34 tayangan

Diunggah oleh gautamdb20

civil services maths optional papers solved from 1979-2006

- Solution of Linear System Theory and Design 3ed for Chi-Tsong Chen
- Essentials of Robust Control
- Classical Control Systems Book
- Ee263 Course Reader
- ET-101(A)
- Antonela Ism Ch08
- 87359342 Lecture Notes on Mathematical Systems Biology
- Canonical Forms
- ECS518 (Review of Basic Solid Mech) - 1
- ddagarc.pdf
- Chapter 3
- Quadratic Equation
- 08_AppendixA
- Un Parámetro Uniforme a Medida Punto Finitos Para Perturbado Singularmente Oda Linear Systems
- Statistics Unfolding Techniques
- 10.1.1.25.5953
- dec 2016.pdf
- Louis H. Kauffman- Take Home Exam - Math 310 - Linear Algebra - Spring 2008
- 02 Handout Linear Algebra
- ICDM07_Papadimitriou_Spiros.pdf

Anda di halaman 1dari 523

Analysis

The Union Public Service Commission (UPSC) is a Government of India organization that is responsible for the selection of

personnel for the Central Civil Services such as the Indian Administrative Service (IAS), the Indian Foreign Service (IFS), the

Indian Police Service (IPS) and many others. It conducts several examinations annually to select candidates, the most important of

which is the Civil Services Examination, usually taken by candidates after graduation with a bachelor's or master's degree.

Candidates choose 2 subjects out of a list, and are required to show competence in these at a graduate level.

Here we have provided solutions to questions in the Mathematics Exam over the years. Only solutions to the Linear Algebra,

Algebra, Calculus and Complex Analysis sections are provided. This page is under construction, please check the change log for

updates.

We hope that these solutions will be useful to candidates preparing for the Civil Services, and will also be useful for anyone

studying these subjects. Please drop us an email at sunderlal.chd (at) gmail.com if you found them useful. We would be glad to

hear comments and criticisms, suggestions for improvement, and any error reports. If you have a different or better solution to

any question, we would be glad to incorporate it, and will credit it to you in the notes.

Retired Professor of Mathematics Software Engineer

Panjab University Google India Inc.

Chandigarh, India. Bangalore, India.

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 License. You

are free to use it for any personal, classroom or other non-commercial purpose, but must seek prior

permission for commercial use.

http://sites.google.com/site/sunderlalchd/home

UPSC Civil Services Main 1979 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(b) Reduce the quadratic expression x2 + 2y 2 + 2z 2 + 2xy + 2xz to the canonical

form.

x + y + z, Y = y − z, Z = z to get the canonical form = X 2 + Y 2 . The expression is positive

semi-definite.

Alternate solution: See 1981 question 1(b) for an alternate method of canonicalization.

Question 2(a) Find the elements p, q, r such that the product BA of the matrices

1 2 1 1 0 0

A = 4 1 2 , B = p 1 0

−10 2 4 q r 1

is of the form

a1 b1 c1

BA = 0 b2 c2

0 0 c3

Hence solve the set of equations Ax = y, where x is the column vector (x1 , x2 , x3 ), and y is

the column vector (0, 8, −4).

1

Solution.

1 2 1 a1 b1 c1

BA = p + 4 2p + 1 p + 2 = 0 b2 c2

q + 4r − 10 2q + r + 2 q + 2r + 4 0 0 c3

Thus p + 4 = 0, q + 4r − 10 = 0, 2q + r + 2 = 0 ⇒ p = −4, r = 22

7

, q = − 18

7

.

Now solving Ax = y is the same as solving BAx = By because |B| 6= 0.

1 2 1 x1 1 0 0 0 0

0 −7 −2 x2 = −4 1 0 8 = 8

54

0 0 7

x3 − 18

7

22

7

1 −4 148

7

Thus 54 x = 148

7 3 7

⇒ x3 = 27 74

. −7x2 − 2x3 = 8 ⇒ −7x2 = 2x3 + 8 = 364

27

, so x2 = − 52

27

.

104 74

x1 + 2x2 + x3 = 0 ⇒ x1 = 27 − 27 = 10

9

.

10 52 74

Thus x1 = 9 , x2 = − 27 , x3 = 27 is the required solution.

Question 3(a) If S and T are subspaces of a finite dimensional vector space, then show

that

dim(S + T ) = dim S + dim T − dim(S ∩ T )

Question 3(b) Determine the value of a for which the following system of equations:

x1 + x2 + x3 = 2

x1 + 2x2 + x3 = −2

x1 + x2 + (a − 5)x3 = a

1 1 1

Solution. 1 2 3 = 2a − 10 − 3 − a + 5 + 3 − 1 = a − 6

1 1 a − 5

which is inconsistent with the first. So there is no solution.

2

Paper II

Question 4(a) Prove that any two finite dimensional vector spaces of the same dimension

are isomorphic.

Solution. See 1987 question 4(b).

Question 4(b) Define the dual space of a finite dimensional vector space V and show that

it has the same dimension as V.

Solution. Let V ∗ = {f : V −→ R, f a linear transformation}. Then V ∗ is a vector space

for the usual pointwise addition and scalar multiplication of functions: for all v ∈ V and all

α ∈ R, (f + g)(v) = f (v) + g(v), (αf )(v) = αf (v).

Let P a basis for V. Define n linear functionals v1∗ , . . . , vn∗ by vi∗ (vj ) = δij ,

v1 , . . . , vn be P

and vi∗ ( j=1 αj vj ) = nj=1 αj vi∗ (vj ) = αi .

n

Then v1∗ , . . . , vn∗ are linearly independent — ni=1 αi vi∗ = 0 ⇒ ( ni=1 αi vi∗ )(vj ) = αj =

P P

0, 1 ≤ j ≤ n. Pn Pn

∗ ∗ ∗ ∗ ∗ ∗

Pn v1 , . . . , v n generate V — if f ∈ V , then f = i=1 f (vi )v i . Clearly ( i=1 f (vi )vi )(vj ) =

∗

i=1 f (vi )vi (vj ) = f (vj ), so the two sides agree on v1 , . . . , vn , and hence by linearity on all

of V.

Thus v1∗ , . . . , vn∗ is a basis of V ∗ , so dim V ∗ = dim V . V ∗ is called the dual of V.

Question 4(c) Show that every finite dimensional inner product space V over the field of

complex numbers has an orthonormal basis.

Solution. Let w1 , . . . , wn be a basis of V. We will convert it into an orthonormal basis of

V by the Gram-Schmidt orthonormalization process.

Starting with i = 1, define

i−1

X hwi , vj i

v i = wi − vj

j=1

||vj ||2

but w1 , . . . , wn are linearly independent.

Now we can prove by induction on i that hvi , vj i = 0 for all j < i — this is enough

because hvi , vj i = hvj , vi i . Suppose it is true for all k < i. Then hvi , vj i = hwi , vj i −

Pi−1 hwi ,vm i hwi ,vj i

m=1 ||vm ||2 hvm , vj i = hwi , vj i − ||vj ||2 hvj , vj i = 0. Thus v1 , . . . , vn are mutually orthog-

onal. They are linearly independent, as ni=1 ai vi = 0 ⇒ h ni=1 ai vi , vj i = aj ||vj ||2 = 0 ⇒

P P

aj = 0 for all i ≤ j ≤ n. Replacing vi by ||vvii || gives us an orthonormal basis of V.

Question 5(a) Define the rank and nullity of a linear transformation. If V and W are

finite dimensional vector spaces over a field, and T is a linear transformation of V into W,

prove that

rank T + nullity T = dim V

Solution. See 1998 question 3(a).

3

Question 5(b) Define a positive definite form. State and prove a necessary and sufficient

condition for a quadratic form to be positive definite.

2z, 2x + y, −x − 2y + 2z) is a linear transformation. Find its nullity.

Solution.

= (ax1 + by1 − ax2 − by2 + 2ax3 + 2by3 , 2ax1 + 2by1 + ax2 + by2 ,

−ax1 − by1 − 2ax2 − 2by2 + 2ax3 + 2by3 )

= aT (x1 , x2 , x3 ) + bT (y1 , y2 , y3 )

If (x, y, z) ∈ the null space of T , then x − y + 2z = 0, 2x + y = 0, −x − 2y + 2z = 0 ⇒ y =

−2x, z = − 3x2

. Thus the null space is {(x, −2x, − 3x2

) | x ∈ R} = {(2, −4, −3)x | x ∈ R}.

Thus nullity T = 1.

4

UPSC Civil Services Main 1980 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Define the rank of a matrix. Prove that a system of equations Ax = b is

consistent if and only if rank (A, b) = rank A, where (A, b) is the augmented matrix of the

system.

2 1

Question 1(b) Verify the Cayley Hamilton Theorem for the matrix A = , and

1 2

hence find A−1 .

its characteristic

x − 2 −1

equation |xI − A| = 0. In the current problem, |xI − A| = = x2 − 4x + 4 − 1 =

−1 x − 2

2 2 2 2 1 2 1

x − 4x + 3. Thus we need to show that A − 4A + 3I = 0. Now A = =

1 2 1 2

5 4 2 5 4 2 1 1 0 0 0

, so A − 4A + 3I = −4 +3 = , verifying the Cayley

4 5 4 5 1 2 0 1 0 0

Hamilton Theorem.

2 1 1 0

A2 − 4A + 3I = 0 ⇒ A(A − 4I) = −3I ⇒ A−1 = − 3 (A − 4I) = − 3 (

1 1 −4 )=

2 1 2 0 1

− 13

3 .

− 13 23

Question 2(a) Prove that if P is any non-singular matrix of order n, then the matrices

P−1 AP and A have the same characteristic polynomial.

Solution. The characteristic polynomial of P−1 AP is |xI−P−1 AP| = |xP−1 P−P−1 AP| =

|P−1 ||xI − A||P| = |xI − A| which is the characteristic polynomial of A.

1

3 4

Question 2(b) Find the eigenvalues and eigenvectors of the matrix A = .

4 −3

3 4 3 − λ 4

Solution. The characteristic equation of A = is = 0 ⇒ −(9 −

4 −3 4 −3 − λ

λ2 ) − 16 = 0 ⇒ λ2 − 25 = 0 ⇒ λ = 5, −5.

−2 4 x1

If (x1 , x2 ) is an eigenvector for λ = 5, then = 0 ⇒ 2x1 − 4x2 = 0 ⇒

4 −8 x2

x1 = 2x2 . Thus (2x, x), x ∈ R, x 6= 0 gives all eigenvectors for λ = 5, in particular, we can

take (2, 1) as an eigenvector for λ = 5.

8 4 x1

If (x1 , x2 ) is an eigenvector for λ = −5, then = 0 ⇒ 4x1 + 2x2 = 0 ⇒ x2 =

4 2 x2

−2x1 . Thus (x, −2x), x ∈ R, x 6= 0 gives all eigenvectors for λ = −5, in particular, we can

take (1, −2) as an eigenvector for λ = −5.

Question 3(a) Find a basis for the vector space V = {p(x) | p(x) = a0 + a1 x + a2 x2 } and

its dimension.

Solution. Let f1 = 1, f2 = x, f3 = x2 , then f1 , f2 , f3 are linearly independent, because

α1 f1 + α2 f2 + α3 f3 = 0 ⇒ α1 + α2 x + α3 x2 = 0(zero polynomial) ⇒ α1 = α2 = α3 = 0.

f1 , f2 , f3 generate V because p(x) = a0 + a1 x + a2 x2 = a0 f1 + a1 f2 + a2 f3 for any p(x) ∈ V.

Thus {f1 , f2 , f3 } is a basis for V and its dimension is 3.

Question 3(b) Find the values of the parameter λ for which the system of equations

x + y + 4z = 1

x + 2y − 2z = 1

λx + y + z = 1

will have (i) unique solution (ii) no solution.

−1

1 1 4 1

Solution. The system will have the unique solution given by 1 2 −2

1 if

λ 1 1 1

1 1 4

1 2 −2 = 1(2 + 2) + 4(1 − 2λ) − 1(1 + 2λ) 6= 0. Thus 4 + 4 − 8λ − 1 − 2λ 6= 0 ⇒ λ 6= 7 .

10

λ 1 1

7

When λ = 10

, the system is

x + y + 4z = 1

x + 2y − 2z = 1

7x + 10y + 10z = 10

This system has no solution as it is inconsistent: 4(x+y+4z)+3(x+2y−2z) = 7x+10y+10z =

7, but the third equation says that 7x + 10y + 10z = 10. Thus there is a unique solution if

7 7

λ 6= 10 , and no solution if λ = 10 .

2

Paper II

show that each vector x ∈ V can be uniquely expressed as x = y + z, where y ∈ M and

z ∈ M⊥ , the orthogonal complement of M.

m

x ∈ V, let y = i=1 hx, vi i vi , and z =Px − y. Clearly y ∈ M, and x = y + z. Now

hz, vi i = hx, vi i − hy, vi i = hx, vi i − m j=1 hx, vj i hvj , vi i = hx, vi i − hx, vi i = 0. So

hz, vi i = 0, i = 1, . . . , m ⇒ hz, mi = 0 for every m ∈ M, so z ∈ M⊥ .

Now if x = y0 +z0 , then y−y0 = z0 −z. But y−y0 ∈ M, z0 −z ∈ M⊥ , so hy − y0 , z0 − zi =

0 ⇒ hy − y0 , y − y0 i = 0 ⇒ ||y − y0 || = 0 ⇒ y − y0 = 0 ⇒ z0 − z = 0. Thus y = y0 , z = z0

and the representation is unique.

Question 3(d) Find one characteristic value and corresponding characteristic vector for

the operators T on R3 defined as

Solution.

1. T (x, y, z) = (z, y, x) because the midpoint of (x, y, z) and (z, y, x) lies on the plane

x = z. T (1, 0, 0) = (0, 0, 1), T (0, 1, 0) = (0, 1, 0), T (0, 0, 1) = (1, 0, 0). Thus it is clear

that 1 is an eigenvalue, and (0, 1, 0) is a corresponding eigenvector.

2. T (1, 0, 0) = (1, 0, 0), T (0, 1, 0) = (0, 1, 0), T (0, 0, 1) = (0, 0, 0). Clearly 1 is an eigen-

value with (1, 0, 0) or (0, 1, 0) as eigenvectors.

3. T (1, 0, 0) = (3, 0, 0), T (0, 1, 0) = (1, 2, 0), T (0, 0, 1) = (1, 1, 1). Clearly (1, 0, 0) is an

eigenvector, corresponding to the eigenvalue 3.

3

UPSC Civil Services Main 1981 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question

1(a) State and prove the Cayley Hamilton theorem and verify it for the matrix

2 3

A= . Use the result to determine A−1 .

3 5

Cayley Hamilton

theorem.

x − 2 −3

The characteristic equation of A is = 0, or (x − 2)(x − 5) − 9 = 0 ⇒

−3 x − 5

x2 − 7x +1 = 0. The

Cayley

Hamilton

2

theorem implies that A − 7A + I = 0.

2 3 2 3 13 21

A2 = = .

3 5 3 5 21 34

2 13 21 2 3 1 0 0 0

Now A − 7A + I = −7 + =

21 34 3 5 0 1 0 0

2 −1

So thetheorem

is verified.

A − 7A + I = 0 ⇒ (A − 7I)A = −I ⇒ A = 7I − A. Thus

1 0 2 3 5 −3

A−1 = 7 − = .

0 1 3 5 −3 2

By using an orthogonal change of variables reduce Q to a form without the cross terms i.e.

with terms of the form aij xi xj , i 6= j.

5 4 2

Solution. The matrix of the qiven quadratic form Q is A = 4 5 2.

2 2 2

1

The characteristic polynomial of A is

5 − λ 4 2

4

5−λ 2 = 0

2 2 2 − λ

⇒ (5 − λ)(5 − λ)(2 − λ) − 4(5 − λ) − 4(8 − 4λ) + 16 + 16 − 4(5 − λ) = 0

⇒ (λ2 − 10λ + 25)(2 − λ) − 20 + 4λ − 32 + 16λ + 12 + 4λ = 0

⇒ −λ3 + 12λ2 + λ(−25 + 4 + 16 + 4 − 20) + 50 − 20 − 32 + 12 = 0

⇒ λ3 − 12λ2 + 21λ − 10 = 0

Thus the eigenvalues are λ = 1, 1, 10. Let (x1 , x2 , x3 ) be an eigenvector for λ = 10, then

−5 4 2 x1 −5x1 + 4x2 + 2x3 = 0 (i)

4 −5 2 x2 = 0 ⇒ 4x1 − 5x2 + 2x3 = 0 (ii)

2 2 −8 x3 2x1 + 2x2 − 8x3 = 0 (iii)

Subtracting (ii) from (i), we get −9x1 + 9x2 = 0 ⇒ x1 = x2 ⇒ x1 = 2x3 . Thus taking

x3 = 1, we get (2, 2, 1) as an eigenvector for λ = 10.

Let (x1 , x2 , x3 ) be an eigenvector for λ = 1, then

4 4 2 x1

4 4 2 x2 = 0 ⇒ 4x1 + 4x2 + 2x3 = 0

2x1 + 2x2 + x3 = 0

2 2 1 x3

x1 = x2 = 1 ⇒ x3 = −4 to get (1, 1, −4) as another eigenvector for λ = 1, orthogonal to the

first.

Thus

√2 √1 √1

√29 2

− √12

18

√1

O=

9 18

√1 0 − √418

9

10 0 0

is an orthogonal matrix such that O0 AO = O−1 AO = 0 1 0. If (X1 , X2 , X3 ) are new

0 0 1

x1 X1

variables, then x2 = O X2 takes Q(x1 , x2 , x3 ) to 10X12 + X22 + X32 .

x3 X3

Note: If the orthogonal transformation was not required for the diagonalization, we

2

could do it easily by completing squares:

8 4

= 5[x21 + x1 x2 + x1 x3 ] + 5x22 + 2x23 + 4x2 x3

5 5

4 2 2 16 4 16

= 5[x1 + x2 + x3 ] + (5 − )x22 + (2 − )x23 + (4 − )x2 x3

5 5 25 5 5

4 2 2 9 2 4 6 2

= 5[x1 + x2 + x3 ] + (x2 + x2 x3 ) + x3

5 5 5 9 5

4 2 2 9 2 2 10 2

= 5[x1 + x2 + x3 ] + [x2 + x3 ] + x3

5 5 5 9 9

9 10

= 5X 2 + Y 2 + Z 2

5 9

where X = x1 + 45 x2 + 52 x3 , Y = x2 + 29 x3 , Z = x3 , or x3 = Z, x2 = Y − 92 Z, x1 = X − 45 Y − 29 Z.

Question 2(a) Define a vector space. Show that the set V of all real-valued functions on

[0, 1] is a vector space over the set of real numbers with respect to the addition and scalar

multiplication of functions.

Question 2(b) If zero is a root of the characteristic equation of a matrix A, show that the

corresponding linear transformation cannot be one to one.

eigenvalue of A, so there is a non-zero eigenvector x such that Ax = 0, thus A is not 1-1.

orthogonal if and only if the matrix corresponding to it with respect to any orthonormal basis

is orthogonal.

Lemma 1. T is orthogonal iff T takes an orthonormal basis to an orthonormal basis.

Proof: Let {v1 , v2 , . . . , vn } be an orthonormal basis. Then

3. If ni=1 αi T(vi ) = 0, then h ni=1 αi T(vi ), vj i = αj = 0 for all j, so T(vi ) are linearly

P P

independent.

3

Thus T(v1 ), . . . , T(vn ) form an orthonormal basis. Pn

Pn Conversely, let T(v 1 ), . .

Pn. , T(v n ) be an orthonormal basisPn of V. Let v =

Pn i=1 αi vi , w =

Pi=1 βi vi , then hv, wi = i=1 αi βi and hT(v), T(w)i = h i=1 αi T(vi ), i=1 βi T(vi )i =

n

i=1 αi βi . Thus hT(v), T(w)i = hv, wi , so T is orthogonal.

Lemma 2. Let T∗ be defined by hT(v), wi = hv, T∗ (w)i . Then T∗ is a linear trans-

formation, and T is orthogonal iff T∗ T = TT∗ = I.

Proof: The fact that T∗ is a linear transformation can be easily checked. If T is

orthogonal, then hv, T∗ T(w)i = hT(v), T(w)i = hv, wi , so T∗ T = I. From this and the

fact that T is 1-1, it follows that TT∗ = I.

Lemma 3. If the matrix of T w.r.t. the orthonormal basis {v1 , v2 , . . . , vn } is A = (aij ),

then the matrix of T∗ P is the transpose, i.e. (aji ). P

n ∗ n ∗

Proof: T(vi )P= j=1 aij vj . Let T (vi ) = j=1 bij vj . Now bij = hT (vi ), vj i =

hvi , T(vj )i = hvi , nk=1 ajk vk i = aji . Since TT∗ = I, A0 A = AA0 = I, so A is orthogonal.

The converse is also obvious now.

Question 3(a) Investigate for what values of λ and µ does the following system of equations

x+y+z = 6

x + 2y + 3z = 10

x + 2y + λz = µ

have (1) a unique solution (2) no solution (3) an infinite number of solutions?

Solution.

1 1 1

1. A unique solution exists when 1 2 3 6= 0, whatever µ may be. Thus 2λ−6−(λ−3) 6=

1 2 λ

0 ⇒ λ 6=

3. Thus−1 λ 6=

forall 3 and for all µ we have a unique solution given by

x 1 1 1 6

y = 1 2 3 10

z 1 2 λ µ

2. A unique solution does not exist if λ = 3. If µ 6= 10, then the second and third

equations are inconsistent. Thus if λ = 3, µ 6= 10, the system has no solution.

3. If λ = 3, µ = 10, then the system is x+y+z = 6, x+2y+3z = 10. The coefficient matrix

is of rank 2, so the space of solutions is one dimensional. y + 2z = 4 ⇒ y = 4 − 2z,

and thus x = 2 + z. The space of solutions is (2 + z, 4 − 2z, z) for z ∈ R.

4

Question 3(b) Let (xi , yi ), i = 1, . . . , n be n points in the plane, no two of them having the

same abscissa. Find a polynomial f (x) of degree n − 1 which takes the value f (xi ) = yi , 1 ≤

i ≤ n.

that

a0 1 x1 . . . xn−1 1 a 0 y1

a1 1 x2 . . . xn−1 a1 y2

2

A .. = .. .. = ..

. . . .

n−1

an−1 1 xn . . . xn an−1 yn

a0 y1

a1 y2

This is possible as |A| 6= 0, as x1 , . . . , xn are distinct. .. = A−1 .. .

. .

an−1 yn

Note: We can also use Lagrange’s interpolation formula from numerical analysis, giving

n Q

j6=i (x − xj )

X

f (x) = yi Q

i−1 j6=i (xi − xj )

Paper II

3 0 √0

Question 4(a) Find a set of three orthonormal eigenvectors for the matrix A = 0 √4 3

0 3 6

3 − λ 0 0

√

|A − λI = 0

4√−λ 3 = 0

0 3 6 − λ

of A are 3, 3, 7.

Let (x1 , x2 , x3 ) be an eigenvector for λ = 7. Then

−4 0 √0 x1

0 −3 3 x2 = 0

√

0 3 −1 x3

√ √ √

Thus −4x1 = 0, −3x2 + 3x3 = 0, 3x2 − x3 = 0. Thus √ x1 = 0, x3 = 3x2 with x2 6=1 0 √gives

any eigenvector for λ = 7. Take x2 = 1 to get (0, 1, 3), and normalize it to get (0, 2 , 23 ).

5

Let (x1 , x2 , x3 ) be an eigenvector for λ = 3. Then

0 0 √0 x1

0 1 3 x2 = 0

√

0 3 3 x3

√ √

Thus x2 + 3x3 = 0. Thus (x1 , − 3x3 , x3 ) with x1 , x3 ∈ R gives any √ eigenvector for λ = 3.

We can take x1 = 1, x3 = 0, and x1 = 0, x3 = 1 to get (1, 0, 0), (0, − 3, 1) as eigenvectors for

λ = 3 — these are orthogonal √ and therefore span the the eigenspace of λ = 3. Orthonormal

3 1

vectors are (1, 0, 0), (0, − 2 , 2 ). √ √

Thus the required orthonormal vectors are (0, 12 , 23 ), (1, 0, 0), (0, − 23 , 21 ).

In fact

√

1 3 0 0√ 1

0 2√ 2

3 0 √0 7 0 0

0 − 3 1 0 4

√ 3 √12 − 23 0 = 0 3 0

2 2

0 3 6 3 1 0 0 3

1 0 0 2 2

0

Question 4(b) Show that if A = X0 AX and B = X0 BX are two quadratic forms one of

which is positive definite and A, B are symmetric matrices, then they can be expressed as

linear combinations of squares by an appropriate linear transformation.

Solution. Let B be positive definite. Then there exists an orthogonal real non-singular

matrix H such that H0 BH = In , the unit matrix of order n. A is real-symmetric ⇒ H0 AH is

real symmetric. There exists

K a real orthogonal

matrix such that K0 H0 AHK is a diagonal

λ1 0 . . . 0

0 λ2 . . . 0

matrix i.e. K0 H0 AHK = .. 0

.. where λ1 , . . . , λn are the eigenvalues of H AH.

..

. . .

0 0 . . . λn

x1 X1

0 0 0 .. ..

Now K H BHK = K In K = In . Then . = HK . diagonalizes A, B simulta-

xn Xn

neously.

x1 x1

. . . xn A ... = λ1 X12 + . . . + λn Xn2 . . . xn B ... = X12 + . . . + Xn2

x1 x1

xn xn

Note that λ1 , . . . , λn are the roots of Q

|H AH − λH BH| = |H AH − λIn | = ni=1 (λ − λi ).

0 0 0

6

UPSC Civil Services Main 1982 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Let V be a vector space. If dim V = n with n > 0, prove that

Solution. From 1983 question 1(a) we get that any two bases of V have n elements.

if v ∈ V is such that v is not a linear combination of v1 , . . . , vn , then v, v1 , . . . , vn are

linearly independent, so dim V > n which is not true. Thus v1 , . . . , vn is a basis of V

— here we have used the technique used to complete any linearly independent set to

a basis.

2. V cannot be generated by fewer than n vectors, because then it will have a basis

consisting of less than n elements, which contradicts the fact that dim V = n.

Question 1(b) Define a linear transformation. Prove that both the range and the kernel of

a linear transformation are vector spaces.

linear transformation if

1

Range of T = T(V), kernel of T = {v | T(v) = 0}. If w1 , w2 ∈ T(V), then w1 =

T(v1 ), w2 = T(v2 ) for some v1 , v2 ∈ V, αw1 + βw2 = αT(v1 ) + βT(v2 ) = T(αv1 + βv2 ).

But αv1 + βv2 ∈ V ∴ αw1 + βw2 ∈ T(V), thus T(V) is a subspace of W. Note that

T(V) 6= ∅ ∵ 0 ∈ T(V) so T(V) is a vector space.

If v1 , v2 ∈ kernel T then T(v1 ) = 0, T(v2 ) = 0. Now T(αv1 + βv2 ) = αT(v1 ) +

βT(v2 ) = 0 ⇒ αv1 + βv2 ∈ kernel T. Thus kernel T is a subspace. kernel T 6= ∅, bf 0 ∈

kernel T so kernel T is a vector space.

2 3 −1 0

1 −1 2 0

1 2 −1 0

1 4 −3 0

Operation R1 − R2 ⇒ A ∼ 1 −1 2 0

1 2 −1 0

1 4 −3 0

Operation R2 − R1 , R3 − R1 ⇒ A ∼ 0 −5 5

0

0 −2 2 0

1 4 −3 0

Operation − 15 R2 , − 12 R3 ⇒ A ∼ 0 1 −1 0

0 1 −1 0

1 4 −3 0

Operation R3 − R2 ⇒ A ∼ 0 1 −1 0

0 0 0 0

1 0 1 0

Operation R1 − 4R2 ⇒ A ∼ 0 1 −1 0

0 0 0 0

1 0 1 0

Thus rank A = 2 and the row echelon form is 0 1 −1 0

0 0 0 0

V of rank r, prove that T has nullity n − r.

2

Question 2(c) Show that the system of equations

3x + y − 5z = −1

x − 2y + z = −5

x + 5y − 7z = 2

is inconsistent.

Solution. From the first two equations, (3x + y − 5z) − 2(x − 2y + z) = −1 − 2(−5) = 9 ⇒

x + 5y − 7z = 9. But this is inconsistent with the third equation, hence the overall system

in inconsistent.

Question 3(a) Prove that the trace of a matrix is equal to the sum of its characteristic

roots.

Thus the sum of the roots of |λI − A| = −p1 = a11 + a22 + . . . + ann = tr A. Thus the trace

of A = sum of the eigenvalues of A.

Question 3(b) If A, B are two non-singular matrices of the same order, prove that AB

and BA have the same eigenvalues.

cos θ sin θ

Question 3(c) Find the eigenvalues and eigenvectors of the matrix A = .

sin θ − cos θ

λ2 − 1 = 0 ⇒ λ = ±1.

If (x1 , x2 ) is an eigenvector for λ = 1, then

cos θ − 1 sin θ x1

=0

sin θ − cos θ − 1 x2

Thus x1 (cos θ−1)+x2 sin θ = 0, x1 sin θ+x2 (− cos θ−1) = 0. We can take x1 = 1+cos θ, x2 =

sin θ.

Similarly if (x1 , x2 ) is an eigenvector for λ = −1, then

cos θ + 1 sin θ x1

=0

sin θ − cos θ + 1 x2

Thus x1 (cos θ+1)+x2 sin θ = 0, x1 sin θ+x2 (− cos θ+1) = 0. We can take x1 = 1−cos θ, x2 =

− sin θ as an eigenvector.

3

Paper II

is finite dimensional and dim W ≤ dim V.

be the space spanned by v1 then W1 is of dimension 1. If W1 = W, then dim W = 1 ≤ dim V.

If W1 6= W, then there exists a v2 ∈ W, v2 6∈ W1 . v1 , v2 are linearly independent —

if av1 + bv2 = 0, then if b 6= 0 then v2 = − ab v1 ⇒ v2 ∈ W1 , which is not true, hence

b = 0 ⇒ a = 0. Now let W2 be the space spanned by v1 , v2 then W2 is of dimension 2. If

W2 = W, then dim W = 2 ≤ dim V.

We continue the same reasoning as above, but this process must stop after at most r

steps where r ≤ n, otherwise we would have found n + 1 linearly independent vectors in V,

which is not possible. After r steps, we would have v1 , . . . , vr which are linearly independent

and span W. Thus dim W ≤ dim V, and W is finite dimensional.

Question 5(a) State and prove the Cayley-Hamilton Theorem when the eigenvalues are all

different.

Question 5(b) When are two real symmetric matrices said to be congruent? Is congruence

an equivalence relation? Also show how you can find the signature of A.

Solution. Two matrices A, B are said to be congruent to each other if there exists a

nonsingular matrix P such that P0 AP = B.

Congruence is an equivalence relation:

• Reflexive: A ≡ A ∵ A = I0 AI, I is the unit matrix.

• Transitive: A ≡ B, B ≡ C ⇒ A ≡ C — P0 AP = B, Q0 BQ = C ⇒ Q0 P0 APQ =

C ⇒ A ≡ C because PQ is nonsingular as both P, Q are nonsingular.

Given a symmetric matrix A, we first prove that there exists a nonsingular matrix P

such that P0 AP = diagonal[α1 , α2 , . . . , αr , 0, . . . , 0] where r is the rank of A.

We will prove this by induction on the order n of the matrix A. If n = 1, there is nothing

to prove. Assume that the result is true for all matrices of order < n.

Step 1. We first ensure that we have a11 6= 0. If it is 0, but some other akk 6= 0, we

interchange the k-th row with the first row and the k-th column with the first column, to

get B = P0 AP, where b11 = akk 6= 0. Note that P is the elementary matrix E1k (see 1983

question 2(a)), and is hence nonsingular and symmetric, so B is symmetric.

If all aii are 0, but some aij 6= 0. We add the j-th row to the i-th row and the j-the column

to the i-th column by multiplying A by Eij (1) and its transpose, to get B = Eij (1)AEij (1)0

4

— now bii = aij + aji 6= 0. B is still symmetric, and we can shift bii to the leading place as

above.

(Note that if all aij = 0, we stop.)

Thus we start with a11 6= 0. We subtract aa1k 11

times the first row from the k-th row and

a1k

a11

times the first column from the k-th column, by performing B = Ek1 (− aa1k11

)AEk1 (− aa1k

11

)0

a11 0

Repeating this for all k, 2 ≤ k ≤ n, we get P01 AP1 = , where A1 is n − 1 ×

0 A1

0

n − 1 and P1 is nonsingular. Now by induction, ∃P2 , n − 1 × n − 1 such that P2 AP2 =

1 0

diagonal[β2 , . . . , βr , 0, . . . , 0], rank A1 = rank A − 1. Now set P = P1 to get the

0 P2

result.

Now that we have P0 AP = diagonal[α1 , α2 , . . . , αr , 0, . . . , 0], let us assume that α1 , . . . , αs

are positive, the rest are negative. Then let αi = βi2 , 1 ≤ i ≤ s, −αj = βj2 , s + 1 ≤ j ≤ r. Set

Q = diagonal[β1−1 , . . . , βr−1 , 1, . . . , 1]. Then x0 Q0 P0 APQx = x21 + . . . + x2s − x2s+1 − x2r . Thus

we can find the signature of A by looking at the number of positive and negative squares of

the RHS.

Question

P 5(c)P Derive a set of necessary and sufficient conditions that the real quadratic

form 3j=1 3i=1 aij xi xj be positive definite.

Is 4x2 + 9y 2 + 2z 2 + 8yz + 6zx + 6xy positive definite?

3 3 9 9 9

= (2x + y + z)2 + 9y 2 + 2z 2 + 8yz + − yz − y 2 − z 2

2 2 2 4 4

3 3 2 27 2 1 2 7

= (2x + y + z) + y − z − yz

2 2 4 4 2

3 3 2 27 2 1 2 14

= (2x + y + z) + (y − z − yz)

2 2 4 27 27

3 3 2 27 7 2 1 2 49 2

= (2x + y + z) + (y − z) − z − z

2 2 4 27 4 108

So set X = 2x + 32 y + 23 z, Y = y − 277

z, Z = z, then Q(x, y, z) is transformed to X 2 + 27

4

Y2−

76

108

Z 2 . Hence Q(x, y, z) is is not positive definite.

5

UPSC Civil Services Main 1983 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Let V be a finitely generated vector space. Show that V has a finite basis

and any two bases of V have the same number of vectors.

Solution. Let {v1 , . . . , vm } be a generating set for V, we assume that vi 6= 0, 1 ≤ i ≤ m.

If {v1 , . . . , vm } is linearly independent, then it is a basis of V. Otherwise, there exists a

vk that depends linearly on {vi | 1 ≤ i ≤ m, i 6= k}. This latter set is also a generating

set, and we rename it {u1 , . . . , um−1 }. We now apply the same reasoning to it — either it

is linearly independent and hence a basis, or we can drop an element from it and it still

remains a generating set. In a finite number of steps, we reach {x1 , . . . , xr } ⊆ {v1 , . . . , vm }

such that {x1 , . . . , xr } is linearly independent and a generating set, thus {x1 , . . . , xr } is a

basis of V.

Note: An alternative approach leading to the same result is to pick the maximal linearly

independent subset of {v1 , . . . , vm }. There are only 2m such subsets, so we can do so in a

finite number of steps (in the above procedure we dropped the dependent elements one at

a time to reach the maximal linearly independent subset). Now to be a basis, the maximal

linearly independent subset S = {x1 , . . . , xr } ⊆ {v1 , . . . , vm } needs to generate V. But this

is immediate, as for each vi , either vi ∈ S or S ∪ {vi } is linearly Pdependent — in that case

P r r

j=1 aj xj + bvi = 0, but not all aj , b are 0. Now if b = 0 then j=1 aj xj = 0 ⇒ aj = 0 for

1 ≤ j ≤ r, as S is linearly independent, and this contradicts the statement that not all aj , b

are 0. So b 6= 0, hence vi is a linear combination of S, hence S generates V and is a basis.

Any two bases have the same number of elements: Let {v1 , . . . , vm } and {w1 , . . . , wn }

be two bases of V. Assume wlogP that m ≤ n. Now since w1 ∈ V, w1 is generated by the

m

basis {v1 , . . . , vm }, thus w1 = j=1 aj vj . There must be at least one non-zero ak , as

w1 6= 0. NowP the set {vi | 1 ≤ i ≤ m, i 6= k} ∪ {w1 } generates the set {v1 , . . . , vm } (since

aj

vk = a1k w1 − m j=1,j6=k aP

k

vj ) and hence generates V.

Now we have w2 = m i=1,i6=k ai vi + bw1 . At least one of the ai 6= 0, otherwise we have a

linear equation between w1 and w2 , but these are linearly independent. We replace vi by w2 ,

1

and the result is also a generating set as above. Continuing, after m steps, we

P get a subset

{w1 , . . . , wm } which is a generating set. Now if n > m, we would have wn = m i=0 ai wi , but

this is not possible as the wi were a basis, and thus linearly independent. Hence n = m, and

the two bases have equal number of elements.

Question 1(b) Let V be the vector space of polynomials of degree ≤ 3. Determine whether

the following vectors of V are linearly dependent or independent: u = t3 − 3t2 + 5t + 1, v =

t3 − t2 + 8t + 2, w = 2t3 − 4t2 + 9t + 5.

a + b + 2c = 0 (1)

−3a − b − 4c = 0 (2)

5a + 8b + 9c = 0 (3)

a + 2b + 5c = 0 (4)

From (4) - (1) we get b + 3c = 0. Substituting b = −3c in (2), c = −3a ⇒ b = 9a. Now from

(1), a + 9a − 6a = 0 ⇒ a = 0 ⇒ b = c = 0. Thus au + bv + cw = 0 ⇒ a = b = c = 0, so the

vectors are linearly independent.

therefore dim T (V1 ) ≤ dim V2 . Let v1 , . . . , vn be a basis of V1 , then T (V1 ) is generated by

T (v1 ), . . . , T (vn ) — If w ∈ T (V1 ), then there

Pn exists v ∈ V1 such that T (v) = w. But v =

P n

i=1 ai vi , ai ∈ R, therefore w = T (v) = i=1 ai T (vi ) ⇒ {T (v1 ), . . . , T (vn )} is a generating

system for T (V1 ) ⇒ dim T (V1 ) ≤ n. Thus rank T = dim T (V1 ) ≤ min(dim V1 , dim V2 ).

Question 2(a) Show that every non-singular matrix can be expressed as a product of ele-

mentary matrices.

1. Eij = the matrix obtained by interchanging the i-th and j-th rows (or the i-th and

j-th columns of the unit matrix. For example, if n = 4, then

1 0 0 0

0 0 1 0

E23 =

0 1 0 0

0 0 0 1

2

2. Ei (α) is the matrix obtained by multiplying the i-th row of the unit matrix by α =

the matrix obtained by multiplying the i-th column of the unit matrix by α.

3. Eij (β) = the matrix obtained by adding β times the j-th row to the i-th row of the

unit matrix.

4. (Eij (β))0 = transpose of Eij (β) = the matrix obtained by adding β times the j-th

column to the i-th column of the unit matrix.

All elementary matrices are non-singular. In fact |Eij | = −1, |Ei (α)| = α, |Eij (β)| =

|(Eij (β))0 | = 1.

We now prove the result.

(1) Let C = AB. Then any elementary row transformation

on AB is equivalent to sub-

R1

..

jecting A to the same row transformation. Let A = . and B = C1 . . . Cp n×p .

Rm m×n

R1 C1 . . . R1 Cp

.. ..

Then AB = . . . Thus if any elementary row transformation i.e. (i)

Rm C1 . . . Rm Cp m×p

Intercanging two rows (ii) Multiplying a row by a scalar (iii) Adding a scalar multiple of a

row to another row, is carried out on A, the same will be carried out on AB and vice versa.

Similarly any column transformation on B is equivalent to the same column transformation

on AB.

(2) Multiplying by an elementary matrix Eij , Ei (α), Eij (β) on the left is the same as per-

forming the corresponding elementary row operation on the matrix. Multiplying the matrix

by an elementary matrix to the right is equal to subjecting the matrix to the correspond-

ing column transformation. We write A = IA. Now interchanging the i-th and j-th row

of A is equivalent to doing the same on I in IA (result (1) above), which is the same as

Eij A. Similar results hold for the other two row transformations. Writing A as AI gives

the corresponding result for column transformations.

(3) We now prove that if A is a matrix

of rank r > 0, then there exist P, Q products of

Ir 0

elementary matrices such that PAQ = where Ir is the unit matrix of order r. Since

0 0

A 6= 0, A has at least one non-zero element, say aij . By interchanging the i-th row with

the first row and the j-th column with the first column, we get a new matrix Bij = (bij )

such that b11 6= 0. This simply means that there exist elementary matrices P1 , Q1 such

−1

that P1 AQ1 =

B. We multiply P1 AQ1 by P2 = E1 (b11 ) to obtain P2 P1 AQ1 = C =

1 ∗ ... ∗

∗

. Subtracting suitable multiples of the first row from the remaining rows of

..

. ∗

∗

C and suitable multiples of the first column from the remaining columns, we get the new

3

1 0 ... 0

0

matrix D of the form .. . Thus we have proved that there exist P∗ , Q∗ products

. ∗

A

0

1 0 ... 0

0

∗ ∗

of elementary matrices such that P AQ = .. . We carry on the same process

. ∗

A

0

∗ ∗∗ ∗∗ Ir 0

on A without affecting the first row and column, and in r steps we get P AQ = ,

0 E

where P∗∗ , Q∗∗ are products of elementary matrices. Note that E = 0 because rank A = r.

Now if A is nonsingular, then P∗∗ AQ∗∗ = I. Inverting the elementary matrices (the

inverse of an elementary matrix is elementary), we get that A is a product of elementary

matrices.

Question 2(b) Reduce the matrix A to its normal form, and hence or otherwise determine

its rank.

0 1 2 1

A = 1 2 3 2

3 1 1 3

1 2 3 2

Solution. Interchange of R1 and R2 ⇒ A ∼ 0 1 2 1

3 1 1 3

1 2 3 2

R3 − 3R1 ⇒ A ∼ 0 1 2 1

0 −5 −8 −3

1 2 3 2

R3 + 5R2 ⇒ A ∼ 0 1 2 1

0 0 2 2

1 2 3 2

− 21 R3 ⇒ A ∼ 0 1 2 1

0 0 −1 −1

1 2 3 2

R3 + R2 ⇒ A ∼ 0 1 2 1

0 1 1 0

1 2 3 2

Interchanging C2 , C4 , A ∼ 0 1 2 1

0 0 1 1

1 0 0 0

C2 − 2C1 , C3 − 3C1 , C4 − 2C1 ⇒ A ∼ 0 1 2 1

0 0 1 1

4

1 0 0 0 1 0 0 0

C3 − 2C2 , C4 − C2 ⇒ A ∼ 0 1 0 0. C4 − C3 ⇒ A ∼ 0 1 0 0

0 0 1 1 0 0 1 0

we have P(3× 3) and Q(4 × 4) both products of elementary matrices such that

Thus

1 0 0 0

PAQ = 0 1 0 0, which is the normal form of A. Clearly the rank of A is 3.

0 0 1 0

x+y+z = 3

3x − 5y + 2z = 8

5x − 3y + 4z = 14

are consistent and solve them.

1 1 1

Solution. The coefficient matrix A = 3 −5 2.

5 −3 4

det A = 1(−20

+ 6) − 1(12 − 10) + 1(−9 + 25) = 0, thus rank A < 3. Actually rank A = 2,

1 1

since 6= 0.

3 −5

1 1 1 3

The augmented matrix B = 3 −5 2 8 .

5 −3 4 14

1 1 1 3

R2 − 3R1 , R3 − 5R1 ⇒ B ∼ 0 −8 −1 −1

0 −8 −1 −1

1 1 1 3

R3 − R2 ⇒ B ∼ 0 −8 −1 −1

0 0 0 0

1 1

Thus rank B = 2, because 6= 0.

0 −8

Since rank A = rank B = 2, the system is consistent, and the space of solutions has

dimension 1.

Now x + y = 3 − z, 3x − 5y = 8 − 2z, subtracting the second from 3 times the first we

get 8y = 1 − z ⇒ y = 1−z 8

. x = 3 − z − 1−z 8

= 23−7z

8

. Thus the solutions are given by

( 23−7z

8

, 1−z

8

, z), z ∈ R.

Question 3(a) Prove that a square matrix satisfies its characteristic equation. Use this

result to find the inverse of

0 1 2

A = 1 2 3

3 1 1

5

Solution. The first part is the Cayley-Hamilton theorem, see 1987 question 3(a).

The characteristic equation of A is

−λ 1 2

|A − λI| = 1 2 − λ 3 = 0

3 1 1 − λ

⇒ −λ(λ2 − 3λ + 2 − 3) − (1 − λ − 9) + 2(1 − 6 + 3λ) = 0

⇒ λ3 − 3λ2 − 8λ + 2 = 0

0 1 2 0 1 2 7 4 5

A2 = 1 2 3 1 2 3 = 11 8 11

3 1 1 3 1 1 4 6 10

7 4 5 0 3 6 8 0 0 1 −1 1

1 1

∴ A−1 = − 11 8 11 + 3 6 9 + 0 8 0 = −8 6 −2

2 2

4 6 10 9 3 3 0 0 8 5 −3 1

Note: In this case, we were required to use this method to find the inverse. An alternate

method of finding the inverse by performing elementary row and column operations is shown

in 1985 question 1(c).

Question 3(b) Find the eigenvalues and eigenvectors of

8 −6 2

A = −6 7 −4

2 −4 3

Solution.

8 − x −6 2

|A − xI| = −6 7 − x −4 = 0

2 −4 3 − x

⇒ (8 − x)(x2 − 10x + 21 − 16) + 6(6x − 18 + 8) + 2(24 − 14 + 2x) = 0

⇒ −x3 + 18x2 − 85x + 40 + 36x − 60 + 20 + 4x = 0

⇒ x3 − 18x2 + 45x = 0

8 −6 2 x1

If (x1 , x2 , x3 ) is an eigenvector for the eigenvalue 0, then −6 7 −4

x2 = 0.

2 −4 3 x3

1

Thus 8x1 − 6x2 + 2x3 = 0, −6x1 + 7x2 − 4x3 = 0, 2x1 − 4x2 + 3x3 = 0 ⇒ x1 = 2 x3 , x2 = x3 .

Thus (1, 2, 2) is an eigenvector for 0, in general (x/2, x, x), x 6= 0 is an eigenvector for 0.

6

5 −6 2 x1

If (x1 , x2 , x3 ) is an eigenvector for the eigenvalue 3, then −6 4 −4

x2 = 0.

2 −4 0 x3

Thus 5x1 − 6x2 + 2x3 = 0, −6x1 + 4x2 − 4x3 = 0, 2x1 − 4x2 = 0 ⇒ x1 = 2x2 , x3 = −2x2 .

Thus (2, 1, −2) is an eigenvector for 3, in general (2x, x, −2x), x 6= 0 is an eigenvector

for 3.

−7 −6 2 x1

If (x1 , x2 , x3 ) is an eigenvector for the eigenvalue 15, then −6 −8 −4

x2 = 0.

2 −4 −12 x3

Thus −7x1 −6x2 +2x3 = 0, −6x1 −8x2 −4x3 = 0, 2x1 −4x2 −12x3 = 0 ⇒ x1 = 2x3 , x2 = −2x3 .

Thus (2, −2, 1) is an eigenvector for 15, in general (2x, −2x, x), x 6= 0 is an eigenvector for

15.

Question 3(c) Show that the eigenvalues of an upper or lower triangular matrix are just

the diagonal elements of the matrix.

Solution. Let A = (aij ), such that aij = 0 for i < j, i.e. A is upper triangular. Now

|xI − A| = (x − a11 )(x − a22 ) . . . (x − ann )

showing that |xI − A| = 0 ⇒ x = a11 , a22 , . . . , ann . Thus the eigenvalues of A are

a11 , a22 , . . . , ann .

Similarly for a lower triangular matrix.

Paper II

Question 4(a) Prove that a necessary and sufficient condition that a linear transformation

A on a unitary space is Hermitian is that hAx, xi is real for all x.

Solution. A unitary space is an old name for an inner product space. Let V be an inner

product space over C, and hAv, vi be real for all v ∈ V. Then since

hA(v + w), v + wi = hAv, vi + hAw, wi + hAv, wi + hAw, vi

hAv, wi +hAw, vi is real (because hA(v + w), v + wi −hAv, vi −hAw, wi is real). Hence

hAv, wi + hAw, vi = hw, Avi + hv, Awi (1)

because z real ⇒ z = z.

Also,

hA(v + iw), v + iwi = hAv, vi + hA(iw), iwi − ihAv, wi + ihAw, vi

thus −ihAv, wi + ihAw, vi is real. Thus

−ihAv, wi + ihAw, vi = −ihw, Avi + ihv, Awi (2)

Multiplying (1) by i and adding to (2), we get

2ihAw, vi = 2ihv, Awi

∗

Thus A = A , so A is Hermitian.

Conversely, if hAw, vi = hv, Awi , then hAv, vi = hv, Avi = hAv, vi ⇒ hAv, vi is

real.

7

Question 4(b) If A is a linear transformation on an n-dimensional vector space, then prove

that

1. rank A = rank A0 .

2. nullity A = n − rank A.

Solution. We know that rank A = r if A has a minor of order r different from 0, and all

minors of order > r are 0. Thus rank A = rank A0 .

For the second part, see 1998 question 3(a).

Question 4(c) Show that a real symmetric matrix A is positive definite if and only if there

exists a real non-singular matrix P such that A = PP0 .

Conversely: Let x1 , x2 , . . . , xn be a basis of Rn . We will use this to construct a new basis

e1 , e2 , . . . , en which satisfies ei Aej = δij , as follows:

x1

e1 = √

x1 Ax1

y2 = x2 − (x2 Ae1 )e1

y

e2 = √ 2

y2 Ay2

...

i−1

X

yi = xi − (xi Aej )ej

j=1

y

ei = √ i

yi Ayi

...

Pn

e1 , e2 , . . . , en are linearly independent — if i=1 ai ei = 0, then take the largest i such

that ai 6= 0, this allows us to express xi in terms of the other basis vectors, which is not

possible. Inductively we can also verify that ei Aei = 1, and ei Aej = 0 if i 6= j. This is

the Gram-Schmidt orthonormalization process - we exploit the property that any positive

definite matrix A gives rise to an inner product hx,yi = x0 Ay.

Now consider the matrix Q = e1 e2 . . . en . Now if B = Q0 AQ then bij = ei Aej ,

thus B = In . Since Q consists of linearly independent columns, it is invertible, and thus

A = Q0 −1 Q−1 . Setting P = Q0 −1 , we have A = PP0 .

then prove that A = (I − S)(I + S)−1 is an orthogonal matrix of order n.

8

Question 5(b) Under what circumstances will the real n × n matrix

x a a ... a

a x a . . . a

A= a a x . . . a

...

a a a ... x

x − λ a a . . . a

a

x−λ a ... a

a

a x − λ ... a = 0

...

a a a . . . x − λ

x − λ a − x + λ a − x + λ . . . a − x + λ

a

x−λ−a 0 ... 0

⇒ a 0 x − λ − a ... 0 = 0

...

a 0 0 . . . x − λ − a

x − λ + (n − 1)a 0 0 . . . 0

a x−λ−a 0 ... 0

⇒ a 0 x − λ − a ... 0 = 0

...

a 0 0 . . . x − λ − a

⇒ (x − λ + (n − 1)a)(x − λ − a)n−1 = 0

Thus the eigenvalues are x − a (repeated n − 1 times) and x + (n − 1)a. For positive definite,

λ > 0 ⇒ x > a, x > (n − 1)(−a). If a > 0, this reduces to x > a, if a ≤ 0, this reduces to

x > (n − 1)(−a).

For positive semi-definite, λ ≥ 0. By the same reasoning, if a > 0, then x ≥ a, otherwise

x ≥ (n − 1)(−a).

9

UPSC Civil Services Main 1984 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) If W1 , W2 are finite dimensional subspaces of a vector space V, then show

that W1 + W2 is finite dimensional and dim W1 + dim W2 = dim(W1 + W2 ) + dim(W1 ∩ W2 ).

Question 1(b) If A and B are n-rowed square non-zero matrices such that AB = 0, then

show that both A and B are singular. If both A and B are singular, and AB = 0, does it

imply that BA = 0? Justify your answer.

similarly B is singular.

1 0 0 0

AB = 0 does not imply that BA = 0. Let A = ,B = . Then AB =

1 0 0 1

0 0 0 0

, but BA = 6= 0.

0 0 1 0

Question 2(a) Show that row-equivalent matrices have the same rank.

is non-singular if and only if the vectors α1 T, α2 T, . . . , αn T are linearly independent in V.

When this is the case, show that T has an inverse T −1 with T T −1 = T −1 T = I.

1

Solution. If T (α1 P ), T (α2 ), . . . , T (αn ) Pare linearly independent, then T P is one-one: Let

n n n

v ∈ V. Then v = a α

i=1 i i , T (v) = a

i=1 i T (α i ). If T (v) = 0, then i=1 ai T (αi ) =

0 ⇒ ai = 0, 1 ≤ i ≤ n, because T (α1 ), T (α2 ), . . . , T (αn ) are linearly independent. Thus

T (v) = 0 ⇒ v = 0, so T is one-one.

T is onto: dim T (v) = dim V = n.

Thus T is invertible, in fact T −1 (T (αi )) = αi .

