Precision-Graded Cohomology and Arithmetic Persistence for Network Sheaves

Robert Ghrist, Cassie Ding · 2025

Read Precision-Graded Cohomology and Arithmetic Persistence for Network Sheaves by Robert Ghrist, Cassie Ding (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-2511-00677.txt.

Introduction

Robert Ghrist and Cassie Ding rebuild persistent homology with a new yardstick: precision instead of scale. Classic persistence tracks how topological features are born and die as geometry grows, summed up in barcodes. Working over the p-adic integers, the authors define arithmetic barcodes for network sheaves, where each bar marks the precision level at which a cohomology class stops lifting through finer and finer approximations.

Their central Digit-SNF Dictionary shows hierarchical precision data from maps between successive levels encodes exactly the Smith normal form exponents of the coboundary operator. For rank-one sheaves, loop holonomy fixes bar lengths through p-adic valuation, and stability holds under small shifts that respect precision. Torsion becomes the signal, not a nuisance, with uses from quantized consensus to sensor synchronization.

Catalog record

CREATORRobert Ghrist, Cassie Ding
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE106 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-00677

Browse the full library · About the Tower of Babel

Full text

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

Precision-Graded Cohomology and Arithmetic Persistence for Network Sheaves Robert Ghrist, Cassie Ding Abstract . Persistent homology tracks topological features across geometric scales, encoding birth and death of cycles as barcodes. We develop a complementary theory where the filtration parameter is algebraic precision rather than geometric scale. Working over the p p -adic integers ℤ p \mathbb{Z}_{p} , we define arithmetic barcodes that measure torsion in network sheaf cohomology: each bar records the precision threshold at which a cohomology class fails to lift through the valuation filtration ℤ p ⊇ p ​ ℤ p ⊇ p 2 ​ ℤ p ⊇ ⋯ \mathbb{Z}_{p}\supseteq p\mathbb{Z}_{p}\supseteq p^{2}\mathbb{Z}_{p}\supseteq\cdots . Our central result – the Digit-SNF Dictionary – establishes that hierarchical precision data from connecting homomorphisms between successive mod- p k p^{k} cohomology levels encodes exactly the Smith normal form exponents of the coboundary operator. Bars of length a a correspond to ℤ p / p a ​ ℤ p \mathbb{Z}_{p}/p^{a}\mathbb{Z}_{p} torsion summands. For rank-one sheaves, cycle holonomy (the product of edge scalings around loops) determines bar lengths explicitly via p p -adic valuation, and threshold stability guarantees barcode invariance when perturbations respect precision. Smith normal form provides integral idempotents projecting onto canonical cohomology representatives without geometric structure. Results extend to arbitrary discrete valuation rings, with p p -adic topology providing ultrametric geometry when available. Applications include distributed consensus protocols with quantized communication, sensor network synchronization, and systems where measurement precision creates natural hierarchical structure. The framework repositions torsion from computational obstacle to primary signal in settings where data stratifies by precision. 2020 Mathematics Subject Classification 55N30, 13F30, 15A21 1. Introduction Network sheaves are cellular sheaves [ Cur14 ] with a 1-dimensional base space. These provide a flexible framework for encoding local data and constraints on graphs, with cohomology measuring global obstructions to consistency. Over fields – typically \mathbb{R} , \mathbb{C} , or finite fields – the theory is well-developed: cohomology groups are vector spaces, and when inner products are given, harmonic analysis via Hodge decomposition provides canonical representatives for cohomology classes. We consider the cohomology of network sheaves on finite graphs whose stalks are finite-rank free modules over a discrete valuation ring (DVR) – a principal ideal domain R with unique nonzero prime ideal (\pi) , where \pi is the uniformizer , and with finite residue field R/\pi . The prototypical example of a DVR to which our methods apply is the ring of p -adic integers R=\mathbb{Z}_{p} , where \pi=p and R/\pi=\mathbb{F}_{p} [ AM69 ] . Other examples include formal power series rings \mathbb{F}_{q}[[t]] and rings of integers in finite extensions of \mathbb{Q}_{p} . Over a DVR, cohomology changes character. The group H^{1} generically carries torsion – elements annihilated by powers of \pi – even when its rationalization H^{1}\otimes_{R}K (where K=\mathrm{Frac}(R) ) is free. This torsion is not pathological but rather the essential feature distinguishing integral from rational cohomology. Classical decompositions based on inner products fail because DVRs lack