The American Mathematical Monthly

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/uamm20

PROBLEMS AND SOLUTIONS

Daniel H. Ullman, Daniel J. Velleman, Douglas B. West & with the

collaboration of Paul Bracken, Ezra A. Brown, Zachary Franco, László Lipták,
Rick Luttmann, Hosam Mahmoud, Frank B. Miles, Lenhard Ng, Kenneth
Stolarsky, Richard Stong, Stan Wagon, Lawrence Washington, and Li Zhou.

To cite this article: Daniel H. Ullman, Daniel J. Velleman, Douglas B. West & with the collaboration
of Paul Bracken, Ezra A. Brown, Zachary Franco, László Lipták, Rick Luttmann, Hosam Mahmoud,
Frank B. Miles, Lenhard Ng, Kenneth Stolarsky, Richard Stong, Stan Wagon, Lawrence
Washington, and Li Zhou. (2021) PROBLEMS AND SOLUTIONS, The American Mathematical
Monthly, 128:4, 376-384, DOI: 10.1080/00029890.2021.1877020

To link to this article: https://doi.org/10.1080/00029890.2021.1877020

Published online: 23 Mar 2021.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=uamm20
 PROBLEMS AND SOLUTIONS
Edited by Daniel H. Ullman, Daniel J. Velleman, and Douglas B. West
with the collaboration of Paul Bracken, Ezra A. Brown, Zachary Franco, László Lipták,
Rick Luttmann, Hosam Mahmoud, Frank B. Miles, Lenhard Ng, Kenneth Stolarsky,
Richard Stong, Stan Wagon, Lawrence Washington, and Li Zhou.

Proposed problems should be submitted online at

americanmathematicalmonthly.submittable.com/submit.
Proposed solutions to the problems below should be submitted by August 31, 2021,
via the same link. More detailed instructions are available online. Proposed prob-
lems must not be under consideration concurrently at any other journal nor be posted
to the internet before the deadline date for solutions. An asterisk (*) after the number
of a problem or a part of a problem indicates that no solution is currently available.

PROBLEMS
12244. Proposed by Rob Pratt, SAS Institute Inc., Cary, NC, Stan Wagon, Macalester Col-
lege, St. Paul, MN, Douglas B. West, University of Illinois, Urbana, IL, and Piotr Zielinski,
Cambridge, MA. A polyomino is a region in the plane with connected interior that is the
union of a finite number of squares from a grid of unit squares. For which integers k and
n with 4 ≤ k ≤ n does there exist a polyomino P contained entirely within an n-by-n grid
such that P contains exactly k unit squares in every row and every column of the grid?
Clearly such polyominos do not exist when k = 1 and n ≥ 2. Nikolai Beluhov noticed that
they do not exist when k = 2 and n ≥ 3, and his Problem 12137 [2019, 756; 2021, 381],
whose solution appears at the end of this column, shows that they do not exist when k = 3
and n ≥ 5.
12245. Proposed by Jiahao Chen, Tsinghua University, Beijing, China. Suppose that two
circles α and β, with centers P and
Q, respectively, intersect orthogo-
nally at A and B. Let CD be a diam-
eter of β that is exterior to α. Let E
and F be points on α such that CE
and DF are tangent to α, with C and
E on one side of P Q and D and F
on the other side of P Q. Let S be
the intersection of CF and QA, and
let T be the intersection of DE and
QB. Prove that ST is parallel to CD
and is tangent to α.
12246. Proposed by Seán Stewart, Bomaderry, Australia. Let ζ be the Riemann zeta func-
∞
tion, defined
n for n ≥ 2 by ζ (n) = n
k=1 1/k . Let Hn be the nth harmonic number, defined
by Hn = k=1 1/k. Prove
∞
 ∞
ζ (n) ζ (n)Hn π2
+ (−1)n = .
n=2
n2 n=2
n 6

doi.org/10.1080/00029890.2021.1877020

376 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 128
12247. Proposed by Prathap Kasina Reddy, Bhabha Atomic Research Centre, Mumbai,
India. For positive real constants a, b, and c, prove
 π ∞
a x 1
2 + a 2 )3/2
√ dx dθ =  .
0 0 π(x x 2 + b 2 + c 2 − 2cx cos θ (a + b)2 + c2

12248. Proposed by Askar Dzhumadil’daev, Almaty, Kazakhstan. Let n be a positive inte-

ger, and let xk be a real number for 1 ≤ k ≤ 2n. Let C be the 2n-by-2n skew-symmetric
matrix with i, j -entry cos(xi − xj ) when 1 ≤ i < j ≤ 2n. Prove

det(C) = cos2 (x1 − x2 + x3 − x4 + · · · + x2n−1 − x2n ).

