Problems and Solutions

Reviewed work(s):
Source: The American Mathematical Monthly, Vol. 118, No. 4 (April 2011), pp. 371-378
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/10.4169/amer.math.monthly.118.04.371 .
Accessed: 23/09/2012 00:17

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Mathematical Association of America is collaborating with JSTOR to digitize, preserve and extend access to
The American Mathematical Monthly.

http://www.jstor.org
 PROBLEMS AND SOLUTIONS
Edited by Gerald A. Edgar, Doug Hensley, Douglas B. West
with the collaboration of Mike Bennett, Itshak Borosh, Paul Bracken, Ezra A. Brown,
Randall Dougherty, Tamás Erdélyi, Zachary Franco, Christian Friesen, Ira M. Ges-
sel, László Lipták, Frederick W. Luttmann, Vania Mascioni, Frank B. Miles, Bog-
dan Petrenko, Richard Pfiefer, Cecil C. Rousseau, Leonard Smiley, Kenneth Stolarsky,
Richard Stong, Walter Stromquist, Daniel Ullman, Charles Vanden Eynden, Sam Van-
dervelde, and Fuzhen Zhang.

Proposed problems and solutions should be sent in duplicate to the MONTHLY

problems address on the back of the title page. Submitted solutions should arrive
at that address before August 31, 2011. Additional information, such as general-
izations and references, is welcome. The problem number and the solver’s name
and address should appear on each solution. An asterisk (*) after the number of
a problem or a part of a problem indicates that no solution is currently available.

PROBLEMS
11564. Proposed by Albert Stadler, Herrliberg, Switzerland. Prove that
√ !
e−x (1 − e−6x )
Z ∞
3+ 5
d x = log .
0 x(1 + e−2x + e−4x + e−6x + e−8x ) 2

11565. Proposed by Shai Covo, Kiryat-Ono, Israel. Let U1 , U2 , . . . be independent

random variables, each uniformly distributed on [0,P1]. √
(a) For 0 < x ≤ 1, let N x be the least n such that nk=1 Uk > x. Find the expected
value of N x .
(b) For 0 < x ≤ 1, let Mx be the least n such that nk=1 Uk < x. Find the expected
Q
value of Mx .
11566. Proposed by Kent Holing, Statoil, Trondheim, Norway. Let q be a monic quartic
polynomial with rational coefficients, that is, q(z) = z 4 + a3 z 3 + a2 z 2 + a1 z + a0 with
all coefficients rational. Let r be the resolvent of q, that is, r (z) = z 3 − a2 z 2 + (a3 a1 −
4a0 )z + (4a2 a0 − a32 a0 − a12 ). Suppose r is irreducible over Q and q and r have a
common zero.
(a) Determine the Galois group of q.
Suppose further that the coefficients of q are integers.
(b) Show that a3 6 = 0.
(c) Show that only two quartics satisfy the hypotheses of part (b), and find them.
11567. Proposed by David Callan, University of Wisconsin-Madison, Madison, WI.
How many arrangements (a1 , . . . , a2n ) of the multiset {1, 1, 2, 2, . . . , n, n} satisfy the
following two conditions: (i) All entries between the two occurrences of any given
value i exceed i, and (ii) No three entries increase from left to right with the last two
adjacent? (When n = 3, one such arrangement is 122133.)
doi:10.4169/amer.math.monthly.118.04.371

April 2011] PROBLEMS AND SOLUTIONS 371

11568. Proposed by Kurt Foster, Colorado Springs, CO. For n ≥ 1, let f (n) be the
least-significant nonzero decimal digit of n!. For n ≥ 2, show that f (625n) = f (n).
11569. Proposed by M. H. Mehrabi, Nahavand, Iran. Let a, b, and c be the lengths
of the sides of a triangle, and let s, r , and R be the semi-perimeter, inradius, and
circumradius, respectively, of that triangle. Show that
(a + b)(b + c)(c + a)
 
