03.08.11 · modern-geometry / k-theory

The group $J (X)$ and the $J$ -homomorphism

shipped3 tiersLean: none

Anchor (Master): Adams *On the groups $J(X)$ I-IV* (Topology 2-5, 1963-1966); Quillen *The Adams conjecture* (Topology 10, 1971); Sullivan *Genetics of homotopy theory and the Adams conjecture* (Ann. Math. 100, 1974); Atiyah *Thom complexes* (Proc. Lond. Math. Soc. 11, 1961); Whitehead *On the homotopy groups of spheres and rotation groups* (Ann. Math. 43, 1942); Lawson-Michelsohn *Spin Geometry* §IV.1

Intuition [Beginner]

Two vector bundles can be different as bundles yet behave the same once you forget enough information. The right amount of "forgetting" for many problems in topology is: throw away the linear structure and remember only the underlying sphere bundle, up to fibrewise homotopy. Bundles that match under this coarser relation are said to be $J$ -equivalent. The group $J (X)$ packages all the $J$ -equivalence classes of stable virtual bundles on a finite cell complex $X$ , and it sits as a quotient of the K-theory group.

The reason this group matters: many geometric invariants depend only on the sphere-bundle data, not on the full bundle. Examples include the existence of vector fields on spheres, the question of whether a bundle is fibre-homotopy equivalent to a simple bundle, and the order of certain homotopy classes in the stable homotopy groups of spheres. $J (X)$ is the receptacle for all such answers.

The associated $J$ -homomorphism is a way to convert classical Lie-group data into stable homotopy data: it sends a homotopy class of maps into the orthogonal group to a homotopy class of maps between spheres, by a clutching construction.

Visual [Beginner]

A schematic picture of the situation. On the left, a vector bundle $E \to X$ over a base space, drawn as a stack of fibres each carrying a small disk. In the middle, the same bundle with the linear structure stripped: each fibre is now a sphere, recording only the radial information. On the right, the resulting object: a sphere bundle whose homotopy type is the only thing $J (X)$ remembers.

The picture also hints at the second story: an element of the homotopy of the rotation group $O (k)$ gives a way to glue two disks into a sphere, and the resulting map between spheres is the $J$ -image of the rotation.

Worked example [Beginner]

Take the sphere $X = S^{2}$ . The K-theory ring is $K (S^{2}) = Z [β] / (β^{2})$ with $β$ the Bott class. The reduced piece is $K (S^{2}) = Z \cdot β$ .

Step 1. Identify the candidate relations. Adams operations on $β$ act by $ψ^{k} (β) = k β$ . The Adams conjecture says $ψ^{k} β - k β$ should map to the identity in stable fibre-homotopy classes after multiplying by a suitable power of $k$ . For $S^{2}$ the relation reads $k β - k β = 0$ in the quotient — empty by itself.

Step 2. Use the full statement: the subgroup $T (S^{2})$ is generated by elements of the form $ψ^{k} (ξ) - ξ$ for $ξ$ a virtual bundle, and these all vanish on $S^{2}$ because $K (S^{2})$ has $ψ^{k} = k \cdot id$ and the actual relation is more subtle (one must use the full $J$ -equivalence relation, not just the Adams-conjecture relation, on the small group $K (S^{2}) = Z$ ).

Step 3. The result is $J (S^{2}) = Z /2$ . The class of the tangent bundle $T S^{2}$ generates this group; it is $J$ -equivalent to neither the simple bundle nor its negative, but its double is $J$ -equivalent to the simple bundle.

What this tells us: K-theory of $S^{2}$ is the integers, but only a $Z /2$ piece of it survives in $J (S^{2})$ . The rest is killed by the equivalence relation. This is the simplest example of how $J (X)$ is a coarser invariant than $K (X)$ , and how its size is governed by arithmetic data — here, by the order of the image of the $J$ -homomorphism in degree two.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $X$ be a finite CW complex. Recall the reduced K-group $K (X)$ from 03.08.01: classes of stable virtual complex vector bundles whose virtual rank vanishes on every component of $X$ .

Definition (sphere bundle). Given a real vector bundle $ξ \to X$ of rank $n$ , equip it with a Euclidean metric and let $S (ξ) \to X$ denote the associated unit sphere bundle, with fibre $S^{n - 1}$ . Up to fibrewise homotopy equivalence, $S (ξ)$ does not depend on the choice of metric. For a complex vector bundle of rank $n$ , the underlying real rank is $2 n$ and the associated sphere bundle has fibre $S^{2 n - 1}$ .

Definition ( $J$ -equivalence). Two stable vector bundles $ξ, η$ over $X$ are fibre-homotopy equivalent in the stable sense, written $ξ \sim_{J} η$ , if there exist simple bundles $ε^{a}, ε^{b}$ such that the sphere bundles $S (ξ \oplus ε^{a})$ and $S (η \oplus ε^{b})$ are fibrewise homotopy equivalent.

Definition ( $T (X)$ ). Let $T (X) \subseteq K (X)$ be the subgroup consisting of classes $ξ \in K (X)$ for which $S (ξ)$ is stably fibrewise homotopy equivalent to a sphere bundle of the same rank from a simple bundle. Equivalently, $T (X)$ is the kernel of the natural homomorphism $K (X) \to [X, B (F / O)]$ , where $F$ is the monoid of self-homotopy equivalences of $S^{\infty}$ and $F / O$ is the homotopy quotient by the orthogonal stabiliser. (The same definition adapts verbatim to complex bundles and $K_{C} (X)$ .)

Definition ( $J (X)$ , Atiyah 1961). The $J$ -group of $X$ is the quotient $$ J(X) = \widetilde K(X) / T(X). $$ It records the fibre-homotopy classes of stable sphere bundles obtained from vector bundles, by a coarser equivalence than isomorphism.

Definition (clutching map). Fix $k \geq 1$ . The sphere $S^{k}$ is the union of two disks $D_{+}^{k} \cup_{S^{k - 1}} D_{-}^{k}$ . A continuous map $α : S^{k - 1} \to O (m)$ extends to two trivialisations of the bundle on the closed disks $D_{\pm}^{k}$ , glued along $S^{k - 1}$ via $α$ . The resulting bundle is the clutching bundle $E_{α} \to S^{k}$ of rank $m$ .

Definition ( $J$ -homomorphism, Whitehead 1942). For $n, k \geq 1$ let $α \in π_{n} (O (k))$ be a homotopy class. The element $α$ acts on $S^{k - 1}$ by rotation, producing a clutching map. Forming the sphere bundle of the clutching construction and contracting the two disk factors yields a map $S^{n + k} \to S^{k}$ . Its homotopy class defines $J (α) \in π_{n + k} (S^{k})$ . Stabilising in $k$ produces $$ J : \pi_n(\mathrm O) \longrightarrow \pi_n^S(S^0) $$ the stable $J$ -homomorphism. The construction is natural in $n$ .

