Page 346 -
P. 346

330    6. Introduction to Pseudodifferential Operators

                           and all other coefficients are zero. Then the symbol matrix (6.2.12) for (2.3.1)
                           reads
                                                    ⎛                       ⎞
                                                         2
                                                     µ|ξ|    0     0    −iξ 1
                                                       0       2   0
                                                    ⎜       µ|ξ|        −iξ 2  ⎟
                                           a(x, ξ)=  ⎜                      ⎟
                                                    ⎝ 0      0    µ|ξ| 2  −iξ 3  ⎠
                                                                         0
                                                      −iξ 1  −iξ 2  −iξ 3
                           For n = 2, the third column and third row are to be discarded. The charac-
                           teristic determinant becames
                                                               2  2n
                                        H(x, ξ) = det a(x, ξ)= µ |ξ|  for n =2, 3 .
                           Hence, the Stokes system is elliptic in the sense of Agmon–Douglis–Nirenberg.

                           Theorem 6.2.3. (see also Chazarain and Piriou [39, Chap.4, Theorem 7.7])
                           A =((A jk )) p×p is elliptic in the sense of Agmon–Douglis–Nirenberg if and
                           only if there exists a properly supported parametrix Q 0 =((Q jk )) p×p with
                                   −t j −s k
                           Q jk ∈L      (Ω).

                           Proof: Assume that A is elliptic. By the definition of ellipticity, the homo-
                           geneous principal symbol is then of the form

                                        jk0                    r
                                       a    (x; ξ)=((|ξ| δ j  ))((a  (x; Θ)))((|ξ| δ rk ))
                                                       s j
                                                                             t r
                                                                      )
                                        s j +t k              s j +t k
                           where Θ =  ξ  and Einstein’s rule of summation is used, and where
                                  )
                                      |ξ|
                                            jk                                     n
                            H(x; Θ) = det ((a   (x; Θ) )) for every x ∈ Ω and Θ ∈ IR , |Θ| =1 .
                                 )
                                                   )
                                                                                      )
                                                                             )
                                            s j +t k
                           This implies that for every fixed x ∈ Ω there exists R(x) > 0 such that
                                jk                                                 jk
                           det((a (x; ξ)))  = 0 for all |ξ|≥ R(x). By setting a(x; ξ):=((a (x; ξ))) p×p
                                                                             n
                           and using an appropriate cut–off function χ ∈ C (Ω × IR ) with χ(x, ξ)=1
                                                                     ∞
                           for all x ∈ Ω and |ξ|≥ 2R(x) and with χ(x, ξ) = 0 in some neighbourhood
                           of the zeros of det a(x, ξ), we define q −m (x, ξ) by (6.2.8) as a matrix–valued
                           function. Recursively, then q −m−j (x, ξ) can be obtained by (6.2.9) providing
                           us with asymptotic symbol expansions of all the matrix elements of the op-
                           erator Q which we can find due to Theorem 6.1.3. The properly supported
                           parts of Q define the desired Q 0 satisfying the second equation in (6.2.6),
                           i.e., Q 0 is a left parametrix. To show that Q 0 is also a right parametrix, i.e.
                           to satisfy the first equation in (6.2.6), the arguments are the same as in the
                           scalar case in the proof of Theorem 6.2.2.
                                                                                −t j −s k
                              Conversely, if Q is a given parametrix for A with Q jk ∈L  (Ω) being
                           properly supported then its homogeneous principal symbol has the form
                                    jk0                  −t j −s k jk0
                                  ((q     (x; ξ)))  = ((|ξ|   q      (x; Θ)))
                                                                        )
                                    −t j −s k                  −t j −s k
                                                         −t j    r0            −s k
                                                  =((|ξ|   δ j  ))q  (x; Θ)))((|ξ|  δ rk )) .
                                                                         )
                                                                −t j −s k
   341   342   343   344   345   346   347   348   349   350   351