2 < log < (a + b + c) a1 + b1 + 1c − 6


abc
and
     
 r 2 b+c c+a a+b 2r
8 < log log log < .
s a b c R

11570. Proposed by Kirk Bresniker, Hewlett-Packard, Granite Bay, CA, and Stan
Wagon, Macalester College, St. Paul, MN. Alice and Bob play a number game. Start-
ing with a positive integer n, they take turns changing the number; Alice goes first.
Each player in turn may change the current number k to either k − 1 or dk/2e. The
person who changes 1 to 0 wins. For instance, when n = 3, the players have no choice,
k proceeds from 3 to 2 to 1 to 0, and Alice wins. When n = 4, Alice wins if and only
if her first move is to 2. For which initial n does Alice have a winning strategy?
11571. Proposed by Finbarr Holland, University College Cork, Cork, R 1 Ireland. Let
f be a nonnegative Lebesgue-measurable function on [0, 1], with 0
f (x) d x = 1.
Let K (x, y) = (x − y) f (x) f (y), F(t) = [0,t]×[0,t] K (x, y) dy d x, and G(t) =
2
R

K (x, y) dy d x. For 0 ≤ t ≤ 1, prove that

R
[t,1]×[t,1]
p p p
F(t) + G(t) ≤ F(1).

SOLUTIONS

An Occasional Congruence
11411 [2009, 179]. Proposed by Alun Wyn-jones, KPMG, London, U.K. For positive
integers k and n, let L k (n) = n−1j=1 (−1) j .
j k
P
(a) Show that L 1 (n) ≡ L 5 (n) (mod n) if and only if n is not a multiple of 4.
(b) Given distinct, odd, positive integers i and j with {i, j} 6 = {1, 5}, show that the set
of n such that L i (n) ≡ L j (n) (mod n) is finite.
Solution to part (a) by A. Kumar, Goleta, CA . By induction on n, one can check that
(n − 2)(n − 1)n(n + 1)
L 5 (n) − L 1 (n) = (−1)n−1 (2n − 1) .
4
The numerator contains the product of four consecutive integers. If n is not divisible
by 4, then a factor other than n is divisible by 4, and n divides L 5 (n) − L 1 (n). If n is
divisible by 4, then L 5 (n) − L 1 (n) = (−1)n−1 (2n − 1)(n − 2)(n − 1)(n + 1)(n/4) ≡
(−1)(−1)(−2)(−1)(1)(n/4) = n/2 6 ≡ 0 (mod n).
Solution to part (b) by Richard Stong, Center for Communications Research, San
Diego, CA . We prove the restriction of the claim to odd n. The functions L k (n) are
related to the Euler polynomials by L k (n) = 12 [E k (0) + (−1)n−1 E k (n)] (see 23.1.4 in

372 c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 118

M. Abramowitz and I. A. Stegun (eds.), Handbook of Mathematical Functions with
Formulas, Graphs, and Mathematical Tables, Dover, 1972.) The sequence of Euler
polynomials is usually defined by its exponential generating function:
∞
X tr 2e xt
Er (x) = t . (1)
r =0
r! e +1

Since
∞
X tr 2 et − 1
Er (0) = t =1− t = 1 − tanh(t/2),
r =0
r! e +1 e +1

and tanh(t/2) is an odd function, E 0 (0) = 1 and E 2m (0) = 0 for m ≥ 1. Also,

0 2
−2et
 
2 −2 2 −2 1 2
= = + = + ,
et + 1 (et + 1)2 et + 1 (et + 1)2 et + 1 2 et + 1
so
∞ ∞ ∞
!2
X tr X tr 1 X tr
Er +1 (0) + Er (0) = Er (0) .
r =0
r ! r =0 r! 2 r =0 r!

Comparing the coefficients of t r yields

r  
1X r
Er +1 (0) + Er (0) = E k (0)Er −k (0). (2)
2 k=0 k

With r = 0, we obtain E 1 (0) = −1/2. Since the even terms vanish, setting r = 2m in
(??) yields a recurrence for the odd terms:
m−1  
1X 2m
E 2m+1 (0) = E 2s+1 (0)E 2(m−s−1)+1 (0). (3)
2 s=0 2s + 1

Induction on m yields E 2m+1 (0) ∈ Z[ 21 ], where Z[ 12 ] is the set of rational numbers with
powers of 2 as denominators.
Let E 0 = (−1)m+1 E 2m+1 (0). Multiplying (??) by (−1)m+1 yields E 2m+1 0
=
Pm−1 2m+1
1
2 s=0 E 0
E 0
2s+1 2(m−s−1)+1 . With E 1 (0) = −1/2, inductively all E 0
2m+1 are positive.
In particular, |E 2m+1 (0)| > m2 |E 2m−1 (0)|, so |E 5 (0)| < |E 7 (0)| < · · · . Since E 1 (0) =
E 5 (0) = − 21 and E 3 (0) = 14 , the values E 2m+1 (0) for m ≥ 0 are distinct except for
E 1 (0) = E 5 (0).
Factoring the generating function (??) as e xt · et 2+1 , we obtain
∞ ∞ ∞
X tr X xr tr X tr
Er (x) = · Er (0) ,
r =0
r! r =0
r ! r =0 r!

and hence Er (x) = k=0 rk Er −k (0)x k . Thus Er is a polynomial with coefficients in

Pr

Z[ 12 ] and constant term Er (0). For odd i and j, we can now write
1
L i (2m + 1) − L j (2m + 1) = (E i (0) + E i (2m + 1) − E j (0) − E j (2m + 1))
2
= E i (0) − E j (0) + (2m + 1)P(2m + 1),

April 2011] PROBLEMS AND SOLUTIONS 373

where x P(x) = 12 [E i (x) − E i (0) − E j (x) + E j (0)]. Note that P is a polynomial with
coefficients in Z[ 12 ]. Thus L i (2m + 1) ≡ L j (2m + 1) (mod 2m + 1) only if 2m + 1
divides the numerator of E i (0) − E j (0), since its denominator is a power of 2. Since
E i (0) 6= E j (0) for {i, j} 6 = {1, 5}, there are only finitely many such odd numbers.
Editorial comment. In part (b) the restriction to odd n was omitted in error. Several
solvers pointed out that the claim is false when n is allowed to be even. Zoltán Vörös
proved this as follows: For k odd and n even, (n − i)k ≡ (−i)k ≡ −i k (mod n), so
the terms in the sum L k (n) cancel in pairs except for the middle term. Thus L k (n) ≡
(n/2)k (mod n). Hence if i and j are odd and n is even but not divisible by 4, then
L i (n) − L j (n) ≡ (n/2)i − (n/2) j ≡ 0 (mod n). If both i and j are at least 3, then
every even n will work.
Also solved by F. Alayont and by R. Chapman (U.K.). Part (a) only solved by D. Beckwith, Z. Vörös (Hun-
gary), BSI Problems Group (Germany), GCHQ Problem Solving Group (United Kingdom), and the proposer.
Falsity for even n also observed by D. Fleischman and J. H. Lindsey II.

Splitting Fields of Cubic Polynomials

11420 [2009, 277]. Proposed by Kent Holing, StatoilHydro, Trondheim, Norway. Let
p be a monic cubic polynomial√with integer coefficients and discriminant D. Show
that if r is a zero of p and d = D, then Q(r + d) is the splitting field of p.
Solution by Achava Nakhash, San Diego, CA. Let K be the splitting field of p. Let r ,
s, and t be the roots of p, so d = ±(r − s)(r − t)(s − t). If p splits into linear factors
over Q, then both r and d are rational, and it follows that K = Q(r + d) = Q. If p
splits into a linear
√ and an irreducible √ quadratic factor,
√ and r is the rational root, then
because d = D and s, t, d ∈ Q( D), K = Q( D) = Q(r + d). A similar proof
works if r is one of the nonrational roots.
If p is irreducible, then the Galois group of K is either A3 or S3 . In the former case,
the discriminant is a rational square, so Q(r ) = Q(r + d) 6 = Q; since K is of degree
3, it follows that K = Q(r ) = Q(r + d). Finally, if the Galois group of K is S3 , then
it must include the permutation that fixes r and transposes s and t. This permutation
does not fix d, so r + d is not in Q(r ), because the latter field is the fixed field of A3 .
Thus, Q(r + d) properly contains Q(r ) and is a subfield of the splitting field K . Since
K is a degree-six extension of the rationals, it follows that Q(r + d) = K , as claimed.
Also solved by R. Chapman (U.K.), P. Corn, P. P. Dályay (Hungary), S. M. Gagola Jr., J. Grivaux (France),
J. H. Lindsey II, J. Simons (U.K.), J. H. Smith, R. Stong, M. Tetiva (Romania), GCHQ Problem Solving Group
(U.K.), NSA Problem Solving Group, and the proposer.

