# TCDC-16
## Failure-Line Distance / Twisted-Sum Provider Decomposition
### RootRough-to-Rough Reduction、Liouville-to-Möbius Square Correction、Divisor-Progression Expansion、Bad-Mass Capacity Substitution 與 Provider Engineering

**母方法論：** TCDC — Target–Condition Dual Closure  
**對接研究線：** RG-SCP / SRBRD  
**輪次：** TCDC-16  
**日期：** 2026-10-02  
**版本：** v0.1  
**作者：** Neo.K；Aletheia（GPT-5.6 Sol）  
**性質：** Twisted-Sum Decomposition / Möbius Reduction / Provider-Substitution Kernel  
**前置：** TCDC-00 ～ TCDC-15；SRBRD-00 ～ SRBRD-26  
**狀態：** 將 failure-line distance分解成 neutral、first-twist、second-twist三層；證明 cube-root RootRough下 Liouville twist與 Möbius twist只差 prime-square sparse correction；將 ordinary rough indicator展開成 divisor-progressions，定位 first twist 為 Möbius-in-AP 類型、second twist 為 fixed-sum binary Möbius correlation；同時建立完全不直接估 twist 的 neutral-lower + semiprime-upper + mixed-upper替代閉合；未取得 uniform twisted provider或 positive bad-mass constant gap；Goldbach仍未證明

---

# 0. 本輪目標

TCDC-15 將 joint parity plane參數化為：

$$
\boxed{
\Lambda_m
=
\frac{\mathcal T_{1,m}}{\mathcal N_m},
}
$$

$$
\boxed{
\mathcal C_m
=
\frac{\mathcal T_{2,m}}{\mathcal N_m},
}
$$

其中：

$$
\boxed{
\mathcal N_m=S_m
}
$$

是 neutral joint cube-root survivor，

$$
\boxed{
\mathcal T_{1,m}
=
\sum_{x+y=2m}
R_z(x)R_z(y)\lambda(x),
}
$$

$$
\boxed{
\mathcal T_{2,m}
=
\sum_{x+y=2m}
R_z(x)R_z(y)\lambda(x)\lambda(y).
}
$$

這裡：

$$
R_z(n)
=
\mathbf1_{\operatorname{RootRough}_z(n)}.
$$

本輪要回答：

> $\mathcal T_1,\mathcal T_2$ 究竟可以拆成哪些可交給現有 analytic tools 的 provider terms？

---

# 1. Failure-Line Distance

TCDC-15 exact：

$$
\boxed{
\frac{M_{PP}}{S_m}
=
\frac{
1-2\Lambda_m+\mathcal C_m
}{4}.
}
$$

所以定義 unnormalized failure-line distance：

## Definition TCDC-16-D01

$$
\boxed{
\mathscr D_m
=
\mathcal N_m
-
2\mathcal T_{1,m}
+
\mathcal T_{2,m}.
}
$$

---

# 2. Exact Distance Identity

## Theorem TCDC-16-T01

$$
\boxed{
\mathscr D_m
=
4M_{PP}.
}
$$

proof：

$$
\begin{aligned}
\mathscr D_m
&=
\sum_{x+y=2m}
R_z(x)R_z(y)
\left(
1-\lambda(x)
\right)
\left(
1-\lambda(y)
\right)
\\
&=
4
\sum_{x+y=2m}
\mathbf1_P(x)\mathbf1_P(y)
\\
&=
4M_{PP}.
\end{aligned}
$$

