diff 626fd2d..b1f9bc4 weighted_dilation_totient_fibers_v2.txt
diff --git a/weighted_dilation_totient_fibers_v2.txt b/weighted_dilation_totient_fibers_v2.txt
index 58b6ba6..c149249 100644
--- a/weighted_dilation_totient_fibers_v2.txt
+++ b/weighted_dilation_totient_fibers_v2.txt
@@ -88,43 +88,6 @@ Theorem B is much stronger than the classical statement "for every eps >
that in the single short interval (2x,5x] the smooth predecessors are as
numerous as x^(1-o(1)), for every fixed smoothness exponent.
-WHY THE TWO THEOREMS ARE ONE PROBLEM. Section 2 derives Theorem A from
-Theorem B in about forty lines, using only unique factorization and a
-pigeonhole. Sections 3 to 11 prove Theorem B.
-
-WHAT WAS KNOWN, WITH THE EXACT NUMBERS.
-
- Erdos (1935, 1956) raised both questions. Pomerance (1980) states the
- totient conjecture as C = 1, where C is the supremum of the c with
- g(n) > n^c infinitely often, attributes it to Erdos, and records the
- transfer from smooth shifted primes reproduced in section 2.
-
- Baker and Harman (1998) obtained P^+(p-1) <= x^(0.2961) for x < p <= 2x,
- with totient multiplicity exponent 0.7039. This improved Friedlander's
- 1/(2 sqrt e) + eps = 0.30326..., and their introduction surveys earlier
- work of Pomerance, Balog, and Fouvry-Grupp.
-
- Lichtman (2022) obtained at least x/(log x)^C primes p in (x,2x] with
- P^+(p-1) <= x^beta for every beta > 15/(32 sqrt e) = 0.284311..., with
- multiplicity exponent 0.7156.
-
- The Dickman function predicts a positive proportion of primes, not just
- x^(1-o(1)), for every fixed beta > 0 (Granville 2008, section 5.3).
-
- Bharadwaj and Rodgers (2026) prove a full Poisson-Dirichlet law for
- sequences with their regularity and congruence-uniformity conditions
- plus level-one distribution; for shifted primes that gives the
- prediction under Elliott-Halberstam, and their Proposition 2 verifies
- level one half unconditionally.
-
- Alford, Granville and Pomerance (1994) use smooth shifted primes to
- build Carmichael numbers, which is one reason the question is not only
- about phi.
-
-Theorem B is unconditional, holds for every fixed delta, and gets the full
-power of x. It is still weaker than a positive proportion.
-
-
2. THE EASY HALF: SMOOTH PRIMES GIVE LARGE TOTIENT FIBERS
@@ -3998,25 +3961,6 @@ delta in (10.1) leaves room for every subpower P-prime, and the factor 2
is eventually below x^delta. This proves Theorem B.
-11.3 REMARK ON THE PROOF SKELETON.
-
-
-The pieces of the proof combine as follows. Section 4 supplied the
-elementary lemmas; sections 5, 6, 7 proved that the physical dilation
-graph has the same moments as its idealised counterpart; section 8
-extracted a cancellation at any prescribed shift, which section 9
-converted into the Type II bilinear estimate. Sections 10.1-10.9 then
-threaded the sieve: Type I to control the congruence distribution, the
-block sieve to establish the presieve mass, proxies to turn roughness
-tests into ordinary-density tests, Type II to remove unbalanced
-composites, and a two-variable sieve to control the balanced window.
-Section 11.1 fixed the parameters, and section 11.2 converted the
-surviving prime mass into a count. Nothing in this chain depends on
-Siegel-Walfitz for the ideal norm or the shift cancellation; the only
-ineffective input is the usual one inside Type I and the prime number
-theorem in the block sieve, and both are absolute.
-
-
12. THEOREM A FROM THEOREM B.
@@ -4078,103 +4022,7 @@ smaller positive eps. The elementary upper bound n^(1+eta) from
Lemma 2.1 shows the exponent 1 in Theorem A is the largest possible.
