Ziane Achour University of Djelfa Course: Applied numerical

Faculty of Science and Technology methods and optimization

Department of Electrical Engineering Instructor: Dr. Ali TETA
Master01 ELT

Chapter 02 Solving partial differential equations (PDE)

2.1 Introduction of Partial Differential Equations (PDE)

2.1.1 Definitions
- A partial differential equation or PDE is a relation involving an unknown function y of Rn in R, the
𝑑𝑓 𝑑𝑓 𝑑2 𝑓 𝑑2 𝑓 𝑑𝑚 𝑓
variables 𝑥1 , 𝑥2 , … , 𝑥𝑛 and its partial derivatives, (𝑑𝑥 , 𝑑𝑥 , … , 𝑑𝑥 2 , 𝑑𝑥 , … , 𝑑𝑥 𝑚 ).
1 2 1 1 𝑑𝑥2 𝑛

- It is generally written as:

𝑑𝑓 𝑑𝑓 𝑑2 𝑓 𝑑2 𝑓 𝑑2 𝑓 𝑑𝑚 𝑦
𝐹(𝑥1 , 𝑥2 , … , 𝑓, 𝑑𝑥 , 𝑑𝑥 , … , 𝑑𝑥 2 , 𝑑𝑥 2 , 𝑑𝑥 , … , 𝑑𝑥 𝑚 ) = 0 (2.1)
1 2 1 2 1 𝑑𝑥2 𝑛
- Equation (2.1) is considered in a domain Ω of Rn.
- The solutions to equation (2.1) are functions that satisfy this equation in Ω.
2.1.2 Example
𝑑2 𝑓 𝑑2 𝑓
− 𝑑𝑥 2 = 0 (2.2)
𝑑𝑥12 2
The function 𝑓(𝑥1 , 𝑥2 ) = (𝑥1 − 𝑥2 )3 is the solution of (2.2).

2.1.3 Classification of PDE:

Second order linear differential equations can be expressed in the following general form:
𝑑2 𝑓 𝑑2 𝑓 𝑑2 𝑓
𝐴 𝑑𝑥 2 + 𝐵 𝑑𝑥 + 𝐶 𝑑𝑥 2 + 𝐷 = 0 (2.3)
1 1 𝑑𝑥2 2

Differential equations of the form (2.3) can be classified into three categories according to the
coefficients A,B,C.

𝐵 2 − 4𝐴𝐶 Category Example

Laplace equation (2D)
<0 Elliptic 𝑑2𝑇 𝑑2𝑇
+ =0
𝑑𝑥 2 𝑑𝑦 2
Heat equation (1D)
=0 Parabolic 𝑑𝑇 𝑑2 𝑇
=𝑐 2
𝑑𝑡 𝑑𝑥
Wave propagation (1D)
>0 Hyperbolic 𝑑2 𝑇 𝑑2𝑇
= 𝑐
𝑑𝑡 2 𝑑𝑥 2

1
Ziane Achour University of Djelfa Course: Applied numerical
Faculty of Science and Technology methods and optimization
Department of Electrical Engineering Instructor: Dr. Ali TETA
Master01 ELT

2.1.4 PDE applications:

Generally speaking, the modeling of physical phenomena is based on the resolution of partial differential
equations. These equations correspond to the mathematical translation of the laws of physics:
- Fluid mechanics: Navier-Stockes equations
- Electromagnetism: Maxwell's equations
- Thermal: heat equation
- Quantum mechanics: Schrödinger equation
In engineering sciences, modeling a physical device (electric motor for example) makes it possible to
predict its behavior and study the influence of parameters on its performance. The validity of the model
must be compared, when possible, to experimental measurements. In a large majority of cases, the partial
differential equations are non-linear (because of the properties of the materials) and we must use the
computer to solve them (numerical calculation software). When possible (linear equations), it is
interesting to solve the model equations analytically (“by hand”). In this case, the solution obtained
allows us to see the influence of the different parameters. It can be used for a first optimization study of
the device studied.

