On jump minimizing liftings for ๐ 1 \mathbb{S}^{1} -valued maps and connections with Ambrosio-Tortorelli-type ฮ \Gamma -limits
Giovanni Bellettini, Roberta Marziani, Riccardo Scala ยท 2025
Read On jump minimizing liftings for ๐ 1 \mathbb{S}^{1} -valued maps and connections with Ambrosio-Tortorelli-type ฮ \Gamma -limits by Giovanni Bellettini, Roberta Marziani, Riccardo Scala (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-2505-08731.txt.
Introduction
Giovanni Bellettini, Roberta Marziani, and Riccardo Scala work on maps that take values on a unit circle, in a 2025 math study. Such maps show up where direction matters but length is fixed. They study a known smoothing method that trades a map's jumpiness against a cost, and ask what limit that trade gives.
Two limits can arise, based on where the cost is set. One limit is nonlocal and ties to the idea of lifting a circle map to a line map with the least jump. The team proves the tight bounds needed to show such a best lifting exists. Those bounds stand as results in their own right. The work links a classic limit method to a fresh view of jumps in circle-valued maps.
Catalog record
| CREATOR | Giovanni Bellettini, Roberta Marziani, Riccardo Scala |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 109 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-08731
Browse the full library ยท About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
On jump minimizing liftings for ๐ 1 \mathbb{S}^{1} -valued maps and connections with Ambrosio-Tortorelli-type ฮ \Gamma -limits Giovanni Bellettini, Roberta Marziani, Riccardo Scala Abstract This paper is concerned with the ฮ \Gamma -limits of Ambrosio-Tortorelli-type functionals, for maps u u defined on an open bounded set ฮฉ โ โ n \Omega\subset{\mathbb{R}}^{n} and taking values in the unit circle ๐ 1 โ โ 2 \mathbb{S}^{1}\subset\mathbb{R}^{2} . Depending on the domain of the functional, two different ฮ \Gamma -limits are possible, one of which is nonlocal, and related to the notion of jump minimizing lifting, i.e., a lifting of a map u u whose measure of the jump set is minimal. The latter requires ad hoc compactness results for sequences of liftings which, besides being interesting by themselves, also allow to deduce existence of a jump minimizing lifting. Key words: Jump minimizing liftings, \Gamma -convergence, {\mathbb{S}}^{1} -valued maps, free boundary problems. AMS (MOS) subject classification: 49Q15, 49Q20, 49J45, 58E12. 1 Introduction This paper is devoted to the asymptotic analysis, via \Gamma -convergence, of regularized free discontinuity functionals for maps defined on a connected bounded open set \Omega\subset{\mathbb{R}}^{n} with Lipschitz boundary and taking values in the unit circle \mathbb{S}^{1}\subset\mathbb{R}^{2} . More specifically, define the two functional domains \displaystyle\widehat{\mathcal{A}}_{{\mathbb{S}}^{1}}:= \displaystyle\left\{(u,v)\in W^{1,1}(\Omega;\mathbb{S}^{1})\times W^{1,2}(\Omega)\colon v|\nabla u|\in L^{2}(\Omega),\ 0\leq v\leq 1\right\}\subset L^{1}(\Omega;{\mathbb{S}}^{1})\times