Laplace Transforms and Fourier Transforms

Laplace Transforms
Laplace Transformation Second Shift Theorem : If F (t) is a function of exponential order
Z
in t , then
f (s) = F (t) est dt L{H(t t0 )F (t t0 )} = est0 f (s)
0

Given a function, its Laplace Transformation is unique

Inverse Laplace Transform : If F (t) has the Laplace transform
The restrictions on the function to have a Laplace transforma- f (s), i.e
tion are | F (x) | M ex , function should be non singular, have L{F (t)} = f (s)
at most a finite number of finite jumps.
then the Inverse Laplace Transform is defined by
Linearity Property
L1 {f (s)} = F (t)
L{aF1 (t) + bF2 (t)} = aL{F1 (t)} + b{F2 (t)}

First Shift Theorem Laplace transforms are unique apart from null functions. Inverse
Laplace functions are also unique apart from null functions.
L{ebt F (t)} = f (s + b)
Initial Value Theorem

lim F (t) = lim sf (s)

Laplace Transformations for a few functions t0 s

1
L{1} =
s
1 Final Value Theorem
L{t} = 2
s
n! lim F (t) = lim sf (s)
t s0
L{tn } =
sn+1
n!
L{tn eat } =
(s a)n+1 Dirac - function
s
L{cos(t)} = 2
s +1 (t) = 0 t, t 6= 0
1 Z
L{sin(t)} = 2 h(t) (t) dt = h(0)
s +1


L{tF (t)} = f (s)
s for any function h(t) continous in (, )
n
L{tn F (t)} = (1)n n f (s) Laplace Transform of function
s
L{F 0 (t)} = sf (s) F (0)
L{(t)} = 1

L{sin(t)} = 2
s + 2
s
L{cos(t)} = 2
s + 2 Filtering Property
s
L{ cos(t)} = 2
s + 2 L{h(t)(t t0 )} = h(t0 )
Z t
f (s)
L{ F (u)du} =
s
0  Z
F (t) Application of Filtering Property
L = f (u)du
t s
L{est f (t)(t a)} = esa f (a)
Z t 
sin(t) 1 1
L = arctan
0 t s s

L{F (n) (t)} = sn f (s) sn1 F (0) sn2 F 0 (0) . . . F (n1) (0) Relationship between (t) and Heaviside function
Z
(u u0 ) du = H(t u0 )
t
Heavisides Unit Step Function
est0 Informally this means that the impulse function is the derivative
L{H(t t0 )} = of the Heaviside Unit Step Function.
s
 Laplace Transforms and Fourier Transforms

Laplace transform of Dirac - function and its derivatives

Z
h(t) (t) dt = h(0) Solving ODEs using Laplace. Typically a first order or a sec-
ond order differential equation can be solved using the following
Z
Laplace transformations
h(t) 0 (t) dt = h0 (0)

Z L{f 0 (t)} = sf (s) f (0)
h(t) 00 (t) dt = h00 (0)

Z
h(t) (n) (t) dt = (1)n h(n) (0) L{f 00 (t)} = s2 f (s) sf (0) f 0 (0)

For solving an ODE, take a Laplace transformation of the ODE,
the equation becomes a simple equation in f (s). Once you solve
Laplace transformation for a periodic function for f (s), you can apply inverse transform to get the particular and
RT complementary solutions to the ODE.
0
est F (t)dt
L{F (t)} =
1 esT Solving the ubiquitous second order differential equation,

where F (t + T ) = F (t),i.e., a function of period T 2x x

a 2
+b + cx = f (t)
t t
The convolution of two given functions f (t) and g(t) is written
as f g and is defined by the integral The standard procedure is to take Laplace transform and convert
Z t in to an equation involving f (s) and then either use convolution
formula or straight forward Laplace inversion to solve for x.
f g = f ( )g(t )d
0

