Alphabeta Math
Pipeline-generated
How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

  • Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
  • AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
  • AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

✓ 7 results · all verified · 1 also independently AI-judged
Every result on this page is machine-checked by a proof checker and read in full and owner-audited; the judge is an additional, independent cross-model AI review of the proofs. The 6 not AI-judged were verified by owner audit (typically over a confirmed judge false positive), not failures.

Artin Presentation Completeness and Braid Combing

1 · Prerequisites

2 · Summary

This page proves that the Artin presentation is not merely a presentation that surjects onto the geometric braid group: the surjection φn ⁣:BnArtin→Bngeom that sends each abstract generator to the class of the elementary geometric half twist is an isomorphism. The proof is the braid-combing argument. The page first fixes the Zariski combing words αi=σiσi+1⋯σn−1 and xi=αi+1−1σi2αi+1, then shows that any word tracing the trivial braid can be rewritten, by free insertion of the cancelling blocks αjαj−1 alone, as a product of combing factors, one per letter, each factor being the record of where the last strand enters and leaves that letter. The six possible shapes of a combing factor then reduce, using only the two Artin relations and free cancellation, to either a lower-rank letter σ1,…,σn−2 or one of the words xi±1, and the conjugation table for the lower-rank letters moving past an x-letter collects the whole word into the standard form W≡W1W2 with W1 in the x-letters and W2 in the first n−2 generators.

The second half of the page identifies the geometric content of that normal form. The words xi are conjugate, inside the free kernel of the forgetting map PBn→PBn−1, to the standard pure generators Ain, and a left-inverse computation in the free group shows that the classes φn(x1),…,φn(xn−1) form a free basis of that kernel; consequently a combed trivial word has both factors trivial, the left one by free cancellation of x-pairs and the right one by the induction hypothesis on the number of strands. Induction on n then proves completeness of the Artin presentation. Because the free-kernel input is the Axiom-of-Choice-dependent Fadell--Neuwirth sequence of the companion pure-braid page, the completeness theorem carries the Axiom of Choice, declared explicitly where it is used. The closing corollary composes the completeness theorem with the published configuration-space isomorphism and the boundary-fixed mapping-class isomorphism, tracking the generator through all four classical models of the braid group.

3 · Logical flowchart

4 · Definitions, theorems and proofs

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (gpt-6.1-sol)audited 2026-10-02Open item page →

The Zariski combing words alpha_i and x_i in the Artin presentation

Definition

Fix n≥2 and let Bn=⟨σ1,…,σn−1⟩ be the abstract braid group of The braid group by Artin presentation, the group presented by the generators σ1,…,σn−1 and the relations

σiσi+1σi=σi+1σiσi+1(1≤i≤n−2),

σiσj=σjσi(1≤i,j≤n−1, ∣i−j∣>1).

Reading convention for words. A word w1w2⋯wm in this alphabet and its inverses is read first letter first: it denotes the product [w1][w2]⋯[wm] in Bn, and under the published stacking convention [γ][β]:=[γ⋆β] of Stacking of geometric braids is a well-defined associative operation on isotopy classes and The isotopy classes of geometric braids based at Q form a group, and the endpoint permutation is a homomorphism, in which the first factor of a stacking is the upper one, its geometric image is the stacking w1⋆w2⋆⋯⋆wm whose first factor is the topmost layer. Here the geometric image is taken through the published surjection φ ⁣:Bn→Gn, φ(σi)=[σi] of The Artin presentation surjects onto the geometric braid group. In particular the empty word is the identity of Bn. No injectivity or completeness of φ is asserted, and nothing below depends on how many factors the stacking has.

The words αi. For 1≤i≤n−1 put

αi:=σiσi+1⋯σn−1∈Bn,andαn:=1,

the last being the empty word. Each αi is a word of length n−i in the generators, and αn−1=σn−1.

The words xi. For 1≤i≤n−1 put

xi:=σn−1−1⋯σi+1−1σi2σi+1⋯σn−1=αi+1−1σi2αi+1∈Bn.

The two displayed words for xi are literally the same word written in two ways: expanding αi+1−1 gives σn−1−1⋯σi+1−1 and expanding αi+1 gives σi+1⋯σn−1, so the middle factor σi2 sits in the same position in both readings. In particular xn−1=σn−12, and xi is a word of length 2(n−i) in the generators and their inverses.

These words are the Zariski combing words of the Artin presentation. The definition is choice-free, it uses no relation of the presentation, and it asserts no property of αi or xi inside any geometric braid model; all of that is established, when needed, by the items that cite this definition.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

Prefix insertion rewrites a trivial braid word into combing factors

Statement

Assume n≥2 and let W=σi1ε1⋯σimεm be a word in the Artin letters and their inverses whose image under the published surjection φ of The Artin presentation surjects onto the geometric braid group is the trivial geometric braid. Write Wk:=σi1ε1⋯σikεk for k=0,…,m, and for k=0,…,m let jk:=π(φ(Wk))−1(n)∈{1,…,n} be the position at the bottom of the sub-braid φ(Wk) of the point that sits at position n at its top, where π ⁣:Gn→Sn is the endpoint permutation homomorphism; here sr denotes the transposition of r and r+1. Then j0=n, the recursion jk=sik(jk−1)(k=1,…,m) holds, and jm=n because φ(W)=1. Equivalently, jk is the position of the point that starts at position n at the top of W after it has passed the first k letters (the library's stacking puts the first letter of a word on top, so this point meets the letters in word order).

Using only free insertions of the words αjαj−1 and free deletions of cancelling pairs (no braid relation), W is equivalent in the group Bn of The braid group by Artin presentation to the product of combing factors W≡∏k=1m(αjk−1−1 σikεk αjk), where the words α1,…,αn are those of The Zariski combing words alpha_i and x_i in the Artin presentation. The k-th factor is the k-th letter decorated by its two connectors: read bottom to top (the library's stacking puts the first letter of a word on top, so the last block of the factor is met first), the tracked point that starts at position n travels through αjk to position jk, is exchanged by the letter σikεk to position jk−1 when jk∈{ik,ik+1} (and is fixed otherwise), and is carried back to position n by αjk−1−1; equivalently, in the source's bottom-up reading the point has position jk below the letter and jk−1 above it. It lies in the letter's support precisely when jk∈{ik,ik+1}; otherwise jk−1=jk and it remains fixed outside that support during the letter. In every case the recursion above is exactly the interchange rule jk−1=sik(jk) used by the six-case analysis, the empty word is allowed (m=0, where W is empty and j0=jm=n), and nothing but φ(W)=1 is assumed about the geometric braid.

Facts & Assumptions

Given: An integer n≥2, a word W=σi1ε1⋯σimεm with φ(W)=1, its prefixes Wk, and the words αj of The Zariski combing words alpha_i and x_i in the Artin presentation.

[F1]

αj=σjσj+1⋯σn−1 for 1≤j≤n−1 and αn=1 is the empty word; all these are words in the generators and their inverses (The Zariski combing words alpha_i and x_i in the Artin presentation).

[F2]

Bn is the quotient of the free group on σ1,…,σn−1 by the normal closure of the two relation families; consequently words differing by insertions or deletions of adjacent inverse pairs w±1w∓1 represent the same element, and a product of words telescopes whenever adjacent connector words cancel (The braid group by Artin presentation, The Zariski combing words alpha_i and x_i in the Artin presentation).

[F3]

The assignment φ(σi)=[σi] extends to a surjective homomorphism φ ⁣:Bn→Gn (The Artin presentation surjects onto the geometric braid group).

[F4]

The endpoint permutation π ⁣:Gn→Sn is a homomorphism, the class of the half twist σr has π([σr])=sr, and under the stacking convention of Stacking of geometric braids is a well-defined associative operation on isotopy classes the class of a word u1⋯ul satisfies π(φ(u1⋯ul))=π(φ(u1))∘⋯∘π(φ(ul)) as functions, the first factor applied last (The elementary geometric half twist, its support disc, and its opposite, The isotopy classes of geometric braids based at Q form a group, and the endpoint permutation is a homomorphism, Stacking of geometric braids is a well-defined associative operation on isotopy classes).

Proof

technique · direct
1.1F1F3F4

The connectors move the n-th point down to position j. By [F3] and [F1], φ(αj)=φ(σj)⋯φ(σn−1)=[σj]⋯[σn−1], so by [F4] the endpoint permutation is π(φ(αj))=sj∘sj+1∘⋯∘sn−1 as a function. Evaluating on positions, this function sends n↦n−1↦n−2↦⋯↦j and fixes every x<j, so it is the cycle tj:=(j j+1 ⋯ n) with tj(n)=j; in particular a point starting at position n ends at position j, and tj−1(j)=n. For j=n the word αn is empty and π(φ(αn))=id⁡ with id⁡(n)=n.

