0% found this document useful (0 votes)
57 views4 pages

Tutorial Sheet: Test 3

This document contains a tutorial sheet for a numerical analysis exam. It presents 12 exercises related to the mixed approximation of general elliptic equations. The exercises cover: proving a solution satisfies the boundary value problem; showing a proposed quantity is a norm; establishing properties of the bilinear form; obtaining estimates; defining discrete problems; and proving stability and error estimates.

Uploaded by

beckerrolandh
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
57 views4 pages

Tutorial Sheet: Test 3

This document contains a tutorial sheet for a numerical analysis exam. It presents 12 exercises related to the mixed approximation of general elliptic equations. The exercises cover: proving a solution satisfies the boundary value problem; showing a proposed quantity is a norm; establishing properties of the bilinear form; obtaining estimates; defining discrete problems; and proving stability and error estimates.

Uploaded by

beckerrolandh
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

Tutorial Sheet Test 3

Diploma: M2 MMS Université de Pau


Lecture: Numerical Analysis Collège E2S
Year: 2020/2021 R. Becker and D. Capatina

Exercice 1 Mixed approximation of general elliptic equations


Let Ω ⊂ Rd be a polyhedral domain with boundary Γ := ∂Ω. Let f ∈ L2 (Ω),
α ∈ L∞ (Ω), β ∈ Rd , γ ∈ R+ such that there exist α0 > 0, α1 > 0 satisfying 0 < α0 6
α(x) 6 α1 for almost all x ∈ Ω. We consider the boundary value problem

− div(α∇u) + β · ∇u + γu = f in Ω,
(1)
u = 0 on Γ .

Let
β
u.
σ = α∇u −
2
We set V := Hdiv (Ω) and Q = L2 (Ω) and write k·k = k·kL2 (Ω,Rm ) for m = 1 or m = d.
We consider the variational problem: (σ, u) ∈ V × Q such that for all (τ, v) ∈ V × Q
 +
a(σ, τ) + b (τ, u) = 0,
Z
(2)
b (σ, v) − c(u, v) = −

fv,

with
Z Z   Z
1 1 2
a(σ, τ) := σ · τ,
c(u, v) := γ+ |β| uv, b(σ, v) := (div σ) v,
α 4α

Z Ω
Z Ω
(3)
+ 1 − 1
b (σ, v) := b(σ, v) + β · σv, b (σ, v) := b(σ, v) − β · σv.
Ω 2α Ω 2α

We define the bilinear form


A((σ, u), (τ, v)) := a(σ, τ) + b+ (τ, u) + b− (σ, v) − c(u, v). (4)
In addition, let H be a regular simplicial mesh family.
1. Prove that a solution to (2) satisfies the boundary value problem (1).
2. Let ! 21
2
−1 1 2 2
|||(σ, u)||| :=

α (σ + 2 βu) + kdiv σk + kuk . (5)
2

Shortly show that |||(·, ·)||| is a norm on V × Q.


3. Establish that for all (σ, u) ∈ V × Q we have
2
−1 β 2
A((σ, u), (σ, −u)) = α (σ + 2 u) + γ kuk . (6)
2

Hint: use the binomial formula (identité remarquable).

1
4. Show that
Z
1
γ u div σ 6 kdiv σk2 + γ2 kuk2 and
Ω 4
Z 2 (7)
|β|2
 
1 β 1 2 − 1 β
β · σ + u div σ 6 kdiv σk + α 2 (σ + u) .
Ω 2α 2 4 4α0 2

Hint: use Young’s inequality


5. Using the previous question, show that for (τ1 , v1 ) = (0, div σ) we have
2
|β|2

1 2 1 β
α 2 (σ + u) − γ2 kuk2 .


A((σ, u), (τ1 , v1 )) > kdiv σk − (8)
2 4α0 2

6. From now on we suppose γ = 1.


Conclude using (8) and (6) that there exists κ > 0 such that

A((σ, u), (τ, v))


sup > κ |||(σ, u)|||. (9)
(τ,v)∈V×Q\{0,0} |||(τ, v)|||

Hint: use as test function (τ, v) = (τ1 , v1 ) + C(σ, −u) with a well-chosen constant C.

