Insulated endpoint conditions(Neumann BVP)

Now consider the BVP.

$\displaystyle u_t$ $\displaystyle =$ $\displaystyle k u_xx~~0< x<L,$ (3.21)
$\displaystyle u_x(0,t) =u_x(L,t)$ $\displaystyle =$ $\displaystyle 0,$ (3.22)
$\displaystyle u(x,0)$ $\displaystyle =$ $\displaystyle f(x)$ (3.23)

which corresponds to heat distribution in a thin (1D) rod of length L with initial temperature $f(x)$ and its endpoints insulated.
Figure:
\includegraphics[width=15.0cm, height=2.0cm]{heat2.pdf}

As in the previous case,follow the method of separation of variables.

u(x,t)=X(x)T(t)
$\frac{X''}{x} = \frac{T'}{kt} = \lambda$
If $u(x,t)=X(x)T(t)$ is solution of (19), then it must satisfy B.C.s given in Eq. ([*]) and Eq. ([*]), then

$\displaystyle X'(0)T(t)=X'(L)T(t)=0, X'(0)=0$    and $\displaystyle X'(L)=0$

Thus $X(x)$ satisfies

$\displaystyle X''-\lambda X=0,X'(0)=0~~X'(L)=0.$ (3.24)

Again we have to consider three possible $\lambda>0,\lambda=0,\lambda<0.$

  1. If $\lambda >0 $

    $\displaystyle X''=\lambda X$ (3.25)
    $\displaystyle X(x)=\alpha exp( \sqrt{\lambda}x )+\beta exp(-\sqrt{\lambda}x)$ (3.26)
    $\displaystyle X(x)=\alpha cosh( \sqrt{\lambda}x )+\beta sinh(\sqrt{\lambda}x)$ (3.27)
    $\displaystyle X'(x)=\sqrt{\lambda}\alpha exp( \sqrt{\lambda}x )-\sqrt{\lambda}\beta exp(-\sqrt{\lambda}x)$ (3.28)

    using B.C's
    $\displaystyle X'(0)$ $\displaystyle =$ $\displaystyle \sqrt{\lambda}\alpha-\sqrt{\lambda}\beta$ (3.29)
    $\displaystyle X'(x)$ $\displaystyle =$ $\displaystyle \sqrt{\lambda}\alpha exp( \sqrt{\lambda}
L)-\sqrt{\lambda}\beta exp(-\sqrt{\lambda}L)
\alpha exp( \sqrt{\lambda}x )-\beta exp(-\sqrt{\lambda}x)=0$ (3.30)

    from $\alpha=\beta$

    $\displaystyle \alpha (exp( \sqrt{\lambda}x )- exp(-\sqrt{\lambda}x))=0$

    This bracket term never be zero for $\lambda =0.$ Which implies that $\alpha=0$ and $\beta=0$ and we get a trivial solution.Thus we can have positive eigenvalues. Which means $\lambda \not>0.$

  2. If $\lambda=0,$

    $\displaystyle X''=0\implies X'= \alpha $

    $\displaystyle X=\alpha x+\beta$

    $\displaystyle X'(0)=0\implies \alpha =0$

    $\displaystyle \implies X=\beta,\beta \neq 0$

    else we get, trivial solution. As a constant multiple of an eigen-functions is also function,we choose $\beta=1.$ Thus for zero eigenvalues $0,$ we have $\lambda_0=0$, $X_0 (x)=1$

  3. If $\lambda < 0,$

    Replace $\lambda$ by $-\lambda$ in Eq. ([*])

    $\displaystyle X''+\lambda X$ $\displaystyle =$ 0 (3.31)
    $\displaystyle X$ $\displaystyle =$ $\displaystyle \alpha \cos(\sqrt{\lambda}x)+\beta \sin(\sqrt{\lambda}x)$ (3.32)
    $\displaystyle X'$ $\displaystyle =$ $\displaystyle \sqrt{\lambda}\alpha \sin(\sqrt{\lambda}x)+\sqrt{\lambda}\beta \cos(\sqrt{\lambda}x)$ (3.33)

    $\displaystyle X'(0)=0\implies -\lambda \alpha . 0+\sqrt{\lambda}\beta .1=0$

    $\displaystyle \beta =0$

    $\displaystyle X'(L)=0\implies-\sqrt{\lambda}sin(\sqrt{\lambda}L)=0$

    $\alpha \neq 0$,  else we get the trivial solution

    $\displaystyle \implies\sqrt{\lambda}L=n \pi,~~n=1,2,3,...$

    $\displaystyle \lambda_n=\frac{n^2 \pi^2}{L^2}$

    are the eigenvalues and corresponding eigen-functions are

    $\displaystyle X_n(x)=\alpha_n \cos(\frac{n\pi}{L}x)$

    we can combine both the case as

    $\displaystyle X_n(x)=\alpha_n \cos(\frac{n\pi}{L}x)~~n=0,1,2,...$

Solving for $T(t)$: For $\lambda =0 $

$\displaystyle T'(t)=0$

$\displaystyle \implies T(t)=\gamma _n exp(\frac{-n^2\pi^2}{L^2}kt)~~n=1,2,...$

we can combine the choose two cases as

$\displaystyle T_n(t)=\gamma_n exp(-\frac{n^2\pi^2}{L^2}kt)~~n=0,1,2,...$

we obtain sequence of separated solutions.Thus
$\displaystyle u_n(x,t)$ $\displaystyle =$ $\displaystyle ((\alpha_n \cos(\frac{n\pi}{L}x)(\gamma_n \exp(-\frac{n^2\pi^2}{L^2}kt))~~n=0,1,2,...$ (3.34)
  $\displaystyle =$ $\displaystyle \chi_n cos(\frac{n\pi}{L}x)( \exp(-\frac{n^2\pi^2}{L^2}kt)$ (3.35)

