betaFeedback

WorksMathematics

Euler's 1747 conjecture on sums of integers with restricted prime factors

Abstract

We prove that every positive integer congruent to 22 modulo 44 is a sum of two positive integers whose prime factors are all congruent to 11 modulo 44, with 11 admitted as a summand. Euler proposed this statement in a letter to Goldbach dated May 6, 1747. The proof counts representations by four squares and bounds those containing a common prime divisor congruent to 33 modulo 44 in either coordinate pair. An average over a nonsplit torus normalizer converts the quaternionic Hecke bound into an error independent of the sieving prime. Elementary estimates then give the explicit threshold 64464^4. An exact finite arithmetic certificate covers the remaining 4,194,3044{,}194{,}304 integers in the required residue class.

Introduction

Let

P3={p prime:p≡3(mod4)},A={a∈Z>0:p∣a⇒p≡1(mod4)}.\mathcal{P}_3 = \{p\ \text{prime} : p \equiv3 \pmod{4}\}, \qquad\mathcal{A} = \{a \in\mathbb{Z}_{>0} : p \mid a \Rightarrow p \equiv1 \pmod{4}\}.

The integer 1 belongs to A\mathcal{A}. In his letter to Goldbach of May 6, 1747, Euler proposed that every integer of the form 4m+24m+2 can be split into two integers of the form 4x+14x+1, each having no divisor of the form 4p−14p-1. This is the assertion proved here. The statement appears in Letter 115 of the collected correspondence, together with a discussion in editorial note 10 [4 pp. 935–936].

Dietmann and Elsholtz proved that every sufficiently large positive integer congruent to 2 modulo 4 has the required decomposition [2]. Their argument uses a result of Greaves on four squares with two prime variables. We obtain an explicit threshold by counting all four-square representations and estimating the unwanted common divisors of the two coordinate pairs.

Theorem 1.1. Every positive integer n≡2(mod4)n \equiv2 \pmod{4} admits a decomposition

n=a+b,a,b∈A.n = a+b, \qquad a,b \in\mathcal{A}.

The quantitative part of the proof establishes the result for n≥644=16,777,216n \ge64^4 = 16,777,216 after removing factors from A\mathcal{A}. § 7 gives an exact arithmetic certificate for the finite range below this threshold. The spectral ingredient is the weight-two Ramanujan–Petersson bound, applied through the Jacquet–Langlands correspondence. All auxiliary constants are explicit.

Four-square representations and prime obstructions

Write σ(u)=∑d∣ud\sigma(u) = \sum_{d\mid u} d and τ(u)=∑d∣u1\tau(u) = \sum_{d\mid u} 1. We first remove a multiplicative factor that already belongs to A\mathcal{A}.

Lemma 2.1. To prove Theorem 1.1, it suffices to consider n=2mn = 2m, where mm is odd and every prime divisor of mm belongs to P3\mathcal{P}_3.

Proof. For any odd mm, write m=qm0m = qm_0, where q∈Aq \in\mathcal{A} and every prime divisor of m0m_0 belongs to P3\mathcal{P}_3. If 2m0=a+b2m_0 = a+b with a,b∈Aa,b \in\mathcal{A}, then 2m0=qa+qb2m_0 = qa+qb has the same property. □\square

For the remainder of the analytic argument, n=2mn=2m satisfies Lemma 2.1. Put

R(n)={(x1,x2,x3,x4)∈Z4:x12+x22+x32+x42=n}.\mathcal{R}(n)=\{(x_1,x_2,x_3,x_4)\in\mathbb{Z}^4:x_1^2+x_2^2+x_3^2+x_4^2=n\}.

These are signed, ordered representations. Jacobi’s four-square formula gives

R(n):=∣R(n)∣=24σ(m).(1)R(n):=|\mathcal{R}(n)|=24\sigma(m). \tag*{(1)}

Exactly two coordinates are odd, since a square is 0 or 1 modulo 4. Let Rmix(n)\mathcal{R}_{\mathrm{mix}}(n) consist of the representations in which each of (x1,x2)(x_1,x_2) and (x3,x4)(x_3,x_4) has opposite parity. Permuting coordinates shows that the six possible placements of the odd coordinates occur equally often. Four of them have the required parity, so

∣Rmix(n)∣=16σ(m)=23R(n).(2)|\mathcal{R}_{\mathrm{mix}}(n)|=16\sigma(m)=\frac{2}{3}R(n). \tag*{(2)}

For p∈P3p\in\mathcal{P}_3, let bp(n)b_p(n) count the elements of Rmix(n)\mathcal{R}_{\mathrm{mix}}(n) for which

(p∣x1 and p∣x2)or(p∣x3 and p∣x4).(p\mid x_1\text{ and }p\mid x_2)\quad\text{or}\quad(p\mid x_3\text{ and }p\mid x_4).

