SUSPENSION HOMOTOPY OF SIMPLY CONNECTED 7-MANIFOLDS

RUIZHI HUANG AND PENGCHENG LI

Abstract. In homotopy category away from 2, we decompose a simply connected closed 7-

manifold into indecomposable pieces after suspension. As an immediate consequence, this shows
arXiv:2208.13145v1 [[Link]] 28 Aug 2022

certain rigidity property of the manifold. Interesting examples include the respective suspension
decompositions of string 7-manifolds and generalized Aloff-Wallach manifolds.

Contents

1. Introduction 2
2. Homology decomposition of M after suspension 7
2.1. Structure of ΣM(3) 7
2.2. Structure of ΣM(4) 7
2.3. Structure of ΣM(5) 9
2.4. Structure of ΣM 10
3. A refined decomposition 11
3.1. 7-Manifolds with bottom cells killed 11
3.2. Z is a suspension when H or T = 0 13
3.3. ΣZ in general case 14
4. A decomposition via Steenrod module structure 15
4.1. A decomposition of ΣY 15
4.2. A decomposition of ΣX 17
5. Proof of Theorem 1.1, 1.2, 1.3, 1.4 and 1.6 18
5.1. A family of building blocks 18
5.2. Proof of the theorems 19
6. Technical lemmas 20
6.1. The Blakers-Massey theorem 20
6.2. Chang complexes 20
6.3. A lemma on suspension decomposition 21
6.4. Odd primary Moore spaces 22
6.5. Homotopy groups with coefficients 25
References 26

2010 Mathematics Subject Classification. Primary 55P15, 55P40, 57R19; Secondary 55P10, 55Q52.
1
2 RUIZHI HUANG AND PENGCHENG LI

1. Introduction

Seven dimensional manifolds are important in geometry and topology. For instance, in geometric
topology, Kreck [Kre2] and Crowley-Nordström [CN] independently gave a complete classification
of closed smooth 2-connected 7-manifolds in terms of a system of algebraic invariants. Based on his
famous work [Kre1], Kreck [Kre2] also made further contributions on the classification of 1-connected
7-manifolds with torsion free second homology. In differential geometry, among others there has been
a steadily increasing interest in G2 -manifolds. Following the seminal work of Donaldson-Thomas
[DT], in the new century the gauge theory of G2 -manifolds [FIC] was developed rapidly. Passing to
algebraic topology of gauge theory, there is a rough principle [The, So, Hua2] that the homotopy of
gauge groups is largely controlled by the suspension homotopy of the underlying manifolds.
We are interested in the suspension homotopy of 7-manifolds in this paper. It is a relatively new
topic in unstable homotopy theory to study the suspension homotopy of manifolds. Huang [Hua2]
determines the suspension homotopy type of (n − 1)-connected 2n-manifolds. In concrete case,
Theriault [The], So [So] and So-Theriault [ST] studied the suspension homotopy of 4-manifolds,
while Huang [Hua1] studied the problem for certain 5-manifolds. More recently, Huang [Hua3]
and Cutler-So [CS] studied the suspension homotopy of 6-manifolds. Besides, Membrillo-Solis [MS]
studied the problem for S 3 -bundle over S 4 at large primes. All these works can be applied to
investigate the homotopy of gauge groups, which is an active topic in unstable homotopy theory.
In this paper, we study the suspension homotopy of 7-manifolds after localization away from 2.
Let M be a simply connected closed 7-manifold. By Poincaré duality and the universal coefficient
theorem we have


 Z⊕r ⊕ H ∗ = 2,





 Z⊕d ⊕ T ∗ = 3,




 Z⊕d ⊕ H ∗ = 4,
(1) H∗ (M ; Z) =

 Z⊕r ∗ = 5,





 Z ∗ = 0, 7




 0 otherwise,

where r, d ≥ 0, and H and T are two finitely generated abelian torsion groups. Note that for Kreck’s
manifold with torsion free second homology, H = 0. Before proceeding to general cases, we consider
the suspension homotopy problem for two interesting families of 7-manifolds.
The first case is when M is a mod-3 string manifold. Recall M is string if the second Stiefel-
1
Whitney class ω2 (M ) = 0 and half of the first Pontryagin class 2 p1 (M ) = 0. It is well known
that string manifolds are fundamental in modern geometry, topology and mathematical physics; for
instance, see Witten [Wit] and Hopkins [Hop], etc. Here, we adopt a weaker version, and define M
is mod-3 string if p1 (M ) ≡ 0 mod 3.
 SUSPENSION HOMOTOPY OF SIMPLY CONNECTED 7-MANIFOLDS 3

Let ΣX denote the suspension of any CW -complex X. Let P n (A) be the Moore space such that
e ∗ (P n (A); Z) is isomorphic to the abelian group A if ∗ = n and 0 otherwise
the reduced cohomology H
[N2]. When A = Z/k is cyclic, denote P n (A) = P n (k).

Theorem 1.1. Let M be a simply connected closed mod-3 string 7-manifold. Then its homology is
of the form (1), and we have the following:

(1). if M has torsion free second homology, that is H = 0, there is a homotopy equivalence after
localization away from 2
r
_ d
_
3 6
ΣM ≃( 12 ) (S ∨ S ) ∨ (S 4 ∨ S 5 ) ∨ P 5 (T ) ∨ S 8 ;
i=1 i=1

(2). if M has torsion free third homology, that is T = 0, there is a homotopy equivalence after
localization away from 2
r
_ d
_
ΣM ≃( 12 ) (S 3 ∨ S 6 ) ∨ (S 4 ∨ S 5 ) ∨ P 4 (H) ∨ P 6 (H) ∨ S 8 ;
i=1 i=1

(3). more generally, there is a homotopy equivalence after localization away from 2
r
_ d
_
Σ2 M ≃( 21 ) (S 4 ∨ S 7 ) ∨ (S 5 ∨ S 6 ) ∨ P 5 (H) ∨ P 6 (T ) ∨ P 7 (H) ∨ S 9 .
i=1 i=1

The second case is when M is a generalized Aloff-Wallach manifold, that is, M has the homology
of the form (1) such that r = 1, d = 0, H = 0 and T = Z/k (k ≥ 1). In this case, M is called of
type k in the literature. Generalized Aloff-Wallach manifolds are deeply investigated in geometry
and topology. For instance, see the famous work of Aloff-Wallach [AW] and Kreck-Stolz [KS] for
Aloff-Wallach manifolds, and of Eschenburg [Esc] for Eschenburg manifolds; more general case was
studied by Kruggel in a series of papers [Kru1, Kru2, Kru3].
We recall a family of classical manifolds, which will be used as new building blocks in our future
decompositions. Following the notation of Crowley and Escher [CE], any pair of integers (ρ, γ) ∈
Z⊕Z∼
= π3 (SO(4)) determines a fibre bundle

S 3 −→ Mρ,γ −→ S 4

over the 4-sphere with structure group SO(4). The total manifolds Mρ,γ in this family are the
simplest examples of 2-connected 7-manifolds.

Theorem 1.2. Let M be a generalized Aloff-Wallach manifold of type k = 3ν · t with ν ≥ 0 and

(3, t) = 1. We have the following:

(1). if p1 (M ) ≡ 0 mod 3, then there is a homotopy equivalence

ΣM ≃( 12 ) S 3 ∨ S 6 ∨ P 5 (k) ∨ S 8 ;
4 RUIZHI HUANG AND PENGCHENG LI

(2). if p1 (M ) 6≡ 0 mod 3, then there is a homotopy equivalence

ΣM ≃( 12 ) S 3 ∨ S 6 ∨ P 5 (t) ∨ ΣMρ,3ν ,

for any ρ ∈ Z with (ρ, 3) = 1.

Remark. The homotopy of Mρ,3ν is accessible. In Lemma 5.1, it is shown that when (ρ, 3) = 1

ΣMρ,3ν ≃ P 5 (3ν ) ∪i◦Σα e8 ,

where the order 3 generator α : S 6 → S 3 is stable, and i : S 3 ֒→ P 4 (3ν ) is the bottom cell inclusion.
In particular, when (ρ, 3) = 1 the complex Mρ,3ν is stably indecomposable.

For more general case of 7-manifolds, we introduce a numerical invariant of M , though it can be
defined generally. Recall there is the long exact sequence [Hat, Chapter 3.E]

ρ3 β
· · · −→ H 3 (M ; Z) −→ H 3 (M ; Z/3) −→ H 4 (M ; Z) −→ · · · ,

where ρ3 is the mod-3 reduction, and β is the Bockstein homomorphism. For any x ∈ H 3 (M ; Z/3),
define its radius to be the number


 −1 x = ρ3 (x), order(x) = 3µ · t, (3, t) = 1,

 µ
rad(x) = 0 x = ρ3 (x), order(x) = ∞,



 1
ν 3ν−1 | β(x), 3ν ∤ β(x).

If p1 (M ) 6≡ 0 mod 3, define the p1 -radius of M to be the number

radp1 (M ) := min{rad(x) | (p1 (M ) mod 3) ∪ x 6= 0, x ∈ H 3 (M ; Z/3)}.

It is clear that the above definitions are well defined. Moreover, as (p1 (M ) mod 3) is a homotopy
invariant by the Wu formula [Wu], radp1 (M ) is a homotopy invariant of M .
The following result concerns Kreck’s 7-manifolds with torsion free second homology. Note that
in this case, H = 0 and then radp1 (M ) is nonnegative.

Theorem 1.3. Let M be a simply connected closed 7-manifold with torsion free second homology.
Then its homology is of the form (1) with H = 0, and we have the following:

(1). if p1 (M ) ≡ 0 mod 3, then there is a homotopy equivalence

r
_ d
_
ΣM ≃( 12 ) (S 3 ∨ S 6 ) ∨ (S 4 ∨ S 5 ) ∨ P 5 (T ) ∨ S 8 ;
i=1 i=1

(2). if radp1 (M ) = 0, then there is a homotopy equivalence

r
_ d−1
_ d
_
ΣM ≃( 12 ) (S 3 ∨ S 6 ) ∨ ( S4) ∨ ( S 5 ) ∨ P 5 (T ) ∨ (S 4 ∪Σα e8 );
i=1 i=1 i=1
 SUSPENSION HOMOTOPY OF SIMPLY CONNECTED 7-MANIFOLDS 5

1
(3). if radp1 (M ) = ν for some ν ∈ Z+ , then there is a homotopy equivalence
r
_ d
_ T
ΣM ≃( 21 ) (S 3 ∨ S 6 ) ∨ (S 4 ∨ S 5 ) ∨ P 5 ( ) ∨ ΣMρ,3ν ,
i=1 i=1
Z/3ν

for any ρ ∈ Z with (ρ, 3) = 1.

