3

I read, in an exposition of Euler's proof of Fermat's theorem on sums of squares, that the $k$-th order finite forward difference of the function $f(x_i)=x_i^k$, relatively to the nodes $x_i=i$ where $i=1,2,3,...$, defined by the recurrence $\Delta^n f(x_i):=\Delta^{n-1} f(x_{i+1})-\Delta^{n-1} f(x_{i}),\quad \Delta^1 f(x_i)=\Delta f(x_i):=f(x_{i+1})-f(x_i),\quad n=1,2,...$

is always $k!$ for any natural number $i=1,2,3,...$.

For $x_{i+1}=x_i +h$ Wikipedia gives the formula (easily proved by induction) $\Delta^{k}f(x_i)=\sum_{j=0}^k (-1)^j \binom{k}{j}f(x_i+(k-j)h)$, which, in our case where $x_i=i$, $h=1$ and $f(x)=x^k$, gives $\Delta^k f(i)=\sum_{j=0}^k (-1)^j\binom{k}{j} (i+k-j)^k$. But I am not able to use that to prove (I've tried by induction, in vain) that for all $i\in\mathbb{N}_{\geq 1}$, for $f(x)=x^k$, we have $\Delta^k f(i)=k!$. Neither am I able to prove it by recursion because it isn't a linear one...

Has anybody any idea on how to prove it? Thank you so much in advance!

1 Answers1

4

First convince yourself that if $P(x)$ is a polynomial of degree $m$ with leading coefficient $a$ then $P(x+1)-P(x)$ is a polynomial of degree $m-1$ with leading coefficient $ma$. Now argue by induction on the degree.

Gerry Myerson
  • 179,216