betaFeedback

WorksMathematics

Weak chromatic splitting fails at n = p for every odd prime p

Abstract

We prove that for every odd prime pp there is a height nn for which the map L_{n-1}S_p \to L_{n-1}\L_{K(n)}S_p is not injective on homotopy groups, hence not the inclusion of a wedge summand. In particular, for such pairs (n,p)(n,p) the weak chromatic splitting conjecture is false. More precisely, we prove the conjecture fails for (n,p)=(3,3)(n,p) = (3,3) and for all pairs (n,p)(n,p) with p≤n<2(p−1)p \le n < \sqrt{2}(p-1). In particular, the conjecture fails for (n,p)=(p,p)(n,p) = (p,p) for every odd prime pp. This disproof was produced autonomously by Claude in a session working on Freyd's generating hypothesis with no human in the loop. This note is a human write-up of the disproof. We did not contribute any ideas.

Introduction

The chromatic splitting conjecture is one of the central open problems in chromatic homotopy theory. Chromatic homotopy theory divides the stable homotopy groups of spheres into periodic families, and the chromatic splitting conjecture is the principle which guides chromatic reassembly. Namely, by the Hopkins–Ravenel chromatic convergence theorem [16 Thm. 7.5.7], for SpS_p the pp-complete sphere,

Sp≃lim⁡←nLnSpS_p \simeq\lim_{\leftarrow n} L_nS_p

where LnL_n is Bousfield localization at K(0)⊕⋯⊕K(n)K(0) \oplus\cdots\oplus K(n). Furthermore, the chromatic fracture square [11] is a homotopy pullback square

LnSp→LK(n)S↓↓Ln−1Sp→ιLn−1LK(n)S\begin{CD} L_nS_p @>>> L_{K(n)}S \\ @VVV @VVV \\ L_{n-1}S_p @>{\iota}>> L_{n-1}L_{K(n)}S \end{CD}

so that the chromatic layers can be rebuilt inductively from LK(n)SL_{K(n)}S and the lower chromatic localizations, according to the gluing data ι\iota. The weak chromatic splitting conjecture states that ι\iota has a retract, so that chromatic reassembly is simple. The strong chromatic splitting conjecture further describes the fiber of ι\iota in terms of lower chromatic localizations. The conjecture is due to Michael Hopkins based on Shimomura and Yabe’s computations [17] of the homotopy groups π∗L2S\pi_*L_2S at the prime p=5p=5. It appears in print in [11] in its strongest form. The terminology weak versus strong was introduced in [3].

The strong conjecture is known to hold for n=1n = 1 and for n=2n = 2 at p≥3p \ge3 [17, 11, 9]; at (n,p)=(2,2)(n,p) = (2,2) the strong form fails [3] but the weak form holds [5]. In this note, we present the proof that for every odd prime pp, there is a height nn at which the weak chromatic splitting conjecture fails.

Theorem 1.1. Let (n,p)(n,p) be such that p≥5p \ge5 and p≤n<2(p−1)p \le n < \sqrt{2}(p-1), or let (n,p)=(3,3)(n,p) = (3,3). The map

Ln−1Sp⟶Ln−1LK(n)SL_{n-1}S_p \longrightarrow L_{n-1}L_{K(n)}S

is not injective on homotopy groups, hence does not admit a retraction.

The proof is short. For each pair (n,p)(n,p), we exhibit an explicit element of the kernel. For p≥5p \ge5 the element is β1(p−1)2\beta_1^{(p-1)^2}, where β1\beta_1 is the first element of the β\beta-family. For p=3p = 3, the element is β15\beta_1^5.

We summarize the argument here, noting that for experts, this summary will be sufficient to understand the failure of the conjecture. A class x∈π∗Spx \in\pi_*S_p whose image in LK(n)SL_{K(n)}S is trivial, but which is detected in π∗Ln−1Sp\pi_*L_{n-1}S_p must be in the kernel of Ln−1Sp→Ln−1LK(n)SL_{n-1}S_p \to L_{n-1}L_{K(n)}S. If such an xx exists, then this map does not admit a retraction, so it suffices to produce such an xx. When (p−1)∤n(p-1) \nmid n, the homotopy fixed point spectral sequence computing π∗LK(n)S\pi_*L_{K(n)}S has a horizontal vanishing line at s>n2s > n^2, and so any element in the homotopy π∗S\pi_*S whose Adams–Novikov filtration is greater than n2n^2 must be zero in π∗LK(n)S\pi_*L_{K(n)}S. For p≥5p \ge5, β1k\beta_1^k has nonzero image in π∗EOp−1\pi_*EO_{p-1} for k≤(p−1)2k \le(p-1)^2 and hence in π∗Lp−1Sp\pi_*L_{p-1}S_p. Therefore, for p≥5p \ge5, we conclude that β1(p−1)2\beta_1^{(p-1)^2}, whose filtration is 2(p−1)2>p22(p-1)^2 > p^2, is nonzero in π∗Lp−1Sp\pi_*L_{p-1}S_p, but zero in π∗LK(p)S\pi_*L_{K(p)}S. For p=3p = 3, we instead use that β15\beta_1^5 is nonzero in π∗LK(2)S/3\pi_*L_{K(2)}S/3.

