0
$\begingroup$

Consider the equation of the form $\epsilon y^{\prime \prime}+a y^{\prime}=0$ on $x\in[0,1]$ with $a\in\mathbb{R}$, $0<\epsilon\ll1$,$y(0) =\alpha$, and $y(1)=\beta$. Show that if $a >0$ then the boundary layer is near $x= 0$ with the method of matched asymptotic expansions.

To find the inner solution, Wikipedia defines a new variable $\tau=x/\epsilon$, but why not $x/\epsilon^{2}$ or $x/\epsilon^{3}$? I already know it depends on the thickness of the boundary layer, but for my problem above, how can I determine the thickness? (How can I define a new variable to find the inner solution? Is $\tau=x/\epsilon$ okay?)


I got another question just now, i.e. how can I prove the boundary layer is near $x=0$? When I have the outer solution, there exists two boundary condition but I don't know which one should be used. How can I determine which one is the "outer"?

$\endgroup$
1
$\begingroup$

You are right that you can not necessarily just "see" what the thickness $\tau$, the exponent in $\delta=ϵ^τ$, of the boundary layer is. You would then need to make the general zooming coordinate transform $x=x_0+δX$ and consider the ODE for $Y(X)=y(x)$ which in the current case gives $$ ϵY''(X)+aδY'(X)=0. $$ The enjoining reasoning is that $Y$, having the same values as $y$, is bounded and that this coordinate transform renders also the derivatives of $Y$ bounded relative to $ϵ\to 0$. For that being the case, the terms in the equation have to balance, $ϵ$ and $δ$ need to be on the same scale. This is easiest achieved by just setting $δ=ϵ$. Thus $Y(X)=Ce^{-aX}+D$, $C\ne 0$ (As $C=0$ implies that no boundary layer exists at this point).

Now one needs to ensure that the boundedness assumption is actually satisfied for this inner solution. If $x_0$ is an inner point, this requires $\lim_{X\to\pm \infty}Y(X)$ being finite/existing, which is not the case here. Thus $x_0$ has to be one of the boundary points, and only the limit towards the inside the interval has to be finite, which is feasible.

$\endgroup$
5
  • $\begingroup$ Thanks. I got another question just now, i.e. how can I prove the boundary layer is near $x=0$? When I have the outer solution, there exists two boundary condition but I don't know which one should be used. How can I determine which one is the "outer"? $\endgroup$ – Thinkpad Nov 21 '19 at 2:37
  • 1
    $\begingroup$ If $a>0$, then $e^{-aX}$ goes to infinity for $X\to -\infty$. This means that $x_0$ has to be $0$, else the inner solution takes ever greater values for $ϵ→0$, it can not be connected to the outer solution, which is (relatively) independent of $ϵ$, in that direction.. $\endgroup$ – Lutz Lehmann Nov 21 '19 at 12:06
  • $\begingroup$ Happy new year. :) Can you kindly take a look at this post and suggest some good books you know on perturbation theory. $\endgroup$ – Hosein Rahnama Jan 1 '20 at 12:39
  • 1
    $\begingroup$ @H.R. I have looked into the online preview of Mark H. Holmes "Introduction to Perturbation Methods" for this question on multi-scale expansions, it looks solid, but I have no idea of how it compares to the other cited books. $\endgroup$ – Lutz Lehmann Jan 1 '20 at 13:08
  • $\begingroup$ Thanks. :) It would help future readers if you just mention your experience with what you have read. $\endgroup$ – Hosein Rahnama Jan 1 '20 at 13:19

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.