Abstract
We obtain a lower bound for \(\#\{x/2<p_n\leq x \colon\, \, p_n\equiv\dots\equiv p_{n+m}\equiv a\pmod{q}\), \(p_{n+m} - p_n\leq y\}\), where \(p_n\) is the \(n\)th prime.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1. Introduction
Let \(p_n\) denote the \(n\)th prime. We prove the following result.
Theorem 1.1.
There are positive absolute constants \(c\) and \(C\) such that the following holds. Let \(\varepsilon\) be a real number with \(0<\varepsilon<1\). Then there is a number \(c_0(\varepsilon)>0,\) depending only on \(\varepsilon,\) such that if \(x\in{\mathbb R},\) \(y\in{\mathbb R},\) \(m\in{\mathbb Z},\) \(q\in{\mathbb Z},\) and \(a\in{\mathbb Z}\) satisfy the conditions
then
Theorem 1.1 extends a result of Maynard [5, Theorem 3.3], who established the same result but with \(y=\varepsilon\ln x\).
From Theorem 1.1 we obtain
Corollary 1.1.
There is an absolute constant \(C>0\) such that if \(m\) is a positive integer and \(x\) and \(y\) are real numbers satisfying \(\exp(Cm)\leq y \leq \ln x,\) then
Let us introduce necessary notation. The expression \(b\mid a\) means that \(b\) divides \(a\). For a fixed \(a\) the sum \(\sum_{b\mid a}\) and the product \(\prod_{b\mid a}\) should be interpreted as being over all positive divisors of \(a\).
We will use I. M. Vinogradov’s notation: \(A\ll B\) means that \(|A|\leq cB\) with a positive absolute constant \(c\).
We reserve the letter \(p\) for primes. In particular, the sum \(\sum_{p\leq K}\) should be interpreted as being over all prime numbers not exceeding \(K\).
We will also use the following notation:
\(\#A\) is the number of elements of a finite set \(A\);
\({\mathbb N}\), \({\mathbb Z}\), \({\mathbb R}\), and \({\mathbb C}\) are the sets of all positive integers, integers, real numbers, and complex numbers;
\(\mathbb{P}\) is the set of all prime numbers;
\([x]\) is the integer part of a number \(x\); i.e., \([x]\) is the largest integer \(n\) such that \(n\leq x\);
\(\{x\}\) is the fractional part of a number \(x\); i.e., \(\{x\}=x-[x]\);
\(\lceil x\rceil\) is the smallest integer \(n\) such that \(n\geq x\);
\(\operatorname{Re} s\) and \(\operatorname{Im} s\) are the real and imaginary parts of a complex number \(s\);
\((a_1,\dots,a_n)\) is the greatest common divisor of integers \(a_1,\dots,a_n\);
\([a_1,\dots,a_n]\) is the least common multiple of integers \(a_1,\dots,a_n\);
\( \varphi (n)\) is the Euler totient function: \( \varphi (n)= \#\{1\leq m \leq n \colon\, (m,n)=1\}\);
\(\mu(n)\) is the Möbius function, which is defined as follows:
-
(i)
\(\mu(1)=1\),
-
(ii)
\(\mu(n)=0\) if there is a prime \(p\) such that \(p^2\mid n\), and
-
(iii)
\(\mu(n)=(-1)^s\) if \(n=q_1\dots q_s\), where \(q_1<\dots<q_s\) are primes;
\( \Lambda (n)\) is the von Mangoldt function:
\(P^-(n)\) is the least prime factor of \(n>1\) (by convention \(P^-(1)=+\infty\));
\(\binom{n}{k} = n!/(k!\,(n-k)!)\) is the binomial coefficient.
For real numbers \(a\) and \(b\) we use \((a,b)\) and \([a,b]\) to denote, respectively, the open and closed intervals with endpoints \(a\) and \(b\). By \((a_1,\dots,a_n)\) we also denote a vector; the meaning of the notation should be clear from the context.
By definition, we put
We define
We will use the following functions:
Let \(m>1\) and \(a\) be integers. If \((a,m)=1\), then \(a^{ \varphi (m)}\equiv 1\pmod{m}\) (the Fermat–Euler theorem). Let \(d\) be the smallest positive value of \(\gamma\) for which \(a^{\gamma}\equiv 1\pmod{m}\). We call \(d\) the order of \(a\pmod{m}\) and say that \(a\) belongs to \(d\pmod{m}\).
Let \(q\) be a positive integer. We recall that a Dirichlet character modulo \(q\) is a function \(\chi \colon\, {\mathbb Z}\to{\mathbb C}\) such that
-
(1)
\(\chi(n+q)=\chi(n)\) for all \(n\in{\mathbb Z}\) (i.e., \(\chi\) is a periodic function with period \(q\));
-
(2)
\(\chi(mn)=\chi(m)\chi(n)\) for all \(m, n\in{\mathbb Z}\) (i.e., \(\chi\) is a totally multiplicative function);
-
(3)
\(\chi(1) =1\);
-
(4)
\(\chi(n)= 0\) for all \(n\in{\mathbb Z}\) such that \((n,q)>1\).
By \(X_q\) we denote the set of all Dirichlet characters modulo \(q\). We recall that \(\# X_q = \varphi (q)\) and that the principal character modulo \(q\) is
Let \(\chi\in X_q\). We say that the character \(\chi\) restricted by \((n,q)=1\) has period \(q_1\) if it has the property that \(\chi(m)=\chi(n)\) for all \(m,n\in{\mathbb Z}\) such that \((m,q)=1\), \((n,q)=1\) and \(m\equiv n\pmod{q_1}\). Let \(c(\chi)\) denote the conductor of \(\chi\), which is the least positive integer \(q_1\) such that \(\chi\) restricted by \((n,q)=1\) has period \(q_1\). We say that \(\chi\) is primitive if \(c(\chi)=q\), and imprimitive if \(c(\chi)<q\). By \(X^*_q\) we denote the set of all primitive characters modulo \(q\). We observe that the principal character modulo \(1\) is primitive. On the other hand, any principal character modulo \(q>1\) is imprimitive, since its conductor is clearly \(1\). For \(\chi\in X_q\) we put
A character \(\chi\) is said to be real if \(\chi(n)\in{\mathbb R}\) for all \(n\in{\mathbb Z}\). A character \(\chi\) is said to be complex if there is an integer \(n\) such that \(\operatorname{Im}(\chi(n))\neq 0\).
We say that characters \(\chi^{}_1\) and \(\chi^{}_2\) (modulo \(q_1\) and modulo \(q_2\), respectively) are equal and write \(\chi^{}_1 = \chi^{}_2\) if \(\chi^{}_1(n)=\chi^{}_2(n)\) for any integer \(n\). Otherwise, we say that characters \(\chi^{}_1\) and \(\chi^{}_2\) are not equal and write \(\chi^{}_1 \neq \chi^{}_2\).
Let \(\chi\) be a Dirichlet character modulo \(q\). The corresponding \(L\)-function is defined by the series
for \(s\in{\mathbb C}\) with \(\operatorname{Re} s>1\). It is well known that if \(\chi\) is not the principal character modulo \(q\), then \(L(s,\chi)\) can be analytically continued to \({\mathbb C}\). If \(\chi\) is the principal character modulo \(q\), then \(L(s,\chi)\) can be analytically continued to \({\mathbb C}\setminus \{1\}\) with a simple pole at \(s=1\).
We say that two linear functions \(L_1(n)=a_1n+b_1\) and \(L_2(n)=a_2n+b_2\) with integer coefficients are equal and write \(L_1 = L_2\) if \(a_1=a_2\) and \(b_1=b_2\). Otherwise, we say that the linear functions \(L_1\) and \(L_2\) are not equal and write \(L_1 \neq L_2\).
Let \( {\mathcal L} =\{L_1,\dots,L_k\}\) be a set of \(k\) linear functions with integer coefficients:
For \(L(n)=a n+b,\) \(a,b\in{\mathbb Z}\), we define
We say that \(L(n)=an+b\) belongs to \( {\mathcal L} \) (\(L\in {\mathcal L} \)) if there is an \(i\), \(1 \leq i \leq k\), such that \(L=L_i\). Otherwise, we say that \(L(n)=an+b\) does not belong to \( {\mathcal L} \) (\(L\notin {\mathcal L} \)).
This paper is organized as follows. In Sections 2–4 we give necessary lemmas. In Section 5 we prove Theorem 1.1 and Corollary 1.1.
2. Preparatory lemmas
In this section we present some well-known lemmas which will be used in the following sections.
Lemma 2.1 (see, for example, [6, Ch. 1]).
Let \(x\) be a real number with \(x\geq 2\). Then
where \(b_i,\) \(i=1,\dots,4,\) are positive absolute constants.
Lemma 2.2 (see, for example, [4, Chs. 1, 2]).
The limits \(\lim_{x\to+\infty}\psi(x)/x,\) \(\lim_{x\to+\infty}\theta(x)/x,\) \(\lim_{x\to+\infty}\pi(x)/(x/ \ln x),\) and \(\lim_{n\to+\infty} p_n/(n\ln n)\) exist and
From Lemma 2.2 we obtain
Lemma 2.3.
It holds that
where \(b_i,\) \(i=5,\dots,12,\) are positive absolute constants.
Lemma 2.4 (see, for example, [7, Ch. 2]).
Let \(n\) be an integer with \(n>1\). Then
From Lemma 2.4 we readily obtain the following two lemmas.
Lemma 2.5.
Let \(m\) and \(n\) be integers with \(m\geq 1\) and \(n\geq 1\). Then
Lemma 2.6.
Let \(m\) and \(n\) be integers with \(m\geq 1,\) \(n\geq 1,\) and \((m,n)=1\). Then
Lemma 2.7.
Let \(n\) be an integer with \(n\geq 1\). Then
Proof.
For \(n=1\), equality (2.3) holds. Let \(n>1\). Let us express \(n\) in the standard form \(n=q_1^{\alpha_1}\dots q_r^{\alpha_r}\), where \(q_1<\dots<q_r\) are prime numbers. Applying Lemmas 2.4 and 2.6, we have
Lemma 2.8 (see, for example, [6, Ch. 1]).
Let \(n\) be an integer with \(n\geq 1\). Then
where \(c>0\) is an absolute constant.
Lemma 2.9 (see, for example, [1, Ch. 28]).
Let \(x\) be a real number with \(x \geq 2\). Then
where \(c>0\) is an absolute constant.
Lemma 2.10 (see, for example, [2, Ch. 5]).
Let \(n\) be an integer with \(n\geq 1\). Then
where \(c>0\) is an absolute constant.
Lemma 2.11.
Let \(a,\) \(b,\) and \(c\) be integers such that \((a,b)\mid c\). Then the equation
has a solution in integers.
Proof.
We put \(d=(a,b)\). Then \(c=dl\) for some \(l\in{\mathbb Z}\). It is well known (see, for example, [7, Ch. 1, Exercise 1]) that the equation
has a solution in integers. Let \(x_0\in{\mathbb Z}\) and \(y_0\in{\mathbb Z}\) be a solution of (2.5). Then the integers \(lx_0\) and \(ly_0\) satisfy (2.4). \(\quad\Box\)
Lemma 2.12.
Let \(n\) and \(k\) be integers such that \(1 \leq k \leq n\). Then
Proof.
For \(k=n\) inequality (2.6) holds. Let \(1\leq k < n\). Then
Lemma 2.13 (see [3, Ch. 0]).
Let \(x\) and \(z\) be real numbers such that \(2 \leq z \leq x/2\). Then
where \(c_0>0\) is an absolute constant.
3. Lemmas on Dirichlet characters
In this section we give some well-known lemmas on Dirichlet characters which will be used in the following sections.
Lemma 3.1.
Let \(a,\) \(b,\) and \(n\) be integers such that \(1\leq a<b,\) \(a\mid b,\) and \((n,a)=1\). Then there is an integer \(t\) such that \((n+ta,b)=1\).
Proof.
If \((n,b)=1\), we take \(t=0\). Let \((n,b)>1\). Then the set \(\Omega = \{ p\mid b \colon\, p \nmid a\}\) is nonempty. Let \(\Omega = \{ q_1,\dots,q_r\}\) with \(q_1<\dots<q_r\). Let \(1 \leq i \leq r\). Since \((a,q_i)=1\), the congruence
has a solution; i.e., there is an integer \(m_i\) such that \(n+a m_i \equiv 1 \pmod{q_i}\). Consider the system
Since the numbers \(q_1,\dots,q_r\) are coprime, the system has a solution. Let an integer \(t_0\) satisfy system (3.1). We claim that \(t_0\) is a desired number, i.e., that \((n+t_0a,b)=1\). Assume the contrary: \((n+t_0a,b)>1\). Then there is a prime \(p\) such that \(p\mid b\) and \(p\mid (n+t_0a)\). If \(p\nmid a\), then \(p\in\Omega\), i.e., \(p=q_i\) for some \(1 \leq i \leq r\). However,
and hence \(p\nmid (n+t_0a)\). We arrive at a contradiction. Thus this case is impossible. Hence, \(p\mid a\). Since \(p\mid (n+t_0a)\), we see that \(p\mid n\). Hence, \((n,a)>1\). This contradicts the hypothesis of the lemma. Therefore, the assumption \((n+t_0a,b)>1\) is false. Hence, \((n+t_0a,b)=1\). \(\quad\Box\)
Lemma 3.2.
