8.

Some Additional Examples

In addition to the Fourier transform and eigenfunction expansions, it is sometimes
convenient to have the use of the Laplace transform for solving certain problems in partial
differential equations. We will quickly develop a few properties of the Laplace transform and
use them in solving some example problems.

Laplace Transform
Definition of the Transform
Starting with a given function of t, ft, we can define a new function !fs of the variable s.
This new function will have several properties which will turn out to be convenient for
purposes of solving linear constant coefficient ODEs and PDEs. The definition of !fs is as
follows:

Definition Let ft be defined for t 0 and let the Laplace transform of ft be defined by,

K
Lft = X e ?st ft dt = !fs
0

For example:

K ?st
ft = 1, -t 0, L1 = X e ?st dt = e?s | t=K 1 !
t=0 = s = fs for s > 0
0

K ?b?st
ft = e bt , -t 0, Le bt = X e ?b?st dt = e | t=K = 1 = !fs, for s > b.
0 s ? b t=0 s?b

The Laplace transform is defined for all functions of exponential type. That is, any function
ft which is

(a) piecewise continuous = has at most finitely many finite jump discontinuities on any
interval of finite length
(b) has exponential growth: for some positive constants M and k

|ft| M e kt for all t 0, .

Properties of the Laplace Transform

The Laplace transform has the following general properties:

1. Linearity LC 1 ft + C 2 gt = C 1 !fs + C 2 s

2. Homogeneity Lfat = 1a !f as for a>0

3. Transform of the Derivative Lf v t = s !fs ? f0

Lf vv t = s 2 !fs ? sf0 ? f v 0 etc

1
4. Derivative of the Transform Lt ft = ? !f v s
Lt 2 ft = ?1 2 !f vv s etc

Some Special Transforms

There are some transform pairs that are useful in solving problems involving the heat
equation. The derivations are given in an appendix.

S.1 ft = k e ?k 2 /4t , t > 0 Lft = e ?k s , k>0

4^t 3
S.2 ft = 1 e ?k 2 /4t , t > 0 Lft = 1 e ?k s , k0
^t s
S.3 ft = erf c k , t>0 Lft = 1s e ?k s , k>0
2 t

Here erf cz = 1 ? erfz and erfz = 2 X z e ?x 2 dx.

^ 0

Additional Properties of the Transform

Let ft be a function of exponential type and suppose that for some b > 0,

0 if 0 < t < b
ht =
ft ? b if t>b

Then ht is just the function ft, delayed by the amount b. Then

K K
Lht = X ht e ?st dt = X ft ? b e ?st dt
0 b

Let z = t ? b so that

K K
Lht = X fz e ?sz+b dz = e ?bs X fz e ?sz dz = e ?bs !fs.
0 0

0 if 0 < t < b
If we define Ht ? b =
1 if t > b
then
ht = Ht ? b ft ? b

and we find

5. Transform of a Delay LHt ? b ft ? b = e ?bs !fs, for b > 0.

2
A related results is the following

K K
Le bt ft = X e bt ft e ?st dt = X ft e ?s?bt dt = !fs ? b.
0 0
i.e.,

6. Delay of a Transform Le bt ft = !fs ? b

Results 5 and 6 assert that a delay in the function induces an exponential multiplier in the
transform and, conversely, a delay in the transform is associated with an exponential
multiplier for the function.

A final property of the Laplace transform asserts that

7. Inverse of a Product Lf D gt = !fs s

t
where f D gt := X ft ? b gb db
0

The product, f D gt , is called the convolution product of f and g. Life would be simpler
if the inverse Laplace transform of !fs s was the pointwise product ft gt, but it isnt, it
is the convolution product. The convolution product has some of the same properties as the
pointwise product, namely

f D gt = g D ft and h D f D gt = h D f D gt.

We will not give the proof of the result 7 but will make use of it nevertheless.