Counterexamples to common slips

$J (X)$ is not a direct summand of $K (X)$ in general. The short exact sequence $0 \to T (X) \to K (X) \to J (X) \to 0$ need not split; for $X = S^{4 k - 1}$ the obstruction is detected by Adams' $e$ -invariant and is honestly cyclic of order tied to Bernoulli denominators.
The relation generating $T (X)$ is not just $ψ^{k} (L) - L^{\otimes k}$ for line bundles. The full subgroup contains all elements of the form $ψ^{k} (ξ) - ξ$ for $ξ \in K (X)$ once these are spherically simple, with arithmetic factors. The Adams conjecture is the statement that the right multiples of $ψ^{k} (ξ) - ξ$ already lie in $T (X)$ for every $ξ$ .
The $J$ -homomorphism is not surjective in general. Its image is a small but identifiable cyclic subgroup of $π_{n}^{S}$ in each dimension; the rest of $π_{n}^{S}$ is captured by other constructions (Adams' filtration, the Hopf map iterates, the $η, ν, σ$ classes).
The clutching construction outputs an element of $π_{n + k} (S^{k})$ , not of $π_{n + k - 1} (S^{k})$ . The dimension shift comes from the suspension implicit in regarding $S^{k}$ as glued from two disks rather than from the boundary $S^{k - 1}$ of one disk.

Key theorem with proof [Intermediate+]

Theorem (well-definedness of $J (X)$ ; Atiyah 1961). Let $X$ be a finite CW complex. The set $T (X) \subseteq K (X)$ is a subgroup, and $J (X) = K (X) / T (X)$ is well-defined as an abelian group. The construction is natural in $X$ : a continuous map $f : Y \to X$ induces $f^ : J(X) \to J(Y) $, an d t h e d ia g r am w i t h$ \widetilde K$-restriction commutes.*

Proof. Direct sum of vector bundles induces direct sum of sphere bundles via the join construction $S (ξ \oplus η) ≅ S (ξ) * S (η)$ at each fibre. Fibre-homotopy equivalences are closed under direct sum: if $ξ \sim_{J} ξ^{'}$ and $η \sim_{J} η^{'}$ then $ξ \oplus η \sim_{J} ξ^{'} \oplus η^{'}$ by combining the two equivalences fibrewise. Hence the relation $\sim_{J}$ on $K (X)$ is preserved under the additive structure; the subgroup $T (X)$ of classes $J$ -equivalent to the simple class is a subgroup of $K (X)$ . Naturality in $X$ is the functoriality of K-theory together with pull-back of fibrewise-homotopy equivalences. $□$

Theorem (well-definedness of the $J$ -homomorphism; Whitehead 1942). The clutching construction defines a homomorphism $J : π_{n} (O (k)) \to π_{n + k} (S^{k})$ for each $k$ and $n$ . Stabilising in $k$ (via the inclusions $O (k) ↪ O (k + 1)$ and the suspension $π_{n + k} (S^{k}) \to π_{n + k + 1} (S^{k + 1})$ ) yields the stable $J$ -homomorphism $J : π_{n} (O) \to π_{n}^{S} (S^{0})$ .

Proof. Two homotopic maps $α_{0}, α_{1} : S^{n} \to O (k)$ define clutching bundles $E_{α_{0}}, E_{α_{1}}$ over $S^{n + 1}$ . The homotopy between them is a map $S^{n} \times I \to O (k)$ , and applying clutching at each time produces a bundle over $S^{n + 1} \times I$ restricting to $E_{α_{0}}$ and $E_{α_{1}}$ at the endpoints. The resulting Thom-like construction on sphere bundles defines a homotopy between the spherical clutching maps. Hence the assignment $α \mapsto [spherical clutching]$ factors through $π_{n} (O (k))$ .

Additivity: for $α, β : S^{n} \to O (k)$ with $α + β$ defined by the loop addition on $O (k)$ (pinching $S^{n}$ to $S^{n} \lor S^{n}$ and sending the two factors to $α$ and $β$ ), the clutching bundle splits as a fibred sum of $E_{α}$ and $E_{β}$ over a pinch sphere $S^{n + 1} \to S^{n + 1} \lor S^{n + 1}$ . The corresponding spherical clutching map decomposes by pinch into the sum in $π_{n + k} (S^{k})$ . Thus $J (α + β) = J (α) + J (β)$ .

Stability is the verification that $J$ commutes with the stabilisation maps on both sides. The inclusion $O (k) ↪ O (k + 1)$ corresponds, on the bundle side, to adding a simple line bundle; on the spherical clutching side, this adds a free suspension, which is exactly the suspension homomorphism on $π_{n + k} (S^{k}) \to π_{n + k + 1} (S^{k + 1})$ . The diagram commutes. $□$

Theorem (Adams, image of $J$ in dimensions $4 k - 1$ ). For every $k \geq 1$ , the image of the stable $J$ -homomorphism $J : π_{4 k - 1} (O) \to π_{4 k - 1}^{S} (S^{0})$ is a cyclic subgroup. Let $m_{k}$ denote the denominator of $B_{2 k} /4 k$ in lowest terms, where $B_{2 k}$ is the $2 k$ -th Bernoulli number. Then $$ |\mathrm{Im}(J : \pi_{4k-1}(\mathrm O) \to \pi_{4k-1}^S)| \in {m_k, 2 m_k}. $$ Adams' $e$ -invariant $e_{C} : π_{4 k - 1}^{S} \to Q / Z$ embeds the image $J$ -cyclic group as the cyclic subgroup of order $m_{k}$ inside $Q / Z$ .

Sketch. Upper bound (Adams 1965, $J (X)$ -II): the $e$ -invariant on a Hopf-like element produces a rational number whose denominator is forced to divide $m_{k}$ by the Atiyah-Hirzebruch-Riemann-Roch computation of the Chern character on the bundle whose clutching realises $J (α)$ . The Bernoulli-denominator output appears through the Todd genus of complex line bundles on $C P^{\infty}$ , whose universal computation involves $1 - x e^{- x}$ inverted as a power series in $x$ , with coefficients $B_{2 k} / (2 k)!$ . Lower bound (Adams 1965, $J (X)$ -III): Adams operations on stunted projective spaces detect a class of order exactly $m_{k}$ inside $π_{4 k - 1}^{S}$ as a $J$ -image, via a virtual-bundle computation on $C P^{N} / C P^{N - r}$ . The two bounds agree up to a factor of $2$ , which depends on the parity of $k$ and was resolved by Adams in $J (X)$ -IV (1966) using the real $e$ -invariant and the Adams conjecture (then conjectural). $□$

Bridge. The group $J (X)$ builds toward the entire infrastructure of stable fibre-homotopy theory for vector bundles. The foundational reason it appears as a quotient of $K (X)$ is exactly the observation that the spherical fibre-homotopy type forgets the linear structure, and the kernel $T (X)$ is what one quotients by to land in the smaller invariant. This is exactly the same algebraic pattern that appears again in 03.13.04 (Atiyah-Hirzebruch spectral sequence) when one passes from $K$ -theory to its associated graded by the skeletal filtration: a coarser invariant is recovered as a quotient with arithmetic denominators. The central insight is that Adams operations $ψ^{k}$ from 03.08.02 generate the kernel modulo arithmetic factors — the Adams conjecture, proved by Quillen 1971 and Sullivan 1974, identifies $ψ^{k} ξ - ξ$ as $J$ -spherically simple after multiplication by a power of $k$ . Putting these together, $J (X)$ is computable from $K (X)$ together with all $ψ^{k}$ , and this identifies $J (X)$ with the cokernel of the universal $ψ^{k} - id$ map. The bridge is the recognition that the topological $J$ -quotient and the arithmetic $ψ^{k}$ -coinvariants agree after localisation, generalises the Bott periodicity of 03.08.07 (the $J$ -homomorphism inherits mod- $8$ periodicity through Bott), and is dual to the cohomological story for spheres — the Adams $e$ -invariant identifies $Im (J)_{4 k - 1}$ as the denominator subgroup, and this arithmetic pattern appears again in 03.13.04 where Bernoulli-number denominators reappear as differentials.

