Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

On the oscillation problem of a rope with fixed extremities, $$\left\{\begin{matrix} \left.\begin{matrix}\left.\begin{matrix} u_{tt}(t,x) = a^2u_{xx}(t,x)\\ u(0,x) = \varphi(x)\\ u_t(0,x) = \psi(x) \end{matrix}\right\}\end{matrix}\right|\; \left\{\begin{matrix} a > 0\\ t > 0\\ 0 < x < L \end{matrix}\right.\\ u(t,0) = u(t,L) = 0 \;|\; t \geq 0 \end{matrix}\right.\qquad,$$

we can show, by assuming that $u$ is separable (id est, $u(t,x) \equiv X(x)T(t)$ for some pair of function $X$ and $T$) and writing

$$\frac{\ddot{T}(t)}{a^2T(t)} = \frac{\ddot{X}(x)}{X(x)} := -\lambda \in \mathbb{R}\;,$$

that $\lambda_n = (n\pi/L)^2$. Therefore,

$$\left.\begin{matrix} u(t,x) = \sum_{n=1}^{\infty} \left(A_n \cos\frac{n\pi a}{L}t + B_n \sin\frac{n\pi a}{L}t\right)\sin\frac{n\pi}{L}x \end{matrix}\;\right|\; A_n \wedge B_n \in \mathbb{R}\quad,$$

and setting $t = 0$ gives

$$\left.\begin{matrix}\left.\begin{matrix} \varphi(x) = \sum_{n=1}^{\infty} A_n \sin\frac{n\pi}{L}x \\ \psi(x) = \sum_{n=1}^{\infty} \frac{n\pi a}{L}B_n\sin\frac{n\pi}{L}x \\ \end{matrix}\right\}\end{matrix}\right| \;0 \leq x\leq L\quad,$$

which means

$$\left\{\begin{matrix} A_n = \frac{2}{L}\int_{0}^{L} \varphi(x)\left(\sin\frac{n\pi}{L}x\right)dx \\ B_n = \frac{2}{n\pi a}\int_{0}^{L} \psi(x)\left(\sin\frac{n\pi}{L}x\right)dx \\ \end{matrix}\right.\quad.$$

When we add viscosity — $f_v(t,x) = -\beta u_t(t,x)$ with $\beta > 0$ —, the first equation of the problem becomes

$$u_{tt}(t,x) + \beta u_t(t,x)= a^2u_{xx}(t,x)\;,$$


$$\frac{\ddot{T}(t)}{a^2\beta T(t)} + \frac{\dot{T}(t)}{a^2T(t)} = \frac{\ddot{X}(x)}{\beta X(x)} := -\lambda \in \mathbb{R}$$

gives $\lambda_n = (n\pi/L)^2/\beta$. This means

$$u(t,x) = \sum_{1 \leq n < \frac{\beta L}{2\pi a}} \left(A_n \exp\left\{\!\!\frac{\beta}{2} \left[\sqrt{1 - \left(\frac{2n \pi a}{\beta L}\right)^{\!\!2}} - 1\right] t\right\}\right. +$$ $$+\; \left.B_n \exp\left\{\!\!- \frac{\beta}{2} \left[\sqrt{1 - \left(\frac{2n \pi a}{\beta L}\right)^{\!\!2}} + 1\right]t\right\}\right) \sin\frac{n\pi}{L}x +$$

$$+\, \mathbb{I}_{n = \frac{\beta L}{2\pi a}} (C_n + D_nt) \exp\left(\!\!-\frac{\beta}{2}t\right) \sin\frac{\beta}{2a}x +$$

$$+ \sum_{n > \frac{\beta L}{2\pi a}} \left[E_n \cos\frac{\beta}{2}\sqrt{\left(\frac{2n \pi a}{\beta L}\right)^{\!\!2} \!- 1}\;t \,+\, F_n \sin\frac{\beta}{2}\sqrt{\left(\frac{2n \pi a}{\beta L}\right)^{\!\!2} \!- 1}\;t\right] \exp\left(\!\!-\frac{\beta}{2}t\right) \sin\frac{n\pi}{L}x\quad.$$

How do I calculate coefficients $A_n$ to $F_n$ in terms of $\varphi(x)$ and $\psi(x)$?

share|cite|improve this question
$\left\{ \begin{align} a + b &= c \\ c + d &= e \\ example &= of \quad \TeX \end{align} \right\}$ – mixedmath Jun 6 '11 at 1:14
@Luke: no problem! I learned that here too, actually. – mixedmath Jun 6 '11 at 15:58
@Luke: good theory! That looks even better, you're right. Good catch. – mixedmath Jun 6 '11 at 21:53
I think your expression for $\psi(x)$ in the damped case is wrong. The separation leads to a damped oscillation equation, $\ddot T+\beta\dot T+(n\pi a/L)^2T=0$, with characteristic values $(\beta\pm\sqrt{\beta^2-(2n\pi a/L)^2})/2$. In any case, taking $\beta$ to $0$ should recover the undamped case, and there should be only one free coefficient per eigenmode; neither is the case in your expression. – joriki Jun 10 '11 at 2:07
@Luke: Yes, but $\exp(-\frac{\beta}{2}(1+\sqrt{\phantom{\cdot}})t)=\exp(-\frac{\beta}{2}t)\exp(-‌​\frac{\beta}{2}\sqrt{\phantom{\cdot}} t)$, not $\exp(-\frac{\beta}{2}t)\exp(\sqrt{\phantom{\cdot}}t)$. I see now that the non-oscillating solutions are also wrong, for a similar reason: $\exp(-\frac{\beta}{2}(1+\sqrt{\phantom{\cdot}})t)\neq\exp(-\frac{\beta}{2}t)\ex‌​p((1+\sqrt{\phantom{\cdot}})t)$. (By the way, I only get notified of your comments if you ping me ("@joriki").) – joriki Jun 11 '11 at 7:06
up vote 3 down vote accepted

Basically, you can calculate the coefficients just like you calculated them in the undamped case: Substitute $t=0$ into both $u$ and $u_t$ and Fourier-transform. The only difference is that in the undamped case you'd already chosen the coefficients such that they were decoupled and each individually was given by a Fourier integral; now, linear combinations, each involving two coefficients, are given by Fourier integrals, and you need to solve the resulting $2\times2$ systems of linear equations to get the individual coefficients.

Is that enough, or do you want more details?

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.