Recurrences for k-Fibonacci Numbers

11421 [2009, 277]. Proposed by Sergio Falcón and Ángel Plaza, University of las
Palmas de Gran Canaria, Las Palmas de Gran Canaria, Spain. Fix a positive integer
k and define the sequence hai by a0 = 0, a1 = 1, and an+1 = kan + an−1 for n ≥ 1.
(a) Show that if n, r , and h are nonnegative integers such that r + h ≤ n, then
an+r an+r +h + (−1)h+1 an−r −h an−r = a2n a2r +h .
(b) Show that if i and j are positive integers with i ≥ j, then
j−1
(
X ai a j+1 , if j is odd;
k ai−r a j−r =
r =0
a a
i j+1 − ai− j , if j is even.

[Editor’s note - In the original printing, “an+r ” read “an+1 ”; we regret the error.]

374 c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 118

Solution I by J. C.√Linders, Eindhoven, The Netherlands.
√
(a) Set φ = (k + k 2 + 4)/2 and ω = (k − k 2 + 4)/2, which are the roots of x 2 −
kx − 1 = 0. Since this is the characteristic equation of the recurrence, checking the
initial conditions confirms that an = (φ n − ωn )/(φ − ω) for n ≥ 0. Thus
φ p − ω p φ q − ωq φ p+q − φ p ωq − ω p φ q + ω p+q
a p aq = · =
φ−ω φ−ω (φ − ω)2
for p, q ≥ 0. Since φω = −1, this yields
φ p+q + ω p+q − (−1)q φ p−q − ω p−q


a p aq = . (4)
(φ − ω)2
It follows that
φ p+q+2s + ω p+q+2s − (−1)s φ p+q + ω p+q


a p+s aq+s − (−1) a p aq =
s
.
(φ − ω)2
Applying (??) again yields, for p, q, s ≥ 0,
a p+s aq+s − (−1)s a p aq = a p+q+s as (5)
The claim follows by setting p = n − r − h, q = n − r , and s = 2r + h.
(b) Setting q = 1 in (??) yields a p+s as+1 − (−1)s a p = a p+s+1 as . Using this after ap-
plying the recurrence,
ka p+s as = (a p+s+1 − a p+s−1 )as = a p+s as+1 − (−1)s a p − a p+s−1 as ,
assuming p + s ≥ 1. For positive j and nonnegative p, we obtain
j j j
X X
X
k a p+s as = a p+s as+1 − a p+s−1 as − (−1)s a p
s=1 s=1 s=1

j
(
X a p+ j a j+1 , if j is odd;
= a p+ j a j+1 − a p a1 − a p (−1) =
s

s=1
a p+ j a j+1 − a p , if j is even.

The claim follows by setting p = i − j and s = j − r .

Solution II to part (a) by the editors. The claim of (a) reduces to an+r a2n + 0 = a2n an+r
when r + h = n, so in what follows we assume r + h < n. The defining recurrence
generates the k-Fibonacci numbers: an is the number of distinguishable tilings of a
unit-width strip of length n − 1 using squares of k colors and uncolored dominos (see
A. T. Benjamin and J. J. Quinn, Fibonacci and Lucas identities through colored tilings,
Utilitas Math. 56 (1999) 137–142). To facilitate combinatorial proofs, let bn = an+1
for n ≥ 0 and bn = 0 for n < 0.
Start the strip at point 0, so br bs tilings of length r + s have a breakpoint at r .
Similarly, br −1 bs+1 have a breakpoint at r − 1. Those with a square in position r are
counted by both products. Thus br bs−1 − br −1 bs = br −2 bs−1 − br −1 bs−2 . For r < s,
this yields br bs−1 − br −1 bs = (−1)r bs−r −1 by induction on r .
Now consider br +t bs − br bs+t . Among the tilings of length r + s + t counted by
the two products, those not counted by both are those in the first set that don’t break at
r and those in the second that don’t break at r + t. For r < s, we have
br +t bs − br bs+t = br −1 bt−1 bs − br bt−1 bs−1
= bt−1 [br −1 bs − br bs−1 ] = (−1)r −1 bt−1 bs−r −1 .

