Variable Exponent Morrey Spaces

• Variable exponent Morrey spaces are function spaces that combine variable local integrability and growth control via Morrey parameters or general control functions.
• They generalize classical Morrey spaces using formulations based on ball norms, control functions, and complementary definitions to support diverse operator theories.
• Applications include analysis of nonstandard growth in PDEs and extensions to Herz–Morrey and Triebel–Lizorkin scales, though standardizing definitions remains an ongoing challenge.

Variable exponent Morrey spaces are Morrey-type function spaces in which local integrability is measured by a variable exponent and local growth is controlled either by a Morrey parameter or by a more general control function. In the literature considered here, the notion of Morrey space with variable exponents “might differ from work to work”: one finds ball-based formulations, generalized spaces with a Morrey control $u(x,r)$, global Morrey-type spaces with a defining function $w(x,r)$, and local “complementary” spaces based on $\Omega\setminus B(x_0,r)$. Across these formulations, the common theme is the combination of variable exponent Lebesgue theory with Morrey localization, which supports a broad operator theory for maximal operators, potentials, singular integrals, commutators, Stein–Weiss inequalities, and Triebel–Lizorkin refinements (Gurbuz, 5 Jul 2026).

1. Definitions and principal formulations

A variable exponent is a measurable function $p(\cdot):\mathbb{R}^n\to[1,\infty)$, with essential bounds

$$p_-:=\operatorname*{ess\,inf}_{x\in\mathbb{R}^n}p(x),\qquad p_+:=\operatorname*{ess\,sup}_{x\in\mathbb{R}^n}p(x).$$

The associated variable exponent Lebesgue space $L^{p(\cdot)}(\mathbb{R}^n)$ is defined by the Luxemburg norm

$$\|f\|_{L^{p(\cdot)}}:= \inf\left\{\lambda>0:\int_{\mathbb{R}^n}\left|\frac{f(x)}{\lambda}\right|^{p(x)}dx\le 1\right\}.$$

When $p(\cdot)\equiv p_0$, this reduces to the classical space $L^{p_0}(\mathbb{R}^n)$.

One widely used generalized Morrey formulation is

$$\mathcal{M}_{p(\cdot),u}(\mathbb{R}^n) := \left\{f\in L^{p(\cdot)}_{\mathrm{loc}}(\mathbb{R}^n): \|f\|_{\mathcal{M}_{p(\cdot),u}}<\infty\right\},$$

with

$w(x,r)$0

where $w(x,r)$1 is a measurable Morrey control function. This definition recovers variable exponent Lebesgue behavior when $w(x,r)$2, and in the constant-exponent case $w(x,r)$3, $w(x,r)$4, it is isomorphic to the classical Morrey space $w(x,r)$5 (Gurbuz, 5 Jul 2026).

A second formulation, used for fully variable Morrey parameters, is

$w(x,r)$6

where $w(x,r)$7 is bounded and

$w(x,r)$8

The norm is

$w(x,r)$9

This is the bounded-domain framework used for Stein–Weiss and Poincaré-type inequalities (Cruz-Uribe et al., 2 Oct 2025).

A third formulation places the exponent at the center of the ball: $\Omega\setminus B(x_0,r)$0 This ball-centered definition is the base space for variable exponent Triebel–Lizorkin–Morrey scales (Caetano et al., 2018).

There are also generalized global Morrey-type spaces on unbounded domains, defined by

$\Omega\setminus B(x_0,r)$1

and local “complementary” spaces

$\Omega\setminus B(x_0,r)$2

which measure the behavior of $\Omega\setminus B(x_0,r)$3 outside shrinking balls around a fixed point $\Omega\setminus B(x_0,r)$4 (Bokayev et al., 2021).

2. Structural hypotheses and basic analytic tools

The standard regularity hypothesis on variable exponents is log-Hölder continuity. In one common global form, one assumes constants $\Omega\setminus B(x_0,r)$5 such that

$\Omega\setminus B(x_0,r)$6

and

$\Omega\setminus B(x_0,r)$7

These conditions prevent oscillation that is too fast locally or at infinity, and they guarantee boundedness of the Hardy–Littlewood maximal operator on $\Omega\setminus B(x_0,r)$8 in the settings considered here (Gurbuz, 5 Jul 2026).

