Arc Crossing Change in Knot Theory

• Arc crossing change is a local operation on knot or link diagrams that exchanges over/under labels at arc endpoints, altering crossing structure.
• Type I acts as a genuine crossing change to fully unknot links, while Type II preserves parity, introducing constraints in unknotting processes.
• The operation has applications in alternating diagrams and connects with grid homology, enriching the taxonomy of local unknotting moves.

Searching arXiv for papers on arc crossing change and related knot-theoretic crossing-change operations. Arc crossing change is a local operation on a knot or link diagram that modifies crossing information at the endpoints of an arc in the underlying diagram. In the formulation extended from knots to links by Cheng–Liao–Song, an arc is a connected segment of the underlying $4$-valent graph whose endpoints lie at undercrossings, and the operation exchanges the over/under labels at those endpoints. When the two endpoints are distinct crossings, the move changes both crossings simultaneously; when they coincide at a single crossing, two non-equivalent versions arise, called arc crossing change I and arc crossing change II (Cheng et al., 10 Sep 2025). Within this framework, the central questions concern when arc crossing change is an unknotting operation, how admissible sets of crossings are characterized, and how these moves interact with diagram classes such as alternating knots.

1. Definition of the local move

Let $L \subset \mathbb R^2$ be a connected link diagram with crossing set

$C(L)=\{c_1,\dots,c_N\}.$

An arc $\alpha \subset L$ is a connected segment of the underlying $4$-valent graph whose endpoints lie at undercrossings of $L$. If $\alpha$ has distinct endpoints at two crossings $p \neq q$, then performing an arc crossing change along $\alpha$ is the local move which exchanges the over/under labels at both $p$ and $L \subset \mathbb R^2$0 (Cheng et al., 10 Sep 2025).

The case in which both endpoints of $L \subset \mathbb R^2$1 coincide at a single crossing $L \subset \mathbb R^2$2 is the source of the two standard variants. In arc crossing change I, one performs exactly one ordinary crossing change at $L \subset \mathbb R^2$3. In arc crossing change II, one formally changes $L \subset \mathbb R^2$4 twice, and the crossing is therefore unchanged (Cheng et al., 10 Sep 2025). The distinction is purely local but has global consequences for unknotting behavior, especially for links with parity obstructions.

A subset $L \subset \mathbb R^2$5 is called arc crossing change admissible if there exists a finite sequence of arc crossing changes, of the chosen type, each supported on some arc of $L \subset \mathbb R^2$6, such that exactly the crossings in $L \subset \mathbb R^2$7 are switched (Cheng et al., 10 Sep 2025). This admissibility notion provides a combinatorial way to ask which prescribed crossing changes can be realized through arc-based local moves rather than by independent ordinary crossing changes.

2. Type I and Type II arc crossing changes

The two versions differ only for arcs whose endpoints coincide, but that difference is decisive. Type I behaves as a genuine crossing-changing move even in the coincident-endpoint case, whereas Type II is inert there (Cheng et al., 10 Sep 2025). The literature summarized in Cheng–Liao–Song therefore treats them as distinct unknotting operations on link diagrams.

The following table records the defining difference.

Situation	Arc crossing change I	Arc crossing change II
Endpoints at distinct crossings $L \subset \mathbb R^2$8	Exchanges over/under labels at both $L \subset \mathbb R^2$9 and $C(L)=\{c_1,\dots,c_N\}.$0	Exchanges over/under labels at both $C(L)=\{c_1,\dots,c_N\}.$1 and $C(L)=\{c_1,\dots,c_N\}.$2
Endpoints coincide at one crossing $C(L)=\{c_1,\dots,c_N\}.$3	Performs exactly one ordinary crossing change at $C(L)=\{c_1,\dots,c_N\}.$4	Formally changes $C(L)=\{c_1,\dots,c_N\}.$5 twice, hence leaves $C(L)=\{c_1,\dots,c_N\}.$6 unchanged

This distinction explains the contrasting behavior of standard examples. For the Hopf link, choosing the short arc whose both ends meet the same crossing flips that crossing in Type I and turns the Hopf link into the unlink, while in Type II the same arc is only doubly changed and the Hopf link is unchanged (Cheng et al., 10 Sep 2025). The example is not merely pedagogical: it exhibits the precise locus where the two theories diverge.

Cericola’s earlier work is described as having shown that on a knot diagram any ordinary crossing change or “1-dimensional” arc-change (semi-arc) operation is an unknotting move, and Cheng–Liao–Song extend his arc crossing change from knots to arbitrary links (Cheng et al., 10 Sep 2025). This suggests that arc crossing change is best understood as part of a broader taxonomy of local unknotting operations organized by the dimensionality and support of the move.

