INQU 6001

§ Most of the natural laws of physics, such as Maxwell’s

equations, Newton’s law of cooling, the Navier-Stokes
equations, Newton’s equations of motion, and
Schrodinger’s equation of quantum mechanics, are
started in term of PDEs, that is, these laws describe
physical phenomena by relating space and time
derivatives
§ A partial differential equation ia an equation that
contains partial derivatives. I contrast to ordinary
differential equations, where the unkown function
depends on one variable, in PDE’s, the unknown function
depends on several variables (like temperature, location,
and time).
 𝜕𝑢 𝜕𝑢 𝜕#𝑢
= 𝑢! = 𝑢" = 𝑢""
𝜕𝑡 𝜕𝑥 𝜕𝑥 #

§ A few well-known PDE’s

𝜕𝑢 𝜕 # 𝑢
𝑢! = 𝑢"" = # Heat equation in one dimensions
𝜕𝑡 𝜕𝑥

𝑢! = 𝑢"" + 𝑢$$ Heat equation in two dimensions

& &
0= 𝑢%% + 𝑢%% + 𝑢'' Laplace’s equation in polar
% %!
coordinates
§ PDEs are classified according to many thing. Classification
is an important concept because the general theory and
methods of solution usually apply to a given class of
equations. Six basic classifications are:
1. Order of the PDE. The order of the PDE is the order of
the highest partial derivative in the equation, for
example

𝜕𝑢 𝜕 # 𝑢
𝑢! = 𝑢"" = # Second order
𝜕𝑡 𝜕𝑥
𝜕𝑢 𝜕𝑢
𝑢! = 𝑢" = First order
𝜕𝑡 𝜕𝑥
() (") Third order
𝑢! = 𝑢𝑢""" + sin 𝑥 =𝑢 + sin 𝑥
(! (" "
2. Number of variables. The number of variables is the
number of independent variables, for example
𝜕𝑢 𝜕 # 𝑢
𝑢! = 𝑢"" = # Two variables: x and t
𝜕𝑡 𝜕𝑥
1 1 Three variables: r, 𝜃, and t
𝑢! = 𝑢%% + 𝑢%% + # 𝑢''
𝑟 𝑟
3. Linearity: PDEs are either liner or nonlinear. In the
linear ones, the dependent variable u and all its
derivatives appear in a linear fashion (they are not
multiplied together or squared, for example). More
precisely a Second order linear equation in two
variables is an equation of the form

𝐴𝑢"" + 𝐵𝑢"$ + 𝐶𝑢$$ + 𝐷𝑢" + E𝑢$ + 𝐹𝑢 = 𝐺

Where A, B, C, D, E, F, and G can be constants or given

functions of x and y; for example
𝑢! = 𝑒 *! 𝑢"" + sin 𝑡 Linear
𝑢𝑢"" + 𝑢! = 0 Nonlinear
𝑢"" + 𝑦𝑢$$ = 0 Linear
𝑥𝑢" + 𝑦𝑢$ + 𝑢# = 0 Nonlinear
4. Homogeneity – The equation bellow is called
homogeneous if the right-hand side G(x,y) is zero for
all x and y. If G(x,y) is not zero. Then the equation is
called nonhomogeneous.

𝐴𝑢"" + 𝐵𝑢"$ + 𝐶𝑢$$ + 𝐷𝑢" + E𝑢$ + 𝐹𝑢 = 𝐺

5. Kind of Coefficient – If the coefficients A, B, C, D, E, F in
the equation bellow are constant, then equation is said
to have constant coefficients (otherwise, variable
coefficients).

𝐴𝑢"" + 𝐵𝑢"$ + 𝐶𝑢$$ + 𝐷𝑢" + E𝑢$ + 𝐹𝑢 = 𝐺

6. Three Basic Types of Linear Equations. All linear PDE’s
like the equation bellow are either

𝐴𝑢"" + 𝐵𝑢"$ + 𝐶𝑢$$ + 𝐷𝑢" + E𝑢$ + 𝐹𝑢 = 𝐺

a) Parabolic
b) Hyperbolic
c) Elliptic

Parabolic equations describe heat flow and diffusion

process and satisfy the property
𝐵# - 4AC = 0
Hyperbolic equations describe vibrating systems and
wave motion and satisfy the property

𝐵# - 4AC > 0

Elliptic equations describe steady state phenomena and

satisfy the property

𝐵# - 4AC < 0

𝐴𝑢"" + 𝐵𝑢"$ + 𝐶𝑢$$ + 𝐷𝑢" + E𝑢$ + 𝐹𝑢 = 𝐺

Examples
a) 𝑢! = 𝑢"" 𝐵# - 4AC = 0 Parabolic
b) 𝑢!! = 𝑢"" 𝐵# - 4AC = 4 Hyperbolic
c) 𝑢+, = 0
𝐵# - 4AC = 1 Hyperbolic
d) 𝑢"" + 𝑢$$ = 0 𝐵# - 4AC = -4 Elliptic
e) 𝑦𝑢"" + 𝑢$$ = 0 𝐵# - 4AC = -4y Elliptic for y > 0
Parabolic for y = 0
Hyperbolic for y < 0

Inthese case of variable coefficients, the solution can

change from point to point

𝐴𝑢"" + 𝐵𝑢"$ + 𝐶𝑢$$ + 𝐷𝑢" + E𝑢$ + 𝐹𝑢 = 𝐺

§ In general, B2 - 4AC is a function of the independent variables,
hence, and equation can be change from one basic type to
another throughout the domain of the equation (although it’s not
common).
§ The general linear equation was written independent of
variables x and y. In many problems, one of the two variables
stands for time and hence would be written in term of x and t.
§ Separation of variables - This technique reduces a PDE in n
variables to n ODEs. This is the most widely used method in
applied mathematics, and its strategy is to break the dependent
variable into component parts, each depending (usually) on a
single independent variable; invariably, it leads to a multiple of
particular solutions.
§ Combination of Variables - Sometimes called a similarity
transform, this technique seeks to combine all independent
variables into one variable, which may then produce a single
ODE. It is applicable only to cases where variables are
unbounded, for example, 0 < t < ∞, 0 < x < ∞. It is suitable for only
one type of boundary or initial condition and produces a single
particular solution.
§ Integral transformation. This procedure reduces a PDE in n
independent variables to one in n-1 variables, a PDE in two
variables could be change to an ODE.
§ Change of coordinates. This method changes the original PDE to
an ODE or else another PDE (an easier one) by changing the
coordinates of the problem.
§ Transformation of the dependent variable. This method
transforms the unknown of a PDE into a New unknown that is
easier to find.
§ Numerical Methods. These methods change a PDE to a system
of difference equations that can be solved by means of iterative
techniques on a computer; in many cases, this is the only
technique that will be work.
§ Perturbation Methods. This methods changes a nonlinear
problem into a sequence of linear ones that approximates the
nonlinear one.
It is clear from these remarks that no general solution is possible,
and the methods that have evolved are closely tied to specific
boundary conditions
§ Type I: Function specified at boundary; for example, for T(r, z)

For example

This means at time zero, temperature T is distributed according to

the function f(x), which does not exclude a simple constant such as
f(x) = Ti. At some boundary, time variation may also occur, so

Neither of these type (1) conditions are homogeneous. However, if

the boundary value is a fixed constant, such as

then we could define a new variable 𝜃 = 𝑇 − 𝑇! and see

which is indeed homogeneous at the boundary.

