In many nonlinear problems, it is possible to take advantage of the system regularity between iterations in order to factorize the system matrix only from time to time. This is the so-called "secant" or “modified Newton" iterative process. This technique is known to work in most problems with potentially significant savings (as the CPU-dominant system solution step is performed less frequently and system assembly can be done at a reduced cost), however, in very rare cases, the secant iteration may direct iterations along a wrong path, therefore, this option is not the default. Note that this technique requires slightly more disk space for storing the factors than is the case for the pure Newton-Raphson scheme, so do not expect memory savings from this technique.