1 Definition and basic concepts

The Fourier–Stieltjes transform extends Fourier analysis from integrable functions to measures. Instead of multiplying by an ordinary density, it integrates a complex character against a bounded measure. This allows point masses, singular measures, and mixtures of different measure types to be studied in the same framework.

1.1 Motivation from the Fourier transform

The classical Fourier transform is designed for functions with enough integrability to make oscillatory integrals meaningful. Many natural objects, however, are better modeled by measures than by functions. A point mass, for example, is not a function in the usual sense, but it still has a well-defined transform. The Fourier–Stieltjes transform was introduced to capture such objects while preserving much of the structure of Fourier analysis.

1.2 Bounded and finite measures

The transform is defined for bounded measures, and in the classical real-variable setting this usually means finite Borel measures. Finite total mass ensures that integration against bounded complex exponentials is well defined. In this setting, the transform produces a bounded function on the dual parameter domain.

1.3 Complex-valued measures

A complex measure can be decomposed into real and imaginary parts, or more precisely into its total variation and phase. The Fourier–Stieltjes transform applies equally to complex measures, because the integral of a bounded measurable function with respect to such a measure is defined through the standard theory of complex integration. This generality is useful in harmonic analysis, where signed and complex measures naturally arise.

1.4 Transform on the real line

On the real line, the Fourier–Stieltjes transform of a finite complex Borel measure \(\mu\) is commonly written as \[ \widehat{\mu}(\xi)=\int_{\mathbb{R}} e^{-i x \xi}\,d\mu(x). \] This formula resembles the classical Fourier transform, but the density \(f(x)\,dx\) is replaced by an arbitrary measure \(d\mu(x)\). The resulting function is bounded and continuous, and it encodes the oscillatory moments of the measure.

1.5 Transform on locally compact abelian groups

For a locally compact abelian group, the role of \(e^{-i x \xi}\) is played by continuous characters. The transform of a measure is then a function on the Pontryagin dual group obtained by integrating each character against the measure. This viewpoint places the Fourier–Stieltjes transform within abstract harmonic analysis and shows that it is not tied to the real line alone.

2 Fundamental properties

The transform inherits many formal properties from the linearity and boundedness of integration. These features make it a stable tool for studying measures through their frequency behavior.

2.1 Linearity

The Fourier–Stieltjes transform is linear in the measure. If \(\mu\) and \(\nu\) are measures and \(a,b\) are scalars, then the transform of \(a\mu+b\nu\) is \(a\widehat{\mu}+b\widehat{\nu}\). This property is essential for decomposing measures into simpler components.

2.2 Boundedness

If \(\mu\) is a finite measure, then \(\widehat{\mu}(\xi)\le \|\mu\|\), where \(\|\mu\|\) denotes the total variation norm. Thus the transform is uniformly bounded by the size of the measure. This estimate is one of the main reasons the transform is so well behaved on finite measures.

2.3 Continuity

The transform of a finite measure is continuous in the frequency variable. This follows from dominated convergence, since the character \(e^{-i x \xi}\) depends continuously on \(\xi\) and is uniformly bounded in absolute value by 1. Continuity is a basic regularity property that persists even when the measure itself is highly irregular.

2.4 Uniform continuity

In the standard Euclidean setting, the transform is in fact uniformly continuous. The finite mass of the measure gives control over differences \(\widehat{\mu}(\xi+h)-\widehat{\mu}(\xi)\), and the boundedness of the exponential factors prevents wild oscillation in the transform. Uniform continuity reflects the smoothing effect of integration.

2.5 Behavior under translation

Translating a measure changes its transform by a phase factor. If \(\mu_a\) is the translate of \(\mu\), then \(\widehat{\mu_a}(\xi)\) differs from \(\widehat{\mu}(\xi)\) by multiplication with the character corresponding to the shift. This is the measure-theoretic analogue of the translation rule for the ordinary Fourier transform.