Remark. It is well known that [Har, Section 1.5.5]

ΣCP 3 ≃( 12 ) S 5 ∨ (S 3 ∪α e7 ).

Hence, Case (2) of Theorem 1.3, and Case (3) of Theorem 1.6 below, can be simplified when r ≥ 1.

To state the next result, in Lemma 6.8 we show that π7 (P 4 (3µ )) ∼

= Z/3{e
α} such that the com-
e
α q
position S 7 −→ P 4 (3µ ) −→ S 4 is the stable map Σα, where q is the pinch map. In particular, the
complex P 4 (3µ ) ∪αe e8 is stably indecomposable.

Theorem 1.4. Let M be a simply connected closed 7-manifold with torsion free third homology.
Then its homology is of the form (1) with T = 0, and we have the following:

(1). if p1 (M ) ≡ 0 mod 3, then there is a homotopy equivalence

r
_ d
_
ΣM ≃( 21 ) (S 3 ∨ S 6 ) ∨ (S 4 ∨ S 5 ) ∨ P 4 (H) ∨ P 6 (H) ∨ S 8 ;
i=1 i=1

(2). if radp1 (M ) = − µ1 for some µ ∈ Z+ , then there is a homotopy equivalence

r
_ d
_ H
ΣM ≃( 12 ) (S 3 ∨ S 6 ) ∨ (S 4 ∨ S 5 ) ∨ P 4 ( ) ∨ P 6 (H) ∨ (P 4 (3µ ) ∪αe e8 );
i=1 i=1
Z/3µ

(3). if radp1 (M ) = 0, then there is a homotopy equivalence,

r
_ d−1
_ d
_
ΣM ≃( 12 ) (S 3 ∨ S 6 ) ∨ ( S4) ∨ ( S 5 ) ∨ P 4 (H) ∨ P 6 (H) ∨ (S 4 ∪Σα e8 ).
i=1 i=1 i=1

Theorem 1.3 and 1.4 imply the following rigidity result immediately. Let Z( 12 ) be the localization
of the integer ring Z away from 2.

Corollary 1.5. Let M and M ′ be two simply connected closed 7-manifolds with torsion free second
or third homology. Then ΣM ≃( 21 ) ΣM ′ if and only if H∗ (M ; Z( 21 ) ) ∼
= H∗ (M ′ ; Z( 12 ) ) as abelian
groups, p1 (M ) ≡ p1 (M ′ ) mod 3, and radp1 (M ) = radp1 (M ′ ) if p1 (M ) ≡ p1 (M ′ ) 6≡ 0 mod 3. 

The most general case of 7-manifolds is studied in the following result, but with one more fold
suspension applied.

Theorem 1.6. Let M be a simply connected closed 7-manifold. Then its homology is of the form
(1), and we have the following:
6 RUIZHI HUANG AND PENGCHENG LI

(1). if p1 (M ) ≡ 0 mod 3, then there is a homotopy equivalence

r
_ d
_
Σ2 M ≃( 12 ) (S 4 ∨ S 7 ) ∨ (S 5 ∨ S 6 ) ∨ P 5 (H) ∨ P 6 (T ) ∨ P 7 (H) ∨ S 9 ;
i=1 i=1

(2). if radp1 (M ) = − µ1 for some µ ∈ Z+ , then there is a homotopy equivalence

r
_ d
_ H
Σ2 M ≃( 21 ) (S 4 ∨ S 7 ) ∨ (S 5 ∨ S 6 ) ∨ P 5 ( ) ∨ P 6 (T ) ∨ P 7 (H) ∨ (P 5 (3µ ) ∪Σe 9
α e );
Z/3µ
i=1 i=1

(3). if radp1 (M ) = 0, then there is a homotopy equivalence,

r
_ d−1
_ d
_
Σ2 M ≃( 12 ) (S 4 ∨ S 7 ) ∨ ( S5) ∨ ( S 6 ) ∨ P 5 (H) ∨ P 6 (T ) ∨ P 7 (H) ∨ (S 5 ∪Σ2 α e9 );
i=1 i=1 i=1
1
(4). if radp1 (M ) = ν for some ν ∈ Z+ , then there is a homotopy equivalence
r
_ d
_ T
Σ2 M ≃( 12 ) (S 4 ∨ S 7 ) ∨ (S 5 ∨ S 6 ) ∨ P 5 (H) ∨ P 6 ( ) ∨ P 7 (H) ∨ Σ2 Mρ,3ν ,
i=1 i=1
Z/3ν

for any ρ ∈ Z with (ρ, 3) = 1.

Theorem 1.6 implies the following rigidity result immediately.

Corollary 1.7. Let M and M ′ be two simply connected closed 7-manifolds. Then Σ2 M ≃( 12 )
Σ2 M ′ if and only if H∗ (M ; Z( 21 ) ) ∼
= H∗ (M ′ ; Z( 12 ) ) as abelian groups, p1 (M ) ≡ p1 (M ′ ) mod 3, and
radp1 (M ) = radp1 (M ′ ) if p1 (M ) ≡ p1 (M ′ ) 6≡ 0 mod 3. 

The paper is organized as follows. In Section 2, we give a detailed procedure to decompose a simply
connected 7-manifold M after suspension by the homology decomposition method. In particular,
after localization away from 2 ΣM is a wedge of Moore spaces with an extra complex Z. In Section
3, we show that Z is suspension when the second or third homology of M is torsion free. In Section
4, we show a further decomposition of Z or ΣZ based on its Steenrod module structure. In Section
5 we prove the main theorems, Theorem 1.1, 1.2, 1.3, 1.4 and 1.6. Section 6 is devoted to reviewing
or proving the various homotopy technical results used in this paper, including the Blakers-Massey
theorem, the classification of the Chang complexes, a useful lemma on suspension decomposition,
homotopy groups of odd primary Moore spaces, and homotopy groups with coefficients. Other used
results in geometrical topology are cited by references.

Acknowledgement Ruizhi Huang was supported in part by the National Natural Science Founda-
tion of China (Grant nos. 11801544 and 12288201), the National Key R&D Program of China (No.
2021YFA1002300), the Youth Innovation Promotion Association of Chinese Academy Sciences, and
the “Chen Jingrun” Future Star Program of AMSS.
Pencheng Li was supported by the National Natural Science Foundation of China (Grant No.
12101290) and China Postdoctoral Science Foundation (Grant No. 2021M691441).
 SUSPENSION HOMOTOPY OF SIMPLY CONNECTED 7-MANIFOLDS 7

2. Homology decomposition of M after suspension

Let M be a simply connected closed 7-manifold. It has the homology of the form (1). We may
apply homology decomposition to study the cell structure of M . For any finitely generated abelian
e ∗ (P n (A); Z) ∼
group A, let P n (A) be the Moore space such that H = A if ∗ = n and 0 otherwise [N2].

Theorem 2.1. [Hat, Theorem 4H.3] Let X be a simply connected CW -complex. Denote Hi =
Hi (X; Z). Then there is a sequence of complexes X(i) (i ≥ 2) such that

• Hj (X(i) ; Z) ∼
= Hj (X; Z) for j ≤ i and Hj (X(i) ; Z) = 0 for j > i;
• X(2) = P 3 (H2 ), and X(i) is defined by a homotopy cofibration
fi−1 ιi−1
P i (Hi ) −→ X(i−1) −→ X(i) ,

where fi−1 induces a trivial homomorphism fi−1∗ : Hi−1 (P i (Hi ); Z) → Hi−1 (X(i−1) ; Z);
2 ι ιi−2 i ιi−1 ι
• X ≃ hocolim{X(2) −→ · · · −→ X(i−1) −→ X(i) −→ · · · }. 

From this theorem, it is clear that the homology decomposition is compatible with the suspension
functor. That is, for X in Theorem 2.1 the sequence of the triples (ΣX(i) , Efi , Eιi ) is a homology
decomposition of ΣX.
From now on, we work in the homotopy category after localization away from 2.

2.1. Structure of ΣM(3) . By Theorem 2.1 and (1), there is a homotopy cofibration
d
_ r
_
f2 ι
(2) ( S 2 ) ∨ P 3 (T ) −→ M(2) −→
2
M(3) , where M(2) ≃ S 2 ∨ P 3 (H).
i=1 i=1

Lemma 2.2. After localization away from 2 the map Σι2 in (2) has a left homotopy inverse, and
hence
r
_ d
_
ΣM(3) ≃( 12 ) ( S3) ∨ ( S 4 ) ∨ P 4 (H) ∨ P 5 (T ).
i=1 i=1

Proof. Notice that ΣM(3) is a A23 -complex. By Lemma 6.3 and the homology structure of ΣM(3) ,
the local homotopy decomposition in the statement holds. Consider the composition
r
_ r
_ d
_ r
_
Σι q
S 3 ∨ P 4 (H) ≃ ΣM(2) −→
2
ΣM(3) ≃( 12 ) ( S3) ∨ ( S 4 ) ∨ P 4 (H) ∨ P 5 (T ) −→ S 3 ∨ P 4 (H),
i=1 i=1 i=1 i=1

where q is the obvious projection. It induces an isomorphism on homology, and then is a homotopy
equivalence by the Whitehead Theorem. Therefore Σι2 has a left homotopy inverse. 