§ Type II: Derivative of function specified at boundary

This is a classical symmetry condition nearly always arises in

cylindrical and spherical coordinate systems, along the
centerline

For electrically heated tube walls, the heat entering is uniform

and constant, so that we would write at a tube boundary

Since flux is a vector quantity, it is clear that 𝜕𝑇⁄𝜕𝑟 > 0 near the
wall, hence Q is a negative quantity, as it should be since it
moves in the anti-r direction. It is not possible to convert the
electrically heated wall condition to a homogeneous boundary
condition for T. However, the insulated wall is homogeneous
without further modification.
§ Type III: Mixed functions at boundary

This includes both a function and its derivative (or its integral).
Which simply says the conduction flux is exactly balanced by
heat transfer at the wall. By replacing 𝜃 = 𝑇 − 𝑇" (provided Tw is
constant), we see this condition is also homogeneous:

This is a powerful boundary condition, since it actually contains

both type I and type II as subsets. Thus, for large values of h (high
velocity), ℎ → ∞, the boundary condition reduces to 𝜃(z, R) = 0,
which is the homogeneous type I condition. On the other hand, if
h gets very small, then in the limit when h approaches 0, we
obtain type II, which is an insulated boundary.
§ For the physicochemical problems occurring in chemical
engineering, most boundary or initial conditions are (or can be
made to be) of the homogeneous type; a condition or equation
is taken to be homogeneous if, for example, it is satisfied by
y(x), and is also satisfied by 𝜆y(x), where 𝜆 is an arbitrary
constant. The three classical types for such homogeneous
boundary conditions at a point, say x0, are the following:

Most often, the boundary values for a derived model are not
homogeneous but can be made to be so.
§ Separation of variables is one of the oldest techniques for
solving initial boundary-value probles (IBVP’s)and applies to
problems where:
§ The PDE is linear and homogeneous (not necessary constant
coefficients)
§ The boundary conditions are the form

𝛼𝑢# 0, 𝑡 + 𝛽𝑢 0, 𝑡 = 0

𝛾𝑢# 1, 𝑡 + 𝛿𝑢 1, 𝑡 = 0

§ Where 𝛼, 𝛽, 𝛾, and 𝛿 are constant (boundary conditions of this form

are called linear homogeneous BCs)
Consider the IBVP

𝑢$ = 𝛼 % 𝑢## 0≤𝑥≤1 0≤𝑡<∞

BCs 𝑢 0, 𝑡 = 0 0≤𝑡<∞
𝑢 1, 𝑡 = 0
IC 𝑢 𝑥, 0 = 𝑓 𝑥 0≤𝑥≤1

Separation ofvariables looks for simple-type solution to the PDE

of the form

𝑢 𝑥, 𝑡 = 𝑋 𝑥 𝑇 𝑡

By substituting 𝑋 𝑥 𝑇 𝑡 into the PDE and solving for 𝑋 𝑥 𝑇 𝑡

Separation ofvariables looks for simple-type solution to the PDE
of the form
𝑢 𝑥, 𝑡 = 𝑋 𝑥 𝑇 𝑡

By substituting 𝑋 𝑥 𝑇 𝑡 into the PDE and solving for 𝑋 𝑥 𝑇 𝑡

𝑢$ = 𝑋 𝑥 𝑇′ 𝑡 𝑢# = 𝑋′ 𝑥 𝑇 𝑡 𝑢## = 𝑋′′ 𝑥 𝑇 𝑡

𝑋 𝑥 𝑇′ 𝑡 = 𝛼 % 𝑋′′ 𝑥 𝑇 𝑡

and obtain is called separated variables

𝑇′ 𝑡 𝑋′′ 𝑥
= =𝑘
𝛼 𝑇 𝑡
# 𝑋 𝑥
 𝑇′ 𝑡 𝑋′′ 𝑥
= =𝑘
#
𝛼 𝑇 𝑡 𝑋 𝑥

𝑇 & − 𝑘𝛼 % 𝑇 = 0
𝑋′′ − 𝑘𝑋 = 0 Two Ordinary Differential Equations (ODE’s)

So now we can solve each of these two ODE’s, multiply together to

get a solution to the PDE (note that we have essentially changed a
second order PDE to two ODE’s).
However, we now make an important observation, namely, that we
want the separation constant k to be negative (or else the T(t)
factor doesn’t go to zero as 𝑡 → ∞). It is a general practice to
rename 𝑘 = −𝜆% , where 𝜆 is nonzero (-𝜆2 is guaranteed to be
negative)
 𝑇 & + 𝜆% 𝛼 % 𝑇 = 0
𝑋 && + 𝜆% 𝑋 = 0 Two Ordinary Differential Equations (ODE’s)

We will solve the equations

𝑑𝑇
= −𝜆% 𝛼 % 𝑇
𝑑𝑡
𝑑𝑇
= −𝜆% 𝛼 % 𝑑𝑡
𝑇
𝑙𝑛 𝑇 = −𝜆% 𝛼 % 𝑡 + 𝐶

𝑇=𝑒 '(! ) ! $*+ +

=𝑒 𝑒 '(! ) ! $ 𝐷 = 𝑒+
'(! ) ! $
𝑇 = 𝐷𝑒
𝑑% 𝑋 %
%
+ 𝜆 𝑋=0 Homogeneous second order ODE
𝑑𝑥

𝑟 % + 𝜆% = 0

𝑟 % = −𝜆%

𝑟 = −𝜆% = 𝑖 % 𝜆% = 𝜆𝑖

𝑋 𝑥 = 𝐸 𝑠𝑖𝑛𝜆𝑥 + 𝐹 𝑐𝑜𝑠𝜆𝑥

Solution
𝑢 𝑥, 𝑡 = 𝑋 𝑥 𝑇 𝑡
'(! ) ! $
𝑢 𝑥, 𝑡 = 𝐷𝑒 𝐸 𝑠𝑖𝑛𝜆𝑥 + 𝐹 𝑐𝑜𝑠𝜆𝑥 𝐴 = 𝐷𝐸 𝐵 = 𝐷𝐹
'(! ) ! $ A and B are arbitrary constants
𝑢 𝑥, 𝑡 = 𝑒 𝐴 𝑠𝑖𝑛𝜆𝑥 + 𝐵 𝑐𝑜𝑠𝜆𝑥
Find the solution to the PDE and the BC’s
We are now to the point where we have many solutions to the PDE
but not all of then satisfy the BC’s or the IC.

