Fourier transform

From Wikipedia, the free encyclopedia
  (Redirected from Fourier Transform)
Jump to: navigation, search
A sinusoidal curve.The peak amplitude. Illustration of phase shift Δt.
In the first row is the graph of the unit pulse function f(t) and its Fourier transform \hat{f}(\omega), a function of frequency \omega. Translation (that is, delay) in the time domain goes over to complex phase shifts in the frequency domain. In the second row is shown g(t), the a delayed unit pulse, beside the real and imaginary parts of the Fourier transform. The Fourier transform decomposes a function into eigenfunctions for the group of translations.
Fourier transforms
Continuous Fourier transform
Fourier series
Discrete-time Fourier transform
Discrete Fourier transform
Fourier analysis
Related transforms

The Fourier transform decomposes a function of time (a signal) into the frequencies that make it up, similarly to how a musical chord can be expressed as the amplitude (or loudness) of its constituent notes. The Fourier transform of a function of time itself is a complex-valued function of frequency, whose absolute value represents the amount of that frequency present in the original function, and whose complex argument is the phase offset of the basic sinusoid in that frequency. The Fourier transform is called the frequency domain representation of the original signal. The term Fourier transform refers to both the frequency domain representation and the mathematical operation that associates the frequency domain representation to a function of time. The Fourier transform is not limited to functions of time, but in order to have a unified language, the domain of the original function is commonly referred to as the time domain. For many functions of practical interest one can define an operation that reverses this: the inverse Fourier transformation, also called Fourier synthesis, of a frequency domain representation combines the contributions of all the different frequencies to recover the original function of time.

Linear operations performed in one domain (time or frequency) have corresponding operations in the other domain, which are sometimes easier to perform. The operation of differentiation in the time domain corresponds to multiplication by the frequency,[note 1] so some differential equations are easier to analyze in the frequency domain. Also, convolution in the time domain corresponds to ordinary multiplication in the frequency domain. Concretely, this means that any linear time-invariant system, such as an electronic filter applied to a signal, can be expressed relatively simply as an operation on frequencies.[note 2] So significant simplification is often achieved by transforming time functions to the frequency domain, performing the desired operations, and transforming the result back to time. Harmonic analysis is the systematic study of the relationship between the frequency and time domains, including the kinds of functions or operations that are "simpler" in one or the other, and has deep connections to almost all areas of modern mathematics.

Functions that are localized in the time domain have Fourier transforms that are spread out across the frequency domain and vice versa. The critical case is the Gaussian function, of substantial importance in probability theory and statistics as well as in the study of physical phenomena exhibiting normal distribution (e.g., diffusion), which with appropriate normalizations goes to itself under the Fourier transform. Joseph Fourier introduced the transform in his study of heat transfer, where Gaussian functions appear as solutions of the heat equation.

The Fourier transform can be formally defined as an improper Riemann integral, making it an integral transform, although this definition is not suitable for many applications requiring a more sophisticated integration theory.[note 3] For example, many relatively simple applications use the Dirac delta function, which can be treated formally as if it were a function, but the justification requires a mathematically more sophisticated viewpoint.[1] The Fourier transform can also be generalized to functions of several variables on Euclidean space, sending a function of 3-dimensional space to a function of 3-dimensional momentum (or a function of space and time to a function of 4-momentum). This idea makes the spatial Fourier transform very natural in the study of waves, as well as in quantum mechanics, where it is important to be able to represent wave solutions either as functions either of space or momentum and sometimes both. In general, functions to which Fourier methods are applicable are complex-valued, and possibly vector-valued.[2] Still further generalization is possible to functions on groups, which, besides the original Fourier transform on or n (viewed as groups under addition), notably includes the discrete-time Fourier transform (DTFT, group = ), the discrete Fourier transform (DFT, group = ℤ mod N) and the Fourier series or circular Fourier transform (group = S1, the unit circle ≈ closed finite interval with endpoints identified). The latter is routinely employed to handle periodic functions. The Fast Fourier transform (FFT) is an algorithm for computing the DFT.

Definition[edit]

There are several common conventions for defining the Fourier transform \hat{f} of an integrable function f : \mathbb R \rightarrow \mathbb C (Kaiser 1994, p. 29), (Rahman 2011, p. 11). This article will use the following definition:

\hat{f}(\xi) = \int_{-\infty}^\infty f(x)\ e^{- 2\pi i x \xi}\,dx,   for any real number ξ.

When the independent variable x represents time (with SI unit of seconds), the transform variable ξ represents frequency (in hertz). Under suitable conditions, f is determined by \hat f via the inverse transform:

f(x) = \int_{-\infty}^\infty \hat f(\xi)\ e^{2 \pi i \xi x}\,d\xi,   for any real number x.

The statement that f can be reconstructed from \hat f is known as the Fourier inversion theorem, and was first introduced in Fourier's Analytical Theory of Heat (Fourier 1822, p. 525), (Fourier & Freeman 1878, p. 408), although what would be considered a proof by modern standards was not given until much later (Titchmarsh 1948, p. 1). The functions f and \hat{f} often are referred to as a Fourier integral pair or Fourier transform pair (Rahman 2011, p. 10).

For other common conventions and notations, including using the angular frequency ω instead of the frequency ξ, see Other conventions and Other notations below. The Fourier transform on Euclidean space is treated separately, in which the variable x often represents position and ξ momentum.

Introduction[edit]

In the first frames of the animation, a function f is resolved into Fourier series: a linear combination of sines and cosines (in blue). The component frequencies of these sines and cosines spread across the frequency spectrum, are represented as peaks in the frequency domain (actually Dirac delta functions, shown in the last frames of the animation). The frequency domain representation of the function, \hat{f}, is the collection of these peaks at the frequencies that appear in this resolution of the function.

The motivation for the Fourier transform comes from the study of Fourier series. In the study of Fourier series, complicated but periodic functions are written as the sum of simple waves mathematically represented by sines and cosines. The Fourier transform is an extension of the Fourier series that results when the period of the represented function is lengthened and allowed to approach infinity (Taneja 2008, p. 192).

Due to the properties of sine and cosine, it is possible to recover the amplitude of each wave in a Fourier series using an integral. In many cases it is desirable to use Euler's formula, which states that e2πiθ = cos(2πθ) + i sin(2πθ), to write Fourier series in terms of the basic waves e2πiθ. This has the advantage of simplifying many of the formulas involved, and provides a formulation for Fourier series that more closely resembles the definition followed in this article. Re-writing sines and cosines as complex exponentials makes it necessary for the Fourier coefficients to be complex valued. The usual interpretation of this complex number is that it gives both the amplitude (or size) of the wave present in the function and the phase (or the initial angle) of the wave. These complex exponentials sometimes contain negative "frequencies". If θ is measured in seconds, then the waves e2πiθ and e−2πiθ both complete one cycle per second, but they represent different frequencies in the Fourier transform. Hence, frequency no longer measures the number of cycles per unit time, but is still closely related.

There is a close connection between the definition of Fourier series and the Fourier transform for functions f that are zero outside of an interval. For such a function, we can calculate its Fourier series on any interval that includes the points where f is not identically zero. The Fourier transform is also defined for such a function. As we increase the length of the interval on which we calculate the Fourier series, then the Fourier series coefficients begin to look like the Fourier transform and the sum of the Fourier series of f begins to look like the inverse Fourier transform. To explain this more precisely, suppose that T is large enough so that the interval [−T/2, T/2] contains the interval on which f is not identically zero. Then the n-th series coefficient cn is given by:

c_n = \frac{1}{T} \int_{-T/2}^{T/2} f(x)\ e^{-2\pi i(n/T) x} dx.

Comparing this to the definition of the Fourier transform, it follows that c_n = (1/T)\hat f(n/T) since f(x) is zero outside [−T/2,T/2]. Thus the Fourier coefficients are just the values of the Fourier transform sampled on a grid of width 1/T, multiplied by the grid width 1/T.

Under appropriate conditions, the Fourier series of f will equal the function f. In other words, f can be written:

f(x)=\sum_{n=-\infty}^\infty c_n\ e^{2\pi i(n/T) x} =\sum_{n=-\infty}^\infty \hat{f}(\xi_n)\ e^{2\pi i\xi_n x}\Delta\xi,

where the last sum is simply the first sum rewritten using the definitions ξn = n/T, and Δξ = (n + 1)/Tn/T = 1/T.

This second sum is a Riemann sum, and so by letting T → ∞ it will converge to the integral for the inverse Fourier transform given in the definition section. Under suitable conditions this argument may be made precise (Stein & Shakarchi 2003).

In the study of Fourier series the numbers cn could be thought of as the "amount" of the wave present in the Fourier series of f. Similarly, as seen above, the Fourier transform can be thought of as a function that measures how much of each individual frequency is present in our function f, and we can recombine these waves by using an integral (or "continuous sum") to reproduce the original function.

Example[edit]

The following images provide a visual illustration of how the Fourier transform measures whether a frequency is present in a particular function. The function depicted f(t) = cos(6πt) e−πt2 oscillates at 3 hertz (if t measures seconds) and tends quickly to 0. (The second factor in this equation is an envelope function that shapes the continuous sinusoid into a short pulse. Its general form is a Gaussian function). This function was specially chosen to have a real Fourier transform that can easily be plotted. The first image contains its graph. In order to calculate \hat f(3) we must integrate e−2πi(3t)f(t). The second image shows the plot of the real and imaginary parts of this function. The real part of the integrand is almost always positive, because when f(t) is negative, the real part of e−2πi(3t) is negative as well. Because they oscillate at the same rate, when f(t) is positive, so is the real part of e−2πi(3t). The result is that when you integrate the real part of the integrand you get a relatively large number (in this case 0.5). On the other hand, when you try to measure a frequency that is not present, as in the case when we look at \hat f(5), the integrand oscillates enough so that the integral is very small. The general situation may be a bit more complicated than this, but this in spirit is how the Fourier transform measures how much of an individual frequency is present in a function f(t).

Properties of the Fourier transform[edit]

Here we assume f(x), g(x) and h(x) are integrable functions, are Lebesgue-measurable on the real line, and satisfy:

\int_{-\infty}^\infty |f(x)| \, dx < \infty.

We denote the Fourier transforms of these functions by \hat{f}(\xi) , \hat{g}(\xi)  and  \hat{h}(\xi) respectively.

Basic properties[edit]

The Fourier transform has the following basic properties: (Pinsky 2002).

Linearity
For any complex numbers a and b, if h(x) = af(x) + bg(x), then  \hat{h}(\xi)=a\cdot \hat{f}(\xi) + b\cdot\hat{g}(\xi).
Translation
For any real number x0, if  h(x)=f(x-x_0),  then  \hat{h}(\xi)= e^{-i\,2\pi \,x_0\,\xi }\hat{f}(\xi).
Modulation
For any real number ξ0 if h(x)=e^{i \, 2\pi \, x \,\xi_0}f(x), then  \hat{h}(\xi) = \hat{f}(\xi-\xi_{0}).
Scaling
For a non-zero real number a, if h(x) = f(ax), then  \hat{h}(\xi)=\frac{1}{|a|}\hat{f}\left(\frac{\xi}{a}\right).     The case a = −1 leads to the time-reversal property, which states: if h(x) = f(−x), then \hat{h}(\xi)=\hat{f}(-\xi).
Conjugation
If  h(x)=\overline{f(x)},  then  \hat{h}(\xi) = \overline{\hat{f}(-\xi)}.
In particular, if f is real, then one has the reality condition\hat{f}(-\xi)=\overline{\hat{f}(\xi)}, that is, \hat{f} is a Hermitian function.
And if f is purely imaginary, then  \hat{f}(-\xi)=-\overline{\hat{f}(\xi)}.
Integration
Substituting \xi=0 in the definition, we obtain
\hat{f}(0) = \int_{-\infty}^{\infty} f(x)\,dx.

That is, the evaluation of the Fourier transform in the origin (\xi=0) equals the integral of f all over its domain.

Invertibility and periodicity[edit]

Under suitable conditions on the function f, it can be recovered from its Fourier transform \hat{f}. Indeed, denoting the Fourier transform operator by \mathcal{F}, so \mathcal{F}(f) := \hat{f}, then for suitable functions, applying the Fourier transform twice simply flips the function: \mathcal{F}^2(f)(x) = f(-x),, which can be interpreted as "reversing time". Since reversing time is two-periodic, applying this twice yields \mathcal{F}^4(f) = f, so the Fourier transform operator is four-periodic, and similarly the inverse Fourier transform can be obtained by applying the Fourier transform three times: \mathcal{F}^3(\hat{f}) = f. In particular the Fourier transform is invertible (under suitable conditions).

