EDUCATIVE COMMENTARY ON

JEE 2015 ADVANCED MATHEMATICS PAPERS

(Last Revised on May 25, 2019)

Contents
Paper 1 3

Paper 2 40

Concluding Remarks 80

The year 2013 represented a drastic departure in the history of the JEE.
Till 2012, the selection of the entrants to the IITs was entirely left to the
IITs and for more than half a century they did this through the JEE which
acquired a world wide reputation as one of the most challenging tests for entry
to an engineering programme. Having cleared the JEE was often a passport
for many lucrative positions in all walks of life (many of them having little to
do with engineering). It is no exaggeration to say that the coveted position
of the IIT’s was due largely to the JEE system which was renowned not
only for its academic standards, but also its meticulous punctuality and its
unimpeachable integrity.
The picture began to change since 2013. The Ministry of Human Re-
sources decided to have a common examination for not only the IITs, but
all NIT’s and other engineering colleges who would want to come under its
umbrella. This common test would be conducted by the CBSE. Serious con-
cerns were raised that this would result in a loss of autonomy of the IITs
and eventually of their reputation. Finally a compromise was reached that
the common entrance test conducted by the CBSE would be called the JEE
(Main) and a certain number of top rankers in this examination would have
a chance to appear for another test, to be called JEE (Advanced), which
would be conducted solely by the IITs, exactly as they conducted their JEE
in the past.
So, in effect, the JEE (Advanced) from 2013 took the role of the JEE in the
past except that the candidates appearing for it are selected by a procedure
over which the IITs have no control. So, this arrangement is not quite the
same as the JEE in two tiers which prevailed for a few years. It was hoped
that now that the number of candidates appearing for the JEE (Advanced)
is manageable enough to permit evaluation by humans, the classic practice
of requiring the candidates to give justifications for their answers would be
revived at least from 2014, if not from 2013 (when there might not have
been sufficient time to make the switch-over). But this has not happened
even in 2015 and after a change of regime at the Union Government. The
JEE (Advanced) 2015 is completely multiple choice type and its pattern
differs little from that of the JEE (Advanced) 2014. Questions with single
digit answers come before rather than after those with one or more correct
answers. Also negative marking has been revived for the latter.
Academically (and socially), the JEE (Advanced) has the same status as
the JEE in the past. So, from 2013, the Educative Commentary to JEE
Mathematics Papers is confined only to JEE (Advanced). This year, the
numbering of the questions will be that in Code 8 of the question papers.
As in the past, unless otherwise stated, all the references made are to the
author’s book Educative JEE (Mathematics) published by Universities Press,
Hyderabad. The third edition was made available online on the author’s blog
http://www.mathjeecommentary.blogspot.in but has now been withdrawn as
the book is easily available in market.
Because of the multiple choice format and many other constraints in paper-
setting, interesting questions in mathematics are getting rarer. The contin-
uation of these annual commentaries has been possible largely because of
the keen interest shown by the readers. These commentaries are prepared
single-handedly and hence are prone to mistakes of spelling, grammar and oc-
casionally, wrong symbols (but, hopefully, not mistakes of reasoning!). Many
alert readers in the past had pointed out some such errors. They were cor-
rected and the corrected versions were uploaded from time to time. But by
that time the time relevance was reduced.
As an experiment, a draft version of this year’s commentary on both the
papers was uploaded. Those readers who noticed any errors in it were invited
to send an email to the author at [email protected] or send an SMS
to the author at 9819961036. Alternate solutions and any other comments
were also solicited. This really paid off. Several readers, notably Siddhesh
Naik and Deepanshu Rajvanshi pointed out several corrections and suggested
some additions. The elegant solution to Q.55 in Paper 1, where the given
determinant is expanded by writing it as a product of two determinants was
given by Siddhesh Naik. I am grateful to all these and other readers.

2
 PAPER 1

Contents

Section - 1 (One Integer Value Correct Type) 3

Section - 2 (One or More than One Correct Choice Type) 13

Section - 3 (Matching the Pairs Type) 32

SECTION 1
One Integer Value Correct Type

This section contains eight questions. The answer to each question is a

SINGLE DIGIT INTEGER ranging from 0 to 9, both inclusive.
Marking scheme : +4 If the bubble corresponding to the answer is dark-
ened, 0 In all other cases

Q.41 Let the curve C be the mirror image of the parabola y 2 = 4x with
respect to the line x+y+4 = 0. If A and B are the points of intersection
of C with the line y = −5, then the distance between A and B is

Answer and Comments: 4. Call the line y = −5 as L.

y L*

C*
A*

x
O
x+y+4=0

B*
C
A B L

A straightforward approach would be to first identify C which is given

3
 to be the reflection of the parabola y 2 = 4x in the line x + y + 4 = 0.
Call this parabola C ∗ . Then C and C ∗ are reflections of each other and
hence are congruent. So C is also a parabola. To find its equation, we
would have to start with a typical point (x0 , y0) on C, find its reflection,
say (x∗0 , y0∗), in the line x + y + 4 = 0 and then put y0 ∗2 = 4x0 ∗2 . The
formula for the reflection of a point (x0 , y0) into a line ax + by + c = 0
is possible but rather complicated.
A better approach is to realise that under reflections all distances
are preserved. So, instead of taking the points of intersection of the
parabola C and the line L, we may as well take the points of intersection
of the parabola C ∗ and the reflection, say L∗ , of the line L into the line
x + y + 4 = 0. L∗ can be found almost by inspection. The line L cuts
the line x + y + 4 = 0 at the point P = (1, −5) at an angle 45 degrees.
So, L∗ must be the line through P making an angle of 45 degrees with
the line x + y + 4 = 0. Clearly this line is x = 1. This line L∗ happens
to lie along the latus rectum of the parabola C ∗ : y 2 = 4x. Hence its
intercept with the parabola has length 4. But even if we miss this,
the points of intersection, say A∗ and B ∗ of L∗ with C ∗ can be found
by merely solving the two equations y 2 = 4x and x = 1. They are
(1, ±2) and so the distance between them is 4. This is also the distance
between A and B since A∗ , B ∗ are the reflections of A, B.
This is an excellent problem which tests the ability of realising
how not to do a problem in the most straightforward way and, instead,
look for alternate ways. Once the idea of transforming the problem to
finding the intersections of C ∗ and L∗ strikes, the calculation required
is minimal.

Q.42 The minimum number of times a fair coin needs to be tossed, so that
the probability of getting at least two heads is at least 0.96, is

Answer and Comments: 8. Yet another problem where it is much

easier to find the answer by transforming the problem. In the present
case, the transformation is to consider the complementary probability,
say q of the given event. The complementary event here is that at
most one head appears. This falls into two mutually exclusive cases, no
heads and exactly one head. If there are n tosses, then their respective
1 1 1 1
probabilities are ( )n and n × × ( )n−1 = n( )n . Together, q =
2 2 2 2

4
 1
(n + 1)( )n . The problem asks for the least integral value of n for
2
which q < 0.04. This reduces to

25(n + 1) < 2n (1)

As n grows, 2n grows much more rapidly than n + 1. But the least such
n has to be found by trial and error. The L.H.S. is at least 50 and so
n has to be at least 6. By trial we get that 8 is the least integer for
which (1) holds.
This is also a good problem. But the technique of complemen-
tary probability is fairly common as compared with using properties of
reflections in the last problem.

Q.43 Let n be the number of ways in which 5 boys and 5 girls can stand in a
queue in such a way that all the girls stand consecutively in the queue.
Let m be the number of ways in which 5 boys and 5 girls can stand in
a queue in such a way that exactly four girls stand consecutively in the
m
queue. Then the value of is
n

Answer and Comments: 5. To find n think of the 5 girls lined

together as a single object. Then the number of ways to arrange this
object along with the 5 other objects (the boys) is 6!. But the 5 girls
can form a single object in 5! ways. So

n = 6!5! (1)

To find m, we consider a single object consisting of four girls in a

row. This object can be formed in 5 × 4! = 5! ways. Now we have
7 objects, this object with four girls, the remaining girl and the 5
boys. They can be arranged in 7! ways. Hence the total number of
arrangements in which at least 4 girls are together is 7!. But we have
to exclude those in which all 5 girls are together. That is already
counted as n. Moreover, each such arrangement has to be excluded
twice because the excluded girl can be the one at the head or at the
tail. [(G1 G2 G3 G4 G5 ) gets excluded twice, once as G1 (G2 G3 G4 G5 ), and
then again as (G1 G2 G3 G4 )G5 .] Hence

m = 7! × 5! − 2n = 7! × 5! − 2 × 6! × 5! (2)

5
 Hence
m 7! − 2 × 6!
= = 7−2 =5 (3)
n 6!

A simple but interesting problem. The ideas of transforming the

problem (by means of forming a single object of the girls) and of com-
plementary counting are present here too. But what makes the problem
tricky is that each unwanted arrangement appears twice. Those who
7! − 6!
miss this will get = 6 as the answer. And since this is also an
6!
integer between 0 and 9, there is no built in alert.
There is an alternate and a less tricky way to calculate m. As
before, form two groups of girls, one with 4 girls and the other with
just one girl. Now arrange the 5 boys on a row in any of the 5! ways.
In each such arrangement, there are six possible places where the two
groups of girls can go, one before the first boy, one after the last boy
and four in between two consecutive boys. We have to insert the two
groups of girls into these 6 places so that they do not go into the same
place. This can be done in 6 × 5 = 30 ways. This would give

m = 5! × 5! × 30 (4)

which is the same as (2).

Q.44 If the normals of the parabola y 2 = 4x drawn at the end points of its
latus rectum are tangents to the circle (x − 3)2 + (y + 2)2 = r 2 , then
the value of r 2 is

Answer and Comments: 2. Clearly r is the distance of the point

(3, −2) (the centre of the circle) from either of the two normals. The
problem is an unnecessarily clumsy way of asking the distance of this
point from these normals. Actually, either one of the two ends of the
latus rectum will suffice since we are given that both the normals at
both the ends are equidistant from (3, −2). We choose the end A =
dy
(1, 2) for this purpose. The slope of the tangent at this point is
dx
dy 2
where y 2 = 4x. This gives = = 1 at A. Hence the slope of the
dx y

6
 normal is −1. The equation of the normal at A is y − 1 = −(x − 1)
i.e.

x+y−3 =0 (1)

|3 − 2 − 3| √
The distance r of the point (3, −2) is √ = 2. Hence r 2 = 2.
1+1
An extremely straightforward problem. A disappointment on the
backdrop of the last three problems. Such problems hardly belong to
an advanced test. Some students may be unnecessarily tempted to use
formulas for equations of normals in terms of their slopes. The direct
approach above is far better.
(
[x], x≤2
Q.45 Let f : IR −→ IR be a function defined by f (x) =
0, x>2
2 xf (x2 )
Z
where [x] is the greatest integer less than or equal to x. If I = dx,
−1 2 + f (x + 1)
then the value of (4I − 1) is

Answer and Comments: 0. The function f changes its formula at

x = 2. Since the integrand involves f (x2 ) and f (x + 1), we have to
keep track of when x2 exceeds 2 and also when x + 1 exceeds 2 in the √
interval of integration, viz. [−1, 2]. The former happens at x = 2
and the latter at x = 1. Moreover [x] has discontinuities at 0 and −1.
So we have to split √ the interval
√ of integration into four subintervals,
[−1, 0], [0, 1], [1, 2] and [ 2, 2] and integrate over each one of them
and add the four integrals. We do so one-by-one. Call the integrand
as g(x) in all cases and denote the four integrals by I1 , I2 , I3 and I4
respectively.
On both the intervals [−1, 0] and [0, 1] 0 < x2 < 1 except possibly
at the end-points and so f (x2 ) = [x2 ] = 0. Hence the first two integrals
are 0. In the third and the fourth intervals, x + 1 exceeds 2 except at
the point 1 and hence f (x + 1) = 0. So in both the cases the integrand
1 √ x √
g(x) simplifies to xf (x2 ). On [1, 2] this becomes while on [ 2, 2],
2 Z √2
2 √
x x2 2 1
it vanishes. So the whole integral I is merely dx = = .
1 2 4 1 4
Therefore 4I − 1 = 0.

7
 A tedious problem which tests little more than some carefulness on
the part of the candidate. The integration part itself is trivial.
Q.46 A cylindrical container is to be made from certain solid material with
the following constraints : It has a fixed inner volume of V mm3 , has a 2
mm thick solid wall and is open at the top. The bottom of the container
is a solid circular disc of thickness 2 mm and is of radius equal to the
outer radius of the container. If the volume of the material used to
make the container is minimum when the inner radius of the container
V
is 10 mm, then the value of is
250π
Answer and Comments: 4. A typical problem about the minimisa-
tion of a function of one variable. The choice of this variable is ours.
We could take it to be the either the inner or the outer radius of the
base or even its inner or outer height. But since the last part of the
data is in terms of the inner radius of the base, that is a more natural
choice. So let r be the inner radius of the base and h the inner height
of the cylinder. Then its (inner) volume V is
V = πr 2 h (1)
As V is fixed, this equation allows us to express the inner height h in
terms of r. The volume, say W , of the material used is the difference
of the volumes of two coaxial cylinders, the inner one of radius r and
height h and the outer one of radius r + 2 and height h + 2 (and not
h + 4 as the container has no top). Therefore
W = π(r + 2)2 (h + 2) − πr 2 h
= π[(r 2 + 4r + 4)(h + 2) − r 2 h]
= π[h(4r + 4) + 2(r + 2)2 ]
V
= 2 (4r + 4) + 2π(r + 2)2 (2)
r
by (1).
As V is a constant, this expresses W as a function of the single variable
dW
r. We are given that it is minimum when r = 10. So must vanish
dr
at r = 10. By a direct calculation,
dW 4V 8V
= − 2 − 3 + 4π(r + 2) (3)
dr r r
8
 As this vanishes for r = 10, we get
4V 8V
+ = 48π (4)
100 1000
48V V
which simplifies to = 48π and gives = 4.
1000 250π
Although this is a problem about minimising a function of one
variable, it is of a different spirit. Here we are not asked to minimise
W . Rather, we are given where it is minimum. That simplifies the
work. If we were asked to find where W is minimum, we would have
to solve a cubic equation and then hunt for the minimum among the
critical points. That makes the problem reasonable. Also the fact that
the answer is an integer between 0 and 9 serves as an alert in case there
are any computational mistakes. This is a good feature of the problem.
(A problem of a similar spirit was asked in Jee 2013 Advanced Paper
1. An open box was to be formed by folding a rectangle after removing
squares of the same size from its four corners and we were given the side
of the square for which this volume is maximum. There is, of course,
nothing wrong in asking problems similar to those in the past years.
But one wishes that the repetition would not have occurred so soon.
Many but not all students must have studied the 2013 papers and those
who did would have an easier time in understanding the problem than
the others. It is all right to repeat an idea that was used last year, or
one that was used a decade ago. But a gap of two years can be unfair.)
Z x2 +π/6
Q.47 Let F (x) = 2 cos2 t dt for all x ∈ IR and f : [0, 1/2] −→ [0, ∞)
x
be a continuous function. For a ∈ [0, 1/2], if F ′ (a) + 2 is the area of
the region bounded by x = 0, y = 0, y = f (x) and x = a, then f (0) is

Answer and Comments: 3. The problem asks the value of the

function f (x) at x = 0. But this function is not given explicitly. In-
stead, we are given that f is defined on [0, 1/2] and takes only non-
negative values. We are also told something about the area bounded
by x = 0, y = 0, x = a and y = f (x). This is precisely the area under
the graph of y = f (x). So, the second piece of information means that
Z a
f (x) dx = F ′ (a) + 2 (1)
0

9
where the R.H.S. needs to be calculated from the first part of the
problem. (We have to assume that (1) holds for all a ∈ [0, 1/2]. This
is not clearly stated in the problem. But the problem cannot be solved
without such an assumption.)
Our interest is in f (x). We can get it from (1) by differentiating both
the sides (w.r.t. a) using the second form of the FTC (Fundamental
Theorem of Calculus). We then get,

f (a) = F ′′ (a) (2)

for all a ∈ [0, 1/2]. So our problem now reduces to find F ′′ (0).
To find F ′′ (x), we need to differentiate F (x) twice. The function
F (x) is defined by an integral in which both the upper and the lower
limits are functions of x. Therefore, by the generalised form of the
second fundamental theorem of calculus,
d 2 d
F ′ (x) = 2 cos2 (x2 + π/6) (x + π/6) − 2 cos2 x (x)
dx dx
= 4x cos2 (x2 + π/6) − 2 cos2 x (3)

We can get F ′′ (x) by differentiating the R.H.S. Instead of doing this

mechanically, let us observe that our interest is only in F ′′ (0). So
there is no need to consider those terms in the derivative which are
sure to vanish at 0. The derivative of cos2 x is one such term since it
will involve a factor sin x. As for the derivative of the first term, viz.
4x cos2 (x2 + π/6), when we apply the product rule, the factor 4x will
vanish at x = 0. So there is no need to take the derivative of the second
factor cos2 (x2 + π/6). The only term in the derivative that remains to
be considered is 4 cos2 (x2√+ π/6), evaluated at x = 0. This comes out
3
to be 3 since cos(π/6) = .
2
Problems based on the second form of the FTC are fairly common
in JEE. The present problem is more a test of a candidate’s ability
to analyse a problem correctly and focus on the essence so as to weed
out unnecessary work. In this sense it is a very good problem. Many
candidates will be tempted to simplify the integrand of F (x) to 1 +
cos 2t. But such a simplification has no role in the solution. In this
respect, the problem is a bit tricky too.