1.2F3F4

The recursion and its endpoints. For each k, [F3] and [F4] give π(φ(Wk))=π(φ(σi1ε1))∘⋯∘π(φ(σikεk))=si1∘⋯∘sik, because sr±1=sr. Taking inverses, π(φ(Wk))−1=sik∘⋯∘si1, so jk:=π(φ(Wk))−1(n) satisfies j0=n (the empty product) and jk=sik(jk−1) for 1≤k≤m. Since Wm=W and φ(W)=1, π(φ(Wm))=id⁡ and therefore jm=id⁡(n)=n.

2.1F1F2step 1.2

The telescoping insertion. For k=1,…,m−1 insert the word αjkαjk−1 between the k-th and (k+1)-st letter of W and bracket the result as W=αj0−1 σi1ε1 αj1 ⋅ αj1−1 σi2ε2 αj2 ⋯ αjm−1−1 σimεm αjm. Each interior position contributes αjkαjk−1=1, which is a free cancellation by [F2], and by [F1] and step 1.2 the two end connectors are αj0=αn=1 and αjm=αn=1; expanding the displayed product therefore returns the original word W by free cancellations alone, and conversely W is obtained from the displayed product by the inverse free moves. No defining relation of the Artin presentation is used.

2.2F3F4step 1.1step 1.2

The bookkeeping inside a factor. Fix k and read the factor Fk=αjk−1−1σikεkαjk from the bottom upward, that is, starting from its last and lowest block αjk and ending with its first and topmost block αjk−1−1 (the word's first letter is the topmost layer in the stacking of [F4]). A point starting at position n at the bottom of the factor is carried by the block αjk to position tjk(n)=jk by step 1.1, then by the letter σikεk to position sikεk(jk)=sik(jk)=jk−1 by [F4] and step 1.2 (if jk=jk−1 the letter fixes it), and then by the block αjk−1−1 to position tjk−1−1(jk−1)=n by step 1.1. Hence inside the k-th factor the letter acts on the tracked point exactly through the interchange rule connecting the two connector positions jk (below the letter) and jk−1 (above it), and the tracked point returns to position n at the top of every factor, as it must because at the top of W it is again at position n by φ(W)=1.

3.1F4step 1.2step 2.1step 2.2

Conclusion. Steps 1.2 and 2.1 establish the recursion, its endpoints, and the telescoping product. Step 2.2 gives the positions jk below and jk−1 above each letter. By the half-twist definition in [F4], the tracked point lies in the letter's support if jk∈{ik,ik+1}; otherwise it is a fixed base point outside that support. For m=0 the product is empty. ∎

Remarks

  • The lemma is a pure bookkeeping statement: the group element is unchanged because each inserted connector is immediately cancelled, and the geometric input is only the published endpoint-permutation homomorphism, which fixes the positions jk by the triviality of φ(W).
  • In the source the same product is displayed with αj0=αjm=1, reading words bottom-up; the recursion jk=sik(jk−1) is identical in both conventions, and the six-case reduction of the next item depends only on this recursion and on the displayed shape of the factors.
LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-02Open item page →

Lower-rank Artin letters conjugate x-letters

Statement

Assume n≥3, let 1≤i≤n−2 and 1≤j≤n−1, and work in the group Bn of The braid group by Artin presentation with the words σ1,…,σn−1 and x1,…,xn−1 of The Zariski combing words alpha_i and x_i in the Artin presentation. Using only the two families of defining relations and free insertions and deletions of adjacent inverse letters:

(i) σi−1xjσi equals xj when j<i or j>i+1, equals xi+1xixi+1−1 when j=i+1, and equals xi+1 when j=i;

(ii) σixjσi−1 equals xj when j<i or j>i+1, equals xi when j=i+1, and equals xi−1xi+1xi when j=i.

Consequently, for every pair of signs ε,δ∈{±1} there is a word w in the letters x1±1,…,xn−1±1 with σiεxjδ=w σiε, so that in any word over the mixed alphabet σ1±1,…,σn−2±1,x1±1,…,xn−1±1 each occurrence of a lower-rank σ-letter can be moved to the right of every x-letter, the x-letters changing only by further x-letters and their inverses. All identities also hold in the geometric braid group under the published surjection φ of The Artin presentation surjects onto the geometric braid group.

Facts & Assumptions

Given: Integers n≥3, 1≤i≤n−2, 1≤j≤n−1, the group Bn of The braid group by Artin presentation, and the elements x1,…,xn−1∈Bn of The Zariski combing words alpha_i and x_i in the Artin presentation.

[F1]

In Bn the two defining families of relations hold: σrσr+1σr=σr+1σrσr+1 for 1≤r≤n−2, and σrσs=σsσr whenever ∣r−s∣>1; words equal in the free group on σ1,…,σn−1 and their inverses represent the same element of Bn, so adjacent inverse letters may be freely inserted and deleted (The braid group by Artin presentation, The Zariski combing words alpha_i and x_i in the Artin presentation).

[F2]

For every r with 1≤r≤n−1 one has the two displayed words xr=σn−1−1⋯σr+1−1σr2σr+1⋯σn−1=αr+1−1σr2αr+1, and xr−1=σn−1−1⋯σr+1−1σr−2σr+1⋯σn−1; also αs=σsσs+1⋯σn−1 for 1≤s≤n−1 and αn=1 (The Zariski combing words alpha_i and x_i in the Artin presentation).

[F3]

The map φ ⁣:Bn→Gn of The Artin presentation surjects onto the geometric braid group is a homomorphism with φ(σr)=[σr], and in the geometric braid group Gn the two families of relations of [F1] hold: [σr][σr+1][σr]=[σr+1][σr][σr+1] and [σr][σs]=[σs][σr] for ∣r−s∣>1 (The geometric three strand braid relation, Far commutativity of elementary geometric half twists).

Proof

technique · direct
1.1F1

Two mixed forms of the braid relation. Let 1≤r≤n−2. Multiplying σrσr+1σr=σr+1σrσr+1 on the left by σr−1 and on the right by σr+1−1 gives σr−1σr+1σr=σr+1σrσr+1−1, and multiplying the braid relation on the left by σr+1−1 and on the right by σr−1 gives σr+1−1σrσr+1=σrσr+1σr−1, whose inverse is σrσr+1−1σr−1=σr+1−1σr−1σr+1. Both are consequences of the defining relations of [F1] alone.

1.2F1F2

The case j>i+1. Every letter σr±1 occurring in the displayed word for xj of [F2] has r≥j≥i+2, so ∣i−r∣≥2 and σi commutes with that letter by the far-commutation relation of [F1]; repeating this letter by letter, σi commutes with the whole word, so σi±1xjσi∓1=xj. In particular both assertions (i) and (ii) hold for j>i+1.

2.1F1F2step 1.1

The case j<i: sliding σi past xj. Decompose the word of [F2] for xj at the index i, which satisfies j<i, as xj=A σi−1KσiB,A:=σn−1−1⋯σi+1−1,K:=σi−1−1⋯σj+1−1σj2σj+1⋯σi−1,B:=σi+1⋯σn−1, where each of A, K, B may be empty and the displayed equality is free cancellation in the two words of [F2]. Since σi commutes with every letter of A except σi+1−1, and σiσi+1−1σi−1=σi+1−1σi−1σi+1 by step 1.1, one has σiAσi−1=Aσi−1σi+1; since σi+1 commutes with every letter of K (all its indices are at most i−1), and since B=σi+1B′ with B′:=σi+2⋯σn−1, the computation σixj=σiAσi−1KσiB=Aσi−1σi+1KσiB=Aσi−1Kσi+1σiσi+1B′=Aσi−1K σiσi+1σi B′=Aσi−1Kσiσi+1B′σi=xjσi uses only far commutation, the braid relation, and the fact that σi commutes with every letter of B′. Left-multiplying by σi−1 gives σi−1xjσi=xj, and right-multiplying by σi−1 gives σixjσi−1=xj; both assertions hold for j<i.

2.2F1F2step 1.1

The case j=i+1. Put C:=σn−1−1⋯σi+2−1 and D:=σi+2⋯σn−1, so that C=D−1 as words and, by [F2], xi+1=Cσi+12D,xi=Cσi+1−1σi2σi+1D,xi+1−1=D−1σi+1−2C−1. Since σi commutes with every letter of C and of D, and using step 1.1, σi−1xi+1σi=C σi−1σi+12σi D=C(σi−1σi+1σi)2D=C(σi+1σiσi+1−1)2D=C σi+1σi2σi+1−1D. Expanding the product xi+1xixi+1−1 with the three displayed words and using D C=D D−1=1 and C−1=D gives xi+1xixi+1−1=Cσi+12(DC)σi+1−1σi2σi+1(DD−1)σi+1−2C−1=Cσi+1σi2σi+1−1D, so σi−1xi+1σi=xi+1xixi+1−1.

