1

I am trying to solve for the Viscous Burgers Equation on a Ring

$$ u_t + u u_x = \nu u_{xx};\; (x,t) \in [0,1) \times \mathbb{R}_+ $$

$$ u(x,0) = \phi(x) $$

$$ u(0,t) = u(1,t) $$

$$ u_x(0,t) = u_x(1,t). $$

From a Cole-Hopf transformation, this equation can be turned into a Heat Equation

$$ u(x,t) \rightarrow \phi_x(x,t);\quad \phi(x,t) \rightarrow -2\nu\log(\psi(x,t)) $$

$$ \psi_t = \nu \psi_{xx} $$

$$ (*) \quad u(x,t) = \phi_x(x,t) = -2\nu \frac{\psi_x(x,t)}{\psi(x,t)}. $$

Here are my questions:

  1. How do I handle the boundary and initial conditions from the Burgers equation into the converted Heat Equation?
  2. The heat equation can be solved via separation of variables, $\psi(x,t) = X(x)T(t)$. If we take this form however and shove it into $(*)$, we get

$$ u(x,t) = -2\nu\frac{\psi_x(x,t)}{\psi(x,t)} = -2\nu\frac{X'(x)T(t)}{X(x)T(t)} = -2\nu\frac{X'(x)}{X(x)} $$

which is independent in time. Is there something being done wrong?

K.defaoite
  • 12,536
  • Use $\in$ (\in) to denote set inclusion, not epsilon. – K.defaoite Dec 10 '23 at 04:02
  • Your last line is wrong because not all solutions of the heat equation are separable. – K.defaoite Dec 10 '23 at 04:07
  • Thanks for the edit @K.defaoite.

    I think my biggest confusion is in how the Boundary Conditions for the Burgers Equation translate to the Boundary Conditions to the Heat Equation. Perhaps if that was clear to me, the inseparability of the Heat Equation would be obvious.

    – user3166083 Dec 10 '23 at 04:48
  • I don't understand what your doubt is on the matter of the BCs. Just restate them in terms of $\psi$? – K.defaoite Dec 11 '23 at 00:41

1 Answers1

1
  1. The boundary conditions $u(0,t)=u(1,t), u_x(1,t)=u_x(1,t)$ are satisfied if the same conditions are imposed on $\psi$: $$ \psi(0,t)=\phi(1,t),\qquad \psi_x(0,t)=\psi_x(1,t). \tag{1} $$ The initial condition for $\psi$, on the other hand, can be derived by integrating your equation $(*)$: $$ u(x,0)=\phi(x)=-2\nu\frac{\psi_x(x,0)}{\psi(x,0)}\implies \psi(x,0)=\exp\left(-\frac{1}{2\nu}\int\phi(x)\,dx\right). \tag{2} $$
  2. The solution to the heat equation $\psi_t=\nu\psi_{xx}$ subject to the boundary conditions $(1)$ is $$ \psi(x,t)=\frac{a_0}{2}+\sum_{n=1}^{\infty}e^{-4n^2\pi^2\nu t}[a_n\cos(2n\pi x)+b_n\sin(2n\pi x)], \tag{3} $$ where the coefficients $a_n, b_n$ are determined by the initial condition $(2)$. Therefore, it is not true, in general, that $$ u(x,t)=-2\nu\frac{\psi_x(x,t)}{\psi(x,t)} \tag{4} $$ is independent of $t$.
Gonçalo
  • 9,312
  • Thank you. What would the limits of integration be in the argument of the exponential in Equation (2)? – user3166083 Dec 17 '23 at 03:07
  • You may write $\int_0^x\phi(x'),dx'$, but, as a matter of fact, the lower limit of integration is arbitrary. Changing it is equivalent to multiplying $\psi$ by a constant, which disappears in the ratio $\psi_x/\psi$. – Gonçalo Dec 17 '23 at 03:44