CLASSIFICATION OF THE CONJUGACY CLASSES OF SLf2(R)
CHRISTIAN TAFULA´
Abstract. In this paper, we classify the conjugacy classes of SLf2(R), the universal covering group of PSL2(R). For any non-central element α ∈ SLf2(R), we show that its conjugacy class may be determined by three invariants: (i) Trace: the trace (valued in the set of positive real numbers R+) of its image α in PSL2(R); (ii) Direction type: the sign behavior of the induced self-homeomorphism of R 1 determined by the lifting SLf2(R) y R of the action PSL2(R) y S ; (iii) Length-sharp: a conjugacy invariant version of the length function in S. Zhang [12], introduced by S. Mochizuki [9].
Contents 0. Introduction...... 1 1. CP-extensions of groups ...... 3 1 2. The group Homeo+(S )...... 14 g 1 3. Invariants in Homeog +(S )...... 21 4. Conjugacy classes of SLf2(R)...... 25 References...... 31
0. Introduction Our goal in this paper is to classify the conjugacy classes of the universal covering group of PSL2(R), namely, SLf2(R). Based on the group-theoretic proof of Szpiro’s theorem as described by Zhang [12] (originally due to Bogomolov et al. [1]), we start 1 by identifying PSL2(R) with a subgroup of Homeo+(S ), the group of orientation- preserving self-homeomorphisms of S1. This is done via the faithful continuous action 1 induced by PSL2(R) on S as described in Section 4. Then, via the natural quotient map R 3 x 7→ x ∈ S1 ' R/Z, we show that the topological fundamental group of 1 1 Homeo+(S ) is π1(Homeo+(S )) ' Z and then describe its universal covering group 1 Homeo] +(S ) as a subgroup of Homeo+(R), the group of self-homeomorphisms of R which are increasing. The group extension
1 1 1 Z Homeo] +(S ) Homeo+(S ) 1 is shown to be a CP-extension, a notion which is defined in the discussion following Definition 1.1. The notion of a CP-extension is characterized in Theorem 1.12 and
2010 Mathematics Subject Classification. Primary 20E45; Secondary 22E46. Key words and phrases. SL(2, R), conjugacy classes, group extensions.
1 2 CHRISTIAN TAFULA´
discussed at length in Section 1. Denote the image of Z in the above sequence by Z. 1 In Lemma 2.5, we show that this subgroup is actually the center of Homeo] +(S ). A particular consequence of Theorem 1.12, which allows us to calculate the conjugacy 1 classes of SLf2(R) explicitly, is that the set of conjugacy classes of Homeo] +(S ) admits 1 a natural structure of Z-torsor over the set of conjugacy classes of Homeo+(S ). This 1 property is inherited by arbitrary subgroups of Homeo] +(S ) that contain the subgroup 1 Z. More explicitly, we show that the conjugacy class of an element α ∈ Homeo] +(S ) is determined uniquely by the following data: • an element ζ ∈ Z, • the canonical lifting (as defined in Proposition 2.10) of a certain representative of 1 the conjugacy class of its image α ∈ Homeo+(S ). This is stated in Corollary 2.11. The two main concepts used in the proof are the notions of CP-homomorphisms of groups (Definition 1.1) and quasi-homomorphisms (Definition 1.14 — cf. Fujiwara [5]). Corollary 2.11 is derived from the somewhat more technical Theorem 1.21, which, roughly speaking, states that “central group extensions whose kernel may be separated from the identity by quasi-homomorphisms are CP-extensions”. Separation is defined in Subsection 1.5. Next, we classify the non-central conjugacy classes of SLf2(R) completely by considering the trace (which is motivated by the usual classification of the conjugacy classes of SL2(R) as elliptic, parabolic, or hyperbolic — cf., for instance, K. Conrad [3]), together with two further invariants of conjugacy classes in SLf2(R), which 1 arise from the structure of SLf2(R) as a subgroup of Homeo] +(S ). The invariants considered are the following: a b (i) Trace: tr(α) := |a + d|, where [ c d ] = α ∈ PSL2(R) is the image of α ∈ SLf2(R) via the natural quotient map;
