Certified rational bounds for log, arctan and sqrt
DefinitionZeta5_SourceCertificatesanalysisformalizationirrationalitynumber-theoryzeta5-mo271
Certified rational bounds used in the numerical verification of the potential inequality: log_le_artanh (upper bound for log r, r ≥ 1, by a partial sum plus an explicit geometric tail), arctan_ge_partial and arctan_le_partial (two-sided bounds for arctan x, 0 ≤ x < 1, by partial sums of the alternating series), the half-angle reduction arctan x = 2 arctan (x / (1 + √(1+x²))), and the square-root enclosures sqrt_le_of_sq_le, le_sqrt_of_sq_le.
Definition code
import Mathlib
import Definitions.Def_Zeta5_SourceEnergyKit
/-!
# Certified rational bounds for `log`, `arctan`, `sqrt`
Ported from mo271/Zeta5 @ 7fe736760f4b96bfdb4334b68e3b3124ecbe10b0, `Apery/Certificates.lean`.
Adaptations: imports re-pointed at the platform definition modules; definitions already
published on the platform are not redeclared; `Real.arctan_le_self` (absent from the pinned
Mathlib) is replaced by the local `arctan_le_self'` from `Apery/Potential.lean`.
-/
open Finset Filter Topology Set MeasureTheory
namespace Apery
/-- Upper bound for `log r`, `r ≥ 1`: partial sum plus the tail `2 z^(2m+1) / ((2m+1)(1 - z²))`. -/
lemma log_le_artanh {r : ℝ} (hr : 1 ≤ r) (m : ℕ) :
Real.log r ≤ 2 * ∑ k ∈ range m, (1 / (2 * (k : ℝ) + 1)) * ((r - 1) / (r + 1)) ^ (2 * k + 1) +
2 * ((r - 1) / (r + 1)) ^ (2 * m + 1) / ((2 * m + 1) * (1 - ((r - 1) / (r + 1)) ^ 2)) := by
set x : ℝ := (r - 1) / (r + 1) with hx
have hr0 : 0 < r + 1 := by linarith
have hx0 : 0 ≤ x := div_nonneg (by linarith) hr0.le
have hx1 : x < 1 := by rw [hx, div_lt_one hr0]; linarith
have habs : |x| < 1 := by rw [abs_of_nonneg hx0]; exact hx1
have hsum := Real.hasSum_log_sub_log_of_abs_lt_one habs
have hr' : (1 + x) / (1 - x) = r := by rw [hx]; field_simp; ring
have hlog : Real.log (1 + x) - Real.log (1 - x) = Real.log r := by
rw [← Real.log_div (by linarith) (by linarith), hr']
rw [hlog] at hsum
-- split the series at `m`
have hsplit := hsum.summable.sum_add_tsum_nat_add m
rw [hsum.tsum_eq] at hsplit
-- the tail is bounded by a geometric series
have hx2 : x ^ 2 < 1 := by nlinarith
have hgeom : Summable fun k : ℕ => 2 * (1 / (2 * (m : ℝ) + 1)) * x ^ (2 * m + 1) * (x ^ 2) ^ k :=
(summable_geometric_of_lt_one (by positivity) hx2).mul_left _
have htail : ∑' k : ℕ, 2 * (1 / (2 * ((k + m : ℕ) : ℝ) + 1)) * x ^ (2 * (k + m) + 1) ≤
∑' k : ℕ, 2 * (1 / (2 * (m : ℝ) + 1)) * x ^ (2 * m + 1) * (x ^ 2) ^ k := by
refine Summable.tsum_le_tsum (fun k => ?_) ((summable_nat_add_iff m).mpr hsum.summable) hgeom
have hk : (1 / (2 * ((k + m : ℕ) : ℝ) + 1)) ≤ 1 / (2 * (m : ℝ) + 1) := by
gcongr; push_cast; linarith [(Nat.cast_nonneg k : (0:ℝ) ≤ k)]
have hp : x ^ (2 * (k + m) + 1) = x ^ (2 * m + 1) * (x ^ 2) ^ k := by
rw [← pow_mul, ← pow_add]; congr 1; ring
rw [hp]
have : 0 ≤ x ^ (2 * m + 1) * (x ^ 2) ^ k := by positivity
calc 2 * (1 / (2 * ((k + m : ℕ) : ℝ) + 1)) * (x ^ (2 * m + 1) * (x ^ 2) ^ k)
≤ 2 * (1 / (2 * (m : ℝ) + 1)) * (x ^ (2 * m + 1) * (x ^ 2) ^ k) := by gcongr
_ = _ := by ring
rw [tsum_mul_left, tsum_geometric_of_lt_one (by positivity) hx2] at htail
-- assemble
have hsum_eq : ∑ k ∈ range m, 2 * (1 / (2 * (k : ℝ) + 1)) * x ^ (2 * k + 1) =
2 * ∑ k ∈ range m, (1 / (2 * (k : ℝ) + 1)) * x ^ (2 * k + 1) := by
rw [Finset.mul_sum]; refine Finset.sum_congr rfl fun k _ => by ring
rw [hsum_eq] at hsplit