T −1 is a linear transformation: Let T −1 (v) = u, T−1 (w) = x. Then T (u) = v, T (x) = w.

Let T −1 (av + bw) = z, then T (z) = av + bw = aT (u) + bT (x) = T (au + bx) ⇒ z = au + bx.

Thus T −1 (av + bw) = aT −1 (v) + bT −1 (w), so T −1 is linear. It is obvious that T T −1 =

T −1 T = I, as this is true for the basis elements by definition, and extends to all vectors by

linearity. Pn

Conversely

Pn if T is non-singular, then a 1 T (α1 )+a 2 T (α 2 )+. . .+a n T (α n ) = 0 ⇒ T ( i=1 ai αi ) =

0 ⇒ i=1 ai αi = 0 ⇒ ai = 0, 1 ≤ i ≤ n because α1 , α2 , . . . , αn are linearly independent.

Thus T (α1 ), T (α2 ), . . . , T (αn ) are linearly independent.

3x1 + 2x2 + 2x3 − 5x4 = 8

2x1 + 5x2 + 5x3 − 18x4 = 9

4x1 − x2 − x3 + 8x4 = 7

3 2 2 −5

A = 2 5 5 −18

4 −1 −1 8

Doubling R1 , and subtracting R2 + R3 , we get

0 0 0 0

A ∼ 2 5 5 −18

4 −1 −1 8

Thus the rank of A is 2.

The augmented matrix

3 2 2 −5 8 0 0 0 0 0

B = 2 5 5 −18 9 ∼ 2 5 5 −18 9

4 −1 −1 8 7 4 −1 −1 8 7

Thus the rank of B is also 2, so the system is consistent.

Since the rank of A is 2, the space of solutions has rank = 4 − 2 = 2. Adding twice

the third equation to the first we get 11x1 + 11x4 = 22 ⇒ x1 = 2 − x4 . Substituting this

in the third equation, we get x2 = 1 − x3 + 4x4 . Thus the required solution system is

(2 − x4 , 1 − x3 + 4x4 , x3 , x4 ), where x3 , x4 take any value in R.

2

Question 3(a) Let V and W be vector spaces over the field F , and let T be a linear trans-

formation from V to W. If V is finite dimensional show that rank T + nullity T = dim V.

Question 3(b) Let A be a square matrix and T be non-singular. Let A

that

1. A and A

e have the same eigenvalues.

2. tr A = tr A.

e

of Ae corresponding to the same eigenvalue.

Solution.

1. The eigenvalues of A are roots of |xI − A| = 0. The eigenvalue of A

e are roots of

−1 −1 −1 −1

0 = |xI − T AT| = |T xIT − T AT| = |T ||xI − A||T| = |xI − A|, so the

eigenvalues are the same.

2. tr AB = tr BA, so tr T−1 AT = tr ATT−1 = tr A.

3. If Ax = λx then T−1 AT(T−1 x) = T−1 (λx) = λT−1 x.

Question 3(c) A 3 × 3 matrix has the eigenvalues 6, 2, -1. The corresponding eigenvectors

are (2, 3, −2), (9, 5, 4), (4, 4, −1). Find the matrix.

2 9 4

Solution. Let P = 3 5 4 , and let A be the required matrix. Then AP =

−2 4 −1

6 0 0 6 0 0

P 0 2 0 , therefore A = P 0 2 0 P−1 . A simple calculation gives P−1 =

0 0 −1 0 0 −1

−21 25 16 2 9 4 6 0 0 12 18 −4

−5 6 4 , note that |P| = 1. Now 3 5 4 0 2 0 = 18 10 −4.

22 −26 −17 −2 4 −1

0 0 −1 −12 8 1

12 18 −4 −21 25 16 −430 512 352

Thus A = 18 10 −4 −5 6 4 = 516 620 396

−12 8 1 22 −26 −17 234 226 −173

x1 x2 x3 6 0 0

A longer way would be to set A = y1 y2 y3 . Then AP = P 0 2 0 . This

z1 z2 z3 0 0 −1

yields three systems of linear equations which have to be solved.

3

Paper II

Question 4(a) Let V be the set of all functions from a non-empty set into a field K. For

any function f, g ∈ V and any scalar k ∈ K, let f + g and kf be functions in V defined by

(f + g)(x) = f (x) + g(x), (kf )(x) = kf (x) for every x ∈ X. Prove that V is a vector space

over K.

Solution. V = {f | f : X −→ K}.

2. The zero function namely 0(x) = 0∀x ∈ X is the additive identity of V i.e. f + 0 =

0 + f = f ∀f ∈ V.

4. (f + g) + h = f + (g + h) for every f, g, h ∈ V.

Question 4(b) Find the eigenvalues and basis for each eigenspace of the matrix

1 −3 3

A = 3 −5 3

6 −6 4

.

|A − λI| = 0

1−λ −3 3

⇒ 3 −5 − λ 3 = 0

6 −6 4−λ

⇒ (1 − λ)(λ + 5)(λ − 4) + 18(1 − λ) + 3(12 − 3λ − 18) + 3(−18 + 30 + 6λ) = 0

⇒ (1 − λ)(λ2 + λ − 20) + 18 − 18λ − 9λ − 18 + 36 + 18λ = 0

⇒ λ2 + λ − 20 − λ3 − λ2 + 20λ − 9λ + 36 = 0

⇒ λ3 − 12λ − 16 = 0

4

Thus λ = −2, 4, −2. Let (x1 , x2 , x3 ) be an eigenvector for λ = 4.

−3 −3 3 x1

3 −9 3 x2 = 0

6 −6 0 x3

Thus −3x1 − 3x2 + 3x3 = 0, 3x1 − 9x2 + 3x3 = 0, 6x1 − 6x2 = 0 ⇒ x1 = x2 , x3 = 2x1 . We

can take (1, 1, 2) as an eigenvector corresponding to λ = 4.

Let (x1 , x2 , x3 ) be an eigenvector for λ = −2.

3 −3 3 x1

3 −3 3 x2 = 0

6 −6 6 x3

Thus 3x1 − 3x2 + 3x3 = 0 ⇒ x2 = x1 + x3 . (1, 1, 0), (0, 1, 1) can be taken as eigenvectors for

λ = −2.

Clearly (1, 1, 2) is a basis for the eigenspace for λ = 4. (1, 1, 0), (0, 1, 1) is a basis for the

eigenspace for λ = −2.

Question 4(c) Let a vector space V have finite dimension and let W be a subspace of V

and W 0 the annihilator of W. Prove that dim W + dim W 0 = dim V.

Solution. Let dim V = n, dim W = m, W ⊆ V. Let {v1 , . . . , vn } be a basis of V so chosen

that {v1 , . . . , vm } is a basis of W. Let {v1∗ , . . . , vn∗ } be the dual basis of V ∗ i.e. vi∗ (vj ) = δij .

∗

We shall show that W 0 has {vm+1 , . . . , vn∗ } as a basis.

By definition of the dual basis vi∗ (vj ) = 0 when 1 ≤ i ≤ m and m + 1 ≤ j ≤ n. Since

∗

vj , m + 1 ≤ j ≤ n annihilate the basis of W, it follows that vj (w) = 0 for all w ∈ W. Thus

∗

{vm+1 , . . . , vn∗ } ⊆ W 0 , and are linearly independent, being a subset of a linearly independent

set.

Let f ∈ W 0 , then f = ni=1 ai vi∗ . We shall show that ai = 0 for 1 ≤ i ≤ m, thus f

P

∗ ∗ 0

is a linear P combination of {vP m+1 , . . . , vn }. By definition of W , f (v1 ) = 0, . . . , f (vm ) = 0,

n ∗ n ∗

therefore ( i=1 ai vi )(vj ) = i=1 ai δij = aj = 0 when 1 ≤ j ≤ m. Thus {vm+1 , . . . , vn∗ } is a

basis of W 0 , hence dim W 0 = n − m, hence dim W + dim W 0 = n = dim V.

Question 5(a) Prove that every matrix satisfies its characteristic equation.

Solution. See 1987 question 3(a).

Question 5(b) Find a necessary and sufficient condition that the real quadratic form

Xn X

n

aij xi xj be a positive definite form.

i=1 j=1

Question 5(c) Prove that the rank of the product of two matrices cannot exceed the rank

of either of them.

Solution. See 1987 question 1(b).

5

UPSC Civil Services Main 1985 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) If W1 and W2 are finite dimensional subspaces of a vector space V, then

show that W1 + W2 is finite dimensional and

1 0 0 1 0 0 0 0

Question 1(b) Let M1 = , M2 = , M3 = , M4 = . Prove

0 0 0 0 1 0 0 1

that the set {M1 , M2 , M3 , M4 } forms the basis of the vector space of 2 × 2 matrices.

1 3 3

Question 1(c) Find the inverse of the matrix A = 1 4 3.

1 3 4

Solution.

1 3 3 1 0 0

1 4 3 = 0 1 0 A

1 3 4 0 0 1

Using operation R2 − R1 , R3 − R1 , we get

1 3 3 1 0 0

0 1 0 = −1 1 0 A

0 0 1 −1 0 1

1

Now with operation R1 − 3(R2 + R3 ) we get

1 0 0 7 −3 −3

0 1 0 = −1 1 0 A

0 0 1 −1 0 1

7 −3 −3

Thus the inverse of A is −1 1 0 .

−1 0 1

space V to a vector space W, then prove that rank(T ) + nullity(T ) = n.

(1, 1, 0), v3 = (1, 0, 0). Express (2, −3, 5) in terms of v1 , v2 , v3 . Let T : R3 −→ R2 be

defined as T (v1 ) = (1, 0), T (v2 ) = (2, −1), T (v3 ) = (4, 3). Find T (2, −3, 5).

−3, a = 5 ⇒ a = 5, b = −8, c = 5. Thus (2, −3, 5) = 5v1 − 8v2 + 5v3 .

T (2, −3, 5) = 5T (v1 ) − 8T (v2 ) + 5T (v3 ) = 5(1, 0) − 8(2, −1) + 5(4, 3) = (9, 23).

6 3 −4

Question 2(c) Reduce the following matrix into echelon form: −4 1 −6.

1 2 −5

6 3 −4 1 2 −5

Solution. A = −4 1 −6 ∼ −4 1 −6 by exchanging R1 and R3 .

1 2 −5 6 3 −4

1 2 −5

Now R2 + 4R1 , R3 − 6R1 ⇒ A ∼ 0 9 −26.

0 −9 26

1 2 −5

R3 + R2 ⇒ A ∼ 0 9 −26.

0 0 0

1 2 −5

Multiply R2 by 91 to get A ∼ 0 1 − 269

.

0 0 0

1 0 97

Now R1 − 2R2 ⇒ A ∼ 0 1 − 26 9

which is the required form.

0 0 0

2

Question 3(a) Show that if λ is an eigenvalue of matrix A, then λn is an eigenvalue of

An , where n is a positive integer.

Solution. If x is an eigenvector for λ, then An x = An−1 Ax = λAn−1 x. Repeating this

process, we get the result.

Question 3(b) Determine if the vectors (1, −2, 1), (2, 1, −1), (7, −4, 1) are linearly indepen-

dent in R3 .

Solution. If possible, let a(1, −2, 1) + b(2, 1, −1) + c(7, −4, 1) = 0. Then a + 2b + 7c =

0, −2a + b − 4c = 0, a − b + c = 0. Adding the last two we get a = −3c, and from the third

we then get b = −2c. These values satisfy the first equation also, hence letting c = −1 we

get 3(1, −2, 1) + 2(2, 1, −1) − (7, −4, 1) = 0. Thus the vectors are linearly dependent.

Question 3(c) Solve

2x1 + 3x2 + x3 = 9 (1)

x1 + 2x2 + 3x3 = 6 (2)

3x1 + x2 + 2x3 = 8 (3)

Solution. 2(2) − (1) ⇒ x2 + 5x3 = 3 ⇒ x2 = 3 − 5x3 . Substituting x2 in (2), x1 = 7x3 .

5

Now substituting x1 , x2 in (3), we get 21x3 + 3 − 5x3 + 2x3 = 8 ⇒ x3 = 18 , x2 = 29

18

, x1 = 35

18

,

which is the required solution.

(Using Cramer’s rule would have been lengthy.)

Paper II

Question 4(a) Let V be the vector space of all functions from R into R. Let Ve be the

subset of all even functions f, f (−x) = f (x), and Vo be the subset of all odd functions

f, f (−x) = −f (x). Prove that

1. Ve and Vo are subspaces of V

2. Ve + Vo = V

3. Ve ∩ Vo = {0}

Solution.

1. Let f, g ∈ Ve , then αf + βg ∈ Ve for all α, β ∈ R, because (αf + βg)(−x) = αf (−x) +

βg(−x) = αf (x) + βg(x) = (αf + βg)(x), thus Ve is a subspace of V. Similarly, if

f, g ∈ Vo , then αf + βg ∈ Vo for all α, β ∈ R, because (αf + βg)(−x) = αf (−x) +

βg(−x) = −αf (x) − βg(x) = −(αf + βg)(x), thus Ve is a subspace of V.

2. Let f (x) ∈ V. Define F (x) = f (x)+f

2

(−x)

, G(x) = f (x)−f

2

(−x)

. Then F (−x) = F (x) ⇒

F ∈ Ve , G(x) = −G(x) ⇒ G ∈ Vo and f (x) = F (x) + G(x). Thus Ve + Vo = V.

3. If f ∈ Ve ∩ Vo , then f (−x) = f (x) ∵ f ∈ Ve , f (−x) = −f (x) ∵ f ∈ Vo . Thus

2f (−x) = 0 for all x ∈ R, so f = 0 ⇒ Ve ∩ Vo = {0}.

3

Question 4(b) Find the dimension and basis of the solution space S of the system

x1 + 2x2 + 3x3 + x4 + x5 = 0

3x1 + 6x2 + 8x3 + x4 + 5x5 = 0

Solution.

1 2 2 −1 3 1 2 2 −1 3

A = 1 2 3 1 1 ∼ 1 2 3 1 1

3 6 8 1 5 0 0 0 0 0

by performing R3 − R1 − 2R2 .

Thus rank A < 3. Actually rank A = 2, because if A = (C1 , C2 , C3 , C4 , C5 ), where Ci are

columns, then C1 and C3 are linearly independent.

Adding the first two equations, we get 4x5 = −2x1 − 4x2 − 5x3 . Subtracting 3 times

the second from the first, we get 4x4 = −2x1 − 4x2 − 7x3 . From these we can see that

X1 = (2, 0, 0, −1, −1), X3 = (0, 1, 0, −1, −1), X3 = (0, 0, 4, −5, −7) are three independent

solutions. Since rank A = 2, the dimension of the solution space S is 3, and {X1 , X2 , X3 }

is its basis.

Question 4(c) Let W1 and W2 be subspaces of a finite dimensional vector space V. Prove

that (W1 + W2 )0 = W10 ∩ W20 .

W 0 = {f | f ∈ V ∗ , ∀w ∈ W.f (w) = 0}. W 0 is a vector subspace of V ∗ of dimension

dim V − dim W.

If W1 ⊆ W2 , then W20 ⊆ W10 , because if f ∈ W20 , f (w) = 0∀w ∈ W2 , and therefore

f (w) = 0∀w ∈ W1 , so f ∈ W10 .

Now W1 ⊆ W1 + W2 and W2 ⊆ W1 + W2 , so (W1 + W2 )0 ⊆ W10 and (W1 + W2 )0 ⊆ W20 ,

thus (W1 + W2 )0 ⊆ W10 ∩ W20 .

Conversely, if f ∈ W10 ∩ W20 , then f (w1 ) = 0, f (w2 ) = 0 for all w1 ∈ W1 , w2 ∈ W2 . Now

any w ∈ W1 + W2 is of the form w = w1 + w2 , so f (w) = f (w1 ) + f (w2 ) = 0, because f is

linear. Thus f ∈ (W1 + W2 )0 .

Thus (W1 + W2 )0 = W10 ∩ W20 .

1 1+i 2i

Question 5(a) Let H = 1 − i 4 2 − 3i. Find P so that P0 HP is diagonal. Find

−2i 2 + 3i 7

the signature of H.

Solution.

1 1+i 2i 1 0 0 1 0 0

1 − i 4 2 − 3i = 0 1 0 H 0 1 0

−2i 2 + 3i 7 0 0 1 0 0 1

4

Subtracting (1 − i)R1 from R2 , and adding 2iR1 to R3 , we get

1 1 + i 2i 1 0 0 1 0 0

0 2 −5i = −1 + i 1 0 H 0 1 0

0 5i 3 2i 0 1 0 0 1

1 0 0 1 0 0 1 −1 − i −2i

0 2 −5i = −1 + i 1 0 H 0 1 0

0 5i 3 2i 0 1 0 0 1

1 −1 − i 52 − 92 i

1 0 0 1 0 0

0 2 0 = −1 + i 1 0 H 0 5

1 2

i

19 5 9 5

0 0 −2 2

+ 2i −2i 1 0 0 1

1 −1 + i 52 + 29 i

Thus P = 0 1 − 52 i

0 0 1

Index = Number of positive entries = 2. Signature = Number of positive entries - Number

of negative entries = 1.

Question 5(b) Prove that every matrix is a root of its characteristic polynomial.

Solution. This is the Cayley Hamilton theorem, proved in Question 5(a), 1987.

Question 5(c) If B = AP, where P is nonsingular and A orthogonal, show that PB−1 is

orthogonal.

Solution. B−1 = P−1 A−1 , so PB−1 = PP−1 A−1 = A−1 . Now (A−1 )0 A−1 = (AA0 )−1 = I.

Similarly A−1 (A−1 )0 = (A0 A)−1 = I, so PB−1 is orthogonal.

5

UPSC Civil Services Main 1986 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) If A, B, C are three n × n matrices, show that A(BC) = (AB)C. Show by

an example that matrix multiplication is non-commutative.

Pn BC = (βij ), AB = (αij ). Then the ij-th

element

P of P the RHS = k=1 αik ckj . But αik = l=1 ail blk , so the ij-th element of the RHS

= nk=1 nl=1 ail blk ckj . Pn Pn Pn

Similarly, the ij-th element of the LHS = l=1 ail βlj = l=1 ail k=1 blk ckj . Thus

A(BC) = (AB)C.

1 1 0 0 1 1 0 0 0 1

Let A = ,B = , then AB = = . But BA =

0 1 0 1 0 1 0 1 0 1

0 0 1 1 0 0

= . Thus AB 6= BA.

0 1 0 1 0 1

2. If A, B are square matrices each of order n, and I is the corresponding unit matrix,

then the equation

AB − BA = I

can never hold.

Solution.

1 0 0 0 0 0 0 0

1. True. Let A = ,B = ,C = . Then AB = = AC, but

0 0 0 1 0 0 0 0

B 6= C.

1

2. True. We have proved in 1987 question 5(c) that AB and BA have the same eigen-

values. Trace of AB − BA = trace of AB - trace of BA = sum of the eigenvalues of

AB - sum of the eigenvalues of BA = 0. But trace of In = n, thus AB − BA = I can

never hold.

1 2 −2 1 0 1

−1 3 0 X = 0 1 0

0 −2 1 0 1 1

1

2 −2

Solution. |A| = −13 0 = 3 − 2(−1) − 2(2) = 1, so A is non-singular. Hence

0

−2 1

1 0 1 3 2 6

−1

X=A 0 1 0. Asimple calculation gives A−1 = 1 1 2. Thus

0 1 1 2 2 5

3 2 6 1 0 1 3 8 9

X = 1 1 2 0 1 0 = 1 3 3

2 2 5 0 1 1 2 7 7

Question 2(a) If M, N are two subspaces of a vector space S, then show that their dimen-

sions satisfy

dim M + dim N = dim (M ∩ N ) + dim (M + N )

Question 2(b) Find a maximal linearly independent subsystem of the system of vectors

v1 = (2, −2, −4), v2 = (1, 9, 3), v3 = (−2, −4, 1), v4 = (3, 7, −1).

Solution. v1 , v2 are linearly independent because av1 +bv2 = (2a+b, −2a+9b, −4a+3b) =

0 ⇒ a = b = 0.

v3 is dependent on v1 , v2 . If v3 = av1 + bv2 , then (2a + b, −2a + 9b, −4a + 3b) =

7 6

(−2, −4, 1) ⇒ a = − 10 , b = − 10 .

Similarly v4 is dependent on v1 , v2 . If v4 = av1 + bv2 , then (2a + b, −2a + 9b, −4a + 3b) =

(3, 7, −1) ⇒ a = b = 1.

Thus the maximally linearly independent set is {v1 , v2 }.

2

Question 2(c) Show that the system of equations

4x + y − 2z + w = 3

x − 2y − z + 2w = 2

2x + 5y − w = −1

3x + 3y − z − 3w = 1

parameter.

4 1 −2 1

1 −2 −1 2

Solution. The coefficient matrix A = .

2 5 0 −1

3 3 −1 −3

4 1 −2 1 3

1 −2 −1 2 2

The augmented matrix B = .

2 5 0 −1 −1

3 3 −1 −3 1

Add R2 to R4 , and subtract R1 , to get

4 1 −2 1 4 1 −2 1 3

1 −2 −1 2

, B ∼ 1 −2 −1 2 2

A∼

2 5 0 −1 2 5 0 −1 −1

0 0 0 0 0 0 0 0 0

1 −2 1

Since −2 −1 2 = 1 − 16 + 5 6= 0, it follows that rank A = rank B = 3, so the system

5 0 −1

is consistent. Since rank A = 3, the space of solutions is of dimension 1.

Subtracting the second equation from the fourth, we get 2x + 5y − 5w = −1. But

2x + 5y − w = −1, so w − 5w = 0 ⇒ w = 0.

Now y − 2z = 3 − 4x, −2y − z = 2 − x ⇒ −5z = 8 − 9x ⇒ z = 9x−8 5

. Now y =

−2x−1 −2x−1 9x−8

3 − 4x + 2 9x−8

5

= 5

. Thus the space of solutions is (x, 5

, 5

, 0). The system does

not have a unique solution.

Question 3(a) Show that every square matrix satisfies its characteristic equation. Using

this result or otherwiseshow that if

1 0 2

A = 0 −1 1

0 1 0

3

Solution. The first part is the Cayley Hamilton theorem. See 1987 Question 5(a).

The characteristic equation of A is |A − xI| = 0, thus

1 − x 0 2

0

−1 − x 1 = (1 − x)(x2 + x − 1) = −x3 + 2x − 1 = 0

0 1 −x

By the Cayley Hamilton Theorem, A3 − 2A + I = 0.

Thus A4 = 2A2 − A, and 2A3 = 4A − 2I. Hence A4 − 2A3 − 2A2 + 6A − 2I =

2A2 − A − 4A + 2I − 2A2 + 6A − 2I = A as required.

Question 3(b) 1. Show that a square matrix is singular if and only if at least one of its

eigenvalues is 0.

2. The rank of an n×n matrix A remains unchanged if it is premultiplied or postmultiplied

by a nonsingular matrix, and that rank(XAX−1 ) = rank(A).

Solution.

1. The characteristic polynomial of A is |A−xI|. Putting x = 0, we see that the constant

term in the characteristic polynomial is |A|. Thus if A has 0 as an eigenvalue iff 0 is

a root of the characteristic polynomial iff |A| = 0.

R1

2. Let A = ... , where each Ri is 1×n, i.e. A is m×n. Now rank(A) is the dimension

Rm

of the row space of A, i.e. the space generated by R1 , . . . , Rm . Let P = (pij ) be

an

p11 R1 + p12 R2 + . . . + p1m Rm

p21 R1 + p22 R2 + . . . + p2m Rm

m × m nonsingular matrix. Then B = PA = .

..

.

pm1 R1 + pm2 R2 + . . . + pmm Rm

Thus the rows of PA ⊂ the row space of A, being linear combinations of rows of

A. Writing A = P−1 B, we get that the row space of A ⊂ the row space of B, so

rank(A) = rank(B).

Let Q be non-singular n × n, and C = AQ. It can be proved as above that the column

space of A = the column space of C, thus rank(A) = rank(C).

Now by using the above results, rank(XAX−1 ) = rank(XA) = rank(A).

Paper II

Question 4(a) If V1 and V2 are subspaces of a vector space V, then show that dim(V1 +V2 ) =

dim(V1 ) + dim(V2 ) − dim(V1 ∩ V2 ).

Solution. See 1998, question 1(b).

4

Question 4(b) Let V and W be vector spaces over the same field F and dim V = n. Let

{e1 , . . . , en } be a basis of V. Show that a map f : {e1 , . . . , en } −→ W, can be uniquely

extended to a linear transformation T : V −→ W whose restriction to the given basis is f

i.e. T (ei ) = f (ei ).

Pn Pn

Solution.

Pn If v =Pn i=1 a i e i , define T (v) = i=1 ai f (ei ). Clearly T (ei ) = f (ei ). If

v = i=1 ai ei , w = i=1 bi ei , then

n

X

T (αv + βw) = T ( (αai + βbi )ei )

i=1

n

X

= (αai + βbi )f (ei )

i=1

n

X n

X

= α ai f (ei ) + β bi f (ei )

i=1 i=1

= αT (v) + βT (w)

PnIf U is any other linear transformation

Pn satisfying

Pn U (ei ) P= f (ei ), then P

n

for any v =

n

a e

i=1 i i , by linearity, T (v) = T ( a e

i=1 i i ) = a

i=1 i T (e i ) = a

i=1 i f (e i ) = i=1 ai U (ei ) =

U (v). Since this is true for every v, we have T = U .

Question 5(a) 1. If A and B are two linear transformations and if A−1 and B −1 exist,

show that (AB)−1 exists and (AB)−1 = B −1 A−1 .

2. Prove that similar matrices have the same characteristic polynomial and hence the

same eigenvalues.

Solution.

B −1 B = I. Thus AB is invertible and its inverse is B −1 A−1 .

2. If B = P−1 AP then |λI−B| = |λP−1 P−P−1 AP| = |P−1 ||λI−A||P| = |λI−A|. Thus

A and B have the same characteristic polynomial and therefore the same eigenvalues.

5

1

Question 5(b) Reduce 2x2 + 4xy + 5y 2 + 4x + 13y − 4

= 0 to canonical form.

Solution.

1

LHS = 2(x + y + 1)2 − 2y 2 − 2 + 5y 2 + 9y −

4

9

= 2(x + y + 1)2 + 3(y 2 + 3y) −

4

3 27 9

= 2(x + y + 1)2 + 3(y + )2 − −

2 4 4

3

= 2X 2 + 3Y 2 − 9 where X = x + y + 1, Y = y +

2

X2 Y2

2X 2 + 3Y 2 − 9 = 0 ⇒ 9/2

+ 3

= 1. Thus the given equation is an ellipse.

0 1 1

Question 5(c) Find the reciprocal of the matrix T = 1 0 1. Then show that the

1 1 0

b+c c−a b−a

transform of the matrix A = 2 c − b c + a a − b by T i.e. TAT−1 is a diagonal matrix.

1

Determine the eigenvalues of the matrix A.

A11 A21 A31 −1 1 1

1 1

T−1 = A12 A22 A32 = 1 −1 1

2 2

A13 A23 A33 1 1 −1

0 1 1 b+c c−a b−a −1 1 1

1 1

TAT−1 = 1 0 1 c − b c + a a − b 1 −1 1

2 2

1 1 0 b−c a−c a+b 1 1 −1

0 2a 2a −1 1 1

1

= 2b 0 2b 1 −1 1

4

2c 2c 0 1 1 −1

4a 0 0 a 0 0

1

= 0 4b 0 = 0 b 0

4

0 0 4c 0 0 c

Thus TAT−1 is diagonal. Now the eigenvalues of A and TAT−1 are the same, so the

eigenvalues of A are a, b, c.

6

UPSC Civil Services Main 1987 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

7 −3

Question 1(a) 1. Find all the matrices which commute with the matrix .

5 −2

2. Prove that the product of two n × n symmetric matrices is a symmetric matrix if and

only if the matrices commute.

Solution.

1.

a b 7 −3 7 −3 a b

=

c d 5 −2 5 −2 c d

⇒ 7a + 5b = 7a − 3c (i)

−3a − 2b = 7b − 3d (ii)

7c + 5d = 5a − 2c (iii)

−3c − 2d = 5b − 2d (iv)

(i) and (iv) ⇒ 5b = −3c. From (ii) we get d = a + 3b, and from (iii) we get the

same

− 9c = 5a + 15b = 5d, or d = a + 3b. Thus the required matrices are

thing: 5a

a b

, a, b arbitrary.

− 53 b a + 3b

is symmetric. Let AB be symmetric. Then AB = (AB)0 = B0 A0 = BA, so A and B

commute. Thus AB is symmetric ⇔ AB = BA when A0 = A, B0 = B.

1

Question 1(b) Show that the rank of the product of two square matrices A, B each of order

n satisfies the inequality

rA + rB − n ≤ rAB ≤ min(rA , rB )

where rC stands for the rank of C, a square matrix.

G

Solution. There exists a non-singular matrix P such that PA = , where G is a

0

G

rA × n matrix of rank rA . Now PAB = B has at most rA non-zero rows obtained on

0

multiplying rA non-zero rows of G with B. Thus rPAB , which is the same as rank rAB as P

is non-singular, ≤ rA .

Similarly there exists a non-singular matrix Q such

that BQ = H 0 , where H is a

n × rB matrix of rank rB . Now ABQ = A H 0 has at most rB non-zero, columns, so

rABQ ≤ rB . Now rABQ = rAB as |Q| 6= 0, so rAB ≤ rB , hence rAB ≤ min(rA , rB ).

Let S(A) denote the space generated by the vectors r1 , . . . , rn where ri is the ith row

of A, then dim(S(A)) = rA , similarly dim(S(B)) = rB . Let S denote the space generated

by the rows of A and B. Clearly dim(S) ≤ dim(S(A)) + dim(S(B)) = rA + rB . Clearly

S(A + B) ⊆ S. Therefore rA+B ≤ dim(S) ≤ rA + rB .

IrA 0 −1 IrA 0

Now there exist non-singular matrices P, Q such that PAQ = or A = P Q−1 .

0 0 0 0

−1 0 0 −1 −1 IrA 0

Let C = P Q . Then A + C = P Q−1 = P−1 Q−1 , so A + C

0 In−rA 0 In−rA

is nonsingular.

Now rank B = rank((A + C)B) ≤ rank(AB) + rank(CB). But rank(CB) ≤ rank(C) =

n−rA . Thus rB ≤ rAB +n−rA ⇒ rA +rB −n ≤ rAB . Hence rA +rB −n ≤ rAB ≤ min(rA , rB ).

1 1 1 1

1 3 −2 a

A= 2 2a − 2 −a − 2 3a − 1

3 a+2 −3 2a + 1

1 0 0 0

1 2 −3 a − 1

Solution. |A| = by carrying out the operations C2 −C1 , C3 −

2 2a − 4 −a − 4 3a − 3

3 a − 1 −6 2a − 2

2

−3 1

0

0 1

C1 , C4 − C1 . Thus |A| = (a − 1) 2a − 4 −a − 4 3 = (a − 1) 2a − 10 −a + 5 3 =

a−1 −6 2 a−5 0 2

(a − 1)(a − 5)2 .

2

Thus |A| 6= 0 when a 6= 1, a 6= 5. So for 1 < a < 5, rank A = 4.

If a = 5,

1 1 1 1

1 3 −2 5

A = 2 8 −7 14

3 7 −3 11

1 0 0 0

1 2 −3 4

= (C2 − C1 , C3 − C1 , C4 − C1 )

2 6 −9 12

3 4 −6 8

1 0 0 0

0 2 −3 4

= (R2 − R1 , R3 − 2R1 , R4 − 3R1 )

0 6 −9 12

0 4 −6 8

1 0 0 0

0 2 −3 4

= 0 0 0 0

(R3 − 3R2 , R4 − 2R2 )

0 0 0 0

1 0

which has rank 2, as 6= 0, showing that rank of A is 2 when a = 5.

0 2

If a = 1,

1 1 1 1

1 3 −2 1

A = 2 0 −3 2

3 3 −3 3

1 0 0 0

1 2 −3 0

= 2 −2 −5 0

(C2 − C1 , C3 − C1 , C4 − C1 )

3 0 −6 0

1 0 0

which has rank 3 since 1 2 −3 =

6 0, showing that rank of A is 3 when a = 1.

2 −2 −5

polynomial in x, show that the eigenvalues of the polynomial f (A) are f (λj ), j = 1, 2, . . . n.

3

Solution. Let xr be an eigenvector of λr . Then Ak xr = Ak−1 (Axr ) = λr Ak−1 xr = . . . =

λkr xr . Thus the eigenvalues of Ak are λkj , j = 1, 2, . . . , n.

Let f (x) = a0 + a1 x + . . . + am xm . Then (a0 I + a1 A + . . . + am Am )xr = (a0 + a1 λr +

. . . + am λmr )xr = f (λr )xr . Thus the eigenvalues of f (A) are f (λj ), j = 1, 2, . . . n.

Question 2(b) If A is skew-symmetric, then show that (I − A)(I + A)−1 , where I is the

corresponding identity matrix, is orthogonal.

0 ab

Hence construct an orthogonal matrix if A = .

− ab 0

−1

Solution. For

theaorthogonality

− A)(I

of (I

a

+ A) , see question 2(a)

of 1999.

1 −b 1 b −1 b b −a

I−A= a , and I + A = ⇒ (I + A) = a2 +b2 .

b

1 − ab 1 a b !

b2 −a2 −2ab

2

−1 1 b −a b −a 1 b − a2 −2ab a2 +b2 a2 +b2

Thus (I−A)(I+A) = a2 +b2 = a2 +b2 = b2 −a2 ,

a b a b 2ab b 2 − a2 2ab

a2 +b2 a2 +b2

which is the required orthogonal matrix.

show that AB and BA have the same characteristic polynomial.

2. Show that a real matrix A is orthogonal if and only if |Ax| = |x| for all x.

Solution.

1. BA = A−1 ABA. Thus the characteristic polynomial of BA is |xI−BA| = |xA−1 A−

A−1 ABA| = |A−1 ||xI − AB||A| = |xI − AB| which is the characteristic polynomial

of AB.

√ √

2. If A is orthogonal, i.e. A0 A = I, then |Ax| = x0 A0 Ax = x0 x = |x|.

Conversely |Ax| = |x| ⇒ x0 A0 Ax = x0 x ⇒ x0 (A0 A − I)x = 0 for all x. Thus

A0 A − I = 0, so A is orthogonal.

Note that is A = (aij ) is symmetric, and ni,j=1 aij xi xj = 0 for all x, then choose x = ei to

P

get e0i Aei = aii = 0, and choose x = ei +ej to get 0 = x0 Ax = aii +2aij +ajj = 2aij ⇒ aij = 0.

(Here ei is the i-th unit vector.) Thus A = 0.

Question 3(a) Show that a necessary and sufficient condition for a system of linear equa-

tions to be consistent is that the rank of the coefficient matrix is equal to the rank of the

augmented matrix. Hence show that the system

x + 2y + 5z + 9 = 0

x − y + 3z − 2 = 0

3x − 6y − z − 25 = 0

is consistent and has a unique solution. Determine this solution.

4

Solution. Let the system be Ax = b where A is m × n, x is n × 1 and b is m × 1. Let

rank A = r. A = [c1 , c2 , . . . , cn ] where each cj is an m × 1 column. We can assume without

loss of generality that c1 , c2 , . . . , cr are linearly independent, r = rank A. The system is now

x1 c1 + x2 c2 + . . . + xn cn = b

, where x0 = (x1 , . . . , xn ). Suppose rank([A b]) = r. This means that out of n + 1 columns,

exactly r are independent. But by assumption, c1 , c2 , . . . , cr are linearly independent, there-

fore these vectors form a basis for the column space of [A b]. Consequently there exist

α1 , . . . , αr such that α1 c1 + α2 c2 + . . . + αr cr = b. This gives us the required solution

{α1 , . . . , αr , 0, . . . , 0} to the linear system.

Conversely, let the system be consistent. Let A = [c1 , c2 , . . . , cn ] as before, with c1 , c2 , . . . , cr

linearly independent, r = rank A. Since the column space of A, i.e. the space generated

by c1 , c2 , . . . , cn has dimension r, each cj for r + 1 ≤ j ≤ n is linearly dependent on

c1 , c2 , . . . , cr . Since there exist α1 , . . . , αn such that α1 c1 + α2 c2 + . . . + αn cn = b, b is a

linear combination of c1 , c2 , . . . , cn . But each cj for r + 1 ≤ j ≤ n is a linear combination of

c1 , c2 , . . . , cr , therefore b is a linear combination of c1 , c2 , . . . , cr . Thus the space generated

by {c1 , c2 , . . . , cn , b} also has dimension

r, sorank([A b]) = r = rank A.

1 2 5

The coefficient matrix A = 1 −1 3 . |A| = 24 6= 0, so rank A = 3. The aug-

3 −6 −1

1 2 5 −9 1 2

5

mented matrix B = 1 −1 3 2 has rank ≤ 3, but since 1 −1 3 6= 0, it has

3 −6 −1 25 3 −6 −1

rank 3. Thus the given system is consistent.

Subtracting the second equation from the first we get 3y + 2z + 11 = 0. Subtracting 3

times the second equation from the third, we get 3y + 10z + 19 = 0. Clearly z = −1, y =

−3 ⇒ x = 2. Thus (2, −3, −1) is the unique solution. In fact the only solution of the system

is

x −9 2

y = A−1 2 = −3

z 25 −1

linearly independent and can be expressed linearly in terms of the vectors yk , k = 1, . . . , s.

Show that r ≤ s.

Find a maximal linearly independent subsystem of the linear forms

f1 = x + 2y + z + 3t

f2 = 4x − y − 5z − 6t

f3 = x − 3y − 4z − 7t

f4 = 2x + y − z

5

Solution. Let W be the subspace spanned by yk , k = 1, . . . , s. Then dim W ≤ s. Since

xj ∈ W, j = 1, . . . , r because xj is a linear combination of yk , k = 1, . . . , s, and xj , j =

1, . . . , r are linearly independent, dim W ≥ r ⇒ r ≤ s.

Clearly f1 and f4 are linearly independent. f2 is linearly expressible in terms of f1 and f4

because f2 = af1 + bf4 ⇒ a + 2b = 4, 2a + b = −1, a − b = 5, 3a = −6 ⇒ a = −2, b = 3 satisfy

all four, hence f2 = −2f1 + 3f4 . Similarly f3 = − 37 f1 + 53 f4 . Thus {f1 , f4 } is a maximally

independent subsystem.

Paper II

show that

rank T + nullity T = dim V

Question 4(b) Prove that two finite dimensional vector spaces V, W over the same field F

are isomorphic if they are of the same dimension n.

W. Define T : V −→ W by T (vi ) = wi and if v ∈ V, v = ni=1 ai vi , ai ∈ R then

basis of P

T (v) = ni=1 ai T (vi ). Then

P P

n

X

T (αv + βu) = T ( (αai + βbi )T (vi )

i=1

n

X

= (αai + βbi )T (vi )

i=1

n

X n

X

= α ai T (vi ) + β bi T (vi )

i=1 i=1

= αT (v) + βT (u)

P P

2. P

n

i=1 ai wi ⇒ ai = 0, i = 1, . . . , n, because w1 , . . . , wn are linearly independent. Thus

T (v) = 0 ⇒ v = 0.

P P

6

Note: The converse of 4(b) is also true i.e. if T : V −→ W is an isomorphism i.e. V, W are

isomorphic, then dim V = dim W.

Let v1 , . . . , vn be a basis of V. Then {w1 = T P is a basis of W.

(v1 ), . . . , wn = T (vn )}P

n n

w1 , . . . , wn are linearly independent. If i=0 bi wi = 0, then i=0 bi T vi = 0 ⇒

T ( ni=0 bi vi ) = 0 ⇒ ni=0 bi vi = 0 ⇒ bi = 0 for 1 ≤ i ≤ n, because v1 , . . . , vn are linearly

P P

independent.

w1 , . . . , wn generate W.P If w ∈ W, then there exists a P v ∈ V such that

Pn T (v) = w,

n n

because

Pn T is onto. Let v = i=0 bi vi , then w = T (v) = T ( i=0 bi vi ) = i=0 bi T (vi ) =

i=0 bi wi .

Question 5(a) Prove that every square matrix is the root of its characteristic polynomial.

Solution. This is the Cayley Hamilton Theorem. Let A be a matrix of order n. Let

|A − xI| = ao + a1 x + . . . + an xn

ao I + a1 A + . . . + an A n = 0

n. Then by definition of the adjoint,

of like powers, we get

AB0 = ao I

AB1 − B0 = a1 I

AB2 − B1 = a2 I

...

ABn−1 − Bn−2 = an−1 I

−Bn−1 = an I

0 = ao I + a1 A + . . . + an An , which was to be proved.

7

Question 5(b) Determine the eigenvalues and the corresponding eigenvectors of

2 2 1

A = 1 3 1

1 2 2

Solution.

λ − 2 −2 −1

|λI − A| = −1 λ − 3 −1 = 0

−1 −2 λ − 2

⇒ (λ2 − 4λ + 4)(λ − 3) − 5λ + 7 = 0

⇒ λ3 − 7λ2 + 11λ − 5 = (λ − 1)(λ2 − 6λ + 5) = 0

⇒ λ = 1, 5, 1

3 −2 −1 x1

−1 2 −1 x2 = 0

−1 −2 3 x3

(1, 1, 1) is an eigenvector for λ = 5. In fact (x, x, x) with x 6= 0 are eigenvectors for λ = 5.

Let (x1 , x2 , x3 ) be an eigenvector for λ = 1. Then

−1 −2 −1 x1

−1 −2 −1 x2 = 0

−1 −2 −1 x3

Thus x1 + 2x2 + x3 = 0. We can take x1 = (1, 0, −1) and x2 = (0, 1, −2) as eigenvectors for

λ = 1. These are linearly independent, and all eigenvectors for λ = 1 are linear combinations

of x1 , x2 .

1 1 0 5 0 0

Let P = 1 0 1 . Then P−1 AP = 0 1 0.

1 −1 −2 0 0 1

Question 5(c) Let A and B be n square matrices over F . Show that AB and BA have

the same eigenvalues.

BA = A−1 ABA ⇒ |xI − BA| = |xA−1 A − A−1 ABA| = |A−1 ||xI − AB||A| = |xI − AB|

8

Thus the characteristic polynomials of AB and BA are the same, so they have the same

eigenvalues.

If A is singular,

then let rank(A) = r. Then there existP, Q non-singular such that

Ir 0 I 0

PAQ = . Now PABP−1 = PAQQ−1 BP−1 = r Q−1 BP−1 . Let Q−1 BP−1 =

0 0 0 0

B1 B2

, where B1 is r × r, B2 is r × n − r, B3 is n − r × r, B4 is n − r × n − r. Then

B3 B4

−1 Ir 0 B1 B2 B1 B2

PABP = = , so the characteristic roots of AB are the

0 0 B3 B4 0 0

same as those of B1 , along with 0 repeated n − r times.

−1 B 1 B 2 Ir 0 B 1 0

Now Q−1 BAQ = Q−1 BP PAQ = = so the character-

B3 B4 0 0 B3 0

istic roots of BA are the same as those of B1 , along with 0 repeated n − r times. Thus BA

and AB have the same characteristic roots.

9

UPSC Civil Services Main 1988 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

into a vector space Vn of dimension n over the same field can be represented as a matrix.

If T is a linear transformation of V2 into V4 such that T(3, 1) = (4, 1, 2, 1) and T(−1, 2) =

(3, 0, −2, 1), then find the matrix of T.

n

X

T(vi ) = aji wj , i = 1, . . . , m

j=1

then T corresponds to the n × m matrix A whose (i, j)’th entry is aij . In fact (v1 , . . . , vm ) =

(w1 , . . . , wn )A.

It can be easily seen that

2 1

e1 = (1, 0) = (3, 1) − (−1, 2)

7 7

1 3

e2 = (0, 1) = (3, 1) + (−1, 2)

7 7

1

and therefore

2 1

(4, 1, 2, 1) − (3, 0, −2, 1)

T(e1 ) =

7 7

1

= (5, 2, 6, 1)

7

1 ∗

= (5e + 2e∗2 + 6e∗3 + e∗4 )

7 1

1 3

T(e2 ) = (4, 1, 2, 1) + (3, 0, −2, 1)

7 7

1

= (13, 1, −4, 4)

7

1

= (13e∗1 + e∗2 − 4e∗3 + 4e∗4 )

7

5 13

2 1

Thus T corresponds to the matrix 17

6 −4 w.r.t. the standard basis.

7 4

Question 1(b) If M, N are finite dimensional subspaces of V, then show that dim(M +

N ) = dim M + dim N − dim(M ∩ N ).

{u1 , u2 , . . . , ur } to a basis {u1 , u2 , . . . , ur , v1 , . . . , vm } of M, where dim M = m + r. Com-

B

plete {u1 , u2 , . . . , ur } to a basis {u1 , u2 , . . . , ur , w1 , . . . , wn } of N , where dim N = n+r. We

shall show that = {u1 , u2 , . . . , ur , v1 , . . . , vm , w1 , . . . , wn } is a basis of M + N , proving

the result.

If u ∈ M + N , then u = v + w for some v ∈ M, w ∈ N . Since B B is a superset of the

We now show that the set

B

bases of M, N , v, w can be written as linear combination of elements of

B . Thus B

is linearly independent. If possible let

generates M + N .

⇒ u can be

n

X m

X r

X

αi vi + βi w i + γi u i = 0

i=1 i=1 i=1

Pn Pm Pr Pn

Since

Pn i=1 αi v i = − i=1 β i w i − i=1 γi u i it follows that i=1 αi vi ∈ N .P Therefore

n r

means that ni=1 αi vi −

P P

α

Pri=1 i iv ∈ M ∩ N ⇒ α

i=1 i i v = η

i=1 i i u for ηi ∈ R. This

i=1 ηi ui = 0. But {u1 , u2 , . . . , ur , v1 , . . . , vm } are linearly independent, so αi = 0, 1 ≤

i ≤ n. Similarly we can show that βi = 0, 1 ≤ i ≤ m. Then the linear indepen-

dence of {u1 , u2 , . . . , ur } shows that γi = 0, 1 ≤ i ≤ r. Thus the vectors in

early independent and form a basis of M + N , showing that the dimension of M + N is

B

are lin-

2

Question 1(c) Determine a basis of the subspace spanned by the vectors v1 = (1, 2, 3), v2 =

(2, 1, −1), v3 = (1, −1, −4), v4 = (4, 2, −2).

Solution. v1 , v2 are linearly independent because if αv1 + βv2 = 0 then α + 2β =

0, 2α + β = 0, 3α − β = 0 ⇒ α = β = 0. If v3 = αv1 + βv2 , then the three linear equations

α + 2β = 1, 2α + β = −1, 3α − β = −4 should be consistent — clearly α = −1, β = 1 satisfy

all three, showing v3 = v2 − v1 . Again suppose v4 = αv1 + βv2 , then the three linear

equations α + 2β = 4, 2α + β = 2, 3α − β = −2 should be consistent — clearly α = 0, β = 2

satisfy all three, showing v4 = 2v2 .

Hence {v1 , v2 } is a basis for the vector space generated by {v1 , v2 , v3 , v4 }.

a1 b

Question 2(a) Show that it is impossible for S = , b 6= 0 to have identical eigen-

b a2

values.

0 λ1 0

Solution. We know given S symmetric ∃O orthogonal so that O SO = , where

0 λ2

λ1 , λ

2 are eigenvalues

of S. If λ1 = λ2 , then we have S = O0 −1 (λI)O−1 = λ(OO0 )−1 = λI ⇒

λ 0

S= . Thus if b 6= 0, S cannot have identical eigenvalues.

0 λ

Question 2(b) Prove that the eigenvalues of a Hermitian matrix are all real and the eigen-

values of a skew-Hermitian matrix are either zero or pure imaginary.

Solution. See question 2(a), year 1998.

Question 2(c) If x0 Ax > 0 for all x 6= 0, A symmetric, then for all y 6= 0 y0 A−1 y > 0.

If λ is the largest eigenvalue of A, then

x0 Ax

λ = sup

x∈Rn x0 x

x6=0

.

Solution. Clearly A = A0 A−1 A ∴ x0 Ax = x0 A0 A−1 Ax = y0 A−1 y where y = Ax for any

x ∈ Rn , x 6= 0. Since |A| 6= 0, any vector y can be written as Ax, by taking x = A−1 y.

Thus x0 Ax > 0 ⇒ y0 A−1 y > 0 for all y 6= 0.

λ1 0 . . . 0

0 λ2 . . . 0

0

0

Let M = supx∈Rn xxAx . Let O be an orthogonal matrix such that O AO = .. .

0x ..

x6=0 . .

0 0 . . . λn

Let 0 6= x = Oy, then x x = y O Oy = y y. Now x Ax = y O AOy = i λi yi2 ≤ λy0 y

0 0 0 0 0 0 0

P

0 0

where λ is the largest eigenvalue of A. Thus λ ≥ xyAx 0y = xxAx 0 x , so λ ≥ M . On the other

0

hand, if x 6= 0 is an eigenvector corresponding to λ, then x Ax = λx0 x ⇒ λ = xxAx

0

0x ≤ M .

Thus λ = M as required.

3

Question 3(a) By converting A to an echelon matrix, determine its rank, where

0 0 1 2 8 9

0 0 4 6 5 3

A= 0 2 3 1 4 7

0 3 0 9 3 7

0 0 5 7 3 1

Solution. Consider

0 0 0 0 0

0 0 2 3 0

0

1 4 3 0 5

A =

2

6 1 9 7

8 5 4 3 3

9 3 7 7 1

Interchange the first row with the third, then third with fourth, fourth with fifth and fifth

with sixth to get

1 4 3 0 5

0 0 2 3 0

0

2 6 1 9 7

A ∼ 8 5

4 3 3

9 3 7 7 1

0 0 0 0 0

Now perform R3 − 2R1 , R4 − 8R1 , R5 − 9R1 to get

1 4 3 0 5

0 0 2 3 0

0

0 −2 −5 9 −3

A ∼ 0 −27 −20 3 −37

0 −33 −20 7 −44

0 0 0 0 0

Interchange the second and the third row, and perform − 12 R2 , 21 R3 to get

1 4 3 0 5

5

0 1

2

− 92 3

2

0 0 3

0 1 0

A ∼ 2

0 −27 −20 3 −37

0 −33 −20 7 −44

0 0 0 0 0

4

Perform R4 + 27R2 , R5 + 33R2 to get

1 4 3 0 5

5

0

1 2

− 92 3

2

0 3

0 0 1 0

A ∼ 95

2

0

0 2

− 237

2

7

2

125

0 0 2

− 283

2

11

2

0 0 0 0 0

95 125

Operation R4 − 2

R3 , R5 − 2

R3

yields

1 4 3 0 5

5

0

1 2

− 92

3

2

0 3

0 0 1 0

A ∼ 2

0

0 0 − 759

4

7

2

0 0 0 − 941

4

11

2

0 0 0 0 0

4

Now multiply R4 with − 759

1 4 3 0 5

5

0

1 2

− 92 3

2

0 3

0 0 1 0

A ∼ 2

14

0

0 0 1 − 759

0 0 0 − 941

4

11

2

0 0 0 0 0

941

Performing R5 + 4

R4 results in

1 4 3 0 5

5

0

1 2

− 29 3

2

0 3

0 0 1 0

A ∼ 2

14

0

0 0 1 − 759

11

0 0 0 0 2

− 941×7

1882

0 0 0 0 0

1 4 3 0 5

5

0

1 2

− 92 3

2

0 3

0 0 1 0

A ∼ 2

14

0

0 0 1 − 759

0 0 0 0 1

0 0 0 0 0

which is an echelon matrix. Its rank is clearly 5, so the rank of A = 5.

5

Question 3(b) Given AB = AC does it follow that B = C? Can you provide a counterex-

ample?

1 0 0 1

A= , B= ⇒ AB = 0

0 0 0 0

C = 0 ⇒ AC = 0, but B 6= C.

0 −1 0

Question 3(c) Find a nonsingular matrix which diagonalizes A = −1 −1 1 , B =

0 1 0

2 1 −2

1 2 −2 simultaneusly. Find the diagonal form of A.

−2 −2 3

Solution.

−2λ −1 − λ 2λ −2λ −1 − λ 0

|A − λB| = 0 ⇒ −1 − λ −1 − 2λ 1 + 2λ = 0 ⇒ −1 + λ 0 0 = 0

2λ 1 + 2λ −3λ 2λ 1 + 2λ −λ

Thus λ = 0, 1, −1. This shows that the matrices are diagonalizable simultaneously.

We now determine x1 , x2 , x3 such that (A − λB)xi = 0, i = 1, 2, 3. For λ = 0, let

x1 0 = (x1 , x2 , x3 ) be such that (A − λB)x1 = 0. Thus

0 −1 0 x1

−1 −1 1 x2 = 0

0 1 0 x3

For λ = 1, let x2 0 = (x1 , x2 , x3 ) be such that (A − λB)x2 = 0. Thus

−2 −2 2 x1

−2 −3 3 x2 = 0

2 3 −3 x3

Thus −2x1 −2x2 +2x3 = 0, −2x1 −3x2 +3x3 = 0, 2x1 +3x2 −3x3 = 0 ⇒ x2 −x3 = 0 ⇒ x1 = 0.

Thus we may take x2 0 = (0, 1, 1).

For λ = −1, let x3 0 = (x1 , x2 , x3 ) be such that (A − λB)x3 = 0. Thus

2 0 −2 x1

0 1 −1 x2 = 0

−2 −1 3 x3

6

Thus 2x1 − 2x3 = 0, x2 − x3 = 0, −2x1 − x2 + 3x3 = 0 ⇒ x1 = x2 = x3 . Thus we may take

x3 0 = (1, 1, 1).

