A note on the embedding theorems for Sobolev-Lorentz spaces (Q2671544)

The authors prove embedding results for Sobolev spaces in which the role of \(L^p\) is taken by a Lorentz space \(L^{p,q}\). The technique uses an integral formula of Muramatu. The classical results being generalized are first if \(p < d\), \(1 \leq q \leq \infty\), one has \[ \| f \|_{L^{ \frac{dp}{d -p}} } \leq C_{p,d} \| \nabla f\|_{L^{p}(\mathbb R^d)}, \] and second if \(d < p < \infty\), \(1 \leq r < \infty\), \(f \in C^{\infty}_c(\mathbb R^d)\), \[ \| f \|_{L^{\infty}(\mathbb R^d) } \leq C_{p,r,d} \| f\|^{ \frac{ 1 - \frac{d}{p} }{ 1 - \frac{d}{p} + \frac{d}{r} }} _{L^p(\mathbb R^d)} \| \nabla f \|^{ \frac{ \frac{d}{p} }{ 1 - \frac{d}{p} + \frac{d}{r} }}_{L^p(\mathbb R^d)}. \] If \(r = p\), this becomes \[ \| f \|_{L^{\infty}(\mathbb R^d) } \leq C_{p,d} \| f\|^{1 - \frac{d}{p} } _{L^p(\mathbb R^d)} \| \nabla f \|^{ \frac{d}{p} }_{L^p(\mathbb R^d)}, \] and is extended to \(W^{1.p}(\mathbb R^d)\) by density of \(C^{\infty}_c(\mathbb R^d)\) in the Sobolev space \(W^{1,p}(\mathbb R^d)\). It implies Morrey's inequality \[ \| f \|_{L^{\infty}(\mathbb R^d) } \leq C_{p,d} |{\operatorname{supp} f}|^{ \frac1{d} - \frac1{p} } \| \nabla f \|^{ \frac{d}{p} }_{L^p(\mathbb R^d)} , \quad f \in C^{\infty}_c(\mathbb R^d). \] For \( p < d\), \(1 \leq q \leq \infty\), the Lorentz space results are \[ \| f \|_ {L^{ \frac{dp}{d -p}, q } } \leq C_{p,q,d} \| \nabla f\|_{L^{p,q}(\mathbb R^d)}, \] and if \( 2 \leq d < p\), one first obtains the results with weak type \(p\), \[ \| f \|_{L^{\infty}(\mathbb R^d) } \leq C_{p,r,d}\| f\|^{ \frac{ 1 - \frac{d}{p} }{ 1 - \frac{d}{p} + \frac{d}{r} }} _{L^{p, \infty}(\mathbb R^d)} \| \nabla f \|^{ \frac{ \frac{d}{p} }{ 1 - \frac{d}{p} + \frac{d}{r} }}_{L^{p, \infty}(\mathbb R^d)} \] which can be improved to, with \(1 \leq q\), \(q_1, q_2 \leq \infty\), \[ \| f \|_{L^{\infty}(\mathbb R^d) } \leq C_{p,r,q_1, q_2,d} \| f\|^{ \frac{ 1 - \frac{d}{p} }{ 1 - \frac{d}{p} + \frac{d}{r} }} _{L^{p, q_1}(\mathbb R^d)} \| \nabla f \|^{ \frac{ \frac{d}{p} }{ 1 - \frac{d}{p} + \frac{d}{r} }}_{L^{p,q_2}(\mathbb R^d)} , \] \[ \| f \|_{L^{\infty}(\mathbb R^d) } \leq C_{p,q,d} |{\operatorname {supp} f}|^{ \frac1{d} - \frac1{p} } \| \nabla f \|^{ \frac{d}{p} }_{L^{p,q}(\mathbb R^d)} , \quad f \in C^{\infty}_c(\mathbb R^d). \] The authors provide new proofs using Muramatu's integral representation formula for \(1 < p < d\), and the recursive technique of \textit{A.~Porretta} [Adv. Nonlinear Stud. 20, No.~2, 361--371 (2020; Zbl 1439.35015)] which uses the result for \(p = 1< d \) and iteration to prove the case \(p > d\). The result is known for \(p = 1\), but does not follow from their current proof.