0% found this document useful (0 votes)
36 views53 pages

Isi 2022

This document provides an educative commentary on the 2022 ISI mathematics entrance exam papers. It begins with an introduction explaining the motivation and history of the commentaries. It then provides detailed commentary and analysis of questions from Paper 1, examining the concepts and approaches. The document concludes by noting some changes in the availability of solutions but maintains there is still value in the educative commentary.

Uploaded by

Varad Malpure
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
36 views53 pages

Isi 2022

This document provides an educative commentary on the 2022 ISI mathematics entrance exam papers. It begins with an introduction explaining the motivation and history of the commentaries. It then provides detailed commentary and analysis of questions from Paper 1, examining the concepts and approaches. The document concludes by noting some changes in the availability of solutions but maintains there is still value in the educative commentary.

Uploaded by

Varad Malpure
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 53

EDUCATIVE COMMENTARY ON

ISI 2022 MATHEMATICS PAPERS

(Last Revised on 24/05/2022)

Introduction 1

Paper 1 3

Paper 2 39

Concluding Remarks 52

Introduction
The original motivation for writing the annual commentaries on ISI B.Stat.
Entrance Tests is now waning. They began in 2016 because it was the year
in which some exciting questions came to the notice of the author. Later on
commentaries from 2014 were added. Unfortunately, after 2017, there was
a noticeable decline in the quality of questions. Some readers also brought
to the notice of the author that a few questions which were hailed in the
commentary were picked from some reputed international contests. This
did not detract from the intrinsic quality of the questions. Nor is it likely
that some candidates had seen the solutions already. But apart from loss of
originality, it is an insult to a good problem (for which originally ample time
was allowed) to expect the candidates to crack it in a couple of minutes.
Still, the annual commentaries on ISI Entrance Test continued more as a
mission, because some students pointed out that unlike the JEE, there are no
sources to get the answers of ISI Entrance tests. But things are changing here
too. In addition to the official key to the multiple choice questions in Paper
1, videos giving worked out solutions are available. As for the second paper,
solutions for some of the problems are presented on the Artofsolvingproblems.
But even with both the justifications for the commentaries losing their
strengths, I believe there is still some room to bring out the educative aspects
of some of the problems and solutions. So, here is the commentary on ISI
2022.
As in the other commentaries, unless otherwise stated, all the references

1
are to the author’s Educative JEE Mathematics, Third Edition, published by
Universities Press, Hyderabad.
I am thankful to the persons who made the question papers available to
me. Siddharth gave the alternate solutions to Q.3 and Q.24 in Paper 1.
Readers are invited to send their comments and point out errors, if any, in
the commentary. They may be sent either by e-mail ([email protected])
or by an SMS or a WhatsApp message on mobile (9819961036).

2
ISI BStat-BMath-UGA-2022 Paper 1 with Comments

N.B. Each question has four options of which ONLY ONE is correct. There
are 4 marks for a correctly answered, 0 marks for an incorrectly answered
question and 1 mark for each unattempted question.

Q.1 Any positive real number x can be expressed as

x = an 2n + an−1 2n−1 + . . . + a1 21 + a0 20 + a−1 2−1 + a−2 2−2 + . . .

for n ≥ 0, where each ai ∈ {0, 1}. In the above-described expansion of


21.1875, the smallest positive integer k such that a−k 6= 0 is

(A) 3 (B) 2 (C) 1 (D) 4

Answer and Comments: (A). The statement of the problem is need-


lessly elaborate. It simply gives the definition of what is popularly
called the binary or dyadic expansion of a positive real number. It
is analogous to the decimal expansion except that the base is not 10
but 2. As a result, the number x is expressed as a sum of powers of
2. The integral part of x involves only the powers with non-negative
exponents. The fractional part requires powers with negative expo-
nents. Instead of writing them as 2−1 , 2−2 , 2−3 etc. we resort to the
more familiar expressions like 21 , 41 , 18 etc. Moreover, since we are given
the number x in its decimal expansion, we shall write these negative
powers as 0.5, 0.25, 0.125 etc. for easy comparison.
Now coming to the problem itself, we ignore the integral part of x
and concentrate only on its fractional part, viz., 0.1875. It is smaller
than 0.5. Hence the first digit of dyadic expansion is 0, i.e. a−1 = 0.
Similarly, a−2 = 0 since 0.1875 is smaller than 0.25 too. However, it is
bigger than 0.125 and so a3 6= 0. Therefore k = 3.

An absolutely simple problem once a candidate deciphers the (un-


necessarily) elaborate statement and realises that what is asked is the
binary expansion of the fractional part of 0.1875. Perhaps the paper
setters wanted to test this very ability of interpreting the given state-
ment in terms of something familiar. The problem could have been
made more interesting, had the entire dyadic expansion of 0.1875 been

3
involved. To find it we subtract 0.125 from 0.1875 and expand the
1 1
difference 0.0625 in terms of further negative powers of 2, viz. 16 , 32
1
and so on. Luckily, 0.0625 equals 16 and so the dyadic expansion of the
fractional part stops at 0.0011. (Note that in the dyadic expansion the
only coefficients are 0 and 1. So we do not have to worry the highest
multiple of a particular power that is less than the earlier power.)
Q.2 Suppose for some θ ∈ [0, π2 ], cos 3θ
cos θ
= 13 . Then (cot 3θ) tan θ equals
1 1 1 1
(A) (B) (C) (D)
2 3 8 7

Answer and Comments: (D). There are well known identities which
express cos 3θ and sin 3θ as cubic polynomials in cos θ and sin θ respec-
tively. We can use the first one out of these and the data to get a
quadratic in cos θ. Solving, we can get the value of cos θ and hence
that of sin θ. We can then use the second identity to get the numerical
value of sin 3θ. From this the values tan θ and thereafter those of tan 3θ
and cot 3θ would follow. This would be a very straightforward way of
getting the answer.
But there is a short cut. The expression, say E, to be evaluated
3θ sin θ
is simply cos
cos θ sin 3θ
. The value of the first fraction is given as 31 . So,
if we can determine the second factor, or its reciprocal sin 3θ
sin θ
, we would
be through.
The identity sin 3θ = 3 sin θ − 4 sin3 θ gives
sin 3θ
= 3 − 4 sin2 θ (1)
sin θ
Thus the task is reduced to finding the value of sin2 θ, without finding
the value of sin θ. (Or else, this is hardly a short cut!). If we write
cos 3θ as cos(2θ + θ), we have
cos 3θ = cos 2θ cos θ − sin 2θ sin θ
= (1 − 2 sin2 θ) cos θ − 2 sin2 θ cos θ
= (1 − 4 sin2 θ) cos θ (2)
As we are given that cos 3θ = 13 cos θ, we get 13 = 1 − 4 sin2 θ and hence
4 sin2 θ = 1 − 31 = 32 , giving, finally, sin2 θ = 16 . Putting this into (1)
gives sin 3θ
sin θ
= 3 − 23 = 37 and hence E = 13 × 73 = 17 .

4
A simple problem based on trigonometric identities for cos 3θ and
sin 3θ. But, instead of applying these blindly, cos 3θ has been found by
writing it as cos(2θ + θ). So here is an instance where it is not the final
formula but a step in its proof that shortens the work.
Q.3 The locus of points z in the complex plane satisfying z 2 + |z|2 = 0 is

(A) A straight line


(B) a pair of straight lines
(C) a circle
(D) a parabola

Answer and Comments: (A). Every complex number z = x + iy is


uniquely determined by two real numbers, x and y. So, in theory, every
problem about complex numbers can be reduced to a problem about
their real and imaginary parts. This is not always a good idea, But
locus problems are an exception, especially when we are asked merely
to identify the nature of the locus, because till the HSC level only a
few loci are studied and they are all in terms of the coordinates.
So, if z = x + iy, then the given equation translates into

x2 − y 2 + 2ixy + x2 + y 2 = 0 (1)

As the real and imaginary parts must vanish separately, this is a pair
of two equations in x and y, viz. (i) 2x2 = 0 and (ii) xy = 0. These
two can hold simultaneously only when x = 0. So, this is the locus of
z. It is a straight line, in fact, the y-axis.

Another absolutely simple problem which can also be done by fac-


torising of the L.H.S. as (z + i|z|)(z − i|z|). Since |z| is real, vanishing
of either factor implies that z is purely imaginary and hence lies on the
y-axis, which is a straight line.
Q.4 Amongst all polynomials p(x) = c0 + c1 x + c2 x2 + . . . + c10 x10 with
real coefficients satisfying |p(x)| ≤ |x| for all x ∈ [−1, 1], what is the
maximum possible value of (2c0 + c1 )10 ?

(A) 410 (B) 310 (C) 210 (D) 1

5
Answer and Comments: (D). The fact that all the four options are
given as 10-th powers suggests that the real problem is to find the
maximum possible value of 2c0 + c1 subject to the constraint on p(x),
viz. |p(x)| ≤ |x| for all x ∈ [−1, 1]. Since p(0) = c0 , the requirement
|p(0)| ≤ |0| = 0 forces c0 to vanish. Hence p(x) = c1 x + c2 x2 + . . . +
c10 x10 . This can be factored as xq(x) where q(x) = c1 +c2 x+. . .+c10 x9 .
Then |p(x)| = |x||q(x)|. So the requirement |p(x)| ≤ |x| now gives
|x||q(x)| ≤ |x| for all x ∈ [−1, 1] and further, |q(x)| ≤ 1 for x 6= 0 since
|x| > 0 for x 6= 0. But since q(x) is a continuous function of x, we must
have |q(0)| ≤ 1 too. This means |c1 | ≤ 1. Hence the maximum possible
value of 2c0 + c1 among all polynomials of the given type is simply 1
and its 10-th power, viz. 1 is the answer.

A simple but unusual problem. Getting c0 = 0 is easy. A rigorous


proof of |c1 | ≤ 1 requires continuity of q(x) at x = 0. Candidates who
miss this are rewarded in terms of time. Ignorance is bliss in an MCQ.

Q.5 Let ZZ denote the set of all integers. Let f : ZZ −→ ZZ be such that
f (x)f (y) = f (x + y) + f (x − y) for all x, y ∈ ZZ. If f (1) = 3, then f (7)
equals

(A) 840 (B) 844 (C) 843 (D) 842

Answer and Comments: (C). The given relationship among the val-
ues assumed by f , viz.

f (x)f (y) = f (x + y) + f (x − y) (1)

for all x, y ∈ ZZ is an example of what is called a functional equation.


