The Lax-Mizohata theorem for nonlinear gauge invariant equations (Q1347438)

The author reviews the Lax-Mizohata (L-M) theorem for linear equations: \[ D^m_tu+ \sum_{j+|\alpha|\leq m, j\leq m} a_{j,\alpha} (t,x)D^j_t,\;D_x^\alpha u=f(t,x), \] with coefficients in \(C^\infty ([0,T] \times\mathbb{R}^n)\). Here \(D_t=-i\partial/ \partial t\), \(D^\alpha_x=(-i\partial/ \partial x_1)^{\alpha_1}+\cdots +(-i\partial/ \partial x_n)^{\alpha_n}\). The characteristics are solutions of the system of equations: \(\tau^m+ \displaystyle\sum_{j+|\alpha|\leq m, j\leq m}a_{j,\alpha} (t,x)\tau^m \xi^\alpha =0\). The Lax-Mizohata theorem states that if the Cauchy problem is \(C^\infty\) well posed, then all characteristics are real valued. The author introduces a nonlinear system of equations: \[ \partial_t^mu+ F\bigl(t,\{\partial^j_t \partial_x^\alpha u\},\{\partial^j_t \partial^\alpha_x \overline u\}\bigr)=0, \tag{1} \] with the same restrictions on values \(i,j,\alpha\). Here \(\overline u\) denotes the complex conjugate of \(u\). The nonlinear term is continuous with respect to \(t\) and to vectors \(\{\partial^j_t \partial_x^\alpha\}\), and their complex conjugates. The standard definition of stability is extended to this setting. The stability with respect to the trivial solution demands that if \(u\equiv 0\) is stable then in this setting \(F(t,0,0)\equiv 0\) for all values of \(t\). The author offers examples when the L-M theorem fails if the system is not hyperbolic, even if it is weakly hyperbolic. For a system that is invariant with respect to the gauge transformation the author proves that stability of solution of the Cauchy problem implies that all roots of the characteristic equation are real. Moreover if the second derivatives are Lipschitz continuous then a gauge invariance and Cauchy stability in the cone of dependence implies that all roots of the (matrix) characteristic equation are real. Proofs by contradiction involve reducing the system to a system of ordinary differential equations. In fact the author demonstrates in this paper that it is not linearity, but gauge invariance that is crucial for the L-M theorem to hold.