This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

\newaliascnt

lemmatheorem \aliascntresetthelemma \newaliascntcortheorem \aliascntresetthecor \newaliascntproptheorem \aliascntresettheprop \newaliascntcontheorem \aliascntresetthecon \newaliascntdefntheorem \aliascntresetthedefn \newaliascntremarktheorem \aliascntresettheremark \newaliascntclaimremark \aliascntresettheclaim \newaliascntquestionremark \aliascntresetthequestion

Virtually special embeddings of integral Lorentzian lattices

Michelle Chu Department of Mathematics, Statistics, and Computer Science, University of Illinois at Chicago, Chicago, IL 60607, USA michu@uic.edu
Abstract.

The automorphism groups of integral Lorentzian lattices act by isometries on hyperbolic space with finite covolume. In the case of reflective integral lattices, the automorphism groups are commensurable to arithmetic hyperbolic reflection groups. However, for a fixed dimension, there is only finitely many reflective integral Lorentzian lattices, and these can only occur in small dimensions. The goal of this note is to construct embeddings of low-dimensional integral Lorentzian lattices into unimodular Lorentzian lattices associated to right-angled reflection groups. As an application, we construct many discrete groups of Isom(n)\mathrm{Isom}(\mathbb{H}^{n}) for small nn which are C-special in the sense of Haglund-Wise.

1. Introduction

Given a finite volume polyhedron PP in hyperbolic space n\mathbb{H}^{n}, let Γ\Gamma be the group generated by the reflections on the sides of PP. If the action of Γ\Gamma tiles n\mathbb{H}^{n} without interiors of copies of PP overlapping, we say that Γ\Gamma is a hyperbolic reflection group and its fundamental polyhedron PP is a hyperbolic Coxeter polyhedron. The quotient n/Γ\mathbb{H}^{n}/\Gamma is a finite-volume hyperbolic orbifold.

The theory of hyperbolic reflection groups provides many examples of finite volume hyperbolic orbifolds. However, in higher dimensions, these cease to exist [Vin84, Kho86, Pro86]. Another way to construct finite volume hyperbolic orbifolds in any dimension is as quotients of HnH^{n} by the automorphism groups Aut(L)\mathrm{Aut}(L) of Lorentzian lattices LL. These automorphism groups are examples of arithmetic groups of simplest type in Isom(n)\mathrm{Isom}(\mathbb{H}^{n}). If the subgroup generated by reflections has finite index in Aut(L)\mathrm{Aut}(L), we say the lattice LL is reflective. Such a subgroup is an arithmetic hyperbolic reflection group.

In this note we construct embeddings of lattices into unimodular lattices of higher dimension. The Lorentzian unimodular lattices 𝐈n,1\mathbf{I}_{n,1} are reflective for 2n192\leq n\leq 19 [Vin72, KV78]. Furthermore, for 2n82\leq n\leq 8, these are associated to reflection groups of hyperbolic right-angled polyhedra, which are geometric right-angled Coxeter groups [PV05]. Right-angled Coxeter groups, or RACGs, are particularly interesting because they have many nice properties which are inherited by their subgroups. For example, virtually embedding hyperbolic 3-manifold groups into RACGs has determined the virtual Haken and the virtual fibering conjectures for all finite volume hyperbolic 3-manifolds as well as LERFness of their fundamental groups [Ago13, Wis11]. In the sense of [HW08] we say a group is C-special if it embeds as a quasi-convex subgroup of a RACG.

We apply the lattice embeddings together with the explicit relationship between the unimodular lattices 𝐈n,1\mathbf{I}_{n,1} and RACGs given by [ERT12] to construct many examples of C-special hyperbolic manifold groups in dimension 3 and 4. The following theorem extends and improves the results in [Chu19] (see also [Chu18]) and [DMP18, Theorem 2.6].

Theorem 1.1.

Let Γ\Gamma be an integral arithmetic group of simplest type in Isom(H3)\mathrm{Isom}(H^{3}) or Isom(H4)\mathrm{Isom}(H^{4}). Then Γ(2)\Gamma_{(2)}, the principal congruence subgroup of level 2, is compact C-special.

For fixed dimensions, the indices of the principal congruence subgroups of level 2 contained in integral arithmetic group of simplest type are uniformly bounded. Theorem 1.1 extends [Chu19, Theorem 1.2, Proposition 5.3][Chu18, Theorem 3.1, Proposition 4.3, Proposition 5.3] and also improves on the bounds for the value of DD found in [DMP18, Proposition 2.6] by removing the dependence on the discriminant. This gives a uniform bound for DD which is independent of anything to get a strengthening of [DMP18, Theorem 2.2]. Prior to these results, Bergeron-Haglund-Wise showed that given an arithmetic group of simplest type in O+(n,1)\mathrm{O}^{+}(n,1), there exist some mm such that the congruence subgroup of level mm is special [BHW11], but Theorem 1.1 shows that m=2m=2 is enough for the cases included in Theorem 1.1.

As a consequence of Theorem 1.1 we can give a slight improvement to the results in [Chu19].

Proposition \theprop.

Let dd be a square-free positive integer and 𝒪d\mathcal{O}_{d} the ring of integers in the quadratic imaginary field (d)\mathbb{Q}(\sqrt{-d}). The principal congruence subgroup PSL(2,𝒪d)(2)\mathrm{PSL}(2,\mathcal{O}_{d})_{(2)} embeds in a RACG and has index

[PSL(2,𝒪d):PSL(2,𝒪d)(2)]={48if d1,2mod(4)60if d3mod(8)36if d7mod(8).[\mathrm{PSL}(2,\mathcal{O}_{d}):\mathrm{PSL}(2,\mathcal{O}_{d})_{(2)}]=\begin{cases}48&\text{if }d\equiv 1,2\mod(4)\\ 60&\text{if }d\equiv 3\mod(8)\\ 36&\text{if }d\equiv 7\mod(8).\end{cases}

In particular, since the figure eight knot group intersects PSL(2,𝒪3)(2)\mathrm{PSL}(2,\mathcal{O}_{3})_{(2)} in a subgroup of index 10, we also get the following corollary.

Corollary \thecor.

The figure eight knot complement has a special cover of degree 10.

This note is organized as follows: In Section 2 we give the necessary preliminary background in integral lattices, their automorphism groups, and arithmetic groups of simplest type. In Section 3, inspired by the lattice gluings in [All18], we construct embeddings of integral lattices into unimodular lattices. In Section 4 we use these embeddings to prove Theorem 1.1. Finally, in Section 5 we give an explicit example.

Acknowledgments

The author thanks Daniel Allcock for introducing her to lattice gluing and encouraging this work. This work was supported by NSF grant DMS-1803094.

2. Preliminaries

2.1. Integral lattices

A lattice LL is a \mathbb{Z}-module equipped with a \mathbb{Q}-valued non-degenerate symmetric bilinear form (,)(\cdot,\cdot) on the vector space V=LV=L\otimes\mathbb{Q}, called the inner product. LL is called Lorentzian if its inner product has signature +n1+^{n}-^{1}. The norm of a vector vv is its inner product with itself (v,v)(v,v). If the inner product of every pair of vectors in LL is \mathbb{Z}-valued, LL is called integral. In what follows, let LL be an integral lattice unless otherwise noted.

The dual of LL is the lattice L={vL:(v,L)}L^{*}=\{v\in L\otimes\mathbb{Q}:(v,L)\in\mathbb{Z}\}. Notice that LL is integral if and only if LLL\subset L^{*}. We define the discriminant group as Δ(L)=L/L\Delta(L)=L^{*}/L, a finite abelian group. We will refer to the minimal number of generators of Δ(L)\Delta(L) as rank(Δ(L))\mathrm{rank}(\Delta(L)).

The determinant of LL, or detL\det L is the determinant of an inner product matrix ALA_{L}, with respect to some \mathbb{Z}-basis of LL. It is independent of the choice of \mathbb{Z}-basis and in fact |detL|=|Δ(L)||\det L|=|\Delta(L)|. If LL^{\prime} is a sublattice of LL of index dd then detL=d2detL\det L^{\prime}=d^{2}\cdot\det L.

The \mathbb{Z}-valued inner product on LL extends to a \mathbb{Q}-valued inner product on LL^{*} and descends to a /\mathbb{Q}/\mathbb{Z}-valued inner product on Δ(L)\Delta(L).

An integral lattice LL is called strongly-square-free, denoted by SSF, if the rank of Δ(L)\Delta(L) is at most 12dim(L)\frac{1}{2}\dim(L) and every invariant factor of Δ(L)\Delta(L) is square-free. In other words, Δ(L)\Delta(L) is a direct product of at most 12dim(L)\frac{1}{2}\dim(L)-many finite cyclic subgroups, each of square-free order. An integral lattice is called unimodular whenever Δ(L)\Delta(L) is trivial.

Lattices may also be defined more generally over totally real number fields.

2.2. Automorphisms and arithmetic groups

Let kk be a totally real number field with ring of integers 𝒪k\mathcal{O}_{k} and let ff be a quadratic form of signature +n1+^{n}-^{1} defined over kk such that for every non-identity embedding σ:k\sigma:k\hookrightarrow\mathbb{R}, the form fσf^{\sigma} is positive definite. Let O(f;)\mathrm{O}(f;\mathbb{R}) denote the orthogonal group that preserves ff and O+(f;)\mathrm{O}^{+}(f;\mathbb{R}) its index-two subgroup which is time-orientation preserving. Then the group Γ=O+(f;𝒪k)=O+(f;)GLn+1(𝒪k)\Gamma=\mathrm{O}^{+}(f;\mathcal{O}_{k})=\mathrm{O}^{+}(f;\mathbb{R})\cap\mathrm{GL}_{n+1}(\mathcal{O}_{k}) is a finite-covolume discrete subgroup of O+(f;)\mathrm{O}^{+}(f;\mathbb{R}) which is identified with Isom(n)\mathrm{Isom}(\mathbb{H}^{n}) via the hyperboloid model. The field kk is called the field of definition for Γ\Gamma. Any discrete subgroup of Isom(n)\mathrm{Isom}(\mathbb{H}^{n}) which is commensurable to some such O+(f;𝒪k)\mathrm{O}^{+}(f;\mathcal{O}_{k}) is called arithmetic of simplest type.

If LL is an integral Lorentzian lattice, then its automorphism group is the group

Aut(L)\displaystyle\mathrm{Aut}(L) ={gGL(V)|Lg=L and fL(xg,yg)=fL(x,y) for all x,yL}\displaystyle=\{g\in\mathrm{GL}(V)|Lg=L\text{ and }f_{L}(xg,yg)=f_{L}(x,y)\text{ for all }x,y\in L\}
={gGLn+1()|gALgtr=AL}\displaystyle=\{g\in\mathrm{GL}_{n+1}(\mathbb{Z})|gA_{L}g^{tr}=A_{L}\}
=O(fL;).\displaystyle=\mathrm{O}(f_{L};\mathbb{Z}).

We call Aut(L)\mathrm{Aut}(L) and any finite index subgroup of it an integral arithmetic group of simplest type.

The principal congruence subgroup of level mm in Aut(L)\mathrm{Aut}(L) is the subgroup

Aut(L)(m)\displaystyle\mathrm{Aut}(L)_{(m)} ={gGL(V)|vgvmodm for all xL}\displaystyle=\{g\in\mathrm{GL}(V)|vg\equiv v\mod m\text{ for all }x\in L\}
={gGLn+1()|gALgtr=AL and gIn+1modm}\displaystyle=\{g\in\mathrm{GL}_{n+1}(\mathbb{Z})|gA_{L}g^{tr}=A_{L}\text{ and }g\equiv I_{n+1}\mod m\}
=O(fL;)(m)\displaystyle=\mathrm{O}(f_{L};\mathbb{Z})_{(m)}

where In+1I_{n+1} is the (n+1)×(n+1)(n+1)\times(n+1) identity matrix.

