Problems in Elementary Number Theory

Volume 1, No. 1, Fall 2008

PEN TEAM: [email protected]
Written by members (2007)
Andrei Frimu Moldova
Yimin Ge Austria
Hojoo Lee Korea
Peter Vandendriessche Belgium
and edited by members (2008)
Daniel Kohen Argentina
David Kotik Canada
Soo-Hong Lee Korea
Cosmin Pohoata Romania
Ho Chung Siu Hong Kong
Ofir Gorodetsky Israel
I
Contents
1 Problems 1
2 Articles 3
2.1 Increasing multiplicative functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Three ways to reach a Diophantine equation . . . . . . . . . . . . . . . . . . . . . . 8
2.3 A theorem on sum-free subsets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.4 A hidden symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.5 On the monotonicity of the divisor function . . . . . . . . . . . . . . . . . . . . . . 18
2.6 Vieta-Jumping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.7 A Combinatorial Congruence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.8 An arithmetic partition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.9 Primitive Roots: Revisited . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.10 Partitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
II
Chapter 1
Problems
Problem 1.0.1 (PEN K12). (Canada 1969) Let N = 1, 2, 3, denote the set of positive
integers. Find all functions f : N N such that for all m, n N: f(2) = 2, f(mn) = f(m)f(n),
f(n + 1) > f(n).
Problem 1.0.2 (PEN H15). (Balkan Mathematical Olympiad 1998) Prove that there are no
integers x and y satisfying x
2
= y
5
4.
Problem 1.0.3 (PEN O53). (Schur Theorem) Suppose the set M = 1, 2, . . . , n is partitioned
into t disjoint subsets M
1
, . . . , M
t
. Show that if n t! e| then at least one class M
z
contains
three elements x
i
, x
j
, x
k
with the property that x
i
x
j
= x
k
.
Problem 1.0.4 (PEN I11). (Korea 2000) Let p be a prime number of the form 4k +1. Show that
p1

i=1

2i
2
p

i
2
p

=
p 1
2
.
Problem 1.0.5 (PEN J11). (a) Saint-Peterburg, 1998 Let d(n) denote the number of positive
divisors of the number n. Prove that the sequence d(n
2
+1) does not become strictly monotonic
from some point onwards.
(b) Prove that d((n
2
+ 1)
2
) does not become monotonic from any given point onwards.
Problem 1.0.6 (PEN A3). (IMO 1988/6) Let a and b be positive integers such that ab+1 divides
a
2
+b
2
. Show that
a
2
+b
2
ab + 1
(1.1)
is the square of an integer.
Problem 1.0.7 (PEN D2). (Putnam 1991/B4) Suppose that p is an odd prime. Prove that
p

j=0

p
j

p +j
j

2
p
+ 1 (mod p
2
).
1
Project PEN http://projectpen.wordpress.com
Problem 1.0.8 (PEN O35). (Romania TST 1998) Let n be a prime and a
1
< a
2
< . . . < a
n
be
integers. Prove that a
1
, a
2
, . . . , a
n
is an arithmetic progression if and only if there exists a partition
of N
0
= 0, 1, 2, . . . , into n sets A
1
, A
2
, . . . , A
n
so that
a
1
+A
1
= a
2
+A
2
= . . . = a
n
+A
n
, (1.2)
where x +A = x +a[a A.
Problem 1.0.9 (PEN B6). Suppose that m does not have a primitive root. Show that
a
(m)
2
1 (mod m) (1.3)
for every a relatively prime to m.
Problem 1.0.10 (PEN O49). ((D. Fomin) [Ams, pp. 12]) Consider the set of all ve-digit
numbers whose decimal representation is a permutation of the digits 1, 2, 3, 4, 5. Prove that this
set can be divided into two groups, in such a way that the sum of the squares of the numbers in
each group is the same.
Problems in Elementary Number Theory 1(2008) No. 1
Chapter 2
Articles
2.1 Increasing multiplicative functions
1
PEN K12
(Canada 1969) Let N = 1, 2, 3, denote the set of positive integers. Find all
functions f : N N such that for all m, n N: f(2) = 2, f(mn) = f(m)f(n),
f(n + 1) > f(n).
First Solution. To get some idea, we rst evaluate f(n) for small positive integers n. It follows
from f (1 1) = f(1) f(1) that f(1) = 1. By the multiplicity, we get f(4) = f(2)
2
= 4. It
follows from the inequality 2 = f(2) < f(3) < f(4) = 4 that f(3) = 3. Also, we compute
f(6) = f(2)f(3) = 6. Since 4 = f(4) < f(5) < f(6) = 6, we get f(5) = 5.
We prove by induction that f(n) = n for all n N. We know that it holds for n = 1, 2, 3.
Now, let n > 2 and suppose that f(k) = k for all k 1, , n. We show that f(n +1) = n +1.
Case 1 n +1 is composite. One may write n +1 = ab for some positive integers a and b with
2 a b n. By the inductive hypothesis, we have f(a) = a and f(b) = b. It follows that
f(n + 1) = f(a)f(b) = ab = n + 1.
Case 2 n + 1 is prime. In this case, n + 2 is even. Write n + 2 = 2k for some positive integer
k. Since n 2, we get 2k = n + 2 4 or k 2. Since k =
n+2
2
n, by the inductive hypothesis,
we have f(k) = k. It follows that f(n+2) = f(2k) = f(2)f(k) = 2k = n+2. From the inequality
n = f(n) < f(n + 1) < f(n + 2) = n + 2 (2.1)
we see that f(n + 1) = n + 1.
By induction, we conclude that f(n) = n for all positive integers n.
Second Solution. As in the previous solution, we get f(1) = 1. From the multiplicativity of f, we
nd that f(2n) = f(2)f(n) = 2f(n) for all positive integers n. This implies that
f

2
k

= 2
k
(2.2)
3
Project PEN http://projectpen.wordpress.com
for all positive integers k. Let k N. From the assumption, we obtain the inequality
2
k
= f

2
k

< f

2
k
+ 1

< < f

2
k+1
1

< f

2
k+1

= 2
k+1
. (2.3)
In other words, the increasing sequence of 2
k
+ 1 positive integers
f

2
k

, f

2
k
+ 1

, , f

2
k+1
1

, f

2
k+1

(2.4)
lies in the set of 2
k
+ 1 consecutive integers 2
k
, 2
k
+ 1, , 2
k+1
1, 2
k+1
. This means that
f(n) = n for all 2
k
n 2
k+1
. Since this holds for all positive integers k, we conclude that
f(n) = n for all n 2.
Third Solution. The assumption that f(mn) = f(m)f(n) for all positive integers m and n is too
strong. We can establish the following proposition.
Proposition 2.1.1. (Putnam 1963/A2) Let f : N N be a strictly increasing function satisfying
that f(2) = 2 and f(mn) = f(m)f(n) for all relatively prime m and n. Then, f is the identity
function on N.
Proof Since f is strictly increasing, we nd that f(n+1) f(n)+1 for all positive integers n.
It follows that f(n +k) f(n) +k for all positive integers n and k. We now determine p = f(3).
On the one hand, we obtain
f(18) f(15) + 3 f(3)f(5) + 3 f(3)(f(3) + 2) + 3 = p
2
+ 2p + 3. (2.5)
On the other hand, we obtain
f(18) = f(2)f(9) 2(f(10)1) = 2f(2)f(5)2 4(f(6)1)2 = 4f(2)f(3)6 = 8p6. (2.6)
Combining these two, we deduce p
2
+ 2p + 3 8p 6 or (p 3)
2
0. So, we have f(3) = p = 3.
We now prove that f

2
l
+ 1

= 2
l
+ 1 for all positive integers l. Since f(3) = 3, it clearly
holds for l = 1. Assuming that f

2
l
+ 1

= 2
l
+ 1 for some positive integer l, we obtain
f

2
l+1
+ 2

= f(2)f

2
l
+ 1

= 2

2
l
+ 1

= 2
l+1
+ 2. (2.7)
Since f is strictly increasing, this means that f

2
l
+k

= 2
l
+ k for all k 1, , 2
l
+ 2. In
particular, we get f

2
l+1
+ 1

= 2
l+1
+ 1, as desired.
Now, we nd that f(n) = n for all positive integers n. It clearly holds for n = 1, 2. Let l be
a xed positive integer. We have f

2
l
+ 1

= 2
l
+1 and f

2
l+1
+ 1

= 2
l+1
+1. Since f is strictly
increasing, this means that f

2
l
+k

= 2
l
+ k for all k 1, , 2
l
+ 1. Since it holds for all
positive integers l, we conclude that f(n) = n for all n 3. This completes the proof.
Fourth Solution. We can establish the following general result.
Proposition 2.1.2. Let f : N R
+
be a function satisfying the conditions:
(a) f(mn) = f(m)f(n) for all positive integers m and n, and
(b) f(n + 1) f(n) for all positive integers n.
Then, there is a constant R such that f(n) = n

for all n N.
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
Proof We have f(1) = 1. Our job is to show that
ln f(n)
ln n
is constant when n > 1. Assume to
the contrary that
ln f(m)
ln m
>
ln f(n)
ln n
(2.8)
for some positive integers m, n > 1. Writing f(m) = m
x
and f(n) = n
y
, we have x > y or
ln n
ln m
>
ln n
ln m

y
x
(2.9)
So, we can pick a positive rational number
A
B
, where A, B N, so that
ln n
ln m
>
A
B
>
ln n
ln m

y
x
. (2.10)
Hence, m
A
< n
B
and m
Ax
> n
By
. One the one hand, since f is monotone increasing, the rst
inequality m
A
< n
B
means that f

m
A

n
B

. On the other hand, since f

m
A

= f(m)
A
=
m
Ax
and f

n
B

= f(n)
B
= n
By
, the second inequality m
Ax
> n
By
means that
f

m
A

= m
Ax
> n
By
= f

n
B

(2.11)
This is a contradiction.
Fifth Solution. It is known that we get the same result when we only assume that f is monotone
increasing and multiplicative. In fact, in 1946, Paul Erdos proved the following result in [1]:
Theorem 2.1.1. Let f : N R be a function satisfying the conditions:
(a) f(mn) = f(m) +f(n) for all relatively prime m and n, and
(b) f(n + 1) f(n) for all positive integers n.
Then, there exists a constant R such that f(n) = ln n for all n N.
This implies the following multiplicative result.
Theorem 2.1.2. Let f : N R
+
be a function satisfying the conditions:
(a) f(mn) = f(m)f(n) for all relatively prime m and n, and
(b) f(n + 1) f(n) for all positive integers n.
Then, there is a constant R such that f(n) = n