A fundamental consequence is control of characteristic functions of balls. For $\Omega\setminus B(x_0,r)$9 in the global log-Hölder class, one has

$p(\cdot):\mathbb{R}^n\to[1,\infty)$0

which is the variable-exponent analogue of $p(\cdot):\mathbb{R}^n\to[1,\infty)$1. More refined estimates of the form

$p(\cdot):\mathbb{R}^n\to[1,\infty)$2

play the role of uniform $p(\cdot):\mathbb{R}^n\to[1,\infty)$3-type control in many Morrey and Herz–Morrey arguments (Wu, 2014).

Variable Hölder inequalities are equally basic. In the formulation used for $p(\cdot):\mathbb{R}^n\to[1,\infty)$4,

$p(\cdot):\mathbb{R}^n\to[1,\infty)$5

and analogous inequalities hold in the weighted and localized settings. In the Triebel–Lizorkin–Morrey theory, these are supplemented by ball-norm asymptotics such as

$p(\cdot):\mathbb{R}^n\to[1,\infty)$6

for $p(\cdot):\mathbb{R}^n\to[1,\infty)$7 (Caetano et al., 2018).

For bounded-domain variable exponent Morrey spaces, there is also a norm–modular equivalence: $p(\cdot):\mathbb{R}^n\to[1,\infty)$8 This reduction is indispensable in the proof of weighted fractional inequalities and Poincaré-type results (Cruz-Uribe et al., 2 Oct 2025).

3. Morrey control functions, weak spaces, and boundedness theory

In the generalized space $p(\cdot):\mathbb{R}^n\to[1,\infty)$9, the control function $$p_-:=\operatorname*{ess\,inf}_{x\in\mathbb{R}^n}p(x),\qquad p_+:=\operatorname*{ess\,sup}_{x\in\mathbb{R}^n}p(x).$$0 is not arbitrary in operator theory. A class $$p_-:=\operatorname*{ess\,inf}_{x\in\mathbb{R}^n}p(x),\qquad p_+:=\operatorname*{ess\,sup}_{x\in\mathbb{R}^n}p(x).$$1 of admissible Morrey control functions is defined by a discrete Hardy-type condition: $$p_-:=\operatorname*{ess\,inf}_{x\in\mathbb{R}^n}p(x),\qquad p_+:=\operatorname*{ess\,sup}_{x\in\mathbb{R}^n}p(x).$$2 together with a local scaling bound

$$p_-:=\operatorname*{ess\,inf}_{x\in\mathbb{R}^n}p(x),\qquad p_+:=\operatorname*{ess\,sup}_{x\in\mathbb{R}^n}p(x).$$3

The main role of $$p_-:=\operatorname*{ess\,inf}_{x\in\mathbb{R}^n}p(x),\qquad p_+:=\operatorname*{ess\,sup}_{x\in\mathbb{R}^n}p(x).$$4 is that it allows dyadic expansions produced by nonlocal operators to be reduced to a single Morrey weight at scale $$p_-:=\operatorname*{ess\,inf}_{x\in\mathbb{R}^n}p(x),\qquad p_+:=\operatorname*{ess\,sup}_{x\in\mathbb{R}^n}p(x).$$5 (Gurbuz, 5 Jul 2026).

Weak versions are built in parallel. For example,

$$p_-:=\operatorname*{ess\,inf}_{x\in\mathbb{R}^n}p(x),\qquad p_+:=\operatorname*{ess\,sup}_{x\in\mathbb{R}^n}p(x).$$6

These spaces are the natural targets when strong-type boundedness fails at critical parameter values (Gurbuz, 5 Jul 2026).

The operator theory on variable exponent Morrey spaces is extensive. For higher-order commutators of rough fractional maximal operators with variable kernels, one has strong boundedness

$$p_-:=\operatorname*{ess\,inf}_{x\in\mathbb{R}^n}p(x),\qquad p_+:=\operatorname*{ess\,sup}_{x\in\mathbb{R}^n}p(x).$$7

under a strict interior condition

$$p_-:=\operatorname*{ess\,inf}_{x\in\mathbb{R}^n}p(x),\qquad p_+:=\operatorname*{ess\,sup}_{x\in\mathbb{R}^n}p(x).$$8

together with rough-kernel and Morrey-weight hypotheses (Gurbuz, 5 Jul 2026).

