Given n + 1 data points

To make the recursive process of computation clearer, the divided differences can be put in tabular form, where the columns correspond to the value of j above, and each entry in the table is computed from the difference of the entries to its immediate lower left and to its immediate upper left, divided by a difference of corresponding x-values:

Notation edit

Note that the divided difference ${\displaystyle [y_{k},\ldots ,y_{k+j}]}$  depends on the values ${\displaystyle x_{k},\ldots ,x_{k+j}}$  and ${\displaystyle y_{k},\ldots ,y_{k+j}}$ , but the notation hides the dependency on the x-values. If the data points are given by a function f,

Divided differences for ${\displaystyle k=0}$  and the first few values of ${\displaystyle j}$ :

Thus, the table corresponding to these terms upto two columns has the following form:

• Linearity
  $${\displaystyle {\begin{aligned}(f+g)[x_{0},\dots ,x_{n}]&=f[x_{0},\dots ,x_{n}]+g[x_{0},\dots ,x_{n}]\\(\lambda \cdot f)[x_{0},\dots ,x_{n}]&=\lambda \cdot f[x_{0},\dots ,x_{n}]\end{aligned}}}$$

   
• Leibniz rule
  $${\displaystyle (f\cdot g)[x_{0},\dots ,x_{n}]=f[x_{0}]\cdot g[x_{0},\dots ,x_{n}]+f[x_{0},x_{1}]\cdot g[x_{1},\dots ,x_{n}]+\dots +f[x_{0},\dots ,x_{n}]\cdot g[x_{n}]=\sum _{r=0}^{n}f[x_{0},\ldots ,x_{r}]\cdot g[x_{r},\ldots ,x_{n}]}$$

   
• Divided differences are symmetric: If ${\displaystyle \sigma :\{0,\dots ,n\}\to \{0,\dots ,n\}}$  is a permutation then
  $${\displaystyle f[x_{0},\dots ,x_{n}]=f[x_{\sigma (0)},\dots ,x_{\sigma (n)}]}$$

   
• Polynomial interpolation in the Newton form: if ${\displaystyle P}$  is a polynomial function of degree ${\displaystyle \leq n}$ , and ${\displaystyle p[x_{0},\dots ,x_{n}]}$  is the divided difference, then
  $${\displaystyle P_{n-1}(x)=p[x_{0}]+p[x_{0},x_{1}](x-x_{0})+p[x_{0},x_{1},x_{2}](x-x_{0})(x-x_{1})+\cdots +p[x_{0},\ldots ,x_{n}](x-x_{0})(x-x_{1})\cdots (x-x_{n-1})}$$

   
• If ${\displaystyle p}$  is a polynomial function of degree ${\displaystyle <n}$ , then
  $${\displaystyle p[x_{0},\dots ,x_{n}]=0.}$$

   
• Mean value theorem for divided differences: if ${\displaystyle f}$  is n times differentiable, then
  $${\displaystyle f[x_{0},\dots ,x_{n}]={\frac {f^{(n)}(\xi )}{n!}}}$$

   
  for a number ${\displaystyle \xi }$  in the open interval determined by the smallest and largest of the ${\displaystyle x_{k}}$ 's.

The divided difference scheme can be put into an upper triangular matrix:

Then it holds

• ${\displaystyle T_{f+g}(x)=T_{f}(x)+T_{g}(x)}$ 
• ${\displaystyle T_{\lambda f}(x)=\lambda T_{f}(x)}$  if ${\displaystyle \lambda }$  is a scalar
• ${\displaystyle T_{f\cdot g}(x)=T_{f}(x)\cdot T_{g}(x)}$ 
  This follows from the Leibniz rule. It means that multiplication of such matrices is commutative. Summarised, the matrices of divided difference schemes with respect to the same set of nodes x form a commutative ring.
• Since ${\displaystyle T_{f}(x)}$  is a triangular matrix, its eigenvalues are obviously ${\displaystyle f(x_{0}),\dots ,f(x_{n})}$ .
• Let ${\displaystyle \delta _{\xi }}$  be a Kronecker delta-like function, that is
  $${\displaystyle \delta _{\xi }(t)={\begin{cases}1&:t=\xi ,\\0&:{\mbox{else}}.\end{cases}}}$$

