EDUCATIVE COMMENTARY ON

JEE 2017 ADVANCED MATHEMATICS PAPERS

Revised Draft uploaded on June 8, 2017

(Apart from correcting some typographical errors, the only changes are in
Q.52-54 of Paper 1, Q.37, Q.47 of Paper 2 and Problem 2 at the end.)

Contents

Paper 1 2

Paper 2 34

Concluding Remarks 71

The pattern of JEE Advanced 2017 is not substantially different from

that of 2016. The number of questions has also remained the same, viz. 18
per subject in each paper. A novel feature is tripartite match making, i.e.
selecting matching entries from three rather than two columns. Some partial
marking is also introduced where more than one answer is correct.
As in the past, unless otherwise stated, all the references made are to the
authors book Educative JEE (Mathematics) published by Universities Press,
Hyderabad. The third edition of this book is now available in the market.
Readers who notice any errors in it are invited to send an email to the
author at kdjoshi314@gmail.com or send an SMS or a WhatsApp message to
the author at 9819961036. Alternate solutions and any other comments are
also solicited.
The questions taken here are from Code 9.
I heartily acknowledge the help rendered by Deepanshu Rajvanshi, Shailesh
Shirali and Siddhesh Naik in preparing this commentary. But most of all, I
thank the countless and often nameless readers who eagerly inquired when
the commentary would be ready. Some of them volunteered to help to ex-
pedite it. Because of the multiple choice format and many other constraints
in paper-setting, interesting questions in mathematics are getting rarer. The
continuation of these annual commentaries has been possible largely because
of the keen interest shown by the readers.
 PAPER 1

Contents

Section - 1 (One or More Answers Correct Type) 2

Section - 2 (Single Digit Integer Answer Type) 20

Section - 3 (Matching Type) 25

SECTION - 1 (Maximum Marks : 28)

This section contains SEVEN questions each of which has FOUR options
out of ONE or MORE is(are) correct..
Marking scheme :
+4 If the bubble(s) corresponding to all correct answers and no others is(are)
darkened
+1 for darkening a bubble corresponding to each correct answer and no others
0 If no bubble is darkened
2 In all other cases.

Q.37 Which of the following is(are) NOT the square of a 3 3 matrix with
real entries?

1 0 0 1 0 0
(A)

0 1 0 (B)

0 1 0

0 0 1
0 0 1
1 0 0 1 0 0
(C)
0 1 0 (D)
0 1 0
0 0 1 0 0 1

Answer and Comments: (A,B). A brute force approachwould be to

x1 y1 z1
take a 33 matrix M with unknown real entries, say, M = x2 y2 z2

,
x3 y3 z3
calculate M 2 and equate it with each of the four given matrices, one at
a time. This will give a system of 9 equations in the 9 unknowns, viz.

2
x1 , y1 , z1 , x2 , y2 , z2 , x3 , y3 and z3 . For example, for the matrix in (A),
the first equation will be x21 + y1 x2 + z1 x3 = 1, while the second one will
be x1 y1 + y1 y2 + z1 y3 = 0. These equations are not linear. And even
if they were, solving a system of 9 equations in as many unknowns is
hardly an enviable task. So this approach is clearly ruled out. We have
to try something else.
When a direct approach to prove the existence of something (in the
present problem, a square root of a given matrix) fails, one resorts to
associating some new objects to the original ones so that the existence
or non-existence of an original object is reduced to that of its associ-
ated object. This approach is more useful for proving non-existence
rather than existence because the information it is based on is usually
a one way street. It is rather like comparing two objects (which we
are invisible to us for some reason) by comparing their shadows (which
we are in a position to see). As a mathematical example, suppose we
are given two triangles, say 4ABC and 4P QR and we have to decide
if they are congruent. If the coordinates of the vertices are given, we
can quickly calculate their areas. Although equality of areas does not
necessarily mean congruency, their inequality definitely rules it out.
In the present problem, it is not easy to directly test if a matrix P
has a square root, i.e. a matrix Q such that P = Q2 . But it is much
easier to tell if a number has a square root. (The answer will depend
upon where the square root is allowed to lie. Thus, 2 has a square
root in IR, the set of real numbers, but not in Q, | the set of rational
numbers. Similarly, 1 has a square root in C | but not in I R. 25 has a
square root in ZZ but 18 doesnt.)
So, we want to associate to a 3 3 matrix P some number, say
(P ) in such a way that the existence of a square root of P will imply
the existence of a square root of (P ) (and, with greater luck the con-
verse would hold too). There are many numbers one can associate to a
matrix, such as its row size, its column size, its determinant, its trace,
the sum of the squares of its elements and so on. Out of these, we want
to pick one which is compatible with the squaring operation, that is
which has the property that (P 2 ) = ((P ))2 . Clearly, the determi-
nant fits the table because of its multiplicative property, viz. that the
determinant of a product equals the product of the determinants. (See
Comment No. 22 of Chapter 2 for clever applications of this property.)

3
 Once this idea strikes, we immediately see that the matrices in
(A) and (B) cannot have any real matrices as square roots. For their
determinant is 1 which has no square root in IR. (They do have
square roots if complex
numbers are allowed as entries. For example,
1 0 0
the matrix 0 1 0 is a square root of the matrix in (A) which is

0 0 i
easily seen as the square of a diagonal matrix D = (dij ) is the diagonal
matrix (d2ij ).
A similar reasoning shows that the matrix in (C) has a square root.
In fact, it has at least eight distinct square roots, viz. all 3 3 diagonal
matrices with diagonal entries 1. (As the question deals with only
the existence of one square root, it is enough to note that the matrix
in (C) is the identity matrix I3 which is its own square root.)
It is only the matrix in (D) which calls for some new thinking. Its
determinant is positive and so we cant rule out its square root on this
ground. But if we try to construct square roots by taking square roots
of the diagonal entries, we run into complex numbers. To check if it
has a square root, we recast it as a partitioned matrix by drawing lines
after the first row and the first column. Thus
..

1 . 0 0
1 0 0
. . . . . . . ..

0 1 0 = .. (1)

0 . 1 0

0 0 1
.
0 .. 0 1
We now treat the R.H.S. as a 2 2 matrix whose entries are themselves
matrices of suitable orders. Specifically,
..

1 . 0 0
... ... ...
" #
A11 A12


. = A (2)

0 .. 1 0 21 A22


.
0 .. 0 1
where
" # " #
0 1 0
A11 = [1], A12 = [0 0], A21 = and A22 = (3)
0 0 1

4
If we have two partitioned matrices such that the sizes of their en-
tries (which are themselves matrices) are suitable for matrix operations,
then these operations can be preformed on them exactly as on matrices
whose entries are numbers. For example, if we have two ordinary ma-
trices B and C (with orders m n and n p for some positive integers
m, n, p) which are partitioned as
" # " #
B11 B12 C11 C12
B= and C = (4)
B21 B22 C21 C22
then we can multiply them and write
" #
B11 C11 + B12 C21 B11 C12 + B12 C22
BC = (5)
B21 C11 + B22 C21 B21 C12 + B22 C22
where the L.H.S. is an ordinary m p matrix with numerical entries
and each of the four entries of the matrix on the R.H.S. is a matrix of
appropriate size. Naturally, for this to happen, we need to impose some
additional restrictions. It is inherent in the definition of a partitioned
matrix that in (4), the row sizes of the matrices in any row are the same
and the column sizes of the matrices in any column are also equal to
each other. But, in order that the matrix B11 C11 + B12 C21 be defined,
it is necessary that the column sizes of B11 and B12 match, respectively,
with the row sizes of C11 and C21 . The other entries of the matrix on
the R.H.S. impose similar additional restrictions.
At first sight, partitioning might not appear to be of any advantage.
After all, if we compare the two sides of (5) the work needed in calcu-
lating all their entries is the same. This is consistent with our day to
day experience that even if we partition a big hall into four (or more)
cabins, the sweeper has to clean exactly the same floor area. But, in
this analogy, some saving can result if some of the cabins are not going
to be needed and hence can be kept locked. Similarly, if some of the
matrices occurring as entries of a partitioned matrix are the zero matri-
ces of appropriate sizes, then the result of the multiplication simplifies.
For example, suppose that all the non-diagonal matrices B12 , B21 , C12
and C21 are zero matrices of appropriate sizes then (5) takes the simple
form
" #" # " #
B11 O C11 O B11 C11 O
BC = = (6)
O B22 O C22 O B22 C22

5
In particular, if B = C and the two partitions are also the same, then
we have
" #
2
2 B11 O
B = 2 (7)
O B22

(This is a special case of the theorem on p. 946 in Appendix I.)

1 0 0
Let us now apply this to the matrix in (D), viz. 0 1 0 . Let

0 0 1
us call this matrix as P . We are checking if there is some 3 3 matrix
Q with real entries such that P = Q2 . Since det(P ) > 0, we cannot
rule out the existence of such a Q the way we ruled out the square
roots of the matrices in (A) and (B). Let us, then, assume tentatively
that Q exists and try to construct it. A brute force construction is
out of question. But in (1) we already cast B as a 2 2 partitioned
matrix with both non-diagonal entries zero. So, let us begin the hunt
for Q" as a partitioned
# matrix of this special type. That is, assume
Q11 O
Q= . Then applying (7), the equality P = Q2 becomes
O Q22

..

1 . 0 0
... ... ...
" #
Q211 O


.. = (8)

0 . 1 0

O Q222

..
0 . 0 1

which is equivalent to two separate matrix equations, viz.

[1] = Q211 (9)

" #
1 0
and = Q222 (10)
0 1

(9) is very easy to solve. The matrix Q11 is a 1 1 matrix, say [q11 ]
2
and so (9) simply means q11 = 1 which has two possible solutions, viz.
q11 = 1.
Thus, we have reduced the problem from finding a square root of
a 3 3 matrix P to that of finding a square root of the 2 2 matrix

6
" #
1 0
. Now the brute force method is not so horrendous. Indeed,
0 1
" #
a b
if we let Q22 = , then (10) becomes
c d
" # " #
1 0 a2 + bc (a + d)b
= (11)
0 1 (a + d)c d2 + bc
Equating the corresponding entries of the matrices on the two sides,
we must have a + d = 0 as otherwise b = c = 0 would give a2 = 1
which has no real solution. So, d = a and the system reduces to
a2 + bc = 1 which has infinitely many solutions, the simplest being
a = d = 0, b = 1, c = 1. Thus we see that (D) is not a correct
answer for the question.
Although the work needed in (D) is considerably more than that in
(C), once we realise that the matrix can be written in a partitioned
form, it is not difficult to" reduce the# problem to that of checking
1 0
whether the 2 2 matrix has a (real) matrix as a square
0 1
root. We solved it by the brute force method. But there is a more
elegant solution. We go back to our earlier observation that sometimes
it pays to associate some new objects to the given ones and to trans-
late the problem in terms of these new objects. In the present case
these new objects are complex numbers. As pointed out at the end of
Comment No. 20 of Chapter " 2, a complex
# number z = x + iy can be
x y
identified with the matrix . Conversely, every 2 2 real ma-
y x
trix in which the two diagonal elements are equal to each other while
the two non-diagonal elements are negatives of each other represents
a complex number. This association is compatible with addition as
well as multiplication. That is, if z1 = x1 + iy1 and z2 = x2 + iy2 are
any two
" complex # numbers
" then the
# sum of the corresponding matrices,
x1 y1 x2 y2
viz. and corresponds to the complex number
y1 x1 y 2 x2
z1 + z2 and similarly their product corresponds to z1 z2 as one checks
by a direct computation. In particular, this holds for" the square# of a
1 0
complex number. Now, in our problem, the matrix cor-
0 1
responds to the complex number 1 + 0i, i.e. to 1. Regarded as

7
a complex number, it has two square roots, viz. i. Although these
square
" roots are
# complex
" numbers,
# the matrices representing them,
0 1 0 1
viz. and are real matrices and both of them are
1 0 1 0
" #
1 0
square roots of the matrix . Of course, this way we do not
0 1
get all square roots. But we want only one. (The other square roots
of this matrix, obtained from (11) do not correspond to any complex
numbers.)
" #
1 0
There is yet another way to show that the matrix
0 1
has square roots. It is based on the concept of rotation. Take a point
P = (x1 , y1 ) in the plane and let Q = (x2 , y2 ) be its image under a
counterclockwise rotation of the plane through an angle . It is easy
to identify Q using polar coordinates. Write P in the polar form,
viz. (r cos , r sin ). Here is the angle OP makes with the positive
x-axis. (We are tacitly assuming here that r > 0. If r = 0, then
both P and Q degenerate to O.) Then OQ makes an angle +
with the x-axis and therefore, Q in the polar form is simply the point
(r cos( + ), r sin( + )). See the figure below.

Fig. 1a: Rotation of the plane around the origin

Using the trigonometric identities for cos( + ) and sin( + ) and
cancelling r we can write this in a matrix form, viz.
" # " #" #
x2 cos sin x1
= (12)
y2 sin cos y1

8
We denote the 2 2 matrix by A and write this in a more compact
form as
" # " #
x2 x1
= A (13)
y2 y1
Now if we rotate the plane further around O by another angle , the
point Q = (x2 , y2 ) will go to some point R = (x3 , y3 ) and repeating the
reasoning above, we get
" # " #
x3 x2
= A (14)
y3 y2

Putting (13) and (14) together, we get

" # " # " #
x3 x2 x1
= A = A A (15)
y3 y2 y1
But, the net effect of a counterclockwise rotation through followed
by a counterclockwise rotation through (both around the origin O)
is the same as a (counterclockwise) rotation through + around O.
(See the figure below.) So, we also have

Fig. 1b: Composite of two rotations

" # " #
x3 x1
= A+ (16)
y3 y1

Combining (15) and (16) together, we get

" # " #
x1 x1
A+ = A A (17)
y1 y1

9
If this is to hold true for all x1 and y1 , we must have

A+ = A A (18)

(This is a general"property
# of matrices.
" # If B and C are any two 2 2
1 1
matrices, then B and C are simply the first columns of B
0 0
and
" C # respectively.
" # So by taking x1 = 1 and y1 = 0, the equality
x1 x1
B =C implies that the first column of B is the same as
y1 y1
the first column of C. Similarly, by taking x1 = 0 and y1 = 1, we see
that their second columns are equal.)
2
" = in# (18), we get A2 = A . So if we can recognise
If we take
1 0
the matrix as a rotation matrix for some angle, then the
0 1
rotation matrix for half that angle will be a square root of P . This
indeed happens because this matrix is simply A"and so A#/2 is a square
0 1
root of it. If we write out A/2 it comes out as . This is also
1 0
one of the answers we got earlier using the brute force method and then
using complex numbers. Indeed, one may even say "that this is not a#
cos sin
radically new method, because the matrix A =
sin cos
corresponds to the complex number cos + i sin . If we rewrite this
number using complex exponentials as ei then (18) reduces to the
familiar result

ei(+) = ei ei (19)

which is a special case of the general property of exponentials, viz.

ez1 +z2 = ez1 ez2 (20)

One can even go further and derive (18) directly by merely expanding
cos( + ) and sin( + ). Thus it may seem that once we partitioned
the matrix P as in (1), we could get the answer without recourse to
complex numbers or to rotations. The advantage of rotations will be
apparent if we could not think of partitioning P and worked with it as a
33 matrix. Matrices of order 3 do not correspond to complex numbers.

10
But the concept of a rotation makes sense even in the three dimensional
space IR3 , except that we now talk of rotations around an axis through
the origin rather than rotations around the origin. Consider a rotation
of IR3 around the z-axis through an angle , which is counterclockwise
as seen from a point on the positive z-axis. Then points in a plane of
the form z = c are rotated among themselves because if Q is the image
of a point P , then their z-coordinates will be the same. The effect on
the x- and y- coordinates will be the same as that of a rotation of the
xy-plane around O through the angle . (See the figure below.)

