I am not an analytic number theorist. My background is in cryptographic state-space modeling and low-level systems architecture — specifically Cyclic Digit-sum Projection (CDP), a framework of 130+ theorems I developed to analyze carry trees, zero-masking barriers, and bit-weight conservation in ARX (Add-Rotate-Xor) primitives like SHA-256.

Recently, I applied that exact cryptographic lens to the accelerated Collatz (Syracuse) operator ($3n+1$). Instead of treating trajectories as abstract scalar magnitudes, I modeled the transformation directly as a synchronous Binary Full-Adder State Automaton.

The formal mathematical preprint is archived on Zenodo (DOI: 10.5281/zenodo.24311418), building on my foundational CDP v5 framework (DOI: 10.5281/zenodo.22690410).

This post presents the complete CDP–Collatz Master System — an exact, zero-approximation four-part algebraic characterization of how bit rooms, Hamming weights, and carry waves evolve at every step, together with the deterministic bridge identity, the signed corridor decomposition, and non-tautological sliding-window diagnostics.


1. Beyond Base-10 Magnitude: The Binary Full-Adder Engine

The accelerated Collatz rule on positive odd integers $n \in \mathbb{N}_{\text{odd}}$ is simple: multiply $n$ by $3$, add $1$, and divide out the highest power of $2$ so that you land on the next odd integer:

$$ T(n) = \frac{3n + 1}{2^{\delta(n)}}, \qquad \delta(n) = v_2(3n + 1) $$

When you watch a trajectory like $n = 27$ climb to $9232$ before collapsing to $1$ in decimal (base-10), the sequence looks erratic. Decimal representation hides the structural invariants of binary addition because base-10 digits do not align with powers of $2$.

In binary hardware, there is no abstract “multiplication by 3.” Multiplying an odd integer $n$ by $3$ and adding $1$ is identically a single-bit left shift coupled with a binary addition:

$$ 3n + 1 = (2n + 1) + n = ((n \ll 1) \lor 1) + n $$

And dividing by $2^{\delta(n)}$ is a bitwise right-shift ($\gg \delta(n)$) that strips trailing zeros (000...) from the least significant end of the register.

Viewed at the gate level, the Collatz map is a Binary Full-Adder Bit-Flush Engine. Every iteration injects a root carry at the bottom of the register, triggering an internal carry wave that collides with existing bit pairs, modifies the internal Hamming weight, and flushes trailing zeros.


2. Anatomy of Trajectory Growth: Why $\delta(n) = 1$ Cannot Persist Unchecked

In heuristic binary drift arguments, each odd step multiplies the number by $3$ (adding $\log_2(3) \approx 1.585$ bits of length) while stripping an average of $\mathbb{E}[\delta] = 2$ trailing zeros over uniform odd inputs, yielding a net contractive drift of $-0.415$ bits per step.

Why isn’t the heuristic drift argument a rigorous proof for all $n$? Because a global proof must account for adversarial bit configurations that repeatedly force the minimum shift $\delta(n) = 1$ (where $3n+1$ is divisible by $2$ only once), causing the trajectory to climb.

In binary full-adder space, what does it actually take to force $\delta(n) = 1$?

  • As we prove in Section 4, getting a single trailing zero ($\delta(n) = 1$) occurs if and only if the two lowest bits of $n$ are 11 ($b_1 = 1, b_0 = 1$).
  • Sustaining $\delta = 1$ for $k$ consecutive steps requires stacking a contiguous block of $k+1$ ones at the bottom of the number (...0111...11_2).
  • However, every finite integer $n$ has a finite bit-length and must contain a first 0 bit ($b_{k+1} = 0$) bounding the lower 11...1 block. Each step erodes this lower 11...1 block by exactly one bit ($k \mapsto k - 1$) as it converts $2^{k+1} - 1 \mapsto 3 \cdot 2^k - 1$, until the boundary 0 bit reaches index $1$ and forces a multi-bit flush ($\delta \ge 2$).

