Solutions To The 84th William Lowell Putnam Mathematical Competition Saturday, December 2, 2023
Solutions To The 84th William Lowell Putnam Mathematical Competition Saturday, December 2, 2023
Solutions To The 84th William Lowell Putnam Mathematical Competition Saturday, December 2, 2023
A1 If we use the product rule to calculate fn′′ (x), the result Now define the function h : [0, r) → (−π/2, π/2) by
is a sum of terms of two types: terms where two distinct h(x) = tan−1 (g(x)/ f (x)). We compute that
factors cos(m1 x) and cos(m2 x) have each been differen-
f (x)g′ (x) − g(x) f ′ (x)
tiated once, and terms where a single factor cos(mx) has h′ (x) =
been differentiated twice. When we evaluate at x = 0, f (x)2 + g(x)2
all terms of the first type vanish since sin(0) = 0, while and thus
the term of the second type involving (cos(mx))′′ be-
comes −m2 . Thus | f (x)||g′ (x)| + |g(x)|| f ′ (x)| | f (x)|2 + |g(x)|2
|h′ (x)| ≤ ≤ = 1.
f (x)2 + g(x)2 f (x)2 + g(x)2
n
n(n + 1)(2n + 1)
| fn′′ (0)| = − ∑ m2 = . Since h(0) = 0, we have |h(x)| ≤ x < r for all x ∈ [0, r).
m=1 6 Since r < π/2 and tan−1 is increasing on (−r, r), we
conclude that |g(x)/ f (x)| is uniformly bounded above
The function g(n) = n(n+1)(2n+1)
6 is increasing for n ∈ N by tan r for all x ∈ [0, r). But this contradicts the fact
and satisfies g(17) = 1785 and g(18) = 2109. It follows that f (r) = 0 and g(r) ̸= 0, since limx→r− g(x)/ f (x) =
that the answer is n = 18. ∞. This contradiction shows that r < π/2 cannot be
achieved.
A2 The only other real numbers with this property are
±1/n!. (Note that these are indeed other values than A4 The assumption that all vertices of the icosahedron cor-
±1, . . . , ±n because n > 1.) respond to vectors of the same length forces the center
of the icosahedron to lie at the origin, since the icosa-
Define the polynomial q(x) = x2n+2 − x2n p(1/x) = hedron is inscribed in a unique sphere. Since scaling
x2n+2 − (a0 x2n + · · · + a2n−1 x + 1). The statement that the icosahedron does not change whether or not the
p(1/x) = x2 is equivalent (for x ̸= 0) to the state- stated conclusion is true, we may choose coordinates so
ment that x is a root of q(x). Thus we know that that the vertices are the cyclic permutations of the vec-
√
±1, ±2, . . . , ±n are roots of q(x), and we can write 1+ 5
tors (±1, ±φ , 0) where φ = 2 is the golden ratio.
The subgroup of R3 generated by these vectors equals
q(x) = (x2 + ax + b)(x2 − 1)(x2 − 4) · · · (x2 − n2 )
G × G × G where G is the subgroup of R generated by
1 and φ . Since φ is irrational, it generates a dense sub-
for some monic quadratic polynomial x2 + ax + b.
group of R/Z; hence G is dense in R, and so G × G × G
Equating the coefficients of x2n+1 and x0 on both sides
is dense in R3 , proving the claim.
gives 0 = a and −1 = (−1)n (n!)2 b, respectively. Since
n is even, we have x2 + ax + b = x2 − (n!)−2 . We con- A5 The complex numbers z with this property are
clude that there are precisely two other real numbers x √
such that p(1/x) = x2 , and they are ±1/n!. 31010 − 1 31010 − 1 91010 − 1
− and − ± i.
2 2 4
A3 The answer is r = π2 , which manifestly is achieved by
setting f (x) = cos x and g(x) = sin x. We begin by noting that for n ≥ 1, we have the follow-
ing equality of polynomials in a parameter x:
Suppose by way of contradiction that there exist some
f , g satisfying the stated conditions for some 0 < r < 3n −1 n−1
j j
π
2 . We first note that we can assume that f (x) ̸= 0 for ∑ (−2) f (k) xk = ∏ (x2·3 − 2x3 + 1).
x ∈ [0, r). Indeed, by continuity, {x | x ≥ 0 and f (x) = k=0 j=0
0} is a closed subset of [0, ∞) and thus has a minimum This is readily shown by induction on n, using the fact
element r′ with 0 < r′ ≤ r. After replacing r by r′ , we that for 0 ≤ k ≤ 3n−1 − 1, f (3n−1 + k) = f (k) + 1 and
now have f (x) ̸= 0 for x ∈ [0, r). f (2 · 3n−1 + k) = f (k).
Next we note that f (r) = 0 implies g(r) ̸= 0. In- Now define a “shift” operator S on polynomials in z by
deed, define the function k : R → R by k(x) = f (x)2 + S(p(z)) = p(z + 1); then we can define Sm for all m ∈ Z
g(x)2 . Then |k′ (x)| = 2| f (x) f ′ (x) + g(x)g′ (x))| ≤ by Sm (p(z)), and in particular S0 = I is the identity map.
4| f (x)g(x)| ≤ 2k(x), where the last inequality fol- Write
lows from the AM-GM inequality. It follows that 3n −1
d
dx (log k(x)) ≤ 2 for x ∈ [0, r); since k(x) is continu- pn (z) := ∑ (−2) f (k) (z + k)2n+3
ous at x = r, we conclude that k(r) ̸= 0. k=0
2
for n ≥ 1; it follows that Thereafter, as long as Alice chooses j on the j-th turn
(for j ≥ 3 odd), either j + 1 < k, in which case Bob can
n−1
j
pn (z) = ∏ (S2·3 − 2S3 + I)z2n+3
j choose j + 1 to keep the number of fixed points odd; or
j=0 j + 1 = k, in which case k is even and Bob can choose 1
n−1
to transpose into the strategy for n − k (with no moves
n −1)/2 j j
= S(3 − 2I + S−3 )z2n+3 . made).
∏ (S3
j=0 Otherwise, at some odd turn j, Alice does not choose j.
At this point, the number of fixed points is odd, and
Next observe that for any ℓ, the operator Sℓ − 2I + S−ℓ on each subsequent turn Bob can ensure that neither
acts on polynomials in z in a way that decreases degree his own move nor Alice’s next move does not create
by 2. More precisely, for m ≥ 0, we have a fixed point: on any turn j for Bob, if j + 1 is available
Bob chooses it; otherwise, Bob has at least two choices
(Sℓ − 2I + S−ℓ )zm = (z + ℓ)m − 2zm + (z − ℓ)m
available, so he can choose a value other than j.
m 2 m−2 m 4 m−4
=2 ℓ z +2 ℓ z + O(zm−6 ). B1 The number of such configurations is m+n−2
2 4 m−1 .
Initially the unoccupied squares form a path from (1, n)
We use this general calculation to establish the follow- to (m, 1) consisting of m − 1 horizontal steps and n − 1
ing: for any 1 ≤ i ≤ n, there is a nonzero constant Ci vertical steps, and every move preserves this property.
(depending on n and i but not z) such that This yields an injective map from the set of reachable
i configurations to the set of paths of this form.
n− j n− j
∏ (S3 − 2I + S−3 )z2n+3 Since the number of such paths is evidently m+n−2
m−1 (as
j=1 one can arrange the horizontal and vertical steps in any
= Ci z2n+3−2i + (2n+3−2i)(n+1−i)
6 (∑ij=1 9n− j )z2n+1−2i order), it will suffice to show that the map we just wrote
down is also surjective; that is, that one can reach any
+O(z2n−1−2i ). (1) path of this form by a sequence of moves.
Proving (1) is a straightforward induction on i: the This is easiest to see by working backwards. Ending
n−i−1 n−i−1 at a given path, if this path is not the initial path, then
induction step applies S3 − 2I + S−3 to the it contains at least one sequence of squares of the form
right hand side of (1), using the general formula for (i, j) → (i, j − 1) → (i + 1, j − 1). In this case the square
(Sℓ − 2I + S−ℓ )zm . (i + 1, j) must be occupied, so we can undo a move by
Now setting i = n in (1), we find that for some Cn , replacing this sequence with (i, j) → (i + 1, j) → (i +
1, j − 1).
n−1
9n − 1
3j −3 j 2n+3 3
∏ (S − 2I + S )z = Cn z + z . B2 The minimum is 3.
j=0 16
First solution.
