Mapping degrees between spherical 3-manifolds
Daciberg Gonçalves
Dept. de Matemática - IME - USP, Caixa Postal 66.281 - CEP 05314-
970,
São Paulo - SP, Brasil; FAX: 55-11-30916183
[email protected]
,
Peter Wong
Department of Mathematics, Bates College, Lewiston,
ME 04240, U.S.A.; FAX: 1-207-7868331
[email protected]
and
Xuezhi Zhao
Department of Mathematics, Capital Normal University, Beijing 100048, China; FAX: 86-10-68900950
[email protected]
Abstract.
Let D(M,N) be the set of integers that can be realized as the degree of a
map between two closed connected orientable manifolds M and N of the
same dimension. For closed 3-manifolds with S3-geometry M and N,
every such degree degf≡degψ (∣π1(N)∣) where
0≤degψ<∣π1(N)∣ and degψ only
depends on the induced homomorphism ψ=fπ on the fundamental
group. In this paper, we calculate explicitly the set {degψ} when ψ is surjective and then we show how to
determine deg(ψ) for arbitrary
homomorphisms. This leads to the determination of the set D(M,N).
Key words and phrases:
3-manifolds, mapping degrees
2010 Mathematics Subject Classification:
Primary: 55M20; Secondary: 20E45
This work was initiated during the first and second authors’ visit to Capital Normal University February 16 - 23, 2013. The first author was supported in part by Projeto Temático Topologia Algébrica Geométrica e Diferencial 2012/24454-8. The third author was supported in part by the NSF of China (11431009).
1. Introduction
The study of mapping degree has a very long history in topology. In particular, the existence of degree one maps between manifolds is an important aspect in the classification of manifolds.
Given two closed connected orientable manifolds M and N of the same dimension, let D(M,N)={degf∣f:M→N} denote the possible values of mapping degrees of maps from M to N. While the determination of D(M,N) has been of great interest to topologists, in general, very little has been done concerning the computation of D(M,N). For example, the finiteness of the set D(M,N) has been studied by various authors (see e.g. [3, 4, 6, 7, 15, 1]) and the study of D(M):=D(M,M) for self-maps have also been investigated. We refer the reader to the survey paper [22] for the study of degree up to 2000, and to the papers [5, 18, 20] and [14] for mapping degrees on 3-manifolds that are most relevant to our work.
Prompted by our earlier work [13] on the fixed point theory of spherical 3-manifolds, in this paper we determine D(M,N) when M and N are closed 3-manifolds with S3 geometry.
2. Fundamental groups of spherical 3-manifolds
A spherical 3-manifold is an orbit space S3/G, where G is a finite group which acts freely on S3. Hence G is the fundamental group of S3/G. Each spherical 3-manifold is oriented by the canonical orientation of S3.
Proposition 2.1**.**
(cf. [19])
The fundamental group of any spherical 3-manifold is in the form of G or Zm×G, where G belongs to one of the groups listed below and (m,∣G∣)=1.
The binary tetrahedral group is denoted by T24∗={a,b∣a2=b3=(ab)3,a4=1}, which is actually a special case of the generalized tetrahedral group T8⋅3q′ with q=1, i.e. T24∗≅T8⋅31′. One such isomorphism from T8⋅31′ to T24∗ is given by b↦bab−1,a↦a,w↦ba−1.
If n is odd, then the generalized dihedral group D4n∗ is the same as the dicyclic group Dn⋅22′. If n=1, then Dn⋅2q′ is the cyclic group Z2q. If q=1, then Dn⋅2q′ is not the fundamental group of a 3-dimensional spherical manifold except for the case n=1. Thus, all groups in Table 2.1 are distinct.
Here are some more characteristic subgroups and quotients of these groups.
Next, we describe, for each group in Table 2.1 except for the cyclic groups, the conjugacy classes of elements, subgroups and quotients.
Lemma 2.2**.**
The set of conjugacy classes of D4n∗ is given as follows.
Proof.
Note that b(bka)b−1=bk+2a, abla−1=b−l, a(bka)a−1=b−ka.
∎
Lemma 2.3**.**
The non-trival subgroups and quotients of D4n∗ are listed as follows.
Proof.
The presentations of cyclic subgroups follow from Lemma 2.2. Since a2=bn, any non-cyclic subgroup H of D4n∗ must have the form of either ⟨bk,bla⟩ or ⟨bka,bla⟩. Note that bla is conjugate to either a or ba. There are four possibilities: ⟨bk,a⟩, ⟨bka,a⟩, ⟨bk,ba⟩ and ⟨bka,ba⟩.
Clearly, ⟨bka,a⟩=⟨bk,a⟩. Since b has order 2n, we may assume that k∣2n. If k2n is even, i.e. kn is an integer, then we have the presentation ⟨bk,a⟩={bk,a∣a2=(bk)kn=(abk)2,a4=1}. Thus ⟨bk,a⟩ is isomorphic to D4⋅kn∗ if kn is even; to Dkn⋅22′ if kn is odd. In the case that k2n is odd, we have that (n,k)=2k. Since a2=bn, the subgroup ⟨bk,a⟩=⟨bk,bn,a⟩=⟨b2k,a⟩ is isomorphic to Dk2n⋅22′ by the argument above.
Suppose that ⟨bk,a⟩ is normal. Then bab−1=b2a must lie in ⟨bk,a⟩. It follows that b2∈⟨bk,a⟩. Hence k must be 1 or 2. Note that ⟨b,a⟩ is the whole group D4n∗. Thus, k=2 is the unique non-trivial case that ⟨bk,a⟩ is normal.
