Use Laplace transforms to solve the boundary value problem $$Y_{xx}(x,t)-2Y_{tx}(x,t)+Y_{tt}(x,t)=0, \quad 0<x<1, t>0$$ $$Y(x,0)=Y_t(x,0)=0, \quad 0<x<1$$ $$Y(0,t)=0, \ Y(t,1)=F(t), \ t>0.$$

I am supposed to solve the PDE using Laplace transform. I know how to solve such PDE's like the wave equation using Laplace, but I dont know how to solve it for this problem where there is a mixed partial of $\frac{\partial^2}{\partial x\partial t}$. I tried searching online but I couldn't find anything.

What I have so far is $$\frac{d^2U}{dx^2}-2s\frac{\partial U}{\partial x}=-s^2U$$ $$U(0,s)=0,\quad U(1,s)=F(s)$$.

  • $\begingroup$ I haven't read the paper but here is one on the topic. It is a method called the Laplace substitution method. $\endgroup$ – dustin Nov 24 '14 at 22:48
  • $\begingroup$ When I transform $Y_{tx}$, I get $\int_{0}^{\infty}e^{-st}Y_{tx}\,dt = -Y_{x}(x,0)+sU(x,s)$. There are evaluation terms when I transform $Y_{tt}$, too. $\endgroup$ – DisintegratingByParts Nov 26 '14 at 0:13
  • $\begingroup$ When you're integrating in $t$, just write $Y_{xt}=Y_{tx}=\frac{\partial}{\partial t}Y_{x}$, and think of holding the $x$ variables constant while you work in $t$. $\endgroup$ – DisintegratingByParts Nov 26 '14 at 0:33
  • $\begingroup$ If $U$ is the transform of $Y$, then the transform of $Y_{t}$ is $sU+\cdots$ and the transform of $Y_{tx}$ is $s\frac{\partial}{\partial x}U+\cdots$ (where ... indicates evaluation terms.) So you transform without the derivatives in $x$, and then take those derivatives at the end. There are lots of regularity assumptions, but there always are anyway. $\endgroup$ – DisintegratingByParts Nov 26 '14 at 1:25
  • $\begingroup$ @T.A.E. So then will it just be $$\frac{d^2U}{dx^2}-2s\frac{\partial U}{\partial x}+s^2U=0?$$ Which is just an ODE? Hm, I don't think I am getting it still. I should place a bounty on the question. $\endgroup$ – Robben Nov 26 '14 at 1:44


$$y(x,s) = \int_0^{\infty} dt \, Y(x,t) \, e^{-s t}$$

Then, integrating by parts:

$$\int_0^{\infty} dt \, Y_t(x,t) \, e^{-s t} = -Y(x,0) + s y(x,s)$$

$$\int_0^{\infty} dt \, Y_{tt}(x,t) \, e^{-s t} = -Y_t(x,0) + s Y(x,0) + s^2 y(x,s)$$

Then using the initial conditions $Y(x,0)=Y_y(x,0)=0$, the PDE becomes the following ODE:

$$y''-2 s y' + s^2 y = 0$$

where $y(0,s)=0$ and $y(1,s)=f(s)$, where the prime represents derivative with respect to $x$, and where

$$f(s) = \int_0^{\infty} dt \, F(t) \, e^{-s t} $$

The general solution of the ODE is (I will not derive here)

$$y(x,s) = (A + B x) \, e^{s x}$$

Using the boundary conditions, we may find $A$ and $B$ and therefore the LT of the solution to the PDE:

$$y(x,s) = x \, f(s) \, e^{-s (1-x)} $$

We may find the inverse LT by convolution, as we know the individual LT's. The ILT of $f(s)$ is obviously $F(t)$ by definition, and the ILT of $e^{-s (1-x)}$ is $\delta(t-(1-x))$. Therefore, the ILT, and the solution to the equation, is

$$y(x,t) = x \int_0^t dt' \, F(t') \delta(t-t'-(1-x)) = x F(t-(1-x)) \theta(t-(1-x))$$

where $\theta$ is the Heaviside step function, which is necessary because the contribution to the integral from the $\delta$ function for $t < 1-x$ is zero (i.e., $t' \gt 0$ in the integral.)

  • $\begingroup$ Thank you very much!! It is much more clear to me now. $\endgroup$ – Robben Nov 26 '14 at 3:28

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.