1 0 1

Let P = 0 1 1 so that

1 1 1

1 0 1 0 −1 0 1 0 1 1 0 1 0 −1 −1 0 0 0

0

P AP = 0 1 1

−1 −1 1 0 1 1 = 0

1 1 0 0 −1 = 0 1 0

1 1 1 0 1 0 1 1 1 1 1 1 0 1 1 0 0 −1

7

UPSC Civil Services Main 1989 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

3 1 −1

Question 1(a) Find a basis for the null space of the matrix A = .

0 1 2

e∗1 + e∗2 , A(e3 ) = −e∗1 + 2e∗2 , where e1 , e2 , e3 is the standard basis of R3 and e∗1 , e∗2 is the

standard basis of R2 . Thus A(a, b, c) = e∗1 (3a + b − c) + e∗2 (b + 2c). Consequently, (a, b, c) ∈

null space of A if and only if 3a + b − c = 0, b + 2c = 0 ⇒ b = −2c, a = c. Thus null space

of A is {(c, −2c, c) | c ∈ R}. Note that rank A = 2, so the null space has dimension 1.

A basis for the null space is (1, −2, 1), any other multiple of this can also be regarded as a

basis.

Question 1(b) If W is a subspace of a finite dimensional vector space V then prove that

dim V/W = dim V − dim W.

so chosen that v1 , . . . , vn is a basis of V, dim V = n. We will show that vi + W, r + 1 ≤ i ≤ n

is a basis of V/W ⇒ dim V/W = n − r.

1

First we show linear independence:

n

X

αi (vi + W) = 0

i=r+1

Xn

⇒ αi vi + W = 0 + W

i=r+1

Xn

⇒ αi vi ∈ W

i=r+1

Xn r

X

⇒ αi vi = −αi vi (say)

i=r+1 i=1

Xn

⇒ αi vi = 0

i=1

⇒ αi = 0, 1 ≤ i ≤ n (vi are linearly independent.)

Pn

PnIf v + W is anyPnelement of V/W, Pnthen v = i=1 αi vi as v ∈ V. Therefore v + W =

i=1 αi vi +W = i=1 αi (vi +W) = i=r+1 αi (vi +W) because v1 +W = . . . = vr +W = W.

Thus vi + W, r + 1 ≤ i ≤ n generate V/W. Hence dim V/W = n − r = dim V − dim W

Question 1(c) Show that all vectors (x1 , x2 , x3 , x4 ) in the vector space V4 (R) which obey

x4 −x3 = x2 −x1 form a subspace V. Show further that V is spanned by ξ1 = (1, 0, 0, −1), ξ2 =

(0, 1, 0, 1), ξ3 = (0, 0, 1, 1).

because

= α(y4 − y3 ) + β(z4 − z3 )

= α(y2 − y1 ) + β(z2 − z1 ) ∵ y4 − y3 = y2 − y1 , z4 − z3 = z2 − z1

= a2 − a1

Clearly ξ1 , ξ2 , ξ3 are linearly independent ⇒ dim V ≥ 3. But V =

6 V4 (R) because

(1, 0, 0, 0) 6∈ V ∴ dim V < 4 ⇒ dim V = 3.

Hence ξ1 , ξ2 , ξ3 is a basis of V and therefore span V.

Question 2(a) Let P be a real skew-symmetric matrix and I the corresponding unit matrix.

Show that I − P is non-singular. Also show that Q = (I + P)(I − P)−1 is orthogonal.

2

Solution. We have proved (question 2(a), year 1998) that the eigenvalues of a skew-

Hermitian and therefore of a skew-symmetric matrix are zero or pure imaginary. This means

|I − P| =

6 0 because 1 cannot be an eigenvalue of P.

Q Q = [(I − P)−1 ]0 (I + P)0 (I + P)(I − P)−1 = (I + P)−1 (I − P)(I + P)(I − P)−1 . But

0

orthogonal.

Related Results:

then A can be written as A = (I + S)(I − S)−1 for some skew-Hermitian matrix S

provided −1 is not an eigenvalue of A.

Proof:

0 0 0 0

A = ((I − S)−1 )0 (I + S) = (I − S )−1 (I + S )

= (I + S)−1 (I − S)

0

∴ AA = (I + S)(I − S)−1 (I + S)−1 (I − S)

= (I + S)(I + S)−1 (I − S)−1 (I − S) = I

∵ (I − S)−1 (I + S)−1 = (I − S2 )−1 = (I + S)−1 (I − S)−1

0

Similarly A A = I, so A is unitary.

Now A(I − S) = I + S ⇒ A − I = (A + I)S ⇒ S = (A + I)−1 (A − I). It can be checked

as above that S is skew-Hermitian. Note that |A + I| =

6 0.

can be thus represented provided it does not have −1 as its eigenvalue.

3. If S is real, S0 = −S and S2 = −I, then S is orthogonal and of even order, and there

exist non-null vectors x, y such that x0 x = y0 y = 1, x0 y = 0, Sx + y = 0, Sy = x.

Proof: S0 S = −SS = I, so S is orthogonal, |S| 6= 0 ⇒ S is of even order.

Choose y such that y0 y = 1. Then y0 Sy = (y0 Sy)0 = y0 S0 y = −y0 Sy ⇒ y0 Sy = 0.

Set x = Sy, then y0 x = 0, Sx + y = 0. In addition, x0 x = y0 S0 Sy = y0 y = 1.

Question 2(b) Show that an n × n matrix A is similar to a diagonal matrix if and only if

the set of eigenvectors of A includes a set of n linearly independent vectors.

0

corresponding to ri , i = 1, 2. If A is Hermitian, show that ξ1 ξ2 = 0.

3

1 1 , and B =

Question

3(a) Find the roots of the equation |xA − B| = 0 where A = 1 4

0 3 . Use the result to show that the real quadratic forms F = x2 + 2x x + 4x2 , G = 6x x

3 0 1 1 2 2 1 2

can be simultaneously reduced by a non-singular linear substitution to y12 + y22 , y12 − 3y22 .

x x − 3

= 4x2 − (x − 3)2 ⇒ ±2x = x − 3 ⇒ x = −3, 1.

Solution. |xA − B| =

x − 3 4x

Let x1 = (x1 , x2 ) be a row vector such that (A − B) xx12 = 0.

1 −2 x1

= 0 ⇒ x1 − 2x2 = 0

−2 4 x2

Let x2 = (x1 , x2 ) be a row vector such that (−3A − B) xx12 = 0.

−3 −6 x1

= 0 ⇒ x1 + 2x2 = 0

−6 −12 x2

x1 Ax01 = (2, 1) 11 14 21 =(2, 1) 36 = 12.

x2 Ax02 = (−2, 1) 11 14 −2 1 = (−2, 1) −12 = 4.

0

Note that x1 Ax2 = 0.

x1 Bx01 = (2, 1) 03 30 21 =(2, 1) 36 = 12.

= −12.

0

Note that x1 Bx2 = 0.

√1 0

Thus if P = [x1 0 , x2 0 ], then P0 AP = 12 0 , and P0 BP = 12 0

0 4 0 −12 . Let Q = 12

0 12

,

then Q0 P0 APQ = 10 01 and Q0 P0 BPQ = 10 −3

0

as desired. Thus the required non-singular

linear transformation is PQ.

−1

− tan 2θ tan 2θ

cos θ − sin θ 1 1

Question 3(b) Show that = .

sin θ cos θ tan 2θ 1 − tan 2θ 1

Solution.

− tan 2θ cos2 2θ − sin 2θ cos 2θ

1

R.H.S = θ

tan 2 1 sin 2θ cos 2θ cos2 2θ

cos2 2θ − sin2 2θ −2 sin 2θ cos 2θ

= = L.H.S

2 sin 2θ cos 2θ − sin2 2θ + cos2 2θ

4

0 1

Question 3(c) Verify the Cayley-Hamilton theorem for A = .

−2 3

−λ 1

Solution. The characteristic equation for A is = 0 ⇒ −3λ + λ2 + 2 = 0

−2 3 − λ

2

Thus according

theorem A − 3A + 2I = 0.

to the Cayley-Hamilton

0 1 0 1 −2 3

A2 = =

−2

3 −2 3 −6 7

−2 3 0 1 1 0 0 0

−3 +2 =

−6 7 −2 3 0 1 0 0

0 1

Thus the Cayley Hamilton theorem is verified for A = .

−2 3

5

UPSC Civil Services Main 1990 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

1 Linear Algebra

Question 1(a) State any definition of the determinant of an n×n matrix and show that the

determinant function is multiplicative i.e. det AB = det A det B for any two n × n matrices

A, B. You may assume the matrices to be real.

Solution. Let π be a permutation of 1, . . . , n. Define sign(π) as follows: count the number

of pairs of numbers that need to be interchanged to get to π from the identity permutation.

If this is even, the sign is 1, and if it is odd, the sign is −1. Now if Π is the set of all

permutations of 1, . . . , n, define

X Y

det A = sign(π) aiπ(i)

π∈Π i

Note that the det A is n-linear i.e. if we perform any row or column operation on A the

determinant is unchanged. Also, if any two rows are swapped, the sign of the determinant

changes. These are simple consequences of the above definition.

Consider the 2n × 2n matrix

a11 a12 . . . a1n 0 . . . 0

a21 a22 . . . a2n 0 . . . 0

. .. .. .. ..

.

. . . . .

a a . . . ann 0 . . . 0

P = n1 n2

−1 0 . . . 0 b11 . . . b1n

0 −1 . . . 0 b21 . . . b2n

. .. .. .. ..

.. . . . .

0 0 . . . −1 bn1 . . . bnn

1

Then det P = det A det B, because if for any permutation π, π(i) > n for i ≤ n, then

the corresponding element of the sum is 0 as aiπ(i) = 0. Thus π(i) ≤ n if i ≤ n, and

consequently π(j) > n if j > n. So each permutation consists of a permutation of 1, . . . , n

and a permutation of n + 1, . . . , 2n, consequently we can factor the sum, to get det P =

det A det B.

Now we perform a series of column operations to P — add b11 C1 + . . . + bn1 Cn to Cn+1 ,

to get

a11 a12 . . . a1n c11 0 . . . 0

a21 a22 . . . a2n c21 0 . . . 0

. .. .. .. ..

.

. . . . .

an1 an2 . . . ann cn1 0 . . . 0

−1 0 . . . 0 0 b12 . . . b1n

0 −1 . . . 0 0 b22 . . . b2n

. .. .. .. .. ..

.. . . . . .

0 0 . . . −1 0 bn2 . . . bnn

where C = AB = (cij ). Similarly add b12 C1 + . . . + bn2 Cn to Cn+2 , . . ., b1n C1 + . . . + bnn Cn

to C2n to get

A C

−1 0 . . . 0

0 −1 . . . 0

0

...

0 0 . . . −1

We can now verify that det P = det C. Any permutation π that leads to a non-zero term

in the determinant sum must have π(j) = j − n for j > n, thus piπ(i) = −1, i > n. Also

π(j) > n for j ≤ n, so any such π can be written as a permutation of 1, . . . , n followed by a

series of swaps of the i-th number with the (n + i)-th number, which is n + i. Also sign(π)

is the same as the sign of the corresponding permutation π 0 of 1, . . . , n — we first do π 0 by

exchanges and then additionally swap the i-th element with the (i + n)-th element, for each

i ≤ n. Now if n is even, this involves an even number of additional swaps, and multiply by

(−1)n corresponding to piπ(i) for i > n, otherwise we get an odd number of additional swaps,

flipping the sign, but we still multiply by (−1)n = −1.

Thus det P = det C = det A det B.

Question 1(b) Prove Laplace’s formula for simulataneous expansion of the determinant by

the first row and column; that given an (n+1)×(n+1) matrix in the block form M = αγ D β

,

where α is a scalar, β is a 1 × n matrix (a row vector), γ is a n × 1 matrix (a column vector),

and D is an n × n matrix, then det M = α det D − βD0 γ 0 , where D0 is the matrix of cofactors

of D and βD0 γ 0 stands for the matrix product of size 1 × 1.

2

a21 a22 . . . a2,n+1

γ = ... and D = ... ..

.

.

an+1,1 an+1,2 . . . an+1,n+1

det M = a11 |A11 | − a12 |A12 | + . . . + (−1)n a1,n+1 |A1,n+1 | where Aij is the minor corre-

sponding to aij (formed Pby deleting the i-th row and j-th column of A). Clearly D = A11 ,

n+1

so det M = α det D − j=2 (−1)j a1j det A1j . Now

a21 a 22 . . . a 2,j−1 a 2,j+1 . . . a 2,n+1

a31 a32 . . . a3,j−1 a3,j+1 . . . a3,n+1

|A1j | = ..

.. .. .. ..

. . . . .

an+1,1 an+1,2 . . . an+1,j−1 an+1,j+1 . . . an+1,n+1

Let Bij be the minor of aij in D. Expanding |A1j | in terms of the first column, we get

n+1 X

X n+1

det M = α det D − a1j ai1 |Bij |(−1)i (−1)j

j=2 i=2

a21

= α det D − (a12 a13 . . . a1,n+1 )(cij ) ...

an+1,1

= α det D − βD0 γ

Question 1(c) For M as in 1(b), if D is invertible, show that det M = det D(α − βD−1 γ).

det M = α det D − βD0 γ = α det D − βD−1 det Dγ = det D(α − βD−1 γ).

Question 2(a) Write the definition of the characteristic polynomial, eigenvalues and eigen-

vectors of a square matrix. Also say briefly something about the importance and/or applica-

tions of these notions.

the characteristic polynomial of A. The roots of this polynomial are called the eigenvalues

of A. If λ is an eigenvalue of A, then all the non-zero vectors x such that Ax = λx are

called eigenvectors of A corresponding to λ.

Many problems in mathematics and other sciences require finding eigenvalues and eigen-

vectors of an operator.

3

• Eigenvalues can be used to find a very simple matrix for an operator — either diagonal

or a block diagonal form. This can be used to compute powers of matrices quickly.

• If one wishes to solve a linear differential system like x0 = Ax, or study the local

properties of a nonlinear system, finding the diagonal form of the matrix can give us

a decoupled form of the system, allowing us to find the solution or understand its

qualitative behavior, like its stability and oscillatory behavior.

eigenvector (corresponding to the eigenvalue with the largest absolute value) of a very

large matrix (the adjacency matrix of the web graph) — this is used to find the relative

importance of documents on the World Wide Web. Similar calculations are used to

compute the stationary distribution of a Markov system.

• In mechanics, the eigenvectors of the inertia tensor are used to define the principal

axes of a rigid body, which are important in analyzing the rotation of the rigid body.

• Eigenvalues can be used to compute low rank approximations to matrices, which help

in reducing the dimensionality of various problems. This is used in statistics and

operations research to explain a large number of observables in terms of a few hidden

variables + noise.

• Eigenvalues can help us determine the form of a quadric or higher dimensional surface

— see the relevant section in year 1999.

• In quantum mechanics, states are represented by unit vectors, while observable quan-

tities (like position and energy) are represented by Hermitian matrices. The basic

problem in any quantum system is the determination of the eigenvalues and eigenvec-

tors of the energy matrix. The eigenvalues are the observed values of the observable

quantity, and discreteness of the eigenvalues leads to the quantization of the observed

values.

Question 2(b) Show that a Hermitian matrix possesses a set of eigenvectors which form

an orthonormal basis. State briefly how or why a general n × n complex matrix may fail to

possess n linearly independent eigenvectors.

Solution. Let H be Hermitian, and λ1 , . . . , λn its eigenvalues, not necessarily distinct. Let

x1 with norm 1 be an eigenvector corresponding to λ1 . Then there exists (from a result

analogous to the result used in question 3(a), year 1995) a unitary matrix U such that x1 is

its first column. Therefore

−1 0 λ1 L

U1 HU1 = U1 HU1 =

0 H1

4

0

where H1 is (n − 1) × (n − 1) and L is (n − 1) × 1. Since U1 HU1 is Hermitian, it follows

that L = 0. Consequently

0 λ1 0

U1 HU1 =

0 H1

Now H1 is Hermitian with eigenvalues λ2 , . . . , λn . Repeating the above argument, we find

U∗2 an (n − 1) × (n − 1) unitary matrix such that

0 ∗ λ 2 0

U∗2 H1 U2 =

0 H2

1 0

If U2 = then U2 is unitary, and

0 U∗2

λ1 0 0

0 0

U2 U1 HU1 U2 = 0 λ2 0

0 0 H2

Repeating this process or by induction, we can get U unitary such that

λ1 · · · 0

0 .. ..

U HU = . .

0 ··· λn

system.

A complex matrix A would fail to have n eigenvectors which are linearly independent

if A is not diagonalizable i.e. we cannot find P such that P−1 AP is a diagonal matrix.

For example if A = 10 1c , c 6= 0, and x1, x2 are two independent eigenvectors of A, then

−1 1 0 1 0

P = [x1 x2 ] would lead to P AP = 0 1 ⇒ A = 0 1 which is false.

Question 2(c) Define the minimal polynomial and show that a complex matrix is diago-

nalizable (i.e. conjugate to a diagonal matrix) if and only if the minimal polynomial has no

repeated root.

complex coefficients of least degree such that f (A) = 0, then f (x) is called the minimal

polynomial of A. The Cayley-Hamilton therem tells us that any n × n complex matrix A

satisfies the degree n polynomial equation |A − xI| = 0, so the minimal polynomial exists

and is of degree ≤ n.

A complex n × n matrix can be thought of as a linear transformation from Cn to Cn . Let

T : V −→ V, dim V Q = n. Let the minimal polynomial of T be p(x), having distinct roots

c1 , . . . , ck , so p(x) = kj=1 (x − cj ). We shall show that T is diagonalizable.

If k = 1, then the minimal polynomial is x−c, thus T−cI = 0, so T = cI is diagonalizable.

So assume k > 1.

5

k

x−ci

Q

Consider the polymonials pj = cj −ci

. Clearly pj (ci ) = 0 for i 6= j, and pi (ci ) = 1. This

i=1

i6=j

implies that the polynomials p1 , . . . , pk are linearly independent, and each one is of degree

k − 1 < k. Thus these form a basis of the space Pof polynomials of degree ≤ k − 1. Thus

k

given any polynomial g of degree ≤ k − 1, g = i=1 αi pi , where αi = g(ci ). In particular,

1 = ki=1 pi , x = ki=1 ci pi . Thus

P P

k

X k

X

I= pi (T), T = ci pi (T)

i=1 i=1

Moreover pi (T)pj (T) = 0, i 6= j because pi (x)pj (x) is divisible by the minimal polynomial

of T. Also pj (T) 6= 0, 1 ≤ j ≤ k, because the degree of pj is less than k, the degree of the

minimal polynomial of T.

Set Vi = pi (T)V, then V = I(V) = ki=1 pi (T)V = V1 + . . . + Vk . We shall now show that

P

Vi = Vci , the eigenspace of T with respect to ci .

v ∈ Vi ⇒ v = pi (T)w for some w ∈ V. Since (x−ci )pi is divisible by p, (T−ci I)v = 0, so

Tv = ci v so v ∈ Vci . Conversely, if v ∈ Vci , then Tv = ci v, or (T−ci I)v = 0 ⇒ pj (T)v = 0

for j 6= i. Since

Pv = pi (T)v + . . . + pk (T)v, we get v = pi (T)v ⇒ v ∈ Vi .

Thus V = ki=1 Vci so V has a basis consisting of eigenvectors, so T is diagonalizable.

Conversely let T be diagonalizable, then we shall show that the minimal polynomial of

T has distinct roots. Let

λ1 0 . . . 0

0 λ2 . . . 0

−1

P TP = ..

..

. .

0 0 . . . λn

and out of λ1 , . . . , λn , let λ1 , . . . , λk be distinct. Let g(x) = (x − λ1 ) . . . (x − λk ). Then

v ∈ V ⇒ v = v1 + . . . + vk where vi ∈ Vλi , the eigenspace of λi . Thus g(T)(v) = 0, so

g(T) = 0. Thus g(x) is divisible by the minimal polynomial of T. Since g(x) has all distinct

roots, it immediately follows that the minimal polynomial also has all distinct roots.

a b is expressible in the form LDU, where

Question 3(a) Show that a 2 × 2 matrix M = c d

L has the form α1 01 , D is diagonal and U has the form 10 β1 If and only if either a 6= 0

or a = b = c = 0. Also show that when a 6= 0 the factorization M = LDU is unique.

a1 a1 β

a1 α a1 αβ+a2 . Thus M = LDU ⇒ a1 = a, a1 β = b, a1 α = c, a αβ + a = d. Thus if a = 0,

010 1 β 2

0 0 1 0

then b = c = 0 and d = a2 , In this case, M = 0 d = α 1 0 d 0 1 whatever α, β may

be, i.e. M can be represented as LDU in infinitely many ways.

If a 6= 0, then a1 = a, β = ab , α = ac , a2 = d − bca are uniquely determined. Thus

a 0 b

M = ac db = 1ac 01 0 d− bca 10 1a and has a unique representation.

6

Conversely, if M = 00 d0 , i.e. a = b = c = 0, then M = α1 01 00 d0 10 β1 for any

a 0 b

α, β ∈ R. If M = ac db , a 6= 0, then M = LDU with L = 1ac 01 , D = 0 d− bca , U = 10 1a

as shown above.

Question 3(b) Suppose a real matrix has eigenvalue λ, possibly complex. Show that there

exists a real eigenvector for λ if and only if λ is real.

Rn to Rn . Since |A − λI| = 0, the rows are linearly dependent, so there exists x ∈ Rn , x 6= 0

such that (A − λI)x = 0 ⇒ Ax = λx. Thus there exists a real eigenvector for λ.

Conversely, suppose Ax = λx, x ∈ Rn , x 6= 0. then λx = Ax = Ax = λx = λx ⇒

(λ − λ)x = 0 ⇒ λ − λ = 0 ∵ x 6= 0 ⇒ λ = λ i.e. λ is real.

n

Question 3(c) If a 2×2 matrix

A has order n, i.e. A = I2 , then show that A is conjugate

cos θ sin θ

to the matrix where θ = 2πm for some integer m.

− sin θ cos θ n

Solution. Note: A has to be real, otherwise the result is false: if α1 , α2 are two distinct

α1 0

n-th roots of unitysuch that α1 6= α2 , then A = 0 α2 has order n, but A is not conjugate

cos θ sin θ

to whose eigenvalues are complex conjugates of each other.

− sin θ cos θ

An = I ⇒ eigenvalues of A are n-th roots of unity. If A has repeated eigenvalues, then

these can be 1 or −1, because eigenvalues of real matrices are complex conjugates of each

other, so the repeated eigenvalues must be real, and they also must be roots of 1.

−1

1 c .

Case 1: A has eigenvalues 1, 1. There exists P non-singular such that P AP = 0 1

Now (P−1 AP)n = 10 nc = P−1An P = P−1 I2 P = I2 , so nc = 0 ⇒ c = 0. Thus A is

1

cos θ sin θ

conjugate to I2 = , θ = 2πn .

− sin θ cos θ n

Case 2: A has eigenvalues −1, −1. There exists P non-singular such that P−1 AP =

−1 c −1 n −1 nc

1 nc

n

0 −1 . Now P A P = 0 −1 or 0 1 , according as nis odd or even. But A = I2 ,

cos θ sin θ

therefore n is even and c = 0. Thus A is conjugate to −I2 = , θ = 2πm ,m =

− sin θ cos θ n

n

2

.

Case 3: A hasdistinct eigenvalues

λ1 , λ2 . Then λ1 = λ2 . If λ1 = cos θ + i sin θ, with

2πm cos θ sin θ cos θ − λ sin θ

θ = n , set B = . The eigenvalues of B are roots of =

− sin θ cos θ − sin θ cos θ − λ

0 ⇒ λ = cos θ ± i sin θ. Since A and B have the same eigenvalues λ1 , λ2 distinct, both are

λ1 0

conjugate to 0 λ2 and are therefore conjugate to each other.

7

UPSC Civil Services Main 1991 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Let V(R) be the real vector space of all 2×3 matrices with real entries. Find

a basis of V(R). What is the dimension of V(R).

1 0 0 0 1 0 0 0 1

Solution. Let A1 = , A2 = , A3 =

0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0

and B1 = , B2 = , B3 = . Clearly Ai , Bi , i = 1, 2, 3 ∈

1 0 0 0 1 0 0 0 1

V(R). These generate V(R) because

a1 a2 a3

A= = a1 A1 + a2 A2 + a3 A3 + b1 B1 + b2 B2 + b3 B3

b1 b2 b3

They are linearly independent because if the RHS in the above equation was equal to

0 0 0

, then ai = 0, bi = 0 for i = 1, 2, 3. Thus Ai , Bi , i = 1, 2, 3 is a basis for V(R) and

0 0 0

the dimension of V(R) is 6.

Question 1(b) Let C be the field of complex numbers and let T be the function from C3 to

C3 defined by

What is the rank of T?

1

3. What are the conditions on a, b, c so that (a, b, c) is in the null space of T? What is

the nullity of T?

Solution. T(e1 ) = (1, 2, −1), T(e2 ) = (−1, 1, −2), T(e3 ) = (2, 0, 2). Clearly T(e1 ) and

T(e3 ) are linearly independent. If

of T(e1 ) and T(e3 ). Thus rank of T is 2, nullity of T is 1.

If (a, b, c) is in the range of T, then (a, b, c) = α(1, 2, −1) + β(2, 0, 2). Thus α + 2β =

a− b

a, 2α = b, −α + 2β = c. From the first two equations, α = 2b , β = 2 2 . The equations would

be consistent if − 2b + a − 2b = c, or a = b + c. So the condition for (a, b, c) to belong to the

range of T is a = b + c.

If (a, b, c) ∈ null space of T, then a − b + 2c = 0, 2a + b = 0, −a − 2b + 2c = 0. Thus

3a + 2c = 0, so a = − 2c 3

, b = 4c

3

. Thus the conditions for (a, b, c) to belong to the null space of

T are 3a + 2c = 0, 3b = 4c. Thus the null space consists of the vectors {(− 2c , 4c , c) | c ∈ R},

3 3

showing that the nullity of T is 1.

Question 1(c) If A = 1 2

1 3 , express A6 − 4A5 + 8A4 − 12A3 + 14A2 as a linear polynomial

in A.

Solution. Characteristic polynomial of A is 1−λ 2 2

1 3−λ = (λ − 3)(λ − 1) − 2 = λ − 4λ + 1.

By the Cayley Hamilton theorem, A2 − 4A + I = 0. Dividing the given polynomial by

A2 − 4A + I, we have

= A4 (A2 − 4A + I) + 7A4 − 12A3 + 14A2

= (A4 + 7A2 )(A2 − 4A + I) + 16A3 + 7A2

= (A4 + 7A2 + 16A)(A2 − 4A + I) + 71A2 − 16A

= (A4 + 7A2 + 16A + 71I)(A2 − 4A + I) + 268A − 71I

(−x2 , x1 ).

2. What is the matrix of T in the ordered basis B = {α1 , α2 } where α1 = (1, 2), α2 =

(1, −1)?

2

Solution. T(e1 ) = (0, 1) = e2 , T(e2 ) = (−1, 0) = −e1. Thus (T(e1 ), T(e2 )) =

(e1 e2 ) 01 −1 0 −1

0 . So the matrix of T in the standard basis is 1 0 .

T(α1 ) = (−2, 1), T(α2 ) = (1, 1). If (a, b) = xα1 + yα2 , then x + y = a, 2x − y = b, so

x = a+b

3

, y = 2a−b

3

. This shows that

T(α1 ) = (−2, 1) = − 13 α1 − 35 α2

T(α2 ) = (1, 1) = 23 α1 + 13 α2

1 2

−3 3

Thus (T(α1 ) T(α2 )) = (α1 α2 ) 5 1 . Consequently the matrix of T in the ordered

1 2 − 3 3

−3 3

basis B is .

− 53 31

Question 2(b)

Determine

a non-singular matrix P such that P0 AP is a diagonal matrix,

0 1 2

where A = 1

0 3. Is the matrix congruent to a diagonal matrix? Justify your answer.

2 3 0

Solution. The quadratic form associated with A is Q(x, y, z) = 2xy + 4xz + 6yz. Let

x = X, y = X + Y, z = Z (thus X = x, Y = y − x, Z = z). Then

Q(X, Y, Z) = 2X 2 + 2XY + 4XZ + 6XZ + 6Y Z

= 2X 2 + 2XY + 10XZ + 6Y Z

Y 5 2 Y 2 25 2

= 2(X + + Z) − − Z +YZ

2 2 2 2

Y 5 2 1

= 2(X + + Z) − (Y − Z)2 − 12Z 2

2 2 2

Put

Y 5 x y 5z

ξ = X+ + Z= + +

2 2 2 2 2

η = Y − Z = −x + y − z

ζ = Z=z

1 1 5 −1

1 − 12 −3

x 2 2 2

ξ ξ

y = −1 1 −1 η = 1 1 −2 η

2

z 0 0 1 ζ 0 0 1 ζ

Q(x, y, z) transforms to 2ξ 2 − 21 η 2 − 12ζ 2 . Thus

2 0 0

P0 AP = 0 − 12 0

0 0 −12

1 − 12 −3

0 0 1

3

Question 2(c) Reduce the matrix

1 3 4 −5

−2 −5 −10 16

5 9 33 −68

4 7 30 −78

1 3 4 −5

0 1 −2 6

A≈ 0 −6 13

−43

0 −5 14 −58

1 3 4 −5

0 1 −2 6

A≈

0 0 1 −7

0 0 4 −28

Operations R4 − 4R3 ⇒

1 3 4 −5

0 1 −2 6

A≈

0 0 1 −7

0 0 0 0

Operation R1 − 3R2 ⇒

1 0 10 −23

0 1 −2 6

A≈

0 0 1 −7

0 0 0 0

Operations R1 − 10R3 , R2 + 2R3 ⇒

1 0 0 47

0 1 0 −8

A≈

0 0 1 −7

0 0 0 0

4

Question 3(a) U is an n-rowed unitary matrix such that |I − U| 6= 0, show that the matrix

H defined by iH = (I + U)(I − U)−1 is Hermitian. If eiα1 , . . . , eiαn are the eigenvalues of U

then cot α21 , . . . , cot α2n are eigenvalues of H.

Solution.

(iH)(I − U) = (I + U)

0 0 0

⇒ (I − U )(iH) = (I + U )

0 0 0 0

Substituting I = U U, we have from the second equation that U (U − I)(iH) = U (U + I).

0 0 0

So (iH) = −iH = −(I + U)(I − U)−1 = −iH, so H = H, thus H is Hermitian.

If an eigenvalue of a nonsingular matrix A is λ, then λ−1 is an eigenvalue of A−1 ∵ Ax =

λx ⇒ λ−1 x = A−1 x, note that λ 6= 0 ∵ |A| 6= 0. Thus the eigenvalues of H are

1 1 + eiαj

,1 ≤ j ≤ n

i 1 − eiαj

eiαj /2 + e−iαj /2

= −i −iαj /2 ,1 ≤ j ≤ n

e − eiαj /2

eiαj /2 +e−iαj /2

2

= ,1 ≤ j ≤ n

e−iαj /2 −eiαj /2

2i

cot αj

= ,1 ≤ j ≤ n

2

A is non-singular then there exist 2n matrices X such that X2 = A. What happens in case

A is a singular matrix?

Solution. There exists √P non-singular

√ such that P−1 AP = diagonal[λ1 , . . . , λn ].

Let Y1 = diagonal[ λ1 , . . . , λn ], and let X = PYP−1 . Then X2 = PYP−1 PYP−1 =

−1

PY2 P = A. Thus any of the 2n matrices √ formed

√ by choosing a sign for each of the

−1

diagonal entries from X = P diagonal[± λ1 , . . . , ± λn ] P has the same property (note

that they are all distinct).

If one of the eigenvalues is zero, the number of matrices X would become 2n−1 , since we

would have one less choice.

Question 3(c) Show that a real quadratic x0 Ax is positive definite if and only if there exists

a non-singular matrix B such that A = B0 B.

Solution. If A = B0 B, then x0 Ax = x0 B0 Bx = X0 X, where X = Bx. Now if x 6= 0, then

Bx 6= 0, as B is nonsingular, and 0 is not its eigenvalue. Thus x0 Ax = X0 X > 0, so x0 Ax

is positive definite.

Conversely, see the result used in the solution of question 2(c), year 1992.

5

UPSC Civil Services Main 1992 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Let U and V be vector spaces over a field K and let V be of finite dimension.

Let T : V −→ U be a linear transformation, prove that dim V = dim T(V) + dim nullity T.

of R3 . Find a basis of S.

α2 (x2 , y2 , z2 ) ∈ S because (α1 x1 + α2 x2 ) + (α1 y1 + α2 y2 ) + (α1 z1 + α2 z2 ) = α1 (x1 + y1 + z1 ) +

α2 (x2 + y2 + z2 ) = 0. Thus S is a subspace of R3 .

Clearly (1, 0, −1), (1, −1, 0) ∈ S and are linearly independent. Thus dim S ≥ 2. However

(1, 1, 1) 6∈ S, so S = 6 R3 . Thus dim S = 2 and {(1, 0, −1), (1, −1, 0)} is a basis for S.

Solution.

1

1.

= (2αx, −αx) + (2βy, −βy)

= αT(x) + βT(y)

2. T(2(1, 1)) = T(2, 2) = (4, 2, 2) 6= 2T(1, 1) = 2(1, 1, 1) Thus T is not a linear transfor-

mation.

3.

= (αx1 + βx2 + αy1 + βy2 , αy1 + βy2 , αx1 + βx2 )

= α(x1 + y1 , y1 , x1 ) + β(x2 + y2 , y2 , x2 )

= αT(x1 , y1 ) + βT(x2 , y2 )

4. T(2(0, 0)) = T(0, 0) = (1, −1) 6= 2T(0, 0) Thus T is not a linear transformation.

Question 2(a) Let T : M2,1 −→ M2,3 be a linear transformation defined by (with the usual

notation)

1 2 1 3 1 6 1 0

T = ,T =

0 4 1 5 1 0 0 2

x

Find T .

y

Solution.

x 1 1 1

= x −y +y

y 0 0 1

x 2 1 3 6 1 0 2x + 4y x 3x − 3y

T = (x − y) +y =

y 4 1 5 0 0 2 4x − 4y x − y 5x − 3y

2

Question 2(b) For what values of η do the following equations

x+y+z = 1

x + 2y + 4z = η

x + 4y + 10z = η 2

1 1 1

Solution. Since the determinant of the coefficient matrix 1 2 4 is 0, the system has to

1 4 10

be consistent to be solvable.

Clearly x + 4y + 10z = 3(x + 2y + 4z) − 2(x + y + z). Thus for the system to be consistent

we must have η 2 = 3η − 2, or η = 1, 2.

If η = 1, then x + y + z = 1, x + 2y + 4z = 1 so y + 3z = 0, or y = −3z, x = 1 + 2z. Thus

the space of solutions is {(1 + 2z, −3z, z) | z ∈ R}. Note that the rank of the coefficient

matrix is 2, and consequently the space of solutions is one dimensional.

If η = 2, then x + y + z = 1, x + 2y + 4z = 2, so y + 3z = 1 or y = 1 − 3z, hence x = 2z.

Consequently, the space of solutions is {(2z, 1 − 3z, z) | z ∈ R}.

Question 2(c) Prove that a necessary and sufficient condition of a real quadratic form

x0 Ax to be positive definite is that the leading principal minors of A are all positive.

Solution. Let all the principal minors be positive. We have to prove that the quadratic

form is positive definite. We prove the result by induction.

If n = 1, then a11 x2 > 0 ⇔ a11 > 0. Suppose as induction hypothesis the result is true

B B1

for n = m. Let S = B01 k be a matrix of a quadratic form in m + 1 variables, where

B is m × m, B1 is m × 1 and k is a single element. Since all principle minors of B are

leading principal minors of S, and are hence positive, the induction hypothesis gives that B

is positive definite. This means that there exists a non-singular m × m matrix P such that

P0 BP = Im (We shall prove this presently). Let C be an m-rowed column to be determined

soon. Then

0

P0 BP P0 BC + P0 B1

P 0 B B1 P C

=

C0 1 B0 1 k 0 1 C0 B0 P + B01 P C0 BC + C0 B1 + B0 1 C + k

0 0

P 0 B B1 P C P BP 0

=

C0 1 B0 1 k 0 1 0 B0 1 C + k

0 0 2

a single element. Since |S|

> 0,itfollows that

B 1 C + k > 0, so let B 1 C + k = α . Then

P C Im 0

Q0 SQ = Im+1 with Q = . Thus the quadratic forms of S and Im+1 take

0 1 0 α−1

the same values. Hence S is positive definite, so the condition is sufficient.

3

The condition is necessary - Since x0 Ax is positive definite, there is a non-singular matrix

P such that P0 AP = I ⇒ |A||P|2 = 1 ⇒ |A| > 0.

Let 1 ≤ r < n. Let xr+1 = . . . = xn = 0, then we obtain a quadratic form in r variables

which is positive definite. Clearly the determinant of this quadratic form is the r×r principal

minor of A which shows the result.

Proof of the result used: Let A be positive definite, then there exists a non-singular P

such that P0 AP = I.

We will prove this by induction. If n = 1, then the form corresponding to A is a11 x2 and

√

a11 > 0, so that P = ( a11 ).

Take

1 −a−1 11 a12 0 ... 0

0

P1 = ..

. (n − 1) × (n − 1)

0

then

a11 0 a13 ... a1n

0

P01 AP1 = a13

..

. (n − 1) × (n − 1)

a1n

Repeating this process, we get a non-singular Q such that

a11 0 ... 0

Q0 AQ = ...

(n − 1) × (n − 1)

0

Given the (n − 1) × (n − 1) matrix on the lower right, we get by induction P∗ s.t. P∗ 0 ((n −

1) × (n − 1) matrix)P∗ is diagonal. Thus ∃P, |P| 6= 0, P0 AP = [α1 , . . . , αn ] say. Take R =

√ √

diagonal[ α1 , . . . , αn ], then R0 P0 APR = In .

2 1

Question 3(a) State the Cayley-Hamilton theorem and use it to find the inverse of 4 3 .

characteristic equation of A, then the Cayley-Hamilton theorem says that An + a1 An−1 +

. . . + an I = 0 i.e. a matrix satisfies its characteristic

equation.

The characteristic equation of A = 4 3 is2 1

2 − λ 1

= λ2 − 5λ + 2 = 0

4 3 − λ

4

By the Cayley-Hamilton theorem, A2 − 5A + 2I = 0, so A(A − 5I) = −2I, thus A−1 =

− 12 (A − 5I). Thus

3

− 21

−1 1 2 1 5 0

A =− − = 2

2 4 3 0 5 −2 1

x2 + 2xy, 8x2 − 4xy + 5y 2

and give the transformation employed.

8 −2

Solution. Let A = 11 10 , B = −2 5

1 − 8λ 1 + 2λ

Let 0 = |A − λB| = = −5λ + 40λ2 − 4λ2 − 4λ − 1

1 + 2λ −5λ

√

Thus 36λ2 − 9λ − 1 = 0, so λ = 9± 81+144 72

= 13 , − 12

1

.

x1 1 5 5

Let (x1 , x2 ) be the vector such that (A − λB) x2 = 0 with λ = 3 . Thus − 3 x1 + 3 x2 =

0 ⇒ x1 = x2 . We take x1 = 11 so that (A − λB)x1 = 0 with λ = 31 . Similarly, if (x1 , x2 ) is

, then 53 x1 + 56 x2 = 0, so 2x1 + x2 = 0.

1

We take x2 = −2 .

Now

x01 Ax1 = ( 1 1 ) 11 10 11 = ( 1 1 ) 21 = 3

x02 Ax2 = ( 1 −2 ) 11 10 −2 = ( 1 −2 ) −1

1

1 = −3

x01 Ax2 = ( 1 1 ) 11 10 −2 = ( 1 1 ) −1

1

1 =0

If P = (x1 x2 ), then P0 AP = 30 −3

0

, thus x2 + 2xy ≈ 3X 2 − 3Y 2 by P = 11 −2 1

.

Similarly

x01 Bx1 = ( 1 1 ) −2 8 −2

1 6

5 1 = (1 1) 3 = 9

x02 Bx2 = ( 1 −2 ) −2 8 −2

1 12

= ( 1 −2 ) −12 = 36

0 8 −2

5

1−2 12

x1 Bx2 = ( 1 1 ) −2 5 −2 = ( 1 1 ) −12 = 0

Y y

Question 3(c) Prove that the characteristic roots of a Hermitian matrix are all real, and

the characteristic roots of a skew Hermitian matrix are all zero or pure imaginary.

Solution. For Hermitian matrices, see question 2(c), year 1995.

0 0

If H is skew-Hermitian, then iH is Hermitian, because (iH) = iH = −iH = iH as

0

H = −H . Thus the eigenvalues of iH are real. Therefore the eigenvalues of H are −ix

where x ∈ R. So they must be 0 (if x = 0) or pure imaginary.

5

UPSC Civil Services Main 1993 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Show that the set S = {(1, 0, 0), (1, 1, 0), (1, 1, 1), (0, 1, 0)} spans the vector

space R3 but is not a basis set.

Solution. The vectors (1, 0, 0), (0, 1, 0), (1, 1, 1) are linearly independent, because α(1, 0, 0)+

β(1, 1, 0) + γ(1, 1, 1) = 0 ⇒ α + γ = 0, β + γ = 0, γ = 0 ⇒ α = β = γ = 0.

Thus (1, 0, 0), (1, 1, 0), (1, 1, 1) is a basis of R3 , as dimR R3 = 3.

Any set containng a basis spans the space, so S spans R3 , but it is not a basis because

the four vectors are not linearly independent, in fact (1, 1, 0) = (1, 0, 0) + (0, 1, 0).

Question 1(b) Define rank and nullity of a linear transformation. If V is a finite dimen-

sional vector space and T is a linear operator on V such that rank T2 = rank T, then prove

that the null space of T is equal to the null space of T2 , and the intersection of the range

space and null space of T is the zero subspace of V.

Solution. The dimension of the image space T(V) is called rank of T. The dimension of

the vector space kernel of T = {v | T(v) = 0} is called the nullity of T.

Now v ∈ null space of T ⇒ T(v) = 0 ⇒ T2 (v) = 0 ⇒ v ∈ null space of T2 . Thus null

space of T ⊆ null space of T2 . But we are given that rank T = rank T2 , so therefore nullity

of T = nullity of T2 , because of the nullity theorem — rank T + nullity T = dim V. Thus

null space of T = null space of T2 .

Finally if v ∈ range of T, and v ∈ null space of T, then v = T(w) for some w ∈ V. Now

⇒ w ∈ null space of T

⇒ 0 = T(w) = v

1

2

Question 1(c) If the matrix of a linear operator T on R relative to the standard basis

1 1

{(1, 0), (0, 1)} is 1 1 , find the matrix of T relative to the basis B = {(1, 1), (−1, 1)}.

Solution. Let v1 = (1, 1), v2 = (−1, 1). Then T(v1 ) = (11) 11 11 = (2, 2) = 2v1 .

T(v2 ) = (−11) 11 11 = (0, 0) = 0. So (T(v1 ), T(v2 ) = (v1 ) v2 ) 20 00 , so the matrix of T

relative to the basis B is 20 00 .

A−1

A 0 0

Question 2(a) Prove that the inverse of is where A, C are

B C −C−1 BA−1 C−1

1 0 0 0

1 1 0 0

nonsingular matrices. Hence find the inverse of

1 1 1 0.

1 1 1 1

Solution.

A−1

A 0 0 I 0

−1 = = Identity matrix.

B C −C−1 BA C−1 BA−1 − BA−1 I

A−1

0 A 0 I 0

−1 −1 −1 = −1 −1 = Identity matrix, which shows

−C BA C B C −C B + C B I

the result.

B = 11 11 . Then A−1 = C−1 = −1 and C−1 BA−1 =

1 0

1 0

Let A = C = 1 1 and

1

1 0 1 1 1 0 1 0 0 1 = 0 1 Thus

−1 1 1 1 −1 1 = −1 1 0 1 0 0

−1

1 0 0 0 1 0 0 0

1 1 0 0

−1 1 0 0

1 =

0 −1 1

1 1 0 0

1 1 1 1 0 0 −1 1

Question 2(b) If A is an orthogonal matrix with the property that −1 is not an eigenvalue,

then show that A = (I − S)(I + S)−1 for some skew symmetric matrix S.

or AS + S = I − A or (I + A)S = I − A. Let S = (I + A)−1 (I − A), note that I + A is

invertible because if |I + A| = 0, then −1 will be an eigenvalue of A.

Note that the two factors of S commute, because (I+A)(I−A) = I−A2 = (I−A)(I+A),

so (I − A)(I + A)−1 = (I + A)−1 (I − A).

2

Now

S0 = (I − A)0 ((I + A)−1 )0

= (I − A0 )(I + A0 )−1

= (AA0 − A0 )(A0 A + A0 )−1

−1

= (A − I)A0 A0 (A + I)−1

= −(I − A)(I + A)−1

= −(I + A)−1 (I − A)

= −S

Thus S is skew symmetric, so A = (I − S)(I + S)−1 where S = (I + A)−1 (I − A)

Question 2(c) Show that any two eigenvectors corresponding to distinct eigenvalues of (i)

Hermitian matrices (ii) unitary matrices are orthogonal.

Solution. We first prove that the eigenvalues of a Hermitian matrix, and therefore of a

symmetric matrix, are real.

Let H be Hermitian, and λ be one of its eigenvalues. Let x 6= 0 be an eigenvector

0 0

corresponding to λ. Thus Hx = λx, so x0 Hx = x0 λx. But (x0 Hx) = (x0 Hx)0 = x0 H x =

0

x0 Hx, because H = H. Note that (x0 Hx)0 = x0 Hx, since it is a single element, therefore

0

x0 Hx is real. Similarly x0 x 6= 0 is real, so λ = xxHx 0 x is real.

0 0

x01 Hx2 = λ2 x01 x2 . But (x02 Hx1 )0 = x01 H x2 = λ1 x01 x2 . So λ2 x01 x2 = x01 Hx2 = x01 H x2 =

0

λ1 x01 x2 because H = H. Since λ1 6= λ2 , x01 x2 = 0, so x1 , x2 are orthogonal.

Let U be unitary, Ux1 = λ1 x1 , Ux2 = λ2 x2 , where λ1 , λ2 are distinct eigenvalues of

0 0

U with corresponding eigenvectors x1 , x2 . Thus x02 U Ux1 = λ2 x02 λ1 x1 . Since U U = I,

λ2 x02 λ1 x1 = x02 x1 , so (1 − λ2 λ1 )(x02 x1 ) = 0. But 1 − λ2 λ1 = λ2 λ2 − λ2 λ1 = λ2 (λ2 − λ1 ) 6= 01 .

Thus x02 x1 = 0, so x1 , x2 are orthogonal.

Question 3(a) A matrix B of order n is of the form λA, where λ is a scalar and A has 1

everywhere except the diagonal, which has µ. Find λ, µ so that B may be orthogonal.

µ 1 ... 1

1 µ . . . 1

Solution. A = . . .

. B = λA. Thus

...

1 1 ... µ

λµ λ . . . λ λµ λ . . . λ

λ λµ . . . λ λ λµ . . . λ

B0 B = . . .

= BB0 = B2

... . . . ...

λ λ . . . λµ λ λ . . . λµ

1

We used here the fact that all eigenvalues of a unitary matrix have modulus 1. If Ux = λx, then

0 0

x U = λx0 . Thus x0 U Ux = λλx0 x, so x0 x = λλx0 x. Now x0 x 6= 0, so λλ = 1.

0

3

Clearly each diagonal element of BB0 is λ2 µ2 + (n − 1)λ2 , and each nondiagonal element is

2λ2 µ + (n − 2)λ2 . Thus B will be orthogonal if 2λ2 µ + (n − 2)λ2 = 0, λ2 µ2 + (n − 1)λ2 = 1.

Since λ 6= 0, µ = 2−n

2

= 1 − n2 , and λ2 = (1− n )12 +n−1 = 1

n2

= n42 , thus λ = ± n2 .

2 1−n+ 4

+n−1

1 −1 3 6

A = 1 3 −3 −4

5 3 3 11

Solution.

1 0 0 0

1 −1 3 6 1 0 0 0 1 0 0

A = 1 3 −3 −4 = 0 1 0 A

0

0 1 0

5 3 3 11 0 0 1

0 0 0 1

Operation C2 + C1 , C3 − 3C1 , C4 − 6C1 ⇒

1 1 −3 −6

1 0 0 0 1 0 0 0

1 4 −6 −10 = 0 1 0 A 1 0 0

0 0 1 0

5 8 −12 −19 0 0 1

0 0 0 1

Operation R2 − R1 ⇒

1 1 −3 −6

1 0 0 0 1 0 0 0

0 4 −6 −10 = −1 1 0 A 1 0 0

0 0 1 0

5 8 −12 −19 0 0 1

0 0 0 1

Operation R3 − 2R2 ⇒

1

1 −3 −6

1 0 0 0 1 0 0 0

0 4 −6 −10 = −1 1 0 A 1 0 0

0 0 1 0

5 0 0 1 2 −2 1

0 0 0 1

1

1 −6 −3

1 0 0 0 1 0 0 0

0 4 −10 −6 = −1 1 0 A 1 0 0

0 0 0 1

5 0 1 0 2 −2 1

0 0 1 0

4

R3 − 5R1 ⇒

1 1 −6 −3

1 0 0 0 1 0 0 0

0 4 −10 −6 = −1 1 0 A 1 0 0

0 0 0 1

0 0 1 0 −3 −2 1

0 0 1 0

Operation 41 R2 ⇒

1 1 −6 −3

1 0 0 0 1 0 0

0 1 − 5

0 1 0 0

2

− 32 = − 14 41 0 A

0

0 0 1

0 0 1 0 −3 −2 1

0 0 1 0

Operation C3 + 52 C2 , C4 + 32 C2 ⇒

1 − 27 − 23

1

1 0 0 0 1 0 0

0 1 0 0 = − 1 1 0 A

0 1 52 3

2

4 4 0 0 0 1

0 0 1 0 −3 −2 1

0 0 1 0

1 0 0 0 1 0 0

Thus the normal form of A is 0 1 0 0 so rank A = 3. P = − 14 14 0 and

0 0 1 0 −3 −2 1

7 3

1 1 −2 −2

0 1 5 3

Q= 2 2 and PAQ is the normal form.

0 0 0 1

0 0 1 0

Solution. Completing the squares of the given form (say Q(x1 , x2 , x3 )):

1 3

Q(x1 , x2 , x3 ) = 2(x1 + x2 − x3 )2 + x22 + x23 − 2x2 x3

2 2

1 1

= 2(x1 + x2 − x3 )2 + (x3 − x2 )2 + x22

2 2

Thus Q can be written as the sum of 3 squares with positive coefficients, so it is positive

definite.

5

UPSC Civil Services Main 1994 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Show that f1 (t) = 1, f2 (t) = t − 2, f3 (t) = (t − 2)2 forms a basis of P3 =

{Space of polynomials of degree ≤ 2}. Express 3t2 −5t+4 as a linear combination of f1 , f2 , f3 .

coefficient of t is α2 so it must be 0, hence α1 = 0. Thus f1 , f2 , f3 are linearly independent.

Since {1, t, t2 } is a basis for P3 , its dimension is 3, hence f1 , f2 , f3 is a basis of P3 .

Now by Taylor’s expansion p(t) = 3t2 − 5t + 4 = p(2) + p0 (2)(t − 2) + p”(2) 2!

(t − 2)2 =

6 + 7(t − 2) + 3(t − 2)2 = 6f1 + 7f2 + 3f3 .

T(a, b, c, d) = (a − b + c + d, a + 2c − d, a + b + 3c − 3d), a, b, c, d ∈ R

Solution. Let

T(1, 0, 0, 0) = (1, 1, 1) = v1

T(0, 1, 0, 0) = (−1, 0, 1) = v2

T(0, 0, 1, 0) = (1, 2, 3) = v3

T(0, 0, 0, 1) = (1, −1, −3) = v4

will form a basis of T(R4 ). v1 , v2 are linearly independent because if αv1 + βv2 = 0, then

α − β = 0, α = 0 so α = β = 0.

v3 is dependent on v1 , v2 , because if v3 = αv1 + βv2 , then α − β = 1, α = 2, α + β =

3 ⇒ α = 2, β = 1 ∴ v3 = 2v1 + v2 .

1

v4 is dependent on v1 , v2 , because if v4 = αv1 + βv2 , then α − β = 1, α = −1, α + β =

−3 ⇒ α = −1, β = −2 ∴ v4 = −v1 − 2v2 .

Thus v1 , v2 is a basis of T(R4 ), so rank T = 2.

Now (a, b, c, d) ∈ ker T ⇔ a − b + c + d = 0, a + 2c − d = 0, a + b + 3c − 3d = 0

Choosing particular values of a, b, c, d, we see that (1, 2, 0, 1), (−1, 1, 1, 1) ∈ ker T and are

linearly independent, so dim ker T ≥ 2. But (1, 2, 0, 1), (−1, 1, 1, 1) generate ker T, because

if (a, b, c, d) ∈ ker T, and (a, b, c, d) = α(1, 2, 0, 1) + β(−1, 1, 1, 1), then α − β = a, 2α + β =

b, β = c, α + β = d, so α = a + c, β = c and these satisfy the remaining equations 2α + β =

b, α + β = d, because (a, b, c, d) ∈ ker T and therefore a − b + c + d = 0, a + 2c − d = 0. Thus

(a, b, c, d) = (a + c)(1, 2, 0, 1) + c(−1, 1, 1, 1), so dim ker T = nullity T = 2

Hence rank T + nullity T = 4 = dim(R4 ), as required.

