The distance between homotopy classes of Sobolev maps on spheres
Rupert L. Frank, Paata Ivanisvili · 2026
Read The distance between homotopy classes of Sobolev maps on spheres by Rupert L. Frank, Paata Ivanisvili (2026) 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-2606-29382.txt.
Introduction
Frank and Ivanisvili study maps from a sphere to itself that lie in the critical Sobolev space, a space of functions with just enough smoothness for the problem. Each map carries a Brouwer degree, an integer that counts how many times the map wraps around the sphere. They ask how far apart maps of different degrees must sit. Distance here means the cheapest cost of deforming one map toward another.
Their main result gives that directed distance exactly. It equals an explicit constant multiplied by the difference between the two degrees. For maps on the ordinary two-dimensional sphere, the formula settles an open problem posed by Brezis. The answer shows the gap between degree classes grows in a clean, linear way. No cheaper shortcut between different degrees exists. The work sits between analysis, geometry, and the calculus of variations.
Catalog record
| CREATOR | Rupert L. Frank, Paata Ivanisvili |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.29382
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
The distance between homotopy classes of Sobolev maps on spheres Rupert L. Frank, Paata Ivanisvili Abstract . We consider self-maps of a sphere in the critical Sobolev space with a given Brouwer degree. Our main result is that the (directed) distance between maps of different degrees is equal to an explicit constant times the difference in degrees. In the case of the 2-sphere this resolves an open problem by Brezis. 2020 Mathematics Subject Classification 46E35, 58D15, 49Q20, 58E20 1. Introduction and main result 1.1. The critical degree classes In this paper we are interested in maps from the unit sphere \mathbb{S}^{n}=\{x\in\mathbb{R}^{n+1}:|x|=1\} , n\geq 2 , to itself that belong to the Sobolev space (1.1) W^{1,n}(\mathbb{S}^{n};\mathbb{S}^{n}):=\{u\in W^{1,n}(\mathbb{S}^{n};\mathbb{R}^{n+1}):|u|=1\text{ a.e.}\}. As we will recall in detail in Section 2 below, such maps have a well-defined Brouwer degree. For d\in\mathbb{Z} , set (1.2) \mathcal{E}_{d}^{(n)}:=\{u\in W^{1,n}(\mathbb{S}^{n};\mathbb{S}^{n}):\deg u=d\}. Our main result gives an explicit formula for the min-max quantity \sup_{f\in\mathcal{E}_{d_{1}}^{(n)}}\inf_{g\in\mathcal{E}_{d_{2}}^{(n)}}\int_{\mathbb{S}^{n}}|\mathrm{D}_{\tau}f-\mathrm{D}_{\tau}g|^{n}\,\,\mathrm{d}\sigma\,, where D_{\tau}u denotes the weak (tangential) derivative of u\in W^{1,n}(\mathbb{S}^{n};\mathbb{S}^{n}) and we employ the Hilbert–Schmidt norm, that is, if (\tau_{1},\dots,\tau_{n}) is a local orthonormal frame of T\mathbb{S}^{n} , then |\mathrm{D}_{\tau}f-\mathrm{D}_{\tau}g|^{2}=\sum_{i=1}^{n}|\partial_{\tau_{i}}f-\partial_{\tau_{i}}g|^{2}. Also, \,\mathrm{d}\sigma denotes surface measure on \mathbb{S}^{n} . Introducing the constant \kappa_{n}:=n^{n/2}|\mathbb{S}^{n}|=n^{n/2}\frac{2\pi^{(n+1)/2}}{\Gamma((n+1)/2)}, we can state our main result. Theorem 1.1 . Let n\geq 2 and d_{1},d_{2}\in\mathbb{Z} . Then (1.3) \sup_{f\in\mathcal{E}_{d_{1}}^{(n)}}\inf_{g\in\mathcal{E}_{d_{2}}^{(n)}}\int_{\mathbb{S}^{n}}|\mathrm{D}_{\tau}f-\mathrm{D}_{\tau}g|^{n}\,\,\mathrm{d}\sigma=\kappa_{n}\,|d_{1}-d_{2}|\,. The equality ( 1.3 ) is also true for