betaFeedback

WorksMathematics

A necessary condition for a Romanoff type theorem

Abstract

Let {ak}k≥1\{a_k\}_{k\ge1} be an infinite strictly increasing sequence of positive integers with ak∣ak+1a_k\mid a_{k+1} for every integer k≥1k\ge1. Dai and Pan proved that the set of integers of the form p2+q2+ak+alp^2+q^2+a_k+a_l, where p,qp,q are primes and k,l≥1k,l\ge1, has positive lower asymptotic density if lim sup⁡k→∞log⁡akk<∞\limsup_{k\to\infty}\frac{\log a_k}{k}<\infty and sup⁡k≥1∑d∣ak1d<∞\sup_{k\ge1}\sum_{d\mid a_k}\frac1d<\infty. They also asked whether the latter condition is necessary. In this paper, we show that this condition is necessary by proving that if sup⁡k≥1∑d∣ak1d=∞\sup_{k\ge1}\sum_{d\mid a_k}\frac1d=\infty, then this set has asymptotic density zero.

Introduction

Let N\mathbb{N} be the set of all positive integers. For a set A⊆NA\subseteq\mathbb{N}, write A(x)=∣A∩[1,x]∣A(x)=|A\cap[1,x]|. Its lower and upper asymptotic densities are defined by

d‾(A)=lim inf⁡x→∞A(x)x,d‾(A)=lim sup⁡x→∞A(x)x,\underline{d}(A)=\liminf_{x\to\infty}\frac{A(x)}{x},\qquad\overline{d}(A)=\limsup_{x\to\infty}\frac{A(x)}{x},

respectively. If d‾(A)=d‾(A)\underline{d}(A)=\overline{d}(A), then its asymptotic density exists and is denoted by d(A)d(A).

In 1934, Romanoff [17] proved that the set

{p+2k:p is prime and k≥0}\{p+2^k:p\text{ is prime and }k\geq0\}

has positive lower asymptotic density. This result has been generalized in various directions; see [1, 2, 3, 4, 6, 8, 9, 10, 12, 13, 14, 16, 18, 19, 20].

In 1950, Erdős [11 Theorem 4] extended Romanoff’s theorem to strictly increasing sequences {ak}k≥1\{a_k\}_{k\geq1} of positive integers satisfying ak∣ak+1a_k\mid a_{k+1}. He proved that the set {p+ak:p is prime and k≥1}\{p+a_k:p\text{ is prime and }k\geq1\} has positive lower asymptotic density if and only if

lim sup⁡k→∞log⁡akk<∞(1)\limsup_{k\to\infty}\frac{\log a_k}{k}<\infty \tag*{(1)}

and

sup⁡k≥1∑d∣ak1d<∞.(2)\sup_{k\geq1}\sum_{d\mid a_k}\frac{1}{d}<\infty. \tag*{(2)}

Liu, Liu and Zhan [15] considered an analogue of Romanoff’s theorem for sums of two prime squares and proved that

{p2+q2+2k+2l:p,q are primes and k,l≥0}\{p^2+q^2+2^k+2^l:p,q\text{ are primes and }k,l\geq0\}

has positive lower asymptotic density. Recently, Dai and Pan [5 Theorem 1.1] extended the result of Liu, Liu and Zhan to strictly increasing sequences {ak}k≥1\{a_k\}_{k\geq1} of positive integers with ak∣ak+1a_k \mid a_{k+1} satisfying (1.1) and (1.2). They proved the following result.

Theorem 1.1 ([5], Theorem 1.1). Let {ak}k≥1\{a_k\}_{k\geq1} be an infinite strictly increasing sequence of positive integers such that ak∣ak+1a_k \mid a_{k+1} for every integer k≥1k \geq1. If {ak}k≥1\{a_k\}_{k\geq1} satisfies (1.1) and (1.2), then the set

{p2+q2+ak+al:p,q are primes and k,l≥1}\{p^2+q^2+a_k+a_l:p,q\text{ are primes and }k,l\geq1\}

has positive lower asymptotic density.

In [5 Remark], they noted that (1.1) is necessary and asked whether (1.2) is necessary. In this paper, we answer this question by proving the following result.

Theorem 1.2. Let {ak}k≥1\{a_k\}_{k\geq1} be an infinite strictly increasing sequence of positive integers such that ak∣ak+1a_k \mid a_{k+1} for every integer k≥1k \geq1. If

sup⁡k≥1∑d∣ak1d=∞,\sup_{k\geq1}\sum_{d\mid a_k}\frac{1}{d}=\infty,

then the set

{p2+q2+ak+al:p,q are primes and k,l≥1}\{p^2+q^2+a_k+a_l:p,q\text{ are primes and }k,l\geq1\}

has asymptotic density zero.

