You are on page 1of 10

Therefore, for m = 1 we must have k = 2 in (4).

2) is true for p _ 5, but p2 -


a generally.
Therefore for m = 2 one has k = 1. For 3) we will prove that p2ja, but p3 - a
generally.
Let us suppose _rst that m is odd. We will prove that for su_ciently large p,
one has p2ja.
228
If m is even, then for su_ciently large p one has pja. Therefore, one can
say that for m
odd we have k = 2, while for m even, k = 1.
In the proof the method shown in the Niven-Zuckerman book ([1]) will be
used. Let
m = 2n � 1 (n _ 1). Then
a=
((p � 1)!)2n�1
12n�1 +
((p � 1)!)2n�1
22n�1 + : : : +
((p � 1)!)2n�1
(p � 1)2n�1 :
To each member of a corresponds an element ex 2 Zp. In fact
x_
((p � 1)!)2n�1
j2n�1 (mod p);
with 1 _ i _ p � 1 satisfying
xi2n�1 _ �1 (mod p):
Therefore ex = (�ei
2n�1)�1 2 Zp, and since the inverse is unique, �ex is one of the 22n�1
powers ofe1;e2; : : : ;]p � 1. Therefore,
5) a _ �(1 + 22n�1 + : : : + (p � 1)2n�1) (mod p)
But the expression in the right-hand side can be evaluated with the
Bernoulli numbers:
12n�1 + 22n�1 + : : : + (p � 1)2n�1 =
2Xn�1
i=0
1
2n � i
Ci2
n�1Bi;
where B0 = 1, B1 = �
1
2
, B2n+1 = 0 are the Bernoulli numbers. This gives
2n(12n�1 + 22n�1 + : : : + (p � 1)2n�1) =
2Xn�1
i=0
2n
2n � i
Ci2
k�1p2n�iBi:
The right-hand side is divisible by p2 by B2n�1 = 0, and hence p2j2n(12n�1 + :
::+
(p � 1)2n�1). Let now p � 1 _ 2n. Then p2ja, so the assertion is proved. When
m is even,
m = 2n (n _ 1), the proof is very similar, and we omit the details.
Bibliography
1. I. Niven, H.S. Zuckerman, An introduction to the theory of numbers,
Hungarian
translation 1978, Budapest (see pp. 58, 60).
2. G.H. Hardy, E.M. Wright, An introduction to the theory of numbers, 4th
ed., 1960,
Oxford.
229
2 On a non-divisibility property
Let n > 1 be a positive integer. An old problem states that 2n �1 cannot be
divisible
by n. We will prove that 2n�'(n) �1 is not divisible by n, too (where ' is Euler's
totient);
and in fact a more general result will be obtained. Our method is based on
the following
Lemma. Let a > b, (a; b) = 1. Then
(an � bn; am � bm) = a(n;m) � b(n;m); (1)
where (u; v) denotes the g.c.d. of u and v.
For a proof of (1), see [1]; and in more general contexts, see [2] (which
shows that
f(n) = an � bn is a g.c.d. preserving function).
Theorem. Let a > b, (a; b) = 1. Suppose that (a; n) = (b; n) = (a � b; n) =
1. Then
1) n - (an � bn)
2) n - (an�'(n) � bn�'(n)).
Proof. Let us suppose on the contrary that there exists n > 1 such that 1)
holds.
By Euler's divisibility theorem, for (a; n) = 1 one has nj(a'(n) � 1). Now,
since a('(n) �
b'(n) = (a'(n) � 1) � (b'(n) � 1), for (a; n) = (b; n) = 1, n divides a'(n) � b'(n), so
by the Lemma n divides also ad � bd, where d = (n; '(n)). For d = 1 we
would have
nj(a � b), which is impossible by (a � b; n) = 1. Therefore d > 1. On the
other hand,
(a; d) = (b; d) = (a � b; d) = 1 (d being a divisor of n) and d < n by '(n) < n
for
n > 1. Then continuing inde_nitely, one obtains an in_nite sequence (dk) of
positive
integers which is strictly decreasing. This is impossible (i.e. we have
applied the wellknown
Fermat descent method). This proves 1).
Ramarking that (n � '(n); '(n)) = (n; '(n)), by the Lemma
(an�'(n) � bn�'(n); a'(n) � b'(n)) = (an � bn; a'(n) � b'(n))
so 2) follows at once.
Remarks. Let b = 1. Then n - (an � 1) if (a; n) = (a � 1; n) = 1. This is true
for
a = 2. In this case n - (2n�'(n) � 1) is also true.
230
Bibliography
1. J. S_andor, On a divisibility problem (Hungarian), Mat. Lapok 1/1981,
pp.4-5.
2. J. S_andor, A note on certain arithmetic functions, Octogon Math. Mag.
9(2001),
No.2, 793-796.
231
3 On two properties of Euler's totient
Let ' be Euler's totient. The following two properties of ' are well-known:
(1) If (a; n) = 1, then nj[a'(n) � 1] (Euler's theorem); and
(2) If a > 1 then nj'(an � 1)
which as far as we know was _rst discovered by U. Scarpis [1] (see also
Guderson [3],
Rotkiewicz [2]). The aim of this note is the characterization of functions f :
N_ ! N_
such that:
(3) For (a; n) = 1 we have nj[af(n) � 1] and
(4) For all a > 1, njf(an � 1).
We shall prove that (3) and (4) are valid if and only if
(5) '(n)jf(n) for all n.
First, let us suppose that (5) holds. Then since (ak �1)j(am�1) for kjm, from
(1) and
(5) it follows that (3) is true. Now, by (2) and (5) we have nj'(an �1)jf(an �1),
so (4) is
also true. Reciprocally, let us suppose that (3) is true. Let us consider the
multiplicative
group Z_
n of reduced classed (mod n), i.e. ba 2 Z_
n if (a; n) = 1. This is a _nite group
of order '(n). Now it is well-known that in a _nite group if ak = e (e = unity
element),
then k is multiple of order of the group. In our case (3) means that baf(n) =
1 in Z_
n,
therefore '(n)jf(n). By (2) we get mj'(am � 1)jf(am � 1) (here n = am � 1), so
(4) is a
consequence of (3). In fact, (1) is a consequence of the known fact that in
a _nite group G
of order t, one has xt = e (x 2 G). For a proof of (2) let us consider the
group Z_
an�1. Then
(a; an �1) = 1 and an _ 1 (mod an �1), but as 6_ 1 (mod an �1) for s < n.
Therefore
the order of ba is n. This must divide the order of the group, as it is well-
known. (This
follows also from Lagrange's theorem, which says that the order of a
subgroup divides the
order of the group - here one considers cyclic subgroups). Other similar
properties of '
are included in [4].
Bibliography
1. U. Scarpis, Period Mat., 29(1913), p.138.
2. N.G. Guderson, Some theorems of Euler's function, Bull. A.M.S.
49(1943), 278-280.
232
3. A. Rotkiewicz, On the numbers '(an _ bn), Proc. A.M.S. 12(1961), 419-
421.
4. M. Deaconescu, J. S_andor, A divisibility property (Romanian), Gaz. Mat.
XCI(1986), no.2, 48-49.
233
4 On a conjecture of Smarandache on prime
numbers
Let pn denote the n-th prime number. One of Smarandache conjectures in
[3] is the
following inequality:
pn+1=pn _ 5=3; with equality for n = 2: (1)
Clearly, for n = 1, 2, 3, 4 this is true and for n = 2 there is equality. Let n >
4. Then
we prove that (1) holds true with strict inequality. Indeed, by a result of
Dressler, Pigno
and Young (see [1] or [2]) we have
p2
n+1
_ 2p2
n: (2)
Thus pn+1=pn _
p
2 _ 5=3, since 3
p
3 < 5 (i.e. 18 < 25). This _nishes the proof of (1).
Bibliography
1. R.E. Dressler, L. Pigno, R. Young, Sums of squares of primes, Nordisk
Mat. Tidskrift
24(1976), 39.
2. D.S. Mitrinovi_c, J. S_andor (in coop. with B. Crstici), Handbook of
Number Theory,
Kluwer Acad. Publ. 1995.
3. M.L. Perez, ed., Five Smarandache Conjectures on Primes,
http://www.gallup.unm.edu/_smarandache/conjprim.txt.
234
5 On consecutive primes
Let pn denote the nth prime (p1 = 2; p2 = 3; : : :). In paper [2] it is
conjectured that:
pn+1
pn
_
5
3
(1)
with equality only for n = 2. In what follows we shall prove the stronger
relation:
pn+1
pn
_
3
2
; (n 6= 2; 4) (2)
with equality for n = 1. Since
3
2
<
5
3
, clearly (2) improves relation (1) for n 6= 2; 4. We
have
p2
p1
=
3
2
;
p3
p2
=
5
3
>
3
2
;
p4
p3
=
7
5
<
3
2
;
p5
p4
=
11
7
>
3
2
:
Thus (2) is not valid for n = 2; 4.
Let now n > 4. By a result of Dressler, Pigno and Young [1] one has:
p2
n+1
_ 2p2
n: (3)
Thus, by (3) we can write
pn+1
pn
_
p
2<
3
2
for n > 4.
Clearly, relation (1) holds true for all n, with equality only for n = 2.
In paper [2] it is conjectured also that:
1
pn