3.1F1F2step 1.1

The case j=i. With the same words C,D of step 2.2, [F2] gives xi=Cσi+1−1σi2σi+1D, and σi commutes with every letter of C and of D, so σi−1xiσi=C σi−1σi+1−1σi2σi+1σi D=C(σi−1σi+1−1σi)(σiσi+1σi)D. Inverting the first identity of step 1.1 gives σi−1σi+1−1σi=σi+1σi−1σi+1−1, and the defining braid relation gives σiσi+1σi=σi+1σiσi+1; substituting, σi−1xiσi=C σi+1σi−1σi+1−1σi+1σiσi+1D=C σi+1σi−1σiσi+1D=Cσi+12D=xi+1, the penultimate equality deleting the adjacent inverse pairs and the last equality being [F2] again.

4.1step 2.2step 3.1

The second orientation for j=i and j=i+1. The map ci ⁣:Bn→Bn, ci(g):=σi−1gσi, is the conjugation automorphism by σi, with inverse ci−1(g)=σigσi−1. Steps 2.2 and 3.1 give ci(xi)=xi+1 and ci(xi+1)=xi+1xixi+1−1, hence ci(xi−1xi+1xi)=xi+1−1(xi+1xixi+1−1)xi+1=xi; therefore ci−1(xi)=xi−1xi+1xi, that is σixiσi−1=xi−1xi+1xi, and ci−1(xi+1)=xi, that is σixi+1σi−1=xi. This is assertion (ii) in the two remaining cases.

5.1F2step 1.2step 2.1step 2.2step 3.1step 4.1

All four signs. Let ε,δ∈{±1}. If δ=1, then σiεxj=(σiεxjσi−ε)σiε; by steps 1.2, 2.1, 2.2, 3.1 and 4.1 the middle factor equals xj (for j<i or j>i+1), or xi+1, xi+1xixi+1−1, xi, xi−1xi+1xi (for j=i or j=i+1, according to the sign of ε), so in every case it is a word in x1±1,…,xn−1±1. If δ=−1, then σiεxj−1=(σiεxjσi−ε)−1σiε, and the same case list applies with the inverse word. Hence for all signs σiεxjδ=wσiε with w a word in the x-letters and their inverses, and a leftmost occurrence of σi±1 in any mixed word can therefore be moved one x-letter at a time to the right of all x-letters, only x-letters changing.

6.1F3step 1.1step 5.1

Transfer to the geometric braid group. Since φ is a homomorphism with φ(σr)=[σr] by [F3], and since the relations used in steps 1.1-5.1 are exactly the two families of [F1], applying φ to each identity yields the corresponding identity in Gn between φ(σi)±1 and φ(xj)±1: the images satisfy the braid relation and far commutation by the published geometric lemmas, and the free cancellations map to cancellations in the group Gn.

7.1step 1.2step 2.1step 2.2step 3.1step 4.1step 5.1step 6.1

Assertion (i) is steps 1.2, 2.1, 2.2 and 3.1, assertion (ii) is steps 1.2, 2.1 and 4.1, the collection statement is step 5.1, and the geometric transfer is step 6.1. ∎

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

Each combing factor reduces to a lower-rank letter or an x-letter

Statement

Assume n≥2, work in the group Bn=⟨σ1,…,σn−1⟩ of The braid group by Artin presentation, and use the words α1,…,αn and x1,…,xn−1 of The Zariski combing words alpha_i and x_i in the Artin presentation. Let 1≤j≤n be a position, 1≤k≤n−1 an index and ε∈{±1} a sign, and let the combing factor be the word F=αj−1 σkε αj′,j′:={k+1,j=k,k,j=k+1,j,otherwise, so that j′=sk(j); in the bottom-to-top reading of this factor, j′ is the position below the letter and j is the position above it, as in Prefix insertion rewrites a trivial braid word into combing factors. Then, using only the two Artin relations and free insertions and deletions of adjacent inverse letters:

(a) F≡1 (the empty word) if j=k and ε=1;

(b) F≡xk−1 if j=k and ε=−1;

(c) F≡xk if j=k+1 and ε=1;

(d) F≡1 if j=k+1 and ε=−1;

(e) F≡σkε if k<j−1;

(f) F≡σk−1ε if k>j.

The six cases are mutually exclusive and exhaustive, and in every one of them the reduced form is a word in σ1±1,…,σn−2±1,x1±1,…,xn−1±1 alone. In particular the letter σn−1 and its inverse never survive the reduction outside an x-letter, and all identities also hold in the geometric braid group Gn under the published surjection φ of The Artin presentation surjects onto the geometric braid group.

Facts & Assumptions

Given: An integer n≥2, the group Bn of The braid group by Artin presentation, the words αj and xk of The Zariski combing words alpha_i and x_i in the Artin presentation, a position 1≤j≤n, an index 1≤k≤n−1, a sign ε∈{±1}, and the word F=αj−1σkεαj′ with j′=sk(j) as in the statement.

[F1]

In Bn the two defining families of relations hold, σrσr+1σr=σr+1σrσr+1 for 1≤r≤n−2 and σrσs=σsσr for ∣r−s∣>1, and two words that differ by insertions or deletions of adjacent inverse pairs w±1w∓1 represent the same element of Bn (The braid group by Artin presentation). Below we write u≡v when the words u and v can be connected by these two families of relations together with such free insertions and deletions.

[F2]

αj=σjσj+1⋯σn−1 for 1≤j≤n−1, αn is the empty word, and consequently the word identity αk=σkαk+1 and its inverse form αk−1=αk+1−1σk−1 hold for every 1≤k≤n−1. For every 1≤k≤n−1 one has xk=σn−1−1⋯σk+1−1σk2σk+1⋯σn−1=αk+1−1σk2αk+1 and hence xk−1=αk+1−1σk−2αk+1 (The Zariski combing words alpha_i and x_i in the Artin presentation).

[F3]

The factor F is exactly the shape of a combing factor in Prefix insertion rewrites a trivial braid word into combing factors: read bottom to top, the tracked point starts at position n at the bottom, is carried by αj′ to position j′ below the letter, is exchanged by σkε to j=sk(j′) when j′∈{k,k+1} and is fixed otherwise, and is carried by αj−1 back to position n at the top. Since sk is an involution, this is the same relation j′=sk(j) used in the statement.

[F4]

The map φ ⁣:Bn→Gn of The Artin presentation surjects onto the geometric braid group is a surjective homomorphism with φ(σr)=[σr], and in Gn the two families of relations of [F1] hold: [σr][σr+1][σr]=[σr+1][σr][σr+1] for 1≤r≤n−2 and [σr][σs]=[σs][σr] for ∣r−s∣>1 (The geometric three strand braid relation, Far commutativity of elementary geometric half twists).

Proof

technique · direct
1.1F1F2F3

The four cases in which the letter moves the tracked point. Assume j=k or j=k+1; by [F2] we have the word identities αk=σkαk+1, αk−1=αk+1−1σk−1, xk=αk+1−1σk2αk+1 and xk−1=αk+1−1σk−2αk+1. Substituting αk or αk+1 for the two connectors and cancelling the adjacent inverse pair σk−1σk, or its inverse pair, by [F1]: for j=k and ε=1, F=αk−1σkαk+1=αk+1−1σk−1σkαk+1≡αk+1−1αk+1≡1, the empty word; for j=k and ε=−1, F=αk−1σk−1αk+1=αk+1−1σk−2αk+1=xk−1; for j=k+1 and ε=1, F=αk+1−1σkαk=αk+1−1σkσkαk+1=αk+1−1σk2αk+1=xk; and for j=k+1 and ε=−1, F=αk+1−1σk−1αk=αk+1−1σk−1σkαk+1≡αk+1−1αk+1≡1. This gives (a), (b), (c) and (d).

1.2F1F2

The case k<j−1: far commutation. Here j∉{k,k+1}, so j′=j and F=αj−1σkεαj. Every letter σr±1 occurring in the word αj of [F2] has index r≥j≥k+2, hence ∣r−k∣≥2 and σk commutes with that letter by the far-commutation relation of [F1]; iterating over the letters of αj (whose length is n−j, possibly 0 when j=n), we get σkεαj≡αjσkε, whence F=αj−1σkεαj≡αj−1αjσkε≡σkε by free cancellation. This is (e); note k≤j−2 and j≤n, so k≤n−2.