3. Unknotting results for link diagrams

For Type I, the principal result is unconditional. If $C(L)=\{c_1,\dots,c_N\}.$7 is any connected link diagram, then every crossing $C(L)=\{c_1,\dots,c_N\}.$8 is arc crossing change I-admissible. Equivalently, by a finite sequence of Type I arc crossing changes one can transform $C(L)=\{c_1,\dots,c_N\}.$9 into an ascending diagram, hence into the unlink (Cheng et al., 10 Sep 2025). In this sense, Type I is always an unknotting operation on connected link diagrams.

The proof proceeds component by component. One fixes an ordering $\alpha \subset L$0, first uses Type I moves on each under-strand of $\alpha \subset L$1 to convert all its inter-component undercrossings into overcrossings, thereby making $\alpha \subset L$2 “on top,” and then applies Cericola’s Gauss-diagram argument to turn $\alpha \subset L$3 into an ascending knot by Type I moves only on its self-arcs. The same procedure is repeated inductively on $\alpha \subset L$4 (Cheng et al., 10 Sep 2025). The argument combines global control of component ordering with local arc operations confined to prescribed strands.

For Type II, the situation is subtler and governed by total linking parity. Writing

$\alpha \subset L$5

Cheng–Liao–Song prove the following trichotomy (Cheng et al., 10 Sep 2025):

1. If $\alpha \subset L$6 is even, then Type II moves suffice to unknot $\alpha \subset L$7.
2. If $\alpha \subset L$8 is odd and at least one component of $\alpha \subset L$9 has a self-crossing, then Type II moves still unknot $4$0.
3. If $4$1 is odd and no component has a self-crossing, then all components are pairwise linked once and $4$2 can only be reduced to an unlink plus a single Hopf link.

A key structural input is Lemma 3.2, which shows that Type II moves preserve the parity of each $4$3, hence preserve $4$4 (Cheng et al., 10 Sep 2025). Thus the difference between Type I and Type II is not cosmetic: Type II carries an intrinsic parity invariant that can obstruct complete unlinking.

The proof strategy isolates components without self-crossings and with exactly one undercrossing, called “type C” components; these are precisely the arcs on which Type I and Type II differ (Cheng et al., 10 Sep 2025). One then inductively pulls components to the top or bottom until they become simple unknots, at which point parity determines whether a residual Hopf-link obstruction remains. A plausible implication is that Type II is naturally adapted to parity-sensitive link simplification rather than universal unknotting.

4. Admissibility phenomena in alternating knot diagrams

A separate result addresses not full unknotting but the realizability of prescribed pairs of crossing changes. If $4$5 is an alternating knot diagram with at least two crossings, then for any two distinct crossings $4$6, the pair $4$7 is arc crossing change admissible for Cericola’s version (Cheng et al., 10 Sep 2025). This is Theorem 4.1 of Cheng–Liao–Song and gives a strong positive admissibility statement for alternating diagrams.

The proof passes through the shadow of the diagram. One forgets the over/under information to obtain a $4$8-valent planar graph $4$9, then re-orients each edge so that at every vertex two edges point in and two point out; this is possible exactly because the knot diagram is alternating (Cheng et al., 10 Sep 2025). One then chooses an in-edge at $L$0, uses a directed-graph argument to find a directed trail from that edge to $L$1 that does not pass again through $L$2, and interprets the resulting trail in the knot diagram as a sequence of arcs turning only at undercrossings or going straight at overcrossings. Such a trail is called an admissible $L$3 trail, and Lemma 4.4 implies that its existence makes $L$4 arc crossing change admissible (Cheng et al., 10 Sep 2025).

This theorem links arc crossing change to planar graph orientation and directed-trail methods. It also separates alternating knots as a class with particularly strong admissibility behavior. The paper explicitly notes an open question: to find a complete necessary-and-sufficient combinatorial condition on a knot diagram under which any two given crossings may be simultaneously realized by Cericola-type arc crossing changes (Cheng et al., 10 Sep 2025). That question places admissibility, rather than mere unknotting, at the center of the theory’s future development.

5. Representative examples and obstructions

Several examples clarify how the move behaves in practice. Cericola’s original trefoil example uses a single Type I move along a long arc running under two crossings; swapping its endpoints simultaneously changes both crossings and yields the unknot (Cheng et al., 10 Sep 2025). This demonstrates that the operation can package multiple crossing changes into one local arc-supported move.

