EDUCATIVE COMMENTARY ON

ISI 2017 MATHEMATICS PAPERS

(Uploaded on 24/09/2017)

Contents

Introduction 1

Paper 1 3

Paper 2 63

Concluding Remarks 81

Introduction
The commentary on ISI B.Sc. Entrance Test for 2015 was written after
those for 2016 and 2014. The quality of the questions in these two years
had raised expectations. But the first paper of 2015 turned out to be very
disappointing. Also, a look at the question papers prior to 2014 revealed
certain common patterns which were repeated. So the element of novelty
was not so strong as it was when the commentaries for 2016 and 2014 were
written. To some extent such repetitions are to be expected, given the limited
syllabus and the large number of questions to be asked. Instead of asking 30
questions in Paper 1 to be solved in two hours, it would be far better to ask
only about 15 really good questions.
As a result, the plan to write exhaustive commentaries on ISI Entrance
Tests was curtailed to only the four recent ones, viz. from 2014 onwards as
these are the most relevant. A few students admitted through them could
still be studying at ISI.
For reasons far from clear, the official version of the 2017 question papers
has not been uploaded on the ISI website. The present commentary is based
on unofficial versions of the two papers. Revisions will be made when the
official versions are released.
As in the other commentaries, unless otherwise stated, all the references

1
are to the author’s Educative JEE Mathematics, published by Universities
Press, Hyderabad.
I am thankful to Deepanshu Rajvanshi who corrected an omission and
several errors in an earlier draft of the present commentary. Readers are in-
vited 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-2017 Paper 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 Consider a quadratic equation ax2 +2bx+c = 0 where a, b, c are positive

real numbers. If the equation has no real roots, then which of the
following is true?

(A) a, b, c cannot be in AP or HP, but can be in GP.

(B) a, b, c cannot be in GP or HP, but can be in AP.
(C) a, b, c cannot be in AP or GP, but can be in HP.
(D) a, b, c cannot be in AP, GP or HP.

Answer and Comments: (C). The connection with quadratic equa-

tions is superficial. The first statement is a fancy way of saying that
b2 < ac. The success of the solution lies in the observation that three
numbers are in a particular type of a progression if and only if the
middle one is the corresponding type mean of the other two. So, let
A, G and H be the arithmetic, the geometric and the harmonic means
respectively of a and c. Then G2 = ac and so the inequality b2 < ac
implies b2 < G2 and hence, keeping in mind that all the numbers are
positive, that
b<G (1)
This proves that a, b, c cannot be in GP. We can similarly calculate A
and H in terms of a and b and decide how they compare with b. But
we can as well use the inequality that
H ≤G≤A (2)
The second part of this inequality is popularly called the A.M.-G.M.
inequality. The first half is comparatively less known. It follows by
1 1
applying the A.M.-G.M. inequality to the reciprocals and .
a b
In view of (1), (2) automatically rules out that b = A. So, a, b, c are
neither in G.P. nor in A.P. Whether they are in H.P. or not depends

3
on whether b = H. There is nothing to rule this possibility out. So,
2ac
by taking a, c to be any positive real numbers and setting b =
a+c
(which is the harmonic mean of a and c), we see that it is possible for
a, b, c to be in HP.

A good, simple problem based on the A.M.-G.H.-H.M. inequality,

requiring very little computation once the essential idea strikes that
three numbers are in a progression of a particular type if and only if
the middle one is the corresponding type of mean of the other two.
The full A.M.-G.M.-H.M. inequality (3) rarely figures in applications.
As noted above its second half is far more popular than the first one.
Interestingly, there is one situation where all the three means figure
and their inequality assumes a natural interpretation. And, even at
the risk of a digression, it deserves to be mentioned. Surprisingly, it is
from music. The arithmetic mean is more popularly called the average
and we use it all the time in day to day life. The geometric mean of
two numbers, say a and c has a geometric significance as the side of a
square whose area equals that of a rectangle with sides a and c. Some
old books mention that the term ‘harmonic mean’ comes from music.
This is plausible because harmony of notes is very important in music.
But rarely is the connection explained in detail.
In music, when two notes of different frequencies are played together,
maximum harmony is reached when one of them is a multiple of the
other. A well known fact in physics is that the frequency of sound pro-
duced by a vibrating string is inversely proportional to its length. So, if
two strings of lengths L1 and L2 produce notes of frquencies f1 and f2 ,
L1 + L2
then a string of length will produce a frequency which equals
2
2f1 f2
their harmonic mean . If f1 is taken as some base frequency,
f1 + f2
called sa or shadja, in Indian classical music, and f2 = 2f1 , then f2 is
the frequency of the note called the upper sa or taar shadja. When
the string is plucked at a point exactly at the middle, it produces the
4
frequency f1 which is simply the harmonic mean of f1 and 2f1 . It is
3
called the madhyam or more precisely the shudh madhyam while the
3
note whose frequency is the A.M. f1 is called the pancham. Thus, in
2

4
 the musical octave, the pancham is sharper than the madhyam because
of the A.M.-H.M. inequality.
√ √
What about the geometric
√ mean 2f1 ? 2 is an irrational num-
ber. But the frequency 2f1 comes very close to the note called the
teevra madhyam. Summing up, the madhyam, the teevra madhyam
and the pancham are, respecively, the harmonic, the geometric and the
arithmetic means of the shadja and the taar shadja. And the musical
interpretation of the inequality is that the pancham is sharper than the
teevra madhyam, which, in turn, is sharper than the shudh madhyam.

Q.2 A unit square has its corners chopped off to form a regular polygon
with eight sides. What is the area of this polygon?
√
√ √ √ 2 7
(A) 2( 3 − 2) (B) 2 2 − 2 (C) (D)
2 9

Answer and Comments: (B). A good diagram is a key to understand

the problem as well as to solve it.
It is not given that the corners b 1−2b b
are of equal size or even that they b b
are isosceles triangles. But such
things are to be assumed in the
very concept of a corner. So let
b be the side of the corner lying
along an edge of the square. Then
we get a diagram as shown.
Equating the portion of the side left with the hypotenuse of the
corner, we get
√
1 − 2b = 2b (1)

Solving,
1
b= √ (2)
2+ 2
√
2
We now get the length of the side of the octagon as √ . There is
2+ 2
a formula for the area of a regular n-gon in terms of the length of its

5
 side and using it we can answer the question. But that would be too
elaborate. In the present case, the octagon is the figure that results by
removing the four corners from the unit square. So its area, say A, is
obtained by subtracting the removed area which is simply 2b2 from 1.
Hence,
2
A = 1− √
(2 + 2)2
2 1
= 1− √ =1− √
6+4 2 3+2 2
√
2+2 2
= √ (3)
3+2 2
As it stands, this answer does not match with any of the given options.
To remedy this, we rationalise the denominator. Thus
√ √
(2 + 2 2)(3 − 2 2)
A = √ √
(3 + 2 2)(3 − 2 2)
√ √
= (2 + 2 2)(3 − 2 2)
√
= −2 + 2 2 (4)
which matches with (B).

A straightforward problem, but not without some room for elegance.

Had we applied the formula for the area of a regular n-gon with each
d2
side d, viz., n × cot(π/n), we would have got the same answer by
4 √
√ 2
putting n = 8 and d = 2b = √ . But we would have to calculate
2+ 2
cot(π/8) i.e. cot(22.5◦ ) which is not a very standard figure. Instead, we
got the answer effortlessly by simply subtracting the removed area. The
discrete analogue of this technique is called complementary count-
ing (see Comment No. 1 of Chapter 1).
A little work in rationalising a surd is also needed at the end. But
it is not very laborious and so this problem is a reasonable one.
Q.3 A solid cube of side five centimeters has all its faces painted. The cube
is sliced into smaller cubes each of side one centimeter. How many of
these cubes will have paint on exactly one of its faces?

6
 (A) 25 (B) 54 (C) 126 (D) 150

Answer and Comments: (B). Another simple geometric problem

where half the success lies in visualising it correctly. We show the cube
in (a) of the figure below. In all the cube gets divided into 53 = 125
smaller cubes. Out of these, 33 = 27 will lie in the interior of the cube
and none of their faces will be painted. For each of the remaining 98
smaller cubes, at least one face will lie along a face of the bigger cube
and hence will get painted. We want the number of those smaller cubes
which have exactly one face lying along a face of the original cube.
5
5
5 1 1 1 1 1
5 1
1

5 1 X X X 1

5
1 X X X 1
5
1 X X X 1

5 1 1
5 1 1 1 1 1

(a) (b)
In (b) of the figure we show one face of the original cube divided into
25 squares of side 1 each. Out of these 9 lie in the interior while 16
lie along the boundary. The latter will also have another face lying
along some other face of the cube and hence is painted. So there are
only 9 smaller cubes which have exactly one face painted. They are
different for the different faces of the bigger cube. As the big cube has
six faces, the number of smaller cubes having exactly one face painted
is 9 × 6 = 54.

Another good, simple problem. It would have been more interesting

to ask the number of smaller cubes which have exactly two of their faces
painted. On each face, these are the ones which appear on the border
and there are 16 of them (those that are not marked in part (b) of
the figure). So, it is tempting to think that in all there are 16 × 6
smaller cubes which have two of their faces painted. But in doing so,

7
 we are counting each cube twice because it lies along an edge which is
the common boundary of two adjacent faces. So, the correct answer
is not 96 but only 48. But even this is not correct, because cubes
which are at the eight vertices of the big cube have three of their faces
painted. Together they count for 24 smaller painted squares. So we
need to subtract 24 from 96 and then divide by 2. That gives 36.
A slightly easier approach is to count only the small cubes that lie
along an edge but not at either of its two ends. There are three such
small cubes along each edge. As there are twelve edges, the number
of small cubes which have exactly two of their faces painted is 36.
Summing up, each of the 125 smaller cubes can have at most three
of its faces painted. The number of small cubes with 0, 1, 2 and 3
faces painted is, respectively, 27, 54, 36 and 8. Their total is 125 as
expected. Also the total number of painted faces (of smaller cubes) is
27 × 0 + 54 × 1 + 36 × 2 + 8 × 3 = 54 + 72 + 24 = 150, also as expected
because each of the six faces of the big cube contains 25 faces of the
small cubes.
z−i
Q.4 Let z be a complex number such that is purely imaginary. Then
z−1
the minimum value of |z − (2 + 2i)| is
√ √ 3 1
(A) 2 2 (B) 2 (C) √ (D) √
2 2

Answer and Comments: (B). Let us first interpret what is asked.

Geometrically, |z − (2 + 2i)| is the distance of the point z from the fixed
point 2 + 2i in the complex plane. The point z is a variable point which
satisfies a certain given condition. If we can interpret this condition
suitably, it will translate into the equation of some curve, say C, in the
complex plane. Once we identify this curve C, the problem reduces to
finding the point on it which is closest to the point 2 + 2i. This is an
easy task if C comes out to be a straight line or a circle. Otherwise it is,
by itself, another problem. (For example, if C comes out to be a conic,
we shall have to take its parametric equations with some parameter t,
express the distance from the point 2 + 2i as a function of t and then
minimise it.)
So, let us first identify this curve C on which z moves if it satisfies the

8
 z−i
condition that the ratio is a purely imaginary number. A brute
z−1
force method in dealing with problems about complex numbers is to
resolve them into real and imaginary parts. In the present problem,
for example, if we write z = x + iy as usual, then the given condition
means
x + i(y − 1)
= ib (1)
(x − 1) + iy
for some real number b. Multiplying both the sides by the complex
conjugate of the denominator,

(x + i(y − 1))((x − 1) − iy)) = ib((x − 1)2 + y 2 ) (2)

Equating the real parts of both the sides as well as their imaginary
parts will give a system of two equations in the variables x and y with
b as a parameter. Eliminating b between these two equations will give
us the equation of the curve C.
It is obvious that method will entail considerable algebraic manipula-
tions. But there is a way to interprete the data geometrically. Let z, 1, i
and 2 + 2i represent the points P, A, B and Q in the Argand diagram.
Then the complex numbers z − i represents the line segment from B to
P . Similarly, the complex number z − 1 represents the segment from
A to P . The condition that the ratio of these two numbers is purely
imaginary is equivalent to these two segments being perpendicular to
each other as shown in the figure below.

Q. 2+2i

Pz
i B N 1+ i
.
M C
A x
O 1

9
 Since 6 AP B = 90◦ , the locus of P is the circle with AB as a diameter
1+i
and hence M = as its centre. We have called it C. Since it is a
2
circle, the point, say N, on it which is closest to the point Q = 2 + 2i
is the closer end of the diameter passing through Q. As the centre of
1+i
C is the point , it is clear that N is the point 1 + i. Its distance
2 √
from Q is |1 + i| i.e. 2.

Another simple problem once the characterisation of perpendicu-

larity of two segments in terms of complex numbers strikes. The paper
setters have been careful to give the numerical data in such a way that
the closest point N to Q can be identified merely by inspection. When
a problem is designed to test one particular idea, it is unfair to com-
plicate it with laborious calculations not relevant to its main theme.

Q.5 Let f : IR −→ IR be a continuous function such that for any two real
numbers x and y,

|f (x) − f (y)| ≤ 7|x − y|201

Then
(A) f (101) = f (202) + 8 (B) f (101) = f (201) + 1
(C) f (101) = f (200) + 2 (D) None of the above.

Answer and Comments: (D). The problem is poorly designed. The

choice of the particular figures 201 and 7 is highly arbitrary to say
the least. One wonders if they were obtained by splitting 2017, the
calendar year in which the test is held!
More seriously, the condition which the function f is given to satisfy
trivially implies that it is continuous. Actually, more is true as will be
proved later. So it is superfluous to give continuity of f .
The options given are matchingly silly. Any constant function
meets all the conditions in the data. So that rules out (A), (B) and
(C). Hence (D) is correct.

A slightly non-trivial problem would have been to ask to show that

under the given conditions, f (x) has to be a constant. This can be

10
 done as follows. Fix any x ∈ IR. Then for any h 6= 0, we have
 
 f (x + h) − f (x) 
0≤ − 0 ≤ 7|h|200 (1)
 
h

 

as we see by taking y = x + h in the given condition. As h → 0,

the R.H.S. tends to 0 and hence so does the L.H.S. by the popular
Sandwich Theorem for limits. (Actually the redundant left half of
(1) was added only to make the Sanwich Theorem applicable. After
all, you need two pieces of bread to make a sandwich!)
Thus we have proved that f (x) is differentiable everywhere and
further that its derivative vanishes identically. Concluding that f is
a constant function from this is often taken as obvious by those who
confuse a statement with its converse. The converse, viz. that the
derivative of a constant function vanishes everywhere is indeed trivial.
But to go the other way, you need Lagrange’s Mean Value Theorem.
For any two a, b with a < b, there is some c ∈ (a, b) such that f (b) −
f (a) = f ′ (c)(b − a). Since f ′ (c) = 0 regardless of where c lies, we have
f (b) = f (a). So, f is a constant.
Note that the constant 7 and the exponent 201 have little role here.
Suppose, more generally, that f : IR −→ IR satisfies

|f (x) − f (y)| ≤ A|x − y|α (2)

where A and α are some positive constants. Then the argument above
will apply equally well if α > 1. For 0 < α < 1, little can be concluded
beyond continuity of f (x). The case α = 1 is most interesting. It is
tempting to think that in this case f (x) must be of the form ±Ax + B
for some constant B. But there are other examples too which we shall
not go into.
It is entirely possible that there is some hidden agenda in asking
this problem. Perhaps the idea is to compare the performance on this
question with that on the rest. If candidates who do well elsewhere fail
in this problem, there is a cause for concern for their training.

Q.6 In the Mathematics department of a college, there are 60 first year

students, 84 second year students and 108 third year students. All of
these students are to be divided into project groups such that each

11
 group has the same number of first year students, the same number
of second year students and the same number of third year students.
What is the smallest possible size of each group?

(A) 9 (B) 12 (C) 19 (D) 21

Answer and Comments: (D). Let n be the number of groups. All

60 students of the first year have to be distributed equally into these n
groups. So 60 is a multiple of n, or equivalently, n divides 60. Similarly
it divides 84 and 108. So it is a common divisor of 60, 84 and 108. The
largest value we can take is their g.c.d. (greatest common divisor). By
inspection n = 12. To finish, we divide the total number of students,
60 + 84 + 108 by 12 to get 21 as the smallest possible size of each group.