2.2 Finite Difference method for PDE-Elliptic Equations-

2.2.1 Definitions:
- Elliptic partial differential equations can be of the form:

𝑑2 𝑉 𝑑2 𝑉
+ =0 (2.4)
𝑑𝑥 2 𝑑𝑦 2

Where the sum of the second derivative of the dependent variable V with respect to x and y is zero (the
Laplace equation). This elliptic PDE represents the distribution of the electrostatic potential (V) in
regions where there are no charges.
- The second form of elliptic PDE is:
𝑑2 𝑉 𝑑2 𝑉
+ 𝑑𝑦 2 = 𝑓(𝑥, 𝑦) (2.5)
𝑑𝑥 2

Where the sum of the second derivative of the dependent variable V with respect to x and y is a given
function f(x, y) which describes electrostatic potential in the presence of charges. Equation (6.5) is called
(the Poisson equation).

2
Ziane Achour University of Djelfa Course: Applied numerical
Faculty of Science and Technology methods and optimization
Department of Electrical Engineering Instructor: Dr. Ali TETA
Master01 ELT

2.2.2 Finite Difference method (FDM):

For solving the Laplace equation (6.6) using FDM we follow the following steps:

1- We divide the two-dimensional region into small regions with increments in the x and y directions
given by ∆x and ∆y, as shown in Figure 6.1. Each node is denoted by a numbering scheme i and j,
where i indicates the increment of x and j indicates the increment of y.

Figure 2.1 Grid of the finite difference solution of the Laplace equation.

2- Afterwards we discretize the PDE by replacing the partial derivatives by the approximate centered
derivatives for each node.
A finite difference equation fitted to the interior nodes of a stable two-dimensional system can be
obtained by considering the Laplace equation at the nodal point i, j as:
𝑑2 𝑉 𝑑2 𝑉
| + 𝑑𝑦 2 | =0 (2.6)
𝑑𝑥 2 𝑖,𝑗 𝑖,𝑗
The second derivatives at node (i, j) can be approximated (using Taylor series) as:
𝑑2 𝑉 𝑉𝑖+1,𝑗 −2𝑉𝑖,𝑗 +𝑉𝑖−1,𝑗
| ≈ (2.7)
𝑑𝑥 2 𝑖,𝑗 (𝛥𝑥)2

𝑑2 𝑉 𝑉𝑖,𝑗+1 −2𝑉𝑖,𝑗 +𝑉𝑖,𝑗−1

| ≈ (2.8)
𝑑𝑦 2 𝑖,𝑗 (𝛥𝑦)2

Therefore, the equation (6.6) becomes:

𝑉𝑖+1,𝑗 −2𝑉𝑖,𝑗 +𝑉𝑖−1,𝑗 𝑉𝑖,𝑗+1 −2𝑉𝑖,𝑗 +𝑉𝑖,𝑗−1

+ =0 (2.9)
(𝛥𝑥)2 (𝛥𝑦)2

Assuming ∆x=∆y, the finite difference approximation of Laplace's equation for interior points can be
expressed as:
𝑉𝑖+1,𝑗 + 𝑉𝑖−1,𝑗 +𝑉𝑖,𝑗+1 + 𝑉𝑖,𝑗−1 − 4𝑉𝑖,𝑗 = 0 (2.10)

3- We use equation (6.12) for each interior point and apply the boundary conditions to obtain a linear
system, which can be solved by the Jacobi or Gauss-Seidel methods.
3
Ziane Achour University of Djelfa Course: Applied numerical
Faculty of Science and Technology methods and optimization
Department of Electrical Engineering Instructor: Dr. Ali TETA
Master01 ELT

2.2.3 PDE Boundary Conditions

A- Dirichlet Boundary Conditions:

This condition consists of imposing the value of the solution (the electrostatic potential) on the boundary.
The primary variable value is therefore specified at the limits, like the example (2.10):
𝑉𝑖,0 = 50𝑜 𝑐 (2.11)

Example 2.1:
We consider the dielectric system (with Dirichlet boundary conditions) presented in figure 2.2 with
Δx=Δy.

Figure 2.2 dielectric system with 4 interior points.

