Putnam 1997 (Problems and Solutions)

A1. A rectangle, HOM F , has sides HO = 11 and OM = 5. A triangle ABC

has H as the intersection of the altitudes, O the center of the circumscribed
circle, M the midpoint of BC, and F the foot of the altitude from A. What is
the length of BC?

Solution. In the figure below, let G be such that M G = 11, let x = BF = GC,
and let y = AH.

Since O is the circumcenter, we have y2 + 112 = 52 + (11 + x)2 . The slope of

the line AB is (5 + y) /x and the slope of HC is −5/ (22 + x). Since these lines
meet at right angles, the product of the slopes is −1. Thus, 5+y 5
x 22+x = 1 or
2
25 + 5y = 22x + x . Hence

y2 + 112 = 52 + (11 + x)2 = 25 + 112 + 22x + x2 = 25 + 112 + 25 + 5y,

or y 2 − 5y − 50 = 0 or (y − 10) (y + 5) = 0. Thus, y = 10, x = 3 and BC =

22 + 6 = 28.

1
A2. Players 1, 2, 3, ..., n are seated around a table and each has a single penny.
Player 1 passes a penny to Player 2, who then passes two pennies to Player 3.
Player 3 then passes one penny to player 4, who passes two pennies to Player 5,
and so on, players alternately passing one penny or two to the next player who
still has pennies. A player who runs out of pennies drops out of the game and
leaves the table. Find an infinite set of numbers n for which some player ends
up with all n pennies.

Solution. Suppose that at some point there are an odd number greater than
1 of remaining players, say p1 , p2 , ..., p2k+1 (k ≥ 1), none in danger of dropping
out on their next pass. Then there is an odd number of passes starting with
p1 ’s pass to p2 and ending with p2k+1 ’s pass to p1 . Thus, if p1 passes 2 coins
he will receive 2 coins, and if he passes 1 he will receive 1. The same is true
for any other player and the game will not terminate. Thus, we must avoid
this situation, or we are “stuck” with a nonterminating game. Suppose that to
begin with there are n = 2m (where for notational convenience, m ≥ 3) players,
say p1 , ..., p2m . Note that p1 drops out and p2 passes his two and drops out, p3
passes 1 and ends up with 2, p4 drops out, but p5 ends up with 2. Continuing,
p3 , p5 , ..., p2m−1 each have 2 coins and p2m drops out when he passes 2 to p3 who
now has 4. There are now m − 1 players and we are stuck if m − 1 is odd. Thus,
we need m − 1 = 2m1 for some m1 . Relabel the players q1 , ..., q2m1 . Now q1 has
4 coins, the rest have 2, and q1 starts this “stage” by passing 1 to q2 . Since there
are an even number (2m1 ) of players, q1 will eventually receive 2 (for a total of
5) and pass 1 again to q2 to begin the second “round” of the stage, but then q2
drops out, as do q4 , q6 , ..., q2m1 . When q2m1 passes 2 to q1 and drops out, q1 will
have 7 and the rest q3 , q5 , ..., q2m1 −1 each have 4. At the beginning of this new
stage, there are now m1 players left which must be even, or else m1 = 1 (i.e., we
have a winner) or we are stuck. Continuing, we see that the game terminates
if and only if m1 is a power of 2, since half of the players (those repeatedly
passing 2) eventually drop out at each stage which necessarily must have an
even number of players if not a winner. Observe that all the players passing 2
in a stage, have the same number of coins, and so they all drop out in the same
final round of the stage. Thus, a game with n = 2m players¡ terminates ¢ if and
only if m − 1 = 2m1 = 2k for some k ≥ 0, or n = 2m = 2 2k + 1 = 2h + 2,
h = 2, 3, 4, ... will yield terminating games. Of course, n = 2 and n = 4 games
also terminate. Incidentally, for n = 2h + 2, the original player p3 is the winner
since each stage ends with p3 receiving 2.