It takes more time to read the problem than to work it out. The only
‘thoughtful’ step is to realise that the number of groups is a common di-
visor of 60, 84 and 108. After that, it is a problem of elementary school.
There they fit in nicely, because one of the criteria of good education
is the ability to apply it to real life problems. An alarmingly large per-
centage of college entrants simply dread ‘word problems’. Perhaps it
is to prevent their entry that such a simple problem is designed, which
properly belongs as a puzzle in the recreation section of a newspaper.
(A similar remark was made about Q.20 in Paper 1 of 2015. But that
problem required at least some deductive reasoning. The present one
requires little more than finding the g.c.d. of three numbers which are
from the multiplication table of 12.)
Looking at the social aspects of the data, it would have been better
if the figures 60, 84 and 108 stood for students enrolled for degree
programs in Mathematics, Physics and Chemistry respectively. That
will be a realistic reflection of the relative popularity of these subjects.
As the problem stands, one gets the impression that the Mathematics
program of the college is declining in popularity!

Q.7 Let a, b, c be real numbers each greater than 1 such that

2 3 5
logb a + logc b + loga c = 3
3 5 2

12
If the value of b is 9, then the value of a must be
√
(A) 3 81 (B) 27
2
(C) 18 (D) 27

Answer and Comments: (D). Like Q.2 in Paper 1 of 2014, the

present problem involves logarithms w.r.t. different bases. In such
problems, it is convenient to express all such logarithms in terms of
logarithms w.r.t. some common base. The choice of this common base
is unimportant and it need not even be specified. We take this common
base as 3 because we are given that the value of b is 9 which is a power
of 3. We now write log3 simply as log. Denote log a, log b and log c by
x, y, z respectively. Then by the formula for the change of base,
log a x y z
logb a = = , logc b = and loga c = (1)
log b y z x
Hence the given equation becomes
2x 3y 5z
+ + =3 (2)
3y 5z 2x
We are given the value of b as 9 and hence

y = log 9 = 2 (3)

So, effectively, (2) is an equation in two unknowns x and z. In gen-

eral such an equation will have infininitely many solutions and so it
is impossible to determine x and hence a uniquely from it. A single
equation in several unknowns has a unique solution only in degenerate
cases. A simple example is the equation u2 + v 2 = c in the unknowns u
and v. For c > 0 it has infinitely many solutions while for c < 0 it has
no (real) solutions. But for the degenerate case c = 0, it has a unique
solution u = 0, v = 0.
So the only thing we can hope for is that (2) falls in some degenerate
or extreme case whereby it has a unique solution. Let us write the
three terms on the L.H.S. as u, v, w respectively. Then (2) says that
their A.M. is 1. But by a direct calculation, uvw = 1 and hence the
G.M. of u, v, w is also 1. So this is an extreme case where the A.M.
of three positive real numbers equals their G.M. This can happen only

13
 when all the numbers are equal and hence each one of them is 1 here.
So now we get three separate equations:

2x = 3y, 3y = 5z and 5z = 2x (4)

As y = 2, the first equation gives x = 3 and hence a = 3x = 33 = 27.

The problem is a combination of the worn out A.M.-G.M. inequal-

ity and some elementary facts about logarithms. The former already
appeared in Q.1 above. The latter permits only a limited variety of
problems. So the present problem, although a challenge for the untu-
tored, is a dull one for those who are familiar with such problems.

Q.8 Consider a triangle ABC. The sides AB and AC are extended to points
D and E respectively, such that AD = 3AB and AE = 3AC. Then
one diagonal of BDEC divides the other diagonal in the ratio
√ √
(A) 1 : 3 (B) 1 : 3 (C) 1 : 2 (D) 1 : 2

Answer and Comments: (A). The data is pictured in the figure

below where the diagonals of the quadrilateral BDEC intersect at the
point M.

B C

D E

Since AD : AB = AE : AC = 3, the triangles ADE and ABC are

similar and DE : BC = 3. We also get that BC and DE are parallel
to each other. This makes the triangles BMC and EMD similar.
Hence both BM : MC and CM : MD are BC : DE = 1 : 3.

An absolutely simple problem based on similar triangles.

14
Q.9 The area of the region bounded by the curve y = tan x, the x-axis and
the tangent to the curve at x = π4 is
1 1
(A) loge 2 − 2
(B) loge 2 + 2
(C) 21 (loge 2 − 12 ) (D) 21 (loge 2 + 12 )

Answer and Comments: (C). A straightforward problem about cal-

culating the area below the graph of a function y = f (x). The only
catch here is that the vertical boundaries (which correspond to the end
points of the interval over which f (x) is to be integrated) are not given.
Also the lower boundary of the region, say R, is partly the x-axis and
partly the tangent at the point ( π4 , 1) to the curve y = tan x. So, before
we carry out the integration, we have to look for this missing informa-
tion and the best way is to sketch the region R as we do in the figure
below. The tangent to the curve y = tan x at the point B = (π/4, 1)
is shown by the line L. It cuts the x-axis at the point C. The region
R is bounded above by the curve y = tan x and below by the segments
OC and CB. It is shown by shading by horizontal segments.

y
L

B( π , )
/4 1
E y=c
F A
C π/4 x
O

R T

The upper boundary of the region R is an arc of a single curve y = tan x

from O = (0, 0) to B = ( π4 , 1). But the lower boundary consists of two
parts, the segments OC and CB. To evaluate the area of R, we can
proceed in three ways listed below,

(i) We can divide the region R into two subregions, one lying on the
left of the vertical line through the point C and the other to its
right. For this, we shall first have to identify the point C and for
this we shall first have to find the equation of the line L. The area

15
 of each subregion will have to be calculated by an appropriate
integral.
(ii) Instead of splitting the region R, we add to it the triangle ABC
(shown as T in the figure) to get a bigger region S, which is
bounded above by the curve y = tan x and below by the segment
OA. We calculate the area of S and subtract from it the area of
the triangle T .
(iii) We can evaluate the area of R by horizontal slicing. A typical
horizontal line y = c intersects the region R in a segment EF
where E = (tan−1 c, c) and F will come out after we find the
equation of the line L. If this point is (g(c), c), then the area of R
Z 1
will be tan−1 y − g(y)dy.
0

No matter which approach we take, we have to identify the point C.

For this we first need the equation of the line L, i.e. the tangent to the
d
curve y = tan x at the point ( π4 , 1). Since (tan x) = sec2 x, the slope
dx
of L is sec2 (π/4) = 2. And so its equation is
π
y = 1 + 2(x − ) (1)
4
Hence
π 1
C=( − , 0) (2)
4 2

It is now instructive to compare the merits and demerits of the three

methods above and choose the most convenient one.

(i) For the first approach, we shall have to calcuate two integrals
π
Z π/4−1/2 Z π/4
tan x dx and (tan x − 1 − 2(x − )dx.
0 π/4−1/2 4
Z π/4
(ii) Here we shall have to calculate the integral tan x dx and sub-
0
tract from it the area of the triangle T .
(iii) From (1), the function g(y) comes out as + y−1
2
π
4
. The area of the
1 π 1 −y
Z
region R is then the integral tan−1 y − − dy.
0 4 2

16
There is not much conceptual difference between (i) and (ii). If we split
Z π/4
π
the second integral, viz. (tan x − 1 − 2(x − )dx into two parts
π/4−1/2 4
by splitting the integrand and add the first part to the first integral,
Z π/4−1/2
viz. tan x dx, then by the additivity property of the integrals
0 Z π/4
we get the integral tan x dx which is precisely the integral in (ii).
0
The integral of the remainder is the area of the triangle T . The third
approach, however, is qualitatively different. Here we get the answer
by a single integration. But we shall have to find an antiderivative for
tan−1 y. This can be done using integration by parts. But that means
we shall be spending what we saved.
This often happens in mathematics. No matter which approach you
take, the work involved remains essentially the same. It is only in rare
cases that we can bypass some arduous work by cleverly transform-
ing the problem. In the solution to Q.2, we mentioned complementary
counting as an example of such a transformation. In the present prob-
lem, the second approach resembles complementary counting because
the area of the triangle T can be found without integration. Its base is
π π 1 1
− ( − ) which is simply while its height is 1. So the area of the
4 4 2 2
1
triangle T is . We need to subtract this from the area of S which is
Z 4π/4
the integral tan x dx. Hence the area of the region R is given by
0

π/4 1
Z
A(R) = tan x dx −
0 4
π/4 1 √ 1
= loge (sec x) − = loge 2 − loge 1 −

0 4 4
1 1 1 1
= loge 2 − 0 − = (loge 2 − ) (3)
2 4 2 2
which tallies with (C).

A straightforward problem, but the work needed is time consuming.

But then perhaps that balances Q.5 (where no work is needed) and
Q.6 (where the only work needed is to find the g.c.d. of 60, 84 and
108). So the overall package deal is not bad. But since all questions

17
 carry equal marks and it is next to impossible to finish all the questions
in the given time, those who pick the less demanding problems stand
to gain. Exactly the same problem has been asked in JEE 1988. See
Exercise (17.3)(ix). Admittedly, most problems on finding areas are
similar. Still such an outright duplication should have been avoided.

Q.10 Let V be the set of vertices of a regular polygon with twenty sides.
Three distinct vertices are chosen at random from V . Then the prob-
ability that the chosen triplet of vertices forms a right angled triangle
is
7 1 3 1
(A) 19
(B) 19
(C) 38
(D) 38

Answer and Comments: (B). Probability theory is not in the ISI

syllabus. But the present problem, although superficially a probability
problem, is really two counting problems put together, because the
desired probability, say p, is the ratio of two numbers given by

|F |
p= (1)
|S|

where S stands for the set of all possible outcomes and F stands for the
subset of F consisting of the outcomes that are ‘favourable’, i.e. those
outcomes in which the given event occurs. Here |X| stands for the
number of elements in a set X. The set S is often given a fancy name,
the sample space. This simple minded definition of probability is
applicable only when the sample space is finite and, more importantly,
when all elements of it are equally likely. This is the meaning given to
expressions like ‘at random’.
In the present problem, the sample space S is the set of all ways to
select three elements of the vertex set V . As |V | is given as 20, we get
!
20 20 × 19 × 18
|S| = = = 19 × 60 (2)
3 3!

Note that we have not multiplied 19 and 60. To get p, we have to divide
by |S| and at that time the factorisation 19×60 will be more convenient,
especially so since in all the four options given, the denominators are
multiples of 19. So it is foolish to waste precious time in multiplying 60

18
 and 19 prematurely and also thereby increase the chances of a numerical
slip.
Now comes the real task of finding |F |, the number of favourable
cases. In the present problem, this is the set of those triples {A, B, C}
in which A, B, C are vertices of a right angled triangle. As the order is
unimportant, we suppose that ABC is right angled at B. We are given
that the angular distance between any two consecutive vertices is the
same, say, α. (Actually, α = 2π 20
π
= 10 = 18◦ . But that is not crucial
here.) Hence the angular distance between any two (distinct) vertices
will be a multiple of α, ranging from α to 10α.
Consider a triangle ABC right angled at B. Without loss of general-
ity, assume that the side AB is smaller than (or possibly equal to) the
side BC. Then the angular distance between A and B is nα for some
n = 1, 2, 3, 4 or 5. For each n, there are twenty such segments whose
end points are at an angular distance nα. Each segment determines
two possible locations for the third vertex C. (These two positions
along with A and B form a rectangle.) So, for every n, there are 40
right angled triangles in which the shorter side has angular length nα.
However, for n = 5, AB and BC are equal in length and so the same
triangle gets counted twice. So the total count of triangles is not 40 × 5
but 40 ×4 + 20 = 180. So the favourable set F has 180 elements. Hence
180 3
by (1) and (2), p = = .
19 × 60 19

Although the solution above is reasonably simple, there is a more

elegant way to count |F |. For ABC to be right angled, the longest
side must be a diameter. There are in all 10 diameters whose ends are
from V . For each such diameter, the third vertex can be any of the
remaining 18 vertices. So, |F | = 10 × 18 = 180.
The second solution is based on the geometric result that the angle
in a semi-circle is a right angle. As the solution to Q.4 also hinged on
this idea, there is some duplication.
Q.11 A “basic row operation” on a matrix means adding a multiple of one
row to another row. Consider the matrices
   
x 5 x 0 0 21
A= 1 3 −2   and B =  1 −1 −14 
 

−2 −2 2 0 43 4

19
It is given that B can be obtained from A by finitely many basic row
operations. Then, the value of x is:

(A) 3 (B) −3 (C) −1 (D) 2

Answer and Comments: (A) This is one of those problems where

the success depends upon whether the statement of the problem rings
a familiar bell in some way. What is called a “basic row operation” is
more customarily called an elementary row operation. But whatever be
the name, the technique of adding a multiple of one row to another is
familiar as a tool to simplify a determinant. Under each such operation,
the value of the determinant remains unchanged. So the same is true if
one matrix is obtained from the other by applying any (finite) number
of basic row operations. Here A and B are given matrices. Denote
their determinants by |A| and |B| respectively. Then |B| = |A|. The
latter can be calculated directly as
4
|B| = 21 × = 28 (1)
3
|A| comes out to be a polynomial in the indeterminate x.

|A| = 2x + 5(4 − 2) + 4x = 6x + 10 (2)

Equating |A| with |B| gives 6x = 18 and hence x = 3.

Another simple problem, once the key idea strikes. And the key
idea is not at all foreign. But it is one thing to know a fact. It is
quite another to realise where it will be applicable. In the elegant
solution to the last problem, the key fact needed was that the angle in
a semi-circle is a right angle. It is easy to hit this if you start with a
semi-circle. But it is not so easy to hit it if you start with a right angled
triangle. In the present problem, however, the moment somebody sees
the words ‘adding a multiple of one row to another’ the bell ought to
ring that the determinant is invariant. If it doesn’t, there is no other
way to solve the problem. In the last question, on the other hand,
even if you missed the short cut, the answer can be obtained by other
means. This is probably the dividing line between an elegant solution
and a tricky one. An elegant solution does not preclude other, possibly

20
 more laborious approaches. In fact its elegance stands out more in
comparison to them. In a tricky solution, however, if the trick eludes
you, you are helpless.

Q.12 Let C be a circle of area A and centre at O. Let P be a moving point

such that its distance from O is always twice the length of a tangent
drawn from P to the circle. Then the point P moves along

(A) the sides of a square centred at O, with area 43 A

(B) the sides of an equilateral triangle centred at O, with area 4A
(C) a circle centred at O, with area 34 A
(D) a circle centred at O, with area 4A

Answer and Comments: (C). The area of the circle C is given as

some constant A. It would have been more natural, instead, to give its
radius as some constant, say r. Of course, the two are related by

A = πr 2 (1)

But the radius is a more directly appealing attribute of a circle and it

is not clear why this twist was given.
So, let us take r as the radius of the circle C. Let P T be a tangent
from P to C. Suppose P T = t. We are given that OP = 2t.

T t
P

r
2t

21
 From the right angled triangle OT P we get
r 2 + t2 = 4t2 (2)
So t2 = 13 r 2 . As a r is a constant so is t and hence 2t. Therefore P traces
a circle of radius 2t centred at O. Its area is 4πt2 = 43 π3t2 = 34 πr 2 = 43 A.