They are dealt with in Chapter 20. But because of the huge variation
in the types of relations, there is not much of a coherent theory for
for solving functional equations involving two variables, as there is, for
example, for recurrence relations with only one variable (e.g. A.P.’s, the
G.P.’s or the Fibonacci relation). So every problem requires methods
tailor made for it. In the present case we notice that if we interchange
x and y, only the last term in (1) changes and we get

f (x − y) = f (y − x) (2)

6
for all x, y ∈ ZZ. Since any integer can be expressed as a difference of
two integers, we get that f is an even function.
We are further given that f (1) = 3. Putting y = 1 in (1), we have

3f (x) = f (x + 1) + f (x − 1) (3)

for all x ∈ ZZ. If we were given f (1) as 2 instead of 3, then this equation
would mean that f (x) is an A.P., a most simple linear recurrence rela-
tion of order 2. Even as (3) stands, it is a linear recurrence relation of
order 2. To solve it we need to know two initial values. In the present
problem we are given, only one, viz. f (1) = 3. But, putting x = 0 into
(3), and using that f (−1) = f (1) = 3, we get f (0) = 2.
Replacing x by x − 1, we can rewrite (3) as

f (x) = 3f (x − 1) − f (x − 2) (4)

for all x ∈ ZZ. For those who are familiar with solving linear recurrence
relations with constant coefficients,
√ the characteristic polynomial here
3± 5
is λ2 −3λ+1. Its roots, viz. are the characteristic roots. Hence
2
the general solution of (4) is
√ √
3+ 5 x 3− 5 x
f (x) = A( ) + B( ) (5)
2 2
where A and B are some constants to be determined from the initial
conditions f (0) = 2 and f (1) = 3. Once this is done, f (7) can be found
by a mere substitution.
But since 7 is a small number, such an elaborate procedure can be
avoided by simply calculating f (2), f (3), f (4), f (5), f (6) and f (7) one
by one. So,

f (2) = 3f (1) − f (0) = 9 − 2 = 7


f (3) = 3f (2) − f (1) = 21 − 3 = 18
f (4) = 3f (3) − f (2) = 54 − 7 = 47
f (5) = 3f (4) − f (3) = 141 − 18 = 123
f (6) = 3f (5) − f (4) = 369 − 47 = 322

Hence, finally, f (7) = 3f (6) − f (5) = 966 − 123 = 843.

7
The calculations above are like a passenger train from 1 to 7 which
halts at every intermediate station 2,3,4,5 and 6. We could instead
take an express train halting only at 3 and 5, assuming that we have
already calculated f (2) as 7. Note that (3) was obtained by putting
y = 1 in (1). If, instead, we put y = 2, we get

7f (x) = f (x + 2) + f (x − 2) (6)

Starting from x = 1, we get 21 = f (3) + f (−1) = f (3) + 3. Hence


f (3) = 18. So far, there is not much saving. But now put x = 3 into
(6) to get 7f (3) = f (5) + f (1) which gives f (5) = 7 × 18 − 3 = 123.
Finally, knowing f (5) and f (3), we get f (7) quickly by putting x = 5.
Of course, there is a super express train if we put y = 3 in (1) to get

18f (x) = f (x + 3) + f (x − 3) (7)

Now the only intermediate stop needed is f (4) = 47.


A very good problem on functional equations where the functional re-
lation involves two variables. Reduction to a functional equation with
only one parameter itself involves a trick. The paper setters hardly
expected the general solution in the form (5). Instead, repeated appli-
cations of the recurrence relation gives the answer. But this could have
been tested equally well even with finding f (3) or f (4). Calculations
beyond that are repetitious and prone to error, even if we take the
express train above.
Overall, the question is more suited as a full length question in Paper
2.

Q.6 Let A and B be two 3 × 3 matrices such that (A + B)2 = A2 + B 2 .


Which of the following must be true?

(A) A and B are zero matrices.


(B) AB is the zero matrix.
(C) (A − B)2 = A2 − B 2 .
(D) (A − B)2 = A2 + B 2

8
Answer and Comments: (D). As matrix multiplication is not com-
mutative, the correct expansion of (A + B)2 is (A + B)(A + B) =
A2 + AB + BA + B 2. So the hypothesis means that AB + BA = O, the
zero matrix. Neither of the first two options would follow conclusively
from this. The other two options deal with (A−B)(A−B) which comes
out to be A2 −AB −BA+B 2 . This equals A2 +B 2 since AB +BA = O.

Another trivial problem.

Q.7 Let {n1 , n2 , . . . , n12 } be a permutation of the numbers 1, 2, . . . , 12. The


number of arrangements with

n1 > n2 > n3 > n4 > n5 > n6

and
n6 < n7 < n8 < n9 < n10 < n11 < n12
equals
     
12 12 11 11!
(A) 5
(B) 6
(C) 6
(D) 2

Answer and Comments: (C). Clearly n6 is the smallest element of


the set {1, 2, . . . , 12}, viz. 1. The elements n1 , n2 , n3 , n4 and n5 can
be any five of the remaining 11 elements. They can be chosen in 11 5
ways. Once the choice is made, they must be ordered in a descending
order. And after that the remaining six elements have to be ordered in a
strictly increasing order to get n7 to n12 . Summing up, the five elements
chosen from the set {2, 3, . . . , 12} uniquely determine the permutation.
Hence
  the number of permutations   which satisfy the given requirements
11
is 5 which is the same as 11 6
.

A simple problem once the idea that n6 = 1 and that the choice of
the subset {n1 , . . . , n5 } completely determines the permutation strikes.

Q.8 The sides of a regular hexagon ABCDEF are extended by doubling


them to form a bigger hexagon A′ B ′ C ′ D ′ E ′ F ′ as in the figure below.

9
F’
A’

A
60

B F

B’ E’

C E

C’ D’

Then the ratio of the area of the bigger to the smaller hexagon is
√ √
(A) 3 (B) 3 (C) 2 3 (D) 4

Answer and Comments: (B). Being regular, both the two hexagons
are similar to each other. Hence the ratio of their areas is the square
of the ratio of their corresponding sides. It is tempting to think that
this ratio is 2 (and hence that the ratio of the areas is 4) because
the bigger hexagon is obtained by doubling the sides of the original
hexagon. But the catch is that we are only extending the sides of the
original hexagon. These extended segments are not the sides of the
bigger hexagon. For example, in the figure above, A′ F is twice the
side AF of the smaller hexagon. But it is not a side of the bigger
hexagon. Instead, A′ F ′ is a side. Its length can be determined from
the right angled triangle A√′ F ′ F √
in which the angle at F is 60◦ . Hence
′ ′ ′ ◦ 3
A F = A F sin 60 = 2AF 2 = 3AF . The ratio of the corresponding

sides being 3, that of the areas is 3.

The paper setters have done well to give a diagram to expel the
possible misconception that the sides of the bigger hexagon are double
those of the smaller one. Calculating the ratio of the sides is very easy.
The only catch is to realise that the ratio of the areas is the square of
the ratio of the corresponding sides.

10
Q.9 In how many ways can we choose a1 < a2 < a3 < a4 from the set
{1, 2, . . . , 30} such that a1 , a2 , a3 , a4 are in arithmetic progression ?

(A) 135 (B) 145 (C) 155 (D) 165

Answer and Comments: (A). Let d be the common difference of the


A.P. Then a4 − a1 = 3d which is a multiple of 3. In order that both
a1 and a4 are in the set {1, 2, . . . , 30}, 3d can be at most 27. Hence
the possible values of d are 1, 2, . . . , 9. If d = 1, then the possible
values of the initial term a1 are from 1 to 27. For d = 2, they are
from 1 to 24. Continuing, the total number of desired progressions is
27 + 24 + 21 + 18 + 15 + 12 + 9 + 6 + 3. This itself is the sum of an
A.P. and equals 3(1 + 2 + . . . + 9) = 3 9×102
= 135.

A simple counting problem. It could have been made a little more


interesting by asking the probability that four distinct integers picked
at random from the set {1, 2, . . . , 30} form an A.P.
Q.10 Suppose the numbers 71, 104 and 159 leave the same remainder r when
divided by a certain number N > 1. Then the value of 3N + 4 is

(A) 53 (B) 48 (C) 37 (D) 23

Answer and Comments: (A). The hypothesis implies that there are
some integers a, b, c such that

71 = aN + r (1)
104 = bN + r (2)
and 159 = cN + r (3)

The first two equations give 33 = (b − a)N. Similarly, the second and
the third equations give 55 = (c − b)N. Hence N is a common divisor
of 33 and 55. Since N > 1, we must have N = 11. The first equation
along with the restriction on the remainder, viz. that 0 ≤ r ≤ 10 gives
a = 6 and r = 5. There is no point in trying the other two equations,
since they will give the same value of r. Since N = 11 and r = 5, we
have 3N + 4r = 33 + 20 = 53.

A very simple number theory problem. In fact, a disappointment on

11
the backdrop of some of the past years when more challenging questions
based on divisibility and the euclidean division algorithm were asked.
( e.g. see Problem 13 of Paper 1 of last year.)

Q.11 If x, y are positive real numbers such that 3x + 4y < 72, then the
maximum possible value of 12xy(72 − 3x − 4y) is :

(A) 12240 (B) 13824 (C) 10056 (D) 8640

Answer and Comments: (B). Positive constant coefficients have lit-


tle role in problems of maximisation. The paper-setters could as well
have dropped 12 and asked to maximise xy(72 − 3x − 4y). If nothing
else, at least the options would not have looked so horrendous. So,
perhaps there is some purpose in retaining it. And indeed there is,
if we recognise that 12xy is the product of 3x and 4y, about which
something is given, viz. that their sum is less than 72.
So, if we let u = 3x, v = 4y and w = 72 − 3x − 4y, then u, v, w
are positive real numbers which add to 72 and we have to find the
maximum value of their product uvw. By the A.M.-G.M. inequality,
uvw is maximum when u, v, w are equal. This gives u = v = w = 24
and the maximum value of the product is (24)3 . The paper-setters have
given one more reward. Even without calculating it fully, (24)3 has 4
as its last digit. Out of the four options, only (B) has this property. So
the right answer must be 13824.

The problem amounts to maximising a function of two variables, viz.


12xy(72−3x−4y) subject to the inequality constraints x > 0, y > 0 and
3x + 4y < 72. Doing this using calculus falls outside the scope of HSC.
But with the A.M.-G.M. inequality, the problem becomes simple. The
paper setters deserve to be commended for giving a veiled hint through
the factor 12 which is otherwise redundant and further for giving only
one option whose last digit matches that of (24)3 , thereby sparing the
drudgery of calculating it.

Q.12 What is the minimum value of the function |x−3|+|x+2|+|x+1|+|x|


for real x?