2
A3. Evaluate
Z ∞ µ ¶µ ¶
x3 x5 x7 x2 x4 x6
x− + − +··· 1 − 2 + 2 2 + 2 2 2 + · · · dx
0 2 2·4 2·4·6 2 2 ·4 2 ·4 ·6

1 2
Solution. The first series is xe− 2 x . For the second series, we use the fact that
for n = 0, 1, 2, ...,
Z π
1 · 3 · 5 · · · (n − 1) 1 · 2 · 3 · 4 · 5 · · · (2n − 1) 2n
cos2n θ dθ = π= π
0 2·4·6···n 22 · 42 · 62 · · · (2n)2
(2n)!
= π.
22 · 42 · 62 · · · (2n)2
Rπ
Then second series is (using 0
cos2n+1 θ dθ = 0),
∞ Rπ ∞ Rπ Z ∞
X 1
cos2n θ dθ 2n X 1
cosk θ 1 πX
(x cos θ)k
π 0
x = π 0
dθ xk = dθ
n=0
(2n)! n=0
k! π 0 n=0
k!
Z π
1
= ex cos θ dθ.
π 0

Thus, the desired integral is

Z ∞ Z π Z Z
1 21 1 π ∞ − 1 r2 r cos θ
xe− 2 x ex cos θ dθ dx = e 2 e r drdθ
0 π 0 π 0 0
Z ∞Z ∞ Z ∞ Z ∞
1 1
− 2 (x2 +y2 ) x 1 1 1
− 2 (x2 −2x+1)
1 2
= e e dxdy = e 2 e dx e− 2 y dy
π 0 −∞ π −∞ 0
Z Z ∞ ³ √ ´ √
1 1 ∞ − 1 (x−1)2 1 2 1 1 √
= e2 e 2 dx e− 2 y dy = e 2 2π 12 2π = e.
π −∞ 0 π

3
A4. Let G be a group with identity e and φ : G → G be a function such that
φ (g1 ) φ (g2 ) φ (g3 ) = φ (h1 ) φ (h2 ) φ (h3 )

whenever g1 g2 g3 = e = h1 h2 h3 . Prove that there exists an element a in G such

that ψ (x) = aφ (x) is a homomorphism (that is, ψ (xy) = ψ (x) ψ (y) for all x
and y in G).

Solution. Since x−1 xe = e3 , we have

¡ ¢ ¡ ¢ ¡ ¢−1
φ x−1 φ (x) φ (e) = φ (e)3 or φ x−1 φ (x) = φ (e)2 or φ x−1 = φ (x) φ (e)−2 .

Since yy −1 e = e3 , we have
¡ ¢ ¡ ¢ ¡ ¢−1
φ (y) φ y −1 φ (e) = φ (e)3 or φ (y) φ y −1 = φ (e)2 or φ y −1 = φ (e)−2 φ (y) .

Since x−1 (xy) y −1 = eee,we have

¡ ¢ ¡ ¢
φ x−1 φ (xy) φ y −1 = φ (e)3 ,

and so
¡ ¢−1 ¡ ¢−1
φ (xy) = φ x−1 φ (e)3 φ y−1 = φ (x) φ (e)−2 φ (e)3 φ (e)−2 φ (y)
= φ (x) φ (e)−1 φ (y) .

4
A5. Let Nn denote the number of ordered n-tuples of positive integers
(a1 , a2 , . . . , an ) such that 1/a1 + 1/a2 + . . . + 1/an = 1. Determine whether
N10 is even or odd.

Solution. Consider the involution of the set of solutions

(a1 , a2 , a3 , a4 , a5 , a6 , a7 , a8 , a9 , a10 ) 7→ (a2 , a1 , a4 , a3 , a6 , a5 , a8 , a7 , a10 , a9 )

The number of points that are not fixed is even. Thus, we need only to con-
sider the solutions of the form (a1 , a1 , a3 , a3 , a5 , a5 , a7 , a7 , a9 , a9 ). On this set of
remaining solutions, consider the involution

(a1 , a1 , a3 , a3 , a5 , a5 , a7 , a7 , a9 , a9 ) 7→ (a3 , a3 , a1 , a1 , a7 , a7 , a5 , a5 , a9 , a9 )

The points that are fixed are of the form

(a1 , a1 , a1 , a1 , a5 , a5 , a5 , a5 , a9 , a9 ).

By one more involution we need only consider the number of solutions of the
form

(a1 , a1 , a1 , a1 , a1 , a1 , a1 , a1 , a9 , a9 ).

