Strand Diagrams in Topology & Computation

• Strand diagrams are a family of diagrammatic formalisms that encode combinatorial, topological, algebraic, categorical, and molecular data using interconnected strands.
• They use local operations such as concatenation, smoothing, and reduction to transform visual representations into algebraic or computational objects, with applications in Floer homology and contact topology.
• Advanced strand diagram techniques underpin constructions in mapping class groups, quiver potentials, and strand-displacement calculi, bridging theoretical frameworks with practical applications in DNA computing and beyond.

Strand diagrams are diagrammatic formalisms in which strands, wires, or arcs encode combinatorial, topological, algebraic, categorical, or molecular data. The term does not designate a single universal construction. In combinatorial tangle Floer homology it denotes planar encodings of partial permutations interacting with an oriented tangle decomposition; in bordered Floer and contact topology it denotes generators of strand algebras built from monotone strand maps; in Thompson-like and asymptotically rigid mapping class groups it denotes forest-and-braid pictures reduced by dipoles; in knot theory it has a decorated trivalent variant called a trace diagram; and in DNA theory it denotes both literal diagrams of multi-stranded complexes and a rigid string-diagram calculus for noncrossing Watson–Crick pairings (Petkova et al., 2016, Mathews, 2016, Genevois et al., 2021, Nelson et al., 2017, Nowicka et al., 2023, Ortiz-Muñoz, 12 May 2026).

1. Terminological scope and recurrent structure

Several major uses of the term can be organized by the type of data carried by the strands and the operations allowed on them.

Domain	Basic strand-diagram object	Characteristic operations
Tangle Floer homology	Black strands on an oriented orange tangle background	Concatenation, smoothing, endpoint exchange
Strand algebras and contact categories	Monotone strand maps between marked places	Concatenation, crossing resolution, homology
Chambord groups	Triple $(A,\beta,B)$ of forests and a braid	Dipole reduction, vertical concatenation
Trace-diagram skein theory	Crossings plus signed trivalent traces	Recursive expansion, trace moves
DNA secondary structures	Typed noncrossing planar partial matchings	Tensoring, bending, zip-and-transfer

In the tangle Floer setting, the algebra $\mathcal A_P$ is generated by partial bijections $[n]\to[n]$, equivalently by strand diagrams consisting of black strands connecting $(0,i)\to(1,\phi(i))$ against horizontal orange strands determined by a sign sequence $P$ (Petkova et al., 2016). In the arc-diagram strand algebra, an unconstrained strand map is a triple $\mu=(S,T,\phi)$ with $\phi:S\to T$ a nondecreasing bijection, drawn as strands between marked places on an interval (Mathews, 2016, Mathews, 2018). In braided diagram-group theory, a braided strand diagram over $(\mathcal P,w)$ is a triple $(A,\beta,B)$, where $A,B$ are $\mathcal A_P$0-forests and $\mathcal A_P$1 induces a label-preserving bijection between leaves (Genevois et al., 2021). In the categorical DNA setting, a morphism $\mathcal A_P$2 is an isotopy class of typed noncrossing planar partial matchings in a rectangle, built from through-strands and Watson–Crick-typed source and target arcs (Ortiz-Muñoz, 12 May 2026).

This suggests that “strand diagram” is best understood as a family of related diagrammatic languages. What remains stable across these usages is the role of strands as carriers of interface data together with a local calculus—concatenation, reduction, isotopy, smoothing, or transfer—that turns pictures into algebraic or computational objects.

2. Floer-theoretic and contact-algebraic strand diagrams

In combinatorial tangle Floer homology, the starting datum is a sign sequence

$\mathcal A_P$3

The sequence is drawn as $\mathcal A_P$4 horizontal orange strands, oriented left-to-right if $\mathcal A_P$5 and right-to-left if $\mathcal A_P$6. Generators of the algebra $\mathcal A_P$7 are partial bijections, equivalently black-strand diagrams with minimal crossings and no triple points. Multiplication is concatenation subject to forbidden-crossing rules, and the differential is defined by smoothing one crossing at a time, followed by Reidemeister II simplifications. The algebra is bigraded by Maslov and Alexander gradings defined by crossing counts, and the tangle module is built from sequences of partial bijections through a decomposition $\mathcal A_P$8 into elementary tangles. Its structure maps $\mathcal A_P$9 are all diagrammatic: smoothing black-black crossings, introducing black-black crossings, exchanging endpoints, adding a left-boundary contribution, and concatenating on the right by algebra elements (Petkova et al., 2016).

