On free generators in the Grothendieck-Teichm\"uller Lie Algebra

Thomas Willwacher

On free generators in the Grothendieck-Teichm\"uller Lie Algebra

T0 review · 0 major / 4 minor · reviewed 2026-08-01 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Over any field of characteristic zero, any family of homogeneous odd-weight elements σ_{2k+1} in the Grothendieck–Teichmüller Lie algebra with nonzero coefficient of x^{2k}y generates a free Lie subalgebra.

desk verdict Careful, honest re-proof of Brown's freeness theorem with a genuine extension to arbitrary normalized families; the proof hinges on Terasoma's identity, which is the main external input. read the letter →

arxiv 2607.18793 v2 pith:2NW4NGZS submitted 2026-07-21 math.QA

Thomas Willwacher This is my paper · ORCID

classification math.QA MSC 17B0111M3214G32

keywords Grothendieck–TeichmüllerLiealgebrafreesubalgebraDrinfeldassociatorBrown–ZagieridentityLyndonwords2-adicvaluationDeligne–DrinfeldconjectureIharabracket

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proves a strengthened half of the Deligne–Drinfeld conjecture about the Grothendieck–Teichmüller Lie algebra grt1. The theorem states that over any field of characteristic zero, if you pick elements σ_3, σ_5, ... of weights 3,5,7,... with arbitrary higher terms but a nonzero coefficient of x^{2k}y in each σ_{2k+1}, these elements always generate a free Lie subalgebra. In other words, the natural map from the abstract completed free Lie algebra on symbols s_{2k+1} to grt1 sending s_{2k+1} to σ_{2k+1} is injective. This removes the earlier need to choose special 'motivic' generators (as in Brown's theorem) and matters for applications where only the leading coefficients of the generators are accessible.

What carries the argument

The proof builds a matrix P_{N,p} indexed by Lyndon words in the alphabet X = {3,5,7,...}, whose entries are evaluations at the identity of iterated derivations ∂ acting on orbit functions Z(ϕ(u')) derived from a fixed even Drinfeld associator. The decisive input is Terasoma's universal Brown–Zagier identity (Theorem 4.2), which determines certain rational coefficients c_{a,b} (coefficients of (xy)^b x^2 y (xy)^a in σ_{2k+1}) and their 2-adic valuations. These valuations make the matrix triangular in the Lyndon order, with strictly larger valuations above the diagonal; a standard valuation criterion then shows P_{N,p} is invertible over every characteristic-zero field. Invertibility of this

What would settle it

Compute the determinant of the pairing matrix P_{N,p} (65) for the smallest case with more than one Lyndon word, e.g., N=13, p=3 with the two Lyndon words (3,3,7) and (3,5,5), using the universal coefficients c_{a,b} from (41)-(43); the proof predicts a nonzero determinant over Q. A zero determinant, or an explicit nonzero Lie polynomial in the kernel of ρ_Lie, would falsify Theorem 1.2. At a more fundamental level, exhibiting a single associator over a characteristic-zero field for which the Brown–Zagier relation (44) fails would invalidate the cornerstone of the proof.

Watch

Extended reading notes

Core claim

The central discovery is that freeness of a family of odd-weight generators in the Grothendieck–Teichmüller Lie algebra is determined solely by the leading monomial coefficients. More precisely, Theorem 1.2 asserts: for every field K of characteristic zero and every choice of homogeneous elements σ_{2k+1} ∈ grt1(K) of weight 2k+1 satisfying [x^{2k}y]σ_{2k+1} ≠ 0 for all k≥1, the σ's generate a free Lie subalgebra of grt1(K). This is an 'any choice' strengthening of a theorem of Brown, which had exhibited one particular generating family with the freeness property.

Load-bearing premise

The whole proof hangs on Terasoma's theorem that the Brown–Zagier identity holds for every Drinfeld associator over every characteristic-zero field with the stated rational coefficients; if that identity failed for even one associator or its coefficients had different 2-adic valuations, the triangularity of the pairing matrix — and with it the freeness conclusion — would collapse.

