From PDE Evans, 2nd edition, page 148...

Observe \begin{align} \int_0^\infty \int_{-\infty}^\infty w_t v_x \, dx dt &= -\int_0^\infty \int_{-\infty}^\infty wv_{tx} \, dx dt - \int_{-\infty}^\infty wv_x \, dx \vert_{t=0} \\ &= \int_0^\infty \int_{-\infty}^\infty w_x v_{t} \, dx dt + \int_{-\infty}^\infty w_x v \, dx \vert_{t=0} \end{align}

For the first line, I believe the formula to the integration by parts was employed. (Please correct me if I'm mistaken.) But for the last step, may I ask how can we just move the subscript of $x$ from $v$ to $w$, in both the terms, and expect to remove the negative?

(Note: This step is part of proving the Lax-Oleinik formula as an integral solution to a certain PDE with initial conditions.)

  • $\begingroup$ Are we to assume compact support on $w$? $\endgroup$ – DanZimm Jun 23 '14 at 8:52
  • $\begingroup$ Yes. Perhaps I should note also that the mapping x↦w(x,t) is Lipschitz continuous from earlier proofs (thus, absolutely continuous). Similar with t↦w(x,t) -- absolutely continuous for each x∈R. $\endgroup$ – Cookie Jun 23 '14 at 9:05

Assuming $w,v$ are in the necessary spaces you are correct about the first line. Remember that the general idea for IBP is $$ \int_S u v' \, \mathrm{d} \mu = \int_{\partial S} uv \, \mathrm{d} \mu - \int_S u'v \, \mathrm{d} \mu $$ and that the first integral vanishes as long as either $v$ or $u$ has compact support (i.e. the functions are in the correct space).

So now consider $$ \begin{eqnarray} \int_0^\infty \int_{-\infty}^\infty w_t v_x \, \mathrm{d} x \, \mathrm{d} t & = & \int_{-\infty}^\infty \left. w v_x \right|_{t=0}^{t\to\infty} \,\mathrm{d} x - \int_0^{\infty} \int_{-\infty}^\infty w v_{xt} \, \mathrm{d} x \, \mathrm{d} t \\ & = & - \int_{-\infty}^\infty \left. w v_x \right|_{t=0} \, \mathrm{d}x - \int_0^{\infty} \int_{-\infty}^\infty w v_{xt} \, \mathrm{d} x \, \mathrm{d} t \end{eqnarray} $$ Here we are integrating by parts in the $t$, and the second equality comes from an assumption that $w$ has compact support in $t$. Now let's do this again but this time in the $x$ integral: $$ - \int_{-\infty}^\infty \left. w v_x \right|_{t=0} \, \mathrm{d}x - \int_0^{\infty} \int_{-\infty}^\infty w v_{xt} \, \mathrm{d} x \, \mathrm{d} t \\ = - \left. \left. wv \right|_{t=0} \right|_{x\to -\infty}^{x \to \infty} + \int_{-\infty}^{\infty} \left. w_x v \right|_{t=0} \, \mathrm{d} x - \int_0^{\infty} \left. w v_t \right|_{x\to-\infty}^{x\to\infty} \, \mathrm{d}t + \int_0^{\infty} \int_{-\infty}^\infty w_x v_t \, \mathrm{d} x \, \mathrm{d} t \\ = \int_{-\infty}^\infty \left. w_x v \right|_{t=0} \, \mathrm{d} x + \int_0^{\infty} \int_{-\infty}^\infty w_x v_t \, \mathrm{d} x \, \mathrm{d} t $$ where the last equality follows from the assumption that $w$ has compact support in $x$.

  • $\begingroup$ Note that you'll also need some derivatives of some sort here, i.e. $w,v \in \mathcal{C}^1$ or $w,v \in H^1$ $\endgroup$ – DanZimm Jun 23 '14 at 9:14
  • $\begingroup$ Thanks for providing your answer rather quickly. If it's okay with you though, I plan to take a careful look at this later, as right now I need to get some sleep -- it's late in the night in my timezone (which is PDT). $\endgroup$ – Cookie Jun 23 '14 at 9:21
  • $\begingroup$ @glace no problem! I'm central, actually just woke up :D $\endgroup$ – DanZimm Jun 23 '14 at 9:30
  • $\begingroup$ For the work of $\int_0^\infty \int_{-\infty}^{\infty} w_t v_x \, dx dt$, the first step is using the integration by parts formula which I understand. In between the next two steps, you asserted that $$\int_{-\infty}^\infty \left. w v_x \right|_{t=0}^{t\to\infty}=\int_{-\infty}^\infty \left. w v_x \right|_{t=0}$$ That is what I still have trouble understanding; can you explain how the negative sign comes out when you simply remove the "$t \rightarrow \infty$"? $\endgroup$ – Cookie Jun 23 '14 at 17:42
  • $\begingroup$ @glace well note that I assert $$ \int_{-\infty}^\infty \left. w v_x \right|_{t=0}^{t \to \infty} = - \int_{-\infty}^{\infty} \left. w v_x \right|_{t=0} $$ and this follows since the side becomes $$ \lim_{b \to \infty} w(x,b) v_x(x,b) - w(x,0) v_x(x,0) $$ but this first term drops out due to compact support, thus leaving $$ \left. w v_x \right|_{t=0} $$ which is then inside the integral $\endgroup$ – DanZimm Jun 23 '14 at 18:32

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.