EXTENDING ADAMS’ THEOREM FROM SINGLY GENERATED

TO PERIODIC COHOMOLOGY

JOHN R. HARPER AND LEE KENNARD

arXiv:2412.16340v1 [math.AT] 20 Dec 2024

Abstract. In 1960, J.F. Adams introduced secondary cohomology operations that

are defined on cohomology elements on which sufficiently many Steenrod algebra
elements vanish. This led to his theorem on singly generated cohomology rings, which
in turn led to his celebrated resolution of the Hopf invariant one problem. Here we
advertise a conjecture that would extend Adams’ result and prove it in a special case.

1. Introduction

J.F. Adams’ theorem on singly generated cohomology rings was the key step in his
resolution of the Hopf invariant one problem. The theorem is as follows (see [Ada60]):
Theorem 1.1 (Adams, 1960). Let X be a topological space. If H ∗ (X; Z2 ) is isomorphic
to the polynomial ring Z2 [x], then x has degree k ∈ {1, 2, 4}.
Adams’ theorem also holds for truncated polynomial algebras on one generator whose
square is non-trivial. It allows for k = 8 in addition, so long as the third power of the
generator vanishes. All values of k permitted by Adams’ theorem were long known to
be realized. Examples include the classifying spaces BO(1) = RP∞ , BU(1) = CP∞ ,
and BSp(1) = HP∞ , their truncated analogues, and the Cayley plane.
The conjectured extension has the same conclusion but a weaker assumption. For
simplicity, we restrict our attention to an extension in the non-truncated case. We
consider k–periodic cohomology rings:
Definition 1.2. For a connected topological space X, we say that a non-zero element
x ∈ H k (X; Z2 ) induces periodicity if the map H i (X; Z2 ) → H i+k (X; Z2 ) induced by
multiplication by x is an isomorphism for all i ≥ 0. When such an element exists, we
say that H ∗ (X; Z2 ) is k-periodic.
Example 1.3 (Acyclic spaces). If we were to only require surjectivity in degree zero
and allow x = 0 in the definition, we would be including mod 2 acyclic spaces (or mod
2 homology spheres in the manifold setting, properly formulated). For our purposes,
these cases are trivial, so we require here that x 6= 0 and hence generates H k (X; Z2 ).
Example 1.4. In addition to singly generated cohomology rings, the spaces S1 × CP∞ ,
N 2 × HP∞ , and N 3 × HP∞ have 4-periodic Z2 -cohomology rings, where N is any
connected manifold of dimension 2 or 3 in the last two cases, respectively.
Observe that, if H ∗ (X; Z2 ) ∼
= Z2 [x], then x is nonzero and induces periodicity. More-
over, x has minimal degree among all such elements. The following conjecture therefore
extends Adams’ theorem (see [Ken13, Conjecture 6.3]):
1
2 JOHN R. HARPER AND LEE KENNARD

Conjecture 1.5 (Z2 Periodicity Conjecture). Let X be a topological space such that
H ∗ (X; Z2 ) is k-periodic for some k ≥ 1. If k is the minimum period, then k ∈ {1, 2, 4}.
Evidence for this conjecture includes the proof that k is a power of two (see [Ken13,
Proposition 1.3]). This result extended J. Adem’s contribution to the Hopf invariant
one problem to the case of periodic cohomology as in this conjecture (see [Ade52]).
Additional evidence exists for the odd prime analogue of the conjecture in the manifold
case when k = 2p (see Section 6). In this article, we provide additional evidence:
Theorem A. The periodicity conjecture holds for spaces X satisfying the property that
k
H 2 (X; Z2 ) = 0, where k is the minimum period.
One recovers Adams’ theorem from Theorem A by restricting to the case where
i
H (X; Z2 ) = 0 for all 0 < i < k. Our proof actually works under the more general
k
assumption that the Steenrod square Sq 2 evaluates to zero on the group H k (X; Z2 ) ∼=
Z2 . This condition is implied by the one in Theorem A because the image lands in the
k k
group H k+ 2 (X; Z2 ), which by periodicity is isomorphic to H 2 (X; Z2 ).
The proof of Theorem A follows Adams’ strategy and, in particular, uses secondary
cohomology operations. Unlike the singly generated case, however, even proving that
these operations are defined requires proving certain vanishing results for the primary
operations, i.e., Steenrod squares. The assumption in Theorem A gives us one such
vanishing statement, and the main technical work in this article is to derive additional
vanishing results for the action of the Steenrod algebra. Once this is done, we may
apply Adams’ basic strategy, together with results in [Ken13], to finish the proof.
The odd prime analogue of Conjecture 1.5 is the following:
Conjecture 1.6 (Zp Periodicity Conjecture). Let X be a topological space and p be an
odd prime. If a non-zero element x ∈ H k (M; Zp ) induces isomorphisms H i (X; Zp ) →
H i+k (X; Zp ) by multiplication for all i ≥ 0, and if x has minimal degree among all such
elements, then k = 2λ for some divisor λ of p − 1.
By [Ken13, Proposition 2.1], it is known that the minimum period is of the form
k = 2λpa where λ | p − 1 and a ≥ 0. Here we prove a = 0 under an additional vanishing
assumption similar in analogy with Theorem A.
Theorem B. In the setting of Conjecture 1.6, if k = 2λpa for some divisor λ of p − 1
a−1
and some a ≥ 1, then H 2(p−1)p (X; Zp ) 6= 0.
In the special case where H i (X; Zp ) = 0 for all 0 < i < k, where k is the minimum
period, one recovers the classical work of Liulevicius [Liu62] and Shimada-Yamanoshita
[SY61] who proved the odd prime analogue of Adams’ theorem on singly generated
cohomology rings.
More generally, the proof of Theorem B shows that any counterexample X to Con-
a a−1
jecture 1.6 satisfies k = 2λpa for some λ | p − 1 and a ≥ 1 and P (l−1)p +p 6= 0 on the
group H k (X; Zp ) ∼= Zp for some 1 ≤ l ≤ λ. This assumption is more general since the
a a a−1
images of these Steenrod powers land in the group H 2λp +2(p−1)(l−1)p +2(p−1)p (X; Zp ),
a−1
which by periodicity is isomorphic to H 2(p−1)p (X; Zp ).
 EXTENDING ADAMS’ THEOREM 3