More precisely, defining the parity operator \mathcal{P} that inverts time, \mathcal{P}[f]\colon t \mapsto f(-t),:

\mathcal{F}^0 = \mathrm{Id}, \qquad \mathcal{F}^1 = \mathcal{F}, \qquad \mathcal{F}^2 = \mathcal{P}, \qquad \mathcal{F}^4 = \mathrm{Id}
\mathcal{F}^3 = \mathcal{F}^{-1} = \mathcal{P} \circ \mathcal{F} = \mathcal{F} \circ \mathcal{P}

These equalities of operators require careful definition of the space of functions in question, defining equality of functions (equality at every point? equality almost everywhere?) and defining equality of operators – that is, defining the topology on the function space and operator space in question. These are not true for all functions, but are true under various conditions, which are the content of the various forms of the Fourier inversion theorem.

This four-fold periodicity of the Fourier transform is similar to a rotation of the plane by 90°, particularly as the two-fold iteration yields a reversal, and in fact this analogy can be made precise. While the Fourier transform can simply be interpreted as switching the time domain and the frequency domain, with the inverse Fourier transform switching them back, more geometrically it can be interpreted as a rotation by 90° in the time–frequency domain (considering time as the x-axis and frequency as the y-axis), and the Fourier transform can be generalized to the fractional Fourier transform, which involves rotations by other angles. This can be further generalized to linear canonical transformations, which can be visualized as the action of the special linear group SL2(R) on the time–frequency plane, with the preserved symplectic form corresponding to the uncertainty principle, below. This approach is particularly studied in signal processing, under time–frequency analysis.

Units and Duality[edit]

In mathematics, one often does not think of any units as being attached to the two variables t and \xi. But in physical applications, \xi must have inverse units to the units of t. For example, if t is measured in seconds, \xi should be in cycles per second for the formulas here to be valid. If the scale of t is changed and t is measured in units of 2\pi seconds, then either \xi must be in the so-called "angular frequency", or one must insert some constant scale factor into some of the formulas. If t is measured in units of length, then \xi must be in inverse length, e.g., wavenumbers. That is to say, there are two copies of the real line: one measured in one set of units, where t ranges, and the other in inverse units to the units of t, and which is the range of \xi. So these are two distinct copies of the real line, and cannot be identified with each other. Therefore the Fourier transform goes from one space of functions to a different space of functions: functions which have a different domain of definition.

In general, \xi must always be taken to be a linear form on the space of ts, which is to say that the second real line is the dual space of the first real line. See the article on linear algebra for a more formal explanation and for more details. This point of view becomes essential in generalisations of the Fourier transform to general symmetry groups, including the case of Fourier series.

That there is no one preferred way (often, one says "no canonical way") to compare the two copies of the real line which are involved in the Fourier transform---fixing the units on one line does not force the scale of the units on the other line---is the reason for the plethora of rival conventions on the definition of the Fourier transform. The various definitions resulting from different choices of units differ by various constants. If the units of t are in seconds but the units of \xi are in angular frequency, then the angular frequency variable is often denoted by one or another Greek letter, for example, \omega = 2\pi \xi is quite common. Thus (writing \hat x_1 for the alternative definition and \hat x for the definition adopted in this article)

\hat x_1(\omega) = \hat x \left({\omega\over2\pi}\right) = \int_{-\infty}^\infty x(t) e^{-i\omega t} dt

as before, but the corresponding alternative inversion formula would then have to be

 x(t) = {1\over{2\pi}}\int_{-\infty}^\infty \hat x_1(\omega) e^{it\omega} d\omega .

To have something involving angular frequency but with greater symmetry between the Fourier transform and the inversion formula, one very often sees still another alternative definition of the Fourier transform, with a factor of \sqrt 2\pi, thus

\hat x_2 (\omega) = {1\over\sqrt{2\pi}}\int_{-\infty}^\infty x(t) e^{-i\omega t} dt ,

and the corresponding inversion formula then has to be

 x(t) = {1\over\sqrt{2\pi}}\int_{-\infty}^\infty \hat x_2(\omega) e^{it\omega } d\omega .

Furthermore, there is no way to fix which square root of negative one will be meant by the symbol i (it makes no sense to speak of "the positive square root" since only real numbers can be positive, similarly it makes no sense to say "rotation counter-clockwise", because until i is chosen, there is no fixed way to draw the complex plane), and hence one occasionally sees the Fourier transform written with i in the exponent instead of -i, and vice versa for the inversion formula, a convention that is equally valid as the one chosen in this article, which is the more usual one.

For example, in probability theory, the characteristic function \phi of the probability density function f of a random variable X of continuous type is defined without a negative sign in the exponential, and since the units of x are ignored, there is no 2\pi either:

\phi (\lambda) = \int_{-\infty}^\infty f(x) e^{i\lambda x} dx.

(In probability theory, and in mathematical statistics, the use of the Fourier—Stieltjes transform is preferred, because so many random variables are not of continuous type, and do not possess a density function, and one must treat discontinuous distribution functions, i.e., measures which possess "atoms".)

From the higher point of view of group characters, which is much more abstract, all these arbitrary choices disappear, as will be explained in the later section of this article, on the notion of the Fourier transform of a function on an Abelian locally compact group.

Uniform continuity and the Riemann–Lebesgue lemma[edit]

The sinc function, which is the Fourier transform of the rectangular function, is bounded and continuous, but not Lebesgue integrable.

The Fourier transform may be defined in some cases for non-integrable functions, but the Fourier transforms of integrable functions have several strong properties.

The Fourier transform, \hat f, of any integrable function f is uniformly continuous and \|\hat{f}\|_{\infty}\leq \|f\|_1 (Katznelson 1976). By the Riemann–Lebesgue lemma (Stein & Weiss 1971),

\hat{f}(\xi)\to 0\text{ as }|\xi|\to \infty.

However, \hat f need not be integrable. For example, the Fourier transform of the rectangular function, which is integrable, is the sinc function, which is not Lebesgue integrable, because its improper integrals behave analogously to the alternating harmonic series, in converging to a sum without being absolutely convergent.

It is not generally possible to write the inverse transform as a Lebesgue integral. However, when both f and \hat f are integrable, the inverse equality

f(x) = \int_{-\infty}^\infty \hat f(\xi) e^{2 i \pi x \xi} \, d\xi

holds almost everywhere. That is, the Fourier transform is injective on L1(R). (But if f is continuous, then equality holds for every x.)

Plancherel theorem and Parseval's theorem[edit]

Let f(x) and g(x) be integrable, and let \hat{f}(\xi) and \hat{g}(\xi) be their Fourier transforms. If f(x) and g(x) are also square-integrable, then we have Parseval's Formula (Rudin 1987, p. 187):

\int_{-\infty}^{\infty} f(x) \overline{g(x)} \,{\rm d}x = \int_{-\infty}^\infty \hat{f}(\xi) \overline{\hat{g}(\xi)} \,d\xi,

where the bar denotes complex conjugation.

The Plancherel theorem, which follows from the above, states that (Rudin 1987, p. 186)

\int_{-\infty}^\infty \left| f(x) \right|^2\,dx = \int_{-\infty}^\infty \left| \hat{f}(\xi) \right|^2\,d\xi.

Plancherel's theorem makes it possible to extend the Fourier transform, by a continuity argument, to a unitary operator on L2(R). On L1(R)∩L2(R), this extension agrees with original Fourier transform defined on L1(R), thus enlarging the domain of the Fourier transform to L1(R) + L2(R) (and consequently to Lp(R) for 1 ≤ p ≤ 2). Plancherel's theorem has the interpretation in the sciences that the Fourier transform preserves the energy of the original quantity. The terminology of these formulas is not quite standardised. Parseval's theorem was proved only for Fourier series, and was first proved by Liapounoff. But Parseval's formula makes sense for the Fourier transform as well, and so even though in the context of the Fourier transform it was proved by Plancherel, it is still often referred to as Parseval's formula, or Parseval's relation, or even Parseval's theorem.

See Pontryagin duality for a general formulation of this concept in the context of locally compact abelian groups.

Poisson summation formula[edit]

The Poisson summation formula says that for an  L^1 function  f,

 \sum_n \hat f(n) = \sum_n f (n).

This formula is true provided that  \hat f is also  L^1 and that either sum converges to a continuous function.

Although it appears, at first sight, to be merely a curiosity, it has proven to be very useful and important in both engineering applications and number theory.

By using the functional properties of the Fourier transform under translation and scaling, the apparently more general formula  \sum_n \hat f( a + n\lambda) =  {\frac 1\lambda } \sum_n f ({ n\over\lambda } )e^{- 2\pi i {n  \over \lambda }  a } follows immediately.

Poisson's summation formula is sometimes reformulated in terms of a distribution which can be constructed from the Dirac delta function. Write  \delta _y for the distribution which sends any test function  f to its value at  y. Then  \sum_n \delta_n is sometimes called the Dirac Comb. The Poisson summation formula is equivalent to saying that the Fourier transform of the Dirac Comb is another Dirac comb. (The normalisation constants used in this article arrange that its Fourier transform is also   \sum_n \delta_n.)

A special case of the Poisson summation formula is, in the engineering literature, called the Nyquist Sampling Theorem. A function  f is said to be "band-limited" if no frequencies outside a finite interval contribute to its Fourier transform, i.e., there exists a cut-off frequency  f_o such that  \hat f(\xi) = 0 unless  -f_o < \xi < f_o. When the second version of the Poisson summation formula is applied to a band-limited function, there is only one term on the left-hand side if we set  \lambda = f_o:

 \hat f (\xi) =     {1\over f_o } \sum_n f ({ n\over f_o  } )e^{- 2\pi i {n  \over f_o  }  \xi }

which is an explicit formula for  \hat f that only depends on the values of  f at a discrete (but infinite) set of points, all of which are multiples of  1\over f_o. In the statistical study of time-series, if  f is a function of time, then looking only at its values at equally spaced points of time is called "sampling."

This theorem has the practical application that if we know the cut-off frequency for the Fourier transform of  f, we can choose a sampling rate  1\over f_o that will guarantee that no information is lost: since  \hat f can be reconstructed from these sampled values, then, by Fourier inversion, so can  f.

Poisson summation, as it is also called, is used in number theory to prove the transformation properties of theta functions and of automorphic forms, and hence the functional equations of zeta functions. (Pioneers in this include Fourier, Poisson, Jacobi, and Riemann.)

In number theory, a slight variant of a function studied by Fourier in connection with the heat equation is studied for applications to modular forms: by putting  q= e^{i\pi \tau } , for  \tau a complex number in the upper half plane, we can define a kind of theta function:

 \theta ( \tau) =  \sum_n q^{n^2}.

The relation between  \theta ( {-1\over\tau}) and  \theta (\tau) turns out to be important for number theory, since this kind of relation is one of the defining properties of a modular form. By choosing  f= e^{-\pi x^2} in the second version of the Poisson summation formula (with a = 0), and using the fact that  \hat f = e^{-\pi \xi ^2} , one gets immediately

  \theta ({-1\over\tau}) = \sqrt{ \tau \over i} \theta (\tau)

by putting  {1 \over \lambda} = \sqrt{ \tau \over i} .

It follows from this that  \theta^8 has a simple transformation property under  \tau \mapsto {-1 \over \tau} and this can be used to prove Jacobi's formula for the number of different ways to express an integer as the sum of eight perfect squares.

A series of mathematicians applying harmonic analysis to number theory, most notably Martin Eichler, Atle Selberg, Robert Langlands, and James Arthur, have generalised the Poisson summation formula to the Fourier transform on non-commutative locally compact reductive algebraic groups  G with a discrete subgroup  \Gamma such that  G/\Gamma has finite volume. For example,  G can be the real points of  GL_n and  \Gamma can be the integral points of  GL_n. In this setting,  G plays the role of the real number line in the classical version of Poisson summation, and  \Gamma plays the role of the integers  n that appear in the sum. The generalised version of Poisson summation is called the Selberg Trace Formula, and has played a role in proving many cases of Artin's conjecture and in Wiles's proof of Fermat's Last Theorem. The left-hand side of (1) becomes a sum over irreducible unitary representations of  G, and is called "the spectral side," while the right-hand side becomes a sum over conjugacy classes of  \Gamma, and is called "the geometric side." The Poisson summation formula is no longer considered a mere curiosity, but rather the archetype for vast developments in harmonic analysis and number theory.

