Let the initial value problem : $$y'=(x-y)^{1/2}, y(2)=2$$ Study the existence of solutions in a domain around the initial value and study its uniqueness too.

Attempt :

In order to show the uniqueness, I'll work around the function $f(x,y) = (x-y)^{1/2}$ and it's Lipschitz condition with respect to $y$ :

$$|f(x,y_2) - f(x,y_1)| = |(x-y_2)^{1/2} - (x-y_1)^{1/2}|$$

For the function $g(y) = (x-y)^{1/2}$, we can use the Mean Value Theorem since it's continuous and differentiable for $x-y \geq 0 \Leftrightarrow x \geq y$, for some $ξ \in (y_1,y_2) $ :

$$\big((x-y)^{1/2}\big)'|_{y=ξ} = \frac{(x-y_2)^{1/2} - (x-y_1)^{1/2}}{y_2 - y_1}$$


$$-\frac{1}{2\sqrt{x-ξ}}(y_2-y_1) = (x-y_2)^{1/2}-(x-y_1)^{1/2}$$


$$\bigg|-\frac{1}{2\sqrt{x-ξ}}\bigg|\bigg|y_2-y_1\bigg| = \bigg| (x-y_2)^{1/2}-(x-y_1)^{1/2}\bigg|$$

Now, I am not sure if I can make an assumption about a constant $L$ with respect to $y$. I observe that $k=\bigg|-\frac{1}{2\sqrt{x-ξ}}\bigg|$ is independent of $y_2,y_1$ but I cannot see how to make an inequality out of this, since it's value would change for any different $x$. Can you give me some help regarding how to show that there exists a Lipschitz constant or not ?

Also, regarding the solution, how would I study the existence of solutions in a domain around $y(2) = 2$ ? Can you please explain thoroughly with a complete solution regarding this, because it's an issue I have in a lot of similar problems.

  • 1
    $\begingroup$ At $y=x$ you have a vertical tangent in direction $y$ and thus no Lipschitz constant. $\endgroup$ – Dr. Lutz Lehmann Oct 23 '17 at 7:46
  • $\begingroup$ @LutzL ty for the reply. What about the existence of solution though around the domain of the initial value ? $\endgroup$ – Rebellos Oct 23 '17 at 9:31
  • $\begingroup$ @LutzL By the way, I suppose you mean a vertical tangent at $x=ξ$ ? $\endgroup$ – Rebellos Oct 23 '17 at 9:45
  • $\begingroup$ At any point $(x,y)$ where $y=x$, so also at the given initial point. However, contrary to $y'=\sqrt{y}$ the zero curve of the radicant is not a solution itself, so that the possibilities for multiple solutions are reduced. $\endgroup$ – Dr. Lutz Lehmann Oct 23 '17 at 13:39

Let $z(x)=x-y(x)$, then $z'(x)=1-y'(x)$ hence $$z'(x)=F(z(x))\qquad z(2)=0$$ with $$F:u\mapsto1-\sqrt{u}$$ This is an autonomous differential equation with $F(z(2))\ne0$ hence, for every $x\geqslant2$, $$\int_0^{z(x)}\frac{du}{1-\sqrt{u}}=x-2$$ that is, using the change of variable $v=\sqrt u$, $$x-2=\int_0^{\sqrt{z(x)}}\frac{2vdv}{1-v}=\left.-2\log(1-v)-2v\right|_0^{\sqrt{z(x)}}=-2\log(1-\sqrt{z(x)})-2\sqrt{z(x)}$$ or, equivalently, $$(1-\sqrt{z(x)})\,e^{\sqrt{z(x)}}=e^{1-x/2}$$ One sees that $z(x)$ increases from $z(2)=0$ to $z(+\infty)=1$, and that, for every $x\geqslant2$, $y(x)$ is the unique solution of the identity $$(1-\sqrt{x-y(x)})\,e^{\sqrt{x-y(x)}}=e^{1-x/2}$$ On the other hand, again using $F(z(2))>0$, one sees that the solution is not defined on $x<2$.

Nota: The fact that $F(z(2))>0$ implies that every solution $z(x)$ is instantaneously positive, that is, that $z(x)>0$ for every $x>2$, and then Cauchy-Lipschitz uniqueness and existence criterion applies. However, a rigorous argument showing uniqueness and existence starting at $z(2)=0$ is lacking (Cauchy-Peano-Arzelà provides existence, confirmed by the explicit computations above, but the hypotheses of Osgood's criterion, for example, are not met).


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.