Applications to PDEs
1. Consider the IBVP / t ux, t = k / xx ux, t x > 0, t > 0,
ux, 0 = 0, x > 0,
u0, t = ft, f0 = 0, t > 0.
Let
x, s = Lux, t Laplace transform in t.
Then
2
sx, s ? 0 = k U x, s, x > 0, U x, s = d 2 x, s
dx
0, s = !fs.

Solving this ODE in x, we find

x, s = Ae ?x s/k
+ Be x s/k
x > 0.

We want x, s to remain bounded for all positive x, which requires that B = 0. Then the
boundary condition at x=0 leads to

3
 x, s = !fs e ?x s/k
, x > 0.

Then S.1 together with property 7 of the Laplace transform, gives

ux, t = f D Kx, 6 = X
t x 2
e ?x /4kt?b fb db.
0
4^kt ? b 3

as the unique solution of the IBVP. Suppose now that we wish to compute the flux through
x=0,
Flux at 0 = ?k/ x u0, t.

Differentiating the integral expression for u does not seem like a pleasant prospect.
However, note that

L?k/ x ux, t = ?k/ x x, s = ks !fs e ?x s/k

,
and
Flux at 0 = ?k/ x u0, t = L ?1 ks !fs .

In order to use our inversion formulas, we write

k !
ks !fs = s fs
s

and, recalling that f0 = 0, we have s !fs = Lf v t. In addition, S.2 with k=0, gives

L ?1 1 = 1 ,
s ^t
and so,
k ! k t f v b
Flux at 0 = L ?1 s fs = X db.
s ^ 0 t ? b

Note that if ft = At, then

k t A
Flux at 0 = X db = 2A kt
^ .
^ 0 t ? b

Problem 1 Show that if vx, t solves

/ t vx, t = k / xx vx, t x > 0, t > 0,

vx, 0 = 0, x > 0,
?k/ x v0, t = gt, t > 0,
then

4
 t gb
v0, t = X db, t > 0,
0
^kt ? b

and if gt = B, then v0, t = 2B t .

^k

2. Consider the IBVP / tt ux, t = a 2 / xx ux, t x > 0, t > 0,

ux, 0 = / t ux, 0 = 0, x > 0,
u0, t = ft, t > 0.
Let
x, s = Lux, t Laplace transform in t.
Then
s 2 x, s ? 0 = k U x, s, x > 0,
0, s = !fs.

Solving this ODE in x, we find a general solution of the form,

x x
x, s = Ae a s + Be ? a s ,
and both
x x
x, s = !fs e a s x, s = !fs e ? a s

solve the transformed equation and the boundary condition at x=0. To see how to choose
x
the correct solution, recall that for x > 0, property 5 implies that !fs e ? a s is the transform of
x
ft delayed by the amount ax > 0. On the other hand, !fs e a s is the transform of ft
advanced in time by the amount ax > 0. Another way to say this is to say

x
ux, t = L ?1 !fs e ? a s = f t ? ax H t ? ax ,

represents the wave form f6 propagating from L to R into the region x > 0 while

x
ux, t = L ?1 !fs e a s = f t + ax H t + ax

represents the wave form f6 propagating from R to L out of the region x > 0. Then the
solution that is relevant for our problem is the wave that travels from L to R into the region
x > 0.

3. Consider the IBVP

/ t ux, t = K / xx ux, t 0 < x < 1, t > 0,
ux, 0 = 0, 0 < x < 1,
u0, t = ft / x u1, t = 0 t > 0.

5
Here, we may use the Laplace transform, or if we prefer, we can use eigenfunction
expansion after a suitable modification of the problem. We will solve the problem first by
this means. Since the boundary conditions are not homogeneous, the method does not
apply directly but if we let
vx, t = ux, t ? ft
then
/ t vx, t = / t ux, t ? f v t
= K / xx ux, t ? f v t
= K / xx vx, t ? f v t,
vx, 0 = ux, 0 ? f0 = 0,
v0, t = u0, t ? ft = 0, / x v1, t = / x u1, t = 0.