Exercises [Intermediate+]

Exercise 5 (medium, short-answer).

Sketch why the $J$ -homomorphism $J : π_{n} (O) \to π_{n}^{S} (S^{0})$ is independent (after stabilisation) of the dimension $k$ of the rotation group $O (k)$ used in the clutching construction.

Hint

The inclusion $O (k) ↪ O (k + 1)$ adds a simple direction; in the clutching picture this corresponds to suspending the spherical map.

Answer

A class $α \in π_{n} (O (k))$ has image in $π_{n} (O (k + 1))$ under the standard inclusion. The corresponding clutching bundles on $S^{n + 1}$ differ by a simple summand: $E_{α} \oplus ε^{1}$ . The associated sphere bundles satisfy $S (E \oplus ε^{1}) ≅ Σ S (E)$ at each fibre (the unreduced suspension of the sphere $S^{k - 1}$ is $S^{k}$ ). Hence the clutching spherical map for the image of $α$ in $π_{n} (O (k + 1))$ is the suspension of that for $α$ , and lands in $π_{n + k + 1} (S^{k + 1})$ as the image of $π_{n + k} (S^{k})$ under suspension. In the stable range $k \geq n + 2$ (Freudenthal), suspension is an isomorphism. Hence stabilising in $k$ gives a well-defined map into $π_{n}^{S}$ independent of $k$ . Rubric: full credit for the suspension-suspension identification with a Freudenthal mention.

Exercise 6 (medium, symbolic).

Compute $J (S^{2})$ using the Adams conjecture and the fact that $K (S^{2}) = Z$ .

Hint

The relation $ψ^{k} - 1$ on $K (S^{2})$ is multiplication by $k - 1$ . Take the gcd over varying $k$ to find $T (S^{2})$ .

Answer

The Adams conjecture together with explicit identification of small multiples yields $T (S^{2}) = 2 Z \subseteq Z = K (S^{2})$ . The relation $ψ^{k} (β) - β = (k - 1) β \in T (S^{2})$ for every $k$ gives $g cd_{k \geq 2} (k - 1) = 1$ on the face of it, but the correct statement involves $k^{N}$ -multiples and arithmetic factors. For $S^{2}$ the precise result is $T (S^{2}) = 2 Z$ , generated by the tangent-bundle class $[T S^{2}] = 2 β$ which is $J$ -equivalent to the simple bundle (the sphere bundle of $T S^{2}$ is the unit tangent bundle of $S^{2}$ , fibrewise homotopy equivalent to $S^{1} \times S^{2} / S^{1}$ via Hopf-fibration arguments). Hence $J (S^{2}) = Z /2 Z$ . Rubric: full credit for the identification of $T (S^{2}) = 2 Z$ with any valid argument.

Exercise 7 (hard, short-answer).

Define Adams' $e$ -invariant $e_{C} : π_{2 n - 1}^{S} (S^{0}) \to Q / Z$ via the K-theoretic Chern character. Sketch why its image on $Im (J : π_{4 k - 1} (O) \to π_{4 k - 1}^{S})$ has order exactly the denominator of $B_{2 k} /4 k$ .

Hint

Apply $ch$ to the K-theory Thom class of the clutching bundle and reduce modulo integers. Bernoulli numbers appear via the Todd-class expansion $x / (1 - e^{- x})$ .

Answer

For $f : S^{2 n - 1} \to S^{0}$ stable, form the mapping cone $C_{f}$ . The K-theory of $C_{f}$ fits in a short exact sequence $0 \to K (S^{2 n}) \to K (C_{f}) \to K (S^{0}) \to 0$ , and the Chern character of a chosen lift of the generator produces a rational class whose reduction mod $Z$ is independent of the lift. Call this $e_{C} (f)$ . For $f = J (α)$ with $α \in π_{4 k - 1} (O)$ corresponding to a Bott-class generator, the Chern character of the associated clutching bundle on $S^{4 k}$ involves $Td (ξ)$ to compute the Hirzebruch-Riemann-Roch correction, and the coefficient of the top class is $B_{2 k} / (4 k)$ by the universal Bernoulli expansion of the Todd genus. Hence $e_{C} (J (α)) \equiv \pm B_{2 k} /4 k (mod Z)$ , generating the cyclic subgroup of $Q / Z$ of order equal to the denominator of $B_{2 k} /4 k$ in lowest terms. Adams' upper-bound result is that $Im (J)_{4 k - 1}$ injects into $Q / Z$ under $e_{C}$ ; the lower-bound result is that the injection hits this subgroup. Rubric: full credit for the chain $ch \to Td \to B_{2 k}$ with the cyclic-subgroup conclusion.

Exercise 8 (hard, short-answer).

State and discuss the Adams conjecture: for $ξ \in K (X)$ a real virtual vector bundle and $k \geq 2$ an integer, there exists $N$ such that $k^{N} (ψ^{k} (ξ) - ξ)$ lies in $T (X)$ . Outline the strategy of Quillen's 1971 proof.

Hint

Quillen's method uses étale homotopy theory of complex algebraic varieties: he reduces to the universal case $X = B U (n)$ , then realises $ψ^{k}$ as a profinite Galois action on the étale homotopy type after $p$ -localisation.

Answer

Adams conjecture (Adams 1963 $J (X)$ -I): for every finite CW complex $X$ , every real virtual bundle $ξ \in K_{O} (X)$ , and every $k \geq 2$ , there exists $N = N (X, ξ, k)$ such that the class $k^{N} (ψ^{k} (ξ) - ξ) \in K_{O} (X)$ vanishes in $J (X)$ , equivalently lies in $T (X)$ . The conjecture asserts $ψ^{k} ξ$ and $ξ$ are stably fibrewise equivalent up to a power of $k$ .

Quillen's proof (Topology 10, 1971, 67-80) ^{[Quillen Adams conjecture]}: localise at a prime $p ∤ k$ . Reduce to the universal example $X = B U (n)$ or $B O (n)$ by classifying-space functoriality. Use étale homotopy theory: $B U (n)$ has an étale-homotopy presentation modelled by the algebraic stack of rank- $n$ vector bundles on $Spec (Z)$ . The Adams operation $ψ^{k}$ corresponds, at $p$ -completion with $p ∤ k$ , to the Galois automorphism $Frob^{k}$ in the Galois group $Gal (\overline{F}_{p} / F_{p}) ≅ Z$ . Galois action on étale cohomology is automatic; lifting to étale homotopy types and applying Artin-Mazur étale homotopy theory gives a map $ψ^{k} (ξ) \to ξ$ at the level of the spherical fibration after $p$ -completion. Hence $ψ^{k} ξ - ξ$ vanishes in $J (X)_{p}^{\land}$ for every prime $p ∤ k$ ; multiplication by $k^{N}$ kills the $p ∣ k$ part. The two pieces assemble to the integer statement.