Editorial extensions

If this is right

The 'freeness' half of the Deligne–Drinfeld conjecture now holds for any choice of odd-weight generators with nonzero leading coefficients, not just a specially selected set.
Applications that use free subalgebras of grt1 — such as lower bounds on cohomology of GL_n and moduli spaces of abelian varieties — can extend from a limited weight range to all weights, since only the leading coefficients need to be verified.
The proof gives a rational, motivically elementary route to Brown's theorem, so the statement can be checked and re-used without deep motivic machinery.
Exponentiating, the corresponding homomorphism from the completed free prounipotent group to GRT1 is also injective, so the group-level freeness statement follows as well.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A natural stress-test is whether the same valuation argument works when the leading monomial is changed (e.g., a different coefficient in σ_{2k+1}); the proof's structure suggests the freeness phenomenon may be tied to the 2-adic valuations of universal coefficients, not to the particular monomial x^{2k}y.
Since Terasoma's identity is the only external ingredient, a direct elementary proof of the Brown–Zagier identity for associators would remove the motivic black box entirely and lead to a fully self-contained freeness theorem.
The technique may transfer to related pronilpotent Lie algebras (e.g., versions of grt associated to other operads or to genus-one curves) whenever a universal coefficient identity with the same valuation properties is available.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The paper does what it says: it proves that any homogeneous family in grt1 with nonzero leading coefficients generates a free Lie subalgebra, extending Brown's existence result. The main technical machinery is Brown's, as the author freely acknowledges, but the formulation for arbitrary families is new and has a real payoff—removing the N=11 restriction in Brown–Chan–Galatius–Payne.

What's good: the proof is detailed and coherent. The cut formula, the kernel lemma, the level-lowering argument, and the 2-adic triangularity of the Lyndon pairing matrix are all written out with enough care that a reader can check them. The paper is refreshingly honest about provenance: it states plainly that the main ideas are Brown's and that Theorem 1.2 is a modest improvement. That is accurate.

The soft spot is exactly what the stress-test note identifies: the load-bearing external input is Terasoma's universal Brown–Zagier identity, quoted rather than reproved. The 2-adic valuation calculations that make the pairing matrix triangular depend entirely on the exact formula (44). I see no reason to doubt the formula—Terasoma's paper is well-established—but if it contained a mis-stated binomial index or factor, the proof would break. That is a residual risk, not a flaw; citing a theorem is normal practice, and this dependency could have been flagged more prominently but is not hidden.

Minor quibbles: the exposition is terse in a few places, especially the transition from Proposition 5.2 to Proposition 6.2, but it is followable. The paper does not rederive the rational even associator, again a standard external input.

I agree with the reader's take: accept at moderate confidence. Novelty and significance are in the 5–6 range because the improvement is genuinely useful for applications where only leading coefficients are controllable, but it is not a paradigm shift. The self-assessment is accurate.

Who is this for? Specialists in Grothendieck–Teichmüller theory, motivic periods, and cohomology of GL_n and A_g. It deserves a serious referee; I would send it to an expert who can check the 2-adic computations and the transcription of (44). My recommendation: engage with it and cite it. If a referee asks for anything, it should be a careful derivation or independent verification of the quoted Terasoma identity in the form used here.

Referee Report

0 major / 4 minor

Summary. The paper proves Theorem 1.2: for any characteristic-zero field K, any homogeneous family σ3, σ5, ... in the Grothendieck–Teichmüller Lie algebra grt1(K), with nonzero coefficient of x^{2k}y in σ_{2k+1} for every k, generates a free Lie subalgebra. The proof follows Brown's strategy: it builds a source free Lie algebra on symbols s_{2k+1}, defines orbit functions through an even Drinfeld associator and the action of the source group, imports Terasoma's universal Brown–Zagier identity to compute the relevant level-one coefficients and their 2-adic valuations, establishes a level-lowering cut formula, and proves 2-adic triangularity of a Lyndon pairing matrix. The final section converts this triangularity into injectivity of the map from the completed free Lie algebra into grt1(K).

Significance. If correct, the theorem is a genuine improvement over the form of Brown's theorem commonly quoted: it removes any need to control the higher coefficients of the chosen generators, so it applies to arbitrary normalized families. This is useful for applications such as the Brown–Chan–Galatius–Payne cohomology bounds mentioned in Remark 1.3. The proof is detailed, internally coherent, and mostly elementary; the use of the cut formula, the kernel lemma, and the 2-adic valuation criterion is clean. The paper is transparent about its external inputs: Drinfeld's existence theorems and Terasoma's identity (Theorem 4.2) are quoted rather than proved. The stress-test concern about Theorem 4.2 lands in the sense that (44) is the single concentrated external dependency: a mis-stated coefficient would propagate through Propositions 4.1–4.3 and break the triangularity. However, this is a dependency and not an internal gap; the authors identify it explicitly, and I found no circularity or missing internal step.

minor comments (4)