for all n N.
Proof
1
It is enough to show that the function f is completely multiplicative: f(mn) =
f(m)f(n) for all m and n. We split the proof in three steps.
Step 1 Let a 2 be a positive integer and let
a
= x N [ gcd(x, a) = 1. Then, we
obtain
L := inf
x
a
f(x +a)
f(x)
= 1 (2.12)
and
f

a
k+1

a
k

f(a) (2.13)
for all positive integers k.
1
We present a slightly modied proof in [2]. For another short proof, see [3].
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
Proof of Step 1 Since f is monotone increasing, it is clear that L 1. Now, we notice
that f(k +a) Lf(k) whenever k
a
. Let m be a positive integer. We take a suciently large
integer x
0
> ma with gcd (x
0
, a) = gcd (x
0
, 2) = 1 to obtain
f(2)f (x
0
) = f (2x
0
) f (x
0
+ma) Lf (x
0
+ (m1)a) L
m
f (x
0
) (2.14)
or
f(2) L
m
. (2.15)
Since m is arbitrary, this and L 1 force to L = 1. Whenever x
a
, we obtain
f

a
k+1

f(x)
f (a
k
)
=
f

a
k+1
x

f (a
k
)

f

a
k+1
x +a
k

f (a
k
)
= f (ax + 1) f

ax +a
2

= f(a)f(x +a) (2.16)

or
f(x +a)
f(x)

f

a
k+1

f(a)f (a
k
)
. (2.17)
It follows that 1 = inf
x
a
f(x+a)
f(x)

f(a
k+1
)
f(a)f(a
k
)
so that f

a
k+1

a
k

f(a).
Step 2 Similarly, we have
U := sup
x
a
f(x)
f(x +a)
= 1 (2.18)
and
f

a
k+1

a
k

f(a) (2.19)
for all positive integers k.
Proof of Step 2 The rst result immediately follows from Step 1.
sup
x
a
f(x)
f(x +a)
=
1
inf
x
a
f(x+a)
f(x)
= 1. (2.20)
Whenever x
a
and x > a, we have
f

a
k+1

f(x)
f (a
k
)
=
f

a
k+1
x

f (a
k
)

f

a
k+1
x a
k

f (a
k
)
= f (ax 1) f

ax a
2

= f(a)f(xa). (2.21)
It therefore follows that
1 = sup
x
a
f(x)
f(x +a)
= sup
x
a
, x>a
f(x a)
f(x)

f

a
k+1

f(a)f (a
k
)
. (2.22)
Step 3 From the two previous results, whenever a 2, we have f

a
k+1

= f

a
k

f(a). Then,
the straightforward induction gives that
f

a
k

= f(a)
k
(2.23)
for all positive integers a and k. Since f is multiplicative, whenever
n = p
1
k
1
p
l
k
l
(2.24)
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
gives the standard factorization of n, we obtain
f(n) = f

p
1
k
1

p
l
k
l

= f (p
1
)
k
1
f (p
l
)
k
l
. (2.25)
We therefore conclude that f is completely multiplicative.
References
1 P. Erdos, On the distribution function of additive functions, Ann. of Math., 47(1946), 1-20
2 E. Howe, A new proof of Erdoss theorem on monotone multiplicative functions, Amer. Math.
Monthly 93(1986), 593-595
3 L. Moser and J. Lambek, On monotone multiplicative functions, Proc. Amer. Math. Soc.,
4(1953), 544-545
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
2.2 Three ways to reach a Diophantine equation
2
PEN H15
(Balkan Mathematical Olympiad 1998) Prove that there are no integers x and y
satisfying x
2
= y
5
4.
First Solution. Assume to the contrary that a
2
= b
5
4 for some integers a and b. First consider
when a is even:
Since b
5
= a
2
+ 4 is even, b is also even. Since a
2
+ 4 = b
5
is divisible by 2
5
, we have
a
2
4

mod 2
5

. However, this is a contradiction. This is because even x

2
4 (mod 16) is
not possible.
Now, consider the case when a is odd. We rewrite the equation in the form
b
5
= a
2
+ 4 = (a + 2i)(a 2i) (2.26)
and work on Z[i] = x +yi [ x, y Z, the ring of Gaussian integers. Since a is odd, we nd that
a + 2i and a 2i are relatively prime in Z[i]. (Indeed, if Z[i] divides both a + 2i and a 2i,
then also divides (a + 2i) (a 2i) = 4i. In other words, is a divisor of 4i. Since also
divides a + 2i and since a is odd, this implies that is a unit in Z[i].)
We recall that Z[i] is a unique factorization domain. Since a + 2i and a 2i are relatively prime,
b
5
= (a + 2i)(a 2i) guarantees that
a + 2i =
1

1
5
, a 2i =
2

2
5
, (2.27)
where
1
,
1
Z[i] and
1
,
2
are units in Z[i]. Since
1
1, 1, i, i, we get
1
=
1
5
. Hence,
we can write
a + 2i =
1

1
5
= (
1

1
)
5
. (2.28)
After setting
1

1
= p +qi, where p, q Z, it becomes
a + 2i = (p +qi)
5
. (2.29)
Taking conjugates, we also get a 2i = (p qi)
5
. It follows that
4i = (a + 2i) (a 2i) = (p +qi)
5
(p qi)
5
= 2

5p
4
q 10p
2
q
3
+q
5

i (2.30)
or
2 = q

5p
4
10p
2
q
2
+q
4

. (2.31)
Now, we get back in the game on Z. Since q divides 2, we get q 2, 1, 1, 2. Reading
the above equation modulo 5, 2 q
5
(mod 5). Since Fermats Little Theorem says that
q
5
q (mod 5), we have 2 q (mod 5) or q = 2. However, plugging q = 2 into the above
equation, we obtain 2 = 2

5p
4
40p
2
+ 16

or 3 = p
2
(8 p
2
). Since p
2
divides 3, we get p = 1.
However, p = 1 means that p
2
(8 p
2
) = 7. This is a contradiction.
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
Second Solution. Now, assume to the contrary that a
2
= b
5
4 for some integers a and b. As in
the rst solution, it is easy to show that the case when a is even is impossible. We consider the
case when a is odd. So, b is also odd. Since 4 = 6
2
2
5
, one may rewrite the equation in the form
a
2
+ 6
2
= b
5
+ 2
5
= (b + 2)

b
4
+ (2)b
3
+ (2)
2
b
2
+ (2)
3
b + (2)
4

. (2.32)
Letting d
1
= b + 2 and d
2
= b
4
+ (2)b
3
+ (2)
2
b
2
+ (2)
3
b + (2)
4
, we get
a
2
+ 6
2
= b
5
+ 2
5
= d
1
d
2
. (2.33)
We can exclude the case when d
1
= 1 or when d
2
= 1. Indeed, d
1
= 1 or b = 3 implies that
a
2
+ 6
2
= b
5
+ 2
5
= (3)
5
+ 2
5
< 0, (2.34)
which is a contradiction. If d
2
= 1, then b
5
+2
5
= d
1
d
2
= d
1
= b2 or b
5
+b = (2)
5
+(2).
Since the function t t
5
+ t is strictly increasing, we have b = 2 or a
2
= b
5
4 = 36 < 0,
which is a contradiction.
We now clam that the integer b
5
+ 2
5
= d
1
d
2
has a prime divisor q = 3 with q 1 (mod 4).
Step 1 We show that it is not possible that both d
1
and d
2
are divisible by 3. Indeed, if d
1
is divisible by 3, since b d
1
2 2 (mod 3), we nd that
d
2
b
4
+ (2)b
3
+ (2)
2
b
2
+ (2)
3
b + (2)
4
5(2)
4
0 (mod 3). (2.35)
Step 2 If b 1 (mod 4), then we get a
2
b
5
4 1 (mod 4), which is impossible. Hence,
b 1 (mod 4). Since d
1
b + 2 1 (mod 4) and since d
1
= 1, we see that [d
1
[ > 1. Since
d
1
1 (mod 4) and since [d
1
[ > 1, we see that d
1
has at least one prime divisor congruent to 1
modulo 4.
Step 3 It follows from d
1
d
2
b
5
+ 2
5
1 (mod 4) and from d
1
b + 2 1 (mod 4) that
d
2
1 (mod 4). It follows from this and from d
2
= 1 that [d
2
[ > 1. Since d
2
1 (mod 4),
this implies that d
2
also has at least one prime divisor congruent to 1 modulo 4.
Combining results from Step 1 through Step 3, we conclude that at least one of d
1
or d
2
has a
prime divisor q = 3 with q 1 (mod 4). Since q divides b
5
+ 2
5
= d
1
d
2
, our claim is proved.
Now, we employ the following well-known result.
Proposition 2.2.1. Let p 1 (mod 4) be a prime. Let a and b are integers such that a
2
+ b
2
is divisible by p . Then, both a and b are divisible by p.
Since a
2
+ 6
2
= b
5
+ 2
5
, this means that q also divides a
2
+ 6
2
. From Proposition 2.2.1, we
see that both a and 6 are divisible by q. Since q 1 (mod 4) and since q divides 6, we get q = 3.
This is a contradiction for the choice of q.
Now, we oer two dierent ways to establish Proposition 2.2.1.
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
First Proof Assume to the contrary that at least one of them is not divisible by p. Since p
divides a
2
+ b
2
, we see that none of them are divisible by p. Since p divides a
2
+ b
2
, we obtain
a
2
b
2
(mod p). Raise both sides of the congruence to the power
p1
2
and apply Fermats
Little Theorem to obtain
1 a
p1
(1)
p1
2
b
p1
b
p1
1 (mod p). (2.36)
This is a contradiction because p is an odd prime.
Second Proof Again, assume to the contrary that none of them are divisible by p. Since p
divides a
2
+b
2
, we have the congruence a
2
b
2
(mod p) or

ab
1

2
1 (mod p). This means
that 1 is a quadratic residue modulo p, which is a contradiction for p 1 (mod 4).
Third Solution. Just toss the Diophantine equation x
2
= y
5
4 on the eld Z/11Z! It turns out
that x
2
y
5
4 (mod 11) has no solutions. Here is an example of straightforward generalizations:
Proposition 2.2.2. Let p 1, 11, 7, 17 (mod 60) be a prime. Then, the equation
y
p1
2
= x
2
+ 4 (2.37)
has no integral solutions.
Hint. Read the equation modulo p!
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
2.3 A theorem on sum-free subsets
3
PEN O53
(Schur Theorem) Suppose the set M = 1, 2, . . . , n is partitioned into t disjoint
subsets M
1
, . . . , M
t
. Show that if n t! e| then at least one class M
z
contains three
elements x
i
, x
j
, x
k
with the property that x
i
x
j
= x
k
.
First Solution.
Fact 2.3.1. Using Taylor Series approximation for the function f(x) = e
x
at point 0 for x = 1,
we obtain the well-known identity
e = 1 +
1
1!
+
1
2!
+. . . +
1
n!
+. . . , (2.38)
hence
t! e = t!

