PUTNAM TRAINING

NUMBER THEORY
(Last updated: February 23, 2021)

Remark. This is a list of exercises on Number Theory. —Miguel A. Lerma

Exercises
1. Show that the sum of two consecutive primes is never twice a prime.
2. Can the sum of the digits of a square be (a) 3, (b) 1977?

3. Prove that there are infinitely many prime numbers of the form 4n + 3.
4. Prove that the fraction (n3 + 2n)/(n4 + 3n2 + 1) is in lowest terms for every possible
integer n.
5. Let p(x) be a non-constant polynomial such that p(n) is an integer for every positive
integer n. Prove that p(n) is composite for infinitely many positive integers n. (This
proves that there is no polynomial yielding only prime numbers.)

6. Show that if a2 + b2 = c2 , then 3|ab.

1 1 1
7. Show that 1 + + + · · · + can never be an integer for n ≥ 2.
2 3 n
8. Let f (n) denote the sum of the digits of n. Let N = 44444444 . Find f (f (f (N ))).
9. Show that there exist 2020 consecutive numbers, each of which is divisible by a cube
greater than 1.
10. Find all triples of positive integers (a, b, c) such that

1 1 1
1+ 1+ 1+ = 2.
a b c
11. Find all positive integer solutions to abc − 2 = a + b + c.
12. (USAMO, 1979) Find all non-negative integral solutions (n1 , n2 , . . . , n14 ) to
n41 + n42 + · · · + n414 = 1599 .
13. The Fibonacci sequence 0, 1, 1, 2, 3, 5, 8, 13, . . . is defined by F0 = 0, F1 = 1, Fn =
1010
Fn−1 + Fn−2 for n ≥ 2. Prove that for some k > 0, Fk is a multiple of 1010 .
1
PUTNAM TRAINING NUMBER THEORY 2

14. Do there exist 2 irrational numbers a and b greater than 1 such that bam c =
6 bbn c for
every positive integers m, n?
15. The numbers 22005 and 52005 are written one after the other (in decimal notation).
How many digits are written altogether?
16. If p and p2 + 2, are primes show that p3 + 2 is prime.

17. Suppose n > 1 is an integer. Show that n4 + 4n is not prime.

√ √
18. Let m and n be positive integers such that m < b n + 12 c. Prove that m + 21 < n.
√
19. Prove that the function f (n) = bn + n + 1/2c (n = 1, 2, 3, . . . ) misses exactly the
squares.
20. Prove that there are no primes in the following infinite sequence of numbers:
1001, 1001001, 1001001001, 1001001001001, . . .
21. (Putnam 1975, A1.) For positive integers n define d(n) = n − m2 , where m is the
greatest integer with m2 ≤ n. Given a positive integer b0 , define a sequence bi by
taking bk+1 = bk + d(bk ). For what b0 do we have bi constant for sufficiently large i?

22. Let an = 10 + n2 for n ≥ 1. For each n, let dn denote the gcd of an and an+1 . Find
the maximum value of dn as n ranges through the positive integers.

23. Suppose that the positive integers x, y satisfy 2x2 + x = 3y 2 + y. Show that x − y,
2x + 2y + 1, 3x + 3y + 1 are all perfect squares.

24. If 2n + 1 and 3n + 1 are both perfect squares, prove that n is divisible by 40.
25. How many zeros does 1000! ends with?
100

26. For how many k is the binomial coefficient k
odd?

27. Let n be a positive integer. Suppose that 2n and 5n begin with the same digit. What
is the digit?

28. Prove that there are no four consecutive non-zero binomial coefficients nr , r+1n

,
n
n
r+2
, r+3 in arithmetic progression.

29. (Putnam 1995, A1) Show that every positive integer is a sum of one or more numbers
of the form 2r 3s , where r and s are nonnegative integers and no summand divides
another.

30. (Putnam 2003, A1) Let n be a fixed positive integer. How many ways are there to
write n as a sum of positive integers, n = a1 + a2 + · · · + ak , with k an arbitrary
positive integer and a1 ≤ a2 ≤ · · · ≤ ak ≤ a1 + 1? For example, with n = 4 there are
four ways: 4, 2+2, 1+1+2, 1+1+1+1.
PUTNAM TRAINING NUMBER THEORY 3

31. (Putnam 2001, B-1) Let n be an even positive integer. Write the numbers 1, 2, . . . , n2
in the squares of an n × n grid so that the kth row, from right to left is
(k − 1)n + 1, (k − 1)n + 2, . . . , (k − 1)n + n .
Color the squares of the grid so that half the squares in each row and in each column
are read and the other half are black (a chessboard coloring is one possibility). Prove
that for each such coloring, the sum of the numbers on the red squares is equal to the
sum of the numbers in the black squares.
32. How many primes among the positive integers, written as usual in base 10, are such
that their digits are alternating 1s and 0s, beginning and ending with 1?
33. Prove that if n is an integer greater than 1, then n does not divide 2n − 1.

