Professional Documents
Culture Documents
AND ARITHMETICITY
A BSTRACT. In this paper we study crystallographic sphere packings and Kleinian sphere
arXiv:2203.01973v1 [math.GT] 3 Mar 2022
packings, introduced first by Kontorovich and Nakamura in 2017 and then studied further by
Kapovich and Kontorovich in 2021. In particular, we solve the problem of existence of crystal-
lographic sphere packings in certain higher dimensions posed by Kontorovich and Nakamura.
In addition, we present a geometric doubling procedure allowing to obtain sphere packings
from some Coxeter polyhedra without isolated roots, and study “properly integral” packings
(that is, ones which are integral but not superintegral).
§ 1. Introduction
1.1. Sphere packings. For n ≥ 2, let Hn+1 be the ( n + 1)–dimensional hyper-
bolic space. Then the set of all ideal points of hyperbolic space, or its ideal
boundary, is ∂Hn+1 ≈ Sn and can be identified topologically with Rn := Rn ∪{∞}.
One can consider different coordinate systems in Sn , and thus some of the
definitions below depend on this choice of coordinates.
Definition 1.1. By a sphere packing (or just packing) P of Sn we mean an
infinite collection {Bα } of round balls in Sn so that:
(1) The interiors of the balls are disjoint, and
(2) The union of the balls is dense in Sn .
Recall that, given a packing P = {Bα } of Sn ≈ ∂Hn+1 , every sphere S α = ∂Bα
also is the boundary of a hyperplane Hα ∼ = Hn of Hn+1 , i.e., S α = ∂∞ Hα . Any
such hyperplane Hα is associated with a reflection R α ∈ Isom(Hn+1 ), i.e., Hα
is the mirror of the reflection R α . Let ΓR be the group generated by the
reflections in these hyperplanes, ΓR := 〈R α | Bα ∈ P 〉.
Definition 1.2. The superpacking of P is its orbit under the ΓR –action:
f = ΓR · P .
P
Bogachev was partially supported by the Russian Science Foundation, grant no. 21-41-09011.
Kolpakov was partially supported by the Swiss National Science Foundation, project no. PP00P2-202667 .
Kontorovich was partially supported by NSF grant DMS-1802119, BSF grant 2020119, and the 2020-2021
Distinguished Visiting Professorship at the National Museum of Mathematics.
1
2 NIKOLAY BOGACHEV, ALEXANDER KOLPAKOV, AND ALEX KONTOROVICH
We should stress the fact that answering the above question in the nega-
tive even in the more narrow superintegral setting will require a complete
classification of reflective quadratic Lorentzian forms over Q, which seems
rather difficult at the moment.
Another interesting direction to pursue is the possible connection between
arithmetic reflection groups and sphere packings. Namely, Kontorovich and
Nakamura [KN] asked the following question:
For n = 2, Kontrovich and Nakamura [KN] confirmed the same for the fi-
nite list of reflective extended Bianchi groups. However, the Borcherds poly-
hedron in dimension 21 provides a clear counterexample.
§ 2. Preliminaries
2.1. Hyperbolic Lobachevsky space and convex polyhedra. Let Rn,1 be
the ( n+1)-dimensional real Minkowski space equipped with the inner product
( x, y) = − x0 y0 + x1 y1 + . . . + xn yn
cosh ρ ( x, y) = −( x, y).
Let On,1 (R) be the group of orthogonal transformations of the space Rn,1 ,
and let POn,1 (R) be its subgroup of index 2 preserving Hn . The isometry group
of the hyperbolic n-space Hn is Isom(Hn ) ∼= POn,1 (R).
Suppose that e ∈ R is a non–zero vector. Then the set
n,1
H e = { x ∈ Hn | ( x, e) = 0}
H −e = { x ∈ Hn | ( x, e) ≤ 0}, H +e = { x ∈ Hn | ( x, e) ≥ 0}.
with Hn+1 , and use the inner product ( x, y) = f ( x, y) to induce the hyperbolic
metric on it. Here we use the same notation for the quadratic form and the
induced bilinear form as it does not create much ambiguity.
Given a vector e α ∈ {v ∈ V : (v, v) > 0}, we have the hyperplane Hα = e⊥ α ∩P C + ,
( x,e α )
and a reflection R α : x 7→ x − 2 ( e α ,e α ) e α with mirror Hα . This reflection induces
an inversion in the sphere S α = ∂ Hα .
Dual to V is the vector space V ∗ which is a quadratic space under f ∗ . Then
to construct an inversive coordinate system, we simply proceed as follows.
To choose coordinates, we first choose any two f ∗ –isotropic (over R) vectors
b, b̂ ∈ V ∗ , such that f ∗ ( b, b̂) = −2.
Then take any system of orthonormal vectors b j , j ∈ {1, . . . , n} in the space
orthogonal to that generated by b and b̂. The inversive coordinate system of
a sphere S α corresponding to the vector e α ∈ {v ∈ V : (v, v) > 0} is then given
by
vα = ( b( e α ), b̂( e α ), b 1 ( e α ), . . . , b n ( e α )).
This vector vα contains all of the data of the root e α . Indeed, consider the
sphere S α in Sn corresponding to the ideal boundary of the hyperplane Hα
which is f −orthogonal to e α ; the “bend” (inverse radius) of the sphere is b( e α )
and its co-bend is b̂( e α ). Moreover, its center is ( b 1 ( e α ), . . . , b n ( e α ))/ b( e α ). If the
bend b( e α ) = 0, that is, S α is a hyperplane, then this “center” is re-interpreted
to mean the limit as spheres converge to this hyperplane; in that case, this
ratio becomes the (Euclidean) unit normal to this hyperplane in ∂Hn+1 .
Theorem 2.4 (Structure Theorem [KN, K2]).
(1) Let P be a Kleinian packing of Sn ' ∂Hn+1 . Then its supergroup Γ e is a
lattice in Isom(Hn+1 ).
(2) Suppose that Γ e < Isom(Hn+1 ) is a lattice (containing at least one reflec-
tion) with a convex fundamental polyhedron D and a minimal generat-
ing set Se for Γ
e , where elements of Se correspond to face-pairing transfor-
mations of D . Let R ⊂ Se be a non-empty set of generators corresponding
to reflections γα in certain facets Fα , α ∈ A , of D . For each α ∈ A let
S α = ∂∞ Hα ⊂ ∂Hn+1 = Sn be the invariant sphere for the reflection γα
with respect to the supporting hyperplane Hα of the facet Fα (see Fig. 1).
Assume that all the hyperplanes Hα are pairwise disjoint or parallel to
each other, and that, if they meet other supporting hyperplanes of facets
of D , then they do so orthogonally. Let S = Se \ R and let Γ = 〈S 〉.
8 NIKOLAY BOGACHEV, ALEXANDER KOLPAKOV, AND ALEX KONTOROVICH
then let G (P ) = { g i j }N N
i, j =1 = {( e i , e j )} i, j =1 denote its Gram matrix, and let the
field K = Q({ g i j }N
i, j =0 ) be generated by its entries.
The set of all cyclic products of entries of a matrix A (i.e. the set consist-
ing of all possible products of the form a i 1 i 2 a i 2 i 3 . . . a i k i 1 ) will be denoted by
Cyc( A ). Let k = Q(Cyc(G (P ))) ⊂ K , and let O k be the ring of integers of k.
Given the lattice Γ generated by reflections in the facets of a Coxeter poly-
hedron P , we can determine if Γ is arithmetic, quasi–arithmetic, or neither.
Theorem 2.5 (Vinberg’s arithmeticity criterion [Vin67]). Let Γ be a reflec-
tion group acting on Hn with finite volume fundamental Coxeter polyhedron
P . Then Γ is arithmetic if and only if the following conditions hold:
(1) K is a totally real algebraic number field;
(2) for any embedding σ : K → R, such that σ |k 6= id, G σ (P ) is positive semi-
definite;
(3) Cyc(2 · G (P )) ⊂ O k .
The group Γ is quasi–arithmetic if and only if it satisfies conditions (1)–(2),
but not necessarily (3).
Whenever Γ is quasi–arithmetic or arithmetic, k is called its field of defini-
tion (or ground field). The above criterion simplifies greatly if k = Q. Then
the lattice Γ will always be quasi–arithmetic, and it will be arithmetic if and
only if the cyclic products in Theorem 2.5 (3) are integers.
2.4. Vinberg’s algorithm. In addition to his study of arithmeticity of re-
flective lattices, Vinberg also produced an algorithm that produces the set of
generating reflections for a reflective quadratic from. Namely, if f is a qua-
dratic form of signature ( n, 1) over a totally real number field k with the ring
of integers O k .
Moreover, assume that f is admissible, that is for each non–trivial Galois
embedding σ : k → R the form f σ (obtained by applying σ to the coefficients of
f ) is positive definite. Then the group O f (O k ) of integer points of the orthog-
onal group O f (R) is a lattice [AVS, II, Chapter 6].
The form f as above is called reflective if O f (O k ) is generated by a finite
number of reflections, up to a finite index. The algorithm devised by Vinberg
in [Vin72] produces a set of generating reflections for any reflective quadratic
form. We shall refer to it as the Vinberg algorithm. Several computer real-
izations of the algorithm are available [Gug, BP, Bot].
10 NIKOLAY BOGACHEV, ALEXANDER KOLPAKOV, AND ALEX KONTOROVICH
In contrast, if the form f is not reflective, the algorithm never halts. The
later work by Bugaenko [Bug] also provided a way to certify that a given
Lorentzian form f is non–reflective. This “method of infinite symmetry” has
been also implemented in [Gug, BP, Bot] to various extent.
§ 3. Proofs
3.1. Proof of Lemma 1.8. Let P be a finite volume Coxeter polyhedron in
Hn and P0 be a facet that meets other facets of P at angles either π/4 or π/2.
Let P1 , . . . P k be all the facets of P that meet P0 at π/4. Assume that each P j
meets its neighbors at “even” angles of the form π/2m , with m ≥ 1, an integer.
If P is a polyhedron as above, and F is a facet of P , let R F denote the
reflection in the hyperplane of F .
Then the following inductive steps will produce a polyhedron with an or-
thogonal facet.
1. The double P (1) of P along the facet F1 = P1 has the facet R F1 (P0 ) that
forms ( k − 1) angles π/4 with its neighbours. Observe that P (1) is again a Cox-
eter polyhedron, since doubling of even Coxeter angles produces even Coxeter
angles again or, if we double a right angle, it produces a new facet.
2. The facet F2 = R F1 (P2 ) of P (1) forms even Coxeter angles with its neigh-
bours. Thus the double P (2) of P (1) along F2 has the facet R F2 R F1 (P0 ) that
forms ( k − 2) angles π/4 with its neighbours. All the conclusions of the previ-
ous step apply respectively to P (2) .
3. Continuing in this way, we obtain a polyhedron P (k) from P (k−1) by dou-
bling along F k = R Fk−1 (P k ), and the facet R Fk · · · R F1 (P0 ) of P (k) will be orthog-
onal to all of its neighbours. Such a polyhedron produces a crystallographic
sphere packing by the Structure Theorem (Theorem 2.4).
3.2. Proof of Theorem A. The proof of Theorem A follows from the results
of Esselmann [E] and the following simple lemma which is well known to
experts in hyperbolic reflection groups (and also follows from more general
results of [BK21]).
Lemma 3.1. Let Γ < Isom(Hn+1 ) be the symmetry group of a crystallographic
packing P of Sn , and let Γ0 be the stabilizer of a sphere in P . Then Γ0 is a
reflective lattice acting on Hn . If Γ is arithmetic, then Γ0 is also arithmetic
Proof. Indeed, the sphere corresponds to a facet P0 of the Coxeter polyhe-
dron P associated to the packing P , and P0 is orthogonal to all the adjacent
KLEINIAN SPHERE PACKINGS, REFLECTION GROUPS, AND ARITHMETICITY 11
facets. Then Γ0 coincides with the finite covolume hyperbolic group generated
by reflections in the facets of P0 . The orthogonality of P0 to its neighbours to-
gether with Vinberg’s criterion (Theorem 2.5) implies that Γ being arithmetic
makes Γ0 arithmetic too. ■
3.4. Proof of Theorem C. In [SW] Sharlau and Walhorn provide the list of
maximal arithmetic non–uniform reflective lattices (up to conjugation) acting
on Hn+1 for n = 2, 3. In [Tur] Turkalj performs an analogous classification in
Hn+1 for n = 4.
It is worth mentioning that in the list of lattices acting on H4 has a few
entries where the H direct summand apparently has to be replaced by 2 H.
Such entries are easily identifiable by running the Vinberg algorithm (the
number of roots indicated in the respective tables seems to be correct). The
lattices described by Turkalj require more work for the information to be
extracted from the list provided in [Tur]. We used instead an equivalent list
provided by Scharlau and Kirschmer [SK].
In each case we reduce each list up to rational isometry between the re-
spective lattices. Then the proof amounts to a lengthy computation carried
our by using the Julia [BEKS] implementation of the Vinberg algorithm due
to Bottinelli [Bot].
For each lattice, we consider its reflection subgroup: if the corresponding
polyhedron has an orthogonal facet (or, in term of Vinberg’s work [Vin15],
the lattice has an isolated root), then we are done. Otherwise, we choose
another lattice, up to an appropriate rational isometry, and use it instead: we
get a commensurable lattice having an isolated root. The necessary rational
isometry is usually guessed by trial and error: reflectivity is checked by using
the Vinberg algorithm.
All the data necessary to verify our computation [Git1] and Bottinelli’s im-
plementation of the Vinberg algorithm [Git2] are available on Github.
p
4 5 8/7 6
1
F IGURE 3. Some of the spheres of the packing P shown in blue, and the walls
of the generators e 1 , . . . , e 5 of Γ shown in red.
1 0 0 0 0 1 0 0 0 0
0 1 0 0 0
0 0 1 0 0
−1 −1
V R3V = 0 0 0 0 1 , V R4V = 0 1 0 0 0 ,
0 0 −1 1 1 0 0 0 1 0
0 0 1 0 0 0 −1 1 0 1
KLEINIAN SPHERE PACKINGS, REFLECTION GROUPS, AND ARITHMETICITY 15
and
0 1 0 0 0
1 0 0 0 0
V R 5 V −1 = 0 0 1 0 0 .
0 0 0 1 0
0 0 0 0 1
These act on the left on the set of bends of the packing; the bends which
arise in the packing are those in the action of the above matrices on the
initial bends vector (1, 1, 1, 1, 1) of all bends equal to 1. All but the first matrix
are themselves integral, and hence evidently preserve integrality. Moreover,
inspection shows that they preserve the congruence b 1 + b 2 + 2 b 3 + 3 b 4 + b 5 ≡
0 (mod 4), which is satisfied by the initial bends vector (1, 1, 1, 1, 1). This
congruence ensures that the action of the first matrix V R 1 V −1 always gives
integer bends in the orbit, thus proving the proper integrality of the packing.
Finally, following the proof of the Arithmeticity Theorem, we exhibit an
arithmetic lattice Ω containing a conjugate of Γ. We find in this case that Ω
is not reflective. To begin, we record that the Gram matrix of V :
−7 9 9 9 9
9 −7 9 9 9
T 1
G = G (V ) = V QV = − 9 9 −7 9 25 ,
7
9 9 9 −7 41
9 9 25 41 −7
References
[A] D. Allcock. Infinitely many hyperbolic Coxeter groups through dimension 19. Geom. Topol., Vol. 10,
pp. 737–758, 2006. 11
[AVS] D. Alekseevskij, È. Vinberg, A. Solodovnikov. Geometry of spaces of constant curvature. Encycl.
Math. Sci., Vol. 29, pp. 1–138, 1993. 9
[BEKS] J. Bezanson, A. Edelman, S. Karpinski, V. B. Shah. Julia: A fresh approach to numerical computing.
SIAM Review, Vol. 59, No. 1, pp. 65–98, 2017. 12
[BK21] N. Bogachev, A. Kolpakov. On faces of quasi–arithmetic Coxeter polytopes. IMRN, Vol. 2021, No. 4,
pp. 3078–3096, 2021. 10
[BK22] N. Bogachev, A. Kolpakov. Thin hyperbolic reflection groups. Preprint, arXiv:2112.14642 17
[Git1] N. Bogachev, A. Kolpakov, A. Kontorovich. Crystallographic packings.
https://github.com/sashakolpakov/crystallographic-packings 11, 12
[BP] N. Bogachev, A. Perepechko. Vinberg’s algorithm for hyperbolic lattices. Math. Notes, Vol. 103, No. 5,
pp. 836–840, 2018. 9, 10
[Bor] R. Borcherds. Automorphism groups of Lorentzian lattices. J. Algebra, Vol. 111, pp. 133–153, 1987.
