The wave equation with transverse force is given as following:

$u_{tt}-c^2u_{xx}+ku=0$ $(k>0)$

Define the total energy $$E(t)=\frac{1}{2} \int_{-\infty}^{\infty}((u_t)^2+c^2(u_x)^2)dx$$ There are three assumptions:

  1. $u(x,0)=0$ for all $-\infty < x < \infty$
  2. $\lim\limits_{\vert x \vert \rightarrow \infty}u_t(x,0)=0$
  3. $\lim\limits_{\vert x \vert \rightarrow \infty}u(x,t)=0$ for all $t>0$

I want to determine the sign of $E(\tau)-E(0)$ for some $\tau>0$.

Observe that $$\frac{d}{dt}E(t)=\frac{d}{dt}\frac{1}{2} \int_{-\infty}^{\infty}((u_t)^2+c^2(u_x)^2)dx=\frac{1}{2} \int_{-\infty}^{\infty}\frac{\partial}{\partial t}((u_t)^2+c^2(u_x)^2)dx$$$$=\int_{-\infty}^{\infty}(u_tu_{tt}+c^2u_xu_{xt})dx$$ Then by substituting $u_{tt}=c^2u_{xx}-ku$, we have $$\frac{d}{dt}E(t)=\int_{-\infty}^{\infty}((c^2u_{xx}-ku)u_t+c^2u_xu_{xt})dx$$$$=c^2\int_{-\infty}^{\infty}(u_{xx}u_t+u_xu_{xt})dx-\int_{-\infty}^{\infty} kuu_t \ dx$$$$=c^2\int_{-\infty}^{\infty}\frac{\partial}{\partial x}(u_xu_t)dx-\int_{-\infty}^{\infty} kuu_t \ dx$$$$=c^2(u_xu_t)\Big\vert_{x=-\infty}^{x=\infty}-\int_{-\infty}^{\infty} kuu_t \ dx$$$$=-\int_{-\infty}^{\infty} kuu_t \ dx$$ Thus, $$E(\tau)-E(0)=-\int_0^\tau \int_{-\infty}^{\infty} kuu_t \ dx dt$$ By Fubini's theorem (but I'm not sure that $\int_{-\infty}^{\infty} kuu_t \ dx$ converges), $$E(\tau)-E(0)=-k\int_{-\infty}^{\infty}\int_0^\tau uu_t \ dt dx$$ Here, by integration by parts, $$\int_0^\tau uu_t \ dt=u^2\Big\vert_{t=0}^{t=\tau}-\int_0^\tau uu_t \ dt$$$$=u^2(x,\tau)-\int_0^\tau uu_t \ dt$$ Thus $$E(\tau)-E(0)=-k\int_{-\infty}^{\infty}\frac{1}{2}u^2(x,\tau) dx<0$$

I am curious about three parts:

  1. Does the third assumption imply $(u_xu_t)\Big\vert_{x=-\infty}^{x=\infty}=0$?
  2. Can I justify that Fubini's theorem holds for $\int_0^\tau \int_{-\infty}^{\infty} kuu_t \ dx dt=\int_{-\infty}^{\infty}\int_0^\tau kuu_t \ dt dx$?
  3. Is the second assumption necessary? (It is mentioned in my textbook, but I didn't use it in my proof.)

I would appreciate it if someone gives me any intuitive answer.

  • 1
    $\begingroup$ Intuitively from a physics perspective : If your system has the extra transverse force term $ku$ there will also be potential Energy coming from that. It can't be true that the total conserved energy is the same as if $k=0$. $\endgroup$
    – Kurt G.
    Mar 27 at 7:37
  • $\begingroup$ Related question $\endgroup$
    – Dan Doe
    Mar 29 at 17:06

1 Answer 1

  1. In general $\lim_{x \to \pm \infty} u (x,t) = 0$ does not imply that $\lim_{x \to \pm \infty} u_x (x,t) = 0$ holds true. This is nicely explained in this post. In particular, $u_x(x, t)$ might approach $\pm \infty$ (or oscillate between $\infty$ and $-\infty$) for $x \to \pm \infty$. A good example is $u(x, t) = \sin\big(x^3\big) / x $ with $u_x(t, t) = \cos\big(x^3\big) 3x - \sin(x^3)/x^2$.
  2. You already provided the link to Wikipedia, where it is stated that $$\int_0^\tau \int_{\mathbb{R}} \vert k u u_t \vert \mathrm d(x, t) \overset{!}{<} \infty$$ has to be fulfilled to allow the application of Fubini's theorem. You know that $\lim_{x \to \pm \infty} u(x, t) = 0$ holds for all $t$, so this is no problem. Intuitively, it is somewhat clear that $\lim_{x \to \pm \infty} u(x, t) = 0 \: \forall t >0 $ implies that also $\lim_{x \to \pm \infty} u_t(x, t) = 0 \: \forall t >0 $, since for $\vert x \vert$ large enough $u$ effectively no longer varies with $t$ anymore. A rigorous proof would require that you can interchange the limits $$ \lim_{x \to \pm \infty} \lim_{h \to 0} \underbrace{\frac{u(x, t+h) - u(x, t)}{h}}_{=:g(x, h)},$$ which can be done if the conditions of the Moore-Osgood theorem are met. See one of these [1], [2] resources for a more in-depth treatment. So in our case, you need that

$\lim_{x \to \pm \infty} g(x, h) $ converges uniformly for $h \neq 0$ and $\lim_{h \to 0} g(x, h)$ converges point-wise for $x \neq \pm \infty$.

If this condition is met, you can interchange the limits and obtain that $u_t(x \to \pm \infty, t) = 0$. In this case, the integral is bounded and you are good with applying Fubini.

  1. I see also no immediate consequence because of the second statement. In contrast, I see the difficulties with some missing information on the derivatives $u_t, u_x$. Assuming e.g. compact support of $u(t, x)$ would solve these, for instance.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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