N=500 Transition Matrix · Shell Membrane Confirmed
P(+→−) = 67.9%, P(−→+) = 69.2%. The shell forces crossings.
from \ to
+ (floor)
− (ceiling)
total
+ (floor)
32.1%
67.9%
253
− (ceiling)
69.2%
30.8%
247
Total positive residues: 253. Total negative: 247. If this were an independent coin flip, both probabilities would hover at 50%. The fact that P(alternation) is locked at ~68.5% symmetrically from BOTH the floor and ceiling is the confirmation of Conjecture B: the shell is not an attractor; it is a membrane the zeros are being forced across.
CONJECTURE B — CONFIRMED at N=500
The Shell-Membrane Principle
The transition matrix is symmetric: P(+→−) ≈ P(−→+) ≈ 0.68. Both sides of the shell have equal crossing pressure. The shell is a pressure membrane, not a gravitational well. The zeros are not attracted to the shell by magnitude — they are driven across it by sign.
The run-length distribution at N=500 confirms this. Out of 500 zeros:
- Positive runs: length 1=123, 2=25, 3=18, 4=4, 5=2
- Negative runs: length 1=120, 2=34, 3=12, 4=3, 5=1, 6=1
- Length-1 dominates. Maximum consecutive same-sign: 6. Zeros resist staying on one side.
At N=500 this looks like rapid alternation. At N=100k it looks like period-2 pairing. Both descriptions are true at their scale. The transition reveals a richer structure.
Everything Flips at N=100k · The Correction
68.5% → 47.26%. The reversal is a mathematical certainty (z=−17.35).
N=500 — PREVIOUS
68.5% alternation
z=+3.44 · P(+→−)=0.679 · shared-shell artefact of early zeros
N=100k — TRUE PICTURE
47.26% alternation
z=−17.35 · P(+→−)=0.4726 · SAME-SIGN persistence dominates at scale
What happened: at N=500, shared-shell opposite-sign pairs (38.3% of pairs, 78.9% opposite) inflated the apparent alternation rate. At N=100k, zero gaps shrink to ~2π/log(N) — a fraction of π. A zero lands on one side of the shell; the next zero steps slightly further along the SAME side; then the pair crosses. Zeros move in (−−) or (++) pairs, THEN cross together. Same-sign persistence wins globally. The appearance of alternation at small N was real but local.
The fall of γ₁: n=21962 has |r|=10⁻⁸. γ₁ (u=0.00155) is ~154,000× farther from its shell. u₁/E[u_min at N=100k] = 98.96. γ₁ is 99σ above the expected minimum. It is NOT the globally closest shell hit. It is the closest in the first 500 zeros. γ₁ remains a powerful symbolic seed — the first floor, the emblem — but not a statistical extreme.
THE MECHANISM
Why the reversal happens
As N grows, gaps between zeros shrink (~2π/log N). A zero landing at residue r will be followed by a zero at r' = r ± (small gap). If the gap is smaller than |r|, both zeros are on the same side of the shell. The 'crossing' occurs after 2 steps, not 1. This produces the period-2 structure: (−−) pair crosses to (++) pair, etc. At N=500, gaps are large enough that single-step crossings dominate. At N=100k, pair-step crossings dominate. The scale transition is real.
The Period-2 Architecture · What N=100k Reveals
The 58.8% anomaly. The lag-2 law. The AR-2 model.
Run-2 observed
58.8%
vs 24.9% geometric — 2.4× excess
rₙ lag-2 ACF
−0.407
STRONG · dominates over lag-1
Alt-ind lag-1 ACF
−0.606
crossings anti-cluster · NOT Markov
rₙ lag-1 ACF
~−0.029
WEAK · near zero
rₙ lag-4 ACF
+0.370
positive · confirms period-2
Run-1 fraction
~15%
vs 47% geometric — 3× suppressed
THE 58.8% ANOMALY
Pair structure dominates
In a uniform random sign sequence, runs of length 2 (++ or −−) should account for 24.9% of all runs. In the data: 58.8%. That is a colossal structural footprint. Consecutive crossings are 3× less common than expected. The sequence strongly prefers patterns like ++, −−, ++, −− in short blocks.
