The Sieve of Eratosthenes is the ancient algorithm for finding prime numbers by repeatedly crossing out multiples of each prime. This paper treats that sieve not just as a computational tool but as a dynamical system, meaning it tracks how patterns in the remaining numbers evolve step by step as each new prime is applied. The key objects of study are "gaps" and "constellations," which are patterns describing the spacings between numbers that survive the sieve. At each stage, these patterns form a cycle with a precise structure, and there is a well-defined rule for how one stage's cycle transforms into the next. The authors show that the evolution of how often a given pattern appears can be modeled exactly using Markov chains, which are mathematical tools that describe systems where the next state depends only on the current one.
A central finding is that the number of times a given pattern appears grows at a super-exponential rate as the sieve progresses, but in a very predictable way. By dividing out that predictable growth, the authors extract a "relative population" that converges to a finite constant depending only on the arithmetic structure of the pattern. This is an exact result, not an approximation, and it characterizes the long-run frequency of any admissible constellation within the sieve process. Admissible here means the pattern is not ruled out by simple divisibility reasons, which is the standard necessary condition for a pattern of primes to appear infinitely often.
The authors then use these exact results to build first-order estimates for how often prime constellations, including prime gaps, should appear in specific intervals between consecutive perfect squares. They define a quantity called the quadratic density to measure this, and they verify their predictions against actual prime data up to roughly 568 trillion. The agreement is strong throughout that range. This work provides a rigorous probabilistic and combinatorial framework grounded in the actual mechanics of the sieve, offering new tools for studying longstanding questions about the distribution of primes and prime gaps.