An absolutely simple problem requiring nothing beyond the perpen-

dicularity of the tangent and the radius and the theorem of Pythagoras.
Q.13 A moving line intersects the lines x + y = 0 and x − y = 0 at the
points A and B such that the triangle with vertices (0, 0), A and B has
a constant area C. The locus of the midpoint of AB is given by the
equation

(A) (x2 + y 2 )2 = C 2 (B) (x2 − y 2)2 = C 2

(C) (x + y)2 = C 2 (D) (x − y)2 = C 2

Answer and Comments: (B). This problem is of a similar spirit as

the last one. But the last one was a pure geometry problem. In the
present one, the data as well as the conclusion involve coordinates.
One of the advantages of coordinates is that many times it is unnec-
essary to draw a diagram. We can simply translate the data in terms
of algebraic equations and then manipulate it. So, in the present prob-
lem, denote (0, 0) by O. Let the midpoint of AB be P = (h, k). We
take A as (a, −a) for some a and B as (b, b) for some b. Here a, b are
two real parameters. We shall first express h and k in terms of these
two parameters and then eliminate them with some additional piece of
data to get the locus of P .
The first part is easy. Since P = (h, k) is the midpoint of AB, we
have
a + b = 2h (1)
and b − a = 2k (2)
The lines x + y = 0 and x − y = 0 intersect at O are at right angles.
So, AOB is a right angled triangle. Equating its area with C gives
|OA||OB| = 2C (3)

22
Squaring and noting that |OA|2 = 2a2 while |OB|2 = 2b2 , we get

a2 b2 = C 2 (4)

We get the locus of P by eliminating a and b from (1), (2) and (4).
Solving (1) and (2) for a and b,

b = h+k (5)
and a = h − k (6)

Putting these into (4) and replacing h, k by x, y we get the locus of P

as

(x + y)2(x − y)2 = C 2 (7)

i.e. (x2 − y 2 )2 = C 2 .

the standard method to obtain the locus of a point P moving in

the plane subject to some condition is to express its coordinates (often
called current coordinates and usually denoted by (h, k) rather than
by (x, y) which may be involved in some other equations in the data) in
terms of a single parameter, and then eliminate this parameter using the
condition which P is supposed to satisfy. The choice of the parameter
is ours. But when the problem itself contains some 1-parameter family,
the most natural choice is to take the parameter in such a way that as
it varies we get different members of the given family. In the present
problem, this family is the family of all lines through the point (h, k).
The equation of a typical member of this family is of the form

y − k = m(x − h) (8)

In other words, we are taking the slope of the moving line as a param-
eter. Solving (8) simultaneously with x + y = 0 and x − y = 0 we get
the points A and B. They come out to be
mh − k k − mh
A = ( , ) (9)
m+1 m+1
mh − k mh − k
and B = ( , ) (10)
m−1 m−1

23
 respectively. As P = (h, k) is the midpoint of AB, we get
1 1
2h = (mh − k)(+ )
m+1 m−1
1 1
and 2k = (mh − k)( − ) (11)
m−1 m+1
both of which simplify to

h = mk (12)

As a result, we have
(mh − k)2 (h2 − k 2 )2
|OA|2 = 2 2
= 2 2
= 2(h − k)2 (13)
(m + 1) (h + k)
2
(mh − k) (h2 − k 2 )2
and |OB|2 = 2 2
= 2 2
= 2(h + k)2 (14)
(m − 1) (h − k)
Putting these into (3) will give us the same locus as before. But this
derivation is considerably more complicated than the earlier one which
was based on two parameters a and b instead of a single parameter m.
As a general rule, one should minimise the number of parameters. But
as this problem shows, sometimes a clever choice of several parameters
simplifies the work.

Q.14 Let A = {1, 2, 3, 4, 5, 6} and B = {a, b, c, d, e}. How many functions

f : A −→ B are there such that for every x ∈ A, there is one and
exactly one y ∈ A with y 6= x and f (x) = f (y) ?

(A) 450 (B) 540 (C) 900 (D) 5400

Answer and Comments: (C). This question is of the same spirit as

Q.2 of Paper 1 of ISI 2015 which asked for the number of all one-to-one
functions from a set with 3 elements to a set with 6 elements. But the
latter was utterly trivial, being a direct application of the product rule
for counting. In the present problem, we have to do some thinking.
Let F be the set of all functions from A to B which satisfy the
given condition. If f ∈ F , then by very design, f cannot be one-to-
one. Still its failure to be a one-to-one function is well regulated. For
example, it cannot map three elements of A to the same element of B.

24
The condition on f means that the six elements 1, 2, 3, 4, 5 and 6 of
the domain are paired off into three pairs and f takes the same value
on the two elements of the same pair, but different value on those of
the other pairs. (Often such function is called a two-to-one function.)
For example, consider the pairing {1, 5}, {2, 3}, {4, 6} of A. Then f
is uniquely determined by the values f (1), f (2) and f (4), 1, 2, 4 being
the representatives of the three pairs. But these three values must be
different. So, for each pairing, f is like a one-to-one function from a set
with 3 elements (viz., the set of the three pairs in that pairing) to the
set B = {a, b, c, d, e}. Now this problem is very similar to the utterly
trivial problem mentioned above. The value on the first pair can be
chosen in 5 ways. For each such choice the value on the second pair
can be chosen in 4 ways and continuing, that on the third in 3 ways.
So, for each pairing of the domain set, we have 5 × 4 × 3 = 60
elements in F . We now only need to multiply this figure by the number
of possible pairings. In any pairing, the ‘mate’ of 1 can be chosen in
5 ways. Once selected, the remaining 4 elements can be paired in 3
different ways.
Summing up, there are 15 possible pairings and for each pairing
there are 60 functions in F . So the total count is 15 × 60 = 900.

A good problem, certainly far better than the 2015 question which
is absolutely unbecoming for a reputed institute. The reasoning above
looks long because we have given it in detail. But it does not take much
time to conceive it mentally and thereafter the calculations involved are
minimal.
The problem of pairing an even number, say 2n of objects into
n mutually disjoint pairs is interesting. Denote this number by an .
Then it is the number of ways 2n participants in a tournament can
be paired for the first round of matches. Clearly, a1 = 1 and a2 = 3.
In the solution above, we calculated a3 as 15. The formula for an can
be obtained by first writing down a recurrence relation for it and then
solving it. Let S be a set with 2n objects. Fix any one object, say x of
S. In any pairing, the ‘mate’, say y of x can be chosen in 2n − 1 ways.
Once it is chosen, the problem reduces to pairing off the elements of
the set S − {x, y} which is a set with 2n − 2 elements. There are an−1

25
 pairings of it. So, {an } satisfies the recurrence relation
an = (2n − 1)an−1 (1)
for n ≥ 2 with the initial condition a1 = 1. This is very easy to solve
by inspection. We start with 1 and thereafter, every time we multiply
the earlier term with the next odd number. So
an = (2n − 1)(2n − 3) . . . × 5 × 3 × 1 (2)
By supplying the missing even factors, this can also be written as
(2n)!
an =
(2n)(2n − 2)(2n − 4) . . . 4 × 2
(2n)!
= (3)
2n n!
6! 5! 120
In particular, a3 = = = = 15 as we already calculated.
8×6 8 8
Q.15 Two persons, both of height h are standing at a distance h from each
other. The shadow of one person cast by a vertical lamp-post placed
between the two persons is double the length of the shadow of the other.
If the sum of the lengths of the shadows is h, then the height of the
lamp-post is
√  √  √ 
3 1+ 2 3+1
(A) 2
h (B) 2h (C) 2
h (D) √
2 2
h

Answer and Comments: (B). A typical problem about heights and

distances. A good diagram is half the solution.
F

P x Q y R

h h h

C 2h/3 A x E y B h/ 3 D

26
Suppose that the persons are standing at the points A and B and the
foot E of the lamp-post EF is at a distance x from A and y from B.
Then

x+y = h (1)

Let P and R be the heads of the persons and let their shadows fall at
points C and D on the ground. Then F, P, C are collinear and so are
F, R, D. We are given that the lengths of the shadows add to h while
one of them is double the length of the other. We take AC to be the
larger shadow of length 2h/3 and BD to be the shorter one of length
h/3.
We are interested in the length of EF . Normally, we should denote
it by a single symbol. However, here we let Q be the point where
EF meets P R and let z be the length QF . Then the length of the
lamp-post is

h+z (2)

So we would be through if we can find z in terms of h. The triangles

F P Q and P CA are similar. Therefore
z h 3
= = (3)
x 2h/3 2
Hence
3
z= x (4)
2
Similarly, the similarity of the triangles F QR and RBD gives

z = 3y (5)

From (4) and (5), x = 2y. Hence from (1), we have x = 2h/3 and
y = h/3. So, from (4) (or (5)) z = h. Therefore by (2), the height of
the lamp-post is h + h = 2h.

A simple problem which tests a candidate’s ability to visualise

the geometric data correctly and to introduce appropriate notations to
translate the data. The problem could have been made a little more

27
 demanding by giving the distance between the persons to be different
than their common height, and still more demanding by giving the
two persons to have different heights. But that would only increase
the labour involved without testing any new ability. The paper setters
have wisely refrained from such generalisations.

Q.16 Let S be the set of all points z in the complex plane such that
4
1

1+ = 1.
z
Then, the points of S are

(A) vertices of a rectangle

(B) vertices of a right-angled triangle
(C) vertices of an equilateral triangle
(D) collinear

Answer and Comments: (D). Every non-zero complex number has

n distinct n-th roots. Here we want the fourth roots of 1. They are
±1, ±i. Each of these roots will give a separate equation for z and
hence (possibly) four different values of z. Let us obtain them one by
one.
For the root 1, the equation for z is
1
1+ =1 (1)
z
which has no solution. The next equation is
1
1+ = −1 (2)
z
whose only solution is z = − 12 . For the other two roots, we skip the
1 1
details and write only the solutions as z = and z = . Put
i−1 −i − 1
together,
1 1 1
S = {− , , } (3)
2 i − 1 −i − 1

28
As S has only three elements, option (A) is ruled out. To decide which
of the remaining options holds, let us write the second and the third
elements of S in the standard form a + ib with a, b real.
1 −i − 1 −1 − i 1 1
= = =− − i (4)
i−1 (i − 1)(−i − 1) 2 2 2
and similalry,
1 i−1 −1 + i 1 1
= = =− + i (5)
−i − 1 (−i − 1)(i − 1) 2 2 2
Hence, with these cosmetics,
1 1 1 1 1
S = {− , − − i, − + i} (6)
2 2 2 2 2
Clearly all three elements have the same real part, viz. − 21 . So all of
them lie on the vertical line x = − 12 . hence they are collinear.

Another simple good problem requiring little more than elementary

properties of complex numbers. It is interesting that although the
complex number 1 has four distinct fourth roots, the set S has only
three elements. This happens because the fourth root 1 of 1 leads to
(1) which has no solution in C. | Sometimes a hypothetical point ∞
is added to the complex plane to get what is called the extended
∗
complex plane C | ∪ {∞} which is often denoted by C | . This extra
point ∞ is not a complex number. But we allow it to appear in many
algebraic expressions by defining

z+∞=z−∞=∞ (7)

for all z ∈ C,
|

z∞ = ∞ (8)

for all z 6= 0 and

z
=0 (9)
∞
∞
for all z ∈ C.
| Note that the expressions 0 × ∞, ∞ − ∞ and are
∞
undefined. Also care has to be taken in handling equations involving

29
 ∞. For example, even though z1 + ∞ = z2 + ∞, we may not have
| ∗ is a
z1 = z2 . Despite these limitations, the extended complex plane C
very convenient device to take care of some anomalies that arise in C.|

For example, although (1) has no solution in C, it has a solution, ∞

|

in C| ∗ . With this extra member the solution set S now becomes

1 1 i 1 i
S ∗ = {− , − − , − + , ∞} (10)
2 2 2 2 2

| ∗ , there is
Collinearity is not lost because ∞ lies on all lines. In fact in C
no difference between lines and circles. A line is a circle which contains
the point ∞. The points ±1, ±i all lie on the unit circle |z| = 1.
This circle is the image of the line x = − 12 under the transformation
1 z+1
which takes z to 1 + = . This transformation belongs to an
z z
important class of transformations of C | ∗ to C| ∗ called the fractional

az + b
linear transformations because they are of the form i.e. ratios
cz + d
of two linear transformations. We could go on in this wonderland of
C| ∗ . But that is not relevant to the present problem. We mentioned it

because those who study complex numbers in some depth are sure to
∗
encounter C | .

Q.17 A circular lawn of diameter 20 meters on a horizontal plane is to be

illuminated by a light source placed vertically above the centre of the
lawn. It is known that the illuminance at a point P on the lawn is
C sin θ
given by the formula I = for some constant C, where d is the
d2
distance of P from the light source and θ is the angle between the
line joining the centre of the lawn to P and the line joining the light
source to P . Then the maximum possible illuminance at a point on
the circumference of the lawn is
C C C C
(A) √ (B) √ (C) √ (D) √
75 3 100 3 150 3 250 3

Answer and Comments: (C). As usual, we begin with a diagram.

30
 P
N

d
h
α

θ Q
10
M

If a light source is located at a point P , then the illuminance of it at

a point Q is directly proportional to the intensity of the source (which
is denoted by the constant C here and has little role in the problem)
and inversely proportional to the square of the distance between P and
Q (which is denoted here by d). It is also proportional directly to the
cosine of the angle (shown by α) the line P Q makes with the normal
(shown by QN) to the surface on which Q lies. As a result, the smaller
this angle the more is the illuminance. (The same is true for sources
that radiate other forms of energy such as heat. A common application
of this is in the design of solar heaters where the panel that captures
the heat from the sun rays is kept sloping so that it would capture more
energy during the winter months.)
In the present problem, this surface is the horizontal plane and
therefore the normal to it at any point Q is simply the vertical line
through Q. Instead of expressing the law of illuminance in terms of α,
it is given in terms of the complementary angle θ.
It should have been made clear in the data that the vertical height
(denoted by h in the figure) of the light source is variable. Normally,
we deal with a light source which is fixed. In the present problem, if
h is very small (i.e. if the light source is closer to the ground), then
d and hence d2 will also be small. But we may not get maximum
illuminance because the angle θ and hence its sine will also be small.
The higher the height h, the greater is the distance d but so is sin θ.
It would have been less confusing to pose the problem as the problem
of determining the height for which every point on the circumference
will have maximum illuminance. Already the text of the problem is
long. Also to understand the data, a candidate has to draw a diagram.

31
If, further, the candidate has to figure out exactly the purpose of the
problem, it is unfair when the time allotted is meagre (four minutes).
Fortunately, in the present problem, once these hurdles are crossed,
the rest of the work is simple. We express the illuminance as a function
of θ and maximise it. Clearly

d = 10 sec θ (1)

and so the illuminance, as a function of θ is

C sin θ
I = I(θ) =
100 sec2 θ
Simplifying,
C
I(θ) = sin θ cos2 θ
100
C
= (sin θ − sin3 θ) (2)
100

So the problem reduces to maximising the function sin θ − sin3 θ as θ

varies from 0 to π2 . As we are not interested in the value of θ at which
the maximum occurs, we might as well put x = sin θ and maximise the
function f (x) = x − x3 for x ∈ [0, 1]. The only critical point of f in
1
[0, 1] is given by 1 = 3x2 whence x = √ . By a direct substitution,
3
2C
f ( √13 ) = √13 − 3√1 3 = 3√2 3 . Hence the maximum illuminance is √ =
300 3
C
√ .
150 3

A reasonable problem on maximisation, except for the longish and

slightly confusing wording. The problem would have a more natural
appeal if it was posed as a problem of determining the location of the
light source (directly above the centre of the lawn) that will maximise
the illuminance at all points on the circumference of the lawn. Nor
would it entail much extra work. After determining sin θ as √13 for
maximisation, h = 10 tan θ = √102 meters.

32
Q.18 Let f and g be two real-valued, continuous functions defined on the
closed interval [a, b], such that f (a) < g(a) and f (b) > g(b). Then the
area enclosed between the graphs of the two functions and the lines
x = a and x = b is always given by

Rb Rb
(A) |f (x) − g(x)|dx (B) | f (x) − g(x)dx|
a  a
Rb  Rb
(C)  |f (x)| − |g(x)|dx (D) ||f (x)| − |g(x)||dx
 
a  a

Answer and Comments: (A). The classical geometric interpretation

Z b
of a definite integral f (x) dx is that it is the ‘area A(R) of the
a
region, say R, below the graph of f (x)’. Here it is understood that R is
bounded on the sides by the lines x = a and x = b. More importantly,
it is understood that the lower boundary of the R is the x-axis. This
presupposes that f (x) is non-negative throughout the interval [a, b].
Z b
If f (x) ≤ 0 for all x ∈ [a, b], then the integral f (x) dx is negative.
a
In such a case, in order that it equals the area, we have to take the
area below the x-axis as negative. But as a geometric entity the area is
Z b
non-negative. So, in such a case, the area is the integral −f (x) dx
a
and we call it the area ‘below’ rather than ‘above’ the graph y = f (x).
Things get more complicated when f (x) is non-negative on some
parts of [a, b] and negative on the others. Then the graph y = f (x) lies
above the x-axis in some parts and below it in others and the integral
Z b
of f has little geometric significance. In fact, the integral f (x) dx
a
may even vanish (as it always does when f (x) is an odd function and
Z b
b = −a). However, if we take |f (x)| dx, then it measure the area
a
‘between’ the graph y = f (x) and the x-axis. Of course, to evaluate
this area, we need to split [a, b] into several subintervals, over each of
which f (x) maintains its sign. This can be done by identifying the
zeros of f (x) in [a, b]. If they are c1 , c2 , . . . , cr in ascending order, we
set c0 = a, cr+1 = b. Then on each subinterval [ci−1 , ci ], either |f | = f
or |f | = −f .
Note that the x-axis can be thought of as the graph of the identically

33
 zero function g(x). More generally, when f (x) and g(x) are any two
functions, in some parts of [a, b] we may have f (x) ≥ g(x) while in
the rest we have f (x) ≤ g(x). But no matter which possibility holds
Z b
where, |f (x) − g(x)|dx is the area between the graphs of f (x) and
a
g(x). So (A) is correct. Once again, this trick merely gives a succinct
formula for the area. When it comes to actually calculate the area
between the graphs y = f (x) and y = g(x), there is no short cut to
splitting the interval [a, b] into subintervals, in each of which f (x) lies
either entirely above or entirely below g(x). For example, to calculate
the area between the graphs of y = sin x and y = cos x for 0 ≤ x ≤ π,
we need to spilt it into [0, π/4] and [π/4, π].

The present problem tests whether the candidates have been taught
this care. The hypothesis f (a) < g(a) and f (b) > g(b) is really redun-
dant as the question asks to pick up the choice which is always correct.
The hypothesis merely draws attention to the fact in this case it is
mandatory to take the integrand as |f (x) − g(x)|.

Q.19 Consider the function f : IR −→ IR defined as follows:

(
(x − 1) min{x, x2 } if x ≥ 0
f (x) =
x min{x, x1 } if x < 0

Then f is

(A) differentiable everywhere

(B) not differentiable at exactly one point
(C) not differentiable at exactly two points
(D) not differentiable at exactly three points.

Answer and Comments: (C). Yet another problem where a function

f has different definitions at different points of the domain. While
testing the differentiability of such a function, one has to first identify
these turning points where a change of formula occurs and then at
each of these possible trouble makers, check differentiability from the
left and from the right. Usually, separate arguments are needed at each
such point and on each of the two sides.

34
 In the present problem, one obvious point where the function f
changes its definition is, of course, 0. It is given right in the statement
of the problem. But usually there are some hidden points too. For
x > 0, x and x2 are equal at 1. But in the interval (0, 1), x2 is smaller
than x while for x > 1, x is the smaller. So, for x > 0 the definition of
f (x) becomes
(
x3 − x2 if 0 < x ≤ 1
f (x) = (1)
x2 − x if x > 1

We now have all the information we need to test differentiability of

f at 1 and also its right differentiability at 0. Both the functions
x3 − x2 and x2 − x are continuously differentiable everywhere and so
their derivatives at any point c can be safely obtained by considering
their derivatives at x and letting x → c. (This reasoning fails for
functions whose derivatives are not continuous. For example, if g(x) =
x2 sin(1/x) for x 6= 0 and g(0) = 0, then g ′(0) can be shown to exist
from first principles. But it cannot be obtained by letting x → 0 in
2x sin(1/x) − cos(1/x2 ) because the second term does not tend to any
limit.) So, we have

f−′ (1) = lim 3x2 − 2x = 1 (2)

x→1−
f+′ (1) = lim 2x − 1 = 1 (3)
x→1+
and f+′ (0) = lim 3x2 − 2x = 0 (4)
x→0+

From (2) and (3), we see that f is differentiable at 1. Also since f (0) =
0, f (x) is continuous from the right at 0 and (4) gives us one sided
information about its differentiability at 0. To see what happens on
the other side of the border, i.e. for x < 0. We first need the analogue
of (1). At x = −1, x and x1 equal each other. But if −1 < x < 0, then
1
x
< x while for x < −1, x < x1 . (To see such things more easily, give
some particular values to x, e.g. x = − 12 in the first case and x = −10
in the second.) So analogously to (1) we have
(
x2 if x < −1
f (x) = (5)
1 if − 1 ≤ x < 0

35
 Clearly, f is continuous at −1. But at x = 0, lim− f (x) = 1 while
x→1
f (0) = 0, So f is not continuous at 0 and hence automatically not
differentiable at 0 either.
It remains to check differentiability of f at −1. Analogously to (2) and
(3) we have

f−′ (−1) = lim 2x = −2 (6)

x→−1−
and f+′ (−1) = lim 0 = 0 (7)
x→−1+

From (6) and (7), f is not differentiable at −1.

All put together f is differentiable everywhere except at the two
points −1 and 0.

The problem is simple once we identify the intervals on which f is

given by a single formula, viz., are (−∞, −1), [−1, 0), [0, 1] and (1, ∞).
The only catch is that those who miss that f is discontinuous at 0
may falsely conclude its differentiability at 0 from (4) and the fact that
lim− f ′ (x) = lim− 0 = 0. When a function f (x) is discontinuous at a
x→0 x→0
point c, and lim− f ′ (x) exists and equals L (say), it does not follow that
x→c
f−′ (c) = L. This would be the case if f is known to be continuous at
c. Even then, the proof is not trivial. It needs Lagrange’s Mean Value
Theorem applied to f on an interval of the form [c − h, c] for h > 0.

Q.20 Let f : [0, 2] −→ IR be a continuous function such that

1 2
Z
f (x) dx < f (2).
2 0

Then which of the following statements must be true?

(A) f must be strictly increasing.

(B) f must attain a maximum value at x = 2.
(C) f cannot have a minimum at x = 2.
(D) None of the above.

36
Answer and Comments: (C). The L.H.S. of the given inequality is
the average value, say A, of f over the interval [0, 2], while the R.H.S.
is the value of f at a particular point c in the domain. (Here c happens
to be the right end point of the interval [0, 2]. But that makes no
difference.)
When the average is less than a particular functional value f (c), it
does not follow that the function is less than f (c) throughout. This is
sheer common sense. If the average income of some community is, say
Rs.10, 000.00 and some particular person x has an income of Rs. 50,000
it does not follow that he is the richest member of the community. The
community may very well have some persons richer than x, but many
other persons poorer than x can pull the average down. All that can
be said definitely is that such a person cannot be the poorest person,
for if he were then everybody’s income will be at least Rs. 50,000.00
and hence so will be the average income for the community.
The present problem requires the continuous analogue of the rea-
soning in this example. Suppose (C) fails, that is f has a minimum at
2. Then

f (x) − f (2) ≥ 0 (1)

for all x ∈ [0, 2]. So

Z 2
f (x) − f (2) dx ≥ 0 (2)
0

whence
Z 2 Z 2
f (x) dx ≥ f (2) dx = 2f (2) (3)
0 0

Dividing both the sides by the positive number, we get

1 2
Z
f (x) dx ≥ f (2) (4)
2 0

contradicting the hypothesis. So (C) holds.

An absolutely simple problem once the idea of the average value

strikes.

37
Q.21 In a triangle ABC, 3 sin A + 4 cos B = 6 and 4 sin B + 3 cos A = 1 hold.
Then the angle C equals

(A) 30◦ (B) 60◦ (C) 120◦ (D) 150◦

Answer and Comments: (A). The three angles A, B, C always add

to 180 degrees. So, there are only two independent angles and in general
two equations in them will determine them. In the present problem,
these equations are given to be

3 sin A + 4 cos B = 6 (1)

and 4 sin B + 3 cos A = 1 (2)

We can eliminate either one of A and B to get an equation for the

other angle. For example, writing 3 sin A = (6 − 4 cos B) and 3 cos A =
(1 − 4 sin B), squaring and adding gives

9 = (6 − 4 cos B)2 + (1 − 4 sin B)2 (3)

which simplifies to

2 sin B + 12 cos B = 11 (4)

Similalrly by squaring 4 cos B = 6 − 3 sin A and 4 sin B = 1 − 3 cos A

and adding we get

16 = (6 − 3 sin A)2 + (1 − 3 cos A)2 (5)

which reduces to

6 sin A + cos A = 5 (6)

√
We can divide both the sides of (4) by 148 and rewrite it as
1 6 11
√ sin B + √ cos B = √ (7)
37 37 148
and further as
11
cos(B − α) = √ (8)
148

38
where α is the unique angle for which cos α = √6 and sin α = √1 .
37 37
Then we would get
11
B = cos−1 ( √ )+α (9)
148
Similarly, from (6) we can determine A. Having found A and B, C
will come out as π − A − B. But it will require a lot of calculation to
determine C in degrees. So we look for a simpler solution, keeping in
mind that

sin C = sin A cos B + cos A sin B (10)

always holds in any triangle because sin C = sin(π −(A+B)) = sin(A+

B). We simply multiply (1) by cos B and (2) by sin B and add to get

3(sin A cos B + cos A sin B) = 6 cos B + sin B − 4 (11)

Combining this with (10) and (4), we get

11 3
3 sin C = −4= (12)
2 2
which gives
1
sin C = (13)
2
So either C = 30◦ or C = 150◦ . If the second possibility holds then
A + B = 30◦ . That means both the angles A and B√would be smaller
3
than 30◦ . Hence their cosines will be greater than
√ 2
. But that will
3 3
contradict (2) because sin B > 0 and 3 cos A > 2 > 1. So we have
C = 30◦ .

A good computational problem which tests the skill to choose the

right method. Note that although we abandoned (8) which followed
from (4), we did use (4). Instead of (4) we could have used (6) which
is slightly simpler. It is usually difficult to tell right at the start which
approach would be simpler and do only the work needed for it. One
has to be prepared for the possibility that some of the work will turn
out to be superfluous later. But that is the name of the game. If

39
 somebody is really bent on using both (4) and (6), a salvaging feature
is that from (6) we get cos(A − β) = √537 where β is the complementary
angle of the angle α in (9). So, adding, A + B would come out to be
π
2
+ cos−1 ( √11
148
) + cos−1 ( √637 ). This can indeed be simplified to 150◦ .
It is important to note that such a succinct answer as 30◦ was
possible only because the numerical data permitted it. If, instead of
(1) and (2), we had a more general system of equations, viz.
a1 sin A + b1 cos B = c1 (14)
and a2 sin B + b2 cos A = c2 (15)
our first method would still work and will give B and A and finally C.
But the answers would be horrendous. The paper setters deserve to
be commended for giving the numerical data which makes the solution
manageable.
π
Q.22 Let θ = 7
and consider the following matrix
!
cos θ − sin θ
A=
sin θ cos θ

If An means A × A × . . . × A (n times), then A100 is

! !
cos 2θ − sin 2θ cos θ − sin θ
(A) (B)
sin 2θ! cos 2θ sin θ !cos θ
1 0 0 −1
(C) (D)
0 1 1 0

Answer and Comments: (A). As the matrix A depends on θ, denote

it by Aθ . We can similarly define the matrices Aα , Aβ etc. for any real
numbers α, β etc. A direct calculation gives
! !
cos α − sin α cos β − sin β
Aα Aβ =
sin α cos α sin β cos β
!
cos α cos β − sin α sin β − sin α cos β − cos α sin β
=
sin α cos β + cos α sin β cos α cos β − sin α sin β
!
cos(α + β) − sin(α + β)
=
sin(α + β) cos(α + β)
= Aα+β (1)

40
for all α, β. In particular, taking α = β = θ, we have

A2θ = A2θ (2)

and more generally, by induction on n,

Anθ = Anθ (3)

In particular, for θ = π/7,

!
100 cos(100π/7) − sin(100π/7)
A = A100
π/7 = (4)
sin(100π/7) cos(100π/7)
100π
From 7
= 14π+ 2π7
and the periodicity of the trigonometric
!
functions,
cos(2π/7) − sin(2π/7)
A100 is the same as . So (B) is correct.
sin(2π/7) cos(2π/7)

Property (1) is fairly well-known. It is pivotal in deriving (3) on which

the solution to the present problem is based. An alternate derivation
of (3) is possible
! using complex numbers. Every matrix of the form
x −y
where x, y are real numbers, corresponds to the complex
y x
number z = x + iy. This association preserves complex addition and
multiplication. That is, if z1 = x1 + iy1 and z2 = x2 + iy2 are any two
complex numbers, and we denote their corresponding matrices by Az1
and Az2 respectively, then a direct calculation gives

Az1 +z2 = Az1 + Az2 (5)

and Az1 z2 = Az1 Az2 (6)

Note that the matrix Aθ in the statement of the problem is the same
as the matrix Acos θ+i sin θ associated with the complex number cos θ +
i sin θ. Therefore, by repeated applications of (6), (Aθ )n corresponds to
the complex number (cos θ + i sin θ)n . By DeMoivre’s rule,

(cos θ + i sin θ)n = cos nθ + i sin nθ (7)

Therefore, (3) holds for every positive integer n. (Some persons denote
the number cos θ + i sin θ by eiθ . Then a quick proof of (7) comes from
(eiθ )n = einθ . But unless we have independently defined and proved

41
the basic properties of the complex exponential function, this proof is
nothing more than a notational gimmick.)
Yet another interpretation of the matrix Aθ is provided by the
concept of a rotation. Let Tθ : IR2 −→ IR2 denote an anticlockwise
rotation of the plane around the origin O, through an angle θ. Clearly
Tθ (O) = O. For any other point P = (x, y) in IR2 , Tθ (P ) can be
calculated
√ as follows. Let r and α be the polar coordinates of P . Then
r = x2 + y 2 and α is the angle OP makes with the positive x-axis.
The cartesian coordinates x, y can be obtained from r and α by

x = r cos α, y = r sin α (8)

Now let Q = Tθ (P ). Then OQ = OP = r. But OQ makes an angle

α + θ with the positive x-axis. Hence the polar coordinates of Q are r
and θ + α. Therefore, if the cartesian coordinates of Q are x′ , y ′, then

x′ = r cos(θ + α)
= r(cos θ cos α − sin θ sin α)
= cos θx − sin θy
!
x
= (cos θ − sin θ) (9)
y

where the R.H.S. of (9) is the product of a 1 × 2 matrix and a 2 × 1

matrix. By an entirely analogous calculation, we get
!
′ x
y = (sin θ cos θ) (10)
y

(9) and (10) can be combined together to give

! ! !
x′ cos θ − sin θ x
=
y′ sin θ cos θ y
!
x
= Aθ (11)
y

So, if! we denote!the points P and Q of IR2 by the column vectors

x x′
and respectively, rather than by the ordered pairs (x, y)
y y′

42
and (x′ , y ′) which is more usual, we see that the effect of applying the
!
x
rotation Tθ on P is the same as premultiplying the column vector
y
by the matrix Aθ .
Now consider two rotations Tα and Tβ . It is clear that if we first
apply Tβ and then Tα , the net result is the rotation Tα+β . In other
words

Tα+β = Tα ◦ Tβ (12)
!
x
Given any point P = ∈ IR2 , let Q = Tβ (P ) and R = Tα (Q).
y
! !
x′ x′′
Denote Q and R by the column vectors ′ and respectively.
y y ′′
We apply (11) repeatedly, first with θ = α, then with θ = β to get
! !
x′′ x′
= Aα
y ′′ y′
!
x
= Aα Aβ (13)
y

But on the other hand, applying (11) with θ = α + β,

! !
x′′ x
= Aα+β (14)
y ′′ y

Combining (13) and (14),

! !
x x
Aα+β = Aα Aβ (15)
y y

for all x, y ∈ IR2 . From this we cannot hastily conclude that Aα+β =
Aα Aβ , because for matrices it may very well happen that AC = BC
and still A 6= B, even when the matrix C is not the zero matrix. But in
the present case, we are given that (15) holds for all x, y. For any 2!× 2
1
matrix C, it is easy to check by a direct computation that C is
0
simply the first column of C. Hence applying (15) with x = 1, y = 0 we

43
 get that the first column of Aα+β is the same as that of Aα Aβ . Similarly
by applying (15) with x = 0, y = 1 we get that their second columns
are the same. We are now justified in saying that we have an alternate
proof of (1).
Of course, a proof based on a direct computation is much shorter.
But interpreting matrices as transformations and their product as the
composite of transformations opens a whole new branch of mathematics
called linear algebra. The theory of determinants or that of systems
of linear equations properly belongs to this branch of mathematics.

Q.23 Consider all the permutations of the twenty six English letters that
start with z. In how many of these permutations the number of letters
between z and y is less than those between y and x?

(A) 6 × 23! (B) 6 × 24! (C) 156 × 23! (D) 156 × 24!

Answer and Comments: (B). The dependence on the English al-

phabet is superficial. The knowledge that the last three letters of the
alphabet are x, y, z is nowhere needed in the solution. The question
could as well have been posed for permutations of any set S with 26
elements which satisfy the given condition for three specified elements
of S. Such an unwarranted involvement of something extraneous is, in
fact, likely to cause some confusion (which sometimes leads to a contro-
versy) because candidates who have opted for some other language as
their medium of instruction and are diffident about their knowledge of
English may get dissuaded by (incorrectly) thinking that the problem
is beyond their scope.
It is argued, and with some justification, that such settings impart
a certain liveliness to a problem. If every problem in mathematics is
to be posed only after stripping it of these real life links, mathematics
would appear more dull and would be feared more than it is today. But
in an examination, care has to be taken to ensure that these links do
not add to a candidate’s burden. For example, there is nothing wrong
in asking a question about dealing the cards of an ordinary deck. But
it should not require a candidate to know the rules of a particular card
game.
And, if the paper setters are so much in love with the English alpha-

44
bet, one certainly wishes that they used the correct English grammar!
The problem is in the form of a direct question. The verb then has
to be split and the auxiliary part of it must come before the subject.
Thus the word ‘is’ ought to have come before the phrase ‘the number’.
Anyway coming to the mathematical essence of the problem, suppose
the gap between z and y is m letters. Here m can be 0, if there is no
gap, i.e. y comes immediately after z. Now since the gap between y
and x is more than m, x cannot come between z and y. So it has to
come after x and after a gap of at least m + 1 letters.
We classify the permutations according to the value of m. When m = 0,
the permutation begins with

zy×

where the place × is to be filled by any of the 23 letters (i.e. those

other than x, y, z). But after that the remaining 23 places can be filled
freely by the remaining 23 letters (of which x is one). So the number
of permutations where m = 0 is

23 × (23 × 22 × 21 × . . . × 2 × 1) = 23 × 23! (1)

Let us now suppose that m = 1. Then the permutation must begin

with

z×y×× (2)

where the three places marked with an × are to be filled by any of

the 23 symbols (letters other than x, y, z). This can be done in 23 ×
22 × 21 ways. We now have 21 letters left (of which x is one) and they
can occupy these remaining 21 places in any order. So the number of
permutations of the desired type with m = 1 is

23 × 22 × 21 × (21!) = 21 × 23! (3)

The general pattern should be clear from (1) and (3). We expect that
for m = 2 the number of permutations will be

19 × 23! (4)

45
 which can be verified by observing that every such permutation must
begin with
z × ×y × ×× (5)
Thus every time we are getting 23! times an odd number. These odd
numbers start from 23, 21, 19, . . .. So they will end with 3 and finally
1. (We can determine the m for which the coefficient of 23! would be 1.
But that is only an additional labour.) So the answer to the problem
is
23! × (23 + 21 + 19 + . . . + 5 + 3 + 1) (6)
The expression in the parentheses is the sum of an A.P. of length 12.
There is, of course, a formula for it. But the best way is to add them
the way the great mathematician Gauss did when he was 7 years old.
We add together the two end figures 23 and 1 to get 24. Then we add
the ends of the remaining, i.e. 21 and 3 to still get 24. This will go on
till we add the middle two numbers 13 and 11 which also gives 24. So
the total count is
6 × 24 × 23! = 6 × 24! (7)
which tallies with (B).

This is an excellent problem which tests the ability to experiment,

see some pattern and come up with a correct guess. The mathemat-
ical excellence of the problem outweighs the minor slip of grammar
mentioned earlier.
   
Q.24 Let P = √12 , √12 and Q = − √12 , − √12 be two vertices of a regular
polygon having 12 sides such that P Q is a diameter of the circle cir-
cumscribing the polygon. Which of the following points is not a vertex
of this polygon ?
√ √  √ √ 
3−1
√ , 3+1 3+1
√ , 3−1
(A) √ (B) √
 √2 2 2 √2   2 2 √ 2 2
3+1
√ , 1−√ 3
(C) 2 2 2 2
(D) − 21 , 23

Answer and Comments: (D). Both P and Q are at a distance 1

from the origin O which must also be the centre of the polygon since P

46
 and Q are given to be diametrically opposite to each other. Hence all
vertices of it are at a distance 1 from the origin. By a direct calculation,
this is the case for the points in all the four options. So, we cannot
get the odd man out using this preliminary criterion. We need to dig
deeper.
The angular distance between any two consecutive vertices of the
2π π π
polygon is = . The ray OP makes an angle with the x-axis.
12 6 4
So for every other vertex, say V of the polygon the angle OV makes
π kπ
with the x-axis will be of the form + where the integer k will run
4 6
π
from 1 to 11. (For k = 6 we get the angle π + and the corresponding
4
vertex is Q as expected.)
So, to nab the culprit, for each of the given four options, say V ,
we determine the angle which OV makes with the x-axis. Here we are
lucky. The number in the option (D) is more popularly denoted by
ω. It is an imaginary cube root of 1. It can be written in the form
(cos(2π/3), sin(2π/3)). So if V is the vertex in (D), then the angle
2π 2π π
OV is . It this were to be a vertex of the polygon, then −
3 3 4
π
will have to be an integral multiple of . But by a direct calculation,
6
2π π 5π π
− = which is not an integral multiple of . So (D) is correct.
3 4 12 6

A good, but somewhat unusual problem. An exhaustive trial of

all the four options, one by one, until we identify the culprit would
have been very time consuming. The paper setters have been helpful
to those who recognise one of the options as a more familiar complex
number than the others and luckily that being the correct option.
Q.25 Let a, b, c be real numbers such that a = a2 + b2 + c2 . What is the
smallest possible value of b?

(A) 0 (B) −1 (C) − 41 (D) − 12

Answer and Comments: (D). Let us write the given equation as

a − a2 = b2 + c2 (1)

47
 The R.H.S. is non-negative. The L.H.S. is a quadratic in a with a
negative leading coefficient. So its value is non-negative only when a
lies between the two roots 0 and 1. Also the maximum value of the
1
L.H.S. occurs at a = 12 and this maximum is . So this is the maximum
4
b2 + c2 can be. Subject to this, we have to minimise b. This will be the
case when b2 is the maximum and b is the negative square root of this
maximum. For b2 to be maximum, the L.H.S. should be maximum and
1 1
c2 should be minimum. So, we take a = and c = 0. Then b2 = .
2 4
Therefore
1 1
− ≤b≤ (2)
2 2
1
Hence the minimum of b is − . For any smaller value of b, b2 will
2
1
exceed and then (1) cannot hold for any a and c.
4

A simple problem on minimisation but does not fall into any standard
type such as the calculus method or the A.M.-G.M. inequality. It is a
good test of the ability to think fresh rather than be guided by a large
number of problems of the familiar type.

Q.26 Consider the function f : IR −→ IR defined as

 x
 if x 6= 0
f (x) = ex −1

1 if x = 0

Then which of the following statements is correct?

(A) f is not continuous at x = 0.

(B) f is continuous but not differentiable at x = 0.
(C) f is differentiable at x = 0 and f ′ (0) = − 12 .
(D) f is differentiable at x = 0 and f ′ (0) = 12 .

Answer and Comments: (C). To test the continuity of f at 0, we

x
need to find lim x . It is much easier to find the limit of the recipro-
x→0 e − 1

48
 ex − 1
cal, i.e. lim because this limit is, by very definition, the deriva-
x→0 x
1
tive of the function ex at 0. It exists and equals 1. So lim f (x) = = 1.
x→0 1
Hence f is continuous at 0.
For differentiability too, we can consider the reciprocal of f (x). But
unlike limits, the derivative of a function is not so easily related to that
of its reciprocal. So we proceed directly. We have to consider the limit,
say L, defined by

f (x) − f (0)
L = lim
x→0 x
x
− 1
= lim e −1
x

x→0 x
x − ex + 1
= lim (1)
x→0 x(ex − 1)

0
This is a limit of the indeterminate form . So we try l’Hôpital’s rule
0
to evaluate it. Taking the ratio of the derivatives of the numerator and
the denominator, the limit L will be
1 − ex
L = lim (2)
x→0 xex + ex − 1

0
provided this limit exists. But this is again of the form. So, we
0
invoke l’Hôpital’s rule again to get
−ex
L = lim (3)
x→0 xex + 2ex

provided, again, that this limit exists. But this time we have no trouble.
As x → 0, the numerator tends to −1 while the denominator tends to
1
2. So L = − 12 . Therefore f ′ (0) exists and equals − .
2

A good problem requiring two applications of the l’Hôpital’s rule.

It is to be expected that the rule will be needed twice, because in the
ratio (1) the denominator has a zero of order 2 at 0 since both the
factors x and ex − 1 vanish.

49
 The problem can also be solved by advanced methods, which are often
used at the Junior College level without justification. For example, we
can take the Taylor series of ex at 0 (also called MacLaurin series) viz.

x2 x3 xn
ex = 1 + x + + + ...+ + ... (4)
2! 3! n!
If we substitute this into (1), the numerator becomes

x2 x3 xn
− − − ...− − ... (5)
2! 3! n!
while the denominator becomes

2 x3 x4 xn+1
x + + + ...+ + ... (6)
2! 3! n!
In both (5) and (6), the dominating power near 0 is x2 . So the ratio of
(5) by (6) tends to the ratio of the coefficients of this power, i.e. − 21 .
What is involved in this argument is not only the power series
expansion of ex , but also the operation of equating the limit of a sum
of infinitely many terms with the sum of the limits of the individual
terms. Such arguments need justifications. So, it would have been
better to ask this question in Paper 2 which would have shown what
type of justification the candidate has given.

Q.27 Let the function f : [0, 1] −→ IR be defined as

( )
|x − y|
f (x) = max :0≤y≤1 for 0 ≤ x ≤ 1.
x+y+1

Then which of the following statements is correct?

(A) f is strictly increasing on [0, 12 ] and strictly decreasing on [ 12 , 1].

(B) f is strictly decreasing on [0, 12 ] and strictly in creasing on [ 21 , 1].
√
3−1
(C) f√ is strictly increasing on [0, 2
] and strictly decreasing on
[ 3−1
2
, 1].
√
3−1
(D) f√ is strictly decreasing on [0, 2
] and strictly increasing on
[ 3−1
2
, 1].

50
Answer and Comments: (D). All the options involve the increas-
ing/decreasing behaviour of f (x). So the foremost method is to find
f ′ (x). If that fails (e.g. if f happens to be non-differentiable), then we
have to look for other methods.
But, in order to find f ′ (x) we must first find f (x)! And, in the
present problem, the way f (x) is defined, there is no way to apply
some standard formulas to differentiate it. So, we first have to find
some familiar expression for f (x).
Note that for a fixed x, f (x) is defined as the maximum of a function
g(y) of y defined by
|x − y|
g(y) = (1)
x+y+1
for 0 ≤ y ≤ 1. Because the absolute value changes its definition at 0,
and x lies in the same interval as y does, we have to consider two cases
depending upon whether y ≤ x or y ≥ x. Doing so, (1) splits as
( y−x
x+y+1
, if 0 ≤ y ≤ x
g(y) = x−y (2)
x+y+1
, if x ≤ y ≤ 1
which is highly comfortable if for no reason than its familiar form.
There is hardly a test at the 10+2 level where there is no problem
which involves a function defined by different formulas for different
subintervals. So the very sight of the big left curly bracket makes us
feel that we are now on a familiar turf.
We have to maximise this for 0 ≤ y ≤ 1, holding x as a constant.
But since g(y) has a different definitions on [0, x] and [x, 1], we have to
separately find the maximum of g(y) on each of these two subintervals.
Let us call the maximum of g(y) on [0, x] as M1 (x) and its maximum
on [x, 1] as M2 (x). After we calculate them, the greater of the two
will be the maximum of g(y) on the entire interval [0, 1]. (This takes as
much intelligence as is needed to understand that to decide the national
champion in some sport, if the country is divided into two zones, then
a match is played between the two zonal champions and the winner is
the national champion.)
Let us first tackle M1 (x), the maximum of g(y) on [0, x]. Here
x−y
g(y) = . We can use calculus to find the maximum on [0, x].
x+y+1

51
But there is a simpler way. The numerator and the denominator are
both positive. But as y increases the numerator decreases (because of
the negative coefficient of y) while the denominator increases. So the
ratio must decrease. Hence the maximum of g(y) on [0, x] occurs at
the left end point 0. In other words,
x
M1 (x) = g(0) = (3)
x+1

With the second interval [x, 1] we are not so lucky. This time
y−x
g(y) = . Now both the numerator and the denominator in-
x+y+1
crease with y and so our earlier argument breaks down. A perceptive
person can still claim that the ratio is increasing because in the nu-
merator y is clubbed with −x while in the denominator it is clubbed
with the bigger number x + 1. So even though both the numerator and
the denominator increase with y, the same increment in y will cause a
greater proportionate increase in the numerator than the denominator.
(A real life analogy would be that if the price of some commodity is
hiked, between two families that use the same amounts of that com-
modity, the poorer family will feel the impact more.)
What if you lack such fine perceptivity? There is still a simple
way out. We recast g(y) by writing the numerator as y + x + 1 −
y + x + 1 − 2x − 1 2x + 1
2x − 1. Then g(y) = = 1− . The first
x+y+1 x+y+1
term is a constant. In the second term, the numerator is a constant
while the denominator increases with y. So the ratio in the second
term decreases as y increases. But as we are taking its negative, g(y)
increases. (The incorrigible lovers of calculus can differentiate to get
2x + 1
g ′(y) = and observe that it is positive as x > 0. Like
(x + y + 1)2
many other convenient devices such as vehicles and calculators, calculus
makes you forget, and sometimes lose, some of your abilities which
would have been needed in absence of those devices.)
So, no matter which method we follow, g(y) is increasing on the interval
[x, 1]. Hence
1−x
M2 (x) = g(1) = (4)
x+2

52
The function f (x) is the larger of M1 (x) and M2 (x) as noted above.
So,
x 1−x
 
f (x) = max , (5)
x+1 x+2
for all x ∈ [0, 1].
We are still a long way off. We have to decide which of the two num-
bers M1 (x) and M2 (x) is greater. The answer will depend on x. When
it is not easy to directly compare two expressions, we subject them to
a series of simple operations which preserve inequalities. For example,
we can add a constant to both. Or we can multiply both by a positive
constant. We prepare two columns one headed by M1 (x) and the other
by M2 (x). In each column we add the two ‘descendants’ to which the
comparison is reduced. We go on doing this till we reach a stage where
one side can be declared greater than the other by inspection or by
some other means. (The process is reminiscent of a prolonged court
case which goes through hearings after hearings and finally is decided
in favour of one of the parties, usually after several generations of the
original combatants have passed.)

M1 (x) M2 (x)
x 1−x
x+1 x+2
x(x + 2) (1 − x)(x + 1)
2
x + 2x 1 − x2
2
2x + 2x − 1 0
The justifications in each transition are usually obvious and are not
specified, For example, in the very first step we multiplied both the
sides by (x + 1)(x + 2) which is positive for x ∈ [0, 1].
We have thus reduced the problem to deciding whether the quadratic
2x2 + 2x − 1 is positive√or negative. Its leading coefficient is positive
−1 ± 3
and its roots are . We discard the negative square root as x
2
takes√only non-negative values. So, by properties
√ of the quadratics, for
3−1 3−1
x< , M1 (x) < M2 (x), while for x > , M1 (x) > M2 (x).
2√ 2
3−1
For x = , M1 (x) and M2 (x) equal each other.
2

53
 With this information, f (x) now becomes
 √
1−x

x+2
, if 0√ ≤ x ≤ 3−1
2
f (x) = x 3−1
(6)

x+1
, if 2 < x ≤ 1

So, at long last, we have expressed f (x) in a tangible, familiar form.

