1

I have to solve $u_{tt}-u_{xx}=0$ with the given I.C.s

\begin{cases} u_x(0,t)=u_x(\pi,t)=0\\ u(x,0)=\cos x \\ u_t(x,0)=-\cos x \end{cases}

Solving the PDE with separation of variables : \begin{equation} \begin{array} e-\frac{1}{X}u_{xx}=-k^2 \\ \frac{1}{T}u_{tt}=-k^2 \end{array} \end{equation}

which give the following solutions: \begin{gather} u(x,t)= \begin{cases} Ae^{kx}\cos kt\\ Ae^{-kx}\cos kt \\ Ae^{kx}\sin kt\\ Ae^{-kx}\sin kt \end{cases} \end{gather} Given the I.C. nr 2, I can see that the solution must have the cosine term, so I disregard the sine solutions. Then, since this is an oscillatory problem, I disregard the solution with $e^{kt}$ since $e^{kt}\rightarrow \infty$ (since this is not a perturbed system ). I am left with

\begin{equation} u(x,t)= Ae^{-kx}\cos kt \end{equation}

When I want to find A, I use I.C. 2:

\begin{equation} cosx= Ae^{-kx}\cos 0t \rightarrow A=\cos xe^{-kx} \end{equation}

This gives

\begin{equation} u(x,t)= \cos x\cos kt \end{equation}

But from here, I am not sure about $k$. If I use the other conditions, I just get $0=0$ or I use the third IC:

\begin{equation} \begin{array} fu_t(x,t)= -k\cos x\sin kt\\ IC: u_t(x,0)=-\cos x:\\ -\cos x= -k\cos x\sin 0\\ \frac{1}{k}=0\\ k=0? \end{array} \end{equation}

This gives

\begin{equation} u(x,t)= cosx \end{equation}

However, this is a strange result, since the PDE is for a function $u(x,t)$, having only one variable. Is there an apparent error here?

Thanks

UPDATE:

Separation of variables was not properly written:

Solving the PDE with separation of variables : \begin{equation} \begin{array} e\frac{1}{F}F_{xx}=k^2 \\ \frac{1}{T}T_{tt}=k^2 \end{array} \end{equation}

which gives the two ODEs:

\begin{equation} \begin{array} eF_{xx}-Fk^2=0 \\ T_{tt}-Tk^2=0 \end{array} \end{equation}

with the solutions:

\begin{cases} Ae^{kx} \\ Ae^{-kx}\\ Be^{kt}\\ Be^{-kt} \end{cases}

Here we can disregard from the positive exponentials, and have:

$u(x,t)=Ae^{-kx}Be^{-kt}$ which we can write as:

$u(x,t)=Ae^{-k(x+t)}$

$u_t(x,t)=-Ake^{-k(x+t)}$

$u_x(x,t)=-Ake^{-k(x+t)}$

Using second IC:

\begin{equation} \begin{array} f u(x,t)=Ae^{-k(x+t)}\\ IC: u(x,0)=\cos x \\ Ae^{-kx}=\cos x \\ A=e^{kx}\cos x \end{array} \end{equation}

So we have now $u(x,t)=e^{kx}\cos x e^{-k(x+t)}\rightarrow \cos x e^{-kt}$

Therefore: $u(x,t)= \cos x e^{-kt}$

We find k by the 3d IC:

\begin{equation} \begin{array} f u(x,t)= \cos x e^{-kt}\\ IC: u_t(x,0)=-\cos x \\ -k\cos x e^{0}=-cosx\\ k= 1 \end{array} \end{equation}

So the final answer is:

\begin{equation} u(x,t)= \cos x e^{-t} \end{equation}

A double check show that this is correct:

\begin{equation} \begin{array} fu_{xx}(x,t)= \cos x e^{-t} \\ u_{tt}(x,t)= \cos x e^{-t} \\ u_{tt}-u_{xx}=\cos x e^{-t}-\cos x e^{-t}=0 \end{array} \end{equation}

So the original problem $u_{tt}-u_{xx}=0$ is satisfied by $u(x,t)= \cos x e^{-t} $.

A plot of $u(x,t)$ is given below

enter image description here

Luthier415Hz
  • 2,739
  • 6
  • 22

1 Answers1

2

This problem can be solved using separation of variables. Setting $u=X(x)T(t)$ gives $$ \frac{T''}{T}=\lambda = \frac{X''}{X},\;\; X'(0)=X'(\pi)=0. $$ The solutions are $X_n(x)=\cos(n x),\;T_n(t)=A_n\cos(n t)+B_n\sin(n t)$, which correspond to $\lambda_n=-n^2$ for $n=0,1,2,3,\cdots$. So the general solution may be written as $$ u(t,x)=\sum_{n=0}^{\infty}(A_n\cos(n t)+B_n\sin(n t))\cos(n x). $$ The initial conditions are $$ u(0,x)=\cos(x),\;\; u_{t}(0,x)=-\cos(x). $$ The first condition determines the $A_n$'s and the second determines the $B_n$'s: $$ \cos(x)=\sum_{n=0}^{\infty}A_n\cos(n x),\;\; \cos(x)=\sum_{n=1}^{\infty}B_nn\cos(n x) $$

Disintegrating By Parts
  • 87,459
  • 5
  • 65
  • 149
  • Disintegration by Parts! What an honor to see you again. Thanks fort his. I will study your solution. But can you show me how you got $T_n(t)$? – Luthier415Hz Mar 22 '22 at 07:23
  • @Luthier : Good to hear from you again! The separation constant $\lambda$ is determined by the $X$ equation and the constraints $X'(0)=0=X'(\pi)$, whose solutions are $X_n(x)=\cos(n\pi x)$, $n=0,1,2,3,\cdots$, which all have $0$ derivatives at the endpoints $0,\pi$. The corresponding $\lambda_n$ values are $\lambda_n=X_n''/X_n=-n^2\pi^2$. So, the $T_n$ solutions must satisfy $T_n''=-n^2\pi^2 T_n$, without any endpoint restrictions at this point. The solutions are $T_n=A_n\cos(n\pi t)+B_n\sin(n\pi t)$. – Disintegrating By Parts Mar 22 '22 at 07:55
  • Hmm, I will give this a try! Thanks for this! – Luthier415Hz Mar 22 '22 at 07:58
  • 1
    @Luthier : If you have any other questions, feel free to ask, my friend. – Disintegrating By Parts Mar 22 '22 at 08:00
  • @Luthier : You, as a musician, can appreciate this. https://math.stackexchange.com/questions/3285188/motivation-behind-orthogonality-of-functions/3285409#3285409 – Disintegrating By Parts Mar 22 '22 at 08:11
  • I will redo this now, and get back to you. I think what I am not seeing is caused by the fact that I have written this in a different way. I listed the solutions as components (see gather cases formula). – Luthier415Hz Mar 22 '22 at 08:11
  • Yes, I am working on some instruments too. I usually don't think in mathematical terms when making instruments, more following aesthetics actually! – Luthier415Hz Mar 22 '22 at 08:13
  • Disintegration by Parts, equation 2 in the original post has an error. It should be $\frac{1}{X}F_{xx}=k^2$ and $\frac{1}{T}G_{tt}=k^2$ – Luthier415Hz Mar 22 '22 at 08:20
  • @Disintegration by Parts, have a look at the update. – Luthier415Hz Mar 22 '22 at 09:26
  • @Luthier : I don't believe exponentials in $t$. I believe sinusoids in time and space. It should be sinusoids in time and space, which is what $\frac{X''}{X}=-k^2=\frac{T''}{T}$ would give you. – Disintegrating By Parts Mar 22 '22 at 23:06
  • I see, you mean that it is not naturally to have an exponential behaviour in time? Can I switch that then to $u(x,t)=\cos t e^{-x}$ ? – Luthier415Hz Mar 23 '22 at 09:42
  • Where do you live? "Over the hill". Haha, ingenious!! – Luthier415Hz Mar 23 '22 at 09:43
  • The thing is that I cannot find a comprehensive explanation of your Neumann condition method, I have looked here too, but it is not easy to find. Can you elaborate more on your solution, so it is easy to see for those who live "in the valley" ? I have Boas M, "Mathematical Methods in Physical Sciences" but it does not show this. It rather shows my suggestion, which is to be honest, much more simple and intuitive. – Luthier415Hz Mar 23 '22 at 09:46
  • @Luthier : Sorry, I corrected what I had. I had started mixing up constants, and made the whole solution incomprehensible. Hopefully I have it right now. Too many long nights up caring for my Mother who has dementia. – Disintegrating By Parts Mar 23 '22 at 10:47
  • Thanks for that DISINTEGRATING BY PARTS!! – Luthier415Hz Mar 23 '22 at 10:51