2.6 Behavior under reflection and conjugation

Reflection of a measure corresponds to reversing the sign in the frequency variable, while complex conjugation of the measure leads to conjugation of the transform. For real measures, symmetry properties of the measure are reflected in symmetry properties of \(\widehat{\mu}\). These identities are frequently used to infer qualitative information about the original measure.

3 Examples

Concrete examples show how the transform distinguishes between atomic, absolutely continuous, and singular components. They also illustrate that the same transform framework can describe very different kinds of measures.

3.1 Dirac measure

The Dirac measure at a point \(a\) has transform \(e^{-i a \xi}\) on the real line. This is a pure character, so its magnitude is constant. The example shows that a single point mass produces no decay at infinity.

3.2 Absolutely continuous measures

If a measure has density \(f\in L^1\), then its Fourier–Stieltjes transform agrees with the classical Fourier transform of \(f\). In this case the transform often decays at infinity, and many results from ordinary Fourier analysis apply directly. Absolutely continuous measures therefore represent the most familiar part of the theory.

3.3 Pure point measures

A pure point measure is a countable sum of weighted Dirac masses. Its transform is a corresponding sum of exponentials, often resembling a trigonometric series. Such transforms may exhibit almost periodic behavior and can fail to decay, reflecting the discrete structure of the underlying measure.

3.4 Singular continuous measures

Singular continuous measures are neither atomic nor absolutely continuous. Their transforms can behave in subtle ways: they may show partial decay, irregular oscillation, or fractal-like structure. Classical examples include measures supported on sets of zero Lebesgue measure but without point masses.

3.5 Mixed measures

Many measures decompose into atomic, absolutely continuous, and singular continuous parts. The Fourier–Stieltjes transform combines the contributions of all three. This makes it a useful diagnostic tool, since the different parts often leave distinct signatures in frequency space.

4 Relation to the Fourier transform

The Fourier–Stieltjes transform may be viewed as a direct extension of the classical Fourier transform. It retains the same oscillatory kernel while enlarging the class of admissible inputs.

4.1 Extension of the classical Fourier transform

When a function \(f\) is integrable, the measure \(d\mu(x)=f(x)\,dx\) is a finite absolutely continuous measure, and its Fourier–Stieltjes transform is exactly the classical Fourier transform of \(f\). This compatibility is one of the central motivations for the definition. The transform thus unifies function-based and measure-based Fourier analysis.

4.2 Comparison with \(L^1\) functions

For \(L^1\) functions, the Fourier transform often enjoys stronger decay and inversion properties than for arbitrary measures. The passage from functions to measures expands the class of objects but weakens some analytic conclusions. Nevertheless, many estimates valid for \(L^1\) functions remain true at the level of total variation.

4.3 Failure of inversion in general

Unlike the classical Fourier transform on suitable function spaces, the Fourier–Stieltjes transform does not generally admit a simple inversion formula. Different measures can share broad frequency features, and the transform alone may not determine fine measure-theoretic structure without additional hypotheses. Inversion is therefore more delicate in the measure setting.

4.4 Approximation by smooth functions

Many measures can be approximated by smooth functions or by convolutions with approximate identities. Their transforms then arise as limits of ordinary Fourier transforms of smooth approximants. This approach is often used to transfer intuition from classical analysis to more general measure-theoretic contexts.

5 Analytic and measure-theoretic properties

The transform reflects both the size and the distribution of the measure. Its analytic behavior depends on total variation, support, and the type of measure present.

5.1 Total variation of a measure

The total variation provides the natural norm for finite complex measures. It bounds the transform uniformly and controls convergence properties. In many arguments, estimates on the Fourier–Stieltjes transform are ultimately estimates in total variation.

5.2 Support and decay properties

The support of a measure influences the oscillatory structure of its transform, though not always in a simple way. Compact support does not by itself guarantee rapid decay, but it often improves regularity properties. Conversely, widely spread or highly singular measures may produce transforms with slow decay or persistent oscillation.

5.3 Riemann–Lebesgue-type results

For absolutely continuous measures with \(L^1\) densities, the transform vanishes at infinity, as in the Riemann–Lebesgue lemma. This property does not extend to all measures. Point masses, for instance, provide immediate counterexamples, since their transforms do not decay.

