Bochner Integral Representation

• Bochner integral representation is a method to express vector-valued integrals as summable series of simple functions in Banach spaces.
• It enables compatibility with continuous linear functionals, providing scalar characterizations and dual representation in various integration theories.
• Extensions of Bochner integration apply to locally convex, ordered spaces and operator-stochastic frameworks, broadening its practical significance.

Bochner integral representation denotes a family of constructions in which vector-valued objects are realized through integrals taken in a Banach space or a more general topological or ordered vector space. In the classical Banach-space setting, the central representation principle is that an integrable function can be written as an absolutely summable series of simple functions,

$f\sim f_1+f_2+\cdots,$

and its integral is represented by

$\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$

In later developments, the same idea appears as operator-valued integral identities, dual representation formulas, and parameterized feature superpositions in analysis, control, and approximation theory (Mikusinski, 2014, Bogdan, 2010).

1. Banach-space foundations

The basic Banach-space theory starts from a measure space and a Banach target $E$, with simple $E$-valued functions of the form

$s=\sum_{k=1}^n x_k \mathbf 1_{A_k},$

and integral

$\int_X s\,d\mu = x_1\mu(A_1)+\cdots+x_n\mu(A_n).$

At this level one already has the fundamental estimate

$$\left\|\int_X s\,d\mu\right\| \le \int_X \|s\|\,d\mu,$$

which is the norm inequality from which the later Bochner theory is developed (Mikusinski, 2014).

A characteristic representation theorem is that a function $f:X\to E$ is Bochner integrable when it admits an absolutely summable simple-function expansion

$$f\sim f_1+f_2+\cdots,\qquad \sum_{n=1}^\infty \|f_n\|_1<\infty,$$

and then

$\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$

The integral is independent of the chosen expansion, and $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$0 is a Banach space in the Banach-valued case (Mikusinski, 2014).

A closely related prering-based construction develops the same theme in a different order. Instead of starting from a complete measure on a $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$1-algebra, it begins with a positive measure on a prering, constructs simple functions on that prering, and defines $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$2 as the class of almost-everywhere limits of basic sequences of simple $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$3-valued functions. In that framework, the Bochner integral is represented by

$\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$4

with density of simple functions in the $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$5-type norm and eventual recovery of the classical theory after passage to the minimal complete Lebesgue extension (Bogdan, 2010).

A formalized variant in Coq uses the same constructive core. A function $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$6, with $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$7 Banach, is declared Bochner-integrable when there exists a sequence of integrable simple functions $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$8 such that

$\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$9

and then

$E$0

This makes the representation by simple approximants explicit at the level of proof objects as well as mathematics (Boldo et al., 2022).

2. Duality, scalarization, and representing functionals

One of the strongest structural facts about the Bochner integral is compatibility with continuous linear functionals. In the Banach-space theory, if $E$1 and $E$2, then $E$3 is scalar integrable and

$E$4

This gives a scalar characterization of the vector integral and is one of the standard bridges between Bochner and Pettis viewpoints (Mikusinski, 2014).

A further extension replaces scalar-valued norms by $E$5-valued norms, where $E$6 is a Dedekind complete unital $E$7-algebra. In that setting, for an $E$8-Banach space $E$9, one defines $E$0-Bochner integrability again through approximation by $E$1-simple functions, and under an $E$2-Radon–Nikodým property for $E$3 obtains an exact dual representation theorem: $E$4 with

$E$5

This is a direct $E$6-valued analogue of the classical Bochner duality formula (Zhang et al., 2024).

In the finitely additive setting, the representation problem becomes one of transferring vector integration to a countably additive model. For finitely additive Bochner or Pettis integrals, the paper on finitely additive vector integration constructs a standard representation $E$7 on a compact Hausdorff probability space such that

$E$8

for the prescribed test class. In the norm-integrable case this produces an actual $E$9; in the separable Pettis case it characterizes Pettis integrability by the existence of such a concrete countably additive representation (Cassese, 2 Feb 2025).

3. Extensions of the ambient space and of the measure-theoretic setting

The Banach-space construction extends seminorm-by-seminorm to complete locally convex spaces. For a complete locally convex space $s=\sum_{k=1}^n x_k \mathbf 1_{A_k},$0 with defining seminorm family $s=\sum_{k=1}^n x_k \mathbf 1_{A_k},$1, integrability is defined by approximation with simple functions such that, for every seminorm,

$s=\sum_{k=1}^n x_k \mathbf 1_{A_k},$2

and the partial sums converge pointwise in each seminorm. The resulting integral is still represented by the same series formula,

$s=\sum_{k=1}^n x_k \mathbf 1_{A_k},$3

with seminormwise control and compatibility with continuous linear functionals. In this locally convex version, completeness of the resulting $s=\sum_{k=1}^n x_k \mathbf 1_{A_k},$4 is explicitly left open (Mikusinski, 2014).

