REVIEW 3 major objections 4 minor 1 cited by
The instanton homology of the $(-2,3,q)$ pretzel knots and computed bounding cochains in the pillowcase
T0 review · 3 major / 4 minor · reviewed 2026-08-03 · deepseek-v4-flash
Pith's one-line read For every odd q≥3, the pretzel knot P(−2,3,q) has reduced singular instanton homology of rank exactly q+2.
desk verdict Theorem 1.1 is a clean, new, unconditional rank formula; the pillowcase half is honest, conditional computation that should be read as evidence within a model, not as theorem. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The proof rests on the squeeze inequality ℓ(K) ≤ rank I♮(K) ≤ dim Kh_r(K), fed by two closed forms: the Alexander polynomials of the family, derived from a Chebyshev-type skein recursion and having L1-norm q+2, and the reduced Khovanov homology of 3-strand pretzels, with total rank q+2. On the pillowcase side, the machinery is immersed-curve polygon counting by winding numbers: the differential and higher products μ_k are computed as counts of immersed bigons, triangles, quadrilaterals, and larger polygons with convex corners and winding conditions, and the deformed differential is corrected by bounding cochains satisfying a Maurer–Cartan equation.
What would settle it
For the unconditional theorem, find an odd q≥3 where the Alexander polynomial L1-norm or the reduced Khovanov homology rank is not q+2, or compute rank I♮ directly and obtain a different value; for the pillowcase side, compute the true immersed Floer homology of the pillowcase Lagrangians for q=11 analytically, or compute the naive rank for q=17 and check whether naive−I♮ equals +2.
Extended reading notes
Core claim
The central discovery is the unconditional rank formula rank I♮(P(−2,3,q)) = q+2 for all odd q≥3, obtained by combining a closed-form family of Alexander polynomials (Lehmer-like, with coefficients in {0,±1}) with closed-form reduced Khovanov homology; the two bounds coincide. On the symplectic side, within the immersed-curve combinatorial model of the pillowcase, the paper finds an experimental deficiency law: the naive Lagrangian–Floer rank differs from the true instanton rank by 2·sgn(det K−3), so the naive theory is too small by one differential for q=5,7, exact for q=3, and too large by one differential for q≥11. It then computes bounding cochains that correct the naive theory: a unique
Load-bearing premise
The pillowcase computations and the deficiency law rest on the unproved identification that winding-number counts of immersed polygons compute the higher operations of the immersed Fukaya algebra of the pillowcase orbifold; if that hypothesis fails, the computed ranks and bounding cochains do not describe genuine Floer theory.
Editorial extensions
If this is right
- The rank formula gives the exact size of reduced singular instanton homology for every odd q≥3, extending the previously known q=5 case to the whole family without any conjecture.
- If the deficiency law holds beyond the tested members, the naive pillowcase Floer homology is never off by more than one differential for this family, with the sign determined by det K−3 (equivalently, by the number of binary-dihedral traceless characters).
- The computed bounding cochains are the first explicit nonzero cochains on Conway-sum tangles, and the first acting by cancellation; they realize both directions of the conjectured repair within one family.
- Cancellation cochains are unique, while creation cochains are numerous, so rank-matching alone cannot pin down the tangle's canonical cochain at excess members; additional naturality data must be invoked.
- The correction acts through different polygon orders within one family (a quadrilateral for q=5, a triangle for q=7), so the polygon order is not an invariant of the direction of the correction.
Reading between the lines
- If the deficiency law holds for all odd q coprime to 3, the discrepancy between naive pillowcase Floer homology and true instanton homology would be a value in {−1,0,1} controlled by the determinant, possibly provable by a curve-level argument once the combinatorial model is upgraded to a theorem.
- The rigidity asymmetry suggests that at excess members the deformed Floer homology may depend on the choice of bounding cochain, so the invariant-theoretic cochain (if it exists) must be selected by additional structure; this could be tested by computing a naturality-fixed cochain for q=11 from a different tangle decomposition and comparing ranks.
- A direct analytic computation of the pillowcase Lagrangian Floer homology for q=11 or q=13 would either confirm or refute Hypothesis 4.1 and the deficiency law; computing the naive rank for q=17 (det=11) in the model would test the law's extrapolation.