12249. Proposed by Florin Stanescu, Serban Cioculescu School, Gaesti, Romania. Prove


n−1 
n−k  
k + m k + 1 k−m n
(−1)m−1 2 =
k=n/2 m=1
k+1 m−1 2

for any positive integer n.

12250. Proposed by Dorin Mărghidanu, Colegiul National A. I. Cuza, Corabia, Romania.
With n ≥ 4, let a1 , . . . , an be the lengths of the sides of a polygon. Prove
  
a1 a2 an 2n
+ + ··· + > .
−a1 + a2 + · · · + an a1 − a2 + · · · + an a1 + a2 + · · · − an n−1

SOLUTIONS

A Triangle Inequality from Cauchy–Schwarz

12130 [2019, 659]. Proposed by Dan Ştefan Marinescu, Hunedoara, Romania, and Mihai
Monea, Deva, Romania. Let P be a point in the interior of triangle ABC. Suppose that
the lines AP , BP , and CP intersect the circumcircle of ABC again at A , B , and C ,
respectively. Prove
S(BP C) S(AP C) S(AP B) S(BP C) S(AP C) S(AP B)
+ + ≥ + + ,
AP BP CP A P B P C P
where S(XY Z) denotes the area of triangle XY Z.
Solution by Li Zhou, Polk State College, Winter Haven, FL.
As usual, we let a, b, and
c denote the lengths of
the sides of
ABC oppo-
site the vertices A, B, and
C. Let x = S(CP B), y =
S(AP C), z = S(BP A),
and u = S(A CB).
Suppose that AP inter-
sects BC at D. We have

BD BD
z = S(BP A) = S(BDA) − S(BDP ) = (S(BCA) − S(BCP )) = (y + z).
a a

April 2021] PROBLEMS AND SOLUTIONS 377

Therefore BD = az/(y + z), and similarly CD = ay/(y + z). By the law of cosines,
c2 = AD 2 + BD 2 − 2AD · BD cos ∠ADB,
b2 = AD 2 + CD 2 + 2AD · CD cos ∠ADB.
Eliminating the last terms of these equations, we get c2 CD + b2 BD = aAD 2 +
aCD · BD, so (y + z)2 AD 2 = (c2 y + b2 z)(y + z) − a 2 yz. Therefore
u A D BD · CD a 2 yz
= = = ,
x+y+z AD AD 2 (c2 y + b2 z)(y + z) − a 2 yz
and

xAP x(y + z) x(y + z) (c2 y + b2 z)(y + z) − a 2 yz

= = 2
A P x+u (c xy + b2 zx)(y + z) − a 2 xyz + a 2 yz(x + y + z)
(c2 xy + b2 zx)(y + z) − a 2 xyz
= .
a 2 yz + b2 zx + c2 xy
Adding this to the other two analogous equations and simplifying, we obtain
xAP yBP zCP
+ + = x + y + z.
A P BP CP
Finally, let O and R be the circumcenter and circumradius of ABC. Since AP · A P =
BP · B P = CP · C P = R 2 − OP 2 , two applications of the Cauchy–Schwarz inequality
yield
 x y z  x y z 
+ + + +
AP BP CP A P B P C P
 2
x y z
≥ √ +√ +√
AP · A P BP · B P CP · C P
 
x y z
= (x + y + z) + +
R 2 − OP 2 R 2 − OP 2 R 2 − OP 2
  
xAP yBP zCP x y z
= + + + +
A P B P C P AP · A P BP · B P CP · C P
 x y z 2
≥ + + ,
A P B P C P
completing the proof.
Also solved by M. Bataille (France), G. Fera (Italy), L. Giugiuc (Romania), A. Stadler (Switzerland), R. Stong,
and the proposer.

