MIT differential equation---11,12,13

2nd ode

why are c1y1+c2y2  the solution?(linear combination)

why all are solution?

Q1: principle of superposition

First we introduce D linear operator:

for y'' + py' + qy = 0, we can write it as D^2y + p Dy + qy = 0

We can factorize y:  (D^2 + pD + q)y=0. We are not multiplying y but applying the expression to y

Ly = 0, L = D^2 + pD + q . L is a black box. 

L(u1 + u2) = L(u1)+L(u2)

Well, obviously, you can prove the first question by plugging the solution to the original ODE but it is basically calculation instead of proving. 

You should understand the key underlying is that the operator is linear(what are some non-linear operator?)

Q2: first let's rephrase Q2.

if we are given an ODE with IVP, we can find c1y1+c2y2 which satisfies the initial condition.

Why is that?

Wronski

Theorem

If Wronskian === 0 or Wronskian != 0

c1y1+c2y2=c1u1+c2u2

u1=c1_y1+

normalized solution

yo = 1; y1 = 0;

then you can get directly the solution yoy1 + yo'y2

ex+uniq theorem 

p(x),q(x) continuous

THere's one and only one solution that satisfies, s.t. yo=a;yo' = b

inhomogeneous second ODE

y'' + p(x)y' + q(x)y = f(x). f(x) could be signal, driving term, input

y'' + p(x)y' + q(x)y reduced/associated homogeneous equation

yh is the solution to associated homogeneous equation ; complementary solution

ex:

y'' + m(t)y' + k(x)y = f(t)  spring-mass-daport

this is the forced system where f(t) != 0;

y'' + p(x)y' + q(x)y = 0 

this is the passive system

another model:

Li' + Ri + q/c = voltage(t)

L'' + Ri' + i/c = v' ;  

Ly = f(x)

solution is y = yp + yh

Remind that we talk about the steady-state and transient when solving first order ODE.

In second order ODE, yp is steady-state and yh is steady state

ODE is stable when all characteristic roots have negative real part

Lesson 13

Let's talk about the situation when f(x) = e^alpha x; alpha is a complex number

express ODE in linear operator

(D^2 + AD + B) y = f(x)

p(D) y = f(x)

substitution rule: p(D) e^ax = p(a) e^ax

exponential input theorem 

particular solution = e^ax / p(a) where p(a) != 0

what happen if p(a) == 0

Exponential shift rule:

p(D) e^ax u(x) = e^ax p(D+a)u(x)

Proof: well, let p(D) to be D, and check whether it's true;

we can also do p(D) = D^2. Here comes a simple trick. Don't take the second derivative from scratch, do it from the proof at the first step(induction)

Now let's continue our theorem

when p(a) == 0; yp = x e^ax / p'(a) when a is a simple root

yp = x^2 e^ax /p''(a) when a is a double root (p'(a) ==0)