L^{1}(\Omega), (1.1) \displaystyle\mathcal{A}_{{\mathbb{S}}^{1}}:= \displaystyle\left\{(u,v)\in W^{1,2}(\Omega;\mathbb{S}^{1})\times W^{1,2}(\Omega)\colon\ 0\leq v\leq 1\right\}\subset L^{1}(\Omega;{\mathbb{S}}^{1})\times L^{1}(\Omega). For \varepsilon\in(0,1] let us consider the corresponding family of Ambrosio-Tortorelli-type functionals {\widehat{\rm AT}_{\varepsilon}^{{\mathbb{S}}^{1}}},{{\rm AT}_{\varepsilon}^{{\mathbb{S}}^{1}}}\colon L^{1}(\Omega;\mathbb{S}^{1})\times L^{1}(\Omega)\to[0,+\infty] given by {\widehat{\rm AT}_{\varepsilon}^{{\mathbb{S}}^{1}}}(u,v):=\begin{cases}{\rm AT}_{\varepsilon}(u,v)&\text{ if }(u,v)\in\widehat{\mathcal{A}}_{{\mathbb{S}}^{1}},\\[10.00002pt] +\infty&\text{ otherwise in }L^{1}(\Omega;{\mathbb{S}}^{1})\times L^{1}(\Omega),\end{cases} (1.2) and {{\rm AT}_{\varepsilon}^{{\mathbb{S}}^{1}}}(u,v):=\begin{cases}{\rm AT}_{\varepsilon}(u,v)&\text{ if }(u,v)\in\mathcal{A}_{{\mathbb{S}}^{1}},\\[10.00002pt] +\infty&\text{ otherwise in }L^{1}(\Omega;{\mathbb{S}}^{1})\times L^{1}(\Omega),\end{cases} (1.3) where {\rm AT}_{\varepsilon}(u,v):=\int_{\Omega}\left(v^{2}|\nabla u|^{2}+\varepsilon|\nabla v|^{2}+\frac{(v-1)^{2}}{4\varepsilon}\right){\rm d}x, (1.4) with |\nabla u|^{2} indicating the Frobenius norm of \nabla u . It is clear that \mathcal{A}_{{\mathbb{S}}^{1}}\subset\widehat{\mathcal{A}}_{{\mathbb{S}}^{1}},\qquad{{\rm AT}_{\varepsilon}^{{\mathbb{S}}^{1}}}\geq{\widehat{\rm AT}_{\varepsilon}^{{\mathbb{S}}^{1}}}. Our main results, connecting these functionals with Mumford-Shah-type functionals, read as follows. Theorem 1.1 ( \Gamma -convergence of {\widehat{\rm AT}_{\varepsilon}^{{\mathbb{S}}^{1}}} ) . Let \Omega\subseteq\mathbb{R}^{n} be a connected bounded open set with Lipschitz boundary. We have \Gamma-L^{1}\lim_{\epsilon\to 0^{+}}{\widehat{\rm AT}_{\varepsilon}^{{\mathbb{S}}^{1}}}={\rm MS}_{{\mathbb{S}}^{1}}, where {\rm MS}_{{\mathbb{S}}^{1}}\colon L^{1}(\Omega;\mathbb{S}^{1})\times L^{1}(\Omega)\to[0,+\infty] is given by {\rm MS}_{{\mathbb{S}}^{1}}(u,v):=\begin{cases}\displaystyle\int_{\Omega}|\nabla u|^{2}{\rm d}x+\mathcal{H}^{n-1}(S_{u})&{\rm if}~u\in SBV^{2}(\Omega;\mathbb{S}^{1}),\,v=1\text{ a.e.}\,,\\[10.00002pt] +\infty&{\rm otherwise~in}~L^{1}(\Omega;{\mathbb{S}}^{1})\times L^{1}(\Omega).\end{cases} (1.5) Theorem 1.2 ( \Gamma -convergence of {{\rm AT}_{\varepsilon}^{{\mathbb{S}}^{1}}} ) . Let \Omega\subseteq\mathbb{R}^{n} be a connected and simply-connected bounded open set with Lipschitz boundary. We have \Gamma-L^{1}\lim_{\epsilon\to 0^{+}}{{\rm AT}_{\varepsilon}^{{\mathbb{S}}^{1}}}={\rm MS}_{{\rm lift}}, where {\rm MS}_{{\rm lift}}\colon L^{1}(\Omega;\mathbb{S}^{1})\times L^{1}(\Omega)\to[0,+\infty] is given by {\rm MS}_{{\rm lift}}(u,v):=\begin{cases}\displaystyle\int_{\Omega}|\nabla u|^{2}\,{\rm d}x+m_{2}[u]&{\rm if~}u\in SBV^{2}(\Omega;\mathbb{S}^{1}),\,v=1\text{ a.e.