1 (ii) Direction type: Elements α ∈ Homeo] +(S ) will be constructed in such a way that the function R 3 x 7→ α(x) − x ∈ R is invariant with respect to the translation R 3 x 7→ x + 1 ∈ R. The sign behavior of α(x) − x can be classified as follows: strictly positive (the forward direction type), strictly negative (the backward direction type), positive but not strictly positive (the semi-forward direction type), negative but not strictly negative (the semi-backward direction type), or alternating in sign (the alternating direction type). (iii) Length-sharp: As was mentioned above, α(x) − x is invariant under translation by 1; hence, in order to determine the behavior of α, it suffices to restrict our
attention to x ∈ [0, 1], which is compact. In particular, `(α) := supx∈[0,1] |α(x)− x| is a well-defined real number. We then define the length-sharp function by taking `](α) := {b`(α)c, d`(α)e} ⊆ N. The direction type (cf. (ii)) and length-sharp function (cf. (iii)) are invariant under 1 conjugation in Homeo] +(S ) (Lemma 3.2 and Proposition 3.4, respectively). The length-sharp function, which we describe in Subsection 3.2, is based on the exposition of Bogomolov’s proof of Szpiro’s theorem by Mochizuki [9]. A table at the end of this paper shows how (i), (ii), and (iii) determine the conjugacy classes of non-central elements of SLf2(R). CLASSIFICATION OF THE CONJUGACY CLASSES OF SLf 2(R) 3
0.1. Notation. Write N for the set of non-negative integers. For x ∈ R, we use the standard notations bxc := max{k ∈ Z | k ≤ x} and dxe := min{k ∈ Z | k ≥ x} for the floor and ceiling functions, respectively. 0.1.1. Groups. For a group G, we denote its identity element by e ∈ G and the center of G by Z(G). For a subset S ⊆ G, we denote the commutant (or centralizer) of S in G by ZG(S) := {g ∈ G | gs = sg, ∀s ∈ S}. Given g1, g2 ∈ G, we say that g1 −1 is conjugate to g2 if there exists h ∈ G such that g1 = hg2h , in which case we write g1 ∼ g2. Conjugacy is an equivalence relation; we denote the conjugacy class of g ∈ G by [g] := {h ∈ G | g ∼ h} and the set of all conjugacy classes of G by C`(G). For any two elements g, h ∈ G, we denote their commutator by [g, h] := ghg−1h−1. For elements gi ∈ G, indexed by some set I 3 i, the subgroup generated by the gi’s will be denoted by hgi | i ∈ Ii. For N E G a normal subgroup, we denote the associated natural quotient map by G 3 g 7→ g ∈ G/N when no ambiguity is present.
a b 0.1.2. Special linear group. An element ( c d ) ∈ SL2(R) is a matrix with real entries satisfying ad − bc = 1, and we denote its image in PSL2(R) := SL2(R)/{±Id}, where 1 0 a b −a −b Id := ( 0 1 ), by [ c d ] = −c −d ∈ PSL2(R). Since the trace function on SL2(R) is a b given by tr(( c d )) = a + d, the function M 7→ |tr(M)| naturally factors through a b a b PSL2(R); thus, we define the trace function on PSL2(R) by tr([ c d ]) := tr(( c d )) . 0.1.3. Groups of self-homeomorphisms. For a topological space X, write Homeo(X) for the group of self-homeomorphisms of X and Homeo0(X) ⊆ Homeo(X) for the subgroup of self-homeomorphisms that are isotopic to the identity. In general, given two topological spaces X,Y , the compact-open topology on C(X,Y ), the set of continuous maps from X to Y , is defined to be the topology generated by sets of the form V (K,U) := {f ∈ C(X,Y ) | f(K) ⊆ U}, where K ⊆ X is compact and U ⊆ Y is open. We regard Homeo(X) ⊆ C(X,X) as being equipped with the topology induced by the compact-open topology on C(X,X). Thus, Homeo0(X) is the connected component of Homeo(X) that contains the identity. In the particular 1 1 1 case of the circle S := R/Z and the real line R, we write Homeo+(S ) := Homeo0(S ) and Homeo+(R) := Homeo0(R). 0.1.4. Torsors. Let C be a category that admits finite products. For an object Y and a group object G in C, we shall say Y is a (left) G-torsor if there is a (left) action G y Y , written σ : G × Y → Y , such that the morphism
σ × pG×Y/Y : G × Y −→ Y × Y, where pG×Y/Y : G × Y → Y denotes the projection to the second factor, is an isomorphism. This is a natural generalization of the notion of principal G-bundles for smooth manifolds.
