Hardy–Sobolev–Maz'ya Inequalities

• Hardy–Sobolev–Maz'ya inequalities are functional inequalities that quantify integrability gains after subtracting sharp Hardy singular terms, applicable in various geometric settings.
• They compare an energy term with a singular potential, often defined by distance functions, to a critical Sobolev L^p norm while determining best constants and extremals.
• Recent work extends these inequalities to higher order, fractional, and trace formats using tools like Fourier analysis, concentration–compactness, and ground-state transforms to ensure stability.

Searching arXiv for recent and foundational papers on Hardy–Sobolev–Maz'ya inequalities. Hardy–Sobolev–Maz’ya inequalities are functional inequalities that quantify how a Sobolev gain in integrability survives after subtraction of a sharp Hardy singular term. In their basic form, they compare an energy with a singular potential—typically involving the distance to a boundary or to a lower-dimensional singular set—to a critical or subcritical $L^p$ norm. In the modern literature, they are studied in Euclidean domains, half-spaces, hyperbolic spaces, Heisenberg groups, and fractional or higher-order settings, with sharp constants, attainability, symmetry, stability, and associated Euler–Lagrange equations forming the principal themes (Pinchover et al., 2010).

1. Definition and canonical forms

The classical second-order framework starts from the competition between Hardy and Sobolev inequalities. For a domain $\Omega \subset \mathbb{R}^n$, $n \ge 3$, and $u \in C_c^\infty(\Omega)$, a general Hardy–Sobolev–Maz’ya inequality has the form

$$\int_{\Omega} |\nabla u|^2\,dx \;-\; \lambda \int_{\Omega} \frac{|u(x)|^2}{d(x)^2}\,dx \;\ge\; C_{\mathrm{HSM}(\Omega,\lambda)}\, \|u\|_{L^{2^*}(\Omega)}^2,$$

where $d(x)=\operatorname{dist}(x,\partial\Omega)$, $2^* = 2n/(n-2)$, and $0<\lambda\le \Lambda_{\mathrm{Hardy}(\Omega)}$ (Pinchover et al., 2010). The point is that the sharp Hardy term is removed from the energy, yet the remaining quadratic form still controls the critical Sobolev norm.

A model codimension-$m$ version due to Maz’ya replaces boundary distance by distance to a subspace. Writing $\mathbb{R}^N=\mathbb{R}^{N-m}\times\mathbb{R}^m$ with coordinates $\Omega \subset \mathbb{R}^n$0, Maz’ya proved

$\Omega \subset \mathbb{R}^n$1

for $\Omega \subset \mathbb{R}^n$2, $\Omega \subset \mathbb{R}^n$3, $\Omega \subset \mathbb{R}^n$4 (Pinchover et al., 2010). This formulation already exhibits a characteristic feature of the subject: the singularity may lie on a boundary, at a point, or on a flat or curved lower-dimensional set.

In the high-order half-space setting, the $\Omega \subset \mathbb{R}^n$5-th order critical Hardy–Sobolev–Maz’ya inequality on

$\Omega \subset \mathbb{R}^n$6

takes the form

$\Omega \subset \mathbb{R}^n$7

with

$\Omega \subset \mathbb{R}^n$8

for $\Omega \subset \mathbb{R}^n$9 and $n \ge 3$0; in the critical existence theory of the 2026 paper, the relevant range is $n \ge 3$1 and $n \ge 3$2 (Lu et al., 5 Feb 2026). The same source emphasizes that the abstract’s displayed formula contains typographical errors, and that the corrected inequality is the one above, posed on $n \ge 3$3, not $n \ge 3$4 (Lu et al., 5 Feb 2026).

A general functional perspective treats

$n \ge 3$5

and studies when $n \ge 3$6 admits a weighted Hardy–Sobolev–Maz’ya inequality or, in the critical case, a Hardy–Sobolev–Maz’ya–Poincaré inequality involving an additional ground-state correction term (Pinchover et al., 2010). This formulation is central for understanding criticality, ground states, and the natural energy space.