For these a81 + a29 = 1, and so a9 = a2a 1

1 −8
. The set of possible pairs (a1 , a9 ) is
{(9, 18) , (10, 10) , (12, 6) , (16, 4), (24, 3)}. Thus, N10 is odd.

5
A6. For a positive integer n and any real number c, define xk recursively by
x0 = 0, x1 = 1, and for k ≥ 0,

cxk+1 − (n − k)xk
xk+2 = .
k+1
Fix n and then take c to be the largest value for which xn+1 = 0. Find xk in
terms of n and k, 1 ≤ k ≤ n.

Solution. The recurrence relation is as one might find while solving a differen-
tial equation using power series. We have

(k + 1) xk+2 = cxk+1 − (n − k)xk

∞
X ∞
X X∞ ∞
X
xk+2 (k + 1) tk = c k
xk+1 t − n k
xk t + kxk tk
k=0 k=0 k=0 k=0
X∞ X∞ ∞
X ∞
X
xk+2 (k + 1) tk = c xk+1 tk − (n − 1) xk tk + t2 (k − 1) xk tk−2
k=0 k=0 k=0 k=1
P∞ k
P∞ k+1
P∞
Let f (t) = k=0 xk+1 t = 1+ k=0 xk+2 t = k=1 xk tk−1 . Then the above
says

f 0 (t) = cf (t) − (n − 1) tf (t) + t2 f 0 (t)

f 0 (t) c − (n − 1)t B A
= = +
f (t) 1 − t2 1+t 1−t
1 1
(n − 1 + c) (n − 1 − c)
= 2 − 2 .
1+t 1−t
Hence
1 1
f (t) = (1 − t) 2 (n−1−c) (1 + t) 2 (n−1+c) .

Now f (t) will be a polynomial of degree n − 1 if 12 (n − 1 + c) and 12 (n − 1 − c)

are nonnegative
P integers, say j and n − 1 − j (i.e., j = 0, ..., n − 1). Since
f(t) = ∞ k 1
k=0 k+1 t , we would then have xn+1 = 0. Since j = 2 (n − 1 + c) ,
x
c = 2j − n + 1. Thus, some possible values for c for which xn+1 (n, c) = 0 are
the n values

−n + 1, −n + 3, ..., 2 (n − 1) − n + 1 = n − 1.

However, it is clear that xk (n, c) is a polynomial of degree at most k − 1 in c.

Indeed, x2 (n, c) = c, and from xk+2 = cxk+1 −(n−k)x
k+1
k
the result follows from
induction. Thus, the values the above values are the only values of c for which
xn+1 (n, c) = 0. The largest of these values is c = n − 1. Then
1 1
f (t) = (1 − t) 2 (n−1−c) (1 + t) 2 (n−1+c) = (1 + t)n−1 ,

6
 P∞ ¡ ¢
and f (t) = k=0 xk+1 t
k
⇒ xk+1 = n−1
k or


 0 k=0
 1 k =1
xk = ¡n−1¢ .

 k = 2, ..., n
 k−1
0 k>n

7
B1. Let {x} denote the distance between the real number x and the nearest
integer. For each positive integer n, evaluate
6n−1
X m m
Fn = min({ }, { }).
m=1
6n 3n

(Here min(a, b) denotes the minimum of a and b.)

Solution. We have
½ m
m 6n 1 ≤ m ≤ 3n
{ }= m
6n 1− 6n 3n ≤ m ≤ 6n − 1

 m

 1 ≤ m ≤ 3n/2
3n
m 
m
1 − 3n 3n/2 ≤ m ≤ 3n
{ }= m .
3n 
 − 1 3n ≤ m ≤ 3n 32 = 9n
 3n m 9n
2
2 − 3n 2 ≤ m ≤ 6n − 1

Now
 m

 ¡ m6n ¢ 1 ≤ m ≤ 3n/2
m m  m
min , 1 −
¡ 6n m m 3n ¢ 3n/2 ≤ m ≤ 3n
min({ }, { }) =
6n 3n 
 min ¡1 − ,
6n 3n − 1¢ 3n ≤ m ≤ 3n 32 = 9n
2
 m m 9n
min 1 − 6n , 2 − 3n 2 ≤ m ≤ 6n − 1

