A Computational (Co)homological Approach to Contiguity Distance
Enrique Macías-Virgós, Ángel Méndez-Vázquez, David Mosquera-Lois · 2025
Read A Computational (Co)homological Approach to Contiguity Distance by Enrique Macías-Virgós, Ángel Méndez-Vázquez, David Mosquera-Lois (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-2510-18356.txt.
Introduction
Enrique Macías-Virgós, with Ángel Méndez-Vázquez and David Mosquera-Lois, works on a problem in topology, the branch of mathematics that studies shapes and maps between them. The authors want better tools for measuring how far apart two continuous maps can be, a quantity called homotopic distance. Direct measures are hard to compute, so lower bounds that are easier to find carry real value.
The paper introduces two new algebraic measures, called cohomological and homological distances, and shows they give computable lower bounds for homotopic distance. These new measures also improve on older estimates based on cup-length. A version for maps between simple building-block shapes is then defined, and a theorem shows it matches the continuous version after enough repeated subdivision. Worked examples close the argument.
Catalog record
| CREATOR | Enrique Macías-Virgós, Ángel Méndez-Vázquez, David Mosquera-Lois |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 32 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-18356
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A Computational (Co)homological Approach to Contiguity Distance Thanks: The second author acknowledges support of the Deputación Provincial da Coruña (Exp. 2025000022903). Enrique Macías-Virgós, Ángel Méndez-Vázquez, David Mosquera-Lois Abstract . We introduce a new algebraic invariant, the (co)homological distance , which provides computable lower bounds for the homotopic distance and refines the classical cup-length estimates. We then define the simplicial cohomological distance and prove a convergence theorem showing that, after sufficiently many barycentric subdivisions, it approximates the cohomological distance of the corresponding continuous maps. Several explicit computations are presented to illustrate the effectiveness of the proposed approach. Introduction In [ 5 ] , the notion of homotopic distance \mathrm{D}(f,g) between two continuous maps f,g\colon X\rightarrow Y was introduced. This invariant is defined as the least integer n\geq 0 for which there exists an open cover \{U_{0},\dots,U_{n}\} of X such that the restrictions f_{|U_{j}},g_{|U_{j}}\colon U_{j}\rightarrow Y are homotopic, f_{|U_{j}}\simeq g_{|U_{j}} , for all j=0,\dots,n . The homotopic distance extends classical invariants from spaces to maps. In particular, it provides a unified framework that generalizes the Lusternik–Schnirelmann category and the topological complexity. In the same work [ 5 ] , two of the authors also introduced the contiguity or simplicial distance \mathrm{s}\mathrm{D}(\varphi,\psi) between two simplicial maps \varphi,\psi\colon K\rightarrow L . It is the least integer n\geq 0 such that there exists a cover of K by subcomplexes K_{0},\dots,K_{n} with the property that \varphi_{|K_{j}} and \psi_{|K_{j}} lie in the same contiguity class, \varphi_{|K_{j}}\sim\psi_{|K_{j}} , for every j=0,\dots,n . This invariant was further investigated in [ 2 ] . By construction, the homotopic distance is always bounded above by the contiguity distance. However, a major drawback of the contiguity distance is that it does not converge to the homotopic distance under subdivision, as illustrated in [ 3 , Example 4.6] . In this article we introduce a new auxiliary invariant, the (co)homological distance , which enjoys very good computational properties. We further define its simplicial version, the simplicial (co)homological distance , and prove that, unlike the simplicial distance, it converges to the (co)homological distance: after sufficiently many subdivisions, the simplicial (co)homological distance approximates the (co)homological one (Theorem 2.3 ). Another key feature of the (co)homological distance is that it provides a lower bound for the homotopic distance. Moreover, it is sharper than the classical cup-length (see Example 1.8 ). Conventions Throughout this article, R denotes a commutative ring with a unit. All topological spaces and simplicial complexes are assumed to be path-connected. All simplicial complexes are assumed to be finite, as we are primarily interested in a computational perspective. 1. (Co)Homological Distance We introduce the cohomological and homological analogues of the homotopic distance. These invariants retain essential topological information while being more suitable for algebraic computation and provide better lower bounds than cup-length invariants. 1.1. The (co)homological distance We define the (co)homological distance between two continuous maps f,g\colon X\rightarrow Y . Definition 1.1 . The cohomological (homological) distance between f and g with coefficients in R , denoted by \mathrm{H^{\,\begin{picture}(-1.0,-3.0)\circle*{3.0}\end{picture}\ }\mathrm{D}}(f,g;R) (resp. \mathrm{H_{\,\begin{picture}(-1.0,-3.0)\circle*{3.0}\end{picture}\ }\mathrm{D}}(f,g;R) ), is the least integer n\geq 0 for which there exists an open cover \{U_{0},\dots,U_{n}\} of X such that (f_{|U_{j}})^{*}=(g_{|U_{j}})^{*}\colon H^{m}(Y;R)\rightarrow H^{m}(U_{j};R) (resp. (f_{|U_{j}})_{*}=(g_{|U_{j}})_{*}\colon H_{m}(U_{j};R)\rightarrow H_{m}(Y;R) ) for all m\geq 0 and for all j=0,\dots,n . Remark 1 . We always have \mathrm{H^{\,\begin{picture}(-1.0,-3.0)\circle*{3.0}\end{picture}\ }\mathrm{D}}(f,g;R)\leq\mathrm{D}(f,g) and \mathrm{H_{\,\begin{picture}(-1.0,-3.0)\circle*{3.0}\end{picture}\ }\mathrm{D}}(f,g;R)\leq\mathrm{D}(f,g) . Example 1.2 . Let p\colon\mathbb{S}^{3}\rightarrow\mathbb{S}^{2} be the Hopf fibration and let \mathrm{c}\colon\mathbb{S}^{3}\rightarrow\mathbb{S}^{2} be