The condition of periodic cohomology arises in Riemannian geometry. In a foun-

dational paper, Wilking [Wil03] proved that totally geodesic inclusions of positively
curved Riemannian manifolds are highly connected. In special situations, this yields
n-connected inclusions of closed, orientable manifolds N n → M n+k where k ≤ n. Com-
bined with Poincaré duality, this implies the existence of an element x ∈ H k (M; Z)
such that multiplication by x induces an isomorphism H i (M; Z) → H i+k (M; Z) for all
0 < i < n. Moreover the map from H 0 (M; Z) is surjective and the map into H n+k (M; Z)
is injective. Passing to coefficients in Z2 , one finds that M is a Z2 -homology sphere or
has k-periodic cohomology in a sense similar to Definition 1.2.
In [Ken13, Theorem C], the second author used Steenrod’s cohomology operations to
refine the Zp periodicity statements for all primes p and proved, assuming further that
2k ≤ n, that M is a rational homology sphere or has four-periodic rational cohomology.
Together with vanishing results for the first and third Betti numbers in this setting,
one obtains vanishing results for all odd Betti numbers. These are key steps in the
verification by the second author, Wiemeler, and Wilking of Hopf’s Euler Characteristic
Positivity Conjecture for Riemannian metrics with isometry group of rank at least
five (see [KWW1]). The conjecture goes back to the 1930s, and this was the first
evidence involving symmetry groups whose rank is not required to grow to infinity in
the manifold dimension.
In [Nie], Nienhaus improved [Ken13, Theorem C] in the special case where the peri-
odicity is induced by a n-connected inclusion N n ⊆ M n+k of closed, oriented manifolds.
His technique uses characteristic classes and in particular allows one to drop the as-
sumption of simply connected. In the further special case where the normal bundle to
N admits a complex structure, which can often be arranged in geometric applications,
Nienhaus also relaxes the codimension assumption from 2k ≤ n in [Ken13, Theorem C]
to k ≤ n. As an application, Nienhaus proved one can relax the torus rank assumption
in [KWW1] from five to four. The significance of this advance is highlighted by the fact
that the proof technique in both papers fails for rank two. Therefore Nienhaus’ proof
is in a sense either optimal already or only off by one.
Acknowledgements. This work was supported by NSF Grant DMS-2402129, the
Simons Foundation’s TSM program, and NSF Grant DMS-1928930 while the second
author was in residence at SLMath in Berkeley, California, in Fall 2024.

2. Classical proof in the singly generated case

Let X be a space with H ∗ (X; Z2 ) ∼ = Z2 [x]/(xq+1 ) for some q ≥ 2 or q = ∞. Adem

a
showed that, if k 6= 2 , then there is a universal decomposition
X
Sqk = Sqi ◦ai
0<i<k

for some Steenrod algebra elements ai . Evaluating on x, we conclude

X
x2 = Sqi (wi )
0<i<k
4 JOHN R. HARPER AND LEE KENNARD

for some cohomology elements wi ∈ H 2k−i (X; Z2 ). All of these groups are zero for
0 < i < k, so x2 = 0, a contradiction to q ≥ 2.
i
Now assume k = 2a ≥ 16. Note that Sq2 (x) = 0 for all 0 < 2i < k. Adams showed
that there is a secondary decomposition
a i
X
Sq2 = Sq2 ◦Φi
0<2i <2a

that holds on the subspace

a i
\
H 2 (X; Z2 ) ∩ ker(Sq2 )
0<2i <2a

and lands in the quotient

a+1 i
M
H2 (X; Z2 )/ im(Sq2 ).
0<2i <2a

Evaluating on x, we get a conditional relation of the form

i
X
x2 = Sq2 (wi ),
0<2i <k
a+1 i
where wi ∈ H 2 −2 (X; Z2 ). Once again, since these groups are zero, the wi are zero,
and we have a contradiction.
For the case k = 8, it suffices to derive a contradiction if x3 6= 0. For degree reasons,
Sq1 (x) = Sq2 (x) = Sq4 (x) = 0. Following Gonçalves [Gon77], we find that there are
secondary cohomology operations Φ′0,3 and Φi satsifying the following two properties:
i i
(1) Φ′0,3 (Sq8 (u)) = Sq16 (u) + 0<2i <8 Sq2 (Φi (u)) for all u ∈ 0<2i <8 ker(Sq2 ) that
P T
i
holds modulo 0<2i ≤8 im(Sq2 ).
L
i
(2) Φ′0,3 (u2 ) = u3 modulo 0<2i ≤8 im(Sq2 ) for u ∈ H 8 (X; Z2 ) ∩ ker(Sq1 , Sq2 , Sq4 ).
L

Taking u = x in the first and second equations, we conclude that

i
X
x3 = Sq2 (wi )
0<2i ≤8
i
for some wi ∈ H 24−2 (X; Z2 ) (see also [LW90, Corollary 1.3] for a refined decomposi-
tion). Most terms on the right-hand side vanish as in the previous case, so we have
x3 = Sq8 (w) for some w ∈ H 16 (X; Z2 ). This group is generated by x2 , so we again have

2
a contradiction since Sq8 (x2 ) = Sq4 (x) = 0.

