This document summarizes the saddle-layer approach to colossally abundant (CA) numbers, Robin's inequality, and a possible route toward a finite-to-infinite proof strategy for RH via windowed Lyapunov control.
The central philosophy:
-
CA numbers are generated by a layered marginal optimization process.
-
The layer system naturally produces the RH critical scale:
[ \frac{1}{\sqrt{x}\log x}. ]
-
The all-layer saddle system generates a positive asymptotic barrier constant:
[ 2(\sqrt{2}-1). ]
-
The main remaining challenge is proving that discrete prime-event fluctuations cannot overcome the saddle barrier.
For
[ n=\prod p^{a_p}, ]
we have
\prod_{p^a\Vert n} \frac{1-p^{-(a+1)}}{1-p^{-1}}. ]
Define the marginal event:
[ (p,j): \quad p^{j-1}\mapsto p^j. ]
Its exact gain is
\log\left( \frac{1-p^{-(j+1)}}{1-p^{-j}} \right). ]
Define the event threshold:
\frac{g_j(p)}{\log p}. ]
CA numbers are generated by admitting events in decreasing order of (\varepsilon_{p,j}).
For large (p),
p^{-j}-p^{-(j+1)}+O(p^{-2j}). ]
Thus
[ \varepsilon_{p,j} \sim \frac{1}{p^j\log p}. ]
Layer cutoffs (P_j) satisfy
[ \varepsilon P_j^j\log P_j \approx 1. ]
Hence
[ P_j \sim j^{1/j}P_1^{1/j}. ]
This is the fundamental saddle-layer scaling law.
Define
j(j^{1/j}-1). ]
Examples:
[ c_2=2(\sqrt2-1)\approx0.828427, ]
[ c_3\approx1.326352, ]
[ c_4=4(\sqrt2-1)\approx1.656854. ]
Asymptotically:
[ c_j\sim \log j. ]
Define
\frac1{x\log x} \sum_{j\ge2} c_jx^{1/j}. ]
Normalized form:
\sum_{j\ge2} c_jx^{1/j-1/2}. ]
As (x\to\infty):
2(\sqrt2-1). } ]
Also:
[ \sum_{j\ge2}P_j \sim \sqrt{2P_1}. ]
Normalized:
\sqrt2. } ]
These constants arise directly from the saddle scaling law.
Define Robin ratio:
\frac{\sigma(n)}{e^\gamma n\log\log n}. ]
Define the Lyapunov function:
-\log R(n). ]
Robin is equivalent to:
[ \mathcal L(n)>0. ]
For CA candidates:
\Phi(x)-\Omega(x), ]
where:
- (\Phi(x)) is the positive all-layer saddle barrier.
- (\Omega(x)) is the prime/discretization residual.
Normalized:
\sqrt{x}\log x,\mathcal L(x). ]
Numerically:
- (\Lambda(x)) remains positive.
- (\Lambda(x)) appears bounded away from zero.
- (\Lambda(x)) fluctuates but remains inside the all-layer barrier.
A CA transition adds either:
[ Q=p ]
or
[ Q=pq. ]
Semiprime transitions occur only when two marginal events tie:
[ \varepsilon_{p,j}=\varepsilon_{q,k}. ]
Exact jump formula:
\sum_{(p,j)\in E}g_j(p). ]
Define:
\frac1{\log N\log\log N}. ]
Then:
\operatorname{Curv}_N(Q). ]
The local no-crossing condition is:
[ \boxed{ \Delta\mathcal L^- < \mathcal L(N). } ]
The strongest symbolic progress comes from the continuous saddle frontier.
Let (u) parameterize the event frontier.
Continuous layer cutoffs satisfy:
[ uX_j(u)^j\log X_j(u)=1. ]
The continuous frontier imbalance satisfies:
(u-\lambda(u))+,dA(u) + (\lambda(u)-u)+,dA(u) + E(u),du. ]
Integrating over a window:
[ \boxed{ \sum_W D^- \le \sum_W D^+ + \Phi(x_0)-\Phi(x_1) + \int_W E(u),du. } ]
This is the key telescoping identity.