April 2011] PROBLEMS AND SOLUTIONS 375

Setting r = 2m + h − 1, s = n + m − 1, and t = n − m, where m + h < n, converts
this to the desired identity with m written in place of the original r .
Solution II to part (b) by the proposers. To give a combinatorial argument, we keep the
notation from the previous solution to (a) and rewrite the desired identity (for i ≥ j)
as
j
(
X bi−1 b j , if j is odd;
k bi−r b j−r =
r =1
b b
i−1 j − bi− j−1 , if j is even.

There are bi−1 b j pairs of tilings such that one has length i − 1 in positions 2 through i
and the other has length j in positions 1 through j.
To count these another way, group the pairs by the first fault, where a fault is a
position where both tilings have a breakpoint. The first fault immediately follows the
leftmost square in the two tilings, say at position r , and there are k ways to color that
square. This determines both tilings through position r , and there are bi−r b j−r ways to
complete the two tilings.
This counts all the pairs when j is odd, but when j is even there may be no fault. In
this case both tilings begin with j/2 dominos, and there are bi− j−1 ways to complete
the longer tiling.
Editorial comment. The special case of the identity in Solution II of (a) obtained by
setting r = n − 1, t = 1, and s = n + ` was proved (for k = 1) and applied in Br.
A. Brousseau, Summation of infinite Fibonacci series, Fibonacci Quarterly 7 (1969)
143–168.
Also solved by M. Bataille (France), D. Beckwith, R. Chapman (U.K.), P. Corn, C. Curtis, P. P. Dályay (Hun-
gary), Y. Dumont (France), D. Fleischman, J. Grivaux (France), O. Kouba (Syria), H. Kwong, J. H. Lindsey II,
O. P. Lossers (Netherlands), A. Nakhash, J. P. Robertson, J. Simons (U.K.), A. Stadler (Switzerland), R. Stong,
M. Tetiva (Romania), Microsoft Research Problems Group, and the proposer. Part (b) solved by G. C. Greubel
and GCHQ Problem Solving Group (U.K.).

Equally Many Repetitions of Each Type

11424 [2009, 277]. Proposed by Emeric Deutsch, Polytechnic University, Brooklyn,
NY. Find the number of bit strings of length n in which the number of 00 substrings
is equal to the number of 11 substrings. For example, when n = 4 we have 4 such bit
strings: 0011, 0101, 1010, and 1100.
Solution by Christopher Carl Heckman,
Arizona State University, Tempe, AZ. The
n−2
answer is 2 when n = 1 and is 2 dn/2e−1 when n > 1.
Consider n > 1. Given a bit string of length n, let ai be the number of 0s in the ith
run of 0s, and let bi be the number of 1s in the ith run of 1s. For example, the string
00011011 yields a1 = 3, b1 = 2, a2 = 1, and bP 2 = 2. The number of copies of 00 is
(a i (bi − 1).
P
i i − 1),Pand the numberPof copies of 11 is
Let a = i ai and b = i bi . If the first and last bits differ, then the number of
runs of each type is the same, and the desired condition holds if and only if a = b,
which requires n to be even. If the first and last bits are 0, then there is an extra run
of 0, and the desired condition holds if and only if a = b + 1, which requires n to be
odd. Similarly, the condition is a + 1 = b and n odd when the first and last bits are 1.
There are two possibilities for the first bit. Once this bit
is chosen, the number of
n−2
ways to satisfy the needed condition on a and b is (n−2)/2 when n is even, and it is
n−2 n−2



(n−1)/2
when n is odd. Combining the computations yields 2 dn/2e−1 for the answer
in both cases.