3. Results on the general case

Throughout this section, let X be a connected topological space with k-periodic

Z2 -cohomology. We abbreviate H i (X; Z2 ) by H i . We make the following assumptions:
(1) x ∈ H k is non-zero and induces periodicity. In particular,
= Hk ∼
(a) H 0 ∼ = H 2k ∼
= . . ., and these groups are generated by powers of x.
i
(b) H is arbitrary for 0 < i < k.
 EXTENDING ADAMS’ THEOREM 5

(c) All other groups are determined by the isomorphisms H i → H i+k given by
multiplication by x. In particular, any y ∈ H i with i ≥ k factors as xy ′ for
some y ′ ∈ H i−k .
(2) The degree k is minimal.
Detailed proofs of the first three lemmas below can be found in [Ken13].
Lemma 3.1. By minimality, neither x nor x2 factors as yz with 0 < deg(y) < k.
Proof sketch. The definition of periodicity implies that, if x or x2 equals yz, then y too
induces periodicity. (Multiplication by x equals multiplication by y and then by z.) 
Lemma 3.2. There is no y with deg(y) < k such that x = Sqi (y). Similarly, there is
no y with deg(y) 6∈ {k, 2k} such that x2 = Sqi (y).
Proof sketch. If there is such a decomposition, choose one with maximal i. Square both
sides, and apply the Cartan formula and periodicity multiple times to conclude that x
factors. By the previous lemma, this is a contradiction.
To prove the second claim, note that x2 6= 0 by periodicity. Hence deg(y) > k and
y 2 = xy ′ by periodicity with 0 < deg(y ′) < k. By using the Cartan formula and
periodicity, we then factor out and cancel an x. This reduces the claim to the first part
of the lemma. 
Lemma 3.3. The degree k is a power of two.
Proof sketch. If k is not a power of two, then Sqk decomposes by the Adem relations.
Evaluate such a relation on x. Use the fact that x2 generates H 2k and Lemma 3.2 to
derive a contradiction. 
Guided by Adams [Ada60], we have a two-step strategy to prove the main estimate
k = 2a ≤ 8:
• (Step 1) Prove that Sqi (x) = 0 for all 0 < i < k.
• (Step 2) Assuming k ≥ 16, conclude the existence of a conditional relation
Sqk (x) = 0<i<k Sqi (wi ) coming from a decomposition of Sqk (x) = x2 in terms
P
of secondary cohomology operations, and then conclude a contradiction.
As we show below, Step 2 can be carried out with no problem. The issue is to prove
that Sqi (x) = 0 for all 0 < i < k in Step 1. As it turns out, we can prove this for
all i under the assumption that it holds for one particular i, namely i = k2 . The main
technical lemma that implies our theorem is the following:
Lemma 3.4. If Sqk/2 (x) = 0, then k ∈ {1, 2, 4, 8}. Moreover, k = 8 only if x3 = 0.
Proof. Assume k = 2a ≥ 16 and Sqk/2 (x) = 0. First, we induct over i to prove that
Sqi (x) = 0 for k2 < i < k, then we induct over i again to prove that Sqi (x) = 0 for
0 < i < k2 (see Section 4 for details.) Given these calculations, there is a conditional
relation of the form X
x2 = Sqk (x) = Sqi (yi )
0<i<k
for some cohomology elements yi (see [Har02]). Since x2 generates H 2k by periodicity,
we conclude x2 = Sqi (yi ) for some i, a contradiction to Lemma 3.2.
6 JOHN R. HARPER AND LEE KENNARD

Now assume k = 8 and that x3 6= 0. The assumption of the lemma states that
Sq4 (x) = 0, and it follows as in the previous case that Sqi (x) = 0 for 4 < i < 8 and
then for 0 < i < 4. Hence there is a decomposition
i
X
x3 = Sq2 (wi ).
0<2i ≤8

If some term is non-trivial with 0 < 2i < 8, then we have a contradiction to the
minimality of 8 = deg(x) by Lemma 3.2. The 2i = 8 term is also zero since H 16 (X; Z2 )
is generated by x2 , which implies that Sq8 (x2 ) = x2 x24 where Sq4 (x) = xx4 . But again
x24 = 0 by minimality since x does not factor. 
k
To complete the proof of the conjecture, it would suffice to prove that Sq 2 (x) = 0.
On the other hand, perhaps the conjecture is false, and we take a moment to discuss
what a counterexample might look like. Suppose X is a space that has, for example, 16-
periodic but not 8-periodic cohomology. Periodicity implies the existence of generators
1, x, and x2 in degrees 0, 16, and 32, respectively. In addition, Lemma 3.4 implies that
Sq8 (x) 6= 0, since otherwise the period of 16 is not minimal. Applying periodicity once
more, we find that Sq8 (x) = xy8 6= 0 for some element y8 in degree 8. Moreover, there
must be additional non-zero groups H i (X; Z2 ) with i 6≡ 0 mod 8, because otherwise
one has xy8 ∈ ker(Sqi ) for i ∈ {1, 2, 4, 8} and hence a conditional relation of the form
i
X
Sq16 (xy8 ) = Sq2 (x2 z8−2i ),
0≤i≤3

which implies that x y8 = Sq (x ) = Sq4 (x)2 = 0, a contradiction to periodicity. This

2 8 2

shows that a counterexample to Conjecture 1.5 with k = 16 would need to have non-
trivial cohomology in some degrees i 6≡ 0 mod 8. We further explore the structure of
potential counterexamples in Section 7.