BCs 𝑢 0, 𝑡 = 0 0≤𝑡<∞
𝑢 1, 𝑡 = 0

'(! ) ! $
𝑢 𝑥, 𝑡 = 𝑒 𝐴 𝑠𝑖𝑛𝜆𝑥 + 𝐵 𝑐𝑜𝑠𝜆𝑥
'(! ) ! $
𝑢 0, 𝑡 = 𝑒 𝐵 =0 𝐵=0
'(! ) ! $
𝑢 𝑥, 𝑡 = 𝑒 𝐴 𝑠𝑖𝑛𝜆𝑥
'(! ) ! $ sin 𝜆 = 0
𝑢 1, 𝑡 = 𝑒 𝐴 𝑠𝑖𝑛𝜆
 '(! ) ! $ sin 𝜆 = 0
𝑢 1, 𝑡 = 𝑒 𝐴 𝑠𝑖𝑛𝜆

This BC restrict the separation constant 𝜆 from being any nonzero

number, it must be a root of the equation sin 𝜆 = 0. In other
words, in order that u(1,t) = 0, it necessary
𝜆 = ±𝜋, ±2𝜋, ±3𝜋, …

𝜆, = ±𝑛𝜋 𝑛 = 1, 2, 3, …

' ,-) ! $
𝑢, 𝑥, 𝑡 = 𝐴, 𝑒 𝑠𝑖𝑛 𝑛𝜋𝑥 𝑛 = 1, 2, 3, …

Each one satisfying the PDE and the BC’s

0
' ,-) ! $
𝑢 𝑥, 𝑡 = Q 𝐴, 𝑒 𝑠𝑖𝑛 𝑛𝜋𝑥 𝑛 = 1, 2, 3, …
,./
Find the solution to the PDE, BC’s, and the IC

IC 𝑢 𝑥, 0 = 𝑓 𝑥 0≤𝑥≤1

𝑢 𝑥, 0 = 𝑓 𝑥 = Q 𝐴, 𝑠𝑖𝑛 𝑛𝜋𝑥 𝑛 = 1, 2, 3, …
,./
This equation leads us to the question asked by the French
mathematician Joseph Fourier, is it possible to expand the initial
temperature f(x) as the sum of the elementary functions as
follows:
𝐴/ 𝑠𝑖𝑛 𝜋𝑥 + 𝐴% 𝑠𝑖𝑛 2𝜋𝑥 + 𝐴1 𝑠𝑖𝑛 3𝜋𝑥 +…
The answer to this question is yes, provided f(x) is a reasonably
nice continuous function. Sin(n𝜋x) for n = 1, 2, 3, … know an
orthogonality. It turn out these functions are orthogonal to each
other in the sence
/
0 𝑚≠𝑛
R 𝑠𝑖𝑛 𝑚𝜋𝑥 𝑠𝑖𝑛 𝑛𝜋𝑥 𝑑𝑥 = T1
V2 𝑚 = 𝑛
!
Two functions f1 and f2 are said orthogonal on an interval [a, b] if
3

𝑓/ , 𝑓% = R 𝑓/ 𝑥 𝑓% 𝑥 𝑑𝑥 = 0
2
𝑓 𝑥 = 𝐴/ 𝑠𝑖𝑛 𝜋𝑥 + 𝐴% 𝑠𝑖𝑛 2𝜋𝑥 + 𝐴1 𝑠𝑖𝑛 3𝜋𝑥 +…

We multiply each side of this equation by sin(m𝜋x) (m is an

arbitrary integer) and integrate from zero to one, doing this, we
get
For m = n
/ /

R 𝑓 𝑥 𝑠𝑖𝑛 𝑚𝜋𝑥 𝑑𝑥 = 𝐴4 R 𝑠𝑖𝑛% 𝑚𝜋𝑥 𝑑𝑥 = 1V2 𝐴4

! !
/

𝐴4 = 2 R 𝑓 𝑥 𝑠𝑖𝑛 𝑚𝜋𝑥 𝑑𝑥
!
We’re done; the solution is
0
' ,-) ! $
𝑢 𝑥, 𝑡 = Q 𝐴, 𝑒 𝑠𝑖𝑛 𝑛𝜋𝑥 𝑛 = 1, 2, 3, …
,./

Where
/

𝐴, = 𝐴4 = 2 R 𝑓 𝑥 𝑠𝑖𝑛 𝑛𝜋𝑥 𝑑𝑥
!
 0 /
' ,-) ! $
𝑢 𝑥, 𝑡 = 2 Q R 𝑓 𝑥 𝑠𝑖𝑛 𝑛𝜋𝑥 𝑑𝑥 𝑒 𝑠𝑖𝑛 𝑛𝜋𝑥 𝑛 = 1, 2, 3, …
,./ !

For the boundary conditions

𝑢$ = 𝛼 % 𝑢## 0≤𝑥≤𝐿 0≤𝑡<∞
BCs 𝑢 0, 𝑡 = 0 0≤𝑡<∞
𝑢 𝐿, 𝑡 = 0
IC 𝑢 𝑥, 0 = 𝑓 𝑥 0≤𝑥≤𝐿
Solution
0 5
2 𝑛𝜋𝑥 ,-) ! 𝑛𝜋𝑥
' $
𝑢 𝑥, 𝑡 = Q R 𝑓 𝑥 𝑠𝑖𝑛 𝑑𝑥 𝑒 5 𝑠𝑖𝑛 𝑛 = 1, 2, 3, …
𝐿 𝐿 𝐿
,./ !

f(x) is a cotinuous function in the interval [0, L]

 𝑢$ = 𝑘𝑢## 0≤𝑥≤𝐿 0≤𝑡<∞

BCs 𝑢 0, 𝑡 = 0 0≤𝑡<∞
𝑢 𝐿, 𝑡 = 0
IC 𝑢 𝑥, 0 = 𝑓 𝑥 0≤𝑥≤𝐿
u is the temperature in a rod of length L
Solution
0 5
2 𝑛𝜋𝑥 '6 , ! - !7 $ 𝑛𝜋𝑥
𝑢 𝑥, 𝑡 = Q R 𝑓 𝑥 𝑠𝑖𝑛 𝑑𝑥 𝑒 5! 𝑠𝑖𝑛 𝑛 = 1, 2, 3, …
𝐿 𝐿 𝐿
,./ !

Example: for k = 1, L = 𝜋, and f(x) = 100

0 -
2 ', ! $
𝑢 𝑥, 𝑡 = Q R 100𝑠𝑖𝑛 𝑛𝑥 𝑑𝑥 𝑒 𝑠𝑖𝑛 𝑛𝑥 𝑛 = 1, 2, 3, …
𝜋
,./ !
 u is the temperature in a rod of length L
Example: for k = 1, L = 𝜋, and f(x) = 100

0 -
2 ', ! $
𝑢 𝑥, 𝑡 = Q R 100𝑠𝑖𝑛 𝑛𝑥 𝑑𝑥 𝑒 𝑠𝑖𝑛 𝑛𝑥 𝑛 = 1, 2, 3, …
𝜋
,./ !

-
,
100 1 − −1
𝐴, = R 100𝑠𝑖𝑛 𝑛𝑥 𝑑𝑥 =
𝜋 𝑛
!

0
,
200 1 − −1 ', ! $
𝑢 𝑥, 𝑡 = Q 𝑒 𝑠𝑖𝑛 𝑛𝑥 𝑛 = 1, 2, 3, …
𝜋 𝑛
,./
u(x, t) graphed as a function of x for various fixed times
u(x, t) graphed as a function of t for various fixed positions
Steady State temperature u(x,y) in a rectangular plate

𝑢## + 𝑢88 = 0 0<𝑥<𝑎 0<𝑦<𝑏

Baundary Conditions

𝜕𝑢
[ =0 0<𝑦<𝑏
𝜕𝑥 #.!

