Exactly three survivors in the prime-three Kervaire family

← All digests

Family 309 · Topology
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-22

1. The question

Stable homotopy theory studies maps between spheres after allowing repeated suspension. Its algebraic approximations produce many candidate classes, but some candidates disappear before they represent actual maps. Even a surviving candidate can leave uncertainty about the additive order of the represented element.

The prime-three Kervaire problem asks these questions for a distinguished sequence of candidates in the mod-three Adams spectral sequence. The manuscript claims a complete answer: exactly the classes with indices $0,2,3$ survive, and each has a representative of additive order three.

The new positive case lies in stable stem 322. The negative result excludes every index at least four. This is the standard odd-primary Adams-family problem at the prime three, not the classical prime-two Kervaire invariant-one problem with its different indexing and dimensions.

2. Background and essential definitions

Stable stems and prime-three information

For sufficiently large $N$, repeated suspension stabilizes the groups of based maps

$$[S^{N+k},S^N].$$

Their stable value is the $k$-th stable homotopy group of spheres, written $\pi_k^S$. The integer $k$ is the stem. Addition comes from the stable homotopy structure, and an element has order three if it is nonzero but three times it is zero.

Working at the prime three isolates the part associated with powers of three. Positive stable stems are finite, so the relevant three-local and three-complete versions carry the same finite three-primary information for this question.

A spectral sequence is a sequence of tests

The mod-three Adams spectral sequence starts from algebra built from mod-three cohomology operations, organized in the Steenrod algebra $\mathcal A_3$. Its first computational page is

$$E_2^{s,t}=\operatorname{Ext}_{\mathcal A_3}^{s,t} (\mathbb F_3,\mathbb F_3).$$

Here $\mathbb F_3$ is the three-element field. The Ext group measures algebraic extension data; no calculation of Ext is needed to understand the statement. The indices are filtration $s$ and internal degree $t$, with stable stem $t-s$.

Successive pages have differentials that eliminate candidates. A nonzero class surviving to $E_\infty$ is called a permanent cycle and detects an associated-graded part of actual stable homotopy.

Associated-graded information is not the full group. For example, a filtered group of order nine can have two successive quotients of order three without itself being a direct sum of two order-three groups. Thus survival of a mod-three class does not automatically give a representative killed by three. The possible actual representatives form a detection coset, differing by higher-filtration elements.

The standard Kervaire candidates

The family studied here consists of classes

$$b_j\in E_2^{2,\,4\cdot3^{j+1}},\qquad j\geq0.$$

They lie on the Adams two-line, meaning filtration two. Their candidate stable stems are

$$4\cdot3^{j+1}-2.$$

These distinguished classes arise from relations among primary cohomology operations. The problem is not to classify every element of three-primary stable homotopy, only the fate of this particular sequence and the order of its surviving representatives.

3. Development history and previous partial results

The terminology reflects an analogy with the classical Kervaire invariant. Browder related the prime-two framed-manifold question to particular squares on the Adams two-line. At odd primes, the standard classes $b_j$ play a corresponding algebraic role, but the family and geometric conventions differ.

Toda proved that $b_1$ supports a nonzero differential at every odd prime. Ravenel subsequently used formal-group and Morava-theoretic methods to exclude every $b_j$ with $j\geq1$ at odd primes at least five. Prime three behaves differently.

The classes in stems 10 and 106 were known to survive with order-three representatives. The stem-106 case illustrates an important distinction between two spectral sequences. Its mod-three Adams class is detected by a corrected combination of Adams–Novikov beta-family terms, even though one obvious beta-family representative by itself supports a differential. Killing one chosen preimage does not necessarily kill the Adams class when other corrected preimages remain.

Selick formulated the strong odd-primary condition through maps involving a Moore spectrum and a nonzero primary operation. Amelotte connected this order-sensitive condition to decompositions of loop spaces and to multiplications on Anick spaces. These applications require actual order-three representatives, not merely permanence of an associated-graded class.

Hill, Hopkins and Ravenel proposed examining an action of the cyclic group $C_9$ on height-six Morava E-theory at three. Their equivariant norm and fixed-point machinery provides tools for this program independently of its famous prime-two application.

Belmont and Ray established detection results and calculated relevant group cohomology under an explicit coefficient-action hypothesis. The present manuscript claims the needed completed coefficient model and finite-page differential constraints. It also departs from one proposed route: it claims that a literal full vanishing gap in degree $-2$ is false, and uses outgoing differentials instead. This distinction prevents a broad vanishing assumption from replacing the particular obstruction actually needed.

4. What the new manuscripts claim

The September 24, 2026 manuscript defines two sets: the indices whose $b_j$ survive, and the indices whose detection cosets contain an element of additive order three. It claims that both sets equal

$$\{0,2,3\}.$$

The resulting classification is:

Index	Candidate stem	Claimed outcome
$0$	10	Survives; an order-three representative exists
$1$	34	Does not survive
$2$	106	Survives; an order-three representative exists
$3$	322	Survives; an order-three representative exists
$j\geq4$	$4\cdot3^{j+1}-2$	Does not survive

The order statement is existential within each detection coset. It does not assert that every representative of the same associated-graded class has order three.

Using Selick’s and Amelotte’s results, the paper also claims a complete product-decomposition criterion for a family of unstable spaces. Let $S^{2n+1}\{3\}$ be the homotopy fiber of the degree-three map of a sphere, and let $\Omega$ denote the based loop space. For every $n>1$, after localization at three,

$$\Omega S^{2n+1}\{3\}$$

has a nontrivial product decomposition exactly when $n\in\{3,27,81\}$. Nontrivial means that both factors are noncontractible. A homotopy fiber records points together with paths witnessing their image at a chosen basepoint, so it retains the homotopical failure of the degree-three map to be an equivalence.