The subgroup ⟨bk,ba⟩ is isomorphic to ⟨bk,a⟩, by using the automorphism given by a↦b−1a, b↦b. This situation is the same as that of ⟨bk,a⟩.
Note that ⟨bka,ba⟩=⟨bk−1,ba⟩. This is also the case we have already considered.
∎
Next, we consider the binary octahedral group O48∗ and the binary icosahedral group I120∗. The subgroups and quotients of these groups are well-known and the results can be found, for example, in [9, Appendix].
Lemma 2.4**.**
The subgroups of O48∗ are Z2,Z3,Z4,Z6,Z8, Q8, D3⋅22′, Q16 and T24∗.
The normal subgroups are Z2,Q8 and T24∗.
The only quotient which is the fundamental group of a spherical 3-manifold is Z2, which can be embedded in O48∗.
Lemma 2.5**.**
The set of conjugacy classes of O48∗ is given as follows.
Lemma 2.6**.**
The subgroups of I120∗ are Z2,Z3,Z4,Z5,Z6, Q8, Z10, D12′, D20′ and T24∗.
The only normal subgroup is Z2.
The only quotient is A5 which is not the fundamental group of a spherical 3-manifold.
Lemma 2.7**.**
The set of conjugacy classes of I120∗ is given as follows.
Lemma 2.8**.**
The set of conjugacy classes of T8⋅3q′ is given as follows.
Here, t=0,1,…,3q−1−1.
Proof.
Since bwb−1=b(ab)−1w=a−1w and awa−1=ab−1w=baw, two conjugate elements must lie in the same coset of [T8⋅3q′,T8⋅3q′]. Note that w3 commutes with all elements of T8⋅3q′. It is sufficient to consider the conjugate relations in cosets [T8⋅3q′,T8⋅3q′], [T8⋅3q′,T8⋅3q′]w and [T8⋅3q′,T8⋅3q′]w2.
(1) The coset [T8⋅3q′,T8⋅3q′]. Since [T8⋅3q′,T8⋅3q′] is isomorphic D4⋅2∗, from Lemma 2.2, we have that b∼b3, a∼b2a and ba∼b3a by conjugation induced by b and a. Note that waw−1=b and wbw−1=ab=b3a. Thus, the coset [T8⋅3q′,T8⋅3q′] contains three conjugacy classes: [1]={1}, [b]={b,b3,a,ba,b2a,b3a} and [b2]={b2}.
(2) The coset [T8⋅3q′,T8⋅3q′]w. Note that
[TABLE]
[TABLE]
These relations imply that
[TABLE]
Moreover,
[TABLE]
[TABLE]
They give no new conjugacy relations. Thus, the coset [T8⋅3q′,T8⋅3q′]w contains two conjugacy classes: [w] and [bw].
(3) The coset [T8⋅3q′,T8⋅3q′]w2. Note that
[TABLE]
[TABLE]
These relations imply that
[TABLE]
Thus, the coset [T8⋅3q′,T8⋅3q′]w2 contains two conjugacy classes: [w2] and [b2w2].
∎
Lemma 2.9**.**
The non-trival subgroups and quotients of T8⋅3q′ are listed as follows.
Proof.
Note that ⟨ws⟩=⟨w(s,3q)⟩. The first three rows come from those in Lemma 2.8, where 3r=(3q,3t).
Since both 3t+1 and 3t+2 are coprime to 3q, we obtain that
⟨w3t+1⟩=⟨w3t+2⟩=⟨w⟩. Thus, all
cyclic subgroups generated by the elements of order 3q listed
in Lemma 2.8 are conjugate. We obtain the row 4. This
fact follows from Sylow’s theorems. Clearly, these subgroups are not
normal.
Consider the elements of order 2⋅3q. Note that
(b2w)6s+1=b2w6s+1, (b2w)6s+5=b2w6s+5.
Since both 6s+1 and 6s+5 are relatively prime to 2⋅3q,
we have ⟨b2w⟩=⟨b2w6s+1⟩=⟨b2w6s+5⟩. Note that
(b2w)3q−(6s+2)=b2w−6s−3, (b2w)3q−(6s+4)=b2w−6s−4. Since both 3q−(6s+2) and 3q−(6s+4) are
relatively prime to 2⋅3q, we have ⟨b2w⟩=⟨b2w−6s−2⟩=⟨b2w−6s−4⟩. Since
b2w−6s−2=(b2w6s+2)−1 and b2w−6s−4=(b2w6s+4)−1, we obtain that ⟨b2w⟩=⟨b2w6s+2⟩=⟨b2w6s+4⟩. Thus, there are four
cyclic subgroups, all conjugate to each other.
Let us consider the non-cyclic subgroup H of T8⋅3q′. Note that the abelianization of T8⋅3q′=⟨w⟩≅Z3q. The abelianization H/[H,H] will be a subgroup of ⟨w⟩. We have the following three cases:
(1) If H/[H,H] is trivial then H lies in the commutator [T8⋅3q′,T8⋅3q′]=D4⋅2∗. By Lemma 2.3, any proper subgroup of D4⋅2∗ is cyclic, and hence H must be the commutator itself. This is the row 6 for r=q.
(2) If H/[H,H]=⟨w⟩, then by Lemma 2.8 and up to a conjugation, H must contain w or b2w. If b2w∈H, then (b2w)2=w2∈H. Since (2,3q)=1, we still have that w∈H. Since H is non-cyclic, H contains an element of the form xwk, where x is a non-trivial element in the commutator. Hence x=(xwk)w−k also lies in H. Note that either x or x2 is b2. If x=b2 is the unique non-trivial element in the commutator, then H=⟨b2,w⟩ would be cyclic. Thus, there must an element x in H with x2=b2. Thus, one of the six elements: x,wxw−1,w2xw−2,x3,x2wxw−1,x2w2xw−2 must be a. Note that ⟨a,w⟩=T8⋅3q′. Then H must be the whole group.
