You are on page 1of 11

Fourier series

In mathematics, a Fourier series (English pronunciation: a superposition of the corresponding eigensolutions. This
/frie/) is a way to represent a (wave-like) function as superposition or linear combination is called the Fourier
the sum of simple sine waves. More formally, it decom- series.
poses any periodic function or periodic signal into the From a modern point of view, Fouriers results are some-
sum of a (possibly innite) set of simple oscillating func-
what informal, due to the lack of a precise notion of
tions, namely sines and cosines (or, equivalently, complex function and integral in the early nineteenth century.
exponentials). The discrete-time Fourier transform is a Later, Peter Gustav Lejeune Dirichlet[3] and Bernhard
periodic function, often dened in terms of a Fourier se- Riemann[4][5][6] expressed Fouriers results with greater
ries. The Z-transform, another example of application, precision and formality.
reduces to a Fourier series for the important case |z|=1.
Fourier series are also central to the original proof of Although the original motivation was to solve the heat
the NyquistShannon sampling theorem. The study of equation, it later became obvious that the same tech-
Fourier series is a branch of Fourier analysis. niques could be applied to a wide array of mathematical
and physical problems, and especially those involving lin-
ear dierential equations with constant coecients, for
which the eigensolutions are sinusoids. The Fourier se-
1 History ries has many such applications in electrical engineering,
vibration analysis, acoustics, optics, signal processing,
See also: Fourier analysis History image processing, quantum mechanics, econometrics,[7]
thin-walled shell theory,[8] etc.
The Fourier series is named in honour of Jean-Baptiste
Joseph Fourier (17681830), who made important con-
tributions to the study of trigonometric series, after 2 Denition
preliminary investigations by Leonhard Euler, Jean le
Rond d'Alembert, and Daniel Bernoulli.[nb 1] Fourier in-
troduced the series for the purpose of solving the heat In this section, s(x) denotes a function of the real variable
equation in a metal plate, publishing his initial results in x, and s is integrable on an interval [x0 , x0 + P], for real
his 1807 Mmoire sur la propagation de la chaleur dans numbers x0 and P. We will attempt to represent s in that
les corps solides (Treatise on the propagation of heat in interval as an innite sum, or series, of harmonically re-
solid bodies), and publishing his Thorie analytique de la lated sinusoidal functions. Outside the interval, the series
chaleur (Analytical theory of heat) in 1822. The Mmoire is periodic with period P (frequency 1/P). It follows that
introduced Fourier analysis, specically Fourier series. if s also has that property, the approximation is valid on
Through Fouriers research the fact was established that the entire real line. We can begin with a nite summation
an arbitrary (continuous)[1] function can be represented (or partial sum):
by a trigonometric series. The rst announcement of this
great discovery was made by Fourier in 1807, before the
French Academy.[2] Early ideas of decomposing a pe- A0

riodic function into the sum of simple oscillating func- sN (x) = + An sin( 2nx
P +n ), integer for N 1.
2 n=1
tions date back to the 3rd century BC, when ancient as-
tronomers proposed an empiric model of planetary mo-
sN (x) is a periodic function with period P. Using the
tions, based on deferents and epicycles.
The heat equation is a partial dierential equation. Prior
to Fouriers work, no solution to the heat equation was
known in the general case, although particular solutions sin( 2nx + n ) sin(n ) cos( 2nx ) + cos(n ) sin( 2nx )
were known if the heat source behaved in a simple way, { ( )} ( ) (
1 i 2nx + 1 i
P +n +
in particular, if the heat source was a sine or cosine sin( 2nx
+ n ) Re e P n
= e
P i 2i 2i
wave. These simple solutions are now sometimes called
eigensolutions. Fouriers idea was to model a complicated
heat source as a superposition (or linear combination) of we can also write the function in these equivalent forms:
simple sine and cosine waves, and to write the solution as where:


This is the same formula as before except c and c are

no longer complex conjugates. The formula for c is also

x0 +P x0 +P
1 i
2nx 1 2
cn = Re{s(x)} e P dx + i Im{s(x)} ei
P x0 P x0
x0 +P x0 +P
1 2nx 1
= (Re{s(x)} + i Im{s(x)}) ei P dx =
P x0 P x0

Function s(x) (in red) is a sum of six sine functions of dierent

amplitudes and harmonically related frequencies. Their sum- 2.2 Convergence
mation is called a Fourier series. The Fourier transform, S(f) (in
blue), which depicts amplitude vs frequency, reveals the 6 fre-
In engineering applications, the Fourier series is generally
quencies (at odd harmonics) and their amplitudes (1/odd num-
presumed to converge everywhere except at discontinu-
ities, since the functions encountered in engineering are
more well behaved than the ones that mathematicians can
provide as counter-examples to this presumption. In par-
ticular, the Fourier series converges absolutely and uni-
An in
2i e = 12 (an ibn ) forn > 0 formly to s(x) whenever the derivative of s(x) (which may
cn = 12 A0 = 12 a0 forn = 0 not exist everywhere) is square integrable.[10] If a func-

c tion is square-integrable on the interval [x0 , x0 +P], then
|n| forn < 0.
the Fourier series converges to the function at almost ev-
The inverse relationships between the coecients are: ery point. Convergence of Fourier series also depends on
the nite number of maxima and minima in a function
which is popularly known as one of the Dirichlets con-
dition for Fourier series. See Convergence of Fourier se-
An = a2n + b2n n = atan2 (an , bn ) .
ries. It is possible to dene Fourier coecients for more
When the coecients (known as Fourier coecients) general functions or distributions, in such cases conver-
are computed as follows:[9] gence in norm or weak convergence is usually of interest.

sN (x) approximates s(x) on [x0 , x0 +P ], and the approxi-

mation improves as N . The innite sum, s (x), is
called the Fourier series representation of s.