Even this is not quite rosy, because f (x) has
√
a different definition in√the
two parts of the interval, one from 0 to 2 and the other from 3−1
3−1
2
to 1. But, as noted earlier, candidates taking competitive examinations
at the 10+2 level feel no discomfort with such splitting. More impor-
tantly, the logic we adopted earlier to study the increasing/decreasing
behaviour of g(y) (for a particular constant x) can also be applied √
to
3−1
f (x), which is only a function of x. In the first subinterval [0, 2 ],
the numerator decreases√while the denominator increases. √
So f (x) is
3−1 3−1
strictly decreasing on [0, 2 ]. In the second subinterval [ 2 , 1], even
though both the numerator and the denominator increase with x, the
numerator increases more rapidly and so the ratio increases. (If this
x 1
is still not clear to you, rewrite x+1 as 1 − x+1 and then apply some
brain, or, failing which, differentiate!) Hence option (D) is correct.

There are essentially two parts in the problem. The first is to convert
the given formula for f (x) to the manageable (5) and the second is to
decide the increasing/decreasing behaviour of f (x). The spadework
needed in the first part is considerable. But the reasoning needed in
it is the same as that in the second part of the problem. As a result,
the problem has become very repetitious and laborious. It would have
been a fair question in Paper 2 where you get 15 minutes per question.
In Paper 1, you get only four minutes. So, it would have been a far
better problem to give f (x) as in (5) and ask to identify the intervals
where it is increasing/decreasing.
Q.28 For a positive real number α, let Sα denote the set of points (x, y)
satisfying
|x|α + |y|α = 1.
A positive number α is said to be good if the points in Sα that are
closest to the origin lie only on the coordinate axes. Then
(A) all α ∈ (0, 1) are good and others are not good.

54
(B) all α ∈ (1, 2) are good and others are not good.
(C) all α > 2 are good and others are not good.
(D) all α > 1 are good and others are not good.

Answer and Comments: (C). Let us first note that the adjective
‘good’ here is purely mathematical. It has very little to do with any nice
or desirable qualities of the sets Sα . Unlike adjectives like ‘continuous’,
‘differentiable’, ‘increasing’ or ‘isosceles’ which are universally used, this
particular adjective ‘good’ is only of a local usage. It will not be found,
at least used in this sense, anywhere else in mathematics. The paper
setters could as well have called it ‘bad’ or ‘green’ or anything of their
choice.
The point to note is that the question is designed to test, in part, the
ability to grasp a new concept and answer questions about it. Usually,
in this process, the earlier knowledge we have helps. The ability to
relate a new concept with what you already know is an invaluable
asset and the present question deserves to be commended for testing
this.
Now, coming to the problem, we note that for each α > 0, whether
a point (x, y) is in Sα or not depends only on |x| and |y|. So, if a point
(x, y) is in Sα , so are its ‘mates’ (±x, ±y) in the four quadrants of
the plane. The same holds for the distance of a point from the origin.
Hence to test if some α is good or not, it suffices to assume that (x, y)
is in the first quadrant, where x, y are both non-negative and hence
|x|α + |y|α is the same as xα + y α .
Having got riddance of the absolute value, we can now identify the
portions of Sα for some values of α. The easiest case is α = 1. Here we
have

S1 = {(x, y) ∈ IR2 : x + y = 1} (1)

(Note that we have not specified that x ≥ 0 and y ≥ 0 because it is

our standing hypothesis.) This is the straight line segment joining the
points (1, 0) and (0, 1). Both these points lie on the coordinate axes
and are at a distance 1 from the origin O. Clearly every point on it
other than the two ends is closer to the origin than. So S1 does not

55
satisfy the given condition. Hence 1 is not good. The full S1 is shown
in the figure below.
y

(0,1) B C(1,1)
Sr
Sq
S1

Sp

A x
O (1,0)

S2

The next familiar Sα is S2 . This is simply the circle {(x, y) : x2 + y 2 =

1}. It is also shown in the figure. Here all points on it lie at the same
distance from O. So S2 is not good either. But the comparison of S1
and S2 shows that while S1 is miserably ‘bad’, S2 is only marginally
so. Points on S1 other than those on the axes are much closer to the
origin than 1. In fact, the midpoint of the segment, viz. (1/2, 1/2) is
1
only at a distance √ from O.
2
It is obvious that every Sα is symmetric about the line y = x and also
that the points A = (1, 0) and B = (0, 1) lie on it. Take Sp for some
0 < p < 1. Only the first quadrant portion of it is shown in the figure.
It will lie inside the triangle bounded by the axes and S1 . So again all
its interior points will be closer to O. As p → 0, Sp will become more
and more ‘squarish’ and approaches the union of the segments OA and
OB. When α = q for some q ∈ (1, 2), Sq will lie between S1 and S2
and will be bad.
When α > 2, Sα will bulge more than S2 . One such Sr is shown
above. It is good because it lies outside the circle S2 except for the
points A and B. As α → ∞, Sα will approach the union of the other
two sides AC and BC of the square OACB.
Summing up, option (C) is correct.

56
 The statements made above require a justification which can be
given as follows. A typical point P on Sα can be taken as (r cos θ, r sin θ)
where r is its distance from O and θ ∈ [0, π2 ]. For θ = 0, P = A while
for θ = π2 , P = B. If we exclude these values, then both cos θ and sin θ
are positive real numbers less than 1. So their powers cosα θ and sinα θ
decrease as α increases. Now, for P to lie on Sα we must have

r α (cosα θ + sinα θ) = 1 (2)

For α = 2, the expression in the parentheses is 1 and so r = 1, con-

firming that all points of S2 are at a distance 1 from O. But for α > 2,
cosα θ < cos2 θ while sinα θ < sin2 θ. Hence the parenthetical expres-
sion is less than 1 and therefore r > 1. Hence Sα is good. For α < 2,
the opposite happens and so all points on Sα other than A and B are
at a distance less than 1 from O.
Although the justification needed is simple (requiring little more
than the trigonometric identity cos2 θ + sin2 θ = 1 and the fact that the
powers of a positive real number less than 1 decrease as the exponent
increases), its articulation is unimportant. In a problem like this, one
should ‘feel’ the answer from a few examples. The justification is more
a ritual. So this is a very good problem and belongs to Paper 1.
In passing we mention a possible genesis of the problems. If
2
(x, y) is a point
√ 2 in I2R , by its distance from the origin, we normally
understand x + y . Let us call this the norm or the length of the
point (x, y) now treated as a vector and denote it by ||(x, y)||. Then
for any two points P1 = (x1 , y1 ) and P2 = (x2 , y2 ) in IR2 , the distance
between them can be written as ||v|| where v is the difference vector
−→
P1 P2 = (x2 − x1 , y2 − y1 ). We can rewrite ||(x, y)|| as (x2 + y 2 )1/2 and
further as (|x|2 +|y|2)1/2 . This may seem redundant because |x|2 always
equals x2 . The reason for this rewriting is that we can generalise the
concept of norm (and hence of the distance between two points of the
plane) by replacing 2 by any positive real number α. So, we define the
α-norm of (x, y) to be (|x|α + |y|α)1/α and denote it by ||(x, y)||α. The
usual norm (often called the euclidean norm) ||(x, y)|| now is simply
the 2-norm, ||(x, y)||2.
Note that as α varies,
√ so does the α-norm. For example,√ √if v =
(1, −2), then ||v||2 = 5. But ||v||1 = 3, while ||v||1/2 = ( 1 + 2)2 =

57
 √
3 + 2 2 and ||v||10 = (1025)1/10 which is only slightly greater than 2.
It is, in fact, possible (but not so trivial) to show that as α increases
||v||α decreases for any vector v = (x, y) and tends to max{|x|, |y|} as
α → ∞. This number is therefore denoted by ||v||∞ and called the
∞-norm. Note that for a point P on either of the coordinate axes if
we let v be the position vector of P , then ||v||α is independent of α.
With this terminology, the set Sα in the problem is precisely the
set of those points (x, y) for which ||(x, y)||α = 1. Since ||v||α decreases
as α increases, for α > 2, the points which are at an α-distnace 1 from
O will be at a greater euclidean distance from it. That is why all α’s
bigger than 2 are ‘good’ in the terminology of the question.
One may wonder whether for α 6= 2, the concept of the α-norm and
hence also the derived concept of the α-distance between two points
has any application. Certainly, in the physical world, the euclidean
norm dominates all others. But there are occasions where some other
distance is more relevant. It will take us too far afield to illustrate
them. We only mentioned them to show that the present problem is
not as weird as it may appear at first sight.
Q.29 A water pitcher has a hemispherical bottom and a neck in the shape
of two truncated cones of the same size. The vertical cross section of
the pitcher with relevant dimensions is shown in the figure below.
40

30 . 10
10
40

Suppose that the pitcher is filled with water to the brim. If a solid
cylinder with diameter 24 cm and height greater 60 cm is inserted
vertically into the pitcher as far down to the bottom as possible, how
much water should remain in the pitcher?

(A) 6316π cm3 (B) 6116π cm3 (C) 6336π cm3 (D) 6136πcm3

58
Answer and Comments: (A). Problems on mensuration were very
common in JEE a long time ago. They were based on the formulas
for the volumes of some standard solids such as cones and cyclinders.
Proofs of these formulas were not expected as they require calculus.
Later, the JEE syllabus was revised to include calculus. After that, the
mensuration problems were dropped to make room for problems testing
the continuity or differentiability of some clumsily defined functions.
For a working engineer, the former are far more relevant than the latter.
Mensuration can, of course, be studied under integration. In fact,
finding the volumes of solids, their moments of inertia and areas of
curved surfaces etc. are among the standard applications of integrals.
But the integrals needed are double or triple integrals, which are too
advanced for the JEE and comparable examinations. So, nowadays, the
mensuration problems are relegated to physics where they are usually
expected to be tackled using standard formulas.
The present problem, therefore, deserves a welcome. But the time
given is inadequate. The paper setters have been kind enough to supply
a diagram. But even then the very text of the problem takes more than
a minute to read carefully and to figure out a way to solve it.

8 8

3 24 3 . 10
10
12
M B
h
20 h

24
A

It is given that the cylinder is inserted vertically into the pitcher. This
means that its axis is vertical and hence parallel to the axis of the
pitcher. But this by itself, does not mean that the two axes are the
same, i.e. that the axis of the cylinder passes through the centre, say
M, of the hemispherical portion of the pitcher. This has to be inferred
from the remaining part of the data that the cylinder is pushed as far
down as possible. Although it seems obvious by common sense (and
real life experience), that for this to happen, the axis must coincide

59
with that of the bowl, it needs a proof. We skip this proof. When
the essence of the problem is mensuration, such collateral problems of
geometric optimisation take a back seat. And, in any case, in MCQ’s
justifications hardly matter.
Now coming to the solution, the volume, say V , of the water left
in the pitcher is given by

V = V1 − V2 (1)

where V1 is the volume of the pitcher and V2 is the volume of the

portion of the cylinder immersed into the water. We calculate both
these volumes separately.
The volume of the pitcher is the sum of the volumes of its hemispher-
ical portion and that of its neck. As all distances are in centimeters, the
volumes will be in cubic centimeters. We shall drop the units cm and
cm3 in expressing the lengths and the volumes numerically. The hemi-
sphere has radius 20 and so the volume, say H, of the hemispherical
part is
2π 16000
H= (20)3 = π (2)
3 3
The neck is given to consist of two equal truncated cones. Each trun-
cated cone has height 10 and the radii of its circular faces are 20 and 15.
So the volume, say N, of the neck will be twice that of each truncated
cone, and hence
20π
N = × ((20)2 + (15)2 + 20 × 15)
3
20 × 925 18500
= π= π (3)
3 3
Adding (2) and (3),
34500
V1 = π = 11500π (4)
3
To calculate V2 , we are given that the radius of the cylinder is 12. To
find the volume of the portion inserted, we need to know its height.
Obviously this will be less than the depth of the bowl, which is 20 +

60
 10 + 10 = 40. To find out exactly what it is, we let h be the height of
the portion of the cylinder lying inside the hemisphere. To find h, take
any point A on the lower rim of the cylinder and let B be the point
directly above it lying in the diametrical plane of the hemisphere as
shown in the figure. Then h = AB. The triangle MBA is right angled
at B. MB equals the radius of the cylinder, 12 while MA is a radius
of the hemisphere and hence equals 20. Hence by the Pythagorean
theorem,
q √
h = (20)2 − (12)2 = 4 52 − 32 = 4 × 4 = 16 (5)

(Note the time saved by taking out the common factor 4 from 20 and
12 before squaring them hastily.)
So the total height of the immersed portion of the cylinder is 16 + 20 =
36. Hence

V2 = 36π × (12)2 = 5184π (6)

(Here too, some saving of time is possible by the fact that 123 is a
familiar number 1728.)
To finish the proof we subtract (6) from (4) to get

V = 9500π − 5184π = 6316π (7)

which tallies with (A).

A good problem on mensuration. But the work involved is too much,

especially for those who have forgotten the formulas for the volumes of
the various solids involved. The formulas for the volumes of a sphere
and a cylinder are hard to forget, but that for the volume of a truncated
hπ 2
cone (viz. (r1 + r22 + r1 r2 ) where h is the height and r1 , r2 are the
3
radii of the two circular faces) is likely to have been forgotten because
of long non-use. Deriving it fresh by subtracting the volume of one
cone from another with the same axis is time consuming.
 
Q.30 Let f : [−1, 1] −→ IR be a function such that f sin x2 = sin x + cos x
for all x ∈ [−π, π]. The value of f ( 35 ) is
24 31 33 7
(A) 25
(B) 25
(C) 25
(D) 5

61
Answer and Comments: (B). The function f in the problem is
defined only on the interval [−1, 1]. The data gives its value at a point
if that point is expressed as the sine of some angle. We have to find
the value of f ( 35 ). So we have to begin by expressing 53 as sin x2 for
some x ∈ [−π, π]. This is a simple trigonometric equation. Since x
varies in [−π, π], x2 varies in [−π/2, π/2]. In this interval α = sin−1 35 is
uniquely defined. Moreover it lies in (0, π/2) since 53 > 0. So cos α is
also positive and equals 45 .
So, we now have x = 2α with sin α = 53 and cos α = 45 . Now it is
only a matter of using the relation and expressing sin 2α and cos 2α in
terms of sin α and cos α whose values we know. So,
3 2α
f ( ) = f (sin )
5 2
= sin 2α + cos 2α
= 2 sin α cos α + cos2 α − sin2 α
3 4 16 9
= 2× × + −
5 5 25 25
24 + 7 31
= = (1)
25 25
So (B) is correct.

An extremely simple problem about inverse trigonometric functions.

Normally, when sin α = 53 , there is a tendency to simply assume that
cos α = 45 without discarding the other possibility, viz. − 54 . The data
should have been chosen to penalise such a lack of scruples.

62
 ISI BStat-BMath-UGB-2017 Paper with Comments

Q. 1 Let the sequence {an }n≥1 be defined by an = tan(nθ) where tan θ = 2.

Show that for all n, an is a rational number which can be written with
an odd denominator.

Solution and Comments: Formulas which express tan nθ as a ratio-

nal function of tan θ (i.e. as a ratio of two polynomials in tan θ) are
well known for lower values of n. For example,
2 tan θ
tan 2θ = (2)
1 − tan2 θ
3 tan θ − tan3 θ
and tan 3θ = (3)
1 − 3 tan2 θ
So, one way to tackle the problem is to have a formula which expresses
tan nθ as the ratio of two polynomials in tan θ and then put tan θ = 2.
Such a formula indeed exists and is shown below.
   
n n
1
tan θ − 3
tan3 θ + . . .
tan nθ =  
n
 
n
1− 2
tan2 θ + 4
tan4 θ + . . .

