Ma · Mathematics

When Does the Unilateral Fourier Transform Exist?

Published By the 9to5Science editorial team

At a glance

  1. 1. Define the transform

    Integrate the function only over nonnegative time: F₊(ω) = ∫₀∞ f(t)e⁻ⁱωt dt.

  2. 2. Choose convergence type

    Decide whether you need ordinary, absolute, L², or distributional existence.

  3. 3. Check sufficient conditions

    Absolute integrability on [0,∞) guarantees existence for every real frequency.

  4. 4. Use Laplace theory

    A unilateral Fourier transform is the Laplace transform evaluated on the imaginary axis when that boundary value exists.

The unilateral Fourier transform of a function defined for t≥0t\ge 0 is commonly written

F+(ω)=∫0∞f(t)e−iωt dt,F_+(\omega)=\int_0^\infty f(t)e^{-i\omega t}\,dt,

where ω∈R\omega\in\mathbb{R}. The word “unilateral” means that the integration starts at zero rather than extending over the whole real line.

There is no single answer to whether it “exists” until the type of convergence is specified. In elementary applications, existence usually means that this improper integral converges as an ordinary complex integral. In Fourier analysis, one also uses L2L^2 transforms and distributions, which cover a wider class of functions.

The simplest sufficient condition

If

∫0∞∣f(t)∣ dt<∞,\int_0^\infty |f(t)|\,dt<\infty,

then f∈L1(0,∞)f\in L^1(0,\infty), and the unilateral Fourier transform exists absolutely for every real ω\omega.

Indeed,

∣f(t)e−iωt∣=∣f(t)∣,\left|f(t)e^{-i\omega t}\right|=|f(t)|,

because ∣e−iωt∣=1|e^{-i\omega t}|=1. Therefore,

∫0∞∣f(t)e−iωt∣ dt=∫0∞∣f(t)∣ dt<∞.\int_0^\infty \left|f(t)e^{-i\omega t}\right|\,dt =\int_0^\infty |f(t)|\,dt<\infty.

This condition is sufficient, not necessary. Some functions that are not absolutely integrable still have conditionally convergent Fourier integrals at selected frequencies.

Relation to the Laplace transform

The unilateral Fourier transform is closely related to the one-sided Laplace transform

L{f}(s)=∫0∞f(t)e−st dt.\mathcal{L}\{f\}(s)=\int_0^\infty f(t)e^{-st}\,dt.

Formally, putting s=iωs=i\omega gives

F+(ω)=L{f}(iω),F_+(\omega)=\mathcal{L}\{f\}(i\omega),

provided the Laplace integral converges on the imaginary axis.

This last qualification matters. If ff is of exponential order, its Laplace transform may exist only when Re⁡(s)\operatorname{Re}(s) is sufficiently large. The Fourier transform requires the boundary line Re⁡(s)=0\operatorname{Re}(s)=0, so exponential-order growth alone does not guarantee a unilateral Fourier transform.

For example, if f(t)=eatf(t)=e^{at} with a>0a>0, then

∫0∞eate−iωt dt=∫0∞e(a−iω)t dt\int_0^\infty e^{at}e^{-i\omega t}\,dt =\int_0^\infty e^{(a-i\omega)t}\,dt

diverges for every real ω\omega. Its Laplace transform exists only for Re⁡(s)>a\operatorname{Re}(s)>a.

Worked examples

An absolutely convergent example

Let f(t)=e−2tf(t)=e^{-2t} for t≥0t\ge 0. Since

∫0∞∣e−2t∣ dt=12<∞,\int_0^\infty |e^{-2t}|\,dt=\frac12<\infty,

the transform exists for every real ω\omega. Direct calculation gives

F+(ω)=∫0∞e−2te−iωt dt=∫0∞e−(2+iω)t dt=12+iω.\begin{aligned} F_+(\omega) &=\int_0^\infty e^{-2t}e^{-i\omega t}\,dt\\ &=\int_0^\infty e^{-(2+i\omega)t}\,dt\\ &=\frac{1}{2+i\omega}. \end{aligned}

More generally, for a>0a>0,

∫0∞te−ate−iωt dt=1(a+iω)2.\int_0^\infty t e^{-at}e^{-i\omega t}\,dt =\frac{1}{(a+i\omega)^2}.