- The coincidence of the deficit regime with the integral-homology-sphere members (det=1) hints that the correction mechanism may be tied to the presence of binary-dihedral characters at the pillowcase seams, and could be studied via the cut-and-paste resolution of the seam circles.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper has two distinct parts. The first proves Theorem 1.1: for every odd q ≥ 3, the reduced singular instanton knot homology of the pretzel knot P(-2,3,q) has rank q+2. The proof is a squeeze: Proposition 2.1 computes the Alexander polynomials of the family via a Conway skein recursion, showing their L1 norm is q+2, and Manion's closed-form reduced Khovanov homology gives the matching upper bound q+2. The second part works in the immersed-curve pillowcase model of Herald--Kirk--Smith. It reconstructs the relevant Lagrangians, computes naive Lagrangian--Floer ranks for q = 3,5,7,11,13, formulates a deficiency law (Table 1, Computation 1.2), and computes bounding cochains for q = 5,7,11 that repair the rank to q+2 (Computation 1.3). The paper explicitly separates what is proved unconditionally, what is computed within the combinatorial model, and what remains conjectural.
Significance. Theorem 1.1 is a clean, family-wide rank formula and a good illustration of the Alexander-polynomial/Khovanov squeeze; it is unconditional, elementary modulo [Man18] and [Hir01], and appears to be new in this uniform form. The pillowcase part is potentially valuable: these would be the first explicitly computed nonzero bounding cochains on Conway-sum tangles, and the q=5/q=7 examples give the first instances acting by cancellation. The manuscript is unusually candid about the status of its claims and ships reproducible code. However, the pillowcase ranks, the deficiency law, and the bounding-cochain computations all depend on the unproved Hypothesis 4.1 and on perturbation-dependent enumerations checked at only two perturbations per member. Their status is experimental and model-dependent, not theorem-level, and the presentation should not let the unconditional theorem lend borrowed certainty to the pillowcase claims.
major comments (3)
- [§4.1, Hypothesis 4.1] All pillowcase results — the naive ranks of Table 1, Computation 1.2, and Computation 1.3 — are computed under Hypothesis 4.1, which asserts that the winding-number polygon counts of §4.1 compute the higher operations µ_k of the immersed Fukaya algebra of the pillowcase. This extends de Silva–Robbin–Salamon from embedded loops in surfaces to immersed curves in an orbifold and to k ≥ 2, and the paper states that no proof at this generality is known. Since the pillowcase computations are headline contributions, they should not be presented as facts about the actual pillowcase Floer homology. I recommend either adding a theorem that reduces the needed cases to a checkable condition, or explicitly relabeling the pillowcase results as model-dependent experimental evidence throughout the abstract, introduction, and Section 5. Theorem 1.1 is not affected.
- [§5.2, §5.3; Eq. (2)] For q=7 and q=11, the Maurer–Cartan verification is reported only as µ1-closedness of the support crossing. Equation (2) requires the vanishing of all higher terms µ_k(b,...,b). Since b is a single self-intersection point for q=7, the only possible higher terms involve repeated occurrences of the same vertex. Section 4.1 does not explicitly state that polygon counts require distinct vertices, and the code description does not resolve this. Such terms could be nonzero. Please give explicit counts for µ_k(s,...,s), k ≥ 2, for the chosen cochain support s, or prove that no such polygons exist under the definition used.
- [Computation 1.2, Table 1] The sign law 'naive rank − rank I♮ = 2 sgn(det K − 3)' is presented as a 'sharp experimental law' but is fitted to exactly five members q=3,5,7,11,13. The row q=13 is labeled 'predicted' even though its naive rank is already computed. There is no out-of-sample check: the sign and threshold were read off after the computations. This is honest but weaker than the wording suggests. I recommend either computing additional members (e.g. q=17,19) after fixing the prediction, or rephrasing the statement as a conjecture supported by five examples rather than a law.
minor comments (4)
- [§2.1] The elimination leading to the three-term recursion (5) is not shown. The displayed relations a_q = a_{q−2} + sz b_{q−1} and b_{q−1} = b_{q−3} + sz a_{q−2} require an additional use of the skein relation for b_{q−3}; please include the two-line algebra so the reader can verify the recursion without reconstructing it.
- [§3.2, §6.4] For reproducibility, state the perturbation amplitudes and the locality-window size used in the polygon enumerations. Figure 1 gives ε ≈ 0.05–0.07 for one case, but the window of consecutive blue-arc positions in §4.2 is not quantified. The code is provided, but these parameters should be in the text.
- [Table 1] The q=13 row should distinguish what is computed (naive rank 17, rank I♮ = 15) from what is predicted (the sign of the correction). The word 'predicted' in the correction column is confusing because the naive rank is computed in the same table.