1 +
1
1!
+
1
2!
+. . . +
1
t!

+
1
t + 1
+
1
(t + 1)(t + 2)
+. . . (2.39)
Note that, for t 2,
1
t + 1
+
1
(t + 1)(t + 2)
+. . . <
1
t + 1
+
1
(t + 1)
2
+
1
(t + 1)
3
+. . . = 1 +
1
1
1
t + 1
=
1
t
, (2.40)
hence S
t
= t!e| = t!

1 +
1
1!
+
1
2!
+. . . +
1
t!

. It is easy to see that the sequence (S

t
)
t0
satises
the recurrence relation
S
t
= tS
t1
+ 1 (2.41)
for t 1, dening S
0
= 1.
Now we can proceed with the solution of the problem. Assume no subset of the partition
contains three elements a, b, c so that a + b = c. From the recurrence relation we have t [S
t
hence by Pigeonhole Principle, at least

S
t
t

+ 1 = S
t1
+ 1 elements of M are found in the
same subset of the partition. Denote this subset by M
1
= x
1
, x
2
, . . . , x
k
so that x
1
< . . . < x
k
,
and k S
t1
+ 1. Consider the set Y = y
1
, . . . , y
k1
, dened by y
i
= x
i+1
x
1
. Clearly
[Y [ = k 1 S
t1
and no element of Y is in M
1
(otherwise, if y
i
M
1
, then y
i
+ x
1
= x
i+1
,
contradiction). Consequently all elements of Y lie in the remaining t 1 subsets. Using similar
arguments, at least

k 1
t 1

+ 1

S
t1
1
t 1

+ 1 = S
t2
+ 1 elements of Y are found in the
same subset from the partition of M. Without loss of generality, let M
2
be this subset. Then
M
2
= y
1
, . . . , y
s
= x
2
x
1
, . . . , x
s+1
x
1
, where s S
t2
+ 1. Because y
i
y
1
= x
i+1
x
2
,
we obtain y
i
y
1
/ M
1
M
2
. Let Z = y
2
y
1
, . . . , y
s
y
1
= x
3
x
2
, x
4
x
2
, . . . , x
s+1
x
2
.
Then the [Z[ S
t2
elements of Z are in the remaining t 2 subsets of the partition. By an easy
induction, we get that the subset M
i
= x
i
x
i1
, x
i+1
x
i1
, . . . , = y
i1
y
i2
, y
i
y
i2
, . . .
of the partition contains at least S
ti
+ 1 elements, using at the induction step the observation
that the dierence of any two elements of the set M
i
, i > 1, is the dierence of some 2 elements
of each of the sets M
1
, . . . , M
i1
. Moreover for each j < i there is an z M
j
so that for each
c M
i
, there is a d M
j
so that c = d z.
In the end, the set M
t
will contain at least S
0
+ 1 = 2 elements. Assume M
t
= a, b with
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
a < b. Then the number b a must be in one of the subsets M
1
, . . . , M
t1
. Assume b a M
i
.
But b and a, are, again by the construction of the sets (M
j
), of the form z
m
z
k
and z
n
z
k
, where
z
m
, z
n
, z
k
M
i
. We obtained a contradiction because (b a) +z
n
= (z
m
z
k
) (z
n
z
k
) +z
n
=
z
m
.
Second Solution. We use a theorem of Ramsey:
Theorem 2.3.1. Let a
1
, . . . , a
k
1 be positive integers and k 2. There exists a smallest positive
integer n = R
k
(a
1
, . . . , a
k
) so that for any coloring with k colors of the complete graph K
n
there
is an index i, 1 i k and a complete subgraph K
a
i
of K
n
with all edges of the same color.
A proof of this theorem can be found in almost any book on Combinatorics or Graph Theory.
Now we will show that
Proposition 2.3.1. R
t

3, 3, . . . , 3
. .. .
t times

t! e| + 1
Proof We proceed by induction on t 2. For t = 2 we easily get R
2
(3, 3) = 6. Indeed,
R
2
(3, 3) > 5 as a regular pentagon having edges of one color, and diagonals of the other contains
no monochromatic triangle. On the other side, every vertex of a K
6
has at least 3 neighbors
to which it is joined by edges of the same color, say 1. If any of the edges between these three
neighbors has color 1, we are done, otherwise they form a monochromatic triangle with edges of
color 2.
Assume the statement true for some t 2. Let n = t! e|. We will show it holds for t + 1.
Let m = (t +1)! e| +1. Each vertex of K
m
is endpoint for m1 edges. Using the Fact, we have
m1 = (t +1)! e| = 1 +(t +1)t! e| = 1 +(t +1)n, so any vertex V of K
m
has at least n +1
neighbors with which it is joined by edges of the same color, say color t +1. Consider the complete
graph G formed by these n +1 vertices. If some vertices A, B of this graph are joined by an edge
of color t +1, then A, B, V form a monochromatic triangle. Otherwise all edges of G have one of t
colors. Since G has n+1 = t!e|+1 vertices, by the induction hypothesis, it has a monochromatic
triangle. Consequently K
m
has a monochromatic triangle, so R
t+1

3, 3, . . . , 3
. .. .
t+1times

(t+1)!e|+1,
and the induction step is over.
The statement of Schur Theorem follows easily from the Proposition 2.3.1. Indeed, let n =
t! e|. Now assign to the vertices of a complete graph with n + 1 vertices K
n+1
the numbers
1, 2, . . . , n, n + 1. Color each edge (i, j) of K
n+1
with the color c, where [i j[ M
c
. By Propo-
sition 1 R
t
(3, 3, . . . , 3) t! e| + 1 = n + 1, hence K
n+1
contains a monochromatic triangle. Let
x < y < z be the vertices of this monochromatic triangle. Then y x, z x and z y belong
to the same set M
i
, for some 1 i t. Since (y x) + (z y) = (z x) the proof of Schurs
Theorem is completed.
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
Remark 2.3.1 (Schur Number). The Schur Number S(t) is dened as the largest positive integer
n so that there exists a partition in t subsets of the set 1, 2, . . . , n, no subsets containing three
integers x, y, z so that x + y = z (x, y, z need not be dierent). As of now, only the rst 4
exact values of the Schur Number are known, namely S(1) = 1, S(2) = 4, S(3) = 13 and S(4) =
44. We have proved that S(t) t! e| 1. This upper bound can be slightly improved to
S(t)

t!

e
1
24

1. From among the lower bounds, the following estimations are known:
S(t) 2
t
1, S(t)
3
t
1
2
and S(t) c 321
t
5
for t > 5 and some constant c.
References
1 H. L. Abbott and D. Hanson, A Problem of Schur and Its Generalizations, Acta Arith.,
20(1972), 175-187.
2 H. L. Abbott and L. Moser, Sum-free Sets of Integers, Acta Arith., 11(1966), 392-396.
3 T. C. Brown, P. Erdos, F.R.K. Chung and R. L. Graham, Quantitative forms of a theorem
of Hilbert, J. Combin. Theory Ser. A, 38(1985), No. 2, 210-216.
4 F. R. K. Chung, On the Ramsey Numbers N(3, 3, , 3; 2). Discrete Math., 5(1973), 317-
321.
5 F. R. K. Chung and C. M. Grinstead, A Survey of Bounds for Classical Ramsey Numbers,
J. Graph Theory, 7(1983), 25-37.
6 A. Engel, Problem Solving Strategies, Chapter 4, The Box Principle.
7 G. Exoo, A Lower Bound for Schur Numbers and Multicolor Ramsey Numbers of K
3
, Elec-
tron. J. Combin., 1(1994), #R8.
8 H. Fredricksen, Five Sum-Free Sets, Proceedings of the Sixth Southeastern Conference on
Combinatorics, Graph Theory and Computing (Florida Atlantic Univ., Boca Raton, Fla.,
1975), 309-314.
9 H. Fredrickson, Schur Numbers and the Ramsey Numbers N(3, 3, ..., 3; 2), J. Combin. Theory
Ser. A, 27(1979), 371-379.
10 H. Fredricksen and M. M. Sweet, Symmetric Sum-Free Partitions and Lower Bounds for
Schur Numbers, Electron. J. Combin., 7(2000), #R32.
11 G. Giraud, Une generalisation des nombres et de linegalite de Schur, C.R. Acad. Sc. Paris,
Serie A, 266(1968), 437-440.
12 G. Giraud, Minoration de certains nombres de Ramsey binaires par les nombres de Schur
generalises, C.R. Acad. Sc. Paris, Serie A, 266(1968), 481-483.
13 L. Moser, An Introduction to the Theory of Numbers, Chapter 7, Combinatorial Number
Theory
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
14 L. Moser; G. W. Walker, Problem E985, Amer. Math. Monthly, 59(1952), No. 4, 253.
15 J. Nesetril and M. Rosenfeld, I. Schur, C.E. Shannon and Ramsey Numbers, a short story,
Discrete Math., 229(2001), 185-195.
17 A. Robertson, New Lower Bounds for Some Multicolored Ramsey Numbers, Electron. J.
Combin., 6(1999), #R12.
18 A. Robertson, New Lower Bounds Formulas for Multicolored Ramsey Numbers, Electron. J.
Combin., 9(2002), #R13.
16 S. P. Radziszowski, Small Ramsey numbers, Electron. J. Combin., Dynamic Survey 1, July
2002, revision #9.
19 I. Tomescu, Probleme de Combinatorica si Teoria Grafurilor, Chapter 14, Probleme de tip
Ramsey.
20 J. Fox and D. J. Kleitman, On Rados Boundedness Conjecture, J. Combin. Theory Ser. A,
113(2006), 84-100.
21 E. G. Whitehead, The Ramsey Number N(3, 3, 3, 3; 2), Discrete Math., 4(1973), 389-396.
22 X. Xiaodong, X. Zheng, G. Exoo and S. Radziszowski, Constructive Lower Bounds on Clas-
sical Multicolor Ramsey Numbers, Electron. J. Combin., 11(2004), #R35.
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
2.4 A hidden symmetry
4
PEN I11
(Korea 2000) Let p be a prime number of the form 4k + 1. Show that
p1

i=1

2i
2
p

i
2
p

=
p 1
2
.
First Solution. We begin with an example. We list all quadratic residues of 17:

1
2
16
2
1, 2
2
15
2
4, 3
2
14
2
9, 4
2
13
2
16,
5
2
12
2
8, 6
2
11
2
2, 7
2
10
2
15, 8
2
9
2
13.
(2.42)
Hence a list of all quadratic residues of 17 is given by
1, 2, 4, 8, 9, 13, 15, 16. (2.43)
Can you see the number theoretical symmetry? Yeap! Indeed, in the eyes of modulo 17, it becomes
1, 2, 4, 8, 8, 4, 2, 1. (2.44)
In general, the set of quadratic residues of a prime of the form 4k+1 is symmetric. This observation
is the key idea of the solution. We now give two simple lemmas, which proofs we shall omit.
Lemma 2.4.1. Let a and b integers such that a b (mod p).Then,

2a
p

a
p

2b
p

b
p

. (2.45)
Lemma 2.4.2. Let R. Then, we have
2| 2| =

0, |

0,
1
2

,
1, |

1
2
, 1

.
(2.46)
Since p is prime and since p 1 (mod 4), we see that 1 is a quadratic residue modulo p.
It thus follows that an integer k is a quadratic residue modulo p if and only if k is a quadratic
residue modulo p. So we take
p1
4
integers
1
, , p1
4
in 1, ,
p1
2
so that
p1
4
, ,
1
,
1
, , p1
4
(2.47)
is the complete list of all quadratic residues modulo p. In other words, this is a permutation of
1
2
= (p 1)
2
, 2
2
= (p 2)
2
, ,

p 1
2

2
=

p + 1
2

2
(2.48)
modulo p. It follows that

2
i
p
| 2

i
p
| = 0,

2
i
p
| 2
2
i
p
| = 1,
(2.49)
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
when i

1, ,
p1
4

. Therefore
p1

i=1

2i
2
p

i
2
p

= 2
p1
2

i=1

2i
2
p

i
2
p

= 2
p1
4

i=1

2
i
p

i
p

+ 2
p1
4

i=1

2
i
p

i
p

= 2
p1
4

i=1
0 + 2
p1
4

i=1
1
=
p 1
2
.
Second Solution. Since 1 is a quadratic residue modulo p, we can write
2
1 (mod p) for
some M
p
:= 1, , p 1. The number theoretical idea we employ here is the fact that the
map k k yields an equivalence relation on the set M
p
. We split the proof in two steps.
Lemma 2.4.3. Let , Z with + Z. Then, we obtain
| +| = + 1. (2.50)
In addition to Lemma 2.4.1 notice that when a +b 0 (mod p) and if p does not divide a and b,
we have

2a
p

a
p

2b
p

b
p

2a
p

2b
p

a
p

b
p

= 1. (2.51)
Lemma 2.4.4. Let : M
p
M
p
be the function with (k) k (mod p).
(A) Since the function satises the equation
(2)
(k) = ( (k)) = k for all k M
p
, we see
that
(4)
= is the identity function on M
p
. The bijection naturally oers a partition
of the set M
p
into sets of the type

k, (k),
(2)
(k),
(3)
(k)

= k, (k), k, (k).
2
(B) When k M
p
, we obtain

2k
2
p

k
2
p

2(k)
2
p

(k)
2
p

= 1,

2(k)
2
p

(k)
2
p

2(k)
2
p

(k)
2
p

= 1.
(2.52)
Hence each equivalence class k, (k), k, (k) which has four distinct elements contributes
2 in the sum.
Since there are
p1
4
quadruples, we conclude that the total sum is
p1
2
.
Proof. The rst claim comes easily from the denition of the map . For the second part, since
k
2
+(k)
2
0 (mod p), Lemma 3 implies the result.
2
In other words, M
p
is a union of distinct orbits of .
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
Third Solution. Since 1 is a quadratic residue modulo p, we can write
2
1 (mod p) for some
M
p
:= 1, , p 1.
Lemma 2.4.5. Let : M
p
M
p
be the function with (k) k (mod p). Then, the map is a
bijection. When k M
p
, we obtain

k
2
p

(k)
2
p

=
k
2
p
+
(k)
2
p
1,

2k
2
p

2(k)
2
p

=
2k
2
p
+
2(k)
2
p
1.

2k
2
p

k
2
p

2(k)
2
p

(k)
2
p

= 1.
Proof. The rst claim comes easily from the denition of the map . Consider the second part.
Since p is prime, it is clear that
k
2
p
Z. Since k
2
+(k)
2
0 (mod p) or
k
2
p
+
(k)
2
p
Z, this and
Lemma 2.4.3 give the rst identity. Similarly, Lemma 2.4.3 yields the second one. The third one
follows from these two identities.
Now, we compute the sum. Set
S(p) =
p1

i=1

2k
2
p

k
2
p

, (2.53)
T(p) =
p1

i=1

2(k)
2
p

(k)
2
p

. (2.54)
First, since (1), , (p 1) is a permutation of 1, , p 1, we obtain
S(p) = T(p).
On the other hand, the last summation identity in Lemma 2.4.5 implies that
S(p) +T(p) = p 1.
It therefore follows that S(p) = T(p) =
S(p)+T(p)
2
=
p1
2
.
After seeing these approaches, we invite the reader to think on the following variations:
Proposition 2.4.1. Let p be a prime number of the form 4k + 1. Show that
(a)
p1

i=1

i
2
p

=
(p 1)(p 2)
3
;
(b)
(p1)/2

i=1

i
2
p

=
(p 1)(p 5)
24
;
(c)
p1

i=(p+1)/2

i
2
p

=
(p 1)(7p 11)
24
.
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
2.5 On the monotonicity of the divisor function
5
PEN J11
(a) Saint-Peterburg, 1998 Let d(n) denote the number of positive divisors of the
number n. Prove that the sequence d(n
2
+1) does not become strictly monotonic
from some point onwards.
(b) Prove that d((n
2
+ 1)
2
) does not become monotonic from any given point on-
wards.
Solution for (a). Intuitively, the sequence being required to be strictly monotonic points that it
will eventually grow rather fast. This is a hint to the solution. Note that if n is even, then the set
of divisors of n
2
+ 1 can be partitioned into pairs

d,
n
2
+1
d

, where d <
n
2
+ 1
d
. Clearly d is odd
and less than n. Hence we have at most
n
2
pairs, consequently d(n
2
+ 1) n.
Assuming to the contrary that the sequence becomes strictly monotonic starting with an N,
its obvious that it must be increasing (otherwise d(n
2
+1) would be forced to take negative values
from some point n > N onwards). Note that since n
2
+ 1 is not a perfect square for any n > 0,
hence d(n
2
+1) is an even number for every positive integer n. Since d(n
2
+1) is strictly monotonic
for n N, we deduce
d((n + 1)
2
+ 1) d(n
2
+ 1) + 2.
A straightforward induction proves that
d((n +k)
2
+ 1) d(n
2
+ 1) + 2k. (2.55)
By the inequality established in the beginning of the solution, for N + t even we obtain the
inequalities
N +t > d((N +t)
2
+ 1) d(N
2
+ 1) + 2t, (2.56)
or
N > d(N
2
+ 1) +t (2.57)
for any t > 0 which is impossible, since the rest of the terms of the inequality are constant.
AFTERTHOUGHTS 2.5.1. It must be mentioned that the problem was proposed for the Elem-
ination Round of Saint-Peterburg Mathematical Olympiad in 1998 for 11th grade. The problems
author is A. Golovanov. Only 2 contestants solved the problem at the contest.
Solution for (b). Note that since the sequence is not required to be strictly monotonic, we cannot
infer that it will grow very fast, so the argument used at (a) fails. We will prove the following
generalization:
Claim 2.5.1. Let t and m be two positive integers. Then the sequence d((n
2
+ m
2
)
t
) does not
become monotonic from any given point onwards.
Suppose, to the contrary, that from some point onwards, the sequence becomes monotonic.
We will rstly show that it must be increasing. Indeed, take a prime p of the form 4k +1. Clearly
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
1 is a quadratic residue (mod p), hence so is m
2
, so there is an integer r so that p[r
2
+ m
2
.
Take now d dierent primes s
1
, . . . , s
d
of the form 4u + 1 and let r
i
Z so that s
i
[ r
2
i
+ m
2
.
Using the Chinese Remainder Theorem there is an integer N, so that N r
i
(mod s
i
), for
every i = 1, . . . , d. Then N
2
+ m
2
r
2
i
+ m
2
(mod s
i
), hence s
1
. . . s
d
[N
2
+ m
2
. This implies
that d((N
2
+m
2
)
t
) is unbounded, consequently it must be increasing from some point x
0
onwards.
For shortness of notations let f
n
= f(n) = d((n
2
+ m
2
)
t
). We will use the following very simple
result.
Lemma 2.5.1. gcd(a
2
+m
2
, (a 1)
2
+m
2
) = 1 if gcd(2a 1, 4m
2
+ 1) = 1.
Proof. Let gcd(2a 1, 4m
2
+ 1) = 1 and suppose there is a prime p dividing both a
2
+ m
2
and
(a 1)
2
+m
2
. By subtraction, we obtain p[2a 1. Then 2a 1 (mod p), so 4a
2
1 (mod p), or
4a
2
+4m
2
1+4m
2
(mod p). Since p [ 4(a
2
+m
2
) we obtain 0 1+4m
2
(mod p), contradicting
gcd(2a 1, 4m
2
+ 1) = 1.
Take x > x
0
so that gcd(2x 1, 1 + 4m
2
) = 1. Then from Lemma 2.5.1 and the identity
[x
2
+m
2
][(x 1)
2
+m
2
] = (x
2
x +m
2
)
2
+m
2
we get the inequality f
x1
f
x
f
x
2
x+m
2, since
d(uv) = d(u) d(v) if gcd(u, v) = 1.
We now state the following result, which we are going to use a bit later.
Lemma 2.5.2. Let M be an integer. Then there exists a positive integer so that the polynomial
h(x) = 4x
2
satises
gcd(2h(x) + 1, M) = 1, x Z (2.58)
Proof. Since 2h(x) +1 is odd, we need only prove the lemma for odd M. So assume M is odd and
let b
1
, . . . , b
s
be the set of prime divisors of M. We are looking for so that b
i
[2h(x) + 1 =
8x
2
2 +1, i = 1, s. Since b
i
s are odd, the last condition is equivalent to b
i
[(4x)
2
(4 2).
It is enough to nd a so that 4 2 is a quadratic nonresidue (mod b
i
). For every prime b
i
there exists a quadratic non-residue r
i
(actually there are
b
i
1
2
of them). We will apply once
again the Chinese Remainder Theorem in the following way:
We are looking for an integer L satisfying the following system of equations:
L r
i
(mod b
i
), i = 1, s (2.59)
L 2 (mod 4) (2.60)
and take =
L 2
4
. Clearly we can assume > 0.
Lets continue with the problem. Take M = 1 + 4m
2
in Lemma 2.5.2 to obtain such and
h(x). Using the monotonicity of f we deduce the chain of inequalities
f
2
x1
f
x1
f
x
f
x
2
x+m
2 f
4(x1)
2

,
for suciently large x > x
0
. Here, we may also assume that x
0
is suciently large so that x > x
0
guarantees that h(x) > x
0
. Note that the inequality f
x1
f
x
f
x
2
x+m
2 provides another proof
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
that if f is monotonic, then it must be increasing. Hence f
q
2
f
h(q)
, where q = x 1 x
0
,
and gcd(2q + 1, 1 + 4m
2
) = 1. Because by Lemma 2.5.2 we have gcd(2h(q) + 1, 4m
2
+ 1) = 1 we
further get f(q)
4
f(h(q))
2
f[h(h(q))]. By an easy induction we obtain the inequalities
f(q)
2
k
f

h(h(. . . h(q) . . .))