A word on notation. For every homotopy ring RR, and x∈π∗Sx \in\pi_*S, we denote the image of xx in π∗R\pi_*R via the unit by the same name. We also abuse notation and use the same name for classes in the Adams–Novikov or homotopy fixed points spectral sequences detecting xx. Theorem 1.1 follows from Theorems 2.6 and 2.10 below.

Acknowledgments. The disproof of the weak chromatic splitting conjecture given in this note was produced autonomously by Claude. We did not contribute any ideas and Claude was not prompted on this problem but on Freyd’s generating hypothesis in stable homotopy theory. Although the relationship between Freyd’s generating hypothesis and the chromatic splitting conjecture is well established, for example, in work of Hopkins–Devinatz [6, 7] (see also [12, 1]), we did not expect Claude to find a counter example to the weak chromatic splitting conjecture, which we both believed to be true. We note recent progress on Freyd’s generating hypothesis recently appeared in Ma–Xu [14].

The result arose in the context of paid consulting with Anthropic. Claude was run internally by Anthropic, using its own computing resources. The transcript was shared with us September 8, 2026. The mathematical exposition was drafted by us, Agnès Beaudry and Tomer Schlank, after we verified the details of Claude’s disproof, on September 12, 2026. The result was announced at 2026 Clay Research Workshop on Telescopic Homotopy Theory and Algebraic K-theory, on September 21, 2026. This text is not AI generated. It was produced by us, using Claude only for copyediting.

The Disproof

2.1. Vanishing in π∗LK(n)S\pi_*L_{K(n)}S. Let ιK(n):Sp→LK(n)S\iota_{K(n)}: S_p \to L_{K(n)}S be the K(n)K(n)-localization.

Lemma 2.1. Let j<nj<n and x∈π∗Spx\in\pi_* \mathbb{S}_p. If ιK(n)(x)=0\iota_{K(n)}(x)=0 and ιK(j)(x)≠0\iota_{K(j)}(x)\ne0, then ιK(j)(x)\iota_{K(j)}(x) maps to zero under the map

LK(j)ιK(n):LK(j)S⟶LK(j)LK(n)S.L_{K(j)}\iota_{K(n)}:L_{K(j)}\mathbb{S}\longrightarrow L_{K(j)}L_{K(n)}\mathbb{S}.

Consequently, this map does not admit a retraction, and neither does

Ln−1ιK(n):Ln−1Sp⟶Ln−1LK(n)S.L_{n-1}\iota_{K(n)}:L_{n-1}\mathbb{S}_p\longrightarrow L_{n-1}L_{K(n)}\mathbb{S}.

Proof. This follows from the commutative diagram

Sp→ιK(n)LK(n)S↓↓Ln−1Sp→Ln−1ιK(n)Ln−1LK(n)S↓↓LK(j)S→LK(j)ιK(n)LK(j)LK(n)S\begin{CD} \mathbb{S}_p @>{\iota_{K(n)}}>> L_{K(n)}\mathbb{S}\\ @VVV @VVV\\ L_{n-1}\mathbb{S}_p @>{L_{n-1}\iota_{K(n)}}>> L_{n-1}L_{K(n)}\mathbb{S}\\ @VVV @VVV\\ L_{K(j)}\mathbb{S} @>{L_{K(j)}\iota_{K(n)}}>> L_{K(j)}L_{K(n)}\mathbb{S} \end{CD}

where the bottom row is the LK(j)L_{K(j)}-localization of the middle row. The left vertical composite is ιK(j)\iota_{K(j)}. By assumption, the image of xx is nonzero in π∗LK(j)S\pi_*L_{K(j)}\mathbb{S} and thus in π∗Ln−1Sp\pi_*L_{n-1}\mathbb{S}_p, but is zero in π∗LK(n)S\pi_*L_{K(n)}\mathbb{S} and thus in π∗Ln−1LK(n)S\pi_*L_{n-1}L_{K(n)}\mathbb{S} and π∗LK(j)LK(n)S\pi_*L_{K(j)}L_{K(n)}\mathbb{S}. Thus the two bottom horizontal maps are not injective on homotopy groups and so do not admit retractions. □

