While solving a PDE problem involving the Laplace equation in 3D, I arrive at the following summation relation when i substitute the only non-homogeneous boundary condition available

$$ \sum_{m=1}^{\infty}\sum_{n=1}^{\infty}A_{n,m}\cos(\frac{n\pi x}{L})\cos(\frac{m\pi y}{l})\sqrt{\frac{n^2\pi^2}{L^2}+\frac{m^2\pi^2}{l^2}}\sinh\bigg(w\sqrt{\frac{n^2\pi^2}{L^2}+\frac{m^2\pi^2}{l^2}}\bigg) = p_h\bigg(\sum_{m=1}^{\infty}\sum_{n=1}^{\infty}A_{n,m}\cos(\frac{m\pi y}{l})\cos(\frac{n\pi x}{L})\cosh\bigg(w\sqrt{\frac{n^2\pi^2}{L^2}+\frac{m^2\pi^2}{l^2}}\bigg) - \sum_{m=1}^{\infty}\sum_{n=1}^{\infty} \frac{A_{n,m}b_h}{b_h^2 + n^2 \pi^2 }\cos(\frac{m\pi y}{l})\cosh\bigg(w\sqrt{\frac{n^2\pi^2}{L^2}+\frac{m^2\pi^2}{l^2}}\bigg)\bigg[b_h\cos(\frac{n\pi x}{L}) + n\pi\sin(\frac{n\pi x}{L}) \bigg]\bigg) $$

I need to find the Fourier coefficients here, denoted by $A_{n,m}$.


As a starting point i know that if $\cos(\frac{m\pi x}{L})$ and $\cos(\frac{n\pi y}{l})$ is multiplied on the RHS and LHS of the whole equation i could use the known orthogonality of

$\int_{0}^{L} \cos(\frac{n\pi x}{L})\cos(\frac{m\pi y}{L})$ and $\int_{0}^{L} \cos(\frac{n\pi x}{l})\cos(\frac{m\pi y}{l})$

But the problem is, I will get a term like $\int_{0}^{L} \cos(\frac{m\pi x}{L})\sin(\frac{n\pi x}{L})$.

I do not know what to do with this term. Additionally, I have another doubt as to how the $A_{n,m}$ occurring on both sides of the equation are to be handled. Also, I have done 2D problems but how to handle this 3D case ? Is a double integration required ?


The question asked has its origin as follows:-

From a solution of the form

$$ T(x,y,z)=\sum_{m=1}^{\infty}\sum_{n=1}^{\infty}A_{n,m}\cos(\frac{n\pi x}{L})\cos(\frac{m\pi y}{l})\cosh\left(\sqrt{\frac{n^2\pi^2}{L^2}+\frac{m^2\pi^2}{l^2}}z\right). $$

with bc(s) as:

$\frac{\partial T(0,y,z)}{\partial x}=\frac{\partial T(L,y,z)}{\partial x}=0 $

$\frac{\partial T(x,0,z)}{\partial y}=\frac{\partial T(x,l,z)}{\partial y}=0$

$$\frac{\partial T(x,y,-w)}{\partial z}=p_h\bigg( T(x,y,-w) - \frac{e^\frac{-b_h x}{L}b_h}{L}\int e^\frac{b_h x}{L}T\mathrm{d}x \bigg) $$

$ \frac{\partial T(x,y,0)}{\partial z} = 0. $

Extra information From the physical problem available at hand the following is too known:-

$$\frac{\partial T(0,y,-w)}{\partial z}=p_h\bigg( T(0,y,-w) - T_i\bigg) $$


Multiplying both sides by $\cos(\frac{k\pi x}{L})$ and $\cos(\frac{j\pi y}{l})$ assuming that the series converges, converting the summation to an integration, first for $x$ and then for $y$, followed by using orthogonality.

$$ \sum\sum A_{n,m}\sqrt{()}\sinh(w\sqrt{()})\bigg(\frac{L}{2} \vee n=k\bigg)\bigg(\frac{l}{2} \vee m=j\bigg) = p_h\bigg[\sum\sum A_{n,m}\cosh(w\sqrt{()})\bigg(\frac{L}{2} \vee n=k\bigg)\bigg(\frac{l}{2} \vee m=j\bigg)- \sum\sum\frac{A_{n,m}b_h^2}{b_h^2 + n^2 \pi^2}\cosh(w\sqrt{()})\bigg(\frac{L}{2} \vee n=k\bigg)\bigg(\frac{l}{2} \vee m=j\bigg)\bigg] $$

leads to

$$ A_{k,j}\sqrt{()}\sinh(w\sqrt{()})\bigg(\frac{L}{2}\bigg)\bigg(\frac{l}{2} \bigg) = p_h\bigg[A_{k,j}\cosh(w\sqrt{()})\bigg(\frac{L}{2}\bigg)\bigg(\frac{l}{2}\bigg)-\frac{A_{k,j}b_h^2}{b_h^2 + n^2 \pi^2}\cosh(w\sqrt{()})\bigg(\frac{L}{2}\bigg)\bigg(\frac{l}{2}\bigg)\bigg] $$

Hence, $A_{k,j}$ needs to be evaluated. But still I have the problem of $A_{k,j}$ cancelling out from the resulting equation.

