Segregated Step
The Segregated Step node () handles settings for one substep of a segregated iteration. This attribute uses a damped version of Newton’s method and can be used together with a Segregated attribute node.
For more background information about the method and termination settings, see The Segregated Solver and Damped Newton Methods.
General
Use the Variables list to specify variables to be solved for in this step.
By default, all field or state components are included. Select Manual from the Components list to specify which fields or states to include.
The variables are the ones that appear under the Dependent Variables node, and the components are the components listed by each of these (for instance, u is the variable and u, v, and w are the components for a Solid Mechanics model); see also the Field and State nodes.
Select a Linear solver for the linear systems associated with the quantities specified by Variables. The available solvers are attribute nodes of the types Direct and Iterative.
Jacobian matrices requested by the linear solver during solution can be stored in a sparse or a filled format, which you choose as Sparse or Filled, respectively, from the Matrix format list. The default setting is Automatic. In this case, the setting from the Advanced node is used.
In addition, you can choose a matrix-free format, which you choose as Free. The matrix-free representation allows evaluation of matrix-vector products without assembling the matrix. This can reduce the memory usage significantly for solver algorithms that only use the matrix to multiply a vector, specifically:
A warning is issued when the matrix-free format is used with other solvers. COMSOL Multiphysics then assembles the matrix when needed but does not store it between repeated requests from the solver algorithm. This may reduce peak memory usage at the cost of multiple assembly calls.
Method and Termination
See the Fully Coupled node’s Method and Termination section for all settings except for the following, which has a slightly different behavior as described:
For a Time-Dependent Solver, if Constant (Newton) is selected as the Nonlinear method, choose a Jacobian update: Minimal (the default), On every iteration, or Once per time step:
On every iteration computes a new Jacobian for all iterations of Newton’s method.
Minimal updates the Jacobian at least once and then only when the nonlinear solver fails during time stepping. It reuses the Jacobian for several nonlinear systems whenever deemed possible.
Once per time step updates the Jacobian once per time step.
For a Stationary Solver or a parametric solver, if Constant (Newton) is selected as the Nonlinear method, choose a Jacobian update: Minimal (the default), On every iteration, or Once first iteration:
On every iteration computes a new Jacobian for all iterations of Newton’s method.
Minimal updates the Jacobian at least once and then only when the nonlinear solver fails during parameter stepping. It reuses the Jacobian for several nonlinear systems whenever deemed possible.
On first iteration updates the Jacobian for the first subiteration for this segregated step.