1.3F1F2

The case k>j: sliding the letter to the right. Here again j∉{k,k+1}, so j′=j and F=αj−1σkεαj. First take ε=1. Since j<k≤n−1, the word αj=σj⋯σn−1 splits, with the three groups possibly empty, as the word αj=(σj⋯σk−2) σk−1 σk (σk+1⋯σn−1), where the first group contains exactly the letters with indices r≤k−2 and the last exactly those with indices r≥k+1. Every letter of the first group has k−r≥2, so σk commutes with each of them and moves right past them; every letter of the last group has r−(k−1)≥2, so σk−1 commutes with each of them and moves right past them; and the three middle letters satisfy the braid relation σkσk−1σk≡σk−1σkσk−1 by [F1]. Combining the three moves gives the chain σkαj≡(σj⋯σk−2) σkσk−1σk (σk+1⋯σn−1)≡(σj⋯σk−2) σk−1σkσk−1 (σk+1⋯σn−1)≡(σj⋯σk−2) σk−1σk (σk+1⋯σn−1) σk−1=αjσk−1. For ε=−1, left-multiply this identity in the group Bn by σk−1: it becomes αj=σk−1αjσk−1, hence σk−1αj=αjσk−1−1. In both signs, therefore, σkεαj≡αjσk−1ε and F=αj−1σkεαj≡αj−1αjσk−1ε≡σk−1ε by free cancellation. This is (f); here k>j≥1 gives k−1≥1 and k≤n−1 gives k−1≤n−2.

2.1F1F4step 1.1step 1.2step 1.3

Transfer to the geometric braid group. Since φ is a homomorphism with φ(σr)=[σr] by [F4], applying φ to each of the reductions of steps 1.1, 1.2 and 1.3 turns it into the corresponding identity in Gn: the free cancellations become [γ][γ]−1=1, and the two Artin relations used are the geometric relations supplied by the published The geometric three strand braid relation and Far commutativity of elementary geometric half twists.

3.1step 1.1step 1.2step 1.3step 2.1

Conclusion. The conditions of the six cases are exactly: j=k (with either sign), j=k+1 (with either sign), j∉{k,k+1} with k≤j−2, and j∉{k,k+1} with k≥j+1; if j∉{k,k+1} then either j>k+1, that is k<j−1, or j<k, that is k>j, so the list is exhaustive, and the conditions are visibly mutually exclusive. Steps 1.1, 1.2 and 1.3 establish the reductions (a)-(f), and the reduced forms are the empty word, xk±1, or σkε with k≤n−2, or σk−1ε with k−1≤n−2, so all of them are words in σ1±1,…,σn−2±1,x1±1,…,xn−1±1; in particular no copy of σn−1 survives the reduction outside an x-letter. Step 2.1 transfers each reduction to Gn. ∎

Remarks

  • The case list is exactly the source's list for the factors (σn−1−1⋯σi−1)σk±1(σi⋯σn−1), written with the library's first-letter-first convention; the slide identity σkαj≡αjσk−1 is the source's displayed relation (3.2), and it is the only place where the braid relation is used in cases (e) and (f).
  • Cases (a)-(f) are the mechanism by which a combing factor that meets the trivial point either disappears, becomes an x-letter, or degenerates to a letter of rank at most n−2; the surviving lower-rank letters are collected to the right of the x-letters by Lower-rank Artin letters conjugate x-letters.
LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

Every trivial braid word combs as W_1W_2

Statement

Assume n≥2, and let W be a word in σ1±1,…,σn−1±1 whose geometric image under the surjection φ of The Artin presentation surjects onto the geometric braid group is the trivial geometric braid. Then W is equivalent to a product W1W2, using only the two Artin relations and free insertions and deletions of adjacent inverse pairs σ±1σ∓1, in which

Both W1 and W2 may be empty, and no letter σn−1 or σn−1−1 occurs in either of the two words.

Facts & Assumptions

Given: An integer n≥2, the group Bn=⟨σ1,…,σn−1⟩ of The braid group by Artin presentation, a word W in the Artin letters and their inverses with φ(W)=1, and the words α1,…,αn and x1,…,xn−1 of The Zariski combing words alpha_i and x_i in the Artin presentation.

[F1]

In Bn the two families of defining relations hold, and two words differing by insertions or deletions of adjacent inverse pairs w±1w∓1 represent the same element; write u≡v when the words u,v can be connected by these moves and the two relation families (The braid group by Artin presentation). The moves are symmetric, so ≡ is an equivalence relation, and it is compatible with concatenation in the sense that u≡v implies put≡pvt for words p,t.

[F2]

φ ⁣:Bn→Gn is the surjective homomorphism of The Artin presentation surjects onto the geometric braid group with φ(σi)=[σi].

[F3]

Prefix insertion (Prefix insertion rewrites a trivial braid word into combing factors): if W=σi1ε1⋯σimεm and φ(W)=1, then there are positions j0,j1,…,jm∈{1,…,n} with j0=jm=n and jk=sik(jk−1) such that W≡∏k=1m(αjk−1−1σikεkαjk), where each factor is a combing factor in the sense of Each combing factor reduces to a lower-rank letter or an x-letter with j=jk−1, k↦ik, ε↦εk and j′=jk.

[F4]

Each combing factor reduces to a word of at most one letter in the mixed alphabet σ1±1,…,σn−2±1,x1±1,…,xn−1±1: the empty word, or xik±1, or σikεk with ik≤n−2, or σik−1εk with ik−1≤n−2 (Each combing factor reduces to a lower-rank letter or an x-letter).

[F5]

Conjugation table (Lower-rank Artin letters conjugate x-letters): for every 1≤i≤n−2, 1≤j≤n−1 and signs ε,δ∈{±1} there is a word w in the letters x1±1,…,xn−1±1 with σiεxjδ=w σiε in Bn. In particular a contiguous pair consisting of a lower-rank σ-letter followed immediately by an x-letter can be replaced by a word of x-letters followed by that same σ-letter.

Proof

technique · direct
1.1F2F3

Prefix insertion. By [F3] and φ(W)=1, there are positions j0=jm=n with jk=sik(jk−1) and W≡∏k=1mFk, Fk:=αjk−1−1σikεkαjk.

1.2F1F4

Reducing the factors. Fix k and apply [F4] to Fk, whose letter has index ik and whose connector position is jk−1 with jk=sik(jk−1); the factor is equivalent to the empty word, or to a single letter xik or xik−1, or to a single letter σikεk with ik≤n−2, or to a single letter σik−1εk with ik−1≤n−2. Deleting the factors that reduce to the empty word and choosing one such reduced word in each remaining factor, we obtain a word U in the mixed alphabet σ1±1,…,σn−2±1,x1±1,…,xn−1±1 with W≡U, because ≡ is compatible with concatenation by [F1].

1.3F1F5

Collection of the lower-rank letters. We show: for every word U in the mixed alphabet σ1±1,…,σn−2±1,x1±1,…,xn−1±1 there are a word u in the x-letters and a word v in σ1±1,…,σn−2±1 with U≡uv. Proceed by induction on the number r of σ-letters occurring in U. If r=0, take u:=U and v empty. If r≥1, let s be the last (rightmost) σ-letter of U and write U=P s A, where A is the (possibly empty) x-word following s, so that no σ-letter occurs in A. While A is nonempty, let x be its first letter and replace the adjacent pair sx by ws, where s=σiε with i≤n−2, x=xjδ and σiεxjδ=wσiε is the identity of [F5]; this is a permitted rewrite, and the new word again has s as its rightmost σ-letter, now followed by A with its first letter deleted, because the x-word w stands immediately to the left of s. Hence the number of letters strictly to the right of s decreases by exactly one at each rewrite, so after finitely many steps we obtain a word U′≡U whose letters strictly to the right of the rightmost σ-letter s are none, that is, U′=P′ s where P′ is a word in the mixed alphabet with exactly r−1 σ-letters. By the induction hypothesis applied to P′, there are an x-word u′ and a σ-word v′ with P′≡u′v′; then U≡U′≡u′(v′s) by [F1], where u′ is an x-word and v′s is a word in the σ-letters of rank at most n−2, so the induction is complete.

2.1F1step 1.2step 1.3

Assembly. By step 1.2 there is a mixed word U with W≡U, and by step 1.3 there are an x-word W1 and a lower-rank σ-word W2 with U≡W1W2; since ≡ is transitive by [F1], W≡W1W2, which is the required product.