34. The digital root of a number is the (single digit) value obtained by repeatedly adding
the (base 10) digits of the number, then the digits of the sum, and so on until obtaining
a single digit—e.g. the digital root of 65,536 is 7, because 6 + 5 + 5 + 3 + 6 = 25 and
2 + 5 = 7. Consider the sequence an = integer part of 10n π, i.e.,
a1 = 31 , a2 = 314 , a3 = 3141 , a4 = 31415 , a5 = 314159 , ...
and let bn be the sequence
a
a a 4
a 3 a 3
b 1 = a1 , b2 = aa12 , b3 = a1 2 , b4 = a1 2 , ...
Find the digital root of b106 .

35. (IMO, 1988) Let a, b positive integers such that ab + 1 divides a2 + b2 . Prove that
a2 + b 2
is the square of an integer.
ab + 1
36. Prove that there are no positive integers x, y, such that x2y + y2−x = x + y.
37. It is known that 13 + 23 + · · · + n3 = (1 + 2 + · · · + n)2 , hence 13 + 23 + · · · + n3 is a
multiple of 1 + 2 + · · · + n. This problem asks to prove a generalization of this fact.
Show that if n, m are positive integers and m is odd then 1m + 2m + · · · + nm is a
multiple of 1 + 2 + · · · + n.
PUTNAM TRAINING NUMBER THEORY 4

THIS PAGE LEFT BLANK INTENTIONALLY

PUTNAM TRAINING NUMBER THEORY 5

Hints
1. Contradiction.
2. If s is the sum of the digits of a number n, then n − s is divisible by 9.

3. Assume that there are finitely many primes of the form 4n + 3, call P their product,
and try to obtain a contradiction similar to the one in Euclid’s proof of the infinitude
of primes.
4. Prove that n3 + 2n and n4 + 3n2 + 1 are relatively prime.

5. Prove that p(k) divides p(p(k) + k).

6. Study the equation modulo 3.
7. Call the sum S and find the maximum power of 2 dividing each side of the equality

n
X n!
n!S = .
k=1
k

8. f (n) ≡ n (mod 9).

9. Chinese Remainder Theorem.

10. The minimum of a, b, c cannot be very large.

11. Try changing variables x = a + 1, y = b + 1, z = c + 1.
12. Study the equation modulo 16.

13. Use the Pigeonhole Principle to prove that the sequence of pairs (Fn , Fn+1 ) is eventu-
1010
ally periodic modulo N = 1010 .
√ √
14. Try a = 6, b = 3.
15. —

16. If p is an odd number not divisible by 3, then p2 ≡ ±1 (mod 6).

17. Sophie Germain’s identity: a4 + 4b4 = (a2 + 2b2 + 2ab)(a2 + 2b2 − 2ab).
√
18. The number n is irrational or an integer.
√
19. If m 6= bn + n + 1/2c, what can we say about m?
20. Each of the given numbers can be written pn (103 ), where pn (x) = 1 + x + x2 + · · · + xn ,
n = 1, 2, 3, . . . .
PUTNAM TRAINING NUMBER THEORY 6

21. Study the cases bk = perfect square, and bk = not a perfect square. What can we
deduce about bk+1 being or not being a perfect square in each case?
22. gcd(a, b) = gcd(a, b − a).

23. What is (x − y)(2x + 2y + 1) and (x − y)(3x + 3y + 1)?

24. Think modulo 5 and modulo 8.
25. Think of 1000! as a product of prime factors and count the number of 2’s and the
number of 5’s in it.

26. Find the exponent of 2 in the prime factorization of nk .

27. If N begins with digit a then a · 10k ≤ N < (a + 1) · 10k .

28. The desired sequence of binomial numbers must have a constant difference.

29. Induction. The base case is 1 = 20 30 . The induction step depends on the parity of n.
If n is even, divide by 2. If it is odd, subtract a suitable power of 3.

30. If 0 < k ≤ n, is there any such sum with exactly k terms? How many?
31. Interpret the grid as a ’sum’ of two grids, one with the terms of the form (k − 1)n,
and the other one with the terms of the form 1, . . . , n.
32. Each of the given numbers can be written pn (102 ), where pn (x) = 1 + x + x2 + · · · + xn ,
n = 1, 2, 3, . . . .
33. If n is prime Fermat’s Little Theorem yields the result. Otherwise let p be the smallest
prime divisor of n. . .
34. The digital root of a number is its reminder modulo 9. Then show that an1 (n =
1, 2, 3, · · · ) modulo 9 is periodic.
35. Vieta-jumping.

36. The given equation is equivalent to x(2y − 1) = y(1 − 2−x ).

37. Reason mod n and mod n + 1.
PUTNAM TRAINING NUMBER THEORY 7

Solutions
1. If p and q are consecutive primes and p + q = 2r, then r = (p + q)/2 and p < r < q,
but there are no primes between p and q.
2. (a) No, a square divisible by 3 is also divisible by 9.
(b) Same argument.
3. Assume that the set of primes of the form 4n + 3 is finite. Let P be their product.
Consider the number N = P 2 − 2. Note that the square of an odd number is of the
form 4n + 1, hence P 2 is of the form 4n + 1 and N will be of the form 4n + 3. Now,
if all prime factors of N where of the form 4n + 1, N would be of the form 4n + 1, so
N must have some prime factor p of the form 4n + 3. So it must be one of the primes
in the product P , hence p divides N − P 2 = 2, which is impossible.
4. That is equivalent to proving that n3 + 2n and n4 + 3n2 + 1 are relatively prime for
every n. These are two possible ways to show it:
- Assume a prime p divides n3 + 2n = n(n2 + 2). Then it must divide n or n2 + 2.
Writing n4 + 3n2 + 1 = n2 (n2 + 3) + 1 = (n2 + 1)(n2 + 2) − 1 we see that p cannot
divide n4 + 3n2 + 1 in either case.
- The following identity
(n2 + 1)(n4 + 3n2 + 1) − (n3 + 2n)2 = 1
(which can be checked algebraically) shows that any common factor of n4 + 3n2 + 1
and n3 + 2n should divide 1, so their gcd is always 1. (Note: if you are wondering
how I arrived to that identity, I just used the Euclidean algorithm on the two given
polynomials.)
5. Assume p(x) = a0 + a1 x + · · · + an xn , with an 6= 0. We will assume without loss of
generality that an > 0, so that p(k) > 0 for every k large enough—otherwise we can
use the argument below with −p(x) instead of p(x).
We have
n
X i
p(p(k) + k) = ai p(k) + k .
i=0
For each term of that sum we have that
i
ai p(k) + k = multiple of p(k) + ai k i ,

and the sum of the ai k i is precisely p(k), so p(p(k)+k) is a multiple of p(k). It remains
only to note that p(p(k) + k) 6= p(k) for infinitely many positive integers k, otherwise
p(p(x) + x) and p(x) would be the same polynomial, which is easily ruled out for non
constant p(x) (compare degrees).
6. For any integer n we have that n2 only can be 0 or 1 mod 3. So if 3 does not divide
a or b they must be 1 mod 3, and their sum will be 2 modulo 3, which cannot be a
square.
PUTNAM TRAINING NUMBER THEORY 8

7. Assume the sum S is an integer. Let 2i be the maximum power of 2 not larger than
n, and let 2j be the maximum power of 2 dividing n! Then
n
n! i X n!
2 S = .
2j k=1
k2j−i

For n ≥ 2 the left hand side is an even number. In the right hand side all the terms
of the sum are even integers except the one for k = 2i which is an odd integer, so the
sum must be odd. Hence we have an even number equal to an odd number, which is
impossible.
8. Since each digit cannot be greater than 9, we have that f (n) ≤ 9 · (1 + log10 n), so in
particular f (N ) ≤ 9 · (1 + 4444 · log10 4444) < 9 · (1 + 4444 · 4) = 159993. Next we
have f (f (N )) ≤ 9 · 6 = 54. Finally among numbers not greater than 54, the one with
the greatest sum of the digits is 49, hence f (f (f (N ))) ≤ 4 + 9 = 13.
Next we use that n ≡ f (n) (mod 9). Since 4444 ≡ 7 (mod 9), then
44444444 ≡ 74444 (mod 9) .
We notice that the sequence 7n mod 9 for n = 0, 1, 2, . . . is 1, 7, 4, 1, 7, 4, . . . , with
period 3. Since 4444 ≡ 1 (mod 3), we have 74444 ≡ 71 (mod 9), hence f (f (f (N ))) ≡ 7
(mod 9). The only positive integer not greater than 13 that is congruent with 7 modulo
9 is 7, hence f (f (f (N ))) = 7.
9. Pick 2020 different prime numbers p1 , p2 , . . . , p2020 (we can do that because the set of
prime numbers is infinite) and solve the following system of 2020 congruences:


 x ≡ 0 (mod p31 )
 x ≡ −1 (mod p32 )


x ≡ −2 (mod p33 )


 ...
(mod p32020 )

 x ≡ −2019

According to the Chinese Remainder Theorem, that system of congruences has a

solution x (modulo M = p31 . . . p32020 ). For k = 0, . . . , 2019 we have that x + k ≡ 0
(mod p3k+1 ), hence x + k is in fact a multiple of p3k+1 .
10. Assume a ≥ b ≥ c. Then
3
1 1 1 1
2= 1+ 1+ 1+ ≤ 1+ .
a b c c
From here we get that c < 4, so its only possible values are c = 1, 2, 3.
For c = 1 we get (1 + 1/c) = 2, hence

1 1
1+ 1+ = 1,
a b
which is impossible.
PUTNAM TRAINING NUMBER THEORY 9

For c = 2 we have (1 + 1/c) = 3/2, hence

1 1 4
1+ 1+ = ,
a b 3
and from here we get
3(b + 1)
a= ,
b−3
with solutions (a, b) = (15, 4), (9, 5) and (7, 6).
Finally for c = 3 we have 1 + 1/c = 1 + 1/3 = 4/3, hence