2.1 Complex-valued functions

Both components of a complex-valued function are real-
valued functions that can be represented by a Fourier se-
ries. The two sets of coecients and the partial sum are
given by:

x0 +P 2nx
CRn = 1
P x0
Re{s(x)} ei P dx and
x0 +P 2nx
CIn = 1
Im{s(x)} ei P dx
P x0 Another
visualisation of an approximation of a square wave

N by taking the rst 2nx
1, 2, 3 and 4 terms of its Fourier
sN (x) = CRn ei P +i CIn ei P = (Cseries. CIn
Rn + i(An ) ei P animation
interactive . can be seen here)
| {z }
n=N n=N n=N
2.4 Example 2: Fouriers motivation 3

an = s(x) cos(nx) dx = 0, n 0.

bn = s(x) sin(nx) dx

2 2
= cos(n) + 2 2 sin(n)
n n
2 (1)n+1
= , n 1.
It can be proven that Fourier series converges to s(x) at
every point x where s is dierentiable, and therefore:

A When x = , the Fourier series converges to 0, which is the

visualisation of an approximation of a sawtooth half-sum of the left- and right-limit of s at x = . This is
wave of the same amplitude and frequency for a particular instance of the Dirichlet theorem for Fourier
comparison series.

2.3 Example 1: a simple Fourier series


3 2 2 3

Plot of the sawtooth wave, a periodic continuation of the linear

function s(x) = x/ on the interval (, ]

Heat distribution in a metal plate, using Fouriers method

This example leads us to a solution to the Basel problem.

2.4 Example 2: Fouriers motivation

Animated plot of the rst ve successive partial Fourier series
The Fourier series expansion of our function in Example
1 looks more complicated than the simple formula s(x) =
We now use the formula above to give a Fourier series x/, so it is not immediately apparent why one would need
expansion of a very simple function. Consider a sawtooth the Fourier series. While there are many applications,
wave Fouriers motivation was in solving the heat equation. For
example, consider a metal plate in the shape of a square
whose side measures meters, with coordinates (x, y)
x [0, ] [0, ]. If there is no heat source within the plate,
s(x) = , for < x < , and if three of the four sides are held at 0 degrees Celsius,

while the fourth side, given by y = , is maintained at
the temperature gradient T(x, ) = x degrees Celsius, for
s(x + 2k) = s(x), for < x < and k Z. x in (0, ), then one can show that the stationary heat
distribution (or the heat distribution after a long period
In this case, the Fourier coecients are given by of time has elapsed) is given by

( )

(1)n+1 sinh(ny)
n ) i2f x
T (x, y) = 2 sin(nx) . F {S(f )} = S[n] f e df,
n sinh(n) n= P
n ) i2f x

Here, sinh is the hyperbolic sine function. This solution = S[n] f e df,
of the heat equation is obtained by multiplying each term n= P
of Eq.1 by sinh(ny)/sinh(n). While our example func-
2nx def
tion s(x) seems to have a needlessly complicated Fourier = S[n] ei P = s (x).
series, the heat distribution T(x, y) is nontrivial. The n=
function T cannot be written as a closed-form expression.
This method of solving the heat problem was made pos- The constructed function S(f) is therefore commonly
sible by Fouriers work. referred to as a Fourier transform, even though the
Fourier integral of a periodic function is not convergent
at the harmonic frequencies.[nb 2]
2.5 Other applications

Another application of this Fourier series is to solve the 3 Beginnings

Basel problem by using Parsevals theorem. The example
generalizes and one may compute (2n), for any positive This immediately gives any coecient ak of the trigono-
integer n. metrical series for (y) for any function which has such
an expansion. It works because if has such an expan-
sion, then (under suitable convergence assumptions) the
2.6 Other common notations integral

The notation cn is inadequate for discussing the Fourier

coecients of several dierent functions. Therefore, it is 1
customarily replaced by a modied form of the function ak = (y) cos(2k + 1) dy
1 2
(s, in this case), such as s or S, and functional notation 1 ( )
y y y y
often replaces subscripting: = a cos cos(2k + 1) + a cos 3 cos(2k + 1) +
1 2 2 2 2

can be carried out term-by-term. But all terms involv-

s (x) = s(n) e i P ing cos(2j + 1) y y
2 cos(2k + 1) 2 for j k vanish when
n= integrated from 1 to 1, leaving only the kth term.