3.1step 1.1step 1.2step 1.3step 2.1

Conclusion. Given W with φ(W)=1, steps 1.1-1.3 rewrite it, using only the two Artin relations and free insertions and deletions of adjacent inverse pairs, first into the product of its combing factors, then into a word U in the mixed alphabet, and finally into a product W1W2 with W1 a word in x1±1,…,xn−1±1 and W2 a word in σ1±1,…,σn−2±1; the name n−2 in the statement is justified because [F4] bounds every surviving σ-index by n−2. ∎

Remarks

  • The collection step is the source's "we can collect all the σi±1 on the right". The naive measure "number of pairs (an x-letter left of a σ-letter)" is not monotone, because a pair sx can be replaced by a word ws in which w has up to three letters, for instance σixi=xi−1xi+1xiσi; the proof above instead processes the σ-letters from right to left, and each swap strictly shortens the segment to the right of the processed letter.
  • Only the two Artin relations, the conjugation table of Lower-rank Artin letters conjugate x-letters, and the tracking of Prefix insertion rewrites a trivial braid word into combing factors are used; in particular the geometric input is only φ(W)=1.
  • For n=2 the mixed alphabet contains no σ-letters of rank at most 0, so the conclusion reads W≡W1 with W1 a word in x1±1; the lemma is not used to determine how many x1-factors occur, and the induction of the completeness theorem below supplies that in the trivial case.
LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

The combed geometric decomposition is unique

Statement

Assume AC and n≥2, and work with the abstract Artin group Bn of The braid group by Artin presentation, the words x1,…,xn−1 of The Zariski combing words alpha_i and x_i in the Artin presentation and the published surjection φ ⁣:Bn→Gn of The Artin presentation surjects onto the geometric braid group. Let W1 be a word in x1±1,…,xn−1±1 and W2 a word in σ1±1,…,σn−2±1 such that φ(W1W2) is the trivial geometric braid; such a pair exists for every combed word by Every trivial braid word combs as W_1W_2. Write Ψn ⁣:Gnpure→PBn for the canonical isomorphism of Pure geometric braids and ordered configuration loops, and ρ ⁣:PBn→PBn−1 for the forgetting map of The Fadell-Neuwirth short exact sequence for pure braids. Write Aij∈Gn for the geometric generators of Standard geometric pure braid generators A_ij, and write A~ij:=σj−1⋯σi+1σi2σi+1−1⋯σj−1−1∈Bn for their Artin-word lifts, so φ(A~ij)=Aij. Put P~i:=A~i+1,n⋯A~n−1,n in Bn and Pi:=Ai+1,n⋯An−1,n in Gn, with both empty when i=n−1. Then:

(i) every φ(xi) is a pure braid and φ(xi)=Pi−1 Ain Pi, so that the conjugating element lies in the subgroup generated by the standard generators with larger first index; at the word level this is xi≡P~i−1A~inP~i in Bn. Moreover Ψn(φ(x1)),…,Ψn(φ(xn−1)) form a free basis of the free kernel ker⁡ρ of the forgetting map of The Fadell-Neuwirth short exact sequence for pure braids;

(ii) φ(W1)=1, and W1, as a word in the alphabet x1±1,…,xn−1±1, reduces to the empty word by free cancellations of adjacent inverse pairs xi±1xi∓1 (each of which, after expanding both xi's, is a permitted deletion of σ-pairs);

(iii) φ(W2)=1, and, viewing the same letter word as an element of Bn−1, its geometric image under the rank-(n−1) surjection φn−1 ⁣:Bn−1→Gn−1 is trivial: φn−1(W2)=1∈Gn−1. This asserts triviality of the geometric image, not that W2=1 in Bn−1.

Facts & Assumptions

Given: An integer n≥2, the groups Bn and Gn of The braid group by Artin presentation and The isotopy classes of geometric braids based at Q form a group, and the endpoint permutation is a homomorphism, the words αj,xj of The Zariski combing words alpha_i and x_i in the Artin presentation, words W1 in x1±1,…,xn−1±1 and W2 in σ1±1,…,σn−2±1 with φ(W1W2)=1, the standard pure braid generators Aij of Standard geometric pure braid generators A_ij, and the identifications of The Fadell-Neuwirth short exact sequence for pure braids and Pure geometric braids and ordered configuration loops.

[F1]

In Bn the two Artin relations hold and adjacent inverse σ-pairs may be freely inserted and deleted. For 1≤i<j≤n, set A~ij:=σj−1⋯σi+1σi2σi+1−1⋯σj−1−1∈Bn; its image under φ is the geometric standard generator Aij of Standard geometric pure braid generators A_ij, since φ(σr)=[σr] and φ preserves the displayed stacking product (The braid group by Artin presentation, The Artin presentation surjects onto the geometric braid group, Standard geometric pure braid generators A_ij). The words αj satisfy αj=σj⋯σn−1 and αn=1, and xi=αi+1−1σi2αi+1 (The Zariski combing words alpha_i and x_i in the Artin presentation).

[F2]

Assumption AC, and the published consequences that are used to identify the free kernel: AC implies dependent choice and countable choice (The Axiom of Choice, AC implies DC implies countable choice), so that the Fadell--Neuwirth fibrations of The Fadell-Neuwirth short exact sequence for pure braids are available: forgetting the last strand is a homomorphism ρ ⁣:PBn→PBn−1 with kernel a free group of rank n−1, and under the fiber-inclusion identification the elements Ψn(A1n),…,Ψn(An−1,n) are a free basis of that kernel (The Ain are meridian generators of the forgetful free kernel). Forgetting a strand is realized by the coordinate projection: under the identifications [F3] at ranks n and n−1, a pure braid whose coordinate loop is (Z1,…,Zn) satisfies ρ(Ψn([β]))=(ι∗F[(Z1,…,Zn−1)])−1, with target basepoint (q1,…,qn−1). This follows because coordinate projection commutes with the open-to-closed inclusion and its induced homomorphism preserves inverses.

[F3]

For every rank m≥0, Ψm ⁣:Gmpure=ker⁡πgeo→PBm is an isomorphism, and for a pure braid [β] with coordinate path zβ it is Ψm([β])=(ι∗F[zβ])−1, the inverse of the class of the coordinate loop carried to the closed-disc configuration space; the same sign convention is used at every rank, so the inverse cancels in the comparisons between ranks below. At every interior base configuration the open-to-closed inclusion induces an isomorphism on fundamental groups (The interior-disc and closed-disc configuration spaces are homotopy equivalent). Consequently [β]=1 if and only if [zβ]=1 (Pure geometric braids and ordered configuration loops, Ordered configuration spaces Fn(X), The homomorphism on fundamental groups induced by a pointed continuous map). Moreover πconf(Φ([β]))=πgeo([β]), so a geometric braid is pure exactly when its class lies in PBn under these identifications (The geometric endpoint permutation matches covering monodromy).

[F4]

Geometric conventions. The base configuration is Q=(q1,…,qn) with qj=((2j−n−1)h,0), h=14(n+1), and the (n−1)-strand base configuration is Q(n−1)=(q1′,…,qn−1′) with qj′=((2j−n)h′,0), h′=14n (Geometric braids in the disc with setwise endpoints). The elementary half twist σi acts as mi+ρ(t), mi−ρ(t) on the two points qi,qi+1 and fixes all others, where mi=qi+(h,0) and ρ is the diamond path of size h (The elementary geometric half twist, its support disc, and its opposite); stacking is first-under-second with π(γ⋆β)=π(γ)∘π(β) (Stacking of geometric braids is a well-defined associative operation on isotopy classes, The isotopy classes of geometric braids based at Q form a group, and the endpoint permutation is a homomorphism). The two configurations are compared through the similarity A(w):=aw−14(n+1) with a:=nn+1. Since 14(n+1)=1−a4, this is the dilation about c=−14, namely A(w)=c+a(w−c). It satisfies A(Q(n−1))=(q1,…,qn−1); its image A(D∘) is the open disc of radius a centered at A(0)=−14(n+1), whose closure lies in D∘ because a+14(n+1)<1.

[F5]

Moving-homotopy transport: for a path γ ⁣:x0→x1 in a space X and each n≥1 there is a transport isomorphism βγ ⁣:πn(X,x1)→πn(X,x0) depending only on the endpoint-fixed class of γ; if H ⁣:f≃g is a homotopy with basepoint track γ, then f∗=βγ∘g∗. In degree one βγ[a]=[γ∗a∗γˉ] (Higher homotopy basepoint transport and moving homotopies).

Proof

technique · direct
1.1F1

The combing identity, by free cancellation. For 1≤j≤n−1 put δj:=σn−1σn−2⋯σj+1 and γj:=σj+1σj+2⋯σn−1 (empty words for j=n−1); then δj=δj+1σj+1 and γj=σj+1γj+1 as words, and by [F1] the Artin word A~jn=δjσj2δj−1 lifts the geometric generator Ajn, while xj=γj−1σj2γj is the combing word. First, for every j the Artin word Q~j:=A~j,nA~j+1,n⋯A~n−1,n freely reduces to δjσj2γj: this is a downward induction on j, in which both words are σn−12 for j=n−1, and in which the substitution of A~j,n=δjσj2δj−1 and of the reduced form Q~j+1≡δj+1σj+12γj+1 gives Q~j≡δjσj2(δj−1δj+1)σj+12γj+1=δjσj2σj+1−1(δj+1−1δj+1)σj+12γj+1, where the block δj+1−1δj+1 cancels freely to the empty word -- its middle pair σn−1−1σn−1 cancels first, after which the next pair is adjacent, and so on outwards -- leaving δjσj2γj. For 1≤i≤n−2 this gives P~i:=Q~i+1≡δi+1σi+12γi+1, and the inverse word is P~i−1≡γi+1−1σi+1−2δi+1−1; substituting these reduced forms into P~i−1A~inP~i, and using δi+1−1δi≡σi+1 and δi−1δi+1≡σi+1−1 (the same middle-outward cancellation applied to δi=δi+1σi+1), gives P~i−1A~inP~i≡γi+1−1σi+1−2(δi+1−1δi)σi2(δi−1δi+1)σi+12γi+1≡γi+1−1σi+1−1σi2σi+1γi+1=γi−1σi2γi=xi. The remaining index i=n−1 has the empty product P~n−1 and xn−1=A~n−1,n=σn−12, so xn−1=P~n−1−1A~n−1,nP~n−1 holds trivially. No Artin relation is used in this step, only free insertions and deletions. Applying φ gives the stated geometric identity.

