PSEUDOPRIMES MORE STRONG THAN STRONG

PSEUDOPRIMES
GILBERTO GARC

IA-PULGAR

IN, JUAN MIGUEL VEL

ASQUEZ SOTO,
AND JOHN H. CASTILLO
Abstract. We continue the study of the Midys property, establish
new properties of the same and dene Midy pseudoprimes. In this work
we characterize Midy pseudoprimes, give some of their properties and
established interesting connections with other known pseudoprimes, in
particular we show that every divisor of a Midy pseudoprime is a strong
pseudoprime.
1. Introduction
We organize this work as follows. In the rst section, we present the
denition of the Midys property and recall some known results about it. In
the second section, we study new characterizations of the Midys property,
we get some results that will be used in the third section and we obtain
a new proof of a especial case of the Dirichlets Theorem about primes in
arithmetic progressions. Finally, in the last section we introduce the Midy
pseudoprime concept, we prove some properties of Midy pseudoprimes and
we establish some connections between this new concept and other former
concepts of pseudoprimes.
2. Midys Property
Let us x some notation. Let b be a positive integer greater than 1, b
will denote the base of numeration, N a positive integer relatively prime
to b, i.e (N, b) = 1, |b|
N
the order of b in the multiplicative group U
N
of
positive integers less than N and relatively primes to N, and x U
N
. It
is well known that when we write the fraction
x
N
in base b, it is periodic.
By period we mean the smallest repeating sequence of digits in base b in
such expansion, it is easy to see that |b|
N
is the length of the period of the
fractions
x
N
(see Exercise 2.5.9 in [Nat00]). Let d, k be positive integers with
|b|
N
= dk, d > 1 and
x
N
= 0.a
1
a
2
a
|b|
N
where the bar indicate the period
2000 Mathematics Subject Classication. 11A05, 11A07, 11A15, 11A63, 16U60.
Key words and phrases. Period, decimal representation, order of an integer, multiplica-
tive group of units modulo N, pseudoprime, strong pseudoprime, overpseudoprime.
The authors are members of the Grupo de

Algebra, Teora de N umeros y Aplicaciones,
ERM. J.M.Velasquez-Soto was partially supported by CONICET and Universidad del
Valle. J.H. Castillo was partially supported by CAPES, CNPq from Brazil and Universi-
dad de Nari no.
1
2 G. GARC

IA-PULGAR

IN, J.M VEL

ASQUEZ SOTO, AND J.H. CASTILLO

and a
i
s are digits in base b. We separate the period a
1
a
2
a
|b|
N
in d blocks
of length k and let
A
j
= [a
(j1)k+1
a
(j1)k+2
a
jk
]
b
be the number represented in base b by the j-th block and S
d
(x) =
d

j=1
A
j
.
If for all x U
N
, the sum S
d
(x) is a multiple of b
k
1 we say that N has
the Midys property for b and d. It is named after E. Midy (1836), to read
historical aspects about this property see [Lew07] and its references.
We denote with M
b
(N) the set of positive integers d such that N has the
Midys property for b and d and we will call it the Midys set of N to base b.
As usual, let
p
(N) be the greatest exponent of p in the prime factorization
of N.
For example 13 has the Midys property to the base 10 and d = 3, because
|13|
10
= 6, 1/13 = 0.076923 and 07 + 69 + 23 = 99. Also, 49 has the
Midys property to the base 10 and d = 14, since |49|
10
= 42, 1/49 =
0.020408163265306122448979591836734693877551 and 020 + 408 + 163 +
265 +306 +122 +448 +979 +591 +836 +734 +693 +877 +551 = 7 999.
But 49 does not have the Midys property to 10 and 7. Actually, we can see
that M
10
(13) = {2, 3, 6} and M
10
(49) = {2, 3, 6, 14, 21, 42}.
In [GPG09] is given the following characterization of Midys property.
Theorem 1. If N is a positive integer and |b|
N
= kd, then d M
b
(N) if
and only if
p
(N)
p
(d) for all prime divisor p of (b
k
1, N).
The next theorem is a dierent way to write Theorem 1.
Theorem 2. Let N be a positive integer and d a divisor of |b|
N
. The
following statements are equivalent
(1) d M
b
(N)
(2) For each prime divisor p of N such that
p
(N) >
p
(d), there exists
a prime q divisor of |b|
N
that satises
q
_
|b|
p
_
>
q
(|b|
N
)
q
(d).
In [GPCVS11] the authors prove the following theorem.
Theorem 3. Let d
1
, d
2
be divisors of |b|
N
and assume that d
1
| d
2
and
d
1
M
b
(N), then d
2
M
b
(N).
The following result has a big inuence on our work, it is Theorem 3.6 in
[Nat00].
Theorem 4. Let p be an odd prime not divisor of b, m =
p
(b
|b|
p
1) and
t is a positive integer, then
|b|
p
t =
_