Fig. 2: Rotation around the z-axis through angle

So, if P = (x1 , y1 , z1 ) and Q = (x2 , y2 , z2 ), then instead of (12) we
shall have

x2 cos sin 0 x1
y2 = sin cos 0 y1 (21)


z2 0 0 1 z1
and hence by the same reasoning as used above the 3 3 matrix in the
cos(/2) sin(/2) 0
middle will be the square of the matrix sin(/2) cos(/2) 0 .
0 0 1
The matrix P in (D) in our problem represents a rotation around the
x-axis through the angle and hence has a square root. In both these
examples, the 3 3 matrices could be easily partitioned. Part (D)
could
have been made a little more challenging by giving the matrix
1 0 0
0 1 0 . This matrix cannot be conveniently partitioned and

0 0 1
so the earlier methods will not apply. But this matrix represents a rota-
tion around the y-axis through the angle and therefore is the square

11
of the matrix which

represents the rotation through /2, viz. the ma-
0 0 1
trix 0 1 0 . (Alternately, through an interchange of the first

1 0 0

1 0 0
two rows and then the first two columns, the matrix 0 1 0

0 0 1

1 0 0
transforms into the matrix 0 1 0 for which we already found

0 0 1
a square root. By interchanging the two rows and then the two columns
this square root can be converted into a square root of the given ma-
trix. The point to note is that in a visual partition the submatrices
occurring as entries consist of blocks of adjacent rows and columns.
But in a mental partition, this need not be so. A mental partition
can be converted to a visual one
by suitably shuffling
the rows and

1 0 0 1 0 0 1 0 0
columns. Out of 0 1 0 , 0 1 0 and 0 1 0 ,

0 0 1 0 0 1 0 0 1
no matter which matrix we take, one of the three coordinate axes sits
idle and the real action takes place in the plane spanned by the other
two axes. So, although these matrices are visually different, mentally
they are the same.)
In all these examples, the axis of rotation was one of the coordinate
axes. The formula for the matrix of a rotation through an angle
around an axis L through O which is parallel to a unit vector ai+bj+ck
is fairly complicated. This rotation is shown in the figure below.

12
 We leave it as a challenging exercise to show that it is the matrix

a2 (1 cos ) + cos ab(1 cos ) c sin ac(1 cos ) + b sin

ab(1 cos ) + c sin b2 (1 cos ) + cos bc(1 cos ) a sin

ac(1 cos ) b sin bc(1 cos ) + a sin c2 (1 cos ) + cos

Q.38 If a chord, which is not a tangent, of the parabola y 2 = 16x has the
equation 2x + y = p and midpoint (h, k), then which of the following
is(are) possible value(s) of p, h and k ?

(A) p = 5, h = 4, k = 3 (B) p = 1, h = 1, k = 3
(C) p = 2, h = 2, k = 4 (D) p = 2, h = 3, k = 4

Answer and Comments: (D). There could hardly be a problem

which demands less thinking from the candidates. The appearance of
the words parabola, chord, mid-point and equation is enough to
tell even a dumb search engine that what is needed in the solution is
the formula for the chord of a parabola in terms of its midpoint. And
for those who remember this formula (often in the form T = S1 , which
is applicable to any conic), the problem is a cakewalk. In the present
case, this equation comes out to be

ky 8(x + h) = k 2 16h (1)

i.e.

8x + ky = k 2 8h (2)

If this is to be the same line as 2x + y = p, we must have

8 k k 2 8h
= = (3)
2 1 p
There are only two equations here in the three unknowns h, k and p.
So, the system cannot be solved uniquely. Nor does the question ask
for it. It only asks in which of the four given cases these equations hold
true. The first equation gives k = 4. That immediately rules out (A)
and (B). With k = 4, the second equation becomes p = 2h 4. This
is satisfied only in (D).

13
 The paper setters have also given that the chord is not a tangent.
This is a useless piece of information. Normally, by a chord one means
a line passing through two distinct points on a curve and a tangent is
the limiting position of a chord when these two points are infinitesi-
mally close. So, a chord of a conic can never be a tangent. For more
complicated curves, such as graphs of quartics (e.g. y = x4 x2 ), it may
happen that some chord also touches the curve at some other point.
One fails to see what is gained by this useless rider, except possibly to
make a discerning student waste some precious time.
Overall, the problem hardly belongs to an advanced test.

Q.39 Let a, b, x and y be real numbers such that a b ! = 1 and y 6= 0. If the

az + b
complex number z = x + iy satisfies Im = y, then which of
z+1
the following is(are) possible value(s) of x?
2

(A) 1
1 2y 1 + y2
(B) 1 +
(C) 1 1 + y (D) 1 + 1 y 2

Answer and Comments: (A,D). In the problem, x, y (along with

a and b) are some real numbers and they are given to satisfy some
conditions. These conditions do not really give the value of x as a
number. All that they possibly
imply is some relationship
between x
2 2
and y such as x = 1 1 y in (A), x = 1 + 1 + y in (B) and
so on. So a better wording for the question would have been to give
these four relationships as the four options and to ask which of them
hold(s) true. But this is a fine point of diction and the given wording
is not likely to confuse anybody.
Now, coming to the solution, the most straightforward approach
az + b
would be to express the imaginary part of in terms of x, y, a, b
z+1
and equate it with y. Writing z as x + iy and multiplying both
the numerator and the denominator by the complex conjugate of the
az + b
denominator, viz. (x + 1) iy, the ratio comes out to be
z+1
(ax + b) + iay)(x + 1 iy)
. As the denominator is real, the imagi-
(x + 1)2 + y 2

14
 ay(x + 1) (ax + b)y
nary part of this ratio is simply which simpli-
(x + 1)2 + y 2
(a b)y y
fies to 2 2
and further to since it is given that
(x + 1) + y (x + 1)2 + y 2
a b = 1. Equating this with y and using that y 6= 0 we get that
(x + 1)2 + y 2 = 1 from which we conclude that x = 1 1 y 2 .
Although this solution is quite simple (because of the given rela-
tionship between a and b), there is a slightly more elegant way of ap-
proaching the problem. The R.H.S., viz. y, is also the imaginary part
of z. So the data is equivalent to saying that the imaginary part of the
az + b (az + b) z(z + 1)
difference z is zero, or equivalently, that
z+1 z+1
z2
is a real number. Because a = b + 1, this ratio simplifies to b .
z+1
z2
But b is given to be real. So, we get that the ratio is real. Note
z+1
that we have now got rid of both a and b. We further multiply both
the numerator and the denominator by the complex conjugate of the
z 2 (z + 1)
denominator to get that is real. As the denominator is real,
|z + 1|2
this means that the numerator, viz. z 2 (z + 1) is real, i.e. its imaginary
part is 0. It is only at this stage that we write z as x + iy. Doing so,
we get that the imaginary part of (x2 y 2 + 2ixy)(x + 1 iy) is 0. This
gives 2xy(x + 1) = y(x2 y 2 ) and hence 2x(x + 1) = x2 y 2 as y 6= 0.
So, finally we get that (x + 1)2 + y 2 = 1 as before.
The advantage of this alternate solution is that it begins by in-
terpreting the data as saying that a certain number is real. We then
go on simplifying this number, subtracting some real number (viz. b)
from it and multiplying it by a real number (viz. |z + 1|2 ). In all these
operations, the number goes on changing but remains real. So, instead
of working with the imaginary part of a complicated number, we are
reduced to working with the imaginary part of a relatively simple num-
ber. In terms of time there may not be much saving, because whatever
be the disadvantages of the straightforward approach, one undeniable
advantage is that it does not waste any time thinking of any alter-
natives. And in a severly time constrained examination, this is quite
important. In the good old days, when the candidates had to show
their work, such occasional elegant deviations from the straightforward

15
 approach used to give refreshing breezes in the otherwise tiring job of
evaluation.
1 1
Q.40 Let X and Y be two events such that P (X) = , P (X|Y ) = and
3 2
2
P (Y |X) = . Then
5
1 1
(A) P (X 0 |Y ) = (B) P (X Y ) =
2 5
2 4
(C) P (X Y ) = (D) P (Y ) =
5 15

Answer and Comments: (A,D). (A) follows directly from P (X|Y ) =

1
because the events X|Y and X 0 |Y are always mutually complemen-
2
tary. To tackle the remaining options, we translate the conditional
probabilities as ratios and get
P (X Y ) 1
= (1)
P (Y ) 2
P (Y X) 2
and = (2)
P (X) 5
1
Since X Y = Y X, these two equations along with P (X) = yield
3
2 2 1 4
P (Y ) = 2P (X Y ) = 2 P (X) = 2 = (3)
5 5 3 15
Thus we see that (D) is also true. In this derivation, we already got
P (X Y ) = 52 P (X) = 15
2
which shows that (B) is false. Finally, to get
P (X Y ) we apply the relation
P (X Y ) = P (X) + P (Y ) P (X Y ) (4)
As all the terms on the R.H.S. are known, we get
1 4 2 7
P (X Y ) = + = (5)
3 15 15 15
So (C) is false too.
A very simple problem on conditional probabilities. Again one
wonders if it is suitable for an advanced test.

16
Q.41 Let [x] be the greatest integer less than or equal to x. Then, at which
of the following point(s) the function f (x) = x cos((x + [x])) is dis-
continuous?

(A) x = 1 (B) x = 0 (C) x = 2 (D) x = 1

Answer and Comments: (A,C,D). f (x) is the product of two func-

tions. The first factor x is continuous everywhere. The second factor
cos(x + [x]) equals cos x if [x] is even and cos(x) if [x] is odd.
When x is an integer, [x] has different parity in the intervals [n 1, n)
and [n, n + 1). (That is, it is odd in one and even in the other.) As
a result f (x) = x cos x on one side of n and f (x) = x cos x on
the other side. As cos n 6= 0, the left and right handed limits of
cos(x + [x]) as x n will be negatives of each other. So, the
second factor cos(x + [x]) will be discontinuous at every integer n.
However, when n = 0, because of the first factor, both sided limits of
x cos(x + [x]) will be 0 and so f (x) is continuous at 0.
A JEE paper without any question containing the integer part
function [x] is like Hamlet without the prince of Denmark! As a result,
it is very difficult for the examiners to come up with radically new
questions. The only catch in the present problem is that even though
the second factor of f (x) is discontinuous at every integer n, at n = 0,
this discontinuity gets masked because of the factor x which tends to
0. But even this feature has been tested several times.
x2 y 2
Q.42 If 2x y + 1 = 0 is tangent to the hyperbola 2 = 1, then which
a 16
of the following CANNOT be the sides of a right angled triangle?

(A) 2a, 4, 1 (B) 2a, 8, 1 (C) a, 4, 1 (D) a, 4, 2

Answer and Comments: (B,C,D). The paper-setters have been a

little less kind to the candidates in this question than in Q.38 above.
In that question, even the dumbest candidate would realise that the
solution would need the equation of a chord in terms of its midpoint.
In the present problem, it takes some infinitesimally small intelligence
to realise that the solution would need the condition of tangency of a
line to a hyperbola, because these specific words do not appear in the
statement of the problem!

17
 So, assuming that the candidate has an iota of intelligence and
an iota of memory to remember the condition for tangency, the line
y = mx + c is a tangent to the hyperbola if and only if c2 = a2 m2 16.
(Lack of intelligence is not curable. But the lack of memory is. Put
x2 (mx + c)2
y = mx+c in the equation of the hyperbola to get 2 = 1.
a 16
Then take the condition for this quadratic in x to have equal roots.)
In the present problem, m = 2 and c = 1. So, we get

4a2 16 = 1 (1)

17
which gives a = . (It is inherent in the equation of a hyperbola
2
in the standard form that a, being the length of the major axis, is
positive. Anyway, a negative value of a is also untenable because there
can never be a triangle in which one side is negative.)
From this point on, the problem has nothing to do with the hy-
perbola. All we have to do is to see which of the four given triples
satisfy the condition for a right angled triangle, viz. that the sum of
the squares of two of them equals the square of the third. We do this
one by one.
In (A), the squares are 17, 16 and 1. Clearly, this is a right angled
triangle.
In (B), the squares are 17, 64 and 1. This cannot be a right angled
triangle. (In fact, it cant even be a triangle as the sum of 17 and 1
is less than 8. But that is not asked.)
In (C), the squares are 17/4, 16 and 1. No two of them can add up to
the third for the simple reason that two of them are integers and the
third one is not.
In (D), too the squares are 17/4, 16 and 4. So same reasoning as (C)
works.
Another question which is a total misfit in an advanced test. When
a question is basically trivial, little improvement is possible. But one
wishes that at least the
some of the three numbers in the given triplets
were expressions like a 1 or 1 + a2 + 1. So surds, which are no
longer in vogue in JEE, could have been paid a lip service.

18
Q.43 Let f : IR (0, 1) be a continuous function. Then, which of the
following function(s) has(have) the value zero at some point in the
interval (0, 1)?

Rx
(A) ex f (t) sin t dt (B) x9 f (x)
0
/2 /2x
(D) x
R R
(C) f (x) + f (t) sin t dt f (t) cos t dt
0 0

Answer and Comments: (B,D). There are two standard ways of

showing that a given continuous function g(x) has at least one zero in
an interval (a, b). The simpler one is to show that g(a) and g(b) are of
opposite signs and apply the Intermediate Value Property (IVP). The
other is to find an antiderivative G(x) for g(x) for which G(a) = G(b)
and apply Rolles theorem.
In the present problem, in each case we denote the function by g(x).
Then g(x) is continuous in all four options (A) to (D) as f (x) is given
to be continuous. (In (A) and (D), g(x) is also differentiable. But that
may not hold in (B) and (C).) So, we try the simpler approach first,
viz. to apply the IVP to g(x) on the interval [0, 1].
In (A), g(x) > 0 for all x (0, 1) because ex > 1 for x > 0 while the
Rx
integral f (t) sin t dt is at most 1 because the interval of integration
0
has length less than 1 and the integrand is less than 1 throughout this
interval. Hence g(x) has no zeros in (0, 1).
In (B), g(0) = 0 f (0) < 0 since f (0) > 0 while g(1) = 1 f (1) > 0
since f (x) < 1 for all x. So a direct application of IVP gives that g(x)
has at least one zero in (0, 1).
In (C), the integrand is positive throughout (0, /2) and so the
integral is positive too. As f (x) is also positive for all x, g(x) > 0 for
all x and hence has no zeros anywhere on IR.
/2
Finally, in (D), g(0) =
R
f (t) cos t dt < 0 because the integrand
0
is positive on the interval of integration except at the end point /2.
/21
On the other hand, g(1) = 1
R
f (t) cos t dt > 0 because once again
0

19
 the interval of integration has length less than 1 (here we are using that
< 4 so that /2 1 < 1) and the integrand is at most 1 throughout.
So by IVP, g(x) has at least one zero in (0, 1).
A simple, but highly repetitious problem, requiring the same rea-
soning in the estimation of all three integrals.

SECTION 2 : (Maximum Marks : 15)

This section contains FIVE questions.
The answer to each question is a SINGLE DIGIT INTEGER ranging from
0 to 9, both inclusive.
Marking Scheme:
Full Marks : +3 If only the bubble corresponding to the correct answer is
darkened.
Zero Marks : 0 In all other cases.
Q.44 The sides of a right angled triangle are in an arithmetic progression. If
the triangle has area 24, then what is the length of its smallest side?