- [§5.1] The sentence 'This proves Computation 1.3(i)' would be less likely to confuse if it read 'establishes within the model' or 'verifies', in keeping with the paper's own register distinction.
Circularity Check
No significant circularity: Theorem 1.1 is an independent squeeze against external results, and the pillowcase-side claims are explicitly conditional on a stated hypothesis, not circular.
full rationale
Theorem 1.1 does not reduce to its inputs. The lower bound rank I^# >= q+2 comes from the L1 norm of the Alexander polynomial computed in Proposition 2.1 by an explicit Conway skein recursion, with the normalization checked against Hironaka's independent Lehmer polynomial result; the upper bound rank I^# <= q+2 is Manion's closed-form Khovanov homology. Both bounds are external to the paper's own fitted values, and the spectral sequence and Euler-characteristic facts are standard and not derived from the target theorem. The pillowcase-side computations are explicitly placed in register (b), 'computed within the model', and depend on Hypothesis 4.1, which the paper itself labels as an unproved assumption covering exactly the gap between the winding-number counts and the immersed Fukaya algebra operations. This makes those claims conditional, but conditionality is not circularity. The bounding cochains are found by enumerating supports until the deformed rank matches the already-known target q+2; the paper transparently describes this as rank-matching and does not present the rank equality as an independent prediction. The Maurer-Cartan checks, the uniqueness claims, and the polygon-order content are computed rather than assumed from the target. There is no load-bearing self-citation chain: the companion paper [Wue26] is only contextual, and the main theorem relies on Hironaka and Manion, not on the author's prior work. Overall, the paper's central unconditional claim is self-contained and non-circular.
Assumptions & free parameters
free parameters (3)
- pillowcase perturbation amplitude ε =
two values ≈0.05–0.07 per member
- blue-arc locality window =
bounded window of consecutive blue-arc positions, width not stated
- bounding cochain supports =
q=5: sA+sB; q=7: one crossing; q=11: 55 crossings
assumptions (4)
- ad hoc to paper Hypothesis 4.1: immersed-polygon winding counts compute the μ_k operations of the immersed Fukaya algebra of the pillowcase for k>=2.
- standard math Kronheimer-Mrowka spectral sequence from reduced Khovanov homology to I^♮, and the identification of the graded Euler characteristic of I^♮ with the Alexander polynomial.
- domain assumption Manion's closed-form reduced Khovanov homology for non-quasi-alternating 3-strand pretzels, giving dim Kh^r = q+2 for P(-2,3,q).
- domain assumption CHKK Conjecture 1.1: the deformed pillowcase Floer homology with bounding cochains equals I^♮.
Cite this review
Pith. "Pith review of The instanton homology of the $(-2,3,q)$ pretzel knots and computed bounding cochains in the pillowcase." pith.science (2026). https://pith.science/paper/XIL7OAZ3
@misc{pith2026260726096,
author = {Pith},
title = {Pith review of: The instanton homology of the $(-2,3,q)$ pretzel knots and computed bounding cochains in the pillowcase},
year = {2026},
howpublished = {\url{https://pith.science/paper/XIL7OAZ3}},
note = {Machine review of arXiv:2607.26096}
}
abstract
We prove that the reduced singular instanton knot homology of the pretzel knots $P(-2,3,q)$ has rank $q+2$ for every odd $q\ge 3$: the Alexander polynomials of the family, computed in closed form by a skein recursion (Hironaka's Lehmer-like polynomials), give the lower bound $q+2$, and Manion's closed-form reduced Khovanov homology gives the matching upper bound. We then turn to the pillowcase (symplectic) side of the knot Atiyah-Floer program. In the immersed-curve combinatorial model of Herald-Kirk and Smith we reconstruct the pillowcase Lagrangians of the natural tangle decomposition of the family and compute the naive Lagrangian-Floer homology of its members through $q=13$. The outcome is a sharp experimental law: the naive rank differs from $\operatorname{rank} I^\natural$ by exactly one differential, the difference being $2\operatorname{sgn}(\det K-3)$, vanishing for the torus member $q=3$ and changing direction as the determinant $\det=|q-6|$ crosses $3$ (equivalently, as binary-dihedral traceless characters appear). Finally we compute the bounding cochains conjectured by Cazassus-Herald-Kirk-Kotelskiy to repair the deficiency: for $q=5$ a unique two-crossing cochain acting through an immersed quadrilateral, for $q=7$ a unique single crossing acting through a triangle (each cancelling a bigon and raising the rank by $2$), and for $q=11$ single-crossing cochains acting in the opposite direction (creating a differential and lowering the rank by $2$). To our knowledge these are the first computed nonzero bounding cochains on Conway-sum tangles, and the first anywhere acting by cancellation; they realize both directions of the conjectured correction within one family, with a rigidity asymmetry: cancellation admits a unique minimal cochain, creation many. We separate throughout what is proved unconditionally, what is computed within the model, and what remains conjectural.
