Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
Concentration problem within micromixers: absurd concentrations, no mass conservation
Posted 24 mar 2010, 04:57 GMT-4 Fluid & Heat, Computational Fluid Dynamics (CFD), Microfluidics, Studies & Solvers 10 Replies
Please login with a confirmed email address before reporting spam
Hello everybody.
I am modelling micromixers and want to check their efficiencies.
For that purpose, I first used the Incompressible Navier-Stokes model, to calculate the flow inside my device. I set it up that way:
- Pressure, no viscous stress, 20 Pa at my inlets
- Pressure, no viscous stress, 0 Pa at my outlet.
- everything else is "Wall".
I simulate this with default microfluidics settings.
Then I use the Convection/Diffusion model, with these settings:
- Diffusion coef: 1e-9
- x-y-z velocity : u-v-w
- Concentration : 10 at one of my inlet, 0 at the other.
- Boundary condition at outlet: Convective flux.
The problem is: the resuts are absurd. First, I have negative concentrations and >10 concentrations, which is physically impossible.
Then, when I observe the outlet domain, the concentration is very homogeneous (to the hundredth), but not around 5, as it should be. On some of my designs, it is as low as 2 (or as high as 8 if I inverse the concentration at the inlets). It seems there is no mass/matter conservation.
Therefore I tried the "Conservative Equation Form" (in the Properties of the Convection/Diffusion model - Non conservative is default). I am not sure it is a good idea. The results are completely different, and now, I don't know which model I should trust.
Can someone help me with this ? I've already read the KnowledgeBase about negative concentration (www.comsol.com/support/knowledgebase/952/) and tried the "Refine Mesh" solution, with no success (except making my computer freeze!). I would like to try the "smoothing function" but cannot figure out what to do or where (I am new to Comsol AND microfluidics, so I don't know/understand what/where every feature is). And it is still no solution to my mass/matter conservation issue.
Thanks a lot in advance for any advice.
I am modelling micromixers and want to check their efficiencies.
For that purpose, I first used the Incompressible Navier-Stokes model, to calculate the flow inside my device. I set it up that way:
- Pressure, no viscous stress, 20 Pa at my inlets
- Pressure, no viscous stress, 0 Pa at my outlet.
- everything else is "Wall".
I simulate this with default microfluidics settings.
Then I use the Convection/Diffusion model, with these settings:
- Diffusion coef: 1e-9
- x-y-z velocity : u-v-w
- Concentration : 10 at one of my inlet, 0 at the other.
- Boundary condition at outlet: Convective flux.
The problem is: the resuts are absurd. First, I have negative concentrations and >10 concentrations, which is physically impossible.
Then, when I observe the outlet domain, the concentration is very homogeneous (to the hundredth), but not around 5, as it should be. On some of my designs, it is as low as 2 (or as high as 8 if I inverse the concentration at the inlets). It seems there is no mass/matter conservation.
Therefore I tried the "Conservative Equation Form" (in the Properties of the Convection/Diffusion model - Non conservative is default). I am not sure it is a good idea. The results are completely different, and now, I don't know which model I should trust.
Can someone help me with this ? I've already read the KnowledgeBase about negative concentration (www.comsol.com/support/knowledgebase/952/) and tried the "Refine Mesh" solution, with no success (except making my computer freeze!). I would like to try the "smoothing function" but cannot figure out what to do or where (I am new to Comsol AND microfluidics, so I don't know/understand what/where every feature is). And it is still no solution to my mass/matter conservation issue.
Thanks a lot in advance for any advice.
10 Replies Last Post 30 apr 2013, 12:44 GMT-4
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
No one can help me with this issue ?
Not even to understand what's wrong in my model?
Not even to understand what's wrong in my model?
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hello Blaise Mulliez
Please attach your comsol *.mph to to understand your problem.
Please attach your comsol *.mph to to understand your problem.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Did you make any progress? Can a finite element method be absolutely conservative for diffusion/convection problems?
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
from my experience, it would appear not. Although I did find research papers on exactly this topic that did suggest different approaches computationally do. It would just appear that these are not available in comsol.
I have tried implementation several different ways -- and in the end they all result in having negative concentrations at some point. Also I found that if one plotted the smoothing terms that are recommended -- the models can have issues with convergence as the smoothing has significant noise in the low order of manitude range of 10^-12 etc... and you will see spikes or discontinuities in the smoothing profile.
I have tried implementation several different ways -- and in the end they all result in having negative concentrations at some point. Also I found that if one plotted the smoothing terms that are recommended -- the models can have issues with convergence as the smoothing has significant noise in the low order of manitude range of 10^-12 etc... and you will see spikes or discontinuities in the smoothing profile.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi all,
I've been having a similar issue to the one above.
I'm developing passive microfluidic mixers in COMSOL, but I am starting to question some of my results. I usually specify inlet concentrations of 1 and 0 with a resulting fully mixed stream of 0.5. However, the final concentration throughout the mixer rarely stays between 0 and 1. At lower values of fluid velocity, and hence lower Reynolds numbers (Re = 0.01 -1), the concentration may only break these bounds by a fraction ~ 5%, however, as I increase the fluid velocity for higher Reynolds numbers (Re = 10-20) the concentration may exceed +/- 10, which does not make sense to me.
I've tried smoothing the inlet profile to eliminate the negative inlet concentration profile. I've also tried using artificial diffusion, but I'm weary of tinkering with this as I am unsure how it effects the accuracy of my results.
To try and verify my results, I have attempted to model a micromixer from the literature and reproduce the results - so far I have been unsuccessful. I have attached the relevant paper. I have also included a link to my version of the model from the paper (It is too large to directly link here so I have uploaded it to sendspace)
www.sendspace.com/file/dbpkjd
Any help you can give me would be greatly appreciated.
