(September 13, 2011)

Fourier expansions of polynomials and values of

Paul Garrett garrett@math.umn.edu http://www.math.umn.edu/egarrett/

1. Fourier expansions of polynomials

2. Sample computations
This is an example of the impact of harmonic analysis on number theory. We start from the fact that Fourier
series expansions of sufficiently good functions represent the functions pointwise. More precisely, for present
purposes, it suffices to know that for piecewise-C o periodic functions f , at points xo where f is C o and has
left and right derivatives,
X R1
f (xo ) = fb(n) e2inxo (where fb(n) = 0
e2inx f (x) dx)
nZ

and it is implicit that the two-sided series converges. Applying this to polynomials restricted to [0, 1] gives
a systematic approach to summing series [1] such as

1 1 1 1 1 1 2
(2) = + + + + + + . . . =
12 22 32 42 52 62 6

1 1 1 1 1 1 4
(4) = 4
+ 4 + 4 + 4 + 4 + 4 + ... =
1 2 3 4 5 6 90
an /ns at
P
Small extensions of the same idea evaluate certain more complicated Dirichlet series n1
positive integers of suitable parity.
More elementary methods can produce a few results. For example, from the geometric series

1
= 1 x2 + x4 x6 + x8 x10 + . . .
1 + x2
integrate to obtain
x3 x5 x7
arctan x = x + + ...
3 5 7
Letting x = 1 gives Leibniz result

1 1 1 1
1 + + ... =
3 5 7 9 4
However, elementary methods were powerless against (2).

1. Fourier expansions of polynomials

On the unit interval [0, 1], there are two simple types of functions: polynomials, and exponentials e2inx .
They are very unlike each other, and expressing one in terms of the other produces interesting information.

The sawtooth function

1
f (x) = x [[x]] (where [[x]] is the greatest integer x)
2
[1] Euler summed these series circa 1750, although at first he had more a heuristic than proof. Even a heuristic had
eluded the Bernoullis, and Eulers success was one of the things that made a big impression.

1
 Paul Garrett: Fourier expansions of polynomials and values of (September 13, 2011)

is a linear polynomial made periodic. The subtraction of the 1/2 conveniently makes the 0th Fourier coefficient
0. All the other Fourier coefficients are readily computed by integration by parts:
1 1 Z 1 2inx
1 e2inx
Z 
1 e 1
fb(n) = (x ) e2inx dx = (x ) dx =
0 2 2 2inx 0 0 2inx 2in

This begins an inductive description of a family of polynomials [2] B` (x), arranged to have easily computed
Fourier coefficients,

[1.0.1] Definition: (Beware: not quite compatible with other normalizations of Bernoulli polynomials)
Z 1
1 d
B1 (x) = x Bk (x) = Bk1 (x) Bk (x) dx = 0
2 dx 0

Taking Bk (x) so that its derivative is Bk1 (x) is entirely motivated by wanting an integration by parts to
work nicely: Z 1 Z 1
2inx e2inx 1 e2inx
Bk (x) e dx = [Bk (x) ]0 Bk0 (x) dx
0 2in 0 2inx
Z 1 Z 1 Z 1
1 0 e2inx 1 0 e2inx
= [Bk (x) 1
]0 + Bk (x) dx = Bk (x) dx + Bk0 (x) dx
2in 0 2inx 2in 0 0 2inx
1 1 1 R1
= 0+ k1
= k
(using 0 Bk1 (x) dx = 0)
(2in) 2in (2in)

Since Bk0 (x) = Bk1 (x), the only new thing to be determined in Bk is its constant, which is completely
determined by the integral condition. Thus, this description does unambiguously specify a sequence of
polynomials.

[1.0.2] Remark: In computing Fourier coefficients, we think either of the restriction of Bk (x) to [0, 1], or
of the periodicized version Bk (x [[x]]).

[1.0.3] Claim: The Fourier series of Bk (x) is

1 X e2inx
Bk (x [[x]])
(2i)k nk
06=nZ

For k > 1, Bk (0) = Bk (1).

[1.0.4] Remark: The equality Bk (0) = Bk (1) for k > 1 assures that the periodicized version Bk (x [[x]])
of Bk (x) is continuous. Piecewise-polynomial functions have left and right derivatives everywhere, so the
Fourier series for Bk (x) with k > 1 will converge to Bk (x [[x]]) for all x. Thus, for even k, we evaluate (k)
in terms of Bk (x):

[1.0.5] Corollary:
B2k (0) 2 (2k)
= (for 2k even)
(2i)2k

B2k+1 (0) = 0 (for 2k + 1 > 1 odd)
(For k odd, in the previous claim the n terms cancel.) ///

[2] These polynomials are roughly Bernoulli polynomials. We are not worrying about conventional normalization or
indexing, nor other definitions.