Sullivan's alternative proof (Ann. Math. 100, 1974) uses profinite localisation in algebraic topology directly, replacing étale homotopy theory by Sullivan's own genetics-of-homotopy machinery. Both proofs identify the Adams operations on $K_{O} (X)$ with profinite Galois actions on the spherical-fibration classifying space, with the integer denominators on the K-theory side absorbed into the spherical-fibration side by the conjecture. Rubric: full credit for the statement of the conjecture and at least one of the two proof strategies sketched.

Lean formalization [Intermediate+]

Mathlib has Bernoulli numbers and the categorical infrastructure for derived functors, but lacks topological K-theory, stable fibre-homotopy classes, and the $J$ -homomorphism. The intended formalisation reads schematically:

import Mathlib.NumberTheory.Bernoulli
import Mathlib.AlgebraicTopology.SimplicialSet
import Mathlib.Topology.VectorBundle.Basic

/-- For a finite CW complex $X$, $T(X)$ is the subgroup of $\widetilde K(X)$
of classes whose associated sphere bundles are stably fibrewise-homotopy
equivalent to a simple sphere bundle of the same rank. -/
def TGroup (X : TopCat) [FiniteCW X] : Subgroup (reducedKTheory X) :=
  sorry  -- stable fibre-homotopy equivalence is the equivalence relation

/-- The $J$-group of $X$ is the quotient of reduced K-theory by $T(X)$. -/
def JGroup (X : TopCat) [FiniteCW X] : AddCommGroup :=
  reducedKTheory X ⧸ TGroup X

/-- The stable $J$-homomorphism from the homotopy of the stable orthogonal
group to the stable homotopy groups of spheres. -/
def jHomomorphism (n : ℕ) :
    πₙ (stableOrthogonalGroup) →+ stableHomotopyGroupOfSpheres n :=
  sorry  -- clutching construction + stabilisation

/-- Adams' theorem: for $k \geq 1$, the image of the $J$-homomorphism in
dimension $4k - 1$ has order in $\{m_k, 2 m_k\}$ where $m_k$ is the
denominator of $B_{2k} / 4k$ in lowest terms. -/
theorem adams_image_of_J (k : ℕ) (hk : 1 ≤ k) :
    let m := (bernoulli (2*k) / (4*k : ℚ)).den
    (jHomomorphism (4*k - 1)).range.card ∈ ({m, 2*m} : Set ℕ) :=
  sorry  -- Adams 1965-66 $J(X)$ papers I-IV

/-- Quillen-Sullivan: Adams' conjecture. For $k \geq 2$ and any virtual
real bundle $\xi$ over a finite CW complex, there exists $N$ such that
$k^N(\psi^k \xi - \xi)$ lies in $T(X)$. -/
theorem adams_conjecture (X : TopCat) [FiniteCW X]
    (ξ : reducedKOTheory X) (k : ℕ) (hk : 2 ≤ k) :
    ∃ N : ℕ, k^N • (adamsOp k ξ - ξ) ∈ TGroupReal X :=
  sorry  -- Quillen 1971 étale homotopy proof; Sullivan 1974 profinite proof

The proof gap is substantive. Mathlib needs topological K-theory in the Grothendieck-group sense, a notion of stable fibre-homotopy equivalence of sphere bundles, the clutching construction on continuous orthogonal-group maps, and the Adams $e$ -invariant via the K-theoretic Chern character mod integers. Each piece is formalisable from existing infrastructure; the consolidation as a named statement and proof of the Adams conjecture and the Bernoulli-denominator computation of $Im (J)_{4 k - 1}$ is the formalisation target. Quillen's proof would require étale homotopy theory infrastructure not currently in Mathlib; Sullivan's profinite-localisation approach is closer to existing model-category machinery.

Advanced results [Master]

Theorem (Atiyah, originator definition; Thom complexes, Proc. Lond. Math. Soc. 11, 1961). Let $X$ be a finite CW complex and let $J (X) = K (X) / T (X)$ . Two stable real virtual bundles $ξ, η$ over $X$ have the same image in $J (X)$ if and only if there exist simple bundles $ε^{a}, ε^{b}$ such that the sphere bundles $S (ξ \oplus ε^{a})$ and $S (η \oplus ε^{b})$ are fibrewise homotopy equivalent over $X$ . The relation is the equivalence relation on the additive monoid of stable bundles giving rise to a quotient group $J (X)$ .

The originator definition is the spherical-fibre-homotopy-equivalence formulation; the equivalent description via $T (X) \subseteq K (X)$ as the kernel of the map to stable spherical fibrations is the modern packaging. Both are due to Atiyah; the original 1961 paper introduced the $J$ -group in the context of Thom-complex computations for cobordism.

Theorem (Adams, axiomatic characterisation; On the groups $J (X)$ I, Topology 2, 1963). Let $X$ be a finite CW complex. The quotient $J (X) = K (X) / T (X)$ is the largest quotient of $K (X)$ on which the family of natural endomorphisms ${ψ^{k}}_{k \geq 1}$ acts by multiplication by an integer depending only on the dimension. Equivalently, $T (X)$ is the smallest subgroup of $K (X)$ containing $ψ^{k} (ξ) - k^{n} ξ$ for $ξ$ a line bundle of degree- $n$ component and all $k$ , closed under the Adams-conjecture relations.

The axiomatic description is the operational tool: $J (X)$ is computable from $K (X)$ together with the action of Adams operations. This is exactly the conjectural picture before Adams 1963; the precise statement requires the Adams conjecture proved by Quillen 1971 to be unconditional.

Theorem (Quillen-Sullivan, Adams conjecture; Quillen Topology 10, 1971; Sullivan Ann. Math. 100, 1974). Let $X$ be a finite CW complex and $ξ \in K_{O} (X)$ a stable real virtual vector bundle. For every integer $k \geq 2$ , there exists $N \geq 0$ such that the element $k^{N} (ψ^{k} (ξ) - ξ)$ lies in $T (X)$ . Equivalently, after multiplying by a suitable power of $k$ , the bundles $ψ^{k} (ξ)$ and $ξ$ become stably fibrewise-homotopy equivalent.

The Adams conjecture, conjectured by Adams in 1963 ( $J (X)$ -I) and proved by Quillen in 1971 using étale homotopy theory of algebraic varieties and again by Sullivan in 1974 using profinite localisations of topological spaces, is the foundational consistency check for the $J (X)$ -theory: it identifies $J (X)$ as the coinvariants of $K (X)$ under all Adams operations, up to torsion arithmetic. After the Adams conjecture, the calculation of $J (X)$ for many specific $X$ reduces to a calculation of $K (X)$ together with the eigenvalues of $ψ^{k}$ , which is purely algebraic.

Theorem (Adams, $J$ -homomorphism in dimensions $4 k - 1$ ; $J (X)$ -II-III-IV). For $k \geq 1$ , the image of the stable $J$ -homomorphism $J : π_{4 k - 1} (O) \to π_{4 k - 1}^{S}$ is a cyclic group of order in ${m_{k}, 2 m_{k}}$ , where $m_{k}$ is the denominator of $B_{2 k} /4 k$ in lowest terms. The factor of $2$ ambiguity is resolved by the real $e$ -invariant $e_{R} : π_{4 k - 1}^{S} \to Q / Z$ : the image of $J$ equals the cyclic subgroup of order $m_{k}$ if $k$ is even and of order $2 m_{k}$ if $k$ is odd.

