Page 415 -
P. 415

7.2 Coordinate Changes and Pseudohomogeneous Kernels  399

                           and
                                           1                        α
                               b α (x )  :=    p.f.  k(x, x − y)(y − x) dy| x=Ψ(x   )

                                           α!
                                                  Ω

                                                                                α





                                           − p.f.  k(x ,x − y )J(y ) Ψ(y ) − Ψ(x )  dy     . (7.2.22)



                                                Ω
                           For m< 0, the differential operator terms with the coefficients a α and b α do
                           not appear and






                                            A u(x )=  k(x ,y − x )J(y ) u(y )dy .      (7.2.23)




                                                    Ω
                              In comparison with A in (7.2.14), we note that for m ≥ 0 the coefficients

                           of the differential operator part contain the extra terms b α (x ), which for
                           m ∈ IN 0 , in general, do not vanish. In the special case when m =0, see
                           Mikhlin and Pr¨ossdorf [215, Formulae (8) p.223 and (11) p.226]. This is a
                           very important difference from the coordinate transformation for operators
                           with weakly singular or regular kernel functions as in (7.2.23). Fore more
                           general m see also Sellier [281].
                              In the case 0 <m  ∈ IN 0 we shall show that the b α (x ) in (7.2.22) will

                           vanish. For this purpose we first need the following lemma concerning the
                           coordinate transformation Ψ.

                           Lemma 7.2.3. Let   ε (ω) denote the distance between x = Ψ(x ) and y =

                           Ψ(y ) where |y −x | = ε> 0 and y = x+  ε (ω)ω is the image of the sphere in


                           terms of polar coordinates (  ε ,ω) about x. Then, for sufficiently small ε> 0
                           and Ψ a C –diffeomorphism, one admits the asymptotic representation
                                     ∞
                                                       N
                                                               k      N+1
                                                 ε (ω)=   c k (ω)ε + O(ε  )            (7.2.24)
                                                      k=1
                           for any N ∈ IN. The coefficients c k (ω) satisfy the parity conditions
                                                  c k (−ω)=(−1) k+1 c k (ω) .          (7.2.25)

                                             ∞

                           Proof: With the C –diffeomorphism Φ(x)= x ,Φ(y)= y ,inverse to Ψ,

                           the Taylor expansion about x reads
                                                            1
                                                               α
                                                                      α |α|


                                              y − x =        D Φ(x)ω      .
                                                           α!
                                                       |α|≥1
                           Hence,
                                                         α     α       β    β
                                           2
                               2

                              ε = |y − x | =           D Φ(x)ω    · D Φ(x)ω     |α|+|β|  (7.2.26)
                                                                               ε
                                             |α|≥1,|β|≥1
   410   411   412   413   414   415   416   417   418   419   420