(3) If H/[H,H]=⟨w3r⟩ with 0<r<q, then H=H′×⟨w3r⟩ for some subgroup H′ of the commutator D4⋅2∗ since w3r lies in the center. By Lemma 2.3, any proper subgroup of D4⋅2∗ is cyclic, and hence H′ must be the commutator itself. This is the row 6 for 0<r<q.
∎
Lemma 2.10**.**
The set of conjugacy classes of Dn⋅2q′ with 2∣n is given as follows.
Proof.
Note that w(ukwl)w−1=u−kwl, and note that u(ukwl)u−1=uk+2wl if l is odd; and u(ukwl)u−1=ukwl if l is even. Moreover ukwl=uk+n′wl.
∎
Lemma 2.11**.**
The non-trival subgroups and quotients of Dn⋅2q′ with 2∣n are listed as follows.
Proof.
Consider cyclic subgroups. For the conjugacy classes of the form [w2t+1], since 2t+1 is coprime with the order 2q of w, we have that ⟨w2t+1⟩=⟨w⟩. By Lemma 2.10, we obtain the first row.
For the conjugacy classes of the form [w2t], we have that ⟨w2t⟩=⟨u2r⟩ for some r with 0<r<q. These are the second and third row.
For the conjugacy classes of the form [us], we have that ⟨us⟩=⟨uk⟩ for some k with k∣n, giving row 4 and 5.
For the conjugacy classes of the form [usw2t], we have that ⟨usw2t⟩=⟨ukw2r)⟩≅Zkn×Z2q−r for some k and r with k∣n and 0<r<q, which is a normal subgroup. We have the rows from 6 to 8, which are respectively the case k=1, k>1 and r=1, and k,r>1.
Assume that H is a non-cyclic subgroup of Dn⋅2q′, H must contains an element of the form usw2t+1. Up to a conjugation, we may have that w2t+1∈H. It follows that w∈H because w∈⟨w2t+1⟩. Since H is not cyclic, H contain some element of the form us. Let uk be the element in H such that k is the minimal positive integer. Thus, H=⟨uk,w⟩. This is the final row. If such a subgroup is normal, then uwu−1=u2w must lies in H, and hence u2∈H. Note that k is odd. We have that u(k,2)=u lies in H. Thus, H=⟨u,w⟩ is the whole group.
∎
2.1. Embedding G in SO(4)
By definition of spherical 3-manifolds, we may regard the fundamental groups of spherical 3-manifolds as discrete subgroups of the orientation-preserving isometry group SO(4) of S3.
Proposition 2.12**.**
Let G be the fundamental group of a spherical 3-manifold. If G is not cyclic, then the generators in the presentation of Proposition 2.1 can be identified as elements in SO(4) as follows.
(1) if G=Zm×D4n∗,
[TABLE]
(2) if G=Zm×O48∗,
[TABLE]
(3) if G=Zm×I120∗,
[TABLE]
(4) if G=Zm×T8⋅3q′,
[TABLE]
(5) if G=Zm×Dn⋅2q′,
[TABLE]
In all cases, the generator of Zm is given by v=\left(\begin{array}[]{cc}R(\frac{2\pi}{m})&0\\
0&R(\frac{2\pi}{m})\\
\end{array}\right).
Here, we use R(θ) to denote the matrix (block)
\left(\begin{array}[]{cc}\cos\theta&-\sin\theta\\
\sin\theta&\cos\theta\\
\end{array}\right).
From now on, all groups are subgroups of SO(4). By Olum’s results [16, 17], every homomorphism ψ:G1→G2 can be realized by a map f:S3/G1→S3/G2 with ψ=fπ. Moreover, if f and g induce the same homomorphism ψ, then degf≡degg mod ∣G2∣. Therefore, we have a well-defined mod ∣G2∣ integer associated with any homomorphism ψ which we denote by deg(ψ)∈Z∣G2∣.
It is known that the orbit spaces of the form S3/Zm are lens spaces. The homotopy types and homeomorphism type of 3-dimensional lens spaces depend on the Zm action, which can be regarded as embeddings of Zm in SO(4).
3. Lens paces covering spherical manifolds
The generalized 3-dimensional lens space L(m;r1,r2) is defined to be the orbit space S3/Zm, where Zm is the finite cyclic subgroup of SO(4), which is generated by
[TABLE]
Here r1 and r2 are relatively prime to m, but m is not necessarily prime. If we regard S3 as the unit sphere of complex space C2, the element c turns out to be the complex matrix
[TABLE]
By identifying the covering transformation group with the fundamental group, we call this element c the standard generator of π1(L(m;r1,r2)).
It is known that L(m;r1,r2) and L(m;r1′,r2′) are homeomorphic if and only if {±r2r1,±r1r2} and {±r2′r1′,±r1′r2′} are the same subset of Zm∗, the multiple group of invertible elements in Zm (see e.g. [2, Ch. V]). Usually, L(m;1,r) is written as L(m;r). Clearly, L(m;r1,r2)=L(m;1,r′) if r′=r1[r2](m)−1. Here, [r2](m)−1 denotes the inverse of the multiplication of Zm.
The lens spaces that can cover L(m;r1,r2) must be L(k;r1,r2) for some k∣m.
Example 3.1**.**
Consider the group O48∗ and H=⟨b⟩ the cyclic group of order 6 generated by b.
The eigenvalues of the matrix b are {e3πi,e−3πi,e3πi,e−3πi}, meaning either e3πi or e−3πi has multiplicity two. It follows that b is conjugate to
\left(\begin{array}[]{cc}R(\frac{\pi}{3})&0\\
0&R(\frac{\pi}{3})\\
\end{array}\right).
