In this manuscript, we introduce the Adams operators in K-theory, and one of its

applications in solving the Hopf invariant one problem.

1 The splitting principle

In this article, fix a compact Hausdorff topological space X, denote by K(X) the K-
theory ring of X. We present the splitting principle for K(X), with a notable remark
that the proof in the classical book of Atiyah does not extend to the equivariant case.

Definition 1 Let π : E → X be a vector bundle of constant rank d. We inductively

define vector bundles πj : Ej → Xj and projective bundles ρj : Xj → Xj−1 : Starting
with π0 = π, define

ρj : Xj = P (Ej−1 ) → Xj−1 , Ej = ρ∗j (Ej−1 )/Tj ,

where Tj is the tautological bundle over Xj = P (Ej−1 ). The composition ρ = ρ1 ◦ · · · ◦

ρd : Xd → X0 is called the flag bundle F (E) → X.

Theorem 2 The pullback ρ∗ : K ∗ (X) → K ∗ (F (E)) is injective, and ρ∗ (E) is iso-

morphic to a sum of d line bundles.

Proof: The first property is true using Bott periodicity, with the note that ρ∗ is a
composition of embeddings from the K-theory ring of a space to the K-theory ring
of its projective space (this argument does not generalize to the equivariant case).
For the second fact, pulling back all the tautological bundles to F (E), we obtain
line bundles Lj = (ρj+1 ◦ · · · ◦ ρd )(Tj ) → F (E). Using the fact that X is compact,
one can actually realize each Ej as a subbundle complementary to Tj , which makes
Tj+1 a subbundle of the pullback of of E to Xj+1 = P (Ej ), and thus, Lj becomes
a subbundle of ρ∗ (E). It is clear from the construction that we get a direct sum
decomposition E = L1 ⊕ · · · ⊕ Ld . □

2 Definitions
For each vector bundle E over X, the bundle rank(E) is the bundle such that in
each connected component U of X, the bundle rank(E)|U ∼= U × CdU , where dU =
dim(E|U ). It induces a map

rank : K(X) → K(X),

which is obviously an idempotent ring endomorphism.

1
Now we define ψ 0 (x) = rank(x). In the ring K(X)[[t]], we define
X d
ψt (x) = ti ψ i (x) = ψ 0 (x) − t log(λ−t (x)).
i=0
dt

The idea of taking the log is that one is able to transfer back multiplication to addition.

Proposition 3 For any x, y ∈ K(X), ψ k (x + y) = ψ k (x) + ψ k (y). If x is a line

bundle, then ψ k (x) = xk . These properties uniquely determine ψ k .

Before proceeding with a proof, it is worth noting that ψ k is functorial, meaning that
it commutes with pullbacks.
Proof: By definition, we see that ψt (x + y) = ψt (x) + ψt (y); if x is a line bundle,
d −x
then ψ−t (x) = 1 − tx, which implies that log(1 − tx) = = −x − t2 x2 − · · ·
dt 1 − tx
The last part follows from the splitting principle. □

Proposition 4 For x, y ∈ K(X), we have

ψ k (xy) = ψ k (x)ψ k (y).

ψ k (ψ l (x)) = ψ kl (x).
If p is prime, then ψ p (x)−xp is zero mod p. Finally, if u ∈ K̃(S2n ), then ψ k (u) = k n u
for all k.

Proof: The first two assertions follow immediately from the splitting principle.
Also, from the splitting principle, ψ p (x) = xp + pf (λ1 (x), . . . , λp (x)), where f is some
polynomial with integer coefficients. Finally, if h is the generator of K̃(S2 ), then we
know that h = H −1, where H is a line bundle that satisfies the relation (H −1)2 = 0.
Hence for k ≥ 2, we have

ψ k (h) = H k −1 = H k−2 (H−1)2 +2H k−1 −H k−2 −1 = (2(k−1)−k+2)(H−1) = k(H−1).

Since the generator of K̃(S2n ) = K̃(S2 ∧ S2 ∧ · · · ∧ S2 ) is h ⊗ h · · · ⊗ h, we obtain the

desired formula. □

3 The main theorem

Theorem 5 The following statements are true only for n = 1, 2, 4, 8:

1. Rn is a division algebra over R (the reals, complex numbers, quaternions, octo-

nions).

2
 2. Sn−1 is parallelizable, i.e., the tangent bundle to Sn−1 is trivial.

The first step in the proof of the theorem is to reduce it to showing when the sphere
Sn−1 is an H-space, which means there is a continuous multiplication map Sn−1 ×Sn−1
having a two-sided identity element e.

Lemma 6 If Rn is a division algebra or Sn−1 is parallelizable, then Sn−1 is an H-

space.