The Dirichlet boundary conditions for this system are:

𝑉10 = 𝑉20 = 0𝑉
𝑉01 = 𝑉02 = 750𝑉
𝑉13 = 𝑉23 = 1000𝑉
𝑉31 = 𝑉32 = 500𝑉
The electrostatic potential of the interior points can be described as:
𝑉𝑖+1,𝑗 + 𝑉𝑖−1,𝑗 +𝑉𝑖,𝑗+1 + 𝑉𝑖,𝑗−1 − 4𝑉𝑖,𝑗 = 0

- Node (1,1):
𝑉21 + 𝑉01 +𝑉12 + 𝑉10 − 4𝑉11 = 0
With boundary conditions:
𝑉21 + 750+𝑉12 + 0 − 4𝑉11 = 0
Therefore:
𝟒𝑽𝟏𝟏 − 𝑽𝟐𝟏 −𝑽𝟏𝟐 = 𝟕𝟓𝟎

4
Ziane Achour University of Djelfa Course: Applied numerical
Faculty of Science and Technology methods and optimization
Department of Electrical Engineering Instructor: Dr. Ali TETA
Master01 ELT

- Node (2,1)
𝑉31 + 𝑉11 +𝑉22 + 𝑉20 − 4𝑉21 = 0
With boundary conditions:
500 + 𝑉11 +𝑉22 + 0 − 4𝑉21 = 0
Therefore:
𝟒𝑽𝟐𝟏 − 𝑽𝟏𝟏 −𝑽𝟐𝟐 = 𝟓𝟎𝟎
We apply the same development for the other interior nodes. We finally find a linear system of four
equations with four unknowns:
4𝑉11 − 𝑉21 −𝑉12 = 750
4𝑉21 − 𝑉11 −𝑉22 = 500
4𝑉12 − 𝑉11 − 𝑉22 = 1750
4𝑉22 − 𝑉21 − 𝑉12 = 1500
We can write the linear system (6.21) in matrix form as follows:

4 −1 −1 0 𝑉11 750
−1 4 0 −1 𝑉 500
[ ] [ 21 ] = [ ]
−1 0 4 −1 𝑉12 1750
0 −1 −1 4 𝑉22 1500
We solve the linear system above using the iterative Jacobi method (discussed in chapter 01)
In Jacobi technique, we have:
𝑉𝑖+1,𝑗 (𝑖) + 𝑉𝑖−1,𝑗 (𝑖) + 𝑉𝑖,𝑗+1 (𝑖) + 𝑉𝑖,𝑗−1 (𝑖)
𝑉𝑖,𝑗 (𝑖+1) =
4
We assume that the initial values for the interior nodes are 𝑉𝑖,𝑗 (0) = 0
First iteration (i=0)

- Node (1,1) :
(1) 750 + 𝑉21 (0) + 𝑉12 (0) 750 + 0 + 0
𝑉11 = = = 187.5𝑉
4 4
- Node (2,1) :
(1) 500 + 𝑉11 (0) + 𝑉22 (0) 500 + 0 + 0
𝑉21 = = = 125𝑉
4 4
- Node (1,2) :
(1) 1750 + 𝑉11 (0) + 𝑉22 (0) 1750 + 0 + 0
𝑉12 = = = 437.5𝑉
4 4

- Node (2,2) :
(1) 1500 + 𝑉21 (0) + 𝑉12 (0) 1500 + 0 + 0
𝑉22 = = = 375𝑉
4 4

5
Ziane Achour University of Djelfa Course: Applied numerical
Faculty of Science and Technology methods and optimization
Department of Electrical Engineering Instructor: Dr. Ali TETA
Master01 ELT

B- Neumann Boundary Conditions:

Neumann boundary conditions specify the derivative of the unknown function at the boundary rather
than specifying the function itself, as in the example 2.1:
𝑑𝑉
𝐷 = −𝜀 (2.12)
𝑑𝑥

Where D is electric flux density, 𝜀 is the permittivity of the material and V is the electrostatic potential.

Neumann Boundary Conditions for elliptic equation: For the electrostatic potential problem, this
conditions is by specifying the electric flux density D rather than the boundary electrostatic potential. If
one edge of a dielectric system is isolated, in this case the derivative is zero. To find the potential at the
node (i,0) on the lower edge of the dielectric system presented in figure (2.3), we use equation (2.10)
which becomes:
𝑉𝑖+1,0 + 𝑉𝑖−1,0 +𝑉𝑖,1 + 𝑉𝑖,−1 − 4𝑉𝑖,0 = 0 (2.13)

Figure 2.3: Node (i, 0) on the lower edge.

We can notice that we have an imaginary node (i,-1) necessary for equation (2.13). But, this node is
located outside the system. To solve this problem we use the Neumann boundary condition, which can
be done by representing the first derivative in the y dimension at node (i, 0) by:
𝑑𝑉 𝑉𝑖,1 −𝑉𝑖,−1
≃ (2.14)
𝑑𝑦 2𝛥𝑦

The equation of the potential 𝑉𝑖,−1 at the imaginary node is:

𝑑𝑉
𝑉𝑖,−1 ≃ 𝑉𝑖,1 − 2𝛥𝑦 𝑑𝑦 (2.15)

We substitute the equation of 𝑉𝑖,−1 (2.15) in the equation (2.13):

𝑑𝑉
𝑉𝑖+1,0 + 𝑉𝑖−1,0 +2𝑉𝑖,1 − 2𝛥𝑦 𝑑𝑦 − 4𝑉𝑖,0 = 0 (2.16)

6
Ziane Achour University of Djelfa Course: Applied numerical
Faculty of Science and Technology methods and optimization
Department of Electrical Engineering Instructor: Dr. Ali TETA
Master01 ELT

Example 2.2: Giving a dielectric system with an insulated lower edge presented in the figure 2.4 with
𝛥𝑥 = 𝛥𝑦

Figure 2.4: A dielectric system with an insulated bottom edge

1- For the interior points we use the general equation which is

𝑉𝑖+1,𝑗 + 𝑉𝑖−1,𝑗 +𝑉𝑖,𝑗+1 + 𝑉𝑖,𝑗−1 − 4𝑉𝑖,𝑗 = 0

2- For the points located at the insulated edge we use the following equation:
𝑑𝑉
𝑉𝑖+1,0 + 𝑉𝑖−1,0 +2𝑉𝑖,1 − 2𝛥𝑦 − 4𝑉𝑖,0 = 0
𝑑𝑦
𝑑𝑉
For an insulated edge 𝑑𝑦 = 0

The equation (2.16) becomes:

𝑉𝑖+1,0 + 𝑉𝑖−1,0 +2𝑉𝑖,1 − 4𝑉𝑖,0 = 0

3- The boundary conditions for this system are:

𝑉00 = 𝑉01 = 750𝑉
𝑉12 = 𝑉22 = 𝑉32 = 1000𝑉
𝑉40 = 𝑉41 = 500𝑉

4- We find the potential equation for the interior and insulated nodes
 Interior points:
- Node (1,1):
𝑉21 + 𝑉01 +𝑉12 + 𝑉10 − 4𝑉11 = 0
With boundary conditions:
𝑉21 + 750+1000 + 𝑉10 − 4𝑉11 = 0
Therefore:
𝟒𝑽𝟏𝟏 − 𝑽𝟐𝟏 −𝑽𝟏𝟎 = 𝟏𝟕𝟓𝟎

- Node (2,1)
𝟒𝑽𝟐𝟏 − 𝑽𝟑𝟏 − 𝑽𝟏𝟏 − 𝑽𝟐𝟎 = 𝟏𝟎𝟎𝟎

- Node (3,1)
𝟒𝑽𝟑𝟏 − 𝑽𝟐𝟏 − 𝑽𝟑𝟎 = 𝟏𝟓𝟎𝟎
7
Ziane Achour University of Djelfa Course: Applied numerical
Faculty of Science and Technology methods and optimization
Department of Electrical Engineering Instructor: Dr. Ali TETA
Master01 ELT

 Points at the insulated edge:

We use equation (2.16)
- Node (1,0) :
𝑉20 + 𝑉00 +2𝑉11 − 4𝑉1,0 = 0
With boundary conditions:
𝑉20 + 750+2𝑉11 − 4𝑉1,0 = 0
Therefore:
𝟒𝑽𝟏,𝟎 − 𝑽𝟐𝟎 −𝟐𝑽𝟏𝟏 = 𝟕𝟓𝟎

