Turaev Genus 2 Links

• Turaev genus 2 links are defined as links whose Turaev surface has genus 2, representing a two-step complexity leap beyond alternating links.
• Their classification employs alternating decomposition graphs and canonical templates, establishing five distinct equivalence classes through combinatorial analysis.
• Analysis uses state-circle counts, ribbon graph embeddings, and knot polynomials to connect diagrammatic features with topological invariants and Heegaard splittings.

The Turaev genus measures the topological complexity of a link diagram by associating to each diagram a canonical orientable surface, the Turaev surface, and recording its genus. Links of Turaev genus two ($g_T=2$) possess diagrams for which this surface has genus two, situating them precisely two steps beyond alternating diagrams in the topology of links. The study of Turaev genus two links combines diagrammatic, combinatorial, and topological techniques, and has yielded complete classifications of associated graphs, explicit structural characterizations, and connections to knot polynomials, homological invariants, and Heegaard diagrams.

1. The Turaev Surface and Genus Formula

Given a non-split link diagram $D \subset S^2$ with $c(D)$ crossings, the Turaev surface $F(D)$ is constructed by resolving each crossing into its all-$A$ and all-$B$ Kauffman states, connecting these state circles by saddles at each crossing, and capping off the resulting boundaries with disks. $F(D)$ is a closed, unknotted, orientable surface containing a projection of the link. The genus is given by

$g_T(D) = \frac{c(D) + 2 - |s_A| - |s_B|}{2}$

where $|s_A|, |s_B|$ are the counts of state circles in the all-$A$ and all-$D \subset S^2$0 states, respectively. The Turaev genus $D \subset S^2$1 of a link $D \subset S^2$2 is the minimum $D \subset S^2$3 over all diagrams $D \subset S^2$4 of $D \subset S^2$5 (Champanerkar et al., 2014, Armond et al., 2014).

Equivalently, if $D \subset S^2$6 is the all-$D \subset S^2$7 ribbon graph, $D \subset S^2$8 equals its ribbon genus. This surface is always a Heegaard surface for $D \subset S^2$9, and reverting to $c(D)$0 recovers alternating links.

2. Alternating Decomposition Graphs and Classification

Thistlethwaite's alternating decomposition partitions a diagram into maximally alternating regions separated by non-alternating edges. The associated alternating decomposition graph $c(D)$1 encodes this combinatorics: its vertices correspond to curves bordering these regions, and edges to non-alternating connections (Armond et al., 2015). For a reduced diagram, $c(D)$2 is planar, bipartite, and each vertex has even degree.

The genus of the Turaev surface is recovered graph-theoretically via the "twisted" ribbon embedding $c(D)$3: $c(D)$4 where $c(D)$5 is the number of components, $c(D)$6 the number of vertices, $c(D)$7 edges, and $c(D)$8 faces in $c(D)$9 as an embedded ribbon graph.

The central classification (Theorem 1.4 in (Armond et al., 2015)) asserts: a reduced alternating decomposition graph $F(D)$0 is of Turaev genus two if and only if it is doubled path equivalent to one of five canonical graphs:

1. $F(D)$1 —two disjoint doubled 2-cycles
2. $F(D)$2 —vertex-identified doubled 2-cycles
3. $F(D)$3—two 3-cycles sharing one edge, all edges doubled
4. $F(D)$4—complete graph $F(D)$5 with two nonadjacent edges replaced by doubled paths of length 2
5. $F(D)$6—two $F(D)$7's each containing a doubled path of length 2, joined at the spare edge

Doubled path equivalence allows contraction or extension of pairs of parallel edges through vertices of degree 4. Only these five equivalence classes arise for $F(D)$8.

3. Diagrammatic and Structural Descriptions

The tangle structure of Turaev genus two diagrams is classified by decomposing the link projection into maximally connected alternating tangles separated by non-alternating connections. Seungwon Kim (Kim, 2015) enumerates eight abstract templates (“genus 2 templates”), each representing a canonical arrangement of alternating tangles (ranging from cycles of alternating 2-tangles to configurations involving higher-order tangles such as 3- or 4-tangles and specific patterns of non-alternating connections). Each template supports exactly two non-separating cutting loops intersecting the link diagram in two points, realizing genus 2. This combinatorial template approach is exhaustive and exclusive for prime, connected diagrams.

Explicit construction of minimal genus 2 diagrams within these templates involves choosing alternating tangles and ensuring the counts $F(D)$9 satisfy $A$0. Concrete enumerative examples reveal intricate dependence of the minimal crossing number on the detailed local structure of the tangles.

4. Heegaard Diagrams and Topological Constraints

Every Turaev surface of genus two naturally serves as a genus two Heegaard splitting surface for $A$1 (Armond et al., 2014). The associated triple $A$2, with $A$3 and $A$4 representing handlebody-attaching curves, is adapted so that:

• $A$5 embeds as an alternating diagram cutting $A$6 into disks.
• The number of curves satisfies $A$7, where $A$8 counts cap disks.
• The minimal case $A$9 yields two $B$0- and two $B$1-curves, and $B$2 realizes a quadrangulation of $B$3.

