Insertion Chain Complex: Word Homology

• Insertion Chain Complex is a topological structure that extends insertion graphs by using higher-dimensional cells called valid blocks.
• It encodes compatibility among multiple symbol insertions via an explicit boundary operator, ensuring that ∂² = 0 for its chain complex.
• Its homology groups serve as invariants to detect missing words and measure insertion pattern completeness, with concrete minimal examples and universality results.

The Insertion Chain Complex is a higher-dimensional topological structure attached to a finite set of words $W\subset \Sigma^*$, introduced as a generalization of the insertion graph, in which vertices are words and edges record single-symbol insertions. Its basic purpose is to encode not only isolated insertion relations but also higher-order compatibility among multiple insertions. In the construction of Balchin, Delpeuch, and Vicary, valid higher-dimensional “blocks” serve as cells, the associated chain groups are free abelian groups on those blocks, and the resulting homology groups are proposed as invariants measuring the complexity, non-randomness, or lack of completeness of finite word sets (Jonoska et al., 16 Sep 2025).

1. Conceptual foundation

The starting point is the insertion relation on the free monoid $\Sigma^*$, including the empty word $1$. For words $w,w'\in\Sigma^*$, one writes

$w\sim w'$

if and only if $w'$ is obtained from $w$ by a single symbol insertion, that is, if there exist words $x,y\in\Sigma^*$ and a symbol $a\in\Sigma$ such that

$w=xy \text{ and } w'=xay.$

For a finite word set $\Sigma^*$0, the associated insertion graph $\Sigma^*$1 has vertices $\Sigma^*$2 and edges given by this single-insertion relation. Two basic properties are immediate: the graph is bipartite, since insertion changes word length by $\Sigma^*$3, and it is acyclic as a directed graph, since lengths strictly increase along directed edges (Jonoska et al., 16 Sep 2025).

The insertion chain complex extends this graph-theoretic picture by filling higher-dimensional cells whenever several insertions are mutually compatible. The motivating principle is that if a word set is “complete” with respect to insertion patterns, then the expected higher-dimensional cells should be present and homology should vanish. If intermediate words are missing, insertion paths that should fit together fail to do so, creating holes. In this sense, homology is interpreted as a measure of missing-word complexity.

A central observation is that a 4-cycle in the insertion graph need not have a uniform meaning. Some 4-cycles arise from two commuting insertion operations and should therefore be filled by a square, while others do not correspond to compatible insertions and remain unfilled. The insertion chain complex is designed precisely to distinguish these cases.

2. Blocks, vertices, and validity

The cells of the theory are blocks. Let

$\Sigma^*$4

For each $\Sigma^*$5,

$\Sigma^*$6

with $\Sigma^*$7 factors of $\Sigma^*$8. An element of $\Sigma^*$9 is an $1$0-dimensional block, written

$1$1

where each $1$2 and each $1$3. Thus $1$4, so 0-blocks are words, while 1-blocks represent single insertions.

For an $1$5-block

$1$6

and a subset $1$7, the corresponding vertex is

$1$8

The full vertex set is

$1$9

In particular, the minimal and maximal vertices are

$w,w'\in\Sigma^*$0

An $w,w'\in\Sigma^*$1-block therefore packages all words obtained by independently deciding whether to realize each distinguished insertion.

Because repeated symbols can lead to multiple presentations of the same combinatorial cell, the theory imposes an equivalence relation on blocks. Equivalent blocks have the same vertices, and for valid blocks the converse also holds: the vertex set determines the block uniquely. Each equivalence class has a unique canonical form, characterized by the condition that for each $w,w'\in\Sigma^*$2,

$w,w'\in\Sigma^*$3

Not every formal block is admitted as a cell. The basic degeneracy is exemplified by $w,w'\in\Sigma^*$4, which has only three distinct vertices $w,w'\in\Sigma^*$5. A block in canonical form

$w,w'\in\Sigma^*$6

is declared valid if whenever $w,w'\in\Sigma^*$7 for some $w,w'\in\Sigma^*$8, one has

$w,w'\in\Sigma^*$9

Otherwise it is invalid or degenerate. A key equivalent characterization is that $w\sim w'$0 is invalid if and only if there exists $w\sim w'$1 such that

$w\sim w'$2

The notation $w\sim w'$3 is then reused for the set of valid $w\sim w'$4-blocks (Jonoska et al., 16 Sep 2025).

Faces are described by fixing some insertion slots to be realized and others to be deleted. For disjoint $w\sim w'$5,

$w\sim w'$6

where

$w\sim w'$7

Its dimension is $w\sim w'$8. Codimension-1 faces are called facets. There are two kinds: upper facets $w\sim w'$9, which are always valid, and lower facets $w'$0, which may be invalid and are then omitted in the boundary theory.