positive-definite pairings: orthogonal complements need not split, and \im d need not be a direct summand of C^{1} . The obstacle is precisely the torsion in H^{1}=C^{1}/\im d . The algebraic structure of DVRs provides natural tools for understanding this torsion. The filtration by powers of \pi stratifies modules by precision: M\supseteq\pi M\supseteq\pi^{2}M\supseteq\cdots , with successive quotients \pi^{k}M/\pi^{k+1}M capturing data at the k -th level of refinement. For the p -adics R=\mathbb{Z}_{p} , this hierarchy has geometric meaning. The p -adic integers carry a natural ultrametric absolute value |x|_{p}=p^{-\mathrm{val}_{\pi}(x)} satisfying the strong triangle inequality |x+y|_{p}\leq\max\{|x|_{p},|y|_{p}\} with equality when |x|_{p}\neq|y|_{p} [ Sch85 , Gou20 ] . Multiplication by p shifts elements one level deeper in this hierarchy, just as multiplying a decimal by 10 shifts digits. The filtration C^{1}\supseteq pC^{1}\supseteq p^{2}C^{1}\supseteq\cdots thus encodes not merely an algebraic tower but a geometric refinement by precision, where each quotient p^{k}C^{1}/p^{k+1}C^{1} represents data at a specific resolution [ Sch85 , Gou20 , Kob84 ] . This precision stratification organizes cohomology hierarchically. Passing from cochains to cohomology, one obtains connecting homomorphisms – the digit maps – between successive levels, whose images measure obstructions to lifting cohomology classes through the filtration. The dimensions of these images encode the complete torsion structure, and the relationship between this hierarchical data and the algebraic invariants from Smith normal form is our central result: the Digit-SNF Dictionary . This all might seem a thin scenario – only H^{0} and H^{1} are of interest for network sheaves. Yet, as H^{0} classifies global sections and H^{1} the obstructions thereunto, there is something nontrivial to be gained, especially for problems of network consensus and local-to-global integration. Our theory will develop with such applications in mind, and a prototypical example will follow at the conclusion of the theory. The paper has three main themes. First, we show that the filtration by powers of \pi organizes cohomology into a hierarchy of precision levels, with connecting homomorphisms between levels whose dimensions encode the complete torsion structure. Second, we prove that this hierarchical information coincides exactly with the algebraic invariants from Smith normal form, establishing a dictionary that makes torsion both conceptually transparent and algorithmically accessible. Third, we show that the Smith normal form exponents define an arithmetic barcode where torsion summands correspond to finite bars measuring precision thresholds, and prove stability results ensuring these barcodes are robust under high-precision perturbations. Section 2 establishes foundations: discrete valuation rings, graphs, and network sheaves with free stalks. Section 3 develops the precision hierarchy through digit sequences and proves spectral sequence collapse for graphs. Section 4 constructs explicit integral decompositions via Smith normal form and proves the Digit-SNF Dictionary. Section 5 develops the arithmetic barcode interpretation, determines stability, and provides explicit formulas via cycle holonomy. Section 6 demonstrates applications to distributed consensus with quantized communication. 1.1. Related Work and Context Network sheaves on graphs and their cohomology and Hodge theory over fields are now well established. Applications of network sheaf cohomology with field coefficients include network coding [ GH11 ] , distributed sensing [ HG19b ] , distributed optimization [ HG19a ] , spectral graph theory [ HG19c ] , opinion dynamics [ HG21 ] , sheaf neural networks [ HG20 , BGC+22 ] , graphic statics [ CGH23 , CG23a ] , protein folding [ Hay+25 ] origami [ CG25 ] , and visual paradoxes [ GC25 ] : in all cases, the network sheaf encodes how local measurements relate across edges, and H^{1} gives obstructions to the global sections captured by H^{0} . While there are examples of network sheaves with non-field coefficients, these have often been either non-algebraic (e.g., set-valued network sheaves for applications to logic [ Gog92 ] , sensing [ Rob17 ] or Reeb graphs [ dSMP16 ] ) or rather specialized and esoteric (such as order lattices [ GR22 , RG22 ] or categories enriched over quantales [ GLN+25 ] ). The investigation of p-adic coefficients for network sheaves is both general and novel. In persistent homology, the standard barcode classification relies on working over a field; the one-parameter classification as a multiset of intervals is a manifestation of quiver-representation theory for the A_{n} quiver over a field [ CdS10 , Oud15 , CB15 ] . Over non-field coefficients such as \mathbb{Z} or more general PIDs, torsion appears

Read the complete text