Viscoelastic benchmark model gives the wrong output

ASPECT
1

Hello ASPECT team,

ASPECT simulations (ASPECT 2.6.0-pre) with viscoelastic rheology give me wrong outputs such as the effective viscosity. Let’s take benchmarks/free_surface_tractions/viscoelastic/free_surface_VE_cylinder_3D_loading.prm as an example. The input prm file sets viscosity as 3e20 Pa s, whereas the simulation output shows 2.85534e19 Pa s and 3.12285e18 Pa s at the beginning and last time step of the simulation.

Weird. I found similar results if viscoelasticity was considered.

Do other users encounter similar issues?

As an aside, the use of docker does not allow us to pull the older versions of ASPECT.
This does not work at all: docker pull geodynamics/aspect:2.5.0

Best,
SZ

2

Dear SZ,

Thanks for posting!

Effective viscosity is not the same as viscous viscosity. In general, if you add a deformation mechanism (elasticity, plasticity, …), the effective viscosity will always be lower than the pure viscous viscosity. So I have a couple of questions for you:

  • What is the value of the effective viscosity that you expect at the start and end of this simulation?
  • Have you checked ASPECT’s results against the analytical solutions provided in the benchmark directory? Do they fail to pass the benchmark?

I am aware that the files in that directory are very old and not documented in the manual, and I’ve opened a PR to address this here. But we would be very grateful if you could confirm for us that the results are incorrect.

Please make a separate post about the use of docker.

Best wishes,
Bob

3

@ShaozhuoLiu - FYI, I made a forum post here about pulling the docker container for older versions of ASPECT.

Cheers,
John

4

Hi Bob,

Thank you for your quick answer.

Oh! The output, as shown by ‘viscosity’ in ParaView, is the effective viscosity. We get the correct effective viscosities after time step 1 to 15, which is viscous viscosity multiplying time step divided by time step and characteristic time (3e20 * 10/(10+3e20/1e10/365/24/3600) Pa s ==3.12285e18 Pa s). Still, the effective viscosity of 2.85534e19 Pa s at time 0 (Time = 0 in ParaView) is not understandable. At time 0, the viscoelastic stress has not been updated, and hence we don’t need to update the effective viscosity.

Best,
SZ

5

Hi John,

Thank you so much!

So, the issue lies in missing ‘v’ in “docker pull geodynamics/aspect:2.5.0”.

Best,
SZ

6

Dear SZ,

We’re making some updates to ASPECT that will hopefully clarify the behaviour in Timestep 0, but the difference in resulting strain is usually negligible; I wouldn’t worry about it. For what it’s worth, the starting condition for viscoelastic models is always somewhat artificial anyway (an unstressed initial state is highly unlikely in the Earth).

Let us know if you still have any questions about this.

Best wishes,
Bob