Combining Theorems 1.1 and 1.2 with the necessity of (1.1), we obtain the following result.

Corollary 1.3. Let {ak}k≥1\{a_k\}_{k\geq1} be an infinite strictly increasing sequence of positive integers such that ak∣ak+1a_k \mid a_{k+1} for every integer k≥1k \geq1. Then the set

{p2+q2+ak+al:p,q are primes and k,l≥1}\{p^2+q^2+a_k+a_l:p,q\text{ are primes and }k,l\geq1\}

has positive lower asymptotic density if and only if (1.1) and (1.2) hold.

Throughout this paper, pp and qq denote primes, π(x)\pi(x) denotes the number of primes not exceeding xx, and φ\varphi is Euler’s totient function. For functions f(x)f(x) and g(x)g(x) with g(x)>0g(x)>0, we write f(x)=O(g(x))f(x)=O(g(x)) if there exists a constant C>0C>0 such that ∣f(x)∣≤Cg(x)|f(x)|\leq Cg(x) for all sufficiently large xx, and f(x)=o(g(x))f(x)=o(g(x)) if f(x)/g(x)→0f(x)/g(x)\to0 as x→∞x\to\infty.

Preliminary lemmas

For an odd prime qq, let

χ(q)={1,q≡1(mod4),−1,q≡3(mod4).\chi(q)= \begin{cases} 1, & q\equiv1\pmod{4},\\ -1, & q\equiv3\pmod{4}. \end{cases}

Lemma 2.1. Let F\mathcal{F} be a nonempty finite set of odd primes, and let

H=∑q∈F1q,Q=∏q∈Fq,T(n)=∑q∈Fq∣nχ(q)−∑q∈Fχ(q)q.H=\sum_{q\in\mathcal{F}}\frac{1}{q},\qquad Q=\prod_{q\in\mathcal{F}}q,\qquad T(n)=\sum_{\substack{q\in\mathcal{F}\\q\mid n}}\chi(q)-\sum_{q\in\mathcal{F}}\frac{\chi(q)}{q}.

Then

∣{1≤n≤Q:T(n)≥H2}∣≤4QH.(3)\left|\left\{1\leq n\leq Q:T(n)\geq\frac{H}{2}\right\}\right|\leq\frac{4Q}{H}. \tag*{(3)}

and

1φ(Q)2∣{(u,v):1≤u,v≤Q, (uv,Q)=1, T(u2+v2)<H2}∣≤10H.(4)\frac{1}{\varphi(Q)^2}\left|\left\{(u,v):1\leq u,v\leq Q,\ (uv,Q)=1,\ T(u^2+v^2)<\frac{H}{2}\right\}\right|\leq\frac{10}{H}. \tag*{(4)}

Proof. For any primes p,q∈Fp,q \in\mathcal{F} with p≠qp \ne q, we have

∣{1≤n≤Q:p∣n}∣=Qp,∣{1≤n≤Q:pq∣n}∣=Qpq.\left|\{1 \le n \le Q : p \mid n\}\right| = \frac{Q}{p}, \qquad\left|\{1 \le n \le Q : pq \mid n\}\right| = \frac{Q}{pq}.

It follows that

1Q∑n=1QT(n)2=1Q∑n=1Q(∑q∈Fq∣nχ(q)−∑q∈Fχ(q)q)2=1Q∑n=1Q(∑q∈Fq∣nχ(q))2−(∑q∈Fχ(q)q)2=1Q∑p,q∈Fχ(p)χ(q)∑1≤n≤Qq∣n, p∣n1−∑p,q∈Fχ(p)χ(q)pq=∑q∈F1q+∑p,q∈Fp≠qχ(p)χ(q)pq−∑q∈F1q2−∑p,q∈Fp≠qχ(p)χ(q)pq=∑q∈F1q(1−1q)≤H,\begin{aligned} \frac{1}{Q}\sum_{n=1}^{Q} T(n)^2 &= \frac{1}{Q}\sum_{n=1}^{Q}\left(\sum_{\substack{q\in\mathcal{F}\\q\mid n}}\chi(q)-\sum_{q\in\mathcal{F}}\frac{\chi(q)}{q}\right)^2 \\ &= \frac{1}{Q}\sum_{n=1}^{Q}\left(\sum_{\substack{q\in\mathcal{F}\\q\mid n}}\chi(q)\right)^2-\left(\sum_{q\in\mathcal{F}}\frac{\chi(q)}{q}\right)^2 \\ &= \frac{1}{Q}\sum_{\substack{p,q\in\mathcal{F}}}\chi(p)\chi(q)\sum_{\substack{1\le n\le Q\\q\mid n,\ p\mid n}}1-\sum_{p,q\in\mathcal{F}}\frac{\chi(p)\chi(q)}{pq} \\ &= \sum_{q\in\mathcal{F}}\frac{1}{q}+\sum_{\substack{p,q\in\mathcal{F}\\p\ne q}}\frac{\chi(p)\chi(q)}{pq}-\sum_{q\in\mathcal{F}}\frac{1}{q^2}-\sum_{\substack{p,q\in\mathcal{F}\\p\ne q}}\frac{\chi(p)\chi(q)}{pq} \\ &= \sum_{q\in\mathcal{F}}\frac{1}{q}\left(1-\frac{1}{q}\right)\le H, \end{aligned}