Figures
Forward citations
Cited by 1 Pith paper
-
Traceless $\mathrm{SU}(2)$ characters and $\mathbb{Z}/4$ instanton gradings for two-bridge and $(3,n)$-torus knots
For two-bridge knots every traceless SU(2) character is binary-dihedral; for (3,n)-torus knots the characters are mostly non-dihedral, with gradings that predict when knot-instanton homology shrinks below the chain complex.
Reference graph
Works this paper leans on
- [1]
-
[2]
D. Auroux, Lagrangian Floer theory, from geometry to algebra and back again, survey, to appear in Proceedings of the ICM 2026. arXiv:2510.22476
arXiv 2026
-
[3]
G. Cazassus, C. Herald, P. Kirk, A. Kotelskiy, The correspondence induced on the pillowcase by the earring tangle, J.\ Topol.\ 15 (2022), 2472--2543. arXiv:2010.04320
arXiv 2022
-
[4]
de Silva, J
V. de Silva, J. Robbin, D. Salamon, Combinatorial Floer homology, Mem.\ Amer.\ Math.\ Soc.\ 230 (2014), no.\ 1080
2014
- [5]
- [6]
- [7]
-
[8]
Hironaka, The Lehmer polynomial and pretzel links, Canad.\ Math.\ Bull.\ 44 (2001), 440--451
E. Hironaka, The Lehmer polynomial and pretzel links, Canad.\ Math.\ Bull.\ 44 (2001), 440--451
2001
Show all 18 references
-
[9]
Kronheimer, T
P. Kronheimer, T. Mrowka, Knot homology groups from instantons, J.\ Topol.\ 4 (2011), 835--918. arXiv:0806.1053
2011 arXiv
-
[10]
Kronheimer, T
P. Kronheimer, T. Mrowka, Khovanov homology is an unknot-detector, Publ.\ Math.\ IH\'ES 113 (2011), 97--208. arXiv:1005.4346
2011 arXiv
-
[11]
Lim, Instanton homology and the Alexander polynomial, Proc.\ Amer.\ Math.\ Soc.\ 138 (2010), 3759--3768
Y. Lim, Instanton homology and the Alexander polynomial, Proc.\ Amer.\ Math.\ Soc.\ 138 (2010), 3759--3768
2010
-
[12]
Lin, A knot invariant via representation spaces, J.\ Differential Geom.\ 35 (1992), 337--357
X.-S. Lin, A knot invariant via representation spaces, J.\ Differential Geom.\ 35 (1992), 337--357
1992
-
[13]
Manion, The Khovanov homology of 3-strand pretzels, revisited, New York J.\ Math.\ 24 (2018), 1076--1100
A. Manion, The Khovanov homology of 3-strand pretzels, revisited, New York J.\ Math.\ 24 (2018), 1076--1100. arXiv:1303.3303
2018 arXiv
-
[14]
Smith, Perturbed traceless character varieties of tangle sums, arXiv:2412.06066
K. Smith, Perturbed traceless character varieties of tangle sums, arXiv:2412.06066
-
[15]
Wehrheim, C
K. Wehrheim, C. Woodward, Floer cohomology and geometric composition of Lagrangian correspondences, arXiv:0905.1368
-
[16]
B. J. Wuebben, Traceless characters and /4 instanton gradings for two-bridge and (3,n) -torus knots, companion paper, arXiv:2607.26095
-
[17]
Zhang, Construction of holomorphic quilts in Cartesian product of closed surfaces, arXiv:2409.19744
Z. Zhang, Construction of holomorphic quilts in Cartesian product of closed surfaces, arXiv:2409.19744
-
[18]
Zhang, Uniqueness of holomorphic quilts lifted from holomorphic bigons on surfaces, arXiv:2504.09284
Z. Zhang, Uniqueness of holomorphic quilts lifted from holomorphic bigons on surfaces, arXiv:2504.09284
Reviewed August 3, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.