Example of locating a saddlepoint with the Newton-Raphson method

 

Remember that Newton-Raphson geometry optimization methods will converge to a saddlepoint if the starting structure is close to it (it should find the nearest stationary point). 

 

For example, for the function

 

                (1)

 

there are minima at (1,0) and (-1,0). 

 

·                 We can use the Newton-Raphson method to step towards the saddlepoint between these minima if we start fairly close to it. 

 

·                 We need the second derivative (Hessian) matrix), which is:

 

              (2)

 

At the point (x,y) = (1/4, 0):

 

                              (3)

So the inverse Hessian is:

 

                            (4)

 

The gradient (first derivatives) is:

 

                                           (5)

 

Next: optimizing to a saddlepoint using the Newton-Raphson method (cont.)