I have a theorem:

For a linear homogeneous system: $$\dfrac{dx}{dt}=A(t)x \tag{I}$$

Where $A(t)=(a_{ij}(t))_{n \times n} \in C(\mathbb{R}^+,\mathbb{R}^{n \times n})$

Suppose that $X(t)$ be the fundamental matrix solution of the following reference system $(I)$.

Let $K(t,s)=X(t)X^{-1}(s)$ be the Cauchy matrix of the following reference system $(I)$.

Prove that:

  • a/ System $(I)$ is stable iff $X(t)$ is bounded, it means $\exists M>0$ such that $$\|X(t)\| \le M, \forall t \ge 0 $$

  • b/ System $(I)$ is asymptotically stable iff $$\lim_{t \to +\infty}X(t)=0$$


I have stuck when I try to show this theorem.

I have tried using the definition of stable, asymptotically stable. But I have no solution.


Ps: (Or) if somebody knows/reads this theorem (book/pdf/djvu...) then you can post it.

Can anyone help me!

Any help will be appreciated! Thanks/

  • $\begingroup$ Maybe a hint $X(t)$ is formed by linearly independent solutions. Then just follow the definitions. I can/will elaborate an answer in case you still need it. $\endgroup$ – PepeToro Sep 25 '13 at 17:08
  • $\begingroup$ Of course, user58533. I need it. If you have an ANS, it's great. You're welcome! $\endgroup$ – kimtahe6 Sep 25 '13 at 17:18

Here's my solution for Question a:

a/ A linear homogeneous system $\dot{x}=A(t)x$ is stable iff $X(t)$ is bounded.

Without loss of generality we can assume that $X(t_0)=E$, then $x(t)=X(t)x_0$.

  • If $X(t)$ is bounded. Whence $$\left \| x(t) \right \|=\left \| X(t)x_0 \right \| \le \left \| X(t) \right \| \left \| x_0 \right \| \le M\left \| x_0 \right \|<\epsilon, \text{for}\left \| x_0 \right \|<\delta:=\frac{\epsilon}{M}$$ Hence, $x \equiv 0$ is stable.

  • If $x \equiv 0$ is stable. Whence

$$\forall \varepsilon >0, \exists \delta >0 \ \text{such that}\ \left \| x_0 \right \| \le \delta \implies \left \| x(t) \right \| = \left \| X(t)x_0 \right \| < \epsilon, \forall t \ge t_0$$ Thus $$\left \| X(t)\alpha \right \|=\left \| X(t)\alpha \cdot \dfrac{\delta}{\left \| \alpha \right \|}\right \|\cdot \dfrac{\left \| \alpha \right \|}{\delta}<\dfrac{\epsilon}{\delta}\left \| \alpha \right \|, \forall \alpha \in \mathbb{R}^n$$ Hence, $\left \| X(t) \right \| \le M:=\dfrac{\epsilon}{\delta} \blacksquare $.


My solution for Question b:

  • If $\lim_{t \to \infty}X(t)=0$ then since $x(t)=X(t)X^{-1}(t_0)x_0$ we have: $$\left \| x(t) \right \| \le \left \| X(t) \right \|\cdot \left \| X^{-1}(t_0)\right \|\cdot \left \| x_0 \right \|$$ Thus, $\lim_{t \to \infty}\left \| x(t;t_0,x_0) \right \| =0$, forall $x(t)$ of $(I)$.

Hence, $(I)$ is asymptotically stable.

  • If $(I)$ is asymptotically stable. How we can show that $\lim_{t \to \infty}X(t)=0$?

The key step is to note that $x(t)=K(t,s)x_0$ solves the initial value problem

$$ \dot x(t)=A(t)x(t)\\ x(s)=x_0. $$

Now just take the definitions of stability and use $x(t)=K(t,s)x_0=X(t)X^{-1}(s)x_0$. Furthermore, note that $X^{-1}(s)x_0$ is a constant vector so it can be absorbed in some other constant when you take the norm.

Assume the equilibrium point is at the origin (you can always do it).

Stable $\iff$ $||x(t)||< M $ $\iff$ $|| X(t)X^{-1}(s)x_0 ||=||X(t)||\cdot||X^{-1}(s)x_0||<M$ $\iff$ $ || X(t)||<\frac{M}{||X^{-1}(s)x_0||}=P $.

The division is well defined, so $P$ is another constant.

Exponential Stability $\iff$ $||x(t)||< Me^{-ct}, \, c>0, \, t>s $ $\iff$ $|| X(t)X^{-1}(s)x_0 ||=||X(t)||\cdot||X^{-1}(s)x_0||<Me^{-ct}$ $\iff$ $ || X(t)||<\frac{Me^{-ct}}{||X^{-1}(s)x_0||}=Pe^{-ct} $. Next, $\displaystyle\lim_{t\to\infty}Pe^{-ct}=0$.

  • $\begingroup$ I agree with you: $x(t)=K(t,t_0)x_0=X(t)X^{-1}(t_0)x_0$, ok! But Why do you have: $x(t)=K(t,s)x_0=\boxed{X(t)X^{-1}(s)x_0 \cdot X^{-1}(s)x_0}$? $\endgroup$ – kimtahe6 Sep 25 '13 at 17:27
  • $\begingroup$ Sorry, the dot is not multiplication, it is the end of the sentence. I will edit. $\endgroup$ – PepeToro Sep 25 '13 at 17:30
  • $\begingroup$ You mean $$x(t)=X(t) \cdot C \implies X^{-1}(s)x_0=C=const$$. So $$\|x(t)\| \le \|X(t)\| \|X^{-1}(s)x_0\| =C \cdot \|X(t)\| <\epsilon$$, $\forall t \le t_0, \exists \delta >0, \|x_0\|< \delta$. Hence $M:=\epsilon /C$ $\endgroup$ – kimtahe6 Sep 25 '13 at 17:40
  • $\begingroup$ The first $\leq$ is equality, everything else seems correct. $\endgroup$ – PepeToro Sep 25 '13 at 17:46
  • 1
    $\begingroup$ I mean you can show that asymptotically stable iff $\|x(t)\|< Me^{-ct}, \, c>0, \, t>s$, when you get it in your proof. $\endgroup$ – kimtahe6 Sep 25 '13 at 18:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.