Let \(q\ge 2\) be an integer and \(\chi\in X_q\). Suppose that \(\chi\) restricted by \((n,q)=1\) has period \(q_1\). Then \(\chi\) restricted by \((n,q)=1\) also has period \((q,q_1)\).
Proof.
We put \(\delta = (q,q_1)\). Let \(m\) and \(n\) be integers such that \((m,q)=1\), \((n,q)=1\), and \(m\equiv n\pmod{\delta}\). We need to prove that \(\chi(m) = \chi(n)\). By Lemma 2.11, there are integers \(k\) and \(l\) such that
We put \(A=m+ q_1 k= n+ql\). Since \((n,q)=1\), we have \((n+ql, q)=1\). Hence, \((A,q) =1\). Since \(\chi\) has period \(q\), it follows that
Since \((A,q)=1\), \((m,q)=1\), and \(A\equiv m\pmod{q_1}\), we have \(\chi(A) = \chi(m)\). Hence, \(\chi(m)= \chi(n)\). \(\quad\Box\)
Lemma 3.3.
Let \(q\geq 1\) and \(\chi\in X_q\). Then \(c(\chi)\) divides \(q\).
Proof.
If \(q=1\), then \(c(\chi)=1\) and the statement is obvious. Let \(q\geq 2\). By Lemma 3.2, \(\chi\) restricted by \((n,q)=1\) has period \(\delta = (c(\chi), q)\). If \(c(\chi)\) is not a divisor of \(q\), then \(\delta< c(\chi)\), which contradicts the definition of the conductor. \(\quad\Box\)
Lemma 3.4.
Let \(q\geq 1\) and \(\chi\in X_q\). Then there exists a unique Dirichlet character \(\chi^{}_1\in X_{c(\chi)}\) such that
Furthermore, \(\chi^{}_1\) is primitive.
We say that \(\chi^{}_1\) induces \(\chi\).
Proof of Lemma 3.4 .
I. \(\,\)Let \(q=1\). Then \(c(\chi)=1\), \(\# X_1=1\), and \(\chi^{}_1 = \chi\), so the statement is obvious.
II. \(\,\)Let \(q \geq 2\) and \(\chi\) be a primitive character modulo \(q\). Then \(c(\chi)=q\) and we can take \(\chi^{}_1=\chi\). Let us prove the uniqueness. Suppose that there are two different characters \(\chi^{}_1, \chi^{}_2\in X_q\) satisfying (3.2). Then for any \(n\) such that \((n,q)>1\) we have \(\chi^{}_1(n)=0=\chi^{}_2(n)\). For any \(n\) such that \((n,q)=1\), we have \(\chi^{}_1(n)=\chi(n)=\chi^{}_2(n)\). Therefore, \(\chi^{}_1(n)=\chi^{}_2(n)\) for any integer \(n\); i.e., \(\chi^{}_1=\chi^{}_2\), a contradiction.
III. \(\,\)Let \(q\geq 2\) and \(\chi\) be an imprimitive character modulo \(q\). Then \(1\leq c(\chi) < q\) and by Lemma 3.3 we have \(c(\chi)\mid q\). We define \(\chi^{}_1\). Let \(n\in{\mathbb Z}\). Consider several cases.
If \((n,c(\chi))>1\), then we put \(\chi^{}_1(n)=0\).
If \((n,c(\chi))=1\), then by Lemma 3.1 there is an integer \(t\) such that \((n+tc(\chi),q)=1\). We put
The choice of \(t\) subject to the indicated condition is immaterial, since \(\chi\) restricted by \((n,q)=1\) has period \(c(\chi)\). Thus, \(\chi^{}_1(n)\) is defined for any integer \(n\). We claim that \(\chi^{}_1\) is a character modulo \(c(\chi)\). By construction,
By Lemma 3.1, there is an integer \(t\) such that \((1+tc(\chi),q)=1\). Since the choice of such a \(t\) is immaterial, we take \(t=0\). We have \(\chi^{}_1(1)=\chi(1)=1\). Now we prove that
If \((n,c(\chi))>1\), then we have \((n+c(\chi),c(\chi))>1\). Hence,
Let \((n,c(\chi))=1\). Then we have \((n+c(\chi), c(\chi))=1\). By Lemma 3.1, there are integers \(t_1\) and \(t_2\) such that \((n+t_1c(\chi),q)=1\) and \((n+c(\chi)+t_2c(\chi),q)=1\). By construction, we have
Since \(\chi\) restricted by \((n,q)=1\) has period \(c(\chi)\), we have \(\chi(n+t_1 c(\chi))= \chi(n+c(\chi)+t_2 c(\chi))\). Hence, \(\chi^{}_1(n) = \chi^{}_1(n+c(\chi))\) and (3.3) is proved. Now we prove that
If \((m,c(\chi))>1\), then we have \((mn, c(\chi))>1\). Hence, \(\chi^{}_1(mn)=0\) and \(\chi^{}_1(m)=0\). Therefore, relation (3.4) holds. Similarly, (3.4) holds if \((n,c(\chi))>1\). Let \((m,c(\chi))=1\) and \((n,c(\chi))=1\). Then \((mn,c(\chi))= 1\). By Lemma 3.1, there are integers \(t_1\), \(t_2\), and \(t_3\) such that \((m+t_1c(\chi),q)=1\), \((n+t_2c(\chi),q)=1\), and \((mn+t_3c(\chi),q)=1\). We put \(m_1 = m+t_1c(\chi)\), \(n_1 = n+t_2c(\chi)\), and \(u = mn+t_3c(\chi)\). By construction,
Since \(\chi\) is a totally multiplicative function, it follows that
Since \((m_1,q)=1\) and \((n_1,q)=1\), we have \((m_1 n_1,q)=1\). It is clear that \(m_1 n_1 \equiv u\pmod{c(\chi)}\). Since \(\chi\) restricted by \((n,q)=1\) has period \(c(\chi)\), we find that \(\chi(u)=\chi(m_1 n_1)\). Therefore, \(\chi^{}_1(mn)= \chi^{}_1(m)\chi^{}_1(n)\) and (3.4) is proved. Thus, we have proved that \(\chi^{}_1\) is a character modulo \(c(\chi)\), i.e., \(\chi^{}_1 \in X_{c(\chi)}\).
Now we prove that \(\chi^{}_1\) satisfies (3.2). It suffices to show that
Since \((n,q)=1\), we have \((n,c(\chi))=1\) (see Lemma 3.3). By Lemma 3.1, there is an integer \(t\) such that \((n+t c(\chi),q)=1\). By construction \(\chi^{}_1(n) = \chi(n+t c(\chi))\). Since \((n+t c(\chi),q)=1\), \((n,q)=1\), and \(n+tc(\chi) \equiv n\pmod{c(\chi)}\), we have \(\chi(n+t c(\chi))=\chi(n)\). Hence, \(\chi^{}_1(n)=\chi(n)\) and (3.5) is proved.
Now we prove that \(\chi^{}_1\) is a primitive character. Suppose that there is a positive integer \(q_2\) such that \(\chi^{}_1\) restricted by \((n,c(\chi))=1\) has period \(q_2\). Let \(m\) and \(n\) be integers such that \((m,q)=1\), \((n,q)=1\), and \(m\equiv n\pmod{q_2}\). By Lemma 3.3, we have \((m,c(\chi))=1\) and \((n,c(\chi))=1\). Then (see (3.5))
Hence, \(\chi\) restricted by \((n,q)=1\) has period \(q_2\). From the definition of a conductor it follows that \(q_2 \geq c(\chi)\). Hence, \(\chi^{}_1\) is a primitive character.
Now we prove the uniqueness. Suppose that there are two different characters \(\chi^{}_1, \chi^{}_2 \in X_{c(\chi)}\) satisfying (3.2). If \((n,c(\chi))>1\), then \(\chi^{}_1(n)=0=\chi^{}_2(n)\). Let \((n,c(\chi))=1\). By Lemma 3.1, there is an integer \(t\) such that \((n+t c(\chi),q)=1\). Since \(\chi^{}_1\) and \(\chi^{}_2\) are periodic functions with period \(c(\chi)\), we have
Thus, \(\chi^{}_1(n)=\chi^{}_2(n)\) for any \(n\in{\mathbb Z}\), and so \(\chi^{}_1=\chi^{}_2\). We obtain a contradiction. The uniqueness is proved. \(\quad\Box\)
Lemma 3.5.
Let \(q>1\) be an integer expressed in the standard form as \(q=q_1^{\alpha_1}\dots q_r^{\alpha_r},\) where \(q_1<\dots<q_r\) are primes and \(\alpha_1,\dots,\alpha_r\) are positive integers. Let \(\chi\) be a Dirichlet character modulo \(q\). Then there exist unique characters \(\chi^{}_i\) modulo \(q_i^{\alpha_i},\) \(i=1,\dots,r,\) such that
Furthermore, if the character \(\chi\) is real, then all characters \(\chi^{}_i,\) \(i=1,\dots,r,\) are real. If the character \(\chi\) is primitive, then all characters \(\chi^{}_i,\) \(i=1,\dots,r,\) are primitive.
Proof.
For any \(1\leq i\leq r\) we take \(A_i\) such that
Since the moduli of these congruences are coprime, the system has a solution (see, for example, [7, Ch. 4]). Thus, integers \(A_1,\dots,A_r\) are defined.
Let \(1\leq i \leq r\) and \(n\in{\mathbb Z}\). We put
It is easy to show that \(\chi^{}_i\) is a Dirichlet character modulo \(q_i^{\alpha_i}\).
Now we prove that (3.6) holds. Let \(n\in{\mathbb Z}\). Setting
we have
From (3.7) we obtain
Hence, \(n_1\dots n_r - n\) is divisible by \(q\), i.e.,
Hence, \(\chi(n_1\dots n_r)=\chi(n)\) and (3.6) is proved.
Now we prove the uniqueness of the representation of \(\chi\) in the form (3.6). Suppose that
where \( \widetilde{\chi}{} _i\) is a Dirichlet character modulo \(q_i^{\alpha_i}\), \(i= 1,\dots,r\). Let \(1 \leq i \leq r\) and \(n\in{\mathbb Z}\). We have (see (3.7))
and
Hence,
and
From (3.9) we obtain
Therefore (see (3.8)), \( \widetilde{\chi}{} _i(n)= \chi^{}_i(n)\). Since this equation holds for any \(n\in{\mathbb Z}\), we have \( \widetilde{\chi}{} _i = \chi^{}_i\), \(i=1,\dots,r\). Thus, the uniqueness of the representation of \(\chi\) in the form (3.6) is proved.
We see from (3.8) that if the character \(\chi\) is real, then all characters \(\chi^{}_i\), \(i=1,\dots,r\), are real. We claim that if the character \(\chi\) is primitive, then all characters \(\chi^{}_i\), \(i=1,\dots,r\), are primitive. Assume the contrary: there is an \(i\), \(1\leq i \leq r\), such that the character \(\chi^{}_i\) is imprimitive. Then \(c(\chi^{}_i)<q_i^{\alpha_i}\). Since \(c(\chi^{}_i)\mid q_i^{\alpha_i}\) (see Lemma 3.3), we have
We put
Let us show that the character \(\chi\) restricted by \((n,q)=1\) has period \( \widetilde{q}{} \). Take integers \(m\) and \(n\) such that \((m,q)=(n,q)=1\) and \(m\equiv n\pmod{ \widetilde{q}{} \,}\). Let \(1\leq j \leq r\), \(j\neq i\). Since
we have \(\chi^{}_j(m)=\chi^{}_j(n)\). Since \((m,q_i^{\alpha_i})=(n,q_i^{\alpha_i})=1\),
and \(\chi^{}_i\) restricted by \((n,q_i^{\alpha_i})=1\) has period \(q_i^{\beta}\), we have \(\chi^{}_i(m)=\chi^{}_i(n)\). This implies
We have proved that \(\chi\) restricted by \((n,q)=1\) has period \( \widetilde{q}{} \). But then \(c(\chi)\leq \widetilde{q}{} <q\). This contradicts the fact that the character \(\chi\) is primitive. Hence, all characters \(\chi^{}_i\), \(i=1,\dots,r\), are primitive. Lemma 3.5 is proved. \(\quad\Box\)
Lemma 3.6.
Let \(q\) be a positive integer such that there exists a real primitive character \(\chi\) modulo \(q\). Then the number \(q\) is of the form \(2^\alpha k,\) where \(\alpha\in\{0,\dots,3\}\) and \(k\geq 1\) is an odd square-free integer.
Proof.
Modulo \(q=1\) there exists a real primitive character; namely, \(\chi(n)=1\) for all \(n\in{\mathbb Z}\). The number \(1\) is of the form \(2^\alpha k\); namely, \(\alpha = 0\) and \(k=1\).
