3
$\begingroup$

Let $b: \mathbb R_+\times\mathbb R\to \mathbb R$ and $\sigma: \mathbb R_+\times\mathbb R\to (0,\infty)$ be functions as nice as possible (e.g. bounded and of bounded partial derivatives, and $\inf_{(t,x)}\sigma(t,x)>0$). Consider the Kolmogorov forward and backward PDEs:

\begin{eqnarray} \partial_tu(t,x) + \partial_x (bu)(t,x)-\frac{1}{2}\partial^2_{xx}(\sigma^2u)(t,x)=0,\quad \forall t>0, x>0,\\ \partial_tv(t,x) - b(t,x)\partial_x v(t,x)-\frac{\sigma(t,x)^2}{2}\partial^2_{xx}v(t,x)=0,\quad \forall t>0, x\in\mathbb R \end{eqnarray}

with $u(0,\cdot)=f$, $u(\cdot,0)=0$ and $v(0,\cdot)=g$.

Assuming both admit a classical solution. Then a straightforward computation yields

$$\partial_t(uv)(t,x) + \partial_x(buv)(t,x) - \partial_x\left(\frac{\partial_x(\sigma^2u)v}{2}-\frac{\sigma^2u\partial_xv}{2}\right)(t,x)=0,\quad \forall t>0, x>0,$$

and further by Fubini's theorem and integration by parts

\begin{eqnarray} 0&=&\int_0^T \int_0^\infty \left(\partial_t(uv)(t,x) + \partial_x(buv)(t,x) - \partial_x\left(\frac{\partial_x(\sigma^2u)v}{2}-\frac{\sigma^2u\partial_xv}{2}\right)(t,x)\right)dxdt \\ &=& \int_0^\infty uv(T,x)dx - \int_0^\infty fg(x)dx +\int_0^T \frac{\partial_x\big(\sigma^2u)v(t,0)}{2}dt\quad (\ast). \end{eqnarray}

My question is, if $u$ is a weak solution instead of the classical solution, does $(\ast)$ still hold? Here we may assume the initial conditions $f,g$ are also as nice as possible.

$\endgroup$
4
  • 1
    $\begingroup$ This question seems too broad to be answered as currently stated. For instance, what is your definition here of a weak solution? Often for PDE, you have sort of weak-strong uniqueness principle. So if your ID f,g are nice, then the equations admit a classical solution, which therefore must coincide with the weak solution. $\endgroup$ Commented Jun 3, 2022 at 21:35
  • $\begingroup$ @MattRosenzweig Thanks for your comment. Indeed, the sufficient condition ensuring the existence of a classical solution $u$ is unknown to me, as my previous question posted mathoverflow.net/questions/423734/… $\endgroup$
    – GJC20
    Commented Jun 3, 2022 at 22:20
  • $\begingroup$ @MattRosenzweig The definition of the weak solution seems to be the one given at mathoverflow.net/questions/423301/… $\endgroup$
    – GJC20
    Commented Jun 3, 2022 at 22:22
  • $\begingroup$ @MattRosenzweig Btw, if you know some references (as recent as possible) on the existence of the classical solution $u$ (in view of the conditions on $b,\sigma, f$), please let me know and I do appreciate $\endgroup$
    – GJC20
    Commented Jun 3, 2022 at 22:24

0

You must log in to answer this question.

Browse other questions tagged .