Question 1(c) If T is an operator on R3 whose basis is B = {(1, 0, 0), (0, 1, 0), (0, 0, 1)}

such that

0 1 1

[T : B] = 1 0 −1

−1 −1 0

find a matrix of T w.r.t. a basis B1 = {(0, 1, −1), (1, −1, 1), (−1, 1, 0)}.

Solution. The basis B is the standard basis, hence the representation of B1 in this basis is

as given. (Note that if B were some other basis, we would write B1 in that basis, and then

continue as below.) Let v1 = (0, 1, −1), v2 = (1, −1, 1), v3 = (−1, 1, 0). Then

T(v2 ) = (0, 0, 0) = 0

T(v3 ) = (1, −1, 0) = −v3

Thus

1 0 0

[T : B1 ] = 0 0 0

0 0 −1

Note: The main idea behind the above solution is to express T(vi ) = ni=0 αii vi . Now

P

we solve for αii to get the matrix for T in the new basis.

An alternative is to compute P−1 [T : B]P, where P is given by [v1 , . . . , vn ] = [e1 , . . . , en ]P

or P = [v1 0 , . . . , vn 0 ]. Show that this is true.

Question 2(a) If A = haij i is an n × n matrix such that aii = n, aij = r if i 6= j, show that

[A − (n − r)I][A − (n − r + nr)I] = 0

Hence find the inverse of the n × n matrix B = hbij i where bii = 1, bij = ρ, i 6= j and

1

ρ 6= 1, ρ 6= 1−n .

2

Solution. Let C = A − (n − r)I, then every entry of C is r. Let D = A − (n − r + nr)I =

C − nrI. Thus CD = C2 − nrC. Each entry of C2 is nr2 , which is the same as each entry

of nrC, so CD = 0 as required.

The given equation implies

Let A = nB, where r = ρn. Thus A satisfies the conditions for the equation to hold, so

substituting A and r in the above equation

B[B − (2 − 2ρ + nρ)I] = −(1 − ρ)(1 − ρ − nρ)I

1

B−1 = − [B − (2 − 2ρ + nρ)I]

(1 − ρ)(1 − ρ − nρ)

1−2ρ+nρ

Thus the diagonal elements of B−1 are all (1−ρ)(1−ρ−nρ)

, while the off-diagonal elements are

2−3ρ+nρ

all (1−ρ)(1−ρ−nρ) .

Question 2(b) Prove that the eigenvectors corresponding to distinct eigenvalues of a square

matrix are linearly independent.

of a matrix A.Q Let a1 x1 + . . . + ar xr = 0, ai ∈ R. We shall show that ai = 0, 1 ≤ Q i ≤ r.

Let L1 = ri=2 (A − λi I). Note that the factors of L1 commute. Thus L1 x2 = ri=3 (A −

λi I)(A − λ2 I)x2 = 0 because Ax2 = λ2 x2 . Similarly L1 x3 = . . . = L1 xr = 0. Moreover

L1 x1 = (λ1 − λ2 ) . . . (λ1 − λr )x1 .

Consequently

0 = L1 (a1 x1 + . . . + ar xr )

= a1 L 1 x 1

= a1 (λ1 − λ2 ) . . . (λ1 − λr )x1

λ1 − λi 6= 0, 2 ≤ i ≤ r, and Q

x1 6= 0 so a1 = 0.

Similarly taking Li = rj=1 (A − λj I), we show that ai = 0 for 1 ≤ i ≤ r. Thus

i6=j

x1 , x2 , . . . , xr are linearly independent.

3 1 4

A = 0 2 6

0 0 5

3

Solution. The characteristic polynomial of A is |λI − A| = (λ − 3)(λ − 2)(λ − 5).1 Thus

the eigenvalues of A are 3, 2, 5.

If x = (x1 , x2 , x3 ) is an eigenvector corresponding to λ = 3 then

0 1 4 x1

(A − 3I)x = 0 −1 6

x2 = 0

0 0 2 x3

eigenvector for λ = 3. All the eigenvectors are (x1 , 0, 0), x1 6= 0.

If x = (x1 , x2 , x3 ) is an eigenvector corresponding to λ = 2 then

1 1 4 x1

(A − 2I)x = 0 0 6

x2 = 0

0 0 3 x3

Thus x1 + x2 + 4x3 = 0, 6x3 = 0, 3x3 = 0, take x1 = 1 to get (1, −1, 0) as an eigenvector for

λ = 2. All the eigenvectors are (x1 , −x1 , 0), x1 6= 0.

If x = (x1 , x2 , x3 ) is an eigenvector corresponding to λ = 5 then

−2 1 4 x1

(A − 5I)x = 0 −3 6 x2 = 0

0 0 0 x3

Thus −2x1 + x2 + 4x3 = 0, −3x2 + 6x3 = 0, take x3 = 1 to get (3, 2, 1) as an eigenvector for

λ = 5. All eigenvectors are (3x3 , 2x3 , x3 ), x3 6= 0.

Question 3(a) Show that the matrix congruent to a skew symmetric matrix is skew sym-

metric. Use the result to prove that the determinant of a skew symmetric matrix of even

order is the square of a rational function of its elements.

−P0 AP = −B, so B is skew symmetric.

We prove the second result by induction on m, where n = 2m is the order of the skew

0 a 2

symmetric matrix under consideration. If m = 1 then A = −a 0 , |A| = a , so the result is

true for m = 1.

Assume by induction that the result is true for all skew symmetric matrices of even order

< 2m. If A ≡ 0, there is nothing to prove. Otherwise there exists at least one non-zero

element aij . Changing the first row with row j, we move aij in the first row. Changing

1

Note that the determinant of an upper diagonal or a lower diagonal matrix is just the product of the

elements on the main diagonal.

4

column 1 and column j, we get −aij in the symmetric position. Now by multiplying the new

matrix by suitable elementary matrices on the left and right, we get

0 aij ∗ ∗ ... ∗

−aij 0 ∗ ∗ ... ∗

0

P AP = ∗ ∗

... A2m−2

∗ ∗

Now we can find P∗ a product of elementary matrices such that

0 aij 0 0 ... 0

−aij 0 0 0 ... 0

0

P∗ P0 APP∗ =

0 0

... A2m−2

0 0

Thus det A = determinant of a skew symmetric matrix of order 2 × determinant of a

skew symmetric matrix of order 2(m − 1). The induction hypothesis now gives the result.

c −b a0

0

−c 0 a b0

A= b −a 0 c0

0 c

Solution. The rank of a skew symmetric matrix is always even2 . Since −c 2

0 = c >

0, rank A ≥ 2.

If c0 6= 0, then

0

0

0 c −b a

0

cc 0 ac0 b0 c0

−c |A| =

0

b 0 −a0 0 0 c

−a −b −c 0

Adding b0 R3 − aR4 to R2 , all the entries of the second row become 0, so −c0 |A| = 0 ⇒

|A| = 0 ⇒ rank A < 4 ⇒ rank A = 2.

2

We can prove this by induction on the order of the skew symmetric matrix S. It is true if S is a 1 × 1

be the 0-matrix, thus has rank 0. Now given an (n + 1) × (n + 1) matrix S, we can

matrix, since it must

C b

write it as −b 0

0 , where C is skew symmetric, and hence has even rank by the induction hypothesis. If b

is linearly dependent on

the columns of C, then by a series of elementary operations on S, we can transform

it into P0 SP = 0C0 00 , so rank S = rank P0 SP = rank C. If b is linearly independent of the columns of

C b 0

C, then b’ is linearly independent of the rows of C, so rank S = rank −b 0

0 = rank[C b] + 1(∵ [−b 0] is

independent of the rows of [C b]) = rank C + 2, which is also even.

5

If c0 = 0,

0

c −b a0 a

c 0 a b0 a

a|A| =

b0 −a 0 0

−a a −b0 a 0 0

Adding −b0 R3 to R4 , we see that the fourth row has all 0’s, hence rank A = 2 as before.

Alternate solution:

c −b a0

0

−c 0 a b0

−a|A| = b −a 0 c0

Add −c0 R2 and b0 R3 to R4 , then all entries of the last row become 0. So rank A < 4, and

by the reasoning above, rank A ≥ 2, rank A 6= 3 so rank A = 2.

Question 3(c) Reduce the following symmetric matrix to a diagonal form and interpret the

results in terms of quadratic forms.

3 2 −1

A= 2 2 3

−1 3 1

Solution.

x

(x y z)A y

z

= 3x2 + 2y 2 + z 2 + 4xy − 2xz + 6yz

2 1 2 2 22

= 3(x + y − z)2 + y 2 + z 2 + yz

3 3 3 3 3

2 1 2 2 11 2 117 2

= 3(x + y − z) + (y + z) − z

3 3 3 2 6

2 117 2

= 3X 2 + Y 2 − Z

3 6

where X = x − 23 y − 13 z, Y = y + 112

z, Z = z. This implies z = Z, y = Y − 11 2

Z, x =

2 11 1 2

X + 3 (Y − 2 Z) − 3Z = X + 3 Y − 4Z.

1 32 −4

3 0 0

Then if P = 0 1 − 11 2

, we have P0 AP = 0 2

3

0 .

0 0 1 0 0 − 117 6

The quadratic form associated with A is indefinite as it takes both positive and negative

values. Note that x0 Ax and x0 P0 APx0 take the same values.

6

UPSC Civil Services Main 1995 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

1 Linear Algebra

Question 1(a) Let T(x1 , x2 , x3 ) = (3x1 + x3 , −2x1 + x2 , −x1 + 2x2 + 4x3 ) be a linear trans-

formation on R3 . What is the matrix of T w.r.t. the standard basis? What is a basis of the

range space of T? What is a basis of the null space of T?

Solution.

T(e2 ) = T(0, 1, 0) = (0, 1, 2) = e2 + 2e3

T(e3 ) = T(0, 0, 1) = (1, 0, 4) = e1 + 4e3

3 0 1

T ⇐⇒ A = −2 1 0

−1 2 4

Clearly T(e2 ), T(e3 ) are linearly independent. If (3, −2, −1) = α(0, 1, 2) + β(1, 0, 4), then

β = 3, α = −2, but 2α + 4β 6= −1, so T(e1 ), T(e2 ), T(e3 ) are linearly independent. Thus

(3, −2, −1), (0, 1, 2), (1, 0, 4) is a basis of the range space of T.

Note that T(x1 , x2 , x3 ) = 0 ⇔ x1 = x2 = x3 = 0, so the null space of T is {0}, and the

empty set is a basis. Note that the matrix of T is nonsingular, so T(e1 ), T(e2 ), T(e3 ) are

linearly independent.

Question 1(b) Let A be a square matrix of order n. Prove that Ax = b has a solution

⇔ b ∈ Rn is orthogonal to all solutions y of the system A0 y = 0.

1

Solution. If x is a solution of Ax = b and y is a solution of A0 y = 0, then b0 y = x0 A0 y = 0,

thus b is orthogonal to y.

Conversely, suppose b0 y = 0 for all y ∈ Rn which is a solution of A0 y = 0. Let

W = A(Rn ) = the range space of A, and W ⊥ its orthogonal complement. If A0 y = 0 then

x0 A0 y = 0 ⇒ (Ax)0 y = 0 for every x ∈ Rn ⇒ y ∈ W ⊥ . Conversely y ∈ W ⊥ ⇒ ∀x ∈

Rn .(Ax)0 y = 0 ⇒ x0 A0 y = 0 ⇒ A0 y = 0. Thus W ⊥ = {y | A0 y = 0}. Now b0 y = 0 for all

y ∈ W ⊥ , so b ∈ W ⇒ b = Ax for some x ∈ Rn ⇒ Ax = b is solvable.

Question 1(c) Define a similar matrix and prove that two similar matrices have the same

characteristic equation. Write down a matrix having 1, 2, 3 as eigenvalues. Is such a matrix

unique?

Solution. Two matrices A, B are said to be similar if there exists a matrix P such that

B = P−1 AP. If A, B are similar, say B = P−1 AP, then characteristic polynomial of B is

|λI − B| = |λI − P−1 AP| = |P−1 λIP − P−1 AP| = |P−1 ||λI − A||P| = |λI − A|. (Note that

|X||Y| = |XY|.) Thus the characteristic polynomial of B is the same as that of A.

1 0 0

Clearly the matrix A = 0 2 0 has eigenvalues 1,2,3. Such a matrix is not unique, for

0 0 3

1 1 0

example B = 0 2 0 has the same eigenvalues, but B 6= A.

0 0 3

5 −6 −6

A = −1 4 2

3 −6 −4

is diagonalizable and hence determine A5 .

Solution.

|A − λI| = 0

5 − λ −6 −6

⇒ −1 4 − λ 2 = 0

3 −6 −4 − λ

⇒ (5 − λ)[(4 − λ)(−4 − λ) + 12] + 6[4 + λ − 6] − 6[6 − 3(4 − λ)] = 0

⇒ (5 − λ)[λ2 − 4] + 6[λ − 2 − 3λ + 6] = 0

⇒ −λ3 + 5λ2 + 4λ − 20 − 12λ + 24 = 0

⇒ λ3 − 5λ2 + 8λ − 4 = 0

Thus λ = 1, 2, 2.

If (x1 , x2 , x3 ) is an eigenvector for λ = 1, then

4 −6 −6 x1

−1 3 2 x2 = 0

3 −6 −5 x3

⇒ 4x1 − 6x2 − 6x3 = 0

−x1 + 3x2 + 2x3 = 0

3x1 − 6x2 − 5x3 = 0

2

Thus x1 = x3 , x3 = −3x2 , so (−3, 1, −3) is an eigenvector for λ = 1.

If (x1 , x2 , x3 ) is an eigenvector for λ = 2, then

3 −6 −6 x1

−1 2 2 x2 = 0

3 −6 −6 x3

⇒ 3x1 − 6x2 − 6x3 = 0

−x1 + 2x2 + 2x3 = 0

3x1 − 6x2 − 6x3 = 0

Taking x1 = 4, x2 = 1, (4, 1, 1) is another eigenvector for λ = 2, and these two are linearly

independent.

−3 0 4 2 −4 −4

Let P = 1 1 1. A simple calculation shows that P−1 = 12 −4 9 7 .

−3 −11 2 −3 −3

1 0 0

−1

Clearly P AP = 0 2 0.

0 0 2

1 0 0

Now P−1 A5 P = (P−1 AP)5 = 0 32 0 .

0 0 32

1 0 0

A5 = P 0 32 0 P−1

0 0 32

−3 0 4 1 0 0 2 −4 −4

1

= 1 1 1 0 32 0 −4 9 7

2

−3 −1 1 0 0 32 2 −3 −3

−3 0 128 2 −4 −4

1

= 1 32 32 −4 9 7

2

−3 −32 32 2 −3 −3

125 −186 −186

= −31 94 62

93 −186 −154

Note: Another way of computing A5 is given below. This uses the characteristic poly-

nomial of A : A3 = 5A2 − 8A + 4I and not the diagonal form, so it will not be permissible

here.

3

A5 = A2 (5A2 − 8A + 4I)

= 5A(5A2 − 8A + 4I) − 8(5A2 − 8A + 4I) + 4A2

= 25(5A2 − 8A + 4I) − 76A2 + 84A − 32I

= 49A2 − 116A + 68I

is also invertible and

(I − BA)−1 = I + B(I − AB)−1 A

Show that AB and BA have the same characteristic values.

Solution.

(I + B(I − AB)−1 A)(I − BA)

= I − BA + B(I − AB)−1 A − B(I − AB)−1 ABA

= [I + B(I − AB)−1 A] − B[I + (I − AB)−1 AB]A (1)

−1 −1 −1

Now (I − AB) (I − AB) = (I − AB) − (I − AB) AB = I

∴ (I − AB)−1 = I + (I − AB)−1 AB

Substituting in (1) (I + B(I − AB)−1 A)(I − BA)

= I + B(I − AB)−1 A − B(I − AB)−1 A = I

We shall show that λI − AB is invertible if and only if λI − BA is invertible. This means

that if λ is an eigenvalue of AB, then |λI − AB| = 0 ⇒ |λI − BA| = 0 so λ is an eigenvalue

of BA.

If λI − AB is invertible, then

= λI − BA + λB(λI − AB)−1 A − B(λI − AB)−1 ABA

= λ[I + B(λI − AB)−1 A] − B[I + (λI − AB)−1 AB]A (2)

Now (λI − AB)−1 (λI − AB) = λ(λI − AB)−1 − (λI − AB)−1 AB = I

∴ λ(λI − AB)−1 = I + (λI − AB)−1 AB

Substituting in (2) (I + B(λI − AB)−1 A)(λI − BA)

= λI + λB(λI − AB)−1 A − λB(λI − AB)−1 A = λI

is symmetric, thus AB and BA have the same eigenvalues.

We give another simple proof of the fact that AB and BA have the same eigenvalues.

1. Let 0 be an eigenvalue of AB. This means that AB is singular, i.e. 0 = |AB| =

|A||B| = |BA|, so BA is singular, hence 0 is an eigenvalue of BA.

4

2. Let λ 6= 0 be an eigenvalue of AB and let x 6= 0 be an eigenvector corresponding to

λ, i.e. ABx = λx. Let y = Bx. Then y 6= 0, because Ay = ABx = λx 6= 0 as λ 6= 0.

Now BAy = BABx = B(ABx) = λBx = λy. Thus λ is an eigenvalue of BA.

Question 2(c) Let a, b ∈ C, |b| = 1 and let H be a Hermitian matrix. Show that the

eigenvalues of aI + bH lie on a straight line in the complex plane.

a + tb is an eigenvalue of aI + bH. Conversely, if λ is an eigenvalue of aI + bH, then λ−ab

(note b 6= 0 as |b| = 1) is an eigenvalue of H.

Clearly a + tb lies on the straight line joining points a and a + b:

For the sake of completeness, we prove that the eigenvalues of a Hermitian matrix H are

real. Let z 6= 0 be an eigenvector corresponding to the eigenvalue t.

Hz = tz

⇒ z0 Hz = tz0 z

0

⇒ z0 Hz = tz0 z

0

But z0 Hz = z0 H z = z0 Hz = tz0 z

⇒ tz0 z = tz0 z

⇒t = t ∵ z0 z 6= 0

Question 3(a) Let A be a symmetric matrix. Show that A is positive definite if and only

if its eigenvalues are all positive.

eigenvalues of A, not necessarily distinct. Let x1 be an eigenvector corresponding to λ1 .

Since λ1 and A are real,

p x1 is also real. Replacing x1 if necessary by µx1 , µ suitable, we can

assume that ||x1 || = x01 x1 = 1.

Let P1 be an orthogonal matrix with x1 as its first column. Such a P1 exists, as will be

shown at the end of this result. Clearly the first column of the matrix P1 −1 AP1 is equal

λ

to P1 −1 Ax = λ1 P1 −1 x = 00 , because P1 −1 x is the first column of P1 −1 P = I. Thus

0

P1 −1 AP1 = λ01 B = P01 AP1 where B is (n − 1) × (n − 1) symmetric. Since P01 AP1 is

L

symmetric, it follows that P1 −1 AP1 = P01 AP1 = λ01 B0 . Induction now gives that there

λ2 0 . . . 0

exists an (n − 1) × (n − 1) orthogonal matrix Q such that Q0 BQ = . . .

0 0 . . . λn

5

where λ2 , λ3 , . . . , λn are eigenvalues of B. Let P2 = 01 Q 0

, then P2 is orthogonal and

P02 P01 AP1 P2 = diagonal[λ 1 , . . . , λ n ]. Set P = P P

1 2 . . . P 0

n and (y1 , . . . , yn )P = x then

,

n

x0 Ax = y0 P0 APy = i=0 λ2i yi2 .

P

0

Pn 2 2

Since P is non-singular, quadratic forms x Ax and i=0 λi yi assume the same values.

Hence A is positive definite if and only if ni=0 λ2i yi2 is positive definite if and only if λi > 0

P

for all i. p

Result used: If x1 is a real vector such that ||x1 || = x01 x1 = 1 then there exists an

orthogonal matrix with x1 as its first column.

Proof: We have to find real column vectors x2 , . . . , xn such that ||xi || = 1, 2 ≤ i ≤ n

and x2 , . . . , xn is an orthonormal system i.e. x0i xj = 0, i 6= j. Consider the single equation

x01 x = 0, where x is a column vector to be determined. This equation has a non-zero solution,

in fact the space of solutions is of dimension n − 1, the rank of the coefficient matrix being

1. If y2 is a solution, we take x2 = ||yy22 || so that x01 x2 = 0.

We now consider the two equations x01 x = 0, x02 x = 0. Again the number of unknowns

is more than the number of equations, so there is a solution, say y3 , and take x3 = ||yy33 || to

get x1 , x2 , x3 mutually orthogonal.

Proceeding in this manner, if we consider n − 1 equations x01 x = 0, . . . , x0n−1 x = 0, these

will have a nonzero solution yn , so we set xn = ||yynn || . Clearly x1 , x2 , . . . , xn is an orthonormal

system, and therefore P = [x1 , . . . , xn ] is an orthogonal matrix having x1 as a first column.

Question 3(b) Let A and B be square matrices of order n, show that AB − BA can never

be equal to the identity matrix.

Xn Xn n X

X n

= aik bki = bki aik = tr BA

i=1 k=1 k=1 i=1

Thus tr(AB − BA) = tr AB − tr BA = 0. But the trace of the identity matrix is n, thus

AB − BA can never be equal to the identity matrix.

n

X

Question 3(c) Let A = haij i, 1 ≤ i, j ≤ n. If |aij | < |aii |, then the eigenvalues of A lie

j=1

i6=j

in the disc n

X

|λ − aii | ≤ |aij |

j=1

i6=j

6

Solution. See the solution to question 2(c), year 1997. We showed that if |λ − aii | >

n

X

|aij | then |λI − A| =

6 0, so λ is not an eigenvalue of A. Thus if λ is an eigenvalue, then

j=1

i6=j

n

X

|λ − aii | ≤ |aij |, so λ lies in the disc described in the question.

j=1

i6=j

7

UPSC Civil Services Main 1996 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) In R4 let W1 be the space generated by {(1, 1, 0, −1), (2, 4, 6, 0)} and let W2

be space generated by {(−1, −2, −2, 2), (4, 6, 4, −6), (1, 3, 4, −3)}. Find a basis for the space

W1 + W2 .

Solution. Let v1 = (1, 1, 0, −1), v2 = (2, 4, 6, 0), v3 = (−1, −2, −2, 2), v4 = (4, 6, 4, −6), v5 =

(1, 3, 4, −3). Since w ∈ W1 + W2 can be written as w = w1 + w2 , and w1 = α1 v1 + α2 v2

and w2 = α3 v3 + α4 v4 + α5 v5 , it follows that w is a linear combination of vi ⇒ W1 + W2

is generated by {vi , 1 ≤ i ≤ 5}. Thus a maximal independent subset of {vi , 1 ≤ i ≤ 5} will

be a basis of W1 + W2 .

Clearly v1 and v2 are linearly independent. If possible, let v3 = λ1 v1 + λ2 v2 , then the

four equations

λ1 + 2λ2 = −1

λ1 + 4λ2 = −2

0λ1 + 6λ2 = −2

−λ1 + 0λ2 = 2

should be consistent and provide us λ1 , λ2 . Clearly the third and fourth equations give us

λ1 = −2, λ2 = − 31 which do not satisfy the first two equations. Thus v1 , v2 , v3 are linearly

independent.

If possible let v4 = λ1 v1 + λ2 v2 + λ3 v3 . Then

λ1 + 2λ2 − λ3 =4 (1a)

λ1 + 4λ2 − 2λ3 =6 (1b)

0λ1 + 6λ2 − 2λ3 =4 (1c)

−λ1 + 0λ2 + 2λ3 = −6 (1d)

1

Adding (1b) and (1d) we get 4λ2 = 0, so λ2 = 0. Solving (1a) and (1b) we get λ3 = −2, λ1 =

2. These values satisfy all the four equations, so v4 = 2v1 − 2v3 .

If possible let v5 = λ1 v1 + λ2 v2 + λ3 v3 . Then

λ1 + 2λ2 − λ3 =1 (2a)

λ1 + 4λ2 − 2λ3 =3 (2b)

0λ1 + 6λ2 − 2λ3 =4 (2c)

−λ1 + 0λ2 + 2λ3 = −3 (2d)

Adding (2b) and (2d) we get 4λ2 = 0, so λ2 = 0. (2c) then gives us λ3 = −2, and

(2a) now gives λ1 = −1, which satisfies all equations. Thus v5 = −v1 − 2v3 . Hence

{(1, 1, 0, −1), (2, 4, 6, 0), (−1, −2, −2, 2)} is a basis of W1 + W2 .

Question 1(b) Let V be a finite dimensional vector space and v ∈ V, v 6= 0. Show that

there exists a linear functional f on V such that f (v) 6= 0.

n

X n

X

f (vj ) = δ1j and f ( aj v j ) = aj f (vj ).

j=1 j=1

functional over V, and f (v) = f (v1 ) = 1. Note that f (vj ) = 0, j > 1

Clearly f is a linear P

and if any w ∈ V, w = i ai vi , f (w) = a1 .

T(vi ) = v1 + v2 + v3 , 1 ≤ i ≤ 3. By writing the matrix of T w.r.t. another basis show that

the matrices

1 1 1 3 0 0

A = 1 1 1 and B = 0 0 0

1 1 1 0 0 0

are similar.

Let

w1 = v1 + v 2 + v 3

w2 = v1 − v2

w3 = v2 − v3

⇒ T(w1 ) = 3w1 , T(w2 ) = T(w3 ) = 0

We now show that w1 , w2 , w3 is a basis for V, i.e. these are linearly independent.

2

Let αw1 + βw2 + γw3 = 0, then (α + β)v1 + (α − β + γ)v2 + (α − γ)v3 = 0. But

v1 , v2 , v3 are linearly independent, therefore α + β = 0, α − β + γ = 0, α − γ = 0 ⇒ α =

β = γ = 0 ⇒ w1 , w2 , w3 are linearly independent.

The matrix of T w.r.t. the basis w1 , w2 , w3 is clearly B. Note that the choice of

w1 , w2 , w3 is suggested by the shape of B.

6 0 then B = P−1 AP, so A and B are similar.

If (w1 , w2 , w3 ) = (v1 , v2 , v3 )P, |P| =

T(x, y, z) = (x + z, −2x + y, −x + 2y + z)

What is the matrix of T w.r.t. the basis (1, 0, 1), (−1, 1, 1), (0, 1, 1)? Using this matrix write

down the matrix of T with respect to the basis (0, 1, 2), (−1, 1, 1), (0, 1, 1).

Solution. Let v1 = (1, 0, 1), v2 = (−1, 1, 1), v3 = (0, 1, 1). T(x, y, z) = (x+z, −2x+y, −x+

2y+z) = αv1 +βv2 +γv3 , say. This means α−β = x+z, β+γ = −2x+y, α+β+γ = −x+2y+

z. This implies α = x + y + z, β = y, γ = −2x. Thus T(x, y, z) = (x + y + z)v1 + yv2 − 2xv3 .

Hence

2 1 2

[T(v1 ) T(v2 ) T(v3 )] = [v1 v2 v3 ] 0 1 1

−2 2 0

Let w1 = (0, 1, 2), w2 = (−1, 1, 1), w3 = (0, 1, 1). Then

1 0 0

[w1 w2 w3 ] = [v1 v2 v3 ] 1 1 0

0 0 1

Hence

= [v1 v2 v3 ]AP

= [w1 w2 w3 ]P−1 AP

where

2 1 2 1 0 0

A = 0 1 1 , P = 1 1 0

−2 2 0 0 0 1

Thus the matrix of T w.r.t. basis w1 , w2 , w3 is

1 0 0 2 1 2 1 0 0 3 1 2

−1

P AP = −1 1 0 0 1 1 1 1 0 = −2 0 −1

0 0 1 −2 2 0 0 0 1 0 2 0

3

Question 2(b) Let V and W be finite dimensional vector spaces such that dim V ≥ dim W.

Show that there is always a linear map of V onto W.

Solution. Let w1 , w2 , . . . , wm be a basis of W, and v1 , v2 , . . . , vn be a basis of V, n ≥ m.

Define

T(vi ) = wi , i = 1, 2, . . . , m

T(vi ) = 0, i = m + 1, . . . , n

and for any v ∈ V, v = i=1 αi vi , T(v) = m

Pn P

i=1 αi T(vi ). Pm

Clearly

Pm T : V

Pm−→ W is linear. T is onto, since if w ∈ W, w = i=1 ai wi , then

T( i=1 ai vi ) = i=1 ai T(vi ) = w, proving the result.

x + y − 2z = 1

2x − 7z = 3

x+y−z = 5

Solution.

1 1 −2 −1 −1 −2

D = 2 0 −7 = −5 −7 −7 = −2

1 1 −1 0 0 −1

˛ ˛ ˛ ˛

˛1

˛

˛

1 −2˛˛˛ ˛1

˛

˛

1 −2˛˛˛

˛3

˛

˛

0 −7˛˛˛ ˛3

˛

˛

0 −7˛˛˛

˛5

˛

1 −1˛˛ ˛4

˛

0 1 ˛˛ −31 31

x= D

= D

= −2

= 2

˛ ˛ ˛ ˛

˛1

˛

˛

1 −2˛˛˛ ˛1

˛

˛

0 0 ˛˛˛

˛2

˛

˛

3 −7˛˛˛ ˛2

˛

˛

1 −3˛˛˛

˛1

˛

5 −1˛˛ ˛1

˛

4 1 ˛˛ 13

y= D

= D

= −2

= − 13

2

˛ ˛ ˛ ˛

˛1 1 1˛˛˛ ˛1 1 1˛˛˛

˛ ˛

˛ ˛

˛2 0 3˛˛˛ ˛2 0 3˛˛˛

˛ ˛

˛ ˛

˛1 1 5˛˛ ˛0 0 4˛˛

˛ ˛

−8

z= D

= D

= −2

=4

0 1 0 0

0 0 1 0

A= 0 0

0 1

1 0 0 0

by computing its characteristic polynomial.

4

Solution. The characteristic polynomial of A is

−λ 1 0 0

0 −λ 1 0

|A − λI| =

0 0 −λ 1

1 0 0 −λ

= −λ[−λ3 ] − 1[1] = λ4 − 1 = 0

0 1 0 0 0 1 0 0 0 0 1 0

0 0 1 0 0 0

1 0 0 0 0 1

A2 = 0 0 0 1 0 0

=

0 1 1 0 0 0

1 0 0 0 1 0 0 0 0 1 0 0

0 0 1 0 0 1 0 0 0 0 0 1

0 0 0 1 0 0 1 0 1 0 0 0

= A−1

A3 =

1

=

0 0 0 0 0 0 1 0 1 0 0

0 1 0 0 1 0 0 0 0 0 1 0

Question 3(b) If A and B are n × n matrices such that AB = BA, show that AB and

BA have a common characteristic vector.

We show that B(Vλ ) ⊆ Vλ . Let v ∈ Vλ , then A(Bv) = B(Av) = B(λv) = λBv ⇒ Bv ∈ Vλ .

Consider B∗ : Vλ −→ Vλ such that B∗ (v) = B(v) — note that B∗ is a restriction of B

to Vλ and we have already shown that B(Vλ ) ⊆ Vλ .

Let µ be an eigenvalue of B∗ , then µ is also an eigenvalue of B (because a basis of Vλ can be

∗ C

extended to a basis of V, and in this basis B = B0 D for some matrices C, D). Let v ∈ Vλ

be an eigenvector of B∗ corresponding to µ, by definition v 6= 0. Then Bv = B∗ v = µv.

Thus A and B have a common eigenvector v, note that Av = λv as v ∈ Vλ .

2

− 23 13

3

O = 32 13 − 23

1 2 2

3 3 3

Solution. Before solving this particular problem, we present a general discussion about

orthogonal matrices. An orthogonal matrix satisfies O0 O = I, so its determinant is 1 or -1,

here we focus on the case where |O| = 1. If λ is an eigenvalue of O and x a corresponding

eigenvector, then |λ|2 x0 x = (Ox)0 Ox = x0 O0 Ox = x0 x, so |λ| = 1. Since the characteristic

5

polynomial has real coefficients, the eigenvalues must be real or in complex conjugate pairs.

Thus for a matrix of order 3, at least one eigenvalue is real, and must be 1 or -1. Since

|O| = 1, one real value must be 1, and the three possibilities are {1, 1, 1}, {1, −1, −1} and

{1, eiθ , e−iθ }. √

Here we consider the third case, as the given matrix has 1 and 13 ± i 2 3 2 as eigenvalues,

proved later.

Let Z = X1 + iX2 be an eigenvector corresponding to the eigenvalue eiθ . Let X3 be

the eigenvector corresponding to the eigenvalue 1. Since Z and X3 correspond to different

eigenvalues, these are orthogonal, i.e. Z0 X3 = (X01 + iX02 )X3 = 0 ⇒ X01 X3 = 0, X02 X3 = 0.

Note that X1 , X2 , X3 are real vectors. Since OZ = eiθ Z = (cos θ + i sin θ)(X1 + iX2 ).

Equating real and imaginary parts we get

OX2 = X1 sin θ + X2 cos θ

0 0

∴ X1 O OX1 = (X01 cos θ − X02 sin θ)(X1 cos θ − X2 sin θ)

⇒ X01 X1 = X01 X1 cos2 θ − X02 X1 cos θ sin θ − X01 X2 sin θ cos θ + X02 X2 sin2 θ

⇒ 0 = X01 X1 sin2 θ − X02 X2 sin2 θ + 2X01 X2 cos θ sin θ

⇒ 0 = X01 X1 sin θ − X02 X2 sin θ + 2X01 X2 cos θ (1)

(Note that sin θ 6= 0 since we are considering the case where eiθ is complex.) Similarly

⇒ X02 X1 = X01 X1 sin θ cos θ − X01 X2 sin2 θ − X02 X2 sin θ cos θ + X02 X1 cos2 θ

⇒ 0 = X01 X1 cos θ − X01 X2 cos θ − 2X01 X2 sin θ (2)

Multiplying (1) by sin θ and (2) by cos θ and adding, we get X01 X1 − X02 X2 = 0 or X01 X1 =

X02 X2 , so from (2), X1 X2 = 0, i.e. X1 , X2 are orthogonal.

Thus X1 , X2 , X3 are mutually orthogonal. We can assume that X01 X1 = X02 X2 = 1,

replacing Z by λZ, λ ∈ R if necessary. Similarly we can take X03 X3 = 1. Let P = [X1 X2 X3 ]

so that P0 P = I. Now

cos θ sin θ 0

= [X1 X2 X3 ] − sin θ cos θ 0

0 0 1

cos θ sin θ 0

−1 0

⇒ P OP = P OP = − sin θ cos θ 0

0 0 1

which is the canonical form of O when the eigenvalues are 1, eiθ , e−iθ .

6

Solution of given problem.

2

− 23 13

3

O = 32 13 − 32

1 2 2

2 3 3 3

− λ −2 1 2 − 3λ −2 1

3 3 3 1

|O − λI| = 23 1

3

− λ 2

−3 = 2 1 − 3λ −2

1 2 2 27

3 3 3

−λ 1 2 2 − 3λ

1

= [(2 − 3λ)2 (1 − 3λ) + 4(2 − 3λ) + 1(3 + 3λ) + 2(6 − 6λ)]

27

1

= − [27λ3 − 45λ2 + 45λ − 27]

27

1

= − [(λ − 1)(3λ2 − 2λ − 3)]

3

√

Thus λ = 1, 13 ± i 2 3 2 are eigenvalues of O. √

2 2

Thus the canonical form of O is derived from above, where cos θ = 13 , sin θ = 3

:

√

1 2 2

3√ 3

0

− 2 2 1

0

3 3

0 0 1

The matrix P can be determined as follows (this is not needed for this problem, but is

given for completeness):

1 2 1

− x1 − x2 + x3 = 0

3 3 3

2 2 2

x1 − x2 − x3 = 0

3 3 3

1 2 1

x1 + x2 − x3 = 0

3 3 3

Thus x2 = 0, x1 − x3 = 0, so we can take (1, 0, 1) as an eigenvector.

OX2 = X1 sin θ + X2 cos θ

7

√

2 2

where cos θ = 13 , sin θ = 3

. This gives us the following equations

√

2x11 − 2x12 + x13 = x11 − x21 2 2 (3)

√

2x11 + x12 − 2x13 = x12 − x22 2 2 (4)

√

x11 + 2x12 + 2x13 = x13 − x23 2 2 (5)

√

2x21 − 2x22 + x23 = x11 2 2 + x21 (6)

√

2x21 + x22 − 2x23 = x12 2 2 + x22 (7)

√

x21 + 2x22 + 2x23 = x13 2 2 + x23 (8)

√

Adding the √ last 3 equations, we get 2x21 √ = x11 + x12 + x13 . Subtracting equation (6)

from (8), 2x22 = x13 − x11 , and from (7) 2x23 = x11 − x12 + x13 . Substituting these

in the first 3 equations and simplifying, we get x11 = −x13 . Setting x11 = 0, x12 = 1,

we get (0, 1, 0), ( √12 , 0, − √12 ) as a possible solution for X1 , X2 .

0 √12 1

P = 1 0 0

0 − √12 1

√

2 2

1

3√ 3

0

We can now verify that OP = P − 2 3 2 1

3

0

0 0 1

8

UPSC Civil Services Main 1997 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

1 Linear Algebra

Question 1(a) Let V be the vector space of polynomials over R. Find a basis and the

dimension of W ⊆ V spanned by

v1 = t3 − 2t2 + 4t + 1

v2 = 2t3 − 3t2 + 9t − 1

v3 = t3 + 6t − 5

v4 = 2t3 − 5t2 + 7t + 5

0, −2α − 3β = 0, 4α + 9β = 0, α − β = 0 ⇒ α = β = 0.

v3 depends linearly on v1 , v2 — if αv1 + βv2 = v3 , then α + 2β = 1, −2α − 3β =

0, 4α + 9β = 6, α − β = −5 ⇒ α = −3, β = 2 which satisfy all the equations. Thus

v3 = −3v1 + 2v2 .

v4 depends linearly on v1 , v2 — if αv1 + βv2 = v4 , then α + 2β = 2, −2α − 3β =

−5, 4α + 9β = 7, α − β = 5 ⇒ α = 4, β = −1 which satisfy all the equations. Thus

v4 = 4v1 − v2 .

Thus dimR W = 2 and v1 , v2 is a basis of W.

R2 to R3 . Find its range, rank, null space and nullity.

1

Solution. Let x = (x1 , x2 ), y = (y1 , y2 ). Then

T(αx + βy) = T(αx1 + βy1 , αx2 + βy2 )

= (αx1 + βy1 + αx2 + βy2 , αx1 + βy1 − αx2 − βy2 , αx2 + βy2 )

= (α(x1 + x2 ), α(x1 − x2 ), αx2 ) + (β(y1 + y2 ), β(y1 − y2 ), βy2 )

= αT(x1 , x2 ) + βT(y1 , y2 )

Thus T is linear.

T(e2 ) = T(0, 1) = (1, −1, 1)

Clearly T(e1 ), T(e2 ) are linearly independent. Since T (R2 ) is generated by T(e1 ) and T(e2 ),

the rank of T is 2.

The range of T = {αT(e1 ) + βT(e2 ), α, β ∈ R}

= {α(1, 1, 0) + β(1, −1, 1)}

= {(α + β, α − β, β) | α, β ∈ R}

To find the null space of T, if T(x1 , x2 ) = (0, 0, 0), then x1 + x2 = 0, x1 − x2 = 0, x2 = 0,

so x1 = x2 = 0. Thus the null space of T is {0}, and nullity T = 0.

Question 1(c) Let V be the space of 2×2 matrices over R. Determine whether the matrices

A, B, C ∈ V are dependent where

1 2 3 −1 1 −5

A= B= C=

3 1 2 2 −4 0

Solution. If αA + βB + γC = 0, then

α + 3β + γ = 0 (1)

2α − β − 5γ = 0 (2)

3α + 2β − 4γ = 0 (3)

α + 2β = 0 (4)

From (4), we get α = −2β. This, together with (3) gives γ = −β. These satisfy (1) and

(2) also, so taking β = 1, α = −2, γ = −1 gives us −2A + B − C = 0. Thus A, B, C are

dependent.

Question 2(a) Let A be an n × n matrix such that each diagonal entry is µ and each

off-diagonal entry is 1. If B = λA is orthogonal, determine λ, µ.

P n 2

k=1 λ aik akj = δij

Taking i = j = 1, we get λ2 (µ2 +n−1) = 1 Taking i = 1, j = 2, we get λ2 (2µ+n−2) = 0.

Thus µ = −(n−2)/2 and λ2 [(n−2)2 /4+n−1] = 1. Simplifying, λ2 [n2 −4n+4+4n−4]/4 = 1,

which means λ2 = n42 , or λ = ± n2 .

2

Question 2(b) Show that

2 −1 0

A = −1 2 0

2 2 3

is diagonalizable over R. Find P such that P−1 AP is diagonal and hence find A25 .

2 − x −1 0

−1 2 − x 0 = 0

2 2 3−x

⇒ (2 − x)(2 − x)(3 − x) + 1(−(3 − x)) = 0

(3 − x)(4 − 4x + x2 − 1) = 0

Let (x1 , x2 , x3 ) be an eigenvector for λ = 1.

1 −1 0 x1

−1 1 0 x2 = 0

2 2 2 x3

(1, 1, −2) is an eigenvector with eigenvalue 1.

Let (x1 , x2 , x3 ) be an eigenvector for λ = 3.

−1 −1 0 x1

−1 −1 0 x2 = 0

2 2 0 x3

with eigenvalue 3. Take x1 = 0, x3 = 1, then x2 = 0, so (0, 0, 1) is also an eigenvector for

eigenvalue 3.

1 1 0 1 0 0 1 0 0

Let P = 1 −1 0 then AP = P 0 3 0 or P−1 AP = 0 3 0

−2 0 1 0 0 3 0 0 3

1 0 0 1 0 0

−1 25 −1

Now P A P = (P AP) = 25 0 3 25

0 25

. Thus A = P 0 3 25

0 P−1

25

1 1 0 0 3 0 0 325

2 2

0

−1 1 1

|P| = −2, so P =

2

− 2 0 .

1 1 1

3

1 1

1 1 0 1 0 0 2 2

0

A25 = 1 −1 0 0 3 25

0 1

2

− 12 0

−2 0 1 0 0 325 1 1 1

1 1

1 325 0 2 2

0

= 1 −325 0 12 − 21 0

−2 0 325 1 1 1

1+325 1−325

2 2

0

25 1+325

= 1−32 2

0

−1 + 325 −1 + 325 325

Question 2(c) Let A = [aij ] be a square matrix of order n such that |aij | ≤ M . Let λ be

an eigenvalue of A, show that |λ| ≤ nM .

X n

Lemma: If A = [aij ] and |aij | ≤ aii then |A| 6= 0.

j=1

i6=j

If |A| = 0 then there exist x1 , . . . , xn ∈ C not all zero such that

...

ai1 x1 + ai2 x2 + . . . + ain xn = 0

...

an1 x1 + an2 x2 + . . . + ann xn = 0

x

Let |xi | = max(|x1 |, |x2 |, . . . , |xn |), so | xji | ≤ 1 for all j.

x 1 x 2 x n

0 = aii − (−ai1 − ai2 − . . . − ain )

xi xi xi

x1 x2 xn

≥ |aii | − ai1 + ai2 + . . . + ain

xi xi xi

≥ |aii | − |ai1 | − |ai2 | − . . . − |ain |

n

X

which contradicts the premise |aij | ≤ aii Thus |A| 6= 0.

j=1

i6=j

4

n

X

Now the lemma tells us that if |λ − aii | > |aij | then |λI − A| =

6 0, so λ is not an

j=1

i6=j

n

X

eigenvalue of A. Thus |λ| ≤ |λ − aii | + |aii | ≤ |aij | ≤ nM as desired.

j=1

Question 3(a) Define a positive definite matrix and show that a positive definite matrix is

always non-singular. Show that the converse is not always true.

the associated quadtratic form

x1

x2

x1 x2 . . . xn A

. . . > 0

xn

If |A| = 0 then rank A < n, which means that columns of A i.e. c1 , c2 , . . . , cn are

linearly dependent i.e. there exist real numbers x1 , x2 , . . . , xn not all zero such that

x1 x1

x2 x2

x1 c1 + x2 c2 + . . . + xn cn = A . . . = 0 =⇒ x1 x2 . . . xn A . . . = 0

xn xn

where (x1 , x2 , . . . , xn ) 6= (0, 0, . . . , 0), which means that A is not positive definite. Thus A

is positive definite =⇒ |A| 6= 0.

The converse is not true. Take

1 0 0

A = 0 1 0

0 0 −1

0

0 0 1 A 0 = −1

1

so A is not positive definite.

5

Question 3(b) Find the eigenvalues and their corresponding eigenvectors for the matrix

6 −2 2

−2 3 −1

2 −1 3

0 = |A − xI|

6 − x −2 2

= −2 3 − x −1

2 −1 3 − x

= (6 − x)((3 − x)2 − 1) + 2(−6 + 2x + 2) + 2(2 − 6 + 2x)

= (6 − x)(9 − 6x + x2 ) − 6 + x − 8 + 4x − 8 + 4x

0 = x3 − 12x2 + 36x − 32

= (x − 2)(x2 − 10x + 16)

Let (x1 , x2 , x3 ) be an eigenvector for λ = 2.

4 −2 2 x1

−2 1 −1 x2 = 0

2 −1 1 x3

Thus 4x1 − 2x2 + 2x3 = 0, −2x1 + x2 − x3 = 0, 2x1 − x2 + x3 = 0. Take x1 = 1, x2 = 0, then

x3 = −2, so (1, 0, −2) is an eigenvector with eigenvalue 2. Take x1 = 0, x2 = 1, then x3 = 1,

so (0, 1, 1) is an eigenvector with eigenvalue 2.

Let (x1 , x2 , x3 ) be an eigenvector for λ = 8.

−2 −2 2 x1

−2 −5 −1 x2 = 0

2 −1 −5 x3

Thus −2x1 − 2x2 + 2x3 = 0, −2x1 − 5x2 − x3 = 0, 2x1 − x2 − 5x3 = 0. From the last two, we

get x2 + x3 = 0, and from the first we get x1 = 2x3 . Take x3 = 1, then x2 = −1, x1 = 2, so

(2, −1, 1) is an eigenvector with eigenvalue 8.

Question 3(c) Find P invertible such that P reduces Q(x, y, z) = 2xy + 2yz + 2zx to its

canonical form.

0 1 1

A = 1 0 1

1 1 0

6

which has all diagonal entries 0, so we cannot complete squares right away.

0 1 1 1 0 0 1 0 0

1 0 1 = 0 1 0 A 0 1 0

1 1 0 0 0 1 0 0 1

Add the second row to the first and the second column to the first.

2 1 2 1 1 0 1 0 0

1 0 1 = 0 1 0 A 1 1 0

2 1 0 0 0 1 0 0 1

C

2 1

from C2 .

1 − 12 0

2 0 2 1 1 0

0 − 21 0 = − 12 21 0 A 1 12 0

2 0 0 0 0 1 0 0 1

1 − 12 −1

2 0 0 1 1 0

0 − 1 0 = − 1 1 0 A 1 1 −1

2 2 2 2

0 0 −2 −1 −1 1 0 0 1

1 − 12 −1

2 0 0

Thus P = 1 21 −1 and P0 AP = 0 − 12 0

0 0 1 0 0 −2

2 1 2 2

So Q(x, y, z) −→ 2X − 2 Y − 2Z .

Alternative Solution. Let x = X, y = X + Y, z = Z

= 2[X 2 + XY + 2ZX + ZY ]

Y Y2

= 2[(X + + Z)2 − − Z 2]

2 4

Put ξ = X + Y /2 + Z, η = Y, ζ = Z, so X = ξ − η/2 − ζ, Y = η, Z = ζ.

1 − 12 −1 1 − 21 −1

x 1 0 0 X 1 0 0 ξ ξ

y = 1 1 0 Y = 1 1 0 0 1 0 η = 1 12 −1 η

z 0 0 1 Z 0 0 1 0 0 1 ζ 0 0 1 ζ

1 − 21 −1

0 0 1

x = X, y = X + Y, z = Z to create one square term to complete the squares.

7

UPSC Civil Services Main 1998 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Given two linearly independent vectors (1, 0, 1, 0) and (0, −1, 1, 0) of R4 ,

find a basis of R4 which includes them.

Solution. Let v1 = (1, 0, 1, 0), v2 = (0, −1, 1, 0). Clearly these are linearly independent.

Let e1 , e2 , e3 , e4 be the standard basis. Then v1 , v2 , e1 , e2 , e3 , e4 generate R4 . We have to

find four vectors out of these which are linearly independent and include v1 , v2 .

If αv1 + βv2 + γe1 = 0, then α + γ = 0, −α = 0, α + β = 0 ⇒ α = β = γ = 0. Therefore

v1 , v2 , e1 are linearly independent.

We now show that v1 , v2 , e1 , e4 are linearly independent. Let αv1 + βv2 + γe1 + δe4 = 0

then δ = 0, and therefore α = β = γ = 0 because v1 , v2 , e1 are linearly independent.

Thus v1 , v2 , e1 , e4 is a basis of R4 .

Note that e2 = v1 − v2 − e1 , e3 = v1 − e1 .

Question 1(b) If V is a finite dimensional vector space over R and if f and g are two

linear transformations from V to R such that f (v) = 0 implies g(v) = 0, then prove that

g = λf for some λ ∈ R.

If g 6= 0, then f 6= 0. Thus ∃v ∈ V such that f (v) 6= 0 ⇒ ∃w ∈ V such that f (w) = 1

v

(Note that f ( f (v) ) = 1).

Thus V/ ker f w R, or dim(ker f ) = n − 1. Similarly ker g has dimension n − 1. In fact,

ker f = ker g ∵ ker f ⊆ ker g and dim(ker f ) = dim(ker g). Let {v2 , . . . , vn } be a basis of

ker f and extend it to {v1 , v2 , . . . , vn } a basis of V. Then g = λf with λ = g(v1 )/f (v1 ) ∵

if v = α1 v1 + . . . + αn vn , then g(v) = α1 g(v1 ) = α1 λf (v1 ) = λf (v).

1

Question 1(c) Let T : R3 −→ R3 be defined by T(x1 , x2 , x3 ) = (x2 , x3 , −cx1 − bx2 − ax3 )

where a, b, c are fixed real numbers. Show that T is a linear transformation of R3 and that

A3 + aA2 + bA + cI = 0 where A is the matrix of T w.r.t. the standard basis of R3 .

Solution. Let x = (x1 , x2 , x3 ), y = (y1 , y2 , y3 ). Then

T(αx + βy) = (αx2 + βy2 , αx3 + βx3 , −c(αx1 + βy1 ) − b(αx2 + βy2 ) − a(αx3 + βy3 ))

= α(x2 , x3 , −cx1 − bx2 − ax3 ) + β(y2 , y3 , −cy1 − by2 − ay3 )

= αT(x) + βT(y)

Thus T is linear.

Clearly

T(1, 0, 0) = (0, 0, −c)

T(0, 1, 0) = (1, 0, −b)

T(0, 0, 1) = (0, 1, −a)

0 1 0

A= 0 0 1

−c −b −a

The characteristic equation of A is |A − λI| = 0.

−λ 1 0

0 λ 1 = 0

−c −b −a − λ

−λ2 (a + λ) − bλ − c = 0

λ3 + aλ2 + bλ + c = 0

Now by the Cayley-Hamilton theorem A3 + aA2 + bA + cI = 0.

Question 2(a) If A and B are two matrices of order 2 × 2 such that A is skew-Hermitian

and AB = B then show that B = 0.

Solution. We first of all prove that eigenvalues of skew-Hermitian matrices are 0 or pure

0

imaginary. Let A be skew-Hermitian, i.e. A = −A and let λ be its characteristic root. If

x is an eigenvector of λ, then

Ax = λx

⇒ x0 λx = x0 Ax

0

= −x0 A x

0

= −Ax x

= −λx0 x

Thus λ = −λ ∵ x0 x 6= 0, showing that the real part of λ is 0.

Now if B 6= 0 and c1 , c2 are the columns of B, then c1 6= 0 or c2 6= 0. AB = B means

that Ac1 = c1 and Ac2 = c2 . Since either c1 6= 0 or c2 6= 0, 1 must be an eigenvalue of A,

which is not possible. Hence c1 = 0 and c2 = 0, which means B = 0.

2

Question 2(b) If T is a complex matrix of order 2 × 2 such that tr T = tr T2 = 0, then

show that T2 = 0.

Solution. Let λ1 , λ2 be the eigenvalues of T, then λ21 , λ22 are the eigenvalues of T2 . Given

that

tr T = λ1 + λ2 = 0

tr T2 = λ21 + λ22 = 0

characteristic equation of T is (x − λ1 )(x − λ2 ) = 0, or x2 = 0. By Cayley-Hamilton

theorem, we immediately get T2 = 0.

Question 2(c) Prove that a necessary and sufficient condition for an n × n real matrix A

to be similar to a diagonal matrix is that the set of characteristic vectors of A includes a set

of n linearly independent vectors.

Solution.

Necessity: By hypothesis there exists a nonsingular matrix P such that

λ1 0 . . . 0

0 λ2 . . . 0

P−1 AP = D = ... ... ... ...

0 0 . . . λn

sponding to the eigenvalues λ1 , . . . , λn . Since P is nonsingular, c1 , c2 , . . . , cn are linearly

independent. Thus the set of characteristic vectors of A includes a set of n linearly indepen-

dent vectors.