Let \(q>1\) be an integer such that there exists a real primitive character \(\chi\) modulo \(q\). Suppose that \(q=p^rs\), where \(p\geq 3\) is a prime number, \((p,s)=1\), and \(r\geq 2\). Let \( \widetilde{q}{} =p^{r-1}s\). We claim that the character \(\chi\) restricted by \((n,q)=1\) has period \( \widetilde{q}{} \). Let \(m\) and \(n\) be integers such that \((m,q)=(n,q)=1\) and \(m\equiv n\pmod{ \widetilde{q}{} \,}\). We have \(m=n+ \widetilde{q}{} \kern1pt t\), \(t\in{\mathbb Z}\), and
where
Let \(2\leq i \leq p^{r-1}\). Then
It is clear that \(i-1\geq 1\). We claim that
or, which is equivalent, \(i(r-1)\geq r\). Indeed, since \(i\geq 2\) and \(r\geq 2\), we have
Hence, \(A_i=p^rsN\), where \(N\in{\mathbb N}\). Thus, for any \(2\leq i \leq p^{r-1}\),
We have \(A_1=p^{r-1}(p^{r-1}s)=p^rs p^{r-2}\). Since \(r\geq 2\), we obtain
Hence (see (3.10)),
Using the properties of a character, we obtain
Since \((m,q)= (n,q)=1\) and the character \(\chi\) is real, we have \(\chi(m),\chi(n)\in\{-1,1\}\). Since \(p\geq3\) is a prime number and \(r\geq 2\) is an integer, it follows that \(p^{r-1}\) is an odd positive integer. Therefore, if \(\chi(m)=1\), then \(\chi(n)=1\), while if \(\chi(m)=-1\), then \(\chi(n)=-1\) as well. Thus, \(\chi(m)=\chi(n)\). We have proved that the character \(\chi\) restricted by \((n,q)=1\) has period \( \widetilde{q}{} \). Consequently,
This contradicts the fact that \(\chi\) is a primitive character. Hence, the number \(q\) is of the form \(2^\alpha k\), where \(\alpha \geq 0\) is an integer and \(k\geq 1\) is an odd square-free integer.
We claim that \(\alpha\leq 3\). Assume the contrary: \(\alpha \geq 4\). Let \(k=q_1\dots q_r\), where \(q_1<\dots<q_r\) are odd primes. By Lemma 3.5, we have
where \(\chi^{}_1\) is a real primitive character modulo \(2^\alpha\) and \(\chi^{}_i\) is a real primitive character modulo \(q_{i-1}\), \(i=2,\dots,r+1\) (if \(k=1\), then \(\chi^{}_2,\dots,\chi^{}_{r+1}\) are omitted in (3.12)). It is well known (see, for example, [7, Ch. 6]) that if numbers \(\nu\) and \(\gamma\) run independently through the sets \(\{0,1\}\) and \(\{0,\dots,2^{\alpha-2}-1\}\) respectively, then \((-1)^{\nu}\cdot5^{\gamma}\) runs (without repetitions) through a reduced residue system modulo \(2^\alpha\). Hence, for any \(n\) with \((n,2) = 1\) there are unique numbers \(\nu(n) \in \{0,1\}\) and \(\gamma(n) \in \{0,\dots,2^{\alpha-2}- 1\}\) such that
Since \((-1)^2=1\), we have \((\chi^{}_1(-1))^2=1\). Thus,
It is well known (see, for example, [7, Ch. 6]) that the number \(5\) belongs to \(2^{\alpha-2}\pmod{2^\alpha}\); in particular, \(5^{2^{\alpha-2}}\equiv 1\pmod{2^\alpha}\). Hence,
We obtain
We see from (3.13) that if \(n\) is such that \((n,2)=1\), then
We claim that \((b,2)=1\). Indeed, assume the contrary: \((b,2)>1\). We show that then \(\chi^{}_1\) restricted by \((n,2^\alpha)=1\) has period \(2^{\alpha-1}\). Let \(m\) and \(n\) be integers such that \((m,2^\alpha)=(n,2^\alpha)=1\) and \(m\equiv n\pmod{2^{\alpha-1}}\). We have
Since these congruences also hold modulo \(2^{\alpha-1}\), we have
Since \(\alpha\geq 4\), we obtain
It is clear that
Hence,
If \(\nu(m)=0\), then \(\nu(n)=0\); if \(\nu(m)=1\), then \(\nu(n)=1\). Thus,
Therefore (see (3.15)),
Suppose, for definiteness, that \(\gamma(m) \geq \gamma(n)\). We have
Since \((5^{\gamma(n)}, 2^{\alpha-1})=1\), we obtain
Hence,
Since \(5\) belongs to \(2^{\alpha-3}\pmod{2^{\alpha-1}}\), we have (see [7, Ch. 6])
Therefore,
where \(t\geq 0\) is an integer. Since \((b,2)>1\), we have
where \( \widetilde{b}{} \geq 0\) is an integer. We obtain (see (3.14) and (3.16)–(3.18))
Thus, we have proved that \(\chi^{}_1\) restricted by \((n,2^\alpha)=1\) has period \(2^{\alpha-1}\). Hence,
This contradicts the fact that \(\chi^{}_1\) is a primitive character. Hence, \((b,2)=1\).
For \(n=5\) we have \(\nu(5)=0\) and \(\gamma(5)=1\). Therefore (see (3.14)),
Since \(\alpha\geq 4\) and \((b,2)=1\), we have \(\operatorname{Im}(\chi^{}_1(5))\neq 0\). This contradicts the fact that \(\chi^{}_1\) is a real character. Hence, \(0 \leq \alpha\leq 3\). Lemma 3.6 is proved. \(\quad\Box\)
Lemma 3.7.
Let \(q_1\) and \(q_2\) be positive integers with \(q_1 \neq q_2,\) \(\chi^{}_1\) be a primitive character modulo \(q_1,\) and \(\chi^{}_2\) be a primitive character modulo \(q_2\). Then \(\chi^{}_1 \neq \chi^{}_2\).
Proof.
Assume the contrary: \(\chi^{}_1=\chi^{}_2\). Let \(m\) and \(n\) be integers such that \((m,q_1)=(n,q_1)=1\) and \(m\equiv n\pmod{q_2}\). Then
Hence, \(\chi^{}_1\) restricted by \((n,q_1)=1\) has period \(q_2\). Hence, \(c(\chi^{}_1) \leq q_2\). Since \(\chi^{}_1\) is a primitive character modulo \(q_1\), we have \(c(\chi^{}_1)=q_1\). Thus, \(q_1\leq q_2\). Similarly, it can be proved that \(q_2 \leq q_1\). Hence, \(q_1=q_2\). We have arrived at a contradiction, which means that \(\chi^{}_1 \neq \chi^{}_2\). \(\quad\Box\)
4. Lemmas on \(\psi(x,\chi)\)
In this section we present some lemmas on \(\psi(x,\chi)\). Most of these lemmas are well known. The proof of Lemma 4.6 is based on Maynard’s ideas (see the proof of Theorem 3.2 in [5]). The proof of Lemma 4.9 follows a standard proof of the Bombieri–Vinogradov theorem (see, for example, [1, Ch. 28]).
Lemma 4.1.
Let \(u\geq 2\) be a real number, and let \(Q\geq 2\) and \(W\) be integers with \((W,Q)=1\). Then
(the overbar denotes complex conjugation).
Proof.
We define
Since (see, for example, [1, Ch. 4])
we have
Let \(\chi^{}_0\) be the principal character modulo \(Q\). Since \((W,Q)=1\), it follows that \(\chi^{}_0(W)=1\). We have
Hence,
Lemma 4.1 is proved. \(\quad\Box\)
Lemma 4.2 (see, for example, [1, Ch. 14]).
There is a positive absolute constant \(a>0\) such that if \(\chi\) is a complex character modulo \(q,\) then \(L(s,\chi)\) has no zeros in the region
(here \(s=\sigma + i t,\) \(\sigma = \operatorname{Re} s,\) and \(t=\operatorname{Im} s\)). If \(\chi\) is a real nonprincipal character modulo \(q,\) the only possible zero of \(L(s,\chi)\) in this region is a single (simple) real zero. Furthermore, \(L(s,\chi)\) can have a zero in the region \(\Omega\) for at most one of the real nonprincipal characters \(\chi\pmod{q}\).
Remark.
It is easy to see that the constant \(a\) can be replaced by any constant \(a^*\) such that \(0<a^*<a\).
Lemma 4.3 (see [1, Ch. 20]).
Let \(\chi\) be a nonprincipal character modulo \(q\) and \(2\leq T\leq u\). Then
where
Here \(C>0\) is an absolute constant and \(a>0\) is the absolute constant in Lemma 4.2. The term \(-u^{\beta_1}/\beta_1\) should be omitted unless \(\chi\) is a real character for which \(L(s,\chi)\) has a zero \(\beta_1\) (which is necessarily unique, real, and simple) satisfying
Lemma 4.4 (Page’s theorem; see, for example, [1, Ch. 14]).
There are absolute constants \(a_1>0\) and \(a'_1>0\) such that the following holds. Let \(z\geq 3\) be a real number. Then there is at most one real primitive character \(\chi\) to a modulus \(q_0,\) \(3\leq q_0 \leq z,\) for which \(L(s,\chi)\) has a real zero \(\beta\) satisfying
If such a character \(\chi\) exists, then
Such a modulus \(q_0\) is said to be an exceptional modulus in the interval \([3,z]\).
Lemma 4.5.
Let \(z\geq 3\) be a real number. If an exceptional modulus \(q_0\) in the interval \([3,z]\) exists, then the number \(q_0\) is of the form \(2^\alpha k,\) where \(\alpha\in\{0,\dots,3\}\) and \(k\geq 1\) is an odd square-free integer.
Proof.
Suppose an exceptional modulus \(q_0\) in the interval \([3,z]\) exists. In particular, this means that there exists a real primitive character \(\chi\) modulo \(q_0\). By Lemma 3.6, the number \(q_0\) is of the form \(2^\alpha k\) with \(\alpha\in\{0,\dots,3\}\) and an odd square-free integer \(k\geq 1\). \(\quad\Box\)
Lemma 4.6.
There are positive absolute constants \(c_0,\) \(c_1,\) \(\gamma_0,\) and \(C\) such that the following holds. Let \(x\geq c_0\) be a real number, \(q_0\) be an exceptional modulus in the interval \([3,\exp(2c_1\sqrt{\ln x})],\) \(Q\) be an integer such that \(3\leq Q\leq \exp(2 c_1 \sqrt{\ln x})\) and \(Q\neq q_0\) (the last inequality should be interpreted as follows: if \(q_0\) exists, then \(Q\neq q_0;\) if \(q_0\) does not exist, then \(Q\) is any integer in the indicated interval), and \(\chi\) be a primitive character modulo \(Q\). Then
Proof.
We will choose \(c_1\) and \(\gamma_0\) later. The number \(c_0\) depends on \(c_1\) and \(\gamma_0\) and is large enough, and \(x\geq c_0(c_1,\gamma_0)\). We put
We have \(z\geq 3\) if the number \(c_0(c_1,\gamma_0)\) is chosen large enough. By Lemma 4.4, there is at most one real primitive \(\chi\) to a modulus \(q_0\), \(3\leq q_0 \leq z\), for which \(L(s,\chi)\) has a real zero \(\beta\) satisfying
If such a character \(\chi\) exists, then
provided that \(c_0(c_1,\gamma_0)\) is chosen large enough. Let \(Q\) be an integer such that \(3 \leq Q\leq \exp(2 c_1 \sqrt{\ln x})\) and \(Q\neq q_0\), and let \(\chi\) be a primitive character modulo \(Q\). Since \(Q>1\), we see that \(\chi\) is a nonprincipal character. By Lemma 4.3, if \(2 \leq T \leq u\), then
where
The term \(-u^{\beta_1}/\beta_1\) is to be omitted unless \(\chi\) is a real character modulo \(Q\) for which \(L(s,\chi)\) has a zero \(\beta_1\) (which is necessarily unique, real, and simple) satisfying
Let
Let \(u \geq c_2(c_1)\), where \(c_2(c_1)>0\) is a number depending only on \(c_1\). We choose
Then \(2 \leq T \leq u\) if \(c_2(c_1)\) is chosen large enough.
I. \(\,\)Now we estimate the quantity
If \(c_0(c_1,\gamma_0)\) is chosen large enough, then
Hence,
If \(c_0(c_1,\gamma_0)\) is chosen large enough, then
Therefore,
We have
Consider two cases.
(1) Let \(x^{1/4} \leq u \leq x^{1+ \gamma_0/\sqrt{\ln x}}\). Then
Let
Hence,
and
provided that \(c_0(c_1,\gamma_0)\) is chosen large enough. If \(0<\gamma_0 \leq c_1/2\), then
(2) Let \(c_2(c_1) \leq u < x^{1/4}\) (we may assume that \(c_0(c_1,\gamma_0)>(c_2(c_1))^4\) and \(c_2(c_1)\geq 10\)). We have
provided that \(c_0(c_1,\gamma_0)\) is chosen large enough.
Thus, if \(0<c_1<\sqrt{a/160}\), \(0<\gamma_0 \leq c_1/2\), \(x \geq c_0(c_1,\gamma_0)\), and \(c_2(c_1) \leq u \leq x^{1+\gamma_0/\sqrt{\ln x}}\), then
II. \(\,\)Now we estimate the quantity
From (4.5) and (4.8) we obtain
Consider two cases.
