Example using Green's Functions for 1D Poisson Equation with Forced Oscillations

1.4k Views Asked by At

I've found lots of resources deriving the Green's function for the 1D Poisson Equation, but no examples actually solving a DE with them. As a result, I'm very confused setting up the integrals for different forcing functions.

Consider the ODE with fixed boundary conditions $u(0)=0$ and $u(L)=L$: $$\frac{d^2u}{dx^2}=\frac{-f(x)}{T}$$ It can be shown that the Green's function for this problem is $$ G(x|\xi) = \begin{cases} \frac{(L-\xi)}{LT}x, & \text{for $x<\xi$} \\ \frac{(L-x)}{LT}\xi, & \text{for $x>\xi$} \end{cases}$$ Hence $$u(x)=\int_0^LG(x|\xi)f(\xi) \ d\xi=\int_0^xG(x|\xi)f(\xi) \ d\xi+\int_x^LG(x|\xi)f(\xi) \ d\xi$$ Now consider the forcing function $$f(x) = \begin{cases} f_0, & \text{for $0<x<\frac{L}{2}$} \\ -f_0, & \text{for $\frac{L}{2}<x<L$} \end{cases}$$ find $u(x)$

So far, I've broken the problem into 2 regions and set up $u(x)$ separately in both regions. I get a reasonable answer except it's discontinuous at $\frac{L}{2}$. When deriving the Green's function, I thought we used continuity of the string as a condition.

It's a lot to type out, so if anyone is interested in seeing the details, I'll link to a photo and an interactive graph of the solution.

1

There are 1 best solutions below

2
On BEST ANSWER

Green's function

Your Green's function only makes sense if you reformulate the differential equation as $$ -Tu''(x)=f(x). $$ Then indeed the solution to $-T\frac{\partial^2 G(x|\xi)}{\partial\xi^2}=\delta(\xi-x)$, $0=G(x|0)=G(x|L)$ is $$ G(x|\xi)=\begin{cases} \frac{(L-x)\xi}{LT} &\text{ for }\xi\le x, \\ \frac{(L-\xi)x}{LT} &\text{ for }\xi\ge x. \end{cases} $$

Homogeneous boundaries

To apply the Green function, you need homogeneous boundary conditions. To that end, select a function $u_0$ satisfying the boundary conditions and construct the equation for the difference. Here $u_0(x)=x$ and $v=u-u_0$ satisfies the same differential equation, as $v''=u''$.

Integrating the solution

The function $f$ that you insert is not a constant. It is a piecewise constant. This you have to include in your interval division.

Thus for $0\le x\le L/2\le L$ you get \begin{align} u(x)&=x+\int_0^L G(x|\xi)f(\xi)\,d\xi \\ &=x+\int_0^x \frac{(L-x)\xi}{LT}f_0\,d\xi+\int_x^{L/2} \frac{(L-\xi)x}{LT}f_0\,d\xi+\int_{L/2}^L \frac{(L-\xi)x}{LT}(-f_0)\,d\xi \\ &=x+\frac{f_0(L-x)x^2}{2LT}+\frac{f_0[-(L/2)^2+(L-x)^2]x}{2LT}-\frac{f_0(L/2)^2x}{2LT} \\ &=x+\frac{f_0x(L-2x)}{4T} \end{align} and similarly for $L/2\le x\le L$ \begin{align} u(x)&=x+\int_0^L G(x|\xi)f(\xi)\,d\xi \\ &=x+\int_0^{L/2} \frac{(L-x)\xi}{LT}f_0\,d\xi+\int_{L/2}^x \frac{(L-x)\xi}{LT}(-f_0)\,d\xi+\int_x^L \frac{(L-\xi)x}{LT}(-f_0)\,d\xi \\ &=x+\frac{f_0(L-x)(L/2)^2}{2LT}-\frac{f_0(L-x)[x^2-(L/2)^2]}{2LT}-\frac{f_0(L-x)^2x}{2LT} \\ &=x+\frac{f_0(L-x)(L-2x)}{4T} \end{align}