(A) 3 (B) 5 (C) 6 (D) 8

12
Answer and Comments: (C). Call the function as f (x). Then f (x)
is continuous everywhere and differential at all points except −2, −1, 0
and 3. Clearly, f (x) → ∞ as x → ±∞. Hence f (x) has no maximum.
Its minimum must occur at a critical point. If we sketch the graph
of f (x), we can tell its minimum value. But it is very time consum-
ing to do so because the formula for f (x) changes over the intervals
(−∞, −2], [−2, −1], [−1, 0], [0, 3] and [3, ∞). But on each of these in-
tervals it is a linear function of x, i.e. a function of the form Ax + B
for some constants A and B (which are generally different for the five
intervals). The maximum as well as the minimum of f (x) over each
such interval occurs at an end point. (If A = 0 for some interval, then
f (x) is constant over that interval and the extrema also occur at the
end points.)
Summing up, even without drawing the graph of f (x), we know that
its minimum must occur at one of the points −2, −1, 0 and 3. By a
direct calculation, f (−2) = 2 + 1 + 5 = 8, f (−1) = 1 + 1 + 4 = 6, f (0) =
1 + 2 + 3 = 6 and f (3) = 3 + 4 + 5 = 12. Hence the minimum occurs at
both −1 and 0. (In fact, it occurs at every point in the interval [−1, 0]
since the function is constant over this interval.) The minimum value
of f is f (−1) = f (0) = 6.

The search for the minimum of a non-negative continuous function


f (x) on IR narrows down to its critical points. These are the points
where f ′ (x) either does not exist or exists but vanishes. In the present
problem, the latter possibility is taken care of by the constancy of f (x)
in a neighbourhood of such points. Once again, the MCQ format fails to
distinguish whose who can give this justification from those who don’t
even realise that a justification is needed. So this question deserves a
better place in the second paper.

Q.13 Consider a differentiable function u : [0, 1] −→ IR. Assume the function


u satisfies
1 a+r
Z
u(a) = u(x)dx, for all a ∈ (0, 1) and all r < min{a, 1 − a}.
2r a−r

Which of the following statements must be true?

(A) u attains its maximum but not its minimum on the set {0, 1}.

13
(B) u attains its minimum but not maximum on the set {0, 1}.
(C) If u attains either its maximum or its minimum on the set {0, 1},
then it must be a constant.
(D) u attains both its maximum and its minimum on the set {0, 1}.

Answer and Comments: (D). The property which u is given to sat-


1 a+r
Z
isfy may appear clumsy. The expression u(x)dx is called the
2r a−r
average value of u(x) over the interval [a − r, a + r]. More generally,
the average or the mean value of a continuous real valued function
1
Z b
f (x) over an interval [a, b] is defined as f (x)dx. It is the con-
b−a a
tinuous analogue of the average value of a real valued function defined
on a finite set and shares some common properties. For example, it
always lies between the minimum and the maximum of the function
over that interval. The average value of a real valued function on a
finite set may not always be attained. (In a class of students, there
may be no student whose height is exactly the mean height for that
class.) In sharp contrast, a continuous function f (x) always attains its
average. This follows by applying the Lagrange’s Mean Value theorem
to any antiderivative F of f and justifies the name mean value.
Coming to the problem, we are given that the average value of u
over any interval contained in (0, 1) occurs at the midpoint of that in-
terval. This is a very strong hypothesis and from it we can show that
u is a linear function, i.e. a function of the form u(x) = Ax + B
for some constants A and B. Depending on the sign of A, u at-
tains its maximum on [0, 1] at one of the end points and minimum
at the other. That is, (D) holds. But the proof of linearity of u is
non-trivial. (The converse is trivial, because if u(x) = Ax + B, then
A a+b
Z b
u(x)dx = (b2 − a2 ) + B(b − a) = (b − a)u( ).)
a 2 2
As our goal is merely to prove (D), we can give a simpler argument
based merely on continuity of u. u has a maximum, say M, at some
point a ∈ [0, 1]. If either u(0) or u(1) equals M, we are done. So
assume that both u(0) < M and u(1) < M. Choose ǫ > 0 such that
u(0) < M − ǫ and u(1) < M − ǫ. By continuity of u at 0 and 1, there
exists some h > 0 such that u(x) < M − ǫ/2 for all x ∈ [0, h] and

14
also for all x ∈ [1 − h, 1]. Now let r = min{a − h2 , 1 − h2 − a}. Then
[a − r, a + r] ⊂ (0, 1). The situation is shown in the figure below where
it is assumed, without loss of generality, that a lies in the left half of
[0, 1] and hence a − r = h2 . The interval [a − r, a + r] is shown in thick.

M− ε / 2
M− ε

a −r a +r
0 h/ 2 h 1
a

Z a+r
Now let I = u(x)dx. Then u(x) ≤ M for all x ∈ [a − r, a + r] and
a−r
hence I ≤ 2rM. We split the interval of integration at the point h. The
first part is of length h2 while the second part is of length 2r − h2 . On
Z h
h ǫ
the first subinterval, u(x) ≤ M − 2ǫ . Hence u(x)dx ≤ (M − ).
h/2 2 2
h
The integral of u(x) over the second interval is at most (2r − 2 )M.
Adding,
h ǫ h
I ≤ (M − ) + (2r − )M
2 2 2

= 2rM − (1)
4
So, the average value of u over the interval [a − r, a + r] is M − hǫ 8r
.
This is less than M which equals u(a) by assumption. Thus we get a
contradiction. This proves that u attains its maximum at an end point
of [0, 1]. By a similar reasoning it attains its minimum at an end point.
Hence (D) is true.

15
The essential idea behind the argument above is merely that if
M is an upper bound on a continuous integrand over an interval of
length L, then the integral is at most ML. But even if the upper
bound on the integrand is strict at one point then by continuity it is
so over a subinterval of positive length and therefore the integral is
less than ML. Using essentially the same argument, we can prove a
slightly stronger result that u(x) is either monotonically increasing or
decreasing. This is, of course, a far cry from proving its linearity. Note
that differentiability of u has never been used. The purpose of giving
it in the problem is not clear. Perhaps with it the linearity of u can be
proved. At any rate, this is an excellent problem on theoretical calculus
and more suited for Paper 2.
Q.14 A straight road has walls on both sides of heights 8 feet and 4 feet
respectively. Two ladders are placed from the top of one wall to the
foot of the other as in the figure below. What is the height (in feet) of
the maximum clearance x below the ladders?
A

C
8 E
4
x

B a b D
F
√ 8
(A) 3 (B) 2 2 (C) 3
(D) 1

Answer and Comments: (C). It might appear at first sight that


there is no unique answer since the width of the road, i.e. the distance
between the two walls, is not specified. So let us try to solve the problem
with some unknown parameters. Call the walls as AB and CD. Let
the ladders meet at E and EF be the vertical clearance. Denote BF
and F D by a and b respectively. From the similar triangles BEF and
BCD, we get
x a
= (1)
4 a+b

16
Similarly, the similar triangles EF D and ABD give
x b
= (2)
8 a+b
Adding these two equations,
x x
+ =1 (3)
4 8
which simplifies to 12x = 32 and gives x = 83 .

An absolutely simple problem requiring nothing more than elemen-


tary properties of similar triangles.
Q.15 Let y = x + c1 , y = x + c2 be two tangents to the ellipse x2 + 4y 2 = 1.
What is the value of |c1 − c2 | ?
√ √ √
(A) 2 (B) 5 (C) 25 (D) 1
.

Answer and Comments: (B). We could begin by casting the equa-


tion of the ellipse in the standard form and then tapping on the right
tools such as the formulas for equations of tangents or representing the
ellipse in a parametric form. But it is often better to resort to general
principles and proceed directly. The slope of the tangent at a point
(x0 , y0 ) on any curve whose equation is given in terms of x and y is the
dy
value of dx at that point. Here y is not given as an explicit function of
x. But we can apply implicit differentiation and get
dy
2x + 8y =0 (1)
dx
dy
If (x1 , y1 ) is the point of contact of the tangent y = x + c1 , then dx at
this point is 1 and we get x = −4y. Solving this simultaneously with
x2 + 4y 2 = 1, we have 20y 2 = 1. Hence y = ± 2√1 5 . Correspondingly,
x = ∓ √25 . So, the points of contact of the two tangents are ( √ −2
, √1 )
5 2 5
and ( √25 , − 2√1 5 ). Call the first one as (x1 , y1 ) and the second one as
(x2 , y2 ). Then the equation of the tangent at (x1 , y1) is
1 2
y− √ =x+ √ (2)
2 5 5

17
This gives c1 = 2√5 5 . By a similar calculation, c2 = − 2√5 5 . Therefore

|c1 − c2 | = √55 = 5.

A simple problem on conics which can be solved more easily through


the general principles. In fact, the knowledge that the given curve is
an ellipse is never needed.

Q.16 In the figure below, ABCD is a square and △CEF is a triangle with
given sides inscribed as in the figure. Find the length of BE.

A E x B

3
β
5
d

y
4 α

β
d C
D

(A) √13 (B) √14 (C) √15 (D) √16


17 17 17 17

Answer and Comments: (A). Whenever a problem asks to find the


numerical value of something, the standard procedure is to denote the
quantity by an unknown (often x), get an equation for x and solve it.
But usually things are not so simple. Often we have to introduce some
other auxiliary unknowns, y, z etc. and get a system of equations in
all these unknowns. We solve it even though we want the value of only
one of the unknowns.
Art lies in choosing these auxiliary variables. In the present problem,
we can begin by calling BE as x. We can take the side of the square
as d and the distance F D as y. The three right-angled triangles with
hypotenuses along the sides of the given triangle CEF will give a system

18
of three equations in the three unknowns x, y and d. We can then solve
it. This will be a mechanical way of attempting the solution. And
its success will depend upon how easy it is to solve this system. In
the present problem, the system is algebraic but non-linear and ad-hoc
methods will be needed to solve it.
A more imaginative way is to use some elementary trigonometry.
Instead of using the Pythagoras theorem, we can introduce suitable
angles and get a system in terms of these angles. So we let α = 6 BCE.
Then we have

x = 5 sin α (1)

If we can get hold of some trigonometric equation for α, we can get


sin α and thence x from it. But there is no such obvious equation. So
we introduce an auxiliary variable. Specifically, we let β = 6 DCF .
Equating the two expressions for the side d, we get

5 cos α = 4 cos β (2)

As there are two unknowns, we need one more equation relating α


and β. It is tempting to get this by applying the Pythagoras theorem
to △F AB, since we have AE = d − 5 sin α = 5(cos α − sin α) and
F A = d − 4 sin β = 4(cos β − sin β). But squaring and adding will give
a complicated equation in α and β and then this method would be as
laborious as the algebraic one.
There is an ingenious way out if we observe that the triangle EF C
is right angled at F . Hence 6 AF B also equals β. Therefore, AF =
3 cos β. Adding this to F D we get d and hence a trigonometric equation
for β, viz.

3 cos β + 4 sin β = d = 4 cos β (3)

This gives 4 sin β = cos β. Hence tan β = 41 , whence sin β = √1


17
and
cos β = √417 . From (2), we now have

16
5 cos α = √ (4)
17

19
Hence

x = 5 sin α = s25 − 25 cos2 α
256
= 25 −
17
s s
425 − 256 169
= = (5)
17 17

which comes to √13 .


17

An excellent problem testing the ability to select the right variables


taking advantage of the specifics of the data. In the present problem
△EF C is not just some triangle, but a right angled one and this led
to a trigonometric equation in β.