where we used χ(q)2=1\chi(q)^2=1 in the penultimate equality. Thus

H24∣{1≤n≤Q:T(n)≥H2}∣≤∑n=1QT(n)2≤QH,\frac{H^2}{4}\left|\left\{1\le n\le Q:T(n)\ge\frac{H}{2}\right\}\right|\le\sum_{n=1}^{Q}T(n)^2\le QH,

which proves (2.1).

We next consider the integer pairs (u,v)(u,v) with 1≤u,v≤Q1\le u,v\le Q and (uv,Q)=1(uv,Q)=1. For every prime q∈Fq\in\mathcal{F}, since

∣{t:1≤t≤q−1,t2≡−1(modq)}∣=1+χ(q),\left|\left\{t:1\le t\le q-1,t^2\equiv-1\pmod q\right\}\right|=1+\chi(q),

it follows that

∣{(u,v):1≤u,v≤q−1,u2+v2≡0(modq)}∣=∣{(u,v):1≤u,v≤q−1,(uv−1)2≡−1(modq)}∣=(q−1)(1+χ(q)).\begin{aligned} \left|\left\{(u,v):1\le u,v\le q-1,u^2+v^2\equiv0\pmod q\right\}\right| &=\left|\left\{(u,v):1\le u,v\le q-1,(uv^{-1})^2\equiv-1\pmod q\right\}\right|\\ &=(q-1)(1+\chi(q)). \end{aligned}

By the Chinese remainder theorem, for any prime q∈Fq\in\mathcal{F},

1φ(Q)2∣{(u,v):1≤u,v≤Q,(uv,Q)=1, q∣u2+v2}∣=(q−1)(1+χ(q))∏p∈Fp≠q(p−1)2φ(Q)2=1+χ(q)q−1,(5)\begin{aligned} \frac{1}{\varphi(Q)^2}\left|\left\{(u,v):1\le u,v\le Q,(uv,Q)=1,\ q\mid u^2+v^2\right\}\right| &=\frac{(q-1)(1+\chi(q))\prod_{\substack{p\in\mathcal{F}\\p\ne q}}(p-1)^2}{\varphi(Q)^2}\\ &=\frac{1+\chi(q)}{q-1}, \tag*{(5)} \end{aligned}

and for any distinct primes p,q∈Fp,q \in\mathcal{F}

1φ(Q)2∣{(u,v):1≤u,v≤Q, (uv,Q)=1, pq∣u2+v2}∣=(p−1)(1+χ(p))(q−1)(1+χ(q))∏ℓ∈Fℓ≠p,q(ℓ−1)2φ(Q)2=(1+χ(p))(1+χ(q))(p−1)(q−1),(6)\begin{aligned} \frac{1}{\varphi(Q)^2}\left|\left\{(u,v):1\leq u,v\leq Q,\ (uv,Q)=1,\ pq\mid u^2+v^2\right\}\right| &=\frac{(p-1)(1+\chi(p))(q-1)(1+\chi(q))\prod_{\substack{\ell\in\mathcal{F}\\ \ell\ne p,q}}(\ell-1)^2}{\varphi(Q)^2}\\ &=\frac{(1+\chi(p))(1+\chi(q))}{(p-1)(q-1)}, \tag*{(6)} \end{aligned}

where we used φ(Q)=∏p∈F(p−1)\varphi(Q)=\prod_{p\in\mathcal{F}}(p-1). Write

m=1φ(Q)2∑1≤u,v<Q(uv,Q)=1T(u2+v2).m=\frac{1}{\varphi(Q)^2}\sum_{\substack{1\leq u,v<Q\\(uv,Q)=1}}T(u^2+v^2).

