ThreePrimeRunningLcm
DefinitionErdosProblems_Erdos269_ThreePrimeRunningLcmdefinitionerdos-269
Defines three-prime smooth values p^i q^j r^k, the product of the maximal pure prime powers below a cutoff, its reciprocal kernel, finite smooth prefixes and their LCM, height fibers, and dyadic block bases. The definitions alone do not claim irrationality.
Definition code
import Mathlib.Algebra.BigOperators.Group.Finset.Basic
import Mathlib.Algebra.GCDMonoid.Finset
import Mathlib.Algebra.Ring.Parity
import Mathlib.Data.Finset.Card
import Mathlib.Data.Finset.Prod
import Mathlib.Data.Nat.Log
import Mathlib.Data.Nat.Prime.Basic
import Mathlib.Data.Nat.Prime.Int
import Mathlib.Tactic.Linarith
import Mathlib.Tactic.NormNum
import Mathlib.Tactic.Ring
/-!
# Erdős #269: the three-prime running-LCM coordinate
This module starts the problem-owned formalization of the first unresolved
three-prime case. It records the exact computational height used by the
running-LCM representation, its cubic majorant, the smallest non-separation
fixture for `{2,3,5}`, the variable-base tail-state update, and the uniform
quadratic bound for actual filtered smooth-number shells.
No declaration here asserts irrationality or transcendence of a three-prime
value. The missing producer is still an infinite residue-escape or genuinely
higher-dimensional analytic theorem.
-/
namespace ErdosProblems.Erdos269
open scoped BigOperators
/-- The `{p,q,r}`-smooth lattice point with exponent vector `(i,j,k)`. -/
def smooth3Val (p q r i j k : ℕ) : ℕ :=
p ^ i * q ^ j * r ^ k
/-- The product of the largest pure `p`-, `q`-, and `r`-powers not exceeding
`x`. For a `{p,q,r}`-smooth `x`, this is the running LCM of the smooth prefix. -/
def threePrimeHeight (p q r x : ℕ) : ℕ :=
p ^ Nat.log p x * q ^ Nat.log q x * r ^ Nat.log r x
/-- The exact rational lattice kernel attached to the running-LCM height. -/
def threePrimeKernelQ (p q r i j k : ℕ) : ℚ :=
(threePrimeHeight p q r (smooth3Val p q r i j k) : ℚ)⁻¹
/-- Exponent vectors of the actual `{p,q,r}`-smooth prefix up to `x`. The
logarithmic box makes the prefix finite; the final filter keeps only products
which really lie below `x`. -/
def smoothPrefixExponents (p q r x : ℕ) : Finset (ℕ × ℕ × ℕ) :=
((Finset.range (Nat.log p x + 1)).product
((Finset.range (Nat.log q x + 1)).product
(Finset.range (Nat.log r x + 1)))).filter
fun e => smooth3Val p q r e.1 e.2.1 e.2.2 ≤ x
/-- The literal running LCM of the finite smooth prefix. -/
def smoothPrefixLcm (p q r x : ℕ) : ℕ :=
(smoothPrefixExponents p q r x).lcm
fun e => smooth3Val p q r e.1 e.2.1 e.2.2
/-! ## Finite pure-power jump enumeration -/
/-- The first `count` positive powers of one prime base. Exponent zero is
omitted because it is the common initial value `1` in every channel. -/
def positivePrimePowers (p count : ℕ) : Finset ℕ :=
(Finset.range count).image fun e => p ^ (e + 1)
/-- The finite union of the first `count` positive powers in each of the three
prime channels. -/
def threePrimePositiveJumpSet (p q r count : ℕ) : Finset ℕ :=
(positivePrimePowers p count ∪ positivePrimePowers q count) ∪
positivePrimePowers r count
/-- The finite jump set including the unique common origin `1`. -/
def threePrimeJumpSetWithOrigin (p q r count : ℕ) : Finset ℕ :=
insert 1 (threePrimePositiveJumpSet p q r count)
/-! ## Single-coordinate jump ratios -/
/-! ## Finite jump grouping -/
/-- A finite rectangular exponent box. This is the exact finite domain used
before passing to the infinite three-prime smooth series. -/
def smoothExponentBox (hp hq hr : ℕ) : Finset (ℕ × ℕ × ℕ) :=
(Finset.range (hp + 1)).product
((Finset.range (hq + 1)).product (Finset.range (hr + 1)))
/-- The running-LCM height attached to one smooth exponent triple. -/
def smoothPointHeight (p q r : ℕ) (e : ℕ × ℕ × ℕ) : ℕ :=
threePrimeHeight p q r (smooth3Val p q r e.1 e.2.1 e.2.2)
/-- The points of a finite exponent box with one fixed running-LCM height. -/
def smoothHeightFiber
(p q r hp hq hr H : ℕ) : Finset (ℕ × ℕ × ℕ) :=
(smoothExponentBox hp hq hr).filter fun e => smoothPointHeight p q r e = H
/-! The first exact `{2,3,5}` kernel values. -/
/-! ## Exact short-shell multiplicity bounds -/
/-- Exponent triples in a half-open multiplicative shell, with explicit
coordinate bounds. The coordinate bounds are the interface to the pure-power
jump heights; the interval filter is the actual smooth-number shell. -/
def smoothExponentShell
(p q r lo hi hp hq hr : ℕ) : Finset (ℕ × ℕ × ℕ) :=
((Finset.range (hp + 1)).product
((Finset.range (hq + 1)).product (Finset.range (hr + 1)))).filter
fun e => lo ≤ smooth3Val p q r e.1 e.2.1 e.2.2 ∧
smooth3Val p q r e.1 e.2.1 e.2.2 < hi
/-! ## Exact dyadic block geometry for `{2,3,5}` -/
/-- A positive `p`-power lies strictly inside the dyadic block
`(2^a, 2^(a+1))`. Endpoints are excluded because the dyadic jump itself is
the distinguished terminal factor of the compressed block. -/
def DyadicInternalPower (p a e : ℕ) : Prop :=
2 ^ a < p ^ e ∧ p ^ e < 2 ^ (a + 1)
/-- The exact radix of the dyadic block after compressing all internal
`3`- and `5`-power jumps. Each internal channel contributes its prime once,
and the terminal dyadic jump contributes the factor `2`. -/
noncomputable def dyadicBlockBase235 (a : ℕ) : ℕ :=
by
classical
exact
2 *
(if ∃ e, DyadicInternalPower 3 a e then 3 else 1) *
(if ∃ e, DyadicInternalPower 5 a e then 5 else 1)
/-! ## Exact rank-two certificate -/
end ErdosProblems.Erdos269
Source
Lean source (Apache-2.0): https://github.com/wcook04/plectis-erdos-lean/blob/cc7e541cf2081c6fef5a5e377d52e365e33b01eb/ErdosProblems/Erdos269/ThreePrimeRunningLcm.lean#L1-L912
Related paper by Will Cook (CC-BY-4.0): https://github.com/wcook04/plectis-erdos/blob/551bae6dc6e732cf85172d66323c8d2bc77ba962/paper/269/erdos-269-three-prime-running-lcm.tex#L1-L260
Paper prior-art bibliography: https://github.com/wcook04/plectis-erdos/blob/551bae6dc6e732cf85172d66323c8d2bc77ba962/paper/269/erdos-269-three-prime-running-lcm.tex#L681-L689