Faithful inner-homogeneous RR-representations via HNN extensions with cardinal bound max⁡{ℵ0,∣R∣,∣G∣,∣V∣}\max\{\aleph_0,\allowbreak|R|,\allowbreak|G|,\allowbreak|V|\}

17.101

Problem

Does every representation of a group over a field embed in a homogeneous representation: one in which every isomorphism between finitely generated subrepresentations extends to an automorphism?

(A,U),(B,U′)≤fg(Γ,W),(θ,φ):(A,U)≅(B,U′)⟹(θ,φ)⊆Ψ∈Aut⁡(Γ,W)  ?(A,U),(B,U')\le_{\mathrm{fg}}(\Gamma,W),\quad (\theta,\varphi):(A,U)\cong(B,U') \quad\Longrightarrow\quad (\theta,\varphi)\subseteq\Psi\in\operatorname{Aut}(\Gamma,W)\;?

Finite generation of UU is over the group ring R[A]R[A]. The result below works over every nonzero commutative ring and makes the extending automorphism inner on both sorts.

Representations and the embedding theorem

Fix a commutative ring RR with 1≠01\ne0. Write X=(G,V,ρ)X=(G,V,\rho) for an RR-representation, where ρ:G→Aut⁡R(V)\rho:G\to\operatorname{Aut}_R(V), and abbreviate ρ(g)v\rho(g)v to gvgv. An embedding (i,j):X↪Y=(H,W,σ)(i,j):X\hookrightarrow Y=(H,W,\sigma) consists of

i:G↪H,j:V↪W,j(gv)=σ(i(g))j(v),(1) \tag{1} i:G\hookrightarrow H,\qquad j:V\hookrightarrow W, \qquad j(gv)=\sigma(i(g))j(v),

with ii a group homomorphism and jj an RR-linear map. All morphisms fix RR. The action is faithful if ker⁡ρ=1\ker\rho=1.

A subrepresentation S=(A,U)≤XS=(A,U)\le X consists of A≤GA\le G and an RR-submodule U≤VU\le V with AU⊆UAU\subseteq U. Write S≤fgXS\le_{\mathrm{fg}}X when

A=⟨a1,…,am⟩≤G,U=∑j=1nR[A]uj≤V(2) \tag{2} A=\langle a_1,\ldots,a_m\rangle\le G, \qquad U=\sum_{j=1}^n R[A]u_j\le V

for finite tuples (ai)(a_i) and (uj)(u_j). Finite generation is over R[A]R[A]: for a field RR, the cyclic R[Z]R[\mathbb Z]-module R[z,z−1]R[z,z^{-1}] has infinite RR-dimension.

An isomorphism f:S→∼Tf:S\xrightarrow{\sim}T consists of

θ:A→∼B,φ:U→∼RU′,φ(au)=θ(a)φ(u).(3) \tag{3} \theta:A\xrightarrow{\sim}B,\qquad \varphi:U\xrightarrow{\sim}_R U',\qquad \varphi(au)=\theta(a)\varphi(u).

Put

D(X)={(S,T,f):S,T≤fgX, f:S→∼T}.\mathcal D(X)=\{(S,T,f):S,T\le_{\mathrm{fg}}X,\ f:S\xrightarrow{\sim}T\}.

For t∈Gt\in G, the pair

(ct,Lt),ct(g)=tgt−1,Lt(v)=tv,(5) \tag{5} (c_t,L_t),\qquad c_t(g)=tgt^{-1},\qquad L_t(v)=tv,

is an automorphism of XX, since Lt(gv)=ct(g)Lt(v)L_t(gv)=c_t(g)L_t(v). The representation is inner homogeneous if each member of D(X)\mathcal D(X) is a restriction of some (ct,Lt)(c_t,L_t).

Theorem 1. Every RR-representation (G,V)(G,V) embeds in a faithful inner homogeneous representation (Γ,W)(\Gamma,W) with

∣Γ∣,∣W∣≤κ,κ=max⁡{ℵ0,∣R∣,∣G∣,∣V∣}.(6) \tag{6} |\Gamma|,|W|\le\kappa, \qquad \kappa=\max\{\aleph_0,|R|,|G|,|V|\}.

Thus, for (S,T,(θ,φ))∈D(Γ,W)(S,T,(\theta,\varphi))\in\mathcal D(\Gamma,W), some t∈Γt\in\Gamma satisfies

tat−1=θ(a)(a∈A),tu=φ(u)(u∈U).(7) \tag{7} tat^{-1}=\theta(a)\quad(a\in A), \qquad tu=\varphi(u)\quad(u\in U).