Since −1-1 is a quadratic nonresidue modulo pp, the implication p∣x2+y2⇒p∣x,yp\mid x^2+y^2\Rightarrow p\mid x,y holds. Consequently, any representation outside all these exceptional sets supplies two positive odd summands in A\mathcal{A}. It is therefore enough to prove

∑p∈P3bp(n)<23R(n).(3)\sum_{p\in\mathcal{P}_3}b_p(n)<\frac{2}{3}R(n). \tag*{(3)}

A uniform quaternionic estimate

Proposition 3.1. For every p∈P3p\in\mathcal{P}_3,

bp(n)R(n)≤2p(p−1)+1p>3τ(m)m3σ(m).(4)\frac{b_p(n)}{R(n)}\leq\frac{2}{p(p-1)}+\mathbf{1}_{p>3}\frac{\tau(m)\sqrt{m}}{3\sigma(m)}. \tag*{(4)}

The estimate follows from a Hecke action on a finite quotient associated with the Hurwitz order.

The Hurwitz order and its finite quotient

Let B=Q+Qi+Qj+QkB=\mathbb{Q}+\mathbb{Q}i+\mathbb{Q}j+\mathbb{Q}k be the Hamilton quaternion algebra, where i2=j2=−1i^2=j^2=-1 and k=ij=−jik=ij=-ji. Its reduced norm is

nrd⁡(a+bi+cj+dk)=a2+b2+c2+d2.\operatorname{nrd}(a+bi+cj+dk)=a^2+b^2+c^2+d^2.

The Hurwitz order is

H=Z4∪(Z+12)4,\mathcal{H}=\mathbb{Z}^4\cup\left(\mathbb{Z}+\frac{1}{2}\right)^4,

with coordinates interpreted in the basis 1,i,j,k1,i,j,k. Its 24 units are the eight coordinate units and the sixteen elements (±1±i±j±k)/2(\pm1\pm i\pm j\pm k)/2.

The order H\mathcal{H} is norm-Euclidean. Indeed, rounding four real coordinates to integers gives a remainder of norm at most 1. Equality requires all four coordinates to be half-integers, in which case a Hurwitz half-integer gives remainder zero. Euclidean division implies that every right ideal is principal. In particular, the right ideal class number of H\mathcal{H} is one.

Let Qm={α∈H:nrd⁡(α)=m}Q_m=\{\alpha\in\mathcal{H}:\operatorname{nrd}(\alpha)=m\}. Right multiplication by 1+i1+i is a bijection

Qm⟶R(2m).(5)Q_m\longrightarrow\mathcal{R}(2m). \tag*{(5)}

For the inverse, divide an integral quaternion of norm 2m2m by 1+i1+i. The new coordinates are halves of the sums and differences in two coordinate pairs. The two odd coordinates of the original quaternion ensure that all four new coordinates are integers or all four are strict half-integers. Thus the inverse belongs to H\mathcal{H}.

Fix p∈P3p\in\mathcal{P}_3 with p∤mp\nmid m. The algebra BB splits at every odd prime, and reduction identifies H/pH\mathcal{H}/p\mathcal{H} with M2(Fp)M_2(\mathbb{F}_p), with reduced norm corresponding to determinant. Write

Gp=PGL⁡2(Fp),ρp:{α∈H:p∤nrd⁡(α)}⟶Gp.G_p=\operatorname{PGL}_2(\mathbb{F}_p),\qquad \rho_p:\{\alpha\in\mathcal{H}:p\nmid\operatorname{nrd}(\alpha)\}\longrightarrow G_p.

Let Γp=ρp(H×)\Gamma_p=\rho_p(H^\times) and Yp=Γp\GpY_p=\Gamma_p\backslash G_p. The only units with scalar reduction are ±1\pm1, so

∣Γp∣=12,∣Yp∣=p(p2−1)12.(6)|\Gamma_p|=12,\qquad|Y_p|=\frac{p(p^2-1)}{12}. \tag*{(6)}

We equip C[Yp]\mathbb{C}[Y_p] with the counting inner product.

The Hecke operator and its spectral bound

For mm odd and p∤mp\nmid m, define

(Tmf)(Γpg)=124∑α∈Qmf(Γpρp(α)g).(7)(T_mf)(\Gamma_pg)=\frac{1}{24}\sum_{\alpha\in Q_m}f(\Gamma_p\rho_p(\alpha)g). \tag*{(7)}

Right multiplication by a unit permutes the left-unit classes in QmQ_m, so this is well-defined. It commutes with right translation by GpG_p. Quaternion conjugation permutes QmQ_m and gives projective inversion, which proves self-adjointness. Equations (1) and (5) give Tm1=σ(m)1T_m1=\sigma(m)1.

The function

χp(Γpg)=(det⁡(g)p)\chi_p(\Gamma_pg)=\left(\frac{\det(g)}{p}\right)

is well-defined because a scalar changes the determinant by a square and every element of Γp\Gamma_p has square determinant. It satisfies Tmχp=(mp)σ(m)χpT_m\chi_p=\left(\frac{m}{p}\right)\sigma(m)\chi_p.

Lemma 3.2. On {1,χp}⊥\{1,\chi_p\}^{\perp}, the operator norm of TmT_m is at most τ(m)m\tau(m)\sqrt{m}.

Proof. Let PB×\mathrm{PB}^{\times} denote the algebraic group of projective units of BB. At a finite prime ℓ≠p\ell\ne p, take the projective image of Hℓ×H_\ell^{\times} as the local level. At pp, take the kernel of reduction from PGL2(Zp)\mathrm{PGL}_2(\mathbb{Z}_p) to GpG_p. Denote the product of these compact open subgroups by U(p)U(p). The class-number-one property gives

PB×(Q)\PB×(Af)/U(p)≃Yp.(8)\mathrm{PB}^{\times}(\mathbb{Q})\backslash\mathrm{PB}^{\times}(\mathbb{A}_f)/U(p)\simeq Y_p. \tag*{(8)}

To see the stabilizer, choose an integral representative at every finite place. A rational quaternion carrying one such representative to another can, after rational rescaling, be taken to be a Hurwitz unit. Its remaining action at pp is precisely Γp\Gamma_p.

The usual unramified Hecke correspondence of index mm on (8) is (7). In fact, its local right cosets at ℓe∥m\ell^e\mathbin{\|}m specify integral right ideals of reduced norm ℓe\ell^e. Combining these local ideals gives the integral right ideals of reduced norm mm. Each is αH\alpha H for an α∈Qm\alpha\in Q_m, unique up to right multiplication by a unit. Returning to integral representatives gives ρp(α−1)\rho_p(\alpha^{-1}) at pp. Conjugation sends these right-unit classes to the left-unit classes in (7), since projectively α−1=α‾\alpha^{-1}=\overline{\alpha}.

One-dimensional automorphic representations factor through the reduced norm. Their restrictions to YpY_p depend only on the determinant in Fp×/(Fp×)2\mathbb{F}_p^\times/(\mathbb{F}_p^\times)^2, so lie in the span of 11 and χp\chi_p. The Jacquet–Langlands correspondence transfers the remaining representations to weight-two cuspidal representations, unramified at every prime dividing mm [5 §16]. This use of the correspondence and the weight-two spectral bound is also described in [3 §8.4].

By Ramanujan–Petersson, at an unramified prime ℓ\ell the normalized Satake parameters uℓ,vℓu_\ell,v_\ell have absolute value 11 [1]. In the normalization of (7), the index-ℓe\ell^e eigenvalue is

ℓe/2∑r=0euℓrvℓe−r.\ell^{e/2}\sum_{r=0}^{e}u_\ell^r v_\ell^{e-r}.

