Monodromy Ideal Triangulation

• Monodromy Ideal Triangulation is a combinatorial structure for once-punctured surface bundles that encodes the monodromy via layered ideal tetrahedra.
• The construction employs standard transvections and tetrahedron layering to ensure the fiber is realized as an embedded spun-normal surface.
• An explicit algorithm verifies minimality and compatibility through inflation, shortening, and crushing, supported by normal-surface enumeration and computational tools.

A monodromy ideal triangulation is an ideal triangulation of a once-punctured surface bundle, most classically a once-punctured torus bundle over $S^1$, whose combinatorics are determined by the bundle monodromy. In the punctured-torus case, it is the layered ideal triangulation obtained by factoring the monodromy $A\in SL(2,\mathbb Z)$ into standard transvections and layering one tetrahedron per letter of the word (Jaco et al., 2018). In Bryant’s 2025 treatment, the same notion is placed in a broader algorithmic framework: starting from an arbitrary ideal triangulation of a one-cusped fibered 3-manifold, one modifies the triangulation and then crushes it back to an ideal triangulation in which the fiber is realized as an embedded spun-normal surface; in the once-punctured torus case, the resulting combinatorial edge-pairings encode the same monodromy matrix and recover the monodromy ideal triangulation (Bryant, 27 May 2025).

1. Ambient manifolds and bundle structure

The ambient manifolds under consideration are orientable, compact, irreducible, $\partial$-irreducible 3-manifolds $M$ whose single boundary component is a torus. The standing hypotheses also require $M$ to be atoroidal and acylindrical, so that it carries no essential embedded torus or annulus. Many such manifolds occur as once-punctured surface bundles

$M \cong F \times [0,1]/(x,0)\sim(h(x),1),$

where $F$ is a compact surface with one boundary circle and $h\in \mathrm{Aut}(F)$ is the monodromy; in this picture $F$ embeds as an incompressible, $\partial$-incompressible surface in $A\in SL(2,\mathbb Z)$0. Conversely, any atoroidal, acylindrical, irreducible manifold with torus boundary which admits a properly embedded once-punctured surface as a fiber is automatically of this form (Bryant, 27 May 2025).

This class is important because it is simultaneously constrained enough for explicit combinatorial constructions and broad enough to include hyperbolic one-cusped manifolds. Bryant recalls that a classical result of Thurston, later supplemented by Agol–Wise, implies that any hyperbolic once-cusped manifold virtually fibers. That observation situates monodromy ideal triangulations at the interface of bundle structures, ideal triangulations, and normal-surface theory.

2. Construction from the monodromy word

For a hyperbolic once-punctured torus bundle with monodromy $A\in SL(2,\mathbb Z)$1, where hyperbolic means $A\in SL(2,\mathbb Z)$2, the monodromy ideal triangulation begins from the standard ideal triangulation $A\in SL(2,\mathbb Z)$3 of the once-punctured torus: a two-triangle cellulation with one ideal vertex and three edges $A\in SL(2,\mathbb Z)$4, corresponding to the three slopes on the torus. One factors a conjugate of $A\in SL(2,\mathbb Z)$5 as a word in the standard transvections

$A\in SL(2,\mathbb Z)$6

so that

$A\in SL(2,\mathbb Z)$7

Each letter $A\in SL(2,\mathbb Z)$8 or $A\in SL(2,\mathbb Z)$9 prescribes layering one tetrahedron onto the current punctured-torus triangulation along the corresponding edge. After one tetrahedron has been layered for each letter, the top punctured torus is identified back to the bottom one by the original mapping class $\partial$0, completing the ideal triangulation of the total 3-manifold (Jaco et al., 2018).

Combinatorially, this produces a one-cusped ideal triangulation with $\partial$1 tetrahedra. The final triangulation has three edge-classes, all incident to the single ideal vertex, and each ideal edge is an orbit of one of the three slopes under the layered face-gluings. The bottom pair of faces of each tetrahedron is glued into the previous fiber, the top pair forms the next fiber, and the terminal gluing by $\partial$2 closes the bundle. All edges have even degree; in fact, the degree of an edge equals twice its translation distance under $\partial$3 in the Farey graph. The paper on minimal ideal triangulations further states that these triangulations are canonical Epstein–Penner decompositions and hence geometric ideal triangulations, although they are not regular (Jaco et al., 2018).

3. Ideal triangulations, normal coordinates, and spun-normal surfaces

An ideal triangulation $\partial$4 of $\partial$5 is a union of abstract tetrahedra $\partial$6 with orientation-reversing face pairings $\partial$7, so that the identification space is $\partial$8 minus its ideal vertex: $\partial$9 In any triangulation, a normal disc in a tetrahedron is either a triangle, meeting three edges, or a quadrilateral, meeting four edges. A normal surface meets each tetrahedron in a finite collection of such discs, matching across faces, and is encoded by its disc-count vector

$M$0

subject to the standard matching equations and the quadrilateral-admissibility condition that at most one quadrilateral type occurs in each tetrahedron (Bryant, 27 May 2025).