10
 Note that the second term (viz. 2) on the R.H.S. of (1) plays no
role in the solution since we are only dealing with the derivative of the
R.H.S. It would thus appear that in the data of the problem, the area of
the region could as well have been given as F ′ (a) + k for any constant
k. But then there would have been an inconsistency. From (3), we
know that F ′ (0) = −2. On the other hand, by putting a = 0 in (1), we
would have gotten F ′ (0) = −k. So, the data would be consistent only
if k = 2. A somewhat similar inconsistency in the data had occurred
in a problem in JEE 2011 Paper 1 and was commented upon. (See
Q.22 of that year’s
Z commentary.) It was given in the statement of the
x
problem that 6 f (t)dt = 3xf (x) − x3 for all x ≥ 1. Also f (1) was
1
given to be 2. Together we get a contradiction that 0 = 5 (by putting
x Z= 1). The lapse could have been corrected by changing the data to
x
6 f (t)dt = 3xf (x) − x3 − 5. This would not affect the rest of the
1
problem or its solution. Indeed that is probably what led to the lapse.
This year the paper-setters have been careful. The problem could
have been made a little more interesting by giving in the data that the
area of the region was F ′ (a) + k for some constant k and then asking
the candidates to find the value of f (0) + k. That would have forced
the candidates to determine k as 2 from (1) and (3). The answer to
the problem would have been 5 instead of 3, still a single digit number.

Q.48 The number of distinct solutions of the equation

5
cos2 2x + cos4 x + sin4 x + cos6 x + sin6 x = 2
4
in the interval [0, 2π] is

Answer and Comments: 8. Superficially, this is a problem of solving

a trigonometric equation. But there is no way to do this unless we first
simplify the expression, say E on the L.H.S. The key idea is to note
that all terms are expressible in terms of sin2 x and cos2 x. So the good
old identity sin2 x + cos2 x = 1 may be useful. Indeed, if we square this,
we get

sin4 x + cos4 x = 1 − 2 sin2 x cos2 x (1)

11
while if we take the cubes of both the sides, we get

sin6 x + cos6 x = 1 − 3 sin4 x cos2 x − 3 sin2 x cos4 x

= 1 − 3 sin2 x cos2 x(sin2 x + cos2 x)
= 1 − 3 sin2 x cos2 x (2)

With these substitutions, the expression E on the L.H.S. of the given

equation becomes
5
E= cos2 2x + 2 − 5 sin2 x cos2 x (3)
4
and so the equation simplifies to

cos2 2x − 4 sin2 x cos2 x = 0 (4)

or equivalently,

cos2 2x − sin2 2x = 0 (5)

and still further to

cos 4x = 0 (6)

which has eight solutions in the interval [0, 2π]. Specifically, the solu-
tions are of the form where 4x = π/2 and where 4x = 3π/2. But the
problem only asks for the number of solutions.
This is a fairly easy problem once the idea of taking the powers of
the basic identity for sin2 x + cos2 x strikes. See Comment No. 14 of
Chapter 7 for a problem where a similar trick is used.

12
 SECTION 2

This section contains TEN questions.

Each question has FOUR options (A), (B), (C) and (D). ONE OR MORE
THAN ONE of these four option (s) is (are) correct.
Marking scheme : +4 If only the bubble(s) corresponding to all the correct
option(s) is (are) darkened, 0 if none of the bubbles is darkened and −2 in
all other cases.

Q.49 Let y(x) be a solution of the differential equation (1 + ex )y ′ + yex = 1.

If y(0) = 2, then which of the following statements is (are) true?

(A) y(−4) = 0
(B) y(−2) = 0
(C) y(x) has a critical point in the interval (−1, 0)
(D) y(x) has no critical point in the interval (−1, 0).

Answer and Comments: (A), (C). This is an extremely standard

problem of solving a first order linear differential equation. Normally,
one would begin by recasting the equation in the standard form y ′ +
p(x)x = q(x) and find an integrating factor. In the present problem,
that is hardly necessary. The equation is exact as it stands, because
the L.H.S. is simply the derivative of y(1 + ex ). So, integrating both
the sides, the general solution is

y(1 + ex ) = c + x (1)

where c is an arbitrary constant. The initial condition y(0) = 2 gives

4 = c. So the function y is given by
4+x
y(x) = (2)
1 + ex
Clearly y(−4) = 0 while y(−2) 6= 0. For critical points, we need the
dy
derivative .
dx
dy (1 + ex ) − (4 + x)ex
= (3)
dx (ex + 1)2

13
 At a critical point the numerator must vanish. This gives the equation

3ex + xex − 1 = 0 (4)

We cannot solve this equation explicitly. Nor is it needed. All we are

asked is if it has a solution in the interval (−1, 0). For this we apply the
Intermediate Value Property. Call the L.H.S. as g(x). It is continuous
2
everywhere. Also g(−1) = − 1 < 0 since e > 2. On the other hand
e
g(0) = 2 > 0. So, by the Intermediate Value Property, g(x) has at
least one root in (−1, 0). Therefore y(x) has at least one critical point
in (−1, 0).
The problem is a combination of two unrelated parts. First, solving
an initial value problem and secondly testing if a given function has a
root in an interval. Both are very standard. There is little point in
asking such questions in an advanced test.

Q.50 Consider the family of all circles whose centers lie on the straight line
y = x. If this family of circles is represented by the differential equation
dy
P y ′′ +Qy ′ +1 = 0, where P, Q are functions of x, y and y ′ (here y ′ =
dx
′′ d2 y
and y = 2 ), then which of the following statements is (are) true?
dx
(A) P = y + x (B) P = y − x
(C) P + Q = 1 − x + y + y + (y ) (D) P − Q = x + y − y ′ − (y ′)2
′ ′ 2

Answer and Comments: (B), (C). This problem is about finding

the differential equation of a family of curves. In the present case, a
typical member of the given family is a circle of the form

(x − h)2 + (y − h)2 = r 2 (1)

where h and r are arbitrary constants. So this is a two parameter

family and the differential equation representing it will be of order 2.
To get it we differentiate (1) to get

2(x − h) + 2(y − h)y ′ = 0 (2)

14
 Solving this for h we get

x + yy ′
h= (3)
y′ + 1
One more differentiation yields

(y ′ + 1)(1 + yy ′′ + y ′2 ) − (x + yy ′)y ′′
=0 (4)
(y ′ + 1)2

Hence the differential equation representing the given family of circle

is

(y ′ + 1)(1 + yy ′′ + y ′2 ) − (x + yy ′)y ′′ = 0 (5)

When expanded, this may contain terms involving the product y ′y ′′ . In

the statement of the problem, the coefficients P and Q are not allowed
to contain y ′′ . So we have to recast this equation collecting all the
terms in y ′′ together. That gives

(y − x)y ′′ + (1 + y ′ + y ′2 )y ′ + 1 = 0 (6)

(Luckily, the terms involving y ′ y ′′ have cancelled.) Comparing this with

the form given in the statement of the problem,

P = y−x (7)
and Q = 1 + y ′ + y ′2 (8)

Hence (B) and (C) are correct.

A very mechanical problem. The solution essentially ends at (5).
The remainder is a useless addendum. It might have served some pur-
pose if there were any terms involving y ′y ′′ because then the candidate
would have to think whether to include them as multiples of y ′ or of
y ′′. But since these terms get cancelled, this fine thinking is not tested
anyway.

Q.51 Let g : IR −→ IR be a differentiable

 x
function with g(0) = 0, g ′(0) = 0
 g(x), x 6= 0
and g ′(1) 6= 0. Let f (x) = |x| and h(x) = e|x| for all
0, x=0


15
x ∈ IR . Let (f ◦ h)(x) denote f (h(x)) and (h ◦ f )(x) denote h(f (x)).
Then which of the following is (are) true?
(A) f is differentiable at x = 0 (B) h is differentiable at x = 0
(C) f ◦ h is differentiable at x = 0 (D) h ◦ f is differentiable at x = 0

Answer and Comments: (A), (D). This is a question about the

differentiability of two functions and also of their composites. The
composite of two differentiable functions (when defined) is always dif-
ferentiable. But sometimes the composite may be differentiable even
when one of the functions fails to be so. An extreme counterexample
is when one of the functions is constant. Then the composite is also
a constant and differentiable regardless of the other function. So such
cases have to be handled carefully.
Let us begin with the differentiability of the function h(x) = e|x|
at x = 0. This is the composite of the absolute value function and the
exponential function. The former is not differentiable at 0. But that is
no reason to hastily declare that the composite e|x| is not differentiable
at 0 because as we just saw, the composite of a differentiable function
with a non-differentiable one can be differentiable sometimes. But we
can put the non-differentiability of |x| to use as follows. We want to
e|x| − e0
consider whether lim exists. For x 6= 0 we rewrite this ratio
x→0 x
as

e|x| − e0 e|x| − 1 |x|

= × (1)
x |x| x

As x → 0, the first factor tends to 1, this being the right handed

derivative of ex at x = 0. But the second factor is 1 for x > 0 and −1
for x < 0. So the product tends to 1 as x → 0+ and to −1 as x → 0− .
Therefore the product of the two ratios tends to 1 as x → 0+ and to −1
as x → 0− . Hence h is not differentiable at 0. (There is a slicker way
to see this by observing that |x| = ln(h(x)). The logarithm function is
differentiable everywhere in its domain. So, if h were differentiable at
0, then this composite |x| would be differentiable at 0, a contradiction.)

16
 Next, we turn to the differentiability of f at 0. Note that


 g(x), x > 0
f (x) = 0, x=0 (2)
−g(x), x < 0



Therefore, f+′ (0), i.e. the right handed derivative of f at 0, will equal
′
g+ (0) which is 0 since it is given that g ′ (0) = 0. Similarly, f−′ (0) will
′
equal −g− (0) which is also 0. Hence f is differentiable at 0. (It would
not be so if g ′ (0) were non-zero.)
We now tackle the differentiability of the composite functions f ◦ h
and h ◦ f at 0. Note that h(x) is always positive and so by (2)

(f ◦ h)(x) = f (h(x)) = g(h(x)) (3)

for all x. Further h(0) = 1. So, by a reasoning similar to that in (A),

(f ◦ h)′+ (0) would equal g+′
(h(0)), i.e. g+′
(1) = g ′(1) as g is given to
be differentiable at 1, while (f ◦ h)′− (0) would equal −g− ′
(1) = −g ′ (1).
′
Since g (1) 6= 0, these two numbers are unequal and so f ◦ h is not
differentiable at 0.
Finally, we consider the differentiability of h ◦ f at 0. By definition,

(h ◦ f )(x) = e|f (x)| (4)

for all x. Also, by (2), |f (x)| = |g(x)| for all x. Hence e|f (x)| is the
same as e|g(x)| . The exponential function is differentiable everywhere.
So if we can show that |g(x)| is differentiable at 0, then it would follow
that e|g(x)| , and hence h ◦ f is differentiable. We are given that g is
differentiable at 0. This does not by itself imply that |g| is differentiable
at x, as one sees from the fact that x is differentiable at 0 but |x| is
not. But now we are also given that g(0) = 0 and g ′ (0) = 0. So, for
x 6= 0 we can write

|g(x)| − |g(0)| g(x)

= ± (5)
x x

with the + sign holding for x > 0 and the minus sign holding for
g(x)
x < 0. But as we are given that → 0 as x → 0, it follows
x

17
 g(x)
that → 0 as x → 0, no matter from which side. This proves
x
that |g(x)| is differentiable at 0, and, as noted earlier that h ◦ f is
differentiable at 0.
This is a very good conceptual question which demands very fine
thinking and little computations from the candidate. Unfortunately
such questions are unsuitable as multiple choice questions because there
is no way to test if the candidate has understood the subtlety of the
problem, even if he has ticked the correct answer. For example, a
candidate who rules out (B) on the superficial ground that |x| is not
differentiable at 0 cannot be distinguished from one who really proves
it as above.
Negative credit is very unfair for such questions. There is no
reason why a candidate who has done the fine thinking for, say (B),
should be penalised for failure to do a similar thinking for (D). Most
candidates who are adept at the JEE strategy will stay away from such
questions and in the time saved, safely bag 12 points by doing some
absolutely routine questions. So, such questions serve little purpose in
the selection when they are clubbed together with a large number of
mediocre questions.
π π π
  
Q.52 Let f (x) = sin sin sin x and g(x) = sin x for all x ∈ IR. Let
6 2 2
(f ◦ g)(x) denote f (g(x)) and (g ◦ f )(x) denote g(f (x)). Then which
of the following is (are) true?
1 1 1 1
   
(A) Range of f is − , (B) Range of f ◦ g is − ,
2 2 2 2
f (x) π
(C) lim = (D) There is an x ∈ IR such that (g ◦ f )(x) = 1
x→0 g(x) 6

Answer and Comments: (A), (B), (C). Yet another question in-
volving the composites of two functions. In both the questions the
paper-setters have been careful enough to indicate which of the two
possible interpretations of a composite they have in mind. Although
the interpretation given here is more standard, candidates who follow
the other interpretation should not suffer solely for that reason.
Now, coming to the question itself, all parts except (C) are based on
the images of various intervals under the sine function. and the fact that

18
 the range of the composite is the image of the range of the first function.
π π π
Since sin x maps IR onto [−1, 1], sin x maps it onto [− , ]. The sine
2 2 2
π π
function maps this interval onto [−1, 1] and so sin( sin x)) has range
6 2
π π
[− , ]. Under the sine function this interval goes to [−1/2, 1/2] which
6 6
π π
proves (A). In (B), we first have to find the range of g which is [− , ].
2 2
Under the sine function, this interval has the same image viz. [−1, 1]
as the image of the sine function over the entire IR. So f ◦ g has the
same range as f , which is [−1/2, 1/2]. In (D), the range of g ◦ f is the
image under g of the range of the f which we already know to be the
interval [−1/2, 1/2]. So (D) will be true if and only if there is some
2
x ∈ [−1/2, 1/2] for which g(x) = 1, i.e. sin x = . As the sine function
π
is strictly increasing on the interval [−1/2, 1/2], the answer depends
upon which of the two numbers sin(1/2) and 2/π is bigger. To do this
without calculators, we use the inequality sin x < x for all x > 0. In
particular, sin 1/2 < 1/2. But 1/2 < 2/π since π < 4. So (D) is false.
Part (C) is of a totally different spirit than the others. The limit
in question can be calculated rather mechanically using the L’Hôpital’s
π
rule. But a better way out is to put u = g(x) = sin x and v = π6 sin u.
2
Then u → 0 and v → 0 as x → 0. Therefore,
f (x) sin( π6 sin u)
=
g(x) u
π π
sin( 6 sin u) 6
sin u
= π ×
6
sin u u
sin v π sin u
= × × (1)
v 6 u
π
Both the first and the last factors tend to 1 and so the limit is.
6
A simple but highly repetitious problem based on the range of the
sine function. Part (D) requires the approximate value of π. Part (C)
is a useless addendum.
−→ −→ −→
Q.53 Let △P QR be √ a triangle. Let ~a = QR, ~b = RP and ~c = P Q. If
~
|~a| = 12, |b| = 4 3 and ~b · ~c = 24, then which of the following is (are)
true?

19
 |~c|2 |~c|2
(A) − |~a| = 12 (B) + |~a| = 30
2 √ 2
(C) |~a × ~b + ~c × ~a| = 48 3 (D) ~a · ~b = −72

Answer and Comments: (A), (C), (D). A routine problem about

computations involving vectors, their lengths and various products.
The lengths of the two sides of the triangle are given. If we were given
the dot product of the vectors representing them, viz. ~a and ~b, we
would also know the angle between them and then we would know the
triangle completely. But we are not given that. We are given |~a|, |~b|,
but not ~a · ~b. Instead, we are given ~b · ~c. We are also given |~b|. So, if
we could get |~b + ~c|, we would know |~c| and that would also determine
the triangle completely.
The basic idea is that
since the vectors ~a, ~b and ~c P
are the sides of the triangle
c θ/2
(directed appropriately), their θ/2 b
vector sum is 0. That is, 30
Q 30
~a + ~b + ~c = ~0 (1) 6 a O 6 R

Therefore
~b + ~c = −~a (2)

Taking lengths of both the sides

|~b|2 + |~c|2 + 2(~b · ~c) = |~a|2 = 144 (3)

This gives
√ √ √
|~c| =
144 − 48 − 48 = 48 = 4 3 (4)
√
Having known |~a| = 12 and |~c| = 4 3 we immediately dispose of (A)
and (B).
For the other two statements, we need to know more about the
triangle. Since |~b| = |~c| the triangle P QR is isosceles with P Q = P R.

20
Let the angle between these sides be θ. Then θ is the angle between
the vectors −~b and ~c. So,

−~b · ~c −24 1
cos θ = = =− (5)
~
|b| |~c| 48 2
which gives θ = 120◦ . As the triangle is isosceles, the remaining two
angles are 30◦ each. Since the angle between the vectors ~b and −~a is
30◦ the angle between the vectors ~a and ~b is 150◦ . So we get
√
√ 3
~a · ~b = |~a| |~b| cos 150 = −12 × 4 3 ×
◦
= −72 (6)
2
which shows that (D) is true. Finlly, for (C), we note that by (1)
~b = −~a − ~c. Since ~a × ~a = ~0, and the angle between ~a and ~c is also
150◦, we have
|~a × ~b + ~c × ~a| = ~a × (−~a − ~c) + ~c × ~a
= |2~c × ~a|
= 2|~c| × |~a| × sin 150◦
√ √
= 4 3 × 12 = 48 3 (7)
Hence (C) is true too.
Our solution is purely geometric. The second half of it could have
been shortened a little by resolving the vectors along a suitable pair
of mutually orthogonal unit vectors. We take the midpoint of the side
QR as the origin O and two unit vectors ~i and ~j along OR and OP
1
respectively. Since 6 OP Q = 6 OP R = 60◦ we have OP = P Q =
√ 2
2 3. This allows us to express all the three vectors ~a, ~b and ~c as linear
combinations of the mutually orthogonal vectors ~i and ~j as
~a = 12~i (8)
√
~b = −6~i + 2 3 ~j (9)
√
and ~c = −6~i − 2 3 ~j (10)
By a direct calculation,
|~a × ~b + ~c × ~a| = |~a × (~b − ~c)|
√ √
= |12~i × 4 3 ~j| = 48 3 (11)

21
 which shows that (C) is correct. Also
√
~a · ~b = 12~i · (−6~i + 2 3 ~j) = −72 + 0 = −72 (12)

so that (D) is also true.