Then

m=1φ(Q)2∑1≤u,v≤Q(uv,Q)=1(∑q∈Fq∣u2+v2χ(q)−∑q∈Fχ(q)q)=∑q∈F(1+χ(q)q−1−χ(q)q)(7)\begin{aligned} m&=\frac{1}{\varphi(Q)^2}\sum_{\substack{1\leq u,v\leq Q\\(uv,Q)=1}}\left(\sum_{\substack{q\in\mathcal{F}\\q\mid u^2+v^2}}\chi(q)-\sum_{q\in\mathcal{F}}\frac{\chi(q)}{q}\right)\\ &=\sum_{q\in\mathcal{F}}\left(\frac{1+\chi(q)}{q-1}-\frac{\chi(q)}{q}\right) \tag*{(7)} \end{aligned}
m=∑q∈F(1q+1+χ(q)q(q−1))=H+∑q∈F1+χ(q)q(q−1)≥H,(8)\begin{aligned} m&=\sum_{q\in\mathcal{F}}\left(\frac{1}{q}+\frac{1+\chi(q)}{q(q-1)}\right)\\ &=H+\sum_{q\in\mathcal{F}}\frac{1+\chi(q)}{q(q-1)}\geq H, \tag*{(8)} \end{aligned}

where we used (2.3) and χ(q)2=1\chi(q)^2=1 in the second equality and 1+χ(q)≥01+\chi(q)\geq0 for every q∈Fq\in\mathcal{F} in the last inequality. Thus, by (7),

T(u2+v2)−m=∑q∈Fq∣u2+v2χ(q)−∑q∈Fχ(q)q−∑q∈F(1+χ(q)q−1−χ(q)q)=∑q∈Fq∣u2+v2χ(q)−∑q∈F1+χ(q)q−1.\begin{aligned} T(u^2+v^2)-m &=\sum_{\substack{q\in\mathcal{F}\\q\mid u^2+v^2}}\chi(q)-\sum_{q\in\mathcal{F}}\frac{\chi(q)}{q}-\sum_{q\in\mathcal{F}}\left(\frac{1+\chi(q)}{q-1}-\frac{\chi(q)}{q}\right)\\ &=\sum_{\substack{q\in\mathcal{F}\\q\mid u^2+v^2}}\chi(q)-\sum_{q\in\mathcal{F}}\frac{1+\chi(q)}{q-1}. \end{aligned}

Hence,

1φ(Q)2∑1≤u,v≤Q(uv,Q)=1(T(u2+v2)−m)2=1φ(Q)2∑1≤u,v≤Q(uv,Q)=1(∑q∈Fq∣u2+v2χ(q))2−(∑q∈F1+χ(q)q−1)2=∑q∈F1+χ(q)q−1+∑p,q∈Fp≠q(1+χ(p))(1+χ(q))(p−1)(q−1)−(∑q∈F1+χ(q)q−1)2=∑q∈F(1+χ(q)q−1−(1+χ(q))2(q−1)2)≤∑q∈F1+χ(q)q−1≤52H,(9)\begin{aligned} \frac{1}{\varphi(Q)^2}\sum_{\substack{1\leq u,v\leq Q\\(uv,Q)=1}}(T(u^2+v^2)-m)^2 &=\frac{1}{\varphi(Q)^2}\sum_{\substack{1\leq u,v\leq Q\\(uv,Q)=1}}\left(\sum_{\substack{q\in\mathcal{F}\\q\mid u^2+v^2}}\chi(q)\right)^2-\left(\sum_{q\in\mathcal{F}}\frac{1+\chi(q)}{q-1}\right)^2\\ &=\sum_{q\in\mathcal{F}}\frac{1+\chi(q)}{q-1}+\sum_{\substack{p,q\in\mathcal{F}\\p\ne q}}\frac{(1+\chi(p))(1+\chi(q))}{(p-1)(q-1)}-\left(\sum_{q\in\mathcal{F}}\frac{1+\chi(q)}{q-1}\right)^2\\ &=\sum_{q\in\mathcal{F}}\left(\frac{1+\chi(q)}{q-1}-\frac{(1+\chi(q))^2}{(q-1)^2}\right)\\ &\leq\sum_{q\in\mathcal{F}}\frac{1+\chi(q)}{q-1}\leq\frac{5}{2}H, \tag*{(9)} \end{aligned}

where the first equality follows from (2.3), the second equality follows from (2.3)–(2.4), and the last inequality follows from 1+χ(3)=01+\chi(3)=0 and

1+χ(q)q−1≤2q−1≤52q(q≥5).\frac{1+\chi(q)}{q-1}\leq\frac{2}{q-1}\leq\frac{5}{2q}\qquad(q\geq5).

Noting that H≤mH\leq m in (2.6), if T(u2+v2)<H/2T(u^2+v^2)<H/2, then

∣T(u2+v2)−m∣=m−T(u2+v2)>m−H2≥H2.|T(u^2+v^2)-m|=m-T(u^2+v^2)>m-\frac{H}{2}\geq\frac{H}{2}.

