Scribd Logo
Upload

Welcome to Scribd!

  • Cohen 7

    Uploaded by

    berkane
    11 views9 pages

    Original Title

    cohen7

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Report this Document

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    11 views9 pages

    Cohen 7

    Uploaded by

    berkane

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    of 9
    Elementary estimates for certain types of integers By ECKFORD COHEN and K, JOSEPH DAYIS in Greenville (North Carolina, U.S.A.) 1. Introduction, For each integer k=2, let L, represent the set of positive integers n such that each prime factor of m occurs with multiplicity at least k. Let J,(n) denote the characteristic function of the set Ly, and for real x21, let L,(x) be the number of integers contained in Z, and not exceeding x. Let Q be the set of squarefree integers and q(n) the characteristic function of Q. The Riemann zeta- function will be denoted {(s) for real s. The starred references of this paper refer to the bibliography of the paper [2] by the first author, All O-constants which occur are understood to depend upon k. In 1934 Erpds and Szexeres [5*] obtained the following estimate for L,(x): (1.1) Ty (oe) = cyt! + 0 (x10) where ¢ is a constant. This was proved by elementary means without any essential use of Dirichlet series, Later BATEMAN and GrosswAtp obtained (1. 1) in the stronger form. (1.2) Tale) = ogxt tcf xte42) 4 Q(gtIOa+ID, where cf, like cj, is independent of x. While the Bateman—Grosswald proof is elementary, it makes use of the uniqueness theorem for Dirichlet series (see Remark I below). . It is the purpose of the present paper to establish certain weaker estimates for 1,(x) by strictly elementary methods. In particular, we show in § 6, without appealing to the uniqueness theorem, that (1.3) (3) = 6438 eh x VOD 4 O (e142), ‘The argument used in the paper is an elaboration of the method of Expds and SzexeREs [5*]. We require, in addition, estimates for some special sums (§ 4) and an asymptotic formula for the average of a certain divisor function (5). In §7 we give a simple, independent proof of the slightly weaker form of (1. 3) with the O-term O(e!&+Iog x), ma Al rights reserved © Bolyal Institute, University of Saaged 364 B, Cohon—K, J. Davis - Remark 1. The case k=2 is oxceptional with respect to the above discussion of (1.2). In fact, an clementary proof of (1.2) in this case has been given by Barmman [1%]; also see [2] and [3, § 3). 2. Density of Z. Our first estimate for Z,(x) is given in the following theorem, Let L,=L, Theorem 1. The set L has density 03 that is, lim 2@) sre x Proofs of this result have been given by Frnier and Tournim [6*, § 9] and Scuommnre [10*, § 12]. The corresponding result for Ly, k 2 follows immediately. 3. O-estimate for Z,(x). We first prove a characterization of the set Z,. Lemma 1. A necessary and sufficient condition that an integer n be in Iy is that it admit a representation of the form 1 n= dd}. dBA, dydy dy 4ld. Proof. Suppose 7 can be written in the form (3.1), and let pln, p prime. Then pld and hence p*|n. This proves the sufficiency. Now suppose n€I,, n=p{'...p®, 4 =k (=1,...,8) where p; the distinct prime divisors of n. Now ¢ = gk+r, 4>0, 0Sr Ik, 1 (4.3) y F< 0G), xe. Proof. By partial summation and the definition of (rn), vate hi Ta), a = Zh (5 wey) ae hence, by Theorem 2, since (I+1/n)' = 1+0(1/n), Din = O( Snel) 4 O(etit-, oa ae If = 1/k, the first O-term in the last expression is O(x"/*-) or O(log x) according as s1, we have with y>x, 1 } Ly) La) 1 = Ae Gaur B+) OH O( Sate 9) +0(staa), and since both O-terms are O(x'/*~*) the lemma results as y—> <>, We define o*(s, n) to be the sum of the s-th powers of the square free divisors of n, and o(s,n) to be the sum of the s-th powers of all divisors of n, Place An) =0%(0, n). we Al rights reserved © Bolyal Institute, University of Saaged 366 £, Cohen—K, J, Davis Lemma 3, If 0k-+2, (k-+2) docs not divide e, p, does not divide m, for each i. Therefore, Sena Be > eT oie where the second summation is over all ordered r-tuples of natural numbers €15 €25 +++ € such that ¢ does not divide ¢,¢>e (=1,...,r) and all retuples of prime numbers p;, po, ...sP, such that py~4e)-1 for all p, it follows that the product, and hence the series, converges. 5, A divisor function. We first recall a known estimate for the Legendre totient function (x, n), which denotes the number of positive integers. =x prime ton. Lemma 4 @f. [I). f0Sa0,s#1, (5.2) N.@r) = 2 Lat = C(s) @e()/r*— p(A)|r(s— Lx 1 + OCA #0*(—a, 1), n= tniformly in x and r, where 9(r) = 3 ad) (ld), ex) = 9), gw denoting the Mobius function. Now suppose a, 6, i and m to be positive integers. For positive integers n, let 1 (nv) denote the number of decompositions of n in the form n=d"/* where @m)=(f,h)=1. We now are ready to prove the main result of this section, an estimate for the summatory function 7"}"(x) of et4i(n). 1, then for 1>a=0, Al rights reserved © Bolyal Institute, University of Saaged 368 B, Cohon—K. J, Davis Put c= a+b, r=a/b, s=bfa. Theorem 3. (ef. 4, Theorem 3.1 in case m=) If b=azl, r>az0, then for x= 1 Tin) = Ay, y 8 tb yyy Xl? + O(R* DI, (h, m)), where eq(h, m)=max (o%(—a, h), o*(—c, m)), ns = EE) P CR) MAMA, Onn = L0H) @-Cn)lhm" Proof, We have TIC) = Dave =, 2 I Prax whore in the last sum, (4, m)=(J, h)=1. Thus (6.3) THE) = Flt Btn Zt tye aseere Since d and fin the summation cannot both simultaneously be >x**, Each sum of course still has the conditions d'/*=x, (d, m)=(f, i) =1, Let these three sums be denoted by Z,, 22, 2, respectively. For the first summation one obtains by Lemma 6, since aa/b, PO OC x210 4. O(ec(h, mx"), and on applying a similar argument to 2, and Z3, 6.) Za = CLAD pany 4 9OD Ozh Oech, mpxe4"9, 66 3= 2) oc 2216-4 Oxf! Q5(h, m)). The theorem results on the basis of (5. 3), (5. 4), (5.5), and (5. 6). 6. Asymptotic estimation of L,(x). We first introduce some notation and point ‘out a few elementary facts that will be useful for the later discussion. We denote by 4;, By the sets of those positive integers all of whose prime divisors have mul- tiplicity on the ranges k+ 1 = ¢~<2k,and k+2 =t~=2k, respectively, with By = {1}. Note that BS Lys Al rights reserved © Bolyal Institute, University of Saaged Elementary estimates for integers 369 Remark 2. If a€Z,, then a has a unique factorization a=d*e where e€ A,. Remark 3, If e€4,, then e has a unique factorization e=g**"A, where 8€Q, heB, and (g, h)=1. The following result is well known. Remark 4. For positive integers 1, g(n)= > ule). . wean ‘The proof of our main result depends upon the following representation of h(7). Pan LO akere Df thn where the summation is over integers d, e, f and h such that h€ By and (e, hi) =(f, h) = 1. Proof, By Remarks 2 and 3 4@= Zl= 7 1 _ A) denn tg Tpen ceae 062 the last two sums with the conditions héB,, (g, A)=1. Remark 4. ‘The folowing expansion will be needed (Cf. [4], (3. 4)): (6.1) 2 fool =rK(s)or) — (s> 1). net Since 94(0) = nt JT (1—1/p') we have Lemma 8, 4) The lemma results by Lemma 9. If s=1, then 9,(n)/n’ is bounded; for s=1, ,(n)[n’ is bounded away from zero. In particular, for each s>1, (n) has the order of magnitude of W as n+, Put r =k+1,¢ = 2k-+1. Theorem 4. If x22, then 2) Ly) = e424 ext 4 OCH), where o, and ¢, are defined by © *@nlk) a eal: OO) 2 5 (5 “| heBD, and 4 =4y,y, By =b,,, are defined as in’ Theorem 3.. Remark 5, Note that a, and 6, are bounded, Al rights reserved © Bolyal Institute, University of Saaged 370 , Cohen—K. J. Davis Proof, By Lemma 8. (6.3) Lw)= 2 2 Ke) iS raha with 4B, in the first sum and (¢, ht) =(f, h)= sum of (6,3) be denoted by 2*, h=x. Then B= XO ThGfhe, irae! in the second sum. Let the inner from which, by (6. 1) and by Theorem 3, (m=1, a=k, b=k-+1) with kir>a=0, aL iat Phil HOLE + O(n Mo*(—a, H)) = Iie say beet 025] Substituting this into é 3) one deduces by Lemma 9, Remark 5, and the fact that BrSLn+25 L4(2) = ex 4 O(a Am) 4 ete roa ndicts OCR Sho roan 3 2 = atlhagl “Qi = +O(o* a, Iyeays+), oa, D) jetta ; heles2 cle By Lemma 4c (with k replaced by k+-2) the first two — are O(c?) and by Lemma 5 (restricting « further to 0

    You might also like