1. CP-extensions of groups 1.1. CP-homomorphisms. We start with a definition. Definition 1.1 (CP-homomorphism). A homomorphism of groups φ : G → H is called commutant-preserving homomorphism (or a CP-homomorphism, for short) if ∀x, y ∈ G, φ(x)φ(y) = φ(y)φ(x) =⇒ xy = yx. 4 CHRISTIAN TAFULA´
Remark 1.2. For any group homomorphism φ : G → H, it holds that ZG(x) ⊆ −1 φ ZH (φ(x)) for all x ∈ G; thus, φ is a CP-homomorphism exactly when the −1 T converse holds as well, i.e., ZG(x) = φ ZH (φ(x)) . Since ZG(S) = x∈S ZG(x) for any subset S ⊆ G, a CP-homomorphism φ : G → H satisfies −1 (1.1) ∀S ⊆ G, ZG(S) = φ ZH (φ(S)) ,
hence its name. In particular, ZG(ker(φ)) = G, i.e., ker(φ) ⊆ Z(G). We briefly recall the following basic facts on group extensions: Given two groups Q, N, a group extension G of Q by N consists of an injective homomorphism from N to G and a surjective homomorphism φ from G to Q such that the resulting sequence φ (1.2) 1 N G Q 1 is exact. If we identify N with its image in G, then N = ker(φ). An extension is called central if N ⊆ Z(G). Definition 1.3 (CP-extensions and CP-groups). We shall say that an extension 1 (1.2) is a CP-extension if φ : G Q is a CP-homomorphism. (Thus, by Remark 1.2, every CP-extension is central.) A group G is called a CP-group if the quotient map G G/Z(G) is a CP-extension. Several characterizations of CP-extensions are given in Theorem 1.12. Notice that CP-extensions are necessarily central. The first relevant properties of CP- homomorphisms are the following: Lemma 1.4 (Basic properties of CP-homomorphisms). Let φ : G → H and ψ : H → K be group homomorphisms. Then the following hold: (i) If φ and ψ are CP-homomorphisms, then the composite ψ ◦ φ : G → K is also a CP-homomorphism. (ii) If the composite ψ ◦ φ is a CP-homomorphism, then φ is a CP-homomorphism. Moreover, if φ is surjective, then ψ is also a CP-homomorphism. (iii) If φ is CP-homomorphism, then for any group homomorphism ξ : A → H, the map A ×H G → A induced by φ is a CP-homomorphism. (iv) If G is a CP-group, then any central extension 1 → N → G → Q → 1 is a CP-extension. Proof. We prove the items separately, starting from item (i). The condition that φ be a CP-homomorphism is equivalent to the condition that φ([x, y]) = eH =⇒ [x, y] = eG, for any commutator [x, y] ∈ G, where eG, eH , eK denote the identity elements of the respective groups. Thus, since both φ and ψ are CP-homomorphisms and φ([x, y]) = [φ(x), φ(y)],
ψ(φ([x, y])) = eK =⇒ φ([x, y]) = eH =⇒ [x, y] = eG, so ψ ◦ φ is a CP-homomorphism. Next, we consider item (ii). Suppose that ψ ◦ φ is a CP-homomorphism. Then, for any x, y ∈ G, we have
φ([x, y]) = eH =⇒ ψ(φ([x, y])) = eK =⇒ [x, y] = eG,
1In other words, a CP-extension amounts to a surjective CP-homomorphism. CLASSIFICATION OF THE CONJUGACY CLASSES OF SLf 2(R) 5
so φ is a CP-homomorphism. If, in addition, φ is surjective, then for any a, b ∈ H there exist x, y ∈ G such that φ(x) = a and φ(y) = b. Thus,
ψ([a, b]) = eK =⇒ ψ(φ([x, y])) = eK =⇒ [x, y] = eG, so ψ is a CP-homomorphism. Next, we consider item (iii). Consider the commutative diagram
A ×H G G
p φ ξ A H.