4. Calculations

Fix x ∈ H k (M; Z2 ) as in Lemma 3.4. We first show by induction that Sqi (x) = 0 for
k
2
≤ i < k. Note that Sqk/2 (x) = 0 by assumption. Write i = k2 + δ where 0 < δ < k2 .
Since k is a power of two, there is an Adem relation of the form
X
Sqδ Sqk/2 = Sqi + cj Sqi−j Sqj .
0<j≤δ/2

Evaluating on x, we see that Sqi (x) = 0 if Sqi−j (Sqj (x)) = 0 for all 0 < j ≤ 2δ . To
prove this, we use periodicity to write Sqj (x) = xxj for some xj ∈ H j (M; Z2 ) and then
the Cartan formula to obtain
X
Sqi−j (Sqj (x)) = Sqi−j−h (x) Sqh (xj ).
0≤h≤j
k
Observe that k2 ≤ i − j − h < i, so the assumption Sq 2 (x) = 0 and the induction
hypothesis imply that every term in this sum is zero. This concludes the proof.
 EXTENDING ADAMS’ THEOREM 7

Second, we show by induction that Sqi (x) = 0 for 0 < i < k2 . We may assume that
Sqj (x) = 0 for all 0 < j < i. Since 2i < 2(k − i), there is an Adem relation of the form
X
Sq2i Sqk−i = c0 Sqk+i + cj Sqk+i−j Sqj + Sqk Sqi .
0<j<i

Evaluating on x, the left-hand side vanishes by the first step of the proof. The first
term on the right-hand side vanishes because k + i is larger than the degree of x, and
the terms in the middle vanish for all 0 < j < i by the induction hypothesis. Therefore
we have Sqk (Sqi (x)) = 0. On the other hand, we can write Sqi (x) = xxi by periodicity
and apply the Cartan formula to obtain
X
0 = Sqk (Sqi (x)) = x2 xi + Sqk−j (x) Sqj (xi ).
0<j≤i

Since k2 < k − j < k, the first step of the proof implies that 0 = x2 xi . By periodicity,
we conclude 0 = xxi = Sqi (x).

5. Odd primes

Fix an odd prime p, and let X be a topological space with periodic Zp -cohomology.
Specifically, we assume k ≥ 1 is minimal such that there exists a non-zero x ∈
H k (X; Zp ) such that the maps H i (X; Zp ) → H i+k (X; Zp ) induced by multiplication by
x are isomorphisms for all i ≥ 0.
By the Zp Periodicity Theorem [Ken13, Proposition 2.1], we have
k = 2λpa
for some divisor λ of p − 1 and some a ≥ 0. The Zp Periodicity Conjecture claims that
a = 0. To prove Theorem B, we proceed by contradiction. Specifically, we assume that
a ≥ 1 and that
a
H 2(p−1)p (X; Zp ) = 0,
and we derive a contradiction to the minimality of k.
By k-periodicity and the fact that λ | p − 1, the vanishing of this cohomology group
implies
a a−1
H k+2(p−1)(l−1)p +2(p−1)p (X; Zp ) = 0
for all 1 ≤ l ≤ λ. In particular,
a +pa−1
(5.1) P(l−1)p (x) = 0.
for all 1 ≤ l ≤ λ since this element lies in a vanishing cohomology group. Similar to the
proof strategy for Theorem A, we bootstrap this vanishing of P i (x) for certain degrees
to derive additional vanishing in other degrees.
Claim 1: P i (x) = 0 for all (l − 1)pa + pa−1 < i < lpa and for all 1 ≤ l ≤ λ.
8 JOHN R. HARPER AND LEE KENNARD

Proof. Fix i as in the claim, and write i = (l − 1)pa + pa−1 + δ for some 1 ≤ l ≤ λ and
some 0 < δ < (p − 1)pa−1 . By induction, we may assume the result holds for all i′ of
the form i′ = (l′ − 1)pa + pa−1 + δ ′ with 1 ≤ l′ ≤ λ and 0 < δ ′ < δ.
A calculation shows that δ < p(i − δ). Hence there exists an Adem relation
X
P δ P i−δ = c0 P i + cj P i−j P j
0<j≤ pδ

for some cj ∈ Zp . Moreover we have by Lucas’ theorem that

  Y nj 
(p − 1)(i − δ) − 1
c0 = ± =± mod p
δ j≥0
δj

j
P P j
where nj p and δj p are the p-adic expansions of (p − 1)(i − δ) − 1 and δ, respec-
tively. Notice the coefficients nj satisfy na−1 = p − 2 and na−2 = . . . = a0 = p − 1, and
the coefficients δj satisfy δj =
0 for j ≥ a and δa−1 ≤ p − 2 as a result of the bound
δ < (p − 1)pa−1 . Therefore nδjj 6≡ 0 mod p for all j ≥ 0, and so c0 6= 0.
Evaluating the above Adem relation on x, we have zero on the left-hand side by
Equation (5.1). On the right-hand side, the first term is a non-zero multiple of P i (x),
so finishing the proof requires that we show P i−j (P j (x)) = 0 for all 0 < j ≤ δ/p. To
do this, write P j (x) = xx2(p−1)j using periodicity and then
X
P i−j (P j (x)) = P i−j−h(x)P h (x2(p−1)j )
0≤h≤i−j