√
The roots of this polynomial are 0 and ± 94−1 i, and it
n
We record the factorization 2023 = 7 · 172 . We first rule
follows that the roots of pn (z) are these three numbers out k(n) = 1 and k(n) = 2. If k(n) = 1, then 2023n = 2a
n
minus 3 2−1 . In particular, when n = 1010, we find that for some a, which clearly cannot happen. If k(n) = 2,
the roots of p1010 (z) are as indicated above. then 2023n = 2a + 2b = 2b (1 + 2a−b ) for some a > b.
Then 1 + 2a−b ≡ 0 (mod 7); but −1 is not a power of 2
A6 (Communicated by Kai Wang) For all n, Bob has a win- mod 7 since every power of 2 is congruent to either 1,
ning strategy. Note that we can interpret the game play 2, or 4 (mod 7).
as building a permutation of {1, . . . , n}, and the number We now show that there is an n such that k(n) = 3.
of times an integer k is chosen on the k-th turn is exactly It suffices to find a > b > 0 such that 2023 divides
the number of fixed points of this permutation. 2a + 2b + 1. First note that 22 + 21 + 1 = 7 and 23 ≡ 1
For n even, Bob selects the goal “even”. Divide (mod 7); thus if a ≡ 2 (mod 3) and b ≡ 1 (mod 3)
{1, . . . , n} into the pairs {1, 2}, {3, 4}, . . . ; each time Al- then 7 divides 2a + 2b + 1. Next, 28 + 25 + 1 = 172 and
ice chooses an integer, Bob follows suit with the other 216·17 ≡ 1 (mod 7) by Euler’s Theorem; thus if a ≡ 8
integer in the same pair. For each pair {2k − 1, 2k}, we (mod 16 · 17) and b ≡ 5 (mod 16 · 17) then 172 divides
see that 2k − 1 is a fixed point if and only if 2k is, so the 2a + 2b + 1.
number of fixed points is even. We have reduced the problem to finding a, b such that
For n odd, Bob selects the goal “odd”. On the first turn, a ≡ 2 (mod 3), a ≡ 8 (mod 16 · 17), b ≡ 1 (mod 3),
if Alice chooses 1 or 2, then Bob chooses the other one b ≡ 5 (mod 16 · 17). But by the Chinese Remainder
to transpose into the strategy for n − 2 (with no moves Theorem, integers a and b solving these equations exist
made). We may thus assume hereafter that Alice’s first and are unique mod 3 · 16 · 17. Thus we can find a, b sat-
move is some k > 2, which Bob counters with 2; at this isfying these congruences; by adding appropriate mul-
point there is exactly one fixed point. tiples of 3 · 16 · 17, we can also ensure that a > b > 1.
3
Second solution. We rule out k(n) ≤ 2 as in the first (s01 , . . . , s0n+1 ). Observe that there cannot be another
solution. To force k(n) = 3, we first note that 24 ≡ −1 (s1 , . . . , sn′ +1 ) with the same sum, ∑nk=1
′ +1
sk = T0 , satis-
(mod 17) and deduce that 268 ≡ −1 (mod 172 ). (By fying g(s1 , . . . , sn′ +1 ) > 4045; otherwise, the function f
writing 268 = ((24 + 1) − 1)17 and expanding the bino- for (s1 , . . . , sn′ +1 ) would satisfy f (t0 + T0 ) > 4045 and
mial, we obtain −1 plus some terms each of which is there would be some T < T0 such that f (t0 + T ) = 4045
divisible by 17.) Since (28 − 1)2 is divisible by 172 , by the intermediate value theorem.
0 ≡ 216 − 2 · 28 + 1 ≡ 216 + 2 · 268 · 28 + 1 We claim that s0n+1 ≥ 1 and s0k = 1 for 1 ≤ k ≤ n. If
s0n+1 < 1 then
= 277 + 216 + 1 (mod 172 ).
g(s01 , . . . , s0n−1 , s0n + s0n+1 ) − g(s01 , . . . , s0n−1 , s0n , s0n+1 )
On the other hand, since 23 ≡ −1 (mod 7),
= s0n+1 (2ns0n − s0n+1 ) > 0,
277 + 216 + 1 ≡ 22 + 21 + 1 ≡ 0 (mod 7).