1 1 3
1+ 1+ = .
a b 2
So
2(b + 1)
a= .
b−2
The solutions are (a, b) = (8, 3) and (5, 4).
So the complete set of solutions verifying a ≥ b ≥ c are
(a, b, c) = (15, 4, 2), (9, 5, 2), (7, 6, 2), (8, 3, 3), (5, 4, 3) .
The rest of the triples verifying the given equation can be obtained by permutations
of a, b, c.
11. The change of variables x = a + 1, y = b + 1, z = c + 1, transforms the equation into
the following one:
1 1 1
+ + = 1.
x y z
Assuming x ≤ y ≤ z we have that x ≤ 3 ≤ z.
For x = 1 the equation becomes
1 1
+ = 0.
y z
which is impossible.
For x = 2 we have
1 1 1
+ = ,
y z 2
or
2y
z= ,
y−2
with solutions (y, z) = (3, 6) and (4, 4).
For x = 3 the only possibility is (y, z) = (3, 3).
So the list of solutions is
(x, y, z) = (2, 3, 6), (2, 4, 4), (3, 3, 3) ,
and the ones obtained by permuting x, y, z.
With the original variables the solutions are (except for permutations of variables);
(a, b, c) = (1, 2, 5), (1, 3, 3), (2, 2, 2) .
PUTNAM TRAINING NUMBER THEORY 10

12. We look at the equation modulo 16. First we notice that n4 ≡ 0 or 1 (mod 16)
depending on whether n is even or odd. On the other hand 1599 ≡ 15 (mod 16).
So the equation can be satisfied only if the number of odd terms in the LHS is 15
modulo 16, but that is impossible because there are only 14 terms in the LHS. Hence
the equation has no solution.
1010
13. Call N = 1010 , and consider the sequence an = remainder of dividing Fn by N .
Since there are only N 2 pairs of non-negative integers less than N , there must be
two identical pairs (ai , ai+1 ) = (aj , aj+1 ) for some 0 ≤ i < j. Let k = j − i. Since
an+2 = an+1 + an and an−1 = an+1 − an , by induction we get that an = an+k for every
n ≥ 0, so in particular ak = a0 = 0, and this implies that Fk is a multiple of N .
(In fact since there are N 2 + 1 pairs (ai , ai+1 ), for i = 0, 1, . . . , N 2 , we can add the
restriction 0 ≤ i < j ≤ N above and get that the result is true for some k such that
0 < k ≤ N 2 .)
√ √
14. The answer is affirmative. Let a = 6 and b = 3. Assume bam c = bbn c = k
for some positive integers m, n. Then, k 2 ≤ 6m < (k + 1)2 = k 2 + 2k + 1, and
k 2 ≤ 3n < (k + 1)2 = k 2 + 2k + 1. Hence, subtracting the inequalities and taking into
account that n > m:
2k ≥ |6m − 3n | = 3m |2m − 3n−m | ≥ 3m .
9m m
Hence ≤ k 2 ≤ 6m , which implies 41 ≤ 23 . This holds only for m = 1, 2, 3. This
4
values of m can be ruled out by checking the values of
bac = 2, ba2 c = 6, ba3 c = 14,
bbc = 1, bb2 c = 3, bb3 c = 5, bb4 c = 9, bb5 c = 15 .
Hence, bam c =
6 bbn c for every positive integers m, n.

15. There are integers k, r such that 10k < 22005 < 10k+1 and 10r < 52005 < 10r+1 . Hence
10k+r < 102005 < 10k+r+2 , k + r + 1 = 2005. Now the number of digits in 22005 is k + 1,
and the number of digits in 52005 is r + 1. Hence the total number of digits is 22005
and 52005 is k + r + 2 = 2006.
16. For p = 2, p2 + 2 = 6 is not prime.
For p = 3, p2 + 2 = 11, and p3 + 2 = 29 are all prime and the statement is true.
For prime p > 3 we have that p is an odd number not divisible by 3, so it is congruent
to ±1 modulo 6. Hence p2 + 2 ≡ 3 (mod 6) is multiple of 3 and cannot be prime.

