Sobolev–Lorentz Embedding Theorems

• Sobolev–Lorentz embedding is a refined theorem that characterizes the optimal inclusion of Sobolev spaces into Lorentz and rearrangement-invariant spaces with precise constants.
• It quantitatively describes non-compactness—particularly at critical scaling—using notions like Bernstein numbers and mass concentration near the cone vertex.
• Extensions of these embeddings include weighted settings, vector differential operators, and non-Euclidean geometries, linking Hardy inequalities with spectral theory.

Sobolev-Lorentz embedding refers to the class of optimal and refined embedding theorems for Sobolev spaces where the target space is a Lorentz or allied rearrangement-invariant function space. These theorems provide a quantitative and often sharp characterization of how smoothness quantities (measured by gradients or derivatives in Sobolev–type norms) control function space membership at critical scaling—often specifying norm constants, optimality, non-compactness, and extremal profiles. Considerations of weights, cone domains, group structures, non-Euclidean geometries, and various generalizations (Zygmund, Morrey, Besov, Triebel–Lizorkin) are central.

1. Foundational Definitions and Weighted Setting

Let $n\ge2$ and $\mathcal C\subset\mathbb{R}^{n}$ be an open convex cone with vertex at the origin. For $\alpha>0$ and $D=n+\alpha$, define the homogeneous weight $w_{\alpha}(x) = |x|^{\alpha}$. The associated weighted measure is $d\mu(x) = w_{\alpha}(x)dx$ on $\mathcal C$.

Weighted Lorentz spaces $L^{p,q}(\mathcal C, w_\alpha)$ consist of measurable $f:\mathcal C\to\mathbb{R}$ with finite quasi-norm

$$\|f\|_{L^{p,q}(\mathcal C,w_\alpha)} = \left(\int_0^\infty [t^{1/p}f^{*}(t)]^q\,\frac{dt}{t}\right)^{1/q},$$

where $f^*(t)$ is the nonincreasing rearrangement of $|f|$ w.r.t. $\mu$. For $q=p$ this recovers the weighted Lebesgue norm.

The corresponding weighted Sobolev-Lorentz space is

$$W^{1,(p,q)}(\mathcal C, w_\alpha) = \overline{C^\infty_c(\mathbb{R}^n)}^{\|\cdot\|_{W^{1,(p,q)}}},$$

equipped with

$$\|u\|_{W^{1,(p,q)}} = \|\nabla u\|_{L^{p,q}(\mathcal C, w_\alpha)}.$$

The “Sobolev critical” exponent is

$p^* = \frac{Dp}{D - p}.$

The optimal embedding theorem states for $1 \le q < p < D$ and $u \in C_c^\infty(\mathbb{R}^n)$,

$$\|u\|_{L^{p^*,q}(\mathcal C,w_\alpha)} \leq C_{\rm opt} \| \nabla u \|_{L^{p,q}(\mathcal C,w_\alpha)},$$

with the sharp constant

$$C_{\rm opt} = (p^*)^{1/q} D^{-1} \mu(B_1 \cap \mathcal C)^{1/D}.$$

No smaller constant is possible (Gurka et al., 2023).

2. Quantitative Non-Compactness and Singular Structures

When $q \leq r$, the embedding

$$W^{1,(p,q)}(\mathcal C,w_\alpha) \hookrightarrow L^{p^*,r}(\mathcal C,w_\alpha)$$

is continuous but never compact—the property fails especially at critical scaling. Quantitative descriptions use:

• The measure of non-compactness:

$$\beta(T) = \inf \left\{ r > 0 : T(B_X) \subset \bigcup_{i=1}^N B_Y(y_i, r), \text{ for some } N < \infty \right\}$$

for operator $T$.

• Bernstein (strict $s$-) numbers:

$$b_n(T) = \sup_{E \subset X,\,{\rm dim}\,E = n} \inf_{x\in S_E} \|Tx\|_Y,$$

where $S_E$ is the unit sphere of $E$.

For the embedding operator $E: W^{1,(p,q)} \to L^{p^*,q}$,

$$\beta(E) = \|E\|, \quad b_n(E) = \|E\| \;\; \forall n \in \mathbb{N}$$

i.e., the embedding is maximally non-compact. In contrast to translation-invariant scenarios, the non-compactness arises purely from mass concentration near the cone vertex and cannot be addressed by simple shifts (Gurka et al., 2023, Chuah et al., 7 Feb 2025).

3. Extremal Functions and Hardy-Type Optimality

Functions which saturate the embedding are radially symmetric, nonincreasing rearrangements of Euler–Lagrange extremals for associated Hardy inequalities:

$$u(x) = \left( |x|^{D-p} - R^{D-p} \right)^{-1/p} \cdot \mathbf{1}_{ \{ r < |x| < R \} }.$$

As $r\to0$, $R\to\infty$, this recovers the profile $u(x) \sim |x|^{-\frac{D-p}{p}}$.

Key properties:

• Homogeneous of degree $-(D-p)/p$ outside the origin.
• Support shrinks toward the cone vertex in the extremal limit, illustrating localization-driven non-compactness.
• These functions realize equality in the one-dimensional Hardy inequality after using the cone’s Pólya–Szegő principle (Gurka et al., 2023).