1.2F4

The two models of a lower-rank word. Put a:=n/(n+1) and consider the similarity A(w)=aw−hn with hn=14(n+1), so that A(w)=aw−14(n+1). One computes A(qj′)=qj for 1≤j≤n−1, A(mi′)=mi for 1≤i≤n−2 where mi′=qi′+(h′,0) is the (n−1)-strand midpoint, and aρ′(t)=ρ(t) for the two displacement paths, since ah′=h and each diamond-path coordinate is linear in its size. The reflected paths for the negative letters satisfy the same scaling identity. Since A is affine, A(mi′±ρ′(t))=A(mi′)±aρ′(t)=mi±ρ(t). Thus for every 1≤i≤n−2 and both signs, A((σi(n−1))j(t))=(σi(n))j(t)(1≤j≤n−1),A(qj′)=qj, where σi(k) is the elementary half twist of the k-strand model of [F4]. Every lower-rank moving point stays in the support disc Ui′ of the letter currently being run, or is one of the fixed base points. Relative to c:=(−14,0), these base points have radii ∣qj′−c∣=j/(2n)≤(n−1)/(2n), and the support disc Ui′ has center mi′ with ∣mi′−c∣=(2i+1)/(4n) and radius 3/(8n); hence every such path point has radius at most (4n−3)/(8n). This local agreement is all that will be used: A maps the lower standard-generator paths to the corresponding n-strand paths, but it is not used to map arbitrary lower braids into the complement of qn. Consequently, for every word V in σ1±1,…,σn−2±1, if z=(z1,…,zn−1) is its (n−1)-strand coordinate motion and Z=(Z1,…,Zn) its n-strand coordinate motion, then Zj=A∘zj(1≤j≤n−1),Zn≡qn, by induction on its letters: the generator paths agree by the displayed identity, and stacking has the same coupling permutation on {1,…,n−1} in both models and fixes n. Thus φn(V) is represented by (A∘z1,…,A∘zn−1,qn).

2.1F2F3F4F5

The straight-strand extension is a well-defined injective homomorphism. Let c:=(−14,0), a:=n/(n+1), R0:=12−14n and rn:=n/(2(n+1)). For θ∈R let uθ:=(cos⁡θ,sin⁡θ) and let R(θ):=14cos⁡θ+1−116sin⁡2θ, the distance from c to the boundary of D∘ along the ray c+ruθ; in particular R(θ)≥34>R0. Define the strictly increasing radial function λ(r,θ):={ar,0≤r≤R0,aR0+(rn−aR0)r−R0R(θ)−R0,R0≤r<R(θ), and define E(c+ruθ):=c+λ(r,θ)uθ, with E(c):=c. Since aR0<rn, each ray is mapped increasingly onto the ray segment of length rn; thus E is a homeomorphism from D∘ onto the open disk Ln:=B(c,rn). The closure of Ln lies in D∘ because ∥c∥2+rn<34<1, and qn lies on ∂Ln because ∣qn−c∣=n/(2(n+1))=rn. Moreover, E=A whenever ∣w−c∣≤R0. The basepoints satisfy ∣qj′−c∣=j/(2n)≤(n−1)/(2n)<R0, and for n≥3 every lower support disc satisfies max⁡1≤i≤n−2sup⁡w∈Ui′∣w−c∣≤4n−38n<R0 (for n=2 there are no such support discs). Therefore E(Q(n−1))=(q1,…,qn−1), and [F4] with step 1.2 shows that E maps every lower-rank standard-generator path to its corresponding n-strand path. If β∈Gn−1pure with representative coordinate loop z=(z1,…,zn−1), define sˇ(β) to be the class of (E∘z1,…,E∘zn−1,qn). The tuple is a pure n-strand braid: E is injective, its image Ln avoids qn, and the endpoints are E(Q(n−1))=(q1,…,qn−1). Applying E to a braid isotopy preserves pairwise distinctness and avoids qn, so sˇ is well defined on isotopy classes; applying E to the stacking formula shows it is a homomorphism. To prove injectivity, first note that the induced map E∗:π1(Fn−1(D∘),Q(n−1))→π1(Fn−1(D∘),E(Q(n−1))) is injective. Indeed, for 0≤t≤1 let λt(r,θ):=(1−t)r+tλ(r,θ) and Et(c+ruθ):=c+λt(r,θ)uθ. For each t, λt is strictly increasing along every ray and, when r<R(θ), λt(r,θ)<(1−t)R(θ)+trn≤R(θ), so Et is an embedding D∘→D∘, with E0=id⁡ and E1=E. Hence Fn−1(Et) is a homotopy on the ordered configuration space from the identity to Fn−1(E), whose basepoint follows η(t):=Fn−1(Et)(Q(n−1)); by [F5], id⁡∗=βη∘E∗, and since βη is an isomorphism, E∗ is injective. If sˇ(β)=1, [F3] and the coordinate-projection description of [F2] imply E∗[z]=1; injectivity gives [z]=1, and [F3] implies β=1.

2.2F2F3step 1.1

The conjugating identity in the geometric group. Applying φ to the word identity of step 1.1 gives φ(xi)=Pi−1AinPiin Gn, where Pi=Ai+1,n⋯An−1,n is the geometric product defined in the Statement. Each factor Aj,n of Pi and Ain itself lies in the pure subgroup and maps under Ψn into ker⁡ρ by [F2, F3]. Since ker⁡ρ is a subgroup, Ψn(φ(xi)) lies in ker⁡ρ; in particular φ(xi) is pure.

3.1F2F3step 2.1

The straight-strand extensions meet the kernel trivially. By [F2] the forgetting map acts on coordinate loops by dropping the last coordinate. For the straight-strand extension of step 2.1 the coordinate loop is (E∘z1,…,E∘zn−1,qn), whose dropped loop is E∘z. Thus ρ(Ψn(sˇ(β)))=ι∗F(E∗[z])−1 at the basepoint (q1,…,qn−1)=E(Q(n−1)). If γ=Ψn(sˇ(β)) also lies in ker⁡ρ, this class is trivial. The inclusion map ι∗F is injective at that basepoint by [F3], so E∗[z]=1. Injectivity of E∗ (step 2.1) gives [z]=1, and [F3] gives β=1 and γ=1. In particular Ψn(im⁡sˇ)∩ker⁡ρ={1}, which is the uniqueness statement used below.

3.2F2step 2.2

The free basis by the left-inverse argument. Let K:=ker⁡ρ≤PBn be the free kernel, and let F:=F(a1,…,an−1) be the abstract free group. The map ȷ:F→K given by ȷ(aj):=Ψn(Ajn) is an isomorphism by the free-basis clause of [F2]. Write wi:=ai+1ai+2⋯an−1∈F and ti:=wi−1aiwi∈F. By step 2.2 and the homomorphism property of Ψn, ȷ(ti)=Ψn(φ(xi)). First, the elements t1,…,tn−1 generate F: downward induction on i shows ⟨ti,ai+1,…,an−1⟩=⟨ai,…,an−1⟩, because ai=witiwi−1 lies in the left side and ti lies in the right side; at i=n−1 one has wn−1=1 and tn−1=an−1. Second, they are independent: define an endomorphism θ of F by descending recursion on i by θ(ai):=θ(wi)aiθ(wi)−1, where θ(wi) is the image of the word wi under the already defined θ on the generators ai+1,…,an−1 (at i=n−1 this gives θ(an−1)=an−1); this prescription specifies an element θ(ai)∈F for every free generator, hence defines a unique endomorphism θ ⁣:F→F. Then θ(ti)=θ(wi)−1θ(ai)θ(wi)=θ(wi)−1θ(wi)aiθ(wi)−1θ(wi)=ai for every i. If ρ′ ⁣:F→F denotes the homomorphism with ρ′(ai)=ti, then θ(ρ′(ai))=θ(ti)=ai, so θ∘ρ′ is the identity on a free basis, hence θ∘ρ′=id⁡F and ρ′ is injective. Therefore t1,…,tn−1 freely generate F, and applying Ψn−1∘ȷ carries this free basis to the free basis φ(x1),…,φ(xn−1) of Ψn−1(K).