Convolution theorem[edit]

Main article: Convolution theorem

The Fourier transform translates between convolution and multiplication of functions. If f(x) and g(x) are integrable functions with Fourier transforms \hat{f}(\xi) and \hat{g}(\xi) respectively, then the Fourier transform of the convolution is given by the product of the Fourier transforms \hat{f}(\xi) and \hat{g}(\xi) (under other conventions for the definition of the Fourier transform a constant factor may appear).

This means that if:

h(x) = (f*g)(x) = \int_{-\infty}^\infty f(y)g(x - y)\,dy,

where ∗ denotes the convolution operation, then:

\hat{h}(\xi) =  \hat{f}(\xi)\cdot \hat{g}(\xi).

In linear time invariant (LTI) system theory, it is common to interpret g(x) as the impulse response of an LTI system with input f(x) and output h(x), since substituting the unit impulse for f(x) yields h(x) = g(x). In this case, \hat{g}(\xi) represents the frequency response of the system.

Conversely, if f(x) can be decomposed as the product of two square integrable functions p(x) and q(x), then the Fourier transform of f(x) is given by the convolution of the respective Fourier transforms \hat{p}(\xi) and \hat{q}(\xi).

Cross-correlation theorem[edit]

Main article: Cross-correlation

In an analogous manner, it can be shown that if h(x) is the cross-correlation of f(x) and g(x):

h(x)=(f\star g)(x) = \int_{-\infty}^\infty \overline{f(y)}\,g(x+y)\,dy

then the Fourier transform of h(x) is:

\hat{h}(\xi) = \overline{\hat{f}(\xi)} \,\cdot\, \hat{g}(\xi).

As a special case, the autocorrelation of function f(x) is:

h(x)=(f\star f)(x)=\int_{-\infty}^\infty \overline{f(y)}f(x+y)\,dy

for which

\hat{h}(\xi) = \overline{\hat{f}(\xi)}\,\hat{f}(\xi) = |\hat{f}(\xi)|^2.

Eigenfunctions[edit]

One important choice of an orthonormal basis for L2(R) is given by the Hermite functions

{\psi}_n(x) = \frac{2^{1/4}}{\sqrt{n!}} \, e^{-\pi x^2}\mathrm{He}_n(2x\sqrt{\pi}),

where Hen(x) are the "probabilist's" Hermite polynomials, defined by

\mathrm{He}_n(x) = (-1)^n e^{\frac{x^2}{2}}\left(\frac{d}{dx}\right)^n e^{-\frac{x^2}{2}}

Under this convention for the Fourier transform, we have that

 \hat\psi_n(\xi) = (-i)^n {\psi}_n(\xi) .

In other words, the Hermite functions form a complete orthonormal system of eigenfunctions for the Fourier transform on L2(R) (Pinsky 2002). However, this choice of eigenfunctions is not unique. There are only four different eigenvalues of the Fourier transform (±1 and ±i) and any linear combination of eigenfunctions with the same eigenvalue gives another eigenfunction. As a consequence of this, it is possible to decompose L2(R) as a direct sum of four spaces H0, H1, H2, and H3 where the Fourier transform acts on Hek simply by multiplication by ik.

Since the complete set of Hermite functions provides a resolution of the identity, the Fourier transform can be represented by such a sum of terms weighted by the above eigenvalues, and these sums can be explicitly summed. This approach to define the Fourier transform was first done by Norbert Wiener (Duoandikoetxea 2001). Among other properties, Hermite functions decrease exponentially fast in both frequency and time domains, and they are thus used to define a generalization of the Fourier transform, namely the fractional Fourier transform used in time-frequency analysis (Boashash 2003). In physics, this transform was introduced by Edward Condon (Condon 1937).

Fourier transform on Euclidean space[edit]

The Fourier transform can be defined in any arbitrary number of dimensions n. As with the one-dimensional case, there are many conventions. For an integrable function f(x), this article takes the definition:

\hat{f}(\boldsymbol{\xi}) = \mathcal{F}(f)(\boldsymbol{\xi}) = \int_{\R^n} f(\mathbf{x}) e^{-2\pi i \mathbf{x}\cdot\boldsymbol{\xi}} \, d\mathbf{x}

where x and ξ are n-dimensional vectors, and x·ξ is the dot product of the vectors. The dot product is sometimes written as \left\langle \mathbf x, \boldsymbol \xi \right\rangle.

All of the basic properties listed above hold for the n-dimensional Fourier transform, as do Plancherel's and Parseval's theorem. When the function is integrable, the Fourier transform is still uniformly continuous and the Riemann–Lebesgue lemma holds. (Stein & Weiss 1971)

Uncertainty principle[edit]

For more details on this topic, see Gabor limit.

Generally speaking, the more concentrated f(x) is, the more spread out its Fourier transform \hat f(\xi) must be. In particular, the scaling property of the Fourier transform may be seen as saying: if we "squeeze" a function in x, its Fourier transform "stretches out" in ξ. It is not possible to arbitrarily concentrate both a function and its Fourier transform.

The trade-off between the compaction of a function and its Fourier transform can be formalized in the form of an uncertainty principle by viewing a function and its Fourier transform as conjugate variables with respect to the symplectic form on the time–frequency domain: from the point of view of the linear canonical transformation, the Fourier transform is rotation by 90° in the time–frequency domain, and preserves the symplectic form.

Suppose f(x) is an integrable and square-integrable function. Without loss of generality, assume that f(x) is normalized:

\int_{-\infty}^\infty |f(x)|^2 \,dx=1.

It follows from the Plancherel theorem that \hat f(\xi) is also normalized.

The spread around x = 0 may be measured by the dispersion about zero (Pinsky 2002, p. 131) defined by

D_0(f)=\int_{-\infty}^\infty x^2|f(x)|^2\,dx.

In probability terms, this is the second moment of |f(x)|2 about zero.

The Uncertainty principle states that, if f(x) is absolutely continuous and the functions x·f(x) and f′(x) are square integrable, then

D_0(f)D_0(\hat{f}) \geq \frac{1}{16\pi^2}    (Pinsky 2002).

The equality is attained only in the case f(x)=C_1 \, e^{{-\pi x^2}/{\sigma^2}} (hence \hat{f}(\xi)= \sigma C_1 \, e^{-\pi\sigma^2\xi^2}) where σ > 0 is arbitrary and C_1 = \sqrt[4]{2} / \sqrt{\sigma} so that f is L2–normalized (Pinsky 2002). In other words, where f is a (normalized) Gaussian function with variance σ2, centered at zero, and its Fourier transform is a Gaussian function with variance σ−2.

In fact, this inequality implies that:

\left(\int_{-\infty}^\infty (x-x_0)^2|f(x)|^2\,dx\right)\left(\int_{-\infty}^\infty(\xi-\xi_0)^2|\hat{f}(\xi)|^2\,d\xi\right)\geq \frac{1}{16\pi^2}

for any x0, ξ0R  (Stein & Shakarchi 2003, p. 158).

In quantum mechanics, the momentum and position wave functions are Fourier transform pairs, to within a factor of Planck's constant. With this constant properly taken into account, the inequality above becomes the statement of the Heisenberg uncertainty principle (Stein & Shakarchi 2003, p. 158).

A stronger uncertainty principle is the Hirschman uncertainty principle, which is expressed as:

H(|f|^2)+H(|\hat{f}|^2)\ge \log(e/2)

where H(p) is the differential entropy of the probability density function p(x):

H(p) = -\int_{-\infty}^\infty p(x)\log(p(x))dx

where the logarithms may be in any base that is consistent. The equality is attained for a Gaussian, as in the previous case.

Version without complex numbers[edit]

Version Without Complex Numbers

The original version, which some still prefer, of Fourier's integrals did not use complex numbers, but rather sines and cosines. Statisticians and others still use this form. An absolutely integrable function f for which Fourier inversion holds good can be expanded in terms of genuine frequencies (avoiding negative frequencies, which are sometimes considered hard to interpret physically\plainfootnote{Chatfield, The Analysis of Time Series, 6th ed., London, 2004, p. 113.}) \lambda by

f(t) = \int_0^\infty \left[ a(\lambda ) \cos2\pi \lambda t + b(\lambda ) \sin2\pi \lambda t \right] d\lambda.

This is called an expansion as a trigonometric integral, or a Fourier integral expansion. The coefficient functions a and b can be found by using variants of the Fourier cosine transform and the Fourier sine transform (the normalisations are, again, not standardised):

 a (\lambda) = 2\int_{-\infty}^\infty f(t) \cos2\pi\lambda t dt

and

 b (\lambda) = 2\int_{-\infty}^\infty f(t) \sin 2\pi\lambda t dt.

It can be seen that there is no one Fourier transform function (and, indeed, the literature earlier than about 1920 did not usually use the phrase "Fourier transform"), but rather two transform functions, the Fourier cosine transform, a, and the Fourier sine transform, b.

To get around this duplicity, Fourier employed a trigonometric identity to rewrite these equations as

 f(t) = 2\int_0  ^{\infty} \int_{-\infty}^{\infty} f(\tau) \cos 2\pi \lambda(\tau-t) d\tau d\lambda.

This is referred to as Fourier's Integral Formula, (see, e.g., Fourier, Théorie Analytique de la Chaleur, Paris, 1822, p. 441; Poincaré, Théorie Analytique de la Propagation de Chaleur, Paris, 1895, p. 102; Whittaker and Watson, A Course of Modern Analysis, 4th ed., Cambridge, 1927, p. 188), or "Fourier's Integral Theorem" (although it was not proved in this generality by Fourier) and is equivalent to "Fourier Inversion". See Camille Jordan, Cours d'Analyse de l'École Polytechnique, vol. II, Calcul Intégral: Intégrales définies et indéfinies. 2nd ed., Paris, 1883, pp. 216–226, who, in fact, proves Fourier's Integral Theorem before studying Fourier Series. See also Kolmogorov and Fomin, Elementos de la teoría de funciones y del análisis funcional. Moscow, 1972, traducido por Carlos Vega, pp. 466–9.

The operational properties with respect to convolution, differentiation, etc., are awkward to express in this setting, but were well known nevertheless.

The relations between \hat f, a, and b are obvious.

The real part of \hat f is \frac a2, and its imaginary part is b\over 2. (This is because the contribution of the frequency \lambda to f is divided evenly between \hat f(\lambda) and \hat f(-\lambda)---remember, since f is real-valued, \hat f(-\xi) is simply the complex conjugate of \hat f(\xi) and so does not contain any new information.)

If f is even, then its sine transform vanishes, and so does the imaginary part of \hat f, and so \hat f = \frac a2.

If f is odd, then its cosine transform vanishes, and so does the real part of \hat f, and so \hat f = \frac b2.

Spherical harmonics[edit]

Let the set of homogeneous harmonic polynomials of degree k on Rn be denoted by Ak. The set Ak consists of the solid spherical harmonics of degree k. The solid spherical harmonics play a similar role in higher dimensions to the Hermite polynomials in dimension one. Specifically, if f(x) = e−π|x|2P(x) for some P(x) in Ak, then \hat{f}(\xi)=i^{-k}f(\xi). Let the set Hk be the closure in L2(Rn) of linear combinations of functions of the form f(|x|)P(x) where P(x) is in Ak. The space L2(Rn) is then a direct sum of the spaces Hk and the Fourier transform maps each space Hk to itself and is possible to characterize the action of the Fourier transform on each space Hk (Stein & Weiss 1971). Let f(x) = f0(|x|)P(x) (with P(x) in Ak), then \hat{f}(\xi)=F_0(|\xi|)P(\xi) where

F_0(r)=2\pi i^{-k}r^{-(n+2k-2)/2}\int_0^\infty f_0(s)J_{(n+2k-2)/2}(2\pi rs)s^{(n+2k)/2}\,ds.

Here J(n + 2k − 2)/2 denotes the Bessel function of the first kind with order (n + 2k − 2)/2. When k = 0 this gives a useful formula for the Fourier transform of a radial function (Grafakos 2004). Note that this is essentially the Hankel transform. Moreover, there is a simple recursion relating the cases n+2 and n (Grafakos & Teschl 2013) allowing to compute, e.g., the three-dimensional Fourier transform of a radial function from the one-dimensional one.