因 cube-root boundary上：

$$
\frac{1-\lambda(n)}2
=
\mathbf1_P(n).
$$

---

# 3. Exact Failure Criterion

因此：

## Corollary TCDC-16-C01

$$
\boxed{
M_{PP}>0
\iff
\mathscr D_m>0.
}
$$

Goldbach failure：

$$
\boxed{
\mathscr D_m=0.
}
$$

所以所有 provider architecture最終都只是要證：

$$
\boxed{
\mathcal N_m
-
2\mathcal T_{1,m}
+
\mathcal T_{2,m}
>0.
}
$$

---

# 4. RootRough vs Ordinary Rough

令：

$$
\boxed{
P(z)
=
\prod_{\substack{p\le z\\p\in\mathbb P}}p
}
$$

以及 ordinary rough indicator：

$$
\boxed{
\rho_z(n)
=
\mathbf1_{(n,P(z))=1}.
}
$$

RootRough：

$$
R_z(n)
$$

不是完全等於：

$$
\rho_z(n),
$$

因 prime：

$$
p\le z
$$

仍是 RootRough。

---

# 5. Exact RootRough Correction

## Theorem TCDC-16-T02

對：

$$
n\ge2,
$$

$$
\boxed{
R_z(n)
=
\rho_z(n)
+
\mathbf1_{\mathbb P}(n)
\mathbf1_{n\le z}.
}
$$

---

# 6. Proof of T02

### Prime $n\le z$

ordinary：

$$
\rho_z(n)=0,
$$

but RootRough true，

因：

$$
n^2\le n
$$

false。

### Prime $n>z$

both indicators equal：

$$
1.
$$

### Composite $n\le z^2$

smallest prime factor：

$$
p\le\sqrt n\le z,
$$

so：

$$
p^2\le n,
$$

hence not RootRough and not ordinary rough。

### Composite $n>z^2$

any prime divisor：

$$
p\le z
$$

automatically satisfies：

$$
p^2<n,
$$

so RootRough and ordinary rough conditions coincide。

證畢。

---

# 7. Small-Prime Correction Size

define：

$$
\boxed{
E_z(n)
=
\mathbf1_{\mathbb P}(n)
\mathbf1_{n\le z}.
}
$$

then：

$$
R_z=\rho_z+E_z.
$$

number of possible corrected endpoints：

$$
\boxed{
\ll
\pi(z)
}
$$

and with：

$$
z\asymp X^{1/3},
$$

$$
\boxed{
\pi(z)
\asymp
X^{1/3}/\log X.
}
$$

So relative to expected joint main scale：

$$
X/\log^2X,
$$

this is lower order。

---

# 8. Liouville vs Möbius on the Cube-Root Boundary

for：

$$
R_z(n)=1,
$$

we know：

$$
n
$$

is prime or semiprime。

For：

### prime

$$
\lambda(n)=\mu(n)=-1.
$$

### distinct semiprime $pq$

$$
\lambda(n)=\mu(n)=+1.
$$

### square semiprime $p^2$

$$
\lambda(p^2)=+1,
$$

$$
\mu(p^2)=0.
$$

---

# 9. Prime-Square Correction

define：

## Definition TCDC-16-D02

$$
\boxed{
Q_z(n)
=
\mathbf1_{
\{
n=p^2:
p\in\mathbb P,
\ p>z
\}
}.
}
$$

Then：

## Theorem TCDC-16-T03
### Liouville-to-Möbius Reduction

$$
\boxed{
\lambda(n)R_z(n)
=
\mu(n)R_z(n)
+
Q_z(n).
}
$$

exactly for：

$$
2\le n\le X,
$$

$$
z^3\ge X.
$$

---

# 10. Sparse Size of the Square Correction

$$
Q_z(n)
$$

requires：

$$
z<p\le\sqrt X.
$$

therefore：

## Theorem TCDC-16-T04

$$
\boxed{
\sum_{n\le X}Q_z(n)
\le
\pi(\sqrt X)
\ll
\frac{\sqrt X}{\log X}.
}
$$

So any joint sum with one $Q_z$ factor is at most：

$$
\boxed{
O(\sqrt X/\log X).
}
$$

up to endpoint multiplicity one。

---

# 11. First Twist as Möbius Twist + Sparse Squares

define：

$$
\boxed{
\mathcal M_{1,m}
=
\sum_{x+y=2m}
R_z(x)R_z(y)\mu(x).
}
$$

and：

$$
\boxed{
\mathcal Q_{1,m}
=
\sum_{x+y=2m}
Q_z(x)R_z(y).
}
$$

T03 gives：

## Theorem TCDC-16-T05

$$
\boxed{
\mathcal T_{1,m}
=
\mathcal M_{1,m}
+
\mathcal Q_{1,m}.
}
$$

with：

$$
\boxed{
|\mathcal Q_{1,m}|
\ll
\sqrt X/\log X.
}
$$

---

# 12. Second Twist as Binary Möbius Correlation + Sparse Squares

define：

$$
\boxed{
\mathcal M_{2,m}
=
\sum_{x+y=2m}
R_z(x)R_z(y)
\mu(x)\mu(y).
}
$$

mixed square-Möbius correction：

$$
\boxed{
\mathcal Q_{\mu,m}
=
\sum_{x+y=2m}
Q_z(x)R_z(y)\mu(y).
}
$$

double-square：

$$
\boxed{
\mathcal Q_{2,m}
=
\sum_{x+y=2m}
Q_z(x)Q_z(y).
}
$$

Then：

## Theorem TCDC-16-T06

$$
\boxed{
\mathcal T_{2,m}
=
\mathcal M_{2,m}
+
2\mathcal Q_{\mu,m}
+
\mathcal Q_{2,m}.
}
$$

and：

$$
\boxed{
|\mathcal Q_{\mu,m}|
+
|\mathcal Q_{2,m}|
\ll
\sqrt X/\log X.
}
$$

---

# 13. Consequence

provided neutral survivor has：

$$
\boxed{
S_m
\gg
X/\log^2X,
}
$$

T04–T06 imply：

## Theorem Template TCDC-16-T07

$$
\boxed{
\Lambda_m
=
\frac{
\mathcal M_{1,m}
}{
S_m
}
+
o(1),
}
$$

$$
\boxed{
\mathcal C_m
=
\frac{
\mathcal M_{2,m}
}{
S_m
}
+
o(1).
}
$$

So asymptotically the Liouville parity problem can be replaced by a Möbius parity problem，up to sparse square corrections。

---

# 14. Ordinary Rough Divisor Expansion

ordinary rough indicator has exact inclusion-exclusion：

## Theorem TCDC-16-T08

$$
\boxed{
\rho_z(n)
=
\sum_{\substack{
d\mid P(z)\\
d\mid n
}}
\mu(d).
}
$$

This expansion is exact but exponentially long。

sieve methods replace it by truncated / weighted versions。

---

# 15. Neutral Divisor-Progression Expansion

replace RootRough by ordinary rough first，

ignoring T02 small-prime correction。

neutral bulk：

$$
\widetilde{\mathcal N}_m
=
\sum_{x+y=2m}
\rho_z(x)\rho_z(y).
$$

T08：

## Theorem TCDC-16-T09

$$
\boxed{
\widetilde{\mathcal N}_m
=
\sum_{d,e\mid P(z)}
\mu(d)\mu(e)
A_{d,e}(2m),
}
$$

where：

$$
\boxed{
A_{d,e}(N)
=
\#\left\{
x:
x+y=N,
\ d\mid x,
\ e\mid y
\right\}.
}
$$

---

# 16. CRT Compatibility

conditions：

$$
x\equiv0\pmod d,
$$

$$
x\equiv N\pmod e.
$$

are soluble iff：

## Theorem TCDC-16-T10

$$
\boxed{
(d,e)\mid N.
}
$$

when soluble，

$x$ lies in one residue class modulo：

$$
\boxed{
[d,e].
}
$$

Thus neutral sieve reduces to counting integers in CRT progressions。

This is exactly the setting where positive sieve weights and mean-distribution technology are structurally natural。

