A Solution of Sierpinski Problem Based m

Chi Zhang
arXiv:2106.07376v1 [math.NT] 8 Jun 2021

Academy of Mathematics and Systems Science, University of Chinese Academy of

Sciences, Beijing 100049, People’s Republic of China

zhangchi171@mails.ucas.ac.cn

Abstract
In 1960, W. Sierpinski proved that there are infinitely many positive odd numbers
k, such that for any positive integer n, k × 2n + 1 is a composite number. Such
numbers are called “Sierpinski numbers”. In this study, by using covering systems
and the theory of cyclotomic polynomials, the following theorem is proved: for any
integer m > 1, there are infinitely many integers k satisfying k 6≡ −1 (mod q) for
any prime number q|(m − 1), such that for any positive integer n, kmn + 1 is a
composite number. These positive integers k are called “Sierpinski numbers based
m”. The theorem can be regarded as a generalization of Sierpinski problem.
Key words: Sierpinski number, Covering system, Cyclotomic polynomial.

1. Introduction
In 1934, P. Erdös[1] raised the concept of covering systems when he proved that
any integer congruent to 2036812 modulo 5592405 and 3 modulo 62 can not be the
sum of a power of two and a prime. The definition of covering systems is given as
follows.

Definition 1. If every integer satisfies at least one congruence in

x ≡ a1 (mod n1 ), x ≡ a2 (mod n2 ) ,..., x ≡ ak (mod nk ). (1)

Then (1) is called “covering system” and it can be denoted as

A = {as (ns )}ks=1 = {a1 (n1 ), a2 (n2 ), . . . , ak (nk )}.

Some problems on number theory can be solved by using covering systems. In

1956, H. Riesel[2] proved that there are infinitely many positive odd numbers k,
such that for any positive integer n, k × 2n − 1 is a composite number. Then, In
1960, W. Sierpinski[3] proved that there are infinitely many positive odd numbers k,
such that for any positive integer n, k ×2n +1 is a composite number. Such numbers
are called “Sierpinski numbers”. For example, 78557 is a Sierpinski number and it
 2

may be the minimum Sierpinski number. 78557 × 2n + 1 is always divisible by one

of 3,5,7,13,19,37,73.
Inspired by G. Xungui’s study[4], this study tries to generalize Sierpinski prob-
lem. Similar conclusion still holds if the number 2 is replaced by any other integer
m > 2. Notice that for any prime number q|(m − 1), if we take integer k such that
k ≡ −1 (mod q), then kmn + 1 ≡ 0 (mod q) for any n ∈ N. These conditions are
trivial. Excluding the trivial conditions, we proved the following theorem.

Theorem 1. For any integer m > 1, there are infinitely many integers k satisfying
k 6≡ −1 (mod q) for any prime number q|(m − 1), such that for any positive integer
n, kmn + 1 is a composite number.

This theorem may be known since there is a website[5] which focuses on the
problem of minimum Sierpinski number based m.

2. Proof of Theorem 1
First, we introduce two covering systems that we will use in the proof of Theorem
1.

Lemma 1. {0(2), 0(3), 1(4), 5(6), 7(12)} and {2(4), 4(8), 8(16), 8(24), 0(48), 1(3),
5(6), 3(12), 1(5), 7(10), 3(15), 9(20), 15(30)} are both covering systems.

Proof. It is obvious that {0(2),0(3),1(4),5(6),7(12)} is a covering system by verifying

one by one from 1 to 12.
0(16) can be split into 16(48), 32(48), 0(48). They are covered by 1(3), 8(24),
0(48) respectively. So, 0(16) is covered by 1(3), 8(24), 0(48). Therefore, 0(2) is
covered by 2(4), 4(8), 8(16), 1(3), 8(24), 0(48).
9(12) can be split into 21(60), 57(60), 33(60), 9(60), 45(60). They are covered
by 1(5), 7(10), 3(15), 9(20), 15(30) respectively. So, 9(12) is covered by 1(5), 7(10),
3(15), 9(20), 15(30). Therefore, 1(2) is covered by 1(3), 5(6), 3(12), 1(5), 7(10),
3(15), 9(20), 15(30).
Hence {(2(4),4(8),8(16),8(24),0(48),1(3),5(6),3(12),1(5),7(10),3(15),9(20),15(30)}
is a covering system.

