Professional Documents
Culture Documents
Isi 2017
Isi 2017
(Uploaded on 24/09/2017)
Contents
Introduction 1
Paper 1 3
Paper 2 63
Concluding Remarks 81
Introduction
The commentary on ISI B.Sc. Entrance Test for 2015 was written after
those for 2016 and 2014. The quality of the questions in these two years
had raised expectations. But the first paper of 2015 turned out to be very
disappointing. Also, a look at the question papers prior to 2014 revealed
certain common patterns which were repeated. So the element of novelty
was not so strong as it was when the commentaries for 2016 and 2014 were
written. To some extent such repetitions are to be expected, given the limited
syllabus and the large number of questions to be asked. Instead of asking 30
questions in Paper 1 to be solved in two hours, it would be far better to ask
only about 15 really good questions.
As a result, the plan to write exhaustive commentaries on ISI Entrance
Tests was curtailed to only the four recent ones, viz. from 2014 onwards as
these are the most relevant. A few students admitted through them could
still be studying at ISI.
For reasons far from clear, the official version of the 2017 question papers
has not been uploaded on the ISI website. The present commentary is based
on unofficial versions of the two papers. Revisions will be made when the
official versions are released.
As in the other commentaries, unless otherwise stated, all the references
1
are to the author’s Educative JEE Mathematics, published by Universities
Press, Hyderabad.
I am thankful to Deepanshu Rajvanshi who corrected an omission and
several errors in an earlier draft of the present commentary. Readers are in-
vited to send their comments and point out errors, if any, in the commentary.
They may be sent either by e-mail ([email protected]) or by an SMS or
a WhatsApp message on mobile (9819961036).
2
ISI BStat-BMath-UGA-2017 Paper with Comments
N.B. Each question has four options of which ONLY ONE is correct. There
are 4 marks for a correctly answered, 0 marks for an incorrectly answered
question and 1 mark for each unattempted question.
3
on whether b = H. There is nothing to rule this possibility out. So,
2ac
by taking a, c to be any positive real numbers and setting b =
a+c
(which is the harmonic mean of a and c), we see that it is possible for
a, b, c to be in HP.
4
the musical octave, the pancham is sharper than the madhyam because
of the A.M.-H.M. inequality.
√ √
What about the geometric
√ mean 2f1 ? 2 is an irrational num-
ber. But the frequency 2f1 comes very close to the note called the
teevra madhyam. Summing up, the madhyam, the teevra madhyam
and the pancham are, respecively, the harmonic, the geometric and the
arithmetic means of the shadja and the taar shadja. And the musical
interpretation of the inequality is that the pancham is sharper than the
teevra madhyam, which, in turn, is sharper than the shudh madhyam.
Q.2 A unit square has its corners chopped off to form a regular polygon
with eight sides. What is the area of this polygon?
√
√ √ √ 2 7
(A) 2( 3 − 2) (B) 2 2 − 2 (C) (D)
2 9
Solving,
1
b= √ (2)
2+ 2
√
2
We now get the length of the side of the octagon as √ . There is
2+ 2
a formula for the area of a regular n-gon in terms of the length of its
5
side and using it we can answer the question. But that would be too
elaborate. In the present case, the octagon is the figure that results by
removing the four corners from the unit square. So its area, say A, is
obtained by subtracting the removed area which is simply 2b2 from 1.
Hence,
2
A = 1− √
(2 + 2)2
2 1
= 1− √ =1− √
6+4 2 3+2 2
√
2+2 2
= √ (3)
3+2 2
As it stands, this answer does not match with any of the given options.
To remedy this, we rationalise the denominator. Thus
√ √
(2 + 2 2)(3 − 2 2)
A = √ √
(3 + 2 2)(3 − 2 2)
√ √
= (2 + 2 2)(3 − 2 2)
√
= −2 + 2 2 (4)
which matches with (B).
6
(A) 25 (B) 54 (C) 126 (D) 150
5 1 X X X 1
5
1 X X X 1
5
1 X X X 1
5 1 1
5 1 1 1 1 1
(a) (b)
In (b) of the figure we show one face of the original cube divided into
25 squares of side 1 each. Out of these 9 lie in the interior while 16
lie along the boundary. The latter will also have another face lying
along some other face of the cube and hence is painted. So there are
only 9 smaller cubes which have exactly one face painted. They are
different for the different faces of the bigger cube. As the big cube has
six faces, the number of smaller cubes having exactly one face painted
is 9 × 6 = 54.
7
we are counting each cube twice because it lies along an edge which is
the common boundary of two adjacent faces. So, the correct answer
is not 96 but only 48. But even this is not correct, because cubes
which are at the eight vertices of the big cube have three of their faces
painted. Together they count for 24 smaller painted squares. So we
need to subtract 24 from 96 and then divide by 2. That gives 36.
A slightly easier approach is to count only the small cubes that lie
along an edge but not at either of its two ends. There are three such
small cubes along each edge. As there are twelve edges, the number
of small cubes which have exactly two of their faces painted is 36.
Summing up, each of the 125 smaller cubes can have at most three
of its faces painted. The number of small cubes with 0, 1, 2 and 3
faces painted is, respectively, 27, 54, 36 and 8. Their total is 125 as
expected. Also the total number of painted faces (of smaller cubes) is
27 × 0 + 54 × 1 + 36 × 2 + 8 × 3 = 54 + 72 + 24 = 150, also as expected
because each of the six faces of the big cube contains 25 faces of the
small cubes.
z−i
Q.4 Let z be a complex number such that is purely imaginary. Then
z−1
the minimum value of |z − (2 + 2i)| is
√ √ 3 1
(A) 2 2 (B) 2 (C) √ (D) √
2 2
8
z−i
condition that the ratio is a purely imaginary number. A brute
z−1
force method in dealing with problems about complex numbers is to
resolve them into real and imaginary parts. In the present problem,
for example, if we write z = x + iy as usual, then the given condition
means
x + i(y − 1)
= ib (1)
(x − 1) + iy
for some real number b. Multiplying both the sides by the complex
conjugate of the denominator,
Equating the real parts of both the sides as well as their imaginary
parts will give a system of two equations in the variables x and y with
b as a parameter. Eliminating b between these two equations will give
us the equation of the curve C.
It is obvious that method will entail considerable algebraic manipula-
tions. But there is a way to interprete the data geometrically. Let z, 1, i
and 2 + 2i represent the points P, A, B and Q in the Argand diagram.
Then the complex numbers z − i represents the line segment from B to
P . Similarly, the complex number z − 1 represents the segment from
A to P . The condition that the ratio of these two numbers is purely
imaginary is equivalent to these two segments being perpendicular to
each other as shown in the figure below.
Q. 2+2i
Pz
i B N 1+ i
.
M C
A x
O 1
9
Since 6 AP B = 90◦ , the locus of P is the circle with AB as a diameter
1+i
and hence M = as its centre. We have called it C. Since it is a
2
circle, the point, say N, on it which is closest to the point Q = 2 + 2i
is the closer end of the diameter passing through Q. As the centre of
1+i
C is the point , it is clear that N is the point 1 + i. Its distance
2 √
from Q is |1 + i| i.e. 2.
Q.5 Let f : IR −→ IR be a continuous function such that for any two real
numbers x and y,
Then
(A) f (101) = f (202) + 8 (B) f (101) = f (201) + 1
(C) f (101) = f (200) + 2 (D) None of the above.
10
done as follows. Fix any x ∈ IR. Then for any h 6= 0, we have
f (x + h) − f (x)
0≤ − 0 ≤ 7|h|200 (1)
h
where A and α are some positive constants. Then the argument above
will apply equally well if α > 1. For 0 < α < 1, little can be concluded
beyond continuity of f (x). The case α = 1 is most interesting. It is
tempting to think that in this case f (x) must be of the form ±Ax + B
for some constant B. But there are other examples too which we shall
not go into.
It is entirely possible that there is some hidden agenda in asking
this problem. Perhaps the idea is to compare the performance on this
question with that on the rest. If candidates who do well elsewhere fail
in this problem, there is a cause for concern for their training.
11
group has the same number of first year students, the same number
of second year students and the same number of third year students.
What is the smallest possible size of each group?
It takes more time to read the problem than to work it out. The only
‘thoughtful’ step is to realise that the number of groups is a common di-
visor of 60, 84 and 108. After that, it is a problem of elementary school.
There they fit in nicely, because one of the criteria of good education
is the ability to apply it to real life problems. An alarmingly large per-
centage of college entrants simply dread ‘word problems’. Perhaps it
is to prevent their entry that such a simple problem is designed, which
properly belongs as a puzzle in the recreation section of a newspaper.
(A similar remark was made about Q.20 in Paper 1 of 2015. But that
problem required at least some deductive reasoning. The present one
requires little more than finding the g.c.d. of three numbers which are
from the multiplication table of 12.)
Looking at the social aspects of the data, it would have been better
if the figures 60, 84 and 108 stood for students enrolled for degree
programs in Mathematics, Physics and Chemistry respectively. That
will be a realistic reflection of the relative popularity of these subjects.
As the problem stands, one gets the impression that the Mathematics
program of the college is declining in popularity!