Q.17 Let p and q be two non-zero polynomials such that the degree of p is less
than or equal to the degree of q and p(a)q(a) = 0 for a = 0, 1, 2, . . . , 10.
Which of the following must be true?

(A) degree of q 6= 10
(B) degree of p 6= 10
(C) degree of q 6= 5
(D) degree of p 6= 5

Answer and Comments: (C). Let m and n be the degrees of p and


q respectively. Then the degree of the product pq is m + n. We are
given that pq has at least 11 distinct roots, viz. 0, 1, 2, . . . , 10. Hence
the degree of pq, i.e. m + n is at least 11. We are further given that
m ≤ n. So, the larger summand, viz. n is at least 6. Hence (C) is
necessarily true. The falsity of the other options can be proved with
easy counterexamples. But that is hardly necessary because of the
stipulation that exactly one of the four given options is true.

A shockingly simple problem requiring only that the degree of a


product of polynomials is the sum of their degrees and that the number
of roots of a polynomial cannot exceed its degree.

20
Q.18 For n ∈ IN , let an be defined by
Z n 1
an = dx.
0 1 + nx2

Then lim an
n→∞

(A) equals 0 (B) equals π4


(C) equals π2 (D) does not exist.

Answer and Comments: (A). √ The integral can be evaluated explic-


itly with the substitution y = nx. Then dx = √1n dy and we get

1 n n 1
Z
an = √ dy
n 0 1 + y2

1 n n
= √ tan−1 y
n 0
−1

tan (n n)
= √ (1)
n

As n → ∞, n n → ∞ and so the numerator tends to π2 . But the
denominator tends to ∞. Hence the ratio tends to 0. So, lim an = 0.
n→∞

A very simple problem since the integral can be evaluated explicitly.


Limits of integrals with an integer parameter n are interesting when
the integral cannot be evaluated explicitly but can only be estimated.
See, for example, Exercise (18.18).

Q.19 A 3 × 3 magic square is a 3 × 3 rectangular array of positive integers


such that the sum of the three numbers in any row, any column or any
of the two major diagonals is the same. For the following incomplete
magic square

21
27 36

31

the column sum is

(A) 90 (B) 96 (C) 94 (D) 99

Answer and Comments: (B). Looks like a typical number puzzle in


the recreational section of a newspaper. Anyway the problem is purely
mathematical. There are 9 entries of which 3 are given. The remaining
6 can be found by writing and solving a system of equations. As the
only operations involved are addition and subtraction, the system is
linear. But instead of writing it elaborately, we can solve parts of it
one-by-one. For example, since the first row and the first column both
add to the same number, if we call the last entry in the first row as
x then the middle entry in the first column must be x + 5. Similarly,
equality of 27 + 36 + x with the sum of the entries in the non-principal
diagonal gives the central entry of the square as 32 and we get a picture
below

27 36 x

x+5 32 z

31 u v

where z, u, v are the remaining three unknowns. Equating the first two
row sums determines z as 26. Similarly equality of the first two column

22
sums determines u as x − 5. Finally, since both the diagonals have
equal sums we get v = x + 4 and we get the more filled picture

27 36 x

x+5 32 26

31 x−5 x+4

where now there is only one unknown, viz. x. To find its value, there
is no point in using equalities which are already used. But if we equate
the first row sum with the third, we get x + 63 = 2x + 30 which gives
x = 33. All the sums now are 96 each.

A conceptually simple problem which is best done mentally on a


piece of paper. The integrality of the entries was never used. The
most general 3 × 3 magic square has 9 unknown entries which are best
denoted as the entries aij of a 3 × 3 matrix. There are in all 8 sums
(three row sums, three column sums and two diagonal sums) and their
mutual equality gives a system of 7 homogeneous linear equations in
the 9 unknowns, viz. the aij ’s. It would thus appear superficially that
this system has 9 − 7, i.e. 2 degrees of freedom. In simpler terms,
this means that any two entries of a 3 × 3 magic square determine it
uniquely. But that is not correct. There is some redundancy in these 7
linear equations. For example, equality of each of the three row sums
along with their equality with the first two column sums automatically
implies that the third column sum also equals this common sum. The
corresponding equation gives no new information. That reduces the
number of ‘really different’ equations by 1 and increases the degrees of
freedom from 2 to 3. So, 3 entries are needed to determine a magic
square of order 3 uniquely. That is why in the present problem, three
entries of the magic square are given. (The loose terminology of ‘really
different’ equations and degrees of freedom can be made precise with
the concept of the rank of a system of linear equations. But that is
beyond the HSC level.)

23
Q.20 The number of positive integers n less than or equal to 22 such that 7
divides n5 + 4n4 + 3n3 + 2022 is

(A) 7 (B) 8 (C) 9 (D) 10

Answer and Comments: (D). The number 2022 is just a showpiece,


catering to the modern trend of giving at least one problem whose
statement involves (often only superficially) the number of the calendar
year in which a particular test is given. When divided by 7, 2022 leaves
the remainder 6. So, the requirement about n is equivalent to saying
that n5 + 4n4 + 3n3 leaves a remainder of 1 when divided by 7. In a
somewhat sophisticated language (given in Comment No. 15 of Chapter
4), n5 + 4n4 + 3n3 is congruent to 1 modulo 7. Or, in other words, the
residue class of n5 + 4n4 + 3n3 modulo 7 is 1.
Using elementary properties of residue classes, this can be further
simplified by factoring n5 + 4n4 + 3n3 as n3 (n2 + 4n + 3) and further
as n3 (n + 1)(n + 3). The residue class modulo 7 of a sum and of a
product are simply the sum and the product of the respective classes.
So, if we denote n3 (n + 1)(n + 3) by an , then the residue class of a1
modulo 7 is 13 × 2 × 4 = 8 = 1 since modulo 7, 8 is congruent to 1.
Similarly, for n = 2, the residue of a2 modulo 7 is 8 × 3 × 5 = 15 = 1.
For n = 3, a3 has the residue class of 27 × 4 × 6 and hence that of
(−1)4(−1) = 4. The next case, viz. a4 is much simpler, because when
n = 4, n + 3 = 7 and so n3 (n + 1)(n + 3) is divisible by 7, whence
modulo 7 it is congruent to 0 and not to 1. For a similar reason, a6 and
a7 are congruent to 0 modulo 7. As for a5 , its congruence class modulo
7 is that of 125 × 6 × 8 and hence of 6 × 6 × 1 = 1.
Summing up, out of a1 to a7 only three, viz. a1 , a2 and a5 fit the
table. It would be torturous to calculate the residue classes modulo 7
of an for all n from 8 to 22. Fortunately, that is not necessary. Note
that an+7 has the same residue class modulo 7 as an because n + 7 is
congruent to n modulo 7. So once we determine for how many values
of n between 1 to 7, n divides n5 + 4n4 + 3n3 , the number would be the
same for the next set of 7 integers, viz. from 8 to 14, and then again
the next one, viz. the integers from 15 to 21. Only the case n = 2.
But again, a22 is congruent to a1 which is congruent to 1 modulo 7.
So, in all the number of integers n from 1 to 22 for which 7 divides

24
n5 + 4n4 + 3n3 is 3 + 3 + 3 + 1 = 10.

A good number theory problem based on residue arithmetic. The


book keeping work could have been avoided. The essential idea of the
problem is merely that the residue class of n modulo 7 determines that
of n5 + 4n4 + 3n3 . And this could have been tested by asking to identify
which of some given integers has the desired property.

Q.21 In a class of 45 students, three students can write well using either
hand. The number of students who can write well only with their right
is 24 more than the number of those who write well only with their left
hand. Then the number of students who can write well with their right
hand is

(A) 33 (B) 36 (C) 39 (D) 41

Answer and Comments: (B). A typical problem on the cardinalities


of various subsets of a finite set. Let S, R, L and B be, respectively,
the sets of all students, those who can write well only with their right
hand, those who can write well only with their left hand, and those who
can write well with either hand. Let s, r, l and b denote their respective
cardinaities. Then S is the union of the three mutually disjoint subsets
R, L, B. Hence s = r + l + b. We are given that s = 45 and b = 3.
That gives r + l = 45 − 3 = 42. It is also given that r − l = 24. Hence
2r = 42+24
2
= 33. The question asks for r + b which equals 33 + 3 = 36.

Another shockingly simple problem. With only two subsets involved,


there is not much point in drawing a Venn diagram either. It is not
given explicitly that every student can write with at least one hand.
This is left to common sense.

Q.22 Let 11, ω, ω 2 be the cube roots of unity. Then the product
2 2 3 9 10
(1 − ω + ω 2)(1 − ω 2 + ω 2 )(1 − ω 2 + ω 2 ) . . . (1 − ω 2 + ω 2 )

is equal to

(A) 210 (B) 310 (C) 210 ω (D) 310 ω 2

25
Answer and Comments: (A). All powers of ω are roots of unity.
They recur in cycles of length 3. Thus ω 4 = ω, ω 5 = ω 2 etc. If α = ω k ,
then α = 1 if and only if k is a multiple of 3. Otherwise it is a
complex cube root of unity and satisfies 1 + α + α2 = 1 and hence
1 − α + α2 = −2α.
The r-th factor of the product, say P , is 1 − ω 2 −1 + ω 2 . This is of
r r

the form 1 − α + α2 where α = ω 2 . Since 2r−1 is not a multiple of 3,


r−1

we get that the r-th factor equals −2α = −2ω 2 . Therefore we have
r−1

10
P = (−2)10 (ω 2 )2
Y r−1

r=1
10
= 210 ω2
Y r

r=1
10 k
= 2 ω (1)

where
10
2r
X
k =
r=1
2(210 − 1
= (2)
2−1
using the formula for the sum of a G.P. with common ratio 2. Since
210 −1 = 1023 is divisible by 3, k is a multiple of 3 and therefore ω k = 1.
Hence the product P equals 210 .

A good problem involving properties of ω. Such problems have


become routine because many problems involving powers of ω have
been asked in so many examinations. It would have been better to
replace ω with a fifth or sixth root of unity. Then the same principles
would apply as for ω but the calculations would demand some fresh
work.

Q.23 The function x2 loge x in the interval (0, 2) has:

(A) exactly one point of local maximum and no points of local mini-
mum.

26
(B) exactly one point of local minimum and no points of local maxi-
mum.
(C) points of local maximum and as well as points of local minimum.
(D) neither a point of local maximum nor a point of local minimum.

Answer and Comments: (B). Let f (x) = x2 ln x. Then f is unde-


fined at 0. But on the interval (0, 2) it is defined and differentiable
everywhere. Further,

f ′ (x) = 2x ln x + x = x(2 ln x + 1) (1)

which vanishes when ln x = − 21 , i.e. when x = e−1/2 = √1e . Further,


f ′ (x) > 0 for x > √1e and f ′ (x) < 0 for x < √1e . Hence f is decreasing
on (0, √1e ) and increasing on ( √1e , 2). Therefore √1e is a point of local
minimum. As there are no other critical points, the correct answer is
(B).