2
 Paul Garrett: Fourier expansions of polynomials and values of (September 13, 2011)

Proof: (of claim) First, the matching at endpoints is easy:

Z 1
Bk (1) Bk (0) = Bk1 (t) dt = 0 (for k > 1)
0

since, by construction, Bk1 (x) has vanishing 0th Fourier coefficient. Integrating by parts, recapitulating
earlier observations,
1 1 Z 1
e2inx e2inx
Z 
2inx
B
bk (n) = Bk (x) e dx = Bk (x) Bk1 (x) dx
0 2in 0 0 2in

Bk (1) Bk (0) 1 1 1
= =
2in (2in)k1 2in (2in)k
bk1 (n) = 1/(2in)k1 .
by the inductive hypothesis that B ///

[1.0.6] Claim: The polynomials Bk (x) can be characterized by a generating function:

t etx
1 + t B1 (x) + t2 B2 (x) + t3 B3 (x) + . . . =
et 1

[1.0.7] Remark: It is wise to say that we consider this as a formal power series in x and t. This is not to
say that the discussion is a mere heuristic! The formal power series ring R[[x, t]] over a coefficient ring R is
a (projective) limit [3]
lim C[x, t]/I n (with ideal I generated by x, t))
n

[1.0.8] Remark: The latter relation shows that the polynomials Bk (x) have rational coefficients. Rationality
also follows from the recursive definition.

Proof: Let
f (t, x) = 1 + t B1 (x) + t2 B2 (x) + t3 B3 (x) + . . .

[3] Recall (!) that a projective limit X of a family . . . X X X is an object with maps X X such that
2 1 0 i
the natural diagram of curvy trangles commutes:

X ... ,X
2
/+ X1 /* X0

and, for all collections Y Xi with commuting

Y ... ,X
2
/+ X1 /* X0

there exists a unique Y X giving commuting

,X /+ /*
= 2 mmm6 X1hhhhhh4 X0
X `A ...
A { { mm hh
{{ mm hhhh
A
A {{m{mmhmhmhmhhhh
hm{mhhh
Y

The most basic naive-category-theory ideas show that the object X (and maps X Xi ) is unique up to unique
isomorphism. This should be very reassuring. Unfortunately, this way of thinking is not as common as it should be,
so one may be put off by this description of formal power series rings. If so, one should correct this weakness, since
such descriptions/explanations are essential in higher mathematics.

3
 Paul Garrett: Fourier expansions of polynomials and values of (September 13, 2011)

Applying a standard idea, differentiate [4] in x, use B`0 (x) = B`1 (x), and see what happens:

f (t, x) = t 1 + t2 B1 (x) + t3 B2 (x) + . . . = t f (t, x)
x

Therefore, [5]
f (t, x) = C(t) etx (for some C(t) C[[t]], independent of x)
1
The conditions [6] 0 Bk (x) dx = 0 give, on one hand,
R

Z 1 Z 1 Z 1 X Z 1
C(t) etx dx = (1 + t B1 (x) + . . .) dx = 1 dt + t` B` (x) dx = 1
0 0 0 `1 0

On the other hand,

1 1
et 1
Z Z
C(t) etx dx = C(t) etx dx = C(t)
0 0 t
Thus,
t
C(t) =
et 1
Therefore,
t etx
1 + t B1 (x) + t2 B2 (x) + . . . =
et 1
as claimed. ///

[1.0.9] Corollary:
2 (2) 2 (4) 2 (6) 2 (8) t t
t2 2
+ t4 4
+ t6 6
+ t8 8
+ ... = 1 t
(2i) (2i) (2i) (2i) 2 e 1

Proof: Subtract 1 + tB1 (x) = 1 + t(x 12 ) from both sides of the identity

t etx
1 + t B1 (x) + t2 B2 (x) + t3 B3 (x) + . . . =
et 1

evaluate at x = 0, and multiply through by 1, noting that, from above, Bk (0) = 0 for odd k > 1, while
B2k (0) = 2(2k)/(2i)2k . ///

[4] When weve dodged convergence issues by saying formal power series ring, we cant talk about differentiation as
the usual sort of limit. Instead, define a derivation D on C[x, t] or C[[x, t]] by requiring that D be a linear map such
that Dx = 1 and Dt = 0, and satisfying Leibniz rule

D(f g) = Df g + f Dg

An induction proves Dxn = nxn1 . A bit of thought shows that the limit-taking version of derivative was not
essential. Some of the properties are essential: linearity, Leibniz rule, annihilation of constants.
[5] The differential equation f (t, x)/x = tf (t, x) has the convergent power series solution etx . Further, the obvious
induction on coefficients proves that, given C(t) C[[t]], there is a unique solution f (t, x) with f (t, 0) = C(t).
[6] These definite integrals can be rewritten as linear functionals on vector spaces of polynomials, again avoiding any
genuine limit-taking.

4
 Paul Garrett: Fourier expansions of polynomials and values of (September 13, 2011)

[1.0.10] Remark: Note that the above gives no information about the (2k + 1) for positive odd integers
2k + 1.

