Torus-packet equidistribution in prime, quartic, and sextic degrees

The explanation

Take the numbers a + b√2, where a and b are whole numbers. Pair each one with its mirror twin, a − b√2. The number 1 becomes the point (1, 1). The number √2 becomes about (1.414, −1.414). All such points form a regular grid in the plane, called a lattice. Its basic cell has area 2√2, about 2.83.

Now stretch the plane sideways by about 5.83 and squash it upright by the same factor. The point (1, 1) moves to about (5.83, 0.172). That is exactly the point for 3 + 2√2. Every other grid point also lands on a grid point. Multiplying by 3 + 2√2 sends a + b√2 to (3a + 4b) + (2a + 3b)√2. The trick works because the twin 3 − 2√2 is exactly 1 divided by 3 + 2√2. So the whole grid comes back to itself.

Now picture the space of all grids whose cell has area 1. Shrink our grid to area 1 and stretch it slowly, from factor 1 up to 5.83. It traces a closed loop in that space. In the natural logarithmic units, the loop has length about 1.76, the logarithm of 5.83.

Larger number systems give larger versions of this picture. A totally real field of degree n is a number system built from a root of a degree-n whole-number equation. All n of its mirror copies are real. It gives a grid in n dimensions. The loop becomes a closed torus, a doughnut-like shape of dimension n − 1. One field can give several genuinely different grids, one for each ideal class. A torus packet collects the tori from every class, adds their sign-flipped copies, and weights each torus by its volume.

The question is whether packets spread out evenly as the number systems grow more intricate. Their size is measured by a whole number called the discriminant. For the grid above it is 8, the square of the cell area. Spreading evenly means each region gets its fair share of the packet's weight. Fair is judged by the natural uniform measure, called Haar measure. The papers also show that no share of the weight drifts off toward extremely squashed gridsSource 1.

For degree two, the loops are closed paths on a classical surface. In 1988, William Duke proved that these loops, weighted by length, spread evenly for the fundamental real-quadratic discriminants. In 2011, Einsiedler, Lindenstrauss, Michel and Venkatesh proved a cubic version. Their extension to higher prime degrees depended on an unproven analytic bound called subconvexity. In 2024, Lemke Oliver, Thorner and Zaman removed that condition in each fixed prime degree from five on. Their result still left out a small exceptional family of fields.

The first paper proves the result in every fixed prime degree from five on, with no exceptional fieldsSource 2. The argument has two parts. A counting bound keeps weight away from squashed grids and from tiny neighbourhoodsSource 3. A known classification then forces the uniform measureSource 4. The obstacle was making the count work for every field without the unproven bound. The paper uses a 1974 result of Stark in its placeSource 5. It also allows smaller systems of whole numbers inside the field, and grids grouped by their behaviour at each primeSource 6.

Degrees four and six bring a different obstacle. Four and six are not prime. In such degrees, the limit could pile up on smaller pieces of the space instead of spreading evenlySource 7,Source 8. Both papers ask for primitive fields, which have no number system strictly between the ordinary fractions and the field itselfSource 9. The quartic paper claims to rule out the smaller pieces with a count in a related cubic fieldSource 10. The sextic paper claims to do it with a new argument about how the weight is spread at many scalesSource 11.

The scopes differ. The quartic paper allows any order, meaning any smaller ring of whole numbers inside the fieldSource 12. The sextic paper covers only the full ring of whole numbers, as the field discriminant growsSource 13. None of the results says anything about one torus on its ownSource 14. The first paper also gives no rate of convergenceSource 14.

How sure is this? A computer proof checker called Lean has checked a formal version of the prime-degree theorem. It is matched against a fixed reference statement by a tool called Comparator. The quartic and sextic theorems are outside the Lean scope note. No specialist review of any of the three papers has been published.

What it doesn't show. This does not say anything about a single torus, give a rate of convergence, or cover other composite degrees.