Answer and Comments: 6. The information that the answer is

an integer from 0 to 9 gives a sneaky way to solve the problem. Let
the sides be a, b, c with a b < c. Then the area is 24 which means
ab = 48. The only factorisation of 48 with single digit factors gives
a = 6 and b = 8. This makes c = 10 because of the A.P. requirement.
And that indeed gives a right angled triangle. Bingo.
For an honest solution, let the sides be as above, Then we have
c = 2b a and also c2 = a2 + b2 , i.e. (2b a)2 = a2 + b2 . This simplifies
to 3b2 = 4ab = 4 48 because ab = 48 as already noted. So we get
b2 = 64 whence b = 8 and a = 6.
Either way, the problem is simple. Such cute little problems are
suitable as newspaper riddles or for a school scholarship examination
but not for JEE, much less for advanced JEE.
Q.45 For how many values of p does the circle x2 + y 2 + 2x + 4y p = 0 and
the coordinate axes have exactly three points in common?

Answer and Comments: 2. The given equation represents a family

of concentric circles all centred at the point M = (1, 2). Distinct

20
values of p give distinct circles. (For p < 5 there is no circle while
for p = 5 the circle degenerates into a single point.) So the question
amounts to asking how many members of this family of circles have
exactly three points in common with the coordinate axes. Now, given
a line L and a circle C of radius r centred at a fixed point M at a
perpendicular distance d from L, three things happen depending upon
whether r < d, r = d or r > d. In the first case C has no point
in common with L, in the second it touches L (i.e. has one point in
common with it) and in the third it cuts L in two distinct points.
The solution is based on this simple idea. The x- and y-axis
have one point in common, the origin. Consider a typical circle C :
(x + 1)2 + (y + 2)2 = r2 of this family. The only way C will have
three points in common with these two lines is when it touches one of
them and cuts the other in two distinct points or when it cuts both
the lines in two distinct points each, of which the origin is one. The
perpendicular distances of the centre (1, 2) from the two axes are
2 and 1. For r < 1, C will have no point in common with either axes.
For r = 1, C will touch the y-axis but will not meet the x-axis. For
1 < r < 2, C will meet the y-axis in two points but not meet the x-
axis. For r = 2, it will cut the y-axis in two distinct points and touch
the x-axis. So this is one of the possibilities we are looking for. For
r > 2, C will meet each axis in two distinct points and the only way
one of these will be common is when C passes through O, i.e. when
r = 5. Hence there are only two circles in the family which meet the
requirement. They are shown in (a) and (b) of the figure below.

y
y

O x O x

. .M
M

(a) r = 2 (b) r = 5

Fig. 4: Intersection of a variable circle with the axes

21
 This is a good problem. Once the essential idea is understood,
the calculations are minimal. Actually, we need not have identified the
two circles which meet the requirement. The question merely asks how
many of them are there and that could have been answered simply by
studying how the behaviour of C changes as r increases. The reasoning
would have worked equally well if instead of the two axes, we had any
pair of intersecting lines. So, this problem rewards a candidate who has
the ability to get the gist without getting entangled into calculations.

Q.46 For a real number , if the system

1 2

x 1
1 y = 1

2 1 z 1

of linear equations, has infinitely many solutions, then 1 + + 2 =

Answer and Comments: 1. Let be the determinant of the coef-

ficient matrix of the system. For the system to have infinitely many
solutions must vanish and the system be consistent (i.e. has at least
one solution).
It is clear that if = 1 then vanishes as all three rows become iden-
tical. So 1 and indeed ( 1)2 will be a factor of when written as
a polynomial in . However, for = 1, the first two equations becomes
x + y + z = 1 and x + y + z = 1 which are mutually inconsistent. So,
we have to look for other values of which will make vanish and for
this the best way is to expand directly.

= 1(1 2 ) ( 3 ) + 2 (2 2 )
= (1 2 )(1 2 ) (1)

So = 0 for = 1 or = 1. The case = 1 was already rejected

above. For = 1 all the three equations become x y + z = 1
which has at least one solution, e.g. x = 1, y = 0, z = 0. (Actually,
the solution set is a plane in IR3 , but that is not needed here. It is
important to note that mere vanishing of the three determinants that
appear as numerators in the Cramers rule does not imply consistency.
One has to produce an actual solution.)

22
 So the only value of which meets the requirement is = 1. With
this value, 1 + + 2 equals 1.
An absolutely straightforward problem. Not suitable for an advanced
test. But the paper setters are under a compulsion to cover all areas of
the syllabus. And at the JEE level, not much variety of questions about
systems of linear equations is possible because the syllabus requires that
the number of variables be not more than 3.

Q.47 Words of length 10 are formed using the letters A, B, C, D, E, F, G,

H, I, J, Let x be the number of such words where no letter is repeated
and y be the number of such words where exactly one letter is repeated
y
twice and no other letter is repeated. Then =
9x

Answer and Comments: 5. The counting of x is standard. It is the

number of permutations of ten distinct symbols without repetition. So

x = 10! (1)

The counting of y requires some thought. The letter to be repeated

can be chosen in 10 distinct ways and for each such choice, the number
of permissible words formed will be the same. So

y = 10z (2)

where z is the number of permissible words in which one particular

letter, say A, is repeated exactly twice but no other letter is repeated.
So, the 8 places not occupied by A are to be filled in by 8 distinct
letters from the remaining 9 letters. As a result, some letter does not
appear in such words and this absent letter can be chosen in 9 ways.
Without loss of generality let this letter be B. Then

z = 9w (3)

where w is the number of words of length 10 formed by A, C, D, E,

F, G, H, I, J in which A occurs twice and every other letter appears
once. This is also a standard count. It is the number of permutations
with specified repetitions of the symbols. More generally, if there are
mi objects of type i for i = 1, 2, . . . , k, then the total number of their

23
 (m1 + m2 + . . . + mk )!
permutations is . In the present case k = 9,
m1 !m2 ! . . . mk !
m1 = 2 and mi = 1 for i = 2, 3, . . . , 9. So,
10! 10!
w= = (4)
2! 2
Putting (2), (3) and (4) together,

y = 45 10! (5)

y 45 10!
So we now get = = 5.
9x 9 10!
A simple counting problem. Although the explanation looks long
when written out in full, it does not take much time to do it mentally.
Such problems are therefore ideal for a fill in the blank type. In the
present format, they are asked as multiple choice questions, because the
answer has to be an integer from 0 to 9. And the question sometimes
has to be artificially modified to make it satisfy this requirement. In
the present problem, it would be more logical to ask simply the value
of y. But that is not an integer between 0 and 9.
Efficiency and uniformity of evaluation are two major justifications
given for the multiple choice format. These could still be served by
asking the questions as fill into the blank type where the examiner,
even though human, is not supposed to look into the reasoning. The
nature of the question is often such that it is highly unlikely that a
correct answer would come through a wrong reasoning or by fluke.

Q.48 Let f : IR IR be a differentiable function such that f (0) = 0, f ( ) =
2
3 and f 0 (0) = 1. If
Z /2
g(x) = [f 0 (t)cosec t cot t cosec tf (t)] dt
x

for x (0, /2], then lim g(x) =

x0

Answer and Comments: 2. The trick here is to observe that the

d
integrand f 0 (t)cosec t cot t cosec tf (t) is simply (f (t)cosec t) and
dt

24
 consequently,
/2
g(x) = f (t)cosec t
x
f (/2) f (x) f (x)
= =3 (1)
sin(/2) sin x sin x

(Even
Z those who miss this trick will get (1)Z if they begin by integrating
f 0 (t)cosec t dt by parts as f (t)cosec t + f (t) cot tcosec t dt because
this second integral will cancel.)
It is now easy to complete the solution. It is given that f (0) = 0.
As f (x) is continuous (being differentiable), f (x) 0 as x 0. So
f (x) 0
lim is an indeterminate form of the type. An application of
x0 sin x 0
f (x) f 0 (0)
the lHopitals rule gives lim = = 1 as f 0 (0) is given to be
x0 sin x 1
f (x)
1. Those (if any!) who prefer to avoid lHopitals rule can rewrite
sin x
f (x) f (0) x
as and take the limit of each factor.
x0 sin x
(While we are at this question, a point about notations deserves to
be addressed. It is very common to denote a function by f (x) and its
derivative by f 0 (x). Depending on the resolution of printing, the symbol
f and the prime (0 ) often come very close and sometimes overlap so that
f 0 looks f0 , almost like f when read hurriedly. This can cost a candidate
who is often in a hurry. The JEE papersetters are instructed to reserve
the symbols A, B, C, D for options. That is why a triangle in a JEE
question is denoted by P QR rather than by ABC which is universal
in textbooks. Perhaps they should also be instructed to avoid f as a
symbol for a function in a problem which involves the derivative of f
too.)

SECTION 3 : (Maximum Marks : 18)

This section contains SIX questions of matching type.

This section contains TWO tables (each having 3 columns and 4 rows).
Based on each table, there are THREE questions.
Each question has FOUR options (A), (B), (C) and (D). ONLY ONE of these

25
four options is correct.
For each question, darken the bubble corresponding to the correct option in
the ORS.
For each question, marks will be awarded in one of the following categories :
Marking Scheme:
+3 If only the bubble corresponding to the correct option is darkened. Zero
Marks
0 If none of the bubbles is darkened.
Negative Marks : 1 In all other cases

Column 1, 2 and 3 contain conics, equation of tangents to the conics and

points of contact, respectively.

Column 1 Column 2 Column 3

a 2a
 
2 2 2 2
(I) x + y = a (i) my = m x + a (P) P = ,
m2 m

!
2 2 2 2 ma a
(II) x + a y = a (ii) y = mx + a m2 + 1 (Q) 2 ,
m + 1 m2 + 1
a2 m
!
2 1
(III) y = 4ax (iii) y = mx + a m 1 (R) 2 2
2 2 ,
a m + 1 a2 m2 + 1!
a2 m 1
(IV) x2 a2 y 2 = a2 (iv) y = mx + a2 m2 + 1 (S) 2 2 , 2 2
a m 1 a m 1

Q.49 Thetangent to a suitable conic (Column 1) at ( 3, 1/2) is found to
be 3x + 2y = 4. Then which of the following options is the only
CORRECT combination?

(A) (II) (iii) (R) (B) (IV) (iv) (S)

(C) (IV) (iii) (S) (D) (II) (iv) (R)

Answer and Comments: (D). We are given that the point ( 3, 1/2)
lies on one of the conics in Column 1 and further that the equation
of the tangent to that conic at this point is 3x + 2y = 4. Out of
the four options given, in two of them the conic is (II), i.e. the ellipse
x2 2 x2
+y = 1 while in the other two it is (IV), the hyperbola y 2 = 1.
a2 a2
Depending which of these two holds, the value of a will be different.
3 1
If (II) holds, we get 2 + = 1 which gives a = 2. If (IV) holds, we
a 4
26

3 1 2 3
get 2 = 1 which gives a = . With a = 2, the tangent to
a 4 5
(II) at ( 3, 1/2) would be 3x + 2y = 4 which is consistent with the
2 3
data. However, with a = , the tangent to (IV) at ( 3, 1/2) would
5
6 12
be 3x y = which is not the given line. So the possibilities (B)
5 5
and (C) are ruled out. The reamining two possibilities, viz. (A) and
(D) match with each other in terms of their Column 1 entries as well as
in terms of their Column 3 entries. So, the distinction will have to be
made on the basis of the middle entries,
i.e. those from Column
2. As
3 3
we already know a = 2 and m = , (iii) becomes y = x+ 2
2 2
3
while (iv) becomes y = x + 2, which is indeed the given line. So
2
the correct option is (D). (For an honest answer, one should, of course,
check that the point of contact is indeed (R). But in the present case,
3
this follows by merely substituting a = 2 and m = , because the
2
point of contact is ( 3, 1/2) itself.)

Q.50 If a tangent to a suitable conic (in Column 1) is found to be y = x + 8

and its point of contact is (8, 16), then which of the following options
is the only CORRECT combination?

(A) (III) (i) (P) (B) (III) (ii) (Q)

(C) (II) (iv) (R) (D) (I) (ii) (Q)

Answer and Comments: (A). This is of the same spirit as the last
question. The point (8, 16) is given to lie on one of the four conics in
Column 1. From this we first determine the value of a for all possible
conics in Column 1, except (IV) since it does not appear in any of the
given four options. A direct substitution into the equations in Column
1 gives
that for the point (8, 16) to lie on I and III the value of a should
be 8 5 and 8 respectively, while for II there is no value. So (C) is
ruled out. The equation of the tangent to x2 + y 2 = 320 at (8, 16) is
8x + 16y = 320, i.e. x + 2y = 40. This is inconsistent with the data.
So (D) is also ruled out. (This can also be predicted by observing that

27
 the slope of the tangent at the point (8, 16) to the circle x2 + y 2 = 320
must be negative as the tangent is perpendicular to the radius.) That
leaves only (A) and (B) in both of which the entry from Column 1 is
y 2 = 32x. The equation of the tangent to this curve at this point is
16y = 16(x + 8) i.e.y = x + 8 which tallies with the data. So, let us
now see which entries in Columns 2 and 3 are consistent with a = 32
and m = 1. Luckily, the very first entry (i) is so. Therefore, if only one
combination is correct, it has to be (A). Again, for the sake of honesty
(often not the best policy at the present JEE!), we see that with a = 8
and m = 1 (P) in Column 3 comes as (8, 16), which is indeed the point
of contact.

Q.51 For a = 2, if a tangent is drawn to a suitable conic (Column 1) at
the point of contact (1, 1), then which of the following options is the
only CORRECT combination for obtaining its equation?

(A) (II) (ii) (Q) (B) (III)(i)(P)

(C) (I) (i) (P) (D) (I) (ii) (Q)

Answer and Comments: (D). This time we are given the value of
a. As a result, all the conics in Column 1 are
now known. They are (I)
x + y = 2, (II) x + 2y = 2, (III) y = 4 2x and (IV) x 2y 2 = 2.
2 2 2 2 2 2

By a direct substitution we see that the point (1, 1) lies only on (I).
So the correct combination has to be either (C) or (D). The equation
of the tangent to x2 + 2
y = 2 at (1, 1) is x + y = 2, i.e. y = x + 2.
So, we now have a = 2 and m = 1. This matches with (ii) of Column
2 and (Q) of Column 3.
The essential idea in all three questions is merely to find the
equation of the tangent at a given point on a conic. The paper-setters
seem to have an obsession with tangents to conics. Already Q. 42 is
based on the condition for tangency of a given line with a hyperbola.
And now we have three questions each of which involving a tangent to
some conic (listed in Column 1). As if this were not enough, in Q. 38
the paper-setters have specified that the chord there is not a tangent,
a totally useless specification as commented there. It is really shocking
that out of the total 61 marks for the mathematics section of Paper 1,

28
 as many as 13 marks be allotted for a single topic, which is inherently
elementary and hence suitable only for the screening round.
Although questions asking to match entries in two columns have
been asked in JEE many times, what we have here is matching entries
in three columns. Questions like this have not been asked in the JEE
before and chances are that many students will struggle hard to under-
stand exactly what is asked even though the questions themselves are
inherently trivial and based on a common single idea. It would have
been far better to ask three separate questions. Artificially giving them
a common set up only serves to make them clumsy. One is reminded
of a popular puzzle in which a train carrying some men and women as
passengers enters a tunnel. In the dark of the tunnel, the sounds of a
kiss, a shriek and a slap are heard. Given some information about the
passengers, you are asked to find out who kissed whom, who slapped
whom and who shrieked! Good recreation, possibly, but not good in
an examination meant to test some other and more serious kind of
intelligence.

