, which easily determines the equivariant capacities according to [7, Equation (2)]. For more discussion about equivariant capacities, see [5], [7], [14],[15],[16] and the references therein.

A symplectic toric manifold (M2n,ω,Tn,Φ)(M^{2n},\omega,T^{n},\Phi) is a connected, closed 2n2n-dimensional symplectic manifold equipped with an effective, Hamiltonian nn-torus action (see Definition 2.1). Delzant [4] classified symplectic toric manifolds by their moment polytopes which are Delzant polytopes (see Definition 2.2). These manifolds admit many different S1S^{1}-actions and the information of actions is encoded by simple combinatorial data, so it is possible to get a reasonably good estimate of the Hofer-Zehnder capacity from the combinatorial data. The main goal of this paper is to describe an explicit algorithm for how to find a lower bound of the Hofer-Zehnder capacity by simply using the information obtained from the Delzant polytope.

In [13], Lu also gave a lower bound of the Hofer-Zehnder capacity of a toric manifold from the combinatorial data of the Delzant polytope. Lu’s technique is to find symplectic embeddings of balls whose radius is related to the combinatorial data and use the fact that the Hofer-Zehnder capacity is always bounded from below by the Gromov width, while our proof depends on a careful analysis of how to read the maximal (in terms of the cardinality) stabilizer group of non-fixed points from the combinatorial data.

Let (M2n,ω,Tn,Φ)(M^{2n},\omega,T^{n},\Phi) be a symplectic toric manifold with the Delzant polytope Δ=Φ(M)\Delta=\Phi(M). We can write Δ\Delta as an intersection of half-spaces

Δ=i=1d{xn|x,viλi},\Delta=\bigcap_{i=1}^{d}\{x\in\mathbb{R}^{n}\big{|}\,\langle x,v_{i}\rangle\leq\lambda_{i}\},

where vinv_{i}\in{\mathbb{Z}}^{n} are primitive outward-pointing normal vectors. (Recall that a vector vnv\in{\mathbb{Z}}^{n} is primitive if for any unu\in{\mathbb{Z}}^{n}, the equation v=nuv=nu for nn\in{\mathbb{Z}} implies that n=±1n=\pm 1.)

Let EE denote an edge of Δ\Delta. Delzant’s classification [4] implies that there exists a unique JE{1,2,.,d}J_{E}\subset\{1,2,....,d\} with |JE|=n1|J_{E}|=n-1 such that

E=ΔjJE{xn|x,vj=λj}.E=\Delta\cap\bigcap_{j\in J_{E}}\{x\in\mathbb{R}^{n}\big{|}\,\langle x,v_{j}\rangle=\lambda_{j}\}.

Let unu\in{\mathbb{Z}}^{n} be a primitive vector. To each edge EE of Δ\Delta, we associate a lattice polytope PEu={tujJEtjvj| 0t,tj1}P^{u}_{E}=\{tu-\sum_{j\in J_{E}}t_{j}v_{j}\big{|}\,0\leq t,t_{j}\leq 1\}, i.e. PEuP^{u}_{E} is the convex hull of {vj:jJE}{u}\{-v_{j}:j\in J_{E}\}\cup\{u\}. Let kEu:=|nint(PEu)|+1k^{u}_{E}:=|{\mathbb{Z}}^{n}\cap\textrm{int}(P^{u}_{E})|+1 denote the number of interior lattice points plus one. Notice that if uu is a linear combination of vjv_{j}, then kEu=1k^{u}_{E}=1.

Define

mu(Δ):=maxE{kEu|E is an edge of Δ},m_{u}(\Delta):=\max_{E}\{k^{u}_{E}\big{|}\,E\textrm{ is an edge of }\Delta\},
𝒯u(Δ):=1mu(Δ)(maxxΔx,uminxΔx,u).\mathcal{T}_{u}(\Delta):=\frac{1}{m_{u}(\Delta)}(\smash{\max_{x\in\Delta}}\langle x,u\rangle-\smash{\min_{x\in\Delta}}\langle x,u\rangle).

The toric width of a Delzant polytope Δn\Delta\subset{\mathbb{R}}^{n} is

wT(Δ):=supu{𝒯u(Δ)|un is a primitive vector}.w_{T}(\Delta):=\sup_{u}\{\mathcal{T}_{u}(\Delta)\big{|}\,u\in{\mathbb{Z}}^{n}\textrm{ is a primitive vector}\}.

Our main result states that for a symplectic toric manifold, the toric width of its Delzant polytope gives a lower bound of the Hofer-Zehnder capacity.

Theorem 1.1.

Let (M2n,ω,Tn,Φ)(M^{2n},\omega,T^{n},\Phi) be a symplectic toric manifold with a Delzant polytope Δ=Φ(M)\Delta=\Phi(M). Let wT(Δ)w_{T}(\Delta) be the toric width of Δ\Delta, then

cHZ(M,ω)wT(Δ).c_{\textrm{HZ}}(M,\omega)\geq w_{T}(\Delta).
Proof.

For any primitive vector unu\in{\mathbb{Z}}^{n}, let Su1TnS^{1}_{u}\subset T^{n} be the circle whose Lie algebra is u{\mathbb{R}}u. Su1S^{1}_{u} acts on M2nM^{2n} in a Hamiltonian fashion with the moment map μ=Φ,u\mu=\langle\Phi,u\rangle. By Corollary 2.6,

mu(Δ)=max{|stabp|:pMMSu1}m_{u}(\Delta)=\max\{\,|\textrm{stab}_{p}|:\,p\in M\setminus M^{S^{1}_{u}}\}

Since uu is primitive, this action is effective. By Lemma 3.4, there exists an admissible function HH such that

HmaxHmin=1mu(μmaxμmin)=𝒯u.H_{\max}-H_{\min}=\frac{1}{m_{u}}(\mu_{\max}-\mu_{\min})=\mathcal{T}_{u}.

By definition, cHZ(M,ω)wT(Δ)c_{\textrm{HZ}}(M,\omega)\geq w_{T}(\Delta). ∎

Recall that two toric actions ρ1,ρ2:Tn×M2nM2n\rho_{1},\rho_{2}:T^{n}\times M^{2n}\to M^{2n} are equivalent if there exist a symplectomorphism Ψ:MM\Psi:M\to M and an automorphism h:TnTnh:T^{n}\to T^{n} such that the following diagram commutes:

Tn×M2n{T^{n}\times M^{2n}}M2n{M^{2n}}Tn×M2n{T^{n}\times M^{2n}}M2n{M^{2n}}ρ1\scriptstyle{\rho_{1}}(h,Ψ)\scriptstyle{(h,\Psi)}f\scriptstyle{f}ρ2\scriptstyle{\rho_{2}}

Karshon, Kessler and Pinsonnault [12] proved that two toric actions are equivalent if and only if their Delzant polytopes differ by an affine transformation. Indeed, the toric width is invariant under the affine transformations, so equivalent toric actions will give the same bound in Theorem 1.1.

