Introduction to the Finite Element Method -

Lecture 2

Neela Nataraj

Department of Mathematics,
Indian Institute of Technology Bombay,
Powai, Mumbai 76
[email protected]
Adaptive FEM Worshop, IIST, Tvm - March 2012

March 18, 2012

Neela Nataraj IITB

A priori Error Estimates

AIM : To obtain the estimates ku − uh kV .

Neela Nataraj IITB

A priori Error Estimates

AIM : To obtain the estimates ku − uh kV .

The important steps involved are:

Neela Nataraj IITB

A priori Error Estimates

AIM : To obtain the estimates ku − uh kV .

The important steps involved are:
1. Cea’s lemma

Neela Nataraj IITB

A priori Error Estimates

AIM : To obtain the estimates ku − uh kV .

The important steps involved are:
1. Cea’s lemma
2. Interpolation operator

Neela Nataraj IITB

A priori Error Estimates

AIM : To obtain the estimates ku − uh kV .

The important steps involved are:
1. Cea’s lemma
2. Interpolation operator
3. Affine mapping,

Neela Nataraj IITB

A priori Error Estimates

AIM : To obtain the estimates ku − uh kV .

The important steps involved are:
1. Cea’s lemma
2. Interpolation operator
3. Affine mapping, (Inverse),

Neela Nataraj IITB

A priori Error Estimates

AIM : To obtain the estimates ku − uh kV .

The important steps involved are:
1. Cea’s lemma
2. Interpolation operator
3. Affine mapping, (Inverse), |J(FT )|, |J(FT−1 )| estimates

Neela Nataraj IITB

A priori Error Estimates

AIM : To obtain the estimates ku − uh kV .

The important steps involved are:
1. Cea’s lemma
2. Interpolation operator
3. Affine mapping, (Inverse), |J(FT )|, |J(FT−1 )| estimates
4. Change of scale

Neela Nataraj IITB

A priori Error Estimates

AIM : To obtain the estimates ku − uh kV .

The important steps involved are:
1. Cea’s lemma
2. Interpolation operator
3. Affine mapping, (Inverse), |J(FT )|, |J(FT−1 )| estimates
4. Change of scale
5. Bramble-Hilbert lemma

Neela Nataraj IITB

A priori Error Estimates

AIM : To obtain the estimates ku − uh kV .

The important steps involved are:
1. Cea’s lemma
2. Interpolation operator
3. Affine mapping, (Inverse), |J(FT )|, |J(FT−1 )| estimates
4. Change of scale
5. Bramble-Hilbert lemma
6. Semi-norm inner product

Neela Nataraj IITB

A priori Error Estimates

AIM : To obtain the estimates ku − uh kV .

The important steps involved are:
1. Cea’s lemma
2. Interpolation operator
3. Affine mapping, (Inverse), |J(FT )|, |J(FT−1 )| estimates
4. Change of scale
5. Bramble-Hilbert lemma
6. Semi-norm inner product
7. Finite element interpolation over reference triangle

Neela Nataraj IITB

A priori Error Estimates

AIM : To obtain the estimates ku − uh kV .

The important steps involved are:
1. Cea’s lemma
2. Interpolation operator
3. Affine mapping, (Inverse), |J(FT )|, |J(FT−1 )| estimates
4. Change of scale
5. Bramble-Hilbert lemma
6. Semi-norm inner product
7. Finite element interpolation over reference triangle
8. Estimate k u − uh kV .

Neela Nataraj IITB

Step 1. Cea’s Lemma

Let u and uh be the solutions of (PG ) and (PGh ) respectively.

Neela Nataraj IITB

Step 1. Cea’s Lemma

Let u and uh be the solutions of (PG ) and (PGh ) respectively. Then,

ku − uh kV ≤ C inf ku − vh kV . (1)
vh ∈Vh

Neela Nataraj IITB

Step 1. Cea’s Lemma

Let u and uh be the solutions of (PG ) and (PGh ) respectively. Then,

ku − uh kV ≤ C inf ku − vh kV . (1)
vh ∈Vh

Proof: We have, ∀vh ∈ Vh ,

Neela Nataraj IITB

Step 1. Cea’s Lemma

Let u and uh be the solutions of (PG ) and (PGh ) respectively. Then,

ku − uh kV ≤ C inf ku − vh kV . (1)
vh ∈Vh

Proof: We have, ∀vh ∈ Vh ,

a(u, vh ) = l(vh )

Neela Nataraj IITB

Step 1. Cea’s Lemma

Let u and uh be the solutions of (PG ) and (PGh ) respectively. Then,

ku − uh kV ≤ C inf ku − vh kV . (1)
vh ∈Vh

Proof: We have, ∀vh ∈ Vh ,

a(u, vh ) = l(vh )
a(uh , vh ) = l(vh )

Neela Nataraj IITB

Step 1. Cea’s Lemma

Let u and uh be the solutions of (PG ) and (PGh ) respectively. Then,

ku − uh kV ≤ C inf ku − vh kV . (1)
vh ∈Vh

Proof: We have, ∀vh ∈ Vh ,


a(u, vh ) = l(vh )
⇒ a(u − uh , vh ) = 0 ∀vh ∈ Vh .
a(uh , vh ) = l(vh )

Neela Nataraj IITB

Step 1. Cea’s Lemma

Let u and uh be the solutions of (PG ) and (PGh ) respectively. Then,

ku − uh kV ≤ C inf ku − vh kV . (1)
vh ∈Vh

Proof: We have, ∀vh ∈ Vh ,


a(u, vh ) = l(vh )
⇒ a(u − uh , vh ) = 0 ∀vh ∈ Vh . (2)
a(uh , vh ) = l(vh )

(2) is called the Galerkin orthogonality property .

Neela Nataraj IITB

Step 1. Cea’s Lemma

Let u and uh be the solutions of (PG ) and (PGh ) respectively. Then,

ku − uh kV ≤ C inf ku − vh kV . (1)
vh ∈Vh

Proof: We have, ∀vh ∈ Vh ,


a(u, vh ) = l(vh )
⇒ a(u − uh , vh ) = 0 ∀vh ∈ Vh . (2)
a(uh , vh ) = l(vh )

(2) is called the Galerkin orthogonality property .

i.e., the finite element solution uh is the orthogonal projection of
the exact solution u onto Vh with respect to the inner product
a(·, ·).

Neela Nataraj IITB

αku − uh k2V ≤ a(u − uh , u − uh ) (By coercivity of a(·, ·))

Neela Nataraj IITB

αku − uh k2V ≤ a(u − uh , u − uh ) (By coercivity of a(·, ·))
= a(u − uh , u − vh ) + a(u − uh , vh − uh ) (using (2))
| {z }
=0

Neela Nataraj IITB

 αku − uh k2V ≤ a(u − uh , u − uh ) (By coercivity of a(·, ·))
= a(u − uh , u − vh ) + a(u − uh , vh − uh ) (using (2))
| {z }
=0
M
i.e., ku − uh k2V ≤ ku − uh kV ku − vh kV (Boundedness of a(·, ·)).
α
In case u = uh , the result is trivial.

Neela Nataraj IITB

 αku − uh k2V ≤ a(u − uh , u − uh ) (By coercivity of a(·, ·))
= a(u − uh , u − vh ) + a(u − uh , vh − uh ) (using (2))
| {z }
=0
M
i.e., ku − uh k2V ≤ ku − uh kV ku − vh kV (Boundedness of a(·, ·)).
α
In case u = uh , the result is trivial.
M
For u 6= uh , ku − uh kV ≤ ku − vh kV ∀ vh ∈ Vh .
α

Neela Nataraj IITB

 αku − uh k2V ≤ a(u − uh , u − uh ) (By coercivity of a(·, ·))
= a(u − uh , u − vh ) + a(u − uh , vh − uh ) (using (2))
| {z }
=0
M
i.e., ku − uh k2V ≤ ku − uh kV ku − vh kV (Boundedness of a(·, ·)).
α
In case u = uh , the result is trivial.
M
For u 6= uh , ku − uh kV ≤ ku − vh kV ∀ vh ∈ Vh .
α
M
⇒ ku − uh kV ≤ C inf ku − vh kV with C = .
vh ∈Vh α

Neela Nataraj IITB

 αku − uh k2V ≤ a(u − uh , u − uh ) (By coercivity of a(·, ·))
= a(u − uh , u − vh ) + a(u − uh , vh − uh ) (using (2))
| {z }
=0
M
i.e., ku − uh k2V ≤ ku − uh kV ku − vh kV (Boundedness of a(·, ·)).
α
In case u = uh , the result is trivial.
M
For u 6= uh , ku − uh kV ≤ ku − vh kV ∀ vh ∈ Vh .
α
M
⇒ ku − uh kV ≤ C inf ku − vh kV with C = .
vh ∈Vh α
Remark: uh is the best approximation of u in V w.r.t
k · ka = a(·, ·)1/2 .

Neela Nataraj IITB

Step 2 : Interpolation operator

Define a map:
Πh : V ∩ C 0 (Ω) −→ Vh
u 7−→ uI
by

Neela Nataraj IITB

Step 2 : Interpolation operator

Define a map:
Πh : V ∩ C 0 (Ω) −→ Vh
u 7−→ uI
by

(Πh u)(x1 , x2 )

Neela Nataraj IITB

Step 2 : Interpolation operator

Define a map:
Πh : V ∩ C 0 (Ω) −→ Vh
u 7−→ uI
by

(Πh u)(x1 , x2 ) = uI (x1 , x2 )

Neela Nataraj IITB

Step 2 : Interpolation operator

Define a map:
Πh : V ∩ C 0 (Ω) −→ Vh
u 7−→ uI
by
N
u(→
−
X
(Πh u)(x1 , x2 ) = uI (x1 , x2 ) = ai )φih (x1 , x2 )
i=1

Neela Nataraj IITB

Step 2 : Interpolation operator

Define a map:
Πh : V ∩ C 0 (Ω) −→ Vh
u 7−→ uI
by
N
u(→
−
X
(Πh u)(x1 , x2 ) = uI (x1 , x2 ) = ai )φih (x1 , x2 ) (3)
i=1

where {→
−
ai }N
i=1 are the nodes of the triangulation

Neela Nataraj IITB

Step 2 : Interpolation operator

Define a map:
Πh : V ∩ C 0 (Ω) −→ Vh
u 7−→ uI
by
N
u(→
−
X
(Πh u)(x1 , x2 ) = uI (x1 , x2 ) = ai )φih (x1 , x2 ) (3)
i=1

where {→
−
ai }N
i=1 are the nodes of the triangulation
i N
and {φh }i=1 is a canonical basis for Vh .

