---
title: 'Turaev Genus: Diagrammatic Invariant & Insights'
url: https://www.emergentmind.com/topics/turaev-genus
type: topic
---

# Turaev Genus: Diagrammatic Invariant & Insights

Searching arXiv for recent and foundational papers on Turaev genus.
The **Turaev genus** of a link is a diagrammatic and topological invariant defined from the **Turaev surface** associated to a link diagram. It measures how far a link is from being alternating: for a non-split link \(L\), one has \(g_T(L)=0\) if and only if \(L\) is alternating [1406.1945]. The invariant is constructed from the all-\(A\) and all-\(B\) Kauffman states of a diagram, and then minimized over all diagrams of the link. Because the projection becomes alternating on the Turaev surface, the invariant occupies a central position between planar diagrammatics, polynomial invariants, ribbon graphs, and knot homology theories [1406.1945].

## 1. Definition via the Turaev surface

Let \(D\) be a link diagram. At each crossing one has an \(A\)-resolution and a \(B\)-resolution, and a **state** is any choice of smoothing at every crossing. Write \(s_A(D)\) and \(s_B(D)\) for the all-\(A\) and all-\(B\) states. Turaev’s construction places the all-\(A\) state on one side of the projection sphere and the all-\(B\) state on the other, connects them by saddles at the crossings, and caps the resulting boundary components with disks to obtain a closed orientable surface \(F(D)\) [1703.02506].

For a connected diagram \(D\), the genus of the Turaev surface is
\[
g_T(D)=\frac{1}{2}\bigl(2+c(D)-s_A(D)-s_B(D)\bigr),
\]
where \(c(D)\) is the number of crossings of \(D\) [1703.02506]. More generally, for a diagram with \(k(D)\) split components regarded as a \(4\)-valent graph, one has
\[
g_T(D)=\frac{1}{2}\bigl(2k(D)+c(D)-s_A(D)-s_B(D)\bigr)
\]
[1507.02771]. The **Turaev genus** of a link \(L\) is then
\[
g_T(L)=\min\{g_T(D)\mid D \text{ is a diagram of }L\}
\]
[1406.1945].

The basic structural properties are fundamental. The surface \(F(D)\) is an **unknotted closed orientable surface** in \(S^3\), the diagram \(D\) is alternating on \(F(D)\), and \(g_T(L)=0\) exactly when \(L\) is alternating [1406.1945]. This gives the invariant its standard interpretation as a quantitative measure of non-alternatingness. A plausible implication is that Turaev genus is best viewed not as an isolated diagrammatic statistic, but as the genus cost required to recover alternating behavior.

## 2. Alternating distance and related diagrammatic invariants

The Turaev genus is one of several invariants designed to measure deviation from alternation. A closely related quantity is the **dealternating number** \(dalt(L)\), defined as the minimum number of crossing changes needed to turn some diagram of \(L\) into an alternating diagram. Abe and Kishimoto proved the inequality
\[
g_T(L)\le dalt(L)
\]
[1703.02506]. Thus Turaev genus always provides a lower bound for dealternating number.

The same viewpoint underlies the role of **almost alternating** links. A diagram is almost alternating if one crossing change makes it alternating, and a knot is almost alternating if it admits such a diagram but has no alternating diagram [1404.4967]. Every almost alternating knot has Turaev genus \(1\) [1404.4967]. This places genus one links at the first nontrivial level beyond the alternating class.

Subadditivity under connected sum is also part of the structure:
\[
g_T(K_1\sharp K_2)\le g_T(K_1)+g_T(K_2)
\]
[2010.00031]. For adequate knots, the survey literature records additivity under connected sum and mutation invariance within the adequate class [1406.1945]. More recently, an additivity criterion was obtained using concordance invariants: if a knot \(K\) satisfies
\[
s(K)+\limsup_{n\to\infty}\frac{s_n(K)}{n}\ge 2g_T(K),
\]
then equality holds, and if two knots satisfy this condition, Turaev genus is additive on their connected sum [2010.00031].

These relations show that Turaev genus behaves simultaneously as a diagrammatic invariant, an alternating-distance invariant, and a structural invariant for operations such as connected sum. This suggests that its utility lies partly in the fact that it interfaces with several distinct complexity measures without reducing to any single one of them.

## 3. Polynomial and adequacy characterizations

The earliest and most persistent connection is with the **Jones polynomial**. For a non-split link \(L\), one has
\[
\operatorname{span}V_L(t)\le c(L)-g_T(L),
\]
with equality for adequate links [1406.1945]. Equivalently, for adequate links,
\[
g_T(K)=c(K)-\operatorname{span}V_K(t)
\]
[1406.1945]. This places Turaev genus in direct relation with the failure of the Jones span to reach the crossing number.

