|
3/2
⊂ B
h
with
dist(A
3/2
, ∂B
h
\∂Ω) ≤ h such that
β
ξ
q L
2
(A
3/2
)
≤
sup
z∈V
h
supp z⊂B
h
(ξ
q
,
· z)
z
H
1
(B
h
)
.
Since d ≥ κh and we can choose κ > 2 then we have that B
h
⊂ A
2
, and so
β
ξ
q L
2
(A
3/2
)
≤
sup
z∈V
h
supp z⊂A
2
(ξ
q
,
· z)
z
H
1
(A
2
)
Now using equation (3.1a) and a-g.m. inequality to obtain
MAX-NORM STABILITY OF LOW ORDER TAYLOR-HOOD ELEMENTS
IN THREE DIMENSIONS
7
I
5
≤
C
d
sup
z∈V
h
supp z⊂A
2
(ξ
q
,
· z)
z
H
1
(A
2
)
ξ
L
2
(A
2
)
≤
C
d
η
q L
2
(A
2
)
+
sup
z∈V
h
supp z⊂A
2
(e
q
,
· z)
z
H
1
(A
2
)
ξ
L
2
(A
2
)
≤
C
d
η
q L
2
(A
2
)
+
sup
z∈V
h
supp z⊂A
2
( e
q
, z)
z
H
1
(A
2
)
ξ
L
2
(A
2
)
≤
C
d
η
q L
2
(A
2
)
+
sup
z∈V
h
supp z⊂A
2
( e,
z)
z
H
1
(A
2
)
ξ
L
2
(A
2
)
≤
C
d
η
q L
2
(A
2
)
+
e
L
2
(A
2
)
ξ
L
2
(A
2
)
≤
η
q
2
L
2
(A
2
)
+ ε
e
2
L
2
(A
2
)
+
C
d
2
(1 + ε
−1
)( e
2
L
2
(A
2
)
+ η
2
L
2
(A
2
)
)
Until now, combining the estimates for I
2
, I
4
and I
5
in (3.6) we have
1
4
ω e
L
2
(Ω)
≤ I
3
+ C
η
2
L
2
(A
2
)
+ η
q
2
L
2
(A
3/2
)
+
C
εd
2
( η
2
L
2
(A
2
)
+ e
2
L
2
(A
2
)
) + ε
e
2
L
2
(A
2
)
It remains to estimate I
3
. Again we use that e
q
= η
q
+ ξ
q
decomposing I
3
into two terms
I
3
= −(e
q
,
· (ω
2
ξ)) = −(η
q
,
· (ω
2
ξ)) − (ξ
q
,
· (ω
2
ξ)) =: I
6
+ I
7
The estimate for I
6
is obtained applying C-S. inequality, property (2.6a) for s = 0 and s = 1,
and the a-g.m. inequality, resulting
I
6
≤ C η
q
2
L
2
(A
3/2
)
+
1
8
ω e
2
L
2
(A
3/2
)
+ C
η
2
L
2
(A
3/2
)
+
C
d
2
η
2
L
2
(A
3/2
)
+
C
d
2
e
2
L
2
(A
3/2
)
.
In order to estimate I
7
we note that, by definition of ω
(c,
· (ω
2
ξ)) = 0
for c constant. Set ˆ
ξ
q
= ξ
q
− c and choose c such that ˆ
ξ
q
has zero mean on A
3/2
. Then by
product rule and adding and subtracting R(ω
2
( ˆ
ξ
q
)) we have
I
7
=
−( ˆ
ξ
q
,
(ω
2
) · ξ) − ( ˆ
ξ
q
, ω
2
· ξ)
=
−( ˆ
ξ
q
,
(ω
2
) · ξ) − (ω
2
ˆ
ξ
q
− R(ω
2
ˆ
ξ
q
),
· ξ) − (R(ω
2
ˆ
ξ
q
),
· ξ) =: I
8
+ I
9
+ I
10
We estimate I
8
using C-S. inequality and property (2.6a)
I
8
≤
C
d
ˆ
ξ
q L
2
(A
3/2
)
ξ
L
2
(A
3/2
)
.