For ordered vector spaces that are not Banach spaces, one approach is to cover the space by Banach pieces. If $s=\sum_{k=1}^n x_k \mathbf 1_{A_k},$5 is an Archimedean directed ordered vector space, the theory of Banach covers defines $s=\sum_{k=1}^n x_k \mathbf 1_{A_k},$6 to be integrable when it is almost everywhere represented by a Bochner-integrable function with values in one Banach member of the cover. The integral is then

$s=\sum_{k=1}^n x_k \mathbf 1_{A_k},$7

where $s=\sum_{k=1}^n x_k \mathbf 1_{A_k},$8 is any Banach-valued representative of $s=\sum_{k=1}^n x_k \mathbf 1_{A_k},$9, and this is independent of the chosen Banach piece. When the order dual separates points, the integral is further characterized by

$\int_X s\,d\mu = x_1\mu(A_1)+\cdots+x_n\mu(A_n).$0

This gives a representation theory for Bochner integration in non-Banach ordered spaces (Rooij et al., 2016).

An even more abstract generalization treats vector-lattice-valued functions integrated against vector-lattice-valued measures under axiomatized convergence notions. There the integral of $\int_X s\,d\mu = x_1\mu(A_1)+\cdots+x_n\mu(A_n).$1 over $\int_X s\,d\mu = x_1\mu(A_1)+\cdots+x_n\mu(A_n).$2 is defined by choosing a defining sequence of simple functions $\int_X s\,d\mu = x_1\mu(A_1)+\cdots+x_n\mu(A_n).$3 and a map $\int_X s\,d\mu = x_1\mu(A_1)+\cdots+x_n\mu(A_n).$4 such that

$\int_X s\,d\mu = x_1\mu(A_1)+\cdots+x_n\mu(A_n).$5

and then setting

$\int_X s\,d\mu = x_1\mu(A_1)+\cdots+x_n\mu(A_n).$6

The construction is independent of the defining sequence and reduces to the ordinary Lebesgue integral in the scalar $\int_X s\,d\mu = x_1\mu(A_1)+\cdots+x_n\mu(A_n).$7-finite regular case (Boccuto et al., 2021).

4. Operator, semigroup, and stochastic forms of Bochner representation

In operator theory, Bochner integral representation often appears as an interchange formula. For Krivine’s function calculus, if

$\int_X s\,d\mu = x_1\mu(A_1)+\cdots+x_n\mu(A_n).$8

is continuous and

$\int_X s\,d\mu = x_1\mu(A_1)+\cdots+x_n\mu(A_n).$9

is integrable, then for the Banach lattice calculus associated with $$\left\|\int_X s\,d\mu\right\| \le \int_X \|s\|\,d\mu,$$0,

$$\left\|\int_X s\,d\mu\right\| \le \int_X \|s\|\,d\mu,$$1

where the right-hand side is a Bochner integral in the lattice $$\left\|\int_X s\,d\mu\right\| \le \int_X \|s\|\,d\mu,$$2 (Troitsky et al., 2017).

A related commutation theorem is Hille’s theorem in complete locally convex spaces. If $$\left\|\int_X s\,d\mu\right\| \le \int_X \|s\|\,d\mu,$$3 is closed, $$\left\|\int_X s\,d\mu\right\| \le \int_X \|s\|\,d\mu,$$4 almost everywhere, and both $$\left\|\int_X s\,d\mu\right\| \le \int_X \|s\|\,d\mu,$$5 and $$\left\|\int_X s\,d\mu\right\| \le \int_X \|s\|\,d\mu,$$6 are Bochner integrable, then for every measurable $$\left\|\int_X s\,d\mu\right\| \le \int_X \|s\|\,d\mu,$$7,

$$\left\|\int_X s\,d\mu\right\| \le \int_X \|s\|\,d\mu,$$8

This provides the standard domain-stability statement needed once a locally convex Bochner representation is already available (Sullivan, 2024).

For operator-valued Riccati equations on a Hilbert space, the Bochner integral representation becomes an operator identity. Under the assumptions of the Riccati well-posedness theorem, the mild form

$$\left\|\int_X s\,d\mu\right\| \le \int_X \|s\|\,d\mu,$$9

is equivalent to the Bochner integral form

$f:X\to E$0

The proof uses representation of bounded quadratic forms by bounded self-adjoint operators, so the vectorwise mild identity upgrades to a genuine operator-valued Bochner integral identity (Cheung, 2023).

In stochastic analysis, Itô integrals can themselves be represented by Bochner or Lebesgue integrals. If $f:X\to E$1 is Hilbert and

$f:X\to E$2

then there exists

$f:X\to E$3

such that

$f:X\to E$4

The underlying proof is based on a Riesz-type duality theorem for mixed-norm stochastic process spaces (Lü et al., 2010).

