Q1 main-term absorption #
Scope #
This file proves the q¹ main-term bound q1MainTermAbsorption
from SwitchingPrinciple.lean:
∃ C₁ > 0, ∃ N₁, ∀ N ≥ N₁ Even N:
q1MainTermSum N ≤ C₁·𝔖_trunc(N, z−1)·N/log²N
Here z = correctedChenZ N, y = correctedChenY N, P(N) = correctedChenSiftingProduct N (odd primes r < z with r ∤ N), F(N) = correctedChenForbiddenProduct N (primes r < z with r ≤ 2 or r | N), and Fodd(N) = F(N)/2 (the product of the odd prime factors of F when z ≥ 3).
Mathematical structure (sieve main terms in Halberstam--Richert Ch. 10; Chen 1973) #
For a fixed prime q ∈ [z, y), d | P, and e | F:
q1APMainValue N (lcm(lcm q d) e) = if Even (lcm(lcm q d) e) then [lcm(lcm q d) e | N−2] else li(N)/φ(lcm(lcm q d) e).
The numbers q, d, e are pairwise coprime: q ≥ z, all prime factors of P·F are < z, and P and F have disjoint prime-factor sets. In the odd part, q and d are odd, and e is odd exactly when e | Fodd. Thus lcm(lcm q d) e = q·d·e, and
q1CandidateAPMain N q = li(N)·Σ_{d|P} Σ_{e|Fodd} μ(d)μ(e)/φ(q·d·e) + (even-modulus correction, ≤ 0).
The odd part factors through the Möbius Euler-product identity
(Mathlib's prodPrimeFactors_one_sub_of_squarefree):
Σ_{d|P} μ(d)/φ(d) = ∏{p|P} (1 − 1/(p−1)) =: q1SieveProduct N, Σ{e|Fodd} μ(e)/φ(e) = ∏_{p|Fodd} (1 − 1/(p−1)) =: q1ForbiddenOddProduct N,
The even part is exactly zero or negative, by Möbius inversion Σ_{d|m} μ(d) = [m = 1]:
EvenPart = −[q | N−2]·[gcd(P, N−2) = 1]·[gcd(Fodd, (N−2)/2) = 1] ≤ 0.
Consequently, for even N (where p ∤ N automatically excludes p = 2),
q1MainTermSum N ≤ li(N)·goldbachSieveProduct(N, z)·Σ_{q∈[z,y)} 1/φ(q).
Next:
- goldbachSieveProduct(N, z) = primeProduct(z−1)·𝔖_trunc(N, z−1) (MertensTheorem.sieveProduct_identity);
- primeProduct(z−1) ≤ a₂/log(z−1), log(z−1) ≥ (1/20)·log N (Mertens' theorem and an elementary estimate);
- Σ_{q∈[z,y)} 1/φ(q) ≤ 2·Σ 1/q ≤ 2·(primeReciprocalSum y − primeReciprocalSum (z−1))
≤ 2·(log(log y/log z) + E) ≤ 2·(log 7 + E)
(
AnalyticNumberTheory.Sieve.primeReciprocalSum_range_leand log y/log z ≤ 7 fromhTripleMain_log_y_div_log_z_leinTripleMain.lean).
Combining these estimates gives q1MainTermAbsorption,
with C₁ = 20·a₂·(2·(log 7 + E)).
The analytic ingredients are the proved prime-reciprocal range bound and
the goldbachNu / singularSeriesTruncated definitions in
AnalyticNumberTheory.Sieve, the theorems
MertensTheorem.primeProduct_asymptotic_order and sieveProduct_identity,
and the logarithmic parameter bounds from TripleMain.lean.
Notation and definitions #
The logarithmic-integral main term for q¹: li(N) = N/log N,
using the working definition in AnalyticNumberTheory.Sieve.
Equations
Instances For
Real-valued Möbius weights.
Equations
Instances For
The odd part of F(N): F(N)/2, exactly the product of its odd prime factors when 3 ≤ z.
Equations
Instances For
Main-term product over sifting primes: ∏_{p | P(N)} (1 − 1/(p−1)).
Equations
Instances For
Main-term product over the odd forbidden prime factors: ∏_{p | Fodd(N)} (1 − 1/(p−1)).
Equations
Instances For
Even-modulus contribution to the q¹ main term: Σ_{d|P} Σ_{e|F} μ(d)μ(e)·q1APMainEvenValue(lcm(lcm q d) e).
Equations
- MathlibNt.SieveTheory.SwitchingPrinciple.q1CandidateAPMainEvenPart N q = ∑ d ∈ (MathlibNt.SieveTheory.SwitchingPrinciple.correctedChenSiftingProduct N).divisors, MathlibNt.SieveTheory.SwitchingPrinciple.q1Mu d * ∑ e ∈ (MathlibNt.SieveTheory.SwitchingPrinciple.correctedChenForbiddenProduct N).divisors, MathlibNt.SieveTheory.SwitchingPrinciple.q1Mu e * MathlibNt.SieveTheory.SwitchingPrinciple.q1APMainEvenValue N ((q.lcm d).lcm e)
Instances For
1. Möbius sums: Σ_{d|n} μ(d) = [n = 1], Σ_{d|n, d|m} μ(d) = [gcd(m,n) = 1] #
2. Parity and coprimality lemmas #
If 3 ≤ z, then 2 | F(N): 2 is a prime factor of F.
