Sunteți pe pagina 1din 6

Some New Results on Bounds for the Abundancy

Indices of the Components of Odd Perfect Numbers


Jose Arnaldo B. Dris
Graduate, Master of Science in Mathematics
De La Salle University
Manila, Philippines 1004
Email: jabdris@yahoo.com.ph

AbstractIn the M. S. thesis titled Solving the Odd Perfect Number Problem: Some Old and New Approaches and
completed in 2008 [2], the author conjectured that the inequality
q k < n holds for an odd perfect number N given in the Eulerian
form N = q k n2 . In this paper, the authors original approach is
re-set in a more rigorous framework, thereby revealing additional
information about the structure of an odd perfect number.
In particular, we give some sufficient conditions for Sorlis
Conjecture that k = ordq (N ) = 1 to be true.
Index TermsAbundancy Index, Perfect Number, OPN Conjecture, Sorlis Conjecture, Euclid-Euler Heuristics

I. I NTRODUCTION
Perfect numbers are positive integral solutions to the
number-theoretic equation (N ) = 2N , where = 1 is the
sum-of-divisors function. Euclid derived the general form for
the even case; Euler proved that every even perfect number
(EPN) is given in the Euclidean form M = 2p1 (2p 1).
(Numbers of the form 2m 1 are called Mersenne numbers,
and Mersenne primes Mp = 2p 1 give rise to EPNs.)
On the other hand, it is still an open question to determine
existence (or otherwise) for an odd perfect number (OPN).
Euler proved that every OPN is given in the Eulerian form
N = q k n2 where q k 1 (mod 4) and gcd(q, n) = 1.
(Analogously, we call q the Euler prime of the OPN N , and
the component q k will be called the Eulers factor of N .)
As of 2010, only 47 EPNs are known (13 of which were
found by the distributed computing project GIMPS [9]),
while no single example of an OPN has been found. (Pascal
Ochem and Michael Rao of CNRS, France are currently
orchestrating an effort to push the lower bound for an OPN
from the previously known 10300 to a significantly improved
101500 [http://www.lri.fr/ochem/opn/].) Pace Nielsen has
also obtained the lower bound: (N ) 9, for the number of
(N )
distinct prime factors of N ; and the upper bound: N < 24
([9], [10]).
In the Ph. D. dissertation [7] titled Algorithms in
the Study of Multiperfect and Odd Perfect Numbers and
completed in 2003, Ronald Sorli conjectured that the exponent
k = ordq (N ) on the Euler prime q for an odd perfect number
N given in the Eulerian form N = q k n2 is one (i.e. we can
drop k).

Throughout this paper, we will let


(N )

N = q k n2 =

qj j

j=1

denote the canonical factorization of the OPN N = q k n2 . That


is,
min(qj ) = q1 < q2 < q3 < < q(N ) = max(qj )
Note that q 6= q1 because (q + 1)|(q k )|(N ) = 2N and this
forces q+1
2 |N since q is prime and q 1 (mod 4).
II. OPN S C IRCA 2008
The author (in his masters thesis completed in 2008 [2])
made the following conjecture:
Conjecture II.1. Suppose there is an OPN given in the
Eulerian form N = q k n2 where q k 1 (mod 4) and
gcd(q, n) = 1. Then q k < n.
The author formulated Conjecture II.1 on the basis of the
following result:
Lemma II.1. If an OPN is given in the Eulerian form
N = q k n2 and
I(x) =

(x)
x

is the abundancy index of x, then I(q k ) < I(n).


Proof: Since q is the Euler prime and
I(N ) = 2 = I(q k )I(n2 )
we appeal to some quick numerical results. From
I(q k ) <

q
q1

and the fact that q 1 (mod 4), we know that q 5 consequently, we have:
1 < I(q k ) <

5
= 1.25
4

for q an Euler prime. Consequently:

On the other hand,


I(n2 ) =

I(q k ) < I(n2 )

2
I(q k )

a result which was mentioned earlier in Remark II.1.