It follows from (2.7) that

1φ(Q)2∣{(u,v):1≤u,v≤Q, (uv,Q)=1, T(u2+v2)<H2}∣≤1φ(Q)2∣{(u,v):1≤u,v≤Q, (uv,Q)=1, ∣T(u2+v2)−m∣>H2}∣≤4H2φ(Q)2∑1≤u,v≤Q(uv,Q)=1(T(u2+v2)−m)2≤4H2⋅52H=10H,\begin{aligned} \frac{1}{\varphi(Q)^2}\left|\left\{(u,v):1\leq u,v\leq Q,\ (uv,Q)=1,\ T(u^2+v^2)<\frac{H}{2}\right\}\right| \leq\frac{1}{\varphi(Q)^2}\left|\left\{(u,v):1\leq u,v\leq Q,\ (uv,Q)=1,\ |T(u^2+v^2)-m|>\frac{H}{2}\right\}\right| \\ &\leq\frac{4}{H^2\varphi(Q)^2}\sum_{\substack{1\leq u,v\leq Q\\(uv,Q)=1}}(T(u^2+v^2)-m)^2 \\ &\leq\frac{4}{H^2}\cdot\frac{5}{2}H=\frac{10}{H}, \end{aligned}

which proves (2.2). This completes the proof of Lemma 2.1. □

Lemma 2.2. Let F\mathcal{F} be a fixed nonempty finite set of odd primes, and let H,Q,TH,Q,T be as in Lemma 2.1. As X→∞X\to\infty, we have

∣{1≤n≤X:T(n)≥H2}∣≤4XH+OF(1)(10)\left|\left\{1\leq n\leq X:T(n)\geq\frac{H}{2}\right\}\right|\leq\frac{4X}{H}+O_{\mathcal{F}}(1) \tag*{(10)}

and

∣{(p,q):p,q≤X, T(p2+q2)<H2}∣≤(10H+oF(1))π(X)2.(11)\left|\left\{(p,q):p,q\leq\sqrt{X},\ T(p^2+q^2)<\frac{H}{2}\right\}\right|\leq\left(\frac{10}{H}+o_{\mathcal{F}}(1)\right)\pi(\sqrt{X})^2. \tag*{(11)}

Proof. Clearly, T(n)T(n) is periodic modulo QQ. It follows from (3) that

∣{1≤n≤X:T(n)≥H2}∣≤4XH+OF(1),\left|\left\{1\le n\le X:T(n)\ge\frac{H}{2}\right\}\right|\le\frac{4X}{H}+O_F(1),

which proves (2.8).

For any integer uu with (u,Q)=1(u,Q)=1, the prime number theorem for arithmetic progressions gives

∣{p≤X:p≡u(modQ)}∣=π(X)φ(Q)+oF(π(X))(X→∞);\left|\{p\le\sqrt{X}:p\equiv u\pmod{Q}\}\right|=\frac{\pi(\sqrt{X})}{\varphi(Q)}+o_F(\pi(\sqrt{X}))\qquad(X\to\infty);

see [7 Chapter 20]. Then, for any integers u,vu,v with (uv,Q)=1(uv,Q)=1,

∣{(p,q):p,q≤X, p≡u(modQ), q≡v(modQ)}∣=∣{p≤X:p≡u(modQ)}∣∣{q≤X:q≡v(modQ)}∣=(1φ(Q)+oF(1))(1φ(Q)+oF(1))π(X)2=(1φ(Q)2+oF(1))π(X)2.(12)\begin{aligned} &\left|\{(p,q):p,q\le\sqrt{X},\ p\equiv u\pmod{Q},\ q\equiv v\pmod{Q}\}\right|\\ &=\left|\{p\le\sqrt{X}:p\equiv u\pmod{Q}\}\right|\left|\{q\le\sqrt{X}:q\equiv v\pmod{Q}\}\right|\\ &=\left(\frac{1}{\varphi(Q)}+o_F(1)\right)\left(\frac{1}{\varphi(Q)}+o_F(1)\right)\pi(\sqrt{X})^2\\ &=\left(\frac{1}{\varphi(Q)^2}+o_F(1)\right)\pi(\sqrt{X})^2. \tag*{(12)} \end{aligned}

For any pair of primes (p,q)(p,q), if (pq,Q)=1(pq,Q)=1, then there exist integers 1≤u,v<Q1\le u,v<Q with (uv,Q)=1(uv,Q)=1 such that p≡u(modQ)p\equiv u\pmod{Q} and q≡v(modQ)q\equiv v\pmod{Q}, and so p2+q2≡u2+v2(modQ)p^2+q^2\equiv u^2+v^2\pmod{Q} and T(p2+q2)=T(u2+v2)T(p^2+q^2)=T(u^2+v^2). Hence,