17. If n is even then n4 + 4n is even and greater than 2, so it cannot be prime.

If n is odd, then n = 2k + 1 for some integer k, hence n4 + 4n = n4 + 4 · (2k )4 . Next,
use Sophie Germain’s identity: a4 + 4b4 = (a2 + 2b2 + 2ab)(a2 + 2b2 − 2ab).
√ √
18. From the hypothesis we have that m + 1 ≤ b n + 12 c ≤ n + 12 . But the second
√
inequality
√ must be strict because n is irrational or an integer, and consequently
n + 12 cannot be an integer. From here the desired result follows.
PUTNAM TRAINING NUMBER THEORY 11
√
19. Assume m 6= bn + n + 1/2c for every n = 1, 2, 3, . . . Then for some n, f (n) < m <
f (n + 1). The first inequality implies
√ 1
n + n + < m.
2
The second inequality implies m + 1 ≤ f (n + 1), and
√ 1
m+1<n+1+ n+1+
2
(Note that equality is impossible because the right hand side cannot be an integer.)
Hence
√ 1 √
n < m − n − < n + 1,
2
1
n < (m − n)2 − (m − n) + < n + 1
4
1 3
n − < (m − n)2 − (m − n) < n +
4 4
2
(m − n) − (m − n) = n .
m = (m − n)2 .
So, m is a square.
We are not done yet, since we still must prove that f (n) misses all the squares. To
do so we use a counting argument. Among all positive integers ≤√k 2 + k there are
exactly k squares, and exactly k 2 integers of the form f (n) = bn + n + 1/2c. Hence
f (n) is the nth non square.
Another way to express it: in the set A(k) = {1, 2, 3, . . . , k 2 + k} consider√the two
subsets S(k) = squares in A(k), and N (k) = integers of the form f (n) = bn+ n+1/2c
in A(k). The set S(k) has k elements, N (k) has k 2 elements, and A(k) = S(k) ∪ N (k).
Since
|S(k) ∪ N (k)| = |S(k)| + |N (k)| −|S(k) ∩ N (k)|
| {z } | {z } | {z }
k2 +k k k2
we get that |S(k) ∩ N (k)| = 0, i.e, S(k) ∩ N (k) must be empty.
20. Each of the given numbers can be written
1 + 1000 + 10002 + · · · + 1000n = pn (103 )
where pn (x) = 1 + x + x2 + · · · + xn , n = 1, 2, 3, . . . . We have (x − 1)pn (x) = xn+1 − 1.
If we set x = 103 , we get:
999 · pn (103 ) = 103(n+1) − 1 = (10n+1 − 1)(102(n+1) + 10n+1 + 1) .
If pn (103 ) were prime it should divide one of the factors on the RHS. It cannot divide
10n+1 − 1, because this factor is less than pn (103 ), so pn (103 ) must divide the other
factor. Hence 10n+1 − 1 must divide 999, but this is impossible for n > 2. In only
remains to check the cases n = 1 and n = 2. But 1001 = 7 · 11 · 13, and 1001001 =
3 · 333667, so they are not prime either.
PUTNAM TRAINING NUMBER THEORY 12

21. We will prove that the sequence is eventually constant if and only if b0 is a perfect
square.
The “if” part is trivial, because if bk is a perfect square then d(bk ) = 0, and bk+1 = bk .
For the “only if” part assume that bk is not a perfect square. Then, suppose that
r2 < bk < (r + 1)2 . Then, d(bk ) = bk − r2 is in the interval [1, 2r], so bk+1 = r2 + 2d(bk )
is greater than r2 but less than (r + 2)2 , and not equal to (r + 1)2 by parity. Thus bk+1
is also not a perfect square, and is greater than bk . So, if b0 is not a perfect square,
no bk is a perfect square and the sequence diverges to infinity.

22. The answer is 41. In fact, we have:

gcd(an , an+1 ) = gcd(an , an+1 − an ) = gcd(n2 + 10, 2n + 1) = · · ·
(since 2n + 1 is odd we can multiply the other argument by 4 without altering the
gcd)
· · · = gcd(4n2 + 40, 2n + 1) = gcd((2n + 1)(2n − 1) + 41, 2n + 1)
= gcd(41, 2n + 1) ≤ 41 .
The maximum value is attained e.g. at n = 20.
23. The given condition implies:
(x − y)(2x + 2y + 1) = y 2 .
Since the right hand side is a square, to prove that the two factors on the left hand
side are also squares it suffices to prove that they are relatively prime. In fact, if p if
a prime number dividing x − y then it divides y 2 and consequently it divides y. So p
also divides x, and x + y. But then it cannot divide 2x + 2y + 1.
An analogous reasoning works using the following relation, also implied the given
condition:
(x − y)(3x + 3y + 1) = x2 .
24. It suffices to prove that n is a multiple of 5 and 8, in other words, that n ≡ 0 (mod 5),
and n ≡ 0 (mod 8).
We first think modulo 5. Perfect squares can be congruent to 0, 1, or 4 modulo 5
only. We have:
2n + 1 ≡ 0 (mod 5) =⇒ n≡2 (mod 5)
2n + 1 ≡ 1 (mod 5) =⇒ n≡0 (mod 5)
2n + 1 ≡ 4 (mod 5) =⇒ n≡4 (mod 5)
3n + 1 ≡ 0 (mod 5) =⇒ n≡3 (mod 5)
3n + 1 ≡ 1 (mod 5) =⇒ n≡0 (mod 5)
3n + 1 ≡ 4 (mod 5) =⇒ n≡1 (mod 5) .
So the only possibility that can make both 2n + 1 and 3n + 1 perfect squares is n ≡ 0
(mod 5), i.e., n is a multiple of 5.
PUTNAM TRAINING NUMBER THEORY 13

Next, we think modulo 8. Perfect squares can only be congruent to 0, 1, or 4 modulo

