Chapter 11 definitions: Rogers--Ramanujan convergents in scalar and power-series form
DefinitionQseriesFormalization_Pending_Chapter11_RRCF_Convergentchan-invitation-qseriescontinued-fractionsrogers-ramanujan
This bundle defines the numerator and denominator sequences of the Rogers--Ramanujan continued-fraction convergents, with , , and , , both as complex functions of and as power series over . It defines the convergent series and and the coefficientwise limit with .
Definition code
-- Generated by qseries-p2m/p2m/generate.py from QseriesFormalization.Pending.Chapter11_RRCF_Convergent: skeleton subtraction
-- keeps 15 declaration(s); proofs kept; project imports -> Definitions bundles.
import Mathlib.Analysis.SpecialFunctions.Pow.Real
import Mathlib.Data.Complex.Basic
import Mathlib
import Mathlib.Topology.Algebra.InfiniteSum.NatInt
import Mathlib.Analysis.SpecificLimits.Normed
import Mathlib.Analysis.Normed.Ring.InfiniteSum
import Mathlib.Analysis.SpecialFunctions.Log.Summable
import Mathlib.Analysis.Normed.Group.Tannery
import Mathlib.Combinatorics.Enumerative.Partition.GenFun
import Mathlib.RingTheory.PowerSeries.Basic
import Mathlib.RingTheory.PowerSeries.Expand
import Mathlib.RingTheory.PowerSeries.Inverse
import Mathlib.Combinatorics.Enumerative.Partition.Basic
import Mathlib.Tactic.Linarith
/-!
# Chapter 11 — Rogers-Ramanujan continued fraction: correct convergents
⚠️ **AUX-LEVEL SCAFFOLDING** for Ch11 (currently AUX).
`Chapter11.R_trunc` builds the RRCF "outside-in" with `q^k` on the *outer*
factor at step `k`, which makes `R_trunc q n → 1` trivially as `n → ∞`
(since `q^n → 0` for `|q| < 1`). That truncation is NOT a convergent of
the true RRCF.
This file defines the **standard convergents** of
`R(q) = q^{1/5} · 1 / (1 + q / (1 + q² / (1 + q³ / (⋯))))`
via the **backward recurrence** for continued-fraction convergents
`p_n = a_n·p_{n-1} + b_n·p_{n-2}` with `a_n = 1`, `b_n = q^n`, and the
"trivial 0-th convergent = 1" initial condition `(p_{-1}=1, p_0=1)`, giving
`p_0 = 1, p_1 = 1 + q, p_{n+1} = p_n + q^{n+1}·p_{n-1}`
`q_0 = 1, q_1 = 1, q_{n+1} = q_n + q^{n+1}·q_{n-1}`
The `n`-th convergent (stripped of the `q^{1/5}` prefactor) is
`T_n(q) := p_n(q) / q_n(q)`; numerically `T_n → r(q) = (q;q^5)(q^4;q^5)/((q^2;q^5)(q^3;q^5))`
(Chan Theorem 11.1).
## What this file contributes (corrected 2026-05-27)
Note: yesterday's commit of this file had a base-case bug (`rrcf_A 1 := 1`
instead of `1+q`, `rrcf_B 0 := 0` instead of `1`) that made the recurrence
compute a different CF. Today's version is the corrected one matching
Chan §11 / Theorem 11.1.
- `rrcf_A`, `rrcf_B` : ℕ → ℂ → ℂ definitions, scalar form, for the
numerator / denominator convergent sequences.
- `rrcf_APS`, `rrcf_BPS` : ℕ → ℚ⟦X⟧ formal-PS lift.
- `rrcf_TPS n := rrcf_APS (n+1) * (rrcf_BPS (n+1))⁻¹` formal-PS convergent.
- Small explicit values + base-case lemmas.
The chapter-main result (X-adic convergence of `rrcf_TPS n` to the formal
power series `r(q)`, equivalent to Chan Theorem 11.1) remains open; this
file provides the correct combinatorial scaffolding.