7. Let k ∈ N0 . We set Vh := RT kh and Qh := Dkh . Define the discrete problem.


8. Briefly show the uniform discrete inf-sup condition for A.
9. Deduce that the discrete problem has a unique solution (σh , uh ).
10. Give an estimate for the interpolation error in |||(·, ·)||| under sufficient regularity
assumptions on σ and u.
11. Deduce an error estimate for |||(σ − σh , u − uh )|||.
12. BONUS: Any ideas for the case γ = 0?
Solution Exercice 1
1. By the first equation of (2), we have for all τ ∈ V
Z Z Z
1 1
− (div τ) u = σ·τ+ β · τu
Ω Ω α Ω 2α

Since D(Ω, Rd ) ⊂ V, it follows that the distributional gradient of u is in L2 (Ω, Rd )


and
1 1 β
∇u = σ+ βu, ⇒ σ = α∇u − u.
α 2α 2
It follows that u ∈ H1 (Ω) and integration by parts gives
Z Z Z
+ 1 1
0 =a(σ, τ) + b (τ, u) = σ·τ+ (div τ) u + β · τu
Ω α Ω 2α
Z   Ω
1 1
= σ − ∇u + βu · τ + hτn , uiH− 21 (Γ )×H 12 (Γ ) ⇒ u Γ = 0.
Ω α 2α

2
Now the second equation reads
Z Z   
1 2 1
fv = γ+ |β| u − div σ + β·σ v
Ω Ω 4α 2α
 
1 1
⇒ div σ = γ + u+ β·σ−f
4α 2α
     
β 1 2 1 β
⇒ div α∇u − u = γ + |β| u + β · α∇u − u − f
2 4α 2α 2
⇒ div (α∇u − βu) = γu − f
⇒ div (α∇u) − β · ∇u − γu = −f

2. If |||(σ, u)||| = 0, first u = 0 and the remaining part is equivalent to kσkHdiv (Ω)
thanks to α > 0.
2
3. We have τ + β2 v = |τ|2 + τ · βv + 14 |β|2 v2 and then

A((σ, u), (σ, −u)) =a(σ, σ) + b+ (σ, u) − b− (σ, u) + c(u, u)


Z Z ! Z
1 2 |β|2 1
= |σ| + γ+ 2
u + β · σu
Ω α Ω 4α Ω α
Z 2 Z
1 β
= σ + 2 u + γu2
Ω α Ω
2
−1 β 2
= α (σ + 2 u) + γ kuk .
2

4.
Z
1
γu div σ 6 kdiv σk2 + γ2 kuk2
Ω 4
Z 2
|β|2
 
1 β 1 2 − 1 β
β · σ + u div σ 6 kdiv σk + α 2 (σ + u)
Ω 2α 2 4 4α0 2

5.

A((σ, u), (τ1 , v1 )) =b− (σ, v1 ) − c(u, v1 ) = b− (σ, div σ) − c(u, div σ)
Z Z  
2 1 1 2
= kdiv σk − β · σ div σ − γ+ |β| u div σ
Ω 2α 4α
Z  Ω
Z
2 1 β
= kdiv σk − β · σ + u div σ − γu div σ
Ω 2α 2 Ω
2
|β|2

2 1 2 1 β 1
> kdiv σk − kdiv σk − (σ + u) − kdiv σk2 − kγuk2
4 4 α 2 4
2 2
1 |β| α− 12 (σ + β u) − γ2 kuk2
> kdiv σk2 −

2 4α0 2

3
6. Taking the test function (τ, v) = (τ1 , v1 ) + C(σ, −u), we get
! 2
2
1 2 |β| −1 β
+ (C − 1) kuk2

A((σ, u), (τ, v)) > kdiv σk + C − α 2 (σ + u)
2 4α0 2


|β|2
So, we get with C := 1 + max 4α0
,1 the result.
7.
Z
(σh , uh ) ∈ Vh × Qh : A((σ, u), (τ, v)) = − fv, ∀(τ, v) ∈ Vh × Qh .

8. Same argument as in the continuous case, since Qh ⊂ Q, Vh ⊂ V and div Vh ⊂


Qh .
9. By the Babuška theorem.
10. With Ih the Raviart-Thomas interpolator and πh the L2 (Ω) projection on Qh

|β|
 
1
|||(σ − Ih σ, u − πh u)||| 6 kσ − Ih σk + 1 + ku − πh uk + kdiv(σ − Ih σ)k
α0 2α0
|β|
   
1
k
6CI dh |σ| + 1+ |u|k+1,Ω + |div σ|k+1,Ω
α0 k+1,Ω 2α0

11. According to the Babuška theorem, there exists C independent of H such that

|||(σ − σh , u − uh )||| 6C inf {|||(σ − τh , u − vh )||| | (τh , vh ) ∈ Vh × Qh }


6C|||(σ − Ih σ, u − πh u)|||
6C1 dkh |σ|k+1,Ω + |u|k+1,Ω + |div σ|k+1,Ω .


You might also like