betaFeedback

WorksMathematics

Permanence and 144144-periodicity of x153,3x_{153,3} in the stable stems at the prime three

Abstract

The class x153,3x_{153,3} lies in stable degree 153153 and filtration 33 on the algebraic starting page of the 33-primary Adams–Novikov spectral sequence of the sphere. We show that it survives and begins a family with period 144144, answering both parts of a question of Carrick and Davies. A comparison with periodic topological modular forms and a bound on the possible differential targets prove permanence. A divisibility calculation for the mod-33 Moore spectrum supplies a lift to a finite spectrum with a degree-144144 self-map. The resulting iterates have nonzero image in the height-two detector of Carrick and Davies.

Introduction

Write SS for the sphere spectrum and S(3)S_{(3)} for its localization at the prime 3, which inverts integers prime to 3. For a spectrum YY, ΣnY\Sigma^nY denotes its nn-fold suspension (desuspension for negative nn) and πnY=[ΣnS,Y]\pi_nY=[\Sigma^nS,Y] its nnth stable homotopy group. Here [A,B][A,B] denotes stable homotopy classes of maps from AA to BB, and π∗Y\pi_*Y denotes the graded collection of homotopy groups. An integer nn indexing πn\pi_n is called a stem. We work at the prime 3 throughout.

Chromatic homotopy theory organizes periodic families in the stable stems by height. At height one, the image of the stable JJ-homomorphism, which is induced by one-point compactification of orthogonal transformations, is closely related to the Adams self-map of the mod-3 Moore spectrum. For a positive integer mm, write Y/mY/m for the cofiber of multiplication by mm on a spectrum YY, which fits into the sequence Y→mY→Y/m→ΣYY \xrightarrow{m} Y \to Y/m \to\Sigma Y. For m=3m=3 the last two maps are denoted by q0q_0 and ∂0\partial_0, respectively. The mod-3 Moore spectrum is M=S/3M=S/3. The Adams self-map v1:Σ4M→Mv_1:\Sigma^4M\to M induces multiplication by the degree-4 generator of K(1)K(1)-homology, where K(i)K(i) denotes the iith Morava KK-theory [8]. For an integer j≥1j\geq1, write Mj=S/(3,v1j)M_j=S/(3,v_1^j) for the generalized Moore spectrum given by the cofiber of v1j:Σ4jM→Mv_1^j:\Sigma^{4j}M\to M. A power of a self-map here uses composition with the appropriate suspensions. The Smith–Toda complex V(1)V(1) is M1M_1.

At height two, let v2v_2 denote the degree-16 periodicity generator of the coefficients of K(2)K(2). A v29v_2^9 self-map has degree 144 and induces multiplication by v29v_2^9 on K(2)K(2)-homology. Behrens and Pemmaraju constructed such a map on V(1)V(1) [2]. Belmont printed: 2020 Mathematics Subject Classification. Primary 55Q45. Secondary 55Q51, 55T15, 55P42. printed: Key words and phrases. Adams–Novikov spectral sequence, stable stems, chromatic homotopy. and Shimomura constructed such maps on MjM_j for 2≤j≤82 \le j \le8 and used them to obtain families in the stable stems [4 Corollary 4.7 and §5].

The Adams–Novikov spectral sequence uses Brown–Peterson homology, denoted by BP∗Y\mathrm{BP}_*Y for a spectrum YY. Its initial page is the comodule cohomology group

E2f,t(Y)=Ext⁡BP∗BPf,t(BP∗,BP∗Y),E_2^{f,t}(Y)=\operatorname{Ext}_{\mathrm{BP}_*\mathrm{BP}}^{f,t}(\mathrm{BP}_*,\mathrm{BP}_*Y),

where BP∗=BP∗S\mathrm{BP}_*=\mathrm{BP}_*S is the coefficient ring and BP∗BP\mathrm{BP}_*\mathrm{BP} is the ring of cooperations [8]. The cohomological degree ff is the Adams–Novikov filtration and tt is the internal degree. We write Er(n,f;Y)=Erf,n+f(Y)E_r(n,f;Y)=E_r^{f,n+f}(Y) for the group on page ErE_r in stem n=t−fn=t-f and filtration ff. An omitted spectrum means the sphere. The differential drd_r on page ErE_r sends (n,f)(n,f) to (n−1,f+r)(n-1,f+r). A drd_r-cycle is a class killed by drd_r, and a drd_r-boundary is a class in its image. A permanent cycle survives nontrivially to the limiting page E∞E_\infty.

Let FfπnYF^f\pi_nY be the subgroup of Adams–Novikov filtration at least ff. The limiting page gives the associated graded group gr⁡ANfπnY=FfπnY/Ff+1πnY\operatorname{gr}_{\mathrm{AN}}^f\pi_nY=F^f\pi_nY/F^{f+1}\pi_nY of the 3-completed homotopy groups, where AN stands for Adams–Novikov. For the sphere and the finite Moore spectra in the positive stems used below these groups agree with the 3-local groups. A class in πnY\pi_nY that maps to a given class of E∞E_\infty is called a representative of that class. A hidden extension by three occurs when a class is killed by 3 on E∞E_\infty but multiplying a representative by 3 gives a nonzero class in higher filtration.