A necessary condition for a pair $B$4 to arise as a Turaev surface is the existence of an essential simple loop in $B$5 meeting $B$6 exactly twice and bounding a disk in one handlebody—a property ensuring that the combinatorics of the surface and diagram obey the Turaev construction (Kim, 2015).

5. Algorithmic and Polynomial Criteria

Several algorithmic and invariant-based methods exist for detecting and verifying Turaev genus two:

• State-Circle Count: Direct computation of $B$7.
• Ribbon Graph Genus: Compute genus of the all-$B$8 ribbon graph $B$9 embedded on $F(D)$0.
• Bollobás-Riordan-Tutte Polynomial: The BR-Tutte polynomial encodes the enumeration of spanning quasi-trees in each genus; maximal exponent gives $F(D)$1 (Champanerkar et al., 2014).
• Jones Polynomial Bounds: For adequate diagrams, $F(D)$2; in general, $F(D)$3 (Goyal et al., 14 Jan 2026).
• Width Bounds: The reduced Khovanov homology and knot Floer homology widths provide lower bounds: $F(D)$4, $F(D)$5.
• Concordance Invariants: Inequalities involving signature $F(D)$6, Ozsváth–Szabó $F(D)$7, and Rasmussen $F(D)$8 invariants furnish additional lower bounds.

For adequate links, verifying $F(D)$9 or relevant homological widths can suffice to establish $g_T(D) = \frac{c(D) + 2 - |s_A| - |s_B|}{2}$0. The Abe–Kishimoto dealternating number $g_T(D) = \frac{c(D) + 2 - |s_A| - |s_B|}{2}$1 further bounds $g_T(D) = \frac{c(D) + 2 - |s_A| - |s_B|}{2}$2.

6. Infinite Families, Examples, and the QCL Conjecture

Numerous infinite families of Turaev genus two links are known:

• 3-strand pretzel links $g_T(D) = \frac{c(D) + 2 - |s_A| - |s_B|}{2}$3 for $g_T(D) = \frac{c(D) + 2 - |s_A| - |s_B|}{2}$4 even, $g_T(D) = \frac{c(D) + 2 - |s_A| - |s_B|}{2}$5
• Non-alternating Montesinos links with four tangles where standard diagrams realize the genus computation
• Specific torus knots, such as $g_T(D) = \frac{c(D) + 2 - |s_A| - |s_B|}{2}$6 with $g_T(D) = \frac{c(D) + 2 - |s_A| - |s_B|}{2}$7, $g_T(D) = \frac{c(D) + 2 - |s_A| - |s_B|}{2}$8, $g_T(D) = \frac{c(D) + 2 - |s_A| - |s_B|}{2}$9
• Connected sums of non-alternating genus one knots, e.g., $|s_A|, |s_B|$0
• The five-twist-region family $|s_A|, |s_B|$1, with specific parameter bounds ensuring $|s_A|, |s_B|$2 but failing adequacy (Goyal et al., 14 Jan 2026)

The Qazaqzeh–Chbili–Lowrance (QCL) conjecture, stating $|s_A|, |s_B|$3 is adequate iff $|s_A|, |s_B|$4, is verified within these families. In particular, for the five-twist-region links satisfying $|s_A|, |s_B|$5, $|s_A|, |s_B|$6, $|s_A|, |s_B|$7, the defect $|s_A|, |s_B|$8 confirms the strict form of the Jones-span bound, with both leading and trailing coefficients of $|s_A|, |s_B|$9 equal to $A$0, and hence these links are not adequate.

Illustrative table of five-twist-region genus 2 links:

Parameters	Crossing Condition	$A$1
(3,3,5,-3,-3)	$A$2 (barely)	17
(3,3,6,-3,-3)	$A$3	18
(4,4,7,-4,-4)	$A$4	23
(3,3,8,-4,-4)	$A$5	22

This family is non-adequate but achieves $A$6 and realizes the expected minimal crossing numbers.

7. Cutting Loops, Surgery, and Genus Reduction

Key to the structure theory is the existence of cutting loops and corresponding surgery operations (Kim, 2015). For every prime non-alternating diagram $A$7, a cutting arc in the plane gives rise to a non-separating cutting loop $A$8 in $A$9 intersecting $D \subset S^2$00 in two points and bounding a disk in a handlebody. Surgery simultaneously on $D \subset S^2$01 and $D \subset S^2$02 replaces $D \subset S^2$03 with a new diagram $D \subset S^2$04 (by band replacement) and reduces the genus $D \subset S^2$05. All genus two diagrams admit exactly two independent such cutting loops; this property underlies both the graph-theoretic and tangle-decomposition classifications.

Corollaries from this circle of ideas yield necessary obstructions for a pair $D \subset S^2$06 (surface, diagram) to arise as a Turaev surface from a planar diagram: there must exist an essential simple loop meeting $D \subset S^2$07 in exactly two points and bounding a disk in a handlebody.

---

Turaev genus two links thus constitute a well-delineated class characterized by their combinatorial decomposition, graph-theoretic invariants, tangle templates, and explicit inequalities relating polynomial and homological data. The convergence of diagrammatic, topological, and algebraic characterizations for $D \subset S^2$08 provides foundational insight for the study of non-alternating links and their invariants. (Armond et al., 2015, Champanerkar et al., 2014, Goyal et al., 14 Jan 2026, Armond et al., 2014, Kim, 2015)