Limiting configurations for the SU(1,2)-Hitchin equation

From MaRDI portal
Publication:6901297





\textit{N. J. Hitchin} [Proc. Lond. Math. Soc. (3) 55, 59--126 (1987; Zbl 0634.53045)] introduced the self-duality equations on a compact Riemann surface and the associated moduli spaces of Higgs bundles, exhibiting the Hitchin fibration and its role in complex integrable systems. \textit{S. K. Donaldson} [Proc. Lond. Math. Soc. (3) 55, 127--131 (1987; Zbl 0634.53046)] related the same equations to harmonic maps, while \textit{K. Corlette} [J. Differ. Geom. 28, No. 3, 361--382 (1988; Zbl 0676.58007)] and \textit{C. T. Simpson} [Publ. Math., Inst. Hautes Étud. Sci. 75, 5--95 (1992; Zbl 0814.32003)] established the analytic and algebro-geometric foundations of nonabelian Hodge theory, identifying polystable Higgs bundles with reductive surface-group representations via harmonic metrics. A natural analytic problem, motivated both by the geometry at infinity of Higgs moduli and by WKB-type considerations, is to understand the behaviour of the harmonic metric solving Hitchin's equations under the scaling \(\Phi\mapsto t\Phi\) as \(t\to\infty\). For generically regular semisimple Higgs fields, \textit{R. Mazzeo} et al. [Duke Math. J. 165, No. 12, 2227--2271 (2016; Zbl 1352.53018)] developed the framework of \textit{limiting configurations} and a gluing theory describing decoupling away from the discriminant and controlled singularities at branch points. In the presence of singular/irregular local models, one needs wild harmonic-bundle technology as developed by \textit{O. Biquard} and \textit{P. Boalch} [Compos. Math. 140, No. 1, 179--204 (2004; Zbl 1051.53019)] and \textit{T. Mochizuki} [Wild harmonic bundles and wild pure twistor \(D\)-modules. Paris: Société Mathématique de France (SMF) (2011; Zbl 1245.32001)]. The paper under review adapts and extends this analytic circle of ideas to the real form \(\mathrm{SU}(1,2)\), where the Higgs data naturally factor through a quadratic differential and where new local behaviours occur.\N\NLet \(X\) be a compact connected Riemann surface of genus \(g(X)\ge 2\) with canonical bundle \(K_X\). An \(\mathrm{SU}(1,2)\)-Higgs bundle may be presented as a triple \((F,\beta,\gamma)\) where \(F\) is a rank \(2\) holomorphic bundle, \(L=\det(F)^{*}\) is a line bundle, and\N\[\N\beta \in H^0\bigl(X,\Hom(L\otimes K_X^{-1},F)\bigr),\qquad \gamma \in H^0\bigl(X,\Hom(F,L\otimes K_X)\bigr),\N\]\Nsubject to the usual stability condition. The Hitchin map sends \((F,\beta,\gamma)\) to the quadratic differential \( q=\gamma\circ\beta \in H^0(X,K_X^2). \) For \(t>0\) consider the scaled family \((F,t\beta,t\gamma)\) and let \(h_t\) be the unique Hermitian metric on \(F\) solving the \(\mathrm{SU}(1,2)\) Hitchin equation (equivalently the harmonic-metric equation for the associated representation). The paper assumes that \(q\) has only simple zeros, with divisor \(D=\sum_j p_j\).\N\NThe main theorem constructs a canonical limiting configuration on \(X\setminus D\) and proves convergence of the normalized harmonic metrics toward it. The key structural input is that the spectral data of \((F,\beta,\gamma)\) with fixed \((L,q)\) can be described by a Hecke modification \( \iota:F\hookrightarrow V \) of the rank-\(2\) bundle \( V=L^{-2}K_X \oplus LK_X \) at the divisor \(D\). Fix a metric \(h_{K_X}\) on \(K_X\) normalized so that \(|q|_{h_{K_X}^2}\equiv 1\) on \(X\setminus D\). The limiting configuration is given by a singular metric on \(V\) of the form\N\[\Nh_{\infty,V}= \bigl(h_{L,\infty}^{-2}h_{K_X}\bigr)\ \oplus\ \bigl(h_{L,\infty}h_{K_X}\bigr),\N\]\Nwhere \(h_{L,\infty}\) is a harmonic metric on \(L|_{X\setminus D}\) adapted to a parabolic structure along \(D\), and the limiting metric on \(F\) is its pullback \( h_\infty=\iota^{*}h_{\infty,V}. \) The theorem asserts that, after applying an explicit diagonal gauge normalization (needed because the family \((t\beta,t\gamma)\) introduces a preferred scaling), one has \(C^\infty\) convergence on compact subsets: for every \(X_0\Subset X\setminus D\) and every \(k\ge 0\) there exists \(C_k\) such that \( \|\psi_t-\mathrm{Id}\|_{C^k(X_0)}\le C_k/(\log t) \), where the normalized metric \(\widetilde h_t\) satisfies \(\widetilde h_t=h_\infty\cdot \psi_t\). In particular, \(h_t\) converges to \(h_\infty\) on \(X\setminus D\) modulo the prescribed rescaling, and \(h_\infty\) solves the decoupled system on \(X\setminus D\) obtained by imposing the vanishing of the commutator term in Hitchin's equations.