At the level of weighted fractional integrals, the bounded-domain Stein–Weiss inequality gives

$$p_-:=\operatorname*{ess\,inf}_{x\in\mathbb{R}^n}p(x),\qquad p_+:=\operatorname*{ess\,sup}_{x\in\mathbb{R}^n}p(x).$$9

provided

$L^{p(\cdot)}(\mathbb{R}^n)$0

$L^{p(\cdot)}(\mathbb{R}^n)$1

This extends Stein–Weiss theory from variable exponent Lebesgue spaces and constant-exponent Morrey spaces to $L^{p(\cdot)}(\mathbb{R}^n)$2 (Cruz-Uribe et al., 2 Oct 2025).

For global Morrey-type spaces on unbounded sets, the Hardy–Littlewood maximal operator and potential operators are bounded under Hardy-type integral conditions on the defining functions $L^{p(\cdot)}(\mathbb{R}^n)$3. A representative condition is

$L^{p(\cdot)}(\mathbb{R}^n)$4

which yields boundedness of $L^{p(\cdot)}(\mathbb{R}^n)$5 from $L^{p(\cdot)}(\mathbb{R}^n)$6 to $L^{p(\cdot)}(\mathbb{R}^n)$7 (Bokayev et al., 2021).

In the complementary setting, one has

$L^{p(\cdot)}(\mathbb{R}^n)$8

and, when

$L^{p(\cdot)}(\mathbb{R}^n)$9

also

$$\|f\|_{L^{p(\cdot)}}:= \inf\left\{\lambda>0:\int_{\mathbb{R}^n}\left|\frac{f(x)}{\lambda}\right|^{p(x)}dx\le 1\right\}.$$0

with conditions formulated by Zygmund-type integral inequalities on $$\|f\|_{L^{p(\cdot)}}:= \inf\left\{\lambda>0:\int_{\mathbb{R}^n}\left|\frac{f(x)}{\lambda}\right|^{p(x)}dx\le 1\right\}.$$1, without monotonicity assumptions on these functions (Guliyev et al., 2011).

4. Endpoint phenomena and interpolation

A characteristic feature of variable exponent Morrey theory is the presence of critical lines at which the target exponent degenerates. For the higher-order commutator problem, the critical configuration is

$$\|f\|_{L^{p(\cdot)}}:= \inf\left\{\lambda>0:\int_{\mathbb{R}^n}\left|\frac{f(x)}{\lambda}\right|^{p(x)}dx\le 1\right\}.$$2

At points where $$\|f\|_{L^{p(\cdot)}}:= \inf\left\{\lambda>0:\int_{\mathbb{R}^n}\left|\frac{f(x)}{\lambda}\right|^{p(x)}dx\le 1\right\}.$$3, the formula

$$\|f\|_{L^{p(\cdot)}}:= \inf\left\{\lambda>0:\int_{\mathbb{R}^n}\left|\frac{f(x)}{\lambda}\right|^{p(x)}dx\le 1\right\}.$$4

gives $$\|f\|_{L^{p(\cdot)}}:= \inf\left\{\lambda>0:\int_{\mathbb{R}^n}\left|\frac{f(x)}{\lambda}\right|^{p(x)}dx\le 1\right\}.$$5. In that regime, the target variable exponent Lebesgue space $$\|f\|_{L^{p(\cdot)}}:= \inf\left\{\lambda>0:\int_{\mathbb{R}^n}\left|\frac{f(x)}{\lambda}\right|^{p(x)}dx\le 1\right\}.$$6 no longer makes sense globally, and strong-type mapping fails because the Luxemburg norm blows up (Gurbuz, 5 Jul 2026).

The replacement is a weak endpoint bound. Under the critical scaling and an Adams-type relation between Morrey weights,

$$\|f\|_{L^{p(\cdot)}}:= \inf\left\{\lambda>0:\int_{\mathbb{R}^n}\left|\frac{f(x)}{\lambda}\right|^{p(x)}dx\le 1\right\}.$$7

the commutator maps

$$\|f\|_{L^{p(\cdot)}}:= \inf\left\{\lambda>0:\int_{\mathbb{R}^n}\left|\frac{f(x)}{\lambda}\right|^{p(x)}dx\le 1\right\}.$$8

where

$$\|f\|_{L^{p(\cdot)}}:= \inf\left\{\lambda>0:\int_{\mathbb{R}^n}\left|\frac{f(x)}{\lambda}\right|^{p(x)}dx\le 1\right\}.$$9

This shows that the critical line remains accessible, but only in an $p(\cdot)\equiv p_0$0-type weak Morrey scale (Gurbuz, 5 Jul 2026).