(Πh u)(→
−
ai ) = u(→
−
ai ) 1 ≤ i ≤ N. (4)

Neela Nataraj IITB

Step 3 : Affine mapping

Let T
b represent reference triangle and T , any triangle of τh .

Neela Nataraj IITB

Step 3 : Affine mapping

Let T
b represent reference triangle and T , any triangle of τh .
Then, define a map :

Neela Nataraj IITB

Step 3 : Affine mapping

Let T
b represent reference triangle and T , any triangle of τh .
Then, define a map :
b −→ T by
FT : T

Neela Nataraj IITB

Step 3 : Affine mapping

Let T
b represent reference triangle and T , any triangle of τh .
Then, define a map :
b −→ T by
FT : T

FT (b
x)

Neela Nataraj IITB

Step 3 : Affine mapping

Let T
b represent reference triangle and T , any triangle of τh .
Then, define a map :
b −→ T by
FT : T

FT (b
x) = [BT ]2×2 b
x + [bT ]2×1

Neela Nataraj IITB

Step 3 : Affine mapping

Let T
b represent reference triangle and T , any triangle of τh .
Then, define a map :
b −→ T by
FT : T

FT (b
x) = [BT ]2×2 b
x + [bT ]2×1 where BT = (bi,j )2×2 and
bT = (bi )2×1 are obtained using :

Neela Nataraj IITB

Step 3 : Affine mapping

Let T
b represent reference triangle and T , any triangle of τh .
Then, define a map :
b −→ T by
FT : T

FT (b
x) = [BT ]2×2 b
x + [bT ]2×1 where BT = (bi,j )2×2 and
bT = (bi )2×1 are obtained using :

FT (abi ) = →
−
ai 1 ≤ i ≤ 3.

Neela Nataraj IITB

Step 3 : Affine mapping

Let T
b represent reference triangle and T , any triangle of τh .
Then, define a map :
b −→ T by
FT : T

FT (b
x) = [BT ]2×2 b
x + [bT ]2×1 where BT = (bi,j )2×2 and
bT = (bi )2×1 are obtained using :

FT (abi ) = →
−
ai 1 ≤ i ≤ 3.

If FT (b
x , yb) = (x, y ) then, in matrix form:

Neela Nataraj IITB

Step 3 : Affine mapping

Let T
b represent reference triangle and T , any triangle of τh .
Then, define a map :
b −→ T by
FT : T

FT (b
x) = [BT ]2×2 b
x + [bT ]2×1 where BT = (bi,j )2×2 and
bT = (bi )2×1 are obtained using :

FT (abi ) = →
−
ai 1 ≤ i ≤ 3.

If FT (b
x , yb) = (x, y ) then, in matrix form:
      
x b11 b12 x b1
= + (5)
b
y b21 b22 yb b2

Neela Nataraj IITB

Using : abi −→ →
−
ai , we obtain :

Neela Nataraj IITB

Using : abi −→ →
−
ai , we obtain :

BT

Neela Nataraj IITB

Using : abi −→ →
−
ai , we obtain :
 
b11 b12
BT =
b21 b22

Neela Nataraj IITB

Using : abi −→ →
−
ai , we obtain :
   
b11 b12 a11 − a13 a12 − a13
BT = =
b21 b22 a21 − a23 a22 − a23

Neela Nataraj IITB

Using : abi −→ →
−
ai , we obtain :
   
b11 b12 a11 − a13 a12 − a13
BT = =
b21 b22 a21 − a23 a22 − a23

bT

Neela Nataraj IITB

Using : abi −→ →
−
ai , we obtain :
   
b11 b12 a11 − a13 a12 − a13
BT = =
b21 b22 a21 − a23 a22 − a23
 
a13
bT =
a23

Neela Nataraj IITB

Using : abi −→ →
−
ai , we obtain :
   
b11 b12 a11 − a13 a12 − a13
BT = =
b21 b22 a21 − a23 a22 − a23
 
a13
bT =
a23
The Jacobian of the transformation FT = J(FT ) will be :

Neela Nataraj IITB

Using : abi −→ →
−
ai , we obtain :
   
b11 b12 a11 − a13 a12 − a13
BT = =
b21 b22 a21 − a23 a22 − a23
 
a13
bT =
a23
The Jacobian of the transformation FT = J(FT ) will be :

|J(FT )| =

Neela Nataraj IITB

Using : abi −→ →
−
ai , we obtain :
   
b11 b12 a11 − a13 a12 − a13
BT = =
b21 b22 a21 − a23 a22 − a23
 
a13
bT =
a23
The Jacobian of the transformation FT = J(FT ) will be :

∂x ∂x
|J(FT )| = Abs ∂b
x ∂b
y =
∂y ∂y
∂b
x ∂b
y

Neela Nataraj IITB

Using : abi −→ →
−
ai , we obtain :
   
b11 b12 a11 − a13 a12 − a13
BT = =
b21 b22 a21 − a23 a22 − a23
 
a13
bT =
a23
The Jacobian of the transformation FT = J(FT ) will be :

∂x ∂x
∂b
x ∂b
y b11 b12
|J(FT )| = Abs ∂y ∂y = Abs
b21 b22
∂b
x ∂b
y

Neela Nataraj IITB

Also,
 
x
b
yb

Neela Nataraj IITB

Also,
   −1
x b11 b12
=
b
yb b21 b22

Neela Nataraj IITB

Also,
   −1  
x b11 b12 x − b1
=
b
yb b21 b22 y − b2

Neela Nataraj IITB

Also,
   −1    ∗ ∗
 
x b11 b12 x − b1 b11 b12 x − b1
= =
b
∗ ∗
yb b21 b22 y − b2 b21 b22 y − b2

Neela Nataraj IITB

Also,
   −1    ∗ ∗
 
x b11 b12 x − b1 b11 b12 x − b1
= =
b
∗ ∗
yb b21 b22 y − b2 b21 b22 y − b2

where:

BT−1

Neela Nataraj IITB

Also,
   −1    ∗ ∗
 
x b11 b12 x − b1 b11 b12 x − b1
= =
b
∗ ∗
yb b21 b22 y − b2 b21 b22 y − b2

where:
 
1 b22 −b12
BT−1 =
|BT | −b21 b11

Neela Nataraj IITB

Also,
   −1    ∗ ∗
 
x b11 b12 x − b1 b11 b12 x − b1
= =
b
∗ ∗
yb b21 b22 y − b2 b21 b22 y − b2

where:
∗ ∗
   
1 b22 −b12 b11 b12
BT−1 = = ∗ ∗
|BT | −b21 b11 b21 b22

Neela Nataraj IITB

Also,
   −1    ∗ ∗
 
x b11 b12 x − b1 b11 b12 x − b1
= =
b
∗ ∗
yb b21 b22 y − b2 b21 b22 y − b2

where:
∗ ∗
   
1 b22 −b12 b11 b12
BT−1 = = ∗ ∗
|BT | −b21 b11 b21 b22

The Jacobian of transformation FT−1 = JF −1 will be :

T

Neela Nataraj IITB

Also,
   −1    ∗ ∗
 
x b11 b12 x − b1 b11 b12 x − b1
= =
b
∗ ∗
yb b21 b22 y − b2 b21 b22 y − b2

where:
∗ ∗
   
1 b22 −b12 b11 b12
BT−1 = = ∗ ∗
|BT | −b21 b11 b21 b22

The Jacobian of transformation FT−1 = JF −1 will be :

T

∂b
x ∂b
x
|J(FT −1 )| = Abs ∂x ∂b
y =
∂b
y ∂b
y
∂x ∂y

Neela Nataraj IITB

Also,
   −1    ∗ ∗
 
x b11 b12 x − b1 b11 b12 x − b1
= =
b
∗ ∗
yb b21 b22 y − b2 b21 b22 y − b2

where:
∗ ∗
   
1 b22 −b12 b11 b12
BT−1 = = ∗ ∗
|BT | −b21 b11 b21 b22

The Jacobian of transformation FT−1 = JF −1 will be :

T

∂b
x ∂b
x
∗ ∗
∂x ∂b
y b11 b12
|J(FT −1 )| = Abs ∂b
y ∂b
y = Abs ∗ ∗
b21 b22
∂x ∂y

Neela Nataraj IITB

Also,
   −1    ∗ ∗
 
x b11 b12 x − b1 b11 b12 x − b1
= =
b
∗ ∗
yb b21 b22 y − b2 b21 b22 y − b2

where:
∗ ∗
   
1 b22 −b12 b11 b12
BT−1 = = ∗ ∗
|BT | −b21 b11 b21 b22

The Jacobian of transformation FT−1 = JF −1 will be :

T

∂b
x ∂b
x
∗ ∗
∂x ∂b
y b11 b12
|J(FT −1 )| = Abs ∂b
y ∂b
y = Abs ∗ ∗
b21 b22
∂x ∂y

Thus clearly, |J(FT−1 )| = Abs|[J(FT )]−1 |.