Restriction problems[edit]

In higher dimensions it becomes interesting to study restriction problems for the Fourier transform. The Fourier transform of an integrable function is continuous and the restriction of this function to any set is defined. But for a square-integrable function the Fourier transform could be a general class of square integrable functions. As such, the restriction of the Fourier transform of an L2(Rn) function cannot be defined on sets of measure 0. It is still an active area of study to understand restriction problems in Lp for 1 < p < 2. Surprisingly, it is possible in some cases to define the restriction of a Fourier transform to a set S, provided S has non-zero curvature. The case when S is the unit sphere in Rn is of particular interest. In this case the Tomas-Stein restriction theorem states that the restriction of the Fourier transform to the unit sphere in Rn is a bounded operator on Lp provided 1 ≤ p(2n + 2) / (n + 3).

One notable difference between the Fourier transform in 1 dimension versus higher dimensions concerns the partial sum operator. Consider an increasing collection of measurable sets ER indexed by R ∈ (0,∞): such as balls of radius R centered at the origin, or cubes of side 2R. For a given integrable function f, consider the function fR defined by:

f_R(x) = \int_{E_R}\hat{f}(\xi) e^{2\pi ix\cdot\xi}\, d\xi, \quad x \in \mathbf{R}^n.

Suppose in addition that fLp(Rn). For n = 1 and 1 < p < ∞, if one takes ER = (−R, R), then fR converges to f in Lp as R tends to infinity, by the boundedness of the Hilbert transform. Naively one may hope the same holds true for n > 1. In the case that ER is taken to be a cube with side length R, then convergence still holds. Another natural candidate is the Euclidean ball ER = {ξ : |ξ| < R}. In order for this partial sum operator to converge, it is necessary that the multiplier for the unit ball be bounded in Lp(Rn). For n ≥ 2 it is a celebrated theorem of Charles Fefferman that the multiplier for the unit ball is never bounded unless p = 2 (Duoandikoetxea 2001). In fact, when p ≠ 2, this shows that not only may fR fail to converge to f in Lp, but for some functions fLp(Rn), fR is not even an element of Lp.

Fourier transform on function spaces[edit]

On Lp spaces[edit]

On L1

The definition of the Fourier transform by the integral formula

\hat{f}(\xi) = \int_{\mathbf{R}^n} f(x)e^{-2\pi i \xi\cdot x}\,dx

is valid for Lebesgue integrable functions f; that is, fL1(Rn).

The Fourier transform \mathcal{F}: L1(Rn) → L(Rn) is a bounded operator. This follows from the observation that

\vert\hat{f}(\xi)\vert \leq \int_{\mathbf{R}^n} \vert f(x)\vert \,dx,

which shows that its operator norm is bounded by 1. Indeed it equals 1, which can be seen, for example, from the transform of the rect function. The image of L1 is a subset of the space C0(Rn) of continuous functions that tend to zero at infinity (the Riemann–Lebesgue lemma), although it is not the entire space. Indeed, there is no simple characterization of the image.

On L2

Since compactly supported smooth functions are integrable and dense in L2(Rn), the Plancherel theorem allows us to extend the definition of the Fourier transform to general functions in L2(Rn) by continuity arguments. The Fourier transform in L2(Rn) is no longer given by an ordinary Lebesgue integral, although it can be computed by an improper integral, here meaning that for an L2 function f,

\hat{f}(\xi) = \lim_{R\to\infty}\int_{|x|\le R} f(x) e^{-2\pi i x\cdot\xi}\,dx

where the limit is taken in the L2 sense. Many of the properties of the Fourier transform in L1 carry over to L2, by a suitable limiting argument.

Furthermore \mathcal{F}: L2(Rn) → L2(Rn) is a unitary operator (Stein & Weiss 1971, Thm. 2.3). For an operator to be unitary it is sufficient to show that it is bijective and preserves the inner product, so in this case these follow from the Fourier inversion theorem combined with the fact that for any f,gL2(Rn) we have

\int_{\mathbf{R}^n} f(x)\mathcal{F}g(x)\,dx = \int_{\mathbf{R}^n} \mathcal{F}f(x)g(x)\,dx.

In particular, the image of L2(Rn) is itself under the Fourier transform.

On other Lp

The definition of the Fourier transform can be extended to functions in Lp(Rn) for 1 ≤ p ≤ 2 by decomposing such functions into a fat tail part in L2 plus a fat body part in L1. In each of these spaces, the Fourier transform of a function in Lp(Rn) is in Lq(Rn), where q=p/(p-1) is the Hölder conjugate of p. by the Hausdorff–Young inequality. However, except for p = 2, the image is not easily characterized. Further extensions become more technical. The Fourier transform of functions in Lp for the range 2 < p < ∞ requires the study of distributions (Katznelson 1976). In fact, it can be shown that there are functions in Lp with p > 2 so that the Fourier transform is not defined as a function (Stein & Weiss 1971).

Tempered distributions[edit]

One might consider enlarging the domain of the Fourier transform from L1+L2 by considering generalized functions, or distributions. A distribution on Rn is a continuous linear functional on the space Cc(Rn) of compactly supported smooth functions, equipped with a suitable topology. The strategy is then to consider the action of the Fourier transform on Cc(Rn) and pass to distributions by duality. The obstruction to do this is that the Fourier transform does not map Cc(Rn) to Cc(Rn). In fact the Fourier transform of an element in Cc(Rn) can not vanish on an open set; see the above discussion on the uncertainty principle. The right space here is the slightly larger space of Schwartz functions. The Fourier transform is an automorphism on the Schwartz space, as a topological vector space, and thus induces an automorphism on its dual, the space of tempered distributions (Stein & Weiss 1971). The tempered distributions include all the integrable functions mentioned above, as well as well-behaved functions of polynomial growth and distributions of compact support.

For the definition of the Fourier transform of a tempered distribution, let f and g be integrable functions, and let \hat{f} and \hat{g} be their Fourier transforms respectively. Then the Fourier transform obeys the following multiplication formula (Stein & Weiss 1971),

\int_{\mathbf{R}^n}\hat{f}(x)g(x)\,dx=\int_{\mathbf{R}^n}f(x)\hat{g}(x)\,dx.

Every integrable function f defines (induces) a distribution Tf by the relation

T_f(\varphi)=\int_{\mathbf{R}^n}f(x)\varphi(x)\,dx   for all Schwartz functions φ.

So it makes sense to define Fourier transform \hat{T}_f of Tf by

\hat{T}_f (\varphi)= T_f(\hat{\varphi})

for all Schwartz functions φ. Extending this to all tempered distributions T gives the general definition of the Fourier transform.

Distributions can be differentiated and the above-mentioned compatibility of the Fourier transform with differentiation and convolution remains true for tempered distributions.

Generalizations[edit]

Fourier–Stieltjes transform[edit]

The Fourier transform of a finite Borel measure μ on Rn is given by (Pinsky 2002, p. 256):

\hat\mu(\xi)=\int_{\mathbf{R}^n} \mathrm{e}^{-2\pi i x \cdot \xi}\,d\mu.

This transform continues to enjoy many of the properties of the Fourier transform of integrable functions. One notable difference is that the Riemann–Lebesgue lemma fails for measures (Katznelson 1976). In the case that dμ = f(x)dx, then the formula above reduces to the usual definition for the Fourier transform of f. In the case that μ is the probability distribution associated to a random variable X, the Fourier-Stieltjes transform is closely related to the characteristic function, but the typical conventions in probability theory take eix·ξ instead of e−2πix·ξ (Pinsky 2002). In the case when the distribution has a probability density function this definition reduces to the Fourier transform applied to the probability density function, again with a different choice of constants.

The Fourier transform may be used to give a characterization of measures. Bochner's theorem characterizes which functions may arise as the Fourier–Stieltjes transform of a positive measure on the circle (Katznelson 1976).

Furthermore, the Dirac delta function is not a function but it is a finite Borel measure. Its Fourier transform is a constant function (whose specific value depends upon the form of the Fourier transform used).

Locally compact abelian groups[edit]

Main article: Pontryagin duality

The Fourier transform may be generalized to any locally compact abelian group. A locally compact abelian group is an abelian group that is at the same time a locally compact Hausdorff topological space so that the group operation is continuous. If G is a locally compact abelian group, it has a translation invariant measure μ, called Haar measure. For a locally compact abelian group G, the set of irreducible, i.e. one-dimensional, unitary representations are called its characters. With its natural group structure and the topology of pointwise convergence, the set of characters \hat G is itself a locally compact abelian group, called the Pontryagin dual of G. For a function f in L1(G), its Fourier transform is defined by (Katznelson 1976):

\hat{f}(\xi)=\int_G \xi(x)f(x)\,d\mu\qquad\text{for any }\xi\in\hat G.

The Riemann-Lebesgue lemma holds in this case; \hat{f}(\xi) is a function vanishing at infinity on \hat G.

Gelfand transform[edit]

The Fourier transform is also a special case of Gelfand transform. In this particular context, it is closely related to the Pontryagin duality map defined above.

Given an abelian locally compact Hausdorff topological group G, as before we consider space L1(G), defined using a Haar measure. With convolution as multiplication, L1(G) is an abelian Banach algebra. It also has an involution * given by

f^*(g) = \overline{f(g^{-1})}.

Taking the completion with respect to the largest possibly C*-norm gives its enveloping C*-algebra, called the group C*-algebra C*(G) of G. (Any C*-norm on L1(G) is bounded by the L1 norm, therefore their supremum exists.)

Given any abelian C*-algebra A, the Gelfand transform gives an isomorphism between A and C0(A^), where A^ is the multiplicative linear functionals, i.e. one-dimensional representations, on A with the weak-* topology. The map is simply given by

a \mapsto ( \varphi \mapsto \varphi(a)  )

It turns out that the multiplicative linear functionals of C*(G), after suitable identification, are exactly the characters of G, and the Gelfand transform, when restricted to the dense subset L1(G) is the Fourier-Pontryagin transform.

Compact non-abelian groups[edit]

The Fourier transform can also be defined for functions on a non-abelian group, provided that the group is compact. Removing the assumption that the underlying group is abelian, irreducible unitary representations need not always be one-dimensional. This means the Fourier transform on a non-abelian group takes values as Hilbert space operators (Hewitt & Ross 1970, Chapter 8). The Fourier transform on compact groups is a major tool in representation theory (Knapp 2001) and non-commutative harmonic analysis.

Let G be a compact Hausdorff topological group. Let Σ denote the collection of all isomorphism classes of finite-dimensional irreducible unitary representations, along with a definite choice of representation U(σ) on the Hilbert space Hσ of finite dimension dσ for each σ ∈ Σ. If μ is a finite Borel measure on G, then the Fourier–Stieltjes transform of μ is the operator on Hσ defined by

\langle \hat{\mu}\xi,\eta\rangle_{H_\sigma} = \int_G \langle \overline{U}^{(\sigma)}_g\xi,\eta\rangle\,d\mu(g)

where \overline{U}^{(\sigma)} is the complex-conjugate representation of U(σ) acting on Hσ. If μ is absolutely continuous with respect to the left-invariant probability measure λ on G, represented as

d\mu = fd\lambda

for some fL1(λ), one identifies the Fourier transform of f with the Fourier–Stieltjes transform of μ.

The mapping \mu\mapsto\hat{\mu} defines an isomorphism between the Banach space M(G) of finite Borel measures (see rca space) and a closed subspace of the Banach space C(Σ) consisting of all sequences E = (Eσ) indexed by Σ of (bounded) linear operators Eσ: HσHσ for which the norm

\|E\| = \sup_{\sigma\in\Sigma}\|E_\sigma\|

is finite. The "convolution theorem" asserts that, furthermore, this isomorphism of Banach spaces is in fact an isometric isomorphism of C* algebras into a subspace of C(Σ). Multiplication on M(G) is given by convolution of measures and the involution * defined by

f^*(g) = \overline{f(g^{-1})},

and C(Σ) has a natural C*-algebra structure as Hilbert space operators.

The Peter–Weyl theorem holds, and a version of the Fourier inversion formula (Plancherel's theorem) follows: A class function on a group G is any function invariant under group conjugation. For a class function fL2(G), then

f(g) = \sum_{\sigma\in\Sigma} d_\sigma \operatorname{tr}(\hat{f}(\sigma)U^{(\sigma)}_g)

where the summation is understood as convergent in the L2 sense.

The generalization of the Fourier transform to the noncommutative situation has also in part contributed to the development of noncommutative geometry.[citation needed] In this context, a categorical generalization of the Fourier transform to noncommutative groups is Tannaka–Krein duality, which replaces the group of characters with the category of representations. However, this loses the connection with harmonic functions.