A straightforward problem since the end points 0 and 2 are not


included and the question deals with only local extrema. If 0, were
included in the domain, then as x → 0+ , ln x → −∞, but because of
the factor x2 , the product x2 ln x tends to 0.

Q.24 The number of triples (a, b, c) of positive integers satisfying the equation
1 1 1 2
+ + =1+
a b c abc
and such that a < b < c equals:

(A) 3 (B) 2 (C) 1 (D) 0

Answer and Comments: (C). If a, b, c were real numbers, then a


single equation in three variables will generally have infinitely many
solutions. But the stipulation that a, b, c are positive integers puts a
strong restriction on the solutions. For example the equation xy =
10 represents a hyperbola with infinitely many points on it. But if
x, y can take only positive integers as values then (x, y) can only be
(10, 1), (1, 10), (2, 5) and (5, 2). Similarly, x + y + z = 1 is an infinite

27
plane, but if x, y, z are non-negative integers, then the only possible
solutions are (1, 0, 0), (0, 1, 0) and (0, 0, 1).
So, somewhere in the solution, we will need the restriction that a, b, c
are positive e integers. But before invoking it, let us rewrite the given
equation as

ab + bc + ca = abc + 2 (1)

To recast this further we note the expansion

(1 − a)(1 − b)(1 − c) = 1 − (a + b + c) + ab + bc + ca − abc (2)

Adding 1 to both the sides of (1) we can rewrite it as

1 + ab + bc + ca − abc = 3 (3)

which, in view of (2) becomes

(1 − a)(1 − b)(1 − c) = 3 − (a + b + c) = (1 − a) + (1 − b) + (1 − c) (4)

or equivalently,

(a − 1)(b − 1)(c − 1) = (a − 1) + (b − 1) + (c − 1) (5)

We have not yet solved the equation. We have merely recast it. But
now we are close to the solution. Let us call a − 1, b − 1, c − 1 as x, y, z
respectively. Then x, y, z are non-negative integers. Moreover, a <
b < c is equivalent to x < y < z. Thus the problem is now reduced to
finding the number of triples (x, y, z) of non-negative integers satisfying
the equation

xyz = x + y + z (6)

and such that x < y < z. This is a much more manageable task. Note
that x cannot be 0. For it it is, then by (6), x+ y + z would be 0 forcing
x = y = z = 0. But then the inequality x < y < z fails. So x and
therefore y and z are positive integers.
(1, 2, 3) is an obvious solution of (6). We claim that it is the only
solution. In any solution (x, y, z), we have 1 ≤ x < y < z. Hence
x + y + z < 3z. That means xyz < 3z which implies xy < 3 as z > 0.

28
As x, y are integers, this means xy ≤ 2. But the only way this can
happen with 1 ≤ x < y is if x = 1 and y = 2.
So, there is only one solution of (6) and hence of the original equation
(1) too. (The actual solution is a = 2, b = 3, c = 4.)

A tricky problem, where the success depends upon the conversion of


(1) to (6). Last year Q.15 was also a problem where the requirement
that max{a, b, c, d} = a2 + b2 + c2 + d2 on the variables a, b, c, d ∈ [0, 1]
reduced the solution set from an infinite to a finite one. And that
problem was tricky too.
A more methodical solution is also possible. As a, b, c are positive
integers with a < b < c, they cannot be very large becuase in that case
their reciprocals will be so small that they will add to something less
than 1. That shows that the number of solutions is finite and we can
find them by giving possible values to a.
For a = 1 the equation reduces to 1b + 1c = bc2 , which simplifies to
b + c = 2. But this is impossible since b ≥ 2 and c ≥ 3. So there are
no solutions with a = 1.
Suppose a = 2. Then the equation reduces to 1b + 1c = 21 + bc1 which
simplifies to 2(b + c) = bc + 2. This can be recast as 2(b − 1)(c − 1) = bc.
So c − 1 divides bc. But c and c − 1 are relatively prime. So we must
have that c − 1 divides b. But c − 1 ≥ b. This is possible only when
c − 1 = b. Then the equation reduces to 2b − 2 = b + 1 which gives
b = 3 and hence c = 4. Thus (2, 3, 4) is a solution.
For a ≥ 3, we must have b ≥ 4 and c ≥ 5. But then a1 + 1b + 1c ≤
1
3
+ 14 + 15
= 47
60
< 1. Hence the given equation cannot hold as the second
term on the R.H.S. is positive.
Summing up, (a, b, c) = (2, 3, 4) is the only solution.

Q.25 An urn contains 30 balls out of which one is special. If 6 of these balls
are taken out at random, what is the probability that the special ball
is chosen?
1 1 1 1
(A) 30
(B) 6
(C) 5
(D) 15

Answer and Comments: (C). It is intuitively obvious that the an-

29
swer is 15 because to be special among 30 is five times as remarkable as
being special among 6. For a rigorous solution, note that it is easier
to find the complementary probability, i.e. the probability that the
special ball is not in the draw. For this to happen, all the 6 balls
picked must be from the 29 balls   that are not special. Hence the
29
6
complementary probability is  .
30
On expansion, this ratio equals
6
29 × 28 × 27 × 26 × 25 × 24/6! 24
= 30
= 45 . Hence the desired probabil-
30 × 29 × 28 × 27 × 26 × 25/6!
ity is 1 − 45 = 15 .

A simple problem once the idea of complementary probability strikes.


√ √ √
Q.26 A triangle has sides of lengths 5, 2 2, 3 units. Then the radius of
its inscribed circle is
√ √ √ √ √ √
(A) 5+ 23+2 2 (B) 5+ 3+2 2
√ √ √ √ √3 √
5+ 3−2 2
(C) 5 + 3 + 2 2 (D) 2

Answer and Comments: (D). The radius of the incircle of a triangle


is called the inradius of the triangle and generally denoted by r. There
are several formulas for r. In the present problem, as we are given
the three sides, say a, b, c, a formula in terms of the sides will be most
directly applicable. One such formula is
S
r= (1)
s
where S is the area of the triangle and s is its semi-perimeter, defined
as
a+b+c
s= (2)
2
We also need a formula for S in terms of a, b, c. There is a well-known
formula, Heron’s formula for this.
q
S= s(s − a)(s − b)(s − c) (3)

30
√ √ √
It is now merely a matter of substituting a = 5, b = 3, c = 2 2 into
(2) and (3) and then get the numerical value of r from (1). The catch
is that when we write (3) directly in terms of a, b, c, it becomes
1q
S= (a + b + c)(a + b − c)(b + c − a)(c + a − b) (4)
4
As a, b, c are surds, the resulting expression for s after the substitution
will be very complicated.
√ √
But there√is a√way out √ if we rewrite 2 2 as 8. For, now the three
sides are 5, 3 and √ 8. √
Since 5 + 3 = 8, we see that the triangle is
right
√ angled, with 5 and 3 as its shorter sides. So, its area is simply
1
2
15. So, instead of applying (3) we get from (1) and (2)

15
r = √ √ √
5+ 3+2 2
√ √ √ √
15( 5 + 3 − 2 2)
= √ √
( 5 + 3)2 − 8
√ √ √ √
15( 5 + 3 − 2 2)
= √
2 15
√ √ √
5+ 3−2 2
= (5)
2
which is (D).

Like Q.16, in this problem too, the specifics of the numerical data
provides a short cut. Curiously, in both the problems, the key was
to recognise a certain triangle as a right-angled one. In Q.16, this
was easier because a triangle with sides 3, 4 and 5 is the most familiar
example of a right angled triangle with integer sides.√ In the
√ present
problem, the recognition comes only after rewriting 2 2 as 8.

Q.27 Two ships are are approaching a port along straight routes of constant
velocities. Initially the two ships and the port formed an equilateral
triangle. After the second ship traveled 80 km., the triangle becomes
right angled.

31
PORT

60

30

80 km

SHIP I SHIP II

When the first ship reaches the port, the second ship was still 120 km
from the port. Find the initial distance of the ships from the port.

(A) 240 km. (B) 300 km. (C) 360 km. (D) 180 km.

Answer and Comments: (A). Let x be the initial distance between


the ships and the port. To find x, we introduce an auxiliary variable.
Even though each ship is traveling at a uniform speed, these two speeds
are different, or else, they would reach the port simultaneously. Ship
II is slower than Ship I. Let λ be the ratio of the speed of Ship II to
that of Ship I. Then λ < 1.
We construct a system of two equations in these two unknowns. When
Ship I covers the entire distance x, Ship II has covered only x − 120.
Hence

x − 120 = λx (1)

When Ship II has traveled 80 km., it is still x − 80 km. away from


the port. From the right angled triangle with the vertex angle 60◦ , the
distance between Ship I and the port is half of the distance between
Ship II and the port. That means at that time Ship I has covered a
distance x − x−80
2
= x+80
2
km. This gives another equation

x + 80
80 = λ (2)
2

32
As our interest is only in x, we eliminate λ by dividing (1) by (2) to
get
x − 120 2x
= (3)
80 x + 80
On simplification, this gives a quadratic in x, viz.

x2 − 200x − 9600 = 0 (4)



whose roots are 100 ± 19600 = 100 ± 140. Taking the positive root,
x = 240 km.

A simple but good problem testing the ability to interpret the


data correctly and to translate it into a system of equations. The
trigonometric part needed is minimal, only that sin 30◦ = 21 , which too,
can be bypassed by a standard geometric property of a 30 − 60 − 90
degrees triangle, studied in schools. Incidentally, from either (1) or (2),
we get λ = 21 , that means Ship I is twice as fast as Ship II. In fact, it
would have been a better problem to ask the ratio of the speeds of the
two ships because eliminating x between (1) and (2) is not as easy as
eliminating λ. So, most candidates would find x before finding λ.