by the Cartan formula. For h > (p − 1)j, we have P h (x2(p−1)j) ) = 0. For h < (p − 1)j
or for h = (p − 1)j and 0 < j < δ/p, we have (l − 1)pa + pa−1 < i − j − h < i and so
our induction hypothesis implies P i−j−h (x) = 0. Finally for h = (p − 1)j and j = δ/p,
we have P i−j−h(x) = P i−δ (x) = 0 by another application of Equation (5.1). 
Claim 2: P i (x) = 0 for all (l − 1)pa < i < (l − 1)pa + pa−1 and for all 1 ≤ l ≤ λ.
Proof. Let i = (l − 1)pa + ǫ for some 1 ≤ l ≤ λ and 0 < ǫ < pa−1 . By induction assume
′
P i (x) = 0 for all i′ of the form (l′ − 1)pa + ǫ′ with 1 ≤ l′ ≤ λ and 0 < ǫ′ < ǫ.
Note that pǫ < p (λpa − (p − 1)ǫ), so there is an Adem relation of the form
a a
X ′ ′
P pǫ P λp −(p−1)ǫ = cǫ′ P λp +ǫ−ǫ (P ǫ (x))
0≤ǫ′ ≤ǫ

for some constants cǫ′ . The left-hand side vanishes by Claim 1 since
(λ − 1)pa + pa−1 < λpa − (p − 1)ǫ < λpa .
On the right-hand side, the ǫ′ = 0 term vanishes because 2 (λpa + ǫ) = k + 2ǫ is larger
than the degree of x, and the terms with 0 < ǫ′ < ǫ vanish by our induction hypothesis.
Therefore we have
a
0 = P λp (P ǫ (x)) .
Writing P ǫ (x) = xx2(p−1)ǫ by periodicity, we can apply the Cartan formula to conclude
a
X
0 = xp x2(p−1)ǫ + P λp −h (x)P h (x2(p−1)ǫ ).
0<h≤λpa
 EXTENDING ADAMS’ THEOREM 9
a
For the terms on the right-hand side with 0 < h < (p − 1)pa−1 , we have P p −h (x) = 0
by Claim 1, and for terms with h ≥ (p − 1)pa−1 , we have 2h > 2(p − 1)ǫ and hence
P h (x2(p−1)ǫ ) = 0. Therefore 0 = xp x2(p−1)ǫ . By periodicity, we have 0 = xx2(p−1)ǫ =
P ǫ (x). This finishes the proof that P i (x) = 0 if l = 1.
Now we assume 1 < l ≤ λ. We have ǫ < p(l − 1)pa , so we have an Adem relation
a
X ′ ′
(5.2) P ǫ P (l−1)p = cǫ′ P i−ǫ P ǫ .
0≤ǫ′ ≤ǫ/p

Evaluating on x, we find that all terms on the right-hand side with 0 < ǫ′ ≤ ǫ/p vanish
by our induction hypothesis. Moreover, the ǫ′ = 0 term has coefficient
(p − 1)(l − 1)pa − 1
 
c0 = ± ,
ǫ
which can be shown to be non-zero by an argument similar to the argument in Case
1. Therefore the proof of P i (x) = 0 reduces to the claim that the left-hand side of this
equation evaluates to zero. To prove this, we apply periodicity and the Cartan formula
to see that
 
(l−1)pa 1+l p−1
X
ǫ ǫ
P ǫ1 (x) · · · P ǫm (x),


P P (x) = cP x λ =c

for some c ∈ Zp , where m = 1 + l p−1

λ
and the sum runs over ǫj ≥ 0 whose sum is ǫ. By
induction, P ǫj (x) = 0 unless ǫj ∈ {0, ǫ}. Similarly P ǫ (x) = 0 by the l = 1 case, so the
right-hand side vanishes. 
Claim 3: β(x) = 0, where β is the Bockstein homomorphism associated to the short
exact sequence 0 → Z2 → Z4 → Z2 → 0 of coefficient groups.
Proof. Because k = 2λpa with a ≥ 1, we have 1 ≤ p(k/2 − 1). Therefore there is an
Adem relation
k k k
P 1 βP 2 −1 = −βP 2 + P 2 β.
Evaluating on x, the left-hand side vanishes by Claim 1, as does the first term on the
right-hand side since
k
β(P 2 (x)) = β(xp ) = pxp−1 β(x) = 0.
Therefore
k
0 = P 2 (xx1 ) = xp x1
by Cartan’s formula and Claim 1, where we have written β(x) = xx1 using periodicity.
Therefore 0 = xx1 = β(x), as claimed. 
Given Claims 1, 2, and 3, we derive a contradiction and therefore finish the proof
of Theorem B. Taking l = 1 in Claims 1 and 2 as well as in Equation (5.1), we have
P i (x) = 0 for all 0 < i < pa . Since we also have β(x) = 0 by Claim 3, [Har02, Theorem
6.2.1.b] implies that we have a conditional relation of the form
a
X
P p (x) = β(w) + P i (wi )
0<i<pa

for some cohomology elements w and wi .