_
|b|
p
if t m,
p
tm
|b|
p
if t > m.
PSEUDOPRIMES MORE STRONG THAN STRONG PSEUDOPRIMES 3
3. Other characterizations of Midys Property.
In this section, we will study some consequences of Theorem 2.
Theorem 5. Let N be a positive integer and |b|
N
= kd. If, for all prime
divisor p of N, we have
p
(N) >
p
(d), then the followings statements are
equivalent
(1)
_
b
k
1, N
_
= 1
(2) d M
b
(N)
(3) For each prime divisor p of N, there exists a prime q divisor of d
such that
q
(|b|
p
) >
q
(|b|
N
)
q
(d) .
Proof. The equivalence between (2) and (3) is immediate from Theorem 2.
By Theorem 1 we get that (1) implies (2). Now we prove that (2) implies
(1). Suppose that d M
b
(N) and let g =
_
b
k
1, N
_
. If there exists a
prime divisor p of g, from Theorem 1, we have 0 <
p
(N)
p
(d) and this
is impossible because
p
(d) <
p
(N). Therefore
_
b
k
1, N
_
= 1.
We now will study Theorem 4 of [Lew07], which is attributed by its author
to M. Jenkins (1867). Let p
1
, p
2
. . . , p
t
be dierent primes such that d
M
b
(p
i
) for each i and let h
1
, h
2
, . . . , h
t
be positive integers, when does
N = p
h
1
1
p
h
2
2
p
ht
t
have the Midys property for b and d?. The Jenkins
Theorem gives the answer to this question and the same is independent
from the h
i
s.
By simplicity we go to study the case when t = 3, but the argument
is true for any t. Let p
1
, p
2
, p
3
be dierent primes such that d M
b
(p
i
),
|b|
p
i
= dk
i
, m
i
=
p
i
_
b
|b|
p
i
1
_
for i = 1, 2, 3 then
|b|
N
= d
_
p
h
1
m
1
1
k
1
, p
h
2
m
2
2
k
2
, p
h
3
m
3
3
k
3
_
= dk.
We have to check up the prime divisors of
_
b
k
1, N
_
, to determine when
d M
b
(N). As d is a divisor of |b|
p
1
thus d p
1
1 and if, say, p
1
|
_
b
k
1, N
_
we get that h
1
=
p
1
(N) > 0 =
p
1
(d) and so N does not have
the Midys property for b and d. In consequence d M
b
(N) if and only if
_
b
k
1, N
_
= 1. It is clear that
_
b
k
1, N
_
= 1 is equivalent to say that
for each i, |b|
p
i
k. We will see when this fact is veried. Let d =
s

i=1
q
r
i
i
be
the prime decomposition of d, for each i = 1, 2, 3 take c
i
=
d
(k
i
) and
p
h
1
m
1
k
1
=
_
d
c
1
s

i=1
q

(1)
i
i
_
y
1
p
h
2
m
2
k
2
=
_
d
c
2
s

i=1
q

(2)
i
i
_
y
2
p
h
3
m
3
k
3
=
_
d
c
3
s

i=1
q

(3)
i
i
_
y
3
4 G. GARC

IA-PULGAR

IN, J.M VEL

ASQUEZ SOTO, AND J.H. CASTILLO

where (q
i
, y
j
) = 1 for i = 1, 2, . . . , s and j = 1, 2, 3. Then
k =
_
p
h
1
m
1
k
1
, p
h
2
m
2
k
2
, p
h
3
m
3
k
3
_
=
_
d
c
1
s

