A new approach to odd perfect numbers via GCDs

Jose Arnaldo Bebita Dris

josearnaldobdris@gmail.com
February 10, 2022
arXiv:submit/4151607 [math.NT] 10 Feb 2022

Abstract
k 2
Let q n be an odd perfect number with special prime q. Define the GCDs
 
G = gcd σ(q k ), σ(n2 )

 
H = gcd n2 , σ(n2 )

and  
2
I = gcd n, σ(n ) .

We prove that G × H = I 2 . (Note that it is trivial to show that G| I and I | 

H both hold.) We then
k k
compute expressions for G, H, and I in terms of σ(q )/2, n, and gcd σ(q )/2, n . Afterwards, we prove
that if G = H = I, then σ(q k )/2 is not squarefree. Other natural and related results are derived further.
Lastly, we conjecture that the set

A = {m : gcd(m, σ(m2 )) = gcd(m2 , σ(m2 ))}

has asymptotic density zero.

1 Introduction
In what follows, we let σ(x) denote the sum of divisors of the positive integer x. We will denote the
deficiency of x by D(x) = 2x − σ(x), the aliquot sum of x by s(x) = σ(x) − x, and the abundancy index of
x by I(x) = σ(x)/x.
A number M satisfying σ(M ) = 2M is called a perfect number. For example, 6 and 28 are perfect since

σ(6) = 1 + 2 + 3 + 6 = 2 · 6

σ(28) = 1 + 2 + 4 + 7 + 14 + 28 = 2 · 28.
The Euclid-Euler Theorem states that M is an even perfect number if and only if

M = (2t − 1)2t−1 ,

where 2t − 1 (and therefore t) is prime. (If t is prime, then 2t − 1 is not necessarily prime.) Primes of the
form 2t − 1 are called Mersenne primes. Currently, there are 51 known Mersenne primes (with the latest
being discovered by the Great Internet Mersenne Prime Search in December of 2018), corresponding to 51
even perfect numbers [5].
It is currently unknown whether there are infinitely many even perfect numbers. It has been conjectured,
and is widely believed, that no odd perfect numbers exist. (There are no odd perfect numbers less than
101500 [8], making the existence of an odd perfect number appear very unlikely.)
Euler proved that an odd perfect number N must necessarily have the so-called Eulerian form

N = q k n2

1
where q is the special prime satisfying q ≡ k ≡ 1 (mod 4) and gcd(q, n) = 1. Descartes, Frenicle, and
subsequently Sorli conjectured that k = νq (N ) = 1 always holds [1]. Sorli predicted that k = 1 after testing
large numbers with eight distinct prime factors for perfection [9]. √
Dris conjectured that q k < n [4], on the basis of the result I(q k ) < 3 2 < I(n).
We state these conjectures here for ease of reference later on.
Conjecture 1.1. If q k n2 is an odd perfect number given in Eulerian form, then k = 1.
Conjecture 1.2. If q k n2 is an odd perfect number given in Eulerian form, then q k < n.
Dris [3] showed that the equation

σ(n2 ) 2n2 D(n2 ) 2s(n2 )

i(q) = k
= k
= k
= = gcd(n2 , σ(n2 ))
q σ(q ) s(q ) D(q k )

holds. We also know that the index i(q) is an integer which is at least 3 by a result of Dris [4]. (The lower
bound on i(q) has since been improved by several authors.)
Furthermore, we can express i(q) as

i(q) = qσ(n2 ) − 2(q − 1)n2 .

Set E = n, F = σ(q k )/2, and K = gcd(E, F ).

In this note, we compute expressions for the following GCDs
 
G = gcd σ(q k ), σ(n2 )

 
2 2
H = gcd n , σ(n )

and  
2
I = gcd n, σ(n ) .

It turns out that it is possible to express all of them in terms of E, F , and gcd(E, F ).
As far as the author is aware, the approach presented in this paper is new and has not been considered
before in the literature.

2 Preliminaries
Define  
G = gcd σ(q k ), σ(n2 )
 
2 2
H = gcd n , σ(n )

and  
I = gcd n, σ(n2 ) .