Lemma 1.2.

The toric width of a Delzant polytope is invariant under the action of the group of affine transformations Aff(n,)=GL(n,)n\textrm{Aff}(n,{\mathbb{R}})=\textrm{GL}(n,{\mathbb{Z}})\ltimes{\mathbb{R}}^{n}.

Proof.

For any matrix MGL(n,)M\in\textrm{GL}(n,{\mathbb{Z}}), let Δ=MΔ\Delta^{\prime}=M\Delta be the image of Δ\Delta under MM. For any edge EΔE\subset\Delta, let EΔE^{\prime}\subset\Delta^{\prime} be the corresponding edge.

Notice for any yΔy\in\Delta^{\prime}, M1y,viλi\langle M^{-1}y,v_{i}\rangle\leq\lambda_{i}, i.e. y,(M1)Tviλi\langle y,(M^{-1})^{T}v_{i}\rangle\leq\lambda_{i}. Hence, for any primitive vector unu\in{\mathbb{Z}}^{n} and u=MTuu^{\prime}=M^{T}u,

PEu={tujJEtj(M1)Tvj| 0t,tj1}=(M1)TPEu.P^{u}_{E^{\prime}}=\{tu-\sum_{j\in J_{E}}t_{j}(M^{-1})^{T}v_{j}\big{|}\,0\leq t,t_{j}\leq 1\}=(M^{-1})^{T}P^{u^{\prime}}_{E}.

Since MM preserves the lattice n{\mathbb{Z}}^{n},

kEu=|nint(PEu)|+1=|(M1)T(nint(PEu))|+1=|nint(PEu)|+1=kEu.k_{E^{\prime}}^{u}=|{\mathbb{Z}}^{n}\cap\textrm{int}(P^{u}_{E^{\prime}})|+1=|(M^{-1})^{T}({\mathbb{Z}}^{n}\cap\textrm{int}(P^{u^{\prime}}_{E}))|+1=|{\mathbb{Z}}^{n}\cap\textrm{int}(P^{u^{\prime}}_{E})|+1=k_{E}^{u^{\prime}}.

Therefore, mu(Δ)=mu(Δ)m_{u}(\Delta^{\prime})=m_{u^{\prime}}(\Delta).

Furthermore, since for any yΔy\in\Delta^{\prime}, there exists a unique xΔx\in\Delta such that y=Mxy=Mx and y,u=Mx,u=x,MTu=x,u\langle y,u\rangle=\langle Mx,u\rangle=\langle x,M^{T}u\rangle=\langle x,u^{\prime}\rangle. It follows that 𝒯u(Δ)=𝒯u(Δ)\mathcal{T}_{u}(\Delta^{\prime})=\mathcal{T}_{u^{\prime}}(\Delta).

Since MM maps the set of primitive vectors bijectively to itself, wT(Δ)=wT(Δ)w_{T}(\Delta)=w_{T}(\Delta^{\prime}).

On the other hand, translations change maxxΔx,u\smash{\max_{x\in\Delta}}\langle x,u\rangle and minxΔx,u\smash{\min_{x\in\Delta}}\langle x,u\rangle by the same amount while leaving the normal vectors vjv_{j} and thus kEuk^{u}_{E} unchanged. Hence, translations do not change the toric width.

Therefore, wT(Δ)=wT(MΔ+v)w_{T}(\Delta)=w_{T}(M\Delta+v) for any MGL(n,)M\in\textrm{GL}(n,{\mathbb{Z}}) and any vnv\in{\mathbb{R}}^{n}. ∎

However, a symplectic manifold could admit many inequivalent toric actions. For instance, Karshon, Kessler and Pinsonnault [12] proved that each compact symplectic 44-manifold admits finitely many inequivalent toric actions, so we immediately have the following corollary.

Corollary 1.3.

Let (M,ω)(M,\omega) be a symplectic manifold that admits a toric action. Then

cHZ(M,ω)supΔ{wT(Δ)|Δ is a Delzant polytope of a toric action.}.c_{\textrm{HZ}}(M,\omega)\geq\sup_{\Delta}\{w_{T}(\Delta)|\Delta\textrm{ is a Delzant polytope of a toric action.}\}.

In Section 2, we review the basic notions of symplectic toric manifolds and study the sub-circle actions. In Section 3, we prove the key lemma that the moment map of a Hamiltonian circle action can be modified to an admissible function. In Section 4, we give an explicit formula for the lower bound of the Hofer-Zehnder capacity of 44-dimensional symplectic toric manifolds and use the formula to find lower bounds of some familiar manifolds. Furthermore, we compare the lower bound with the result of [11] to show our estimate is sharp in many cases.

Convention

Throughout this paper, if not otherwise specified, we will always assume the symplectic manifold (M,ω)(M,\omega) in discussion is compact and connected. For the Hofer-Zehnder capacity of non-compact symplectic manifold, one needs to further require the admissible functions to be compactly supported.

Acknowledgements

This paper is the result of an REU project at the University of Michigan, Ann Arbor, during the summer semester, 2019. I would like to thank Professor Alejandro Uribe for teaching me symplectic geometry, for introducing this problem to me, and for his invaluable help with many aspects of this project. I am very grateful to the Department of Mathematics at the University of Michigan for supporting this project. I also want to thank Jun Li for encouraging me to compile my results into this paper and Shengzhen Ning for pointing out a mistake in an earlier version and suggesting a way to fix it.

2. Sub-circle Actions on Symplectic Toric Manifolds

In this section, we recall some facts about symplectic toric manifolds and study sub-cirlce actions. Throughout this section, we will use the convention that T=n/nT={\mathbb{R}}^{n}/{\mathbb{Z}}^{n} and its Lie algebra is 𝔱n\mathfrak{t}\cong{\mathbb{R}}^{n}. Furthermore, we fix an inner product on 𝔱\mathfrak{t} that identifies 𝔱\mathfrak{t} with 𝔱\mathfrak{t}^{*}.

2.1. Symplectic Toric Manifolds

Definition 2.1.