2.2. Structure of ΣM(4) . By Theorem 2.1 and Lemma 2.2, there is a homotopy cofibration
d
_ f3 ι
S 3 ∨ P 4 (H) −→ M(3) −→
3
M(4) ,
i=1
(3)
r
_ d
_
where ΣM(3) ≃( 12 ) ( S3) ∨ ( S 4 ) ∨ P 4 (H) ∨ P 5 (T ).
i=1 i=1
8 RUIZHI HUANG AND PENGCHENG LI

Lemma 2.3. After localization away from 2, there is a homotopy equivalence

r
_ d
_
ΣM(4) ≃( 12 ) ( S3) ∨ (S 4 ∨ S 5 ) ∨ P 4 (H) ∨ P 5 (T ) ∨ P 6 (H).
i=1 i=1

Proof. Consider the homotopy commutative diagram

ΣM(2)
7
g3
Σι2

W
d Σf3  Σι3
S 4 ∨ P 5 (H) / ΣM(3) / ΣM(4)
(4) i=1
▼▼▼
▼▼▼Σf3′
▼▼▼ Σκ3
▼& 
Wd
S 4 ∨ P 5 (T ), ,
i=1

where the second row is the suspension of the cofibration (3), κ3 is the quotient map obtained by
extending the cofibration (2) to the right, f3′ := κ3 ◦ f3 by definition, and the map g3 will be defined
momentarily.
W
d W
d
′
Indeed, by Theorem 2.1, Σf3∗ : H4 ( S 4 ∨ P 5 (H); Z) → H4 ( S 4 ∨ P 5 (T ); Z) is trivial. The
i=1 i=1
W
d
Hurewicz theorem implies that the restriction of the suspension map Σf3′ on S 4 is null homotopic,
i=1
while Lemma 6.11 and 6.12 implies that its restriction on P 5 (H) is also null homotopic. Hence Σf3′
W
d
is null homotopic. Note that ΣM(2) is 2-connected and S 4 ∨ P 5 (T ) is 3-connected. By the
i=1
Blakers-Massey theorem (Lemma 6.1), the homotopy cofibration in the middle column of Diagram
(4) is a homotopy fibration up to degree 5. Hence, the map Σf3 can be lifted to a map g3 to ΣM(2) .
We have completed the construction of the diagram.
We claim that by abuse of notation each component of the composition
d
_ r
_
g3 Σ(≃)
g3 : S 4 ∨ P 5 (H) −→ ΣM(2) −→ S 3 ∨ P 4 (H)
i=1 i=1

is null homotopic. Indeed, as π4 (S ) ∼ 3

= Z/2, and by Lemma 6.11 [P 5 (H), S 3 ]( 12 ) = 0, the compo-
nent of g3 on each S 3 is null homotopic. Also, as π4 (P 4 (pr ))( 12 ) = 0 by [Hua3, Lemma 6.3], and
[P 5 (H), P 4 (pr )]( 12 ) = 0 by Lemma 6.11, the component of g3 on each summand of P 4 (H) is null
homotopic. Hence, the claim is true, and then each component of Σf3 ≃ Σι2 ◦ g3 is null homotopic.
By Lemma 6.4, there is a homotopy equivalence
d
_
ΣM(4) ≃( 12 ) ΣM(3) ∨ S 5 ∨ P 6 (H)
i=1
r
_ d
_ d
_
≃( 12 ) ( S3) ∨ ( S 4 ) ∨ P 4 (H) ∨ P 5 (T ) ∨ S 5 ∨ P 6 (H),
i=1 i=1 i=1

where the second homotopy equivalence follows from Lemma 2.2. This proves the lemma. 
 SUSPENSION HOMOTOPY OF SIMPLY CONNECTED 7-MANIFOLDS 9

2.3. Structure of ΣM(5) . By Theorem 2.1 and Lemma 2.3, there is a homotopy cofibration
r
_ f4 ι
S 4 −→ M(4) −→
4
M(5) ,
i=1
(5)
r
_ d
_
3
where ΣM(4) ≃( 21 ) ( S )∨ (S 4 ∨ S 5 ) ∨ P 4 (H) ∨ P 5 (T ) ∨ P 6 (H).
i=1 i=1

Lemma 2.4. After localization away from 2, there is a homotopy equivalence

r
_ d
_
ΣM(5) ≃( 12 ) (S 3 ∨ S 6 ) ∨ (S 4 ∨ S 5 ) ∨ P 4 (H) ∨ P 5 (T ) ∨ P 6 (H).
i=1 i=1

Proof. Consider the homotopy commutative diagram

ΣM(3)
9
g4
Σι3

W
r Σf4  Σι4
S5 / ΣM(4) / ΣM(5)
(6) i=1 ❍❍
❍❍ Σf ′
❍❍ 4
❍❍ Σκ4
❍❍
$ 
Wd
S 5 ∨ P 6 (H),
i=1

where the second row is the suspension of the cofibration (5), κ4 is the quotient map obtained by
extending the cofibration (3) to the right, f4′ := κ4 ◦ f4 by definition, and the map g4 will be defined
momentarily.
W
d W
d
′
Indeed, by Theorem 2.1, Σf4∗ : H5 ( S 5 ; Z) → H5 ( S 5 ∨ P 6 (H); Z) is trivial. The Hurewicz
i=1 i=1
theorem implies that the suspension map Σf4′ is null homotopic. Note that ΣM(2) is 2-connected
W
d
and S 5 ∨ P 6 (H) is 4-connected. By the Blakers-Massey theorem (Lemma 6.1), the homotopy
i=1
cofibration in the middle column of Diagram (6) is a homotopy fibration up to degree 6. Hence, the
map Σf4 can be lifted to a map g4 to ΣM(3) . We have completed the construction of the diagram.
Now since π5 (S 3 ) ∼
= π5 (S 4 ) ∼
= Z/2 and π5 (P 4 (pr ))( 12 ) = π5 (P 5 (pr ))( 21 ) = 0 by [Hua3, Lemma 6.3,
6.4], by abuse of notation each component of the map
r
_ r
_ d
_
g4 ≃
g4 : S 5 −→ ΣM(3) −→ ( S3) ∨ ( S 4 ) ∨ P 4 (H) ∨ P 5 (T )
i=1 i=1 i=1

is null homotopic away from 2, where the second homotopy equivalence is from Lemma 2.2. Then so
is each component of the map Σf4 ≃ Σι3 ◦ g4 . Hence, by Lemma 6.4 there is a homotopy equivalence
r
_
ΣM(5) ≃( 12 ) ΣM(4) ∨ S6
i=1
r
_ d
_ d
_ r
_
≃( 12 ) ( S3) ∨ ( S 4) ∨ ( S 5 ) ∨ P 4 (H) ∨ P 5 (T ) ∨ P 6 (H) ∨ S6,
i=1 i=1 i=1 i=1

where the second homotopy equivalence follows from Lemma 2.3. This proves the lemma. 
10 RUIZHI HUANG AND PENGCHENG LI

2.4. Structure of ΣM. By Theorem 2.1 and Lemma 2.4, there is a homotopy cofibration
f6 ι
S 6 −→ M(5) = M(6) −→
6
M,
(7) r
_ d
_
3 6
where ΣM(5) ≃( 12 ) (S ∨ S ) ∨ (S 4 ∨ S 5 ) ∨ P 4 (H) ∨ P 5 (T ) ∨ P 6 (H).
i=1 i=1

Lemma 2.5. Let Z be the homotopy cofibre of the composition

d
_
Σf6 qZ
hZ : S 7 −→ ΣM(5) −→ S 4 ∨ P 4 (H) ∨ P 5 (T ),
i=1

where qZ is the obvious projection under the homotopy equivalence of ΣM(5) in (7). After localization
away from 2, there is a homotopy equivalence
r
_ d
_
3 6
ΣM ≃ ( 12 ) (S ∨ S ) ∨ S 5 ∨ P 6 (H) ∨ Z.
i=1 i=1

Proof. Consider the homotopy commutative diagram

ΣM(4)
<
g6
Σι4

Σf6  Σι6
S7 ❅ / ΣM(6) / ΣM
(8) ❅❅
❅❅ Σf6′
❅❅ Σκ6
❅❅

Wr
S6,
i=1

where the second row is the suspension of the cofibration (7), κ6 is the quotient map obtained by
extending the cofibration (5) to the right, f6′ := κ6 ◦ f6 by definition, and the map g6 will be defined
momentarily.
W
r L
r
Indeed, by the Hilton-Milnor Theorem [Whi, Chapter XI. 6] we have π7 ( S6) ∼
= π7 (S 6 ) ∼
=
i=1 i=1
L
r W
r
′
Z/2. Thus Σf6∗ is null homotopic away from 2. Note that ΣM(4) is 2-connected and S 6 is
i=1 i=1
5-connected. By the Blakers-Massey theorem (Lemma 6.1), the homotopy cofibration in the middle
column of Diagram (8) is a homotopy fibration up to degree 7. Hence, the map Σf6 can be lifted to
a map g6 to ΣM(4) . We have completed the construction of the diagram.
g6 ≃ Wr W
d W
d
Consider the composition map g6 : S 7 −→ ΣM(4) −→ ( S 3 ) ∨ ( S 4 ) ∨ ( S 5 ) ∨ P 4 (H) ∨
i=1 i=1 i=1
P 5 (T ) ∨ P 6 (H) by abuse of notation, where the second homotopy equivalence is from Lemma 2.3.
Since π7 (S 3 ) ∼
= Z/2, π7 (S 5 ) ∼
= Z/2 [Tod], and π7 (P 6 (pr ))( 12 ) = 0 by [Hua3, Lemma 6.4], we see that
the component of g6 on each S 3 , S 5 , and each wedge summand of P 6 (H) is null homotopic, and so
is the composition Σf6 ≃ Σι4 ◦ g6 . Hence by Lemma 6.4, there is a homotopy equivalence away from
2
r
_ d
_ d
_

ΣM ≃( 12 ) (S 3 ∨ S 6 ) ∨ ( S 5 ) ∨ P 6 (H) ∨ ( S 4 ∨ P 4 (H) ∨ P 5 (T )) ∪hZ e8 ,
i=1 i=1 i=1
 SUSPENSION HOMOTOPY OF SIMPLY CONNECTED 7-MANIFOLDS 11

where hZ is the composition

r
_ d
_ d
_
Σf6 qZ
hZ : S 7 −→ (S 3 ∨ S 6 ) ∨ (S 4 ∨ S 5 ) ∨ P 4 (H) ∨ P 5 (T ) ∨ P 6 (H) −→ S 4 ∨ P 4 (H) ∨ P 5 (T )
i=1 i=1 i=1

with qZ the obvious projection. The lemma is proved. 

