On the Lavrentiev gap for manifold-valued maps

Carlo Alberto Antonini, Filomena De Filippis, Cintia Pacchiano Camacho · 2025

Read On the Lavrentiev gap for manifold-valued maps by Carlo Alberto Antonini, Filomena De Filippis, Cintia Pacchiano Camacho (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-18447.txt.

Introduction

Carlo Alberto Antonini, Filomena De Filippis, and Cintia Pacchiano Camacho study when smooth maps can approximate rougher ones between curved spaces. The question is the Lavrentiev gap, a setting where the best energy found with smooth trial maps can differ from the true minimum over less regular maps. Their arXiv abstract is brief: they investigate when modular density of smooth maps on compact manifolds holds and when it fails. The updated theorems also cover manifolds whose reference space has a boundary.

Catalog record

CREATORCarlo Alberto Antonini, Filomena De Filippis, Cintia Pacchiano Camacho
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE104 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2512-18447

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

On the Lavrentiev gap for manifold-valued maps Carlo Alberto Antonini, Filomena De Filippis, Cintia Pacchiano Camacho Abstract . We investigate the validity and the failure of modular density of smooth maps on compact manifolds. 2020 Mathematics Subject Classification 35B65, 46E35, 58D15 1. Introduction This work is devoted to establishing the density of smooth maps between Riemannian manifolds in nonhomogeneous spaces characterized by the finiteness of certain anisotropic energies. Let (1.1) \mathsf{M}\text{ be an $\textnormal{{m}}$-dimensional oriented compact manifold}, (1.2) \mathsf{N}\text{ be an $\textnormal{{n}}$-dimensional oriented compact manifold}, and consider a Sobolev space W^{1,\varphi} . The problem of approximating Sobolev maps between manifolds is classical, and its resolution depends critically on both the function space and the topology of the manifolds. For maps with values in \mathbb{R}^{N} and Young function \varphi(t)=t^{p} , the approximation argument is straightforward: standard convolution with a smooth kernel produces a sequence of smooth maps that converge strongly. When the target is a manifold \mathsf{N} , the situation is more delicate. Convolution now takes values in the convex hull of \mathsf{N} , so one must subsequently project these values back onto \mathsf{N} , for instance by using the nearest-point projection. This procedure works in the classical Sobolev setting \varphi(t)=t^{p} for the superdimensional case p\geq\textnormal{{m}} . Indeed, by Morrey-Sobolev embedding, maps in W^{1,p}(\mathsf{M},\mathsf{N}) are continuous when p>\textnormal{{m}} or belong to the space of functions with vanishing mean oscillation when p=\textnormal{{m}} (see [ 8 ] ), so projecting them back onto \mathsf{N} then yields a sequence in C^{\infty}(\mathsf{M},\mathsf{N}) converging strongly in W^{1,p}(\mathsf{M},\mathsf{N}) . The case p<\textnormal{{m}} is more subtle, and the possibility of strong approximation depends on the topology of both manifolds. Even for the unit sphere \textnormal{{S}}^{2} the problem is nontrivial. A classical example due to Schoen & Uhlenbeck [ 47 ] is the map u:\textnormal{{B}}^{3}\to\textnormal{{S}}^{2},\qquad u(x)=\frac{x}{|x|}. Then, u\in W^{1,p}(\textnormal{{B}}^{3},\textnormal{{S}}^{2}) , for 1\leq p<3 , but it cannot be strongly approximated in W^{1,p}(\textnormal{{B}}^{3},\textnormal{{S}}^{2}) by smooth maps with values in \textnormal{{S}}^{2} whenever 2\leq p<3 . Seminal work by Bethuel [ 7 ] and Hang & Lin [ 34 ] clarified the approximation problem in the subcritical regime. They showed that for maps in W^{1,p}(\mathsf{M},\mathsf{N}) , the space C^{\infty}(\mathsf{M},\mathsf{N}) is dense if and only if \mathsf{M} satisfies the ([p]-1) -extension property with respect to \mathsf{N} (see [ 34 , Definition 2.3] ) and the [p] -homotopy group of \mathsf{N} is trivial, where [p] denotes the integer part of p . Hajłasz [ 30 ] later extended these results to more general manifold domains showing that density holds provided \mathsf{N} is [{p}] -connected. The space W^{1,\textnormal{{m}}}(\mathsf{M},\mathsf{N}) is thus borderline if one wishes to approximate with smooth maps avoiding topological restrictions on the manifolds. However, this property persists in slightly larger Sobolev-type spaces, such as Orlicz spaces W^{1,A}(\mathsf{M},\mathsf{N}) built upon Young functions A satisfying suitable conditions depending on m . The key insight, originating in the work of Hajłasz, Iwaniec, Malý & Onninen [ 32 ] , is that for maps in Sobolev spaces slightly larger than W^{1,\textnormal{{m}}}(\mathsf{M},\mathsf{N}) , it is possible to detect certain sets on which a given map is still continuous. This leads to the property of vanishing web oscillations. The answer in the Orlicz setting was provided by Carozza & Cianchi [ 11 ] , who showed that if the Young function A satisfies the \Delta_{2} and \nabla_{2} conditions near infinity together