Let $f:\mathbb{R}\times\mathbb{R}^n\to \mathbb{R}^n$ be a continuously differentiable function. Suppose that exists $v:\mathbb{R}\to[0,\infty)$ continuous such that $||f(t,x)||\leq v(t)||x||\; \forall (t,x)\in \mathbb{R}\times\mathbb{R}^n$. Prove that all maximal solutions for $\dot{x}=f(t,x)$ are defined in $\mathbb{R}$; what happens if $f$ is bounded?

I believe that the Picard–Lindelöf theorem is the way to go, but I don't know how to prove that $f$ satisfies the Lipschitz condition. In order to use it, if I understand correctly, I should prove that $v$ is bounded for every $t$ in a neighborhood of $(t_0,x_0)$ being the last a solution for the differential equation.

If $\dot{x}=f(t,x)$ then $x(t)=\displaystyle\int f(t,x)dt$. Considering the hypothesis, $$\displaystyle{||x||\leq\int||f(t,x)||dt\leq\int v(t) ||x||dt}\\1\leq\int||f(t,x)||dt\leq\int v(t)dt$$

which doesn't seem useful at all....

First update

Alright, let's try it this way: The goal is to show that we always have a is defined in $\mathbb{R}$ for any IVP. Suppose $(t_0,x_0)$ is a maximal solution for $\dot{x}=f(t,x)$ with $x(t_0)=a$, then I should prove that exists an interval $[t_0-c,t_0+c]$ in which the solution exists.

$v$ is continuos, then for any $t_0$ we can pick $c$ and $[t_0-c,t_0+c]$ will be be bounded. Define $K=\operatorname{max}\{v(t):t\in [t_0-c,t_0+c]\}$. Then $f$ is Lipschitz for every $(t,x)\in [t_0-c,t_0+c]\times \mathbb{R}^n$, this is, $||f(t,x)||\leq v(t)||x||\leq K ||x||\;\forall (t,x)\in [t_0-c,t_0+c]\times\mathbb{R}^n$

By the Picard-Lindelöf theorem, the IVP has an unique maximal solution. And $t_0,c$ were arbitrary, can I conclude that all maximal solutions are defined?

Second update

Looking about extensibility I found these notes about differential equation. The extensibility theorem and the corollary seem to be enough to prove that the solutions exists for all $\mathbb{R}$.

I'll quote both here:

  • Suppose that $f$ is $C^1$ on $\mathbb{R}^n$. Denote the unique solution by $x(t)$ and suppose $J:=(a,b)$ is the maximal interval of existence.

    Theorem 1.20 (Extensibility Theorem): For each compact set $K\subset \mathbb{R}^n$ there is a $t\in J$ such that $x(t)\notin K$; thus, $$\lim_{t\to b^-}|x(t)|=\lim_{t\to a^+}|x(t)|=+\infty .$$

    Corollary 1.21: Without loss of generality, if $f:\mathbb{R}^n\to \mathbb{R}^n$ is continuous, then the solutions to $\dot{x}=f(x)$, $x(0)=x_0$ exist $\forall t\in \mathbb{R}$.

With the corollary and the continuity of $f:\mathbb{R}\times\mathbb{R}^n\to\mathbb{R}^n$, can I affirm that all maximal solutions can be defined in $\mathbb{R}$ for the equation $\dot{x}=f(t,x)$?

  • 1
    $\begingroup$ You need some mild assumption on $v$, for example that it is continuous. If it is continuous, then it is bounded on compact intervals, which is exactly the local Lipschitz property that you need for Picard-Lindelof. $\endgroup$ – Ian Nov 10 '14 at 0:44
  • $\begingroup$ @Ian You're right, I missed that part when I was copying the problem. I updated the post with new attempt to solve it. $\endgroup$ – Cure Nov 10 '14 at 1:21
  • 1
    $\begingroup$ You are asked to prove that solutions are defined globally, for all $\mathbb R$. The Picard theorem is essentially local, it guarantees the uniqueness and existence of solution on a some small interval. To work through your problem you need several facts about extensibility of solutions. $\endgroup$ – Artem Nov 10 '14 at 2:15
  • $\begingroup$ @Artem Thanks. Could you be a bit more explicity about the $\text{several}$ facts about extensibility? I looked for it and I found the theorem I posted in the second update of the post, but I'm unaware of what other facts are needed to answer the question. $\endgroup$ – Cure Nov 10 '14 at 3:02
  • 1
    $\begingroup$ Not really. Again, the estimate $|f(t,x)|\leq T|x|$ is essential, without it any proof doomed to be wrong (see my example above) $\endgroup$ – Artem Nov 10 '14 at 3:18

By the Picard-Lindelöf theorem, the IVP has an unique maximal solution. And $t_0,\,c$ were arbitrary, can I conclude that all maximal solutions are defined?

No, you cannot conclude this, because, as I mentioned, Picard's theorem is local.

To finish your prove you need to have $$ x(t)=x_0+\int_{t_0}^t f(\tau,x)d\tau, $$ which implies that $$ |x(t)|\leq |x_0|+K|x|(t-t_0), $$ using your notations. After this you will need a Grownwall's inequality, to find that $$ |x(t)|\leq\ldots, $$ which, together with my previous remarks, finished the proof. I think you should be able to fill in the necessary details.


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.