Neela Nataraj IITB

Also,
   −1    ∗ ∗
 
x b11 b12 x − b1 b11 b12 x − b1
= =
b
∗ ∗
yb b21 b22 y − b2 b21 b22 y − b2

where:
∗ ∗
   
1 b22 −b12 b11 b12
BT−1 = = ∗ ∗
|BT | −b21 b11 b21 b22

The Jacobian of transformation FT−1 = JF −1 will be :

T

∂b
x ∂b
x
∗ ∗
∂x ∂b
y b11 b12
|J(FT −1 )| = Abs ∂b
y ∂b
y = Abs ∗ ∗
b21 b22
∂x ∂y

Thus clearly, |J(FT−1 )| = Abs|[J(FT )]−1 |.

Neela Nataraj IITB

Properties of Affine mapping FT

• FT is invertible.

Neela Nataraj IITB

Properties of Affine mapping FT

• FT is invertible.
• FT (T
b) = T.

Neela Nataraj IITB

Properties of Affine mapping FT

• FT is invertible.
• FT (T
b) = T.
• FT carries vertices and sides of T
b to vertices and sides of T .

Neela Nataraj IITB

Properties of Affine mapping FT

• FT is invertible.
• FT (T
b) = T.
• FT carries vertices and sides of T
b to vertices and sides of T .

We would need the following estimates of bij , bij∗ , J(FT ) and JF −1 :

T

Neela Nataraj IITB

Properties of Affine mapping FT

• FT is invertible.
• FT (T
b) = T.
• FT carries vertices and sides of T
b to vertices and sides of T .

We would need the following estimates of bij , bij∗ , J(FT ) and JF −1 :

T
Consider a triangle T . Let α1 , α2 , α3 denote angles of T and
h1 , h2 , h3 denote the sides opposite to these angles.

Neela Nataraj IITB

Properties of Affine mapping FT

• FT is invertible.
• FT (T
b) = T.
• FT carries vertices and sides of T
b to vertices and sides of T .

We would need the following estimates of bij , bij∗ , J(FT ) and JF −1 :

T
Consider a triangle T . Let α1 , α2 , α3 denote angles of T and
h1 , h2 , h3 denote the sides opposite to these angles.
Let
α1 ≤ α2 ≤ α3 ⇒ h1 ≤ h2 ≤ h3

Neela Nataraj IITB

Properties of Affine mapping FT

• FT is invertible.
• FT (T
b) = T.
• FT carries vertices and sides of T
b to vertices and sides of T .

We would need the following estimates of bij , bij∗ , J(FT ) and JF −1 :

T
Consider a triangle T . Let α1 , α2 , α3 denote angles of T and
h1 , h2 , h3 denote the sides opposite to these angles.
Let
α1 ≤ α2 ≤ α3 ⇒ h1 ≤ h2 ≤ h3

Let α = min{αi } = α1

Neela Nataraj IITB

Properties of Affine mapping FT

• FT is invertible.
• FT (T
b) = T.
• FT carries vertices and sides of T
b to vertices and sides of T .

We would need the following estimates of bij , bij∗ , J(FT ) and JF −1 :

T
Consider a triangle T . Let α1 , α2 , α3 denote angles of T and
h1 , h2 , h3 denote the sides opposite to these angles.
Let
α1 ≤ α2 ≤ α3 ⇒ h1 ≤ h2 ≤ h3

Let α = min{αi } = α1 and h = max{hi } = h3

Neela Nataraj IITB

Properties of Affine mapping FT

• FT is invertible.
• FT (T
b) = T.
• FT carries vertices and sides of T
b to vertices and sides of T .

We would need the following estimates of bij , bij∗ , J(FT ) and JF −1 :

T
Consider a triangle T . Let α1 , α2 , α3 denote angles of T and
h1 , h2 , h3 denote the sides opposite to these angles.
Let
α1 ≤ α2 ≤ α3 ⇒ h1 ≤ h2 ≤ h3

Let α = min{αi } = α1 and h = max{hi } = h3

Now, h1 + h2 > h3 and h2 > h1

Neela Nataraj IITB

Properties of Affine mapping FT

• FT is invertible.
• FT (T
b) = T.
• FT carries vertices and sides of T
b to vertices and sides of T .

We would need the following estimates of bij , bij∗ , J(FT ) and JF −1 :

T
Consider a triangle T . Let α1 , α2 , α3 denote angles of T and
h1 , h2 , h3 denote the sides opposite to these angles.
Let
α1 ≤ α2 ≤ α3 ⇒ h1 ≤ h2 ≤ h3

Let α = min{αi } = α1 and h = max{hi } = h3

h3
Now, h1 + h2 > h3 and h2 > h1 ⇒ h2 >
2

Neela Nataraj IITB

Properties of Affine mapping FT

• FT is invertible.
• FT (T
b) = T.
• FT carries vertices and sides of T
b to vertices and sides of T .

We would need the following estimates of bij , bij∗ , J(FT ) and JF −1 :

T
Consider a triangle T . Let α1 , α2 , α3 denote angles of T and
h1 , h2 , h3 denote the sides opposite to these angles.
Let
α1 ≤ α2 ≤ α3 ⇒ h1 ≤ h2 ≤ h3

Let α = min{αi } = α1 and h = max{hi } = h3

h3
Now, h1 + h2 > h3 and h2 > h1 ⇒ h2 >
2
h2 h3 sinα1
Area of triangle T =
2

Neela Nataraj IITB

 h2 sinα h2 h3 sinα1 h2 sinα
⇒ ≤ ≤ .
4 2 2

Neela Nataraj IITB

 h2 sinα h2 h3 sinα1 h2 sinα
⇒ ≤ ≤ .
4 2 2
h2 sinα
Thus, ≤ |J(FT )| ≤ h2 sinα
2

Neela Nataraj IITB

 h2 sinα h2 h3 sinα1 h2 sinα
⇒ ≤ ≤ .
4 2 2
h2 sinα
Thus, ≤ |J(FT )| ≤ h2 sinα
2
1 2 1
2
≤ |J(FT−1 )| ≤ 2 (since|J(FT−1 )| = ).
h sinα h sinα |J(FT )|

Neela Nataraj IITB

 h2 sinα h2 h3 sinα1 h2 sinα
⇒ ≤ ≤ .
4 2 2
h2 sinα
Thus, ≤ |J(FT )| ≤ h2 sinα
2
1 2 1
2
≤ |J(FT−1 )| ≤ 2 (since|J(FT−1 )| = ).
h sinα h sinα |J(FT )|
Again,
|b11 | = |a11 − a13 | ≤ h2 ≤ h.

Neela Nataraj IITB

 h2 sinα h2 h3 sinα1 h2 sinα
⇒ ≤ ≤ .
4 2 2
h2 sinα
Thus, ≤ |J(FT )| ≤ h2 sinα
2
1 2 1
2
≤ |J(FT−1 )| ≤ 2 (since|J(FT−1 )| = ).
h sinα h sinα |J(FT )|
Again,
|b11 | = |a11 − a13 | ≤ h2 ≤ h.
Similarly,
|bij | ≤ h ∀ 1 ≤ i, j ≤ 2 .

Neela Nataraj IITB

 h2 sinα h2 h3 sinα1 h2 sinα
⇒ ≤ ≤ .
4 2 2
h2 sinα
Thus, ≤ |J(FT )| ≤ h2 sinα
2
1 2 1
2
≤ |J(FT−1 )| ≤ 2 (since|J(FT−1 )| = ).
h sinα h sinα |J(FT )|
Again,
|b11 | = |a11 − a13 | ≤ h2 ≤ h.
Similarly,
|bij | ≤ h ∀ 1 ≤ i, j ≤ 2 .

|bij | 2h
|bij∗ | = ≤ 2
|J(FT )| h sinα
Thus,
2
|bij∗ | ≤
hsinα

Neela Nataraj IITB

Step 4 : Change of scale

(We will derive the estimates in reference triangle T

b and then do a
change of scale to get back to T .)
Theorem
Let
•Tb be the reference triangle,

Neela Nataraj IITB

Step 4 : Change of scale

(We will derive the estimates in reference triangle T

b and then do a
change of scale to get back to T .)
Theorem
Let
•T b be the reference triangle,
• T be any generic triangle of τh which is obtained from T
b by the
affine mapping FT defined above,

Neela Nataraj IITB

Step 4 : Change of scale

(We will derive the estimates in reference triangle T

b and then do a
change of scale to get back to T .)
Theorem
Let
•T b be the reference triangle,
• T be any generic triangle of τh which is obtained from T
b by the
affine mapping FT defined above,
• u(x, y ) ∈ H s (T ), u x , yb) ∈ H s (T
b(b b ) satisfy

Neela Nataraj IITB

Step 4 : Change of scale

(We will derive the estimates in reference triangle T

b and then do a
change of scale to get back to T .)
Theorem
Let
•T b be the reference triangle,
• T be any generic triangle of τh which is obtained from T b by the
affine mapping FT defined above,
• u(x, y ) ∈ H s (T ), ub(bx , yb) ∈ H s (T
b ) satisfy
u
b(bx , yb) = u(x, y ), (x, y ) ∈ T , FT (bx , yb) = (x, y ),

Neela Nataraj IITB

Step 4 : Change of scale

(We will derive the estimates in reference triangle T

b and then do a
change of scale to get back to T .)
Theorem
Let
•T b be the reference triangle,
• T be any generic triangle of τh which is obtained from T b by the
affine mapping FT defined above,
• u(x, y ) ∈ H s (T ), ub(bx , yb) ∈ H s (T
b ) satisfy
u
b(bx , yb) = u(x, y ), (x, y ) ∈ T , FT (bx , yb) = (x, y ),
Then, there exists constants Cs , Ds > 0 (independent of u, h and
α) such that:

Neela Nataraj IITB

Step 4 : Change of scale

(We will derive the estimates in reference triangle T

b and then do a
change of scale to get back to T .)
Theorem
Let
•T b be the reference triangle,
• T be any generic triangle of τh which is obtained from T b by the
affine mapping FT defined above,
• u(x, y ) ∈ H s (T ), ub(bx , yb) ∈ H s (T
b ) satisfy
u
b(bx , yb) = u(x, y ), (x, y ) ∈ T , FT (bx , yb) = (x, y ),
Then, there exists constants Cs , Ds > 0 (independent of u, h and
α) such that:

u |s,Tb ≤ Cs (sin α)−1/2 hs−1 |u|s,T

Ds (sin α)s−1/2 hs−1 |u|s,T ≤ |b
(6)

