No label defined (Q4385487)

This paper is a study of the initial-boundary value problem \[ u_t = \Delta u \text{ in } \Omega\times (0,\infty),\;au_y + bu = 0 \text{ on }\partial\Omega\times(0,\infty),\;u(\cdot,0) = u_0 \text{ in } \Omega, \] where \(\Omega\) is the half-plane \(\{(x,y): y < 0\}\), \(u_y\) denotes the derivative with respect to \(y\), and \(u_0\) is a given \(C^\infty\) function. The main assumptions on the functions \(a\) and \(b\) are that \(a^2+b^2 \equiv 1\) and that they are in \(C^\infty\). If \(a\) and \(b\) are nonnegative, then \textit{S. Itò} [Jap. J. Math. 27, 55-102 (1957; Zbl 0092.31101)] showed that this initial boundary value problem has a unique solution and that this solution is \(C^\infty\) provided the obvious compatibility condition is satisfied between the initial and boundary data. In the present setting, the nonnegativity of \(a\) and \(b\) is replaced by the boundedness of these functions and all their derivatives. The main result is that the unique solvability of this problem implies that, if \(a\) vanishes to some finite order at each of its zeros, then \(a\) does not change sign anywhere. Assuming that \(a \geq 0\), it also follows that \(b=1\) wherever \(a=0\). (The basic equation \(a^2+b^2=1\) implies, of course, that \(b=\pm 1\) wherever \(a=0\).) The proof is by contradiction. The author shows that if \(a\) vanishes only to finite order, and if either the vanishing order is odd or \(b =-1\) at a point where \(a\) vanishes, then there is a function \(u_0\) for which the problem has no solution. The construction of this initial function relies on a delicate analysis using the truncated Laplace transform and a result of \textit{T. Komura} [J. Funct. Anal. 2, 258-296 (1968; Zbl 0172.40701)] on infinitesimal generators of semigroups in Fréchet spaces. For the entire collection see [Zbl 0884.00040].