Asymptotics of the Teichmüller harmonic map flow (Q2437430)

The authors consider the Teichmüller harmonic flow, introduced by the first two authors in 2012 [\textit{M. Rupflin} and \textit{P. M. Topping}, ``Flowing maps to minimal surfaces'', Preprint, \url{arXiv:1205.6298}; \textit{M. Rupflin}, ``Flowing maps to minimal surfaces: Existence and uniqueness of solutions'', Preprint, \url{arXiv:1205.6982}], \[ \frac{\partial u}{\partial t}=\tau_g(u),\quad\quad \frac{\partial g}{\partial t}=\frac{\eta^2}{4}\text{Re}\big(P_g(\Phi(u,g))\big), \] where \(\eta\) is a fixed positive constant, \((M,g)\) is a closed Riemannian surface of genus \(\gamma\geq 2\) and hyperbolic metric \(g\), \(u:M\to N\) is a map with values on a compact Riemannian manifold \((N,G)\) and \(\tau_g(u)\) its tension field, \(\Phi(u,g)\) is the quadratic Hopf differential and \(P_g(\Phi(u,g))\) its \(L^2\)-orthogonal projection onto the space of holomorphic quadratic differentials. Then the harmonic map energy \(E(u,g)\) decays according to \[ \frac{dE}{dt}= -\int_M |\tau_g(u)|^2 +\Big(\frac{\eta}{4}\Big)^2\big|\text{Re}\big(P_g(\Phi(u,g))\big)\big|^2. \] If the initial data is a sphere, the flow is just the usual Eells-Sampson harmonic map flow. In the previous mentioned works, it is shown that given initial data \((u_0,g_0)\in H^1(M,N)\times \mathcal{M}_{-1}\), a maximal solution \((u(t), g(t))\) of the flow exists in \(H^1_{\text{loc}}(M\times [0, T), N)\times C^0([0, T), \mathcal{M}_{-1})\), in a weak sense. Moreover, if \(T\) is finite, the flow of metrics degenerates in moduli space, and the length \(l(g(t))\) of the shortest closed geodesic of \((M, g(t))\) tends to zero when \(t\to T\). On the other hand, if such degeneration does not occur, then, whether \(T\) is finite or infinite, \(u(t)\) subconverges either to a constant map or to a brunched minimal immersion with the same action on the fundamental group of \(M\) as \(u_0\). In this paper, the authors study the case when the metric degenerates at \(T=\infty\), i.e. \(l(g(t))\to 0\) when \(t\to \infty\). In Theorem 1.1 the authors prove that in this case there exists a sequence \(t_i\to \infty\) and a hyperbolic punctured surface \((\Sigma, h, c)\) with \(2k\) punctures for some \(1\leq k \leq 3(\gamma-1)\), where \(c\) is the complex structure, such that the evolved surfaces \((M, g(t_i), c(t_i))\) converge to \((\Sigma, h, c)\) by collapsing \(k\) simple closed geodesics, whose complementary set in \(M\) is diffeomorphic to \(\Sigma\) by a a diffeomorphism \(f_i\) such that \(f^*_ig(t_i)\) converges to \(h\), \(f^*_ic(t_i)\) converges to \(c\), and \(u(t_i)\circ f_i\) converges to a limit map \(u_{\infty}:\Sigma \to N\) in \(H^1_{\text{loc}}(\Sigma)\) and weakly in \(H^2\) away from a finite set \(S\subset \Sigma\), as well in some suitable Sobolev space. This map \(u_{\infty}\) either is constant or by filling the punctures it extends to a branched minimal immersion. The proof relies on the Deligne-Munford compactness theorem for such degenerating sequences of complex structures and hyperbolic metrics and an inspection of a sequence of suitable subspaces of the space \(\mathcal{H}(M, g_i)\) of \(L^1\) holomorphic quadratic differentials on the evolved surfaces, that decay along each degenerating collar, and are uniformly \(L^p(\Sigma)\)-bounded, for any \(p\geq 1\), and will converge to \(\mathcal{H}(\Sigma, h)\) in certain sense.