2. Geometric settings and operator-theoretic reformulations

The subject is strongly shaped by geometry. In Euclidean half-spaces and convex domains, the singularity is usually expressed in terms of boundary distance. In codimension-$n \ge 3$7 problems it is the distance to a submanifold or axis. In hyperbolic and sub-Riemannian settings, the Hardy term may emerge through conformal or group-theoretic structure rather than as an explicit Euclidean inverse-square weight.

A central modern observation is that the high-order half-space inequality is equivalent to a Poincaré–Sobolev inequality on hyperbolic space. In the Poincaré ball model

$n \ge 3$8

one considers the GJMS operator

$n \ge 3$9

and the inequality

$u \in C_c^\infty(\Omega)$0

for $u \in C_c^\infty(\Omega)$1, $u \in C_c^\infty(\Omega)$2 (Lu et al., 5 Feb 2026). The equivalence rests on conformal covariance of $u \in C_c^\infty(\Omega)$3 and conformal maps between the ball and the half-space; under this correspondence, the hyperbolic “Poincaré shift” $u \in C_c^\infty(\Omega)$4 becomes the boundary Hardy singular potential $u \in C_c^\infty(\Omega)$5 (Lu et al., 5 Feb 2026).

This hyperbolic rephrasing has antecedents in the higher-order theory on half-spaces, where Fourier analysis on hyperbolic and symmetric spaces was used to derive Hardy–Sobolev–Maz’ya inequalities for higher derivatives and to relate half-space inequalities to hyperbolic Poincaré–Sobolev and Hardy–Littlewood–Sobolev inequalities (Lu et al., 2017). It also underlies sharp results for the critical order $u \in C_c^\infty(\Omega)$6 in odd dimensions through Green’s functions of Paneitz and GJMS operators on hyperbolic spaces (Lu et al., 2019).

Beyond real hyperbolic space, analogous inequalities have been constructed on complex, quaternionic, and Cayley hyperbolic spaces using factorization theorems, Helgason–Fourier analysis, and spectral-gap subtraction (Lu et al., 2021, Flynn et al., 2021). In these settings the Hardy term is encoded by the bottom of the spectrum of the relevant invariant Laplacian rather than by an explicit Euclidean distance function. A comparable phenomenon appears for fractional orders on hyperbolic space: the fractional GJMS operators $u \in C_c^\infty(\Omega)$7 and the auxiliary operators $u \in C_c^\infty(\Omega)$8 give rise to fractional Poincaré–Sobolev and fractional Hardy–Sobolev–Maz’ya inequalities on $u \in C_c^\infty(\Omega)$9, $$\int_{\Omega} |\nabla u|^2\,dx \;-\; \lambda \int_{\Omega} \frac{|u(x)|^2}{d(x)^2}\,dx \;\ge\; C_{\mathrm{HSM}(\Omega,\lambda)}\, \|u\|_{L^{2^*}(\Omega)}^2,$$0, and $$\int_{\Omega} |\nabla u|^2\,dx \;-\; \lambda \int_{\Omega} \frac{|u(x)|^2}{d(x)^2}\,dx \;\ge\; C_{\mathrm{HSM}(\Omega,\lambda)}\, \|u\|_{L^{2^*}(\Omega)}^2,$$1 (Lu et al., 2023).

3. Extremals, best constants, and symmetry

A fundamental question is whether the best constant in a Hardy–Sobolev–Maz’ya inequality is attained. The answer depends sensitively on dimension, order, geometry, and symmetry class.