𝜕𝑢
[ =0 0<𝑦<𝑏
𝜕𝑥 #.2

𝑢 𝑥, 0 = 0 0<𝑥<𝑎 Steady-state temperatures in a rectangular plate

𝑢 𝑥, 𝑏 = 𝑓 𝑥 0<𝑥<𝑎
 Separation ofvariables looks for simple-type solution to the PDE
of the form
𝑢 𝑥, 𝑦 = 𝑋 𝑥 𝑌 𝑦

By substituting 𝑋 𝑥 𝑌 𝑦 into the PDE and solving for 𝑋 𝑥 𝑌 𝑦

𝑢# = 𝑋′ 𝑥 𝑌 𝑦 𝑢## = 𝑋′′ 𝑥 𝑌 𝑦 𝑢8 = 𝑋 𝑥 𝑌′ 𝑦 𝑢88 = 𝑋 𝑥 𝑌′′ 𝑦

𝑋′′ 𝑥 𝑌 𝑦 + 𝑋 𝑥 𝑌′′ 𝑦 = 0

and obtain is called separated variables

𝑋′′ 𝑥 𝑌 -- $
=− = −𝜆
𝑋 𝑥 𝑌 𝑦
 𝑋′′ 𝑥 𝑌 -- $
=− = −𝜆
𝑋 𝑥 𝑌 𝑦

𝑋′& + 𝜆𝑋 = 0
Two Ordinary Differential Equations (ODE’s)
Y′′ − 𝜆𝑋 = 0
BC’s: X’(0) = 0, X’(a) = 0, and Y(0) = 0.
The Sturn-Liouville problem associated with equation of X.

𝑋′& + 𝜆𝑋 = 0 𝑋& 0 = 0 𝑋& 𝑎 = 0

Three possible cases: 𝜆 = 0, 𝜆 < 0,and 𝜆 > 0

Case I: 𝜆 = 0

𝑋′& + 𝜆𝑋 = 0 𝑋& 0 = 0 𝑋& 𝑎 = 0

𝑋′& = 0

𝑟% = 0 𝑟=0

X = 𝐶/ + 𝐶% x

BC’s
X′ 0 = 𝐶% 𝐶% = 0

X = 𝐶/ X′ 𝑎 = 0

X = 𝐶/ 𝐶/ ≠ 0 Note that for any C1, this constant solution

satisfies the second boundary condition X’(a) = 0
Case II: 𝜆 = -𝛼2 < 0

𝑋′& + 𝜆𝑋 = 0 𝑋& 0 = 0 𝑋& 𝑎 = 0

𝑋′& − 𝛼 % 𝑋 = 0

𝑟 % −𝛼 % = 0 𝑟 = ±𝛼

X = 𝐶/ cosh 𝛼𝑥 + 𝐶% sinh 𝛼𝑥

BC’s
X & = 𝐶/ 𝛼 sinh 𝛼𝑥 + 𝐶% 𝛼 cosh 𝛼𝑥

X & 0 = 0 + 𝐶% 𝛼 = 0 𝐶% = 0

X = 𝐶/ cosh 𝛼𝑥
 X = 𝐶/ cosh 𝛼𝑥

X′ 𝑎 = 0 X & = 𝐶/ 𝛼 sinh 𝛼𝑥

0 = 𝐶/ 𝛼 sinh 𝛼𝑥
sinh 𝛼𝑥 = 0

Only when 𝛼 = 0 C1 = 0

For 𝜆 = -𝛼2 < 0 trivial solution

Case III: 𝜆 = 𝛼2 > 0

𝑋′& + 𝜆𝑋 = 0 𝑋& 0 = 0 𝑋& 𝑎 = 0

𝑋′& + 𝛼 % 𝑋 = 0

𝑟 % +𝛼 % = 0 𝑟 = 𝑖𝛼

X = 𝐶/ cos 𝛼𝑥 + 𝐶% sin 𝛼𝑥

BC’s
X & = −𝐶/ 𝛼 sin 𝛼𝑥 + 𝐶% 𝛼 cos 𝛼𝑥

X & 0 = 0 + 𝐶% 𝛼 = 0 𝐶% = 0

X = 𝐶/ cos 𝛼𝑥
 X = 𝐶/ cos 𝛼𝑥

X′ 𝑎 = 0 X & = −𝐶/ 𝛼 si𝑛 𝛼𝑥

0 = 𝐶/ 𝛼 sin 𝛼𝑎

sin 𝛼𝑎 = 0 𝛼>0 C1 ≠ 0

sin 𝛼𝑎 = 0 𝑛𝜋
𝛼𝑎 = 𝑛𝜋 𝛼= 𝑛 = 1, 2, 3, …
𝑎

The eigenvalues of the equation are the 𝜆0 and 𝜆n =𝛼,% = , ! - !V , n = 1,2,3,…

2!

For 𝜆0 = 0 with n = 0, the eigenvalues are

𝑛𝜋
X = 𝐶/ X = 𝐶/ cos 𝑥
𝑎
 Y′′ − 𝜆𝑌 = 0
For 𝜆0 = 0 Y′′ = 0
Solution

𝑌 = 𝐶1 + 𝐶9 𝑦

BC 𝑌 0 =0 𝐶1 = 0

𝑌 = 𝐶9 𝑦

, !-!
For 𝜆n = V2!

Solution

Y′′ − , ! - !V 𝑌=0
2!

% %
%
𝑟 − 𝑛 𝜋 V =0 𝑟 = ± 𝑛𝜋⁄𝑎
𝑎 %
Becouse 0 < y < b is a finite interval, we write the general
solution in terms of hypeobolic functions
𝑛𝜋 𝑛𝜋
𝑌 = 𝐶1 𝑐𝑜𝑠ℎ 𝑦 + 𝐶9 𝑠𝑖𝑛ℎ 𝑦
𝑎 𝑎
𝑛𝜋
BC 𝑌 0 =0 𝑌 = 𝐶1 𝑐𝑜𝑠ℎ 𝑦 =0 𝐶1 = 0
𝑎
𝑛𝜋
𝑌 = 𝐶9 𝑠𝑖𝑛ℎ 𝑦
𝑎

This product solution un = X(x)Y(y) that satisfy the Laplace’s

equation and the tree homogeneous boundary conditions

𝑢! 𝑥, 𝑦 = 𝐴! 𝑦 𝑛=0 𝐴! = 𝐶/ 𝐶9
𝑛𝜋 𝑛𝜋
𝑢, 𝑥, 𝑦 = 𝐴, 𝑠𝑖𝑛ℎ 𝑦 𝑐𝑜𝑠 𝑥 𝑛 = 1, 2, 3, …
𝑎 𝑎
The superposition principle yields another solution
0
𝑛𝜋 𝑛𝜋
𝑢 𝑥, 𝑦 = 𝐴! 𝑦 + Q 𝐴, 𝑠𝑖𝑛ℎ 𝑦 𝑐𝑜𝑠 𝑥
𝑎 𝑎
,./
Substituting the y = b 𝑢 𝑥, 𝑏 = 𝑓 𝑥 0 < 𝑥 < 𝑎)
0
𝑛𝜋 𝑛𝜋
𝑢 𝑥, 𝑦 = 𝑓 𝑥 = 𝐴! 𝑏 + Q 𝐴, 𝑠𝑖𝑛ℎ 𝑏 𝑐𝑜𝑠 𝑥
𝑎 𝑎
,./