-/
namespace QseriesFormalization
namespace Pending
namespace Ch11RRCFConvergent
/-- Numerator sequence of the RRCF convergents (scalar):
`A_0 = 1, A_1 = 1+q, A_{n+2} = A_{n+1} + q^{n+2}·A_n`. -/
noncomputable def rrcf_A : ℕ → ℂ → ℂ
| 0, _ => 1
| 1, q => 1 + q
| n + 2, q => rrcf_A (n + 1) q + q ^ (n + 2) * rrcf_A n q
/-- Denominator sequence of the RRCF convergents (scalar):
`B_0 = 1, B_1 = 1, B_{n+2} = B_{n+1} + q^{n+2}·B_n`. -/
noncomputable def rrcf_B : ℕ → ℂ → ℂ
| 0, _ => 1
| 1, _ => 1
| n + 2, q => rrcf_B (n + 1) q + q ^ (n + 2) * rrcf_B n q
@[simp] theorem rrcf_A_zero (q : ℂ) : rrcf_A 0 q = 1 := rfl
@[simp] theorem rrcf_A_one (q : ℂ) : rrcf_A 1 q = 1 + q := rfl
@[simp] theorem rrcf_B_zero (q : ℂ) : rrcf_B 0 q = 1 := rfl
@[simp] theorem rrcf_B_one (q : ℂ) : rrcf_B 1 q = 1 := rfl
/-! ## Formal-power-series version of the RRCF convergents
To talk about convergence `T_n → r(q)` as `n → ∞`, we lift the recurrence
from `ℂ → ℂ` to `ℚ⟦X⟧` (where the `X`-adic topology makes "stabilization
at fixed degree" precise). Each `B_n` has constant coefficient `1`, hence
is a unit, hence `T_n := A_n · B_n⁻¹` is a genuine formal power series.
-/
open PowerSeries
/-- Numerator sequence of RRCF convergents, as formal power series over `ℚ`. -/
noncomputable def rrcf_APS : ℕ → ℚ⟦X⟧
| 0 => 1
| 1 => 1 + X
| n + 2 => rrcf_APS (n + 1) + X ^ (n + 2) * rrcf_APS n
/-- Denominator sequence of RRCF convergents, as formal power series over `ℚ`. -/
noncomputable def rrcf_BPS : ℕ → ℚ⟦X⟧
| 0 => 1
| 1 => 1
| n + 2 => rrcf_BPS (n + 1) + X ^ (n + 2) * rrcf_BPS n
@[simp] theorem rrcf_APS_zero : rrcf_APS 0 = 1 := rfl
@[simp] theorem rrcf_APS_one : rrcf_APS 1 = 1 + X := rfl
@[simp] theorem rrcf_BPS_zero : rrcf_BPS 0 = 1 := rfl
@[simp] theorem rrcf_BPS_one : rrcf_BPS 1 = 1 := rfl
/-- The `n`-th formal-PS convergent of the reciprocal RRCF `1/r(q) =
1 + q/(1+q²/(1+q³/⋯))`. This converges to `1/r(q)`, NOT directly to `r(q)`:
`T_n(X) := A_n(X) · B_n(X)⁻¹` in `ℚ⟦X⟧`.
For the convergent of `r(q)` itself, use `rrcf_RPS` below. -/
noncomputable def rrcf_TPS (n : ℕ) : ℚ⟦X⟧ :=
rrcf_APS n * (rrcf_BPS n)⁻¹
/-! ## Convergence-toward-`r(q)` infrastructure: the CF cross-determinant
The standard structural identity for CF convergents:
`A_{n+1}·B_n − A_n·B_{n+1} = (−1)^n · X^{(n+1)(n+2)/2}`
drives the convergence of `T_n = A_n/B_n` in the X-adic topology of `ℚ⟦X⟧`,
because the difference `T_{n+1} − T_n = (−1)^n · X^{(n+1)(n+2)/2}/(B_n·B_{n+1})`
has X-adic order ≥ `(n+1)(n+2)/2 → ∞`, hence the sequence is Cauchy and
converges to a limit (`r(q)`, Chan Theorem 11.1).
We prove the cross-determinant identity by induction. -/
/-- The `n`-th formal-PS convergent of the RRCF `r(q) = 1/(1+q/(1+q²/⋯))`:
`R_n(X) := B_n(X) · A_n(X)⁻¹` in `ℚ⟦X⟧`.