The notation viv_i also denotes the usual chromatic generators of BP∗\mathrm{BP}_*, of internal degree 2(3i−1)2(3^i-1) for i≥1i\ge1. The Greek-letter class α1\alpha_1 is the algebraic connecting class of v1v_1 modulo 3, in bidegree (3,1)(3,1). The divided class βi/j\beta_{i/j} is defined, where the construction applies, by taking the connecting classes of v2iv_2^i modulo (3,v1j)(3,v_1^j) for multiplication by v1jv_1^j and then by 3. Its bidegree is (16i−4j−2,2)(16i-4j-2,2), and we abbreviate βi/1\beta_{i/1} to βi\beta_i [4 §5]. Juxtaposition and superscripts on classes denote multiplication and its iterated powers. We use the same symbols for homotopy representatives when the classes are permanent and we are discussing products in homotopy.

Carrick and Davies use the connective height-two image-of-JJ spectrum j2j^2 to detect classes in the stable stems [6]. In high degrees its homotopy groups are calculated from the fiber of ψ2−1\psi^2-1 on tmf\mathrm{tmf}, the connective spectrum of topological modular forms. Here ψ2\psi^2 is the stable Adams operation indexed by 2 and 1 is the identity map. The unit map S→j2S\to j^2 gives the Hurewicz homomorphism on stable homotopy groups, whose image is called the Hurewicz image. A class is detected by j2j^2 when its image under this homomorphism is nonzero. Let TMF\mathrm{TMF} denote periodic topological modular forms and J2J^2 the fiber of ψ2−1\psi^2-1 on TMF\mathrm{TMF}, or equivalently the homotopy equalizer of ψ2\psi^2 and the identity [7 §3]. We write ∂\partial for the connecting map πn+1TMF→πnJ2\pi_{n+1}\mathrm{TMF}\to\pi_nJ^2 from this fiber sequence. We use the same symbol for its connective counterpart πn+1tmf→πnj2\pi_{n+1}\mathrm{tmf}\to\pi_nj^2 in the high degrees where j2j^2 agrees with the fiber.

In topological modular forms, α\alpha and β\beta denote the images of α1\alpha_1 and β1\beta_1, or their spectral sequence classes when working on a page. The modular discriminant Δ\Delta has weight 12 and bidegree (24,0)(24, 0). In homotopy, powers of Δ\Delta with exponent divisible by three denote powers of the degree-72 class detected by Δ3\Delta^3 [1 §§5–6]. We use J2J^2 to compare differentials and j2j^2 to detect the periodic family. The modified Adams–Novikov spectral sequence of J2J^2 uses the same stem and filtration convention as the ordinary one.

Carrick and Davies [6 Question 3.15] ask whether the class x153,3x_{153,3} in stem 153 and Adams–Novikov filtration 3 survives in the 3-primary sphere, and whether it is v29v_2^9-periodic. They identify the class on the E2E_2-page by the relation α1x153,3≐β5β6/3\alpha_1x_{153,3} \doteq\beta_5\beta_6/3, where ≐\doteq denotes equality up to multiplication by a unit at the prime 3. They report that its counterpart in the K(2)K(2)-local sphere, obtained by localizing with respect to K(2)K(2)-homology, is permanent. Naturality then constrains a possible differential on x153,3x_{153,3} to have target with zero image on the corresponding local page. Permanence in the sphere requires controlling those possible targets there.

For the periodicity question, let ∂j:Mj→Σ4j+1M\partial_j:M_j\to\Sigma^{4j+1}M be the connecting map of the cofiber sequence defining MjM_j for j≥1j\geq1. Together with the mod-3 boundary ∂0:M→ΣS\partial_0:M\to\Sigma S, it gives

π155+4jMj→∂jπ154M→∂0π153S.\pi_{155+4j}M_j \xrightarrow{\partial_j} \pi_{154}M \xrightarrow{\partial_0} \pi_{153}S.

To construct a periodic family from a representative XX of x153,3x_{153,3}, we need a lift z∈π155+4jMjz\in\pi_{155+4j}M_j through these two boundaries. Such a lift requires 3X=03X=0 and a choice of y∈π154My\in\pi_{154}M with ∂0y=X\partial_0y=X and v12y=0v_1^2y=0. Given zz and a v29v_2^9 self-map V:Σ144Mj→MjV:\Sigma^{144}M_j\to M_j, we write VkzV^kz for the class obtained by applying the suspended kk-fold iterate of VV to zz for each integer k≥0k\geq0. The family requires the double boundaries ∂0∂j(Vkz)\partial_0\partial_j(V^kz) to be nonzero.

Write Z/m\mathbb{Z}/m for the cyclic group of order mm.

Theorem 1.1. The class x153,3x_{153,3} is a permanent cycle in the Adams–Novikov spectral sequence for S(3)S_{(3)}. In particular, gr⁡AN3π153S(3)≅Z/3\operatorname{gr}^{3}_{\mathrm{AN}}\pi_{153}S_{(3)}\cong\mathbb{Z}/3, generated by the class detected by x153,3x_{153,3}. Every representative of x153,3x_{153,3} has order three and begins a v29v_2^9-periodic family of order-three, filtration-three classes. The class in stem 153+144k153+144k is detected by ∂(βΔ6(k+1))\partial(\beta\Delta^{6(k+1)}) in π∗j2\pi_*j^2.

Sections 2 and 3 compute the possible differential targets and rule them out using comparison with J2J^2, multiplication in the sphere, and a finite-page vanishing line, which bounds the bidegrees that can still be nonzero on a fixed page. Section 4 resolves the hidden extension by three. Section 5 uses β1\beta_1-divisibility in the Moore spectrum to choose a lift with v13y=0v_1^3y=0. Section 6 then detects every iterate in j2j^2. These stems are the exceptional degrees in Carrick and Davies’s description of the Hurewicz image [6 Theorem A].

