On the formation of shocks for quasilinear wave equations (Q510169)

In their long paper, the authors present a situation of shock formation in the case of the 3D quasilinear wave equation: \(-(1+3G^{\prime \prime }(0)(\partial _{t}\phi )^{2})\partial _{t}^{2}\phi +\Delta \phi =0\), where \( G^{\prime \prime }(0)\) is a non-zero constant, for which they look for a solution \(\phi \in C^{\infty }(\mathbb{R}_{t}\times \mathbb{R}_{x}^{3}; \mathbb{R})\). The main result proves that if a given pair \((\phi _{1},\phi _{2})\in C^{\infty }((0,1]:\mathbb{S}^{2})\) of seed data satisfies \( G^{\prime \prime }(0)\partial _{r}\phi _{1}(r,\theta )\phi _{1}(r,\theta )\leq -1/6\) for at least one \((r,\theta )\in (0,1]\times \mathbb{S}^{2}\) and if the initial data are taken equal to the no-outgoing-radiation initial data, there exists \(\delta _{0}\) which only depends on the seed data \((\phi _{1},\phi _{2})\) such that for every \(\delta <\delta _{0}\) shocks form for the corresponding solution \(\phi \) of the above equation before \(t=-1\). Moreover \(\phi \) will no longer be smooth in the sense that \(\nabla \partial _{t}\phi \) blows up when approaching the shocks, while \(\phi \) and \(\partial \phi \) remain bounded. The authors first recall the Lagrangian formulation leading to the above equation, considering the perturbation of the Lagrangian density of linear waves \(L=-\frac{1}{2}G((\partial _{t}\phi )^{2})+\frac{1}{2}\left| \nabla \phi \right| ^{2}\) where \(G\) is a smooth function on \(\mathbb{R}\). For the proof of the main result, the authors use the short pulse method introduced by \textit{D. Christodoulou} [The formation of shocks in 3-dimensional fluids. Zürich: European Mathematical Society Publishing House (2007; Zbl 1117.35001)] and developed by \textit{S. Klainerman} and \textit{I. Rodnianski} [Acta Math. 208, No. 2, 211--333 (2012; Zbl 1246.83028)]. They rewrite the above equation as \( \square _{m}\phi =-3G^{\prime \prime }(0)(\partial _{t}\phi )^{2}\partial _{t}^{2}\phi \) where \(\square _{m}\) is the wave operator on the Minkowski space time and they observe that the nonlinearity is cubic. This allows considering only the case of large initial data with amplitude \(\sqrt{\delta }\) and which are supported in the annulus \(\Sigma _{-r_{0}}^{\delta }:=\{x\in \Sigma _{-r_{0}}\mid r_{0}\leq r(x)\leq r_{0}+\delta \}\), where \( \Sigma _{t}\) denotes the level set of the time function \(t\) in the Minkowski spacetime. Here \(r_{0}=2\). The authors then observe that the Lagrangian, hence the above equation, is invariant under time translation and isometries of \(\mathbb{R}^{3}\). They can thus linearize the equation. They prove energy estimates for linearized wave equations \(\square _{\widetilde{g}}\phi =\rho \), where \(\rho \) is a smooth function, that they link to a control of the underlying geometry. They prove a bootstrap argument. They then analyze the inverse density function \(\mu \) of the characteristic surfaces, proving asymptotic expansions for \(\mu \) and its derivatives. The main parts of the paper are devoted to the derivation of estimates on different physical or geometrical quantities associated to this equation. The final main tool is a descent scheme which expresses the energy identity in terms of the product between an integral \(\int_{-r_{0}}^{t}\mu ^{-(a+1)}\frac{\partial \mu }{ \partial t}d\tau \) for large \(a\) and \(\widetilde{E}(t)=\sup_{\tau \leq t}\mu ^{a}E(\tau )\). The authors here prove that the right-hand side of this inequality can be absorbed by its left-hand side. The verification of the different steps of the proof are given with lengthy details.