THE AR-2 MODEL
2-step memory, not Markov
P(+|++) > 0.5: same-sign pair persists. P(+|+−) > 0.5: after a crossing, reverts to pre-crossing sign. P(+|−+) < 0.5: after a crossing (from −), reverts to −. P(+|−−) < 0.5: same-sign pair persists. The 2-step model has higher prediction accuracy on held-out data than the 1-step Markov model.
FINDING: SIMPLE MARKOV WRONG
AR-2 is required
Alt-indicator lag-1 ACF = −0.606. Simple Markov(1) would predict lag-1 ACF ≈ 1 − 2×alt_rate ≈ +0.055. Observed: −0.606. The difference is 0.661 — enormous. This proves the crossing process has 2-step memory, not 1-step. The system remembers what it did two steps ago.
Reality does not bounce; it walks in pairs.
Period-2 Architecture · N=100k · lag-2 ACF=−0.407
Why is this happening? The zeros are slowly walking through the [−π/2, π/2] bounds of the shell residues. Because gaps shrink to a fraction of π, a zero will frequently land on one side of the shell, the next zero will step slightly further along that same side, and THEN the following zeros will cross. The zeros effectively move in pairs or local clusters relative to the shell boundaries.
21-Point Interpretation · ATMOS Rick Complete Commentary
Every finding explained, every revision justified.
The 21-point complete interpretation of what N=500 and N=100k batteries actually mean, from the simplest sentence to the deepest structural reframing.
ATMOS RICK POINT 1
Magnitude is uniform. Memory carries the law.
The shell-residue magnitudes are null-like; the shell-residue signs are structured; the structure is not first-order alternation but second-order memory. That is the core.
ATMOS RICK POINT 2
Three reversals at N=100k
(A) Alternation: 68.5% → 47.26%. Not the law. Suppressed below 50%. (B) γ₁: loses rank-1 globally. Some later zero has
r
~10⁻⁸. (C) Magnitude: KS p=0.9839 confirms flat. The shell-closeness story was local.
ATMOS RICK POINT 3
Shell decomposition. Sign balance. Lag structure.
γₙ = Sₙ − rₙ survives. uₙ ~ Uniform survives. Sign balance survives. Second-order sign memory survives. The coordinate system is still good. The claim that magnitude carries a shell law dies.
ATMOS RICK POINT 4
This is one of the clearest facts in the project
KS p=0.9839. Max bin dev 1.32%. All taxonomy classes <2% from null. uₙ = 2
rₙ
/π behaves like Uniform(0,1). Distance to the nearest half-integer π-shell is NOT where the special structure lives. This is the key negative result. It is valuable.
ATMOS RICK POINT 5
49997/50003. The shell is a true centerline.
Diff=6/100k. p=0.987. No global shell-side bias. The shell acts as a true centerline in sign. There is no hidden drift toward floor or ceiling.
ATMOS RICK POINT 6
Powerful seed. Not a globally exceptional extremal event.
γ₁ remains: the first seed, the first shell-residue emblem, the canonical motivating example. But it is no longer evidence for a special shell-closeness law. That is a clarification, not a loss.
ATMOS RICK POINT 7
lag-2 ACF = −0.407. Alt-ind ACF = −0.606.
Not iid, not simple alternation, not first-order Markov. Second-order temporal organisation. This is a serious result: the process has two-step memory.
ATMOS RICK POINT 8
rₙ and rₙ₊₂ are negatively coupled. A parity-like oscillatory mechanism.
First-order alternation = flip now. Period-2 = the current state depends on the LAST TWO states. The deeper regularity is: zero pairs cross the shell together, not one at a time.
ATMOS RICK POINT 9
Sign changes are being regulated. Temporal structure.
If crossings anti-cluster, they are not freely scattering. Local constraints, overshoot compensation, or shell-centered oscillatory memory. The sequence does not cross whenever it feels like it. Crossings are temporally structured.