If n1n-1 is not divisible by 8, there is, up to isomorphism over \mathbb{Z}, a unique unimodular Lorentzian lattice of signature +n1+^{n}-^{1} denoted 𝐈n,1\mathbf{I}_{n,1}. Let qnq_{n} be the standard Lorentzian quadratic form

(2.1) qn:=x02+x12++xn2.q_{n}:=-x_{0}^{2}+x_{1}^{2}+\dots+x_{n}^{2}.

The unimodular lattice 𝐈n,1\mathbf{I}_{n,1} has automorphism group Aut(𝐈n,1)=O(qn,)Aut(\mathbf{I}_{n,1})=\mathrm{O}(q_{n},\mathbb{Z}) and is reflective for n19n\leq 19 [Vin72, KV78].

2.3. Invariants and existence of integral lattices

This section assumes familiarity with Conway-Sloane pp-adic symbols [CS99, Chapter 15].

Over the pp-adic integers, a form ff associated to a padicp-adic lattice LpL_{p} can be decomposed as a direct sum

(2.2) f=f1pfpp2fp2qfqf=f_{1}\oplus pf_{p}\oplus p^{2}f_{p^{2}}\oplus\dots\oplus qf_{q}\oplus\dots

where qq is a pp-power and fqf_{q} is a pp-adic integral form with determinant prime to pp. This Jordan decomposition is unique for odd primes.

For pp odd, the pp-adic symbol of ff is the formal product of factors qϵqnqq^{\epsilon_{q}n_{q}} with

ϵq=(detfqp) and nq=dimfq\epsilon_{q}=\left(\frac{\det f_{q}}{p}\right)\text{ and }n_{q}=\dim f_{q}

where (αp)\left(\frac{\alpha}{p}\right) denotes the Kronecker symbol.

For p=2p=2, the 22-adic symbol of ff is the formal product of factors qtqϵqnqq^{\epsilon_{q}n_{q}}_{t_{q}} or qϵqnqq^{\epsilon_{q}n_{q}} where the former indicates fqf_{q} is of type I and the later indicates fqf_{q} is of type II and with

ϵq=(detfq2) , nq=dimfq and tq=oddity(fq)\epsilon_{q}=\left(\frac{\det f_{q}}{2}\right)\text{ , }n_{q}=\dim f_{q}\text{ and }t_{q}=\mathrm{oddity}(f_{q})

where the Kronecker symbol (a2)\left(\frac{a}{2}\right) is +1+1 if a±1mod8a\equiv\pm 1\mod 8 or 1-1 if a±3mod8a\equiv\pm 3\mod 8.

Unfortunately, the 2-adic symbol is not unique, since a 22-adic form can have essentially different Jordan decompositions. However, Conway-Sloane define an abreviated 22-adic symbol using compartments and trains. Two abreviated 22-adic symbols represent the same form if and only they are related by sign walking see [CS99, Chapter 15, §7.5].

By [CS99, Theorem 11, Chapter 15], there exist an integral lattice LL of determinant dd having specified local forms LpL_{p} and signature +rs+^{r}-^{s} if and only if the determinant condition, the oddity formula, and the Jordan blocks conditions displayed below hold.

  • 1.

    The determinant condition: for each pp, the ϵq\epsilon_{q} from the pp-adic symbol satisfy

    (2.3) ϵq=(ap)\prod\epsilon_{q}=\left(\frac{a}{p}\right)

    where det(L)=pαa\det(L)=p^{\alpha}a.

  • 2.

    The oddity formula:

    (2.4) signature(L)+p oddpexcess(Lp)oddity(K2)mod8\mathrm{signature}(L)+\sum_{p\text{ odd}}p\mathrm{-excess}(L_{p})\equiv\mathrm{oddity}(K_{2})\mod 8

    where

    signature(L)=rs,\mathrm{signature}(L)=r-s,
    pexcess(Lp)qnq(q1)+4#(odd powers q with ϵq=1),p\mathrm{-excess}(L_{p})\equiv\sum_{q}n_{q}(q-1)+4\cdot\#(\text{odd powers $q$ with $\epsilon_{q}=-1$}),
    and oddity(L2)=tq+4#(odd powers q with ϵq=1).\text{and }\mathrm{oddity}(L_{2})=\sum t_{q}+4\cdot\#(\text{odd powers $q$ with $\epsilon_{q}=-1$}).
  • 3.

    The Jordan blocks conditions: the 22-adic Jordan blocks satisfy the following

    (2.5) if type II, tq0mod8\text{if type II, }t_{q}\equiv 0\mod 8
    (2.6) if nq=1,{ϵq=+1 then tq±1mod8ϵq=1 then tq±3mod8\text{if }n_{q}=1,\begin{cases}\epsilon_{q}=+1\text{ then }t_{q}\equiv\pm 1\mod 8\\ \epsilon_{q}=-1\text{ then }t_{q}\equiv\pm 3\mod 8\end{cases}
    (2.7) if type I and nq=2,{ϵq=+1 then tq0 or ±2mod8ϵq=1 then tq4 or ±2mod8\text{if type I and }n_{q}=2,\begin{cases}\epsilon_{q}=+1\text{ then }t_{q}\equiv 0\text{ or }\pm 2\mod 8\\ \epsilon_{q}=-1\text{ then }t_{q}\equiv 4\text{ or }\pm 2\mod 8\end{cases}
    (2.8) and tqnqmod2.\text{and }t_{q}\equiv n_{q}\mod 2.