∣{(p,q):p,q≤X, (pq,Q)=1, T(p2+q2)<H2}∣=∑1≤u,v≤Q(uv,Q)=1T(u2+v2)<H/2∣{(p,q):p,q≤X, p≡u(modQ), q≡v(modQ)}∣=(1φ(Q)2∑1≤u,v≤Q(uv,Q)=1T(u2+v2)<H/21+oF(1))π(X)2≤(10H+oF(1))π(X)2,\begin{aligned} \left|\left\{(p,q):p,q\le\sqrt{X},\ (pq,Q)=1,\ T(p^2+q^2)<\frac{H}{2}\right\}\right| \\ &=\sum_{\substack{1\le u,v\le Q\\(uv,Q)=1\\T(u^2+v^2)<H/2}}\left|\{(p,q):p,q\le\sqrt{X},\ p\equiv u\pmod{Q},\ q\equiv v\pmod{Q}\}\right| \\ &=\left(\frac{1}{\varphi(Q)^2}\sum_{\substack{1\le u,v\le Q\\(uv,Q)=1\\T(u^2+v^2)<H/2}}1+o_F(1)\right)\pi(\sqrt{X})^2 \\ &\le\left(\frac{10}{H}+o_F(1)\right)\pi(\sqrt{X})^2, \end{aligned}

where the last equality follows from (2.10), and the last inequality follows from (2.2). Since

∣{(p,q):p,q≤X, (pq,Q)>1, T(p2+q2)<H2}∣≤∣{(p,q):q≤X, p∣Q}∣+∣{(p,q):p≤X, q∣Q}∣\begin{aligned} \left|\left\{(p,q):p,q\le\sqrt{X},\ (pq,Q)>1,\ T(p^2+q^2)<\frac{H}{2}\right\}\right| \\ &\le\left|\{(p,q):q\le\sqrt{X},\ p\mid Q\}\right|+\left|\{(p,q):p\le\sqrt{X},\ q\mid Q\}\right| \end{aligned}
≤2∣F∣π(X)=o(π(X)2),\le2|\mathcal{F}|\pi(\sqrt{X})=o\left(\pi(\sqrt{X})^2\right),

it follows that

∣{(p,q):p,q≤X, T(p2+q2)<H2}∣≤(10H+oF(1))π(X)2,\left|\left\{(p,q):p,q\le\sqrt{X},\ T(p^2+q^2)<\frac{H}{2}\right\}\right| \le\left(\frac{10}{H}+o_F(1)\right)\pi(\sqrt{X})^2,

which proves (2.9). This completes the proof of Lemma 2.2.

Proofs of the main results

Throughout this section, let A={ak:k≥1}A=\{a_k:k\ge1\} and let

S(A)={p2+q2+ak+al:p,q are primes and k,l≥1}.S(A)=\{p^2+q^2+a_k+a_l:p,q\text{ are primes and }k,l\ge1\}.

Since the sequence is strictly increasing, we have A(X)=∣{k:ak≤X}∣A(X)=|\{k:a_k\le X\}|.

Proof of Theorem 1.2. Let

P={q:q∣ak for some k≥1}.\mathcal{P}=\{q:q\mid a_k\text{ for some }k\ge1\}.

For every integer k≥1k\ge1, write

ak=∏q∣akqνq,νq≥1.a_k=\prod_{q\mid a_k}q^{\nu_q},\qquad\nu_q\ge1.

Since

∑j=0∞1qj=(1−1q)−1,\sum_{j=0}^{\infty}\frac{1}{q^j}=\left(1-\frac{1}{q}\right)^{-1},

it follows that

log⁡(∑d∣ak1d)=log⁡(∏q∣ak(1+1q+1q2+⋯+1qνq))≤log⁡(∏q∣ak(1−1q)−1)=∑q∣aklog⁡(1+1q−1)≤∑q∣ak1q−1≤2∑q∣ak1q.\begin{aligned} \log\left(\sum_{d\mid a_k}\frac{1}{d}\right) &=\log\left(\prod_{q\mid a_k}\left(1+\frac{1}{q}+\frac{1}{q^2}+\cdots+\frac{1}{q^{\nu_q}}\right)\right)\\ &\le\log\left(\prod_{q\mid a_k}\left(1-\frac{1}{q}\right)^{-1}\right)\\ &=\sum_{q\mid a_k}\log\left(1+\frac{1}{q-1}\right)\\ &\le\sum_{q\mid a_k}\frac{1}{q-1}\le2\sum_{q\mid a_k}\frac{1}{q}. \end{aligned}

