The solution to that PDE is then used as a phase function of a parametrix U(t) to the time evolution e itH . More precisely, Hörmander defines U(t)φ (x) := e i[x·ξ−S(t,ξ)] φ(ξ)dξ (1.9)for Schwartz functions φ ∈ S(R d ) with supp( φ) ⊂ R d \ {0}. A novel feature of the scattering theory of (1.2) is the introduction of a time-dependent amplitude function in the parametrix. Since D 2 V(x) does not decay like |x| −2−ε , we can easily show that the parametrix (1.9) does not apply here. Rather, we definewith a suitable a(t, ξ).Generally speaking, the gain of 1/4 in terms of the conditions on the potential is achieved throughout this paper by systematically exploiting averaging in the potential.As an example, consider the Hamilton-Jacobi PDE (1.8). It is well known that initial conditions can be chosen such that DS(t, ξ) is a trajectory that escapes at a linear rate to infinity. Therefore, any integral like 2T T V DS(t, ξ) dt (1.11)is going to be significantly smaller (roughly by T −1/4 if β = 1/2) than the one with absolute values inside due to the mean-zero assumption on the randomness. Similarly, integrating the ODE (1.7) will lead to the expression ∞ t ∇V(x(s))ds. In contrast to the deterministic theory that only exploits the size of ∇V(x(s)) we need to invoke cancellations.It is well known that the existence of (modified) wave operators has strong spectral implications. More precisely, Ω + is an isometry that gives a unitary equivalence between and the restriction of H to a subspace of L 2 (R d ). The Weyl criteria implies that even the deterministic essential spectrum σ ess (H) of the Schrödinger operator with potential (1.2) coincides with the essential spectrum of the unperturbed operator −∆,. Therefore, we obtain that with probability onewhere σ a.c. denotes the absolutely continuous spectrum of H. It seems natural to believe at Cornell University Library on July 23, 2015 http://imrn.oxfordjournals.org/ Downloaded from at Cornell University Library on July 23, 2015 http://imrn.oxfordjournals.org/ Downloaded from Classical and Quantum Scattering 247 Our interest in this question arose in connection with a conjecture about square integrable potentials, see Simon's review [20]: if14) then − + V has a.c. spectrum essentially supported on [0, ∞). That conjecture would imply that the 1/2-model should have a.c. spectrum essentially supported on [0, ∞) for any realization of the ω n . In one dimension this has been shown by Deift and Killip [8]. In a series of several papers, Christ and Kiselev [6, 7] had previously settled the one-dimensional problem for potentials satisfying ∞ −∞ |V(x)| p dx < ∞, where 1 ≤ p < 2. Moreover, Christ and Kiselev have recently constructed wave operators in the onedimensional case under basically optimal conditions [5]. Another interesting approach to a.c. spectrum under optimal conditions is Molchanov, Novitskii, and Vainberg [17].This paper is organized as follows: in Section 2, we construct classical scattering trajectories for the 1/2-model with probability one, see Proposition 2.3 (th...