I would like to solve the $\triangle u(x,y) = 0$ in the unit square, with periodic BC when $x=0,1$ and Neumann condition when $y=0,1$ $$\partial_y u(x,0) = \begin{cases} A \quad &\text{for } 0\leq x\leq L\\ B \quad &\text{for } L<x\leq 1\\ \end{cases}\quad \partial_y u(x,1) = \begin{cases} B \quad &\text{for } 0\leq x\leq 1-L\\ A \quad &\text{for } 1-L<x\leq 1\\ \end{cases} $$ We also fix $u(0,0) = 0$ to make the solution unique.

The hint said instead of considering Green's functions or separation of variables, think about simple solutions of $\triangle u = 0$, draw a picture consider the solution in different parts of the domain and derive jump conditions to connect the solutions.

My attempt: Here let us call $$g(x) = \begin{cases} A \quad &\text{for } 0\leq x\leq L\\ B \quad &\text{for } L<x\leq 1\\ \end{cases}$$ and note that the Nuemann condition at $z=0$ is $g(x)$, and at $z=1$ is $g(1-x)$.

If provided $g(x) = \sum_n a_n \cos(n\pi x)$ (point-wise or uniform), we can define $$u(x,y) = \sum_n a_n \frac{\sin(n\pi x - n\pi y)}{-n\pi}.$$ (I realized this $u$ is not harmonic) $u$ is periodic when $x=0,1$, and
$$\partial_y u(x,0) = \sum _n a_n \cos(n\pi x) = g(x) $$ and $$\partial_y u(x,1) = \sum _n a_n \cos(n\pi (x - 1))=\sum _n a_n \cos(n\pi (1-x)) = g(1-x).$$

But I am not sure if we can find a cosine series which pointwise converges to $g(x)$, how could we fix this? I think I have $L^2$ convergence $$\sum_n a_n \cos(n\pi x) \rightarrow g \text{ in } L^2$$ which means I have pointwise a.e. convergence because of the summation.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.