(1) Let \(x^{9/10} \leq u \leq x^{1+ \gamma_0/\sqrt{\ln x}}\). Then
Since \(0<\gamma_0 \leq c_1/2\), we have
provided that \(c_0(c_1,\gamma_0)\) is chosen large enough.
(2) Let \(c_2(c_1) \leq u < x^{9/10}\). Then
provided that \(c_0(c_1,\gamma_0)\) is chosen large enough.
Thus, if \(0<c_1<\sqrt{a/160}\), \(0<\gamma_0 \leq c_1/2\), \(x \geq c_0(c_1,\gamma_0)\), and \(c_2(c_1) \leq u \leq x^{1+\gamma_0/\sqrt{\ln x}}\), then
III. \(\,\)Now we estimate the quantity
we have
provided that \(c_0(c_1,\gamma_0)\) is chosen large enough.
Finally, we obtain the following (see (4.4)): if \(0<c_1<\sqrt{a/160}\), \(0<\gamma_0 \leq c_1/2\), \(x \geq c_0(c_1,\gamma_0)\), and \(c_2(c_1) \leq u \leq x^{1+\gamma_0/\sqrt{\ln x}}\), then
where \(C>0\) is an absolute constant.
IV. \(\,\)Now we estimate the quantity (see (4.3))
If \(\chi\) is not a real character modulo \(Q\) for which \(L(s,\chi)\) has a zero \(\beta_1\) (which is necessarily unique, real, and simple) satisfying
then the term \(-u^{\beta_1}/\beta_1\) in (4.3) is to be omitted, and there is nothing to estimate. Let \(\chi\) be such a character. Then \(\chi\) is a real primitive character modulo \(Q\). Since \(Q\neq q_0\), we have (see Lemma 3.7 and (4.1))
Hence,
By the remark made after Lemma 4.2, we may assume that \(0<a < 1/2\). Since \(Q\geq 3\), we have
Hence, \(0<1/{\beta_1}\leq 2\). Thus,
Consider two cases.
(1) Let \(x^{1/2} \leq u \leq x^{1+\gamma_0/\sqrt{\ln x}}\). We have (see (4.6))
We take
Then,
Since \(0<\gamma_0 \leq c_1/2\), we obtain (see (4.10))
(2) Let \(c_2(c_1) \leq u < x^{1/2}\). Then (see (4.10))
provided that \(c_0(c_1,\gamma_0)\) is chosen large enough. Combining the estimates found at steps I–IV together, we obtain the following (see (4.3) and (4.9)): if \(0<c_1<\sqrt{\min\{a,a_1\}/160}\), \(0<\gamma_0 \leq c_1/2\), \(x \geq c_0(c_1,\gamma_0)\), and \(c_2(c_1) \leq u \leq x^{1+\gamma_0/\sqrt{\ln x}}\), then
where \(C>0\) is an absolute constant.
There is a number \(d(c_1)>0\), depending only on \(c_1\), such that
We may assume that \(c_0(c_1,\gamma_0) > d(c_1)\). Hence, if \(2 \leq u < c_2(c_1)\), then (see (2.1))
Thus, if \(0<c_1<\sqrt{\min\{a,a_1\}/160}\), \(0<\gamma_0 \leq c_1/2\), and \(x \geq c_0(c_1,\gamma_0)\), then
where \(C>0\) is an absolute constant. We take
Since \(a>0\) and \(a_1>0\) are absolute constants, we see that \(c_1\), \(\gamma_0\), \(c_0(c_1,\gamma_0)\) and \(c_2(c_1)\) are positive absolute constants. Lemma 4.6 is proved. \(\quad\Box\)
Lemma 4.7 (see [1, Ch. 19]).
Let \(u \geq 2\) be a real number, \(Q\geq 2\) be an integer, \(\chi \in X_Q,\) and \(\chi^{}_1\) be a primitive character modulo \(q_1\) inducing \(\chi\). Then
Lemma 4.8 (see [1, Ch. 28]).
Let \(Q_1,\) \(Q_2,\) and \(t\) be real numbers such that \(1\leq Q_1 < Q_2\) and \(t\geq 2\). Then
where \(C>0\) is an absolute constant.
Lemma 4.9.
Let \(\varepsilon\) and \(\delta\) be real numbers such that \(0<\varepsilon<1\) and \(0<\delta<1/2\). Then there exists a number \(c(\varepsilon,\delta)>0,\) depending only on \(\varepsilon\) and \(\delta,\) such that if \(x\in{\mathbb R}\) and \(q\in{\mathbb Z}\) satisfy the conditions \(x\geq c(\varepsilon,\delta)\) and \(1 \leq q \leq (\ln x)^{1-\varepsilon},\) then there is a positive integer \(B\) for which the following relations hold:
and
Here \(c_1,\) \(\gamma,\) \(c_2,\) and \(c_3\) are positive absolute constants.
Proof.
Let \(c_0\), \(c_1\), \(\gamma_0\) and \(C\) be the positive absolute constants in Lemma 4.6. We will choose \(\gamma\) and \(c(\varepsilon,\delta)=c(\varepsilon,\delta,\gamma)\) later; they are assumed to be small and large enough, respectively; for now, let \(0<\gamma \leq \gamma_0\), \(c(\varepsilon,\delta,\gamma) \geq c_0\), and \(x \geq c(\varepsilon,\delta,\gamma)\). Let \(q_0\) be the exceptional modulus in the interval \([3,\exp(2 c_1 \sqrt{\ln x})]\). If \(q_0\) does not exist, then we take \(B=1\). If \(q_0\) exists, then (see (4.2))
where \(c_4>0\) is an absolute constant. We have \(q_0\geq 24\) if \(c(\varepsilon,\delta,\gamma)\) is chosen large enough. By Lemma 4.5, the number \(q_0\) is of the form \(2^\alpha k\), where \(\alpha\in\{0,\dots,3\}\) and \(k\geq 3\) is an odd square-free integer. We put
Let \(\tau = (M_1,q)\) and \(M_2=M_1/\tau\). Then \((M_2, q)=1\). Since \(\tau \leq q \leq (\ln x)^{1-\varepsilon}\), we have
If \(c(\varepsilon,\delta,\gamma)\) is chosen large enough, then \(M_2 \geq 3\). Hence, \(M_2\geq 3\) is an odd square-free integer. Furthermore, we have \((M_2, q)=1\) and \(M_2\) divides \(q_0\). Let \(B\) be the largest prime divisor of \(M_2\). Hence, \(B\geq 3\) is a prime number and \(B\) divides \(q_0\). We have (see Lemma 2.4)
Thus, \(1 \leq B \leq \exp(2c_1 \sqrt{\ln x})\) is an integer, \((B,q)=1\), \(1\leq{B}/{ \varphi (B)} \leq 2\), and \(B\geq 3\) is a prime divisor of \(q_0\) if \(q_0\) exists.
Let \(u\) be a real number such that \(2 \leq u \leq x^{1+\gamma/ \sqrt{\ln x}}\), and let \(Q\) and \(W\) be integers such that \(2\leq Q \leq x^{1/2 - \delta}\), \((Q, B)=1\), and \((W,Q)=1\). By Lemma 4.1, we have
Therefore,
Since the right-hand side of this inequality does not depend on \(W\), we have
Let \(\chi\in X_Q\), and let \(\chi^{}_1\) be a primitive character modulo \(q_1\) inducing \(\chi\). From Lemma 3.4 and the definition of the inducing character (which is given below Lemma 3.4), we have \(q_1 = c(\chi)\), and hence \(q_1\mid Q\) (see Lemma 3.3). Applying Lemma 4.7, we find
Since \(\# X_Q = \varphi (Q)\), we obtain
We can assume that
provided that \(c(\varepsilon,\delta,\gamma)\) is chosen large enough. Hence,
We obtain
Therefore,
where
Let us estimate the sum \(S'\). Let \(Q\) be an integer with \(2 \leq Q \leq x^{1/2 - \delta}\) and \((Q,B)=1\), let \(\chi\in X_Q\), and let \(\chi^{}_1\) be the primitive character modulo \(q_1\) inducing \(\chi\). Since \(q_1\mid Q\), we have \(1\leq q_1 \leq x^{1/2 - \delta}\) and \((q_1,B)=1\). Hence,
Applying Lemmas 2.5 and 2.9, we obtain
where \(C>0\) is an absolute constant. We have
Redenoting \(q_1\) by \(Q\) and \(\chi^{}_1\) by \(\chi\), we find
where
and \(c_1>0\) is the absolute constant in Lemma 4.6.
I. \(\,\)Now we estimate \(S'_1\). We have
(1) Let us estimate \(R_1\). Since \(\# X^*_1=1\), we have
where \(\chi\in X^*_1\), i.e., \(\chi(n)=1\) for any \(n\in{\mathbb Z}\). Since \(\chi\) is the principal character modulo \(1\), it follows that
We have
It is well known (see, for example, [1, Ch. 18]) that
where \(C>0\) and \(c>0\) are absolute constants. Consider two cases.
(i) Let \(x^{1/4} \leq u \leq x^{1+ \gamma/\sqrt{\ln x}}\) (we may assume that \(c(\varepsilon,\delta,\gamma)>16\)). Then (see (4.11))
Hence,
provided that \(0<\gamma \leq c/4\).
(ii) Let \(2 \leq u < x^{1/4}\). Then
provided that \(c(\varepsilon,\delta,\gamma)\) is chosen large enough.
We obtain
(2) Now we estimate
Let \(Q\) be an integer such that \(2 \leq Q \leq \ln x\), and let \(\chi \in X^*_Q\). Then \(\chi\) is a nonprincipal character modulo \(Q\), and hence \(\psi'(u,\chi) = \psi(u,\chi)\). Consider two cases.
(i) Let \(x^{1/4} \leq u \leq x^{1+ \gamma/\sqrt{\ln x}}\). Then (see (4.11))
We may assume that \(c(\varepsilon,\delta,\gamma) \geq e^{16}\). Hence, \(\ln u \geq (\ln x)/4\geq 4\). We have
Therefore (see, for example, [1, Ch. 22]),
where \(C>0\) and \(c(2)>0\) are absolute constants. We have
and
provided that \(0<\gamma \leq c(2)/4\).
(i) Let \(2 \leq u < x^{1/4}\). Then (see (2.1))
and
provided that \(c(\varepsilon,\delta,\gamma)\) is chosen large enough. Hence,
Substituting this estimate into (4.17) and using the fact that \(\#X^*_Q\leq \#X_Q= \varphi (Q)\), we obtain
provided that \(c(\varepsilon,\delta,\gamma)\) is chosen large enough.
Substituting (4.16) and (4.18) into (4.14), we find
where \(C>0\) and \(c>0\) are absolute constants.
II. \(\,\)Now we estimate the quantity
Let \(Q\) be an integer with \(\exp(c_1\sqrt{\ln x}) < Q \leq x^{1/2 - \delta}\) and \((Q,B)=1\), and let \(\chi\in X^*_Q\). Since \(Q>1\), we see that \(\chi\) is a nonprincipal character modulo \(Q\). Hence,
We have
Applying Lemma 4.8 with \(Q_1 = \exp(c_1\sqrt{\ln x})\), \(Q_2= x^{1/2 - \delta}\), and \(t=x^{1+\gamma/\sqrt{\ln x}}\), we obtain
We can assume that
if \(c(\varepsilon,\delta,\gamma)\) is chosen large enough. Increasing \(C\) if necessary, we have
Then
provided that \(0<\gamma\leq c_1/2\). We obtain
provided that \(c(\varepsilon,\delta,\gamma)\) is chosen large enough. Redenoting \(3C\) by \(C\) and \(c_1/4\) by \(c\), we arrive at
where \(C>0\) and \(c>0\) are absolute constants.
III. \(\,\)Now we estimate the quantity
Let \(Q\) be an integer with \(\ln x < Q \leq \exp(c_1\sqrt{\ln x})\) and \((Q,B)=1\), and let \(\chi\in X^*_Q\). Since \(Q>1\), we see that \(\chi\) is a nonprincipal character modulo \(Q\), and hence \(\psi'(u,\chi)= \psi(u,\chi)\). We recall that if an exceptional modulus \(q_0\) in the interval \([3,\exp(2c_1\sqrt{\ln x})]\) does not exist, then \(B=1\); if \(q_0\) exists, then \(B \geq3\) is a prime divisor of \(q_0\), and so \(Q\neq q_0\). Since \(0<\gamma \leq \gamma_0\) and \(c(\varepsilon,\delta,\gamma) \geq c_0\), we see from Lemma 4.6 that
Since \(\#X^*_Q\leq \#X_Q= \varphi (Q)\), we obtain
where \( \widetilde{C}{} >0\) and \( \widetilde{c}{} > 0\) are absolute constants. Substituting (4.22) into (4.13), we obtain
provided that \(c(\varepsilon,\delta,\gamma)\) is chosen large enough. Redenoting \(C'\) by \(C\) and \( \widetilde{c}{} /2\) by \(c\), we arrive at
where \(C>0\) and \(c>0\) are absolute constants.
IV. \(\,\)We have
provided that \(c(\varepsilon,\delta,\gamma)\) is chosen large enough (here \(c>0\) is the absolute constant in (4.23)).
V. \(\,\)Now we estimate the quantity
Let \(W\in{\mathbb Z}\). We have
Hence,
Using (4.15) and arguing as in cases I(1), (i) and I(1), (ii), we obtain
where \(C>0\) and \(c>0\) are absolute constants.
Substituting (4.23)–(4.25) into (4.12), we find
where \(C>0\) and \(c>0\) are absolute constants. Thus, if \(\gamma\) is a sufficiently small positive absolute constant, \(x \geq c(\varepsilon,\delta,\gamma)\) is a real number, and \(q\) is an integer such that \(1\leq q \leq (\ln x)^{1-\varepsilon}\), then there is an integer \(B\) such that
and
where \(c_1\), \(C\), and \(c\) are positive absolute constants. Let us redenote \(2c_1\) by \(c_1\), \(C\) by \(c_2\), and \(c\) by \(c_3\). Since \(\gamma\) is an absolute constant, we see that the positive number \(c(\varepsilon,\delta,\gamma)= c(\varepsilon,\delta)\) depends only on \(\varepsilon\) and \(\delta\). Lemma 4.9 is proved. \(\quad\Box\)
Lemma 4.10.
Let \(\varepsilon\) and \(\delta\) be real numbers such that \(0<\varepsilon<1\) and \(0<\delta<1/2\). Then there is a number \(c(\varepsilon,\delta)>0,\) depending only on \(\varepsilon\) and \(\delta,\) such that if \(x\in{\mathbb R}\) and \(q\in{\mathbb Z}\) satisfy the conditions \(x\geq c(\varepsilon,\delta)\) and \(1 \leq q \leq (\ln x)^{1-\varepsilon},\) then there is a positive integer \(B\) for which the following relations hold:
and
Here \(c_1,\) \(\gamma,\) \(c_2,\) and \(c_3\) are positive absolute constants.
Proof.
We will choose the number \( \widetilde{c}{} (\varepsilon,\delta)\) later; it is assumed to be large enough. Let \( \widetilde{c}{} (\varepsilon,\delta) \geq c(\varepsilon,\delta)\), where \(c(\varepsilon,\delta)\) is the number in Lemma 4.9. Let \(x\in{\mathbb R}\) and \(q\in{\mathbb Z}\) be such that \(x \geq \widetilde{c}{} (\varepsilon,\delta)\) and \(1 \leq q \leq (\ln x)^{1-\varepsilon}\). Then, by Lemma 4.9, there is a positive integer \(B\) such that
and
where
and \(c_1\), \(\gamma\), \(c_2\), and \(c_3\) are positive absolute constants.
We put
Let \(Q\in{\mathbb Z}\), \(W\in{\mathbb Z}\), and \(u\in{\mathbb Z}\) be such that \(1 \leq Q \leq x^{1/2 - \delta}\), \((Q,B)=1\), \((W,Q)=1\), and \(3\leq u \leq x^{1+\gamma/\sqrt{\ln x}}\). We claim that
where \(C_1>0\) is an absolute constant. We define
Let us show that
where \(C>0\) is an absolute constant. Let \(u \geq 8\). Then
We have
and
where \(C'>0\) is an absolute constant. If \(3 \leq u <8\), then
Thus, (4.30) is proved.
Since
we have
Further,
Since
we obtain
We have (see (4.30))
where
We can estimate this quantity as
Since \(u \geq 3\), we have
Since
it follows (see (4.30)) that
Let \(f(x)=-\ln^{-1}x\) and \(n \geq 2\) be an integer. By the mean value theorem, there is a \(\xi\in (n,n+1)\) such that
Substituting (4.32)–(4.34) into (4.31), we obtain (4.29). Hence,
I. \(\,\)Now we estimate \(S_1\). We have (see (4.27))
II. \(\,\)Let us estimate \(S_2\). We can assume that
provided that \( \widetilde{c}{} (\varepsilon,\delta)\) is chosen large enough. We have
provided that \( \widetilde{c}{} (\varepsilon,\delta)\) is chosen large enough.
III. \(\,\)Now we estimate \(S_3\). Let \(Q\), \(W\), \(u\), and \(n\) be integers such that \(1 \leq Q \leq x^{1/2 - \delta}\), \((Q,B)=1\), \((W,Q)=1\), \(3 \leq u \leq x^{1+ \gamma/\sqrt{\ln x}}\), and \(2\leq n \leq u-1\). Then
Hence,
We have
Therefore (see (4.27)),
Substituting (4.36)–(4.38) into (4.35), we obtain (see (4.28))
where \(c_4 = c_2+1+ c_0 c_2> 0\) is an absolute constant.
Let \(Q\) and \(W\) be integers such that \(1 \leq Q \leq x^{1/2 - \delta}\), \((Q,B)=1\), and \((W,Q)=1\), and let \(u\) be a real number with \(2 \leq u \leq x^{1+\gamma/\sqrt{\ln x}}\). Consider two cases.
(1) Let \(2 \leq u \leq 3\). Then
and so
(2) Let \(3< u \leq x^{1+\gamma/\sqrt{\ln x}}\). Then
Hence,
From (4.40) and (4.41) we obtain
We can assume that
provided that \( \widetilde{c}{} (\varepsilon,\delta)\) is chosen large enough. From (4.39), (4.42), and (4.43) we obtain
Thus, if \(x \geq \widetilde{c}{} (\varepsilon,\delta)\) is a real number and \(q\) is an integer such that \(1 \leq q \leq (\ln x)^{1-\varepsilon}\), then there is a positive integer \(B\) for which (4.26) and (4.44) hold. Let us redenote \( \widetilde{c}{} (\varepsilon,\delta)\) by \(c(\varepsilon,\delta)\) and \(c_4 + 2 \operatorname{li} (3)+2\) by \(c_2\). Lemma 4.10 is proved. \(\quad\Box\)
5. Proof of Theorem 1.1 and Corollary 1.1
Let us introduce some additional notation. Let \( {\mathcal A} \) be a set of integers, \( {\mathcal P} \) a set of primes, and \(L(n)=l_1 n + l_2\) a linear function with integer coefficients. We define
Let \( {\mathcal L} =\{L_1,\dots,L_k\}\) be a set of distinct linear functions \(L_i (n) = a_i n+b_i\), \(i=1,\dots,k\), with positive integer coefficients. We say such a set is admissible if for every prime \(p\) there is an integer \(n_p\) such that \(\bigl(\prod_{i=1}^kL_i(n_p),p\bigr)=1\).
We focus on sets satisfying the following hypothesis, which is given in terms of \(( {\mathcal A} , {\mathcal L} , {\mathcal P} ,B,x,\theta)\), where \( {\mathcal L} \) is an admissible set of linear functions, \(B\in{\mathbb N}\), \(x\) is a large real number, and \(0<\theta<1\).
Hypothesis 1.
For \(( {\mathcal A} , {\mathcal L} , {\mathcal P} ,B,x,\theta)\) and \(k=\# {\mathcal L} \), the following holds.
(1) \( {\mathcal A} \) is well distributed in arithmetic progressions:
(2) The primes in \(L( {\mathcal A} )\cap {\mathcal P} \) are well distributed in most arithmetic progressions: for any \(L\in {\mathcal L} \) we have
(3) \( {\mathcal A} \) is not too concentrated in any arithmetic progression: for any \(1\leq q< x^{\theta}\) we have
Maynard proved the following result (see [5, Proposition 6.1]).
Proposition 5.1.
Let \(\alpha\) and \(\theta\) be real numbers such that \(\alpha>0\) and \(0<\theta <1\). Let \( {\mathcal A} \) be a set of integers, \( {\mathcal P} \) a set of primes, and \( {\mathcal L} =\{L_1,\dots,L_k\}\) an admissible set of \(k\) linear functions, and let \(B\) and \(x\) be integers. Let the coefficients of \(L_i(n)= a_in+b_i\in {\mathcal L} \) satisfy \(1\leq a_i,b_i\leq x^\alpha\) for all \(1 \leq i \leq k,\) and let \(k \leq (\ln x)^{1/5}\) and \(1 \leq B \leq x^\alpha\). Let \(x^{\theta/10}\leq R \leq x^{\theta/3}\). Let \(\rho\) and \(\xi\) satisfy \(k(\ln \ln x)^2/ \ln x \leq \rho, \xi \leq \theta/10,\) and define
Then there is a number \(C>0\) depending only on \(\alpha\) and \(\theta\) such that the following holds. If \(k \geq C\) and \(( {\mathcal A} , {\mathcal L} , {\mathcal P} ,B,x,\theta)\) satisfy Hypothesis 1, then there exist nonnegative weights \(w_n=w_n( {\mathcal L} )\) satisfying
such that the following statements hold.
(1) We have
(2) For \(L(n)=a_Ln+b_L\in {\mathcal L} \) we have
(3) For \(L(n)=a_0n+b_0\notin {\mathcal L} \) and \(D \leq x^\alpha,\) if \(\Delta_L\neq 0,\) we have
where
(4) For \(L\in {\mathcal L} \) we have
Here \(I_k\) and \(J_k\) are quantities depending only on \(k,\) and \( {\mathfrak S} _B( {\mathcal L} )\) is a quantity depending only on \( {\mathcal L} ,\) and these satisfy
for a smooth function \(F=F_k \colon\, {\mathbb R}^k\to{\mathbb R}\) depending only on \(k\). The implied constants here depend only on \(\alpha,\) \(\theta,\) and the implied constants from Hypothesis 1. The constant \(c\) in inequality (5.6) is positive and absolute.
Proof of Theorem 1.1 .
First we prove the following
Lemma 5.1.
Let \(k\) be a positive integer. Let \(a,\) \(q,\) and \(b_1,\dots,b_k\) be positive integers such that \(b_1<\dots<b_k\) and \((a,q) =1\). Let \(L_i(n) = q n+a+ q b_i,\) \(i=1,\dots,k\). Then \( {\mathcal L} =\{L_1,\dots,L_k\}\) is an admissible set if and only if for any prime \(p\) such that \(p\nmid q\) there is an integer \(m_p\) with \(m_p\not\equiv b_i\pmod{p}\) for all \(1\leq i\leq k\).
Proof.
(1) Let \( {\mathcal L} =\{L_1,\dots,L_k\}\) be an admissible set. Let \(p\) be a prime such that \(p \nmid q\). Since \( {\mathcal L} \) is an admissible set, there is an integer \(n_p\) such that \(\bigl(\prod_{i=1}^kL_i(n_p),p\bigr)=1\). Since \((q,p)=1\), there is an integer \(q'\) such that \(q q' \equiv 1\pmod{p}\). We put \(m_p = -(n_p+q'a)\). Let \(i\) be an integer with \(1 \leq i \leq k\). Since \((q',p)=1\) and \((L_i(n_p),p)=1\), it follows that \((q'L_i(n_p), p)=1\). We have
Hence, \(m_p \not\equiv b_i\pmod{p}\).
(2) Suppose that for any prime \(p\) with \(p\nmid q\) there is an integer \(m_p\) such that \(m_p\not\equiv b_i\pmod{p}\) for all \(1 \leq i\leq k\). Let us show that then \( {\mathcal L} \) is an admissible set. First we observe that \( {\mathcal L} =\{L_1,\dots,L_k\}\) is a set of distinct linear functions \(L_i (n) = q n+l_i\), \(i=1,\dots,k\), with positive integer coefficients. Thus, we need to prove that for any prime \(p\) there is an integer \(n_p\) such that \(\bigl(\prod_{i=1}^k L_i(n_p),p\bigr)=1\). Let \(p\) be a prime number. Consider two cases.
(i) Let \(p\mid q\). Since \((a,q)=1\), we have \((a,p)=1\). Let \(i\) be an integer with \(1 \leq i \leq k\). For any integer \(n\) we have
and so \(L_i(n) \not\equiv 0\pmod{p}\). Hence, \(\bigl(\prod_{i=1}^k L_i(n), p\bigr)=1\). Therefore, in this case we may take any integer as \(n_p\).
(ii) Let \(p\nmid q\). Then \((q,p)=1\), and so there is an integer \(c\) such that
By assumption, there is an integer \(m_p\) such that \(m_p\not\equiv b_i\pmod{p}\) for all \(1 \leq i \leq k\). We put \(n_p=-m_p - c\). Let \(i\) be an integer with \(1 \leq i \leq k\). We have
Since \((q,p)=1\), we obtain
In view of (5.9) this yields \(L_i(n_p)\not\equiv 0\pmod{p}\). Hence, \((L_i(n_p), p) =1\). Since this holds for all \(1 \leq i \leq k\), we have \(\bigl(\prod_{i=1}^kL_i(n_p),p\bigr)=1\). Lemma 5.1 is proved. \(\quad\Box\)
The proof of the following lemma is based on Maynard’s ideas used in the proof of Lemma 8.1 in [5] (the notation \(L\in {\mathcal L} \) was explained in the Introduction).
Lemma 5.2.
There are positive absolute constants \(c\) and \(C\) such that the following holds. Let \(x\) and \(\eta\) be real numbers with \(x\geq c\) and \((\ln x)^{-9/10}\leq \eta \leq 1\). Let \(k\) and \(a\) be positive integers. Let \(b_1,\dots,b_k\) be integers with \(1\leq b_i \leq \ln x,\) \(i=1,\dots,k\). Let \( {\mathcal L} =\{L_1,\dots,L_k\}\) be a set of \(k\) linear functions \(L_i(n)=an+b_i,\) \(i=1,\dots,k\). For \(L(n)=an+b,\) \(b\in{\mathbb Z},\) we define
Then
Proof.