4. Extensions: Vector Differential Operators and Other Settings

Lorentz-refined Sobolev inequalities extend to general vector-valued operators $A(D)$ satisfying ellipticity and “canceling” conditions. Spector–Van Schaftingen proved for such operators $A(D)$, the embedding

$$\|u\|_{L^{\frac{n}{n-1},1}(\mathbb{R}^n; V)} \leq C \|A(D)u\|_{L^1(\mathbb{R}^n; E)},$$

holds if $A(D)$ is elliptic and $(n-1)$-canceling. For $A(D) = \nabla$, this recovers Alvino's sharp Sobolev–Lorentz inequality; for $A(D) = \operatorname{div}$, $\operatorname{curl}$, etc., similar endpoint Lorentz embeddings follow (Spector et al., 2018).

In non-Euclidean and geometric contexts, optimal Sobolev–Lorentz embeddings persist under suitable isoperimetric and Pólya–Szegő hypotheses on Cartan–Hadamard manifolds, with the sharp constant:

$$\|u\|_{L^{2^*, 2}(M)} \leq S_{N, 2^*, 2} \| \nabla u \|_{L^2(M)}, \quad S_{N, 2^*, 2} = \frac{2}{N - 2} (\Gamma(1 + N/2))^{1/N} \pi^{-1/2}$$

(Banerjee et al., 20 Jan 2026). No smaller rearrangement-invariant space than $L^{2^*,2}$ admits a bounded embedding in this regime.

5. Generalizations: Lorentz-Zygmund, Besov, Morrey, and Campanato Spaces

Sobolev–type embeddings extend beyond classical Lorentz targets to generalized Lorentz-Zygmund (GLZ) spaces $L^{p,q;\alpha,\beta}$:

$$\|u\|_{L^{p,q;\alpha,\beta}(\Omega)} = \left\| s^{1/p - 1/q} \ell(s)^\alpha \ell\ell(s)^\beta u^*(s) \right\|_{L^q(0,1)},$$

where $\ell(s) = 1 + |\ln s|$, $\ell\ell(s) = 1 + \ln(1 + |\ln s|)$.

Optimal r.i. targets for Sobolev-type spaces $W^mL^{p,q;\alpha,\beta}$ on a John domain are specified:

• For $1 < p < n/m$, $Y = L^{p^*, q; \alpha, \beta}$, $p^* = np/(n-mp)$.
• At critical scaling ($p = n/m$), targets may be Zygmund-type or $L^\infty$.

Similarly, precise criteria for Hölder, Morrey, and Campanato embeddings are delineated: $$W^mL^{p,q;\alpha,\beta}(\Omega) \hookrightarrow C^{0,\sigma(\cdot)}(\Omega)$$ holds with explicit modulus $\sigma(\cdot)$, which may include logarithmic corrections in the critical regime (Cavaliere et al., 20 Aug 2025).

6. Strict Singularity, Operator Theory, and Spectral Consequences

A mapping $T$ is strictly singular if it is not an isomorphism on any infinite-dimensional subspace. In the non-translation-invariant, weighted cone scenario, one can produce infinite orthogonal sequences of extremal functions such that the embedding

$$W^{1,(p,q)}(\mathcal C,w_\alpha) \to L^{p^*,q}(\mathcal C,w_\alpha)$$

is not strictly singular (Gurka et al., 2023). Operator-theoretically, all injective s- and Bernstein numbers attain the maximum $\|E\|$—no compactness improvement or finite-rank approximability occurs. Spectrally, this manifests as a nontrivial essential spectrum for the corresponding weighted $p$-Laplacian.

7. Optimality, Failure of Compactness, and Further Directions

All sharp Sobolev–Lorentz embeddings in weighted, unweighted, and generalized function space settings are optimal in their respective rearrangement-invariant scales. Compactness fails precisely at the Lorentz endpoint ($q=p^*$, or at critical logarithmic cases). As soon as target spaces are strictly larger, one quantitatively gains strict singularity (Bernstein numbers decay polynomially), and these exponents are sharp (Chuah et al., 7 Feb 2025).

The paradigm generalizes to group settings and fractional embedding—embedding theorems align with capacity/Hardy/perimeter inequalities and symmetrization techniques. This broad and deep theory, built on Lorentz spaces, rearrangement, Hardy and Pólya–Szegő principles, and operator-theoretic methods, remains central to modern analysis in PDE, functional analysis, geometric measure theory, and spectral theory.

---

Key References:

• Quantitative analysis for optimal weighted Sobolev–Lorentz embeddings (Gurka et al., 2023)
• Non-compactness and strict singularity (Bernstein numbers) for Sobolev–Lorentz embeddings (Chuah et al., 7 Feb 2025)
• Extension to vector differential operators (Spector et al., 2018)
• Sharp embeddings on Cartan–Hadamard manifolds (Banerjee et al., 20 Jan 2026)
• Generalizations to Lorentz-Zygmund and Besov–Morrey–Campanato spaces (Cavaliere et al., 20 Aug 2025)