Since A ×H G = {(a, g) ∈ A × G | ξ(a) = φ(g)}, given (a1, g1), (a2, g2) ∈ A ×H G, we have
a1a2 = a2a1
=⇒ ξ(a1)ξ(a2) = ξ(a2)ξ(a1)
=⇒ φ(g1)φ(g2) = φ(g2)φ(g1)
=⇒ g1g2 = g2g1 (because φ is a CP-homomorphism),
therefore a1a2 = a2a1 implies (a1, g1)(a2, g2) = (a2, g2)(a1, g1). Finally, item (iv) follows from item (ii), by considering the composite of surjections G G/N G/Z(G). 1.2. Induced map on conjugacy classes. Consider the exact sequence (1.2). We denote conjugacy equivalence in G by ∼G, and G-conjugacy classes by a subscript G. Thus, for g ∈ G, we have [g]G = {a ∈ G | a ∼G g}. We use similar notational conventions for Q. The surjection φ induces a surjective map of sets Φ : C`(G) 3 −1 [g]G 7→ [φ(g)]Q ∈ C`(Q). Our goal (Proposition 1.8) is to describe Φ ([q]Q) for each conjugacy class [q]Q ∈ C`(Q). We start with the general case, where N is an arbitrary normal subgroup of G, and later we impose the hypothesis that N ⊆ Z(G). −1 For each q ∈ Q, for the rest of this subsection, choose and fix a lifting qe ∈ φ (q) ⊆ G. Thus, every element in G may be written in the form kqe, for a uniquely determined (k, q) ∈ N × Q. Write I := C`(Q). Choose and fix a representative element qi ∈ Q for each i ∈ I, so C`(Q) = {[qi]Q | i ∈ I}.
Lemma 1.5. For every g ∈ G, there exists a unique i ∈ I such that g ∼G kqei, for some (not necessarily unique) k ∈ N.
Proof. Write g = k1qe. By definition, there is a unique i ∈ I such that q ∼Q qi, i.e., −1 −1 −1 ∃p ∈ Q such that q = pqip . Since φ(pq^ip ) = φ(peqei pe ), there exists k2 ∈ N such −1 −1 −1 that pq^ip = k2 · peqei pe . Furthermore, since N is normal, k3 := pe k1k2 pe ∈ N. −1 −1 −1 Hence, g = k1qe = k1 · pq^ip = k1k2 · peqei pe = pe(k3qei) pe , so g ∼G k3qei, as desired. Therefore, in order to determine the fibers of Φ, it suffices to analyse, for each i ∈ I, the conditions under which kqei ∼ jqei for k, j ∈ N. To this end, we consider the following. 6 CHRISTIAN TAFULA´
Definition 1.6 (Twisted conjugation). Let N E G be a fixed normal subgroup of a group G. For each g ∈ G, we define g-twisted conjugation in N to be the following equivalence relation: for k, j ∈ N, g −1 (1.3) k ≈g j ⇐⇒ ∃a ∈ G | k = aj(a ) , g −1 where we write a := gag . When k ≈g j, we say that k and j are g-twisted conjugate in N, and, when there is no ambiguity with respect to N, we write [[k]]g ∈ N/≈g for the set of elements of N that are g-twisted conjugate to k. Here, we note that this relation is well-defined, for if k, j, ` ∈ N are such that k = aj(ag)−1, j = b`(bg)−1 for some a, b ∈ G, then k = ek(eg)−1, j = a−1k((a−1)g)−1, k = (ab)` (ab)g−1,
so ≈g is reflexive, symmetric and transitive. The following are basic properties of twisted conjugation. Lemma 1.7. Let g ∈ G. Then, for any k, j ∈ N and p ∈ G, the following hold:
(i) kg ∼G jg ⇐⇒ k ≈g j; (ii) k ≈pg j ⇐⇒ kp ≈g jp; −1 −1 (iii) k ≈pgp−1 j ⇐⇒ p kp ≈g p jp; Proof. For item (i), we have −1 kg ∼G jg ⇐⇒ ∃a ∈ G | kg = ajga ⇐⇒ ∃a ∈ G | k = ajga−1g−1 = aj(ag)−1
⇐⇒ k ≈g j. Item (ii) follows immediately from item (i):
k ≈pg j ⇐⇒ kpg ∼G jpg ⇐⇒ kp ≈g jp. −1 Item (iii) follows immediately by conjugating by p . −1 We can now apply Lemmas 1.5 and 1.7 (i) to describe the fibers Φ ([qi]Q) of the map Φ : C`(G) 3 [g]G 7→ [φ(g)]Q ∈ C`(Q), as follows: For each i ∈ I, write Ψ : N 3 k 7→ [[k]] ∈ Λ := N/ ≈ . Then, for each λ(i) ∈ Λ , choose and fix a i qei i qei i −1 (i) representative element κλ(i) ∈ Ψi (λ ). Thus, we have: Proposition 1.8. It holds that (i) C`(G) = [κλ(i) qei]G | i ∈ I, λ ∈ Λi . In particular, if the q -twisted conjugacy relations are trivial (i.e., N/≈ = N for ei qei every i ∈ I), then the map N × C`(Q) 3 (k, [q ] ) 7→ κ q ∈ C`(G) is bijective. i Q Ψi(k) ei G Remark 1.9. One way to interpret Proposition 1.8 is by viewing the qei-twisted conjugacy relations as encoding the obstructions preventing the map N × C`(Q) 3 (k, [q ] ) 7→ κ q ∈ C`(G) from being injective. i Q Ψi(k) ei G CLASSIFICATION OF THE CONJUGACY CLASSES OF SLf 2(R) 7
