EXPLANATION FILE OF PROGRAM NEWTON1
                  ===================================


    Newton's Method
    ---------------


    If the function and its derivative are known analytically, then a simple
  routine can be written for Newton's method. As Hamming notes (see Ref. 16),
  when Newton's method closes in on a zero, the number of digits accuracy rough-
  ly doubles with each iteration step. This is termed quadratic convergence.
  However, there are conditions under which the direct application of Newton's
  method does not converge. This may seem confusing in that we used the idea of
  maximizing the convergence rate to obtain Newton's method. We can examine the
  problem by recalling that, for Newton's method, 

                                     y
                        g(x) = x - -----
                                   dy/dx

  The derivative of g(x) is

                    dg(x)        dy/dx     y(d²y/dx²)   y(d²y/dx²)
                    ----- = 1 - -------- + ---------- = ----------   (6.8.1)
                    d(x)         dy/dx      (dy/dx)²     (dy/dx)²

  If y(x) is nearly a straight line near the location of the zero, we have dy/dx
  =0, and convergence is guaranteed. However, if there is considerable curvature
  to y(x) [d²y/dx² >= (dy/dx)²/y], the method may converge poorly, or sirnply
  diverge. 

    Although Newton's method is very good when it works, there is a problem in
  making it converge quickly. The convergence rate problem can usually be over-
  come if the starting guess is sufficiently near the root so that dg/dx = 0 in
  equation (6.8.1). Thus, Newton's method is most applicable as the final rou-
  tine in finding a root, where the first routine approximately locates the po-
  sition of that root before-hand. 

    The Newton's method has been implemented in program Newton1. The variables
  passed to the subroutine are the initial guess, XO, the maximum number of ite-
  rations, M, and the desired accuracy, E. Returned is the approximate root, X,
  the number of iterations, N, and the last value of Y calculated. The subrou-
  tine itself calls another routine which contains the evaluation of y(x), Y,
  and its derivative, YI. An example of a function subroutine is given in the 
  program listing.


  From [BIBLI 01].
-----------------------------------------
End of file Newton1.txt