Plotkin’s Problem 17.101 asks for homogeneous extensions of group representations, with finite generation as in (2) (6). Aladova–Gvaramiya–Plotkin pose the fixed-ring version in (1, Definition 5.8 and Problem 5.11). Their discussion identifies the obstruction: extending GG and inducing VV need not extend a prescribed φ\varphi. Theorem 1 gives an inner extension on both sorts.

The group construction is due to Higman–Neumann–Neumann (4). Kegel’s permutation-action analogue already gives inner implementation and cardinal control (5, Theorem 1). For modules, one must also preserve the additive and scalar relations in VV. Dicks develops the relevant Mayer–Vietoris constructions (2; 3); his induced grading applies to arbitrary modules for group-ring embeddings (3, Theorem 16 and p. 455). Here the relation module has the form of (3, §5, equation (16)), and the tree argument proves im⁡δ∩(1⊗V)=0\operatorname{im}\delta\cap(1\otimes V)=0 for the prescribed linear relations.

The HNN group and its coset tree

Let A,B≤GA,B\le G and let θ:A→B\theta:A\to B be an isomorphism. Set

H=⟨G,t∣tat−1=θ(a) (a∈A)⟩,(8) \tag{8} H=\langle G,t\mid tat^{-1}=\theta(a)\ (a\in A)\rangle,

the quotient of G∗⟨t⟩G*\langle t\rangle by the normal closure of these relations.

Lemma 2. The canonical homomorphism G→HG\to H is injective. The graph with vertex set H/GH/G, edge set H/AH/A, and endpoints

hA:hG⟶ht−1G(9) \tag{9} hA:\quad hG\longrightarrow ht^{-1}G

is a tree.

Proof. For ε∈{+1,−1}\varepsilon\in\{+1,-1\} put

C+=B,C−=A,ψ+=θ−1,ψ−=θ.C_+=B,\quad C_-=A, \qquad \psi_+=\theta^{-1},\quad\psi_-=\theta.

Choose sets TεT_\varepsilon of representatives such that every g∈Gg\in G has a unique expression g=rcg=rc, with r∈Tεr\in T_\varepsilon and c∈Cεc\in C_\varepsilon, and such that 1∈Tε1\in T_\varepsilon represents CεC_\varepsilon. Consider formal strings

r1tε1⋯rmtεmg,ri∈Tεi,g∈G,(11) \tag{11} r_1t^{\varepsilon_1}\cdots r_mt^{\varepsilon_m}g, \quad r_i\in T_{\varepsilon_i},\quad g\in G,

subject to

i>1,εi−1=−εi⟹ri≠1.(12) \tag{12} i>1, \varepsilon_{i-1}=-\varepsilon_i \quad\Longrightarrow\quad r_i\ne1.

The case m=0m=0 is the string gg.

Write Pm=r1tε1⋯rmtεmP_m=r_1t^{\varepsilon_1}\cdots r_mt^{\varepsilon_m}, with P0=1P_0=1. For g=rcg=rc, r∈Tεr\in T_\varepsilon, c∈Cεc\in C_\varepsilon, define

(Pmg)⋅b=Pm(gb)(b∈G),(Pmg)⋅tε={Pm−1rmψε(c),m>0, εm=−ε, r=1,Pmrtεψε(c),otherwise.\begin{align*} (P_mg)\cdot b&=P_m(gb)\qquad(b\in G),\\ (P_mg)\cdot t^\varepsilon &=\begin{cases} P_{m-1}r_m\psi_\varepsilon(c), &m>0,\ \varepsilon_m=-\varepsilon,\ r=1,\\ P_mrt^\varepsilon\psi_\varepsilon(c),&\text{otherwise}. \end{cases} \end{align*}

Both outputs satisfy (12). Since ψε(c)∈C−ε\psi_\varepsilon(c)\in C_{-\varepsilon}, reversing a noncancelling step deletes the added block and restores rψ−εψε(c)=rcr\psi_{-\varepsilon}\psi_\varepsilon(c)=rc. After a cancelling step, rmψε(c)r_m\psi_\varepsilon(c) is already in T−εC−εT_{-\varepsilon}C_{-\varepsilon} form; reversal restores the deleted block. A second cancellation would violate (12). Hence the two stable-letter operations are inverse.

