Chapter

10

Partial Differential
Equations

10.1 Introduction
Most scientific and engineering problems are governed by
partial differential equations for which the dependent variable u
varies with the three coordinates of x, y, z, and time t. Solving the
partial differential equations is more difficult than the ordinary
differential equations learned earlier in the preceding chapters.
Their exact solutions are not available, in general, and numerical
methods are used to obtain the approximate solutions.
The widely used numerical methods are the finite
difference, finite element and finite volume methods. The finite
element method is the popular one because it can handle
complicated boundary conditions and domain geometries
effectively. MATLAB contains a toolbox that uses the finite
element method to solve for approximate solutions of the partial
differential equations in two dimensions.
332 Chapter 10 Partial Differential Equations

This chapter begins with the classification of the partial

differential equations. The MATLAB toolbox for solving these
differential equations is explained. The toolbox is then used to
solve the elliptic, parabolic and hyperbolic partial differential
equations, respectively. Exact solutions for simple domain
geometries are derived so that the finite element solutions can be
compared to measure their accuracy. More complicated examples,
for which their exact solutions are not available, are then used to
demonstrate the efficiency of the finite element method imple-
mented in the toolbox.

10.2 Classification of Partial Differential Equations

The Partial Differential Equation toolbox (PDE toolbox) is
used to solve the problems in two dimensions. The unknown
dependent variable u is function of x- and y- coordinates and may
be time t. The partial differential equations are classified into three
types:
(a) Elliptic equation,
 u  u
   c    c    a u  f
 x  x  y  y  
(b) Parabolic equation,
u    u    u  
d  c  c  au  f
t  x  x  y  y  

and (c) Hyperbolic equation,

 2u    u    u  
d  c  c  au  f
t 2  x  x  y  y  
where d, c, a and f are constants or may be function of x, y and u.
The boundary conditions consist of,
(a) Specifying value of the dependent variable u along the
boundary (Dirichlet condition) ,
10.3 The Finite Element Toolbox 333

hu  r
where h and r are constants or function of x, y and u.
(b) Specifying gradient value of the dependent variable u
along the boundary (Neumann condition) ,
u
c  qu  g
n
where c, q and g are constants or function of x, y and u. The letter
n denotes a unit vector normal to the boundary.
Solving the elliptic equation does not need the initial
condition while the parabolic equation needs an initial condition of
u x, y,0 at time t  0 . The hyperbolic equation needs two initial
conditions of u x, y,0 and  u  x, y,0 t at time t  0 .

10.3 The Finite Element Toolbox

The finite element method is used to solve the partial
differential equations. The main advantage of the method is its
flexibility for modelling arbitrary geometry easily. The problem
domain is first divided into small elements connected at nodes
where the unknowns are located. The finite element equations for
each element are derived from the governing differential equation
of the problems. These finite element equations are assembled
together to form a set of simultaneous algebraic equations. The
boundary conditions are then imposed prior to solving for solutions
of the unknowns at nodes.
Understanding the finite element procedure above leads to
the development of the PDE toolbox which contains commands
under the menus, tool bars, etc. via the Graphical User Interface
(GUI). The PDE toolbox can be initialized by simply typing
>> pdetool on the Command Window, a graphic interface as
shown in the figure is displayed.
334 Chapter 10 Partial Differential Equations

On the graphic interface, the menu commands are:

File New, Open…, Save, Save As…, Print…, Exit.

Edit Undo, Cut, Copy, Paste…, Clear, Select All.
Options Grid, Grid Spacing…, Snap, Axes Limits…,
Axes Equal, Zoom, etc.
Draw Draw Mode, Rectangle/square, Ellipse/circle,
Polygon, etc.
Boundary Boundary Mode, Specify Boundary Conditions, etc.
PDE PDE Mode, PDE Specification…, etc.
Mesh Mesh Mode, Initialize Mesh, Refine Mesh,
Show Node Labels, Show Triangle Labels, etc.
Solve Solve PDE, Parameters…, etc.
Plot Plot Solution, Parameters…, Export Movie….

Icons in the tool bar under the menu bar perform the same
tasks through the graphic interface. We will be familiar with this
graphic interface by solving examples in the following sections.
 10.4 Elliptic Equations 335

10.4 Elliptic Equations

The elliptic partial differential equation is in the form,
 u  u
   c    c    a u  f
 x  x  y  y  
where c, a and f are constants or function of x, y and u. The partial
differential equations in the elliptic form above arise in many
applications. These include steady-state heat transfer, potential
flow, cross-sectional stress in a bar under torsion, electric potential,
underground seepage flow, etc. We will use the PDE toolbox to
solve for the solution of u(x,y) from the elliptic differential
equation with the specified boundary conditions.
P1- Example Use the PDE toolbox to solve the elliptic partial differen-
200119790
tial equation,
 2u  2 u
  0
x 2 y 2
for the solution of u(x,y) in a rectangular domain with the boundary
conditions as shown in the figure.
y
u  sin  x 2 

1
u=0 u=0

x
u=0
2

It is noted that the exact solution for this problem which is derived
in the next example is,
sin x 2 sinh y 2
u x, y  
sinh 2
336 Chapter 10 Partial Differential Equations

By comparing coefficients of the governing differential

equation with the standard form of the elliptic differential equation,
we find that c  1 , a  0 and f  0 . This problem is equivalent to
a steady-state heat conduction in a rectangular plate with the size of
2  1 . Zero temperature ( u  0 ) is specified along the left, right
and bottom edges while the temperature is u  sin x 2 along the
top edge.
To use the PDE toolbox, we start by typing,

>> pdetool pdetool

on the Command Window. We select the menu Options, choose

the sub-menu Grid and click at the Snap option so that the points
we create will be located at the round off coordinates. We select
the Axes Limits as [-0.2 2.2] and [-0.2 1.2] for x- and y-axis,
respectively. Then, we click Apply and Axes Equal options so
that the model has equal scalings in both x- and y-directions. Note
that if the background grid overflows the screen area, the screen
resolution should be changed to 1024 768 .
10.4 Elliptic Equations 337

To create the rectangular domain of 2  1 , we select the

menu Draw and sub-menu Rectangle/square. Then, move the
cursor from the coordinate of (0,1) to the coordinate of (2,0) and
click at the mouse, a gray rectangle will appear with the letter of R1
as shown in the figure.

