Newton

Source: solvers/newton.py

The standard Newton-Raphson solver which always takes the ‘full’ Newton step.

Other options

linear_solver — str

The linear solver to use within the nonlinear solver

abs_tol — float · default 1e-10

Absolute tolerance in the convergence criteria

rel_tol — float · default 1e-08

Relative tolerance in the convergence criteria

max_its — int · default 25

Maximum number of iterations allowed before issuing an error/exception

substep_del_tol — float · default 1e-06

Relative Newton-step tolerance gating the relative convergence branch of the masked (substepping) solve: a sub-step is accepted via rel_tol only if its step also satisfies ||du|| < substep_del_tol*(||u||+abs_tol). This bisects instead of freezing a still-moving, non-physical iterate that merely dips under a loose rel_tol threshold when the predictor residual is inflated. Set large to recover the pure relative test. Ignored by the single-shot (non-substepping) solve.