Compression boundary condition incorrect

[bclyons12]

If I understand it right, com_bc=1 should set grad^2 chi = 0, but this seems to be setting Del^* chi = 0. Should we flip -x to x?

M3DC1/unstructured/model.f90

Lines 382 to 385 in cc6e982

	! no compression
	if(com_bc.eq.1 .and. numvar.ge.3) then
	call set_laplacian_bc(i_chi,rhs,temp,normal,curv,izonedim,-x,mat)
	endif

[nferraro]

I think it's right the way it is. v = ... + (1/R^2) grad_perp(chi) so div(v) = Del*(chi) / R^2

[nferraro]

Although I do see that this seems to be at odds with boundary_com in boundary.f90, where there's a +x. I think it might be the case that the "com" field that is calculated when there is hyperviscosity, for example, is indeed del_perp^2(chi), probably as a vestige from the ivform=0 coding. So we need to be careful in recognizing that the "com" field is not the same as div(v).

If the goal of the com_bc=1 boundary condition in boundary_vel is to be consistent with a dirichlet condition on the "com" field, then you are right that we need to change this to +x. But if the goal of com_bc=1 is to set div(v)=0 at the boundary, then we should leave boundary_vel the way it is and change the definition of the "com" field from del_perp^2(chi) to del*(chi). This would require some changes to the metricterms that use com, which probably are only related to hyperviscosity.

The latter approach is probably more physically motivated, although I can't really think of a physical effect that would enforce div(v) = 0 at the boundary.

[bclyons12]

I'll admit I don't have a specific application for com_bc = 1 here, so I haven't thought about it much. I was just looking at how the inocurrent_tor = 1 boundary condition was implemented, as I had a case where jy was not zero on the boundary even with it imposed (see my email from today). I noticed that com_bc was implemented the same way. I haven't checked the new LaTeX documentation, but the old documentation said it should be grad^2 chi:

[nferraro]

Right, that might be the definition of the variables, but those definitions don't correspond to the actual vorticity or compressibility given the decomposition of the fluid velocity we are using now. It sounds like the hyperviscosity is not functioning properly at the moment, and maybe this is related. It's worth a look.