Is a half-range expansion of f(x) in a Fourier cosine series. If we

make the identification A0b = a0/2 and An sinh(n𝜋b/a) = an, n = 1, 2,
3,…
2 2 1 2
2𝐴! 𝑏 = R 𝑓 𝑥 𝑑𝑥 𝐴! = R 𝑓 𝑥 𝑑𝑥
𝑎 ! 𝑎𝑏 !
 The superposition principle yields another solution
𝑛𝜋 2 2 𝑛𝜋
𝐴, 𝑠𝑖𝑛ℎ 𝑏 = R 𝑓 𝑥 𝑐𝑜𝑠 𝑥 𝑑𝑥
𝑎 𝑎 ! 𝑎
2
2 𝑛𝜋
𝐴, = 𝑛𝜋 R 𝑓 𝑥 𝑐𝑜𝑠 𝑎 𝑥 𝑑𝑥
𝑎𝑠𝑖𝑛ℎ 𝑏 !
𝑎

Solution
" & "
1 2 𝑛𝜋 𝑛𝜋 𝑛𝜋
𝑢 𝑥, 𝑦 = * 𝑓 𝑥 𝑑𝑥 𝑦 + . 𝑛𝜋 * 𝑓 𝑥 𝑐𝑜𝑠 𝑥 𝑑𝑥 𝑠𝑖𝑛ℎ 𝑦 𝑐𝑜𝑠 𝑥
𝑎𝑏 ! 𝑎𝑠𝑖𝑛ℎ 𝑏 ! 𝑎 𝑎 𝑎
#$% 𝑎
Steady State temperature u(x,y) in a rectangular plate

𝑢## + 𝑢88 = 0 0<𝑥<𝑎 0<𝑦<𝑏

y
Baundary Conditions

𝑢 0, 𝑦 = 0 0<𝑦<𝑏 u = f(x)
(a, b)

𝑢 𝑎, 𝑦 = 0 0<𝑦<𝑏 u=0 u=0

x
𝑢 𝑥, 0 = 0 0<𝑥<𝑎 u=0

𝑢 𝑥, 𝑏 = 𝑓 𝑥 0<𝑥<𝑎
 Separation ofvariables looks for simple-type solution to the PDE
of the form
𝑢 𝑥, 𝑡 = 𝑋 𝑥 𝑌 𝑦

By substituting 𝑋 𝑥 𝑌 𝑦 into the PDE and solving for 𝑋 𝑥 𝑌 𝑦

𝑢# = 𝑋′ 𝑥 𝑌 𝑦 𝑢## = 𝑋′′ 𝑥 𝑌 𝑦 𝑢8 = 𝑋 𝑥 𝑌′ 𝑦 𝑢## = 𝑋 𝑥 𝑌′′ 𝑦

𝑋′′ 𝑥 𝑌 𝑦 + 𝑋 𝑥 𝑌′′ 𝑦 = 0

and obtain is called separated variables

𝑋′′ 𝑥 𝑌 -- $
=− = −𝜆
𝑋 𝑥 𝑌 𝑦
 𝑋′′ 𝑥 𝑌 -- $
=− = −𝜆
𝑋 𝑥 𝑌 𝑦

𝑋′& + 𝜆𝑋 = 0
Two Ordinary Differential Equations (ODE’s)
Y′′ − 𝜆𝑋 = 0
BC’s: X(0) = 0, X(a) = 0, and Y(0) = 0.
The Sturn-Liouville problem associated with equation of X.

𝑋′& + 𝜆𝑋 = 0 𝑋 0 =0 𝑋 𝑎 =0

Three possible cases: 𝜆 = 0, 𝜆 < 0,and 𝜆 > 0

Case I: 𝜆 = 0

𝑋′& + 𝜆𝑋 = 0 𝑋 0 =0 𝑋 𝑎 =0

𝑋′& = 0

𝑟% = 0 𝑟=0

X = 𝐶/ + 𝐶% x

BC’s
X 0 = 𝐶/ 𝐶/ = 0

X = 𝐶% 𝑥 X 𝑎 =0

X = 𝐶% 𝑎 𝐶% = 0 Trivial solution
Case II: 𝜆 = -𝛼2 < 0

𝑋′& + 𝜆𝑋 = 0 𝑋 !
=0 𝑋 2
=0

𝑋′& − 𝛼 % 𝑋 = 0

𝑟 % −𝛼 % = 0 𝑟 = ±𝛼

X = 𝐶/ cosh 𝛼𝑥 + 𝐶% sinh 𝛼𝑥

BC’s: X(0) = 0, X(a) = 0,

𝑋 0 = 𝐶/ + 0 = 0 𝐶/ = 0

X = 𝐶% sinh 𝛼𝑥
Case III: 𝜆 = 𝛼2 > 0

𝑋′& + 𝜆𝑋 = 0 𝑋& 0 = 0 𝑋& 𝑎 = 0

𝑋′& + 𝛼 % 𝑋 = 0

𝑟 % +𝛼 % = 0 𝑟 = 𝑖𝛼

X = 𝐶/ cos 𝛼𝑥 + 𝐶% sin 𝛼𝑥

BC’s: X(0) = 0, X(a) = 0,

X 0 = 𝐶/ + 0 = 0 𝐶/ = 0

X = 𝐶% sin 𝛼𝑥
X = 𝐶% sin 𝛼𝑥

X 𝑎 =0

0 = 𝐶% sin 𝛼𝑎

sinh 𝛼𝑎 = 0 𝛼>0 C1 ≠ 0

sinh 𝛼𝑎 = 0 𝑛𝜋
𝛼𝑎 = 𝑛𝜋 𝛼= 𝑛 = 1, 2, 3, …
𝑎
The eigenvalues of the equation is 𝜆n =𝛼,% = , ! - !V , n = 1,2,3,…
2!

𝑛𝜋
X = 𝐶/ sin 𝑥
𝑎
 Y′′ − 𝜆𝑌 = 0
, !-!
For 𝜆n = V2!

Solution

Y′′ − , ! - !V 𝑌=0
2!

% %
%
𝑟 − 𝑛 𝜋 V =0 𝑟 = ± 𝑛𝜋⁄𝑎
𝑎 %

Becouse 0 < y < b is a finite interval, we write the general

solution in terms of hypeobolic functions
𝑛𝜋 𝑛𝜋
𝑌 = 𝐶1 𝑐𝑜𝑠ℎ 𝑦 + 𝐶9 𝑠𝑖𝑛ℎ 𝑦
𝑎 𝑎
𝑛𝜋
BC 𝑌 0 =0 𝑌 = 𝐶1 𝑐𝑜𝑠ℎ 𝑦 =0 𝐶1 = 0
𝑎
𝑛𝜋
𝑌 = 𝐶9 𝑠𝑖𝑛ℎ 𝑦
𝑎
This product solution un = X(x)Y(y) that satisfy the Laplace’s
equation and the tree homogeneous boundary conditions