---

# 17. First Möbius-Twisted Divisor Expansion

define ordinary-rough first twist：

$$
\boxed{
\widetilde{\mathcal M}_{1,m}
=
\sum_{x+y=N}
\rho_z(x)\rho_z(y)\mu(x).
}
$$

Then：

## Theorem TCDC-16-T11

$$
\boxed{
\widetilde{\mathcal M}_{1,m}
=
\sum_{d,e\mid P(z)}
\mu(d)\mu(e)
B_{d,e}(N),
}
$$

where：

$$
\boxed{
B_{d,e}(N)
=
\sum_{\substack{
x\equiv0\pmod d\\
x\equiv N\pmod e
}}
\mu(x).
}
$$

So after CRT：

# First Twist Provider

must control Möbius sums in arithmetic progressions with moduli：

$$
[d,e].
$$

---

# 18. Second Möbius-Twisted Divisor Expansion

define：

$$
\boxed{
\widetilde{\mathcal M}_{2,m}
=
\sum_{x+y=N}
\rho_z(x)\rho_z(y)
\mu(x)\mu(y).
}
$$

Then：

## Theorem TCDC-16-T12

$$
\boxed{
\widetilde{\mathcal M}_{2,m}
=
\sum_{d,e\mid P(z)}
\mu(d)\mu(e)
C_{d,e}(N),
}
$$

with：

$$
\boxed{
C_{d,e}(N)
=
\sum_{\substack{
x\equiv0\pmod d\\
x\equiv N\pmod e
}}
\mu(x)\mu(N-x).
}
$$

So after CRT：

# Second Twist Provider

must control a fixed-sum binary Möbius correlation inside arithmetic progressions。

---

# 19. Hierarchy of Analytic Difficulty

T09：

$$
A_{d,e}(N)
$$

is neutral progression counting。

T11：

$$
B_{d,e}(N)
$$

needs one Möbius sign。

T12：

$$
C_{d,e}(N)
$$

needs two correlated Möbius signs at fixed sum。

So provider hardness naturally orders：

$$
\boxed{
\text{Neutral}
<
\text{One Twist}
<
\text{Two Twists}.
}
$$

---

# 20. Why Full Inclusion-Exclusion Is Not the Algorithm

T09–T12 contain all：

$$
d,e\mid P(z).
$$

This is not computationally or analytically usable directly。

Instead choose sieve weights：

$$
\boxed{
\lambda_d,
\lambda'_e
}
$$

supported：

$$
d,e\le D.
$$

Then approximate：

$$
\rho_z
$$

from above / below。

Neutral and capacity terms can use：

- linear sieve；
- Selberg sieve；
- well-factorable weights。

Twisted terms require cancellation after those weights are inserted。

---

# 21. Direct Twisted Provider Route

Suppose proof-admissible：

$$
\boxed{
\mathcal N_m\ge L_N,
}
$$

$$
\boxed{
\mathcal T_{1,m}\le U_1,
}
$$

$$
\boxed{
\mathcal T_{2,m}\ge L_2.
}
$$

Then T01：

## Theorem TCDC-16-T13
### Direct Twisted Closure

$$
\boxed{
M_{PP}
\ge
\frac{
L_N-2U_1+L_2
}{4}.
}
$$

So sufficient：

$$
\boxed{
L_N-2U_1+L_2>0.
}
$$

---

# 22. Capacity Substitute for the Second Twist

TCDC-15：

$$
\boxed{
\mathcal T_{2,m}
=
S_m-4M_{PS}.
}
$$

because：

$$
\mathcal C_m
=
1-4M_{PS}/S_m.
$$

If：

$$
\boxed{
M_{PS}\le U_{PS},
}
$$

then：

## Theorem TCDC-16-T14

$$
\boxed{
\mathcal T_{2,m}
\ge
\mathcal N_m
-
4U_{PS}.
}
$$

Thus no direct two-Liouville theorem is required。

---

# 23. One-Twist + Capacity Closure

Combine T13–T14：

## Theorem TCDC-16-T15

if：

$$
\mathcal N_m\ge L_N,
$$

$$
\mathcal T_{1,m}\le U_1,
$$

$$
M_{PS}\le U_{PS},
$$

then：

$$
\boxed{
M_{PP}
\ge
\frac{
2L_N-2U_1-4U_{PS}
}{4}
}
$$

or：

$$
\boxed{
M_{PP}
\ge
\frac{
L_N-U_1
}{2}
-
U_{PS}.
}
$$

sufficient：

$$
\boxed{
U_1+2U_{PS}<L_N.
}
$$

---

# 24. Semiprime Marginal Substitutes the First Twist

exact：

$$
\mathcal T_{1,m}
=
n_{SP}-n_0.
$$

and：

$$
S_m=n_0+n_{SP}.
$$

so：

## Theorem TCDC-16-T16

$$
\boxed{
\mathcal T_{1,m}
=
2n_{SP}-S_m.
}
$$

Therefore an upper bound：

$$
\boxed{
n_{SP}\le U_{SP}
}
$$

gives：

$$
\boxed{
\mathcal T_{1,m}
\le
2U_{SP}-S_m.
}
$$

So first-order Liouville control can also be replaced by a positive semiprime-marginal upper。