2nx In these few lines, which are close to the modern
= S[n] ej P notation engineering common
formalism used in Fourier series, Fourier revolutionized
both mathematics and physics. Although similar trigono-
In engineering, particularly when the variable x repre- metric series were previously used by Euler, d'Alembert,
sents time, the coecient sequence is called a frequency Daniel Bernoulli and Gauss, Fourier believed that such
domain representation. Square brackets are often used to trigonometric series could represent any arbitrary func-
emphasize that the domain of this function is a discrete tion. In what sense that is actually true is a somewhat
set of frequencies. subtle issue and the attempts over many years to clarify
this idea have led to important discoveries in the theories
Another commonly used frequency domain representa- of convergence, function spaces, and harmonic analysis.
tion uses the Fourier series coecients to modulate a
Dirac comb: When Fourier submitted a later competition essay in
1811, the committee (which included Lagrange, Laplace,
Malus and Legendre, among others) concluded: ...the
( manner in which the author arrives at these equations is

def not exempt of diculties and...his analysis to integrate
S(f ) = S[n] f ,
P them still leaves something to be desired on the score of
generality and even rigour.
where f represents a continuous frequency domain.
When variable x has units of seconds, f has units of
hertz. The teeth of the comb are spaced at multiples 3.1 Birth of harmonic analysis
(i.e. harmonics) of 1/P, which is called the fundamental
frequency. s (x) can be recovered from this representa- Since Fouriers time, many dierent approaches to den-
tion by an inverse Fourier transform: ing and understanding the concept of Fourier series have
4.2 Fourier series of Bravais-lattice-periodic-function 5

been discovered, all of which are consistent with one an- may write any arbitrary vector r in the coordinate-system
other, but each of which emphasizes dierent aspects of of the lattice:
the topic. Some of the more powerful and elegant ap-
proaches are based on mathematical ideas and tools that
a1 a2 a3
were not available at the time Fourier completed his orig- r = x1 + x2 + x3 ,
inal work. Fourier originally dened the Fourier series a1 a2 a3
for real-valued functions of real arguments, and using the where ai = |ai|.
sine and cosine functions as the basis set for the decom-
Thus we can dene a new function,
Many other Fourier-related transforms have since been ( )
dened, extending the initial idea to other applications. g(x , x , x ) := f (r) = f x a1 + x a2 + x a3 .
1 2 3 1 2 3
This general area of inquiry is now sometimes called a1 a2 a3
harmonic analysis. A Fourier series, however, can be
This new function, g(x1 , x2 , x3 ) , is now a function of
used only for periodic functions, or for functions on a
three-variables, each of which has periodicity a1 , a2 , a3
bounded (compact) interval.
respectively: g(x1 , x2 , x3 ) = g(x1 + a1 , x2 , x3 ) =
g(x1 , x2 + a2 , x3 ) = g(x1 , x2 , x3 + a3 ) . If we write
a series for g on the interval [0, a1 ] for x1 , we can dene
4 Extensions the following:

4.1 Fourier series on a square a1

1 m1
hone (m1 , x2 , x3 ) := g(x1 , x2 , x3 )ei2 a1 x1 dx1
a1 0
We can also dene the Fourier series for functions of two
variables x and y in the square [, ] [, ]: And then we can write:

f (x, y) = cj,k eijx eiky , g(x1 , x2 , x3 ) = hone (m1 , x2 , x3 ) ei2 a1 x1
j,kZ(integers) m1 =

Further dening:
cj,k = f (x, y)eijx eiky dx dy.
4 2
Aside from being useful for solving partial dierential htwo (m , m , x ) := 1

1 2 3 hone (m1 , x2 , x3 ) ei2 a2 x2 dx2

equations such as the heat equation, one notable appli- a2 0
cation of Fourier series on the square is in image com-
a2 a1 (
pression. In particular, the jpeg image compression stan- 1 1 m
i2 a 1 x1 +
= dx2 dx1 g(x1 , x2 , x3 ) e 1
dard uses the two-dimensional discrete cosine transform, a2 0 a1 0
which is a Fourier transform using the cosine basis func-
tions. We can write g once again as:

m1 m2
4.2 Fourier series of Bravais-lattice- g(x1 , x2 , x3 ) = htwo (m1 , m2 , x3 )ei2 a1 x1 ei2 a2 x2
periodic-function m1 = m2 =

Finally applying the same for the third coordinate, we de-

The Bravais lattice is dened as the set of vectors of the ne:
1 m3
hthree (m1 , m2 , m3 ) := htwo (m1 , m2 , x3 ) ei2 a3 x3 dx3
R = n1 a1 + n2 a2 + n3 a3 a3 0

where ni are integers and ai are three linearly independent a3 a2 a1

1 1 1
vectors. Assuming we have some function, f(r), such that = dx 3 dx 2 dx1 g(x1 , x2 , x3 )
a3 0 a2 0 a1 0
it obeys the following condition for any Bravais lattice
vector R: f(r) = f(r + R), we could make a Fourier se- We write g as:
ries of it. This kind of function can be, for example, the
eective potential that one electron feels inside a peri-
odic crystal. It is useful to make a Fourier series of the g(x , x , x ) = hthree (m1 , m2 , m3 )ei2 a1 x1 ei2
1 2 3
potential then when applying Blochs theorem. First, we m1 = m2 = m3 =