3. A refined decomposition

In Section 2, we have showed that for a simply connected closed 7-manifold M with homology of
the form (1), there is a homotopy equivalence after localization away from 2
r
_ d
_
3 6
ΣM ≃ ( 12 ) (S ∨ S ) ∨ S 5 ∨ P 6 (H) ∨ Z,
i=1 i=1

W
d
where Z is the homotopy cofibre of a map hZ : S 7 −→ S 4 ∨ P 4 (H) ∨ P 5 (T ). In this section, we
i=1
refine this decomposition by showing that Z is a suspension when H or T = 0. The method is to
kill the bottom cells of M , and then study the structure of the resulted complex.
n
3.1. 7-Manifolds with bottom cells killed. Suppose H = ⊕ Z/pti . By [N3] or [N2], we have
i=1
W
n
P 3 (H) ≃ P 3 (3ti ). Define the complex W by a homotopy cofibration
i=1
r
_ n
_
2 i π
(9) S ∨ S 2 −→
W W
M −→ W,
i=1 i=1

W
r W
n W
r
where iW is the inclusion of the bottom cells S2 ∨ S 2 ֒→ S 2 ∨ P 3 (H) ≃ M(2) ֒→ M . By
i=1 i=1 i=1
Theorem 2.1, the homology decomposition of W can be described as follows:
n
_ d
_
W(3) ≃ S3 ∨ S 3 ∨ P 4 (T ),
i=1 i=1
d
_ fW ιX
S 3 ∨ P 4 (H) −→
3 3
W(3) −→ W(4) ,
i=1
_r
fW ιX
S 4 −→
4 4
W(4) −→ W(5) ,
i=1
fW ιX
S 6 −→
6 6
W(5) = W(6) −→ W.

This homology decomposition is trivial up to the 5-th stage:

Lemma 3.1. After localization away from 2, there is a homotopy equivalence

n
_ d
_ r
_
W(5) ≃( 12 ) S3 ∨ (S 3 ∨ S 4 ) ∨ S 5 ∨ P 4 (T ) ∨ P 5 (H).
i=1 i=1 i=1

W
d W
n W
d
Proof. Consider f3W : S 3 ∨ P 4 (H) −→ S3 ∨ S 3 ∨ P 4 (T ). By the Hilton-Milnor theorem
i=1 i=1 i=1
and Lemma 6.6 it has no Whitehead product as component. Then since f3W is trivial on H3 , its
restriction from each S 3 is null homotopic, and the component of its restriction from P 4 (H) to
12 RUIZHI HUANG AND PENGCHENG LI

P 4 (T ) is null homotopic by Lemma 6.12. Also by Lemma 6.11, [P 4 (H), S 3 ]( 21 ) = 0. Hence, f3W is
Wn W
d W
d
null homotopic, and W(4) ≃ S3 ∨ S 3 ∨ P 4 (T ) ∨ S 4 ∨ P 5 (H).
i=1 i=1 i=1
W
r
4
W
n
3
Wd
3 4
Wd
Consider f4W : S −→ S ∨ S ∨ P (T ) ∨ S 4 ∨ P 5 (H). Again by the Hilton-Milnor
i=1 i=1 i=1 i=1
theorem and Lemma 6.6 it has no Whitehead product as component. Then since f4W is null ho-
W
d
motopic on H4 , its component on S 4 ∨ P 5 (H) is null homotopic. Also as π4 (S 3 ) ∼
= Z/2 and
i=1
4
W
n W
d
r
π4 (P (p ))( 12 ) = 0 by [Hua3, Lemma 6.3], its component on S3 ∨ S 3 ∨ P 4 (T ) is null homo-
i=1 i=1
topic. Hence, f4W is null homotopic and the lemma is proved. 

The nice structure of W allows us to show that the attaching map of Z in Lemma 2.5 is almost
a suspension.

Lemma 3.2. For the complex Z in Lemma 2.5, its attaching map hZ satisfies that
d
_ d
_ n
_
h id∨q∨id
S 7 −→
Z
S 4 ∨ P 4 (H) ∨ P 5 (T ) −→ S4 ∨ S 4 ∨ P 5 (T )
i=1 i=1 i=1

a suspension map, where q is the obvious pinch map of Moore space.

Proof. By the definition of W (9), it is clear that there is a diagram of homotopy cofibrations

f6
S6 / M(5) / M

πW (5) πW

f6W  
S6 / W(5) / W,

W
r W
n
where both πW and πW (5) kill the bottom cells S2 ∨ S 2 . Let XW be the homotopy cofibre of
i=1 i=1
the composition map
d
_ n
_
6 6 fW qW
hW 6
6 : S −→ W(5) −→ S3 ∨ S 3 ∨ P 4 (T ),
i=1 i=1

where q6W is the obvious projection under the homotopy equivalence in Lemma 3.1. There is a
diagram of homotopy cofibrations

W
r W
d
(S 3 ∨ S 6 ) ∨ S 5 ∨ P 6 (H) / ΣM / Z
i=1 i=1

π3 ΣπW πX ′

W
r W
d  
S6 ∨ S 5 ∨ P 6 (H) / ΣW / ΣXW ,
i=1 i=1

where the top and bottom homotopy cofibrations are naturally induced from Lemma 2.5 and 3.1, π3
W
r
is the obvious projection to kill S 3 , and πX ′ is the induced map. The restriction of the diagram
i=1
 SUSPENSION HOMOTOPY OF SIMPLY CONNECTED 7-MANIFOLDS 13

on the 5-stage of homology decomposition is simply a diagram of natural projections

W
r W
d
/ W S 4 ∨ P 4 (H) ∨ P 5 (T )
d
(S 3 ∨ S 6 ) ∨ (S 4 ∨ S 5 ) ∨ P 4 (H) ∨ P 5 (T ) ∨ P 6 (H)
i=1 i=1 i=1

ΣπW (5) πX ′ (5) =id∨q∨id

 
W
r W
d W
n W
d W
n
S6 ∨ (S 4 ∨ S 5 ) ∨ S 4 ∨ P 5 (T ) ∨ P 6 (H) / S4 ∨ S 4 ∨ P 5 (T ),
i=1 i=1 i=1 i=1 i=1

W
n
where the map πX ′ (5) kills the bottom cells S 3 of P 4 (H) in Z. The above together shows that
i=1
there is a homotopy commutative diagram of homotopy cofibrations

hZ W
d
S7 / S 4 ∨ P 4 (H) ∨ P 5 (T ) / Z
i=1

id∨q∨id

ΣhW
6 W
d W
n 
S7 / S4 ∨ S 4 ∨ P 5 (T ) / ΣXW .
i=1 i=1

It particularly proves the lemma. 

3.2. Z is a suspension when H or T = 0. For any finite generated abelian torsion group A,
denote A = A3 ⊕ A6=3 , where A3 consists of the 3-primary torsion of A, and A6=3 consists of other
torsions.

Lemma 3.3. Suppose H or T = 0. For the complex Z in Lemma 2.5, there exists a CW -complex
d
_
X ≃( S 3 ∨ P 3 (H3 ) ∨ P 4 (T3 )) ∪hX e7
i=1

for some attaching map hX such that

Z ≃( 21 ) ΣX ∨ P 4 (H6=3 ) ∨ P 5 (T6=3 ).

Proof. By Lemma 3.2, the attaching map hZ of Z satisfies that the composition
d
_ d
_ n
_
h id∨q∨id
S 7 −→
Z
S 4 ∨ P 4 (H) ∨ P 5 (T ) −→ S4 ∨ S 4 ∨ P 5 (T )
i=1 i=1 i=1

is a suspension. We claim that hZ itself is a suspension map. When H = 0, this is obviously true.
When T = 0, by the Hilton-Milnor theorem we may write hZ as

hZ = hH + ΣhS + ωH ,

where hH and ΣhS are the linear combinations of the components on summands of P 4 (H) and
Wd
S 4 respectively, and ωH consists of possible Whitehead products with at least one entry from
i=1
a wedge summand of P 4 (H). However, with the facts that π7 (P 7 (pr ))( 21 ) = π7 (P 6 (pr ))( 12 ) = 0 by
[Hua3, Lemma 6.3 and 6.4], it is easy to compute that ωH = 0. For the map hH , by Lemma 6.8
e
α
π7 (P 4 (pr )) = 0 if p ≥ 5, and π7 (P 4 (3r )) ∼
= Z/3{e e satisfies that the class of S 7 −→
α}, where α
14 RUIZHI HUANG AND PENGCHENG LI

q
P 4 (3r ) −→ S 4 is Σα, and α is detected by the reduced power P 1 . Thus even though we have not
computed π6 (P 3 (3r )), we still obtain that hH is a suspension. Combining the above argument, we
have showed that the attaching map hZ is a suspension map. In particular, Z is a suspension.
E
Moreover, since when p ≥ 5 π7 (P 4 (pr )) = 0 and the suspension morphism π6 (P 4 (pr )) −→
∗

π7 (P 5 (pr )) is trivial by [Hua3, Lemma 6.6], we see that the suspension map hZ factors as
d
_ d
_
Σh
hZ : S 7 −→
X
S 4 ∨ P 4 (H3 ) ∨ P 5 (T3 ) ֒→ S 4 ∨ P 4 (H) ∨ P 5 (T ),
i=1 i=1

where ΣhX is the restricted attaching map. Let X be the homotopy cofibre of hX . The lemma
follows easily. 

3.3. ΣZ in general case. In general case, the complex Z in Lemma 2.5 can be decomposed further
after suspension.

Lemma 3.4. For the complex Z in Lemma 2.5, there exists a CW -complex
d
_
Y ≃( S 4 ∨ P 4 (H3 ) ∨ P 5 (T3 )) ∪h e8
i=1

for some attaching map h such that

ΣZ ≃( 21 ) ΣY ∨ P 5 (H6=3 ) ∨ P 6 (T6=3 ).

Furthermore, when H or T = 0, ΣX ≃ Y .

Proof. Consider the suspension of the attaching map hZ of Z

d
_
ΣhZ : S 8 −→ S 5 ∨ P 5 (H) ∨ P 6 (T ).
i=1

Since when p ≥ 5, π8 (P 5 (pr )) = π8 (P 6 (pr )) = 0 by Lemma 6.9 and 6.7, we see that the suspension
map hZ factors as
d
_ d
_
Σh
ΣhZ : S 8 −→ S 5 ∨ P 5 (H3 ) ∨ P 6 (T3 ) ֒→ S 5 ∨ P 5 (H) ∨ P 6 (T ),
i=1 i=1

where Σh is the restricted attaching map. Let Y be the homotopy cofibre of h. The lemma follows
easily. 