Interpolation restores interior strong estimates from weak endpoints. In an abstract Grafakos–Martell-type framework, if a sublinear operator $p(\cdot)\equiv p_0$1 is bounded both

$p(\cdot)\equiv p_0$2

and

$p(\cdot)\equiv p_0$3

then for $p(\cdot)\equiv p_0$4,

$p(\cdot)\equiv p_0$5

and

$p(\cdot)\equiv p_0$6

is bounded. In the commutator setting this yields a full continuum of intermediate strong Morrey estimates between the weak endpoint and the interior strong case (Gurbuz, 5 Jul 2026).

Endpoint questions also appear in other branches of the theory. In Herz–Morrey spaces with variable exponent, parameter restrictions such as

$p(\cdot)\equiv p_0$7

or

$p(\cdot)\equiv p_0$8

arise from convergence of the geometric series generated by dyadic decompositions. The papers explicitly note that at the boundary some geometric series fail to converge, so the admissible intervals are not merely formal (Wu, 2014).

5. Herz–Morrey, Triebel–Lizorkin–Morrey, and other extensions

A major extension of variable exponent Morrey theory is the Herz–Morrey scale. With dyadic annuli $p(\cdot)\equiv p_0$9 and characteristic functions $L^{p_0}(\mathbb{R}^n)$0, the homogeneous Herz–Morrey space with variable exponent is

$L^{p_0}(\mathbb{R}^n)$1

When $L^{p_0}(\mathbb{R}^n)$2, it reduces to the Herz space $L^{p_0}(\mathbb{R}^n)$3. These spaces are Morrey-type because the parameter $L^{p_0}(\mathbb{R}^n)$4 controls the growth of radial partial sums, while $L^{p_0}(\mathbb{R}^n)$5 supplies radial decay or growth (Wu, 2014).

This radialized framework supports extensive operator theory. Sublinear operators, fractional integrals, fractional Hardy-type operators of variable order, commutators generated by Riesz potentials, multilinear Hausdorff commutators, singular integrals with variable kernels, and fractional-differentiation commutators have all been shown to be bounded on suitable weighted or unweighted variable exponent Herz–Morrey spaces under explicit exponent, weight, and kernel conditions (Wu, 2013).

A more recent extension is the weighted grand Herz–Morrey scale with variable exponents,

$L^{p_0}(\mathbb{R}^n)$6

whose quasi-norm contains an additional supremum over a grand parameter $L^{p_0}(\mathbb{R}^n)$7. On top of this, weighted grand Herz–Morrey–Triebel–Lizorkin spaces

$L^{p_0}(\mathbb{R}^n)$8

are defined by Littlewood–Paley decompositions and admit equivalent quasi-norms via maximal functions and Peetre-type maximal functions. This places variable exponent Morrey-type control inside a smoothness scale (Wang et al., 19 Feb 2025).

The fully variable Triebel–Lizorkin–Morrey scale takes a different route. Starting from

$L^{p_0}(\mathbb{R}^n)$9

one defines

$$\mathcal{M}_{p(\cdot),u}(\mathbb{R}^n) := \left\{f\in L^{p(\cdot)}_{\mathrm{loc}}(\mathbb{R}^n): \|f\|_{\mathcal{M}_{p(\cdot),u}}<\infty\right\},$$0

through the sequence norm

$$\mathcal{M}_{p(\cdot),u}(\mathbb{R}^n) := \left\{f\in L^{p(\cdot)}_{\mathrm{loc}}(\mathbb{R}^n): \|f\|_{\mathcal{M}_{p(\cdot),u}}<\infty\right\},$$1

The key technical advance here is a vector-valued convolution inequality in $$\mathcal{M}_{p(\cdot),u}(\mathbb{R}^n) := \left\{f\in L^{p(\cdot)}_{\mathrm{loc}}(\mathbb{R}^n): \|f\|_{\mathcal{M}_{p(\cdot),u}}<\infty\right\},$$2 that replaces the Hardy–Littlewood maximal inequality in the fully variable setting and leads to Peetre maximal characterizations and independence from the admissible dyadic system (Caetano et al., 2018).

These extensions show that variable exponent Morrey spaces serve not only as stand-alone objects but also as the base layer of richer scales carrying radial, weighted, grand, and frequency-localized structure. This suggests a broad architecture in which Morrey control, variable local integrability, and smoothness can be combined systematically.