Lemma 2.2. Suppose (p−1)∤n(p-1)\nmid n. If x∈π∗S(p)x\in\pi_*\mathbb{S}_{(p)} in the homotopy of the pp-local sphere S(p)\mathbb{S}_{(p)} is of Adams–Novikov filtration s>n2s>n^2, then ιK(n)(x)=0\iota_{K(n)}(x)=0 in π∗LK(n)S\pi_*L_{K(n)}\mathbb{S}.

Proof. When (p−1)∤n(p-1)\nmid n, the Morava stabilizer group Sn\mathbb{S}_n is a pp-adic analytic group of dimension n2n^2 with no pp-torsion, and so has cohomological dimension n2n^2 ([13]; see [2 Thm. 5.3.12]). As a consequence, the cohomology Hcs(Sn,πtEn)H_c^s(\mathbb{S}_n,\pi_tE_n) vanishes for s>n2s>n^2, and hence so does Hcs(Gn,πtEn)H_c^s(G_n,\pi_tE_n), since by [4 Lemma 1.32],

Hcs(Gn,πtEn)≅Hcs(Sn,πtEn)Gal.H_c^s(G_n,\pi_tE_n)\cong H_c^s(\mathbb{S}_n,\pi_tE_n)^{\mathrm{Gal}}.

By [8 Thm. 1], the homotopy fixed point spectral sequence

E2s,t=Hcs(Gn,πtEn)⟹πt−sLK(n)SE_2^{s,t}=H_c^s(G_n,\pi_tE_n)\Longrightarrow\pi_{t-s}L_{K(n)}\mathbb{S}

is the K(n)K(n)-local EnE_n-Adams spectral sequence for π∗LK(n)S\pi_*L_{K(n)}\mathbb{S}. So there is a map of spectral sequences

Ext⁡BP∗BPs,t(BP∗,BP∗)⟶Hcs(Gn,πtEn)\operatorname{Ext}_{BP_*BP}^{s,t}(BP_*,BP_*)\longrightarrow H_c^s(G_n,\pi_tE_n)
Ext⁡BP∗BPs,t(BP∗,BP∗)→Hcs(Gn,πtEn)↓↓πt−sS(p)→πt−sLK(n)S\begin{CD} \operatorname{Ext}_{BP_*BP}^{s,t}(BP_*,BP_*) @>>> H_c^s(G_n,\pi_tE_n)\\ @VVV @VVV\\ \pi_{t-s}\mathbb{S}_{(p)} @>>> \pi_{t-s}L_{K(n)}\mathbb{S} \end{CD}

where the bottom arrow is the map induced by localization. The right-hand spectral sequence has a horizontal vanishing line at s>n2s>n^2, which proves the claim. □

In the next result, we let

β1∈Ext⁡BP∗BP2,2(p2−p)(BP∗,BP∗)\beta_1\in\operatorname{Ext}_{BP_*BP}^{2,2(p^2-p)}(BP_*,BP_*)

be the first element of the β\beta-family. We also denote by the same name the element of

π2(p2−p)−2S(p)\pi_{2(p^2-p)-2}\mathbb{S}_{(p)}

it detects.

Corollary 2.3. If pp is odd and (p−1)∤n(p-1)\nmid n, then ιK(n)(β1k)=0\iota_{K(n)}(\beta_1^k)=0 whenever 2k>n22k>n^2.

Proof. Since β1\beta_1 has Adams–Novikov filtration 22, the class β1k\beta_1^k has Adams–Novikov filtration at least 2k>n22k > n^2. The claim then follows from Lemma 2.2. □

The case p≥5p \ge5

For an odd prime pp, let Ep−1E_{p-1} be Lubin–Tate theory associated to the pair (Fpp−1,Γp−1)(\mathbb{F}_{p^{p-1}}, \Gamma_{p-1}) where Γp−1\Gamma_{p-1} is the Honda formal group law of height p−1p-1. The Morava stabilizer group Sp−1S_{p-1} is the group of Fpp−1\mathbb{F}_{p^{p-1}}-automorphisms of Γp−1\Gamma_{p-1}. Let Φ⊂Sp−1\Phi\subset S_{p-1} be a maximal finite subgroup containing an element of order pp, so that Φ≅Z/p×Z/(p−1)2\Phi\cong\mathbb{Z}/p \times\mathbb{Z}/(p-1)^2 up to conjugation [15 §2.2]. Let

EOp−1:=Ep−1hΦ.EO_{p-1}:=E_{p-1}^{h\Phi}.