Telescoping in Disguise
12134 [2019, 755]. Proposed by Paul Bracken, University of Texas, Edinburg, TX. Evaluate
the series
∞   ∞  
1 1
n −1− .
n=1 k=n
k2 2n

Solution by Kyle Gatesman, student, Johns Hopkins University, Baltimore, MD. The value
is 1/4. For n ∈ N,
 ∞   ∞
1 1 1
1=n − =n
k=n
k k+1 k=n
k(k + 1)

378 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 128
and
∞   ∞ ∞
1 n 1 1 n  2k + 1  k + 1/2
= − = =n .
2n 2 k=n k 2 (k + 1)2 2 k=n k 2 (k + 1)2 k=n
k 2 (k + 1)2

Therefore,
∞
∞  ∞
∞  
  1 1   1 1 k + 1/2
n −1− = n − −
n=1 k=n
k2 2n n=1 k=n
k2 k(k + 1) k 2 (k + 1)2
∞
∞ 
  (k + 1)2 − k(k + 1) − (k + 1/2)
= n
n=1 k=n
k 2 (k + 1)2
∞ ∞ ∞
1  1 k
1 1
= n = n
2 n=1 k=n k 2 (k + 1)2 2 k=1 k 2 (k + 1)2 n=1
∞ ∞ ∞  
1 1 k(k + 1) 1 1 1 1 1 1
= · = = − = .
2 k=1 k (k + 1)
2 2 2 4 k=1 k(k + 1) 4 k=1 k k+1 4

Interchanging the order of summation in the double sum is justified because all terms are
positive.
Also solved by B. Abdelhak (Algeria), U. Abel & V. Kushnirevych (Germany), B. Ahn (Korea), K. F. Andersen
(Canada), M. Bataille (France), N. Batir (Turkey), K. N. Boyadzhiev, B. Bradie, B. S. Burdick, R. Chapman
(UK), H. Chen, Ó. Ciaurri (Spain), C. Curtis, K. Egamberganov (France), G. Fera (Italy), P. Fulop (Hun-
gary), O. Geupel (Germany), M. Goldenberg & M. Kaplan, R. A. Gordon, J.-P. Grivaux (France), J. Gross-
man, J. A. Grzesik, E. A. Herman, N. Hodges (UK), F. Holland (Ireland), B. Huang, W. Janous (Austria),
J. C. Kieffer, B. Kim, O. Kouba (Syria), S. S. Kumar, P. Lalonde (Canada), K.-L. Lau (China), J. H. Lindsey II,
O. P. Lossers (Netherlands), L. Matejı́c̆ka (Slovakia), M. Omarjee (France), K. F. Schilling, N. C. Singer,
A. Stadler (Switzerland), A. Stenger, S. M. Stewart (Australia), R. Stong, R. Tauraso (Italy), M. Tetiva (Roma-
nia), M. Vowe (Switzerland), E. White & R. White, T. Wiandt, T. Wilde (UK), M. Wildon (UK), L. Zhou,
GCHQ Problem Solving Group (UK), and the proposer.

Bounds on a Function of the Medians

12135 [2019, 755]. Proposed by George Apostolopoulos, Messolonghi, Greece. Suppose
that a triangle in the plane has inradius r, circumradius R, angles A, B, and C, and corre-
sponding medians mA , mB , and mC . Prove
√
√ r2 sin3 A sin3 B sin3 C 3 3(R − r)
3 3 3 ≤ + + ≤ .
R mA mB mC 8r 2

Composite solution by Subhankar Gayen, Vivekananda Mission Mahavidyalaya, Haldia,

India and Oliver Geupel, Brühl, Germany. Let a, b, and c be the sides of the triangle
opposite the angles A, B, and C, and let s be the semiperimeter. By the law of sines, the
requested inequality is equivalent to
√
√ 2 a3 b3 c3 3 3R 3 (R − r)
24 3r ≤ + + ≤ . (1)
mA mB mC r2
By two applications of the AM–GM inequality, we have
a3 b3 c3 3abc 9abc
+ + ≥ √ ≥ .
mA mB mC 3 m m m
A B C m A + mB + mC