a constant map. Then \mathrm{D}(p,\mathrm{c})=1 , whereas \mathrm{H^{\,\begin{picture}(-1.0,-3.0)\circle*{3.0}\end{picture}\ }\mathrm{D}}(p,\mathrm{c};R)=0 and \mathrm{H_{\,\begin{picture}(-1.0,-3.0)\circle*{3.0}\end{picture}\ }\mathrm{D}}(p,\mathrm{c};R)=0 . Hence, the above inequalities can be strict. Let \mathrm{id}\colon X\rightarrow X be the identity map, \mathrm{c}\colon X\rightarrow X a constant map and \pi_{1},\pi_{2}\colon X\times X\rightarrow X the canonical projections. We define the cohomological (resp. homological) Lusternik-Schnirelmann category as \mathrm{H^{\,\begin{picture}(-1.0,-3.0)\circle*{3.0}\end{picture}\ }cat}(X;R)=\mathrm{H^{\,\begin{picture}(-1.0,-3.0)\circle*{3.0}\end{picture}\ }\mathrm{D}}(\mathrm{id},\mathrm{c}) (resp. \mathrm{H_{\,\begin{picture}(-1.0,-3.0)\circle*{3.0}\end{picture}\ }\mathrm{D}}(\mathrm{id},\mathrm{c}) ) and the cohomological (resp. homological) topological complexity \mathrm{H^{\,\begin{picture}(-1.0,-3.0)\circle*{3.0}\end{picture}\ }TC}(X;R)=\mathrm{H^{\,\begin{picture}(-1.0,-3.0)\circle*{3.0}\end{picture}\ }\mathrm{D}}(\pi_{1},\pi_{2};R) (resp. \mathrm{H_{\,\begin{picture}(-1.0,-3.0)\circle*{3.0}\end{picture}\ }\mathrm{D}}(\pi_{1},\pi_{2};R) ). As a consequence, mimicking the approach in [ 5 ] , for any pair of maps f,g\colon X\rightarrow Y , we have (1) \mathrm{H^{\,\begin{picture}(-1.0,-3.0)\circle*{3.0}\end{picture}\ }\mathrm{D}}(f,g;R)\leq\mathrm{H^{\,\begin{picture}(-1.0,-3.0)\circle*{3.0}\end{picture}\ }cat}(X;R). (2) \mathrm{H^{\,\begin{picture}(-1.0,-3.0)\circle*{3.0}\end{picture}\ }\mathrm{D}}(f,g;R)\leq\mathrm{H^{\,\begin{picture}(-1.0,-3.0)\circle*{3.0}\end{picture}\ }TC}(Y;R). Remark 2 . The homological Lusternik-Schnirelmann category was introduced by Fox ( [ 4 ] ) and we recover it as a particular case. 1.2. Cup-length In [ 5 ] , the authors established the following cohomological lower bound for the homotopic distance. Let \mathcal{J}(f,g;R)\subset H^{*}(X;R) be the image of f^{*}-g^{*}\colon H^{*}(Y;R)\rightarrow H^{*}(X;R). Definition 1.3 . We denote by \mathrm{l.c.p.}\,\mathcal{J}(f,g;R) the least integer n\geq 0 such that every product of n+1 elements of \mathcal{J}(f,g;R) vanishes. Theorem 1.4 . It holds that \mathrm{l.c.p.}\,\mathcal{J}(f,g;R)\leq\mathrm{H^{\,\begin{picture}(-1.0,-3.0)\circle*{3.0}\end{picture}\ }\mathrm{D}}(f,g;R) . Proof. Suppose \mathrm{H^{\,\begin{picture}(-1.0,-3.0)\circle*{3.0}\end{picture}\ }\mathrm{D}}(f,g;R)=n , and let \{U_{0},\dots,U_{n}\} be an open cover of X such that (f_{|U_{j}})^{*}=(g_{|U_{j}})^{*} for every j=0,\dots,n . Consider an arbitrary product x_{0}\smile\cdots\smile x_{n} of n+1 elements with x_{k}\in\mathcal{J}(f,g;R) for each k=0,\dots,n . For each open set U=U_{k} in the cover, consider the long exact sequence of the pair (X,U) : {\lx@inpgf@ignorespaces\cdots} {\lx@inpgf@ignorespaces{H^{m}(X,U;R)}} {\lx@inpgf@ignorespaces H^{m}(X;R)} {\lx@inpgf@ignorespaces