Journal of Inequalities in Pure and Applied Mathematics http://jipam.vu.edu.au/ Volume 3, Issue 4, Article 53, 2002 ON THE SEQUENCE (p2n − pn−1 pn+1 )n≥2 LAURENŢIU PANAITOPOL FACULTY OF M ATHEMATICS 14 ACADEMIEI S T. RO-70109 B UCHAREST, ROMANIA pan@al.math.unibuc.ro Received 17 December, 2001; accepted 24 May, 2002 Communicated by L. Toth A BSTRACT. Let pn be the n-th prime number and xn = p2n − pn−1 pn+1 . In this paper, we study sequences containing the terms of the sequence (xn )n≥1 . The main result asserts that the series P∞ 2 x /p n n is convergent, without being absolutely convergent. n=1 Key words and phrases: Prime Numbers, Sequences, Series, Asymptotic Behaviour. 2000 Mathematics Subject Classification. 11A25, 11N05, 11N36. 1. I NTRODUCTION We shall use the following notation: pn the n-th prime number xn = p2n − pn−1 pn+1 for n ≥ 2, dn = pn+1 − pn for n ≥ 1, pn+1 qn = for n ≥ 1, pn f (x) g(x) if there exist c1 , c2 , M > 0 such that c1 f (x) < g(x) < c2 f (x) for every x > M . It will be our aim here to study the sequence (xn )n≥2 defined above. It was proved in [1] that the sequence (dn )n≥1 is not nonotone. A similar result holds for the sequence (qn )n≥1 as well. This means that the sequence (xn )n≥2 has infinitely many positive terms, and infinitely many negative terms, hence it is not monotone. In [1], the so-called method of the triple sieve (due to Vigo Brun) was used to prove that X q n log x. log (1.1) qn−1 pn ≤x ISSN (electronic): 1443-5756 c 2002 Victoria University. All rights reserved. 090-01 2 L AUREN ŢIU PANAITOPOL This result plays an essential role in the following paragraph of the present paper. Another useful result is proved in [4]: n ∞ X dn is convergent. (1.2) the series pn n=1 2. T HE S ERIES P∞ xn n=2 p2n P xn Theorem 2.1. The series ∞ n=2 p2n is convergent, but it is not absolutely convergent. In order to prove this fact, we need the following lemmas. Lemma 2.2. For x ≥ − 12 , we have x2 . 2 Proof. The inequalities are well known for x > 0. When x ∈ − 21 , 0 , they take on the form x2 + |x| ≥ | log(1 + x)| ≥ |x| − x2 . 2 2 Let f, g : − 21 , 0 → R be defined by f (x) = log(1+x)−x− x2 and g(x) = log(1+x)−x+x2 , respectively. We have f 0 (x) = − x(x+2) ≥ 0, and g 0 (x) = x(2x+1) ≤ 0. Since f is increasing 1+x 1+x and f (0) = 0, we get f (x) ≤ 0. On the other hand, we have g 0 (x) < 0 and g(0) = 0, so that g(x) ≥ 0. P∞ dn −dn−1 Lemma 2.3. The series n=2 pn is convergent. P k−1 Proof. Denote Sn = nk=2 dk −d , so that pk x2 − x ≥ − log(1 + x) ≥ −x − n 1 dn X d2k−1 + − . Sn = pn k=2 pk pk−1 2 P∞ d2k−1 Since limn→∞ pn+1 = 1, it suffices to prove that the series k=2 pk pk−1 is convergent. Since pn 2 2 P∞ d2k−1 dk−1 dk−1 ∼ and the terms of the series are positive, it follows that the series k=2 pk pk−1 pk pk−1 pk−1 P∞ dk−1 2 and k=2 pk−1 are simultaneously convergent or not. Now just use (1.2) and the proof ends. P∞ x2n Lemma 2.4. The series n=2 p4 is convergent. n Proof. Since xn = dn dn−1 + pn (dn−1 − dn ), it follows that xn dn dn−1 dn−1 − dn (2.1) = + 2 pn p2n pn hence 2 2 dn dn−1 (dn−1 − dn )2 x2n (2.2) ≤2 + . p4n p4n p2n P P∞ d2n−1 d2n−1 d2n−1 d2n Since the series ∞ is convergent and ∼ , it follows that the series 2 2 2 n=1 pn n=2 p2n is pn pn−1 P∞ max(d2n ,d2n−1 ) convergent as well. This implies that the series n=2 is also convergent. Since p2 n d2n d2n−1 p4n < max(d2n , d2n−1 ) p2n J. Inequal. Pure and Appl. Math., 3(4) Art. 53, 2002 and max(d2n , d2n−1 ) (dn−1 − dn )2 < , p2n p2n http://jipam.vu.edu.au/ O N THE S EQUENCE (p2n − pn−1 pn+1 )n≥2 we deduce by (2.2) that the series x2n n=2 p4n P∞ 3 is convergent. Lemma 2.5. For x > 0 we have X qn − qn−1 qn−1 log x. pn ≤x Proof. In view of Lemma 2.2, we have 2 qn − qn−1 qn − qn−1 1 qn − qn−1 2 qn − qn−1 q n ≥ log ≥ − + . qn−1 qn−1 qn−1 qn−1 2 qn−1 Since pn +1 pn − pn−1 qn − qn−1 xn pn = =− 2, pn qn−1 pn pn−1 we have 1 x2n qn qn − qn−1 qn x2n (2.3) · + log > ≥ log − . 