Main theorem: the box-counting dimension of the prime fractal is `1`.
ProvedPrimeFractal.tendsto_boxCount_log_divaether-catalognumbertheory
Main theorem: the box-counting dimension of the prime fractal is 1.
theorem PrimeFractal.tendsto_boxCount_log_div:
Tendsto (fun m : ℕ => Real.log (boxCount m) / Real.log m) atTop (𝓝 1) := by sorry
Formalization Note Transplanted verbatim from the Aether Catalog source NumberTheory/PrimeFractalBoxDimension.lean; the statement is byte-identical to the source declaration, elaborated with autoImplicit disabled in the platform environment.
Preamble
-- Thm stub generated from NumberTheory/PrimeFractalBoxDimension.lean
import Mathlib
import Definitions.Def_NumberTheory_PrimeFractalBoxDimension
import Definitions.Def_NumberTheory_PrimeFractalHausdorff
/-!
# The box-counting dimension of the prime fractal is exactly `1`
`NumberTheory.PrimeFractalHausdorff` shows that the Hausdorff dimension of the
prime fractal `{1 / log p : p prime} ⊆ ℝ` is `0`, refuting the mission
conjecture. Here we prove the *positive* half of the story: the notion that
actually sees the conjectured value is the **box-counting (Minkowski)
dimension**, and for the prime fractal it equals `1` on the nose.
At scale `1/m` we count the boxes `[k/m, (k+1)/m)` that meet the prime fractal,
i.e. the cardinality `boxCount m` of the set of values `⌊m / log p⌋` over primes
`p`. The two halves are:
* `boxCount_le` : `boxCount m ≤ 2m + 1` — a trivial upper bound valid for any
subset of an interval, which is what forces the dimension to be `≤ 1`. In
particular *no* configuration of primes — twin primes included — can produce
a dimension `1 + ε` with `ε > 0`.
* `eventually_boxCount_ge` : `boxCount m ≥ m / (16 (log m)^4)`. This is the
arithmetic input: it uses the Chebyshev-type lower bound
`PrimeFractal.le_primeCounting_mul_log` together with the observation that
`p ↦ ⌊m / log p⌋` is injective on primes `p ≤ Y` as soon as
`2 Y (log Y)^2 ≤ m` (primes below `Y` are spread more than `1/m` apart in the
`d`-metric).
Together they give `tendsto_boxCount_log_div`: `log (boxCount m) / log m → 1`,
hence `upperBoxDim = lowerBoxDim = 1` (`upperBoxDim_eq_one`,
`lowerBoxDim_eq_one`), while the Hausdorff dimension is `0`
(`dimH_lt_boxDim`). The prime fractal is therefore a *dimension-irregular*
set: box and Hausdorff dimensions disagree maximally.
-/
open PrimeFractal
open Filter Topology
/-! ### Elementary bounds -/
/-! ### Separation of primes in the `d`-metric -/
/-! ### From Chebyshev to a lower bound on the box count -/
/-! ### Asymptotics -/Formal statement
theorem PrimeFractal.tendsto_boxCount_log_div:
Tendsto (fun m : ℕ => Real.log (boxCount m) / Real.log m) atTop (𝓝 1) := by sorrySource