Metric (Graph) Bundles

• Metric (graph) bundles are a coarse-geometric framework that organizes a total space over a base with uniformly hyperbolic fibers and quasi-isometric sections.
• The canonical graph bundle discretizes the structure, relying on a flaring condition that forces exponential divergence of quasi-isometric sections along the base.
• They underpin combination theorems and Cannon–Thurston maps, enabling hyperbolicity inheritance and applications in geometric group theory.

Searching arXiv for papers on metric (graph) bundles to ground the article in published work. Metric (graph) bundles are a coarse-geometric framework, introduced by Mj–Sardar and extended in subsequent work, for organizing a total space $X$ over a base $B$ so that the fibers are uniformly hyperbolic and the geometry of $X$ is controlled by quasi-isometric coherence along the base. In the geometric group theory literature, the framework is designed to study hyperbolicity, combination theorems, pullbacks, and boundary phenomena such as Cannon–Thurston maps. A central theme is that hyperbolicity of the total space is governed not only by hyperbolicity of the fibers and the base, but also by a flaring condition expressing exponential divergence of quasi-isometric sections (Krishna et al., 2020).

1. Core definition and geometric structure

A metric bundle $p:X\to B$ consists of geodesic metric spaces $(X,d_X)$ and $(B,d_B)$, together with fibers $X_b=p^{-1}(b)$ carrying the path metric inherited from $X$, and coherence data ensuring that nearby fibers are uniformly quasi-isometric and that there exist uniform quasi-isometric sections. In the setting developed for pullbacks and Cannon–Thurston maps, the fibers are assumed uniformly hyperbolic and nonelementary: there are $\delta\ge 0$ and $R\ge 0$ such that any fiber $B$0 is $B$1-hyperbolic and the barycenter map $B$2 is $B$3-coarsely surjective (Krishna et al., 2020).

The relevant large-scale morphisms are quasi-isometries and Lipschitz maps. A map $B$4 is a $B$5-quasi-isometry if

$B$6

and every point of $B$7 lies within $B$8 of $B$9. A map is $X$0-Lipschitz if $X$1 (Krishna et al., 2020).

The graph model is a metric graph bundle. In the Mj–Sardar formulation recalled in later work, a simplicial, surjective, $X$2-Lipschitz map $X$3 between connected metric graphs is an $X$4-metric graph bundle if each fiber $X$5 is a connected subgraph whose inclusion in $X$6 is metrically $X$7-proper, and adjacency in the base is reflected by edges joining corresponding fibers. For adjacent vertices $X$8, the induced map $X$9 is a $p:X\to B$0-quasi-isometry for a uniform $p:X\to B$1 (Halder, 9 Jul 2025).

A metric bundle and a metric graph bundle are related by a quasi-isometric “dictionary”: metric bundles can be approximated by metric graph bundles, with fibers and bases quasi-isometric and the total space quasi-isometric to the approximating graph model. This permits arguments to move between continuous and combinatorial settings without changing the coarse geometry (Halder, 9 Jul 2025).

2. Canonical metric graph bundles and flaring

Associated to a metric bundle $p:X\to B$2 is its canonical metric graph bundle, obtained by discretizing the base and fibers via nets. One chooses a net $p:X\to B$3 with mesh $p:X\to B$4, and for each $p:X\to B$5 a net $p:X\to B$6 with mesh $p:X\to B$7. The vertex set is $p:X\to B$8. Vertical edges connect points in a common fiber within a fixed bound, while horizontal edges connect fibers using the uniform quasi-isometries supplied by bundle coherence. Endowed with the path metric, this graph is a metric graph bundle over the graph induced by $p:X\to B$9 (Krishna et al., 2020).

The decisive geometric condition is flaring. If $(X,d_X)$0 is a geodesic or quasigeodesic in the base and $(X,d_X)$1 are $(X,d_X)$2-quasi-isometric sections, the width function is

$(X,d_X)$3

Flaring requires exponential divergence of distinct sections along $(X,d_X)$4: there exist $(X,d_X)$5, $(X,d_X)$6, and $(X,d_X)$7 such that, whenever $(X,d_X)$8 is sufficiently large,

$(X,d_X)$9

Equivalent formulations permit single-sided growth or a threshold $(B,d_B)$0 beyond which the inequality is imposed (Krishna et al., 2020).

In prior work of Mj–Sardar, flaring is used to derive hyperbolicity of the total space. A key technical contribution of the pullback paper is the appendix theorem showing that flaring is equivalent for a metric bundle and for its canonical metric graph bundle. If $(B,d_B)$1 satisfies flaring with constants $(B,d_B)$2, then the canonical metric graph bundle satisfies flaring with constants depending only on $(B,d_B)$3, the quasi-isometry constants of sections, the fiber hyperbolicity constant, and the discretization parameters; conversely, flaring in the canonical graph model implies flaring in the original metric bundle (Krishna et al., 2020).