Re-arranging: in which it just so happens that a1 is parallel to the x

axis, a2 lies in the x-y plane, and a3 has components of
all three axes). ) The denominator is exactly the volume
i2 a 1 xof
1 +the
a2 xprimitive
2 + a x3 unit cell which is enclosed by the three
m m2 m3
g(x1 , x2 , x3 ) = h (m1 , m2 , m3 )e 1 3 .
primitive-vectors a1 , a2 and a3 . In particular, we now
m1 ,m2 ,m3 Z
know that
Now, every reciprocal lattice vector can be written as
K = l1 g1 + l2 g2 + l3 g3 , where li are integers and gi
a1 a2 a3
are the reciprocal lattice vectors, we can use the fact that dx1 dx2 dx3 = dx dy dz.
gi aj = 2ij to calculate that for any arbitrary recipro- a 1 (a2 a3 )
cal lattice vector K and arbitrary vector in space r, their We can write now h(K) as an integral with the traditional
scalar product is: coordinate system over the volume of the primitive cell,
instead of with the x1 , x2 and x3 variables:
( ) ( )
a1 a2 a3 l1 l2 l3
Kr = (l1 g1 + l2 g2 + l3 g3 ) x1 + x2 + x3 = 2 x1 + x2 + x3 .
a1 a2 a3 a 1 a2 a3
h(K) = 1 drf (r) eiKr
a1 (a2 a3 ) C
And so it is clear that in our expansion, the sum is actually
over reciprocal lattice vectors: And C is the primitive unit cell, thus, a1 (a2 a3 ) is the
volume of the primitive unit cell.

f (r) = h(K) eiKr ,
4.3 Hilbert space interpretation
where Main article: Hilbert space

a3 a2 a1 ( In the language) of Hilbert spaces, the set of functions

1 1 1 a1 a2 inxa3 iKr
h(K) = dx3 dx2 dx1 f x1 + {e
x2n =+ex 3 : ne Z}. is an orthonormal basis for the
a3 0 a2 0 a1 0 a1 space
a2 L2 ([,
a3 ]) of square-integrable functions on [,
Assuming ]. This space is actually a Hilbert space with an inner
product given for any two elements f and g by

a1 a2 a3
r = (x, y, z) = x1 + x2 + x3 , 1
a1 a2 a3 def
f, g =
f (x)g(x) dx.
we can solve this system of three linear equations for x,
y, and z in terms of x1 , x2 and x3 in order to calculate The basic Fourier series result for Hilbert spaces can be
the volume element in the original cartesian coordinate written as
system. Once we have x, y, and z in terms of x1 , x2 and
x3 , we can calculate the Jacobian determinant:

f= f, en en .
x x x n=
1 1 1
x y z This corresponds exactly to the complex exponential for-

x x2 mulation given above. The version with sines and cosines
2 x2
is also justied with the Hilbert space interpretation. In-
x y z
deed, the sines and cosines form an orthogonal set:
x x3 x3
x y z
which after some calculation and applying some non- cos(mx) cos(nx) dx = mn , m, n 1,

trivial cross-product identities can be shown to be equal
sin(mx) sin(nx) dx = mn , m, n 1

a1 a2 a3 (where mn is the Kronecker delta), and
a1 (a2 a3 )

(it may be advantageous for the sake of simplifying cal-
culations, to work in such a cartesian coordinate system, cos(mx) sin(nx) dx = 0;

5.1 Compact groups 7

Plancherels theorem.
If c0 , c1 , c2 , . . . are co-
ecients and n= |cn |2 < then there is a
unique function f L2 ([, ]) such that f(n) =
cn for every n.
The rst convolution theorem states that if f and g
are in L1 ([, ]), the Fourier series coecients of
the 2-periodic convolution of f and g are given by:

2 g](n) = 2 f(n) g(n),
[nb 4]
Sines and cosines form an orthonormal set, as illustrated above.
The integral of sine, cosine and their product is zero (green and where:
red areas are equal, and cancel out) when m, n or the functions
are dierent, and pi only if m and n are equal, and the function
used is the same.
( def ( ))
[f 2 g] (x) = f (u) g[pv(x u)]du, and pv(x) =Arg eix
| {z }
furthermore, the sines and cosines are orthogonal to the value principal
constant function 1. An orthonormal basis for L2 ([,])
consisting of real functions is formed by the functions 1 = f (u) g(x u) du, 2 is g(x) when-periodic.

and 2 cos(nx), 2 sin(nx) with n = 1, 2,... The density

of their span is a consequence of the StoneWeierstrass = f (u) g(x u) du, 2 are functions both when2 any over i
theorem, but follows also from the properties of classical 2

kernels like the Fejr kernel. The second convolution theorem states that the
Fourier series coecients of the product of f and
g are given by the discrete convolution of the f and
5 Properties g sequences:

We say that f belongs to C k (T) if f is a 2-periodic

function on R which is k times dierentiable, and its kth [fd
g](n) = [f g](n).
derivative is continuous.

