Three modified quotient limits with cleared Omega factors
OpenRamanujanLostNotebook.PartI.entry_8_2_2Here (a;q)n=product(k=0..n-1)(1-aq^k), (a;q)infinity=product(k>=0)(1-aq^k), and [t,k]q=(q;q)t/((q;q)k*(q;q)(t-k)) for integers 0<=k<=t, zero otherwise. The continuants have P_0=0,P_1=1,Q_0=Q_1=1, and T_N=(1+q^(N-1))*T(N-1)-T(N-2) for N>=2, for T=P,Q at a=0; at N>=2 the modified T_N(a) is T_N(0)+aT_(N-1)(0). For |q|<1, primitive w, arbitrary complex a and epsilon in {-1,0,1}, put A=(w^2q;q)infinity, B=(wq;q)infinity and F_j=(1-aw^2)A-(1-aw)w^j B. If F(epsilon-1) is nonzero, Q(3v+epsilon+1)(a) is eventually nonzero and P/Q converges to -w^2F_(epsilon+1)/F_(epsilon-1)(q^2;q^3)_infinity/(q;q^3)_infinity. Differs from the printed source: following the editors we replace the n-dependent informal limit by explicit residue-class subsequences. We clear the common (1-aw)(wq;q)infinity factors in Omega, thereby including a=w^2, and require F(epsilon-1) to be nonzero for an ordinary complex quotient. The clearing and regularity condition are ours; poles and general convergence on the sphere are not assigned a totalized complex value. The independent pre-compile audit confirms the cleared formula, including a=w^2: maximum continuant residual 6.60e-90 at 90-digit precision, numerical evidence, not a proof. Differs from the printed source: the print fixes w=e^{2πi/3}; we state both primitive cube roots, the second following by complex conjugation.
Discrepancy from the printed source. Differs from the printed source: following the editors we replace the n-dependent informal limit by explicit residue-class subsequences. We clear the common (1-aw)*(wq;q)infinity factors in Omega, thereby including a=w^2, and require F(epsilon-1) to be nonzero for an ordinary complex quotient. The clearing and regularity condition are ours; poles and general convergence on the sphere are not assigned a totalized complex value. The independent pre-compile audit confirms the cleared formula, including a=w^2: maximum continuant residual 6.60e-90 at 90-digit precision, numerical evidence, not a proof. Differs from the printed source: the print fixes w=e^{2πi/3}; we state both primitive cube roots, the second following by complex conjugation.
import Mathlib import Definitions.Def_RamanujanLostNotebook_PartI_ch08_ch08P import Definitions.Def_RamanujanLostNotebook_PartI_ch08_ch08PQ0 import Definitions.Def_RamanujanLostNotebook_PartI_ch08_ch08Q import Definitions.Def_RamanujanLostNotebook_PartI_shared_qPochInf
namespace RamanujanLostNotebook.PartI
theorem entry_8_2_2 (q w : ℂ) (hq : ‖q‖ < 1) (hw : w ^ 3 = 1) (hw1 : w ≠ 1) (a : ℂ) (epsilon : ℤ) (he : epsilon = -1 ∨ epsilon = 0 ∨ epsilon = 1) (hregular : ((1 - a * w ^ 2) * qPochInf (w ^ 2 * q) q - (1 - a * w) * w ^ (epsilon - 1) * qPochInf (w * q) q) ≠ 0) :
(∀ᶠ v : ℕ in Filter.atTop, ch08Q q a (3 * (v : ℤ) + epsilon + 1).toNat ≠ 0) ∧
Filter.Tendsto (fun v : ℕ => ch08P q a (3 * (v : ℤ) + epsilon + 1).toNat / ch08Q q a (3 * (v : ℤ) + epsilon + 1).toNat)
Filter.atTop (nhds (-w ^ 2 * ((1 - a * w ^ 2) * qPochInf (w ^ 2 * q) q - (1 - a * w) * w ^ (epsilon + 1) * qPochInf (w * q) q) / ((1 - a * w ^ 2) * qPochInf (w ^ 2 * q) q - (1 - a * w) * w ^ (epsilon - 1) * qPochInf (w * q) q) * qPochInf (q ^ 2) (q ^ 3) / qPochInf q (q ^ 3))) := by sorry
end RamanujanLostNotebook.PartI