I(n2 )
Consider the product
I(q k ) q+1
q

so that we get the bounds:


8
1.6 = < I(n2 ) < 2
5

q+1
q

But it is also (fairly) well-known ([4], [5], [6]) that the


abundancy index (as a function) satisfies the inequality:
I(ab) I(a)I(b)
with equality occurring if and only if gcd(a, b) = 1.
In particular, by setting a = b = n, we get:
2
= I(n2 ) < (I(n))2
I(q k )
whereupon we get the lower bound:
s
r

p
8
2
2
<p
=
= I(n2 ) < I(n)
k
5
I(q )
I(q k )
q
Per WolframAlpha, 85 1.264911.
Remark II.1. When Conjecture II.1 was formulated in 2008,
the author was under the naive impression that the divisibility
constraint gcd(q, n) = 1 induced an ordering property
for
q
N
k
the Euler prime-power q and the component n = qk , in the
sense that the related inequality q k < n2 followed from the
result I(q k ) < I(n2 ). (Indeed, the author was able to derive
the [slightly] stronger result q k < (q k ) 23 n2 [2]).
The author also obtained the following results in the same
year (2008):
Lemma II.2. Let N be an OPN given in the Eulerian form
N = q k n2 , with q k 1 (mod 4) and gcd(q, n) = 1.
2
4q+2
and
Then we have the following bounds L(q) = 3qq(q1)
U (q) =

3q 2 +2q+1
q(q+1)

A quick double-check using WolframAlpha gives you that,


indeed, the lower bound L(q) is less than the upper bound
U (q), if q is an Euler prime.
Remark II.2. Notice that, from the proof of Lemma II.2, we
have:
q
2(q 1)
<
q1
q

2
q
which implies that q1
< 2. Thus:

2
2(q 1)
q
< 2= <
< I(n2 ) < 2
q1
q
2
Also, observe from Lemma II.1 that:
s
r

5
8
2
2
I(q k ) < <
<p
=
k
4
5
I(q k )
I(q )
p

which implies that I(q k ) I(q k ) < 2, from which we get:


1 < I(q k ) <

I(q k )

 32

<

which then gives:


I(q k ) <

with L(q) < I(q ) + I(n ) U (q).


Per WolframAlpha,

2 1.259921.

2(q 1)
2
2q
< I(n2 ) =

k
q
I(q )
q+1

Corollary II.1. Let N be an OPN given in the Eulerian form


N = q k n2 , with q k 1 (mod 4) and gcd(q, n) = 1. Then
we have the following (explicit) numerical bounds:
57
2.85 =
< I(q k ) + I(n2 ) < 3
20
with the further result that they are best-possible.

2(q 1)
q
<
q1
q

Proof: This Corollary can be proved using Lemma II.2


and basic differential calculus, and is left as an exercise to the
interested reader.

we get:

Notice that:

This product is nonnegative since


I(q ) < I(n ).
Expanding the product and simplifying using the equation
2
+2q+1
I(q k )I(n2 ) = 2, we get the upper bound U (q) = 3qq(q+1)
k
2
for the sum I(q ) + I(n ).



q
q
I(n2 ) q1
.
Next, consider the product I(q k ) q1
q
k
2
This product is negative since I(q ) < q1 < I(n ). Again,
expanding the product and simplifying using the equation
2
4q+2
I(q k )I(n2 ) = 2, we get the lower bound L(q) = 3qq(q1)
for the same sum I(q k ) + I(n2 ).

Proof: Starting from the (trivial) inequalities:


q+1
q
I(q k ) <
q
q1

q+1
q
2

Remark II.3. As remarked by Joshua Zelinsky a few years


back: Any improvement on the upper bound of 3 would
have [similar] implications for all arbitrarily large primes and
thus would be a very major result. (e.g. U (q) < 2.99 implies
q 97.) Thus, (assuming Sorlis conjecture that k = 1)
heuristically we can have the following approach for the
OPN problem: Fix an upper bound u = U (q) for
I(q k ) + I(n2 ) = I(q) + I(n2 ) = U (q)

and then use factor/sigma chains to loop until you get the
contradiction U 1 (u) 6= q.
Remark II.4. Note that, from Lemma II.2,
L(q) =

q2
3q 2 4q + 2
=3
q(q 1)
q(q 1)

U (q) =

3q 2 + 2q + 1
q1
=3
q(q + 1)
q(q + 1)

and

Observe that, when L(x) and U (x) are viewed as functions


on the domain D = R \ {1, 0, 1}, then L(x + 1) = U (x)
and U (2) = U (3) = L(3) = 17
6 < 2.84.

