I have solved numerically the following system: \begin{cases} \partial_t{u} + \partial_x{v} = 0 \\ \partial_t{v} + \frac{1}{\varepsilon^2}\partial_x{u} = -\frac{1}{\varepsilon^2}(v-f(u)) \end{cases} with periodic boundary conditions by finite volume method (IMEX-Runge Kutta).

I was wondering if, as it happens in a lot of other systems, in this case, I should expect to have a mass conservation property.

In the scalar case that kind of property can be verified by checking if $\forall n\in\mathbb{N}$ the following holds:

$$ \sum_{i\in\mathbb{N}} u_i^n = \sum_{i\in\mathbb{N}} u_i^{n+1}$$

but in my vector case I don't know how to eventually proceed. Should I consider any particular norm?

  • $\begingroup$ The first equation is still in divergence form and if you use a finite volume method, you still have conservation both at pde and numerical level. $\endgroup$ – cfdlab Mar 7 at 2:23
  • $\begingroup$ @cfdlab I think that the OP is also interested in how to see numerically the conservation, since in the scalar case is trivial, but in this case the solution is a vector of two components, so what quantity should one check for conservation? $\endgroup$ – VoB Mar 7 at 9:11
  • $\begingroup$ The $v$ equation is not a conservation law, so it will not be a conserved quantity. $u$ is clearly conserved if you use a finite volume method. We dont know what is the scheme for $v$, so it is not possible to say. $\endgroup$ – cfdlab Mar 9 at 11:53
  • $\begingroup$ I have used Rusanov numerical flux, which in this linear case is just Conservative Lax-Friedrichs and the $u$ variable is in fact conserved. Thanks for the answer for the $v$ variable, which in fact is not conserved by my scheme and hence now I understand why it is correct. $\endgroup$ – Dadeslam Mar 9 at 12:14
  • $\begingroup$ @Dadeslam can you close this question by accepting one of the answers that is best in your view. $\endgroup$ – cfdlab Mar 15 at 4:58

Yes, mass conservation holds, because in fact you are solving some sort of a relativistic diffusion equation. Why:

$$\partial_{t} u + \partial_{x} v = 0$$


$$\partial_{t} v + \frac{1}{\epsilon^{2}} \partial_{x} u = -\frac{1}{\epsilon^{2}} (v - f(u))$$

But from first equation:

$$\partial_{tt} u + \partial_{xt} v = 0$$

and from the second one:

$$\partial_{tx} v + \frac{1}{\epsilon^{2}} \partial_{xx} u = -\frac{1}{\epsilon^{2}} (\partial_{x} v - f^{'}(u) \partial_{x} u)$$

But $\partial_{xt} v = \partial_{tx} v$, so:

$$-\epsilon^{2} \partial_{tt} u + \partial_{xx} u = \partial_{t} u + f^{'}(u) \partial_{x} u$$

or by taking $\epsilon^{2} = \frac{1}{c_{s}^{2}}$, where $c_{s}$ is some sort of speed of sound:

$$\frac{1}{c_{s}^{2}} \frac{\partial^{2} u}{\partial t^{2}} + \frac{\partial u}{\partial t} - \frac{\partial^{2} u}{\partial x^{2}} + f^{'}(u) \frac{\partial u}{\partial x} = 0$$

If I take the convective velocity as $w = f^{'}(u)$:

$$\frac{1}{c_{s}^{2}} \frac{\partial^{2} u}{\partial t^{2}} + \frac{\partial u}{\partial t} - \frac{\partial ^{2} u}{\partial x^{2}} + w \frac{\partial u}{\partial x} = 0$$

Now define: $$\frac{D}{Dt} = \frac{\partial}{\partial t} + w \frac{\partial }{\partial x}$$

and d'Alembert (wave) operator:

$$\Box^{2} = -\frac{1}{c_{s}^{2}} \frac{\partial^{2}}{\partial t^{2}} + \frac{\partial^{2}}{\partial x^{2}}$$

So finally:

$$\frac{D u}{D t} = \Box^{2} u$$

Taking integration from both side ($\Omega$ is your computational domain, in 1D case it is just a line):

$$\frac{D}{D t} \int_{\Omega} u(x,t) dx = \int_{\Omega} \Box^{2} u dx = \int_{\partial \Omega} \Box u da = 0$$

If you have Neumann boundary condition of $\Box u = 0$ at the wall. So:

$$\frac{D}{D t} \int_{\Omega} u dx = 0$$


$$\int_{\Omega} u dx = C$$

Where $C$ is just a constant equal to total mass of your system.

Update: Note that same thing is true for $v$, because you have this as well for $v$ variable:

$$\frac{D v}{D t} = \Box^{2} v$$

| cite | improve this answer | |
  • 1
    $\begingroup$ It looks clear to me now that $u$ needs to be preserved, but since we are dealing with a balance law in the second equation of the system and I am getting some dissipative effect in the numerical solution, I am not understanding now why we should have a conservation even in $v$. Can you please explain it to me? $\endgroup$ – Dadeslam Mar 8 at 12:37

It's clear that the mass of $v$ is not conserved in general. Just take $u(x,t=0) = 0$ and choose $f$ so that $f(0)=0$. Then

$$\left. \frac{\partial v}{\partial t}\right|_{t=0} = -\frac{v}{\epsilon^2}.$$

| cite | improve this answer | |

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.