Abstract Measure Grand Lebesgue Spaces

• Abstract Measure Grand Lebesgue Spaces are Banach function spaces that encode integrability via an interval of Lᵖ norms controlled by a generating function.
• They extend classical Lᵖ spaces by incorporating rearrangement invariance, sharp fundamental function estimates, and operator transfer principles across varied measure spaces.
• The framework unifies tools from harmonic analysis, Sobolev embeddings, and sparse domination, providing versatile methods for analyzing operator behavior and function space interpolation.

Abstract measure Grand Lebesgue spaces are Banach function spaces in which integrability is encoded by an entire interval of $L^p$-norms rather than a single exponent. In the ordinary Grand Lebesgue formulation, one works on a measure space $(X,\mathcal B,\mu)$ and controls a measurable function $f$ by a generating function $\psi$ through

$$\|f\|_{G(\psi)}=\sup_{p\in(a,b)}\frac{\|f\|_{L^p(X,\mu)}}{\psi(p)}<\infty.$$

In a parallel, limiting-scale formulation, abstract grand Lebesgue spaces $L^{r)}(X,\mu)$ are defined on finite measure spaces by taking a supremum over exponents $r-\varepsilon$ as $\varepsilon\downarrow 0$. The subject therefore combines Banach function space theory, rearrangement invariance, interpolation, harmonic analysis on abstract groups, and more recent sparse-domination methods on ball-basis measure spaces (Ostrovsky et al., 2015, Ostrovsky et al., 2016, Wang et al., 5 Jul 2026).

1. Definitions and principal frameworks

A standard abstract-measure definition fixes numbers $1\le a<b\le\infty$ and a strictly positive generating function $\psi$ on $(X,\mathcal B,\mu)$0, with $(X,\mathcal B,\mu)$1. The associated Grand Lebesgue space $(X,\mathcal B,\mu)$2 consists of measurable functions satisfying

$(X,\mathcal B,\mu)$3

This construction appears on nontrivial $(X,\mathcal B,\mu)$4-finite measure spaces, on probability spaces, and on general measurable spaces endowed with Lebesgue, Haar, or other measures. A degenerate choice of generator,

$(X,\mathcal B,\mu)$5

recovers the classical space $(X,\mathcal B,\mu)$6, so Grand Lebesgue spaces extend the Lebesgue–Riesz scale rather than replace it (Ostrovsky et al., 2016, Ostrovsky et al., 2015).

The same literature also uses generalized grand Lebesgue spaces of Iwaniec–Sbordone type. On a finite measure space, the abstract grand space $(X,\mathcal B,\mu)$7 is defined by

$(X,\mathcal B,\mu)$8

In the 2026 ball-basis framework, this definition is placed on a finite measure space $(X,\mathcal B,\mu)$9 equipped with an abstract family of measurable sets $f$0 satisfying axioms $f$1–$f$2, including finite positive measure for each ball, approximation of measurable sets by countable unions of balls, and a hull operation $f$3 with $f$4 (Wang et al., 5 Jul 2026).

These two frameworks are related but not identical. The ordinary $f$5 spaces are moment-envelope spaces indexed by an arbitrary generating function, while $f$6 is a logarithmically limiting scale around a fixed exponent $f$7. Both are treated as Banach function spaces, and both support abstract operator theory, but the 2026 theory ties the spaces to ball-basis geometry and sparse domination rather than to rearrangement alone (Wang et al., 5 Jul 2026).

2. Structural invariants and function-space relations

A central invariant of $f$8 is its fundamental function. For a measurable set $f$9 with $\psi$0,

$\psi$1

This quantity records the norm of indicators and is the natural replacement for the factor $\psi$2 that appears in classical $\psi$3-estimates. Closely related is the natural function of a measurable $\psi$4,

$\psi$5

defined on the interval where these moments are finite; then $\psi$6 with norm $\psi$7. This makes Grand Lebesgue spaces canonical “moment profile” spaces rather than ad hoc weighted suprema (Ostrovsky et al., 2015, Ostrovsky et al., 2015).