Preamble for Q.52 to 54

Let f (x) = x + loge x x loge x, x (0, ).

Q. 52 to 54 are based on the following table in which

Column 1 contains information about the zeros of f (x), f 0 (x) and f 00 (x).
Column 2 contains information about the limiting behaviour of f (x), f 0 (x)
and f 00 (x) at infinity.
Column 3 contains information about increasing/decreasing nature of f (x)
and f 0 (x).

Column 1 Column 2 Column 3

(I) f (x) = 0 for some x (1, e2 ) (i) x
lim f (x) = 0 (P) f is increasing in (0, 1)
(II) f 0 (x) = 0 for some x (1, e) (ii) lim f (x) = (Q) f is decreasing in (e, e2 )
x
(III) f 0 (x) = 0 for some x (0, 1) lim f 0 (x) =
(iii) x (R) f 0 is increasing in (0, 1)
(IV) f 00 (x) = 0 for some x (1, e) (iv) lim f 00 (x) = 0 (S) f is decreasing in (e, e2 )
x

29
Q.52 Which of the following options is the only CORRECT combination ?

(A) (IV) (i) (S) (B) (I) (ii) (R)

(C) (III) (iv) (P) (D) (II) (iii) (S)

Q.53 Which of the following options is the only CORRECT combination ?

(A) (III) (iii) (R) (B) (I) (i) (P)

(C) (IV) (iv) (S) (D) (II) (ii) (Q)

Q.54 Which of the following options is the only INCORRECT combination?

(A) (II) (iii) (P) (B) (II) (iv) (Q)

(C) (I) (iii) (P) (D) (III) (i) (R)

Answers and Comments: (D) for all.

Nowadays, it is more common to denote the natural logarithm loge x by
ln x and we shall do so. The word information in the preamble is misleading.
The problems deal with a particular given function f (x). No additional
information can be given about it. It would have been better to say that each
column contains some statements about f (x), some of which will be true and
the others false. And the problems in essence ask which of these are true.
But instead of asking this for each of the 12 statements, in each question,
four combinations of three statements, one coming from each column, are
given. The problem is to decide in which of these four combinations, all the
three selected statements are true.
It is difficult to understand what advantage arises by this novel posing of
the problems. It would be far better to ask three separate questions, one for
each column, asking which of the four statements in it are true about the
given function f (x). Once this is answered, all the three questions can be
answered by inspection.
So we shall prepare a table listing if the 12 given statements are true or
false. Generally most questions of the types here about a function can be
answered simply by an inspection of a well drawn graph of the function. But
that is not of much help unless the function happens to be a very standard
function (such as ex , ln x or the standard trigonometric functions) or some
modification of it. In all other cases, we have to do considerable spade

30
work to draw a truly revealing graph of the function. And this spadework
consists largely of answering the types of questions that are given in the three
columns. So, there is really no short cut, except possibly when a candidate
has already seen the graph of this or a substantially similar function or, by
luck, happens to be seated so close to an intelligent candidate as be able
to peek at his/her rough work! Usually, candidates are seated far enough
to make it next to impossible for them to see the texts of each other, but
figures can be seen and their crucial features grasped, at least qualitatively,
even from a longer distance.
We begin with the statements in Column 1. As pointed out in the
comments on Q.43, there are two standard ways of showing that a given
function has at least one zero in an interval. One is based on the IVP
(Intermediate Value Property) and the other on Rolles theorem. Since
f (x) = x + ln x x ln x, we have
1 1
f 0 (x) = 1 + ln x 1 = ln x (1)
x x
1 1
and f 00 (x) = 2 (2)
x x
Also f (1) = 1 and f (e2 ) = e2 + 2 2e2 = 2 e2 < 0 (since e > 2). So by
the IVP, we see that (I) is true. Further f (e) = e + 1 e = 1 = f (0). So,
by Rolles theorem (II) is also true. From (2), f 00 (x) < 0 for all x > 0 and
so (IV) is false. Further, by Lagrange Mean Value theorem, f 0 (x) is strictly
decreasing on (0, ). But by (1) f 0 (1) = 1. So, f 0 (x) > 1 for all x < 1.
Hence (III) cannot be true either. Summing up, (I) and (II) are true while
(III) and (IV) are false.
Coming to the second column, it is obvious from (1) that f 0 (x) as
x (because ln x as x ). Truth of (iv) is even more obvious.
But (i) and (ii) are not so easy to dispose off. Of course, both cannot be
true. A careful look at f (x) = x + ln x x ln x shows that its limit as x
is of the type. In such cases what matters is which term tends to
ln x
more rapidly. Let us write f (x) as xg(x) where g(x) = 1 + ln x. Then
x
the middle term tends to 0 as x as can be seen by applying lHopitals
rule. Hence g(x) as x . Therefore f (x) = xg(x) also tends to
as x . So, (ii) is true and (i) false.
Finally, in Column 3, we already observed that f 0 (x) is strictly decreasing
on (0, ). So (R) is false and (S) is true. As for (P), ln x < 0 for x (0, 1)

31
and so by (1) f 0 (x) > 0 for all x (0, 1). Hence by Lagranges MVT again,
(P) is true. For (Q), too, we check the sign of f 0 (x) on (e, e2 ). By (1), we
1
have f 0 (e) = 1 < 0 since e > 1. As f 0 is strictly decreasing on (0, ),
e
f 0 (x) < f 0 (e) < 0 for all x > e. Hence f 0 (x) < 0 for all x (e, e2 ). So again
by LMVT, (Q) is true.
The results of this analysis are summarised in the table below.
Column 1 Column 2 Column 3
(I) True (i) False (P) True
(II) True (ii) True (Q) True
(III) False (iii) True (R) False
(IV) False (iv) True (S) True
All three questions 52 to 54 can now be answered instantaneously by look-
ing at this table. (However, that leaves open the possibility that somebody
who has prepared a correct table hastily marks some wrong option, like a
passenger with a valid ticket for one train catching a wrong train in a hurry.
The ability to prepare a table like the above is mathematical. The ability
to look at the table (possibly prepared by someone else) and answer the
three questions correctly is essentially a clerical matter requiring little more
than presence of mind. It would be ironical if a candidate who has the first
ability fumbles later and thereby gets a negative credit. Of course, this dan-
ger is inherent in every multiple choice question. But it increases multifold
when you have to simultaneously look at the entries of three columns. So,
whatever be the creative accomplishment on the part of the paper setters
in designing such tripartite matching questions, it would be far better to
form a paragraph (another creative genius active at JEE) with the function
f (x) = x + ln x x ln x as its preamble and ask three separate questions,
each corresponding to one of the three columns.
We remarked that all questions could have been answered almost instanta-
neously if we had a good graph of the function f (x). Now with the spadework
done above we are in a position to draw such a graph. From (2) it follows
that the graph of f (x) is concave downward on (0, ). So qualitatively it
will look like the graph of an inverted parabola. But it is not symmetric
about any vertical line. From (1) we see that f 0 (x) is positive and very
large for x (0, 1). So the graph increases very steeply in (0, 1). Even af-
ter 1, f (x) continues to rise but not for very long because because although
1
f 0 (1) = 1 > 0, f 0 (e) = 1 < 0. So somewhere in the interval (1, e), f (x)
e
32
has a critical point and changes its behaviour from increasing to decreas-
ing. The determination of this critical point would require the solution of
1
ln x = 0. This can be done only approximately. For example, from e < 4,
x
1
we see that f 0 (2) = ln 2 < 0 and so the critical point is somewhere in
2
(1, 2). After that the function decreases increasingly rapidly because f 0 (x) is
strictly decreasing and tends to as x . So, there are no asymptotes
except the negative y-axis.
With this analysis we are in a position to sketch the graph of f (x). It is
shown in the figure below along with the graphs of f 0 (x) and f 00 (x) too. This
figure was prepared using GeoGebra. (Note that even though the question
deals with f (x) only for x > 0, the formula for f 00 (x) in (2) makes sense even
1 1
for x < 0. So the figure shows the graph of y = 2 for x < 0 too.)
x x

Fig. 5: Graph of y = x + ln x x ln x

33
 PAPER 2

Contents

Section - 1 (Only One Correct Type) 34

Section - 2 (One or More Correct Type) 44

Section - 3 (Paragraph Type) 64

SECTION I : (Maximum Marks : 21)

This section contains SEVEN questions.
Each question has FOUR options (A), (B), (C) and (D). ONLY ONE of these
four options is correct.
For each question, darken the bubble corresponding to the correct option in
the ORS.
For each question, marks will be awarded in one of the following categories :
+3 If only the bubble corresponding to the correct option is darkened.
0 If none of the bubbles is darkened.
1 In all other cases.

Q.37 Three randomly chosen nonnegative integers x, y and z are found to

satisfy the equation x + y + z = 10. Then the probability that z is
even, is
36 6 5 1
(A) (B) (C) (D)
55 11 11 2

Answer and Comments: (B). A standard probability problem. The

non-negativity restriction implies that none of the numbers can exceed
10. So the sample space, say S, is the set
S = {(x, y, z) : x, y, z {0, 1, 2, . . . , 10}, x + y + z = 10} (1)
Every triplet (x, y, z) is determined uniquely by writing 10 dots in a
row and inserting two walls anywhere between them (including the two
ends of the row).
. .
.. ..

34
 so that x is the number of dots before the first wall, y is the number of
dots between the two walls and z is the number of dots after the second
wall. (In the illustration, x = 3, y = 5 and z = 2. Placing both the
walls between two consecutive dots gives an arrangement where y = 0
while placing both of them before the first dot gives x = 0, y = 0 and
z = 10.)
To find |S|, note that in all there are 12 objects in a row, 10 identical
dots
  and 2 identical walls. So the number of all such arrangements is
12
2
= 66. hence

|S| = 66 (2)

Another way to look at this is to let a = x + y + 1. Then x and a

determine x, y, z uniquely. Also 0 x< a 11. So x and a can be
chosen from the set {0, 1, 2, . . . , 11} in 12
2
ways.
We now classify S into 11 mutually disjoint subsets S0 , S1 , . . . , S9
and S10 where for each k = 0, 1, . . . , 10, Sk consists of those arrange-
ments in which z = k. In every such arrangement there are 10 k
dots before the second wall and it is uniquely determined by the lo-
cation of the first wall, which can be anywhere in the row of these
10 k dots including the two ends. So there are 11 k elements
in Sk for k = 0, 1, 2, . . . , 10. (This is confirmed by the fact that
10 11
X X 11 12
(11 k) = r= = 66 = |S|.)
k=0 r=1 2
Now the favourable set is T = S0 S2 S4 S6 S8 S10 . So

|T | = 11 + 9 + 7 + 5 + 3 + 1 = 36 (3)
36 6
Hence the desired probability is the ratio 66
= 11
.
A simple problem. The work is somewhat repetitious. Some
variation could have been introduced by asking the probability that all
three x, y, z are different or that they are the sides of a triangle etc.

Q.38 Let S = {1, 2, 3, . . . , 9}. For k = 1, 2, . . . , 5, let Nk be the number of

subsets of S, each containing five elements, out of which exactly k are
odd. Then N1 + N2 + N3 + N4 + N5 =
(A) 125 (B) 252 (C) 210 (D) 126

35
Answer and Comments: (D). Decompose the set S into two subsets
A = {1, 3, 5, 7, 9} and B = {2, 4, 6, 8}. Then, each Nk consists of all
subsets C of S for which |C| = 5 and |C A| = k. Every such subset
is obtained by taking exactly k elements from A and remaining 5 k
elements from the complement B. So, we have

! !
5 4
Nk = (1)
k 5k
for k = 1, 2, 3, 4, 5. As the numbers involved are small, we can get the
answer by directly calculating each Nk and adding. Calling the sum as
N,
N = N1 + N2 + N3 + N4 + N5
= 5 1 + 10 4 + 10 6 + 5 4 + 1 1
= 126 (2)
5
! !
X 5 4
Another solution is to evaluate the sum using a bi-
k=1 k 5k
! !
5 4
nomial identity. If we add one more term (which is 0 any-
0 5
5
! !
X 5 4
way), the resulting sum, viz. is precisely the coeffi-
k=0 k 5k
cient of !x5 in the 5 4 9
! product (1 + x) (1 + x) , i.e. in (1 + x) . Hence
9 9 9876
N= = = = 63 2 = 126, the same answer as
5 4 24
before.
In the present problem, however, even the use of the binomial
identity can be avoided. The question does not ask for the numbers
N1 , N2 , . . . , N5 individually but only for their sum. As we already saw,
their sum is the number of subsets C of S = {1, 2, . . . , 9} having 5
elements for which C A has 1, 2, 3, 4 or 5 elements. But since the
complement B of A in the set S has only four elements, every subset
C of S having five elements must contain at least one element from
A. So the answer is the number of all subsets of S having 5 elements.
The restriction that!it has 1, 2, 3, 4 or 5 elements from A is superfluous.
9
Therefore, N = = 126 as before.
5

36
 Problems based on binomial identities are rare in the Multiple Choice
Format. The present problem would have been one such problem had
there not been this short cut. Actually, it is not exactly a short cut.
Rather it is a combinatorial proof of the binomial identity. Suppose,
more generally, that we have a set S of people in which there are m men
and n women. We want to form a committee of k persons from these
people. Here m, n, k are some positive integers. For r = 0, 1, 2, . . . , k,
let Nr be the number of such committees in which there are ! r men and
k
!
X m n
kr women. Then the total number of committees is .
r=0 r kr
But, on the other hand, we also have the total number ! of committees
m+n
of k persons from m + n pesrons as simply . So we have a
k
combinatorial proof of the binomial identity
k
! ! !
X m n m+n
= (3)
r=0 r kr k
which can also be written as
!
m+n
N0 + N1 + . . . + Nk = (4)
k
When the committee size k exceeds n, the number of women, then
the committee must have at least one man. This is obvious by sheer
common sense and also follows mathematically because in this case
! !
m n
N0 = =10=0 (5)
0 k
!
n
since = 0 whenever k > n. So, in this case we have
k
!
m+n
N1 + N2 + . . . + Nk = (6)
k
The present problem falls in this category, where m = 5, n = 4 and
k = 5. More generally, if k n = p, then instead of (6) we have
!
m+n
Np + Np+1 + . . . + Nk = (7)
k

37
Q.39 If f : IR IR is a twice differentiable function such that f 00 (x) > 0
1 1
for all x IR, and f ( ) = , f (1) = 1, then
2 2
1 1
(A) 0 < f 0 (1) (B) f 0 (1) < 0 (C) f 0 (1) > 1 (D) < f 0 (1) 1
2 2

Answer and Comments: (C). The Lagrange Mean Value Theorem

applied to f (x) on the interval [1/2, 1] gives some c (1/2, 1) such that
1 12
f 0 (c) = = 1.
1 12
The hypothesis f 00 (x) > 0 means (by applying Lagrange MVT to f 0 (x))
that f 0 (x) increases strictly on (, ). Since c < 1 and f 0 (c) = 1,
we must have f 0 (1) > f 0 (c) = 1. So (C) holds.
A very straightforward problem on two applications of LMVT, once
to f (x) and then to f 0 (x).

