3

I have a equation like this:

$\dfrac{\partial y}{\partial t} = -A\dfrac{\partial y}{\partial x}+ B \dfrac{\partial^2y}{\partial x^2}$

with the following I.C
$y(x,0)=0$

and boundary conditions $y(0,t)=1$ and $y(\infty , t)=0$

I tried to solve the problem as follows: Taking Laplace transform on both sides,

$\mathcal{L}(\dfrac{\partial y}{\partial t}) = - A \mathcal{L}(\dfrac{\partial y}{\partial x})+B \mathcal{L}(\dfrac{\partial^2 y}{\partial x^2})$

Now, on the L.H.S we have, $sY-y(x,0)=sY$

$\mathcal{L}(\dfrac{\partial^2 y}{\partial x^2}) =\displaystyle \int e^{-st} \dfrac{\partial^2 y}{\partial x^2} dt$

Exchanging the order of integration and differentiation

$\displaystyle\mathcal{L}(\frac{\partial^2 y}{\partial x^2}) =\frac{\partial^2}{\partial x^2} \int e^{-st} y(x,t) dt$

$\displaystyle\mathcal{L}(\frac{\partial^2 y}{\partial x^2}) =\frac{\partial^2}{\partial x^2}\mathcal{L}(y) = \frac{\partial^2Y}{\partial x^2} $

$\displaystyle\mathcal{L}\frac{\partial y}{\partial x} = \frac{\partial Y}{\partial x}$

Now, combing L.H.S and R.H.S, we have,

$\displaystyle sY = - A \frac{\partial Y}{\partial x} + B \frac{\partial^2Y}{\partial x^2}$

Above equation might have three solutions:

If $b^2 - 4ac > 0 $ let $r_1=\frac{-b-\sqrt{b^2-4ac}}{2a}$ and $r_2 = \frac{-b+\sqrt{b^2-4ac}}{2a}$

The general solution is $\displaystyle y(x) = C_1e^{r_1x}+C_2 e^{r_2x}$

if $b^2 - 4ac = 0 $, then the general solution is given by

$ y(x)=C_1e^{-\frac{bx}{2a}}+C_2xe^-{\frac{bx}{2a}}$

if $b^2 - 4ac <0$ , then the general solution is given by

$y(x) = C_1e^{\frac{-bx}{2a}}\cos(wx) + C_2 e^{\frac{-bx}{2a}}\sin(wx)$

Since, A, and B are always positive in my problem, the first solution seems to be appropriate.

Now, from this point I am stuck and couldn't properly use the boundary conditions.

If anyone could offer any help that would be great.

"Solution added" The solution of the problem is

$y(x,t)= \dfrac {y_0}{2} [exp(\dfrac {Ax}{B}erfc(\dfrac{x+At}{2\sqrt{Bt}}) + erfc(\dfrac{x-At}{2\sqrt{Bt}})$

Jdbaba
  • 197
  • Is it a must of solving this question by using laplace transform? I get a simpler procedure without using laplace transform. – doraemonpaul Apr 01 '13 at 01:37
  • No, I can use any procedure to solve the problem. I would highly appreciate your solution with any method. Thanks. – Jdbaba Apr 01 '13 at 01:47
  • Please note that when $A=0$ , the result should be comparable with the results obtained for examples from http://math.stackexchange.com/questions/112845/random-diffusion-coefficient-in-the-fourier-equation, http://math.stackexchange.com/questions/184110 and http://inside.mines.edu/~abunge/classes/chen507/MathStuff/PDEcombvar.pdf. – doraemonpaul Apr 04 '13 at 02:49
  • Also, the result obtained should make sure it can really satisfly for $u(\infty,t)=0$ . – doraemonpaul Apr 04 '13 at 02:54

2 Answers2

2

Using your notation, we have

$$B Y''(x,s) - A Y'(x,s) - s Y(x,s) = 0$$

$$Y(0,s)=\int_0^{\infty} dt\: 1 \cdot e^{-s t} = \frac{1}{s}$$

$$\lim_{x \rightarrow \infty} Y(x,s) = 0$$

The general solution to the equation is

$$Y(x,s) = M(s) e^{r_+ x} + N(s) e^{r_- x}$$

where

$$r_{\pm} = \frac{A \pm \sqrt{A^2+ 4 B s}}{2 B}$$

where

$$M(s) + N(s) = \frac{1}{s}$$

$$M(s)=0$$

The latter equation was determined by the limit at $\infty$. (I am of course assuming that $B > 0$; the other case may be considered as well.)

Therefore

$$Y(x,s) = \frac{1}{s} \exp{\left[-\frac{\sqrt{A^2+4 B s}-A}{2 B} x \right ]}$$

It is this quantity that must be inverse LT'ed to get the solution to your equation, viz.

$$y(x,t) = \frac{1}{i 2 \pi} \int_{c-i \infty}^{c+i \infty} ds \: \frac{1}{s} \exp{\left[-\frac{\sqrt{A^2+4 B s}-A}{2 B} x + s \, t\right ]}$$

Ron Gordon
  • 138,521
  • Hi Ron Gordon, thank you so much for your answer. `The solution of the problem is

    $y(x,t)= \dfrac {y_0}{2} [exp(\dfrac {Ax}{B}erfc(\dfrac{x+At}{2\sqrt{Bt}}) + erfc(\dfrac{x-At}{2\sqrt{Bt}})$` I didn't understand the last equation in your solution. Could you please explain that little bit ?

    – Jdbaba Mar 24 '13 at 18:48
  • The last equation is the equation for the inverse Laplace transform - what you need to derive from the LT'ed solution. See http://en.wikipedia.org/wiki/Inverse_Laplace_transform – Ron Gordon Mar 24 '13 at 18:49
  • @ Ron, thank you for explaining things so nicely. But I am stuck solving this complex integral also. Can you please help me solve this integral by any suggestion ? Thanks. – Jdbaba Mar 24 '13 at 19:26
  • @Jdbaba: this one is a tough one. I am currently looking around for a solution. Mathematica offers none. I do not think that standard residue techniques will work here. Your answer looks plausible to be sure - kind of a damped diffusion equation. And my solution checks out as well in the limit of $A$ going to $0$. Stay tuned here, I will come up with something. – Ron Gordon Mar 24 '13 at 19:38
  • @ Ron: Thank you so much for your reply. I really appreciate your time and support. – Jdbaba Mar 24 '13 at 19:40
  • @ Ron : The solution I posted uses the boundary condition $y(0,t)=C_0$. I think that is the reason there is $C_0$ in the solution. – Jdbaba Mar 24 '13 at 20:00
  • 1
    OK, just remember that we will solve (or try) the problem you post, so hopefully you can translate what I wrote into the initial condition you meant. BTW my pleasure posting this solution, I love this stuff. I am still looking for that ILT. – Ron Gordon Mar 24 '13 at 20:05
  • Hi I was wondering whether you got the solution of the integral or not. – Jdbaba Mar 25 '13 at 00:10
  • 1
    @Jdbaba: not yet; an exhaustive search of the literature has not turned up anything yet. Really trying to figure out how to evaluate the integral myself at this point. Not easy, I'm going to have to ask you to hang on...it may be a few days. – Ron Gordon Mar 25 '13 at 00:27
  • Hi Ron, I was trying to solve the problem without the diffusive term and tried to solve with the remaining part of the equation. But still I couldn't get the solution. I am stuck on getting the inverse laplace transform of $1/s(e^(-s/v)x)$. Can you give any advice ? – Jdbaba Mar 26 '13 at 01:58
1

I get a simpler procedure that without using laplace transform.

Note that this PDE is separable.

Let $y(x,t)=X(x)T(t)$ ,

Then $X(x)T'(t)=-AX'(x)T(t)+BX''(x)T(t)$

$X(x)T'(t)=(BX''(x)-AX'(x))T(t)$

$\dfrac{T'(t)}{T(t)}=\dfrac{BX''(x)-AX'(x)}{X(x)}=\dfrac{4B^2s^2-A^2}{4B}$

$\begin{cases}\dfrac{T'(t)}{T(t)}=\dfrac{4B^2s^2-A^2}{4B}\\BX''(x)-AX'(x)-\dfrac{4B^2s^2-A^2}{4B}X(x)=0\end{cases}$

$\begin{cases}T(t)=c_3(s)e^\frac{t(4B^2s^2-A^2)}{4B}\\F(x)=\begin{cases}c_1(s)e^\frac{Ax}{2B}\sinh xs+c_2(s)e^\frac{Ax}{2B}\cosh xs&\text{when}~s\neq0\\c_1xe^\frac{Ax}{2B}+c_2e^\frac{Ax}{2B}&\text{when}~s=0\end{cases}\end{cases}$

$\therefore y(x,t)=\int_{-\infty}^\infty C_1(s)e^\frac{2Ax+t(4B^2s^2-A^2)}{4B}\sinh xs~ds+\int_{-\infty}^\infty C_2(s)e^\frac{2Ax+t(4B^2s^2-A^2)}{4B}\cosh xs~ds$

$y(0,t)=1$ :

$\int_{-\infty}^\infty C_2(s)e^\frac{t(4B^2s^2-A^2)}{4B}~ds=1$

$C_2(s)=\dfrac{1}{2}\left(\delta\left(s-\dfrac{A}{2B}\right)+\delta\left(s+\dfrac{A}{2B}\right)\right)$

$\therefore y(x,t)=\int_{-\infty}^\infty C_1(s)e^\frac{2Ax+t(4B^2s^2-A^2)}{4B}\sinh xs~ds+\int_{-\infty}^\infty\dfrac{1}{2}\left(\delta\left(s-\dfrac{A}{2B}\right)+\delta\left(s+\dfrac{A}{2B}\right)\right)e^\frac{2Ax+t(4B^2s^2-A^2)}{4B}\cosh xs~ds=\int_{-\infty}^\infty C_1(s)e^\frac{2Ax+t(4B^2s^2-A^2)}{4B}\sinh xs~ds+e^\frac{Ax}{2B}\cosh\dfrac{Ax}{2B}$

doraemonpaul
  • 16,178
  • 3
  • 31
  • 75