Now the method of eigenfunction expansion applies directly to the problem for vx, t since
the boundary conditions are homogeneous. Since v0, t = / x v1, t = 0, it follows that the
eigenfunctions are the eigenfunctions of example 6.3, namely

1 2 1
Vn = n? 2
^ 2 = W 2n , d n x = sin n ? 2
^x = sin W n x, n = 1, 2, ...

K K
We write vx, t = > v n t d n x and f v t = f v t 1 = f v t > C n d n x,
n=1 n=1
where
1
1, d n X sin n ? 1
^xdx
4 = W2n .
2
Cn = = 10 =
d n , d n X sin 2 n ? 1
^xdx 2n ? 1^
0 2

Substituting these expansions into the IBVP for vx, t, we conclude that

v vn t + KW 2n v n t = ?f v t C n , v n 0 = 0, -n.

t
v n t = ?C n X e ?KW n t?s f v s ds.
2
and
0

For example, if ft = mt, then

v n t = ?mC n X e ?KW n t?s ds. = mC2n e ?KW n t ? 1

t 2 2

0 KW n
and
K
ux, t = ft + > mC2n e ?KW n t ? 1 d n x,
2

n=1 KW n
K
= mt + 2m > 13 e ?KW n t ? 1 d n x.
2

K n=1 W n

Alternatively, we may use the Laplace transform to solve this same problem. Let x, s
denote the Laplace transform of ux, t. Then,

6
 s x, s ? 0 = Kx, s, 0, s = !fs, v 1, s = 0.

The general solution of the equation may be written as

x, s = A exp ?x s/K + B exp x s/K ,

and then the boundary conditions lead to

exp s/K !fs, exp ? s/K !fs.

A= B=
exp ? s/K + exp s/K exp ? s/K + exp s/K
Then
exp ?x ? 1 s/K !fs + exp ?1 ? x s/K !fs,
x, s =
exp ? s/K + exp s/K exp ? s/K + exp s/K

exp ?x s/K !fs + exp ?2 ? x s/K !fs,

=
exp ?2 s/K +1 exp ?2 s/K + 1

and since the formula for the sum of a geometric series implies that,

K
1 = >?1 n exp ?2n s/K ,
exp ?2 s/K +1 n=0

this becomes
K K
x, s = !fs >?1 n exp ?x + 2n s/K + !fs >?1 n exp 2n + 2 ? x s/K .
n=0 n=0

It follows from (S.1) that

2
L ?1 exp ? x s = x exp ? x := Gx, Kt, t > 0,
K 4^Kt 3 4Kt

Then, using property 7, we find

K K
t t
ux, t = >?1 n X Gx + 2n, Kt ? b fb db + >?1 n X G2n + 2 ? x, Kt ? b fb db
0 0
n=0 n=0

t t t
= X Gx, Kt ? b fb db + X G2 ? x, Kt ? b fb db ? X Gx + 2, Kt ? b fb db
0 0 0

Notice that this representation for the solution looks nothing at all like the eigenfunction

7
expansion obtained previously. However, this problem has a unique solution so the two
representations must produce identical results. What can, in fact, be seen is that since each
representation involves an infinite series of which only a finite number of terms can actually
be computed, each representation produces only an approximate solution. Moreover, it can
be shown that the Laplace transform approximation is most accurate at small values of t
while the eigenfunction expansion is more accurate as t grows large.

Appendix Some Laplace Transform Formulas

We will derive some Laplace transform formulas which are useful in solving problems
involving the heat equation.