[Section 3.1] The notation for the source algebra, especially "K⟨f3,f5,f7,...⟩ ∐", is confusing. The accompanying sentence says juxtaposition denotes a basis word and multiplication is the shuffle product, but the symbol "∐" is not defined. Please replace it with a standard notation and explicitly define the shuffle product on words.
[Section 4.1, Theorem 4.2] Because the entire 2-adic valuation argument depends on the exact coefficients in (44), the reader needs to verify the quotation. Please give the precise theorem number or equation in [6] and state explicitly that (44) is quoted verbatim from Terasoma's paper. This is particularly important since the identity is not rederived here.
[Section 4.1, proof of Proposition 4.1] The sentence "with a cut of the full length 2k+1" is terse. It would help to spell out how formula (28) identifies [f_{2k+1}] Z(2^a,3,2^b) with c_{a,b}, since this is the key coefficient comparison.
[General editorial] The arXiv text contains several typographical artifacts, including "existenvce" and recurring garbled symbols such as "leftr⫯g⊸tl⫯ne". These should be cleaned before publication.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; all load-bearing inputs are external theorems about associators, not about the freeness being proved.

full rationale

The paper's conclusion (Theorem 1.2) is that arbitrary homogeneous sigma_3, sigma_5, ... with nonzero leading coefficients generate a free Lie subalgebra. The proof does not invoke Theorem 1.1 or any equivalent freeness statement as a premise. Instead its chain is: (1) choose a normalized family, giving the source map rho (Def. 3.1, (10)-(11)); (2) use the cut formula (Prop. 3.2), an internal combinatorial identity, to obtain the action of partial_m on coefficient functions; (3) import Terasoma's universal Brown-Zagier identity (Theorem 4.2, quoted from [6]) for associators; (4) derive from it the universal coefficients c_{a,b} and c^partial_k (Prop. 4.1) and their 2-adic valuations (Prop. 4.3); (5) prove triangularity/invertibility of the Lyndon pairing matrix P_{N,p} (Prop. 6.4) from those valuations; (6) apply this invertibility to kill any hypothetical kernel element (Sec. 7). Each step is a concrete algebraic implication. The only load-bearing external input, Terasoma's identity, concerns associator coefficients, not grt_1 freeness; it is not derived from the target theorem. Drinfeld's existence of normalized elements and even associators (Lemmas 2.1, 2.2) is likewise external. The citation of Brown [2] is for the 'deconcatenation modulo I' strategy and is not used as an unproved premise; the self-citation [7] is only an expository reference for an existence construction whose source is Drinfeld. No fitted parameters are renamed as predictions, and no equation of the paper equates the freeness conclusion with its input. Thus there is no circular step.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

The proof introduces no fitted constants or new postulated entities. The arbitrary inputs are the choice of homogeneous family σ and an even associator, but the theorem shows the freeness conclusion is independent of all choices beyond the leading coefficient. The universal coefficients c_{a,b} and c^∂ are derived, not tuned.

assumptions (6)

standard math The defining equations (3)–(5) and Ihara bracket make grt1(K) a weight-graded pronilpotent Lie algebra; weight spaces satisfy grt1(K)_n = grt1(Q)_n ⊗_Q K.
Background from Drinfeld [4]; used throughout to pass from Q to arbitrary characteristic-zero fields (eq. 6).
domain assumption There exist elements σ_{2k+1} ∈ grt1(Q) with [x^{2k}y] σ ≠ 0 (Lemma 2.1).
Existence of the family; the theorem would be vacuous without it. Quoted from Drinfeld and the author's lecture notes [7].
domain assumption There exists a rational even Drinfeld associator Φ_ev (Lemma 2.2).
Fixes the base point used to build coefficient functions; evenness gives parity identities such as (34).
domain assumption GRT1(K) acts freely and transitively on the set of Drinfeld associators.
Used to define orbit functions (23) and in the proof of the cut formula Proposition 3.2.
domain assumption Terasoma's Brown–Zagier identity (Theorem 4.2) holds universally for every associator, with rational coefficients.
The load-bearing external input; it fixes c_{a,b} in Proposition 4.1 and the 2-adic valuations in Proposition 4.3.
standard math Lyndon basis theorem: the br(u) are a basis of the free Lie algebra and expand as u plus higher words (eq. 56).
Used in Section 6 to set up the invertible pairing matrix P_{N,p}; standard result cited to Reutenauer [5].

how reviews work

0 comments

Cite this review

Pith. "Pith review of On free generators in the Grothendieck-Teichm\"uller Lie Algebra." pith.science (2026). https://pith.science/paper/2NW4NGZS

@misc{pith2026260718793,
  author       = {Pith},
  title        = {Pith review of: On free generators in the Grothendieck-Teichm\"uller Lie Algebra},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2NW4NGZS}},
  note         = {Machine review of arXiv:2607.18793}
}

abstract

We prove that any homogeneous family $\sigma_3,\sigma_5,\ldots$ in $\mathfrak{grt}_1(\mathbb K)$ with nonzero coefficients of $x^{2k}y$ in $\sigma_{2k+1}$ generates a free Lie subalgebra, building on the work of Brown.

Discussion (0). Continue with ORCID to comment.

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.