Preprint
Article

This version is not peer-reviewed.

On Representations of Even Integers as a Sum of Two Semiprimes

Submitted:

30 October 2025

Posted:

31 October 2025

You are already at the latest version

Abstract
Representations of a large even integer N as a sum of two semiprimes (products of two primes, squares allowed) are studied. Using a smooth bilinear weight W localized on uv ≍ N and the Hardy–Littlewood circle method, an asymptotic formula of order N with a power saving in log N is obtained: \( R(N) \;=\; \sum_{\begin{array}{c}s_1+s_2=N \\ s_1,s_2\in S_2 \end{array}} W(s_1)W(s_2) \;=\; \mathfrak{S}(N)\,N \;+\; O\!\Big(\frac{N}{(\log N)^{1+\delta}}\Big) \) for some absolute δ > 0, where the multiplicative singular series \( \mathfrak{S}(N) \) is bounded and bounded away from zero along even N. The argument uses a bilinear Bombieri–Vinogradov type input (via the dispersion method and the multiplicative large sieve) together with minor-arc bounds based on Gallagher’s lemma. As a direct consequence, every sufficiently large even integer is the sum of two semiprimes (unweighted existence). A computational check (unweighted) finds no counterexamples up to N = 106 except 2, 4, 6, 22, in agreement with the asymptotic formula.
Keywords: 
;  ;  ;  ;  ;  ;  
Historical note and scope.
Almost-prime Goldbach results have a long history, notably Chen’s theorem N = p + P 2 for large even N and its expositions. An explicit published statement asserting the unconditional existence of P 2 + P 2 representations for all sufficiently large even N in the exact form obtained here (via a weighted asymptotic with positive singular series) has not been located by the author; it may be implicit in the classical literature. The note isolates a transparent circle-method route, with a bilinear Bombieri–Vinogradov input, to a positive singular series and the unweighted existence.

1. Introduction

Let S 2 = { n 2 : n = p q with p q primes } . Consider the weighted representation function
R ( N ) : = s 1 + s 2 = N s 1 , s 2 S 2 W ( s 1 ) W ( s 2 ) ,
where W is a nonnegative smooth weight localizing s i N . The main result is an asymptotic of order N with a power saving in log N . The method uses the major/minor arc decomposition and a bilinear Bombieri–Vinogradov type estimate for semiprimes.

2. Statement of Results

Theorem 1. 
Let W be a smooth nonnegative bilinear weight supported on u v N with u v , arising from a finite smooth dyadic partition of unity, and set R ( N ) as in (1). There exist absolute constants δ > 0 and a multiplicative singular series S ( N ) = p σ p ( N ) such that, for all sufficiently large even N,
R ( N ) = S ( N ) N + O N ( log N ) 1 + δ .
Moreover, there exist constants 0 < c 1 c 2 < with c 1 S ( N ) c 2 for all even N.
Corollary 1 
(Semiprime Goldbach for large even integers). For all sufficiently large even integers N, there exist semiprimes s 1 , s 2 S 2 with s 1 + s 2 = N .
Proof. 
Choose W to majorize the indicator on a fixed positive-proportion window (e.g. W 1 on [ N / 4 , 3 N / 4 ] and supported on x N ). By Theorem 1, R ( N ) = S ( N ) N + O ( N ( log N ) 1 δ ) with S ( N ) bounded below by a positive constant along even N; hence R ( N ) > 0 for N large, which forces the existence of a representation.    □

3. Circle Method Outline

Let S ( α ) = s S 2 W ( s ) e 2 π i α s . Then
R ( N ) = 0 1 S ( α ) 2 e 2 π i α N d α .
Major arcs are handled by a local main term plus an error; minor arcs are bounded via Gallagher’s lemma and a multiplicative large sieve adapted to the semiprime model. The required distribution up to Q N θ ε with θ > 1 / 2 is supplied by the bilinear estimate below.

4. A Bilinear Bombieri–Vinogradov Input

Let X be large and U , V be dyadic with U V X , U V . Let ϖ ( u , v ) be a smooth nonnegative cutoff supported on u U , v V , with derivatives bounded up to a fixed order, and let a u , b v be bounded coefficients modeling prime-like weights.
Proposition 1. 
There exists θ > 1 / 2 and η > 0 such that, uniformly for Q X θ ε ,
q Q max ( a , q ) = 1 u , v 1 u v a ( mod q ) ϖ ( u , v ) a u b v 1 φ ( q ) u , v 1 ( u v , q ) = 1 ϖ ( u , v ) a u b v ε , ϖ X ( log X ) 1 η .
Remark 1. 
The estimate follows from the dispersion method and the multiplicative large sieve in standard bilinear forms.

5. Finite Verification (Computational Note)

A numerical check (unweighted) indicates that the only even integers N 10 6 that are not a sum of two semiprimes are 2 , 4 , 6 , 22 . In particular, every even N 24 up to 10 6 admits a representation, consistent with Theorem 1. A standalone Python script is provided in Appendix A to reproduce and extend this verification.

