Arithmetic Margulis Lemma Overview

• Arithmetic Margulis Lemma is an arithmetic refinement of the classical Margulis lemma, establishing a uniform identity neighborhood in which arithmetic lattices contain only finite-order elements.
• It converts the problem of controlling small infinite-order elements in Lie groups into a Diophantine lower-bound problem on the Mahler measure, connecting to weak Lehmer’s conjecture.
• The lemma underpins quantitative results in higher-rank arithmetic groups and hyperbolic settings, affecting estimates on injectivity radius and trace-field degrees.

The Arithmetic Margulis Lemma is a modern label for an arithmetic enhancement of the Margulis-lemma paradigm in which small elements in lattices are controlled not merely by local Lie or Riemannian geometry, but by arithmetic data. In the most precise sense supported by current usage, it refers to a uniform identity-neighborhood statement for arithmetic lattices: for a fixed semisimple real group, there should exist a neighborhood of the identity containing no infinite-order elements of irreducible cocompact arithmetic lattices. In higher rank, the central result of "Arithmetic Groups and the Lehmer Conjecture" identifies this phenomenon with a weak form of Lehmer’s conjecture on Mahler measure, thereby converting a small-element problem in Lie groups into a Diophantine lower-bound problem (Pham et al., 2020). Later papers also use the same label for arithmetic refinements of the classical Margulis lemma in hyperbolic settings, where the small-displacement threshold scales linearly with the trace-field degree (Belolipetsky et al., 2023, Fisher et al., 2022).

1. Terminology and basic formulation

The term does not originate as a formal theorem name in (Pham et al., 2020). That paper instead isolates a statement which functions as an arithmetic analogue of Margulis’s conjectural cocompact higher-rank neighborhood theorem. Let $G$ be a connected semisimple $\mathbb R$-group with $\operatorname{rank}_{\mathbb R}(G)\ge 2$. Margulis had proved for irreducible non-cocompact lattices that, if $G$ has no $\mathbb R$-anisotropic factor, there exists a neighborhood $U\subset G$ of $1$ such that for every irreducible non-cocompact lattice $\Gamma\subset G$, the intersection $U\cap \Gamma$ consists of unipotent elements. He conjectured the cocompact counterpart: $\text{Let %%%%9%%%% be a connected semisimple %%%%10%%%%-group with %%%%11%%%%. Then there exists a neighborhood %%%%12%%%% of the identity such that for any irreducible cocompact lattice %%%%13%%%%, the intersection %%%%14%%%% consists of elements of finite order.}$

The arithmetic version isolated in (Pham et al., 2020) is formulated for a family $\mathbb R$0 of semisimple $\mathbb R$1-groups: $\mathbb R$2 Because irreducible higher-rank lattices are arithmetic, this arithmetic version is equivalent to Margulis’s original conjecture in higher rank.

This is the exact sense in which one may speak of an arithmetic Margulis lemma: a uniform identity neighborhood, depending only on the ambient group, in which arithmetic cocompact lattices contain no infinite-order elements. In torsion-free situations this becomes

$\mathbb R$3

equivalently a uniform positive lower bound on injectivity radius or displacement near the identity. The paper simultaneously stresses a limitation specific to the real case: torsion cannot be removed from the general statement, because cocompact arithmetic lattices may contain torsion elements tending to $\mathbb R$4. Thus the precise conclusion is “finite order,” not “trivial intersection” (Pham et al., 2020).

2. Equivalence with weak Lehmer

The decisive arithmetic input in (Pham et al., 2020) is a weak Lehmer conjecture stated level by level. For $\mathbb R$5, the conjecture asserts the existence of $\mathbb R$6 such that for any irreducible monic $\mathbb R$7 with $\mathbb R$8, either

$\mathbb R$9

where

$\operatorname{rank}_{\mathbb R}(G)\ge 2$0

with $\operatorname{rank}_{\mathbb R}(G)\ge 2$1 the roots of $\operatorname{rank}_{\mathbb R}(G)\ge 2$2, and $\operatorname{rank}_{\mathbb R}(G)\ge 2$3 the number of roots with $\operatorname{rank}_{\mathbb R}(G)\ge 2$4, counted with multiplicity. For an algebraic integer $\operatorname{rank}_{\mathbb R}(G)\ge 2$5, $\operatorname{rank}_{\mathbb R}(G)\ge 2$6 denotes the Mahler measure of its minimal polynomial.