   
  Obviously ${\displaystyle f\cdot \delta _{\xi }=f(\xi )\cdot \delta _{\xi }}$ , thus ${\displaystyle \delta _{\xi }}$  is an eigenfunction of the pointwise function multiplication. That is ${\displaystyle T_{\delta _{x_{i}}}(x)}$  is somehow an "eigenmatrix" of ${\displaystyle T_{f}(x)}$ : ${\displaystyle T_{f}(x)\cdot T_{\delta _{x_{i}}}(x)=f(x_{i})\cdot T_{\delta _{x_{i}}}(x)}$ . However, all columns of ${\displaystyle T_{\delta _{x_{i}}}(x)}$  are multiples of each other, the matrix rank of ${\displaystyle T_{\delta _{x_{i}}}(x)}$  is 1. So you can compose the matrix of all eigenvectors of ${\displaystyle T_{f}(x)}$  from the ${\displaystyle i}$ -th column of each ${\displaystyle T_{\delta _{x_{i}}}(x)}$ . Denote the matrix of eigenvectors with ${\displaystyle U(x)}$ . Example
  $${\displaystyle U(x_{0},x_{1},x_{2},x_{3})={\begin{pmatrix}1&{\frac {1}{(x_{1}-x_{0})}}&{\frac {1}{(x_{2}-x_{0})(x_{2}-x_{1})}}&{\frac {1}{(x_{3}-x_{0})(x_{3}-x_{1})(x_{3}-x_{2})}}\\0&1&{\frac {1}{(x_{2}-x_{1})}}&{\frac {1}{(x_{3}-x_{1})(x_{3}-x_{2})}}\\0&0&1&{\frac {1}{(x_{3}-x_{2})}}\\0&0&0&1\end{pmatrix}}}$$

   
  The diagonalization of ${\displaystyle T_{f}(x)}$  can be written as
  $${\displaystyle U(x)\cdot \operatorname {diag} (f(x_{0}),\dots ,f(x_{n}))=T_{f}(x)\cdot U(x).}$$

   

Polynomials and power series edit

The matrix

Now consider increasing the degree of ${\displaystyle p}$  to infinity, i.e. turn the Taylor polynomial into a Taylor series. Let ${\displaystyle f}$  be a function which corresponds to a power series. You can compute the divided difference scheme for ${\displaystyle f}$  by applying the corresponding matrix series to ${\displaystyle J}$ : If

Expanded form edit

With the help of the polynomial function ${\displaystyle \omega (\xi )=(\xi -x_{0})\cdots (\xi -x_{n})}$  this can be written as

Peano form edit

If ${\displaystyle x_{0}<x_{1}<\cdots <x_{n}}$  and ${\displaystyle n\geq 1}$ , the divided differences can be expressed as^[4]

This is a consequence of the Peano kernel theorem; it is called the Peano form of the divided differences and ${\displaystyle B_{n-1}}$  is the Peano kernel for the divided differences, all named after Giuseppe Peano.

Forward and backward differences edit

When the data points are equidistantly distributed we get the special case called forward differences. They are easier to calculate than the more general divided differences.

Given n+1 data points

The relationship between divided differences and forward differences is^[5]

• Difference quotient
• Neville's algorithm
• Polynomial interpolation
• Mean value theorem for divided differences
• Nörlund–Rice integral
• Pascal's triangle

• Louis Melville Milne-Thomson (2000) [1933]. The Calculus of Finite Differences. American Mathematical Soc. Chapter 1: Divided Differences. ISBN 978-0-8218-2107-7.
• Myron B. Allen; Eli L. Isaacson (1998). Numerical Analysis for Applied Science. John Wiley & Sons. Appendix A. ISBN 978-1-118-03027-1.
• Ron Goldman (2002). Pyramid Algorithms: A Dynamic Programming Approach to Curves and Surfaces for Geometric Modeling. Morgan Kaufmann. Chapter 4:Newton Interpolation and Difference Triangles. ISBN 978-0-08-051547-2.

• An implementation in Haskell.