2
$\begingroup$

I need to solve the following PDE, given the boundary conditions and initial condition, using Laplace Transforms:

$$ \begin{aligned} \displaystyle K_1\frac{\partial^2u}{\partial x^2}+K_2u&=\frac{\partial u}{\partial t} \\ u(0,t)&=u_0 \\ u(\infty,t)&=0 \\ u(x,0)&=0 \end{aligned} $$

So far, I've taken the Laplace transform of the PDE to get the following result:

$$K_1\frac{\partial ^2U}{\partial x^2}+K_2U=sU$$

Solving the above ODE results in the following:

$$U(x,s)=C_1\exp\left(\sqrt{\frac{k_2+s}{k_1}}x\right)+C_2\exp\left(-\sqrt{\frac{k_2+s}{k_1}}x\right)$$

By taking the Laplace transform of the two boundary conditions, I get the following:

$$U(0,s)=\frac{u_0}{s}$$

$$U(\infty,s)=0$$

Using the second boundary condition, I can calculate that $C_2=0$, and that the PDE in terms of $x$ and $s$ is:

$$U(x,s)=\frac{u_0}{s}\exp\left(\sqrt{\frac{K_2+s}{K_1}}x\right)$$

Now, I have to take the inverse Laplace transform, but I'm having problems at this step. I have the result which is:

$$ u(x,t)=\frac{2u_0}{\sqrt{\pi}}\int_{\frac{x}{2\sqrt{K_1t}}}^{\infty} \exp\left(-\lambda^2-\frac{K_2x^2}{4K_1\lambda^2}\right) \ d\lambda $$

It's suggested that I use one of the shifting theorems to obtain this solution, but I'm not sure how. Thanks!

$\endgroup$
  • $\begingroup$ Are you allowed to use complex analysis? $\endgroup$ – Dylan Nov 28 '17 at 2:34
  • $\begingroup$ We haven't been taught that in the course...but how would you do it using complex analysis? $\endgroup$ – fred12 Nov 28 '17 at 4:24
  • $\begingroup$ Generally the inverse transform is defined as an contour integral parallel to the imaginary line. $\endgroup$ – Dylan Nov 28 '17 at 4:38
2
$\begingroup$

There is some simplification if you rewrite the PDE as $$ K_1\frac{\partial^2u}{\partial x^2}=\frac{\partial u}{\partial t}-K_2u \\ K_1 \frac{\partial^2 (e^{-K_2t}u)}{\partial^2 x}=\frac{\partial (e^{-K_2t}u)}{\partial t} $$ Let $v(x,t)=e^{-K_2t}u(x,t)$. Then the new PDE becomes $$ v_t = K_1 v_{xx} \\ v(0,t) = u_0e^{-K_2t} \\ v(\infty,t) = 0 \\ v(x,0) = 0. $$ Now when you transform in $t$: \begin{align} \mathscr{L}\{v_t\} & =\int_{0}^{\infty}v_t(x,t)e^{-st}dt \\ & =v(x,t)e^{-st}|_{t=0}^{\infty}-\int_{0}^{\infty}v(x,t)e^{-st}(-s)dt \\ & = s\mathscr{L}\{v\}. \end{align} So the transformed equations for $\mathscr{L}\{v\}(s,x)$ are \begin{align} s\mathscr{L}\{v\}(s,x) &=K_1\mathscr{L}\{v\}_{xx}(s,x), \\ \mathscr{L}\{v\}(s,0)&= \mathscr{L}\{u_0e^{-K_2t}\}=\frac{u_0}{s+K_2}. \end{align} The solution in $s,x$ is $$ \mathscr{L}\{v\}(s,x) = \frac{u_0}{s+K_2}\exp\left(-\sqrt{\frac{s}{K_1}}x\right). $$ The inverse Laplace transform is a convolution in $t$ of inverse transforms of the two terms multiplying each other on the right above: \begin{align} v(t,x) & = \left( u_0 e^{-K_2 t}\right)*\left( \frac{1}{\pi}\int_{0}^{\infty}e^{-st}\sin\left(\sqrt{\frac{s}{K_1}}x\right)ds\right) \\ & = \frac{u_0}{\pi}\int_{0}^{\infty}\int_{0}^{t}e^{-K_2t'}e^{-s(t-t')}dt'\sin\left(\sqrt{\frac{s}{K_1}}x\right)ds \end{align} The integral in $t'$ is easily simplified and the solution $u(x,t)=e^{K_2t}v(x,t)$ of the original equation recovered, but I'll leave it there, before I compound my errors. :)

$\endgroup$

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.