Two systems are topologically equivalent if there exists a homeomorphism of the phase space which takes phse curves of one system to phase curves of another.

Theorem Let $f\colon \mathbb{R}^n\to\mathbb{R}^n, \lambda\colon\mathbb{R}^n\to\mathbb{R}$ smooth, $\lambda>0$. Then the two systems $\dot{x}=f(x)$ and $\dot{x}=\lambda(x)f(x)$ are topologically equivalent and the homeomorphism is the identity map.

Unfortunately, I do not understand the following proof:


(The proof consists of proving that a solution $x=x(t)$ which is a solution of the first system, is also a solution of the second system.)

We show that there exists a monotonic function $s(t)$ of $t$, such that $x(s(t))$ is a solution of the second system.

Let $x=x(t)$ be a solution of the first system. We want $x(s(t))$ to be a solution of the second system, hence

$$ \frac{dx(s(t))}{ds}\cdot\frac{ds(t)}{dt}=\lambda(x(s(t))\cdot f(x(s(t)).~~~(*) $$

We cannot - as one might think at first sight - cancel $\frac{dx(s(t))}{ds}$ from the left side and $f(x(s(t))$ from the right side. But we can write the following:

$$ \left(\frac{ds(t)}{dt}-\lambda(x(s(t))\right)\cdot f(x(s(t)) =0.~~~~(**) $$

Then the first or the second factor is zero.

The second factor is zero if the solution $x(t)$ is constant. That is, if $x$ is an equilibrium, we are done.

Let the first factor be zero. Then we can separate variables and get $$ \int_{0}^s \frac{ds}{\lambda(x(s(t))}=\int_{t_0}^t \, dt\Leftrightarrow t=t_0+\int_{0}^s\frac{dw}{\lambda(x(w))}=F(s) $$ and we get $s(t)=F^{-1}(t)$.

There are three main points I do not understand.

1) What does the expression $\frac{dx(s(t))}{ds}$ mean, to what are we derivating? $s$ is a function...

(2) Why can't we delete/ cancel $\frac{dx(s(t))}{ds}$ from the left side and $f(x(s(t))$ from the right side?

(3) How do we get from $(*)$ to $(**)$?

(4) And how do we know that $s(t)=F^{-1}(t)$ is within the existence interval of the solution?

Very confusing....

Maybe you know a source where I can read that more clearly as we did it in the reading...


1) This is just the chain rule. See https://en.wikipedia.org/wiki/Chain_rule. $s$ itself becomes a free variable in that case.

2) Because $f(x(s(t)))$ might be 0. If we try to cancel, it would be undefined.

3) $dx(t)/dt = f(x(t))$ is the same thing as $dx(s)/ds = f(x(s))$. Just think $s$ is a free variable like $t$, and indeed it is as you can select $s$ arbitrarily and the equation is satisfied.

4) $t = F(s)$ means that $s = F^{-1}(t)$. Event though it is not stated explicitly in this equation, $s$ itself depends on $t$, so we can write $s(t) = F^{-1}(t)$.

I understand that your main confusion is about using a function as a free variable. Well, a free variable is, by definition, that you can select the values of it freely, and some other function may depend on it. Selecting values freely includes selecting values using a rule (or a function), so you can also use a function as a free variable, as is done in (*).


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.