Combining Lemma 2.5 with Lemma 3.4 and 3.3, we obtain the following proposition immediately.

Proposition 3.5. Let M be a simply connected closed 7-manifold with homology of the form (1).
There is a homotopy equivalence after localization away from 2
r
_ d
_
Σ2 M ≃( 12 ) (S 4 ∨ S 7 ) ∨ S 6 ∨ P 5 (H6=3 ) ∨ P 6 (T6=3 ) ∨ P 7 (H) ∨ ΣY,
i=1 i=1
 SUSPENSION HOMOTOPY OF SIMPLY CONNECTED 7-MANIFOLDS 15

W
d
where Y ≃ ( S 4 ∨ P 4 (H3 ) ∨ P 5 (T3 )) ∪h e8 for some attaching map h. Further, if H or T = 0 there
i=1
is a homotopy equivalence after localization away from 2
r
_ d
_
ΣM ≃( 12 ) (S 3 ∨ S 6 ) ∨ S 5 ∨ P 4 (H6=3 ) ∨ P 5 (T6=3 ) ∨ P 6 (H) ∨ ΣX,
i=1 i=1

W
d
where X ≃ ( S 3 ∨ P 3 (H3 ) ∨ P 4 (T3 )) ∪hX e7 for some attaching map hX , and ΣX ≃ Y . 
i=1

4. A decomposition via Steenrod module structure

In this section, we decompose the complexes ΣY and ΣX in Proposition 3.5 based on their
Steenrod module structures.

W
d
4.1. A decomposition of ΣY. Recall by Proposition 3.5, Y ≃ ( S 4 ∨ P 4 (H3 ) ∨ P 5 (T3 )) ∪h e8 .
i=1
Let
m d+m ℓ+d+m
H3 = ⊕ Z/3µi , Z= ⊕ Zj , T3 = ⊕ Z/3νk ,
i=1 j=1+m k=1+d+m

where 1 ≤ µ1 ≤ · · · ≤ µm , ν1+m+d ≥ · · · ≥ νℓ+m+d ≥ 1, and each Zj is a copy of Z. By [N3]

W
m W
ℓ+d+m
or [N2], we have P 4 (H3 ) ≃ P 4 (3µi ), P 5 (T3 ) ≃ P 5 (3νk ). Note that the canonical order
i=1 k=1+d+m
on the subset {1, · · · , ℓ + d + m} ⊆ N of natural numbers induces a partial order on the index set
d+m ℓ+d+m
{µi }m
i=1 ∪ {j}j=1+m ∪ {νk }k=1+d+m , and then on the wedges summands of the lower skeleton of Y

and their 5-th cohomology groups with coefficient Z/3. Hence we can write
ℓ+d+m
_ m
_ d+m
_ ℓ+d+m
_

(10) Y ≃ Yt ∪h e8 ≃ ( P 4 (3µi )) ∨ ( Sj4 ) ∨ ( P 5 (3νk )) ∪h e8 ,
t=1 i=1 j=1+m k=1+d+m

where Yt is the t-th wedge summand under the partial order.

Recall there is the classical Steenrod reduced power P 1 : H 4 (Y ; Z/3) −→ H 8 (Y ; Z/3). It is
known that π6 (S 3 )( 12 ) ∼
= Z/3{α} and the stable map α is detected by P 1 [Har, Section 1.5.5].
By Lemma 6.7 π8 (P 6 (3ν )) ∼ = Z/3{i ◦ Σ2 α} with i the bottom cell inclusion, and by Lemma 6.9
e
α q
π8 (P 5 (3µ )) ∼ α} such that the composition S 7 −→ P 4 (3µ ) −→ S 4 is Σα.
= Z/3{Σe
Define a P 1 -index of Y by

IndP 1 (Y ) := min{t | P 1 (x) 6= 0, x ∈ H 5 (Yt ; Z/3), 1 ≤ t ≤ ℓ + d + m},

if P 1 (H 5 (Y ; Z/3)) 6= 0.

Lemma 4.1. For the complex Y (10), we have the following:

• if P 1 = 0, then there is a homotopy equivalence

d
_
ΣY ≃( 12 ) S 5 ∨ P 5 (H3 ) ∨ P 6 (T3 ) ∨ S 9 ;
i=1
16 RUIZHI HUANG AND PENGCHENG LI

• if P 1 6= 0 and IndP 1 (Y ) = i ≤ m, then there is a homotopy equivalence

d
_ H3
ΣY ≃( 21 ) S5 ∨ P 5( ) ∨ P 6 (T3 ) ∨ (P 5 (3µi ) ∪Σe 9
α e ),
i=1
Z/3µi

• if P 1 6= 0 and 1 + m ≤ IndP 1 (Y ) ≤ d + m, then there is a homotopy equivalence

d−1
_
ΣY ≃ ( 12 ) S 5 ∨ P 5 (H3 ) ∨ P 6 (T3 ) ∨ (S 5 ∪Σ2 α e9 ),
i=1

• if P 1 6= 0 and IndP 1 (Y ) = k ≥ 1 + d + m, then there is a homotopy equivalence

d
_ T3
ΣY ≃( 21 ) S 5 ∨ P 5 (H3 ) ∨ P 6 ( ) ∨ (P 6 (3νk ) ∪i◦Σ2 α e9 ),
i=1
Z/3νk

Proof. There is a chain of standard maps

3 3 3 3 q i 3 3 3
P 5 (3) → · · · → P 5 (3s ) → P 5 (3s+1 ) → · · · → S 5 → · · · P 6 (3r+1 ) → P 6 (3r ) → · · · −→ P 6 (3),

where each 3 is of degree 3 on the bottom cell and of degree 1 on the top cell, each 3 is of degree 3 on
the top cell and of degree 1 on the bottom cell, and q and i are the obvious pinch and inclusion maps.
This chain defines a partial order on all the involved Moore spaces, which is consistent with the partial
order for the wedge summands Yt of Y . Recall π8 (P 5 (3µ )) ∼ = Z/3{Σe α}, π8 (S 5 )(3) ∼
= Z/3{Σ2 α}, and
π8 (P 6 (3ν )) ∼
= Z/3{i ◦ Σ2 α}, and notice that all the three generators can be desuspended. It implies
that each map in the chain induces an isomorphism on π8 at prime 3 and preserves the assigned
generators. Therefore, for the wedge summands of ΣY there are compatible maps

fa,b : ΣYa −→ ΣYb

whenever a < b such that fa,b∗ : π8 (Ya )(3) → π8 (Yb )(3) is an isomorphism of stable homotopy groups
preserving the generators.
We may apply the matrix method in [Hua2, Section 3] to simplify the attaching map
ℓ+d+m
_
Σh : S 8 −→ ΣYt
t=1

of the top cell of ΣY . With the ordering on Yt and the previous discussion, we can represent the map
Σh by a row vector c = (c1 , · · · , cℓ+d+m )τ over the field Z/3, where ct represents the component of
Σh on the t-th wedge summand ΣYt .
Let us consider a typical transformation as follows. Suppose a < b. Consider the composition
map
µ′ ∨id id∨Σfa,b ∨id id∨∇
Ea,b : ΣYa ∨ ΣYb −→ ΣYa ∨ ΣYa ∨ ΣYb −→ ΣYa ∨ ΣYb ∨ ΣYb −→ ΣYa ∨ ΣYb ,

where µ′ is the standard co-multiplication, and ∇ is the folding map. It can be viewed as rank 2
square matrix  
1 0
Ea,b =  .
Σfa,b 1
 SUSPENSION HOMOTOPY OF SIMPLY CONNECTED 7-MANIFOLDS 17

In particular, Ea,b induces an isomorphism on homology, and then is a homotopy equivalence by the
Whitehead theorem. We can form the diagram of homotopy cofibrations

(ca ,cb )τ
S8 / ΣYa ∨ ΣYb / ΣYa,b

Ea,b ≃ φ

Ea,b ◦(ca ,cb )  τ 

S8 / ΣYa ∨ ΣYb / Y′ ,
a,b

′
where ΣYa,b and Ya,b are defined by the homotopy cofibrations, and φ is a homotopy equivalence
away from 2 by the five lemma and the Whitehead theorem. By the definition of fa,b , Ea,b ◦(ca , cb ) ≃
(ca , ca + cb ), which is an elementary row addition operation on (ca , cb ) with a < b. Hence, on the
vector representation c = (c1 , · · · , cℓ+d+m )τ of Σh, we are allowed to add ca to cb whenever a < b
without changing the homotopy type of ΣY . It is clear we can also apply row multiplication by −1.
Recall the class α ∈ π6 (S 3 ) is detected by P 1 . Let t = IndP 1 (Y ). It means that ct 6= 0 but cw = 0
for any w ≤ t − 1. Hence, by allowable operations we can transform c to be c′ = (0, · · · , 0, c′t =
1, 0, · · · 0) such that
_
ΣY ≃( 12 ) ΣYw ∨ Σ(Yt ∪c′t e8 ).
w6=t

The lemma can be proved by writing down the three different cases of Yt separately. 

4.2. A decomposition of ΣX. For the case when H or T = 0, recall by Proposition 3.5, X ≃
Wd
( S 3 ∨ P 3 (H3 ) ∨ P 4 (T3 )) ∪hX e7 , and ΣX ≃ Y . For the suspended attaching map ΣhX , it is known
i=1
that π6 (S 3 )( 12 ) ∼
= Z/3{α} is stable, π7 (P 4 (3µ )) ∼
= Z/3{e
α} by Lemma 6.8 such that the composition
e
α q E
S 7 −→ P 4 (3µ ) −→ S 4 is Σα, and the suspension image π6 (P 4 (3ν )) −→
∗
π7 (P 5 (3ν )) is generated by
i ◦ Σα by [CS, Lemma 3.1]. Hence, the argument for proving Lemma 4.1 can be directly applied to
prove the following lemma for ΣX ≃ Y .