Adequacy is defined diagrammatically. A diagram \(D\) is \(A\)-adequate if no crossing has its two \(A\)-resolved arcs lying in the same component of the all-\(A\) state, and \(B\)-adequate is defined similarly; a link is adequate if it admits a diagram that is both \(A\)- and \(B\)-adequate [2507.18855]. Every reduced alternating diagram is adequate, but adequacy is strictly broader than alternation [2507.18855]. Adequate diagrams minimize crossing number, and Abe showed that they also minimize Turaev genus [2507.18855].

A sharp genus-one characterization was obtained in 2025. A non-split link \(L\subset S^3\) is adequate and has \(g_T(L)=1\) if and only if
\[
\operatorname{span} V_L(t)=c(L)-1
\]
[2507.18855]. This refines the classical alternating characterization \(\operatorname{span}V_L(t)=c(L)\) by identifying the first non-alternating adequate level exactly one unit below the crossing number [2507.18855]. The same work derives further consequences: if a non-split link satisfies \(\operatorname{span}V_L(t)=c(L)-1\), then it is not quasi-alternating, has arc index \(\alpha(L)=c(L)\), has Jones diameter \(jd_L=2c(L)\), and the leading and trailing coefficients of \(V_L(t)\) have absolute value one [2507.18855].

Related genus-one results were obtained earlier for the extreme coefficients of the Jones polynomial. If
\[
V_L(t)=a_m t^m+\cdots + a_M t^M
\]
with \(a_m,a_M\neq 0\), then for an almost alternating link or a link of Turaev genus one, either \(|a_m|=1\) or \(|a_M|=1\) [1604.03501]. This provides a practical obstruction: if both extreme coefficients have absolute value greater than \(1\), then the link cannot be almost alternating and cannot have Turaev genus one [1604.03501].

Together these results show that the Jones polynomial does not merely correlate with Turaev genus; in the adequate setting, and especially in genus one, it often detects it sharply.

## 4. Homological and concordance bounds

A second major theme is the relationship between Turaev genus and knot homology theories. The survey literature records the width bounds
\[
w_{KH}(K)\le 1+g_T(K), \qquad w_{HF}(K)\le g_T(K)+1
\]
for reduced Khovanov homology and knot Floer homology, respectively [1406.1945]. These inequalities imply that homological thickness yields lower bounds for Turaev genus.

A more refined perspective comes from spanning-tree and ribbon-graph models. For a diagram \(D\), the Turaev surface genus equals the \(\delta\)-width of the spanning-tree complex:
\[
g(\Sigma_D)=\delta_{\max}(D)-\delta_{\min}(D)
\]
[1002.0898]. Lowrance’s inequality then gives
\[
\delta_{\max}(K)-\delta_{\min}(K)\le g_T(K)
\]
and this was used effectively in the study of torus knots [1703.02506].

The invariant also controls concordance-type quantities. For any knot \(K\),
\[
\left |\tau(K)+\frac{\sigma(K)}{2} \right | \leq g_T(K), \qquad 
\frac{|s(K) + \sigma(K)|}{2} \leq g_T(K), \qquad 
\left |\tau(K) - \frac{s(K)}{2} \right | \leq g_T(K)
\]
[1002.0898]. In particular, if \(g_T(K)=0\), then
\[
2\tau(K)=s(K)=-\sigma(K)
\]
[1002.0898]. This recovers the classical alternating pattern from the vanishing of Turaev genus.

The 2020 work on concordance invariants recast these bounds in a broader framework. A concordance invariant \(\nu\) belongs to a class \(\mathcal{DL}\) if it changes by at most one under an oriented band surgery and satisfies
\[
s_B(D)-n_-(D)-1\le \nu(L)\le 1+n_+(D)-s_A(D)
\]
[2010.00031]. Rasmussen’s \(s\)-invariant, \(\frac{s_n}{1-n}\) for \(n\ge 2\), \(-\sigma\), and invariants of the form \(2\nu_0-\ell+1\) coming from slice-torus link invariants all lie in \(\mathcal{DL}\) [2010.00031]. The central theorem is
\[
\frac12\,|\mu(K)-\nu(K)|\le g_T(K)
\qquad\text{for any } \mu,\nu\in\mathcal{DL}
\]
[2010.00031]. A notable consequence is that these bounds can be nontrivial for quasi-alternating knots, unlike many earlier lower bounds [2010.00031].

This homological and concordance picture establishes Turaev genus as a bridge invariant: it is defined diagrammatically, but it is often most effectively bounded or detected by Floer-theoretic, Khovanov-theoretic, and concordance data.

## 5. Low-genus structure and classification

Low Turaev genus imposes rigid diagrammatic structure. One approach uses **alternating decomposition graphs**. Starting from a diagram \(D\), one marks non-alternating edges and constructs the alternating decomposition \(\{\gamma_1,\dots,\gamma_k\}\), then defines a graph \(G\) whose vertices correspond to the curves \(\gamma_i\) and whose edges correspond to non-alternating edges of \(D\) [1507.02771]. The graph is planar, bipartite, and every vertex has even degree [1507.02771].