The relevant E2E_2-groups and products

We use revision f530898 of the data of Belmont and Wang [3], generated with Belmont’s prime-3 adaptation of Wang’s MinimalResolution program [11]. Wang describes the underlying minimal-resolution algorithm in [10]. The data extend through internal degree 185, which contains every algebraic group used in the permanence argument. The program computes the algebraic Novikov spectral sequence and its multiplicative information, which determine the indicated groups and products on the Adams–Novikov E2E_2-page.

Write F3\mathbb{F}_3 for the field of three elements. The notation [f−i][f-i] in Belmont’s generator names labels the basis element with index ii in homological degree ff of the chosen resolution. The hyphen separates the two indices. A prefix such as v26v_2^6 in v26[8−2]v_2^6[8-2] records the coefficient monomial of the leading term used to name the generator. In these names v0v_0 denotes 3 and viv_i for i≥1i \ge1 denotes the Brown–Peterson generator of that index.

Lemma 2.1. The groups in the following table have the indicated F3\mathbb{F}_3-bases, with ∅\varnothing denoting the empty basis of the zero group. Here bb is the generator v31[12−5]v_3^1[12-5] in (152,12)(152,12) and ww is the generator [4−8][4-8] in (92,4)(92,4).

(n,f)(n,f)Basis of E2(n,f)E_2(n,f)
(153,3)(153,3)x153,3x_{153,3}
(152,8)(152,8)β12β4β5\beta_1^2\beta_4\beta_5
(152,12)(152,12)bb
(152,16)(152,16)β16w, β17β6/3\beta_1^6w,\ \beta_1^7\beta_{6/3}
(152,20)(152,20)∅\varnothing
(91,9)(91,9)β13α1β4\beta_1^3\alpha_1\beta_4
(91,13)(91,13)∅\varnothing
(90,18)(90,18)β19\beta_1^9

Table 2.

The product α1b\alpha_1b in E2(155,13)E_2(155,13) is nonzero and the map given by multiplication by α1\alpha_1,

α1⋅−:E2(153,7)⟶E2(156,8)\alpha_1\cdot-: E_2(153,7)\longrightarrow E_2(156,8)

is surjective.

Proof. The following are the generator names in Belmont’s data [3], in the order of the named elements in the table. Coefficients of the single-generator products are suppressed.

x153,3x_{153,3}v01v11v21v32[3−2]v_0^1v_1^1v_2^1v_3^2[3-2]β12β4β5\beta_1^2\beta_4\beta_5v26[8−2]v_2^6[8-2]
β16w\beta_1^6w[16−9][16-9]β17β6/3\beta_1^7\beta_{6/3}v23[16−3]v_2^3[16-3]
β13α1β4\beta_1^3\alpha_1\beta_4v23[9−0]v_2^3[9-0]β19\beta_1^9[18−0][18-0]

Table 3.

To reproduce the lookups, initialize ANSSData() from beta.py at the root of Belmont’s repository [3]. The method in_deg(n,f) lists the generators in stem nn and filtration ff, and three multiplies a class by 3. The β\beta-products use the multiplication tables for the mod-3 Moore spectrum and the algebraic boundary, as implemented in beta.py. To compute a product of named classes, start with the unit [0−0][0-0] and compose the corresponding multiplication methods. For the last two claims the alpha1 table gives

α1b=v21[13-9]+2v23[13-5],\alpha_1 b = v_2^1[13\text{-}9] + 2v_2^3[13\text{-}5],
α1[7-11]=[8-10],\alpha_1[7\text{-}11] = [8\text{-}10],
α1v31[7-8]=v31[8-8].\alpha_1v_3^1[7\text{-}8] = v_3^1[8\text{-}8].

The last two images form the basis in (156,8)(156, 8). □

Permanence

The BP Hopf algebroid (BP∗,BP∗BP)(\mathrm{BP}_*, \mathrm{BP}_*\mathrm{BP}) is concentrated in internal degrees divisible by four. Since drd_r changes the internal degree by r−1r-1, a nonzero differential can occur only for r≡1(mod4)r \equiv1 \pmod{4}. Thus none can hit x153,3x_{153,3}, whose filtration is three. Outgoing differentials have target in stem 152 and filtration 3+r3+r. By [5 Proposition 11.6 and Theorem 12.2], the sphere E10E_{10}-page vanishes at f>14+(n+9)/23f > 14 + (n + 9)/23. For n=152n = 152 this excludes r≥21r \ge21, and Lemma 2.1 excludes r=17r = 17. Figure 1 shows the remaining possibilities r=5,9,13r = 5, 9, 13.

Possible outgoing differentials from $x_{153,3}$ in the sphere Adams–Novikov spectral sequence

Figure 1. Possible outgoing differentials from x153,3x_{153,3} in the sphere Adams–Novikov spectral sequence. The E2E_2 entries come from Belmont’s data [3]. Each filled dot denotes one generator, and the differential labels are beside their targets. Dashed arrows show the three candidates checked below. Crosses mark the zero groups that exclude d17d_{17} and d21d_{21} on the indicated pages. The shaded band starts above filtration 21, so the E10E_{10} vanishing line also excludes every drd_r with r>21r > 21.

The d5 comparison

Davies’s filtered model of J2J^2 [7 §3] gives the comparison from the sphere and the maps of spectral sequences induced by the equalizer boundary ∂\partial and by the projection p ⁣:J2→TMFp\colon J^2 \to\mathrm{TMF}. On pages written in stem and filtration coordinates, ∂\partial has degree (−1,1)(-1, 1), meaning it sends (n,f)(n, f) to (n−1,f+1)(n-1, f+1). The modified spectral sequence is also supported in internal degrees divisible by four.