12
If the value of b is 9, then the value of a must be
√
(A) 3 81 (B) 27
2
(C) 18 (D) 27
y = log 9 = 2 (3)
13
when all the numbers are equal and hence each one of them is 1 here.
So now we get three separate equations:
Q.8 Consider a triangle ABC. The sides AB and AC are extended to points
D and E respectively, such that AD = 3AB and AE = 3AC. Then
one diagonal of BDEC divides the other diagonal in the ratio
√ √
(A) 1 : 3 (B) 1 : 3 (C) 1 : 2 (D) 1 : 2
B C
D E
14
Q.9 The area of the region bounded by the curve y = tan x, the x-axis and
the tangent to the curve at x = π4 is
1 1
(A) loge 2 − 2
(B) loge 2 + 2
(C) 21 (loge 2 − 12 ) (D) 21 (loge 2 + 12 )
y
L
B( π , )
/4 1
E y=c
F A
C π/4 x
O
R T
(i) We can divide the region R into two subregions, one lying on the
left of the vertical line through the point C and the other to its
right. For this, we shall first have to identify the point C and for
this we shall first have to find the equation of the line L. The area
15
of each subregion will have to be calculated by an appropriate
integral.
(ii) Instead of splitting the region R, we add to it the triangle ABC
(shown as T in the figure) to get a bigger region S, which is
bounded above by the curve y = tan x and below by the segment
OA. We calculate the area of S and subtract from it the area of
the triangle T .
(iii) We can evaluate the area of R by horizontal slicing. A typical
horizontal line y = c intersects the region R in a segment EF
where E = (tan−1 c, c) and F will come out after we find the
equation of the line L. If this point is (g(c), c), then the area of R
Z 1
will be tan−1 y − g(y)dy.
0
(i) For the first approach, we shall have to calcuate two integrals
π
Z π/4−1/2 Z π/4
tan x dx and (tan x − 1 − 2(x − )dx.
0 π/4−1/2 4
Z π/4
(ii) Here we shall have to calculate the integral tan x dx and sub-
0
tract from it the area of the triangle T .
(iii) From (1), the function g(y) comes out as + y−1
2
π
4
. The area of the
1 π 1 −y
Z
region R is then the integral tan−1 y − − dy.
0 4 2
16
There is not much conceptual difference between (i) and (ii). If we split
Z π/4
π
the second integral, viz. (tan x − 1 − 2(x − )dx into two parts
π/4−1/2 4
by splitting the integrand and add the first part to the first integral,
Z π/4−1/2
viz. tan x dx, then by the additivity property of the integrals
0 Z π/4
we get the integral tan x dx which is precisely the integral in (ii).
0
The integral of the remainder is the area of the triangle T . The third
approach, however, is qualitatively different. Here we get the answer
by a single integration. But we shall have to find an antiderivative for
tan−1 y. This can be done using integration by parts. But that means
we shall be spending what we saved.
This often happens in mathematics. No matter which approach you
take, the work involved remains essentially the same. It is only in rare
cases that we can bypass some arduous work by cleverly transform-
ing the problem. In the solution to Q.2, we mentioned complementary
counting as an example of such a transformation. In the present prob-
lem, the second approach resembles complementary counting because
the area of the triangle T can be found without integration. Its base is
π π 1 1
− ( − ) which is simply while its height is 1. So the area of the
4 4 2 2
1
triangle T is . We need to subtract this from the area of S which is
Z 4π/4
the integral tan x dx. Hence the area of the region R is given by
0
π/4 1
Z
A(R) = tan x dx −
0 4
π/4 1 √ 1
= loge (sec x) − = loge 2 − loge 1 −
0 4 4
1 1 1 1
= loge 2 − 0 − = (loge 2 − ) (3)
2 4 2 2
which tallies with (C).
17
carry equal marks and it is next to impossible to finish all the questions
in the given time, those who pick the less demanding problems stand
to gain. Exactly the same problem has been asked in JEE 1988. See
Exercise (17.3)(ix). Admittedly, most problems on finding areas are
similar. Still such an outright duplication should have been avoided.
Q.10 Let V be the set of vertices of a regular polygon with twenty sides.
Three distinct vertices are chosen at random from V . Then the prob-
ability that the chosen triplet of vertices forms a right angled triangle
is
7 1 3 1
(A) 19
(B) 19
(C) 38
(D) 38
|F |
p= (1)
|S|
where S stands for the set of all possible outcomes and F stands for the
subset of F consisting of the outcomes that are ‘favourable’, i.e. those
outcomes in which the given event occurs. Here |X| stands for the
number of elements in a set X. The set S is often given a fancy name,
the sample space. This simple minded definition of probability is
applicable only when the sample space is finite and, more importantly,
when all elements of it are equally likely. This is the meaning given to
expressions like ‘at random’.
In the present problem, the sample space S is the set of all ways to
select three elements of the vertex set V . As |V | is given as 20, we get
!
20 20 × 19 × 18
|S| = = = 19 × 60 (2)
3 3!
Note that we have not multiplied 19 and 60. To get p, we have to divide
by |S| and at that time the factorisation 19×60 will be more convenient,
especially so since in all the four options given, the denominators are
multiples of 19. So it is foolish to waste precious time in multiplying 60
18
and 19 prematurely and also thereby increase the chances of a numerical
slip.
Now comes the real task of finding |F |, the number of favourable
cases. In the present problem, this is the set of those triples {A, B, C}
in which A, B, C are vertices of a right angled triangle. As the order is
unimportant, we suppose that ABC is right angled at B. We are given
that the angular distance between any two consecutive vertices is the
same, say, α. (Actually, α = 2π 20
π
= 10 = 18◦ . But that is not crucial
here.) Hence the angular distance between any two (distinct) vertices
will be a multiple of α, ranging from α to 10α.
Consider a triangle ABC right angled at B. Without loss of general-
ity, assume that the side AB is smaller than (or possibly equal to) the
side BC. Then the angular distance between A and B is nα for some
n = 1, 2, 3, 4 or 5. For each n, there are twenty such segments whose
end points are at an angular distance nα. Each segment determines
two possible locations for the third vertex C. (These two positions
along with A and B form a rectangle.) So, for every n, there are 40
right angled triangles in which the shorter side has angular length nα.
However, for n = 5, AB and BC are equal in length and so the same
triangle gets counted twice. So the total count of triangles is not 40 × 5
but 40 ×4 + 20 = 180. So the favourable set F has 180 elements. Hence
180 3
by (1) and (2), p = = .
19 × 60 19
19
It is given that B can be obtained from A by finitely many basic row
operations. Then, the value of x is:
Another simple problem, once the key idea strikes. And the key
idea is not at all foreign. But it is one thing to know a fact. It is
quite another to realise where it will be applicable. In the elegant
solution to the last problem, the key fact needed was that the angle in
a semi-circle is a right angle. It is easy to hit this if you start with a
semi-circle. But it is not so easy to hit it if you start with a right angled
triangle. In the present problem, however, the moment somebody sees
the words ‘adding a multiple of one row to another’ the bell ought to
ring that the determinant is invariant. If it doesn’t, there is no other
way to solve the problem. In the last question, on the other hand,
even if you missed the short cut, the answer can be obtained by other
means. This is probably the dividing line between an elegant solution
and a tricky one. An elegant solution does not preclude other, possibly
20
more laborious approaches. In fact its elegance stands out more in
comparison to them. In a tricky solution, however, if the trick eludes
you, you are helpless.
A = πr 2 (1)
T t
P
r
2t
21
From the right angled triangle OT P we get
r 2 + t2 = 4t2 (2)
So t2 = 13 r 2 . As a r is a constant so is t and hence 2t. Therefore P traces
a circle of radius 2t centred at O. Its area is 4πt2 = 43 π3t2 = 34 πr 2 = 43 A.
22
Squaring and noting that |OA|2 = 2a2 while |OB|2 = 2b2 , we get
a2 b2 = C 2 (4)
We get the locus of P by eliminating a and b from (1), (2) and (4).
Solving (1) and (2) for a and b,
b = h+k (5)
and a = h − k (6)
i.e. (x2 − y 2 )2 = C 2 .
y − k = m(x − h) (8)
In other words, we are taking the slope of the moving line as a param-
eter. Solving (8) simultaneously with x + y = 0 and x − y = 0 we get
the points A and B. They come out to be
mh − k k − mh
A = ( , ) (9)
m+1 m+1
mh − k mh − k
and B = ( , ) (10)
m−1 m−1
23
respectively. As P = (h, k) is the midpoint of AB, we get
1 1
2h = (mh − k)(+ )
m+1 m−1
1 1
and 2k = (mh − k)( − ) (11)
m−1 m+1
both of which simplify to
h = mk (12)
As a result, we have
(mh − k)2 (h2 − k 2 )2
|OA|2 = 2 2
= 2 2
= 2(h − k)2 (13)
(m + 1) (h + k)
2
(mh − k) (h2 − k 2 )2
and |OB|2 = 2 2
= 2 2
= 2(h + k)2 (14)
(m − 1) (h − k)
Putting these into (3) will give us the same locus as before. But this
derivation is considerably more complicated than the earlier one which
was based on two parameters a and b instead of a single parameter m.
As a general rule, one should minimise the number of parameters. But
as this problem shows, sometimes a clever choice of several parameters
simplifies the work.