-13. WHERE EACH HYPOTHESIS IS USED.
-
-
-Every result of the paper is stated with the parameters on which it
-depends, and the following table records where each external
-hypothesis is introduced and where it is discharged. All constants
-are fixed before x tends to infinity, except where a parameter
-depends on x, which is stated.
-
-Hypothesis | Introduced | Used / discharged in
--------------------------------+----------------+--------------------------------
-Prime number theorem with | sec 4.1 | 4.2, 8.5, 10.7 (proxy
-logarithmic error | | densities), 11.2
-Mertens asymptotics | sec 4.1 | 4.2, 4.5 (block sieve),
- | | 10.5 (Sel, V(W))
-Siegel-Walfitz (Thm 4.x) | sec 4.1 | 4.4 (harmonic character
- | | estimates), Lemma 10.1
- | | (proxy discrepancy), and
- | | the small-conductor case
- | | of Type I (Prop 10.2)
-Multiplicative large sieve | sec 4.1 | 7.3 (character frame),
- | | 10.4 (Type I, large f)
-Block sieve (Lemma 4.5) | sec 4.5 | Cor 4.6 (rough
- | | polynomial), Lemma 10.1
- | | (proxy discrepancy),
- | | Lemma 10.2 (rough
- | | densities), 10.5 (pre-
- | | sieve), 10.8 (second
- | | sieve for unbalanced
- | | composites), 10.9
- | | (two-variable sieve for
- | | balanced composites)
-Ideal-norm hypothesis (5.12) | sec 5.2 | required by Thm 5.2
- | | (transference); proved
- | | by Thm 7.1
-Distinctness of birth values | sec 6.7 | restored in 6.7 by the
- | | rank expansion
-Discrepancy condition (8.5) | sec 8.1 | required by Thm 9.1
- | | (Type II); supplied by
- | | Lemma 10.1
-Geometric data (5.1) | sec 5.1 | fixed before 6-9;
- | | used by 10.1 and 10.9
-Candidate geometry (10.1) | sec 10.1 | fixed at the start of
- | | single-prime-extraction
- | | step
-Band gaps (9.1) | sec 9.1 | fixed before Type II,
- | | controlled by j_0 in
- | | (10.1)
-
-The quantifier order. Throughout, the choices are made in the order:
-
- delta, Psi, r_0, Q/P geometry
- -> finite prefix for the candidate lower bound, then j' and theta
- -> kappa (with C_bal kappa < Sel/4)
- -> b_1 (with b_1(4h+2) < theta in (10.22)) and b_* = b_1/3
- -> j_** (with c_2 s' 2^(-j_**) < b_*/4)
- -> finite exponent mesh for (10.46)
- -> Type II saving target and the constants it needs (J, E_id,
- A_0, C_1, m, B)
- -> S in Lemma 10.1 (with S > 2B+3)
- -> prime-number-theorem accuracies for this S
- -> x -> infinity.
-
-The only apparent feedback is the requirement c_2 s' < delta for the
-final smooth-image growth in (11.2), and that inequality is
-discharged by fixing r_0 sufficiently large. Everything else is
-fixed in one direction.
-
-Dependence on J. The pattern length J is chosen last among the
-graph analytic parameters: after the ideal norm target E_id and the
-comparison kernels. Its only appearance outside the graph
-construction is in the small-group gain in (6.37) and the small-group
-mask of (10.9); both are invisible once J is fixed.
-
-Dependence on the discrepancy exponent B. B is chosen last in the
-Type II chain: after J and the Mellin cutoffs, and before S. Its
-only role is to state the discrepancy hypothesis (8.5), so it is
-also the only constraint on the cell precision in (10.25).
-
-
-13.1 A ONE-PARAGRAPH SKELETON.
-
-A reader who wants only the shape of the proof can hold the following
-in mind. The candidate set supplies about x/L integers u whose prime
-factors all lie below x^delta. The presieve removes u with a small
-odd prime dividing 2u+1, leaving x/L mass. Type I ensures that
-2u+1 is equidistributed among residue classes up to the sieve level.