(Indices swapped from `T_n` because `r(q) = 1/T(q)` and the convergent of a
reciprocal is the reciprocal of the convergent.) This converges to
`rrcf_r := pentagonal014SeriesPS · pentagonal023SeriesPS⁻¹` (Chan Theorem
11.1, the formal-PS version with `q^{1/5}` prefactor stripped). -/
noncomputable def rrcf_RPS (n : ℕ) : ℚ⟦X⟧ :=
rrcf_BPS n * (rrcf_APS n)⁻¹
/-! ## Coefficient stabilization (X-adic Cauchy form)
The X^N divisibility theorems give a direct, no-topology form of the
Cauchy property: for fixed coefficient degree `k`, the sequence
`(rrcf_TPS n).coeff k` is **eventually constant** as `n → ∞`.
Specifically, for `n ≥ k`, `(rrcf_TPS (n+1)).coeff k = (rrcf_TPS n).coeff k`.
This is because `X^{(n+1)(n+2)/2} ∣ (T_{n+1}−T_n)` forces the `k`-th
coefficient to vanish whenever `k < (n+1)(n+2)/2`, which holds for `n ≥ k`
(since `(n+1)(n+2)/2 ≥ n+1 > k`).
-/
/-- **Definition of `r(q)` via the CF limit**: the formal power series
whose `k`-th coefficient is `(rrcf_RPS k).coeff k`. Well-defined: any
convergent index `n ≥ k` gives the same `k`-th coefficient by
`rrcf_RPS_coeff_eq_of_le`. -/
noncomputable def rrcf_r_via_CF : ℚ⟦X⟧ :=
PowerSeries.mk (fun k => (rrcf_RPS k).coeff k)
/-! ## Match with product-form `rrcf_r` at small coefficient orders
These verify Chan §11 Theorem 11.1 unconditionally at small coefficient
indices: the CF-limit form `rrcf_r_via_CF` and the product form
`rrcf_r := pentagonal014SeriesPS · pentagonal023SeriesPS⁻¹` (from
`Ch13_RRCF_RForm`) agree at degrees 0 and 1.
-/
/-! ## Key identity: `rrcf_RPS n * rrcf_APS n = rrcf_BPS n`
This is the multiplicative form of `R_n = B_n · A_n⁻¹`: clearing the
denominator gives `R_n · A_n = B_n`, which makes coefficient extraction
a finite antidiagonal sum over known polynomial coefficients. -/
/-! ## Higher-degree coefficient computations via the product identity
For degree `k`, `rrcf_r_via_CF.coeff k = (rrcf_RPS k).coeff k`. We compute
`(rrcf_RPS k).coeff k` by extracting degree `k` from `R_k · A_k = B_k`:
`∑_{(i,j) ∈ antidiagonal k} R_k.coeff i · A_k.coeff j = B_k.coeff k`
Since `A_k.coeff 0 = 1`, this gives
`R_k.coeff k = B_k.coeff k − ∑_{i<k} R_k.coeff i · A_k.coeff (k−i)`,
a finite computation using the already-known lower coefficients. -/
/-!
### Recurrence-backed coefficient calculators for larger CF checks
The next coefficient verifications continue to use `R_k * A_k = B_k`.
These private calculators package the same recurrence-peeling used above:
`A_{n+2}=A_{n+1}+X^{n+2} A_n` and `B_{n+2}=B_{n+1}+X^{n+2} B_n`,
with `PowerSeries.coeff_X_pow_mul'` handling the shifted term.
-/
end Ch11RRCFConvergent
end Pending
end QseriesFormalization
Source
Lean formalization (Xiang Huang) of Hei-Chi Chan, An Invitation to q-Series: From Jacobi's Triple Product Identity to Ramanujan's "Most Beautiful Identity" (World Scientific, 2011), Chapter 11. Lean source: https://github.com/xiangyazi24/invitation-to-qseries-lean/blob/5df34b280b6a9c4a68ab09105b4e5d3c7bbe88d1/QseriesFormalization/Pending/Chapter11_RRCF_Convergent.lean (definitions of this module that the uploaded theorems need; ported to Lean v4.33.1 / Mathlib 0df444a).