April 2021] PROBLEMS AND SOLUTIONS 379

Now we use the known √ result abc = 4sRr together with the two inequalities mA + mB +
mC ≤ 9R/2 and s ≥ 3 3r (inequalities 8.3 and 5.11 in O. Bottema, R. Ž. Djordjević,
R. R. Janić, D. S. Mitrinović, P. M. Vasić (1969), Geometric Inequalities, Groningen:
Wolters-Noordhoff) to conclude
a3 b3 c3 36sRr √
+ + ≥ = 8sr ≥ 24 3r 2 ,
mA mB mC 9R/2
which is the first inequality in (1).
To prove the second inequality, we begin with the observation that the median mA is at
√ A, which we denote by hA . The area of the triangle is both
least as long as the altitude from
rs and ahA /2, so using s ≥ 3 3r again we see
a3 a4 a4 a4 a4
= ≤ = ≤ √ .
mA amA ahA 2rs 6 3r 2
Combining this with similar inequalities for B and C, we obtain
a3 b3 c3 a 4 + b4 + c4
+ + ≤ √ .
mA mB mC 6 3r 2
The second inequality in (1) now follows from a 4 + b4 + c4 ≤ 54R 3 (R − r), which is item
5.26 on p. 177 of D. S. Mitrinović, J. E. Pečarić, V. Volenec (1989), Recent Advances in
Geometric Inequalities, Dordrecht: Kluwer Academic Publishers.
Editorial comment. Two solvers discovered
√ sharper upper bounds than required. Richard
Stong replaced the upper bound with 3 3R/(16r 2 ), and Michael Vowe replaced it with
(R − r)(4R + r)2 /(6R 2 rs).
Also solved by K. Egamberganov (France), G. Fera (Italy), N. Hodges (UK), K.-W. Lau (China), M. Lukarevski
(North Macedonia), A. Stadler (Switzerland), R. Stong, M. Vowe (Switzerland), J. Zacharias, and the proposer.

A Consequence of Schur’s Inequality

12136 [2019, 756]. Proposed by Albert Stadler, Herrliberg, Switzerland. Prove
  
4 b +c 4 c +a 4 a +b
4 4 4 4 4 4
a +b +c ≥a
2 2 2
+b +c
2 2 2
for all positive real numbers a, b, and c.
Solution by Roberto Tauraso, Università di Roma “Tor Vergata,” Rome, Italy. Since
2(x 2 − xy + y 2 )2 − (x 4 + y 4 ) = (x − y)4 ≥ 0, we have

 4 x +y
4 4
x − xy + y ≥
2 2
2
for all real x and y. Thus it suffices to show the stronger inequality
  
a 2 + b2 + c2 ≥ a b2 − bc + c2 + b c2 − ca + a 2 + c a 2 − ab + b2 .
√
Moreover, the concavity of x → x implies

a(b2 − bc + c2 ) + b(c2 − ca + a 2 ) + c(a 2 − ab + b2 )
a+b+c
√ √ √
a b2 − bc + c2 + b c2 − ca + a 2 + c a 2 − ab + b2
≥ .
a+b+c

380 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 128
That is,
√ 
a + b + c a(b2 + c2 ) + b(c2 + a 2 ) + c(a 2 + b2 ) − 3abc
  
≥ a b2 − bc + c2 + b c2 − ca + a 2 + c a 2 − ab + b2 .
Hence it suffices to show

(a 2 + b2 + c2 )2 ≥ (a + b + c) · a(b2 + c2 ) + b(c2 + a 2 ) + c(a 2 + b2 ) − 3abc .

However, this is just
a 2 (a − b)(a − c) + b2 (b − a)(b − c) + c2 (c − a)(c − b) ≥ 0,
which holds by Schur’s inequality.
Also solved by B. Ahn (Korea), F. R. Ataev (Uzbekistan), P. Bracken, R. Chapman (UK), G. Fera (Italy),
K. O. Gennadievich (Russia), J. A. Grzesik, E. A. Herman, N. Hodges (UK), P. Ivády (Hungary), W. Janous
(Austria), O. Kouba (Syria), K.-W. Lau (China), L. Matejı́c̆ka (Slovakia), D. J. Moore, Á. Plaza (Spain), A. Tza-
vella (Greece), E. A Weinstein, T. Wiandt, T. Wilde (UK), L. Zhou, and the proposer.

There Are No 3-Regular Polyominos

12137 [2019, 756]. Proposed by Nikolai Beluhov, Stara Zagora, Bulgaria. A polyomino is
a region with connected interior that is a union of a finite number of squares from a grid of
unit squares. Do there exist a positive integer n with n ≥ 5 and a polyomino P contained
entirely within an n-by-n grid such that P contains exactly 3 unit squares in every row and
every column of the grid?
Solution by Joseph DeVincentis, Salem, MA. (The editors have omitted some details due to
space considerations.) Let P be such a polyomino.
The five tetrominos are shown below; the first (in any orientation) is called a T .

We split the analysis into two parts, depending on whether P contains a T . When n = 4,
solutions exist with and without a T .