. .. .
k times

(4q)
2
k

. (2.61)
Here we have iteratively used the fact that h(z) < 4z
2
. We are going now to summarize the
obtained results. Let c = f(q) and dene g(z) to be the positive integer satisfying
(4q)
2
g(z)
z < (4q)
2
g(z)+1
. (2.62)
We easily obtain g(z) = log
2
log
4q
z||. Then the above inequality and the monotonicity of f
implies
c
2
g(z)
f(z) (2.63)
for suciently large z. With this, we have found a lower estimate for f(z).
Lets nd an upper estimate for f(x) which would contradict, for large enough x the lower es-
timate obtained above. For this, let (p
i
)
i1
be the sequence of prime numbers, not containing
the prime divisors of m. Lets take a closer look at f(p
1
. . . p
k
). Let (p
1
. . . p
k
)
2
+ m
2
=
s

i=1
q

i
i
.
Using divisibility arguments, we have q
i
> p
j
for all i = 1, s and j = 1, s. This clearly implies
s

i=1

i
2k. Note that
f (p
1
p
k
) = d

(p
1
. . . p
k
)
2
+m
2

= (t
1
+ 1) . . . (t
s
+ 1) =
def
h(
1
, . . . ,
s
) (2.64)
Using the already stated inequality
s

i=1

i
2k we will prove that h(
1
, . . . ,
s
) (t + 1)
2k
.
Indeed, note that if a > 1 then (t + 1)(t(a 1) + 1) ta + 1. Hence if there is some
i
> 1,
Without Loss Of Generality,
1
> 1, we have h(
1
,
2
, . . . ,
s
) h(
1
1,
2
, . . . ,
s
, 1). By
repeated applications of this inequality until
i
= 1, for all i, we obtain the following inequality
f (p
1
p
k
) = h(
1
, . . . ,
s
) h

1, 1, . . . , 1
. .. .
P

(t + 1)
P

i
(t + 1)
2k
= T
k
, (2.65)
where T = (t + 1)
2
. Dene now the function l(x) to be equal v + 1, where v is the unique
positive integer for which p
1
. . . p
v
< x p
1
. . . p
v+1
. Using once again the monotonicity of f, we
establish the following upper bound for the function f:
f(x) f

p
1
. . . p
l(x)

T
l(x)
(2.66)
Now, since g(x) = log
2
log
4q
x||, we have g(x) > log
2
log
4q
x| 1, hence
2
g(x)
> 2
log
2
log
4q
x1
=
1
2
log
4q
x|. (2.67)
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
It thus follows that
T
l(x)
f(x) c
2
g(x)
>

c
log
4q
x
. (2.68)
By the fact that f is unbounded, we can choose c as large as we want, hence we can assume

c > T
2
. Then, for reaching a contradiction, we will show that l(x) < 2log
4q
x| for large
enough x. Since 1 + log
4t
x < 2 + 2 log
4t
x < 2log
4t
x| for log
4q
x > 3, it is sucient to
prove l(x) 1 < log
4q
x for large enough x. The last inequality is equivalent to (4q)
l(x)1
< x.
Recall that 4q is a constant value. We nd that the primes grow very fast so that the inequality
(4q)
l(x)1
< p
1
. . . p
l(x)1
holds for large enough x. By the denition of l(x), we have then, indeed,
p
1
. . . p
l(x)1
< x, obtaining l(x) 1 < log
4q
x, what we wanted.
AFTERTHOUGHTS 2.5.2 (About the sequence p
i

i1
). We omitted proof of the validity of
the above inequality (4q)
l(x)1
< p
1
. . . p
l(x)1
for large enough x. Our sequence p
i

i1
, though
not equal to the sequence of prime numbers P
i

i1
, is obtained from the set P of all primes by
removing a nite number of primes - those dividing m, hence when x goes to it behaves just as
P does.
AFTERTHOUGHTS 2.5.3. A polynomial f Z[X] is called a Bouniakowsky Polynomial if f
is irreducible, deg f > 1 and gcd(f(1), f(2), . . .)) = 1.
Theorem 2.5.1 (Bouniakowsky Conjecture). A Bouniakowsky polynomial takes prime values for
innitely many valuos of x.
If the Bouniakowsky Conjecture is true, then we can easily prove that d((n
2
+ 1)
t
), where t
is a xed positive integer, doesnt eventually become monotonic. Indeed, assume the contrary and
suppose d((n
2
+1)
t
) is monotonic from some point n n
0
. If the Conjecture is true, n
2
+1 > n
0
is
a prime for innitely many values of n. For such values, n
2
+1 = p, and d((n
2
+1)
t
) = d(p
t
) = t+1.
This, together with the monotonicity of the sequence would imply that d((n
2
+1)
t
) t+1, n n
0
.
However we have proven before that n
2
+ 1 can have an arbitrarily large number of divisors.
References
1 S. L. Berlov, S. V. Ivanov, K. P. Kohasi, St. Peterburg Mathematical Olympiads
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
2.6 Vieta-Jumping
6
PEN A3
(IMO 1988/6) Let a and b be positive integers such that ab +1 divides a
2
+b
2
. Show
that
a
2
+b
2
ab + 1
(2.69)
is the square of an integer.
[PEN A4] (CRUX, Problem 1420, Shailesh Shirali) If a, b, c are positive integers such
that
0 < a
2
+b
2
abc c, (2.70)
show that a
2
+b
2
abc is a perfect square.
Solution. Suppose that a, b are positive integers so that ab + 1 divides a
2
+b
2
and let
k :=
a
2
+b
2
ab + 1
. (2.71)
We have to prove that k is a perfect square. The very fundamental idea of this and similar
problems is to give up the idea of proving properties of a and b directly. Instead, we are going
to prove the desired property of k (i.e. that k is a perfect square) by xing k and considering all
positive integers a, b which satisfy k =
a
2
+b
2
ab+1
, that is, we consider
S(k) :=

(a, b) Z
+
Z
+
:
a
2
+b
2
ab + 1
= k

. (2.72)
By dening this set, we leave the concrete values of a, b and instead take the whole environment
of k into consideration.
The next step is to assume the required statement to be wrong (for the sake of contradiction), that
is, we suppose that k is not a perfect square. The rest of the problem goes by the method of Innite
Descent: We take any pair (a, b) S(k) and show the existence of another pair (a
1
, b
1
) S(k)
which is smaller than (a, b) where (a
1
, b
1
) is said to be smaller than (a, b) if a
1
+b
1
< a +b. This
however is a contradiction because S(k) Z
+
Z
+
implies that there exists a lower bound for
a +b which is also achieved by at least one pair (a, b) S(k).
Suppose that (a, b) S(k) is any pair which satises k =
a
2
+b
2
ab+1
. Wlog assume that a b.
Consider the equation
x
2
+b
2
xb + 1
= k (2.73)
as a quadratic equation in x. This equation is equivalent to
x
2
kxb +b
2
k = 0. (2.74)
We know that x = a is a root of x
2
kxb +b
2
k = 0 since x = a solves
x
2
+b
2
xb+1
= k. Let a
1
be the
other solution of x
2
kxb +b
2
k = 0. Notice that by using the fact that this quadratic equation
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
has another solution, we have found another pair (a
1
, b) that solves
x
2
+b
2
xb+1
= k. The rst step is
to show that (a
1
, b) S(k).
Lemma 2.6.1. a
1
is a positive integer.
Proof. We know from Vietas theorem that a
1
= kb a. Thus, a
1
is an integer. We still have to
prove that a
1
is positive. First, assume that a
1
= 0. But this implies that k = b
2
since we know
that x = a
1
solves the equation
x
2
+b
2
xb+1
= k, a contradiction to our assumption that k is not a
perfect square. Now, assume that a
1
< 0. Then from x
2
kxb +b
2
k = 0 we infer that
k = a
2
1
ka
1
b +b
2
a
2
1
+kb +b
2
> k, (2.75)
clearly impossible. Notice that the last step follows from b > 0. We therefore know that a
1
> 0
and thus a
1
Z
+
.
Corollary 2.6.1. (a
1
, b) S(k).
We hence have constructed another pair in S(k) from any given pair (a, b). If we are able to
prove that this new pair is smaller than the old one, we can use the argument of innite descent
to reach our contradiction and we are done. The next step is to prove that the new pair is indeed
smaller than (a, b).
Lemma 2.6.2. a
1
< a.
Proof. We know that x = a and x = a
1
are the roots of x
2
kxb +b
2
k = 0. It therefore follows
from Vietas theorem that
a
1
=
b
2
k
a
. (2.76)
However, since we assumed that a b, we infer that
b
2
k
a
< a (2.77)
from which a
1
< a follows.
We thus have proved the existence of a pair (a
1
, b
1
) S(k) that is smaller than (a, b), i.e. that
a
1
+b
1
< a +b. Iterating this procedure for (a
1
, b
1
), we can construct another pair (a
2
, b
2
) S(k)
that is smaller than (a
1
, b
1
) and another pair (a
3
, b
3
) S(k) that is smaller than (a
2
, b
2
) and so
on. In other words, we can construct pairs (a
j
, b
j
) for j = 1, 2, . . . so that
a +b > a
1
+b
1
> a
2
+b
2
> a
3
+b
3
> . . . (2.78)
which is impossible since all (a
j
, b
j
) S(k) Z
+
Z
+
.
Revising this method, we rst assumed the existence of a pair (a, b) that does not satisfy the
statement we want to prove. We then went away from this concrete pair (a, b) and instead consid-
ered pairs with the same property as (a, b). The next step is to dene a size of a pair (a, b) which
in our case was simply a + b. It is trivial that this size has a lower bound. Using the theorem of
Vieta, we constructed another pair (a
1
, b
1
) from any given pair (a, b) and we proved that the new
pair is smaller than the old one. This method is called Vieta-Jumping or Root Flipping. Applying
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
the method of innite descent, we obtain our desired contradiction.
With the same ideas, we can also prove A4:
[PEN A4] (CRUX, Problem 1420, Shailesh Shirali) If a, b, c are positive integers such
that
0 < a
2
+b
2
abc c, (2.79)
show that a
2
+b
2
abc is a perfect square.
Indeed, the rst problem is a special case of this one since
0 < a
2
+b
2
abc = c (2.80)
implies that
a
2
+b
2
ab + 1
= c (2.81)
which must be a perfect square.
Solution. Again, as in the rst problem, we assume that there exist positive integers a, b, c so that
k := a
2
+b
2
abc (2.82)
is not a perfect square. We know that k > 0 and k c. We now x k and c and consider all pairs
(a, b) of positive integers which satisfy the equation k = a
2
+b
2
abc, that is, we consider
S(c, k) =