contradicting our observation from the previous para-
Hence n = (277 + 216 + 1)/2023 is an integer with graph. Thus s0n+1 ≥ 1. If s0k > 1 for some 1 ≤ k ≤ n then
k(n) = 3. replacing (s0k , s0n+1 ) by (1, s0n+1 + s0k − 1) increases g:
Remark. A short numerical calculation shows that the
value of n with k(n) = 3 found in the second solution is g(s01 , . . . , 1, . . . , s0n+1 + s0k − 1) − g(s01 , . . . , s0k , . . . , s0n+1 )
the smallest possible. = (s0k − 1)((n + 1 − k)(s0k + 1) + 2(n + 1)(s0n+1 − 1)) > 0,
B3 The expected value is (2n + 2)/3. again contradicting the observation. This establishes
If we divide the sequence X1 , . . . , Xn into increasing and the claim.
decreasing segments, then a(X1 , . . . , Xn ) can be com- Given that s0k = 1 for 1 ≤ k ≤ n, we have T = s0n+1 + n
puted as 1 plus the number of segments. For n ≥ 3, and
we can compare the number of segments in the se-
quences X1 , . . . , Xn and X2 , . . . , Xn by examining the n(n + 1)
g(s01 , . . . , s0n+1 ) = + (n + 1)(T − n)2 .
terms X1 , X2 , X3 : if these three terms form a monotone 2
sequence then the numbers of segments coincides, and
Setting this equal to 4045 and solving for T yields
otherwise they differ by 1.
r
Let E(n) denote the desired expected value. Since the 4045 n
six possible orderings of X1 , X2 , X3 are equally likely, T = n+ − .
n+1 2
the previous analysis shows that E(n) = E(n−1)+2/3.
We also have E(2) = 2 because any sequence of two For n = 9 this yields T = 29; it thus suffices to show
distinct elements is a zigzag. By induction on n, we that for all n,
obtain that E(n) = (2n + 2)/3 as claimed. r
4045 n
B4 The minimum value of T is 29. n+ − ≥ 29.
n+1 2
Write tn+1 = t0 + T and define sk = tk −tk−1 for 1 ≤ k ≤
n + 1. On [tk−1 ,tk ], we have f ′ (t) = k(t − tk−1 ) and so This is evident for n ≥ 30. For n ≤ 29, rewrite the claim
as
f (tk ) − f (tk−1 ) = 2k s2k . Thus if we define
r
4045 n
n+1 − ≥ 29 − n;
g(s1 , . . . , sn+1 ) = ∑ ks2k , n+1 2
k=1
we then obtain an equivalent inequality by squaring
then we want to minimize ∑n+1 both sides:
k=1 sk = T (for all pos-
sible values of n) subject to the constraints that 4045 n
g(s1 , . . . , sn+1 ) = 4045 and sk ≥ 1 for k ≤ n. − ≥ n2 − 58n + 841.
n+1 2
We first note that a minimum value for T is in-
deed achieved. To see this, note that the constraints Clearing denominators, gathering all terms to one side,
g(s1 , . . . , sn+1 ) = 4045 and sk ≥ 1 place an upper bound and factoring puts this in the form
on n. For fixed n, the constraint g(s1 , . . . , sn+1 ) = 4045
95
places an upper bound on each sk , whence the set of (9 − n)(n2 − n + 356) ≥ 0.
(s1 , . . . , sn+1 ) on which we want to minimize ∑ sk is a 2
compact subset of Rn+1 . The quadratic factor Q(n) has a minimum at 95 4 = 23.75
n+1
Now say that T0 is the minimum value of ∑k=1 sk (over and satisfies Q(8) = 40, Q(10) = −19; it is thus positive
all n and s1 , . . . , sn+1 ), achieved by (s1 , . . . , sn+1 ) = for n ≤ 8 and negative for 10 ≤ n ≤ 29.
4
B5 The desired property holds if and only if n = 1 or n ≡ 2 – If n = 2m is even, then Sm j = 3 for j = 1, m, cor-
(mod 4). responding to (a, b) = (2, 0), (1, 2nj ), (0, nj ).
Let σn,m be the permutation of Z/nZ induced by mul- – For n2 < i ≤ n, Si j = #(Z ∩ { n−i n
j , j }), correspond-
tiplication by m; the original problem asks for which n ing to (a, b) = (1, n−i n
does σn,m always have a square root. For n = 1, σn,m is j ), (0, j ).
the identity permutation and hence has a square root. Let T be the matrix obtained from S by performing row
We next identify when a general permutation admits a and column operations as follows: for d = 1, . . . , n − 2,
square root. subtract Snd times row n − 1 from row d and subtract
Snd times column n − 1 from column d. Evidently T is
Lemma 1. A permutation σ in Sn can be written as the square again symmetric and det(T ) = det(S). We have:
of another permutation if and only if for every even positive
Ti j = Si j − Sin S(n−1) j − Sn j Si(n−1) (i, j ≤ n − 2),
integer m, the number of cycles of length m in σ is even.