Lemma 4.2. When H or T = 0, for the complex Y ≃ ΣX (10), we have the following:

• if P 1 = 0, then there is a homotopy equivalence

d
_
ΣX ≃ ( 21 ) S 4 ∨ P 4 (H3 ) ∨ P 5 (T3 ) ∨ S 8 ;
i=1

• if P 1 6= 0 and IndP 1 (Y ) = i ≤ m, then there is a homotopy equivalence

d
_ H3
ΣX ≃( 21 ) S4 ∨ P 4( ) ∨ P 5 (T3 ) ∨ (P 4 (3µi ) ∪αe e8 ),
i=1
Z/3µi

• if P 1 6= 0 and 1 + m ≤ IndP 1 (Y ) ≤ d + m, then there is a homotopy equivalence

d−1
_
ΣX ≃( 21 ) S 4 ∨ P 4 (H3 ) ∨ P 5 (T3 ) ∨ (S 4 ∪Σα e8 ),
i=1
18 RUIZHI HUANG AND PENGCHENG LI

• if P 1 6= 0 and IndP 1 (Y ) = k ≥ 1 + d + m, then there is a homotopy equivalence

d
_ T3
ΣX ≃( 12 ) S 4 ∨ P 4 (H3 ) ∨ P 5 ( ) ∨ (P 5 (3νk ) ∪i◦Σα e8 ), 
i=1
Z/3νk

5. Proof of Theorem 1.1, 1.2, 1.3, 1.4 and 1.6

In this section, we prove the main theorems of the paper.

5.1. A family of building blocks. Before proving the main theorems, we would like to replace
certain terms in the homotopy decompositions of ΣY and Y in Lemma 4.1 and 4.2 by geometric
manifolds.
Following the notation of Crowley-Escher [CE], any pair of integers (ρ, γ) ∈ Z ⊕ Z ∼
= π3 (SO(4))
determines a fibre bundle

S 3 −→ Mρ,γ −→ S 4

over the 4-sphere with structure group SO(4) and the total manifold Mρ,γ . We are interested in the
case when γ = 3ν .

Lemma 5.1. For the total manifold Mρ,3ν with ρ ∈ Z, ν ≥ 1, its cell structure is of the form

Mρ,3ν ≃ P 4 (3ν ) ∪ e7 .

Moreover, we have the following:

• if 3 | ρ, there is a homotopy equivalence

ΣMρ,3ν ≃ P 5 (3ν ) ∨ S 8 ;

• if 3 ∤ ρ, there is a homotopy equivalence

ΣMρ,3ν ≃ P 5 (3ν ) ∪i◦Σα e8 .

Proof. By [CE] it is known that as a CW -complex

Mρ,3ν ≃ P 4 (3ν ) ∪g e7 ,

for some attaching map g. By [CS, Lemma 3.1], it is known that

π6 (P 4 (3ν )) ∼
= Z/3{i ◦ α} ⊕ Z/3ν {θ},

where the order 3 generator α : S 6 → S 3 is detected by P 1 , i : S 3 ֒→ P 4 (3ν ) is the bottom cell

inclusion, and the generator θ satisfies that Σθ = 0.
On the other hand, by the Wu formula [Wu] P 1 is trivial if and only if p1 (Mρ,3ν ) ≡ 0 mod 3. By
Milnor’s calculation [CE, Fact 3.1], the first Pontryagin class p1 (Mρ,3ν ) = 4ρ ∈ Z/3ν ∼
= H 4 (Mρ,3ν ; Z).
It follows that 3 | ρ if and only if p1 (Mρ,3ν ) ≡ 0 mod 3, and then if and only if P 1 is trivial.
 SUSPENSION HOMOTOPY OF SIMPLY CONNECTED 7-MANIFOLDS 19

Hence, if 3 | ρ, the attaching map g ≃ aθ for some a ∈ Z/3ν . Then Σg is null homotopic, and the
first suspension homotopy equivalence follows. If 3 ∤ ρ, g ≃ ±i ◦ α + a′ θ for some a′ ∈ Z/3ν . Then
Σg ≃ ±Σα is stable, and the second suspension homotopy equivalence follows. 

5.2. Proof of the theorems. We are ready to prove Theorem 1.1, 1.2, 1.3, 1.4 and 1.6, and we
start from the most general case.

Proof of Theorem 1.6. We know M has homology of the form (1). By Proposition 3.5 and (10),
there are homotopy equivalences
r
_ d
_
Σ2 M ≃( 21 ) (S 4 ∨ S 7 ) ∨ S 6 ∨ P 5 (H6=3 ) ∨ P 6 (T6=3 ) ∨ P 7 (H) ∨ ΣY,
i=1 i=1
ℓ+d+m
_ m
_ d+m
_ ℓ+d+m
_
8 4


Y ≃ Yt ∪h e ≃ ( µi
P (3 )) ∨ ( Sj4 ) ∨( P 5 (3νk )) ∪h e8 ,
t=1 i=1 j=1+m k=1+d+m

where 1 ≤ µ1 ≤ · · · ≤ µm and ν1+m+d ≥ · · · ≥ νℓ+m+d ≥ 1. In Lemma 4.1 the homotopy

decomposition of ΣY depends on P 1 and IndP 1 (Y ).
In the statement of Theorem 1.6, the homotopy equivalences are stated depending on the value
radp1 (M ). By the mod-3 Wu formula [Wu]

P 1 (x) = (p1 (M ) mod 3) ∪ x,

for any x ∈ H 3 (M ; Z/3). In particular, P 1 is trivial if and only if the cup product homomorphism
(p1 (M ) mod 3) ∪ (−) : H 3 (M ; Z/3) −→ H 7 (M ; Z/3) is trivial, and then if and only if p1 (M ) ≡
0 mod 3 by Poincaré duality.
Consider the canonical partial order

1 1 1 1 1 1
− ≤ ··· ≤ − ≤− ≤ 01+m ≤ · · · ≤ 0d+m ≤ ≤ ≤ ··· ≤ ,
µ1 µm−1 µm νℓ+d+m νℓ−1+d+m ν1+d+m

where 0j are d-copies of 0. From the definition of rad(x), the radius of a class x ∈ H 3 (M ; Z/3), it
gives a partial order on all the possible values of rad(x), which is isomorphic to the partial order on
its index set, and then to the order on Yt . In particular, to choose the minimal number of rad(x)
such that (p1 (M ) mod 3) ∪ x 6= 0 is equivalent to choose the minimal index t ∈ [1, ℓ + d + m] ∩ Z
such that P 1 (H 5 (Yt ; Z/3)) is non-trivial. In other words, if p1 (M ) 6≡ 0 mod 3 we have

radp1 (M ) = rad(x)

for some x ∈ H 3 (M ; Z/3), where the double suspension of x is a generator of H 5 (Yt ; Z/3) and
t = IndP 1 (Y ).
Hence, the conditions of Lemma 4.1 and Theorem 1.6 are equivalent. With the above homotopy
equivalences, the homotopy equivalences in the statement of the theorem can be obtained by applying
Lemma 4.1 with Lemma 5.1. This completes the proof of the theorem. 
20 RUIZHI HUANG AND PENGCHENG LI

Proof of Theorem 1.3 and 1.4. In this case, M has homology of the form (1) with H or T = 0. With
Proposition 3.5, Lemma 4.2 and 5.1, the two theorems can be proved by similar argument as in the
proof of Theorem 1.6. 

Theorem 1.1 and 1.2 are special cases of Theorem 1.3, 1.4 and 1.6.

Proof of Theorem 1.1. This special case follows immediately from Case (1) of Theorem 1.3, 1.4 and
1.6. 

Proof of Theorem 1.2. By the definition of generalized Aloff-Wallach manifold, M has the homology
of the form (1) such that r = 1, d = 0, H = 0 and T = Z/k with k = 3ν · t. Notice that in this case if
p1 (M ) 6≡ 0 mod 3, we have radp1 (M ) = ν1 . With this information, the theorem follows immediately
from Theorem 1.3. 

6. Technical lemmas

In this section, we provide the various technical results used in the paper.

6.1. The Blakers-Massey theorem. The following version of the Blakers-Massey theorem [BM]
answers the question that to what extent a homotopy cofibration is a homotopy fibration.

Lemma 6.1 ([BM]). Given a diagram of CW -complexes

FO ❅
❅❅
❅❅ i
g ❅❅
❅
f
A / X / Y,

where the second row is a homotopy cofibration, F is the homotopy fibre of f , A is k-connected and i
is m-connected (i.e., π∗ (i) is an isomorphism when ∗ < m and an epimorphism when ∗ = m). Then
the induced map g : A → F is (k + m)-connected, and there is a long exact sequence of homotopy
groups

πk+m (A) → πk+m (X) → πk+m (Y ) → πk+m−1 (A) → · · · → πk+1 (X) → πk+1 (Y ) → 0. 

6.2. Chang complexes. Recall that a finite CW -complex X is called an A2n -complex if X is (n−1)-
connected and dim(X) ≤ n+2. Denote by CX the cone of a space X. Denote by η = ηn : S n+1 → S n
the iterated suspension of the Hopf map η : S 3 → S 2 . Let P n+1 (pr ) be the Moore space defined as
the mapping cone of a degree map pr : S n → S n . In 1950 Chang classified A2n -complexes.

Lemma 6.2 ([Cha]). Let n ≥ 3. Every A2n -complex is homotopic equivalent to a wedge sum of the
Moore spaces S n+i (i = 0, 1, 2), P n+j (pr )(j = 1, 2) with p prime, and the following four types of the
 SUSPENSION HOMOTOPY OF SIMPLY CONNECTED 7-MANIFOLDS 21

fundamental Chang-complexes (r, t ≥ 1):

Cηn+2 = S n ∪η CS n+1 = Σn−2 CP 2 ,

Crn+2 = S n ∪(2r ,η) C(S n ∨ S n+1 ),

C n+2,t = (S n ∨ S n+1 ) ∪( ηt ) CS n+1 ,

2

Crn+2,t n
= (S ∨ S n+1
) ∪ 2r η
 C(S n ∨ S n+1 ).
0 2t

Moreover, the homotopy decomposition of an A2n -complex is unique up to the order of wedge sum-
mands. 

The above classification theorem implies the following immediately.

Lemma 6.3. Let n ≥ 3. After localization away from 2, every A2n -complex is homotopy equivalent
to a wedge sum of the Moore spaces

S n+i (i = 0, 1, 2), P n+j (pr ) (j = 1, 2, prime p ≥ 3, r ≥ 1).

Moreover, the homotopy decomposition of an A2n -complex is unique up to the order of wedge sum-
mands. 

6.3. A lemma on suspension decomposition. The following lemma is useful for suspension
decomposition; cf. [ST, Lemma 5.6].

Lemma 6.4. Let

f _ g
S −→ Ai ∨ B −→ ΣC
i
be a homotopy cofibration of simply connected CW -complexes. Suppose the component of the attach-
ing map f on each Ai is null homotopic. Then there is a homotopy equivalence
_
ΣC ≃ Ai ∨ D,
i

f W qB
where D is the homotopy cofibre of the composition S −→ Ai ∨ B −→ B with qB the obvious
i
projection.

Proof. Consider the diagram of homotopy cofibrations

f W g
S / Ai ∨ B / ΣC
i
qi hi

fi  gi 
S / Ai / Fi ,

where qi is the obvious projection onto Ai , Fi is the homotopy cofibre of fi = qi ◦ f , and gi and hi
are the induced maps. By the condition fi is null homotopic, and thus Fi ≃ Ai ∨ ΣS. In particular,
22 RUIZHI HUANG AND PENGCHENG LI

gi has a left homotopy inverse ̺i . It means that the restriction of g on each Ai has a left homotopy
inverse ̺i ◦ hi . Define the composition
W W
_ hi _ ̺i _
µ′ i i
̺ : ΣC −→ ΣC −→ Fi −→ Ai ,
i i i

where µ′ is the iterated comultiplication of ΣC. From the above, we see that ̺ is a left homotopy
W iA W g
inverse of gA : Ai −→ Ai ∨ B −→ ΣC, where iA is the obvious inclusion. Consider the diagram
i i
of homotopy cofibrations

W gA
/ ΣC
hD ′
/ D′
Ai
i ✏
✏✏ ✏
iA
✏✏
f W  g ✏✏ ✏
S / Ai ∨ B / ΣC ✏ φ
i ✏✏
qB
✏✏ ✏
hD ✏
qB ◦f  gD  ✏✏
S / B / D,

where D is the homotopy cofibre of qB ◦ f by definition with the induced maps gD and hD , D′ is the
homotopy cofibre of gA with the induced map hD′ , and the homotopy extension φ of hD exists since
qB ◦ iA is null homotopic. From the diagram, it is easy to see that φ induces an isomorphism on
homology, and hence is a homotopy equivalence by the Whitehead theorem. Further, since ̺ ◦ gA is
homotopic to the identity map, the top homotopy cofibration splits to give a homotopy equivalence
_ _
ΣC ≃ Ai ∨ D′ ≃ Ai ∨ D.
i i

The lemma is proved. 

6.4. Odd primary Moore spaces. Consider the homotopy fibration

q
(11) F 2n+1 {pr } −→ P 2n+1 (pr ) −→ S 2n+1 ,

where q is the pinch map of the bottom cell. Cohen-Moore-Neisendorfer proved the following the
famous decomposition theorem.

Theorem 6.5. [CMN, N] Let p be an odd prime. There is a p-local homotopy equivalence
∞
Y _
k
ΩF 2n+1 {pr } ≃(p) S 2n−1 × S 2p n−1
{pr+1 } × ΩΣ P nα (pr ),
k=1 α
W
where S i {pr } is the homotopy fibre of the degree map pr : S i → S i , and P nα (pr ) is an infinite
α
bouquet of mod pr Moore spaces, with only finitely many Moore spaces in each dimension and the
least value of nα being 4n − 1. 

We also need the following classical result.

 SUSPENSION HOMOTOPY OF SIMPLY CONNECTED 7-MANIFOLDS 23

Lemma 6.6. [N2, Proposition 6.2.2] Let p be an odd prime. Then there is a homotopy equivalence

P m (pr ) ∧ P n (ps ) ≃ P m+n (pmin(r,s) ) ∨ P m+n−1 (pmin(r,s) ).

Further, if q1 and q2 are distinct primes then P m (q1r ) ∧ P n (q2s ) is contractible. 

Lemma 6.7. When n ≥ 6,

πn+2 (P n (pr )) = 0, for p ≥ 5, πn+2 (P n (3r )) ∼

= Z/3{i ◦ Σn−4 α},

where α ∈ π6 (S 3 )(3) ∼
= Z/3 is the generator [Tod], and i : S n−1 ֒→ P n (3r ) is the bottom cell
inclusion.

Proof. The statement for p ≥ 5 is proved by [Hua3, Lemma 6.5]. Let p = 3. By the Freudenthal
suspension theorem, it suffices to show π9 (P 7 (3r )) = Z/3{i ◦ Σ3 α}.
×3r i
Consider the cofibration S 6 −→ S 6 −→ P 7 (3r ). By the Blakers-Massey theorem (6.1), it is a
homotopy fibration up to degree 10. Hence, there is an exact sequence

×3r i
Z/3 ∼
= π9 (S 6 )(3) −→ π9 (S 6 )(3) −→ π9 (P 7 (3r ))(3) −→ π8 (S 6 )(3) = 0,
∗

where ×3r is trivial. Hence π9 (P 7 (3r ))(3) ∼

= π9 (S 6 )(3) ∼
= Z/3{Σ3 α}. Since P 7 (3r ) is contractible
at any prime p 6= 3 and rationally, π9 (P 7 (3r )) ∼
= π9 (P 7 (3r ))(3) ∼
= Z/3{i ◦ Σ3 α} as required. This
completes the proof of the lemma. 

Lemma 6.8.

π7 (P 4 (pr )) = 0, for p ≥ 5, π7 (P 4 (3r )) ∼

= Z/3{e
α},
e
α q
e satisfies that the class of S 7 −→ P 4 (3r ) −→ S 4 is Σα.
where α

Proof. For an odd prime p, by [CMN, N] there is a p-local homotopy equivalence

∞
_
4 3
r
ΩP (p ) ≃(p) S {p } × Ω( r
P 2k+7 (pr )).
k=0

It follows that

π7 (P 4 (pr )) ∼
= π6 (S 3 {pr }) ⊕ π7 (P 7 (pr )) = π6 (S 3 {pr })

by [Hua3, Lemma 6.3]. Consider the long exact sequence of p-primary homotopy groups

×pr
· · · −→ 0 = π6 (ΩS 3 )(p) −→ π6 (S 3 {pr })(p) −→ π6 (S 3 )(p) −→ π6 (S 3 )(p) −→ · · · .

When p ≥ 5, π6 (S 3 )(p) = 0 and hence π6 (S 3 {pr })(p) = 0. Then π7 (P 4 (pr )) ∼

= π7 (P 4 (pr ))(p) = 0, and
the lemma is proved in this case. When p = 3, π6 (S 3 )(3) ∼= Z/3, and the map ×3r is trivial. Then
the exact sequence implies that π6 (S 3 {3r })(3) ∼
= Z/3. Hence π6 (ΩP 4 (3r )) ∼
= π6 (ΩP 4 (3r ))(3) ∼
= Z/3.
Its generator is described in Lemma 6.9 below. This completes the proof of the lemma. 
24 RUIZHI HUANG AND PENGCHENG LI

Lemma 6.9.
π8 (P 5 (pr )) = 0, for p ≥ 5, π8 (P 5 (3r )) ∼
= Z/3{Σe
α},
e
α q
e satisfies that the class of S 7 −→ P 4 (3r ) −→ S 4 is Σα.
where α

Proof. Consider the homotopy fibration (11) with n = 2. Theorem 6.5 implies the homotopy equiv-
alence
∞
Y _
k
ΩF 5 {pr } ≃(p) S 3 × S 4·p −1
{pr+1 } × ΩΣ P nα (pr ),
k=1 α
W nα r
where the bouquet (p ) is explicitly described in [CMN, Notation before Theorem 1.6]. By
P
α W
straightforward calculation, it can be showed that the two bottom pieces of P nα (pr ) are P 7 (pr )
α
and P 11 (pr ). Then as π7 (S 3 ) ∼
= Z/2, π6 (S 3 )(3) ∼
= Z/3{α} [Tod] and π7 (P 7 (pr )) = 0 by [Hua3,
Lemma 6.3], the Hilton-Milnor theorem implies that

π7 (ΩF 5 {pr })(p) ∼

= π7 (S 3 )(p) ⊕ π7 (P 7 (pr )) = 0.

There is a homotopy commutative diagram

×pr q
S3 / S3 / P 4 (pr ) / S4

E j1
  q
∂
ΩS 4 / F 4 {pr } / P 4 (pr ) / S4

ΩE j2 E E
   Ωq

Ω∂
Ω2 S 5 / ΩF 5 {pr } / ΩP 5 (pr ) / ΩS 5 ,

where the first row is the homotopy cofibre sequence for P 4 (pr ), the second and third rows are
the homotopy fibre sequence induced from the pinch maps q with ∂ the connecting maps, E is the
suspension map, j1 is the inclusion of the bottom cell, and j2 is the induced map. It induces a
diagram of homotopy groups
×pr
π6 (S 3 )(p) / π6 (S 3 )(p)

E∗ j1∗

(Ωq)∗
 ∂∗

π6 (ΩP 4 (pr ))(p) / π6 (ΩS 4 )(p) / π6 (F 4 {pr })(p)

(ΩE)∗ (ΩE)∗ j2∗

 2
(Ω q)∗
 (Ω∂)∗

0 = π6 (Ω2 F 5 {pr })(p) / π6 (Ω2 P 5 (pr ))(p) / π6 (Ω2 S 5 )(p) / π6 (ΩF 5 {pr })(p) ,

where the last two rows are exact. When p ≥ 5, π6 (Ω2 S 5 )(p) = 0. Then the exact sequence implies
that π8 (P 5 (pr ))(p) = 0, and thus π8 (P 5 (pr )) = 0. The first statement is proved. When p = 3,
consider the rightmost squares in the above diagram. It is known that both E∗ and (ΩE)∗ are
isomorphisms, while the morphism ×3r is trivial. It follows that both (Ω∂)∗ and ∂∗ are trivial.
Then the exact sequencew implies that (Ωq)∗ is surjective and (Ω2 q)∗ is an isomorphism. It means
 SUSPENSION HOMOTOPY OF SIMPLY CONNECTED 7-MANIFOLDS 25

that π8 (P 5 (3r )) ∼
= π8 (P 5 (3r ))(3) ∼
= π6 (ΩS 4 )(3) ∼
= Z/3{α}, and the generator comes from the top cell
and can be desuspended once as required. 

6.5. Homotopy groups with coefficients. Let A be a finitely generated abelian group. Recall
there is the universal coefficient theorem of homotopy groups with coefficients.

Lemma 6.10. [N2, Theorem 1.3.1 and Exercise (1)] Let n ≥ 2. For a complex X there is a natural
short exact sequence

0 −→ πn (X) ⊗ A −→ [P n (A), X] −→ TorZ (πn−1 (X), A) −→ 0.

Lemma 6.11. Let n ≥ 4. After localization away from 2, there are isomorphisms

[P n (A), S n−1 ]( 21 ) = 0, [P n (A), S n−2 ]( 21 ) = 0, [P n+1 (A), P n (pr )]( 12 ) = 0.

Proof. For the first statement, Lemma 6.10 implies a short exact sequence

0 −→ πn (S n−1 ) ⊗ A −→ [P n (A), S n−1 ] −→ TorZ (πn−1 (S n−1 ), A) −→ 0.

Since πn−1 (S n−1 ) ∼

= Z is free, TorZ (πn−1 (S n−1 ), A) = 0. Also πn (S n−1 ) ⊗ A = 0 with n ≥ 4 after
localization away from 2. Hence, [P n (A), S n−1 ]( 12 ) = 0.
For the second statement, Lemma 6.10 implies a short exact sequence

0 −→ πn (S n−2 ) ⊗ A −→ [P n (A), S n−2 ] −→ TorZ (πn−1 (S n−2 ), A) −→ 0.

Since πn−1 (S n−2 ) ∼

= Z when n = 4, and Z/2 when n ≥ 5, TorZ (πn−1 (S n−2 ), A) = 0 after localization
away from 2. Also πn (S n−2 ) ∼
= Z/2 with n ≥ 4. Hence, [P n (A), S n−2 ]( 12 ) = 0.
For the third statement, Lemma 6.10 implies a short exact sequence

0 −→ πn+1 (P n (pr )) ⊗ A −→ [P n+1 (A), P n (pr )] −→ TorZ (πn (P n (pr )), A) −→ 0.

However, by [Hua3, Lemma 6.3 and 6.4] πn (P n (pr )) = πn+1 (P n (pr )) = 0 with n ≥ 4 after localiza-
tion away from 2. Hence, [P n+1 (A), P n (pr )]( 12 ) = 0. 

Lemma 6.12. Let n ≥ 4. Let h : P n (A) −→ P n (pr ) be a map such that it is trivial on homology.
Then after localization away from 2, h is null homotopic.

Proof. We only need to prove the lemma when A = Z/q s with q prime. By Lemma 6.10 there is a
morphism of short exact sequences

0 / πn (P n (A)) ⊗ A / [P n (A), P n (A)] / TorZ (πn−1 (P n (A)), A) / 0

h∗ h∗ h∗
  
0 / πn (P n (pr )) ⊗ A / [P n (A), P n (pr )] / TorZ (πn−1 (P n (pr )), A) / 0.

Since πn (P n (A)) = πn (P n (pr )) = 0 by [Hua3, Lemma 6.3] after localization away from 2, the left-
most h∗ is trivial. Since h is trivial on homology, the morphism h∗ : πn−1 (P n (A)) → πn−1 (P n (pr ))
26 RUIZHI HUANG AND PENGCHENG LI

is trivial by the Hurewicz theorem. It follows that the rightmost h∗ is trivial. Hence the middle
h∗ : [P n (A), P n (A)] → [P n (A), P n (pr )] is trivial. In particular, [h] = h∗ [id] is trivial. 

References

[AW] S. Aloff and N. R. Wallach, An infinite family of distinct 7-manifolds admitting positively curved Riemannian
structures, Bull. Amer. Math. Soc. 81 (1975), 93-97. 1
[BM] A. L. Blakers and W. S. Massey, The homotopy groups of a triad II, Ann. of Math. 55 (1952), 192-201. 6.1,
6.1
[Cha] S. Chang, Homotopy invariants and continuous mappings, Proc. Roy. Soc. London Ser. A 202 (1950) 253-263.
6.2
[CMN] F. R. Cohen, J. C. Moore and J. A. Neisendorfer, Torsion in homotopy groups, Ann. Math. 109 (1979),
121-168. 6.5, 6.4, 6.4
[CE] D. Crowley and C. Escher, A classification of S 3 -bundles over S 4 , Differential Geom. Appl. 18 (2003), 363-380.
1, 5.1, 5.1
[CN] D. Crowley and J. Nordström, The classification of 2-connected 7-manifolds, Proc. Lond. Math. Soc. (3) 119
(2019), no. 1, 1-54. 1
[CS] T. Cutler and T. So, The homotopy type of a once-suspended 6-manifold and its applications, to appear in
Topol. Appl., 2022. arXiv:2109.12415. 1, 4.2, 5.1
[DT] S. K. Donaldson and R. P. Thomas, Gauge Theory in higher dimensions, from: “The geometric universe”,
(S. A. Huggett, L. J. Mason, K. P. Tod, S. T. Tsou, N. M. J. Woodhouse, editors), Oxford University Press
(1998) 31-47. 1
[Esc] J.-H. Eschenburg, New examples of manifolds with strictly positive curvature, Invent. Math. 66 (1982), 469-
480. 1
[FIC] Lectures and surveys on G2 -manifolds and related topics, including lectures from the Minischool and Workshop
on G2 -Manifolds held as part of the Major Thematic Program on Geometric Analysis at the Fields Institute,
Toronto, August 19-25, 2017, edited by Spiro Karigiannis, Naichung Conan Leung and Jason D. Lotay, Fields
Institute Communications, 84. Springer, New York, (2020), xxii+382 pp. 1
[Har] J. Harper, Secondary cohomology operations, Graduate Studies Math., vol. 49, Amer. Math. Soc. (2002). 1,
4.1
[Hat] A. Hatcher, Algebraic Topology, Cambridge University Press, 2002. 1, 2.1
[Hua1] R. Huang, Homotopy of gauge groups over non-simply connected five-dimensional manifolds, Sci. China Math.
64 (2021) no. 5, 1061-1092. 1
[Hua2] R. Huang, Homotopy of gauge groups over high-dimensional manifolds, Proc. Roy. Soc. Edinburgh Sect. A
152 (2022) no. 1, 182-208. 1, 4.1
[Hua3] R. Huang, Suspension homotopy of 6-manifolds, to appear in Alg. Geom. Topol., 2021. arXiv:2104.04994 1,
2.2, 2.3, 2.4, 3.1, 3.2, 6.4, 6.4, 6.4, 6.5, 6.5
[Hop] M. Hopkins, Algebraic Topology and Modular Forms, Plenary talk, ICM, Beijing, 2002. 1
[Kre1] M. Kreck, Surgery and duality, Ann. of Math. (2) 149 (1999), 707-754. 1
[Kre2] M. Kreck, On the classification of 1-connected 7-manifolds with torsion free second homology, J. Topol. 11
(2018), 720-738. 1
[KS] M. Kreck and S. Stolz, A diffeomorphism classification of 7-dimensional homogeneous Einstein manifolds with
SU (3) × SU (2) × U (1)-symmetry, Ann. of Math. 127 (1988), 373-388. 1
 SUSPENSION HOMOTOPY OF SIMPLY CONNECTED 7-MANIFOLDS 27

[Kru1] B. Kruggel, A homotopy classification of certain 7-manifolds, Trans. Amer. Math. Soc. 349 (1997), 2827-2843.
1
[Kru2] B. Kruggel, Kreck-Stolz invariants, normal invariants and the homotopy classification of generalised Wallach
spaces, Quart. J. Math. Oxford Ser. (2) 49 (1998), no. 196, 469-485. 1
[Kru3] B. Kruggel, Homeomorphism and diffeomorphism classification of Eschenburg spaces, Q. J. Math. 56 (2005),
no. 4, 553-577. 1
[MS] I. Membrillo-Solis, Homotopy types of gauge groups related to S 3 -bundles over S 4 , Topology Appl. 255 (2019),
56-85. 1
[N] J. A. Neisendorfer, 3-primary exponents, Math. Proc. Camb. Philos. Soc. 90 (1981), 63-83. 6.5, 6.4
[N2] J. A. Neisendorfer, Algebraic methods in unstable homotopy theory, New Mathematical Monograph: 12,
Cambridge (2010). 1, 2, 3.1, 4.1, 6.6, 6.10
[N3] J. A. Neisendorfer, Primary homotopy theory, Mem. Amer. Math. Soc. 232, 1980. 3.1, 4.1
[So] T. L. So, Homotopy types of gauge groups over non-simply-connected closed 4-manifolds, Glasgow Math. J.
61 (2) (2019), 349-371. 1
[ST] T. L. So and S. Theriault, The suspension of a 4-manifold and its applications, arXiv: 1909.11129, (2019). 1,
6.3
[The] S. Theriault, Odd primary homotopy decompositions of gauge groups, Algebr. Geom. Topol. 10 (2010), no. 1,
535-564. 1
[Tod] H. Toda, Composition methods in homotopy groups of spheres, Ann. Math. Studies 49, Princeton Univ. Press,
Princeton, N. J. (1962). 2.4, 6.7, 6.4
[Whi] G. Whitehead, Elements of homotopy theory, GTM 62, Berlin-Heidelberg, New York: Springer-Verlag, 1978.
2.4
[Wit] E. Witten, The index of the Dirac operator in loop space, in P.S. Landweber, ed., Elliptic Curves and Modular
Forms in Algebraic Topology (Proceedings, Princeton 1986), Lecture Notes in Math., 1326, pp. 161-181,
Springer, 1988. 1
[Wu] W.-T. Wu, On Pontrjagin classes II, IV, Acta Math. Sinica 4 (1954), 171-199; Acta Math. Sinica 5 (1955),
37-63. 1, 5.1, 5.2

Institute of Mathematics, Academy of Mathematics and Systems Science, Chinese Academy of Sciences,
Beijing 100190, China
Email address: huangrz@[Link]
URL: [Link]

Department of Mathematics, Southern University of Science and Technology, Shenzhen 518055, China
Email address: lipc@[Link]