A n+-1 Solutions - Yufei Zhao - Trinity Training 2011

Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

an 1

Trinity Training 2011

Yufei Zhao

an 1
Solutions
Yufei Zhao
Trinity College, Cambridge
yufei.zhao@gmail.com
April 2011

Practice problems:
1. A primitive root mod n is a number g such that the smallest positive integer k for which
g k 1 mod n is (n).
(a) Show that 2 is a primitive root mod 3n for any n 1.
(b) Show that if g is an odd primitive root mod p such that p2 - g p1 1, then g is also
a primitive root mod pn and 2pn for any n 1.
n1

Solution. (a) Since (3n ) = 2 3n1 , the problem amounts to showing that 3n - 23
1
n2
and 3n - 223
1 (when n 2). The first claim follows from reduction mod 3, and the
n2
second claim follows from the exponent lifting trick, as 3 k 22 1, so that 3n1 k 223 1.
n2

(b) Since (pn ) = (2pn ) = (p 1)pn1 , it suffices to show pn - g (p1)p


1 and
n1
pn - g dp
1 for any divisor d of p 1 with d < p 1. The first claim follows from
n2
pn1 k g (p1)p
1 by the exponent lifting trick as p k g p1 1 by assumption, and
the second claim follows from the fact that p | g m 1 if and only if (p 1) | m as g is a
primitive root mod p.
2. (Cyclotomic polynomials) For a positive integer n, define the polynomial n (x) by
Y
2ik
n (x) =
(x e n ).
1kn
gcd(n,k)=1

(a) Prove the polynomial identity


all divisors d of n.

d|n d (x)

= xn 1, where the product is taken over

(b) Prove that n (x) is an integer polynomial.


(c) Let m and n be positive integers, and let p be a prime divisor of n (m). Prove that
either p | n or n | p 1.
(d) (Special case of Dirichlets theorem) Prove that for every positive integer n, there
are infinitely many primes p with p 1 (mod n).
Q
2ik
Solution. (a) The right-hand side polynomial xn 1 can be factored as nk=1 (x e n ).
2ik
For1 k n, each factor x e n appears exactly once in the left hand side (in d (x)
n
for d = gcd(n,k)
) and all factors in the left hand side are of this form.
(b) Use induction on d. We have 1 (x) = x 1. Suppose d (x) is an integer polynomial
for all d < n. Then by (a) n (x) is the quotient of two monic integer polynomials, and
hence it must also be an integer polynomial.
(c) Suppose p - n and n - p 1. We have p | n (m) | mn 1 by (a). So p - m, and
hence p | mp1 1 by Fermats little theorem. Thus p | mgcd(p1,n) 1. Since n - p 1,
1

an 1

Trinity Training 2011

Yufei Zhao

n
k
gcd(p1,n) 1.
gcd(p1,n) . Suppose p k m
mn 1. However, from (a) we know that

gcd(p 1, n) < n. Since p - n, we know p pk

By

the exponent lifting trick, we have


k
the
xn 1
polynomial n (x) is a factor of xgcd(p1,n) 1 . Setting x = m gives us a contradiction, since
mn 1
on one hand we have p | n (m) but on the other hand p does not divide mgcd(p1,n)
as
1
k
it is the quotient of two numbers both exactly divisible by p .
(d) It suffices to show that for every positive integer n 2, there is at least one prime p
with p 1 (mod n), since then we can find infinitely many such p by finding primes pk
such that pk 1 (mod kn) for each k 1.
For n 2, |n (n)| > 1. Let p be a prime divisor of n (n). Since n (n) | nn 1 by (a),
p - n, so n | p 1 by part (b). This gives us the desired prime.
3. (IMO 2003) Let p be a prime number. Prove that there exists a prime number q such
that for every integer n, np p is not divisible by q.
p

1
= pp1 + pp2 + + p + 1 with
Solution. Let q be a prime divisor of p (p) = pp1
p2 - q 1 (this must exist since p (p) 6 1 (mod p2 )). By problem 2, p | q 1. If np p
q1

q1

gcd( q1 ,p)

p
(mod q), then p p nq1 1 (mod q). We have q | gcd(p p 1, pp 1) = p
1,
2
which equals to p 1 since p - q 1. However, we cannot simultaneously have q | p 1
and p | q 1. Thus np p is not divisible by q.

4. (a) Prove that m (x) and n (x) are always relatively prime as polynomials for m 6= n.
(b) Show that if for some integer x, m (x) and n (x) are not relative prime, then m/n
is an integer power of a prime.
Solution. (a) The zeros of n (x) and m (x) are distinct, since the zeros of n (x) are
precisely the primitive n-th roots of unity. Thus the polynomials are relatively prime.
(b) Suppose some prime p divides both m (x) and n (x). By replacing x by x + p if
necessary, we may assume that x > 1. Let us deal with the p = 2 case separately. We
claim that if m (x) is even then m must be a power of 2. Indeed, otherwise let q be an
odd prime divisor of m, and let m = qs, then by the previous problem, m (x) divides
xm 1
(q1)s + x(q2)s + + xs + 1, which is always odd. The p = 2 case follows.
xs 1 = x
Now assume that p > 2. By the previous problem, p divides xm 1 and xn 1, and hence
m
n
p | xgcd(m,n) 1. Let pk k xgcd(m,n) 1. One of gcd(m,n)
and gcd(m,n)
is not divisible by
p, and assume that it is the latter. Then by the exponent lifting trick, pk k xn 1. If
xn 1
gcd(m, n) < n, then n (x) divides xgcd(m,n)
, which is not divisible by p by the above
1
analysis. This contradicts p | n (x). Hence gcd(m, n) = n, i.e., n | m.
m
We claim that m
n is a power of p. If not, then pick some prime q dividing n . We have
p | n (x) | xn 1 | xm/q 1. By the exponent lifting trick, the same power of p divides
m
both xm 1 and xm/q 1. But m (x) divides xxm/q1
, which contradicts p | m (x). Thus
1
m
n is a power of p.

5. Let p1 , p2 , . . . , pk be distinct primes greater than 3. Let N = 2p1 p2 pk + 1.


(a) (IMO Shortlist 2002) Show that N has at least 4n divisors.
k1

(b) Show that N has at least 22

divisors. (Hint: use cyclotomic polynomials)

Solution. (a) Observe that if a and b are coprime odd numbers, then gcd(2a + 1, 2b + 1) =
3, since their gcd must divide gcd(22a 1, 22b 1) = 2gcd(2a,2b) 1 = 22 1 = 3. Since
2ab + 1 is divisible by both 2a + 1 and 2b + 1, it must also be divisible by 31 (2a + 1)(2b + 1).
2

Trinity Training 2011

an 1

Yufei Zhao

We use induction on k. When k = 1, 2p1 + 1 is divisible by 3 and greater than 9, so it must


have at least 4 divisors. Let a = p1 pk1 and b = pk . Suppose that 2a + 1 has at least
4k1 divisors. Since 2a + 1 is coprime to 31 (2b + 1), the number M = 13 (2a + 1)(2b + 1) must
have at least 24k1 divisors (for each divisor d of 2a +1, we get two divisors d and 13 (2b +1)d
of M ). Also M | N and N = 2ab + 1 > M 2 (since 2ab + 1 > 2ab > 22(a+b+1) > M 2 ). So
N has at least 4k divisors (for each divisor d of M , we have divisors d and N/d). This
completes the induction.
(b) It suffices to show that N is divisible by at least 2k1 distinct prime. We have
Q
Y
22p1 pk 1
d|2p p d (x)
p1 pk
= Q 1 k
N =2
+ 1 = p1 p
2d (2).
=
k 1
2
d|p1 pk d (x)
d|p1 pk

Consider the set of divisors d of p1 pk with an odd number of prime factors. There are
2k1 such divisors d, and they provide mutually coprime d (2) by Problem 4. Take one
prime divisor from each such d (2) and we get what we want.
6. (IMO 1990) Determine all positive integers n such that