Next, we introduce some lemmas about cyclotomic polynomials.

Lemma 2. Let Φn (x) be the nth cyclotomic polynomial. For any prime number p,
k k−1 k
if (p, n) = 1, then Φn (xp ) = Φn (xp )Φpk n (x); if p|n, then Φpk n (x) = Φn (xp ).
k
Proof. For any primitive pk nth root of unity ω, ω p is a primitive nth root of unity.
k
Thus all roots of Φpk n (x) are roots of Φn (xp ).
 3

If p|n, then
   k

deg Φpk n (x) = ϕ(pk n) = pk ϕ(n) = deg Φn (xp ) .

k k
Since both Φpk n (x) and Φn (xp ) are monic polynomials, we have Φpk n (x) = Φn (xp ).
If (p, n) = 1, then for any primitive nth root of unity ω1 , ω1p is also a primitive
k−1 k
n root of unity. Thus all roots of Φn (xp )Φpk n (x) are roots of Φn (xp ). Since
th

 k−1

deg Φn (xp )Φpk n (x) = pk−1 ϕ(n) + ϕ(pk n)
= pk−1 ϕ(n) + (pk − pk−1 )ϕ(n)
 k

= pk ϕ(n) = deg Φn (xp )

k−1 k
and both Φn (xp )Φpk n (x) and Φn (xp ) are monic polynomials. We have
k k−1
Φn (xp ) = Φn (xp )Φpk n (x).

Lemma 3. Suppose n ∈ N, d is a true factor
of n and p is a prime. If there is an

integer x 6= ±1 such that p gcd Φd (x), Φn (x) , then p nd .
xn −1 n
Proof. Since d is a true factor of n, so Φn (x) xd −1
. Therefore, p gcd(xd − 1, xxd −1
−1
).
Let X = xd − 1, then
n
xn − 1 (X + 1) d − 1 n
d
= = AX + , (A ∈ Z).
x −1 X d

Therefore, p gcd(X, AX + nd ) ⇒ p gcd(X, nd ) ⇒ p n

d.

Lemma 4. If there is an integer x 6= ±1 such that gcd Φa (x), Φb (x) > 1, then ab

is a prime power.


Proof. Suppose p is a prime and p gcd Φa (x), Φb (x) . Let a = pα m, b = pβ n,
gcd(p, mn) = 1. We just need to prove that m = n.
First, we claim that p|Φm (x). When α = 0, the conclusion is obvious. When
α ≥ 1, according to lemma 2,
α α−1
Φm (x) ≡ Φm (xp ) ≡ Φm (xp )Φa (x) ≡ 0 (mod p).

Thus p|xm − 1. Similarly, p|xn − 1. Let d = gcd(m, n). Then p|xd − 1. If m 6= n,

suppose m > n. Then d < m. So, there is a factor of d, denoted as d1 , such that
p|Φd1 (x). Since d1 is true factor of m, according to lemma 3, we have p|m, which
is a contradiction. Therefore, m = n and ab = pα−β is a prime power.

Lemma 5. Suppose integer n ≥ 3 is a power of a prime number q. For any integer

x ≥ 2, if Φn (x) > q, then Φn (x) has a prime factor p 6= q.
 4

Proof. Let n = q α . When α = 1, we have q 6= 2 and Φn (x) = xq−1 + · · · + x + 1.

Since q|xq−1 + · · · + x + 1 ⇔ x ≡ 1 (mod q), if x 6≡ 1 (mod q), then Φn (x) has a
prime factor p 6= q. If x ≡ 1 (mod q), suppose x = qr + 1, then

q−1 q(q − 1)
xq−1 + · · · + x + 1 ≡ (qr + 1) + · · · + qr + 1 + 1 ≡ qr + q ≡ q (mod q 2 ).
2
Since Φn (x) > q, Φn (x) has a prime factor p 6= q.
α−1 α−1
When α ≥ 2, according to lemma 2, Φn (x) = Φq (xq ). Since q|Φq (xq ) ⇔
q|Φq (x) ⇔ x ≡ 1 (mod q), when x 6≡ 1 (mod q), Φn (x) has a prime factor p 6= q.
Now suppose that x ≡ 1 (mod q).
α−1
If q 6= 2, then Φq (xq ) ≡ q (mod q 2 ). Φn (x) has a prime factor p 6= q.
α−1 α−1
If q = 2, then Φn (x) = Φq (xq ) = x2 + 1 ≡ 2 (mod 4). So Φn (x) has a
prime factor p 6= 2.