In an ideal triangulation one also allows infinitely many normal triangles accumulating on the boundary torus. The resulting properly embedded surface is a spun-normal surface. Bryant, following Kang–Rubinstein, records that a spun-normal surface is determined purely by its quadrilateral counts

$M$1

which satisfy the $M$2-matching equations

$M$3

where $M$4 is the sense of quadrilateral type $M$5 around edge $M$6 (Bryant, 27 May 2025).

The decisive obstruction is Walsh’s theorem. If $M$7 is atoroidal, acylindrical, irreducible with torus boundary and $M$8 is any ideal triangulation, then every properly embedded, two-sided, incompressible, $M$9-incompressible surface $M$0 is $M$1-isotopic to a spun-normal surface, except exactly when $M$2 is isotopic to a fiber or a virtual fiber. Thus ordinary spun-normal-surface methods miss the fiber in an arbitrary ideal triangulation; the monodromy ideal triangulation construction is designed to remove precisely that exceptional case (Bryant, 27 May 2025).

4. Bryant’s inflation–shortening–crushing construction

Bryant gives an explicit algorithm for a 3-manifold with a single boundary component that fibers over $M$3 with fiber $M$4. The input is an arbitrary ideal triangulation $M$5 of $M$6 with one ideal vertex, and the output is a new ideal triangulation $M$7 in which $M$8 is isotopic to a spun-normal surface. The construction proceeds in five stages (Bryant, 27 May 2025).

First, $M$9 is inflated to a material-boundary triangulation $M \cong F \times [0,1]/(x,0)\sim(h(x),1),$0 of $M \cong F \times [0,1]/(x,0)\sim(h(x),1),$1 by Jaco–Rubinstein inflations. This introduces band tetrahedra, crossing tetrahedra, and branch-point tetrahedra arranged along a chosen frame spine in the boundary-linking torus of $M \cong F \times [0,1]/(x,0)\sim(h(x),1),$2. The result has exactly one vertex and two unglued boundary faces triangulating $M \cong F \times [0,1]/(x,0)\sim(h(x),1),$3. Second, $M \cong F \times [0,1]/(x,0)\sim(h(x),1),$4 is shortened by special 2–3 Pachner moves across quadrilateral blocks, reducing the number of band and crossing tetrahedra and producing a short inflation of boundary-length $M \cong F \times [0,1]/(x,0)\sim(h(x),1),$5. Third, the fiber is located inside the short inflation as a normal surface: by Tollefson–Wang and Jaco–Sedgwick, any fiber in a knot exterior appears among the vertex solutions of the normal-surface cone, so one enumerates vertex-solution surfaces and selects the one whose boundary class agrees with $M \cong F \times [0,1]/(x,0)\sim(h(x),1),$6. Fourth, one performs at most two additional 2–3 Pachner moves, described as site-swaps at band–branch interfaces, followed by further shortening moves, so that $M \cong F \times [0,1]/(x,0)\sim(h(x),1),$7 becomes compatible with the short inflation and uses at most one quadrilateral type per band tetrahedron. Fifth, one crushes $M \cong F \times [0,1]/(x,0)\sim(h(x),1),$8 along its material boundary $M \cong F \times [0,1]/(x,0)\sim(h(x),1),$9 back to an ideal triangulation $F$0. Because $F$1 is now compatible, the crushing sends $F$2 to a spun-normal surface $F$3 in $F$4, and the quadrilateral vector $F$5 of $F$6 is exactly the projection of the normal-coordinate vector of $F$7 in $F$8 onto the subspace of tetrahedra coming from $F$9. By Theorem 3.8 of Bryant’s paper, $h\in \mathrm{Aut}(F)$0 is properly embedded in $h\in \mathrm{Aut}(F)$1 and isotopic to $h\in \mathrm{Aut}(F)$2 (Bryant, 27 May 2025).

This construction answers affirmatively Cooper–Tillmann–Worden’s Question 5.1: every once-punctured surface bundle with one cusp admits an ideal triangulation in which the fiber is realized as an embedded spun-normal surface, and there is a concrete, terminating algorithm to construct it.

5. Monodromy encoding and the once-punctured torus case

In the special case of a once-punctured torus bundle, $h\in \mathrm{Aut}(F)$3 may be chosen to be the layered triangulation determined by the continued fraction, or equivalently by the factorization of the monodromy $h\in \mathrm{Aut}(F)$4 into flips. Bryant describes this in terms of two tetrahedra $h\in \mathrm{Aut}(F)$5 with explicit face-pairings, and states that the monodromy matrix

$h\in \mathrm{Aut}(F)$6

dictates how the triangulation of $h\in \mathrm{Aut}(F)$7 at level $h\in \mathrm{Aut}(F)$8 is glued to that at level $h\in \mathrm{Aut}(F)$9. After inflating and crushing back, one recovers an ideal triangulation whose combinatorial edge-pairings encode the same matrix $F$0; in this precise sense the output may be viewed as the monodromy ideal triangulation (Bryant, 27 May 2025).