If f is a 2-periodic odd function, then an = 0 for 5.1 Compact groups

all n.
Main articles: Compact group, Lie group, and Peter
If f is a 2-periodic even function, then bn = 0 for Weyl theorem
all n.
If f is integrable, lim|n| f(n) = 0 , One of the interesting properties of the Fourier transform
limn+ an = 0 and limn+ bn = 0. This re- which we have mentioned, is that it carries convolutions
sult is known as the RiemannLebesgue lemma. to pointwise products. If that is the property which we
seek to preserve, one can produce Fourier series on any
A doubly innite sequence {an} in c0 (Z) is the se-
1 compact group. Typical examples include those classical
quence of Fourier coecients of a function in L ([0,
groups that are compact. This generalizes the Fourier
2]) if and only if it is a convolution of two se-
2 [12] transform to all spaces of the form L2 (G), where G is a
quences in (Z) . See
compact group, in such a way that the Fourier transform
If f C 1 (T) , then the Fourier coecients fb (n) carries convolutions to pointwise products. The Fourier
of the derivative f can be expressed in terms of the series exists and converges in similar ways to the [,]
Fourier coecients f(n) of the function f, via the case.
formula fb (n) = inf(n) . An alternative extension to compact groups is the Peter
Weyl theorem, which proves results about representations
If f C k (T) , then fd (k) (n) = (in)k f(n) .
of compact groups analogous to those about nite groups.
In particular, since fd(k) (n) tends to zero, we have

that |n|k f(n) tends to zero, which means that the

Fourier coecients converge to zero faster than the 5.2 Riemannian manifolds
kth power of n.
Main articles: Laplace operator and Riemannian mani-

Parsevals theorem. If f belongs
to L 2
([, ]), then fold
(n)|2 = 1 |f (x)|2 dx .
n= | f 2

6.1 Least squares property

We say that p is a trigonometric polynomial of degree N
when it is of the form

The atomic orbitals of chemistry are spherical harmonics and

can be used to produce Fourier series on the sphere. p(x) = pn einx .

If the domain is not a group, then there is no intrinsi- Note that fN is a trigonometric polynomial of degree N.
cally dened convolution. However, if X is a compact Parsevals theorem implies that
Riemannian manifold, it has a LaplaceBeltrami oper-
ator. The LaplaceBeltrami operator is the dierential Theorem. The trigonometric polynomial
operator that corresponds to Laplace operator for the Rie- fN is the unique best trigonometric polynomial
mannian manifold X. Then, by analogy, one can consider of degree N approximating f(x), in the sense
heat equations on X. Since Fourier arrived at his basis that, for any trigonometric polynomial p fN
by attempting to solve the heat equation, the natural gen- of degree N, we have
eralization is to use the eigensolutions of the Laplace
Beltrami operator as a basis. This generalizes Fourier se- fN f 2 < p f 2 ,
ries to spaces of the type L (X), where X is a Riemannian where the Hilbert space norm is dened as:
manifold. The Fourier series converges in ways similar to
the [, ] case. A typical example is to take X to be the
sphere with the usual metric, in which case the Fourier 1
g2 = |g(x)|2 dx.
basis consists of spherical harmonics. 2

5.3 Locally compact Abelian groups 6.2 Convergence

Main article: Pontryagin duality Main article: Convergence of Fourier series
See also: Gibbs phenomenon
The generalization to compact groups discussed above
does not generalize to noncompact, nonabelian groups. Because of the least squares property, and because of the
However, there is a straightfoward generalization to Lo- completeness of the Fourier basis, we obtain an elemen-
cally Compact Abelian (LCA) groups. tary convergence result.
This generalizes the Fourier transform to L1 (G) or L2 (G), Theorem. If f belongs to L2 ([, ]), then f converges
where G is an LCA group. If G is compact, one also to f in L2 ([, ]), that is, fN f 2 converges to 0 as
obtains a Fourier series, which converges similarly to the N .
[, ] case, but if G is noncompact, one obtains instead
We have already mentioned that if f is continuously dif-
a Fourier integral. This generalization yields the usual
ferentiable, then (i n)f(n) is the nth Fourier coe-
Fourier transform when the underlying locally compact
cient of the derivative f. It follows, essentially from
Abelian group is R.
the CauchySchwarz inequality, that f is absolutely
summable. The sum of this series is a continuous func-
tion, equal to f, since the Fourier series converges in the
6 Approximation and convergence mean to f:
of Fourier series Theorem. If f C 1 (T) , then f converges to f
uniformly (and hence also pointwise.)
An important question for the theory as well as applica-
This result can be proven easily if f is further assumed
tions is that of convergence. In particular, it is often
to be C 2 , since in that case n2 f(n) tends to zero as n
essary in applications to replace the innite series
. More generally, the Fourier series is absolutely
by a nite one,
summable, thus converges uniformly to f, provided that
f satises a Hlder condition of order > . In the
absolutely summable case, the inequality supx |f (x)

fN (x) = inx
f (n)e . fN (x)| |n|>N |f(n)| proves uniform convergence.
Many other results concerning the convergence of Fourier
This is called a partial sum. We would like to know, in series are known, ranging from the moderately simple re-
which sense does fN(x) converge to f(x) as N . sult that the series converges at x if f is dierentiable at x,

