Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Another qual problem that is causing me some difficulty...

Consider the PDE $$ \left\{\begin{array}{rl} u_{xxt}+u_{xx}-u^{3}=0&\text{in}\;[0,1]\times(0,\infty)\\ u(0,t)=u(1,t)&\text{on}\{x=0\}\times(0,\infty)\bigcup\{x=1\}\times(0,\infty)\\ u(x,0)=g(x).\end{array}\right. $$ Take $g(x)=x(x-1)$ and show that solutions tend to zero uniformly in $t$ as $t\infty$.

To solve the problem I used the usual energy argument, multiplying the PDE by $u$ and integrating both sides in space (applying to the periodic boundary conditions to eliminate boundary terms from the integration by parts) in order to easily obtain $$\frac{de}{dt}(t):=\frac{d}{dt}\int\limits_{0}^{1}|u_{x}(x,t)|^{2}\;dx=-2e(t)-2\int\limits_{0}^{1}|u(x,t)|^{4}\;dx\leq0$$ so that the energy decreases. Then, using the initial condition, I also have $$0\leq e(t)\leq e(0)=\frac{1}{3}$$ for all $t>0$. So I have a fairly sharp bound and the fact that $e(t)$ is decreasing in $t$. If I could show further that $e(t)$ actually tends to $0$ uniformly, then that would imply $u_{x}(x,t)\to0\;\text{as}\;t\to\infty$ uniformly and that would give me $u(x,t)\to\;\text{C}\;\text{as}\;t\to\infty$. Applying the periodic boundary conditions once more would then give me $C=0$ as required.

However, I do not know how to show $e(t)\to0$ as $t\to\infty$. I've tried computing additional derivatives of $e$ to see if I could glean any additional information on the decay of $e$, but the computations did not seem to lead to anything fruitful (even after playing around/substituting terms using the PDE).

Any hints are appreciated (a fully fleshed answer is not needed). Thanks!

share|improve this question
    
Thanks for pointing that out - it should have been $=u^{3}$ in the equation. –  Taylor Feb 24 '13 at 8:14
add comment

2 Answers

up vote 0 down vote accepted

Hint: You can rewrite

$$ u_{xxt} + u_{xx} = e^{-t} \partial_t (e^t u_{xx}) $$

by integrating factors. So let $v = e^t u$ and $v_{xx} = e^{t} u_{xx}$ you can rewrite the equation in terms of $v$. If you can show that $v$ has bounded energy, then since the energy of $v$ is $e^{2t}$ times that of $u$, the energy of $u$ must decay exponentially.

share|improve this answer
    
Thanks for the tip. It is easy to see the energy of $v$ is bounded since a similar computation shows $\frac{d}{dt}E_{v}(t)=-2e^{-2t}\int_{0}^{1}v^{4}dx\leq0$ (interesting that the middle term $u_{xx}=e^{-t}v_{xx}$ drops out in the PDE) and $0\leq E_{v}(t)\leq E_{v}(0)=e^{2t}\int_{0}^{1}|u_{x}(x,t)|^{2}dx\bigg|_{t=0}=E_{u}(0)=\frac{1}{3}$ as before. Is there any methodology you used in order to come up with this trick/substitution? I've had to use a similar substitution in a problem dealing with a modified diffusion equation, but sadly such ideas don't often come to me.... –  Taylor Feb 24 '13 at 8:38
1  
"Is there any methodology..." It is called heuristics. It you spend a few more years looking at PDEs you'll also pick up these types of intuitions. –  Willie Wong Feb 24 '13 at 22:39
add comment

You don't just have $\dfrac{de}{dt} \le 0$, you have $\dfrac{de}{dt} \le -2e$. That (with $e > 0$) implies $e(t)$ decreases at least as fast as the solutions to $\dfrac{dy}{dt} = - 2y$, which are $C \exp(-2t)$. In fact, $$\dfrac{d}{dt} (e(t) \, \exp(2t)) = \exp(2t) \left(\dfrac{de}{dt} + 2 e \right) \le 0$$ so $e(t) \le e(0) \exp(-2t)$.

share|improve this answer
    
Good point. I actually noticed this, and in retrospect should have applied Gronwall's inequality. –  Taylor Feb 24 '13 at 9:38
add comment

Your Answer

 
discard

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.