The Bernoulli-denominator formula identifies $Im (J)_{4 k - 1}$ with an arithmetically computable cyclic group: $m_{1} = 24, m_{2} = 240, m_{3} = 504, m_{4} = 480, m_{5} = 264, m_{6} = 65520$ , and so on. These are exactly the orders of the cyclic groups giving the entire stable homotopy of spheres in those dimensions, up to the small contribution from the rest of $π_{*}^{S}$ : $π_{3}^{S} = Z /24$ , $π_{7}^{S} = Z /240$ are entirely the $J$ -image.

Theorem ( $J$ -homomorphism in dimensions $\neq \equiv 3 mod 4$ ). In dimensions $n \equiv 0 mod 4$ the homotopy group $π_{n} (O) = 0$ (Bott), so $J$ is the simple map. In dimensions $n \equiv 1, 2 mod 8$ , $π_{n} (O) = Z /2$ and $J$ injects the $Z /2$ generator into $π_{n}^{S}$ as the Hopf-style element $η^{a}$ (specifically $η \in π_{1}^{S}$ and $η^{2} \in π_{2}^{S}$ in low dimensions). In all other dimensions $π_{n} (O)$ vanishes or is $Z$ in the unstable range; the stable picture is dictated by Bott periodicity $π_{n} (O) ≅ π_{n + 8} (O)$ .

Together with the $4 k - 1$ dimension result, this determines $Im (J)$ in every dimension, and the cumulative size pattern is the source of the standard tabulation of stable homotopy groups of spheres in low dimensions.

Theorem ( $J$ -homomorphism on classifying spaces). Let $G$ be a compact Lie group with classifying space $B G$ . The $J$ -homomorphism $J : \pi_(BO) = \widetilde K_O(\mathrm{pt}) = 0 $in p os i t i v e d e g r ee, b u tt h er e l a t i v e$ J $- h o m o m or p hi s m o n s p a ceso f t h e f or m$ BG $f or$ G $co m p a c t g i v es, a f t er s u i t ab l es u s p e n s i o n, c l a sses in$ \pi_*^S(BG) $. T h ese a r e d e t ec t e d b y t h e$ e $- in v a r ian t o n$ K(BG) $an d a sse mb l e d in t o t h e A t i y ah - H i r z e b r u c h s p ec t r a l se q u e n ceco n v er g in g t o$ \pi_*^S(BG)$.*

The $J$ -homomorphism on classifying spaces is the structural tool relating $K$ -theory of $B G$ to stable cohomotopy of $B G$ , and is the source of $K$ -theoretic obstructions to existence of $G$ -structures on stably framed manifolds.

Synthesis. The group $J (X)$ is the foundational quotient of $K$ -theory by fibrewise-homotopy equivalence of sphere bundles, and the central insight is exactly the Quillen-Sullivan identification of $J (X)$ as the coinvariants of $K (X)$ under Adams operations modulo arithmetic powers. Putting these together, $J (X)$ is computable from $K (X)$ plus the universal Adams-operation data, with Bernoulli denominators dictating the integer factors. The bridge to stable homotopy of spheres is the $J$ -homomorphism: a map from $π_{n} (O)$ — itself given by Bott periodicity of 03.08.07 mod $8$ — into $π_{n}^{S} (S^{0})$ , and Adams' $e$ -invariant identifies $Im (J)_{4 k - 1}$ as cyclic of order tied to $B_{2 k} /4 k$ , generalises the eigenvalue picture of 03.08.02 to a quantitative arithmetic invariant, and is dual to the cohomological story of even-dimensional spheres. The same arithmetic pattern appears again in 03.13.04 where Bernoulli denominators reappear as obstruction differentials, identifies $Im (J)_{4 k - 1}$ with the denominator subgroup of $Q / Z$ as the Hopf-invariant analogue, and putting these together the Adams operations from 03.08.02, Bott periodicity from 03.08.07, and the AHSS from 03.13.04 collapse into one arithmetic identity: $Im (J)_{4 k - 1}$ is the cyclic group whose order Bernoulli arithmetic forces. This is exactly the same organising principle throughout K-theoretic index theory, the foundational reason being that the Chern character converts K-theory eigenvalue data into rational cohomology, where Bernoulli denominators live through the Todd genus.

Full proof set [Master]

Proposition (clutching construction is well-defined). Let $α : S^{n} \to O (k)$ be a continuous map. The clutching construction $E_{α} \to S^{n + 1}$ is a well-defined rank- $k$ vector bundle, with class depending only on $[α] \in π_{n} (O (k))$ .

Proof. Write $S^{n + 1} = D_{+}^{n + 1} \cup_{S^{n}} D_{-}^{n + 1}$ . Define $E_{α}$ by trivialising over each closed disk and gluing along the equatorial $S^{n}$ via $α$ : a point $(x, v) \in S^{n} \times R^{k}$ on the $D_{+}^{n + 1}$ side is identified with $(x, α (x) v) \in S^{n} \times R^{k}$ on the $D_{-}^{n + 1}$ side. The local triviality is automatic on each disk, and the gluing on the equatorial sphere is continuous. The total space is a fibre bundle of rank $k$ .

If $α_{0} ≃ α_{1} : S^{n} \to O (k)$ via a homotopy $H : S^{n} \times I \to O (k)$ , then applying the clutching at each time produces a bundle $E_{H} \to S^{n + 1} \times I$ . By the homotopy invariance of vector bundles over CW complexes, $E_{H} ∣_{S^{n + 1} \times {0}} ≅ E_{H} ∣_{S^{n + 1} \times {1}}$ , i.e. $E_{α_{0}} ≅ E_{α_{1}}$ . Hence the class $[E_{α}] \in K (S^{n + 1})$ depends only on $[α]$ . $□$

Proposition (clutching is additive). The clutching construction $π_{n} (O (k)) \to K (S^{n + 1})$ is a group homomorphism with respect to the loop-addition on $π_{n} (O (k))$ and the direct-sum structure on K-theory restricted to the kernel of $K (S^{n + 1}) \to K (pt) = Z$ .

Proof. Addition in $π_{n} (O (k))$ corresponds to pinching $S^{n}$ to $S^{n} \lor S^{n}$ and sending one wedge factor to $α$ , the other to $β$ . The clutching construction is functorial in this pinch: gluing along the wedge produces a bundle $E_{α \lor β}$ that splits as a fibred sum $E_{α} \oplus_{simple} E_{β}$ over a pinch $S^{n + 1} \to S^{n + 1} \lor S^{n + 1}$ . In K-theory after the pinch this reads $[E_{α + β}] = [E_{α}] + [E_{β}]$ modulo simple terms (the simple bundle of rank $k$ subtracted). Restricting to the reduced K-theory $K (S^{n + 1})$ kills the simple-bundle term, giving the homomorphism property. $□$

Theorem ( $J$ -homomorphism well-definedness), full proof. Combine the two propositions and pass to the limit in $k$ . The clutching construction sends $α \in π_{n} (O (k))$ to $[E_{α}] - [ε^{k}] \in K (S^{n + 1})$ . The associated spherical fibration $S (E_{α}) \to S^{n + 1}$ is a sphere bundle, and its class in $π_{n + k} (S^{k})$ is the homotopy class of the resulting map between sphere fibres at the two poles. Stability in $k$ is the suspension agreement from Exercise 5. $□$

Theorem ( $J (X)$ -subgroup). $T (X) \subseteq K (X)$ is a subgroup.