The principal theorem fixes an absolutely (almost) simple isotropic $\operatorname{rank}_{\mathbb R}(G)\ge 2$7-group $\operatorname{rank}_{\mathbb R}(G)\ge 2$8, and for $\operatorname{rank}_{\mathbb R}(G)\ge 2$9 considers

$G$0

It then proves two implications. First, Margulis’ conjecture for any such family $G$1 implies Lehmer’s conjecture at level $G$2. Second, full Margulis conjecture in higher rank is equivalent to the weak Lehmer conjecture at all levels. In this form, uniform discreteness of cocompact arithmetic lattices in higher-rank semisimple groups is not merely a consequence of weak Lehmer; it is equivalent to it (Pham et al., 2020).

This equivalence clarifies the arithmetic content of the neighborhood statement. The obstruction to a cocompact higher-rank Margulis neighborhood is exactly the possible existence of algebraic integers with bounded $G$3 and Mahler measure arbitrarily close to $G$4. Conversely, if weak Lehmer holds, then adjoint characteristic polynomials of infinite-order arithmetic lattice elements must stay uniformly away from Mahler measure $G$5.

The rank-one prototype is Sury’s result for $G$6, where the relevant number-theoretic statement is the Salem conjecture. Since Salem numbers are exactly the $G$7 instance of weak Lehmer, the case $G$8 recovers that correspondence. The paper also records that $G$9, in the purely complex case, links analogous uniform-discreteness statements for $\mathbb R$0 to the complex Salem conjecture (Pham et al., 2020).

3. Arithmetic mechanism from weak Lehmer to a Margulis neighborhood

The direction “weak Lehmer $\mathbb R$1 arithmetic Margulis lemma” is the clearest arithmetic formulation in (Pham et al., 2020). Let $\mathbb R$2 be a semisimple $\mathbb R$3-group and $\mathbb R$4 an irreducible arithmetic lattice. After reduction away from the center and anisotropic factors, the arithmeticity package provides a semisimple $\mathbb R$5-group $\mathbb R$6, an $\mathbb R$7-epimorphism $\mathbb R$8 with compact kernel, and an embedding $\mathbb R$9 such that $U\subset G$0 is commensurable with $U\subset G$1. Consequently $U\subset G$2 preserves a lattice in $U\subset G$3, so the characteristic polynomials of $U\subset G$4 have integer coefficients.

For $U\subset G$5, write $U\subset G$6 for the characteristic polynomial of $U\subset G$7, and define

$U\subset G$8

This function is continuous. Compact factors contribute only eigenvalues of modulus $U\subset G$9, so for $1$0,

$1$1

If weak Lehmer holds at level $1$2, there exists $1$3 such that any integral polynomial with at most $1$4 roots outside the unit circle satisfies either $1$5 or $1$6. Hence

$1$7

is an open neighborhood of $1$8 in $1$9 such that every $\Gamma\subset G$0 has $\Gamma\subset G$1.

Kronecker’s theorem then forces $\Gamma\subset G$2 to be a product of cyclotomic polynomials. Since cocompact lattices contain no nontrivial unipotents, $\Gamma\subset G$3 is semisimple; all its eigenvalues are roots of unity, so $\Gamma\subset G$4, hence $\Gamma\subset G$5, has finite order. This yields exactly the desired neighborhood statement (Pham et al., 2020).

The resulting lower bound is naturally described as a lower bound on adjoint spectral size. What is ruled out uniformly is not merely short translation length in a chosen metric, but the possibility that an infinite-order arithmetic lattice element has adjoint characteristic polynomial with Mahler measure arbitrarily close to $\Gamma\subset G$6. This suggests an arithmetic version of the usual Margulis-lemma intuition: small semisimple behavior is forbidden by integrality plus a Diophantine gap.

4. Converse mechanism from small Mahler measure to small lattice elements

The converse implication, “uniform discreteness $\Gamma\subset G$7 weak Lehmer,” is established in (Pham et al., 2020) by constructing cocompact arithmetic lattices with infinite-order elements tending to the identity from algebraic integers whose Mahler measures tend to $\Gamma\subset G$8. This generalizes Sury’s rank-one $\Gamma\subset G$9 argument to higher-rank families.

Assume weak Lehmer fails at some level $U\cap \Gamma$0. After using Smyth’s theorem, the paper reduces to a sequence of irreducible monic palindromic polynomials $U\cap \Gamma$1 with bounded $U\cap \Gamma$2, with

$U\cap \Gamma$3

and, after passing to a subsequence, fixed $U\cap \Gamma$4 and fixed number $U\cap \Gamma$5 of real roots outside the unit circle. The failure is encoded as

$U\cap \Gamma$6

For such a polynomial $U\cap \Gamma$7, with root $U\cap \Gamma$8, let $U\cap \Gamma$9 and $\text{Let %%%%9%%%% be a connected semisimple %%%%10%%%%-group with %%%%11%%%%. Then there exists a neighborhood %%%%12%%%% of the identity such that for any irreducible cocompact lattice %%%%13%%%%, the intersection %%%%14%%%% consists of elements of finite order.}$0, the fixed field of the involution $\text{Let %%%%9%%%% be a connected semisimple %%%%10%%%%-group with %%%%11%%%%. Then there exists a neighborhood %%%%12%%%% of the identity such that for any irreducible cocompact lattice %%%%13%%%%, the intersection %%%%14%%%% consists of elements of finite order.}$1. Using suitable $\text{Let %%%%9%%%% be a connected semisimple %%%%10%%%%-group with %%%%11%%%%. Then there exists a neighborhood %%%%12%%%% of the identity such that for any irreducible cocompact lattice %%%%13%%%%, the intersection %%%%14%%%% consists of elements of finite order.}$2-forms of the chosen simple group $\text{Let %%%%9%%%% be a connected semisimple %%%%10%%%%-group with %%%%11%%%%. Then there exists a neighborhood %%%%12%%%% of the identity such that for any irreducible cocompact lattice %%%%13%%%%, the intersection %%%%14%%%% consists of elements of finite order.}$3, the paper constructs arithmetic lattices $\text{Let %%%%9%%%% be a connected semisimple %%%%10%%%%-group with %%%%11%%%%. Then there exists a neighborhood %%%%12%%%% of the identity such that for any irreducible cocompact lattice %%%%13%%%%, the intersection %%%%14%%%% consists of elements of finite order.}$4 in

$\text{Let %%%%9%%%% be a connected semisimple %%%%10%%%%-group with %%%%11%%%%. Then there exists a neighborhood %%%%12%%%% of the identity such that for any irreducible cocompact lattice %%%%13%%%%, the intersection %%%%14%%%% consists of elements of finite order.}$5

together with elements $\text{Let %%%%9%%%% be a connected semisimple %%%%10%%%%-group with %%%%11%%%%. Then there exists a neighborhood %%%%12%%%% of the identity such that for any irreducible cocompact lattice %%%%13%%%%, the intersection %%%%14%%%% consists of elements of finite order.}$6 whose archimedean eigenvalues are governed by the conjugates $\text{Let %%%%9%%%% be a connected semisimple %%%%10%%%%-group with %%%%11%%%%. Then there exists a neighborhood %%%%12%%%% of the identity such that for any irreducible cocompact lattice %%%%13%%%%, the intersection %%%%14%%%% consists of elements of finite order.}$7 of $\text{Let %%%%9%%%% be a connected semisimple %%%%10%%%%-group with %%%%11%%%%. Then there exists a neighborhood %%%%12%%%% of the identity such that for any irreducible cocompact lattice %%%%13%%%%, the intersection %%%%14%%%% consists of elements of finite order.}$8. In the split $\text{Let %%%%9%%%% be a connected semisimple %%%%10%%%%-group with %%%%11%%%%. Then there exists a neighborhood %%%%12%%%% of the identity such that for any irreducible cocompact lattice %%%%13%%%%, the intersection %%%%14%%%% consists of elements of finite order.}$9 model case,

$\mathbb R$00

The remaining issue is not only $\mathbb R$01, but also control of the arguments of complex conjugates. Dirichlet’s simultaneous approximation theorem is used to choose exponents $\mathbb R$02, with $\mathbb R$03, such that

$\mathbb R$04

where

$\mathbb R$05

This is ensured by taking

$\mathbb R$06

so that

$\mathbb R$07

Then $\mathbb R$08 tends to $\mathbb R$09 in $\mathbb R$10, while remaining of infinite order because one conjugate still has modulus $\mathbb R$11. Any purported Margulis neighborhood for the family is therefore contradicted (Pham et al., 2020).

The paper isolates the arithmetic mechanism explicitly: small Mahler measure produces small semisimple elements in arithmetic lattices. The count $\mathbb R$12 controls how many archimedean directions can expand, while the product of the expanding moduli is exactly the Mahler measure.

5. Geometric content, quantitative form, and limitations

The geometric interpretation in (Pham et al., 2020) is clear but deliberately narrower than a classical effective Margulis constant. In a semisimple Lie group or symmetric space, a semisimple element is close to $\mathbb R$13 precisely when its eigenvalues in a faithful linear representation, or in the adjoint representation, are close to the unit circle. Therefore a lower bound

$\mathbb R$14

for every infinite-order $\mathbb R$15 gives a uniform positive lower bound on displacement near the identity, and thus on injectivity radius for torsion-free arithmetic quotients. What is proved exactly, however, is the neighborhood statement via the continuous function

$\mathbb R$16

not an explicit Riemannian lower bound (Pham et al., 2020).