8
MAX-NORM STABILITY OF LOW ORDER TAYLOR-HOOD ELEMENTS
IN THREE DIMENSIONS
Using the superapproximation property (2.6c) and the inverse estimate assumption A4 we
estimate I
9
as follows
I
9
≤
Ch
d
ˆ
ξ
q L
2
(A
3/2
)
ξ
L
2
(A
3/2
)
≤
C
d
ˆ
ξ
q L
2
(A
3/2
)
ξ
L
2
(A
2
)
.
To estimate I
10
we apply the equation (3.1b), the property of R (2.4b), (2.6a), the local
inf-sup condition A5 and C-S. inequality obtaining
I
10
= (R(ω
2
ˆ
ξ
q
),
· η) ≤ C ˆ
ξ
q L
2
(A
3/2
)
η
L
2
(A
3/2
)
.
We claim that
ˆ
ξ
q L
2
(A
3/2
)
≤ C(
e
L
2
(A
2
)
+ η
q L
2
(A
2
)
). We prove this claim in Lemma 3.1.
Therefore, we have
I
7
≤ C( η
q L
2
(A
2
)
+
e
L
2
(A
2
)
)(
η
L
2
(A
3/2
)
+
1
d
ξ
L
2
(A
2
)
)
≤ ε( η
q
2
L
2
(A
2
)
+
e
2
L
2
(A
2
)
) +
C
εd
2
( η
2
L
2
(A
2
)
+ e
2
L
2
(A
2
)
) +
C
ε
η
2
L
2
(A
2
)
.
The estimates for I
6
and I
7
yield
1
8
ω e
L
2
(Ω)
≤ C(
1
ε
η
2
L
2
(A
2
)
+ η
q
2
L
2
(A
2
)
+
1
εd
2
η
2
L
2
(A
2
)
) +
C
εd
2
e
2
L
2
(A
2
)
+ ε
e
2
L
2
(A
2
)
.
The exact statement of Theorem 2 is attached using ε
2
. This completes the proof under
Assumption A7.
Now we extend the result for general sets A
1
⊂ A
2
⊂ Ω with dist(A
1
, ∂A
2
\∂Ω) ≥ d ≥ κh. It
is not difficult to construct a covering {G
i
}
M
i=1
of A
1
, where G
i
= B
d/2
(x
i
) ∩ Ω with the following
properties:
(1) A
1
⊂
M
i=1
G
i
.
(2) x
i
∈ A
1
for each 1 ≤ i ≤ M .
(3) Let H
i
= B
d
(x
i
) ∩ Ω. There exists a fixed number L such that each point x ∈
M
i=1
H
i
is
contained in at most L sets from {H
j
}
M
j=1
.
(4) There exists a ρ > 0 such that for each 1 ≤ i ≤ M there exists a ball B ⊂ G
i
such that
diam(G
i
) ≤ ρdiam(B).
Since dist(A
1
, ∂A
2
\∂Ω) ≥ d, using property 2 we have that
M
i=1
H
i
⊂ A
2
.
Applying the result proved above and using properties 1 and 4 we have
(v − v
h
)
2
L
2
(A
1
)
≤
M
i=1
(v − v
h
)
2
L
2
(G
i
)
≤
M
i=1
C
(v − Pv)
2
L
2
(H
i
)
+ q − Rq
2
L
2
(H
i
)
+ (
1
εd
)
2
v − Pv
2
L
2
(H
i
)
+ ε
2
(v − v
h
)
2
L
2
(H
i
)
+ (
C
εd
)
2
v − v
h
2
L
2
(H
i
)
.
Using property 3 we have
MAX-NORM STABILITY OF LOW ORDER TAYLOR-HOOD ELEMENTS
IN THREE DIMENSIONS
9
(v − v
h
)
2
L
2
(A
1
)
≤ CL
(v − Pv)
2
L
2
(A
2
)
+ q − Rq
2
L
2
(a
2
)
+ (
L
εd
)
2
v − Pv
2
L
2
(A
2
)
+ Lε
2
(v − v
h
)
2
L
2
(A
2
)
+ (
CL
εd
)
2
v − v
h
2
L
2
(A
2
)
.
The exact statement of Theorem 2 is attached using ε
2
.
The next result is exactly the same as Lemma 3.2 in [5]. However, the proof in [5] used the
existence of a quasi-local Fortin projection.