The Hopf link furnishes the standard contrast between Types I and II. Under Type I, a short arc with both ends at the same crossing flips that crossing and turns the Hopf link into the unlink; under Type II, the same configuration leaves the diagram unchanged (Cheng et al., 10 Sep 2025). This is the minimal example exhibiting the parity rigidity later abstracted by the total-linking criterion.

A $L$5-component link with $L$6 shows that Type II can still unknot nontrivial links when the parity obstruction vanishes in the appropriate way: if one component has two undercrossings, then two successive Type II moves along those arcs each switch one undercrossing, and overall one can reduce $L$7 by $L$8 and unknot (Cheng et al., 10 Sep 2025). Conversely, a $L$9-component link with odd total linking but one component knotted illustrates the role of self-crossings: Type II moves on the knotted component inject enough flexibility to unknot the whole link, by first introducing a self-crossing and then cancelling it (Cheng et al., 10 Sep 2025).

These examples make clear that the theory is not solely about whether a move changes a crossing. The decisive parameters are the interaction between endpoint coincidence, self-crossing structure, and the parity of pairwise linking. Common misconceptions therefore arise if Type I and Type II are conflated or if results for knots are transferred uncritically to links.

6. Relation to crossing-change maps in filtered grid homology

Arc crossing change is a diagrammatic local move, whereas Kendall’s work on crossing-change maps in filtered grid homology studies algebraic maps induced by a single crossing change between knots represented by grid diagrams (Kendall, 2023). The two subjects are distinct, but they are closely adjacent in the broader study of crossing modifications and unknotting complexity.

In Kendall’s framework, if $\alpha$0 and $\alpha$1 are two grids of the same size representing knots $\alpha$2 and $\alpha$3 differing by a single crossing change realized as a cross-commutation of two adjacent columns, then one defines chain maps

$\alpha$4

by counting empty pentagons with exactly one distinguished vertex at $\alpha$5 or $\alpha$6 (Kendall, 2023). These maps are bona fide chain maps; $\alpha$7 is homogeneous of degree $\alpha$8, while $\alpha$9 has degree $p \neq q$0, matching multiplication by $p \neq q$1 (Kendall, 2023). Their compositions are filtered-chain-homotopic to multiplication by powers of $p \neq q$2, and concatenating such maps along unknotting sequences yields a combinatorial formulation of the invariant $p \neq q$3, satisfying $p \neq q$4 the Gordian distance and in particular $p \neq q$5 (Kendall, 2023).

The relevance to arc crossing change is indirect but conceptually important. Arc crossing change organizes certain simultaneous or constrained crossing modifications at the diagrammatic level, while filtered grid homology packages single crossing changes into chain maps and numerical invariants. This suggests a broader perspective in which local crossing operations may be studied both combinatorially, through admissibility and diagram moves, and homologically, through induced maps on knot Floer-type complexes. The current data do not state a direct homological model for arc crossing change itself; however, the proximity of the two theories indicates a natural interface between local move classification and Floer-theoretic unknotting bounds.

7. Position within the study of local unknotting operations

The available results place arc crossing change within a broader program of classifying local operations on knots and links. Cheng–Liao–Song explicitly situate their work beyond Cericola’s arc crossing change and Kinuno’s semi-arc moves, as contributing to the classification of local operations of “$p \neq q$6-, $p \neq q$7-, $p \neq q$8-dimensional” type that are or are not unknotting moves on knots and links (Cheng et al., 10 Sep 2025). Within that program, Type I and Type II furnish a controlled comparison between a universally unknotting arc-based move and a parity-constrained variant.

Several structural themes emerge. First, the extension from knots to links introduces linking parity as an essential obstruction for Type II (Cheng et al., 10 Sep 2025). Second, alternating diagrams admit especially strong admissibility statements through directed-planar-graph methods (Cheng et al., 10 Sep 2025). Third, the distinction between local diagram modification and algebraic crossing-change formalism, exemplified by filtered grid homology, shows that crossing operations can be studied simultaneously as moves on planar diagrams and as morphisms between chain complexes (Kendall, 2023).

The current state of the subject therefore consists of a precise local definition, a complete unknotting characterization for the two standard arc-crossing-change types on connected link diagrams, a strong admissibility theorem for alternating knot diagrams, and an explicit set of open combinatorial questions about which prescribed crossing sets can be realized by arc-supported moves (Cheng et al., 10 Sep 2025). This suggests that arc crossing change now occupies a well-defined place in contemporary knot theory: as a local move whose subtle endpoint behavior generates nontrivial distinctions between knots and links, between unconditional and parity-obstructed unknotting, and between diagrammatic and homological approaches to crossing modification.