have h2 : 2 * (1 / (2 * (m : ℝ) + 1)) * x ^ (2 * m + 1) * (1 - x ^ 2)⁻¹ =
2 * x ^ (2 * m + 1) / ((2 * m + 1) * (1 - x ^ 2)) := by
have : (1 : ℝ) - x ^ 2 ≠ 0 := by linarith
field_simp
rw [h2] at htail
linarith
/-- Lower bound `arctan x ≥ ∑_{i < 2k} (-1)^i x^(2i+1)/(2i+1)` for `0 ≤ x < 1`. -/
lemma arctan_ge_partial {x : ℝ} (hx0 : 0 ≤ x) (hx1 : x < 1) (k : ℕ) :
∑ i ∈ range (2 * k), (-1 : ℝ) ^ i * (x ^ (2 * i + 1) / (2 * (i : ℝ) + 1)) ≤ Real.arctan x := by
have hs := Real.hasSum_arctan (x := x) (by rw [Real.norm_eq_abs, abs_of_nonneg hx0]; exact hx1)
have hanti : Antitone fun i : ℕ => x ^ (2 * i + 1) / (2 * (i : ℝ) + 1) := by
intro i j hij
have h1 : x ^ (2 * j + 1) ≤ x ^ (2 * i + 1) := pow_le_pow_of_le_one hx0 hx1.le (by omega)
have h2 : (2 * (i : ℝ) + 1) ≤ 2 * (j : ℝ) + 1 := by
have : (i : ℝ) ≤ j := by exact_mod_cast hij
linarith
exact div_le_div₀ (by positivity) h1 (by positivity) h2
have htend := hs.tendsto_sum_nat
have := Antitone.alternating_series_le_tendsto (f := fun i : ℕ => x ^ (2 * i + 1) / (2 * (i : ℝ) + 1))
(by
refine htend.congr fun n => ?_
refine Finset.sum_congr rfl fun i _ => ?_
push_cast; ring) hanti k
exact this
/-- Upper bound `arctan x ≤ ∑_{i < 2k+1} (-1)^i x^(2i+1)/(2i+1)` for `0 ≤ x < 1`. -/
lemma arctan_le_partial {x : ℝ} (hx0 : 0 ≤ x) (hx1 : x < 1) (k : ℕ) :
Real.arctan x ≤ ∑ i ∈ range (2 * k + 1), (-1 : ℝ) ^ i * (x ^ (2 * i + 1) / (2 * (i : ℝ) + 1)) := by
have hs := Real.hasSum_arctan (x := x) (by rw [Real.norm_eq_abs, abs_of_nonneg hx0]; exact hx1)
have hanti : Antitone fun i : ℕ => x ^ (2 * i + 1) / (2 * (i : ℝ) + 1) := by
intro i j hij
have h1 : x ^ (2 * j + 1) ≤ x ^ (2 * i + 1) := pow_le_pow_of_le_one hx0 hx1.le (by omega)
have h2 : (2 * (i : ℝ) + 1) ≤ 2 * (j : ℝ) + 1 := by
have : (i : ℝ) ≤ j := by exact_mod_cast hij
linarith
exact div_le_div₀ (by positivity) h1 (by positivity) h2
have htend := hs.tendsto_sum_nat
exact Antitone.tendsto_le_alternating_series (f := fun i : ℕ => x ^ (2 * i + 1) / (2 * (i : ℝ) + 1))
(by
refine htend.congr fun n => ?_
refine Finset.sum_congr rfl fun i _ => ?_
push_cast; ring) hanti k
/-- The half-angle reduction `arctan x = 2 arctan (x / (1 + √(1 + x²)))` for `x ≥ 0`. -/
lemma arctan_eq_two_mul_arctan {x : ℝ} (hx : 0 ≤ x) :
Real.arctan x = 2 * Real.arctan (x / (1 + Real.sqrt (1 + x ^ 2))) := by
set s := Real.sqrt (1 + x ^ 2) with hs
have hs2 : s ^ 2 = 1 + x ^ 2 := Real.sq_sqrt (by positivity)
have hs1 : 1 ≤ s := by
rw [hs]; exact Real.le_sqrt_of_sq_le (by nlinarith : (1 : ℝ) ^ 2 ≤ 1 + x ^ 2)
set y := x / (1 + s) with hy
have hy0 : 0 ≤ y := by positivity
have hy1 : y < 1 := by
rw [hy, div_lt_one (by positivity)]
nlinarith [Real.sqrt_nonneg (1 + x ^ 2)]
have hyy : y * y < 1 := by nlinarith
have hadd := Real.arctan_add (x := y) (y := y) (by nlinarith)
have hval : (y + y) / (1 - y * y) = x := by
rw [div_eq_iff (by nlinarith : (1 : ℝ) - y * y ≠ 0), hy]
have h1s : (1 + s) ≠ 0 := by positivity
field_simp
nlinarith [hs2, hs1]
rw [hval] at hadd
linarith
/-- Rational enclosure of a square root: `√q ≤ u` if `q ≤ u²`, `0 ≤ u`. -/
lemma sqrt_le_of_sq_le {q u : ℝ} (hu : 0 ≤ u) (h : q ≤ u ^ 2) : Real.sqrt q ≤ u :=
Real.sqrt_le_iff.mpr ⟨hu, h⟩
/-- Rational enclosure of a square root: `l ≤ √q` if `l² ≤ q`, `0 ≤ l`. -/
lemma le_sqrt_of_sq_le {q l : ℝ} (hl : 0 ≤ l) (h : l ^ 2 ≤ q) : l ≤ Real.sqrt q :=
Real.le_sqrt_of_sq_le h
end Apery
Source
mo271/Zeta5 @ 7fe736760f4b96bfdb4334b68e3b3124ecbe10b0, `Apery/Certificates.lean` (verbatim, imports adapted).