Theory and Proof

for

Halley's Method

 

Background

Definition (Order of a Root)
  Assume that  f(x)  and its derivatives  [Graphics:Images/Halley'sMethodProof_gr_1.gif]  are defined and continuous on an interval about  [Graphics:Images/Halley'sMethodProof_gr_2.gif].  We say that  [Graphics:Images/Halley'sMethodProof_gr_3.gif]  has a root of order  m  at  [Graphics:Images/Halley'sMethodProof_gr_4.gif]  if and only if  

    
[Graphics:Images/Halley'sMethodProof_gr_5.gif].  

A root of order  [Graphics:Images/Halley'sMethodProof_gr_6.gif]  is often called a simple root, and if  [Graphics:Images/Halley'sMethodProof_gr_7.gif]  it is called a multiple root.  A root of order  [Graphics:Images/Halley'sMethodProof_gr_8.gif].  is sometimes called a double root, and so on.  The next result will illuminate these concepts.

 

Definition (Order of Convergence)  Assume that  [Graphics:Images/Halley'sMethodProof_gr_9.gif] converges to  p,  and set  [Graphics:Images/Halley'sMethodProof_gr_10.gif].    If two positive constants  [Graphics:Images/Halley'sMethodProof_gr_11.gif]  exist, and  

    
[Graphics:Images/Halley'sMethodProof_gr_12.gif]

then the sequence is said to converge to  p  with
order of convergence R.  The number  A  is called the asymptotic error constant.  The cases  [Graphics:Images/Halley'sMethodProof_gr_13.gif]  are given special  consideration.

(i)   If  [Graphics:Images/Halley'sMethodProof_gr_14.gif], the convergence of  [Graphics:Images/Halley'sMethodProof_gr_15.gif]  is called linear.

(ii)  If  [Graphics:Images/Halley'sMethodProof_gr_16.gif], the convergence of  [Graphics:Images/Halley'sMethodProof_gr_17.gif]  is called quadratic.

(ii)  If  [Graphics:Images/Halley'sMethodProof_gr_18.gif], the convergence of  [Graphics:Images/Halley'sMethodProof_gr_19.gif]  is called cubic.

 

Theorem (Newton-Raphson Iteration).  Assume that [Graphics:Images/Halley'sMethodProof_gr_20.gif] and there exists a number [Graphics:Images/Halley'sMethodProof_gr_21.gif], where [Graphics:Images/Halley'sMethodProof_gr_22.gif].  If   [Graphics:Images/Halley'sMethodProof_gr_23.gif], then there exists a [Graphics:Images/Halley'sMethodProof_gr_24.gif] such that the sequence [Graphics:Images/Halley'sMethodProof_gr_25.gif] defined by the iteration  

        [Graphics:Images/Halley'sMethodProof_gr_26.gif]    for    [Graphics:Images/Halley'sMethodProof_gr_27.gif]  

will converge to [Graphics:Images/Halley'sMethodProof_gr_28.gif] for any initial approximation  [Graphics:Images/Halley'sMethodProof_gr_29.gif].  

Furthermore, if  [Graphics:Images/Halley'sMethodProof_gr_30.gif] is a simple root, then  [Graphics:Images/Halley'sMethodProof_gr_31.gif]  will have order of convergence  [Graphics:Images/Halley'sMethodProof_gr_32.gif],  i.e.  [Graphics:Images/Halley'sMethodProof_gr_33.gif].

Proof .

    Use the Taylor polynomial of degree  n=2  for  f(x)  expanded about  [Graphics:Images/Halley'sMethodProof_gr_34.gif]:

(1)            
[Graphics:Images/Halley'sMethodProof_gr_35.gif],  

where  [Graphics:Images/Halley'sMethodProof_gr_36.gif]  lies somewhere between  [Graphics:Images/Halley'sMethodProof_gr_37.gif]  and  x.  Since  f(p) = 0  and  [Graphics:Images/Halley'sMethodProof_gr_38.gif],  truncate this formula to the second term and obtain the approximation:

(2)            [Graphics:Images/Halley'sMethodProof_gr_39.gif]


This can be rearranged in the form:

(3)            [Graphics:Images/Halley'sMethodProof_gr_40.gif].

Since we want to iteratively solve for the root, replace  
p  with  [Graphics:Images/Halley'sMethodProof_gr_41.gif]  and  [Graphics:Images/Halley'sMethodProof_gr_42.gif]  with  =   in (3) and then solve for  [Graphics:Images/Halley'sMethodProof_gr_43.gif]  and obtain the Newton-Raphson iteration formula:

(4)            
[Graphics:Images/Halley'sMethodProof_gr_44.gif].  

Q. E. D.

Proof  Newton-Raphson Method  Newton-Raphson Method  

 

Theorem (Convergence Rate for Newton-Raphson Iteration)  Assume that Newton-Raphson iteration produces a sequence  [Graphics:Images/Halley'sMethodProof_gr_45.gif] that converges to the root  p  of the function  [Graphics:Images/Halley'sMethodProof_gr_46.gif].  

If  p  is a simple root, then convergence is quadratic and  [Graphics:Images/Halley'sMethodProof_gr_47.gif]  for  k  sufficiently large.

If  p  is a multiple root of order  m,  then convergence is linear and  [Graphics:Images/Halley'sMethodProof_gr_48.gif]  for  k  sufficiently large.
Proof.

 


    The Newton-Raphson iteration function is

(5)        [Graphics:Images/Halley'sMethodProof_gr_71.gif].

 

Halley's Method

    
It is possible to speed up convergence significantly when the root is simple.  A popular method is known as Halley's method uses the iteration function:  
    

(6)        [Graphics:Images/Halley'sMethodProof_gr_72.gif] ,

The term in brackets in (6) shows where formula (5) is modified to yield (6).    

Theorem (Halley's Iteration).  Assume that [Graphics:Images/Halley'sMethodProof_gr_73.gif] and there exists a number [Graphics:Images/Halley'sMethodProof_gr_74.gif], where [Graphics:Images/Halley'sMethodProof_gr_75.gif].  If   [Graphics:Images/Halley'sMethodProof_gr_76.gif], then there exists a [Graphics:Images/Halley'sMethodProof_gr_77.gif] such that the sequence [Graphics:Images/Halley'sMethodProof_gr_78.gif] defined by the iteration  

        [Graphics:Images/Halley'sMethodProof_gr_79.gif]    for    [Graphics:Images/Halley'sMethodProof_gr_80.gif]  

will converge to [Graphics:Images/Halley'sMethodProof_gr_81.gif] for any initial approximation  [Graphics:Images/Halley'sMethodProof_gr_82.gif].  

Furthermore, if  [Graphics:Images/Halley'sMethodProof_gr_83.gif] is a simple root, then  [Graphics:Images/Halley'sMethodProof_gr_84.gif]  will have order of convergence  [Graphics:Images/Halley'sMethodProof_gr_85.gif],  i.e.  [Graphics:Images/Halley'sMethodProof_gr_86.gif].

Proof.

Use the Taylor polynomial of degree  n=3  for  f(x)  expanded about  [Graphics:Images/Halley'sMethodProof_gr_87.gif]:

(7)        
[Graphics:Images/Halley'sMethodProof_gr_88.gif],  

where  
[Graphics:Images/Halley'sMethodProof_gr_89.gif]  lies somewhere between  [Graphics:Images/Halley'sMethodProof_gr_90.gif]  and  x.  Since  f(p) = 0  and  [Graphics:Images/Halley'sMethodProof_gr_91.gif],  truncate this formula to the third term and obtain the approximation:

(8)        [Graphics:Images/Halley'sMethodProof_gr_92.gif].

This equation is non-linear in the unknown p.  Use the substitution [Graphics:Images/Halley'sMethodProof_gr_93.gif] in the bracketed term in (8).   The result is:

(9)        
[Graphics:Images/Halley'sMethodProof_gr_94.gif].

Then replace the  
[Graphics:Images/Halley'sMethodProof_gr_95.gif]  with  =   and  p  with  [Graphics:Images/Halley'sMethodProof_gr_96.gif]  in (9) and get:

(10)        
[Graphics:Images/Halley'sMethodProof_gr_97.gif],

Now solve for  
[Graphics:Images/Halley'sMethodProof_gr_98.gif]  and obtain the Halley iteration formula:

(11)        
[Graphics:Images/Halley'sMethodProof_gr_99.gif].  

Q. E. D.

 

   

Return to Numerical Methods - Numerical Analysis

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(c) John H. Mathews 2004