Thus, the corresponding lens space is L(6;1,1)
More generally, consider the group Zm×O48∗ and H=⟨vb⟩, the eigenvalues of the matrix vb are {e3(6+m)πi,e−3(6+m)πi,e3(6−m)πi,e−3(6−m)πi}. Thus, vb is conjugate to
\left(\begin{array}[]{cc}R(\frac{(6+m)\pi}{3})&0\\
0&R(\frac{(6-m)\pi}{3})\\
\end{array}\right).
The example above shows in general how one determines the lens space corresponding to a maximal cyclic subgroup. We summarize the results below.
Lemma 3.2**.**
Let H be a maximal cyclic subgroup of some G, where G is the fundamental group of some spherical 3-manifold. Then the lens spaces of the form S3/H are listed as follows:
Proof.
By Lemma 2.1, the generator of a cyclic group can be written as a matrix in SO(4). We can compute its eigenvalues for any given matrix. Note that the corresponding lens spaces are completely determined by the eigenvalues of the matrices corresponding to the generators. This proof is a straightforward computation.
∎
Lemma 3.3**.**
Let f:L(m1;r11,r12)→L(m2;r21,r22) be a map between lens spaces,
inducing a homomorphism fπ on fundamental group given by fπ(c1)=c2l, where c1 and c2 are respectively standard generators of π1(L(m1;r11,r12)) and π1(L(m2;r21,r22)). Then the degree of f is
[TABLE]
where [∗](m1)−1 denotes the inverse of multiplication of Zm1.
Proof.
Let us consider the case when fπ is surjective. Then we have that m2∣m1. We construct a map g~:S3→S3 by \tilde{g}\left(\begin{array}[]{c}r_{1}e^{\theta_{1}i}\\
r_{2}e^{\theta_{2}i}\end{array}\right)=\left(\begin{array}[]{c}r_{1}e^{k_{1}\theta_{1}i}\\
r_{2}e^{k_{2}\theta_{2}i}\end{array}\right).
Note that
[TABLE]
[TABLE]
In order to obtain that g~c1=c2lg~, it is sufficient to assume that
m1r11k1−m2r21l and m1r12k2−m2r22l are both integers. Thus, we may pick k1=lr21m2m1[r11](m1)−1 and k2=lr22m2m1[r12](m1)−1. It is obvious that g~ has degree
[TABLE]
The equality g~c1=c2lg~ also implies that g~ induces a map g:L(m1;r11,r12)→L(m2;r21,r22). Since g~ is a lifting of g with respect to the universal covering S3 of both L(m1;r11,r12) and L(m2;r21,r22), we have that deg(g~)m2=m1deg(g). Here m1 and m2 are respectively the degrees of the covering maps S3→L(m1;r11,r12) and S3→L(m2;r21,r22). Thus,
[TABLE]
Note that f and g have the same induced homomorphism on π1, which is given by c1↦c2l. By Olum results [16, 17], we have that deg(f)≡deg(g)(m2). Thus, we give the proof of deg(f) when fπ is surjective.
Now, we turn to the general case. The image Imfπ of fπ is the cyclic group generated by c2(l,m2), which is isomorphic to Z(l,m2)m2. Note that the covering space of L(m2,r21,r22) corresponding to the group ⟨c2(l,m2)⟩ is L((l,m2)m2,r21,r22). There must be a lifting f~ of f satisfying the following commutative diagram:
[TABLE]
Note that c2l=(c2(l,m2))(l,m2)l. Apply our previous argument to f~ which induces a surjection on π1, we have that
[TABLE]
Since deg(f)=deg(p)deg(f~)=(l,m2)deg(f~), we obtain our conclusion.
∎
It should be mentioned that the existence of the homomorphism c1→c2l forces that m1l≡0(m2). In other word, m2lm1 must be an integer.
Theorem 3.4**.**
The set of all degrees D(L(m1;r11,r12),L(m2;r21,r22)) of maps from L(m1;r11,r12) to L(m2;r21,r22) is:
[TABLE]
Proof.
For any map f between these two lens spaces, the homomorphism fπ on fundamental groups is given by c1→c2l. Note that m1l≡0(m2). It follows that l=j(m1,m2)m2 for j=0,1,…,(m1,m2)−1. This theorem holds by Lemma 3.3.
∎
Remark 3.1*.*
When we write a lens space L(m;r1,r2), its orientation is specified, inheriting the natural orientation of S3. Thus, by above theorem D(L(3;1,1),L(3;1,2))={3k,−1+3k∣k∈Z}, although L(3;1,1) and L(3;1,2) are homeomorphic.
4. Mapping degrees
In this section, we determine all mapping degrees between spherical 3-manifolds. Suppose f:M→N is a map between two such manifolds with G1=π1(M) and G2=π1(N). If G1=Zm then the degree is already given in the last section. For the other cases of G1, we only need to consider the case when the induced homomorphism fπ is surjective.
Lemma 4.1**.**
Let ψ:G1→G2 be a homomorphism and G∗ the quotient
ker(ψ)G1. Then
[TABLE]
where φ:G1→G∗ is the projection and ξ:G∗→Imψ
is the isomorphism given by φ(x)↦ψ(x).
Proof.
Let us consider the following three maps:
i) f:S2k+1/G1→S2k+1/G∗ be a map such that fπ=φ
ii) h:S2k+1/G∗→S2k+1/Imψ such that hπ=ξ
iii) p2:S2k+1/Imψ→S2k+1/G2.