In fact, the entire solution could have been algebraic. We take ~i to
be a unit vector along ~a and ~j to be a unit vector perpendicular to ~i.
Then, ~a = 12~i and ~b = α~i + β~j where√α, β are some scalars. Then by
(1), ~c = (−12 − α)~i − β~j. Since ~b = 4 3, we have
√
α2 + β 2 = (4 3)2 = 48 (13)

Similarly, ~b · ~c = 24 gives

−α(α + 12) − β 2 = 24 (14)

Solving this system simultaneously, we get

√
α = −6, β = 2 3 (15)

This way we get (9) and (10) more efficiently. Having known all three
vectors in terms of ~i and ~j, all the four statements can be tested one
by one. Thus we see that in the present problem the purely algebraic
solution is fastest. But the gain is not so significant as the problem
itself is simple.
The problem is simple, once the essential idea, viz. Equation (1)
strikes. Unfortunately, there is too much numerical work. Even a single
mistake is costly. So this problem is more a test of speed and numerical
accuracy than reasoning.

Q.54 Let X and Y be two arbitrary, 3×3, non-zero skew-symmetric matrices

and Z be an arbitrary 3 × 3 non-zero symmetric matrix. Then which
of the following matrices is (are) skew-symmetric?

(A) Y 3 Z 4 − Z 4 Y 3 (B) X 44 + Y 44
(C) X 4 Z 3 − Z 3 X 4 (D) X 23 + Y 23

Answer and Comments: (C), (D). Parts (B) and (D) are based
on some simple properties of skew-symmetric and symmetric matrices.

22
Specifically, we need : (i) the sum of two symmetric (skew-symmetric)
matrices is symmetric (respectively, skew-symmetric) and (ii) all pow-
ers of symmetric matrices are symmetric (iii) all even powers of skew-
symmetric matrices are symmetric while all odd powers of them are
skew-symmetric. With these properties, we have X 44 , Y 44 and hence
their sum symmetric and similarly X 23 + Y 23 skew-symmetric. Note
the analogy of these properties with the properties about the signs of
real numbers. For example, the sum of two positive (negative) real
numbers is also positive (negative). Also the even powers of a negative
number are positive while the odd powers are negative.
However, in (A) and (C), we are dealing with the products of
symmetric and a skew-symmetric matrices. Here the analogy with the
real numbers breaks down. For example, the product of two negative
numbers is positive. But little can be said about the product of two
skew-symmetric matrices unless they commute with each other. So,
(A) and (C) have be handled by directly taking transposes and using
the elementary properties of transposes, viz. the anticommutativity
and the self-reciprocity. (In simpler terms, this means (AB)T = B T AT
and (AT )T = A, for all A, B.)
With these rules in mind, we have

(Y 3 Z 4 − Z 4 Y 3 )T = (Y 3 Z 4 )T − (Z 4 Y 3 )T
= (Z 4 )T (Y 3 )T − (Y 3 )T (Z 4 )T
= −Z 4 Y 3 + Y 3 Z 4 (1)

because Y 3 is skew-symmetric and Z 4 is symmetric. So, Y 3 Z 4 − Z 4 Y 3

is symmetric. If it were to be skew-symmetric too, it would have to
vanish, which means that Y 3 and Z 4 must commute with each other.
As this is not given, we discard (A).
By an analogous computation,

(X 4 Z 3 − Z 3 X 4 )T = (Z 3 )T (X 4 )T − (X 4 )T (Z 3 )T = Z 3 X 4 − X 4 Z 3

since Z 3 and X 4 are both symmetric. This shows that X 4 Z 3 − Z 3 X 4

is skew-symmetric.
Because of their limited scope, questions about matrices tend to be
repetitious. The present one is a little unusual. The calculations are

23
 simple once the essential idea is understood. However, in (A), there is a
possibility that the matrix may be skew-symmetric as well, as pointed
out above. It is given in the question that the matrices Y, Z are non-
zero. But that does not preclude the possibility that Z 4 commutes
with Y 3 . For example, Z could be the identity matrix. It would have
been better if the question had asked to identify those matrices that
are necessarily skew-symmetric.
Q.55 Which of the following values of α satisfy the equation
(1 + α)2 (1 + 2α)2 (1 + 3α)2
(2 + α)2 (2 + 2α)2 (2 + 3α)2 = −648α ?
(3 + α)2 (3 + 2α)2 (3 + 3α)2

(A) −4 (B) 9 (C) −9 (D) 4

Answers and Comments: (B), (C). The given determinant, say D

is a polynomial in α. It would be horrendous to compute D by direct
expansion. But if we subtract R2 from R3 and then R1 from R2 we get
1 + 2α + α2 1 + 4α + 4α2 1 + 6α + 9α2
D= 3 + 2α 3 + 4α 3 + 6α (1)
5 + 2α 5 + 4α 5 + 6α
Next, we subtract the middle row from the other two to get
α2 − 2 4α2 − 2 9α2 − 2
D = 3 + 2α 3 + 4α 3 + 6α (2)
2 2 2
We now add the last row to the first to get
α2 4α2 9α2
D = 3 + 2α 3 + 4α 3 + 6α (3)
2 2 2
To simplify D further, we subtract the first column from the other two
to get
α2 3α2 8α2
D = 3 + 2α 2α 4α (4)
2 0 0

24
Now D is simple enough to be expanded by its last row. We get
D = 2(12α3 − 16α3) = −8α3 (5)
So the given equation, viz. D = −648α reduces to α3 = 81α whose
roots are 0, 9 and −9.
In the past there used to be many interesting problems based on
evaluation of determinants by various manipulations. But they used to
be full length questions, allowing five to ten minutes. In the present set-
up, there are hardly 3 minutes. A saving feature is that the candidates
do not have to explain their work. In the present problem, it is unlikely
that a candidate would get the correct answer by mere substitution. So
this is a good, classic type problem on determinants.
There is an alternate, albeit a trickier solution to the problem.
We first expand all the squares and then write the determinant as the
product of two determinants, viz.
1 + 2α + α2 1 + 4α + 4α2 1 + 6α + 9α2
D = 4 + 4α + α2 4 + 8α + 4α2 4 + 12α + 9α2
9 + 6α + α2 9 + 12α + 4α2 9 + 18α + 9α2
1 1 1 1 1 1
= 4 2 1 2α 4α 6α
9 3 1 α2 4α2 9α2
1 1 1 1 1 1
3
= 2α 4 2 1 1 2 3 (6)
9 3 1 1 4 9
Both the determinants can be evaluated directly or by subtracting the
last column from the others for the first and subtracting the first row
from the remaining ones for the second. Their values are 2 and −2.
(It is not an accident that their values are the negatives of each other,
because if we interchange the first and the last column of the first de-
terminant and then take its transpose, we get the second determinant.)
So, we finally get D = −8α3 which is the same as (5). The rest of the
solution is the same.
More generally one can consider a determinant D of the form
(a + x)2 (a + y)2 (a + z)2
D = D(x, y, z, a, b, c) = (b + x)2 (b + y)2 (b + z)2 (7)
(c + x)2 (c + y)2 (c + z)2

25
where x, y, z, a, b, c are variables. When fully expanded this is a homo-
geneous polynomial in these six variable in which each term has total
degree 6. By expanding the squares and taking steps similar to those
above, one can show that

a2 a 1 1 1 1
D(x, y, z, a, b, c) = 2 b2 b 1 x y z (8)
c2 c 1 x2 y 2 z 2

The second determinant is a Vandermonde determinant (see Exercise

(3.26)) and hence has value (y − x)(z − y)(z − x). If we interchange
the first and the last columns of the first determinant and take its
transpose, that is also a Vandermonde determinant. As a result, we
get

D = −2(a − b)(b − c)(c − a)(x − y)(y − z)(z − x) (9)

There is also an easier way to see this if we observe from (7) that D
vanishes if, say a = b. Hence (a − b) is a factor of D. And so are
(b − c), (c − a), (x − y), (y − z) and (z − x). So the product of these
six factors also divides D(x, y, z, a, b, c). But this product is already a
polynomial of total degree 6 in x, y, z, z, b, c. Hence we must have

D(x, y, z, a, b, c) = k(a − b)(b − c)(c − a)(x − y)(y − z)(z − x) (10)

for some constant k. The value of k can be determined by giving some

special, simple values to a, b, c, x, y, z, e.g. a = x = 0, b = y = 1 and
c = z = −1. (A similar technique is also possible for evaluating the
Vandermonde determinant.)
There is a certain formal resemblance between the determinant in
(7) and the determinant

cos(A − P ) cos(A − Q) cos(A − R)

D(P, Q, R, A, B, C) = cos(B − P ) cos(B − Q) cos(B − R) (11)
cos(C − P ) cos(C − Q) cos(C − R)

A 1994 JEE problem (see Comment No. 22 of Chapter 2) asked to

show that this determinant vanishes for all A, B, C, P, Q, R. Again, a
direct expansion is simply ruled out. But a solution is possible using

26
 elementary row operations and trigonometric identities. The best solu-
tion, however, is to expand all the entries of the determinant and then
show that it equals the product
cos A sin A 0 cos P cos Q cos R
cos B sin B 0 sin P sin Q sin R (12)
cos C sin C 0 0 0 0

Q.56 In IR3 , consider the planes P1 : y = 0 and P2 : x + z = 1. Let P3 be a

plane, different from P1 and P2 which passes through the intersection
of P1 and P2 . If the distance of the point (0, 1, 0) from P3 is 1 and the
distance of a point (α, β, γ) from P3 is 2, then which of the following
relations is (are) true ?
(A) 2α + β + 2γ + 2 = 0 (B) 2α − β + 2γ + 4 = 0
(C) 2α + β − 2γ − 10 = 0 (D) 2α − β + 2γ − 8 = 0

Answer and Comments: (B), (D). This is a problem about a

parametrised family of planes. Let L be the line of intersection of
the planes P1 and P2 . Write the equations of P1 and P2 in the form
E1 = 0 and E2 = 0 where E1 , E2 are linear expressions in x, y, z.
Then the equation of every plane passing through L is of the form
λE1 + µE2 = 0 for some values of the parameters λ and µ. We can
dispense with one of the parameters, say µ and consider an equation of
the form λE1 + E2 = 0. This will represent all possible planes through
L for various values of λ, except the plane P1 . (For the plane P1 ,
we need λ = 1 and µ = 0.) Similarly, the equation E1 + µE2 = 0
will represent all planes through L except P2 . In the present case we
are given that P3 is different from both P1 and P2 . So we are free to
take either approach. (See Comment No. 13 of Chapter 9) for more
examples of this technique.)
We take E1 as y, E2 as x + z − 1. Then the equation of P3 is of the
form
x + z − 1 + λy = 0 (1)
for some value of λ. To determine it, we use the condition that the
distance of the point (0, 1, 0) is 1. This gives
0+λ+0−1
√ =1 (2)
1 + λ2 + 1

27
which simlifies to

(λ − 1)2 = λ2 + 2 (3)
1
and determines λ as − . So the equation of the plane P3 is x + z −
2
1
1 − y = 0 or, equivalently,
2
2x − y + 2z − 2 = 0 (4)

We are further given that the distance of the point (α, β, γ) from P3 is
1. This means
2α + β + 2γ − 2
√ =2 (5)
4+1+4
This means

2α − β + 2γ − 2 = ±6 (6)

The two signs correspond to (D) and (B) respectively.

A routine problem once the idea of a parametrised family strikes.
A discerning student will observe that normally, (2) would reduce to a
quadratic in λ, which is consistent with the fact in the family of planes
containing the line L, there are two planes whose distance from the
point (0, 1, 0) is 1. In the present case, one of these two planes is P1
itself and is discarded by the data. Since we took the equation of P3
in the form λE1 + E2 = 0, which represents all planes containing L
except P1 , (2) degenerated into a linear equation in λ. Had we taken
the equation of P3 in the form

y + µ(x + z − 1) = 0 (7)

where µ is a parameter, then instead of (2) we would have gotten

1−µ
√ =1 (8)
1 + 2µ2

which would reduce to the quadratic µ2 + 2µ = 0, having 0 and −2 as

its roots. The root µ = 0 gives the plane P1 and has to be discarded.

28
 The second root, viz. −2 will give the equation of the plane P3 as
2x − y + 2z − 2 = 0 which is the same as (4). So the answer does
not change if we take the other parametrisation of the family, but the
work involved does slightly. So, we were rather lucky to start with the
equation λE1 +E2 = 0 rather than with E1 +µE2 = 0. A sharp student
will, however, not leave this choice to luck. He will observe that P1 is
already at a distance 1 from (0, 1, 0) and since it is to be excluded, it
is safe (and numerically easier) to take the equation of P3 as (1) rather
than (7). But in the present problem the advantage gained is minor.
Most candidates would anyway prefer to start with (1) rather than (7)
because it is simpler. If the advantage were substantial and the easier
option not so tempting, then this would have been a good problem
which rewards the sharp candidates.

Q.57 In IR3 , let L be a straight line passing through the origin. Suppose
that all the points on L are at a constant distance from the two planes
P1 : x + 2y − z + 1 = 0 and P2 : 2x − y + z − 1 = 0. Let M be the
locus of the feet of the perpendicular drawn from the points on L to
the plane P1 . Which of the following points lie(s) on M ?
(A) (0, −5/6, −2/3) (B) (−1/6, −1/3, 1/6)
(C) (−5/6, 0, 1/6) (D) (−1/3, 0, 2/3)

Answer and Comments: (A), (B). The first part of the data simply
means that the line L is parallel to both P1 and P2 and hence to the
line of their intersection. From the two equations of the planes, viz.

x + 2y − z = −1 (1)
and 2x − y + z = 1 (2)

we see that the direction numbers of their line of intersection are

1, −3, −5. (This is a standard result. For those who don’t know it,
these are the components of the vector u × v where u = i + 2j − k and
v = 2i − j + k are normals to the planes P1 , P2 respectively.) Since L
passes through the origin, its parametric equations are

x = t, y = −3t, z = −5t (3)

where t is a parameter.

29
 As L is parallel to the plane P1 , the feet of the perpendiculars from
the points on L to the plane P1 will form a line M parallel to L. To
determine M, it suffices to know any one point P0 = (x0 , y0 , z0 ) on it.
We take it to be the foot of the perpendicular from the point (0, 0, 0)
(which is given to lie on L) to P1 . To determine this point we need
three equations in x0 , y0 , z0 . one of them comes of course, from the
equation of the plane P1 . That is

x0 + 2y0 − z0 = −1 (4)
−→
The other two equations come from the fact that the vector OP0 =
x0 i + y0 j + z0 k is perpendicular to P1 and hence parallel to the normal
vector u = i + 2j − k. This gives

x0 = r, y0 = 2r, z0 = −r (5)

for some real number r. Substituting this into (4), we get r = −1/6.
Hence P0 = (−1/6, −1/3, 1/6). Therefore the locus M is the line
x + 1/6 y + 1/3 z − 1/6
= = =λ (6)
1 −3 −5
where λ is any real number. λ = 0 gives (−1/6, −1/3, 1/6) as a point on
M. For the point (0, −5/6, −2/3) to lie on M, we must have λ = 1/6
from x + 1/6 = λ. This value also satisfies the other two equations
in (6). But for the other two given points, we have y = 0, whence
λ = −1/9. But that would make x = −1/9 − 1/6 = −5/18. So the
points in (C) and (D) do not lie on M.
A fairly simple problem once the idea strikes that L is parallel to
the line of intersection of the two given planes. In the conventional
examination the solution would end with (6), i.e. finding the equation
of the locus M. Asking which of the given four points satisfy (6) is
sheer arithmetic and prone to numerical errors. Also it is anybody’s
guess what is the reason for making the candidates do this work four
times. The only answer is that the constraints on the paper-setters
stipulate that every MCQ must have four choices. Thank God it was
not 10.
Q.58 Let P and Q be distinct points on the parabola y 2 = 2x such that
a circle with P Q as a diameter passes through the vertex O of the

30
parabola. If√P lies in the first quadrant and the area of the triangle
△OP Q is 3 2, then which of the following are the coordinates of P ?
√ √ √ √
(A) (4, 2 2) (B) (9, 3 2) (C) (1/4, 1/ 2) (D) (1, 2)

Answer and Comments: (A), (D). There is a minor anomaly in the

wording of the problem. Given any two distinct points P and Q, there
is only one circle having P Q as a diameter. So instead of saying ‘a
circle with P Q as a diameter’, the wording should have been ‘the circle
with P Q as a diameter. This might be just a lapse on the part of the
paper-setters. But it might confuse a discerning student, maybe only
for a few seconds. But in a severely competitive test even a few seconds
are precious.
Now, coming to the problem itself, take the points P and Q in the
parametric form as
t2
P = ( 1 , t1 ) (1)
2
t2
and Q = ( 2 , t2 ) (2)
2
2 2
Then the slopes of OP and OQ are and respectively. As O lies
t1 t2
on the circle with P Q as a diameter, we have OP ⊥ OQ which means
4
= −1 and hence
t1 t2
t1 t2 = −4 (3)
√
We are further given that the area of the triangle △OP Q is 3 2. Since
OP Q is right angled at O, we get
√ 1
3 2 = × OP × OQ
2
1 |t1 | q 2 |t2 | q 2
= × t1 + 4 × t2 + 4
2 2 2
1q 2
= (t1 + 4)(t22 + 4) (4)
2
using (3). Squaring both the sides
(t21 + 4)(t22 + 4) = 72 (5)

31
 Expanding and using (3) again, this gives

t21 + t22 = 10 (6)

Hence (t1 − t2 )2 = 10 − 2t1 t2 = 18. This gives

√
t1 − t2 = ±3 2 (7)

We are given that P is in the first quadrant. This makes t1 positive

and hence t2 negative by (3) again. So the positive sign must hold.
Now that we know both t1 − t2 and t1 t2 , we can find t1 by solving the
equation
4 √
t1 + =3 2 (8)
t1
which becomes a quadratic
√
t21 − 3 2t1 + 4 = 0 (9)
√ √
3 2± 2 √ √
The roots are i.e. 2 2 and 2. The corresponding point
2
t21 √ √
P = ( , t1 ) then is (4, 2 2) and (1, 2).
2
A routine problem, once the idea of taking the points in their
parametric forms strikes.