- Node (2,0) :
𝟒𝑽𝟐,𝟎 − 𝑽𝟏,𝟎 −𝑽𝟑𝟎 −𝟐𝑽𝟐𝟏 = 𝟎

- Node (3,0)
𝟒𝑽𝟑,𝟎 − 𝑽𝟐𝟎 −𝟐𝑽𝟑𝟏 = 𝟓𝟎𝟎

We finally find a linear system of 06 equations with 06 unknown potentials (03 interior nodes and 03
nodes at the insulated edge):
4𝑉11 − 𝑉21 − 𝑉10 = 1750
4𝑉21 − 𝑉31 − 𝑉11 − 𝑉20 = 1000
4𝑉31 − 𝑉21 − 𝑉30 = 1500
4𝑉1,0 − 𝑉20 −2𝑉11 = 750
4𝑉2,0 − 𝑉1,0 −𝑉30 −2𝑉21 = 0
4𝑉3,0 − 𝑉20 −2𝑉31 = 500

This system can be written using the matrix form as follows:

4 −1 0 −1 0 0 𝑉11 𝑉10
−1 4 −1 0 −1 0 𝑉21 𝑉20
0 −1 4 0 0 −1 𝑉31 𝑉30
=
−2 0 0 4 −1 0 𝑉10 𝑉11
0 −2 0 −1 4 −1 𝑉20 𝑉31
[0 0 −2 0 −1 4 ] [ 𝑉30 ] [ 𝑉12 ]
5- We solve this linear system using Jacobi method
For the interior nodes we use:
𝑉𝑖+1,𝑗 (𝑘) + 𝑉𝑖−1,𝑗 (𝑘) + 𝑉𝑖,𝑗+1 (𝑘) + 𝑉𝑖,𝑗−1 (𝑘)
𝑉𝑖,𝑗 (𝑘+1) =
4
For the insulated nodes we use:
(𝑘+1) 𝑉𝑖+1,0 (𝑘) + 𝑉𝑖−1,0 (𝑘) + 2𝑉𝑖,1 (𝑘)
𝑉𝑖,0 =
4
Giving an initial vector of electrostatic potentials equal to 𝑉𝑖,𝑗 (0) = 0
After one iteration (k=0) we find:
𝑉11 = 437.5𝑉, 𝑉21 = 250𝑉, 𝑉31 = 375𝑉
𝑉10 = 187.5𝑉, 𝑉20 = 0𝑉, 𝑉31 = 125𝑉
8
Ziane Achour University of Djelfa Course: Applied numerical
Faculty of Science and Technology methods and optimization
Department of Electrical Engineering Instructor: Dr. Ali TETA
Master01 ELT

C- Secondary variables:
The electrostatic potential is considered to be the main variable in the electrostatic potential problem in
a dielectric system, because its distribution is described by Laplace's equation. For this case, as well as
for other problems involving PDEs, the secondary variables may also be of interest. In fact, in some
engineering contexts, the secondary variable may actually be more important.
In this case the secondary variable is the electric flux density, this quantity can be calculated from
Fourier's law. The central finite difference approximations for the first derivatives can be calculated to
give the following values for the electric flux density in the x and y dimensions:
𝑉𝑖+1,𝑗 −𝑉𝑖−1,𝑗
𝐷𝑥 = −𝜀 (2.17)
2𝛥𝑥
And
𝑉𝑖,𝑗+1 −𝑉𝑖,𝑗−1
𝐷𝑦 = −𝜀 (2.18)
2𝛥𝑦
Where D is the electric flux density, ε is the permittivity of the material, and V is the electrostatic
potential.
The resulting electric flux density can be calculated from these two quantities by:
𝐷𝑛 = √𝐷𝑥 2 + 𝐷𝑦 2 (2.19)
Where the direction of 𝐷𝑛 is given by:
- If 𝐷𝑥 > 0 et 𝐷𝑦 > 0
𝐷𝑦
𝜃 = 𝑡𝑎𝑛−1 (𝐷 ) (2.20)
𝑥
- If 𝐷𝑥 < 0 et 𝐷𝑦 < 0
𝐷𝑦
𝜃 = 𝑡𝑎𝑛−1 (𝐷 ) + 𝜋 (2.21)
𝑥
- If 𝐷𝑥 = 0 et 𝐷𝑦 > 0 ⇒ 𝜃 = 𝜋/2
- If 𝐷𝑥 = 0 et 𝐷𝑦 < 0 ⇒ 𝜃 = 3𝜋/2
Example 2.3:
Suppose the two-dimensional “empty” system shown in Figure 2.5 is 30m x 30m (Δx=Δy=10m) with
permittivity ε=1. Use the previous example results to find the electric flux density (D) and their angles
(θ) for the interior points.
Using the pervious example results we have:

Figure 2.5: A dielectric system with electrostatic potential values.

