Solutions to the Twenty-Eighth University of Melbourne/Hewlett-Packard

Mathematics Competition

Senior Division, 1999

1 Two mathematicians, Walter and Hyam, are working at their respective homes on
a problem on which they are collaborating. They both get a good idea, and decide
to walk over to the home of the other to discuss their insights. They leave home
simultaneously, and walk, at uniform velocity, along the same road to the home of
the other. Being engrossed in the problem, they fail to notice that they actually pass
one another! However, 3 minutes after passing one another, Hyam arrived at Walter's
home, whereas Walter arrived at Hyam's home one minute after their meeting. How
long had each of them walked?
The distance apart doesn't matter | it just scales the speed of the two walkers.
However, let's call it d metres. Let Hyam walk with speed h m/min. and Walter
with speed w m/min. Let them meet on the path after t min. of walking. Thus
ht + wt = d: But we are told that ht + 3h = d; and also that wt + t = d: Thus
wt + t =pht + 3h = ht + wt: Hence 3h = ht
p2 : This gives h = 0; a non-physical solution,
p
or t = 3: Hence Hyam walks for 3 + 3 minutes, while Walter walks for 1 + 3
minutes.
Alternatively, let Walter and Hyam meet after t minutes of walking. Let d1 and d2 be
the distances that Walter and Hyam each walk respectively from their homes to the
point they pass each other. Then Walter covers distance d1 in t minutes and d2 in 1
minute, while Hyam covers distance
d2 in t minutes and d1 inp3 minutes. As they are
each walking at constant speeds, dd = 1t = 3t , from which t = 3.
p
p
Hence Hyam walks for 3 + 3 minutes, while Walter walks for 1 + 3 minutes.
2 Three grasshoppers A, B and C are placed on a line. Grasshopper B sits at the midpoint between A and C. Every second, one of the grasshoppers jumps over one of the
others to the symmetrical point on the other side (if X jumps over Y to point X', then
XY=YX'). After several jumps it so happened that they returned to the three initial
points (but maybe in dierent order). Prove that B returns to her initial position.
Label the points corresponding to the initial positions of A; B; C as 1; 0; 1 respectively.
Then by the rules of the game, A and B will continue to occupy points with odd
numbered coordinates, while B will occupy points with even numbered coordinates.
This is because every grasshopper moves an even number of steps on each move, leaving
their parity unchanged. Thus if they return to the initial coordinates, B must occupy
her original spot, being the only grasshopper whose initial coordinate was even.
3 A right circular cone has base of radius 1 and height 3. The largest possible cube is
inscribed in the cone so that one face of the cube is contained in the base of the cone.
What is the side-length of the cube?
1
2

Consider the plane containing both the axis of the cone and two opposite vertices of the
cube's bottom face. The cross
p section of the cone and the cube in this plane consists of
a rectangle of sides s and s 2 inscribed in anpisosceles triangle of base 2 and height 3,
where s is the side-length of the cube. (The s 2 side
p of the rectangleplies on the base of
the triangle.) Similar triangles yield s=3 = (1 s 2=2)=1, or s = (9 2 6)=7: (Many
students visualised in cross section a square rather than a rectangle being inscribed in
the triangle. This led to the incorrect answer of s = 6=5.)
4 If Sn denotes the sum of the decimal digits of the number 2n; either show that Sn 6= Sn+1
for all positive integers n; or nd a counter example.
Pr
r
Let 2n = a0 +10a1 + : : : +10
ar = i=0 ai 10i , where the ai 's are the digits of 2n . Then
P
r
Sn = a0 + a1 + : : : + ar = i=0 ai .
Pr
As 10iPr1 is divisible
by
9
for
any
non-negative
integer
i, 9 is a factor of i=0 ai (10i
P
1) = i=0 ai10i ri=0 ai = 2n Sn. Similarly 9 divides 2n+1 Sn+1. Therefore 9
divides their dierence which is (2n+1 Sn+1) (2n Sn) = 2n (as Sn = Sn+1 by
assumption). But 9 cannot divide 2n, which only has powers of two as factors.
From this contradiction we conclude that there is no such n for which Sn = Sn+1.
5 Let s be any arc of the unit circle lying entirely in the rst quadrant. Denote its
endpoints by D and E. Let A be the area of the region lying below s; bounded by s;
the x-axis and vertical lines through D and E. Let B be the area of the region lying to
the left of s, and bounded by the y axis and horizontal lines through D and E. Prove
that A + B depends only on the arc length, and not on the position, of s.
First solution, without calculus: to x notation, let A be the area of region DEF G,
and B be the area of DEIH ; further let C denote the area of sector ODE , which
only depends on the arc length of s. If [XY Z ] denotes the area of triangle [XY Z ],
then we have A = C + [OEG] [ODF ] and B = C + [ODH ] [OEI ]. But clearly
[OEG] = [OEI ] and [ODF ] = [ODH ], and so A + B = 2C .
D
H
E