Alternatives[edit]

In signal processing terms, a function (of time) is a representation of a signal with perfect time resolution, but no frequency information, while the Fourier transform has perfect frequency resolution, but no time information: the magnitude of the Fourier transform at a point is how much frequency content there is, but location is only given by phase (argument of the Fourier transform at a point), and standing waves are not localized in time – a sine wave continues out to infinity, without decaying. This limits the usefulness of the Fourier transform for analyzing signals that are localized in time, notably transients, or any signal of finite extent.

As alternatives to the Fourier transform, in time-frequency analysis, one uses time-frequency transforms or time-frequency distributions to represent signals in a form that has some time information and some frequency information – by the uncertainty principle, there is a trade-off between these. These can be generalizations of the Fourier transform, such as the short-time Fourier transform or fractional Fourier transform, or other functions to represent signals, as in wavelet transforms and chirplet transforms, with the wavelet analog of the (continuous) Fourier transform being the continuous wavelet transform. (Boashash 2003).

Applications[edit]

Some problems, such as certain differential equations, become easier to solve when the Fourier transform is applied. In that case the solution to the original problem is recovered using the inverse Fourier transform.

Analysis of differential equations[edit]

Perhaps the most important use of the Fourier transformation is to solve partial differential equations. Many of the equations of the mathematical physics of the nineteenth century can be treated this way. Fourier studied the heat equation, which in one dimension and in dimensionless units is

{\partial^2 y(x,t)\over \partial^2 x } ={\partial y(x,t) \over \partial t}.

The example we will give, a slightly more difficult one, is the wave equation in one dimension,

{\partial^2y(x,t)\over\partial^2 x }={\partial^2y(x,t)\over\partial^2t}.

As usual, the problem is not to find a solution: there are infinitely many. The problem is that of the so-called "boundary problem": find a solution which satisfies the "boundary conditions"

y(x,0) = f(x) , {\partial y(x,t) \over \partial t}= g(x).

Here, f and g are given functions. For the heat equation, only one boundary condition can be required (usually the first one). But for the wave equation, there are still infinitely many solutions y which satisfy the first boundary condition. But when one imposes both conditions, there is only one possible solution.

It is easier to find the Fourier transform \hat y of the solution than to find the solution directly. This is because the Fourier transformation takes differentiation into multiplication by the variable, and so a partial differential equation applied to the original function is transformed into multiplication by polynomial functions of the dual variables applied to the transformed function. After \hat y is determined, we can apply the inverse Fourier transformation to find y.

Fourier's method is as follows. First, note that any function of the forms

 \cos 2\pi\xi(x\pm t)  \mbox{or} \sin 2\pi\xi(x \pm t)

satisfies the wave equation. These are called "the elementary solutions."

Second, note that therefore any integral

 y(x,t) = \int_{0}^{\infty} a_+(\xi)\cos 2\pi\xi(x +t) +  a_-(\xi)\cos 2\pi\xi(x -t) + b_+(xi)\sin 2\pi\xi(x +t) + b_-(\xi)\sin 2\pi\xi(x -t) d\xi

(for arbitrary a_+, a_-, b_+, and b_-) satisfies the wave equation. (This integral is just a kind of continuous linear combination, and the equation is linear.)

Now this resembles the formula for the Fourier synthesis of a function. In fact, this is the real inverse Fourier transform of a_\pm and b_\pm in the variable x.

The third step is to examine how to find the specific unknown coefficient functions a_\pm and b_\pm that will lead to y's satisfying the boundary conditions. We are interested in the values of these solutions at t=0. So we will set t=0. Assuming that the conditions needed for Fourier inversion are satisfied, we can then find the Fourier sine and cosine transforms (in the variable x) of both sides and obtain

 2\int_{-\infty}^{\infty} y(x,0) \cos 2\pi\xi x dx = a_++a_-

and

2\int_{-\infty}^{\infty} y(x,0) \sin 2\pi\xi x dx = b_++b_-.

Similarly, taking the derivative of y with respect to t and then applying the Fourier sine and cosine transformations yields

 2\int_{-\infty}^{\infty} {\partial y(u,0) \over \partial t} \sin (2\pi\xi x) dx = (2\pi\xi)(-a_++a_-)

and

2\int_{-\infty}^{\infty} {\partial y(u,0) \over \partial t} \cos (2\pi\xi x) dx = (2\pi\xi)(b_+-b_-).

These are four linear equations for the four unknowns a_\pm and b_\pm, in terms of the Fourier sine and cosine transforms of the boundary conditions, which are easily solved by elementary algebra, provided that these transforms can be found.

In summary, we chose a set of elementary solutions, parametrised by \xi, of which the general solution would be a (continuous) linear combination in the form of an integral over the parameter \xi. But this integral was in the form of a Fourier integral. The next step was to express the boundary conditions in terms of these integrals, and set them equal to the given functions f and g. But these expressions also took the form of a Fourier integral because of the properties of the Fourier transform of a derivative. The last step was to exploit Fourier inversion by applying the Fourier transformation to both sides, thus obtaining expressions for the coefficient functions a_\pm and b_\pm in terms of the given boundary conditions f and g.

George Mackey has called this procedure "Fourier's algorithm." "Harmonic Analysis as the Exploitation of Symmetry---A Historical Survey." Rice University Studies Volume 64 (1978), Numbers 2 and 3, pp. 73–228. Reprinted in Bulletin of the American Mathematical Society (New Series) Volume 3 (1980), Number 1 (July—November), pp. 543–698.

In fact, one can perform further manipulations on the answer obtained, but those belong rather to the subject of partial differential equations than to the subject of the Fourier transformation itself.

From a higher point of view, Fourier's procedure can be reformulated more conceptually. Since there are two variables, we will use the Fourier transformation in both x and t rather than operate as Fourier did, who only transformed in the spatial variables. Note that \hat y must be considered in the sense of a distribution since y(x,t) is not going to be L^1: as a wave, it will persist through time and thus is not a transient phenomenon. But it will be bounded and so its Fourier transform can be defined as a distribution. The operational properties of the Fourier transformation that are relevant to this equation are that it takes differentiation in x to multiplication by 2\pi i\xi and differentiation with respect to t to multiplication by 2\pi if where f is the frequency. Then the wave equation becomes an algebraic equation in \hat y:

\xi^2 \hat y (\xi, f) = f^2 \hat y (\xi, f).

This is equivalent to requiring \hat y(\xi,f) = 0 unless \xi = \pm f. Right away, this explains why the choice of elementary solutions we made earlier worked so well: obviously \hat f = \delta(\xi\pm f) will be solutions. Applying Fourier inversion to these delta functions, we obtain the elementary solutions we picked earlier. But from the higher point of view, one does not pick elementary solutions, but rather considers the space of all distributions which are supported on the (degenerate) conic \xi^2-f^2 = 0.

We may as well consider the distributions supported on the conic that are given by distributions of one variable on the line \xi = f plus distributions on the line \xi = -f as follows: if \phi is any test function,

\int\int \hat y \phi(\xi,f) d\xi df
= \int s_+ \phi(\xi,\xi) d\xi + \int s_- \phi(\xi,-\xi) d\xi,

where s_+, and  s_-, are distributions of one variable.

Then Fourier inversion gives, for the boundary conditions, something very similar to what we had more concretely above (put \phi(\xi,f)= e^{2\pi i (x\xi+tf)}, which is clearly of polynomial growth):

 y(x,0) = \int\{s_+(\xi) + s_-(\xi)\} e^{2\pi i\xi x+0} d\xi

and

 {\partial y(x,0) \over \partial t} =
\int\{s_+(\xi) - s_-(\xi)\} 2\pi i \xi e^{2\pi i\xi x+0} d\xi.

Now, as before, applying the one-variable Fourier transformation in the variable x to these functions of x yields two equations in the two unknown distributions s_\pm (which can be taken to be ordinary functions if the boundary conditions are L^1 or L^2).

From a calculational point of view, the drawback of course is that one must first calculate the Fourier transforms of the boundary conditions, then assemble the solution from these, and then calculate an inverse Fourier transform. Closed form formulas are rare, except when there is some geometric symmetry that can be exploited, and the numerical calculations are difficult because of the oscillatory nature of the integrals, which makes convergence slow and hard to estimate. For practical calculations, other methods are often used.

The twentieth century has seen the extension of these methods to all linear partial differential equations with polynomial coefficients, and by extending the notion of Fourier transformation to include Fourier integral operators, some non-linear equations as well.

Fourier transform spectroscopy[edit]

The Fourier transform is also used in nuclear magnetic resonance (NMR) and in other kinds of spectroscopy, e.g. infrared (FTIR). In NMR an exponentially shaped free induction decay (FID) signal is acquired in the time domain and Fourier-transformed to a Lorentzian line-shape in the frequency domain. The Fourier transform is also used in magnetic resonance imaging (MRI) and mass spectrometry.

Quantum mechanics[edit]

The Fourier transform is useful in Quantum Mechanics in two different ways. To begin with, the basic conceptual structure of Quantum Mechanics postulates the existence of pairs of complementary variables, connected by the Heisenberg uncertainty principle. For example, in one dimension, the spatial variable q of, say, a particle, can only be measured by the quantum mechanical "position operator" at the cost of losing information about the momentum p of the particle. Therefore the physical state of the particle can either be described by a function, called "the wave function", of q or by a function of p but not by a function of both variables. The variable p is called the conjugate variable to q. In Classical Mechanics, the physical state of a particle (existing in one dimension, for simplicityof exposition) would be given by assigning definite values to both p and q simultaneously. Thus, the set of all possible physical states is the two-dimensional real vector space with a p-axis and a q-axis.

In contrast, Quantum Mechanics chooses a polarisation of this space in the sense that it picks a subspace of one-half the dimension, for example, the q-axis alone, but instead of considering only points, takes the set of all complex-valued "wave function" on this axis. Nevertheless, choosing the p-axis is an equally valid polarisation, yielding a different representation of the set of possible physical states of the particle which is related to the first representation by the Fourier transformation


\phi_(p) = \int \psi (q) e^{2\pi i pq\over h} dq.

Physically realisable states are L^2, and so by the Plancherel theorem, their Fourier transforms are also L^2. (Note that since q is in units of distance and p is in units of momentum, the presence of Planck's constant in the exponent makes the exponent dimensionless, as it should be.)

Therefore, the Fourier transform can be used to pass from one way of representing the state of the particle, by a wave function of position, to another way of representing the state of the particle: by a wave function of momentum. Infinitely many different polarisations are possible, and all are equally valid. Being able to transform states from one representation to another is sometimes convenient.

The other use of the Fourier transform in both Quantum Mechanics and Quantum Field Theory is to solve the applicable wave equation. In non-relativistic Quantum Mechanics, Schroedinger's equation for a time-varying wave function in one-dimenion, not subject to external forces, is


{\partial^2\over\partial x^2} \psi(x,t)  = i \frac h{2\pi} {\partial\over\partial t} \psi(x,t).

This is the same as the heat equation except for the presence of the imaginary unit i. Fourier methods can be used to solve this equation.

In the presence of a potential, given by the potential energy function V(x), the equation becomes


{\partial^2\over\partial x^2} \psi(x,t) + V(x)\psi(x,t)  = i \frac h{2\pi} {\partial\over\partial t} \psi(x,t).

The "elementary solutions", as we referred to them above, are the so-called "stationary states" of the particle, and Fourier's algorithm, as described above, can still be used to solve the boundary value problem of the future evolution of \psi given its values for t=0. Neither of these approaches is of much practical use in Quantum Mechanics. Boundary value problems and the time-evolution of the wave function is not of much practical interest: it is the stationary states that are most important.

In relativistic Quantum Mechanics, Schroedinger's equation becomes a wave equation as was usual in classical physics, except that complex-valued waves are considered. A simple example, in the absence of interactions with other particles or fields, is the free one-dimensional Klein--Gordon--Schroedinger--Fock equation, this time in dimensionless units,


{\partial^2\over\partial x^2} \psi(x,t)  =  {\partial^2\over\partial t^2} \psi(x,t).

This is, from the mathematical point of view, the same as the wave equation of classical physics solved above (but with a complex-valued wave, which makes no difference in the methods). This is of great use in Quantum Field Theory: each separate Fourier component of a wave can be treated as a separate harmonic oscillator and then quantised, a procedure known as "second quantisation". Fourier methods have been adapted to also deal with non-trivial interactions.

Signal processing[edit]

The Fourier transform is used for the spectral analysis of time-series. The subject of statistical signal processing does not, however, usually apply the Fourier transformation to the signal itself. Even if a real signal is indeed transient in that it began after 4004 B.C. and will end before the next millenium, it has been found in practice advisable to model a signal by a function (or, alternatively, a stochastic process) which is stationary in the sense that its characteristic properties are constant over all time. The Fourier transform of such a function does not exist in the usual sense, and it has been found more useful for the analysis of signals to instead take the Fourier transform of its auto-correlation function.

