Review / Part I — Explicit & Qualitative Methods / Ch 2

Existence and Uniqueness

Ch 2 — Part I — Explicit & Qualitative Methods

When does \(y' = f(t,y),\; y(t_0) = y_0\) have a solution, and when is it unique? The Picard–Lindelöf theorem gives a clean sufficient condition.

Picard's theorem

If \(f(t,y)\) is continuous and \(\partial f/\partial y\) is continuous on a rectangle around \((t_0, y_0)\), then the IVP has a unique solution on some open interval containing \(t_0\).

Things that can go wrong

Linear case is nicer

For linear ODEs \(y' + p(t) y = g(t)\) with \(p,g\) continuous on an interval \(I\), solutions exist and are unique on all of \(I\). No blow-up surprises.

Practice Problems

Problem 1medium
Does \(y' = y^{1/3},\; y(0) = 0\) have a unique solution?
Hint
Check \(\partial f/\partial y\) at \(y = 0\).
Solution

\(\partial_y(y^{1/3}) = \tfrac{1}{3} y^{-2/3}\) is undefined at \(y=0\). Uniqueness fails: both \(y\equiv 0\) and \(y(t) = (2t/3)^{3/2}\) are solutions.

Answer: No — non-unique.

Problem 2medium
For \(y' = 1 + y^2,\; y(0)=0\), find the largest \(t\) for which the solution is defined.
Hint
Separable; solution is \(\tan t\).
Solution

\(\int dy/(1+y^2) = \int dt\Rightarrow \arctan y = t + C\). \(C=0\), so \(y = \tan t\). Blows up at \(t = \pi/2\).

Answer: \(t \in (-\pi/2, \pi/2)\).

Problem 3easy
State the hypotheses you need on \(p(t), g(t)\) in the linear first-order IVP \(y' + p(t)y = g(t)\) to guarantee a unique solution on an interval \(I\).
Hint
Continuity is enough in the linear case.
Solution

We need \(p, g\) continuous on \(I\). The solution exists and is unique on all of \(I\).

Answer: \(p, g\) continuous on \(I\).