0
$\begingroup$

I have a Cauchy problem for the differential equation \begin{equation} y' = f(t, y), \end{equation} with initial condition $y(0) = y^0$; here, $y$ and $f$ are two-dimensional vector-functions. The function $f$ is smooth with respect to $(t, y) \in \mathbb R \times \mathbb R^2$. So, in accordance with classical results, the solution to this problem exists and is unique in some vicinity of $t = 0$.

I also have auxiliary Cauchy problems for differential equations \begin{equation} u' = U(u), \end{equation} and \begin{equation} u' = -U(u), \end{equation} with initial conditions $u(0) = u^0$ and $u(0) = u_0$, respectively; here, $u$ and $U$ are scalar (one-dimensional) functions and $U(u)$ is continuous with respect to $u \in \mathbb R$. Besides this, $U(u) \geqslant 0$ for $u \in \mathbb R$ and it is true that $$ |y'(t)| = |f(t, y)| \leqslant U(|y(t)|) $$ (that's why these Cauchy problems are called "auxiliary" for the original one).

Let $u^0 (t)$ and $u_0 (t)$ be solutions of the auxiliary Cauchy problems; it is supposed that they both exist on the segment $[0, \theta]$.

Finally, the following estimate \begin{equation} u_0 (t) \leqslant |y(t)| \leqslant u^0(t), \end{equation} is given, where $y(t)$ is the solution to the first (two-dimensional) Cauchy problem and $|\cdot|$ is the euclidean norm, i.e. $|y(t)| = \sqrt{y_1^2 + y_2^2}$.

So, I am interested if there is any theorem guaranteeing that my original problem has a solution defined on the same segment where $u^0 (t)$ and $u_0 (t)$ exist (that is $[0, \theta]$) and that on this segment the estimate mentioned above is true.

$\endgroup$
8
  • 1
    $\begingroup$ Is there any relationship at all between $u_0$, $u^0$, and $y_0$? Also, any relationship between $f$ and $U$? As it stands I don't see how there can be any theorem that does what you want. $\endgroup$ Commented Nov 17, 2020 at 22:05
  • $\begingroup$ Where (for what $t$) does the estimate $u_0(t)\leq |y(t)|\leq u^0(t)$ holds? $\endgroup$ Commented Nov 17, 2020 at 23:38
  • 1
    $\begingroup$ "so (...) the solution to this problem exists and is unique ". It is not unique, in general, with just continuous $f$. $\endgroup$ Commented Nov 18, 2020 at 0:27
  • $\begingroup$ @WillieWong, sorry for missing that in my question. I added some additional details to my question. $\endgroup$
    – StaTik
    Commented Nov 18, 2020 at 13:30
  • $\begingroup$ @AlexandreEremenko, this estimate holds on the interval where all these functions exist. It is known that $u_0(t)$ and $u^0 (t)$ exist at least on the segment $[0, \theta]$ and, as for $y(t)$, it is known that it exists in some vicinity of $0$. So, the estimate holds at least in some vicinity of $0$. I want to prove that $y(t)$ exists for $t \in [0, \theta]$ and that the estimate mentioned holds for $t \in [0, \theta]$. $\endgroup$
    – StaTik
    Commented Nov 18, 2020 at 13:36

0

You must log in to answer this question.