24
The condition on f means that the six elements 1, 2, 3, 4, 5 and 6 of
the domain are paired off into three pairs and f takes the same value
on the two elements of the same pair, but different value on those of
the other pairs. (Often such function is called a two-to-one function.)
For example, consider the pairing {1, 5}, {2, 3}, {4, 6} of A. Then f
is uniquely determined by the values f (1), f (2) and f (4), 1, 2, 4 being
the representatives of the three pairs. But these three values must be
different. So, for each pairing, f is like a one-to-one function from a set
with 3 elements (viz., the set of the three pairs in that pairing) to the
set B = {a, b, c, d, e}. Now this problem is very similar to the utterly
trivial problem mentioned above. The value on the first pair can be
chosen in 5 ways. For each such choice the value on the second pair
can be chosen in 4 ways and continuing, that on the third in 3 ways.
So, for each pairing of the domain set, we have 5 × 4 × 3 = 60
elements in F . We now only need to multiply this figure by the number
of possible pairings. In any pairing, the ‘mate’ of 1 can be chosen in
5 ways. Once selected, the remaining 4 elements can be paired in 3
different ways.
Summing up, there are 15 possible pairings and for each pairing
there are 60 functions in F . So the total count is 15 × 60 = 900.
A good problem, certainly far better than the 2015 question which
is absolutely unbecoming for a reputed institute. The reasoning above
looks long because we have given it in detail. But it does not take much
time to conceive it mentally and thereafter the calculations involved are
minimal.
The problem of pairing an even number, say 2n of objects into
n mutually disjoint pairs is interesting. Denote this number by an .
Then it is the number of ways 2n participants in a tournament can
be paired for the first round of matches. Clearly, a1 = 1 and a2 = 3.
In the solution above, we calculated a3 as 15. The formula for an can
be obtained by first writing down a recurrence relation for it and then
solving it. Let S be a set with 2n objects. Fix any one object, say x of
S. In any pairing, the ‘mate’, say y of x can be chosen in 2n − 1 ways.
Once it is chosen, the problem reduces to pairing off the elements of
the set S − {x, y} which is a set with 2n − 2 elements. There are an−1
25
pairings of it. So, {an } satisfies the recurrence relation
an = (2n − 1)an−1 (1)
for n ≥ 2 with the initial condition a1 = 1. This is very easy to solve
by inspection. We start with 1 and thereafter, every time we multiply
the earlier term with the next odd number. So
an = (2n − 1)(2n − 3) . . . × 5 × 3 × 1 (2)
By supplying the missing even factors, this can also be written as
(2n)!
an =
(2n)(2n − 2)(2n − 4) . . . 4 × 2
(2n)!
= (3)
2n n!
6! 5! 120
In particular, a3 = = = = 15 as we already calculated.
8×6 8 8
Q.15 Two persons, both of height h are standing at a distance h from each
other. The shadow of one person cast by a vertical lamp-post placed
between the two persons is double the length of the shadow of the other.
If the sum of the lengths of the shadows is h, then the height of the
lamp-post is
√ √ √
3 1+ 2 3+1
(A) 2
h (B) 2h (C) 2
h (D) √
2 2
h
P x Q y R
h h h
C 2h/3 A x E y B h/ 3 D
26
Suppose that the persons are standing at the points A and B and the
foot E of the lamp-post EF is at a distance x from A and y from B.
Then
x+y = h (1)
Let P and R be the heads of the persons and let their shadows fall at
points C and D on the ground. Then F, P, C are collinear and so are
F, R, D. We are given that the lengths of the shadows add to h while
one of them is double the length of the other. We take AC to be the
larger shadow of length 2h/3 and BD to be the shorter one of length
h/3.
We are interested in the length of EF . Normally, we should denote
it by a single symbol. However, here we let Q be the point where
EF meets P R and let z be the length QF . Then the length of the
lamp-post is
h+z (2)
z = 3y (5)
From (4) and (5), x = 2y. Hence from (1), we have x = 2h/3 and
y = h/3. So, from (4) (or (5)) z = h. Therefore by (2), the height of
the lamp-post is h + h = 2h.
27
demanding by giving the distance between the persons to be different
than their common height, and still more demanding by giving the
two persons to have different heights. But that would only increase
the labour involved without testing any new ability. The paper setters
have wisely refrained from such generalisations.
Q.16 Let S be the set of all points z in the complex plane such that
4
1
1+ = 1.
z
Then, the points of S are
28
As S has only three elements, option (A) is ruled out. To decide which
of the remaining options holds, let us write the second and the third
elements of S in the standard form a + ib with a, b real.
1 −i − 1 −1 − i 1 1
= = =− − i (4)
i−1 (i − 1)(−i − 1) 2 2 2
and similalry,
1 i−1 −1 + i 1 1
= = =− + i (5)
−i − 1 (−i − 1)(i − 1) 2 2 2
Hence, with these cosmetics,
1 1 1 1 1
S = {− , − − i, − + i} (6)
2 2 2 2 2
Clearly all three elements have the same real part, viz. − 21 . So all of
them lie on the vertical line x = − 12 . hence they are collinear.
z+∞=z−∞=∞ (7)
for all z ∈ C,
|
z∞ = ∞ (8)
29
∞. For example, even though z1 + ∞ = z2 + ∞, we may not have
| ∗ is a
z1 = z2 . Despite these limitations, the extended complex plane C
very convenient device to take care of some anomalies that arise in C.|
| ∗ , there is
Collinearity is not lost because ∞ lies on all lines. In fact in C
no difference between lines and circles. A line is a circle which contains
the point ∞. The points ±1, ±i all lie on the unit circle |z| = 1.
This circle is the image of the line x = − 12 under the transformation
1 z+1
which takes z to 1 + = . This transformation belongs to an
z z
important class of transformations of C | ∗ to C| ∗ called the fractional
az + b
linear transformations because they are of the form i.e. ratios
cz + d
of two linear transformations. We could go on in this wonderland of
C| ∗ . But that is not relevant to the present problem. We mentioned it
because those who study complex numbers in some depth are sure to
∗
encounter C | .
30
P
N
d
h
α
θ Q
10
M
31
If, further, the candidate has to figure out exactly the purpose of the
problem, it is unfair when the time allotted is meagre (four minutes).
Fortunately, in the present problem, once these hurdles are crossed,
the rest of the work is simple. We express the illuminance as a function
of θ and maximise it. Clearly
d = 10 sec θ (1)
32
Q.18 Let f and g be two real-valued, continuous functions defined on the
closed interval [a, b], such that f (a) < g(a) and f (b) > g(b). Then the
area enclosed between the graphs of the two functions and the lines
x = a and x = b is always given by
Rb Rb
(A) |f (x) − g(x)|dx (B) | f (x) − g(x)dx|
a a
Rb Rb
(C) |f (x)| − |g(x)|dx (D) ||f (x)| − |g(x)||dx
a a
33
zero function g(x). More generally, when f (x) and g(x) are any two
functions, in some parts of [a, b] we may have f (x) ≥ g(x) while in
the rest we have f (x) ≤ g(x). But no matter which possibility holds
Z b
where, |f (x) − g(x)|dx is the area between the graphs of f (x) and
a
g(x). So (A) is correct. Once again, this trick merely gives a succinct
formula for the area. When it comes to actually calculate the area
between the graphs y = f (x) and y = g(x), there is no short cut to
splitting the interval [a, b] into subintervals, in each of which f (x) lies
either entirely above or entirely below g(x). For example, to calculate
the area between the graphs of y = sin x and y = cos x for 0 ≤ x ≤ π,
we need to spilt it into [0, π/4] and [π/4, π].
The present problem tests whether the candidates have been taught
this care. The hypothesis f (a) < g(a) and f (b) > g(b) is really redun-
dant as the question asks to pick up the choice which is always correct.
The hypothesis merely draws attention to the fact in this case it is
mandatory to take the integrand as |f (x) − g(x)|.
Then f is
34
In the present problem, one obvious point where the function f
changes its definition is, of course, 0. It is given right in the statement
of the problem. But usually there are some hidden points too. For
x > 0, x and x2 are equal at 1. But in the interval (0, 1), x2 is smaller
than x while for x > 1, x is the smaller. So, for x > 0 the definition of
f (x) becomes
(
x3 − x2 if 0 < x ≤ 1
f (x) = (1)
x2 − x if x > 1
From (2) and (3), we see that f is differentiable at 1. Also since f (0) =
0, f (x) is continuous from the right at 0 and (4) gives us one sided
information about its differentiability at 0. To see what happens on
the other side of the border, i.e. for x < 0. We first need the analogue
of (1). At x = −1, x and x1 equal each other. But if −1 < x < 0, then
1
x
< x while for x < −1, x < x1 . (To see such things more easily, give
some particular values to x, e.g. x = − 12 in the first case and x = −10
in the second.) So analogously to (1) we have
(
x2 if x < −1
f (x) = (5)
1 if − 1 ≤ x < 0
35
Clearly, f is continuous at −1. But at x = 0, lim− f (x) = 1 while
x→1
f (0) = 0, So f is not continuous at 0 and hence automatically not
differentiable at 0 either.