Next, to apply the boundary conditions, we select the

menu Boundary and sub-menu Boundary Mode. The gray
rectangle disappears while the rectangle edges become red arrows
as shown in the figure.
To specify the boundary condition along the left edge with
the value of u  0 , we double click at that edge, the Boundary
Condition dialog box with the equation,
h u  r
appears. We enter h  1 and r  0 which mean u is zero along this
edge and click OK. We repeat applying the same boundary
conditions of u  0 for the right and bottom edges. For the top
edge, we enter h  1 and r  sin pi  x. / 2 , then click OK. Note
that the letter x in the r expression must follow by a period due to
the expression format requirement in MATLAB.
338 Chapter 10 Partial Differential Equations

The next step is to choose the type of the differential

equation. We select the menu PDE and sub-menu PDE
Specification, the PDE Specification dialog box appears. After
selecting Type of PDE as Elliptic, we enter c  1 , a  0 and
f  0 , and click OK.
10.4 Elliptic Equations 339

A finite element mesh can now be constructed. After

selecting the menu Mesh and sub-menu Initialize Mesh, a mesh
with triangular elements is created. The mesh can be refined by
clicking at the sub-menu Refine Mesh. The mesh without the
background grid is shown in the figure.

Then, we can solve for solution of the problem. We select

the menu Solve and sub-menu Solve PDE so that the problem is
executed. The computed solution is plotted by selecting the menu
Plot and sub-menu Plot Solution. To display the colors similar to
those in commercial finite element software, the Colormap which is
under the sub-menu Parameter in the Plot Selection dialog box
should be selected as jet as shown in the figure.

We can display the computed solution in the form of color

fringe plot, line contour plot, deformed plot, plot with arrows, as
well as generating an animation. Displaying solution in the form of
animation is useful to show the solution behaviors of the parabolic
and hyperbolic problems. Most of the solutions shown herein are
in form of the line contour plots for clarity. Distribution of the
340 Chapter 10 Partial Differential Equations

computed solution in the form of contour lines for this example is

shown in the figure.

The exact solution in the preceding example can be

derived by using the method of separation of variable. Detailed
derivation of the exact solution is presented in the following
example.
10.4 Elliptic Equations 341

Example Derive the exact solution of the elliptic differential

equation which is in form of the Laplace equation,
 2u  2u
  0
x 2 y 2
for a rectangular domain with dimensions of   w as shown in the
figure. The boundary conditions along the four edges are,
u0, y   0 , u, y   0
u x,0  0 , u x, w  f  x

y u  f x

 2u  2u
w u0  0 u0
x 2 y 2

x
u0


The method of separation of variables is applied by first

assuming that the solution is in the product form of the functions
X  x  and Y  y  ,
u x, y   X  x  Y  y 
By substituting the assumed solution into the differential equation,
we obtain,
X Y  X Y   0
where the prime symbol (  ) denotes the derivative order. The
equation above can be written as,
342 Chapter 10 Partial Differential Equations

1 d2X 1 d 2Y
 
X dx 2
Y dy 2
Since X is only function of x while Y is only function of y, then
both sides of this equation must be equal to a constant,
1 d2X 1 d 2Y
    2
X dx 2
Y dy 2
Thus, the given partial differential equation becomes two ordinary
differential equations which should be solved easier.
If we consider the first ordinary differential equation,
d2X
 2 X  0
dx 2
Its general solution is in the form of sine and cosine functions,
X  x   A sin  x  B cos  x
where A and B are constants. These two constants are determined
as follows. Since the left edge boundary condition is u0, y   0 ,
then X 0 Y  y   0 , or X 0  0 , so that B  0 . Similarly, the
right edge boundary condition is u, y   0 , then X   Y  y   0 ,
or X    0 . Then, X  x  above becomes,

X   0  A sin  
But A cannot be zero, thus,
n
   n and  

where n is an integer. Hence, in general,
n
n  n  1, 2, 3, ...

which are called the eigenvalues.
Similarly, if we consider the second ordinary differential
equation,
d 2Y
 2 Y  0
dy 2
which has the general solution of,
10.4 Elliptic Equations 343

Y  y   E sinh  y  F cosh  y
where E and F are constants. The constant F is determined from
the bottom edge boundary condition of u x,0  0 , i.e.,
X  x  Y 0  0 . Then Y 0  0 , which leads to F  0 . Thus, the
general solution of the partial differential equation is,

u x, y   X  x  Y  y    A sin  x   E sinh  y 
Or, u x, y   C sin  x sinh  y
where C  AE denotes a new constant. Since there are n values of
 , i.e, 1 , 2 , 3 , ..., n , thus, the general solution can be written
in the summation form for n values of  as,

u x, y   C
n 1
n sin n x sinh n y

The constants Cn are determined from the top edge

boundary condition at y  w as,

u x, w  f  x   C
n 1
n sinh n w sin n x


Or, D
n 1
n sin n x  f  x

where Dn  Cn sinh n w . This means, once we obtain Dn from the

above equation, we can find Cn and also the exact solution of
u x, y  .
To solve for Dn , we multiply both sides of the equations
by sinm x  where m denotes an integer, and perform integra-
tion from   to  as follows,
 

n x m x m x
n 1
Dn  sin

sin

dx   f  x  sin 
dx
 

But from the integration formula,

344 Chapter 10 Partial Differential Equations


n x m x  , m  n


sin

sin

dx  
0, m  n
which is known as the orthogonal properties.
Then, the equation above reduces to,

