RESEARCH / Focused target
arxiv 2506 09912 shkolnikov sandpile limit divisible
RELEASED FOUNDATIONS / PROPOSED CONNECTION
Research connections
All source anchors (1)
- D5/S3/StatisticalMechanics/Sandpiles/ShkolnikovSandpileLimitDivisible.resultAbsent from this graph
Problem
The question
Mikhail Shkolnikov, Algebraic Limits of Sandpiles, arXiv:2506.09912v3, Section 3, states that the image of the direct limit of rectangular sandpile groups may be a rational subtorus or a divisible group. The settled question is the divisibility alternative: for every positive rectangle, every circle-valued sandpile function on it, and every positive integer $k$, a later rectangular stage contains a sandpile function whose $k$-fold sum equals the signed reflected image of the original function.
Statement echo
The Lean declaration D5.S3.StatisticalMechanics.Sandpiles.ShkolnikovSandpileLimitDivisible.claim preserves the six binding conventions from issue #20011: the carrier is AddCircle (1 : ℝ); InRect is the open integer rectangle; IsSandpile is zero extension plus the four-neighbour equation; unfold is the parity-reflected coordinate; embed is the signed reflected zero extension; and the quantifiers require positive $p,q,k$, a sandpile $f$, and a later $g$ with the exact equality. The theorem result proves this proposition without weakening its quantifiers.
Motivation
Our foothold
Divisibility supplies a structural property of the direct limit that is stronger than the previously known existence of elements of every finite order. The source's dual torsion question and the full image description are related but require additional identifications.
Gap
Missing bridges
The source left divisibility as an expectation. The literature screen recorded in issue #20011 found the expectation still open and found no repository declaration settling the same claim. Literature completeness is ASSUMED-UNVERIFIED.
Route
Proposed approach
The proof uses reversible row dynamics. A finite-denominator lift of the first row lies in a finite subgroup, so the injective transfer orbit is periodic. Reflection at a zero boundary row gives a closing period and the signed reflected tile formula. The resulting rows define the later sandpile, and recurrence uniqueness gives the required $k$-fold equality. The construction uses $m=1$ and only increases the vertical parameter.
Falsifier
What would falsify this route
A counterexample would be a positive rectangle, sandpile $f$, and $k\ge1$ for which no positive later height and no sandpile $g$ satisfy the exact equality in claim. A proof failure would also invalidate the settlement if any recurrence, boundary, finite-period, or embedding equality used by result were false.
Evidence
Evidence to collect
The kernel-checked declaration is D5.S3.StatisticalMechanics.Sandpiles.ShkolnikovSandpileLimitDivisible.result. Its axiom closure is contained in {propext, Classical.choice, Quot.sound}. The module state is Golden/Frozen/state/D5/S3/StatisticalMechanics/Sandpiles/ShkolnikovSandpileLimitDivisible.lean.json.
The independent exact-arithmetic experiment is pinned to the experiment repository. The command python3 docs/reports/shkolnikov-2025-sandpile-limit-divisibility/check.py exits 0; the script SHA-256 is 8b5716cb00667a95f59381cda86b8ab69d97445eca6cc13418bf6b4406b29617. It checks 45 harmonic-root and embedding cases for $(p,q)\in{(2,2),(3,2),(2,3),(3,3),(4,3)}$, $k\in{2,3,5}$, and three sampled elements per pair, together with the six source Table 1 group orders.
Triage
Scope assessment
| item | status | evidence |
| --- | --- | --- |
| Exact preregistered divisibility claim | proved | Kernel-checked result; the claim is the full proposition above. |
| Reversible finite-denominator row dynamics, zero-row reflection, and a closing period | proved | The live proof path of result uses rows_reflect, rows_double_period, rows_tile, and finite_row_lift. |
| Only vertical enlargement is required | proved | result constructs witnesses with $m=1$ and a positive vertical multiplier. |
| Complete image in the inverse limit of extended groups | open | No declaration in this delivery identifies the image. |
| Rational-subtorus alternative | open | No declaration in this delivery proves the alternative. |
| Pontryagin-dual/inverse-limit identification and the torsion-free corollary | open | The dual translation is not formalized by result. |
| Explicit numerical upper bound on the chosen period | open | finite_row_lift proves existence of a period but gives no numerical bound. |
| Review the standalone proof and source fidelity | proved | The delivered Lean definitions, Scribe formulas, and source note preserve the preregistered conventions. |
Declaration judgement
| declaration | proof_shape | escape_witness | consumers |
| --- | --- | --- | --- |
| D5.S3.StatisticalMechanics.Sandpiles.ShkolnikovSandpileLimitDivisible.result | bind-only | none; the external named open-problem resolution is the admission basis | none |
The module admission basis is open-problem-resolution (#20011; Proved). The public result is the designated settling theorem and is not a four-slot registration claim. Its typed Reg mirror compiles, but the lawful enrolled readout and source-coordinate evidence remain unfinished under issue #20360.
ASSUMED-UNVERIFIED
Unverified assumptions
The literature search establishes the source's stated expectation and the searched-scope absence of a settlement; it does not establish exhaustive worldwide completeness. The publisher text cited by the preregistration was not independently inspected beyond the arXiv source. The Reg mirror has no declared_validated DTR reading at the delivered head, so the escape audit remains unfinished under issue #20360.