Proof. Two bundles $ξ, η$ with simple-equivalent sphere bundles ( $S (ξ \oplus ε^{a}) ≃_{fhe} S (ε^{r + a})$ and similarly for $η$ with rank- $s$ ) satisfy $$ S((\xi \oplus \eta) \oplus \varepsilon^{a+b}) \cong S(\xi \oplus \varepsilon^a) \times_{X} S(\eta \oplus \varepsilon^b) / \mathrm{join} \simeq S(\varepsilon^{r+a}) \times_X S(\varepsilon^{s+b}) / \mathrm{join} = S(\varepsilon^{r+s+a+b}). $$ Hence $ξ + η \in T (X)$ . Similarly $- ξ \in T (X)$ because reversing direction in the fibre is a fibrewise homotopy equivalence. So $T (X)$ is closed under addition and negation; it is a subgroup. $□$

Theorem (Adams, image of $J$ in $π_{4 k - 1}^{S}$ ), proof outline. The image of the stable $J$ -homomorphism in dimension $4 k - 1$ is cyclic of order in ${m_{k}, 2 m_{k}}$ where $m_{k} = denom (B_{2 k} /4 k)$ in lowest terms.

Proof outline. Adams' four-paper sequence $J (X)$ -I-IV (Topology 2-5, 1963-1966) carries out the argument:

(i) $e$ -invariant construction. For each $f \in π_{2 n - 1}^{S}$ , the K-theory of the mapping cone $C_{f}$ fits in a short exact sequence $0 \to K (S^{2 n}) \to K (C_{f}) \to K (S^{0}) \to 0$ . The Chern character produces a rational class whose reduction mod $Z$ is independent of choices, defining $e_{C} (f) \in Q / Z$ . Adams' $e$ -invariant is a stable homomorphism $π_{2 n - 1}^{S} \to Q / Z$ .

(ii) Upper bound (Adams 1965 $J (X)$ -II). For $f = J (α)$ with $α \in π_{4 k - 1} (O)$ , the Chern character of the clutching bundle on the mapping cone $C_{J (α)} ≃ Th (E_{α})$ is computed via the Hirzebruch-Riemann-Roch formula: it equals the Todd class $Td (E_{α})$ pushed forward. The top-degree coefficient of $Td$ on a stably-framed rank- $2 k$ bundle has the universal expansion involving $B_{2 k} / (2 k)!$ . After mod- $Z$ reduction and absorbing the factorial into the Chern-character denominators, the resulting $e_{C} (J (α))$ has denominator dividing the denominator of $B_{2 k} /4 k$ in lowest terms. Hence $∣ Im (J)_{4 k - 1} ∣ \leq m_{k}$ (up to factor $2$ ambiguity from real-vs-complex).

(iii) Lower bound (Adams 1965 $J (X)$ -III). Adams operations on the K-theory of stunted projective spaces $C P^{N} / C P^{N - r}$ detect a class of order exactly $m_{k}$ as a $J$ -image. The argument constructs a specific element $α_{k} \in π_{4 k - 1} (O)$ whose $J$ -image is hit by an explicit virtual-bundle computation on $C P^{N} / C P^{N - r}$ , and shows by the Adams-operation eigenvalue analysis that the order is at least $m_{k}$ .

(iv) Conditional resolution (Adams 1966 $J (X)$ -IV). Combining (ii) and (iii) yields the precise order assuming the Adams conjecture. The conjecture was proved in (Quillen 1971; Sullivan 1974). Hence $∣ Im (J)_{4 k - 1} ∣ = m_{k}$ or $2 m_{k}$ unconditionally, with the parity factor resolved by the real $e$ -invariant $e_{R}$ . $□$

Theorem (Quillen, Adams conjecture; Topology 10, 1971), proof sketch. For every finite CW complex $X$ , every $ξ \in K_{O} (X)$ , and every $k \geq 2$ , there exists $N$ with $k^{N} (ψ^{k} ξ - ξ) \in T (X)$ .

Proof sketch. Reduce to the universal case $X = B U (n)$ or $B O (n)$ . Localise at a prime $p$ . For $p ∤ k$ , Quillen's étale-homotopy theory of complex varieties identifies $B U (n)_{p}^{\land}$ with the $p$ -completed étale homotopy type of the algebraic stack $B GL_{n}$ over $Spec (\overline{F}_{p})$ . The Galois group $Z = Gal (\overline{F}_{p} / F_{p})$ acts on this étale homotopy type. The Frobenius $Frob^{k}$ corresponds, on the underlying étale K-theory, to the Adams operation $ψ^{k}$ — this is the standard identification of Frobenius and Adams operations through the comparison $K_{0}^{\overset{e}{ˊ} t} (Spec (\overline{F}_{p})) ≅ R (Z)$ . Hence at $p ∤ k$ , the difference $ψ^{k} (ξ) - ξ$ is realised as the difference between $ξ$ and a Galois translate of $ξ$ , which is automatically simple in the spherical-fibration sense after $p$ -completion: Galois translates of an algebraic variety preserve the étale homotopy type.

For primes $p ∣ k$ , the relation is automatic after multiplying by $k^{N}$ for $N$ large (the $p$ -primary part of $K_{O} (X)$ has bounded exponent, and multiplication by $k^{N}$ kills it). Combining the two pieces by the local-global Hasse principle for finitely generated abelian groups, $k^{N} (ψ^{k} ξ - ξ)$ is simple in $J (X)$ for $N$ sufficiently large. $□$

Theorem (Sullivan, Adams conjecture; Ann. Math. 100, 1974), proof sketch. Sullivan's proof replaces étale homotopy by his own profinite-localisation theory. The argument is: the profinite completion of $B O$ at $p$ , after inverting $p$ , admits a self-map $ψ^{k}$ for $p ∤ k$ which is shown to factor through a fibre-homotopy equivalence on the spherical-fibration side by direct construction in the model category of profinite simplicial sets. The Galois action of $Z$ enters through the comparison with algebraic K-theory after $p$ -completion. Both proofs identify the Adams conjecture as a statement about the compatibility of Adams operations on K-theory with the profinite/Galois action on the spherical fibration. $□$

Connections [Master]