Consider two cases.
(1) Let \(k> \ln\ln x\). We can assume that \(\ln\ln x \geq 100\) provided that \(c\) is chosen large enough. Therefore, \(k\geq 100\). Let \(b\) be an integer such that \(1 \leq b \leq \eta\ln x\) and \(L=an+b\notin {\mathcal L} \). Then \(\Delta_L\in{\mathbb N}\). Applying Lemma 2.8, we see that
where \(c_0>0\) is an absolute constant. Further,
For any \(1 \leq i \leq k\) we have \(|b_i - b| \leq \ln x\). Hence,
Since
we have
We observe that if \(u\geq 2\) and \(v\geq 2\) are real numbers, then
Applying (5.11), we obtain
Applying (5.11) again, we have
Substituting this estimate into (5.10), we obtain
where \(c_1=100 c_0>0\) is an absolute constant. Thus,
(2) Let \(1 \leq k \leq \ln\ln x\). For an integer \(b\) we define
Let \(b\) be an integer such that \(1\leq b \leq \eta\ln x\) and \(L=an+b\notin {\mathcal L} \). Applying Lemmas 2.5 and 2.4, we obtain
Hence,
Applying Lemma 2.7, we have
First we estimate the sum \(S_2\). Let \(b\) and \(d\) be positive integers such that \(1\leq b \leq \eta\ln x\), \(L=an+b\notin {\mathcal L} \), \(d> \eta \ln x\), and \(d\mid\Delta(b)\). We claim that
We can assume that
provided that \(c\) is chosen large enough. If \(\mu^2(d)=0\), then inequality (5.15) holds. Let \(\mu^2(d)\neq 0\). Then \(d\) is square-free. Therefore, \(\sum_{p\mid d}\ln p = \ln d\). Inequality (5.15) is equivalent to the inequality
which obviously holds. Thus, (5.15) is proved. We have
Let \(b\in{\mathbb N}\), \(d\in{\mathbb N}\), and \(p\in\mathbb{P}\) be such that \(1 \leq b \leq \eta\ln x\), \(L=an+b\notin {\mathcal L} \), \(p\mid \Delta(b)\), \(d>\eta\ln x\), \(d\) is a multiple of \(p\), and \(d\mid\Delta (b)\). Then \(d=pt\), where \(t\in{\mathbb N}\), \(t>(\eta\ln x)/p\), and \(t\mid\Delta(b)\). We have (see Lemmas 2.5 and 2.4)
Hence,
We obtain (see Lemma 2.7)
Therefore,
Since \(\eta\geq (\ln x)^{-9/10}\), we have
Thus,
Let \(b\) be an integer such that \(1 \leq b \leq \eta\ln x\) and \(L=an+b\notin {\mathcal L} \). Applying Lemmas 2.8 and 2.10, we obtain
where \(c_2>0\) and \(c_3>0\) are absolute constants. We have
Hence,
provided that \(c\) is chosen large enough. It follows from (5.17) and (5.19) that
where \(c_4=9c_2c_3>0\) is an absolute constant. Substituting this estimate into (5.16), we obtain
We can assume that
provided that \(c\) is chosen large enough. Hence,
Now we estimate \(S_1\). We have
Let \(d\) be an integer such that \(1\leq d \leq \eta\ln x\) and \(d\in {\mathcal M} \). We claim that
If \(d=1\), then the inequality is obvious. Let \(d>1\). We define
Then \(\Delta(b)=|R(b)|\). We have
Let \(d\) be expressed in the standard form as \(d=q_1\dots q_r\), where \(q_1<\dots <q_r\) are prime numbers. It is well known (see, for example, [7, Ch. 4]) that the congruence \(R(b)\equiv 0\pmod{d}\) is equivalent to the system of congruences
Let \(1\leq j \leq r\). Let \(\Omega_j\) be the set of numbers of a complete system of residues modulo \(q_j\) satisfying the congruence \(R(b)\equiv 0\pmod{q_j}\). Since \(R(b_1)=0\), we see that \(\Omega_j\neq\varnothing\). Since the leading coefficient of the polynomial \(R(b)\) is \(1\) and the degree of the polynomial \(R(b)\) is \(k\), we have \(\#\Omega_j\leq k\) (see, for example, [7, Ch. 4]). It is also clear that \(\#\Omega_j \leq q_j\). Thus,
System (5.23) is equivalent to the union of \(T=\#\Omega_1\dots\#\Omega_r\) systems
where \(\tau_1\in\Omega_1,\dots,\tau_r\in\Omega_r\). It is well known (see, for example, [7, Ch. 4]) that the system of congruences (5.24) is equivalent to the congruence
where \(x_0=x_0(\tau_1,\dots,\tau_r)\). It is also known that the numbers \(x_0(\tau_1,\dots,\tau_r)\), \(\tau_1\in\Omega_1,\dots,\tau_r\in \Omega_r\), are incongruent modulo \(d\). Thus,
Let \(\tau_1\in\Omega_1,\dots,\tau_r\in\Omega_r\) and \(x_0=x_0(\tau_1,\dots,\tau_r)\). We have
where \(\theta_1\) and \(\theta_2\) are real numbers with \(0 \leq \theta_1<1\) and \(0 \leq \theta_2<1\). Since \(1 \leq d \leq \eta\ln x\), we obtain
Thus,
Inequality (5.22) is proved.
Substituting (5.22) into (5.21), we obtain
Let \(d\) be an integer such that \(1\leq d \leq \eta\ln x\) and \(d\in {\mathcal M} \). We have (see Lemmas 2.6 and 2.4)
Hence,
We have (see Lemma 2.1)
where \(c_5>0\) is an absolute constant.
Now we estimate \(B\). Since \(\ln(1+u)\leq u\) and \(u>0\), we get
We define
Then,
We obtain \(\ln B\leq 1\), i.e.,
It follows from (5.26)–(5.28) that \(S_3 \leq c_6\ln(k+1)\), where \(c_6>0\) is an absolute constant. Substituting this estimate into (5.25), we obtain
where \(c_7>0\) is an absolute constant. Therefore (see (5.14), (5.20), and (5.29)),
where \(c_8=c_7+2>0\) is an absolute constant. We obtain (see (5.13) and Lemma 2.8)
where \(c_9>0\) is an absolute constant. We put \(C=c_1+c_9\), where \(c_1\) is the constant in (5.12). Then \(C>0\) is an absolute constant and in both cases, \(1\leq k \leq \ln\ln x\) and \(k>\ln\ln x\), we have
Lemma 5.2 is proved. \(\quad\Box\)
Lemma 5.3.
Let \( {\mathcal A} ={\mathbb N},\) \( {\mathcal P} =\mathbb{P},\) \(\alpha=1/5,\) and \(\theta=1/3,\) and let \(C_0=C(1/5,1/3)>0\) be the absolute constant in Proposition 5.1. Let \(\varepsilon\) be a real number with \(0<\varepsilon<1\). Then there is a number \(c_0(\varepsilon)>0\) such that the following holds. Let \(x\in{\mathbb N},\) \(y\in{\mathbb R},\) and \(q\in{\mathbb N}\) satisfy the conditions \(x\geq c_0(\varepsilon),\) \(1\leq y \leq \ln x,\) and \(1\leq q \leq y^{1-\varepsilon}\). Then there is a positive integer \(B\) such that
Furthermore, let \(k\in{\mathbb N},\) \(\rho\in{\mathbb R},\) \(\xi\in{\mathbb R},\) \(R\in{\mathbb R},\) \(\eta\in{\mathbb R},\) and \(a\in{\mathbb Z},\) be such that
Let \( {\mathcal L} =\{L_1,\dots,L_k\}\) be an admissible set of \(k\) linear functions, where \(L_i(n) = q n + a+ q b_i,\) \(i=1,\dots,k,\) \(b_1,\dots,b_k\) are positive integers with \(b_1<\dots<b_k,\) and \(q b_k \leq \eta y\). Then the hypothesis of Proposition 5.1 holds and there exist nonnegative weights \(w_n= w_n( {\mathcal L} )\) with the properties stated in Proposition 5.1; the implied constants in (5.1)–(5.5) are positive and absolute. In (5.31), \(\vartheta>0\) is also an absolute constant.
Proof.
We will choose \(c_0(\varepsilon)\) later; this number is assumed to be large enough. We take \(\delta =1/10\) and let \(c_0(\varepsilon)\geq c(\varepsilon,\delta)=c(\varepsilon,1/10)\), where \(c(\varepsilon,\delta)\) is the quantity in Lemma 4.10. Let \(x\in{\mathbb N}\), \(y\in{\mathbb R}\), and \(q\in{\mathbb N}\) be such that \(x\geq c_0(\varepsilon)\), \(1 \leq y \leq \ln x\), and \(1 \leq q \leq y^{1-\varepsilon}\). By Lemma 4.10, there is a positive integer \(B\) such that
and
where \(c_1\), \(\gamma\), \(c_2\), and \(c_3\) are positive absolute constants. Let (5.32)–(5.35) hold. Let \( {\mathcal L} =\{L_1,\dots,L_k\}\) be an admissible set of \(k\) linear functions \(L_i(n) = q n + a+ q b_i\), \(i=1,\dots,k\), where \(b_1,\dots,b_k\) are positive integers with \(b_1<\dots<b_k\) and \(q b_k \leq \eta y\). Let us show that the hypothesis of Proposition 5.1 holds. First we show that the set \(( {\mathcal A} , {\mathcal L} , {\mathcal P} ,B,x,1/3)\) satisfies Hypothesis 1.
I. \(\,\)Let us show that condition (2) of Hypothesis 1 holds. Let \(L(n)=l_1 n+l_2\in {\mathcal L} \). Clearly, we have
Let us show that
It is not hard to see that
and hence
We obtain
where
Let us show that
for any \(b\in{\mathbb Z}\). Assume the contrary: there is an integer \(b\) such that \((L(b),l_1)>1\). Then there is a prime \(p\) such that \(p\mid l_1\) and \(p\mid L(b)\). Hence \(p\mid l_2\), and we see that \(p\mid L(n)\) for any integer \(n\). Since \(L\in {\mathcal L} \), we see that \(p\mid L_1(n)\dots L_k(n)\) for any integer \(n\). But this contradicts the fact that \( {\mathcal L} =\{L_1,\dots,L_k\}\) is an admissible set. Thus, (5.41) is proved. We also observe that since \((B,q)=1\) and \(l_1=q\), we have
Let \(r\) be an integer with \(1\leq r \leq x^{1/3}\) and \((r,B)=1\). Applying (5.37), we have
provided that \(c_0(\varepsilon)\) is chosen large enough. Hence, we obtain (see (5.41), (5.42) and (5.36))
Applying Lemmas 2.5 and 2.9, we get
where \( \widetilde{c}{} >0\) is an absolute constant. Since \(l_1 x+ l_2-1\geq l_1x\geq x\) (see (5.37)), we obtain
Hence (see, for example, [1, Ch. 22]),
where \(C\) and \(c\) are positive absolute constants. We have
We can assume that
if \(c_0(\varepsilon)\) is chosen large enough. Hence,
where \( \widetilde{C}{} =2 \widetilde{c}{} \kern1pt C\) is a positive absolute constant. Similarly, it can be shown that
where \(C\) and \(c\) are positive absolute constants. Substituting (5.43), (5.48), and (5.49) into (5.40), we obtain
where \(c_4\) and \(c_5\) are positive absolute constants. Applying (5.44)–(5.47), we have
where \(C\) and \(c\) are positive absolute constants. Similarly, it can be shown that
where \(C\) and \(c\) are positive absolute constants. Therefore (see (5.39)),
where \(c_6\) and \(c_7\) are positive absolute constants. We have
provided that \(c_0(\varepsilon)\) is chosen large enough. Hence,
Let us show that
Since \(l_1/ \varphi (l_1)\geq 1\), we see from (5.52) that it is sufficient to show that
This inequality holds if \(c_0(\varepsilon)\) is chosen large enough. Thus, (5.54) is proved. From (5.51), (5.53), and (5.54) we obtain
Now we prove (5.38). Since \(l_1/ \varphi (l_1)\geq 1\), we see from (5.50) and (5.55) that it suffices to establish the estimate
Taking logarithms, we obtain
or, which is equivalent,
Since \(k \leq (\ln x)^{1/5}\), we have
The inequality
holds if \(c_0(\varepsilon)\) is chosen large enough. Inequality (5.56) is proved. Thus, (5.38) is proved.
II. \(\,\)Let us show that condition (1) of Hypothesis 1 holds. We show that
Let \(1 \leq r \leq x^{1/3}\) and \(b\in{\mathbb Z}\). We have
Hence,
We obtain
Hence, \(S \leq x^{1/3}\). Thus, to prove (5.57), it suffices to show that
or, which is equivalent, \((\ln x)^{100 k^2} \leq x^{2/3}\). Taking logarithms, we obtain
Since \(k \leq (\ln x)^{1/5}\), we have
The inequality
holds if \(c_0(\varepsilon)\) is chosen large enough. Thus, (5.57) is proved.