to Lennart Carleson's much more sophisticated result that 8 Notes

the Fourier series of an L2 function actually converges
almost everywhere. [1] These three did some important early work on the wave
equation, especially D'Alembert. Eulers work in this
These theorems, and informal variations of them that
area was mostly comtemporaneous/ in collaboration with
don't specify the convergence conditions, are sometimes Bernoulli, although the latter made some independent
referred to generically as Fouriers theorem or the contributions to the theory of waves and vibrations (see
Fourier theorem.[13][14][15][16] here, pg.s 209 & 210, ).
[2] Since the integral dening the Fourier transform of a pe-
riodic function is not convergent, it is necessary to view
6.3 Divergence the periodic function
{ 2nxand} its transform as distributions. In
this sense F ei P is a Dirac delta function, which is
Since Fourier series have such good convergence prop-
an example of a distribution.
erties, many are often surprised by some of the nega-
tive results. For example, the Fourier series of a contin- [3] These words are not strictly Fouriers. Whilst the cited ar-
uous T-periodic function need not converge pointwise. ticle does list the author as Fourier, a footnote indicates
The uniform boundedness principle yields a simple non- that the article was actually written by Poisson (that it was
constructive proof of this fact. not written by Fourier is also clear from the consistent use
of the third person to refer to him) and that it is, for rea-
In 1922, Andrey Kolmogorov published an article ti- sons of historical interest, presented as though it were
tled Une srie de Fourier-Lebesgue divergente presque Fouriers original memoire.
partout in which he gave an example of a Lebesgue-
[4] The scale factor is always equal to the period, 2 in this
integrable function whose Fourier series diverges almost case.
everywhere. He later constructed an example of an inte-
grable function whose Fourier series diverges everywhere
(Katznelson 1976).
9 References
[1] John Stillwell, Logic and Philosophy of mathematics in
7 See also the nineteenth century, Routledge History of Philosophy
Volume VII (2013) p. 204.

ATS theorem [2] Florian Cajori, A History of Mathematics (1893) p. 283.

[3] Lejeune-Dirichlet, P. (1829). "Sur la convergence des
Dirichlet kernel sries trigonomtriques qui servent reprsenter une fonc-
tion arbitraire entre des limites donnes" [On the con-
Discrete Fourier transform vergence of trigonometric series which serve to represent
an arbitrary function between two given limits]. Journal
Fast Fourier transform fr die reine und angewandte Mathematik (in French). 4:
Fejrs theorem [4] Ueber die Darstellbarkeit einer Function durch eine
trigonometrische Reihe [About the representability of a
Fourier analysis function by a trigonometric series]. Habilitationsschrift,
Gttingen; 1854. Abhandlungen der Kniglichen
Fourier sine and cosine series Gesellschaft der Wissenschaften zu Gttingen, vol. 13,
1867. Published posthumously for Riemann by Richard
Dedekind (in German). Archived from the original on 20
Fourier transform
May 2008. Retrieved 19 May 2008.

Gibbs phenomenon [5] D. Mascre, Bernhard Riemann: Posthumous Thesis on

the Representation of Functions by Trigonometric Se-
Laurent series the substitution q = eix transforms a ries (1867). Landmark Writings in Western Mathematics
Fourier series into a Laurent series, or conversely. 16401940, Ivor Grattan-Guinness (ed.); pg. 492. Else-
vier, 20 May 2005. Accessed 7 Dec 2012.
This is used in the q-series expansion of the j-
invariant. [6] Theory of Complex Functions: Readings in Mathematics,
by Reinhold Remmert; pg 29. Springer, 1991. Accessed
Multidimensional transform 7 Dec 2012.
[7] Nerlove, Marc; Grether, David M.; Carvalho, Jose L.
Spectral theory (1995). Analysis of Economic Time Series. Economic The-
ory, Econometrics, and Mathematical Economics. Else-
SturmLiouville theory vier. ISBN 0-12-515751-7.

[8] Flugge, Wilhelm (1957). Statik und Dynamik der Schalen. Walter Rudin (1976). Principles of mathematical
Berlin: Springer-Verlag. analysis (3rd ed.). New York: McGraw-Hill, Inc.
ISBN 0-07-054235-X.
[9] Dorf, Richard C.; Tallarida, Ronald J. (1993-07-15).
Pocket Book of Electrical Engineering Formulas (1 ed.). A. Zygmund (2002). Trigonometric series (third
Boca Raton,FL: CRC Press. pp. 171174. ISBN ed.). Cambridge: Cambridge University Press.
ISBN 0-521-89053-5. The rst edition was pub-
[10] Georgi P. Tolstov (1976). Fourier Series. Courier-Dover. lished in 1935.
ISBN 0-486-63317-9.

[11] Gallica Fourier, Jean-Baptiste-Joseph (17681830).