Part 1: P contains no T . Here also P contains no 2-by-2 square, because any square
connecting it to the rest creates a T . Also P cannot use four squares in a line (a “line” is a
row or a column of unit squares). Hence all tetrominos in P are of the last two types. Since
each square in P has at most two neighbors, P follows a one-cell-wide path or cycle in
which any four consecutive squares are arranged like one of the last two tetrominos.
Let L be one of the four boundary lines. When P is a cycle, or when P is a path that
does not begin or end in L, the squares of P in L must be consecutive, since an isolated
square in L can only be an end of P . Furthermore, if the cells of P in L are not consecutive,
then P begins and ends in L, since the square tetromino is forbidden.
Whether P is a path or a cycle, let B be a boundary line that contains neither an end of
P nor a square adjacent to an end of P . Such a boundary line B must exist, since P has at

April 2021] PROBLEMS AND SOLUTIONS 381

most two ends, and an analysis of cases examining the possibilities of these ends being at
or adjacent to a corner shows that at least one of the four boundary lines has the required
property. By symmetry, we may let it be the top row. The squares abc in B are consecutive,
and the squares 1 and 2 below a and c in the next row must also be in P and must not be
ends of P . We consider several cases.

a b c a b c
1 2 1 2 4
3 5 4 5 3 6
7 9 8
Figure 1 Figure 2
Suppose first that neither column a nor column c is a boundary column and P continues
straight down from 1 and 2 to reach squares 3 and 4 in those columns. If the third row does
not use square 5 in Figure 1, then we may assume by symmetry that P uses the square to
the left of 3. Now P can neither continue from 4 nor cross column a to complete column
b. Hence 5 must be used to complete the third row, and now the addition of any contiguous
square completes a T .
Hence by symmetry and the limit of three squares per row, we may assume that square
3 is used below 1, and then square 4 must be used to the right of 2 to permit completion of
the second row.
If column a is not leftmost, as in Figure 2, then P can only reach columns leftward from
a by using square 5 to the left of 3. Now ending at 4 would not permit completing column
b, so P must continue down from 4 to 6. Now the third row is finished, and P must use
squares 7 below 5 and 8 below 6. If P goes to the right from 8, then P cannot cross this
column or column a to finish columns b and c. Hence P must use square 9 but then can
neither continue from 9 nor cross column a from the left to complete column b.
Hence column a must be leftmost. The remaining cases depend on whether square 3 is
an end of P . Suppose first that it is. Now P must continue on below square 4, with square
5. Since we can only continue P on this end, the conditions successively force squares 6
through 10 as in Figure 3 (if P moves rightward after square 7 or 8, then P will not be able
to return to finish columns c and b). After square 10 there is nowhere to go.

a b c a b c
1 2 4 1 2 4
3 5 6 3 5 7
7 6 8 9
10 9 8
Figure 3 Figure 4
Finally, we have the case where a is the leftmost column and P continues after square
3, which can only be to square 5 in Figure 4. If P ends there, then the other end must
reach the fifth column and cannot return to column b. This forces square 6, and then by
symmetry of rows and columns we may assume that P continues below square 4 with 7.
Now 8 is forced, and 9 is forced (else we complete the example for n = 4), and now we
cannot return to finish column c (the other end cannot now extend beyond square 6).
Part 2: P contains a T . There are six cells around a T that may be included in P . We
label them a through f as in Figure 5. We consider cases for the usage of these cells.
Case A: a, b, c ∈ P . The shaded squares in Figure 6 can extend contiguously only to the
four  cells. Columns a and c both must be completed in this way, because P cannot move

382 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 128
across column b outside the three shaded cells. We cannot extend both columns upward,
because the horizontal completed rows would then prevent completing the bottom row,
and extending both downward would create a 3-by-3 block, which cannot extend. Hence
by symmetry we have the configuration shown in Figure 7.