Q.40 If y = y(x) satisfies the differential equation

r !1
q  q
8 x 9 + x dy = 4+ 9+ x dx, x > 0

and y(0) = 7, then y(256) =

(A) 80 (B) 3 (C) 16 (D) 9

Answer and Comments: (B). The differential equation can be brought

to the separate variable form as
dy 1
= ! (1)
dx
q  r q
8 x 9+ x 4+ 9+ x

So, the problem really is of integrating the function appearing on the

R.H.S. We first evaluate the indefinite integral
1Z dx
I= ! (1)
8
q  r q
x 9+ x 4+ 9+ x

38
The substitution
q
u= 9+ x (2)

suggests itself. With this substitution,

dx
du = q


(3)
2 9+ x 2 x

and the integral becomes

Z
du
I =
2 4+u

= 4+u+C (4)

where C is an arbitrary constant. So

y = 4+u+C
r q
= 4+ 9+ x+C (5)

The initial condition y(0) = 7 determines C as 0. Then by a direct
calculation (using (16)2 = 256),
q
y(256) = 4+ 9 + 16 = 9=3 (6)

A straightforward problem once the correct substitution strikes.

Those
who cannot think of it at one go can begin with the less ambitious
t = x to get
1Z dt
I=   q  (7)
4 9+t 4+ 9+t

and now put u = 9 + t to get (4).
Although ostensibly a problem about differential equation, it is really
a problem about finding antiderivatives. Such problems are safe havens
for the hard workers. The numerical data has been carefully chosen to
alert against numerical slips.

39
Q.41 How many 3 3 matrices M with entries from {0, 1, 2} are there for
which the sum of the diagonal entries of M T M is 5 ?

(A) 198 (B) 126 (C) 135 (D) 162

Answer and Comments: (A). Although M T is a standard notation

for the transpose of a matrix M , other notations such as M t and M 0 are
also used. It would have been better if the paper setters had clarified
that M T stands for the transpose of M .

x1 y1 z1 x1 x2 x3
Now suppose M = x2 y2 z2 . Then M T = y1 y2 y3 . There

x3 y3 z3 z1 z2 z3
T
is no need to write the matrix M M elaborately. We are interested
only in its diagonal elements. But there is an easy way to remember
any entry, say the (i, j)-th entry of it. It is simply the dot product
of the i-th and the j-th columns of M treated as vectors. Thus, for
example, the entry in the first row and the third column of M T M will
be x1 z1 + x2 z2 + x3 z3 obtained by taking the dot product of the vectors
x1 i + x2 j + x3 k and z1 i + z2 j + z3 k corresponding to the first and the
third columns of M respectively. With this understanding, the three
3 3
x2i , yi2 and
X X
diagonal entries of the matrix M T M will be simply
1 1
3
zi2 respectively. So their sum (often called the trace of M T M ) will
X

1
simply be the sum of the squares of the nine entries of M . We are given
that this number is 5 and each entry is from the set {0, 1, 2}.
After this the matrices have no role to play. If we unfold the matrix
M by witing its nine entries as (x1 , y1 , z1 , x2 , y2 , z2 , x3 , y3 , z3 ) we get an
ordered 9-tuple. Conversely every such ordered 9-tuple determines a
unique matrix (in which the first three entries form the first row, the
next three the second row and so on). So the problem now reduces
to asking how many ordered 9-tuples (n1 , n2 , . . . , n9 ) there are whose
squares add to 5 and in which each ni is either 0, 1 or 2. This is a
simple counting problem.
9
n2i as N . Each n2i is either 0, 1 or 4. As
X
Let us call the sum
i=1
N = 5 and all summands are non-negative, at most one summand in

40
 this summation is 4. When that happens, there is one more summand
which is 1 and the rest are all 0. These lone entries 2 and 1 can occur
in 9 8 = 72 ways. So there are 72 ordered 9-tuples in which one entry
is 2. When none of the entries is 2, exactly 5 ! of them are 1 and the
9
remaining 4 are 0 each. This can happen in = 126 ways (already
5
counted in the answer to Q.38). So the total count is 72 + 126 = 198.
The problem is a good combination of elementary facts about
matrices and combinatorics. The calculations are reasonable once the
key idea strikes.

Q.42 Let O be the origin and let P QR be any arbitrary triangle. The point
S is such that

OP OQ + OR OS = OR OP + OQ OS = OQ OR + OP OS

Then the triangle P QR has S as its

(A) incentre (B) orthocentre (C) circumcentre (D) centroid

Answer and Comments: (B). The very fact that the question in-
volves the dot products of the position vectors of the vertices of a tri-
angle P QR suggests that somewhere perpendicularity will be involved
and therefore chances are that the correct answer is (B).
This guess turns out to be correct. In fact, as shown in Comment No.
5 of Chapter 8, the most elegant proof of the concurrency of the three
altitudes of a triangle is using vectors, specifically their dot product
because perpendicularity of two vectors is equivalent to the vanishing
of their dot product.
Let us begin by recasting the given equations. The equality of the
first two expressions can be recast as

OR OS OR OP = OQ OS OQ OP (1)

where on the R.H.S. we have used the commutativity of the dot product

to rewrite OP OQ as OQ OP . Now using distributivity of the dot

product, we can write the L.H.S. of (1) as OR (OS OP ). But

41

OS OP = P S. So the L.H.S. of (1) equals

OR P S (2)
By a similar reasoning the R.H.S. of (1) equals

OQ P S (3)
Subtracting (3) from (2) and again using distributivity we get

(OR OQ) P S = 0 (4)
and hence

QR P S = 0 (5)
As a result we have that the line P S is perpendicular to the line QR.
This shows that the point S lies on the altitude of 4P QR passing
through P . Using the equality of the second and the third expression
in the data, we get RS P Q which means that S also lies on the
altitude through R. So S is at the intersection of two altitudes of
4P QR and is, therefore, the orthocentre of it.

Q R

Fig. 6 : Concurrency of altitudes

This is a good problem if asked as a full length question. In essence
the solution is the proof of the concurrency of the three altitudes of a
triangle. But as indicated at the beginning, the correct answer can also
be guessed by the mere presence of the dot product. It is hard to say
how many of the candidates who answer this question correctly will be
able to prove their answer in the manner indicated here. Another sad
casualty of the multiple choice format of the JEE.

42
Q.43 The equation of the plane passing through the point (1, 1, 1) and per-
pendicular to the planes 2x + y 2z = 5 and 3x 6y 2z = 7, is

(A) 14x + 2y + 15z = 31 (B) 14x + 2y 15z = 1

(C) 14x + 2y + 15z = 3 (D) 14x 2y + 15z = 27

Answer and Comments: (A). An extremely standard problem. Call

(1, 1, 1) as A. We are given that it lies on the desired plane. We now
need a vector n normal to the desired plane. Once we have such a
vector, say n = ai + bj + ck, then for any point P = (x, y, z) in the
plane, the line P A must be perpendicular to n. Therefore we have

a(x 1) + b(y 1) + c(z 1) = 0 (1)

which will give the equation of the desired plane.

To identify this normal vector n, let n1 and n2 be vectors per-
pendicular to the two given planes respectively. As the desired plane
is perpendicular to the first plane, we must have n n1 because two
planes are perpendicular to each other if and only if their normals are
perpendicular to each other. Similarly, n n2 . So we may take n as
n1 n2 . The vectors n1 and n2 can be written down from the equations
of these planes by inspection. Thus

n1 = 2i + j 2k
and n2 = 3i 6j 2k (1)

Therefore,

n = n1 n2
 
 i j k 

=  2 1 2 
 
 3 6 2 
 

= 14i 2j 15k (2)

As the point (1, 1, 1) is on the plane, its equation is

14(x 1) + 2(y 1) + 15(z 1) = 0 (3)

which is the same as (A).

43
 The crux of the solution is the observation that n can be taken as
n1 n2 . But this is a well-known result and students are exposed to
many drill problems based on it. So this question is not suitable in an
advanced test.

SECTION 2 : (Maximum Marks 28)

This section contains SEVEN questions.
Each question has FOUR options (A), (B), (C) and (D). ONE OR MORE
THAN ONE of these four options is (are) correct.
For each question, darken the bubble(s) corresponding to all the correct
option(s) in the ORS
For each question, marks will be awarded in one of the following categories :
+4 If only the bubble(s) corresponding to all the correct option(s) is (are)
darkened.
+1 For darkening a bubble corresponding to each correct option, Provided
NO incorrect option is darkened. Zero Marks
0 If none of the bubbles is darkened.
-2 In all other cases. for example, if (A), (C) and (D) are all the correct
options for a question, darkening all these three will get +4 marks; darkening
only (A) and (D) will get +2 marks; and darkening (A) and (B) will get -2
marks, as a wrong option is also darkened
98 Z k+1
X k+1
Q.44 If I = dx, then
k=1 k x(x + 1)

49 49
(A) I < (B) I < loge 99 (C) I > (D) I > loge 99
50 50

Answer and Comments: (B,C). An important feature of a good

question is that it often carries a built in hint for the solution. Art
lies not only in recognising the hint but even more so, in designing a
problem with just the right degree of hint. As commented (with an
illustration) in Comment No. 4 of Chapter 7, too faint a hint would
make the problem dicey while too loud a hint makes it dull.
The present problem is an excellent example of the right degree
of hint. The given expression I is not a single integral but a sum of
98 terms each of which is an integral. We are asked to compare it

44
 49
with two numbers, one and the other ln 99. The former can be
50
98
written as which, in turn, is a sum of 98 terms each of which is
100
1
. The other number, ln 99, is not a sum of 98 terms as it stands.
100
But we can rewrite it as ln 99 ln 1 and then recast it as a telescopic
98
X
series ln(k + 1) ln k. Further the k-th term ln(k + 1) ln k of
k=1
Z k+1
dx
this series is precisely the integral . And since I is also given
k x
as a Zsum of 98 integrals this suggests that we should perhaps com-
k+1 k + 1 Z k+1
1
pare dx with dx. With greater luck, two integrals
k x(x + 1) k x
whose intervals of integration are identical can be compared by com-
paring their integrands. (Things are, of course, not so rosy always.
That is, while it is true that f (x) g(x) for all x [a, b] implies that
Z b Z b
f (x)dx g(x)dx, the converse is far from true.)
a a
So, let us see if luck is with us. That is, let us compare the
k+1 1
integrands and for all values of x in the interval [k, k + 1].
x(x + 1) x
k+1
For every x [k, k + 1], x + 1 k + 1 and so the factor is at
x+1
most 1. So, we do have
k+1 1
(1)
x(x + 1) x
for all x [k, k + 1] and hence
Z k+1 Z k+1
k+1 dx
dx = ln(k + 1) ln k (2)
k x+1 k x
for every k = 1, 2, . . . , 98. As already noted, by adding these 98 in-
equalities we get I ln 99. This is not quite the same as (B), because
(B) demands a strict inequality. An unscrupulous student will simply
not bother about such fine points. But, for the scrupulous ones, we
remark that if f (x), g(x) are continuous on [a, b] and f (x) g(x) for
all x [a, b], with the inequality being strict even at one point of [a, b],
Z b Z b Z b Z b
then not only f (x)dx g(x)dx but f (x)dx < g(x)dx. The
a a a a

45
proof follows by considering the continuous function g(x)f (x) and ap-
plying the result that if the integral of a non-negative continuous func-
tion is 0 then the function is identically 0. (The proof requires that a
continuous function which is positive at some point c is bounded below
by a positive number over a sufficiently small interval [c, c+] around
c.) Now, the inequality (1) is strict not only for some x [k, k + 1],
but, in fact, for all x > k and so each of the inequalties in (2) is strict.
As a result, what we get by adding them is also a strict inequality.
49
Let us now address the other task, viz. comparing I with , i.e.
50
98 k+1
with . In (1) we bounded the integrand from above on
100 x(x + 1)
the interval [k, k + 1]. Let us now go for a lower bound on it. As both
the factors in the denominator are positive and increase with x, the
miminum of the integrand will occur at x = k + 1. This minimum is
1
simply . So we get
k+2
k+1 1
(3)
x(x + 1) k+2
for all x [k, k + 1]. Integrating both the sides over [k, k + 1] we get
Z k+1 Z k+1
k+1 1 1
dx dx = (4)
k x(x + 1) k k+2 k+2
for every k = 1, 2, . . . , 98. Adding these 98 inequalities, we get
1 1 1 1
I + + ... + + (5)
3 4 99 100
The terms of the R.H.S. are descending and so the sum is at least 98
98 49 49
times the last term, i.e. = . So, we have proved that I .
100 50 98
Again for the fussy ones, strict inequality holds, because it holds in (4)
(with the same justification as that given for strictness in (2)). But
this time, there is a much simpler way out. We got the inequality by
1
replacing all terms on the R.H.S. of (5) by the lowest one, viz. .
100
The remaining 97 terms are all bigger than it. So, strict inequality
49
holds and we get I > , thereby proving (C).
50
46
 Z k+1
k+1
In this solution, we never evaluated the integrals dx
k x(x + 1)
which form the terms of the series that defines I, If we do, we would
get the exact value of I and not just a lower and an upper bound.
Although this extra work is unnecessary as far as the present problem
is concerned, it is tempting and may have some educative value and so
1 1 1
we carry it out. Resolving into partial fractions as
x(x + 1) x x+1
we get
Z k+1 Z k+1
k+1 1 1
 
dx = (k + 1) dx
k x(x + 1) k x x+1
k+1
= (k + 1)[ln x ln(x + 1)]

k
= (k + 1)[ln(k + 1) ln(k + 2) ln k + ln(k + 1)]

So,
98
X
I= (k + 1)[ln(k + 1) ln(k + 2) ln k + ln(k + 1)] (6)
k=1

There is usually no easy way to get a closed form expression for the
sum of a series like this. One option is to replace the integer 98 (which
has little special significance) by a variable positive integer n, call the
sum as Sn , guess a formula for it by evaluating Sn for some low values
of n and then prove the guess by induction on n. But sometimes
certain tricks such as telescopic series provide an unexpected way out.
(See Comment No. 11 of Chapter 5 for an illustration of both the
approaches.) In the present problem, the idea of a telescopic series is
not so foreign. Already we recast ln 99 as the sum of a telescopic series
98
X 1
(ln(k + 1) ln k). The integrand also resembles the terms
k=1 x(x + 1)
n
X 1
of a very standard series which is evaluated by expressing it
r=1 r(r + 1)
n 
1 1 1
X 
as a telescopic series . The answer is simply 1 .
r=1 r r+1 n+1
X
The trick in expressing a given series ak as a telescopic series
is to rewrite ak as bk+1 bk for a suitable sequence {bk }. This is not
always possible. Let us see if we can do this to the typical term of

47
the series (6). To begin with we would like to subtract k ln k from the
first term (k + 1) ln(k + 1). This can be done by splitting (k + 1) ln k
as k ln k + ln k. Similarly, we split the last term (k + 1) ln(k + 1) as
k ln(k + 1) + ln(k + 1). Luckily the two left over terms can be combined
to give ln(k + 1) ln k which will serve as a term of another telescopic
series. We are now left with k ln(k + 1) (k + 1) ln(k + 2), which is
precisely the negative of the k-th term of another telescopic series. (We
reiterate that such a regrouping is not always possible. But here we
are lucky.)
With this regrouping, we now have
98
X 98
X
I = [(k + 1) ln(k + 1) k ln k] + [ln(k + 1) ln k]
k=1 k=1
X98
[(k + 1) ln(k + 2) k ln(k + 1)] (7)
k=1

All the three series on the R.H.S. are telescopic series. Therefore

I = 99 ln 99 + ln 99 99 ln 100 + ln 2
= 100 ln 99 99 ln 100 + ln 2 (8)

Let us now see whether this hard work pays to answer the question at
hand. That is, whether we can give alternate proofs of the inequalities
49
I < ln 99 and I > which we proved earlier using elementary (albeit
50
rather tricky) arguments. Let us begin with the first one. We can
rewrite (8) as

