1
$\begingroup$

I'm solving the heat equation on a two dimensional square domain. The problem is defined as: $$ u_{xx}+u_{yy} = 0 \hskip{0.5cm}\text{for} \hskip{0.5cm} 0 \leq x \leq 1, 0 \leq y \leq 1 $$ with the following boundary conditions: $$ u(0,y)= 0, \hskip{0.5cm} u(1,y)= 0, \hskip{0.5cm} u(x,1)= 0, \hskip{0.5cm} u(x,0)= sin(\pi x) $$ I'm able to solve this eqations both numerically and analytically, however I'm having trouble with the following: I would like to convert the boundary condition on $u(x,0)$ to an equivalent Neumann boundary condition. I've tried the following: $$ u_x(x,0) = \pi cos(\pi x) $$ However this is not resulting in a satisfactory result. Is my thinking correct?

edit:

The exact solution I get is: $$ u(x,y)= sin(\pi y) \cdot \left[ cosh(\pi x) - coth(\pi) \cdot sinh(\pi x) \right] $$

Edit:

I've got it working now using $u_y(x,y) = \pi cosh(\pi) / sinh(\pi) sin(\pi x)$ as the neumann boundary condition.

Thank you for your help,

Jan Willem

$\endgroup$
0
$\begingroup$

Look at a solution $v(x,y)$ of the Neumann problem \begin{align*} v_{xx}+v_{yy}& =0\text{ in the unit square} \\ v_{x}(0,y)=v_{x}(1,y)& =0 \\ v_{y}(x,1)& =0 \\ v_{y}(x,0)& =-\pi \cos \pi x. \end{align*} Now obtain a solution $u(x,y)$ of your Dirichlet problem as the curve integral \begin{equation*} u(x,y)=\int_{\varphi }\left( -v_{y}\mathrm{\,d}x+v_{x}\,\mathrm{d}y\right) , \end{equation*} where $\varphi $ is any curve in the unit square connecting $(0,0)$ to $(x,y) $. By the Gauss-Green theorem and the heat equation satisfied by $v$, the value of this integral is independent of the path. For example, \begin{equation*} u(x,y)=-\int_{0}^{x}v_{y}(s,0)\mathrm{\,d}s+\int_{0}^{y}v_{x}(x,t)\,\mathrm{d }t. \end{equation*} From this follows that \begin{align*} u_{x}& =-v_{y} \\ u_{y}& =v_{x}, \end{align*} and hence \begin{equation*} u_{xx}+u_{yy}=-v_{yx}+v_{xy}=0. \end{equation*} Also the Dirichlet boundary values are easily checked.

$\endgroup$
  • $\begingroup$ I see what you mean. I messed up and got it to work (see my edit). Thanks. $\endgroup$ – Jan Willem Feitsma Oct 9 '15 at 9:49

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.