IntroductionThe direct and inverse scattering problems for the n X n system (generalized the AKNS system) d -m + z [ J , m ] = q ( x ) m dx were solved by Beals and Coifman [3] for J diagonal with distinct eigenvalues and for potentials q, off-diagonal and generic. Partial solutions to the problems may be found in [13] and [7]. Certain bounded solutions m are piecewise meromorphic on the Riemann sphere relative to a contour 2, consisting of finitely many straight lines passing through the origin determined by J. Generic potentials are defined such that m extends continuously to the boundary from each component of C\Z (this implies that m has only finitely many poles) and only has simple poles of a special type. The Beals-Coifman scattering data are then defined by the multiplicative jump of m across I: and certain normalization constants at the poles. Following [8], we call z' E Z a spectral singularity of the scattering problem of q if the limit of m as z -+ z' from one of the components of C\X does not exist. Even for the potentials of Schwartz class, the poles may accumulate to spectral singularities and the spectral singularities themselves may accumulate (see Example 3.3.16). These problems (after the significant progress made by [3]) constitute the major obstacle to the complete solution of the direct and inverse scattering problems for first-order systems (including the 2 X 2 systems).When the inverse scattering method is applied in solving the Cauchy problem for the nonlinear evolution equations, the time evolution can be obtained by inverse scattering transform only if the initial q happens to be generic. Indeed, an a priori determination of all the generic potentials has not been obtained. This unpleasant situation apparently should not be ascribed to the evolution equations but to a deficiency of the existing theory of direct and inverse scattering transforms.In this paper, we find a way to overcome the deficiency by dropping the boundedness requirement for m. In our method, the bounded solutions of (1.1) are only required in a neighborhood of z = 00 (which is the only pole of zJ).According to [3], for q E L', the bounded solutions always exist there. Away