The following lemma gives an identity that relates the values of G, H, and I.
Lemma 2.1. If N = q k n2 is an odd perfect number given in Eulerian form, then G × H = I 2 .
Proof. We have
2n2
σ(q k ) =
i(q)
and
σ(n2 ) = q k i(q),

2
so that we get
 2 !  !2
2
gcd n , i(q) gcd n, i(q)
2n2 k
 
G = gcd σ(q k ), σ(n2 ) = gcd


, q i(q) = =
i(q) i(q) i(q)
 !2  !2  !2
2 2 2 2 2
gcd n, gcd(n , σ(n )) gcd gcd(n, n ), σ(n ) gcd n, σ(n )
I2
= = = = ,
i(q) i(q) i(q) H
2
since σ(n ) is odd and gcd(q, n) = 1.
Using the identity in Lemma 2.1, we can now derive the following divisibility conditions.
Lemma 2.2. Suppose that N = q k n2 is an odd perfect number given in Eulerian form. Then G divides I
and I divides H.
Proof. The proof of the divisibility constraint I | H follows from the GCD property
   
a | b =⇒ gcd(a, c) | gcd(b, c) .

Afterwards, the proof of the divisibility constraint G | I then follows from Lemma 2.1.
Set
I H
J=
= .
G I
By Lemma 2.1 and Lemma 2.2, J is an (odd) integer.
The following lemma computes the value of J, in terms of E, F, and gcd(E, F ). (The proof is due to the
anonymous MSE user mathlove [6].)
Lemma 2.3. If N = q k n2 is an odd perfect number given in Eulerian form, then we obtain
n
J=  .
gcd σ(q k )/2, n

Proof. We have
n2
H= .
σ(q k )/2
Hence, we obtain

H n2 n2 n2
J= =  =  =  
I
σ(q k )/2 · gcd n, σ(n2 ) gcd n · σ(q k )/2, σ(q k )σ(n2 )/2 gcd n · σ(q k )/2, q k n2

n n
=  =  .
gcd σ(q k )/2, q k n gcd σ(q k )/2, n

Remark 2.1. Notice that we then have

r
I H H
J= = =
G I G
so that
H = G × J 2.
We can now compute expressions for I and G, using Lemma 2.3.

3
Lemma 2.4. If N = q k n2 is an odd perfect number given in Eulerian form, then we obtain
!  
n k
I= · gcd σ(q )/2, n
σ(q k )/2

and
 !2
k
gcd σ(q )/2, n
G= .
σ(q k )/2
Proof. The proof is trivial.

3 What happens when J = 1?

Let us examine the case J = 1 to see whether it is interesting.
First, we prove the following unconditional lemma.
Lemma 3.1. Suppose that N = q k n2 is an odd perfect number given in Eulerian form. Then J = 1 holds if
and only if n | σ(q k )/2.
Proof. Recall that from Lemma 2.3, we have
n
J=  .
k
gcd σ(q )/2, n

This is equal to one if and only if  

gcd σ(q k )/2, n = n,

which holds if and only if n | σ(q k )/2.

Remark 3.1. Note that J = 1 if and only if G = H = I holds.
If J = 1, then from Lemma 3.1, we get n | σ(q k )/2, from which we obtain n < q k . But Brown [2] proved
the estimate q < n in 2016. Hence, J = 1 implies that k > 1. We record this in the succeeding proposition.
Lemma 3.2. Suppose that N = q k n2 is an odd perfect number given in Eulerian form. If J = 1, then both
Conjecture 1.1 and Conjecture 1.2 are false.
Recall from Remark 2.1 that H = G × J 2 . Since H ≥ 3 holds [4], G = J = 1 is not true. We therefore
get the following proposition.
Lemma 3.3. Suppose that N = q k n2 is an odd perfect number given in Eulerian form. If J = 1, then G 6= 1
holds.
Theorem 3.4. Suppose that N = q k n2 is an odd perfect number given in Eulerian form. The following
conditions are equivalent to J = 1:
1. n | σ(q k )/2
2. σ(n2 ) | q k n.
Proof. The proof follows from Lemma 3.1, and by writing the equation