---

# 25. Full Bad-Mass Substitution

substitute T16 into T15。

Or directly use the transport matrix：

$$
S_m
=
M_{PP}
+
2M_{PS}
+
M_{SS},
$$

$$
n_{SP}
=
M_{PS}+M_{SS}.
$$

subtract：

## Theorem TCDC-16-T17
### Bad-Mass Identity

$$
\boxed{
M_{PP}
=
S_m
-
n_{SP}
-
M_{PS}.
}
$$

This is exact。

---

# 26. Neutral-Lower + Two-Upper Closure

If：

$$
\boxed{
S_m\ge L_S,
}
$$

$$
\boxed{
n_{SP}\le U_{SP},
}
$$

$$
\boxed{
M_{PS}\le U_{PS},
}
$$

then：

## Theorem TCDC-16-T18
### Full Capacity-Substitution Closure

$$
\boxed{
M_{PP}
\ge
L_S
-
U_{SP}
-
U_{PS}.
}
$$

Therefore：

$$
\boxed{
L_S
>
U_{SP}+U_{PS}
}
$$

implies Goldbach success。

---

# 27. Why T18 Is Important

T18 requires：

### one lower

$$
\boxed{
\text{neutral rough survivor lower}.
}
$$

### two uppers

$$
\boxed{
\text{semiprime marginal upper},
}
$$

$$
\boxed{
\text{mixed prime-semiprime upper}.
}
$$

It does **not** directly require：

- first Liouville cancellation；
- second Liouville correlation cancellation。

So the parity problem is converted into：

# Bad-Mass Capacity Saturation

---

# 28. Parity Has Not Magically Disappeared

If ordinary positive sieve weights could always make：

$$
U_{SP},
U_{PS}
$$

sharp enough to satisfy T18，

Goldbach would already be solved。

So classical parity obstruction must reappear as：

- upper constants too large；
- inability to lower-bound $S_m$ and upper both bad classes simultaneously with enough precision；
- shared remainder constraints。

Thus：

## Research Firewall TCDC-16-R01

$$
\boxed{
\text{T18 changes the provider polarity;
it does not invalidate the parity obstruction}.
}
$$

---

# 29. Least-Factor Decomposition of the Semiprime Marginal

cube-root semiprime endpoint：

$$
x=rp,
$$

with：

$$
z<r\le p.
$$

so：

## Theorem TCDC-16-T19

$$
\boxed{
n_{SP}
=
\sum_{r>z}n_r.
}
$$

This is the SRBRD semiprime-row decomposition。

Each：

$$
n_r
$$

requires：

- $p$ prime；
- opposite endpoint RootRough。

Therefore：

# Semiprime Marginal Upper

is a positive prime-cofactor + rough-complement sieve problem。

---

# 30. Mixed Branch Decomposition

mixed branch：

$$
M_{PS}
$$

is：

## Theorem TCDC-16-T20

$$
\boxed{
M_{PS}
=
\sum_{r>z}M_{r,0}.
}
$$

Each term requires：

$$
\boxed{
2m=rp+q,
}
$$

with：

$$
r,p,q
$$

prime。

So：

# Mixed Upper

is an aggregate two-prime / three-prime-form upper problem。

This is exactly the TCDC-09 ～ 11 capacity line。

---

# 31. Provider Decomposition Table

| Object | Exact role | Preferred polarity | Natural tools |
|---|---|---|---|
| $S_m$ | neutral total | lower | linear/Chen sieve, convolution distribution |
| $n_{SP}$ | semiprime marginal | upper | least-factor/Buchstab shells, upper sieve |
| $M_{PS}$ | mixed bad branch | upper | two-prime Selberg, convolution upper, transport LP |
| $\mathcal T_1$ | first parity twist | upper | Möbius-in-AP / direct parity theorem |
| $\mathcal T_2$ | second parity twist | lower | binary Möbius / Chowla-type correlation |
| $\Gamma,G_{\rm mir}$ | capacity reserve | lower | LP / mirror transport |