Thus, a trajectory can never climb forever using a single static block of ones. To sustain growth across a long orbit, the upper bits of the number must continually synthesize new 11...1 blocks faster than lower-corridor flushes can erode them. To analyze how such regeneration occurs, we first decouple the carry tree of $(2n+1)+n$.


3. Carry-Tree Decoupling: Orthogonal Bit-Room Masks

A natural systems question arises when moving from fixed-width cryptography (like 32-bit words in SHA-256) to arbitrary-precision integers: In Collatz, an integer can be millions of bits long. Can a carry wave originating at bit $0$ ripple uncontrollably across the entire bit-length of $n$?

CDP Theorem 116 (Carry-Kill Bulkhead Law) proves that it cannot.

Let any odd integer be written in binary as $n = \sum_{i=0}^{L} b_i 2^i$ (with $b_0 = 1$, leading bit $b_L = 1$, $b_i = 0$ for $i > L$, boundary condition $b_{-1} = 0$, and initial root carry $c_0 = 1$). By inspecting adjacent bit pairs $(b_i, b_{i-1})$ at every index $i \ge 0$, we project the number into three mutually orthogonal Bit-Room Masks:

$$ U_i = b_i \oplus b_{i-1} \qquad \text{(Open Propagate Room: } \mathtt{01} \text{ or } \mathtt{10}\text{)} $$

$$ G_i = b_i \land b_{i-1} \qquad \text{(Carry-Generate Wall: } \mathtt{11}\text{)} $$

$$ Z_i = \neg b_i \land \neg b_{i-1} \qquad \text{(Zero-Absorb Wall: } \mathtt{00}\text{)} $$

Intuitive Hardware Picture (Corridors and Bulkheads)

Think of the binary representation of $n$ as a corridor partitioned into rooms, with the carry bit $c_i$ flowing from right to left (from LSB to MSB):

  • Open Propagate Rooms ($U_i = 1$): Whenever adjacent bits differ (01 or 10), the room is open. An incoming carry $c_i = 1$ passes straight through to room $i+1$ while flipping the room’s sum bit $s_i$ to 0.
  • Bulkhead Walls ($U_i = 0$): Whenever adjacent bits are equal (00 or 11), $U_i = 0$. This forms an impenetrable bulkhead. Carry information from lower rooms can never pass through a 00 or 11 wall. A 00 wall ($Z_i = 1$) absorbs the incoming carry and outputs $c_{i+1} = 0$. A 11 wall ($G_i = 1$) blocks the incoming carry and initiates a fresh carry $c_{i+1} = 1$ of its own.

Formal Algebraic Formulation (Partition of Unity)

At every bit index $i \ge 0$, the masks $U_i, G_i, Z_i \in {0, 1}$ are pairwise disjoint and partition unity:

$$ U_i + G_i + Z_i \equiv 1, \qquad U_i \land G_i \equiv G_i \land Z_i \equiv Z_i \land U_i \equiv 0 $$

Furthermore, because $b_i + b_{i-1} = U_i + 2G_i$, summing over all $i \ge 0$ yields the exact Hamming weight relation:

$$ 2w(n) = w(U) + 2w(G) \qquad \text{(Lemma 1)} $$

Under the full-adder decomposition $3n + 1 = (2n + 1) + n = \sum_{i \ge 0} s_i 2^i$, the sum bit $s_i$ and carry-pipe bit $c_{i+1}$ satisfy:

$$ s_i = U_i \oplus c_i, \qquad c_{i+1} = \operatorname{Maj}(b_i, b_{i-1}, c_i) = G_i + (U_i \land c_i) \equiv G_i \oplus (U_i \land c_i) $$

Because $U_i \land G_i \equiv 0$, the carry transition $c_{i+1} = G_i + (U_i \land c_i)$ produces zero secondary carries, establishing an exact linear equivalence between Boolean ($\mathbb{F}_2$) logic and integer arithmetic. At any wall index $r$ where $U_r = 0$, the propagation term $(U_r \land c_r)$ vanishes identically:

$$ U_r = 0 \implies c_{r+1} = G_r = b_r \qquad \text{(CDP Theorem 116)} $$

Every 00 and 11 pair completely severs the carry tree into independent, finite segments.


4. The Zero-Search Flush Law: Exact $\delta(n)$ Without Multiplication

At the root room $i = 0$, the boundary conditions $b_0 = 1, b_{-1} = 0, c_0 = 1$ lock $U_0 = 1$ and $G_0 = 0$, producing:

$$ s_0 = U_0 \oplus c_0 = 1 \oplus 1 = 0, \qquad c_1 = G_0 + (U_0 \land c_0) = 1 $$

The injected carry $c_1 = 1$ propagates unattenuated through every consecutive open room $U_i = 1$ ($1 \le i < v$), annihilating each sum bit into zero ($s_i = 1 \oplus 1 = 0, ; c_{i+1} = 1$) until it collides with the very first wall room $U_v = 0$, where it terminates as $s_v = 0 \oplus 1 = 1$.

This establishes CDP Theorems 46 & 99 (The Zero-Search Lower-Corridor Flush Law): the exact 2-adic valuation $\delta(n) = v_2(3n+1)$ is determined with zero multiplication by the index of the first identical adjacent bit pair:

$$ \delta(n) = v_2(3n + 1) = \min \left{ i \ge 1 : U_i = 0 \right} = \min \left{ i \ge 1 : b_i = b_{i-1} \right} $$

  • Example 1 ($n = 5 = \mathtt{00101}_2$): Scanning from LSB to MSB (1, 0, 1, 0, 0), the bits alternate until the first 00 pair at index $i = 4$. Thus $\delta(5) = 4$ ($3(5)+1 = 16 = 2^4$).
  • Example 2 ($n = 13 = \mathtt{01101}_2$): Scanning from LSB to MSB (1, 0, 1, 1), the first identical pair 11 occurs at index $i = 3$. Thus $\delta(13) = 3$ ($3(13)+1 = 40 = 2^3 \times 5$).

5. The Unified CDP–Collatz Master Equation & Bridge Identity

By combining the Zero-Search Flush Law with CDP Theorem 124 (Weight-Carry Conservation) and CDP Theorem 136 (Carry-Free Multiplexer Evolution), we lock the step-to-step evolution ($t \mapsto t+1$) of any odd Collatz state into an exact four-part closed-form system:

  1. Exact Lower-Corridor Flush (Zero-Search Law): $$ \delta(n_t) = \min\left{ i \ge 1 : U_i^{(t)} = 0 \right} $$

  2. Global Weight-Carry Conservation (CDP Theorem 124): Because right-shifting by $\delta(n_t)$ strips only trailing zeros, $w(n_{t+1}) = w(3n_t+1)$. Summing the single-bit full-adder identity $s_i + c_{i+1} - c_i = U_i^{(t)} + G_i^{(t)} - \left(U_i^{(t)} \land c_i^{(t)}\right)$ across the upper corridor $i > \delta(n_t)$ telescopes the carry chain to the bulkhead boundary $c_{\delta+1}^{(t)} = G_{\delta(n_t)}^{(t)}$ (by Theorem 116) plus the sentinel bit $s_{\delta(n_t)} = 1$: $$ w(n_{t+1}) = w(3n_t + 1) = 1 + G_{\delta(n_t)}^{(t)} + \sum_{i=\delta(n_t)+1}^{L_t} U_i^{(t)} + \sum_{i=\delta(n_t)+1}^{L_t} G_i^{(t)} - \sum_{i=\delta(n_t)+1}^{L_t} \left(U_i^{(t)} \land c_i^{(t)}\right) $$ Splitting the upper open rooms into uncarried open rooms $U_{0,\text{upper}}^{(t)} = \sum_{i>\delta} (U_i^{(t)} \land \neg c_i^{(t)})$ and carried (vacuumed) open rooms $U_{1,\text{upper}}^{(t)} = \sum_{i>\delta} (U_i^{(t)} \land c_i^{(t)})$ yields the simplified positive-definite form: $$ w(n_{t+1}) = 1 + G_{\delta(n_t)}^{(t)} + G_{\text{upper}}^{(t)} + U_{0,\text{upper}}^{(t)} $$