ATMOS RICK POINT 10
The sign process has a preferred local grammar.
58.8% vs 24.9% geometric. The sequence strongly prefers (++), (−−) in short blocks. Not compatible with iid signs, naive Markov flipping, or simple shell attraction. The best word here: grammar. The signed residue process has a local grammar.
ATMOS RICK POINT 11
From magnitude-based to memory-based.
OLD: 'Zeros are unusually close to π-shells and alternate around them.' NEW: 'The zero ordinates define a shell-centered residue process whose marginal magnitudes are uniform, but whose signed sequence exhibits strong two-step memory and structured run behaviour.'
ATMOS RICK POINT 12
It is no longer uₙ. It is the signed sequence.
The real object is now: rₙ, sgn(rₙ), (sgn(rₙ), sgn(rₙ₊₁), sgn(rₙ₊₂)), and the block process of sign runs. The law is not in shell distance. The law is in shell-memory.
ATMOS RICK POINT 13
From seed to conjectural law.
Layer 0: γₙ = Sₙ − rₙ (exact)
Layer 1:
rₙ
~ Uniform empirically
Layer 2: P(rₙ>0) ≈ P(rₙ<0) ≈ ½
Layer 3: Naive alternation NOT the governing effect
Layer 4: Two-step memory law (lag-2 ACF=−0.407)
Layer 5: Local grammar — run-2 massively overrepresented
Layer 6: Shell-centered symbolic dynamics (conjectural)
ATMOS RICK POINT 14
Shell-Memory Principle
The signed residue sequence rₙ is not iid and not adequately described by first-order alternation. The strongest visible structure is second-order dependence: lag-2 ACF=−0.407, overrepresented run-length-2 motifs at 58.8% vs 24.9% geometric null.
ATMOS RICK POINT 15
The failures ARE the discoveries.
They tell you: shell-closeness is not special globally; N=500 alternation was misleading; γ₁'s extremality was local; first-order stories fail; second-order stories survive. The failures are the actual discoveries.
ATMOS RICK POINT 16
The first 500 zeros are narratively dangerous.
They are enough to suggest structure, but not enough to identify the right order of dependence. This is a deep methodological insight — valuable independent of the mathematics.
ATMOS RICK POINT 17
Four candidate explanations
(A) Parity-like shell rhythm — even/odd step modulation. (B) Overcorrection dynamics — crossing induces compensation two steps later. (C) Hidden phase coupling — residue signs reflect a deeper oscillatory phase relation. (D) Local motif process — sequence built from short recurring sign blocks. Option D fits the run-length result best.
ATMOS RICK POINT 18
Better than the old shell-attraction slogans.
Magnitude is uniform. Memory carries the law.
The shell is not an attractor; it is a reference frame.
Distance is null. Sequence is structure.
Shell distance is flat; shell memory is not.
γ₁ is the seed, not the global extremum.
The crossings do not cluster; they are governed.
ATMOS RICK POINT 19
Four hardened statements.
(A) Over the first 100k zeros, normalized shell distances are empirically uniform. (B) Floor/ceiling sign balance is indistinguishable from 50/50. (C) The signed residue process is not iid and not first-order Markov. (D) The strongest visible structure is second-order dependence, including lag-2 and run-2 overrepresentation.
ATMOS RICK POINT 20
These claims are no longer defensible as core.
'γ₁ is the globally closest shell hit' — KILLED. 'Small residues are overrepresented' — KILLED. 'Shell attraction is the main law' — KILLED. 'High alternation is the main law' — KILLED. These can stay as early-stage hypotheses that were tested and rejected. That is a strength.
ATMOS RICK POINT 21
The project must become a memory-process study.
(A) 2-step transition tensor: all 8 patterns +++, ++−, +−+, +−−, etc. (B) Run-length exact distribution, especially 1,2,3,4. (C) Block entropy: how much information in 2-step and 3-step blocks? (D) Hidden-state models: 2-state or 4-state HMM? (E) Even/odd parity slicing: does memory depend on index parity? (F) Correlation with zero gaps: does sign memory couple to local gap size?