This equivalence ensures that combination arguments and hyperbolicity proofs may be carried out interchangeably in the metric and graph settings. A plausible implication is that the graph model is not merely a convenient discretization, but a coarse-geometric avatar of the original bundle.

3. Pullbacks and combination theorems

Given a map $(B,d_B)$4, the pullback of a metric bundle $(B,d_B)$5 is

$(B,d_B)$6

with projection $(B,d_B)$7, $(B,d_B)$8. The fiber over $(B,d_B)$9 is $X_b=p^{-1}(b)$0. The metric on $X_b=p^{-1}(b)$1 is chosen so that vertical fibers inherit the metric from $X_b=p^{-1}(b)$2, while horizontal movement along $X_b=p^{-1}(b)$3 is controlled by the bundle coherence and by the Lipschitz and quasi-isometric constants of $X_b=p^{-1}(b)$4 (Krishna et al., 2020).

If $X_b=p^{-1}(b)$5 is a metric (graph) bundle and $X_b=p^{-1}(b)$6 is a Lipschitz quasi-isometric embedding, then the pullback $X_b=p^{-1}(b)$7 is again a metric (graph) bundle with the same uniform fiber hyperbolicity and nonelementary hypotheses, and any flaring condition on $X_b=p^{-1}(b)$8 pulls back to a flaring condition on $X_b=p^{-1}(b)$9 with uniform constants depending only on the original bundle constants and on the Lipschitz/quasi-isometry constants of $X$0. Under these assumptions, if $X$1 and all fibers are uniformly hyperbolic and nonelementary, then $X$2 is hyperbolic (Krishna et al., 2020).

This pullback theorem belongs to a broader combination-theorem lineage. A distinct extension is the theory of trees of metric bundles, which subsumes both trees of metric spaces and metric bundles. In that framework, a tree of metric bundles is hyperbolic if the fibers are uniformly hyperbolic metric spaces and the base is hyperbolic, barycenter maps for the fibers are uniformly coarsely surjective, edge spaces are uniformly qi embedded in the corresponding fibers, and Bestvina–Feighn’s hallway flaring condition is satisfied (Halder, 2022).

An analogous relative theory replaces hyperbolic fibers by strongly relatively hyperbolic fibers. Under hyperbolicity of the base, uniformly mutually cobounded peripheral structures, uniformly coarsely surjective barycenter maps on coned-off fibers, type-preserving and qi-preserving electrocution hypotheses, flaring of the induced coned-off bundle, and cone-bounded hallways strictly flare, the total space $X$3 is strongly hyperbolic relative to a canonical family of maximal cone-subbundles of horosphere-like spaces (Krishna, 2022).

4. Cannon–Thurston maps and lamination structure

The principal boundary theorem for pullbacks states that if $X$4 is a metric (graph) bundle such that $X$5 is hyperbolic and all fibers are uniformly hyperbolic and nonelementary, and if $X$6 is a Lipschitz quasi-isometric embedding with pullback $X$7, then the natural map $X$8 admits a Cannon–Thurston map: $X$9 is continuous (Krishna et al., 2020).

The proof strategy uses hyperbolicity of $\delta\ge 0$0, ladder and hallway constructions, flaring, and bounded backtracking or stability of geodesics in hyperbolic spaces. In the relatively hyperbolic setting, the analogous statement is formulated for Bowditch relative boundaries: if $\delta\ge 0$1 is the pullback of a qi-embedded $\delta\ge 0$2, then $\delta\ge 0$3 admits a Cannon–Thurston map between relative boundaries (Krishna, 2022).

For a fiber $\delta\ge 0$4 and its inclusion $\delta\ge 0$5, the Cannon–Thurston lamination is

$\delta\ge 0$6

If $\delta\ge 0$7, the directional sublamination $\delta\ge 0$8 is defined by restricting to pairs mapped to the endpoint of a quasi-isometric lift in $\delta\ge 0$9 of a geodesic or quasigeodesic ray in $R\ge 0$0 converging to $R\ge 0$1. The paper records several structural properties: $R\ge 0$2; both $R\ge 0$3 and $R\ge 0$4 are closed in $R\ge 0$5; leaves from distinct directions are coarsely transverse; directional laminations vary continuously with $R\ge 0$6; and under the pullback hypotheses, $R\ge 0$7 for all $R\ge 0$8 (Krishna et al., 2020).