3. The chain complex and its boundary

Given $w'$1, the Insertion Block Complex is

$w'$2

Its dimension is

$w'$3

Because faces have smaller vertex sets, this collection is closed under taking faces. Its 1-skeleton is exactly the insertion graph: $w'$4 For finite $w'$5, the complex is finite, and

$w'$6

The resulting chain complex is defined for any face-closed set $w'$7 of valid blocks. If

$w'$8

then

$w'$9

and $w$0 otherwise. For a word set $w$1, the Insertion Chain Complex is

$w$2

Thus $w$3-chains are formal integer combinations of valid $w$4-blocks.

The boundary operator is defined on a valid $w$5-block $w$6, $w$7, by

$w$8

where

$w$9

and

$x,y\in\Sigma^*$0

The published definition is printed with a different sign expression, but the subsequent computations and the proof of $x,y\in\Sigma^*$1 use the indexed formula above. For a 1-block $x,y\in\Sigma^*$2,

$x,y\in\Sigma^*$3

For the 2-block $x,y\in\Sigma^*$4,

$x,y\in\Sigma^*$5

The identity

$x,y\in\Sigma^*$6

is proved by a cancellation argument analogous to the standard simplicial or cubical proof, but with additional case analysis for invalid lower faces. The double boundary separates into upper-after-upper, lower-after-upper, upper-after-lower, and lower-after-lower terms. The first three classes cancel by commutation of face operations, while the lower-lower terms require tracking exactly when lower facets disappear because of invalidity. This establishes that $x,y\in\Sigma^*$7 is a genuine chain complex (Jonoska et al., 16 Sep 2025).

The paper explicitly notes that the resulting object is not simplicial and in general not cubical. In particular, two squares may intersect in two edges, which cannot happen in a cubical complex. It is therefore best understood as a custom face-based chain complex built from valid insertion blocks.

4. Homology and principal theorems

Homology is defined in the standard way: $x,y\in\Sigma^*$8 These homology groups are interpreted as invariants of the word set. If the set is complete with respect to insertion patterns, then the complex should be filled in and higher homology should vanish. Nontrivial homology detects missing intermediate words or incompatible insertion patterns, and higher-dimensional homology detects larger missing families.

One of the first structural results classifies minimal nontrivial 1-cycles. If $x,y\in\Sigma^*$9 is finite and $a\in\Sigma$0 has $a\in\Sigma$1, then $a\in\Sigma$2 only in a very restricted set of cases: up to reflection, symbol permutation, or affixes, the complex must be isomorphic to one of six families $a\in\Sigma$3, namely

$a\in\Sigma$4

$a\in\Sigma$5

$a\in\Sigma$6

$a\in\Sigma$7

$a\in\Sigma$8

$a\in\Sigma$9

This theorem isolates exactly the 4-vertex insertion cycles that survive as nontrivial $w=xy \text{ and } w'=xay.$0-classes.

A second major theorem establishes homological universality. Let $w=xy \text{ and } w'=xay.$1 be a finitely generated abelian group and $w=xy \text{ and } w'=xay.$2. Then there exists a word set $w=xy \text{ and } w'=xay.$3 such that

$w=xy \text{ and } w'=xay.$4

The proof proceeds by passing from finite simplicial complexes to finite cubical complexes and then encoding cubes by words. The key intermediate statement is that every finite cubical complex $w=xy \text{ and } w'=xay.$5 can be represented, at the level of homology, by an insertion chain complex.

The word encoding used in that theorem is explicit. Vertices $w=xy \text{ and } w'=xay.$6 are sent to

$w=xy \text{ and } w'=xay.$7

and an elementary cube

$w=xy \text{ and } w'=xay.$8

is sent to a block

$w=xy \text{ and } w'=xay.$9

where

$\Sigma^*$00

This correspondence preserves dimensions, vertices, and boundaries.

The theory also yields sharp minimality results for sphere homology. Writing $\Sigma^*$01 for the smallest cardinality of a word set $\Sigma^*$02 whose insertion chain complex has the same homology as $\Sigma^*$03, the paper proves

$\Sigma^*$04

and for $\Sigma^*$05,

$\Sigma^*$06

The values for $\Sigma^*$07 are sharp.

A strong vanishing theorem complements these realizability results. For words $\Sigma^*$08, define the interval

$\Sigma^*$09

If $\Sigma^*$10 embeds uniquely into $\Sigma^*$11, then $\Sigma^*$12 is connected and

$\Sigma^*$13

This theorem formalizes the intuition that when all intermediate words between a minimal and maximal word are present and the embedding structure is unique, no higher holes remain (Jonoska et al., 16 Sep 2025).