Sufficiency: Let c1 , c2 , . . . , cn be n linearly independent eigenvectors of A corresponding

to eigenvalues λ1 , . . . , λn . Thus Aci = λi ci for i = 1, . . . , n. Let P = [c1 , c2 , . . . , cn ], then

P is nonsingular (otherwise 0 is an eigenvalue of P, so ∃x = (x1 , . . . , xn ) 6= 0 such that

Px = 0 ⇒ x1 c1 + . . . + xn cn = 0 ⇒ c1 , c2 , . . . , cn are not linearly independent.). Clearly

λ1 0 . . . 0

0 λ2 . . . 0

P−1 AP = D = ... ... ... ...

0 0 . . . λn

3

Question 3(a) Let A be a m × n matrix. Show that the sum of the rank and nullity of A

is n.

F is the field to which the entries of A belong, and the bases for F n , F m are standard bases.

Let T : V −→ W be a linear transformation, where dim(V) = n, dim(W) = m. We shall

show that dim(T(V)) + dim(kernel T) = n.

Take vn−r+1 , . . . , vn to be any basis of kernel T, where dim(kernel T) = r. Complete it

to a basis v1 , . . . , vn−r+1 . . . , vn of V. We shall show that T(v1 ), . . . , T(vn−r ) are linearly

independent and generate T(V), thus dim(T(V)) = n − r.

If w ∈ T(V), then ∃v ∈ V such that T(v) = w. If v = α1 v1 + . . . + αn vn , αi ∈ F, then

w = T(v) = α1 T(v1 ) + . . . + αn−r T(vn−r ) because T(vi ) = 0 for i > n − r. Thus T(V) is

generated by T(v1 ), . . . , T(vn−r ).

If α1 T(v1 ) + . . . + αn−r T(vn−r ) = 0, then T(α1 v1 + . . . + αn−r vn−r ) = 0. This implies

α1 v1 +. . .+αn−r vn−r ∈ kernel T ⇒ α1 v1 +. . .+αn−r vn−r = αn−r+1 vn−r+1 +. . .+αn vn . But

v1 , . . . , vn are linearly independent, so αi = 0 for i = 1, . . . , n. Hence T(v1 ), . . . , T(vn−r ) are

linearly independent, so they form a basis for T(V). Thus dim(T(V)) + dim(kernel T) = n.

Question 3(b) Find all real 2 × 2 matrices A with real eigenvalues which satisfy AA0 = I.

a b a c 1 0

=

c d b d 0 1

c cos θ + d sin θ = 0 c cos θ sin θ + d sin2 θ = 0

⇒ ⇒ d = cos θ, c = − sin θ

−c sin θ + d cos θ = 1 −c sin θ cos θ + d cos2 θ = cos θ

cos θ sin θ

Thus A = , θ is real.

− sin θ cos θ

Now the eigenvalues of A are given by

cos θ − λ sin θ

|A − λI| = =0

− sin θ cos θ − λ

√

2 cos θ ± 4 cos2 θ − 4

λ= = cos θ ± i sin θ

2

Since the eigenvalues of A are real, sin θ = 0, so cos θ = ±1. Thus

1 0 −1 0

A= ,

0 1 0 −1

4

0 1

If |A| = −1, J = , then |JA| = 1. Also JA(JA)0 = JAA0 J0 = JJ0 = I. Thus

1 0

cos θ sin θ

JA =

− sin θ cos θ

−1 cos θ sin θ 0 1 cos θ sin θ − sin θ cos θ

A=J = =

− sin θ cos θ 1 0 − sin θ cos θ cos θ sin θ

Now the eigenvalues of A are given by

λ + sin θ − cos θ

0 = |λI − A| = = λ2 − sin2 θ − cos2 θ = λ2 − 1

− cos θ λ − sin θ

Hence λ = ±1, so the eigenvalues are always real. Thus the possible values of A are

1 0 −1 0 − sin θ cos θ

, , for all real θ

0 1 0 −1 cos θ sin θ

Question 3(c) Reduce to diagonal matrix by rational congruent transformation the sym-

metric matrix

1 2 −1

A= 2 0 3

−1 3 1

= (x + 2y − z)2 − 4y 2 + 10yz

5 25

= (x + 2y − z)2 − 4(y − z)2 + z 2

4 4

25

= X 2 − 4Y 2 + Z 2

4

where X = x + 2y − z, Y = y − 5z/4, Z = z. From this we get z = Z, y = Y + 5Z/4, x =

X − 2Y − 32 Z. Thus

1 −2 − 32

1 0 0 1 0 0 1 2 −1

0 −4 0 = −2 1 5

0 2 0 3 0 1 4

0 0 25 4

− 23 54 1 −1 3 1 0 0 1

5

UPSC Civil Services Main 1999 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Let V be the vector space of functions from R to R. Show that f, g, h ∈ V

are linearly independent where f (t) = e2t , g(t) = t2 and h(t) = t.

Solution. Let a, b, c ∈ R and let ae2t + bt2 + ct = 0 for all t. Setting t = 0 shows that

a = 0. From t = 1 we get b + c = 0, and t = −1 gives b − c = 0, hence b = c = 0. Thus

f, g, h are linearly independent.

Question 1(b) If the matrix linear transformation T on V2 (R) with respect to the

of the

1 1

basis B = {(1, 0), (0, 1)} is , then what is the matrix of T with respect to the ordered

1 1

basis B1 = {(1, 1), (1, −1)}.

e2 . Then T(v1 ) = T((1, 1)) = (2, 2) = 2e1 + 2e2 = 2v1 . T(v2 ) = T((1, −1)) =(0, 0) = 0.

2 0

Thus the matrix of T with respect to the ordered basis B1 = {(1, 1), (1, −1)} is .

0 0

4 2 2

Question 1(c) Diagonalize the matrix A = 2 4 2.

2 2 4

1

Solution. The characteristic equation is

4 − x 2 2

0 = 2

4−x 2

2 2 4 − x

= (4 − x)((4 − x)2 − 4) + 2(4 − 8 + 2x) − 2(8 − 2x − 4)

= (4 − x)(12 − 8x + x2 ) − 8 + 4x − 8 + 4x

= 48 − 32x + 4x2 − 12x + 8x2 − x3 − 16 + 8x

= −(x3 − 12x2 + 36x − 32)

= −(x − 2)(x2 − 10x + 16) ∵ 2 is a root

= −(x − 2)(x − 2)(x − 8)

The characteristic roots are 2, 2, 8.

If (x1 , x2 , x3 ) is an eigenvector for λ = 8, then

−4 2 2 x1 0

2 −4 2 x2 = 0

2 2 −4 x3 0

Thus x1 = x2 = x3 , so (1, 1, 1) is an eigenvector for λ = 8.

Similarly for λ = 2,

2 2 2 x1 0

2 2 2 x2 = 0

2 2 2 x3 0

Thus x1 + x2 + x3 = 0. Take x1 = 1, x2 = 0, so (1, 0, −1) is an eigenvector. Take x1 = 0, x2 =

1, so (0, 1, −1)

is an eigenvector

for λ = 2. These

eigenvectors

are linearly independent.

1 1 0 8 0 0

Thus if P = 1 0 1 , then P−1 AP = 0 2 0

1 −1 −1 0 0 2

8 0 0

To check, verify that AP = P 0 2 0

0 0 2

1 0 0 i

Question 2(a) Test for congruency the matrices A = and B = . Prove

0 −1 −i 0

that A2n = B2m = I where m, n are positive integers.

Solution. A and B are not congruent, because A is symmetric and B is not. If A ≡ B

then ∃P non-singular such that P0 AP = B which implies that B should be symmetric.

2 1 0 1 0 1 0

A = =

0 −1 0 −1 0 1

0 i 0 i 1 0

B2 = =

−i 0 −i 0 0 1

2

Hence A2n = (A2 )n = I, and B2m = (B2 )m = I.

Question 2(b) If A is a skew symmetric matrix of order n then prove that (I−A)(I+A)−1

is orthogonal.

Solution.

O = (I − A)(I + A)−1

OO0 = (I − A)(I + A)−1 ((I + A)−1 )0 (I − A)0

= (I − A)(I + A)−1 (I − A)−1 (I + A) as A0 = −A

= (I − A)[(I − A)(I + A)]−1 (I + A)

= (I − A)[I − A2 )]−1 (I + A)

= (I − A)[(I + A)(I − A)]−1 (I + A)

= (I − A)(I − A)−1 (I + A)−1 (I + A)

= I

Question 2(c) Test for positive definiteness the quadratic form 2x2 + y 2 + 2z 2 + 2xy − 2zx.

1

2

(4x2 + 3y 2 + 4z 2 + 4xy − 4zx)

= 21 ((2x + y − z)2 + y 2 + 3z 2 + 2yz)

= 21 ((2x + y − z)2 + (y + z)2 + 2z 2 )

Now 2x + y − z = 0, y + z = 0, z = 0 implies x = y = z = 0. Hence the form is positive

definite.

1 −1 1

S = −1 1 −1

1 −1 1

1 − λ −1 1

−1 1 − λ −1 =0

1 −1 1 − λ

3

or λ3 − 3λ2 = 0. Thus λ = 0, 0, 3.

We next determine the characteristic vectors. For λ = 0, we get

1 −1 1 x1 0

−1 1 −1 x2 = 0

1 −1 1 x3 0

corresponding to λ = 0.

For λ = 3, we get

−2 −1 1 x1 0

−1 −2 −1 x2 = 0

1 −1 −2 x3 0

This yields x1 = −x2 = x3 . Take (1, −1, 1) as the characteristic vector for λ = 3.

Thus if 1

√ √1 − √1

3 2 6

O = − √13 √12 √16

√1 0 √2

3 6

3 0 0

Then O0 SO = 0 0 0 .

0 0 0

Let

X x

O Y = y

Z z

or

X Y Z

x = √ +√ −√

3 2 6

X Y Z

y = −√ + √ + √

3 2 6

X 2Z

z = √ +√

3 6

Thus the given equation can be transformed to

X Y

3X 2 + √ + √ − √Z6 + √X3 − √Y2 − √Z6 − 2X

√ − 4Z

√ +6=0

3 2

2

√ 3 6

⇒ 3X − Z 6 + 6 = 0

q √

⇒ 32 X 2 = Z − 6

√ 2

q

2

Shifting the origin to (0, 0, 6), we get X = 3

Z, showing that the equation is a parabolic

cylinder.

4

1 Reduction of Quadrics

For the sake of completeness, we give the complete theoretical discussion for the above

question.

Let

a h g u x

h b f v y

x y z 1

g

=0

f c w z

u v w d 1

Step I. Consider

a h g x x

x y z h b f y = x y z S y

g f c z z

These on normalization give us

O = ||vv11 || ||vv22 || ||vv33 ||

X x

O Y = y

Z z

F (X, Y, Z) = λ1 X 2 + λ2 Y 2 + λ3 Z 2 + 2U X + 2V Y + 2W Z + d = 0

X x

Note 1 Since O is orthogonal, the transformation O Y = y is just a rotation

Z z

of the axes, and therefore the nature of the quadric is unaffected.

5

Step II. We now consider 3 possibilities (ρ is rank of the matrix):

1. ρ(S) = 3 ⇒ λ1 λ2 λ3 6= 0. Shift the origin to (− λU1 , − λV2 , − W

λ3

), i.e. x = X + λU1 , y =

Y + λV2 , z = Z + W

λ3

. (Actually we are just completing the squares.) F gets transformed

to

λ1 x2 + λ2 y 2 + λ3 z 2 + d2 = 0

2. ρ(S) = 2. One characteristic root, say λ3 = 0. Shift the origin to (− λU1 , − λV2 , 0), and F

gets transformed to

λ1 x2 + λ2 y 2 + 2w2 z + d2 = 0

3. ρ(S) = 1. Two characteristic roots, say λ2 = λ3 = 0. Shift the origin to (− λU1 , 0, 0),

and F gets transformed to

λ1 x2 + 2v2 y + 2w2 z + d2 = 0

plane.

Step III. Observe that ρ(S) ≤ ρ(Q) ≤ 4, ρ(S) ≤ 3.

1. Let ρ(Q) = 4, ρ(S) = 3. As shown above, F (x, y, z) = 0 is transformed to

λ1 x2 + λ2 y 2 + λ3 z 2 + d2 = 0

|Q| = λ1 λ2 λ3 d2 ⇒ d2 = |Q||S|

. Thus the quadric is λ1 x2 + λ2 y 2 + λ3 z 2 = − |Q|

|S|

, which

is a central quadric i.e. a quadric surface with a center, e.g., a sphere, ellipsoid, or

hyperboloid, depending upon the signs and magnitudes of the eigenvalues. If the right

hand side has positive sign (maybe by multiplying the equation with -1), then look at

the signs of the coefficients of the l.h.s. If all are positive, it is an ellipsoid, further

if all are equal, it is a sphere. If 1 or 2 are negative, it is a hyperboloid. If all 3 are

negative, the surface is the empty set.

2. ρ(Q) = 3, ρ(S) = 3. |Q| = λ1 λ2 λ3 d2 = 0 ⇒ d2 = 0, because λ1 λ2 λ3 6= 0. Thus the

quadric becomes λ1 x2 + λ2 y 2 + λ3 z 2 = 0, which is a cone.

3. ρ(Q) = 4, ρ(S) = 2

F (x, y, z) = λ1 x2 + λ2 y 2 + 2w2 z + d2 = 0

λ1 0 0 0

0 λ2 0 0

Q= 0 0 0 w2

0 0 w 2 d2

d2

|Q| = −λ1 λ2 w22 . Since ρ(Q) = 4, w2 6= 0. Shifting the origin to (0, 0, − 2w 2

) we get

F (x, y, z) = λ1 x2 + λ2 y 2 + 2w2 z = 0

where w22 = −|Q|/λ1 λ2 . The surface is a paraboloid.

6

4. ρ(Q) = 3, ρ(S) = 2

F (x, y, z) = λ1 x2 + λ2 y 2 + 2w2 z + d2 = 0

λ1 0 0 0

0 λ2 0 0

Q= 0 0

0 0

0 0 0 d2

F (x, y, z) = λ1 x2 + λ2 y 2 + d2 = 0

5. ρ(Q) = 2, ρ(S) = 2

λ1 0 0 0

0 λ2 0 0

Q=

0 0

0 0

0 0 0 d2

ρ(Q) = 2 ⇒ d2 = 0, and F (x, y, z) = λ1 x2 + λ2 y 2 = 0. The quadric is a pair of distinct

planes or a point, if λ1 = λ2 6= 0.

6. ρ(Q) = 4, ρ(S) = 1

λ1 0 0 0

0 0 0 v2

Q= 0 0 0 w2

0 v2 w2 d2

which shows that ρ(Q) = 4 is not possible.

7. ρ(Q) = 3, ρ(S) = 1

d2

(0, − 2v2

, 0).

F (x, y, z) = λ1 x2 + 2v2 y + 2w2 z = 0

7

Rotate the axes by

x = X

v2 w2

y = p 2 2

Y −p 2 Z

v2 + w2 v2 + w22

w2 v

z = p 2 Y + p 2 Z

v2 + w22 v22 + w22

q

2

F (x, y, z) = λ1 X + 2v3 Y = 0 v3 = v22 + w22

Thus the quadric is a parabolic cylinder.

F (x, y, z) = λ1 X 2 + d2 = 0

The quadric is two parallel planes.

represents two coincident planes x = 0.

ρ(Q) ρ(S) Surface Canonical form

4 3 central quadric λ1 x + λ2 y 2 + λ3 z 2 = − |Q|

2

|S|

2 2 2

3 3 cone λ1 x + λ2 y + λ3 z = 0

4 2 paraboloid λ1 x + λ2 y 2 + 2w2 z = 0, w22 = − λ|Q|

2

1 λ2

3 2 elliptic or hyperbolic cylinder λ1 x2 + λ2 y 2 + d2 = 0

pair of distinct planes if λ1 λ2 < 0

2 2 λ1 x2 + λ2 y 2 = 0

point if λ1 λ2 > 0

4 1 Not possible p

3 1 parabolic cylinder λ1 X 2 + 2v3 Y = 0, v3 = v22 + w22

2 1 pair of parallel planes λ1 X 2 + d2 = 0

1 1 Two coincident planes λ1 X 2 = 0

8

UPSC Civil Services Main 2000 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

T = {(x, y) | x, y ∈ V}

for every α, β ∈ R. Show that T is a vector space over C.

Solution.

1. v1 , v2 ∈ T ⇒ v1 + v2 ∈ T

4. Clearly v1 +v2 = v2 +v1 and (v1 +v2 )+v3 = v1 +(v2 +v3 ) as addition is commutative

and associative in V.

5. z ∈ C, v ∈ T ⇒ zv ∈ T

6. 1v = (1 + i0)(x, y) = (x, y)

7.

(α + iβ)((x1 , y1 ) + (x2 , y2 ))

= (α(x1 + x2 ) − β(y1 + y2 ), β(x1 + x2 ) + α(y1 + y2 ))

= (αx1 − βy1 , βx1 + αy1 ) + (αx2 − βy2 , βx2 + αy2 )

= (α + iβ)(x1 , y1 ) + (α + iβ)(x2 , y2 )

1

8.

= (αγ − βδ + i(αδ + βγ))(x, y)

= ((αγ − βδ)x − (αδ + βγ)y, (αγ − βδ)y + (αδ + βγ)x)

= (α(γx − δy) − β(δx + γy), β(γx − δy) + α(δx + γy))

= (α + iβ)((γx − δy), (δx + γy))

= (α + iβ)((γ + iδ)(x, y))

Question 1(b) Show that if λ is a characteristic root of a non-singular matrix A, then λ−1

is a characteristic root of A−1 .

Solution.

Av = λv v 6= 0

−1 −1

⇒ A Av = λA v

⇒ A−1 v = λ−1 v

Thus λ−1 is a characteristic root of A−1 .

if and only if

1 2 3

A = BB0 for some non-singular B. Show also that A = 2 5 7 is positive definite

3 7 11

0 0

and find B such that A = BB . (Here B is the transpose of B.)

column vector. Since |B| 6= 0, B0 x 6= 0 =⇒ x0 B.(B0 x) is the sum on n squares, at least one

of which is non-zero. Thus x0 Ax > 0 whenever x 6= 0, showing that A is positive definite.

Conversely, if A is positive definite, then ∃P non-singular such that P0 AP = In . Thus

A = P0−1 P−1 . Letting B = P0−1 we get A = BB0 as required.

The existence of P can be found by induction on n. Let

a11 a12 . . . a1n

a21 a22 . . . a2n

A=

...

an1 an2 . . . ann

Define

1 − aa11 . . . − aa1n

12

11

0 1 ... 0

Q=

...

0 0 ... 1

2

0 a11 0

Then Q is non-singular, and Q AQ = , where S is (n − 1) × (n − 1) positive

0 S

definite. Let Q∗ be a (n − 1) × (n − 1) non-singular

matrix such that Q∗ 0 SQ∗ is diagonal,

1 0

by induction. Then let Q1 = , and let P = Q1 Q. Then P0 AP is diagonal

0 Q∗

(b11 , b22 , . . . , bnn ). Let B = diagonal ( √b111 , . . . , √b1nn ). Then B0 P0 APB = In .

The quadratic form Q(x, y, z) associated with the given matrix A is given by

1 2 3 x

x y z 2 5 7 y = x2 + 5y 2 + 11z 2 + 4xy + 6xz + 14yz

3 7 11 z

definite, as z = 0, y + z = 0, x + 2y + 3z = 0 =⇒ x = y = z = 0.

If B is a 3 × 3 matrix such that

x x + 2y + 3z

B0 y = y+z

z z

then x0 BB0 x = Q = x0 Ax, so A = BB0 as A and BB0 are both symmetric. Clearly

1 0 0

B= 2 1 0

3 1 1

has a unique solution provided the coefficient matrix is non-singular.

Solution. If A is non-singular, then the system is consistent because the rank of the

coefficient matrix A = n = rank of the n × n + 1 augmented matrix (A, B). If x1 , x2 are

two solutions, then

Ax1 = B = Ax2

=⇒ A(x1 − x2 ) = 0

=⇒ A−1 A(x1 − x2 ) = 0

=⇒ x1 = x2

Thus the unique solution is given by the column vector x = A−1 B.

3

Question 2(c) Prove that two similar matrices have the same characteristic roots. Is the

converse true? Justify your claim.

P−1 AP| = |P−1 λIP − P−1 AP| = |P−1 ||λI − A||P| = |λI − A|. (Note that |X||Y| = |XY|.)

Thus the characteristic polynomial of B is the same as that of A, so both A and B have the

same characteristic roots.

The converse is not true. Let

1 1 1 0

A= , B=

0 1 0 1

Then A and B have the same characteristic polynomial (λ − 1)2 and thus the same charac-

teristic roots. But B can never be similar to A because P−1 BP = B whatever P may be.

4

UPSC Civil Services Main 2001 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Show that the vectors (1, 0, −1), (0, −3, 2) and (1, 2, 1) form a basis of the

vector space R3 (R).

Solution. Since dimR (R3 ) = 3, it is enough to prove that these are linearly independent.

If possible, let

a(1, 0, −1) + b(0, −3, 2) + c(1, 2, 1) = 0

This implies

a + c = 0, −3b + 2c = 0, −a + 2b + c = 0

Solving for c, c + 34 c + c = 0, so c = 0, hence a = b = 0. (Note that if these linearly

independent vectors were not a basis, they could be completed into one, but in R3 any four

vectors are linearly dependent, so this is a maximal linearly independent set, hence it is a

basis.

Alternate Solution. Since dim(R3 ) = 3, to show that (1, 0, −1), (0, −3, 2) and (1, 2, 1)

form a basis it is enough to show that these vectors generate R3 . In fact, given (x1 , x2 , x3 ),

we can always find a, b, c s.t. (x1 , x2 , x3 ) = a(1, 0, −1) + b(0, −3, 2) + c(1, 2, 1) as follows:

a + c = x1 , −3b + 2c = x2 , −a + 2b + c = x3 . Thus (c − x1 ) + 2(2c − x2 )/3 + c = x3 ,

so c + 34 c + c = x1 + 23 x2 + x3 . Thus c = 3x1 +2x 10

2 +3x3

, a = x1 − c = 7x1 −2x10

2 −3x3

, and

2c−x2 x1 −x2 +x3

b= 3 = 5

.

|A|

Question 1(b) If λ is a characteristic root of a non-singular matrix A, then prove that λ

is a characteristic root of Adj A.

Solution. If µ is a characteristic root of A, then aµ is a characteristic root of aA for a

constant a, because if Av = µv, v 6= 0 a vector, then aAv = aµv. Hence the result.

If λ is the characteristic root of A, |A| 6= 0, then λ 6= 0, and λ−1 is a characteristic root

of A−1 , because Av = λv =⇒ A−1 v = λ−1 v.

Since Adj A = A−1 |A|, it follows that |A| λ

is a characteristic root of Adj A.

1

1 0 0

Question 2(a) If A = 1 0 1 show that for all integers n ≥ 3, An = An−2 + A2 − I.

0 1 0

50

Hence determine A .

λ−1 0 0

1 λ 1 = 0

0 1 λ

or (λ−1)(λ2 −1) = λ3 −λ2 −λ+1 = 0. From the Cayley-Hamilton theorem, A3 −A2 −A+I =

0 ⇒ A3 = A + A2 − I. Thus the result is true for n = 3. Suppose the theorem is true for

n = m i.e. Am = Am−2 + A2 − I. We shall prove it for m + 1.

Am+1 = Am A

= (Am−2 + A2 − I)A

= Am−1 + A3 − A

= Am−1 + A2 + A − A − I

= Am−1 + A2 − I

Let n = 2m. Using successively An = An−2 + A2 − I, we get A2m = mA2 − (m − 1)I.

Now

1 0 0 1 0 0 1 0 0

A2 = 1 0 1 1 0 1 = 1 1 0

0 1 0 0 1 0 1 0 1

so

A50 = 25A

2

− 24I

25 0 0 24 0 0

= 25 25 0 − 0 24 0

25 0 25 0 0 24

1 0 0

= 25 1 0

25 0 1

Prove that every matrix congruent to a skew-symmetric matrix is skew-symmetric.

−(P0 SP), so P0 SP is also skew-symmetric.

2

Question

2(c) Determine

the orthogonal matrix P such that P−1 AP is diagonal where

7 4 −4

A= 4 −8 −1 .

−4 −1 −8

λ − 7 −4 4

−4 λ + 8 1 = 0

4 1 λ + 8

(λ − 7)((λ + 8)2 − 1) + 4(−4 − 4λ − 32) + 4(−4 − 4λ − 32) = 0

λ3 + 9λ2 − 81λ − 729 = 0

(λ + 9)(λ2 − 81) = 0

−4x1 + 17x2 + x3 = 0

4x1 + x2 + 17x3 = 0

From the second and third we get 18x2 +18x3 = 0. Take x2 = 1. Then x3 = −1, x1 = 4,

so (4, 1, −1) is an eigenvector for λ = 9.

−4x1 − x2 + x3 = 0

4x1 + x2 − x3 = 0

so (0, 1, 1) is an eigenvector. Take x1 = −1, x2 = 2, then x3 = −2, so (−1, 2, −2) is

another eigenvector. These two are orthogonal to each other and are eigenvectors for

λ = −9. Note that to make the second vector orthogonal to the first, we needed to

ensure x2 = −x3 , then the equation suggested values for x1 , x2 .

Let

0 − 13 √4

18

P= √1 2 √1

2 3 18

√1 − 32 − √118

2

3

Clearly P0 P = I, since the columns of P are mutually orthogonal unit vectors.

Moreover from

−9 0 0

Ax = xλ for the eigenvalues and eigenvectors, it follows that AP = P 0 −9 0 .

0 0 9

−9 0 0

−1

Thus P AP = 0 −9 0 , which is diagonal as required.

0 0 9

E = (X − x1 )2 + . . . + (X − xn )2

= nX 2 − 2X(x1 + . . . + xn ) + (x21 + x22 + . . . + x2n )

A= B=

n n

Then E = n(X 2 − 2AX + B) = n((X − A)2 + B − A2 ). When X = A, E = n(B − A2 ) = Φ,

and since E ≥ 0, Φ ≥ 0.

If x1 = x2 = . . . = xn = 1, then Φ = 0 showing that Φ is actually positive semi-definite.

Alternate solution. By Cauchy’s inequality

n

! n ! n

!2

X X X

a2i b2i ≥ ai bi

i=1 i=1 i=1

Setting b1 = b2 = . . . = bn = 1, we get

n

! n

!2

X X

n a2i − ai ≥0

i=1 i=1

4

UPSC Civil Services Main 2002 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

is linear and non-singular.

Solution.

= λ(a − b, b − c, a + c)

= λT(a, b, c)

T(a, b, c) + T(x, y, z) = (a − b, b − c, a + c) + (x − y, y − z, x + z)

= (a − b + x − y, b − c + y − z, a + c + x + z)

= T(a + x, b + y, c + z)

Thus T is linear.

Now we show that

T(1, 0, 0) = (1, 0, 1)

T(0, 1, 0) = (−1, 1, 0)

T(0, 0, 1) = (0, −1, 1)

⇒ a1 − a2 = 0, a2 − a3 = 0, a1 + a3 = 0

⇒ a1 = a2 = a3 = 0

1

Thus (1, 0, 1), (−1, 1, 0), (0, −1, 1) are linearly independent.

Since (1, 0, 0), (0, 1, 0), (0, 0, 1) generate R3 , (1, 0, 1), (−1, 1, 0), (0, −1, 1) generate T(R3 ),

hence dim(T(R3 )) = 3. Thus T is non-singular.

Alternatively,

T(a, b, c) = (0, 0, 0) ⇐⇒ a − b = 0, b − c = 0, a + c = 0 =⇒ a = b = c = 0

Question 1(b) Prove that a square matrix A is non-singular if and only if the constant

term in its characteristic polynomial is different from 0.

Solution. Let

a11 ... a1n

a21 ... a2n

A=

. . .

... ...

an1 ... ann

Then characteristic polynomial of A = Det(xI − A). I = n × n unit matrix. Clearly

n

X

Det(xI − A) = xn − aii xn−1 + . . . + (−1)n DetA

i=1

e, b, 2d + e, b + e). Obtain bases for its null space and range space.

Solution. Clearly

T (1, 0, 0, 0, 0) = (0, 0, 0, 0, 0)

T (0, 1, 0, 0, 0) = (1, 0, 1, 0, 1)

T (0, 0, 1, 0, 0) = (0, 0, 0, 0, 0)

T (0, 0, 0, 1, 0) = (−1, 1, 0, 2, 0)

T (0, 0, 0, 0, 1) = (0, 1, 0, 1, 1)

are generators of the range space of T . In fact, if v1 = (1, 0, 1, 0, 1), v2 = (−1, 1, 0, 2, 0), v3 =

(0, 1, 0, 1, 1) then v1 , v2 , v3 generate T (R5 ). We now show that v1 , v2 , v3 are linearly inde-

pendent. Let α1 v1 +α2 v2 +α3 v3 = 0. Then α1 −α2 = 0, α2 +α3 = 0, α1 = 0 ⇒ α2 = α3 = 0.

Thus v1 , v2 , v3 are linearly independent over R ⇒ T (R5 ) is of dimension 3 with basis

v 1 , v2 , v3 .

Thus the null space is of dimension 2, because dim(null space) + dim(range space) =

dim(given vector space = R5 ) = 5. Since e1 = (1, 0, 0, 0, 0) and e3 = (0, 0, 1, 0, 0) belong to

the null space of T , and both are linearly independent over R, e1 , e3 is a basis of the null

space of T .

2

Question 2(b) Let A be a 3 × 3 real symmetric matrix with eigenvalues 0, 0, 5. If the

corresponding eigenvectors are (2, 0, 1), (2, 1, 1), (1, 0, −2) then find the matrix A.

2 2 1 0 0 0 0 0 0

Solution. Let P = 0 1 0 , then P−1 AP = 0 0 0, so A = P 0 0 0 P−1 .

1 1 −2 2 0 0 5 0 0 5

1

5

−1 5

−1

A simple calculation shows that P = 0 1 0 , therefore

1

5

0 − 25

2

−1 15

2 2 1 0 0 0 5

1 0 −2

A = 0 1 0 0 0 0 0 1 0 = 0 0 0

1

1 1 −2 0 0 5 5

0 − 25 −2 0 4

1 0 −2

Thus 0 0 0 is the required symmetric matrix with 0, 0, 5 as eigenvalues.

−2 0 4

−x1 + 3x2 + 5x3 − 5x4 − 2x5 = 0

2x1 + x2 − 2x3 + 3x4 − 4x5 = 17

in 5 unknowns, therefore the rank of the coefficient

1 −2 −3

matrix ≤ 3. Since −1 3 5 = 1(−6 − 5) + 2(2 − 10) + (−3)(−1 − 6) = −6, the rank

−2 1 −2

of the coefficient matrix is 3. Using Cramer’s rule we solve the system

−x1 + 3x2 + 5x3 = 5x4 + 2x5 (2)

2x1 + x2 − 2x3 = 17 − 3x4 + 4x5 (3)

3

−1 − 4x4 −2 −3

1

x1 = − 5x4 + 2x5 3 5

6

17 − 3x4 + 4x5 1 −2

1

= − [(−1 − 4x4 )(−11) − 3(5x4 + 2x5 − 51 + 9x4 − 12x5 ) + 2(−10x4 − 4x5 − 85 + 15x4 − 20x5 )]

6

1

= − [−6 + 44x4 − 42x4 + 10x4 + 30x5 − 48x5 ]

6

= 1 − 2x4 + 3x5

1 −1 − 4x 4 −3

1

x2 = − −1 5x4 + 2x5 5

6

2 17 − 3x4 + 4x5 −2

1

= − [−10x4 − 4x5 − 85 + 15x4 − 20x5 − 8 − 32x4 + 51 − 9x4 + 12x5 + 30x4 + 12x5 ]

6

1

= − [−42 − 6x4 ]

6

= 7 + x4

1 −2 −1 − 4x 4

1

x3 = − −1 3 5x4 + 2x5

6

2 1 17 − 3x4 + 4x5

1

= − [51 − 9x4 + 12x5 − 5x4 − 2x5 − 34 + 6x4 − 8x5 − 20x4 − 8x5 + 7 + 28x4 ]

6

1

= − [24 − 6x5 ]

6

= −4 + x5

Note that the vector space of solutions is of dimension 2.

Alternate Method.

7x2 + 8x3 = 17 + 7x4 + 8x5 adding 2×(2) and (3) (5)

6x3 = −24 + 6x5 using 7×(4) - (5) (6)

x2 = 7 + x4 from (4) and (6) (7)

x1 = 1 − 2x4 + 3x5 using (1), (6), (7) (8)

4

Question 2(d) Use Cayley-Hamilton theorem to find the inverse of the following matrix

0 1 2

A= 1 2 3

3 1 1

matrix.

x

−1 −2

−1 x − 2 −3 = 0

−3 −1 x − 1

x[x2 − 3x + 2 − 3] + 1[−x + 1 − 9] − 2[1 + 3x − 6] = 0

x3 − 3x2 − 8x + 2 = 0

1

A−1 = − (A2 − 3A − 8I)

2

7 4 5 0 3 6 8 0 0

1

= − 11 8 11 − 3 6 9 − 0 8 0

2

4 6 10 9 3 3 0 0 8

−1 1 −1

1

= − 8 −6 2

2

−5 3 −1

1 1 1

2

− 2 2

= −4 3 −1

5

2

− 23 12

Check A−1 A = I.

5

UPSC Civil Services Main 2003 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Let S be any non-empty subset of a vector space V over the field F . Show

that the set {a1 x1 + . . . + an xn | a1 , . . . , an ∈ F, x1 , . . . xn ∈ S, n ∈ N} is the subspace

generated by S.

w1 = a1 x1 + . . . ar xr , where a1 , . . . , ar ∈ F, x1 , . . . xr ∈ S and w2 = b1 y1 + . . . bs ys where

b1 , . . . bs ∈ F, y1 , . . . , ys ∈ S. Now αw1 +βw2 = c1 z1 +. . .+cr+s zr+s , where ci = αai , 1 ≤ i ≤

r, cj+r = βbj , 1 ≤ j ≤ s, and zi = xi , 1 ≤ i ≤ r, zj+r = yj , 1 ≤ j ≤ s. Clearly cj ∈ F, zj ∈ S

for 1 ≤ j ≤ r + s, showing that for any w1 , w2 ∈ W, α, β ∈ F, αw1 + βw2 ∈ W, moreover

W 6= ∅ as S ⊆ W and S = 6 ∅. Thus W is a subspace of V.

2 1 1

Question 1(b) If A = 0 1 0, then find the matrix represented by 2A10 − 10A9 +

1 1 2

14A − 6A − 3A + 15A − 21A4 + 9A3 + A − I.

8 7 6 5

2 − x 1 1

|A − xI| = 0 1−x 0 = (2 − x)2 (1 − x) − (1 − x) = 0

1 1 2 − x

1

Cayley-Hamilton theorem, we get A3 − 5A2 + 7A − 3I = 0. Now

= 2A7 [A3 − 5A2 + 7A − 3I] − 3A3 [A3 − 5A2 + 7A − 3I] + A − I

= 2A7 0 − 3A 3

0 + A−I

1 1 1

= A − I = 0 0 0

1 1 1

Question 2(a) Prove that the eigenvectors corresponding to distinct eigenvalues of a square

matrix are linearly independent.

of the square matrix A.

We will show that if any subset of the vectors x1 , x2 , . . . , xk is linearly dependent, then

we can find a smaller set that is also linearly dependent — but this leads to a contradiction

as the eigenvectors are all non-zero.

Suppose, without loss of generality, that x1 , x2 , . . . , xr are linearly dependent. Thus there

exist α1 , . . . αr ∈ R, not all zero, such that

α 1 x1 + . . . + α r x r = 0 (1)

subtracting, we have α2 (λ2 − λ1 )x2 + . . . + αr (λr − λ1 )xr = 0. Now αi 6= 0 ⇒ αi (λi − λ1 ) 6=

0, so not all αi (λi − λ1 ) can be zero, so we have a smaller set x2 , . . . , xr which is also

linearly dependent. This leads us to the contradiction mentioned above, hence the vectors

x1 , x2 , . . . , xk must be linearly independent.

Question 2(b) If H is a Hermitian matrix, then show that (H + iI)−1 (H − iI) is a unitary

matrix. Also show that every unitary matrix A can be written in this form provided 1 is not

an eigenvalue of A.

6 −2 2

Question 2(c) If A = −2

3 −1 then find a diagonal matrix D and a matrix B

2 −1 3

such that A = BDB where B0

0

denotes the transpose of B.

2

x1

Solution. Let Q(x1 , x2 , x3 ) = x1 x2 x3 A x2 be the quadratic form associated with

x3

A. Then

1 1 7 7 2

= 6[x1 − x2 + x3 ]2 + x22 + x23 − x2 x3

3 3 3 3 3

1 1 2 7 1 2 16 2

= 6[x1 − x2 + x3 ] + [x2 − x3 ] + x3

3 3 3 7 7

6 0 0

Let X1 = x1 − 31 x2 + 31 x3 , X2 = x2 − 71 x3 , X3 = x3 and D = 0 73 0 . Then

0 0 16 7

x1 x1 X1

0

x1 x2 x3 A x2 = x1 x2 x3 BDB x2 = X1 X2 X3 D X2

x3 x3 X3

1 − 31 31

X1 x1 x1 6 0 0

1

where X2 = 0 1 − 7

x2 = B x2 . Thus A = BDB where D = 0 73

0 0 0

16

X3 0 0 1 x3 x3 0 0 7

1 0 0

1

and B = − 3 1 0

1

3

− 17 1

Question 2(d) Reduce the quadratic form given below to canonical form and find its rank

and signature:

x2 + 4y 2 + 9z 2 + u2 − 12yx + 6zx − 4zy − 2xu − 6zu

Solution. Let

= (x − 6y + 3z − u)2 − 32y 2 + 32yz − 12yu

3

= (x − 6y + 3z − u)2 − 32(y 2 − yz + yu)

8

1 3 2

= (x − 6y + 3z − u) − 32(y − z + u) + 8z 2 + 18u2 − 24uz

2

2 4

1 3 2 3

= (x − 6y + 3z − u) − 32(y − z + u) + 8(z − u)2

2

2 4 2

3

Put

X = x − 6y + 3z − u

1 3

Y = y− z+ u

2 4

3

Z = z− u

2

U = u

2

√ that ∗Q(x,√y, z, u)∗ is transformed∗2 to X∗2 − 32Y

so 2

+ 8Z 2 . We now put X ∗ = X, Y ∗ =

32Y, Z = 8Z, U = U to get X − Y + Z ∗2 as the canonical form of Q(x, y, z, u).

Rank of Q(x, y, z, u) = 3 = rank of the associated matrix. Signature of Q(x, y, z, u) =

number of positive squares - number of negative squares = 2 − 1 = 1.

4

UPSC Civil Services Main 2004 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Let S be the space generated by the vectors {(0, 2, 6), (3, 1, 6), (4, −2, −2)}.

What is the dimension of S? Find a basis for S.

Solution. (0, 2, 6), (3, 1, 6) are linearly independent, because α(0, 2, 6) + β(3, 1, 6) = 0 ⇒

3β = 0, 2α + β = 0 ⇒ α = β = 0. Thus dim S ≥ 2.

If possible let (4, −2, −2) = α(0, 2, 6)+β(3, 1, 6), then 4 = 3β, −2 = 2α+β, −2 = 6α+6β

should be consistent. Clearly β = 43 , α = 12 (−2 − 43 ) = − 53 from the first two equations, and

these values satisfy the third. Thus (4, −2, −2) is a linear combination of (0, 2, 6) and (3, 1, 6).

Hence dim S = 2 and {(0, 2, 6), (3, 1, 6)} is a basis of S, being a maximal linearly inde-

pendent subset of a generating system.

mation. What is the dimension of the kernel? Find a basis for the kernel.

Solution.

= 3(αx1 + βx2 ) + αy1 + βy2 − (αz1 + βz2 )

= α(3x1 + y1 − z1 ) + β(3x2 + y2 − z2 )

= αf (x1 , y1 , z1 ) + βf (x2 , y2 , z2 )

Easy solution for this particular example. Clearly (1, 0, 0) does not belong to the

kernel, therefore the dimension of the kernel is ≤ 2. A simple look at f shows that (0, 1, 1)

and (1, −1, 2) belong to the kernel and are linearly independent, thus the dimension of the

kernel is 2 and {(0, 1, 1), (1, −1, 2)} is a basis for the kernel.

1

General solution. Clearly f : R3 −→ R is onto, thus the dimension of the range of

f is 1. From question 3(a) of 1998, dimension of nullity of f + dimension of range of f =

dimension of domain of f , so the dimension of the nullity of f = 2. Given this, we can pick

a basis for the kernel by looking at the given transformation.

Question 2(a) Show that T the linear transformation from R3 to R4 represented by the

matrix

1 3 0

0 1 −2

2 1 1

−1 1 2

is one to one. Find a basis for its image.

T(e2 ) = (3, 1, 1, 1) = v2

T(e3 ) = (0, −2, 1, 2) = v3

0, −a + b + 2c = 0 ⇒ a = b = c = 0. Thus T is one-one. Also {v1 , v2 , v3 } forms a basis for

the image, since {e1 , e2 , e3 } generates R3 , and {v1 , v2 , v3 } is a linearly independent set.

x + 3z = 5

−2x + 5y − z = 0

−x + 4y + z = 4

Solution. The first equation gives x = 5 − 3z, the second now gives 5y = z + 10 − 6z = 10 −

5z ⇒ y = 2−z. Putting these values in the third equation we get 4 = −5+3z+8−4z+z = 3,

hence the given system is inconsistent.

1 0 3 1 0 3 5

Alternative. Let A = −2 5 −1 be the coefficient matrix and B = −2 5 −1 0

−1 4 1 −1 4 1 4

be the augmented matrix, then it can be shown that rank A = 2 and rank B = 3, which

implies that the system is inconsistent. For consistency the ranks should be equal. This

procedure will be longer in this particular case.

2

1 1

Question 2(c) Find the characteristic polynomial of the matrix A = . Hence find

−1 3

A−1 and A6 .

x − 1 −1

Solution. The characteristic polynomial of A is given by |xI − A| = =

1 x − 3

(x − 1)(x − 3) + 1 = x2 − 4x + 4.

2

The Cayley-Hamilton theorem states that A satisfies its characteristic equation

i.e. A −

−3 1

4A + 4I = 0 ⇒ (A − 4I)A = A(A − 4I) = −4I. Thus A−1 = − A−4I = − 14 =

3

4 −1 −1

− 41

4

1 1

4 4

From A2 − 4A + 4I = 0 we get

A2 = 4A − 4I

A3 = 4A2 − 4A = 4(4A − 4I) − 4A = 12A − 16I

A6 = (12A − 16I)2 = 144A2 − 384A + 256I = 144(4A − 4I) − 384A + 256I

−128 192

= 192A − 320I =

−192 256

Question 2(d) Define a positive definite quadratic form. Reduce the quadratic form x21 +

x23 + 2x1 x2 + 2x2 x3 to canonical form. Is this quadratic form positive definite?

Solution. If Q(x1 , . . . , xn ) = ni=1 aij xi xj , aij = aji is a quadratic form in n variables with

P

j=1

aij ∈ R, thenPit is said to be positive definite if Q(α1 , . . . , αn ) > 0 whenever αi ∈ R, i =

1, . . . , n and i αi2 > 0.

Let the qiven be Q(x1 , x2 , x3 ). Then

= (x1 + x2 )2 + x23 + 2x2 x3 − x22

= (x1 + x2 )2 + (x2 + x3 )2 − 2x22

Let X1 = x1 + x2 , X2 = x2 , X3 = x2 + x3 i.e.

x1 1 −1 0 X1

x2 = 0 1 0 X2

x3 0 −1 1 X3

then Q(x1 , x2 , x3 ) is transformed to X12 − 2X22 + X32 . Since Q(x1 , x2 , x3 ) and the transformed

quadratic form assume the same values, Q(x1 , x2 , x3 ) is an√indefinite form. The canonical

form of Q(x1 , x2 , x3 ) is Z12 − Z22 + Z32 where Z1 = X1 , Z2 = 2X2 , Z3 = X3 .

3

UPSC Civil Services Main 2005 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Find the values of k for which the vectors (1, 1, 1, 1), (1, 3, −2, k), (2, 2k −

2, −k − 2, 3k − 1) and (3, k − 2, −3, 2k + 1) are linearly independent in R4 .

1 1 1 1

1

3 −2 k

2 2k − 2 −k − 2 3k − 1

3 k+2 −3 2k + 1

is non-singular.

Now

1 1 1 1 1 0 0 0

2 −3 k − 1

1 3 −2 k 1 2 −3 k − 1

2 2k − 2 −k − 2 3k − 1 = 2 2k − 4 −k − 4 3k − 3 = 2k − 4 −k − 4 3k − 3

k−1 −6 2k − 2

3 k + 2 −3 2k + 1 3 k − 1 −6 2k − 2

2

−3 k − 1

= 2k − 4 −k − 4 3k − 3 = (k − 5)[−9k + 9 + (k − 1)(k + 4)] 6= 0

k−5 0 0

independent when k 6= 1, 5.

1

Question 1(b) Let V be the vector space of polynomials in x of degree ≤ n over R. Prove

that the set {1, x, x2 , . . . , xn } is a basis for V. Extend this so that it becomes a basis for the

set of all polynomials in x.

where ai ∈ R, 0 ≤ i ≤ n, then we must have ai = 0 for every i because the non-zero

polynomial a0 +a1 x+. . .+an xn can have at most n roots in R whereas a0 +a1 x+. . .+an xn = 0

for every x ∈ R.

Every polynomial in x of degree ≤ n is clearly a linear combination of 1, x, x2 , . . . , xn

S

with coefficients from R. Thus {1, x, x2 , . . . , xn } is a basis for V.

We shall show that = {1, x, x2 , . . . , xn , xn+1 , . . .} is a basis for the space of all polyno-

mials.

S

(i) Linear Independence: Let {xi1 , . . . , xir } be a finite subset of . Let n = max{i1 , . . . , ir },

S

then {xi1 , . . . , xir } being a subset of the linearly independent set {1, x, x2 , . . . , xn } is linearly

independent, which shows the linear independence of .

S S

(ii) Let f be any polynomial. If degree of f is m, then f is a linear combination of

{1, x, x2 , . . . , xm }, which is a subset of . Thus is a basis of W, the space of all

polynomials over R.

3

Question 2(a) Let T bea linear transformation on R whose matrix relative to the standard

3

2 1 −1

basis of R is 1 2 2 . Find the matrix of T relative to the basis

3 3 4

B

= {(1, 1, 1), (1, 1, 0), (0, 1, 1)}.

of the given

basis be v1 , v2 , v3 . (T(v1 ), T(v2 ), T(v3 )) =

2 1 −1 1 1 0 2 3 0

1 2 2 1 1 1 = 5 3 4.

3 3 4 1 0 1 10 6 7

If (a, b, c) = αv1 + βv2 + γv3 , then α + β = a, α + β + γ = b, α + γ = c therefore

α = a − b + c, β = b − c, γ = b − a. Consequently

T(v1 ) = 7v1 − 5v2 + 3v3

T(v1 ) = 6v1 − 3v2 + 0v3

T(v1 ) = 3v1 − 3v2 + 4v3

B

7 6 3

This shows that the matrix of T with respect to given basis is −5 −3 −3

3 0 4

unitary matrix. Show that a unitary matrix A can be expressed in the above form provided

−1 is not an eigenvalue of A.

2

Question 2(c) Reduce the quadratic form

to a sum of squares. Also find the corresponding linear transformation, index and signature.

Solution.

2 2 1 1 2

= 6[x21 − x1 x2 + x1 x3 + x22 + x23 − x2 x3 ]

3 3 9 9 9

7 2 7 2 8

+ x2 + x3 − x2 x3

3 3 3

1 1 2 7 4 33

= 6[x1 − x2 + x3 ] + [x2 − x3 ]2 + x23

3 3 3 7 21

Put X1 = x1 − 13 x2 + 31 x3 , X2 = x2 − 47 x3 , X3 = x3 , so that

1

1 3 − 17

x1 X1

x2 = 0 1 4 X2 (1)

7

x3 0 0 1 X3

X .

21 3

Let

√1 Z1

X1 q 6

3

X2 = 7 Z2

q

X3 7

Z

11 3

then Q(x1 , x2 , x3 ) is transformed to Z12 + Z22 + Z32 , which is its canonical form. Thus

Q(x1 , x2 , x3 ) is positive definite. The Index of Q(x1 , x2 , x3 ) = Number of positive squares in

its canonical form = 3. The signature of Q(x1 , x2 , x3 ) = Number of positive squares - the

number of negative squares in its canonical form = 3.

The required linear transformation which transforms Q(x1 , x2 , x3 ) to sums of squares is

given by (1), and the linear transformation which transforms it to its canonical form is given

by

√1

1 1 0 0

x1 1 3 −7 6 q Z1

3

x2 = 0 1 4 0 7

0 Z2

7 q

x3 0 0 1 0 0 7 Z3

11

3

UPSC Civil Services Main 2006 - Mathematics

Linear Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Let V be a vector space of all 2 × 2 matrices over the field F. Prove that V

has dimension 4 by exhibiting a basis for V.

Solution. Let M1 = 10 00 , M2 = 00 10 , M3 = 01 00 , M4 = 00 01 . We will show that

{M1 , M2 , M3 , M4 } is a basis of V over F.

{M1 , M2 , M3 , M4 } generate V. Let A = ac db ∈ V. Then A = aM1 + bM2 + cM3 +

dM4 , where a, b, c, d ∈F. Thus {M1 , M2 , M3 , M4 } is a set of generators for V over F.

{M1 , M2 , M3 , M4 } are linearly independent over F. If aM1 + bM2 + cM3 + dM4 =

a b = 0 for a, b, c, d ∈F, then clearly a = b = c = d = 0, showing that {M , M , M , M }

c d 1 2 3 4

are linearly independent over F.

Hence {M1 , M2 , M3 , M4 } is a basis of V over F and dim V = 4.

1 3

Question 1(b) State the Cayley-Hamilton theorem and using it find the inverse of 2 4 .

Solution. Let A be an n × n matrix and let In be the n × n identity matrix. Then the n-

degree polynomial |xIn −A| is called the charateristic polynomial of A. The Cayley-Hamilton

theorem states that every matrix is a root of its characteristic polynomial:

if |xIn − A| = xn + a1 xn−1 + . . . + an

then An + a1 An−1 + . . . + an In = 0

−3

Let A = 12 34 . The characteristic equation of A is 0 = x−1

−2 x−4 = (x − 1)(x − 4) − 6 =

x2 − 5x − 2.

By the Cayley-Hamilton Theorem, A2 −5A−2I2 = 0 ⇒ A(A−5I2 ) = (A−5I2 )A = 2I2 .

−2 3

Thus A is invertible and A−1 = 21 (A − 5I2 ), so A−1 = 12 12 34 − 50 05 = 1 −21

2

1

B

Question 2(a) If T : R2 −→ R2 is defined by T(x, y) = (2x − 3y, x + y), compute the

matrix of T with respect to the basis = {(1, 2), (2, 3)}.

T(v2 ) = T(2, 3) = (−5, 5)

2a − b. Thus = 18v1 − 11v2 , T(v2 ) = 25v1 − 15v1 , so (v1 , v2 )T = (T(v

T(v1 ) 1 ), T(v2 ))=

B

18 25 18 25

(v1 , v2 ) . Thus the matrix of T with respect to the basis is

−11 −15 −11 −15

3 −2 0 −1

0 2 2 1

Question 2(b) Using elementary row operations, find the rank of A =

1 −2 −3 −2

0 1 2 1

1 2 6 3

0 1 0 0

A∼ 1 −2 −3 −2

0 1 2 1

Operation R3 − R1 gives

1 2 6 3

0 1 0 0

A∼

0 −4 −9 −5

0 1 2 1

Operations R3 + 4R2 , R4 − R2 ⇒

1 2 6 3

0 1 0 0

A∼

0 0 −9 −5

0 0 2 1

R4 + 92 R3 ⇒

1 2 6 3

0 1 0 0

A∼

0 0 −9 −5

0 0 0 − 91

Clearly |A| = 1 ⇒ rank A = 4.

2

Question 2(c) Investigate for what values of λ and µ the equations

x+y+z = 6

x + 2y + 3z = 10

x + 2y + λz = µ

have (1) no solution (2) a unique solution (3) infinitely many solutions.

Solution.

(2) The equations will have a unique solution

for all values of µ if the coefficient

1 1 1 1 1 1 1 0

0

matrix 1 2 3 is non-singular. i.e. 1 2 3 = 1 1 2 = λ − 1 − 2 6= 0 i.e. λ 6= 3.

1 2 λ 1 2 λ 1 1 λ − 1

Thus for λ 6= 3 and for all µ we have a unique solution which can be obtained by Cramer’s

rule or otherwise.

(1) If λ = 3, µ 6= 10 then the system is inconsistent and we have no solution.

(3) If λ = 3, µ = 10, the system will have infinitely many solutions obtained by solving

x + y = 6 − z, x + 2y = 10 − 3z ⇒ x = 2 + z, y = 4 − 2z, z is any real number.

Question 2(d) Find the quadratic form q(x, y) corresponding to the symmetric matrix

5 −3

A=

−3 8

5 −3 x

q(x, y) = x y

−3 8 y

= 5x2 − 6xy + 8y 2

6 8

= 5[x2 − xy + y 2 ]

5 5

3 2 31 2

= 5[(x − y) + y ]

5 25

6 (0, 0), (x, y) ∈ R2 . Thus q(x, y) is positive definite. In

Clearly q(x, y) > 0 for all (x, y) =

3

fact, q(x, y) = 0 ⇒ x − 5 y = 0, y = 0 ⇒ x = y = 0.