Suppose TT acts on (M,ω)(M,\omega) via symplectomorphisms. This action is Hamiltonian if there exists a TT-invariant map Φ:M𝔱n\Phi:M\rightarrow\mathfrak{t}^{*}\cong{\mathbb{R}}^{n} such that for each X𝔱X\in\mathfrak{t},

dΦX=ω(X#,)d\Phi^{X}=\omega(X^{\#},\cdot)

where ΦX(p):=Φ(p),X\Phi^{X}(p):=\langle\Phi(p),X\rangle, and X#X^{\#} is the fundamental vector field corresponding to XX.

The tuple (M,ω,T,Φ)(M,\omega,T,\Phi) is called a Hamiltonian TT-space and Φ\Phi is called a moment map.

For any compact connected Hamiltonian TT-space, its moment image Φ(M)\Phi(M) is a convex polytope [1, 9], called the moment polytope.

Since the orbits of the TT-action are isotropic submanifolds, if TT acts effectively, i.e. every non-identity element acts non-trivially, then dim(T)12dim(M)\dim(T)\leq\frac{1}{2}\dim(M). For a compact, connected Hamiltonian TT-space (M,ω,T,Φ)(M,\omega,T,\Phi), if TT acts effectively and dim(T)=12dim(M)\dim(T)=\frac{1}{2}\dim(M), we call the tuple (M,ω,T,Φ)(M,\omega,T,\Phi) a symplectic toric manifold. The moment polytope of a symplectic toric manifold satisfies the following definition.

Definition 2.2.

A Delzant polytope Δn\Delta\subset{\mathbb{R}}^{n} is a polytope satisfying:

  • simplicity: there are n edges meeting at each vertex;

  • rationality: the edges meeting at the vertex p are rational in the sense that each edge is of the form p+tui,t0p+tu_{i},t\geq 0, with uinu_{i}\in\mathbb{Z}^{n};

  • smoothness: for each vertex, the corresponding u1,,unu_{1},...,u_{n} can be chosen to be a \mathbb{Z}-basis of n\mathbb{Z}^{n}.

Delzant [4] classified symplectic toric manifolds by their moment images.

Theorem 2.3 (Delzant).

There is a one-to-one correspondence bewteen the following two sets.

{Symplectic Toric Manifolds}/{\{\text{Symplectic Toric Manifolds}\}/{\sim}}{Delzant Polytopes}/{\{\text{Delzant Polytopes}\}/{\sim}}11\scriptstyle{1-1}

The equivalence relation is given by equivariant symplectormorhism on the left and by translation on the right.

Example 2.4.

Consider the symplectic manifold (S2,c1(TS2))(S^{2},c_{1}(TS^{2})). In cylindrical coordinate, the symplectic form is 12πdθdh\frac{1}{2\pi}d\theta\wedge dh. Let S1S^{1} act on it by rotation:

t.(θ,h)=(θ+2πt,h).t.(\theta,h)=(\theta+2\pi t,h).

The moment map is h:S2h:S^{2}\rightarrow\mathbb{R} and the moment polytope is [1,1][-1,1].

Φ=h\Phi=h1-1

2.2. Sub-circle Actions

Let (M,ω,T,Φ)(M,\omega,T,\Phi) be a symplectic toric manifold with the Delzant polytope

Δ=i=1d{xn|x,viλi}.\Delta=\bigcap_{i=1}^{d}\{x\in\mathbb{R}^{n}\big{|}\,\langle x,v_{i}\rangle\leq\lambda_{i}\}.

We will give a formula for computing the stabilizer group under a sub-circle action of a non-fixed point whose moment image is on an edge of Δ\Delta. We need the following lemma.

Lemma 2.5.

Let u,v1,v2,,vn1nu,v_{1},v_{2},...,v_{n-1}\in\mathbb{Z}^{n} be linearly independent primitive vectors. Let TTT^{\prime}\subset T be the (n1)(n-1) dimensional subtorus whose Lie algebra is generated by v1,v2,,vn1v_{1},v_{2},...,v_{n-1} and Su1TS^{1}_{u}\subset T be the circle whose Lie algebra is u{\mathbb{R}}u. Let P={tui=1n1tivi| 0t,ti1}P=\{tu-\sum_{i=1}^{n-1}t_{i}v_{i}\big{|}\,0\leq t,t_{i}\leq 1\}. Then

|Su1T|=|nint(P)|+1.|S^{1}_{u}\cap T^{\prime}|=|\mathbb{Z}^{n}\cap\textrm{int}(P)|+1.
Proof.

Write u=(u1,,un),vi=(vi1,,vin)u=(u^{1},...,u^{n}),v_{i}=(v_{i}^{1},...,v_{i}^{n}). Then

Su1={[tu1,,tun]n/n|t[0,1)},S^{1}_{u}=\{[tu^{1},...,tu^{n}]\in{\mathbb{R}}^{n}/{\mathbb{Z}}^{n}\big{|}\,t\in[0,1)\},

and

T={[j=1n1tjvj1,,j=1n1tjvjn]n/n|tj[0,1)}.T^{\prime}=\big{\{}[\sum_{j=1}^{n-1}t_{j}v_{j}^{1},...,\sum_{j=1}^{n-1}t_{j}v_{j}^{n}]\in{\mathbb{R}}^{n}/{\mathbb{Z}}^{n}\big{|}\,t_{j}\in[0,1)\big{\}}.

Thus, there is a one-to-one correspondence between (Su1T){[0,,0]}(S^{1}_{u}\cap T^{\prime})\setminus\{[0,...,0]\} and

Q:={(t,t1,,tn1)(0,1)n|tukj=1n1tjvjk(mod) for all k=1,,n}.Q:=\{(t,t_{1},...,t_{n-1})\in(0,1)^{n}\big{|}\,tu^{k}\equiv\sum_{j=1}^{n-1}t_{j}v_{j}^{k}\pmod{{\mathbb{Z}}}\textrm{ for all }k=1,...,n\}.

Now, we show that the matrix A=(u,v1,,vn1)A=(u,-v_{1},...,-v_{n-1}) maps QQ bijectively to nint(P)\mathbb{Z}^{n}\cap\textrm{int}(P).

For any (t,t1,,tn1)Q(t,t_{1},...,t_{n-1})\in Q, there exist unique k1,,knk_{1},...,k_{n}\in{\mathbb{Z}} such that

(2.1) {tu1j=1n1tjvj1=k1tu2j=1n1tjvj2=k2tunj=1n1tjvjn=kn\begin{cases}tu^{1}-\sum_{j=1}^{n-1}t_{j}v_{j}^{1}=k_{1}\\ tu^{2}-\sum_{j=1}^{n-1}t_{j}v_{j}^{2}=k_{2}\\ \cdots\cdots\cdots\cdots\cdots\cdots\cdots\\ tu^{n}-\sum_{j=1}^{n-1}t_{j}v_{j}^{n}=k_{n}\end{cases}

Thus, (k1,,kn)T=A(t,t1,,tn1)Tnint(P)(k_{1},...,k_{n})^{T}=A(t,t_{1},...,t_{n-1})^{T}\in\mathbb{Z}^{n}\cap\textrm{int}(P).

Since u,v1,,vn1u,v_{1},...,v_{n-1} are linearly independent, AA is invertible. For any (k1,,kn)Tnint(P)(k_{1},...,k_{n})^{T}\in\mathbb{Z}^{n}\cap\textrm{int}(P), there exist unique (t,t1,,tn1)T=A1(k1,,kn)T(0,1)n(t,t_{1},...,t_{n-1})^{T}=A^{-1}(k_{1},...,k_{n})^{T}\in(0,1)^{n}. Now, equation (2.1) implies that (t,t1,,tn1)Q(t,t_{1},...,t_{n-1})\in Q, i.e. nint(P)=AQ\mathbb{Z}^{n}\cap\textrm{int}(P)=AQ.

Since AGL(n,)A\in\textrm{GL}(n,{\mathbb{Z}}), we conclude that |Su1T|=|Q|+1=|nint(P)|+1|S^{1}_{u}\cap T^{\prime}|=|Q|+1=|\mathbb{Z}^{n}\cap\textrm{int}(P)|+1. ∎

Now we apply this lemma to the sub-circle action.

Recall that for an edge EE of Δ\Delta, there exists a unique JE{1,2,.,d}J_{E}\subset\{1,2,....,d\} with |JE|=n1|J_{E}|=n-1 such that

E=ΔjJE{xn|x,vj=λj}.E=\Delta\cap\bigcap_{j\in J_{E}}\{x\in\mathbb{R}^{n}\big{|}\,\langle x,v_{j}\rangle=\lambda_{j}\}.
Corollary 2.6.

Let unu\in{\mathbb{Z}}^{n} be a primitive vector and Su1TS^{1}_{u}\subset T be the circle whose Lie algebra is u{\mathbb{R}}u. Let PEu={tujJEtjvj| 0t,tj1}P^{u}_{E}=\{tu-\sum_{j\in J_{E}}t_{j}v_{j}\big{|}\,0\leq t,t_{j}\leq 1\}, kEu=|nint(PEu)|+1k^{u}_{E}=|{\mathbb{Z}}^{n}\cap\textrm{int}(P^{u}_{E})|+1, and mu(Δ)=maxE{kEu|E is an edge of Δ}m_{u}(\Delta)=\max_{E}\{k^{u}_{E}\big{|}\,E\textrm{ is an edge of }\Delta\}. Then

mu(Δ)=maxp{|stabp|:pMMSu1}.m_{u}(\Delta)=\max_{p}\{\,|\textrm{stab}_{p}|:\,p\in M\setminus M^{S^{1}_{u}}\}.
Proof.

By Delzant’s construction[4], for any xMMTx\in M\setminus M^{T}, there exists an edge EΔE\subset\Delta and a pMMTp\in M\setminus M^{T} such that Φ(p)E\Phi(p)\in E and stabxstabp\textrm{stab}_{x}\subset\textrm{stab}_{p}. Furthermore, stabp\textrm{stab}_{p} for the TT-action is an (n1)(n-1) dimensional subtorus TET_{E} whose Lie algebra is generated by {vj:jJE}\{v_{j}:j\in J_{E}\}. Therefore, for the Su1S^{1}_{u}-action,

maxp{|stabp|:pMMSu1}=maxp{|stabp|:pMMSu1,Φ(p)E}.\max_{p}\{\,|\textrm{stab}_{p}|:\,p\in M\setminus M^{S^{1}_{u}}\}=\max_{p}\{\,|\textrm{stab}_{p}|:\,p\in M\setminus M^{S^{1}_{u}},\Phi(p)\in E\}.

Let pMMTp\in M\setminus M^{T} such that Φ(p)E\Phi(p)\in E for some edge EΔE\subset\Delta. The stabilizer group of pp under the Su1S^{1}_{u}-action is Su1TES^{1}_{u}\cap T_{E}. If uSpan{vj:jJE}u\in\textrm{Span}\{v_{j}:j\in J_{E}\}, then Su1TES^{1}_{u}\subset T_{E}, so pMSu1p\in M^{S^{1}_{u}}. In this case, int(PEu)=\textrm{int}(P^{u}_{E})=\emptyset, so kEu=1k^{u}_{E}=1. If uSpan{vj:jJE}u\not\in\textrm{Span}\{v_{j}:j\in J_{E}\}, by Lemma 2.5, stabp=Su1TE\textrm{stab}_{p}=S^{1}_{u}\cap T_{E} has cardinality kEuk^{u}_{E}. Since kEu1k^{u}_{E}\geq 1,

maxp{|stabp|:pMMSu1,Φ(p)E}=maxE{kEu|E is an edge of Δ}=mu(Δ).\max_{p}\{\,|\textrm{stab}_{p}|:\,p\in M\setminus M^{S^{1}_{u}},\Phi(p)\in E\}=\max_{E}\{k^{u}_{E}\big{|}\,E\textrm{ is an edge of }\Delta\}=m_{u}(\Delta).

3. The Hofer-Zehnder Capacity

In this section, we review the definition of the Hofer-Zehnder capacity and prove the key lemma used in the proof of the main theorem. Recall that H:MH:M\to{\mathbb{R}} is admissible if

  • there exist open sets U,VU,V such that H|U=HmaxH|_{U}=H_{\max} and H|V=HminH|_{V}=H_{\min}, and

  • XHX_{H} has no non-constant periodic orbits of period less than 11.

We denote the set of admissible functions by ad(M,ω)={HC(M)\mathcal{H}_{ad}(M,\omega)=\{H\in C^{\infty}(M) HH is admissible}\}.

Definition 3.1.

The Hofer-Zehnder capacity of (M,ω)(M,\omega) is defined by

cHZ(M,ω)=sup{HmaxHmin|Had(M,ω)}.c_{\textrm{HZ}}(M,\omega)=\textrm{sup}\{H_{\max}-H_{\min}\big{|}\,H\in\mathcal{H}_{ad}(M,\omega)\}.

An equivalent description of the Hofer-Zehnder capacity is given below.

For a smooth function H:MH:M\to{\mathbb{R}}, let

𝒯H:=inf{𝒯|𝒯 is the period of a non-constant periodic orbit of XH}.\mathcal{T}_{H}:=\textrm{inf}\{\mathcal{T}\big{|}\,\mathcal{T}\textrm{ is the period of a non-constant periodic orbit of }X_{H}\}.
Proposition 3.2.

