My question is about Finite Element Method.

I want to know how to implement "gradient zero" conditions to advection-diffusion equations in conservative form like,

$\frac{\partial \rho}{\partial t} + \nabla \cdot \mathbf{\Gamma} = 0$

where, $\mathbf{\Gamma} = \mathbf{u}\rho - k\nabla \rho$

$\mathbf{u}$ and $k$ is not constant variables.

The boundary integral part of weak-form can be written as follows,

$\int_{\Gamma^N} \rho^* \mathbf{\Gamma} \cdot \mathbf{n} d\Gamma^N$

where, $\mathbf{n}$ is normal vector and $\rho^*$ is weighed function.

But it seems that this integral is specialized to "Mixed boundary condition".

So my question is how to modify this integral to that of "Gradient zero conditon".

I beg your kindness.

  • $\begingroup$ The term "gradient zero condition" is not commonly used. What do you mean by it? $\endgroup$ Dec 26, 2013 at 14:47
  • $\begingroup$ Thank you for your speedy response. I mean, $\frac{\partial \rho}{\partial \mathbf{n}} = 0 $ as "gradient zero condition". $\endgroup$ Dec 26, 2013 at 16:48

1 Answer 1


For the advection-diffusion equation you can't just apply a Neumann boundary condition (the 'gradient zero condition' as you called it) because there is are two components to the flux: advection and diffusion. Therefore you are correct to apply Robin boundary conditions (mixed boundary conditions as you called it).

If you want the edges of your domain to prevent mass from leaving then you simply apply the zero flux condition,

$$\int_{\Gamma^N} \rho^* \mathbf{\Gamma} \cdot \mathbf{n} d\Gamma^N = 0$$

For more information see theses question/answers:

  1. Conservation of a physical quantity when using Neumann boundary conditions applied to the advection-diffusion equation

  2. How should boundary conditions be applied when using finite-volume method?

If you were to apply the condition $\frac{\partial n}{\partial x} = 0$ at the boundary this would be equivalent to preventing diffusion through the boundary, but it would be an open boundary condition for the advection component of the flux. This may or may not be what you want.

  • $\begingroup$ Thank you for your answer. I can clearly understand what you mean. $\endgroup$ Jan 26, 2014 at 8:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.