Hi Tao,
Basically the answer to your question is : You can't get rid of it with the current code. Let me explain this more.
We use three color functions indeed, one for each fluid. But at some points there are 3 fluids in a cell. How will the code reconstruct the interface ? In two phases we have only one color function, then there is only 1 interface between the 2 fluids. If you solve a 2-fluid problem with 2 color functions you will have 2 interfaces (which are equals). Then for 3 color functions you have three interfaces in the same cell. At this stage we are not able to reconstruct only part of the segment in a cell, this leads to some overlap zone and some empty zone : this is the white you see.
Then due to the advection we propagate this error around the triple point. This is why you see it grow.
If you look in my sandbox there is a patch on the advection scheme to help to reduce the spreading of the error :
We are currently writing a paper to characterise this method. These white might be important or not depending on the problem you want to study.
I hope this helps, let me know if you need more details on this.
Best regards
Pierre-Antoine