F(N) = 2·Fodd(N) when 2 | F.
The prime q is coprime to P: q ≥ z and all prime factors of P are < z.
The prime q is coprime to Fodd: q ≥ z and its prime factors are < z.
P and Fodd are coprime because their prime-factor sets are disjoint: p ∤ N for p | P, whereas p | N for p | Fodd.
The prime q is coprime to the forbidden-prime product F: all prime factors of F are < z ≤ q.
The sifting product P and the forbidden-prime product F are coprime: their prime-factor sets are disjoint.
3. Odd-even decomposition of the q¹ main term #
Exact odd-even decomposition: q1CandidateAPMain N q = li(N)·Σ_{d|P}Σ_{e|Fodd} μ(d)μ(e)/φ(q·d·e) + q1CandidateAPMainEvenPart N q.
4. Exact evaluation of the even-modulus contribution (Möbius inversion) #
Exact even-modulus contribution:
−[q | N−2]·[gcd(P,N−2)=1]·[gcd(Fodd,(N−2)/2)=1].
The even-modulus contribution is nonpositive: its exact value is the negative of a product of three nonnegative indicator factors.
5. Möbius Euler-product factorization: Σ_{d|P} μ(d)/φ(d) = ∏_{p|P} (1 − 1/(p−1)) #
Möbius Euler product for P: Σ_{d | P} μ(d)/φ(d) = q1SieveProduct N.
Möbius Euler product for Fodd:
Σ_{e | Fodd} μ(e)/φ(e) = q1ForbiddenOddProduct N.
Factorization of the odd double sum:
Σ_{d|P} Σ_{e|Fodd} μ(d)μ(e)/φ(q·d·e) = (1/φ(q))·q1SieveProduct·q1ForbiddenOddProduct, since q, d, e are pairwise coprime.
6. Bounds for each q and decomposition of the main-term sum #
Bound for each q in the q¹ main term:
q1CandidateAPMain N q ≤ li(N)/φ(q)·q1SieveProduct·q1ForbiddenOddProduct.
The main-term product over sifting primes is nonnegative.
The main-term product over odd forbidden primes is at most 1: each factor lies in (0, 1].
The main-term sifting product equals the Goldbach sieve product. For even N, the condition p ∤ N excludes p = 2 on both sides.
Main-term sum decomposition:
q1MainTermSum N ≤ li(N)·goldbachSieveProduct(N, z)·Σ_{q∈[z,y)} 1/φ(q).
7. Combining analytic estimates: prime reciprocals, log(z−1), and absorption #
Reciprocal-totient sum bound: for N > 2^110,
Σ_{q∈[z,y)} 1/φ(q) ≤ C₀, by Mertens' second theorem and log y/log z ≤ 7.
The truncated singular-series definitions in MathlibNt and
AnalyticNumberTheory agree: their local-factor formulas are identical.
q¹ main-term absorption: the decomposition, Mertens' theorem, and
the prime-reciprocal sum bound combine into the form C₁·𝔖_trunc·N/log²N.
The finite delta-one Pan bridge for the q¹ error #
The switching primes occurring in the q¹ error.
Equations
Instances For
The product containing both small-prime divisor lanes in the q¹ Möbius expansion. Its two factors are coprime and squarefree.
Equations
Instances For
The reduced-residue delta-one Pan error used by the q¹ bridge. A non-reduced residue is set to zero and left to the explicit truncation input.
Equations
- MathlibNt.SieveTheory.SwitchingPrinciple.q1PanReferenceError N m = if (N % m).Coprime m then AnalyticNumberTheory.Sieve.panDistributionError (N - 2) 1 m (N % m) else 0
Instances For
On an even modulus and even source, the Pan reduced-residue reference
vanishes because 2 divides both the residue and the modulus.
The complete diagnostic difference vanishes on every even-modulus lane;
the obstruction in q1PanDeltaOneTruncationSum is therefore entirely in the
odd non-reduced lanes (apart from the logarithmic-integral shift).
Exact obstruction on an odd non-reduced lane: the diagnostic "truncation"
summand is the full AP error, including its li(N) / φ(m) main term.
The exact Möbius triple sum after replacing each q¹ AP error by Pan's delta-one maximum.