Neela Nataraj IITB

Proof :
(for s=0) : To prove that
D0 (sinα)−1/2 h−1 |u|0,T ≤ |b
u |0,Tb ≤ C0 (sinα)−1/2 h−1 |u|0,T

Neela Nataraj IITB

Proof :
(for s=0) : To prove that
D0 (sinα)−1/2 h−1 |u|0,T ≤ |b
u |0,Tb ≤ C0 (sinα)−1/2 h−1 |u|0,T
We know that

Neela Nataraj IITB

Proof :
(for s=0) : To prove that
D0 (sinα)−1/2 h−1 |u|0,T ≤ |b
u |0,Tb ≤ C0 (sinα)−1/2 h−1 |u|0,T
We know that
Z
|u(x)|2 dT
T

Neela Nataraj IITB

Proof :
(for s=0) : To prove that
D0 (sinα)−1/2 h−1 |u|0,T ≤ |b
u |0,Tb ≤ C0 (sinα)−1/2 h−1 |u|0,T
We know that
Z Z
2
|u(x)| dT = |b x )|2 |J(FT )|d T
u (b b
T T
b

Neela Nataraj IITB

Proof :
(for s=0) : To prove that
D0 (sinα)−1/2 h−1 |u|0,T ≤ |b
u |0,Tb ≤ C0 (sinα)−1/2 h−1 |u|0,T
We know that
Z Z
2
|u(x)| dT = |b x )|2 |J(FT )|d T
u (b b
T T
b

⇒ |u|20,T

Neela Nataraj IITB

Proof :
(for s=0) : To prove that
D0 (sinα)−1/2 h−1 |u|0,T ≤ |b
u |0,Tb ≤ C0 (sinα)−1/2 h−1 |u|0,T
We know that
Z Z
2
|u(x)| dT = |b x )|2 |J(FT )|d T
u (b b
T T
b

u |20,Tb |J(FT )|
⇒ |u|20,T = |b

Neela Nataraj IITB

Proof :
(for s=0) : To prove that
D0 (sinα)−1/2 h−1 |u|0,T ≤ |b
u |0,Tb ≤ C0 (sinα)−1/2 h−1 |u|0,T
We know that
Z Z
2
|u(x)| dT = |b x )|2 |J(FT )|d T
u (b b
T T
b

u |20,Tb |J(FT )|
⇒ |u|20,T = |b
⇒ |J(FT )|−1 |u|20,T

Neela Nataraj IITB

Proof :
(for s=0) : To prove that
D0 (sinα)−1/2 h−1 |u|0,T ≤ |b
u |0,Tb ≤ C0 (sinα)−1/2 h−1 |u|0,T
We know that
Z Z
2
|u(x)| dT = |b x )|2 |J(FT )|d T
u (b b
T T
b

u |20,Tb |J(FT )|
⇒ |u|20,T = |b
⇒ |J(FT )|−1 |u|20,T = |b
u |20,Tb .

Neela Nataraj IITB

Proof :
(for s=0) : To prove that
D0 (sinα)−1/2 h−1 |u|0,T ≤ |b
u |0,Tb ≤ C0 (sinα)−1/2 h−1 |u|0,T
We know that
Z Z
2
|u(x)| dT = |b x )|2 |J(FT )|d T
u (b b
T T
b

u |20,Tb |J(FT )|
⇒ |u|20,T = |b
⇒ |J(FT )|−1 |u|20,T = |b
u |20,Tb .
Using estimates for the inverse of Jacobian, we have
|u|20,T 2|u|20,T
u |20,Tb ≤
≤ |b
h2 sinα h2 sinα

Neela Nataraj IITB

Proof :
(for s=0) : To prove that
D0 (sinα)−1/2 h−1 |u|0,T ≤ |b
u |0,Tb ≤ C0 (sinα)−1/2 h−1 |u|0,T
We know that
Z Z
2
|u(x)| dT = |b x )|2 |J(FT )|d T
u (b b
T T
b

u |20,Tb |J(FT )|
⇒ |u|20,T = |b
⇒ |J(FT )|−1 |u|20,T = |b
u |20,Tb .
Using estimates for the inverse of Jacobian, we have
|u|20,T 2|u|20,T
u |20,Tb ≤
≤ |b
h2 sinα h2 sinα
√
|u|0,T 2|u|0,T
⇒ √ u |0,Tb ≤ √
≤ |b
h sinα h sinα

Neela Nataraj IITB

Proof :
(for s=0) : To prove that
D0 (sinα)−1/2 h−1 |u|0,T ≤ |b
u |0,Tb ≤ C0 (sinα)−1/2 h−1 |u|0,T
We know that
Z Z
2
|u(x)| dT = |b x )|2 |J(FT )|d T
u (b b
T T
b

u |20,Tb |J(FT )|
⇒ |u|20,T = |b
⇒ |J(FT )|−1 |u|20,T = |b
u |20,Tb .
Using estimates for the inverse of Jacobian, we have
|u|20,T 2|u|20,T
u |20,Tb ≤
≤ |b
h2 sinα h2 sinα
√
|u|0,T 2|u|0,T
⇒ √ u |0,Tb ≤ √
≤ |b i.e., the results holds with
h sinα h sinα
√
D0 = 1, C0 = 2.
Neela Nataraj IITB
s=1
To prove that ∃ D1 , C1 such that:

Neela Nataraj IITB

s=1
To prove that ∃ D1 , C1 such that:

u |1,Tb ≤ C1 (sinα)−1/2 h0 |u|1,T

D1 (sinα)1/2 h0 |u|1,T ≤ |b
Z
|u|21,T = (ux2 + uy2 )dT
T
We know u(x, y ) = u
b(b
x , yb)

Neela Nataraj IITB

s=1
To prove that ∃ D1 , C1 such that:

u |1,Tb ≤ C1 (sinα)−1/2 h0 |u|1,T

D1 (sinα)1/2 h0 |u|1,T ≤ |b
Z
|u|21,T = (ux2 + uy2 )dT
T
We know u(x, y ) = u
b(b
x , yb)
ux = u xx + u
bbx b byb ybx

Neela Nataraj IITB

s=1
To prove that ∃ D1 , C1 such that:

u |1,Tb ≤ C1 (sinα)−1/2 h0 |u|1,T

D1 (sinα)1/2 h0 |u|1,T ≤ |b
Z
|u|21,T = (ux2 + uy2 )dT
T
We know u(x, y ) = u
b(b
x , yb)
ux = u xx + u
bbx b byb ybx

uy = u xy + u
bbx b byb yby

Neela Nataraj IITB

s=1
To prove that ∃ D1 , C1 such that:

u |1,Tb ≤ C1 (sinα)−1/2 h0 |u|1,T

D1 (sinα)1/2 h0 |u|1,T ≤ |b
Z
|u|21,T = (ux2 + uy2 )dT
T
We know u(x, y ) = u
b(b
x , yb)
ux = u xx + u
bbx b byb ybx

uy = u xy + u
bbx b byb yby
Hence,
Z
|u|21,T = bbx2 + c2 u
[c1 u byb2 + 2c3 u
bbx u
byb ]|J(FT )|d T
b
T
b

Neela Nataraj IITB

s=1
To prove that ∃ D1 , C1 such that:

u |1,Tb ≤ C1 (sinα)−1/2 h0 |u|1,T

D1 (sinα)1/2 h0 |u|1,T ≤ |b
Z
|u|21,T = (ux2 + uy2 )dT
T
We know u(x, y ) = u
b(b
x , yb)
ux = u xx + u
bbx b byb ybx

uy = u xy + u
bbx b byb yby
Hence,
Z
|u|21,T = bbx2 + c2 u
[c1 u byb2 + 2c3 u
bbx u
byb ]|J(FT )|d T
b
T
b

where
c1 = bxx2 + b
xy2 ,
Neela Nataraj IITB
s=1
To prove that ∃ D1 , C1 such that:

u |1,Tb ≤ C1 (sinα)−1/2 h0 |u|1,T

D1 (sinα)1/2 h0 |u|1,T ≤ |b
Z
|u|21,T = (ux2 + uy2 )dT
T
We know u(x, y ) = u
b(b
x , yb)
ux = u xx + u
bbx b byb ybx

uy = u xy + u
bbx b byb yby
Hence,
Z
|u|21,T = bbx2 + c2 u
[c1 u byb2 + 2c3 u
bbx u
byb ]|J(FT )|d T
b
T
b

where
c1 = b xy2 , c2 = ybx2 + yby2 ,
xx2 + b
Neela Nataraj IITB
s=1
To prove that ∃ D1 , C1 such that:

u |1,Tb ≤ C1 (sinα)−1/2 h0 |u|1,T

D1 (sinα)1/2 h0 |u|1,T ≤ |b
Z
|u|21,T = (ux2 + uy2 )dT
T
We know u(x, y ) = u
b(b
x , yb)
ux = u xx + u
bbx b byb ybx

uy = u xy + u
bbx b byb yby
Hence,
Z
|u|21,T = bbx2 + c2 u
[c1 u byb2 + 2c3 u
bbx u
byb ]|J(FT )|d T
b
T
b

where
c1 = b xy2 , c2 = ybx2 + yby2 , c3 = b
xx2 + b xx ybx + b
xy yby .
Neela Nataraj IITB
 a2 + b 2
Using ab ≤ ,
2

Neela Nataraj IITB

 a2 + b 2
Using ab ≤ ,
2
Z " bbx2 + u
u byb2
!#
|u|21,T ≤ c1 u byb2 + 2c3
bbx2 + c2 u |J(FT )|dT
T
b 2

Neela Nataraj IITB

 a2 + b 2
Using ab ≤ ,
2
Z " bbx2 + u
u byb2
!#
|u|21,T ≤ c1 u byb2 + 2c3
bbx2 + c2 u |J(FT )|dT
T
b 2
Z
= ubx2 + (c2 + c3 )b
[(c1 + c3 )b uyb2 )]|J(FT )|dT
T
b

Neela Nataraj IITB

 a2 + b 2
Using ab ≤ ,
2
Z " bbx2 + u
u byb2
!#
|u|21,T ≤ c1 u byb2 + 2c3
bbx2 + c2 u |J(FT )|dT
T
b 2
Z
= ubx2 + (c2 + c3 )b
[(c1 + c3 )b uyb2 )]|J(FT )|dT
T
b
Z
≤ max{c1 + c3 , c2 + c3 } ubx2 + u
(b byb2 )|J(FT )|dT
T
b

Neela Nataraj IITB

 a2 + b 2
Using ab ≤ ,
2
Z " bbx2 + u
u byb2
!#
|u|21,T ≤ c1 u byb2 + 2c3
bbx2 + c2 u |J(FT )|dT
T
b 2
Z
= ubx2 + (c2 + c3 )b
[(c1 + c3 )b uyb2 )]|J(FT )|dT
T
b
Z
≤ max{c1 + c3 , c2 + c3 } ubx2 + u
(b byb2 )|J(FT )|dT
T
b

⇒ |u|21,T ≤ C |b
u |21,Tb |J(FT )|

Neela Nataraj IITB

 a2 + b 2
Using ab ≤ ,
2
Z " bbx2 + u
u byb2
!#
|u|21,T ≤ c1 u byb2 + 2c3
bbx2 + c2 u |J(FT )|dT
T
b 2
Z
= ubx2 + (c2 + c3 )b
[(c1 + c3 )b uyb2 )]|J(FT )|dT
T
b
Z
≤ max{c1 + c3 , c2 + c3 } ubx2 + u
(b byb2 )|J(FT )|dT
T
b

⇒ |u|21,T ≤ C |b
u |21,Tb |J(FT )|
16
Now, C = max{c1 + c3 , c2 + c3 }, where : c1 + c3 ≤ ,
h2 sin2 α

Neela Nataraj IITB

 a2 + b 2
Using ab ≤ ,
2
Z " bbx2 + u
u byb2
!#
|u|21,T ≤ c1 u byb2 + 2c3
bbx2 + c2 u |J(FT )|dT
T
b 2
Z
= ubx2 + (c2 + c3 )b
[(c1 + c3 )b uyb2 )]|J(FT )|dT
T
b
Z
≤ max{c1 + c3 , c2 + c3 } ubx2 + u
(b byb2 )|J(FT )|dT
T
b

⇒ |u|21,T ≤ C |b
u |21,Tb |J(FT )|
16
Now, C = max{c1 + c3 , c2 + c3 }, where : c1 + c3 ≤ 2 2 ,
h sin α
16
c2 + c3 ≤ 2 2 .
h sin α

Neela Nataraj IITB

 a2 + b 2
Using ab ≤ ,
2
Z " bbx2 + u
u byb2
!#
|u|21,T ≤ c1 u byb2 + 2c3
bbx2 + c2 u |J(FT )|dT
T
b 2
Z
= ubx2 + (c2 + c3 )b
[(c1 + c3 )b uyb2 )]|J(FT )|dT
T
b
Z
≤ max{c1 + c3 , c2 + c3 } ubx2 + u
(b byb2 )|J(FT )|dT
T
b

⇒ |u|21,T ≤ C |b
u |21,Tb |J(FT )|
16
Now, C = max{c1 + c3 , c2 + c3 }, where : c1 + c3 ≤ 2 2 ,
h sin α
16
c2 + c3 ≤ 2 2 . Using the estimates for Jacobian, we have
h sin α

Neela Nataraj IITB

 a2 + b 2
Using ab ≤ ,
2
Z " bbx2 + u
u byb2
!#
|u|21,T ≤ c1 u byb2 + 2c3
bbx2 + c2 u |J(FT )|dT
T
b 2
Z
= ubx2 + (c2 + c3 )b
[(c1 + c3 )b uyb2 )]|J(FT )|dT
T
b
Z
≤ max{c1 + c3 , c2 + c3 } ubx2 + u
(b byb2 )|J(FT )|dT
T
b

⇒ |u|21,T ≤ C |b
u |21,Tb |J(FT )|
16
Now, C = max{c1 + c3 , c2 + c3 }, where : c1 + c3 ≤ 2 2 ,
h sin α
16
c2 + c3 ≤ 2 2 . Using the estimates for Jacobian, we have
h sin α
2 16
|u|1,T ≤ u |2
|b
sinα 1,Tb

Neela Nataraj IITB

 a2 + b 2
Using ab ≤ ,
2
Z " bbx2 + u
u byb2
!#
|u|21,T ≤ c1 u byb2 + 2c3
bbx2 + c2 u |J(FT )|dT
T
b 2
Z
= ubx2 + (c2 + c3 )b
[(c1 + c3 )b uyb2 )]|J(FT )|dT
T
b
Z
≤ max{c1 + c3 , c2 + c3 } ubx2 + u
(b byb2 )|J(FT )|dT
T
b

⇒ |u|21,T ≤ C |b
u |21,Tb |J(FT )|
16
Now, C = max{c1 + c3 , c2 + c3 }, where : c1 + c3 ≤ 2 2 ,
h sin α
16
c2 + c3 ≤ 2 2 . Using the estimates for Jacobian, we have
h sin α
2 16 1
|u|1,T ≤ u |2 ⇒ (sinα)1/2 |u|1,T ≤ |b
|b u |1,Tb .
sinα 1,Tb 4

Neela Nataraj IITB

 a2 + b 2
Using ab ≤ ,
2
Z " bbx2 + u
u byb2
!#
|u|21,T ≤ c1 u byb2 + 2c3
bbx2 + c2 u |J(FT )|dT
T
b 2
Z
= ubx2 + (c2 + c3 )b
[(c1 + c3 )b uyb2 )]|J(FT )|dT
T
b
Z
≤ max{c1 + c3 , c2 + c3 } ubx2 + u
(b byb2 )|J(FT )|dT
T
b

⇒ |u|21,T ≤ C |b
u |21,Tb |J(FT )|
16
Now, C = max{c1 + c3 , c2 + c3 }, where : c1 + c3 ≤ 2 2 ,
h sin α
16
c2 + c3 ≤ 2 2 . Using the estimates for Jacobian, we have
h sin α
2 16 1
|u|1,T ≤ u |2 ⇒ (sinα)1/2 |u|1,T ≤ |b
|b u |1,Tb .
sinα 1,Tb 4
Thus, LHS of the required result holds with D1 = 1/4.
Neela Nataraj IITB
Similarly, we can get the right hand side estimate :

u |21,Tb ≤ |J(FT−1 )|C 0 |u|21,T

|b

Neela Nataraj IITB

Similarly, we can get the right hand side estimate :

u |21,Tb ≤ |J(FT−1 )|C 0 |u|21,T

|b

where:

C 0 = max{c10 +c20 , c30 +c20 },

Neela Nataraj IITB

Similarly, we can get the right hand side estimate :

u |21,Tb ≤ |J(FT−1 )|C 0 |u|21,T

|b

where:

C 0 = max{c10 +c20 , c30 +c20 }, c10 = xbx2 +xyb2 ,

Neela Nataraj IITB

Similarly, we can get the right hand side estimate :

u |21,Tb ≤ |J(FT−1 )|C 0 |u|21,T

|b

where:

C 0 = max{c10 +c20 , c30 +c20 }, c10 = xbx2 +xyb2 , c20 = ybx2 +yyb2 ,

Neela Nataraj IITB

Similarly, we can get the right hand side estimate :

u |21,Tb ≤ |J(FT−1 )|C 0 |u|21,T

|b

where:

C 0 = max{c10 +c20 , c30 +c20 }, c10 = xbx2 +xyb2 , c20 = ybx2 +yyb2 , c30 = xbx ybx +xyb yyb .

Neela Nataraj IITB

Similarly, we can get the right hand side estimate :

u |21,Tb ≤ |J(FT−1 )|C 0 |u|21,T

|b

where:

C 0 = max{c10 +c20 , c30 +c20 }, c10 = xbx2 +xyb2 , c20 = ybx2 +yyb2 , c30 = xbx ybx +xyb yyb .

From from estimates obtained for BT and its inverse,

Neela Nataraj IITB

Similarly, we can get the right hand side estimate :

u |21,Tb ≤ |J(FT−1 )|C 0 |u|21,T

|b

where:

C 0 = max{c10 +c20 , c30 +c20 }, c10 = xbx2 +xyb2 , c20 = ybx2 +yyb2 , c30 = xbx ybx +xyb yyb .

From from estimates obtained for BT and its inverse, C 0 ≤ 4h2 .

Also,
8
u |21,Tb ≤
|b |u|2
sinα 1,T
√
8
⇒ |bu |1,Tb ≤ √ |u|1,T
sinα
Thus, RHS of the required inequality also holds.

Neela Nataraj IITB

To establish the result for s > 1 we need an analog of ux , uy for
derivatives of order 0 s‘.

Neela Nataraj IITB

To establish the result for s > 1 we need an analog of ux , uy for
derivatives of order 0 s‘.
Note that every partial derivative of u of order s is a polynomial of
degree s w.r.t. b
xx , b
xy , ybx , yby with no term having degree less than s.

Neela Nataraj IITB

To establish the result for s > 1 we need an analog of ux , uy for
derivatives of order 0 s‘.
Note that every partial derivative of u of order s is a polynomial of
degree s w.r.t. b
xx , b
xy , ybx , yby with no term having degree less than s.
Summing the squares of all s th order derivatives and using
a2 + b 2
ab ≤ to eliminate cross-products, we get:
2

Neela Nataraj IITB

To establish the result for s > 1 we need an analog of ux , uy for
derivatives of order 0 s‘.
Note that every partial derivative of u of order s is a polynomial of
degree s w.r.t. b xx , b
xy , ybx , yby with no term having degree less than s.
Summing the squares of all s th order derivatives and using
a2 + b 2
ab ≤ to eliminate cross-products, we get:
2
|u|2s,T ≤ |J(FT )|C |b u |2s,Tb ,

Neela Nataraj IITB

To establish the result for s > 1 we need an analog of ux , uy for
derivatives of order 0 s‘.
Note that every partial derivative of u of order s is a polynomial of
degree s w.r.t. b xx , b
xy , ybx , yby with no term having degree less than s.
Summing the squares of all s th order derivatives and using
a2 + b 2
ab ≤ to eliminate cross-products, we get:
2
|u|2s,T ≤ |J(FT )|C |b u |2s,Tb ,
where C = O((hsinα)−2s ).

Neela Nataraj IITB

To establish the result for s > 1 we need an analog of ux , uy for
derivatives of order 0 s‘.
Note that every partial derivative of u of order s is a polynomial of
degree s w.r.t. b xx , b
xy , ybx , yby with no term having degree less than s.
Summing the squares of all s th order derivatives and using
a2 + b 2
ab ≤ to eliminate cross-products, we get:
2
|u|2s,T ≤ |J(FT )|C |b u |2s,Tb ,
where C = O((hsinα)−2s ).
Similarly, reversing the process we get :

Neela Nataraj IITB

To establish the result for s > 1 we need an analog of ux , uy for
derivatives of order 0 s‘.
Note that every partial derivative of u of order s is a polynomial of
degree s w.r.t. b xx , b
xy , ybx , yby with no term having degree less than s.
Summing the squares of all s th order derivatives and using
a2 + b 2
ab ≤ to eliminate cross-products, we get:
2
|u|2s,T ≤ |J(FT )|C |b u |2s,Tb ,
where C = O((hsinα)−2s ).
Similarly, reversing the process we get :

u |2s,Tb ≤ |J(FT−1 )|C 0 |u|2s,T ,

|b

where C 0 = O(h2s ).

Neela Nataraj IITB

To establish the result for s > 1 we need an analog of ux , uy for
derivatives of order 0 s‘.
Note that every partial derivative of u of order s is a polynomial of
degree s w.r.t. b xx , b
xy , ybx , yby with no term having degree less than s.
Summing the squares of all s th order derivatives and using
a2 + b 2
ab ≤ to eliminate cross-products, we get:
2
|u|2s,T ≤ |J(FT )|C |b u |2s,Tb ,
where C = O((hsinα)−2s ).
Similarly, reversing the process we get :

u |2s,Tb ≤ |J(FT−1 )|C 0 |u|2s,T ,

|b

where C 0 = O(h2s ).
Using upper bounds for |J(FT )| and |J(FT−1 )|, we get the required
result.

Neela Nataraj IITB

Step 5 : Bramble- Hilbert Lemma

Let
• Ω be a sufficiently smooth domain,

Neela Nataraj IITB

Step 5 : Bramble- Hilbert Lemma

Let
• Ω be a sufficiently smooth domain,
• L : H k+1 (Ω) −→ R is a bounded linear funcional,

Neela Nataraj IITB

Step 5 : Bramble- Hilbert Lemma

Let
• Ω be a sufficiently smooth domain,
• L : H k+1 (Ω) −→ R is a bounded linear funcional,

i.e., |L(u)| ≤ M k u kk+1,Ω ∀u ∈ H k+1 (Ω), M > 0

Neela Nataraj IITB

Step 5 : Bramble- Hilbert Lemma

Let
• Ω be a sufficiently smooth domain,
• L : H k+1 (Ω) −→ R is a bounded linear funcional,

i.e., |L(u)| ≤ M k u kk+1,Ω ∀u ∈ H k+1 (Ω), M > 0

• L(p) = 0 ∀p ∈ Pk (Ω).
Then, there exists a constant C > 0 depending only on Ω such
that:

Neela Nataraj IITB

Step 5 : Bramble- Hilbert Lemma

Let
• Ω be a sufficiently smooth domain,
• L : H k+1 (Ω) −→ R is a bounded linear funcional,

i.e., |L(u)| ≤ M k u kk+1,Ω ∀u ∈ H k+1 (Ω), M > 0

• L(p) = 0 ∀p ∈ Pk (Ω).
Then, there exists a constant C > 0 depending only on Ω such
that:
|L(u)| ≤ CM|u|k+1,Ω ∀u ∈ H k+1 (Ω).

Neela Nataraj IITB

Step 6 : Semi-norm inner product

We will define an inner product which will give rise to a semi-norm.

Neela Nataraj IITB

Step 6 : Semi-norm inner product

We will define an inner product which will give rise to a semi-norm.

For u ∈ H k+1 (Ω), (k ≥ 1)
XZ
[u, v ]s,Ω = D α u(x)D α v (x)dΩ
|α|=s Ω

Now, [u, u]s,Ω = |u|2s,Ω .

Neela Nataraj IITB

Step 7 : Finite element interpolation over the reference
triangle
b ∈ H k+1 (T
Given some integer k ≥ 1 and some u b ), consider the
interpolant :

Neela Nataraj IITB

Step 7 : Finite element interpolation over the reference
triangle
b ∈ H k+1 (T
Given some integer k ≥ 1 and some u b ), consider the
interpolant :
n
X
u
bI (b
x , yb) = u
b(b
ai )φbi (b x , yb) ∈ T
x , yb) , (b b
i=1

Neela Nataraj IITB

Step 7 : Finite element interpolation over the reference
triangle
b ∈ H k+1 (T
Given some integer k ≥ 1 and some u b ), consider the
interpolant :
n
X
u
bI (b
x , yb) = u
b(b
ai )φbi (b x , yb) ∈ T
x , yb) , (b b
i=1

where b
ai are nodes in the reference triangles and n denotes the
number of nodes in a triangle of the triangulation.

Neela Nataraj IITB

Step 7 : Finite element interpolation over the reference
triangle
b ∈ H k+1 (T
Given some integer k ≥ 1 and some u b ), consider the
interpolant :
n
X
u
bI (b
x , yb) = u
b(b
ai )φbi (b x , yb) ∈ T
x , yb) , (b b
i=1

where bai are nodes in the reference triangles and n denotes the
number of nodes in a triangle of the triangulation.
In this case n = 3, φbi (b
x , yb) denotes the element nodal canonical
basis functions in T .
b
Differentiating ,
n
X
D αu
bI (b
x , yb) = u ai )D α φbi (b
b(b x , yb) ∈ T
x , yb) , (b b.
i=1

Neela Nataraj IITB

Step 7 : Finite element interpolation over the reference
triangle
b ∈ H k+1 (T
Given some integer k ≥ 1 and some u b ), consider the
interpolant :
n
X
u
bI (b
x , yb) = u
b(b
ai )φbi (b x , yb) ∈ T
x , yb) , (b b
i=1

where bai are nodes in the reference triangles and n denotes the
number of nodes in a triangle of the triangulation.
In this case n = 3, φbi (b
x , yb) denotes the element nodal canonical
basis functions in T .
b
Differentiating ,
n
X
D αu
bI (b
x , yb) = u ai )D α φbi (b
b(b x , yb) ∈ T
x , yb) , (b b.
i=1

Since the standard local basis functions in T

b are polynomials,
derivatives of higher order vanish.
Neela Nataraj IITB
So, ∃M > 0 such that :

Neela Nataraj IITB

So, ∃M > 0 such that :

|D α φbi (b
x , yb)| ≤ M

Neela Nataraj IITB

So, ∃M > 0 such that :

|D α φbi (b
x , yb)| ≤ M ∀(b
x , yb) ∈ T
b , i = 1, 2, ..., n

Neela Nataraj IITB

So, ∃M > 0 such that :

|D α φbi (b
x , yb)| ≤ M ∀(b
x , yb) ∈ T
b , i = 1, 2, ..., n

Also, since H k+1 ,→ C 0 as k ≥ 1

|b ai )| ≤ max(bx ,by )∈Tb |b

u (b x , yb)| ≤ η k u
u (b u ∈ H k+1 (T
b kk+1,Tb ∀b b ).

Neela Nataraj IITB

So, ∃M > 0 such that :

|D α φbi (b
x , yb)| ≤ M ∀(b
x , yb) ∈ T
b , i = 1, 2, ..., n

Also, since H k+1 ,→ C 0 as k ≥ 1

|b ai )| ≤ max(bx ,by )∈Tb |b

u (b x , yb)| ≤ η k u
u (b u ∈ H k+1 (T
b kk+1,Tb ∀b b ).

Here η depends on k and T

b . Hence,

|D α u x , yb)| ≤ ηnM k u
bI (b u ∈ H k+1 (T
b kk+1,Tb ∀b b ).

Neela Nataraj IITB

So, ∃M > 0 such that :

|D α φbi (b
x , yb)| ≤ M ∀(b
x , yb) ∈ T
b , i = 1, 2, ..., n

Also, since H k+1 ,→ C 0 as k ≥ 1

|b ai )| ≤ max(bx ,by )∈Tb |b

u (b x , yb)| ≤ η k u
u (b u ∈ H k+1 (T
b kk+1,Tb ∀b b ).

Here η depends on k and T

b . Hence,

|D α u x , yb)| ≤ ηnM k u
bI (b u ∈ H k+1 (T
b kk+1,Tb ∀b b ).

Forming the Sobolev norms of ubI (of which only a finite number
are non-zero),we see that ∃M
b independent of u b and s, such that:
k+1
|b
uI | b ≤ M
s,T
b kubk b ∀b
k+1,T
u∈H (T
b ) , s = 0, 1, 2, ...

Neela Nataraj IITB

Theorem
Let
•Tb be the reference finite element.

Neela Nataraj IITB

Theorem
Let
•Tb be the reference finite element.
• {φbi }ni=1 be the standard local basis functions on T
b.

Neela Nataraj IITB

Theorem
Let
•Tb be the reference finite element.
• {φbi }ni=1 be the standard local basis functions on Tb.
• Let k be the largest integer such that Pk (T b ) ⊆ SPAN{φbi }n
i=1

Neela Nataraj IITB

Theorem
Let
•Tb be the reference finite element.
• {φbi }ni=1 be the standard local basis functions on Tb.
• Let k be the largest integer such that Pk (T b ) ⊆ SPAN{φbi }n
i=1
Then, there exists a constant C b , independent of u b such that

|b
u−u
bI |s,Tb ≤ C
b |b u ∈ H k+1 (T
u |k+1,Tb ∀b b ) s = 0, 1, ..., k + 1.

Proof:
Let 0 ≤ s ≤ k + 1 .

Neela Nataraj IITB

Theorem
Let
•Tb be the reference finite element.
• {φbi }ni=1 be the standard local basis functions on Tb.
• Let k be the largest integer such that Pk (T b ) ⊆ SPAN{φbi }n
i=1
Then, there exists a constant C b , independent of u b such that

|b
u−u
bI |s,Tb ≤ C
b |b u ∈ H k+1 (T
u |k+1,Tb ∀b b ) s = 0, 1, ..., k + 1.

Proof:
Let 0 ≤ s ≤ k + 1 .
Fix vb ∈ H s (T
b ).

Neela Nataraj IITB

Theorem
Let
•Tb be the reference finite element.
• {φbi }ni=1 be the standard local basis functions on Tb.
• Let k be the largest integer such that Pk (T b ) ⊆ SPAN{φbi }n
i=1
Then, there exists a constant C b , independent of u b such that

|b
u−u
bI |s,Tb ≤ C
b |b u ∈ H k+1 (T
u |k+1,Tb ∀b b ) s = 0, 1, ..., k + 1.

Proof:
Let 0 ≤ s ≤ k + 1 .
Fix vb ∈ H s (T
b ).
k+1 b
Let L : H (T ) −→ R be a linear functional defined by:

Neela Nataraj IITB

Theorem
Let
•Tb be the reference finite element.
• {φbi }ni=1 be the standard local basis functions on Tb.
• Let k be the largest integer such that Pk (T b ) ⊆ SPAN{φbi }n
i=1
Then, there exists a constant C b , independent of u b such that

|b
u−u
bI |s,Tb ≤ C
b |b u ∈ H k+1 (T
u |k+1,Tb ∀b b ) s = 0, 1, ..., k + 1.

Proof:
Let 0 ≤ s ≤ k + 1 .
Fix vb ∈ H s (T
b ).
k+1 b
Let L : H (T ) −→ R be a linear functional defined by:
L(b u−u
u ) = [b bI , vb]s,Tb

Neela Nataraj IITB

Also, |L(b
u )| =

Neela Nataraj IITB

Also, |L(b
u )| = |[b
u−u
bI , vb]s,Tb |

Neela Nataraj IITB

Also, |L(b
u )| = |[b
u−u
bI , vb]s,Tb | ≤ |b
u−u
bI |s,Tb |b
v |s,Tb

Neela Nataraj IITB

Also, |L(b
u )| = |[b
u−u
bI , vb]s,Tb | ≤ |b
u−u
bI |s,Tb |b
v |s,Tb

≤ (|b
u |s,Tb + |b
uI |s,Tb )(|b
v |s,Tb )

Neela Nataraj IITB

Also, |L(b
u )| = |[b
u−u
bI , vb]s,Tb | ≤ |b
u−u
bI |s,Tb |b
v |s,Tb

≤ (|b
u |s,Tb + |b
uI |s,Tb )(|b
v |s,Tb )

≤ (1 + M)|b
b v| b k u
s,T
b kk+1,Tb .

Neela Nataraj IITB

 Also, |L(b
u )| = |[b
u−u
bI , vb]s,Tb | ≤ |b
u−u
bI |s,Tb |b
v |s,Tb

≤ (|b
u |s,Tb + |b
uI |s,Tb )(|b
v |s,Tb )

≤ (1 + M)|b
b v| b k u
s,T
b kk+1,Tb .

So, L is a bounded linear functional on H k+1 (T

b ).

Neela Nataraj IITB

 Also, |L(b
u )| = |[b
u−u
bI , vb]s,Tb | ≤ |b
u−u
bI |s,Tb |b
v |s,Tb

≤ (|b
u |s,Tb + |b
uI |s,Tb )(|b
v |s,Tb )

≤ (1 + M)|b
b v| b k u
s,T
b kk+1,Tb .

So, L is a bounded linear functional on H k+1 (T

b ).
Now,

b ∈ Pk (T
u b) ⇒

Neela Nataraj IITB

 Also, |L(b
u )| = |[b
u−u
bI , vb]s,Tb | ≤ |b
u−u
bI |s,Tb |b
v |s,Tb

≤ (|b
u |s,Tb + |b
uI |s,Tb )(|b
v |s,Tb )

≤ (1 + M)|b
b v| b k u
s,T
b kk+1,Tb .

So, L is a bounded linear functional on H k+1 (T

b ).
Now,

b ∈ Pk (T
u b) ⇒ u
bI = u
b

Neela Nataraj IITB

 Also, |L(b
u )| = |[b
u−u
bI , vb]s,Tb | ≤ |b
u−u
bI |s,Tb |b
v |s,Tb

≤ (|b
u |s,Tb + |b
uI |s,Tb )(|b
v |s,Tb )

≤ (1 + M)|b
b v| b k u
s,T
b kk+1,Tb .

So, L is a bounded linear functional on H k+1 (T

b ).
Now,

b ∈ Pk (T
u b) ⇒ u b ⇒ L(b
bI = u u−u
u ) = [b bI , vb]s,Tb

Neela Nataraj IITB

 Also, |L(b
u )| = |[b
u−u
bI , vb]s,Tb | ≤ |b
u−u
bI |s,Tb |b
v |s,Tb

≤ (|b
u |s,Tb + |b
uI |s,Tb )(|b
v |s,Tb )

≤ (1 + M)|b
b v| b k u
s,T
b kk+1,Tb .

So, L is a bounded linear functional on H k+1 (T

b ).
Now,

b ∈ Pk (T
u b) ⇒ u b ⇒ L(b
bI = u u−u
u ) = [b bI , vb]s,Tb = 0.

Neela Nataraj IITB

 Also, |L(b
u )| = |[b
u−u
bI , vb]s,Tb | ≤ |b
u−u
bI |s,Tb |b
v |s,Tb

≤ (|b
u |s,Tb + |b
uI |s,Tb )(|b
v |s,Tb )

≤ (1 + M)|b
b v| b k u
s,T
b kk+1,Tb .

So, L is a bounded linear functional on H k+1 (T

b ).
Now,

b ∈ Pk (T
u b) ⇒ u b ⇒ L(b
bI = u u−u
u ) = [b bI , vb]s,Tb = 0.

So, by Bramble-Hilbert lemma , ∃ a contant C such that

Neela Nataraj IITB

 Also, |L(b
u )| = |[b
u−u
bI , vb]s,Tb | ≤ |b
u−u
bI |s,Tb |b
v |s,Tb

≤ (|b
u |s,Tb + |b
uI |s,Tb )(|b
v |s,Tb )

≤ (1 + M)|b
b v| b k u
s,T
b kk+1,Tb .

So, L is a bounded linear functional on H k+1 (T

b ).
Now,

b ∈ Pk (T
u b) ⇒ u b ⇒ L(b
bI = u u−u
u ) = [b bI , vb]s,Tb = 0.

So, by Bramble-Hilbert lemma , ∃ a contant C such that

|L(b
u )| ≤ C (1 + M)|b
b v | b |b
s,T
u ∈ H k+1 (T
u |k+1,Tb ∀b b ).

Neela Nataraj IITB

 Also, |L(b
u )| = |[b
u−u
bI , vb]s,Tb | ≤ |b
u−u
bI |s,Tb |b
v |s,Tb

≤ (|b
u |s,Tb + |b
uI |s,Tb )(|b
v |s,Tb )

≤ (1 + M)|b
b v| b k u
s,T
b kk+1,Tb .

So, L is a bounded linear functional on H k+1 (T

b ).
Now,

b ∈ Pk (T
u b) ⇒ u b ⇒ L(b
bI = u u−u
u ) = [b bI , vb]s,Tb = 0.

So, by Bramble-Hilbert lemma , ∃ a contant C such that

|L(b
u )| ≤ C (1 + M)|b
b v | b |b
s,T
u ∈ H k+1 (T
u |k+1,Tb ∀b b ).

Neela Nataraj IITB

 b−u
Put vb = u bI to obtain:

Neela Nataraj IITB

 b−u
Put vb = u bI to obtain:

|b bI |2s,Tb ≤ C (1 + M)|b
u−u b u−ubI |s,Tb |b
u |k+1,Tb

Neela Nataraj IITB

 b−u
Put vb = u bI to obtain:

|b bI |2s,Tb ≤ C (1 + M)|b
u−u b u−ubI |s,Tb |b
u |k+1,Tb

|b
u−u
bI |s,Tb ≤ C
b |b
u |k+1,Tb u ∈ H k+1 (T
∀b b ), s = 0, 1, ..., k + 1.

Neela Nataraj IITB

 b−u
Put vb = u bI to obtain:

|b bI |2s,Tb ≤ C (1 + M)|b
u−u b u−ubI |s,Tb |b
u |k+1,Tb

|b
u−u
bI |s,Tb ≤ C
b |b
u |k+1,Tb u ∈ H k+1 (T
∀b b ), s = 0, 1, ..., k + 1.

Neela Nataraj IITB

Step 8 : Error Estimates

Theorem
Let
• Ω be a polygon in R2

Neela Nataraj IITB

Step 8 : Error Estimates

Theorem
Let
• Ω be a polygon in R2
• τh be a triangulation of Ω into closed triangles.

Neela Nataraj IITB

Step 8 : Error Estimates

Theorem
Let
• Ω be a polygon in R2
• τh be a triangulation of Ω into closed triangles.
• The associated canonical nodal basis functions be {φi }N
i=1 .

Neela Nataraj IITB

Step 8 : Error Estimates

Theorem
Let
• Ω be a polygon in R2
• τh be a triangulation of Ω into closed triangles.
• The associated canonical nodal basis functions be {φi }N
i=1 .
• h, α denote the greatest edge length and the smallest angle
respectively in the mesh.

Neela Nataraj IITB

Step 8 : Error Estimates

Theorem
Let
• Ω be a polygon in R2
• τh be a triangulation of Ω into closed triangles.
• The associated canonical nodal basis functions be {φi }N
i=1 .
• h, α denote the greatest edge length and the smallest angle
respectively in the mesh.
• k be the largest integer for which
Pk (Ω) ⊆ Vh = SPAN{φ1 , φ2 , ..., φN }.

Neela Nataraj IITB

Step 8 : Error Estimates

Theorem
Let
• Ω be a polygon in R2
• τh be a triangulation of Ω into closed triangles.
• The associated canonical nodal basis functions be {φi }N
i=1 .
• h, α denote the greatest edge length and the smallest angle
respectively in the mesh.
• k be the largest integer for which
Pk (Ω) ⊆ Vh = SPAN{φ1 , φ2 , ..., φN }.
Then, ∃ a value C , independent of u and the mesh, such that

|u − uI |s ≤ C (sin α)−s hk+1−s |u|k+1 ∀u ∈ H k+1 (Ω), s = 0, 1,

where uI is the interpolant of u.

Neela Nataraj IITB

Step 8 : Error Estimates

Theorem
Let
• Ω be a polygon in R2
• τh be a triangulation of Ω into closed triangles.
• The associated canonical nodal basis functions be {φi }N
i=1 .
• h, α denote the greatest edge length and the smallest angle
respectively in the mesh.
• k be the largest integer for which
Pk (Ω) ⊆ Vh = SPAN{φ1 , φ2 , ..., φN }.
Then, ∃ a value C , independent of u and the mesh, such that

|u − uI |s ≤ C (sin α)−s hk+1−s |u|k+1 ∀u ∈ H k+1 (Ω), s = 0, 1,

where uI is the interpolant of u.

Neela Nataraj IITB

Remark : The global basis functions for Vh belong to C 0 (Ω).

Neela Nataraj IITB

Remark : The global basis functions for Vh belong to C 0 (Ω).
Hence, it is typical that the first order derivative of global basis
functions and of the interpolant uI , have step discontinuties across
element boundaries;

Neela Nataraj IITB

Remark : The global basis functions for Vh belong to C 0 (Ω).
Hence, it is typical that the first order derivative of global basis
functions and of the interpolant uI , have step discontinuties across
element boundaries;
with the result that,in general uI does not belong to H 2 (Ω).

Neela Nataraj IITB

Remark : The global basis functions for Vh belong to C 0 (Ω).
Hence, it is typical that the first order derivative of global basis
functions and of the interpolant uI , have step discontinuties across
element boundaries;
with the result that,in general uI does not belong to H 2 (Ω).
Hence, the Sobolev-seminorm |u − uI |s must be restricted to
s = 0, 1.
(In the earlier theorem,this restriction was not necessary, as the
domain here is a single element.)

Neela Nataraj IITB

Proof:
We apply change of scale result twice and Finite element
interpolation estimate once to obtain the required result.

Neela Nataraj IITB

Proof:
We apply change of scale result twice and Finite element
interpolation estimate once to obtain the required result.
Let T ∈ τh where h and α are the greatest edge length and
smallest angle, respectively of that element.

Neela Nataraj IITB

Proof:
We apply change of scale result twice and Finite element
interpolation estimate once to obtain the required result.
Let T ∈ τh where h and α are the greatest edge length and
smallest angle, respectively of that element.
Then,

|u − uI |2s,T ≤ Ds−2 (sin α)1−2s h2−2s |b bI |2s,Tb

u−u

Neela Nataraj IITB

Proof:
We apply change of scale result twice and Finite element
interpolation estimate once to obtain the required result.
Let T ∈ τh where h and α are the greatest edge length and
smallest angle, respectively of that element.
Then,

|u − uI |2s,T ≤ Ds−2 (sin α)1−2s h2−2s |b bI |2s,Tb

u−u

≤ Ds−2 (sin α)1−2s h2−2s C

b 2 |b
u |2k+1,Tb

Neela Nataraj IITB

Proof:
We apply change of scale result twice and Finite element
interpolation estimate once to obtain the required result.
Let T ∈ τh where h and α are the greatest edge length and
smallest angle, respectively of that element.
Then,

|u − uI |2s,T ≤ Ds−2 (sin α)1−2s h2−2s |b bI |2s,Tb

u−u

≤ Ds−2 (sin α)1−2s h2−2s C

b 2 |b
u |2k+1,Tb

≤ Ds−2 (sin α)1−2s h2−2s C

b 2 C 2 (sin α)−1 h2k |u|2
k+1 k+1,T

Neela Nataraj IITB

Proof:
We apply change of scale result twice and Finite element
interpolation estimate once to obtain the required result.
Let T ∈ τh where h and α are the greatest edge length and
smallest angle, respectively of that element.
Then,

|u − uI |2s,T ≤ Ds−2 (sin α)1−2s h2−2s |b bI |2s,Tb

u−u

≤ Ds−2 (sin α)1−2s h2−2s C

b 2 |b
u |2k+1,Tb

≤ Ds−2 (sin α)1−2s h2−2s C

b 2 C 2 (sin α)−1 h2k |u|2
k+1 k+1,T

Neela Nataraj IITB

Proof:
We apply change of scale result twice and Finite element
interpolation estimate once to obtain the required result.
Let T ∈ τh where h and α are the greatest edge length and
smallest angle, respectively of that element.
Then,

|u − uI |2s,T ≤ Ds−2 (sin α)1−2s h2−2s |b bI |2s,Tb

u−u

≤ Ds−2 (sin α)1−2s h2−2s C

b 2 |b
u |2k+1,Tb

≤ Ds−2 (sin α)1−2s h2−2s C

b 2 C 2 (sin α)−1 h2k |u|2
k+1 k+1,T

(s=0 or 1)

≤ Ds−2 CK2 +1 (sin α)−2s h2k+2−2s C

b 2 |u|2
k+1,T

Neela Nataraj IITB

Summing over all elements, we get

Neela Nataraj IITB

Summing over all elements, we get

|u − uI |2s,Ω ≤ C 2 (sin α)−2s h2k+2−2s |u|2k+1,Ω

Neela Nataraj IITB

Summing over all elements, we get

|u − uI |2s,Ω ≤ C 2 (sin α)−2s h2k+2−2s |u|2k+1,Ω

where C 2 = Ck+1
2 b2
C /min{D0 , D1 } = 4C
b Ck+1

Neela Nataraj IITB

Summing over all elements, we get

|u − uI |2s,Ω ≤ C 2 (sin α)−2s h2k+2−2s |u|2k+1,Ω

where C 2 = Ck+1
2 b2
C /min{D0 , D1 } = 4C
b Ck+1
i.e.

|u − uI |s,Ω ≤ C (sin α)−s hk+1−s |u|k+1,Ω ∀u ∈ H k+1 (Ω) , s = 0, 1.

Neela Nataraj IITB

Summing over all elements, we get

|u − uI |2s,Ω ≤ C 2 (sin α)−2s h2k+2−2s |u|2k+1,Ω

where C 2 = Ck+1
2 b2
C /min{D0 , D1 } = 4C
b Ck+1
i.e.

|u − uI |s,Ω ≤ C (sin α)−s hk+1−s |u|k+1,Ω ∀u ∈ H k+1 (Ω) , s = 0, 1.

k u − uh kV can be computed using this.

Neela Nataraj IITB

Summing over all elements, we get

|u − uI |2s,Ω ≤ C 2 (sin α)−2s h2k+2−2s |u|2k+1,Ω

where C 2 = Ck+1
2 b2
C /min{D0 , D1 } = 4C
b Ck+1
i.e.

|u − uI |s,Ω ≤ C (sin α)−s hk+1−s |u|k+1,Ω ∀u ∈ H k+1 (Ω) , s = 0, 1.

k u − uh kV can be computed using this.

We obtain, k u − uh kV ≤ Ch|u|2,Ω ∀u ∈ H 2 (Ω).

Neela Nataraj IITB

Summing over all elements, we get

|u − uI |2s,Ω ≤ C 2 (sin α)−2s h2k+2−2s |u|2k+1,Ω

where C 2 = Ck+1
2 b2
C /min{D0 , D1 } = 4C
b Ck+1
i.e.

|u − uI |s,Ω ≤ C (sin α)−s hk+1−s |u|k+1,Ω ∀u ∈ H k+1 (Ω) , s = 0, 1.

k u − uh kV can be computed using this.

We obtain, k u − uh kV ≤ Ch|u|2,Ω ∀u ∈ H 2 (Ω).
Using Aubin-Nitsche duality arguments, we can show

Neela Nataraj IITB

Summing over all elements, we get

|u − uI |2s,Ω ≤ C 2 (sin α)−2s h2k+2−2s |u|2k+1,Ω

where C 2 = Ck+1
2 b2
C /min{D0 , D1 } = 4C
b Ck+1
i.e.

|u − uI |s,Ω ≤ C (sin α)−s hk+1−s |u|k+1,Ω ∀u ∈ H k+1 (Ω) , s = 0, 1.

k u − uh kV can be computed using this.

We obtain, k u − uh kV ≤ Ch|u|2,Ω ∀u ∈ H 2 (Ω).
Using Aubin-Nitsche duality arguments, we can show
k u − uh kL2 (Ω) ≤ Ch2 |u|2,Ω ∀u ∈ H 2 (Ω).

Neela Nataraj IITB