1elhod

through the following examples.

ExaTnple 9.16 Using the method of
separation of variables, solve the parabolic
equation partial differential
du
= 16 ay
dy
Solution Using u(x, y) = X(*) Y(y), we obtain
X" Y= 16XY' Or X"Y
16X
The variables X, Yare now separated. The
left hand side is a function of x alone and the
side is a function of y alone. Therefore, both the right hand
left and right hand sides must be equal to a
say k or - k. We have the following three constant,
cases.
Case 1. k=0. Then, we obtain

X" = 0 and Y'=0.

Integrating, we obtain X = Ajx+ B and Y= Cj.
Hence, u (x, y) = (Ajx + B) Cj= Ax + B.
where A = A,C and B = BC are arbitrary constants.
Case 2. When the constant is k, we obtain
X" - 16k²X =0, Y'-kY=0.
The solutions of these equations are

X(x) = A,ekx + Be = A, cosh (4kx) + B, sinh (4kx) and Y(y) = Cjek'y,

Hence, u(x, y) = [A, cosh (4kx) + B, sinh (4kx)] Cjek´y
=[A cosh (4kx) + B sinh (4kx)] ek*y
where A = A, C and B = B,C, are arbitrary constants.
 9.32 Engineering Mathenatics

Case 3. When the constant is -k, we obtain

X" + 16k'X =0, Y'+ ky= 0.
The solutions of these equations are

X(«) = A, cos(4 kx) + B, sin (4kx) and Y(y) = Cje-k'y,

Hence, u(x, y) = [A, cos (4kx) + B, sin (4kx)] Cje-k'y
= [Acos (4kx) + Bsin (4kx)]eky
where A and B are arbitrary constants.
In the remaining part of this section, we shall be considering the solution of the heat equation.
wave cquation and Laplace equation by the Fourier method.
9.5.3 Fourier Series Solution of the Heat Equation
One dimensional heat equation
Consider a thin, homogeneous bar or wire of length I. Let the bar coincide with the x-axis on the
interval (0, 71. Let (x, 1) denote the temperature distribution or heat flow within the bar. We shall
Study the flow of heat which is produced under various situations.
Ends of the bar kept at zero temperature
We assume that the initial temperature in the bar is f()and the ends of the bar are maintained at zero
temperature for alltime (Fig. 9.14). The boundary value problem (byp) modelling this temperature
distribution u(x, ) is given by
du
0<x<l,t> 0
with the initial condition: u(x, 0) = f(o), 0 <x<l and
the boundary conditions: u(0, t) = u(l, t) = 0, t > 0
where c is the thermal diffusivity (sometimes, this problem is also called an initial-boundary value
problem).

Fig. 9.14. Flow of heat in abar.

To apply the Fourier method, let

u(x, t) = X(*) T(). (9.52)

Substituting in the differential equation, we obtain

T'
XT'=cx"T or X" =

where prime sign(s) denote differentiation with respect to the corresponding variables. Since the lett
variables).
hand side is a function of x and the right side a function of r(where x and t are independent
2, that is
each side must be equal to a constant
 c2T
Therefore,
X" - X = 0 and T'- cT =0. (9.53)
we obtain
Using the boundary conditions,
u(0, t) = X(0) T () = 0, for all t
and
u(l, t) = X() T() =0, for all 1.
Hence, we obtain X(0) = 0 and X(I)
= 0. (9.54)
Depending on the value of 2, the following three cases arise.
Case 1 = 0. Then, we have X" = 0. The solution is X(r) = Ax + B. Substituting in the conditions
obtain
given in Eq. (9.54), we
X(0) = 0= B and X () = 0 = Al+ B.

The solution is A=0, B =0, that is X() =0and u(x, t) = 0.

be considered.
But, u(x, t) =0is not a solution of the bvp. Therefore, the case =0cannot
Case 2 = k.Then, we have X" - kX= 0. The solution of this equation is
X(*) = Aek + Be-kx

Applying the boundary conditions, we obtain

+ Bek1
X(0) = 0= A + B and X) = Ae
a positive
solution is A= 0, B=0 which gives the trivial solution u(x, t) =0. Hence, is not
Again, the
constant.
equation is
+ k X= 0. The solution of this
Case 3 A=-k. Then, we have X"
(kx).
X(x) =A cos (kx) + B sin

Applying the boundary conditions,

we obtain
X(0) = 0 =A and X(l) = B sin (kl) = 0.