}\\[10.00002pt] +\infty&{\rm otherwise~in~}L^{1}(\Omega;{\mathbb{S}}^{1})\times L^{1}(\Omega),\end{cases} (1.6) with m_{2}[u]:=\inf\{\mathcal{H}^{n-1}(S_{\varphi})\colon\varphi\in GSBV^{2}(\Omega),\,e^{i\varphi}=u\text{ a.e. in }\Omega\}. (1.7) In ( 1.5 ) and ( 1.7 ) the symbol \mathcal{H}^{n-1}(S_{\varphi}) stands for the ({n-1}) -dimensional Hausdorff measure of the jump set S_{\varphi} of \varphi , a lifting 1 1 1 See Section 2.3 . of u , and SBV^{2}(\Omega;{\mathbb{S}}^{1}) (resp. GSBV^{2}(\Omega) ) is the space of {\mathbb{S}}^{1} -valued maps with special bounded variation in \Omega (resp. the space of generalized special bounded variation functions in \Omega ) whose absolutely continuous part of the gradient is square integrable and whose jump set has finite (n-1) -dimensional Hausdorff measure. To better understand the above results, some comments are in order. (i) In Theorem 1.1 the \Gamma -limit in ( 1.5 ) is the classical Mumford-Shah functional for \mathbb{S}^{1} -valued maps, and indeed part of the proof is an adaptation of known results, together with some applications of the properties of liftings of nonsmooth maps; (ii) the second, more interesting, \Gamma -limit in Theorem 1.2 is nonlocal, it does not have an integral representation and depends on the minimization problem in ( 3.1 ). The singular term m_{2}[u] is the penalization of a particular lifting of u , the one which minimizes the measure of the jump. It is clear that {\rm MS}_{{\rm lift}}\geq{\rm MS}_{{\mathbb{S}}^{1}} . These two theorems show the huge difference made by the choice of the two domains ( 1.1 ) where defining the approximating functionals, in particular the requirements u\in W^{1,1}(\Omega;{\mathbb{S}}^{1}) , v|\nabla u|\in L^{2}(\Omega) , as opposite to the more standard requirement u\in W^{1,2}(\Omega;{\mathbb{S}}^{1}) . The reason of the corresponding different limit behaviours is topological in nature, and it is better explained by the following example, see also [ 26 , pag. 30] . The example of the vortex map. Let n=2 , \Omega=B_{1}(0)\subset\mathbb{R}^{2} , and consider the vortex map u_{V}(x)=\frac{x}{|x|} for any x\in B_{1}(0)\setminus\{0\} . Then u_{V}\in W^{1,p}(B_{1}(0);{\mathbb{S}}^{1}) for any p\in[1,2) and u\notin W^{1,2}(B_{1}(0);{\mathbb{S}}^{1}) . Thus in particular {\rm MS}_{{\mathbb{S}}^{1}}(u_{V},1)=+\infty\qquad{\text{since }u_{V}\notin SBV^{2}(\Omega;\mathbb{S}^{1}).} Also, any lifting of u_{V} jumps (in B_{1}(0) ) at least on some curve connecting the origin to \partial B_{1}(0) . Now, let us modify u_{V} in a small neighbourhood of the origin to get a function which is in SBV^{2}(\Omega;\mathbb{S}^{1}) , by introducing a small jump. To this purpose consider the lifting of u_{V} given by the argument function 2 2 2 I.e., the imaginary part of the complex logarithm. \theta jumping on the positive real axis. Hence, in \Omega , S_{\theta}=(0,1)\times\{0\} with jump opening \text{{\rm\textlbrackdbl}}{\theta}\text{{\rm\textrbrackdbl}}=2\pi . Now, for 0<r<R\leq 1 define the annulus A_{r,R}=B_{R}(0)\setminus\overline{B}_{r}(0) , and let \sigma\in(0,1) . Consider the following perturbation of \theta and u_{V} on B_{1}(0)\setminus\{0\} : let \chi_{\sigma}\in C^{\infty}(B_{\sigma}(0),[0,1]) be such that \chi_{\sigma}\equiv 0 in B_{\sigma/4}(0) , \chi_{\sigma}\equiv 1 in A_{3/4\sigma,\sigma} , 0<\chi_{\sigma}<1 in A_{\sigma/4,3\sigma/4} , |\nabla\chi_{\sigma}|\leq\frac{C}{\sigma} for