1
$\begingroup$

I wish to solve for $u(x,y)$ from the PDE $u_x + y u_y = -u$, $u(x,1) =x^2$. Since this is a first order linear PDE, I think I should try the method of characteristics. Let $u = u(x(s), y(s))$.

Then the char equations are $\frac{dx}{ds} = 1$, $\frac{dy}{ds} = y$ and $\frac{du}{ds} = -u$.

Solving: $x = s+c_1$, $y = c_2 e^{s}$, and $u = c_3e^{-s}$. Now I eliminate the $s$ parameter: $s = x - c_1$ which implies $y = c_2 e^{x- c_1}$ and $u = c_3 e^{-x + c_1}$. If I let $x(0) = x_0$, then $x_0 = c_1$. But i'm a little confused on how to proceed. Could anyone advize on this?

$\endgroup$
1
$\begingroup$

Using the fact that $x(0) = x_0$ and $y(0) = 1$, then we see that \begin{align} x= x_0 +s, \ \ y= e^s, \ \ \ u(x, y) = u(x_0, 1)e^{-s} = x_0^2e^{-s}. \end{align} Finally, since $s= \log y$ and $x_0 = x-s = x-\log y$, then we have that \begin{align} u(x, y) = (x-\log y)^2\exp(-\log y) = \frac{(x-\log y)^2}{y}. \end{align}

Check: Observe \begin{align} u_x= \frac{2(x-\log y)}{y}, \ \ u_y = \frac{-2(x-\log y)-(x-\log y)^2}{y^2} = \frac{-(x-\log y)(2+x-\log y)}{y^2} \end{align} which means \begin{align} u_x+yu_y = \frac{2(x-\log y)-(x-\log y)(2+x-\log y)}{y} = -\frac{(x-\log y)^2}{y}=-u. \end{align}

$\endgroup$
  • $\begingroup$ Why is $y(0) = 1$? $\endgroup$ – Taln Apr 4 '18 at 4:19
  • $\begingroup$ Since you have the initial condition $u(x, 1) = x^2$, then it's natural to set $y=1$ at $s=0$. $\endgroup$ – Jacky Chong Apr 4 '18 at 4:20
  • 1
    $\begingroup$ @Taln see my answer; you want to pick characteristic curves that pass through the region where you have data $\endgroup$ – qbert Apr 4 '18 at 4:21
  • $\begingroup$ I understand now. Thank you to both of you. $\endgroup$ – Taln Apr 4 '18 at 4:22
0
$\begingroup$

Basically, you want your characteristic curves to pass through the line $(x,1)$ in $\mathbb{R}^2$.

Checking your solutions, $$ \begin{cases} x(s)=s+x_0\\ y(s)=y_0e^s\\ u(s)=u_0e^{-s} \end{cases} $$ where it would be convenient to take $y_0=1$, since this brings us to the line on which we have boundary data. Then, we get $$ u_0=u(0)=u(x_0,1)=x_0^2 $$ and we derived the following system $$ \begin{cases} x=s+x_0\\ y=e^s\\ u=x_0^2e^{-s} \end{cases}\\ \implies u(x,y)=\frac{x_0^2}{y}=\frac{(x-\log y)^2}{y} $$

$\endgroup$

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.