Lemma 6. Suppose n = pα q β , α, β ≥ 1. For any integer x ≥ 2, if Φn (x) > pq,

then Φn (x) has a prime factor r 6= p, q.

Proof. According to lemma 2,

α α β−1
Φqβ (xp ) Φq (xp q )
Φn (x) = α−1 = .
Φqβ (xp ) Φqβ (xpα−1 )
α β−1
The proof of lemma 5 implies that q 2 ∤ Φq (xp q ) . Hence q 2 ∤ Φn (x) . Similarly,
p2 ∤ Φn (x). Since Φn (x) > pq, Φn (x) has a prime factor r 6= p, q.

Lemma 7. Suppose m ∈ Z, A = {as (ns )}ts=1 is a covering system, p1 , p2 , . . . , pt

are distinct prime factors of mn1 − 1, mn2 − 1, . . . , mnt − 1 respectively. Then there
are infinitely many integers k, such that kmn + 1 is a composite number for any
n ∈ N.

Proof. According to Chinese remainder theorem, there are infinitely many integers
k > p1 , p2 , . . . , pt , satisfying the following congruence equations:


 kma1 + 1 ≡ 0 (mod p1 )
kma2 + 1 ≡ 0 (mod p2 )


 ···
kmat + 1 ≡ 0 (mod pt ).


For any n ∈ N, there is an integer i (1 ≤ i ≤ t), such that n ≡ ai (mod ni ). Since

pi |mni − 1, we have mni ≡ 1 (mod pi ). Thus kmn + 1 ≡ kmai + 1 ≡ 0 (mod pi ),
which implies that kmn + 1 is a composite number.

Proof of Theorem 1. m = 2 is the condition which Sierpinski proved. We assume

m ≥ 3 in the rest of the proof.
 5

When m 6= 2l − 1, l ∈ N, we can use the covering system

{ai (ni )}5i=1 = {0(2), 0(3), 1(4), 5(6), 7(12)}

to complete the proof. Snice m 6= 2l − 1, so Φ2 (m) = m + 1 has an odd prime factor

p1 . Thus gcd(m − 1, p1 ) = 1. Since m ≥ 3, according to lemma 5 and lemma 6, we
can take prime factors p2 , p3 , p4 , p5 from Φ3 (m), Φ4 (m), Φ6 (m), Φ12 (m) respectively
such that p2 6= 3, p3 6= 2, p4 6= 2, 3, p5 6= 2, 3. Since Φ1 (m) = m − 1, according to
lemma 3 and lemma 4, it follows that

gcd(m − 1, p2 ) = gcd(m − 1, p3 ) = gcd(m − 1, p4 ) = gcd(m − 1, p5 ) = 1,

which implies that p1 , p2 , p3 , p4 , p5 are not the factors of m − 1.

According to lemma 3, if gcd Φa (x), Φb (x) > 1, (a ≤ b), then its prime factor
p ab . Since we always take a prime factor pi from Φni (m) such that gcd(pi , ni ) = 1
for every i ∈ {1, 2, 3, 4, 5}, according to lemma 4, we have gcd(pi , pj ) = 1, which
implies that pi 6= pj , (i 6= j).
So, p1 , p2 , p3 , p4 , p5 are distinct. According to lemma 7, k just need to satisfy the
following congruence equations


 k + 1 ≡ 0 (mod p1 )
k + 1 ≡ 0 (mod p2 )




mk + 1 ≡ 0 (mod p3 )




 5
 m k + 1 ≡ 0 (mod p4 )

m7 k + 1 ≡ 0 (mod p5 )
k 6≡ −1 (mod q1 )




k 6≡ −1 (mod q2 )