The triangular cosets with diagonal (ℓr,ℓe−r)(\ell^r,\ell^{e-r}) have multiplicity ℓr\ell^r and spherical factor ℓ(e−2r)/2\ell^{(e-2r)/2}. Their product is ℓe/2\ell^{e/2}, giving the displayed normalization. The eigenvalue is at most (e+1)ℓe/2(e+1)\ell^{e/2} in absolute value. Multiplication over ℓe∥m\ell^e\mathbin{\|}m proves the lemma. ∎

Averaging over a torus normalizer

Since p≡3(mod4)p \equiv3 \pmod{4}, the subalgebra Fp[i]\mathbb{F}_p[i] is the field Fp2\mathbb{F}_{p^2}. Let

Cp=Fp[i]×/Fp×⊂Gp,Kp=Cp∪Cpρp(j).C_p=\mathbb{F}_p[i]^\times/\mathbb{F}_p^\times\subset G_p,\qquad K_p=C_p\cup C_p\rho_p(j).

Conjugation by jj induces the nontrivial field automorphism, so KpK_p is a subgroup of order 2(p+1)2(p+1). An invertible quaternion has its first or its second coordinate pair zero modulo pp exactly when its projective reduction belongs to KpK_p. Since ρp(1+i)∈Cp\rho_p(1+i)\in C_p, multiplication by 1+i1+i preserves this condition.

The intersection Γp∩Kp\Gamma_p\cap K_p consists of the four projective classes of 1,i,j,k1,i,j,k. Let

Op={Γpk:k∈Kp},fp=1Op,μp=∣Op∣∣Yp∣=6p(p−1).\mathcal{O}_p=\{\Gamma_p k:k\in K_p\},\qquad f_p=1_{\mathcal{O}_p},\qquad\mu_p=\frac{|\mathcal{O}_p|}{|Y_p|}=\frac{6}{p(p-1)}.

There are (p+1)/2(p+1)/2 points in Op\mathcal{O}_p. Each left-unit class whose reduction lies in Op\mathcal{O}_p has exactly eight elements reducing into KpK_p. Counting all representations therefore gives

bp(n)≤8(Tmfp)(Γp).(9)b_p(n)\leq8(T_mf_p)(\Gamma_p). \tag*{(9)}

The function fpf_p is invariant under right translation by KpK_p. The same is true of TmfpT_mf_p, hence

(Tmfp)(Γp)=⟨fp,Tmfp⟩∣Op∣.(10)(T_mf_p)(\Gamma_p)=\frac{\langle f_p,T_mf_p\rangle}{|\mathcal{O}_p|}. \tag*{(10)}

The norm map Fp2×→Fp×\mathbb{F}_{p^2}^\times\to\mathbb{F}_p^\times is surjective. The orbit Op\mathcal{O}_p therefore has equally many points in the two determinant classes. Thus fp−μp1f_p-\mu_p1 is orthogonal to both 11 and χp\chi_p, and

∥fp−μp1∥22=∣Op∣(1−μp).\lVert f_p-\mu_p1\rVert_2^2=|\mathcal{O}_p|(1-\mu_p).

Lemma 3.2 and (10) imply

(Tmfp)(Γp)≤σ(m)μp+τ(m)m(1−μp).(11)(T_mf_p)(\Gamma_p)\leq\sigma(m)\mu_p+\tau(m)\sqrt{m}(1-\mu_p). \tag*{(11)}

For p=3p=3 one has μp=1\mu_p=1, so the second term vanishes. Combining (9) and (11) with R(n)=24σ(m)R(n)=24\sigma(m) proves Proposition 3.1 when p∤mp\nmid m.

Completion of the proof of Proposition 3.1. Suppose p∣mp\mid m. If a coordinate pair is divisible by pp, the other pair has sum of squares zero modulo pp, so all four coordinates are divisible by pp. Hence bp(n)=0b_p(n)=0 when p2∤mp^2\nmid m and bp(n)=16σ(m/p2)b_p(n)=16\sigma(m/p^2) otherwise. Writing m=peum=p^e u gives

σ(m/p2)σ(m)=pe−1−1pe+1−1<1p2.\frac{\sigma(m/p^2)}{\sigma(m)}=\frac{p^{e-1}-1}{p^{e+1}-1}<\frac{1}{p^2}.