The manuscript further gives corresponding associativity results for Anick spaces and an explicit decomposition at $n=81$, through the cited earlier equivalences.

5. Why the advance matters

Importance: 5/5, provisional. If correct, the theorem completes the standard prime-three Kervaire-family problem, supplying both a new high-dimensional survivor and a uniform exclusion of all later candidates.

The stem-322 element is not merely an algebraic entry in a chart. The claimed construction produces an actual stable sphere element killed by three. That order refinement is what activates the unstable-space consequences.

The negative branch also addresses a subtle detection issue. It retains all relevant corrected preimages rather than treating the fate of one beta-family expression as decisive. This matters precisely because the known stem-106 survivor already demonstrates that corrections can change the outcome.

The result organizes an infinite family with a finite list of exceptions. Its reach is substantial within stable homotopy theory, while remaining distinct from a computation of all three-primary stable stems.

6. The main idea

Use a detector sensitive to every possible survivor

The exclusion branch uses height-six Morava E-theory $E_6$, a generalized cohomology theory encoding deformations of a height-six formal group. Its coefficients carry a $C_9$-action. The homotopy fixed-point object $E_6^{hC_9}$ records invariance together with the necessary coherent homotopies, rather than taking only ordinary fixed coefficients.

The paper first claims that an actual sphere class detected by $b_j$, for $j\geq2$, must have a nonzero finite-order image in this detector. This prevents a hypothetical survivor from simply escaping the test.

It constructs the completed coefficient action and analyzes the geometry of fixed deformations. Character and determinant constraints rule out certain possible maps in coefficient cohomology. Equivariant norms and geometric fixed points give finite Euler-class bounds, which impose deadlines for differentials after Euler localization.

Euler localization is the operation called Tate localization in this argument. It is not rationalization, and the proof does not rely on an unrestricted claim that every relevant localized homotopy group vanishes.

An outgoing differential is the decisive obstruction

A subtle point is the direction of a differential. Passing to a localized spectral sequence can introduce an incoming differential that kills an image even when the original class was genuine. Mere disappearance after localization would therefore not exclude a sphere class.

The paper instead forces every higher detector to support a nonzero outgoing differential. It organizes the remaining possibilities into a finite array of one-dimensional groups with tightly restricted degrees and characters. Translation classes are used only on pages where they and their inverses have been shown to survive.

The known $b_2$ survivor fixes one incoming pairing in the array. The resulting differential constraints then force outgoing differentials on the pattern containing every detector with $j\geq4$. This contradicts the required image of an actual sphere class and supplies the infinite exclusion.

Construct the stem-322 element with its order intact

The positive branch starts from the known order-three stem-106 class. Let $M=S/3$ be the Moore spectrum, the cofiber in

$$S\xrightarrow{\,3\,}S\longrightarrow M\longrightarrow\Sigma S.$$

The exact sequence associated with this cofiber keeps track of elements killed by three. This makes it useful for producing the strong order statement, not just an Adams permanent cycle.

A truncated extended cube, based on Toda’s construction, turns a Moore-spectrum map from the known class into a new map involving a four-cell spectrum. Its cone carries a nonzero primary operation $P^{81}$, an operation on mod-three cohomology that detects the desired Kervaire behavior.

The initial map has obstruction components preventing the required lift. The manuscript constructs a correction of Adams–Novikov filtration at least two with the same quotient image. Subtracting it makes the map lift to the Moore spectrum while preserving the nonzero primary operation. The resulting actual Moore boundary is a nonzero sphere element killed by three, hence of exact order three, detected by $b_3$.

Finite calculations support the correction

The correction needs particular finite-page Adams–Novikov information: a cubic boundary, a multiplication that vanishes on an indicated page, and higher-filtration vanishing estimates.

The paper presents finite algebraic resolutions, product rules and elimination procedures for these assertions. It constrains possible differentials by affine equations and by the condition that a differential squares to zero. The stated bounds apply to every differential satisfying those constraints; the exposition does not select an arbitrary favorable differential from the possibilities.

These finite calculations serve the specific lifting argument. They are not presented as a full computation of all homotopy in the relevant range, and the positive Moore-spectrum construction is separate from the height-six exclusion branch.

7. Scope, limitations, and open questions

The theorem concerns the standard classes $b_j$ in the mod-three Adams spectral sequence. It does not classify all order-three elements, all beta-family elements, or the classical prime-two Kervaire invariant-one classes.

Survival and additive order remain separate conclusions. The manuscript claims that each surviving coset contains an order-three representative, not that its associated-graded appearance alone determines the order of every representative.

The unstable product classification assumes $n>1$ and localization at three. Its index conversion is $n=3^{j+1}$, giving $3,27,81$ from $j=0,2,3$; confusing these two index systems would shift the claimed dimensions.

The source’s rejection of a proposed full vanishing gap does not reject every other conjectural feature of the detector. It explicitly leaves separate periodicity and torsion assertions untouched. Its proof route uses finite-page outgoing differentials instead.

8. Sources and further reading

Ravenel’s odd-primary nonexistence theorem and Miller–Ravenel–Wilson’s Adams–Novikov calculations provide the classical algebraic background. Toda’s extended-power construction underlies the positive geometric branch.

Amelotte’s degree-three fiber and Anick-space results explain the order-sensitive unstable consequences. Belmont–Ray’s height-six detector study gives the recent program and conditional coefficient calculations addressed by the source manuscript.

Source manuscripts

• M1. OpenAI, The Kervaire invariant problem at the prime three (September 24, 2026).

Sources consulted on 22 October 2026 at repository revision 1581d4c. Download the manuscript citations.