It is easy to see that the homomorphism ψ is p2∘ξ∘φ, the composite of the three homomorphisms. Since the degree of the covering of p2 is ∣G∗∣∣G2∣
the formula follows.
∎
By using Lemma 4.1, since the degree of endomorphisms are almost listed in [5], we can compute the degree deg(ψ), where ψ:G1→G2 is a surjection and G2 is a subgroup of G1.
Lemma 4.2**.**
Each endomorphism of D4n∗ is conjugate to a composition of some endomorphisms in following table.
Proof.
Let ψ be an endomorphism on D4n∗.
By Lemma 2.2, up to a conjugation, we may assume ψ(a)=1,a,ba,bk.
(1) ψ(a)=1.
(1.1) ψ(b)=bs. Then ψ(a2)=1, ψ(bn)=bns and ψ((ab)2)=b2s. The relation a2=bn=(ab)2 yields that 0≡ns≡2s(2n). Thus, s≡0,n(2n). Both are in the table (ψ1 and ψ2).
(1.2) ψ(b)=bsa. Note that ψ((ab)2)=bsabsa=a2=1=ψ(a2). Thus, this case is impossible.
(2) ψ(a)=a.
(2.1) ψ(b)=bs. Then ψ(a2)=a2=bn, ψ(bn)=bns and ψ((ab)2)=absabs=a2. The relation a2=bn=(ab)2 yields that ns≡n(2n). It follows that s is odd. Note that s=s′s′′, where s′∣2n and (2n,s′′)=1. We have that ψ=ψ5,s′′ψ4,s′.
(2.2) ψ(b)=bsa. Then ψ(a2)=a2=bn, ψ(bn)=(bsabsa)2n=an=bn2/2 and ψ((ab)2)=absaabsa=b−2s. The relation a2=bn=(ab)2 yields that n≡n2/2≡−2s(2n). It follows that 2n is odd and s=2n,23n. Since ab2naa−1=b−2na=b23na, we can consider only the case ψ(b)=b2na, which is ψ7.
(3) ψ(a)=ba. Note that there is an isomorphism ψ6 with b↦b and a↦ba in this case. By composing with such an isomorphism, this case is the same as case (2).
(4) ψ(a)=bk. Since a4=b2n=1, we must have that 4k≡0(2n). Up to conjugation by a, we may assume that k=2n or n.
(4.1) ψ(a)=b2n.
(4.1.1) ψ(b)=bs. Then ψ(a2)=bn, ψ(bn)=bns and ψ((ab)2)=(b2nbs)2=bn+2s. The relation a2=bn=(ab)2 yields that n≡ns≡n+2s(2n), which implies that s is odd and s≡0(n). This case is impossible because n is even.
(4.1.2) ψ(b)=bsa. Then ψ(a2)=bn, ψ(bn)=(bsabsa)2n=an and ψ((ab)2)=(b2nbsa)2=a2. The relation a2=bn=(ab)2 yields that n≡2(4). By Lemma 2.2, up to conjugation, there are two possibilities: either ψ(b)=a or ψ(b)=ba. They are respectively ψ8 and ψ6ψ8.
(4.2) ψ(a)=bn.
(4.2.1) ψ(b)=bs. Then ψ(a2)=1, ψ(bn)=bns and ψ((ab)2)=(bnbs)2=b2n+2s. The relation a2=bn=(ab)2 yields that 0≡ns≡2n+2s(2n), which implies that s=0 or n. They are respectively ψ3 and ψ2ψ6.
(4.2.2) ψ(b)=bsa. Note that ψ((ab)2)=(bnbsa)2=a2=1=ψ(a2). Thus, this case is also impossible.
Now we turn to the degree.
Consider the endomorphism ψ6 with ψ6(b)=b and ψ6(a)=ba. A direct computation shows that a linear map on R4 determined by the following matrix induces a self-map on S3/D4n∗, whose induced endomorphism on π1 is exactly ψ′.
[TABLE]
Since the determinant of the above matrix is 16, the degree of the corresponding self-map must be 1. Thus, we obtain that deg(ψ6)=1.
Let us write ψ2′ and ψ3′ for two homomorphisms from D4n∗ to Z2=⟨c⟩, defined by ψ2′(b)=c, ψ2′(a)=1, ψ3′(b)=1 and ψ3′(a)=c. They are two non-zero elements in H1(S3/D4n∗,Z2)=Z2⊕Z2. The other non-zero element is given by ψ2′ψ6. If 2∣2n, we know from [21, Theorem 2.2] that the cube of one element of these three is zero and the cube of the other two are both 1. Since H∗(S3/Z2)=Z2[c∗], we obtain that deg(φ)=φ∗(c∗)3[S3/D4n∗] for any φ:D4n∗→Z2, where [S3/D4n∗]∈H3(S3/D4n∗,Z2) is the fundamental class. Note that degψ6=1(4n). It follows that degψ2′ψ6=degψ2′=1 and hence degψ3′=0. These mean that degψ2=2n and degψ3=0. Recall from [21, Theorem 2.2] again that the cube of any element in H1(S3/D4n∗,Z2) vanishes if 2∣2n. Thus, degψ2=degψ3=0.
For ψ7 with b↦b2na and a↦a. It follows that b2n↦(b2na)2n which is b2na if n≡2(8), is b23na if n≡6(8). Note that ⟨b2n,a⟩=Q8. We may identify a as i, b2n as j in quaternion number. Thus, ψ7∣⟨b2n,a⟩(i)=i and ψ7∣⟨b2n,a⟩(j)=±k. Hence, ψ7∣⟨b2n,a⟩ is an isomorphism. By [14], we have that deg(ψ7∣⟨b2n,a⟩)=1. Clearly, deg(ψ7∣⟨b4⟩)=0. Hence, deg(ψ7)=4n2.