The auto-correlation function R of a function f is defined by

\R_f (\tau) = \lim_{T\rightarrow \infty} {1\over 2T} \int_{-T}^T f(t) f(t+\tau) dt.


This function is a function of the time-lag \tau elapsing between the values of f to be correlated.

For most functions f that occur in practice, R is a bounded even function of the time-lag  \tau and for typical noisy signals it turns out to be uniformly continuous with a maximum at  \tau = zero.

The auto-correlation function, more properly called the auto-covariance function unless it is normalised in some appropriate fashion, measures the strength of the correlation between the values of f separated by a time-lag. This is a way of searching for the correlation of f with its own past. It is useful even for other statistical tasks besides the analysis of signals. For example, if f(t) represents the temperature at time t, one expects a strong correlation with the temperature at a time-lag of 24 hours.

It possesses a Fourier transform,

 P_f(\xi) =  \int_{-\infty}^\infty  R_f (\tau) e^{-2\pi i \xi\tau} d\tau.

This Fourier transform is called the power spectral density function of f. (Unless all periodic components are first filtered out from f, this integral will diverge, but it is easy to filter out such periodicities.)

The power spectrum, as indicated by this density function P, measures the amount of variance contributed to the data by the frequency  \xi. In electrical signals, the variance is proportional to the average power (energy per unit time), and so the power spectrum describes how much the different frequencies contribute to the average power of the signal. This process is called the spectral analysis of time-series and is analogous to the usual analysis of variance of data that is not a time-series (ANOVA).

Knowledge of which frequencies are "important" in this sense is crucial for the proper design of filters and for the proper evaluation of measuring apparatuses. It can also be useful for the scientific analysis of the phenomena responsible for producing the data.

The power spectrum of a signal can also be approximately measured directly by measuring the average power that remains in a signal after all the frequencies outside a narrow band have been filtered out.


Spectral analysis is carried out for visual signals as well. The power spectrum ignores all phase relations, which is good enough for many purposes, but for video signals other types of spectral analysis must also be employed, still using the Fourier transform as a tool.

Other notations[edit]

Other common notations for \hat f(\xi) include:

\tilde{f}(\xi),\  \tilde{f}(\omega),\  F(\xi),\  \mathcal{F}\left(f\right)(\xi),\  \left(\mathcal{F}f\right)(\xi),\  \mathcal{F}(f),\  \mathcal F(\omega),\ F(\omega),\  \mathcal F(j\omega),\  \mathcal{F}\{f\},\  \mathcal{F} \left(f(t)\right),\ \mathcal{F} \{f(t)\}.

Denoting the Fourier transform by a capital letter corresponding to the letter of function being transformed (such as f(x) and F(ξ)) is especially common in the sciences and engineering. In electronics, the omega (ω) is often used instead of ξ due to its interpretation as angular frequency, sometimes it is written as F(jω), where j is the imaginary unit, to indicate its relationship with the Laplace transform, and sometimes it is written informally as F(2πf) in order to use ordinary frequency.

The interpretation of the complex function \hat f(\xi) may be aided by expressing it in polar coordinate form

\hat f(\xi) = A(\xi) e^{i\varphi(\xi)}

in terms of the two real functions A(ξ) and φ(ξ) where:

A(\xi) = |\hat f(\xi)|,

is the amplitude and

\varphi (\xi) = \arg \big( \hat f(\xi) \big),

is the phase (see arg function).

Then the inverse transform can be written:

f(x) = \int  _{-\infty}^{\infty} A(\xi)\ e^{ i(2\pi \xi x +\varphi (\xi))}\,d\xi,

which is a recombination of all the frequency components of f(x). Each component is a complex sinusoid of the form e2πixξ  whose amplitude is A(ξ) and whose initial phase angle (at x = 0) is φ(ξ).

The Fourier transform may be thought of as a mapping on function spaces. This mapping is here denoted \mathcal F and \mathcal F(f) is used to denote the Fourier transform of the function f. This mapping is linear, which means that \mathcal F can also be seen as a linear transformation on the function space and implies that the standard notation in linear algebra of applying a linear transformation to a vector (here the function f) can be used to write \mathcal F f instead of \mathcal F(f). Since the result of applying the Fourier transform is again a function, we can be interested in the value of this function evaluated at the value ξ for its variable, and this is denoted either as \mathcal{F} f(\xi) or as (\mathcal F f)(\xi). Notice that in the former case, it is implicitly understood that \mathcal F is applied first to f and then the resulting function is evaluated at ξ, not the other way around.

In mathematics and various applied sciences, it is often necessary to distinguish between a function f and the value of f when its variable equals x, denoted f(x). This means that a notation like \mathcal{F}(f(x)) formally can be interpreted as the Fourier transform of the values of f at x. Despite this flaw, the previous notation appears frequently, often when a particular function or a function of a particular variable is to be transformed.

For example, \mathcal F( \mathrm{rect}(x) ) = \mathrm{sinc}(\xi) is sometimes used to express that the Fourier transform of a rectangular function is a sinc function,

or \mathcal F(f(x + x_0)) = \mathcal F(f(x)) e^{2\pi i \xi x_0} is used to express the shift property of the Fourier transform.

Notice, that the last example is only correct under the assumption that the transformed function is a function of x, not of x0.

Other conventions[edit]

The Fourier transform can also be written in terms of angular frequency: ω = 2πξ whose units are radians per second.

The substitution ξ = ω/(2π) into the formulas above produces this convention:

\hat{f}(\omega) = \int_{\mathbf R^n} f(x) e^{-i\omega\cdot x}\,dx.

Under this convention, the inverse transform becomes:

f(x) = \frac{1}{(2\pi)^n} \int_{\mathbf R^n} \hat{f}(\omega)e^{i\omega \cdot x}\,d\omega.

Unlike the convention followed in this article, when the Fourier transform is defined this way, it is no longer a unitary transformation on L2(Rn). There is also less symmetry between the formulas for the Fourier transform and its inverse.

Another convention is to split the factor of (2π)n evenly between the Fourier transform and its inverse, which leads to definitions:

 \hat{f}(\omega) = \frac{1}{(2\pi)^{n/2}} \int_{\mathbf{R}^n} f(x) e^{- i\omega\cdot x}\,dx
f(x) = \frac{1}{(2\pi)^{n/2}} \int_{\mathbf{R}^n} \hat{f}(\omega) e^{ i\omega \cdot x}\,d\omega.

Under this convention, the Fourier transform is again a unitary transformation on L2(Rn). It also restores the symmetry between the Fourier transform and its inverse.

Variations of all three conventions can be created by conjugating the complex-exponential kernel of both the forward and the reverse transform. The signs must be opposites. Other than that, the choice is (again) a matter of convention.

Summary of popular forms of the Fourier transform
ordinary frequency ξ (hertz) unitary \displaystyle \hat{f}_1(\xi)\ \stackrel{\mathrm{def}}{=}\ \int_{\mathbf{R}^n} f(x) e^{-2 \pi i x\cdot\xi}\, dx = \hat{f}_2(2 \pi \xi)=(2 \pi)^{n/2}\hat{f}_3(2 \pi \xi)

\displaystyle f(x) = \int_{\mathbf{R}^n} \hat{f}_1(\xi) e^{2 \pi i  x\cdot \xi}\, d\xi \

angular frequency ω (rad/s) unitary \displaystyle \hat{f}_3(\omega) \ \stackrel{\mathrm{def}}{=}\  \frac{1}{(2 \pi)^{n/2}} \int_{\mathbf{R}^n} f(x) \ e^{-i \omega\cdot x}\, dx = \frac{1}{(2 \pi)^{n/2}} \hat{f}_1\left(\frac{\omega}{2 \pi} \right) = \frac{1}{(2 \pi)^{n/2}} \hat{f}_2(\omega)

\displaystyle f(x) = \frac{1}{(2 \pi)^{n/2}} \int_{\mathbf{R}^n} \hat{f}_3(\omega)e^{i \omega\cdot x}\, d \omega \

non-unitary \displaystyle \hat{f}_2(\omega) \ \stackrel{\mathrm{def}}{=}\int_{\mathbf{R}^n} f(x) e^{-i\omega\cdot x} \, dx \ = \hat{f}_1 \left ( \frac{\omega}{2 \pi} \right ) = (2 \pi)^{n/2}\ \hat{f}_3(\omega)

\displaystyle f(x) = \frac{1}{(2 \pi)^n} \int_{\mathbf{R}^n} \hat{f}_2(\omega) e^{i \omega\cdot x} \, d \omega \

As discussed above, the characteristic function of a random variable is the same as the Fourier–Stieltjes transform of its distribution measure, but in this context it is typical to take a different convention for the constants. Typically characteristic function is defined E(e^{it\cdot X})=\int e^{it\cdot x}d\mu_X(x).

As in the case of the "non-unitary angular frequency" convention above, there is no factor of 2π appearing in either of the integral, or in the exponential. Unlike any of the conventions appearing above, this convention takes the opposite sign in the exponential.

Tables of important Fourier transforms[edit]

The following tables record some closed-form Fourier transforms. For functions f(x), g(x) and h(x) denote their Fourier transforms by \hat{f}, \hat{g}, and \hat{h} respectively. Only the three most common conventions are included. It may be useful to notice that entry 105 gives a relationship between the Fourier transform of a function and the original function, which can be seen as relating the Fourier transform and its inverse.

Functional relationships[edit]

The Fourier transforms in this table may be found in Erdélyi (1954) or Kammler (2000, appendix).

Function Fourier transform
unitary, ordinary frequency
Fourier transform
unitary, angular frequency
Fourier transform
non-unitary, angular frequency
Remarks
\displaystyle f(x)\, \displaystyle \hat{f}(\xi)=

\displaystyle \int_{-\infty}^{\infty}f(x) e^{-2\pi i x\xi}\, dx

\displaystyle \hat{f}(\omega)=

\displaystyle \frac{1}{\sqrt{2 \pi}} \int_{-\infty}^{\infty} f(x) e^{-i \omega x}\, dx

\displaystyle \hat{f}(\nu)=

\displaystyle \int_{-\infty}^{\infty}f(x) e^{-i \nu x}\, dx

Definition
101 \displaystyle a\cdot f(x) + b\cdot g(x)\, \displaystyle a\cdot \hat{f}(\xi) + b\cdot \hat{g}(\xi)\, \displaystyle a\cdot \hat{f}(\omega) + b\cdot \hat{g}(\omega)\, \displaystyle a\cdot \hat{f}(\nu) + b\cdot \hat{g}(\nu)\, Linearity
102 \displaystyle f(x - a)\, \displaystyle e^{-2\pi i a \xi} \hat{f}(\xi)\, \displaystyle e^{- i a \omega} \hat{f}(\omega)\, \displaystyle e^{- i a \nu} \hat{f}(\nu)\, Shift in time domain
103 \displaystyle e^{ 2\pi iax} f(x)\, \displaystyle \hat{f} \left(\xi - a\right)\, \displaystyle \hat{f}(\omega - 2\pi a)\, \displaystyle \hat{f}(\nu - 2\pi a)\, Shift in frequency domain, dual of 102
104 \displaystyle f(a x)\, \displaystyle \frac{1}{|a|} \hat{f}\left( \frac{\xi}{a} \right)\, \displaystyle \frac{1}{|a|} \hat{f}\left( \frac{\omega}{a} \right)\, \displaystyle \frac{1}{|a|} \hat{f}\left( \frac{\nu}{a} \right)\, Scaling in the time domain. If \displaystyle |a|\, is large, then \displaystyle f(a x)\, is concentrated around 0 and \displaystyle \frac{1}{|a|}\hat{f} \left( \frac{\omega}{a} \right)\, spreads out and flattens.
105 \displaystyle \hat{f}(x)\, \displaystyle f(-\xi)\, \displaystyle f(-\omega)\, \displaystyle 2\pi f(-\nu)\, Duality. Here \hat{f} needs to be calculated using the same method as Fourier transform column. Results from swapping "dummy" variables of x and \xi or \omega or \nu.
106 \displaystyle \frac{d^n f(x)}{dx^n}\, \displaystyle  (2\pi i\xi)^n  \hat{f}(\xi)\, \displaystyle (i\omega)^n  \hat{f}(\omega)\, \displaystyle (i\nu)^n  \hat{f}(\nu)\,
107 \displaystyle x^n f(x)\, \displaystyle \left (\frac{i}{2\pi}\right)^n \frac{d^n \hat{f}(\xi)}{d\xi^n}\, \displaystyle i^n \frac{d^n \hat{f}(\omega)}{d\omega^n} \displaystyle i^n \frac{d^n \hat{f}(\nu)}{d\nu^n} This is the dual of 106
108 \displaystyle (f * g)(x)\, \displaystyle \hat{f}(\xi) \hat{g}(\xi)\, \displaystyle \sqrt{2\pi} \hat{f}(\omega) \hat{g}(\omega)\, \displaystyle \hat{f}(\nu) \hat{g}(\nu)\, The notation \displaystyle f * g\, denotes the convolution of f and g — this rule is the convolution theorem
109 \displaystyle f(x) g(x)\, \displaystyle (\hat{f} * \hat{g})(\xi)\, \displaystyle (\hat{f} * \hat{g})(\omega) \over \sqrt{2\pi}\, \displaystyle \frac{1}{2\pi}(\hat{f} * \hat{g})(\nu)\, This is the dual of 108
110 For \displaystyle f(x) \, purely real \displaystyle \hat{f}(-\xi) = \overline{\hat{f}(\xi)}\, \displaystyle \hat{f}(-\omega) = \overline{\hat{f}(\omega)}\, \displaystyle \hat{f}(-\nu) = \overline{\hat{f}(\nu)}\, Hermitian symmetry. \displaystyle \overline{z}\, indicates the complex conjugate.
111 For \displaystyle f(x) \, a purely real even function \displaystyle \hat{f}(\omega), \displaystyle \hat{f}(\xi) and \displaystyle \hat{f}(\nu)\, are purely real even functions.
112 For \displaystyle f(x) \, a purely real odd function \displaystyle \hat{f}(\omega), \displaystyle \hat{f}(\xi) and \displaystyle \hat{f}(\nu) are purely imaginary odd functions.
113 \displaystyle \overline{f(x)} \displaystyle \overline{\hat{f}(-\xi)} \displaystyle \overline{\hat{f}(-\omega)} \displaystyle \overline{\hat{f}(-\nu)} Complex conjugation, generalization of 110