σ(n2 ) qk n
=
n σ(q k )/2
in the form
qk n σ(q k )/2
= .
σ(n2 ) n

4
4 What happens when F is squarefree?
We rewrite the equation
σ(n2 ) n2
k
=
q σ(q k )/2
in the form
σ(n2 ) qk n
=
n σ(q k )/2
to get the succeeding proposition.
The following theorem is similar in spirit to Theorem 3.4.
Theorem 4.1. If N = q k n2 is an odd perfect number given in Eulerian form, then the following conditions
are equivalent:
1. σ(q k )/2 | n
2. n | σ(n2 )
3. G = σ(q k )/2
4. I = n
Proof. The equivalence of the first two conditions follows from the fact that gcd(q k , σ(q k )) = 1.
Next, we show that the third condition is equivalent to the first. Recall from Lemma 2.4 that
 !2
k
gcd σ(q )/2, n
G= .
σ(q k )/2
We then see that G = σ(q k )/2 if and only if σ(q k )/2 | n.
Lastly, we show that the fourth condition is equivalent to the second. To this end, suppose that
n = I = gcd(n, σ(n2 )).
By the definition of GCD, it follows that n | σ(n2 ). Conversely, assume that n | σ(n2 ). Then we obtain
I = gcd(n, σ(n2 )) = n
by the definition of GCD, and we are done.
Under the condition that σ(q k )/2 is squarefree, we get the following result.
Theorem 4.2. Suppose that N = q k n2 is an odd perfect number given in Eulerian form. If F = σ(q k )/2 is
squarefree, then J 6= 1.
Proof. Suppose that σ(q k )/2 is squarefree. (Note that, since σ(q k )/2 | n2 holds in general, then this hy-
pothesis implies that σ(q k )/2 | n is true.) Assume to the contrary that J = 1. By Theorem 3.4, we have
n | σ(q k )/2. Since σ(q k )/2 and n are both positive, we obtain σ(q k )/2 = n. This contradicts a result of
Steuerwald in 1937 [10], who proved that n must contain a square factor.

4.1 What happens when H is squarefree?

Suppose that H is squarefree. By Lemma 2.1, we have the equation
G × H = I 2.
Since this implies H | I 2 , it follows that H | I. However, by Lemma 2.2, we have I | H. Since I and H are
both positive, we obtain I = H. This means that J = 1. By the contrapositive to Theorem 4.2, we finally
have σ(q k )/2 is not squarefree.
We record the immediately preceding results in the following propositions.
Theorem 4.3. Suppose that N = q k n2 is an odd perfect number given in Eulerian form. If H = G × J 2 is
squarefree, then J = 1.
Corollary 4.3.1. Suppose that N = q k n2 is an odd perfect number given in Eulerian form. If H is squarefree,
then F = σ(q k )/2 is not squarefree.

5
5 On the equation H = I
Recall that we have the biconditional
J = 1 ⇐⇒ H = I
from Remark 3.1.
In this section, we shall attempt a naive determination of the asymptotic density of positive integers m
satisfying the equation gcd(m, σ(m2 )) = gcd(m2 , σ(m2 )).
The author tried searching for examples and counterexamples via Sage Cell Server.
All positive integers from 1 to 100 (except for the integer 99) satisfy the equation.
The following integers in the range 1 ≤ m ≤ 1000 do not satisfy gcd(m, σ(m2 )) = gcd(m2 , σ(m2 )).

99 = 32 · 11

154 = 2 · 7 · 11
198 = 2 · 32 · 11
273 = 3 · 7 · 13
322 = 2 · 7 · 23
396 = 22 · 32 · 11
399 = 3 · 7 · 19
462 = 2 · 3 · 7 · 11
469 = 7 · 67
495 = 32 · 5 · 11
518 = 2 · 7 · 37
546 = 2 · 3 · 7 · 13
553 = 7 · 79
620 = 22 · 5 · 31
651 = 3 · 7 · 31
693 = 32 · 7 · 11
741 = 3 · 13 · 19
742 = 2 · 7 · 53
770 = 2 · 5 · 7 · 11
777 = 3 · 7 · 37
792 = 23 · 32 · 11
798 = 2 · 3 · 7 · 19
903 = 3 · 7 · 43
938 = 2 · 7 · 67
966 = 2 · 3 · 7 · 23
990 = 2 · 32 · 5 · 11
A simple inspection yields that primes and prime powers satisfy the equation, so that there are infinitely
many solutions.
The following Pari/GP-routines efficiently determine the numbers and percentages of solutions, up to a
certain search limit. One can easily adjust the range.
c =0; f o r (m=1 ,10 , i f ( gcd (m, sigma (m^2))==gcd (m^2 , sigma (m^ 2 ) ) , c=c + 1 ) ) ; p r i n t ( c , " ␣ ␣ " , ( ( c / 1 0 ) ∗ 1 . 0 ) )

c =0; f o r (m=1 ,100 , i f ( gcd (m, sigma (m^2))==gcd (m^2 , sigma (m^ 2 ) ) , c=c + 1 ) ) ; p r i n t ( c , " ␣ ␣ " , ( ( c / 1 0 0 ) ∗ 1 . 0 ) )

