Page 334 -
P. 334

318    6. Introduction to Pseudodifferential Operators

                           and we have the asymptotic expansion for the complete symbol class:

                                           1       ∂    α           ∂    α
                                σ A◦B ∼              σ A (x, ξ)  − i     σ B (x, ξ)  .  (6.1.37)
                                           α!   ∂ξ                  ∂x
                                       α≥0
                           Here, σ A (x, ξ)and σ B (x, ξ) denote one of the respective representatives in
                           the corresponding equivalence classes of the complete symbol classes.
                              For the proof, we remark that with Theorem 6.1.9 we have either
                           A = A 0 + R and B = B 0 or A = A 0 and B = B 0 + Q. Then

                                               A ◦ B  = A 0 ◦ B 0 + R ◦ B 0
                              or
                                               A ◦ B  = A 0 ◦ B 0 + A 0 ◦ Q.

                           Since A 0 and B 0 are properly supported, R ◦ B 0 or A 0 ◦ Q are continuous

                           mappings from E (Ω)to C (Ω) and, hence, are smoothing operators. (Note
                                                   ∞
                           that for A and B both not properly supported, the composition generates
                           the term R ◦ Q, the composition of two regularizers, which is not defined in
                           general.)
                              For the remaining products A 0 ◦ B 0 , we can use that without loss of
                           generality, A 0 = A 0 (x, D) ∈ OPS m 1 (Ω)and B 0 = B 0 (x, D) ∈ OPS m 2 (Ω).
                           Hence, one obtains for C := A 0 ◦ B 0 a representation in the form (6.1.26)
                           with the amplitude function

                                      c(x, z, η)=     a(x, ξ)e i(x−y)·ξ b(y, η)e i(y−z)·η dydξ
                                                IR n  Ω
                                                                         n
                           for which one needs to show c ∈ S m 1 +m 2 (Ω × Ω × IR ) and, with the Leibniz
                           rule, to evaluate the asymptotic expansion (6.1.28) for c. For details see e.g.
                           [306, Chap. I, Theorem 3.2].
                              As an immediate consequence of Theorem 6.1.14, the following corollary
                           holds.

                           Corollary 6.1.15. Let A ∈L  m 1 (Ω) ,B ∈L m 2 (Ω) and one of them be
                           properly supported. Then the commutator satisfies
                                                                   m 1 +m 2 −1
                                          [A, B]:= A ◦ B − B ◦ A ∈L        (Ω) .       (6.1.38)
                           Proof: For A ◦ B and B ◦ A we have Theorem 6.1.14 and the commutator’s
                           symbol has the asymptotic expansion
                                                1       ∂    α           ∂    α

                                    σ [A,B] ∼              σ A (x, ξ)  − i    σ B (x, ξ)
                                                α!    ∂ξ                 ∂x
                                           |α|≥1
                                              ∂                  ∂
                                                  α                  α
                                          −        σ B (x, ξ)  − i   σ A (x, ξ)  .
                                              ∂ξ                 ∂x
                           Hence, the order of [A, B] equals m 1 − 1+ m 2 .
   329   330   331   332   333   334   335   336   337   338   339