Oeuvres de Fourier. 1888, pp. 218219 (in French). 10 External links 2007-10-15. Retrieved 2014-08-08. Fourier Series as a prelude
[12] fa.functional analysis - Characterizations of a linear sub-
to the Fourier Transform
space associated with Fourier series. MathOverow.
2010-11-19. Retrieved 2014-08-08. Characterizations of a linear subspace associated
[13] William McC. Siebert (1985). Circuits, signals, and sys- with Fourier series
tems. MIT Press. p. 402. ISBN 978-0-262-19229-3.
An interactive ash tutorial for the Fourier Series
[14] L. Marton; Claire Marton (1990). Advances in Electronics
and Electron Physics. Academic Press. p. 369. ISBN
Phasor Phactory Allows custom control of the har-
978-0-12-014650-5. monic amplitudes for arbitrary terms

[15] Hans Kuzmany (1998). Solid-state spectroscopy. Fourier Series 3D interactive demonstration
Springer. p. 14. ISBN 978-3-540-63913-8. HTML5 and JavaScript webpage: Interactive
Fourier Series demonstration (time, frequency,
[16] Karl H. Pribram; Kunio Yasue; Mari Jibu (1991). Brain magnitude and phase axes in 3D view)
and perception. Lawrence Erlbaum Associates. p. 26.
ISBN 978-0-89859-995-4. Java applet shows Fourier series expansion of an ar-
bitrary function
9.1 Further reading Example problems Examples of computing
Fourier Series
William E. Boyce; Richard C. DiPrima (2005). Ele-
mentary Dierential Equations and Boundary Value Hazewinkel, Michiel, ed. (2001), Fourier series,
Problems (8th ed.). New Jersey: John Wiley & Encyclopedia of Mathematics, Springer, ISBN 978-
Sons, Inc. ISBN 0-471-43338-1. 1-55608-010-4

Joseph Fourier, translated by Alexander Free- Weisstein, Eric W. Fourier Series. MathWorld.
man (published 1822, translated 1878, re-released
2003). The Analytical Theory of Heat. Dover Pub- Fourier Series Module by John H. Mathews
lications. ISBN 0-486-49531-0. Check date values
in: |date= (help) 2003 unabridged republication of Joseph Fourier A site on Fouriers life which was
the 1878 English translation by Alexander Freeman used for the historical section of this article at the
of Fouriers work Thorie Analytique de la Chaleur, Wayback Machine (archived December 5, 2001)
originally published in 1822. 'Fourier Theorem'
Enrique A. Gonzalez-Velasco (1992). Connections
Example of the mechanical generation of a Fourier
in Mathematical Analysis: The Case of Fourier Se-
series to draw a plane curve, Fourier curve tracing
ries. American Mathematical Monthly. 99 (5):
427441. doi:10.2307/2325087.
This article incorporates material from example of Fourier
Katznelson, Yitzhak (1976). An introduction to series on PlanetMath, which is licensed under the Creative
harmonic analysis (Second corrected ed.). New Commons Attribution/Share-Alike License.
York: Dover Publications, Inc. ISBN 0-486-63331-
Felix Klein, Development of mathematics in the 19th
century. Mathsci Press Brookline, Mass, 1979.
Translated by M. Ackerman from Vorlesungen ber
die Entwicklung der Mathematik im 19 Jahrhundert,
Springer, Berlin, 1928.