n x
Dn  

 f  x sin 
dx
 
1 n x 2 n x
Or, Dn  
 
f  x  sin

dx 
  f  x  sin
0

dx

But Dn  Cn sinh n w  Cn sinhn w   , therefore,


2 n x
Cn 
n w  f  x sin 
dx
 sinh 0

Hence, the exact solution of this problem is,
n y 
2  sinh  n x n x
u  x, y   
 n1 n w
sin
 0  f  x  sin

dx
sinh

For the special case in the preceding example, if
x
f  x  sin
2
with the domain sizes of   2 and w  1 , the integral in the exact
solution reduces to,

n x
2
x n x 1, n  1

0
f  x  sin

dx   sin
0
2
sin
2
dx  
0, n  1
Thus, the exact solution of the preceding example is,
y
2 sinh 2 x
u x, y   sin
2  2
sinh
2
10.4 Elliptic Equations 345

sin x 2 sinh y 2
Or, u x, y  
sinh 2
This exact solution can be plotted to show the distribution
of u x, y  by using the meshgrid and mesh commands as follows,
xx = 0:.05:2; yy = 0:.05:1;
[x,y] = meshgrid(xx,yy);
u = sin(pi.*x/2).*sinh(pi.*y/2)/sinh(pi/2);
mesh(x,y,u); view(-37.5,30);

Distribution of the exact solution u x, y  is shown in the figure.

0.5

0
1
2
0.5 1.5
1
0.5
0 0

Exact solution

The finite element solution obtained from the preceding

example can be plotted in the same fashion as shown in the figure.
Height: u

0.5

0
1
2
0.5 1.5
1
0.5
0 0

Approximate solution
 346 Chapter 10 Partial Differential Equations

P-2
201214540 Example Use the PDE toolbox to solve the elliptic partial differen-
tial equation,
 2u  2u
 3 2  2   5u  8
 x y 
for an arbitrary two-dimensional domain as shown in the figure.
The domain has the specified boundary condition of u  0 for all
the edges. The problem statement is equivalent to the two-
dimensional steady-state heat transfer in a plate. The plate material
has its thermal conductivity coefficient of 3 units and is subjected
to surface convection with the convection coefficient of 5 units.
The plate is also subjected to a surface heating with the magnitude
of 8 units.
u0

u0
u0

u0
u0
u0

u0

We begin solving this problem by typing,

>> pdetool pdetool
on the Command Window similar to the preceding example. We
select the menu Axes Limits as [-0.2 2.2] and [-0.2 1.6] for x- and
y-axis, respectively, then click the sub-menu Grid Spacing to enter
-0.2:0.1:2.2 and -0.2:0.1:1.6 into the X-axis and Y-axis boxes,
respectively, and click Apply.
Next, we draw the geometry of the domain by creating the
rectangle R1, the circle E1 (lower left of the figure), the ellipse E2
and the square SQ1. Then, we modify the command in the Set
10.4 Elliptic Equations 347

formula box to R1-E1-E2-SQ1 which means we subtract the areas

of E1, E2 and SQ1 from R1 as shown in the figure.

After that, we select the menu Boundary followed by the

sub-menu Boundary Mode, the red arrows representing the
domain boundaries appear as shown in the figure.
348 Chapter 10 Partial Differential Equations

Next, we select the type of the differential equation that

governs the problem. Here, we select the menu PDE and sub-
menu PDE Specification, then click on Elliptic in PDE
Specification dialog box and enter c  3, a  5 and f  8 as
shown in the figure, and click OK.

A mesh can now be constructed by selecting the menu

Mesh and sub-menu Initialize Mesh. If the element sizes are too
large, we click at the sub-menu Refine Mesh so that smaller
elements are generated as shown in the figure.

The final step is to perform the analysis to find solutions

of u at nodes. We select the menu Solve and sub-menu Solve PDE,
the color fringe plot of the computed solution appears on the
10.5 Parabolic Equations 349

screen. If other plotting style is preferred, we select the menu Plot

followed by sub-menu Parameters and choose different options
there. The line contour plot representing the distribution of the
computed solution is shown in the figure.

10.5 Parabolic Equations

The parabolic partial differential equation is in the form,

u    u    u  
d  c  c  au  f
t  x  x  y  y  

where d , c , a and f are constants or may be function of x, y and

u. The parabolic equation is more complicated than the elliptic
equation because of an additional independent variable of t.
Finding the exact solution is more difficult, especially for domain
with arbitrary geometry. The PDE toolbox using the finite element
method can alleviate such difficulty by finding the approximate
solution instead.
 350 Chapter 10 Partial Differential Equations

P-3 Example Use the PDE toolbox to solve the parabolic partial
201361710 differential equation in the form,
u  2u u 2
2  3 2  2   8
t  x y 
for a circular domain with a unit radius. The domain has the
specified boundary condition of u  0 along the edge as shown in
the figure and the initial condition of u x, y,0  0 at time t  0 .

u0

The problem statement is equivalent to transient heat

conduction in a circular plate with specified zero temperature along
the edge and zero initial temperature. The plate material has the
thermal conductivity coefficient of 3 units and the specific heat of 2
units with a unit of material density. The plate is also subjected to
a specified surface heating of 8 units.

We can create a finite element model for the entire plate

but, due to symmetry of the solution, only the lower half of the
plate can be modelled. As shown in the figure, the boundary
condition of u  0 is applied along the outer edge. Along the
center line, the gradient of u normal to the edge must be equal to
zero representing the insulated boundary condition.
10.5 Parabolic Equations 351

u
n̂ 0
n

u0

Again, to analyze this problem by using the PDE toolbox,

we begin by entering,
>> pdetool pdetool

on the Command Window. We select the menu Option followed

by sub-menu Grid to show the grid background. We choose the
sub-menu Snap so that the constructed points are located at the
round off coordinates. Then, we select the sub-menu Axes Equal
for equal scalings in both x- and y-directions.
To construct the domain with half circle geometry, we
start by clicking at the ellipse icon in the tool bar. We place the
cursor arrow head at the coordinates of (0,0) and move it to the
coordinates of (1,-1), click the mouse, then a circle with the letter
C1 appears. We follow the same procedure by choosing the
rectangle icon in the tool bar. We place the cursor arrow head at
the coordinates of (-1,0), drag it to the point at the coordinates of
(1,-1), click the mouse, a rectangle with the letter R1 appears as
shown in the figure. Then, we modify the expression in the Set
formula box to C1R1, click the icon ?, a half circle domain
indicated by the red arrows appears as shown in the figure.
The next step is to apply the boundary conditions. We
double click at the circle centerline, the Boundary condition
dialog box appears, then choose the Neumann condition type
followed by OK. Note that the default boundary condition of
u  0 is automatically applied along the outer edge.
352 Chapter 10 Partial Differential Equations

