Civil Engineering Reference
In-Depth Information
Note that the cone condition excludes cusps in the domain. The domain
2
0 <y<x 5 < 1
:
={ (x, y) ∈ R
;
}
has a cusp at the origin, and H 1 () contains the function
u(x, y) = x 1 ,
whose trace is not square-integrable over .
We would like to point out that Green's formula (2.9) also holds for functions
u, w H 1 () , provided that satisfies the hypotheses of the trace theorem.
The space H 1 () is isomorphic to a direct sum
H 1 () H 0 () γ(H 1 ()).
Specifically, every u H 1 () can be decomposed as
u = v + w,
according to the following rule. Let w be the solution of the variational problem
| w |
2
1
min ! More exactly, suppose
H 0 (),
(
w,
v) 0 , =
0
for all v
w u H 0 ().
= u w H 0 () . Here γ is an injective mapping on the set of functions
w which appear in the decomposition.
Let v :
We now consider the connection with continuous functions. As usual, the
norm
u = u , is based on the essential supremum of
| u |
over . [It is
not a Sobolev norm.]
2 be a convex polygonal domain, or a domain with
Lipschitz continuous boundary. Then H 2 () is compactly imbedded in C( ¯ ) , and
3.4 Remarks. (1) Let ⊂ R
for all v H 2 (),
v c v 2
( 3 . 11 )
for some number c = c() .
(2) For every open connected domain ⊂ R
2 , H 2 () is compactly imbedded
in C() .
The above results are not the sharpest possible in this framework. Because of
their importance, and because they follow simply from the trace theorem, we now
give their proofs.
Search WWH ::




Custom Search