we require B 0. Hence,
non-trivial solutions,
To obtain k= nTI, n= 1,2,....
n=1,2,..., or
(kl)=0= sin (nt),
sin
n=1,2,....
x,(x) = B, sin
We obtain

of the equation T' + c'kT= O is

The solution e-n're12
T() = De-c'i or I,(1) = D,
solution as
integer n, we have the
For each positive
= d, sin
sin
u, (x,t) =B, D,
 9.34 Engineering Mathematics
where dn = B, D,. By the superposition principle, we
have the solution as
u(x, t) = n=1 d, sin 0<x<l.
(9.55)
Applying the initial condition, we obtain

nUX
u(x, 0) =f(x) =2 i 4n sin 0sxsl.
This is a Fourier half-range sine series in [0, ].
Hence, d,, are the Fourier coefficients given by
(see Eq. (9.26)

d, = fw) sind. (9.56)

Therefore, the solution of the bvp is given by Eq. (9.55) where
the coefficients d, are given by
Eq. (9.56).

Temperature distribution in a thin bar or wire with a radiating end

We assume thc following
(i) left end of the bar, x= 0, is kept at zero
temperature.
(ii) right end of the bar x=lis poorly insulated and radiates
energy into the surrounding medium.
This end radiates at arate proportional to the temperature at that
end of the bar.
(iii) initial temperature of the rod is f (x), 0sxsl.
The temperature distribution u(x, t) in the bar can be modelled by the
following initial bvp

dt
=
O,0<x<l, t>0
ou
with the boundary conditions: u(0, t) = 0, (l, t) = - au(ll, t), t > 0, a positive constant
and the initialcondition: u(x, 0) =fx), 0 <sl.
Let u(x, t) = X%) T() be the solution. Substituting in the differential equation, we
obtain
XT'= c'x"T or =A.
Therefore, we obtain the equations
X"-AX= 0 and T'-dcT= 0. (9.57)
Using the boundary conditions, we get
u{0, t) = 0= X(0) T() for all , or X(0) = 0
du
(1,)= - au(l, t) gives X(/) T() + aX() T() = 0, for all 1.
Hence, X(U) + aX(l)= 0.
 consider
the following three cases.
We
0. From Eq. (9.57), we get X" = 0, whose solution is X(«) = Ax + B. Using the
Case 1 d=0
conditions, we obtain
boundary
X(0) = 0= B, X'(U) + aX(/) =A+ aAl= A(1 + al) = 0.
we have A = 0, B = 0, that is, X = 0and u(x, ) =0. Therefore, À= Ois not a valid value.
Hence,

Let a = 2 From Eq. (9.57), we get X" - kX = 0 whose solution is

Case 2
X(x) = Acosh (kx) + Bsinh (kx).
second boundary condition gives
Ttoingthe boundary condition X(0) =0, we get A= 0. The
X'U) + CX(U) =0= B[k cosh (kl) + a sinh (kl)] =0.
solution.
we obtain B = 0. Again, we have a trivial
Since kl > 0. cosh (kl)> 0, sinh (kl) > 0,
Hence, =k is not possible.
get X* + kX = 0, whose solution is
= - k. From Eq. (9.57), we
Case 3 Let
X(x) =A cos (k) + B sin (kx).
second condition gives
Now, X(0) = 0 gives A = 0. The
(kl)] = 0.
X(U) + aX(l) = B[k cos (kl) + a sin
satisfies the equation
solutions are obtained when k
Therefore, non-trivial (9.58)
a sin (kl) =0 Or tan (kl) =-(k/).
k cos (kl) +
number of
equation can be determined. There are infinite
roots of this
Since l and a are given, the as
equation which can be written
roots for this
kl.
tan v =- ul(oal), u =
that is
Let the roots of this equation be Vj, U2, U3, ·

k,l= v, or k, = n= 1,2,....

1,2,...
Therefore ay = -k = - , n =

and X, (x) =B, sin n= 1,2,... .

kcT= 0, whose solution is
The equation for T (t) is T' +
d,e-vic'u'

T() = de-k1 or T, () =

The solution of the heat equation is therefore

e-uem² ,n= l, 2,...

u, (x, t) = X, (x) T, (t) = D, sin (

 where D, = B, d,:
initial bvp as
Using the superposition principle, we obtain the solution of the given

u(x,t) =ED, sin

n=l
e (9.59)
Applying the initial condition, we obtain

u(x, 0) =f(x) =n=l D, sin (9.60)

This series is not the standard Fourier half-range series, because u,'s are solutions of atranscendental
equation. We shall use the Sturm-Liouville theorem (Theorem 7.1) for the bvp
X"+ k'X = 0, X(0) = 0, X'(@) + a X() = 0.
Using this theorem,we have that if y,() = sin (vmx/) and y,(x) = sin (U, x/l) are the eigenfunctions
corresponding to distinct eigenvalues vm, U, then y), y,(x) are orthogonal to each other, on the
interval [0, I] with respect to the weight function p(r) = 1, that is

sin sin -dx =0, m # n. (9.61)