We now can select the type of the differential equation to

be solved. We click the menu PDE and sub-menu PDE
Specification. We choose the Parabolic Type of PDE in the PDE
Specification dialog box, enter c  3, a  0, f  8 and d  2 as
shown in the figure, then click OK.

A finite element mesh consisting of triangular elements

can now be constructed. We select the menu Mesh and sub-menu
Initialize Mesh, the mesh as shown in the figure is generated. To
solve for the transient solution, we choose the menu Solve and
click at the Parameters icon. In the Solve Parameters dialog box
as shown in the figure, we enter the values of 0:0.05:1 in the Time
box which means we perform the calculation for the interval of
0  t  1 and save the solution at every t  0.05 .
 10.5 Parabolic Equations 353

By clicking the  icon in the tool bar, the computed

solutions at different times are obtained. These solutions can be
plotted as contour lines in two dimensions or the carpet plots in
three dimensions as shown in the figure. The transient solution
which varies with time t can be animated by clicking at the
Animation icon in the sub-menu Plot Selection under the menu
Plot  Parameters. Users may try different commands in this
sub-menu to display solutions in various ways.

P-4 Example Use the PDE tool to solve the parabolic partial
201421800 differential equation,
u  2u u 2
 2 2  2   3u  4
t  x y 
on the 3  2 rectangular domain with small ellipse and rectangular
holes inside. The initial condition is u x, y,0  0 while the
boundary conditions are shown in the figure.
The procedure for solving this problem is similar to that
explained the preceding example. The domain geometry, however,
is more complicated with the boundary condition along the left
edge changes abruptly from u  0 to u  1 at t  0 . A finite
element mesh can be constructed as shown in the figure. We select
Parabolic as the type of PDE to be solved in the PDE Specification
dialog box under the PDE menu with the parameters of c  2,
a  3, f  4 and d  1 . To obtain the transient solution, we enter
0:0.01:0.5 in the Time box of the Solve Parameters dialog box as
shown in the figure.
354 Chapter 10 Partial Differential Equations

n̂

u
0 u  x , y ,0   0 n̂
n

u 1 u0

u
n̂ 0
n
u
0
n

n̂

The computed solutions of u x, y, t  at t  0.1 and 0.3 are

plotted using contour lines as shown in the figures.
10.5 Parabolic Equations 355

t  0 .1

t  0 .3

Animation of the transient solution u x, y, t  can help us to

understand the solution behavior clearly. A sample of the transient
solution u x, y  at time t  0.1 is shown as a carpet plot in the
figure.
This example demonstrates that the finite element method
can provide approximate solution to the differential equation with
complicated boundary conditions and geometry effectively. The
next example shows detailed derivation of the exact solution for the
parabolic differential equation with simple boundary conditions
and domain geometry. The example will show that the exact
solution is not easy to derive even for a simple problem. These two
examples thus highlight benefits of using the finite element method
to obtain approximate solutions for complex problems.
356 Chapter 10 Partial Differential Equations

Example Derive the exact solution of the parabolic partial

differential equation,
u   2u u 2 
   0
t  x 2 y 2 
for the   w rectangular domain as shown in the figure. The
boundary conditions are,
u  x, 0, t   0, u  x, w, t   0, 0 x
u 0, y, t   0, u , y, t   0, 0 yw
and the initial condition is,
u  x, y, 0  f  x, y  t 0

The method of separation of variables is used to solve for

the exact solution. We start by assuming the solution u  x, y, t  as
the product of the three functions X  x  , Y  y  and T t  ,
u  x, y, t   X  x  Y  y  T t 
By substituting it into the differential equation, we obtain,
X Y T    X Y T  X Y T   0
10.5 Parabolic Equations 357

u0

u   2u u 2 
    0
t  x 2 y 2 
w u0 u0
u  x, y , 0   f  x, y 

x
u0


where the prime symbol (  ) denotes the derivative order. We

divide the equation through by X Y T and move some terms to get,
T  Y  X 
 
T Y X
Since the terms on the left-hand-side of the equation are only
functions of t and y while the term on the right-hand-side of the
equation is only function of x, then they must be equal to a
constant,
T  Y  X 
    2
T Y X
Similarly,
T Y 
 2    2
T Y
Hence, the method of separation of variables changes the given
partial differential equation into three ordinary differential
equations as,
358 Chapter 10 Partial Differential Equations

X   2 X  0, Y    2Y  0 and T   2   2 T  0
At the same time, the given boundary conditions become,
X 0  0, X    0, Y 0  0 and Y  w  0
By using the same procedure as shown in the preceding example of
the elliptic differential equation, the eigenvalues and eigenvectors
are obtained,
m m x 
m  , X m  x   sin 
   
n n y 
and n  , Yn  y   sin 
w  w 
The ordinary differential equation related to the time t is,
T   2   2 T  0
Its general solution is in the form,
T  Amn e  mn t
2

m   n 
2 2
where  2
     
mn
2
m  
2
n 
    w 
Thus, the general solution is,
 
u x, y, t    X
m 1 n 1
m Yn T

 
m x   n y   mn2 t
  A sin   sin  e
    w 
mn
m 1 n 1

where the constants Amn are determined from the initial condition
of u x, y,0  f  x, y  by using the orthogonal properties,
 w
m x   n y 
f  x, y  sin 
4
Amn   
w 0 0  sin 
    w 
 dx dy

As an example, if the initial condition is,

u x, y,0  f  x, y   x  x 2  y w  y 
then, the constants Amn are,
10.5 Parabolic Equations 359

 w
m x   n y 
x  x 2  y  w  y  sin 
4
Amn   
w 0 0  sin 
    w 
 dx dy

 1m  1n  1

48  2 w2

mn  2 3

Hence, the exact solution for this particular case is,

 
 1m   1n  1 
48  2 w2
u x, y, t     mn
m 1 n 1 
2 3

m x   n y   m2
sin  
 2  n 2 w2  2t
 sin  e
    w 

The exact solution above is in the form of infinite series.