It follows that bp(n)/R(n)≤2/(3p2)≤2/[p(p−1)]b_p(n)/R(n)\leq2/(3p^2)\leq2/[p(p-1)]. □

Large primes and elementary lattice counting

For t≥0t\geq0, let Nmix(t)N_{\mathrm{mix}}(t) be the number of signed, ordered pairs (x,y)∈Z2(x,y)\in\mathbb{Z}^2 with opposite parity and x2+y2≤tx^2+y^2\leq t.

Lemma 4.1. For every t≥0t\geq0, one has Nmix(t)≤2(t+1)N_{\mathrm{mix}}(t)\leq2(t+1).

Proof. Each orbit under quarter-turns has four elements and a unique representative with x>0x>0 and y≥0y\geq0. Write

x+y=2u+1,x−y=2v+1,u≥0,−u≤v≤u.x+y=2u+1,\qquad x-y=2v+1,\qquad u\geq0,\quad-u\leq v\leq u.

Assign to this pair the positive integer L=u2+u+v+1L=u^2+u+v+1. For fixed uu, these labels run from u2+1u^2+1 to (u+1)2(u+1)^2. The labels are therefore distinct. Moreover,

x2+y2+12=L+v2≥L.\frac{x^2+y^2+1}{2}=L+v^2\geq L.

At most ⌊(t+1)/2⌋\lfloor(t+1)/2\rfloor labels can occur. Multiplication by four gives the claimed upper bound. □

Define

M(n)=max⁡1≤u≤n∏q∣uq≡1 (4)(e+1).M(n)=\max_{1\le u\le n}\prod_{\substack{q\mid u\\q\equiv1\ (4)}}(e+1).

The two-square formula implies r2(u)≤4M(n)r_2(u)\le4M(n) for 1≤u≤n1\le u\le n, where r2r_2 counts signed, ordered representations.

Lemma 4.2. For every P≥3P\ge3,

1R(n)∑p>Pp∈P3bp(n)≤43M(n)(∑p>Pp∈P31p2+#{p∈P3:P<p≤n}n).(12)\frac{1}{R(n)}\sum_{\substack{p>P\\p\in\mathcal{P}_3}}b_p(n)\le\frac{4}{3}M(n)\left(\sum_{\substack{p>P\\p\in\mathcal{P}_3}}\frac{1}{p^2}+\frac{\#\{p\in\mathcal{P}_3:P<p\le\sqrt{n}\}}{n}\right). \tag*{(12)}

Proof. If pp divides the first coordinate pair, divide that pair by pp. Lemma 4.1 leaves at most 2(n/p2+1)2(n/p^2+1) possibilities. The other pair has positive norm, because its coordinates have opposite parity, and has at most 4M(n)4M(n) representations. Accounting for either pair gives

bp(n)≤16M(n)(n/p2+1).b_p(n)\le16M(n)(n/p^2+1).

Such a prime is at most n\sqrt{n}, since the pair it divides is nonzero. Sum this estimate and use R(n)=24σ(m)≥12nR(n)=24\sigma(m)\ge12n. □

Explicit arithmetic estimates

Lemma 5.1. Under the hypotheses of Lemma 2.1,

τ(m)mσ(m)≤2310m1/5,M(n)≤115n1/5.(13)\frac{\tau(m)m}{\sigma(m)}\le\frac{23}{10}m^{1/5},\qquad M(n)\le\frac{11}{5}n^{1/5}. \tag*{(13)}

Proof. For pe∥mp^e\parallel m, e≥1e\ge1, one has

(e+1)peσ(pe)pe/5≤pp+1e+1pe/5.\frac{(e+1)p^e}{\sigma(p^e)p^{e/5}}\le\frac{p}{p+1}\frac{e+1}{p^{e/5}}.

For the second bound, the corresponding local factor is (e+1)/qe/5(e+1)/q^{e/5}. At a prime at least 3232, both factors are at most 11, since e+1≤2e≤pe/5e+1\le2^e\le p^{e/5}.