(a, b) Z
+
Z
+
: a
2
+b
2
abc = k

. (2.83)
Suppose that (a, b) is any pair in S(c, k). Wlog assume that a b. Consider the equation
x
2
xbc +b
2
k = 0 (2.84)
as a quadratic equation in x. We know that x = a is a root of this equation. Let a
1
be the other
root of this equation.
Lemma 2.6.3. a
1
is a positive integer.
Proof. Since a
1
and a are the roots of x
2
xbc +b
2
k = 0, we know from the theorem of Vieta
that
a
1
= bc a. (2.85)
It therefore follows that a
1
is an integer. If a
1
= 0 then x
2
xbc +b
2
k = 0 implies that b
2
= k
is a perfect square, a contradiction. If a
1
< 0 then x
2
xbc +b
2
k = 0 implies that
k = a
2
1
+b
2
a
1
bc a
2
1
+b
2
+bc > c, (2.86)
a contradiction to k c. Thus, a
1
is a positive integer.
Corollary 2.6.2. (a
1
, b) S(c, k).
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
Again, it remains to be proven that the new pair (a
1
, b) is smaller than (a, b).
Lemma 2.6.4. a
1
< a.
Proof. We know that a
1
and a are the roots of x
2
xbc +b
2
k = 0, so by Vietas theorem,
a
1
=
b
2
k
a
. (2.87)
Since we assumed that a b, it follows that
b
2
k
a
< a (2.88)
which implies a
1
< a.
We have therefore constructed another pair (a
1
, b
1
) in S(c, k) with a
1
+ b
1
< a + b. However,
S(c, k) Z
+
Z
+
, so using the argument of innite descent, we obtain our desired contradiction.
Remark: There exists a bunch of problems which can be solved with these ideas. Here are
some of them:
1. (IMO 2007/5) Let a, b be positive integers so that 4ab 1 divides (4a
2
1)
2
. Show that
a = b.
Hint: First prove that if 4ab 1|(4a
2
1)
2
, then 4ab 1|(a b)
2
.
2. (A5) Let x and y be positive integers such that xy divides x
2
+y
2
+ 1. Show that
x
2
+y
2
+ 1
xy
= 3. (2.89)
3. Let a, b be positive integers with ab = 1. Suppose that ab 1 divides a
2
+b
2
. Show that
a
2
+b
2
ab 1
= 5. (2.90)
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
2.7 A Combinatorial Congruence
7
PEN D2
(Putnam 1991/B4) Suppose that p is an odd prime. Prove that
p

j=0

p
j

p +j
j

2
p
+ 1 (mod p
2
).
First Solution. We rst oer three well-known properties on binomial coecients.
Lemma 2.7.1. Let p be a prime and let k 1, , p 1. Then, we have

(a)

p
k

0 (mod p),
(b)

p+k
k

1 (mod p),
(c)

2p
p

2

mod p
2

.
(2.91)
Proof. For (a) and (b), we work on the eld Z/pZ, also denoted as Z
p
, and identify the coset
a = a +pZ with a Z. We compute
(a)

p
k

p
k

p 1
k 1

0
k

p 1
k 1

= 0, (2.92)
and
(b)

p +k
k

=
(p +k)!
k!p!
=
1
k!
k

i=1
(p +i) =
1
k!
k

i=1
i =
1
k!
k! = 1. (2.93)
It follows from Vandermondes Identity and (a) that
(c)

2p
p

k=0

p
k

p
p k

1 +
p1

k=1

p
k

2
+ 1 2

mod p
2

. (2.94)
Now, we prove the congruence in the problem. By (a) and (b) in Lemma 12, whenever
j 1, , p 1, the integer

p+j
j

p
j

is divisible by p
2
, in other words,

p
j

p+j
j

p
j

mod p
2

.
It follows from the above, Lemma 12 and the Binomial Theorem that
p

j=0

p
j

p +j
j

1 +

p1

j=1

p
j

p +j
j

2p
p

(mod p
2
)
1 +
p1

j=1

p
j

+ 2 (mod p
2
)
1 + (2
p
2) + 2 (mod p
2
)
2
p
+ 1 (mod p
2
).
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
Second Solution. We establish the following combinatorial identity.
Lemma 2.7.2. For all positive integers n, we have
n

k=0

n
k

n +k
k

=
n

k=0

n
k

2
2
k
. (2.95)
Proof. We rst expand the polynomial (2 + x)
n
(1 + x)
n
= ( (1 + x) + 1)
n
(1 + x)
n
in two ways.
On the one hand, we compute
f(x) =

k=0

n
k

2
k
x
nk

j=0

n
j

x
j

=
2n

l=0

k+j=l,0k,jn

n
k

n
j

2
k

x
l
. (2.96)
On the other hand, we compute
f(x) =

k=0

n
k

(1 +x)
k

(1 +x)
n
=
n

k=0

n
k

(1 +x)
n+k
=
n

k=0

n
k

n+k

j=0

n +k
j

x
j

=
2n

j=0

k=max(0,nj)

n
k

n +k
j

x
j
Now, we can nd the coecient of x
n
in f(x) in two ways. The rst identity gives
x
n
[f(x)] =

k+j=n

n
k

n
j

2
k
=
n

k=0

n
k

n
n k

2
k
=
n

k=0

n
k

2
2
k
(2.97)
and the second identity gives
x
n
[f(x)] =
n

k=0

n
k

n +k
n

=
n

k=0

n
k

n +k
k

. (2.98)
Equating the coecients x
n
[f(x)], we get the desired result.
Now, we go back to the original problem. We take n = p in Lemma 12 and use the fact that

p
k

is divisible by p(established in the previous oered solution), where k 1, , p1. We obtain

p

j=0

p
j

p +j
j

1 +
p1

j=1

p
j

2
2
j
+ 2
p
1 + 2
p
(mod p
2
). (2.99)
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
2.8 An arithmetic partition
8
PEN O35
(Romania TST 1998) Let n be a prime and a
1
< a
2
< . . . < a
n
be integers. Prove
that a
1
, a
2
, . . . , a
n
is an arithmetic progression if and only if there exists a partition
of N
0
= 0, 1, 2, . . . , into n sets A
1
, A
2
, . . . , A
n
so that
a
1
+A
1
= a
2
+A
2
= . . . = a
n
+A
n
, (2.100)
where x +A = x +a[a A.
Vasile Pop
First Solution. Assume rstly that a
1
, a
2
, . . . , a
n
is an arithmetic progression. Dene A
i
= knr+
ir +j[k N
0
, 0 j n1. It is easy to see that N
0
= A
1
A
2
. . . A
n
and A
i
A
j
= for i = j.
The converse part is much more dicult. For convenience of notations, let B
i
= A
ni
and
r
i
= a
n
a
ni
. Hence B
i
= B
0
+ r
i
and N
0
= B
0
B
1
. . . B
n1
. Call a segment of length k
of a subset B
i
a set S B
i
of the form m+ 1, . . . , m+k, where m, m+k + 1 B
i
.
Lemma 2.8.1. Any segment of any subset B
i
has length r = r
1
.
Proof. Note that if B
i
for some i > 0 contains a segment of length dierent from r, then so must
B
0
, since B
i
= B
0
+ r
i
. Hence it is enough to show that B
0
consists only of segments of length
r. Indeed, note that if m B
0
then m + r B
1
, hence any segment of B
0
has length at most
r. Assume to the contrary that there is at least one segment of length less than r in B
0
. Among
all such segments, let S = m + 1, . . . , m + k B
0
with k < r be the one with smallest m (the
rst one). Then m+1 +r, m+2 +r, . . . , m+k +r is a segment of B
1
. Since m+k +1 B
0
,
m + r B
1
, and the set m + k + 1, . . . , m + r has r k > 0 elements it follows that there
is a segment T m + k + 1, . . . , m + r of some B
i
, i > 0 of length at most r k. Hence
T r
i
= m + k + 1 r
i
, . . . , m + r r
i
is a segment of length at most r k < r of B
0
. Since
m+k + 1 r
i
< m, this contradicts the denition of S.
Lemma 2.8.2. Each B
i
starts with the segment S
i
= ir, ir + 1, . . . , ir +r 1.
Proof. We prove the statement by induction on i. It is clear that S
0
= 0, 1, . . . , r 1 B
0
,
which is the base of our induction. Assume the statement true for 0 i < k. We are going to
show the statement for i = k.
So S
i
B
i
, i = 0, k 1. Let S
k
B
j
and assume, to the contrary, that j = k. From
a
1
< a
2
< . . . < a
n
we get r
1
< r
2
< . . . < r
n
. This implies j < k. But then S
k
is already
the second segment of B
j
(after S
j
) which is impossible for j > 0, since we havent reached the
second segment of B
0
yet. Hence j = 0, and so S
k
B
0
. Note that for j < k we have r
j
= jr.
Then it follows that S
i
B
ik
for i = k, k + 1, . . . , 2k 1. Again, S
2k
must be a subset of
either B
0
or B
k
. If S
2k
B
0
then we apply the above argument again to obtain S
i
B
i2k
for
i = 2k, 2k +1, . . . , 3k 1. Repeating this process, we obtain that the rst segment of B
k
must be
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
of the form S
tk
, for some t. This implies r
k
= tk.
Lets prove now by induction on l, that if lk < n then each apparition of a segment from B
lk
is followed by a sequence of segments belonging to the sets B
lk+1
, . . . , B
lk+1
, implying that
(l + 1)k n. Moreover, if (l + 1)k n then the rst segment of B
(l+1)k
is of the form S
tk
for
some t.
For l = 0 the statement of trivial. Assume now the statement true for all l < u and lets prove it
for l = u. Assume lk < n. Applying the inductive hypothesis for l = u 1, we get that the rst
segment of B
(u+1)k
= B
lk
is of the form S
tk
for some t. Take the segment S
tk+i
, 0 < i < k. Lets
prove that it belongs to a new segment B
lk+i
. Indeed assume S
tk+i
B
j
for some j. Assume
B
j
has appeared before. The inductive hypothesis shows that each B
xk+i
, x < l, 0 i < k has
only segments of the form S
x