For ψ8 with b↦a and a↦b2n. Similar to the case above, ψ8∣⟨b2n,a⟩ is an isomorphism on Q8. By [14], we have that deg(ψ8∣⟨b2n,a⟩)=1. Clearly, deg(ψ8∣⟨b4⟩)=0. Hence, deg(ψ8)=4n2.
The degrees for ψ4,k and ψ5,l follow from [5, p.1248, Case II].
∎
It should be mentioned that ψ3=ψ8ψ2, if 2∣2n. The degree for ψ2,ψ7,ψ8 seem to be missing in [5].
Corollary 4.3**.**
If n>2, then the group Out(D4n∗)≅Z2×Zn∗ is generated by η1,η2, where η1(b)=b, η1(a)=ba and η2(b)=bs, η2(a)=a. Here s is an element generating
Zn∗≅Zφ(n).
The group Out(D4⋅2∗)≅S3 is generated by η3,η4, where η3(b)=ba, η3(a)=a and η4(b)=a, η2(a)=b.
Theorem 4.4**.**
Let φ be a surjective non-trivial homomorphism from D4n∗ to the fundamental group G of some 3-dimensional spherical manifold S3/G. Then φ=η′ψη′′, where η′′∈Aut(D4n∗) and η′∈Aut(G). Moreover, all possible G and ψ are listed as follows:
Proof.
The first row comes from the degree of ψ4,k in Lemma 4.2, where m=kn. The other two come from the degrees of ψ2 and ψ3 there.
∎
For O48∗, we know that the only non-trivial proper quotient is Z2. Hence,
Theorem 4.5**.**
Let φ be a non-trivial surjective homomorphism from O48∗ to G induced by a map from S3/O48∗ to some 3-dimensional spherical manifold S3/G. Then φ is conjugate to one of the following.
[TABLE]
Proof.
The first two rows comes from [14]. Let f:S3/O∗48→L(2;1,1) be a map, inducing a homomorphism indicated in third row. By [21, Theorem 4.13], H3(S3/O48∗,Z2)
is generated by β13 for the unique non-zero element β1∈H1(S3/O48∗,Z2). It follows that the deg(f)≡1(2) because f∗(c3)=β13. We obtain the third row.
∎
We conclude that D(O48∗)={m+48k∣m=0,1,24,25,k∈Z}. Note that the values 24+48k,k∈Z were missing in [5].
For I120∗ the only possible proper quotient is the trivial group.
Theorem 4.6**.**
Let φ be a non-trivial surjective homomorphism from I120∗ to G induced by a map from S3/I120∗ to some 3-dimensional spherical manifold S3/G, then φ is conjugate to either the identity or the automorphism η of I120∗ determined by a↦a−1, b↦b−1abab−1a, where G is also I120∗. Here, deg(η)≡49(120).
Lemma 4.7**.**
Each endomorphism of T8⋅3q′ is conjugate to a composition of some endomorphisms in the following table
Proof.
By Lemma 2.9, the non-trivial normal subgroups of T8⋅3q′ are Z3q−r=⟨w3r⟩, Z2⋅3q−r=⟨b2w3r⟩, or D4⋅2∗×Z3q−r=⟨b,a,w3r⟩ with 0<r≤q. Since neither T8⋅3q′/⟨w3r⟩ nor T8⋅3q′/⟨b2w3r⟩ is a subgroup of T8⋅3q′ except for the trivial case T8⋅3q′/⟨w3q⟩=T8⋅3q′, any endomorphism on T8⋅3q′ has kernel 1, D4⋅2∗×Z3q−r, or T8⋅3q′.
(1) kerψ=1, i.e. ψ is an automorphism. Note that ψ(w) has order 3q. By Lemma 2.8, we may assume that ψ(w)=w3t+1 or w3t+2. Since a has order 4, ψ(a) has also order 4, and hence ψ(a)∈{b,a,ba,b−1,b2a,b−1a}. Note that waw−1=b, wbw−1=ab=b−1a, wb2aw−1=b−1 and wb−1w−1=ba. Up to a conjugation, we may also assume that ψ(a)=a or b−1. There four subcases:
(1.1) ψ(w)=w3t+1 and ψ(a)=a. It follows that ψ(b)=ψ(waw−1)=ψ(w)ψ(a)ψ(w−1)=b. From number theory, we know that the set Z3q∗ of all multiple invertible elements in Z3q is a cyclic group of order 3q−3q−1, i.e. Z3q∗≅Z3q−3q−1. Moreover, 2 is a generator of Z3q∗. Thus, ψ=ψ1r, where 2r≡3t+1(3q).
(1.2) ψ(w)=w3t+1 and ψ(a)=b−1. It follows that ψ(b)=ψ(waw−1)=ψ(w)ψ(a)ψ(w−1)=ba. Hence, ψ(wbw−1)=w3t+1baw−(3t+1)=abb=b2a, but ψ(ab)=b−1ba=a. Thus, this is an impossible case.
(1.3) ψ(w)=w3t+2 and ψ(a)=a. It follows that ψ(b)=ψ(waw−1)=ψ(w)ψ(a)ψ(w−1)=b−1a. Hence, ψ(wbw−1)=w3t+2b−1aw−(3t+2)=a−1ab=b, but ψ(ab)=ab−1a=ba2=b3. Thus, this is an impossible case.
(1.4) ψ(w)=w3t+2 and ψ(a)=b−1. It follows that ψ(b)=ψ(waw−1)=ψ(w)ψ(a)ψ(w−1)=b2a. Here, ψ=ψ1k, where 2k≡3t+2(3q).