n=1 , as shown in [ BMS16 , Theorem 1.4(1)] . For n=2 , the inequality \leq in ( 1.3 ) was already known (see [ Bre23 ] and references therein) and the inequality \geq was known when 0\leq d_{1}<d_{2} (see [ BMS16 , Proposition 7.3(2)] ), but the inequality \geq in the general case is new and gives an affirmative answer to Problem 5.9 on Brezis’s list ‘Some of my favorite open problems’ [ Bre23 ] . To the best of our knowledge, before the present work no exact all-degree formula was known beyond the one-dimensional case and the two-dimensional ranges above. Theorem 1.1 has the following consequence. If we set, following [ BMS16 ] , (1.4) \operatorname{Dist}_{n}(\mathcal{E}_{d_{1}}^{(n)},\mathcal{E}_{d_{2}}^{(n)}):=\sup_{f\in\mathcal{E}_{d_{1}}^{(n)}}\inf_{g\in\mathcal{E}_{d_{2}}^{(n)}}\left(\int_{\mathbb{S}^{n}}|\mathrm{D}_{\tau}f-\mathrm{D}_{\tau}g|^{n}\,\,\mathrm{d}\sigma\right)^{1/n}, then we obtain the symmetry relation (1.5) \operatorname{Dist}_{n}(\mathcal{E}_{d_{1}}^{(n)},\mathcal{E}_{d_{2}}^{(n)})=\operatorname{Dist}_{n}(\mathcal{E}_{d_{2}}^{(n)},\mathcal{E}_{d_{1}}^{(n)})\,. Indeed, by Theorem 1.1 , both sides are equal to (\kappa_{n}\,|d_{1}-d_{2}|)^{1/n} . The equality ( 1.5 ) is new for all n\geq 3 . In particular, it gives an affirmative answer to Open Problem 1 in [ BMS16 ] in the case s=1 , p=n . Also, Theorem 1.1 answers Open Problem 2 in [ BMS16 ] for p=n . Background Before describing the strategy of our proof, let us mention some previous results on this and related problems. The interaction between degree and concentration in critical Sobolev spaces arose prominently in the work of Brezis and Coron on harmonic maps [ BC83 , Lemmas 1–2] ; related relaxation phenomena were developed by Bethuel, Brezis, and Coron [ BBC90 , Theorems 2–3] . Brezis and Nirenberg subsequently developed degree theory in \mathrm{VMO} [ BN95 , Section I.3 and Theorem 1] . Brezis, Mironescu, and Shafrir [ BMS16 ] introduced the ‘directed distance’ (1.6) \operatorname{Dist}_{p}(\mathcal{E}_{d_{1}}^{(n,p)},\mathcal{E}_{d_{2}}^{(n,p)}):=\sup_{f\in\mathcal{E}_{d_{1}}^{(n,p)}}\inf_{g\in\mathcal{E}_{d_{2}}^{(n,p)}}\|f-g\|_{\dot{W}^{n/p,p}(\mathbb{S}^{n};\mathbb{S}^{n})}\,, extending ( 1.4 ) to more general Sobolev spaces W^{n/p,p}(\mathbb{S}^{n};\mathbb{S}^{n}) with 1\leq p<\infty . Moreover, \mathcal{E}_{d}^{(n,p)} consists of maps in W^{n/p,p}(\mathbb{S}^{n};\mathbb{S}^{n}) with degree d (which is well-defined in W^{n/p,p}(\mathbb{S}^{n},\mathbb{S}^{n})\subset VMO(\mathbb{S}^{n};\mathbb{S}^{n}) ). In their Open Problem 1 they asked whether \operatorname{Dist}_{p}(\mathcal{E}_{d_{1}}^{(n,p)},\mathcal{E}_{d_{2}}^{(n,p)})=\operatorname{Dist}_{p}(\mathcal{E}_{d_{2}}^{(n,p)},\mathcal{E}_{d_{1}}^{(n,p)}) or, even better, whether it only depends on |d_{1}-d_{2}| . In their Open Problem 2 they asked whether one has a lower bound by some constant times |d_{1}-d_{2}|^{1/p} . Since the directed distance is bounded above by the corresponding Hausdorff distance, [ BMS16 , Theorem 1.7(1)] gives (1.7) \operatorname{Dist}_{p}(\mathcal{E}_{d_{1}}^{(n,p)},\mathcal{E}_{d_{2}}^{(n,p)})\leq C_{p,n}|d_{1}-d_{2}|^{1/p}. For a fixed source degree, [ BMS16 , Proposition 7.8] gives (1.8) \operatorname{Dist}_{p}(\mathcal{E}_{d_{1}}^{(n,p)},\mathcal{E}_{d_{2}}^{(n,p)})\geq c_{p,n,d_{1}}|d_{1}-d_{2}|^{1/p}, where the constant may depend on d_{1} . When 1\leq p\leq n+1 and d_{1}d_{2}\leq 0 , they proved the