1: \begin{abstract}

2: %

3: A new Jacobian approximation is developed for use in

4: quasi-Newton methods for solving systems of nonlinear

5: equations. The new hypersecant Jacobian approximation is

6: intended for the special case where the evaluation of the

7: functions whose roots are sought dominates the computation time,

8: and additionally the Jacobian is sparse. One example of such

9: a case is the solution of the discretized transport equation

10: to calculate particle and energy fluxes in a fusion plasma.

11: %

12: The hypersecant approximation of the Jacobian is calculated

13: using function values from previous Newton iterations, similarly

14: to the Broyden approximation. Unlike Broyden, the hypersecant

15: Jacobian converges to the finite-difference approximation of

16: the Jacobian. The calculation of the hypersecant Jacobian

17: elements requires solving small, dense linear systems, where

18: the coefficient matrices can be ill-conditioned or even exactly

19: singular. Singular-value decomposition (SVD) is therefore used.

20: %

21: Convergence comparisons of the hypersecant method, the

22: standard Broyden method, and the colored finite differencing

23: of the PETSc SNES solver are presented.

24: %

25: \end{abstract}

26: