Scribd
Upload
20 views2 pages

1961 01erdos

The document proves that for almost all random walks, the number of prime numbers encountered grows proportionally to the logarithm of the total numbers encountered. It establishes this by analyzing the expected value and variance of the number of primes encountered and applying the weak law of large numbers. It also discusses generalizing the result to other sequences beyond primes.

Uploaded by

vahidmesic45
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
20 views2 pages

1961 01erdos

The document proves that for almost all random walks, the number of prime numbers encountered grows proportionally to the logarithm of the total numbers encountered. It establishes this by analyzing the expected value and variance of the number of primes encountered and applying the weak law of large numbers. It also discusses generalizing the result to other sequences beyond primes.

Uploaded by

vahidmesic45
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

A PROBLEM ABOUT PRIME NUMBERS AND THE

RANDOM WALK II

1 am going t.0 prove y = 1. Denot,e by ~(a., 6) the probability of the


random walk passing through a if it st’arts at 6. It is known and easy to
prove that
(1) u(n, 6) - Cl / 6 - a 1-l
(We 4(. IT6 AND 13. 1’. i%IiEAK, JR., P0tentiU~s cxnd the ?Tu&H?Z WC& Illinois
J. Mat,h., vol. 4 (1960), pp. 119-132; a’lso a paper of Murdoch cited therein
where a sharper result’ is obt,ained). In t,he sequel, the letters p and Q de-
not!e primes and u(p, y) = ~(a, 6) in case a. = (p, 0, 0) and 6 = (q, 0, 0).
Consider the number e(n) of points (p, 0, 0) (p 5 n) t,hat t,he pat,h hits.
We have t,o prove tha.t for almost all paths e(~z) T 0 as n T 03.
By (1) and Mertens’ estimate cP5, p-’ - lg, -r~(lgz = lg lg), me evidently
have
(2) E[e(n)l = cppz tr(O, p) - Cl c*<n p-1 - Cl lgz n.

Next, we prove by a customary argument

(3) E[(e(n) - cl Ig, nj’] = O(lgz 4~)‘,


which establishes the weak law of large numbers for e(n), i.e., it shows that)
e(n) = cl lg? n + 0(lg2 n) except for a set of small measure, and this is enough
for our purpose.
Clea’rly by (2)
(41 E[(e - Cl lg, n)?] = R’(2) - &lgz )2)y + “(lg, /1)2.
Furt#her we evident’ly have
me21 = CP~=,x 40, PI + C9<PZllMO, PMP, 4) + 4% YMY, PII
(5) zz act Ctl<p$n [l,‘p(p - 4) + l/p(p - 911 + m2 d2.
Mertens’ estimat’e cited above gives cp<PS,L l/( gp) = 3( lgz YL)~-i- 0( lgz )l.),
and so
(6) Cen l/h - 4) = Cn<m llbj + lLpsn Wdio - d - Lb1
= i(lg2 n>’ + C*<pgz llp(p - (1) + O(k2n).
Thus we have only to e&mate x4cDSn l/p(p - 4).
Received May 19, 1960. See the preceding paper for a st,atement of the problem.
This paper is from a letter sent by I?. Erd6s to the Illinois Journal of Mathematics. II
was edited for publication by H. P. McLean, Jr., who t.hen wrote the preceding paper
which is a treat.ment of the same problem.
352
PRIME SUMBERR ASD THE RANDOM WALK II 353

Put cb = 0 if k is not prime and Q, = c,+ l/(p - q). We have


(ij Cqcps.i, l,lp(p - q) = CL ek/k = CL1 sJk(k + 1) + O(1)
by partial summat,ion (81; = Cf,l Ei). A well-known theorem of Schnirel-
mann stat,es that the number of solutions of p - q = a (p $ k) is less than
cz k(lg k)-‘n,Ia (1 + p-‘) where c2 is an absolute constant’. Thus

(8) Sk < CPk(lg A.>-’ CL1 u-l n[p,a (1 + p-l) < c3 k/g k
since by ink-changing the order of summation we have t’he well-known
CL1 0-l & (1 + p-‘1 = CL d-l zaxmod d)&P
< cd ~L’=l lg k,‘d* < cj lg k.
Thus from (7) and (8)
(9) Cp<pn UP(P - Q) < C6 lgz P-2.
From (O), (6), and (5), we finally obt’ain E(e2) = cf(lg2 n)’ + o(lg, n)’
which proves (3)) and thus the proof of our theorem is complet,e.
By using a sharper estimat,e t,han (1)) it is easy to show t.hat for almost all
paths
(10) lim n+* e(n)/s lg,n = 1.
By t’he same method one can prove that if t#he int,egers q = q1 < q2 < ‘ . .
satisfy

(111 p~-&-l>cilgrt (nSZ), cm= m,


then almost all paths pass through infinitely many points (y, 0, 0). The
primes probably do not satisfy (11) since probably there are an infinite num-
ber of prime twins, but one can prove by Brun’s method that one can select
a subsequence that does satisfy ( 11) .
A~TRAL~N XATIONAL UNIVERSITY
CANBERRA, Xwm4LI.4

You might also like