Therefore, multiplying both sides of Eq. (9.60) by sin (vm xll) and integrating on the interval [0, IJ,
we obtain

f(x) sin Jdx = D, Sin Sin dx

=Dm Sin

Hence, the coefficients D,, in the series solution (Eq. (9.59)) are given by

D, = Sin sin 2 (9.62)

Therefore, the solution of the initial bvp is given by Eq. (9.59) where the coefficientsD, are given by
Eq. (9.62).The roots of the equation tan v=- (v/al) can be obtained by a numerical method.
 which is fastened at its ends x=0and x =
atExample pointAn
its centre9.17 x=/2 tostring
elastic a height of l/2 and released from rest. Find the displacementlisof the stringup
of length
picked
at any instant of time.
The initial-boundary value problem modelling the motion of the string is given by
Solution
Eqs. (9.73) to (9.75) where
0sxs l/2,
g )= 0 and f(*) =
U2 <x sI.
The initial position of the string is given in Fig. 9. 15. The soltion is given by Eq. (9.82) as
nNCt
u(x, t)= n=l
2 D, COS sin

where D, = (x) sin (see Eq. 9.80)

f()4

/2

Fig. 9.15. Function f (r) in Example 9.17.

Substituting for f (x), we obtain

elU2

xsn"Ját
X Sin
(l- x) sin nUX
dx
 W2
sin (nTx/l)
I COs ( n N x I l )
+
+-0-) cos(nz/l)
(n Ttxll) sin (nTxll)

- (nrll)

(/2) cos (nn/2)

(nn/l)?

sin (nt/2)
0

+
(//2) cos (nn/2) +
sin (nt/2)
(nzlly?

(nn/) (nrll)? (nTll) (nnll)2

41 sin
2
and for n odd we have
for n even
Hence,
D, = 0
4l(-1)k+!
D, = D2k-1 = (2k - 1)²n2 k= 1,2,3,....

Therefore,

41 (2k - 1) TCt T(2k - 1) Tx

COS Sin
u(x, t) =T2 kal (2k - 1)
released
of length which is fastened at its ends x = 0 and x = l is
t Exazole 9.18 Anelastic string velocity gr) given as
horizontal position (zero initial displacement) with initial
from its
|x, 0<xs l/3,
&(x) = U3 <x <l.
|0,
at any instant of time.
Find the displacemnent of the string
modelling the motion of the string
is given by Eqs. (9.73) to
Solution The boundary value problem
obtain D,= 0and the solution
is given
given. Since f(x) = 0, we
(9./5) wheref(r) =0and g(x)
is as
by
nTX
Sin
E, sin
u(x, t) = 2 n=l

where
nTX (see Eq. 9.81)
2 g(x) sin dx.
E, = n NC

Subsituting for g(x), we obtain t cos (nTxll)

sin (nT/l) "
ell3 2 (nz/1)?

E,=2 nTtC
xsin dx =nc (nTl)

(/3)cos (nn/3)
2 sin (nt/3)
(nnll)
nTTC (nnll)2
Therefore, the solution is given by
 212
()-)n(),
1 nNct
Sin COS sin Sin
u(x,t) = Nn3 3 3n2
Vibrations in an infinitely long elastic string (D'Alembert solution)
The vertical displacement u(x, t) of an infinitely long elastic string is governed by the initial value
problem
2
= o <X < oo, 1>0 (9.87)

du
u(x, 0) =f(x). (x, 0) = g(x). (9.88)

This problem can be solved directly by change of variables. Consider the transformation of variables
as

=X + ct and n=x- ct. (9.89)

Then, by chain rule, we have

du du du
+

du
+ + + +2 u
dx? dn x Ogon n?

du du dn du du
=C

+ C
du = c? 2 u
dn dn or
 9.44 Engineering Mathemnatics
Substituting in Eq. (9.87), we obtain
= 0.
(9.90)
Integrating with respect to 1, we obtain
du
= h(5) (9.91)
where h()is an arbitrary function of . Integrating Eq. (9.91) with respect to 5, we obtain

u(a. n)= E) d~ +v() (9.92)

where y () is an arbitrary function of K. Denoting h() d= o5), we obtain

u(5, n) = ) + y(n) or u (x, t) = o(x + ct) + v(x- ct). (9.93)
Substituting this solution in the initial conditions, we obtain
u(x, 0) = (x) + y(x) =f(x)
du
(x, 0) = c [o'(x) - y'(x)] =g (x).
Solving the two equations

0'(x) + y'(r) =f'(r) and o'(x) - u'(x) =g(x) C

we obtain

o' (x) =f (*)+8(x) and w'(r)= f ) &(x).