We can create a MATLAB script file to compute the solution. For
example, the script file for determining the solution of u x, y, t 
at t  0.02 for a unit square domain   w  1 consists of the
following statements,

xx = 0:1/60:1; yy = 0:1/60:1;
[x,y] = meshgrid(xx,yy); meshgrid
t = 0.02; u = 0;
for m = 1:50;
for n = 1:50;
a = (-1)^m; b = (-1)^n - 1;
c = sin(m*pi.*x); d = sin(n*pi.*y);
ee= (m*m+n*n)*pi*pi*t; e = exp(-ee);
fac = 48/(m^3*n^3*pi^6);
u = u + fac*a*b.*c.*d*e;
end
end
mesh(x,y,u); view(-37.5,30); mesh

The same problem is solved by using the PDE toolbox for

the approximate solution. The exact and approximate solutions are
compared in form of the carpet plots as shown in the figures.
360 Chapter 10 Partial Differential Equations

0.07

0.06

0.05

0.04

0.03

0.02

0.01

0
1
0.8 1
0.6 0.8
0.4 0.6
0.4
0.2 0.2
0 0

Exact solution

Time=0.02 Height: u

0.07

0.06

0.05

0.04

0.03

0.02

0.01

0
1
0.8 1
0.6 0.8
0.4 0.6
0.4
0.2 0.2
0 0

Approximate solution
 10.6 Hyperbolic Equations 361

10.6 Hyperbolic Equations

The hyperbolic partial differential equation is in the form,
 2u    u    u  
d  c  c  au  f
t 2  x  x  y  y  
where d, c, a and f are constants or function of x, y and u. The
hyperbolic equation is more complicated than the elliptic and
parabolic equations. The dependent variable u is function of the
independent variables x, y and t. Two initial conditions of
u x, y,0 and u x, y,0 t are needed for solving the problem.
In this section, we will use the PDE toolbox to solve the
hyperbolic partial differential equation for both simple and
complex domains.

P-5 Example Use the PDE toolbox to solve the hyperbolic partial
201426040 differential equation in the form,
 2u   2u  2u 
   0
t 2  x 2 y 2 
for a 2  2 unit square domain. The boundary conditions are u  0
on the left and right edges while u y  0 on the top and bottom
edges as shown in the figure.
y
u
0
 1, 1 y 1, 1

u0 u0
x

 1,  1 u 1,  1
0
y
362 Chapter 10 Partial Differential Equations

The initial conditions are,

u x, y,0  tan1 cos x 2
u x, y,0 t  4 sin x esin y 2 
To use the PDE toolbox, we start by typing,
>> pdetool pdetool
on the Command Window. We select the menu Options and click
the Grid and Snap options to show the background grids and to
place the points exactly at the round off coordinates. We also
select Axes Equal option to have equal scalings for both x- and y-
axes.
Next, we create a square domain with the dimensions of
2  2 units. We click at the rectangle icon in the tool bar, place the
cursor arrow head at the coordinate of (-1,1), drag the mouse to
move the cursor arrow head to the coordinate of (1,-1), then click
the mouse again, a square domain with the letter SQ1 appears.
Then, we apply the boundary conditions for the four edges, one at a
time. We select the Dirichlet condition for the left and right edges,
and the Neumann condition for the top and bottom edges.
To specify the type of the differential equation, we select
the menu PDE and sub-menu PDE Specification. We choose the
type of PDE as Hyperbolic, and enter the values of c  1 , a  0 ,
f  0 and d  1 , and click OK as shown in the figure.

A finite element mesh can now be constructed. We select

the menu Mesh followed by sub-menu Initialize Mesh, a mesh
with triangular elements appears. The initial mesh may be
10.6 Hyperbolic Equations 363

too crude, we can click the sub-menu Refine Mesh for couple
times so that elements become smaller as shown in the figure.
To solve the problem, we choose the menu Solve and sub-
menu Parameters, and enter values in the Time box as 0:2:40. We
also need to provide the initial conditions of u x, y,0 and
u x, y,0 t in the next two small boxes as shown in the figure.

Then, we execute the problem by choosing the menu

Solve and the sub-menu Solve PDE. The computed solutions of
u x, y,10 and u x, y,20 at time t  10 and 20 are shown as the
contour line plots in the figures.

t  10 t  20
 364 Chapter 10 Partial Differential Equations

To display the solution that changes with time, we can

select the Animation icon. Herein, a typical solution at t  40 is
shown in the figure in form of a carpet plot.

P-6 Example Use the PDE toolbox to solve the hyperbolic partial
201578370
differential equation,
 2u  2 u  2u
5  4  2  2   3u  2
t 2
 x y 
for a unit square domain with two small holes inside as shown in
the figure. The boundary conditions are u  0 on the left and right
edges, and u y  0 on the top and bottom edges while u n  0
along inner edges of the holes. The two initial conditions are,

u x, y,0  tan1 cos  x 2

u x, y,0 t  4 sin x esin y 2 

10.6 Hyperbolic Equations 365

y
u
0
 1, 1 y 1, 1

u0 n̂ u0
x

n̂

 1,  1 u
0
 1, 1
y

Since this example is similar to the preceding one except

the differential equation and the two holes inside the domain, we
can follow the same procedure to solve the problem. To provide
information of the differential equation, we select the menu PDE
and sub-menu PDE Specification, choose the Type of PDE as
Hyperbolic and enter the values of c  4, a  3, f  2 and d  5 ,
then click OK.
366 Chapter 10 Partial Differential Equations

A finite element mesh can be constructed by using the

same procedure as explained in the preceding example. The Solve
Parameter dialog box remains the same as shown in the figure.

