Laplace Transforms

Introduction

Recall in Proof for Euler's Identity using Maclaurin Series, and Proof for Euler's Identity using ODEs. We derived that g(t)=eit=cost+isint. where g maps t (time) to a position on the complex plane. g:RC

Suppose a particle follows the path of g(t). It's velocity is given by ddtg(t)=ig(t). Thus its velocity vector is always perpendicular to its position (Recall that i is a 90 degree rotation). As such, it will move in a circular motion. x=cost and y=sint.

Generalisation

Suppose a particle follows the path of f(t)=est where sC and f:RC. Let s=a+bi.

f(t)=est

Intuition

Consider initial position at f(0)=1. The velocity of the particle is ddtf(t)=sf(t). Thus, the particle at 1+0i will have a velocity s. If a<0, s is pulling the particle inwards, causing it to spiral inwards. If a>0, s is pushing the particle outwards, causing it to spiral outwards. As |b| increases, it's angular frequency ω=dθdt=dbtdt=b increases, causing it to move faster in a circular motion. Thus x=eatcos(bt) , y=eatsin(bt).

e(a+ib)t=eatMagnitudeeibtω=eat[cos(bt)+isin(bt)]

Re(s) determines the rate of change of magnitude.
Im(s) determines the angular frequency of the particle (ω).

Why do we care?

Consider a scenario of a mass on spring placed horizontally.
The force acting on the mass can be modelled as such.

F=kxμv

k : spring constant
μ : damping term (friction, air resistance)
x : displacement from equilibrium.
v : velocity

It describes how the spring and damping will exert a force F on the mass to resist its direction of motion. We can rewrite it as

mx(t)+kx(t)+μx(t)=0

Let's guess that the solution of x(t) is est. Thus

ms2est+kest+μsest=0est(ms2+μs+k)=0ms2+μs+k=0

Solving for s gives

s=μ±μ24(mk)2m

For simplicity, let's assume that μ is 0.

ms2=ks=±km=±ikm=ω

What does this mean? It basically means that the natural frequency of the system is Im(s)=ω=±km. Since Re(s)=0, its amplitude does not change.

BUT WAIT!!! There are two solutions for s, but in fact, there are a whole family of solutions for x(t)!

Linearity

If x1 solves

mx1(t)+kx1(t)=0

and x2 solves

mx2(t)+kx2(t)=0

Then (c1x1+c2x2) solves it as well.

m[c1x1(t)+c2x2(t)]+k[c1x1(t)+c2x2(t)]=0

Thus, there are a whole family of equations x(t) that satisfy the conditions mx(t)+kx(t)=0

x(t)=c1es1t+c2es2t

Thus, the equation is just the linear combination of the complex exponential of s.

In physics, a lot of functions can be expressed as

f(t)=n=1Ncnesnt

Introduction

A Laplace transform is a "function" that maps one function in time-domain to another function in the s domain. where tR,sC

F(s)=0f(t)estdt

And

L{f(t)}=F(s)

Lets consider the simplest case where f(t)=1

0estdt=[1sest]0=1s

Consider now, f(t)=eat

eatL0eatestdt=1sa

The Laplace transform of eat has a pole at a where F(a)=. Thus we have figured out what values of a are embedded in the function f(t).

Consider now f(t)=cos(ωt). cos(ωt)=12eiωt+12eiωt. (Recall Complex Numbers)
Its Laplace transform is

cos(ωt)L0[12eiωt+12eiωt]estdtL{cosωt}=012eiωtest+012eiωtest=12(sωi)+12(s+ωi)=ss2+ω2

The poles in its Laplace transform at s=±iω helps us identify the family of equations that describes cos(ωt). Typically, the expression cos(ωt)=12eiωt+12eiωt is unknown, thus integration by parts is usually used to get the poles of the equation.