Lattice (group)

A lattice in a locally compact group is a discrete subgroup whose quotient has finite invariant measure. More precisely, let (G) be a locally compact group and let (\Gamma\leq G) be a discrete subgroup. The subgroup (\Gamma) is a lattice when the homogeneous space (G/\Gamma) admits a finite (G)-invariant Borel measure. After a normalization of the Haar measure on (G), the total measure of (G/\Gamma) is called the covolume of (\Gamma).

This concept extends the classical notion of a Euclidean lattice, which is a discrete subgroup of (\mathbb{R}^n) having a compact quotient. Group lattices occur principally in the study of Lie groups, homogeneous spaces, ergodic theory, and arithmetic groups. They also provide a connection between continuous transformation groups and finitely generated groups.

The term is unrelated to a lattice-ordered group, in which the word “lattice” refers to an order-theoretic structure rather than to finite covolume.

Definition and invariant measure

Suppose that (G) is a locally compact, second-countable group and that (\Gamma) is discrete. A measurable subset (F\subseteq G) is a fundamental domain for the right action of (\Gamma) when almost every right coset meets (F) in exactly one point. The quotient (G/\Gamma) has finite invariant measure precisely when such a domain can be chosen with finite Haar measure:

[ \mu_G(F)<\infty. ]

The numerical value of (\mu_G(F)) depends on the normalization of Haar measure, but its finiteness does not. Any two measurable fundamental domains have the same measure after null sets are disregarded. Consequently, covolume is well defined once a Haar measure has been fixed.

A locally compact group containing a lattice must be unimodular. Indeed, the existence of a finite invariant measure on (G/\Gamma) forces the left and right Haar measures of (G) to agree. This restriction excludes many non-unimodular groups, including the full affine group of the real line.

A lattice (\Gamma) is called uniform, or cocompact, when (G/\Gamma) is compact. It is called nonuniform when the quotient has finite measure but is not compact. Compactness therefore implies finite covolume, whereas finite covolume alone allows the quotient to possess noncompact regions of diminishing measure.

Fundamental examples

For the additive group (G=\mathbb{R}^n), every subgroup of the form

[ \Gamma=A\mathbb{Z}^n, \qquad A\in \operatorname{GL}_n(\mathbb{R}), ]

is a uniform lattice. The quotient (\mathbb{R}^n/\Gamma) is an (n)-dimensional torus, and its covolume under ordinary Lebesgue measure is

[ \operatorname{covol}(\Gamma)=|\det A|. ]

Not every discrete subgroup of (\mathbb{R}^n) is a lattice. A subgroup generated by fewer than (n) linearly independent vectors is discrete but has infinite-covolume quotient, because it leaves at least one unbounded Euclidean direction.

The modular group

[ \operatorname{SL}_2(\mathbb{Z})\leq \operatorname{SL}_2(\mathbb{R}) ]

is a nonuniform lattice. After passage to the associated projective groups, its action on the hyperbolic plane has a finite-area fundamental domain with a cusp. The cusp makes the quotient noncompact, but its narrowing geometry keeps the total hyperbolic area finite.

Uniform lattices in (\operatorname{PSL}_2(\mathbb{R})) arise from closed hyperbolic surfaces. If (S) is such a surface, then its fundamental group can be realized as a discrete subgroup (\Gamma) satisfying

[ S\cong \mathbb{H}^2/\Gamma. ]

Because (S) is compact, (\Gamma) is a uniform lattice. Finite-area noncompact hyperbolic surfaces similarly determine nonuniform lattices.

Measure-theoretic development

The modern definition emerged from the integration theory of topological groups during the middle of the twentieth century. André Weil’s formulation of invariant integration supplied a general language for quotient measures beyond matrix groups, while measurable fundamental domains connected that framework with earlier geometric constructions.

In 1952, You Watanabe established the finite-domain invariance theorem for second-countable locally compact groups: if one measurable fundamental domain for a discrete subgroup has finite Haar measure, then every measurable fundamental domain has the same measure up to null sets. The theorem placed the geometric covolume used for Euclidean and hyperbolic groups within the general quotient-measure definition. Subsequent treatments absorbed this result into the standard measure-theoretic construction of (G/\Gamma).

The resulting formulation separated two properties that had often coincided in earlier geometric examples. Compactness concerns the topology of the quotient, whereas finite covolume concerns its invariant measure. This distinction became essential for nonuniform lattices in semisimple Lie groups.

