Controllability of a string under tension (Q1422480)

The most common studies of vibration of an elastic string with fixed end points are such that transverse oscillations are caused by a transverse load. Here the authors assume that the tension is a sum of a constant term and a variable load, and study controllability. The first author followed in the past a very effective technique known as method of moments, which is also used here. The following equation is introduced: \[ \rho(x) y_{tt}= (P(y)_x)_x+ F(x,t),\quad t> 0,\quad 0< x< \ell,\tag{1} \] where \(F\) is an external transverse force, and it is assumed, to make life easy, that the variables can be separated, thus: \(F(x,t)= f(x)g(t)\). Assuming that \(p\) varies ``slowly'' (the authors define a bound on the variation claiming that it is ``slow'' below that bound. Also oscillations are assumed to be ``small'', thus the equation (1) is replaced by the one-dimensional wave equation: \(\rho(x)y_{tt}= P(t)y_{xx}+ f(t) g(x)\). Here \(f(t)\) is regarded as the control. The authors prove uniqueness of solutions by decomposing them into Fourier (sine) series in \(x\) with variable coefficients \(c_n(t)\). They do not say where the proof of the theorem 1 (uniqueness) ends. It really ends with the display of the second-order ordinary differential equation satisfied by \(c_n(t)\), with initial conditions \(c_n(0)= c^0_n\), \(\partial c_n/\partial t(0)= c^1_n\), which of course has a unique solution. Then a word about a bound on sums of these solutions would convince everybody that uniqueness has been proved. That bound is produced later as a sufficient and necessary condition for controllability (in equations labelled A1). The authors point out relations between a crucial time \(T^*\) in which the system is brought to rest and the convergence of infinite sums, which look like relations between \(L_2\) and \(\ell_2\) summability. The authors comment that similar results are available for \(L_2\) controls, which enable us to reach various states starting with zero initial data. The reviewer would like to point out that serious difficulties may be encountered in these so-called excitation problems. Solutions are generally unstable. The authors handle a problem that has been largely ignored in studies of classical vibrations of strings or membranes, including control of such vibrations initiated in 1964 by D. L. Russell. The authors should be commended for making this effort. Aside from the main theme of this article, the authors provide a badly needed discussion of the Liouville transform, which has been useful in some ODE applications, but its possible generalizations were hardly ever attempted in higher dimensions to the best of the reviewer's knowledge.