For smaller primes, the ratio of successive values of (e+1)/pe/5(e+1)/p^{e/5} is (e+2)/((e+1)p1/5)(e+2)/((e+1)p^{1/5}), which decreases with ee. The maximizing positive exponent is therefore determined by (e+2)5≤p(e+1)5(e+2)^5\le p(e+1)^5. The required values are

p∈P3p\in\mathcal{P}_3, p<32p<323711192331
Maximizing positive ee421111
q≡1(mod4)q\equiv1\pmod4, q<32q<325131729
Maximizing positive ee2111

Taking the maximum with 11 allows each prime to be absent. The products of the fifth powers of the resulting local bounds are

C15=1831431200273643432849934139195392<(2310)5,C_1^5=\frac{183143120027364343}{2849934139195392}<\left(\frac{23}{10}\right)^5,
C25=7962624160225<(115)5.C_2^5=\frac{7962624}{160225}<\left(\frac{11}{5}\right)^5.

Multiplication over the prime factors proves both assertions. □

Lemma 5.2. The following bounds hold for real t≥0t \ge0 and P≥3P \ge3.

#{p∈P3:3<p≤t}≤2t15+1,(14)\#\{p \in\mathcal{P}_3 : 3 < p \le t\} \le\frac{2t}{15} + 1, \tag*{(14)}
∑p>Pp∈P31p2≤215P+2P2,(15)\sum_{\substack{p>P\\p\in\mathcal{P}_3}} \frac{1}{p^2} \le\frac{2}{15P} + \frac{2}{P^2}, \tag*{(15)}
2∑p∈P31p(p−1)<2150.(16)2\sum_{p\in\mathcal{P}_3}\frac{1}{p(p-1)} < \frac{21}{50}. \tag*{(16)}

Proof. Every prime in P3\mathcal{P}_3 exceeding 3 belongs to one of the eight residue classes

S={7,11,19,23,31,43,47,59}(mod60).S = \{7,11,19,23,31,43,47,59\}\pmod{60}.

Let A(t)A(t) count the positive integers at most tt in these classes. Checking the eight jumps in one period gives

−1415≤A(t)−2t15≤1415.-\frac{14}{15} \le A(t)-\frac{2t}{15}\le\frac{14}{15}.

Periodicity extends the bound to every t≥0t\ge0, and proves (14). For t≥Pt\ge P it also gives

A(t)−A(P)≤215(t−P)+2.A(t)-A(P)\le\frac{2}{15}(t-P)+2.

Partial summation yields

∑k>Pkmod60∈S1k2=2∫P∞A(t)−A(P)t3 dt≤215P+2P2.\sum_{\substack{k>P\\k\bmod60\in S}}\frac{1}{k^2} =2\int_P^\infty\frac{A(t)-A(P)}{t^3}\,dt \le\frac{2}{15P}+\frac{2}{P^2}.

For P≥3P\ge3, this set contains all the primes in (15). Finally, direct summation through 60 gives

2∑p∈P31p(p−1)≤∑p∈{3,7,11,19,23,31,43,47,59}2p(p−1)+12059(215⋅60+2602)<2150.2\sum_{p\in\mathcal{P}_3}\frac{1}{p(p-1)} \le\sum_{p\in\{3,7,11,19,23,31,43,47,59\}}\frac{2}{p(p-1)} +\frac{120}{59}\left(\frac{2}{15\cdot60}+\frac{2}{60^2}\right) <\frac{21}{50}.

Here p/(p−1)≤60/59p/(p-1)\le60/59 for p>60p>60, so (15) bounds the remainder. □

The analytic range

Proposition 6.1. Suppose n=2m≥644n=2m\ge644, where mm is odd and all its prime divisors belong to P3\mathcal{P}_3. Then n∈A+An\in\mathcal{A}+\mathcal{A}.

Proof. Set X=n1/4≥64X=n^{1/4}\ge64 and P=2XP=2X. Proposition 3.1, Lemma 5.1, and (14) give

1R(n)∑p<Pp∈P3bp(n)≤2150+τ(m)m3σ(m)(4X15+1)≤2150+233023/10X−1/5(415+1X).(17)\frac{1}{R(n)}\sum_{\substack{p<P\\p\in\mathcal{P}_3}}b_p(n) \le\frac{21}{50}+\frac{\tau(m)\sqrt{m}}{3\sigma(m)}\left(\frac{4X}{15}+1\right) \le\frac{21}{50}+\frac{23}{30}2^{3/10}X^{-1/5}\left(\frac{4}{15}+\frac{1}{X}\right). \tag*{(17)}