3

UPSC Civil Services Main 1979 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Define the centre of a group and show that a group of prime power has a

non-trivial centre. Hence or otherwise prove that a group of order p2 is abelian for every

prime p.

Solution. See theorem 2.11.2 Page 86 of Algebra by Herstein. See also the corollary to the

theorem.

Question 1(b) Show that any two cyclic groups of the same order are isomorphic.

Case 1. G1 and G2 are of infinite order. Define T : G1 −→ G2 by T (ar ) = br .

T (αβ) = T (ar+s ) = br+s = br bs = T (α)T (β).

3. T is onto is obvious.

Thus G1 ' G2 .

Case 2. ord(G1 ) = ord(G2 ) = n. Again define T (ar ) = br . Let α, β ∈ G1 ⇒ α =

a , β = as ⇒ T (αβ) = T (ar+s ). If r + s ≡ t mod n, then T (αβ) = T (at ) = bt = br+s =

r

T is onto is obvious, so G1 ' G2 .

1

√ √ √ √

Question 1(c) If Q is the field of rational numbers, prove that Q( 2, 3) = Q( 2 + 3),

where Q(a1 , a2 ) is the smallest subfield of the real field containing Q, a1 , a2 .

√ √ √ √ √ √

Solution. Clearly Q( √2, 3) √ ⊇ Q( 2 + 3). √ We√ shall prove

√ that√ the degree of√Q( 2, √ 3)

over Q =√degree√of Q( 2 + 3) = 4 ⇒ Q( 2, 3) = Q( √ 2 + 3). Let x = 2 +√ 3 so

that x − 2 = 3. Squaring both sides, we get x2 − 2 2x + 2 = 3 ⇒ x2 − 1 = 2 2x ⇒

x4 − 2x2 + 1 = 8x2 ⇒ x4 − 10x2 + 1 = 0. Now x4 − 10x2 + 1 is irreducible over Q — it has no

rational roots, as ±1 are not roots of the polynomial. Another way√to see √ this is to observe

that t2 −√10t + 1 √ has no rational roots,

√ so is irreducible.

√ Thus (Q( 2 + √3) : Q) = 24.

2

Now 3 6∈ Q( √2), because if 3 √ = a√ + b 2, a,√b ∈ Q, then 3 = √a √ + 2 2ab + 2b , which

would imply that 2 ∈ Q. Thus (Q( 2, 3) : Q( 2) = 2 ⇒ (Q( 2, 3) : Q) = 4.

2

UPSC Civil Services Main 1980 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Let R be a relation on a non-empty set A and for every y ∈ A let

By = {x | (x, y) ∈ R}

relation.

partition of A. Let a, b ∈ A. It is enough to show that if x ∈ Ba ∩ Bb , then Ba = Bb . Let

y ∈ Ba , then (y, a) ∈ R. As x ∈ Ba , (x, a) ∈ R, and x ∈ Bb , (x, b) ∈ R. Now by symmetry

and transitivity of R, (x, a) ∈ R ⇒ (a, x) ∈ R, and (y, a) ∈ R, (a, x) ∈ R ⇒ (y, x) ∈ R, and

now (x, b) ∈ R ⇒ (y, b) ∈ R ⇒ y ∈ Bb so Ba ⊆ Bb . SSimilarly Bb ⊆ Ba ⇒ Ba = Bb . So

Ba ∩ Bb = ∅ or Ba = Bb for all a, b ∈ A. Since A = a∈A Ba , the set {By | y ∈ A} is a

partition of A.

The converse is false. Let A = {a, b} and let R = {(a, b), (b, a)}. Then Ba = {b}, Bb =

{a}, which is a partition of A, but R is not an equivalence relation, as it is not reflexive or

transitive.

Question 1(b) Let I be the set of integers and let R be a relation on I × I defined by

(m, n)R(p, q) if and and only if mq = np. Show that R is an equivalence relation and

identify the partition generated by R.

Solution. The question as stated is not correct. (2, 3)R(0, 0) and (0, 0)R(1, 2), but

(2, 3)R(1, 2) is false, as 2 × 2 6= 3 × 1. However if we define I ∗ = I − {0}, and R is de-

fined as a relation over I × I ∗ with the above definition, then we can show that it is an

equivalence relation.

1

Reflexive: Clearly (m, n)R(m, n), as mn = nm. Symmetric: (m, n)R(p, q) ⇒ (p, q)R(m, n)

because mq = np ⇒ qm = pn. Transitive: If (m, n)R(p, q), then mq = np. If (p, q)R(r, s)

then ps = qr. Thus mqs = nps = nqr ⇒ ms = nr because q 6= 0. Thus (m, n)R(r, s).

Clearly B(m,n) = {(a, b) | mb/n ∈ I, a = mb/n}. In fact this is isomorphic to the set of

all rational numbers.

Question 1(c) Let A and B be any two non-empty sets. Show that the collections of all

mappings from A to B is a proper subset of the collection of all relations from A to B.

Solution. Let f be a function from A to B. Then the set {(a, f (a)) | a ∈ A} is a relation

from A to B. Thus the set of all functions is a subset of the set of all relations. It is a proper

subset if B has more than one element — the R = A × B is a relation from A to B, but is

not a function.

cx+d

, a, b, c, d ∈ R, ad−bc = 1

is a group. Examine whether the restricted set of transformations where a, b, c, d are integers

subject to the same constraint will form a group.

Solution. Let G be the set of all transformations, and H be the set of all transformations

where a, b, c, d are integers.

• G is nonempty. I : x −→ x is in G, with a = d = 1, b = c = 0.

a x+b

a1 x+b1 a2 x+b2 a2 c 1x+d1 +b2 (a1 a2 +b2 c1 )x+a2 b1 +b2 d1

• If T1 (x) = c1 x+d1

, T2 (x) = c2 x+d2

, then (T2 ◦ T1 )(x) = 1

a x+b

1

= (c2 a1 +d2 c1 )x+c2 b1 +d1 d2

c2 c 1x+d1 +d2

1 1

is an element of G because (a1 a2 + b2 c1 )(c2 b1 + d1 d2 ) − (a2 b1 + b2 d1 )(c2 a1 + d2 c1 ) =

(a1 d1 − b1 c1 )(a2 d2 − b2 c2 ) = 1.

• T ◦ I = I ◦ T = T for every T ∈ G.

cx+d

, T −1 (x) = −cx+a

dx−b

is the inverse of T . In fact T ◦ T −1 = I = T −1 ◦ T ,

this can be checked by substituting in the above formula and setting ad − bc = 1.

−1

Thus G is a group. H is clearlyclosed, and T ∈ H ⇒ T ∈ H, so H is a group.

a b

In fact G is isomorphic to { | ad − bc = 1, a, b, c, d ∈ R} and H is isomorphic to

c d

a b

{ | ad − bc = 1, a, b, c, d ∈ Z}.

c d

2

Question 2(b) Show that a group of order 15 is cyclic.

Solution. By Cauchy’s theorem, a group of order 15 has an element a such that the order of

a = o(a) = 5 and G has an element b such that o(b) = 3. Let H = hai , the group generated

by a, and K = hbi , the group generated by b. Then H ∩ K = {e}, because x ∈ H ⇒ x = e

or o(x) = 5, and x ∈ K ⇒ x = e or o(x) = 3.

Since G has a unique subgroup of order 5, and a unique subgroup of order 3, by Sylow’s

theorems, H, K are normal subgroups of G. Then ab = ba because aba−1 ∈ H, ba−1 b−1 ∈

K ⇒ aba−1 b−1 ∈ H ∩ K ⇒ ab = ba. Hence o(ab) = 15 ⇒ G is cyclic.

Here we have used the fact that if in a group G, o(x) = l, o(y) = m, (l, m) = 1, xy =

yx ⇒ o(xy) = lm.

Question 2(c) Let G be a finite group and F = {f : G −→ C} be the set of all complex

valued functions on G. If addition and multiplication in F are defined for f, g ∈ F by

X

(f g)(x) = f (xy −1 )g(y)

y∈G

Solution. F is a commutative group for addition is obvious, as the 0 function is the additive

identity, −f is the inverse of f , and + is commutative and associative in C.

Let e∗ : F −→ C be defined by e∗ (x) = 1 if x = e, 0 otherwise. Then

X

(f.e∗ )(x) = f (xy −1 )e∗ (y) = f (xe−1 ) = f (x)

y∈G

and X

(e∗ .f )(x) = e∗ (xy −1 )f (y) = f (x)

y∈G

P

3. (f + g)h = f h + gh

3

UPSC Civil Services Main 1981 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

f (A ∩ B) ⊆ f (A) ∩ f (B) and f (A ∪ B) = f (A) ∪ f (B). Examine whether f −1 (A ∩ B) is

properly contained in f −1 (A) ∩ f −1 (B).

Solution.

A ∩ B ⊆ A ⇒ f (A ∩ B) ⊆ f (A)

A ∩ B ⊆ B ⇒ f (A ∩ B) ⊆ f (B)

⇒ f (A ∩ B) ⊆ f (A) ∩ f (B)

A ⊆ A ∪ B ⇒ f (A) ⊆ f (A ∪ B)

B ⊆ A ∪ B ⇒ f (B) ⊆ f (A ∪ B)

⇒ f (A) ∪ f (B) ⊆ f (A ∪ B)

f (A) or f (x) ∈ f (B) ⇒ f (x) ∈ f (A) ∪ f (B). This shows that f (A ∪ B) ⊆ f (A) ∪ f (B),

hence f (A ∪ B) = f (A) ∪ f (B).

f −1 (A∩B) is not properly contained in f −1 (A)∩f −1 (B), as f −1 (A∩B) = f −1 (A)∩f −1 (B).

x ∈ f −1 (A) ∩ f −1 (B) ⇒ f (x) ∈ A, f (x) ∈ B ⇒ f (x) ∈ A ∩ B ⇒ x ∈ f −1 (A ∩ B).

Conversely x ∈ f −1 (A ∩ B) ⇒ f (x) ∈ A ∩ B ⇒ f (x) ∈ A, f (x) ∈ B ⇒ x ∈ f −1 (A), x ∈

f (B) ⇒ x ∈ f −1 (A) ∩ f −1 (B).

−1

Thus the two sides are contained in each other, hence must be equal.

1

Question 1(b) Define a binary relation on a set A. Give examples of relations which are

Solution.

1. Reflexive, symmetric but not transitive. A = {a, b, c}, and R = {(a, a), (b, b), (c, c), (a, b),

(b, c), (b, a), (c, b)}.

2. reflexive, transitive but not symmetric. A = {a, b, c}, and R = {(a, a), (b, b), (c, c), (a, b),

(b, c), (a, c)}.

3. symmetric, transitive but not reflexive. A = {a}, R = ∅. Note that if (x, y) ∈ R, then

by symmetry (y, x) ∈ R, and then by transitivity (x, x) ∈ R, (y, y) ∈ R. Hence for any

element a ∈ A, (a, a) 6∈ R ⇒ (a, b) 6∈ R, (b, a) 6∈ R for all b ∈ A.

1+2n

| m, n ∈ Z} forms a

group under multiplication.

Note that the set of non-zero rational numbers forms a group under multiplication. To

prove that G is its subgroup, all we need to show is that given x, y ∈ G, x−1 y ∈ G. Let y =

1+2m

1+2n

1+2q

, x = 1+2p , then x−1 y = 1+2(p+m+2mp)

1+2(q+n+2qn)

. x−1 y ∈ G because m + p + 2mp, q + n + 2qn ∈ Z.

Thus G is a group.

±1 0 0 ±1

{ , }

0 ±1 ±1 0

1 0 −1 0 1 0 −1 0

Solution. Let A1 = , A2 = , A3 = , A4 = , A5 =

0 1 0 1 0 −1 0 −1

0 1 0 −1 0 1 0 −1

, A6 = , A7 = , A8 = . These are all the elements of

1 0 1 0 −1 0 −1 0

the given set G.

2

2. G is closed w.r.t. multiplication, as shown in the following table.

A1 A2 A3 A4 A5 A6 A7 A8

A1 A1 A2 A3 A4 A5 A6 A7 A8

A2 A2 A1 A4 A3 A6 A5 A8 A7

A3 A3 A4 A1 A2 A7 A8 A5 A6

A4 A4 A3 A2 A1 A8 A7 A6 A5

A5 A5 A7 A6 A8 A1 A3 A2 A4

A6 A6 A8 A5 A7 A2 A4 A1 A3

A7 A7 A5 A8 A6 A3 A1 A4 A2

A8 A8 A6 A7 A5 A4 A2 A3 A1

is clear from the above table.

3

UPSC Civil Services Main 1982 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

November 1, 2008

1. If ∗ is any binary operation on any set S, then a ∗ a = a for all a ∈ S.

2. If ∗ is any commutative binary operation on any set S, then a ∗ (b ∗ c) = (b ∗ c) ∗ a.

3. If ∗ is any associative binary operation on any set S, then a ∗ (b ∗ c) = (b ∗ c) ∗ a.

4. Every binary operation defined on a set having exactly one element is both commutative

and associative.

5. A binary operation on a set S assigns at least one element to each ordered pair of

elements of S.

6. A binary operation on a set S assigns at most one element to each ordered pair of

elements of S.

7. A binary operation on a set S may associate more than one element to some ordered

pair of elements of S.

8. A binary operation on a set S may assign exactly one element to each ordered pair of

elements of S.

Solution.

1. If ∗ is any binary operation on any set S, then a ∗ a = a for all a ∈ S.

False. Take S = Z, ∗ = +, 2 + 2 6= 2.

2. If ∗ is any commutative binary operation on any set S, then a ∗ (b ∗ c) = (b ∗ c) ∗ a.

True.

1

3. If ∗ is any associative binary operation on any set S, then a ∗ (b ∗ c) = (b ∗ c) ∗ a.

1 0 1 0

False. Take S = 2×2 matrices, ∗ = multiplication. Let a = ,b = ,c =

0 0 1 0

1 0

. Then a ∗ (b ∗ c) = a, (b ∗ c) ∗ a = b 6= a.

0 1

4. Every binary operation defined on a set having exactly one element is both commutative

and associative.

True.

5. A binary operation on a set S assigns at least one element to each ordered pair of

elements of S.

True, since a binary operation assigns exactly one element to each ordered pair.

6. A binary operation on a set S assigns at most one element to each ordered pair of

elements of S.

True, same reason as above.

7. A binary operation on a set S may associate more than one element to some ordered

pair of elements of S.

False.

8. A binary operation on a set S may assign exactly one element to each ordered pair of

elements of S.

True.

Does it imply that if N is a normal subgroup, then ∀n ∈ N.∀g ∈ G.g −1 ng = n? Give an

example.

Solution. For the first part see Lemma 2.6.1, Page 50 of Algebra by Herstein.

The second statement is false. Let G = S3 , the symmetric group on 3 symbols. Let

N = {Identity permutation, (123), (132)}, then [G : N ] = Index of N in G = 2, so N is

normal in G.

Clearly (123)(12) = (13) and (12)(123) = (23). Thus (12)−1 (123)(12) = (12)(123)(12) =

(132) 6= (123).

Question 1(c) Let φ be a homomorphism of a group G into G with kernel K. Show that

Gφ = φ(G) is a group and there is an isomorphism of Gφ with G/K.

Solution. φ(e) = e ⇒ φ(G) 6= ∅. Let y1 , y2 ∈ φ(G). Then there exist x1 , x2 ∈ G, φ(x1 ) =

y1 , φ(x2 ) = y2 . Thus φ(x−1 −1 −1

1 x2 ) = φ(x1 ) φ(x2 ) = y1 y2 ∈ φ(G). Hence φ(G) is a subgroup

of G, so is a group.

See Question 1(a) year 1985 for the second part.

2

Question 1(d) Let I be an ideal in a ring R and let the coset of the element x ∈ R be

defined by x + I = {x + i | i ∈ I}. Then the distinct cosets form a partition of R, and if

addition and multiplication are defined by (x+I)+(y +I) = x+y +I, (x+I)(y +I) = xy +I,

then the cosets constitute a ring R/I in which the 0 element is 0 + I and the inverse of x + I

is (−x) + I.

Let j = x − y ∈ I. Let t ∈ x + I. Then t = x + i, i ∈ I. t = x + i = y + i − j ∈ y + I

because i − j ∈ I, so x + I ⊆ y + I. Since x − y ∈ I ⇒ y − x ∈ I ⇒ y + I ⊆ x + I, hence

x + I = y + I.

Conversely if t ∈ (x + I) ∩ (y + I), then t = x + i = y + j ⇒ x − y = j − i ∈ I. Hence

if x − y 6∈ I, then x + I ∩ y + I = ∅. Note that z ∈ z + I as 0 ∈ I. So the cosets form a

partition of R.

1. 0 + I = I ∈ R/I ⇒ R/I 6= ∅. R/I is clearly closed for addition. Note that addition is

well-defined i.e. if x + I = x1 + I, y + I = y1 + I ⇒ x + y + I = x1 + y1 + I.

2. (x + I) + (0 + I) = (x + 0) + I = x + I = (0 + I) + (x + I) for every x ∈ R, so 0 + I is

the additive identity of R/I.

R/I has an additive inverse.

4. Addition is associative and commutative, follows from the fact that this is so in R.

x1 + I, y + I = y1 + I ⇒ xy + I = x1 y1 + I, because x = x1 + i, y = y1 + j, i, j ∈ I, so

xy = (x1 + i)(y1 + j) = x1 y1 + x1 j + y1 i + ij ⇒ xy − x1 y1 ∈ I because x1 j + y1 i + ij ∈ I.

3

UPSC Civil Services Main 1983 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Show that the set I × I of integers is a commutative ring with unity with

respect to addition and multiplication defined as under:

(a, b) + (c, d) = (a + c, b + d)

(a, b)(c, d) = (ac, bd)

2. (0, 0) + (a, b) = (a, b) + (0, 0) = (a, b). Thus (0, 0) is the additive identity of I × I.

3. (a, b)+(−a, −b) = (0, 0) = (−a, −b)+(a, b), thus every element has an additive inverse.

an additive abelian group.

1

Question 1(b) Prove that the relation of isomorphism on the collection of groups is an

equivalence relation.

Question 1(c) Prove that a polynomial domain K[x] over a field K is a principal ideal

domain.

Solution. Let A be an ideal of K[x], A 6= h0i , A 6= K[x]. Let f (x) be a polynomial of the

smallest degree in A − {0}. Then f (x) is a generator of A, proved as follows: if g(x) 6= 0 be

any element of A. If f (x) does not divide g(x), then since K[x] is a Euclidean domain (proof

below), there exist polynomials q(x) and r(x) in K[x] such that g(x) = f (x)q(x) + r(x),

where deg(r(x)) < deg(f (x)). Since g(x), f (x) ∈ A and A is an ideal, it follows that

r(x) = g(x) − f (x)q(x) ∈ A, which contradicts the fact that f (x) has smallest degree in A.

Thus any element of A is divisible by f (x) ⇒ A = hf (x)i , the ideal generated by f (x) (note

that 0 = f (x)0).

Proof of the fact that K[x] is a Euclidean domain. The Euclidean function d : K[x] −

{0} −→ Z is the degree of a polynomial.

3. Given g(x) and f (x) 6= 0, we show that there exist q(x), r(x) such that g(x) =

q(x)f (x) + r(x) where r(x) = 0 or d(r(x)) < d(f (x)). The proof is by induction

on the degree n of g(x). If n < d(f (x)) = m, then q(x) = 0, r(x) Pn = g(x). Suppose

i

the result

P holds ifor all polynomials of degree < n. Let g(x) = i=0 bi x , bn 6= 0 and

f (x) = m a

i=0 i x . Let h(x) = g(x) − (bn /am )xn−m

f (x). Then degree of h(x) < n, and

therefore there exist q1 (x) and r(x) such that h(x) = q1 (x)f (x) + r(x) where r(x) = 0

or d(r(x)) < m. Hence g(x) = (q1 (x) + (bn /am )xn−m )f (x) + r(x).

2

UPSC Civil Services Main 1984 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

a − b ∈ S and ab ∈ S for all a, b ∈ S.

Solution. Let a, b ∈ S.

addition.

• 0 ∈ S.

• ab ∈ S, given.

subring of R, then a, b ∈ S ⇒ ab ∈ S, −b ∈ S ⇒ a + (−b) ∈ S ⇒ a − b ∈ S.

1

Question 1(c) Prove that for any two ideals A and B of a ring R, the product AB is an

ideal of R.

P

Solution. Note first that AB = { i∈F ai bi | ai ∈ A, bP

i ∈ B, F finite}.

AB isPa subring of R. If x ∈ AB, y ∈ AB, then x = i∈F ai bi | ai ∈ A, bi ∈ B, F finite

and y = i∈G ci di | ci ∈ A, di ∈ B, G finite. Clearly

X X

x−y = ai bi + ci di ∈ AB

i∈F i∈G

X X X

xy = ai bi ci di = xi yi where xi ∈ A, yi ∈ B

i∈F i∈G i∈H

Thus xy can be expressedP as above.

For any r ∈ R, xr = i∈F ai (bi r) = i∈F ai b0i , b0i ∈ B as B is an ideal. So xr ∈ AB.

P

Similarly rx ∈ AB. Thus AB is an ideal.

2

UPSC Civil Services Main 1985 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) State and prove the fundamental theorem of homomorphisms of groups.

kernel N . Then N is a normal subgroup of G, and G/N is isomorphic to G0 .

Proof: N = {x | x ∈ G, f (x) = e0 }, where e0 is the identity of G0 . Clearly N 6= ∅, as

e ∈ N . If x, y ∈ N , then f (x) = e0 , f (y) = e0 , so f (xy −1 ) = f (x)f (y)−1 = e0 ⇒ xy −1 ∈ N ⇒

N is a subgroup of G.

Now let g ∈ N , then for any x ∈ G, f (xgx−1 ) = f (x)e0 f (x)−1 = e0 ⇒ xgx−1 ∈ N , so N

is a normal subgroup of G.

Let φ : G/N −→ G0 defined by φ(gN ) = f (g).

• φ is well defined: We need to show that φ does not depend on the choice of coset

representative i.e. if g1 N = g2 N then φ(g1 N ) = φ(g2 N ). Now g1 N = g2 N ⇒ g2−1 g1 ∈

N ⇒ f (g2−1 g1 ) = e0 ⇒ f (g1 ) = f (g2 ) ⇒ φ(g1 N ) = φ(g2 N ).

• φ is onto: Let y ∈ G0 , then f being onto, there exists x ∈ G such that f (x) = y.

Clearly φ(xN ) = f (x) = y.

φ(g1 N ) 6= φ(g2 N ).

Note 1: If η : G −→ G/N is the natural homomorphism i.e. η(g) = gN , then f = φ ◦ η.

Note 2: If f is not assumed to be onto, we can only say that G/N ' f (G).

1

Question 1(b) Prove that the order of each subgroup of a finite group divides the order of

the group.

group.

4. The relation of isomorphism in the class of all groups is not an equivalence relation.

Solution.

m

True. (am )n = a . . .am} = a.a.a.a.

| . .{z |

mn

{z . . . .a} = a .

n times mn times

2. Every subgroup of an abelian group is not necessarily abelian.

False. If G is abelian, and H is a subgroup of G, then for any a, b ∈ H, a, b ∈ G ⇒

ab = ba ⇒ H is abelian.

group.

True. Let a ∈ G, then ax = a is solvable ⇒ there exists an element e ∈ G such that

ae = a. Now let b ∈ G, then there exists y ∈ G, ya = b. Then be = yae = ya = b.

Thus e is the right identity for G. For any a ∈ G, the equation ax = e is solvable, thus

a has a right inverse.

Since G has a right identity and a every element of G has a right inverse, G is a group.

Note that since G is a semigroup, it is already closed the semigroup operation, and the

operation is associative.

4. The relation of isomorphism in the class of all groups is not an equivalence relation.

False.

• The relation ' is reflexive, as the identity map is an isomorphism from any group

to itself.

2

• If σ : G −→ G0 is an isomorphism, then σ −1 : G0 −→ G is an isomorphism, so '

is symmetric.

• If G1 ' G2 and σ1 : G1 −→ G2 , and G2 ' G3 , σ2 : G2 −→ G3 , then σ2 ◦ σ1 :

G1 −→ G3 is also an isomorphism, so G1 ' G3 , thus ' is transistive.

True. The two groups are the cyclic group of order 6 and S3 , the symmetric group on

3 symbols.

If G is abelian, then G has x, an element of order 2 and y, an element of order 3. Since

xy = yx, o(xy) = 6 so G is cyclic.

If G is non-abelian, let a, b ∈ G, where o(a) = 2, o(b) = 3 (such elements exist because

of Cauchy’s theorem), then G = {e, a, b, b2 , ab, ba} ' S3 .

3

UPSC Civil Services Main 1986 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Prove that a map f : X −→ Y is injective if and only if f can be left

cancelled in the sense that f ◦ g = f ◦ h ⇒ g = h. Prove that f is surjective if and only if it

can be right cancelled in the sense that g ◦ f = h ◦ f ⇒ g = h.

Solution. Suppose that f is injective (1-1). Let f ◦ g = f ◦ h, where g, h are mappings

from Z −→ X, where Z can be arbitrary. We have to prove g(z) = h(z) ∀z ∈ Z. If

g(z) 6= h(z) then f (g(z)) 6= f (h(z)), since f is injective, consequently f ◦ g 6= f ◦ h. Thus

g(z) = h(z)∀z ∈ Z, so f ◦ g = f ◦ h ⇒ g = h.

Conversely let f ◦ g = f ◦ h ⇒ g = h. We have to prove that f is 1-1. Suppose it is

not 1-1. Then there exist x1 , x2 ∈ X, x1 6= x2 , f (x1 ) = f (x2 ). Define g, h : X −→ X by

g(x) = x1 , h(x) = x2 for all x ∈ X. Then ∀x ∈ X f (g(x)) = f (x1 ) = f (x2 ) = f (h(x)), so

f ◦ g = f ◦ h, but g 6= h, which is a contradiction. Thus f must be 1-1.

Suppose that f is onto. Given (g ◦ f )(x) = (h ◦ f )(x), we need to show that g(y) = h(y)

for all y ∈ Y . For any y ∈ Y , there exists x ∈ X such that f (x) = y, since f is onto. Thus

g(y) = (g ◦ f )(x) = (h ◦ f )(x) = h(y), showing that g ◦ f = h ◦ f ⇒ g = h.

Conversely let g ◦ f = h ◦ f ⇒ g = h. We have to prove that f is onto. Suppose it is

not onto. Then there exists y0 ∈ Y such that there is no x ∈ X such that f (x) = y0 . Let

g, h : Y −→ Z (for any Z) be defined by g(y) = h(y) if y 6= y0 , and g(y0 ) 6= h(y0 ). Clearly

(g ◦ f )(x) = (g ◦ f )(x) for every x ∈ X, but g 6= h, which is a contradiction. Thus f is onto.

and only if HK = KH.

Solution. See Lemma 2.5.1 Page 44 of Algebra by Herstein.

Solution. See lemma 3.2.2 Page 127 of Algebra by Herstein.

1

UPSC Civil Services Main 1987 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

and (g ◦ f )−1 = f −1 ◦ g −1 .

g(f (x1 )) 6= g(f (x2 )) because g is 1-1. Thus g ◦ f is 1-1.

For any z ∈ Z, ∃y ∈ Y such that g(y) = z, because g is onto. Now y ∈ Y , so there exists

x ∈ X such that f (x) = y, because f is onto. Thus for any z ∈ Z, we have determined

x ∈ X such that (g ◦ f )(x) = g(f (x)) = g(y) = z, thus g ◦ f is onto. Thus g ◦ f is invertible.

(g ◦ f ) ◦ (f −1 ◦ g −1 )(z) = (g ◦ f )f −1 (g −1 (z))

= (g ◦ f )(f −1 (g −1 (z)))

= g(f (f −1 (g(z))))

= g(g −1 (z)) = z ∀z ∈ Z

Thus (g ◦ f ) ◦ (f −1 ◦ g −1 ) = IZ .

= f −1 (g −1 (g(f (x))))

= f −1 (f (x)) = x ∀x ∈ X

Thus (f −1 ◦ g −1 ) ◦ (g ◦ f ) = IX , so f −1 ◦ g −1 = g ◦ f .

1

Question 1(c) If G is a finite group of order g and H is a subgroup of G of order h , then

prove that h is a factor of g.

2

UPSC Civil Services Main 1988 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) If H and K are normal subgroups of a group G such that H ∩ K = {e},

show that hk = kh for h ∈ H, k ∈ K.

hkh−1 k −1 ∈ K. Similarly, H is normal in G, so khk −1 ∈ H ⇒ hkh−1 k −1 ∈ H. Thus

hkh−1 k −1 ∈ H ∩ K ⇒ hkh−1 k −1 = e ⇒ hk = kh as required.

Question 1(b) Show that the set of even permutations on n symbols, n > 1, is a normal

subgroup of the symmetric group Sn and has order n!/2.

Solution. Let An be the set of all even permutations, then An 6= ∅ because the identity

permutation is in An . If α, β ∈ An , then αβ ∈ An because α even, β even ⇒ αβ is an even

permutation. α ∈ An ⇒ α−1 ∈ An , because α even ⇒ α−1 even. Thus An is a subgroup of

Sn .

An is a normal subgroup. Let β ∈ An , then αβα−1 is even for all α ∈ Sn , because if α is

odd, then α−1 is odd, so αβα−1 is even, and if α is even, then α−1 is even, so αβα−1 is even.

Thus αβα−1 ∈ An for all α ∈ Sn , β ∈ An , so An is normal in Sn .

Since n > 1, Sn has an odd permutation, for example (1, 2). Then Sn = An ∪ An (1, 2) is

the coset decomposition of Sn modulo An : Let α ∈ Sn . If α is even, then α ∈ An . If α is

odd, then β = α(1, 2) is even, so β ∈ An . Now α = β(1, 2) ∈ An (1, 2), so Sn = An ∪ An (1, 2).

It is obvious the An ∩ An (1, 2) = ∅, so the index of An in Sn is 2, i.e. the order of An = n!/2

as required.

1

Question 1(c) Prove that the set of inner automorphisms of a group G is a normal subgroup

of the group of automorphisms of G.

Solution. Let A be the group of automorphisms of G and let I be the set of all inner

automorphisms (which are automorphisms of the form αa (x) = axa−1 for a ∈ G).

then α(x) = x = exe−1 = αe (x), so α ∈ I.

Thus I is a subgroup of G.

Let σ ∈ A be arbitrary. Then for any αa ∈ I,

= σ(a)σ(σ −1 (x))σ(a−1 )

= σ(a)xσ(a)−1 = ασ(a) (x)

Question 2(a) Show that the numbers 0, 2, 4, 6, 8 with addition and multiplication modulo

10 form a field isomorphic to J5 the ring of integers modulo 5. Give the isomorphism explic-

itly.

+ 0 2 4 6 8

0 0 2 4 6 8

2 2 4 6 8 0

4 4 6 8 0 2

6 6 8 0 2 4

8 8 0 2 4 6

3. Clearly 2, 8 and 4, 6 are inverses of each other, so F is an abelian group under addition.

2

4. F is closed under multiplication. The multiplication table is

× 0 2 4 6 8

0 0 0 0 0 0

2 0 4 8 2 6

4 0 8 6 4 2

6 0 2 4 6 8

8 0 6 2 8 4

Isomorphism. For α ∈ Z, let [α] be the residue class of α modulo 5. Let σ : F −→ J5

be defined by σ(0) = [0], σ(6) = [1], σ(2) = σ(6 + 6) = σ(6) + σ(6) = [1] + [1] = [2], σ(4) =

σ(6 + 6 + 6 + 6) = 4σ(6) = [4], σ(8) = σ(6 + 6 + 6) = 3σ(6) = [3]. It can be easily checked

that σ is an isomorphism.

Question 2(b) If R is a commutative ring with identity and U is an ideal of R. Show that

U is maximal if and only if R/U is a field.

Solution. Let U be a maximal ideal. We already know that R/U is a commutative ring

with identity given by 1 + U . R/U will be a field if we show that every nonzero element

a + U ∈ R/U is invertible. Now a + U 6= U ⇒ a 6∈ U . Thus the ideal generated by a

and U contains U properly ⇒ ha, U i = R as U is maximal. Thus 1 = ab + u for some

b ∈ R, u ∈ U ⇒ (a + U )(b + U ) = 1 + U ⇒ a + U is invertible. Thus R/U is a field.

Conversely, let R/U be a field, and M ⊇ U be an ideal. We shall show that if M 6= U ,

then M = R ⇒ U is maximal. Now if M ) U ⇒ ∃a ∈ M, a 6= U . Thus a + U is a

non-zero element of R/U and therefore invertible i.e. there exists b + U ∈ R/U such that

(a + U )(b + U ) = 1 + U , or ab − 1 ∈ U ⇒ ab − 1 ∈ M . But ab ∈ M as a ∈ M , therefore

1 ∈ M ⇒ M = R ⇒ U is maximal.

See theorem 3.5.1 Page 138 of Algebra by Herstein for a different proof.

Question 2(c) Determine all the ideals U in J12 , the ring of integers modulo 12. In each

case describe J12 /U by finding a familiar ring with which the quotient ring is isomorphic.

Which of these ideals is maximal?

Solution. Let h[α]i denote the ideal generated by [α] ∈ J12 = {[0], [1], . . . [11]}. Since J12

3

is a principal ideal domain, we have to determine the distinct principal ideals. These are

h[0]i = {[0]}

h[1]i = J12 = h[11]i = h[5]i = h[7]i

h[2]i = {[0], [2], [4], [6], [8], [10]} = h[10]i

h[3]i = {[0], [3], [6], [9]} = h[9]i

h[4]i = {[0], [4], [8]} = h[8]i

h[6]i = {[0], [6]}

If U = h[2]i , then R/U = {[0]+U, [1]+U } ' Z/2Z, the isomorphism being σ([0]+U ) = 0

mod 2, σ([1] + U ) = 1 mod 2.

If U = h[3]i , then R/U = {[0] + U, [1] + U, [2] + U } ' Z/3Z, the isomorphism being

σ([0] + U ) = 0 mod 3, σ([1] + U ) = 1 mod 3, σ([2] + U ) = 2 mod 3.

If U = h[4]i , then R/U = {[0] + U, [1] + U, [2] + U, [3] + U } ' Z/4Z, the isomorphism

being σ([0] + U ) = 0 mod 4, σ([1] + U ) = 1 mod 4, σ([2] + U ) = 2 mod 4, σ([3] + U ) = 3

mod 4.

If U = h[6]i , then R/U = {[α] + U | 0 ≤ α ≤ 5} ' Z/6Z, the isomorphism being

σ([α] + U ) = α mod 6.

Clearly ideals generated by [2], [3] are maximal as R/h[2]i ' Z/2Z, and R/h[3]i ' Z/3Z,

which are fields.

Thus J12 has 6 ideals, namely h[0]i , h[1]i , h[2]i , h[3]i , h[4]i , h[4]i of which h[2]i , h[3]i are

maximal.

4

UPSC Civil Services Main 1989 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Let G be a finite group of order 2p, p a prime. Show that G has a normal

subgroup of order p.

Solution. Assume that G has a subgroup H of p elements. We shall show that H is normal

in G. Clearly [G : H] i.e. the index of H in G is 2. Let G = H ∪ Hx, where H and Hx

are distinct right cosets i.e. x 6∈ H. Consider xH, xH 6= H because x 6∈ H ⇒ xH ∩ H =

∅ ⇒ xH ⊆ Hx. Similarly, Hx ⊆ xH. Thus if x 6∈ H, then Hx = xH. If x ∈ H, then

xH = H = Hx. Thus xHx−1 = H for every x ∈ G, so H is normal in G.

Existence of H: State Cauchy’s theorem, or better yet, prove it (See theorem 2.11.3 Page

87 of Algebra by Herstein). Let a be an element of G of order p, then H, the subgroup

generated by a is of order p.

Question 1(b) Give an example of an infinite group in which every element is of finite

order.

Let G = ∪∞ n

n=1 Ωn = {α | α ∈ C, α = 1 for some n}. G is a subgroup of C − {0}. If

α ∈ G, β ∈ G, then αm = 1, β n = 1 for some m, n ⇒ (αβ)mn = 1 ⇒ αβ ∈ G. α ∈ G ⇒

α−1 ∈ G ∵ αn = 1 ⇒ α−n = 1. Clearly every element of G is of finite order. If G were finite,

2πi

say order M , then αM = 1 for every α ∈ G. But β = e M +1 ∈ G, β M 6= 1. Thus G is not

finite.

Another example: Consider the set of all infinite sequences of bits, under the operation

bitwise exclusive or: 0 ⊕ 0 = 1 ⊕ 1 = 0, 0 ⊕ 1 = 1 ⊕ 0 = 1. The identity element is the all 0

sequence, every element is its own inverse, and the operation is associative and commutative.

The group is clearly infinite, but every element has order 2.

1

Question 1(c) Let G be a group and let H be the smallest group containing elements of the

form x−1 y −1 xy, x, y ∈ G. Show that H is normal in G and the factor group G/H is abelian.

Solution. Let x ∈ G, h ∈ H, then x−1 hx = x−1 hxh−1 h. But x−1 hxh−1 ∈ H by definition,

therefore x−1 hx = x−1 hxh−1 h ∈ H ⇒ x−1 Hx = H for every x ∈ G. Thus H is normal in G.

Now in the factor group G/H, xH.yH = xyH. Since x−1 y −1 xyH = H as x−1 y −1 xy ∈ H,

it follows that xyH = yxH = yH.xH, thus G/H is abelian.

Question 2(a) If each element of a ring is idempotent, show that the ring is commutative.

Question 2(b) If a finite field F has q elements, then show that q = pn , where p is the

characteristic of F .

by φ(n) = ne. Then φ is a homomorphisms of rings as φ(m + n) = (m + n)e = me + ne =

φ(m) + φ(n) and φ(mn) = mne = mne2 = me.ne = φ(m)φ(n). Now ker φ = {n | φ(n) =

ne = 0 ⇔ p | n} = hpi , the ideal generated by p. Thus the field Z/pZ is isomorphic to a

subfield of F . In other words, F contains a subfield say Λ containing p elements. Now F is

finite, therefore F as a vector space over Λ is of finite dimension. Let (F : Λ) = n, and let

{v1 , . . . , vn } be a basis of F over Λ. Then F = {a1 v1 + . . . + an vn | a1 , . . . an ∈ Λ}. Since

each ai has p values, F has pn elements. Actually, F is isomorphic to Λn as a vector space.

Question 2(c) Let A be a ring and I be a two-sided ideal generated by the subset of all

elements of the form ab − ba, a, b ∈ A. Prove that the residue class ring A/I is commutative.

Solution.

⇔ ab + I = ba + I

⇔ ab − ba ∈ I which is true.

2

UPSC Civil Services Main 1990 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Let G be a group having no proper subgroup. Show that G should be a finite

group of order which is a prime, or unity.

Solution. See question 1(a), 1991. Once we have proved that G is finite, then we observe

that G has exactly one element if and only if the order of G is 1. If the order of G > 1, then

we show that it is a prime number.

Question 1(b) If the order of a group is 20, show that its 5-Sylow subgroup is a normal

subgroup. Also prove that a group of order 16 has a proper normal subgroup.

Solution. We know from various Sylow theorems that the numbr of 5-Sylow subgroups

≡ 1 mod 5 and is a divisor of 20 and therefore 4. Thus G, a group of order 20, has exactly

one Sylow subgroup of order 5, say H. Now aHa−1 for any a ∈ G is also a subgroup of order

5, therefore by uniqueness, aHa−1 = H. Thus H is normal in G.

For the second part, we prove a general theorem of which this is a special case.

Theorem. Let G be a group of order pr , p a prime, then G has a normal subgroup of

order ps for every s, 0 ≤ s < r.

Proof: By induction on r. If r = 1, then G is cyclic of prime order, hence the result

is true. Assume true for groups of order pm , m < r. Since G is a group of order pr , the

power of a prime, its center is non-trivial. Since the order of the center is pn , n ≥ 1, the

center has an element, say a, of order p (Cauchy’s theorem, Theorem 2.11.3 of Algebra

by Herstein). Let H = hai be the group generated by a. Since a ∈ center of G, H is a

normal subgroup of G. Now G/H is a group of order pr−1 . Using the induction hypothesis,

we see that G/H has a normal subgroup N ∗ of order ps−1 , 0 ≤ s − 1 < r − 1. Let η :

G −→ G/H be the natural homomorphism. Set N = η −1 (N ∗ ), we show that N is a

normal subgroup of G of order ps . η −1 (N ∗ ) 6= ∅. If x, y ∈ N , then η(x), η(y) ∈ N ∗ ,

1

then η(x)(η(y))−1 ∈ N ∗ ⇒ η(xy −1 ) ∈ N ∗ ⇒ xy −1 ∈ N , so N is a subgroup of G. For

x ∈ N, a ∈ G, η(x) ∈ N ∗ ⇒ η(a)η(x)η(a)−1 = η(axa−1 ) ∈ N ∗ as N ∗ is a normal subgroup

of G/H. Thus axa−1 ∈ N , so N is a normal subgroup of G. N ⊇ H is immediate as

∀h ∈ H.η(h) = H, the identity element of G/H. Consider η : N −→ N ∗ , then η is a

homomorphism with kernel H ⇒ N/H ' N ∗ ⇒ o(N ) = o(N ∗ )o(H) = ps .

Now for a group of order 16, p = 2, r = 4, and the above theorem shows that it has

normal groups of order 2, 4, and 8.

Question 1(c) If C is the center of a group G, and G/C is cyclic, prove that G is abelian.

Question 2(a) Show that the set of Gaussian integers is a Euclidean ring. Find an HCF

of 5i and 3 + i.

complex numbers.

An integral domain R is said to be a Euclidean domain if there exists a function N :

R → Z (the ring of integers) such that

1. N (a) ≥ 0

N (r) < N (b).

For Z[i], let N (α) = N (a + ib) = a2 + b2 . Clearly

1. N (α) ≥ 0 for every α ∈ Z[i].

2. N (αβ) ≥ N (α) for all α, β ∈ Z[i] because N (αβ) = N (α)N (β) and N (β) ≥ 1 if β 6= 0.

a+ib

3. Let α = a + ib, β = m + ni, β 6= 0. Then f racαβ = m+ni = x + iy, x ∈ Q, y ∈ Q.

Determine p, q ∈ Z such that |x−p| ≤ 2 , |y−q| ≤ 2 (take p = [x] if x = [x]+θ, 0 ≤ θ ≤ 21

1 1

α

Now β

− (p + qi) = x − p + i(y − q). Thus N ( αβ − (p + qi)) = (x − p)2 + (y − q)2 < 1.

Now α = (p + qi)(m + ni) + γ where γ = (x − p + i(y − q))(m + ni). Clearly γ ∈ Z[i]

and N (γ) = N (β)((x − p)2 + (y − q)2 ) < N (β), which is what we wanted to prove.

Thus Z[i] is a Euclidean ring.

Now 5i = (3 + i)(2i) + (2 − i), and 3 + i = (2 − i)(1 + i) ⇒ (5i, 3 + i) = 2 − i.

Note: In this case writing the division algorithm was easy, otherwise N (5) = 25, N (3 +

i) = 10 ⇒ GCD is a factor of 5 = (25, 10). Thus the GCD can be 1, 2 − i, 2 + i, 5. We rule

out 2 + i, 5 by showing that 2 + i 6 | 3 + i. 2 − i then fits the bill.

2

Question 2(b) If K is a finite extension of a field F of degree n, prove taht any element

of K is algebraic over F with degree m where m divides n.

F , because (K : F ) = degree of K over F = n. Thus there exist aP 0 , a1 , . . . , an ∈ F , not all

0, such that a0 + a1 α + . . . + an αn = 0 ⇒ α is a root of f (x) = ni=0 ai xi ∈ F [x] ⇒ α is

algebraic over F .

Let p(x) be the minimal polynomial of α over F , deg p(x) = m. Then (F (α) : F ) = m

— first of all 1, α, . . . , αm−1 are linearly independent over F , because otherwise α will be

the root of a non-zero polynomial of degree less than m. We know that α algebraic over F

implies F (α) = F [α] as F (α) is the smallest field containing F and α, and F [α] is a field1 .

Now any element of F [α] is a linear combination of 1, α, . . . αm−1 . Take f (α) again.

f (x) = q(x)p(x) + r(x) where r(x) = 0 or deg r(x) < m. Thus f (α) = r(α), hence (F (α) :

F ) = m. We also know that (K : F ) = (K : F (α))(F (α) : F ) (See 2(c), 1993 — if

{v1 , . . . , vr } is a basis of K over F (α), and {w1 , . . . , wm } is a basis of F (α) over F , then

{vi wj | 1 ≤ i ≤ r, 1 ≤ j ≤ m} is a basis for K over F ).

Thus m divides n.

p √

Question√ 2(c) Find the minimum polynomial over Q (the field of rationals) of 5 − 2

and i + 3.

√

Solution. Let x = i + 3, then (x − i)2 = 3 ⇒ x2 − 2ix + i2 = 3 ⇒ x2 − 4 = 2ix ⇒

(x2 − 4)2 = −4x2 ⇒ x4 − 4x2 + 16 = 0. We shall show that x4 − 4x2 + 16 is irreducible over

Q. If possible, let x4 − 4x2 + 16 = (x2 + ax + b)(x2 + cx + d), then a + c = 0, ac + b + d =

−4, ad + bc = 0, db = 16. Using a + c = 0, ac + bd = 0, we get c(b − d) = 0. If c = 0, then

a = 0, so b + d = −4, bd = 16 so b, d are roots of x2 + 4x + 16, thus b, d are not real numbers.

Thus b = d ⇒ b = d = ±4 ⇒ ac = −12 or ac = 0 (not possible). Thus a, c are roots of

x2 − 12 = 0, thus are not rationals. Hence x4 − 4x2 + 16 is not reducible.

A simpler way of seeing the above is that t2 −4t+16 has non-real roots, hence is irreducible

4

over Q, so xp − 4x2 + 16 is not reducible over Q.

√ 2

√ 4

p Let√x = 5 − 2. Then x − 5 = − 2 ⇒ p x −√ 10x2 + 23 = 0 is a polynomial satisfied by

5 − 2. It is the minimal polynomial of 5 − 2 because it is irreducible over Q, since

2

t − 10t + 23 has non real p roots.

√ √

Hence the degree of 5 − 2 and i + 3 is 4.

1

Let f (α) = a0 + a1 α + . . . + ar αr be any non-zero element of F [α]. Then the polynomial p(x) 6 | f (x) ⇒

(f (x), p(x)) = 1 ⇒ there exist b(x), c(x) ∈ F [x] such that p(x)b(x) + f (x)c(x) = 1 ⇒ f (α)c(α) = 1 ⇒ f (α)

is invertible

3

UPSC Civil Services Main 1991 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) If the group G has no non-trivial subgroups, show that G must be finite of

prime order.

G is cyclic: Let a ∈ G, a 6= e. Let H be the cyclic group generated by a. Then H 6= {e},

therefore H = G, so G is cyclic.

G has finite order: If order of G is infinite, then the group K generated by a2 is a

non-trivial subgroup of G, because K 6= {e}, K 6= G as a 6∈ K — note that a ∈ K, a = (a2 )m

for some m shows that a is of finite order ⇒ G is of finite order. This is a contradiction,

hence order of G is finite.

The order of G is a prime number: If the order is pq, p > 1, q > 1, then order of ap

or equivalently the order of the group generated by ap is q ⇒ G has a nontrivial subgroup,

which is a contradiction. Hence order of G is a prime number.

Solution. We first prove that if G is a group with centre C such that G/C is cyclic, then

G is abelian. Let G/C be generated by the coset aC. Let x, y ∈ G, then xC = (aC)r

and yC = (aC)s for some integers r, s. This means that x ∈ ar C, y ∈ as C and therefore

x = ar c1 , y = as c2 , c1 , c2 ∈ C. Now xy = ar c1 as c2 = ar as c1 c2 since c1 ∈ C, so it commutes

with every element of G. Similarly, c2 ∈ C so it commutes with ar , so

xy = ar+s c1 c2 = as+r c2 c1 = as c2 ar c1 = yx

Hence G is abelian.

Now we prove that a group G of order p2 , p prime, is abelian. In particular, a group of

order 9 will be abelian. Let C be the center of G. Then C is of order p or p2 as the center

of a prime power group is non-trivial (Theorem 2.11.2 page 86 of Algebra by Herstein).

1

If the order of C is p2 , and G = C so G is abelian.

If order of C is p, then G/C is of order p and therefore is a cyclic group. Thus G must

be abelian as shown above. In either case G is abelian.

Question 1(c) If the characteristic of an integral domain D is finite, show that it is a prime

number.

a ∈ D, a 6= 0. Then 0 = ma2 = pa.qa. But D is an integral domain, therefore either pa = 0

or qa = 0. Suppose without loss of generality that pa = 0. If b ∈ D is arbitrary, then

0 = (pa)b = (pb)a. But a 6= 0, therefore pb = 0 ⇒ m is not the smallest positive integer such

that ma = 0 for every a ∈ D. Thus the assumption m has a proper factorization is wrong,

hence m is a prime number.

Question 2(a) Find the greatest common divisor (GCD) in J[i], the ring of Gaussian in-

tegers of (i) 3 + 4i and 4 − 3i (ii) 11 + 7i and 18 − i.

Solution. (i) 4 − 3i = (−i)(3 + 4i), and −i is a unit in J[i] as i(−i) = 1. It follows that

4 − 3i and 3 + 4i are associates of each other. Thus the GCD of 4 − 3i and 3 + 4i can be

taken to be either of them.

(ii) N (11 + 7i) = (11 + 7i)(11 − 7i) = 170, N (18 − i) = 325. Since (170, 325) = 5, we can

find integers x, y such that 170x + 325y = 5, or

showing that if α divides 11 + 7i, 18 − i in J[i], then α divides 5. Therefore the GCD of

11 + 7i, 18 − i is a factor of 5, i.e. 1, 2 − i, 2 + i, 5.

Now 11+7i

2+i

= (11+7i)(2−i)

5

= 295

+ 35 i. Thus 2 + i 6 | 11 + 7i.

11+7i

2−i

= (11+7i)(2+i)

5

= 3 + 5i. Thus 2 − i | 11 + 7i. 18−i 2−i

= (18−i)(2+i)

5

= 37

5

+ 16

5

i, so

2 − i 6 | 18 − i.

Thus the GCD of 11 + 7i and 18 − i is 1.

Note: We could have got this by Euclid’s Algorithm also.

11 + 7i = (7 − 8i)i + 3 N (3) < N (7 − 8i)

7 − 8i = (2 − 3i)3 + (1 + i) N (1 + i) < N (3)

3 = (1 + i)(1 − i) + 1 N (1) < N (1 + i)

2

Question 2(b) Show that every maximal ideal of a commutative ring R with unit element

is a prime ideal.

a 6∈ M i.e. a 6≡ 0 mod M . We shall show that b ≡ 0 mod M , proving that M is a

prime ideal. Consider hM, ai , the ideal generated by M and a. Clearly M ⊆ hM, ai and

M 6= hM, ai as a 6∈ M , therefore hM, ai = R as M is maximal. Thus e ∈ hM, ai , where

e is the unit element of R. Thus e = m + xa where m ∈ M, x ∈ R, so b = mb + xab.

mb ∈ M, xab ∈ M because ab ∈ M . Hence mb + xab = b ∈ M , which was to be proved,

showing that M is a prime ideal.

Remark. The converse of the above statement is not true. Let R = Z[x], P = h2i , the

ideal generated by 2, then P is prime but not maximal — in fact h2i ( h2, xi ( R.

Question 2(c) The field K is an extension of the field F . If α, β ∈ K are both algebraic

over F , show that α ± β, αβ, α/β (if β 6= 0) are all algebraic over F .

Solution. Let p(x) be the minimal polynomial of α over F , then F [x]/hp(x)i ' F [α], the

homomorphism from F [x] to F [α] being f (x) = f (α) with kernel hp(x)i . Thus F [α] = F (α)

(the smallest field containing F and α in K). If deg p(x) = n, then 1, α, . . . , αn−1 are linearly

independent over F and generate F (α). Hence (F (α) : F ) = n ⇒ if γ ∈ F (α), γ is algebraic

over F as 1, γ, . . . , γ n are linearly dependent over F , so γ is a root of a polynomial of degree

≤ n.

Now β being algebraic over F , is algebraic over F (α) ⇒ F (α.β) is a finite extension of

F (α), and (F (α, β) : F (α)) = degree of the minimal polynomial of β over F (α) ≤ degree

of the minimal polynomial of β over F . Since (F (α, β) : F ) = (F (α, β) : F (α))(F (α) : F )

(see question 2(c) of 1993), it follows that F (α, β) is an algebraic extension over F . In fact

if (F (α, β) : F ) = m and ζ ∈ F (α, β), then 1, ζ, ζ 2 , . . . , ζ m are linearly dependent, so ζ is a

root of a polynomial of degree ≤ m. Thus α ± β, αβ, α/β, being elements of F (α, β) are all

algebraic over F .

3

UPSC Civil Services Main 1992 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) If H is a cyclic normal subgroup of a group G, then show that every subgroup

of H is a normal subgroup of G.

Solution. Let K be a normal subgroup of H. Let H = hai, and let K = har i, where r is

the least positive integer such that ar ∈ K.

Then k ∈ K ⇒ k = (ar )m for some m.

| {z }

r times

Now H is normal in G, so gam g −1 ∈ H ⇒ gam g −1 = at for some t. Thus gkg −1 = (ar )t =

(a ) ⇒ gkg −1 ∈ K. Thus K is normal in G.

r t

Note: Cyclic subgroups need not be normal. G = S3 , H = {I, (1, 2)} is cyclic but not

normal in S3 .

Solution. o(G) = 3 · 2 · 5.

n5 = number of Sylow groups of order 5 is 1 or 6 because n5 ≡ 1 mod 5 and n5 | 30.

n3 = number of Sylow groups of order 3 is 1 or 10 because n3 ≡ 1 mod 3 and n3 | 30.

If G has 6 Sylow groups of order 5, then G has 24 elements of order 5, because if H and

K are two subgroups of order 5, then H ∩ K {e} when H 6= K. Thus each Sylow subgroup

of order 5 gives rise to 4 distinct elements of order 5.

If G has 10 subgroups of order 3, then G has 20 elements of order 3. Thus either n3 = 1

or n5 = 1. So G has a unique Sylow subgroup of order 3 or 5, which has to be a normal

subgroup of G. Thus G is not simple.

Note that n5 > 1, n3 > 1 means that G must have at least 45 elements.

1

Question 1(c) Let p be the smallest prime factor of the order of a group G, then prove that

any subgroup of index p is normal in G.

Solution. Let G/H = {x1 H, x2 H, . . . , xp H}. For any x ∈ G consider the mapping πx :

G/H −→ G/H defined by πx (xj H) = xxj H = xk H for some k, 1 ≤ k ≤ p. Clearly πx is

one-one and therefore gives rise to a permutation on p symbols. Let Sp denote the symmetric

group on p symbold. Define φ : G −→ Sp by φ(x) = πx . Then φ is a homomorphism as

πxy (xj H) = xy(xj (H)) = x(yxj H) = πx (πy (xj H) ⇒ φ(xy) = φ(x)φ(y)

Thus by the fundamental theorem of homomorphisms G/K is isomorphic to a subgroup of

Sp , where K is the kernel of φ.

K ⊆ H. Proof: Let x ∈ K. Then πx is the identity permutation in Sp i.e. πx (xj H) =

xxj H = xj H for every j, 1 ≤ j ≤ p. Let xr be such that xr H = H, such an xr exists then

xH = xxr H = xr H = H ⇒ x ∈ H. Thus K ⊆ H.

(G : K) = (G : H)(H : K) — This follows immediately from (G : K) = o(G)/o(K).

(Note that all groups are of finite order here. This statement also holds for groups of infinite

order).

Let (H : K) = r. Then (G : K) = pr and therefore pr | p!, because G/K is isomorphic

to a subgroup of Sp , so order of G/K = (G : K) divides o(Sp ) = p!. Thus r | (p − 1)!. But

r divides o(G) also, because K is a subgroup of H which is a subgroup of G. Consequently

r divides ((p − 1)!, o(G)). But ((p − 1)!, o(G)) = 1 as p is the smallest prime factor of o(G).

Thus r = 1 ⇒ K = H. Hence H being a kernel of a homomorphism φ : G −→ Sp is a

normal subgroup of G.

Remark: We don’t need it in the above proof, but it is worth noticing that

K = ∩a∈G aHa−1

For x ∈ K ⇔ xxj H = xj H ∀j.1 ≤ j ≤ p

⇔ x ∈ xj Hx−1

j ∀j.1 ≤ j ≤ p

−1

⇔ x ∈ aHa ∀a ∈ G

(Note that aHa = xj Hx−1

−1

j if a = xj H).

Proof of (G : K) = (G : H)(H : K). Let G/H = {x1 H, x2 H, . . . , xn H} and H/K =

{y1 K, . . . , ym K}. Then we will show that G/K = {xi yj K | 1 ≤ i ≤ m, 1 ≤ j ≤ n}.

xi y j ≡ xk y l mod K ⇒ yl−1 x−1

k xi y j ∈ K

⇒ yl−1 x−1

k xi y j ∈ H

−1

⇒ xk xi ∈ H (∵ yl , yj ∈ H)

⇒ x k H = xi H ⇒ k = i

⇒ yl−1 yj ∈ K

⇒ yl K = yj K ⇒ l = k

Given x ∈ G, xH = xj H for some j, 1 ≤ j ≤ n. Since x−1 −1

j x ∈ H, xj xK = yk K for some

k, 1 ≤ k ≤ m. Therefore xK = xj yk K, so {xi yj K | 1 ≤ i ≤ m, 1 ≤ j ≤ n} is a complete

system of representation of cosets of G/K. This implies (G : K) = mn = (G : H)(H : K).

2

Question 2(a) If R is a unique factorization domain, then prove that any f ∈ R[x] is an

irreducible element of R[x] if and only if either f is an irreducible element of R or f is an

irreducible polynomial in R[x].

Solution. We first observe that units of R and R[x] are the same — let f, g ∈ R[x] be such

that f g = 1 then deg f + deg g = 0 ⇒ deg f = 0, deg g = 0 ⇒ f, g ∈ R and both are units

in R.

If f is an irreducible element of R, then f is an irreducible element of R[x] — if f = gh

then deg g + deg h = 0 ⇒ deg g = 0, deg h = 0 ⇒ g, h ∈ R, but since f is irreducible in R,

either g is a unit in R or f is a unit in R, and therefore in R[x].

Conversely, if f is an irreducible element in R[x] and f ∈ R, then f has to be irreducible

in R also, because if f = gh is a proper factorization of f ∈ R, then this would be a proper

factorization of f in R[x] also, because units of R and R[x] are the same, so g, h cannot be

units in R[x].

Now let f ∈ R[x] be an irreducible element of R[x] and f 6∈ R, then f is an irreducible

polynomial. But an irreducible polynomial need not be an irreducible element of R[x]. For

example, 2x2 + 2 is an irreducible polynomial in Z[x] but is not an irreducible element. Thus

the correct question would be — f ∈ R[x] is an irreducible element of R[x] if and only if

either f is an irreducible element of R or f is an irreducible primitive polynomial in R[x].

Question 2(b) Prove that the polynomials x2 + 1 and x2 + x + 4 are irreducible over F , the

field of integers modulo 11. Prove that F [x]/hx2 + 1i and F [x]/hx2 + x + 4i are isomorphic

fields each having 121 elements.

If possible let x2 + 1 ≡ (x + a)(x + b) mod 11 where a, b are integers. This implies that

a + b ≡ 0 mod 11, ab ≡ 1 mod 11 ⇒ a2 ≡ −1 mod 11, which is not possible, since the

only quadratic residues of 11 are 0, 1, 4, 9, 5 and 3. Thus x2 + 1 has no linear factors modulo

11 i.e. x2 + 1 is irreducible modulo 11.

Let p(x) be an irreducible polynomial over a field F and α be a root of p(x) in some

extension of F . Then the field F [x]/hp(x)i is isomorphic to F [α]. Proof: Consider the

mapping ρ : F [x] −→ F [α] defined by ρ(f (x)) = f (α). It can be easily seen that ρ is a

homomorphism, onto with kernel hp(x)i . If deg p(x) = n, then (F [α] : F ) = n. Clearly

1, α, α2 , . . . , αn−1 are independent over F , otherwise α will be the root of a polynomial of

degree < n. Let β ∈ F (α) = F [α], then β = a0 +a1 α+. . . ar αr , let f (x) = a0 +a1 x+. . .+ar xr ,

then there exist q(x), s(x) such that f (x) = q(x)p(x) + r(x) where s(x) = 0 or deg s(x) <

deg p(x). Thus β = f (α) = s(α) as p(α) = 0, showing that β is a linear combination of

1, α, α2 , . . . , αn−1 .

In case p(x) = x2 + 1, F = field of integers modulo 11, then F [x]/hx2 + 1i ' F [α] with

α2 + 1 = 0. Now (F [α] = F (α) : F ) = 2 with 1, α as its basis. Thus F (α) = {a0 + a1 α |

a0 , a1 ∈ F }. Clearly F (α) has 121 elements. Similarly, F [x]/hx2 + x + 4i has 121 elements.

Consider the mapping σ : F [x] −→ F [x] defined by σ(x) = x − 5 and σ(a) = a for a ∈ F .

It is obvious that σ is an isomorphism. Now σ(x2 + 1) = (x − 5)2 + 1 = x2 − 10x + 26 ≡

3

x2 + x + 4 mod 11. This shows that σ gives rise to a map from K1 = F [x]/hx2 + 1i to

K2 = F [x]/hx2 + x + 4i . Any typical element of K1 is of the form a0 + a1 x + hx2 + 1i where

a0 , a1 ∈ F . Then σ(a0 + a1 x + hx2 + 1i ) = a0 + a1 (x − 5) + hx2 + x + 4i .

We now check that σ is an isomorphism. We write αx + β = α + βx + hx2 + 1i . Then

= (α + γ)x + β + δ − 5((α + γ) + hx2 + x + 4i

= σ(αx + β) + σ(γx + δ)

= σ((αδ + βγ)x + βδ − αγ) as αγx2 ≡ −αγ mod x2 + 1

= (αδ + βγ)x − 5(αδ + βγ) + βδ − αγ + hx2 + x + 4i

Now

= αγx2 + αδx − 5αγx + βγx + βδ − 5βγ − 4αγx − 5αδ + 25γα + hx2 + x + 4i

= αγ(−x − 4) + αδx − 5αγx + βγx + βδ − 5βγ − 4αγx − 5αδ + 3γα + hx2 + x + 4i

= x[−αγ + αδ + βγ − 5αγ − 5αγ] + βδ − 5βγ − 5αδ − αγ + hx2 + x + 4i

≡ x[αδ + βγ] + βδ − 5βγ − 5αδ − αγ + hx2 + x + 4i mod 11

Thus σ((αx + β)(γx + δ)) = σ((αx + β))σ((γx + δ)) showing that σ is a homomorphism.

σ is 1 − 1: The kernel of σ is an ideal of K1 , but K1 is a field, therefore the only ideals

of K1 are the trivial ideal h0i and K1 . Since σ is not a zero map, it follows that the kernel

of σ is h0i , thus σ is 1 − 1.

σ is onto: Since K1 and K2 have 121 elements each, and sigma is one-one, σ(K1 ) = K2 .

Thus σ is an isomorphism from K1 to K2 .

Question 2(c) Find the degree of the splitting field of f (x) = x5 − 3x3 + x2 − 3 over Q, the

field of rationals.

Solution. f (x) has -1 as a root, so f (x) = (x + 1)(x4 − x3 − 2x2 + 3x − 3). It does not

have any other linear factors as −1, 1, 3, −3 are not roots of x4 − x3 − 2x2 + 3x + 3.

Let x4 − x3 − 2x2 + 3x + 3 = (x2 + bx + c)(x2 + dx + e), where b, c, d, e ∈ Z. Then

b + d = −1, c + e + bd = −2, be + dc = 3, ce = −3. From ce = −3, we get c = −1, e = 3 or

c = 1, e = −3 (the other choices are symmetric). Using c = 1, e = −3, we get −3b + d = 3,

and now from b + d = −1, we get b = −1, d = 0. Thus we get

4

√ √

Consequently, the splitting field of f (x) over Q is the smallest field containing ± 3, 1±i2 3 ,

namely the √roots of x2 − 3 and x2 − x + 1. √ √ √

Thus Q( 3, i)√ is the required

√ splitting field. Since Q( 3, i) ⊇ Q( 3) ⊇

√ Q, and (Q( 3) :

Q) = 2 and (Q( 3, i) : Q( 3)) = 2 it follows that the splitting field Q( 3, i) of f (x) has

degree 4 over Q.

5

UPSC Civil Services Main 1993 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Let G be a cyclic group of order n and p | n. Prove that there exists a

homomorphism of G onto a cyclic group of order p. What is the kernel of the homomorphism?

Define φ : G −→ G0 by φ(ar ) = bt where r ≡ t mod p, t = 1, 2, . . . p.

φ(ar · as ) = bu where r + s ≡ u mod p. If φ(ar ) = bx where r ≡ x mod p and φ(as ) = by

where s ≡ y mod p, then x + y ≡ r + s ≡ u mod p ⇒ bx+y = bu . So φ(ar · as ) = φ(ar )φ(as ),

thus φ is a homomorphism.

ker φ = {ar | φ(ar ) = bp ⇔ r ≡ p mod p}. Thus the kernel of φ is {ap , a2p , . . . , amp , mp =

n}.

Solution. The number of 7-Sylow subgroups of ≡ 1 mod 7 and divides 56, so can be 1 or

8. If 1, then the 7-Sylow group is normal in G. If the number is 8, then G has 48 elements of

order 7, because if H, K are different Sylow subgroups of order 7, then H ∩ K = {e} because

7 is a prime.

Thus the Sylow subgroups of order 8 can come from the remaining elements which are

only 8 in number (including e). Thus there is a unique Sylow subgroup of order 8, which

has to be normal.

Thus any group of order 56 has a normal subgroup of order 7 or of order 8, so it cannot

be simple.

1

Question 1(c) Let H and K be normal subgroups of G (finite), with H a normal subgroup

of K. If P = K/H, S = G/H, show that G/K ' S/P .

• φ is well-defined: aH = bH ⇔ b−1 a ∈ H ⇒ b−1 a ∈ K(∵ H ⊆ K) ⇒ aK = bK ⇒

φ(aH) = φ(bH).

ker φ = {xH | xK = K}. But xK = K ⇔ x ∈ K. Thus ker φ = K/H. By the fundamental

G/H

theorem of homomorphisms, S/P = K/H ' G/K.

√ √

Question 2(a) If Z is the set of integers, then show that Z[ −3] = {a + b −3 | a, b ∈ Z}

is not a UFD.

√ √ 2 2

Solution. Let α = a + b −3 ∈ Z[ −3]. Then α is a unit √ iff N (α) = a + 3b = 1, because

if N (α) = 1, then αα = 1 ⇒ α is a unit with α = a − b −3 as its inverse. Conversely, if

√ N (αβ) = N (α)N (β) = 1 ⇒ N (α) = 1. In fact N (α) = 1 ⇒ α = ±1, the only

αβ = 1, then

units of Z[ −3].

2 is irreducible. Let 2 = αβ. We will prove that either α or β is a unit. N (2) = 4 ⇒

N (α) = 1, 2, 4. But N (α) = a2 + 3b2 = 2 is not possible for a, b ∈ Z. If N (α) = 1, then α is

a unit, otherwise N (α) = 4 ⇒ N (β) = 1√⇒ β is a √ unit. Thus 2 is irreducible. √

Similarly it can √ that 1 + −3, 1 − −3 are irreducible (N (1 + −3) √

√ be shown = 4).

Moreover, 2, 1 + −3, 1 − −3 are √ not associates

√ of each other as the only units in Z[ −3]

are ±1. Now 4 = 2 ·√2 = (1 + −3)(1 − −3) are two different factorizations of 4 into

irreducibles, hence Z[ −3] is not a UFD.

Question 2(b) Construct the addition and mutiplication table for Z3 [x]/hx2 + 1i, where Z3

is the set of integers modulo 3 and hx2 + 1i is the ideal generated by 1 + x2 .

(−1)r/2

r

x ≡ r−1 mod x2 + 1

(−1) 2 x

it follows that

f (x) ≡ a0 + a1 x − a2 − a3 x + a4 + . . . mod x2 + 1

= [bo ] + [b1 ][x]

where [x] is the residue class of x modulo x2 + 1 and [b0 ], [b1 ] are residue classes in Z3 .

Conversely, [bo ] + [b1 ][x] always belongs to Z3 [x]/hx2 + 1i.

2

Thus Z3 [x]/hx2 + 1i has 9 elements, namely

{0, 1, 2, x, x + 1, x + 2, 2x, 2x + 1, 2x + 2}

Note that we have listed representative elements of distinct residue classes modulo x2 + 1.

Now addition is simply: (a0 + a1 x) + (b0 + b1 x) = c0 + c1 x where ci ≡ ai + bi mod 3 for

i = 0, 1.

Multiplication is defined by (a0 + a1 x)(b0 + b1 x) = c0 + c1 x where c0 ≡ a0 b0 − a1 b1 mod 3

and c1 ≡ a0 b1 +a1 b0 mod 3. The reader can now expand these into the appropriate addition

and multiplication tables.

Notice the strong resemblance between the addition and multiplication rules derived

above and the corresponding rules for complex numbers. This is a consequence of the fact

that i is a root of x2 + 1, in fact the ring of Gaussian integers is isomorphic to Z[x]/hx2 + 1i.

Question 2(c) Let Q be the rational number field, and Q(21/2 , 21/3 ) by the smalled extension

field containing 21/2 , 21/3 . Find a basis of Q(21/2 , 21/3 ) over Q.

In fact if {v1 , . . . , vm } is a basis of K over L, {w1 , . . . , wn } is a basis of L over k, then

{vi wj | 1 ≤ i ≤ m, 1 ≤ j ≤ n} is a basis of K over k.

X m X n X m X n m X

X n

Proof: Let aij vi wj = 0 with aij ∈ k. Then aij vi wj = ( aij wj )vi =

i=1 j=1 i=1 j=1 i=1 j=1

n

X Xn

0. But aij wj ∈ L, and as v1 , . . . , vm are linearly independent over L, aij wj = 0 for

j=1 j=1

each i, 1 ≤ i ≤ m. However w1 , . . . , wn are linearly independent over k, thus aij = 0 for

1 ≤ i ≤ m, 1 ≤ j ≤ n.

Xm Xn

Let α ∈ K. Then α = αi vi , αi ∈ L. Now let αi = aij wj with aij ∈ k, then

i=1 j=1

m X

X n

α= aij vi wj ⇒ {vi wj | 1 ≤ i ≤ m, 1 ≤ j ≤ n} generate K over k. This completes

i=1 j=1

the proof. √

Now Q(21/2 ) has {1, 2} as a basis over Q, and (Q(21/2 , 21/3 ) : Q(21/2 )) = 3 with x3 −2 as

the irreducible polynomial of 21/3 over Q(21/2 ). Thus 1, 21/3 , 22/3 is a basis of (Q(21/2 , 21/3 )

over Q(21/2 ). Thus by the above result, {1, 21/2 , 21/3 , 21/2+1/3 , 22/3 , 22/3+1/2 } is a basis for

Q(21/2 , 21/3 ) over Q.

3

UPSC Civil Services Main 1994 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) If G is a group such that (ab)n = an bn for three consecutive integers for all

a, b ∈ G, then show that G is abelian.

Solution. We are given that (ab)i = ai bi , (ab)i+1 = ai+1 bi+1 , (ab)i+2 = ai+2 bi+2 .

Now (ab)i+1 = abai bi = abai bi = ai+1 bi+1 . Thus ai b = bai .

Also, (ab)2 (ab)i = ai+2 bi+2 = a2 ai b2 bi = a2 ai bbbi = a2 bai bbi = a2 b2 ai bi , because ai b = bai .

But (ab)i = ai bi , hence (ab)2 = a2 b2 ⇒ abab = a2 b2 ⇒ ba = ab. Thus G is abelian.

Note that the result is false if we only have two consecutive integers e.g. G = S3 has

(ab)6 = e = a6 b6 , and (ab)7 = (ab)6 ab = ab = a7 b7 .

Solution. By Sylow theorems, the number of 7-Sylow groups is ≡ 1 mod 7, and divides

42, and therefore divides 6 ⇒ there is only 1 Sylow group of order 7, which has to be normal,

thus a group of order 42 cannot be simple.

Question 1(c) Show that the additive group of integers modulo 4 is isomorphic to the mul-

tiplicative group of the non-zero elements of integers modulo 5. State the two isomorphisms.

Solution.

Z/h5i = {[1], [2], [3], [4]} = h[2]i

= {[2], [2]2 = [4], [2]3 = [3], [2]4 = [1]}

Two cyclic groups of the same order are isomorphic. φ : Z/(4) −→ Z/h5i:

1

φ([1]) = [2]

φ([1] + [1]) = φ([2]) = [2]2 = [4]

φ([3]) = φ(3.[1]) = [2]3 = [3]

φ([4]) = φ(4.[1]) = [2]4 = [1]

f : Z/h5i −→ Z/(4).

f ([2]) = [1]

f ([4]) = f ([2]2 ) = f ([2]) + f ([2]) = [2]

f ([3]) = f ([2]3 ) = f ([2]) + f ([2]) + f ([2]) = [3]

f ([1]) = f ([2]4 ) = f ([2]) + f ([2]) + f ([2]) + f ([2]) = [4]

Question 2(a) Find all the units of the integral domain of Gaussian integers.

Solution. Let Z[i] = {a + ib | a, b ∈ Z}. Let N (a + ib) = a2 + b2 . We will show that

α ∈ Z[i] is a unit ⇔ N (α) = 1.

If α is a unit then αβ = 1 for some β ∈ Z[i] ⇒ N (αβ) = N (α)N (β) = 1 ⇒ N (α) = 1

because N (α), N (β) are positive integers.

Conversely, N (α) = 1 ⇒ a2 + b2 = 1 ⇒ (a + ib)(a − ib) = 1 ⇒ α is a unit.

Now the only integer solutions to N (α) = a2 + b2 = 1 are a = ±1, b = 0 or a = 0, b = ±1.

Thus the only units are {±1, ±i}.

Question 2(b) Prove or disprove: The polynomial ring I[x] over the ring of integers is a

Principal Ideal Domain (PID).

Solution. It is not a PID. The ideal generated by 2 and x is not a principal ideal. Suppose

h2, xi = hf (x)i. Then 2 ∈ hf (x)i ⇒ f (x)g(x) = 2 for some g(x). This means that f (x) is a

constant and divides 2, so f (x) = 1or2.

f (x) = 2 ⇒ x 6∈ hf (x)i ∵ 2g(x) = x is not possible for any g(x) ∈ I[x].

f (x) = 1 ⇒ 1 ∈ h2, xi ⇒ 1 = 2p(x) + xq(x) ⇒ 2× the constant term of a(x) = 1, which

is not possible. Thus h2, xi is not a principal ideal.

Question 2(c) Let R be an integral domain (not necessarily a unique factorization domain),

and F its field of quotients. Show that any element f (x) ∈ F [x] is of the form f (x) = f0a(x)

where f0 (x) ∈ R[x] and a ∈ R.

Solution.Qf (x) = a0 + a1 x + . . . am xm , where ai ∈ F . Now ai = bi /ci , where bi , ci ∈ R.

Then f (x) i ci = A0 + A1 x + . . . + Am xm where Ai ∈ R.

Thus f (x) = f0a(x) , where f0 (x) = A0 + A1 x + . . . + Am xm , and a = i ci .

Q

2

UPSC Civil Services Main 1995 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Let G be a finite set closed under an associative binary operation such that

ab = ac ⇒ b = c, ba = ca ⇒ b = c for all a, b, c ∈ G. Prove that G is a group.

These sets have distinct elements because aj ai = ak ai ⇒ aj = ak , and ai aj = ai ak ⇒ aj = ak .

Thus G = {a1 , a2 , . . . , an } = {a1 a1 , a2 a1 , . . . , an a1 } = {a1 a1 , a1 a2 , . . . , a1 an }. Thus there ex-

ists r, 1 ≤ r ≤ n such that a1 = a1 ar . Now for any aj ∈ G, aj = as a1 for some s, therefore

aj ar = as a1 ar = as a1 = aj . Hence we have proved that G has a right identity. As seen above,

for any aj ∈ G, the set {aj a1 , aj a2 , . . . , aj an } = G, hence therefore there exists k, 1 ≤ k ≤ n

such that aj ak = ar , thus every element has a right inverse.

Similarly, we can show that G has a left identity and every element in G has a left

inverse. Let as be the left identity. Then ar = as ar = as , so the left identity is the same as

the right identity. If ai aj = ar and ak ai = ar , then ak = ak ar = ak ai aj = ar aj = aj (using

associativity), hence the left inverse is the same as the right inverse. Thus G has an identity,

every element of G has an inverse, and the operation is associative, so G is a group.

Alternatively, let x ∈ G and let xy = e, where e is the right identity. Then exy = ee =

e = xy ⇒ ex = x, so e is also the left identity. Now yxy = ye = ey ⇒ yx = e, thus the right

inverse is the same as the left inverse.

Question 1(b) Let G be a subgroup of order pn , where p is a prime number and n > 0. Let

H be a proper subgroup of G and N (H) = {x ∈ G | x−1 hx ∈ H for every h ∈ H} = {x ∈

G | x−1 Hx = H}. Prove that N (H) 6= H.

Solution. The proof is by induction over n. If n = 1, then H = {e} is the only possibility

for a proper subgroup, since G is cyclic. N (H) = G 6= H. If n = 2, it is well known that G

is abelian, and therefore for any proper subgroup H of G, N (H) = G 6= H.

1

Assume as induction hypothesis that the result is true for all groups of order pm where

m < n.

Let G be a group of order pn and let H be a proper subgroup of G. We consider the

following two possible cases

Case (i): H does not contain C, the center of G, then there exists an element z ∈ C − H.

Clearly z ∈ N (H) and therefore N (H) ⊃ H properly.

Case (ii): H ⊇ C. In this case H = H/C is a proper subgroup of G = G/C. Since G

is a prime power group, it is known that the center C of G is nontrivial, therefore |G| =

order of G = pm where m < n. Thus by the induction hypothesis the normalizer of H in G

contains H properly, i.e. there exists an element b ∈ G such that b 6∈ H and b ∈ N (H) i.e.

−1

b Hb = H. It is now obvious that b 6∈ H and b−1 Hb ⊆ HC = H i.e. b ∈ N (H). Hence

N (H) 6= H.

Alternative presentation: Let Co = {e}, C1 = center of G. If C1 6= G, let Z1 be

the center of G/C1 . Let C2 = η −1 (Z1 ), where η : G −→ G/C1 is the natural map. Thus

C2 /C1 = Z1 . If C2 6= G, we define C3 = η −1 (center of G/C2 ), where η is now the natural

map from G ont G/C2 .

Clearly C0 ( C1 ( C2 ( . . . because the center of a prime power group is non-trivial.

Since G is finite, we have Cr = G for some r. Thus C0 ( C1 ( C2 ( . . . ( Cr = G. Now

each Ci is normal in G, because Z1 is normal in G/C1 ⇒ η −1 (Z1 ) = C2 is normal in G and

so on.

Since C0 ⊆ H, and Cr 6⊆ H, there is a k, 0 ≤ k < r such that Ck ⊆ H, Ck+1 6⊆ H. Let

x ∈ Ck+1 , x 6∈ H. For any g ∈ G, x−1 g −1 xg ∈ Ck , because xCk ∈ center of G/Ck , x ∈ Ck+1 ,

which means that xgCk = xCk gCk = gCk xCk = gxCk . Thus x−1 g −1 xg ∈ Ck .

In particular x−1 h−1 xh ∈ Ck ∀h ∈ H. Thus x−1 h−1 xh ∈ H because Ck ⊆ H, or

x−1 h−1 x ∈ H for all h ∈ H. Thus x ∈ N (H). But x 6∈ H, so N (H) 6= H.

If the Sylow 2-subgroup, which is of order 16, is unique, then it is automatically a normal

subgroup of G and we have nothing to prove.

Let us therefore assume that G has more than one Sylow 2-subgroups. By one of Sylow

theorems, the number of such subgroups is ≡ 1 mod 2, and is a divisor of 112 and therefore

of 7. Thus G has 7 subgroups say H1 , H2 , . . . , H7 such that |Hi | = 16, 1 ≤ i ≤ 7.

Observe that Hi ∩ Hj for i 6= j must have at least 4 elements because if not |Hi Hj | ≥ 128

|H ||H |

as |Hi Hj | = |Hii∩Hjj | , which is not possible.

We now consider the following two cases.

Case 1: Suppose (without loss of generality) that |H| = |H1 ∩ H2 | = 8. This means

that H is a normal subgroup of H1 as well as H2 and therefore N (H) contains H1 H2 . But

|H1 ||H2 |

|H1 H2 | = |H 1 ∩H2 |

= 32, therefore |N (H)| ≥ 32 and 16 divides |N (H)| as N (H) ⊃ H.

Consequently |N (H)| = 112 i.e. N (H) = G. Thus H is a normal subgroup of G showing

that G is not simple.

2

Case 2: Let |Hi ∩ Hj | = 4 for i 6= j. Let H = H1 ∩ H2 , then |H| = 4. We have proved

in question 1(b) that NH1 (H) (the normalizer of H in H1 ) contains H properly and also

NH2 (H) contains H properly. Thus each of NH1 (H) and NH2 (H) have 8 or 16 elements.

Case 2(a): One of the normalizers has 16 elements — suppose without loss of generality

that NH1 (H) = H1 , then NG (H) contains H1 and NH2 (H) and therefore NG (H) contains

at least 16 × 8/4 elements, and 16 divides |NG (H)| as H1 ⊂ NG (H) — note that |Hi | =

16, |NH2 (H)| ≥ 8 and H1 ∩ NH2 (H) being a subgroup of H1 ∩ H2 has at most 4 elements.

Thus as in case 1, we get NG (H) = G, so H is a normal subgroup of G, showing that G is

not simple.

Case 2(b): NH1 (H) 6= H1 and NH2 (H) 6= H2 , then |NH1 (H)| = |NH2 (H)| = 8. In this

case NG (H) contains at least 8 × 8/4 elements and 8 divides |NG (H)|. Thus |NG (H)| = 16

or 56. If |NG (H)| = 16, then it is one of the Hi , say NG (H) = H3 , in this case |H1 ∩ H3 | = 8,

which contradicts the precondition for case 2 i.e. |Hi ∩Hj | = 4 for i 6= j. Thus |NG (H)| = 56,

and in this case G is not simple as NG (H) is a proper normal subgroup of G.

This completes the proof.

Alternative Presentation. o(G) = 24 · 7. The number of 7-Sylow subgroups ≡ 1

mod 7 and divides o(G). Thus the number of 7-Sylow subgroups is 1 or 8. If 1, then the

7-Sylow subgroup of G is normal in G, and G is not simple. Otherwise we will show that G

has a unique 16-Sylow subgroup, which will be normal in G and hence G will not be simple.

Let the number of 7-Sylow subgroups be 8. This accounts for 49 elements, 48 of order 7,

and the identity. Note that if H and K are Sylow subgroups of order 7, then H ∩ K = {e}

if H 6= K because order of H is prime.

We are now left with 63 elements + identity. The number of 2-Sylow groups is ≡ 1

mod 2 and divides 7. Thus out of these 64 elements we should get 7 16-Sylow subgroups

(because if there is only one 16-Sylow subgroup, it is normal, hence G is not simple). These

7 subgroups of order 16 will have a unique subgroup of order 8, which would be normal in

G.

Thus in all cases, G is not simple.

Question 2(a) Let R be a ring with identity. If an element of R has more than one right

inverse, show that it has infinitely many right inverses.

y = x). Consider x, (xa − e) + x, (ya − e) + x. Then

Thus we get three distinct right inverses (if xa − e + x = ya − e + x then xax = yax ⇒

x = y). So given n inverses a1 , a2 , . . . , an of a, by considering a1 , a1 a − e + a1 , a2 a − e +

a1 , . . . , an a − e + a1 we can get n + 1 distinct right inverses. Hence there must be infinitely

many right inverses.

3

Question 2(b) Let hp(x)i be an ideal generated by an irreducible polynomial in F [x], F a

field. Prove that it is a maximal ideal.

Let g(x) ∈ M, g(x) 6∈ hp(x)i ⇒ p(x) 6 | g(x). Thus (g(x), p(x)) = 1 i.e. g(x) and p(x) are

coprime. Then there exist a(x), b(x) ∈ F [x] such that a(x)g(x) + p(x)b(x) = 1 ⇒ 1 ∈ M ⇒

M = F [x].

Note that F [x] is a principal ideal domain. Therefore hp(x), g(x)i is a principal ideal and

it has to be generated by 1, because p(x) has no other divisors.

F [x]. Define f 0 (x) = a1 + 2a2 x + . . . + nan xn−1 . If f 0 (x) = 0, then prove that there exists

g(x) = F [x] such that f (x) = g(xp ) = g(x)p .

Solution. f 0 (x) = 0 ⇔ rar = 0 ⇔ ar = 0 when r 6≡ 0 mod p. Thus f (x) = tm=0 amp xmp

P

where t = [n/p]. Let g(y) = a0 + ap y + . . . atp y t . Then g(xp ) = f (x) = (g(x))p .

4

UPSC Civil Services Main 1996 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Let R be the set of all real numbers and G = {(a, b) | a, b ∈ R, a 6= 0}. Let

x : G × G −→ G be defined by (a, b) ∗ (c, d) = (ac, bc + d). Show that (G, ∗) is a group. Is it

abelian? Is (H, ∗) a subgroup of (G, ∗) where H = {(1, b) | b ∈ R}?

Solution.

1. α = (a, b) ∈ G, β = (c, d) ∈ G ⇒ α ∗ β ∈ G ∵ ac 6= 0 when a 6= 0, c 6= 0.

2. (a, b) ∗ (1, 0) = (a, b.1 + 0) = (a, b). (1, 0) ∗ (a, b) = (a, 0.a + b) = (a, b). Hence (1, 0) is

identity of G.

3. (a, b) ∗ (1/a, −b/a) = (1, b/a − b/a) = (1, 0). (1/a, −b/a) ∗ (a, b) = (1, (−ab/a) + b) =

(1, 0). Thus the inverse exists for every (a, b) ∈ G.

4. (a, b) ∗ [(c, d) ∗ (e, f )] = (a, b) ∗ (ce, de + f ) = (ace, bce + de + f ). [(a, b) ∗ (c, d)] ∗ (e, f ) =

(ac, bc + d) ∗ (e, f ) = (ace, bce + de + f ). Thus ∗ is associative.

Hence G is a group.

G is not abelian: (c, d) ∗ (a, b) = (ac, da + b). Thus if (a, b) ∗ (c, d) = (c, d) ∗ (a, b), then

bc + d = da + b. This need not be true, for example if a = d = 1, b = 0.

If (1, a), (1, b) ∈ H, then (1, a) ∗ (1, b) = (1, a + b) ∈ H. (1, 0) ∈ H. Finally (1, a)−1 =

(1, −a) ∈ H, hence H is a subgroup of G.

Question 1(b) Let f be a homomorphism of a group G onto a group G0 with kernel H. For

each subgroup K 0 of G0 , define K as K = {x | x ∈ G, f (x) ∈ K 0 }. Prove that

1. K 0 is isomorphic to K/H.

2. G/K is isomorphic to G0 /K 0 .

1

Solution.

because given y ∈ K 0 , y ∈ G0 ⇒ ∃x ∈ G, f (x) = y. This x ∈ K.). ker(f 0 ) =

{x | x ∈ K, f (x) = e0 , the identity of G0 }. ker(f ∗ ) ⊆ H, but H ⊆ K, therefore

f (x) = f ∗ (x) = e0 for x ∈ H. Thus H ⊆ ker(f ∗ ), that is H = ker(f ∗ ). Thus

K/H ' K 0 by the fundamental theorem of homomorphisms.

φ(x)φ(y).

• φ is onto: Let yK 0 ∈ G0 /K 0 . f is onto ⇒ there exists x ∈ G such that f (x) = y.

Note that y ∈ G0 . Then φ(x) = f (x)K 0 = yK 0 . ker(φ) = K ∵ x ∈ ker(φ) ⇔

f (x)K 0 = K 0 ⇔ f (x) ∈ K 0 ⇔ x ∈ K. Thus G/K ' G0 /K 0 .

Question 1(c) Prove that a normal subgroup H of a group G is maximal ⇔ the quotient

G/H is simple.

K 6= G. Then K/H is a normal subgroup of G/H. G/H is simple so K/H is identity or

K/H = G/H. If K/H is identity, then K = H. If K/H = G/H then K = G. Hence H is

maximal.

Conversely, let H be maximal. Let H 0 be a normal subgroup of G/H. Assume H 0 is

different from the identity of G/H, i.e. H 0 contains at least one element different from the

identity of G/H. We shall show that H 0 = G/H, showing that G/H is simple.

Let η : G −→ G/H be the natural homomorphism. Then

η −1 (H 0 ) = {x | x ∈ G, η(x) = Hx ∈ H 0 }

that xH ∈ H 0 ⇒ x ∈ η −1 (H 0 ) but x 6∈ H. Since H is maximal, η −1 (H 0 ) = G ⇒ H 0 = G/H.

Question 2(a) In a ring R, prove that the cancellation law holds in R ⇔ R has no zero

divisors.

a0, a 6= 0 ⇒ b = 0.

Conversely ab = ac ⇒ a(b − c) = 0, a 6= 0 ⇒ b − c = 0 ⇒ b = c.

2

Question 2(b) If S is an ideal of R and T any subring of R then prove that S is an ideal

of S + T = {s + t | s ∈ S, t ∈ T }.

condition a, b ∈ S ⇒ a−b ∈ S is true). α = s+t, s ∈ S, t ∈ T , thus αa = (s+t)a = sa+ta ∈ S

because s ∈ S, a ∈ S ⇒ sa ∈ S, t ∈ T ⇒ t ∈ R, a ∈ S ⇒ ta ∈ S ⇒ sa + ta ∈ S.

Question 2(c) Prove that the polynomial x2 + x + 4 is irreducible over the field of integers

modulo 11.

Solution. If x2 + x + 4 were reducible modulo 11, then it would have a linear factor i.e. it

would have a root in the field Z11 . But

x=1 ⇒ 1+1+4 6 ≡ 0 mod 11

x=2 ⇒ 4+2+4 6 ≡ 0 mod 11

x=3 ⇒ 9+3+4 6 ≡ 0 mod 11

x=4 ⇒ 16 + 4 + 4 6 ≡ 0 mod 11

x=5 ⇒ 25 + 5 + 4 6 ≡ 0 mod 11

x=6 ⇒ 36 + 6 + 4 6 ≡ 0 mod 11

x=7 ⇒ 49 + 7 + 4 6 ≡ 0 mod 11

x=8 ⇒ 64 + 8 + 4 6 ≡ 0 mod 11

x=9 ⇒ 81 + 9 + 4 6 ≡ 0 mod 11

x = 10 ⇒ 100 + 10 + 4 6 ≡ 0 mod 11

3

UPSC Civil Services Main 1997 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Show that a necessary and sufficient condition for a subset H of a group G

to be a subgroup is HH −1 = H.

Solution. Let H be a subgroup. Clearly H ⊆ HH −1 because h ∈ H can be written as

h = he where h ∈ H, e ∈ H −1 ⇒ h ∈ HH −1 . If x ∈ HH −1 , then x = hk −1 where h, k ∈ H.

But H is a group, so hk −1 ∈ H, thus HH −1 ⊆ H ⇒ HH −1 = H.

Conversely, let H = HH −1 and assume H 6= ∅.

1. a ∈ H ⇒ a−1 ∈ H −1 ⇒ aa−1 ∈ HH −1 = H ⇒ e ∈ H.

2. x, y ∈ H ⇒ xy −1 ∈ H. Thus x ∈ H ⇒ x−1 = ex−1 ∈ H. x, y ∈ H ⇒ y −1 ∈ H ⇒ y ∈

H −1 ⇒ xyinH.

Thus H is a subgroup of G.

Question 1(b) Show that the order of each subgroup of a finite group is a divisor of the

order of the group.

Solution. Lagrange’s theorem, see Theorem 2.4.1 page 41 of Algebra by Herstein.

Question 1(c) In a group G, the commutator of (a, b), a, b ∈ G is the element aba−1 b−1

and the smallest subgroup containing all commutators is called the commutator subgroup of

G. Show that a quotient group G/H is abelian ⇔ H contains the commutator subgroup of

G.

Solution. Let G/H be abelian. Then HaHb = HbHa ⇒ Hab = Hba ⇒ Haba−1 b−1 =

H ⇒ aba−1 b−1 ∈ H. This means H contains all the commutators, and therefore contains

the group generated by them (i.e. the commutator subgroup).

Conversely, if H contains the commutator subgroup, then for any a, b ∈ G, aba−1 b−1 ∈

H ⇒ Haba−1 b−1 = H ⇒ Hab = Hba ⇒ HaHb = HbHa ⇒ G/H is abelian.

1

Question 2(a) If x2 = x for all x in a ring R, show that R is commutative. Give an

example to show that the converse is not true.

ab + ba = 0. Setting a = b, we get 2b2 = 0 ⇒ 2b = 0. Thus ab = −2ba + ba ⇒ ab = ba. Thus

R is commutative.

Converse is not true — Z is commutative but n2 6= n for n 6= 0, 1.

Question 2(b) Show that an ideal S of the ring of integers Z is a maximal ideal ⇔ S is

generated by a prime integer.

Solution. Let S be maximal. Since Z is a PID, we have S = hqi for some q ∈ Z, q 6= 0, 1, −1.

We will prove that if q | ab, q - a then q | b showing that q is prime. Since q - a, we

have a 6∈ S. Consider the ideal generated by S and a. It is Z, because S is maximal.

hS, ai = Z ⇒ 1 = α + ta, α ∈ S. Thus 1 = xq + ta, α = xq. Hence b = xbq + tab. Clearly

q | RHS, so q | b ⇒ q is a prime.

Conversely let S = hpi where p is a prime. We wish to show that S is maximal. Let A

be an ideal, A ⊃ S and A 6= S, then we shall show that A = Z. Since A ⊃ S, ∃a ∈ A, a 6∈ S.

Now a 6∈ S ⇔ p - a ⇔ (a, p) = 1 ⇔ xa + yp = 1 for some x, y ∈ Z ⇒ 1 ∈ A(∵ a ∈ A ⇒ xa ∈

A, p ∈ A ⇒ yp ∈ A). Hence Z = A, so S is a maximal ideal.

Question 2(c) Show that in an integral domain every prime element is irreducible. Give

an example to show that the converse is not true.

Solution. Let R be an integral domain with unity. Let p be a prime element of R i.e.

if p | ab then p | a or p | b. We have to show that if a | p then either a is an associate

of p or a is a unit. If a | p, then p = ab for some b ∈ R. But p is a prime, therefore

p = ab ⇒ p | ab ⇒ p | a or p | b. If p | a, then p is an associate of a as a | p. If p | b, then

b = px for some x ∈ R. Thus p = pax ⇒ ax = 1 as R is an integral domain, thus a is a unit.

Hence a prime element is irreducible.

The converse is not

√ true. Let R be an √integral domain which is not a unique factorization

domain e.g. R = Z[ −5]. For α = a + b −5, N (α) = a2 + 5b2 . 2 is an irreducible element

of R — 2√= αβ ⇒ N (α)N (β) = 4 ⇒ N (α) = 1, 2, 4. N (α) = 1 ⇒ α is a unit, because if

α = a + b −5 then a2 + 5b2 = 1 ⇒ b = 0, a = ±1 ⇒ a is a unit. If N (α) = 4, then N (β) = 1

2 2

so β is a unit. N (α) = 2 is impossible

√ as a +√5b = 2 is not possible.

√ √

Now 2 is not

√ prime — 2 | (1 + −5)(1 − −5). But 2 - 1 + −5 because

√ (1 + −5) =

2α = 2(a + b −5) ⇒ 2a = 1, which is not possible. Similarly 2 - 1 − −5. So 2 is not

prime.

2

UPSC Civil Services Main 1998 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Prove that if a group has only 4 elements then it must be abelian.

Solution. Let G be a group of order 4. If it has an element of order 4, then G is cyclic

and therefore abelian. If G has no elements of order 4, then the order of all elements other

than identity is 2 because the order of an element must be a divisor of 4. Let x, yinG, then

(xy)2 = xyxy = e ⇒ yx = x−1 ey −1 = x−1 y −1 = xy because x−1 = x, y −1 = y. Hence

xy = yx for every x, y ∈ G so G is abelian.

Question 1(b) If H and K are subgroups of G then show that HK is a subgroup of G if

and only if HK = KH.

Solution. See Lemma 2.5.1 page 44 of Algebra by Herstein.

Question 1(c) Show that every group of order 15 has a normal subgroup of order 5.

Solution. By Sylow’s theorem a group G of order 15 has a subgroup of order 5. Again by

one of Sylow’s theorems the number of subgroups is ≡ 1 mod 5, and this number divides 3.

Therefore there is exactly 1 subgroup of order 5, say H. Now aHa−1 is also a subgroup of G

of order 5, but H is the only such subgroup, so aHa−1 = H, hence H is a normal subgroup.

Hence every group of order 15 has a normal subgroup of order 5.

Question 2(a) Let (R, +, .) be a system satisfying all the axioms for a ring with unity with

the possible exception of a + b = b + a. Prove that (R, +, .) is a ring.

Solution. Let e denote unity of R. Then (a+b)(e+e) = a(e+e)+b(e+e) = ae+(a+b)e+be.

Also (a+b)(e+e) = (a+b)e+(a+b)e = ae+be+ae+be. Thus ae+be = be+ae ⇒ a+b = b+a.

Thus R is a ring.

A similar question is the following. Let (R, +, .) be a system satisfying all the axioms for

a ring with the possible exception of a + b = b + a. If there is an element c ∈ R such that

ac = bc ⇒ a = b for every a, b ∈ R, then show that R is a ring.

1

Question 2(b) If p is a prime then prove that Zp is a field. Discuss the case when p is not

a prime.

Solution. Zp is a commutative ring with unity. Let [a] ∈ Zp such that a 6≡ 0 mod p

i.e. [a] 6= [0]. Let {[x1 ], . . . , [xp ]} = Zp . Then [a][x1 ], . . . , [a][xp ] are all distinct, since

[a][xi ] = [a][xj ] ⇒ a(xi − xj ) ≡ 0 mod p ⇒ xi ≡ xj mod p because a 6≡ 0 mod p. Thus

there exists k such that [a][xk ] = [1] ⇒ every non-zero element in Zp has an inverse. Thus

Z∗p = Zp − {[0]} is a group, so Zp is a field.

If p is not prime, then Zp is not even an integral domain — if p = n1 n2 , n1 > 1, n2 > 1,

then [n1 ][n2 ] = [0], but [n1 ] 6= [0], [n2 ] 6= [0] in Zp .

See corollary to Lemma 3.2.2 page 128 of Algebra by Herstein.

Question 2(c) Let D be a principal ideal domain. Show that every element that is neither

0 nor a unit in D is a product of irreducible elements.

Solution.

an integer m thus that Am = Am+1 = . . ..

Proof: Let A = ∞

S

i=1 Ai , then we will show that A is an ideal — If a, b ∈ A, then

a ∈ Ar for some r, and b ∈ As for some s. Hence a, b ∈ As if s ≥ r (say), thus a−b ∈ As

because As is an ideal ⇒ a − b ∈ A. Let a ∈ A, d ∈ D ⇒ a ∈ Ar ⇒ ra ∈ Ar because

Ar is an ideal ⇒ ra ∈ A. Thus A is an ideal. Since D is a PID, A = hai, i.e.a generates

A. By definition of A, there exists m s.t. a ∈ Am . Thus A = Am = Am+1 = . . . ⊂= A.

Proof: Let a ∈ D, a 6= 0, a non-unit. If a is not irreducible then we have nothing to

prove. If a is irreducible, then a has a proper divisor, say a1 ⇒ ha1 i ⊂ hai. Continuing

this process, we have a2 , a3 , . . . , such that as divides as−1 for s = 1, 2, . . ., where a0 = a.

But this sequence must terminate i.e. ∃m such that ham i = ham+1 i = . . . because of

step 1. But this means that am has no proper factors i.e.am is irreducible.

2, a = p1 a1 where p1 is irreducible, and a1 | a properly. If a1 is a unit, then a is a

product of irreducible factors. If not, then a1 = p2 a2 where a2 | a1 properly. But this

process cannot go on forever, by the same argument as in step 2. Thus we must have

an integer k such that a = p1 p2 . . . pk ak where ak is a unit. Thus a is a product of

irreducible elements.

2

UPSC Civil Services Main 1999 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

normal subgroup of G.

Question 1(b) If p is a prime number pα | o(G), then prove that G has a subgroup of order

pα .

Solution. This is Sylow’s theorem. See theorem 2.12.1 page 92 of Algebra by Herstein.

Question 1(c) Let R be a commutative ring with unit element whose only ideals are the

zero ideal and R itself. Show that R is a field.

1

UPSC Civil Services Main 2000 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Let n be a fixed positive integer and let Zn be the ring of integers modulo n.

Let

G = {[a] ∈ Zn | a 6= 0, (a, n) = 1}

Show that G is a group under multiplication defined in Zn . Hence or otherwise show that

aφ(n) ≡ 1 mod n for all integers a relatively prime to n, where φ(n) denotes the number of

positive integers that are less than n and relatively prime to n.

Solution. The only thing we have to show is that every element in G is invertible as we

already know that G is multiplicatively closed, and has identity element namely [1]. Let

a1 , . . . , am , m = φ(n) be representatives of prime residue classes modulo n. Let a be any

integer such that (a, n) = 1, i.e. a is coprime with n. Then [aa1 ], [aa2 ], . . . , [aam ] are all

distinct because aai ≡ aaj mod n ⇒ a(ai − aj ) ≡ 0 mod n, but (a, n) = 1, therefore

ai − aj ≡ mod n, which is not true. Thus there exists a j such that aaj ≡ 1 mod n, note

that G = {[a1 ], [a2 ], . . . , [am ]} = {[aa1 ], [aa2 ], . . . , [aam ]} and [1] ∈ G. Consequently [a] is

invertible, in fact [a][aj ] = [1].

We know that if G is a group of order n, then x ∈ G ⇒ xn = e for every x ∈ G, where e

is the identity of G — consider H the cyclic subgroup of G generated by x. Using Lagrange’s

theorem, which says that the order of a subgroup divides the order of a group if the group

is finite, we get o(x) = o(H) | o(G) = n. Thus n = o(x)k, so xn = xo(x)k = ek = e.

Thus if a is any integer such that (a, n) = 1, then [a] ∈ G ⇒ [a]φ(n) = [1] because

o(G) = φ(n). Hence aφ(n) ≡ 1 mod n.

Question 1(b) Let M be a subgroup and N a normal subgroup of a group G. Show that

M N is a subgroup of G and M N/N is isomorphic to M/M ∩ N .

Solution.

1

1. M N 6= ∅

2. x, y ∈ M N ⇒ x = m1 n1 , y = m2 n2 where m1 , m2 ∈ M, n1 , n2 ∈ N . Then xy =

m1 n1 m2 n2 = m1 m2 m−1 −1

2 n1 m2 n2 . Since N is a normal subgroup, m2 n1 m2 ∈ N , there-

fore xy = m1 m2 n∗1 n2 where n∗1 = m−12 n1 m2 , showing that xy ∈ M N .

3. x ∈ M N ⇒ x−1 = n−1 −1 −1 −1 −1

1 m1 = m1 m1 n1 m1 ∈ M N

Thus M N is a subgroup of G.

Consider the function f : M −→ M N/N defined by f (m) = mN . Then

f (m).

3. ker f = {m | f (m) = mN = N ⇔ m ∈ N } = M ∩ N .

N ' M N/N .

Question 2(a) Let F be a finite field. Show that the characteristic of F must be a prime

integer p and the number of elements in F must be pm for some positive integer m.

0 ⇒ λa = 0 or µa = 0. Suppose λa = 0. Then for any b ∈ F, b 6= 0, λab = a.λb = 0 ⇒ λb = 0

because a 6= 0. Thus λx = 0 for every x ∈ F , so λ = n because n is the smallest such integer.

Thus if n = λµ, then λ = n or µ = n, so n is prime, say p.

Consider the mapping f : Z −→ F defined by f (n) = ne where e is the multiplicative

identity of F . It is obvious that f is a homomorphism, and that ker f is hpi, the ideal

generated by p. Thus Z/hpi is isomorphic to a subfield of F . In other words F contains a

field Λ having p elements. If (F : Λ) = m, then F has pm elements. For details see question

2(c)(ii) year 2002.

Question 2(b) Let F be a field and F [x] denote the set of all polynomials defined over F .

If f (x) is an irreducible polynomial in F [x], show that the ideal hf (x)i generated by f (x) in

F [x] is maximal and F [x]/hf (x)i is a field.

Solution. Let A by an ideal, A ⊃ hf (x)i. Since F [x] is a principal ideal domain, let

A = hg(x)i. Then A ⊃ hf (x)i ⇒ f (x) = g(x)h(x). But f (x) is irreducible, so either g(x) is

a unit or g(x) is an associate of f (x). Thus hg(x)i = F [x] or hg(x)i = hf (x)i ⇒ hf (x)i is

maximal.

In order to show that F [x]/hf (x)i is a field, the only thing we have to show is that any

non-zero element in F [x]/hf (x)i is invertible. Let g(x) + hf (x)i be any non-zero element in

F [x]/hf (x)i i.e.f (x) - g(x). This means that f (x), g(x) are comprime, therefore there exist

2

a(x), b(x) such that a(x)g(x) + b(x)f (x) = 1. Consequently a(x)g(x) ≡ 1 mod f (x), thus

g(x) + hf (x)i has a(x) + hf (x)i as an inverse in F [x]/hf (x)i.

Alternately, let g(x) be as above. Consider M = ideal generated by f (x), g(x). Since

f (x) - g(x), M 6= hf (x)i, and as hf (x)i is maximal, M = F [x]. Thus there exists a(x), b(x) ∈

F [x] such that a(x)g(x) + b(x)f (x) = 1 and we get the same conclusion as above.

Question 2(c) Show that a finite commutative ring with no zero divisors must be a field.

3

UPSC Civil Services Main 2001 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Let K be a field and G be a finite subgroup of the multiplicative group of

the non-zero elements of K. Show that G is a cyclic group.

Solution. Let a ∈ G be chosen that o(a) ≥ o(b) ∀b ∈ G — this is possible because G is

finite. We shall show that G = hai.