1
pn+1
_
1
6
: (4)
By (2) one has:
1
pn

1
pn+1
_
3
2
_
1
pn+1

1
pn+1
=
1
2
_
1
pn+1
for n 6= 2; 4:
By
1
pn+1
<
1
3
, this yields relation (4), which holds true also for n = 1; 3 (direct
veri_cation), with equality only for n = 1.
As an application of (2) and (1) note that an improvement of the well-
known relation:
pn + pn+1 _ pn+2 (5)
can be deduced. Indeed,
pn + pn+1 _
2
3
pn+1 + pn+1 =
5
3
pn+1 _ pn+2
235
by (2) and (1). Thus, one has:
pn + pn+1 _
5
3
pn+1 _ pn+1 (n 6= 2; 4) (6)
which sharpens relation (5).
2. A similar relation to (5) and (6) follows by the above proved inequality:
1
pn

1
pn+1
_
1
2
1
pn+1
and the known relation pn+2 < 2pn+1 (which is a consequence of the
Chebyshev Theorem,
that there exists a prime between a and 2a for a _ 2, see [5]).
We can deduce:
1
pn
_
3
2
_
1
pn+1
<
1
pn+1
+
1
pn+2
: (7)
3. The Open Question 171 by Mih_aly Bencze (Octogon Mathematical
Magazine,
vol.6(1998), No.2, pp.219) asks for the inequality
r
pn2+1
pn

p
pn < 1; 34: (8)
In fact, we shall prove that
r
pn2+1
pn

p
pn < 0 for n _ 5.
This is equivalent to:
pn2+1 < p2
n; n _ 5: (9)
We note that by the prime number theorem, pn _ n log n (n ! 1) (see e.g.
[5]) it
follows easily
pn2+1
pn
! 0 as n ! 1, so
pn2+1
pn
< 1 for n _ n0, but this "n0" cannot be
determined by this way. Thus inequality (9) can be much improved for
greater values of
n. However, we shall apply here a general simple method based on the
Rosser-Schoenfeld
inequalities ([3]):
pm < mlogm + mlog logm (m _ 6) (10)
pm > mlogm (m _ 2)
Now, since
log(a + 1) � log a <
1p
a(a + 1)
236
(see e.g. [4]) and
log x _ x � 1 (x > 0);
we have
log(n2 + 1) < log n2 +
1p
n2(n2 + 1)
= 2 log n +
1
n
p
n2 + 1
:
By using the _rst part of (10) with m = n2 + 1 (n _ 3), and the second part
of (10)
with m = n, it is immediate that to prove (9) it is su_cient to deduce an
inequality
4(n2 + 1) log n < n2(log n)2:
Putting n2 = t this becomes equivalent to
log t > 8
_
1+
1
t
_
: (11)
With the increasing function f(t) = log t�8
_
1+
1
t
_
it can be proved that (11) holds
true e.g. for t _ e9. Thus (9) is proved for n _
p
e9 = e4;5; for 5 _ n < e4;5 a direct
computation can be done, and this _nishes the proof of (9).
Bibliography
1. R.E. Dressler, L. Pigno, R. Young, Sums of sequences of primes, Nordisk
Mat.
Tidskrift 24(1976), 39.
2. F. Smarandache, Conjectures which generalize Andrica's conjecture,
Octogon Mathematical
Magazine, Vol.7(1999), No.1, pp.173-176.
3. J.B. Rosser, L. Schoenfeld, Approximate formulae for some functions of
prime numbers,
Illinois J. Math. 6(1962), 64-94.
4. D.S. Mitrinovic, in cooperation with P.M. Vasic, Analytic inequalities,
Springer Verlag,
1970.
5. D.S. Mitrinovic and J. S_andor, in cooperation with B. Crstici, Handbook
of number
theory, Kluwer Acad. Publ., 1995.
237
6 On Bonse-type inequalities
1. Let pn be the nth prime. Euclid's proof of the in_nitude of primes implies
pn+1 _
p1p2 : : : pn � 1 (indeed, the least prime divisor of the right side is distinct
from each of
p1; p2; : : : ; pn, so is at least pn+1).
In what follows, by similar simple considerations we shall deduce some
results, which
particularly contain the Bonse inequalities ([1]). G. P_olya applied the
method for Fermat
numbers Fn = 22n+1 and deduced pn+2 _ 22n + 1 ([2]).
P. Erdos has proved the following (see [3]): If 1 < a1 < a2 < : : : < ak _ x is
a sequence
of integers such that none of them divides the product of others, then k _
_(x), where
_(x) denotes the number of primes _ x.
This implies pk _ ak. Indeed, if ak < pk, then _(ak) _ _(pk) = k and applying
Erdos'
theorem for x = ak one can write k _ _(ak) _ k, i.e. _(ak) = k, impossible.
2. Let us consider now the sequence of general term ak = kp1p2 : : : pn�1 �
pn (n _ 2
_xed). Clearly a1 < a2 < : : : < as with s = pn�1 and Euclid's theorem implies
a1 _ 1. On
the other hand, (ai; aj) = 1 for 1 _ i; j _ s; since

You might also like