-Type II replaces roughness of each factor of 2u+1 by a local density,
-which the two-variable sieve and the Buchstab identity then subtract
-from the presieve mass. The remainder is positive on prime values,
-and the count of distinct primes follows by dividing the surviving
-mass by the largest summand. Everything else is bookkeeping to make
-these five steps uniform in the parameters and correct at every
-scale.
-
-
-14. REFERENCES AND PROOF-DEPENDENCY MAP.
+14. REFERENCES.
14.1 SHORT REFERENCE LIST.
@@ -4219,80 +4067,5 @@ orientation:
Pilatte, C. Improved bounds for the two-point logarithmic Chowla
conjecture. arXiv:2310.19357, v3 (2026).
-The role each of these plays: Erdos (1935), Erdos (1956), and
-Pomerance (1980) were the origin of Theorem A and the smoothness
-expectation in Theorem B. Baker-Harman (1998) were the strongest
-earlier shifted-prime bounds and highlight why a fixed exponent
-delta with x^{1-o(1)} count is a change of kind. Montgomery-Vaughan
-(1974) is the source of the Dirichlet-polynomial mean-value lemma
-in 4.3. Ford-Halberstam (2000) is the source of the block sieve
-inequality in 4.5. Soundararajan (2009) is the source of the
-prime-polynomial moment argument in 8.6. Buchstab (1937) is the
-source of the rough-integer density in 10.7. Matomaki-Radziwill
-(2016) is the modern model for Type II estimates of this kind.
-Alford-Granville-Pomerance and Pilatte are included as context for
-the totient-fiber problem; neither is used in the proof.
-
-
-14.2 PROOF-DEPENDENCY MAP.
-
-The following list is the plain-text version of the paper's diagram
-of dependencies. A result may use anything to its left.
-
- Theorem 5.2 (local transference)
- depends on Section 6 (rare transfers, clean edges) and
- Section 7 (ideal kernels).
-
- Theorem 7.1 (ideal norm bound)
- depends on Sections 7.3 (character frame) and 7.4 (symmetric
- probes).
-
- Theorem 8.1 (shift cancellation)
- depends on Theorem 5.2 and its pairing corollary, and on the
- Mellin and local-energy input of Section 8.5.
-
- Theorem 9.1 (Type II)
- depends on Theorem 8.1.
-
- Proposition 10.1 (candidate mass)
- depends on Section 4 only (prime number theorem).
-
- Proposition 10.2 (Type I)
- depends on Section 4.1 (Siegel-Walfitz and large sieve).
-
- Lemma 10.1 (proxy discrepancy)
- depends on Lemma 4.5 (block sieve) and Theorem 4.4 (Siegel-
- Walfitz).
-
- Lemma 10.2 (rough densities)
- depends on Lemma 4.5 and the prime number theorem.
-
- Sections 10.8 and 10.9 (removing unbalanced and balanced
- composites)
- depend on Theorem 9.1 and Lemma 10.1.
-
- Theorem B
- depends on Propositions 10.1-10.2, Lemmas 10.1-10.2, and
- Sections 10.8-10.9.
-
- Theorem A
- depends on Theorem B and Lemma 2.1.
-
-Section 3 (the shape of the proof) reads these in order. Section 13.1
-compresses the same chain into a single paragraph.
-
-
-14.3 A NOTE ON THE STYLE OF THIS FILE.
-Because the file is meant to be read without LaTeX, several conventions
-differ from a conventional paper.
- - "e(t) = exp(2 pi i t)"; "L = log x"; "T = log L".
- - "z << w" means |z| <= C w for an absolute C, occasionally for a
- constant depending on fixed parameters; those parameters are
- stated.
- - "for large x" means "for all x >= x_0(<fixed parameters>)".
- - "x^(1-o(1))" is defined in the header.
- - A display's equation number, e.g. (10.23), is meant to be cited
- only from later in this file. Section 13 has the index of
- where each hypothesis is actually consumed.