In bidegree (153,3)(153,3) the TMF term is zero. The boundary from bidegree (154,2)(154,2) of TMF therefore spans the J2J^2 group in (153,3)(153,3), which is generated by ∂(βΔ6)\partial(\beta\Delta^6). Indeed, βΔ6\beta\Delta^6 is fixed by ψ2\psi^2 on E2E_2, and is the generator of that bidegree. The class ∂(βΔ6)\partial(\beta\Delta^6) is a d5d_5-cycle. On TMF, d5(Δ)=αβ2d_5(\Delta) = \alpha\beta^2 [1 Equation (6.1)] and d5(β)=0d_5(\beta)=0, so d5(βΔ6)=0d_5(\beta\Delta^6)=0 by the Leibniz rule and 3α=03\alpha=0. Hence the image of x153,3x_{153,3} is a d5d_5-cycle.

By [7 Propositions 3.10 and 3.11], the classes β1\beta_1 and β4\beta_4 project to β\beta and βΔ2\beta\Delta^2, respectively, and β5\beta_5 maps to ∂(αΔ3)\partial(\alpha\Delta^3), up to units. The boundary is linear over the equalizer via the projection on its TMF term. Consequently the image of the generator of the possible d5d_5 target group is

β12β4β5⟼∂(αβ3Δ5)\beta_1^2\beta_4\beta_5 \longmapsto\partial(\alpha\beta^3\Delta^5)

up to a unit.

Here αβ3Δ5\alpha\beta^3\Delta^5 generates the TMF term in (153,7)(153,7). It has order three, and ψ2\psi^2 fixes it, since ψ2\psi^2 fixes α\alpha and β\beta and acts on Δ\Delta by 2122^{12} [7 Remark 3.4]. The equalizer exact sequence therefore shows that its boundary is nonzero. The comparison is injective on E5(152,8)E_5(152,8), so naturality forces d5(x153,3)=0d_5(x_{153,3})=0.

The d9 obstruction

Suppose d9(x153,3)≠0d_9(x_{153,3})\ne0. Its target is represented by a unit multiple of bb. The group E2(152,12)E_2(152,12) is one-dimensional by [3], so bb can survive to E9E_9 only if every d5d_5 from (153,7)(153,7) is zero. The surjectivity in Lemma 2.1 and the permanence of α1\alpha_1 then imply that every d5d_5 from (156,8)(156,8) is zero. This is the entire source group for a d5d_5 hitting α1b\alpha_1b. Since bb is a d5d_5-cycle, this proves α1b≠0\alpha_1b\ne0 on E9E_9.

On E2E_2, α1x153,3≐β5β6/3\alpha_1x_{153,3} \doteq\beta_5\beta_6/3 by [6 Remark 3.14]. The right-hand side is a nonzero product of permanent cycles, detected by j2j^2 [6 Proposition 3.13]. By the Leibniz rule our hypothetical d9d_9 on x153,3x_{153,3} would give a nonzero d9d_9 on this permanent product, a contradiction.

The d13 target group

Ravenel’s differential is d9(α1β4)≐β16d_9(\alpha_1\beta_4)\doteq\beta_1^6 [5 Lemma 12.10]. We check that ww is a d9d_9-cycle. The class β19\beta_1^9 is a nonzero d5d_5-cycle, and its possible d5d_5 source E5(91,13)E_5(91,13) is zero by the E2E_2 data [3]. Hence it is nonzero on E9E_9. The Leibniz rule then gives

d9(β13α1β4)≐β19.d_9(\beta_1^3\alpha_1\beta_4)\doteq\beta_1^9.

In particular, the generator of E5(91,9)E_5(91,9) survives to E9E_9, so it cannot be a d5d_5-boundary. This is the only possible d5d_5 target for ww. The d9d_9 target for ww is itself zero by Lemma 2.1, and filtration rules out incoming differentials to ww through E9E_9.

Multiplying d9(α1β4)d_9(\alpha_1\beta_4) by ww and by the permanent cycle β1β6/3\beta_1\beta_6/3 kills both generators of E2(152,16)E_2(152,16) that remain on E9E_9. It follows that E10(152,16)=0E_{10}(152,16)=0. This excludes d13(x153,3)d_{13}(x_{153,3}) and proves the permanence assertion of Theorem 1.1.

The hidden extension by three

Let X∈π153S(3)X\in\pi_{153}S_{(3)} represent x153,3x_{153,3}. In this section we abbreviate Ffπ153S(3)F^f\pi_{153}S_{(3)} to FfF^f. Since 3x153,3=03x_{153,3}=0 on E2E_2, we have 3X∈F73X\in F^7. Also α1(3X)=0\alpha_1(3X)=0 because 3α1=03\alpha_1=0. Sparseness, Belmont’s data [3], and the E10E_{10} vanishing line leave possible filtration groups in F7F^7 only at 7, 11, and 19.

