1
$\begingroup$

Consider the initial value problem (IVP) \begin{cases} y'(t)= t + \sin(y(t)), \\ y(2) = 1. \\ \end{cases} Find the largest interval $\mathcal{I}\subset \mathbb{R}$ containing $t_0=2$ so that the problem has a unique solution $y$ in $\mathcal{I}$.


My proof attempt:

Let $L>\frac{\pi}{2}>0$. Define $$ R := \{(t,y)\in \mathbb{R}^2:|t-2|\leq L, |y-1|\leq L \} $$ Then $$ 2-L\leq t \leq 2+L \text{ and } 1-L\leq y \leq 1+L $$ Since $L>\frac{\pi}{2} \Rightarrow \exists y_0\in (1-L, 1+L)$ so that $\sin(y_0)=1$. Hence, $$ M = \underset{R}{\sup}|F(t,y)|=3+L $$ Then $$ |\partial_y F(t,y)|=|\cos(y)|\leq 1 $$ Put $c = 1$, then by the Mean Value theorem, for $(t,y),(t,u)\in R \Rightarrow$ $$ |F(t,y)-F(t,u)|\leq c|y-u| $$ Let $a_{*}=\min \left(L, \frac{L}{M}\right)=\min \left(L, \frac{L}{3+L}\right)=\frac{L}{3+L}$. Using Picard-Lindelof's theorem, there should exist a unique solution on the interval $\mathcal{I}=\left[2-\frac{L}{3+L}, 2+\frac{L}{3+L}\right]$.

Since $\underset{L\rightarrow \infty}{\lim}\frac{L}{3+L}=1$, our largest interval should be $\mathcal{I}=\left[1+\varepsilon,3-\varepsilon\right]$ where $\varepsilon \in (0,1)$.

Am I on the right track here?

$\endgroup$
  • 1
    $\begingroup$ I have also read somewhere that there is a global solution (i.e. $\mathcal{I}=\mathbb{R}$) if F(t,y) is globally lipschitz. Which I believe it is since $\nabla F=(1, \cos(y))$. Which should be bounded under the operator norm. So how do I reconcile this fact with what I did up above? Is what I did wrong? $\endgroup$ – Joe Man Analysis Nov 19 '18 at 4:49
2
$\begingroup$

From the Picard theorem one can infer the following: Given an ODE $y'=f(t,y)$ with $f$ defined in an open set $\Omega\subset{\mathbb R}^2$ and fulfilling the assumptions of the theorem in the neighborhood of each point $(t,y)\in\Omega$, any solution of an IVP $y(t_0)=y_0$ can be extended to a maximal solution $$\tilde y:\quad J\mapsto{\mathbb R},\quad x\mapsto \tilde y(x)$$ in an unique way. Here $J$ is an open interval which may depend on the given initial point $(t_0,y_0)$. Furthermore the graph ${\cal G}(\tilde y)\subset\Omega$ of this maximal solution will "ultimately" leave any compact set $K\subset\Omega$ given in advance. (For example, the solution cannot develop a $x\mapsto\sin{1\over x}$ singularity in the interior of $\Omega$.)

In the case at hand we have $f(t,y)=t+\sin y$ and $\Omega={\mathbb R}^2$. It follows that for any solution $t\mapsto y(t)$ one has $|y'(t)|\leq |t|+1$. This allows to conclude that $|\tilde y(t)|\leq C(1+t^2)$ for a suitable $C>0$; hence $\tilde y$ cannot drift away to $\pm \infty$ in finite time. Since ${\cal G}(\tilde y)$ will ultimately leave any compact rectangle $K:[-M,M]\times[-C(2+M^2),C(2+M^2]$, and cannot do so across the bottom and top edges, it follows that $\tilde y$ is defined on all of ${\mathbb R}$.

$\endgroup$
  • 1
    $\begingroup$ Could you please explain how the bound on the maximal solution was obtained? $\endgroup$ – Joe Man Analysis Nov 19 '18 at 10:49

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.