H^{m}(U;R)} {\lx@inpgf@ignorespaces\cdots} {\lx@inpgf@ignorespaces H^{m}(Y;R)} \scriptstyle{\lx@inpgf@ignorespaces j_{U}} \scriptstyle{\lx@inpgf@ignorespaces i^{*}} \scriptstyle{\lx@inpgf@ignorespaces f^{*}-g^{*}} \scriptstyle{\lx@inpgf@ignorespaces(f_{|U})^{*}-(g_{|U})^{*}} Since (f_{|U})^{*}-(g_{|U})^{*}=0 , we have \mathcal{J}(f,g;R)\subseteq\ker i^{*}=\mathrm{im\,}j_{U} . Thus, for each element x_{k} of the product x_{0}\smile\cdots\smile x_{n} , there exists an element \widetilde{x}_{k}\in H^{*}(X,U_{k};R) such that j_{U_{k}}(\widetilde{x}_{k})=x_{k} . Consider the following commutative diagram: {\lx@inpgf@ignorespaces{H^{*}(X,U_{0};R)\otimes\cdots\otimes H^{*}(X,U_{n};R)}} {\lx@inpgf@ignorespaces{H^{*}(X,\bigcup_{k=0}^{n}U_{k};R)}} {\lx@inpgf@ignorespaces H^{*}(X;R)\otimes\cdots\otimes H^{*}(X;R)} {\lx@inpgf@ignorespaces H^{*}(X;R)} \scriptstyle{\lx@inpgf@ignorespaces\smile} \scriptstyle{\lx@inpgf@ignorespaces{j_{U_{0}}}\otimes\cdots\otimes{j_{U_{n}}}} \scriptstyle{\lx@inpgf@ignorespaces j_{X}} \scriptstyle{\lx@inpgf@ignorespaces\smile} Since \bigcup_{k=0}^{n}U_{k}=X , we have H^{*}(X,\bigcup_{k=0}^{n}U_{k};R)=H^{*}(X,X;R)=0 . Consequently, \begin{split}x_{0}\smile\cdots\smile x_{n}&=j_{U_{0}}(\widetilde{x}_{0})\smile\cdots\smile j_{U_{n}}(\widetilde{x}_{n})\\ &=j_{X}(\widetilde{x}_{0}\smile\cdots\smile\widetilde{x}_{n})=j_{X}(0)=0.\end{split} Therefore \mathrm{l.c.p.}\,\mathcal{J}(f,g;R)\leq n=\mathrm{H^{\,\begin{picture}(-1.0,-3.0)\circle*{3.0}\end{picture}\ }\mathrm{D}}(f,g;R) , as required. ∎ For the LS-category, taking f=\mathrm{id}_{X} and g=\mathrm{c} (a constant map), we have \mathcal{J}(\mathrm{id}_{X},\mathrm{c};R)=H^{>0}(X;R), so that \mathrm{l.c.p.}\,\mathcal{J}(\mathrm{id}_{X},\mathrm{c};R) coincides with the usual cup-length of X , denoted by \mathrm{l.c.p.}\,(X;R) . This shows that the cohomological category improves the classical cup-length lower bound: \mathrm{l.c.p.}\,(X;R)\leq\mathrm{H^{\,\begin{picture}(-1.0,-3.0)\circle*{3.0}\end{picture}\ }cat}(X;R)\leq\mathop{\mathrm{cat}}(X). For the topological complexity, we have the zero divisors cup-length of X , denoted \mathrm{l.c.p.}\,(\ker\smile;R) , defined as the least integer n\geq 0 such that every product x_{0}\cdot\ldots\cdot x_{n} of n+1 elements in the kernel of the cup product \smile\colon H^{*}(X;R)\otimes H^{*}(X;R)\mathrel{\mathchoice{{}\hbox{$\displaystyle{\meno}$}}{{}\hbox{$\textstyle{\meno}$}}{{}\hbox{$\scriptstyle{\meno}$}}{{}\hbox{$\scriptscriptstyle{\meno}$}}}\mathrel{\mkern-3.0mu}\rightarrow H^{*}(X;R) vanishes. Theorem 1.5 . It holds that \mathrm{l.c.p.}\,\mathcal{J}(f,g;R)=\mathrm{l.c.p.}\,\langle\mathcal{J}(f,g;R)\rangle , where \langle\mathcal{J}(f,g;R)\rangle denotes the ideal generated by \mathcal{J}(f,g;R) . Proof. The inequality \mathrm{l.c.p.}\,\mathcal{J}(f,g;R)\leq\mathrm{l.c.p.