11 Text and image sources, contributors, and licenses

11.1 Text
Fourier series Source: Contributors: The Anome, Tarquin, PierreAbbat,
DrBob, Michael Hardy, Nixdorf, Delirium, Loisel, Looxix~enwiki, Ahoerstemeier, Stevenj, Drz~enwiki, Charles Matthews, Dysprosia,
Samsara, Robbot, Hankwang, MathMartin, Sverdrup, Prara, Wile E. Heresiarch, Tea2min, Pdenapo, Giftlite, MathKnight, Waltpohl,
Mboverload, CryptoDerk, Knutux, LucasVB, Antandrus, Anythingyouwant, Lumidek, Ribamar, Cvalente, ELApro, Thorwald, Edudobay,
D6, Rich Farmbrough, Gadykozma, DonDiego, Paul August, Bender235, ZeroOne, Gaussmarkov, Maqsarian, Bobo192, Johnteslade, Matt
Britt, La goutte de pluie, Samadam, Cherlin, Hectigo, LutzL, Msh210, Arthena, PAR, RJFJR, Oleg Alexandrov, Tbsmith, Jacobolus, MFH,
Waldir, Palica, Pfalstad, Kesla, BD2412, Chubby Chicken, Kevmitch, Mathbot, Nihiltres, Je02, RexNL, Fresheneesz, Chobot, Krish-
navedala, DVdm, Bgwhite, FrankTobia, YurikBot, Wavelength, Bambaiah, Sceptre, RussBot, Hede2000, Grubber, Archelon, Rsrikanth05,
Theshadow27, LMSchmitt, Gareth Jones, Mccready, Jperret, Bota47, Ms2ger, Tetracube, Petri Krohn, Tyomitch, Raeyin, Cmglee,
Sbyrnes321, SmackBot, Deimos 28, Dav2008, Melchoir, Pgk, Jagged 85, Eskimbot, Skizzik, Oli Filth, Silly rabbit, Metacomet, Nbarth,
Colonies Chris, Bob K, Tamfang, Dreadstar, Eliyak, Jim.belk, Lim Wei Quan, Dicklyon, Waggers, JoeBot, Martin Kozk, AlsatianRain,
Paul Matthews, CRGreathouse, CmdrObot, Jackzhp, Shorespirit, HenningThielemann, Unmitigated Success, Myasuda, BigGoose2006,
A.kverma, Kupirijo, Wrwrwr, ChrisKennedy, Entangledphotons, Quibik, Doug Weller, Robertinventor, Hanche, Kablammo, The Hy-
brid, Jomoal99, Futurebird, Escarbot, BigJohnHenry, Sbandrews, Eleuther, Trlkly, Thenub314, Spamicles, Coee2theorems, Richard
Giuly, VoABot II, AuburnPilot, Hypergeek14, Jaakobou, JJ Harrison, 28421u2232nfenfcenc, Tenniszaz, MartinBot, Mrten Berglund,
Nono64, JonathonReinhart, Lld2006, MistyMorn, Mike.lifeguard, Falquaddoomi, Jacksonwalters, Gombang, Policron, DorganBot, Jux-
tapos99, Natl1, LovaAndriamanjay, PowerWill500, Idioma-bot, VolkovBot, JohnBlackburne, LokiClock, AlnoktaBOT, Bovineboy2008,
Philip Trueman, Beng186, One zero one, Hesam7, DennyColt, Mr. PIM, GirasoleDE, SieBot, Zbvhs, Faradayplank, RSStockdale, OKBot,
Anchor Link Bot, Denisarona, Loren.wilton, ClueBot, Stokito, Mild Bill Hiccup, Excirial, Justin545, PixelBot, Brews ohare, Wikeepedian,
Danielsimonjr, Ant59, DumZiBoT, AlanM1, XLinkBot, Galoisgroupie, Charles Sturm, Simplix, RyanCross, Addbot, Olli Niemitalo,
Fgnievinski, Chris19910, Fluernutter, , LaaknorBot, Ginosbot, Quercus solaris, Lightbot, Legobot, Publicly Visible, Luckas-bot,
Yobot, The Earwig, Sarrus, SwisterTwister, Mlewko, AnomieBOT, Jim1138, Pete463251, Ehsfball78, Citation bot, Taeshadow, And-
mats, PavelSolin, Obersachsebot, Xqbot, Bdmy, Adrian Wiemer, Jhbdel, RibotBOT, AliceNovak, Charithjayanada, Constructive editor,
FrescoBot, Anterior1, Sawomir Biay, PiratePi, Citation bot 1, Lost-n-translation, Alipson, Jonesey95, Ashok567, Tcnuk, Rthimmig,
TobeBot, Lotje, Afreiden, 777sms, Gzorg, Bj norge, Dalba, Kakahw, Fblasqueswiki, Newty23125, Teckcheong, EmausBot, KHamsun,
Tawsifkhan, Dcirovic, Slawekb, , SlimDeli, VishnuHaridas, A Thousand Doors, Maschen, Zueignung, Orange Suede Sofa,
ChuispastonBot, Support.and.Defend, ClueBot NG, Jack Greenmaven, Dylan Moreland, Ilovejersey, Helpful Pixie Bot, Ricardohz, Martin
Berka, BG19bot, Walrus068, MadameBruxelles, Prof McCarthy, Mark Arsten, Taylanmath, Glacialfox, BattyBot, Justincheng12345-bot,
Jdogzz, JYBot, Jiejie9988, Ecce simia, ZX95, Yashkes, Darvii, Federicofuentes, Cupitor, Tentinator, Pavel Bezdek, Friedlicherkoenig, Dp-
sangwal, 968u30u7, DominicPrice, Elenpach, Luocheng99, Loraof, Govkub, Philologick, Kavya l, KasparBot, Mikeyrichardson, ,
Sebasgonky, GreenC bot and Anonymous: 336

11.2 Images
File:AtomicOrbital_n4_l2.png Source: License: CC-
BY-SA-3.0 Contributors: Transferred from en.wikipedia to Commons by Zinder using CommonsHelper. Original artist: The original
uploader was DMacks at English Wikipedia
File:Fourier_Series.svg Source: License: Public domain Con-
tributors: Own work Original artist: Jim.belk
File:Fourier_heat_in_a_plate.png Source: Li-
cense: Public domain Contributors: Own work Original artist: Loisel at English Wikipedia
File:Fourier_series_and_transform.gif Source:
transform.gif License: CC0 Contributors: File:Fourier transform time and frequency domains (small).gif Original artist: Lucas V. Barbosa
File:Fourier_series_integral_identities.gif Source:
identities.gif License: Public domain Contributors: Own work Original artist: LucasVB
File:Fourier_series_sawtooth_wave_circles_animation.gif Source:
series_sawtooth_wave_circles_animation.gif License: CC BY-SA 3.0 Contributors: Own work Original artist: Cmglee
File:Fourier_series_square_wave_circles_animation.gif Source:
series_square_wave_circles_animation.gif License: CC BY-SA 3.0 Contributors: Own work Original artist: Cmglee
File:Periodic_identity_function.gif Source: Li-
cense: Public domain Contributors: No machine-readable source provided. Own work assumed (based on copyright claims). Original artist:
No machine-readable author provided. LucasVB assumed (based on copyright claims).
File:Sawtooth_pi.svg Source: License: CC BY-SA 3.0 Contrib-
utors: Own work Original artist: Krishnavedala
File:SquareWaveFourierArrows,rotated.gif Source:
SquareWaveFourierArrows%2Crotated.gif License: CC0 Contributors:
SquareWaveFourierArrows.gif Original artist: Bob K

11.3 Content license

Creative Commons Attribution-Share Alike 3.0