i=1
q

(1)
i
i
, d
c
2
s

i=1
q

(2)
i
i
, d
c
3
s

i=1
q

(3)
i
i
_
[y
1
, y
2
, y
3
] .
We want, as was said before, that k not be divisible for any |b|
p
i
, i = 1, 2, 3.
Now if, say, k = |b|
p
1
l with l integer, then k = |b|
p
1
l = (k
1
d) l therefore
k
k
1
is
a multiple of d, and as the y
i
s are relatively primes with d this is equivalent
to
_
d
c
1
s

i=1
q

(1)
i
i
, d
c
2
s

i=1
q

(2)
i
i
, d
c
3
s

i=1
q

(3)
i
i
_
d
c
1
s

i=1
q

(1)
i
i
0 mod d
From the above analysis follows the next theorem.
Theorem 6 (Jenkins Theorem). Let p
1
, p
2
, . . . , p
t
be dierent primes such
that d M
b
(p
i
) for each i, let h
1
, h
2
, . . . , h
t
be positive integers and N =
p
h
1
1
p
h
2
2
p
ht
t
. With the notations introduced above we obtain that d M
b
(N)
if and only if, for each j = 1, 2, . . . , t is satised
_
d
c
1
s

i=1
q

(1)
i
i
, d
c
2
s

i=1
q

(2)
i
i
, . . . , d
ct
s

i=1
q

(t)
i
i
_
d
ct
s

i=1
q

(j)
i
i
0 mod d
Our next result has a similar avor of the Jenkins Theorem, although its
statement and proof are simpler.
Theorem 7. Let N, q, v be integers with q prime and v > 0. Then q
v

M
b
(N) if and only if N = q
n
p
h
1
1
p
h
2
2
p
h
l
l
where n is a non-negative integer,
p
i
s are dierent primes and h
i
s are non-negatives integers not all zero,
verifying 0 n v,
q
(|b|
p
i
) > 0 and
max
1il
_
n m,
q
(|b|
p
i
)
_
v < min
1il
_

q
(|b|
p
i
)
_
where m =
q
(b
|b|
q
1).
Proof. We write |b|
N
= q
t
k, with (q, k) = 1 and t v. Let us denote
g =
_
b
kq
tv
1, N
_
. Suppose that q
v
M
b
(N). By Theorem 1 we know
that g can not be divisible by other prime dierent from q and that
q
(N)
v. Let p = q be a prime divisor of N. Because p not divides g, we have
|b|
p
kq
tv
and thus
q
(|b|
p
) > t v 0 and it is easy to see that t =
max
1il
_
n m,
q
(|b|
p
i
)
_
. Therefore for all prime divisor p of N we get that

q
(|b|
p
) > max
1il
_
n m,
q
(|b|
p
i
)
_
v.
PSEUDOPRIMES MORE STRONG THAN STRONG PSEUDOPRIMES 5
Conversely, from the hypothesis the unique prime divisor of N that could
be a divisor of g is q and thus Theorem 1 implies that q
v
M
b
(N) .
As a special case, if in the above theorem we do v = 1 and since for any p
prime divisor of N we have
q
(|b|
N
)
q
(|b|
p
) and thus we obtain the next
corollary.
Corollary 8. Let N be a positive integer and let q be a prime divisor of
|b|
N
, then q M
b
(N) if and only if
(1) If (N, q) = 1, then
q
(|b|
p
) =
q
(|b|
N
) for all p prime divisor of N.
(2) If (N, q) > 1, then q
2
not divides N and
q
(|b|
p
) =
q
(|b|
N
) for all
p prime divisor of N dierent from q.
Note that, from the last theorem, the smallest number N such that q
v

