Let $P,Q,f:[-1,1]\to\mathbb{R}$ continuous and $a,b\in\mathbb{R}$. Then I want to prove that the IVP

$$\begin{cases} u''(x)+P(x)u'(x)+Q(x)u(x)=f(x)\\u(0)=a,\qquad u'(0)=b\end{cases}$$

has a unique solution in a neighborhood of x=0.

To this end I have to show that the function $F(x,y,t)$ is lipschitz in the first variables and the constant independent of the parameter $t$.

(This comes when this problem is converted to solve the integral equation and invoke the condition that the operator $\Phi(x)(t)=x_0 + \int_0^tF(x(s),s)ds$ has only one fixed point and the use the Banach fixed point theorem)

So I chose $F(x,y,t)=f(x)-P(x)-Q(x)$ then:


Now, by continuity of $f,P,Q$ in a compact set we have that they are uniformly continuous so we pick $\delta = \{ \delta_1,\delta_2, \delta_3 \}$ so if $|x-y|<\delta$ we get

$$|F(x,y,t)-F(u,v,t)|< \epsilon<\epsilon |x-y|$$

but the thing is that I don't know how to extend this property to all $x,y \in [-1,1]$

Can someone help me to fix this problem or provide another way to prove this Lipschitz condition ? because I don't know other way to proceed.

Thanks a lot in advance.

  • $\begingroup$ So, do you have to reconstruct all steps of existence and uniqueness theorem or you can just use it? Because it's pretty easy to just use it here. $\endgroup$ – Evgeny Oct 7 '15 at 4:51
  • $\begingroup$ I only have to use them :) so I only need to prove that the function I decided is Lipschitz in the first variables :). $\endgroup$ – user162343 Oct 7 '15 at 11:21
  • $\begingroup$ Just observe that derivatives w.r.t. to $u$ and $u'$ ($Q(x)$ and $P(x)$ respectively) are continuous functions. And it's well known property that if function is continuously differentiable then it's locally Lipschitz :) and that's all that you really need here. $\endgroup$ – Evgeny Oct 7 '15 at 11:31
  • $\begingroup$ Also, you should show that function $F(x, y, t) = f(t) - P(t) y - Q(t) x$ is Lipschitz in the variables $x$ and $y$, not the one that you've written. $\endgroup$ – Evgeny Oct 7 '15 at 11:33
  • 1
    $\begingroup$ You can transform your second order ODE to the system of first order equations $\frac{dx}{dt} = y$, $\frac{dy}{dt} = f(t) - P(t)y - Q(t)x$. For this system existence and uniqueness theorem tells that for having unique solution it's enough for RHS to be continuous w.r.t. to $t$ and Lipschitz-continuous w.r.t. to $x$ and $y$. $\endgroup$ – Evgeny Oct 7 '15 at 12:44

Expanding the comment about transforming second order equation to the system:

$$ \frac{dx}{dt} = y, \; \frac{dy}{dt} = f(t) - P(t)y- Q(t)x . $$

RHS is a vector now. Its first component is smooth w.r.t. all variables. The second component $F(x, y, t) = f(t) - P(t)y- Q(t)x$ is continuous w.r.t. $t$ and continuously differentiable in variables $x$ and $y$, because $ \frac{\partial F}{\partial x} = -Q(t), \frac{\partial F}{\partial y} = -P(t) $ and both are continuous functions. Hence $F(x, y, t)$ is Lipschitz in variables $x$ and $y$. Therefore, we can apply existence and uniqueness theorem in this case.

If function $F(x)$ is continuously differentiable w.r.t. $x$, then it's Lipschitz continuous w.r.t. to them

Since $F(x)$ is continuously differentiable with, we can write $F(x_1) - F(x_2)$ as $(x_1 - x_2) \cdot F'_{x} (x_1 + \theta (x_2 - x_1))$ where $\theta \in \lbrack 0, 1 \rbrack$. From here we have an estimation: $$\vert F(x_1) - F(x_2)\vert = \vert (x_1 - x_2) \cdot F'_{x} (x_1 + \theta (x_2 - x_1)) \vert = \vert x_1 - x_2 \vert \cdot \vert F'_{x} (x_1 + \theta (x_2 - x_1)) \vert \leqslant C \cdot \vert x_1 - x_2 \vert,$$ where $C$ is the supremum of $F'_x (x)$ in some neighbourhood of point $x_1$ (it exists for compact neighbourhoods because $F'_x$ is continuous).

  • $\begingroup$ I am back, what is your notation in $F'_x$? $\endgroup$ – user162343 Oct 9 '15 at 13:00
  • $\begingroup$ Of course it's derivative of $F(x)$ w.r.t. to $x$. What else it could be in this context? $\endgroup$ – Evgeny Oct 9 '15 at 14:23
  • $\begingroup$ Right :), and what do you think about my proof in the question math.stackexchange.com/questions/1471164/… ? How can I fix those bounds to get it of the form $L||w_1-w_2||$? $\endgroup$ – user162343 Oct 9 '15 at 14:25

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.