Bernoulli Numbers and Faulhaber’s Formula

What is $200^5 + 201^5 + 202^5 + \cdots + 500^5$ ?

Well, if a formula could be found for $F_t(n) = \sum_{k=1}^n k^t$ , then this would just be $F_5(500) - F_5(199)$ . Enter Faulhaber…

I find the following sequence of formulae fascinating:

$F_0(n)$	=	$\sum_{k=1}^n 1$
	=	$n$ ;
$F_1(n)$	=	$\sum_{k=1}^n k$
	=	$n(n+1)/2$
	=	$\frac{1}{2}n^2 + \frac{1}{2}n$
$F_2(n)$	=	$\sum_{k=1}^n k^2$
	=	$n(n+1)(2n+1)/6$
	=	$n(n+\frac{1}{2})(n+1)/12$
	=	$\frac{1}{3}n^3 + \frac{1}{2}n^2 + \frac{1}{6}n$
$F_2(n)$	=	$\sum_{k=1}^n k^2$
	=	$n(n+1)(2n+1)/6$
	=	$n(n+\frac{1}{2})(n+1)/12$
	=	$\frac{1}{3}n^3 + \frac{1}{2}n^2 + \frac{1}{6}n$
$F_3(n)$	=	$\sum_{k=1}^n k^3$
	=	$n^2(n+1)^2/4$
	=	$\frac{1}{4}n^4 + \frac{1}{2}n^3 + \frac{1}{4}n^2$
$F_4(n)$	=	$\sum_{k=1}^n k^4$
	=	$n(n+1)(2n+1)(3n^3+3n-1)/30$
	=	$\frac{1}{5}n^5 + \frac{1}{2}n^4 + \frac{1}{3}n^3 - \frac{1}{30}n$
	=	$n(n+\frac{1}{2})(n+1)\left(n+\dfrac{3+\sqrt{21}}{6}\right)\left(n+\dfrac{3-\sqrt{21}}{6}\right)$
$F_5(n)$	=	$\sum_{k=1}^n k^5$
	=	$n^2(n+1)^2(2n^2+2n-1)/12$
	=	$\frac{1}{6}n^6 + \frac{1}{2}n^5 + \frac{5}{12}n^4 - \frac{1}{12}n^2$

… and so on …

(The right-hand sides may be expressed as polynomials in many forms: with one or many integral polynomial factors over an integer; with one or many monic polynomial factors over an integer; as separated terms with rational coefficients.)

What is fascinating is this: the left-hand sides of these equations have $n$ terms, where $n$ could be very large; but, the right-hand sides have a number of terms fixed by, and no greater than, $t+1$ .

Faulhaber (perhaps; this is disputed) found the following general formula:

$F_t(n) = \sum_{k=1}^n k^t$

$F_{t-1}(n) = \sum_{k=1}^n k^{t-1} = \dfrac{1}{t} \sum_{j=1}^{t} \dbinom{t}{j}B^*_{t-j}n^j$

i.e.

$F_t(n) = \sum_{k=1}^n k^t = \sum_{j=1}^{t+1} c_j n^j$

where

$c_j = \dfrac{1}{t+1} \dbinom{t+1}{j}B^*_{t+1-j}$

where

$\dbinom{s}{j} = \dfrac{s!}{j!(s-j)!}$ is a binomial coefficient, and

$B^*_i = (-1)^i B_i$ where $B_i$ is a Bernoulli number.

Finally, the Bernoulli numbers are generated by the recurrence

$\sum_{k=0}^n \dbinom{n+1}{k}B_k = [n = 0]$

i.e.

$B_0 = 1$ and $\sum_{k=0}^n \dbinom{n+1}{k}B_k = 0 \;\bullet\;\forall n > 0$ .

It so happens that $B_i = 0 \;\bullet\;\forall \text{ odd } i \ge 3$ , so

$B^*_1 = -B_1 = \frac{1}{2}$ ; and
$B^*_i = B_i \;\bullet\;\forall i \ne 1$ .

Here’s a table of the first few (sign-adjusted) Bernoulli numbers:

$i$	$B^*_i$
0	1
1	(+) 1/2
2	1/6	(corrected 2009/10/17)
3	0
4	−1/30
6	1/42
8	−1/30	(again)
10	5/66
12	−691/2730

(Oh, the answer is $F_5(500) - F_5(199)$
$= 2,619,817,708,312,500 - 10,507,333,330,000$
$= 2,609,310,374,982,500$ .)

This entry was posted on Thursday, 23 April 2009 at 00:26 and is filed under Mathematics. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.

6 Responses to “Bernoulli Numbers and Faulhaber’s Formula”

David Woodford Says:
Thursday, 23 April 2009 at 17:46 | Reply
Great Post,
I’m guessing the Bernoulli numbers have other uses as well.

David
- Rob Says:
  Friday, 24 April 2009 at 07:55 | Reply
  Hello David,
  
  Yes, the sequence of Bernoulli numbers crops up all over the place. I first came across it in the context of Taylor series (power series) for trigonometric functions: Taylor series of tan.
  
  Here are a couple of links for Bernoulli numbers:
  on Wikipedia;
  on Mathworld.
  
  Beware that some mathematicians define the Bernoulli numbers differently. (Some use “B_n” to mean B_2n or (−1)^n B_n, for example.)
  
  In fact, you’ll find the name Bernoulli cropping up elsewhere too, because there were many mathematicians in the Bernoulli family.
  
  Some other sequences and series you might wish to read about include the the binomial coefficients, the Eulerian numbers, the Fibonacci sequence and the Catalan numbers. Again, these are numbers that crop up in many places (even in nature).
  
  Thanks for your comment. (It’s the first from someone other than the person who suggested that I should start a blog.)
  
  How did you discover my post please? I’d guess that it was through WordPress rather than Google. I see that you have some useful posts too (also on WordPress—presumably one of these is a mirror of the other); I shall be browsing through those.
  
  Rob.
David Woodford Says:
Friday, 24 April 2009 at 16:25 | Reply
I found it through wordpress, a search for maths I think, though I first arrived at a different page but looked at this one because it looked interesting.

Your links are also usefull.

With my blog I initially set it up on wordpress.com but am now trying to launch a self hosted version.

Thanks,
David
Rob Says:
Saturday, 17 October 2009 at 16:21 | Reply
(I have made a correction [typo; B_2 is +1/6, not -1/6]
and have slightly improved the formatting.)
Interesting Series and Sequences | B. Doyle Says:
Saturday, 2 January 2016 at 17:46 | Reply
[…] about Johann Faulhaber, I will refer you to [1] and the introductory remarks of [2]. [Also see this post for excellent examples and discussion of Faulhaber’s formula and Bernoulli numbers.] Just by […]
- Rob Says:
  Saturday, 2 January 2016 at 22:09 | Reply
  I highly recommend reading B. Doyle’s post in the above link:
  http://aperiodicity.com/2016/01/02/interesting-series-and-sequences/, and the references mentioned there, in particular: D. Knuth’s https://ia601000.us.archive.org/18/items/arxiv-math9207222/math9207222.pdf.
  
  One comment that B. Doyle makes is that B₁ should be −½ rather than +½. Whether B₁ should be ±½ seems to vary across the literature, and seems to be a matter of taste. Sometimes one is preferred over another in order to avoid an awkward (−1)^k factor in a sum (for the sake of the single oddly-indexed Bernoulli number for which it makes a difference: B₁).
  
  As I mentioned in an earlier comment, some authors even prefer to index over just our evenly-indexed values.
  
  As a result, when reading about the Bernoulli numbers, you should take careful note of which variant has been chosen.

Rob Hubbard's Blog