Let ~(M):={HC(M)\tilde{\mathcal{H}}(M):=\{H\in C^{\infty}(M) H(M)=[0,1]H(M)=[0,1] , HH attains 0,10,1 on open sets}\}. Then

cHZ(M,ω)=sup{𝒯H|H~(M)}.c_{\textrm{HZ}}(M,\omega)=\textrm{sup}\{\mathcal{T}_{H}\big{|}\,H\in\tilde{\mathcal{H}}(M)\}.

We first establish a lemma that will be used in the proof of both Proposition 3.2 and a later result.

Lemma 3.3.

Let (M,ω)(M,\omega) be a symplectic manifold and f:Mf:M\to{\mathbb{R}} be a non-constant smooth function. Then for any a>0a\in{\mathbb{R}}>0, let g=afg=af, then 𝒯g=𝒯fa\mathcal{T}_{g}=\frac{\mathcal{T}_{f}}{a}.

Proof.

Since dg=adfdg=adf, Xg=aXfX_{g}=aX_{f}. Let γ(t)\gamma(t) be an integral curve of XfX_{f} starting from a point pMp\in M. Let η(t):=γ(at)\eta(t):=\gamma(at), then

η(t)=aγ(at)=aXf(γ(at))=aXf(η(t))=Xg(η(t)).\eta^{\prime}(t)=a\gamma^{\prime}(at)=aX_{f}(\gamma(at))=aX_{f}(\eta(t))=X_{g}(\eta(t)).

The flow ϕft\phi^{t}_{f} for XfX_{f} and the flow ϕgt\phi^{t}_{g} for XgX_{g} are related by ϕgt=ϕfat.\phi^{t}_{g}=\phi^{at}_{f}. Therefore, there is a one-to-one correspondence between the non-constant periodic orbits of XfX_{f} and the non-constant periodic orbits of XgX_{g}. In addition, for any non-constant periodic orbits of XfX_{f} with period 𝒯\mathcal{T}, the corresponding periodic orbits of XgX_{g} has period 𝒯a\frac{\mathcal{T}}{a}. The lemma follows. ∎

Proof of Proposition 3.2.

For any non-constant function Had(M,ω)H\in\mathcal{H}_{ad}(M,\omega), let H:=HmaxHmin||H||:=H_{\max}-H_{\min}, then

H~:=HHminH~(M).\tilde{H}:=\frac{H-H_{\min}}{||H||}\in\tilde{\mathcal{H}}(M).

By Lemma 3.3, 𝒯H~H𝒯H\mathcal{T}_{\tilde{H}}\geq||H||\mathcal{T}_{H}. Since HH is admissible, 𝒯H1\mathcal{T}_{H}\geq 1, so 𝒯H~H\mathcal{T}_{\tilde{H}}\geq||H||.

Taking supremum over ad(M,ω)\mathcal{H}_{ad}(M,\omega),

sup{𝒯H~|Had(M,ω)}sup{HmaxHmin|Had(M,ω)}=cHZ(M,ω).\textrm{sup}\{\mathcal{T}_{\tilde{H}}\big{|}\,H\in\mathcal{H}_{ad}(M,\omega)\}\geq\textrm{sup}\{H_{\max}-H_{\min}\big{|}\,H\in\mathcal{H}_{ad}(M,\omega)\}=c_{\textrm{HZ}}(M,\omega).

Since {𝒯H~|Had(M,ω)}{𝒯H|H~(M)}\{\mathcal{T}_{\tilde{H}}\big{|}\,H\in\mathcal{H}_{ad}(M,\omega)\}\subset\{\mathcal{T}_{H}\big{|}\,H\in\tilde{\mathcal{H}}(M)\}, we conclude that

sup{𝒯H|H~(M)}sup{𝒯H~|Had(M,ω)}cHZ(M,ω).\textrm{sup}\{\mathcal{T}_{H}\big{|}\,H\in\tilde{\mathcal{H}}(M)\}\geq\textrm{sup}\{\mathcal{T}_{\tilde{H}}\big{|}\,H\in\mathcal{H}_{ad}(M,\omega)\}\geq c_{\textrm{HZ}}(M,\omega).

Conversely, for any H~~(M)\tilde{H}\in\tilde{\mathcal{H}}(M) such that 𝒯H~>0\mathcal{T}_{\tilde{H}}>0, let H:=𝒯H~H~H:=\mathcal{T}_{\tilde{H}}\tilde{H}. By Lemma 3.3, 𝒯H𝒯H~𝒯H~=1.\mathcal{T}_{H}\geq\frac{\mathcal{T}_{\tilde{H}}}{\mathcal{T}_{\tilde{H}}}=1. Thus, Had(M,ω)H\in\mathcal{H}_{ad}(M,\omega) and H=𝒯H~||H||=\mathcal{T}_{\tilde{H}}. This implies that

{𝒯H~:H~~(M)}{HmaxHmin|Had(M,ω)}.\{\mathcal{T}_{\tilde{H}}:\tilde{H}\in\tilde{\mathcal{H}}(M)\}\subset\{H_{\max}-H_{\min}\big{|}\,H\in\mathcal{H}_{ad}(M,\omega)\}.

Therefore,

sup{𝒯H~|H~~(M)}cHZ(M,ω).\textrm{sup}\{\mathcal{T}_{\tilde{H}}\big{|}\,\tilde{H}\in\tilde{\mathcal{H}}(M)\}\leq c_{\textrm{HZ}}(M,\omega).

We conclude that

cHZ(M,ω)=sup{𝒯H|H~(M)}.c_{\textrm{HZ}}(M,\omega)=\textrm{sup}\{\mathcal{T}_{H}\big{|}\,H\in\tilde{\mathcal{H}}(M)\}.

For a Hamiltonian circle action on a symplectic manifold, its moment map does not attain its max and min on an open set, so it is not an admissible function. We now show that we can always modify the moment map of a Hamiltonian S1S^{1}-action to an admissible function while keeping track of its magnitude, i.e. the difference between its max and its min. In [11], the authors modified the moment map of a semi-free Hamiltonian S1S^{1}-action to get an admissible function. Specifically, they modified the function locally in an S1S^{1}-invariant neighborhood of a minimal fixed point. The similar technique will not work here, as their modification is based on the fact that near a minimal fixed point, the isotropy weights are all 11. Instead, Lemma 4.0.1 in [2] does the work. We remark that in our proof we need to first divide the moment map by a constant to ensure that the non-constant periodic orbits have periods at least 11.

