Toronto Math Forum

APM346-2012 => APM346 Math => Home Assignment 4 => Topic started by: Victor Ivrii on October 20, 2012, 06:28:59 AM

Title: Bonus Web Problem--2
Post by: Victor Ivrii on October 20, 2012, 06:28:59 AM
Oscillations of the beam  (with both its ends  having fixed positions but not directions, imagine beam lying on supports) are described by an equation
\begin{equation*}
u_{tt} + K u_{xxxx}=0, \qquad 0<x<l
\end{equation*}
with $K>0$ and the boundary conditions
\begin{equation*}
u(0,t)=u_{xx}(0,t)=u(l,t)=u_{xx}(l,t)=0.
\end{equation*}


Compare with eigenvalues of Problem 2 of HA2
Title: Re: Web Bonus Problem--3
Post by: Victor Ivrii on October 27, 2012, 02:19:01 PM
Actually my comments are related to this problem (so I shuffled posts -- splitting and merging)


 One can see easily the relation to $X''+\lambda X=0$ with the Dirichlet b.c.: the same eigenfunctions and squared eigenvalues. Reason is simple: if $A$ denotes $\frac{d^2\ }{dx^2}$ with Dirichlet b.c. then our operator $B$ is just $A^2$ including b.c.: $u\in \mathfrak{D}(B)$  iff $u\in \mathfrak{D}(A) $ and also $Au\in \mathfrak{D}(A) $. Here $\mathfrak{D}(A) $ denotes domain of definition of $A$.
Title: Re: Re: Web Bonus Problem--3
Post by: Calvin Arnott on October 27, 2012, 05:10:45 PM
Hmm, interesting. It seems as if that perspective gives some insight as to why the result was an even and odd eigenfunction for the respective problems. What does the generalization to this problem in several spacial dimensions look like?
Title: Re: Re: Web Bonus Problem--3
Post by: Victor Ivrii on October 27, 2012, 07:23:46 PM
Hmm, interesting. It seems as if that perspective gives some insight as to why the result was an even and odd eigenfunction for the respective problems. What does the generalization to this problem in several spacial dimensions look like?

Oscillations of membranes is described by $u_{tt}-K\Delta u=0$ while oscillations by plates by
$u_{tt}+K\Delta^2 u=0$.

The simplest b.c. for membrane are Dirichlet $u|_S=0$ and Neumann $\partial_nu|_S=0$ where $n$ is a normal to the boundary which is $S$; so $\partial_n u:= \sum_i u_{x_i}n_i$.

The simplest b.c. for plate  are Dirichlet $u|_S=\partial_n u |_S=0$ and free
$\sum_{i,j} u_{x_ix_j}n_in_j|_S =\partial_n\Delta u|_S=0$.

THis particular problem generalizes to $u|_S=\Delta u|_S=0$ and again we have a square of Laplacian.
Title: Re: Re: Web Bonus Problem--3
Post by: Calvin Arnott on October 27, 2012, 07:43:30 PM
Suppose we somehow have several time-like variables. Does it make sense to consider this kind of system? How would we represent such a structure in the language of equations we're developing in this course?- perhaps in this specific case a Laplacian on time as well? What kind of properties do systems of this kind have that don't appear with a single time-like variable?
Title: Re: Re: Web Bonus Problem--3
Post by: Victor Ivrii on October 28, 2012, 12:17:05 AM
Suppose we somehow have several time-like variables. Does it make sense to consider this kind of system? How would we represent such a structure in the language of equations we're developing in this course?- perhaps in this specific case a Laplacian on time as well? What kind of properties do systems of this kind have that don't appear with a single time-like variable?


There are so called ultra-hyperbolic equations http://forum.math.toronto.edu/index.php?topic=37.msg176#msg176 (http://forum.math.toronto.edu/index.php?topic=37.msg176#msg176) but they are much less developed and needed
Title: Re: Bonus Web Problem--2
Post by: Victor Ivrii on October 28, 2012, 08:26:09 AM
So, when we have this problem we can see easily that general solution
$$
X= A\cosh (cx)+  B\cos (cx) +C\sinh (cx)  +D\sin (cx)
$$
satisfies conditions at $0$ as $A=B=0$ as
$$
c^{-2}X''= A\cosh (cx)-  B\cos (cx) +C\sinh (cx)  -D\sinh (cx)
$$
and therefore
$$
X=C\sinh (cx)  +D\sinh (cx)
$$
satisfies conditions at $0$ as $A=B=0$ as
$$
c^{-2}X''= C\sinh (cx)  -D\sin (cx).
$$
To satisfy conditions on the right end we need
\begin{align*}
& C\sinh (cl)  +D\sin  (cl) =0,\\
& C\sinh (cl)  -D\sin (cl) =0
\end{align*}
which is possible for $(X,D)\ne 0$ iff $C=0$, $\sin (cl)=0$ i.e. we get $\lambda_n= n^2 \pi^2/l^2$, $X_n=\sin (n\pi x/l)$.