Page 339 -
P. 339

6.1 Basic Theory of Pseudodifferential Operators  323

                              Since the Schwartz kernel of A Φ,0 satisfies

                                                                            x − y


                                           (x ,x − y )= K Φ ∗ A 0 Φ ∗(x ,x − y )ψ





                                      K A Φ,0
                                                                              ε
                           and Φ ∗ A 0 Φ is properly supported, so is A Φ,0 . Now Theorem 6.1.11 can be
                                     ∗
                           applied to A Φ,0 .
                                                                         −1

                              Since the function v(y ) 1−ψ  x −y  det  ∂φ   vanishes in the vicin-

                                                            ε         ∂y


                           ity of x = y identically, we can proceed in the same manner as in the proof
                           of Theorem 6.1.2 and obtain


                                                                          x − y

                                 Rv(x )  =  (2π) −n     e i(x−y)·ξ v(y ) 1 − ψ     ×

                                                                             ε
                                                    n
                                                  IR Ω
                                                              ∂φ           N
                                                                  −1

                                                       × det         (−∆ ξ ) a(x, ξ) dy dξ .
                                                              ∂y
                           Here, for N> m + n + 1 we may interchange the order of integration and
                                                                           ∞
                           find that R is a smoothing integral operator with C –kernel. Therefore,
                                                  n
                                         m
                           Φ ∗ A 0 Φ ∈ OPS (Ω × IR ) and the asymptotic formula (6.1.45) is a conse-

                                 ∗
                           quence of Theorem 6.1.11. This completes the proof.
                              The transformed symbol in (6.1.45) is not very practical from the compu-
                           tational point of view since one needs the explicit form of the matrix Ξ(x ,y )


                           in (6.1.47) and its inverse. Alternatively, one may use the following asymp-
                           totic expansion of the symbol.
                           Lemma 6.1.17. Under the same conditions as in Theorem 6.1.16, the com-
                           plete symbol class of A Φ also has the asymptotic expansion
                                            1      ∂    α                ∂    α
                                                                              iξ ·r
                             a Φ (x ,ξ ) ∼           a(x, ξ)         − i     e         (6.1.50)
                                           α!   ∂ξ         |            ∂z        | z=x
                                                                 ξ
                                        0≤α                 ξ= ∂Φ
                                                              ∂x
                           where x = Φ(x) and

                                                               ∂Φ
                                              r = Φ(z) − Φ(x) −  (x)(z − x) .          (6.1.51)
                                                               ∂x
                           Moreover,

                                   ∂     iξ ·r
                                       α             0                          for |α| =1 ,

                                − i     e    | z=x =                α
                                   ∂z                (−i) |α|−1  ∂  Φ(x) · ξ    for |α|≥ 2 .
                                                                ∂x
                                                                                       (6.1.52)
                           Proof: In view of Theorem 6.1.16 it suffices to consider properly supported
                           operators A 0 and A Φ,0 . Therefore the symbol class to A Φ,0 can be found by
                           using formula (6.1.29), i.e.,
   334   335   336   337   338   339   340   341   342   343   344