v0.4.0 — from-scratch `ring` + ℚ ordered field + ℝ additive group
v0.4.0 — a from-scratch ring, ℚ as an ordered field, ℝ as an ordered additive group
This release completes the v0.3.0 ring normalizer into a real tactic and uses it to give the
constructive reals their arithmetic — all pure Lean 4 core, no Mathlib, no sorry. RH remains
open; the substrate makes the analytic half statable and checkable, never proven.
Added
ring_uor— a from-scratchringtactic (F1Square/Analysis/RingTac.lean). A genuine Lean
tactic in core metaprogramming (Lean.Elab.Tactic— not Mathlib): it reifies an integer equality
goal into thePExprsyntax, applies the v0.3.0 soundness lemmanf_eq, and discharges the
residualnorm lhs = norm rhsbydecide. Reification is fuel-bounded (nopartial def), and the
tactic only builds annf_eqproof term — so every goal it closes is as axiom-clean asnf_eq.- ℚ as a verified ordered field (
F1Square/Analysis/QOrder.lean): reflexivity/transitivity of≤,
Qeq → Qle, additive monotonicity, the absolute-value triangle inequality,|·|respecting
value-equality, order transport along≈, and the telescoping triangle|(a+b)−(c+d)| ≤ |a−c|+|b−d|— from the core ℤ order/natAbslemmas plusring_uor. - ℝ as an ordered additive group (
F1Square/Analysis/Real.lean): negationRneg(an isometry)
and the reindexed Bishop additionRadd((x⊕y)ₙ = x₍₂ₙ₊₁₎+y₍₂ₙ₊₁₎), each with its
regularity proof — the addition's bound is exactly the2·1/(2k+2) = 1/(k+1)identity, discharged
byring_uor. TheRealstructure now carriesden_pos.
Provenance note (purity)
ring is confirmed absent from Lean 4 core (it compiles with zero imports → "unknown tactic"),
which is why we built ring_uor. push_cast and omega, used for the cast/linear steps, are
core (they compile with zero imports, no Mathlib/Std) — same tier as decide/simp.
Honesty
The mechanized gate (scripts/honesty_audit.sh) is green: every proof-layer theorem is axiom-clean
(#print axioms ⊆ {propext, Classical.choice, Quot.sound}). The crux (Hodge index on 𝕊 = RH) stays
none because it is open, never because it is forbidden.
Next (v0.5.0)
ℝ multiplication, ≈-transitivity (an Archimedean argument), completeness, ℂ = ℝ×ℝ, and the
transcendentals (exp/log/cos via convergent series with rigorous error bounds).