Page 241 -
P. 241

5.2 Abstract Existence Theorems for Variational Problems  225

                           Lemma 5.2.5. Let the sesquilinear form a(·, ·) satisfy G˚arding’s inequality
                           (5.2.2) and, in addition, have the property
                                          Re a(v, v) > 0 for all v ∈H with v  =0 .

                           Then a(·, ·) satisfies (5.2.4) and, consequently, is strongly H–elliptic.
                           Proof: The proof rests on the well known weak compactness of the unit
                           sphere in reflexive Banach spaces and in Hilbert spaces (Schechter [270, VIII
                           Theorem 4.2]), i. e. every bounded sequence {v j } j∈IN ⊂H with 
v j 
 H ≤ M
                           contains a subsequence v j   with a weak limit v 0 ∈H such that

                                           lim (g, v j  ) H =(g, v 0 ) H for every g ∈H .
                                          j   →∞
                              We now prove the lemma by contradiction. If a were not strongly H–
                           elliptic then there existed a sequence {v j } j∈IN ⊂H with 
v j 
 H =1 and

                                                    lim Re a(v j ,v j )=0 .
                                                   j→∞
                           Then {v j } contained a subsequence {v j  } converging weakly to v 0 ∈H.
                           G˚arding’s inequality then implied

                                        2
                             α 0 
v j   − v 0 
  ≤ Re{a(v j   − v 0 ,v j   − v 0 )+ C(v j   − v 0 ),v j   − v 0  }
                                        H
                                                                                         H
                                            ≤ Re{a(v j  ,v j  ) − a(v 0 ,v j  ) − a(v j  ,v 0 )
                                                + a(v 0 ,v 0 )+(Cv j  ,v j   − v 0 ) H − (Cv 0 ,v j   − v 0 ) H } .
                           Since C is compact, there existed a subsequence {v j   }⊂{v j  }⊂H such that

                           Cv j    → w ∈H for j →∞. Hence, due to the weak convergence v j     v 0
                           we would have

                                   (Cv j   ,v j   ) H − (Cv j   ,v 0 ) H  →  (w, v 0 ) H − (w, v 0 ) H =0 ,

                                                 ∗                    ∗
                                 a(v 0 ,v j   )= (jA) v 0 ,v j     →  (jA) v 0 ,v 0  = a(v 0 ,v 0 )
                                                         H                  H
                           and corresponding convergence of the remaining terms on the right–hand
                           side. This yielded

                                                                2
                                         0 ≤ lim j    →∞ α 0 
v j    − v 0 
  ≤−Re a(v 0 ,v 0 )
                                                                H
                           and, consequently, with Re a(v 0 ,v 0 ) ≥ 0, we could find
                                     lim 
v j    − v 0 
 H = 0  together with Re a(v 0 ,v 0 )=0 .
                                    j    →∞
                           The latter implied
                                        v 0 = 0 ; however, 
v 0 
 H = lim 
v j   
 H =1 ,
                                                                 j    →∞
   236   237   238   239   240   241   242   243   244   245   246