376 c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 118

Editorial comment. Let s(n, k) be the number of binary n-tuples beginning with 0
in which the number of copies of 00 exceeds the number of copies of 11 by exactly
n−2


k. Several solvers showed that s(n, k) = d(n−k)/2e−1 ; setting k = 0 gives the desired
result. Generalizing the problem in a different direction, Heckman asks how many
binary n-tuples have the same number of copies of 00 and 01. Computational results
suggest that this may be harder.
Also solved by T. Barcume, D. Beckwith, N. Caro (Colombia), R. Chapman (U.K.), P. Corn, C. Curtis,
P. P. Dályay (Hungary), D. Grinberg, N. Grivaux (France), P. Landweber, J. H. Lindsey II, O. P. Lossers
(Netherlands), R. Martin (Germany), D. Nacin, C. M. Nicolas, Á. Plaza & A. Saure & J. P. Pedro (Spain),
C. R. Pranesachar (India), R. E. Prather, J. Riley, T. Rucker, B. Schmuland (Canada), J. Simons (U.K.),
R. Staum, R. Stong, S. Sullivan, R. Tauraso (Italy), M. Tetiva (Romania), B. Tomper, H. Widmer (Switzer-
land), L. Zhou, CMC 328, GCHQ Problem Solving Group (U.K.), Hofstra University Problem Solvers, NSA
Problems Group, Problem Solving Group Magdeburg (Germany), Texas State University Problem Solvers
Group, and the proposer.

Sum of Two Squares in Every Congruence Class

11425 [2009, 366]. Proposed by Erwin Just (emeritus), Bronx Community College,
City University of New York, Bronx, NY. For which positive integers m does every
congruence class mod m contain the sum of two squares?
Solution by John H. Lindsey II, Cambridge, MA . Every congruence class modulo m
contains a sum of two squares if and only if every prime whose square divides m is
congruent to 1 modulo 4.
Necessity. Let p be a prime with p 2 | m. If every congruence class modulo m con-
tains a sum of two squares, then the same holds modulo p 2 . Modulo 4 we cannot get 3
as a sum of squares, so p > 2. If a 2 + b2 ≡ p mod p 2 , then a and b are not both di-
visible by p. Hence in the field F p we have a 2 + b2 = 0 and may assume a is nonzero;
thus (b/a)2 = −1, and b/a has order 4 in the multiplicative group of p − 1 elements.
This requires 4 | ( p − 1), as desired.
Sufficiency. Let p be a prime. We claim that every congruence class modulo p con-
tains a sum of two squares. This is clearly true for p = 2, so suppose p is odd, and let
h be the first quadratic nonresidue among 1, . . . , p − 1. Thus h ≡ a 2 + 1 mod p for
some integer a. The quadratic nonresidues are the p−1 2
numbers of the form b2 h, where
b 6= 0. Since b2 h = (ba)2 + b2 , the quadratic nonresidues are sums of two squares.
The quadratic residues have the form b2 + 02 . Hence every congruence class mod-
ulo p contains a sum of two squares.
If also p ≡ 1 mod 4, then each class contains a sum of two squares that are not
both divisible by p. This is obvious except for class 0; since −1 is a quadratic residue
modulo p, we have a 2 + 1 ≡ 0 mod p for some a with p - a .
We show next that if all classes modulo p occur in the form a 2 + b2 with p not
dividing a, then for i ≥ 2 all classes modulo pi occur in the form (a + pj)2 + b2 for
some j with 0 ≤ j ≤ pi−1 − 1. For this, it suffices to show that for fixed a, the values
(a + pj)2 lie in pi−1 distinct classes modulo pi , all of which are congruent to a 2
modulo p. If (a + pj)2 and (a + pk)2 lie in the same class modulo pi , then pi divides
their difference, which equals (2a + pj + pk) p( j − k). We have assumed that p - a
and pi−1 - ( j − k), so (a + pj)2 6 ≡ (a + pk)2 mod pi . Also, (a + pj)2 ≡ a 2 mod p.
i
Now suppose that m has prime factorization nj=1 p j j such that i j ≥ 2 implies p j ≡
Q
1 mod 4. Fix a congruence class r mod m. For each j with 1 ≤ j ≤ n, there exist a j
i
and b j such that a 2j + b2j ≡ r mod p j j . By the Chinese Remainder Theorem, there
i i
exist a and b such that a ≡ a j mod p j j and b ≡ b j mod p j j for 1 ≤ j ≤ n. Hence