6
 To summarize, we have the table below which shows the counts and percentages of the number of solutions
to the equation
gcd(m, σ(m2 )) = gcd(m2 , σ(m2 )),
up to 10, 102 , 103 , 104 , 105 , and 106 , respectively:
Upper limit Count Percentage
10 10 100%
100 99 99%
1000 974 97.4%
10000 9561 95.61%
100000 93845 93.845%
1000000 923464 92.3464%

The author was only able to test until 106 because the Pari/GP interpreter of Sage Cell Server begins to
crash as soon as a search limit of 107 is specified.
The author thinks this is not a rigorous proof, but it is definitely evidence to suggest that the asymptotic
density in question is less than one.
We state and prove this assertion in the following theorem, which the author first conjectured in the year
2020:
Theorem 5.1. The asymptotic density A of positive integers m with
gcd(m, σ(m2 )) = gcd(m2 , σ(m2 ))
satisfies
A < 1.
Proof. Generalizing the first (counter)example of 99 is trivial.
If 32 · 11 k m, then 11 k gcd(m, σ(m2 )) and 112 k gcd(m2 , σ(m2 )). So the asymptotic density in question
is less than
2 10 3247
1− 3 · 2 = ≈ 0.993878.
3 11 3267
Also, if 3 k m, then with probability 1 there exist two distinct primes y and z congruent to 1 modulo
3 such that y k m and z k m. In this case, we get 3 k gcd(m, σ(m2 )) and 32 k gcd(m2 , σ(m2 )). So the
asymptotic density in question is less than
2 7
1− = ≈ 0.777.
32 9

The real open problem is whether the asymptotic density A is 0. We state this in the succeeding
conjecture:
Conjecture 5.1. The asymptotic density A of positive integers m with
gcd(m, σ(m2 )) = gcd(m2 , σ(m2 ))
satisfies
A = 0.
Remark 5.1. In an answer to one of the author’s questions in MathOverflow, Aaron Meyerowitz [7]
(https: // mathoverflow. net/ users/ 8008 ) made the following assertions regarding Conjecture 5.1: "I
think the density does go to zero, but quite slowly. If p ≡ 1 (mod 6) is prime then there are two solutions
0 < r < s < p − 1 of
u2 + u + 1 ≡ 0 (mod p).
If p k m then, with probability 1, there are two distinct primes u and v, each congruent to r (mod p),
with u k m and v k m. (Either or both could be congruent to s as well.) Then p k gcd(m, σ(m2 )) while
p2 k gcd(m2 , σ(m2 )). So the asymptotic density for this not to happen is 1 − p−1 1
p2 < 1 − p+2 . If we can argue
that the chance that none of these events happen is asymptotically (1 − p−1
Q
p2 ) over the primes congruent to
1 (mod 6), then that asymptotic density is 0."

7
6 Some Further Considerations
6.1 Bounds for K, G, I and J
Recall from Section 1 that we have set E = n and F = σ(q k )/2. In this section, we compute a lower bound
for K = gcd(E, F ).
We begin with the following proposition.
Theorem 6.1. If N = q k n2 is an odd perfect number given in Eulerian form, then K 6= 1.
Proof. Assume to the contrary that K = 1. Then, we can simplify the expression for J (from Lemma 2.3) as
n E
J=  = = E.
K
gcd σ(q k )/2, n

Recall from Remark 2.1 that

H
J2 = .
G
We solve for G and then obtain
H E2 1 1
= · G= = ,
J2 F E2 F
whereupon we get a contradiction from F = σ(q k )/2 ≥ 3 and G ≥ 1.
Remark 6.1. By Theorem 6.1, K ≥ 2 must hold. Since E and F are both odd, then we also obtain K 6= 2.
Consequently, we have the lower bound
K = gcd(E, F ) ≥ 3.
As a corollary, we obtain the following (unconditional) bounds for G, I and J. (We have also used the
definitional property of K = gcd(E, F ), i.e. that K divides both E and F .)
Corollary 6.1.1. If N = q k n2 is an odd perfect number given in Eulerian form, then the following bounds
hold:
9
1. F ≤G≤F
3E
2. F ≤I≤E
E E
3. F ≤J ≤ 3