I
O

Second solution, with calculus: We may parameterize a point in s by any of x, y,

or  = tan 1(y=x). Then A and B are just the integrals of y dx and x dy over the
appropriate intervals; thus A + B is the integral of x dy y dx (minus because the
limits of integration are reversed). But d = x dy y dx, and so A + B =
 is
precisely the radian measure of s. (Of course, one can perfectly well do this problem
by computing the two integrals separately. But what's the fun in that?)
2

6 (a) A 6  6 board is tiled with 2  1 dominoes. Prove that we can always divide the
board into two rectangles each of which is tiled separately (with no domino crossing
the dividing line).
(b) Is this true for an 8  8 board?
Solution provided by Chaitanya Rao: (a) Assume that no such division is possible. We
say a domino bridges two columns if half the domino is in each column. We show by
induction that for 0 < n < 6 the number of dominoes bridging columns n and n + 1
must be at least 2 and even.
Consider rst n = 1: There cannot be 3 dominoes entirely in column 1, or it would be
separately tiled. So there must be at least one domino bridging columns 1 and 2. The
number must be even, because it must equal the number of squares in column 1 (even)
less twice the number of dominoes (entirely) in column 1.
Now suppose it is true for n < 5 and consider column n + 1: There must be at least
one domino bridging columns n + 1 and n + 2; or columns 1 through n + 1 would
be separately tiled. The number must be even, because it must equal the number of
squares in column n + 1 (even) less the number bridging n and n + 1 (even) less twice
the number entirely in the column. By induction, it follows that at least 2 dominoes
bridge columns n and n + 1 for 0 < n < 6:
So in total there are at least 5  2 = 10 dominoes bridging columns. By the same
argument there are at least another 10 bridging rows, but there are only 18 dominoes
in total, a contradiction.
Solution based on that presented by Andrew Cheeseman:
Assume that no such division is possible. Then any separation of the board into two
rectangles must have at least one domino being cut in half. Let the number of such
dominoes for a particular separation be k, where k > 0. The total number of squares
in one of the rectangles is then twice the number of whole dominoes within it, plus k.
But this total has to be even as the rectangle has 6 rows or columns. Hence k is even
and so the number of dominoes \bridging" any row or column must be at least 2.
Therefore in total there are at least 5  2 = 10 dominoes bridging columns and at least
another 10 bridging rows. But there are only 18 dominoes in total, a contradiction.
(b) No. Many examples can be constructed, here is one:
7 A polynomial P (x) is said to be boring if it takes rational values for rational x and
irrational values for irrational x: Prove that the coecients of a boring polynomial
must be rational. Hence, or otherwise, prove that all boring polynomials are linear.
Solutions provided by Frank Calegari:
Solution 1 Here is a solution of a slightly easier problem. Let f (x) be a polynomial
with real coecients of degree  1 such that for innitely many rational numbers R,
f (R) is also rational. Prove that f (x) has rational coecients.
3