It remains to check differentiability of f at −1. Analogously to (2) and
(3) we have
36
Answer and Comments: (C). The L.H.S. of the given inequality is
the average value, say A, of f over the interval [0, 2], while the R.H.S.
is the value of f at a particular point c in the domain. (Here c happens
to be the right end point of the interval [0, 2]. But that makes no
difference.)
When the average is less than a particular functional value f (c), it
does not follow that the function is less than f (c) throughout. This is
sheer common sense. If the average income of some community is, say
Rs.10, 000.00 and some particular person x has an income of Rs. 50,000
it does not follow that he is the richest member of the community. The
community may very well have some persons richer than x, but many
other persons poorer than x can pull the average down. All that can
be said definitely is that such a person cannot be the poorest person,
for if he were then everybody’s income will be at least Rs. 50,000.00
and hence so will be the average income for the community.
The present problem requires the continuous analogue of the rea-
soning in this example. Suppose (C) fails, that is f has a minimum at
2. Then
whence
Z 2 Z 2
f (x) dx ≥ f (2) dx = 2f (2) (3)
0 0
37
Q.21 In a triangle ABC, 3 sin A + 4 cos B = 6 and 4 sin B + 3 cos A = 1 hold.
Then the angle C equals
which simplifies to
which reduces to
38
where α is the unique angle for which cos α = √6 and sin α = √1 .
37 37
Then we would get
11
B = cos−1 ( √ )+α (9)
148
Similarly, from (6) we can determine A. Having found A and B, C
will come out as π − A − B. But it will require a lot of calculation to
determine C in degrees. So we look for a simpler solution, keeping in
mind that
39
somebody is really bent on using both (4) and (6), a salvaging feature
is that from (6) we get cos(A − β) = √537 where β is the complementary
angle of the angle α in (9). So, adding, A + B would come out to be
π
2
+ cos−1 ( √11
148
) + cos−1 ( √637 ). This can indeed be simplified to 150◦ .
It is important to note that such a succinct answer as 30◦ was
possible only because the numerical data permitted it. If, instead of
(1) and (2), we had a more general system of equations, viz.
a1 sin A + b1 cos B = c1 (14)
and a2 sin B + b2 cos A = c2 (15)
our first method would still work and will give B and A and finally C.
But the answers would be horrendous. The paper setters deserve to
be commended for giving the numerical data which makes the solution
manageable.
π
Q.22 Let θ = 7
and consider the following matrix
!
cos θ − sin θ
A=
sin θ cos θ
40
for all α, β. In particular, taking α = β = θ, we have
Note that the matrix Aθ in the statement of the problem is the same
as the matrix Acos θ+i sin θ associated with the complex number cos θ +
i sin θ. Therefore, by repeated applications of (6), (Aθ )n corresponds to
the complex number (cos θ + i sin θ)n . By DeMoivre’s rule,
Therefore, (3) holds for every positive integer n. (Some persons denote
the number cos θ + i sin θ by eiθ . Then a quick proof of (7) comes from
(eiθ )n = einθ . But unless we have independently defined and proved
41
the basic properties of the complex exponential function, this proof is
nothing more than a notational gimmick.)
Yet another interpretation of the matrix Aθ is provided by the
concept of a rotation. Let Tθ : IR2 −→ IR2 denote an anticlockwise
rotation of the plane around the origin O, through an angle θ. Clearly
Tθ (O) = O. For any other point P = (x, y) in IR2 , Tθ (P ) can be
calculated
√ as follows. Let r and α be the polar coordinates of P . Then
r = x2 + y 2 and α is the angle OP makes with the positive x-axis.
The cartesian coordinates x, y can be obtained from r and α by
x′ = r cos(θ + α)
= r(cos θ cos α − sin θ sin α)
= cos θx − sin θy
!
x
= (cos θ − sin θ) (9)
y
42
and (x′ , y ′) which is more usual, we see that the effect of applying the
!
x
rotation Tθ on P is the same as premultiplying the column vector
y
by the matrix Aθ .
Now consider two rotations Tα and Tβ . It is clear that if we first
apply Tβ and then Tα , the net result is the rotation Tα+β . In other
words
Tα+β = Tα ◦ Tβ (12)
!
x
Given any point P = ∈ IR2 , let Q = Tβ (P ) and R = Tα (Q).
y
! !
x′ x′′
Denote Q and R by the column vectors ′ and respectively.
y y ′′
We apply (11) repeatedly, first with θ = α, then with θ = β to get
! !
x′′ x′
= Aα
y ′′ y′
!
x
= Aα Aβ (13)
y
for all x, y ∈ IR2 . From this we cannot hastily conclude that Aα+β =
Aα Aβ , because for matrices it may very well happen that AC = BC
and still A 6= B, even when the matrix C is not the zero matrix. But in
the present case, we are given that (15) holds for all x, y. For any 2!× 2
1
matrix C, it is easy to check by a direct computation that C is
0
simply the first column of C. Hence applying (15) with x = 1, y = 0 we
43
get that the first column of Aα+β is the same as that of Aα Aβ . Similarly
by applying (15) with x = 0, y = 1 we get that their second columns
are the same. We are now justified in saying that we have an alternate
proof of (1).
Of course, a proof based on a direct computation is much shorter.
But interpreting matrices as transformations and their product as the
composite of transformations opens a whole new branch of mathematics
called linear algebra. The theory of determinants or that of systems
of linear equations properly belongs to this branch of mathematics.
Q.23 Consider all the permutations of the twenty six English letters that
start with z. In how many of these permutations the number of letters
between z and y is less than those between y and x?
(A) 6 × 23! (B) 6 × 24! (C) 156 × 23! (D) 156 × 24!
44
bet, one certainly wishes that they used the correct English grammar!
The problem is in the form of a direct question. The verb then has
to be split and the auxiliary part of it must come before the subject.
Thus the word ‘is’ ought to have come before the phrase ‘the number’.
Anyway coming to the mathematical essence of the problem, suppose
the gap between z and y is m letters. Here m can be 0, if there is no
gap, i.e. y comes immediately after z. Now since the gap between y
and x is more than m, x cannot come between z and y. So it has to
come after x and after a gap of at least m + 1 letters.
We classify the permutations according to the value of m. When m = 0,
the permutation begins with
zy×
z×y×× (2)
The general pattern should be clear from (1) and (3). We expect that
for m = 2 the number of permutations will be
19 × 23! (4)
45
which can be verified by observing that every such permutation must
begin with
z × ×y × ×× (5)
Thus every time we are getting 23! times an odd number. These odd
numbers start from 23, 21, 19, . . .. So they will end with 3 and finally
1. (We can determine the m for which the coefficient of 23! would be 1.
But that is only an additional labour.) So the answer to the problem
is
23! × (23 + 21 + 19 + . . . + 5 + 3 + 1) (6)
The expression in the parentheses is the sum of an A.P. of length 12.
There is, of course, a formula for it. But the best way is to add them
the way the great mathematician Gauss did when he was 7 years old.
We add together the two end figures 23 and 1 to get 24. Then we add
the ends of the remaining, i.e. 21 and 3 to still get 24. This will go on
till we add the middle two numbers 13 and 11 which also gives 24. So
the total count is
6 × 24 × 23! = 6 × 24! (7)
which tallies with (B).
46
and Q are given to be diametrically opposite to each other. Hence all
vertices of it are at a distance 1 from the origin. By a direct calculation,
this is the case for the points in all the four options. So, we cannot
get the odd man out using this preliminary criterion. We need to dig
deeper.
The angular distance between any two consecutive vertices of the
2π π π
polygon is = . The ray OP makes an angle with the x-axis.
12 6 4
So for every other vertex, say V of the polygon the angle OV makes
π kπ
with the x-axis will be of the form + where the integer k will run
4 6
π
from 1 to 11. (For k = 6 we get the angle π + and the corresponding
4
vertex is Q as expected.)
So, to nab the culprit, for each of the given four options, say V ,
we determine the angle which OV makes with the x-axis. Here we are
lucky. The number in the option (D) is more popularly denoted by
ω. It is an imaginary cube root of 1. It can be written in the form
(cos(2π/3), sin(2π/3)). So if V is the vertex in (D), then the angle
2π 2π π
OV is . It this were to be a vertex of the polygon, then −
3 3 4
π
will have to be an integral multiple of . But by a direct calculation,
6
2π π 5π π
− = which is not an integral multiple of . So (D) is correct.
3 4 12 6
47
The R.H.S. is non-negative. The L.H.S. is a quadratic in a with a
negative leading coefficient. So its value is non-negative only when a
lies between the two roots 0 and 1. Also the maximum value of the
1
L.H.S. occurs at a = 12 and this maximum is . So this is the maximum
4
b2 + c2 can be. Subject to this, we have to minimise b. This will be the
case when b2 is the maximum and b is the negative square root of this
maximum. For b2 to be maximum, the L.H.S. should be maximum and
1 1
c2 should be minimum. So, we take a = and c = 0. Then b2 = .
2 4
Therefore
1 1
− ≤b≤ (2)
2 2
1
Hence the minimum of b is − . For any smaller value of b, b2 will
2
1
exceed and then (1) cannot hold for any a and c.