5.4 Differentiability issues

The transform of a finite measure need not be differentiable, and higher smoothness requires stronger moment conditions. If a measure has finite moments of a given order, then derivatives of the transform can often be expressed through corresponding integrals. Without such assumptions, the transform may be only continuous.

5.5 Singular measures and oscillatory behavior

Singular measures often produce transforms with intricate oscillations. These oscillations reflect irregular geometric or fractal structure in the original measure. The transform may fail to have a simple asymptotic law, making singular measures a rich source of examples and counterexamples.

6 Convergence and limits

Because the transform is defined through integration, it interacts naturally with various modes of convergence of measures. This makes it especially useful in approximation theory and weak convergence analysis.

6.1 Weak-* convergence of measures

If measures converge weakly-* against continuous compactly supported test functions, their transforms often converge pointwise under appropriate hypotheses. Since characters are bounded and continuous, they serve as natural test functions in many settings. Weak-* convergence is therefore closely connected to convergence of Fourier–Stieltjes transforms.

6.2 Pointwise convergence of transforms

Pointwise convergence of transforms may follow from convergence of measures, but the converse is much less direct. A sequence of transforms can converge pointwise while the measures themselves behave differently in stronger topologies. Pointwise limits are useful but do not fully determine measure convergence without additional structure.

6.3 Uniform convergence on compact sets

Under suitable boundedness assumptions on the measures, convergence of the transforms can often be strengthened to uniform convergence on compact subsets of the dual group. This is a consequence of equicontinuity and boundedness. Such results are common in compactness arguments.

6.4 Dominated convergence for measures

Dominated convergence applies because the exponential kernels are uniformly bounded. If a sequence of measures converges in a manner compatible with a dominating bound, then the transforms often converge by the same principle used for ordinary integrals. This makes the transform robust under limit operations.

6.5 Continuity with respect to measure norm

The transform depends continuously on the measure in the total variation norm. Small changes in a measure produce uniformly small changes in its transform. This stability is one of the reasons the Fourier–Stieltjes transform is a natural object in Banach space settings.

7 Positive-definite functions and harmonic analysis

The Fourier–Stieltjes transform is closely tied to positivity and representation theory. These connections give it a central role in harmonic analysis on groups.

7.1 Bochner’s theorem

Bochner’s theorem characterizes continuous positive-definite functions on abelian groups as Fourier transforms of finite positive measures. This result links positivity on the dual side with measure representation on the original side. It is one of the foundational theorems connecting measure theory and harmonic analysis.

7.2 Positive measures

When the measure is positive, the transform satisfies additional structural constraints. In particular, the transform is positive-definite in the appropriate sense. These constraints are often used to identify probability distributions and spectral objects.

7.3 Positive-definite transforms

A transform arising from a positive measure produces a positive-definite function. Such functions generate kernels with nonnegative quadratic forms, which makes them important in analysis and probability. The Fourier–Stieltjes transform provides a natural source of examples.

7.4 Spectral measures

Spectral measures encode how a function or operator decomposes into frequency components. Their Fourier–Stieltjes transforms summarize this spectral information in a compact analytic form. This viewpoint is central in the study of unitary representations and ergodic phenomena.

7.5 Role in harmonic analysis on groups

On locally compact abelian groups, the transform is a basic bridge between measures on the group and functions on the dual group. It supports convolution identities, positivity arguments, and structural decompositions. In this setting, the Fourier–Stieltjes transform is one of the main tools for translating between spatial and frequency descriptions.

8 Fourier–Stieltjes algebra

The Fourier–Stieltjes algebra organizes transforms of measures into a function space with algebraic and Banach space structure. It is an important object in abstract harmonic analysis.

8.1 Definition of the algebra

The Fourier–Stieltjes algebra consists of coefficient functions arising from unitary representations, or equivalently from Fourier–Stieltjes transforms of bounded measures in the appropriate group setting. It extends the Fourier algebra and captures a broader class of bounded functions. The algebraic viewpoint clarifies how transforms interact under addition and multiplication.