3, 11
[Bor2] R. Borcherds. Reflection groups of Lorentzian lattices. Duke Math J., Vol. 104 no. 2, pp. 319–366,
2000. 4
[Bot] R. Bottinelli. On computational aspects of hyperbolic reflection groups & two homology theories for
graphs. PhD thesis, 2021. 9, 10, 12
[Git2] R. Bottinelli. Vinny the Coxboy: Vinberg’s algorithm over number fields.
https://github.com/bottine/VinbergsAlgorithmNF 12
[Bug] V. O. Bugaenko. Arithmetic crystallographic groups generated by reflections, and reflective hyper-
bolic lattices. In: Lie groups, their discrete subgroups, and invariant theory. Adv. Soviet Math., Vol.
8, Amer. Math. Soc., Providence, RI, 1992, pp. 33–55. 10
[Col] G. Collinet. Personal communication to A. Kolpakov. 10 November 2014 17
[E] F. Esselmann. Über die maximale Dimension von Lorentz–Gittern mit coendlicher Spiegelungs-
gruppe. J. Num. Theory, Vol. 61, pp. 103–144, 1996. 2, 3, 10, 11
[Gug] R. Guglielmetti. Hyperbolic isometries in (in–)finite dimensions and discrete reflection groups: the-
ory and computations. PhD Thesis, Université de Fribourg, 2017. 9, 10
[KN] A. Kontorovich, K. Nakamura. Geometry and arithmetic of crystallographic sphere packings. PNAS,
Vol. 116, No. 2, pp. 436–441, 2019. 2, 3, 4, 7, 12, 14, 15
[KV] I. Kaplinskaya, È. Vinberg. On the groups O18,1 (Z) and O19,1 (Z). Dokl. Akad. Nauk SSSR, Vol. 238,
No. 6, pp. 1273–1275, 1978. 11
[K2] M. Kapovich, A. Kontorovich. On superintegral Kleinian sphere packings, bugs, and arithmetic
groups. Preprint, arXiv:2104.13838 2, 7, 12
[PV] L. Potyagailo, È. Vinberg. On right-angled reflection groups in hyperbolic spaces. Comment. Math.
Helv., Vol. 80, No. 1, pp. 63–73, 2005. 11
18 NIKOLAY BOGACHEV, ALEXANDER KOLPAKOV, AND ALEX KONTOROVICH
[Sage] SageMath, the Sage Mathematics Software System. Version 9.4, The Sage Developers, 2021,
https://www.sagemath.org 11
[SK] R. Scharlau, M. Kirschmer. Personal communication to N. Bogachev. 17 February 2022 12
[SW] R. Scharlau, C. Walhorn. Integral lattices and hyperbolic reflection groups. Astérisque, Tome 209, pp.
279–291, 1992. 12
[Tur] I. Turkalj. Reflective Lorentzian lattices of signature (5,1). PhD Thesis, Universität Dortmund, 2017.
12
[Vin67] È. B. Vinberg. Discrete groups generated by reflections in Lobachevski spaces. Mathematics of the
USSR – Sbornik, Vol. 1, No. 3, pp. 429–444, 1967. 6, 8, 9, 11, 16
[Vin72] È. B. Vinberg. On groups of unit elements of certain quadratic forms. Mathematics of the USSR –
Sbornik, Vol. 16. No. 1, pp. 17–35, 1972. 9
[Vin85] È. B. Vinberg. The absence of crystallographic groups of reflections in Lobachevsky spaces of large
dimension. In: Trans. Moscow Math. Soc., 1985, pp. 75–112. 6
[Vin15] È. B. Vinberg. Non–arithmetic hyperbolic reflection groups in higher dimensions. Moscow Math. J.,
Vol. 15, No. 3, pp. 593–602, 2015. 12