This formula will give the exact value of tan nθ if we put tan θ = 2. It is
also clear that the value will be a rational number because all binomial
coefficients are integers. Moreover, all the terms in the numerator and
all except the first term in the denominator will be even. So the tan nθ
is a rational number which can be written with an odd denominator.
But this formula is not so well known, as its special cases above. So
we give an argument which does not use it, which is possible because
our goal is much less ambitious than finding the exact value of tan nθ.
We proceed by induction on n. For the inductive step, we need to relate
tan(n + 1)θ to tan θ.
For this we use the more basic identity (which is instrumental to derive
both (1) and (2)), viz.

tan α + tan β
tan(α + β) = (4)
1 − tan α tan β

63
Putting α = nθ and β = θ and then substituting tan θ = 2 we get
tan nθ + tan θ
an+1 = tan(n + 1)θ = (5)
1 − tan nθ tan θ
tan nθ + 2
= (6)
1 − 2 tan nθ

2
For tan θ = 2, denote tan kθ by ak . Evidently a1 = tan θ = is a
1
rational number with odd denominator. So the statement is true for
n = 1. Assume that the result is true for n = k, i.e.
p
ak = tan kθ = (7)
q

where p, q are integers and q is odd. Putting this into (5) with n = k,
we get
p
q
+2
ak+1 =
1 − 2 pq
p + 2q
= (8)
q − 2p
which is a ratio of two integers and the denominator, viz. q − 2p is odd
because q is odd. Hence the statement is true for n = k + 1. So, it is
true for all n ∈ IN.

A simple problem once it is realised that we do not need to derive

a formula for tan nθ for a general θ since we are dealing with a specific
angle θ and the assertion to be proved is comparatively weak.
The paper setters have also made the problem too straightforward.
They could have given a slight twist by merely asking to show that
tan(2017θ) is a rational of the desired type, given that tan θ = 2. This
would be perfectly in line with the modern practice of designing at least
one problem in which the data involves the number of the calendar year
in which the test is held. Never mind how superficial this involvement
is. In fact, the candidates would first have to realise that 2017 has no
role and could be replaced by any positive integer n, so that a proof by
induction becomes possible.

64
Q. 2 Consider a circle of radius 6. Let B, C, D and E be points on the circle
such that BD and CE, when extended, intersect at A. If AD and AE
have lengths 5 and 4 respectively,√and DBC is a right angle, then show
12 + 9 15
that the length of BC is .
5
Solution and Comments: The solution to any good geometry prob-
lem (and geometry problems are presumed to be good unless proved
otherwise!) ought to begin with a good diagram. Such a diagram is, in
fact, often needed to understand the problem. And then, if it is well
drawn, it might also suggest a solution. However, it is not necessary,
especially when time is severly limited, that the diagram be drawn to
scale. Certain vital features should not be compromised. For example,
an isosceles triangle should indeed look like an isosceles one. Similarly,
perpendicularity of two lines in the data should not be compromised.
But, if the angle between two lines is given to be, say, 30◦ , it is hardly
necessary to use a protractor in showing this. Similarly, some leeway
can be taken as far as the lengths of some of the segments are concerned.
What matters more is not so much the neatness and mathematical ac-
curacy of the diagram as its ability to convey to you the essence of the
problem, and possibly, a line of attack.
The following diagram fits this prescription.
A

4
5
E

3
D

z
y 12

B x C

65
The angle 6 DBC is clearly shown as a right angle. But the segment
AD which is supposed to have length 5 appears much longer in pro-
portionate comparison with the diameter CD which is given to be 12.
But that does not affect the vital calculations.
Now, coming to the √ solution, the problem asks us to show that
the length of BC is 12+95 15 . If instead of this horrible number, we
had some simple number, say 4, then one way to solve the problem
would reduce to showing that BC = AE and then we can look for
some intermediary (usually not given in the figure) which equals BC
on one hand and AE on the other. The search for such intermediary
would make the problem interesting. That is usually the beauty of
many geometry problems. Things which look unrelated on the face of
it, turn out to be closely related through such an intermediary. The
excitement is akin to that in discovering that the total stranger who
sits next to you on a flight turns out to be your cobrother’s brother!
The
√ present problem is not meant for such excitement. The figure
12+9 15
5
will have to be derived through some hard calculation, possi-
bly involving a quadratic equation. In such a problem, it is best to
introduce symbols for the lengths of the various line segments (other
than those whose lengths are given) and reduce the problem to solving
a system of equations. We begin by calling the sides BC, BD and CE
as x, y, z respectively. Our interest is only in x. But we need other
variables as auxiliary variables. Generally, one should minimise the
number of such auxiliary variables. For example, we could introduce
one more variable, say w, for the length of the segment DE. But that
can be obviated if we observe that since DBCE is a cyclic quadrilateral
and 6 DBC is a right angle, 6 DEC is also a right angle. But then so
is 6 AED and so by Pythagoras theorem we get DE = 3.
Thus we have three unknowns x, y, z and to determine them we need
a system of three (independent) equations. These are easily obtained
from the right angled triangles △DBC, △DEC and △ABC. Keeping
in mind that CD is a diameter and hance has length 12, we get,

x2 + y 2 = 144 (1)
z 2 + 9 = 144 (2)
and x2 + (y + 5)2 = (z + 4)2 (3)

66
From this point onwards, this is purely a problem of algebra. Using (1)
and (2), (3) can be simplified to give

10y + 25 = 8z + 7 (4)

i.e.

5y + 9 = 4z (5)

Substituting from (1) and (2) into this we get

√ √ √
5 144 − x2 + 9 = 4 135 = 12 15 (6)

So, we have finally got an equation in x, in whose value we are inter-

ested. Simplifying and squaring, we have
√
2 (12 15 − 9)2
144 − x = (7)
25
and hence
√
2 (12 15 − 9)2
x = 144 −
25 √
144 × 25 − 144 × 15 − 81 + 216 15
= (8)
√25 √
1440 − 81 + 216 15 1359 + 216 15
= = (9)
25 25
Note that in (8), we have left the products 144 × 25 and 144 × 15 as
they are. This helps in deriving (9) quickly. Minor as they are, such
savings have some elegance. (In the present case, the saving is based
on the distributivity of multiplication over subtraction.)
To finish the solution,
√ we merely need to show that the numerator in
2
(9) equals (12 + 9 15) . Here also a minor saving of time is possible
by taking out 3 as a common factor from 12 and 9. Then
√ √ √ √
(12 + 9 15)2 = 9(4 + 3 15)2 = 9(151 + 24 15) = 1359 + 216 15

which completes the solution.

In the solution above, the results needed from geometry were only the
Pythagoas theorem (which was used four times) and a very elementary

67
 property of cyclic quadrilaterals, viz. that their opposite angles add
to 180◦ . Those who know a little more about circles can get an easier
derivation of (5). The result needed is that for any point A outside a
circle, if a line through A cuts the circle at points X and Y , then the
product AX.AY is independent of the line. This constant, is called the
power of that point w.r.t. that circle. In the present problem, we can
apply it to the lines ADB and AEC to get

5(y + 5) = 4(z + 4) (10)

which implies (5). The proof of this result is based on the fact that an
external angle of a cyclic quadrilateral equals the opposite angle. As
a result, the triangles △AED and △ABC are similar to each other.
Equating the ratios of the corresponding sides we get (10).
Whichever way look at it, this is a simple geometric problem. But
the computations involved hijack it to the domain of algebra. So, this is
a geometric problem which is not ‘so’ good as a ‘purely’ pure geometry
problem!
Q. 3 Suppose f : IR −→ IR is a function given by
(
1 if x = 1
f (x) = 10 1
e(x −1) + (x − 1)2 sin x−1 if x =
6 1

(a) Find f ′ (1)

100
" !#
X k
(b) Evaluate lim 100u − u f 1+ .
u→∞
k=1 u

Solution and Comments: (a) is easy. The second term is differen-

tiable at 1 (if we set its value equal to 0 at 1) with derivative 0 because
for x 6= 1,
1
(x − 1)2 sin x−1 −0 1
= (x − 1) sin (1)
x−1 x−1
which tends to 0 as x tends to 1 because the first factor tends to 0
while the second one is bounded. So the problem reduces to finding
10
the derivative of the first term, viz., e(x −1) at x = 1. By the chain
10
rule this comes out as the value of e(x −1) 10x9 at 1. So f ′ (1) = 10.

68
 For (b), we first recast the expression whose limit is to be taken to
get
100 100
" !#
X k X k
lim 100u − u f 1+ = lim u 1 − f (1 + )
u→∞
k=1 u u→∞
k=1 u
100
" #
X k
= lim u f (1) − f (1 + ) (2)
k=1
u→∞ u
100
X f (1) − f (1 + uk )
= k lim (3)
k=1
u→∞ k/u
where (2) is obtained by moving the limit inside a summation, which
is valid since only a finite sum is involved.
The problem is now reduced to finding, for each k from 1 to
100, the limit in (3) and adding these limits. From the nature of the
ratio appearing in (3), it is tempting to apply Lagrange’s Mean Value
Theorem and to write the ratio as −f ′ (c) for some c in the interval
(1, 1 + uk ). The trouble is that this intermediate point c will depend not
only on k but on u as well. It might be argued that such variation does
not matter because we are only interested in the limit of the expression
as u → ∞ and no matter what c is, as long as it lies in the interval
(1, 1 + uk ), it will tend to 1 from the right as u → ∞. So the limit in
(3) equals lim+ −f ′ (c) and hence it is −f ′ (1) i.e. −10, for every k.
c→1
The catch here is that this argument will require the continuity of
1
f ′ (x) at x = 1. But because of the second term, viz., (x − 1)2 sin( ),
x−1
in the expression for f (x), although f is differentiable at 1, the deriva-
tive is not continuous at 1. (It is helpful to recall here that if we write
y for x − 1, then the hierarchy sin y1 , y sin y1 , y 2 sin y1 , y 3 sin y1 , y 4 sin y1 , . . .
of functions (all of which are set to 0 at y = 0) is a well known example
of the progressively improved behaviour at 0. The first function is not
even continuous at 0, the second one is continuous but not differen-
tiable at 0, the third one (which is involved in the present problem) is
differentiable but not continuously differentiable at 0, the fourth one is
continuously differentiable at 0 but fails to have a second derivative at
0 and so on.)
It is tempting to try to salvage the situation using the all popular
l’Hôpital’s rule. For a fixed k, put y = uk . Then y → 0+ as u → ∞

69
 f (1) − f (y)
and so the limit in (3) is the same as lim+ . By l’Hôpital’s
y→0 y
−f ′ (1 + y)
rule, this equals lim+ . But that does not help because as
y→0 1
noted above, even though f ′ (1) exists, lim+ f ′ (1 + y) does not exist.
y→0
So, l’Hôpital’s rule is not of much help either.
Fortunately, there is a much easier way out. Putting y = uk as above,
the limit in (3) equals −f ′ (1) from the very definition of a derivative.
No fancy theorems are needed for it. As we already calculated f ′ (1)as
100
X
10 in (a), the sum in (3)is simply −10 k which comes out to be
k=1
−500 × 101 = −50500.

The paper setters have done a wise thing by not asking this problem
in Paper 1. It does not take much time to arrive at the correct answer.
Nor is the correct justification very subtle. The problem is more a test
of the ability to realise how a tempting justification will not apply and
it is only in Paper 2 that this ability can be tested.

Q. 4 Let S be a square formed by the four vertices (1, 1), (1. − 1), (−1, 1)
and (−1, −1). Let the region R be the set of points inside S which are
closer to the center than any of the four sides. Find the area of the
region R.
Solution and Comments: This problem is a straight √ replica of a 1995
16 2 − 20
JEE problem. The answer comes out to be . It is obtained
3
by dividing the region into eight mutually congruent subregions each of
which is bounded by a pair of straight lines and a parabola. The reason
the parabolas enter the picture is that each parabola is the locus of a
point which is equidistant from a fixed point (the focus) and a fixed
line, the directrix. So, for a point on one side of the parabola the focus
is closer than the directrix while for a point on the other side of the
parabola, the directrix is closer than the focus. In the present problem,
we have to deal with four parabolas all having their foci at the origin
O but their directrices are the four sides of the square.
For the full solution see Comment No. 4 of Chapter 17. (There is
a minor difference of notation. The region there is denoted by S and

70
 not R. We reproduce here the diagram in the solution.)

y
(−1,1) (1,1)

R1
x = −1
C4
C1
(−1,0) S
V1 S O x
1 C3
y=c
P C2
2
y=x y = 2x +1
(−1,−1) (1,−1)

Although the problem is mathematically a replica of the JEE

1995 problem, it is probably picked from some American source as
suspected from the spelling ‘center’ rather than ‘centre’. The latter is
used consistently elsewhere.

Q. 5. Let g : IN −→ ZZ with g(n) being the product of the digits in n.

(a) Prove that g(n) ≤ n for all n ∈ IN

(b) Find all n for which n2 − 12n + 36 = g(n)

Solution and Comments: The digits of a number n depend on the

base chosen. Usually, we take the base as 10. But this should have
been specified in the definition of g(n). Also it is to be assumed that
there are no leading zeros as otherwise g(n) will not be well defined.
The function g is not a frequently occurring function. When a new
function is defined for a problem, it is customary to illustrate it with
one or two examples. Trivially, g(n) = 0 if even one digit of n is 0. Also
the g(n) remains the same if the digits are reshuffled in any manner.
Thus g(1224) = g(2142) = 16.

71
 Trivially, if n is a single digit number i.e. a number from 1 to 9, then
g(n) = n. Beyond 9, g(n) cannot grow as fast as n, because multiples
of 10 will have to be moved to the 10’s place where they will be only
single digits. But the digit in the unit’s place is at most 9. So if n is
a two digit number, say ab, then n = 10a + b. But g(n) = a × b ≤ 9a.
Hence g(n) < n.
A similar argument applies for any number with more than one
digit. Let the digits of n, read from the right (i.e. the unit’s place)
be x0 , x1 , x2 , . . . , xr where xr 6= 0. We may call r + 1 the ‘length’ of
n. (Actually, r is the integral part of log10 n. But that is not so vital
here.) We apply induction on r. For r = 0, n is a single digit number
x0 and both g(n) and n equal x0 . Suppose that the assertion is proved
for all numbers whose length is k + 1. Now suppose n is a number with
k + 2 digits x0 , x1 , . . . , xk and xk+1 , read from the right to left. Let m
be the number whose digits are x1 , x2 , . . . , xk+1 , again read backwards
from the unit’s place digit which is now x1 . (For example, if n = 2337
then m is simply 233.) Clearly,

n = 10m + x0 (1)

Since m has length k + 1, by the induction hypothesis, we have

g(m) = x1 × x2 × . . . × xk+1 ≤ m (2)

But g(n) is simply x0 g(m) and x0 is at most 9. So,

g(n) = x0 g(m) ≤ 9g(m) ≤ 9m (3)

Putting (1) and (3) together,

g(n) ≤ 9m < 10m + x0 (4)

So the assertion is true for all numbers of length k + 2. Hence by

induction, it holds for all n. (In fact, strict inequality holds for all
numbers except single digit ones.)

(b) If n2 − 12n + 36 = g(n), then combining this with the inequality

in (a), we have

n2 − 13n + 36 ≤ 0 (5)

72
 Factorising the L.H.S.,
(n − 4)(n − 9) ≤ 0 (6)
which is possible only when 4 ≤ n ≤ 9, i.e. for n = 4, 5, 6, 7, 8 and 9.
In all these cases, g(n) = n. But since n2 − 12n + 36 is also a perfect
square, for it to equal g(n) and hence n, n must also be a perfect
square. Hence the only possibilities are n = 4 and n = 9. By actual
substitution, (n − 6)2 equals 4 and 9 respectively for these values. So
(b) holds only for n = 4 and n = 9.