Bryant’s worked example is the once-punctured torus bundle of slope $F$1, the trefoil complement. Starting from the 2-tetrahedron layered triangulation of $F$2trefoil with isomorphism signature “cPcbbbadu”, one chooses a frame $F$3 of length $F$4 on the vertex-linking torus. Inflating along $F$5 adds $F$6 band tetrahedra $F$7, $F$8 branch tetrahedra $F$9, and one crossing tetrahedron $\partial$0, yielding an 8-tetrahedron inflation $\partial$1. Shortening moves at faces $\partial$2 produce a short inflation of length $\partial$3. Enumerating vertex normal surfaces identifies surface $\partial$4 as the once-punctured torus fiber $\partial$5; a site-swap at face $\partial$6 in tetrahedron $\partial$7 and one final shortening move make $\partial$8 compatible. Crushing $\partial$9 along $A\in SL(2,\mathbb Z)$00 yields an ideal triangulation $A\in SL(2,\mathbb Z)$01 in which the spun-normal quadrilateral vector is

$A\in SL(2,\mathbb Z)$02

Bryant states that one checks directly that every edge-matching equation $A\in SL(2,\mathbb Z)$03 is satisfied, so $A\in SL(2,\mathbb Z)$04 is the desired monodromy spun-normal surface (Bryant, 27 May 2025).

6. Minimality, complexity, and computational verification

A central structural result for monodromy ideal triangulations of hyperbolic once-punctured torus bundles is minimality. The 2018 study of minimal ideal triangulations develops a topological lower bound for any ideal triangulation of a cusped 3-manifold $A\in SL(2,\mathbb Z)$05: for any rank $A\in SL(2,\mathbb Z)$06 subgroup $A\in SL(2,\mathbb Z)$07, the number of tetrahedra satisfies

$A\in SL(2,\mathbb Z)$08

Here the authors construct three canonical normal surfaces $A\in SL(2,\mathbb Z)$09, dual to the nonzero classes in $A\in SL(2,\mathbb Z)$10, each meeting each tetrahedron in either a single quad or a single triangle, and define the $A\in SL(2,\mathbb Z)$11-taut norm $A\in SL(2,\mathbb Z)$12 by minimizing the negative Euler characteristic. In the case of equality, all edges have even degree, each $A\in SL(2,\mathbb Z)$13 meets every tetrahedron in exactly one quadrilateral, and there are no degree-3 edges except those forced by $A\in SL(2,\mathbb Z)$14. For the rank-2 subgroup $A\in SL(2,\mathbb Z)$15 arising from the three core loops on the fiber, equality holds for the monodromy triangulation; combined with the covering-trick of Section 5.2, this shows that every monodromy ideal triangulation of a hyperbolic once-punctured torus bundle is globally minimal in the sense of Matveev complexity (Jaco et al., 2018).

The geometric lower bound $A\in SL(2,\mathbb Z)$16, with $A\in SL(2,\mathbb Z)$17, also applies, but the same paper emphasizes that monodromy triangulations are not built out of regular ideal tetrahedra. Edges can have arbitrarily large degree, some tetrahedra can have very small hyperbolic volume, and the gap between $A\in SL(2,\mathbb Z)$18 and $A\in SL(2,\mathbb Z)$19 can therefore be arbitrarily large (Jaco et al., 2018).

Bryant’s algorithm adds an explicit computational layer to this structural picture. The number of tetrahedra in the final ideal triangulation satisfies

$A\in SL(2,\mathbb Z)$20

with $A\in SL(2,\mathbb Z)$21, giving a linear-in-$A\in SL(2,\mathbb Z)$22 bound. The computational bottleneck is normal-surface enumeration in the short inflation $A\in SL(2,\mathbb Z)$23, which is exponential in the number of tetrahedra. Bryant notes that in practice one uses Burton’s tree-traversal or Regina’s vertex-enumeration, and that the final output can be verified in SnapPy+Tnorm by checking that $A\in SL(2,\mathbb Z)$24 appears among the spun-normal vertex or quadrilateral surfaces of $A\in SL(2,\mathbb Z)$25 (Bryant, 27 May 2025).

A common misconception is that any ideal triangulation of a fibered one-cusped manifold should already exhibit the fiber as spun-normal. Walsh’s theorem shows that the opposite is typical for a fixed arbitrary ideal triangulation: the fiber is precisely the exceptional case. The significance of the monodromy ideal triangulation is therefore not merely that it triangulates the bundle, but that it aligns the combinatorics of the triangulation with the bundle monodromy and the fibered normal-surface structure simultaneously.