In filtration 7, the α1\alpha_1-kernel on E5E_5 is spanned by z7=v26[7-3]z_7 = v_2^6[7\text{-}3], by the multiplication calculation in § 2. The product tables [3] identify it on E5E_5 with β1β6/3(α1β4)\beta_1\beta_{6/3}(\alpha_1\beta_4) up to a unit. Its d9d_9 is nonzero, since β17β6/3\beta_1^7\beta_{6/3} maps to β8Δ3\beta^8\Delta^3 in the Adams–Novikov spectral sequence of TMF. The latter is a nonzero d5d_5-cycle and E5(153,11;TMF)=0\mathcal{E}_5(153,11;\mathrm{TMF}) = 0, so it remains nonzero on E9E_9. Here we use that β6/3\beta_{6/3} maps to βΔ3\beta\Delta^3 [7 Proposition 3.10]. In stem 156 and filtration 8, the only possible incoming differential is d5d_5 from (157,3)(157,3), which is spanned by α1β10\alpha_1\beta_{10} by the E2E_2 data [3]. This is a permanent cycle because β10\beta_{10} is permanent [2 Corollary 1.2]. Thus multiplication by α1\alpha_1 has zero kernel on the surviving filtration-7 associated graded of stem 153. It follows that 3X3X lies in F11F^{11}.

For filtration 11, Belmont’s data [3] identify the generator as β4u\beta_4u up to a unit, with u=[9-6]∈E5(95,9)u = [9\text{-}6] \in\mathcal{E}_5(95,9). The classical differential d5(β4)=α1β12β3/3d_5(\beta_4) = \alpha_1\beta_1^2\beta_{3/3} [4 Table 6] gives a nonzero component of d5(β4u)d_5(\beta_4u) along β16w\beta_1^6w. Indeed, the products in [3] show that α1β12β3/3u\alpha_1\beta_1^2\beta_{3/3}u is a unit multiple of [16-9][16\text{-}9]. The only possible d5d_5 target for uu is [14-3][14\text{-}3] in (94,14)(94,14), whose product with β4\beta_4 is a unit multiple of the other generator v23[16-3]v_2^3[16\text{-}3] in (152,16)(152,16). The two components are independent, so the filtration-11 generator supports d5d_5.

In filtration 19, the data [3] give the sole generator [19-7]=β16u′[19\text{-}7] = \beta_1^6u', where u′=[7-7]u' = [7\text{-}7] lies in (93,7)(93,7). If d5u′d_5u' is nonzero, it is a unit multiple of [12-4][12\text{-}4] in (92,12)(92,12), and its product with β16\beta_1^6 is the nonzero class [24-4][24\text{-}4]. Hence β16u′\beta_1^6u' supports d5d_5 in this case. Otherwise β16u′\beta_1^6u' is zero on E10E_{10} because u′u' has no d9d_9 target in (92,16)(92,16) and β16=0\beta_1^6 = 0 on E10E_{10}. This exhausts the possible leading terms of 3X3X, which proves 3X=03X = 0.

The generalized Moore lift

For M=S/3M = S/3, the mod-3 map q0:S→Mq_0 : S \to M includes the bottom cell. The sphere element β1\beta_1 of degree 10 acts on MM by a map Σ10M→M\Sigma^{10}M \to M. In the composite β1v12\beta_1v_1^2 we first apply the suspension of v12v_1^2 by 10 and then this action. Similarly, β1π148M\beta_1\pi_{148}M denotes the image of multiplication by β1\beta_1 on π148M\pi_{148}M.

Lemma 5.1. The self-map β1v12:Σ18M→M\beta_1v_1^2 : \Sigma^{18}M \to M is nullhomotopic, and

F6π158M⊆β1π148M.F^6\pi_{158}M \subseteq\beta_1\pi_{148}M.

Proof. Write LK(1)YL_{K(1)}Y for localization of a spectrum YY with respect to K(1)K(1)-homology. The low stem groups [8 Table A3.4], together with the image of JJ, give π18M=0\pi_{18}M = 0 and π19M≅F3\pi_{19}M \cong\mathbb{F}_3 from the image of JJ, mapping isomorphically to π19LK(1)M\pi_{19}L_{K(1)}M. Also π18LK(1)M=0\pi_{18}L_{K(1)}M = 0. The local groups and the comparison on the image of JJ follow from Ravenel’s height-one calculation [9 Theorem 8.10 and Remark 8.12] and the mod-3 cofiber sequence. Applying the stable mapping functors into MM and LK(1)ML_{K(1)}M to the suspended Moore cofiber sequence therefore shows that [Σ18M,M]→[Σ18M,LK(1)M][\Sigma^{18}M,M] \to[\Sigma^{18}M,L_{K(1)}M] is injective. The map β1v12\beta_1v_1^2 localizes to zero because π10LK(1)S=0\pi_{10}L_{K(1)}S = 0.

For an abelian group GG, write G/3G/3 for its quotient by 3G3G and G[3]G[3] for its subgroup killed by 3. The sequence of BP∗BPBP_*BP-comodules for MM gives the short exact sequence

0⟶E2(n,f)/3→q0E2(n,f;M)→∂0E2(n−1,f+1)[3]⟶0,(1)0 \longrightarrow E_2(n,f)/3 \xrightarrow{q_0} E_2(n,f;M) \xrightarrow{\partial_0} E_2(n-1,f+1)[3] \longrightarrow0, \tag*{(1)}

The maps in this sequence are induced by the bottom-cell inclusion and the mod-3 boundary. They commute with β1\beta_1 multiplication with the indicated bidegree shift of ∂0\partial_0. In Table 1, the bottom group at (n,f)(n,f) is the left term of (1) and the top group is the right term, coming from the top cell of MM. Each generator is named in Belmont’s notation, and 0 denotes the zero group. We obtain the two end groups from in_deg and three, and use beta1(g,k) for multiplication of a generator gg by β1k\beta_1^k in Belmont’s data [3]. Reading across each pair of columns gives the product up to a unit. In particular the first row is a surjection on the Moore E2E_2-page, and rows two through six are isomorphisms.