On finite diffuse measure spaces, the fundamental function is not merely a derived object: it determines the generating function and is determined by it. The 2015 paper “Fundamental function for Grand Lebesgue Spaces” proves a one-to-one and mutually continuous accordance between $\psi$8 and $\psi$9 by identifying $$\|f\|_{G(\psi)}=\sup_{p\in(a,b)}\frac{\|f\|_{L^p(X,\mu)}}{\psi(p)}<\infty.$$0 with an Orlicz space $$\|f\|_{G(\psi)}=\sup_{p\in(a,b)}\frac{\|f\|_{L^p(X,\mu)}}{\psi(p)}<\infty.$$1, using the Young–Fenchel transform, and recovering $$\|f\|_{G(\psi)}=\sup_{p\in(a,b)}\frac{\|f\|_{L^p(X,\mu)}}{\psi(p)}<\infty.$$2 from $$\|f\|_{G(\psi)}=\sup_{p\in(a,b)}\frac{\|f\|_{L^p(X,\mu)}}{\psi(p)}<\infty.$$3 through inverse-function and convex-duality formulas (Ostrovsky et al., 2015). In the infinite-measure setting, the same paper only asserts the weaker lower bound

$$\|f\|_{G(\psi)}=\sup_{p\in(a,b)}\frac{\|f\|_{L^p(X,\mu)}}{\psi(p)}<\infty.$$4

so the exact finite-measure correspondence does not survive unchanged (Ostrovsky et al., 2015).

Grand Lebesgue spaces are repeatedly described as rearrangement-invariant Banach function spaces, and several papers place them among “moment rearrangement invariant spaces.” They may coincide with exponential Orlicz spaces when $$\|f\|_{G(\psi)}=\sup_{p\in(a,b)}\frac{\|f\|_{L^p(X,\mu)}}{\psi(p)}<\infty.$$5 and $$\|f\|_{G(\psi)}=\sup_{p\in(a,b)}\frac{\|f\|_{L^p(X,\mu)}}{\psi(p)}<\infty.$$6 is convex; when the support endpoint $$\|f\|_{G(\psi)}=\sup_{p\in(a,b)}\frac{\|f\|_{L^p(X,\mu)}}{\psi(p)}<\infty.$$7, they generally do not coincide with the classical Orlicz, Lorentz, or Marcinkiewicz spaces (Ostrovsky et al., 2015). At a more abstract level, the 2022 operator-estimation paper defines a moment rearrangement invariant space $$\|f\|_{G(\psi)}=\sup_{p\in(a,b)}\frac{\|f\|_{L^p(X,\mu)}}{\psi(p)}<\infty.$$8 by requiring $$\|f\|_{G(\psi)}=\sup_{p\in(a,b)}\frac{\|f\|_{L^p(X,\mu)}}{\psi(p)}<\infty.$$9 to be an r.i. norm of the function $L^{r)}(X,\mu)$0, and ordinary $L^{r)}(X,\mu)$1 becomes the special case where that norm is a weighted $L^{r)}(X,\mu)$2-type supremum (Formica et al., 2022).

Localized comparison principles also belong to the structure theory. On a nontrivial $L^{r)}(X,\mu)$3-finite measure space, localized GLS norms on measurable subsets $L^{r)}(X,\mu)$4 with $L^{r)}(X,\mu)$5 lead to the “double ratio”

$L^{r)}(X,\mu)$6

When the supports are separated by $L^{r)}(X,\mu)$7, the exact value is $L^{r)}(X,\mu)$8, which generalizes Lyapunov’s inequality from ordinary $L^{r)}(X,\mu)$9-spaces to abstract-measure GLS in a normalized, sharp form (Ostrovsky et al., 2014).

3. Operator transfer principles on abstract measure spaces

A recurring theme is that a family of $r-\varepsilon$0 estimates can be lifted to a single Banach-space estimate in $r-\varepsilon$1 language. In the most general abstract form, if an operator $r-\varepsilon$2, not necessarily linear, satisfies

$r-\varepsilon$3

then the corresponding GLS estimate is

$r-\varepsilon$4

and the moment rearrangement invariant version replaces the GLS fundamental functions by the general fundamental-function ratio $r-\varepsilon$5 (Formica et al., 2022). A closely related 2023 paper formulates the same mechanism for operators with $r-\varepsilon$6, obtaining a GLS estimate controlled by $r-\varepsilon$7 (Formica et al., 2023).

The 2016 paper on composition and multiplicative operators gives exact $r-\varepsilon$8 norms on two $r-\varepsilon$9-finite measure spaces $\varepsilon\downarrow 0$0 and $\varepsilon\downarrow 0$1. For a measurable transformation $\varepsilon\downarrow 0$2 with Radon–Nikodým density

$\varepsilon\downarrow 0$3

the composition operator $\varepsilon\downarrow 0$4 has exact norm

$\varepsilon\downarrow 0$5

and the multiplicative operator $\varepsilon\downarrow 0$6 has exact norm

$\varepsilon\downarrow 0$7

These formulas induce explicit GLS bounds by minimizing the Lebesgue-operator norm times the source generator: $\varepsilon\downarrow 0$8 The paper emphasizes that the constant $\varepsilon\downarrow 0$9 is best possible (Ostrovsky et al., 2016).