10 JOHN R. HARPER AND LEE KENNARD
p−1 a
First we claim the left-hand side is a non-zero multiple of x1+ λ . Indeed, P p (x)
lives in the one-dimensional group spanned by this power of x, so the claim follows if
a a
P p (x) is non-zero. To see this, we use that P λp (x) = xp 6= 0 together with the Adem
relation of the form
a a a a
X
P (λ−1)p P p = c0 P λp + ci P λp −i P i
0<i≤(λ−1)pa−1

where
(p − 1)pa − 1
   
p−2
c0 = ± ≡± 6≡ 0 mod p.
(λ − 1)pa λ−1
Since P i (x) = 0 by Claim 1 for all i < pa and hence for all i ≤ (λ − 1)pa−1 , the claim
a p−1
that P p (x) is a non-zero multiple of x1+ λ holds.
As for the right-hand side, we can furthermore use Adem relations to decompose
j
the Steenrod powers P i using elements of the form P p with 0 ≤ j ≤ a − 1. Since this
equation holds in a one-dimensional group, we derive an equation of the form
p−1 j
x1+ λ = β(z) or P p (z)
for some cohomology element z. In the latter case, we note that pj satisfies the estimate
2(p − 1)pj ≤ 2(p − 1)pa−1 < 2pa ≤ 2λpa = k,
so the proof of Lemma 2.4 in [Ken13] implies that x decomposes non-trivially as either
a product or a Steenrod power of an element of smaller degree. As in the Z2 setting,
this implies a contradiction to the minimality of k (see [Ken13, Lemmas 1.2 and 2.3]).
p−1
Similarly, if x1+ λ = β(z) for some cohomology element z, then a similar argument
leads to a contradiction to the minimality of k (see [Ken17, Lemma 3.5]).

6. Truncated and manifold cases

For an n-dimensional Poincaré duality manifold M, periodic Z2 -cohomology is usu-

ally required to satisfy the condition that the minimum period k satisfies 2k ≤ n.
This condition ensures that x2 6= 0, which is essential to proving restrictions on the
degree even in the singly generated case. Moreover, if dim M ≡ m mod k, then we have
automatically that 2k + m ≤ n.
Similarly, for odd primes p, periodic Zp -cohomology for an n-manifold is not expected
to be restricted unless pk ≤ n. Again letting m be the congruence class of n modulo p,
Poincaré duality implies that periodicity goes from degree 0 to pk + m. We therefore
suggest the following as analogues in the truncated case of the periodicity conjecture.
Conjecture 6.1 (Periodicity Conjecture, Truncated Case). Let X n be a space and
p be a prime. Assume H ∗ (X; Zp ) is k-periodic up to degree c, in the sense that the
last isomorphism lands in degree c. Assume that k is the minimum period and that
0 ≤ m < k is the maximum degree with H m (X; Z2 ) 6= 0.
(1) If p = 2 and c ≥ 2k + m, then k ∈ {1, 2, 4} or (k, c) = (8, 16 + m).
(2) If p ≥ 3 and c ≥ pk + m, then k = 2λ for some divisor λ of p − 1.
 EXTENDING ADAMS’ THEOREM 11

As an example, the product of the Cayley plane OP2 and any closed, orientable
manifold N m with m ≤ 7 has dimension 16 + m and has 8-periodic but not 4-periodic
Z2 -cohomology. Similarly, it has 8-periodic but not 4-periodic Z3 cohomology, which is
permitted by the conjecture since pk + m > dim(OP2 × N m ).
In applications where the space with periodic cohomology is a closed, orientable
manifold, Poincaré duality imposes additional multiplicative structure. This was used
in the proof of [Ken17, Lemma 3.3], which verifies the truncated conjecture for p ≥ 3
and k = 2p in the odd-dimensional Poincaré duality manifold case. Similarly, in the
further special case where periodicity arises from n-connected inclusions N n → M n+k
of Poincaré duality manifolds, perhaps other geometric input could be applied to prove
the conjecture.

7. Structure of potential counterexamples

Theorem A implies that a counterexample X to Conjecture 1.5 with k = 16 has

non-trivial cohomology in degrees i ≡ 8 mod 16. Moreover, the discussion following
the proof of Lemma 3.4 implies that such a counterexample has additional non-trivial
cohomology in some degrees i 6≡ 0 mod 8. In this section, we further analyze the struc-
ture of a potential counterexample with k = 16 and the property that H i (X; Z2 ) = 0
for all i 6≡ 0 mod 4. In fact, our analysis works more generally under the assumption
that Sqi = 0 for all i 6≡ 0 mod 4.
Lemma 7.1. Let X be a space for which H ∗ (X; Z2 ) is 16-periodic but not 8-periodic.
Assume in addition that Sq1 = 0 and Sq2 = 0 on H ∗ (X; Z2 ). Let x ∈ H 16 (X; Z2 )
denote the element inducing periodicity, and let yi ∈ H i (X; Z2 ) denote the elements
satisfying the equation Sqi x = xyi for all i ∈ {4, 8, 12}. All of the following hold:
(1) The map x−1 Sq16 equals the identity on H 16+i (X; Z2 ) for all i ∈ {0, 4, 8, 12},
where x−1 denotes the inverse of the isomorphism given by multiplication by x.
(2) Products of the form u8 v12 , u12 v12 , and u8 Sq8 (v12 ) vanish, where subscripts
denote degrees.
(3) Products of the form yi yj vanish for all i and j, including i = j.
(4) Sq4 y8 = y12 and Sq8 y12 = xy4 .
Proof. Part (1) is immediate for i = 0 and is straightforward for i = 4 since x2 does
not non-trivially factor. The proof for i ∈ {8, 12} requires results on products, so we
prove these next.
To prove the first claim of Part (2), suppose that u8 v12 = xw4 6= 0. Applying Sq16 to
both sides and applying (1), we have Sq4 (u8 ) Sq12 (v12 ) = x2 w4 6= 0. Pulling out the Sq4
using the Cartan formula, we have that u8 v12 2
= x2 since u8 Sq4 (v122
) = u8 Sq2 (v12 )2 = 0.
This is a contradiction since x2 does not non-trivially factor.
We now prove Part (1) for i = 8. The Cartan formula implies Sq16 (xt8 ) = x2 t8 +
xy12 Sq4 (t8 ) since t28 = 0. The claim follows because y12 Sq4 (t8 ) = Sq4 (y12 t8 )−Sq4 (y12 )t8 ,
which vanishes by the first claim of (2) together with the fact that x is not in the image
of a non-trivial Steenrod operation.
12 JOHN R. HARPER AND LEE KENNARD