  • $\begingroup$ en.wikipedia.org/wiki/… $\endgroup$ – Paul Sinclair Feb 17 at 0:09
  • $\begingroup$ @PaulSinclair I followed the link you suggested. I have thought about writing $\cos(\frac{m\pi x}{L})\sin(\frac{n\pi x}{L})$ as $\frac{1}{2}\bigg(\sin(\frac{m\pi x}{L} + \frac{n\pi x}{L}) +\sin(\frac{m\pi x}{L} - \frac{n\pi x}{L})\bigg)$. But now the arguments of these $\sin$ functions are different from the rest of the equation. Can you elaborate a bit more ? $\endgroup$ – Indrasis Mitra Feb 17 at 1:36
  • $\begingroup$ Ypu are integrating them from $0$ to $L$. What happens when you integrate those two terms over a this range? $\endgroup$ – Paul Sinclair Feb 17 at 1:49
  • $\begingroup$ @PaulSinclair $\cos(\frac{m\pi x}{L})\sin(\frac{n\pi x}{L})$ .On integrating ove this range they just result in $0$ ? $\endgroup$ – Indrasis Mitra Feb 17 at 1:59
  • $\begingroup$ Exactly. $\cos m\theta$ and $\sin n\theta$ are orthogonal to each other even when $m = n$. $\endgroup$ – Paul Sinclair Feb 17 at 2:05

I admit that I had failed to check this at the start: you have $A_{m,n}$ on both sides. One solution to your original equation is just $A_{m,n} = 0$ for all $m,n$. And in fact, that is the only solution.

To see this, multiply by $\sin\left(\frac{k\pi x}L\right)$ and take the $\int_0^{2L}\ \ dx$ integral of both sides (note the $2L$ - you have to integrate over a full period, and that is $2L$, not $L$). Every term with $\cos\left(\frac{n\pi x}L\right)$ in it becomes $0$. And the only $\sin\left(\frac{n\pi x}L\right)$ term that survives is $n = k$. Now, $\int_0^{2L}\sin^2\left(\frac{k\pi x}L\right)\,dx = L$, so the result is

$$0 = -p_h\left[\sum_{m=1}^{\infty}\frac{A_{k,m}b_h}{b_h^2 + k^2 \pi^2 }\cos(\frac{m\pi y}{l})\cosh\left(w\sqrt{\frac{k^2\pi^2}{L^2}+\frac{m^2\pi^2}{l^2}}\right)k\pi L \right]$$

Then we multiply through by $\cos\left(\frac{j\pi y}l\right)$ and take the $\int_0^{2l}\ \ dy$ integral, which similarly leaves us with

$$0 =-p_h\left[\frac{A_{k,j}b_h}{b_h^2 + k^2 \pi^2 }l\cosh\left(w\sqrt{\frac{k^2\pi^2}{L^2}+\frac{j^2\pi^2}{l^2}}\right)k\pi L \right]$$ $\cosh$ is never $0$. If we assume that $p_h, b_h, L$, and $l$ are all non-zero, this solves to $A_{k,j} = 0$.

Either you have an error in deriving the equation posted, or else your solution is $0$.

  • $\begingroup$ Thankyou Paul. I will recheck my derivations. Actually the only non-homogeneous BC in the problem (if you see the note in the question) has $A_{m,n}$ in all the terms which is what causes the problem. Although i do have an extra information which tells that at $(0,y,-w)$ (instead of $(x,y,-w)$ in the original problem) the bc is $\frac{\partial T(0,y,-w)}{\partial x} = p_h(T(0,y,-w)-T_i)$ where $T_i$ is a constant.This does remove the $A_{m,n}$ cancelling out problem. But now this reduces the bc to an edge type condition in a 3D problem. $\endgroup$ – Indrasis Mitra Feb 17 at 17:35
  • $\begingroup$ I have added this information to the original question. Thankyou anyway. Your comments and answer really made a lot of muddy points clear. $\endgroup$ – Indrasis Mitra Feb 17 at 17:42
  • $\begingroup$ By $\frac{\partial T(x,y,-w)}{\partial z}$, do you mean $\frac{\partial T}{\partial z}(x,y,-w)$, or are any or all of $x, y, w$ supposed to be a function of $z$? (This is an order of operations question - the way you wrote it, T is evaluated at (x,y,-w) first, then the result is differentiated by $z$. But unless the other variables are functions of $z$, the differentiation will be $0$.) $\endgroup$ – Paul Sinclair Feb 17 at 17:49
  • $\begingroup$ Sorry, my bad. It is $\frac{\partial T}{\partial x}$ at $(x,y,-w)$. In words : the derivative of $T$ evaluated at $(x,y,-w)$ $\endgroup$ – Indrasis Mitra Feb 17 at 18:00
  • $\begingroup$ Added another attempt to math.stackexchange.com/questions/3117489/… . the exponential is still what causes problems. have a look if you get time. $\endgroup$ – Indrasis Mitra Feb 23 at 16:27

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.