6.2 On the constraint σ(w 2 ) ≡ 0 (mod w)

Lastly, the author also tried checking for examples of numbers 2 ≤ w ≤ 106 satisfying the divisibility
constraint
w | σ(w2 )
using the following Pari-GP script, via Sage Cell Server:
f o r (w=2 , 1 0 0 0 0 0 0 , i f ( ( Mod( sigma (w^ 2 ) ,w) == 0 ) , p r i n t (w, f a c t o r (w ) ) ) )

Here is the output:

39 = 3 · 13
793 = 13 · 61
2379 = 3 · 13 · 61
7137 = 32 · 13 · 61
13167 = 32 · 7 · 11 · 19
76921 = 13 · 61 · 97
78507 = 32 · 11 · 13 · 61
230763 = 3 · 13 · 61 · 97

8
 238887 = 32 · 11 · 19 · 127
549549 = 32 · 7 · 11 · 13 · 61
692289 = 32 · 13 · 61 · 97
863577 = 32 · 112 · 13 · 61
Note that all of the known examples are odd. The author double-checked the list of the first 199 terms
of OEIS sequence A232354 (https://oeis.org/A232354/b232354.txt) and verified that all of them are
odd. Additionally, all of the terms w in that list do not satisfy σ(w2 )/w = de (where d is prime), except for
w = 39.

7 Future Research
We leave the following problem for other researchers to solve.
Conjecture 7.1. If N = q k n2 is an odd perfect number given in Eulerian form, then unconditionally we
have
gcd(σ(q k ), σ(n2 )) 6= gcd(n2 , σ(n2 )).

8 Acknowledgments
The author thanks the anonymous MSE user Peter (https://math.stackexchange.com/u/82961) for shar-
ing Pari/GP-routines. The author also would like to give credit to the anonymous MSE user mathlove
(https://math.stackexchange.com/users/78967) for providing a proof of Lemma 2.3 [6]. The author is
likewise indebted to Aaron Meyerowitz of MathOverflow [7].

References
[1] Beasley, B. D. (2013) Euler and the ongoing search for odd perfect numbers, Proc. of ACMS 19-th
Biennial Conference Proceedings, Bethel University, 29 May-1 June, 2013, pp. 21–31. Available online:
https://pillars.taylor.edu/acms-2013/11/.
[2] Brown, P. A. (2016) A Partial Proof of a Conjecture of Dris, Preprint,
https://arxiv.org/abs/1602.01591.
[3] Dris, J. A. B. (2017) Conditions equivalent to the Descartes–Frenicle–Sorli Conjecture on odd perfect
numbers, Notes on Number Theory and Discrete Mathematics, 23(2), 12–20.
[4] Dris, J. A. B. (2012) The abundancy index of divisors of odd perfect numbers, Journal of Integer
Sequences, 15, Issue 4, Article 12.4.4.
[5] GIMPS - Various contributors (2018) List of known Mersenne prime numbers, Great Internet Mersenne
Prime Search, https://www.mersenne.org/primes/. Last checked on: February 10, 2022.
[6] mathlove (2022) Answer to "On odd perfect numbers and a GCD - Part VII",
https://math.stackexchange.com/a/4372880/28816. Last updated on: February 3, 2022.
[7] Meyerowitz, A. (2020) Answer to "Is the asymptotic density of positive integers n satisfying
gcd(n, σ(n2 )) = gcd(n2 , σ(n2 )) equal to zero?", https://mathoverflow.net/a/369521/10365. Last
updated on: August 18, 2020.
[8] Ochem, P., & Rao, M. (2012) Odd perfect numbers are greater than 101500 , Math. Comp., 81, 1869–1877.
[9] Sorli, R. M. (2003) Algorithms in the study of multiperfect and odd perfect numbers, Ph. D. Thesis,
University of Technology, Sydney.
[10] Steuerwald, R. (1937) "Verschärfung einer notwendigen Bedingung für die Existenz einer ungeraden
vollkommenen Zahl", S.-B. Math.-Nat. Abt. Bayer. Akad. Wiss, 68–73.