By the principle of superposition of solutions,we get the general solution
$\displaystyle u(x,t)$ $\displaystyle =$ $\displaystyle \sum_{n=0}^{\infty}\chi_n u_n(x,t)$ (3.36)
  $\displaystyle =$ $\displaystyle \sum_{n=0}^{\infty}\chi_n \cos(\frac{n\pi}{L}x)( exp(-\frac{n^2\pi^2}{L^2}kt)$ (3.37)

$\displaystyle u(x,t)=\chi_0 +\sum_{n=1}^{\infty}\chi_n \cos(\frac{n\pi}{L}x)( \exp(-\frac{n^2\pi^2}{L^2}kt)$ (3.38)

The unknown coefficients ${\chi_n}$ are found by using the initial condition

$\displaystyle u(x,0)=\chi_0 +\sum_{n=1}^{\infty}\chi_n \cos(\frac{n\pi}{L}x)=f(x)$ (3.39)

multiply both sides by $\cos(\frac{n\pi}{L})x$ and integrate both sides $[0,L]$

$\displaystyle \int_{0}^{L}\chi_0 \cos(\frac{n\pi}{L}x dx +\int_{0}^{L}\sum_{n=1...
...frac{m\pi}{L}x)\cos(\frac{n\pi}{L})xdx=\int_{0}^{L}f(x)\cos(\frac{n\pi}{L})x dx$ (3.40)

Now we use the orthogonality property
$\int_{0}^{L}\cos(\frac{m\pi}{L}x)\cos(\frac{n\pi}{L}x dx=0~~m\not =n$ $\frac{L}{2}~~m=n \not =0$ L   m=n=0

$\displaystyle L\chi_0=\int_{0}^{L}f(x) dx$

$\displaystyle \chi_0=\frac{\int_{0}^{L}f(x) dx}{L}$

For m=1 

$\displaystyle \chi_1\frac{L}{2}=\int_{0}^{L}f(x)\cos(\frac{2\pi}{L}x dx$

$\displaystyle \chi_1=\frac{2}{L}\int_{0}^{L}f(x)\cos(\frac{2\pi}{L}x dx$

$\displaystyle \chi_n = \frac{2}{L} \int_{0}^{L} f(x)\cos( \frac{n\pi}{L} x dx$ (3.41)

Thus ${\chi_n}$ are the Fourier-cosine series coefficients of $f(x)$ on $[0,L]$ and the formal series solution of the model problem Eq. ([*]),Eq. ([*]) and Eq. ([*]) is

$\displaystyle u(x,t)=\chi_0 +\sum_{n=1}^{\infty} \chi_n \cos( \frac{n\pi}{L} x)\exp(- \frac{n^2\pi^2}{L^2} kt)$ (3.42)

where ${\chi}$ are coefficients of Fourier cosine expansion of initial condition $f(x)$ on the initial $[0,L]$.

Remark Note that $\lim_{n\to\infty}$u(x,t)=$\chi_0$= $\frac{1}{L}$ $\int_{0}^{L}f(x)dx=$ average values of the initial temp. Thus with lateral surfaces and end points insulated,the initial heat content ultimately distributes itself uniformly throughout the rod and attains the average values $\chi_0.$

Illustration

We consider the same $50cm$ rod as in the previous example with the following modifications

  1. At time $t=0$, the rod's lateral surfaces and its two ends are insulated.
  2. The initial temp. profile is given in angular function.

The model equation for this setting for $~0< x<L$ is
$\displaystyle u_t$   $\displaystyle = k u_{xx},$ (3.43)
$\displaystyle u_x(0,t) =u_x(50,t)$   $\displaystyle = 0,$ (3.44)
$\displaystyle u(x,0)$   $\displaystyle = f(x)$ (3.45)

where

\begin{displaymath}f(x) =
\begin{cases}
4x &0\leq x \leq 25\\
100-4x& 25\leq x \leq 50
\end{cases}\end{displaymath} (3.46)

\includegraphics[width=10cm, height=7.0cm]{iamge1.pdf}

Solution For this Neumann problem, the solution is given by

$\displaystyle u(x,t)=\frac{\chi_0}{2} +\sum_{n=1}^{\infty}\chi_n cos(\frac{n\pi}{L}x)exp(-\frac{n^2\pi^2}{L^2}kt)$ (3.47)

where ${\chi_n}$ are Fourier coefficients of Fourier-cosine expansion of the initial temp. profile $f(x)$ in the interval $[0,L=50].$ These coefficients are given by
$\frac{\chi_n}{2}=\frac{1}{L}\int_{0}^{L}f(x)cos(\frac{n\pi}{L}x dx~n=0,1,2,...$
Thus to find the required solution,all we are do is find Fourier-cosine expansion of $f(x).$

$\displaystyle \frac{\chi_n}{2} =\frac{1}{L}\int_{0}^{50}f(x)\cos(\frac{n\pi}{L}...
...)\cos(\frac{n\pi}{L}x dx+\int_{25}^{50}(200-4x)\cos\big(\frac{n\pi}{L}x dx\big)$

$\displaystyle f(x)=50-\frac{1600}{\pi^2}\sum_{n=1}^{\infty}\frac{1}{(n)^2} \cos...
...{\pi^2}\sum_{n=1}^{\infty}\frac{1}{(4n-2)^2}\cos\big(\frac{(4n-2)\pi}{50}x\big)$

$\displaystyle u(x,t)=50-\frac{1600}{\pi^2}\sum_{n=2,6,10}^{}\frac{1}{(n)^2} cos\big(\frac{n\pi}{50}\big)x\exp\big(\frac{-n^2\pi^2}{2500}kt\big)$