  3. Shifted Orthogonal Room Evolution ($\forall m \ge 1, ; i = m + \delta(n_t)$): $$ G_m^{(t+1)} = \underbrace{\left(U_i^{(t)} \land U_{i-1}^{(t)} \land \neg c_{i-1}^{(t)}\right)}{\text{Channel 1: Carry-Free Corridor}} + \underbrace{\left(U_i^{(t)} \land Z{i-1}^{(t)} \land c_{i-1}^{(t)}\right)}{\text{Channel 2: Wall Collision}} + \underbrace{\left(\neg U_i^{(t)} \land G{i-1}^{(t)} \land c_{i-1}^{(t)}\right)}_{\text{Channel 3: Block Survival}} $$

    $$ Z_m^{(t+1)} = \left(U_i^{(t)} \land U_{i-1}^{(t)} \land c_{i-1}^{(t)}\right) + \left(\neg U_i^{(t)} \land Z_{i-1}^{(t)} \land \neg c_{i-1}^{(t)}\right) + \left(U_i^{(t)} \land G_{i-1}^{(t)} \land \neg c_{i-1}^{(t)}\right) $$

    $$ U_m^{(t+1)} = 1 - G_m^{(t+1)} - Z_m^{(t+1)} $$

  4. Telescopic Trajectory Identity ($N$ Odd Steps): $$ n_N = \frac{3^N n_0 + \sum_{j=1}^{N} 3^{N-j} 2^{\Sigma_{j-1}}}{2^{\Sigma_N}}, \qquad \Sigma_N = \sum_{t=0}^{N-1} \min\left{ i \ge 1 : U_i^{(t)} = 0 \right} $$

Corollary: Bridge Identity, Signed Corridor Decomposition & Cycle Condition

Combining Lemma 1 ($2w(n_t) = \delta(n_t) + U_{0,\text{upper}}^{(t)} + U_{1,\text{upper}}^{(t)} + 2G_{\delta(n_t)}^{(t)} + 2G_{\text{upper}}^{(t)}$) with Equation (2) eliminates the generator counts $G$ entirely and links the lower-corridor shift $\delta(n_t)$ directly to the upper open-room imbalance and Hamming weight change $\Delta w_t = w(n_{t+1}) - w(n_t)$:

$$ \delta(n_t) = 2 + \left(U_{0,\text{upper}}^{(t)} - U_{1,\text{upper}}^{(t)}\right) - 2\Delta w_t $$

Furthermore, by Theorem 116 ($c_{r+1} = b_r$ at every wall $U_r = 0$), the carry bit $c_i$ is constant across each contiguous open corridor $k \in \mathcal{C}{\text{upper}}(n_t)$ of length $L_k$ above index $\delta(n_t)$ and equals the bit value $b{r_k} \in {0, 1}$ of the wall immediately beneath it. This yields the exact geometric decomposition:

$$ U_{1,\text{upper}}^{(t)} - U_{0,\text{upper}}^{(t)} = \sum_{k \in \mathcal{C}{\text{upper}}(n_t)} (2b{r_k} - 1),L_k $$

In plain terms: open corridors sitting above 11 walls contribute $+L_k$, while open corridors sitting above 00 walls contribute $-L_k$.

Summing the Bridge Identity over any periodic cycle of length $N$ ($n_N = n_0$, where $\sum \Delta w_t = 0$ and $\Sigma_N / N = \log_2 3 + \epsilon_N$) yields the exact necessary condition for any cycle:

$$ \langle U_{1,\text{upper}} - U_{0,\text{upper}} \rangle_{\text{cycle}} = 2 - \log_2 3 - \epsilon_N, \qquad \epsilon_N = \frac{1}{N} \sum_{j=0}^{N-1} \log_2!\left(1 + \frac{1}{3n_j}\right) > 0 $$

At the unique known cycle $n_0 = 1$, $\epsilon_1 = \log_2(4/3) = 2 - \log_2 3 \approx 0.41504$, giving $\langle U_{1,\text{upper}} - U_{0,\text{upper}} \rangle = 0$ identically.


6. Channel Dynamics and the Adversarial Family $n_k = 1 + 8(4^k-1)/3$

Equation (3) decomposes the creation and survival of 11 carry-generating walls ($G_m^{(t+1)} = 1$) into three mutually exclusive Boolean channels:

  1. Channel 3 (Block Survival & Mandatory Boundary Erosion): For an existing contiguous 11 block of length $L$ to survive at index $i$, it requires both $G_{i-1}^{(t)} = 1$ and $c_{i-1}^{(t)} = 1$. At the lower boundary of any isolated 11 block, the preceding room has $G_{i-1}^{(t)} = 0$; at the upper boundary where the block meets an open room ($U_{k+1}^{(t)} = 1$), $\neg U_{k+1}^{(t)} = 0$. Consequently, existing 11 blocks cannot self-expand via Channel 3 and erode by at least one room per step ($L \mapsto L - 1$).

  2. Pure Channel 2 Isolation ($\text{Ch2}m \land \text{Ch2}{m-1} \equiv 0$): Two adjacent rooms cannot both become 11 walls purely via Channel 2 carry-wall collisions, because index $i-1$ cannot simultaneously satisfy $Z_{i-1}^{(t)} = 1$ (required for $\text{Ch2}m$) and $U{i-1}^{(t)} = 1$ (required for $\text{Ch2}{m-1}$): $$ G{m,\text{Ch2}}^{(t+1)} \land G_{m-1,\text{Ch2}}^{(t+1)} \equiv 0 $$

  3. Cross-Channel Cooperativity ($\text{Ch2} + \text{Ch1}^*$ Synthesis & Inductive Erosion Proof): While pure Channel 2 births are strictly isolated, contiguous $G^{(t+1)}$ blocks of arbitrary single-step length $2k$ are synthesized when Channel 2 couples with Channel 1 across a 00 bulkhead ($Z_{i-1}^{(t)} = 1$) positioned directly beneath an alternating open propagate corridor of length $2k$ ($U_i^{(t)} = \dots = U_{i+2k-1}^{(t)} = 1$). Specifically, for all $k \ge 1$, the seed family $$ n_k = 1 + \frac{8(4^k - 1)}{3} \quad (\mathtt{0101\dots01001}_2) $$ contains zero initial 11 walls ($G^{(t)} = 0$ everywhere), yet satisfies $3n_k + 1 = 4(2^{2k+1} - 1)$, converting its upper alternating corridor into $x_0 = 2^{2k+1} - 1$ in a single $\delta = 2$ step via $\text{Ch2} + \text{Ch1}^{2k-1}$.

    Inductive Erosion Proof ($\forall k \ge 1$): Define $x_j = 3^j \cdot 2^{2k+1-j} - 1$ for $0 \le j \le 2k$. Then $3x_j + 1 = 2(3^{j+1} \cdot 2^{2k-j} - 1) = 2x_{j+1}$. For $0 \le j \le 2k - 1$, $x_{j+1}$ is odd, forcing $\delta(x_j) = 1$ for $2k$ consecutive steps until $x_{2k} = 2 \cdot 3^{2k} - 1$. At the next step, $3x_{2k} + 1 = 2(3^{2k+1} - 1)$; because $3^{2k+1} - 1 \equiv 2 \pmod 4$, $\delta(x_{2k}) = 2$, yielding the exact multi-step transition for all $k \ge 1$:

$$ n_k = 1 + \frac{8(4^k - 1)}{3} \xrightarrow{\delta=2} 2^{2k+1} - 1 \xrightarrow{2k \text{ steps of } \delta=1} 2 \cdot 3^{2k} - 1 \xrightarrow{\delta=2} \frac{3^{2k+1} - 1}{2} $$

