Homework Help: Separation of variables

1. Nov 13, 2013

sarahqwert

1. The problem statement, all variables and given/known data

utt = uxx -(25/4)cos((5/2)x)
ux(0,t) =1
u(pi,t)= pi
u(x,0)=x
ut(x,0)=0

2. Relevant equations

u(x,t)=v(x) + w(x,t)

3. The attempt at a solution

This is what I did so far:

u(x,t)=v(x) + w(x,t)
u(x,0) = v(x) +w(x,0)

when t is large:
vxx - (25/4)cos((5/2)x) = 0
vx = (5/2)sin((5/2)x)
v(x) = -cos((5/2)x) +x

when t is not large
wtt = wxx
w(x,t)= x-v(x) = cos((5/2)x)

and I'm not sure what to do after this

Maybe: X''+ λX = 0
X'(0) = 1 and X(pi)= pi

2. Nov 13, 2013

LCKurtz

What does $t$ being large have to do with anything?

No. The point of your original substitution is to let $v(x)$ take care of the non-homogeneous terms. I will call $f(x) = \frac {25}4\cos(\frac 5 2 x)$ to save typing. What you want to do is substitute $u(x,t) = v(x) + w(x,t)$ into the equation $u_{tt}=u_{xx}-f(x)$ and its boundary conditions:

$u_{tt}=u_{xx}-f(x)$ becomes $w_{tt} = w_{xx}+v''(x) -f(x)$ That will be satisfied if we take $v''(x) = f(x)$ and $w_{tt} = w_{xx}$

Now $u_x(0,t) = 1$ becomes $w_x(0,t) + v'(0) = 1$. That will be satisfied if we take $v'(0)=1$ and $w_x(0,t)=0$.

$u(\pi,t)=\pi$ becomes $w(\pi,t) +v(\pi)=1$. That will be satisfied if we take $v(\pi)=\pi$ and $w(\pi,t) = 0$.

Now with those conditions on $v(x)$ you should be able to solve for $v(x)$ and you have a homogeneous equation and boundary conditions in $w$. Try to take it from there.

3. Nov 14, 2013

sarahqwert

Thank you for your help, I m almost done solving this problem and got:

u(x,0)= Ʃ (cos((2n+1)/2)x)*Dn = x

sigma from 0 to infinity and Dn is a constant.

How do I find Dn?

4. Nov 14, 2013

LCKurtz

That isn't correct. You need to show your work so I can see where you fell off the tracks.

Remember, when you substitute in the initial condition you get$$x=u(x,0) = v(x) + w(x,0)$$so your initial condition for $w(x,t)$ is $w(x,0)=x - v(x)$. What did you get for $v(x)$?

Also, did you remember to check whether the eigenvalue zero works?

[Edit: Added] Your focus should be on solving the $w(x,t)$ system completely. That's where I need to see your work. Once you have figured that out, all you have to do to solve the $u$ system is write $u(x,t) = v(x) + w(x,t)$ because you know both $v$ and $w$.

Last edited: Nov 14, 2013
5. Nov 14, 2013

sarahqwert

Sorry that was supposed to be w(x,t) not u(x,t). So:

Xn=C*cos((2n+1)/2)x
Tn=D*cos((2n+1)/2)t)+(B*sin((2n+1)/2)t)

w(x,t) = Ʃ Xn * Tn

w(x,t) = Ʃcos((2n+1)/2)x)*((D*cos((2n+1)/2)t)+(B*sin((2n+1)/2)t))

so

w(x,0)=Ʃcos((2n+1)/2)x)*D

and

w(x,0) = cos(5/2)x

Last edited: Nov 14, 2013
6. Nov 14, 2013

LCKurtz

You need subscripts on the constants, as $D_n$ and $B_n$. You didn't answer some of my questions in post #4. What did you get for $v(x)$? Did you check for zero eigenvalues? (This matters as to whether there is a constant term). What does the initial condition for $w(x,0)$ become?

[Edit: added] Also you need to show what your T(t) eigenvalue problem becomes, with boundary condition, and how you got $T_n(t)$, because what you have isn't correct.

Last edited: Nov 14, 2013