Abstract Generated abstract
This paper studies representations of natural numbers in the form n equals p plus a power of a fixed integer a, where p is prime and the exponent is nonnegative. Using a counting function for such representations and Romanov’s identity relating its second moment to prime differences and differences of powers, the author derives lower bounds for the number of integers having a bounded number of representations. The main result gives a quantitative estimate for integers up to 2x with between one and k representations, implying in particular that, for sufficiently large a, many integers up to x have a unique representation, and that for arbitrary a many have at most a fixed number of representations. Additional results show that some integers have more than a constant multiple of log log n representations, while integers with representation count exceeding any function tending to infinity have density zero.
Full Text
MATHEMATICS
A. F. LAVRIK
ON THE REPRESENTATION OF NUMBERS IN THE FORM OF THE SUM OF A PRIME NUMBER AND A POWER OF A GIVEN INTEGER
(Presented by Academician I. M. Vinogradov on 7 III 1957)
§ 1. The present paper is devoted to the study of the question of the number of representations of natural numbers \(n\) in the form
\[ n=p+a^i, \]
where \(p\) is a prime number; \(a \geqslant 2\) is a given integer; \(i \geqslant 0\) is an integer.
Theorem 1. In the interval \((0,x)\) there are more than \(\dfrac{\alpha x}{\lg a}\) numbers representable in one and only one way in the form of the sum of a prime number and a power of the given integer \(a>a_0\), where \(\alpha\) is an absolute positive constant.
Without a restriction on the number \(a\), Theorem 2 is valid.
Theorem 2. There exists a constant number \(k\), independent of \(x\) and \(a\), such that the number of numbers \(n \leqslant x\) for which the equation \(n=p+a^i\), where \(p\) is prime, \(a \geqslant 2\) is a given integer, \(i \geqslant 0\) is an integer, has \(1,2,\ldots,k\) solutions, will be more than \(\dfrac{\gamma x}{\lg a}\), where \(\gamma>0\) is a constant.
Theorems 1 and 2 refine known results of N. P. Romanov \((^1)\), E. Landau \((^3)\), and are obtained from a more general theorem.
Theorem 3. Let \(\psi(n,x)\) be the number of solutions of the equation
\[ n=p+a^i, \]
where \(p \leqslant x\) is a prime number, \(a \geqslant 2\) is a given integer, \(i \geqslant 0\) is an integer, \(a^i \leqslant x\). Further, let \(F_m(x)\) be the number of numbers \(n \leqslant 2x\) for which \(\psi(n,x)=m\), \(m>0\) an integer; let \(k\) be any odd positive number. Put also
\[ \Phi_k(x)=F_1(x)+\cdots+F_k(x). \]
Then
\[ \Phi_k(x) \geqslant \frac{4kx}{(k+1)^2 \lg a} \left(1-\frac{c_1 \lg^3 \lg 2a}{k \lg a}-\frac{c_2}{\lg x}\right), \]
where \(c_1,c_2\) are positive absolute constants.
We outline the proof of Theorem 3. We have
\[ \sum_{m=1}^{k} mF_m(x) = \sum_{n=1}^{2x}\psi(n,x) - \sum_{\substack{n=1\\ \psi(n,x)>k}}^{2x}\psi(n,x). \tag{1} \]
Next, we note that the subtracted term in equality (1) does not exceed
\[ \sum_{n=1}^{2x}\psi(n,x)[\psi(n,x)-k] + \sum_{m=1}^{k-1} m(k-m)F_m(x). \tag{2} \]
From (1) and (2) it follows that
\[ \max_{1\le m<k}\{m(k-m+1)\}\Phi_k(x)\ge \sum_{n=1}^{2x}\psi(n,x)[k+1-\psi(n,x)]. \tag{3} \]
Now we shall use the identity of N. P. Romanov \({}^{1}\)
\[ \sum_{n=1}^{2x}\psi^2(n,x)= \sum_{n=1}^{2x}\psi(n,x)+2\sum_{n=1}^{x}A_1(n,x)A_2(n,x), \tag{4} \]
where \(A_1(n,x)\) and \(A_2(n,x)\) are, respectively, the numbers of solutions of the equations:
\[ p_i-p_j=n,\qquad p_i,p_j\le x\text{ — prime numbers;} \]
\[ a^u-a^t=n,\qquad a^u,a^t\le x,\qquad u,t\ge0\text{ — integers.} \]
From (3) and (4) we obtain
\[ \Phi_k(x)\ge \frac{4k}{(k+1)^2}\sum_{n=1}^{2x}\psi(n,x) - \frac{8}{(k+1)^2}\sum_{n=1}^{x}A_1(n,x)A_2(n,x). \tag{5} \]
It is not difficult to see that
\[ \sum_{n=1}^{2x}\psi(n,x)=\pi(x)N(x), \tag{6} \]
where \(\pi(x)\) is the number of primes \(\le x\); \(N(x)\) is the number of numbers \(a^i\le x\).
On the basis of the investigations of Viggo Brun, L. G. Schnirelmann \({}^{2}\), N. P. Romanov \({}^{1}\), and E. Landau,
\[ \sum_{n=1}^{x}A_1(n,x)A_2(n,x)<cx\frac{\lg^3\lg 2a}{\lg^2 a}, \tag{7} \]
where \(c>0\) is an absolute constant.
Combining the estimates (5)—(7), we obtain Theorem 3.
§ 2. The following two propositions are a supplement to Theorems 1 and 2.
Theorem 4. There exists an infinite set of numbers \(n\) for which (in the notation of Theorem 3), as \(x\to\infty\), we have
\[ \psi(n,x)>\delta\lg\lg n, \]
where \(\delta>0\) is some constant.
Theorem 5. If \(\gamma(x)\) is any positive function, increasing without bound as \(x\to\infty\), and \(M(x)\) is the number of numbers \(n\le x\) for which
\[ \psi(n,x)>\gamma(x), \]
then
\[ M(x)=o(x). \]
Tashkent State Pedagogical Institute
named after Nizami
Received
4 III 1957
CITED LITERATURE
\({}^{1}\) N. P. Romanov, Uspekhi Mat. Nauk, 7, 47 (1940).
\({}^{2}\) L. G. Schnirelmann, Uspekhi Mat. Nauk, 7, 7 (1940).
\({}^{3}\) E. Landau, Acta Arithmetica, 1, 43 (1935).
\({}^{4}\) K. Prachar, J. London Math. Soc., 29, No. 3 (1954).