Typical solutions of u x, y, t  at times t  10 and 40 are

plotted in the form of contour lines as shown in the figures.

t  10 t  40

The solution can be displayed as a carpet plot in three

dimensions. A typical solution of u x, y, t  at time t  40 in form
of the carpet plot is shown in the figure.
10.6 Hyperbolic Equations 367

To highlight the advantage of using the finite element

method to solve for approximate solution of the hyperbolic
problem, we will use the example below to show the derivation of
exact solution for a very simple problem. We will see that the
derivation is rather lengthy and complicated.

Example Derive the exact solution of the hyperbolic partial

differential equation,
 2u   2u u 2 
   0
t 2  x 2 y 2 
for a rectangular domain with the dimensions of   w as shown in
the figure. The boundary conditions are,
u  x, 0, t   0, u  x, w, t   0 , 0 x
u 0, y, t   0, u , y, t   0 , 0 yw
and the initial conditions are,
u
u  x, y, 0  f  x, y ,  x, y,0  0 t 0
t
The method of separation of variables is again used to
derive the exact solution. We first assume the exact solution
u  x, y, t  in form of the product of the three functions X  x, Y  y
and T t  as,
368 Chapter 10 Partial Differential Equations

u0

 2u   2u u 2 
    0
t 2  x 2 y 2 

w u0 u  x, y , 0   f  x, y  u0
u
 x, y , 0   0
t

x
u0


u  x, y, t   X  x Y  y  T t 
By substituting it into the differential equation, we obtain,
X Y T    X Y T  X Y T   0
where the prime symbol (  ) denotes the derivative order. We
divide the equation through by X Y T and move some terms to get,
T  Y  X 
 
T Y X
Since the terms on the left-hand-side of the equation are only
functions of t and y while the term on the right-hand-side of the
equation is only function of x, then they must be equal to a
constant,
T  Y  X 
    2
T Y X
T  Y 
which leads to,  2    2
T Y
10.6 Hyperbolic Equations 369

Hence, the method of separation of variables changes the partial

differential equation into three ordinary differential equations as,
X   2 X  0 , Y    2Y  0 and T   2   2 T  0
while the boundary conditions become,
X 0  0 , X   0 , Y 0  0 and Y  w  0
The process leads to the eigenvalues and the eigenvectors similar to
those explained in the elliptic and parabolic problems,
m m x 
m  , X m  x  sin 
   
n n y 
and n  , Yn  y   sin 
w  w 
The ordinary differential equation for the time t is,
T   2   2 T  0
The general solution of this differential equation is in the form of
sine and cosine functions. But after applying the initial condition
of u x, y,0 t  0 , the solution is only in the form of cosine
function as,
T t   Amn cos mnt 
m   n  
2 2
where  mn
2
 m2  n2     
    w 
Thus, the general solution of the partial differential
equation is,
 
u x, y, t    X m Yn T
m 1 n 1
 
m  x   n y 
  A mn sin   sin   cos  mn t 
m 1 n 1     w 
where the constant Amn can be determined from the initial
condition of u x, y,0  f  x, y  by using the orthogonal properties
as explained in the preceding examples. These constants Amn are,
370 Chapter 10 Partial Differential Equations

 w
m x   n y 
f  x, y  sin 
4
Amn   
w 0 0  sin
    w 
 dx dy

As an example, if the initial condition is,

u x, y,0  f  x, y   x  x y w  y 
then, the constants Amn are,
 w
m x   n y 
x  x  y w  y  sin 
4
Amn   
w 0 0  sin
    w 
 dx dy


16  2 w 2
 1m  1  1n  1
m n 2 3

So that the exact solution for this case is,

 
u x, y, t     mn
m 1 n 1
16  2 w2

2 3
 1
m
 
 1  1  1 
n

m x   n y 
sin   sin   cos  mnt 
    w 
The exact solution above is in the form of infinite series.
To solve for the solution of u x, y, t  at t  0.35 for a unit square
domain   w  1 , a MATLAB script file can be created with
details as follows,
xx = 0:1/60:1; yy = 0:1/60:1;
[x,y] = meshgrid(xx,yy); meshgrid
t = 0.35; u = 0;
for m = 1:50;
for n = 1:50;
a = (-1)^m - 1; b = (-1)^n - 1;
c = sin(m*pi.*x); d = sin(n*pi.*y);
al= sqrt(m^2*pi^2+n^2*pi^2); e = cos(al*t);
fac = 16/(m^3*n^3*pi^6);
u = u + fac*a*b.*c.*d*e;
end
end
mesh(x,y,u); view(-30,-10); mesh
10.6 Hyperbolic Equations 371

The same problem is solved by using the PDE toolbox.

The approximate solution obtained from the PDE toolbox is
compared with the exact solution in the form of carpet plot as
shown in the figures. The comparison indicates that the finite
element method in the PDE toolbox can provide accurate
approximate solution by agreeing very well with the exact solution.
-3
x 10

-1

-2
0
-3
0 0.2 0.5
0.4 0.6 0.8 1 1

Exact solution

Time=0.35 Height: u
-3
x 10

-1

-2
0
-3
0 0.2 0.5
0.4 0.6 0.8 1 1

Approximate solution
372 Chapter 10 Partial Differential Equations

10.7 Concluding Remarks

In this chapter, we learned three types of the partial
differential equations which are in the elliptic, parabolic and
hyperbolic forms. We solved these partial differential equations for
their solutions in two dimensions. Exact solutions can be derived
only for simple differential equations with plain boundary
conditions and geometry. For more complicated problems, we
have to employ the numerical methods to solve for approximate
solutions.

