Stability of critical points of a system given by a constant matrix

69 Views Asked by At

We are given the homogeneous system $$y'=\begin{bmatrix}5 & -3 & 2 \\ 15 & -9 & 6 \\ 10 & -6 & 4\end{bmatrix}y$$ with the initial value $y(0)=(1,1,0)$ and are supposed to check whether the critical points are (asymptotically) stable.

I tried to directly apply the formal definitions of stability but failed to arrive at something useful.

Is this a use case for Lyapunov? If so how do I find an explicit Lyapunov function? Lyapunov has only just been introduced and I can't find any examples for a case as simple as this. Most use some physical intuition about energy...

1

There are 1 best solutions below

2
On BEST ANSWER

It is worth practicing Jordan normal form, especially when they give you something where all numbers are integers, with the possible exception of a denominator needed for the inverse matrix. That is, I am constructing $P^{-1}A P = J,$ where your coefficient matrix is called $A.$ The minimal polynomial is $x^2,$ so there will be a 2 by2 block and a singleton. We start with the right hand column, calling it $w,$ where the only requirement is that $w$ not itself be an eigenvector. I like $$ w = \left( \begin{array}{c} 0 \\ 0 \\ 1 \end{array} \right) $$ For the middle column, we are required to take $v = A w,$ so $$ v = \left( \begin{array}{c} 2 \\ 6 \\ 4 \end{array} \right) $$

For the left hand column, any eigenvector not a multiple of $v$ is acceptable, and I like $$ u = \left( \begin{array}{c} 0 \\ 2 \\ 3 \end{array} \right) $$ for $$ P = \left( \begin{array}{ccc} 0 & 2&0 \\ 2 & 6 & 0 \\ 3 & 4 & 1 \end{array} \right) $$ Next $$ P^{-1} = \frac{1}{2} \left( \begin{array}{ccc} -3 & 1&0 \\ 1 & 0 & 0 \\ 5 & -3 & 2 \end{array} \right) $$ and we have $P^{-1}A P = J,$ with $$ J = \left( \begin{array}{ccc} 0 & 0&0 \\ 0 & 0 & 1 \\ 0 & 0 & 0 \end{array} \right) $$ The reason for finding actual matrices $P$ and $P^{-1}$ is simply that you get explicit $$ P J P^{-1} = A $$ It is from that expression that we can find explicit $e^A$ and $e^At$