Basically I got a simple wave equation with an extra twist. The PDE is $$\frac {\partial^2 y}{\partial t^2} = c^2\frac {\partial^2 y}{\partial x^2} + L$$ with homogeneous boundary condition. As usual, I use the ansatz $ Y(x,t) = F(x)G(t) $ and I have $\frac {\partial^2 y}{\partial t^2} =F''G $ and $\frac {\partial^2 y}{\partial x^2} =FG''$. I substitute this in, and now I have $$FG''= c^2F''G +L$$ I don't know how to make this separable--obviously if I divide both sides with $FG$ as usual, then I will have that annoying constant $\frac {L}{FG} $ and I don't know how to make it separable then.

Can anybody help me? I know I have to separate it somehow, but I don't know exactly how to.

{The actual problem is an inhomogeneous boundary condition type of problem, but I have found the steady state solution, so I just have to find the transient solution, thus the homogeneous boundary condition}

The actual question:

  • PDE: $$\frac {\partial^2 y}{\partial t^2} = c^2\frac {\partial^2 y}{\partial x^2} + L,\quad 0 \leq x \leq J,\quad t>0$$ where $L$ is constant and $c$ is constant wave speed.
  • Boundary Condition: $ u(0,t)=0$, $t>0$ and $u(J,t)=h$, $t>0$
  • Initial Condition: $u(x,0) = 0$, $0<x<J$ and $\frac {\partial u}{\partial t} (x,0)=0$, $0<x<J$

I did the steady state solution bit, and got it as $w(x)= \frac{-Lx^2}{c} + \frac {Hx}{L} +\frac{GLx}{c} $. So I modified the boundary condition and initial condition to get the transient part, and I got the boundary condition homogeneous. However when I try to solve it, the question above arises.


Let $y(x,t) = w(x,t) + \phi(x)$ where $\phi(x) = a x^2 + b x + c$. From this it is seen that $\phi'(x) = 2 a x + b$, $\phi''(x) = 2a$ and \begin{align} \partial_{t}^{2} w = c^2 \, \partial_{x}^{2} w + 2a c^2 + L. \end{align} In order to cancel the $L$ term let $2 a c^2 + L = 0$ which leads to \begin{align} \phi(x) = - \frac{L \, x^{2}}{2 \, c^{2}} + b \, x + c_{1}. \end{align} Now it is understood that the transformation \begin{align} y(x,t) = w(x,t) - \frac{L \, x^{2}}{2 \, c^{2}} + b \, x + c_{1} \end{align} will make the p.d.e more solvable without "complications".

How this applies to the original problem is the following: \begin{align} u_{tt} = c^{2} u_{xx} + L \hspace{5mm} u(0,t) = 0 , \, u(a,t) = h\\ u(x,0) = 0, \, u_{t}(x,0)= 0 \end{align} The solution form proposed is \begin{align} u(x,t) = w(x,t) - \frac{L \, x^{2}}{2 \, c^{2}} + b \, x + c_{1} \end{align} for which \begin{align} u(0,t) &= 0 = w(0,t) + c_{1} \\ u(a,t) &= h = w(a,t) - \frac{L \, a^{2}}{2 \, c^{2}} + b a + c_{1} \\ u(x,0) &= 0 = w(x,0) - \frac{L \, x(x-a)}{2 \, c^{2}} + \frac{h \, x}{a} = w(x,0) - \phi_{1}(x) \\ u_{t}(x,0) &= 0 = w_{t}(x,0) \end{align} From the first equation it is seen that $w(0,t) = 0$ and $c_{1} = 0$. From the second let $w(a,t) = 0$ to obtain \begin{align} b = \frac{h}{a} + \frac{L \, a}{2 \, c^{2}}. \end{align} From all of this it can now be said that the original differential set is transformed by \begin{align} u(x,t) &= w(x,t) - \frac{L \, x(x-a)}{2 \, c^{2}} + \frac{h \, x}{a} \end{align} into \begin{align} w_{tt} = c^{2} w_{xx} \hspace{5mm} w(0,t) = w(a,t) = 0 , w(x,0) = \phi_{1}(x), w_{t}(x,0) = 0 \end{align} where \begin{align} \phi_{1}(x) = \frac{L \, x(x-a)}{2 \, c^{2}} - \frac{h \, x}{a}. \end{align} The equation for $w(x,t)$ is then an easier equation to solve. Since the boundaries for $x$ are zero at each end it suggests a sine solution and can be stated as \begin{align} w(x,t) = \sum_{n=1}^{\infty} \left( A_{n} \cos\left( \frac{n \pi c t}{a} \right) + B_{n} \sin\left( \frac{n \pi c t}{a} \right) \right) \, \sin\left( \frac{n \pi x}{a} \right) \end{align}

Note: the boundary conditions $w(x,0) = w_{t}(x,0) = 0$ leads to $A_{1}$ and $B_{1}$ being zero. This suggests that $w_{t}(x,0) \neq 0$.

Returning to the original form the solution is of the form \begin{align} u(x,t) &= \frac{h \, x}{a} - \frac{L \, x(x-a)}{2 \, c^{2}} + \sum_{n=1}^{\infty} \left( A_{n} \cos\left( \frac{n \pi c t}{a} \right) + B_{n} \sin\left( \frac{n \pi c t}{a} \right) \right) \, \sin\left( \frac{n \pi x}{a} \right) \end{align}

  • $\begingroup$ I'm very lost. So with the ϕ(x) known, do we just essentially do the w(x,t) part of the pde? Do we just ignore the L constant and then just plugging in it back once we got into the solution? $\endgroup$ – Skipe Apr 25 '15 at 18:11
  • $\begingroup$ @user234395 yes. Since the p.d.e. depends on $x$ and $t$ the remaining $w(x,t)$ would satisfy $w_{tt} = c^2 w_{xx}$. $\endgroup$ – Leucippus Apr 25 '15 at 18:33
  • $\begingroup$ So, when we do the w(x,t) pde part, does the boundary condition changes? I do kind of understand your method, but it is still quite fuzzy to me. Thanks a lot for your help by the way! $\endgroup$ – Skipe Apr 25 '15 at 19:56
  • $\begingroup$ Or does the boundary condition changes--like do we subtract the 0 boundary condition with ϕ(0)=c? If it is, then it would be pretty complicated wouldn't it? I would have to do the steady state again and things like that? And also does the initial condition changes? $\endgroup$ – Skipe Apr 25 '15 at 21:18
  • $\begingroup$ @Skipe edit the question/problem to include the boundary conditions of the problem and I'll use them to show how to utilize them. $\endgroup$ – Leucippus Apr 25 '15 at 22:47

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.