[CIG-MC] Layered viscosity in CitcomCU

[Eh Tan]

Joya Tetreault wrote:

> Hello Eh,
>
> Shijie pointed me in your direction for help with CitcomCU.  I am 
> trying to produce simple Stokes steady-state flow for couette flow 
> systems with citcomCU in cartesian coordinates.  I am having trouble 
> with understanding how the layers are constructed and the interface 
> boundary conditions involved.  It seems that if I give the input file 
> 2 layers in z direction, but the same viscosity for both layers that I 
> do not get the same velocity profile as if I actually did construct 
> the region with only 1 layer.  I'm enclosing the input files and 
> velocity profiles for both the 1 layer case (input_layer1, 
> Vx_profile_nonlayers.eps) and for the 2 layer case (input_layer2, 
> Vx_profile.eps).  I'm doing this for a simple test case where there is 
> only velocity on the surface (40 in X direction), no temperature BCs, 
> and allowing for periodic flow.  My question is why don't those 
> velocity profiles look the same if I am essentially creating 1 layer 
> by creating 2 identical layers?  Obviously there are other input 
> parameters that I need to modify.  Could you help me?
>
> thanks so much,
> Joya Tetreault
>
>  
>

Hi Joya,

(Cc to cig-mc as well)

CitcomCU (and CitcomS) always needs 4 layers of viscosity, that is, 
num_mat must be 4. The layer interface is defined by these parameters.

# for cart3d
z_lmantle=0.76655052		# 2870 km-670 kM
z_410=0.857143			# 2870 km-410 km
z_lith=0.9651568		# 2870-100 km

# for Rsphere
r_lmantle=0.89482       # (Ro-670 km)/Ro
r_410=0.9356358
r_lith=0.984301         # (Ro-100 km)/Ro  


If num_mat is more than 4, the extra parameters (e.g. the fifth and 
later elements of viscE) are ignored. If num_mat is less than 4, default 
value will be used for the layer, which gives constant viscosity for the 
layer. This behavior is surprising and not documented. I consider this 
as a bug. Fututer version of CitcomCU and CitcomS should enforce that 
num_mat=4.


Eh