(E-2) The Mersenne primes Mp


satisfy Mp 3 (mod 4).
(Known)
(E-3) The exponent
s = ordMp (M ) is one.
(Known)
(E-4) If M 0 (mod 2)
is perfect, then given the
Euclidean form
2
Y
M = 2p1 (2p 1) =
pi i ,

(O-2) The Euler primes q


satisfy q 1 (mod 4).
(Known)
(O-3) The exponent
k = ordq (N ) is one.
(Sorlis Conjecture, 2003 [7])
(O-4) If N 1 (mod 2)
is perfect, then given the
Eulerian form
(N )
Y
N = q k n2 =
qj j ,
j=1

i=1

pi i (pi i )
then
= i,
M
for i = 1, 2.
(Observation, Dris 2011)
(E-5) There are infinitely
many EPNs.
(EPN Conjecture)
(E-6) The density of EPNs
is zero. (Kanold)
(E-7) 1 < I(Mp ) 87
for p 3.
Mp +1
2
7
4 I(Mp ) = I( 2 ) < 2
In particular, q
q
Mp +1
8
7
<
7
4 < I(
2 ) < 2.

qj j (qj j )
2
then
< j,
N
3
for all j, 1 j (N ).
(Theorem, Dris 2008 [2])
(O-5) There do not exist
any OPNs.
(OPN Conjecture)
(O-6) The density of OPNs
is zero. (Kanold)
(O-7) 1 < I(q k ) < 45
for q 5.
2
8
2
5 < I(q k ) = I(n ) < 2
In particular,
q
5
8
<
4
5 < I(n) < 2.

III. E UCLID -E ULER H EURISTICS FOR P ERFECT N UMBERS


So far, we have focused our discussions on a hypothetical
OPN and its Euler prime. We now turn our attention to a
re-evaluation of the theory of OPNs in light of the empirical
(and more concrete) evidence for EPNs.
To this end, we have what we call the Euclid-Euler heuristics for EPNs and OPNs presented through a tabular side-byside comparison:
(E-1) (Euclid-Euler Theorem)
The Mersenne primes Mp
are in one-to-one
correspondence with
the EPNs.

(O-1) (Conjecture, 2010)


The Euler primes q
are in one-to-one
correspondence with
the OPNs.

Euler proved that a hypothetical OPN must have a unique


Euler prime. However, it is currently unknown whether two
distinct hypothetical OPNs can share the same Euler prime.
This is because:

In the most general case, it is not known whether squares


are friendly or solitary.
It is not even known if the smallest possible Euler prime
5 does divide an odd perfect number.
It is conceivable that two distinct numbers P and Q
which satisfy a particular number-theoretic equation do
not necessarily have to satisfy gcd(P, Q) = 1.

Remark III.1. We excluded p1 = 2 from (E-7) because


M p1 + 1
= 2p1 1 = 221 = 2 is squarefree.
2
From (E-4) and (O-4), we have the following result:
(K)

Theorem III.1. If a number K =

rl l has a component

l=1

C(r, a) = ra a satisfying one of the following conditions for


some a, 1 a (K):
2
Condition L[eft] ::
< i(C(r, a)) < 1
3
Condition M[iddle] :: 1 < i(C(r, a)) < 2
Condition R[ight] :: 2 < i(C(r, a))
ra a (ra a )
, then K is not perfect.
where i(C(r, a)) =
K
Proof: Take the contrapositives of (E-4) and (O-4), and
then combine the two results.
IV. S ORLI S C ONJECTURE [2003]
We now state Sorlis Conjecture on OPNs:
Conjecture IV.1. If N is an odd perfect number with Euler
prime q then q||N .
Remark IV.1. In other words, if the OPN N is given in the
Eulerian form N = q k n2 , then Sorlis Conjecture predicts

that k = ordq (N ) = 1. Note that, trivially we have

q<

V. S OME N EW R ESULTS ON OPN S


A. Preliminaries

N = q k/2 n

which gives q (2k)/2 < n. Furthermore, the inequality


(q k ) 23 n2 in page 111 of
the authors M. S. thesis

6 k/2
(see Theorem 4.2.5) gives
us 2 q
< n. Together, these

two inequalities imply 26 q < n2 . WolframAlpha gives the


approximation 26 1.22474487.

We now assign the values of the following fractions to the


indicated variables, for ease of use later on:
(q k )
= I(q k )
qk
(q k )
qk
2 =
=
I(q k )
n2
n2
(q k )
qk
3 =
= I(q k )
n
n
(n2 )
1 =
= I(n2 )
n2
n2
(n2 )
=
I(n2 )
2 =
qk
qk
(n)
3 =
= I(n)
n
(n)
n
4 = k = k I(n)
q
q
1 =

We give here three sufficient conditions for Sorlis Conjecture to be true. (In that direction, recall that the components q k
and n2 of the OPN N = q k n2 are related via the inequality
q k < n2 , as mentioned in Remark II.1. Also, observe that
gcd(q, n) = 1 forces q 6= n and q k 6= n.)
Theorem IV.1. Let N be an OPN given in the Eulerian form
N = q k n2 , with q k 1 (mod 4) and gcd(q, n) = 1.

If q k < n and n < q 2 , then k = 1.


If n < q, then k = 1.
Proof: The first implication follows from:
q qk < n < q2

B. Initial Lemmas
Lemma V.1. 1 < 1 < 1.25
1.6 < 1 < 2 < 2.85 < 1 + 1 < 3

and which subsequently gives:


1k<2
whence k = 1 because k 1 (mod 4). For the second
implication, we proceed as follows: First, we show that n < q
implies that n < q k . Trivially, q q k since k 1. Thus,
n < q q k . Still under the assumption n < q, we get:
n < q q k < n2 < q 2
which gives:
1k<2

Proof: Refer to Lemmas II.1 and II.2, and Corollary II.1.


q
Lemma V.2. 1.2649 < 85 < 3 < 1 3 < 2
q
2.2649 < 1 + 85 < 1 + 3 < 3
Proof: Refer to the proofs for Lemmas II.1 and II.2. (Note
that q k n is deficient.)

2 4 1000
< 3 + 4
Lemma V.3. 2.2493653 <
5
Proof: Use the Arithmetic Mean-Geometric Mean Inequality.

whence [again] k = 1 because k 1 (mod 4).

Lemma V.4. 3 6= 4

Corollary IV.1. Let N be an OPN given in the Eulerian form


N = q k n2 , with q k 1 (mod 4) and gcd(q, n) = 1. If
n < q 2 , then k = 1.

(q k )
(n)
=
. Thus,
n
qk
q k (q k ) = n(n). But gcd(q, n) = 1. Consequently, q k |(n)
and n|(q k ) by Euclids Lemma.
q Therefore, 3 , 4 N. This
contradicts Lemma V.2: 1 < 85 < 1 3 = 3 4 < 2.

Proof: We know that q k < (q k ) 23 n2 [2]. In particular,


q < n2 . Consequently, if n < q 2 then q q k < n2 < q 4 .
This implies 1 k < 4, whereupon we get k = 1 since k 1
(mod 4).
On the other hand:
k

Corollary IV.2. Let N be an OPN given in the Eulerian form


N = q k n2 , with q k 1 (mod 4) and gcd(q, n) = 1. If
q < n and n < q k , then k 5.
k

Proof: Since q < n < q implies that 1 < k, k 1


(mod 4) then gives k 5.

Proof: Suppose 3 = 4 . Then

Lemma V.5.

