Let $f:\mathbb{R}\times\mathbb{R}^{n}$ continuous such that $x'=f(t,x)$ has uniqueness of solution, and $|f(t,x)|\leq 10$. I wnat to prove that every solution for the ODE is defined for all $t\in\mathbb{R}$.

Seems like the hypothesis $|f(x,t)|<10$ is just to say that $f$ is bounded.

A question: $f$ having uniqueness of solution implies $f$ is Lipschitz?

I'm asking that because, if that is true, it contradicts what the Picard Theorem says about the interval of solution.

How can I solve that?

  • 1
    $\begingroup$ The answer to the question is: NO. Indeed, in a sense a "generic" RHS has the property that we have local uniqueness of the IVP. How can that contradict the Picard theorem? $\endgroup$
    – user539887
    Apr 20, 2019 at 18:25
  • $\begingroup$ Just some thoughts, I'm wrong. How can I solve that? $\endgroup$ Apr 21, 2019 at 1:46

1 Answer 1


I saw an answer to a similar question on MSE some time ago, but as I cannot find it I write the following.

Assume that $f \colon \mathbb{R} \times \mathbb{R}^n \to \mathbb{R}^n$ is continuous and that there exists $M > 0$ such that $\lvert f(t, x) \rvert \le M$ for all $(t, x) \in \mathbb{R} \times \mathbb{R}^n$.

Suppose to the contrary that some nonextedible (to the right, for the definiteness' sake) solution $\varphi$ of $x' = f(t, x)$ is defined on an open interval whose right endpoint, $\beta$, is $< \infty$. Observe that for any two $t_1, t_2$ belonging to that interval there holds $$ \tag{$*$} \lvert \varphi(t_1) - \varphi(t_2) \lvert = \Bigl\lvert \int\limits_{t_1}^{t_2} \varphi'(s) \, \mathrm{d}s \Bigr\rvert \le \int\limits_{t_1}^{t_2} \lvert \varphi'(s) \rvert \, \mathrm{d}s \le M \lvert t_2 - t_1 \rvert. $$ Now, let $(t_k)$, $t_k < \beta$, be a sequence convergent to $\beta$ as $k \to \infty$. $(t_k)$ is a Cauchy sequence, so, by $(*)$, $(\varphi(t_k))$ is a Cauchy sequence, too. By the completeness of $\mathbb{R}$, $(\varphi(t_k))$ has a (finite) limit, $\tilde{\varphi}$. Indeed, this limit is the same for all sequences converging to $\beta$ from the left (if there were sequences $\check{t}_k$, $\hat{t}_k$ for which the limits are different, then the limit for the sequence $(\check{t}_1, \hat{t}_1, \check{t}_2, \hat{t}_2, \dots)$ would not exist), and we can legitimately denote it as $\varphi(\beta)$. We can easily prove that the extended function $\varphi$ has left-hand derivative, $\varphi'_{-}(\beta)$, at $\beta$, equal to $f(\beta, \varphi(\beta))$.

We have thus extended $\varphi$ to a solution of $x' = f(t, x)$ on a right-closed interval with right endpoint $\beta$. Apply the (local) existence theorem (for instance, the Peano theorem) to the IVP $$ \begin{cases} x' = f(t, x) \\ x(\beta) = \varphi(\beta) \end{cases} $$ to extend the solution $\varphi$ to the right of $\beta$. But this contradicts our assumption that $\varphi$ cannot be extended to the right. So the right endpoint of the interval of existence of $\varphi$ is $\infty$.

In a similar way we prove that the left endpoint of the interval of existence of $\varphi$ must be $-\infty$. Notice that we nowhere assume the uniqueness property.


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.