The invariant symplectic action and decay for vortices (Q1032722)

Let \((M,\omega)\) be a symplectic manifold without boundary, \(G\) a compact connected Lie group with Lie algebra \({\mathfrak g}\) acting on \(M\) in a Hamiltonian way with moment map \(\mu: M\to{\mathfrak g}^*\). For a pair \((x,\xi)\in C^\infty(S^1,M\times{\mathfrak g})\), the invariant action \({\mathcal A}(x,\xi)\) is defined by \[ {\mathcal A}(x,\xi)= A(gx)+ \int^1_0 \langle\mu\circ x,\,\xi- g^{-1}g\rangle\,dt. \] Assume that there exists a \(G\)-invariant \(\omega\)-compatible almost complex structure \(J\) such that \(g_{\omega,J}= \omega(\cdot,J.)\) becomes a \(G\)-invariant Riemannian metric on \(M\). Then the following sharp isoperimetric inequality is shown: For every compact subset \(K\subseteq M^*\), \(M^*\subset M\) the subset of all points on which \(G\) acts freely, and constant \(c>{1\over 4\pi}\), there exists a \(\delta> 0\) such that if \(\overline\ell(Gx)<\delta\), then \[ |{\mathcal A}(x,\xi)|\leq c\|\dot x+ L_x\xi\|^2_p+ {\pi\over m^2_K}\|\mu\circ x\|^2_{p/(p-1)}. \] Here, \(L_x:{\mathfrak g}\to T_xM\) is the infinitesimal action of \({\mathfrak g}\) on the tangent space to \(M\) at \(x\), \(\overline\ell(Gx)\) is the arc length of \(Gx/G\) with respect to the induced metric of \(g_{\omega,J}\) of \(M^*/G\) (Th.1.2). As an application, the following optimal decay at \(\infty\) of symplectic vortices [\textit{K. Cieliebak}, \textit{A. R. Gaio} and \textit{D. A. Salamon}, \(J\)-holomorphic curves, moment maps, and invariants of Hamiltonian group actions, Int. Math. Res. Not. 2000, No. 16, 831--882 (2000; Zbl 1083.53084); \textit{I. Mundet i Riera}, Hamiltonian Gromov-Witten invariants, Topology 42, No. 3, 525--553 (2003; Zbl 1032.53079)] on the half cylinder \([0,\infty)\times S^1\) is shown under the hypothesis \(({\mathbf H})\), which assumes the existence of an \(\varepsilon> 0\) such that \(\{x\in M\|\mu(x)\|\leq\varepsilon\}\) is compact and \(G\) acts freely on \(\mu^{-1}(0)\): \[ e_w(s+ it)\leq C\lambda^{-2} e^{(-{4\pi\over a}+\varepsilon)s},\quad s\geq 1,\;t\in\mathbb{R}/a\mathbb{Z}, \] (Th.1.3). Here \(e_w={1\over 2}(|d_A u|^2+ |F_A|^2+|\mu\circ u|^2)\) is the energy density of \(w= (u,A)\), \(u\) an equivariant map from a principal \(G\)-bundle \(P\) over a Riemannian surface \(\Sigma\) equipped with a compatible area form \(\omega_\Sigma\), and \(A\) is a connection 1-form on \(P\). \(w\) is assumed to be a locally \(W^{1,p}\)-solution of the vortex equation \[ \overline\partial_{J,A}(u)= 0,\quad F_A+ (\mu\circ u)\omega_\Sigma= 0. \] The author claims that if \(G= \{1\}\), Th.1.2 generalizes the isoperimetric inequality for the usual action [\textit{D. McDuff} and \textit{D. A. Salamon}, \(J\)-holomorphic curves and symplectic topology, Colloquium Publications. American Mathematical Society 52. Providence, RI: American Mathematical Society (AMS) (2004; Zbl 1064.53051), Th.4.4.1]. An example showing that \(c>{1\over 4\pi}\) is best possible, is also given. To prove Th.1.2, under suitable assumptions on diameters of images of loops, the symplectic action of a loop \(x: S^1\to M\) is defined by \[ {\mathcal A}(x)=- \int_{\mathbb{D}} u^*\omega. \] Here \(\mathbb{D}\) is the closed unit disk and \(u:\mathbb{D}\to M\) is any smooth map such that \(u(e^{2\pi it})= x(t)\). Then, if the diameters of \(x\) and \(gx\) are small, it is shown that \[ {\mathcal A}(gx)-{\mathcal A}(x)= \int^1_0 \langle\mu\circ x,\,g^{-1}\dot g\rangle\,dt, \] (Lemma 2.4 (Key Lemma)). Th.1.2 is proved by using this fact. It is remarked if \(p= 2\), there is a simpler proof of Th. l.2 (Remark 2.6). To prove Th.1.3. for a pair of maps \(\Phi,\Psi:\Sigma\to{\mathfrak g}\), the \(\lambda\)-vortex equation \[ \partial_s u+ L_u\Phi+ J(\partial_t u+ L_u\Psi)= 0,\quad (\partial_s\Psi- \partial_t\Phi+ [\Phi,\Psi])+ \lambda^2\mu\circ u= 0, \] which is equivalent to the vortex equation, is introduced. Let \(\Sigma\) be \(\{s+ it\in\mathbb{C}\mid s\geq 0\}/i\mathbb{Z}\cong\mathbb{R}\times S^1\), and \(U\subseteq\Sigma\). The energy density and the energy of a solution \(w= (u,\Phi,\Psi)\) of the \(\lambda\)-vortex equation are defined by \[ \widetilde e_w= |\partial_s u+ L_u\Phi|^2\text{ and } E(w,U)= \int_U\widetilde e_w ds\wedge dt, \] respectively. If \(w\) is a solution of the \(\lambda\)-vortex equation and small, it is shown that \[ E(w,\Sigma)=-{\mathcal A}((u,\Psi)(s_+,\cdot))+{\mathcal A}((u,\Psi)(s_-,\cdot)), \] (Prop.3.1). Then, assuming hypothesis \(({\mathbf H})\), the estimates \[ \widetilde e_w(z)\leq{32\over\pi} E(w,B_{{1\over 2}}(z)),\quad {d\over ds} E(s)\leq -(4\pi- \varepsilon) E(s), \] are shown (Lemma 3.3, Claim 1 in \S3.2). Th.1.3 is proved by these facts and Th.1.2. In the Appendix, the inequality \(d(1,h)\leq C\| F_A\|_{L^\infty(X)}\ell(x)^2\) for the holonomy of the connection is proved.