Square-integrable functions[edit]

The Fourier transforms in this table may be found in (Campbell & Foster 1948), (Erdélyi 1954), or the appendix of (Kammler 2000).

Function Fourier transform
unitary, ordinary frequency
Fourier transform
unitary, angular frequency
Fourier transform
non-unitary, angular frequency
Remarks
\displaystyle f(x) \displaystyle \hat{f}(\xi)=

\displaystyle \int_{-\infty}^{\infty}f(x) e^{-2\pi ix\xi}\,dx

\displaystyle \hat{f}(\omega)=

\displaystyle \frac{1}{\sqrt{2 \pi}} \int_{-\infty}^{\infty} f(x) e^{-i \omega x}\, dx

\displaystyle \hat{f}(\nu)=

\displaystyle \int_{-\infty}^{\infty} f(x) e^{-i\nu x}\, dx

201 \displaystyle \operatorname{rect}(a x) \, \displaystyle \frac{1}{|a|}\cdot \operatorname{sinc}\left(\frac{\xi}{a}\right) \displaystyle \frac{1}{\sqrt{2 \pi a^2}}\cdot \operatorname{sinc}\left(\frac{\omega}{2\pi a}\right) \displaystyle \frac{1}{|a|}\cdot \operatorname{sinc}\left(\frac{\nu}{2\pi a}\right) The rectangular pulse and the normalized sinc function, here defined as sinc(x) = sin(πx)/(πx)
202 \displaystyle \operatorname{sinc}(a x)\, \displaystyle \frac{1}{|a|}\cdot \operatorname{rect}\left(\frac{\xi}{a} \right)\, \displaystyle \frac{1}{\sqrt{2\pi a^2}}\cdot \operatorname{rect}\left(\frac{\omega}{2 \pi a}\right) \displaystyle \frac{1}{|a|}\cdot \operatorname{rect}\left(\frac{\nu}{2 \pi a}\right) Dual of rule 201. The rectangular function is an ideal low-pass filter, and the sinc function is the non-causal impulse response of such a filter. The sinc function is defined here as sinc(x) = sin(πx)/(πx)
203 \displaystyle \operatorname{sinc}^2 (a x) \displaystyle \frac{1}{|a|}\cdot \operatorname{tri} \left( \frac{\xi}{a} \right) \displaystyle \frac{1}{\sqrt{2\pi a^2}}\cdot \operatorname{tri} \left( \frac{\omega}{2\pi a} \right) \displaystyle \frac{1}{|a|}\cdot \operatorname{tri} \left( \frac{\nu}{2\pi a} \right) The function tri(x) is the triangular function
204 \displaystyle \operatorname{tri} (a x) \displaystyle \frac{1}{|a|}\cdot \operatorname{sinc}^2 \left( \frac{\xi}{a} \right) \, \displaystyle \frac{1}{\sqrt{2\pi a^2}} \cdot \operatorname{sinc}^2 \left( \frac{\omega}{2\pi a} \right) \displaystyle \frac{1}{|a|} \cdot \operatorname{sinc}^2 \left( \frac{\nu}{2\pi a} \right) Dual of rule 203.
205 \displaystyle e^{- a x} u(x) \, \displaystyle \frac{1}{a + 2 \pi i \xi} \displaystyle \frac{1}{\sqrt{2 \pi} (a + i \omega)} \displaystyle \frac{1}{a + i \nu} The function u(x) is the Heaviside unit step function and a>0.
206 \displaystyle e^{-\alpha x^2}\, \displaystyle \sqrt{\frac{\pi}{\alpha}}\cdot e^{-\frac{(\pi \xi)^2}{\alpha}} \displaystyle \frac{1}{\sqrt{2 \alpha}}\cdot e^{-\frac{\omega^2}{4 \alpha}} \displaystyle \sqrt{\frac{\pi}{\alpha}}\cdot e^{-\frac{\nu^2}{4 \alpha}} This shows that, for the unitary Fourier transforms, the Gaussian function exp(−αx2) is its own Fourier transform for some choice of α. For this to be integrable we must have Re(α)>0.
207 \displaystyle \operatorname{e}^{-a|x|} \, \displaystyle \frac{2 a}{a^2 + 4 \pi^2 \xi^2} \displaystyle \sqrt{\frac{2}{\pi}} \cdot \frac{a}{a^2 + \omega^2} \displaystyle \frac{2a}{a^2 + \nu^2} For a>0. That is, the Fourier transform of a two-sided decaying exponential function is a Lorentzian function.
208 \displaystyle \operatorname{sech}(a x) \, \displaystyle \frac{\pi}{a} \operatorname{sech} \left( \frac{\pi^2}{ a} \xi \right) \displaystyle \frac{1}{a}\sqrt{\frac{\pi}{2}}\operatorname{sech}\left( \frac{\pi}{2 a} \omega \right) \displaystyle \frac{\pi}{a}\operatorname{sech}\left( \frac{\pi}{2 a} \nu \right) Hyperbolic secant is its own Fourier transform
209 \displaystyle e^{-\frac{a^2 x^2}2} H_n(a x)\, \displaystyle \frac{\sqrt{2\pi}(-i)^n}{a}

  \cdot e^{-\frac{2\pi^2\xi^2}{a^2}} H_n\left(\frac{2\pi\xi}a\right)

\displaystyle \frac{(-i)^n}{a}

  \cdot e^{-\frac{\omega^2}{2 a^2}} H_n\left(\frac \omega a\right)

\displaystyle \frac{(-i)^n \sqrt{2\pi}}{a}

  \cdot e^{-\frac{\nu^2}{2 a^2}} H_n\left(\frac \nu a \right)

H_n is the Hermite's polynomial. If a = 1 then the Gauss-Hermite functions are eigenfunctions of the Fourier transform operator. For a derivation, see Hermite polynomial. The formula reduces to 206 for n = 0.

Distributions[edit]

The Fourier transforms in this table may be found in (Erdélyi 1954) or the appendix of (Kammler 2000).

Function Fourier transform
unitary, ordinary frequency
Fourier transform
unitary, angular frequency
Fourier transform
non-unitary, angular frequency
Remarks
\displaystyle f(x) \displaystyle \hat{f}(\xi)=

\displaystyle \int_{-\infty}^{\infty}f(x) e^{-2\pi ix\xi}\,dx

\displaystyle \hat{f}(\omega)=

\displaystyle \frac{1}{\sqrt{2 \pi}} \int_{-\infty}^{\infty} f(x) e^{-i \omega x}\, dx

\displaystyle \hat{f}(\nu)=

\displaystyle \int_{-\infty}^{\infty} f(x) e^{-i\nu x}\, dx

301 \displaystyle 1 \displaystyle \delta(\xi) \displaystyle \sqrt{2\pi}\cdot \delta(\omega) \displaystyle 2\pi\delta(\nu) The distribution δ(ξ) denotes the Dirac delta function.
302 \displaystyle \delta(x)\, \displaystyle 1 \displaystyle \frac{1}{\sqrt{2\pi}}\, \displaystyle 1 Dual of rule 301.
303 \displaystyle e^{i a x} \displaystyle \delta\left(\xi - \frac{a}{2\pi}\right) \displaystyle \sqrt{2 \pi}\cdot \delta(\omega - a) \displaystyle 2 \pi\delta(\nu - a) This follows from 103 and 301.
304 \displaystyle \cos (a x) \displaystyle \frac{\displaystyle \delta\left(\xi - \frac{a}{2\pi}\right)+\delta\left(\xi+\frac{a}{2\pi}\right)}{2} \displaystyle \sqrt{2 \pi}\cdot\frac{\delta(\omega-a)+\delta(\omega+a)}{2}\, \displaystyle \pi\left(\delta(\nu-a)+\delta(\nu+a)\right) This follows from rules 101 and 303 using Euler's formula: \textstyle \cos(a x) =

(e^{i a x} + e^{-i a x})/2.

305 \displaystyle \sin( ax) \displaystyle \frac{\displaystyle\delta\left(\xi-\frac{a}{2\pi}\right)-\delta\left(\xi+\frac{a}{2\pi}\right)}{2i} \displaystyle \sqrt{2 \pi}\cdot\frac{\delta(\omega-a)-\delta(\omega+a)}{2i} \displaystyle -i\pi\left(\delta(\nu-a)-\delta(\nu+a)\right) This follows from 101 and 303 using \textstyle \sin(a x) =

(e^{i a x} - e^{-i a x})/(2i).

306 \displaystyle \cos ( a x^2 ) \displaystyle \sqrt{\frac{\pi}{a}} \cos \left( \frac{\pi^2 \xi^2}{a} - \frac{\pi}{4} \right)  \displaystyle \frac{1}{\sqrt{2 a}} \cos \left( \frac{\omega^2}{4 a} - \frac{\pi}{4} \right) \displaystyle \sqrt{\frac{\pi}{a}} \cos \left( \frac{\nu^2}{4 a} - \frac{\pi}{4} \right)
307 \displaystyle \sin ( a x^2 ) \, \displaystyle - \sqrt{\frac{\pi}{a}}  \sin \left( \frac{\pi^2 \xi^2}{a} - \frac{\pi}{4} \right)  \displaystyle \frac{-1}{\sqrt{2 a}} \sin \left( \frac{\omega^2}{4 a} - \frac{\pi}{4} \right) \displaystyle -\sqrt{\frac{\pi}{a}}\sin \left( \frac{\nu^2}{4 a} - \frac{\pi}{4} \right)
308 \displaystyle x^n\, \displaystyle \left(\frac{i}{2\pi}\right)^n \delta^{(n)} (\xi)\, \displaystyle i^n \sqrt{2\pi} \delta^{(n)} (\omega)\, \displaystyle 2\pi i^n\delta^{(n)} (\nu)\, Here, n is a natural number and \textstyle \delta^{(n)}(\xi) is the n-th distribution derivative of the Dirac delta function. This rule follows from rules 107 and 301. Combining this rule with 101, we can transform all polynomials.
309 \displaystyle \frac{1}{x} \displaystyle -i\pi\sgn(\xi) \displaystyle -i\sqrt{\frac{\pi}{2}}\sgn(\omega) \displaystyle -i\pi\sgn(\nu) Here sgn(ξ) is the sign function. Note that 1/x is not a distribution. It is necessary to use the Cauchy principal value when testing against Schwartz functions. This rule is useful in studying the Hilbert transform.
310 \displaystyle \frac{1}{x^n} :=