In the first-order half-space setting, the best constant equals the Sobolev constant in dimension $$\int_{\Omega} |\nabla u|^2\,dx \;-\; \lambda \int_{\Omega} \frac{|u(x)|^2}{d(x)^2}\,dx \;\ge\; C_{\mathrm{HSM}(\Omega,\lambda)}\, \|u\|_{L^{2^*}(\Omega)}^2,$$2, with non-attainment in the natural space; for $$\int_{\Omega} |\nabla u|^2\,dx \;-\; \lambda \int_{\Omega} \frac{|u(x)|^2}{d(x)^2}\,dx \;\ge\; C_{\mathrm{HSM}(\Omega,\lambda)}\, \|u\|_{L^{2^*}(\Omega)}^2,$$3, the best Hardy–Sobolev–Maz’ya constant is strictly smaller than the Sobolev constant, and extremals exist (Pinchover et al., 2010). This first-order pattern has now been extended to higher order. For the $$\int_{\Omega} |\nabla u|^2\,dx \;-\; \lambda \int_{\Omega} \frac{|u(x)|^2}{d(x)^2}\,dx \;\ge\; C_{\mathrm{HSM}(\Omega,\lambda)}\, \|u\|_{L^{2^*}(\Omega)}^2,$$4-th order critical inequality on $$\int_{\Omega} |\nabla u|^2\,dx \;-\; \lambda \int_{\Omega} \frac{|u(x)|^2}{d(x)^2}\,dx \;\ge\; C_{\mathrm{HSM}(\Omega,\lambda)}\, \|u\|_{L^{2^*}(\Omega)}^2,$$5, extremals exist for $$\int_{\Omega} |\nabla u|^2\,dx \;-\; \lambda \int_{\Omega} \frac{|u(x)|^2}{d(x)^2}\,dx \;\ge\; C_{\mathrm{HSM}(\Omega,\lambda)}\, \|u\|_{L^{2^*}(\Omega)}^2,$$6 and $$\int_{\Omega} |\nabla u|^2\,dx \;-\; \lambda \int_{\Omega} \frac{|u(x)|^2}{d(x)^2}\,dx \;\ge\; C_{\mathrm{HSM}(\Omega,\lambda)}\, \|u\|_{L^{2^*}(\Omega)}^2,$$7, while for $$\int_{\Omega} |\nabla u|^2\,dx \;-\; \lambda \int_{\Omega} \frac{|u(x)|^2}{d(x)^2}\,dx \;\ge\; C_{\mathrm{HSM}(\Omega,\lambda)}\, \|u\|_{L^{2^*}(\Omega)}^2,$$8 the best constant equals the Euclidean Sobolev constant $$\int_{\Omega} |\nabla u|^2\,dx \;-\; \lambda \int_{\Omega} \frac{|u(x)|^2}{d(x)^2}\,dx \;\ge\; C_{\mathrm{HSM}(\Omega,\lambda)}\, \|u\|_{L^{2^*}(\Omega)}^2,$$9 and is not attained (Lu et al., 5 Feb 2026).

The 2026 existence result proves more than attainability. In hyperbolic space, extremals are positive, radially symmetric, and monotone decreasing about some point $d(x)=\operatorname{dist}(x,\partial\Omega)$0. On the half-space, existence follows by equivalence, and positivity carries over, but Euclidean radial symmetry is not asserted because standard rearrangement arguments are not available on $d(x)=\operatorname{dist}(x,\partial\Omega)$1 (Lu et al., 5 Feb 2026). The corresponding Euler–Lagrange equation on hyperbolic space is

$d(x)=\operatorname{dist}(x,\partial\Omega)$2

while the formal half-space Euler–Lagrange equation is

$d(x)=\operatorname{dist}(x,\partial\Omega)$3

with homogeneous Dirichlet boundary traces up to order $d(x)=\operatorname{dist}(x,\partial\Omega)$4 inherited from the closure of $d(x)=\operatorname{dist}(x,\partial\Omega)$5 (Lu et al., 5 Feb 2026).

