Small-scale magnitude below one for cyclic two-chunk finite metric spaces
Tsubasa Kamiyama · 2026
Read Small-scale magnitude below one for cyclic two-chunk finite metric spaces by Tsubasa Kamiyama (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-2606-00707.txt.
Introduction
Kamiyama studies a size-like measure for finite metric spaces in this 2026 maths paper. The measure is called magnitude. When a space is shrunk, its copies collapse to a single point. Yet the magnitude can still record how that collapse took place.
Kamiyama builds a family of cyclic two-chunk spaces, set by four parameters, and works out their magnitude in exact form. With that formula he finds balanced cases whose small-scale magnitude falls below one. One case has twelve points, and its limit is 44/59, well under one. The question grows from Leinster's work on magnitude and from a small-scale view set out by Roff and Yoshinaga. The result shows that a space can look like a point yet keep a rich trace of its shape.
Catalog record
| CREATOR | Tsubasa Kamiyama |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 30 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.00707
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Small-scale magnitude below one for cyclic two-chunk finite metric spaces Tsubasa Kamiyama Abstract . Motivated by the small-scale viewpoint of Roff and Yoshinaga, we study finite metric spaces whose scaled copies collapse to a single point while their magnitude remembers how the collapse takes place. The limit metric space is geometrically indistinguishable from a point, but the magnitude function can detect differences in the path of collapse. We introduce a four-parameter family of cyclic two-chunk finite metric spaces, compute their magnitude explicitly, and use the formula to construct balanced examples whose small-scale magnitude is less than one. In particular, we exhibit a twelve-point finite metric space satisfying lim t → 0 + Mag ( t X ) = 44 / 59 < 1 . \lim_{t\to 0+}\Mag(tX)=44/59<1. The guiding question and the terminology around the one-point property come from Leinster’s magnitude of finite metric spaces and from Roff–Yoshinaga’s work on small-scale limits. 2020 Mathematics Subject Classification 51F99, 05C12, 15A18 1. Introduction Magnitude was introduced by Leinster as a size invariant of enriched categories and, in particular, of finite metric spaces [ 1 ] . For a finite metric space X , its scaled copy tX is obtained by multiplying all distances by t>0 . As t\to 0^{+} , the spaces tX collapse to a one-point metric space. The one-point property asks whether this geometric collapse is reflected by magnitude, namely whether (1.1) \lim_{t\to 0^{+}}\Mag(tX)=1. The small-scale behaviour of magnitude was recently studied systematically by Roff and Yoshinaga [ 3 ] . Their work gives a sharp picture of the one-point property for finite metric spaces. They proved that, for each fixed cardinality, the space of finite metric spaces contains a dense open subset on which the one-point property holds. In this sense, the one-point property is generic. They also showed that this generic behaviour is best possible: every metric space with at most four points has the one-point property, while the property already fails for a certain five-point metric space. In their example, the space is obtained from a two-point space of distance 4/3 and a three-point equilateral space of distance 2 , and its small-scale magnitude is 7/6 . An important insight of Roff and Yoshinaga’s work is that failure of the one-point property is not merely a technical pathology. Although such failure is non-generic, it can be quantitatively large: the small-scale limit of magnitude can be made to take arbitrary prescribed real values greater than one. Thus their results reveal a striking contrast. Magnitude is generically