Q.28 If x1 > x2 > . . . > x10 are real numbers, what is the least possible value
of
x1 − x10 x1 − x10 x1 − x10
    
... ?
x1 − x2 x2 − x3 x9 − x10

(A) 1010 (B) 109 (C) 99 (D) 910

Answer and Comments: (C). The numbers xi are not given. It is not
even given if they are positive. But that does not matter. The expres-
sion depends only on the differences of the successive xi ’s. Specifically,
let di = xi − xi+1 for i = 1, 2, . . . 9. Then x1 − x10 = d1 + d2 + . . . + dn
(d1 + d2 + . . . + d9 )9
and the expression to be minimised is . As all di ’s
d1 d2 . . . d9
are positive, by a direct application of the A.M.-G.M. inequality, this
is minimum when the d’s are all equal. But in that case, the expression
(9d1 )9
is = 99 .
d91

33
A simple problem once the idea that each numerator is the sum of
all denominators strikes.

Q.29 The range of values that the function

x2 + 2x + 4
f (x) =
2x2 + 4x + 9
takes as x varies over all real numbers in the domain of f is

3 1 3 1
(A) 7
< f (x) ≤ 2
(B) 7
≤ f (x) < 2
3 4 3 1
(C) 7
< f (x) ≤ 9
(D) 7
≤ f (x) ≤ 2

Answer and Comments: (B). Normally, finding the range of a func-


tion requires calculus methods to identify the intervals on which it is
increasing/decreasing and consolidating the information gained from
all such intervals. But other methods are available when the function
f (x) has a particularly simple form. The present problem falls in this
category. In fact, several methods are available which do not use cal-
culus or use it only marginally.
The stipulation about the domain is redundant because since 2x2 +
4x + 9 = 2(x + 1)2 + 7, the denominator never vanishes and so f (x) is
defined for all x ∈ IR. In fact, if we put u = (x + 1)2 , then
u+3
f (x) = (1)
2u + 7
As x varies over IR, u varies over [0, ∞). So the range of f (x) is the
u+3
same as that of g(u) = 2u+7 as u ∈ [0, ∞). Finding the range of g(u)
is simpler than finding the range of f (x). We can actually go further.
Since g(u) takes only positive values, we can take its reciprocal, say
h(u) = 2u+7
u+3
. Then the range of g(u) consists of the reciprocals of the
elements in the range of h(u). We rewrite h(u) as
2u + 7 2u + 6 + 1
h(u) = =
u+3 u+3
1
= 2+ (2)
u+3

34
1
As u varies over [0, ∞), u+3 decreases from 31 and assumes values as
1
near to 0 as we like but not 0. Hence the range of u+3 is (0, 13 ]. Adding
2, the range of h(u) is (2, 73 ]. Hence the range of the reciprocal function
g(u) is [ 37 , 21 ). As noted earlier, this is also the range of f (x).

Notice how the range was found by gradually reducing the problem
1
to the problem of finding the range of u+3 for u ≥ 0, for which calculus
1
was needed only to the extent that u+3 → 0 as u → ∞.
The special feature of f (x) that it is a ratio of two quadratic poly-
nomials enables another method based solely on the criterion for a
quadratic equation to have real roots. To say that a real number y is
in the range of f (x) is equivalent to saying that the equation

x2 + 2x + 4
=y (3)
2x2 + 4x + 9
has at least one real solution. For a fixed y this can be rewritten as a
quadratic in x, viz.

(2y − 1)x2 + (4y − 2)x + 9y − 4 = 0 (4)


1
We have to exclude y = 2
because in that case there are no solutions.
For y 6= 2, (4) will have a real solution in x if and only if the discriminant
is non-negative, i.e.

(2y − 1)2 ≥ (2y − 1)(9y − 4) (5)

As we are assuming that 2y − 1 6= 0, there are two possibilities depend-


ing on the sign of 2y − 1. If 2y − 1 is positive we can cancel it from (5)
to get 2y − 1 ≥ 9y − 4 or equivalently, 7y ≤ 3, i.e. y ≤ 37 . But this is
inconsistent with 2y − 1 > 0. So this possibility gives no points in the
range of f (x). Assume, however, that 2y − 1 < 0. Then after canceling
it from (5), the inequality is reversed and gives

2y − 1 ≤ 9y − 4 (6)

or equivalently, y ≥ 73 . Combining this with 2y < 1, we get that (3)


has a real solution if and only if 73 ≤ y < 12 . So the range of f (x) is the
interval [ 73 , 21 ).

35
Q.30 In the following diagram, four triangles and their sides are given. Areas
of three of these are also given. Find the area of the remaining triangle.

c d d a
a b b c

6 6 6 6
Area = 4 Area = 5 Area = x Area = 13

(A) 12 (B) 13 (C) 14 (D) 15

Answer and Comments: (C). The problem is ill-posed. We can


express the areas of the triangles in terms of their sides using Heron’s
formula mentioned (but not used) in the solution to Q.26. That gives a
system of three equations in the four unknowns a, b, c, d. For example,
the first triangle gives

(a + b + 6)(a + b − 6)(a + 6 − b)(b + 6 − a) = (4S)2 = 256 (1)

The second and the fourth triangle will give similar equations. If we
can find the values of c and d from these, Heron’s formula will give the
area of the third triangle.
But in general, a system of three equations in four unknowns has in-
finitely many solutions. In the solution to Q.24, we remarked that if the
variables are to take only non-negative integers as values, then some-
times the solution set becomes finite and in some cases the solution is
unique too. But here the problem does not specify that a, b, c, d are
integers. Nor does it follow from the data.
Nevertheless we can explicitly obtain one particular solution of this
system of three equations by a simple geometric construction. All the
four given triangles have 6 as one side. Let h1 , h2 , h3 , h4 be the altitudes
of the triangles falling on this side. Then the known values of the areas
of the first, the second and the last triangle imply three equations, viz.
4 5 13
h1 = , h2 = and h4 = (2)
3 3 3
The value of h3 is not given. If we know it we can find x immediately.

36
If we express the altitudes in terms of sides, the system of equations we
shall get will be the same as the one above (of which only one equation
was shown explicitly in (1)). But the advantage of the present system
is that a particular solution can be constructed very easily. We first
construct a square ABCD of side 6 and mark a point P inside it which
is at a perpendicular distance h1 from AB, h2 from BC and h4 from
AD as shown below. (This is feasible since h2 + h4 = 6 which is the
perpendicular between the sides AD and BC.)

D 6 C

h
3
d

6 c
6

h4 P h2

a b
h1

A B
6

If we call P A, P B, P C and P D as a, b, c, d respectively, then the trian-


gles AP B, BP C, CP D and DP A are congruent to the four given trian-
gles. So the third triangle CP D has altitude h3 = 6 − h1 = 6 − 34 = 14 3
.
Therefore its area x equals 21 ×6× 14
3
= 14. (Alternately, x can be found
by subtracting from 36 the sum of the areas of the other triangles.)
q
By Pythagoras theorem, the sides come out as a = h21 + h24 =
√ √ √ √
185 41
3
,b= 3
,c = 2213
and d = 365
3
.
It is doubtful if anybody can get,
or even guess, these by solving the system consisting of (1) and two
similar equations. And, even if someone can do that, unless it is shown
that this is the only solution, we cannot conclusively prove that x = 14.
All we can say is that 14 is a possible value of x. Sometimes solutions

37
are unique in degenerate cases. For example the single equation x2 +
y 2 = c in two variables x and y has infinitely many solutions for c > 0.
But for c = 0, there is only one solution, viz. x = 0, y = 0. It is unlikely
that in the presentq problem, because

of √some degeneracy

in the√data,
the solution a = h1 + h4 = 3 , b = 3 , c = 3 and d = 365
2 2 185 41 221
3
is
the only solution.
Another possibility is that sometimes the nature of the problem is
such that despite the plurality of solutions, the answer is unique. For
example, suppose we are asked to find 3x + 3y given that x + y = 7.
This single equation does not determine the values of x and y. But that
does not matter as far as the answer is concerned, which is 21. If this
scenario is to happen in the present problem, then it must be shown
that regardless of which solution of the system of three equations in
the four unknowns a, b, c, d we take, the value of x remains the same.
But that is also unlikely to happen.
Those familiar with jigsaw puzzles may think that the square in
the second figure is obtained by putting together the four triangular
pieces given in the first figure. But that is not so. On the contrary, we
started with the square in the second figure and partitioned it suitably
into triangles which satisfy the data in the problem. If we start with
some other quadruple of triangles in the first diagram, we may not be
able to assemble them into a square. Let α be the angle between the
sides 6 and b of the first triangle. Then its area is 3b sin α. As this
equals 4 we get 3b sin α = 4. Similarly, if β is the angle between the
sides 6 and b in the second triangle, we would get 3b sin β = 5. But if
these two angles are to be abutted along the common side b so as to
form a corner of a square, α and β must be complementary angles, i.e.
sin β should equal cos α. Although this happens to be the case in the
solution above (because tan 6 ABP = hh21 = 54 ), it may not be the case
with every quadruple of triangles in the first figure.

38
ISI BStat-BMath-UGA-2022 Paper 2 with Comments

Q.1 Consider a board having 2 rows and n columns. Thus there are 2n cells
in the board. Each cell is to be filled in by 0 or 1.
(a) In how many ways can this be done such that each row sum and
each column sum is even ?
(b) In how many ways can this be done such that each row sum and
each column sum is odd ?

Solution and Comments: (a) 2n−1 , (b) 2n−1 if n is even and 0 if n


is odd.
Let S be the set {1, 2, . . . , n}. Every placement of 0’s and 1’s in the cells
of the first row determines a unique subset A of S, consisting of those i’s
for which the i-th cell in the first row has 1 in it. Similarly, every filling
of the sells in the second row determines a subset B of S defined by
{i : i-th cell in the second row is filled with 1}. For example, suppose
we have n = 9 and the following filling of the cells.

1 2 3 4 5 6 7 8 9
0 1 0 0 1 1 1 0 1

0 0 1 0 1 0 1 0 1

Then A = {2, 5, 6, 7, 9} and B = {3, 5, 7, 9}. These two subsets deter-


mine the placement completely.
Now in each column there are two entries. The column sum is even
when both of them are 1 or both are 0. So, if it is the i-th column,
then either i is in both A and B or in neither of them. If every column
sum is to be even, then for every i = 1, 2, . . . , n, either both A and B
contain i or neither contains i. This is equivalent to saying that A = B.
So, in (a), the sets A and B should be equal and moreover each should
have an even number of elements, since the row sum is also to be even.
The number of subsets of S having an even number of elements is the
k
!
n
where the upper index k of the summation is n2
X
binomial sum
r=0 2r
if n is even and n−12
if n is odd. It is well known that this sum equals

39
2n−1 , i.e. half the total number of subsets of S. (See Equation (24) in
Chapter 5.) So this is the answer to (a).
For (b), both A and B must have an odd number of elements since
each row sum is odd. As for column sums, for the i-th column, for the
sum to be odd, i is in exactly one of A and B. That means B is the
complement of A w.r.t. the set S. Summing up, the answer to (b) is
the number of subsets A of S for which both A and S − A have odd
number of elements. If n is odd there is no such subset. If n is even,
then any subset of even cardinality has this property and the number
of such subsets is 2n−1 as already noted. So the answer to (b) is 0 if n
is odd and 2n−1 if n is even.

A simple problem once the idea of relating filling of cells with a pair
of subsets of S strikes. The binomial identity needed is well-known.

Q.2 Consider the function


m
(x − k)4 , x ∈ IR,
X
f (x) =
k=1

where m is an integer. Show that f has a unique minimum and find


the point where the minimum is attained.