This distinction explains the quantitative shape of the result. The constants appearing are arithmetic rather than metric: $\mathbb R$17 from weak Lehmer, the Dobrowolski–Voutier bound

$\mathbb R$18

and the approximation parameter $\mathbb R$19. The neighborhood

$\mathbb R$20

is explicit only through the adjoint-Mahler function $\mathbb R$21, not as a metric ball of known radius. Thus these are arithmetic spectral neighborhoods, not classical effective Margulis constants.

The scope is also sharply delimited. For irreducible non-cocompact higher-rank lattices, Margulis had already proved a neighborhood theorem in which all sufficiently small elements are unipotent; no Lehmer input is required. The cocompact case is different because cocompact lattices contain no nontrivial unipotents, so the conjectural replacement is “near the identity there should be only torsion.” Arithmeticity is essential in the proof because integrality of adjoint characteristic polynomials comes from the arithmetic model. The paper proves nothing analogous for general non-arithmetic lattices (Pham et al., 2020).

A common misconception is therefore that the arithmetic Margulis lemma is merely the classical Margulis lemma restated for arithmetic lattices. The actual statement is different in both hypothesis and conclusion: the classical lemma yields virtual nilpotence of the subgroup generated by small elements, whereas the arithmetic statement forbids small infinite-order elements altogether in the cocompact arithmetic setting.

6. Later uses and related but distinct meanings

Later papers use the phrase arithmetic Margulis lemma in a more explicit quantitative form adapted to arithmetic hyperbolic geometry. In "Geometry and arithmetic of semi-arithmetic Fuchsian groups," the lemma is imported from Frączyk’s work, following Breuillard, and used as the key input converting bounded coarea and bounded stretch into a bound on the degree of the invariant trace field. The version used states that there exists $\mathbb R$22, depending only on $\mathbb R$23, such that for an arithmetic group $\mathbb R$24 acting on $\mathbb R$25, with invariant trace field $\mathbb R$26, the subgroup generated by

$\mathbb R$27

is virtually nilpotent. Combined with Yamada’s radius estimate and the stretch map $\mathbb R$28, this yields

$\mathbb R$29

for constants depending only on $\mathbb R$30 and $\mathbb R$31, and then a systole lower bound and finiteness results for semi-arithmetic Fuchsian groups (Belolipetsky et al., 2023).

In "A new proof of finiteness of maximal arithmetic reflection groups," the central imported statement is Lemma 1.3: for every $\mathbb R$32, there exists $\mathbb R$33 such that if $\mathbb R$34 is an arithmetic group whose trace field has degree $\mathbb R$35 over $\mathbb R$36, then for every $\mathbb R$37,

$\mathbb R$38

is virtually solvable. The paper uses this arithmetic scaling to force large embedded half-balls in a $\mathbb R$39-face of a Coxeter polytope, thereby obtaining a degree bound for the trace field and a new proof of finiteness of maximal arithmetic reflection groups without trace formulas or automorphic forms (Fisher et al., 2022).

These later formulations are not identical to the Lehmer-equivalent neighborhood theorem of (Pham et al., 2020), but they belong to the same thematic class: arithmetic data enlarge the scale on which one can control small-displacement behavior. This suggests that the term functions as a family label for arithmetic small-element principles rather than as a uniquely standardized theorem name.

Several nearby results should be distinguished from this usage. Cerocchi–Sambusetti prove an explicitly quantitative abelian analogue for free cocompact $\mathbb R$40-actions on length spaces, relating asymptotic volume and stable systole: $\mathbb R$41 and use it to bound the defect $\mathbb R$42 (Cerocchi et al., 2014). By contrast, the Alexandrov-space paper (Xu et al., 2019) proves a geometric generalized Margulis lemma in the usual sense—small loops generate a subgroup that is virtually nilpotent with uniform index bound—and contains no arithmetic-group content. Benoist–Miquel’s theorem on discrete groups containing horospherical lattices proves a Margulis-conjectured arithmeticity statement in higher rank, but it is not the classical or arithmetic Margulis lemma in the small-element sense (Benoist et al., 2018).

Taken together, these works place the Arithmetic Margulis Lemma at the intersection of higher-rank arithmeticity, Diophantine height gaps, and quantitative hyperbolic geometry. Its most precise current formulation is the cocompact arithmetic neighborhood theorem equivalent to weak Lehmer (Pham et al., 2020); its later applications show that the same idea can also be recast as a field-degree-sensitive displacement threshold in arithmetic hyperbolic settings (Belolipetsky et al., 2023, Fisher et al., 2022).