9
Ziane Achour University of Djelfa Course: Applied numerical
Faculty of Science and Technology methods and optimization
Department of Electrical Engineering Instructor: Dr. Ali TETA
Master01 ELT

The electric flux density for the interior points is calculated by the equations 2.17:

Node (1,1):
𝑉𝑖+1,𝑗 − 𝑉𝑖−1,𝑗
𝐷𝑥 = −𝜀
2𝛥𝑥
So :
125−750
𝐷𝑥 = −1 =31.25
2(10)

And
𝑉𝑖,𝑗+1 − 𝑉𝑖,𝑗−1
𝐷𝑦 = −𝜀
2𝛥𝑦
So:
437.5−0
𝐷𝑦 = −1 = -21.875
2(10)
Then the electric flux density is:
𝐷1,1 = √𝐷𝑥 2 + 𝐷𝑦 2 = 38.145 𝐶/𝑚2
And the angle 𝜃1,1 is:
𝐷𝑦 −21.875
𝜃1,1 = 𝑡𝑎𝑛−1 ( ) = 𝑡𝑎𝑛−1 ( ) = −34.99𝑜
𝐷𝑥 31.25
Node (2,1) :
𝐷𝑥 = −15.625
𝐷𝑦 = −18.75
Then
𝐷2,1 = 24.407 𝐶/𝑚2

For the angle 𝜃2,1 we have 𝐷𝑥 < 0 et 𝐷𝑦 < 0 therefore:

𝐷𝑦
𝜃1,2 = 𝑡𝑎𝑛−1 ( ) + 𝜋 = 230.19𝑜
𝐷𝑥

Node (1,2) :
𝐷1,2 = 43.895 𝐶/𝑚2
And
𝜃1,2 = −67.739𝑜

Node (2,2) :
𝐷2,2 = 43.861 𝐶/𝑚2

And for the angle 𝜃2,2 we have 𝐷𝑥 < 0 et 𝐷𝑦 < 0 therefore:

10
Ziane Achour University of Djelfa Course: Applied numerical
Faculty of Science and Technology methods and optimization
Department of Electrical Engineering Instructor: Dr. Ali TETA
Master01 ELT

𝐷𝑦
𝜃1,2 = 𝑡𝑎𝑛−1 ( ) + 𝜋 = 265.91𝑜
𝐷𝑥
The electric flux density vectors (D) and their angles (𝜃) for the interior points are shown in the
following figure

Figure 2.6 Electric flux density values and angles.

D-Irregular boundaries:
Although the rectangular system of Figure 2.1 has served well to illustrate the fundamental aspects of
solving elliptic PDEs, many engineering problems do not have such an idealized geometry. For example,
many systems have irregular boundaries (Figure 2.7).

Figure 2.7 Grid for a system with irregular boundaries.

Figure 2.7 is a system that can be used to illustrate how non-rectangular boundaries can be handled. As
shown, the lower left boundary of the system is circular. The first derivatives in the x dimension can be
approximated as:
𝑑𝑉 𝑉𝑖,𝑗 −𝑉𝑖−1,𝑗
(𝑑𝑥 ) ≅ (2.22)
𝑖−1,𝑖 𝛼1 𝛥𝑥
And
𝑑𝑉 𝑉𝑖+1,𝑗 −𝑉𝑖,𝑗
(𝑑𝑥 ) ≅ (2.23)
𝑖,𝑖+1 𝛼2 𝛥𝑥

11
Ziane Achour University of Djelfa Course: Applied numerical
Faculty of Science and Technology methods and optimization
Department of Electrical Engineering Instructor: Dr. Ali TETA
Master01 ELT