Solution and Comments: The case m = 1 is trivial for in this case


f (x) = (x − 1)4 is minimum at the point 1. More generally, for any
a ∈ IR, (x − a)4 attains its unique minimum at a.
When m = 2, f (x) = (x − 1)4 + (x − 2)4 . It is symmetric about
the midpoint 32 of [1, 2]. f (x) is differentiable everywhere and f ′ (x) =
4[(x − 1)3 + (x − 2)3 ] = 4[(x − 1)3 − (2 − x)3 ] which vanishes only when
x−1 = 2 −x, i.e. at x = 23 . f (x) changes its behaviour from decreasing
to increasing at 23 . Hence there is a unique minimum at 32 .
Before dealing with higher values of m, we note that the argument for
m = 2 can be generalised if f (x) were of the form (x − a)4 + (x − b)4
for any two real numbers a, b, with a < b. There would be a unique
minimum at the midpoint a+b 2
.
Now suppose m is any integer bigger than 2. First suppose m is even,
say m = 2k. We express f (x) as the sum of k functions, by grouping

40
together two terms of f (x) which are symmetric about the midpoint
m+1
2
. That is, we let f1 (x) = (x − 1)4 + (x − m)4 , f2 (x) = (x − 2)4 +
(x − m + 1)4 , . . . , fk (x) = (x − k)4 + (x − (k + 1))4 . Then f (x) =
f1 (x) + f2 (x) + . . . + fk (x). In general, little can be said about the
location of the minimum of the sum of several functions in terms of the
minima of those functions. But in the present case, each fi (x) attains
its minimum at the same point, viz. m+1 2
. Hence so does their sum.
When m is odd, say m = 2k + 1, the point of symmetry is k + 1. we
group together (x − 1)4 with (x − m)2 , (x − 2)4 with (x − m + 1)4 and
so on. Each group attains its minimum at m+1 2
= k + 1. This time
(x − k − 1)4 is left out. But that also attains its minimum at k + 1.
Hence f (x) attains its minimum at x = k + 1 = 1+m 2
.
Thus we have shown that for every m, f (x) attains its unique minimum
at m+1
2
.

A simple problem once the idea of expressing f (x) as a sum of


functions all having their minima at the same point strikes. Unlike the
cases m = 1, 2, the general case cannot be generalised if f (x) were of
m
(x − ak )4 where a1 , a2 , . . . , am are arbitrary real numbers.
X
the form
k=1

Q.3 Consider the parabola C : y 2 = 4x and the straight line L : y = x + 2.


Let P be a variable point on L. Draw the two tangents from P to C
and let Q1 and Q2 denote the two points of contact on C. Let Q be
the mid-point of the line segment joining Q1 and Q2 . Find the locus of
Q as P moves along L.

Solution and Comments: The elaborate description of Q could have


been shortened. The segment Q1 Q2 is called the chord of contact of
the point P w.r.t. the parabola C. So, Q is simply the midpoint of the
chord of contact of P w.r.t. C.
Formulas are available that give the equation of the chord of contact of
a point (h, k) w.r.t. any conic. Here the conic is the parabola y 2 = 4x
and the equation of Q1 Q2 is

ky = 2(x + h) (1)

41
To determine the points of contact Q1 = (x1 , y1 ) and Q2 = (x2 , y2 )
(say) we solve this simultaneously with the equation of C, i.e.

y 2 = 4x (2)

Eliminating x, we have a quadratic in y, viz.

y 2 = 2ky − 4h (3)

The roots, say y1 and y2 can be found. But as we are interested only
in the midpoint of Q1 Q2 , we want only y1 +y
2
2
which is simply k even
without solving (3). To find the x-coordinate of the midpoint, we
2
simply put y = k in (1) to get x = k −2h
2
.
Since the point P = (h, k) lies on the line y = x + 2 to begin with we
have k = h + 2. Therefore the locus of Q in parametric form is

y = h+2
(h + 2)2 − 2h h2 + 2h + 4
and x = = (4)
2 2
The question does not specifically ask the locus in a non-parametric
form. It comes by eliminating the parameter h from these two equa-
tions. Thus the locus of Q is 2x = (y −2)2 + 2(y −2) + 1 = y 2 −2y + 1 =
(y − 1)2 . (Note that this is also a parabola, which has the same axis
and vertex as the original parabola.)

An absolutely routine problem, hardly suitable for Paper 2. Ready-


made formulas are also available for the mid-point of the chord of con-
tact. But it is preferable to get it from (1) which is easier to remember
both as the equation of the tangent (in case the point is on the conic)
and that of the chord of contact. It is a special case of the equation
of the chord of contact of a point (x1 , y1 ) for a conic with equation
ax2 + by 2 + 2hxy + 2gx + 2f y + c = 0 which is axx1 + byy1 + h(xy1 +
yx1 ) + g(x + x1 ) + f (y + y1 ) + c = 0 and popularly remembered as
T = S1 . Of course, one can proceed directly from the basic principles
as in the answer to Problem 15 above. But that amounts to duplicating
the proofs of the standard formulas. The best compromise is to rely on
T = S1 , applicable for any conic.

42
Q.4 Let P (x) be an odd degree polynomial in x with real coefficients. Show
that the equation P (P (x)) = 0 has at least as many distinct real roots
as the equation P (x) = 0.

Solution and Comments: Every polynomial with real coefficients


has a real root. In fact, as a function from IR to IR, it is surjective.
That is, for every y0 ∈ IR, there is at least one x0 ∈ IR such that
P (x0 ) = y0 . (This follows because P (x) − y0 is also a polynomial of
odd degree.)
Now, assume that the distinct roots of P (x) are y1 , y2 , . . . , ym . Then
for every i = 1, 2, . . . , m, there is some xi ∈ IR such that P (xi ) = yi .
Then P (P (xi )) = P (yi ) = 0. Further, for ineqj, xi 6= xj as otherwise
we would have yi = P (xi ) = P (xj ) = yj , contradicting that the y’s are
distinct.
Hence P (P (x)) = 0 has at least m distinct roots.

Another simple question. While the existence of a root of a real


polynomial of odd degree is fairly well-known, the fact that the function
defined by it is onto is not used so frequently. That is the only thought
needed here.

Q.5 For a positive integer n and i = 1, 2, let fi (n) denote the number of
divisors of n of the form 3k + i (including 1 and n). Define for any
positive integer n,
f (n) = f1 (n) − f2 (n).
Find the values of f (52022 ) and f (212022 ).

Solution and Comments: Every odd number is either of the form


3k + 1 for some k or of the form 3k + 2 for some k. Both 52022 and
212022 are odd and hence so are all their divisors. The problem asks to
count how many are of the two given types.
In terms of residue classes, every odd number is congruent modulo 3 to
either 1 or 2. The latter is the same as the residue class of −1, which
is more convenient in taking powers.
As 5 is a prime, the only divisors of 52022 are of the form 5r where r
can be any integer from 0 to 2022. Since 5 is congruent modulo 3 to

43
−1, 5r is congruent to 1 modulo 3 if r is even and to −1 if r is odd. As
r is to lie between 0 and 2022, f1 (52022 ) is the number of even integers
between 0 to 2022. Its count is 1012 since r = 0 is also to be included.
So, f1 (52022 ) = 1012. Contributions to f2 (52022 ) come from odd values
of the exponent. They number 2023 − 1012 = 1011. Hence

f (52022 ) = f1 (52022 ) − f2 (52022 ) = 1012 − 1011 = 1 (1)

The method for f (212022 ) is basically the same. But 21 is not a prime.
It factors as the product of two primes, viz. 3 and 7. So every divisor of
it is of the form 3r 7s where r, s are integers taking values from 0 to 2022.
Since the residue class of 7 modulo 3 is 1, all powers of 7 are congruent
to 1 modulo 3. Hence the residue class modulo 3 of 3r 7s is the same
as that of 3r . This behaves like 5r since, like 5, 3 is also congruent to
−1 modulo 2. Therefore f1 (212022 ) is the number of ordered pairs of
the form (r, s) where both r and s range over 0 to 2022 and r is even.
Hence

f1 (212022 ) = 1012 × 2023 (2)

and similarly,

f2 (212022 ) = 1011 × 2023 (3)

Subtracting,

f (212022 ) = 2023. (4)

So the answer is f (52022 ) = 1 and f (212022 ) = 2023.

Another simple problem based on residue arithmetic. There is little


point in asking it after Problem 20 in Paper 1, which involves a little
more work.
Q.6 Consider a sequence P1 , P2 , . . . of points in the plane such that P1 , P2 , P3
are non-collinear and for every n ≥ 4, Pn is the midpoint of the line
segment joining Pn−2 and Pn−3 . Let L denote the line segment joining
P1 and P5 . Prove the following:
(a) The area of the triangle formed by the points Pn , Pn−1, Pn−2 tends
to 0 as n goes to ∞.

44
(b) The point P9 lies on L.

9 6
7

8
3
2 5

Solution and Comments: (a) is easy. Every time we take a new Pn ,


it divides a side of the earlier triangle with vertices Pn−1 , Pn−2 , Pn−3 .
Hence the area of △Pn Pn−1 Pn−2 is half the area of △Pn−1 Pn−2 Pn−3 .
Therefore the areas form a geometric progression with common ratio
1
2
. Since ( 12 )n → 0 as n → ∞, (a) follows. (See the figure above, where
to avoid cluttering, Pi is shown merely as i.)
For (b), we label the points Pi ’s with complex numbers. The conclu-
sion is independent on the choice of the frame of reference and scaling.
So, without loss of generality, we take P1 = 0, P2 = 1 and P3 = z, some
(unknown) complex number. Then the complex numbers representing
P4 to P9 can be calculated recursively from their construction. Thus

P1 = 0 (1)
P2 = 1 (2)
P3 = z (3)
0+1 1
P4 = = (4)
2 2
1+z
P5 = (5)
2
z + 21 2z + 1
P6 = = (6)
2 4
z+2
P7 = (7)
4

45
4z + 3
P8 = (8)
4
3z + 3
and P9 = (9)
8
The line segment L joins 0 and z+1 2
. Clearly 3z+3
8
lies on it since 3z+3
8
is a positive real multiple of z+1
2
. Hence P9 lies on L. In fact, it divides
L in the ratio 3 : 1.

Another very simple problem not suitable for Paper 2. (a) is trivial
as soon as it is recognised that every time the area is divided by 2. (b)
could have been asked in Paper 1, asking to find the least n > 5 such
that Pn lies on L.

Q.7 Let
P (x) = 1 + 2x + 7x2 + 13x3 , x ∈ IR.
Calculate for all x ∈ IR,
n
x

lim P( ) .
n→∞ n

Solution and Comments: By a direct calculation,

x 2x 7x2 13x3
P( ) = 1 + + 2 + 3 (1)
n n n n
If x = 0, then P ( nx ) = 1 and the sequence (P ( nx ))n is the constant
sequence 1. Hence the limit in question is 1.
Assume x 6= 0. Put h = nx . Then
x
P ( ) = 1 + 2h + 7h2 + 13h3 (2)
n
and so,
n
x

P( ) = (1 + 2h + 7h2 + 13h3 )x/h (3)
n
As n → ∞, h → 0 only through the values of the form x, x2 , x3 , . . .. But
the R.H.S. of (3) makes sense for all h in a sufficiently small deleted

46
neighbourhood of 0 (so small that the expression 1 + 2h + 7h2 + 13h3 is
positive). Hence the limit, say L in the question can also be evaluated
if lim (1 + 2h + 7h2 + 13h3 )x/h can be evaluated. This limit is in an 1∞
h→0
indeterminate form. So, we take the log of the expression and see if it
has a limit.
  ln(1 + 2h + 7h2 + 13h3 )
ln (1 + 2h + 7h2 + 13h3 )x/h = x (4)
h
x is a constant. The limit of the other factor is simply the derivative
(w.r.t. h) of ln(1 + 2h + 7h2 + 13h3 ) at h = 0. By a direct calculation,
2+14h+39h3
it equals 1+2h+7h 2 +13h3 = 2. Hence
h=0

lim (1 + 2h + 7h2 + 13h3 )x/h = 2x (5)


h→0

Taking exponentials, the limit of the R.H.S. of (3) and hence also the
limit asked is 22x .