M
b
(N) has to be a prime P which satises
q
(|b|
P
) > 0 and q
v
| |b|
P
. We
obtain the below corollary, recalling that |b|
P
divides P 1.
Corollary 9. If q is a prime and v is a positive integer, then the smallest
integer N such that q
v
M
b
(N) is a prime congruent with 1 mod q
v
.
As a consequence of the above corollary we have a particular case of the
Dirichlets Theorem about primes in arithmetic progressions.
Corollary 10. If q is a prime and v is a positive integer, there are innitely
many primes which are congruent to 1 modulo q
v
.
Proof. By the last corollary there exists a prime P
1
congruent with 1 modulo
q
v
and which satises q
v
M
b
(P
1
). Take t
1
an integer such that q
t
1
v
> P
1
.
Once again, we can nd a prime P
2
congruent with 1 modulo q
t
1
v
such that
q
t
1
v
M
b
(P
2
) and so on.
4. Midy pseudoprimes
Pomerance and Crandall in their book [CP05], state that:
Suppose we have a theorem, If n is prime, then S is true
about n, where S is some easily checkable arithmetic state-
ment. If we are presented with a large number n, and we wish
to decide whether n is prime or composite, we may very well
try out the arithmetic statement S and see whether it actu-
ally holds for n. If the statement fails, we have proved the
theorem that n is composite. If the statement holds, how-
ever, it may be that n is prime, and it also may be that n is
composite. So we have the notion of S-pseudoprime, which
is a composite integer for which S holds.
Applying the above commentary, to the Fermats little theorem the con-
cepts of pseudoprime and strong pseudoprime are given as follows
Denition 11. The composite integer N is called a pseudoprime (or Fermat
pseudoprime) to base b if (b, N) = 1 and b
N1
1 mod N. An integer
6 G. GARC

IA-PULGAR

IN, J.M VEL

ASQUEZ SOTO, AND J.H. CASTILLO

which is pseudoprime for all possible bases b is called a Carmichael number
or an absolute pseudoprime. An odd composite N such that N 1 = 2
r
s
with s an odd integer and (b, N) = 1, is said to be a strong pseudoprime to
base b if either b
s
1 mod N or b
2
i
s
1 mod N, for some 0 < i < r.
Proposition 12. An odd composite integer N is a strong pseudoprime to
base b if and only if N is pseudoprime to base b and there is a non-negative
integer k such that
2
_
|b|
p
p(N)
_
=
2
_
|b|
p
_
= k for all prime p divisor of
N.
Proof. Let N 1 = 2
t
s. Assume that N is pseudoprime to base b and