III. \(\,\)Let us show that condition (3) of Hypothesis 1 holds. To this end we show that for any integer \(r\) with \(1\leq r < x^{1/3}\) we have
Let \(1\leq r < x^{1/3}\) and \(b\in{\mathbb Z}\). We may assume that \(c_0(\varepsilon)\geq 2\). Hence, \(r \leq x^{1/3} \leq x\). Applying (5.58), we obtain
and (5.60) is proved. Thus, the set \(( {\mathcal A} , {\mathcal L} , {\mathcal P} ,B,x,1/3)\) satisfies Hypothesis 1.
We can assume that
provided that \(c_0(\varepsilon)\) is chosen large enough. Since \(1\leq B \leq \exp(c_1\sqrt{\ln x})\), we obtain \(1 \leq B \leq x^{1/5}\). Let \(L=l_1n+l_2\in {\mathcal L} \). Applying (5.37), we have \(1\leq l_1 \leq x^{1/5}\) and \(1\leq l_2 \leq x^{1/5}\). Thus, the hypothesis of Proposition 5.1 holds and there are nonnegative weights \(w_n= w_n( {\mathcal L} )\) with the properties stated in Proposition 5.1. In that proposition, the implied constants in (5.1)–(5.5) depend only on \(\alpha\), \(\theta\) and on the implied constants from Hypothesis 1, and in our case these constants are absolute (\(\alpha=1/5\), \(\theta=1/3\), and estimates (5.38), (5.57), and (5.60) hold). Therefore, in our case the implied constants in (5.1)–(5.5) are positive and absolute. Finally, let us denote \(c_1\) by \(\vartheta\). Lemma 5.3 is proved. \(\quad\Box\)
Lemma 5.4.
There are positive absolute constants \(c\) and \(C\) such that the following holds. Let \(\varepsilon\) be a real number with \(0<\varepsilon<1\). Then there is a number \(c_0(\varepsilon)>0,\) depending only on \(\varepsilon,\) such that if \(x\in{\mathbb N},\) \(y\in{\mathbb R},\) \(m\in{\mathbb Z},\) \(q\in{\mathbb Z},\) and \(a\in{\mathbb Z}\) satisfy the conditions \(c_0(\varepsilon) \leq y \leq \ln x,\) \(1 \leq m \leq c \kern1pt \varepsilon\ln y,\) \(1 \leq q \leq y^{1-\varepsilon},\) and \((a,q)=1,\) then
Proof.
Let \( {\mathcal A} ={\mathbb N}\), \( {\mathcal P} =\mathbb{P}\), \(\alpha=1/5\), and \(\theta=1/3\), and let \(C_0=C(1/5, 1/3)>0\) be the absolute constant in Proposition 5.1. Let \(c_0(\varepsilon)\) be the quantity in Lemma 5.3. We will choose \(c(\varepsilon)\) later; this number is large enough. Let \(c(\varepsilon) \geq c_0(\varepsilon)\). Let \(x\in{\mathbb N}\), \(y\in{\mathbb R}\), and \(q\in{\mathbb Z}\) be such that
By Lemma 5.3, there is a positive integer \(B\) such that (5.31) holds. We assume that
where \( \widetilde{C}{} _0>0\) is an absolute constant. We will choose \( \widetilde{C}{} _0\) later. For now, we assume that \( \widetilde{C}{} _0\) is large enough; in particular, \( \widetilde{C}{} _0 \geq C_0\). It follows from (5.61) and (5.63) that \(k\leq (\ln x)^{1/5}\). Thus, (5.32) holds. Let (5.33)–(5.35) hold. Let \( {\mathcal L} =\{L_1,\dots,L_k\}\) be an admissible set of \(k\) linear functions \(L_i(n) = q n + a+ q b_i\), \(i=1,\dots,k\), where \(b_1,\dots,b_k\) are positive integers such that \(b_1< \dots< b_k\) and \(q b_k \leq \eta y\). Then (see Lemma 5.3) the hypothesis of Proposition 5.1 holds and there are nonnegative weights \(w_n= w_n( {\mathcal L} )\) with the properties stated in Proposition 5.1; the implied constants in (5.1)–(5.5) are positive and absolute. We write \( {\mathcal L} = {\mathcal L} (\textbf{b})\) for such a set defined by \(b_1,\dots,b_k\). Denote the class of admissible sets by \( \mathrm{AS} \).
Let \(m\) be a positive integer. We consider
Let \( {\mathcal L} =\{L_1,\dots,L_k\}\) and \(n\) be in the range of summation of \(S\) and \(A_n( {\mathcal L} )>0\). Then the following statements hold:
-
(1)
The number of primes among \(L_1(n),\dots,L_k(n)\) is at least \(m+1\).
-
(2)
For any \(1\leq i \leq k\), \(L_i(n)\) has no prime factor \(p\) such that \(p<x^\rho\) and \(p\nmid B\).
-
(3)
For any linear function \(L=qt+b\notin {\mathcal L} \), where \(b\) is an integer with \(1\leq b \leq 2\eta y\), \(L(n)\) has a prime factor \(p\) such that \(p<x^\rho\) and \(p\nmid B\) (we choose \(\rho\) so that \(x^\rho\) is not an integer; therefore, the conditions \(p \leq x^\rho\) and \(p<x^\rho\) are equivalent). Since \(L(n)> n\geq x>x^\rho\), we see that \(L(n)\) is not a prime number.
As a consequence we obtain the following statements:
-
(i)
None of \(n\in {\mathcal A} (x)\) can make a positive contribution to \(S\) from two different admissible sets (since if \(n\) makes a positive contribution for some admissible set \( {\mathcal L} =\{L_1,\dots,L_k\}\), then the numbers \(L_1(n),\dots,L_k(n)\) are uniquely determined as the integers in \([qn+1, qn+2\eta y]\) with no prime factors \(p\) such that \(p<x^\rho\) and \(p\nmid B\)).
-
(ii)
If \( {\mathcal L} =\{L_1,\dots,L_k\}\) and \(n\) are in the range of summation of \(S\) and \(A_n( {\mathcal L} )>0\), then there can be no primes in the interval \([qn+1, qn+2\eta y]\) apart from possibly \(L_1(n),\dots,L_k(n)\), and so the primes counted in this way must be consecutive.
Let \( {\mathcal L} =\{L_1,\dots,L_k\}\) and \(n\) be in the range of summation of \(S\) and \(A_n( {\mathcal L} )>0\). Let \(1 \leq i \leq k\). If \(p\mid L_i(n)\) and \(p\nmid B\), then \(p \geq x^\rho\). Setting
we have
Since
we obtain
provided that \(c(\varepsilon)\) is chosen large enough. Hence, \(\rho\,\#\Omega \leq 2\), i.e., \(\#\Omega \leq {2}/{\rho}\). We have
Thus, if \( {\mathcal L} =\{L_1,\dots,L_k\}\) and \(n\) are in the range of summation of \(S\) and \(A_n( {\mathcal L} )>0\), then (see (5.1))
where \(C>0\) is an absolute constant.
Let \( {\mathcal L} =\{L_1,\dots,L_k\}\) be in the range of summation of \(S\). We consider
Our aim is to obtain a lower bound for \( \widetilde{S}{} ( {\mathcal L} )\). We write \(w_n\) instead of \(w_n( {\mathcal L} )\) for brevity. Let \(1 \leq i \leq k\). Since \(\# {\mathcal A} (x)= x\), we have (see (5.3))
Hence,
since
We have shown (see (5.55)) that if \(x \geq c_0\), where \(c_0> 0\) is an absolute constant, then for any \(L\in {\mathcal L} \)
We may assume that \(c(\varepsilon) \geq c_0\). Since \( \varphi (B)/B \geq 1/2\) (see (5.31)), we obtain
We have \(|o(1)|\leq 1/2\) in \(S_1'\) if \(x \geq c'\), where \(c'>0\) is an absolute constant. We may assume that \(c(\varepsilon)\geq c'\). Since (see (5.8))
where \(c''>0\) is an absolute constant, we get
We have
provided that \(c(\varepsilon)\) is chosen large enough. Therefore,
where \(c>0\) is an absolute constant.
We have (see (5.2))
provided that \(c(\varepsilon)\) is chosen large enough. Applying (5.5), we obtain
where \(c_2>0\) is an absolute constant. Let \(c_3>0\) be an absolute constant such that
We choose an arbitrary number \(\rho\) in the interval
so that \(x^\rho\) is not an integer. It is clear that \(\rho \leq 1/30\). Let us show that the first inequality in (5.33) holds. It suffices to show that
This inequality is equivalent to
Since \(k \leq (\ln x)^{1/5}\), we have
provided that \(c(\varepsilon)\) is chosen large enough. Thus, the inequalities in (5.33) hold. We have (see (5.67))
Now we estimate the quantity
Let \(b\) be in the range of summation of \(S_4\). Then \(L=qt+b\notin {\mathcal L} \) and
Since \(1\leq B \leq x^{1/5}\), we have (see (5.4))
where \(c_4>0\) is an absolute constant. Since \(B/ \varphi (B) \leq 2\) and \(\rho\) lies in the interval (5.68), we obtain
Hence,
We put
where \(C>0\) is the absolute constant in Lemma 5.2, and
Let us show that
The second inequality in (5.73) is equivalent to the inequality
We may assume that \( \widetilde{C}{} _0\geq 3\); therefore, \(\ln k \geq 1\). We have
provided that \( \widetilde{C}{} _0\) is chosen large enough. The first inequality in (5.73) is equivalent to the inequality
Since \(q \leq \ln x\) and \(k\leq (\ln x)^{1/5}\), we have
provided that \(c(\varepsilon)\) is chosen large enough. Thus, (5.73) holds. We can assume that \(x \geq c\), where \(c\) is the absolute constant in Lemma 5.2, provided that \(c(\varepsilon)\) is chosen large enough. Applying Lemma 5.2 and taking into account that \(\ln(k+1) \leq 2 \ln k\), we have
Substituting this estimate into (5.70), we get (see also (5.71), (5.72), and (5.67))
From (5.69) and (5.74) we obtain
We have (see (5.2))
provided that \(c(\varepsilon)\) is chosen large enough. Applying (5.66) with \(c\) replaced by \(3c_1\), we obtain
where \(c_1>0\) is an absolute constant. We put
It is not hard to see that
Since \(m\) is a positive integer, we see that \(3c_1\ln k - 2m \geq 1\). Hence,
Since \(B^k/ \varphi (B)^k\geq 1\), \(\ln R = (1/9)\ln x\), \( {\mathfrak S} _B( {\mathcal L} )\geq \exp(-c_2k)\), and \(I_k\geq c_3(2k\ln k)^{-k}\), where \(c_2\) and \(c_3\) are positive absolute constants (see (5.6) and (5.7)), it follows that
provided that \( \widetilde{C}{} _0\) is chosen large enough. We obtain
Now we derive a lower bound for \(S'\). First let us show that
The first inequality obviously holds, since we may assume that \( \widetilde{C}{} _0 \geq 2\). To prove the second inequality, it suffices to show that
We have (see (5.62) and (5.72))
where \(c_4>0\) is an absolute constant. Thus, to prove (5.79), it suffices to show that
In particular, from (5.62) it follows that \(q\leq y\). Applying (5.63), we have
provided that \(c(\varepsilon)\) is chosen large enough. Thus, (5.78) is proved.
We put
Applying Lemma 2.13, we have
where \(c_0>0\) is an absolute constant. In particular, from (5.78) it follows that \(\eta y/ q \geq 4\), and so
We obtain
where \(c_5>0\) is an absolute constant. Let us show that
Applying (5.62) and (5.72), we have
where \(c_6>0\) is an absolute constant. Therefore, it suffices to show that
Applying (5.63) and taking into account that \(q\leq y\), we have
provided that \(c(\varepsilon)\) is chosen large enough. Thus, (5.81) is proved.