1 K
Lemma 1 If ft = L ?1 !fs , then L ?1 !f s = X 0 z e ?z 2 /4t fz dz.
4^t 3
Proof- if the lemma holds, then

!f s = X K e ?st 1 K
X 0 z e ?z 2 /4t fz dz dt
0
4^t 3

=X
K 1 fz X K z e ?z 2 /4t e ?st dt dz.
0 ^ 0
4t 3

Let b= z then db = ? 12 z dt
4t 4t 3
2
t = z 2 and as t goes from 0 to K, b goes from K to 0.
4b
Then
!f s = X K 1 fz X K 2e ?z 2 s/4b 2 e ?b 2 db dz
0 ^ 0

2 X K fz ^ ?z K
= e s
dz = X fze ?z s dz.
^ 0 2 0

Here we used the following result,

K ^ ?z s
X 0 e ?z 2 s/4b 2 e ?b 2 db = e .n
2

Applications of the Lemma

We have that
for ft = Nt ? a !fs = e ?as .

Then, by lemma 1,

8
 1 K
L ?1 !fs = L ?1 e ?a s
= X 0 z e ?z 2 /4t Nz ? a dz
4^t 3

L ?1 e ?a s
= 1 a e ?a 2 /4t for a 0. (A.1)
4^t 3

Now, if we integrate both sides of (A.1) with respect to the parameter a from b to K,

K 1 K
L ?1 X e ?a s da = X b a e ?a 2 /4t da,
b
4^t 3

we get

L ?1 1 e ?b s
= 1 e ?b 2 /4t , b 0. (A.2)
s ^t
In particular, for b = 0,

L ?1 1 = 1 .
s ^t

Integrate both sides of (A.2) with respect to the parameter b from a to K,

K 1 e ?b s db 1 X K e ?b 2 /4t db.
L ?1 X =
a s ^t a

Let Y= b dY = db
4t 4t

K 1 e ?b s db 2 XK
L ?1 X = L ?1 1s e ?a
2
so that s
= a e ?Y dY.
a s ^
4t

If we define erfcx = 2 X K e ?Y 2 dY,

^ x
then

L ?1 1s e ?a s
= erfc a (A.3)
4t

Proceeding, we integrate both sides of (A.3) with respect to the parameter a from b to K,

K 1 e ?a s da K a
L ?1 X s = X erfc da.
b b 4t

9
We let Y= a dY = da
4t 4t
and then,

K 1 e ?a s da 1 e ?b K
L ?1 X s = L ?1 3/2
s
= 4t X b erfcY dY.
b s
4t
That is,

L ?1 1 e ?b s
= 4t ierfc b (A.4)
s 3/2 4t
Here
K
ierfcx = X erfcY dY
x

denotes the so called, iterated complementary error function. It can be shown that

ierfcx = 1 e ?x 2 ? x erfcx.
^

so the iterated complementary error function can be expressed in terms of other functions.

Finally, there are two Laplace transform pairs that are obtainable by elementary means:

L ?1 1 = 1c 1 ? e ?ct
ss + c
and
L ?1 e ?as = 1c 1 ? e ?ct?a Ht ? a := F a t.
ss + c

Then lemma 1 implies

L ?1 e ?a s = 1 K
X 0 z e ?z 2 /4t F a z dz.
s s + c 4^t 3

1 K
= X a z e ?z 2 /4t 1c 1 ? e ?cz?a dz.
4^t 3

Then integration by parts leads to

K K
X a z e ?z 2 /4t 1c 1 ? e ?cz?a dz = 2te ac X a e ?cz e ?z 2 /4t dz

so
L ?1 e ?a s ac K
= e X e ?cz e ?z /4t dz,
2

s s + c ^t a

10
 ac K
= e X e ?z +4zct+4c t /4t e ?c t dz,
2 2 2 2

^t a

ac+ct 2
K
= e X a e ?z+2ct 2 /4t dz,
^t

= 2 e ac+ct 2 X K 2
e ?Y dY
a
^ 4t
+c t

i.e.,
L ?1 e ?a s 2
= e ac+ct erfc a +c t .
s s + c 4t

11