5. Nonlinear, analytic, and feature-based representation theorems

A nonlinear analogue is the Choquet–Bochner integral with respect to vector capacities. For a vector capacity $f:X\to E$5, the Choquet–Bochner integral of a scalar measurable function $f:X\to E$6 over $f:X\to E$7 is

$f:X\to E$8

In a Banach lattice with order continuous norm, every comonotonic additive and monotone operator

$f:X\to E$9

with $$f\sim f_1+f_2+\cdots,\qquad \sum_{n=1}^\infty \|f_n\|_1<\infty,$$0 is represented by a unique upper continuous vector capacity through

$$f\sim f_1+f_2+\cdots,\qquad \sum_{n=1}^\infty \|f_n\|_1<\infty,$$1

This is a vector-valued Choquet analogue of Riesz-type representation (Gal et al., 2020).

In complex analysis, a strip-holomorphic extension of Bochner’s theorem represents a holomorphic positive definite function on a horizontal strip $$f\sim f_1+f_2+\cdots,\qquad \sum_{n=1}^\infty \|f_n\|_1<\infty,$$2 as

$$f\sim f_1+f_2+\cdots,\qquad \sum_{n=1}^\infty \|f_n\|_1<\infty,$$3

where $$f\sim f_1+f_2+\cdots,\qquad \sum_{n=1}^\infty \|f_n\|_1<\infty,$$4 is a unique non-negative measure that is exponentially finite with respect to $$f\sim f_1+f_2+\cdots,\qquad \sum_{n=1}^\infty \|f_n\|_1<\infty,$$5. The horizontal sections recover classical Bochner representations, and the imaginary section recovers Widder-type Laplace representations, thereby placing both within a single Fourier–Laplace framework (Buescu et al., 2018).

In approximation theory and neural networks, Bochner representation appears as an infinite-width superposition formula. With $$f\sim f_1+f_2+\cdots,\qquad \sum_{n=1}^\infty \|f_n\|_1<\infty,$$6, feature map $$f\sim f_1+f_2+\cdots,\qquad \sum_{n=1}^\infty \|f_n\|_1<\infty,$$7, and coefficient density $$f\sim f_1+f_2+\cdots,\qquad \sum_{n=1}^\infty \|f_n\|_1<\infty,$$8, the resulting function

$$f\sim f_1+f_2+\cdots,\qquad \sum_{n=1}^\infty \|f_n\|_1<\infty,$$9

belongs to $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$0 and satisfies the Bochner identity

$\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$1

This is paired with the norm bound

$\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$2

under the essential boundedness assumption $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$3 almost everywhere, and with a variation-space estimate

$\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$4

The same paper also shows that a pointwise scalar integral may exist without yielding an element of the Banach space, and in that case no Bochner representation in $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$5 is available (Kainen et al., 2023).

6. Scalar specialization, formalization, and scope

In the real-valued finite-measure case, the Bochner integral does not define a genuinely new scalar integral. On a finite measure space, a real-valued measurable and almost everywhere finite function is Bochner integrable if and only if it is integrable in the paper’s real-valued measurable-function sense, and the values coincide: $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$6 This identifies the real-valued Bochner integral with the ordinary scalar theory in that setting (Lo et al., 2019).

Formal proof development has sharpened the constructive content of this equivalence. In the Coq formalization, simple functions are encoded as dependent records, Bochner integrability is itself a dependent type carrying an explicit approximating sequence, and the resulting theory proves dominated convergence and equality with the existing formalization of the nonnegative Lebesgue integral for nonnegative real-valued functions (Boldo et al., 2022).

The lattice structure of Bochner spaces imposes further scope conditions. When $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$7 is a Banach lattice, the space $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$8 is itself a Banach lattice, and structural properties are reflected exactly: $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$9 is a $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$00-space if and only if $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$01 is a $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$02-space, and $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$03 has the sequential Fatou property if and only if $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$04 does. Under $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$05-order continuity, order-theoretic convergence in $\int_X f\,d\mu=\sum_{n=1}^\infty \int_X f_n\,d\mu.$06 yields convergence of Bochner integrals (Zabeti, 2018).

The general theory also has genuine nonuniqueness and boundary phenomena. In Banach-cover integration for ordered spaces, different Banach covers can yield different integrals; this shows that representation by Bochner integration outside a single Banach ambient space requires canonical choices if one wants uniqueness (Rooij et al., 2016). More broadly, the literature surveyed here shows that “Bochner integral representation” is not a single theorem but a family of representation mechanisms: simple-function series in Banach spaces, seminormwise or orderwise extensions beyond Banach spaces, operator and stochastic interchange formulas, nonlinear representation theorems, and concrete countably additive models for finitely additive vector integration.