A central theorem states that if two diagrams have isomorphic alternating decomposition graphs, then they have the same Turaev genus [1507.02771]. Equivalently, the Turaev genus is determined by the graph \(G\). The same paper gives a recursive algorithm based on deletion/contraction of parallel edges and contraction at degree-two vertices, and proves that for every non-negative integer \(k\), there are only finitely many doubled path equivalence classes of reduced alternating decomposition graphs with Turaev genus \(k\) [1507.02771].

In genus one, the reduced alternating decomposition graphs are exactly the doubled cycles of even length, equivalently graphs doubled path equivalent to \(C_2^2\) [1507.02771]. In genus two, there are exactly five doubled path equivalence classes of reduced alternating decomposition graphs [1507.02771].

A complementary classification uses **cutting arcs** and **cutting loops** on the Turaev surface. For a prime non-alternating diagram, there exists a cutting arc; every cutting arc determines a cutting loop; and surgery along it reduces Turaev genus by one:
\[
g_T(D')=g_T(D)-1
\]
[1507.02918]. Using this machinery, prime connected diagrams with \(g_T(D)=1\) are classified as **cycles of alternating \(2\)-tangles**, and prime genus-two diagrams fall into eight alternating-tangle structures [1507.02918].

These classifications lead to further consequences. Every non-split Turaev genus one link has a diagram that is either \(A\)-adequate, \(B\)-adequate, or almost alternating [1812.11387; 2211.10009]. For prime non-split inadequate links of Turaev genus one, the link is almost-alternating [1507.02918]. A plausible implication is that genus one is not merely the first nonzero value of the invariant; it is a highly constrained regime with multiple equivalent-looking structural descriptions.

## 6. Computations, examples, and open phenomena

Explicit computations form a substantial part of the theory. For torus knots with five or fewer strands, exact values and near-exact bounds are known. Among the main formulas are
\[
g_T(T_{4,4n+1}) = 2n,\qquad g_T(T_{4,4n+3})=2n+1,
\]
\[
g_T(T_{5,5n+1}) = 4n,\qquad g_T(T_{6,6n+1}) = 6n,
\]
and
\[
4n+j-2 \le g_T(T_{5,5n+j}) \le 4n+j-1 \qquad (j=2,3,4)
\]
[1703.02506]. Corresponding dealternating bounds show that in these families the dealternating number exceeds the Turaev genus by at most one or two [1703.02506]. The proofs combine lower bounds from knot Floer homology with explicit braid diagrams whose Turaev surfaces realize the bounds [1703.02506].

For low-crossing knots, explicit almost alternating constructions have resolved many genus-one cases. Among non-alternating knots with \(n\le 12\) crossings in KnotInfo, Turaev genus was unknown for \(191\) knots; \(154\) were shown to be almost alternating, hence have Turaev genus \(1\), leaving \(35\) unresolved in that work [1404.4967].

Genus one also exhibits strong homological rigidity. If \(L\) is a non-split Turaev genus one link, then its Khovanov homology is \(\mathbb{Z}\) in at least one extremal quantum grading [1812.11387]. The 2022 refinement showed that for a nonsplit Turaev genus one link, a particular near-extremal summand is trivial, which leads to computations of the Rasmussen \(s\)-invariant and bounds on the smooth four-genus for certain Turaev genus one knots [2211.10009].

Not all structures associated with minimal-genus diagrams are genuine knot invariants. Champanerkar and Kofman had asked whether the polynomial \(q(G;t)\) counting quasi-trees of the all-\(A\) ribbon graph of a minimal Turaev genus diagram is an invariant of the knot. A counterexample using two minimal-genus diagrams of \(8_{21}\) shows the answer is negative: one diagram yields
\[
q(G;t)=6t+21,
\]
while the other yields
\[
q(G;t)=24t+9
\]
[1407.3259]. This demonstrates that even in the minimal-genus setting, ribbon-graph data can depend on the chosen diagram rather than only on the underlying knot.

A current direction links Turaev genus to arc presentations. For a non-split prime link \(L\), the conjecture
\[
c(L)+2-\alpha(L)\ge 2g_T(L)
\]
has been verified for alternating links, links with Turaev genus one, adequate links, closures of positive \(3\)-braids, torus links, and most Kanenobu knots [2509.21274]. For adequate links, equality holds:
\[
c(L)+2-\alpha(L)=2g_T(L)
\]
[2509.21274]. This suggests a broader organizing principle in which Turaev genus mediates between crossing number, arc index, adequacy, and near-alternating structure.

Across these developments, the recurring pattern is that Turaev genus is simultaneously computable, structurally rigid in low genus, sensitive to homological thickness, and subtle enough that diagram-dependent ribbon-graph refinements can fail to descend to knot invariants. That combination explains its central role in the contemporary study of alternating distance.

Source: https://www.emergentmind.com/topics/turaev-genus