···




k 6≡ −1 (mod qt ),


where q1 , q2 , . . . , qt are all distinct prime factors of m − 1. Since p1 , p2 , p3 , p4 , p5 ,

q1 , q2 , . . . , qt are distinct, according to Chinese remainder theorem, there are in-
finitely many such integers k.
When m = 2l − 1, l ∈ N, we can use the covering system
 
13 2(4), 4(8), 8(16), 8(24), 0(48),
{ai (ni )}i=1 =
1(3), 5(6), 3(12), 1(5), 7(10), 3(15), 9(20), 15(30)

to complete the proof. Since m ≥ 3, according to lemma 5 and lemma 6, we can take
a prime factor pi from Φni (m), such that gcd(pi , ni ) = 1 for every i ∈ {1, 2, . . . , 12}.
Since Φ30 (m) = m8 + m7 − m5 − m4 − m3 + m + 1, it is easy to verify that
gcd(30, Φ30 (m)) = 1. Hence we can take a prime factor p13 6= 2, 3, 5 from Φ30 (m).
According to lemma 3 and lemma 4, we have gcd(m − 1, pi ) = 1, gcd(pi , pj ) = 1,
(i 6= j), which implies that p1 , p2 , . . . , p13 are distinct and are not the factors of
m − 1.
 6

According to lemma 7, k just need to satisfy the following congruence equations:



 ma1 k + 1 ≡ 0 (mod p1 )
ma2 k + 1 ≡ 0 (mod p2 )




···




 a13
m k + 1 ≡ 0 (mod p13 )

 k 6≡ −1 (mod q1 )
k 6≡ −1 (mod q2 )




···




k 6≡ −1 (mod qt ),


where q1 , q2 , . . . , qt are all distinct prime factors of m − 1. Since p1 , p2 , . . . , p13 ,

q1 , q2 , . . . , qt are distinct, according to Chinese remainder theorem, there are in-
finitely many such integers k. This completes the proof of Theorem 1.

Such positive integers k are called “Sierpinski numbers based m”.

If k 6≡ −1 (mod qi ) (i = 1, 2, . . . , t) are replaced by k ≡ 0 (mod m − 1) in the
proof of theorem 1, then we get integers k such that m − 1|k. Therefore, we have
the following corollary.

Corollary 1. For any integer m > 1, there are infinitely many integers k, such
that for any positive integer n, k(m − 1)mn + 1 is a composite number.

Replacing k with −k in theorem 1 and corollary 1, we have the following corol-

laries, which are generalizations of Riesel problem[2].

Corollary 2. For any integer m > 1, there are infinitely many integers k satisfying
k 6≡ 1 (mod q) for any prime number q|(m − 1), such that for any positive integer
n, kmn − 1 is a composite number.

Corollary 3. For any integer m > 1, there are infinitely many integers k, such
that for any positive integer n, k(m − 1)mn − 1 is a composite number.

3. Minimum Sierpinski Number Based m

After proving the existence of Sierpinski numbers based m, we can consider how to
find the minimum Sierpinski number based m.
It can be inferred from previous proofs that as long as the covering system has
less congruences and the corresponding primes are smaller, the positive integer k,
which is obtianed from congruence equations, will be smaller.
In fact, the covering system used to obtain k need not have distinct moduli,
as long as mn1 − 1, mn2 − 1, . . . , mnt − 1 have distinct prime factors p1 , p2 , . . . , pt .
Therefore, the simplest covering system {0(2), 1(2)} can be applied in some condi-
tions.
 7

Example 1. Find the minimum integer k, such that for any n ∈ N, k · 34n + 1 is
a composite number, and k 6≡ 2 (mod 3), k 6≡ 10 (mod 11).

Solution. 342 − 1 = 399 = 3 × 5 × 7 × 11. Take prime factors 5 and 3. We have two
sets of congruence equations:
 
k + 1 ≡ 0 (mod 5) k + 1 ≡ 0 (mod 7)
and
34k + 1 ≡ 0 (mod 7) 34k + 1 ≡ 0 (mod 5)

We obtain k ≡ 29 (mod 35) and k ≡ 6 (mod 35) respectively, in which k = 6 is

a nontrivial solution. When k ≤ 5, 2 and 5 are trivial solutions. 344 + 1 = 1336337,
3 · 34 + 1 = 103 and 4 · 34 + 1 = 137 are all primes. Therefore, the minimum
Sierpinski number based 34 is 6.