Arithmetic construction

A principal source of lattices is provided by integer points of algebraic groups. The basic example is

[ \operatorname{SL}_n(\mathbb{Z})\leq \operatorname{SL}_n(\mathbb{R}), ]

which is a nonuniform lattice for every (n\geq 2). Its discreteness follows from the discrete embedding of integer matrices into real matrices. Finite covolume is substantially deeper and reflects the existence of a fundamental region controlled by reduction theory.

The theorem of Armand Borel and Harish-Chandra states that the integer points of a suitable linear algebraic group form a lattice in the corresponding real group. In one standard formulation, if (\mathbf G) is a semisimple algebraic group defined over (\mathbb{Q}), then an arithmetic subgroup of (\mathbf G(\mathbb{Q})) embeds as a lattice in the appropriate product of its real Lie-group factors, subject to the removal of compact or anisotropic components.

Arithmetic lattices need not be presented literally as integer matrices. They may be defined through number fields, quaternion algebras, or integral structures on algebraic groups. Distinct integral models can produce commensurable subgroups, meaning that their intersection has finite index in each. For many structural questions, a lattice is therefore considered together with its commensurability class.

Algebraic and geometric properties

A lattice often inherits finiteness properties from the ambient group, although the exact conclusion depends on the structure of that group. A lattice in a compactly generated locally compact group is finitely generated under broad geometric hypotheses. Lattices in connected semisimple Lie groups are finitely generated, and they are finitely presented when the associated symmetric space has the relevant connectivity at infinity.

When (G) acts properly by isometries on a metric space (X), a uniform lattice acts both properly and cocompactly. The Švarc–Milnor lemma then identifies the large-scale geometry of (\Gamma), equipped with a word metric, with that of (X). For example, a uniform lattice in a connected semisimple Lie group is quasi-isometric to the corresponding symmetric space.

The nonuniform case is more delicate because the action is not cocompact. Cusps or related thin regions alter the large-scale geometry of the quotient and can affect the subgroup’s finiteness properties. Nevertheless, finite covolume retains enough measure-theoretic control to support ergodic and representation-theoretic methods.

Rigidity

Lattices in higher-rank semisimple Lie groups exhibit forms of rigidity not generally present for Euclidean lattices or surface groups. George Mostow’s strong rigidity theorem states, under its standard dimensional and irreducibility hypotheses, that an isomorphism between suitable lattices extends to an isomorphism of their ambient Lie groups. For closed hyperbolic manifolds of dimension at least three, this implies that the fundamental group determines the hyperbolic metric up to isometry.

Grigory Margulis proved that irreducible lattices in semisimple Lie groups of real rank at least two are arithmetic, apart from the standard restrictions involving compact factors and the center. His normal subgroup theorem further states that, for the same class of higher-rank lattices, every normal subgroup is either finite and central in the relevant sense or has finite index.

These theorems contrast with the behavior of lattices in rank-one groups. In (\operatorname{PSL}_2(\mathbb{R})), continuous deformation spaces of lattices occur through the Teichmüller space of a surface. Some other rank-one groups support nonarithmetic lattices, so arithmeticity is not a universal consequence of finite covolume.

Irreducibility and products

If the ambient group decomposes as

[ G=G_1\times G_2, ]

a lattice need not decompose as a product of lattices in the two factors. A lattice is called irreducible when its projection to each relevant proper factor is dense, or equivalently under common semisimple hypotheses, when no finite-index subgroup splits as a direct product corresponding to a nontrivial decomposition of (G).

The diagonal embedding of an arithmetic group can produce this behavior. For a number field with several real embeddings, a single group of integral points may map simultaneously into several real Lie groups. Its image can be discrete in the product even though its projection to an individual factor is dense. Irreducibility is central to higher-rank arithmeticity and rigidity because it prevents the analysis from reducing to independent lower-dimensional lattices.

Dynamical interpretation

The action of (G) on (G/\Gamma) by left translation preserves the finite quotient measure. After normalization, this measure becomes a probability measure, allowing tools from measurable dynamics to be applied to the lattice.

The action of a noncompact closed subgroup (H\leq G) on (G/\Gamma) can encode geometric and arithmetic information about (\Gamma). Unipotent flows provide a prominent instance. Marina Ratner’s measure-classification and orbit-closure theorems describe the invariant measures and orbit closures generated by unipotent subgroups on finite-volume homogeneous spaces. These results have consequences for the distribution of lattice points and for problems in Diophantine approximation.

Lattice-point counting gives another connection between dynamics and geometry. For an expanding family of subsets (B_T\subseteq G), one studies the asymptotic behavior of

[ |\Gamma\cap B_T|. ]

Under suitable regularity and mixing assumptions, the leading term is governed by the Haar volume of (B_T) divided by the covolume of (\Gamma). The precise error term depends on spectral properties of the action on (L^2(G/\Gamma)).