For a∈Aa\in A, write the tail as rcrc, c∈Bc\in B. Appending tata and θ(a)t\theta(a)t uses the same representative rr and the same case above; in either case the new tail is θ−1(c)a\theta^{-1}(c)a, preceded by rmr_m in the cancelling case. Thus

(w⋅t)⋅a=(w⋅θ(a))⋅t.(w\cdot t)\cdot a=(w\cdot\theta(a))\cdot t.

Together with (w⋅b)⋅b′=w⋅(bb′)(w\cdot b)\cdot b'=w\cdot(bb'), this defines a right HH-action on the normal strings.

Every operation is valid in HH, by

ctε=tεψε(c)(c∈Cε).ct^\varepsilon=t^\varepsilon\psi_\varepsilon(c) \qquad(c\in C_\varepsilon).

Reading a normal string from 11 returns that string. Equal elements of HH therefore have equal normal strings, so G→HG\to H is injective. In particular, a word

g0tε1g1⋯tεmgm,m>0,(16) \tag{16} g_0t^{\varepsilon_1}g_1\cdots t^{\varepsilon_m}g_m, \qquad m>0,

with no segment tat−1tat^{-1}, a∈Aa\in A, or t−1btt^{-1}bt, b∈Bb\in B, lies outside GG. Before a first cancellation, the coefficient carried past tεit^{\varepsilon_i} lies in C−εiC_{-\varepsilon_i}. For opposite successive signs, multiplication by this coefficient preserves membership in C−εiC_{-\varepsilon_i}; cancellation would require one of the excluded segments. The normal string consequently retains all mm stable letters.

The endpoints in (9) are well-defined, because

haG=hG,hat−1G=ht−1θ(a)G=ht−1G(a∈A).haG=hG, \qquad hat^{-1}G=ht^{-1}\theta(a)G=ht^{-1}G \qquad(a\in A).

They are distinct, since t∉Gt\notin G. The graph is connected: its edges permit multiplication by a stable letter in either direction, while GG fixes the base vertex, and G∪{t}G\cup\{t\} generates HH.

More explicitly, a step from wGwG can be written wG→wgtεGwG\to wgt^\varepsilon G with g∈Gg\in G. For ε=−1\varepsilon=-1 the edge is wgAwgA; for ε=+1\varepsilon=+1 it is wgtAwgtA traversed backwards. Two consecutive steps containing tat−1tat^{-1}, a∈Aa\in A, traverse the same edge in opposite directions. The same holds for t−1btt^{-1}bt, b∈Bb\in B, using bt=tθ−1(b)bt=t\theta^{-1}(b). A nonbacktracking path consequently gives a word of the form (16) with none of the excluded segments. A positive length closed path based at GG would give such a word lying in GG, contradicting the reduced-word assertion. By translating the base vertex, the same holds at every vertex. The graph is a tree. ◻

Modules on a forest

Let F\mathcal F be an oriented forest, with vertex set V\mathcal V, edge set E\mathcal E, and initial and terminal vertices o(e)o(e) and q(e)q(e) for an edge ee. Assign an RR-module MvM_v to each vertex and an RR-module NeN_e to each edge, together with injective linear maps

αe:Ne⟶Mo(e),βe:Ne⟶Mq(e).(18) \tag{18} \alpha_e:N_e\longrightarrow M_{o(e)}, \qquad \beta_e:N_e\longrightarrow M_{q(e)}.

Write ιv:Mv→⨁w∈VMw\iota_v:M_v\to\bigoplus_{w\in\mathcal V}M_w for the summand inclusion. The boundary map is

∂:⨁e∈ENe⟶⨁v∈VMv,∂(x)=∑e(ιq(e)βe(xe)−ιo(e)αe(xe)).(19) \tag{19} \partial:\bigoplus_{e\in\mathcal E}N_e \longrightarrow\bigoplus_{v\in\mathcal V}M_v, \qquad \partial(x)=\sum_e\bigl(\iota_{q(e)}\beta_e(x_e) -\iota_{o(e)}\alpha_e(x_e)\bigr).

All sums in (19) are finite.

Lemma 3. For every v∈Vv\in\mathcal V,

ker⁡∂=0,im⁡∂∩ιv(Mv)=0.(20) \tag{20} \ker\partial=0,\qquad \operatorname{im}\partial\cap\iota_v(M_v)=0.

Consequently every canonical map Mv→coker⁡∂M_v\to\operatorname{coker}\partial is injective.