a b c
a b c a b c
f d
e
Figure 5 Figure 6 Figure 7
Suppose that there are r rows above square a and s columns to its left. Due to the rows
and columns already completed, all remaining squares of P in these rows and columns
must appear both in a row above square a and a column to the left of square a. There are
3s such squares in P , but due to the square above a there are also 3r − 1 such squares to
be added. Since 3r − 1 = 3s, the configuration cannot be completed.
Case B: P ∩ {a, b, c} = ∅. If e ∈ P , then column b is finished, so also d, f ∈ P , which
is equivalent to Case A. If e ∈
/ P but d, f ∈ P , then the only extensions are below d and f ,
but then column b cannot be completed. The same is true if only one of {d, f } is included.
Case C: P ∩ {a, b, c} = {b}. The completed row and column prevent any additional
squares from being shaded in the row containing b.
Case D: b ∈ P and |P ∩ {a, c}| = 1. By symmetry, we may assume a ∈ P and c ∈ / P.
Note e ∈ / P . If d ∈
/ P , then no other squares can be added to the right of column b, so
column c cannot be completed. Hence d ∈ P . If also f ∈ P , then it is equivalent to Case
A. Hence we may assume the configuration in Figure 8, with c and f absent.
The row above d and column to the left of d are both complete, and the lines containing
d each need one more square. Call a position with these properties a wedge position. A
wedge position cannot occur in the rightmost column or bottom row, because then that
column or row cannot be completed. Since P is connected, the squares completing the
lines through d and the lower rows and rightward columns from d can only be reached
by paths from d through those rows and columns. Since the grid is finite, presence of any
wedge position requires that there is a wedge position having no other wedge position that
is below and to its right; we may assume that d is such a terminal wedge position. Any case
that then generates a wedge position below and to the right of d yields a contradiction.
We can extend from square d only to the right or below. By symmetry, we may assume
that P uses square 1 to the right of d. If also the square below d is used, then the only
extension is to square 2 below 1, but then 2 is another wedge position. Hence we may
assume that the square below d is empty, and again we can extend only to 2.

a b d 1
2 5 6
d 1 4 3
e 2
Figure 8 Figure 9
To extend from square 2, in P we must have a square 3 neighboring 2, either below 2
or to its right. First suppose that square 3 is below 2, as in Figure 9. Now the column of
2 is complete, so finishing the column of d requires square 4 in P . Further extension can
only be to the right of square 2 or square 3, and the latter forces the former, so P contains
square 5. Now 6 must be used whether the square below 5 is used or not.

April 2021] PROBLEMS AND SOLUTIONS 383

 Extension now can only be below square 5 or square 6; we consider these two cases
for square 7. If 7 is below 5 as in Figure 10, then the only available extension is below
7 to square 8, and then rightward from 8 to 9. Now square 9 is a wedge position, which
excludes this case.

d 1 d 1
2 5 6 2 5 6
4 3 7 4 3 7
8 9 9 8 11
10 12

Figure 10 Figure 11
Hence we may assume square 7 is below 6, as in Figure 11. Now the only extension
is below 7 to 8, which requires 9 and 10 as shown to complete the column of 5, since the
column of 6 is already finished. Now the only extension is rightward from 8 to 11 and
below 11 to 12, but now 12 is a wedge position.
Therefore, we may assume that the cell immediately below square 2 is not in P . Hence
the column of 2 is completed by square 3 farther below. Let m be the number of rows
skipped between 2 and 3. We show m = 1 in Figure 12 and m = 3 in Figure 13. Reasoning
as before, the completion of column d must be square 4 to the left of 3, but now since the
column of square 3 is complete we can only reach square 3 from square 5 to its right.
Now the only extension from the portion of P containing 2 is to square 6 on its right,
and since the row of square 5 is complete the only way to reach it is from square 7 above
5. If m = 1 as in Figure 12 then we are now stuck. We must extend to the right from 7
(whether or not we extend to the right from 6), and then we cannot complete that column.

d 1 d 1
2 6 2 6 8
7 9
4 3 5
7 10
4 3 5

Figure 12 Figure 13
If m > 1, then because the column of 6 and 7 is now complete we must extend to the
right from 6 to 8 and down to 9 and extend from 7 to 10. Now the column with 8, 9, 10
has the same properties as the column with 1, 2, 3, only with m smaller by 2. That is, the
columns to the left are complete, and the only way to bring the upper and lower portions
of P in this column together is to reduce m by 2 while moving to the right. However, when
m reduces to 2 or 1 the next column to the right cannot be complete.
The remaining cases use arguments like those above; we omit the details.
Case E: P ∩ {a, b, c} = {a, c}.
/ P and |P ∩ {a, c}| = 1.
Case F: b ∈
Editorial comment. The proposer showed that if P has exactly two squares in every row
and every column, then n must be 2.
Also solved by N. Hodges (UK) and the proposer.

384 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 128