There are also special regimes in which the sharp Hardy–Sobolev–Maz’ya constant coincides with the sharp Sobolev constant. For odd $d(x)=\operatorname{dist}(x,\partial\Omega)$6, Lu and Yang proved that the sharp constant in the $d(x)=\operatorname{dist}(x,\partial\Omega)$7-th order half-space inequality equals the best $d(x)=\operatorname{dist}(x,\partial\Omega)$8-th order Sobolev constant (Lu et al., 2019). Earlier, the $d(x)=\operatorname{dist}(x,\partial\Omega)$9 case for $2^* = 2n/(n-2)$0 had been identified in work of Benguria, Frank, and Loss, and the $2^* = 2n/(n-2)$1, $2^* = 2n/(n-2)$2 case was shown to have the same coincidence for the bi-Laplacian (Lu et al., 2017). By contrast, the 2026 existence theorem shows that in dimensions $2^* = 2n/(n-2)$3 the high-order best constant is strictly smaller than $2^* = 2n/(n-2)$4 and is attained (Lu et al., 5 Feb 2026).

In a different Euclidean geometry, a 2024 classification result identifies extremals for the weighted $2^* = 2n/(n-2)$5-Laplace Hardy–Sobolev–Maz’ya inequality with cylindrical singularity $2^* = 2n/(n-2)$6 for $2^* = 2n/(n-2)$7. The classified positive finite-energy cylindrically symmetric solutions generate the extremal manifold and determine the best constant (Lin et al., 2024). This complements the 2025 stability theory, which takes those classified extremals as the reference manifold for quantitative estimates (Dai et al., 31 Aug 2025).

4. Analytical mechanisms: concentration, rearrangement, Fourier analysis, and ground-state transforms

The main technical obstacle in critical Hardy–Sobolev–Maz’ya theory is lack of compactness. In the high-order half-space problem this is compounded by higher derivatives, the boundary singularity $2^* = 2n/(n-2)$8, lack of translation invariance, and the failure of direct rearrangement on $2^* = 2n/(n-2)$9 (Lu et al., 5 Feb 2026).

The 2026 solution combines several mechanisms. First, it introduces the shifted operator

$0<\lambda\le \Lambda_{\mathrm{Hardy}(\Omega)}$0

and uses the Helgason–Fourier transform to write

$0<\lambda\le \Lambda_{\mathrm{Hardy}(\Omega)}$1

This turns the differential inequality into an $0<\lambda\le \Lambda_{\mathrm{Hardy}(\Omega)}$2-form and, by duality, into a Hardy–Littlewood–Sobolev-type integral inequality involving the kernel $0<\lambda\le \Lambda_{\mathrm{Hardy}(\Omega)}$3 (Lu et al., 5 Feb 2026). The authors then prove an explicit correspondence between extremizing sequences for the differential and integral formulations.

Second, a Lions-type concentration–compactness principle is developed for radially decreasing sequences on hyperbolic space. Vanishing is excluded using radial decay estimates, while dichotomy is ruled out in the dual integral formulation via kernel monotonicity and HLS estimates (Lu et al., 5 Feb 2026). Third, Beckner’s hyperbolic rearrangement inequality is invoked once positivity and monotonicity of the Green kernel $0<\lambda\le \Lambda_{\mathrm{Hardy}(\Omega)}$4 with respect to hyperbolic distance have been established. This justifies restricting to radial decreasing minimizing sequences, which restores compactness (Lu et al., 5 Feb 2026).

Related methods appear in other branches of the theory. In fractional half-space problems, Frank–Seiringer ground-state representations isolate the sharp Hardy term and rewrite the remainder as a weighted fractional seminorm of a transformed function (Dyda et al., 2017, Sloane, 2010). In the cylindrical fractional setting, the ground-state transform

$0<\lambda\le \Lambda_{\mathrm{Hardy}(\Omega)}$5

converts the fractional Hardy–Sobolev–Maz’ya functional into a weighted kernel form and yields a sharp Hardy constant together with attainment of the best weighted HSM constant (Mallick, 2018). In the weighted fractional setting with singularity on a flat submanifold of codimension $0<\lambda\le \Lambda_{\mathrm{Hardy}(\Omega)}$6, sharp Hardy constants and remainder terms are obtained through nonlinear ground-state representations; these then feed into weighted fractional Hardy–Sobolev and Hardy–Sobolev–Maz’ya inequalities (Kijaczko et al., 24 Mar 2025).