Another routine problem for those used to evaluate limits in inde-


terminate form.
Q.8 Find the minimum value of

| sin x + cos x + tan x + cot x + sec x + cosecx|

for real numbers x not multiple of π/2.

Solution and Comments: This problem is a straight replica of Q.27


in Paper 1 of ISI Entrance Test 2014, except for the order of the terms
in the expression for the function and the stipulation that multiples of
π/2√are to be excluded from the domain of the function. The answer
is 2 2 − 1. See the commentary for that year for the solution. The
change of order of the terms and the exclusion of multiples of π/2 were
justified in the commentary. They have been implemented this year.
But such an outright duplication is shocking to say the least.
Q.9 Find the smallest positive real number k such that the following in-
equality holds
1
|z1 + z2 + . . . + zn | ≥ (|z1 | + |z2 | + . . . + |zn |).
k
47
for every positive integer n ≥ 2 and every choice of z1 , z2 , . . . , zn of
complex numbers with non-negative real and imaginary parts.
[Hint: First find k that works for n = 2. Then show that the same k
works for any k ≥ 2.]

Answer and Comments: k = 2. The wording of the question is
somewhat clumsy. It would have been better to ask for the greatest
positive real number M such that
|z1 + z2 + . . . + zn | ≥ M(|z1 | + |z2 | + . . . + |zn |) (1)
holds for all z1 , z2 , . . . , zn lying in the first quadrant. Once such M is
found, k is merely its reciprocal.
Following the hint, we first work out the case n = 2. Now the problem
is to find the largest real number M such that
|z1 + z2 | ≥ M(|z1 | + |z2 |) (2)
for all z1 , z2 in the first quadrant. When z1 , z2 are both 0, equality
holds for any M no matter how large. We must therefore exclude this
case because we want M to work in all cases. When at least one of z1
and z2 is non-zero, |z1 | + |z2 | > 0 and therefore (2) is equivalent to
|z1 + z2 |
≥M (3)
|z1 | + |z2 |
Call the L.H.S. as f (z1 , z2 ). It is a real valued function of two complex
variables defined over the set, say S = {(z1 , z2 ) ∈ C | × C| : (z1 , z2 ) 6=
(0, 0)}. The number M we are looking for is the minimum of f (z1 , z2 )
as (z1 , z2 ) ranges over the subset, say T = {(z1 , z2 ) ∈ C| + × C | +) :
+
(z1 , z2 ) 6= (0, 0)} where C| denotes the set of complex numbers in the
first quadrant, i.e. whose real and imaginary parts are non-negative.
On the entire set S, the minimum of f is 0. It occurs when z1 + z2 =
0. But such pairs (z1 , z2 ) are not in the subset T . So, it is possible
that the minimum of f on the subset T is higher. (Note that when
a function is restricted to a subset, its maximum decreases but its
minimum increases in general.)
Let us analyse when f assumes smaller values. If we view z1 , z2 as
vectors and the angle between them is obtuse, then their components

48
along the angle bisector between them are oppositely directed and serve
to nullify each other. The larger the angle between them, the larger the
extent to which they cancel each other. In the extreme case, when z1
and z2 are oppositely directed, the angle between them is 180 degrees,
the degree of cancellation is highest and we get |z1 + z2 | = ||z1 | − |z2 ||.
At the other extreme, when the angle between z1 and z2 is 0 degrees
they add up and we get |z1 + z2 | = |z1 | + |z2 |.
This suggests that to minimise f (z1 , z2 ) over the subset T , we
should choose the angle between z1 and z2 as large as possible subject
to the constraint that both are in the first quadrant. The largest angle
possible with this restriction is a right angle. For example, if we take
|z1 + z2 | |1 + i| √
z1 = 1 and z2 = i, then the ratio equals = 22 = √12 .
|z1 | + |z2 | 1 + |i|
1
Hence the minimum of f (z1 , z2 ) over T is at least 2 . Since √12 is one

of the values assumed by f on T , the minimum M of f on T is at least


√1 . We claim, however, that M is actually the minimum of f on T ,
2
that is, we show that for any (z1 , z2 ) 6= (0, 0) with z1 , z2 lying in the
first quadrant,
|z1 + z2 | 1
f (z1 , z2 ) = ≥√ (4)
|z1 | + |z2 | 2
This is trivially true if z1 or z2 is 0. So assume that both are non-zero.
Write them in the polar form as z1 = r1 eiθ1 and z2 = r2 eiθ2 where both
θ1 , θ2 are in [0, π2 ]. Without loss of generality assume that θ2 ≤ θ1 . Let
z = z1 /z2 . Then z = reiθ where r = rr21 and θ = θ1 − θ2 . Then z
also lies in the first quadrant. Dividing both the numerator and the
denominator by |z2 |,
|z + 1|
f (z1 , z2 ) = f (z, 1) = (5)
|z| + 1
So, (4) is equivalent to proving that 2|z + 1|2 ≥ (|z| + 1)2 . Since
|z + 1|2 = |z|2 + 1 + z + z = 1 + r 2 + 2r cos θ we are reduced to
proving that 2 + 2r 2 + 4r cos θ ≥ 1 + 2r + r 2 . But that is true because
1 − 2r + r 2 = (1 − r)2 and cos θ are both non-negative (since z is in the
first quadrant).

Thus we have proved (2) with M = √12 . Hence k = 2 works for
n = 2. This completes the first part of the hint given. To extend the

49
inequality for any n ≥ 3, first assume that n = 3. Then z1 , z2 , z3 are
linearly dependent. Without loss of generality, assume z3 lies in the
acute angle between z1 and z2 . Then z3 = αz1 + βz2 where α, β are
non-negative. Hence 1 + α and 1 + β are positive. Using the result in
the case n = 2, we have

|z1 + z2 + z3 | = |(1 + α)z1 + (1 + β)z2 |


1
≥ √ [|(1 + α)z1 | + |(1 + β)z2 |]
2
1
= √ [(1 + α)|z1 | + (1 + β)|z2 |]
2
1
= √ [|z1 + |z2 | + (|αz1 | + |βz2 |)]
2
1
≥ √ [|z1 | + |z2 | + |z3 |] (6)
2
where in the last step we have used the triangle inequality |αz1 |+|βz2| ≥
|αz1 + βz2 |.
The proof for the case n > 3 is similar. Without loss of generality
we assume that z3 , z4 , . . . , zn all lie in the acute angle between z1 and
z2 . (To express this more precisely, suppose z1 has the least and z2 the
greatest argument among z1 , z2 , . . . , zn .) For i = 3, 4, . . . , n, write zi as
αi z1 + βi z2 where αi and βi are non-negative. Then z1 + z2 + . . . + zn =
w1 + w2 where w1 = (1 + α3 + α4 + . . . + αn )z1 and w2 = (1 + β3 + β4 +
. . . + βn )z2 . Applying the result for the case n = 2 to the pair (w1 , w2 ),
regrouping the terms and finally applying the triangle inequality for
αi z1 + βi z2 for each i = 3, 4, . . . , n establishes the result for n.

So M = √12 and the number k asked is 2.

An elementary but excellent, thought provoking problem on abso-


lute values of complex numbers. We are thoroughly familiar with the
triangle inequality |z1 + z2 | ≤ |z1 | + |z2 |. In a multiplicative form, 1 is
|z1 + z2 |
an upper bound on the ratio . We know when equality holds.
|z1 + z2 |
We also know when the upper bound is way too high. The present
problem deals with a situation where the upper bound is not so high
by restricting the z1 and z2 to a subset of the complex plane. The hint

50
given is very helpful because the arguments needed for n = 2 and for
extending the result to n > 2 are very different. The extension hinges
crucially on the linear dependence of three (or more) complex num-
bers. If the result is paraphrased in terms of the triangle inequality
for vectors, then an extension from IR2 to IR3 will not be valid since
three vectors in IR3 , even though given to lie in the first orthant, may
be linearly independent.

51
CONCLUDING REMARKS

The test this year is highly disappointing, especially on the backdrop of


the last year. Paper 2 is especially lacking in any imaginatively designed
problems. The only exception is Q.9. Q.8 is also an excellent problem, but
a straight repetition of a 2014 problem. Q.2 in Paper 2 is a good problem
because of the short cut provided by the symmetry of the function to be
minimised. Solution of Q.1 is based on a certain binomial identity, which,
although fairly well-known, is not among the basic identities. In a full length
question, this leads to a controversy whether a candidate is expected to prove
that identity.
The only (moderately) noteworthy problems in Paper 1 are :

1. Problem 4: About a polynomial P (x) satisfying |P (x)| ≤ |x| for all


real x.
2. Problem 5: About the functional equation f (x)f (y) = f (x + y) + f (x −
y).
3. Problem 11: Maximisation of 12xy(72 − 3x − 4y).
4. Problem 12: Minimisation of |x − 3| + |x + 1| + |x + 2| + |x|
5. Problem 13: About a function whose average over any interval occurs
at the midpoint..
6. Problem 16: About a triangle inscribed in a square.
7. Problem 19: About magic squares of order 3
8. Problem 20: Divisors of n5 + 4n4 + 3n3 + 2022.
1
9. Problem 24: About the integral solutions of a
+ 1b + 1
c
=1+ 2
abc
.
10. Problem 26: Finding the in-radius of a triangle with sides given.

Out of these, Problem 16 and 26 test the ability to simplify work


involved by noticing some special features of the data. Problem 24 is not
likely to be solved without a tricky conversion of the given equation.
There are a few other good problems in Paper 1. But they lack novelty
as similar problems have been asked several times in comparable entrance
tests.

52
Problem 30 of Paper 1 seems ill posed since it asks to determine
solutions of a system of three equations in four unknowns. Analogy with
jigsaw puzzles is tempting but misleading.
As usual, there is duplication of certain ideas while omission of some areas.
For example, both Problem 11 and 28 of Paper 1 are based on the A.M.-G.M.
inequality. Although number theory is represented in three problems, they
are not upto the mark. Probability is paid a lip service. There are no
problems on conditional probability. And, of course none on the integral
part function. But the most shocking omission is differential equations.
The best problem in the entire test is the last problem of Paper 2 about
the triangle inequality for complex numbers.

53

You might also like