8.2 Connection with coefficient functions of unitary representations

Many functions in the algebra can be written as matrix coefficients of unitary representations. This representation-theoretic description links harmonic analysis to operator theory. It also explains why the algebra naturally reflects symmetries of the underlying group.

8.3 Norm structure

The Fourier–Stieltjes algebra carries a norm defined by the infimum of representation sizes or, equivalently in many cases, by the size of the corresponding measure. This norm makes the space into a Banach algebra. The norm controls both analytic behavior and algebraic operations.

8.4 Multiplicative properties

Products of Fourier–Stieltjes functions remain within the algebra in the standard group setting. This closure under multiplication is part of what makes the object an algebra rather than merely a linear space. Multiplicative structure is useful for studying composite frequencies and factorization phenomena.

8.5 Duality and Banach algebra aspects

The Fourier–Stieltjes algebra is deeply connected to duality theory for group algebras and measure algebras. Its Banach algebra structure supports spectral theory, ideal theory, and norm estimates. These features place it at the intersection of functional analysis and harmonic analysis.

9 Applications

The Fourier–Stieltjes transform appears in several branches of analysis and probability. Its flexibility makes it useful whenever measures represent randomness, spectra, or distributions of mass.

9.1 Probability theory

In probability, the transform of a probability measure is a characteristic function. This function uniquely encodes the distribution under mild conditions and is central to limit theorems and distributional convergence. The Fourier–Stieltjes transform therefore provides the analytic language of characteristic functions.

9.2 Characteristic functions of distributions

Characteristic functions are exactly Fourier–Stieltjes transforms of probability measures on the real line or on abelian groups. They are used to study sums of independent random variables, stability, and weak convergence. Their boundedness and continuity make them especially convenient in probabilistic arguments.

9.3 Signal analysis

In signal analysis, measures can model impulses, spikes, or mixed deterministic components. The Fourier–Stieltjes transform then describes frequency content even when the signal is not an ordinary function. This is useful in systems where point events matter as much as smooth variation.

9.4 Spectral analysis of measures

The transform is a natural tool for distinguishing spectral behavior among measures. It can reveal periodicity, quasi-periodicity, and singular structure in frequency space. Analysts often use it to compare different decomposition types within a single measure.

9.5 Ergodic and harmonic analysis

In ergodic theory and harmonic analysis, the transform helps study invariant measures, recurrence, and spectral decomposition. It provides a link between dynamical behavior and frequency structure. This makes it useful for analyzing long-term averages and correlation phenomena.

The Fourier–Stieltjes transform belongs to a broader family of transforms that extend Fourier methods beyond ordinary functions. Many related constructions appear in distribution theory, group representation theory, and multivariable analysis.

10.1 Fourier transform of distributions

Distributions generalize measures by allowing more singular linear functionals on test functions. Their Fourier transform extends the same basic oscillatory principle to even broader objects. In many cases, the measure-theoretic transform can be viewed as a special case of distributional Fourier analysis.

10.2 Bochner transform

The Bochner transform is related to representing positive-definite functions by measures. In that context, it links a function to a measure whose transform reproduces it. The terminology is especially common in harmonic analysis and probability.

10.3 Stieltjes integration framework

The use of Stieltjes integration emphasizes that the transform is built from integration against a measure rather than a density. This framework is well suited to jumps, atoms, and singular components. It also clarifies why the transform naturally handles non-smooth data.

10.4 Transforms on nonabelian groups

On nonabelian groups, characters are replaced by unitary representations and matrix coefficients. The resulting theory is richer and more complicated than in the abelian case. Nevertheless, the same philosophy persists: integrate a representation-theoretic kernel against a measure to obtain frequency information.

10.5 Higher-dimensional versions

In higher dimensions, the transform is defined on \(\mathbb{R}^n\) or on higher-dimensional locally compact abelian groups by integrating the appropriate characters. The basic properties remain similar, but geometric aspects such as support and decay can become more subtle. Multidimensional versions are widely used in analysis, probability, and applied mathematics.