8, and we have:
3n + 1 ≡ 0 (mod 8) =⇒ n≡5 (mod 8)
3n + 1 ≡ 1 (mod 8) =⇒ n≡0 (mod 8)
3n + 1 ≡ 4 (mod 8) =⇒ n≡1 (mod 8) .
The possibilities n ≡ 5 (mod 8) and n ≡ 1 (mod 8) can be ruled out because n must
be even. In fact, if 2n + 1 = a2 , then a is odd, and 2n = a2 − 1 = (a + 1)(a − 1).
Since a is odd we have that a − 1 and a + 1 are even, so 2n must be a multiple of 4,
consequently n is even. So, we have that the only possibility is n ≡ 0 (mod 8), i.e., n
is a multiple of 8.
Since n is a multiple of 5 and 8, it must be indeed a multiple of 40, QED.
25. The prime factorization of 1000! contains more 2’s than 5’s, so the number of zeros
at the end of 1000! will equal the exponent of 5. That will be equal to the number
of multiples of 5 in the sequence 1, 2, 3, . . . , 1000, plus the number of multiples of
52 = 25, plus the number of multiples of 53 = 125, and the multiples of 54 = 625, in
total:

1000 1000 1000 1000
+ + + = 200 + 40 + 8 + 1 = 249 .
5 25 125 625
So 1000! ends with 249 zeros.
26. The answer is 8.
More
generally, for any given positive integer n, the number of binomial coefficients
n
k
that are odd equals 2 raised to the number of 1’s in the binary representation
of n—so, for n = 100, with binary representation 1100100 (three 1’s), the answer is
23 = 8. We prove it by induction in the number s of 1’s in the binary representation
of n.
- Basis step: If s = 1, then n is a power of 2, say n = 2r . Next, we use that the
exponent of a prime number p in the prime factorization of m! is (Legendre’s formula):
X m
i
,
i≥1
p

where bxc = greatest integer ≤ x. Since nk = k!(n−k)!

n!

, the exponent of 2 in the prime
2r

factorization of k is
X 2r X k X 2r − k X r r
k X k
i
− i
− i
= i
− ,
i≥1
2 i≥1
2 i≥1
2 i=1
2 i=1
2i

where dxe = least integer ≥ x. The right hand side is the number of values of i in
the interval from 1 to r for which 2ki is not an integer. If k = 0 or k = 2r then the
r
expression is 0, i.e., 2k is odd. Otherwise, for 0 < k < 2r , the right hand side is
r
strictly positive (at least k/2r is not an integer), and in that case 2k is even. So the
PUTNAM TRAINING NUMBER THEORY 14
r
number of values of k for which 2k is odd is 2 = 21 . This sets the basis step of the
induction process.
- Induction step: Assume the statement is true for a given s ≥ 1, and assume that
the number of 1’s in the binary representation of n is s + 1, so n can be written
n = 2r + n0 , where 0 < n0 < 2r and n0 has s 1’s in its binary representation. By
n0

induction hypothesis the number of values of k for which k is odd is 2s . We must
r 0
of k for which nk = 2 +n

prove that the number of values k
is odd is 2s+1 . To do so
n
we will study the parity of k in three intervals, namely 0 ≤ k ≤ n0 , n0 < k < 2r , and

2r ≤ k ≤ n. 0
(1) For every k such that 0 ≤ k ≤ n0 , nk and nk have the same parity. In fact, using

again the above formula to determine the exponent of 2 in the prime factorization of
binomial coefficients, we get
X j n k X k X n − k X 2r + n0 X k
− − = −
i≥1
2i i≥1
2i i≥1
2i i≥1
2i i≥1
2i
X 2r + n0 − k
−
i≥1
2i
r 0 X r
X
r−i n k
= 2 + i −
i=1
2 i=1
2i
r 0
X
r−i n −k
− 2 +
i=1
2i
X n0 X k X n0 − k

= − − .
i≥1
2i i≥1
2i i≥1
2i

0 n

Hence, the number of values of k in the interval from 0 to n for which k
is odd is 2s .
(2) If n0 < k < 2r ,then nk is even. In fact, we have that the power of 2 in the prime

factorization of nk is:

Xk X n − k r j k
Xjnk X n k n−k
− − = − i − .
i≥1
2i i≥1
2i i≥1
2i i=1
2i 2 2i

using bxc + byc ≤ bx + yc, and given that n/2i = k/2i + (n − k)/2i , we see that all
terms of the sum on the right hand side are nonnegative, and all we have to show
is that at least one of them is strictly positive. That can be accomplished by taking
i = r. In fact, we have 2r < n < 2r+1 , hence 1 < n/2r < 2, bn/2r c = 1. Also,
0 < k < 2r , hence 0 < k/2r < 1, bk/2r c = 0. And 2r = n − n0 > n − k > 0, so
0 < (n − k)/2r < 1, b(n − k)/2r c = 0. Hence,
j n k k n − k
− r − = 1 − 0 − 0 = 1 > 0.
2r 2 2r
PUTNAM TRAINING NUMBER THEORY 15

(3) If 2r ≤ k ≤ n, then letting k 0 = n − k we have that 0 ≤ k 0 ≤ n0 , and nk = kn0 ,

and by (1), the number of values of k in the interval from 2r to n for which nk is odd
is 2s .
The three results (1), (2) and (3) combined show that the number of values of k for
which nk is odd is 2 · 2s = 2s+1 . This completes the induction step, and the result is
proved.

