How arithmetic torus packets spread through spaces of lattices

← All digests

Family 015 · Number theory
Importance 4/5 · Provisional editorial rating
Sources consulted 2026-10-10

1. The question

A lattice in the plane can be a square grid, a stretched rectangular grid, or a tilted one. In higher dimensions the variety of shapes is much larger. Number fields produce particularly structured lattices: apply all real embeddings of a field to its algebraic integers or ideals, then rescale to give a fundamental cell volume one.

These arithmetic lattices belong to compact families obtained by stretching the coordinate directions while preserving volume. The central question is whether complete collections of such families eventually spread uniformly through the space of all lattice shapes as their arithmetic complexity increases.

Family 015 reports three equidistribution results: one for every fixed prime degree at least five, one for primitive quartic fields, and one for primitive sextic fields. The objects being distributed are packets of orbits, with precise arithmetic membership and volume weights. This distinction is as important as the limiting uniformity.

2. Background and essential definitions

Number fields give lattices

A number field $K$ is a finite extension of $\mathbb Q$. It is totally real of degree $n$ if it has $n$ embeddings $\sigma_1,\ldots,\sigma_n$ into $\mathbb R$. A full lattice $M\subset K$ is a free abelian group of rank $n$ spanning $K$ over $\mathbb Q$. Its image

$$\sigma(M)=\{(\sigma_1(a),\ldots,\sigma_n(a)):a\in M\}$$

is an ordinary lattice in $\mathbb R^n$. Dividing it by the $n$-th root of its covolume makes its fundamental cell have volume one.

The space of these unimodular lattices is

$$X_n=\mathrm{SL}_n(\mathbb Z)\backslash\mathrm{SL}_n(\mathbb R).$$

Changing an integral basis leaves the lattice unchanged, which explains the quotient. Its invariant probability $m_n$, induced by Haar measure, is the natural meaning of “uniform distribution” on this space.

The positive diagonal group

$$A_n=\{\operatorname{diag}(e^{u_1},\ldots,e^{u_n}):u_1+\cdots+u_n=0\}$$

stretches coordinates without changing volume. Units of the appropriate number-theoretic order provide periods for this action. Dirichlet’s unit theorem makes the resulting arithmetic orbits compact tori of dimension $n-1$. The phrase “torus packet” refers to a finite arithmetic collection of these compact diagonal orbits.

Orders, ideal classes and local type

The ring of integers $\mathcal O_K$ contains all algebraic integers in $K$. An order $\mathcal O\subseteq\mathcal O_K$ is a subring containing one and having full rank. Its discriminant satisfies

$$|\operatorname{Disc}(\mathcal O)| =|\operatorname{Disc}(K)|[\mathcal O_K:\mathcal O]^2.$$

The discriminant therefore grows either because the field changes or because the order becomes thinner inside a fixed field.

Ideals are grouped into classes by multiplication by nonzero elements of $K$. For a nonmaximal order, an ideal is invertible if it has a fractional ideal inverse. Such ideals are locally free of rank one over that order; arbitrary full lattices need not have this property.

For a general full lattice, its multiplier order is

$$\mathcal O(M)=\{a\in K:aM\subseteq M\}.$$

Two full lattices have the same local homothety type if, after completing at every rational prime $\ell$, they differ by multiplication by an invertible element of $K\otimes_{\mathbb Q}\mathbb Q_\ell$. Here $\mathbb Q_\ell$ is the completion for the $\ell$-adic absolute value. Specifying the multiplier order alone can leave several local types possible.

Finally, a field is primitive if no field lies strictly between $\mathbb Q$ and $K$. Prime-degree fields are automatically primitive. Quartic and sextic fields require this as an additional hypothesis.

What is averaged?

For a packet $\mathcal P$, let $\nu_O$ be uniform probability on an orbit $O$, and let $\operatorname{vol}(O)$ be its diagonal-group volume. The packet measure is

$$\mu_{\mathcal P} =\frac{\sum_{O\in\mathcal P}\operatorname{vol}(O)\nu_O} {\sum_{O\in\mathcal P}\operatorname{vol}(O)}.$$

Each distinct orbit appears once. The packet also includes the prescribed coordinate-sign translates, so ordinary ideal-class representatives do not introduce an artificial sign choice. In even dimensions all sign patterns are included, including determinant-negative sign matrices acting on the underlying lattice. In odd prime degree, determinant-positive sign patterns suffice because a common minus sign does not change a lattice.

3. Development history and previous partial results

For real quadratic fields, the related objects project to closed geodesics on the modular surface. Linnik’s dynamical approach, including Skubenko’s work, established distribution results with an auxiliary splitting condition. Duke removed that restriction for packets associated with fundamental discriminants using analytic methods. These are the two-dimensional origins of the problem, described in the introductions of M1–M3 and Duke’s cited 1988 paper.

Higher dimensions introduce a diagonal group with several independent directions. Einsiedler, Katok and Lindenstrauss proved a rigidity theorem: positive-entropy diagonal-invariant ergodic probabilities are algebraic in the relevant lattice spaces. Entropy measures the rate at which a dynamical system creates distinguishable orbit histories. This turns enough arithmetic spreading into a strong restriction on possible limit measures.

Einsiedler–Lindenstrauss–Michel–Venkatesh connected arithmetic counting to that rigidity and proved the cubic analogue of Duke’s theorem, including packets arising from nonmaximal orders. Their 2011 paper is a central predecessor. It also explains why local homothety is finer than a multiplier-order condition and why composite dimensions allow additional limiting obstructions.

For higher prime degrees, Lemke Oliver, Thorner and Zaman proved equidistribution outside a small exceptional family of fields through progress on Artin $L$-functions. Their primary paper gives the surrounding analytic framework. Khayutin obtained entropy bounds under maximal-order and Galois-group hypotheses, while Wieser and Yang studied certain quartic packets with quadratic subfields and additional conditions.

The gaps addressed here are thus specific. M1 removes the exceptional-field restriction and admits arbitrary local module types. M2 and M3 tackle composite dimensions where positive entropy alone does not force the full uniform measure.

4. What the new manuscripts claim

M1: fixed prime degree, every local type

Fix a prime $n\ge5$. For any sequence of totally real degree-$n$ fields $K_i$, full lattices $M_i\subset K_i$, and ordered real embeddings, M1 claims

$$D(M_i):=|\operatorname{Disc}(\mathcal O(M_i))|\longrightarrow\infty \quad\Longrightarrow\quad \mu_i\longrightarrow m_n.$$

The packet consists of all global homothety classes in the prescribed local type, with the sign and volume conventions above. Nonmaximal orders can have unbounded index, and the lattices need not be invertible modules over their multiplier order. The fields may remain fixed while their multiplier-order discriminants grow. The degree is fixed throughout the limit.

M2: primitive quartic fields and arbitrary orders

For totally real primitive quartic fields $K_i$ and arbitrary orders $\mathcal O_i$, M2 claims the same conclusion on $X_4$ as

$$|\operatorname{Disc}(\mathcal O_i)|\longrightarrow\infty.$$

Here the packet contains every class of proper invertible fractional $\mathcal O_i$-ideals. Proper means that the multiplier order is exactly $\mathcal O_i$. This is narrower than M1’s arbitrary full-lattice local types. It still includes sequences in one fixed primitive quartic field with increasing order index.

M3: primitive sextic fields, maximal orders

For totally real primitive sextic fields $K_i$, M3 claims equidistribution of the complete ordinary ideal-class packets of $\mathcal O_{K_i}$ on $X_6$, provided

$$|\operatorname{Disc}(K_i)|\longrightarrow\infty.$$

This statement concerns maximal orders. It does not extend M2’s arbitrary-order conclusion to degree six.

Uniformity includes keeping all the mass

For every continuous compactly supported test function $f$, each theorem asserts

$$\int f\,d\mu_i\longrightarrow\int f\,dm_n.$$

It also asserts no escape of mass: given $\varepsilon>0$, a compact set contains at least $1-\varepsilon$ of every sufficiently late packet. This is necessary because $X_n$ is noncompact. A lattice can become extremely thin in one direction while keeping volume one, moving into a cusp. Convergence observed only on bounded regions would otherwise allow some probability to disappear there.

5. Why the advance matters

Provisional importance: 4/5. Taken as stated, these theorems make major progress on higher-dimensional Duke equidistribution and on the interaction between number-field arithmetic and homogeneous dynamics.

M1’s force comes from treating every sequence in its degree range, including difficult local types and increasing conductor inside a fixed field. M2 and M3 overcome an additional structural issue: a spread-out limit in composite dimension can still live on a proper algebraic subspace. Excluding that possibility demands more information than a single entropy estimate.

The conclusion says that a highly constrained arithmetic construction eventually samples ordinary lattice geometry according to its invariant measure. It is a global distribution statement about complete packets, not a claim that an individual ideal produces a randomly chosen lattice.