11
3

2 + 2

Proof: Consider the product (2 3)(2 3) in light of


the relations 2 2 = 2 and 2 32 < 3 2 . (The proof is
very similar to that of Lemma II.2.)
Observe that:
1
2
(q k )
1 + 2
5
=
=
=
<
2
2
1
(n )
1 + 2
12

1
3
(q k )
1 + 3
qk
3
3
=
=
=
<
=
=
4
3
(n)
3 + 4
n
1
4
qk
I(q k )  q k 2 I(q k )  q k 2
3
1
= nn
=
=
4
n
I(n)
n

3
I(n)
qk
Furthermore, we have:

1
I(q k )
1
5 10
8 10
3
I(n)
<
=
<
<
<
=
<2
2
I(n)
3
16
25
1
I(q k )
   
2  k 2
1
3
(q k )
q
=
<
4
3
(n)
n

Theorem V.1. Let N = q k n2 be an OPN with Eulers factor


(q k )
(n)
q k . Then
, and the following
= 3 6= 4 =
n
qk
statements hold:

qk
Case 1: If 3 < 4 , then
< 2 < 1.4142136.
n

2 4 40
qk
Case 2: If 4 < 3 , then 1.0059467 <
< .
5
n
Finally, observe that 3 6= 1, because otherwise (q k ) = n
which contradicts (q k ) k + 1 2 (mod 4).
We consider two cases:
Case A: 3 < 1


These results follow directly from Lemmas V.1 and V.2.


We now consider two cases, in view of Lemma V.4:
Case 1: 3 < 4
Case 2: 4 < 3
C. Final Theorems
Case 1: 3 < 4 This means that:
 k 2
I(q k )
3
q
=
<1
n
I(n)
4

qk
n

2

3 3
3
=
<
=
1 4
1 4

1
1

3
=
4

1
1



qk
n

qk
1
<
< 1 by Lemma V.1.
n
1
q
q
Furthermore, since 85 < 3 4 , then 3 < 1 < 85 < 4 .

qk
Case 1 of Theorem V.1 then gives
< 2.
n
Thus,

Observe that Case 2 of Theorem V.1 cannot hold under


Case A. That is, we have either 1 < 3 or 1 < 4 .
Case B: 1 < 3

Consequently, we have:
 k 2
q
I(n)
<2
<
n
I(q k )
We therefore get the explicit numerical bound:

qk
< 2 < 1.4142136
n
Case 2: 4 < 3 This means that:
 k 2
3
I(q k )
q
1<
=
4
n
I(n)
Thus, we have:

 k 2
I(n)
q
8 10
<
<
25
I(q k )
n
We now get the absolute bound:
s

8 10
qk
2 4 40
=
<
5
25
n
WolframAlpha gives the rational approximation:

2 4 40
1.005946744
5
These findings are summarized in the following theorem:

qk
n

Thus,

2

3
3 3
>
=
=
1 4
1 4

1
1

3
=
4

1
1



qk
n

1
qk
4
>
> by Lemma V.1.
n
1
5

Furthermore, since 4 < 3 4 < 2, then we have:


Under Case 1 of Theorem V.1, we get
1 < 3 < 4 < 2.
This then gives the upper bound 3 < 2 and the lower

4
qk
4
<
< 2.
bound 1000
<

.
Hence,
4
5
5
n
Under Case
2 of Theorem V.1, we obtain the improved
2 4 40
qk
lower bound
< .
5
n
All of the preceding calculations are summarized in the
following theorem:
Theorem V.2. Let N = q k n2 be an OPN with Eulers factor
q k . Then we have:
q
qk
8
Case A: If 3 < 1, then
5 < 4 and n < 1.
Case B: If 1 < 3 , then we have:

Case
B.1: If 3 < 4 , then 1 < 3 <
2 and

1000
qk
4
< 4 < 2. Also, 5 <
< 2.
5
n

4
qk
2 40
Case B.2: If 4 < 3 , then
< .
5
n

ACKNOWLEDGMENT

We now predict that:


Conjecture V.1. If there is an OPN given in the Eulerian form
N = q k n2 where q k 1 (mod 4) and gcd(q, n) = 1, then
q k (q k ) < n(n).
Our previous considerations are encapsulated in the
following theorem:
Theorem V.3. Suppose there is an OPN given in the Eulerian
form
(N )

N = q k n2 =

qj j

j=1

where
qk1

(mod 4)

and gcd(q, n) = 1. Then:

Conjecture V.1 implies Conjecture II.1.