SECTION 3

This section contains TWO questions. Each question contains two columns
Column I and Column II. Column I has four entries (A), (B), (C) and
(D). Column II has five entries (P), (Q), (R), (S) and (T). Match the en-
tries in Column I with the entries in Column II. One or more entries in
Column I may match with one or more entries in Column II.
Marking scheme: For each entry in Column I, 2 points if fully correct, 0
points if not attempted and −1 points in all other cases.

32
Q.59 Column I Column II
(A) In IR2 , if the magnitude of√the projection
√ vector (P) 1
of the vector√ αi + βj on 3 i + j is 3 and if
α = 2 + 3β, then possible value(s) of |α| is (Q) 2
(are) (R) 3
(B) Let a and b be(real numbers such that the func- (S) 4
−3ax2 − 2, x < 1
tion f (x) = is differen-
bx + a2 , x≥1 (T) 5
tiable for all x ∈ IR, then the possible value(s)
of a are

(C) Let ω 6= 1 be a complex cube root of unity. If

(3 − 3ω + 2ω 2 )4n+3 + (2 + 3ω − 3ω 2 )4n+3 + (−3 +
2ω + 3ω 2)4n+3 = 0, the possible value(s) of n is
(are)

(D) Let the harmonic mean of two positive real num-

bers a and b be 4. If q is a positive real number
such that a, 5, q, b is an arithmetic progression,
then the value(s) of |q − a| is (are)

Answers and Comments: (A;P,Q), (B;P,Q), (C; P,Q,S,T), (D; Q,T).

The four parts in Column I are quite independent of each other. But
so is the marking for them. This is a merciful departure from the past
where credit would be given only if all parts of Column I were answered
correctly. But then one wonders what was the point of grouping these
questions in such an elaborate manner especially when all of them have
numerical answers. Why not ask four separate questions, each having
five possible answers? This is again a silly constraint on the paper-
setters regarding the format of the question paper.
Now, coming to the questions in Column
√ I, in (A), a unit vector in
√ 3 1
the direction of the vector 3 i + j is i + j. Hence the projection
√ 2 2
3 1 √
of αi + βj has magnitude | α + β|. Equating this with 3 and
2 2
squaring gives
√
( 3 α + β)2 = 12 (1)

33
 α−2
We are further given that β = √ . Substituting this into (1) we get
3

(4α − 2)2 = 36 (2)

which gives 4α − 2 = ±6. So the possible values of α are α = 2 and

α = −1. Hence |α| = 1 or 2.
In (B), there are two unknowns, a and b. To determine them
we need two equations in them. One is provided by the fact that
every differentiable function is continuous. Continuity of f (x) at x = 1
implies

−3a − 2 = b + a2 (3)

As for differentiability of f (x), the right handed and the left handed
derivatives at x = 1 are −6a and b respectively. So we get

b = −6a (4)

Eliminating b from these two equations gives a quadratic in a, viz.

a2 − 3a + 2 = 0 (5)

whose possible solutions are 1 and 2.

In (C), we are dealing with a sum of the (4n + 3)-th powers of three
quadratic expressions in ω, specifically,

E1 = 3 − 3ω + 2ω 2 (6)
E2 = 2 + 3ω − 3ω 2 (7)
and E3 = −3 + 2ω + 3ω 2 (8)

We observe that the same coefficients, viz. 3, −3 and 2 appear in these

polynomials. This suggests that the polynomials must be related to
each other in some simple way. (If no such relationship exists, then the
problem would be extremely hard.) Once this idea strikes, finding the
actual relationship is easy. Since ω 3 = 1, ω 4 = ω, ω 5 = ω 2 we see that

E1 = ωE3 and E2 = ω 2E3 (9)

34
So the given equation simplifies to

(ω 4n+3 + (ω 2 )4n+3 + 1)E34n+3 = 0 (10)

We also have ω 2 +ω +1 = 0. (This is a well known result and follows by

factorising ω 3 − 1 and noting that ω 6= 1.) So E3 = −3 + 2ω − 3 − 3ω =
−6 − ω which is non-zero. Hence E34n+3 is also non-zero. So, only the
first factor of the L.H.S. of (10) vanishes. Further, since ω 3 = 1, ω 3n
and (ω 2 )3n also equal 1 each. So the equation simplifies to

ω n + ω 2n + 1 = 0 (11)

Powers of ω repeat in a cycle of 3. When n is a multiple of 3, the

equation becomes 1 + 1 + 1 = 0 which is impossible. However, for other
values of n, it reduces to ω + ω 2 + 1 = 0 (for n = 1, 4, 7, . . .) or to
ω 2 + ω + 1 = 0 (for n = 2, 5, 8, . . .) both of which are true. Hence n
must not be divisible by 3. In Column II, the possible values of n are
those other than 3.
Finally, in (D), the first part gives the equation

ab = 2(a + b) (12)

We are also given that a, 5, q, b are in an A.P. This gives two more
equations, viz.

q + a = 10 (13)
and b + 5 = 2q (14)

Eliminating q,

b = 2(10 − a) − 5 = 15 − 2a (15)

Putting this into (12) we get

a(15 − 2a) = 2a + 30 − 4a (16)

which simplifies to a quadratic in a, viz.

2a2 − 17a + 30 = 0 (17)

35
 17 ± 7
whose roots are , i.e. 6 and 5/2. Corresponding values of q are 4
4
and 15/2. So |q − a| = 2 in the first case and |q − a| = 5 in the second.
(C) is the only interesting problem in the bunch. The others reduce
to solving systems of two equations in two unknowns. There is nothing
very great either in formulating these equations or in solving them.
These parts hardly belong to an advanced test.

Q.60 Column I Column II

(A) In a triangle △XY Z, let a, b and c be the (P) 1
lengths of the sides opposite to the angles X, Y
and Z respectively. If 2(a2 − b2 ) = c2 and (Q) 2
sin(X − Y )
λ = , then possible values of n for (R) 3
sin Z
which cos(nπλ) = 0 is (are) (S) 5
(B) In a triangle △XY Z, let a, b and c be the (T) 6
lengths of the sides opposite to the angles X, Y
and Z, respectively. If 1 + cos 2X − cos 2Y =
a
2 sin X sin Y , then possible value(s) of is (are)
b
√ √
(C) In IR2 , let 3 i + j, i + 3 j and βi + (1 − β)j be
the position vectors of X, Y and Z with respect
to the origin O, respectively. If the distance of Z
from the bisector of the acute angle of OX with
3
OY is √ , then possible value(s) of |β| is (are)
2
(D) Suppose that F (α) denotes the area of the region
bounded by x = 0, x = 2, y 2 = 4x and y =
|αx − 1| + |αx − 2| + αx, where α ∈ {0, 1}. Then
8√
the value(s) of F (α) + 2 when α = 0 and
3
α = 1 is (are)

Answers and Comments: (A; P,R,S), (B; P), (C; P,Q), (D; S,T).
In this question, items (A) and (B) have a common setting. Since
a, b, c are proportional to the sines of their opposite angles, the equation

36
2(a2 − b2 ) = c2 in (A) translates into

2(sin2 X − sin2 Y ) = sin2 Z (1)

By a well-known identity the L.H.S. factorises and we get

2 sin(X − Y ) sin(X + Y ) = sin2 Z (2)

which further simplifies to

sin(X − Y ) 1
= (3)
sin Z 2
since in any triangle △XY Z, sin(X + Y ) = sin(π − Z) = sin Z. So we
1
get λ = and the equation to be solved reduces to
2
cos(nπ/2) = 0 (4)

which is possible only when n is an odd integer. The odd integers in

Column II are 1, 3 and 5.
In (B), the condition given is in terms of the angles and we have
a sin X
to find the ratio which equals . Let us first recast the given
b sin Y
condition in terms of sin X and sin Y .

1 + cos 2X − 2 cos 2Y = 2 cos2 X − 2(1 − 2 sin2 Y )

= (2 − 2 sin2 X) − 2 + 4 sin2 Y
= 4 sin2 Y − 2 sin2 X (5)

Hence the condition given translates as

2 sin2 Y − sin2 X = sin X sin Y (6)

sin X
Dividing by sin2 Y and calling as λ (which is the ratio we are
sin Y
interested in), we get a quadratic in λ, viz.

λ2 + λ − 2 = 0 (7)

whose roots are λ = 1 and −2. As λ, being the ratio of two sides cannot
be negative, we get λ = 1.

37
 In (C), the vectors are involved only superficially. The first part
of√the data√is just another way of saying that the points X, Y, Z are
( 3, 1), (1, 3) and (β, 1 − β) respectively. Even without drawing a
diagram it is obvious that the points X and Y are symmetrically located
w.r.t. the line y = x. Also OX and OY are inclined at angles 30◦ and
60◦ respectively. So, the line y = x is the acute angle bisector 6 XOY .
3
We are given that the distance of Z from this line y − x = 0 is √ .
2
This implies
|(1 − β) − β| 3
√ =√ (8)
1+1 2
This means 2β − 1 = ±3 and hence β equals 2 or −1. So |β| equals 2
or 1.
Finally, (D) consists of two separate problems, one for α = 0 and
the other for α = 1. The first one is easier because in this case y = 3
and so, F (0) is the area bounded by x = 0, x = 2, y 2 = 4x and y = 3.
It is the shaded area OABCO in the figure below.
y

C B y=3
(0, 3) A

x
O (2, 0)

By a direct calculation,
√
Z 2 √ 4 2 8 2
F (0) = 3 − 2 x dx = 3x − x3/2 = 6− (9)
0 3 0 3
√
8 2
So, F (0) + = 6 which tallies with (T) in Column II.
3
On the other hand, when α = 1,
y = |x − 1| + |x − 2| + x (10)

38
For x ∈ [0, 2], |x − 2| = 2 − x and so (10) simplifies a little to

y = |x − 1| + 2 (11)

whose graph for 0 ≤ x ≤ 2 is the union of two straight line segments,

one from C = (0, 3) to A = (1, 2) and the other from A = (1, 2) to
B = (2, 3). So this time F (1) is the sum of the two shaded areas
OACO and ADBA shown in the figure below.

C B y=3
(0, 3) D
A

x
O (2, 0)

Again, by a direct calculation,

Z 1 √ Z 2 √
F (1) = 3 − x − 2 x dx + x + 1 − 2 x dx (12)
0 1
√
8 2
By a routine integration which we skip, F (1) comes out to be 5 − .
√ 3
8 2
So F (1) + = 5.
3
Parts (A) to (C) are straightforward, but too elementary to be
asked in an advanced test. The wording of (D) is clumsy and many
candidates might not understand the problem and might have skipped
it. They are the clever ones, because those who do struggle success-
fully to realise that the problem involves the calculation of two (in
reality three) unrelated areas will pay a heavy price in terms of time
and the strong possibility of some numerical error. If the idea was
merely to give a problem about identifying and finding the area of a
plane region, the first half of the problem (where α = 0) would have

39
 served the purpose quite well. Adding one more part which hardly
tests anything new (except the ability to draw the graph of the func-
tion y = |x − 1| + |x − 2| + x) is nothing short of torture. It is such
sadistic problems which make success at the JEE a matter of adopt-
ing a clever strategy, whose prime rule is to simply stay away from a
problem which is clumsily worded and utilise the time saved on routine
problems requiring mediocre intelligence.

PAPER 2

Contents

Section - 1 (One Integer Value Correct Type) 40

Section - 2 (One or More than One Correct Choice Type) 54

Section - 3 (Paragraph Type) 71

SECTION 1
One Integer Value Correct Type

This section contains eight questions. The answer to each question is a

SINGLE DIGIT INTEGER ranging from 0 to 9, both inclusive.
Marking scheme : +4 If the bubble corresponding to the answer is dark-
ened, 0 In all other cases

Q.41 Suppose that all the terms of an arithmetic progression (A.P.) are nat-
ural numbers. If the ratio of the sum of the first seven terms to the
sum of the first eleven terms is 6 : 11 and the seventh term lies between
130 and 140, then the common difference of this A.P. is

Answer and Comments: 9. An A.P. is determined by two numbers,

viz. its initial term, say a, and its common difference, often denoted
by d. To determine them, we need two equations in a and d. The
problem gives only one. In such cases some additional restrictions on

40
 the variables such as that they are integers or that they lie in some
specific intervals is needed to determine the unknowns. The present
problem is of this type. (See Exercise (4.24) for an example where
there are fewer equations than unknowns and still a unique solution.)
With the notations just introduced, the sum of the first n terms
(often denoted by Sn ) is

n(n − 1)
Sn = na + d(1 + 2 + . . . + (n − 1)) = na + d (1)
2
S7 6
We are given that S11
= 11
. Because of (1) this means

7a + 21d 6
= (2)
11a + 55d 11
which yields 7a + 21d = 6a + 30d and hence

a = 9d (3)

This single equation cannot determine d (or a) uniquely. We now use

the second condition. The seventh term is a + 6d. So we are given

130 < a + 6d < 140 (4)

which by (2) becomes

130 < 15d < 140 (5)

But d is given to be integer. The only multiple of 15 between 130 and

140 is 135. Setting 135 = 15d gives d = 9.
A simple, but thought provoking problem. The computations
needed are minimal and can be done quickly. Such questions are ideal
for multiple choice tests because the answer is not likely to come by
guessing.

Q.42 The coefficient of x9 in the expansion of (1 + x)(1 + x2 )(1 + x3 ) . . . (1 +

x100 ) is

Answer and Comments: 8. The problem is superficially algebraic.

But in reality it is combinatorial. When the product, say P is fully

41
expanded it will be a sum of 2100 terms, each being a product of 100
terms of the form u1 u2 . . . u100 where each ui has two possibilities , either
1 or xi , for i = 1, 2, . . . , 100. We can combine these two possibilities
by saying that ui = xni where the exponent ni is either 0 or i. So,
u1 u2 . . . u100 will equal un1 +n2 +...+n100 . Clearly only the positive terms
in the exponent matter. So each occurrence of x9 will result from a
sequence of distinct positive integers in ascending order adding to 9.
Once this is understood, the problem reduces to finding the num-
ber of ways to express 9 as a sum of distinct integers (arranged in an
ascending order), ranging from 1 to 100. As 9 is a small number, these
possibilities can be counted by classifying according to the number of
terms in the summation. There is only one way to express 9 as a sum
of a single integer, viz. 9 = 9. To find the number of ways to write 9
as a sum of two distinct integers in ascending order, we go systemati-
cally as 1 + 8, 2 + 7, 3 + 6 and 4 + 5. If there are three summands, say
9 = a + b + c, with 0 < a < b < c, note that a + b is at least 3. So
c can be only 6, 5 or 4 since with c = 3, a < b < c implies a + b can
be at most 3. For c = 4, we have a + b = 5. But since a < b < c this
can happen only a = 2, b = 3. For c = 5, a + b will equal 4 only for
a = 1, b = 3 and finally for c = 6, the only possibility is a = 1, b = 2.
There is no need to go further because the sum of any four distinct
positive integers will be at least 10. So 9 can be expressed as a sum of
distinct positive integers in an ascending order in 1 + 4 + 3 = 8 ways.
The problem could have been formulated as a combinatorial
problem. Suppose there are 100 boxes, numbered 1 to 100. Then the
problem asks for the number of ways to put nine identical balls into
these boxes, so that each box is either empty or contains as many balls
as its number.
In the present case, we have translated the algebraic problem
into a combinatorial one. It could have as well been done the other
way. The work done is essentially the same with either approach. But
there are situations where converting a combinatorial problem into an
algebraic one pays off. Consider, for example, the problem of counting
selections with repetitions. So, let an,k be the number of ways to choose
k objects from n types of objects with repetitions allowed freely. This
is equivalent to placing k identical balls into n distinct boxes, there

42
being no restriction as to how many balls can go into!any box. There
k+n−1
is a tricky way to show that an,k equals by thinking of
k
each such placement as an arrangement of k balls and n − 1 inter-box
partitions. (See Comment No. 5 of Chapter 1.)
The algebraic version of this problem can be constructed as follows.
Each of the n boxes can hold at least 0 and at most k balls. So an,k is
the coefficient of xk in the expansion of (1 + x + x2 + . . . + xk )(1 + x +
. . . + xk ) . . . (1 + x + . . . + xk ), there being n factors in all. Equivalently,
an,k is the coefficient of xk in (1 + x + x2 + . . . + xk )n . This conversion
does not help much by itself. But, instead of taking the polynomial
1 + x + x2 + . . . + xk , let us take the entire infinite series 1 + x + x2 +
. . . + xk + xk+1 + xk+2 + . . .. This may appear useless because the extra
terms we are adding cannot contribute to any selection. (There is no
way to write k as a sum of non-negative integers if one or more terms
is greater than k.) But now the advantage is that the infinite series
1
1 + x+ x2 + . . .+ xk + . . . can be identified as the geometric series .
1−x
Doing thisfor each of the factors, we see that an,k is the coefficient of xk
n
1

in . This too, is not of much use by itself. But if rewrite this
1−x
expression as (1−x)−n and expand it using the binomial theorem where
the exponent is the negative integer −n, we see that the coefficient of
  (−n)(−n − 1) . . . (−n − k + 1)
xk is (−1)k −n k
= (−1)k which comes
k!
n(n + 1) . . . (n + k − 1)
out to be which is nothing but the binomial
k!
!
n+k−1
coefficient .
k
What makes this solution possible is the rich machinery of algebra,
including power series. The combinatorial solution was elementary but
rather tricky. The situation is analogous to the relationship between
pure geometry and coordinate geometry. Pure geometry solutions are
elegant but sometimes tricky. When coordinates are introduced the
powerful machinery of algebra makes the problem more amenable, if
somewhat dull. (See, for example, the second proof of the concurrency
of the three altitudes of a triangle, given in Comment No. 3 of Chapter
8.)

43
 Even without the powerful machinery of power series, the algebraic
recasting of a combinatorial problem in terms of a suitable polynomial
is sometimes useful when the answer is less demanding than finding the
coefficient of a specific power of x. Consider, for example, the problem
of placing n identical balls into, say, r distinct boxes with the restriction
that each box can contain at most k balls. If we denote this number by
an , then as we just saw, an is the coefficient of xn in the polynomial,
say f (x) = (1 + x + x2 + . . . + xk )r and there is no easy of finding
it except for some select values of n. For example, we can say that
ark = 1 and an = 0 for n > rk because f (x) is a polynomial of degree
rk and leading coefficient 1. But this is something obvious by common
sense anyway. Similarly, we can tell a0 = 1 and a1 = r equally easily
with or without the help of the polynomial f (x).
But, suppose that our problem is not to calculate an for a particular
value of n, but to find the entire sum, say S1 = a0 +a1 +a2 +. . .+ark . We
∞
X
could have written S1 ostensibly as an infinite sum S1 = an because
n=0
all except finitely many terms of this series vanish. Since f (x) = a0 +
a1 x + a2 x2 + . . . ark xrk , we have S1 = f (1). From the factorisation of
f (x) as (1 + x + x2 + . . . + xk )r we immediately get S1 = (k + 1)r .
Combinatorially, S1 is the number of all possible ways of putting any
numbers of identical balls into r distinct boxes so that each box contains
at most k balls. Here, too, a direct combinatorial argument is easy
because, for each of the r boxes, there are k + 1 possibilities depending
upon how many balls go into it. So again, this example does not quite
bring home the power of algebraic codification.
To do so, consider S2 = a0 + a2 + +a4 + . . . + a2m + . . ., i.e. the sum
of the coefficients of all even degree terms in f (x). Combinatorially, S2
is the number of all possible ways to put any numbers of balls into the
boxes as before with the additional restriction that the total number
of balls is even. (It is not required that only even numbers of balls
go into the individual boxes. The restriction of evenness is only on
the total number of balls.) This time, the combinatorial count is not
as immediate as for S1 . But the algebraic one is simple. If we add
f (1) and f (−1), the even powers add up while the odd ones cancel
f (1) + f (−1)
out. In other words, S2 = a0 + a2 + a4 + . . . = . This is
2

44
 true for any polynomial. In our example, f (1) = S1 = (k + 1)r while
f (−1) = (1 − 1 + 1 − . . . + (−1)k+1 )r which equals 0 or 1 according as
(k + 1)r (k + 1)r + 1
k is odd or even. Hence S2 = or depending upon
2 2
whether k is odd or even.
The essence of the calculations of S1 and S2 was, respectively, that
for every non-negative integer n, 1n = 1 for all n while 1n +(−1)n equals
2 or 0 depending upon whether n is a multiple of 2 or not. Note that
1 and −1 are the square roots of 1. In the solution to Part (C) of Q.59
of Paper 1, we proved that 1n + ω n + (ω 2 )n equals 3 or 0 depending
upon whether n is a multiple of 3 or not. So, for any polynomial
f (x) = a0 +a1 x+a2 x2 +. . ., the sum, say S3 = a0 +a3 +a6 +. . .+a3m +. . .
f (1) + f (ω) + f (ω 2 )
would equal .
3
The picture is now quite clear. Even if we may not be able to identify
the coefficients an individually for all n, we can calculate the sums of the
coefficients whose suffixes are in an A.P. by adding the values of f (x) at
the d complex d-th roots of unity where d is the common difference of
this progression. Sometimes such sums have some appeal and provide
an unexpected solution to a problem where a direct combinatorial count
may be laborious. Consider, for example, the problem of finding the
number of 6-digit numbers whose digits add to a number of the form
4p + 1. This is equivalent to counting the number of ways to put
identical balls into 6 boxes the first of which can hold 1 to 9 balls and
the remaining 0 to 9 balls each, so that the total number of balls is of
the form 4p + 1. Then the number we want is the sum a1 + a5 + a9 + . . .
for the polynomial f (x) = (x + x2 + . . . + x9 )(1 + x + x2 + . . . + x9 )5 .
Equivalently, this is the sum b0 + b4 + b8 + . . . for the polynomial g(x) =
b0 +b1 x+b2 x2 +. . . = (1+x+x2 +. . .+x8 )(1+x+x2 +. . .+x9 )5 . By our
g(1) + g(i) + g(−1) + g(−i)
work, the count is . By direct substitutions
4
of the powers of these complex numbers into the factors of g(x), we have
g(1) = 900000, g(−1) = 0, g(i) = (1+i)5 = −4−4i and g(−i) = −4+4i.
9000000 − 4 − 4
So the desired number is = 249998.
4
x2 y 2
Q.43 Suppose that the foci of the ellipse + = 1 are (f1 , 0) and (f2 , 0)
9 5
where f1 > 0 and f2 < 0. Let P1 and P2 be two parabolas with a com-

45
mon vertex at (0, 0) and with foci at (f1 , 0) and (2f2 , 0), respectively.
Let T1 be a tangent to P1 which passes through (2f2 , 0) and T2 be a
tangent to P2 which passes through (f1 , 0). If m1 is!the slope of T1 and
1
m2 is the slope of T2 , then the value of + m22 is
m21

Answer and Comments: 4. The problem is about the tangents to

the two parabolas. The role of the ellipse is only to specifys the foci of
5 2
the two parabolas. The eccentricity e of the given ellipse is 1 − =
9 3
and so the foci are at (±3e, 0) i.e. f1 = 2 and f2 = −2.
Since both the parabolas P1 and P2 have their vertices at (0, 0), and
their foci are at (2, 0) and (−4, 0) respectively, their equations are
y 2 = 8x (1)
y 2 = −16x (2)
respectively. We are given that the line y = m1 (x + 4) touches the
parabola y 2 = 8x. So it intersects the parabola in two coincident
points. Therefore, the quadratic m1 (x + 4)2 − 8x = 0 has discriminant
0. On simplification, this gives (8m21 − 8)2 = 64m41 and hence
1
m21 = (3)
2
(Incidentally, this shows that there are two lines through (−4, 0) that
touch the parabola P1 . The problem does not specify which of the two
is to be chosen. But the answer does not depend on the choice since it
involves only m21 .)
By an entirely analogous computation, the slope m2 of T2 satisfies
m22 = 2 (4)
1
Hence the expression + m22 equals 4.
m21
Those who know the equation of a tangent to a parabola in terms of
its slope can shorten the work by observing that a tangent to y 2 = 8x
having slope m1 has its equation of the form
2
y = m1 x + (5)
m1

46
 and get (3) from the fact that this tangent passes through (−4, 0). Sim-
ilarly an alternate derivation of (4) is possible. We have intentionally
given a derivation from first principles, based on the concept of a tan-
gent as a line which intersects a curve in two coincident points, to show
that even if you do not remember a whole lot of formulas, things can
often be salvaged if you go by the basic principles.
A very straightforward problem about identifying the tangents to
a parabola passing through a given point. Perhaps the paper-setters
realised that the problem is too straightforward and hence gave it a
twist by first making the candidates identify the points through which
the tangents are to pass. In the old days, instead of making the problem
numerical, it would have been asked to show that if two parabolas have
their vertices at the centre of an ellipse and their foci at the foci of that
ellipse then the tangents to either of them passing through the focus
of the other are mutually orthogonal. Such geometric results expressed
solely in words, have their own beauty.

Q.44 Let m and n be two! positive integers greater than 1. If

cos(αn )
e −e e m
lim m
= − then the value of is
α→0 α 2 n

Answer and Comments: 2. As α → 0 both the numerator and

the denominator tend to 0. So the limit, say L, in the question if
0
of the indeterminate form . It is, therefore, tempting to apply the
0
L’Hôpital’s rule. But if m > 1 this rule will have to be applied again
and again. Let us temporarily assume that m = 1 and allow n to take
n
non-integral values as well. Define the function f (α) = ecos(α ) − e.
Then L’Hôpital’s rule implies that

f ′ (α)
L = lim (1)
α→0 1
provided this limit exists.
Now, by a direct computation, for α 6= 0 we have,
n
f ′ (α) = − sin(αn )ecos(α ) nαn−1 (2)

47
It is tempting to think that this is 0 at α = 0 because of the first factor,
viz., sin(αn ). This is valid for n > 0. But we have to be careful about
the last factor, viz. αn−1 . If n < 1, then this factor tends to ∞ as
n
α → 0+ . The middle factor ecos(α ) creates no problem as it tends to 1
when α → 0+ as long as n > 0. Nor does n which is a fixed number.
To see the result of the battle between the first and the last factor
when 0 < n < 1, let us rewrite (2) as

sin(αn ) cos(αn ) 2n−1

f ′ (α) = − ne α (3)
αn
Now the first factor tends to 1 as α → 0. So the fate of the limit
depends on the last term, viz. α2n−1 . If n > 1/2, then it will make
f ′ (α) tend to 0. But if 0 < n < 1/2, then it will tend to ∞ as α → 0+ .
We conclude that if m = 1, then it is only for n = 1/2 that the
e
limit in question will exist. When it does, its value will be − . In the
2
problem, of course, n is given to be an integer. But the argument above
m
suggests that if at all the ratio has some fixed value, it is probably
n
2.
An intelligent and a smart gambler will leave the solution at this
point. But a scrupulous person can make the argument valid as follows.
m
Let k = . Call αm as β. Then n = m/k and therefore, αn = β 1/k .
n
Moreover β → 0 as α → 0. So, by a change of variable,
1/k )
ecos(β −e
L = lim (4)
β→0 β
We are now back to the old game with α replaced by β and n replaced by
1/k. By the argument above, L will equal a finite non-zero value (which
will be necessarily −e/2 ) only when 1/k = 1/2. This is equivalent to
saying that k = 2.
A mature person will approach the problem by considering the
comparable orders of magnitude. For example, when u → 0, sin u is of
sin u
the same order as u because tends to a finite non-zero limit as
u
cos u − 1
u → 0. Similarly, cos u − 1 is of the order of u2 because tends
u2

48
 1
to the non-zero limit − . Similarly, eu − 1 is of the order of u. Now
2
factor out e from the numerator of the given expression and consider
n
′ecos(α )−1 − 1
L = lim (5)
α→0 αm
Then clearly, L = eL′ provided L′ exists. By what we just said, the
numerator is of the order of cos(αn ) − 1, which, in turn, is of the order
of α2n . The denominator, on the other hand, is of the order of αm . So
the ratio will tend to a finite non-zero limit only when the numerator
and the denominator have the same orders and that happens precisely
when m = 2n.
If need arises, this reasoning based on comparable orders of mag-
nitude can be made precise by dividing and multiplying the expression
on the R.H.S. of (5) by cos(αn ) − 1 and then again by α2n . But that is
essentially a clerical work.
This is a very good, thought oriented problem. But since it is
a multiple choice question where no reason has to be given, a smart
candidate may get the correct answer by the sneaky path, i.e. by
assuming m = 1.
12 + 9x2
!
Z 1
9x+3 tan−1 x
Q.45 If α = e dx where tan−1 x takes only princi-
0 x2 + 1
3π
pal values, then the value of loge (1 + α) − is
4
Answer and Comments: 9. This is plainly a question about evalu-
ating a definite integral. Since the answer is to be an integer between
0 and 9, the last part operates as a hint to the value, viz. 1 + α is of
the form ek+3π/4 for some integer k. Although this does not help in
evaluating the integral, it serves to alert against numerical errors. In
the conventional form the question would have merely asked the value
of the integral.
The integral itself is easy if we use the substitution
u = 9x + 3 tan−1 x (1)
which yields
du 3 12 + 9x2
=9+ = (2)
dx 1 + x2 x2 + 1

49
 Hence we have
Z 9+3π/4
α = eu du
0
9+3π/4
= e −1 (3)

So ln(1 + α) = 9 − 3π/4.
Too simple a problem once the correct substitution strikes. And
there is little choice about the correct substitution. The paper-setters
2
!
9 x 12 + 9x
at least could have given the integrand as (e ) (etan x )3 ,
−1
2
x +1
when in many other questions (for example, Q. 43 above) they have
given the data in an unnecessarily twisted form.

Q.46 Let f : IR −→ IR be a continuous, odd function whichZ vanishes ex-

1 x
actly at one point and f (1) = . Suppose F (x) = f (t) dt for
Z x2 −1

all x ∈ [−1, 2] and G(x) = t|f (f (t))| dt for all x ∈ [−1, 2]. If
−1
F (x) 1
lim = , then the value of f (1/2) is
x→1 G(x) 14

Answer and Analysis: 7. The problem is a combination of L’Hôpital’s

rule and the second Fundamental Theorem of Calculus (FTC). To ap-
F (x)
ply the former to the limit lim we must first ensure that it is
x→1 G(x)
0
of the form. As both F and G are continuous (being functions de-
0
fined by integrals), this amounts to verifying that F (1) and G(1) both
vanish. This is not given. In fact, all we know is

Z 1
F (1) = f (t) dt (1)
−1
Z 1
and G(1) = = t|f (f (t))| dt (2)
−1

Since f is given to be an odd function, the integral on the R.H.S. of

(1) vanishes. For the same to happen for (2), we must ensure that
x|f (f (x))| is an odd function of x. For this we note that f (f (x)) is an

50
odd function of x, being the composite of two odd functions. Hence
its absolute value |f (f (x))| is an even function. Finally, x is an odd
function of x and so the product x|f (f (x))| is an odd function of x,
being the product of an even function and an odd function. So, the
integral on the R.H.S. of (2) vanishes as well.
We are now justified in applying the L’Hôpital’s rule. By the second
FTC, F ′ (x) = f (x) and G′ (x) = x|f (f (x))| for all x ∈ [−1/2, 2]. Hence
F (x) F ′ (x) F ′ (1) f (1)
lim = lim ′ = ′ =
x→0 G(x) x→1 G (x) G (1) 1 × |f (f (1))|
1/2
=
|f (f (1))|
1/2
= (3)
|f (1/2)|
1
We are given that this limit is . So, we now get
14
14
|f (1/2)| = =7 (4)
2
We are still not quite through, because the problem asks for f (1/2)
and we only know that f (1/2) = ±7. A sneaky reasoning would be
that since the correct answer has to be an integer from 0 to 9, −7 is
excluded. But, for an honest answer, we must show that f (1/2) > 0.
For this we apply the Intermediate Value Property (IVP). f (x) is given
to be continuous everywhere. Also f (1) = 1/2 > 0. So, if f (1/2) were
negative, then by the IVP, f would vanish at least once in the interval
(1/2, 1). But, being an odd function, f already vanishes at 0. As we
are given that f vanishes at exactly one point, we get a contradiction.
So, at long last we have fully justified our answer, viz. f (1/2) = 7.
To arrive at it legitimately, we had to give a lot of reasoning. First to
ensure the applicability of the L’Hôpital’s rule. Many students apply
this rule without ensuring this. Secondly, many students might not
bother to justify that f (1/2) > 0. And there is no way to tell if
they have really done the reasoning. So, this question rewards the
sloppy students and punishes the scrupulous ones. In the conventional
examinations, the two could be differentiated because reasonings would
have to be given.

51
 The essential idea of the question is the combination of L’Hôpital’s
rule and the FTC. Properties of odd functions have also been tested
along the way. But being an MCQ, all this fine creativity on the part
of the paper-setters is wasted. (In fact, the possibility that a student
might pick 7 as the correct answer simply because of the number 14
appearing in the problem cannot be entirely discounted. A safe gamble,
as there is no negative credit for a wrong answer.)
Q.47 Suppose that p~, ~q and ~r are three non-coplanar vectors in IR3 . Let the
components of a vector ~s along ~p, ~q and ~r be 4, 3 and 5, respectively.
If the components of this vector ~s along (−~p + ~q + ~r), (~p − q~ + ~r) and
(−~p − ~q + ~r) are x, y and z respectively, then 2x + y + z is

Answer and Comments: 9.. An extremely straightforward problem

about resolving a vector along three linearly independent vectors. We
are given
~s = 4~p + 3~q + 5~r (1)
But we are also given that
~s = x(−~p + ~q + ~r) + y(~p − ~q + ~r) + z(−~p − q~ + ~r) (2)
= (−x + y − z)~p + (x − y − z)~q + (x + y + z)~r (3)
As p~, ~q, ~r are non-coplanar and hence linearly independent, the coeffi-
cients in the two resolutions must match. This gives a system of three
equations in the three unknowns, viz.
x + y − z = 4, x − y − z = 3, and x + y + z = 5 (4)
Solving this system is child’s play and gives x = 4, y = 9/2, z = −7/2.
A straight substitution gives 2x + y + z = 8 + 1 = 9.
The word ‘component’ in the statement of the problem is likely
to confuse some candidates. Sometimes, by a component of a vector ~s
along a vector ~p one understands, ‘component along the direction of p~’,
i.e. the component along a unit vector in the direction of p~. This usage
is common in expressions like ‘the tangential and normal components
of acceleration’ used in physics. With this meaning, instead of (1) we
would have
4~p 3~q 5~r
~s = + + (5)
||~p|| ||~q|| ||~r||

52
 Similarly, Equation (2) will have to be replaced by a more complicated
equation. This would make the problem not only more complicated,
but impossible to solve since the lengths of the vectors p~, ~q and ~r and
of their various linear combinations are not given.
But the paper-setters should not be blamed for this. The interpreta-
tion of ‘component’ as they have intended is fairly standard. With the
alternate interpretation, the components are generally taken to be vec-
tors along the respective directions. Nevertheless, the JEE authorities
have decided to award four points to every candidate for this question.
For those who interpret the word ‘component’ as we have, the
question is a cakewalk. It is difficult to believe that such a simple
question is asked in an advanced test. It may be argued that this is
an atonement for some other difficult questions. But when these two
types of questions are combined, the selection takes place mostly on
the basis of such mediocre questions. That defeats the very purpose of
selection. And, for this, the paper-setters do deserve some criticism.
√
! !
kπ kπ
Q.48 For any integer k, let αk = cos + i sin where i = −1.
7 7
12
|αk+1 − αk |
P
k=1
Then the value of the expression 3
is
|α4k−1 − α4k−2 |
P
k=1

Answer and Comments: 4. In the Argand diagram, the complex

numbers lie at the vertices of a regular polygon with 14 sides inscribed
in the unit circle {z : |z| = 1}. (This happens because all the given
complex numbers are 14-th roots of unity.) For every k the vertices
αk+1 and αk are consecutive. So, |αk+1 − αk | equals the length, say
d of each side of the heptagon. So the numerator of the expression is
12d. In the denominator also we are summing the distances between
consecutive vertices, except that in the denominator there are only 3
12d
terms. Hence the denominator is 3d. So the given ratio equals = 4.
3d
A good problem based on the geometric representation of complex
numbers. Note that the number 7 has little role in the problem. It
could have been any positive integer r. Then the vertices would repeat

53
 in a cycle of length 2r and even if the summation extended to more
than 2r terms, the calculations would be the same.
The problem is reminiscent of a 1994 JEE problem which was a sort
of a converse to the present problem. There we were given a regular
n-gon with vertices A1 , A − 2, . . . , An and the relationship between the
lengths of its sides and some of its diagonals. Specifically, it was given
that
1 1 1
= + (1)
A1 A2 A1 A3 A1 A4
and the problem asked to determine n. This can be reduced to solving
a trigonometric equation. (Ironically, the answer to that problem was
also 7. See Comment No. 1 of Chapter 10.) However, the regularity
of the n-gon permits us to conclude from (1) a similar relationship for
any four consecutive vertices of the polygon, that is
1 1 1
= + (2)
Ai Ai+1 Ai Ai+2 Ai Ai+3
for every i, with the understanding that An+1 = A1 etc. This fact, along
with an application of what is called Ptolemy’s theorem about cyclic
quadrilaterals, gave an unexpected solution to the problem. (See again
Comment No. 2 of Chapter 10.) A solution using complex numbers is
also possible by taking the vertices to lie at the complex 2n-th roots of
unity. (See Comment No. 3 of the same chapter.) It is possible that
this solution has spurred the present problem. Of course, the present
problem is much simpler than the 1994 problem. How one wishes to
revive the golden days when such nice problems could be asked as full
length questions! All we have today is their weak imitations.