Interpretation:
- positive Robin-rise spikes consume saddle potential,
- but the total potential drop compensates them over windows.
Discrete CA events differ from the continuous flow because of prime gaps.
For layer (j), local threshold spacing is:
[ \Delta\varepsilon_{p,j} \sim \frac{h}{p^{j+1}\log p}, ]
where (h) is the local prime gap.
Discrete event error:
[ E_{p,j} \sim \frac{h}{xp}, ]
with (x=P_1).
Main contribution:
[ E_{\rm disc}(x) \sim \frac{g(P_2)}{xP_2} \sim \frac{g(\sqrt{x})}{x^{3/2}}. ]
Compare to Lyapunov margin:
[ \mathcal L(x)\asymp \frac1{\sqrt{x}\log x}. ]
Ratio:
[ \frac{E_{\rm disc}}{\mathcal L} \sim \frac{g(\sqrt{x})\log x}{x}. ]
Under extremely weak prime-gap behavior this tends to zero.
This suggests:
o\left(\frac1{\sqrt{x}\log x}\right). } ]
Instead of estimating (P_1) externally, estimate it from lower layers.
From the saddle law:
[ P_1\log P_1 \approx P_2^2\log P_2. ]
Define internal scale (X_2):
P_2^2\log P_2. ]
Then:
[ |P_1-X_2| ]
is controlled by local frontier spacing.
This replaces global Chebyshev control with local saddle-frontier control.
Key target inequality:
[ P_1-\vartheta(P_1) < \sum_{j\ge2}\vartheta(P_j). ]
Interpretation:
- any first-layer prime deficit is repaid by forced lower-layer mass.
Using saddle asymptotics:
[ \sum_{j\ge2}\vartheta(P_j) \sim \sqrt{2P_1}. ]
This is one of the central remaining proof bottlenecks.
The strongest current theorem candidate:
[ \boxed{ \sum_{m\in W}(\Delta\mathcal L_m)^- \le \sum_{m\in W}(\Delta\mathcal L_m)^+ + \Phi(x_0)-\Phi(x_1) + E_W. } ]
where:
- (W) is a CA-event window,
- (E_W) is the discrete frontier error.
If:
[ \sum_W E_W<\infty, ]
then finite verification plus telescoping would imply Robin globally.
Observed numerically:
- (\Lambda(x)>0) through tested ranges,
- local monotonicity fails,
- but windowed telescoping appears to hold,
- all-layer barrier dominates observed oscillation.
Representative values:
| (P_1) | (\Lambda) |
|---|---|
| (10^3) | (\sim0.9) |
| (10^5) | (\sim0.82) |
| (10^6) | (\sim0.79) |
Window tests up to (10^6) showed:
[ \sum D^- \le \sum D^+ + \Delta\Phi. ]
No failures observed.
The current best candidate proof route is:
- Exact CA event formulation.
- Continuous frontier telescoping theorem.
- Summable discrete frontier error bound.
- Finite verification.
- Global Robin positivity.
- RH.
The key remaining theorem is likely:
o(\mathcal L(x)). } ]
combined with rigorous continuous telescoping.
Nothing here is yet a proof of RH.
The continuous model appears structurally coherent, but the discrete prime-event step still needs rigorous closure.
The current status is:
- conceptual framework: strong,
- asymptotic structure: strong,
- numerical support: promising,
- rigorous closure: incomplete.
- Prove continuous telescoping rigorously.
- Derive exact Stieltjes frontier formulation.
- Prove summable discretization error.
- Formalize local no-crossing theorem.
- Push CA event simulations much further.
- Test windowed telescoping at large scales.
- Measure discrete frontier error directly.
- Study semiprime tie events separately.
- Formalize event ordering.
- Formalize saddle scaling law.
- Formalize continuous frontier potential.
- Formalize discrete-to-continuous error estimates.