For the large primes, Lemmas 4.2–5.2 give

1R(n)∑p>Pp∈P3bp(n)≤4415X4/5(115X+12X2+215X2+1X4)=4415X−1/5(115+1930X+1X3).(18)\frac{1}{R(n)}\sum_{\substack{p>P\\p\in\mathcal{P}_3}}b_p(n) \le\frac{44}{15}X^{4/5}\left(\frac{1}{15X}+\frac{1}{2X^2}+\frac{2}{15X^2}+\frac{1}{X^4}\right) = \frac{44}{15}X^{-1/5}\left(\frac{1}{15}+\frac{19}{30X}+\frac{1}{X^3}\right). \tag*{(18)}

The prime-count term in (12) was bounded by 2/(15n)+1/n2/(15\sqrt{n}) + 1/n.

For X≥64X \ge64,

X−1/5≤49,23/10<54,X−1≤164.X^{-1/5} \le\frac{4}{9}, \qquad2^{3/10} < \frac{5}{4}, \qquad X^{-1} \le\frac{1}{64}.

The first two inequalities follow by raising to the fifth and tenth powers. Adding (17) and (18) now gives

1R(n)∑p∈P3bp(n)≤2150+2330⋅54⋅49(415+164)\frac{1}{R(n)}\sum_{p\in\mathcal{P}_3} b_p(n) \le\frac{21}{50}+\frac{23}{30}\cdot\frac{5}{4}\cdot\frac{4}{9}\left(\frac{4}{15}+\frac{1}{64}\right)
+4415⋅49(115+1930⋅64+1643)=2123549333177600<23.\begin{aligned} \qquad+\frac{44}{15}\cdot\frac{4}{9}\left(\frac{1}{15}+\frac{19}{30\cdot64}+\frac{1}{64^3}\right) \\ &=\frac{21235493}{33177600}<\frac{2}{3}. \end{aligned}

Equations (2) and (3) therefore supply the required representation. □

The finite range

We give the finite arithmetic certificate in terms of binary indicators and finite sets. Put H=222−1=4,194,303H=2^{22}-1=4,194,303 and define

εj=∏p≤4096p∈P3(1−⌊4j+1p⌋+⌊4jp⌋)(0≤j≤H),εj=0(j<0).(19)\varepsilon_j=\prod_{\substack{p\le4096\\p\in\mathcal{P}_3}}\left(1-\left\lfloor\frac{4j+1}{p}\right\rfloor+\left\lfloor\frac{4j}{p}\right\rfloor\right)\quad(0\le j\le H),\qquad\varepsilon_j=0\quad(j<0). \tag*{(19)}

Each factor is 00 if p∣4j+1p\mid4j+1 and is 11 otherwise. Thus εj∈{0,1}\varepsilon_j\in\{0,1\}.

Lemma 7.1. For 0≤j≤H0\le j\le H, one has εj=1\varepsilon_j=1 exactly when 4j+1∈A4j+1\in A.

Proof. The integer 4j+1≤644−34j+1\le64^4-3 is odd and congruent to 11 modulo 44. Its prime factors belonging to P3\mathcal{P}_3 occur an even number of times, counting multiplicity. If any occur, at least two occur. At least one of these factors is then no larger than 4j+1<4096\sqrt{4j+1}<4096. Formula (19) detects precisely this possibility. □

Let J={0≤i≤78:εi=1}J=\{0\le i\le78:\varepsilon_i=1\}. The corresponding possible summands are

{4i+1:i∈J}={1,5,13,17,25,29,37,41,53,61,65,73,85,89,97,\{4i+1:i\in J\}=\{1,5,13,17,25,29,37,41,53,61,65,73,85,89,97,
101,109,113,125,137,145,149,157,169,173,181,185,101,109,113,125,137,145,149,157,169,173,181,185,
193,197,205,221,229,233,241,257,265,269,277,281,193,197,205,221,229,233,241,257,265,269,277,281,
289,293,305,313}.289,293,305,313\}.

For 1≤a≤3131\le a\le313, a≡1(mod4)a\equiv1\pmod{4}, define the exact integer

U(a)=∑k=0H∏i=0(a−1)/4(1−εiεk−i).(20)U(a)=\sum_{k=0}^{H}\prod_{i=0}^{(a-1)/4}(1-\varepsilon_i\varepsilon_{k-i}). \tag*{(20)}

It counts the indices still unrepresented when the first summand is at most aa. Table 1 records successive values of (20); all quantities are integers.

The evaluation in the table can equivalently be performed by the following finite-set recurrence, which specifies every removal. Start with C0={0,1,…,H}C_0=\{0,1,\ldots,H\}. In increasing order, for each i∈Ji\in J replace the current set CC by

C∖{k:i≤k≤H, εk−i=1}.(21)C\setminus\{k:i\le k\le H,\ \varepsilon_{k-i}=1\}. \tag*{(21)}

The indicators themselves are obtained by starting with all indices retained and, for each of the 287 primes p≤4096p\le4096 in P3\mathcal{P}_3, removing the progression

3p−14,3p−14+p,3p−14+2p,…through H.\frac{3p-1}{4},\qquad\frac{3p-1}{4}+p,\qquad\frac{3p-1}{4}+2p,\qquad\ldots\quad\text{through }H.
aaU(a)U(a)aaU(a)U(a)aaU(a)U(a)
12 849 6431016 5612298
131 261 3341497562772
29364 708193572891
6171 867205233130

Table 1. The finite arithmetic certificate. Each column pair gives an upper bound aa for the first summand and the number U(a)U(a) of indices remaining.

There are 1,344,661 retained indices. The recurrence (21) ends with the empty set, as recorded by U(313)=0U(313)=0.

Proposition 7.2. For every n≡2(mod4)n \equiv2 \pmod{4} with 2≤n<6442 \le n < 64^4, there are a,b∈Aa,b \in\mathcal{A} such that n=a+bn=a+b. The smaller summand can be chosen at most 313.

Proof. Write n=4k+2n=4k+2, where 0≤k≤H0 \le k \le H. Since U(313)=0U(313)=0, one of the factors in the kkth product in (20) vanishes. Thus εi=εk−i=1\varepsilon_i=\varepsilon_{k-i}=1 for some 0≤i≤780 \le i \le78 with i≤ki \le k. Lemma 7.1 gives

n=(4i+1)+(4(k−i)+1)∈A+A.n=(4i+1)+(4(k-i)+1)\in\mathcal{A}+\mathcal{A}.

One of these summands is at most 313. □

Proof of Theorem 1.1. Use Lemma 2.1 to write n=g(2m0)n=g(2m_0) with g∈Ag \in\mathcal{A} and every prime divisor of m0m_0 in P3\mathcal{P}_3. Proposition 6.1 applies if 2m0≥6442m_0 \ge64^4. Proposition 7.2 applies otherwise. In both cases, multiply the resulting two summands by gg. □

Acknowledgment

OpenAI Codex generated the mathematical argument, performed the exact calculations, and drafted the manuscript under human direction.

References

Email address: [email protected]

Email address: [email protected]

Email address: [email protected]

Email address: [email protected]

References

  1. [1]P. Deligne, La conjecture de Weil. I, Publ. Math. Inst. Hautes Études Sci. 43 (1974), 273–307. doi:10.1007/BF02684373.DOI
  2. [2]R. Dietmann and C. Elsholtz, Sums of two numbers having only prime factors congruent to one modulo four, in Number Theory Week 2017, Banach Center Publications 118 (2019), 177–179. doi:10.4064/bc118-11.DOI
  3. [3]J. S. Ellenberg, P. Michel, and A. Venkatesh, Linnik’s ergodic method and the distribution of integer points on spheres, in Automorphic Representations and L-Functions, Tata Institute of Fundamental Research Studies in Mathematics 22 (2013), 119–185. Author manuscript (§8.4).
  4. [4]L. Euler, Letter to Christian Goldbach, May 6, 1747, Letter 115, in F. Lemmermeyer and M. Mattmüller (eds.), Correspondence of Leonhard Euler with Christian Goldbach, Opera Omnia, Series IVA, vol. 4, online edition, Bernoulli–Euler-Gesellschaft, Basel, 2016, 935–936. doi:10.12685/pubblbez.1.2016.
  5. [5]H. Jacquet and R. P. Langlands, Automorphic Forms on GL(2), Lecture Notes in Mathematics 114, Springer, Berlin, 1970, §16. doi:10.1007/BFb0058988.

Paper details

vlnc:W_p3mybhqd
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 13

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

OpenAI Codex

View statement
Looking for related papers…
Contents