A geometric example of the same transfer principle appears in the GLS reformulation of Alberti’s multidimensional Lusin theorem. If $1\le a<b\le\infty$0, $1\le a<b\le\infty$1, and $1\le a<b\le\infty$2, then for every $1\le a<b\le\infty$3 there are an open set $1\le a<b\le\infty$4 and a function $1\le a<b\le\infty$5 with $1\le a<b\le\infty$6 on $1\le a<b\le\infty$7, $1\le a<b\le\infty$8, and

$1\le a<b\le\infty$9

Here the exceptional-set dependence is encoded exactly by the fundamental function, and the dimension constant $\psi$0 is the same sharp constant as in Alberti’s original $\psi$1-estimate (Ostrovsky et al., 2015).

4. Harmonic-analysis realizations: groups, Fourier analysis, and ball-basis spaces

On unimodular locally compact groups equipped with Haar measure, ordinary Grand Lebesgue spaces can carry a genuine algebra structure. If $\psi$2 is unimodular, $\psi$3 is bi-invariant Haar measure, and $\psi$4, then

$\psi$5

for all $\psi$6; hence $\psi$7 is a Banach algebra under convolution. The proof is a direct lifting of Young’s inequality

$\psi$8

through the Grand Lebesgue norm, and the assumption $\psi$9 is essential (Formica et al., 2019).

Fourier analysis on infinite LCA groups provides a complementary realization. For an infinite compact or discrete LCA group $(X,\mathcal B,\mu)$00 with dual $(X,\mathcal B,\mu)$01, Haar measures $(X,\mathcal B,\mu)$02 and $(X,\mathcal B,\mu)$03, and the sharp Madiman–Xu $(X,\mathcal B,\mu)$04 Fourier norms, the GLS bounds are expressed through truncated fundamental functions. In the compact case one obtains

$(X,\mathcal B,\mu)$05

while in the discrete case

$(X,\mathcal B,\mu)$06

These formulas show that the Fourier transform acts on GLS by optimizing the classical $(X,\mathcal B,\mu)$07-$(X,\mathcal B,\mu)$08 inequalities against the generating function and then repackaging the result via truncated fundamental functions (Ostrovsky et al., 2017).

The generalized grand Lebesgue spaces $(X,\mathcal B,\mu)$09 exhibit a sharper algebraic dichotomy. On a locally compact Abelian group with Haar measure, $(X,\mathcal B,\mu)$10 is a Banach algebra under convolution if and only if $(X,\mathcal B,\mu)$11 is compact. In the same setting, multiplier theory links grand and small Lebesgue spaces through

$(X,\mathcal B,\mu)$12

for compact Abelian $(X,\mathcal B,\mu)$13, and

$(X,\mathcal B,\mu)$14

under finite Haar measure (Gurkanli, 2019).

The most explicitly abstract development is the 2026 theory of “Abstract Measure Grand Lebesgue Spaces and Applications.” There, the ambient space is a finite measure space $(X,\mathcal B,\mu)$15 with a separable ball basis $(X,\mathcal B,\mu)$16, and operator theory is organized around $(X,\mathcal B,\mu)$17 operators, defined by shell-control and bounded-oscillation conditions $(X,\mathcal B,\mu)$18-I) and $(X,\mathcal B,\mu)$19-II). The main theorem proves pointwise sparse domination,

$(X,\mathcal B,\mu)$20

and consequently norm estimates on $(X,\mathcal B,\mu)$21. The applications explicitly include maximal operators, Calderón–Zygmund operators on homogeneous spaces, and Carleson operators (Wang et al., 5 Jul 2026).

5. Geometric, Sobolev, and fractional models

Several papers develop Euclidean models that clarify how GLS interact with differentiation, extension, and fractional integration. In fractional Sobolev theory, the Sobolev–Grand Lebesgue space

$(X,\mathcal B,\mu)$22

leads to a sharp embedding

$(X,\mathcal B,\mu)$23

with the same best constant $(X,\mathcal B,\mu)$24 as in the classical fractional Sobolev inequality. The paper also defines derivative Grand Lebesgue spaces $(X,\mathcal B,\mu)$25 and gives weighted variants on convex domains (Ostrovsky et al., 2014).