4.1step 2.2step 3.2step 1.2step 2.1step 3.1

The three claims. (i) is step 2.2 together with step 3.2, the free-basis clause of [F2] being what turns Ψn(A1n),…,Ψn(An−1,n) into a basis of the free kernel. For (ii) and (iii), note first that φ(W1) is a product of the pure braids φ(xi) of step 2.2 and that φ(W2)=φ(W1)−1 because φ(W1W2)=1; by (i) the elements φ(x1),…,φ(xn−1) generate ker⁡ρ, so both φ(W1) and φ(W2) lie in the free kernel. The lower-rank word W2 has the same reading in the two models, so the endpoint permutation of its n-strand realization is the identity on the labels 1,…,n−1 (the last strand is fixed) and the permuted labels 1,…,n−1 agree with those of its (n-1)-strand realization; since the permutation of φ(W2) is trivial, the (n-1)-strand realization is a pure braid and step 1.2 identifies φ(W2)=sˇ(φn−1(W2)) in Gnpure. Hence Ψn(φ(W2))∈Ψn(im⁡sˇ)∩ker⁡ρ={1} by step 3.1, so φ(W2)=1; then injectivity of sˇ (step 2.1) gives φn−1(W2)=1, which is (iii), and φ(W1)=φ(W2)−1=1. Finally, since x1,…,xn−1 are carried to a free basis of ker⁡ρ by (i), the homomorphism from the free group on the letters x1,…,xn−1 to Gnpure sending xi↦φ(xi) is injective; a word in this alphabet whose image is φ(W1)=1 is therefore freely trivial, and each cancellation of an adjacent pair xi±1xi∓1 expands, after writing out both expanded words xi±1, into a sequence of permitted deletions of adjacent inverse σ-pairs. This is (ii). ∎

Remarks

  • The load-bearing in-run inputs are the batch-21 items The Fadell-Neuwirth short exact sequence for pure braids, The Ain are meridian generators of the forgetful free kernel and Standard geometric pure braid generators A_ij; steps 2.1 and 3.1 in particular use the convention that the forgetting map is realized by the coordinate projection, which the current statement of The Fadell-Neuwirth short exact sequence for pure braids asserts (the map induced by (x1,…,xn)↦(x1,…,xn−1)); the suppliers' certification remains the owner-held obligation before the item is accepted.
  • The combing identity of step 1.1 is the free-cancellation identity Q~j≡δjσj2γj, hence P~i=Q~i+1≡δi+1σi+12γi+1, and no relation of the Artin presentation is used in it; applying φ gives the geometric conjugation with Pi in step 2.2.
  • The axiom of choice is used only through the batch-21 Fadell--Neuwirth inputs and the identification of the free kernel; the combing identity, the θ argument and the two-model comparison of steps 1.1, 2.1 and 3.1 are choice-free.
TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

The Artin presentation is complete for geometric braids

Statement

Assume AC. For every n≥1 the published surjection φn ⁣:BnArtin→Bngeom of The Artin presentation surjects onto the geometric braid group is an isomorphism. Equivalently, every word in σ1±1,…,σn−1±1 whose geometric braid is trivial is equivalent to the empty word using only the two Artin relations and free insertions and deletions of adjacent inverse pairs, so that the Artin presentation of The braid group by Artin presentation is a presentation of the geometric braid group.

Facts & Assumptions

Given: A natural number n≥1; the abstract Artin group BnArtin=⟨σ1,…,σn−1∣σiσi+1σi=σi+1σiσi+1 (1≤i≤n−2), σiσj=σjσi (∣i−j∣>1)⟩ of The braid group by Artin presentation, trivial for n≤1; the geometric braid group Bngeom=Gn of The isotopy classes of geometric braids based at Q form a group, and the endpoint permutation is a homomorphism; and the surjective homomorphism φn of The Artin presentation surjects onto the geometric braid group, which sends the abstract letter σi to the class of the elementary geometric half twist.

[F1]

Permitted moves. Two words in the letters σ1±1,…,σn−1±1 are called equivalent when one can be obtained from the other by a finite sequence of the following operations: replacing a subword σiσi+1σi by σi+1σiσi+1 or conversely; replacing a subword σiσj by σjσi or conversely when ∣i−j∣>1; and inserting or deleting a subword σiϵσi−ϵ. Equivalence is an equivalence relation compatible with concatenation, and equivalent words represent the same element of BnArtin and, through φn, the same geometric braid. (The braid group by Artin presentation, Group presentation by generators and relations.)

[F2]

Combing. Assume n≥2 and let W be a word in σ1±1,…,σn−1±1 whose image under φn is the trivial geometric braid. Then W is equivalent, by the permitted moves of [F1], to a product W1W2 in which W1 is a word in x1±1,…,xn−1±1 and W2 is a word in σ1±1,…,σn−2±1, where xj=αj+1−1σj2αj+1 are the combing words of The Zariski combing words alpha_i and x_i in the Artin presentation (Every trivial braid word combs as W_1W_2).

[F3]

Uniqueness. Assume AC, n≥2, and that W1 is a word in x1±1,…,xn−1±1 and W2 a word in σ1±1,…,σn−2±1 with φn(W1W2)=1. Then φn(W1)=1, the word W1 reduces to the empty word by free cancellations of adjacent inverse pairs xj±1xj∓1, each of which expands into permitted deletions of σ-pairs; and φn−1(W2)=1, where φn−1 ⁣:Bn−1Artin→Bn−1geom is the rank n−1 surjection applied to the same word read on n−1 strands (The combed geometric decomposition is unique).

[F4]

AC holds, and AC implies dependent choice and countable choice (The Axiom of Choice, AC implies DC implies countable choice); this is the hypothesis under which [F3] is available. The free kernel of the forgetting map PBn→PBn−1 is free with basis A1n,…,An−1,n for the standard pure braid generators Aij of Standard geometric pure braid generators A_ij (The Ain are meridian generators of the forgetful free kernel, The Fadell-Neuwirth short exact sequence for pure braids), and Ψm ⁣:Gmpure→PBm is the canonical isomorphism at every rank m (Pure geometric braids and ordered configuration loops).

Proof

technique · direct induction on $n$
1.1F1

Base case. For n=1 there is no index i with 1≤i≤n−1, so the only word in the displayed alphabet is the empty word, and it is equivalent to itself by the empty sequence of permitted moves; the claim holds at n=1.

2.1F2F3step 1.1

Induction step. Assume n≥2, that the claim holds at rank n−1, and let W be a word in σ1±1,…,σn−1±1 with φn(W)=1. By [F2] the word W is equivalent to a product W1W2 with W1 in the x-letters x1±1,…,xn−1±1 and W2 in the lower-rank letters σ1±1,…,σn−2±1; since equivalence is compatible with concatenation and does not change the geometric braid, φn(W1W2)=φn(W)=1. By [F3] applied to the pair (W1,W2), the word W1 reduces to the empty word by free cancellations of adjacent inverse x-pairs, each of which expands into permitted deletions of σ-pairs, so W1 is equivalent to the empty word by the moves of [F1]; and φn−1(W2)=1. The word W2 lies in the alphabet σ1±1,…,σn−2±1 of the rank n−1 presentation, so the induction hypothesis applies to it: W2 is equivalent to the empty word using the rank n−1 moves. Every rank n−1 move is also a permitted rank n move of [F1], because the generators σ1,…,σn−2 with the braid and far-commutation relations among them are part of the rank n presentation, and the intermediate free insertions and deletions are the same operation. Hence W≡W1W2≡W2≡1, so the claim holds at rank n.

3.1F4step 2.1

Conclusion. By steps 1.1 and 2.1, for every n≥1 each word in σ1±1,…,σn−1±1 whose image under φn is trivial is equivalent to the empty word; since equivalent words represent the same element of BnArtin, the kernel of φn is trivial. The published proposition gives that φn is surjective, so φn is an isomorphism of groups. ∎

Remarks

  • The first nontrivial rank is n=2: there the free kernel of the forgetting map PB2→PB1 is all of PB2, freely generated by the single standard generator A12=σ12 by The Ain are meridian generators of the forgetful free kernel, and x1=σ12 with P1 the empty product; the uniqueness clause of [F3] therefore has content already at n=2, where it says that a word in x1±1 with trivial geometric image is freely trivial.
  • No injectivity of any Artin presentation is assumed anywhere: the induction reduces words in the kernel to the empty word, and injectivity is a conclusion. The only use of AC is through [F3], whose suppliers invoke dependent and countable choice.
CorollaryStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

All four classical braid models realize the Artin presentation

Statement

Assume AC and n≥1. Write BnArtin for the Artin-presentation group of The braid group by Artin presentation, Gn=Bngeom for the geometric braid group at the base tuple Qn of The elementary geometric half twist, its support disc, and its opposite, Bnconf:=π1(Cn(D2),[Qn]),π1(Cn(int⁡D2),[Qn]) for the unordered configuration-space fundamental groups of Unordered configuration spaces Cn(X) at the same base configuration, the second identified with the first by the open-to-closed inclusion, and Mod⁡(D2,Qn;∂D2) for the boundary-fixed punctured-disk mapping class group of Boundary-fixed mapping class group of a punctured disk. Then:

  1. the four models BnArtin, Gn, Bnconf≅π1(Cn(int⁡D2),[Qn]) and Mod⁡(D2,Qn;∂D2) are pairwise connected by the canonical isomorphisms: the completeness isomorphism φn of The Artin presentation is complete for geometric braids, the published inverse-loop isomorphism of The geometric and configuration braid models agree at the fixed base configuration, the open-to-closed identification, and the AC-dependent boundary-fixed mapping-class isomorphism of Braid group as boundary-fixed punctured-disk mapping classes (an in-run batch-20 scaffold, not a published supplier);
  2. under these identifications, for every 1≤i≤n−1 the Artin generator σi corresponds to the class [σi] of the elementary geometric half twist, to the configuration loop class Φ([σi]) whose endpoint monodromy is the adjacent transposition (i i+1), and to the class [Hi] of the half twist supported near the i-th and (i+1)-st punctures, that is, of the explicit boundary-fixed homeomorphism supported in the disc Ui and exchanging qi and qi+1;
  3. consequently each of the four models carries the Artin presentation with these corresponding generators: for each model the assignment σi↦ its generator extends to a group isomorphism from BnArtin onto the model, so the model is presented by the generators σ1,…,σn−1 subject to the two Artin relations and to no further relations.

For n=1 there is no generator, all four groups are trivial, and clauses 2 and 3 are vacuous.

Facts & Assumptions

Given: AC, an integer n≥1, the Artin-presentation group BnArtin=⟨X∣R⟩ of The braid group by Artin presentation with generating set X={σ1,…,σn−1} and its two families of Artin relators interpreted in the sense of Group presentation by generators and relations, the four models of the statement, and an index i with 1≤i≤n−1.

[F1]

For every n≥1 the surjection φn ⁣:BnArtin→Gn of The Artin presentation surjects onto the geometric braid group is an isomorphism, and it carries each generator σi to the class [σi] of the elementary geometric half twist of The elementary geometric half twist, its support disc, and its opposite; the endpoint permutation of that class is the transposition of i and i+1 (The Artin presentation is complete for geometric braids, The Artin presentation surjects onto the geometric braid group, The elementary geometric half twist, its support disc, and its opposite).

[F2]

The inverse-loop slicing map Φ ⁣:Gn→Bnconf=π1(Cn(D2),[Qn]), Φ([β])=(ι∗C[S(β)])−1, is a group isomorphism intertwining the endpoint maps, πconf(Φ([β]))=πgeo([β]) for every [β]∈Gn; and the open-to-closed inclusion induces an isomorphism ι∗C ⁣:π1(Cn(int⁡D2),[Qn])→π1(Cn(D2),[Qn]) at the same basepoint (The geometric and configuration braid models agree at the fixed base configuration, The geometric endpoint permutation matches covering monodromy, The interior-disc and closed-disc configuration spaces are homotopy equivalent).

[F3]

The composite Ψ:=δ∘(ι∗C)−1∘Φ ⁣:Gn→Mod⁡(D2,Qn;∂D2) is a group isomorphism, and for every 1≤i≤n−1 it satisfies Ψ([σi])=[Hi], where Hi is the explicit boundary-fixed homeomorphism supported in the support disc Ui of The elementary geometric half twist, its support disc, and its opposite and exchanging qi and qi+1. The theorem supplying Ψ is an in-run batch-20 scaffold of this run, not a published supplier (Braid group as boundary-fixed punctured-disk mapping classes, Boundary-fixed mapping class group of a punctured disk).

[F4]

AC holds, and AC implies dependent choice and countable choice (The Axiom of Choice, AC implies DC implies countable choice); this is the hypothesis under which the mapping-class isomorphism of [F3] and the free-kernel suppliers of the completeness theorem of [F1] are available.

[F5]

Presentation transport along an isomorphism. If a group has a presentation B=⟨X∣R⟩=F(X)/⟨ ⁣⟨R⟩ ⁣⟩F(X) and θ ⁣:B→M is a group isomorphism, then the composite qθ ⁣:F(X)→B→ θ M of the quotient map with θ is a surjective homomorphism with kernel ⟨ ⁣⟨R⟩ ⁣⟩F(X): surjectivity is clear, and qθ(w)=eM holds exactly when q(w)∈ker⁡θ={eB}, that is, exactly when w∈⟨ ⁣⟨R⟩ ⁣⟩F(X). The first isomorphism theorem therefore gives M≅F(X)/⟨ ⁣⟨R⟩ ⁣⟩F(X)=⟨X∣R⟩, the isomorphism carrying the class of each x∈X to θ([x]) (Group presentation by generators and relations, First isomorphism theorem for groups: G/ker⁡f≅im⁡f, The braid group by Artin presentation).

Proof

technique · direct
1.1F1

The abstract and geometric models. By [F1] the map φn ⁣:BnArtin→Gn is a group isomorphism and φn(σi)=[σi] for every i, so the abstract model and the geometric model are identified generator by generator.

2.1F1F2step 1.1

The two configuration models. By [F2] the inverse-loop slicing map Φ ⁣:Gn→Bnconf is a group isomorphism with πconf∘Φ=πgeo, and ι∗C is an isomorphism π1(Cn(int⁡D2),[Qn])→Bnconf; hence the composites Φ∘φn and (ι∗C)−1∘Φ∘φn, being composites of group isomorphisms, are group isomorphisms from BnArtin onto Bnconf and onto π1(Cn(int⁡D2),[Qn]) respectively. The generator σi is carried to Φ([σi]), whose endpoint monodromy is πconf(Φ([σi]))=πgeo([σi]), the transposition of i and i+1 by [F1] and [F2]; in the open-disc model it is carried to (ι∗C)−1Φ([σi]), the same configuration loop class read through the inclusion.

2.2F3step 1.1

The mapping-class model. By [F3] the composite Ψ is a group isomorphism Gn→Mod⁡(D2,Qn;∂D2) with Ψ([σi])=[Hi], so Ψ∘φn is a group isomorphism BnArtin→Mod⁡(D2,Qn;∂D2) carrying σi to [Hi], the class of the boundary-fixed half twist supported in Ui that exchanges qi and qi+1.

3.1F5step 1.1step 2.1step 2.2

Presentation transport to each model. Let R be the set of the two families of Artin relators, so that BnArtin=⟨X∣R⟩=F(X)/⟨ ⁣⟨R⟩ ⁣⟩F(X) by The braid group by Artin presentation and Group presentation by generators and relations. Apply [F5] to the identity isomorphism of BnArtin and to the isomorphisms of steps 1.1, 2.1 and 2.2: the identity on BnArtin, φn, Φ∘φn, (ι∗C)−1∘Φ∘φn and Ψ∘φn. Each of the five models is therefore isomorphic to F(X)/⟨ ⁣⟨R⟩ ⁣⟩ through the composite of the quotient map with that isomorphism, with the class of σi mapping to the corresponding generator displayed in steps 1.1, 2.1 and 2.2; in particular each model is generated by those n−1 elements and satisfies no relation among them beyond the Artin relators.

4.1F1F3F4step 1.1step 2.1step 2.2step 3.1

Conclusion and the one-strand case. Steps 1.1, 2.1 and 2.2 identify the four models pairwise through the stated isomorphisms and track σi to the half twist, to the loop of monodromy (i i+1) and to the supported half twist, and step 3.1 transports the presentation to each of them, proving all three clauses for n≥2. For n=1 the index range 1≤i≤n−1 is empty, so the generator clauses are vacuous, and B1Artin is the trivial group given by the empty presentation by The braid group by Artin presentation; the isomorphisms of [F1]–[F3] then identify the other three models with it, so each of the four models is trivial and carries the empty presentation of the trivial group, which is clause 3 at n=1. AC enters only through [F3] and through the free-kernel suppliers of the completeness theorem recorded in [F4]. ∎

Remarks

  • The corollary does not reprove the mapping-class or configuration identifications: it composes them with the completeness theorem and tracks the generator through the composite. Its only genuinely new input beyond the suppliers is the bookkeeping that the generator correspondence survives each composite, which is why the configuration and mapping-class models inherit the Artin presentation.
  • The mapping-class isomorphism is an in-run batch-20 draft (Braid group as boundary-fixed punctured-disk mapping classes, precheck PASS; not a published supplier), flagged in the dispatch report together with the consuming step 2.2 and the cross-batch edge recorded in frontier-37-owner-30-batch-22.cross-batch-dependencies.json; no published theorem supplies it.
  • The construction is choice-free apart from the AC hypothesis: the presentations are finite, the free group on X is explicit, and no connecting path or lift is chosen in the composites, all of which use the fixed basepoint Qn of the suppliers.

5 · Examples, counterexamples and false statements

None yet.

Sources