The bordered Floer/contact-theoretic strand algebra places the same picture in a more abstract arc-diagram setting. For an arc diagram $[n]\to[n]$0, the unconstrained strand algebra $[n]\to[n]$1 is generated by strand maps $[n]\to[n]$2 with $[n]\to[n]$3. Multiplication is

$[n]\to[n]$4

so concatenation is permitted only when endpoints match and no excess inversions are created. Idempotents are horizontal strand diagrams $[n]\to[n]$5. After imposing the matching $[n]\to[n]$6, one obtains $[n]\to[n]$7-constrained and symmetrised strand diagrams, and the full algebra is

$[n]\to[n]$8

Its differential resolves crossings, and the homology $[n]\to[n]$9 is represented by crossingless symmetrised $(0,i)\to(1,\phi(i))$0-constrained strand diagrams under explicit support and endpoint conditions (Mathews, 2016).

The contact-geometric significance is that $(0,i)\to(1,\phi(i))$1 is isomorphic to the contact category algebra $(0,i)\to(1,\phi(i))$2. In that dictionary, arc diagrams correspond to quadrangulated surfaces, idempotents correspond to basic dividing sets, strand diagrams correspond to contact structures, and multiplication of strand diagrams corresponds to stacking of contact structures. The local classification of homology generators near each matched pair mirrors the local classification of tight cubes in a cubulated $(0,i)\to(1,\phi(i))$3-manifold, and multiplication in homology vanishes when two factors share a used step, matching the overtwistedness of the corresponding stacked contact structure (Mathews, 2016).

The $(0,i)\to(1,\phi(i))$4 refinement extends this picture from ordinary multiplication to higher composition laws. The strand algebra $(0,i)\to(1,\phi(i))$5 is a differential graded algebra, so its homology inherits $(0,i)\to(1,\phi(i))$6-operations $(0,i)\to(1,\phi(i))$7 and maps $(0,i)\to(1,\phi(i))$8 via Kadeishvili’s construction. The paper organizes these operations using local types—tight, twisted, crossed, sublime, critical, and singular—and proves explicit vanishing and nonvanishing criteria. In particular, nonzero $(0,i)\to(1,\phi(i))$9 requires precise local criticality conditions, while creation operators insert crossings to invert differentials on twisted local summands. The contact interpretation is that higher operations can reorder bypass data and realize higher-order compositions of contact structures that are invisible to ordinary gluing (Mathews, 2018).

3. Braided, symmetric, and quotient generalizations

A different line of development replaces planar partial permutations by forests decorated with braids. For an arboreal semigroup presentation

$P$0

a braided strand diagram over $P$1 is a triple $P$2, where $P$3 are $P$4-forests with matching leaf and interior-vertex counts and $P$5 is a braid inducing a label-preserving bijection between the leaves. The key local move is a braided dipole reduction: a pair of corresponding interior vertices whose children are all leaves and are matched in order under the braid can be collapsed to a single wire when the connecting strand is parallel to the wires joining the children. Every braided strand diagram is equipotent to a unique reduced one, and multiplication is defined by vertical concatenation after enlarging diagrams by dipole insertions until the middle forests match. The resulting group is the Chambord group $P$6, with identity $P$7 and inverse obtained by turning a diagram upside down and inverting the braid (Genevois et al., 2021).

This diagrammatic calculus is not only combinatorial. The paper identifies Chambord groups with asymptotically rigid mapping class groups of punctured arboreal surfaces, constructs a CAT(0) cube complex $P$8 from clopen braided strand diagrams, and proves that polycyclic subgroups are virtually abelian and undistorted. Vertex stabilisers are braid groups $P$9, and the action on the cube complex controls subgroup geometry (Genevois et al., 2021).

A separate quotient construction appears in the theory of orbifold diagrams. Here one starts with a $\mu=(S,T,\phi)$0-symmetric Postnikov diagram on a disk and takes the quotient by rotation. The resulting orbifold diagram lives on a disk with $\mu=(S,T,\phi)$1 boundary marked points and one orbifold point $\mu=(S,T,\phi)$2 of order $\mu=(S,T,\phi)$3. Its strands satisfy the usual alternating-crossing conditions, but self-crossings and double-crossing loops are governed by winding-number constraints around $\mu=(S,T,\phi)$4. The paper proves that, for sufficiently large $\mu=(S,T,\phi)$5, orbifold diagrams are exactly quotients of $\mu=(S,T,\phi)$6-symmetric Postnikov diagrams, and it associates to each orbifold diagram a quiver with potential whose Jacobian algebra is Morita equivalent to a skew-group algebra built from the covering Postnikov diagram. In the Grassmannian case, this Jacobian algebra is realized as the endomorphism algebra of an explicit cluster-tilting object, extending the Baur–King–Marsh framework from ordinary disks to disks with orbifold points (Baur et al., 2020).

These constructions show that strand-diagram techniques persist under substantial enrichment. Braids add mapping-class-theoretic content, while rotational quotients add orbifold and skew-group structure. The basic diagrammatic logic—local reduction, global composition, and invariance under controlled moves—remains intact.

4. Trace diagrams and skein-theoretic recursion

In knot theory and biquandle-bracket theory, the closest analogue of a strand diagram is the trace diagram. A trace diagram is a planar diagram consisting of ordinary oriented crossings together with signed traces, which are degree-$\mu=(S,T,\phi)$7 vertices with one dashed, unoriented edge and two oriented edges. There are two kinds: Type A traces connect two parallel oriented pass-through vertices, and Type B traces connect a bivalent sink to a bivalent source. Each trace carries a sign $\mu=(S,T,\phi)$8 or $\mu=(S,T,\phi)$9, recording the sign of the crossing that was expanded (Nelson et al., 2017).

The purpose of trace diagrams is recursive evaluation. Instead of expanding every crossing simultaneously into a full state sum, one expands one crossing at a time and leaves behind a signed trace that preserves the coloring data. For an $\phi:S\to T$0-colored trace diagram $\phi:S\to T$1, the recursion stops when no crossings remain, and then

$\phi:S\to T$2

where $\phi:S\to T$3 is the number of negative traces, $\phi:S\to T$4 the number of positive traces, and $\phi:S\to T$5 the number of components obtained after deleting all traces. The paper proves that this recursive evaluation agrees with the original biquandle bracket invariant, so trace diagrams repackage the same invariant in a skein-style form (Nelson et al., 2017).

The essential technical issue is whether ordinary strands may be moved over, under, or through traces. The paper analyzes $\phi:S\to T$6 oriented trace moves. Over-crossing trace moves are allowed exactly when the bracket is over-adequate, meaning the coefficients satisfy

$\phi:S\to T$7

together with a family of equalities among products of $\phi:S\to T$8- and $\phi:S\to T$9-coefficients. Under-crossing trace moves are allowed exactly when the bracket is under-adequate, with a different family of coefficient identities. If both hold, the bracket is adequate. For monochromatic crossings, the formalism simplifies further: the bracket satisfies a Homflypt-style skein relation, and pass-through trace moves are allowed exactly when

$(\mathcal P,w)$0

Trace diagrams therefore preserve the traditional skein-theoretic strategy of local expansion and simplification while incorporating biquandle colorings and new stopping rules (Nelson et al., 2017).

5. DNA computation, rendering, and strand-displacement calculi

In DNA computing, strand diagrams are literal models of nucleic-acid complexes and reactions. One influential restricted formalism studies top-nicked double strands, denoted ndsDNA, together with mobile two-domain single strands of the form $(\mathcal P,w)$1 or $(\mathcal P,w)$2, where $(\mathcal P,w)$3 is a toehold and $(\mathcal P,w)$4 is a recognition domain. The central invariant is that every double-stranded species remains a top-nicked double strand except for fleeting branch-migration transients. Within this discipline, the paper defines transducer, fork, and join gates, such as

$(\mathcal P,w)$5

The formal framework, called nick algebra, provides syntax, structural congruence, and reduction rules for strand-displacement steps, together with may-correctness $(\mathcal P,w)$6 and will-correctness $(\mathcal P,w)$7. The paper explicitly observes that populations of fork and join gates are equivalent to Petri nets (Cardelli, 2010).

