On the non-uniqueness of locally minimizing clusters via singular cones

Lia Bronsard, Robin Neumayer, Michael Novack, Anna Skorobogatova · 2025

Read On the non-uniqueness of locally minimizing clusters via singular cones by Lia Bronsard, Robin Neumayer, Michael Novack, Anna Skorobogatova (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-2507-13995.txt.

Introduction

Lia Bronsard, Robin Neumayer, Michael Novack, and Anna Skorobogatova settle a uniqueness question in the geometry of clusters, in this 2025 mathematics paper. A cluster splits space into regions of set volumes with the least total interface area, much like soap films settling into equilibrium. The standard lens cluster, two caps back to back, was long suspected to be the only local winner of its kind.

The authors build three region partitions of space that locally minimize interface area under volume preserving changes. Far away, each partition looks like a singular cone that itself minimizes area. Because these rivals exist, the lens cluster is not unique in many dimensions, starting from eight. The counterexamples close a question that shape optimization had left open.

Catalog record

CREATORLia Bronsard, Robin Neumayer, Michael Novack, Anna Skorobogatova
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE127 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-13995

Browse the full library · About the Tower of Babel

Full text

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

On the non-uniqueness of locally minimizing clusters via singular cones Lia Bronsard, Robin Neumayer, Michael Novack, Anna Skorobogatova Abstract . We construct partitions of ℝ n \mathbb{R}^{n} into three sets { 𝒳 ⁡ ( 1 ) , 𝒳 ⁡ ( 2 ) , 𝒳 ⁡ ( 3 ) } \{\mathcal{X}(1),\mathcal{X}(2),\mathcal{X}(3)\} that locally minimize interfacial area among compactly supported volume preserving variations and that blow down at infinity to singular area-minimizing cones. As a consequence, we prove the non-uniqueness of the standard lens cluster in a large number of dimensions starting from 8 8 . 1. Introduction 1.1. Overview The classical multiple bubble problem in \mathbb{R}^{n} concerns the existence and structure of configurations of N regions of prescribed finite volumes, together with an exterior chamber of infinite volume, that minimize the surface area of their interfaces. 1 1 1 Note that the case N=1 is the classical isoperimetric problem. This problem has received much attention in recent decades, with an abundance of work concerning the existence [ 5 ] , structure and regularity [ 46 , 11 ] , and characterization [ 16 , 18 , 39 , 40 , 49 , 38 , 41 , 31 , 29 ] of minimizers and critical points. This includes recent breakthrough work in which Milman & Neeman [ 30 ] gave a complete classification of minimizers in the case N\leq\min\{4,n\} , resolving a long-standing conjecture of [ 45 ] . See the introduction of [ 30 ] for a comprehensive review of the literature surrounding this problem. Questions of these types have recently been extended to configurations with multiple infinite chambers, following the introduction of the \boldsymbol{(1,2)} -cluster problem , and more generally, the (N,M) -cluster problem, by the first author with Alama & Vriend in [ 1 ] in connection to tri-block copolymers. A (1,2) -cluster is a partition of \mathbb{R}^{n} into three chambers, each of which is a set of locally finite perimeter, generically denoted by \mathcal{X}=(\mathcal{X}(1),\mathcal{X}(2),\mathcal{X}(3)) , where 0<|\mathcal{X}(1)|<\infty and |\mathcal{X}(2)|=|\mathcal{X}(3)|=\infty . The cluster perimeter of a (1,2) -cluster \mathcal{X} in a ball B_{\rho}(x) is (1.1) \mathcal{P}(\mathcal{X};B_{\rho}(x)):=\frac{1}{2}\sum_{i=1}^{3}P(\mathcal{X}(i);B_{\rho}(x))\,. Here P(\mathcal{X}(i);B_{\rho}(x)) denotes the relative perimeter of \mathcal{X}(i) in B_{\rho}(x) , which simply agrees with the surface area of \partial\mathcal{X}(i)\cap B_{\rho}(x) when \partial\mathcal{X}(i) is piecewise smooth and is defined in section 2 in general. A (1,2) -cluster \mathcal{X} is locally minimizing 2 2 2 When there are two or more infinite-volume chambers, we necessarily need to consider local minimizers, due to the infinite perimeter contribution of the interfaces between neighboring pairs of infinite-volume chambers. if, for every \rho>0 , (1.2) \displaystyle\mathcal{P}(\mathcal{X};B_{\rho}(0))\leq\mathcal{P}(\mathcal{X}^{\prime};B_{\rho}(0)) whenever \mathcal{X}(i)\Delta\mathcal{X}^{\prime}(i)\subset\!