Using
m m
≤ 1− ⇔ m ≤ 2n,
6n 3n
m m
1− ≤ − 1 ⇔ m ≥ 4n
6n 3n
m m
1− ≤ 2− ⇔ m ≤ 6n,
6n 3n
we have
 m

 6n 1 ≤ m ≤ 3n/2 

 m m

 6n 3n/2 ≤ m ≤ 2n 
 1 ≤ m ≤ 2n − 1
m m  m  6n m
1 − 3n 2n ≤ m ≤ 3n 1 − 3n 2n ≤ m ≤ 3n − 1
min({ }, { }) = m = m
6n 3n 

 3n − 1 3n ≤ m ≤ 4n 
 − 1 3n ≤ m ≤ 4n − 1
 3n m

 1 m
− 6n 9
4n ≤ m ≤ 2 n 1 − 6n 4n ≤ m ≤ 6n − 1

 m 9
1 − 6n 2 n ≤ m ≤ 6n − 1

8
 Hence,
6n−1
X m m
Fn = min({ }, { })
m=1
6n 3n
2n−1
X m X ³
3n−1
m´ X ³m
4n−1 X ³
´ 6n−1 m´
= + 1− + −1 + 1−
m=1
6n m=2n 3n m=3n
3n m=4n
6n
Ã2n−1 3n−1 4n−1 6n−1
!
1 X X X X
= m−2 m+2 m− m
6n m=1 m=2n m=3n m=4n
+ (3n − 1 − 2n + 1) − (4n − 1 − 3n + 1) + (6n − 1 − 4n + 1)
1 ¡1 ¢
= 2 (2n) (2n − 1) − (5n − 1) n + (7n − 1) n − 12 (10n − 1) 2n + 2n
6n
= n.

9
B2. Let f be a twice-differentiable real-valued function satisfying

f(x) + f 00 (x) = −xg(x)f 0 (x),

where g(x) ≥ 0 for all real x. Prove that |f(x)| is bounded.

Solution. Multiplying by f 0 (x), we get

d ¡ ¢
1
2 f (x)2 + f 0 (x)2 = f(x)f 0 (x) + f 0 (x)f 00 (x) = −xg(x)f 0 (x)2 .
dx
Thus, f (x)2 + f 0 (x)2 is nondecreasing for x ≤ 0 and nonincreasing for x ≥ 0.
Hence

f(x)2 ≤ f (x)2 + f 0 (x)2 ≤ f (0)2 + f 0 (0)2 .

10
 Pn
B3. For each positive integer n, write the sum m=1 1/m in the form pn /qn ,
where pn and qn are relatively prime positive integers. Determine all n such
that 5 does not divide qn .
Pn
Solution. We may write the sum Sn := m=1 1/m in the form
n
X bn/5c bn/52 c
1 X 1 X
Sn = 1/m + 1/m + 2 1/m + ...
5 5
5-m 5-m 5-m
 
k
XK bn/5
X c K
1   X 1
=  1/m = F k ,
5k 5k bn/5 c
k=0 5-m k=0

where 5K is the largest power of 5 not greater than n and

M
X 1
FM := .
m
5-m

Consider a sum of reciprocals of integers n1 , ..., nk not divisible by 5, say

1 1 1 n2 ...nk + n1 n3 ...nk + ... + n1 n2 ...nk−1 N
+ + ... + = =
n1 n2 nk n1 n2 ...nk D
with D = n1 n2 ...nk . Let ni be the associate of ni ; i.e., ni ni ≡ 1 mod (5) . Note
that
D (n1 + n2 + ... + nk ) = n1 n2 ...nk (n1 + n2 + ... + nk )
≡ n2 ...nk + n1 n3 ...nk + ... + n1 n2 ...nk−1 = N (mod 5)
Thus,
D (n1 + n2 + ... + nk ) ≡ N (mod 5)
Since 5 - D we have
5|N ⇔ n1 + n2 + ... + nk ≡ 0 (mod 5)
Note that
1 = 1, 2 = 3, 3 = 2, 4 = 4
Working mod 5, we have
1 = 1,
1+2 = −1,
1+2+3 = 1,
1+2+3+4 = 0
1+2+3+4+6 = 1
1+2+3+4+6+7 = −1
1+2+3+4+6+7+8 = 1
1+2+3+4+6+7+8+9 = 0, etc..