When working with the abbreviated 22-adic symbol, the Jordan blocks conditions on a compartment of total dimension at least 33 reduce to just one condition: the total oddity in the compartment has the same parity as its total dimension.

2.4. Some facts and observations

We state here some observations.

For odd pp, the pp-excess is always even.

If LL is SSF, then since the rank of any invariant factor is square-free, the pp-adic symbol for LpL_{p} will only contains terms for q=1q=1 and q=pq=p. Furthermore, since the rank of Δ(L)\Delta(L) is at most 12dim(L)\frac{1}{2}\dim(L), also np12dim(L)n_{p}\leq\frac{1}{2}\dim(L).

If we take LnegL^{\mathrm{neg}} to be as LL with all inner products negated, its local forms will change as follows. If pp is odd, then the pp-adic symbol for LpnegL^{\mathrm{neg}}_{p} is got from that of LpL_{p} by multiplying each superscript by ((1)nqp)\left(\frac{(-1)^{n_{q}}}{p}\right). The 22-adic symbol for LpnegL^{\mathrm{neg}}_{p} is got from that of LpL_{p} by negating each subscript.

If pdetLp\nmid\det L or if (1p)=1\left(\frac{-1}{p}\right)=1, then the pp-excess of LpL_{p} and LpnegL_{p}^{\mathrm{neg}} agree. If p|Lp|L and (1p)=1\left(\frac{-1}{p}\right)=1, then the pp-excess of LpL_{p} and LpnegL_{p}^{\mathrm{neg}} differ by 4mod84\mod 8.

3. Lattice embeddings

The goal of this section is to prove the following proposition.

Proposition \theprop.

Let LL be an integral lattice of signature +rs+^{r}-^{s} and let δ=rank(Δ(L))\delta=\mathrm{rank}(\Delta(L)). Let m=max{δ+1,3}m=\max\{\delta+1,3\}. Then LL embeds in a unimodular lattice of signature +r+ms+^{r+m}-^{s}.

The proof will be separated into two cases depending on the parity of det(L)\det(L). The main idea is to use a technique in [All18]. We will construct a lattice KK of signature +δ+10+^{\delta+1}-^{0} with det(K)=(1)sdet(L)\det(K)=(-1)^{s}\det(L) by specifying its local forms KpK_{p}, chosen such that there exist a group isomorphism ϕ:Δ(L)Δ(K)\phi:\Delta(L)\rightarrow\Delta(K) which negates norms and inner products. Gluing LL to KK along the graph of ϕ\phi will then result in a unimodular lattice.

3.1. Case 1: dd is odd

If det(L)\det(L) is odd, the following lemma holds regardless of whether LL is SSF.

Lemma \thelemma.

Let LL be an integral lattice of signature +rs+^{r}-^{s} with det(L)\det(L) odd and rank(Δ(L))=δ\mathrm{rank}(\Delta(L))=\delta. Let m=max{δ+1,3}m=\max\{\delta+1,3\}. Then LL embeds in a unimodular lattice of signature +r+ms+^{r+m}-^{s}.

Proof.

Assume LL is not unimodular and let d:=(1)sdet(L)d:=(-1)^{s}\det(L). Defined the local forms KpK_{p} as follows.

For odd pdp\nmid d, define KpK_{p} by 1(dp)m1^{(\frac{d}{p})m}.

For odd p|dp|d with d=pαad=p^{\alpha}a, define KpK_{p} by the product of qϵqnqq^{\epsilon_{q}n_{q}} where for q=pm>1q=p^{m}>1, the term qϵqnqq^{\epsilon_{q}n_{q}} matches that of LpnegL_{p}^{neg}, and where 1ϵ1n11^{\epsilon_{1}n_{1}} is chosen such that ϵq=(ap)\prod\epsilon_{q}=(\frac{a}{p}) and nq=m\sum n_{q}=m.

Let tm+p oddpexcess(Kp)t\equiv m+\sum_{p\text{ odd}}p\mathrm{-excess}(K_{p}). Since p oddpexcess(Kp)\sum_{p\text{ odd}}p\mathrm{-excess}(K_{p}) is even, tt will have the parity of mm. Define K2K_{2} by 1t(d2)m1^{(\frac{d}{2})m}_{t}.

With these choices of local forms, all conditions (missing) 2.3-(missing) 2.8 are satisfied. So there exist an integral lattice KK of signature +m0+^{m}-^{0} and determinant dd with the prescribed local forms.

Now each local form KpK_{p} differs from LpnegL_{p}^{neg} by a unimodular factor. We have that Δ(Kp)\Delta(K_{p}) and Δ(Lpneg)\Delta(L_{p}^{neg}) are isomorphic and correspond to the Sylow pp-subgroups of Δ(K)\Delta(K) and Δ(Lneg)\Delta(L^{neg}). It follows that there exist a group isomorphism ϕ:Δ(L)Δ(K)\phi:\Delta(L)\rightarrow\Delta(K) which negates norms and inner products. Let G={(x,ϕx)}G=\{(x,\phi x)\} be the graph of ϕ\phi. Then GG is a totally isotropic subgroup of Δ(LK)=Δ(L)Δ(K)\Delta(L\oplus K)=\Delta(L)\oplus\Delta(K), that is, the natural /\mathbb{Q}/\mathbb{Z}-valued inner product on Δ(LK)\Delta(L\oplus K) vanishes on GG.

Write LGKL\oplus_{G}K for the preimage of GG in (LK)=LK(L\oplus K)^{*}=L^{*}\oplus K^{*}. Since GG is totally isotropic, LGKL\oplus_{G}K is an integral lattice containing LKL\oplus K as a sublattice of index |G||G| and therefore its determinant is given by

(3.1) detLGK=detLK|G|2=d2(d)2=1.\det L\oplus_{G}K=\frac{\det L\oplus K}{|G|^{2}}=\frac{d^{2}}{(d)^{2}}=1.