Proof. For 0≠x∈⨁eNe0\ne x\in\bigoplus_eN_e, put F=supp⁡xF=\operatorname{supp}x. A component of the finite forest FF containing an edge has two leaves, the endpoints of a longest simple path. At either leaf ww, let ee be its unique incident edge in FF. Then

(∂x)w={βe(xe),w=q(e),−αe(xe),w=o(e),(∂x)w≠0.\bigl(\partial x\bigr)_w= \begin{cases} \beta_e(x_e),&w=q(e),\\ -\alpha_e(x_e),&w=o(e), \end{cases} \qquad \bigl(\partial x\bigr)_w\ne0.

The last inequality uses xe≠0x_e\ne0 and the injectivity of the endpoint maps. Thus ∣supp⁡(∂x)∣≥2|\operatorname{supp}(\partial x)|\ge2, giving both ∂x≠0\partial x\ne0 and ∂x∉ιv(Mv)\partial x\notin\iota_v(M_v) for every vv. The induced map Mv→coker⁡∂M_v\to\operatorname{coker}\partial is therefore injective. ◻

The argument requires neither finite generation of the modules nor finiteness of F\mathcal F. In particular, it does not choose complements to the endpoint images in (18).

Realizing a prescribed subrepresentation isomorphism

Theorem 4. Let (G,V)(G,V) be an RR-representation and let (θ,φ):(A,U)→(B,U′)(\theta,\varphi):(A,U)\to(B,U') be an isomorphism of subrepresentations. There are a representation (H,V^,σ)(H,\widehat V,\sigma) with ker⁡σ=1\ker\sigma=1, an embedding (i,j):(G,V)→(H,V^)(i,j):(G,V)\to(H,\widehat V), and an element t∈Ht\in H satisfying

ti(a)t−1=i(θ(a)),tj(u)=j(φ(u))(a∈A, u∈U).(22) \tag{22} ti(a)t^{-1}=i(\theta(a)),\qquad tj(u)=j(\varphi(u)) \quad(a\in A,\ u\in U).

No finite-generation hypothesis is imposed, and

∣H∣,∣V^∣≤max⁡{ℵ0,∣R∣,∣G∣,∣V∣}.(23) \tag{23} |H|,|\widehat V|\le\max\{\aleph_0,|R|,|G|,|V|\}.

Proof. Take HH from (8). By Lemma 2 we may identify GG with its image in HH. Form the induced modules

M=R[H]⊗R[G]V,E=R[H]⊗R[A]U.(24) \tag{24} M=R[H]\otimes_{R[G]}V, \qquad E=R[H]\otimes_{R[A]}U.

Define an R[H]R[H]-linear map δ:E→M\delta:E\to M by

δ(h⊗u)=ht−1⊗φ(u)−h⊗u.(25) \tag{25} \delta(h\otimes u)=ht^{-1}\otimes\varphi(u)-h\otimes u.

To check balancing, let a∈Aa\in A. Then

δ(ha⊗u)=hat−1⊗φ(u)−ha⊗u=ht−1θ(a)⊗φ(u)−h⊗au=ht−1⊗φ(au)−h⊗au=δ(h⊗au).\begin{align*} \delta(ha\otimes u) &=hat^{-1}\otimes\varphi(u)-ha\otimes u\\ &=ht^{-1}\theta(a)\otimes\varphi(u)-h\otimes au\\ &=ht^{-1}\otimes\varphi(au)-h\otimes au =\delta(h\otimes au). \end{align*}

The formula is RR-linear and satisfies δ(kh⊗u)=kδ(h⊗u)\delta(kh\otimes u)=k\delta(h\otimes u), so it defines an R[H]R[H]-linear map.

For each x∈H/Gx\in H/G choose a representative rxr_x, with rG=1r_G=1. The right R[G]R[G]-module decomposition R[H]=⨁xrxR[G]R[H]=\bigoplus_x r_xR[G] gives

M=⨁x∈H/GVx,E=⨁y∈H/AUy,(27) \tag{27} M=\bigoplus_{x\in H/G}V_x, \qquad E=\bigoplus_{y\in H/A}U_y,

where each VxV_x is a copy of VV and each UyU_y is a copy of UU. These are decompositions of RR-modules; no basis of VV or UU is assumed.

Choose a representative sys_y for each y∈H/Ay\in H/A. Write

sy=rxg0,syt−1=rzg1,g0,g1∈G,s_y=r_xg_0, \qquad s_yt^{-1}=r_zg_1, \qquad g_0,g_1\in G,

where x=syGx=s_yG and z=syt−1Gz=s_yt^{-1}G are the endpoints of yy. The map from the summand UyU_y in (27) has exactly the two components

u⟼−g0u∈Vx,u⟼g1φ(u)∈Vz.(29) \tag{29} u\longmapsto-g_0u\in V_x, \qquad u\longmapsto g_1\varphi(u)\in V_z.