2. Sample computations
Reasonable efficiency allows numerical computation of some small-index polynomials Bk (x).

[2.1] Symmetry/skew-symmetry We discover that it is convenient to express the polynomials B` (x) as

a sum of monomials (x 12 )m rather than xm . The recursive description, starting with B1 (x) = x 12 , can
be written in two steps: use an auxiliary polynomial C` (x) which needs its constant corrected to be B` (x):
Z x Z 1
C` (x) = B`1 (t) dt B` (x) = C` (x) C` (t) dt
1
2 0

21 )n ,
P
Thus, with B`1 (x) = n cn (x

X cn X cn X cn 1 (1)n+2
C` (x) = (x 12 )n+1 B` (x) = (x 12 )n+1
n
n + 1 n
n+1 n
(n + 1)(n + 2) 2n+2

Looking at the beginning of the numerical computation below, we discover that when expressed in monomials
in x 21 , polynomials B` (x) apparently have only odd-degree monomials for ` odd, and only even-degree
monomials for ` even. That is, B` (x) apparently has the symmetry/skew-symmetry

B` (1 x) = (1)` B` (x) (skew/symmetry about the point x = 12 )

Indeed, the constant-term correction vanishes when B`1 (x) consists entirely of even-degree monomials in
x 12 . This gives an induction in ` proving the skew/symmetric property.

1
[2.2] Rewritten recursive description Expressing B`1 (x) as a sum of monomials in x 2, say
1 n
P
B`1 (x) = cn (x 2) , the odd-even property reduces the complexity:
P cn X cn 1

n (x 12 )n+1 n+1 (` even)
n+1 (n + 1)(n + 2) 2
cn (x 12 )n )
P
B` (x) = n (for B`1 (x) =
P cn 1 n+1
(x ) (` odd)

n 2
n+1

[2.3] Values B2` ( 21 ) We saw above that

(1)`+1 (2`)
B2` (0) =
22`1 2`
1 1
Expressing B2` (x) in powers of x 2 makes evaluation at 2 simpler. Thus, rearrange

X ein (1)` X (1)n+1

B2` ( 12 ) = 2`
= 2`1 2`
(2in) 2 n2`
n6=0 n1

In the region of convergence

X (1)n+1 1 1

s
= (s) 2 s (s) = 1 (s)
n 2 2s1
n1

5
 Paul Garrett: Fourier expansions of polynomials and values of (September 13, 2011)

Thus,
(1)` 1

B2` ( 21 ) = 1 2`1 (2`)
22`1 2` 2

[2.4] Numerical determination of B` , ` 6 Numerical computation of the first few B` (x) is easy:
1
listing coefficients of monomials of in x 2 in ascending order, starting with B1 (x),

0 EE 1 EE
EE EE
EE EE
EE EE
EE EE
1 " " 1
24 0 CC 2 EE
CC CC EE
CC C EE
CC CC EE
CC CC EE
! C! " 1
1
0 CC 24 0 EE 6 EE
CC CC EE EE
CC CC EE EE
CC CC E EE EE
CC CC E EE
! ! E " "
7 1 1
6!8 C 0 CC 48 0 EE 24 C
CC CC EE EE CC
CC CC EE EE CC
CC CC EE EE CC
C! CC EE EEE C!
! " "
7 1 1
0 CC 6!8 C 0 EE 144 0 CC
CC CC EE EE C
120 A
AA
CC CC EE EE CC AA
CC CC EE EE CC
CC C! EE EE CC AA
! E" " C!
31 7 1 1
8!12 0 6!16 0 576 0 720

where, as above, the constant terms for even-degree cases are determined by
X cn 1
cn (x 12 )n )
P
constant coefficient for B2k (x) = n+1 (for B2k1 (x) =
n
(n + 1)(n + 2) 2
1 1 1
1+1 = (going from ` = 1 to ` = 2)


(1 + 1)(1 + 2) 2 24

1 1 1 1 7


= (going from ` = 3 to ` = 4)


24 (1 + 1)(1 + 2) 21+1 6 (3 + 1)(3 + 2) 23+1 6! 8
and
7 1 1 1 1 1
+
6! 8 (1 + 1)(1 + 2) 21+1 144 (3 + 1)(3 + 2) 23+1 120 (5 + 1)(5 + 2) 25+1
7 1 1 49 + 21 3
= + =
6! 3 64 6! 64 7! 64 7! 3 64
31
= (going from ` = 5 to ` = 6)
8! 24

[2.5] Values of Putting this together,

1 1 1
(2)
= B2 ( 12 ) = 2 1 (2) =
24 2 2 4 2
which gives
2
(2) =
6
6
 Paul Garrett: Fourier expansions of polynomials and values of (September 13, 2011)

Next,
7 (1)2 1
7 (4)
= B4 ( 12 ) = 221 22 1 221 (4) =
6! 8 2 2 64 4
which gives
4
(4) =
90
Finally,
31 (1)3 1

= B6 ( 12 ) = 231 23 1 231 (6)
8! 24 2 2
which gives
6 6
(6) = =
33 5 7 945

All this was known long ago.