The last claim of Part (1) follows similarly, so we omit the proof.
Next we prove the second claim in (2), suppose u12 v12 = xw8 6= 0. By Part (1),
applying Sq16 yields
x2 w8 = Sq8 (u12 ) Sq8 (v12 ) = Sq8 u12 Sq8 (v12 ) − Sq4 (u12 ) Sq4 Sq8 v12 − u12 Sq8 Sq8 v12 .

The first two terms on the right-hand side vanish by the minimality properties of x,
and the last also vanishes by using the Adem relation for Sq8 Sq8 .
The last claim of Part (2) follows by the Cartan formula, which implies u8 Sq8 (v12 ) =
Sq8 (u8 v12 ) − Sq4 (u8 ) Sq4 (v12 ) − u28 v12 , which vanishes by the first part of (2) and the
properties of x following from minimality.
To prove Part (3), first consider the xy42 = y4 Sq4 x. Applying Sq16 to both sides and
applying (1), we obtain
x2 y42 = Sq16 (y4 Sq4 x) = y4 Sq16 Sq4 x = y4 Sq8 Sq12 x.
Applying the Cartan formula to the action of Sq8 , we have
y4 Sq8 Sq12 x = Sq8 y4 Sq12 x − y42 Sq4 Sq12 x = 0.

Hence xy42 = 0, which implies y42 vanishes by periodicity.

To finish the proof of Part (3), we only need to prove y4 y8 = 0 since the other
products vanish by minimality of x. For this we follow a similar strategy as before.
Consider the term xy4 y8 = y8 Sq4 x. Applying Sq16 yields
x2 y4 y8 = y8 Sq16 Sq4 x + Sq4 y8 Sq12 Sq4 x.
The first term on the right-hand side vanishes as in the previous argument, using in
addition the fact that y8 Sq12 x = 0 by Part (2). The second term on the right-hand side
also vanishes because x2 does not decompose non-trivially as the image of a Steenrod
operation. Hence the right-hand side is zero in contradiction to Part (1).
To prove Part (4), we first apply Part (3) to calculate
xy12 = Sq12 (x) = Sq4 Sq8 x = Sq4 (xy8 ) = x Sq4 y8 ,
which proves y12 = Sq4 y8 . Similarly, we can apply Part (2) and the minimality prop-
erties of x to calculate
x Sq8 y12 = Sq8 (xy12 ) = Sq8 Sq12 x = Sq20 x + Sq16 Sq4 x = Sq16 (xy4 ).
Multiplying by x−1 and applying Part (1) yields the claim. 
We can push this analysis further with the use of secondary cohomology operations.
Lemma 7.2. If X is a counterexample to Conjecture 1.5 with k = 16, and if Sq1 and
Sq2 vanish on H ∗ (X; Z2 ), then Sq8 Sq4 y8 = xy4 6= 0 where Sqi x = xyi for i ∈ {4, 8}.
Proof. We assume that x ∈ H 16 (X; Z2 ) induces periodicity and that 16 is the minimum
period. We assume also that Sq1 = 0 and Sq2 = 0 everywhere. In particular, Lemma
7.1 applies. If Sq4 (x) 6= 0, then Lemma 7.1 implies that Sq8 Sq4 y8 = Sq8 y12 = xy4 6= 0,
as claimed. To finish the proof, we assume Sq4 (x) = 0 and seek a contradiction.
Lemma 7.1 only used primary operations in the proof. Here we use the following
secondary decompositions:
 EXTENDING ADAMS’ THEOREM 13