For instance, with a=2a=2, the transform of te−2tt e^{-2t} is 1/(2+iω)21/(2+i\omega)^2.

A conditionally convergent example

The integral

∫0∞sin⁡tte−iωt dt\int_0^\infty \frac{\sin t}{t}e^{-i\omega t}\,dt

is not absolutely convergent in general, because ∣sin⁡t∣/t|\sin t|/t is not integrable over the whole half-line. Nevertheless, oscillation can produce conditional convergence for appropriate frequencies. Such examples show why L1L^1 is a convenient sufficient condition rather than a necessary one.

A standard way to prove conditional convergence is Dirichlet’s test: if the oscillatory factor has bounded partial integrals and the remaining factor decreases suitably to zero, the improper integral may converge.

What happens for a constant function?

Take f(t)=1f(t)=1. For ω≠0\omega\ne 0, the truncated integral is

∫0Re−iωt dt=1−e−iωRiω.\int_0^R e^{-i\omega t}\,dt =\frac{1-e^{-i\omega R}}{i\omega}.

As R→∞R\to\infty, the term e−iωRe^{-i\omega R} keeps oscillating and has no limit. Thus the ordinary improper integral does not exist, even when ω≠0\omega\ne 0. At ω=0\omega=0, the integral is simply RR, which also diverges.

However, in distribution theory one assigns the causal constant a Fourier transform. With the convention e−iωte^{-i\omega t}, the transform of the Heaviside function is written formally as

F{u(t)}(ω)=πδ(ω)−iPV⁡(1ω).\mathcal{F}\{u(t)\}(\omega) =\pi\delta(\omega)-i\operatorname{PV}\left(\frac{1}{\omega}\right).

This is not an ordinary function-valued integral. It is a tempered distribution. Therefore, a statement that the unilateral transform of a constant “exists” is correct only if generalized transforms are allowed.

Other meanings of existence

L2L^2 existence

If f∈L2(0,∞)f\in L^2(0,\infty), extend it by zero to negative tt and apply the Plancherel theorem. The Fourier transform then exists as an L2L^2 equivalence class, meaning it is defined up to changes on a set of measure zero. It need not be represented by a convergent integral at every individual frequency.

Abel-regularized existence

One can introduce a damping factor and define

F+,ε(ω)=∫0∞f(t)e−εte−iωt dt,ε>0.F_{+,\varepsilon}(\omega)=\int_0^\infty f(t)e^{-\varepsilon t}e^{-i\omega t}\,dt, \qquad \varepsilon>0.

If the limit as ε↓0\varepsilon\downarrow 0 exists, it is called an Abel-regularized transform. This method is useful for functions that do not have an ordinary Fourier integral, but the result should be labeled as regularized rather than ordinary.

Common mistakes

  • Confusing exponential order with Fourier-transform existence. Growth bounded by CeatCe^{at} guarantees a Laplace transform in a right half-plane, not necessarily on the imaginary axis.
  • Ignoring the endpoint ω=0\omega=0. A transform may converge for nonzero frequencies but fail at zero, or fail everywhere for a nondecaying function.
  • Treating boundedness as integrability. A bounded function need not belong to L1(0,∞)L^1(0,\infty); the constant function is the simplest counterexample.
  • Using a generalized answer without saying so. Distributional and Abel-regularized transforms are valid tools, but they do not mean that the original improper integral converges.
  • Forgetting the zero extension. The unilateral transform is the ordinary Fourier transform of f(t)u(t)f(t)u(t), where uu is the Heaviside function, subject to the chosen convention and convergence interpretation.

References

  1. Fourier Analysis: An Introduction, by Elias M. Stein and Rami Shakarchi
  2. The Fourier Transform and Its Applications, by Ronald N. Bracewell

Spotted an error? Send a correction. Read how we write and check articles in our editorial policy.