In arbitrary Euclidean domains, another mechanism is directional averaging. Frank and Loss established a Hardy–Sobolev–Maz’ya inequality with a constant depending only on the dimension by replacing boundary distance with Davies’ directional pseudodistance $0<\lambda\le \Lambda_{\mathrm{Hardy}(\Omega)}$7, thereby resolving a conjecture of Filippas, Maz’ya, and Tertikas for convex domains (Frank et al., 2011). In the fractional domain setting, Dyda and Frank proved a nonlocal analogue for arbitrary domains using the Loss–Sloane pseudodistance $0<\lambda\le \Lambda_{\mathrm{Hardy}(\Omega)}$8, preserving the sharp Hardy constant in the regional fractional Hardy inequality (Dyda et al., 2011).

5. Variants: fractional, trace, sub-Riemannian, weighted, and logarithmic forms

Hardy–Sobolev–Maz’ya inequalities admit a large family of nonlocal and weighted variants. For the upper half-space $0<\lambda\le \Lambda_{\mathrm{Hardy}(\Omega)}$9, Dyda, Lehrbäck, and Vähäkangas established a fractional Hardy–Sobolev–Maz’ya inequality

$m$0

where $m$1, $m$2, and $m$3 (Dyda et al., 2017). The constant $m$4 is the sharp fractional Hardy constant on the half-space, and the result answers an open problem of Musina and Nazarov, including the previously open range $m$5 (Dyda et al., 2017).

Another fractional line of work studies singularities on lower-dimensional sets. Mallick proved a fractional Hardy–Sobolev–Maz’ya inequality on $m$6 with cylindrical singular weight $m$7, existence of extremals both below and at the critical Hardy strength, cylindrical symmetry of solutions, and explicit asymptotic behavior near the singular set and at infinity (Mallick, 2018). More recently, weighted fractional HSM inequalities with singularities on flat submanifolds of codimension $m$8 were derived, including sharp weighted Hardy constants, remainder terms, and a weighted logarithmic fractional HSM inequality in the origin case $m$9; in that case the non-logarithmic HSM inequality fails (Kijaczko et al., 24 Mar 2025).

Trace inequalities form another major branch. For weakly mean convex or convex domains, Filippas, Moschini, and Tertikas established sharp trace Hardy and trace Hardy–Sobolev–Maz’ya inequalities for the Caffarelli–Silvestre extension, then used them to derive fractional Hardy–Sobolev–Maz’ya inequalities with best Hardy constants for spectral and restricted fractional Laplacians (Filippas et al., 2011). Their half-space results cover the full range $\mathbb{R}^N=\mathbb{R}^{N-m}\times\mathbb{R}^m$0, thereby resolving the Frank–Seiringer open problem mentioned in that work (Filippas et al., 2011). At the spectral half-Laplacian $\mathbb{R}^N=\mathbb{R}^{N-m}\times\mathbb{R}^m$1, a separate trace theory on weakly mean convex domains identifies the sharp Hardy constant $\mathbb{R}^N=\mathbb{R}^{N-m}\times\mathbb{R}^m$2 and proves a trace Hardy–Sobolev–Maz’ya inequality with critical trace exponent $\mathbb{R}^N=\mathbb{R}^{N-m}\times\mathbb{R}^m$3 (Filippas et al., 2014). Polyhedral convex cones admit analogous trace Hardy and trace HSM inequalities with explicit best constants, together with logarithmic trace versions and radial sharp constants (Nguyen, 2016).