Since

sup⁡k≥1∑d∣ak1d=∞,\sup_{k\ge1}\sum_{d\mid a_k}\frac{1}{d}=\infty,

it follows that

∑q∈P1q=∞.(13)\sum_{q\in\mathcal{P}}\frac{1}{q}=\infty. \tag*{(13)}

Fix a nonempty finite set F⊆P∖{2}\mathcal{F}\subseteq\mathcal{P}\setminus\{2\}, and let H,Q,TH,Q,T be as in Lemma 2.1. Since F\mathcal{F} is a finite set and ak∣ak+1a_k\mid a_{k+1} for every integer k≥1k\ge1, it follows that there exists K≥1K\ge1 such that Q∣aKQ\mid a_K, and so

Q∣ak(k≥K),T(p2+q2+ak+al)=T(p2+q2)(k,l≥K).(14)Q\mid a_k\quad(k\ge K),\qquad T(p^2+q^2+a_k+a_l)=T(p^2+q^2)\quad(k,l\ge K). \tag*{(14)}

Since the positive sequence {ak}k≥1\{a_k\}_{k\ge1} is strictly increasing and ak∣ak+1a_k\mid a_{k+1} for every integer k≥1k\ge1, it follows that

ak≥2ak−1≥⋯≥2k−1a1≥2k−1.a_k\ge2a_{k-1}\ge\cdots\ge2^{k-1}a_1\ge2^{k-1}.

Therefore

A(X)≤1+log⁡Xlog⁡2.(15)A(X)\le1+\frac{\log X}{\log2}. \tag*{(15)}

If n≤Xn \le X and n=p2+q2+ak+aln=p^{2}+q^{2}+a_{k}+a_{l}, then p,q≤Xp,q \le\sqrt{X} and ak,al≤Xa_{k},a_{l} \le X. The number of integers of the form n=p2+q2+ak+aln=p^{2}+q^{2}+a_{k}+a_{l} with n≤Xn \le X and either k<Kk<K or l<Kl<K is at most

2(K−1)A(X)π(X)2=OF(Xlog⁡X)=oF(X).(16)2(K-1)A(X)\pi(\sqrt{X})^{2}=O_{F}\left(\frac{X}{\log X}\right)=o_{F}(X). \tag*{(16)}

For any integer of the form n=p2+q2+ak+aln=p^{2}+q^{2}+a_{k}+a_{l} with n≤Xn \le X and k,l≥Kk,l \ge K, we consider two cases.

Case 1. T(n)≥H/2T(n) \ge H/2. By (10), the number of such integers is at most

∣{1≤n≤X:T(n)≥H2}∣≤4XH+OF(1).(17)\left|\left\{1 \le n \le X:T(n) \ge\frac{H}{2}\right\}\right| \le\frac{4X}{H}+O_{F}(1). \tag*{(17)}

Case 2. T(n)<H/2T(n)<H/2. By (14) and (11), the number of such integers is at most

A(X)2∣{(p,q):p,q≤X, T(p2+q2)<H2}∣≤A(X)2(10H+oF(1))π(X)2.(18)A(X)^{2}\left|\left\{(p,q):p,q \le\sqrt{X},\ T(p^{2}+q^{2})<\frac{H}{2}\right\}\right| \le A(X)^{2}\left(\frac{10}{H}+o_{F}(1)\right)\pi(\sqrt{X})^{2}. \tag*{(18)}

By (15) and the prime number theorem,

A(X)2π(X)2X≤(1+log⁡Xlog⁡2)24+o(1)(log⁡X)2=(4+o(1))(1log⁡X+1log⁡2)2=4(log⁡2)2+o(1).\begin{aligned} \frac{A(X)^{2}\pi(\sqrt{X})^{2}}{X}\le\left(1+\frac{\log X}{\log2}\right)^{2}\frac{4+o(1)}{(\log X)^{2}} \\ &= (4+o(1))\left(\frac{1}{\log X}+\frac{1}{\log2}\right)^{2} \\ &= \frac{4}{(\log2)^{2}}+o(1). \end{aligned}

It follows from (16), (17) and (18) that

S(A)(X)X≤1H(4+40(log⁡2)2)+oF(1).\frac{S(A)(X)}{X}\le\frac{1}{H}\left(4+\frac{40}{(\log2)^{2}}\right)+o_{F}(1).

Then

d‾(S(A))≤1H(4+40(log⁡2)2).(19)\overline{d}(S(A))\le\frac{1}{H}\left(4+\frac{40}{(\log2)^{2}}\right). \tag*{(19)}

By (13), the finite set F\mathcal{F} can be chosen such that HH is arbitrarily large. Letting H→∞H\to\infty in (19), we have

d‾(S(A))=0.\overline{d}(S(A))=0.

This completes the proof of Theorem 1.2. □

Proof of Corollary 1.3. The sufficiency follows from Theorem 1.1.