The problem with the description of C`(G) given in Proposition 1.8 lies in the −1 fact that it depends on both the choice of liftings qe ∈ φ (q) and the choice of representatives for [q]Q in Q. The properties of Lemma 1.7 (ii), (iii) describe how −1 different liftings (say q,e qb ∈ φ (q)) and different representatives for conjugacy classes in Q (say r, s ∈ [q]Q) may give rise to different equivalence relations on N (i.e., N/≈ ⇐⇒/N/≈ ). The situation is substantially ameliorated when N is central. re sb 1.3. Properties of CP-extensions. We continue to consider a group extension
φ 1 N G Q 1. If N is a central subgroup of G, then the multiplication by an element in N to the left induces an action (1.4) N × C`(G) 3 (k, [g]) 7→ [kg] ∈ C`(G). On the other hand, for k ∈ G, in order for the multiplication by k to the left to be well defined as a map of conjugacy classes, it is necessary and sufficient that −1 (1.5) ∀g ∈ G, ∀p ∈ G, k(pgp ) ∼G kg, which is clearly the case when N is central. Moreover, the converse holds. Lemma 1.10. An element k ∈ G satisfies (1.5) if and only if k ∈ Z(G). Proof. If k ∈ Z(G), then (1.5) clearly holds. For the other direction, suppose (1.5) holds for k. Since this condition applies to all g, p ∈ G, it applies, in particular, −1 for g = k , which implies [k, p] ∼G e for all p ∈ G. Since [e]G = {e}, that means [k, p] = e for every p ∈ G, therefore k ∈ Z(G). If N is central, then, using the above action, we may describe the fibers of the map Φ : C`(G) C`(Q) without referencing representatives or liftings. Indeed, this is a consequence of the next lemma.
Lemma 1.11. Suppose N ⊆ Z(G), and define Sg := {[a, g] | a ∈ G} ∩ N for each g ∈ G. Then, the following hold:
(i) For each g ∈ G, the set Sg is a subgroup of N, and N/≈g = N/Sg. (ii) We have Sg = Skg for every k ∈ N and g ∈ G. (iii) We have Sg = Spgp−1 for every p, g ∈ G. (iv) For g, h ∈ G, if φ(g) ∼Q φ(h) then Sg = Sh. (v) For a conjugacy class [q] ∈ C`(Q), write S := S , for some lifting q ∈ φ−1(q). Q [q] qe e Then, the action N y C`(G) described in (1.4) induces a free, transitive action −1 N/S[q] y Φ ([q]Q). Proof. For item (i), since N ⊆ Z(G), we have [a, g][b, g] = a[b, g]a−1 [a, g] = [ab, g], [a, g]−1 = a−1[a, g]−1a = [a−1, g], thus, Sg is a group. From the definition of ≈g in (1.3) and the assumption that −1 N ⊆ Z(G), we have k ≈g j ⇐⇒ kj ∈ Sg for k, j ∈ N, and thus N/≈g = N/Sg. Item (ii) (resp. (iii)) follows from Lemma 1.7 (ii) (resp. (iii)). 8 CHRISTIAN TAFULA´
For item (iv), the condition φ(g) ∼Q φ(h) implies that there are k ∈ N, p ∈ G such that g = k php−1. Then, item (iv) follows from the items (ii) and (iii). Finally, for item (v), notice that for each conjugacy class [q]Q ∈ C`(Q), we deduce from item (iv) that the group S depends neither on the choice of liftings q ∈ φ−1(q) qe e nor on the choice of a representative for [q]Q; that is, S[q] is well-defined. From item (i), we have N/≈ = N/S , therefore, by Lemma 1.7 (i), qe [q] StabN ([qe]G) : = {k ∈ N | kqe ∼G qe} = {k ∈ N | k ≈ e} = S , qe [q] −1 implying that the induced action N/S[q] y Φ ([q]Q) is well-defined and free. More- over, in the notation from the discussion leading up to Proposition 1.8, letting qi ∈ Q, i ∈ I be such that q ∼Q qi, Proposition 1.8 says that N/S × {[q ] } 3 k, [q ] 7−→ κ q ∈ Φ−1([q] ) ⊆ C`(G) [q] i Q i Q Ψi(k) ei G Q is a one-to-one correspondence, amounting to the fact that the action is transitive. With Lemma 1.11 (v), we obtain a complete description of C`(G) in terms of N and C`(Q), depending only on φ, which is necessary for determining S[q]. Categorically, −1 we say that Φ ([q]Q) is a N/S[q]-torsor in Set. When G is a CP-extension of Q, it turns out that S[q] is trivial for every [q]Q ∈ C`(Q), as will be shown in the next theorem. Thus, in this situation, N acts freely on C`(G), and the orbits of this action −1 are exactly the Φ ([q]Q) for [q]Q ∈ C`(Q); i.e., C`(G) is an N-torsor over C`(Q) (along Φ), meaning that it is a (N × C`(Q))-torsor in the slice category Set/C`(Q).