If the covering system {0(2), 1(2)} can not be applied. Then we can cosider
about other covering systems which have less congruences such as {0(2), 1(4), 3(4)},
{1(2), 0(4), 2(4)} and {0(3), 1(3), 2(3)}.
Next, taking m = 127 as an example, a method to find a smaller Sierpinski
number is described.
First, take small integers a and factorize 127a − 1. We find that 5 × 1613|1274 − 1,
17 × 137|1278 − 1, 5419|1273 − 1, and 13 × 1231|1276 − 1. Therefore, we can
use the covering system {0(3), 0(4), 2(4), 1(6), 5(6)} with the corresponding primes
5419,5,1613,13,1231. Or we can use the covering system {0(3), 0(4), 1(6), 5(6), 4(8), 0(8)}
with the corresponding primes 5419,5,13,1231,17,137.
For a certain set of moduli {ns }ts=1 , there always exist many covering systems
whose remainders are not all the same, that is {as (ns )}ts=1 and {bs (ns )}ts=1 are
both covering systems but (a1 , a2 , . . . , at ) 6= (b1 , b2 , . . . , bt ). We can get an integer
k from each of these covering systems. Thus, we need to find covering systems with
moduli set {ns }ts=1 as many as possible.
A trivial ideal is to go through all possible remainders and leave those are covering
systems. However, this method is time-consuming. Let N be the least common
multiple of n1 , n2 , . . . , nt . Notice that for any a, b ∈ {0, 1, . . . , N −1}, gcd(a, N ) = 1,
equation ax+ b ≡ as (mod ns ) has unique solution for every s ∈ {1, 2, . . . , t}. Thus,
if {as (ns )}ts=1 is a covering system, then a new covering system can be obtained
by repalcing as with a−1 (as − b). (For some sets of a, b there is as ≡ a−1 (as − b)
(mod ns ) for every s ∈ {1, 2, . . . , t}, which means this set of a, b can not obtain
a new covering system.) Besides, if there is ni = nj with i 6= j in the moduli
set {ns }ts=1 , then a new covering system can be obtaied by swapping ai and aj .
Sometimes the swap has the same effect as a set of a, b, but sometimes not.
Combining these two methods, we can obtain a lot of covering systems with
moduli set {ns }ts=1 . As for the moduli sets {3, 4, 4, 6, 6} and {3, 4, 6, 6, 8, 8}, all cov-
ering systems can be obtained by these two methods from initial covering system
 8

{0(3), 0(4), 2(4), 1(6), 5(6)} and {0(3), 0(4), 1(6), 5(6), 4(8), 0(8)}. Use these cover-
ing systems to obtain k. Then delete trivial solutions. Finally, find the minimum k
in nontrivial soutions.
For the moduli set {3, 4, 4, 6, 6}, the minimum k = 43429139464. For the moduli
set {3, 4, 6, 6, 8, 8}, the minimum k = 11254645362. k = 11254645362 is smaller.
Its covering system is {1(3), 1(4), 0(6), 2(6), 3(8), 7(8)} with corresponding primes
5419,5,13,1231,17,137.

Acknowledgments
I sincerely thank my tutor, professor Yingpu Deng, for introducing me to this
problem and Lixia Luo for suggesting me to use cyclotomic polynomial to solve the
problem, which simplified the proof a lot. They also helped me in modifying this
article.

References
[1] P. Erdös, On integers of the form 2n + p and some related problems, Summa
Brasil. Math, 2: 113-123, 1950.

[2] H. Riesel, Några stora primtal (Swedish: Some large primes), Elementa, 39:
258-260, 1956.

[3] W. Sierpinski, Sur un problème concernant les nombres k · 2n + 1, Elemente

der Mathematik, 15: 73-74, 1960.

[4] G. Xungui, An application of covering systems of congruences, Journal of

Zhoukou Normal University, 33: 1-6, 2016.

[5] Gary Barnes, Sierpinski conjectures and proofs,

http://www.noprimeleftbehind.net/crus/Sierp-conjectures.htm.