Action(n,f)(n,f)Source generators BottomSource generators TopTarget generators BottomTarget generators Top
β1\beta_1(148,4)(148,4)v26[4−4]v_2^6[4-4]v23v3[5−3]v_2^3v_3[5-3]00v23v3[7−3]v_2^3v_3[7-3]
β1\beta_1(148,8)(148,8)v3[8−7]v_3[8-7]00v3[10−7]v_3[10-7]00
β1\beta_1(147,9)(147,9)00v26[10−0]v_2^6[10-0]00v26[12−0]v_2^6[12-0]
β1\beta_1(147,13)(147,13)v3[13−4]v_3[13-4]v3[14−3]v_3[14-3]v3[15−4]v_3[15-4]v3[16−3]v_3[16-3]
β1\beta_1(147,17)(147,17)00[18−7][18-7]00[20−7][20-7]
β15\beta_1^5(108,4)(108,4)v23[4−5]v_2^3[4-5]v3[5−4]v_3[5-4]v23[14−5]v_2^3[14-5]v3[15−4]v_3[15-4]

Table 1. The bottom and top groups and their β1\beta_1 products, computed from Belmont’s data [3] using (1). The target of a row with β1k\beta_1^k has bidegree (n+10k,f+2k)(n+10k,f+2k).

We first check that the source of the last row has permanent representatives. The bottom-cell generator comes from the permanent product β2β6/3\beta_2\beta_6/3 in (108,4)(108,4). For the top-cell generator, let cc be the sphere class v3[5−4]v_3[5-4] in (107,5)(107,5). Ravenel’s table [8 Table A3.4] gives two independent classes modulo the image of JJ in stem 107, namely β2γ2\beta_2\gamma_2 and β2x81\beta_2x_{81}. Here γ2\gamma_2 is the height-three Greek-letter element in stem 81 and x81x_{81} is the other generator in that stem in Ravenel’s table. The two products in stem 107 are killed by 3. Together with the image of JJ they give the direct sum decomposition π107S(3)≅Z/81⊕(Z/3)2\pi_{107}S_{(3)} \cong\mathbb{Z}/81 \oplus(\mathbb{Z}/3)^2. The cyclic order-81 filtration-one group survives by [8 Theorem 5.3.7(a)]. The E2E_2 data [3] leave three other possible filtrations below the E10E_{10} vanishing line, namely 5, 13, and 17, each of dimension one. In filtration 17 the generator is β16[5−3]\beta_1^6[5-3] up to a unit, for [5−3][5-3] in (47,5)(47,5). Its potential d5d_5 and d9d_9 targets (46,10)(46,10) and (46,14)(46,14) are zero in the same data. It follows that β16[5−3]\beta_1^6[5-3] vanishes on E10E_{10}. Comparing with the known group order shows that cc survives. The kernel of the cyclic order-81 filtration-one quotient has exponent three, since the quotient is an isomorphism on the cyclic order-81 summand. Thus cc has an order-three representative and lifts to π108M\pi_{108}M. This lift has filtration four. Indeed the only possible lower filtration is zero, and

SourceTargetPage check
(148,4)(148, 4)⟶\longrightarrow (158,6)(158, 6)surjective on E2E_2
(148,8)(148, 8)⟶\longrightarrow (158,10)(158, 10)isomorphism on E2E_2
(147,9)(147, 9)⟶\longrightarrow (157,11)(157, 11)injective on E5E_5
(147,13)(147, 13)⟶\longrightarrow (157,15)(157, 15)injective on E9E_9
(147,17)(147, 17)⟶\longrightarrow (157,19)(157, 19)injective on E13E_{13}
The d5d_5 from (158,14)(158, 14) is zero.

∂0\partial_0 on E2(108,0;M)\mathcal{E}_2(108, 0; M) is injective onto the order-three subgroup on the filtration-one line, where no differentials occur. Filtration greater than four would place its boundary above filtration five. By the geometric boundary theorem [8 Theorem 2.3.4], the connecting map in (1) therefore sends the filtration-four detector of the lift to cc. Consequently the d5d_5 from (158,14)(158, 14) is zero, and any surviving class there is represented by a homotopy product with β1\beta_1.

We next check the first two rows on later pages, as shown in Figure 2. For the row with source (147,13)(147, 13) the map is injective on E9E_9. All possible d5d_5-boundaries in the target (157,15)(157, 15) come from (158,10)(158, 10), which on E5E_5 is in the image of β1\beta_1 by the second row. Thus if a d5d_5-cycle maps to a d5d_5-boundary, the Leibniz rule and injectivity on E5E_5 show it was already a boundary.

Moore spectrum bidegrees used in the $\beta_1$-divisibility argument

Figure 2. Moore spectrum bidegrees used in the β1\beta_1-divisibility argument. The E2E_2 groups and products are computed from [3] using (1). In (a), each square marks a nonzero E2E_2-group in one of the two candidate differential patterns. A dashed arrow from filtration ff to filtration gg represents a possible dg−fd_{g-f}. In (b), each arrow adds 10 to the stem and 2 to the filtration. The page checks identify the injectivity needed to compare differentials on the two sides.

The row from (147,17)(147, 17) to (157,19)(157, 19) is injective on E9E_9, since the only possible d5d_5-sources at the target are in (158,14)(158, 14), where we have shown d5=0d_5 = 0. It remains injective on E13E_{13}. A possible d9d_9-boundary in (157,19)(157, 19) comes from a class in (158,10)(158, 10). By the second row of the table we can choose a preimage BB in (148,8)(148, 8) on E5E_5 such that β1B\beta_1 B represents this class. Since d5(β1B)=0d_5(\beta_1 B) = 0, injectivity on E5E_5 of the row from (147,13)(147, 13) gives d5B=0d_5 B = 0. Hence the d9d_9-boundary in question is β1d9B\beta_1 d_9 B, and injectivity on E9E_9 shows that any cycle mapping to this boundary was already a d9d_9-boundary.

A surviving class in (158,10)(158, 10) now has a surviving antecedent in (148,8)(148, 8). The just-proved injections force the antecedent’s d5d_5 and d9d_9 to vanish, and an incoming differential would kill its product with β1\beta_1 as well. For a survivor in (158,6)(158, 6), an antecedent in (148,4)(148,4) is a d5d_5, d9d_9, and d13d_{13} cycle by the corresponding injections from (147,9)(147,9), (147,13)(147,13), and (147,17)(147,17). The strong E10E_{10} vanishing line [5 Theorem 12.2] applies to MM and excludes all remaining outgoing differentials from the antecedents, since their next target filtration is at least 21 in stem 147. In stem 158, filtration 18 is zero on E2E_2 by [3], and filtrations at least 22 are above the E10E_{10} line. Thus in every possible filtration at least 6, the surviving associated graded is β1\beta_1-divisible by an actual homotopy class. Successively subtracting such products in the finite filtration gives the claimed inclusion.

By § 4, we may choose y∈π154My \in\pi_{154}M with ∂0y=X\partial_0 y = X. The only possible filtration below 6 of v1y∈π158Mv_1y \in\pi_{158}M is 2. By (1) and Belmont’s data [3], the group E2(158,2;M)E_2(158,2;M) has zero bottom and top generated by α1β10\alpha_1\beta_{10} in (157,3)(157,3). The identity ∂0v1q0(β10)=α1β10\partial_0v_1q_0(\beta_{10}) = \alpha_1\beta_{10} [6 Lemma 2.6] shows that v1q0(β10)v_1q_0(\beta_{10}) represents its generator. Subtracting a multiple of q0(β10)q_0(\beta_{10}) from yy therefore gives a lift still mapping to XX and satisfying v1y∈F6π158Mv_1y \in F^6\pi_{158}M. Lemma 5.1 now implies v13y=0v_1^3y = 0. Exactness gives a lift z∈π167M3z \in\pi_{167}M_3 with ∂3z=y\partial_3z = y.

Detection of the iterates

Fix the degree-144 self-map V:Σ144M3→M3V:\Sigma^{144}M_3 \to M_3 from [4 Corollary 4.7] inducing v29v_2^9 on K(2)K(2)-homology. For the lift z∈π167M3z \in\pi_{167}M_3 just constructed, set

Xk=∂0∂3(Vkz)∈π153+144kS(3),k≥0.X_k = \partial_0\partial_3(V^kz) \in\pi_{153+144k}S_{(3)}, \qquad k \ge0.

Here V0V^0 is the identity, so X0=XX_0 = X. Put R=j2R = j^2 and T=tmf⁡T = \operatorname{tmf}. For Y=RY = R or TT, the notation Y/(3,v1j)Y/(3,v_1^j) means Y∧MjY \wedge M_j, where ∧\wedge is the smash product of spectra. Let hh denote the unit S→R\mathbb{S} \to R and its induced maps on homotopy groups, including those on cofibers obtained by smashing with MM or MjM_j.

In the degrees of interest RR agrees with the fiber of ψ2−1\psi^2-1 on TT. We have h(X)≐∂(βΔ6)h(X) \doteq\partial(\beta\Delta^6) by [6 Proposition 3.13 and Remark 3.14] and the permanence calculation in § 3. In detail the product α1h(X)\alpha_1h(X) is detected there. Write F3{g}\mathbb{F}_3\{g\} for the one-dimensional F3\mathbb{F}_3-span of a generator gg. The equalizer sequence and π153T=0\pi_{153}T = 0 give π153R=F3{∂(βΔ6)}\pi_{153}R = \mathbb{F}_3\{\partial(\beta\Delta^6)\}. We also have π154R≅F3\pi_{154}R \cong\mathbb{F}_3 by π155T=0\pi_{155}T = 0 and π154T=F3{βΔ6}\pi_{154}T = \mathbb{F}_3\{\beta\Delta^6\}. These groups follow from the prime-three tmf calculation [1 §§5–6].

Write q27:R→R/27q_{27}:R \to R/27 and ∂27:R/27→ΣR\partial_{27}:R/27 \to\Sigma R for the quotient and connecting maps in the mod-27 cofiber sequence. The periodicity element a=∂Δ6a = \partial\Delta^6 in π144(R/27)\pi_{144}(R/27) is the unique class satisfying ∂27(a)=∂(Δ6)\partial_{27}(a) = \partial(\Delta^6) [6 Proposition 3.17]. Multiplication by aa uses the module action of R/27R/27 on the relevant mod-3 spectra. Lemma 3.22 of [6] identifies this action on R/(3,v13)R/(3,v_1^3) with the effect of VV, up to sign. Let bR=h(y)∈π154(R/3)b_R = h(y) \in\pi_{154}(R/3) and write akbRa^k b_R for the result of applying this action kk times. Linearity of the v13v_1^3 boundary gives h(Xk)=∂0(akbR)h(X_k) = \partial_0(a^k b_R).

