Hi,all

For problems including both fast and slow reactions, taking a big timestep may results in negative concentrations. This always happens in my study.

Some people said that using a dimensionless concentration(mass fraction) can solve the above problem. Is this correct? Can someone help clarify this and give theoritical explaination?

Thanks!

Best regards!

James

Hi,

It is a boundedness issue. This is spurious numerical results, often caused by high order differencing schemes, but highly non-linear processes like chemistry can cause it too.

How to fix it depends on the exact problem. An alternative approach might work, also simply putting bounds on the variable values (ie fixing it to greater than or equal to zero) can also work.

Glenn Horrocks

Thanks, Glenn

I am actually using step fuctions in cfx to limit the concentration above zero, simialr as you suggested.

However I am interested in if using dimensionless concentrations could be another option.

Best regards!

James

Hi,

How are you using the step functions? Is the concentration an algebraic function or is a convected variable? If it is a convected variable (which is probably what it is) then the step functions will not limit it. You will have to use the variable bonudednesses limiters which I think are available under the solver tab or expert parameters.

Glenn Horrocks

Thanks again, Glenn

I define the concentrations as volumetric additional variables(scalars)(I guess it's convected variable in your term?). I am also using 'step function' to terminate the source term when the concentration becomes less than zero.

What is the 'variable bonudednesses limiters'? Do you mean 'scalar diffusion scheme' or 'bounded bnd diffusion tets', or others? I found following information in help document, but haven't yet understand how to use them.

'scalar diffusion scheme' Default Value 2 (tetrahedral) 1 (hexagonal)

Description Specify whether the standard central scheme default) or the positive definite scheme is applied to the scalar equations. This parameter may be of use in obtaining convergence with poor quality meshes. A value of 2 sets the scalar diffusion scheme to be positive definite.

bounded bnd diffusion tets

Description: Controls whether or not a bounded diffusion scheme is used at boundaries for tetrahedral elements. The diffusion scheme for other element types is bounded by default through the default setting of 5 for the expert parameters "cht diffusion scheme" and "scalar diffusion scheme". Tetrahedral elements are controlled separately because the bounded scheme may be lead to less accurate results. Parameter Type: String Allowed Values: t,f Default Value: f

Do you have more suggestions?

Best regards!

James

Hi,

Yes, I mean volumetric additional variables.

I don't think just limiting the source term will stop negative values. It is not a true boundedness limit. I have had a look around and cannot find a way of setting the boundedness of a variable.

The 'scalar diffusion scheme' or 'bounded bnd diffusion tets' you mention are different things and not what I mean.

Have a look in the book "Computational Fluid Dynamics" by Roache as he has a good discussion of this issue.

Glenn Horrocks