The TCDC strategy is allowed to choose whichever provider column is strongest。

---

# 32. Normalized Bad-Mass Gate

define：

$$
\boxed{
u_{SP}
=
\frac{
U_{SP}
}{
L_S
},
}
$$

$$
\boxed{
u_{PS}
=
\frac{
U_{PS}
}{
L_S
}.
}
$$

T18 sufficient：

## Theorem TCDC-16-T21

$$
\boxed{
u_{SP}+u_{PS}<1.
}
$$

This is a simple two-upper capacity simplex。

---

# 33. Buchstab Rank-One Benchmark

TCDC-14 benchmark：

$$
\boxed{
s_*
=
\frac{\log2}{1+\log2}
\approx
0.409383890850,
}
$$

$$
\boxed{
p_*s_*
\approx
0.241788720763.
}
$$

ideal bad-mass sum：

$$
\boxed{
s_*+p_*s_*
=
1-p_*^2.
}
$$

Therefore ideal residual：

$$
\boxed{
1-s_*-p_*s_*
=
p_*^2
\approx
0.348827388387.
}
$$

exactly the benchmark $PP$ fraction。

---

# 34. Multiplier Budget

suppose upper providers satisfy：

$$
\boxed{
U_{SP}
\le
\kappa_{SP}s_*S_m,
}
$$

$$
\boxed{
U_{PS}
\le
\kappa_{PS}p_*s_*S_m.
}
$$

ignoring neutral-lower loss for the benchmark。

T18 needs：

## Theorem Template TCDC-16-T22

$$
\boxed{
\kappa_{SP}s_*
+
\kappa_{PS}p_*s_*
<
1.
}
$$

---

# 35. If Semiprime Marginal Were Ideal

take：

$$
\kappa_{SP}=1.
$$

then mixed multiplier ceiling：

$$
\boxed{
\kappa_{PS}
<
\frac{
1-s_*
}{
p_*s_*
}
=
\frac1{s_*}
}
$$

numerically：

$$
\boxed{
\kappa_{PS}
<
2.442695040889.
}
$$

This reproduces the TCDC-14 bulk mixed-capacity ceiling。

---

# 36. If Mixed Upper Were Ideal

take：

$$
\kappa_{PS}=1.
$$

then semiprime multiplier ceiling：

$$
\boxed{
\kappa_{SP}
<
\frac{
1-p_*s_*
}{
s_*}
}
$$

numerically：

$$
\boxed{
\kappa_{SP}
<
1.852078931739.
}
$$

So semiprime marginal upper may tolerate a larger multiplier than an exact branch-transfer theorem would suggest。

---

# 37. Equal-Multiplier Benchmark

if：

$$
\kappa_{SP}
=
\kappa_{PS}
=
\kappa,
$$

T22 gives：

$$
\boxed{
\kappa
<
\frac{
1
}{
s_*(1+p_*)
}
}
$$

numerically：

$$
\boxed{
\kappa
<
1.535691124237.
}
$$

This provides a simple target for symmetric provider stacks。

---

# 38. Neutral-Lower Loss

if only：

$$
L_S
=
\alpha S_m,
$$

with：

$$
0<\alpha\le1,
$$

while uppers normalize to true $S_m$：

T18 requires：

$$
\boxed{
\kappa_{SP}s_*
+
\kappa_{PS}p_*s_*
<
\alpha.
}
$$

So every neutral-lower constant loss directly tightens both upper budgets。

This is where Chen-style neutral lower efficiency matters。

---

# 39. Capacity Reserve Can Be Added

TCDC-13 lower：

$$
M_{PP}
\ge
S_m-n_{SP}-M_{PS}
+
\Gamma_{\rm extra}
$$

is not an identity if $\Gamma_{\rm extra}$ double-counts the same matrix mass。

So capacity reserve must be inserted carefully。

A safe form uses improved upper：

$$
\boxed{
U_{PS}^{eff}
=
\min(
U_{PS}^{analytic},
U_{PS}^{transport}
).
}
$$

or directly a coupled bad-mass support：

$$
\boxed{
U_{\mathrm{bad}}^{LP}
\ge
n_{SP}+M_{PS}
}
$$

under the relevant failure outer polytope。

Then：

$$
\boxed{
M_{PP}
\ge
L_S-U_{\mathrm{bad}}^{LP}.
}
$$

---

# 40. Coupled Bad-Mass Support

define：

## Definition TCDC-16-D03

$$
\boxed{
B_{\max}^{fail}
=
\sup_{\mathfrak F}
\left(
n_{SP}+M_{PS}
\right)
}
$$

where $\mathfrak F$ is a proof-admissible outer transport polytope consistent with neutral sender constraints。

Then：

## Theorem TCDC-16-T23

$$
\boxed{
M_{PP}
\ge
L_S-B_{\max}^{fail}.
}
$$

if：

$$
\boxed{
L_S>B_{\max}^{fail},
}
$$

failure impossible。

This is the bad-mass analogue of TCDC-07 support-function closure。

---

# 41. RootRough Small Correction Does Not Change Main Provider Type

T02：

$$
R_z=\rho_z+E_z.
$$

terms involving：

$$
E_z
$$

have at least one endpoint：

$$
\le z.
$$

their total number：

$$
\boxed{
O(z/\log z)
=
O(X^{1/3}/\log X).
}
$$

So under any main-scale lower：

$$
S_m\gg X/\log^2X,
$$

these terms are negligible。

Thus provider engineering can work with ordinary rough weights in the bulk，

then restore RootRough by explicit correction。

---

# 42. Square Corrections Are Also Main-Scale Negligible

T04：

$$
Q_z
$$

support：

$$
O(\sqrt X/\log X).
$$

relative：

$$
\frac{
\sqrt X/\log X
}{
X/\log^2X
}
=
\frac{
\log X
}{
\sqrt X
}
\to0.
$$

So Liouville-to-Möbius replacement is asymptotically safe once neutral main scale is established。

---

# 43. First-Twist Provider After Möbius Reduction

T11 inner term：

$$
B_{d,e}(N)
=
\sum_{x\equiv a\pmod q}
\mu(x),
$$

for：

$$
q=[d,e].
$$

Classical zero-free-region / Siegel-Walfisz technology controls Möbius sums in progressions for relatively small moduli。