Let \(b_1 <\dots<b_k\) be positive integers from the set \(\Omega\). Let us show that for any prime \(p\) with \(p\nmid q\) there is an integer \(m_p\) such that \(m_p\not \equiv b_i\pmod{p}\) for all \(1\leq i \leq k\). Let \(p\) be a prime with \(p\nmid q\). If \(p > k\), then the statement is obvious. If \(p \leq k\), then we may put \(m_p=0\); from the definition of the set \(\Omega\) it follows that \(b_i\not\equiv 0\pmod{p}\) for all \(1\leq i \leq k\). Thus, the statement is proved. By Lemma 5.1, \( {\mathcal L} (\textbf{b})\) is an admissible set. Hence (see also Lemma 2.12, (5.80), (5.81), and (5.72)),
where \(c_6>0\) is an absolute constant. We have
provided that \( \widetilde{C}{} _0\) is chosen large enough. Hence,
Substituting this estimate into (5.77), we obtain
Now we obtain an upper bound for \(S\). Applying (5.64) and (5.65), we get
We have (see assertions (1)–(3), (i), and (ii) at the beginning of the proof)
Hence,
Since \(\rho\) lies in the interval (5.68), we have
where \(c_4>0\) is an absolute constant. Since \(\ln R = (1/9)\ln x\), it follows that
provided that \( \widetilde{C}{} _0\) is chosen large enough. Hence,
From (5.82) and (5.83) we obtain
We define
and put \(N_2=\#\Omega_2\). Since \(x\) is a positive integer, we have \(N_1=\#\Omega_1\). Let us show that
Let \(n\in\Omega_1\). Then there are at least \(m+1\) consecutive primes all congruent to \(a\pmod{q}\) in the interval \([qn+1, qn + y]\). Let \(p\) be the first of them. Then \(p\in\Omega_2\). We put
and claim that
Let \(I_j= [qj+1, qj+y]\), \(j\in{\mathbb Z}\). Since \(p\in I_n\), we have \( \Lambda \neq \varnothing\). Let \(l\) be the minimal element in \( \Lambda \). We put \(t=\lceil y\rceil + 1\). Then \(t>y\) and
Hence, \(p\notin I_j\) for \(j \geq l+t\) and \(j\leq l-1\). We obtain \(\# \Lambda \leq t\). Thus, (5.86) is proved; (5.85) follows from (5.86). We have \(\lceil y\rceil + 1\leq y+2 \leq 2y\) provided that \(c(\varepsilon)\) is chosen large enough. Since
we obtain (see (5.84))
We put
Then
Since \(q \leq y\), we have
Let us show that
Since \(q\leq y^{1-\varepsilon}\leq y\) and \(k \leq y^{\varepsilon/14}\), we have
Therefore, to prove (5.91), it suffices to show that
Taking logarithms, we obtain
or, which is equivalent,
Since \(y \leq \ln x\) and \(0<\varepsilon<1\), we have \(k \leq (\ln x)^{\varepsilon/14}\leq (\ln x)^{1/14}\). Then
provided that \(c(\varepsilon)\) is chosen large enough. Thus, (5.91) is proved. From (5.90) and (5.91) it follows that
Applying (5.87), (5.89), and (5.92), we obtain
We have (see (2.2))
where \(c_1>0\) and \(c_2=2c_1>0\) are absolute constants. Therefore,
Using the inequality \(\ln(1+x)\leq x\), \(x>0\), we obtain \(\ln\ln(q+2)\leq \ln(1+q)\leq q\). Hence,
We can assume that \(4c_2 \leq 2^{3k^5}\) if \( \widetilde{C}{} _0\) is chosen large enough. We have
We can also assume that \(3k^5\leq k^6\) if \( \widetilde{C}{} _0\) is chosen large enough. Hence,
We have \((2e)^{k^6} \leq 2^{k^7}\) if \( \widetilde{C}{} _0\) is chosen large enough. It is clear that \(q^{k^6} \leq q^{k^7}\). Then
Further (see (5.61)),
We obtain
From (5.75) and (5.76) we find
provided that \( \widetilde{C}{} _0\) is chosen large enough. Therefore, \(k^7 \leq \exp(14 \widetilde{c}{} m)\). Since \( \widetilde{C}{} _0\) is a positive absolute constant, we see from (5.75) that \( \widetilde{c}{} \) is a positive absolute constant. We have
where \(C=14 \widetilde{c}{} >0\) is an absolute constant. Combining (5.88), (5.93), and (5.95) we obtain
Applying (5.94), we see that the inequality \(k\leq y^{\varepsilon/14}\) holds if \(\exp(2 \widetilde{c}{} m) \leq y^{\varepsilon/14}\). This inequality is equivalent to
where \(c= 1/(28 \widetilde{c}{} \kern1pt )>0\) is an absolute constant. Let us redenote \(c(\varepsilon)\) by \(c_0(\varepsilon)\). Lemma 5.4 is proved. \(\quad\Box\)
Lemma 5.5.
There are positive absolute constants \(c\) and \(C\) such that the following holds. Let \(\varepsilon\) be a real number with \(0<\varepsilon<1\). Then there is a number \(c_0(\varepsilon)>0,\) depending only on \(\varepsilon,\) such that if \(x\in{\mathbb R},\) \(y\in{\mathbb R},\) \(m\in{\mathbb Z},\) \(q\in{\mathbb Z},\) and \(a\in{\mathbb Z}\) satisfy the conditions \(c_0(\varepsilon)\leq y \leq \ln x,\) \(1 \leq m \leq c \kern1pt \varepsilon\ln y,\) \(1 \leq q \leq y^{1-\varepsilon},\) and \((a,q)=1,\) then
Proof.
Let \(c\), \(C\), and \(c_0(\varepsilon)\) be the quantities mentioned in Lemma 5.4. We will choose a quantity \( \widetilde{c}{} _0(\varepsilon)\) and an absolute constant \( \widetilde{C}{} \) later; they will be large enough. In particular, let \( \widetilde{c}{} _0(\varepsilon) \geq c_0(\varepsilon)\) and \( \widetilde{C}{} \geq C\). Let \(x\in{\mathbb R}\), \(y\in{\mathbb R}\), \(m\in{\mathbb Z}\), \(q\in{\mathbb Z}\), and \(a\in{\mathbb Z}\) be such that \( \widetilde{c}{} _0(\varepsilon)\leq y \leq \ln x\), \(1\leq m \leq c \kern1pt \varepsilon\ln y\), \(1\leq q \leq y^{1-\varepsilon}\), and \((a,q)=1\). We put \(l=\lceil x\rceil\). Then, by Lemma 5.4, we have
Since \(x \leq l < x+1\), we have
Therefore,
We have \(x+1 \leq x^2\) provided that \( \widetilde{c}{} _0(\varepsilon)\) is chosen large enough. Hence,
Since \(\pi(2ql)\geq \pi(2qx)\), we have
Then
provided that \( \widetilde{C}{} \) is chosen large enough. Since \( \widetilde{C}{} \geq C\), we have
Further,
Hence,
Let us denote \( \widetilde{c}{} _0(\varepsilon)\) by \(c_0(\varepsilon)\) and \(2 \widetilde{C}{} \) by \(C\). Lemma 5.5 is proved. \(\quad\Box\)
Let us complete the proof of Theorem 1.1. Let \(c_0(\varepsilon)\), \(c\), \(C\) be the quantities in Lemma 5.5. We will choose a quantity \( \widetilde{c}{} _0(\varepsilon)\) and an absolute constant \( \widetilde{C}{} \) later; they will be large enough. Let \( \widetilde{c}{} _0(\varepsilon) \geq c_0(\varepsilon)\) and \( \widetilde{C}{} \geq C\).
Let us prove the following statement.
Proposition 5.2.
Let \(\varepsilon\) be a real number with \(0<\varepsilon<1\). Let \(t\in{\mathbb R},\) \(y\in{\mathbb R},\) \(m\in{\mathbb Z},\) \(q\in{\mathbb Z},\) and \(a\in{\mathbb Z}\) be such that
Then
Proof.
Indeed, since \(t\geq 100\), we have \(2\ln t \geq 1\). Hence,
We have \(q\leq y^{1-\varepsilon} \leq y \leq \ln t\). Therefore,
We put \(x={t}/{2q}\). Then \(x\in{\mathbb R}\), \(y\in{\mathbb R}\), \(m\in{\mathbb Z}\), \(q\in{\mathbb Z}\), and \(a\in{\mathbb Z}\) are such that
By Lemma 5.5, we have
Returning to the variable \(t\), we obtain (5.99). \(\quad\Box\)
Let us prove the following statement.
Proposition 5.3.
Let \(\varepsilon\) be a real number with \(0<\varepsilon<1\). Let \(t\in{\mathbb R},\) \(m\in{\mathbb Z},\) \(q\in{\mathbb Z},\) and \(a\in{\mathbb Z}\) be such that
Then
Proof.
We need the following
Lemma 5.6.
Let \(t\) be a real number with \(t \geq 100\). Then
The proof of lemma 5.6 is a simple exercise in calculus, and we omit it.
We put
Since \(t\geq 100\), we have (see Lemma 5.6)
Therefore,
We may assume that \( \widetilde{c}{} _0(\varepsilon) \geq 2^{1/\varepsilon}\). Since \(t \geq 100\), we have \(t/(2\ln t)\leq t\) and
Hence,
Therefore,
From (5.100) and the last inequality in Lemma 5.6 we find
Since \(1 \leq q \leq (\ln t)^{1-\varepsilon}\), we have \(1 \leq q \leq y^{1 - \varepsilon/2}\). Applying Proposition 5.2 with \(\varepsilon/2\) and the second inequality of Lemma 5.6, we have
The statement is proved. \(\quad\Box\)
Proposition 5.4.
Let \(\varepsilon\) be a real number with \(0<\varepsilon<1\). Let \(t\in{\mathbb R},\) \(y\in{\mathbb R},\) \(m\in{\mathbb Z},\) \(q\in{\mathbb Z},\) and \(a\in{\mathbb Z}\) be such that
Then
Proof.
Since \(y \leq \ln t\), we have
Applying Proposition 5.3, we obtain
For \(0<\varepsilon<1\) we define the quantity \(t_0(\varepsilon)\) as follows:
Let us prove the following statement.
Proposition 5.5.
Let \(\varepsilon\) be a real number with \(0<\varepsilon<1\). Let \(t\in{\mathbb R},\) \(y\in{\mathbb R},\) \(m\in{\mathbb Z},\) \(q\in{\mathbb Z},\) and \(a\in{\mathbb Z}\) be such that
Then
Proof.
Consider two cases. If
then \(t\), \(y\), \(m\), \(q\), and \(a\) satisfy the hypothesis of Proposition 5.4, which yields
If
then \(t\), \(y\), \(m\), \(q\), and \(a\) satisfy the hypothesis of Proposition 5.2, which yields
For \(0<\varepsilon<1\) we put
Let us prove the following statement.
Proposition 5.6.
Let \(\varepsilon\) be a real number with \(0<\varepsilon<1\). Let \(t\in{\mathbb R},\) \(y\in{\mathbb R},\) \(m\in{\mathbb Z},\) \(q\in{\mathbb Z},\) and \(a\in{\mathbb Z}\) be such that
Then
Proof.
We have
Applying Proposition 5.5, we obtain
We may assume that \( \widetilde{C}{} \geq 2\). Therefore, \(\exp( \widetilde{C}{} m)\geq \widetilde{C}{} m\geq \widetilde{C}{} \geq 2\). Hence, \(2\exp( \widetilde{C}{} m)\leq \exp(2 \widetilde{C}{} m)\). We have
Further,
We obtain
Relations (5.101) and (5.102) imply the required assertion. \(\quad\Box\)
Let us denote \(\rho(\varepsilon)\) by \(c_0(\varepsilon)\), \(c/4\) by \(c\), and \(2 \widetilde{C}{} \) by \(C\). Theorem 1.1 is proved. \(\quad\Box\)
Proof of Corollary 1.1 .
Let \(c_0(\varepsilon)\), \(c\), and \(C\) be the quantities in Theorem 1.1. We put
Let \(m\) be a positive integer. Let \(x\in{\mathbb R}\) and \(y\in{\mathbb R}\) be such that \(\exp(C_1m) \leq y \leq \ln x\). Then
The last inequality implies
Putting \(q=1\) and \(a=1\), we have
Applying Theorem 1.1 with \(\varepsilon=1/2\), we see that
Let us redenote \(C_1\) by \(C\). Corollary 1.1 is proved. \(\quad\Box\)
References
H. Davenport, Multiplicative Number Theory, 3rd ed. (Springer, New York, 2000), Grad. Texts Math. 74.
H. Halberstam and H.-E. Richert, Sieve Methods (Academic, London, 1974), LMS Monogr. 4.
R. R. Hall and G. Tenenbaum, Divisors (Cambridge Univ. Press, Cambridge, 1988), Cambridge Tracts Math. 90.
A. E. Ingham, The Distribution of Prime Numbers (Cambridge Univ. Press, London, 1932), Cambridge Tracts Math. Math. Phys. 30.
J. Maynard, “Dense clusters of primes in subsets,” Compos. Math. 152 (7), 1517–1554 (2016).
K. Prachar, Primzahlverteilung (Springer, Berlin, 1957), Grundl. Math. Wiss. 91.
I. M. Vinogradov, Elements of Number Theory (Gos. Izd. Tekhn.-Teor. Lit., Moscow, 1953; Dover, New York, 1954).
Acknowledgments
The author is deeply grateful to Sergei Konyagin and Maxim Korolev for their attention to this work and useful comments. The author also expresses his gratitude to Mikhail Gabdullin and Pavel Grigor’ev for useful comments and suggestions.
Funding
This work was performed at the Steklov International Mathematical Center and supported by the Ministry of Science and Higher Education of the Russian Federation (agreement no. 075-15-2019-1614).
Author information
Authors and Affiliations
Corresponding author
Additional information
Published in Russian in Trudy Matematicheskogo Instituta imeni V.A. Steklova, 2021, Vol. 314, pp. 152–210 https://doi.org/10.4213/tm4163.
On the occasion of the 130th anniversary of I. M. Vinogradov’s birth
Rights and permissions
About this article
Cite this article
Radomskii, A.O. Consecutive Primes in Short Intervals. Proc. Steklov Inst. Math. 314, 144–202 (2021). https://doi.org/10.1134/S008154382104009X
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1134/S008154382104009X