4
A simple problem on minimisation but does not fall into any standard
type such as the calculus method or the A.M.-G.M. inequality. It is a
good test of the ability to think fresh rather than be guided by a large
number of problems of the familiar type.
48
ex − 1
cal, i.e. lim because this limit is, by very definition, the deriva-
x→0 x
1
tive of the function ex at 0. It exists and equals 1. So lim f (x) = = 1.
x→0 1
Hence f is continuous at 0.
For differentiability too, we can consider the reciprocal of f (x). But
unlike limits, the derivative of a function is not so easily related to that
of its reciprocal. So we proceed directly. We have to consider the limit,
say L, defined by
f (x) − f (0)
L = lim
x→0 x
x
− 1
= lim e −1
x
x→0 x
x − ex + 1
= lim (1)
x→0 x(ex − 1)
0
This is a limit of the indeterminate form . So we try l’Hôpital’s rule
0
to evaluate it. Taking the ratio of the derivatives of the numerator and
the denominator, the limit L will be
1 − ex
L = lim (2)
x→0 xex + ex − 1
0
provided this limit exists. But this is again of the form. So, we
0
invoke l’Hôpital’s rule again to get
−ex
L = lim (3)
x→0 xex + 2ex
provided, again, that this limit exists. But this time we have no trouble.
As x → 0, the numerator tends to −1 while the denominator tends to
1
2. So L = − 12 . Therefore f ′ (0) exists and equals − .
2
49
The problem can also be solved by advanced methods, which are often
used at the Junior College level without justification. For example, we
can take the Taylor series of ex at 0 (also called MacLaurin series) viz.
x2 x3 xn
ex = 1 + x + + + ...+ + ... (4)
2! 3! n!
If we substitute this into (1), the numerator becomes
x2 x3 xn
− − − ...− − ... (5)
2! 3! n!
while the denominator becomes
2 x3 x4 xn+1
x + + + ...+ + ... (6)
2! 3! n!
In both (5) and (6), the dominating power near 0 is x2 . So the ratio of
(5) by (6) tends to the ratio of the coefficients of this power, i.e. − 21 .
What is involved in this argument is not only the power series
expansion of ex , but also the operation of equating the limit of a sum
of infinitely many terms with the sum of the limits of the individual
terms. Such arguments need justifications. So, it would have been
better to ask this question in Paper 2 which would have shown what
type of justification the candidate has given.
50
Answer and Comments: (D). All the options involve the increas-
ing/decreasing behaviour of f (x). So the foremost method is to find
f ′ (x). If that fails (e.g. if f happens to be non-differentiable), then we
have to look for other methods.
But, in order to find f ′ (x) we must first find f (x)! And, in the
present problem, the way f (x) is defined, there is no way to apply
some standard formulas to differentiate it. So, we first have to find
some familiar expression for f (x).
Note that for a fixed x, f (x) is defined as the maximum of a function
g(y) of y defined by
|x − y|
g(y) = (1)
x+y+1
for 0 ≤ y ≤ 1. Because the absolute value changes its definition at 0,
and x lies in the same interval as y does, we have to consider two cases
depending upon whether y ≤ x or y ≥ x. Doing so, (1) splits as
( y−x
x+y+1
, if 0 ≤ y ≤ x
g(y) = x−y (2)
x+y+1
, if x ≤ y ≤ 1
which is highly comfortable if for no reason than its familiar form.
There is hardly a test at the 10+2 level where there is no problem
which involves a function defined by different formulas for different
subintervals. So the very sight of the big left curly bracket makes us
feel that we are now on a familiar turf.
We have to maximise this for 0 ≤ y ≤ 1, holding x as a constant.
But since g(y) has a different definitions on [0, x] and [x, 1], we have to
separately find the maximum of g(y) on each of these two subintervals.
Let us call the maximum of g(y) on [0, x] as M1 (x) and its maximum
on [x, 1] as M2 (x). After we calculate them, the greater of the two
will be the maximum of g(y) on the entire interval [0, 1]. (This takes as
much intelligence as is needed to understand that to decide the national
champion in some sport, if the country is divided into two zones, then
a match is played between the two zonal champions and the winner is
the national champion.)
Let us first tackle M1 (x), the maximum of g(y) on [0, x]. Here
x−y
g(y) = . We can use calculus to find the maximum on [0, x].
x+y+1
51
But there is a simpler way. The numerator and the denominator are
both positive. But as y increases the numerator decreases (because of
the negative coefficient of y) while the denominator increases. So the
ratio must decrease. Hence the maximum of g(y) on [0, x] occurs at
the left end point 0. In other words,
x
M1 (x) = g(0) = (3)
x+1
With the second interval [x, 1] we are not so lucky. This time
y−x
g(y) = . Now both the numerator and the denominator in-
x+y+1
crease with y and so our earlier argument breaks down. A perceptive
person can still claim that the ratio is increasing because in the nu-
merator y is clubbed with −x while in the denominator it is clubbed
with the bigger number x + 1. So even though both the numerator and
the denominator increase with y, the same increment in y will cause a
greater proportionate increase in the numerator than the denominator.
(A real life analogy would be that if the price of some commodity is
hiked, between two families that use the same amounts of that com-
modity, the poorer family will feel the impact more.)
What if you lack such fine perceptivity? There is still a simple
way out. We recast g(y) by writing the numerator as y + x + 1 −
y + x + 1 − 2x − 1 2x + 1
2x − 1. Then g(y) = = 1− . The first
x+y+1 x+y+1
term is a constant. In the second term, the numerator is a constant
while the denominator increases with y. So the ratio in the second
term decreases as y increases. But as we are taking its negative, g(y)
increases. (The incorrigible lovers of calculus can differentiate to get
2x + 1
g ′(y) = and observe that it is positive as x > 0. Like
(x + y + 1)2
many other convenient devices such as vehicles and calculators, calculus
makes you forget, and sometimes lose, some of your abilities which
would have been needed in absence of those devices.)
So, no matter which method we follow, g(y) is increasing on the interval
[x, 1]. Hence
1−x
M2 (x) = g(1) = (4)
x+2
52
The function f (x) is the larger of M1 (x) and M2 (x) as noted above.
So,
x 1−x
f (x) = max , (5)
x+1 x+2
for all x ∈ [0, 1].
We are still a long way off. We have to decide which of the two num-
bers M1 (x) and M2 (x) is greater. The answer will depend on x. When
it is not easy to directly compare two expressions, we subject them to
a series of simple operations which preserve inequalities. For example,
we can add a constant to both. Or we can multiply both by a positive
constant. We prepare two columns one headed by M1 (x) and the other
by M2 (x). In each column we add the two ‘descendants’ to which the
comparison is reduced. We go on doing this till we reach a stage where
one side can be declared greater than the other by inspection or by
some other means. (The process is reminiscent of a prolonged court
case which goes through hearings after hearings and finally is decided
in favour of one of the parties, usually after several generations of the
original combatants have passed.)
M1 (x) M2 (x)
x 1−x
x+1 x+2
x(x + 2) (1 − x)(x + 1)
2
x + 2x 1 − x2
2
2x + 2x − 1 0
The justifications in each transition are usually obvious and are not
specified, For example, in the very first step we multiplied both the
sides by (x + 1)(x + 2) which is positive for x ∈ [0, 1].
We have thus reduced the problem to deciding whether the quadratic
2x2 + 2x − 1 is positive√or negative. Its leading coefficient is positive
−1 ± 3
and its roots are . We discard the negative square root as x
2
takes√only non-negative values. So, by properties
√ of the quadratics, for
3−1 3−1
x< , M1 (x) < M2 (x), while for x > , M1 (x) > M2 (x).
2√ 2
3−1
For x = , M1 (x) and M2 (x) equal each other.
2
53
With this information, f (x) now becomes
√
1−x
x+2
, if 0√ ≤ x ≤ 3−1
2
f (x) = x 3−1
(6)
x+1
, if 2 < x ≤ 1
There are essentially two parts in the problem. The first is to convert
the given formula for f (x) to the manageable (5) and the second is to
decide the increasing/decreasing behaviour of f (x). The spadework
needed in the first part is considerable. But the reasoning needed in
it is the same as that in the second part of the problem. As a result,
the problem has become very repetitious and laborious. It would have
been a fair question in Paper 2 where you get 15 minutes per question.
In Paper 1, you get only four minutes. So, it would have been a far
better problem to give f (x) as in (5) and ask to identify the intervals
where it is increasing/decreasing.
Q.28 For a positive real number α, let Sα denote the set of points (x, y)
satisfying
|x|α + |y|α = 1.
A positive number α is said to be good if the points in Sα that are
closest to the origin lie only on the coordinate axes. Then
(A) all α ∈ (0, 1) are good and others are not good.
54
(B) all α ∈ (1, 2) are good and others are not good.
(C) all α > 2 are good and others are not good.
(D) all α > 1 are good and others are not good.
Answer and Comments: (C). Let us first note that the adjective
‘good’ here is purely mathematical. It has very little to do with any nice
or desirable qualities of the sets Sα . Unlike adjectives like ‘continuous’,
‘differentiable’, ‘increasing’ or ‘isosceles’ which are universally used, this
particular adjective ‘good’ is only of a local usage. It will not be found,
at least used in this sense, anywhere else in mathematics. The paper
setters could as well have called it ‘bad’ or ‘green’ or anything of their
choice.