Both endpoint maps are injective: the inclusions U,U′≤VU,U'\le V are injective, φ\varphi is an isomorphism, and the actions of g0,g1g_0,g_1 are invertible. By Lemma 2, the indexing graph in (27) is a tree. Lemma 3 therefore gives

ker⁡δ=0,im⁡δ∩(1⊗V)=0.(30) \tag{30} \ker\delta=0, \qquad \operatorname{im}\delta\cap(1\otimes V)=0.

Set

V^0=M/im⁡δ,j0(v)=[1⊗v].(31) \tag{31} \widehat V_0=M/\operatorname{im}\delta, \qquad j_0(v)=[1\otimes v].

Equivalently, let D=⨁h∈HVD=\bigoplus_{h\in H}V with summand maps ehe_h, and let C≤DC\leq D be the RR-submodule generated by

ehg(v)−eh(gv),eht(u)−eh(φ(u))(h∈H, g∈G, v∈V, u∈U).(32) \tag{32} e_{hg}(v)-e_h(gv),\qquad e_{ht}(u)-e_h(\varphi(u)) \quad(h\in H,\ g\in G,\ v\in V,\ u\in U).

Then V^0≅D/C\widehat V_0\cong D/C. Indeed, quotienting by the first family gives MM through eh(v)↦h⊗ve_h(v)\mapsto h\otimes v; the inverse is the balanced map h⊗v↦[eh(v)]h\otimes v\mapsto[e_h(v)]. Under this identification, the second family is the negative of δ(ht⊗u)\delta(ht\otimes u) and spans im⁡δ\operatorname{im}\delta. The submodule im⁡δ\operatorname{im}\delta is HH-invariant, so V^0\widehat V_0 is an R[H]R[H]-module. Equation (30) makes j0j_0 injective. Equivariance follows from

gj0(v)=[g⊗v]=[1⊗gv]=j0(gv).gj_0(v)=[g\otimes v]=[1\otimes gv]=j_0(gv).

For the stable letter,

0=[δ(t⊗u)]=[1⊗φ(u)]−[t⊗u]=j0(φ(u))−tj0(u).(34) \tag{34} 0=[\delta(t\otimes u)] =[1\otimes\varphi(u)]-[t\otimes u] =j_0(\varphi(u))-tj_0(u).

The conjugation equation comes from the presentation of HH.

Put

V^=V^0⊕R[H],j(v)=(j0(v),0).(35) \tag{35} \widehat V=\widehat V_0\oplus R[H], \qquad j(v)=(j_0(v),0).

For the diagonal action σ\sigma, the regular basis gives

σ(h)(0,[1])=(0,[h])≠(0,[1])(h≠1).\sigma(h)(0,[1])=(0,[h])\ne(0,[1])\qquad(h\ne1).

Thus ker⁡σ=1\ker\sigma=1, and (34) still holds for jj.

For κ\kappa as in (23), finite words and finite sums give

∣H∣≤max⁡{ℵ0,∣G∣}≤κ,∣M∣,∣E∣,∣R[H]∣≤max⁡{ℵ0,∣R∣,∣H∣,∣V∣}≤κ,∣V^∣≤∣M∣ ∣R[H]∣≤κ.\begin{align*} |H|&\le\max\{\aleph_0,|G|\}\le\kappa,\\ |M|,|E|,|R[H]|&\le\max\{\aleph_0,|R|,|H|,|V|\}\le\kappa,\\ |\widehat V|&\le |M|\,|R[H]|\le\kappa. \end{align*}

 ◻

The construction gives the exact sequence

0⟶R[H]⊗R[A]U→ δ R[H]⊗R[G]V⟶V^0⟶0,(38) \tag{38} 0\longrightarrow R[H]\otimes_{R[A]}U \xrightarrow{\ \delta\ }R[H]\otimes_{R[G]}V \longrightarrow\widehat V_0\longrightarrow0,

with the additional intersection property (30). The target V^0\widehat V_0 need not itself be induced from GG. For example, if G=A=1G=A=1, U=V=RU=V=R, and φ=id\varphi=\mathrm{id}, then

M=R[t,t−1],V^0=R[t,t−1]/(t−1)≅R.M=R[t,t^{-1}],\qquad \widehat V_0=R[t,t^{-1}]/(t-1)\cong R.