Lemma 3.1. Under the assumption A7, there exists a constat C independent of A
3/2
and ˆ
ξ
q
,
but depends on ρ sucht that
ˆ
ξ
q L
2
(A
3/2
)
≤ C(
e
L
2
(A
2
)
+ η
q L
2
(A
2
)
).
Proof. Define w ∈ H
1
0
(A
3/2
) as the solution of the problem
· w =
ˆ
ξ
q
in A
3/2
w
=
0
on ∂A
3/2
We can choose w so that w
H
1
(A
3/2
)
≤ C ˆ
ξ
q L
2
(A
3/2
)
.
By Lemma 3.1 in Chapter III.3 in [11], the constant C is independent of ˆ
ξ
q
and depends only
on the ratio of the diameter of A
3/2
and the radius of the largest ball that can be inscribed
into A
3/2
and hence by our hypothesis only depends on ρ. Let us extend w on all of Ω by zero
outside of A
3/2
. We note that this implies that Pw vanishes outside of A
2
by A3. Then,
ˆ
ξ
q
2
L
2
(A
3/2
)
=
( ˆ
ξ
q
, ˆ
ξ
q
)
A
3/2
= ( ˆ
ξ
q
,
· w) = (ξ
q
,
· w)
=
(e
q
,
· w) − (η
q
,
· w).
Using (3.1a),
(e
q
,
· w) = (e
q
,
· Pw) + (e
q
,
· (w − Pw))
=
( e,
Pw) + (η
q
,
· (w − Pw)) + (ξ
q
,
· (w − Pw))
=
( e,
Pw) + (η
q
,
· (w − Pw)) − ( ξ
q
, w − Pw)
≤
e
L
2
(A
2
)
Pw
L
2
(A
2
)
+ η
q L
2
(A
2
)
(w − Pw)
L
2
(A
2
)
ξ
q L
2
(A
3/2
)
w − Pw
L
2
(A
3/2
)
≤ C(
e
L
2
(A
2
)
+ η
q L
2
(A
2
)
+ h
ξ
q L
2
(A
3/2
)
) w
H
1
(A
3/2
)
Using the local inf-sup condition A5 we have
h
ξ
q L
2
(A
3/2
)
≤ C( η
q L
2
(A
2
)
+
e
L
2
(A
2
)
)
Therefore
ˆ
ξ
q
2
L
2
(A
3/2
)
≤ C( η
q L
2
(A
2
)
+
e
L
2
(A
2
)
) w
H
1
(A
3/2
)
≤ C( η
q L
2
(A
2
)
+
e
L
2
(A
2
)
) ˆ
ξ
q H
1
(A
3/2
)
which implies the result.
10
MAX-NORM STABILITY OF LOW ORDER TAYLOR-HOOD ELEMENTS
IN THREE DIMENSIONS
4. Proof of Theorem 1
4.1. Green’s function estimates. In this section we recall pointwise estimates for the Green’s
matrix. Let φ(z) be an infinitely differentiable function in Ω which vanishes in a neighborhood
of the edges of Ω such that
(4.1)
Ω
φ(x)dx = 1.
Consider the Stokes problem with non-zero divergence. Let (u, p) ∈ [H
1
0
(Ω)]
3
× L
2
φ
(Ω) solve
−∆u +
p = f
in Ω
(4.2a)
· u = q
in Ω
(4.2b)
u = 0
on ∂Ω.
(4.2c)
for arbitrary f ∈ [H
−1
(Ω)]
3
and q ∈ L
2
0
(Ω) with q vanishing on the singular points of Ω (see
[2]). If q ∈ H
1
(Ω) ∩ L
2
0
(Ω) with q vanishing on the edges of Ω and f ∈ [L
2
(Ω)]
3
we have the
following elliptic regularity result (see [2])
(4.3)
u
H
2
(Ω)
+ p
H
1
(Ω)
≤ C( f
L
2
(Ω)
+ q
H
1
(Ω)
).
The Green’s matrix for the problem (4.2) G
j
= (G
1,j
, G
2,j
, G
3,j
)
T
and the functions G
4,j
for
j = 1, 2, 3, 4 are solutions of the problem
−∆
x
G
j
(x, ξ) +
x
G
4,j
(x, ξ) = δ(x − ξ)(δ
1,j
, δ
2,j
, δ
3,j
)
T
for x, ξ ∈ Ω
(4.4a)
x
· G
j
(x, ξ) = (δ(x − ξ) − φ(x))δ
4,j
for x, ξ ∈ Ω
(4.4b)
G
j
(x, ξ)( = 0
for x ∈ ∂Ω, ξ ∈ Ω.