some C>0 , \theta_{\sigma}(x):=\begin{cases}\chi_{\sigma}(x)\theta(x)&{\rm if}~x\in B_{\sigma}(0)\setminus\{0\},\\ \theta(x)&{\rm if}~x\in A_{\sigma,1}(0),\end{cases}\qquad\qquad u^{(\sigma)}:=e^{i\theta_{\sigma}}. [formula] [formula] [formula] [formula] [formula] Figure 1: The dotted segment denotes the set S^{f}_{\theta_{\sigma}} where \text{{\rm\textlbrackdbl}}{\theta_{\sigma}}\text{{\rm\textrbrackdbl}} varies between 0 and 2\pi , the continuous segment denotes the set S^{I}_{\theta_{\sigma}} where \text{{\rm\textlbrackdbl}}{\theta_{\sigma}}\text{{\rm\textrbrackdbl}}=2\pi . Thus we have (see Figure 1 ) S_{\theta_{\sigma}}=S^{f}_{\theta_{\sigma}}\cup S^{I}_{\theta_{\sigma}}\quad\text{where}\quad S^{f}_{\theta_{\sigma}}=\Big(\frac{\sigma}{4},\frac{3\sigma}{4}\Big)\times\{0\}\,,\quad S^{I}_{\theta_{\sigma}}=\Big(\frac{3\sigma}{4},1\Big)\times\{0\}\,, moreover \text{{\rm\textlbrackdbl}}{\theta_{\sigma}}\text{{\rm\textrbrackdbl}} varies from 0 to 2\pi on S^{f}_{\theta_{\sigma}} and \text{{\rm\textlbrackdbl}}{\theta_{\sigma}}\text{{\rm\textrbrackdbl}}\equiv 2\pi on S^{I}_{\theta_{\sigma}} . This in turn implies u^{(\sigma)}=u_{V} on A_{3\sigma/4,1} and S_{u^{(\sigma)}}=S^{f}_{\theta_{\sigma}} . We also have u^{(\sigma)}\in SBV^{2}(B_{1}(0);{\mathbb{S}}^{1}) , and {\rm MS}_{{\mathbb{S}}^{1}}(u^{(\sigma)},1)<+\infty , and m_{2}[u^{(\sigma)}] turns out to be the length of the shortest segment joining the right extremum of S_{u^{(\sigma)}} to the boundary of \Omega . Indeed any lifting \varphi of {u^{(\sigma)}} has the following properties: S_{\varphi}\supset S_{\theta_{\sigma}}^{f} , \varphi is a lifting of u_{V} on A_{3\sigma/4,1} and \varphi is a lifting of (1,0) on B_{\sigma}(0) . Hence, in order to be minimal it must be S_{\varphi}=S_{\theta_{\sigma}} and therefore m_{2}[u^{(\sigma)}]=\mathcal{H}^{1}(S_{\theta_{\sigma}})=1-\sigma/4 . The map u^{(\sigma)} is one of the possible maps on which one may compute the two \Gamma -limits above. It turns out that the gradient integrability of the approximating maps is crucial to devise the two behaviours. In particular in Theorem 1.2 , due to topological obstructions related to the fact that u^{(\sigma)} has non-zero degree on any circle \partial B_{\rho}(0) with 3\sigma/4\leq\rho<1 (as it coincides with the vortex map outside a small neighbourhood of the origin), the recovery sequence gives rise to the contribution m_{2}[u^{(\sigma)}] . Indeed, a naive approach to get the upper bound would be to adapt the classical construction given in [ 7 , 8 ] . More precisely, the latter consists in regularising u^{(\sigma)} in a neighbourhood with width \ll\varepsilon of its jump set S_{u^{(\sigma)}} , being careful to keep values on the unit circle \mathbb{S}^{1} , and then define v_{\varepsilon} accordingly. However this leads to an approximating function u_{\varepsilon}\colon B_{1}(0)\to\mathbb{S}^{1} which coincides with the vortex map far from the origin, and so necessarily u_{\varepsilon} has non-zero degree, which in turn implies 3 3 3 Indeed, if u_{\varepsilon}\in W^{1,2}(B_{1}(0),\mathbb{S}^{1}) ,