But TCDC weighted sieve naturally generates much larger / structured moduli。

Therefore：

## Research Target TCDC-16-H01

need a weighted mean theorem of the form：

$$
\boxed{
\sum_q
\omega_q
\left|
\sum_{x\equiv a_q\pmod q}
\mu(x)
\right|
\ll
\text{acceptable error}
}
$$

for the actual sieve-weight class。

---

# 44. Second-Twist Provider After Möbius Reduction

T12 inner：

$$
C_{d,e}(N)
=
\sum_{x\equiv a\pmod q}
\mu(x)\mu(N-x).
$$

This is not the same as standard small-shift Chowla：

$$
\sum_n\mu(n)\mu(n+h).
$$

Here：

- the relation is fixed sum；
- the second argument is $N-x$；
- congruence restrictions are present；
- rough / sieve weights are present。

So：

## Research Firewall TCDC-16-R02

$$
\boxed{
\text{existing small/growing-shift Chowla theorems do not directly estimate }
C_{d,e}(N).
}
$$

---

# 45. Why Maynard Helps Neutral / Capacity but Not Möbius Twists

Maynard's large-moduli results provide mean distribution for primes in structured arithmetic progressions：

- well-factorable weights up to $x^{3/5-\varepsilon}$；
- uniform residue classes beyond $x^{1/2}$ for moduli with convenient divisors。

These are highly relevant to：

- neutral rough lower；
- semiprime-shell upper；
- mixed prime upper。

But the coefficients：

$$
\mu(x)
$$

or：

$$
\mu(x)\mu(N-x)
$$

are outside the direct theorem statements。

Therefore：

## Research Firewall TCDC-16-R03

$$
\boxed{
\text{Maynard distribution is a neutral/capacity provider,
not a direct first/second Möbius-twist provider}.
}
$$

---

# 46. Current 2026 Liouville Results

recent quantitative logarithmic Chowla work controls logarithmically weighted two-point Liouville correlations uniformly for shifts up to a small power of：

$$
\log x.
$$

This is genuine second-order parity technology。

But TCDC second twist is：

$$
\lambda(x)\lambda(N-x)
$$

inside a fixed-sum rough-conditioned slice。

So theorem geometry and weighting are different。

Similarly recent sign-change results in arithmetic progressions prove occurrence of both Liouville signs，

not the weighted mean required by：

$$
\Lambda_m.
$$

---

# 47. Provider Substitution Lattice

TCDC-16 yields four proof stacks。

## Stack A — Direct Twists

$$
\boxed{
N\text{ lower}
+
T_1\text{ upper}
+
T_2\text{ lower}.
}
$$

## Stack B — One Twist + Capacity

$$
\boxed{
N\text{ lower}
+
T_1\text{ upper}
+
M_{PS}\text{ upper}.
}
$$

## Stack C — Full Bad-Mass Substitution

$$
\boxed{
S\text{ lower}
+
n_{SP}\text{ upper}
+
M_{PS}\text{ upper}.
}
$$

## Stack D — Coupled LP Bad-Mass Support

$$
\boxed{
S\text{ lower}
+
B_{\max}^{fail}\text{ upper}.
}
$$

Proof engineering should choose the cheapest legal stack。

---

# 48. Why Stack C Is Especially Attractive

all three directions align with standard sieve strengths：

- neutral total：
  lower via weighted almost-prime sieve；
- semiprime marginal：
  upper via positive shell counts；
- mixed branch：
  upper via two-prime / transport capacity。

So Stack C is：

# One Lower + Two Uppers

rather than requiring sign cancellation。

This is likely the cleanest parity-substitution architecture currently available inside TCDC。

---

# 49. Why Stack D May Be Stronger

Stack C separately upper-bounds：

$$
n_{SP}
$$

and：

$$
M_{PS}.
$$

These bad masses are not independent。

A coupled failure LP can exploit：

- shared rows；
- prime-sink saturation；
- mirror competition；
- internal symmetry。