\displaystyle\frac{(-1)^{n-1}}{(n-1)!}\frac{d^n}{dx^n}\log |x|

\displaystyle -i\pi \frac{(-2\pi i\xi)^{n-1}}{(n-1)!} \sgn(\xi) \displaystyle -i\sqrt{\frac{\pi}{2}}\cdot \frac{(-i\omega)^{n-1}}{(n-1)!}\sgn(\omega) \displaystyle -i\pi \frac{(-i\nu)^{n-1}}{(n-1)!}\sgn(\nu) 1/xn is the homogeneous distribution defined by the distributional derivative \textstyle\frac{(-1)^{n-1}}{(n-1)!}\frac{d^n}{dx^n}\log|x|
311 \displaystyle |x|^\alpha\, \displaystyle -2 \frac{\sin(\pi\alpha/2)\Gamma(\alpha+1)}{|2\pi\xi|^{\alpha+1}} \displaystyle \frac{-2}{\sqrt{2\pi}}\frac{\sin(\pi\alpha/2)\Gamma(\alpha+1)}{|\omega|^{\alpha+1}} \displaystyle -2\frac{\sin(\pi\alpha/2)\Gamma(\alpha+1)}{|\nu|^{\alpha+1}} This formula is valid for 0 > α > −1. For α > 0 some singular terms arise at the origin that can be found by differentiating 318. If Re α > −1, then |x|^\alpha is a locally integrable function, and so a tempered distribution. The function \textstyle \alpha\mapsto |x|^\alpha is a holomorphic function from the right half-plane to the space of tempered distributions. It admits a unique meromorphic extension to a tempered distribution, also denoted |x|^\alpha for α ≠ −2, −4, ... (See homogeneous distribution.)
 \frac{1}{\sqrt{|x|}} \,  \frac{1}{\sqrt{|\xi|}}  \frac{1}{\sqrt{|\omega|}}  \frac{\sqrt{2\pi}}{\sqrt{|\nu|}} Special case of 311.
312 \displaystyle \sgn(x) \displaystyle \frac{1}{i\pi \xi} \displaystyle \sqrt{\frac{2}{\pi}} \frac{1}{i\omega } \displaystyle \frac{2}{i\nu } The dual of rule 309. This time the Fourier transforms need to be considered as Cauchy principal value.
313 \displaystyle u(x) \displaystyle \frac{1}{2}\left(\frac{1}{i \pi \xi} + \delta(\xi)\right) \displaystyle \sqrt{\frac{\pi}{2}} \left( \frac{1}{i \pi \omega} + \delta(\omega)\right) \displaystyle \pi\left( \frac{1}{i \pi \nu} + \delta(\nu)\right) The function u(x) is the Heaviside unit step function; this follows from rules 101, 301, and 312.
314 \displaystyle \sum_{n=-\infty}^{\infty} \delta (x - n T) \displaystyle \frac{1}{T} \sum_{k=-\infty}^{\infty} \delta \left( \xi -\frac{k }{T}\right) \displaystyle \frac{\sqrt{2\pi }}{T}\sum_{k=-\infty}^{\infty} \delta \left( \omega -\frac{2\pi k}{T}\right) \displaystyle \frac{2\pi}{T}\sum_{k=-\infty}^{\infty} \delta \left( \nu -\frac{2\pi k}{T}\right) This function is known as the Dirac comb function. This result can be derived from 302 and 102, together with the fact that \sum_{n=-\infty}^{\infty} e^{inx}=

2\pi\sum_{k=-\infty}^{\infty} \delta(x+2\pi k) as distributions.

315 \displaystyle J_0 (x) \displaystyle \frac{2\, \operatorname{rect}(\pi\xi)}{\sqrt{1 - 4 \pi^2 \xi^2}} \displaystyle \sqrt{\frac{2}{\pi}} \cdot \frac{\operatorname{rect}\left( \displaystyle \frac{\omega}{2} \right)}{\sqrt{1 - \omega^2}} \displaystyle \frac{2\,\operatorname{rect}\left(\displaystyle\frac{\nu}{2} \right)}{\sqrt{1 - \nu^2}} The function J0(x) is the zeroth order Bessel function of first kind.
316 \displaystyle J_n (x) \displaystyle \frac{2 (-i)^n T_n (2 \pi \xi) \operatorname{rect}(\pi \xi)}{\sqrt{1 - 4 \pi^2 \xi^2}} \displaystyle \sqrt{\frac{2}{\pi}} \frac{ (-i)^n T_n (\omega) \operatorname{rect} \left( \displaystyle\frac{\omega}{2} \right)}{\sqrt{1 - \omega^2}} \displaystyle \frac{2(-i)^n T_n (\nu) \operatorname{rect} \left(\displaystyle \frac{\nu}{2} \right)}{\sqrt{1 - \nu^2}} This is a generalization of 315. The function Jn(x) is the n-th order Bessel function of first kind. The function Tn(x) is the Chebyshev polynomial of the first kind.
317 \displaystyle \log \left| x \right| \displaystyle -\frac{1}{2} \frac{1}{\left| \xi \right|} - \gamma \delta \left( \xi \right) \displaystyle -\frac{\sqrt{\pi / 2}}{\left| \omega \right|} - \sqrt{2 \pi} \gamma \delta \left( \omega \right) \displaystyle -\frac{\pi}{\left| \nu \right|} - 2 \pi \gamma \delta \left( \nu \right) \gamma is the Euler–Mascheroni constant.
318 \displaystyle \left( \mp ix \right)^{-\alpha} \displaystyle \frac{\left(2\pi\right)^\alpha}{\Gamma\left(\alpha\right)}u\left(\pm \xi \right)\left(\pm \xi \right)^{\alpha-1} \displaystyle \frac{\sqrt{2\pi}}{\Gamma\left(\alpha\right)}u\left(\pm\omega\right)\left(\pm\omega\right)^{\alpha-1} \displaystyle \frac{2\pi}{\Gamma\left(\alpha\right)}u\left(\pm\nu\right)\left(\pm\nu\right)^{\alpha-1} This formula is valid for 1 > α > 0. Use differentiation to derive formula for higher exponents. u is the Heaviside function.

Two-dimensional functions[edit]

Function Fourier transform
unitary, ordinary frequency
Fourier transform
unitary, angular frequency
Fourier transform
non-unitary, angular frequency
400 \displaystyle f(x,y) \displaystyle \hat{f}(\xi_x, \xi_y)=

\displaystyle \iint f(x,y) e^{-2\pi i(\xi_x x+\xi_y y)}\,dx\,dy

\displaystyle \hat{f}(\omega_x,\omega_y)=

\displaystyle \frac{1}{2 \pi} \iint f(x,y) e^{-i (\omega_x x +\omega_y y)}\, dx\,dy

\displaystyle \hat{f}(\nu_x,\nu_y)=

\displaystyle \iint f(x,y) e^{-i(\nu_x x+\nu_y y)}\, dx\,dy

401 \displaystyle e^{-\pi\left(a^2x^2+b^2y^2\right)} \displaystyle \frac{1}{|ab|} e^{-\pi\left(\xi_x^2/a^2 + \xi_y^2/b^2\right)} \displaystyle \frac{1}{2\pi\cdot|ab|} e^{\frac{-\left(\omega_x^2/a^2 + \omega_y^2/b^2\right)}{4\pi}} \displaystyle \frac{1}{|ab|} e^{\frac{-\left(\nu_x^2/a^2 + \nu_y^2/b^2\right)}{4\pi}}
402 \displaystyle \mathrm{circ}(\sqrt{x^2+y^2}) \displaystyle \frac{J_1\left(2 \pi \sqrt{\xi_x^2+\xi_y^2}\right)}{\sqrt{\xi_x^2+\xi_y^2}} \displaystyle \frac{J_1\left(\sqrt{\omega_x^2+\omega_y^2}\right)}{\sqrt{\omega_x^2+\omega_y^2}} \displaystyle \frac{2\pi J_1\left(\sqrt{\nu_x^2+\nu_y^2}\right)}{\sqrt{\nu_x^2+\nu_y^2}}
Remarks

To 400: The variables ξx, ξy, ωx, ωy, νx and νy are real numbers. The integrals are taken over the entire plane.

To 401: Both functions are Gaussians, which may not have unit volume.

To 402: The function is defined by circ(r)=1 0≤r≤1, and is 0 otherwise. This is the Airy distribution, and is expressed using J1 (the order 1 Bessel function of the first kind). (Stein & Weiss 1971, Thm. IV.3.3)

Formulas for general n-dimensional functions[edit]

Function Fourier transform
unitary, ordinary frequency
Fourier transform
unitary, angular frequency
Fourier transform
non-unitary, angular frequency
500 \displaystyle f(\mathbf x)\, \displaystyle \hat{f}(\boldsymbol \xi)=

\displaystyle \int_{\mathbf{R}^n}f(\mathbf x) e^{-2\pi i \mathbf x \cdot \boldsymbol \xi }\, d^n \mathbf x

\displaystyle \hat{f}(\boldsymbol \omega)=

\displaystyle \frac{1}{{(2 \pi)}^{n/2}} \int_{\mathbf{R}^n} f(\mathbf x) e^{-i \boldsymbol \omega \cdot \mathbf x}\, d^n \mathbf x

\displaystyle \hat{f}(\boldsymbol \nu)=

\displaystyle \int_{\mathbf{R}^n}f(\mathbf x) e^{-i \mathbf x \cdot \boldsymbol \nu }\, d^n \mathbf x

501 \displaystyle \chi_{[0,1]}(|\mathbf x|)(1-|\mathbf x|^2)^\delta \displaystyle \pi^{-\delta}\Gamma(\delta+1)|\boldsymbol \xi|^{-n/2-\delta}
\displaystyle \times J_{n/2+\delta}(2\pi|\boldsymbol \xi|)
\displaystyle 2^{-\delta}\Gamma(\delta+1)\left|\boldsymbol \omega\right|^{-n/2-\delta}
\displaystyle \times J_{n/2+\delta}(|\boldsymbol \omega|)
\displaystyle \pi^{-\delta}\Gamma(\delta+1)\left|\frac{\boldsymbol \nu}{2\pi}\right|^{-n/2-\delta}
\displaystyle \times J_{n/2+\delta}(|\boldsymbol \nu|)
502 \displaystyle |\mathbf x|^{-\alpha}, \quad 0 < \operatorname{Re} \alpha < n. \displaystyle c_\alpha |\boldsymbol \xi|^{-(n - \alpha)}
503 \displaystyle \frac{1}{\left\|\boldsymbol \sigma\right\|\left(2\pi\right)^{n/2}} e^{-\frac{1}{2} \mathbf x^{\mathrm T} \boldsymbol \sigma^{-\mathrm T} \boldsymbol \sigma^{-1} \mathbf x} \displaystyle e^{-\frac{1}{2} \boldsymbol \nu^{\mathrm T} \boldsymbol \sigma \boldsymbol \sigma^{\mathrm T} \boldsymbol \nu}
504 \displaystyle e^{-2\pi\alpha|\mathbf x|} c_n\frac{\alpha}{(\alpha^2+|\xi|^2)^{(n+1)/2}}
Remarks

To 501: The function χ[0, 1] is the indicator function of the interval [0, 1]. The function Γ(x) is the gamma function. The function Jn/2 + δ is a Bessel function of the first kind, with order n/2 + δ. Taking n = 2 and δ = 0 produces 402. (Stein & Weiss 1971, Thm. 4.15)

To 502: See Riesz potential. The formula also holds for all α ≠ −n, −n − 1, … by analytic continuation, but then the function and its Fourier transforms need to be understood as suitably regularized tempered distributions. See homogeneous distribution.

To 503: This is the formula for a multivariate normal distribution normalized to 1 with a mean of 0. Bold variables are vectors or matrices. Following the notation of the aforementioned page, \boldsymbol \Sigma = \boldsymbol \sigma \boldsymbol \sigma^{\mathrm T} and \boldsymbol \Sigma^{-1} = \boldsymbol \sigma^{-\mathrm T} \boldsymbol \sigma^{-1}

To 504: Here c_n=\Gamma((n+1)/2)/\pi^{(n+1)/2}. See (Stein & Weiss 1971, p. 6).

See also[edit]

Notes[edit]

  1. ^ Up to an imaginary constant factor whose magnitude depends on what Fourier transform convention is used.
  2. ^ The Laplace transform is a generalization of the Fourier transform that offers greater flexibility for many such applications.
  3. ^ Depending on the application a Lebesgue integral, distributional, or other approach may be most appropriate.

References[edit]

External links[edit]