(4.4c)
and G
4,j
satisfies the condition
(4.5)
Ω
G
4,j
(x, ξ)φ(x)dx = 0,
for ξ ∈ Ω,
j = 1, 2, 3, 4.
Here, δ(x) is the delta function, and δ
i,j
is the Kronecker delta symbol. In addition,
G
i,j
(x, ξ) = G
j,i
(ξ, x)
for x, ξ ∈ Ω,
i, j = 1, 2, 3, 4/
The following Theorem, (cf. [7], [8]) gives us the existence and uniqueness of such a matrix.
Theorem 3. There exists a uniquely determined Green’s matrix G(x, ξ) such that the vector
functions
x → ζ(x, ξ)(G
j
(x, ξ), G
4,j
(x, ξ))
belong to the space [H
1
0
(Ω)]
3
× L
2
(Ω) for each ξ ∈ Ω and for every infinitely differentiable
function ζ(·, ξ) equal zero in a neighborhood of the point x = ξ.
Then, we have the following representation (cf. [12]) of the solution of problem 4.2 in terms
of the Green’s matrix
MAX-NORM STABILITY OF LOW ORDER TAYLOR-HOOD ELEMENTS
IN THREE DIMENSIONS
11
u
i
(x) =
3
j=1
Ω
G
i,j
(x, ξ)f
j
(ξ)dξ +
Ω
G
i,4
(x, ξ)q(ξ)dξ
i = 1, 2, 3.
(4.6a)
p(x) =
3
j=1
Ω
G
4,j
(x, ξ)f
j
(ξ)dξ +
Ω
G
4,4
(x, ξ)q(ξ)dξ
(4.6b)
The following estimates were established in papers of [9, 7, 8, 11] ( see also [10] Sec. 11.5).
Theorem 4. Let Ω ⊂ R
3
be a convex domain of polyhedral type. Then there exists a constant
C such that
(4.7)
|∂
α
x
∂
β
ξ
G
i,j
(x, ξ)| ≤ C|x − ξ|
−1−|α|−|β|−δ
i,4
−δ
j,4
,
for |α| ≤ 1 − δ
i,4
, |β| ≤ 1 − δ
j,4
, x, ξ ∈ Ω, x = ξ, and multi-indices 0 ≤ |α|, |β| ≤ 1.
Moreover, for polyhedral domain the Green’s matrix satisfies the H¨
older type estimate
(4.8)
|∂
α
x
∂
β
ξ
G
i,j
(x, ξ) − ∂
α
y
∂
β
ξ
G
i,j
(y, ξ)|
|x − y|
σ
≤ C(|x − ξ|
−1−σ−|α|−|β|−δ
i,4
−δ
j,4
+ |y − ξ|
−1−σ−|α|−|β|−δ
i,4
−δ
j,4
),
for |α| ≤ 1 − δ
i,4
, |β| ≤ 1 − δ
j,4
. Here σ is a sufficiently small positive number which depends
on the geometry of the domain.
4.2. Preliminary results. Let z be an arbitrary point of Ω and let T
z
∈ T
h
be the element
containing z. Our aim is to estimate |∂
x
j
(u
h
)
i
(z)| and |p
h
(z)|, where 1 ≤ i, j ≤ 3 are arbitrary.
We will start representing them in terms of the smooth Green’s function. Then after some
manipulations the problem is reduced to estimate the error of the Green’s function in L
1
(Ω)
norm, that estimate is presented in this section and we leave the rest of the proof for section
4.3. Preliminarily, we define the smooth delta function. Let δ
z
h
(x) = δ
h
∈ C
1
0
(T
z
) be a smooth
function such that
(4.9)
r(z) = (r, δ
h
)
T
z
,
∀r ∈ P
l
(T
z
),
where P
l
(T
z
) is the space of polynomials of degree at most l defined on T
z
, with the following
property
δ
h W
k
q
(T Dostları ilə paylaş: |
|
|