\subset B_{\rho}(0) and |\mathcal{X}(i)|=|\mathcal{X}^{\prime}(i)| for i=1,2,3 3 3 3 As shown in [ 10 ] , local minimality of a (1,2) -cluster \mathcal{X} with respect to compactly supported variations satisfying |\mathcal{X}(i)|=|\mathcal{X}^{\prime}(i)| for i=1,2,3 is equivalent to local minimality taken against variations satisfying |\mathcal{X}(i)\cap B_{\rho}(0)|=|\mathcal{X}^{\prime}(i)\cap B_{\rho}(0)| for i=1,2,3 , where the volume of the infinite chambers is required to be preserved locally instead of only globally. . Clusters arise in the modeling of various physical systems such as immiscible fluid configurations [ 48 , 26 ] in bounded domains or, for our setting with more than one infinite chamber, in tri-block copolymers [ 2 , 3 ] in the small-volume “droplet” regime for multiple phases. Recent progress has been made toward existence, classification, and stability for (1,2) -clusters (and (M,N) -clusters more generally) in [ 1 , 36 , 10 , 9 , 31 ] . In [ 10 ] , the first and third authors give a complete classification of locally minimizing (1,2) -clusters for n\leq 7 , proving that the standard lens cluster \mathcal{X}_{\lens} is the unique locally minimizing (1,2) -cluster in \mathbb{R}^{n} modulo homotheties. They obtain the same result in higher dimensions ( n\geq 8 ) under the assumption of planar growth at infinity. The standard lens cluster, which is locally minimizing in every dimension, is characterized by the properties that |\mathcal{X}_{\lens}(1)|=1 , \partial\mathcal{X}_{{\lens}}(2)\cap\partial\mathcal{X}_{{\lens}}(3)\subset\{x_{n}=0\} and \partial\mathcal{X}_{{\lens}}(1) is the union of pair of equal-radii spherical caps meeting on \{x_{n}=0\} with equal angles of \frac{2\pi}{3} between the three interfaces. An analogous classification was previously shown for n=2 in [ 1 ] . The aforementioned classification in n\leq 7 is fundamentally related to the fact that planes are the only area-minimizing hypercones in these dimensions [ 44 ] . In view of the existence of non-planar area-minimizing hypercones when n\geq 8 [ 8 ] , the following question was raised in [ 10 ] and [ 36 ] : For n\geq 8 , do there exist locally minimizing (1,2) -clusters besides the standard lens? The main goal of this paper is to answer this open question in the affirmative for a large number of dimensions, starting from n=8 . Part of our proof relies on error estimates for rigorous numerical comparison of appropriate renormalized quantities and so our statement goes up to dimension 2700 as will be explained below. Theorem 1.1 . Let n\in\{8,\dots,2700\}. There exists a locally minimizing (1,2) -cluster \mathcal{X} that is not a standard lens. 1.2. Main Results The basic scheme to construct the local minimizing (1,2) -clusters of Theorem 1.1 goes as follows. Fix n\geq 8 and let K be a perimeter minimizing cone in \mathbb{R}^{n} whose boundary is not a plane, e.g., the region K=\{x_{1}^{2}+\dots+x_{4}^{2}<x_{5}^{2}+\cdots+x_{8}^{2}\} bounded by the Simons cone in \mathbb{R}^{8}. For R>0 large, we set up an energy minimization problem in the class of (1,2) -clusters \mathcal{X} with \mathcal{X}(1)\subset B_{3R} of volume 1 for which, outside of B_{3R} , the chambers \mathcal{X}(2) and \mathcal{X}(3) coincide with K and K^{c} respectively. Morally one wants to minimize the cluster perimeter \mathcal{P}(\mathcal{X};B_{4R}) , though by instead minimizing the energy \mathcal{P}(\mathcal{X};B_{4R})+\mathcal{G}_{R}(\mathcal{X}(1)) for a carefully constructed confinement potential \mathcal{G}_{R} , we circumvent technical challenges that would arise from the possibility of \mathcal{X}(1) saturating the constraint \mathcal{X}(1)\subset B_{3R} . For a sequence R_{k}\to\infty , take a sequence of minimizers \mathcal{X}_{k} to this minimization problem with R=R_{k} . We wish to obtain a local minimizer of \mathcal{P} as a limit of the \mathcal{X}_{k} , and to this end we use a “partial concentration compactness approach,” where the volume of \mathcal{X}_{k}(1) is the quantity we wish to preserve in the limit. Recall how classical concentration compactness methods are often used to prove existence in variational problems with critical scaling or on non-compact domains: Step 1: Characterize precisely how a sequence can lose compactness and the cost in energy to exhibit this behavior. Step 2: Establish compactness for a (minimizing) sequence by showing its energy lies below this loss-of-compactness threshold. For instance, to produce constant scalar curvature metrics in a given conformal class in the Yamabe problem, one shows that a noncompact minimizing sequence for the Yamabe functional must concentrate at a point and pay precisely the Yamabe constant of the sphere in energy [ 24 ] . Similarly, in the theory of stationary harmonic maps, loss of strong H^{1} -compactness for equibounded sequences is characterized by the formation of “bubbles”, and the energy loss is realized exactly by the sum of energies of such bubbles; see [ 27 , 33 ] . In the present setting, Step 1 as stated seems out of reach, as there are in principle myriad possible asymptotic behaviors,

Read the complete text