Integrating, we obtain

o(x)=fr) + g(s)ds + k

and 8(s) ds -k (9.94)

\since o (x) + y (r)=f()], where xo is arbitrary and kis the constant of integration. Substituting these
solutions in Eq. (9.93), we obtain
Xtct X-ct

u(x,)= r+ )+f(r -c) + 2c g(s)ds - g(s) ds

=f(* +ct) +f(x- ct)] + 8 (s) ds. (9.95)

 of the Ourier Series,
This form solution
ththe of the Fourier Integrals and Fourier Transforms 9.45
9.88)) is called D' wave
Alembert solution.equation
If the (initial alue problem defined by Eqs. (9.87) and
u(x,t)
initial
velocity g(*) = 0, then we obtain
solution can be =fo + ct) +f (x-
This
interpreted as a ct)\ (9.96)
isf(r - c)/2) and the other moving superposition of two
f (x + travelling waves, one moving to right (that
speed c and have the same to left (that is
initial ct)/2).
Exanple 9.19 Use displacement
D'Alembert solution to find fthe(x).
Both the waves travel with the same

vibrations of, an
infinitely long elastic solution of the initial value problem defining the
8(x)= a, where a is a Constant. string (Eqs. (9.87) and (9.88)) when f() = sin x and
Solution We have from the
D'Alembert solution (Eq.
(9.95))
u(x, 1) = ; [sin (x + ct) + Pxtct
sin (x- ct)] + àds
X-ct
= Sin X COS
(ct) + at.
 9,56 Fourier Series Solution of Laplace's Equation
We want to study the steady-state temperature distribution in a thin, flat, rectangular plate.
any loss of generality let the boundaries of the plate be x = 0, x = a, y=0 and y =b. The Without
equation modelling the steady-state temperature distribution is given by the Laplace cauatinon differenial

dx2
+ = 0, 0<<a, 0<y<b. (9.4
Oy2
Various types of boundary conditions can be imposed to solve this equation.
 example,
let the boundary conditions be
For
u(0, y) =0, u(a, y) = 0, 0< y< b
u(x, 0) =f(), u(x, b) g(*), 0 <x< a.
and

be written as u(x, y) = X() Y(y). equation, we get

Let the
s o l u t i o n

Substituting in the di fferential

X"Y+ XY" = 0 or X"Y1=-k?
X
Hence, we have
XX
and y'/Y are functions of the independent variables x and y respectively.
since
t h e e q u a t i o n s

X" + k2 X= 0 and y"- k2Y= 0

whose solutions are

X(%) = Acos (kx) + B sin (kx) and Y(v) = C cosh (ky) + Dsinh (ky).
boundary conditions,
we obtain the
values of, A do not give a solution. Using the given
Other
conditions X(0) = 0 and X(a) = 0. Therefore, we obtain
X(0) = 0=A and X(a) =0= sin (ka) = sin (nt), n = 1, 2, .

LHence. k= nnla, n =1, 2,. .. The solution

is given by
nTXN

D, sinh (kn y)] sin

u, (x,y) = X, (x) Y, (y) = [C, cosh (k,y) +
we have the solution as
Using the superposition principle,
nTX (9.115)
u(x, y)= [C, cosh (kny) + D, sinh (kn y)] sin a

u(, 0) =f (x), we get

Applying the boundary condition
nUX
sin
u(x, 0) =f(r) = n=l C,
series. Therefore,
which is aFourier half-range sine
nTX
dx.
(9.116)
Ch = f(x) sin

u(x, b) = g(), we get

Pplying the boundary condition
nTX

(k,b) + D, sinh (k, b)]

sin
=2 [C, cosh
u (x, b)=g (x) n=l
Therefore
which is again a Fourier half-range
sine series.
nNX
dx
sinh (k,b) = &(x) sin
C cosh (kyb) + D, 0
 1 nNX
or D, =
sinh (k,b) & (x) sin dx - C, cosh (k, b)
Therefore, the steady-state temperature distribution in the rectangular plate is given by
nTty nTy nTX
u(x, y) = n=l
2 C, cosh +D, sinh Sin

where C, is given in Eq. (9.116) and

1 ntb
Dn = sinh (nnbla) s(x) sin dx - C, cosh
co
0 a

Exercise 9.4
Classify the following partial differential euations