Proof: If Rn is a division algebra with a two-sided identity, an H-space structure

on Sn−1 is given by
x·y
(x, y) 7→ ,
|x · y|
which is continuous because the multiplication is bilinear, which is automatically
continuous. Now suppose Sn−1 is parallelizable, from the Gram-Schmidt process, one
may assume that TSn−1 = v1 ⊕⊥ v2 ⊕⊥ · · · ⊕⊥ vn , where v1 , . . . , vn are tangent vector
fields on Sn−1 . By changing the sign, we can also assume that vj (e1 ) = ej+1 , ∀j.
Let αx ∈ SO(n) send the standard basis to x, v1 (x), . . . , vn−1 (x). Then the map
(x, y) 7→ αx (y) defines an H-space structure on Sn−1 with identity element the vector
e1 , since αe1 is the identity map. □

By investigating the induced ring map µ∗ : K(S2k ) → K(S2k × S2k ) induced from an
H-space multiplication, we see that its existence is contradictory.
Next, for a map g : Sn−1 × Sn−1 → Sn−1 , regard S2n−1 = ∂(Dn × Dn ) = ∂Dn × Dn ∪
Dn × ∂Dn , and S n as the union of two disks Dn and D+ n
with boundaries identified.
n n n
Then gb is defined on ∂D × D by gb(x, y) = |y|g(x, y/|y|) ∈ D+ , and on Dn × ∂Dn
n
by gb(x, y) = |x|g(x/|x|, y) ∈ D . Note that gb is well-defined and continuous.
Now we specialize to the case that n is even, so we replace n by 2n. For a map
f : S4n−1 → S2n , let Cf be S2n with a cell e4n attached by f . Them, the quotient
Cf /S2n is obviously S4n , and since K̃ 1 (S4n ) = K̃ 1 (S2n ) = 0, the exact sequence of the
pair (Cf , S2n ) becomes a short exact sequence

0 → K̃(S4n ) → K̃(Cf ) → K̃(S2n ) → 0.

Let α be the image of the generator (H − 1) ∗ · · · ∗ (H − 1) of K̃(S4n ), and assume

that β maps to the generator (H − 1) ∗ · · · ∗ (H − 1) of K̃(S2n ). Then, the element
β 2 maps to 0 in K̃(S2n ), and thus β 2 = hα for some h, called the Hopf invariant of
f . Suppose one chooses β ′ = β + kα, then β ′2 = β 2 + 2kαβ, hence we only need to
show that αβ = 0. Since α maps to 0 in K̃(S2n ), and the map above is a ring map,
we see that αβ = mα, thus mαβ = hα2 = 0, which implies αβ = 0, as αβ lies in the
image of K̃(S4n ) in K̃(Cf ), which is an infinite cyclic group.

3
Lemma 7 If g : S2n−1 × S2n−1 → S2n−1 is an H-space multiplication, then the asso-
ciated map gb has Hopf invariant ±1.

Proof: Let e be the identity element of g, and f = gb. Let Φ denote the characteristic
map of the 4n-cell of Cf :

Φ : (D2n × D2n , ∂(D2n × D2n )) → (Cf , S2n ).

Consider β ⊗ β ∈ K̃(Cf ) ⊗ K̃(Cf ), then by mapping β ⊗ β through the map Φ∗ ⊗ Φ∗ ,

we see that β 2 is a generator of K̃(Cf ). The gist of the proof lies in the fact that
from the definition of f , the map Φ restricts to a homeomorphism from D2n × {e}
onto D2n and from {e} × D2n to D+ 2n
. Thus β 2 = ±α, as desired. □

The knowledge of the action of the Adams operators on a generator of K̃(S2n ) is

actually sufficient to show that

Theorem 8 If there is a map f : S4n−1 → S2n of Hopf invariant 1, then n ∈ {1, 2, 4}.

Proof: We use the elements α, β as in the above argument. Now, from Proposition
4, we have ψ k (α) = k 2n α, and ψ k (β) = k n β + µk α, because the image of β is a
generator of K̃(S2n ). Therefore,

ψ k ψ l (β) = (kl)n β + (k 2n µl + ln µk )α.

By Proposition 4, it follows that (k 2n − k n )µl = (l2n − ln )µk for all k, l. We also have
ψ 2 (β) = β 2 = hα mod 2, if h is odd, then ψ 2 (β) = 2n β + µ2 α = hα mod 2 implies
that µ2 is odd. We also have (22n − 2n )µ3 = (32n − 3n )µ2 , hence 2n divides 3n − 1. It
is a standard technique (lifting the exponent) to show that 2n divides 3n − 1 if and
only if n = 1, 2, 4, which concludes the proof. □

If you want to help by writing manuscripts, or if you have any questions or comments,
please contact us via the email address lxhoanghsgs@[Link], I really appreciate
your contribution!