Classification of Partial Differential Equation

of the Second order

The most general linear partial differential equation of

second order can be written as

u
2
u
2
u
2
u u
A 2 +B + C 2 + D + E + Fu = 0
x xy y x y
ie., Au xx + Bu xy + Cu yy + Du x + Eu y + Fu = 0
where A, B,C, D, E, F are in general function of x and y.
The above equation of sec ond order (linear)is
said to
(i) elliptic if B − 4AC  0
2

(ii) parabolic if B − 4AC = 0

2

(iii) hyperbolic if B − 4AC  0.

2

Classify the following equations :

u2
u u
2 2
(i) 2 + 2 + 2 =0
x xy y
(ii) x f xx + (1 − y )f yy = 0
2 2
(i) Here, A = 1, B = 2,C = 1
B − 4AC = 4 − 4 = 0, for all x, y.
2

Hence, the equation is parabolic at all point s.

(ii) A = x , B = 0, C = (1 − y )
2 2

B − 4AC = −4x (1 − y ) = 4x (y − 1)
2 2 2 2 2

For all x except x = 0, x is positive

2

If − 1  y  1, y − 1 is negative.
2

 B − 4AC is negative if − 1  y  1, x  0
2
 For −   x   (x  0), − 1  y  1,
the equation is ellipse.
 For −   x   (x  0), y  −1 or y  1,
the equation is hyperbola.
 For x = 0, for all y or for all x, y  1,
the equation is parabolic.
Elliptic Equations:

The elliptic equation is Laplace’s equation is

 u  u
2 2
+ 2 =0 − − − −(1)
x 2
y
ie., u xx + u yy = 0 or  u = 0
2

In this type, we have two methods they are:

1. Liebmann’s iteration process

2. Poisson’s Equation
 Solve the above equation, we get

1
u i, j =  u i −1, j + u i +1, j + u i, j−1 + u i, j+1  − − − −(2)
4
This is called standard five point formula.

1
u i, j =  u i −1, j−1 + u i −1, j+1 + u i +1, j−1 + u i +1, j+1  − − − −(3)
4

This is called the diagonal five - point formula.

Schematic diagram:

Ui,j+1

Ui-1,j Ui,j Ui+1,j

Ui,j-1

Standard Five Point Formula

Schematic diagram

Ui-1,j+1 Ui+1,j+1

Ui,j

Ui-1,j-1 Ui+1,j-1

Diagonal Five Point Formula

Example
Find by the Liebmann’s method the values at
the interior lattice points of a square region of
the harmonic function u whose boundary values
are as shown in the following figure.

11.1 17.0 19.7 18.6

0

0 21.9

0 21.0

0 17.0
0 9.0
8.7 12.1 12.8
Solution:
Since u is harmonic, it satisfies Laplace’s

equation  2
u  2
u in the square.
+ = 0
x 2 y 2

11.1 17.0 19.7

0 18.6
u1 u2 u3
0 21.9
u4 u5 u6
0 21.0
u7 u8 u9
0 17.0
0 9.0
8.7 12.1 12.8
Finding rough values:
u5 = ¼ (0+17+21+12.1)=12.5 (SFPF)
u1 = ¼(0+12.5+17)=7.4 (DFPF)
u3 = ¼(12.5+18.6+17+1)=17.3 (DFPF)
u7 = ¼(12.5+0+0+12.1)=6.2 (DFPF)
u9 = ¼(12.5+9+12.1+21)=13.7 (DFPF)
u2 = ¼ (17+12.5+7.4+17.3)=13.6 (SFPF)
u4 = ¼(7.4+6.2+0+12.5) = 6.5 (SFPF)
u6 = ¼ (12.5+21+17.3+13.7)=16.1 (SFPF)
u8 = ¼(12.5+12.1+6.2+13.7)=11.1 (SFPF)
 Now we have got the rough values at all
interior grid points and already we possess
the boundary values at the lattice points.
We will now improve the values by using
always SFPF.
First iteration: (We obtain all values by
SFPF)
u1(1) = ¼(0+11.1+u2+u4)
= ¼(0+11.1+13.6+6.5)= 7.8
u2(1) = ¼(17+12.5+7.8+17.3) = 13.7
u3(1) = ¼(13.7+21.9+19.7+16.1) = 17.9
u4(1) = ¼ ( 0+12.5+7.8+6.2) = 6.6
u5(1) = ¼ (13.7+11.1+6.6+16.1) =11.9
u6(1) = ¼ (17.9+13.7+11.9+21) = 16.1
u7(1) = ¼ ( 6.6+8.7+0+11.1) = 6.6
u8(1) = ¼ (11.9+12.1+6.6+13.7) =11.1
u9(1) = ¼ (16.1+12.8+17+11.1) = 14.3
Second iteration:
u1(2) = ¼ (0+11.1+13.7+6.6) =7.9
u2(2) = ¼ (17+17.9+7.9+11.9) =13.7
u3(2) = ¼ (13.7+19.7+21.9+16.1) = 17.9
u4(2) = ¼ (7.9+0+11.9+6.6) = 6.6
u5(2) = ¼ (13.7+6.6+16.1+11.1) = 11.9
u6(2) = ¼ (11.9+ 17.9+21+14.3) = 16.3
u7(2) = ¼ (0+6.6+11.1+8.7) = 6.6
u8(2) = ¼ ( 6.6+11.9+14.3+12.1) = 11.2
u9(2) = ¼ (11.2+16.3+17+12.8) = 14.3
Third iteration:
u1(2) =7.9, u2(2) =13.7 , u3(2) =17.9, u4(2) = 6.6,
u5(2) = 11.9, u6(2) = 16.3, u7(2) = 6.6, u8(2) = 11.2,
u9(2) = 14.3.

Now, all the 9 values of u of the third iteration are

same as the corresponding values of the second
iteration. Hence we stop the procedure and accept
u1 = 7.9 , u2 = 13.7 , u3 = 17.9, u4 = 6.6
u5 = 11.9 , u6 = 16.3 , u7 = 6.6 , u8 = 11.2 ,
u9 = 14.3.
Example: 2
Solve u xx + u yy = 0 for the following square mesh
with boundary condition as shown below. Iterate
until the maximum difference between successive
values at any grid point is less than 0.001.

A 1 2 B

u1 u2
1 2

u3 u4
2 1

D C
2 1
Solution:

From the above diagram, we see that it is

symmetrical above the diagonals AC and BD.

Let u1 , u 2 , u 3 , u 4 be the values at the interior grid

points.
By symmetry, u1 = u 4 , u 2 = u 3 .
Therefore, we need to find only two values
u1 and u 2 .
 A 1 2
B

u1 u2
1 2

u2 u1
2 1

D C
2 1

1 1
u1 = (1 + u 2 ) ; u 2 = (2 + u1 )
2 2
4
solving, u1 = = 1.3333
3
5
u 2 = = 1.6666.
3
Poisson’s Equation

An equation of the form

 2 u = f ( x, y )
 2u  2u
i.e., + 2 = f ( x, y )
x 2
y

is called as Poisson’s equation where f ( x, y )

is a function of x and y only.

Solve the above equation, we get
ui −1, j + ui +1, j + ui , j −1 + ui , j +1 − 4ui , j = h 2 f (ih, jh) − − − −(1)
By applying the above formula at each mesh point,
we get a system of linear equation in the pivotal values i, j.
Example : 1
Solve  2
u = −10( x 2
+ y 2
+ 10) over the square mesh with
x = 0, y = 0, x = 3, y = 3 with u = 0 on the boundary and
mesh length 1 unit.

Solution: u=0
A B

D u1 E u2
u=0 u=0
F u3 G u4

C
u=0
The P.D.E is  2u = −10( x 2 + y 2 + 10)
here h = 1
ui −1, j + ui +1, j + ui , j −1 + ui , j +1 − 4ui , j = −10( x 2 + y 2 + 10)
Applying the formula (2) at D(i = 1, j = 2)
0 + 0 + u 2 + u 3 − 4u1 = −10(15) = −150
u 2 + u 3 − 4u1 = −150 − − − −(3)
Applying at E(i = 2, j = 2)
u1 + u 4 − 4u 2 = −180 − − − −(4)
Applying (2) at F(i = 1, j = 1)
u1 + u 4 − 4u 3 = −120 − − − −(5)
Applying (2) at G(i = 2, j = 1)
u 2 + u 3 − 4u 4 = −10(2 + 1 + 10) = −150 − − − −(6)
2 2
 (5) − (4) gives, 4(u 2 − u 3 ) = 60
u 2 − u 3 = 15 − − − −(7)
E lim inate u1 from (3) and (4), (3) + 4(4) gives,
−15u 2 + u 3 + 4u 4 = −870 − − − −(8)
Adding (6) and (8) − 7u 2 + u 3 = −510 − − − (9)
From (7) , (9) adding, u 2 = 82.5
u sin g (7), u 3 = u 2 − 15 = 82.5 − 15 = 67.5
Put in (3), 4u1 = 300  u1 = 75
4u 4 = 150 + 150  u 4 = 75
 u1 = u 4 = 75, u 2 = 82.5, u 3 = 67.5.
Note: we can solve the equation (3),(4),(5),(6) either by direct
elimination or by Gauss-seidel method
PARABOLIC EQUATIONS:
The one dimensional heat equation is

u 2  u
2
k
= where  =
2

t x 2
c

In this type, we have two methods they are:

1. Bender-Schmidt Method
2. Crank-Nicholson Difference Method.
Bender-Schmidt Method

Solve u xx = aut with boundary conditions,

u (0, t ) = T0 , u (l , t ) = Tl and with initial condition
u ( x,0) = f ( x), 0  x  l

Solve the above equation, we get

u i , j +1= ui +1, j + (1 − 2 )ui , j + ui −1, j − − − −(1)
k
where  = 2 .
ah
Equation (1) is called Explicit formula. It is valid if
1
0  .
2
 1
If we take  = , then equation (1) becomes,
2
 1

ui , j +1 = ui −1, j + ui +1, j − − − − − − − −(2)
2
1 k a 2
when  = = 2 ; k = h
2 ah 2
Equation (2) is called Bender – Schmidt recurrence
a 2
equation. This is valid if k = h .
2
Bender-Schmidt Method
Ui,j+1

B C

Ui-1,j Ui,j Ui+1,j

Schematic diagram
Example 1:
 2 f f
Solve = given f (0, t ) = f (5, t ) = 0,
x 2
t
f ( x,0) = x 2 (25 − x 2 ), find f in the range taking h = 1 and
upto 5 seconds.
Solution :
Given : u xx = aut  a = 1, h = 1
a 2 1
To use Bender − Schmidt' s equation, k = h =
2 2
1
Step − size in time = , Step − size of x = 1.
2
f (0,0) = 0, f (1,0) = 24, f (2,0) = 84, f (3,0) = 144
f (4,0) = 144, f (5,0) = 0.

we have, ui , j +1 =
(ui −1, j + ui +1, j )
2
 → x direction

i 0 1 2 3 4 5
j
0 0 24 84 144 144 0
0.5 0 42 84 114 72 0
1 0 42 78 78 57 0
1.5 0 39 60 67.5 39 0

2 0 30 53.25 49.5 33.75 0
t − direction
2.5 0 26.625 39.75 43.5 24.75 0
3 0 19.875 35.0625 32.25 21.75 0
3.5 0 17.5312 26.0625 28.4062 16.125 0
4 0 13.0312 22.9687 21.0938 14.2031 0

4.5 0 11.4843 17.0625 18.5859 10.5469 0

5 0 8.5312 15.0351 13.8047 9.2929 0

Example 2:
 u u
2
Solve = given u (0, t ) = 0, u (4, t ) = 0,
x 2
t
u ( x,0) = x(4 − x), assu min g h = k = 1.
Find the values of u upto 5 seconds.
Solution :
Given : u xx = aut  a = 1, h = 1, k = 1
a 2 1
To use Bender − Schmidt' s equation, k = h =
2 2
These values do not satisfy the condition. Hence
we cannot employ Bender − Schmidt formula.
Hence, we use the equation,
ui , j +1 = ui +1, j + (1 − 2 )ui , j + ui −1, j
k 1
Now,  = 2 = =1
ah 1 1
Hence, equation (1) reduces to
ui , j +1 = ui +1, j − ui , j + ui −1, j
 → x direction

i 0 1 2 3 4
j
0 0 3 4 3 0

1 0 1 2 1 0

2 0 1 0 1 0
t − direction
3 0 -1 2 -1 0

4 0 3 -4 3 0

5 0 -7 10 -7 0
Crank-Nicholson Difference Method:

To solve the parabolic equations u xx = aut

with boundary conditions, u (0, t ) = T0 , u (l , t ) = Tl
and initial condition u ( x,0) = f ( x), 0  x  l

Solve the above equation, we get

1 1
 u i +1, j +1+  u i −1, j +1−( + 1)ui , j +1
2 2
1 1
= − ui +1, j − ui −1, j + ( − 1)ui , j − − − −( I )
2 2
k
where  = 2 .
ah
 The above equation can be rewritten as
 (u i +1, j +1+u i −1, j +1) − 2( + 1)ui , j +1 = 2( − 1)ui , j −  (ui +1, j + ui −1, j ) − − − ( I )

Equation (I) is called Crank-Nicholson difference

scheme or method.

Note:
Setting  = 1 i.e., k = ah 2 . The Crank-Nicholson

formula reduces to
1

ui , j +1 = ui −1, j +1 + ui +1, j +1 + ui −1, j + ui +1, j
4

Crank-Nicholson Difference Method

E D

Ui-1,j Ui+1,j

B A C

Ui-1,j+1 Ui,j+1 Ui+1,j+1

Schematic diagram
Example: 1
Solve by Crank − Nicholson method the
 2u u
equation 2 = subject to u ( x,0) = 0, u (0, t ) = 0,
x t
u (1, t ) = t , for two time steps.
Solution :
1
Given : u xx = aut  a = 1, Take h =
4
2
1 1
 k = ah = 1  =2

 4  16
we have
1

ui , j +1 = ui −1, j +1 + ui +1, j +1 + ui −1, j + ui +1, j
4
 − − − −(A)
 → x direction

i 0 0.25 0.5 0.75 1

j
0 0 0 0 0 0

 1 0 u1 u2 u3 1
t − direction 16 16
0 u4 u5 u6 2
2
16 16
Using the equation (A), we get

u1 = (0 + 0 + 0 + u2 ) (i.e., ) u1 = u2 − − − (1)
1 1
4 4
u2 = (0 + 0 + u1 + u3 )
1 1
u2 = (u1 + u3 ) − − − (2)
4 4
1 1 1 1
u3 =  0 + 0 + u 2 +  u3 =  u2 +  − − − (3)
4 16  4 16 
Solving the three equations given by (1), (2), (3)
we get u1 , u2 , u3 . Substitute u3 , u1 values in (2)
1 1 1 1 
u2 =  u 2 +  u2 + 
4 4 4 16 
 1 1
u2 = = 0.0045 ; u1 = = 0.0011;
224 896
u3 = 0.0168.
Similarly, u4 , u5 , u6 can be got again
getting 3 equations in unknowns u4 , u5 , u6 .
we get u4 = 0.005899, u5 = 0.01913, u6 = 0.05277.
 Example: 2

Using Crank − Nicholson method, solve

 2 u u
= subject to u(x, 0) = 0, u(0, t) = 0, u(1, t) = t,
x 2
t
1 1 1
(i) taking h = 0.5 and k = (ii) h = and k =
8 4 8
Solution:
1
k 8 1
(i) Here a = 1,  = 2 = =
ah 1 2
1
4
1
Since  = , we cannot use simplied formula.
2
1
we use equation (1) , Putting  = the formula
2
becomes, u i +1, j+1 + u i −1, j+1 − 6u i, j+1 = −2u i, j − (u i −1, j + u i +1, j )
 → x direction

i 0 0.5 1
j
0 0 0 0

 (u i, j )
t − direction
1 0
u1 1
8 (u i, j+1 ) 8

Now, using above equation

1
+ 0 − 6u1 = −2(0) − (0 + 0)
8
1
 u1 = = 0.0208333.
48
 1
1 1 k 8
(ii) If h = , k = ,  = 2 = =2
4 8 ah 1
1
16
u sin g in general Crank − Nicholson formula
equation (I), we get
u i +1, j+1 + u i −1, j+1 − 3u i, j+1 = u i, j − u i +1, j − u i −1, j

→ x direction
i 0 0.25 0.5 0.75 1
j

0 0 0 0 0 0
t − direction

1 0
1
u1 u2 u3
8 8