L5 : elliptic PDE (PETSc) solver diverged. Reason: indefinite preconditioning.

Question

Hi,

I have a script that works for monodomain simulation on a 3D block with a planar stimulus. For some reason when I try to run bidomain simulations instead of monodomain the following error occurs during prepacing:

L5 : elliptic PDE (PETSc) solver diverged. Reason: indefinite preconditioning.

What is the problem here and do you have any suggestions on what I could try to fix it?

Thank you a lot in advance!

Nils

Comments

Hi Nils,

Please provide a minimal working example that allows reproducing the error.

---

I found out what caused the error, while trying to provide a minimal working example. The block contains a conduction block area that has a low conductivity.  
The selected conductivity value (g=0.00000001) of this area caused the error. Changing the conductivity to a higher value that is still sufficiently low (g=0.0001) to create a conduction block, fixed the problem.

Answer 1

Copying the comment as answer:

I found out what caused the error, while trying to provide a minimal working example. The block contains a conduction block area that has a low conductivity.  
The selected conductivity value (g=0.00000001) of this area caused the error. Changing the conductivity to a higher value that is still sufficiently low (g=0.0001) to create a conduction block, fixed the problem.

Answer 2

Hi!

Extreme conductivity heterogeneities will cause problems when solving the elliptic PDE. Even if the solver doesnt crash, convergence will be reduced. Conduction blocks can be more easily created by turning of Na in the ionic model parameters (GNa*0).

Cheers, Aurel