April 2011] PROBLEMS AND SOLUTIONS 377

 i i
a 2 + b2 ≡ a 2j + b2j ≡ r mod p j j for 1 ≤ j ≤ n. Since the various p j j are relatively
prime, a 2 + b2 ≡ r mod m.
Also solved by R. Chapman (U.K.), S. M. Gagola Jr., J. Grivaux (France), C. Lanski, O. P. Lossers (Nether-
lands), J. Moreira (Portugal), A. Nakhash, K. Schilling, J. Simons (U.K.), N. C. Singer, R. Stong, M. Tetiva
(Romania), Con Amore Problem Group (Denmark), GCHQ Problem Solving Group (U.K.), Microsoft Re-
search Problems Group, NSA Problems Group, and the proposer.

An Arccos Integral
11457 [2009, 747]. Proposed by M. L. Glasser, Clarkson University, Potsdam, NY. For
real numbers a and b with 0 ≤ a ≤ b, find
Z b  
x
arccos √ d x.
x=a (a + b)x − ab

Solution by FAU Problem Solving Group, Florida Atlantic University, Boca Raton, FL.
ab
We integrate by parts, using x − a+b as the antiderivative of 1. Notice that
 
d x b(x − a) − a(b − x)
arccos √ =−
√ ,
dx (a + b)x − ab 2(a + b) x − a+b ab
(b − x)(x − a)
√
and that arccos(x/ (a + b)x − ab) vanishes for x = a and x = b. Thus
Z b      b
x ab x 
arccos √ dx = x − arccos √ 
a (a + b)x − ab a+b (a + b)x − ab a
" Z √
b Z b√ #
1 x −a b−x
+ b √ dx − a √ dx
2(a + b) a b−x a x −a
" Z √
b Z b√ #
1 x −a b−x
= b √ dx − a √ dx .
2(a + b) a b−x a x −a

Writing B for Euler’s beta function, we have

Z b√ Z b√
(b − a)π
 
x −a b−x 3 1
√ dx = √ d x = (b − a)B , = .
a b−x a x −a 2 2 2
It follows that
b
(b − a)2 π
Z  
x
arccos √ dx = .
a (a + b)x − ab 4(a + b)

Also solved by K. F. Andersen (Canada), G. Apostolopoulos (Greece), M. S. Ashbaugh, M. R. Avidon, R.

Bagby, D. H. Bailey & J. M. Borwein (U.S.A. & Canada), D. Beckwith, K. N. Boyadzhiev, P. Bracken, R.
Chapman (U.K.), H. Chen, C. Curtis, P. P. Dályay (Hungary), P. De (India), M. Goldenberg & M. Kaplan,
J.-P. Grivaux (France), J. A. Grzesik, E. A. Herman, F. Holland (Ireland), M. E. H. Ismail, W. P. Johnson, P.
Khalili, O. Kouba (Syria), C. Koutschan, G. Lamb, O. P. Lossers (Netherlands), K. McInturff, D. K. Nester, M.
Omarjee (France), P. Perfetti (Italy), J. Posch, C. R. Pranesachar (India), O. G. Ruehr, F. Sami, B. Schmuland
(Canada), J. Simons (U.K.), N. C. Singer, S. Singh, R. Stong, T. Tam, A. P. Taraporevala, R. Tauraso (Italy),
M. Tetiva (Romania), T. P. Turiel, D. B. Tyler, E. I. Verriest, M. Vowe (Switzerland), T. Wiandt, H. Widmer
(Switzerland), GCHQ Problem Solving Group (U.K.), Hofstra University Problem Solvers, Missouri State
University Problem Solving Group, and the proposer.

378 c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 118