2
_
|b|
p
p(N)
_
= k, for some non-negative integer k and for any prime divisor
p of N. If k = 0, it follows that |b|
N
is odd and as b
N1
1 mod N, then
b
s
1 mod N. If k > 0, let |b|
p
p(N) = 2
k
s
p
, then b
2
k1
sp
1 mod p
p(N)
and thus b
2
k1
s
1 mod N for each prime divisor p of N. Therefore, in
any case we obtain that N is a strong pseudoprime to base b. The reciprocal
can be prove in a similar way.
The smallest absolute pseudoprime is 561 and generally such numbers are
square-free and product of at least three primes, Alford et al. in [AGP94]
proved that there are innitely many absolute pseudoprimes.
Theorem 1 implies that if N is prime then N veries the Midys property
for any base b and for all divisor d, dierent from 1, of |b|
N
, this fact and the
commentary quoted from Pomerance and Crandall leave us to study Midy
pseudoprimes and we will dedicate the rest of this work to do it.
Theorem 13. If N is a positive integer such that for all d > 1 and divisor
of |b|
N
is satised that d M
b
(N), then (N, |b|
N
) is either 1 or a prime.
Proof. Let p
1
< p
2
be prime divisors of (N, |b|
N
). Write N = p
1
p
2
N
1
for
some integer N
1
, therefore |b|
N
= p
1
p
2
_
|b|
p
1
, |b|
p
2
_
r; with r an integer. Let
d = p
2
, |b|
N
= p
2
k, because |b|
p
1
divides k it follows that p
1
|
_
N, b
k
1
_
and since d M
b
(N) we have
p
1
(N)
p
1
(d), which is a contradiction
as 1 <
p
1
(N) and
p
1
(d) =
p
1
(p
2
) = 0. It is clear that p
2
not divides
(N, |b|
N
).
Theorem 14. Let N be a positive integer, then d M
b
(N) for all divisor
d > 1 of |b|
N
, if and only if
(1) If (N, |b|
N
) = 1, then N = p
e
1
1
p
e
2
2
. . . p
e
l
l
with each p
i
prime and
|b|
N
= |b|
p
i
for i = 1, . . . , l.
(2) If (N, |b|
N
) = r prime, then N = rp
e
1
1
p
e
2
2
. . . p
e
l
l
with each p
i
prime
and |b|
N
= |b|
p
i
= r
s
|b|
r
for i = 1, . . . , l with s a positive integer.
Proof. From Theorem 3, it is clear that d M
b
(N) for all divisor d > 1 of
|b|
N
if and only if q M
b
(N) for each prime divisor q of |b|
N
. The part (1)
is immediate from the rst case of Corollary 8.
PSEUDOPRIMES MORE STRONG THAN STRONG PSEUDOPRIMES 7
To prove the second part, assume that d M
b
(N) for all divisor d > 1
of |b|
N
. From Theorem 13 we have r = (N, |b|
N
) for some prime r. Take
N = rN
1
with N
1
an integer and |b|
N
= r
s
|b|
r
h where (h, r) = 1. We will
prove that h = 1. If there is a prime divisor q of h, from Corollary 8 follows
that
q
(|b|
N
) =
q
_
|b|
p
_
for all prime divisor p of N, particulary to p = r
we obtain
q
(|b|
r
) =
q
(|b|
N
) =
q
(|b|
r
) +
q
(h) and hence
q
(h) = 0. Thus
h = 1.
Let p = r a prime which divides N, we will see that |b|
N
= |b|
p
. Write
|b|
N
= |b|
p
H with H an integer. If p is a divisor of H, then p divides |b|
N
and
consequently p divides (N, |b|
N
) which is absurd because p = r. Suppose
that there exists a prime q dierent from p and divisor of H, so |b|
N
=
|b|
p
H = |b|
p
H
1
q = qk and as, by the assumption, q M
b
(N), it leaves us
to a contradiction from Theorem 1 because p is a divisor of
_
N, b
k
1
_
. So
H = 1 and |b|
p
= |b|
N
= r
s
|b|
r
. Therefore N = rp
e
1
1
p
e
2
2
. . . p
e
l
l
with each p
i
prime and also |b|
N
= |b|
p
i
= r
s
|b|
r
.
Conversely, assume that N = rp
e
1
1
p
e
2
2
. . . p
e
l
l
, (N, |b|
N
) = r and |b|
N
=
|b|
p
i
= r
s
|b|
r
. Take d > 1 a divisor of |b|
N
, |b|
N
= kd and let g =
_
N, b
k
1
_
. Since |b|
p
i
= |b|
N
= kd for each 1 i l, we obtain that
|b|
p
i
is not a divisor of k therefore p
i
does not divide g. Thus either g = 1
or g = r. In any case, by Theorem 1, N has the Midys property for b and
d.
Denition 15. We say that a number N is a Midy pseudoprime to base b
if N is an odd composite number relatively prime to both b and |b|
N
and for
all divisor d > 1 of |b|
N
we get that d M
b
(N).
By this denition the rst part of Theorem 14 can be write in the following
way.
Theorem 16. An odd composite number N = p
e
1
1
p
e
2
2
. . . p
e
l
l
, with p
i
s dif-
ferent primes and N relatively prime with b, is a Midy pseudoprime to base
b if and only if |b|
N
= |b|
p
i
for every 1 i l.
V. Shevelev denes in [She08] the concept of overpseudoprime numbers
and characterized them in Theorem 7. That result is equivalent to our
Theorem 16, so the concepts of overpseudoprime and Midys pseudoprime
agree.
Theorem 16 give us the following characterization.
Corollary 17. An odd composite N is a Midy pseudoprime to base b if and
only if each divisor of N is either a prime or a Midy pseudoprime to base b.
The bellow result, Theorem 2.3 of [Mot95], allows us to give a equivalent
form of Theorem 16. We denote with
n
(x) the n-th cyclotomic polynomial.
Theorem 18 (Theorem 2.3 of [Mot95]). Let m, b 2, n 3 and p be
integers, where p is the greatest prime divisor of n. Then a composite number
8 G. GARC