This is a K(p−1)K(p-1)-local ring spectrum. Non-triviality of an element in π∗LK(p−1)S\pi_*L_{K(p-1)}\mathbb{S} can thus be detected in EOp−1EO_{p-1} by its image under the unit LK(p−1)S→EOp−1L_{K(p-1)}\mathbb{S}\to EO_{p-1}.

Lemma 2.4. Let p≥5p \ge5 and 1≤k≤(p−1)21 \le k \le(p-1)^2. Then β1k\beta_1^k has nonzero image in π∗EOp−1\pi_*EO_{p-1}. Consequently ιK(p−1)(β1k)≠0\iota_{K(p-1)}(\beta_1^k) \ne0 in π∗LK(p−1)S\pi_*L_{K(p-1)}\mathbb{S}.

Proof. The unit S(p)→EOp−1\mathbb{S}_{(p)} \to EO_{p-1} induces a map from the Adams–Novikov spectral sequence computing π∗S(p)\pi_*\mathbb{S}_{(p)} to the homotopy fixed point spectral sequence

H∗(Φ,(Ep−1)∗)⟹π∗EOp−1.H^*(\Phi,(E_{p-1})_*) \Longrightarrow\pi_*EO_{p-1}.

Modulo the image of the transfer (which are permanent cycles) the spectral sequence is computed in [15]. It follows from that computation that β1k\beta_1^k has nonzero image in π∗EOp−1\pi_*EO_{p-1} for k≤(p−1)2k \le(p-1)^2. Since EOp−1EO_{p-1} is K(p−1)K(p-1)-local, the unit factors through LK(p−1)SL_{K(p-1)}\mathbb{S}, and so ιK(p−1)(β1k)≠0\iota_{K(p-1)}(\beta_1^k) \ne0 for kk in this range. □

Corollary 2.5. Let p≥5p \ge5 and p≤n<2(p−1)p \le n < \sqrt{2}(p-1), and let n2<2k≤2(p−1)2n^2 < 2k \le2(p-1)^2. Then ιK(n)(β1k)=0\iota_{K(n)}(\beta_1^k)=0, while ιK(p−1)(β1k)≠0\iota_{K(p-1)}(\beta_1^k)\ne0.

Proof. Since p−1<n<2(p−1)p-1<n<2(p-1), we have (p−1)∤n(p-1)\nmid n. The claim then follows from Lemma 2.2 and Lemma 2.4. □

Theorem 2.6. Let p≥5p \ge5 and p≤n<2(p−1)p \le n < \sqrt{2}(p-1). Then

Ln−1Sp→Ln−1LK(n)SL_{n-1}\mathbb{S}_p \to L_{n-1}L_{K(n)}\mathbb{S}

is not injective on homotopy groups, and so does not admit a retraction.

Proof. Apply Lemma 2.1 with j=p−1j=p-1 and x=β1kx=\beta_1^k as in Corollary 2.5. □

Remark 2.7. Note that we have shown the stronger result that, under the assumptions of Theorem 2.6,

LK(p−1)S→LK(p−1)LK(n)SL_{K(p-1)}\mathbb{S} \to L_{K(p-1)}L_{K(n)}\mathbb{S}

does not admit a retraction.

Remark 2.8. The condition p≥5p \ge5 is required for the range p≤n<2(p−1)p \le n < \sqrt{2}(p-1) to be nonempty. Further, for p≥5p \ge5, we can take n=pn=p.

The case (n,p)=(3,3)(n,p)=(3,3)

At p=3p=3 there is no nn with p≤n<2(p−1)=22p \le n < \sqrt{2}(p-1)=2\sqrt{2}. However, since 2∤32\nmid3 and 10>910>9, Corollary 2.3 gives ιK(3)(β15)=0\iota_{K(3)}(\beta_1^5)=0. So, by Lemma 2.1, it suffices to prove that ιK(2)(β15)≠0\iota_{K(2)}(\beta_1^5)\ne0. The following is a consequence of [10 Thm. 1.8].

Theorem 2.9 (Henn–Karamanov–Mahowald). Let p=3p=3 and let S/3\mathbb{S}/3 be the mod 33 Moore spectrum. Then β15\beta_1^5 has nonzero image in π50LK(2)S/3\pi_{50}L_{K(2)}\mathbb{S}/3 under the map

S(p)→LK(2)S→LK(2)S/3.\mathbb{S}_{(p)} \to L_{K(2)}\mathbb{S} \to L_{K(2)}\mathbb{S}/3.

In particular, $\iota_{K(2)}(\beta_1^5)\ne 0.

Theorem 2.10. Let (n,p)=(3,3)(n,p)=(3,3). The map

Ln−1Sp⟶Ln−1LK(n)SL_{n-1}S_p \longrightarrow L_{n-1}L_{K(n)}S

is not injective on homotopy groups, and so does not admit a retraction.

Proof. Apply Lemma 2.1 with j=2j=2 and x=β15x=\beta_1^5: Corollary 2.3 gives ιK(3)(β15)=0\iota_{K(3)}(\beta_1^5)=0 and Theorem 2.9 gives ιK(2)(β15)≠0\iota_{K(2)}(\beta_1^5)\ne0. □

References

References

  1. [1][Bar20] T. Barthel, A short introduction to the telescope and chromatic splitting conjectures, in: Bousfield classes and Ohkawa’s theorem, Springer Proc. Math. Stat. 309 (2020) 261–273, arXiv:1902.05046.arxiv.org/abs/1902.05046
  2. [2][BB20] T. Barthel, A. Beaudry, Chromatic structures in stable homotopy theory, in: Handbook of Homotopy Theory (H. Miller, ed.), CRC Press, 2020, arXiv:1901.09004. Numbering follows the published version.arxiv.org/abs/1901.09004
  3. [3][Bea17] A. Beaudry, The chromatic splitting conjecture at n = p = 2, Geom. Topol. 21 (2017) 3213–3230, arXiv:1502.02190.DOI
  4. [4][BG18] I. Bobkova, P. G. Goerss, Topological resolutions in K(2)-local homotopy theory at the prime 2, J. Topol. 11 (2018) 917–956, arXiv:1610.00158.DOI
  5. [5][BGH22] A. Beaudry, P. G. Goerss, H.-W. Henn, Chromatic splitting for the K(2)-local sphere at p = 2, Geom. Topol. 26 (2022) 377–476, arXiv:1712.08182.
  6. [6][Dev90] E. S. Devinatz, K-theory and the generating hypothesis, Amer. J. Math. 112 (1990) 787–804.DOI
  7. [7][Dev98] E. S. Devinatz, The generating hypothesis revisited, in: Stable and unstable homotopy (Toronto, 1996), Fields Inst. Commun. 19, Amer. Math. Soc., 1998, 73–92.DOI
  8. [8][DH04] E. S. Devinatz, M. J. Hopkins, Homotopy fixed point spectra for closed subgroups of the Morava stabilizer groups, Topology 43 (2004) 1–47.DOI
  9. [9][GHM14] P. G. Goerss, H.-W. Henn, M. Mahowald, The rational homotopy of the K(2)-local sphere and the chromatic splitting conjecture for the prime 3 and level 2, Doc. Math. 19 (2014) 1271–1290, arXiv:1210.7031.arxiv.org/abs/1210.7031
  10. [10][HKM13] H.-W. Henn, N. Karamonov, M. Mahowald, The homotopy of the K(2)-local Moore spectrum at the prime 3 revisited, Math. Z. 275 (2013) 953–1004, arXiv:0811.0235.
  11. [11][Ho95] M. Hovey, Bousfield localization functors and Hopkins’ chromatic splitting conjecture, in: The Čech centennial, Contemp. Math. 181 (1995) 225–250. Numbering follows the published version; in the author’s preprint the splitting statement is Cor. 4.4.DOI
  12. [12][Ho07] M. Hovey, On Freyd’s generating hypothesis, Q. J. Math. 58 (2007) 31–45.DOI
  13. [13][La65] M. Lazard, Groupes analytiques p-adiques, Publ. Math. IHÉS 26 (1965) 389–603.DOI
  14. [14][MX26] S. Ma, Z. Xu, Generating hypotheses along the motivic deformation, arXiv:2609.15253 (2026).arxiv.org/abs/2609.15253
  15. [15][Na10] L. S. Nave, The Smith–Toda complex V((p + 1)/2) does not exist, Ann. of Math. 171 (2010) 491–509.
  16. [16][Rav92] D. C. Ravenel, Nilpotence and periodicity in stable homotopy theory, Ann. of Math. Stud. 128, Princeton Univ. Press, 1992.
  17. [17][SY95] K. Shimomura, A. Yabe, The homotopy groups π*(L₂S⁰), Topology 34 (1995) 261–289.DOI

Paper details

vlnc:W_s2zzpkb6
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 2

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Claude

View statement
Looking for related papers…
Contents