uniform estimate (1.9) \operatorname{Dist}_{p}(\mathcal{E}_{d_{1}}^{(n,p)},\mathcal{E}_{d_{2}}^{(n,p)})\geq c_{p,n}|d_{1}-d_{2}|^{1/p}\,, see [ BMS16 , Proposition 7.5] . For the case p=n that we are interested in, these estimates give the correct order of growth in every fixed-source regime and uniformly for opposite-sign degrees, but do not identify the sharp constant or imply symmetry for arbitrary pairs. As we have already mentioned, the analogous problem for n=1=p was completely settled by [ BMS16 , Theorem 1.4(1)] . Optimal results for n=1 and 1<p<\infty appear in [ Sha18 , Theorem 1.1 and Corollary 1.2] . In passing, we note that if one replaces the outer supremum in ( 1.6 ) by an infimum, one obtains zero; see [ BMS16 , Theorem 4(1)] . For related results in sub-critical Sobolev spaces we refer to [ LS14 , Theorem 3.4] and [ RS07 , Theorem 1 and Remark 2.1] . A recent development, which is decisive for our proof, is the heterotopic energy of Detaille and Van Schaftingen. Their formula identifies the least W^{1,n} energy of maps in one homotopy class converging almost everywhere to a map in another class; for sphere targets, the defect is exactly \kappa_{n} times the difference of degree [ DVS26 , Theorem 1.1 and Corollary 1.2] . Their bubbling theorem is stronger than the total-energy inequality: it locates the topological defect in atoms of the limiting energy measure [ DVS26 , Theorem 5.1] . In the introduction of their paper, they note explicitly that the heterotopic energy has “no apparent formal mathematical connection” with the earlier directed-distance functionals. A by-product of our paper is to develop such a connection. Outline of the proof To prove Theorem 1.1 we will prove an upper and a lower bound on the left side in ( 1.3 ). The proof of the upper bound is much easier than that of the lower bound and is presented in Section 3 . Similar arguments have appeared before in the literature. The idea is to locally insert a bubble. The minimal energy carried by one unit of Brouwer degree is \kappa_{n} , as is made precise in 3.1 . The intuition is that therefore it costs an energy of at most |d_{2}-d_{1}| times \kappa_{n} to change the degree d_{1} of a given map f to d_{2} of some other map. The proof of the lower bound takes up Sections 4 , 5 and 6 . In Section 5 we first construct a finite collection of rotations, all arbitrarily close to the identity, such that the orbit of every p\in\mathbb{S}^{n} affinely spans \mathbb{R}^{n+1} . Applying these rotations to a fixed smooth map f_{0} in rapidly alternating phases gives sphere-valued maps F_{j} of the same degree as f_{0} . The maps F_{j} are ‘hard to approximate’ in the sense that if a sequence g_{j} of sphere-valued maps satisfies a uniform L^{n} -bound on \mathrm{D}(F_{j}-g_{j}) , compactness and periodic averaging force F_{j}-g_{j}\to 0 in L^{n} ; this is 5.5 . We call this mechanism oscillatory pinning. Now for near-minimizers g_{j} corresponding to F_{j} , we remove the oscillation from the difference and radially normalize: u_{j}:=\frac{f_{0}-(F_{j}-g_{j})}{|f_{0}-(F_{j}-g_{j})|}. Because F_{j} is uniformly close to f_{0} , the denominator stays away from zero and u_{j} has the degree of g_{j} . Pinning gives u_{j}\to f_{0} in L^{n} . Thus, we are in the situation where all the maps u_{j} have a fixed degree d_{2} , while their L^{n} limit has degree d_{1} . A special case of the bubbling theorem of