Thus：

$$
\boxed{
B_{\max}^{fail}
\le
U_{SP}+U_{PS}
}
$$

for any consistent separate uppers。

So Stack D avoids double-counting bad capacity。

---

# 50. Failure-Line Distance as Bad-Mass Slack

T17：

$$
M_{PP}
=
S_m-n_{SP}-M_{PS}.
$$

therefore：

## Theorem TCDC-16-T24

$$
\boxed{
\mathscr D_m
=
4
\left(
S_m
-
n_{SP}
-
M_{PS}
\right).
}
$$

So the parity-plane distance has an exactly equivalent：

# Neutral Mass − Bad Mass

interpretation。

---

# 51. Two Geometries of the Same Distance

## Parity Plane

$$
\boxed{
\mathscr D/S
=
1-2\Lambda+\mathcal C.
}
$$

## Capacity Plane

$$
\boxed{
\mathscr D/4
=
S-n_{SP}-M_{PS}.
}
$$

The first is suitable for：

- Liouville / Chowla technology。

The second is suitable for：

- sieve upper；
- transport LP；
- Buchstab shells。

This is the main methodological result of TCDC-16。

---

# 52. Formal Theorem Queue

```yaml
TCDC_16:

  T01:
    failure_line_distance_equals_four_PP

  T02:
    RootRough_equals_ordinary_rough_plus_small_prime_correction

  T03:
    Liouville_equals_Mobius_plus_prime_square_correction_on_boundary

  T04:
    square_correction_sparse

  T05:
    first_twist_Mobius_reduction

  T06:
    second_twist_Mobius_reduction

  T07:
    normalized_twists_asymptotically_Mobius_under_neutral_main_scale

  T08:
    ordinary_rough_exact_divisor_expansion

  T09:
    neutral_divisor_progression_expansion

  T10:
    CRT_compatibility

  T11:
    first_twist_Mobius_AP_expansion

  T12:
    second_twist_binary_Mobius_expansion

  T13:
    direct_twisted_closure

  T14:
    mixed_capacity_substitutes_second_twist

  T15:
    one_twist_plus_capacity_closure

  T16:
    semiprime_marginal_substitutes_first_twist

  T17:
    bad_mass_identity

  T18:
    neutral_lower_plus_two_upper_closure

  T19:
    semiprime_marginal_least_factor_decomposition

  T20:
    mixed_branch_row_decomposition

  T21:
    normalized_bad_mass_gate

  T22:
    benchmark_multiplier_budget

  T23:
    coupled_bad_mass_support_closure

  T24:
    failure_distance_equals_four_bad_mass_slack
```

---

# 53. Barrier Queue

```yaml
TCDC_16_Barriers:

  B01:
    first_Mobius_twist_large_structured_modulus_mean_not_proved

  B02:
    second_fixed_sum_Mobius_correlation_not_proved

  B03:
    small_shift_logarithmic_Chowla_not_direct_match

  B04:
    neutral_lower_constant_loss

  B05:
    semiprime_marginal_upper_not_sharp_enough_uniformly

  B06:
    mixed_branch_upper_not_sharp_enough_uniformly

  B07:
    coupled_bad_mass_support_constant_not_derived

  B08:
    no_positive_failure_distance_lower
```

---

# 54. Acceptance Gate

TCDC-16 formal core前需：

1. failure-line distance algebra；
2. RootRough vs ordinary rough exact correction；
3. Liouville vs Möbius square correction exact；
4. square correction size；
5. normalized negligibility conditional on neutral main scale；
6. divisor expansion CRT compatibility；
7. first vs second twist analytic difficulty separated；
8. capacity substitution directions correct；
9. bad-mass identity exact；
10. upper/lower provider polarity explicit；
11. multiplier benchmark labeled benchmark；
12. external theorems not treated as direct twist providers；
13. no claim that parity obstruction is solved；
14. no Goldbach proof claim。

---

# 55. Verdict

本輪可以正式給：

$$
\boxed{
\mathrm{TCDC\mbox{-}16\ PROVIDER\mbox{-}SUBSTITUTION\ KERNEL\ PASS}.
}
$$

最重要 exact failure-distance identity：

$$
\boxed{
\mathscr D_m
=
\mathcal N_m
-
2\mathcal T_{1,m}
+
\mathcal T_{2,m}
=
4M_{PP}.
}
$$

cube-root boundary還允許：

$$
\boxed{
\lambda(n)R_z(n)
=
\mu(n)R_z(n)
+
Q_z(n),
}
$$

其中 prime-square correction只有：

$$
\boxed{
O(\sqrt X/\log X)
}
$$

support。

所以 direct parity route可 reduction到：

- Möbius in AP；
- fixed-sum binary Möbius correlation。

但更重要的是 exact bad-mass identity：

$$
\boxed{
M_{PP}
=
S_m
-
n_{SP}
-
M_{PS}.
}
$$

因此只要：

$$
\boxed{
L_S
>
U_{SP}
+
U_{PS},
}
$$

就 closure。

這把直接 Liouville問題替換成：

# Neutral Lower
+
# Semiprime Marginal Upper
+
# Mixed Branch Upper

也就是：

# One Lower + Two Uppers

的 provider stack。

rank-one benchmark中：

$$
s_*
\approx
0.409383890850,
$$

$$
p_*s_*
\approx
0.241788720763,
$$

ideal residual：

$$
p_*^2
\approx
0.348827388387.
$$

若 semiprime marginal upper理想，

mixed upper multiplier只需：

$$
\boxed{
\kappa_{PS}
<
2.442695040889.
}
$$

若 mixed upper理想，

semiprime upper可允許：

$$
\boxed{
\kappa_{SP}
<
1.852078931739.
}
$$

若兩者有同一 multiplier，

需要：

$$
\boxed{
\kappa
<
1.535691124237.
}
$$

所以 parity barrier現在被 quantitatively轉成：

> positive upper providers的常數是否能一起落進 bad-mass simplex。

現有外部 audit顯示：

- Maynard large-moduli成果明確支援 neutral / capacity方向：well-factorable weights可到 $x^{3/5-\varepsilon}$，uniform residue-class工作則超過 $x^{1/2}$於具有合適因子的模數；
- 2026 logarithmic Chowla進展確實控制某些 growing small shifts 的二點 Liouville correlation；
- 但 fixed-sum、cube-root rough-conditioned、unweighted TCDC twist目前仍沒有 direct theorem match。