}\,\langle\mathcal{J}(f,g;R)\rangle is clear. Let us prove the converse. We show that if there exists x_{1},\ldots,x_{n}\in\left\langle\mathcal{J}(f,g;R)\right\rangle such that x_{1}\smile\cdots\smile x_{n}\neq 0 , then \mathrm{l.c.p.}\,\mathcal{J}(f,g;R)\geq n . Each x_{i}\in\left\langle\mathcal{J}(f,g;R)\right\rangle can be written as a finite sum x_{i}=\sum_{k_{i}}(\alpha_{k_{i}}\smile(f^{*}(y_{k_{i}})-g^{*}(y_{k_{i}}))),\quad y_{k_{i}}\in H^{*}(Y;R),\quad\alpha_{k_{i}}\in H^{*}(X;R). As the cup product is distributive, we have that \begin{split}x_{i}\smile x_{j}&=\left(\sum_{k_{i}}\bigl(\alpha_{k_{i}}\smile(f^{*}(y_{k_{i}})-g^{*}(y_{k_{i}}))\bigr)\right)\smile\left(\sum_{k_{j}}\bigl(\alpha_{k_{j}}\smile(f^{*}(y_{k_{j}})-g^{*}(y_{k_{j}}))\bigr)\right)\\ &=\sum_{k_{i},k_{j}}\Bigl(\left(\alpha_{k_{i}}\smile\alpha_{k_{j}}\right)\smile\left((f^{*}(y_{k_{i}})-g^{*}(y_{k_{i}}))\smile(f^{*}(y_{k_{j}})-g^{*}(y_{k_{j}}))\right)\Bigr).\end{split} Since the product x_{1}\smile\cdots\smile x_{n}=\sum_{k_{1},\ldots,k_{n}}\left(\left(\mathop{\raisebox{-1.29167pt}{\scalebox{1.5}{$\smile$}}}\limits_{i=1}^{\raisebox{3.61664pt}{\scriptsize$n$}}\alpha_{k_{i}}\right)\smile\left(\mathop{\raisebox{-1.29167pt}{\scalebox{1.5}{$\smile$}}}\limits_{i=1}^{\raisebox{3.61664pt}{\scriptsize$n$}}(f^{*}(y_{k_{i}})-g^{*}(y_{k_{i}}))\right)\right) is nonzero, there exists at least one tuple (k_{1},\ldots,k_{n}) whose summand is nonzero. In particular, for this tuple, the product \mathop{\raisebox{-1.29167pt}{\scalebox{1.5}{$\smile$}}}\limits_{i=1}^{\raisebox{3.61664pt}{\scriptsize$n$}}(f^{*}(y_{k_{i}})-g^{*}(y_{k_{i}}))=(f^{*}(y_{k_{1}})-g^{*}(y_{k_{1}}))\smile\cdots\smile(f^{*}(y_{k_{n}})-g^{*}(y_{k_{n}})) is nonzero, and therefore we conclude that \mathrm{l.c.p.}\,\mathcal{J}(f,g;R)\geq n . ∎ Corollary 1.6 . Let \mathbb{F} be a field. Then \mathrm{l.c.p.}\,\mathcal{J}(\pi_{1},\pi_{2};\mathbb{F})=\mathrm{l.c.p.}\,(\ker\smile;\mathbb{F}). Proof. By the Künneth formula, if z is an element of \mathcal{J}(\pi_{1},\pi_{2};\mathbb{F}) , then z=x\otimes 1-1\otimes x for some x\in H^{*}(X;\mathbb{F}) so z\in\ker\smile . Conversely, if z=\sum_{i}a_{i}\otimes b_{i}\in\ker\smile , we have \begin{split}z=\sum_{i}a_{i}\otimes b_{i}&=\sum_{i}a_{i}(1\otimes b_{i}-b_{i}\otimes 1)+(a_{i}\smile b_{i})\otimes 1\\ &=\sum_{i}(a_{i}(1\otimes b_{i}-b_{i}\otimes 1))+\left(\sum_{i}(a_{i}\smile b_{i})\right)\otimes 1\\ &=\sum_{i}a_{i}(1\otimes b_{i}-b_{i}\otimes 1)\in\left\langle\mathcal{J}(\pi_{1},\pi_{2})\right\rangle.\end{split} Therefore, \left\langle\mathcal{J}(\pi_{1},\pi_{2})\right\rangle=\ker\smile and the results follows from Proposition 1.5 . ∎ 1.3. Comparison between cohomological and homological distances We compare the homological and cohomological versions of the distance. We prove that they coincide when coefficients are taken in a field, and we exhibit examples showing that no general inequality holds between them. Theorem 1.7 . Let \mathbb{F} be a field and X a topological space with finitely generated cohomology