A distinct but complementary problem is the automatic rendering of multi-stranded DNA complexes. At the domain level, a strand is a sequence of domains, and a species is a multiset of strands with complementary pairings. The paper converts such a species into a graph $(\mathcal P,w)$8 whose vertices are domain boundaries, whose solid edges are domains within a strand, and whose dashed edges are domain pairings between strands. A species is planar iff its graph $(\mathcal P,w)$9 is planar, and planarity can be checked in linear time using standard algorithms such as Boyer–Myrvold. If $(A,\beta,B)$0 is planar, the method computes a straight-line plane drawing in linear time; if $(A,\beta,B)$1 is non-planar, it uses a topology-shape-metrics pipeline with planarisation, orthogonalisation, compaction, force-directed refinement, and simulated annealing. The method minimizes topological depth using an SPQR-tree-based embedding algorithm and, among minimum-depth embeddings, chooses one with maximum external face size. The result is a 2D strand diagram that keeps strands connected, keeps pairings visible, and handles arbitrarily pseudoknotted multi-strand DNA species (Nowicka et al., 2023).

The rendering framework also clarifies a persistent misconception: “nested” and “planar” are not equivalent for multi-strand DNA. The paper states that, for multi-strand DNA, a species is pseudoknotted when no linear ordering of strands makes the pairings nested, but its notion of planar DNA species is broader than “nested” species and includes some pseudoknotted species as long as their graph is planar. This distinction is made precise using book embeddings: some planar graphs need more than $(A,\beta,B)$2 pages, but any planar graph can be embedded on at most $(A,\beta,B)$3 pages (Nowicka et al., 2023).

6. Categorical DNA secondary structures and related protocol abstractions

A categorical formulation turns DNA strand diagrams into the morphisms of a rigid monoidal category. In the strict pivotal monoidal category $(A,\beta,B)$4, objects are DNA sequences, that is finite words over $(A,\beta,B)$5, with tensor product given by concatenation and unit the empty word. A morphism $(A,\beta,B)$6 is an isotopy class of typed noncrossing planar partial matchings in a rectangle whose top boundary is labeled by $(A,\beta,B)$7 and bottom boundary by $(A,\beta,B)$8. The allowed local pieces are through-strands, source arcs, and target arcs, with the typing constraint that every arc connects Watson–Crick complements $(A,\beta,B)$9 and $A,B$0. Duality is reverse complementation, evaluation and coevaluation are canonical duplex pairings, and the snake identities hold by planar isotopy: $A,B$1 A central bending correspondence identifies

$A,B$2

so generalized elements $A,B$3 are exactly non-pseudoknotted secondary structures on $A,B$4. Composition, after bending, becomes the zip-and-transfer operation

$A,B$5

followed by planar isotopy and loop removal. The paper explicitly frames toehold-mediated strand displacement as a kinetically specific instance of this operation (Ortiz-Muñoz, 12 May 2026).

A related, but distinct, use of “strand” appears in the strand-space model of cryptographic protocols. There a strand is a linearly ordered sequence of nodes, each representing a transmission, a reception, or a state-synchronization event labeled by a fact. A bundle is a finite directed acyclic graph with strand edges $A,B$6 for local order and communication arrows $A,B$7 for message delivery. To analyze fair exchange, the model adds mutable state as a multiset of facts updated by rewrite rules and explicit progress assumptions about resilient channels and non-stopping principals. The paper gives a strand-diagram intuition in which vertical strands represent local timelines, arrows between strands represent message delivery, and state events are shown as synchronization points. In this usage, the strand is not a geometric arc but a local behavior line in a global execution structure (0910.4342).

Taken together, these categorical and protocol-theoretic developments show that strand diagrams are not merely heuristic pictures. In some areas they are the morphisms of a rigid monoidal category; in others they are the basic units of a stateful execution semantics. This suggests that the deepest common feature of the notion is compositionality: a strand diagram packages local connectivity or local behavior so that global structure can be built by concatenation, gluing, isotopy, reduction, or state synchronization.