$k$ Seed $n_k$ Step 1 Output ($2^{2k+1}-1$) After $2k$ Steps ($2 \cdot 3^{2k}-1$) Next Step ($(3^{2k+1}-1)/2, ; \delta=2$)
$1$ $9$ $7$ $17$ $13$
$2$ $41$ $31$ $161$ $121$
$3$ $169$ $127$ $1457$ $1093$
$4$ $681$ $511$ $13121$ $9841$
$5$ $2729$ $2047$ $118097$ $88573$
$6$ $10921$ $8191$ $1062881$ $797161$
$7$ $43689$ $32767$ $9565937$ $7174453$

7. Conditional Framework and Reproducible Diagnostics

What does this four-part characterization prove unconditionally, and how does it formulate the open global question?

  1. Unconditional Identities vs. Drift Regimes: Equations (1)–(4), the Deterministic Bridge Identity, the Signed Corridor Decomposition, the bulkhead decoupling law ($U_r = 0 \implies c_{r+1} = G_r$), pure Channel 2 isolation ($\text{Ch2}m \land \text{Ch2}{m-1} \equiv 0$), and the inductive erosion of $n_k$ hold unconditionally for all odd integers. Single-step bulkhead decoupling also holds in generalized $(2^k+1)n+1$ adders such as $5n+1$ ($k=2$), which nonetheless admits cycles ($13 \to 33 \to 83 \to 13$) because its bit growth ($\log_2 5 \approx 2.322$) exceeds $\mathbb{E}[\delta] = 2$.
  2. Avoiding the Convergent-Orbit Tautology: By Equation (4), $2^{\Sigma_N} n_N > 3^N n_0$. Any trajectory that reaches $n_N = 1$ in $N$ steps automatically obeys $\langle \delta \rangle = \Sigma_N / N > \log_2 3 + (\log_2 n_0)/N > 1.5850$. Measuring $\langle \delta \rangle > \log_2 3$ on completed orbits is therefore a tautological consequence of convergence.
  3. Sliding-Window Diagnostic $W_{\max}(B)$ ($B = 4 \dots 14$): Without assuming convergence, we measure the longest contiguous window $W_{\max}(B)$ along any trajectory starting in $2^{B-1} < n_0 < 2^B$ over which the window-averaged flush remains strictly expansive ($\sum_{j=t}^{t+W-1} \delta(n_j) < W \log_2 3$), alongside the maximum odd stopping time $N_{\max}(B)$. The observed growth of $W_{\max}(B)$ across $B \in [4, 14]$ is consistent with the $O(B)$ scaling predicted by standard geometric random-walk models:
Bit-Length $B$ Worst Seed $n_0$ Static $2^B-1$ Window Max Expansive Window $W_{\max}(B)$ Max Odd Steps $N_{\max}(B)$ Ratio $W_{\max}(B) / B$
$4$ $9$ $3$ $4$ $6$ $1.000$
$5$ $27$ $37$ $39$ $41$ $7.800$
$6$ $41$ $33$ $38$ $41$ $6.333$
$7$ $109$ $8$ $38$ $43$ $5.429$
$8$ $231$ $7$ $40$ $46$ $5.000$
$9$ $327$ $10$ $45$ $52$ $5.000$
$10$ $871$ $10$ $56$ $65$ $5.600$
$11$ $1161$ $35$ $57$ $66$ $5.182$
$12$ $2919$ $33$ $67$ $87$ $5.583$
$13$ $6919$ $34$ $68$ $96$ $5.231$
$14$ $9225$ $31$ $69$ $101$ $4.929$
  1. Per-Room Normalized Densities (100% Reproducible via Companion Script): Normalizing cross-channel synthesis $\rho_{\text{synth}} = (\text{Ch2}+\text{Ch1}^*)/L_t$, carry vacuuming $\rho_{\text{vac}} = U_{1,\text{upper}}/L_t$, and uncarried open rooms $\rho_{U_0} = U_{0,\text{upper}}/L_t$ by active bit-length $L_t = \lfloor \log_2 n_t \rfloor + 1$ allows direct comparison against the analytic i.i.d. Bernoulli($1/2$) baseline ($\rho_{\text{synth}}^{\text{iid}} = 1/12 \approx 0.0833$, $\rho_{\text{vac}}^{\text{iid}} = \rho_{U_0}^{\text{iid}} = 1/4 = 0.2500$, with finite-$L$ boundary offset $\mathbb{E}[U_{1,\text{upper}} - U_{0,\text{upper}}] \approx 2 - \log_2 3 \approx 0.415$ and asymptotic weight drift $\mathbb{E}[\Delta w] \approx \frac{1}{2}(\log_2 3 - 2) \approx -0.2075$):
Evaluated Regime / Scale Mean $\langle \delta \rangle$ Mean $\rho_{\text{synth}}$ Mean $\rho_{\text{vac}}$ ($U_1/L$) Mean $\rho_{U_0}$ ($U_0/L$) Mean $\langle \Delta w \rangle$
Asymptotic i.i.d. Theory ($L \to \infty$) $2.0000$ $0.0833$ $0.2500$ $0.2500$ $-0.2075$
Random $256$-Bit Odd Sample ($N = 2,000$, seed $= 42$) $1.9690$ $0.0829$ $0.2503$ $0.2481$ $-0.2840$
Exact $13$-Bit Odd Set ($L$-Matched Baseline, $N = 2,048$) $2.0005$ $0.0897$ $0.2564$ $0.2051$ $-0.3335$
Full Orbits of All $13$-Bit Odd Seeds ($N = 66,061$ steps) $1.9812$ $0.0838$ $0.2489$ $0.1923$ $-0.2015$
Conditioned on $W_{\max}$ Windows ($B=10\dots14$, $N = 317$ steps) $1.5647$ $0.1022$ $0.2649$ $0.2241$ $-0.0063$