T(n−1) j = S(n−1) j (1 ≤ j ≤ n),
Proof. We first check the “only if” direction. Suppose that Tn j = 0 (1 ≤ j ≤ n − 2),
σ = τ 2 . Then every cycle of τ of length m remains a cycle in
Tn(n−1) = 1, Tnn = 2.
σ if m is odd, and splits into two cycles of length m/2 if m is
even. Using these, we read off the following features of T .
We next check the “if” direction. We may partition the cycles – T11 = n − 2 · 2 − 2 · 2 = n − 7.
of σ into individual cycles of odd length and pairs of cycles – If n = 2m is even, then Tm1 = 3 − 2 · 1 − 2 · 2 = −3
of the same even length; then we may argue as above to write and Tmm = 3 − 2 · 1 − 1 · 2 = −1.
each partition as the square of another permutation.
– If n2 < i < n−1, then Ti1 = Tin = 0 and, for 1 < j <
n, Ti j = 1 if j divides n − i and Ti j = 0 otherwise.
Suppose now that n > 1 is odd. Write n = pe k where p
is an odd prime, k is a positive integer, and gcd(p, k) = Now recall (e.g., from the expansion of a determinant in
1. By the Chinese remainder theorem, we have a ring minors) if a matrix contains an entry equal to 1 which is
isomorphism the unique nonzero entry in either its row or its column,
then we may strike out this entry (meaning striking out
Z/nZ ∼
= Z/pe Z × Z/kZ. the row and column containing it) at the expense of mul-
tiplying the determinant by a sign. To simplify notation,
Recall that the group (Z/pe Z)× is cyclic; choose m ∈ Z we do not renumber rows and columns after performing
reducing to a generator of (Z/pe Z)× and to the identity this operation.
in (Z/kZ)× . Then σn,m consists of k cycles (an odd
We next verify that for the matrix T , for i = 2, . . . , ⌈ n2 ⌉−
number) of length pe−1 (p − 1) (an even number) plus
1 in turn, it is valid to strike out (i, n − i) and (n −
some shorter cycles. By Lemma 1, σn,m does not have
i, i) at the cost of multiplying the determinant by -1.
a square root.
Namely, when we reach the entry (n − i, i), the only
Suppose next that n ≡ 2 (mod 4). Write n = 2k with k other nonzero entries in this row have the form (n − i, j)
odd, so that where j > 1 divides n − i, and those entries are in pre-
viously struck columns.
Z/nZ ∼
= Z/2Z × Z/kZ.
We thus compute det(S) = det(T ) as:
Then σn,m acts on {0} × Z/kZ and {1} × Z/kZ with the
n−7 2 0
same cycle structure, so every cycle length occurs an (−1)⌈n/2⌉ det 2 0 1 for n odd,
even number of times. By Lemma 1, σn,m has a square 0 1 2
root.
n − 7 −3 2 0
Finally, suppose that n is divisible by 4. For m = −1, −3 −1 1 0
σn,m consists of two fixed points (0 and n/2) together (−1)⌈n/2⌉ det for n even.
2 1 0 1
with n/2 − 1 cycles (an oodd number) of length 2 (an 0 0 1 2
even number). By Lemma 1, σn,m does not have a
square root. After subtracting 12 times the last column from the next-
to-last, we can then strike out the last row and column
B6 The determinant equals (−1)⌈n/2⌉−1 2⌈ 2n ⌉. to obtain
To begin with, we read off the following features of S.
⌈n/2⌉ n−7 2
(−1) 2 det for n odd,
– S is symmetric: Si j = S ji for all i, j, corresponding 2 − 12
to (a, b) 7→ (b, a)). n − 7 −3 2
– S11 = n + 1, corresponding to (a, b) = (−1)⌈n/2⌉ 2 det −3 −1 1 for n even.
(0, n), (1, n − 1), . . . , (n, 0). 2 1 − 21
5
Remark. One can use a similar approach to compute where τ(n) denotes the number of divisors of n and
some related determinants. For example, let J be the (
matrix with Ji j = 1 for all i, j. In terms of an indetermi- qn−1t + q2 − 2t for n odd,
nate q, define the matrix T by fn (q,t) =
qn−1t + q2 − qt − t for n even.