1 Mersenne Primes and Perfect Numbers
Basic idea: try to construct primes of the form an − 1; a, n ≥ 1. e.g.,
21 − 1 = 3 but 24 − 1 = 3 · 5
23 − 1 = 7
25 − 1 = 31
26 − 1 = 63 = 32 · 7
27 − 1 = 127
211 − 1 = 2047 = (23)(89)
213 − 1 = 8191
Lemma: xn − 1 = (x − 1)(xn−1 + xn−2 + · · · + x + 1)
Corollary: (x − 1)|(xn − 1)
So for an − 1 to be prime, we need a = 2.
Moreover, if n = md, we can apply the lemma with x = ad . Then
(ad − 1)|(an − 1)
So we get the following
Lemma If an − 1 is a prime, then a = 2 and n is prime.
Definition: A Mersenne prime is a prime of the form
q = 2p − 1, p prime.
Question: are they infinitely many Mersenne primes?
Best known: The 37th Mersenne prime q is associated to p = 3021377,
and this was done in 1998. One expects that p = 6972593 will give the next
Mersenne prime; this is close to being proved, but not all the details have
been checked.
Definition: A positive integer n is perfect iff it equals the sum of all its
(positive) divisors < n.
Definition: σ(n) = d|n d (divisor function)
So u is perfect if n = σ(u) − n, i.e. if σ(u) = 2n.
Well known example: n = 6 = 1 + 2 + 3
Properties of σ:
1. σ(1) = 1
1
2. n is a prime iff σ(n) = n + 1
pj+1 −1
3. If p is a prime, σ(pj ) = 1 + p + · · · + pj = p−1
4. (Exercise) If (n1 , n2) = 1 then σ(n1 )σ(n2) = σ(n1 n2 ) “multiplicativity”.
Consequently, if
r
e
n= pi j , ej ≥ 1 ∀j, pj prime,
j=1
r r ej +1
e p −1
σ(n) = σ(pj j ) =
j=1 j=1
p−1
6=1+2+3
28=1+2+4+7+14
Examples of perfect numbers:
496
8128
Questions:
1. Are there infinitely many perfect numbers?
2. Is there any odd perfect number?
Note:
6=(2)(3), 28=(4)(7), 496=(16)(31), 8128=(64)(127)
They all look like
2n−1 (2n − 1),
with 2n − 1 prime (i.e., Mersenne).
Theorem (Euler) Let n be a positive, even integer. Then
n is perfect ⇔ n = 2p−1 (2p − 1), for a prime p, with 2p − 1 a prime.
Corollary. There exists a bijection between even perfect numbers and
Mersenne primes.
Proof of Theorem. (⇐) Start with n = 2p−1 q, with q = 2p − 1 a Mersenne
prime. To show: n is perfect, i.e., σ(n) = 2n. Since 2p−1 q, and since
(2p−1 , q) = 1, we have
σ(n) = σ(2p−1 )σ(q) = (2p − 1)(q + 1) = q2p = 2n.
2
(⇒): Let n be a even, perfect number. Since n is even, we can write
n = 2j m, with j ≥ 1, m odd = n
.
⇒ σ(n) = σ(2j )σ(m) = (2j+1 − 1)σ(m)
Since n is perfect,
σ(n) = 2n = 2j+1 m
Get
2j+1 m = (2j+1 − 1) σ(m)
odd
⇒
2j+1 |σ(m);
so
r2j+1 = σ(m) (1)
for some r ≥ 1
Also
2j+1 m = (2j+1 − 1)r2j+1 ,
so
m = (2j+1 − 1)r (2)
Suppose r > 1. Then
m = (2j+1 − 1)r
will have 1, r and m as 3 distinct divisors. (Explanation: by hypothesis,
1 = r. Also, r = m iff j = 0 iff n = m, which will then be odd!)
Hence
σ(m) ≥1 + r + m
=1 + r + (2j+1 − 1)r
=1 + 2j+1 r
=1 + σ(m)
Contradiction!
3
So r = 1, and so (1) and (2) become
σ(m) = 2j+1 (1’)
m = 2j+1 − 1 (2’)
Since n = 2j m, we will be done if we prove that m is a prime. It suffices to
show that σ(m) = m + 1. But this is clear from (1’) and (2’).
Mn = 2n − 1 Mersenne number. Define numbers Sn recursively by setting
2
Sn = Sn−1 − 2, and S1 = 4.
Theorem: (Lucas-Lehmer Primality Test) Suppose for some n ≥ 1 that Mn
divides Sn−1 . Then Mn is prime.
√ √
Proof. (Very clever) Put α = 2 + 3, β = 2 − 3. Note that α + β = 4,
αβ = 1. So S1 = α + β.
n−1 n−1
Lemma. For any n ≥ 1, Sn = α2 + β2 .
Proof of Lemma: n = 1 : S1 = α + β = 4. So let n > 1, and assume that
the lemma holds for n − 1. Since
2
Sn = Sn−1 −2
we get (by induction)
n−1 n−1
Sn = (α2 + β2 )2 − 2
Note:
(αk + β k )2 = α2k + 2αk β k + β 2k
= α2k + β 2k + 2, as αβ = 1.
So we get (setting k = 2n−2 )
n−1 n−1
Sn = α2 + β2 + 2 − 2.
Hence the lemma.
Proof of Theorem (continued): Suppose Mn |Sn−1 . Then we may write
rMn = Sn−1 , some positive integer. By the lemma, we get
n−2 n−2
rMn = α2 + β2 (3)
4
n−2
Multiply (3) by α2 and subtract 1 to get:
n−1 n−2
α2 = rMn α2 −1 (4)
Squaring (4) we get
n n−2
α2 = (rMn α2 − 1)2 (5)
√
Suppose Mn is not a prime. Then ∃ a prime dividing Mn , ≤ Mn . Let
us work in the number system
√
R = {a + b 3|a, b ∈ Z}
Check: R is closed under addition, subtraction, and multiplication (it is
what√one calls a ring). Equations (4) and (5) happen in R. Define R/ =
{a, b 3|a, b ∈ Z/}.
Note: |R/| = 2
We can view α, β as elements of R/. Since |Mn , (4) becomes the fol-
lowing congruence in R/:
n−1
α2 ≡ −1 (mod ) (6)
Similarly, (5) says
n
a2 ≡ 1 (mod )
Put
X = {αj mod |1 ≤ j ≤ 2n }.
Claim |X| = 2n .
Proof of claim. Suppose not. Then ∃j, k between 1 and 2n , with j = k,
such that αj ≡ αk (mod ).
If r denotes |j − k|, then 0 < r < 2n and αr ≡ 1 (mod ). Let d denote
the gcd of r and 2n , so that ar + b2n = d for some a, b ∈ Z. Then we have
n n
αd = αar+b2 = (αr )a · (α2 )b ≡ 1 (mod ).
But since d|2n , d is of the form 2m for some m < n, and αd ≡ 1 (mod )
n−1
contradicts α2 ≡ −1 (mod ). Hence the claim.
So |X| ≤ 2 − 1, i.e., we need 2n ≤ 2 − 1.
Since
≤ Mn , 2 − 1 < Mn = 2n − 1.
⇒ 2n < 2n − 1, a contradiction!
So Mn is prime.