Second derivatives can be developed from these first derivatives

𝑑𝑉 𝑑𝑉
( ) −( )
𝑑2 𝑉 𝑑 𝑑𝑉 𝑑𝑥 𝑖,𝑖+1 𝑑𝑥 𝑖−1,𝑖
= 𝑑𝑥 (𝑑𝑥 ) = 𝛼1 𝛥𝑥+𝛼2 𝛥𝑥 (2.24)
𝑑𝑥 2
2
Substituting Equations 2.22 and 2.23 into 2.24, we obtain:
𝑉𝑖+1,𝑗−𝑉𝑖,𝑗 𝑉𝑖,𝑗−𝑉𝑖−1,𝑗
𝑑2 𝑉 −
𝛼2 𝛥𝑥 𝛼1 𝛥𝑥
=2 (2.25)
𝑑𝑥 2 𝛼1 𝛥𝑥+𝛼2 𝛥𝑥
Collecting terms gives:
𝑑2 𝑉 2 𝑉 −𝑉 𝑉 −𝑉
= 𝛥𝑥 2 [𝛼 𝑖+1,𝑗 𝑖,𝑗
+ 𝛼 𝑖−1,𝑗 𝑖,𝑗
] (2.26)
𝑑𝑥 2 (𝛼 +𝛼 )
2 1 (𝛼 +𝛼 )
2 1 1 2
A similar equation can be developed for the y dimension:
𝑑2 𝑉 2 𝑉 −𝑉 𝑉 −𝑉
= 𝛥𝑦 2 [𝛽𝑖,𝑗+1 𝑖,𝑗
+ 𝛽𝑖,𝑗−1 𝑖,𝑗
] (2.27)
𝑑𝑦 2 (𝛽 +𝛽 ) 2 1 (𝛽 +𝛽 )2 1 1 2
Using equation (2.9), we find:
2 𝑉𝑖+1,𝑗 −𝑉𝑖,𝑗 𝑉 −𝑉 2 𝑉 −𝑉 𝑉 −𝑉
+ 𝛼 𝑖−1,𝑗 𝑖,𝑗
] + 𝛥𝑦 2 [𝛽𝑖,𝑗+1 𝑖,𝑗
+ 𝛽𝑖,𝑗−1 𝑖,𝑗
[ ]=0 (2.28)
𝛥𝑥 2 𝛼2 (𝛼1 +𝛼2 ) (𝛼 +𝛼 )
1 1 2 (𝛽 +𝛽 ) (𝛽 +𝛽 )
2 1 2 1 1 2
Example 2.4: We have the dielectric system with circular boundaries as shown in Figure 2.8. Find the
electrostatic potential at the interior points given that:
𝛥𝑥 = 𝛥𝑦 = 1
𝛼1 = 𝛽1 = 0.7
𝛼2 = 𝛽2 = 1

Figure 2.8 Dielectric system with irregular boundaries.

For the node (1,1) near the irregular boundary we use the equation 2.28:
2 𝑉21 − 𝑉11 𝑉01 − 𝑉11 2 𝑉12 − 𝑉11 𝑉10 − 𝑉11
[ + ] + [ + ]=0
𝛥𝑥 2 𝛼2 (𝛼1 +𝛼2 ) 𝛼1 (𝛼1 +𝛼2 ) 𝛥𝑦 2 𝛽2 (𝛽1 +𝛽2 ) 𝛽1 (𝛽1 +𝛽2 )
NA:
𝑉21 − 𝑉11 𝑉01 − 𝑉11 𝑉12 − 𝑉11 𝑉10 − 𝑉11
2[ + ]+ 2[ + ]=0
0.7(0.7 + 0.7) 0.7(0.7 + 0.7) 0.7(0.7 + 0.7) 0.7(0.7 + 0.7)

With boundary conditions:

𝑉21 − 𝑉11 750 − 𝑉11 𝑉12 − 𝑉11 750 − 𝑉11
2[ + ]+ 2[ + ]=0
0.98 0.98 0.98 0.98

12
Ziane Achour University of Djelfa Course: Applied numerical
Faculty of Science and Technology methods and optimization
Department of Electrical Engineering Instructor: Dr. Ali TETA
Master01 ELT

8.16𝑉11 − 2.040𝑉21 − 2.040𝑉12 = 3060

For the interior points we use equation 5.12, we find:
4𝑉21 − 𝑉11 − 𝑉22 = 1250
4𝑉12 − 𝑉11 − 𝑉22 = 1750
4𝑉22 − 𝑉12 − 𝑉21 = 1500
We take the initial values for the interior nodes and the isolated edge 𝑉𝑖,𝑗 (0) = 0
In the first iteration (k=0) by the Jacobi method, we find:
𝑉11 = 375𝑉 𝑉21 = 312.5𝑉
𝑉12 = 437.5𝑉 𝑉22 = 375𝑉

13