2n + 1
is an integer.
n2

Solution. We claim that the only solutions are n = 1, 3. Suppose n


/ {1, 3}. Let p be
the smallest prime divisor of n. Then p | 2n + 1, so p | 22n 1. By Fermats little theorem,
we also have p | 2p1 1. Thus p | 2gcd(p1,2n) 1. Since p is the smallest prime divisor
of n, we must have gcd(p 1, 2n) = 2. So p | 22 1 and hence p = 3.
Suppose 3k k n. We have 3 k 22 1. So by the exponent lifting trick, 3k+1 k 22n 1. If
n2 | 2n + 1, then 32k | 22n 1. Thus 2k k + 1, hence k = 1. Thus 3 k n.
Let n = 3m. Suppose m 6= 1. Let q denote the smallest prime divisor of m. By the
same argument as above, we have q | 2gcd(q1,6m) 1, and gcd(q 1, 6m) {2, 6}, so q
divides either 22 1 = 3 or 26 1 = 63 = 7 32 . Since 3 k n, q 6= 3, so q = 7. However,
2n + 1 = (23 )m + 1 2 (mod 7), so 7 cannot divide 2n + 1, contradiction. This shows
that 1 and 3 are the only solutions.
7. (IMO 2000) Does there exist a positive integer N which is divisible by exactly 2000
different prime numbers and such that 2N + 1 is divisible by N ?
Solution. Yes. We will show by induction that for any m 1, there exists a positive
integer N divisibly by exactly m different prime numbers such that N | 2N + 1.
When m = 1, choose N = 3.
We will use the following variant of the exponent lifiting trick: if p is an odd prime, a 2,
k, m 1, ` 0, n odd, pk k a + 1, and p` k n, then pk+` k an + 1. This in fact follows
from our usual exponent lifting trick, as neither a 1 nor an 1 are divisible by p (since
a 1 (mod p) and n is odd), so the claim follows as pk k a2 1 implies pk+` k a2n 1.
Now suppose N = pa11 pamm satisfies N | 2N + 1, where p1 , . . . , pm are distinct prime
and ai 1. Suppose pbi i k 2N + 1 for each i. Write this as pb11 pbmm k 2N + 1. Then
`
by above variant of the exponent lifting trick, we have pb11 +` pb22 pbmm k 2N p1 + 1. For
`
`
` sufficiently large, we also have pb11 +` pb22 pbmm < 2N p1 + 1, so that 2N p1 + 1 has some
`
prime divisor pk+1 distinct from p1 , . . . pk . Then N p`1 pk+1 | 2N p1 pk+1 + 1, and hence we
can choose N 0 = N p`1 pk+1 to complete the induction.
8. Let N be a positive integer ending in digits 25, and m a positive integer. Prove that
for some positive integer n, the rightmost m digits of 5n and N agree in parity (i.e., for

an 1

Trinity Training 2011

Yufei Zhao

1 k m, the k-th digit from the right in n is odd if and only if the k-th digit from the
right in N is odd).
Solution. We will prove by induction on m that there exists infinitely many n that works.
This is trivial when m = 1, 2.
For the inductive step, it suffices to prove the following claim: if n m 2, then the
m2
rightmost m digits of 5n and 5n+2
agree in parity, but the (m + 1)-th digit from the
right differ in parity.
m2

By the exponent lifting trick, we have 2m k 52


1 as 22 k 5 1. It follows that
m2 +n
2
n
m
m
5
5 is divisible by 10 but not 2 10 . The claim follows.
9. (Hensels lemma) Let
f (x) = cn xn + cn1 xn1 + + c2 x2 + c1 x + c0
be a polynomial with integer coefficients. Its derivative f 0 is a polynomial defined by
f 0 (x) = ncn xn1 + (n 1)cn1 xn2 + + 2c2 x + c1 .
Suppose that a Z satisfies p | f (a) and p - f 0 (a). Prove that for any integer k, there
exists an integer b satisfying pk | f (b) and p | b a.
Solution. We use induction on k to find, for each k 1, an integer bk , satisfying b1 = a
and
bk+1 bk (mod pk )
and
f (bk ) 0

(mod pk ).

Note that this implies bk b1 = a (mod p).


When k = 1, we can just take b1 = a. Now assume that k > 1 and bk1 has already been
chosen. Set
bk+1 = bk + pk r
for some integer r to be decided later. We have
f (bk+1 ) = f (bk + pk r) =

n
X

cj (bk + pk r)j

j=0

k
X

k
0
cj (bjk + jpk rbj1
k ) = f (bk ) + p rf (bk )

(mod pk+1 ),

j=0

where the modulo equivalence comes from binomial expansion. (This is related to the
taylor expansion in calculus: f (x + ) f (x) + f 0 (x).) From the induction hypothesis,
we know pk | f (bk ). Also bk a (mod p), so p - f 0 (bk ), and hence f 0 (bk ) has an inverse
mod p, say t Z, satisfying f 0 (bk )t 1 (mod p). Then setting r = f (bpkk )t , we have
f (bk+1 ) f (bk ) f (bk )rf 0 (bk ) = f (bk )(1 rf 0 (bk )) 0

(mod pk+1 ).

since pk | f (bk ) and p | 1 rf 0 (bk ). This completes the induction step.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy