Let $u$ be a solution of the wave equation $u_{tt} = c^2u_{xx}$ on the domain $\Omega=(0,1)$, subject to homogeneous Neumann boundary conditions.

(a) Define $$E(t)=\int_{0}^{1}\big(u_t(t,x)^2+c^2u_x(t,x)^2\big)dx$$ Show that $E(t)$ is constant, i.e., it does not depend on the time $t$. That is, show that the derivative satisfies $dE(t)/dt = 0$.

(b) Use (a) to show that the only differentiable function satisfying the wave equation $u_{tt }= c^2u_{xx}$ such that $u(0,x) = 0$ and $u_t(0,x) = 0$ for all $x\in (0,1)$, as well as $u_x(t,0) = u_x(t,1) = 0$ for all $t > 0$ is the trivial function $u(t,x) \equiv 0$. Hint: Find $E(0)$ for a function with these initial conditions

  • $\begingroup$ Shouldn't your energy integrad have a term in $u_t$? $\endgroup$ – Robert Lewis Sep 27 '17 at 3:48
  • 1
    $\begingroup$ Yes you are correct, fixing that now, sorry. $\endgroup$ – Peetrius Sep 27 '17 at 3:50
  • $\begingroup$ I edited your question to remove a seemingly spurious "\". Also answered it! $\endgroup$ – Robert Lewis Sep 28 '17 at 2:09

In what follows, I assume without further comment that $u(t, x)$ is "sufficiently differentiable"; enough so to make the logic fly.

For part (a):


$u_{tt} = c^2 u_{xx} \tag 1$


$E(t) = \displaystyle \int_0^1 (u_t^2(t, x) + c^2 u_x^2(t, x))dx, \tag 2$

we may perform the "lite" computation of $\dot E(t) = E_t(t) = dE(t)/dt$, where I use the word "lite" since I am not going to worry about or otherwise vindicate differentiating under the integral sign with respect to $t$, an operation which would probably merit some justifying analysis and/or remarks in a more thorough treatment, as follows:

$\dfrac{dE(t)}{dt} = \dfrac{d}{dt}\displaystyle \int_0^1 (u_t^2(t, x) + c^2 u_x^2(t, x))dx= \int_0^1 \dfrac{d}{dt}(u_t^2(t, x) + c^2 u_x^2(t, x))dx$ $= \displaystyle \int_0^1 (2u_t(t, x)u_{tt}(t, x) + 2c^2u_x(t, x)u_{xt}(t, x))dx$ $ = \displaystyle 2\int_0^1 (u_t(t, x)u_{tt}(t, x) + c^2u_x(t, x)u_{xt}(t, x))dx; \tag 3$

using (1), the rightmost integral in (3) becomes

$\displaystyle \int_0^1 (u_t(t, x)u_{tt}(t, x) + c^2u_x(t, x)u_{xt}(t, x))dx$ $= \displaystyle \int_0^1 (c^2u_t(t, x)u_{xx}(t, x) + c^2u_x(t, x)u_{xt}(t, x))dx; \tag 4$


$\displaystyle \int_0^1 (c^2u_t(t, x)u_{xx}(t, x) + c^2u_x(t, x)u_{xt}(t, x))dx$ $= \displaystyle \int_0^1c^2 u_t(t, x)u_{xx}(t, x)dx + \int_0^1 c^2u_x(t, x)u_{xt}(t, x)dx; \tag 5$

as for the second integral on the right of (5), consider the identity

$(u_t(t, x) u_x(t, x))_x = u_{tx}(t, x) u_x(t, x) + u_t(t, s) u_{xx}(t, x), \tag 6$

which results from a simple application of the Leibniz rule for derivatives of products to $u_t(t, x) u_x(t, x)$; (6) may be re-arranged to yield

$u_{tx}(t, x) u_x(t, x) = (u_t(t, x) u_x(t, x))_x - u_t(t, s) u_{xx}(t, x), \tag 7$

which we integrate in $x$ over $[0, 1]$:

$\displaystyle \int_0^1 u_{tx}(t, x) u_x(t, x)dx = \int_0^1(u_t(t, x) u_x(t, x))_x dx - \int_0^1 u_t(t, s) u_{xx}(t, x)dx; \tag 8$

when $t \ge 0$, the leftmost integral on the right-hand side of (8) may be explicitly found:

$\displaystyle \int_0^1(u_t(t, x) u_x(t, x))_x dx= u_t(t, 1)u_x(t, 1) - u_t(t, 0)u_x(t, 0) = 0, \tag 9$

by virtue of the homogeneous Neumann boundary conditions $u_x(t, 1) = u_x(t, 0) = 0$, $t \ge 0$; thus, (8) becomes

$\displaystyle \int_0^1 u_{tx}(t, x) u_x(t, x)dx = - \int_0^1 u_t(t, s) u_{xx}(t, x)dx, \tag {10}$

in the light of which (5) yields

$\displaystyle \int_0^1 (c^2u_t(t, x)u_{xx}(t, x) + c^2u_x(t, x)u_{xt}(t, x))dx$ $= \displaystyle \int_0^1 c^2 u_t(t, x)u_{xx}(t, x)dx - \int_0^1 c^2 u_t(t, x) u_{xx}(t, x)dx = 0; \tag {11}$

now following the logic back from (5) to (3) we see that

$\dfrac{dE(t)}{dt} = \dfrac{d}{dt}\displaystyle \int_0^1 (u_t^2(t, x) + c^2 u_x^2(t, x))dx, \tag{12}$

for $t \ge 0$; the requisite result.

Now as for part (b):

Since $u(0, x) = 0$ we have

$u_x(0, x) = 0, \tag{13}$

and since we are given that

$u_t(0, x) = 0, \tag{14}$

we see that

$E(0) = 0; \tag{15}$

then by the result of part (a) we find

$E(t) = 0 \tag{16}$

for all $t \ge 0$. Since we (though perhaps tacitly) assume $u_t(t, x)$ and $u_x(t, x)$ are continuous, this implies

$u_t(t, x) = u_x(t, x) = 0 \tag{17}$

for $t \ge 0$; since the derivatives of $u(t, x)$ are both $0$, $u(t, x)$ is constant in both time and space for $t \ge 0$; since $u(x, 0) = 0$ it follows that

$u(x, t) = 0 \; \text{for} \; (x, t) \in [0, 1] \times [0, \infty), \tag{18}$

and the hypothesis proposed in part (b) is affirmed.

  • 1
    $\begingroup$ Wow, I managed to write up my answer and confirm what you wrote, but this is written up wonderfully that I now see for certain how (b) is true and how we check (a) thank you so much! $\endgroup$ – Peetrius Sep 29 '17 at 1:34
  • $\begingroup$ Thanks for the kind words. Any chance of adding a vote/acceptance to your written lauds? Cheers! $\endgroup$ – Robert Lewis Sep 29 '17 at 1:38
  • 1
    $\begingroup$ Of course! Don't worry I did, but because my rep is <15 it won't show it, but I gotcha :) still new and learning the ropes around stack exchange. $\endgroup$ – Peetrius Sep 29 '17 at 16:56
  • $\begingroup$ Many thanks good sir, my your reputation grow! $\endgroup$ – Robert Lewis Sep 29 '17 at 17:09

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.