IA-PULGAR

IN, J.M VEL

ASQUEZ SOTO, AND J.H. CASTILLO

m is a divisor of
n
(b) if and only if b
n
1 mod m and every prime divisor
q of m satises that
n =
_
_
_
|b|
q
if q = p,
p
e
|b|
p
if q = p.
The following result is a consequence of Theorems 16 and 18.
Theorem 19. A composite number N with (N, |b|
N
) = 1, is a Midy pseu-
doprime to base b if and only if
|b|
N
(b) 0 mod N and |b|
N
> 1.
Theorem 1 of [PSW80] shows the subsequent result for strong pseudo-
primes. We present here a more wide version which is a direct consequence
of Theorems 16 and 18.
Theorem 20. Let N > 2 and f
N
(b) =

N
(b)
(N,
N
(b))
. If f
N
(b) is composite,
then f
N
(b) is a Midy pseudoprime to base b.
Our next result extends Theorem 3.5.10 of [CP05] .
Theorem 21. Let p be an odd prime and 1 < b < p 1, then N =
b
p
+1
b+1
is
either a Midy pseudoprime to base b or a prime.
Proof. It is well known that n odd implies that
2n
(b) =
n
(b) and from
here N =
b
p
+1
b+1
=
p
(b) =
2p
(b). In consequence, N is odd and congruent
with 1 mod p. Therefore, (2p,
2p
(b)) = 1 and the result follows from the
last theorem.
The set of bases of Midy pseudoprimality is closed respect to powers,
although it is not closed by product as we can see when take N = 91 which
is Midy pseudoprime to bases 9 and 16 but it is not to 53, their product
modulo N.
Theorem 22. If N is a Midy pseudoprime to base b, then N is Midy pseu-
doprime to base b
t
for any positive integer t 1.
Proof. The result is immediate from Theorem 16, since N is Midy pseudo-
prime to base b so |b|
N
= |b|
p
for each prime divisor p of N. Now

b
t

N
=
|b|
N
(t, |b|
N
)
=
|b|
p
(t, |b|
p
)
=

b
t

p
. It shows that N is a Midy pseudoprime to base
b
t
.
Theorem 23. If N is a Midy pseudoprime to base b, then N is a pseudo-
prime to base b.
Proof. Write N = p
e
1
1
p
e
2
2
. . . p
e
l
l
and assume that N is a Midy pseudoprime
to base b. From Theorem 16 follows |b|
N
= |b|
p
i
= |b|
p
e
i
i
= t for each
i = 1, 2, . . . , l. By the assumption we get that t | p
i
1 for each i = 1, 2, . . . , l
and thus b
p
j
1
1 mod p
e
i
i
for all pair i, j. So, b
p
j
b mod p
e
i
i
and from
here b
p
e
j
j
b mod p
e
i
i
and, consequently, b
p
e
1
1
p
e
2
2
...p
e
l
l
b mod p
e
i
i
, namely
b
N
b mod p
e
i
i
for each i and therefore b
N
b mod N.
PSEUDOPRIMES MORE STRONG THAN STRONG PSEUDOPRIMES 9
Theorem 24. If N is a Midy pseudoprime to base b, then N is a strong
pseudoprime to base b.
Proof. Write N = p
e
1
1
p
e
2
2
. . . p
e
l
l
and assume that N is a Midy pseudoprime
to base b. We know that N is a pseudoprime to base b. Since N is a Midy
pseudoprime to base b implies that |b|
N
= |b|
n
for each divisor n of N and
thus there is a non-negative integer k such that for all prime divisor p of
N we get that
2
_
|b|
p
p(N)
_
= k and the result follows from Proposition
12.
The reciprocal is not true. For example N = 91 is a strong pseudoprime
to base 53, but this is not a Midy pseudoprime to this base.
Additionally, from the last theorem and Corollary 17 we get that every
composite divisor of a Midy pseudoprime is a strong pseudoprime, in this
sense the Midy pseudoprimes are more strong than strong pseudoprimes.
Among the rst 58892 strong pseudoprimes to base 2 there are only 31520
Midy pseudoprimes to base 2. Similarly, to base 3 there are 2558 Midy
pseudoprimes in the rst 6087 strong pseudoprimes and we found 582 Midy
pseudoprimes to base 5 in the rst 1288 strong pseudoprime to base 5.
Almost the 47% of the strong pseudoprimes are Midy pseudoprimes.
We denote with
k
and