So LGKL\oplus_{G}K is a unimodular lattice containing LL with orthogonal complement LL^{\perp} isomorphic to KK. ∎

3.2. Case 2: dd is even

Lemma \thelemma.

Let LL be an integral lattice of signature +rs+^{r}-^{s} with det(L)\det(L) even and rank(Δ(L))=δ\mathrm{rank}(\Delta(L))=\delta. Let m=max{δ+1,3}m=\max\{\delta+1,3\}. Then LL embeds in a unimodular lattice of signature +r+ms+^{r+m}-^{s}.

Proof.

Let d:=(1)sdet(L)d:=(-1)^{s}\det(L). Observe first that the SSF assumption guarantees that Δ(L2)\Delta(L_{2}) (the 22-Sylow subgroup of Δ(L)\Delta(L)) is (/2)α(\mathbb{Z}/2\mathbb{Z})^{\alpha} where d=2αad=2^{\alpha}a and 2a2\nmid a. This means that the natural 2/2\mathbb{Q}_{2}/\mathbb{Z}_{2}-valued norms and inner products in Δ(L2)\Delta(L_{2}) are in 122\frac{1}{2}\mathbb{Z}_{2}. In particular, negating norms and inner products in Δ(L2)\Delta(L_{2}) is trivial.

Defined the local forms KpK_{p} as follows.

For odd pdp\nmid d, define KpK_{p} by 1(dp)m1^{(\frac{d}{p})m}.

For odd p|dp|d with d=pαad=p^{\alpha}a, define KpK_{p} by the product of qϵqnqq^{\epsilon_{q}n_{q}} where for q=pm>1q=p^{m}>1, the term qϵqnqq^{\epsilon_{q}n_{q}} matches that of LpnegL_{p}^{neg}, and where 1ϵ1n11^{\epsilon_{1}n_{1}} is chosen such that ϵq=(ap)\prod\epsilon_{q}=(\frac{a}{p}) and nq=m\sum n_{q}=m.

Let tm+p oddpexcess(Kp)t\equiv m+\sum_{p\text{ odd}}p\mathrm{-excess}(K_{p}). Since p oddpexcess(Kp)\sum_{p\text{ odd}}p\mathrm{-excess}(K_{p}) is even, tt will have the parity of mm. Define K2K_{2} by the reduced 22-adic symbol [1(a2)(mα)2+α]t[1^{\left(\frac{a}{2}\right)(m-\alpha)}2^{+\alpha}]_{t} where d=2αad=2^{\alpha}a.

With these choices of local forms, all conditions (missing) 2.3-(missing) 2.8 are satisfied. So there exist an integral lattice KK of signature +m0+^{m}-^{0} and determinant dd with the prescribed local forms.

Now each local form KpK_{p} for p2p\neq 2 differs from LpnegL_{p}^{neg} by a unimodular factor. We then have that for all pp (including p=2p=2) Δ(Kp)\Delta(K_{p}) and Δ(Lpneg)\Delta(L_{p}^{neg}) are isomorphic and correspond to the Sylow pp-subgroups of Δ(K)\Delta(K) and Δ(Lneg)\Delta(L^{neg}). It follows that there exist a group isomorphism ϕ:Δ(L)Δ(K)\phi:\Delta(L)\rightarrow\Delta(K) which negates norms and inner products. Let G={(x,ϕx)}G=\{(x,\phi x)\} be the graph of ϕ\phi. Then GG is a totally isotropic subgroup of Δ(LK)=Δ(L)Δ(K)\Delta(L\oplus K)=\Delta(L)\oplus\Delta(K).

The remaining follows exactly as in the proof of § 3.1. ∎

4. Special subgroups

In this section we gather the necessary ingredients and prove Theorem 1.1 as a direct consequence of Section 4 and Theorem 4.1.

It can be shown that the automorphism group of a non-SSF lattice is always contained in the automorphism group of one which is SSF [Wat62, Wat75, All12]. For SSF lattices of dimension up to 55, the rank of their discriminant groups is at most 22. Therefore, the following corollary follows from Section 3 since LL\otimes\mathbb{R} is the orthogonal complement to the positive subgroups KK\otimes\mathbb{R} in the vector space (LGK)(L\oplus_{G}K)\otimes\mathbb{R}.

Corollary \thecor.

If LL is an integral lattice of signature +31+^{3}-^{1} (resp. +41+^{4}-^{1}), then Aut(L)\mathrm{Aut}(L) embeds as a geometrically finite subgroup of Aut(𝐈6,1)\mathrm{Aut}(\mathbf{I}_{6,1}) (resp. Aut(𝐈7,1)\mathrm{Aut}(\mathbf{I}_{7,1})) which preserves a copy of 3\mathbb{H}^{3} (resp. 4\mathbb{H}^{4}) in 6\mathbb{H}^{6} (resp 7\mathbb{H}^{7}).

The following theorem of Everitt-Ratcliffe-Tschantz provides the explicit relationship between the automorphism groups of the unimodular lattices 𝐈n,1\mathbf{I}_{n,1} for n8n\leq 8 and right-angled Coxeter groups. Recall that qn:=x02+x12++xn2q_{n}:=-x_{0}^{2}+x_{1}^{2}+\dots+x_{n}^{2} and Aut(𝐈n,1)=O(qn,)\mathrm{Aut}(\mathbf{I}_{n,1})=\mathrm{O}(q_{n},\mathbb{Z}).

Theorem 4.1 ([ERT12, Theorem 2.1]).

For 2n72\leq n\leq 7, O+(qn,)(2)\mathrm{O}^{+}(q_{n},\mathbb{Z})_{(2)} is a geometric RACG. It is the reflection group of an all-right hyperbolic polyhedron of dimension nn. The group O+(q8,)(2)\mathrm{O}^{+}(q_{8},\mathbb{Z})_{(2)} contains a geometric RACG as a subgroup of index 2. This subgroup is the reflection group of an all-right hyperbolic polyhedron of dimension 88.

Lemma \thelemma.

Let LL be an integral lattice and KK the corresponding integral lattice with LGKL\oplus_{G}K unimodular as constructed in the proof of Section 3. Then Aut(L)(2)\mathrm{Aut}(L)_{(2)} is contained in Aut(LGK)(2)\mathrm{Aut}(L\oplus_{G}K)_{(2)}.

Proof.

Recall that KK was constructed so that Δ(L)Δ(K)G\Delta(L)\cong\Delta(K)\cong G. An element γAut(L)\gamma\in\mathrm{Aut}(L) is in Aut(L)(2)\mathrm{Aut}(L)_{(2)} if and only if for any vLv\in L, vγ=v+2vv\gamma=v+2v^{\prime} where vv^{\prime} is some other element in LL, or equivalently, vγvmod2v\gamma\equiv v\mod 2. Similarly, γAut(LGK)\gamma\in\mathrm{Aut}(L\oplus_{G}K) if and only if for any vLGKv\in L\oplus_{G}K, vγvmod2v\gamma\equiv v\mod 2. If vLKv\in L\oplus K then clearly for γAut(L)\gamma\in Aut(L), vγvmod2v\gamma\equiv v\mod 2.

For the remainder of the proof, set γAut(L)(2)\gamma\in Aut(L)_{(2)} and uLGKLKu\in L\oplus_{G}K-L\oplus K. Suppose mm is the smallest positive integer so that mu=uLKmu=u^{\prime}\in L\oplus K. Then uγ=u+2xu^{\prime}\gamma=u^{\prime}+2x where xLKx\in L\oplus K. If vv is some other element of LKL\oplus K with uvu^{\prime}\perp v, then vγ=v+2yv\gamma=v+2y for some yLKy\in L\oplus K and since 0=(u,v)=(uγ,vγ)0=(u^{\prime},v)=(u^{\prime}\gamma,v\gamma),

0=(uγ,v+2y)=(uγ,v)+(muγ,2y)=(uγ,v)+2m(uγ,y).0=(u^{\prime}\gamma,v+2y)=(u^{\prime}\gamma,v)+(mu\gamma,2y)=(u^{\prime}\gamma,v)+2m(u\gamma,y).

Thus (uγ,v)0mod2m(u^{\prime}\gamma,v)\equiv 0\mod 2m. Also,

(uγ,u)=(u+2x,u)=(u,u)+(2x,mu)=(u,u)+2m(x,u).(u^{\prime}\gamma,u^{\prime})=(u^{\prime}+2x,u^{\prime})=(u^{\prime},u^{\prime})+(2x,mu)=(u^{\prime},u^{\prime})+2m(x,u).

Thus (uγ,u)(u,u)mod2m(u^{\prime}\gamma,u^{\prime})\equiv(u^{\prime},u^{\prime})\mod 2m. Therefore uγumod2mu^{\prime}\gamma\equiv u^{\prime}\mod 2m and since u=muu^{\prime}=mu we have uγumod2u\gamma\equiv u\mod 2. This shows γAut(LGK)(2)\gamma\in\mathrm{Aut}(L\oplus_{G}K)_{(2)}. ∎

Proof of Theorem 1.1.

The proof follows from Section 4, Theorem 4.1, and Section 4 since the lattice LL embeds as the orthogonal complement of the positive definite sublattice KK and this induces the embedding of Aut(L)\mathrm{Aut}(L) as a geometrically finite subgroup of Aut(LGK)\mathrm{Aut}(L\oplus_{G}K). ∎

We remark here that the principal congruence subgroups of level 2 have uniformly bounded index in integral arithmetic subgroups of simplest type in Isom(Hn)\mathrm{Isom}(H^{n}). Indeed, any such Γ\Gamma is contained in some O+(f;)GLn+1()\mathrm{O}^{+}(f;\mathbb{Z})\subset\mathrm{GL}_{n+1}(\mathbb{Z}) and thus [Γ:Γ(2)]|GLn+1(/2)|[\Gamma:\Gamma_{(2)}]\leq|\mathrm{GL}_{n+1}(\mathbb{Z}/2\mathbb{Z})|.

The Bianchi group PSL2(𝒪d)\mathrm{PSL}_{2}(\mathcal{O}_{d}) is contained in O+(fd;)\mathrm{O}^{+}(f_{d};\mathbb{Z}) where

fd={2x0x1+2x22+2dx32if d1,2mod42x0x1+2x22+2x2x3+d+12x32if d3mod4f_{d}=\begin{cases}2x_{0}x_{1}+2x_{2}^{2}+2dx_{3}^{2}&\text{if }d\equiv 1,2\mod 4\\ 2x_{0}x_{1}+2x_{2}^{2}+2x_{2}x_{3}+\frac{d+1}{2}x_{3}^{2}&\text{if }d\equiv 3\mod 4\end{cases}

[JM96]. It is easy to see via the explicit embedding of PSL2(𝒪d)\mathrm{PSL}_{2}(\mathcal{O}_{d}) into O+(fd;)\mathrm{O}^{+}(f_{d};\mathbb{Z}) described in [Chu19, §3.1] that the image of PSL2(𝒪d)(2)\mathrm{PSL}_{2}(\mathcal{O}_{d})_{(2)} lands in O+(fd;)(2)\mathrm{O}^{+}(f_{d};\mathbb{Z})_{(2)}. Therefore, we get that the principal congruence subgroup PSL(2,𝒪d)(2)\mathrm{PSL}(2,\mathcal{O}_{d})_{(2)} embeds in a RACG. The index of this subgroup is well known to be

[PSL(2,𝒪d):PSL(2,𝒪d)(2)]={48if d1,2mod(4)60if d3mod(8)36if d7mod(8),[\mathrm{PSL}(2,\mathcal{O}_{d}):\mathrm{PSL}(2,\mathcal{O}_{d})_{(2)}]=\begin{cases}48&\text{if }d\equiv 1,2\mod(4)\\ 60&\text{if }d\equiv 3\mod(8)\\ 36&\text{if }d\equiv 7\mod(8),\end{cases}

(see for example [Dic01]), and thus, Section 1 follows.

5. An example

Consider the lattice LL with basis {e0,e1,e2,e3}\{e_{0},e_{1},e_{2},e_{3}\} and associated inner product matrix AL=diag{7,1,1,1}A_{L}=\mathrm{diag}\{-7,1,1,1\}. The dual LL^{*} has \mathbb{Z}-basis {17e0,e1,e2,e3}\{\frac{1}{7}e_{0},e_{1},e_{2},e_{3}\} and the discriminant group Δ(L)/7\Delta(L)\cong\mathbb{Z}/7\mathbb{Z} is generated by the image of the vector 17e0\frac{1}{7}e_{0}. We can take KK to be the lattice with basis {e4,e5,e6}\{e_{4},e_{5},e_{6}\} and associated inner product matrixAK=diag{7,1,1}A_{K}=\mathrm{diag}\{7,1,1\}. The dual KK^{*} has \mathbb{Z}-basis {17e4,e5,e6}\{\frac{1}{7}e_{4},e_{5},e_{6}\} and the discriminant group Δ(K)/7\Delta(K)\cong\mathbb{Z}/7\mathbb{Z} is generated by the image of the vector 17e4\frac{1}{7}e_{4}. Therefore Δ(LK)=/7×/7\Delta(L\oplus K)=\mathbb{Z}/7\mathbb{Z}\times\mathbb{Z}/7\mathbb{Z} and image of the vector u=47e0+37e4u=\frac{4}{7}e_{0}+\frac{3}{7}e_{4} generates a totally isotropic subgroup GG of Δ(LK)\Delta(L\oplus K) of order /7\mathbb{Z}/7\mathbb{Z}. Indeed, uu has norm 1-1 and pairs integrally with LKL\oplus K. If we let v=37e0+47e4v=\frac{3}{7}e_{0}+\frac{4}{7}e_{4} then {u,v,e1,e2,e3,e5,e6}\{u,v,e_{1},e_{2},e_{3},e_{5},e_{6}\} defines an orthogonal \mathbb{Z}-basis for LGKL\oplus_{G}K with associated inner product matrix diag{1,1,1,1,1,1,1}\mathrm{diag}\{-1,1,1,1,1,1,1\}. Therefore LGK=𝐈6,1L\oplus_{G}K=\mathbf{I}_{6,1}.

Since e0=4u3ve_{0}=4u-3v, a copy of LL is the sublattice with \mathbb{Z}-basis {4u3v,e1,e2,e3}\{4u-3v,e_{1},e_{2},e_{3}\} which is the orthogonal complement of a copy of KK with \mathbb{Z}-basis {3u+4v,e5,e6}\{-3u+4v,e_{5},e_{6}\}. The change of basis matrix is

(5.1) B=(470370I3037047I2)B=\begin{pmatrix}\frac{4}{7}&0&\frac{-3}{7}&\\ 0&I_{3}&0&\\ \frac{-3}{7}&0&\frac{4}{7}&\\ &&&I_{2}\end{pmatrix}

where InI_{n} denotes the n×nn\times n identity matrix. Therefore,

Bdiag{7,1,1,1,7,1,1}Btr=diag{1,1,1,1,1,1,1}.B\cdot\mathrm{diag}\{-7,1,1,1,7,1,1\}\cdot B^{tr}=\mathrm{diag}\{-1,1,1,1,1,1,1\}.

Let γ\gamma be in Aut(L)\mathrm{Aut}(L). As a matrix with entries in \mathbb{Z} preserving ALA_{L}, γ\gamma the first row (abcd)\begin{pmatrix}a&b&c&d\end{pmatrix} satisfies a21mod7a^{2}\equiv 1\mod 7 and b,c,d1mod7b,c,d\equiv 1\mod 7. We extend γ\gamma in AutL\mathrm{Aut}_{L} to γ\gamma in Aut(LK)\mathrm{Aut}(L\oplus K) as follows:

(5.2) γ{(γI3)if a1mod7(γI2I1)if a1mod7\gamma\mapsto\begin{cases}\begin{pmatrix}\gamma&\\ &I_{3}\end{pmatrix}&\text{if }a\equiv 1\mod 7\\ \begin{pmatrix}\gamma&&\\ &-I_{2}&\\ &&I_{1}\end{pmatrix}&\text{if }a\equiv-1\mod 7\end{cases}

and such an integral matrix preserves integrality when conjugated by BB, that is, BγB1B\cdot\gamma\cdot B^{-1} is integral and preserves diag{1,1,1,1,1,1,1}\mathrm{diag}\{-1,1,1,1,1,1,1\}. Indeed, if