Adams operations $ψ^{k}$ 03.08.02. Adams operations are the central computational tool for $J (X)$ . The subgroup $T (X) \subseteq K (X)$ that one quotients by is generated, up to arithmetic multiples, by $ψ^{k} (ξ) - ξ$ for $ξ$ a virtual bundle; the Adams conjecture (Quillen 1971, Sullivan 1974) makes this precise after $k^{N}$ -multiplication. Concretely, $J (X)$ is computed by reducing $K (X)$ modulo all $ψ^{k}$ -invariants, which is a finite arithmetic computation once the structure of $K (X)$ as a $λ$ -ring is known.
Bott periodicity 03.08.07. The orthogonal group $O$ has homotopy groups computed by Bott periodicity: $π_{n} (O) ≅ π_{n + 8} (O)$ with the pattern $Z /2, Z /2, 0, Z, 0, 0, 0, Z$ in dimensions $0$ through $7$ . The $J$ -homomorphism $J : π_{n} (O) \to π_{n}^{S}$ inherits this mod- $8$ periodicity in its source, and Adams' $J (X)$ -IV (Topology 5, 1966) uses this to organise the analysis of $Im (J)$ in each residue class mod $8$ . The image in dimensions $4 k - 1$ is the canonical case (when $π_{n} (O) = Z$ ); the other dimensions contribute $Z /2$ classes detected by Bott-periodic generators.
Atiyah-Hirzebruch spectral sequence 03.13.04. The AHSS converging to $K^{*} (X)$ provides the systematic computational tool for $K (X)$ , hence for $J (X)$ via the Adams-operation coinvariants. Differentials in the AHSS are torsion operations of arithmetic origin (Steenrod squares, Massey products), and the survival of integer classes to the $E_{\infty}$ page corresponds to lifting cohomological data to K-theoretic data. Bernoulli denominators reappear as the universal differentials in the Atiyah-Hirzebruch spectral sequence for $K^{*} (B U)$ , and the comparison between $K$ -theoretic differentials and the Bernoulli-denominator formula for $Im (J)_{4 k - 1}$ is the source of the cohomological-K-theoretic translation of Adams' theorem.
Stable homotopy of spheres and the Hopf invariant. The $J$ -homomorphism produces the most computable subgroups of $π_{n}^{S} (S^{0})$ , exhausting the group in dimensions $1, 3, 7$ (where $π_{n}^{S} = Z /2, Z /24, Z /240$ ), and contributing a definite cyclic summand in every dimension $4 k - 1$ . The classical Hopf-invariant-one elements $η \in π_{1}^{S}$ , $ν \in π_{3}^{S}$ , $σ \in π_{7}^{S}$ correspond, under $J$ , to the Hopf fibrations $S^{3} \to S^{2}, S^{7} \to S^{4}, S^{15} \to S^{8}$ — and the non-existence of further Hopf-invariant-one elements (Adams 1960; Adams-Atiyah 1966 via Adams operations) closes the picture.
Spectrum 03.12.04. The $J$ -homomorphism is the map of spectra $J : k O \to S$ from the connective real K-theory spectrum to the sphere spectrum, an essential structure map in stable homotopy theory. The image of $J$ as a sub-spectrum is the $j$ -spectrum, with computable homotopy groups equal to Bernoulli denominators in dimensions $4 k - 1$ . The $j$ -spectrum sits inside the sphere spectrum as a sub-spectrum encoding the entire Adams-computed image-of- $J$ data.
Whitehead tower 03.12.07. The Whitehead tower of the orthogonal group $O$ consists of successive principal $K (π, n)$ -fibrations killing the homotopy groups $π_{n} (O)$ one by one. The $J$ -homomorphism interacts with the Whitehead tower because each $K (π, n)$ -section corresponds to a refinement of $Im (J)_{n}$ inside $π_{n}^{S}$ . The first non-vanishing step kills $π_{0} (O) = Z /2$ producing $SO$ , the second kills $π_{1} (O) = Z /2$ producing $Spin$ , and at each higher stage the corresponding $J$ -image generator becomes lifted to a refined obstruction class.

Historical & philosophical context [Master]

The $J$ -homomorphism originated in J.H.C. Whitehead's 1942 paper On the homotopy groups of spheres and rotation groups (Annals of Mathematics (2) 43, 634-640) ^{[Whitehead 1942 J-homomorphism]}, in which Whitehead defined a homomorphism from the homotopy groups of the rotation group to the stable homotopy groups of spheres through what he called the "clutching" or "characteristic-class" construction. Whitehead's original definition was unstable — landing in $π_{n + k} (S^{k})$ for fixed $k$ — and the stable version emerges by passing to the limit using Freudenthal suspension. Whitehead recognised the map as a fundamental tool for understanding the homotopy of spheres, but did not have at his disposal the K-theoretic machinery that would later determine its image.

The $J$ -group $J (X)$ was introduced by Michael Atiyah in his 1961 paper Thom complexes (Proceedings of the London Mathematical Society (3) 11, 291-310) ^{[Atiyah Thom complexes]}, in connection with computations of cobordism rings and stable normal bundles. Atiyah defined $J (X)$ as the quotient of $K (X)$ by the equivalence relation of stable fibre-homotopy equivalence of associated sphere bundles, and observed that $J (X)$ is the natural target for any invariant of vector bundles depending only on the spherical fibre-homotopy type. The construction was put in its definitive K-theoretic form by Atiyah in his 1962 Aarhus lectures and the 1967 book K-Theory §3.4.

The systematic study of $J (X)$ as an arithmetic computational problem was carried out by J. F. Adams in his four-paper series On the groups $J (X)$ I-IV in Topology: I (volume 2, 1963, 181-195), II (volume 3, 1965, 137-171), III (volume 3, 1965, 193-222), IV (volume 5, 1966, 21-71). ^{[Adams J(X) I]} ^{[Adams J(X) II]} ^{[Adams J(X) III]} ^{[Adams J(X) IV]} The I paper introduced the Adams conjecture: for any real virtual bundle $ξ$ over a finite CW complex $X$ and any integer $k \geq 2$ , the element $k^{N} (ψ^{k} ξ - ξ)$ lies in $T (X)$ for some $N$ . Papers II and III computed the upper and lower bounds for $Im (J)_{4 k - 1}$ via the $e$ -invariant and the Adams operations on stunted projective spaces, establishing the Bernoulli-denominator formula. Paper IV synthesised the picture into the precise statement: the image of $J$ in dimensions $4 k - 1$ is cyclic of order $m_{k}$ or $2 m_{k}$ , with the factor-of- $2$ ambiguity controlled by the real $e$ -invariant.

The Adams conjecture was proved unconditionally by Daniel Quillen in 1971 in The Adams conjecture (Topology 10, 67-80) ^{[Quillen Adams conjecture]}. Quillen's proof was a breakthrough in technique: it introduced the use of étale homotopy theory of algebraic varieties (developed earlier by Artin-Mazur in 1969) to prove a question in classical algebraic topology. The identification of Adams operations on topological K-theory with the Galois action of $Z$ on étale K-theory at primes $p ∤ k$ was the key bridge. A second proof was given by Dennis Sullivan in 1974 in Genetics of homotopy theory and the Adams conjecture (Annals of Mathematics (2) 100, 1-79) ^{[Sullivan Adams conjecture]}. Sullivan's profinite-localisation approach replaced étale homotopy with a direct construction in the model category of pro-finite simplicial sets, and is conceptually closer to the modern formulation of arithmetic fracture squares in homotopy theory.

The philosophical significance of the $J (X)$ story is twofold. First, it identifies the precise quantitative content of the difference between two coarser invariants of vector bundles (K-theory and fibre-homotopy theory of sphere bundles), with the difference being entirely arithmetic and entirely computable in terms of Bernoulli numbers and Adams operations. Second, the proof of the Adams conjecture by Quillen and Sullivan was a foundational moment in the relationship between number theory, algebraic geometry, and algebraic topology: it showed that the same Galois-action structure governing rational points on algebraic varieties also governs spherical-fibration equivalence of vector bundles, and that classical-topology questions about stable homotopy of spheres have answers buried in the arithmetic of Bernoulli numbers. The von Staudt-Clausen theorem (1840) on Bernoulli denominators thereby acquires a topological meaning: the primes $p$ with $p - 1 ∣ 2 k$ are exactly the primes contributing to the order of $Im (J)_{4 k - 1}$ .

