Temperature solver diverges at small timesteps

Extension of Elmer in computational glaciology
Post Reply
chenry
Posts: 10
Joined: 11 Nov 2020, 19:05
Antispam: Yes

Temperature solver diverges at small timesteps

Post by chenry »

Dear all,

I've been running thermo-mechanically coupled simulations of an ice rise and am having trouble with the temperature solver. The solver diverges even at small time steps (0.01 yrs) and I think high temperature gradients at the grounding line may be the cause.

Here's a brief description of the set up: As initial conditions, I prescribe a vertically linear temperature profile, varying from a prescribed surface temperature to the pressure melting point temperature at the lower surface. Temperature is then allowed to evolve subject to the prescribed surface temperature and the pressure melting point temperature where ice is floating. Where it is grounded, a geothermal heat flux is applied. I have attached the sif file and an image of the temperature at the lower ice surface after 13 time steps.

Many thanks in advance for any help or suggestions.

Best,

Clara
Attachments
Screenshot 2022-08-08 at 11.39.35.png
(653.07 KiB) Not downloaded yet
Forward.sif
(24.01 KiB) Downloaded 141 times
raback
Site Admin
Posts: 4801
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Temperature solver diverges at small timesteps

Post by raback »

Hi

Enforcing Dirichlet conditions with very small timesteps may be problematic as the solution may give rise to unphysically large fluxes. Could you reformulate the problem with "External Temperature" and "Heat Transfer Coefficient". If you don't know what value to use for heat transfer coefficient take d=0.1 m of water (or whatever) and use c=k/d.

-Peter
chenry
Posts: 10
Joined: 11 Nov 2020, 19:05
Antispam: Yes

Re: Temperature solver diverges at small timesteps

Post by chenry »

Hi Peter,

Thanks very much for your suggestion. I ran simulations over night with Neumann rather than Dirichlet conditions and am now longer having issues with divergence.

A note to anyone else having these issues: a small time step was necessary when using Neumann boundary conditions. A large time step (0.5 years) diverged.

Best, Clara
raback
Site Admin
Posts: 4801
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Temperature solver diverges at small timesteps

Post by raback »

Hi Clara,

If possible, Robin BCs are usually most robust as they work for extremely short and long timesteps.

-Peter
Post Reply