I have since then resolved this problem by taking a low relative and absolute tolerance. Still curious if there are any other solutions, just in case it happens another time.

Hi Kumpal,

there might actually be several different reasons depending on your case. Could you be a little more specific regarding the case you are trying to simulate?

Best,

J

Hi Jacopo,

I'm trying to simulate mass transport of molecular species across tissue membranes. The error comes in when I modify the rate constants for the chemical reactions involving these molecular species beyond a certain range. Allotting a small tolerance value seems to remove the error and extend the range of usable rate constants. If you have any other solution, please let me know.

Regards,

KM

Hi Kumpal,

this is what comes to my mind: it might be that when you change the rate constants your model becomes unstable for different reasons, one can be that you have a very different range of values for the rate constants and the system becomes stiff, or that the reactions are too fast and the solver struggles a lot to solve them properly. Also, convection-diffusion-reaction models are badly handled by SUPG or GLS stabilization, newer methods are available which promise to give way better results. Unfortunately COMSOL does not have them implemented... yet. Using a smaller tolerance probably helps you because forces the solver to solve every time step very accurately and that avoids the global error to grow too much. I would suggest to try isotropic diffusion (very very small) and adaptive mesh refinement, this should work. Please note that isotropic diffusion is a non-conservative stabilization method, so use it with care.

Please let me know your comments,

Best,

J

I meant "inconsistent" and not "non-conservative", sorry for the mistake.

J

Variable Rate constants for the chemical reaction can be defined using the analytical function or step function. This will certainly help.

It seems to me that changing the parameter you mentioned from 1 to 100 has modified the physics a lot. If that is true, it is not a surprise that you need to modify the meshes. I believe the time scale and the length scales become totally different.

Have you check the dimensionless parameter yet? Every physical problem should be done in dimensionless way. At least, we should THINK in dimensionless way even though we are running dimensionally. If you check with that, you might even find that the problems with the parameter of 1 and of 100 are in two different physical regimes. One typical example is Peclect number for low and high thermal conductivity fluid.