We can assume f (x) 2 Q[x]; by Lagrange's formula. We can assume f (x) 2 Z [x]; since
if the property holds for f (x) it also holds for f (ax + b); a; b 2 Q | a change of variables
then means we can assume that f (x) is monic also. Finally, if f (x) = xn + ::: + c; then
fn (x) = f (x) c n; has a rational solution, which (Gauss' lemma) says that f (Z )
contains c; c + 1; c + 2; :::; which is obvious. 2
An alternative proof is given below.
Solution 2 The solution consists of two parts. (1) Show that p(x) has rational coefcients, (which follows immediately from the solution of the simpler problem above).
(2) Show that if p(x) has rational coecients, then p(x) is linear.
(1) Consists of the following Lemma:
Lemma 1

If p(x) is rational for all but nitely many rational x; then p(x) has rational

coecients.2

by induction (on degree). If deg(p(x)) = 0; then the result is trivial. Otherwise,

let R be a rational number such that p(R) is rational. Let p(R) = C: Then
p(x) = C + (x R)q (x)
and q(x) is rational for all x for which p(x) is rational, except possibly at R; and now
we are done.
(2). Multiplying p(x) by an integer does not aect its properties. Thus we may let
Proof

( ) = anxn + ::::: + a0 ; ai 2 Z:

p x

W.L.O.G, let an be positive. Hence the image of p(x) consists (at least) of all positive
real numbers greater or equal to a0 (by Rolle's theorem). The idea is now that the
polynomials p(x) a0 ; p(x) a0 1; p(x) a0 2; :::p(x) a0 k have to have rational solutions for all x (since they have real solutions, and if they had irrational solutions then
p(x) would be rational), and this will be impossible. Note that for p(x) = "integer ";
one must have an  x an integer. Hence the possible integer values of p(x) are contained
within the set p(n=an); for n 2 Z:
If p(x) is of degree > 1; then its pretty clear that this expression can't equal every
positive integer greater than a0: The reason is that it is growing too fast, and we have
forced the range to be discrete. 2
Based on a solution provided by Garth Gaudry:
Let p(x) = a0 + a1 x +


+ anxn be a \boring" polynomial. Then
Lemma 2

The coecients of p(x) are rational.

There are various possibilities. For instance, use the fact that a polynomial
of degree n is determined by its values at n + 1 points, and write the interpolating
polynomial down. (Digression. The interpolation polynomial is the unique polynomial
which solves the following problem: Find, for a given function f (x) a polynomial of nth
degree Pn(x), which attains for n + 1 dierent values xk ; k = 0; 1; : : : ; n; of the independent variable x, the same values as the given function f (x): That is Pn(xk ) = f (xk )
for k = 0; 1; : : : ; n: The proof of uniqueness and the construction of this polynomial is
discussed in most textbooks on calculus or numerical analysis.)
Alternatively, let x = 0 to get a0 rational. We may suppose without loss of generality
from now on that a0 = 0. Then substituting x = 1; 2; : : : n 1; m where m is a
suciently large integer, we get rational values for p(x) for each such x, and hence a
system of n a system of simultaneous linear equations in n unknowns with rational
right-hand sides. The coecient matrix is nonsingular if m is large enough, (more
simply, we have n independent equations if m is large enough), and so we can solve for
a1 ; : : : ; an . They are rational. 2
Proof.

Theorem 1

A boring polynomial is linear.

Recall that we are assuming that a0 = 0; we may also assume that the
polynomial has integer coecients. Suppose that the degree of the polynomial is
greater than 1. The polynomial takes all suciently large real values (all real values if
n is odd), so the equation
u
(1)
an xn + an 1 xn 1 +


+ a1 x =
v
Proof.

has a solution for all suciently large u=v. Take v to be a prime that does not divide
an , and let u be a large integer not divisible by v . Since (1) has a solution (assuming
u is large enough), that root must be rational, say p=q , where (p; q ) = 1. This yields
the equation
an

pn

+ an
qn

1
1 qn 1
pn

+


+ a1 pq = uv

and hence
( + an 1pn 1q +


+ a1pqn 1) = uqn:
Since v is prime, coprime to u, and divides the left-hand side, it must also divide the
right-hand side, and hence q. There are at least two factors q on the right-hand side;
so after division by v, the right- hand side is still divisible by v. This implies that
v jan pn . But (p; q ) = 1, so v jan . This is a contradiction. 2 (Nobody completely solved
the second part of this question.)
v an pn