\N\NA central feature is an explicit rule for parabolic weights at the points of \(D\). The divisor \(D\) decomposes into three disjoint subsets according to the local behaviour of the factorization \(q=\gamma\circ\beta\): points where \(\beta(p_j)=0\), points where \(\gamma(p_j)=0\), and the remaining points where neither map vanishes. The limiting parabolic weight \(\lambda_{\infty,j}\) of \(h_{L,\infty}\) at \(p_j\) equals \(+1/4\) in the first case and \(-1/4\) in the second case; on the remaining points the weights are all equal to a constant \(\lambda_\infty\in(-1/4,1/4)\) determined by the global constraint \(\sum_j \lambda_{\infty,j}=-\deg(L)\). The paper also explains how the three local types correspond to different normal forms for the Higgs field in frames compatible with the Hecke modification, making the sign \(\pm 1/4\) canonical in the first two cases.\N\NThe proof follows the general MSWW blueprint of gluing limiting configurations, but new analytic ingredients are required for \(\mathrm{SU}(1,2)\) because not all local models admit an explicit rotationally symmetric description. The construction begins by producing local model solutions near each \(p_j\in D\) on a punctured disk. When \(\beta(p_j)=0\) or \(\gamma(p_j)=0\), the local equations reduce (after symmetry reduction) to an ordinary differential equation of Painlevé~III type, yielding explicit 1-parameter families of rotationally symmetric solutions and identifying the two special weights \(\pm 1/4\). For points of the third type, the needed models are obtained using wild harmonic bundle theory: one appeals to existence and uniqueness results for irregular connections/harmonic metrics to produce a symmetric ``wild local model depending on a real parameter \(\lambda\in(-1/4,1/4)\).\N\NOne then glues the local models to the global decoupled configuration away from \(D\) to obtain an approximate solution \(h_t^{\mathrm{app}}\) whose Hitchin-equation defect is small. A major technical issue is that, in the wild (third-type) case, gluing with exponentially small error requires matching not only the leading asymptotic term \(\lambda\log|z|\) but also the next constant term in the model expansion; this constant is uniquely determined by \(\lambda\) and the paper proves its continuous dependence on \(\lambda\). Using this continuity, the author selects a \(t\)-dependent choice of weights \(\lambda(t)\) by a fixed-point argument on a polytope of admissible weight data and shows \(\lambda(t)\to\lambda_\infty\) as \(t\to\infty\).\N\NFinally, one perturbs \(h_t^{\mathrm{app}}\) to the genuine solution \(h_t\) via a contraction mapping argument. Writing \(h_t=h_t^{\mathrm{app}}e^{u}\) with a skew-Hermitian endomorphism \(u\), the Hitchin operator is expanded as \( \mathcal{H}_t(0)+L_t u+Q_t(u), \) where \(L_t\) is the linearization and \(Q_t\) is quadratic. The core analytic estimate is a \(t\)-dependent lower bound for \(L_t\) on suitable weighted spaces. Unlike the rank 2 situation where one can often arrange a uniform spectral gap, an explicit diagonal endomorphism behaves as an approximate kernel on \(X\setminus(\text{small disks})\) but cannot extend across the Hecke modification points; this forces a weaker bound of order \((\log t)^{-1}\) and ultimately explains the \((\log t)^{-1}\) convergence rate in the theorem. With these bounds and standard quadratic estimates for \(Q_t\), the contraction mapping principle produces \(u\) and hence the exact solution, completing the gluing construction and establishing convergence to the limiting configuration with the stated parabolic weights.



Cites work









This page was built for publication: Limiting configurations for the \(\mathrm{SU}(1,2)\)-Hitchin equation

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6901297)