k+i
, for some x

. Hence r
xk+i
= Mrk + ri, for some M. Also, the
inductive hypothesis shows that r
xk
+ ri = r
xk+i
, 0 i < k. It follows that j = t

k + i, for some
t

< l and that r

t

k
= r
j
ri. Since S
tk+i
B
j
has been obtained by adding r
j
to some segment
S
hk
B
0
, it follows that when adding r
t

k
= r
j
ri to S
hk
we should obtain a segment belonging
to B
t

k
. However S
hk
+r
j
r
i
= S
tk+i
ri = S
tk
B
lk
. Contradiction because t

< l.
From the last result, we infer that k[n, which is impossible for 1 < k < n. Hence the proof
of Lemma 2.8.2 is completed.
Lemma 2.8.2 states that S
i
B
i
, for i = 0, 1, . . . , n 1, hence r
i
= ir for all i, implying that
a
1
, a
2
, . . . , a
n
is an arithmetic progression with term dierence r.
Second Solution. If a
1
, . . . , a
n
is an arithmetic progression, proceed like in the previous solution.
Lets prove the converse. Again, let B
i
= A
ni
and r
i
= a
n
a
ni
, hence B
i
= B
0
+r
i
. Let f(m, i)
be the number of nonnegative integers mwhich are in B
i
. Clearly f(m, i) = f(mr
i
, 0). Because
the sets (B
i
) cover the set of nonnegative integers, m+1 = f(m, 0) +f(m, 1) +. . . +f(m, n 1).
Dene x
i
= f(i, 0) for i 0 and x
i
= 0 for i < 0. Using the above remark, we obtain
x
m
+x
mr
1
+. . . +x
mr
n1
= m+ 1, (2.101)
for m 0. Adding the above relation for m1, m+ 1 and subtracting it twice for m, we obtain
t
m
+t
mr
1
+. . . +t
mr
n1
= 0, (2.102)
where t
i
= x
i+1
+x
i1
2x
i
.
From the denition of t
i
and x
i
we observe that t
i
1, 0, 1. This and the recurrence re-
lation for the sequence (t
i
) implies that (t
i
) is a periodic sequence. Let M be the length of its
smallest period. Then t
i+1
+ t
i+2
+ . . . + t
i+M
is a constant value. Lets prove that this value
equals 0. Indeed, let C = t
i+1
+t
i+2
+. . . +t
i+M
. Let N be a positive integer. Summing up the
recurrence relation for m = 0 to N, we obtain
0 = n(t
0
+. . . +t
Nr
n1
) +E, (2.103)
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
where E consists of nitely many t
i
s (for example, less than (r
n1
+1)
2
t
i
s), hence it is bounded:
[E[ < h for some constant h implying that t
0
+. . . +t
Nr
n1
is bounded for all N. On the other
side t
0
+. . . +t
kM1
= kC as it is the sum of k blocks of t
i
s. If C = 0 for large enough k we have
[kC[ > h. Impossible. So C = 0.
Since t
i+1
+ . . . + t
i+M
= (x
i
x
i+1
) (x
i+M
x
i+M+1
) = 0, we have the implication: if
i B
0
then i + M B
0
. Moreover, if i B
j
then i + M B
j
for j = 0, n 1. Let B be the
subset of B
0
having all elements less than M. Lets prove that
B B +r
1
. . . B +r
n1
= 0, 1, . . . , M 1. (2.104)
It is obvious that every m 0, 1, 2, . . . , M 1 belongs to some B + r
i
. For the converse, let
x = y + r
i
, for some y B. Assume, to the contrary, that x M. Let x = qM + r, q 1,
M 1 r 0. Then r B+r
i
, hence r r
i
B, hence r r
i
. Since y +r
i
= qm+r qm+r
i
,
we obtain y qm m. Impossible since y B.
Denote now by R the set 0, r
1
, . . . , r
n1
. Dene set addition as X +Y = x +y[x X, y Y .
We are to show that if B +R = 0, 1, . . . , M 1 and [R[ is a prime number, then 0, r
1
, . . . , r
n1
form an arithmetic progression. We will make use of the following Lemma, which proves better
than anything the power of the Extremal Principle:
Lemma 2.8.3. Let X and Y be two sets so that X+Y = 0, 1, . . . , M1. Let m = min(Y `0).
Then [X[ is a multiple of m and there exist sets X

and Y

so that X

+Y

0, 1, . . . ,
M
m
1

and X = mX

+0, 1, . . . , m1, Y = mY

.
Proof. Lets show rstly that every element of Y is a multiple of m. Indeed, note rstly that
0, 1, . . . , m1 X. Note also that every element from 0, 1, . . . [X[ [Y [ 1 can be uniquely
written as x+y, where x X and y Y . Assume to the contrary that there is an y = qm+r Y ,
with 0 < r < m. Among all such numbers, take the one with smallest q. If qm Y , since r X
then qm+r = (qm+r) + 0 are two representations of the same number as x +y, x X, y Y .
Impossible. Hence qm / Y . Also, qm / X, because otherwise qm + m = (qm + r) + (m r).
Hence qm / X, Y . Since qm < qm+r and all elements less than qm+r in Y are multiples of m,
we deduce the existence of a positive u so that um Y and (q u)m X. Lets prove now that:
if km X for some k < q u then km + r X, for all 0 < r < m; and if km + r X for some
0 < r < m; k < q u then km X. Assume to the contrary that there is a pair (k, r) so that
km X and km+r / X or km / X and km+r X. Among all such pairs take the one with the
smallest k. Assume rstly km X and km + r / X. Consider the number z = km + r. By our
choice z / X. By the minimality of q, we obtain z / Y , hence z / X, Y . Hence there is a positive
x < q such that z = xm+(k x)m+r, where xm Y and (k x)m+r X. By our choice of k
and r, we have (kx)m X. Since km X and xm Y , we obtain two distinct representations:
km = km + 0 = (k x)m + xm. Impossible. The second case is treated in an analogous way.
Consider now the number Z = (q u)m+r. If Z X, then Z +um = (qm+r) + 0, impossible.
Also from the minimality of q, Z / Y . Hence there exist x X and y Y so that Z = x + y.
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
Because Z < qm + r, x = tm for some t > 0, and y = (q u t)m + r. From what was proved
above, we obtain that (q u t)m X. But then (q u t)m + um = (q u)m + 0 are two
distinct representations of (q u)m as sum x +y, x X and y Y . Contradiction.
So every element of Y is a multiple of m and writing Y = mY

, for some set Y

makes sense. Well

now prove in a completely similar way as above that if km X for some k, then km + r X,
for 0 < r < m; and conversely, if km + r X for some 0 < r < m, then km X. Among all
such bad pairs, take the one with the least k. For the sake of completeness we shall now treat
the second case. Assume that km + r X for some 0 < r < m and that km / X. It is easy
to see that km / Y , otherwise km + r = (km + r) + 0. Hence there is a positive x so that
xm Y and (k x)m X. By the choice of our k, we obtain that (k x)m +r X. But then
(km + r) + 0 = [(k x)m + r] + xm are two distinct representations. Contradiction. So we can
write X = mX

+0, 1, . . . , m1.
It remains to prove that X

+ Y

0, 1, 2, . . . ,
M
m
1

. Indeed, let k

0, 1, 2, . . . ,
M
m
1

and take consider z = km 0, 1, . . . , M 1. Since the representation z = x + y in X + Y is

unique and m[y we also have m[x, so the representation k = x/m+y/m = x

+y

where x

,
y

is unique.
Note that Lemma 2.8.3 is symmetrical with respect to X and Y .
Lets now nish the problem. We will prove by induction on [B[ that if B R = 0, 1, . . . , [B[
[R[ 1 then 0, r
1
, . . . , r
n1
form an arithmetic progression. If [B[ = 1, then B = 0 and
B +R = B = 0, 1, . . . , n 1 so r
i
= i and we are done. Assume now the statement true for all
sets B having less than b elements. We have two cases:
If 1 R, then m = min(B ` 0) > 1 and from Lemma 3, m[n. Since n is a prime number,
it follows that m = n. Then it follows that [R

[ = 1, R

= 0, so R = mR

+0, 1, . . . , m1 =
0, 1, . . . , m1 and the statement is true.
If 1 / R, then m = min(R`0) = r
1
[b. By Lemma 3, R = mR

and B = mB

+0, 1, . . . , m1.
R

and B

have the properties that [R

[ = [R[ is a prime, R

+ B

= 0, 1, . . . , [R

[ [B

[ 1
and [B

[ =
b
m
< b, hence by the induction hypothesis the elements of R

form an arithmetic
progression. Because R = mR

the same holds for 0, r

1
, . . . , r
n1
.
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
2.9 Primitive Roots: Revisited
9
PEN B6
Suppose that m does not have a primitive root. Show that
a
(m)
2
1 (mod m) (2.105)
for every a relatively prime to m.
First, we use the well known fact that m has primitive roots if and only if m has the form
2, 4, p
k
or 2p
k
where p is an odd prime number and k is a positive integer. We thus have to prove
the statement for all other numbers.
First Solution. First, notice that if m has no primitive roots and is not a power of 2, then we can
write m as m = m
1
m
2
where m
1
and m
2
are positive integers satisfying 2 [ (m
1
) and 2 [ (m
2
).
Since a
(m
1
)
1 (mod m)
1
for every integer a coprime to m
1
and a
(m
2
)
1 (mod m)
2
for every
integer a coprime to m
2
, we have
a
(m
1
)
(m
2
)
2
1 (mod m
1
) (2.106)
and
a
(m
2
)
(m
1
)
2
1 (mod m
2
) (2.107)
for every integer a coprime to m
1
m
2
= m. Thus, by the chinese remainder theorem, it follows
that
a
(m
1
)(m
2
)
2
= a
(m)
2
1 (mod m
1
m
2
= m) (2.108)
which proves the proposition for all desired m that are not a power of 2.
Suppose now that m = 2
k
is a power of 2 and has no primitive roots. Notice that k 3. The proof
of the claim goes by induction on k. For k = 3, we can simply check that a
(8)
2
= a
2
1 (mod 8)
for all odd a (we have 1
2
3
2
5
2
7
2
1 (mod 8)). Suppose now that a
(2
k
)
2
= a
2
k2
1
(mod 2
k
) for some positive integer k 3 and every odd integer a. Then for every such a, we either
have
a
2
k2
1 (mod 2
k+1
) (2.109)
or
a
2
k2
1 + 2
k
(mod 2
k+1
). (2.110)
In the rst case, we trivially have
a
2
k1
1 (mod 2
k+1
) (2.111)
and in the second case, we have

a
2
k1

1 + 2
k

2
1 + 2
k+1
+ 2
2k
1 + 2
2k
(mod 2
k+1
) (2.112)
and since k 3, we have 2k k + 1,
a
2
k1
1 (mod 2
k+1
) (2.113)
which proves the induction step.
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
Second Solution. From a more advanced and more general point of view, we can analyze the
smallest positive integer t, so that a
t
1 (mod m) for every integer a coprime to m, where m is
a given positive integer.
Denition 2.9.1. Let m be a positive integer. Then (m) denotes the least positive integer t so
that
a
t
1 (mod m) (2.114)
holds for all integers a coprime to m. is called the Carmichael Function.
We start with an easy lemma:
Lemma 2.9.1. Let m and t be positive integers. Then
a
t
1 (mod m) (2.115)
holds for every integer a coprime to m if and only if (m) [ t. In particular, (m) [ (m).
Proof. The claim is trivial if (m) [ t. On the other hand, if
a
t
1 (mod m) (2.116)
holds for every integer a coprime to m, then by the integer division algorithm, there exist integers
q and r so that t = q(m) +r and 0 r (m) 1. Thus, for every integer a coprime to m, we
have
1 a
t
a
q(m)+r
a
q(m)
a
r
a
r
(mod m). (2.117)
But r < (m) and since we have dened (m) as the smallest positive integer with this property,
this implies r = 0 and thus (m) [ t.
In order to solve the problem, we can derive a (well known) formula for (m):
Proposition 2.9.1. Let m 2 be a positive integer. Then
(m) =

(m) if m = 2, 4, p
k
, 2p
k
where p is an odd prime and k Z
+
,
2
k2
if m = 2
k
where k 3 is an integer,
lcm((p
k
1
1
), . . . , (p
k
r
r
)) if m = p
k
1
1
. . . p
k
r
r
is the prime factorization of m.
(2.118)
Proof. The rst line directly follows from the existence of primitive roots modulo m = 2, 4, p
k
, 2p
k
.
For the second one, we can, as in the rst proof of this problem, rst show that if k 3 is an
integer, then
a
2
k2
1 (mod 2
k
) (2.119)
for every odd integer a (which implies that (2
k
) 2
k2
. Next, we can show by induction on k
that there exists an odd integer a which satises ord
2
k(a) = 2
k2
for every integer k 3 (which
implies the minimality of 2
k2
). This is clear if k = 3 or k = 4, simply take a = 3 for example and
observe that ord
8
(3) = 2 and ord
16
(3) = 4. Suppose now that for some integer k 4 and some
odd integer a, we have
ord
2
k1(a) = 2
k3
and ord
2
k(a) = 2
k2
. (2.120)
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
This however implies that
a
2
k3
1 (mod 2
k1
) but a
2
k3
1 (mod 2
k
), (2.121)
so
a
2
k3
1 + 2
k1
(mod 2
k
) (2.122)
and thus, we either have
a
2
k3
1 + 2
k1
(mod 2
k+1
) (2.123)
or
a
2
k3
1 + 2
k1
+ 2
k
(mod 2
k+1
). (2.124)
In the rst case, we have
a
2
k2

1 + 2
k1

2
1 + 2
k
+ 2
2k2
1 + 2
k
1 (mod 2
k+1
) (2.125)
and in the second case, we have
a
2
k2

1 + 2
k1
+ 2
k

2
1+2
2k2
+2
2k
+2
k
+2
k+1
+2
2k
1+2
k
1 (mod 2
k+1
). (2.126)
So in both cases, we have a
2k2
1 (mod 2
k+1
) and since ord
2
k+1(a) [ (2
k+1
) = 2
k
, we have
ord
2
k+1(a) 2
k1
. But we already know that (2
k+1
) 2
k1
and since ord
2
k+1(a) (2
k+1
), this
implies ord
2
k+1(a) = 2
k1
and thus (2
k+1
) = 2
k1
. For the third one, we assume that m = m
1
m
2
where m
1
, m
2
> 1 are coprime positive integers. Then by the chinese remainder theorem,
a
t
1 (mod m) (2.127)
holds for every integer a coprime to m if and only if
a
t
1 (mod m
1
) (2.128)
holds for every integer a coprime to m
1
and
a
t
1 (mod m
2
) (2.129)
holds for every integer a coprime to m
2
. However, by Lemma 2.9.1, the latter holds if and only if
(m
1
) [ t and (m
2
) [ t (2.130)
and since we have dened (m) as the smallest smallest positive integer t which satises a
t
1
(mod m) and thus the smallest positive integer satisfying (m
1
) [ t and (m
2
) [ t, we obtain
(m) = lcm((m
1
), (m
2
)). (2.131)
From Proposition 2.9.1, we immediately infer that
Corollary 2.9.1. Let m 2 be a positive integer. Then (m) = (m) if and only if m =
2, 4, p
k
, 2p
k
where p is an odd prime number and k is a positive integer. For all other m we have
(m) < (m) and since (m) [ (m) by Lemma 2.9.1, we have (m) (m)/2.
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
It thus remains to be proven that if m has no primitive roots, then (m) [ (m)/2. This is
clear if m 8 is a power of two. Otherwise, by Proposition 2.9.1, we have at least two coprime
divisors m
1
and m
2
with (m
1
), (m
2
) being even, which implies that at least one factor 2 drops
out of (m) if compared to (m). This solves the problem.
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
2.10 Partitions
10
PEN O49
((D. Fomin) [Ams, pp. 12]) Consider the set of all ve-digit numbers whose decimal
representation is a permutation of the digits 1, 2, 3, 4, 5. Prove that this set can be
divided into two groups, in such a way that the sum of the squares of the numbers in
each group is the same.
Proof. For this problem, let us denote
: 1, 2, 3, 4, 5 1, 2, 3, 4, 5 : i 6 i
and
S = n N[n > 33333 and n is a digit permutation of 12345. (2.132)
Also, we will write [a, b, c, d, e] for the number 10000a + 1000b + 100c + 10d +e, and we dene
[a, b, c, d, e]

:= [(a), (b), (c), (d), (e)]. (2.133)

Now, we can represent each number s S as s = [a, b, c, d, e], with a, b, c, d, e 1, 2, 3, 4, 5
keeping in mind that abcde > 33333. We dene the dierence T
abcde
as
T
abcde
= [a, b, c, d, e]
2
[(a), (b), (c), (d), (e)]
2
, (2.134)
which rewrites as [6, 6, 6, 6, 6] [a (a), b (b), c (c), d (d), e (d)]. Also, observe that
when [a, b, c, d, e] > 33333, then we automatically have [(a), (b), (c), (d), (e)] < 33333 and
vice versa (since the sum is 66666), hence exactly half of our ve-digit numbers are in S (and is
a bijection between the elements in S and the elements in S
c
, the complement of S.
Now consider the following problem: let T be the set
T := T
abcde
[[a, b, c, d, e] S. (2.135)
Prove that we can split T into two disjoint sets T
1
, T
2
with equal sum of elements.
If we nd such T
1
, T
2
, then
S
1
:= [a, b, c, d, e][[a, b, c, d, e] T
1
[a, b, c, d, e]

[[a, b, c, d, e] T
2

S
2
:= [a, b, c, d, e][[a, b, c, d, e] T
2
[a, b, c, d, e]

[[a, b, c, d, e] T
1
(2.136)
is a valid partition for our original problem.
That leaves us to nd T
1
, T
2
. First observe that, if we dene
: 1, 2, 3, 4, 5 1, 2, 3, 4, 5 : (12)(3)(4)(5). (2.137)
then [a, b, c, d, e] > 33333 if and only if [a, b, c, d, e]

:= [(a), (b), (c), (d), (e)] > 33333, hence

we can split the elements of T into pairs (t
1
, t
2
) with t
1
> t
2
, where the only dierence is the
Problems in Elementary Number Theory 1(2008) No. 1
Project PEN http://projectpen.wordpress.com
switched positions of 1 and 2.
For [a, b, c, d, e] > 33333, we need a 3, so the dierence t
1
t
2
9, 90, 99, 900, 990, 999,
appearing respectively 3!, 3!, 3!, 3! 2!, 3! 2!, 3! 2! times. To nd our T
1
, T
2
, it is sucient to
partition the set of pairs (t
1
, t
2
) into two sets A and B such that

(t
1
,t
2
)T
1
t
1
t
2
=

(t
1
,t
2
)T
2
t
1
t
2
,
since this will cause them to have an equal sum over the elements in the set.
Now note that 999 = 990 + 9, 999 = 900 + 99, 990 = 900 + 90 and 99 = 90 + 9. Hence:
1. Put 3 pairs with t
1
t
2
= 999 into A, 3 pairs with t
1
t
2
= 990 into B, 3 pairs with
t
1
t
2
= 9 into B.
2. Put 3 pairs with t
1
t
2
= 999 into A, 3 pairs with t
1
t
2
= 900 into B, 3 pairs with
t
1
t
2
= 99 into B.
3. Put 3 pairs with t
1
t
2
= 990 into A, 3 pairs with t
1
t
2
= 900 into B, 3 pairs with
t
1
t
2
= 90 into B.
4. Put 1 pair with t
1
t
2
= 99 into A, 1 pair with t
1
t
2
= 90 into B, 1 pair with t
1
t
2
= 9
into B.
By the counting above, we have partitioned all of the pairs into two sets, which have an equal sum
over the elements per construction.
Now, we dene
T
1
:= t
1
[(t
1
, t
2
) A t
2
[(t
1
, t
2
) B
and
T
2
:= t
2
[(t
1
, t
2
) A t
1
[(t
1
, t
2
) B.
Since A and B are partitions of the couples of elements of S, this is a valid partition.
Problems in Elementary Number Theory 1(2008) No. 1