The extension theorem for Sobolev–Grand Lebesgue spaces on Lipschitz domains is another uniform-in-$(X,\mathcal B,\mu)$26 lifting result. If $(X,\mathcal B,\mu)$27 denotes the space with norm

$(X,\mathcal B,\mu)$28

then there exists a linear bounded extension operator $(X,\mathcal B,\mu)$29 with

$(X,\mathcal B,\mu)$30

The generating function is preserved, because the classical extension estimates are controlled uniformly in $(X,\mathcal B,\mu)$31 (Formica et al., 2022).

Fractional integral and derivative estimates show the same pattern. For the Riesz potential $(X,\mathcal B,\mu)$32, the paper defines

$(X,\mathcal B,\mu)$33

and proves

$(X,\mathcal B,\mu)$34

For fractional derivatives of interval indicators and simple functions, the estimates are expressed through the fundamental function, for example

$(X,\mathcal B,\mu)$35

These papers explicitly state that the analysis is mainly Euclidean or half-line based rather than a complete abstract-measure theory, but they provide model cases in which the GLS machinery preserves sharp constants and reveals the correct transformed generating function (Ostrovsky et al., 2015, Ostrovsky et al., 2014).

6. Variants, interpolation theory, probability, and limitations

The parameter set of exponents can itself be restricted. For a Borel set $(X,\mathcal B,\mu)$36 with $(X,\mathcal B,\mu)$37, the restricted GLS norm is

$(X,\mathcal B,\mu)$38

If

$(X,\mathcal B,\mu)$39

then $(X,\mathcal B,\mu)$40 is equivalent to the full GLS norm. In the discrete case $(X,\mathcal B,\mu)$41, the criterion becomes

$(X,\mathcal B,\mu)$42

The same paper shows that restricted GLS remain Banach convolution algebras on unimodular locally compact groups when $(X,\mathcal B,\mu)$43 (Formica et al., 2019).

Interpolation places grand and small spaces inside a wider rearrangement-invariant scale. The 2017 interpolation paper proves that the interpolation spaces between grand, small, and classical Lebesgue spaces are Lorentz–Zygmund spaces or, in critical cases, $(X,\mathcal B,\mu)$44-spaces. Typical formulas are

$(X,\mathcal B,\mu)$45

and

$(X,\mathcal B,\mu)$46

As a consequence, any Lorentz–Zygmund space $(X,\mathcal B,\mu)$47 with $(X,\mathcal B,\mu)$48 and $(X,\mathcal B,\mu)$49 is an interpolation space between two grand spaces or between two small spaces (Fiorenza et al., 2017).

Embeddings between grand, small, and variable Lebesgue spaces require quantitative control of the rearranged exponent. On a finite measure space normalized by $(X,\mathcal B,\mu)$50, logarithmic conditions on $(X,\mathcal B,\mu)$51 and $(X,\mathcal B,\mu)$52 yield

$(X,\mathcal B,\mu)$53

and

$(X,\mathcal B,\mu)$54

while explicit counterexamples show that the reverse inclusions fail in general when $(X,\mathcal B,\mu)$55 (Cruz-Uribe et al., 2017). This situates grand and small spaces as endpoint logarithmic refinements of variable exponent theory rather than mere substitutes for it.

Probability theory supplies another important branch. On a probability space $(X,\mathcal B,\mu)$56, GLS membership implies exponential-type tail control via the Young–Fenchel transform, and factorizable almost sure convergence can be quantified in GLS norms. If

$(X,\mathcal B,\mu)$57

then for $(X,\mathcal B,\mu)$58,

$(X,\mathcal B,\mu)$59

with

$(X,\mathcal B,\mu)$60

The paper states that these estimates are essentially non-improvable up to multiplicative constants (Formica et al., 2024).

Across these variants, one limitation recurs. Several theories are genuinely abstract only under finite-measure assumptions: the 2026 ball-basis definition of $(X,\mathcal B,\mu)$61 assumes $(X,\mathcal B,\mu)$62, and the exact generating-function/fundamental-function correspondence of $(X,\mathcal B,\mu)$63 is proved in the finite diffuse case, while the infinite-measure theory is weaker (Wang et al., 5 Jul 2026, Ostrovsky et al., 2015). A plausible implication is that finite-measure normalization remains structurally central whenever the grand parameter is tied to limiting behavior near a fixed exponent rather than to an arbitrary generator $(X,\mathcal B,\mu)$64.