A later development concerns surjectivity of fiber Cannon–Thurston maps. For metric graph bundles over $R\ge 0$9 with hyperbolic total space, the boundary map $B$00 is surjective in two main settings: when the fibers are uniformly quasi-isometric to a fixed nonelementary hyperbolic group and have uniformly bounded valence, and when the fibers are one-ended, proper hyperbolic metric spaces with uniformly coarsely surjective barycenter maps. A further criterion proves surjectivity under uniform exponential growth of the fibers (Halder, 9 Jul 2025).

5. Group-theoretic realizations and applications

A basic source of metric graph bundles is a short exact sequence of hyperbolic groups

$B$01

Here the metric bundle arises from the short exact sequence, with base $B$02, fiber $B$03, and total space $B$04. If $B$05 is quasi-isometrically embedded and $B$06, then $B$07 is hyperbolic and the inclusion $B$08 admits the Cannon–Thurston map $B$09 (Krishna et al., 2020).

A second class of applications concerns complexes of hyperbolic groups. If $B$10 is a developable complex of nonelementary hyperbolic groups over a finite simplicial complex $B$11, the face groups are nonelementary hyperbolic, face maps are isomorphisms onto finite-index subgroups, the fundamental group $B$12 is hyperbolic, and $B$13 is a good subcomplex in the sense recorded in the pullback theorem, then $B$14 is hyperbolic and the inclusion $B$15 admits a Cannon–Thurston map $B$16 (Krishna et al., 2020).

In the relatively hyperbolic setting, short exact sequences of pairs yield analogous results. If

$B$17

is a short exact sequence of relatively hyperbolic groups satisfying the cusp-preserving and flaring hypotheses recorded in the relative combination theorem, then $B$18 is strongly relatively hyperbolic relative to $B$19. If $B$20 is qi embedded in $B$21 and $B$22, then $B$23 is strongly relatively hyperbolic and its inclusion in $B$24 admits a Cannon–Thurston map (Krishna, 2022).

These applications show that metric (graph) bundles provide a common geometric mechanism behind hyperbolicity inheritance, subgroup combination theorems, and boundary maps.

6. Assumptions, limitations, and other meanings of the term

The geometric-group-theoretic theory depends on a rigid package of hypotheses: uniform hyperbolicity and nonelementarity of fibers, existence of uniform quasi-isometric sections, flaring, and, for pullbacks, Lipschitz quasi-isometric embeddings of bases. Properness is needed in several lamination statements. The dependence of hyperbolicity constants on flaring and quasi-isometry data is qualitative rather than optimized. The literature also emphasizes that Cannon–Thurston maps for arbitrary inclusions of hyperbolic subgroups can fail, with Baker–Riley counterexamples serving as the standard warning that structural hypotheses such as flaring are substantive rather than technical (Krishna et al., 2020).

The phrase “metric graph bundle” also appears in other literatures with different meanings. The following usages are distinct in definition and objective.

Setting	Basic data	Aim
Geometric group theory	Hyperbolic fibers over a base with quasi-isometric coherence and flaring	Hyperbolicity, combination theorems, Cannon–Thurston maps
Tropical geometry	$B$25-torsors and $B$26-torsors on a metric graph $B$27	Principal bundles, vector bundles, moduli, tropicalization (Gross et al., 2022)
Discrete graph theory	Graph bundles $B$28 with connection $B$29	Construction of S-Ricci-flat graphs and non-product phenomena (Li et al., 2023)

In tropical geometry, a tropical principal $B$30-bundle on a metric graph is a $B$31-torsor, and for $B$32 a rank-$B$33 tropical vector bundle is an $B$34-torsor. That theory develops multidivisors, slope stability, a tropical Weil–Riemann–Roch theorem, a tropical Narasimhan–Seshadri correspondence, and skeleton identifications for Tate-curve moduli (Gross et al., 2022). A later extension formulates tropical reductive groups $B$35 and tropical principal bundles on metric graphs for classical groups and $B$36 (Gross et al., 7 Nov 2025).

In discrete graph theory, graph bundles are specified by a base graph $B$37, a fiber graph $B$38, and a twisting datum $B$39 satisfying $B$40. That notion is used to construct S-Ricci-flat graphs and to prove that, under natural constraints, a non-trivial graph bundle cannot be isomorphic to the Cartesian product of its base and fiber (Li et al., 2023).

In the geometric group theory sense, however, metric (graph) bundles remain chiefly a combination-theorem technology: a framework in which uniformly hyperbolic fibers, quasi-isometric sections, and flaring interact to produce hyperbolic total spaces, stable pullbacks, and continuous—sometimes surjective—boundary extensions.