Directed limits and cardinal bounds

Let II be a nonempty directed poset. Consider a system Xi=(Gi,Vi,ρi)X_i=(G_i,V_i,\rho_i) with embeddings eij=(aij,bij)e_{ij}=(a_{ij},b_{ij}) satisfying

eii=1Xi,ejkeij=eik(i≤j≤k).e_{ii}=1_{X_i},\qquad e_{jk}e_{ij}=e_{ik}\qquad(i\le j\le k).

Lemma 5. Every directed system of embeddings has a limit representation X=(G,V,ρ)X=(G,V,\rho) and compatible embeddings ei:Xi→Xe_i:X_i\to X such that every finite collection of group and module elements in XX belongs to the image of one XiX_i. Moreover,

(∀i∈I, ker⁡ρi=1)⟹ker⁡ρ=1.(\forall i\in I,\ \ker\rho_i=1)\quad\Longrightarrow\quad\ker\rho=1.

If κ\kappa is infinite and

∣I∣≤κ,∣Gi∣,∣Vi∣≤κ(i∈I),(42) \tag{42} |I|\le\kappa,\qquad |G_i|,|V_i|\le\kappa\quad(i\in I),

then ∣G∣,∣V∣≤κ|G|,|V|\le\kappa.

Proof. On the two disjoint unions, put

(i,g)∼(j,h)  ⟺  ∃k≥i,j: aik(g)=ajk(h),(i,u)∼(j,v)  ⟺  ∃k≥i,j: bik(u)=bjk(v).\begin{align*} (i,g)\sim(j,h)&\iff\exists k\ge i,j:\ a_{ik}(g)=a_{jk}(h),\\ (i,u)\sim(j,v)&\iff\exists k\ge i,j:\ b_{ik}(u)=b_{jk}(v). \end{align*}

Directedness and compatibility make these equivalence relations. Set G=(∐iGi)/∼G=(\coprod_iG_i)/{\sim} and V=(∐iVi)/∼V=(\coprod_iV_i)/{\sim}. For k≥i,jk\ge i,j, define

[i,g][j,h]=[k,aik(g)ajk(h)],[i,g]−1=[i,g−1],[i,u]+[j,v]=[k,bik(u)+bjk(v)],r[i,u]=[i,ru],\begin{align*} [i,g][j,h]&=[k,a_{ik}(g)a_{jk}(h)],&[i,g]^{-1}&=[i,g^{-1}],\\ [i,u]+[j,v]&=[k,b_{ik}(u)+b_{jk}(v)],&r[i,u]&=[i,ru], \end{align*}

and

[i,g] [j,v]=[k,aik(g)bjk(v)].(45) \tag{45} [i,g]\,[j,v]=[k,a_{ik}(g)b_{jk}(v)].

Passing to a common upper stage proves well-definedness and the representation identities. Injectivity follows from

[i,g]=[i,h] ⟹ aik(g)=aik(h) ⟹ g=h[i,g]=[i,h]\ \Longrightarrow\ a_{ik}(g)=a_{ik}(h)\ \Longrightarrow\ g=h

for some k≥ik\ge i, and the analogous calculation for bikb_{ik}. Finite sets have representatives in one stage. A compatible family (fi)(f_i) factors uniquely by [i,x]↦fi(x)[i,x]\mapsto f_i(x) on each sort.

If a≠ba\ne b in GiG_i and ker⁡ρi=1\ker\rho_i=1, choose v∈Viv\in V_i with av≠bvav\ne bv. Then

[i,a][i,v]=[i,av]≠[i,bv]=[i,b][i,v],[i,a][i,v]=[i,av]\ne[i,bv]=[i,b][i,v],

which proves ker⁡ρ=1\ker\rho=1.

Each quotient has cardinality at most that of the corresponding disjoint union. Under (42),

∣G∣≤∑i∈I∣Gi∣≤∣I∣κ≤κ,∣V∣≤∑i∈I∣Vi∣≤∣I∣κ≤κ.|G|\le\sum_{i\in I}|G_i|\le|I|\kappa\le\kappa, \qquad |V|\le\sum_{i\in I}|V_i|\le|I|\kappa\le\kappa.

These inequalities use only κ⋅κ=κ\kappa\cdot\kappa=\kappa for infinite κ\kappa. ◻

Lemma 6. Suppose an isomorphism between subrepresentations of XX is implemented in an extension YY by t∈GYt\in G_Y. For every embedding (a,b):Y→Z(a,b):Y\to Z, its transported isomorphism is implemented by a(t)a(t).