k
the smallest strong pseudoprime and the
smallest Midy pseudoprime to all the rst k primes taken as bases, respec-
tively. From Theorem 24 we know that
k

k
for every positive integer
k. With some calculations, we can see that

1
= 2047,

2
= 5173601
and

3
= 960946321. We know, by [Jae93], the exact values for
k
, with
1 k 8. Thus,

4
> 3215031751 =
4
,

5
> 2152302898747 =

5
,

6
> 3474749660383 =
6
,

7
> 341550071728321 =
7
and

8
>
341550071728321 =
8
.
References
[AGP94] W. R. Alford, Andrew Granville, and Carl Pomerance, There are innitely
many Carmichael numbers, Ann. of Math. (2) 139 (1994), no. 3, 703722.
MR 1283874 (95k:11114)
[CP05] Richard Crandall and Carl Pomerance, Prime numbers, second ed., Springer,
New York, 2005, A computational perspective. MR 2156291 (2006a:11005)
[GPCVS11] Gilberto Garca-Pulgarn, John H. Castillo, and Juan Miguel Velasquez-Soto,
Structure of associated sets to Midys Property, ArXiv e-prints (2011).
[GPG09] Gilberto Garca-Pulgarn and Hernan Giraldo, Characterizations of Midys
property, Integers 9 (2009), A18, 191197. MR MR2506150
[Jae93] Gerhard Jaeschke, On strong pseudoprimes to several bases, Math. Comp. 61
(1993), no. 204, 915926. MR 1192971 (94d:11004)
[Lew07] Joseph Lewittes, Midys theorem for periodic decimals, Integers 7 (2007), A2,
11 pp. (electronic). MR MR2282184 (2008c:11004)
[Mot95] Kaoru Motose, On values of cyclotomic polynomials. II, Math. J. Okayama
Univ. 37 (1995), 2736 (1996). MR 1416242 (97h:11151)
[Nat00] Melvyn B. Nathanson, Elementary methods in number theory, Gradu-
ate Texts in Mathematics, vol. 195, Springer-Verlag, New York, 2000.
MR 1732941 (2001j:11001)
10 G. GARC

IA-PULGAR

IN, J.M VEL

ASQUEZ SOTO, AND J.H. CASTILLO

[PSW80] Carl Pomerance, J. L. Selfridge, and Samuel S. Wagsta, Jr., The pseudo-
primes to 25 10
9
, Math. Comp. 35 (1980), no. 151, 10031026. MR 572872
(82g:10030)
[She08] V. Shevelev, Overpseudoprimes, Mersenne Numbers and Wieferich primes,
ArXiv e-prints (2008).
Gilberto Garca-Pulgarn, Universidad de Antioquia
E-mail address: gigarcia@ciencias.udea.edu.co
Juan Miguel Vel asquez Soto, Departamento de Matem aticas, Universidad
del Valle
E-mail address: jumiveso@univalle.edu.co
John H. Castillo, Departamento de Matem aticas y Estadstica, Universidad
de Nari no
E-mail address: jhcastillo@gmail.com