-KC
I've been having a similar issue to the one above.
I'm developing passive microfluidic mixers in COMSOL, but I am starting to question some of my results. I usually specify inlet concentrations of 1 and 0 with a resulting fully mixed stream of 0.5. However, the final concentration throughout the mixer rarely stays between 0 and 1. At lower values of fluid velocity, and hence lower Reynolds numbers (Re = 0.01 -1), the concentration may only break these bounds by a fraction ~ 5%, however, as I increase the fluid velocity for higher Reynolds numbers (Re = 10-20) the concentration may exceed +/- 10, which does not make sense to me.
I've tried smoothing the inlet profile to eliminate the negative inlet concentration profile. I've also tried using artificial diffusion, but I'm weary of tinkering with this as I am unsure how it effects the accuracy of my results.
To try and verify my results, I have attempted to model a micromixer from the literature and reproduce the results - so far I have been unsuccessful. I have attached the relevant paper. I have also included a link to my version of the model from the paper (It is too large to directly link here so I have uploaded it to sendspace)
www.sendspace.com/file/dbpkjd
Any help you can give me would be greatly appreciated.
-KC
Attachments:
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi all,
I've been having a similar issue to the one above.
I'm developing passive microfluidic mixers in COMSOL, but I am starting to question some of my results. I usually specify inlet concentrations of 1 and 0 with a resulting fully mixed stream of 0.5. However, the final concentration throughout the mixer rarely stays between 0 and 1. At lower values of fluid velocity, and hence lower Reynolds numbers (Re = 0.01 -1), the concentration may only break these bounds by a fraction ~ 5%, however, as I increase the fluid velocity for higher Reynolds numbers (Re = 10-20) the concentration may exceed +/- 10, which does not make sense to me.
I've tried smoothing the inlet profile to eliminate the negative inlet concentration profile. I've also tried using artificial diffusion, but I'm weary of tinkering with this as I am unsure how it effects the accuracy of my results.
To try and verify my results, I have attempted to model a micromixer from the literature and reproduce the results - so far I have been unsuccessful. I have attached the relevant paper. I have also included a link to my version of the model from the paper (It is too large to directly link here so I have uploaded it to sendspace)
www.sendspace.com/file/dbpkjd
Any help you can give me would be greatly appreciated.
-KC
I redid your model in Comsol 4.1, since I am more comfortable with that.
- Your inlet concentration is based on hight, y, which if you look starts from -1e-4 to 1e-4, so you have actually a negative inlet concentration which messes up everything.
This is the model result I got. Of course there are some oscilaltions int he result (higher concentration than 1), but with mesh refining you can avoid it as much as possible.
--
Comsol 4.1
Ubuntu 10.04.1
Attachments:
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Thank you for taking the time to look at this.
- On your first point re: inlet concentration.
I don't think placing my channel along the 0 y coordinate affects the results. I re-ran with it moved up into the positive y region and it resulted in the same outcome for me.
If I run it for lower Re values I get a good profile between 0 and 1, it's only at higher Re values that it breaks down.
- What mesh size did you use when solving both the N/S and C/D?
- I have been using artificial isotropic diffusion to try help the solution converge, but I find this really affects the mixing efficiency result, and I am unsure whether to trust it. Any thoughts on this?
- On your first point re: inlet concentration.
I don't think placing my channel along the 0 y coordinate affects the results. I re-ran with it moved up into the positive y region and it resulted in the same outcome for me.
If I run it for lower Re values I get a good profile between 0 and 1, it's only at higher Re values that it breaks down.
- What mesh size did you use when solving both the N/S and C/D?
- I have been using artificial isotropic diffusion to try help the solution converge, but I find this really affects the mixing efficiency result, and I am unsure whether to trust it. Any thoughts on this?
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Thank you for taking the time to look at this.
- On your first point re: inlet concentration.
I don't think placing my channel along the 0 y coordinate affects the results. I re-ran with it moved up into the positive y region and it resulted in the same outcome for me.
If I run it for lower Re values I get a good profile between 0 and 1, it's only at higher Re values that it breaks down.
- What mesh size did you use when solving both the N/S and C/D?
- I have been using artificial isotropic diffusion to try help the solution converge, but I find this really affects the mixing efficiency result, and I am unsure whether to trust it. Any thoughts on this?
Hi Kevin,
ok, as I said I used Comsol 4.1 and the results changed for me depending on y. If your profile is ok, then no problem.
as for the mesh, I used normal mesh.
Artificial diffusion is a *very* tricky business in finite element and C-D. Specially, for this kind of cases, where you are interested in mixing behavior.
One method you can try is this. Always use sequential solver, then try adaptive mesh refinement, go to very high mesh densities, then disable artificial diffusion. There you can see the effect of AD. Yet, since you are using Finite element, you must be sure that your mesh is fine enough to avoid oscillations in the result (the result solved without AD may not be the most accurate one).
Danial
--
Comsol 4.1
Ubuntu 10.04.1
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Can you repost your model. I am working on something similar and am curious about your approach
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Dear Danial Tz,
I see your picture on micromixers. I have a problem on control the mixing efficiency because the color of outlet can not see clear enough. Your picture has a nice color range. How can I fix the problem? Can you send me an example model just for study. I'm a student now. Thank you very much.
Best Regards
I see your picture on micromixers. I have a problem on control the mixing efficiency because the color of outlet can not see clear enough. Your picture has a nice color range. How can I fix the problem? Can you send me an example model just for study. I'm a student now. Thank you very much.
Best Regards