27. The answer is 3.

Note that 25 = 32, 55 = 3125, so 3 is in fact a solution. We will prove that it is the
only solution.
Let d be the common digit at the beginning of 2n and 5n . Then
d · 10r ≤ 2n < (d + 1) · 10r ,
d · 10s ≤ 5n < (d + 1) · 10s
for some integers r, s. Multiplying the inequalities we get
d2 10r+s ≤ 10n < (d + 1)2 10r+s ,
d2 ≤ 10n−r−s < (d + 1)2 ,
so d is such that between d2 and (d + 1)2 there must be a power of 10. The only
possible solutions are d = 1 and d = 3. The case d = 1 can be ruled out because that
would imply n = r + s, and from the inequalities above would get
5r ≤ 2s < 2 · 5r ,
2s ≤ 5r < 2 · 2s ,
hence 2s = 5r , which is impossible unless r = s = 0 (implying n = 0).
Hence, the only possibility is d = 3.
28. Assume that the given binomial coefficients are in arithmetic progression. Multiplying
each binomial number by (k + 3)!(n − k)! and simplifying we get that the following
numbers are also in arithmetic progression:
(k + 1)(k + 2)(k + 3),
(n − k)(k + 2)(k + 3),
(n − k)(n − k − 1)(k + 3),
(n − k)(n − k − 1)(n − k − 2) .
Their differences are
(n − 2k − 1)(k + 2)(k + 3),
(n − k)(n − 2k − 3)(k + 3),
(n − k)(n − k − 1)(n − 2k − 5) .
Writing that they must be equal we get a system of two equations:
( 2
n − 4kn − 5n + 4k 2 + 8k + 2 = 0
n2 − 4kn − 9n + 4k 2 + 16k + 14 = 0
PUTNAM TRAINING NUMBER THEORY 16

Subtracting both equations we get

4n − 8k − 12 = 0 ,
i.e., n = 2k + 3, so the four binomial numbers should be of the form

2k + 3 2k + 3 2k + 3 2k + 3
, , , .
k k+1 k+2 k+3
However
2k + 3 2k + 3 2k + 3 2k + 3
< = > ,
k k+1 k+2 k+3
so they cannot be in arithmetic progression.
- Remark: There are sets of threeconsecutive binomial numbers in arithmetic pro-
gression, e.g.: 71 = 7, 72 = 21, 73 = 35.

29. We proceed by induction, with base case 1 = 20 30 . Suppose all integers less than
n − 1 can be represented. If n is even, then we can take a representation of n/2 and
multiply each term by 2 to obtain a representation of n. If n is odd, put m = blog3 nc,
so that 3m ≤ n < 3m+1 . If 3m = n, we are done. Otherwise, choose a representation
(n − 3m )/2 = s1 + · · · + sk in the desired form. Then
n = 3m + 2s1 + · · · + 2sk ,
and clearly none of the 2si divide each other or 3m . Moreover, since 2si ≤ n − 3m <
3m+1 −3m , we have si < 3m , so 3m cannot divide 2si either. Thus n has a representation
of the desired form in all cases, completing the induction.

30. There are n such sums. More precisely, there is exactly one such sum with k terms
for each of k = 1, . . . , n (and clearly no others). To see this, note that if n = a1 +
a2 + · · · + ak with a1 ≤ a2 ≤ · · · ≤ ak ≤ a1 + 1, then
ka1 = a1 + a1 + · · · + a1
≤ n ≤ a1 + (a1 + 1) + · · · + (a1 + 1)
= ka1 + k − 1.
However, there is a unique integer a1 satisfying these inequalities, namely a1 = bn/kc.
Moreover, once a1 is fixed, there are k different possibilities for the sum a1 +a2 +· · ·+ak :
if i is the last integer such that ai = a1 , then the sum equals ka1 +(i−1). The possible
values of i are 1, . . . , k, and exactly one of these sums comes out equal to n, proving
our claim.

31. Let R (resp. B) denote the set of red (resp. black) squares in such a coloring, and
for s ∈ R ∪ B, let f (s)n + g(s) + 1 denote the number written in square s, where
0 ≤ f (s), g(s) ≤ n − 1. Then it is clear that the value of f (s) depends only on the
row of s, while the value of g(s) depends only on the column of s. Since every row
contains exactly n/2 elements of R and n/2 elements of B,
X X
f (s) = f (s).
s∈R s∈B
PUTNAM TRAINING NUMBER THEORY 17

Similarly, because every column contains exactly n/2 elements of R and n/2 elements
of B, X X
g(s) = g(s).
s∈R s∈B
It follows that
X X
f (s)n + g(s) + 1 = f (s)n + g(s) + 1,
s∈R s∈B

