satisfying, for t = 0, the initial condition ~/(x, 0) = 0(-x)~bo(X), where 0 is the Heaviside function, and ~bo(X) = A(x)e is0(x)/h. Here V, S o, and A are infinitely smooth functions, V belonging to the Schwarz space, V and S O real-valued functions, and A a compactly supported function: supp A C (- a, a), a > 0 (at the end of the article we also consider the case of a semi-infinite step), and h > 0 is a small parameter. In the article we will be investigating what form is assumed by q~(x, t) when t = t* > 0. If we take fro(X) as the initial condition, the asymptote of the solution is well known and may be constructed using the Maslov canonical operator [3]. In the present situation this method cannot be applied directly, since the initial condition is represented by a discontinuous function. Suppose that ~b(x, t, k) is the solution of the initial-value problem for the Schroedinger equation with initial condition if(x, 0, k) = ~o(x)e ikx/h. To solve the problem we represent the function ~(x, 0) in the form of a superposition of waves ~(x, 0, k). By the linearity of the Schroedinger equation ,I~(x, t*) may be represented in the form of the same superposition of waves ~b(x, t*, k). Therefore, once we know the asymptote of the function ~b(x, T*,k) as h --, +0, we will be able to determine the asymptote of ff'(x, t*) by computing a corresponding rapidly oscillating integral. Moreover, we fix some k o > 0 and limit ourself to values k E [-ko, ko]. Finally, we will have to estimate the error of the inexact computation of ~b(x, t*, k) and the error due to ignoring k ~ [-k 0, ko]. Let us first recall the technique for constructing the asymptote of ~b(x, t*, k) using the canonical operator. We consider a Hamiltonian system with Hamiltonian H(q, p) = p2/2 + V(q); the corresponding Lagrangian K(q, r = dl2/2-V(q). Suppose that over time t* a point in the phase space (xo, P0), where Po = So'(Xo) + k, turns into the point (X, P)(x o, k), and J(x o, k) = Xxo(X o, k). We denote by I the closed interval with endpoints X(_a, 0). Let us start by limiting ourself to the case in which J(x o, 0) ~ 0 for all x0 E I-a, a].Then there exist k0 > 0 and some neighborhood U of the closed interval [-a, a] such that for all