2 (99)100
!
I = ln (9)
(100)99

2 (99)100
So, comparing I with ln 99 is equivalent to comparing with
(100)99
99, or eqivalently, comparing 2(99)99 with (100)99 . As all the numbers
100 99
 
involved are positive this further reduces to comparing 2 with .
99
99
1

The latter is a familiar figure if we rewrite it as 1 + . It is the
99

48
 n
1

99-th term of the well-known sequence { 1 + } which converges to
n
1 99
 
e as n . As 99 is a fairly large integer, we expect 1 +
99
to be fairly close to e which is much bigger than 2. So we get that
I < ln 99. A rigorous proof is not very difficult either. In showing that
1 n
 
the sequence { 1 + } is convergent, it is proved using the binomial
n
theorem that it is monotonically increasing (and also that it is bounded
1 99
 
above). If we expand 1 + using the binomial theorem, it equals
99
1
1 + 99 plus 98 more terms, all of which are positive. So surely,
99
99
1

1+ > 2, which is equivalent to saying that I < ln 99.
99
49
Let us now turn to the other inequality, viz. I > . This is much
50
easier because in our earlier proof we did this by replacing every term
1 1 1 1
of the series + + . . . + by its lowest term, viz. . So the
3 4 100 100
49
lower bound was a very crude one. By adding the first few terms
50
of this series, we can get a better lower bound, e.g. 1 or even higher.
So, we would be through if using (9) we can prove that I > 1, which
is equivalent to showing that 2 (99)100 > e(100)99 . Once again, we
 99
1
reduce this to showing that 2 99 > e 1 + 99 . The R.H.S. is at
2
most e which is much much less than the L.H.S. 198.
49
Of course, using the powerful calculation (9) of I to prove <I<
50
ln 99 is like using a sledge hammer to kill a fly, or, resorting to a simile
from the engineering parlance, to call an electical engineer to change a
light bulb! The inequalities asked in the problem are not so ambitious.
What makes the problem commendable is the implied hint in its very
formulation. The paper setters deserve to be complimented for coming
up with it.
1 1 1
In passing we mention that the number + + . . . + which
3 4 100
we encountered is closely related to a well known sequence of numbers
called the harmonic numbers. (See p. 240.) The n-th harmonic

49
 n
X 1
number Hn is defined as . So the number we encountered is H100
k=1 k
3
. It can be proved that for large n, Hn is very close to ln n. So the
2
lower bound we got for I in (5) is nearly ln 1001.5 which comes out to
49
be 3.10517019, a lot bigger than but a lot less than the approximate
50
value of I (which comes out to be 4.293283). Incidentally, the upper
bound ln 99 we got is about 4.59511.
Q.45 If f : IR IR is a differentiable function such that f 0 (x) > 2f (x) for
all x IR, and f (0) = 1, then

(A) f (x) > e2x in (0, ) (B) f (x) is decreasing in (0, )

(C) f (x) is increasing in (0, ) (D) f 0 (x) < e2x in (0, )

Answer and Comments: (A,C). It is tempting to recast the inequal-

f 0 (x)
ity f 0 (x) > 2f (x) as > 2 and integrate both the sides w.r.t. x
f (x)
to get that ln f (x) > 2x which will imply (A). But this is fallacious
for two reasons. First, it is not given that f (x) is positive. Secondly,
although an inequality of the integrands implies an inequality of the
integrals (which was the main idea in the solution of the last problem),
that is true only for definite integrals. Nothing can be concluded about
indefinite integrals, because an indefinite integral is not a single func-
tion but a family of functions every two of which differ by a constant.
The constants of integrations appearing on the two sides are unrelated
and arbitrary. For example, we have sin x 1 for all x. But that does
not mean that cos x x for all x. Indeed this is false for x = /2.
The correct approach here is to rewrite the given inequality as
f 0 (x) 2f (x) > 0 (1)
for all x IR. Now in (1), if instead of the inequality sign > we had
an equality, then it would be a differential equation and we could solve
it by multiplying it throughout by its integrating factor e2x . As the
exponential function is always positive, we can still multiply both the
sides of the inequality (1) to get
e2x (f 0 (x) 2f (x)) > 0 (2)

50
 which can be rewritten as
d 2x
(e f (x)) > 0 (3)
dx
for all x IR. For brevity, denote e2x f (x) by g(x). Then this inequal-
ity means that g(x) is strictly increasing on IR. So g(x) > g(0) for all
x (0, ). From the data f (0) = 1 we get g(0) = 1. So So we get

e2x f (x) > 1 (4)

for all x (0, ). Multiplying both the sides by e2x (which is positive
for all x) we now get

f (x) > e2x (5)

for all x (0, ). So (A) is true. In particular, this also means that
f (x) > 0 for all x > 0 (something we would have liked to hold for
the earlier approach to go through). Combining this with the data
that f 0 (x) > 2f (x) to conclude that f 0 (x) > 0 for all x (0, )
which means (C) is true and (B) false. Finally, (D) is false too because
f 0 (x) > 2f (x) along with (5) means f 0 (x) > 2e2x for all x (0, ).
Although we used methods from differential equations (specifically,
the integrating factor), the problem itself is about an inequality. The
problem in Comment No. 13 of Chapter 19 is strikingly similar to the
present problem.
 

cos(2x) cos(2x) sin(2x) 

Q.46 If f (x) = cos x cos x sin x

 ,

then
 
 sin x sin x cos x 

(A) f 0 (x) = 0 at exactly three points in (, )

(B) f (x) attains its maximum af x = 0
(C) f (x) attains its minimum af x = 0
(D) f 0 (x) = 0 at more than three points in (, )

Answer and Comments: (B,D). It was remarked at the beginning of

the solution to Q.44 that a good problem carries a built in hint in its for-
mulation. It is therefore tempting to think that there is some purpose

51
behind defining f (x) by a determinant. But this is more likely to be a
trap than a hint. The question deals with the derivative of f (x). And,
if a function is the the determinant whose entries are (differentiable)
functions, then its derivative is in general not the determinant whose
entries are the derivatives of these functions. The correct formula for
f 0 (x) is that it is the sum of determinants obtained by differentiating
the functions in one row at a time (leaving
 the other rows unchanged).
 f (x) f (x) 
For example, if f (x) =  1 2
, then

 f3 (x) f4 (x) 
     
d  f1 (x) f2 (x)  f 0 (x) f20 (x)
   f1 (x) f2 (x) 
= 1 + (1)
    
dx  f3 (x) f4 (x) f3 (x) f4 (x) f30 (x) f40 (x)
   
    

In the present problem, f (x) is a 3 3 determinant of functions and

so f 0 (x) will be a sum of three 3 3 determinants. Expanding each
one of them and adding will be a horrendous task. It is far better to
express f (x) itself by expanding the determinant. Fortunately, because
of the trigonometric identities for cos 2x and sin 2x, the expansion is
quite manageable. Thus
 

 cos(2x) cos(2x) sin(2x) 

f (x) =

 cos x cos x sin x 

 
 sin x sin x cos x 

= cos 2x + cos 2x(cos2 x sin2 x) + sin 2x(2 sin x cos x)

= cos 2x + cos2 2x sin2 2x
= cos 2x + cos 4x (2)

As a result,

f 0 (x) = 2 sin 2x 4 sin 4x = 2 sin 2x(1 + 4 cos 2x) (3)

The factor sin 2x vanishes at three points in (, ), viz. x = 0 and

1
x = /2. In addition, the second factor vanishes when cos 4x =
4
which has some solutions in (, ). So (A) is false and (D) is true.
(We dont need to count all the zeros of the second factor. It is enough
that it has at least one.)
For (B) and (C), clearly f 0 (0) = 0. Further f 0 (x) changes sign from +
to as x passes through 0. So, there is a local maximum at x = 0.

52
 But this is not what the question asks. It asks for global maximum i.e.
maximum value of f (x) as x varies all over the domain of f (x). The
domain of f (x) is not clearly specified. The wording of (A) and (D)
gives the impression that the domain of f (x) is (, ). Fortunately,
for this function it does not matter. Both the summands cos 2x and
cos 4x attain their maxima at 0. So f (x) has an absolute maximum at
x = 0 no matter what the domain is, as long as it contains the point
0. So (B) is true and (C) false.
A disappointing problem for several counts. First, giving the
function f (x) as a determinant had little to do with the main theme
of the problem. Secondly, the expansion of the determinant was too
straightforward. Thirdly, the trigonometric equation f 0 (x) = 0 was
very simple to solve because of its factorisation and the nature of the
options further made it unnecessary to solve it in full. Finally, the part
involving the maximum of the function was such that even a student
who confuses a local maximum with maximum would go scot free.
In essence, the problem is a combination of problems from three ar-
eas, viz. determinants, trigonometric equations and maxima/minima.
All three parts are trivial and you cant make a good problem merely
by putting together several trivial ones.

Q.47 Let and be non-zero real numbers such that 2(cos cos ) +
cos cos = 1. Then which of the following is/are true ?

! !

   
(A) tan 3 tan =0 (B) 3 tan tan =0
2 2! 2 2!
 

 
(C) tan + 3 tan =0 (D) 3 tan + tan =0
2 2 2 2

Answer and Comments: (A,C). In commending the paper setters

for Q.44 because that problem carried an implied hint for its solution,
it was remarked that if the hint is too faint, the problem becomes dicey
while if it is too loud, the problem becomes dull. The present problem
is an excellent example of the latter. The data is in terms of cos and
cos while all the four options are in terms of tan( 2 ) and tan( 2 ). So

53
it takes no brain to realise that what is needed here is a formula which
expresses cos x in terms of tan( x2 ).
Nor does it take much memory to recall this formula as it is a very
standard formula, often used in integration too when the integrand is
a trigonometric function of x and we use the substitution t = tan( x2 ).
Specifically, the formula we need is

1 tan2 ( x2 )
cos x = (1)
1 + tan2 ( x2 )

With this formula, the given equation, viz.

2(cos cos ) + cos cos = 1 (2)

becomes
1 tan2 ( 2 ) 1 tan2 ( 2 ) 1 tan2 ( 2 ) 1 tan2 ( 2 )
!
2 + =1 (3)
1 + tan2 ( 2 ) 1 + tan2 ( 2 ) 1 + tan2 ( 2 ) 1 + tan2 ( 2 )

For notational brevity denote tan( 2 ) and tan( 2 ) by a and b respectively.

Then (3) becomes

1 b 2 1 a2 1 a2 1 b 2
!
2 + =1 (4)
1 + b2 1 + a2 1 + a2 1 + b 2

Clearing the denominators and simplifying, we get

4(a2 b2 ) + 1 a2 b2 + a2 b2 = 1 + a2 + b2 + a2 b2 (5)

Further simplification yields

a2 3b2 = 0 (6)

Factoring, either a + 3b = 0 or a 3b = 0. These correspond to (C)
and (A) respectively.
The knowledge of the right formulas and the skill needed to ma-
nipulate them algebraically are elementary. But the purpose of stating
in the question that , are non-zero is not clear. The question deals
exclusively with trigonometric functions of and . Because of their
periodicity, merely knowing that , are non-zero gives no relevant

54
 information. In the options given it has to be tacitly assumed that
, are not odd multiples of (or else tan(/2) and tan(/2) are
undefined). But this is not implied by the hypothesis that , are
non-zero. So this additional redundant stipulation is likely to confuse
a candidate. It is probably for this reason that a full score of 4 points
to every candidate has been announced.
sin(2x)
sin1 (t)dt, then
R
Q.48 If g(x) =
sin x

(A) g 0 (/2) = 2 (B) g 0 (/2) = 2

(C) g 0 (/2) = 2 (D) g 0 (/2) = 2

Answer and Comments: (None.) The integrand in the definition

of g(x), viz. sin1 (t) is defined only for t [1, 1]. But this creates
no problem in defining g(x) because both the limits of integration, viz.
sin x and sin 2x lie in this interval. It may happen that for some x, the
lower limit sin x is higher than the upper limit sin 2x. For example, this
happens when /3 < x < . But that need not bother us either. Al-
Z b
though a definite integral f (x)dx is defined initially (using Riemann
a
sums) under the assumption that a < b, later the definition is extended
Z b Z a
even to cover the case where b < a by setting f (x)dx = f (x)dx
a b
provided the function f (x) is integrable over the interval [b, a]. With
this extended definition all the results about integrals continue to hold
regardless of whether the upper limit is larger or smaller than the lower
one, provided that all the integrals appearing in the statement of those
results make sense. For example, the additivity property of integrals
Z b Z c Z b
f (x)dx = f (x)dx + f (x)dx (1)
a a c

normally presupposes that c lies in the interval [a, b], i.e. a b c.

(See the first part of the figure below.)
. . .b . . .
a c a b c

55
Suppose however, that a < b < c. Then c lies outside this interval
as shown in the second part of the figure. But assume that f (x) is
integrable on the interval [a, c]. Then by the additivity property above
we have
Z c Z b Z c
f (x)dx = f (x)dx + f (x)dx (2)
a a b
Z b
Replacing the last term by f (x)dx and bringing it to the L.H.S.
c
we get
Z c Z b Z b
f (x)dx + f (x)dx = f (x)dx (3)
a c a

which means that the additivity property (1) holds true in this case too.
The same is also true of important theorems such as the two forms of
the Fundamental Theorem of Calculus. What we need in this problem
is a slight extension of the second form of Fundamental theorem, i.e. a
formula for the derivative of a function defined by an integral in which
both the upper and the lower limits are some functions. Specifically,
the result we need is
d Z v(x)
f (t)dt = f (v(x))v 0 (x) f (u(x))u0 (x) (4)
dx u(x)
provided f (t) is continuous and u(x), v(x) are differentiable functions
of x. (This is Equation (19) on p. 644.)
Applying this result to the present problem, we have

g 0 (x) = 2 sin1 (sin 2x) cos 2x sin1 (sin x) cos x (5)

for all x IR. For x = , we have sin 2x = 0 and so sin1 sin(2x) =
2
1
sin 0 = 0. Also cos x = 0 for x = . Therefore a direct substitution
2

into (5) gives that g 0 ( ) = 0 and g 0 ( ) = 0 too. Hence none of the
2 2
given four statements is correct.
Apparently, the paper setters took sin1 (sin 2x) to be 2x. With
this interpretation we do get g 0 (/2) = 2 and g 0 (/2) = 2. But
the formula sin1 (sin ) = is not unconditionally valid. It is valid

56
 only for [/2, /2]. As it happens, when x = /2, 2x =
or neither of which lies in the interval [/2, /2]. This lapse on
the part of the paper setters has given a bonus credit to all candidates.
But the scrupulous candidates who have an unflinching faith in the
correctness of the question often spend considerable time in checking
and rechecking their work to see if they have made any mistake. So,
in a competitive examination, a lapse like this serves to punish the
sincere candidates. This is even worse than asking a question for which
a sneaky answer is possible.

Q.49 If the line x = divides the region R = {(x, y) IR2 : x3 y x, 0

x 1} into two equal parts, then

1
(A) <<1 (B) 4 + 42 1 = 0
2
1
(C) 0 < < (D) 24 42 + 1 = 0
2

Answer and Comments: (A,D). It is hardly necessary to draw a

diagram. But we do it nevertheless.

y
x = A (1, 1)

Q R

x
O

x = 1/2

Fig. 8: Region between y = x3 and y = x

The graphs of the functions y = x and y = x3 meet at (0, 0) and (1, 1).
For x (0, 1) the graph of the cubic lies below the line y = x. So, the
area, say A, of the region R (shown by shading) comes out to be
Z 1
1 1 1
A= x x3 dx = = (1)
0 2 4 8

57
Now, for 0 < < 1, the subregion bounded above by y = x and
below by y = x3 and lying between x = 0 and x = is the region
O P Q O where P = (, 3 ) and Q = (, ). Its area is
Z
2 4 22 4
x x3 dx = = (2)
0 2 4 4
The line x = will divide the region into two parts of equal areas if
A 1
and only if this equals = . This gives us an equation for , viz.
2 8
42 24 = 1 (3)

which is the same as (D). So (D) is true and automatically (B) is false,
for if both (B) and (D) were true, 24 = 4 which is clearly false as
6= 0.
Our next task is to choose between (A) and (C). For this we need
to compare the numbers and 12 . Given x, y IR deciding which of
them is greater is not an easy task when the numbers are very close to
each other. Calculators give only approximate values. If the difference
|xy| is less than the smallest positive number the machine can handle,
then it will not be able to tell which of the two numbers is smaller than
the other. In such cases, a variety of methods, often tailored to the
particular problem are used. In general they fall into three categories.
We illustrate them here by comparing with 12 in three different ways.

(i) Direct Comparison: For this method we need to have both the
numbers in an exact form. The number 21 is exact enough. To
determine the exact value of we shall have to solve (3) for .
Normally polynomial equations of degree 4 are not easy to solve.
But in the present case, we can treat (3) as a quadratic in u where
u = 2 . Then (3) gives

2u2 4u + 1 = 0 (4)

whose roots are 22 2 = 1 1 .
2
We discard the positive square
root since u = 2 < 1. So,

1 21
u=1 = (5)
2 2

58
 r
21
Taking square root again, we get =
2
. This is a rather
complicated number. To compare it with 12 , we reduce their com-
parison to that of their squares. This is valid as both the numbers
are positive. So, whether < 21 or > 12 can be decided by
comparing 2 i.e. u with ( 12 )2 , i.e. with 41 . This in turn, is equiv-
2
alent to comparing u2 with 16 1
. From (5), u2 = ( 21) = 322 2 .
1
2
Comparison of this with 16 reduces that between 24 16 2 and
2
1, and
2further between 23 and 16 2. Since (23) =1 529 while
(16 2) = 512 we see that 23 > 16 2 and hence > 2 .
Summing up, the original pair of numbers to be compared
is subjected to a series of reversible operations such as squaring
(assuming that the numbers to be compared are positive), multi-
plying both the numbers by any positive number (such as 16in the
illustration above), adding any number to both (such as 16 2 1
in the illustration above). All these reductions are reversible. We
go on applying them till we reach a stage where the two numbers
that are on the stage can be compared simply by inspection. As
the steps are all reversible, the ancestor of the greater number is
the greater number of the original two numbers.
In a rough work the process can be shown by preparing two
columns headed by the original numbers to be compared and a
column in between them. The entries in these two columns are
the numbers we get by applying some step in the reduction. The
middle column is kept blank or a question mark is put in it till
we know which side is the winner. For example, in the example
above the reduction can be shown as follows.
v
u 21
u
t 1
2
2
21 1

2 4
32 2 1
2 16
24 16 2 1
23 16 2
529 > 512

59
 The procedure is reminiscent of a prolonged battle between
two parties lasting over several generations, whose outcome is ul-
timately decided between the descendants of the original combat-
ants (who may well be dead by then).
(ii) Comparison through Middleman: Sometimes it is not easy
to compare two given numbers x and y directly, even after sub-
jecting both of them to a series of reductions. But we may be able
to find a number z (of our choice) and show that x < z and also
y < z relatively easily. Then we get x < y. (This is called the
transitivity property of the order relation. A layman applies it
in the day to day life and calls it common sense. For example, if
he knows that he is taller than his father but shorter than some
school mate of him, then he concludes that his father is shorter
than his school mate even though there may never be any occasion
for them to stand together.) Often more than one middlemen may
be

needed.
10
For

example,
20
it is not so easy to show directly that
1 1
1 + 10 < 1 + 20 . But using the binomial theorem it is not
 n
hard to show that the sequence an = 1 + n1 is monotonically
increasing by directly comparing an and an+1 for every n. So here
a10 and a20 are compared through the middlemen a11 , a12 , . . . , a18
2m1 1 2n1 1
and a19 . (See also the proof of the inequality >
m n
whenever m, n are positive integers with m > n given in the an-
swer to Q.40 of Paper 2 in the commentary of 2016.)
r
21 1
Let us now compare =
2
with through middlemen.
2
We already reduced the comparison to that between u = 2 =
1
21

2
and and continued it by squaring both the numbers again.
4
Instead, let us use that 2 > 1.4 (which follows because 2 > 1.96).

21 0.4
Then 2 1 > 0.4 and hence u = 2 > . But on the other
2
hand, we also have 2 < 1.5 (which follows from 2 < 2.25) and
0.4 0.4 4 4
hence > . This number is which is greater than ,
2 1.5 15 16
1 1 1
i.e. . So we have proved that > (or rather that 2 > )
4 2 4
0.4 0.4
through the middlemen and .
2 1.5

60
(iii) Calculus based methods: Suppose we can find a strictly in-
creasing function f defined on an interval that contains both x
and y. Then x < y is equivalent to f (x) < f (y). Already we used
some special cases of this in the first method, viz. reduction of the
comparison. For example, we used that if x, y are positive then
x < y if and only if x2 < y 2 . This follows from the fact that the
function f (t) = t2 is strictly increasing. But this fact can also be
proved purely algebraically. But calculus gives us a whole world of
functions whose monononicity can be proved easily using deriva-
tives but not otherwise. By choosing the function suitably, we can
compare two numbers by comparing their functional values.
For example, in the present problem, (2) gives the area of the
subregion of R bounded on the right by x = . As increases,
so does this subregion and hence its area. Therefore the R.H.S. of
(2) is a strictly increasing function of . Call it f (). (One can
also take the derivative f 0 () = 3 and see that it is positive
for (0, 1).) Herev we are taking as a variable and not the
u 21
u
particular number t . To avoid confusion, let us call this
2
1
number . We showed earlier that > by a rather elaborate
2
comparison, involving the exact value of and later through the
middlemen. With calculus the job can be done without identifying
explicitly. All we need is that is the unique number for which
1 1
f () = . As f is a strictly increasing function, whether > or
2 2
1 1
< can be decided by comparing f () and f ( ). We already
2 2
1
know f () = . By a direct calculation,
2
1 2 14 1
16 7
f( ) = = (6)
2 4 64
1 1 1
Evidently, f ( ) < = f (). Therefore < .
2 2 2
So, no matter which method we follow, we get that (A) is true and
(C) false.

61
 As a problem of calculation of areas of plane regions between the
graphs of two functions this problem comes nowhere close to many
other problems in the past where considerable work had to be done
first to identify the region. (See the list of problems in Exercise (17.3)
on p. 668.) The novel feature of this problem is that the region is a
variable subregion of a fixed region R. In this respect, the problem is
of the same spirit as Exercise (17.4) where you are asked to determine
the value of a so that the line x = a divides the region bounded by
8
the x-axis, the curve y = 1 + 2 and the vertical lines x = 2 and
x
x = 4, into two subregions of equal areas. But unlike in that problem
where you had to find the exact value of a, in the present problem you
1 1
are merely asked to determine if > or < . First we did this
2 2
by finding the exact value of (by solving a fourth degree equation)
1
and then directly comparing it with . This approach is analogous to
2
the ambitious approach in the solution to Q.44 above which consists of
finding the exact value of the sum I and then comparing it with ln 99
49
and . But just as in that problem there was a simpler way to do
50
the comparison without actually calculating I exactly, in the present
problem too, we saw that there is a much easier way to compare with
1
without actually determining exactly. This feature of the problem
2
makes it stand out from the routine problems of calculating areas using
definite integrals.
1 x(1 + |1 x|) 1
 
Q.50 Let f (x) = cos for x 6= 1. Then
|1 x| 1x

(A) lim+ f (x) does not exist (B) lim f (x) does not exist
x1 x1
(C) lim f (x) = 0 (D) lim+ f (x) = 0
x1 x1

Answer and Comments: (A,C). Limits where the independent vari-

able tends to 0 are more commonly asked. As a result, in handling
other limits, i.e. limits of the form lim f (x) where a 6= 0, a convenient
xa
substitution is to put x = a + h, express f (x) in terms of h and take
the limit as h 0. We follow this approach, i.e. we put h = x 1, i.e.

62
x = 1 + h. Then the expression for f (x) for x 6= 1 becomes

1 (1 + h)(1 + |h|) 1
 
cos (1)
|h| h

where we have used that | h| = |h| and also that cos() = cos().
Let us call this as g(h). The problem now asks for the right and the
left handed limits of g(h) as h 0.
After simplification, we have
h 1
 
g(h) = ( 1 h) cos (2)
|h| h
1
 
for h 6= 0. As h 0, cos oscillates and does not tend to any
h
1
 
limit. But it is bounded and so h cos 0 as h 0. This is also
h
1
 
true if we multiply cos by any function of h which tends to 0 as
h
h does. As a result, the problem reduces to considering the behaviour
h 1
 
of ( 1) cos as h 0. We consider two cases, depending upon
|h| h
whether h 0+ or h 0 .
When h 0+ , h > 0 and so |h| = h which makes the first factor
vanish. So, the limit exists and equals 0. So (A) is false and (D) is
true.
But when h 0 , h < 0 and so |h| = h which makes the factor
h
1 identically equal to 2 which is a constant and does not tend to
|h|
1
 
0. As cos tends to no limit, we get that lim g(h) does not exist.
h h1
So (B) is true and (D) false.
A very simple problem based on the fact that the product of two
functions one of which is bounded and the other tends to 0 also tends
to 0 regardless of the first factor.

63
 SECTION III : Maximum Marks : 12
This section contains TWO paragraphs.
Based on each paragraph, there are TWO questions.
Each question has FOUR options (A), (B), (C) and (D) ONLY ONE of these
four options is correct.
For each question, darken the bubble corresponding to the correct option in
the ORS.
For each question, marks will be awarded in one of the following categories :
+3 If only the bubble corresponding to the correct option is darkened.
0 In all other cases.

PARAGRAPH 1

Let O be the origin, and OX, OY , OZ be three unit vectors in the direc-

tions of the sides QR, RP , P Q respectively, of a triangle P QR.

Q.51 | OX OY | =
(A) sin(Q + R) (B) sin(P + R) (C) sin 2R (D) sin(P + Q)

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

below where denotes the angle between the vectors OX and OY .

P Y

O X

Z

Q R

Fig. 9: Unit vectors along the sides of a triangle

As both OX and OY are unit vectors, we have | OX OY | = sin .
So the problem boils down to expressing in terms of the angles of the
triangle P QR. Clearly , being an external angle, equals the sum of
the opposite two, viz. P + Q. Hence (D) is the only correct option.
The problem is too simple to require any comment. But the paper
setters deserve a compliment in their choice of notations. As pointed

64
 out in Comment No. 13 of Chapter 7, in a problem where several
triplets are involved, it saves some work if the notations are chosen so
as to be consistent with the natural cyclic orders of these triples. For
example, if D, E, F are points on the sides of a triangle, invariably the
old books will take D to lie on BC, E to lie on CA (not AC, mind
you!) and F on AB. This is so because the triplet (A, B, C) should
correspond to the triple (D, E, F ). So A and D should go together
etc. The side BC is opposite to A and so the point (out of D, E and
F ) which lies on it should be the one that corresponds to A. Also,
(BC, CA, AB) is a triplet in the right order and not (BC, AC, AB).
If a cyclically symmetric expression in a, b, c is to be simplified using
suitable substitutions, the old books will always take these substitutions
as, say, x = b+c, y = c+a and z = a+b rather than x = a+b, y = b+c
and z = a + c. (There are many other standard cyclic triplets of
notations, such as (x, y, z), (P, Q, R), (i, j, k), (, , ) and (, , ).)
Problems where the notations are designed with this aesthetic sense in
mind have a greater appeal than those without it. But this is probably
a dying virtue. So today it would not have been uncommon had the

vectors OX, OY and OZ were given to lie in the directions of P Q,

QR and RP respectively. But the paper-setters have been careful to
preserve the cyclic consistency of (P, Q, R) and (X, Y, Z).
Ironically, it is this very exercise of care which gives an unwar-
ranted sneaky path to the answer for the perceptive candidate. If the

triple (X, Y, Z) corresponds to the triple (P, Q, R), and | OX OY |
is symmetric w.r.t. X and Y , then logically the answer should be sym-
metric w.r.t. P and Q. So the choice boils down to (C) or (D). But if
the angle R is obtuse (and there is nothing in the problem to preclude
this) then sin 2R would be negative and hence cannot possibly equal
the magnitude of any vector. So that leaves only (D).
It is not clear whether the paper setters were aware of this sneaky
path. Maybe they were and they really wanted to reward a perceptive
candidate. Anyway, the problem is so simple even when done honestly,
that there is not much point in debating.

Q.52 If the triangle P QR varies, then the minimum value of cos(P + Q) +

cos(Q + R) + cos(R + P ) is

65
 3 3 5 5
(A) (B) (C) (D)
2 2 3 3

Answer and Comments: (B). This problem has nothing to do with

the preamble of the paragraph, viz. that OX, OY , OZ are three unit

vectors in the directions of the sides QR, RP , P Q respectively, of a
triangle P QR. That preamble was relevant only to the last question
and could have been added as a part of the data of that question. In
fact, this question is totally unrelated to the last one. Clubbing them
together into a single paragraph is silly to say the least. The paper
setters have also done away with the preservation of the natural order in
cyclic notations for which they were complimented in the last question.
Otherwise they would have given the expression as cos(Q+R)+cos(R+
P ) + cos(P + Q) is instead of cos(P + Q) + cos(Q + R) + cos(R + P ) is.
Anyway, coming to the problem itself, it is a problem of triangular
optimisation. In the present case the function
f (P, Q, R) = cos(P + Q) + cos(Q + R) + cos(R + P ) (1)
is to be minimised subject to the constraints P +Q+R = and P, Q, R
be non-negative. Since cos(P + Q) = cos( R) = cos R, etc. the
objective function f (P, Q, R) comes out to be
f (P, Q, R) = (cos P + cos Q + cos R) (2)
In the Main Problem of Chapter 14, it is shown that the maximum of
f (P, Q, R) (and hence the minimum of f (P, Q, R)) occurs when the
3
triangle is equilateral and its value is . In fact, as commented in
2
the introduction to that chapter, in the vast majority of problems of
triangular optimisations, the optimum is attained at an equilateral tri-
angle. The proofs by elementary methods are not always easy. But the
method of Lagranges multipliers makes a mincemeat of the problem
as shown in Comment No. 4 of Chapter 14.
There is, therefore, little point in asking such questions in the multi-
ple choice format. But as commented at the end of the introduction to
Chapter 14, a short question may be asked to test awareness of the fact
that the optimum occurs for an equilateral triangle. And the present
question does just that. The only thing is that this would have been
fine in the screening round but not in an advanced test.

66
 PARAGRAPH 2

Let p, q be integers and let , be the roots of the equation, x2 x1 = 0,

where 6= . For n = 0, 1, 2, . . . , let an = pn +
q n .
FACT : If a, b are rational numbers and a + b 5 = 0, then a = 0 = b.

