Thank you for your answer. Now I see the possible ways.
I want to use option #4 that you mentioned but with another implementation. Instead of setting the stress to zero and using of stretch to model material failure, can I set the first principle stress to an extremely high value, which is not normal for ice-like material (for example 500 MPa), and use IDAM=2 to model the failure? Will be brittle failure more stable in that way?