Lemma 3.4.

Let (M,ω)(M,\omega) be a compact, connected symplectic manifold. Let S1S^{1} act effectively on (M,ω)(M,\omega) with a moment map Φ:M\Phi:M\to{\mathbb{R}}. Let 𝒯=max{|stabp|:pMMS1}\mathcal{T}=\max\{\,|\textrm{stab}_{p}|:\,p\in M\setminus M^{S^{1}}\}. Then there exists an admissible function HH such that HmaxHmin=1𝒯(ΦmaxΦmin)H_{\max}-H_{\min}=\frac{1}{\mathcal{T}}(\Phi_{\max}-\Phi_{\min})

Proof.

Let Ψ=Φ𝒯\Psi=\frac{\Phi}{\mathcal{T}}. We first show that the non-constant periodic orbits of XΨX_{\Psi} are of period at least 11. By Lemma 3.3, it suffices to show that the non-constant periodic orbits of XΦX_{\Phi} are of period at least 1𝒯\frac{1}{\mathcal{T}}.

For any free orbit, its period is 11 which is greater than or equal to 1𝒯1\frac{1}{\mathcal{T}}\geq 1.

Suppose 𝒪p\mathcal{O}_{p} is a non-free orbit with stabp/k\textrm{stab}_{p}\cong{\mathbb{Z}}/k{\mathbb{Z}} for some k>1k>1. Then its period is 1k\frac{1}{k}. Since 𝒯k\mathcal{T}\geq k, we conclude that 1k1𝒯\frac{1}{k}\geq\frac{1}{\mathcal{T}}.

Now, we modify Ψ\Psi so that it achieves its maximum (and its minimum, respectively) on an open set. Let f:[Ψmin,Ψmax][0,)f:[\Psi_{\min},\Psi_{\max}]\to[0,\infty) be a smooth function that satisfies the following properties:

  • (i)

    there exists an ϵ>0\epsilon>0 such that f(x)=0f(x)=0 for x[Ψmin,Ψmin+ϵ)x\in[\Psi_{\min},\Psi_{\min}+\epsilon) and f(x)=ΨmaxΨminf(x)=\Psi_{\max}-\Psi_{\min} for x(Ψmaxϵ,Ψmax]x\in(\Psi_{\max}-\epsilon,\Psi_{\max}].

  • (ii)

    0f(x)<10\leq f^{\prime}(x)<1.

Let H=fΨH=f\circ\Psi. By item (i), HH attains its maximum and minimum on open sets. Since each non-constant periodic orbit of Ψ\Psi has period at least 11, by item (ii), each non-constant periodic orbit of XHX_{H} has period at least 1f(x)1\frac{1}{f^{\prime}(x)}\geq 1. Hence, HH is an admissible function and HmaxHmin=ΨmaxΨmin=1𝒯(ΦmaxΦmin)H_{\max}-H_{\min}=\Psi_{\max}-\Psi_{\min}=\frac{1}{\mathcal{T}}(\Phi_{\max}-\Phi_{\min}). ∎

4. Examples

In this section, we describe an explicit algorithm to find the lower bound of the Hofer-Zehnder capacity provided in Theorem 1.1 for 44-dimensional symplectic toric manifolds. The technical difficulty is to find an explicit formula for the number of lattice points in the interior of a simple lattice polytope. The computation in dimension 4 is feasible thanks to the Pick’s Theorem.

Theorem 4.1 (Pick’s Theorem).

Let Δ2\Delta\subset\mathbb{R}^{2} be a convex lattice polygon. Let AA denote the area of Δ\Delta, ii denote the number of lattice points in the interior of the polygon, bb denote the number of lattice points on the boundary. Then

A=i+b21.A=i+\frac{b}{2}-1.
Corollary 4.2.

Let Δ2\Delta\subset{\mathbb{R}}^{2} be a Delzant polygon. Let EE be an edge of Δ\Delta with the primitive outward normal vector vv. For any primitive vector u2{±v}u\in{\mathbb{Z}}^{2}\setminus\{\pm v\}, kEu=|det(u,v)|k^{u}_{E}=|\det(u,-v)|. In particular, if v1,,vdv_{1},\ldots,v_{d} are primitive outward normal vectors of edges of Δ\Delta, then

mu(Δ)=max{|det(u,v1)|,,|det(u,vd)|}.m_{u}(\Delta)=\max\{|\det(u,-v_{1})|,...,|\det(u,-v_{d})|\}.
Proof.

PEu={tusv:t,s[0,1]}P^{u}_{E}=\{tu-sv:t,s\in[0,1]\} is a parallelogram whose vertices are lattice points. By Theorem 4.1, kEu=|nint(PEu)|+1=i+1=Ab2+2k^{u}_{E}=|{\mathbb{Z}}^{n}\cap\textrm{int}(P^{u}_{E})|+1=i+1=A-\frac{b}{2}+2. Since u,vu,v are primitive, there are four lattice points on the boundary of PEuP^{u}_{E}, namely 0,u,v,uv0,u,-v,u-v. Hence, b=4b=4. On the other hand, A=|det(u,v)|A=|\det(u,-v)|, so

kEu=|det(u,v)|42+2=|det(u,v)|.k^{u}_{E}=|\det(u,-v)|-\frac{4}{2}+2=|\det(u,-v)|.

If u,viu,v_{i} are linearly dependent, then there exists some jj such that u,vju,v_{j} are linearly independent. Hence, kEju=|det(u,vj)|1=kEiuk^{u}_{E_{j}}=|\det(u,-v_{j})|\geq 1=k^{u}_{E_{i}}, which implies

max{kE1u,,kEdu}=max{|det(u,v1)|,,|det(u,vd)|}.\max\{k^{u}_{E_{1}},\ldots,k^{u}_{E_{d}}\}=\max\{|\det(u,-v_{1})|,...,|\det(u,-v_{d})|\}.

Example 4.3.

Let c>0c>0. Consider the symplectic toric manifold (2,cω,T2,Φ)(\mathbb{CP}^{2},c\omega,T^{2},\Phi) whose moment image Δ\Delta is given in Figure 1:

A=(0,0)A=(0,0)B=(c,0)B=(c,0)C=(0,c)C=(0,c)
Figure 1. The moment image of 2{\mathbb{C}}{\mathbb{P}}^{2}.

