Local uniqueness of Cauchy Problem, but can it split into branches?

156 Views Asked by At

The problem I've been struggling is exemplified here:

Is local Lipschitz continuity sufficient for an ODE to have a unique solution?

The problem that they solve is

\begin{align*} x'(t) = x^2,\quad x(0) = x_0 \end{align*}

which in the other thread is said to have a unique solution given by

\begin{align*} x(t) = \frac{1}{x_0^{-1}-t}. \end{align*}

This function is not continuous at $t = x_0^{-1}$ bu they claim in the other thread that it is unique in $(-\infty,x_0^{-1})$, supposing $x_0>0$. Since $f(x,t) = x^2$ the function is only locally Lipschitz and therefore we are only guaranteed a unique solution locally? How can one in general be sure that it does not split into separate branches outside of some small interval?

1

There are 1 best solutions below

1
On BEST ANSWER

Local uniqueness already garantuees uniqueness on the whole intervall of existence. Let me give a short proof:

Let $u_1$ and $u_2$ solve the same initial value problem. Let us also assume there exists a $t\in\mathbb{R}$ such that for every $\varepsilon>0$ we find an $\tilde{t}\in(t,t+\varepsilon)$ such that $u_1(\tilde{t})\neq u_2(\tilde{t})$ and $u_1(t)=u_2(t)$. This means $u_1$ and $u_2$ are equal up to $t$, but after that $t$, they may differ. Local uniqueness now gives us a small open neighborhood of $t$ in which $u_1$ equals $u_2$. Hence we find an $\tilde{\varepsilon}>0$ such that for every $x\in(t,t+\tilde{\varepsilon})$ we have $u_1(x)=u_2(x)$. If you choose you $\varepsilon$ now to be smaller than $\tilde{\varepsilon}$ you obtain a contradiction.

Your example is also unique up to the point of existence. Please note, that your $x(t)$ tends to $\infty$ if $t\uparrow x_0$. Hence your solution cannot be extended further and therefore it cannot branch. (See also https://en.wikipedia.org/wiki/Ordinary_differential_equation#Global_uniqueness_and_maximum_domain_of_solution)