11
 NM
Writing FM = D M
, we have NM ≡ 1, −1, 1, 0, 0 (mod 5) if M ≡ 1, 2, 3, 4, 5 (mod 5) ,
respectively. Note that if M ≡ 5, then FM = FM−1 and NM = NM−1 ≡
0 (mod 5) . Recall
K
X 1
Sn = F k (∗)
5k bn/5 c
k=0

Consider the last term

1
F K
5K bn/5 c
¥ ¦ ¥ ¦
in (∗) . We have n/5K = 1, 2, 3 or 4. If n/5K = 1, 2,or 3, then the numerator
of Fbn/5K c is not divisible by 5. In this case, Sn must have a factor of at least
5K in its denominator. Thus,¥K = 0¦ in this case and n = 1, 2,or 3.
Henceforth, suppose that n/5K = 4. In this case, we have F4 = 1 + 1/2 +
1/3 + 1/4 = 25 1 1
12 and 5K Fbn/5K c = 12·5K−2 . If K = 0, then S4 = F4 = 12
25

and n = 4 has the desired property. If K ≥ 1, the preceding term in (∗) is

1
5K−1 Fbn/5K−1 c , and

¥ ¦ 21 22 23 24 ¥ ¦
n/5K = 4 ⇒ n/5K = 4, , , , ⇒ n/5K−1 = 20, 21, ..., 24.
5 5 5 5
¥ ¦ 1
If n/5K−1 = 21, 22, 23, then 5K−1 Fbn/5K−1 c has numerator not divisible by
5 and the denominator has a factor of 5K−1 . As all other denominators in
the sum have lower powers of 5 (in particular the last term is 12·51K−2 ), in this
case
¥ Sn will
¦ have denominator divisible by¥ 5 unless ¦ K ≤ 1. Thus, in the case
n/5K−1
¥ = 21,
¦ 22, 23, we have n = bnc = n/5 K−1
= 21, 22, 23. Now suppose
that n/5 K−1
= 20, 24. We have F20 = 5173168 and F24 = 399698125
16456225
118982864 . As the
1
numerators of F20 and F24 are divisible by 25, the term 5K−1 Fbn/5K−1 c has
denominator with a power of 5 of at least K − 3. The sum of the two terms in
(∗) before and after this term is
µ ¶
1 1 1 1
F K−2 c + F K = Fbn/5K−2 c +
5K−2 bn/5 5K bn/5 c 5K−2 12

Since 12 ≡ 3 and the numerator of Fbn/5K−2 c can be only be congruent to ±1 or

0, the denominator of this sum has a factor of at least 5K−2 , and all of the other
terms have denominators with lower powers of 5. Thus, we must have K = 1
or 2. For K = 1, n = 20, 24 and for K = 2,we have bn/5c = 20, 24 ⇒ n =
100 − 104, 120 − 124. Thus, in view of all of the above, the only values for n
with the desired property are 1 − 4, 20 − 24, 100 − 104, and 120 − 124.

12
B4. Let am,n denote the coefficient of xn in the expansion of (1 + x + x2 )m .
Prove that for all k ≥ 0,
b 2k
3 c
X
0≤ (−1)i ak−i,i ≤ 1.
i=0

Solution. Note that since the degree of (1 + x + x2 )m is 2m, we have am,n = 0

for n > 2m. Hence ak−i,i = 0 for i > 2 (k − i) or 3i > 2k. Hence the upper limit
on the sum only serves to eliminate terms which would be 0 or undefined. Let
us define am,n = 0 for n < 0 or m < 0. Thus in the sum above and all sums
below, we may take the index to run over all integers.
X
am+1,n xn = (1 + x + x2 )m+1 = (1 + x + x2 )(1 + x + x2 )m
n
X X
= (1 + x + x2 ) am,k xk = am,k (xk + xk+1 + xk+2 )
k k
X X X
n n
= am,n x + am,n−1 x + am,n−2 xn
n n n
X
= (am,n + am,n−1 + am,n−2 ) xn .
n