(2) kerψ=D4⋅2∗×Z3q−r with 0<r<q. Then ψ(a)=ψ(b)=1. Since ψ(w3r)=1, the order of ψ(w) is a divisor of 3r. By Lemma 2.8, up to a conjugation, we may also assume that ψ(w)=w3t. Note that 2 is a generator of Z3q∗. There is some s such that 3t⋅2s≡3r(3q). Thus, ψ1sψ(w)=w3r. It follows that ψ=ψ1−sψ2r if r>0; ψ=ψ2−sψ3 ir r=0. The degree of ψ2,ψ3 were given in [5, p.1250, Case VI].
(3) kerψ=T8⋅3q′. Clearly, ψ=ψ4.
Consider the degree of ψ1. Since ψ1∣⟨b,a⟩ is an automorphism on Q8, we have that deg(ψ1∣⟨b,a⟩)≡1(8) by [14]. Thus, deg(ψ1)≡1(8). On the other hand,
By Lemma 3.3, we have that deg(ψ1∣⟨w⟩)≡22(3q). We obtain that deg(ψ1)≡4−32q+1(8⋅3q).
∎
Corollary 4.8**.**
The group Out(T8⋅3q′)≅Z3q−3q−1 is generated by η, which is defined by η(a)=b−1 and η(w)=w2.
Theorem 4.9**.**
Let φ be a surjective non-trivial homomorphism from T8⋅3q′ to G induced by a map from S3/T8⋅3q′ to some 3-dimensional spherical manifold S3/G. Then φ=η′ψη′′, where η′′∈Aut(T8⋅3q′) and η′∈Aut(G). Moreover, all possible G and ψ are listed as follows:
[TABLE]
Proof.
(1) kerψ=⟨w3r⟩, the homomorphism ψ:T8⋅3q′→T8⋅3r′ can be given by ψ(b)=bˉ, ψ(a)=aˉ ψ(w)=wˉ. By Lemma 3.2, S3/⟨w⟩ and S3/⟨wˉ⟩ are respectively Lens spaces L(3q;2−3q,2+3q) and L(3r;2−3r,2+3r). By Lemma 3.3,
[TABLE]
Since ψ∣⟨b,a⟩ is an isomorphism, degψ∣⟨b,a⟩≡1(8). It follows that
[TABLE]
(2) kerψ=Z2⋅3q−r. Since its quotient (Z2×Z2)⋊Z3r is not a subgroup in our list, this is an impossible case.
(3) kerψ=D4⋅2∗×Z3q−r (0<r≤q). Then G=Imψ=Z3r=⟨c∣c3r⟩. The homomorphism ψ:T8⋅3r′→Z3r can be given by ψ(a)=1 and ψ(w)=c. By identification ι:G→⟨w3q−r⟩, we have that ιψ=ψ2q−r, where ψ2r is listed in Lemma 4.7. Since degψ2q−r=(9−32q+2)q−r, we obtain that degψ=(9−32q+2)q−r/(8⋅3q−r)=81×(3−32q+1)q−r(3r).
∎
Lemma 4.10**.**
Each endomorphism of Dn⋅2q′ is conjugate to a composition of some endomorphisms in the following table
[TABLE]
Proof.
By Lemma 2.11, kerψ must be 1, ⟨uk⟩ (k∣n,1≤k<n), ⟨uw2r⟩ (0<r<q), ⟨ukw2r⟩ (k∣n,1<k<n, 1<r<q), or Dn⋅2q′.
(1) kerψ=1. Since ψ(u) has order n and ⟨u⟩ is the commutator and hence is a fully invariant subgroup, ψ(u)=us for some s with (s,n)=1. Since ψ(w) has order 2q, we have ψ(w)=umwt for some odd t. Note that u(umwt)u−1=um+2wt. Up to a conjugation, we may assume that ψ(w)=wt. Thus, ψ=ψ1,sψ2,t. Moreover, since w(uswt)w−1=u−s, we may assume that 1≤s<2n.
(2) kerψ=⟨uk⟩ with k∣n and 1≤k<n. Then ψ(u)=ukn. Since ψ(w) has order 2q, by Lemma 2.10, we have ψ(w)=wt for some odd t. Thus, ψ is ψ2,tψ3,k.
(3) kerψ=⟨uw2r⟩ with 0<r<q. Then ψ(u)=1. Since ψ(w) has order 2r, by Lemma 2.10, we have ψ(w)=wt⋅2r for some odd t. Thus, ψ is ψ2,tψ4r if 0<r<q.
(4) kerψ=Dn⋅2q′. Clearly, ψ=ψ4q.
Now we consider the degree deg(ψ). Note that both ⟨u⟩ and ⟨w⟩ are invariant subgroups of ψ. Assume that ψ(u)=us and ψ(w)=wt. By Lemma 3.3, we have that deg(ψ∣⟨u⟩)=s2 and deg(ψ∣⟨w⟩)=t2. Thus, deg(ψ) is the unique solution module n⋅2q of the system of congruences:
[TABLE]
By Euler-Fermat theorem, we have that nφ(2q)=n2q−1≡1(2q). Thus, we have
n⋅n2q−1−1+2−q(1−n2q−1)2q=1. By the Chinese remainder theorem, we have
deg(ψ)=(1−n2q−1)s2+n2q−1t2.
We obtain the all deg(ψ)’s.
∎
Corollary 4.11**.**
The group Out(Dn⋅2q′) is isomorphic to
(Zn∗/Z2)×Z2q∗.
Proof.