6. The main idea

Count vectors, then use dynamics

The common first step averages lattice-vector counts over the packet. Ideal classes and local module choices convert this geometric average into an arithmetic ideal count. Short-vector bounds prevent escape into the cusp; bounds for vectors in small boxes show that limiting measures cannot concentrate too strongly in ordinary neighborhoods.

A key normalization is the residue

$$\kappa_K=\operatorname*{Res}_{s=1}\zeta_K(s),$$

where $\zeta_K$ is the Dedekind zeta function. The class-number formula brings this residue into the packet volume. The papers retain the same factor in their counting estimates so that it cancels correctly, rather than replacing it prematurely by a rough field-dependent bound.

M1 handles nonmaximal local modules through containment probabilities: at each prime, how likely is a vector to belong to a unit translate of the prescribed module? A uniform small positive moment controls the local tails. Stark’s exceptional-zero information and short-interval estimates supply the global count; the absence of quadratic subfields is relevant here.

The resulting radius-$r$ ball estimate has order $r^n$. Since the diagonal orbits have dimension $n-1$, this enforces positive entropy in almost every ergodic component. In prime degree, the existing measure-classification theorem then identifies the limit as Haar probability.

Quartic fields: detect the remaining block structure

In degree four, proper homogeneous limits can have two blocks of size two. M2 detects this through the exterior square of a lattice. Exterior coordinates $v_{ij}$ record oriented two-dimensional areas, and the diagonal action preserves three products:

$$(v_{12}v_{34},-v_{13}v_{24},v_{14}v_{23}).$$

These products are organized by a cubic resolvent field associated with the quartic field. A two-block obstruction forces two product coordinates to be very small. Counting such thin configurations rules out concentration on the obstructing homogeneous pieces.

The hard extension to arbitrary orders is uniformity at primes dividing the order index. M2 uses local quadratic Fourier analysis, Poisson summation and a sieve to retain the correct conductor dependence and packet normalization.

Sextic fields: retain information before taking the limit

M3 uses ordinary vectors together with dual vectors to bound the mass of thin two-sided dynamical tubes. Primitivity ensures that every nonrational field element generates the full field, providing the arithmetic separation needed for these bounds.

A subtlety is that exact conditional measures of each finite packet along a transverse root direction are atomic. Taking such conditionals too early would hide spreading visible at growing scales. M3 instead conditions at intermediate resolutions before passing to a limit, proves a product rule, and tracks which root directions have nontrivial spread.

The root groups are elementary shears $I+sE_{jk}$, where $E_{jk}$ has a single nonzero entry. The argument propagates activity among these directions and uses tube decay to force full activity. A proper homogeneous component would leave a direction inactive, giving the required contradiction.

7. Scope, limitations, and open questions

All three statements concern complete packets with the stated weights. They do not assert equidistribution of an individual orbit, an arbitrary subset of ideal classes, or an unweighted collection chosen by another rule. No quantitative convergence rate is part of the headline conclusions.

The permitted degrees and orders should remain separate: arbitrary local types in fixed prime degree at least five; invertible ideals of arbitrary orders in primitive quartic fields; maximal-order ideals in primitive sextic fields. General composite degrees, fields with intermediate subfields, and unrestricted sextic orders are outside these statements.

Primitivity is an arithmetic hypothesis on each field, but it does not automatically prohibit a sequence from approaching a proper homogeneous space. That is precisely why the additional quartic and sextic arguments are needed. Extending those arguments to broader settings is a natural further question.

8. Sources and further reading

Theorem 1.1 of each manuscript gives its packet definition and limiting statement. M1’s introduction and Sections 2–6 explain counting, local types and entropy. M2’s Sections 4–9 describe the quartic obstruction and cubic-resolvent method. M3’s introduction provides a detailed guide to the intermediate-resolution argument.

The cubic theorem and the approximate-Artin paper linked in section 3 are useful primary predecessors. The manuscripts’ bibliographies identify Duke’s original paper and the measure-rigidity and entropy results used here. For optional terminology, see Dirichlet’s unit theorem and equidistribution; the packet convergence above is a measure-valued version of that general idea.

Source manuscripts

• M1. OpenAI, Equidistribution of Prime-Degree Torus Packets with Arbitrary Local Type (September 24, 2026).
• M2. OpenAI, Equidistribution of primitive quartic torus packets for arbitrary orders (October 5, 2026).
• M3. OpenAI, Equidistribution of Primitive Sextic Torus Packets (October 5, 2026).

Sources consulted on 10 October 2026 at repository revision fd4aeeb. Download the manuscript citations.