last day (27 days later) » 

andre
andre
6:42 PM
This is a chatroom about my investigations about "how to make disappear the artificial boundary phi=0" when solving a Laplace equation in polar coordinate. These are investigations are a my way to understand how to impose continuity of flux of heat (or electric field, it's exactly the same problem) in a "internal" boundary
To undestand the context see the beginning of my answer here
... and the comments too.
First a solution that respects the continuity of the potential only (and does not respect flux continuity) : 
solCyl10 = NDSolveValue[
{
laplacianCil == 0, boundaryConditionCil,
PeriodicBoundaryCondition[V[r,p],p==2 Pi,Function[x,x+{0,-2 Pi}]]
},
V,
{r, r1, R},
{p, 0, 2 Pi},
MaxSteps -> Infinity];

potentialSquareRepresentation=ContourPlot[solCyl10[r, p], {r,p} \[Element] solCyl10["ElementMesh"]
, ColorFunction -> "Temperature"
,Contours-> 20
, PlotLegends -> Automatic
];

potentialCylindricalRepresentation=Show[
potentialSquareRepresentation /. GraphicsComplex[array1_, rest___] :>
                  GraphicsComplex[(#[[1]] {Cos[#[[2]]],Sin[#[[2]]]})& /@ array1, rest],
That not the only solution that respects the continuity of the potential. Here is another one : 
solCyl11 = NDSolveValue[
{
laplacianCil == 0, boundaryConditionCil,
PeriodicBoundaryCondition[(*.5 +*) V[r,p],p==0 && 0<r<1,Function[x,x+{0, 2 Pi}]]
},
V,
{r, r1, R},
{p, 0, 2 Pi},
MaxSteps -> Infinity];

potentialSquareRepresentation=ContourPlot[solCyl11[r, p], {r,p} \[Element] solCyl10["ElementMesh"]
, ColorFunction -> "Temperature"
,Contours-> 20
, PlotLegends -> Automatic
];

potentialCylindricalRepresentation=Show[
potentialSquareRepresentation /. GraphicsComplex[array1_, rest___] :>
                  GraphicsComplex[(#[[1]] {Cos[#[[2]]],Sin[#[[2]]]})& /@ array1, rest],
Of course any linear combinaison of theses two solutions are also solution.
and there are other ones ...
The difference in the code of these two solutions are in the PeriodicBoundaryCondition[...]
Indeed, the documentation of PeriodicBoundaryCondition[...] (item "possible issues") explains that there is a notion of "Source" and "Target" boundaries. This imposes a "direction of propagation" to the solution. Between the two examples above I have simply inverted the target and the source boundaries.
(In the first case the source boundary is upper side of horizontal line and the target is the other side. Note that compromises are possible, one can invert the target/source at the half of the line for example, hence my text :"and there are other ones" above)
And now something I find weird :
Here is the sommation of the two solutions above : 
potentialSquareRepresentation=ContourPlot[functionPot00[r, p], {r,p} \[Element] solCyl10["ElementMesh"]
, ColorFunction -> "Temperature"
,Contours-> 20
, PlotLegends -> Automatic
];

potentialCylindricalRepresentation=Show[
potentialSquareRepresentation /. GraphicsComplex[array1_, rest___] :>
                  GraphicsComplex[(#[[1]] {Cos[#[[2]]],Sin[#[[2]]]})& /@ array1, rest],
PlotRange -> Automatic
];

functionGivingAPair12={Derivative[1,0][functionPot00][#1,#2],1/#1 Derivative[0,1][functionPot00][#1,#2]}&;
One has the feeling that this solution is not far from the real solution, the one that respects continuity of potential and flux. (Note that in this problem the real solution is symetric with regard to the diagonal).
Further investigations seems to confirm this.
Furthermore, it seems to be also true is the example given in the documention of PeriodicBoundaryCondition[], "item Possible issues".
I don't think it's true, in general, but in these cases ???
I'm interested in any enlightening...
 
andre
andre
7:42 PM
Otherwise, maybe someone knows if it is possible to use some Periodic Neumann condition + Periodic Dirichlet. It would be the best attack to the problem ?
By the way, I could solve the OP's problem (I mean in the interpretation with only two electrodes in diagonal, as in the sketch. This does not lead to 1.9pF/m) with a method that has nothing to do with finite elements and that is very accurate. But I afraid I will not have enough time to do it.
 
andre
andre
8:10 PM
Maybe there's another approach : is it possible to modify the mesh in the way to sew the lower side and the upper side of the horizontal line ? . I'm thinking of something like a "duplicate vertex fusioning" in the mesh ? Someone inspired ?
 

  last day (27 days later) »