as desired.
32. The answer is only 101.
Each of the given numbers can be written
1 + 100 + 1002 + · · · + 100n = pn (102 ) ,
where pn (x) = 1 + x + x2 + · · · + xn , n = 1, 2, 3, . . . . We have (x − 1)Pn (x) = xn+1 − 1.
If we set x = 102 , we get
99 · pn (102 ) = 102(n+1) − 1 = (10n+1 − 1)(10n+1 + 1) .
If pn (102 ) is prime it must divide one of the factors of the RHS. It cannot divide
10n+1 − 1 because this factor is less than pn (102 ), so pn (102 ) must divide the other
factor. Hence 10n+1 − 1 must divide 99. This is impossible for n ≥ 2. In only remains
to check the case n = 1. In this case we have p1 (102 ) = 101, which is prime.
33. By contradiction. Assume n divides 2n − 1 (note that this implies that n is odd). Let
p be the smallest prime divisor of n, and let n = pk m, where p does not divide m.
Since n is odd we have that p 6= 2. By Fermat’s Little Theorem we have 2p−1 ≡ 1
k−1
(mod p). Also by Fermat’s Little Theorem, (2mp )p−1 ≡ 1 (mod p), hence 2n =
k k−1 k−1 k−1 k−1
2p m = (2p m )p = (2p m )p−1 · 2p m ≡ 2p m (mod p). Repeating the argument
k k−1 k−2
we get 2n = 2p m ≡ 2p m ≡ 2p m ≡ · · · ≡ 2m (mod p). Since by hypothesis 2n ≡ 1
(mod p), we have that 2m ≡ 1 (mod p).
Next, we use that if 2a ≡ 1 (mod p), and 2b ≡ 1 (mod p), then 2gcd(a,b) ≡ 1 (mod p).
If g = gcd(n, p − 1), then we must have 2g ≡ 1 (mod p). But since p is the smallest
prime divisor of n, and all prime divisors of p − 1 are less than p, we have that n
and p − 1 do not have common prime divisors, so g = 1, and consequently 2g = 2,
contradicting 2g ≡ 1 (mod p).
34. In spite of its apparent complexity this problem is very easy, because the digital root
of bn becomes a constant very quickly. First note that the digital root of a number a
is just the reminder r of a modulo 9, and the digital root of an will be the remainder
of rn modulo 9.
For a1 = 31 we have
digital root of a1 = digital root of 31 = 4 ;
digital root of a21 = digital root of 42 = 7;
digital root of a31 = digital root of 43 = 1;
digital root of a41 = digital root of 44 = 4;
PUTNAM TRAINING NUMBER THEORY 18

and from here on it repeats with period 3, so the digital root of an1 is 1, 4, and 7 for
remainder modulo 3 of n equal to 0, 1, and 2 respectively.
Next, we have a2 = 314 ≡ 2 (mod 3), a22 ≡ 22 ≡ 1 (mod 3), a32 ≡ 23 ≡ 2 (mod 3),
and repeating with period 2, so the reminder of an2 depends only on the parity of n,
with an2 ≡ 1 (mod 3) if n is even, and an2 ≡ 2 (mod 3) if n is odd.
And we are done because a3 is odd, and the exponent of a2 in the power tower defining
bn for every n ≥ 3 is odd, so the reminder modulo 3 of the exponent of a1 will be 2,
and the reminder modulo 9 of bn will be 7 for every n ≥ 3.
Hence, the answer is 7.
35. By contradiction. Assume there is some pair (a, b) of positive integers such that
a2 +b2
ab+1
= k, and k is not a square. Without loss of generality we may assume a ≥ b,
and also assume that among all pairs verifying the hypothesis, (a, b) is one in which
b is minimum. Then we have
a2 − abk + b2 − k = 0 .
This means that a is a root of the following quadratic equation in x:
x2 − bkx + b2 − k = 0 .
Since quadratic equations have two roots, let a0 be the other root of this equation.
Then we have a + a0 = bk, hence a0 is an integer, and aa0 = b2 − k. Since by hypothesis
02 +b2
k is not a square we have a0 6= 0. Also aa0 b+1 = k > 0 implies a0 > 0. Finally, since
2
a ≥ b we have a0 = b a−k < b, hence (b, a0 ) is another pair of positive integers verifying
the hypotheses except for the fact that b > a0 , contradicting the assumption that b
was minimum.
36. The given equation is equivalent to x(2y − 1) = y(1 − 2−x ). Since x, y are positive
integers we have 2y − 1 ≤ x(2y − 1) = y(1 − 2−x ) < y, hence 2y − 1 < y. But no
positive integer y satisfies this inequality.

37. Let Sn,m = 1m + 2m + · · · + nm . We have Sn,m = nk=0 k m = nk=0 (n − k)m , hence

P P

X n X n
2Sn,m = (k m + (n − k)m ) ≡ (k m + (−k)m ) ≡ 0 (mod n) ,
k=0 k=0

where we have used that for m odd (−k)m = −k m . Analogously, Sn,m = nk=1 k m =
P
P n m
k=1 (n + 1 − k) , hence
Xn n
X
m m
2Sn,m = (k + (n + 1 − k) ) ≡ (k m + (−k)m ) ≡ 0 (mod n + 1) .
k=1 k=1

So 2Sn,m is a multiple of both n and n + 1, and since gcd(n, n + 1) = 1 we have that

2Sn,m is a multiple of n(n + 1). We know that S1,m = n(n+1)2
, hence 2S1,m = n(n + 1),
consequently 2Sn,m is a multiple of 2S1,m , i.e., 2Sn,m = M · 2S1,m for some integer M .
Dividing both sides by 2 we get Sn,m = M · S1,m and the result follows.