You are on page 1of 5

NEWTON'S FORWARD DIFFERENCE

FORMULA
Making use of forward difference operator and forward difference table
( will be defined a little later) this scheme simplifies the calculations
involved in the polynomial approximation of fuctons which are known at
equally spaced data points. 
Consider the equation of the linear interpolation optained in the earlier
section : 
f1 - f0 f0x1 - f1x0
f(x)P1(x) x
= ax-1b =  + 
x1 - x0 x1 - x0

 
       1 
=  [(x  - x)f0 + (x-x0)f1]
(x1 - x0) 1

 
x1 - x x - x0 x - x0
(f1
-
f0  f
f0) 
+  0
 

x1 - x0 x1 - x0 x1 - x0
x - x0
(f1-
= f0 + 
f0) 
x1 - x0
= f0 + r f0    [ r = (x - x0) / (x1 - x0)  f0 = f1 - f0 ]
since x1 - x0 is the step lenght h, r can be written as (x - x0)/h and will be
between (0, 1).
Error in the linear interpolation :
If  (x)  is the error in the linear interpolation then
(x)  =  P1(x) - f(x)   =   f0 + r(f1 - f0) - f(x)
By Taylor's theorem
f(x)   =   f(x0  +  r h)   =   f0  +  r h f0'   +  1/2 r2 h2 f0'' (f),       x0 < f < x1
f1       =   f(x1)       =    f(x0 + h)    =    f0 + h f0'  + 1/2 h2 f0''(u),                 x0 < u
< x1

(x)  =  f0  +  r ( f0 + h f0' + 1/2 h2 f0'' (u) - f0 )( f0 + r h f0' + 1/2
s2 h2 f0''(t) )
   =  1/2 h2 ( r f0''(u) r2 f0''(t) )
Let us assume that the second derivative of the function is bounded such that
| f ''(x) | < M2and since r < 1 we have | (x) |  <    1/2 h2 ( M2+M2)   =   h2M2
The general formula is very convenient to find the function value at various
points if forward difference at various points are avilable. Similarly the
polynomial approximations of functions of higher degree also can be
expressed in terms of   r  and forward differences of higher order. Instead of
using the method of solving the system as we did earlier it is convenient to
use binomial formulae involving the difference operators to generate the
higher order interpolation formuale.
Newton's forward difference formulae :
Let the function  f  is known at  n+1  equally spaced data points  a = x0 <
x1 < ... <  =  xn = b  in the interval [a,b] as f0, f1, . . . fn. Then the n  the degree
polynomial approximation of  f(x)  can be given as 
n

f(x)      Pn(x)   =   ( i ) f i 
0

i=0
where r = (x-x0 ) / n      x  = x0  +  r h   0 <r <n
r(r - 1) . . . (r -
and (
r    are the binomial coefficients ) =
(r  1,  ( r
 ) =  i + 1)
 for any
i)
defined as integer i > 0
0 i i!
Proof :
To prove that the given result is the n the degree polynomial approximation
of f(x) it is sufficeint to prove that at the node i i.e., at x = xi  the polynomial
approximation Pn(x) gives fi for any tabulated value xi since the
curve f(x) passes through the node points xi,  i = 0, 1, . . ., n
Take r  =  k
     x  =  x0 + r h  =  x0 + k h  = xk 
 
k
k  k 
fk =  ( i )  = f  + k f  + . . . +  ( k ) 
0 0

f0
i=0

 
n
(i )0,  =
the terms after k need not be considered  for  n <
since  i
we can prove this result by mathematical induction.
consider k = 0
0
  hence the result is true for k = 0 and assume that the result is also
f0=   if0  = f0
true upto k = 1, 2, . . . p
i=0

 
p
fp =  f0 +  rf0  + . . . +  (p)  p 
f0
consider
fp+1 =  fp + fp
p p p p p p p
( p )
p+1
f0 +(
p-1 ) f0
p +...+ ( 0 )f 0  +  (
p )  f0
+...+ ( 1 )f 0 + = (
0 )f0
k+1
k+1 
 ( i ) f i 
0

i=0
p+
p p
since (
i ) ( i-1 ) (1 )
+  = 
i
i.e., the nth degree polynomial approximation for f(x) can be written as
 r(r-1)  2  r(r-1) . . . (r - n +1) n
f(x)      Pn(x)   = f0 + rf0 +  f0 + . . . +  f0
2! n!
The formula is called Newton's (Newton-Gregory) forward interpolation
formula.
So if we know the forward difference values of  f at x0 until order n then the
above formula is very easy to use to find the function values of  f at any
non-tabulated value of x in the internal [a,b].The higher order forward
differences can be obtained by making use of forward difference table.
Forward difference table : Consider the function value (xi, fi)  i = 0,1,2,--
5 then the forward difference table is

 
xi   fi fi 2fi 3fi 4fi 5fi

x0   f0          
    f0 = f1- f0        
2f0 = f1- 
x1   f1        
f0
3f0 = 2f1- 
    f1 = f2 - f1      
2f0
2f1 = f2 -  4f0 = 3f1- 
x2  f2      
f1 3f0
3f1 = 2f2 -  5f0 = 4f1- 
    f2 = f3 - f2    
2f1 4f0
 f2 = f3 - 
2
 f1 =  f2 - 
4 3
x3  f3      
f2 3f1
3f2 = 2f3 - 
    f3 = f4 - f3      
2f2
2f3 = f4 - 
x4  f4        
f3
    f4 = f5 - f4        
x5  f5          
Example : If  f(x) is known at the following data points 
xi 0 1 2 3 4
fi 1 7 23 55 109
then find f(0.5) and f(1.5) using Newton's forward difference formula.
Solution :
Forward difference table
xi   fi fi 2fi 3fi 4fi

0  1        
    6      
1  7   10    
    16   6  
2  23   16   0
    32   6  
3  55   22    
    54      
4  109        
(Note : The given data satifies f(x) = x3 + 2x2 + 3x +1, i.e the function is a
third degree polynomial and hence third forward differences are constant by
the result).
By Newton's forward difference formula
 r(r-1)  2  r(r-1)(r-2) 3
f(x)  = f0 + rf0 +  f0  +   f0
2! 3!
at x = 0.5,  r = (x - x0) / h = (0.5 - 0) / 1 = 0.5
0.5(0.5 - 1) x 10 5 0.5(0.5 - 1)(0.5 - 2) x 6 
f(0.5) = 1 + 0.5 x 6 +  + 
 2    6 

         = 1 + 3 + 2.5 x (-0.5) + (-0.25)(-1.5)

         = 3.125

Exact value is 


f(0.5) = (0.5)3 + 2(0.5)2 + 3(0.5) + 1 
          = 0.125 + 0.5 + 1.5 + 1 
          = 3.125

 
Error in the Interpolation :
En(x) = (x - x0)(x - x1) . . .(x - xn)  f(n+1)() / (n+1)!                 x0 <  < xn

So for the Newton's method where the nodel points  xi,  i = 0, 1, . . . n  are


equally spaced, the error is   En(x) = (x - x0)(x - x0 - h) . . .(x - x0 - nh)  f(n+1)
() / (n+1)!

r(r-1). . .(r-n) (n+1) (n+1)


         = h f ()
(n+1)!

 
r
         = ( n+1 ) h(n+1)f(n+1)()

You might also like