Bibliography [Master]

@article{Whitehead1942,
  author  = {Whitehead, J. H. C.},
  title   = {On the homotopy groups of spheres and rotation groups},
  journal = {Annals of Mathematics. Second Series},
  volume  = {43},
  year    = {1942},
  pages   = {634--640}
}

@article{Atiyah1961ThomComplexes,
  author  = {Atiyah, M. F.},
  title   = {Thom complexes},
  journal = {Proceedings of the London Mathematical Society. Third Series},
  volume  = {11},
  year    = {1961},
  pages   = {291--310}
}

@article{AdamsJX1,
  author  = {Adams, J. F.},
  title   = {On the groups {$J(X)$. I}},
  journal = {Topology},
  volume  = {2},
  year    = {1963},
  pages   = {181--195}
}

@article{AdamsJX2,
  author  = {Adams, J. F.},
  title   = {On the groups {$J(X)$. II}},
  journal = {Topology},
  volume  = {3},
  year    = {1965},
  pages   = {137--171}
}

@article{AdamsJX3,
  author  = {Adams, J. F.},
  title   = {On the groups {$J(X)$. III}},
  journal = {Topology},
  volume  = {3},
  year    = {1965},
  pages   = {193--222}
}

@article{AdamsJX4,
  author  = {Adams, J. F.},
  title   = {On the groups {$J(X)$. IV}},
  journal = {Topology},
  volume  = {5},
  year    = {1966},
  pages   = {21--71}
}

@article{Quillen1971Adams,
  author  = {Quillen, Daniel},
  title   = {The {A}dams conjecture},
  journal = {Topology},
  volume  = {10},
  year    = {1971},
  pages   = {67--80}
}

@article{Sullivan1974Genetics,
  author  = {Sullivan, Dennis},
  title   = {Genetics of homotopy theory and the {A}dams conjecture},
  journal = {Annals of Mathematics. Second Series},
  volume  = {100},
  year    = {1974},
  pages   = {1--79}
}

@article{AtiyahTall1969,
  author  = {Atiyah, M. F. and Tall, D. O.},
  title   = {Group representations, $\lambda$-rings and the {$J$}-homomorphism},
  journal = {Topology},
  volume  = {8},
  year    = {1969},
  pages   = {253--297}
}

@book{AtiyahKTheory,
  author    = {Atiyah, M. F.},
  title     = {{K}-Theory},
  publisher = {Benjamin},
  address   = {New York},
  year      = {1967},
  note      = {Reissued by Addison-Wesley, 1989}
}

@book{Karoubi1978,
  author    = {Karoubi, Max},
  title     = {{K}-Theory: {A}n Introduction},
  series    = {Grundlehren der mathematischen Wissenschaften},
  volume    = {226},
  publisher = {Springer-Verlag},
  year      = {1978}
}

@book{LawsonMichelsohn1989,
  author    = {Lawson, H. Blaine and Michelsohn, Marie-Louise},
  title     = {Spin Geometry},
  series    = {Princeton Mathematical Series},
  volume    = {38},
  publisher = {Princeton University Press},
  year      = {1989}
}

@book{MilnorStasheff1974,
  author    = {Milnor, John W. and Stasheff, James D.},
  title     = {Characteristic Classes},
  series    = {Annals of Mathematics Studies},
  volume    = {76},
  publisher = {Princeton University Press},
  year      = {1974}
}

@article{ArtinMazur1969,
  author    = {Artin, M. and Mazur, B.},
  title     = {Etale homotopy},
  series    = {Lecture Notes in Mathematics},
  volume    = {100},
  publisher = {Springer-Verlag},
  year      = {1969}
}

Prerequisites

03.08.01
03.08.02
03.08.07
03.12.04
03.12.07
03.13.04

Tier anchors

beginner: fiberwise stable equivalence of vector bundles as a coarsening of K-theory — Strogatz *Infinite Powers* pacing for the abstract account
intermediate: Atiyah *Thom complexes* (Proc. Lond. Math. Soc. 11, 1961) §3; Atiyah *K-Theory* §3.4; Karoubi *K-Theory: An Introduction* Ch. IV §7
master: Adams *On the groups $J(X)$ I-IV* (Topology 2-5, 1963-1966); Quillen *The Adams conjecture* (Topology 10, 1971); Sullivan *Genetics of homotopy theory and the Adams conjecture* (Ann. Math. 100, 1974); Atiyah *Thom complexes* (Proc. Lond. Math. Soc. 11, 1961); Whitehead *On the homotopy groups of spheres and rotation groups* (Ann. Math. 43, 1942); Lawson-Michelsohn *Spin Geometry* §IV.1

References

fast-track
images/Atiyah-K-Theory-683x1024__52e3382e9f.jpg · Atiyah K-Theory §3.4, $J(X)$ and Adams' conjecture as motivation
TODO_REF
Adams — On the groups J(X) I · Topology 2 (1963) 181-195, originator paper defining $J(X) = \widetilde K(X) / T(X)$ and stating the Adams conjecture
TODO_REF
Adams — On the groups J(X) II · Topology 3 (1965) 137-171, $e$-invariant and the upper bound for $J(\pi_{4k-1} O)$ via Bernoulli denominators
TODO_REF
Adams — On the groups J(X) III · Topology 3 (1965) 193-222, lower bound: image of the $J$-homomorphism is detected by Adams operations on $K$-theory of stunted projective spaces
TODO_REF
Adams — On the groups J(X) IV · Topology 5 (1966) 21-71, full statement of Adams conjecture; completion of $\mathrm{Im} J$ in dimensions $4k-1$ up to a factor of $2$
TODO_REF
Quillen — The Adams conjecture · Topology 10 (1971) 67-80, étale-homotopy proof of the Adams conjecture: for $\xi$ a real vector bundle and $k$ coprime to a prime $p$, $\psi^k \xi - \xi$ becomes spherically trivial after $p$-localisation
TODO_REF
Sullivan — Genetics of homotopy theory and the Adams conjecture · Ann. of Math. (2) 100 (1974) 1-79, profinite-localisation proof of the Adams conjecture and the broader localisation framework
TODO_REF
Atiyah — Thom complexes · Proc. London Math. Soc. (3) 11 (1961) 291-310, the originator paper introducing $J$-equivalence of vector bundles in terms of fibre-homotopy equivalence of associated sphere bundles
TODO_REF
Whitehead — On the homotopy groups of spheres and rotation groups · Ann. of Math. (2) 43 (1942) 634-640, originator paper defining the $J$-homomorphism $\pi_n(\mathrm{SO}) \to \pi_{n}^{S}(S^0)$ via the clutching construction
TODO_REF
Atiyah-Tall — Group representations, lambda-rings and the J-homomorphism · Topology 8 (1969) 253-297, $J$-homomorphism for compact Lie groups via $\lambda$-ring structure on $R(G)$
TODO_REF
Lawson-Michelsohn — Spin Geometry · §IV.1, $J$-homomorphism and image of $J$ in the context of spin bordism and the Atiyah-Singer index theorem
TODO_REF
Karoubi — K-Theory: An Introduction · Springer Grundlehren 226 (1978) Ch. IV §7, $J(X)$ and the topological consequences of the Adams conjecture
TODO_REF
Milnor-Stasheff — Characteristic Classes · Princeton (1974), appendix on the $J$-homomorphism and Bernoulli-number denominators

Reviewer

TBD

Estimated time

beginner: 18m
intermediate: 50m
master: 95m