Appendix A. Python Verification Script

The following script checks which even N N 0 fail to be a sum of two semiprimes (standard definition: product of two primes, allowing squares and the prime 2). It attempts an FFT-based convolution when numpy is available (recommended for N 0 10 6 ); otherwise it falls back to a pure-Python routine and may reduce N 0 to keep runtimes reasonable.
   Usage: python3 verify_semiprimes.py 1000000 (if numpy is installed).  
#!/usr/bin/env python3
# verify_semiprimes.py
# Checks even N <= N0 for representation as a sum of two semiprimes.
import sys
def sieve_primes(n):
    is_prime = bytearray(b"\x01")*(n+1)
    is_prime[:2] = b"\x00\x00"
    import math
    for p in range(2, int(n**0.5)+1):
        if is_prime[p]:
            step = p
            start = p*p
            is_prime[start:n+1:step] = b"\x00"*(((n - start)//step) + 1)
    return [i for i in range(n+1) if is_prime[i]]
def semiprime_indicator(n, primes):
    # Mark semiprimes n = p*q with p <= q (allow squares)
    is_semiprime = bytearray(b"\x00")*(n+1)
    import bisect
    m = bisect.bisect_right(primes, n)
    for i, p in enumerate(primes[:m]):
        if p*p > n:
            break
        is_semiprime[p*p] = 1
        # iterate q >= p
        for q in primes[i+1:]:
            pq = p*q
            if pq > n:
                break
            is_semiprime[pq] = 1
    return is_semiprime
def exceptions_via_fft(is_semiprime):
    # Requires numpy. Uses FFT-based convolution of the indicator.
    import numpy as np
    f = np.array(is_semiprime, dtype=np.float64)
    L = 1
    while L < 2*len(f):
        L <<= 1
    F = np.fft.rfft(f, n=L)
    conv = np.fft.irfft(F * F, n=L)
    eps = 1e-6
    can_sum = conv[:len(f)] > eps
    exc = [k for k in range(2, len(f), 2) if not can_sum[k]]
    return exc
def exceptions_pure_python(is_semiprime, cap=200_000):
    # Fallback: mark sums via nested loops (reasonable up to ~2e5).
    n = min(len(is_semiprime)-1, cap)
    sems = [i for i in range(4, n+1) if is_semiprime[i]]
    can_sum = bytearray(b"\x00")*(n+1)
    m = len(sems)
    for i in range(m):
        a = sems[i]
        for j in range(i, m):
            s = a + sems[j]
            if s > n: break
            can_sum[s] = 1
    exc = [k for k in range(2, n+1, 2) if not can_sum[k]]
    return exc, n
def main():
    try:
        N0 = int(sys.argv[1])
    except Exception:
        N0 = 1_000_000  # default
    try:
        import numpy as _np
        has_numpy = True
    except Exception:
        has_numpy = False
    if not has_numpy and N0 > 200_000:
        print("[warn] numpy not found; reducing N0 to 200000 for pure-Python run.")
        N0 = 200_000
    print(f"[info] building semiprime table up to N0 = {N0} ...")
    primes = sieve_primes(N0)
    isp = semiprime_indicator(N0, primes)
    if has_numpy:
        print("[info] using FFT-based convolution (numpy).")
        exc = exceptions_via_fft(isp)
        print(f"[result] exceptions up to {N0}: {exc}")
    else:
        print("[info] pure-Python marking of sums.")
        exc, up_to = exceptions_pure_python(isp, cap=N0)
        print(f"[result] exceptions up to {up_to}: {exc}")
if __name__ == "__main__":
    main()
Notes. On a machine with numpy available, no even counterexamples were found up to 10 6 beyond { 2 , 4 , 6 , 22 } . Without numpy, the fallback confirms the same up to 2 × 10 5 by default; the bound can be raised cautiously, at the cost of runtime and memory.

References

  1. Chen, J.R. On the representation of a large even integer as the sum of a prime and the product of at most two primes. Sci. Sinica 1973, 16, 157–176. [Google Scholar]
  2. H. L. Montgomery, R. C. Vaughan, Multiplicative Number Theory I, Cambridge University Press, 2007.
  3. H. Iwaniec, E. Kowalski, Analytic Number Theory, AMS Colloquium Publications 53, 2004.
  4. R. C. Vaughan, The Hardy–Littlewood Method, 2nd ed., Cambridge University Press, 1997.
  5. Gallagher, P.X. A large sieve density estimate near σ=1. Invent. Math. 1970, 11, 329–339. [Google Scholar] [CrossRef]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.
Prerpints.org logo

Preprints.org is a free preprint server supported by MDPI in Basel, Switzerland.

Subscribe

Disclaimer

Terms of Use

Privacy Policy

Privacy Settings

© 2025 MDPI (Basel, Switzerland) unless otherwise stated