Civil Engineering Reference
In-Depth Information
Here
τ
=
τ(x)
is defined (almost everywhere) on
∂
as the direction of the tangent
in the counterclockwise direction. We endow the space (6.3) with the norm
1
2
.
2
0
2
η
0
,
rot
:
=
(
η
+
rot
η
0
)
(
6
.
4
)
In terms of the general theory in Ch. III, §4, the bilinear form
b
for the mixed
formulation is given by
b(w, θ, η)
:
=
(
∇
w
−
θ,η)
0
. Now we specialize
η
to be
an element of
L
2
()
2
of the form
η
=
curl
p
. Then because of the orthogonality
of the rotation and the gradient,
b(w, θ
;
η)
=
(
∇
w
−
θ,η)
0
=
0
−
(θ, η)
0
≤
θ
1
η
−
1
,
H
0
(), θ
H
0
()
2
, and thus
for
w
∈
∈
b(w, θ, η)
w
1
+
θ
1
≤
η
−
1
.
sup
w,θ
In order to ensure that the inf-sup condition holds, we have to endow
M
with a
norm which is weaker than the
L
2
-norm, that is with the one which is dual to (6.4),
i.e.,
·
H
−
1
(
div
,)
; see below.
As shown in Brezzi and Fortin [1986], the analyis is simplified if we use the
Helmholtz decomposition of the shear term into a gradient field and a rotational
field. Using the decomposition we get expressions and estimates which involve
the usual Sobolev norms.
The Helmholtz Decomposition
In the following we shall see that the space
H
−
1
(
div
,)
:
={
η
∈
H
−
1
()
2
div
η
∈
H
−
1
()
}
;
with the graph norm (5.17) is the dual space of
H
0
(
rot
,)
. Clearly,
H
0
(
rot
,)
⊂
L
2
()
2
⊂
H
−
1
(
div
, ).
As usual, we identify functions in
L
2
()/
R
which differ only by a constant. The
norm of an element in this space is just the
L
2
-norm of the representer which is
normalized to have zero integral; see Problem III.6.6.
Search WWH ::
Custom Search