with a sharp integral condition (which places W^{1,A} either slightly larger than W^{1,\textnormal{{m}}} or contained in W^{1,\textnormal{{m}}} ) then every map in W^{1,A}(\mathsf{M},\mathsf{N}) possesses vanishing web oscillations, establishing the density of smooth maps without any topological constraints on the manifolds. In the present paper, we go beyond the classical and Orlicz settings, and consider the more general Musielak-Orlicz spaces W^{1,\varphi}(\mathsf{M},\mathsf{N}) , where the function \varphi depends explicitly on the variable x\in\mathsf{M} as well. We show that, under suitable assumptions on \varphi and appropriate topological conditions on \mathsf{M} and \mathsf{N} , the space C^{\infty}(\mathsf{M},\mathsf{N}) is dense in the Musielak-Orlicz space W^{1,\varphi}(\mathsf{M},\mathsf{N}) . To our knowledge, this is the first systematic study of smooth approximation for manifold-valued maps in this kind of framework. We consider a map \varphi:\mathsf{M}\times[0,\infty)\to[0,\infty) such that: (1.3) x\mapsto\varphi(x,t)\quad\text{is measurable for all }t\in[0,\infty), and, for every x\in\mathsf{M} , (1.4) \varphi(x,t)=0\quad\text{if and only if}\quad t=0, (1.5) t\mapsto\varphi(x,t)\quad\text{is convex}, (1.6) \exists\beta,\gamma>1:\ t\mapsto\frac{\varphi(x,t)}{t^{\beta}}\text{ is almost increasing},\quad t\mapsto\frac{\varphi(x,t)}{t^{\gamma}}\text{ is almost decreasing }, (1.7) \exists\ \textnormal{{L}}\geq 1:\ \textnormal{{L}}^{-1}\leq\varphi(x,1)\leq\textnormal{{L}}. Moreover, we assume the following: for all \textnormal{{d}}>1 there exists a constant \textnormal{{c}}\equiv\textnormal{{c}}(\beta,\gamma,\textnormal{{d}}) , where \beta and \gamma are the exponents arising in ( 1.6 ), such that (1.8) \varphi(x_{1},t)\leq\textnormal{{c}}(\varphi(x_{2},t)+1),\text{ for all }t\in\left[0,\textnormal{{d}}\varrho^{-\min\left\{1,\frac{\textnormal{{m}}}{\beta}\right\}}\right], for all x_{1},x_{2}\in\overline{\mathbb{B}}_{\varrho} . Here, \mathbb{B}_{\varrho} is a geodesic ball of radius \varrho<1/2 and \overline{\mathbb{B}}_{\varrho} denotes its closure. These hypotheses are naturally satisfied by a wide class of functionals, most notably the double phase integrand (1.9) \varphi(x,t)=t^{p}+a(x)t^{q}, with 0\leq a(\cdot)\in C^{0,\alpha} , \alpha\in(0,1] and the exponents 1<p<q such that (1.10) q\leq p+\alpha\max\left\{1,\frac{p}{\textnormal{{m}}}\right\}, see [ 10 , Lemma 3.1] . A more general example satisfying our assumptions is provided by the variable-exponent integrand (1.11) \varphi(x,t)=t^{p(x)}+a(x)t^{q(x)}, where 1<p\leq p(x)\leq q(x)\leq q and p(\cdot),q(\cdot) are log-Hölder continuous and 0\leq a(\cdot)\in C^{0,\alpha} , \alpha\in(0,1] and ( 1.10 ) holds, see [ 10 , Lemma 3.3] . Double phase functionals were introduced by Zhikov [ 49 , 48 ] in the framework of homogenization and in relation to the Lavrentiev phenomenon. Moreover, this class of variational integrals represents a model for strongly anisotropic materials, with significant applications in elasticity theory. The interplay between the p - and q -growth regimes is dictated by the vanishing behavior of the coefficient a(\cdot) . Over the past decade, an extensive regularity theory has emerged for double phase and, more generally, for nonuniformly elliptic functionals. In this direction, Marcellini [ 40 , 43 , 41 , 42 ] investigated (p,q) -nonuniformly elliptic integrals, providing the first systematic results in this setting. Later contributions by Baroni, Colombo, and Mingione [ 5 , 14 , 13 ] provided a complete framework for regularity, which subsequently inspired a vast body of literature, see [ 1 , 29 , 37 , 38 , 44 , 45 , 46 , 16 , 21 , 22 , 39 , 23 ] and the references therein. Let us also mention that problems characterized by linear or nearly linear growth have been analyzed, among others, in [ 6 , 27 , 28 ] . Now, condition ( 1.10 ) is connected with the nonuniform ellipticity of the associated Euler-Lagrange system (1.12) -\mathrm{div}\,A(x,Du)=0, where A(x,z)=|z|^{p-2}z+(q/p)\,a(x)|z|^{q-2}z. As shown in [ 10 , Section 3] , for ( 1.9 ) and ( 1.11 ), it can be verified that assumption ( 1.8 ) holds if ( 1.10 ) is true. Observe that the (nonlocal) ellipticity ratio (1.13) \displaystyle\frac{\sup_{x\in\textnormal{{B}}}\text{highest eigenvalue of }\partial_{z}A(x,Du)}{\inf_{x\in\textnormal{{B}}}\text{lowest eigenvalue of }\partial_{z}A(x,Du)}\,\approx\,1+\|a\|_{L^{\infty}(\textnormal{{B}})}|Du|^{q-p} can become unbounded on any ball \textnormal{{B}}\subset\Omega intersecting the transition region \{a(\cdot)=0\} . Then, condition ( 1.10 ), is precisely designed to ensure that a(\cdot) decays sufficiently rapidly to prevent uncontrolled degeneracy, thereby guaranteeing maximal gradient regularity for local minimizers. The necessity of such a condition has been confirmed through explicit counterexamples in [ 2 , 4 , 25 , 26 , 3 ] showing that, when ( 1.10 ) is violated, the so-called Lavrentiev phenomenon can occur, i.e., the infimum of a variational integral over smooth

Read the complete text