Thus,
am+1,n = am,n + am,n−1 + am,n−2 .
Pb 2k
3 c
P
Let sk := (−1)i ak−i,i = i (−1)i ak−i,i . Then
i=0
X X
sk+1 : = (−1)i ak+1−i,i = (−1)i (ak−i,i + ak−i,i−1 + ak−i,i−2 )
i i
X¡ ¢
= (−1)i ak−i,i − (−1)i−1 a(k−1)−(i−1),i−1 + (−1)i−2 a(k−2)−(i−2),i−2
i
= sk − sk−1 + sk−2
We have
X X X
s−2 = (−1)i a−2−i,i = 0, s−1 = (−1)i a−1−i,i = 0, s0 = (−1)i a0−i,i = a0,0 = 1
i i i

Thus,
s1 = s0 − s−1 + s−2 = 1
s2 = s1 − s0 + s−1 = 0
s3 = s2 − s1 + s0 = 0
s4 = s3 − s2 + s1 = 1
s5 = s4 − s3 + s2 = 1

13
Assume that s4k = 1, s4k+1 = 1, s4k+2 = 0, s4k+3 = 0. Then

s4(k+1) = s4k+4 = s4k+3 − s4k+2 + s4k+1 = 0 − 0 + 1 = 1,

s4(k+1)+1 = s4k+5 = s4k+4 − s4k+3 + s4k+2 = 1 − 0 + 0 = 1,
s4(k+1)+2 = s4k+6 = s4k+5 − s4k+4 + s4k+3 = 1 − 1 + 0 = 0, and
s4(k+1)+3 = s4k+7 = s4k+6 − s4k+5 + s4k+4 = 0 − 1 + 1 = 0.

Thus, sk is 0 or 1 for all k ≥ 0, and in particular 0 ≤ sk ≤ 1.

14
B5. Prove that for n ≥ 2,
···2
o ···2
o
22 n ≡ 22 n − 1 (mod n)

Solution. Let t1 := 2 and let tn := 2tn−1 . We are to show tn = 2tn−1 ≡

tn−1 (mod n) . We first check the result in the case n = 2k . Note that
¡ ¢ ¡ ¢
tn−2 ≥ 2k ⇒ 2tn−1 ≡ 2tn−2 mod 2k ⇐⇒ tn ≡ tn−1 mod 2k

Thus, for the case n = 2k , it suffices to check that t2k −2 ≥ 2k . This is true if
k = 1. If it is true for k ≥ 1, then
k
t2k+1 −2 > t2k −1 = 2t2k −2 ≥ 22 ≥ 2k+1 ,

and so t2k −2 ≥ 2k by induction. We handle the case n = 2k d where d is odd,

by using induction on d. The case d = 1 has been done. First note that since
(d, 2) = 1,

2ϕ(d) ≡ 1 (mod d) .

Thus, we have the key observation that

tm−2 ≡ tm−1 (mod ϕ (d)) ⇒ 2tm−1 −tm−2 ≡ 1 (mod d)

⇒ 2tm−2 ≡ 2tm−1 (mod d) ⇒ tm−1 ≡ tm (mod d) .

Now
 ¡ ¢
¡ ¢  tn−1 ≡ tn mod 2k
tn−1 ≡ tn mod 2k d ⇔ and

tn−1 ≡ tn (mod d)
¡ ¢
Since n ≥ 2k , we have already shown tn−1 ≡ tn mod 2k . Thus, we need only
show that tn−1 ≡ tn (mod d) and by the key observation it suffices to show that
tn−2 ≡ tn−1 (mod ϕ (d)) . We have
0
d − 1 ≥ n0 := ϕ (d) = 2k d0 ,

where d0 is odd. Since d0 < d, by induction we have

tn0 −1 ≡ tn0 (mod n0 ) or tn0 −1 ≡ tn0 (mod ϕ (d)) .