𝑛𝜋 𝑛𝜋
𝑢, 𝑥, 𝑦 = 𝐴, 𝑠𝑖𝑛ℎ 𝑦 𝑠𝑖𝑛 𝑥 𝑛 = 1, 2, 3, …
𝑎 𝑎
Substituting the BC 𝑢 𝑥, 𝑏 = 𝑓 𝑥 0<𝑥<𝑎
0
𝑛𝜋 𝑛𝜋
𝑢 𝑥, 𝑦 = 𝑓 𝑥 = Q 𝐴, 𝑠𝑖𝑛ℎ 𝑏 𝑠𝑖𝑛 𝑥
𝑎 𝑎
,./
The expansion of f(x) in a Fourier sine series. If we make the
identification An sinh(n𝜋b/a) = an, n = 1, 2, 3,…

𝑛𝜋 2 2 𝑛𝜋
𝐴, 𝑠𝑖𝑛ℎ 𝑏 = R 𝑓 𝑥 𝑠𝑖𝑛 𝑥 𝑑𝑥
𝑎 𝑎 ! 𝑎
2
2 𝑛𝜋
𝐴, = 𝑛𝜋 R 𝑓 𝑥 𝑠𝑖𝑛 𝑎 𝑥 𝑑𝑥
𝑎𝑠𝑖𝑛ℎ 𝑏 !
𝑎
 Solution
& (
2 𝑛𝜋 𝑛𝜋 𝑛𝜋
𝑢 𝑥, 𝑦 = * 𝑛𝜋 3 𝑓 𝑥 𝑠𝑖𝑛 𝑎 𝑥 𝑑𝑥 𝑠𝑖𝑛ℎ 𝑎 𝑦 𝑠𝑖𝑛 𝑎 𝑥
#$% 𝑎𝑠𝑖𝑛ℎ 𝑎 𝑏 '

In the special case when f(x) = 100, a = 1, b = 1, the coefficients An

are given by

0 /
2 2
𝑢 𝑥, 𝑡 = Q R 100𝑠𝑖𝑛 𝑛𝑥 𝑑𝑥 𝑠𝑖𝑛 𝑛𝑥 𝑠𝑖𝑛ℎ 𝑛𝑦 𝑛 = 1, 2, 3, …
1 𝑠𝑖𝑛ℎ 𝑛𝜋
,./ !

/
2 200 1 − −1 ,
𝐴, = R 100𝑠𝑖𝑛 𝑛𝑥 𝑑𝑥 =
𝑠𝑖𝑛ℎ 𝑛𝜋 𝑛 𝑠𝑖𝑛ℎ 𝑛𝜋
!
0
1 1 − −1 ,
𝑢 𝑥, 𝑡 = 200 Q 𝑠𝑖𝑛 𝑛𝑥 𝑠𝑖𝑛ℎ 𝑛𝑦 𝑛 = 1, 2, 3, …
𝑛 𝑠𝑖𝑛ℎ 𝑛𝜋
,./
 0
1 1 − −1 ,
𝑢 𝑥, 𝑡 = 200 Q 𝑠𝑖𝑛 𝑛𝑥 𝑠𝑖𝑛ℎ 𝑛𝑦 𝑛 = 1, 2, 3, …
𝑛 𝑠𝑖𝑛ℎ 𝑛𝜋
,./

The isotherms can also be visualized as

the curves of intersection (projected into
the xy-plane) of horizontal planes u = 80,
u = 60, u = 40, u = 20, and u = 10
A boundary- value problem is said to be nonhomogeneous when
either the partial differential equation or the boundary conditions
are nonhomegeneous. For example, a typical nonhomogeneous
BVP for the heat equation is

𝜕𝑢 𝜕%𝑢
= 𝑘 % + 𝐹 𝑥, 𝑡 0 < 𝑥 < 𝐿, 𝑡>0
𝜕𝑡 𝜕𝑥

BCs 𝑢 0, 𝑡 = 𝑢! 𝑡 𝑡>0
𝑢 𝐿, 𝑡 = 𝑢/ 𝑡
IC 𝑢 𝑥, 0 = 𝑓 𝑥 0≤𝑥≤𝐿

F(x,t) = rate of heat is being generated

We consider certain types of nonhomogeneous boundary-value
problems that can be solved by changing the dependent variable
u to a new variable 𝜈 by means of the substitution u = 𝜈 + 𝜓,
where 𝜓 is a function to be determined.

Time-independent PDE and BC’s

𝜕𝑢 𝜕%𝑢
=𝑘 % +𝐹 𝑥 0 < 𝑥 < 𝐿, 𝑡>0
𝜕𝑡 𝜕𝑥

BCs 𝑢 0, 𝑡 = 𝑢! 𝑡>0
𝑢 𝐿, 𝑡 = 𝑢/
IC 𝑢 𝑥, 0 = 𝑓 𝑥 0≤𝑥≤𝐿

u0 and u1 are constants

By changing the dependent variable u to a new variable 𝜈 by
means of the substitution u(x,t) = 𝜈(x,t) + 𝜓(x), can be reduced to
two problems
Problem1: ODE
𝜕𝑢
=0
𝜕𝑡

𝜕%𝜓
𝑘 % +𝐹 𝑥 =0 0 < 𝑥 < 𝐿, 𝑡>0
𝜕𝑥

BCs 𝜓 0 = 𝑢! 𝑡>0
𝜓 𝐿 = 𝑢/
Problem 2: homogeneous BVP PDE
𝜕𝜈 𝜕%𝜈
=𝑘 % 0 < 𝑥 < 𝐿, 𝑡>0
𝜕𝑡 𝜕𝑥

BCs 𝜈 0, 𝑡 = 0 𝑡>0
𝜈 𝐿, 𝑡 = 0
IC 𝜈 𝑥, 0 = 𝑓 𝑥 − 𝜓 𝑥 0≤𝑥≤𝐿

Solution for u = solution of Problem 1 + solution of Problem 2

𝜕𝑢 𝜕%𝑢
=𝑘 % +𝑟 0 < 𝑥 < 1, 𝑡>0
𝜕𝑡 𝜕𝑥
𝑢=0 𝑢 = 𝑢!
BCs 𝑢 0, 𝑡 = 0 𝑡>0
𝑢 1, 𝑡 = 𝑢/
IC 𝑢 𝑥, 0 = 𝑓 𝑥 0≤𝑥≤1
𝐿=1

Both PDE and the condition at the right boundary x = 1 are

nonhomogeneous

Substitution u(x,t) = 𝜈(x,t) + 𝜓(x)

𝜕%𝑢 𝜕%𝜈 𝜕%𝜓 𝜕𝑢 𝜕𝜈
%
= + =
𝜕𝑥 𝜕𝑥 % 𝜕𝑥 % 𝜕𝑡 𝜕𝑡

𝜕𝜈 𝜕%𝜈 𝜕%𝜓
=𝑘 +𝑘 % +𝑟
𝜕𝑡 𝜕𝑥 % 𝜕𝑥
𝜕𝜈 𝜕%𝜈 𝜕%𝜓
=𝑘 + 𝑘 + 𝑟
𝜕𝑡 𝜕𝑥 % 𝜕𝑥 %

𝑢 0, 𝑡 = 𝜈 0, 𝑡 + 𝜓 0 = 0 𝑡>0
𝑢 1, 𝑡 = 𝜈 1, 𝑡 + 𝜓 1 = 𝑢/
This equation reduces to a homogeneous PDE if we demand that 𝜓
be a function that satiesfies the ODE

𝑑% 𝜓 𝑢 = 𝑢!
𝑘 % +𝑟 =0 𝑢=0
𝑑𝑥
𝑑% 𝜓 𝑟
%
=−
𝑑𝑥 𝑘 𝐿=1

Solution
𝑟 %
𝜓 = 𝜓= + 𝜓> = 𝐶/ 𝑥 + 𝐶% − 𝑥
2𝑘
we have 𝜓(0) = 0 and 𝜓(1) = u1 𝑢=0 𝑢 = 𝑢!

Applying the BC
𝑟 % 𝐿=1
𝜓 = 𝜓= + 𝜓> = 𝐶/ 𝑥 + 𝐶% − 𝑥
2𝑘
𝜓 0 =0 𝐶% = 0
𝑟 %
𝜓 = 𝜓= + 𝜓> = 𝐶/ 𝑥 − 𝑥
2𝑘
? 𝑟
𝜓 1 = 𝑢/ = 𝐶/ − 𝐶/ = 𝑢/ +
%6 2𝑘
𝑟 𝑟 %
𝜓 = 𝜓= + 𝜓> = 𝑢/ + 𝑥− 𝑥
2𝑘 2𝑘
 𝑢=0 𝑢 = 𝑢!
Finally
𝜕𝜈 𝜕%𝜈
=𝑘
𝜕𝑡 𝜕𝑥 % 𝐿=1

BCs 𝜈 0, 𝑡 = 0 𝑡>0
𝜈 1, 𝑡 = 0
IC 𝑟 𝑟 %
𝜈 𝑥, 0 = 𝑓 𝑥 − 𝑢/ + 𝑥+ 𝑥 0≤𝑥≤𝐿
2𝑘 2𝑘

Solution using separation of variables

0
'6 ,- ! $
𝜈 𝑥, 𝑡 = Q 𝐴, 𝑒 𝑠𝑖𝑛 𝑛𝜋𝑥 𝑛 = 1, 2, 3, …
,./
/
𝑟 𝑟 %
𝐴, = 2 R 𝑓 𝑥 − 𝑢/ + 𝑥+ 𝑥 𝑠𝑖𝑛 𝑛𝜋𝑥 𝑑𝑥
2𝑘 2𝑘
!
 𝑢=0 𝑢 = 𝑢!

0
'6 ,- ! $
𝜈 𝑥, 𝑡 = Q 𝐴, 𝑒 𝑠𝑖𝑛 𝑛𝜋𝑥 𝑛 = 1, 2, 3, … 𝐿=1

,./
/
𝑟 𝑟 %
𝐴, = 2 R 𝑓 𝑥 − 𝑢/ + 𝑥+ 𝑥 𝑠𝑖𝑛 𝑛𝜋𝑥 𝑑𝑥
2𝑘 2𝑘
!

The solution of the original proble is obtained by adding 𝜓(x) and 𝜈(x,t)
% !
𝑟 𝑟 " 𝑟 𝑟 " #) ! *
𝑢 𝑥, 𝑡 = 𝑢! + 𝑥− 𝑥 + + 2 , 𝑓 𝑥 − 𝑢! + 𝑥+ 𝑥 𝑠𝑖𝑛 𝑛𝜋𝑥 𝑑𝑥 𝑒 '( 𝑠𝑖𝑛 𝑛𝜋𝑥
2𝑘 2𝑘 2𝑘 2𝑘
#$! &

Observe that u(x,t) → 𝜓(x) as t → ∞. In the context of the given BVP, 𝜓 is

called the steady-state solution.

𝜈(x,t) → 0 as t → ∞, 𝜈 is called the transient solution.

𝜕𝑢 𝜕%𝑢
= 𝑘 % + 𝐹 𝑥, 𝑡 0 < 𝑥 < 𝐿, 𝑡>0
𝜕𝑡 𝜕𝑥

BCs 𝑢 0, 𝑡 = 𝑢! 𝑡 𝑡>0
𝑢 𝐿, 𝑡 = 𝑢/ 𝑡
IC 𝑢 𝑥, 0 = 𝑓 𝑥 0≤𝑥≤𝐿

Where the heat source term F and the boundary conditions are
time dependent

Substitution u(x,t) = 𝜈(x,t) + 𝜓(x,t)

𝜕%𝑢 𝜕%𝜈 𝜕%𝜓 𝜕𝑢 𝜕𝜈 𝜕𝜓

= + % = +
𝜕𝑥 %
𝜕𝑥 % 𝜕𝑥 𝜕𝑡 𝜕𝑡 𝜕𝑡
 𝜕%𝜈 𝜕%𝜓 𝜕𝜈 𝜕𝜓
𝑘 + 𝑘 % + 𝐹 𝑥, 𝑡 = +
𝜕𝑥 % 𝜕𝑥 𝜕𝑡 𝜕𝑡
𝑢 0, 𝑡 = 𝜈 0, 𝑡 + 𝜓 0, 𝑡 = 𝑢! 𝑡 𝑡>0
𝑢 1, 𝑡 = 𝜈 𝐿, 𝑡 + 𝜓 𝐿, 𝑡 = 𝑢/ 𝑡
𝜈 𝑥, 0 = 𝑓 𝑥 −𝜓 𝑥, 0 0 ≤ 𝑥 ≤ 𝐿

The boundary conditions on 𝜈 will be homogeneous if we

demand that
𝜓 0, 𝑡 = 𝑢! 𝑡
𝜓 𝐿, 𝑡 = 𝑢/ 𝑡
To force the problem to be homogeneous by solving
𝑘𝜓## + 𝐹 𝑥, 𝑡 = 𝜓$
If the defining equation for 𝜓 is itself a nonhomogeneous PDE,
this is an unrealistic expectation. We try an entirely different track
by simply constructing a function 𝜓 that satisfies both conditions.
One such a function is given by
𝑥
𝜓 𝑥, 𝑡 = 𝑢! 𝑡 + 𝑢/ 𝑡 − 𝑢! 𝑡
𝐿
Reinspection of equation and boundary conditions shows that we
have gained some additional simplification with this choice of 𝜓
since 𝜓xx = 0, we now start over. This time if we substitute
𝑥
𝑢 𝑥, 𝑡 = 𝜈 𝑥, 𝑡 + 𝑢! 𝑡 + 𝑢/ 𝑡 − 𝑢! 𝑡
𝐿
𝜕𝜈 𝜕%𝜈
= 𝑘 % + 𝐹 𝑥, 𝑡 0 < 𝑥 < 𝐿, 𝑡>0
𝜕𝑡 𝜕𝑥

BCs 𝜈 0, 𝑡 = 0 𝑡>0
𝜈 𝐿, 𝑡 = 0
IC 𝜈 𝑥, 0 = 𝑓 𝑥 −𝜓 𝑥, 0 0≤𝑥≤𝐿
Where F(x,t) while the problem is still nonhomogeneous (the
boundaru conditions are homogeneous but the partial differential
equation is nonhomogeneous) it is a problem that we can solve.
Basic Strategy
Assume that time-dependent coefficient 𝜈n(t) and Gn(t) can be
found such that both 𝜈(x,t) and G(x,t) can be expandend in series
0 0
𝑛𝜋 𝑛𝜋
𝜈 𝑥, 𝑡 = Q 𝜈, 𝑡 𝑠𝑖𝑛 𝑥 𝐺 𝑥, 𝑡 = Q 𝐺, 𝑡 𝑠𝑖𝑛 𝑥
𝐿 𝐿
,./ ,./

where sin(n𝜋x/L), n = 1, 2, 3, . . . , are the eigenfunctions of X" + 𝜆X

= 0, X(0) = 0, X(L) = 0 corresponding to the eigenvalues 𝜆, = 𝛼,% =
𝑛% 𝜋 % ⁄𝐿% .
The latter problem would have been obtained had separation of
variables been applied to the associated homogeneous PDE. For
𝜈(x,t) and G(x,t) observe that the assumed form for u(x,t) already
satisfies the boundary conditions. The basic idea here is to
substitute the first series in 𝜈(x,t) and G(x,t) into the
nonhomogeneous PDE, collect terms, and equate the resulting
series with the actual series expansion found for F(x, t)
Solve 𝜕𝑢 𝜕%𝑢
= 𝑘 % + 𝐹 𝑥, 𝑡 0 < 𝑥 < 1, 𝑡>0
𝜕𝑡 𝜕𝑥
BCs 𝑢 0, 𝑡 = cos 𝑡 𝑡>0 𝑢 𝐿, 𝑡 = 0

IC 𝑢 𝑥, 0 = 0 0≤𝑥≤1

𝐹𝑜𝑟 𝐹 𝑥, 𝑡 = 0 𝑘=1

u(x,t) = 𝜈(x,t) + 𝜓(x,t)

𝑥
𝜓 𝑥, 𝑡 = 𝑢! 𝑡 + 𝑢/ 𝑡 − 𝑢! 𝑡
𝐿

𝜓 𝑥, 𝑡 = cos 𝑡 − 𝑥 cos 𝑡 = 1 − 𝑥 cos 𝑡

𝑢 𝑥, 𝑡 = 𝜈 𝑥, 𝑡 + 1 − 𝑥 cos 𝑡
To obtain a BVP for 𝜈(x,t)

𝜕%𝜈 𝜕𝜈
%
+ 1 − 𝑥 sin 𝑡 =
𝜕𝑥 𝜕𝑡
𝜈 0, 𝑡 = 0 𝑡>0 𝜈 𝐿, 𝑡 = 0

𝜈 𝑥, 0 = 0 − 1 − 𝑥 cos 0 = 𝑥 − 1

where sin(n𝜋x), n = 1, 2, 3, . . . , are the eigenfunctions of X" + 𝜆X = 0,

X(0) = 0, X(1) = 0 corresponding to the eigenvalues 𝜆, = 𝛼,% = 𝑛% 𝜋 % .
G(x,t) = 1 − 𝑥 sin 𝑡 we assume that the fixed t, 𝜈, G can be written
as Fourier sine series

0 0

𝜈 𝑥, 𝑡 = Q 𝜈, 𝑡 sin 𝑛𝜋𝑥 𝐺 𝑥, 𝑡 = 1 − 𝑥 sin 𝑡 = Q 𝐺, 𝑡 sin 𝑛𝜋𝑥

,./ ,./
 / /
𝐺, 𝑡 = 2 R 1 − 𝑥 sin 𝑡 sin 𝑛𝜋𝑥 𝑑𝑥 = 2 sin 𝑡 R 1 − 𝑥 sin 𝑛𝜋𝑥 𝑑𝑥
! !

2
𝐺, 𝑡 = sin 𝑡
𝑛𝜋
0
2
1 − 𝑥 sin 𝑡 = Q sin 𝑡 sin 𝑛𝜋𝑥
𝑛𝜋
,./

We can determine the coefficients 𝜈, 𝑡 by substituting 𝜈 𝑥, 𝑡 and

1 − 𝑥 sin 𝑡 back into the PDE. To and with the partial differential
derivatives of 𝜈 are
0
0
𝜈 𝑥, 𝑡 = Q 𝜈, 𝑡 sin 𝑛𝜋𝑥 𝜕𝜈
= Q 𝜈, ′ 𝑡 sin 𝑛𝜋𝑥
,./ 𝜕𝑡
,./
0
%
𝜕 𝜈 % %
%
= Q 𝜈, 𝑡 −𝑛 𝜋 sin 𝑛𝜋𝑥
𝜕𝑥
,./
Write the PDE

𝜈$ − 𝜈## = 1 − 𝑥 sin 𝑡

0 0
& $ % %
2
Q 𝜈, + 𝑛 𝜋 𝜈, 𝑡 sin 𝑛𝜋𝑥 = Q sin 𝑡 sin 𝑛𝜋𝑥
𝑛𝜋
,./ ,./

0 0
& $ % %
2
Q 𝜈, + 𝑛 𝜋 𝜈, 𝑡 sin 𝑛𝜋𝑥 = Q sin 𝑡 sin 𝑛𝜋𝑥
𝑛𝜋
,./ ,./

& $ % % %
𝜈, + 𝑛 𝜋 𝜈, 𝑡 = sin 𝑡
,-

Linear first order ODE whose general solution

% , ! - ! @AB $'CD@ $ ', ! - ! $

𝜈, 𝑡 = + 𝐶, 𝑒
,- , " - " */
 5 4* 6* 789 :;<=7 : ;4* 6* :
𝜈4 𝑡 = + 𝐶4 𝑒
46 4+ 6+ >?

Cn denotes the arbitrary constant

A
2 𝑛5 𝜋 5 sin 𝑡 − cos 𝑡 ;4* 6* :
𝜈 𝑥, 𝑡 = ) B B
+ 𝐶4 𝑒 sin 𝑛𝜋𝑥
𝑛𝜋 𝑛 𝜋 +1
4@?

Cn can be found by applying the initial condition 𝜈 𝑥, 0

A
2 −1
𝑥−1= ) B B
+ 𝐶4 sin 𝑛𝜋𝑥
𝑛𝜋 𝑛 𝜋 + 1
4@?

We see that the quantity in the brackets represent the Fourier sine
coefficient bn for x -1. That is
/
2 −1
9 9
+𝐶, = 2 R 1 − 𝑥 sin 𝑛𝜋𝑥 𝑑𝑥
𝑛𝜋 𝑛 𝜋 + 1 !
 /
2 −1
9 9
+𝐶, = 2 R 1 − 𝑥 sin 𝑛𝜋𝑥 𝑑𝑥
𝑛𝜋 𝑛 𝜋 + 1 !

'% %
+ 𝐶, = −
,- , " - " */ ,-

2 2
𝐶, = 9 9
−
𝑛𝜋 𝑛 𝜋 + 1 𝑛𝜋

By substitution
A 5 5 ;4* 6* : ;4* 6* :
2 𝑛 𝜋 sin 𝑡 − cos 𝑡 + 𝑒 𝑒
𝜈 𝑥, 𝑡 = ) - sin 𝑛𝜋𝑥
𝜋 𝑛𝜋 𝑛. 𝜋 . +1 𝑛
4@?
0
% % ', ! - ! $ ', ! - ! $
2 𝑛 𝜋 sin 𝑡 − cos 𝑡 + 𝑒 𝑒
𝑢 𝑥, 𝑡 = 1 − 𝑥 cos 𝑡 + Q − sin 𝑛𝜋𝑥
𝜋 𝑛𝜋 𝑛9 𝜋 9 + 1 𝑛
,./