Page 323 -
P. 323

6.1 Basic Theory of Pseudodifferential Operators  307

                           Remark 6.1.2: The representation (6.1.14) will be also expressed in terms
                           of Hadamard’s finite part integrals later on.

                           Remark 6.1.3: In the theorem we use the identity (6.1.11) to ensure that
                           the integral (6.1.12) decays at infinity with sufficiently high order, whereas
                           for ξ = 0, we employ the regularization given by the Hadamard’s finite part
                           integrals. In fact, the trick (6.1.11) used in (6.1.12) leads to the definition of
                           the oscillatory integrals which allows to define pseudodifferential operators
                           of order m ≥−n (see e.g. Hadamard [117], H¨ormander [131, I p. 238], Wloka
                           et al [323]).


                           Proof:
                           i) n+m< 0: In this case, the integral has a weakly singular kernel, hence,

                                        Au(x)=(2π)  −n   u(y)   a(x, ξ)e i(x−y)·ξ dξdy
                                                      IR n   IR n
                           and

                                          k(x, x − y)=(2π) −n  a(x, ξ)e i(x−y)·ξ dξ .
                                                            IR n
                              For the regularity, we begin with the identity (6.1.11) setting z = x − y.
                                                             n
                                      N
                           Since (−∆ ξ ) a(x, ξ) ∈ S S S m−2N (Ω × IR ) integration by parts yields

                                                  −2N     −n          N         i(x−y)·ξ
                                 k(x, x − y)= |x − y|  (2π)     (−∆ ξ ) a(x, ξ) e     dξ .
                                                            IR n
                              This representation can be differentiated 2N times with respect to x and
                           y for x  = y. Since N ∈ IN is arbitrary, k is C  ∞  for x  = y and continuous
                           for x = y.

                           ii) n + m ≥ 0: Here, with

                                                              1       α  α
                                         v(y):= u(y) −          (x − y) D u(x) ,
                                                              α!
                                                        |α|≤2N
                           (6.1.14) has the form
                                                           α
                                    Au(x)  =        c α (x)D u(x)
                                              |α|≤2N

                                               +(2π) −n     v(y)ψ(y)e i(x−y)·ξ dya(x, ξ)dξ .
                                                       IR n  Ω
                           Note that the latter integral exists since the inner integral defines a C  ∞
                           function which decays faster than any power of ξ due to the Palay–Wiener–
                           Schwartz theorem 3.1.3. With (6.1.11) we obtain
   318   319   320   321   322   323   324   325   326   327   328