stable under collapse to a point, but, in exceptional cases, the small-scale limit can retain delicate information about the manner in which the original finite metric space collapses. The present note continues this line of investigation by showing that the exceptional behaviour is not confined to the side above one. We construct explicit finite metric spaces whose small-scale magnitude is strictly less than one. More precisely, we introduce a four-parameter family of cyclic two-chunk finite metric spaces, compute their magnitude explicitly, and exhibit a twelve-point example satisfying \lim_{t\to 0^{+}}\Mag(tX)=\frac{44}{59}<1. Consequently, a finite metric space can collapse to a point while its magnitude converges to a value below the magnitude of a point. Before giving the definitions, we display the motivating example. Let X be the twelve-point finite metric space whose distance matrix is (1.2) \displaystyle D=\begin{pmatrix}0&\frac{7}{3}&\frac{7}{3}&\frac{7}{3}&\frac{7}{3}&\frac{7}{3}&3&3&\frac{7}{6}&\frac{7}{6}&\frac{7}{6}&\frac{7}{6}\\ \frac{7}{3}&0&\frac{7}{3}&\frac{7}{3}&\frac{7}{3}&\frac{7}{3}&\frac{7}{6}&3&3&\frac{7}{6}&\frac{7}{6}&\frac{7}{6}\\ \frac{7}{3}&\frac{7}{3}&0&\frac{7}{3}&\frac{7}{3}&\frac{7}{3}&\frac{7}{6}&\frac{7}{6}&3&3&\frac{7}{6}&\frac{7}{6}\\ \frac{7}{3}&\frac{7}{3}&\frac{7}{3}&0&\frac{7}{3}&\frac{7}{3}&\frac{7}{6}&\frac{7}{6}&\frac{7}{6}&3&3&\frac{7}{6}\\ \frac{7}{3}&\frac{7}{3}&\frac{7}{3}&\frac{7}{3}&0&\frac{7}{3}&\frac{7}{6}&\frac{7}{6}&\frac{7}{6}&\frac{7}{6}&3&3\\ \frac{7}{3}&\frac{7}{3}&\frac{7}{3}&\frac{7}{3}&\frac{7}{3}&0&3&\frac{7}{6}&\frac{7}{6}&\frac{7}{6}&\frac{7}{6}&3\\ 3&\frac{7}{6}&\frac{7}{6}&\frac{7}{6}&\frac{7}{6}&3&0&\frac{29}{15}&\frac{29}{15}&\frac{29}{15}&\frac{29}{15}&\frac{29}{15}\\ 3&3&\frac{7}{6}&\frac{7}{6}&\frac{7}{6}&\frac{7}{6}&\frac{29}{15}&0&\frac{29}{15}&\frac{29}{15}&\frac{29}{15}&\frac{29}{15}\\ \frac{7}{6}&3&3&\frac{7}{6}&\frac{7}{6}&\frac{7}{6}&\frac{29}{15}&\frac{29}{15}&0&\frac{29}{15}&\frac{29}{15}&\frac{29}{15}\\ \frac{7}{6}&\frac{7}{6}&3&3&\frac{7}{6}&\frac{7}{6}&\frac{29}{15}&\frac{29}{15}&\frac{29}{15}&0&\frac{29}{15}&\frac{29}{15}\\ \frac{7}{6}&\frac{7}{6}&\frac{7}{6}&3&3&\frac{7}{6}&\frac{29}{15}&\frac{29}{15}&\frac{29}{15}&\frac{29}{15}&0&\frac{29}{15}\\ \frac{7}{6}&\frac{7}{6}&\frac{7}{6}&\frac{7}{6}&3&3&\frac{29}{15}&\frac{29}{15}&\frac{29}{15}&\frac{29}{15}&\frac{29}{15}&0\end{pmatrix} The terminology used to describe this matrix will be introduced below. It is a (6,2;7/3,29/15,3,7/6) -two-chunk cyclic metric space; equivalently, its parameters are (1.3) \alpha=\frac{7}{3},\qquad\beta=\frac{29}{15},\qquad\gamma=3,\qquad\delta=\frac{7}{6}. Using the closed formula proved in Proposition 3.2 and applying l’Hopital’s rule twice as in Proposition 4.3 , one obtains (1.4) \lim_{t\to 0+}\Mag(tX)=\frac{44}{59}<1. Thus this example collapses to a one-point space, but its small-scale magnitude does not converge to the magnitude of a point. 2. Preliminaries Definition 2.1 (Finite metric space) . A finite metric space is a pair (X,d) , where X is a finite set and (2.1) d\colon X\times X\longrightarrow\mathbb{R}_{\geq 0} satisfies, for all x,y,z\in X , (2.2) d(x,y)=0\Longleftrightarrow x=y,\qquad d(x,y)=d(y,x),\qquad d(x,z)\leq d(x,y)+d(y,z). Definition 2.2 (Magnitude of a finite metric space) . Let (X,d) be a finite metric space with (2.3) X=\{x_{1},\ldots,x_{m}\}. Its zeta matrix is (2.4) Z_{X}=\bigl(e^{-d(x_{i},x_{j})}\bigr)_{1\leq i,j\leq m}. A weighting on X is a vector w=(w_{1},\ldots,w_{m})^{T}\in\mathbb{R}^{m} satisfying (2.5) Z_{X}w=\mathbf{1}, where (2.6) \mathbf{1}=(1,1,\ldots,1)^{T}\in\mathbb{R}^{m}. If a weighting exists, the magnitude of X is (2.7) \Mag(X)=\sum_{i=1}^{m}w_{i}. When Z_{X} is invertible, the weighting is unique and (2.8) \Mag(X)=\mathbf{1}^{T}Z_{X}^{-1}\mathbf{1}. Definition 2.3 (Scaled