Proof. Apply aa and bb to the two implementation equations. The required identities follow from

a(tgt−1)=a(t)a(g)a(t)−1,b(tv)=a(t)b(v).a(tgt^{-1})=a(t)a(g)a(t)^{-1},\qquad b(tv)=a(t)b(v).

 ◻

Lemma 7. If κ\kappa is infinite and X=(G,V)X=(G,V) satisfies ∣G∣,∣V∣≤κ|G|,|V|\le\kappa, then

∣D(X)∣≤κ.|\mathcal D(X)|\le\kappa.

Proof. By (2),

∣{S:S≤fgX}∣≤∑m,n<ω∣G∣m∣V∣n≤κ.\bigl|\{S:S\le_{\mathrm{fg}}X\}\bigr| \le\sum_{m,n<\omega}|G|^m|V|^n\le\kappa.

Fix S=(⟨a1,…,am⟩,∑j=1nR[A]uj)S=(\langle a_1,\ldots,a_m\rangle,\sum_{j=1}^nR[A]u_j). An isomorphism with domain SS is determined by

(θ(a1),…,θ(am),φ(u1),…,φ(un))∈Gm×Vn:(52) \tag{52} \bigl(\theta(a_1),\ldots,\theta(a_m), \varphi(u_1),\ldots,\varphi(u_n)\bigr)\in G^m\times V^n:

the group images determine θ\theta, and for finite Fj⊆AF_j\subseteq A,

φ(∑j=1n∑a∈Fjrj,aauj)=∑j=1n∑a∈Fjrj,aθ(a)φ(uj).(53) \tag{53} \varphi\left(\sum_{j=1}^n\sum_{a\in F_j}r_{j,a}au_j\right) =\sum_{j=1}^n\sum_{a\in F_j}r_{j,a}\theta(a)\varphi(u_j).

The codomain is (θ(A),φ(U))(\theta(A),\varphi(U)). There are at most κm+n≤κ\kappa^{m+n}\le\kappa choices for (52), hence ∣D(X)∣≤κ⋅κ=κ|\mathcal D(X)|\le\kappa\cdot\kappa=\kappa. ◻

Proposition 8. Suppose κ\kappa is infinite and

X=(G,V,ρ),∣R∣,∣G∣,∣V∣≤κ,ker⁡ρ=1.X=(G,V,\rho),\qquad |R|,|G|,|V|\le\kappa,\qquad\ker\rho=1.

There is an embedding X↪X+=(G+,V+,ρ+)X\hookrightarrow X^+=(G^+,V^+,\rho^+) with

∣G+∣,∣V+∣≤κ,ker⁡ρ+=1,|G^+|,|V^+|\le\kappa,\qquad\ker\rho^+=1,

in which every member of D(X)\mathcal D(X) is implemented by an element of G+G^+.

Proof. By Lemma 7, write D(X)={dα:α<λ}\mathcal D(X)=\{d_\alpha:\alpha<\lambda\}, ∣λ∣≤κ|\lambda|\le\kappa. Construct compatible embeddings by transfinite recursion:

X0=X,Xα↪Xα+1,Xβ=lim→⁡α<βXα(β a nonzero limit).X_0=X,\qquad X_\alpha\hookrightarrow X_{\alpha+1},\qquad X_\beta=\varinjlim_{\alpha<\beta}X_\alpha \quad(\beta\text{ a nonzero limit}).

At a successor, transport dαd_\alpha along X↪XαX\hookrightarrow X_\alpha and realize it by Theorem 4; at a limit, use Lemma 5. Induction gives

∣GXα∣,∣VXα∣≤κ,ker⁡ρXα=1(α≤λ),|G_{X_\alpha}|,|V_{X_\alpha}|\le\kappa, \qquad\ker\rho_{X_\alpha}=1\qquad(\alpha\le\lambda),

since ∣β∣κ≤κ|\beta|\kappa\le\kappa at every limit β≤λ\beta\le\lambda. Lemma 6 preserves all previous realizations, so X+=XλX^+=X_\lambda suffices; for λ=0\lambda=0, take X+=XX^+=X. ◻

Only κ⋅κ=κ\kappa\cdot\kappa=\kappa is used; κ\kappa need not be regular.