Suppose that d‾(S(A))>0\underline{d}(S(A))>0. By Theorem 1.2, condition (1.2) holds. For completeness, we give the short counting argument for (1.1).

If (1.1) does not hold, there are indices j→∞j\to\infty such that log⁡aj/j→∞\log a_{j}/j\to\infty. Let Xj=aj−1X_{j}=a_{j}-1. Since A(Xj)=j−1A(X_{j})=j-1 and log⁡Xj∼log⁡aj\log X_{j}\sim\log a_{j}, it follows that

S(A)(Xj)Xj≤(j−1)2π(Xj)2Xj=o((jlog⁡Xj)2)=o(1),\frac{S(A)(X_{j})}{X_{j}}\le\frac{(j-1)^{2}\pi(\sqrt{X_{j}})^{2}}{X_{j}}=o\left(\left(\frac{j}{\log X_{j}}\right)^{2}\right)=o(1),

a contradiction with d‾(S(A))>0\underline{d}(S(A))>0. Hence (1.1) holds. This completes the proof of Corollary 1.3. □

Acknowledgments

This work was supported by the National Natural Science Foundation of China (Grant No. 12371005).

References

  1. [1]Ballot, C., Luca, F.: On the sumset of the primes and a linear recurrence. Acta Arith. 161, 33–46 (2013).DOI
  2. [2]Chen, Y.-G.: Romanoff theorem in a sparse set. Sci. China Math. 53, 2195–2202 (2010).
  3. [3]Chen, Y.-G., Sun, X.-G.: On Romanoff’s constant. J. Number Theory 106, 275–284 (2004).DOI
  4. [4]Chen, Y.-G., Xu, J.-Z.: On integers of the form p + 2^{k_1}^{r_1} + ··· + 2^{k_t}^{r_t}. J. Number Theory 258, 66–93 (2024).
  5. [5]Dai, L.-X., Pan, H.: On Erdős’ generalization of the Romanoff theorem. J. Number Theory 291, 65–73 (2027).DOI
  6. [6]Dai, L.-X., Pan, H.: Polynomial analogue of Erdős’ extension of the Romanoff theorem. J. Number Theory 279, 45–55 (2026).
  7. [7]Davenport, H.: Multiplicative Number Theory, 3rd edn. Revised by H.L. Montgomery. Graduate Texts in Mathematics, vol. 74. Springer, New York (2000).
  8. [8]Ding, Y.-C., Verwee, J.: Short intervals for the Romanoff-type sumset. J. Number Theory 289, 114–131 (2026).DOI
  9. [9]Ding, Y.-C., Zhou, H.-Y.: Romanoff’s theorem for polynomials over finite fields revisited. Acta Arith. 200, 1–15 (2021).DOI
  10. [10]Dubickas, A.: Sums of primes and quadratic linear recurrence sequences. Acta Math. Sin. (Engl. Ser.) 29, 2251–2260 (2013).
  11. [11]Erdős, P.: On integers of the form 2^k + p and some related problems. Summa Brasil. Math. 2, 113–123 (1950).
  12. [12]Kuan, Y.-L.: A function field analogue of Romanoff’s theorem. Int. J. Number Theory 11, 2161–2173 (2015).DOI
  13. [13]Lee, K.S.E.: On the sum of a prime and a Fibonacci number. Int. J. Number Theory 6, 1669–1676 (2010).arxiv.org/abs/1011.0173
  14. [14]Li, H.-Z., Pan, H.: The Romanoff theorem revisited. Acta Arith. 135, 137–142 (2008).DOI
  15. [15]Liu, J.-Y., Liu, M.-C., Zhan, T.: Squares of primes and powers of 2. Monatsh. Math. 128, 283–313 (1999).DOI
  16. [16]Liu, Z.-X., Xue, M.-Y.: The sum of a prime and a Fibonacci number. Int. J. Number Theory 17, 1815–1823 (2021).DOI
  17. [17]Romanoff, N.P.: Über einige Sätze der additiven Zahlentheorie. Math. Ann. 109, 668–678 (1934).
  18. [18]Shparlinski, I.E., Weingartner, A.J.: An explicit polynomial analogue of Romanoff’s theorem. Finite Fields Appl. 44, 22–33 (2017).arxiv.org/abs/1510.08991
  19. [19]Sun, X.-G., Chen, Y.-G.: On quantitative Romanoff’s theorem. Acta Math. Sin. (Chin. Ser.) 49, 577–582 (2006).
  20. [20]Yang, Q.-H., Chen, Y.-G.: On the integers of the form p + b. Taiwanese J. Math. 18, 1623–1631 (2014).DOI

Paper details

vlnc:W_z9vtaz3k
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

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

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents