TheoremComplete

Brun's Theorem

Theorem7.1Brun's Theorem

The sum of reciprocals of twin primes converges: B=∑p:p+2 prime(1p+1p+2)<∞B = \sum_{\substack{p : p+2 \text{ prime}}} \left(\frac{1}{p} + \frac{1}{p+2}\right) < \infty. The constant B≈1.9022B \approx 1.9022 is called Brun's constant. In particular, ∣{p≤x:p+2 is prime}∣=O ⁣(x(log⁡log⁡x)2(log⁡x)2)|\{p \leq x : p + 2 \text{ is prime}\}| = O\!\left(\frac{x(\log\log x)^2}{(\log x)^2}\right).


Proof

Proof

Step 1: Setup. Let A={n(n+2):n≤x}\mathcal{A} = \{n(n+2) : n \leq x\}, P={p:p prime}\mathcal{P} = \{p : p \text{ prime}\}, z=x1/2z = x^{1/2}. An element nn survives the sieve (i.e., (n(n+2),P(z))=1(n(n+2), P(z)) = 1) only if both nn and n+2n + 2 have no prime factor below zz, hence both are prime (or 1).

Step 2: Brun's combinatorial sieve. Truncate the Mobius inclusion-exclusion at level 2r2r (even, for an upper bound): S(A,z)≤∑d∣P(z)Ω(d)≤2rμ(d)∣Ad∣S(\mathcal{A}, z) \leq \sum_{\substack{d | P(z) \\ \Omega(d) \leq 2r}} \mu(d) |\mathcal{A}_d| where Ad={n≤x:d∣n(n+2)}\mathcal{A}_d = \{n \leq x : d | n(n+2)\}.

Step 3: Counting ∣Ad∣|\mathcal{A}_d|. For squarefree dd with d∣n(n+2)d | n(n+2): the number of residues n mod dn \bmod d with d∣n(n+2)d | n(n+2) is ω(d)=∏p∣dω(p)\omega(d) = \prod_{p|d} \omega(p) where ω(p)=2\omega(p) = 2 for p>2p > 2 (since n≡0n \equiv 0 or n≡−2 mod pn \equiv -2 \bmod p) and ω(2)=1\omega(2) = 1. So ∣Ad∣=ω(d)dx+rd|\mathcal{A}_d| = \frac{\omega(d)}{d}x + r_d with ∣rd∣≤ω(d)≤2Ω(d)|r_d| \leq \omega(d) \leq 2^{\Omega(d)}.

Step 4: Main term estimate. The main term gives: S≤x∑d∣P(z)Ω(d)≤2rμ(d)ω(d)d+∑d∣P(z)Ω(d)≤2r2Ω(d)S \leq x \sum_{\substack{d|P(z) \\ \Omega(d) \leq 2r}} \frac{\mu(d)\omega(d)}{d} + \sum_{\substack{d|P(z) \\ \Omega(d) \leq 2r}} 2^{\Omega(d)}

Choosing r=clog⁡log⁡xr = c\log\log x (Brun's choice), the main term is ≪x∏2<p<z(1−2/p)≪x(log⁡z)2≪x(log⁡x)2\ll x \prod_{2 < p < z}(1 - 2/p) \ll \frac{x}{(\log z)^2} \ll \frac{x}{(\log x)^2} times factors of (log⁡log⁡x)O(1)(\log\log x)^{O(1)} from the truncation.

The remainder sum is bounded by ∑d≤D2Ω(d)≤D⋅(log⁡D)O(1)\sum_{d \leq D} 2^{\Omega(d)} \leq D \cdot (\log D)^{O(1)} with D=z2r=xc′log⁡log⁡xD = z^{2r} = x^{c'\log\log x}, which is o(x/(log⁡x)2)o(x/(\log x)^2) for appropriate cc.

Step 5: Conclusion. π2(x)≤S(A,z)≪x(log⁡log⁡x)2(log⁡x)2\pi_2(x) \leq S(\mathcal{A}, z) \ll \frac{x(\log\log x)^2}{(\log x)^2}. The sum ∑p≤x,p+2 prime1/p\sum_{p \leq x, p+2 \text{ prime}} 1/p is then bounded by partial summation: ∑1/p≤∫2xdπ2(t)t≪∫(log⁡log⁡t)2t(log⁡t)2dt<∞\sum 1/p \leq \int_2^x \frac{d\pi_2(t)}{t} \ll \int \frac{(\log\log t)^2}{t(\log t)^2}dt < \infty. □\square

■

RemarkLower Bounds and the Twin Prime Conjecture

While Brun's sieve gives excellent upper bounds, obtaining lower bounds for π2(x)\pi_2(x) is much harder due to the parity problem. The best unconditional lower bound (from sieve methods) shows infinitely many nn with n(n+2)n(n+2) having at most 3 prime factors (Goldston-Pintz-Yildirim type). Proving π2(x)→∞\pi_2(x) \to \infty remains open.