Fourier Transforms
If f (t) and g(t) are two functions of exponential order and writ-
ing L{f } = f (s) and L{g} = g(s) as the two Laplace transforms, Bessels Inequality : If {e1 , e2 , . . . , en , . . .} is an orthonormal
then basis for the linear space V, then for each a V , the series
1
L {f g} = f g
X
| ha, en i |2
r=1
Whenever you want to find a convolution between two func-
converges. In addition, the inequality
tions, find the Laplace transformation in to frequency domain,
multiply the transformations and then take the inverse Laplace X
transformation to get the convolution in the time domain. | ha, en i |2 kak2
r=1
r
1
L{ } =
t s Fourier Series Representation of f (x) that has a period 2

1 X
The Error function erf (x) is defined by f (x) a0 + (an cos(nx) + bn sin(nx)) < x <
2 n=1
Z x
2 2
erf (x) = et dt where
0 Z
1
an = f (x) cos(nx)dx

The complementary error function erf c(x) is defined by and Z

1
Z bn = f (x) sin(nx)dx
2 2
erf c(x) = et dt
x

Fourier Series Representation of f (x) that has a period 2l

Results useful in diffusion context
1 X
k 2 f (x) a0 + (an cos(nx/l) + bn sin(nx/l)) l < x < l
L1 {ek }= s
ek /4t 2 n=1
2 t 3

k s
! where
1 e k Z l
L { } = erf c 1
s
p an = f (x) cos(nx/l)dx
2 (t) l l
 Laplace Transforms and Fourier Transforms

and Fourier Half Range series, i.e fourier series expansion over the
Z l
1 interval (0, ). If f (x) is assumed as an even function over (, 0),
bn = f (x) sin(nx/l)dx
l l then all bn s are 0. If f (x) is assumed as an odd function over
(, 0) , all an s are 0.
Odd functions are represented by sines and even functions are
2
Z
represented by cosines an = f (x) cos(nx)dx
0
Complex Fourier series representation Z
2
X bn = f (x) sin(nx)dx
f (x) cn einx 0
n=

where Fourier half range series are very useful in specifying the bound-
ary condition for the heat equation. Since the boundary condi-
1 1 1
cn = (an ibn ), cn = (an + ibn ), c0 = a0 tion is specified for [0, L], depending on whether the terms in
2 2 2 the boundary condition are even or odd, an appropriate form of
Z Z
1 1 Fourier half range series can be used.
cn = f (x)einx and cn = f (x)einx
2 2
Solving heat equation by using Separation of Variables technique
Properties of Fourier Series where Laplace transformation can be used to solve the second
If f(x) is represented by the following Fourier Series order ODE ,Half range Fourier series can be used to represent the
initial value condition.

1 X
f 0 (x) a0 + (an cos(nx) + bn sin(nx)) < x < Laplace transformation can be used to solve PDEs. The symbol
2 n=1
t, where denotes time in PDEs that ranges from (0, ) corre-
then sponds neatly to the range of the Laplace transform.

X Use of Fourier and Laplace transform and such analytical meth-
f 0 (x) (nan cos(nx) + nbn sin(nx)) < x <
ods have been surpassed by computers that be solve a PDE using
n=1
numerical methods. However analytical method gives the intu-
and ition behind the solution that is not so obvious from the numerical
solution.
Z x
1
f (t)dt = a0 (x + )+
2

X an bn
( sin(nx) (cos(nx) cos(n)))
n=1
n n
<x<
and the function on the right converges uniformly to the function
on the left.

A fourier series expansion of f (x) can be point wise convergent

or uniformly convergent. If it is uniformly convergent, then you
can differentiate term by term. If it is pointwise convergent, then
differentiating both sides of the equation gives nonsense results

If f (t) and g(t) are continous (, ) and provided

Z Z
| f (t) |2 dt < and | g(t) |2 dt <

if an and bn are the Fourier coefficients of f (t) and n , n those

of g(t), then
Z
1 X
f (t)g(t)dt = a0 b0 + (n an + n bn )
2 n=1

Parseval Identity :If f (t) is continous in the range(, ), is

square integrable and has Fourier coefficients an , bn , then
Z
X
[f (t)]2 dt = 2a20 + (a2n + b2n )
n=1