(a) If u ∈ ker(Sq1 , Sq2 , Sq4 , Sq8 ), then Sq16 u = Sq8 (v)+Sq4 Sq8 (v ′ )+Sq4 Sq2 Sq1 (v ′′ )
modulo im(Sq1 , Sq2 ) for some v, v ′ , and v ′′ .
(b) If u ∈ ker(Sq1 , Sq2 , Sq8 ) ∩ ker(Sq2 Sq4 , Sq8 Sq4 ), then Sq16 u = Sq4 (w) + Sq8 (w ′ )
modulo im(Sq1 , Sq2 ) for some w and w ′ .
The first decomposition follows from [Har02, Theorem 6.2.1]. For the second one, we
refer to [LW91, Section 3]. We note first that the α13 in the notation of [LW91] should
be α13 = Sq3 + Sq2 Sq1 . We also remark that Lin and Williams state the secondary
decomposition only for integral cohomology elements u. However, there is a standard
procedure for padding the matrix B and vector c = (Sq2 , Sq6 , Sq8 , Sq8 Sq4 )T in the
equation at the top of [LW91, page 180] and adding an entry to the vector
a = (α3 , α8 , α9 , α12 , α13 , α15 )
to extend the decomposition to all (not necessarily integral) elements as long as, in
addition, they lie in the kernel of Sq1 . This is sufficient to obtain (b). We omit the
details.
By Parts (2) and (4) of Lemma 7.1, we have xy8 ∈ ker(Sq8 Sq4 ). It is then easy to
check that secondary decomposition (b) applies to xy8 . By Part (1), we have
x2 y8 = Sq16 (xy8 ) = Sq4 (xz4 ) + Sq8 (x2 z0 ),
where z4 ∈ H 4 (X; Z2 ) and z0 ∈ H 0 (X; Z2 ). Since y4 = 0, we conclude
y8 = z42 .
We claim that z43 = 0. Indeed, xz43 ∈ ker(Sq4 , Sq8 ), so conditional relation (a) applies.
By Part (1), we have
x2 z43 = Sq16 (xz43 ) = Sq8 (x2 a4 ) + Sq4 Sq8 (x2 a0 ) + Sq4 Sq2 Sq1 (x2 a5 )
for some a4 ∈ H 4 (X; Z2 ), a0 ∈ H 0 (X; Z2 ), and a5 ∈ H 5 (X; Z2 ). All terms on the right-
hand side vanish by the Cartan formula and the assumptions that y4 = 0 and Sq1 = 0,
so we have that
z43 = 0,
as claimed. It now follows that xz4 ∈ ker(Sq8 ) since y8 = z42 and y4 = 0. But we also
have xz4 ∈ ker(Sq8 Sq4 ) since this lands in the group generated by x2 , which does not
decompose by the minimality properties of x. Hence conditional relation (b) applies to
xz4 and we have a relation of the form
x2 z4 = Sq16 (xz4 ) = Sq4 (x2 b0 ) + Sq8 (xb12 ).
As before the first term on the right-hand side is zero, and the other equals x Sq8 (z12 ).
We now have
xy8 = xz42 = Sq8 (z12 )z4 = Sq8 (z12 z4 ) + Sq4 (z12 )z42 .
Both terms on the right-hand side vanish by the minimality properties of x, so we
y8 = 0.
But now the proof of the main theorem applies. In short, we can now prove that
conditional relation (a) applies to x, and the minimality properties of x imply that
y16 = 0,
14 JOHN R. HARPER AND LEE KENNARD

where for suggestive reasons we have denoted x by y16 since they both satisfy the
equation Sq16 (x) = xy16 . This is a contradiction, so the proof is complete. 
To motivate further investigation, we propose the following as a minimal potential
counterexample to Conjecture 1.5 when k = 16.
Question 7.3. Does there exist a space X with Z2 -cohomology ring isomorphic to
Z2 [y4 , y8 , y12 , x16 ]/(y42 , y4 y8 , y4y12 , y82, y8 y12 , y12
2
) and the properties that Sqi x16 = x16 yi
for all i ∈ {4, 8, 12}, Sq4 y8 = y12 , and Sq8 y12 = xy4 .
References
[Ada60] J.F. Adams. On the non-existence of elements of Hopf invariant one. Ann. of Math.,
72(1):20–104, 1960.
[Ade52] J. Adem. The iteration of the Steenrod squares in algebraic topology. Proc. Nat. Acad. Sci.
USA, 38:720–726, 1952.
[Gon77] D.L. Goncalves. Mod 2 homotopy-associative H-spaces. ProQuest LLC, Ann Arbor, MI,
1977. Thesis (Ph.D.)–University of Rochester.
[Har02] J. Harper. Secondary cohomology operations. American Mathematical Society, 2002.
[Ken13] L. Kennard. On the Hopf conjecture with symmetry. Geom. Topol., 17:563–593, 2013.
[Ken17] L. Kennard. Fundamental Groups of Manifolds with Positive Sectional Curvature and Torus
Symmetry. J. Geom. Anal., 27(4):2894–2925, 2017.
[KWW1] L. Kennard, M. Wiemeler, and B. Wilking. Splitting of torus representations and applica-
tions in the grove symmetry program. preprint, arXiv:2106.14723, .
[Liu62] A. Liulevicius. The factorization of cyclic reduced powers by secondary cohomology opera-
tions. Mem. Amer. Math. Soc., 42, 1962.
[LW90] J.P. Lin and F. Williams. Two torsion and homotopy associative H-spaces. J. Math. Kyoto
Univ., 30(3):523–541, 1990.
[LW91] J.P. Lin and F. Williams. The type of a torsion free finite loop space. Topology Appl.,
42(2):175–186, 1991.
[Nie] J. Nienhaus. An improved four-periodicity theorem and a conjecture of Hopf with symmetry.
[SY61] N. Shimada and T. Yamanoshita. On triviality of the mod p Hopf invariant. Jpn. J. Math.,
31:1–25, 1961.
[Wil03] B. Wilking. Torus actions on manifolds of positive sectional curvature. Acta Math.,
191(2):259–297, 2003.

Department of Mathematics, University of Rochester, Rochester, NY 14627, USA

Email address: [email protected]

Department of Mathematics, Syracuse University, Syracuse, NY 13244, USA

Email address: [email protected]