Choose a lift β~∈π11(T/3)\widetilde{\beta} \in\pi_{11}(T/3) with ∂0(β~)=β\partial_0(\widetilde{\beta}) = \beta. Bauer’s calculation [1 Figure 1] gives π14T=π15T=0\pi_{14}T = \pi_{15}T = 0, so π15(T/3)=0\pi_{15}(T/3) = 0. Thus β~\widetilde{\beta} is the boundary of a class in π16T/(3,v1)\pi_{16}T/(3,v_1). The j=1j=1 case of [6 Lemma 3.22] identifies the actions of aa and Δ6\Delta^6 up to sign on T/(3,v1)T/(3,v_1). By linearity of the v1v_1 boundary this yields akβ~=Δ6kβ~a^k\widetilde{\beta} = \Delta^{6k}\widetilde{\beta} in π∗(T/3)\pi_{*}(T/3). The map ∂ˉ:Σ−1(T/3)→R/3\bar{\partial}: \Sigma^{-1}(T/3) \to R/3 in the degrees of interest is the equalizer boundary with mod-3 coefficients, and it is R/27R/27-linear [6 Lemma 3.23]. Thus the lift b0=∂ˉ(Δ6β~)b_0=\bar{\partial}(\Delta^6\tilde{\beta}) of ∂(βΔ6)\partial(\beta\Delta^6) in π154(R/3)\pi_{154}(R/3) satisfies

∂0(akb0)=∂(βΔ6(k+1)).\partial_0(a^k b_0)=\partial(\beta\Delta^{6(k+1)}).

Here the boundary ∂0\partial_0 on T/3T/3 is linear over TT.

After rescaling b0b_0 by a unit, exactness gives bR−b0=q0uRb_R-b_0=q_0u_R for a class uR∈π154Ru_R\in\pi_{154}R. The inclusion i:R/3→R/27i:R/3\to R/27 of the fiber of R/27→R/9R/27\to R/9 is R/27R/27-linear [6 §3.3] and satisfies ∂0=∂27i\partial_0=\partial_{27}i. Also iq0uR=9q27uR=0iq_0u_R=9q_{27}u_R=0 because 3uR=03u_R=0. Therefore

∂0(ak(bR−b0))=∂27(akiq0uR)=0.\partial_0(a^k(b_R-b_0))=\partial_{27}(a^kiq_0u_R)=0.

This proves

h(Xk)=∂(βΔ6(k+1))≠0h(X_k)=\partial(\beta\Delta^{6(k+1)})\ne0

for all kk. The final nonvanishing follows because βΔ6(k+1)\beta\Delta^{6(k+1)} spans the order-three group in stem 154+144k154+144k of TT, on which ψ2−1\psi^2-1 is zero [6 Theorem 2.1]. The boundary is detected in filtration three in the modified spectral sequence for J2J^2, by the same equalizer computation as in Section 3, with Δ6\Delta^6 replaced by Δ6(k+1)\Delta^{6(k+1)}. The unit comparison therefore puts XkX_k in filtration at most three, and sphere sparseness gives filtration at least three. The classes have order three because they lie in the image of ∂0\partial_0. This completes the proof of Theorem 1.1.

Acknowledgment

The author is supported by the NSFC Grant No. 12405001. The author thanks AI for its help.

References

References

  1. [1]T. Bauer, Computation of the homotopy of the spectrum tmf, Geom. Topol. Monogr. 13 (2008), 11–40. DOI 10.2140/gtm.2008.13.11. arXiv:math/0311328.DOI
  2. [2]M. Behrens and S. Pemmaraju, On the existence of the self map v₂⁹ on the Smith–Toda complex V(1) at the prime 3, Contemp. Math. 346 (2004), 9–49. DOI 10.1090/conm/346/06284. arXiv:math/0303223. Author’s copy.
  3. [3]E. Belmont and G. Wang, Adams–Novikov charts and data at the prime 3, computational data and code, revision f530898.
  4. [4]E. Belmont and K. Shimomura, Beta families arising from a v₂⁹ self-map on S/(3, v₁⁸), Algebr. Geom. Topol. 23 (2023), 4349–4378. DOI 10.2140/agt.2023.23.4349. arXiv:2109.01059.DOI
  5. [5]R. Burklund, J. Hahn, and A. Senger, On the boundaries of highly connected, almost closed manifolds, Acta Math. 231 (2023), 205–344. DOI 10.4310/ACTA.2023.v231.n2.a1. arXiv:1910.14116.DOI
  6. [6]C. Carrick and J. M. Davies, Nonvanishing of products in v₂-periodic families at the prime 3, Adv. Math. 502 (2026), 111183. DOI 10.1016/j.aim.2026.111183. arXiv:2410.02564v3.DOI
  7. [7]J. M. Davies, Revisiting the β₁-action on the 3-primary stable homotopy groups of spheres, arXiv:2511.02324v2, 2026.arxiv.org/abs/2511.02324
  8. [8]D. C. Ravenel, Complex Cobordism and Stable Homotopy Groups of Spheres, 2nd ed., AMS Chelsea Publishing, 2004. DOI 10.1090/chel/347. Author’s copy.DOI
  9. [9]D. C. Ravenel, Localization with respect to certain periodic homology theories, Amer. J. Math. 106 (1984), 351–414. DOI 10.2307/2374308.DOI
  10. [10]G. Wang, Computations of the Adams–Novikov E₂-term, Chinese Ann. Math. Ser. B 42 (2021), 551–560. DOI 10.1007/s11401-021-0277-2.DOI
  11. [11]G. Wang, MinimalResolution, source code. Original repository. Prime-3 version by E. Belmont and G. Wang, revision e8a5482.

Paper details

vlnc:W_q43tvhm7
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 11

downloads · 0

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

  • Mathematics

AI

View statement
Looking for related papers…
Contents