metric space) . Let (X,d) be a finite metric space and let t>0 . The t -scaled metric space is (2.9) tX=(X,td), where (2.10) (td)(x,y)=t\,d(x,y) for all x,y\in X . Definition 2.4 (Distance matrix) . A real m\times m matrix D=(D_{ij}) is called a distance matrix if (2.11) D_{ii}=0,\qquad D_{ij}=D_{ji}>0\quad(i\neq j), and if the triangle inequalities (2.12) D_{ij}\leq D_{i\ell}+D_{\ell j} hold for all 1\leq i,j,\ell\leq m . Definition 2.5 (Metric space generated by a distance matrix) . Let D=(D_{ij}) be an m\times m distance matrix. The finite metric space generated by D is (2.13) X_{D}=\{x_{1},\ldots,x_{m}\}, with metric d_{D} defined by (2.14) d_{D}(x_{i},x_{j})=D_{ij}. Before introducing the cyclic family, we fix the following notation: for s\in\mathbb{R} , \lfloor s\rfloor denotes the greatest integer not exceeding s . Definition 2.6 ( (n,k;\alpha,\beta,\gamma,\delta) -two-chunk cyclic distance matrix) . Let n\geq 2 , let 1\leq k\leq\lfloor n/2\rfloor , and let (2.15) \alpha,\beta,\gamma,\delta>0. Put (2.16) \mathcal{A}=\{a_{0},\ldots,a_{n-1}\},\qquad\mathcal{B}=\{b_{0},\ldots,b_{n-1}\}, where a_{i+n}=a_{i} and b_{i+n}=b_{i} for every integer i . The (n,k;\alpha,\beta,\gamma,\delta) -two-chunk cyclic matrix is the symmetric 2n\times 2n matrix D_{n,k}(\alpha,\beta,\gamma,\delta) whose entries are (2.17) \displaystyle D(a_{i},a_{j}) \displaystyle=\begin{cases}0,&i=j,\\ \alpha,&i\neq j,\end{cases} (2.18) \displaystyle D(b_{i},b_{j}) \displaystyle=\begin{cases}0,&i=j,\\ \beta,&i\neq j,\end{cases} (2.19) \displaystyle D(a_{i},b_{j}) \displaystyle=D(b_{j},a_{i})=\begin{cases}\gamma,&\begin{aligned} &\text{if }j\equiv i+s\pmod{n}\\ &\text{for some }s\in\{0,1,\ldots,k-1\},\end{aligned}\\ \delta,&\text{otherwise}.\end{cases} If this matrix is a distance matrix in the sense of Definition 2.4 , then it is called an (n,k;\alpha,\beta,\gamma,\delta) -two-chunk cyclic distance matrix . Definition 2.7 ( (n,k;\alpha,\beta,\gamma,\delta) -two-chunk cyclic metric space) . The metric space generated by an (n,k;\alpha,\beta,\gamma,\delta) -two-chunk cyclic distance matrix is called an (n,k;\alpha,\beta,\gamma,\delta) -two-chunk cyclic metric space . We denote it by (2.20) X_{n,k}(\alpha,\beta,\gamma,\delta). Lemma 2.8 (Triangle inequalities for the two-chunk matrix) . For n\geq 3 and 1\leq k\leq\lfloor n/2\rfloor , the matrix D_{n,k}(\alpha,\beta,\gamma,\delta) is a distance matrix if and only if the following inequalities hold: (2.21) \displaystyle\max\{\alpha,\beta\} \displaystyle\leq 2\gamma \displaystyle\text{if }k\geq 2, (2.22) \displaystyle\max\{\alpha,\beta\} \displaystyle\leq 2\delta, (2.23) \displaystyle\max\{\alpha,\beta\} \displaystyle\leq\gamma+\delta, (2.24) \displaystyle|\gamma-\delta| \displaystyle\leq\min\{\alpha,\beta\}. Proof. It is enough to check non-degenerate triangles. Triangles contained entirely in \mathcal{A} have all side lengths equal to \alpha , and triangles contained entirely in \mathcal{B} have all side lengths equal to \beta . Hence those triangles automatically satisfy the triangle inequality. Consider next a triangle with two vertices in \mathcal{A} and one vertex in \mathcal{B} , say a_{i},a_{i^{\prime}}\in\mathcal{A} with i\neq i^{\prime} , and b_{j}\in\mathcal{B} . One side has length \alpha . For a vertex a_{\ell}\in\mathcal{A} , the cross-block side d(a_{\ell},b_{j}) has length \gamma exactly when (2.25) j\equiv\ell+s\pmod{n}\quad\text{for some }s\in\{0,1,\ldots,k-1\}, and otherwise it has length \delta . Equivalently, for fixed j , the \gamma -neighbours of b_{j} inside \mathcal{A} are precisely (2.26) a_{j},a_{j-1},\ldots,a_{j-k+1}, where subscripts are understood modulo n . Hence exactly k vertices of \mathcal{A} have distance \gamma from b_{j} , and the remaining n-k vertices have distance \delta from b_{j} . Therefore the two cross-block side lengths are either