5. Examples and computational aspects

Several examples illustrate how the construction behaves on small word sets. For

$\Sigma^*$14

the insertion block complex contains the listed 0-blocks

$\Sigma^*$15

the 1-blocks

$\Sigma^*$16

and the 2-blocks

$\Sigma^*$17

This example shows a finite complex with two compatible insertion squares but no higher cells.

The minimal nontrivial 1-cycle occurs already for

$\Sigma^*$18

Its insertion graph is a 4-cycle with no filling square, so $\Sigma^*$19. This minimality is forced by bipartiteness: no 3-cycles can occur in the insertion graph.

A basic 2-dimensional example is

$\Sigma^*$20

Here the resulting insertion chain complex has the homology of a 2-sphere: $\Sigma^*$21 This 8-word example is the source of the sharp value $\Sigma^*$22.

The complex also records the effect of completing missing words. The set

$\Sigma^*$23

forms a type-1 nontrivial 4-cycle. If it is enlarged to

$\Sigma^*$24

the former cycle becomes filled and the corresponding insertion complex has no cycles. This is the clearest combinatorial manifestation of the interpretation of homology as detecting missing words.

The paper emphasizes that a maximal block need not capture all relations in the full insertion complex on its vertices. For

$\Sigma^*$25

$\Sigma^*$26 is the maximal block in $\Sigma^*$27, but the full complex contains an extra edge

$\Sigma^*$28

joining $\Sigma^*$29 and $\Sigma^*$30, which is not itself an edge of the single block $\Sigma^*$31. Thus the insertion block complex is genuinely richer than the combinatorics of any one maximal block.

From a computational standpoint, the worst-case size bound

$\Sigma^*$32

already indicates exponential growth. The proof of $\Sigma^*$33 used a SageMath implementation. The workflow was to generate all possible 1-skeleta on $\Sigma^*$34 vertices, generate all possible 2-dimensional complexes on those graphs with nontrivial second homology, discard complexes containing forbidden local patterns of overlapping squares, and analyze the remaining cases manually. One key obstruction is that two distinct squares cannot share the same two upper edges (Jonoska et al., 16 Sep 2025).

6. Terminological scope and related constructions

The exact term Insertion Chain Complex is introduced for finite sets of words and their associated homological invariants. Related literature uses “insertion” and “chain” in other senses, but not with the same topological meaning. The distinction is important because several adjacent constructions can be mistaken for the same object.

In the language-modeling literature, the phrase “insertion chain” appears in a completely different sense. A continuous-time Markov chain framework for insertion LLMs casts reverse-time generation as an insertion-only CTMC on variable-length token sequences. That work develops an “insertion chain” view of sequence generation, but the chain is a stochastic process on sequences, not a graded chain complex with a boundary operator (Patel et al., 8 Jun 2026).

In rewriting theory, “Chinese syzygies by insertions” develops a finite convergent semi-quadratic presentation of the Chinese monoid, with rewriting rules defined by insertion algorithms and syzygies given by relations among insertion algorithms. The paper explicitly notes that it does not introduce an object literally named an insertion chain complex, but it provides the data from which a polygraphic resolution or Squier-type homological complex is naturally built (Hage et al., 2019).

In the combinatorics of partitions and graphs, “Insertion and Lie Bracket Concerning Finite Sets” defines insertion, composition, and Lie bracket operations on ordered partitions, and transports them to ordinary graphs, Feynman diagrams, and Kontsevich admissible graphs. It does not define a graded chain complex with a differential $\Sigma^*$35 satisfying $\Sigma^*$36; its main output is insertion-based algebraic structure rather than homology (Zhou, 2021).

In algebraic knot concordance, Powell’s work on the second order algebraic concordance group repeatedly uses mapping cones, algebraic gluing, zero-surgery attachment, and algebraic surgery on symmetric Poincaré chain complexes. Those are chain-level operations that insert, attach, or replace one complex by another, but the papers do not define an object with the exact name “Insertion Chain Complex” (Powell, 2012).

Against that background, the 2025 word-theoretic construction is distinctive in three respects. First, it gives a precise higher-dimensional cell structure built from valid blocks. Second, it supplies an explicit signed boundary operator and proves $\Sigma^*$37. Third, it interprets the resulting homology groups as invariants of finite word sets. A plausible implication is that the insertion chain complex occupies a specific niche at the intersection of combinatorics on words and algebraic topology: it is neither merely an insertion graph nor merely an insertion-based algebraic operation, but a genuine homological object built from insertion compatibility.