We denote the primitive outward normal vectors by v1=(0,1),v2=(1,1),v3=(1,0)v_{1}=(0,-1),v_{2}=(1,1),v_{3}=(-1,0). Let u=(p,q)2{0}u=(p,q)\in{\mathbb{Z}}^{2}\setminus\{0\} be a primitive vector. |det(u,v1)|=|p||\det(u,-v_{1})|=|p|, |det(u,v2)|=|pq|,|det(u,v3)|=|q||\det(u,-v_{2})|=|p-q|,|\det(u,-v_{3})|=|q|. By Corollary 4.2, mu(Δ)=max{|p|,|pq|,|q|}m_{u}(\Delta)=\max\{|p|,|p-q|,|q|\}.

Case 1. pq<0pq<0. Then we can assume without loss of generality that p>0,q<0p>0,q<0, since multiplying uu by 1-1 will not change mu(Δ)m_{u}(\Delta). Then mu(Δ)=pqm_{u}(\Delta)=p-q.

𝒯u(Δ)=1pq(max(x,y)Δ(px+qy)min(x,y)Δ(px+qy)).\mathcal{T}_{u}(\Delta)=\frac{1}{p-q}\big{(}\smash{\max_{(x,y)\in\Delta}}(px+qy)-\smash{\min_{(x,y)\in\Delta}}(px+qy)\big{)}.

If we think of the geometric meaning of px+qypx+qy and write b=px+qyb=px+qy, then bb is maximized when the y-intercept of the line y=pqx+y0y=-\frac{p}{q}x+y_{0} is minimized.

As the blue line in the picture shows, bb is maximized at BB and minimized at CC. So 𝒯u(Δ)=1pq(cpcq)=c\mathcal{T}_{u}(\Delta)=\frac{1}{p-q}(cp-cq)=c.

Case 2. pq>0pq>0. Then we can assume without loss of generality that p,q>0p,q>0.

In this case, mu(Δ)=max(p,q)m_{u}(\Delta)=\max(p,q).

We want to maximize and minimize px+qypx+qy on Δ\Delta. Similarly, when p>qp>q, as the red line shows, it’s maximized at BB and minimized at AA. So 𝒯u(Δ)=1p(cp0)=c\mathcal{T}_{u}(\Delta)=\frac{1}{p}(cp-0)=c. The same argument works for p<qp<q (the orange line) and we still get 𝒯u(Δ)=c\mathcal{T}_{u}(\Delta)=c.

Case 3. u=(0,±1)u=(0,\pm 1) or u=(±1,0)u=(\pm 1,0). It is straightforward to check that 𝒯u(Δ)=c\mathcal{T}_{u}(\Delta)=c.

By Theorem 1.1, we get the estimate cHZ(2,cω)cc_{\textrm{HZ}}(\mathbb{CP}^{2},c\omega)\geq c.

Example 4.4.

For a,b>0a,b\in{\mathbb{R}}_{>0}, consider (S2×S2,aωbω,T2)(S^{2}\times S^{2},a\,\omega\oplus b\,\omega,T^{2}) whose moment image is Δ=[0,a]×[0,b]\Delta=[0,a]\times[0,b], where ω=12c1(TS2)\omega=\frac{1}{2}c_{1}(TS^{2}). We denote the primitive outward normal vectors by v1=(1,0),v2=(0,1),v3=(1,0),v4=(0,1)v_{1}=(-1,0),v_{2}=(0,-1),v_{3}=(1,0),v_{4}=(0,1).

For any primitive vector u=(p,q)2{0}u=(p,q)\in{\mathbb{Z}}^{2}\setminus\{0\}, by Corollary 4.2, mu(Δ)=max{|p|,|q|}m_{u}(\Delta)=\max\{|p|,|q|\}.

A computation similar to the previous example shows that

𝒯u(Δ)=a|p|+b|q|max{|p|,|q|}.\mathcal{T}_{u}(\Delta)=\frac{a|p|+b|q|}{\max\{|p|,|q|\}}.

Notice that 𝒯u(Δ)a+b\mathcal{T}_{u}(\Delta)\leq a+b for all primitive vectors uu. Furthermore, for u=(1,1)u=(1,1), 𝒯u(Δ)=a+b\mathcal{T}_{u}(\Delta)=a+b. Hence, wT(Δ)=a+bw_{T}(\Delta)=a+b and cHZ(S2×S2,aωbω)a+bc_{\textrm{HZ}}(S^{2}\times S^{2},a\,\omega\oplus b\,\omega)\geq a+b.

Remark 4.5.

In [11], the authors proved that if (M,ω)(M,\omega) is a closed Fano symplectic manifold with a semi-free S1S^{1}-action and the Hamiltonian function H:MH:M\to{\mathbb{R}} attains its max at a single point, then cHZ(M,ω)=HmaxHminc_{\textrm{HZ}}(M,\omega)=H_{\max}-H_{\min}. In particular, they showed that cHZ(2,3ω)=3c_{\textrm{HZ}}(\mathbb{CP}^{2},3\omega)=3 and cHZ(S2×S2,aωbω)=a+bc_{\textrm{HZ}}(S^{2}\times S^{2},a\omega\oplus b\omega)=a+b. By the conformality of the symplectic capacity, we further obtain that cHZ(2,cω)=cc_{\textrm{HZ}}(\mathbb{CP}^{2},c\omega)=c for any c>0c>0. Hence, our estimate is sharp in these cases. In fact, our estimate is always sharp in the Fano case.

Example 4.6.

Given n>0n\in{\mathbb{Z}}_{>0}, let 𝒪(n)1\mathcal{O}(n)\to{\mathbb{C}}{\mathbb{P}}^{1} be the holomorphic line bundle whose first Chern class is nn. Consider the nn-th Hirzebruch surface n:=(𝒪𝒪(n))\mathcal{H}_{n}:={\mathbb{P}}(\mathcal{O}\oplus\mathcal{O}(-n)). Under a suitable choice of the torus action and the symplectic form, its moment image Δ\Delta is the trapezoid shown in Figure 2, where a,b>0a,b>0. The primitive outward normal vectors are v1=(0,1),v2=(1,n),v3=(0,1),v4=(1,0)v_{1}=(0,-1),v_{2}=(1,n),v_{3}=(0,1),v_{4}=(-1,0).

(0,0)(0,0)(a+nb,0)(a+nb,0)(0,b)(0,b)(a,b)(a,b)
Figure 2. The moment image of the nn-th Hirzebruch surface

For any primitive vector u=(p,q)2{0}u=(p,q)\in{\mathbb{Z}}^{2}\setminus\{0\}, by Corollary 4.2, mu(Δ)=max{|p|,|npq|,|q|}m_{u}(\Delta)=\max\{|p|,|np-q|,|q|\}.