The homogeneous representation

Proof of Theorem 1. Let κ\kappa be as in (6). The representation

X0=(G,V⊕R[G])X_0=(G,V\oplus R[G])

with diagonal action contains (G,V)(G,V) by v↦(v,0)v\mapsto(v,0) and satisfies ker⁡ρX0=1\ker\rho_{X_0}=1 and ∣GX0∣,∣VX0∣≤κ|G_{X_0}|,|V_{X_0}|\le\kappa. Apply Proposition 8 successively to obtain

X0→e0X1→e1X2→e2⋯ ,(59) \tag{59} X_0\xrightarrow{e_0}X_1\xrightarrow{e_1}X_2 \xrightarrow{e_2}\cdots,

where D(Xn)\mathcal D(X_n) is realized in Xn+1X_{n+1}. Lemma 5 gives an extension (Γ,W,σ)(\Gamma,W,\sigma) with

ker⁡σ=1,∣Γ∣,∣W∣≤κ.\ker\sigma=1,\qquad |\Gamma|,|W|\le\kappa.

Identify each stage with its image in this direct limit. These images are nested, so

Γ=⋃n<ωGn,W=⋃n<ωVn.(61) \tag{61} \Gamma=\bigcup_{n<\omega}G_n, \qquad W=\bigcup_{n<\omega}V_n.

Let (S,T,(θ,φ))∈D(Γ,W)(S,T,(\theta,\varphi))\in\mathcal D(\Gamma,W), with S=(A,U)S=(A,U) and T=(B,U′)T=(B,U'). Choose finite group generating sets SA,SBS_A,S_B and finite module generating sets TU,TU′T_U,T_{U'} over R[A],R[B]R[A],R[B], respectively. By Lemma 5, some XnX_n contains all four finite sets. Closure of that stage under its operations gives

A=⟨SA⟩≤Gn,B=⟨SB⟩≤Gn,U=R[A]TU≤Vn,U′=R[B]TU′≤Vn.(62) \tag{62} A=\langle S_A\rangle\le G_n, \quad B=\langle S_B\rangle\le G_n, \quad U=R[A]T_U\le V_n, \quad U'=R[B]T_{U'}\le V_n.

The stage actions are the restrictions of the limit action, so (S,T,(θ,φ))∈D(Xn)(S,T,(\theta,\varphi))\in\mathcal D(X_n). Choose the implementing t∈Gn+1t\in G_{n+1}. By Lemma 6, (7) holds in (Γ,W)(\Gamma,W). The automorphism (ct,Lt)(c_t,L_t), with inverse (ct−1,Lt−1)(c_{t^{-1}},L_{t^{-1}}), extends the given isomorphism. ◻

Corollary 9. If RR is countable, every representation with countable group and module embeds in a countable inner homogeneous representation (Γ,W,σ)(\Gamma,W,\sigma) with ker⁡σ=1\ker\sigma=1.

Proof. The cardinal κ\kappa in (6) is ℵ0\aleph_0. ◻

Corollary 10. Let KK be a field. Every representation of a group GG on a KK-vector space VV embeds in an inner homogeneous representation (Γ,W,σ)(\Gamma,W,\sigma) satisfying

ker⁡σ=1,∣Γ∣,∣W∣≤max⁡{ℵ0,∣K∣,∣G∣,∣V∣}.\ker\sigma=1,\qquad |\Gamma|,|W|\le\max\{\aleph_0,|K|,|G|,|V|\}.

Proof. Apply Theorem 1 with R=KR=K. ◻

References

Preprint · Lean (GitHub)

  1. E. V. Aladova, A. A. Gvaramiya and B. I. Plotkin, Logic in representations of groups, Algebra and Logic 51 (2012), 1–27. doi:10.1007/s10469-012-9167-8. Russian original: Algebra i Logika 51 (2012), 3–40, MathNet:al520.
  1. W. Dicks, Mayer–Vietoris presentations over colimits of rings, Proc. London Math. Soc. (3) 34 (1977), no. 3, 557–576. doi:10.1112/plms/s3-34.3.557.
  1. W. Dicks, The HNN construction for rings, J. Algebra 81 (1983), no. 2, 434–487. doi:10.1016/0021-8693(83)90198-9.
  1. G. Higman, B. H. Neumann and H. Neumann, Embedding theorems for groups, J. London Math. Soc. 24 (1949), no. 4, 247–254. doi:10.1112/jlms/s1-24.4.247.
  1. O. H. Kegel, Examples of highly transitive permutation groups, Rend. Sem. Mat. Univ. Padova 63 (1980), 295–300.
  1. E. I. Khukhro and V. D. Mazurov (eds.), Unsolved Problems in Group Theory: The Kourovka Notebook, 21st ed., Problem 17.101. arXiv:1401.0300v46.