Q.53 If a4 = 28 , then p + 2q =

(A) 14 (B) 7 (C) 12 (D) 21

Answer and Comments: (C). It would have been better to call the
given fact as a hint. Yes, it isa fact all right and can be proved using
another fact that the number 5 is irrational. But when a fact is given
with some purpose, the implication clearly is thatit will be needed in
the solution. So, let us see what role the number 5 in this fact has
2
present problem. The roots of the quadratic x x 1 = 0 are
in the
1 5
. The question does not specify which of these two roots is
2
and which is . But regardless of which is which we see that 5 has a
role to play in the problem.
Coming to the solution, by definition,

a4 = p4 + q 4 (1)

We calculate 4 and 4 . We have 2 = + 1. Squaring,

4 = ( + 1)2 = 2 + 2 + 1 = 3 + 2 (2)

Similarly,

4 = 3 + 2 (3)

Adding and using the fact (sorry, data) that a4 = 28, we get

28 = a4 = p4 + q 4
= p(3 + 2) + q(3 + 2) (4)

To make use of the hint, we need to writethis in terms of 5. The
1+ 5 1 5
problem does not specify whether = and = or the
2 2
67
other way. But that should not affect the answer which is given to be
unique. So, we take this possibility and substitute into (4) to get

1+ 5 1 5
28 = 3p + 3q + 2p + 2q
2 2
3p + 3q 3p 3q
   
= + 2p + 2q + 5
2 2
7p + 7q 3p 3q
   
= + 5 (5)
2 2
7p + 7q 3p 3q
Taking a = 28 and b = this becomes
2 2

a+b 5=0 (6)

Moreover, both a and b are rational since p, q are given to be integers.

So, from the fact, we get a = 0 = b, and hence

7p + 7q = 56 (7)
and 3p 3q = 0 (8)

solving which we get p = 4 and q = 4 and hence p + 2q = 12. Had we

interchanged and , the first term on the R.H.S.of (5) would remain
3q 3p

the same but the second term would have been 5. But
2
that would not affect (7) and (8). So we would get the same answer.
By recasting (5) as (6) we have been rather obsequious in adhering
to the fact. A mature way would have been to derive from the given
fact, another equivalent but more directly useful fact, viz. if

a1 + b 1 5 = a2 + b 2 5 (9)

where a1 , b1 , a2 , b2 are all rational, then a1 = b1 and a2 = b2 . This

follows by taking a = a1 b1 and b = b1 b2 in the given fact. Then (7)
and (8) would have followed directly from (5). A still more mature way
would be to understand the principle behind the fact, which ismerely
that the number 5 is irrational. The fact would still hold if 5 were
replaced by any irrational number. This can be used to simplify the
work a little by rewriting (4) in terms of , using the relation that

68
 + = 1 because and are the roots of the quadratic equation
x2 x 1 = 0. Then (4) becomes
(3p 3q) + (2p + 5q 28) = 0 (10)

and then the irrationality of (which follows from that of 5) would
imply p = q and 2p + 5q = 28 which give the same solution, viz.
p = 4 = q.
By giving a certain fact as a hint the paper setters have tried to help
a candidate. Had they given it more directly as Use irrationality of the
roots it would have been too loud. The best thing would be not to give
any hint at all, considering that this is an advanced test. Of course,
in that case it would not be clear in a multiple choice test whether a
candidate who answers the question correctly by going through (10)
really understands the reasoning or has taken a shot in the dark. But
then, this is a criticism that can be levelled against many multiple
choice questions. In the conventional tests, the question could besplit
into two parts, the first part asking the candidates to prove that 5 is
irrational and then give the present question as the second part.
Q.54 a12 =
(A) 2a11 + a10 (B) a11 a10 (C) a11 + a10 (D) a11 + 2a10

Answer and Comments: (C). In all four options, a12 is expressed as

a combination of a11 and a10 . It would be horrendous to try to do this
by expressing a12 , a11 and a10 in terms of and the way we could do
it for a4 in the last question because the higher powers of cannot be
converted so easily to linear expressions in . Nor do we need it.
All we need is to express 12 in terms of 11 and 10 and similarly,
to express 12 in terms of 11 and 10 . This can be done simply by
rewriting the quadratic x2 x 1 = 0 as x2 = x + 1. As , are the
roots we have
2 = + 1 (11)
and 2 = + 1 (12)
Multiply both the sides of (11) by 10 and those of (12) by 10 to get
12 = 10 + 10 (13)
and 12 = 11 + 10 (14)

69
Multiplying (13) by p, (14) by q and adding we get

a12 = a11 + a10 (15)

which is (C).
More generally, one can show that for every n 3,

an = an1 + an2 (16)

Note that (16) holds regardless of what p and q are. A relation of this
type where every term of a sequence {an } after some stage (in this
case for n 3 is expressed in terms of the preceding terms is called a
recurrence relation. The particular special case of the relation in this
problem in which p = 1 and q = 1 is popularly called the Fibonacci
relation. See Comment No. 11 of Chapter 4, where a closed form
expression for an (i.e. an expression in which an is expressed directly in
terms of n rather than in terms of an1 and an2 ) is proved by induction
on n. Proofs by induction (and indeed any proofs) are precluded by
the miltiple choice format. The Fibonacci relation is so popular that
chances are that some candidates are already familiar with it. Those
who are, are likely to get an advantage. But perhaps that is intentional.
Unlike in the last paragraph, those of the present one are based on a
common theme. The first question is more number theoretic, while the
second one is purely algebraic. Of course the number theory needed in
the first question appears more in the fact than in the question itself.
The proof of the fact given as a hint requires that 5 is an irrational
number. This, in turn, requires that 5 is a prime. More generally, one
can prove that for any prime p, there is no rational number x whose
square is p. The proof is notsubstantially different from the well-known
proof of the irrationality of 2. The crucial property of a prime needed
is that if p divides mn where m, n are integers, then p must divide either
m or n. Chapter 4 contains many interesting problems based on this
simple fact. But like the binomial identities, number theory is a victim
of the multiple choice format.

70
 CONCLUDING REMARKS

Both the papers are easy and, as compared with the past, below the
expected JEE standard (which is already low after making allowances for
the multiple choice questions).
Many familiar areas and types of questions are missing. There are no
probability problems in real life settings. There is no locus problem, except
perhaps Q.39 of Paper 1 (on complex numbers). But it is certainly not a lo-
cus problem in the conventional sense. Even though there are two interesting
questions in Paper 2, Q.44 (about a sum of integrals) and Q.49 (about a line
dividing a region into equal parts) which ask to compare two numbers, there
are no problems based on standard inequalities such as the A.M.-G.M. in-
equality or the Cauchy Schwarz inequality. There is only one problem (Q.46)
where a trigonometric equation is to be solved. But that equation is a trivial
one (and hence perhaps clubbed with determinants and maxima/minima.)
There is no problem about trigonometric identities in a triangle unless one
considers the Pythagorus theorem as a trigonometric identity, because two
questions (Q.42 and 44 in Paper 1) are based on it. Surely, in at least one of
them the triangle could have had some other given angle. The only problem
involving trigonometric simplification (Q.47 in Paper 2) demands little more
than a most standard identity which expresses cos in terms of tan(/2).
There are no questions about progressions, the only progression appearing
in the two papers is an A.P. of length 3 in Q.44 of Paper 1 (the sides of a
right angled triangle) where little more than the definition of an arithmetic
progression is needed. There are no questions on infinite series.
The two vector problems (Q.42 and Q.49 in Paper 2) are good but allow
sneaky solutions as pointed in the comments on them. The latter requires
little more than the definition of the cross product. Differential equations
appear only through differential inequalities (Q.45 in Paper 2). Q.40 in Paper
2 is ostensibly on solving an initial value problem. But in substance it is a
problem on finding an antiderivative.
On the other hand some areas have got disproportionately large repre-
sentation. For example, there are four counting problems, Q.47 in Paper 1
(about words of length 10), and in Paper 2 (Q.37, 38 and 41). Similarly, there
are as many as five questions on integration, Q.48 in Paper 1 and Q.40, Q.44,
48 and 49 in Paper 2. Some of them touch other areas, but the essence is on
integration. There are also many problems in calculus, based on Lagranges
Mean Value Theorem or the Intermediate Value Property. Surely some du-

71
plication could have been avoided. The paper setters have also dutifully
included one problem each involving limits of the greatest integer function
(Q.41 in Paper 1) and the absolute value function (Q.50 in Paper 2), both
the topics being worn out now to permit any radically new questions.
But what takes the cake is that there are four problems in Paper 1 about
tangents to a conic, viz. Q.42 and all three questions (Q. 49, 50, 51) in
tripartite matching. Coordinate geometry and especially the conics is a rel-
atively fertile area for innovative problems. Instead, what we see here is
some absolutely routine problems about tangents to conics. As pointed out
there and also in the comments on Q.52 to 54 of Paper 1, it is difficult to
see what is gained by posing questions in tripartite matching form. No new
mathematical knowledge or skill is tested. Apart from causing confusion to
candidates (who are not used to such creativity on the part of the paper
setters), their chances of marking the wrong option despite having done the
mathematical part correctly increase. For every unusual format of questions,
it would help the candidates to illustrate with an example what is asked.
(Way back when English used to be one of the subjects for JEE, even a
question in which adjectives were to be matched with nouns, would carry an
illustration like tall boy and large field, the questions themselves involving
words like fastidious and taskmaster !).
The only good problems in Paper 1 are Q.37 (about the existence of
square roots of matrices) and Q.45 (about members of a concentric family of
circles). In Paper 2, the list includes Q.42 (concurrency of altitudes), Q.44
(about finding upper and lower bound for a sum of 98 integrals) and Q.49
(the part asking on which side of the middle line the dividing line lies).
On the other hand, the list of problems which properly belong to an elim-
ination round, rather than to a round where the final selection is made is
alarmingly large this year. These have been identified in the respective com-
ments. But collectively they cast a serious aspersion on the very purpose of
a two tier selection. Once the screening round has tested the knowledge of
basic facts and skills, the advanced test should focus more on testing intelli-
gence and analytical thinking rather than memory, numerical accuracy and
speed. As commented in the solution to Q.47 of Paper 1, two major purposes
of multiple choice formating, viz. efficiency and uniformity of evaluation, can
still be served even with a non-mechanised evaluation, if questions are asked
in the fill into the blanks form where the answer has ro be expressed as a
whole number or as a reduced rational. (The Main Problems in all twen-
tyfour chapters are of this form.) This will still not test the ability to give

72
reasoning in a coherent form. But it will at least test the presence of a correct
reasoning in many problems. (It is relevant to recall here that the first prob-
lem in Comment No. 11 of Chapter 22 on Finitistic Probability appeared as
a full length question in JEE 1994. The evaluation was done by examiners
193
in those days. Whenever a candidate would display the correct answer
792
at the end of the solution, the examiner would heave a sigh of relief because
there was no need to see how the candidate arrived at this crazy answer.
If this question were to be asked as a multiple choice question, one of the
options would have to be this number and that may give an unwarranted
hint to the candidates.)
Even given the compulsion of asking only multiple choice questions, it is
possible to design imaginative questions. The entrance examination of the
Indian Statistical Institute provides excellent examples. Space considerations
forbid a complete listing of all 30 questions in Paper 1 of 2016. In all these 30
questions, only one of the four options is correct. There are no gimmicks like
questions with more than one answer correct answer, or single digit answers
or matching and, last but not least, paragraphs.
We select two samples, one where the success depends on hitting a smart
idea and the other where the work needed is more methodical but can be
shortened with improvisations.

1. Problem 1: Let X = {a + 5b : a, b ZZ}. An element x X is
called special if there exists y X such that xy = 1. The number of
special elements in X is

(A) 2 (B) 4 (C) 6 (D) 8

Answer and Comments: (A). An element x = a + 5b is the
complex number a + 5ib. Suppose there is some y = c + i 5d X
with xy = 1. Taking absolute values of both the sides

|xy|2 = |x|2 |y|2 = (a2 + 5b2 )(c2 + 5d2 ) = 12 = 1 (1)

As a, b, c, d are integers and a2 + 5b2 and c2 + 5d2 are non-negative the

only possibility is a2 + 5b2 = 1. This forces b = 0 and a = 1.
A person who has studied some algebra will recognise
that the
question asks for the number of units in the ring ZZ+i 5ZZ. Candidates
are not expected to know this. But a clever use of the right trick

73
 (multiplicative property of the absolute value of a complex number)
makes the solution. The brute force method would be to equate the
real and imaginary parts of the product (a + i 5b)(c + i 5d) with
1 and 0 respectively and look for solutions. This is laborious. In this
respect Problem 1 resembles Q.44 of Paper 1 where the multiplicativity
property of the determinant quickly disposes off two of the options. But
the remaining two require something else.

2. Problem 2: The largest integer which is less than or equal to (2+ 3)4
is

(A) 192 (B) 193 (C) 194 (D) 195

Answer and Comments: (B). The problem is supposed to be4 done

without calculators. A direct expansion of the number (2 + 3) gives

(2 + 3)4 = ((2 + 3)2 )2 = (7 + 4 3)2 = 97 + 56 3 (2)
So the problem reduces to determining which of the numbers 95, 96, 97
and 98 (obtained by subtracting
97 from the given four options) equals
the integral part of 56 3. This is best done by directly comparing
squares. Because 56 and 98 have a large common factor, viz. 14, it is
easiest
2 to compare 56 3 with 98. Cancelling 14 and squaring
we get
2
(4 3) = 48 on one side and 7 = 49 on the other. So 56 3 < 98 and
hence the given number is less than 195.

The next easier number to compare 56 3 with is 96 because of
the common
factor 8. Since (7 3)2 = 147 while 122 = 144, we see
that 56 3 > 96.That takes us to the most delicate comparison, viz.
that between 56 3 and 97. Now there are no common factors. Fortu-
nately, the squares are easy to calculate without the brute force method.
(56)2 = (50 + 6)2 = 2500 + 600 + 36 = 3136, while 972 = (100 3)2 =
10000 600 + 9 = 9409. Thus (56 3)2 = 3136 3 = 9408 is only
slightly less than 97. So the integral part of 56 3 is 96 and hence that
of (2 + 3)4 is 193.
Just as there was a smart solution to Problem 1 based on complex
conjugation, the present
problem can also be
done if one considers the
conjugate surd of 2 + 3 defined as 2 3. Analogously to (2), we
have

(2 3)4 = ((2 3)2 )2 = (7 4 3)2 = 97 56 3 (3)

74
 Hence
4
(2 + 3) + (2 3)4 = 194 (4)

(More generally, a similar reasoning would apply if the exponent 4

is replaced by any even positive integer. The irrational terms in the
binomial expansions of the two surds cancel out.)

Now the number 2 3 (0, 1). So all its powers lie in (0, 1). Hence
the second term of (4) is some number between 0 and 1. Therefore the
integral part of the first term is 194 1 = 193.
This problem is quite elementary. But since calculators are not
allowed it tests the ability of a candidate to apply elementary formulas
intelligently to shorten the work. The paper setters, in turn, deserve
to be complimented for choosing the numbers carefully so that such
shortening is possible. (The second solution works only because 2

3 (0, 1).) One fails to understand what ability on the part of a
candidate is tested by asking worn out questions based on the continuity
of the integral part function year in and year out.

To be fair, the JEE paper setters over the years have also come up with
imaginative problems from time to time. (The second solution to Problem
2 is akin to the solution of a 1988 JEE problem given in Exercise (6.27)(a).)
Already we cited some examples from this years papers. Sadly, their per-
centage is declining alarmingly. One reason, perhaps, is the compulsion to
leave no area of the syllabus untouched. It is hard to defend this compulsion
when the candidates have already cleared a test which is designed to cover all
areas. The ISI questions, on the other hand, appear to be selected because
they are good questions which will test a certain ability on the part of a
candidate and not because the questions cover some area of the syllabus and
will test a candidates knowledge of that area.
And, one must not forget that in addition to a paper consisting of well
chosen multiple choice questions, ISI also has another paper in which the
candidaters are asked to answer eight conventional questions. That tests
their ability to give logical reasoning.
It is high time that the JEE takes some cue from ISI. Or else it will
continue to be a test where mediocrity passes as excellence.

75