6. Applications, reductions, and open problems

The principal motivation repeatedly stated in the literature is the treatment of nonstandard growth. Variable exponent analysis is used in PDEs with nonstandard growth, including electrorheological fluids, non-Newtonian fluids, elasticity with variable growth, and related inhomogeneous media. Morrey control adds local growth information that is fundamental in elliptic and parabolic regularity theory, so variable exponent Morrey spaces combine spatially varying integrability with localized scale control (Wang et al., 19 Feb 2025).

Several reductions recover classical settings. If $$\mathcal{M}_{p(\cdot),u}(\mathbb{R}^n) := \left\{f\in L^{p(\cdot)}_{\mathrm{loc}}(\mathbb{R}^n): \|f\|_{\mathcal{M}_{p(\cdot),u}}<\infty\right\},$$3 and $$\mathcal{M}_{p(\cdot),u}(\mathbb{R}^n) := \left\{f\in L^{p(\cdot)}_{\mathrm{loc}}(\mathbb{R}^n): \|f\|_{\mathcal{M}_{p(\cdot),u}}<\infty\right\},$$4, then $$\mathcal{M}_{p(\cdot),u}(\mathbb{R}^n) := \left\{f\in L^{p(\cdot)}_{\mathrm{loc}}(\mathbb{R}^n): \|f\|_{\mathcal{M}_{p(\cdot),u}}<\infty\right\},$$5 is isomorphic to the classical Morrey space $$\mathcal{M}_{p(\cdot),u}(\mathbb{R}^n) := \left\{f\in L^{p(\cdot)}_{\mathrm{loc}}(\mathbb{R}^n): \|f\|_{\mathcal{M}_{p(\cdot),u}}<\infty\right\},$$6. If $$\mathcal{M}_{p(\cdot),u}(\mathbb{R}^n) := \left\{f\in L^{p(\cdot)}_{\mathrm{loc}}(\mathbb{R}^n): \|f\|_{\mathcal{M}_{p(\cdot),u}}<\infty\right\},$$7 in Herz–Morrey spaces, one obtains Herz spaces. If all exponents and weights are constant, the variable exponent Triebel–Lizorkin–Morrey and grand Herz–Morrey scales collapse to their classical counterparts (Gurbuz, 5 Jul 2026).

The complementary theory reveals a precise link with weighted Lebesgue spaces. In the constant exponent power-weight case,

$$\mathcal{M}_{p(\cdot),u}(\mathbb{R}^n) := \left\{f\in L^{p(\cdot)}_{\mathrm{loc}}(\mathbb{R}^n): \|f\|_{\mathcal{M}_{p(\cdot),u}}<\infty\right\},$$8

and this identifies complementary Morrey control as a quantified description of point singularity at $$\mathcal{M}_{p(\cdot),u}(\mathbb{R}^n) := \left\{f\in L^{p(\cdot)}_{\mathrm{loc}}(\mathbb{R}^n): \|f\|_{\mathcal{M}_{p(\cdot),u}}<\infty\right\},$$9 (Guliyev et al., 2011).

A recurrent caution is that the definition itself is not completely standardized: the notion of Morrey space with variable exponents might differ from work to work (Caetano et al., 2018). This is not merely terminological. Some frameworks place the exponent at the center of the ball, others use a modular with $w(x,r)$00 inside the integral, others employ general Morrey control functions, and still others use global or complementary norms. A plausible implication is that boundedness theorems are best interpreted within the precise definition adopted by each paper.

One explicit open problem is the Stein–Weiss inequality on variable exponent Morrey spaces over all of $w(x,r)$01. The bounded-domain theory is established, but the proof “does not seem to extend to unbounded domains,” and proving a Stein–Weiss type inequality on variable exponent Morrey spaces defined on all of $w(x,r)$02 remains open (Cruz-Uribe et al., 2 Oct 2025). Other papers point to possible extensions to Calderón–Zygmund operators, commutators, nonlocal operators, grand Herz–Morrey–Besov spaces, and sharper endpoint theories (Gurbuz, 2 Apr 2025).

Variable exponent Morrey spaces therefore form a heterogeneous but coherent field: the common core is a variable exponent $w(x,r)$03 structure localized by Morrey scaling, while the main research directions concern admissible formulations, endpoint control, interpolation, weighted inequalities, and the transfer of harmonic-analysis methods to increasingly refined Morrey-type scales.