Exact Window Equivalence of the Open Convergence Problem

By the Bridge Identity, over any window of $W$ consecutive steps, the window-averaged flush rate satisfies:

$$ \langle U_{1,\text{upper}} - U_{0,\text{upper}} \rangle_W + 2\langle \Delta w \rangle_W = 2 - \langle \delta \rangle_W $$

Notice why retaining the weight-drift term $2\langle \Delta w \rangle_W$ is essential: in a hypothetical divergent trajectory with random-like bit density where Hamming weight grows at half the bit-length rate ($\langle \Delta w \rangle_W \approx \frac{1}{2}(\log_2 3 - \langle \delta \rangle_W) > 0$), the corridor imbalance $\langle U_{1,\text{upper}} - U_{0,\text{upper}} \rangle_W$ also sits right at $2 - \log_2 3 \approx 0.415$. Thus, bounding $\langle U_{1,\text{upper}} - U_{0,\text{upper}} \rangle_W$ alone is sufficient to rule out non-trivial cycles (where $\langle \Delta w \rangle_{\text{cycle}} \equiv 0$), whereas establishing uniform trajectory contraction ($\langle \delta \rangle_W \ge \log_2 3 + \eta$) is equivalent to establishing the joint window bound:

$$ \langle U_{1,\text{upper}} - U_{0,\text{upper}} \rangle_W + 2\langle \Delta w \rangle_W \le 2 - \log_2 3 - \eta $$


8. Preprint Archive