Page 343 -
P. 343

6.2 Elliptic Pseudodifferential Operators on Ω ⊂ IR n  327

                                                                                m
                           Definition 6.2.2. A parametrix Q 0 for the operator A ∈L (Ω) is a prop-
                                                                                       −∞
                           erly supported operator which is a two–sided inverse for A modulo L  (Ω):
                                               A ◦ Q 0 − I  = C 1 ∈L −∞ (Ω)
                           and                                                          (6.2.6)
                                               Q 0 ◦ A − I  = C 2 ∈L −∞ (Ω)
                           Theorem 6.2.2. (H¨ormander[129], Taylor [302, Chap. III, Theorem 1.3])
                                 m
                           A ∈L (Ω) is elliptic of order m if and only if there exists a parametrix
                                  −m
                           Q 0 ∈L   (Ω) satisfying (6.2.6).
                                                                   m
                           Proof: (i) Let A be a given operator in L (Ω) which is elliptic of order
                           m. Then to the complete symbol class of A there is a representative a(x, ξ)
                           satisfying (6.2.2). If Q would be known then (6.1.37) with (6.2.6) would imply

                                                                         α
                                                                   1     ∂
                             σ Q◦A =1 ∼    q −m−j (x, ξ)a(x, ξ)+               q −m−  (x, ξ) ×
                                                                  α! ∂ξ
                                        0≤j                   1≤|α|         0≤
                                                                           ∂
                                                                               α
                                                                    ×   − i     a(x, ξ) .  (6.2.7)
                                                                           ∂x
                                                                  n
                           So, to a we choose a function χ ∈ C (Ω ×IR ) with χ(x, ξ) = 1 for all x ∈ Ω
                                                          ∞
                           and |ξ|≥ C ≥ 1 where C is chosen appropriately large, and χ = 0 in some
                           neighbourhood of the zeros of a. Then define
                                                q −m (x, ξ):= χ(x, ξ)a(x, ξ) −1         (6.2.8)
                           and, in view of (6.2.7), recursively for j =1, 2,...,

                             q −m−j (x, ξ) :=                                           (6.2.9)
                                          1      ∂    α                 ∂    α
                                 −                 q −m−j+|α| (x, ξ)  − i   a(x, ξ) q −m (x, ξ) .
                                          α!  ∂ξ                       ∂x
                                   1≤|α|≤j
                                                                                           n
                           From this construction one can easily show that q −m−j ∈ S S S −m−j (Ω × IR ).
                           The sequence {q −m−j } can be used to define an operator Q ∈ OPS −m (Ω ×
                             n
                           IR ) with  ∞    q −m−j as the asymptotic expansion of σ Q , see Theorem 6.1.3.
                                    j=0
                           With Q available, we use the decomposition of Q = Q 0 + R, Theorem 6.1.9,
                                                    n
                           where Q 0 ∈ OPS  −m (Ω × IR ) is properly supported and still has the same
                                                                                   0
                           symbol as Q. Then we have with Theorem 6.1.14 that Q 0 ◦A ∈L (Ω)and by
                                                                                       −∞
                           construction in view of (6.1.37) σ Q 0 ◦A ∼ 1. Hence, Q 0 ◦A−I = R 2 ∈L  (Ω).
                                                                               −∞
                              In the next step it remains to show that A ◦ Q 0 − I ∈L  (Ω), too.
                              The choice of q −m also shows that Q 0 is an elliptic operator of order
                                                                      (−1)      m
                           −m. Therefore, to Q 0 there exists an operator Q 0  ∈ OPS  which is also
                                                          (−1)
                           properly supported and satisfies Q 0  ◦ Q 0 − I = R 1 ∈L −∞ (Ω), where R 1
                           is properly supported, due to Proposition 6.1.6. Then
   338   339   340   341   342   343   344   345   346   347   348