Equations
- MathlibNt.SieveTheory.SwitchingPrinciple.q1PanDeltaOneRawSum N = ∑ q ∈ MathlibNt.SieveTheory.SwitchingPrinciple.q1SwitchingPrimes N, ∑ d ∈ (MathlibNt.SieveTheory.SwitchingPrinciple.correctedChenSiftingProduct N).divisors, |↑(ArithmeticFunction.moebius d)| * ∑ e ∈ (MathlibNt.SieveTheory.SwitchingPrinciple.correctedChenForbiddenProduct N).divisors, |↑(ArithmeticFunction.moebius e)| * AnalyticNumberTheory.Sieve.panMaxY N ((q.lcm d).lcm e) ⌊↑N⌋₊ MathlibNt.SieveTheory.SwitchingPrinciple.chenPanWeightOne
Instances For
A diagnostic lcm-fibre majorant for the delta-one Pan sum. After both
original coprime divisor lanes are enlarged to all divisors of the squarefree
small-prime product, 3^ω(m) is the exact number of pairs in that enlarged
square whose lcm is m; it is not the multiplicity of the original lanes.
Equations
- MathlibNt.SieveTheory.SwitchingPrinciple.q1PanDeltaOneWeightedSum N = ∑ q ∈ MathlibNt.SieveTheory.SwitchingPrinciple.q1SwitchingPrimes N, ∑ m ∈ (MathlibNt.SieveTheory.SwitchingPrinciple.q1SmallModulusProduct N).divisors, 3 ^ m.primeFactors.card * AnalyticNumberTheory.Sieve.panMaxY N (q.lcm m) ⌊↑N⌋₊ MathlibNt.SieveTheory.SwitchingPrinciple.chenPanWeightOne
Instances For
Everything not supplied by the reduced-residue delta-one Pan maximum:
the li(N) versus li(N-2) main shift, the non-reduced residue lane, and the
even-modulus normalization (which is actually zero for large even N).
Equations
- MathlibNt.SieveTheory.SwitchingPrinciple.q1PanDeltaOneTruncationSum N = ∑ q ∈ MathlibNt.SieveTheory.SwitchingPrinciple.q1SwitchingPrimes N, ∑ d ∈ (MathlibNt.SieveTheory.SwitchingPrinciple.correctedChenSiftingProduct N).divisors, |↑(ArithmeticFunction.moebius d)| * ∑ e ∈ (MathlibNt.SieveTheory.SwitchingPrinciple.correctedChenForbiddenProduct N).divisors, |↑(ArithmeticFunction.moebius e)| * |MathlibNt.SieveTheory.SwitchingPrinciple.q1APError N ((q.lcm d).lcm e) - MathlibNt.SieveTheory.SwitchingPrinciple.q1PanReferenceError N ((q.lcm d).lcm e)|
Instances For
Diagnostic full-divisor-lane delta-one estimate. This is strictly stronger
than the classical Pan mean-value theorem: m runs through every divisor of the
small-prime product and lcm q m has no
N^(1/2) / log(N)^B cutoff. It is retained only to audit the finite lcm
reindexing, not as an active Chen analytic input.
Equations
Instances For
Diagnostic complement to Q1DeltaOnePanMeanValue. Besides the harmless
li(N) versus li(N-2) shift, this sum contains every odd non-reduced residue
lane. On such a lane q1PanReferenceError is zero while q1APError still
contains its li(N) / φ(m) main term, with the full divisor-pair multiplicity.
Consequently this is not a proved truncation estimate and is not a canonical
literature hypothesis.
Equations
Instances For
Diagnostic form of the former q¹ frontier: arbitrary inverse-log saving for
the complete positive Möbius error aggregate. This is not a published
Pan/Bombieri--Vinogradov theorem, because its full divisor lanes have no level
cutoff. It is retained only to record the old implication to
q1APErrorUniformBound.
Equations
Instances For
A q¹ AP error is bounded by the reduced-residue Pan maximum plus the explicit exceptional difference.
Exact finite reduction of the q¹ error to the raw delta-one Pan sum and the explicit exceptional/truncation sum.
The union of the two small-prime products is squarefree.
The union of the two small-prime products is nonzero.
The raw q¹ Möbius sum is bounded by its canonical 3^ω lcm-fibre
packaging.
Fully finite q¹ error bridge: all Möbius multiplicities and lcm fibres are
absorbed into the 3^ω delta-one Pan sum; only the explicit exceptional
truncation remains.
The fixed-delta-one mean-value theorem and its explicit truncation contract produce the q¹ AP error input. This is a valid implication between diagnostic full-lane predicates, but it is not the canonical literature route.
The obsolete full-positive-error aggregate supplies the single A = 3
estimate used by the legacy diagnostic consumer.
The unconditional q¹ main-term absorption and the diagnostic full-lane Pan bridge give a q¹ count bound.
A designated q¹ constant extracted from the delta-one Pan bridge. This lets downstream numerical hypotheses refer to the same witness that supplies the q¹ estimate, rather than quantifying over all admissible (and arbitrarily enlargeable) constants.
Equations
Instances For
Positivity and the eventual q¹ estimate for the designated bridge constant.