However, we need tn−2 ≡ tn−1 (mod ϕ (d)) . Since n − 1 ≥ d − 1 ≥ n0 , we

have n − 2 ≥ n0 − 1. Hence, if we could show the stronger result that not only
tn−1 ≡ tn (mod n) but also

tn−1 ≡ tn ≡ tn+1 ≡ tn+2 ≡ · · · (mod n) ,

15
(ad infinitum), then we would have

tn0 −1 ≡ tn0 ≡ · · · ≡ tn−1 ≡ tn ≡ tn+1 ≡ · · · (mod ϕ (d))

The stronger result tn−1 ≡ tn ≡ tn+1 ≡ tn+2 ≡ · · · (mod n) is true in the base
case n = 2k . Then assuming it holds for d0 < d inductively, we have

tn0 −1 ≡ tn0 ≡ · · · ≡ tn−1 ≡ tn ≡ tn+1 ≡ · · · (mod ϕ (d))

in which case by the key observation,

tn0 ≡ · · · ≡ tn−1 ≡ tn ≡ tn+1 ≡ · · · (mod d) .

We already know that

¡ ¢
tn ≡ tn+1 ≡ · · · mod 2k .

Thus,
¡ ¢
tn ≡ tn+1 ≡ tn+2 ≡ · · · mod 2k d ,

and we have proven this stronger result by induction.

16
B6. The dissection of the 3—4—5 triangle shown below (into four congruent
right triangles similar to the original) has diameter 5/2. Find the least diameter
of a dissection of this triangle into four parts. (The diameter of a dissection is
the least upper bound of the distances between pairs of points belonging to the
same part.)

Solution. Lower bounds on the least diameter of a dissection of the triangle into
4 parts can be obtained by selecting 5 points and taking the smallest distance
between pairs. Indeed, any dissection into four parts must have one region
containing 2 of the 5 points and that region has diameter at least the distance
between the two points. Thus, optimistically, we seek 5 points so that the
minimum distance between pairs is as large as possible. Then we attempt to find
a dissection where no region has diameter larger than this minimum distance.
Let A = (0, 4) ,B = (0, 0),C = (3,¡ 0). The ¢ point on
¡ the segment ¢ AC which is
distance t from (3, 0) is 15 t (0, 4) + 1 − 15 t (3, 0) = 3 − 35 t, 45 t . The square of
the distance from this point to the point (0, 4 − t) on AB (at distance t from A)
is
°µ ¶ °2
° °
° 3 − 3 t, 4 t − (0, 4 − t)° = 25 − 18t + 18 t2
° 5 5 ° 5

Setting this equal to t2 yields

18 2 25
25 − 18t + t = t2 ⇒ t = , 5
5 13
25
¡ ¢
Thus, A is¡ distance
¡ ¢13 from
¡ D:=¡ 0, 27
¢¢ 13 ¢ ≈ (0, 2.0769) which is at distance 13
25
3 25 4 25 24 20
from E:= 3 − 5 13 , 5 13 = 13 , 13 ≈ (1.8462, 1.5385). It is easy to verify
that 25
13 is the minimum distance between pairs of the five points A,B,C,D,E.
Thus, 2513 is a lower bound on the diameter of any dissection of ABC ¡ 3 into
¢ 4
regions.
¡3 ¢ We try to show that this lower bound is realized. Let F = 2 , 0 and
25
G = 2 , h where h > 0 is chosen so that BG = CG = 13 .
µ ¶2 µ ¶2
39 25
+ h2 =
26 13
1
√ ¡ 1√ ¢ ¡ 1√ ¢
We find h = ¡26 979.
¡ Thus,¢ G =¡ 32 , 26 ¢¢ 979 ¡ ≈ (1.5,¢ 1.2034) . Let H = 0, 26 979
and let I = 3 − 35 5 − 25
13 , 4
5 5 − 25
13 = 15 32
,
13 13 ≈ (1.1538, 2.4615) be the
point on AC at distance 25 25
13 from A. Then BG = 13 , CG = 13 .
25

17
 Now, the triangle ADI and the two quadrilaterals BFGH and FGEC clearly
have diameter 25 25
13 . For the pentagon HGEID, we need to check that HE ≤ 13 ,
but the height of E is less than that of the midpoint of DH since
µ ¶
20 1 1√ 27
1.538 5 ≈ <2 979 + ≈ 1.6402.
13 26 13
25
Thus, all 4 regions have diameter equal to the lower bound 13 .

18