Here we give an alternate proof of this reuslt as [8, Theorem 2.4(2)] contains an error which will be corrected. From Proposition 1.3 of [8] we have a short exact sequence,
[TABLE]
where Derα(Z2q,Zn)≅Zn, and
η(Zn⋊αZ2q)≅Zn∗×Z2q∗. Since the group
Zn∗×Z2q∗
embeds naturally in Aut(Zn⋊αZ2q) we have a section. Thus, we obtain
[TABLE]
Denote by (a,(b,c)) a typical element of Aut(Zn⋊αZ2q) where a∈Zn,(b,c)∈Zn∗×Z2q∗. Here, we write Zk as an additive group and Zk∗ as a multiplicative group.
Now let us factor this group by the inner automorphisms. Write ⟨x⟩=Zn and ⟨y⟩=Z2q.
The conjugation of (x,0) by (0,y) yields (−x,0) thus the corresponding inner automorphism is the element (0,(−1,1)). The conjugation of (0,y) by (x,0) yields (2x,y) thus the corresponding inner automorphism is the element (2,(1,1)). It follows that the inner automorphisms form the subgroup
Zn⋊(Z2×{1}). Therefore the quotient is isomorphic to
(Zn∗/Z2)×Z2q∗
and the result follows.
∎
Theorem 4.12**.**
Let φ be a surjective non-trivial homomorphism from Dn⋅2q′ to G induced by a map from S3/Dn⋅2q′ to some 3-dimensional spherical manifold S3/G. Then φ=η′ψη′′, where η′′∈Aut(Dn⋅2q′) and η′∈Aut(G). Moreover, all possible G and ψ are listed as follows:
[TABLE]
Proof.
If kerψ=⟨ukw2r⟩≅Zkn⋅2q−r (k∣n, k>1 and 1<r<q). The homomorphism ψ:Dn⋅2q′→Dk⋅2r′ can be given by ψ(u)=uˉ, ψ(w)=wˉ. By Lemma 3.2, S3/⟨w⟩ and S3/⟨wˉ⟩ are respectively lens spaces L(2q;1+2q−1,1) and L(2r;1+2r−1,1). By Lemma 3.3, degψ∣⟨w⟩≡2q−r(1+2r−1)[(1+2q−1)](2q)−1=2q−r(1+2r−1)(1−2q−1)≡2q−r(2r). On the other hand, by Lemma 3.2 again, S3/⟨u⟩ and S3/⟨uˉ⟩ are respectively lens spaces L(n;1,−1) and L(k;1,−1). By Lemma 3.3, degψ∣⟨u⟩≡1(k). It follows that
[TABLE]
The other cases come from Lemma 4.10, by using Lemma 4.1.
∎
Next Lemma deals with the case where the domain is in the form of a direct product.
Lemma 4.13**.**
Let φ:Zm×G1→G′ be a surjective homomorphism. Then G′=Zk×G2 and φ=φ1×φ2. Moreover, both of φ1:Zm→Zk and φ2:G1→G2 are surjective, and degφ=(∣G1∣degφ2+mdegφ1)[∣G1∣+m](m∣G1∣)−1.
5. Computations
We conclude this paper with some computations of the mapping degree of certain spherical 3-manifolds.
In order to compute D(S3/G1,S3/G2), our strategy is as follows: (1) look for all normal subgroups H of G1 such that G1/H is isomorphic to a subgroup of G2; (2) find the degree of φH:G1→G1/H; (3) look for all embedding ξ:G1/H→G2. The degrees of epimorphisms φH:G1→G1/H were computed in
Lemma 3.3 if G1 is cyclic, and in Theorems 4.4, 4.5, 4.6, 4.9, 4.12 if G1
is one of the groups listed in Table 2.1.
Thus Lemma 4.1 will compute the set
{deg(ξ)deg(φH)}. Here, deg(ξ)=deg(η′)∣G2∣∣H∣/∣G1∣ for some η′∈Out(G1/H). The set D(M,N) can be determined once
deg(ψ) is for all ψ.
Example 5.1**.**
Let G1=Dn⋅2q′ and G2=Dm⋅2r′ with r<q.
Suppose f:M1→M2 is a map between two spherical 3-manifolds with π1(M1)=Gi for i=1,2 and φ=f♯:G1→G2.
By Lemma 2.11, if kerφ is non-trivial, then kerφ is either Zkn⋅2q−r with k∣(n,m) or Zn⋅2l with 1<l≤r.
(1) kerφ=⟨ukw2r⟩≅Zkn⋅2q−r with k∣(n,m). By Theorem 4.12, we have
[TABLE]
where η′′∈Aut(Dn⋅2q′) and η′∈Aut(Dq⋅2r′).
(2) kerφ=⟨uw2q−l⟩≅Zn⋅2l with 1<l≤r. By Theorem 4.12, we have
[TABLE]
where η′′∈Aut(Dn⋅2q′) and η′∈Aut(Dq⋅2r′).
Clearly, if kerφ=Dn⋅2q′, then deg(φ)≡0(m⋅2r). Thus, we can write down the set D(S3/Dn⋅2q′,S3/Dm⋅2r′).
Example 5.2**.**
Let G1=Z120 and G2=O48∗×Z5.
By Lemma 3.2, the manifold S3/(O48∗×Z5) is covered by three lens spaces L(30;1,11), L(40;3,13) and L(40;−7,23). Thus, any map from L(120;r1,r2) can be lifted to a map into one of these three lens spaces.
Note that the degrees from these three lens spaces to S3/(O48∗×Z5) are respectively 8, 6 and 6. By Theorem 3.4, D(L(120;r1,r2),S3/(Q48×Z5)) consists of the following three sets:
[TABLE]
Note that any lens space L(120;r1,r2) must be homeomorphic to one of the form L(120;1,r), where
[TABLE]
We list all the elements of D(L(120;r1,r2), S3/(O48∗×Z5)) as follows.
[TABLE]