When n=1n=1, mu(Δ)=max{|p|,|pq|,|q|}m_{u}(\Delta)=\max\{|p|,|p-q|,|q|\}. A discussion similar to Example 4.3 shows that wT(Δ)=a+bw_{T}(\Delta)=a+b. Hence, cHZ(1,ωΔ)a+bc_{\textrm{HZ}}(\mathcal{H}_{1},\omega_{\Delta})\geq a+b. For example, when (a,b)=(1,2)(a,b)=(1,2), this toric manifold is Fano and the estimate is sharp.

For n2n\geq 2, we put the data in the table below.

uu mu(Δ)m_{u}(\Delta) 𝒯u(Δ)\mathcal{T}_{u}(\Delta)
(0,±1)(0,\pm 1) 11 bb
(±1,0)(\pm 1,0) nn a+nbn\frac{a+nb}{n}
pq<0\frac{p}{q}<0 |npq||np-q| |pa+(npq)b||npq|\frac{|pa+(np-q)b|}{|np-q|}
0<pq1n0<\frac{p}{q}\leq\frac{1}{n} |q||q| |p|a+|q|b|q|\frac{|p|a+|q|b}{|q|}
1n<pq2n\frac{1}{n}<\frac{p}{q}\leq\frac{2}{n} |q||q| (a+nb)|p||q|\frac{(a+nb)|p|}{|q|}
pq>2n\frac{p}{q}>\frac{2}{n} |npq||np-q| (a+nb)|p||npq|\frac{(a+nb)|p|}{|np-q|}

When pq<0\frac{p}{q}<0,

|pa+(npq)b||npq||p|a|npq|+b=a|nqp|+b<an+b\frac{|pa+(np-q)b|}{|np-q|}\leq\frac{|p|a}{|np-q|}+b=\frac{a}{|n-\frac{q}{p}|}+b<\frac{a}{n}+b

When 0<pq1n0<\frac{p}{q}\leq\frac{1}{n},

|p|a+|q|b|q|=apq+ban+b\frac{|p|a+|q|b}{|q|}=a\frac{p}{q}+b\leq\frac{a}{n}+b

When 1n<pq2n\frac{1}{n}<\frac{p}{q}\leq\frac{2}{n},

(4.1) (a+nb)|p||q|2an+2b\frac{(a+nb)|p|}{|q|}\leq\frac{2a}{n}+2b

When pq>2n\frac{p}{q}>\frac{2}{n},

(a+nb)|p||npq|=a+nb|nqp|<2n(a+nb)=2an+2b\frac{(a+nb)|p|}{|np-q|}=\frac{a+nb}{|n-\frac{q}{p}|}<\frac{2}{n}(a+nb)=\frac{2a}{n}+2b

Furthermore, when (p,q)=(2gcd(2,n),ngcd(2,n))(p,q)=\left(\frac{2}{\gcd(2,n)},\frac{n}{\gcd(2,n)}\right), the inequality (4.1) is an equality. Hence, wT(Δ)=2an+2bw_{T}(\Delta)=\frac{2a}{n}+2b.

Therefore, for n2n\geq 2, cHZ(n,ωΔ)2an+2bc_{\textrm{HZ}}(\mathcal{H}_{n},\omega_{\Delta})\geq\frac{2a}{n}+2b. cHZ(n,ωΔ)c_{\textrm{HZ}}(\mathcal{H}_{n},\omega_{\Delta}) is not known in this case.

Remark 4.7.

Though the computation is straightforward in dimension 22 (i.e. when we have symplectic toric 44-manifolds), it is not known whether there is a closed formula for the number of interior lattice points of a simple lattice polytope in higher dimension. Brion and Vergne [3] gave a formula in terms of the derivative of the Todd operator. It is not clear to us how this formula could be applied to extract the exact numbers.

References

  • [1] M. F. Atiyah, Convexity and commuting hamiltonians, Bulletin of the London Mathematical Society 14 (1982), no. 1, 1–15.
  • [2] J. Bimmermann, On the hofer-zehnder capacity of twisted tangent bundles, Ph.D. thesis, Heidelberg University, 2023.
  • [3] Michel Brion and Michèle Vergne, Lattice points in simple polytopes, J. Amer. Math. Soc. 10 (1997), no. 2, 371–392. MR 1415319
  • [4] T. Delzant, Hamiltoniens périodiques et image convex de l’application moment, Bull. Soc. Math. France 116 (1988), 315–339.
  • [5] Yu Du, Gabriel Kosmacher, Yichen Liu, Jeff Massman, Joseph Palmer, Timothy Thieme, Jerry Wu, and Zheyu Zhang, Packing densities of Delzant and semitoric polygons, SIGMA Symmetry Integrability Geom. Methods Appl. 19 (2023), Paper No. 081, 42. MR 4660165
  • [6] I. Ekeland and H. Hofer, Symplectic topology and Hamiltonian dynamics, Math. Z. 200 (1989), no. 3, 355–378. MR 978597
  • [7] A. Figalli, J. Palmer, and Á. Pelayo, Symplectic GG-capacities and integrable systems, Ann. Sc. Norm. Super. Pisa Cl. Sci. (5) 18 (2018), no. 1, 65–103.
  • [8] M. Gromov, Pseudo holomorphic curves in symplectic manifolds, Invent. Math. 82 (1985), no. 2, 307–347. MR 809718
  • [9] V. Guillemin and S. Sternberg, Convexity properties of the moment mapping, Inventiones Mathematicae 67 (1982), no. 3, 491–513.
  • [10] H. Hofer and E. Zehnder, A new capacity for symplectic manifolds, Analysis, et cetera, Academic Press, Boston, MA, 1990, pp. 405–427. MR 1039354
  • [11] Taekgyu Hwang and Dong Youp Suh, Symplectic capacities from Hamiltonian circle actions, J. Symplectic Geom. 15 (2017), no. 3, 785–802. MR 3696591
  • [12] Yael Karshon, Liat Kessler, and Martin Pinsonnault, A compact symplectic four-manifold admits only finitely many inequivalent toric actions, J. Symplectic Geom. 5 (2007), no. 2, 139–166. MR 2377250
  • [13] Guangcun Lu, Symplectic capacities of toric manifolds and related results, Nagoya Math. J. 181 (2006), 149–184. MR 2210713
  • [14] Á Pelayo, Toric symplectic ball packing, Topology and its Appl. 157 (2006), 3633–3644.
  • [15] Á. Pelayo, Topology of spaces of equivariant symplectic embeddings, Proc. Amer. Math. Soc. 135 (2007), no. 1, 277–288.
  • [16] Á Pelayo and B. Schmidt, Maximal ball packings of symplectic-toric manifolds, Intern. Math. Res. Not. (2008), 24p, ID rnm139.