γ1=(7a+17b7c7defghijklmnopI3)γ2=(7a17b7c7defghijklmnopI2I1)\gamma_{1}=\begin{pmatrix}7a+1&7b&7c&7d&\\ e&f&g&h&\\ i&j&k&l&\\ m&n&o&p&\\ &&&&I_{3}\end{pmatrix}\text{, }\gamma_{2}=\begin{pmatrix}7a-1&7b&7c&7d&&\\ e&f&g&h&&\\ i&j&k&l&&\\ m&n&o&p&&\\ &&&&-I_{2}&\\ &&&&&I_{1}\end{pmatrix}

then

Bγ1B1=(16a+14b4c4d12a004efgh3e004ijkl3i004mnop3m0012a3b3c3d19a0000000100000001)B\cdot\gamma_{1}\cdot B^{-1}=\begin{pmatrix}16a+1&4b&4c&4d&12a&0&0\\ 4e&f&g&h&3e&0&0\\ 4i&j&k&l&3i&0&0\\ 4m&n&o&p&3m&0&0\\ -12a&-3b&-3c&-3d&1-9a&0&0\\ 0&0&0&0&0&1&0\\ 0&0&0&0&0&0&1\end{pmatrix}

and

Bγ2B1=(16a14b4c4d12a004efgh3e004ijkl3i004mnop3m0012a3b3c3d9a10000000100000001).B\cdot\gamma_{2}\cdot B^{-1}=\begin{pmatrix}16a-1&4b&4c&4d&12a&0&0\\ 4e&f&g&h&3e&0&0\\ 4i&j&k&l&3i&0&0\\ 4m&n&o&p&3m&0&0\\ -12a&-3b&-3c&-3d&-9a-1&0&0\\ 0&0&0&0&0&-1&0\\ 0&0&0&0&0&0&1\end{pmatrix}.

Furthermore, If γI7mod2\gamma\equiv I_{7}\mod 2 entrywise, so is BγB1B\cdot\gamma\cdot B^{-1}, and thus there is an inclusion of principal congruence subgroups of level 22: Aut(L)(2)Aut(𝐈6,1)(2)\mathrm{Aut}(L)_{(2)}\subset\mathrm{Aut}(\mathbf{I}_{6,1})_{(2)}.

References

  • [Ago13] Ian Agol. The virtual Haken conjecture. Doc. Math., 18:1045–1087, 2013. With an appendix by Agol, Daniel Groves, and Jason Manning.
  • [All12] Daniel Allcock. The reflective Lorentzian lattices of rank 3. Mem. Amer. Math. Soc., 220(1033):x+108, 2012.
  • [All18] Daniel Allcock. Prenilpotent pairs in the E10E_{10} root lattice. Math. Proc. Cambridge Philos. Soc., 164(3):473–483, 2018.
  • [BHW11] Nicolas  Bergeron, Frédéric Haglund, and Daniel T. Wise. Hyperplane sections in arithmetic hyperbolic manifolds. J. Lond. Math. Soc. (2), 83(2):431–448, 2011.
  • [Chu18] Michelle Chu. Quantifying virtual properties of Bianchi groups. Phd dissertation, The University of Texas at Austin, 2018.
  • [Chu19] Michelle Chu. Special subgroups of Bianchi groups. To appear in Groups Geom. Dyn., 2019.
  • [CS99] J. H. Conway and N. J. A. Sloane. Sphere packings, lattices and groups, volume 290 of Grundlehren der Mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences]. Springer-Verlag, New York, third edition, 1999. With additional contributions by E. Bannai, R. E. Borcherds, J. Leech, S. P. Norton, A. M. Odlyzko, R. A. Parker, L. Queen and B. B. Venkov.
  • [DMP18] Jason  DeBlois, Nicholas Miller, and Priyam Patel. Effective virtual and residual properties of some arithmetic hyperbolic 3-manifolds. arXiv:1806.02360, 2018. 30 pages.
  • [Dic01] L.E. Dickson. Linear Groups with an Exposition of the Galois Field Theory. B.G Teubner’s Sammlung von Lehrbuchern auf dem Gebiete der mathematischen Wissenschaften mit Einschluss ihrer Anwendungen. B.G. Teubner, 1901.
  • [ERT12] Brent  Everitt, John G. Ratcliffe, and Steven T. Tschantz. Right-angled Coxeter polytopes, hyperbolic six-manifolds, and a problem of Siegel. Math. Ann., 354(3):871–905, 2012.
  • [HW08] Frédéric  Haglund and Daniel T. Wise. Special cube complexes. Geom. Funct. Anal., 17(5):1551–1620, 2008.
  • [JM96] D. G.  James and C. Maclachlan. Fuchsian subgroups of Bianchi groups. Trans. Amer. Math. Soc., 348(5):1989–2002, 1996.
  • [KV78] I. M.  Kaplinskaja and È. B. Vinberg. The groups O18,1(Z)O_{18,1}(Z) and O19,1(Z)O_{19,1}(Z). Dokl. Akad. Nauk SSSR, 238(6):1273–1275, 1978.
  • [Kho86] A. G. Khovanskiĭ. Hyperplane sections of polyhedra, toric varieties and discrete groups in Lobachevskiĭ space. Funktsional. Anal. i Prilozhen., 20(1):50–61, 96, 1986.
  • [PV05] Leonid  Potyagailo and Ernest Vinberg. On right-angled reflection groups in hyperbolic spaces. Comment. Math. Helv., 80(1):63–73, 2005.
  • [Pro86] M. N. Prokhorov. Absence of discrete groups of reflections with a noncompact fundamental polyhedron of finite volume in a Lobachevskiĭ space of high dimension. Izv. Akad. Nauk SSSR Ser. Mat., 50(2):413–424, 1986.
  • [Vin72] È. B. Vinberg. The groups of units of certain quadratic forms. Mat. Sb. (N.S.), 87(129):18–36, 1972.
  • [Vin84] È. B. Vinberg. Absence of crystallographic groups of reflections in Lobachevskiĭ spaces of large dimension. Trudy Moskov. Mat. Obshch., 47:68–102, 246, 1984.
  • [Wat62] G. L. Watson. Transformations of a quadratic form which do not increase the class-number. Proc. London Math. Soc. (3), 12:577–587, 1962.
  • [Wat75] G. L. Watson. Transformations of a quadratic form which do not increase the class-number. II. Acta Arith., 27:171–189, 1975.
  • [Wis11] Daniel T. Wise. The structure of groups with quasiconvex hierarchy. 2011. 187 pages.