The point to note is that the question is designed to test, in part, the
ability to grasp a new concept and answer questions about it. Usually,
in this process, the earlier knowledge we have helps. The ability to
relate a new concept with what you already know is an invaluable
asset and the present question deserves to be commended for testing
this.
Now, coming to the problem, we note that for each α > 0, whether
a point (x, y) is in Sα or not depends only on |x| and |y|. So, if a point
(x, y) is in Sα , so are its ‘mates’ (±x, ±y) in the four quadrants of
the plane. The same holds for the distance of a point from the origin.
Hence to test if some α is good or not, it suffices to assume that (x, y)
is in the first quadrant, where x, y are both non-negative and hence
|x|α + |y|α is the same as xα + y α .
Having got riddance of the absolute value, we can now identify the
portions of Sα for some values of α. The easiest case is α = 1. Here we
have
55
satisfy the given condition. Hence 1 is not good. The full S1 is shown
in the figure below.
y
(0,1) B C(1,1)
Sr
Sq
S1
Sp
A x
O (1,0)
S2
56
The statements made above require a justification which can be
given as follows. A typical point P on Sα can be taken as (r cos θ, r sin θ)
where r is its distance from O and θ ∈ [0, π2 ]. For θ = 0, P = A while
for θ = π2 , P = B. If we exclude these values, then both cos θ and sin θ
are positive real numbers less than 1. So their powers cosα θ and sinα θ
decrease as α increases. Now, for P to lie on Sα we must have
57
√
3 + 2 2 and ||v||10 = (1025)1/10 which is only slightly greater than 2.
It is, in fact, possible (but not so trivial) to show that as α increases
||v||α decreases for any vector v = (x, y) and tends to max{|x|, |y|} as
α → ∞. This number is therefore denoted by ||v||∞ and called the
∞-norm. Note that for a point P on either of the coordinate axes if
we let v be the position vector of P , then ||v||α is independent of α.
With this terminology, the set Sα in the problem is precisely the
set of those points (x, y) for which ||(x, y)||α = 1. Since ||v||α decreases
as α increases, for α > 2, the points which are at an α-distnace 1 from
O will be at a greater euclidean distance from it. That is why all α’s
bigger than 2 are ‘good’ in the terminology of the question.
One may wonder whether for α 6= 2, the concept of the α-norm and
hence also the derived concept of the α-distance between two points
has any application. Certainly, in the physical world, the euclidean
norm dominates all others. But there are occasions where some other
distance is more relevant. It will take us too far afield to illustrate
them. We only mentioned them to show that the present problem is
not as weird as it may appear at first sight.
Q.29 A water pitcher has a hemispherical bottom and a neck in the shape
of two truncated cones of the same size. The vertical cross section of
the pitcher with relevant dimensions is shown in the figure below.
40
30 . 10
10
40
Suppose that the pitcher is filled with water to the brim. If a solid
cylinder with diameter 24 cm and height greater 60 cm is inserted
vertically into the pitcher as far down to the bottom as possible, how
much water should remain in the pitcher?
(A) 6316π cm3 (B) 6116π cm3 (C) 6336π cm3 (D) 6136πcm3
58
Answer and Comments: (A). Problems on mensuration were very
common in JEE a long time ago. They were based on the formulas
for the volumes of some standard solids such as cones and cyclinders.
Proofs of these formulas were not expected as they require calculus.
Later, the JEE syllabus was revised to include calculus. After that, the
mensuration problems were dropped to make room for problems testing
the continuity or differentiability of some clumsily defined functions.
For a working engineer, the former are far more relevant than the latter.
Mensuration can, of course, be studied under integration. In fact,
finding the volumes of solids, their moments of inertia and areas of
curved surfaces etc. are among the standard applications of integrals.
But the integrals needed are double or triple integrals, which are too
advanced for the JEE and comparable examinations. So, nowadays, the
mensuration problems are relegated to physics where they are usually
expected to be tackled using standard formulas.
The present problem, therefore, deserves a welcome. But the time
given is inadequate. The paper setters have been kind enough to supply
a diagram. But even then the very text of the problem takes more than
a minute to read carefully and to figure out a way to solve it.
8 8
3 24 3 . 10
10
12
M B
h
20 h
24
A
It is given that the cylinder is inserted vertically into the pitcher. This
means that its axis is vertical and hence parallel to the axis of the
pitcher. But this by itself, does not mean that the two axes are the
same, i.e. that the axis of the cylinder passes through the centre, say
M, of the hemispherical portion of the pitcher. This has to be inferred
from the remaining part of the data that the cylinder is pushed as far
down as possible. Although it seems obvious by common sense (and
real life experience), that for this to happen, the axis must coincide
59
with that of the bowl, it needs a proof. We skip this proof. When
the essence of the problem is mensuration, such collateral problems of
geometric optimisation take a back seat. And, in any case, in MCQ’s
justifications hardly matter.
Now coming to the solution, the volume, say V , of the water left
in the pitcher is given by
V = V1 − V2 (1)
60
10 + 10 = 40. To find out exactly what it is, we let h be the height of
the portion of the cylinder lying inside the hemisphere. To find h, take
any point A on the lower rim of the cylinder and let B be the point
directly above it lying in the diametrical plane of the hemisphere as
shown in the figure. Then h = AB. The triangle MBA is right angled
at B. MB equals the radius of the cylinder, 12 while MA is a radius
of the hemisphere and hence equals 20. Hence by the Pythagorean
theorem,
q √
h = (20)2 − (12)2 = 4 52 − 32 = 4 × 4 = 16 (5)
(Note the time saved by taking out the common factor 4 from 20 and
12 before squaring them hastily.)
So the total height of the immersed portion of the cylinder is 16 + 20 =
36. Hence
(Here too, some saving of time is possible by the fact that 123 is a
familiar number 1728.)
To finish the proof we subtract (6) from (4) to get
61
Answer and Comments: (B). The function f in the problem is
defined only on the interval [−1, 1]. The data gives its value at a point
if that point is expressed as the sine of some angle. We have to find
the value of f ( 35 ). So we have to begin by expressing 53 as sin x2 for
some x ∈ [−π, π]. This is a simple trigonometric equation. Since x
varies in [−π, π], x2 varies in [−π/2, π/2]. In this interval α = sin−1 35 is
uniquely defined. Moreover it lies in (0, π/2) since 53 > 0. So cos α is
also positive and equals 45 .
So, we now have x = 2α with sin α = 53 and cos α = 45 . Now it is
only a matter of using the relation and expressing sin 2α and cos 2α in
terms of sin α and cos α whose values we know. So,
3 2α
f ( ) = f (sin )
5 2
= sin 2α + cos 2α
= 2 sin α cos α + cos2 α − sin2 α
3 4 16 9
= 2× × + −
5 5 25 25
24 + 7 31
= = (1)
25 25
So (B) is correct.
62
ISI BStat-BMath-UGB-2017 Paper with Comments
This formula will give the exact value of tan nθ if we put tan θ = 2. It is
also clear that the value will be a rational number because all binomial
coefficients are integers. Moreover, all the terms in the numerator and
all except the first term in the denominator will be even. So the tan nθ
is a rational number which can be written with an odd denominator.
But this formula is not so well known, as its special cases above. So
we give an argument which does not use it, which is possible because
our goal is much less ambitious than finding the exact value of tan nθ.
We proceed by induction on n. For the inductive step, we need to relate
tan(n + 1)θ to tan θ.
For this we use the more basic identity (which is instrumental to derive
both (1) and (2)), viz.
tan α + tan β
tan(α + β) = (4)
1 − tan α tan β
63
Putting α = nθ and β = θ and then substituting tan θ = 2 we get
tan nθ + tan θ
an+1 = tan(n + 1)θ = (5)
1 − tan nθ tan θ
tan nθ + 2
= (6)
1 − 2 tan nθ
2
For tan θ = 2, denote tan kθ by ak . Evidently a1 = tan θ = is a
1
rational number with odd denominator. So the statement is true for
n = 1. Assume that the result is true for n = k, i.e.
p
ak = tan kθ = (7)
q
where p, q are integers and q is odd. Putting this into (5) with n = k,
we get
p
q
+2
ak+1 =
1 − 2 pq
p + 2q
= (8)
q − 2p
which is a ratio of two integers and the denominator, viz. q − 2p is odd
because q is odd. Hence the statement is true for n = k + 1. So, it is
true for all n ∈ IN.
64
Q. 2 Consider a circle of radius 6. Let B, C, D and E be points on the circle
such that BD and CE, when extended, intersect at A. If AD and AE
have lengths 5 and 4 respectively,√and DBC is a right angle, then show
12 + 9 15
that the length of BC is .
5
Solution and Comments: The solution to any good geometry prob-
lem (and geometry problems are presumed to be good unless proved
otherwise!) ought to begin with a good diagram. Such a diagram is, in
fact, often needed to understand the problem. And then, if it is well
drawn, it might also suggest a solution. However, it is not necessary,
especially when time is severly limited, that the diagram be drawn to
scale. Certain vital features should not be compromised. For example,
an isosceles triangle should indeed look like an isosceles one. Similarly,
perpendicularity of two lines in the data should not be compromised.
But, if the angle between two lines is given to be, say, 30◦ , it is hardly
necessary to use a protractor in showing this. Similarly, some leeway
can be taken as far as the lengths of some of the segments are concerned.
What matters more is not so much the neatness and mathematical ac-
curacy of the diagram as its ability to convey to you the essence of the
problem, and possibly, a line of attack.
The following diagram fits this prescription.
A
4
5
E
3
D
z
y 12
B x C
65
The angle 6 DBC is clearly shown as a right angle. But the segment
AD which is supposed to have length 5 appears much longer in pro-
portionate comparison with the diameter CD which is given to be 12.
But that does not affect the vital calculations.
Now, coming to the √ solution, the problem asks us to show that
the length of BC is 12+95 15 . If instead of this horrible number, we
had some simple number, say 4, then one way to solve the problem
would reduce to showing that BC = AE and then we can look for
some intermediary (usually not given in the figure) which equals BC
on one hand and AE on the other. The search for such intermediary
would make the problem interesting. That is usually the beauty of
many geometry problems. Things which look unrelated on the face of
it, turn out to be closely related through such an intermediary. The
excitement is akin to that in discovering that the total stranger who
sits next to you on a flight turns out to be your cobrother’s brother!
The
√ present problem is not meant for such excitement. The figure
12+9 15
5
will have to be derived through some hard calculation, possi-
bly involving a quadratic equation. In such a problem, it is best to
introduce symbols for the lengths of the various line segments (other
than those whose lengths are given) and reduce the problem to solving
a system of equations. We begin by calling the sides BC, BD and CE
as x, y, z respectively. Our interest is only in x. But we need other
variables as auxiliary variables. Generally, one should minimise the
number of such auxiliary variables. For example, we could introduce
one more variable, say w, for the length of the segment DE. But that
can be obviated if we observe that since DBCE is a cyclic quadrilateral
and 6 DBC is a right angle, 6 DEC is also a right angle. But then so
is 6 AED and so by Pythagoras theorem we get DE = 3.
Thus we have three unknowns x, y, z and to determine them we need
a system of three (independent) equations. These are easily obtained
from the right angled triangles △DBC, △DEC and △ABC. Keeping
in mind that CD is a diameter and hance has length 12, we get,
x2 + y 2 = 144 (1)
z 2 + 9 = 144 (2)
and x2 + (y + 5)2 = (z + 4)2 (3)
66
From this point onwards, this is purely a problem of algebra. Using (1)
and (2), (3) can be simplified to give
10y + 25 = 8z + 7 (4)
i.e.
5y + 9 = 4z (5)
In the solution above, the results needed from geometry were only the
Pythagoas theorem (which was used four times) and a very elementary
67
property of cyclic quadrilaterals, viz. that their opposite angles add
to 180◦ . Those who know a little more about circles can get an easier
derivation of (5). The result needed is that for any point A outside a
circle, if a line through A cuts the circle at points X and Y , then the
product AX.AY is independent of the line. This constant, is called the
power of that point w.r.t. that circle. In the present problem, we can
apply it to the lines ADB and AEC to get
which implies (5). The proof of this result is based on the fact that an
external angle of a cyclic quadrilateral equals the opposite angle. As
a result, the triangles △AED and △ABC are similar to each other.
Equating the ratios of the corresponding sides we get (10).
Whichever way look at it, this is a simple geometric problem. But
the computations involved hijack it to the domain of algebra. So, this is
a geometric problem which is not ‘so’ good as a ‘purely’ pure geometry
problem!
Q. 3 Suppose f : IR −→ IR is a function given by
(
1 if x = 1
f (x) = 10 1
e(x −1) + (x − 1)2 sin x−1 if x =
6 1
68
For (b), we first recast the expression whose limit is to be taken to
get
100 100
" !#
X k X k
lim 100u − u f 1+ = lim u 1 − f (1 + )
u→∞
k=1 u u→∞
k=1 u
100
" #
X k
= lim u f (1) − f (1 + ) (2)
k=1
u→∞ u
100
X f (1) − f (1 + uk )
= k lim (3)
k=1
u→∞ k/u
where (2) is obtained by moving the limit inside a summation, which
is valid since only a finite sum is involved.
The problem is now reduced to finding, for each k from 1 to
100, the limit in (3) and adding these limits. From the nature of the
ratio appearing in (3), it is tempting to apply Lagrange’s Mean Value
Theorem and to write the ratio as −f ′ (c) for some c in the interval
(1, 1 + uk ). The trouble is that this intermediate point c will depend not
only on k but on u as well. It might be argued that such variation does
not matter because we are only interested in the limit of the expression
as u → ∞ and no matter what c is, as long as it lies in the interval
(1, 1 + uk ), it will tend to 1 from the right as u → ∞. So the limit in
(3) equals lim+ −f ′ (c) and hence it is −f ′ (1) i.e. −10, for every k.
c→1
The catch here is that this argument will require the continuity of
1
f ′ (x) at x = 1. But because of the second term, viz., (x − 1)2 sin( ),
x−1
in the expression for f (x), although f is differentiable at 1, the deriva-
tive is not continuous at 1. (It is helpful to recall here that if we write
y for x − 1, then the hierarchy sin y1 , y sin y1 , y 2 sin y1 , y 3 sin y1 , y 4 sin y1 , . . .
of functions (all of which are set to 0 at y = 0) is a well known example
of the progressively improved behaviour at 0. The first function is not
even continuous at 0, the second one is continuous but not differen-
tiable at 0, the third one (which is involved in the present problem) is
differentiable but not continuously differentiable at 0, the fourth one is
continuously differentiable at 0 but fails to have a second derivative at
0 and so on.)
It is tempting to try to salvage the situation using the all popular
l’Hôpital’s rule. For a fixed k, put y = uk . Then y → 0+ as u → ∞
69
f (1) − f (y)
and so the limit in (3) is the same as lim+ . By l’Hôpital’s
y→0 y
−f ′ (1 + y)
rule, this equals lim+ . But that does not help because as
y→0 1
noted above, even though f ′ (1) exists, lim+ f ′ (1 + y) does not exist.
y→0
So, l’Hôpital’s rule is not of much help either.
Fortunately, there is a much easier way out. Putting y = uk as above,
the limit in (3) equals −f ′ (1) from the very definition of a derivative.
No fancy theorems are needed for it. As we already calculated f ′ (1)as
100
X
10 in (a), the sum in (3)is simply −10 k which comes out to be
k=1
−500 × 101 = −50500.
The paper setters have done a wise thing by not asking this problem
in Paper 1. It does not take much time to arrive at the correct answer.
Nor is the correct justification very subtle. The problem is more a test
of the ability to realise how a tempting justification will not apply and
it is only in Paper 2 that this ability can be tested.
Q. 4 Let S be a square formed by the four vertices (1, 1), (1. − 1), (−1, 1)
and (−1, −1). Let the region R be the set of points inside S which are
closer to the center than any of the four sides. Find the area of the
region R.
Solution and Comments: This problem is a straight √ replica of a 1995
16 2 − 20
JEE problem. The answer comes out to be . It is obtained
3
by dividing the region into eight mutually congruent subregions each of
which is bounded by a pair of straight lines and a parabola. The reason
the parabolas enter the picture is that each parabola is the locus of a
point which is equidistant from a fixed point (the focus) and a fixed
line, the directrix. So, for a point on one side of the parabola the focus
is closer than the directrix while for a point on the other side of the
parabola, the directrix is closer than the focus. In the present problem,
we have to deal with four parabolas all having their foci at the origin
O but their directrices are the four sides of the square.
For the full solution see Comment No. 4 of Chapter 17. (There is
a minor difference of notation. The region there is denoted by S and
70
not R. We reproduce here the diagram in the solution.)
y
(−1,1) (1,1)
R1
x = −1
C4
C1
(−1,0) S
V1 S O x
1 C3
y=c
P C2
2
y=x y = 2x +1
(−1,−1) (1,−1)
71
Trivially, if n is a single digit number i.e. a number from 1 to 9, then
g(n) = n. Beyond 9, g(n) cannot grow as fast as n, because multiples
of 10 will have to be moved to the 10’s place where they will be only
single digits. But the digit in the unit’s place is at most 9. So if n is
a two digit number, say ab, then n = 10a + b. But g(n) = a × b ≤ 9a.
Hence g(n) < n.
A similar argument applies for any number with more than one
digit. Let the digits of n, read from the right (i.e. the unit’s place)
be x0 , x1 , x2 , . . . , xr where xr 6= 0. We may call r + 1 the ‘length’ of
n. (Actually, r is the integral part of log10 n. But that is not so vital
here.) We apply induction on r. For r = 0, n is a single digit number
x0 and both g(n) and n equal x0 . Suppose that the assertion is proved
for all numbers whose length is k + 1. Now suppose n is a number with
k + 2 digits x0 , x1 , . . . , xk and xk+1 , read from the right to left. Let m
be the number whose digits are x1 , x2 , . . . , xk+1 , again read backwards
from the unit’s place digit which is now x1 . (For example, if n = 2337
then m is simply 233.) Clearly,
n = 10m + x0 (1)
n2 − 13n + 36 ≤ 0 (5)
72
Factorising the L.H.S.,
(n − 4)(n − 9) ≤ 0 (6)
which is possible only when 4 ≤ n ≤ 9, i.e. for n = 4, 5, 6, 7, 8 and 9.