因此：

$$
\boxed{
\mathrm{NO\ POSITIVE\ FAILURE\mbox{-}DISTANCE\ LOWER\ YET}.
}
$$

$$
\boxed{
\mathrm{GOLDBACH\ STILL\ OPEN}.
}
$$

---

# 56. 下一輪建議

# TCDC-17
## Bad-Mass Simplex / Semiprime-Marginal Upper Constant Closure

TCDC-16顯示目前最符合 sieve polarity的 stack是：

$$
\boxed{
L_S-U_{SP}-U_{PS}.
}
$$

下一輪先不要再碰直接 Liouville twist。

### Part A — Neutral Lower

沿用：

$$
u\approx8
$$

Chen-neutral lower，

求 normalized：

$$
\alpha_S.
$$

### Part B — Semiprime Marginal Upper

對：

$$
n_{SP}
=
\sum_{r>z}n_r
$$

建立 least-factor integral upper constant：

$$
\kappa_{SP}.
$$

### Part C — Mixed Upper

對：

$$
M_{PS}
=
\sum_rM_{r,0}
$$

加入：

- two-prime upper；
- center locks；
- LP transport discount。

### Part D — Simplex Gate

直接測：

$$
\boxed{
\alpha_S
>
u_{SP}+u_{PS}.
}
$$

### Part E — Coupled Bad-Mass LP

若 separate constants仍太鬆，

直接計：

$$
B_{\max}^{fail}
$$

避免 semiprime marginal與 mixed branch double-count。

這會是第一次真正不碰 direct parity twist，
而完全依靠：

# positive sieve polarity + TCDC capacity

嘗試逼出 failure-distance positive lower。

---

# 附錄 A：Twisted Distance

```yaml
neutral:
  N

first_twist:
  T1

second_twist:
  T2

failure_distance:
  D:
    N-2*T1+T2

exact:
  D:
    4*M_PP
```

---

# 附錄 B：Möbius Reduction

```yaml
RootRough:
  R_z:
    ordinary_rough
    + prime_le_z_correction

Liouville_on_boundary:
  lambda*R_z:
    mu*R_z
    + prime_square_correction

prime_square_support:
  O(
    sqrt(X)/log(X)
  )
```

---

# 附錄 C：Provider Substitution

```yaml
direct:
  lower:
    N
  upper:
    T1
  lower_second:
    T2

hybrid:
  lower:
    N
  upper:
    T1
  upper_mixed:
    M_PS

full_capacity:
  lower:
    S
  upper:
    n_SP
  upper:
    M_PS

exact_bad_mass_identity:
  M_PP:
    S-n_SP-M_PS
```

---

# 附錄 D：Benchmark Simplex

```yaml
p_star:
  0.590616109150

s_star:
  0.409383890850

ideal_bad_mass:

  semiprime_marginal:
    s_star

  mixed:
    p_star*s_star

ideal_good_residual:
  p_star^2

condition:
  kappa_SP*s_star
  + kappa_PS*p_star*s_star
  <
  1
```

---

# 附錄 E：Commit Decision

```yaml
research_round:
  TCDC-16

decision:

  failure_distance:
    exact

  RootRough_to_ordinary_rough:
    exact_with_sparse_correction

  Liouville_to_Mobius:
    exact_with_prime_square_correction

  first_twist_provider_shape:
    Mobius_in_AP

  second_twist_provider_shape:
    fixed_sum_binary_Mobius

  direct_external_provider:
    not_found

  bad_mass_substitution:
    proved

  positive_bad_mass_gap:
    not_proved

  Goldbach_proved:
    false

next:
  TCDC-17_BadMassSimplex_SemiprimeMarginal_UpperConstant_Closure
```

---

# References / External Technology Audit

1. Terence Tao, **A general parity problem obstruction**, 2014.

   Relevant point:
   exact prime-factor parity properties are obstructed for broad classes of sieve constructions. This motivates separating parity-sensitive twisted sums from parity-neutral positive sieve quantities rather than treating all three providers as equally accessible.

2. James Maynard, **Primes in arithmetic progressions to large moduli II: Well-factorable estimates**, *Memoirs of the American Mathematical Society* 306 (1543), 2025.

   Relevant fact:
   establishes mean value theorems for primes in progressions to moduli as large as $x^{3/5-\varepsilon}$ with suitably well-factorable weights. This is directly relevant to neutral and capacity sieve weights, not to Möbius-twisted signs.

3. James Maynard, **Primes in arithmetic progressions to large moduli III: Uniform residue classes**, *Memoirs of the American Mathematical Society* 306 (1544), 2025.

   Relevant fact:
   obtains mean-value estimates beyond $x^{1/2}$ for moduli with conveniently sized divisors, completely uniform in the residue classes. Again this is a neutral/distribution provider rather than a direct Möbius-correlation theorem.

4. Jizhou Guo, **Quantitative Logarithmic Chowla Correlations Uniformly over Growing Shifts**, arXiv:2608.23500, 2026.

   Relevant fact:
   proves logarithmically weighted two-point Liouville correlation bounds uniformly for shifts up to $(\log x)^\kappa$ for fixed $\kappa<1/700$. It explicitly does not prove the ordinary Cesàro two-point Chowla conjecture. The TCDC second twist is fixed-sum and rough-conditioned, so there is no direct theorem match.

5. Kevin Ford and Maksym Radziwiłł, **Sign changes of the Liouville function in arithmetic progressions**, arXiv:2605.03349, 2026.

   Relevant fact:
   proves existence of both Liouville signs in a residue class within polynomial range. This is occurrence information, not the quantitative fixed-sum mean or pair correlation required by the direct TCDC twist route.