Hi John,

- At the beginning of the model run, the maximum velocity is 80 m/yr in the upper mantle beneath the subducting slab (flow goes around the hanging slab). The high velocity is at the first few step and decrease to 7m/yr which is still pretty high. My smallest mesh is 2.5km.

After I use composite flow law, I keep running into solver converging problem. For my subduction model, there are velocities imposed on the side boundaries (7cm/yr) and my boundary conditions are fixed composition on the left and right boundary and the boundary temperature is fixed as the initial temperature. The bottom is free slip and the top is free surface. I am not sure if it is my boundary condition cause the problem.

I tried to use 'iterated Advection and defect correction Stokes" (+ block GMG). I attached my boundary conditions and the Material model section. Could you spot any obvious error?

```
subsection Boundary composition model
set Fixed composition boundary indicators = right, left
set List of model names = initial composition
end
subsection Boundary temperature model
set Fixed temperature boundary indicators = bottom, top, left, right
set List of model names = initial temperature
end
subsection Mesh deformation
set Mesh deformation boundary indicators = top: free surface, top: diffusion
subsection Free surface
set Surface velocity projection = normal
end
subsection Diffusion
set Hillslope transport coefficient = 1.e-7
end
end
subsection Boundary velocity model
set Tangential velocity boundary indicators = bottom
set Prescribed velocity boundary indicators = right x:function, left x:function
subsection Function
set Function constants = w=2000e3, cm=0.01, year=1, oc_v=5, cont_v=2, blance_out=0.5926 #=(5*80+2*120)/(560+520)
set Function expression = if (x<=w/2 && y>580e3, (oc_v)*cm/year, \
if (x<=w/2 && y<560e3, -(blance_out)*cm/year, \
if (x<=w/2 && y<=580e3 && y>=560e3, \
((y-560e3)*(oc_v+blance_out)/20e3-blance_out)*cm/year, \
if (x>w/2 && y>540e3, -cont_v*cm/year, \
if (x>w/2 && y<520e3, blance_out*cm/year, \
((y-540e3)*(blance_out+cont_v)/(-20e3)-cont_v)*cm/year \
)\
)\
)\
)\
);\
0
set Variable names = x,y
end
end
```

Above are my boundary conditions

```
subsection Material model
set Model name = visco plastic
set Material averaging = harmonic average only viscosity
subsection Visco Plastic
set Minimum strain rate = 1.e-20
set Reference strain rate = 1.e-16
# Limit the viscosity with minimum and maximum values
set Minimum viscosity = 1e18
set Maximum viscosity = 1e25
set Reference temperature = 273
# compositional field: background, OC, OML, UC, LC, CML
set Define thermal conductivities = true
set Thermal conductivities = 54.2, 2.25, 2.25, 2.25, 2.25, 2.25
set Heat capacities = 750.
set Densities = 3337.4, 2950, 3337.4, 2811.2, 2929, 3337.4
set Thermal expansivities = 2e-5
# Harmonic viscosity averaging
set Viscosity averaging scheme = harmonic
set Viscous flow law = composite
# Dislocation creep parameters
# Materials background, OC, OML, UC, LC, CML
set Prefactors for diffusion creep = 1.5e-16, 1e-29, 1.5e-16, 1.5e-16, 1.5e-16, 1.5e-16
set Activation energies for diffusion creep = 375.e3, 0.e3, 375.e3, 375.e3, 375.e3, 375.e3
set Activation volumes for diffusion creep = 3.e-6, 0., 3.e-6, 3.e-6, 3.e-6, 3.e-6
# Dislocation creep parameters
# Note that the viscous pre-factors below are scaled to plane strain from unixial strain experiments.
set Prefactors for dislocation creep = 6.52e-16, 1e-50, 6.52e-16, 8.57e-28, 7.13e-18, 6.52e-16
set Stress exponents for dislocation creep = 3.5, 1.0, 3.5, 4.0, 3.0, 3.5
set Activation energies for dislocation creep = 530.e3, 0., 530.e3, 223.e3, 345.e3, 530.e3
set Activation volumes for dislocation creep = 18.e-6, 0., 18.e-6, 0., 0., 18.e-6
# Plasticity parameters
set Angles of internal friction = 30, 0, 30, 30, 30, 30
set Cohesions = 20.e6, 0, 20.e6, 20.e6, 20.e6, 20.e6
end
end
```

Above is the material model section.

The temperature and compositional field look fine. I think it’s either the boundary conditions or the material model cause the problem.

- That is good to know about particles. I will try to implement it in the models.

Thanks,

Xiaowen