In all these cases, g(n) = n. But since n2 − 12n + 36 is also a perfect
square, for it to equal g(n) and hence n, n must also be a perfect
square. Hence the only possibilities are n = 4 and n = 9. By actual
substitution, (n − 6)2 equals 4 and 9 respectively for these values. So
(b) holds only for n = 4 and n = 9.
73
4 + p21 = u2 which gives p21 = (u + 2)(u − 2). By uniqueness of prime
factorisation, p1 will equal both u + 2 and u − 2, which is impossible.
Thus we assume that p1 , p2 , p3 are three distinct odd primes. Since
4 + p1 p2 = u2 for some u, we once again have p1 p2 = (u + 2)(u − 2),
Hence the smaller of p1 and p2 equals u − 2 and the other u + 2. So the
primes p1 and p2 differ by 4. By a similar reasoning, p1 and p3 differ by
4. Since p2 and p3 are distinct, p1 must be the smaller member of one
pair and the larger one of the other. Without loss of generality, we may
suppose that p2 < p1 < p3 . So the problem reduces to finding all triplets
(p2 , p1 , p3 ) of primes which form an A.P. with common difference 4. All
primes greater than 3 are of the form 6k + 1 or 6k − 1. The only way
p1 and p3 can differ by 4 is if p1 = 6k + 1 and p3 = 6k + 5 for some
integer k. But then p2 has to be 6k + 1 − 4 i.e. 6k − 3, which is a prime
only for k = 1.
Thus the only possible solutions are p1 = 7 and p2 = 3, p3 = 11
(or vice versa). That these primes actually give solutions is seen from
4 + 7 × 3 = 25 = 52 and 4 + 7 × 11 = 81 = 92 .
74
least one special case by hand, for a small value of n. If this value of
n is too small like 1 or 2, the work might not lead to any clue for the
general case. On the other hand, too large a value would entail a lot
of work just for trial. Let us take a reasonably small value, say n = 7.
So we have a permutation P of the symbols 1, 2, 3, 4, 5, 6 and 7. in
all there are 7! permutations. But we have to count only those which
satisfy the conditions (a) and (b). So, let us fix the first element of
P as, say 5. (Again, the choice of the extreme values 1 and 7 should
be avoided as they may not be sufficiently representative. Similarly
P (1) = 4 is ruled out on the ground that the symmetric location of 4
between the extremes 1 and 7 may lead to some simplifications which
will not hold in general.)
So, suppose P = (5, a, b, c, d, e, f ) is a permutation in which
a, b, c, d, e, f are integers from 1 to 7, other than 5. Condition (a)
means that all the integers less than 5 (i.e. the integers 1, 2, 3 and 4)
must appear in a descending order in the sequence (a, b, c, d, e, f ) while
(b) means that the integers 6 and 7 must appear in an ascending order
in this sequence. Once this is realised, it is clear that our freedom is
over the moment we fix the four places among a, b, c, d, e, f which the
integers from 1 to 4 should occupy, becausue once they are fixed, we
must send 1, 2, 3, 4 to then in an descending order and now only two
places are left which will have to be filled by 6 and 7 in an ascending
order.
The net result is that ! with n = 7 and P (1) = 5, the number of
6
desired permutations is . The same reasoning applies if P (1) has
4
any other value from 1 to 7. (In the extreme case where P (1) = 1, the
integers 2 to 7 must appear in the descending
! order, viz. as 7, 6, 5, 4, 3
6
and 2 and this is consistent with = 1. Therefore the total number
0
of permutations of the desired type for n = 7 is
7
!
X 6
(1)
k=1 k−1
6
!
X 6
With a change of index, this sum is simply . It is well known
r=0 r
75
that this sum is 26 . (One way to see this is to put x = 1 in the binomial
expansion of (1 + x)6 .)
We hardly need to give the details elaborately for the general case.
For each m = 1, 2, . . . , n, the number
! of permutations of the desired
n−1
type for which P (1) = m is . Hence the total number of such
m−1!
n
X n−1
permutations is the sum . With a change of index and the
m=1 m − 1
binomial theorem, this comes out to be 2n−1.
76
Q(x)
coefficients. Show that the function is strictly positive for
xk
all real x satisfying
1
0 < |x| < n
1+ |ai |
P
i=1
77
The numerator is a monic polynomial of degree n, say g(y), in y. So,
for all y 6= 0,
g(y)
f (x) = (5)
yn
We first assume x > 0 and hence y > 0. Then the denominator y n is
positive and so proving f (x) > 0 for 0 < x < δ is equivalent to proving
that
whenever,
n
X
y >1+ |ai | (7)
i=1
We are not given anything about the signs of the coefficients a0 , a1 , . . . , an−1 , an .
The positive coefficients will cause no problem. So, it suffices to con-
sider only those terms ai y n−i for which ai < 0. But we shall give a
proof which is independent of the signs.
Note first that (7) ensures that y > 1 and hence y r < y s whenever
r < s. Now since y > 1 + |a1 | + |a2 | + . . . + |an |, we have
yn = yy n−1
> (1 + |a1 | + |a2 | + . . . + |an )y n−1 )y n−1
= y n−1 + |a1 |y n−1 + |a2 |y n−1 + . . . + |an−1 |y n−1 + |an |y n−1 (8)
> |a1 |y n−1 + |a2 |y n−2 + |a3 |y n−3 + . . . + |an−1 |y + |an | (9)
where in going from (8) to (9), we have dropped the first term y n−1
which is positive and from the third term onwards, replaced higher
powers of y by lower ones.
(9) gives a lower bound on the first term of the R.H.S. of (6). Keeping
the other terms as they are, we get
g(y) > (a1 + |a1 |)y n−1 + (a2 + |a2 |)y n−2 + . . .
+(an−1 + |an−1 |)y + (an + |an |) (10)
78
n
X
holds for every term of (10), we get that g(y) > 0 for all y > 1 + |ai |.
i=1
Going back to x we have proved that
Q(x)
f (x) = >0 (11)
xk
for all real x satisfying 0 < x < δ. We still need to prove this for
1
−δ < x < 0. We can again put y = and argue as above. But now
x
y will be negative and so we shall have to be more careful in handling
inequalities involving powers of y. There is a way to avoid all this by
applying what we have already done to a different polynomial. In (2)
we replace x by −x. This gives a new polynomial in x, say h(x), defined
by
h(x) = 1 + b1 x + b2 x2 + . . . + bn xn (12)
where bi = (−1)i for i = 1, 2, . . . , n. Still, |bi | = |ai | for all i and so
1
δ= n (13)
1+ |bi |
P
i=1
If −δ < x < 0, then 0 < −x < δ and so by applying our work to h(x)
we have
h(−x) > 0 (14)
But h(−x) is the same as f (x). So we get that f (x) > 0 for −δ < x < 0
too. This completes the proof in all cases.
(b) This is of the same spirit as (a) but now we are more relaxed
because we merely have to see what happens in a sufficiently small
neighbourhood of 0. The size of this neighbourhood is immaterial as
long as it is positive. The coefficients b0 , b1 , . . . , bm−1 all vanish. But
bm 6= 0. So, we write
P (x) = bm xm (1 + a1 x + a2 x2 + ar−m xr−m ) (15)
bm+i
where ai = for i = 1, 2, . . . , r − m. The last factor is positive
bm
at x = 0 and hence, by continuity, for all x in some neighbourhood of
79
0. So, the sign of P (x) in this neighbourhood is the same as that of
bm xm . bm is a constant. If m is even, then xm > 0 for all x in this
neighbourhood (except at 0) and hence P (x) maintains its sign as x
passes through 0. Depending upon the sign of bm , P (x) will be either
positive on both the sides of 0 in this neighbourhood or negative on
both the sides. But if m is odd, then P (x) will have opposite signs on
the two sides.
Part (b) is absolutely trivial as compared with (a) and does not
need (a) for its solution. One fails to see the purpose behind asking
it. Perhaps, as in Q. 6, the idea is to award some consolation credit to
those who can do (b) but not (a).
But Part (a) deserves some comments. In essence, it shows that all
the real roots of the monic polynomial y n + a1 y n−1 + . . . + an−1 y + an
n
X
lie in the interval (−M, M) where M = 1 + |ai |. This upper bound
i=1
on the size of a real root of a monic polynomial is due to Lagrange.
As noted in the solution, we could have as well dropped the terms with
positive coefficients. In fact, one can concentrate on the ‘worst’ term,
i.e. one for which ai is negative and has the largest absolute value. This
way we get sharper upper bounds. Considerable work has been done in
cornering the roots of a polynomial in as small as intervals as possible,
because that is the next best thing to do when one cannot obtain the
roots exactly. (And for polynomials of degree 5 or above, there is no
formula to express the roots in terms of the radicals of expressions
involving the coefficients.) The quadratic formula is exceptional and
deceptive.
80
CONCLUDING REMARKS
81
So, it is now difficult to rank one test as consistently better than the other.
Let us hope that in the coming years, there will be a healthy competition
between JEE and ISI, each trying to come up with some original problems.
82