Path 2: Sieve Theory Path

Gradually strengthen “almost-prime” results until two true primes remain.

Rationale

Sieve methods aim to isolate prime patterns by systematically eliminating composite factors. Viggo Brun (1920s) pioneered using sieves to show every large even is a sum of two numbers each with at most 9 prime factors. This was improved by Rademacher (7 prime factors, 1924), and finally Chen Jingrun’s theorem (1966) which shows every sufficiently large even number is the sum of a prime and a semiprime (product of two primes). Chen’s result is one of the biggest breakthroughs: it proves “prime + prime” or “prime + prime·prime” covers all large evens. The sieve path attempts to push “prime + semiprime” to “prime + prime” by reducing the size or number of prime factors in the second term. It’s difficult due to the parity problem: traditional sieve techniques have trouble distinguishing an odd count of prime factors from an even count, making it hard to rule out that last extra factor. However, further innovations or combining sieve with analytic ideas (e.g. using the large sieve or parity-breaking weights) might overcome this. The literature shows steady if slow progress (e.g. Pintz & Ruzsa 2020 reducing Linnik’s $K$ of powers of 2 needed to 8), so continued refinement is plausible.

Prerequisite Themes

Elementary and combinatorial number theory; basic and advanced sieve techniques (Eratosthenes sieve, Brun’s method, Selberg sieve, large sieve); understanding of almost primes and distribution of friable (smooth) numbers; parity barrier in sieves.

Dependencies

Relies on primality results which sometimes come from Path 3 (e.g. zero-free regions or estimates of primes in short intervals to feed into sieves). Path 2 and Path 1 can be combined – e.g. using sieve to handle “bad” minor arc regions in the circle method.

Signs of Progress

Reducing the “almost prime” level: e.g. proving every large even is prime + a number with $\le 1$ prime factors (which is exactly Goldbach). Intermediate milestones: showing every even is prime + $P_k$ for smaller $k$ (product of $k$ primes). Chen achieved $k=2$; getting to $k=1$ is the goal. Another sign: eliminating the need for the “sufficiently large” caveat by explicitly bounding where exceptions could lie (shrinking the unknown range). Any drop in the number of allowed 2’s in Linnik’s result (now 8) or improved bounds in Chen’s theorem (e.g. Tomohiro Yamada’s work making Chen’s result explicit) would be concrete progress.

Base Camp 2.1: Classical Sieve Techniques – Brun’s Sieve and Basic Notions

Scope: Learn the principles of sieve methods in their simplest form, starting with Brun’s sieve (combinatorial sieve) and legacies like the Brun constant for twin primes. What you must be able to do: set up upper and lower bound sieves, like the combinatorial sieve that estimates the size of sifted sets (e.g. count of numbers with no small prime factors). Understand how to derive results like “there are infinitely many integers with at most 2 prime factors” or Brun’s result on twin primes (even if he didn’t prove infinitely many, he got a convergence result). Grasp the basic inclusion-exclusion strategy and the concept of sieve weight functions. In context, Brun’s method was used to show every sufficiently large even number is the sum of two $P_9$ numbers (each having at most 9 prime factors). You should be comfortable with such statements and how a sieve is set up to prove them.

Stepping-stones: (1) Prove Brun’s theorem: the sum of reciprocals of twin primes converges (this introduces the notion of using sieves to count almost primes; it’s not directly Goldbach, but twin primes are analogous in difficulty). (2) Using a basic sieve, show the existence of “almost primes” in any interval (e.g. there are $O(x (\ln\ln x)/\ln x)$ primes ≤ $x$, which is weaker than PNT but doable by sieve). (3) Derive the result that any even number can be written as sum of two $P_9$ (9-almost-primes) following Brun’s 1920 method. This involves counting representations with inclusion-exclusion: even $N = a + b$ where $a, b$ avoid having >9 prime factors, ensuring at most 9 each. (4) Practice building upper bound sieves (like Selberg’s sieve) and lower bound sieves (to show existence of primes or almost primes in sequences).

Best resources:

Base Camp 2.2: Selberg’s Sieve and “Almost Prime” Results – Toward Chen’s Theorem

Scope: Study Selberg’s sieve (a more flexible analytic sieve) and its powerful results, culminating in Chen’s theorem (1973): every sufficiently large even $N$ is $p + p_2$ with $p$ prime and $p_2$ either prime or product of two primes. What you must be able to do: apply Selberg’s sieve to count almost primes in various sequences, e.g. show there are $O(x (\ln\ln x)^2/\ln^3 x)$ primes with at most 2 prime factors up to $x$ (this kind of estimate is key in Chen’s argument). Understand how Chen sharpened the remainder error terms to ensure one of the primes is a true prime. This base-camp is technical: you’ll learn to optimize the parity of counts – the so-called parity problem in sieve (why sieves have trouble telling apart sets of primes vs almost primes due to symmetric inclusion-exclusion). Chen circumvented parity issues partially by a clever two-level sieve.

Stepping-stones: (1) Work through Selberg’s sieve derivation: derive the Selberg sieve upper bound for the count of numbers ≤ $x$ with no prime factors ≤ $D$. (2) As an exercise, use Selberg’s sieve to prove Rényi’s result (1947): there exists some fixed $K$ such that every even number is prime + $P_K$ (number with ≤$K$ prime factors). Selberg’s sieve gives such existence of $K$ more systematically than Brun. (3) Study Chen’s original paper or a good exposition of it: identify how he handles the case distinctions (“$N=p_1+p_2$ or $p_1 + (p_3 p_4)$”) and ensures that at least one representation falls into the desired form. (4) Practice a simpler but related proof: Bombieri–Vinogradov theorem (which uses analytic methods, but conceptually similar to sieve outcomes) – it says primes are well-distributed in arithmetic progressions on average, a fact Chen utilized implicitly to control error terms.

Best resources:

Base Camp 2.3: Erdős–Rankin Type Refinements – Breaking Parity and Beyond

Scope: Explore advanced sieve refinements and auxiliary methods devised to overcome or circumvent the parity problem. Erdős’s contributions and later developments (like linear combinations of sieves or using weights to favor certain residues) come into play. What you must be able to do: understand the parity problem clearly: why a sieve can rarely distinguish an odd number of prime factors from an even number, leaving many problems (like Goldbach, twin primes) hanging. Learn about techniques to break parity artificially – e.g. using sequences with known parity (like almost primes of a certain type) or applying weights that are not $\{0,1\}$ (Ramaré’s work that every even is sum of 6 primes in 1995 used a brute-force but notable combinatorial enumeration, not exactly parity-breaking but a different angle). Also look at Pintz’s work on slowly decreasing that “$K$” in Linnik’s theorem (which uses some parity-breaking insights).

Stepping-stones: (1) Study the Erdős–Rankin theorem which gives an infinitely often lower bound on gaps between primes that is stronger than the average behavior – they used sieve with some parity insight. (2) Look at small improvements: e.g. “Chen’s theorem with explicit bounds” (Yamada 2015) which shows how far current techniques can go in making Chen’s result explicit. (3) Understand binary Goldbach attempts with parity consideration: e.g. the concept of using a “bilinear sum” approach (combining two different sequences in a sieve-like way), which appears in some recent work. (4) Familiarize with the idea of majorant principle in sieve: sometimes we sieve a larger set that contains primes but is easier to analyze, then intersect with primes. This is used in combination with analytic methods to handle parity (such as in Vinogradov’s theorem proof, one uses all almost primes up to certain size as a majorant for primes).

Best resources:

Base Camp 2.4: Combining Sieve with Analytic Input – Hybrid Approaches

Scope: At this advanced base-camp, focus on how sieve methods integrate with analytic methods (from Path 1 and 3) to yield stronger results. For example, Bombieri–Vinogradov theorem (analytic) greatly enhances what sieve can do by providing “well-distributed primes” on average. What you must be able to do: replicate some proofs that require both sieve and analytic number theory. A key example: the proof that every even is the sum of at most 4 primes (Ramaré 1995) which relied on exhaustive computation up to a limit, then a mixture of analytic estimates and combinatorial argument beyond that. Another: the Pintz–Ruzsa result (2020) that improved Linnik’s $K$ to 8 – it likely used modern sieve with some heavy-duty input like the distribution of primes in almost all moduli. Develop the skill to take an analytic theorem (like “primes have no Siegel zeros” or “Erdős–Turán inequality for primes”) and plug it into a sieve framework.

Stepping-stones: (1) Re-derive Chen’s theorem assuming Bombieri–Vinogradov as a black box – see how the proof simplifies or the bounds strengthen. (2) Look at Heath-Brown & Puchta (2002) who showed Linnik’s theorem holds with $K=7$ under GRH. Following their approach will exercise using a hypothesis from analytic number theory inside a sieve counting argument. (3) Study a simpler hybrid result: the proof that the primes contain infinitely many pairs at distance ≤ 20 (a weak twin prime result by Chen’s student Pan Chengdong, 1973). It combined Chen’s methods with analytic tricks. (4) Explore the concept of enlarging sieve weights via Fourier analysis – this is used in “dense subset” arguments (for example, in Green–Tao theorem on arithmetic progressions).

Best resources:

Foundational across camps: Halberstam & Richert’s Sieve Methods is used in multiple base-camps as the cornerstone text for understanding sieve theory from the ground up to advanced results like Chen’s theorem.

Full Bibliography (Path 2)

BC2.1 Classical Sieve:

BC2.2 Selberg Sieve & Chen:

BC2.3 Refinements & Parity:

BC2.4 Hybrid Approaches:

← Back to Kangchenjunga – Goldbach’s Conjecture