Conjecture II.1 also implies Conjecture V.1.
If Conjecture II.1 is false, then Conjecture IV.1 is still true
provided that:
n < q (i.e. q = max(qj ), 1 < j (N )) (Strong
Form).
n < q 2 (Weak Form).

Proof: We prove only the second statement, as the first


statement is a direct consequence of Lemma II.1, while the
third statement follows from Corollary IV.1 and Corollary IV.2.
Suppose Conjecture II.1 is true and Conjecture V.1 is false.
Then we have 4 < 3 and q k < n. This
Case 2
contradicts
qk
2 4 40
< .
of Theorem V.1: 4 < 3 implies 1 <
5
n
VI. C ONCLUSION
With a proof for Sorlis Conjecture in sight [as stated in
Conjecture IV.1 and for which sufficient conditions were
given in Theorem IV.1 and Corollary IV.1], every number
theorist will then have an easier time with his/her OPN
computations because k can be dropped and 1 = I(pk )
p+1
collapses to I(p) =
. In this paper, we have seen
p
more compelling evidence, not only for Sorlis Conjecture
and the conjectured nonexistence of OPNs, but also for the
Euclid-Euler heuristics from (E-1)/(O-1) to (E-7)/(O-7) in
page 3.
In a forthcoming paper, the author will present a new
approach to settling the conjectures in the Euclid-Euler
heuristics (E-5) and (O-5), using modern mathematical
machinery such as algebraic number theory and elliptic
curves.

The author would like to thank Severino Gervacio, Fidel


Nemenzo, Jose Balmaceda, Julius Basilla, Blessilda Raposa,
Carl Pomerance, Richard Brent and Gerry Myerson for sharing
their expertise. Lastly, the author wishes to express his deepest
gratitude to Joshua Zelinsky from whom he derived the
inspiration to pursue OPNs as a research topic.
R EFERENCES
[1] J. A. B. Dris, On the Components of an Odd Perfect Number,
Electronic Proceedings of the 9th Science and Technology Congress,
De La Salle University, Manila, Philippines, July 4, 2007.
[2] J. A. B. Dris, Solving the Odd Perfect Number Problem: Some Old and
New Approaches, M. S. Math thesis, De La Salle University, Manila,
Philippines, Aug. 2008. Available online: http://www.scribd.com/doc/
16144034/OPNThesis1.
[3] J. A. B. Dris, Solving the Odd Perfect Number Problem: Some
New Approaches, Electronic Proceedings of the 11th Science and
Technology Congress, De La Salle University, Manila, Philippines,
Sept. 22, 2009. Available online: http://www.scribd.com/doc/19690412/
OPNPaper1.
[4] M. B. Nathanson, Elementary Methods in Number Theory, Graduate
Texts in Mathematics (195), Springer-Verlag, (2000).
[5] J. Sandor and B. Crstici (eds.), Perfect Numbers: Old and New Issues Perspectives, (2004), Handbook of Number Theory vol II, (pp 15-98),
Dordrecht, The Netherlands: Kluwer Academic Publishers, Retrieved
June 10, 2007, from Gale Virtual Reference Library via Thomson Gale.
[6] J. Sandor, D. Mitrinovic, and B. Crstici (eds.), Sum-of-Divisors
Function, Generalizations, Analogues - Perfect Numbers and Related
Problems, (2006), Handbook of Number Theory vol I, (2nd ed, pp 77120), Dordrecht, The Netherlands: Springer, Retrieved June 10, 2007,
from Gale Virtual Reference Library via Thomson Gale.
[7] R. M. Sorli, Algorithms in the Study of Multiperfect and Odd Perfect
Numbers, Ph. D. Thesis, University of Technology, Sydney, (2003).
[8] Various contributors, Great Internet Mersenne Prime Search, Available online: http://www.mersenne.org/prime.htm, Viewed: Jan. 2011.
[9] Various contributors, Perfect Number, From Wikipedia - the free
encyclopedia, Available online: http://en.wikipedia.org/wiki/Perfect
number, Viewed: Jan. 2011.
[10] E. W. Weisstein, Perfect Number. From MathWorldA Wolfram Web Resource. Available online: http://mathworld.wolfram.com/
PerfectNumber.html, Viewed: Jan. 2011.

S-ar putea să vă placă și