MATLAB contains the PDE toolbox that uses the

finite element method to solve these problems for approximate
solutions. The toolbox solves the partial differential equations that
are in different forms. The boundary conditions may be
complicated and the geometry could be arbitrary. The process
starts from discretizing the problem domain into a number of
small triangular elements. These elements are connected at nodes
where the unknowns are located. The finite element equations are
derived for each element and assemble together to form up a set
of simultaneous equations. The boundary conditions are then
imposed on the set of equations before solving them for the
solutions at nodes.

Several examples are used to show derivation of exact

solutions for simple problems and to find approximate solutions for
more complicated ones. The results have demonstrated that the
PDE toolbox can provide approximate solutions with high accuracy
if the finite element meshes are refined with small elements. These
results highlight the advantages of using the finite element method
for solving partial differential equations with complicated boundary
conditions and domain geometries.
 Exercises 373

Exercises

1. Use the PDE toolbox to solve the elliptic partial differential

equation,
 2u  2 u
  0
x 2 y 2
for a unit square domain of 0  x  1 and 0  y  1 with the
boundary conditions of,
u x,0  0 , u x,1  1 0  x 1
u0, y   0 , u1, y   0 0  y 1
Plot to compare the solution with the exact solution of,
4 
sin n x  sinh n y 
u x, y   
 n 1,3 ,5 ,... n sinh n 

2. If the boundary condition along the right edge of the square

domain in Problem 1 is changed to,
u1, y   1 0  y 1
use the PDE toolbox to solve the problem again. Plot to
compare the solution with the exact solution if it can be
derived.
P-7 3. Use the PDE toolbox to solve the elliptic partial differential
201589270 equation,
 2u  2u
  1
x 2 y 2
for a unit square domain of 0  x  1 and 0  y  1 . The four
edges have the specified boundary conditions of u  0 . Plot to
compare the solution with the exact solution of,
16  sin i x  sin  j y 
u x, y  
π 4 i,j 
1,3 ,5 ,...
i3 j 2  i 2 j 3
374 Chapter 10 Partial Differential Equations

4. If there is a circular hole with radius of 0.2 at the center of the

square domain in Problem 3, use the PDE toolbox to solve the
problem again when the hole edge has the: (a) Dirichlet
boundary condition of u  0 and (b) Neumann boundary
condition of u n  0 . Plot to compare the two solutions in
form of the carpet plot.

5. Use the PDE toolbox to solve the elliptic partial differential

equation,
 2u  2 u
  0
x 2 y 2
for a unit square domain of 0  x  1 and 0  y  1 with the
boundary conditions of,
u x,0  0 , u x,1  x 0  x 1
u0, y   0 , u1, y  y 0  y 1
Plot to compare the solution with the exact solution of
u x, y   xy .

6. Use the PDE toolbox to solve Problem 5 again if the

differential equation is changed to,
 2u  2u
 5 2  2   7u  10
 x y 
Explain the physical meaning of the solution which is different
from that obtained in Problem 5.

7. Use the PDE toolbox to solve the elliptic partial differential

equation,
 2u  2 u
  0
x 2 y 2
for a unit square domain of 0  x  1 and 0  y  1 with the
boundary conditions of,
Exercises 375

u
u x,0  1 ,  x,1  u x,1  2 0  x 1
y
u u
0, y   0 , 1, y   0 0  y 1
x x
Plot to compare the solution with the exact solution of,
y
u x, y   1 
2

8. Use the PDE toolbox to solve Problem 7 again if the

differential equation is changed to,
 2 u  2u
  2  2   5u  9
 x y 

Plot the solution u x, y  in form of the carpet plot.

9. Use the PDE toolbox to solve the elliptic partial differential

equation,
 2u  2u
   x2  y 2  ex y
x 2  y 2
for a rectangular domain with the size of 2  1 units, i.e.,
0  x  2 and 0  y  1 . The boundary conditions are,
u  x, 0  1, u  x, 1  ex , 0 x2
u 0, y   1, u 2, y   e2 y , 0  y 1
Plot to compare the solution with the exact solution of,
u x, y  exy

10. Use the PDE toolbox to solve the elliptic partial differential
equation,
 2u  2u
  2x x3  6xy  6xy2  1
x 2  y 2
376 Chapter 10 Partial Differential Equations

for a unit square domain of 0  x  1 and 0  y  1 . The four

edges have the specified boundary conditions of u  0 . Plot to
compare the solution with the exact solution of,
u x, y    x  x4   y  y 2 

11. Use the PDE toolbox to solve the elliptic partial differential
equation,
 2u  2u
   20
x 2  y 2
for a circular domain with a unit radius. The boundary
condition along the edge is u  0 . Compare the solution with
the exact solution of,
u  x, y   5 1  x2  y2 

12. Use the PDE toolbox to solve Problem 11 again if the circular
domain contains a small circular hole with radius of 0.3 unit
and a square hole with dimension of 0.2  0.2 units as shown
in the figure. The boundary conditions for the edges of the
circular hole and square hole are u  0 and u n  0 ,
respectively. Plot the solution u x, y  in form of the carpet
plot.
Exercises 377

.4 .4

13. Use the PDE toolbox to solve the elliptic partial differential
equation,
 2u  2u
  0
x 2  y 2
y
u0

u  100
u0

8 x
2

u0

u0

for a square domain with the size of 8  8 units. The domain

contains a circular hole with the radius of 2 units at its center
378 Chapter 10 Partial Differential Equations

as shown in the figure. If the boundary conditions of the hole

and the outer boundary of the domain are u  100 and u  0 ,
respectively, solve for the solution of u x, y  . Then, plot the
solution in form of the carpet plot.

14. Due to symmetry of the solution in Problem 13, only the upper
right quarter of the domain as shown in the figure can be used
for the analysis. Solve the problem again with the boundary
conditions as shown in the figure. Plot the solution in form of
the carpet plot.
y
u0

u
0
n
u0

2
u  100
x
u
0
n

15. Use the PDE toolbox to solve the elliptic partial differential
equation which is in form of the Helmholtz equation,
 2u  2u
  a x, y  u  f  x, y
x 2 y 2
for a unit square domain of 0  x  1 and 0  y  1 . The four
edges have the specified boundary conditions of u  0 . Find
the solution if a  2 and
f  x, y   xy x2  71  y 2   1  x2  y 2  7
Plot to compare the solution with the exact solution of,
u x, y    x  x3   y  y3 
 Exercises 379

16. Use the PDE toolbox to solve the parabolic partial differential
equation,
u   2 u  2 u 
    0
t  x 2 y 2 
for a rectangular domain with the dimensions of   0.3 units,
i.e., 0  x   and 0  y  0.3 . The boundary conditions along
the four edges are u n  0 and the initial condition is,
u x, y,0  cos x
Plot to compare the solution with the exact solution of,
u x, y, t   et cos x
when t  0.2 .

P-8 17. Use the PDE toolbox to solve the parabolic partial differential
201693480 equation,
u   2u  2u 
   2
t  x 2 y 2 
for the rectangular domain of 0  x  1 and 0  y  0.2 . The
boundary conditions are shown in the figure. The initial
condition is,
u x, y,0  sin x  x1  x
Plot to compare the solution with the exact solution of,
u x, y, t   e 2t sin x  x1  x
when t  0.1 . Study the solution behavior from animation in
the form of carpet plot.
y
u n  0

0 .2 u  0 u0
x
u n  0
1 .0
 380 Chapter 10 Partial Differential Equations

18. Solve Problem 17 again when the domain contains two square
holes and a circular hole as shown in the figure. The boundary
conditions along the edges of the holes are u n  0 . Provide
comments on the solution behavior which is different from that
obtained in Problem 17.
y

0.1 0.1 0.1

0 .2
0 .1
x
0.25 0.25 0.25 0.25
1 .0

P-9 19. Use the PDE toolbox to solve the parabolic partial differential
201734070 equation,
u   2u  2u 
   0
t  x 2 y 2 
for the rectangular domain of 0  x  1 and 0  y  0.2 with
the boundary conditions of,
u u
 x,0, t   0,  x,0.2, t   0 , 0  x 1
y y

u0, y, t   1, u1, y, t   0 , 0  y  0.2

and the initial condition of,
1
u x, y,0  1  x  sin2 x 

Plot to compare the solution with the exact solution of,
1
u x, y, t   1  x  e 4 t sin2 x 
2


when t  0.01, 0.02 and 0.05 .
 Exercises 381

P-10 20. Use the PDE toolbox to solve the parabolic partial differential
202205120 equation,
u  2u  2u
7  4 2  2   3 u  12
t  x y 
for the 4  4 unit square domain with two holes as shown in
the figure. The boundary conditions along the outer four edges
are u  0 . The initial condition is given by u x, y,0  0 . Find
the solutions at t  0.2, 0.5 and 1.0 when the boundary
conditions along the hole edges are: (a) u  0 and (b)
u n  0 . Plot the solutions in form of the carpet plot.

y
1 .2

1 .2
.5

4
.7

1 .5

x
1 .5
4

P-11
21. Use the PDE toolbox to solve the hyperbolic partial differential
202205160
equation,
 2u   2u  2u 
   0
t 2  x 2 y 2 
 382 Chapter 10 Partial Differential Equations

for the domain of 0  x  1 and 0  y  0.2 with the boundary

conditions as shown in the figure. The initial conditions are,
u
u x, y,0  sin x and  x, y,0  0
t
Plot to compare the solution with the exact solution of,
u x, y, t   sin x cos t 
when t  1.0 and 2.0 .

y
u n  0

0 .2 u  0 u0

x
u n  0
1 .0

P-12 22. Solve Problem 21 again if the rectangular domain contains

202205180 three small circular holes as shown in the figure. The
boundary conditions along the hole edges are u n  0 .
Compare the solution with that obtained in Problem 21 by
plotting it in form of the carpet plot.

.07 .07 .07

0 .2
0 .1
x
0.25 0.25 0.25 0.25
1 .0
Exercises 383

23. Use the PDE toolbox to solve the hyperbolic partial differential
equation,
 2u   2u  2u 
  2e sin x
t
 
t 2  x 2 y 2 
for a rectangular domain of   0.3 units, i.e., 0  x   and
0  y  0.3 with the boundary conditions of,
u u
 x,0, t   0 ,  x,0.3, t   0 0  x 
y y
u0, y, t   0 , u , y, t   0 0  y  0.3
and the initial conditions of,
u
u x, y,0  sin x and x, y,0   sin x
t
Plot to compare the solution with the exact solution of,
u x, y, t   et sin x
when t  0.2 and 0.5 .

24. Use the PDE toolbox to solve the hyperbolic partial differential
equation,
 2u   2u  2u 
   0
t 2  x 2 y 2 
for a rectangular domain of 1  0.2 units, i.e., 0  x 1 and
0  y  0.2 with the boundary conditions of,
u u
 x,0, t   0,  x,0.2, t   0 0  x 1
y y
u0, y, t   0, u1, y, t   0 0  y  0.2
and the initial conditions of,
u
u x, y,0  sin2 x and  x, y,0  2 sin2 x
t
Plot to compare the solution with the exact solution of,
u x, y, t   sin2 xsin2 t   cos2 t 
when t  0.3 and 0.6 .
 384 Chapter 10 Partial Differential Equations

P-13 25. Use the PDE toolbox to solve the hyperbolic partial differential
202205200 equation,
 2u   2u   2u   2u
3  5 x 2 y 2   4
t 2  
for a rectangular domain of 6  4 units with the boundary
conditions along the outer edges as shown in the figure. The
boundary conditions along the hole edges are u n  0 . The
initial conditions are given by,
u
u x, y,0  0 and  x, y,0  0
t
Find the solutions for the interval of 0  t  1 . Display the
solutions at t  0.3, 0.5 and 1.0 in form of the carpet plot.

3
.75 u n  0

.75

2.5
u0
4
.75

u 1
1.5

x
u n  0 2
6