SECTION 2

This section contains EIGHT questions. Each questions has FOUR op-
tions (A), (B), (C) and (D). ONE OR MORE THAN ONE of these four
option(s) is(are) correct.
Marking scheme : +4 If fully correct, 0 if not attempted and −2 In all
other cases.

54
Q.49 Let f, g : [−1, 2] −→ IR be continuous functions which are twice differ-
entiable in the interval (−1, 2). Let the values of f and g at the points
−1, 0 and 2 be given by the following table:

x = −1 x=0 x=2
f(x) 3 6 0
g(x) 0 1 -1

In each of the intervals (−1, 0) and (0, 2) the function (f − 3g)′′ never
vanishes. Then the correct statement(s) is (are)

(A) f ′ (x) − 3g ′ (x) = 0 has exactly three solutions in (−1, 0) ∪ (0, 2)

(B) f ′ (x) − 3g ′ (x) = 0 has exactly one solution in (−1, 0)
(C) f ′ (x) − 3g ′ (x) = 0 has exactly one solution in (0, 2)
(D) f ′ (x) − 3g ′ (x) = 0 has exactly two solutions in (−1, 0) and exactly
two solutions in (0, 2)

Answer and Comments: (B), (C). All the four statements are about
the zeros of h′ (x) where h(x) is defined by

h(x) = f (x) − 3g(x) (1)

From the table, we get

h(−1) = 3, h(0) = 3 and h(2) = 3 (2)

Rolle’s theorem implies that h′ (x) vanishes at least once in each of the
intervals (0, 1) and (1, 2). If it had more zeros in either of these two
intervals, then by Rolle’s theorem again (applied to h′ (x) this time),
h′′ (x) would vanish at least once. But we are given that this is not so.
Hence (B) and (C) are correct. That automatically precludes (A) and
(D).
A simple problem based on Rolle’s theorem. One fails to see what
was the idea of introducing the two functions f (x) and g(x) when the
entire problem is about the function h(x) = f (x) − 3g(x). But then,

55
 instead of the table, we would have to be given (2) and the paper-setters
probably thought that that would make the solution too obvious. So,
they worded the problem in terms of two functions. But nothing more
than elementary arithmetic is needed to convert the table to (2). The
question would have been interesting if some other pieces of information
were given from which (2) would follow, but not in such a direct manner.
(Contrast this with Q.46 above, where for an honest solution so many
properties of even and odd functions, and also the FTC had to be used.)
Since both the questions have equal marks, between two competing
candidates with severely limited time, the mediocre one who opts for
the present question will easily beat his possibly more intelligent and
scrupulous rival who might not have time left for this problem.

Q.50 Let f (x) = 7 tan8 x + 7 tan6 x − 3 tan4 x − 3 tan2 x for x ∈ (−π/2, π/2).
Then the correct expression(s) is (are)

π/4 1 π/4
Z Z
(A) xf (x) dx = (B) f (x) dx = 0
Z 0π/4 12 Z0
1 π/4
(C) xf (x) dx = (D) f (x) dx = 1
0 6 0

Answer and Comments: (A) and (B). The paper-setters have care-
fully avoided the end points ±π/2 from the domain because the tan-
gent function is undefined at these points. Anyway that does not mat-
ter because in the problem the interval of integration is [0, π/4] in all
statements. And there is no difficulty at the end points here. (If the
integrand tends to ±∞ at either end points, the integral becomes what
is called an improper integral. Occasionally, improper integrals are
asked in the JEE. See Comment No. 16 of Chapter 18 and Exercise
(18.16). But they are not a part of the syllabus.)
Now, coming to the problem itself, the identity

tan2 x + 1 = sec2 x (1)

simplifies the function to

f (x) = 7 tan6 x sec2 x − 3 tan2 x sec2 x (2)

56
 The substitution

u = tan x (3)

transforms the integrals in (B) and (D) to

Z π/4 Z 1
f (x)dx = 7u6 − 3u2 du
0 0
1
= u7 − u3 = 1 − 1 = 0 (4)
0

The calculation of the integral in (A) and (C) is less direct. The same
substitution gives
Z π Z 1
xf (x)dx = tan−1 u(7u6 − 3u2 ) du (5)
0 0

As we already know an antiderivative of the second factor of the inte-

grand, we resort to integration by parts and get
π/4 1 1 1
Z Z
−1 7 3
xf (x) dx = tan u(u − u ) − (u7 − u3 ) du (6)
0 0 0 u2 +1
Z 1
= 0− u3 (u2 − 1) du
0
Z 1
= u3 − u5 du
0
1 1 1
= − = (7)
4 6 12

The problem is a godsend for the huge mediocrity that banks on

problems asking a direct evaluation of some integral (instead of asking
some areas which first have to be written as integrals). The identity
(1), the substitution (3) and the factorisation of the last integrand in
(6) are so obvious, that a candidate who has cleared JEE Main can be
safely presumed to know them. So this problems tests no new ability.
It hardly belongs to an advanced test.
192x3 1
Q.51 Let f ′ (x) = 4 for all x ∈ IR with f ( ) = 0.
2 + sin πx 2
Z 1
If m ≤ f (x) dx ≤ M , then the possible values of m and M are
1/2

57
 1 1
(A) m = 13, M = 24 (B) m = , M =
4 2
(C) m = −11, M = 0 (D) m = 1, M = 12

Answer and Comments: (D). We are not given f (x) directly. In-
stead we are given f ′ (x) and f (1/2). In theory this determines f (x)
uniquely. In fact
Z x Z x 192t3
f (x) = f ′ (t) dt = dt (1)
1/2 1/2 2 + sin4 πt

But this is of little help because we are not in a position to evaluate

the last integral in a closed form. Fortunately, the problem does not
Z 1
ask for the exact value of the integral f (x) dx, but only for some
1/2
estimates, i.e. some lower and upper bound on it. And these can be
obtained from some estimates on the integrand f (x), which, in turn,
can be obtained from properties of f ′ (x) for x ∈ [1/2, 1]. Note that
f ′ (x) is a ratio of two positive quantities. The numerator increases
strictly on the interval [1/2, 1] (in fact it increases on the entire real
line). But the denominator decreases on [1/2, 1] because of the term
sin4 πx since the sine function decreases on the interval [π/2, π]. So the
192
maximum of f ′ (x) on [1/2, 1] is f ′ (1) = = 96 while its minimum
2+0
24
on [1/2, 1] is f ′ (1/2) = = 8.
2+1
Thus we have proved that
8 ≤ f ′ (x) ≤ 96 (2)
for all x ∈ [1/2, 1]. From this, we now derive estimates for f (x) us-
ing the Fundamental Theorem of Calculus (FTC), and the fact that
f (1/2) = 0. We replace x by t in (2) and integrate over [1/2, x] to get
Z x Z x Z x
8 dt ≤ f ′ (t) dt ≤ 96 dt (3)
1/2 1/2 1/2

BY FTC, the middle integral is simply f (x) − f (1/2) = f (x), while the
first and the last integrals are 8(x − 1/2) and 96(x − 1/2) respectively.
So we have
8x − 4 ≤ f (x) ≤ 96x − 48 (4)

58
 for all x ∈ [1/2, 1]. Integrating again,
Z 1 Z 1 Z 1
8x − 4 dx ≤ f (x) dx ≤ 96x − 48 dx (5)
1/2 1/2 1/2

1
By an easy computation, the first integral equals (4x2 − 4x) = 1
1/2
1
while the last integral equals (48x2 − 48x) = 24 − 12 = 12. These
1/2
are exactly the lower and upper bounds on the given integral as given
in (D).
This is a good problem which tests the ability to estimate when the
exact evaluation is not possible. Apart from some elementary knowl-
edge of trigonometric functions, the entire solution is based on the fact
that inequalities are preserved by integrals, that is to say, if f (x) ≤ g(x)
Z b Z b
for all x ∈ [a, b], then f (x) dx ≤ g(x) dx. It is precisely problems
a a
of this kind and not those of the type of the last problem that are
suitable for an advanced test.
Q.52 Let S be the set of all non-zero real numbers α such that the quadratic
equation αx2 −x+α = 0 has two distinct real roots x1 and x2 satisfying
the inequality |x1 − x2 | < 1. Which of the following intervals is (are) a
subset(s) of S?
√ √ √ √
(A) (−1/2, −1/ 5) (B) (−1/ 5, 0) (C) (0, 1/ 5) (D) (1/ 5, 1/2)

Answer and Comments: (A), (D). For the equation to be quadratic,

its leading coefficient
√ α must be non-zero. When this is the case the
1 ± 1 − 4α2
two roots are . For the roots to be real and distinct, we
2α
must have
4α2 < 1 (1)
Also the difference of the roots x1 and x2 is
√
1 − 4α2
|x1 − x2 | = (2)
|α|
So the second condition about the roots reduces to
1 − 4α2 < α2 (3)

59
 (1) is satisfied if and
√ only if α ∈ √
(−1/2, 1/2) while (3) holds if and only
if α ∈ (−∞, −1/√ 5) ∪ α √ ∈ (1/ 5, ∞). The intersection of these two
sets is (−1/2, −1 5) ∪ (1 5, 1/2).
A very simple problem about the roots of a quadratic.

Q.53 If α = 3 sin−1 ( 11
6
) and β = 3 cos−1 ( 49 ), where the inverse trigonometric
functions take only the principal values, then the correct option(s) is
(are)
(A) cos β > 0 (B) sin β < 0 (C) cos(α + β) > 0 (D) cos α < 0

6 4
Answer and Comments: (B), (C), (D). Denote sin−1 ( ) and cos−1 ( )
11 9
by θ and φ respectively. Then α = 3θ and β = 3φ. We are given
6
sin θ = (1)
11
4
and cos φ = (2)
9

A straightforward approach is to first calculate cos β, sin β etc. from

(1) and (2) using the formulas for the sines and cosines of triples of
angles. But the calculation of sin β = sin 3φ would require the value
4
of sin φ. It can be calculated from cos φ = . But that will involve a
9
radical and hence the calculation will be complicated. The calculations
for (C) will be even more complicated.
So we look for an easier alternative. If we look carefully, we are
interested only in the signs of the various expressions and not so much
in their exact values. And this can be done by locating the quadrants
in which the angles lie. We already know that both θ and φ lie in the
first quadrant, i.e.
π π
0<θ< and 0 < φ < (3)
2 2

But this information is inadequate to locate the quadrants in which

3θ, 3φ and 3(θ + φ) lie. So we need sharper inequalities for θ and φ
than (3). How sharp should they be? If θ lies in an interval [a, b],
then 3θ lies in [3a, 3b] whose length is thrice that of the first interval.
Since the angular measure of each quadrant is π/2, to determine the

60
quadrants in which α, β lie, we should trap θ and φ in intervals of length
π/6 or less.
4
Let us first tackle φ. We are given that cos φ = which is only
9
1 π
slightly less than = cos( ). As the cosine function is decreasing in
2 3
the first quadrant, we have
π
φ> (4)
3
Thus we have found a sharper lower bound on φ than in (3). To get
4
an upper bound, we need an angle whose cosine is slightly less than .
9
But here we need not be very choosy. Since we want to trap φ in an
interval of length π/6, in view of (4), the upper bound π/2 which we
already know from (3) will do. So we now have
π π
<φ< (5)
3 2
The determination of a narrower interval containing θ is similar.
6 1
We are given that sin θ = which is slightly bigger than . Since
11 2
1
sin(π/6) = , we get π/6 as a lower bound on θ. As for an upper
2
1
bound, we try π/4 knowing that sin(π/4) = √ . And, indeed, it works
2
6 1
because does happen to be less than √ as we see by squaring both
11 2
the numbers. So we get
π π
<θ< (6)
6 4
We now have all the ammunition to fire the four shots. Since β = 3φ,
because of (5) β ∈ (π, 3π/2). So β lies in the third quadrant where
both the sines and cosines are negative. This shows that (A) is false
while (B) is true. In a similar vein, because of (6), α lies in (π/2, 3π/4)
which is a part of the second quadrant. So (D) is true. For (C), we
need both (5) and (6). Multiplying them by 3 and adding, we get
3π 9π
<α+β < (7)
2 4
61
 The interval (3π/2, 9π/4) cuts across two quadrants, specifically, the
first and the fourth quadrant. But since the cosine function is positive
in both these quadrants, we see that (C) is true.
Like Q, 41 in Paper I, this question is also an excellent example of
testing the ability to quickly discard the most obvious approach. The
paper-setters have given just the right degree of hint by specifying that
all the inverse trigonometric functions are to be taken to have their
principal values.

Q.54 Let E1 and E2 be two ellipses whose centers are at the origin. The major
axes of E1 and E2 lie along the x-axis and the y-axis, respectively. Let
S be the circle x2 + (y − 1)2 = 2. The straight line x + y = 3 touches
the curves S, E1√and E2 at P, Q and R, respectively. Suppose that
2 2
PQ = PR = . If e1 and e2 are the eccentricities of E1 and E2 ,
3
respectively, then the correct expression(s) is (are)
√
43 7
(A) e21 + e22 = (B) e1 e2 = √
40 2√ 10
5 3
(C) |e21 − e22 | = (D) e1 e2 =
8 4

Answer and Comments: (A), (B). A highly computational problem

in coordinate geometry. The equations of the ellipses E1 and E2 are
not given. We are only given that their major axes lie along the x- and
the y-axes respectively. Hence the equation of E1 is of the form
x2 y 2
+ 2 =1 (1)
a2 b
where a > b > 0. Similarly, the equation of E2 is of the form
x2 y 2
+ 2 =1 (2)
c2 d
where 0 < c < d.
Let us first determine the points of contacts of the line (say L), x+y = 3
with S, E1 and E2 . Putting

y = 3−x (3)

62
into the equation of S gives x2 + (2 − x)2 = 2 or 2x2 − 4x + 2 = 0,
which has x = 1 as its only root, confirming that L is a tangent to S.
From x = 1, y comes out to be 2 by (3). The point of contact, P is

P = (1, 2) (4)

The points Q and R will have to be expressed in terms of the unknowns

a, b, c, d.
Putting y = 3 − x into (1) gives a quadratic

(a2 + b2 )x2 − 6a2 x + a2 (9 − b2 ) = 0 (5)

For tangency, this quadratic has discriminant 0 which simplifies to

a2 + b2 = 9 (6)

3a2 a2
Also the quadratic has only one root, viz. = . Hence the
a2 + b2 3
point of contact comes out to be
a2 9 − a2
Q=( , ) (7)
3 3
√
2 2
As we are given P Q = , we get
3
a2 2 9 − a2 8
( − 1) + ( − 2)2 = (8)
3 3 9
which simplifies to a2 − 3 = ±2. So a2 = 5 or 1 and correspondingly,
from (6), b2 = 4 or 8. But we are given that a > b. So we have
a2 = 5, b2 = 4. Since the eccentricity e1 of the first ellipse is given by

b2 = a2 (1 − e21 ) (9)

we get
s
4 1
e1 = 1− =√ (10)
5 5
The calculation of e2 is entirely similar. We merely replace a, b by c, d
respectively. Since P Q = P R, (8) also holds with a replaced by c. So

63
 once again we get c2 = 5 or 1 and correspondingly, d2 = 4 or 8. But this
time we are assuming d > c. So we have to take the second possibility,
viz. c = 1 and d = 8. Then e2 is given by
s s √
c2 1 7
e2 = 1 − 2 = 1 − = √ (11)
d 8 2 2
Now that we know both e1 and e2 , by a direct calculation we get
√
7 1 7 43 1 7 27
e1 e2 = √ , e21 + e22 = + = , |e21 − e22 | = | − | =
2 10 5 8 40 5 8 40

The point P could as well have been given directtly. After determin-
ing it, Q and R can be determined by solving x + y = 3 simultaneously
wirh (x − 1)2 + (y − 2)2 = 89 . They come out to be (5/3, 4/3) and
(1/3, 8/3) although we do not know which one of them is Q. There is
a short cut (pointed out by Kaustuv Lahiri) to determine the eccen-
tricty because it depends only on the ratio of the axes. Take E1 . The
equation of the tangent to it at a point (x0 , y0 ) is
xx0 yy0
+ 2 =1 (12)
a2 b
2
Comparing this with x + y = 3 gives xy00 = ab 2 . The first ratio is 45
or 8 depending on whether (x0 , y0 ) is Q or R. Since b < a, the first
2
possibility holds. From ab2 = 54 , e1 , the eccentricity of E1 comes out to
2
be as in (10). By a similar reasoning, for E2 , dc2 = 8 and e2 comes out
as in (11).
The symmetry of the data allows us to apply most of the work
for the first ellipse to the second ellipse. The only difference comes in
the last step because of the interchange of the directions of the major
and the minor axes. In this problem, there does not seem to be any
way to directly compute, say e1 e2 , without first identifying e1 and e2
individually (as there would have been if we could write a quadratic
whose roots are e1 and e2 ).

Q.55 Consider the hyperbola H : x2 − y 2 = 1 and a circle S with centre

N(x2 , 0). Suppose H and S touch each other at a point P (x1 , y1 ) with
x1 > 0, y1 > 0. The common tangent to H and S at P intersects the

64
x-axis at a point M. If (l, m) is the centroid of the triangle △P MN,
then the correct expression(s) is (are)

dl 1 dm x1
(A) = 1 − 2 for x1 > 1 (B) = q for x1 > 1
dx1 3x1 dx1 3 x21 − 1
dl 1 dm 1
(C) = 1 + 2 for x1 > 1 (D) = for y1 > 0
dx1 3x1 dy1 3

Answer and Comments: (A), (B), (D). The notations in the problem
are slightly confusing. Usually, P (x1 , y1) would denote a fixed point in
the plane. In the present problem, it is a variable point in the first
quadrant which moves on the hyperbola H. But instead of beginning
the problem by saying so, the point (x1 , y1 ) is specified as the point
of contact of this hyperbola with a circle S whose centre is given as
N = (x2 , 0). So, it is to be assumed that x2 is also not fixed. So, the
picture we gather is that x1 and y1 are functions of the variable x2 .
But the problem asks for derivatives w.r.t. the variables x1 and y1 .

P S

x
O M N

Note that instead of defining P in terms of N, we can as well

express N in terms of P . In fact, the circle S has little role to play
except to tell us that the line P N is normal to S and hence also to H
at P , because S and H have a common tangent at P . So, we could
as well define N as the point where the normal to H at P meets the
x-axis. The point M, on the other hand, is specified directly in terms
of (x1 , y1), as the point where the tangent to H at P cuts the x-axis.

65
 So the essence of the data of the problem is this. Suppose P (x1 , y1 )
is a variable point in the first quadrant portion of the hyperbola H :
x2 − y 2 = 1, M and N are, respectively, the points where the tangent
and the normal to H at P cut the x-axis. And finally, (l, m) is the
centroid of the triangle △P MN. The coordinates of M and N and
hence those of the centroid will be functions of the variables x1 and y1 .
But these two variables are not independent of each other. Either of
them can be expressed in terms of the other by the equations
q
y1 = x21 − 1 (1)
q
and x1 = y12 + 1 (2)
As a result, we can express l both as a function of x1 and also of y1 .
The same goes for m. The problem then asks about the derivatives of
these functions.
Let us now turn to the problem itself. The equation of the tangent
to H at (x1 , y1 ) is
xx1 − yy1 = 1 (3)
1
Its point of intersection with the x-axis is ( , 0). So, we have
x1
1
M =( , 0) (4)
x1
x1 y1
As the slope of the tangent is , that of the normal at P is − .
y1 x1
Hence the equation of the normal to H at P is
y1
y − y1 + (x − x1 ) = 0 (5)
x1
which cuts the x-axis at the point (2x1 , 0). Hence
N = (2x1 , 0) (6)
As (l, m) is the centroid of △P MN, we have
1 1 1
l =(3x1 + ) = x1 + (7)
3 x1 3x1
y1
and m = (8)
3
66
 dl 1
From (7) we get = 1 − 2 which shows that (A) is true and (C)
dx1 3x1
is false. Also from (8), we see that (D) is true. For (B), we first need
to express m as a function of x1 . For this we combine (8) with (1) to
get
q
x21 − 1
m= (9)
3
Differentiating,
dm x1
= q (10)
dx1 3 x21 − 1

which proves that (B) is true.

A simple problem except possibly for its clumsy framing. One
wonders if it was intentional. If so, the idea was probably to test a
candidate’s ability to grab the essence of a problem from a nebulous
data. But it does look silly. It is like a person telling you that he met
his father’s wife’s daughter, instead of saying that he met his sister
(possibly a step-sister)!
Q.56 The option(s) with the values of α and L that satisfy the following
equation is (are)
Z 4π
et (sin6 αt + cos4 αt) dt
Z0 π =L
et (sin6 αt + cos4 αt) dt
0

e4π − 1 e4π + 1
(A) α = 2, L = (B) α = 2, L =
eπ − 1 eπ + 1
e4π − 1 e4π + 1
(C) α = 4, L = π (D) α = 4, L = π
e −1 e +1

Answer and Comments: (A), (C). The given expression is a ratio of

two integrals, say I and J. Both the integrals have the same integrand,
say

f (t) = et (sin6 αt + cos4 αt) (1)

67
The numerator I can be written as

I = I1 + I2 + I3 + I4 (2)

where
Z π Z 2π Z 3π Z 4π
I1 = f (t)dt, I2 = f (t)dt I3 = f (t)dt and I4 = f (t)dt (3)
0 π 2π 3π

Clearly, I1 = J. We need to relate I2 , I3 and I4 to I1 . For this we need

the periodicity of the trigonometric functions. The functions sin2 αt
and cos2 αt and hence all their powers are periodic with a period of π
for both α = 2 and α = 4 (in fact, for all integral values of α) because

sin(α(t + π)) = (−1)α sin αt (4)

and cos(α(t + π)) = (−1)α cos αt (5)

for all t. However, the first factor of the integrand, viz. et is not
periodic. Still we have the identity

et+π = eπ et (6)

for all t. Using (4), (5) and (6), we get

f (t + π) = eπ f (t) (7)

for all t. The simple substitution u = t − π now gives

Z 2π
I2 = f (t) dt
Zππ
= f (u + π) du
Z0π
= eπ f (u) du
0
π
= e I1 (8)

The same reasoning gives I3 = eπ I2 and I4 = eπ I3 . So the terms of the

R.H.S. of (2) are in a geometric progression with common ratio eπ and
the first term I1 . Hence
e4π − 1
I = I1 (9)
eπ − 1

68
This holds regardless of what the integer α is. So both (A) and (C) are
correct.
Problems involving integrals of periodic functions have been asked
in the JEE (e.g. see Exercise (18.8)). But the present one is a novel
one in that it combines periodic functions and a simple property of the
exponential functions. Once the idea strikes, the computations involved
are minimal. So this is a very good problem.
Although it is well beyond the JEE level, we remark that similar
calculations arise in finding what is called the Laplace transform of
a periodic function. Suppose f (t) is a continuous function defined for
all t ≥ 0. Then assuming that the growth of |f (t)| as t → ∞ is not too
wild, it can be shown that the improper integral
Z ∞
F (s) = e−st f (t)dt (10)
0

is convergent for all sufficiently large values of the real parameter s.

We now have a new function, F (s), of a new variable s. It depends on
the function f (t) and is often denoted by L(f ). As easy examples, we
invite you to show, using integration by parts, that
n!
L(tn ) = , n = 0, 1, 2, . . . (11)
sn+1
1
L(eαt ) = (12)
s−α
s
L(cos ωt) = 2 (ω ∈ IR) (13)
s + ω2
ω
and L(sin ωt) = 2 (ω ∈ IR) (14)
s + ω2
What makes the Laplace transforms very useful is the result, easily
proved by integration by parts, that for a continuously differentiable
function f (t),

L(f ′) = sL(f ) − f (0) (15)

As a consequence, a differential equation for a function f (t) is trans-

formed into an algebraic equation for its transform F (s). Algebraic
equations are easier to solve. So just as logarithms are useful in arith-
metic because they transform the multiplication of two positive real

69
numbers (which is time consuming) into the addition of their loga-
rithms (which is much easier), Laplace transforms are very useful in
solving differential equations.
Now suppose f (t) is a periodic function with a period T > 0. That
means

f (t + T ) = f (t) (16)

for all t ∈ IR. Then by taking steps similar Zto those in the solution of
∞
the present problem, for every s, the integral e−st f (t) dt can be split
0
as an infinite series of integrals over intervals of the form [(n − 1)T, nT ]
for n = 1, 2, 3, . . .. As a result, we get

F (s) = I1 + I2 + . . . + In + . . . (17)

where
Z nT
In = e−st f (t)dt (18)
(n−1)T

(17) and (18) are valid for any f (t). But if f (t) is periodic with a period
T , then by taking steps similar to the derivation of (8) above, we get,
that for every fixed s,
Z (n+1)T
In+1 = e−st f (t) dt
nT
Z nT
= e−s(u+T ) f (u + T ) du
(n−1)T
Z nT
= e−sT e−su f (u) du
(n−1)T
−sT
= e In (19)

In other words {In }n≥1 is a geometric progression with common ratio

e−sT . So the the R.H.S. of (17) is an infinite geometric series. As a
result, we get
I1
F (s) = (20)
1 − e−sT

70
 Note that even though F (s) is defined through an improper integral,
the integral I1 is not improper. So periodicity of f (t) allows us to
express its Laplace transform L(f ) in terms of a proper integral over
an interval whose length equals the period of f (t).
The JEE paper-setters are undoubtedly familiar with Laplace
transforms and chances are that the present problem was spurred as a
miniature version of (20).

SECTION 3

This section contains TWO paragraph. Based on each paragraph, there

will be TWO questions Each question has FOUR options (A), (B), (C) and
(D). ONE OR MORE THAN ONE of these four option(s) is(are) correct For
each question, darken the bubble(s) corresponding to all the correct option(s)
in the ORS.
Marking scheme :
+4 If only the bubble(s) corresponding to all the correct option(s) is(are)
darkened
0 If none of the bubbles is darkened
-2 In all other cases.

PARAGRAPH 1
Let F : IR −→ IR be a thrice differentiable function.
Suppose that F (1) = 0, F (3) = −4 and F ′ (x) < 0
for all x ∈ (1/2, 3). Let f (x) = xF (x) for all x ∈ IR.

Q.57 The correct statement(s) is (are)

(A) f ′ (1) < 0 (B) f ′ (2) < 0

(C) f ′ (x) 6= 0 for any x ∈ (1, 3) (D) f ′ (x) = 0 for some x ∈ (1, 3)

Answer and Comments: (A), (B), (C). The problem asks about
f (x) and its derivative f ′ (x). We are given
f (x) = xF (x) (1)

71
 As F (x) is differentiable, so is f (x) and

f ′ (x) = F (x) + xF ′ (x) (2)

Hence

f ′ (1) = F (1) + F ′ (1) = 0 + F ′ (1) = F ′ (1) < 0 (3)

since F ′ (x) < 0 for all x ∈ (1/2, 3). So, (A) is true.
For (B),

f ′ (2) = F (2) + 2F ′(2) (4)

We are not given the values of either F (2) or F ′ (2). But we are given
that F ′ (x) < 0 for all x ∈ (1/2, 3). So F (x) is strictly decreasing
on [1/2, 3]. (This assertion requires the use of Lagrange’s Mean Value
Theorem. But it is possible that this subtlety is missed by many can-
didates, and, being an MCQ, there is no way to tell.) Since F (1) = 0
and 2 > 1, we get F (2) < F (1) = 0. Also since 2 ∈ (1/2, 3), F ′ (2) < 0
straight from the hypothesis. So by (3), (B) is also true.
The options (C) and (D) are the logical negations of each other.
So exactly one of them is true. The reasoning given in (B) for x = 2
applies for any x ∈ (1, 3) and so from (2), f ′ (x) < 0 for all x ∈ (1, 3),
as both the terms on the R.H.S. are negative. So (C) is true, and
automatically, (D) is false.
Z 3 Z 3
2 ′
Q.58 If x F (x) dx = −12 and x3 F ′′ (x) dx = 40, then the correct ex-
1 1
pression(s) is (are)
Z 3
(A) 9f ′ (3) − f ′ (1) − 32 = 0 (B) f (x) dx = 12
Z1 3
(C) 9f ′ (3) − f ′ (1) + 32 = 0 (D) f (x) dx = −12
1

Answer and Comments: (C), (D). All options involve exact val-
ues of some integrals and not just their sign determinations. So, the
inequalities given in the paragraph are not likely to be of much help.
From the nature of the integrands, it is obvious that integration by

72
 Z Z
parts is called for. But when we form f (x)dx = xF (x) dx nothing
is given to us about an antiderivative of F (x). So we have to start from
an antiderivative of the first factor, viz. x of the integrand. We then
get
Z 3 Z 3
f (x) dx = xF (x) dx
1 1
x2 3
Z 3 2
x ′
= ( F (x)) − F (x) dx (5)
2 1 1 2

The value of the integral on the R.H.S. is given to us as −6. By a direct

9 1
substitution, the first term is F (3) − F (1). The values of F (3) and
2 2
F (1) are given as −4 and 0 respectively. Substituting these, the R.H.S.
of (5) comes out to be −18 − 0 + 6 = −12. Hence (D) is correct and
(B) false.
To determine which (if any) of the remaining options is true, we use
the second piece of information, viz.
Z 3
x3 F ′′ (x) dx = 40 (6)
1

The integral on the L.H.S. can be evaluated by parts as

Z 3 3
Z 3
3 ′′ 3 ′
x F (x) dx = (x F (x)) − 3x2 F ′ (x) dx (7)
1 1 1

Again, we are given the value of the integral on the R.H.S. as −36. So,
by a direct substitution we get

27F ′ (3) − F ′ (1) = 40 − 36 = 4 (8)

Since the options (A) and (C) are in terms of f ′ rather than F ′ , we
must convert F ′ (3) and F ′ (1) suitably. Applying (2) and the fact that
F (3) = −4, we have
f ′ (3) − F (3) f ′ (3) + 4
F ′ (3) = = (9)
3 3
and similarly, since F (1) = 0,
f ′ (1) − F (1)
F ′ (1) = = f ′ (1) (10)
1
73
 Putting these values into (8), we get

9f ′ (3) + 36 − f ′ (1) = 4 (11)

which means (C) is true and (A) false.

A simple, but extremely weird pair of questions. Conceptually, the
two questions have little in common. But because of the constraint in
paper-setting, they have been artificially clubbed together into a sin-
gle paragraph. The first question is based on applying the Lagrange’s
Mean Value Theorem to get some inequalities. The second one is based
on integration by parts. The work involved is repetitious and the arith-
metic required is arbitrary and prone to numerical errors.

PARAGRAPH 2

Let n1 and n2 be the number of red and black balls,

respectively, in box I. Let n3 and n4 be the number
of red and black balls, respectively, in box II.

Q.59 One of the two boxes, box I and box II was selected at random and a
ball was drawn at random from out of this box. The ball was found to
be red. If the probability that this red ball was drawn from box II is
1
, then the correct option(s) with the possible values of n1 , n2 , n3 and
3
n4 is (are)

(A) n1 = 3, n2 = 3, n3 = 5, n4 = 15
(B) n1 = 3, n2 = 6, n3 = 10, n4 = 50
(C) n1 = 8, n2 = 6, n3 = 5, n4 = 20
(D) n1 = 6, n2 = 12, n3 = 5, n4 = 20

Answer and Comments: (A), (B). This is a problem of conditional

probability. We are first selecting a box and then drawing a ball from

74
the selected box. We want to find the probability that the ball drawn
is from the second box, given that its colour is red.
There is a chain of two processes here, first selection of a box and
then the draw of a ball. So for a visual representation of the thought
process, tree diagrams are more convenient than Venn diagrams as
elaborated in Comment No. 11 of Chapter 22. The appropriate tree
diagram for the present problem is shown below.

1/ 2 1/ 2

I II

p1 p p p
3 4
2
r b r b

Here A is the starting node. From there, there are two branches lead-
ing to the two nodes I and II corresponding to the two boxes. Each
branch has probability 1/2. From each of these two nodes, there are
two branches leading to the nodes marked r and b depending upon the
colour of the ball drawn. Their probabilities are marked as p1 , p2 , p3 , p4 .
They can be easily calculated from the data of the problem as
n1 n2 n3 n4
p1 = , p2 = , p3 = and p4 = (1)
n1 + n2 n1 + n2 n3 + n4 n3 + n4

Now let R be the event that the ball drawn is red. Then from the
diagram above,
1 1
P (R) = p1 + p3
2 2
1 n1 1 n3
= + (2)
2 n1 + n2 2 n3 + n4
Let E2 be the event that the second box was chosen. Of course P (E2 ) =
1
. But the desired probability is not this, but rather the conditional
2
75
probability P (E2 |R), i.e. the probability that the ball drawn is from
the second box given that it is red. By Bayes theorem (see Comment
No. 5 of Chapter 22), or more easily, by common sense,
P (E2 ∩ R)
P (E2 |R) = (3)
P (R)
The denominator was already calculated in (2). It is clear that the
numerator is simply the second term of the R.H.S. of (2). So, the
desired probability, say p, is
n3
n3 + n4
p= n1 n3 (4)
+
n1 + n2 n3 + n4

The conceptual part of the problem ends here. In a numerical

problem, we would be given the values of n1 , n2 , n3 and n4 and asked
1
to calculate p from (4). But here we are given that p = and are
3
asked to see for which of the given sets of values of n1 , n2 , n3 and n4 ,
1
the answer comes out to be . So, in essence, we are asked to repeat
3
the arithmetical work four times. A slight simplification is possible.
p3 1
Note that p can be written as . So this will equal if and only
p1 + p3 3
if p1 = 2p3 . In (A), p1 = 1/2, p3 = 1/4 and so (A) is true. In (B),
p1 = 1/3, p2 = 1/6 and so (B) is also true. In (C), p1 = 4/7, p3 = 1/5
and so (C) is false. Finally, in (D), p1 = 1/3, p3 = 1/5 and so (D) is
false.
A very standard problem of computing conditional probability.
But one wonders what is gained by forcing the candidate to do the
last bit of arithmetic four times. Why not give only one set of values
of n1 , n2 , n3 and n4 and give four numerical alternatives for the value
of p? This used to be the standard practice in the past. But there
have been cases where, because of a carelessness on the part of the
paper-setters, the fake answers could be eliminated easily without doing
the problem honestly, for example, when the fake answer was given as
a rational number whose denominator did not divide the size of the
sample space. It is probably to avoid this kind of a criticism that this
time the paper-setters have chosen to torture the candidates.

76
Q.60 A ball is drawn at random from box I and transferred to box II. If the
1
probability of drawing a red ball from box I, after this transfer, is ,
3
then the correct option(s) with the possible values of n1 and n2 is(are)

(A) n1 = 4 and n2 = 6 (B) n1 = 2 and n2 = 3

(C) n1 = 10 and n2 = 20 (D) n1 = 3 and n2 = 6

Answer and Comments:. (C), (D). It is given that a ball, randomly

drawn from box I, is transferred to box II. But the second part of the
problem deals only with what is left in box I. So what happens to
this ball after its removal from box I is irrelevant. The paper-setters
could simply have said that one ball at random was removed from box
I, or kept the data as it is except that the second ball is drawn from
box II. That would make the transfer relevant and the problem more
interesting.
Let p1 and p2 be the probabilities that the ball removed was red
and black respectively. Clearly,
n1 n2
p1 = and p2 = (5)
n1 + n2 n1 + n2
Now, if the ball removed is red, then box I is left with n1 − 1 red and
n2 black balls. The probability of drawing a red ball from this box is
p3 , given by
n1 − 1
p3 = (6)
n1 + n2 − 1
Similarly, if the ball removed is black, then box I is left with n1 red
and n2 − 1 black balls and the probability, say p4 , of drawing a red ball
from it now is
n1
p4 = (7)
n1 + n2 − 1
With the help of a tree diagram, if necessary, the probability, say p,
that the second ball drawn from box I is red, is

p = p1 p3 + p2 p4
n1 (n1 − 1) n2 n1
= +
(n1 + n2 )(n1 + n2 − 1) (n1 + n2 )(n1 + n2 − 1)

77
 n1 (n1 + n2 − 1)
=
(n1 + n2 )(n1 + n2 − 1)
n1
= (8)
n1 + n2
1
We are given that p = . This will happen if and only if n2 = 2n1 .
3
Clearly this is the case in options (C) and (D) but not in the others.
This problem is much simpler than the last one. In fact, a perceptive
student will hardly fail to notice that p is the same as p1 , i.e. the same
as the probability of drawing a red ball from the box I at the start. It is
unlikely that he will have the time to ponder if this is just a coincidence
or can be somehow justified. But if he does, he can paraphrase the
problem to say that two balls were drawn from box I, one after the
other, and the problem asks for the probability that the second ball is
red. How do we justify that this is the same as the probability that the
first ball is red?
Intuitively, the answer is that we can interchange the two balls
drawn. For a formal proof, let n = n1 + n2 . Let us suppose that the
red balls as well as the black balls are distinct from each other. (This
does not affect the probabilities of the events involved.) Let X be the
set of all these n balls. Our sample space, say S is the set of all ordered
pairs (x, y) ∈ X × X such that x 6= y. Clearly S has n(n − 1) elements.
Now let F1 and F2 , respectively, be the favourable subsets for the event
that the first ball is red and the event that the second ball is red. In
symbols,

F1 = {(x, y) ∈ S : x is red} (9)

and F2 = {(x, y) ∈ S : y is red} (10)

As the sample space is the same, proving that the two events have the
same probability is equivalent to showing that the sets F1 and F2 have
the same number of elements. This can be done by defining a bijection
f : F1 −→ F2 which interchanges the two balls drawn. Formally,
f (x, y) = (y, x). Then f is its own inverse and hence a bijection.
Note that this argument will also apply if we draw 3 balls, one
after the other without replacement. The probability that the first one
be red equals the probability that the second one be red and also the

78
probability that the third one be red. Indeed this will hold even if the
box contains balls of more than two colours and we draw any number
of balls from it as long as this number does not exceed the total number
of balls in the box. (If it does, then the sample space is vacuous and
the probability is undefined.)
As in every year, the MCQ format makes it impossible to reward a
candidate who has the ability to do this extra bit of thinking. In fact,
he is advised against it because it is a sheer waste of his precious time.

79
 CONCLUDING REMARKS

The paper-setters have been careful to avoid mathematical mistakes and

ambiguities. For example, in both the questions involving the composite
of two functions, the order in which the two functions act has been clearly
specified. In Q. 47 of Paper 1, the symbol a is used to denote a variable.
This should have been made clear, because normally the symbols a, b, c etc.
are used for constants and x, y, z, t etc. for variables. Similarly, in Q. 55 of
Paper 2, it should have been made clear that x2 and x1 are variables. In
Q.54 of Paper 1, instead of saying ‘true’ it would have been better to say
‘necessarily true’ to avoid the degenerate possibility of a zero matrix which
is symmetric as well as skew-symmetric. But these omissions are unlikely to
have caused any serious confusion.
Except for a slight lapse of the article in Q. 58 of Paper 2, (saying, ‘a circle
with a given diameter’ instead of saying ‘the circle’) there are no anomalies
in the wording of the problems. As pointed out in Q.60 of Paper 2, the in-
formation that the ball removed from box I was put into box II is irrelevant.
But again, these are things to be forgotten after a slight chuckle. On the
contrary, as commented in Q. 47 of Paper 1, the paper-setters have com-
mendably avoided an inconsistency in the data which could have resulted if
there was a repetition of carelessness in the drafting of a similar question in
JEE 2011.
A controversy did arise about the word ‘component’ used in Q.47 of Paper
2. In a conventional examination, from the answerbook of a candidate, it
would be possible to see which interpretation he had taken. And then a
decision could be taken as to how much credit be given. The MCQ format
makes this impossible. The JEE organisers have decided to award full marks
to every candidate. As a result, those who interpreted and solved the problem
correctly are at a loss. Fortunately, it is not a serious loss because the
problem was very trivial anyway. The JEE Advanced organisers deserve to
be commended for showing this openness of mind. (In an earlier version of
this commentary, the JEE Advanced organisers were wrongly criticised. The
mistake is regretted.)
Given the constraints imposed on them, the paper-setters have come up
with some good problems. The two that stand out are Q.41 in Paper 1
(about the reflection of a parabola) and Q.53 of Paper 2 (about the sign
determination of the trigonometric functions of angles that are triples of some
given angles). Q. 51 of Paper 1 (about the differentiability of the composite of

80
two functions), Q.44 (about orders of magnitude) and Q.46 (about functions
defined by integrating odd functions) of Paper 2 are also very good questions
of the conceptual type. But their ability to separate the men from the boys
is marred by the multiple choice format.
Some other good questions in Paper 1 are Q.43 (about boys and girls
standing in a row), Q.47 (about a function defined by integrating cos2 t), Q.54
(about symmetric and skew-symmetric matrices), Q.55 (about expanding a
determinant of order 3) and Q.59(C) (about the complex cube roots of unity),
while those in Paper 2 include Q.41 (about the sums of the terms in an
arithmetic progression), Q.48 (about representing complex roots of unity as
vertices of a regular polygon), Q.51 (about finding lower and upper bounds on
an integral), Q.54 (about the eccentricities of two ellipses) and Q.56 (about
integrating products of periodic functions with the exponential function).
Questions which are too straightforward or familiar and hence have no
place in an advanced test include Q.42 (about complementary probability
with a coin toss), Q.46 (about minimising the material needed for a cylin-
drical container), Q.49 and Q.50 (both on differential equations) and Q.56
(about parametrised family of planes) in Paper 1 and Q.47 (about resolv-
ing a vector), Q.49 (about applications of Intermediate Value Property and
Rolle’s theorem), Q.50 (a straight, simple integral), Q.52 (about the roots of
a quadratic) and both the probability problems (Q.59 and 60) in Paper 2.
There are also numerous questions which reduce to writing down and solving
a system of equations, often in two unknowns. As there is nothing very great
in writing these equations or in solving them, such questions only reward the
mediocrity.
Many of these questions could have been dropped and room made to
accommodate certain areas that are singularly absent, such as number theory,
surds, logarithms and inequalities. This is probably the first year when
there is no question based on the A.M.-G.M. inequality. It is true that
the multiple choice format precludes all proofs and that includes proofs by
induction, combinatorial identities, triangular optimisation and many results
in theoretical calculus. Still, in the past the paper-setters have managed to
give at least a token place to the areas just mentioned.
There is also a tendency apparent in many questions to force the can-
didates to do extra, repetitious drudgery after the conceptual part of the
problem is over. For example, Q.54 in Paper 2 logically ends with the deter-
mination of the eccentricities of the two ellipses. Subsequent work is sheerly
clerical. Similarly, in Q.59 and 60 about probabilities, instead of giving some

81
particular values of the parameters n1 , n2 , n3 and n4 and asking for a numer-
ical answer, which is the standard practice, the paper-setters have given four
sets of such values and asked the candidates to identify those sets of values
in which the answer will come out to be a given number. One really fails
to see the purpose of such a twist when what it tests is not even remotely
related to any probability, It is sheer arithmetical drudgery. In Q.53 of Paper
1, once the vectors ~a, ~b and ~c are identified correctly, little purpose is served
by asking the candidates to compute the various expressions involving their
dot or the cross products.
Some questions demand work that is far in excess of what is justified by
their credit. A notable example is Q.60(D) where a candidate has to sketch
three regions and find their areas all for 2 marks.
There is no significant difference either in the standard or the topics cov-
ered in the two papers. One really wonders the rationale behind having two
papers each having all three subjects. Why not have three separate papers,
one for each subject and give the paper-setters of each subject the freedom
to choose the format of that paper as demanded by the peculiarities of the
subject, rather than impose the same format for all the three subjects? In
Chemistry, many of the questions are memory oriented. You simply have to
know the answer beforehand. There is no way you can get it by deductive
reasoning. For example, questions that ask for the compounds that result
from some chemical reactions.
As a sample, as many as eight questions out of the 20 questions in Chem-
istry in Paper 2 of this year’s advanced JEE directly ask to identify the
products of some reactions. Q.39 in Paper 1 is for eight marks and asks to
identify the radicals (from a given list) occurring in each of the five given
ores, viz. siderite, malachite, bauxite, calamine and argentite. This is sheer
memory. If you know the chemical composition of these ores you get 8 marks
within a few seconds. If you don’t, just forget the question. No amount
of deductive reasoning can help you. Its counterpart in mathematics, viz.,
Q.59 in Paper 1, requires you to solve four completely unrelated problems,
each requiring some thought and moderate computation. There is simply no
comparison between the two.
One can argue that even in chemistry, there are questions where some
thought and numerical work is needed. This is true. But the thrust in these
questions is on knowing some reaction. For example, in one of the questions
in Paper 2, a closed vessel with rigid walls containing 1 mol radioactive
uranium and one mol of air is given. The question then asks what will be

82
the ratio of the final pressure to that of the initial one after the uranium
decays completely. The answer comes from the fact that in the radioactive
decay, every atom of uranium produces one atom of lead and eight atoms of
helium. This is pure memory. The ‘thought’ needed afterwards is that now
in the vessel we have 8 mols of gas in addition to the original 1 mol of air.
As the volume and the temperature are the same, the new pressure will be
9 times the original. This bit of mathematics is childishly simple. The same
thing is true about a question in which diborane reacts with methanol.
The mathematical equivalent of such a single idea question, would be, for
example, to give a right angled triangle with hypotenuse 5 and one side 4 and
ask to find the third side, or to ask in how many ways√ 5 boys and√5 girls can
stand in a row. In both the cases, the answers (viz. 52 − 42 = 25 − 16 =
√
9 = 3 and 10! respectively) come with the application of single formulas.
But such questions are not asked in mathematics even in the elimination
round. The second question appears as a part of Q.43 in Paper 1 this year.
But it is only the beginning and not the end. In fact, the real problem then is
to count in how many of these 10! arrangements, all the 5 girls stand together
and also in how many of them exactly four girls stand together. This cannot
be done just by a hand counting. It requires a thought. Once the correct
thought is grasped, the arithmetic needed is elementary. But the thought
has to come from the candidate. No manual can help.
This is certainly not to deride Chemistry as a discipline of science. But
the point is to decide which qualities are more important when they are being
tested by an examination like the JEE whose avowed purpose is to select the
most gifted candidates. It would be great if the IIT’s get a sufficient number
of candidates who are good in all respects. That is, they know the chemical
composition of bauxite as well as have the ability to solve the problem above
of 5 boys and 5 girls. Unfortunately, this does not happen and so the IIT’s
have to make a choice. So, suppose the IIT’s have to choose between two
students, say A, who can do the problem of 5 boys and 5 girls but does not
know, or has forgotten what bauxite is, and another candidate B who knows
what bauxite is but cannot do the thinking needed for the problem of 5 boys
and 5 girls. Whom should they choose? Clearly, the choice should be A,
because he can be taught what bauxite is, or, when need arises, can easily
look it up. But, although B can also be taught how to do this particular
problem, the ability to do such thinking originally cannot be inculcated into
him as it is largely an innate quality.
In Physics the picture is better than in Chemistry. There are virtually no

83
questions which are purely memory oriented. Most questions require some
work to arrive at the answer. But usually, this work consists of applying one,
or in some cases, two standard formulas. Once you know the right formulas
to apply, the rest of the work is purely mathematical. For example, in one
of the Physics questions in Paper 2 this year, two solid spheres of equal
radii R but with different mass density functions are given and we have to
find the ratio of their moments of inertia about axes passing through their
respective centres. This is a standard application of integrals. At one time it
was taught in mathematics under mechanics. There is one question where a
parallel plate capacitor is given, the plates being separated by two dielectric
slabs of equal thickness but different relative permittivities. We are asked
to find its capacitance given its capacitance in air. A diagram is also given.
That makes it abundantly clear that we have to apply the formula for the
capacitance of two capacitors in series. Once this is realised, the mathematics
involved is straightforward.
In fact, in some questions in the Physics paper, the relevance with physics
is superficial. Take, for example, this question from Paper 2 where we are
given that the energy of a system as a function of time t is E(t) = A2 e−αt ,
α being a given constant. We are further given the percentage errors in the
measurements of A and of t and are asked to calculate the percentage error
in E(t) at t = 5. Now, this is a purely mathematical question, involving
first order approximation using derivatives. (Ironically, that is usually done
in the first mathematics course in the IIT’s after the successful candidates
enter them!) There is hardly any physics in this problem except the name.
The problem could as well have been posed as a problem in economics by
replacing the phrase ‘the energy of a system at time t’ by a phrase like ‘the
demand for some commodity at time t’.
The point to note is that once a question in chemistry is understood, it
takes very little time to answer it, provided you know the answer. In physics,
it takes some work to arrive at the answer. But that work is usually some
standard mathematical computation. Moreover, the physics paper-setters
usually make it easier for the candidates to understand the question, by
supplying appropriate diagrams. This year, for example, the Physics section
of Paper 2 has as many as 10 diagrams for the 20 questions.
Contrast this with mathematics. There is not a single diagram in the
mathematics sections of either of the two papers! And this year is not an
exception. In fact, some questions seem to be intentionally obscure. For
example, Q.55 of Paper 2 about the tangent and the normal to a hyperbola.

84
This problem as well as many other problems where the data is geometric
would be easier to understand if appropriate diagrams were given. But then
the trouble is that they will also be easier to solve. For example, if a diagram
were given for Q.41 of Paper 1 (about the reflection of a parabola) that would
also give an unwarranted hint to the solution. And that will kill the beauty
of the problem. One of the qualities considered desirable in mathematics is
the ability to read between the lines.
Summing up, Chemistry questions test the knowledge of facts, Physics
questions test the ability to apply some standard formulas. But Mathematics
is inherently different. A mathematical equivalent of the question about the
capacitor considered above, would be, for example, a question asking for
the area of a triangle, given its two sides and one angle. None of the good
questions listed above are of this type. There is an element of art and sport
in mathematics which is generally absent in physics and chemistry which are
sterile sciences. It becomes very difficult for the paper-setters in mathematics
to cater to this element if they are forced to frame the questions in the
same format as physics and chemistry. Designing and solving a truly good
mathematics problem is an artistic experience. No wonder some of the good
problems in the old JEE’s are remembered even after decades. (Two such
problems from the 1994 JEE have been referred to in this commentary, one
in Q.55 of Paper 1 and the other in Q.48 of Paper 2.) Can that be said about
any of the mathematics problems of the recent JEE’s?
True improvement would come only if the unique position of mathematics
is recognised by the policy makers. When the candidates are already screened
by JEE Main and the number of those eligible for the Advanced test is re-
duced to a fairly manageable size, there is no reason to duplicate the selection
through another test of the same level and the same format. The Advanced
JEE ought to be qualitatively different from the Main one. At present it is
not. Because of the constraints on the paper-setters in mathematics, they
are forced to ask questions which fit only in the screening round but hardly
qualify for the final selection.
Ideally, at the Advanced JEE, there should be two papers of three hours
each, one solely for mathematics and the other for a combination of physics
and chemistry in the proportion, say 2 : 1. Since the Main examination
already tests the degree of exposure to the three subjects, there should be no
compulsion to cover the entire syllabus in the Advanced papers. Moreover,
mathematics paper-setters should have the freedom, as they did many years
ago, to design questions requiring different amounts of work and allot the

85
credit proportionately. The evaluation should be manual, where the quality
of thinking of a candidate can be assessed, as he has to give reasoning for his
work.
One can only pray that this will happen in near future.

86