solving ODE system X' = AX
What if A has duplicate eigenvalues?
EX: Suppose there are three fish tanks where each of one is in heat contact with another two. How to model this problem?
Xi' = a sigma(Xj - Xi)
a [-2 1 1; 1 -2 1; 1 1 -2] x = 0
Solve for eigenvalue. lambda1= 0, lambda2 = -3, lambda3 = -3.
For lambda1, [-2 1 1; 1 -2 1; 1 1 -2] x = 0, solve for x.
We cannot use inverse matrix because the matrix is singular. But physically, we can observe the solution. When lambda = 0, e^-lambda x = 1. It's a constant solution.
It's the situation when three cells have the same temperature. So a1 = a2 = a3 = 1.
For lambda2,3, A = [1 1 1; 1 1 1; 1 1 1].
Here, we have to write out two eigenvectors which are independent.
a1 = [ 1 0 -1]; a2 = [1 - 1 0]
Therefore the general solution will be c1[1 0 -1]' e^3t + c2[1 -1 0]' e^3t + c3[1 1 1]' e^3t
For repeated eigenvalue, if we can find enough independent eigenvectors, we can it complete eigenvalue. Otherwise, we call it defective.
For defective eigenvalue, we have to introduce complex eigenvalue. It forms solution in the form of e^(a+bi)t
Then, we separate real and ima part differently.
x' = x + 2y ;
y' = -x - y.
A = [1 2; -1 -1]. r^2 +1 = 0 r = +- i;
LESSON 27
Sketch the solution
ex:
x' = -x + 2y
y' = - 3y
General solution:
x = c1 [1; -1] e^-3t + c2 [1;0] e^-t;
let c1 be +-1 and c2 be +-1 accordingly
Notice for this general solution, as t->-INF, c1 becomes dominant while as t->INF, c2 becomes dominant
How the graph will look like? Curves will tend to parallel to c1 but as they come close to the origin, they will converge to zero.
The pattern is called NODAL SINK. It's asymptotically stable.
Other patterns can be asymptotically unstable, saddle, and spiral sink(complex eigenvalue).
Eigenvector specifies the direction and the eigenvalue specifies approaching or leaving the origin.
LESSON28
solution to inhomogeneous system
Theorem A:
The general solution to a homogeneous system is c1x1+c2x2
Theorem B:
Wronskian of two solutions W(x1,x2) := | x1 x2|
Either the Wronskian is 0 if x1, x2 are linearly dependent, or W(x1,x2) != 0 if x1, x2 are linearly independent.
Property of Wronskian X_ := [X1 X2] for x1, x2 linearly independent
1.det(X_) != 0
2.X_ = A X_(the differentiation of a matrix is simply the differentiation of each entry)
[X1' X2'] = [AX1 AX2]
Inhomogeneous system:
x' = ax + by + r1(t)
y' = cx + dy + r2(t)
X' = AX + R(t)
Theorem C: General Solution = Xc + Xp
EX: Say there are two tanks of salt water connecting to each other. Meanwhile there are inflows to two tanks and an outflow to one of them
X' = [-3 2 ;3 -4] X = [5e^-t ;0]
This example shows you modelling by inhomogeneous system.
The problem of solution to inhomogenous system is how to get Xp.
Xp = v1(t) x1 + v2(t) x2
Xp' = (X_ V)'
= X_' V + X_ V'
= AX_ V + R
We can get X_V' = R
Because X_ is nonsingular , V' = X_^-1 R
V= $X_^-1 R dt.
Xp = X_ $X_^-1 R dt