Recurrence Structures, Finite State Decomposition, and Statistical Bias in Collatz Path Sequences

Abstract

We investigate the structure of Collatz path sequences $\{F^k(n)\}_{k=0}^{\infty}$ for positive integers $n$, where $F$ denotes the standard Collatz map. By classifying natural numbers into residue classes modulo~4, we establish that the Collatz conjecture reduces to verifying convergence for integers congruent to $3 \pmod{4}$. For this class, we identify six recurrent forms -- residue classes modulo~9 -- through which the path sequence elements cycle, and we prove that these forms are \emph{complete} in the sense that every power of~2 belongs to exactly one of them. We construct a deterministic finite state machine (FSM) whose states correspond to these six forms and whose transitions encode the Collatz dynamics, yielding a system of coupled functional equations involving linear congruences. We prove closed-form characterizations of the power-of-2 elements within three of the six recurrent classes and establish an equivalence between the FSM dynamics and the Syracuse acceleration of the Collatz map. Numerical experiments on the first $10^8$ natural numbers reveal a pronounced statistical bias in the distribution of terminating recurrent forms, with form $9n+8$ accounting for approximately $97.6\%$ of all terminations, and we formulate precise conjectures regarding the asymptotic frequencies. These results provide a structured decomposition of the Collatz problem into a finite system of interlocking recurrences and highlight the non-random character of the Collatz dynamics.