Backward Difference Operator

DEFINITION 11.2.10 (First Backward Difference Operator)   The FIRST BACKWARD DIFFERENCE OPERATOR, denoted by $ \nabla,$ is defined as

$\displaystyle \nabla f(x) =f(x)-f(x-h).$

Given the step size $ h,$ note that this formula uses the values at $ x$ and $ x- h,$ the point at the previous step. As it moves in the backward direction, it is called the backward difference operator.

DEFINITION 11.2.11 ( $ r^{\mbox{th}}$ Backward Difference Operator)   The $ r^{\mbox{th}}$ backward difference operator, $ \nabla^r,$ is defined as
$\displaystyle \nabla^{r} f(x)$ $\displaystyle =$ $\displaystyle \nabla^{r-1} f(x)- \nabla^{r-1}f(x-h),
\qquad\qquad r=1,2, \ldots,$  
$\displaystyle {\mbox{with }}$   $\displaystyle \nabla^0 f(x)=f(x).$  

In particular, for $ x=x_k,$ we get

$\displaystyle \nabla y_k =y_k - y_{k-1}\; {\mbox{ and }} \;
\nabla^2 y_k =y_k -2y_{k-1}+y_{k-2}.$

Note that $ \nabla^2 y_k = \Delta^2 y_{k-2}.$

EXAMPLE 11.2.12   Using the tabulated values in Example 11.2.5, find $ \nabla y_4$ and $ \nabla ^3 y_3.$

Solution: We have $ \nabla y_4=y_4-y_3 = 0.49-0.35=0.14,$ and
$\displaystyle \nabla ^3 y_3$ $\displaystyle =$ $\displaystyle \nabla ^2 y_3 -\nabla^2 y_2=(y_3-2y_2+y_1)-(y_2-2y_1+y_0)$  
  $\displaystyle =$ $\displaystyle y_3-3y_2+3y_1-y_0$  
  $\displaystyle =$ $\displaystyle 0.35-3\times 0.26+3\times 0.11-0.05 =-0.15.$  

EXAMPLE 11.2.13   If $ f(x)=x^2+ax+b,$ where $ a$ and $ b$ are real constants, calculate $ \nabla^r f(x).$

Solution: We first calculate $ \nabla f(x)$ as follows:
$\displaystyle \nabla f(x)$ $\displaystyle =$ $\displaystyle f(x)-f(x-h)=\left[x^2 + ax +b
\right]- \left[ (x-h)^2+a(x-h)+b \right]$  
  $\displaystyle =$ $\displaystyle 2xh-h^2 + ah.$  

Now,
$\displaystyle \nabla^2f(x)$ $\displaystyle =$ $\displaystyle \nabla f(x)-\Delta f(x-h)
= [2xh-h^2 + ah]-[2(x-h)h- h^2+ ah] = 2 h^2,$  
$\displaystyle {\mbox{and }} \hspace{0.15in} \nabla ^3f(x)$ $\displaystyle =$ $\displaystyle \nabla^2
f(x)-\nabla^2 f(x)
= 2 h^2- 2 h^2=0.$  

Thus, $ \nabla ^r f(x)=0$   for all $ \,\, r \geq 3.$




Remark 11.2.14   For a set of tabular values, backward difference table in the horizontal form is written as:
$ x_0$ $ y_0$
$ x_{1}$ $ y_{1}$ $ \nabla y_{1}= y_1 - y_{0} $
$ x_2$ $ y_2$ $ \nabla y_2= y_2 - y_1 $ $ \nabla^2 y_2 = \nabla y_2 -
\nabla y_1 $
&vellip#vdots;
$ x_{n-2} $ $ y_{n-2}$ $ \cdots$ $ \cdots$
$ x_{n-1}$ $ y_{n-1}$ $ \nabla y_{n-1} = y_{n-1} - y_{n-2} $ $ \cdots$ $ \cdots$
$ x_n$ $ y_n$ $ \nabla y_n = y_n - y_{n-1} $ $ \nabla^2 y_n = \nabla y_n -
\nabla y_{n-1} $ $ \cdots$ $ \nabla^n y_n =
\nabla^{n-1} y_n - \nabla^{n-1} y_{n-1} $

EXAMPLE 11.2.15   For the following set of tabular values $ (x_i,y_i),$ write the forward and backward difference tables.
$ x_i$ 9 10 11 12 13 14
$ y_i$ 5.0 5.4 6.0 6.8 7.5 8.7

Solution: The forward difference table is written as
$ x$ $ y$ $ \Delta y $ $ \Delta^2 y$ $ \Delta^3 y$ $ \Delta^4 y$ $ \Delta^5 y$
9 5 0.4 = 5.4 - 5 0.2 = 0.6 - 0.4 0= 0.2-0.2 -.3 = -0.3 - 0.0 0.6 = 0.3 - (-0.3)
10 5.4 0.6 0.2 -0.3 0.3  
11 6.0 0.8 -0.1 0.0    
12 6.8 0.7 -0.1      
13 7.5 0.6        
14 8.1          

In the similar manner, the backward difference table is written as follows:

$ x$ $ y$ $ \nabla y $ $ \nabla^2 y$ $ \nabla^3 y$ $ \nabla^4 y$ $ \nabla^5 y$
9 5          
10 5.4 0.4        
11 6 0.6 0.2      
12 6.8 0.8 0.2 0.0    
13 7.5 0.7 -0.1 - 0.3 -0.3  
14 8.1 0.6 -0.1 0.0 0.3 0.6

Observe from the above two tables that $ \;\Delta^3 y_1 =\nabla^3
y_{4}, \;\Delta^2 y_3 =\nabla^2 y_{5} $ , $ \;\Delta^4 y_1 =\nabla^4
y_{5} $ etc.

EXERCISE 11.2.16  
  1. Show that $ \Delta^3 y_4 =\nabla^3 y_7.$
  2. Prove that $ \Delta(\nabla y_k) = \Delta^2 y_{k+1} = \nabla^2 y_{k-1}.$
  3. Obtain $ \nabla^k y_k$ in terms of $ y_0, y_1,y_2,\ldots,y_k.$ Hence show that $ \;\nabla^k y_k = \Delta^k y_0.$

Remark 11.2.17   In general it can be shown that $ \;\Delta^k f(x) =\nabla^k
f(x+kh)$ or $ \;\Delta^k y_m =\nabla^k y_{k+m} $

Remark 11.2.18   In view of the remarks (11.2.8) and (11.2.17) it is obvious that, if $ y=f(x)$ is a polynomial function of degree $ n,$ then $ \nabla^n f(x)$ is constant and $ \nabla^{n+r}f(x)= 0$ for $ r> 0.$

A K Lal 2007-09-12