2 p4n qn−1 qn−1 qn−1 p4n Now the desired conclusion follows by (1.1) and Lemma 2.4. Proof of Theorem 2.1. By the relation (2.1) we have n n n X xk X dk dk−1 X dk−1 − dk Sn = = + . 2 2 p p p k k k k=2 k=2 k=2 P max(d2n ,d2n−1 ) Since dk dk−1 ≤ max(d2k , d2k−1 ), and since the series ∞ is convergent, see the n=2 p2n P∞ dn dn−1 proof of Lemma 2.4, it follows that the series n=2 p2 is convergent too. Consequently the n P sequence (Sn0 )n≥1 , defined by Sn0 = nk=2 dk dpk−1 is convergent. Lemma 2.3 implies that the 2 k P sequence (Sn00 )n≥1 , defined by Sn00 = nk=2 dk−1pk−dk is convergent as well. It then follows that the P xn sequence (Sn )n≥2 is convergent, that is, the series ∞ is convergent. n=2 p2n |x | n−1 On the other hand, Lemma 2.5 and the relation qnq−q = p2n imply that n−1 n X |xn | log x, 2 p n p ≤x (2.4) n hence the series P∞ xn n=2 p2n is not absolutely convergent. 3. T HE S ERIES P∞ |xn | n=2 p2n logα n P |xn | Since the series ∞ n=2 p2n is divergent, it is natural to study what “correction” does it need to become convergent. In this connection, we prove the following fact. P |xn | Theorem 3.1. The series ∞ n=2 p2 logα n is convergent if and only if α > 1. n Proof. We are going to put to use a technique from [3]. To P begin with, we recall an inequality due P to Abel: Let ak , bk ∈ R, k ∈ 1, n such that, if Si = ik=1 bk , then Si ≥ 0 for i ∈ 1, n. Then ni=1 ai bi = S1 (a1 − a2 ) + S2 (a2 − a3 ) + · · · + Sn−1 (an−1 − an ) + Sn an , which implies the inequalities n X (3.1) ai bi ≥ an Sn provided a1 ≥ · · · ≥ an , i=1 J. Inequal. Pure and Appl. Math., 3(4) Art. 53, 2002 http://jipam.vu.edu.au/ 4 L AUREN ŢIU PANAITOPOL and n X (3.2) ai bi ≤ an Sn when a1 ≤ · · · ≤ an . i=1 It follows by (2.4) that there exist c1 and c2 such that 0 < c1 < c2 and X |xn | (3.3) c1 log x < < c2 log x for all x ≥ 2. 2 p n p ≤x n 0 For α > 0 and n ≥ 1, we denote a1 = 1, b1 = 0 and for i ≥ 2 ai = log1α i and bi = ci · |xp2i | , where P i c0 > 0 is chosen such that S1 , S2 , . . . , Sn ≥ 0. Such a choice is possible because 2≤i≤x 1i ∼ log x and (3.3) holds. 0 P P P 2 i| It now follows by (3.1) that ni=2 log1α i ci − |xp2i | ≥ 0, that is, ni=2 p2|xlog < c0 ni=2 i log1 α i . α i i i P P |xn | 1 Since the series ∞ is convergent for α > 1, we deduce that the series ∞ α i=2 i log i n=2 p2n logα n is convergent as well. One can similarly show that there exists c00 > 0 such that n n X X |xi |2 1 00 > c α α . 2 p log i i log i i i=2 i=2 P 1 Since the series ∞ i=2 i logα i is divergent for α ≤ 1, it follows that in this case the series P∞ |xn | n=2 p2 logα n is in turn divergent. n R EFERENCES [1] P. ERDŐS AND A. RÉNYI, Some problems and results on consecutive primes, Simon Stevin, 27 (1950), 115–125. [2] P. ERDŐS AND P. TURÁN, On some new question on the distribution of prime numbers, Bull. Amer. Math. Soc., 54 (1948), 371–378. [3] L. PANAITOPOL, Properties of the series of differences of prime numbers, Publications du Centre de Recherches en Mathématiques Pures. Univ. Neuchâtel, Sér. I, Fasc., 31 (2000), 21–28. [4] P. VLAMOS, Inequalities involving the sequence of the differences of the prime numbers, Publications du Centre de Recherches en Mathématiques Pures. Univ. Neuchâtel, Sér. I, Fasc., 32 (2001), 32–40. J. Inequal. Pure and Appl. Math., 3(4) Art. 53, 2002 http://jipam.vu.edu.au/