ScalingStacks

2.3. Green’s function on a flat cylinder [03GL]

Original official author HTML, exact retained edition. Historical TeX conversion verdicts remain unchanged. Cited-edition alignment and mathematical self-containment are not assessed.

Complete original source context · Original author HTML

2.3. Green’s function on a flat cylinder

The neck region in our gluing construction is given by a doubly-periodic analog of the Ooguri-Vafa metric. To construct this metric using the Gibbons-Hawking ansatz, we first need to construct a Green’s function V∞V_{\infty} on (𝕋2×ℝ,g0)(\mathbb{T}^{2}\times\mathbb{R},g_{0}), where 𝕋2\mathbb{T}^{2} is any flat 22-torus. We also need to determine the asymptotics of V∞V_{\infty} as |z|→∞|z|\to\infty in order to ensure that the neck matches the Calabi model space from Section 2.2 on both ends.

It is known that the flat cylinder 𝕋2×ℝ\mathbb{T}^{2}\times\mathbb{R} is parabolic, namely, it does not admit a positive Green’s function. In fact, Cheng and Yau proved that any complete non-compact Riemannian manifold with Vol⁡(BR​(p))≤C​R2\Vol(B_{R}(p))\leq CR^{2} must be parabolic. See theorem 1 and corollary 1 in [CY75].

We now construct a particular sign-changing Green’s function V∞V_{\infty}. Fix a point pp on 𝕋2×ℝ\mathbb{T}^{2}\times\mathbb{R} with z⁡(p)=0z(p)=0. For R>0R>0 let VRV_{R} denote the unique function on 𝕋2×[−R,R]\mathbb{T}^{2}\times[-R,R] satisfying

(2.27) −Δg0​VR=2πδp,z∈(−R,R),VR=0,z=±R.\displaystyle\begin{split}-\Delta_{g_{0}}V_{R}&=2\pi\delta_{p},\ z\in(-R,R),\\ V_{R}&=0,\ \hskip 16.0ptz=\pm R.\end{split}

The normalization of the right-hand side is precisely chosen in such a way that VR=12​r+O⁡(1)V_{R}=\frac{1}{2r}+O(1) near pp, where rr denotes g0g_{0}-distance to pp. Thus, the 44-dimensional Gibbons-Hawking metric associated with VRV_{R} (or VR+CV_{R}+C for any constant CC) extends smoothly across pp; cf. Example 2.1.

By the maximum principle, one can see that VRV_{R} is an increasing family as R→∞R\to\infty. Define

(2.28) CR≡sup∂B1​(p)VR.C_{R}\equiv\sup\limits_{\partial B_{1}(p)}V_{R}.

Then it follows from the parabolicity of 𝕋2×ℝ\mathbb{T}^{2}\times\mathbb{R} that CR→∞C_{R}\to\infty as R→∞R\to\infty. By a result of Li and Tam (see theorem 1 in [LT87]), VR​(x,y,z)−CRV_{R}(x,y,z)-C_{R} converges to a function V∞​(x,y,z)V_{\infty}(x,y,z) uniformly on compact subsets of (𝕋2×ℝ)∖{p}(\mathbb{T}^{2}\times\mathbb{R})\setminus\{p\}. Moreover, V∞≤0V_{\infty}\leq 0 on the complement of B1​(p)B_{1}(p) and

(2.29) −Δg0​V∞=2​π​δp​on​𝕋2×ℝ.-\Delta_{g_{0}}V_{\infty}=2\pi\delta_{p}\ \text{on}\ \mathbb{T}^{2}\times\mathbb{R}.

Notice also that V∞V_{\infty} is symmetric in zz by construction.

Theorem 2.6.

There are constants β−,β+∈ℝ\beta_{-},\beta_{+}\in\mathbb{R} and k−,k+∈ℝk_{-},k_{+}\in\mathbb{R} with

(2.30) k−=−k+=πAreag0⁡(𝕋2)>0k_{-}=-k_{+}=\frac{\pi}{\Area_{g_{0}}(\mathbb{T}^{2})}>0

such that for all k∈ℕk\in\mathbb{N},

(2.31) |∇g0k(V∞​(z)−(k−​z+β−))|=O(eλ1​z),z→−∞,|∇g0k(V∞​(z)−(k+​z+β+))|=O(e−λ1​z),z→+∞,\displaystyle\begin{split}|\nabla^{k}_{g_{0}}(V_{\infty}(z)-(k_{-}z+\beta_{-}))|&=O(e^{\sqrt{\lambda_{1}}z}),\ z\to-\infty,\\ |\nabla^{k}_{g_{0}}(V_{\infty}(z)-(k_{+}z+\beta_{+}))|&=O(e^{-\sqrt{\lambda_{1}}z}),\ z\to+\infty,\\ \end{split}

where λ1>0\lambda_{1}>0 is the smallest eigenvalue of −Δ𝕋2-\Delta_{\mathbb{T}^{2}}.

Proof.

Consider the fiberwise average

(2.32) 𝒱∞​(z)≡1Areag0​(𝕋2)​∫𝕋2×{z}V∞​(x,y,z)​dvolg0⁡(x,y).\mathcal{V}_{\infty}(z)\equiv\frac{1}{{\rm Area}_{g_{0}}(\mathbb{T}^{2})}\int_{\mathbb{T}^{2}\times\{z\}}V_{\infty}(x,y,z)\dvol_{g_{0}}(x,y).

This is well-defined, smooth in zz for z≠0z\neq 0, and continuous at z=0z=0. For z≠0z\neq 0 we have

(2.33) 𝒱∞′′(z)=∫𝕋2×{z}d2d​z2V∞=−∫𝕋2×{z}Δ𝕋2V∞=0.\mathcal{V}_{\infty}^{\prime\prime}(z)=\int_{\mathbb{T}^{2}\times\{z\}}\frac{d^{2}}{dz^{2}}V_{\infty}=-\int_{\mathbb{T}^{2}\times\{z\}}\Delta_{\mathbb{T}^{2}}V_{\infty}=0.

This implies that 𝒱∞​(z)\mathcal{V}_{\infty}(z) is a piecewise linear function. Since VR​(x,y,z)≤−CRV_{R}(x,y,z)\leq-C_{R} for |z|≥R|z|\geq R with CR→∞C_{R}\to\infty as R→∞R\to\infty, it follows that limz→+∞V∞​(x,y,z)=−∞\lim\limits_{z\to+\infty}V_{\infty}(x,y,z)=-\infty, and hence for all z>0z>0 that

(2.34) 𝒱∞′​(z)=c​o​n​s​t≡k+<0.\mathcal{V}_{\infty}^{\prime}(z)=const\equiv k_{+}<0.

Denote D0≡Diamg0⁡(𝕋2)D_{0}\equiv\diam_{g_{0}}(\mathbb{T}^{2}). Choose R0>10​D0R_{0}>10D_{0} large enough so that V∞​(x,y,z)≤−1V_{\infty}(x,y,z)\leq-1 in (𝕋2×ℝ)∖BR0​(p)(\mathbb{T}^{2}\times\mathbb{R})\setminus B_{R_{0}}(p). Then for any fixed q∈(𝕋2×ℝ)∖B2​R0​(p)q\in(\mathbb{T}^{2}\times\mathbb{R})\setminus B_{2R_{0}}(p) and r∈(2​D0,4​D0)r\in(2D_{0},4D_{0}), we can apply the Harnack inequality to the harmonic function V∞V_{\infty}, which is negative in the geodesic ball Br​(q)⊂(𝕋2×ℝ)∖BR0​(p)B_{r}(q)\subset(\mathbb{T}^{2}\times\mathbb{R})\setminus B_{R_{0}}(p). More precisely, passing to the universal cover and applying the standard Harnack inequality for positive harmonic functions on a fixed ball in ℝ3\mathbb{R}^{3}, we see that there is a uniform constant C0>0C_{0}>0 depending only on D0>0D_{0}>0 such that for all w1,w2∈Br​(q)w_{1},w_{2}\in B_{r}(q),

(2.35) 1C0≤V∞​(w1)V∞​(w2)≤C0.\frac{1}{C_{0}}\leq\frac{V_{\infty}(w_{1})}{V_{\infty}(w_{2})}\leq C_{0}.

Since the fiber average 𝒱∞​(z)\mathcal{V}_{\infty}(z) of V∞V_{\infty} is linear in zz with slope k+<0k_{+}<0, (2.35) yields that

(2.36) −C2​z≤V∞​(x,y,z)≤−C1​z-C_{2}z\leq V_{\infty}(x,y,z)\leq-C_{1}z

for z≫1z\gg 1, where the constants C1C_{1} and C2C_{2} depend only on the constants C0C_{0} and k+k_{+}.

We denote by Λ𝕋2={λj}j=1∞\Lambda_{\mathbb{T}^{2}}=\{\lambda_{j}\}_{j=1}^{\infty} the positive spectrum of −Δ𝕋2-\Delta_{\mathbb{T}^{2}} and expand V∞V_{\infty} according to the eigenfunctions of Δ𝕋2\Delta_{\mathbb{T}^{2}} along the torus fiber 𝕋2×{z}\mathbb{T}^{2}\times\{z\} for each fixed z>0z>0. This yields

(2.37) V∞​(x,y,z)=(k+​z+β+)+∑j=1∞fj​(z)​hj​(x,y),V_{\infty}(x,y,z)=(k_{+}z+\beta_{+})+\sum_{j=1}^{\infty}f_{j}(z)h_{j}(x,y),

where k+<0k_{+}<0 is the constant of (2.34) and where

(2.38) fj′′​(z)=λj​fj​(z),−Δ𝕋2​hj=λj​hj,∫𝕋2|hj|2=1.\displaystyle f_{j}^{\prime\prime}(z)=\lambda_{j}f_{j}(z),\ -\Delta_{\mathbb{T}^{2}}h_{j}=\lambda_{j}h_{j},\ \int_{\mathbb{T}^{2}}|h_{j}|^{2}=1.

Immediately,

(2.39) fj​(z)=cj​e−λj​z+cj∗​eλj​z.f_{j}(z)=c_{j}e^{-\sqrt{\lambda_{j}}z}+c_{j}^{*}e^{\sqrt{\lambda_{j}}z}.

Notice that

(2.40) ∫𝕋2×{z}|V∞|2=(k+​z+β+)2+∑j=1∞|fj​(z)|2.\int_{\mathbb{T}^{2}\times\{z\}}|V_{\infty}|^{2}=(k_{+}z+\beta_{+})^{2}+\sum_{j=1}^{\infty}|f_{j}(z)|^{2}.

By the linear growth property (2.36), we obtain that cj∗=0c_{j}^{*}=0 for all j∈ℤ+j\in\mathbb{Z}_{+}. Therefore,

(2.41) ∫𝕋2×{z}|V∞−(k+​z+β+)|2=∑j=1∞|cj|2​e−2​λj​z=O⁡(e−2​λ1​z)​as​z→+∞,\int_{\mathbb{T}^{2}\times\{z\}}|V_{\infty}-(k_{+}z+\beta_{+})|^{2}=\sum_{j=1}^{\infty}|c_{j}|^{2}e^{-2\sqrt{\lambda_{j}}z}=O(e^{-2\sqrt{\lambda_{1}}z})\ \text{as}\ z\to+\infty,

where λ1>0\lambda_{1}>0 is the minimum of Λ𝕋2\Lambda_{\mathbb{T}^{2}}. To see the O⁡(e−2​λ1​z)O(e^{-2\sqrt{\lambda_{1}}z}) estimate, note that the series converges for z=1z=1. Applying elliptic regularity to the harmonic function V^∞≡V∞−(k+​z+β+)\widehat{V}_{\infty}\equiv V_{\infty}-(k_{+}z+\beta_{+}),

(2.42) ‖V^∞‖W2,2​(Br/2​(q))≤C​‖V^∞‖L2​(Br​(q))≤C​e−2​λ1​z\|\widehat{V}_{\infty}\|_{W^{2,2}(B_{r/2}(q))}\leq C\|\widehat{V}_{\infty}\|_{L^{2}(B_{r}(q))}\leq Ce^{-2\sqrt{\lambda_{1}}z}

for all balls Br​(q)B_{r}(q) as above, where CC depends only on the diameter and on the injectivity radius of 𝕋2\mathbb{T}^{2}. By the 33-dimensional Sobolev embedding W2,2↪C0,12W^{2,2}\hookrightarrow C^{0,\frac{1}{2}},

(2.43) |V∞​(x,y,z)−(k+​z+β+)|=O⁡(e−λ1​z)​as​z→+∞.|V_{\infty}(x,y,z)-(k_{+}z+\beta_{+})|=O(e^{-\sqrt{\lambda_{1}}z})\ \text{as}\ z\to+\infty.

Standard elliptic regularity then shows that for any k∈ℕk\in\mathbb{N},

(2.44) |∇g0k(V∞​(x,y,z)−(k+​z+β+))|=O⁡(e−λ1​z)​as​z→+∞.|\nabla_{g_{0}}^{k}(V_{\infty}(x,y,z)-(k_{+}z+\beta_{+}))|=O(e^{-\sqrt{\lambda_{1}}z})\ \text{as}\ z\to+\infty.

The same argument applies in the case z→−∞z\rightarrow-\infty.

Now we prove the slope relation (2.30). Fix R>0R>0. Then by Green’s formula,

(2.45) 𝒱∞′​(R)−𝒱∞′​(−R)=∫𝕋2×[−R,R]Δ​V∞.\mathcal{V}_{\infty}^{\prime}(R)-\mathcal{V}_{\infty}^{\prime}(-R)=\int_{\mathbb{T}^{2}\times[-R,R]}\Delta V_{\infty}.

Thus, by the definition of k+k_{+} and the analogous definition of k−k_{-},

(2.46) k+​Areag0⁡(𝕋2)−k−​Areag0⁡(𝕋2)=−2​π.k_{+}\Area_{g_{0}}(\mathbb{T}^{2})-k_{-}\Area_{g_{0}}(\mathbb{T}^{2})=-2\pi.

It follows that

(2.47) k−−k+=2​πAreag0⁡(𝕋2).k_{-}-k_{+}=\frac{2\pi}{\Area_{g_{0}}(\mathbb{T}^{2})}.

Since V∞V_{\infty} is symmetric in zz, it holds that k−=−k+k_{-}=-k_{+} and the claim follows. ∎

It is straightforward to use superposition to extend the above construction to the case of multiple poles. Precisely, we have the following corollary.

Corollary 2.7.

Let (𝕋2×ℝ,g0)(\mathbb{T}^{2}\times\mathbb{R},g_{0}) be a flat cylinder with a flat product metric g0g_{0}. Given a finite set 𝒫m0≡{p1,…,pm0}⊂𝕋2×ℝ\mathcal{P}_{m_{0}}\equiv\{p_{1},\ldots,p_{m_{0}}\}\subset\mathbb{T}^{2}\times\mathbb{R}, there is a sign changing Green’s function V∞V_{\infty} with

(2.48) −Δg0​V∞=2​π​∑k=1m0δpk,-\Delta_{g_{0}}V_{\infty}=2\pi\sum\limits_{k=1}^{m_{0}}\delta_{p_{k}},

and there are linear functions L±​(z)=k±​z+β±L_{\pm}(z)=k_{\pm}z+\beta_{\pm} with

(2.49) k−=−k+=π​m0Areag0⁡(𝕋2)>0k_{-}=-k_{+}=\frac{\pi m_{0}}{\Area_{g_{0}}(\mathbb{T}^{2})}>0

such that for all k∈ℕk\in\mathbb{N},

(2.50) |∇g0k(V∞​(z)−L−​(z))|=O(eλ1​z),z→−∞,|∇g0k(V∞​(z)−L+​(z))|=O(e−λ1​z),z→+∞,\displaystyle\begin{split}|\nabla^{k}_{g_{0}}(V_{\infty}(z)-L_{-}(z))|&=O(e^{\sqrt{\lambda_{1}}z}),\ z\to-\infty,\\ |\nabla^{k}_{g_{0}}(V_{\infty}(z)-L_{+}(z))|&=O(e^{-\sqrt{\lambda_{1}}z}),\ z\to+\infty,\\ \end{split}

where λ1>0\lambda_{1}>0 is the smallest eigenvalue of −Δ𝕋2-\Delta_{\mathbb{T}^{2}}.

Remark 2.8.

The asymptotics in Theorem 2.6 have an intuitive electro-magnetic interpretation. Namely, for |z||z| large, the electric potential determined by the union of point charges on Λ×{z0}\Lambda\times\{z_{0}\} is approximated by an evenly distributed charge on the corresponding plane ℂ×{z0}\mathbb{C}\times\{z_{0}\} in the universal cover. The electric potential of this uniformly charged plate corresponds to the linear term in V∞V_{\infty}. Further, for any number of charged plates parallel to the x​yxy-plane located at z=ziz=z_{i}, i=1,…,m0i=1,\ldots,m_{0}, we can add together the corresponding ViV_{i} to get the potential V=V1+⋯+Vm0V=V_{1}+\dots+V_{m_{0}}, and the potential at large distances looks like the potential due to m0m_{0} uniformly charged plates.

In Section 6, we will use V∞V_{\infty} to construct a potential VβV_{\beta} which is positive in a large region, and such that [12​π∗d​Vβ]∈H2​(U,ℤ)[\frac{1}{2\pi}*dV_{\beta}]\in H^{2}(U,\mathbb{Z}) is a integral class, which will then be used to define our neck metric.

Original mathematics by the credited authors. Source-backed reader collection; mathematical self-containment is not assessed.