Part (b) is a trivial consequence of (a). It is probably meant only to

give some consolation credit to candidates who cannot prove (a), but
can derive (b) from it. The crux of the problem is (a). Even this is fairly
easy once you do some experimentation with small values of n. So, this
problem is a good test of a candidate’s ability to understand a new
concept, do some experimentation and get the clue to the right answer
(in the present problem, induction on the number of digits). These
desirable qualities do not come automatically with drill work. There
are candidates who have huge repertoires of solved problems. Ask
them to prove a trigonometric identity about triangles and they will
immediately tell you which standard identity to start with. Similarly,
they know how to pull the right trick to effortlessly evaluate some
integral. But ask them to think fresh and they draw a blank. So a
few problems of the present type are welcome and Paper 2 is the right
setting for them.
Q. 6 Let p1 , p2 , p3 be primes with p2 6= p3 such that 4 + p1 p2 and 4 + p1 p3
are perfect squares. Find all possible values of p1 , p2 , p3 .

Solution and Comments: It is not given that p1 , p2 , p3 are odd. But

this follows from the rest of the data. The only even prime is 2. So if
one of the three primes is even then at least one of the numbers 4+p1 p2
and 4 + p1 p3 will be either 8 (if two of the three primes are 2) or be of
the form 4 + 2k for some odd integer k. 8 is not a perfect square. Nor
is a number of the form 4 + 2k with k odd because it is divisible by 2
but not by 4.
So, we assume that p1 , p2 , p3 are all odd. It is given that p2 6= p3 . We
claim that they are all distinct. For otherwise, p1 = p2 (say). But then

73
 4 + p21 = u2 which gives p21 = (u + 2)(u − 2). By uniqueness of prime
factorisation, p1 will equal both u + 2 and u − 2, which is impossible.
Thus we assume that p1 , p2 , p3 are three distinct odd primes. Since
4 + p1 p2 = u2 for some u, we once again have p1 p2 = (u + 2)(u − 2),
Hence the smaller of p1 and p2 equals u − 2 and the other u + 2. So the
primes p1 and p2 differ by 4. By a similar reasoning, p1 and p3 differ by
4. Since p2 and p3 are distinct, p1 must be the smaller member of one
pair and the larger one of the other. Without loss of generality, we may
suppose that p2 < p1 < p3 . So the problem reduces to finding all triplets
(p2 , p1 , p3 ) of primes which form an A.P. with common difference 4. All
primes greater than 3 are of the form 6k + 1 or 6k − 1. The only way
p1 and p3 can differ by 4 is if p1 = 6k + 1 and p3 = 6k + 5 for some
integer k. But then p2 has to be 6k + 1 − 4 i.e. 6k − 3, which is a prime
only for k = 1.
Thus the only possible solutions are p1 = 7 and p2 = 3, p3 = 11
(or vice versa). That these primes actually give solutions is seen from
4 + 7 × 3 = 25 = 52 and 4 + 7 × 11 = 81 = 92 .

A good, simple problem based on uniqueness of prime factorisation

and the fact that all odd primes greater than 3 are of the form 6k ± 1.
Initially, the candidate has to do some work to rule out the degenerate
cases where one of the primes is 2 or where two of them are equal. This
work is not relevant to the main theme of the problem. Still, in a test
where candidates have to justify their answers, a scrupulous candidate
may spend time writing the justification, lest he loses credit. If the
paper setters do not really intend to test this, he would be made a fool
of. In that case it would have been better to give p1 , p2 , p3 as three
distinct odd primes right in the statement of the question.

Q. 7 Let A = {1, 2, . . . , n}. For a permutation P = (P (1), P (2), . . . , P (n)),

of the elements of A, let P (1) denote the first element of P . Find the
number of all such permutations P so that for all i, j ∈ A :

(a) if i < j < P (1), then j appears before i in P ; and

(b) if P (1) < i < j, then i appears before j in P .

Solution and Comments: In such problems, it pays to work out at

74
least one special case by hand, for a small value of n. If this value of
n is too small like 1 or 2, the work might not lead to any clue for the
general case. On the other hand, too large a value would entail a lot
of work just for trial. Let us take a reasonably small value, say n = 7.
So we have a permutation P of the symbols 1, 2, 3, 4, 5, 6 and 7. in
all there are 7! permutations. But we have to count only those which
satisfy the conditions (a) and (b). So, let us fix the first element of
P as, say 5. (Again, the choice of the extreme values 1 and 7 should
be avoided as they may not be sufficiently representative. Similarly
P (1) = 4 is ruled out on the ground that the symmetric location of 4
between the extremes 1 and 7 may lead to some simplifications which
will not hold in general.)
So, suppose P = (5, a, b, c, d, e, f ) is a permutation in which
a, b, c, d, e, f are integers from 1 to 7, other than 5. Condition (a)
means that all the integers less than 5 (i.e. the integers 1, 2, 3 and 4)
must appear in a descending order in the sequence (a, b, c, d, e, f ) while
(b) means that the integers 6 and 7 must appear in an ascending order
in this sequence. Once this is realised, it is clear that our freedom is
over the moment we fix the four places among a, b, c, d, e, f which the
integers from 1 to 4 should occupy, becausue once they are fixed, we
must send 1, 2, 3, 4 to then in an descending order and now only two
places are left which will have to be filled by 6 and 7 in an ascending
order.
The net result is that ! with n = 7 and P (1) = 5, the number of
6
desired permutations is . The same reasoning applies if P (1) has
4
any other value from 1 to 7. (In the extreme case where P (1) = 1, the
integers 2 to 7 must appear in the descending
! order, viz. as 7, 6, 5, 4, 3
6
and 2 and this is consistent with = 1. Therefore the total number
0
of permutations of the desired type for n = 7 is
7
!
X 6
(1)
k=1 k−1

6
!
X 6
With a change of index, this sum is simply . It is well known
r=0 r

75
 that this sum is 26 . (One way to see this is to put x = 1 in the binomial
expansion of (1 + x)6 .)
We hardly need to give the details elaborately for the general case.
For each m = 1, 2, . . . , n, the number
! of permutations of the desired
n−1
type for which P (1) = m is . Hence the total number of such
m−1!
n
X n−1
permutations is the sum . With a change of index and the
m=1 m − 1
binomial theorem, this comes out to be 2n−1.

Another good problem testing the ability to understand, focus on the

essentials, do experimentation and get a clue. The problem is similar
to another problem given below.
Find the number of permutations of the set {1, 2, . . . , n} in which
every integer except the one appearing last, is followed, although not
necessarily immediately followed, by another integer which differs from
it by one.
In this problem, if P is a desired permutation, then all the integers
less than P (n) must occur in an ascending order, while those greater
than P (n) must occur in a descending order. So we get the same answer
as above, viz. 2n−1 . But in this problem it takes some time to get this
realisation. The reasoning is as follow. Suppose P (n) = r. If r > 1,
then 1 appears somewhere before r. But 2 must not appear before 1
because it is the only ‘neighbour’ 1 has. So, 1 is followed by 2. Now 3
must appear later because it is the only neighbour of 2 which is now
left (the other neighbour 1 having appeared before 2). This will go on
till r. For integers bigger than r we start from the highest one, viz. n
which has only one neighbour, viz. n − 1. So, n − 1 must come after
n. But then it has only one neighbour, viz. n − 2 left to follow it.
Continuing like this, integers bigger than r must occur in a descending
order.
The present problem is of a similar spirit but simpler. Hence it is a
reasonable question to ask.

Q. 8 Let k, n and r be positive integers.

(a) Let Q(x) = xk + a1 xk+1 + . . . + an xk+n be a polynomial with real

76
 Q(x)
coefficients. Show that the function is strictly positive for
xk
all real x satisfying
1
0 < |x| < n
1+ |ai |
P
i=1

(b) Let P (x) = b0 + b1 x + . . . + br xr be a non zero polynomial with

real coefficients. Let m be the smallest number such that bm 6= 0.
Prove that the graph of y = P (x) cuts the x-axis at the origin
(i.e., P changes signs at x = 0) if and only if m is an odd integer.
Q(x)
Solution and Comments: (a). Denote by f (x). For x 6= 0,
xk
f (x) is simply the polynomial
f (x) = 1 + a1 x + a2 x2 + . . . + an xn (2)
f (0) is not defined. But if we set f (0) = 1 then f (x) is continuous
at 0. Since f (0) = 1 > 0, by continuity, f (x) > 0 for all x in some
neighbourhood (−δ, δ) of 0. But the problem asks us to show that this
is the case for a very specific δ, viz.
1
δ= n (3)
1+ |ai |
P
i=1

For this we need to analyse f (x) closely. The δ given to us is the

n
X
reciprocal of 1 + |ai | and so the condition |x| < δ is equivalent to
i=1
n
1 X
| | > 1+ |ai |. This suggests that it would be more convenient to
x i=1
1
make a substitution x = to express f (x) in terms of some function,
y
say g(y), of y and then show that g(y) is positive for all y for which
n
X
|y| > 1 + |ai |.
i=1
1
The first part is easy. Put y = which is valid for x 6= 0. Then
x
y n + a1 y n−1 + a2 y n−2 + . . . + an−1 y + an
f (x) = (4)
yn

77
The numerator is a monic polynomial of degree n, say g(y), in y. So,
for all y 6= 0,
g(y)
f (x) = (5)
yn
We first assume x > 0 and hence y > 0. Then the denominator y n is
positive and so proving f (x) > 0 for 0 < x < δ is equivalent to proving
that

g(y) = y n + a1 y n−1 + a2 y n−2 + . . . + an−1 y + an > 0 (6)

whenever,
n
X
y >1+ |ai | (7)
i=1

We are not given anything about the signs of the coefficients a0 , a1 , . . . , an−1 , an .
The positive coefficients will cause no problem. So, it suffices to con-
sider only those terms ai y n−i for which ai < 0. But we shall give a
proof which is independent of the signs.
Note first that (7) ensures that y > 1 and hence y r < y s whenever
r < s. Now since y > 1 + |a1 | + |a2 | + . . . + |an |, we have

yn = yy n−1
> (1 + |a1 | + |a2 | + . . . + |an )y n−1 )y n−1
= y n−1 + |a1 |y n−1 + |a2 |y n−1 + . . . + |an−1 |y n−1 + |an |y n−1 (8)
> |a1 |y n−1 + |a2 |y n−2 + |a3 |y n−3 + . . . + |an−1 |y + |an | (9)

where in going from (8) to (9), we have dropped the first term y n−1
which is positive and from the third term onwards, replaced higher
powers of y by lower ones.
(9) gives a lower bound on the first term of the R.H.S. of (6). Keeping
the other terms as they are, we get

g(y) > (a1 + |a1 |)y n−1 + (a2 + |a2 |)y n−2 + . . .
+(an−1 + |an−1 |)y + (an + |an |) (10)

whenever (7) holds. Now, regardless of the sigh of ai , the expression

ai + |ai | and hence also the term (ai + |ai |)y n−i is non-negative. As this

78
 n
X
holds for every term of (10), we get that g(y) > 0 for all y > 1 + |ai |.
i=1
Going back to x we have proved that
Q(x)
f (x) = >0 (11)
xk
for all real x satisfying 0 < x < δ. We still need to prove this for
1
−δ < x < 0. We can again put y = and argue as above. But now
x
y will be negative and so we shall have to be more careful in handling
inequalities involving powers of y. There is a way to avoid all this by
applying what we have already done to a different polynomial. In (2)
we replace x by −x. This gives a new polynomial in x, say h(x), defined
by
h(x) = 1 + b1 x + b2 x2 + . . . + bn xn (12)
where bi = (−1)i for i = 1, 2, . . . , n. Still, |bi | = |ai | for all i and so
1
δ= n (13)
1+ |bi |
P
i=1

If −δ < x < 0, then 0 < −x < δ and so by applying our work to h(x)
we have
h(−x) > 0 (14)
But h(−x) is the same as f (x). So we get that f (x) > 0 for −δ < x < 0
too. This completes the proof in all cases.

(b) This is of the same spirit as (a) but now we are more relaxed
because we merely have to see what happens in a sufficiently small
neighbourhood of 0. The size of this neighbourhood is immaterial as
long as it is positive. The coefficients b0 , b1 , . . . , bm−1 all vanish. But
bm 6= 0. So, we write
P (x) = bm xm (1 + a1 x + a2 x2 + ar−m xr−m ) (15)
bm+i
where ai = for i = 1, 2, . . . , r − m. The last factor is positive
bm
at x = 0 and hence, by continuity, for all x in some neighbourhood of

79
0. So, the sign of P (x) in this neighbourhood is the same as that of
bm xm . bm is a constant. If m is even, then xm > 0 for all x in this
neighbourhood (except at 0) and hence P (x) maintains its sign as x
passes through 0. Depending upon the sign of bm , P (x) will be either
positive on both the sides of 0 in this neighbourhood or negative on
both the sides. But if m is odd, then P (x) will have opposite signs on
the two sides.

Part (b) is absolutely trivial as compared with (a) and does not
need (a) for its solution. One fails to see the purpose behind asking
it. Perhaps, as in Q. 6, the idea is to award some consolation credit to
those who can do (b) but not (a).
But Part (a) deserves some comments. In essence, it shows that all
the real roots of the monic polynomial y n + a1 y n−1 + . . . + an−1 y + an
n
X
lie in the interval (−M, M) where M = 1 + |ai |. This upper bound
i=1
on the size of a real root of a monic polynomial is due to Lagrange.
As noted in the solution, we could have as well dropped the terms with
positive coefficients. In fact, one can concentrate on the ‘worst’ term,
i.e. one for which ai is negative and has the largest absolute value. This
way we get sharper upper bounds. Considerable work has been done in
cornering the roots of a polynomial in as small as intervals as possible,
because that is the next best thing to do when one cannot obtain the
roots exactly. (And for polynomials of degree 5 or above, there is no
formula to express the roots in terms of the radicals of expressions
involving the coefficients.) The quadratic formula is exceptional and
deceptive.

80
 CONCLUDING REMARKS

As is to be expected, the entire test is a mixed bag. Still, Paper 1 is much

better than Paper 1 of 2015 which was highly disappointing.
Unlike in the past three years, there is no question which stands out as
the ‘man of the match’. If a selection has to be made there will be a tie
between Q.23 of Paper 1 (about restricted permutations of the letters of the
English alphabet) and Q.8 of Paper 2 (based on Lagrange’s upper bound on
the absolute values of the roots of a polynomial).
Some areas such as mensuration have appeared after a long gap (Q.29
of Paper 1). Mercifully, there is no question regarding the integral part
function!. Of course, there are at least three questions which involve the other
favourite of paper setters, viz. the absolute value function (Q.3, Q.27 and
Q.28). The first one can be answered in a sneaky manner. Q.27 is laborious,
because even though the essential idea is simple, it has to be applied again
and again. Q.28 is a good question testing a candidate’s ability to grasp a
fresh concept. The same thing can be said about Q.5 in Paper 2.
Overall, geometry has gotten a large representation. Q.2, Q.4, Q.8, Q.10,
Q.12, Q.15 Q.17 and Q.24 of Paper 1 and Q.2 of Paper 2 are all based on pure
geometry, sometimes coupled with something extra such as trigonometry
or complex numbers. There is only one locus problem (Q.13 of Paper 1).
It is different from the ordinary locus problems in that the choice of two
parameters is more advantageous than one. Q.21 in Paper 1 (about finding
an angle of a triangle given two trigonometric equations) is also interesting.
Comparatively, number theory has got a beating. Q.6 of Paper 1 involves
absolutely nothing more than finding the g.c.d. of three numbers. The only
other number theory problem (Q.6 in Paper 2) about three primes is also
way too elementary.
Combinatorics has fared better. Q.14 of Paper 1 about finding the number
of two-to-one functions is far better than Q.2 of ISI 2015 which was absolutely
trivial. Both the problems on permutations are good, the better one (Q.23
of Paper 1) was already mentioned.
There are three problems about finding areas. Two of them (Q.9 from
Paper 1 and Q.4 from Paper 2) require integration. Both are good problems.
But it is shocking that they are direct replicas of old JEE problems. This is
really ironic. At the end of the JEE 2017 commentary, ISI 2016 paper was
mentioned with approval as a model JEE can emulate. But now we have two
earlier JEE problems appearing in 2017 ISI.

81
 So, it is now difficult to rank one test as consistently better than the other.
Let us hope that in the coming years, there will be a healthy competition
between JEE and ISI, each trying to come up with some original problems.

82