The theory also extends to non-Euclidean structures. On the Heisenberg group, the Hardy term becomes

$\mathbb{R}^N=\mathbb{R}^{N-m}\times\mathbb{R}^m$4

where $\mathbb{R}^N=\mathbb{R}^{N-m}\times\mathbb{R}^m$5 is the angle function measuring the horizontal projection of the boundary normal. For $\mathbb{R}^N=\mathbb{R}^{N-m}\times\mathbb{R}^m$6, this yields a Hardy–Sobolev–Maz’ya inequality on Heisenberg half-spaces, with the sharp Hardy constant $\mathbb{R}^N=\mathbb{R}^{N-m}\times\mathbb{R}^m$7 already valid for all $\mathbb{R}^N=\mathbb{R}^{N-m}\times\mathbb{R}^m$8 (Ruzhansky et al., 2018). Weighted orthant analogues with monomial weights and explicit Hardy, Sobolev, Maz’ya, and remainder terms have also been developed (Kömbe et al., 2021).

6. Stability, natural energy spaces, and associated equations

The variational structure of Hardy–Sobolev–Maz’ya inequalities leads naturally to energy spaces and nonlinear equations. For $\mathbb{R}^N=\mathbb{R}^{N-m}\times\mathbb{R}^m$9, one introduces the norm

$\Omega \subset \mathbb{R}^n$00

and completes $\Omega \subset \mathbb{R}^n$01 to obtain the natural energy space $\Omega \subset \mathbb{R}^n$02 whenever the quadratic form is subcritical (Pinchover et al., 2010). In the critical case, a Hardy–Sobolev–Maz’ya–Poincaré correction involving the ground state $\Omega \subset \mathbb{R}^n$03 is needed to define a coercive norm (Pinchover et al., 2010). This framework also extends, in a partial form, to $\Omega \subset \mathbb{R}^n$04 through ground-state transforms and convexified Lagrangians on cones of nonnegative functions (Pinchover et al., 2010).

The Euler–Lagrange equations associated with extremals are critical elliptic equations with singular potentials. For the hyperbolic high-order problem, the equation

$\Omega \subset \mathbb{R}^n$05

is a Brezis–Nirenberg type equation for the GJMS operator $\Omega \subset \mathbb{R}^n$06. The existence of positive radial solutions in both the critical and subcritical regimes follows directly from the existence of extremals for the corresponding Poincaré–Sobolev inequalities (Lu et al., 5 Feb 2026). In the cylindrical $\Omega \subset \mathbb{R}^n$07-Laplace setting, the equation

$\Omega \subset \mathbb{R}^n$08

governs extremals, and the classified family of solutions provides the extremal manifold (Dai et al., 31 Aug 2025).

A recent development is quantitative stability. For the inequality

$\Omega \subset \mathbb{R}^n$09

with $\Omega \subset \mathbb{R}^n$10, the 2025 paper proves non-degeneracy of the Euler–Lagrange equation and a sharp global stability estimate

$\Omega \subset \mathbb{R}^n$11

where $\Omega \subset \mathbb{R}^n$12 is the extremal manifold (Dai et al., 31 Aug 2025). The analysis requires a compact embedding adapted to the partial strong singularity $\Omega \subset \mathbb{R}^n$13, a spectral gap above the tangent space of $\Omega \subset \mathbb{R}^n$14, and refined spectral inequalities in both Hilbert and non-Hilbert regimes (Dai et al., 31 Aug 2025). A plausible implication is that quantitative stability for non-radial extremals is now becoming part of the Hardy–Sobolev–Maz’ya toolkit, not merely an add-on to existence theory.

The broader picture is that Hardy–Sobolev–Maz’ya inequalities now form a unified interface between sharp functional inequalities, criticality theory, singular elliptic operators, and geometric analysis. Their modern developments cover arbitrary domains, half-spaces, hyperbolic and sub-Riemannian geometries, higher-order and fractional operators, sharp constants, extremals, and stability, while continuing to expose delicate threshold phenomena such as non-attainment, logarithmic corrections, and the dependence of sharp behavior on codimension and curvature (Pinchover et al., 2010).