φ Theorem 1.12 (Characterizations of CP-extensions). Let 1 → N → G → Q → 1 be an exact sequence of groups. Then, the following are equivalent: (i) G is a CP-extension of Q by N. −1 −1 (ii) For any subset T ⊆ Q, it holds ZG(φ (T )) = φ (ZQ(T )). (iii) N ∩ {[a, b] | a, b ∈ G} = {e}. (iv) The twisted conjugacy classes in N are trivial; i.e., ∀g ∈ G, N/≈g = N. (v) C`(G) is an N-torsor over C`(Q) (along Φ). (vi) For p, q ∈ Q, it holds
pq = qp (in Q) =⇒ peqe = qepe (in G), −1 −1 for any liftings pe ∈ φ (p), qe ∈ φ (q). In other words, any liftings of elements that commute with each other also commute with each other. Proof. The equivalences (i) ⇐⇒ (iii) ⇐⇒ (vi) follow by definition, and (i) ⇐⇒ (ii) follows from (1.1) together with the fact that φ : G Q is surjective. We prove the remaining implications separately. • (iii) ⇐⇒ (iv): Since N is a normal subgroup of G, we have [g, k] ∈ N ∩ {[x, y] | x, y ∈ G} for any g ∈ G and k ∈ N. Then, the condition (iii) implies that [g, k] = e for any g ∈ G and k ∈ N, hence N ⊆ Z(G). For arbitrary g ∈ G, let k, j ∈ N such that k ≈g j. Then, from the definition of ≈g in (1.3), there is a ∈ G such that k = aj(ag)−1. Since N ⊆ Z(G), we have kj−1 = [a, g] ∈ N ∩ {[x, y] | x, y ∈ G} = {e}, hence k = j. We proved that (iii) =⇒ (iv). CLASSIFICATION OF THE CONJUGACY CLASSES OF SLf 2(R) 9
For the converse, assume (iii) is false. Then, there is a, b ∈ G such that [a, b] = k ∈ N, with k 6= e. This is the same as aba−1 = kb, which, by Lemma 1.7 (i), implies k ≈b e, therefore N/≈b 6= N. Hence, (iv) =⇒ (iii). • (iv) ⇐⇒ (v): The implication (iv) =⇒ (v) follows from Lemma 1.11 (v). For the converse, notice that in order for the action N y C`(G) to be well-defined, by −1 Lemma 1.10, we must have N ⊆ Z(G). Moreover, since N acts freely on Φ ([q]Q) for every [q]Q ∈ C`(Q), we have, for every g ∈ G and k 6= j ∈ N, that [kg]G 6= [jg]G, i.e., kg 6∼G jg. By Lemma 1.7 (i), this implies that for every `, m ∈ N, it holds ` ≈g m if and only if ` = m, which means N/≈g = N. We proved that (v) =⇒ (iv). Remark 1.13 (Examples of CP-extensions). Group extensions with the properties described in Theorem 1.12 are plenty in the literature, generally appearing in the form of item (vi). We give three classes of groups as examples. n n (i)H omeo] 0(T ) : For each integer n ≥ 1, write Homeo0(T ) for the topological group of self-homeomorphisms of the n-dimensional torus which are isotopic to the identity, equipped with the compact-open topology. This group admits a uni- n n n n versal covering Homeo] 0(T ), which, by writing T = R /Z , may be identified with those self-homeomorphisms of Rn which commute with translation by an el- n n n n ement of the lattice Z ⊆ R . The group extension Homeo] 0(T ) Homeo0(T ) is a CP-extension (cf. Remarque 2.12, p. 22 of Herman [8]). In Section 2, we will study in detail the particular case for n = 1, however, we follow different methods, closer to what is presented in Ghys [7].