Step 1. For any b ∈ G, o(b) | o(a). If not, there exists an element b ∈ G s.t. o(b) =

pl r, (p, r) = 1, o(a) = pm s, (p, s) = 1 where l > m ≥ 0, because if all primes occurring in

m

o(b) have power less than that occurring in o(a), then o(b) | o(a). Define x = br , y = ap ⇒

o(x) = pl , o(y) = s ⇒ o(xy) = pl s(∵ (o(x), o(y)) = 1, xy = yx) ⇒ o(xy) > o(a) which is a

contradiction. Hence o(b) | o(a).

Step 2. If o(a) = n, then bn = 1 ∀b ∈ G. Thus all elements of G are roots of xn − 1 = 0.

But this equation has at most n roots in K, thus |G| ≤ n. But o(a) = n ∴ 1, a, . . . , an−1 are

all distinct in G. Therefore o(G) ≥ n.

Thus o(G) = n ⇒ hai = G so G is cyclic.

Question 1(b) Prove that the polynomial 1 + x + . . . xp−1 , where p is a prime number, is

irreducible over the field of rational numbers.

xp −1

Solution. f (x) is irreducible ⇐⇒ f (1 + x) is irreducible. f (x) = x−1

. Thus

(x + 1)p − 1

f (1 + x) =

x

x + p1 xp−1 + p p

p

2

xp−2 + . . . + p−1

x

=

x

p−1 p p−2 p p−3 p p−r−1 p

= x + x + x + ... + x + ... +

1 2 r p−1

p p p!

Now p | r for r = 1, 2, . . . , p − 1, as r = r!(p−r)! , and p | p!, but p - r!, p - (p − r)!. Thus

the Eisenstein criterion gives the result.

1

Question 2(a) Let N be a normal subgroup of a group G. Show that G/N is abelian ⇔ for

all x, y ∈ G, xyx−1 y −1 ∈ N .

N ⇒ x−1 y −1 xy ∈ N .

Conversely, xyx−1 y −1 ∈ N ⇒ xyx−1 y −1 N = N ⇒ x−1 y −1 N = y −1 x−1 N ⇒ x−1 N y −1 N =

y −1 N x−1 N . Thus G/N is abelian.

Question 2(b) If R is a commutative ring with unit element and M is an ideal of R, show

that M is a maximal ideal of R if and only if R/M is a field.

Question 2(c) Prove that every finite extension of a field is an algebraic extension. Give

an example to show that the converse is not true.

space over k is n. Let α ∈ K, α 6= 0, then {1, α, α2 , . . . , αn } are linearly dependent, i.e. there

exist a0 , a1 , . . . , anP∈ k with at least one ai 6= 0 such that a0 + a1 α + . . . + an αn = 0 i.e. α is

a root of f (x) = ni=0 ai xi with coefficients from k. Thus K | k is an algebraic extension of

k as every element of K is algebraic over k.

Example: Let K = Q(21/n , n = 2, 3, 4, . . .). K | Q is algebraic but (K : Q) is not finite.

If (K : Q) = r then 21/n for n > r + 1 is a root of the polynomial of degree ≤ r + 1, which

is not possible because 21/n is a root of xn − 2 = 0 which is an irreducible polynomial over

Q, showing that 21/n cannot be a root of a polynomial of degree < n.

K | Q is algebraic because every element is contained in a field L such that Q ⊆ L ⊂ K

and (L : Q) < ∞ ⇒ α is algebraic over Q.

2

UPSC Civil Services Main 2002 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Solution. Let G be a group of order 35. By Sylow’s theorem, G has a subgroup of order 7

and the number of such groups is ≡ 1 mod 7 and divides 35. If the number is 7t + 1, then

7t + 1 | 35 ⇒ 7t + 1 | 5 ⇒ t = 0. Thus G has a unique Sylow subgroup H of order 7, which

must be a normal subgroup of G. Similarly the number of 5-Sylow subgroups is ≡ 1 mod 5

and divides 35. If this number is 5r + 1, then 5r + 1 | 35 ⇒ 5r + 1 | 7 ⇒ r = 0. Thus G has

a unique 5-Sylow group say K which is normal in G.

This means HK is a subgroup of G. o(H) = 7, o(K) = 5 ⇒ H ∩ K = {e}. Let

x ∈ H, o(x) = 7, y ∈ K, o(y) = 5. Now xyx−1 y −1 = x · yx−1 y −1 ∈ H, xyx−1 · y −1 ∈ K ⇒

xyx−1 y −1 = e ⇒ xy = yx ⇒ o(xy) = 35. Thus G is cyclic, xy being a generator.

More generally, o(G) = pq, p < q, p - q − 1 ⇒ G is cyclic of order pq, using a similar

argument.

Question 1(b) Show that the polynomial 25x4 + 9x3 + 3x + 3 is irreducible over the field of

rational numbers.

is a polynomial with integer coefficients and if there is a prime p such that p | ai , 0 ≤ i <

n, p2 - a0 , p - an then f (x) is irreducible over the rationals. (Proof: see theorem 3.10.2 page

160 of Topics in Algebra by Herstein). In the present case p = 3 does the trick.

Question 2(a) 1. Show that a group of order p2 is abelian where p is a prime number.

Solution.

1

1. o(G) = p2 . Let C be the center of G. Then the center of G is nontrivial ∵ o(G) is a

power of a prime. If o(C) = p2 , then G = C ⇒ G is abelian. If o(C) = p, then G/C is

cyclic of order p. Let G/C be generated by aC. Let x, y ∈ G ⇒ xC = ar C, yC = as C

for some r, s, 1 ≤ r, s < p. Then

⇒ x = ar c1 , y = as c2 for some c1 , c2 ∈ C

⇒ xy = ar c1 as c2

⇒ xy = ar as c1 c2 ∵ c1 as = as c1

⇒ xy = as c2 ar c1 = yx

Thus G is abelian.

≡ 1 mod 7 and divides 42. If this number is 7r + 1, then 7r + 1 | 42 ⇒ 7r + 1 | 6 ⇒

r = 0. Thus this subgroup H is unique. Consider aHa−1 , a ∈ G. o(aHa−1 ) = o(H) =

7 ∴ aHa−1 = H as H is the unique subgroup of order 7. Thus H is a normal subgroup

of order 7.

Question 2(b) Prove that in the ring F [x] of polynomials over a field F , the ideal I = [p(x)]

is maximal ⇐⇒ p(x) is irreducible over F .

Solution. Let I be maximal. If g(x) is any divisor of p(x), then [g(x)] ⊇ I ⇒ [g(x)] = F [x]

or [g(x)] = [p(x)]. Thus g(x) is a unit or g(x) is an associate of p(x). Thus p(x) is irreducible.

Conversely let p(x) be irreducible. Let M be an ideal, M ⊇ [p(x)]. Since F [x] is Euclidean

and therefore a Principal Ideal Domain, M = [f (x)] say. Then p(x) ∈ [f (x)] ⇒ f (x) | p(x).

Thus f (x) is a unit or f (x) is an associate of p(x) ⇒ M = F [x] or M = [p(x)] ⇒ I is

maximal.

Show that E has q n elements.

Solution.

α ∈ E ⇒ α = α1 x1 + . . . + αn xn where α1 , . . . , αn ∈ F are uniquely determined by α.

Thus E = {α1 x1 + . . . + αn xn | αi ∈ F } ' F n =⇒ |E| = q n , as each αi has q choices.

2

UPSC Civil Services Main 2003 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

prove that H is a normal subgroup of G.

Solution. Let h ∈ H, g ∈ G. Then h(h−1 g −1 )2 g 2 (g −1 hg)2 = hh−1 g −1 h−1 g −1 g 2 g −1 hgg −1 hg =

g −1 hg. Now x ∈ G ⇒ x2 ∈ H, therefore (h−1 g −1 )2 , g 2 , (g −1 hg)2 ∈ H Consequently for any

h ∈ H, g −1 hg ∈ H. Thus H is a normal subgroup of G.

Alternative solution. We shal prove that Hx = xH for every x ∈ G. Clearly for

any h ∈ H, xh = xh.xh.h−1 x−1 x−1 x = h1 x, where h1 = (xh)2 h−1 x−2 ∈ H, this shows that

xH ⊆ Hx. Similarly hx = xx−1 x−1 h−1 hxhx = xh1 with h1 = x−2 h−1 (hx)2 ∈ H. Thus

Hx ⊆ xH, so xH = Hx for every x ∈ G. Hence H is a normal subgroup of G.

√

Question 1(b) Show that the ring Z[i] = {a + bi | a, b ∈ Z, i = −1} of Gaussian integers

is a Euclidean domain.

Solution. For α = a + ib ∈ Z[i], we define N (α) = a2 + b2 . Clearly (i) N (α) > 0 for α 6= 0,

(ii) For α 6= 0, β 6= 0, N (αβ) = N (α)N (β). Let α = a + ib, β = c + id 6= 0. We shall find

γ, δ ∈ Z[i] such that α = βγ + δ where δ = 0 or N (δ) < N (β). This will prove Z[i] is a

Euclidean domain for the Euclidean function N (α).

α a + ib (a + ib)(c − id)

= = = p + iq

β c + id c2 + d2

where p, q are rational numbers. We determine integers x, y so that |p−x| ≤ 21 , |q −y| ≤ 1

2

— x, y are the integers nearest to p, q respectively. Let γ = x + iy. Then

α

= γ + (p − x) + (q − y)i ⇒ α = βγ + βη = βγ + δ

β

where δ = βη. Clearly δ = α − βγ is a Gaussian integer, and if δ 6= 0, then N (δ) =

N (β)[(p − x)2 + (q − y)2 ] ≤ N (β)[ 41 + 41 ] < N (β). This completes the proof.

1

Question 2(a) 1. Let R be the ring of all real-valued continuous functions on the closed

interval [0, 1]. Let M = {f (x) ∈ R | f ( 31 ) = 0}. Show that M is a maximal ideal of

R.

either M ⊆ N or N ⊆ M .

Solution.

1. M is an ideal. M 6= ∅ since the function f (x) = 0 clearly belongs to M .

Let f, g ∈ M then the function (f − g)(x) = f (x) − g(x) is continuous everywhere on

[0, 1] and (f − g)( 13 ) = f ( 13 ) − g( 31 ) = 0, so f − g ∈ M . Thus M is a subgroup of the

group (R, +).

If g ∈ M and f ∈ R, then the function (f g)(x) = f (x)g(x) is continuous everywhere

on [0, 1] and (f g)( 31 ) = f ( 13 )g( 13 ) = 0 as g( 31 ) = 0, thus f g ∈ M . Thus M is an ideal of

R. Note that R is a commutative ring with unity I, where I(x) = 1.

Let M ⊆ A ⊆ R where A is an ideal of R. If M 6= A, we shall show that A = R. Let

β ∈ A − M , thus β( 13 ) = c 6= 0. Define α : [0, 1] −→ [0, 1] by α(x) = c for all x ∈ [0, 1].

Then the function µ = β − α ∈ M ⊂ A as µ( 31 ) = 0. Thus α = β − µ ∈ A as β, µ ∈ A.

Now consider γ : [0, 1] −→ [0, 1] defined by γ(x) = 1c for all x. Clearly γ ∈ R. Since A

is an ideal, γα ∈ A. But γα(x) = 1c c = 1, thus γα = I ∈ A. Since I is unity in R, it

follows that A = R, hence M is a maximal ideal of R.

Note: The converse of the above statement is also true i.e. if M is a maximal ideal

of R, then there exists number r ∈ [0, 1] such that M = {f | f ∈ R, f (r) = 0}. The

proof needs compactness of [0, 1] which is not an algebraic concept.

is an ideal of R.

Conversely, let M ∪ N be an ideal of R. If possible, let us assume that M * N and

N * M , this means there exist x ∈ M − N, y ∈ N − M . Now x ∈ M, y ∈ N ⇒ x, y ∈

M ∪ N . But M ∪ N is an ideal, thus x − y ∈ M ∪ N , hence x − y ∈ M or x − y ∈ N . If

x − y ∈ M , then x − (x − y) = y ∈ M as M is an ideal, but this is a contradiction. If

x − y ∈ N , then (x − y) + y = x ∈ N , which is also not possible. Thus our assumption

that M * N and N * M is incorrect, so if M ∪ N is an ideal, either M ⊆ N or

N ⊆ M.

√

Question 2(b) 1. Show that Q( 3, i) is the splitting field for x5 − 3x3 + x2 − 3 where Q

is the field of rational numbers.

2. Prove that x2 + x + 4 is irreducible over F , the field of integers modulo 11 and prove

further that F [x]/hx2 + x + 4i is a field with 121 elements.

2

Solution.

1. x5 − 3x3 + x2 − 3 = x3 (x2 − 3) + x2 − 3 = (x2 − 3)(x3 +√1) = (x2 − 3)(x + 1)(x2 − x + 1).

√

Thus the roots of x5 − 3x3 + x2 − 3 are −1, ± 3, 1±i2 3 . Consequently all the roots of

√

the given polynomial lie in the field Q( 3, i). Conversely, if K√ is any field containing

√ the roots of the √given polynomial, then 3 ∈ K, and therefore

Q, which contains

i ∈ K, thus Q( 3, i) ⊆ K. Thus √ Q( 3, i) is the smallest field containing all the roots

5 3 2

of x − 3x + x − 3. Thus Q( 3, i) is the splitting field of the given polynomial over

Q.

2. See question 2(c) from 1996 for the irreducibility of x2 + x + 4 over F .

See question 2(b) from 1992 for the second part.

Question 2(c) If R is a unique factorization domain (UFD), then prove that R[x] is also

a UFD.

Solution. Let F denote the field of quotients of R.

Result 1. If f (x) ∈ R[x] is irreducible, then f (x) remains irreducible in F [x]. (Note that

the converse is obvious as R[x] ⊆ F [x].) Let f (x) be reducible in F [x] i.e. f (x) = g(x)h(x),

where deg g(x) < deg f (x), deg h(x) < deg f (x) and g(x), h(x) ∈ F [x]. We can write g(x) =

a1 b−1 −1

1 g1 (x), h(x) = a2 b2 h1 (x), where g1 (x), h1 (x) ∈ R[x] and are primitive and a1 , b1 , a2 , b2 ∈

R (b1 is the LCM of all the denominators of g(x), and a1 is the GCD of the numerators). Thus

b1 b2 f (x) = a1 a2 g1 (x)h1 (x). But by Gauss Lemma, the product of two primitive polynomials

is primitive, therefore g1 (x)h1 (x) is primitive. Since f (x) is irreducible in R[x], therefore it

is also primitive. Consequently b1 b2 = content of b1 b2 f (x) = a1 a2 = content of a1 a2 g(x)h(x)

and therefore we get f (x) = g1 (x)h1 (x), thus f (x) is reducible in R[x]. Hence if f (x) is

irreducible in R[x] then it is irreducible in F [x].

Result2. Factorization exists in R[x]. Let f (x) ∈ R[x], f (x) 6= 0 and f (x) not a unit.

Let a = c(f ) = content of f then f = af ∗ where f ∗ is a primitive polynomial in R[x] of the

same degree as f . Since F [x] is a UFD (being a Euclidean domain), we can write f ∗ (x) =

p1 (x) . . . pr (x), where each pi (x) is an irreducible element of F [x]. Let pi (x) = ai b−1i qi (x),

where ai , bi ∈ R, and qi (x) ∈ R[x] is a primitive polynomial. Thus we get

b1 . . . br f ∗ (x) = a1 . . . ar q1 (x) . . . qr (x)

But the product q1 (x) . . . qr (x) is again primitive (Gauss Lemma), therefore equating the

contents of both sides (note that f ∗ (x) is primitive), we get b1 . . . br = a1 . . . ar , therefore

f ∗ (x) = q1 (x) . . . qr (x)

where each qi (x) ∈ R[x] and is irreducible in F [x] and therefore irreducible in R[x]. Since R

is a UFD, a = π1 . . . πt , where π1 , . . . , πt are irreducible in R. Thus

f (x) = π1 . . . πt q1 (x) . . . qr (x)

3

where π1 , . . . , πt , q1 (x), . . . , qr (x) are irreducible elements of R[x]. Note that π1 , . . . , πt being

constants cannot have a proper factorization in R[x] if they do not have one in R. Hence

the result is established.

Result 3. Uniqueness. If possible, let

where π1 , . . . , πt , π10 , . . . , πu0 are irreducible elements in R and q1 (x) . . . qr (x), g1 (x) . . . gs (x) are

irreducible elements of R[x]. Using Gauss Lemma, we get that the products q1 (x) . . . qr (x),

g1 (x) . . . gs (x) are primitive. Comparing the contents of both sides, we get π1 . . . πt =

π10 . . . πu0 . But R is a UFD, so t = u, and we can reorder the πi0 to ensure that each πi

is an associate of πi0 . Thus we are left with q1 (x) . . . qr (x) = g1 (x) . . . gs (x). We consider this

equation in F [x]. By the first result each qi , gj , 1 ≤ i ≤ r, 1 ≤ j ≤ s is irreducible in F [x].

Since F [x] is a UFD, we get r = s and by reordering, we get that qi (x) is an associate of

gi (x) in F [x]. We can assume w.l.o.g. that qi (x) = (unit in F [x])gi (x), 1 ≤ i ≤ r. Since

units in F [x] are non-zero constants, these are of the form cd−1 where c, d ∈ R. Thus we

get di qi (x) = ci gi (x). Using contents, we conclude that di = ci , thus qi (x) is an associate of

gi (x) in R[x], so the factorization is unique.

Thus R[x] is a UFD.

4

UPSC Civil Services Main 2004 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) If p is a prime number of the form 4n + 1, n a natural number, then show

that the congruence x2 ≡ −1 mod p is solvable.

Solution. Consider the multiplicative group G of non-zero residue classes modulo p. In this

group [1] and [p − 1] are the only two elements which are there own inverses as the equation

x2 = [1] has exactly two solutions in the field Z/pZ. Since order of G is φ(p) = p − 1 = 4n,

the remaining 4n − 2 elements form 2n − 1 pairs, in each pair each element is the inverse of

the other. Thus Y

[r] = [1]

1<r<p−1

as each one of the 2n − 1 pairs when multiplied would give us [1]. Consequently

Y

[r] = [p − 1] =⇒ (p − 1)! ≡ p − 1 ≡ −1 mod p

1≤r≤p−1

p − 1 p + 1

(p − 1)! = 1 · 2 · ... · · . . . · (p − 1)

2 2

p − 1 p − 1 p − 3

= 1 · 2 · ... · p− p− . . . (p − 1))

2 2 2

p−1 p − 1 2

≡ (−1) 2 ( )! mod p

2

p−1

Since p ≡ 1 mod 4, (−1) = 1 and we get

2

p − 1 2

( )! ≡ (p − 1)! ≡ −1 mod p

2

showing that the congruence x2 ≡ −1 mod p is solvable.

1

Question 1(b) Let G be a group and let a, b ∈ G. If ab = ba and (O(a), O(b)) = 1 then

show that O(ab) = O(a)O(b).

Solution. Let O(a) = l, O(b) = m and O(ab) = k. Now (ab)lm = alm blm because ab = ba.

But alm = (al )m = e, blm = (bm )l = e therefore (ab)lm = e and consequently k divides lm.

Also e = (ab)k = ak bk ⇒ ak = b−k ⇒ akm = b−km = e ⇒ l | km, but (l, m) = 1, therefore

l | k. Considering e = akl = b−kl , we get m | k. Since (l, m) = 1, we get lm | k. Hence

k = lm completing the proof.

Question 2(a) Verify that the set E of the four roots of x4 − 1 = 0 forms a multiplicative

group. Also prove that a transformation T, T (n) = in is a homomorphism from I+ (group of

all integers with addition) onto E under multiplication.

2πi 4πi 6πi 8πi πi 3πi

Solution. Clearly E = {e 4 , e 4 , e 4 , e 4 } = {e 2 , eπi , e 2 , e2πi } = {1, −1, i, −i}. The

following multiplication table shows that E is a multiplication group.

1 −1 i −i

1 1 −1 i −i

−1 −1 1 −i i

i i −i −1 1

−i −i i 1 −1

In fact

1. α, β ∈ E ⇒ αβ ∈ E

from I+ to E and it is clearly onto (note that T (0) = 1, T (1) = i, T (2) = −1, T (3) = −i).

Moreover T is a homomorphism is obvious as T (m + n) = im+n = im in = T (m)T (n).

Question 2(b) Prove that if the cancellation law holds for a ring R then a(6= 0) ∈ R is not

a zero divisor and conversely.

Solution. Assume the cancellation law holds. If a 6= 0 and ab = 0 for some b ∈ R, then we

get ab = a0 and since a 6= 0, the cancellation law gives us b = 0, showing that a is not a zero

divisor. Conversely assume R has no zero divisors. Let a 6= 0 and ax = ay ⇒ a(x − y) = 0.

This should imply x − y = 0 because otherwise a will be a zero divisor. Thus ax = ay, a 6=

0 =⇒ x = y i.e. cancellation law holds. This completes the proof.

2

Question 2(c) Show that the residue class ring Z/(m) is a field if m is a prime number.

mod m, y ≡ b mod m ⇒ x + y ≡ a + b mod m.

4. If [a] ∈ Z/(m) then [−a] ∈ Z/(m) and [a] + [−a] = [a + (−a)] = [0], hence [−a] is the

additive inverse of [a].

5. [a] + ([b] + [c]) = [a] + [b + c] = [a + b + c] = [a + b] + [c] = ([a] + [b]) + [c] showing the

operation is additive.

mod m. This shows that Z/(m) is closed with respect to the operation of multiplication

of residue classes modulo m. Moreover this operation is commutative.

8. [a]([b] + [c]) = [a][b + c] = [a(b + c)] = [ab + ac] = [a][b] + [a][c] and ([a] + [b])[c] =

[a][c] + [b][c].

Thus Z/(m) is a commutative ring with identity. To show that it is a field, we have to show

that every non-zero element is invertible.

Let [a] ∈ Z/(m), [a] 6= [0]. This means that a 6≡ 0 mod m. Since m is a prime, it

follow that (a, m) = 1, and therefore there exist integers b and c such that ab + cm = 1.

Consequently ab ≡ 1 mod m, or [a][b] = [1] i.e. [a] is invertible and [b] is its inverse.

Hence Z/(m) is a field when m is a prime.

Note: Z/(m) is not even an integral domain when m > 1 and is not prime. m = bc where

1 < b < m, 1 < c < m and therefore [b], [c] 6= [0] but [b][c] = [bc] = [0], showing that Z/(m)

has zero divisors and is not an integral domain.

Question 2(d) Define an irreducible element and a prime element in an integral domain

D with unit. Prove that every prime element in D is irreducible, but the converse of this is

not in general true,

an irreducible element if a = bc implies that either b or c is a unit (consequently either b or

c is an associate of a).

3

Prime. An element a 6= 0, a not a unit is is said to be a prime element if a | bc ⇒ a | b

or a | c.

Every prime is irreducible. Let a be a prime element in D. If possible let a = bc, we

shall show that either b or c is a unit. Since a | bc and a is a prime element, a | b or a | c. If

a | b then there exists x ∈ D such that b = xa =⇒ a = xac. But D is an integral domain

and therefore cancellation holds. Thus a = xac ⇒ 1 = xc i.e. c is a unit. Similarly we can

show that if a | c then b is a unit. Thus a is an irreducible element i.e. it has no proper

divisors.

Example where an irreducible need not be prime.

√ √

D = Z[ −5] = {a + b −5 | a, b ∈ Z}

√

We define for α ∈ D, N (α) = a2 + 5b2 where α = a + b −5. Clearly for α, β ∈ D,

N (αβ) = N (α)N (β). Moreover α ∈ D is a unit if and only if N (α) = 1. Thus D has only

two units namely ±1.

Now we show that 2 is irreducible but not a prime. If possible, let 2 = αβ, then

N (2) = 4 = N (α)N (β), showing that N (α) = 1, 2, 4 ⇒ √ N (β) = 4, 2, 1. If N (α) = 1, then α

is a unit, and if N (α) = 4, then β is a unit. If α = a + b −5, then N (α) = 2 ⇒ a + 5b2 = 2,

which is not possible for√a, b ∈ Z. Hence

√ 2 is irreducible.

√ √

However√2 · 3 = (1 + √ −5)(1 − −5), i.e. 2 | (1 + −5)(1 √ − −5) but 2 does not divide

either (1 + −5) or (1 − −5), √ because N (2) = 4, N (1 ± −5) = 6 and 4 - 6. Hence 2 is

irreducible but not prime in Z[ −5].

4

UPSC Civil Services Main 2005 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) If M and N are normal subgroups of a group G such that M ∩ N = {e},

show that every element of M commutes with every element of N .

and y ∈ N ⊆ G, and M is a normal subgroup of G, therefore yx−1 y −1 ∈ M , consequently

α ∈ M . Similarly since N is a normal subgroup of G and y ∈ N , xyx−1 ∈ N , hence

α = xyx−1 y −1 ∈ N . Thus α ∈ M ∩ N , which means that α = xyx−1 y −1 = e ⇒ xy = yx i.e.

every element of M commutes with every element of N .

Question 1(b) Show that (1 + i) is a prime element in the ring R of Gaussian integers.

Solution. The ring of Gaussian integers is a Euclidean domain with Euclidean function

N (a + ib) = a2 + b2 , therefore any two elements α, β ∈ R have a GCD (greatest common

divisor). If d is the GCD of α, β, then there exist γ, δ ∈ R such that αγ + βδ = d. Moreover

α is a unit in R if and only if N (α) = 1, because if N (α) = 1 then αα = 1, implying that

α is a unit, and conversely, if α is a unit, then there exist β ∈ R such that αβ = 1, and

therefore N (αβ) = N (α)N (β) = 1 ⇒ N (α) = N (β) = 1 as both are positive integers.

First of all we prove that 1 + i is an irreducible element (note that it is not a unit as

N (1 + i) = 2). Let 1 + i = αβ. Taking norm of both sides, we get N (αβ) = N (α)N (β) =

2 ⇒ N (α) = 1 or N (β) = 1, so either α is a unit or β is a unit. Thus 1 + i is an irreducible

element.

Let 1 + i divide αβ and assume that 1 + i does not divide α. We shall show that 1 + i

divides β. Since the only divisors of 1 + i are 1 + i and units, and 1 + i does not divide α, it

follows that GCD of α and 1 + i is 1. Thus there exists γ, δ ∈ R such that γ(1 + i) + δα = 1

or γβ(1 + i) + δαβ = β. Since (1 + i) divides the left hand side of this equation, it follows

that 1 + i divides β. Hence 1 + i is a prime element in R.

1

Question

p 2(a) p H and K be two subgroups of a finite group G, such that |H| >

1. Let

|G| and |K| > |G|. Prove that H ∩ K 6= {e}.

of f (a).

Solution.

|H||K|

1. We prove that |HK| = |H∩K|

.

If H ∩ K = {e}, then hk = h1 k1 ⇔ h−1 1 h = k1 k

−1

⇔ h−1 1 h, k1 k

−1

∈ H ∩ K ⇔ h−11 h =

k1 k −1 = e ⇔ h = h1 , k = k1 . Thus there are no repetitions in HK = {hk | h ∈

|H||K|

H, k ∈ K}, so |HK| = |H||K| = |H∩K| . (This is sufficient to prove the result, but for

completeness we show the result when H ∩ K 6= {e}.)

If H ∩ K 6= {e}, then hk = h1 k1 ⇔ h−1 1 h, k1 k

−1

∈ H ∩ K ⇔ h−11 h = k1 k

−1

=u∈

−1

H ∩ K ⇔ h = h1 u, k = u k1 with u ∈ H ∩ K. Thus hk is duplicated at least |H ∩ K|

times as hk = (hu)(u−1 k) with u ∈ H ∩K. It is duplicated no more than |H ∩K| times,

|H||K|

because hk = h1 k1 ⇒ h = h1 u, k = u−1 k1 with u ∈ H ∩ K. Hence |HK| = |H∩K| .

√ √

|H||K| |G| |G|

Now |G| ≥ |HK| = |H∩K| ≥ |H∩K| Thus |H ∩ K| > 1, so |H ∩ K| 6= {e}.

2. Let o(a) = order of a = m and order of f (a) = o(f (a)) = n. Then e0 = f (am ) = f (a)m ,

where e0 is the identity of G0 , showing that n divides m. Conversely, f (e) = e0 =

f (a)n = f (an ) ⇒ an = e as f is one-one. This means that m divides n. Thus m = n,

which was to be proved.

Question 2(b) Prove that any polynomial ring F [x] over a field F is a UFD.

Solution. We know that F [x] is a Euclidean domain with the Euclidean function being

the degree of the polynomial — the algorithm being: given f (x), g(x) 6= 0 belonging to

F [x], there exist q(x), r(x) ∈ F [x] such that f (x) = q(x)g(x) + r(x) where r(x) = 0 or

deg r(x) < deg g(x).

Step 1. If f (x), g(x) ∈ F [x], both not 0, then they have a GCD d(x), and there exist

λ(x), µ(x) ∈ F [x] such that d(x) = f (x)λ(x) + g(x)µ(x). Let S = {f (x)a(x) + g(x)b(x) |

a(x), b(x) ∈ F [x]}. Then S 6= ∅, as f (x), g(x) ∈ S. Let d(x) be a non-zero polynomial is S

with minimal degree, i.e. deg d(x) ≤ deg h(x) for every nonzero h(x) ∈ S. Clearly if any d0 (x)

divides f (x) and g(x), then d0 (x) divides d(x) because d(x) is of the form f (x)a(x)+g(x)b(x).

Moreover d(x) divides both f (x) and g(x), otherwise we have q(x), r(x) ∈ F [x] such that

f (x) = d(x)q(x) + r(x) where deg r(x) < deg d(x), but this is not possible as r(x) ∈ S as it is

of the form f (x)a(x) + g(x)b(x) so deg r(x) ≥ deg d(x). So d(x) divides f (x), and similarly

d(x) divides g(x).

2

Step 2. An irreducible element of F [x] is a prime element i.e. if f (x) is irreducible and

f (x) | g(x)h(x) and f (x) - g(x) then f (x) | h(x).

If f (x) - g(x), then f (x) is irreducible implies its only divisors are units or associates of

f (x). Therefore the GCD of f (x) and g(x) is 1. By Step 1, we have 1 = f (x)a(x) + g(x)b(x)

for some a(x), b(x) ∈ F [x]. Thus h(x) = h(x)f (x)a(x) + h(x)g(x)b(x). Clearly f (x) divides

the right hand side, so f (x) | h(x), as required.

Step 3. Every non-zero non-unit element in F [x] can be written as the product of

irreducible elements in F [x].

The proof is by induction on the degree of f (x). If deg f (x) = 0, then f (x) is a non-zero

constant, therefore a unit in F [x], so we have nothing to prove.

Let the result be true for all polynomials whose degree is < deg f (x). If f (x) is irre-

ducible, we have nothing to prove. If f (x) is not irreducible, then there exist g(x), h(x),

1 ≤ deg g(x), deg h(x) < deg f (x) such that g(x)h(x) = f (x). Now by induction both g(x)

and h(x) are products of irreducible elements, therefore f (x) is the product of irreducible

elements.

Step 4: Uniqueness. If possible let

the gi ’s can be reordered such that each fi is the associate of gi .

Now f1 (x) divides g1 (x) . . . gs (x), therefore by step 2, f1 (x) must divide one of g1 (x), . . . , gs (x).

Let us assume without loss of generality that f1 (x) | g1 (x), but g1 (x) is also irreducible and

f1 (x) is not a unit, therefore f1 (x) and g1 (x) are associates. Thus we get

If r < s, then after r steps we shall get gr+1 (x) . . . gs (x) = 1, which is not possible, hence

r ≥ s, similarly s ≥ r so r = s. Now by relabelling g1 , . . . , gr we get each fi (x) is an associate

of gi (x), 1 ≤ i ≤ r. Hence F [x] is a UFD.

3

UPSC Civil Services Main 2006 - Mathematics

Algebra

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Let S be the set of all real numbers except −1. Define ∗ on S by

a ∗ b = a + b + ab

2∗x∗3=7

in S.

Solution. Clearly S 6= ∅.

1. S is closed for the operation (∗). If a+b+ab = −1, then a+b+ab+1 = (a+1)(b+1) =

0 ⇒ a = −1 or b = −1. Thus a, b ∈ S ⇒ a 6= −1, b 6= −1 ⇒ a+b+ab 6= −1 ⇒ a∗b ∈ S.

a a a2

3. a 6= −1, then b = − 1+a 6= −1 and a ∗ b = b ∗ a = a − 1+a

− 1+a

= 0, thus S is closed

with respect to inverses for the operation (∗).

4. a ∗ b = b ∗ a for every a, b ∈ S.

a + b + c + bc + ab + ac + abc. Thus (a ∗ b) ∗ c = a ∗ (b ∗ c) thus the operation (∗) is

associative.

2 ∗ x ∗ 3 = 2 + x + 3 + 2x + 3x + 6 + 6x. Therefore we want 12x + 11 = 7, so x = − 31 .

1

Question 1(b) If G is a group of real numbers under addition and N is the subgroup of

G consisting of the integers, prove that G/N is isomorphic to the group H of all complex

numbers of absolute value 1 under multiplication.

2. Let z be any complex number with |z| = 1, then z 6= 0. Let θ = arg z, then

θ

f( ) = eiθ = z

2π

fundamental theorem of homomorphisms G/N is isomorphic to H.

Alternative solution. Let f be as defined above. Define φ : G/N −→ H by φ(α) =

φ(α + N) = f (α) for α ∈ G. Then

1. φ is well defined i.e.. if α = β then φ(α) = φ(β) i.e. φ does not depend on the choice

of representative in the coset. Clearly α = β ⇔ α − β ∈ N ⇒ e2iπα = e2iπβ ⇒ f (α) =

f (β).

φ(α) 6= φ(β).

f (α) = z (as above) then φ(α) = f (α) = z.

Question 2(a) 1. Let O(G) = 108. Show that there exists a normal subgroup of order

27 or 9.

2. Let G be the set of all those ordered pairs (a, b) of real numbers for which a 6= 0 and

define in G an operation ⊗ as follows:

abelian?

Solution.

2

1. According to one of the Sylow theorems, the number of subgroups of G of order 27 is

≡ 1 (modulo 3) and is a divisor of 108 and therefore of 4, thus the number of such

subgroups is 1 or 4. If G has a unique Sylow group H of order 27, then it has to be a

normal subgroup because O(a−1 Ha) = 27 and therefore a−1 Ha = H for every a ∈ G.

Let us therefore assume that G has more than one subgroup of order 27. Then G has

four subgroups of order 27, say H1 , H2 , H3 , H4 .

We first of all observer that Hi ∩ Hj must have at least 9 elements, because if not, then

|H ||H |

|Hi Hj |, the number of elements in Hi Hj , would be at least 243 as |Hi Hj | = |Hii∩Hjj | ,

and this is not possible. Let H = Hi ∩ Hj , i 6= j, then O(H) = 9, because Hi 6= Hj .

Now NHi (H), the normalizer of H in Hi , contains H properly (see 1995 question 1(b)),

showing that NHi (H) = Hi and similarly NHj (H) = Hj . Thus NG (H) ⊇ Hi as well as

Hj and therefore O(NG (H)) ≥ 81 and is divisor of 108. Hence NG (H) = G and H is

a normal subgroup of G. Thus G has a normal subgroup of order 27 or of order 9.

(a) G is closed with respect to the operation ⊗ i.e. (a, b), (c, d) ∈ G ⇒ (a, b) ⊗ (c, d) ∈

G.

(b) (1, 0) is identity of G w.r.t. ⊗ as (a, b)(1, 0) = (a, b) = (1, 0) ⊗ (a, b)

(c) If (a, b) ∈ G, then (a−1 , −ba−1 ) ∈ G as a 6= 0, and (a, b) ⊗ (a−1 , −ba−1 ) = (1, 0) =

(a−1 , −ba−1 )(a, b). Thus every element of G has an inverse w.r.t. the operation

⊗ and it belongs to G.

(d) (a, b) ⊗ ((c, d) ⊗ (e, f )) = (a, b) ⊗ (cd, de + f ) = (ace, bce + de + f ) = ((a, b) ⊗

(c, d)) ⊗ (e, f )

(2a, 2b) whereas (2, 0) ⊗ (a, b) = (2a, b) showing that (2, 0) ⊗ (a, b) 6= (a, b) ⊗ (2, 0) when

b 6= 0.

√ √

Question 2(b) Show that Z[ 2] = {a + 2b | a, b ∈ Z} is a Euclidean domain.

exists a function g : R − {0} −→ Z (ring of integers) such that

1. g(a) ≥ 0 for every a ∈ R∗ = R − {0}.

where r = 0 or g(r) < g(b).

√ √

For α ∈ Z[ 2], α = a + b 2, a, b ∈ Z, we define N (α) = a2 − 2b2 and g(α) = |N (α)|. Clearly

3

√

1. g(α) ≥ 0 for every α ∈ Z[ 2], α 6= 0.

√

2. For α, β ∈ Z[ 2], α 6= 0, β 6= 0, g(αβ) = g(α)g(β) ≥ g(α) because g(β) ≥ 1.

√ √

Note that if α = a + b 2, β = c + d 2, then

= a2 c2 + 4b2 d2 − 2a2 d2 − 2b2 c2

= (ac + 2bd)2 − 2(ad + bc)2

√

= N (ac + 2bd + 2(ad + bc)

= N (αβ)

√ √ √ √

3. Let α = a + b 2 ∈ Z[ 2] and β = c + d 2 ∈ Z[ 2] and β 6= 0. Clearly

√ √

α (a + b 2)(c − d 2) √

= √ √ =p+q 2

β (c + d 2)(c − d 2)

2 −2d2 , q = c2 −2d2 are rational numbers. Let m, n be the integers nearest to

and [p] is the integral part of p, then m = [p] if θ ≤ 21 and m = [p] + 1 if θ > 12 .

Let p − m = r, q − n = s, then |r| ≤ 12 , |s| ≤ 12 . Now

√ √ √

α = a + b 2 = (c + d 2)(p + q 2)

√ √

= (c + d 2)((m + r) + (n + s) 2)

√ √ √ √

= (c + d 2)(m + n 2) + (c + d 2)(r + s 2)

√ √ √ √

Let γ = m + n √2, δ = (c + d 2)(r + s 2), then α = βγ + δ, where γ ∈ Z[ 2] and

δ = α − βγ ∈ Z[ 2].

Now either δ = 0 or g(δ) = |N (β)||r2 − 2s2 |. But |r2 − 2s2 | ≤ 41 + 24 < 1, therefore

√ √

g(δ) < g(β). Thus given α, β ∈ Z[ 2], β 6= 0, we have found γ, δ ∈ Z[ 2] such that

α = βγ + δ where δ = 0 or g(δ) < g(β).

√

This shows that Z[ 2] is a Euclidean domain.

4

UPSC Civil Services Main 1979 - Mathematics

Calculus

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

n Z x (n+1)

X f (k) (a) k f (t)(x − t)n

f (x) = (x − a) + dt

k=0

k! a n!

f (n+1) (X)(x − X)n

(x − a)

n!

where a < X < x.

Solution. Carrying out integration by parts on the given integral repeatedly, we get

Z x (n+1) x Z x (n)

f (t)(x − t)n f (n) (t)(x − t)n f (t)(x − t)n−1

dt = + dt

a n! n! a a (n − 1)!

Z x (n)

f (n) (a)(x − a)n f (t)(x − t)n−1

= − + dt

n! a (n − 1)!

Z x (n−1)

f (n) (a)(x − a)n f (n−1) (a)(x − a)n−1 f (t)(x − t)n−2

= − − + dt

n! (n − 1)! a (n − 2)!

= ...

n Z x

X f (r) (a)(x − a)r

= − + f 0 (t) dt

r=1

r! a

n

X f (r) (a)(x − a)r

= − + f (x) − f (a)

r=1

r!

n

X f (r) (a)(x − a)r

= − + f (x)

r=0

r!

1

Thus n x

f (r) (a)(x − a)r f (n+1) (t)(x − t)n

X Z

f (x) = + dt

r=0

r! a n!

as required. Now we use the result — if f (x) is continuous in [a, b], then

Z b

f (x) dx = f (µ)(b − a) for some µ ∈ (a, b)

a

Rb

This is obvious as m(b − a) ≤ a f (x) dx ≤ M (b − a), where m, M are respectively the

min and max of f (x) over [a, b]. and every value between m and M is attained (by the

intermediate value theorem).

Since f ∈ C n+1 , f (n+1) is continuous and we can use the above result. Therefore

Z x (n+1)

f (t)(x − t)n f (n+1) (X)(x − X)n

dt = (x − a)

a n! n!

where a < X < x.

Question 1(b) If 2 2

2xy x − y ,

(x, y) 6= (0, 0)

f (x, y) = x2 + y 2

0, (x, y) = (0, 0)

∂ 2f ∂ 2f

show that 6= . Explain this result.

∂x ∂y ∂y ∂x

Solution.

f (0, k) − f (0, 0) 0

fy (0, 0) = lim = lim = 0

k→0 k k→0 k

2 −k 2

f (h, k) − f (h, 0) 2hk hh2 +k 2

fy (h, 0) = lim = lim = 2h

k→0 k k→0 k

∂ 2f fy (h, 0) − fy (0, 0) 2h

(0, 0) = lim = lim =2

∂x ∂y h→0 h h→0 h

f (h, 0) − f (0, 0) 0

fx (0, 0) = lim = lim = 0

h→0 h h→0 h

2 2

f (h, k) − f (0, k) 2hk hh2 −k

+k2

fx (0, k) = lim = lim = −2k

h→0 h h→0 h

∂ 2f fx (0, k) − fx (0, 0) −2k

(0, 0) = lim = lim = −2

∂y ∂x k→0 k k→0 k

∂ 2f ∂ 2f

Thus 6= at (0, 0). The reason for this is that neither fxy nor fyx is continuous

∂x ∂y ∂y ∂x

at (0, 0). This implies that neither fx (x, y) nor fy (x, y) is differentiable at (0, 0). Thus the

criteria of Young’s or Schwartz theorem are not satisfied.

2

R1

Question 2(a) If B(x, y) = 0

tx−1 (1 − t)y−1 dt, x > 0, y > 0, then show that B(x, x) =

21−2x B(x, 12 ).

=

Γ(x + 12 ) Γ(2x)

Γ( 21 )Γ(x) 1 [Γ(x)]2

Since = B(x, ) and = B(x, x), we get B(x, x) = 21−2x B(x, 12 ) as required.

Γ(x + 12 ) 2

Γ(2x)

Paper II

Question 3(a) Find the maximum of x21 x22 . . . x2n under the restriction x21 +x22 +. . .+x2n = 1.

Using the result derive the inequality

1 a1 + . . . + an

(a1 a2 . . . an ) n ≤

n

for positive real numbers a1 , . . . , an .

Solution. Let F (x1 , . . . , xn ) = x21 x22 . . . x2n + λ( ni=1 x2i − 1), where λ is Lagrange’s unde-

P

termined multiplier. For extreme values,

∂F

= 2x21 x22 . . . x2i−1 xi x2i+1 . . . x2n + 2λxi = 0, 1 ≤ i ≤ n

∂xi

Since x21 +x22 +. . .+x2n = 1, (x1 , . . . , xn ) 6= (0, . . . , 0), it follows that λ = −x21 x22 . . . x2i−1 x2i+1 . . . x2n

for 1 ≤ i ≤ n. Thus x21 = x22 = . . . = x2n = n1 , and λ = − nn−1 1

.

∂ 2F

= 2x21 x22 . . . x2i−1 x2i+1 . . . x2n + 2λ = 0 for 1 ≤ i ≤ n

∂x2i

n

∂ 2F Y 4 1

= 4xi xj x2r = at x21 = x22 = . . . = x2n =

∂xi ∂xj r=1,r6=i,j

nn−1 n

8 X

⇒ d2 F = dxi dxj (only the cross terms appear)

nn−1 1≤i<j≤n

Now x21 + x22 + . . . + x2n = 1, so 2 ni=1 xi dxi = 0 ⇒ ni=1 dxi = 0 ⇒ dxn = −(dx1 + . . . +

P P

dxn−1 ).

n−1

2 8 h X X 2 i

Thus d F = n−1 dxi dxj − dxi . Clearly d2 F < 0, so we have a

n 1≤i<j≤n−1 i=1

maximum when x21 = x22 = . . . = x2n = n1 .

3

Derivation of inequality:

1 n x2 + x2 + . . . + x2 n

1 2 n

x21 x22 . . . x2n ≤ = as x21 + x22 + . . . + x2n = 1

n n

2 2 2

1 x + x2 + . . . + xn

⇒ (x21 x22 . . . x2n ) n ≤ 1

n

Let x2i = ai , 1 ≤ i ≤ n, to get

1 a1 + . . . + an

(a1 a2 . . . an ) n ≤

n

as required.

x 2 1 2 2

e−x (t +1)

Z Z

−t2

Question 4(a) Define f (x) = e dt , g(x) = 2

dt. Show that g 0 (x) +

0 0 tZ + 1 √

x

0 π −t2 π

f (x) = 0 for all x. Deduce that (i) g(x) + f (x) = 4 (ii) lim e dt = .

x→∞ 0 2

Solution.

Z x

0 −t2

2

f (x) = 2 e dt e−x

0

1 2 2

∂ e−x (t +1)

Z

0

g (x) = dt

0 ∂x t2 + 1

1 2 (t2 +1)

e−x (−2x(t2 + 1))

Z

= dt

0 t2 + 1

2 Rx 2

Let xt = y in the second equation, then dy = x dt, g 0 (x) = −2e−x 0

e−y dy = −f 0 (x),

showing that f 0 (x) + g 0 (x) = 0 for all x.

R1

1. f 0 (x) + g 0 (x) = 0 ⇒ f (x) + g(x) = C, a constant. Clearly f (0) = 0, g(0) = dt

0 t2 +1

=

i1

tan−1 t = π4 . So C = f (0) + g(0) = π4 , thus g(x) + f (x) = π4 .

0

2 2

1

e−x t

Z

2

2. lim g(x) = lim 2

dt e−x .

x→∞ x→∞ 0 t +1

Z 1 −x2 t2 Z 1

e dt π 2 2 2

But 2

dt ≤ 2

= , ∵ e−x t ≤ 1. Thus 0 ≤ g(x) ≤ π4 e−x ⇒

0 t +1 0 1+t 4

limx→∞ g(x) = 0.

Z x 2 Z x √

−t2 π −t2 π

Thus lim f (x) = lim e dt = =⇒ lim e dt = , as both sides are

x→∞ x→∞ 0 4 x→∞ 0 2

positive.

4

UPSC Civil Services Main 1980 - Mathematics

Calculus

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) Let f be a real continuous function on the interval [a, b] which is differen-

tiable in the open interval (a, b). Prove that there exists at least one point x ∈ (a, b) at which

f (b) − f (a) = (b − a)f 0 (x).

Solution. (This is the mean value theorem of calculus.)

Consider the function φ(x) = f (x)−f (a)−(x−a)A, where A is a constant so determined

f (b) − f (a)

that φ(b) = 0 ⇒ A = . Thus φ(b) = φ(a) = 0, φ(x) is continuous in [a, b] as

b−a

it is the sum of functions continuous over [a, b], and φ(x) is differentiable over (a, b) as

f (x), f (a), (x − a)A are all differentiable over (a, b). Thus φ(x) satisfies the conditions of

Rolle’s theorem, so there is at least one point c ∈ (a, b) such that φ0 (c) = 0 i.e. f 0 (c) − A =

f (b) − f (a)

0 ⇒ f 0 (c) = , thus f (b) − f (a) = (b − a)f 0 (c).

b−a

Question 1(b) Find the volume of the solid of revolution formed by rotating the area cut

off from y 2 = 4ax by the line y = x through 2π radians about the x-axis.

Solution. The curve y 2 = 4ax intersects y = x at (0, 0) and (4a, 4a). The required volume

V = Volume of rotating y 2 = 4ax about the x-axis - Volume of rotating y = x about the

x-axis. Thus

Z 4a Z 4a

2

V = πy1 dx − πy22 dx

Z0 4a 0

Z 4a

= π4ax dx − πx2 dx

0 0

2 4a 3 4a

x x

= 4aπ −π

2 0 3 0

2

16a 64a3 32πa3

= 4aπ −π =

2 3 3

1

Question 2(a) A rectangular box open at the top is to have a capacity of 4m3 . Find the

dimensions of the box requiring least material for its construction.

Solution. Let x be the length, y be the breadth, and z the height of the box. Then the

capacity of the box is xyz = 4. The surface area of the box is S = 2xz + 2yz + xy =

4

xy + 2(x + y) xy = xy + x8 + y8 . For extreme values

∂S 8 ∂S 8

= y − 2 = 0, =x− 2 =0

∂x x ∂y y

x4

Thus y = x82 , substituting in the second equation we get x = 8

⇒ x = 0, 2. Since x = 0 is

inadmissible, x = 2 ⇒ y = 2, z = 1.

∂ 2S 16 ∂ 2 S 16 ∂ 2 S

= , = , =1

∂x2 x3 ∂y 2 y 3 ∂x ∂y

2 2

∂ 2S ∂ 2S ∂ S ∂ 2S ∂ 2S

Thus − = 4 − 1 > 0 when x = 2, y = 2, and = 2 > 0, = 2 > 0.

∂x2 ∂y 2 ∂x ∂y ∂x2 ∂y 2

So x = 2, y = 2 is a minimum for S.

Thus the dimensions of the required box are 2 meters × 2 meters × 1 meter.

2

UPSC Civil Services Main 1981 - Mathematics

Calculus

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

|x|

Question 1(a) Let f (x) = . Discuss the value of limx→0 f (x) and f (0). Justify your

x

answer. Discuss the differentiability of f and f 2 at x = 0.

Solution. Clearly (

1, x>0

f (x) =

−1, x < 0

Thus limx→0+ f (x) = 1, limx→0− f (x) = −1, showing that limx→0 f (x) does not exist. Thus

no matter what value is given to f (x) at 0, f (x) cannot be continuous at x = 0.

Since f (x) is not continuous at x = 0, the question of its differentiability does not arise., as

f (x0 + h) − f (x0 )

differentiability implies continuity — note that f (x0 + h) − f (x0 ) = h, so if

h

f (x0 + h) − f (x0 )

f (x) is differentiable at x0 , then limh→0 f (x0 +h)−f (x0 ) = limh→0 limh→0 h =

0

h

f (x0 ) · 0 = 0, so limh→0 f (x0 + h) = f (x0 ).

Clearly f 2 (x) = 1 for every x 6= 0, therefore f 2 would become differentiable at x = 0 if

we define f (x) = 1 or −1, making f 2 (x) = 1 for all x. In any other case f 2 (x) would not be

differentiable at x = 0, as it would not be continuous at x = 0.

x

Question 2(a) Suppose a manufacturer can sell x items at a price P = 200 − paisa

100

per item, and it costs y = 50x + 20000 paise to produce the x items. What is the production

level for maximum profits, and the selling price per item?

x

Solution. Selling price of x items = 200 − x

100

Cost price of x items = 50x + 20000.

1

x

Profict P on the sale of x items = 200 − x − (50x + 20000).

100

For extreme values,

dP x x x

= 200 − − − 50 = 150 − = 0 =⇒ x = 7500

dx 100 100 50

d2 P 1

2

=− <0

dx 50

showing that the profit is maximum when 7500 items are sold at 125 paisa per item.

Z ∞Z ∞

2 2

Question 2(b) Evaluate the double integral e−(x +y ) dx dy.

0 0

1

Question 2(c) State the mean value theorem. If f 0 (x) = for all x, and f (0) = 0,

1 + x2

show that 0.4 < f (2) < 2.0.

Solution. If f (x) is a real valued function defined on the closed interval [a, b], such that (i)

f (x) is continuous on the closed interval [a, b], (ii) f (x) is differentiable on the open interval

(a, b), then there exists a point c ∈ (a, b) such that f (b) − f (a) = (b − a)f 0 (c).

We apply the mean value theorem to f (x) on [0, 2] and get f (2) = f (2) − f (0) =

(2 − 0)f 0 (c) = 1+c

2

2 for some c ∈ (0, 2).

2 2 2

Now 0 < c < 2 so < 2

< , thus 0.4 < f (2) < 2.

1+4 1+c 1+0

√

Question 3(a) The region bounded by the curves y = x3 and y = x is revolved about the

x-axis. Find the volume generated.

Solution. The curves intersect each other at (0, 0 and (1, 1). The required volume V =

Volume obtained by rotating the arc from (0, 0) to (1, 1) of y 2 = x about the x-axis - Volume

obtained by rotating the arc from (0, 0) to (1, 1) of y = x3 about the x-axis. Thus

Z 1 Z 1

2

V = πy1 dx − πy22 dx

0 0

Z 1 Z 1

= π x dx − π x6 dx

0 0

π π 5π

= − =

2 7 14

Γ(p)Γ(q)

Question 3(b) Show that B(p, q) = .

Γ(p + q)

Solution. See 1991 question 2(c).

2

UPSC Civil Services Main 1982 - Mathematics

Calculus

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

(

0, x irrational

f (x) = 2

q2

, x rational, x = pq in lowest terms

Show that f (x) has a removable discontinuity at every rational point, but is continuous for

irrational values.

Solution.

of positive integers q for which q 2 ≤ 2 is finite, we can find a positive number δ > 0

such that (c − δ, c + δ) does not contain any rational number pq for which q 2 ≤ 2 , i.e.

x ∈ (c − δ, c + δ), x = pq , (p, q) = 1 ⇒ q 2 > 2 ⇒ q22 < . Thus for x ∈ (c − δ, c + δ),

• If x is rational, x = pq , (p, q) = 1, then |f (x) − f (c)| = | q22 − 0| < .

whatever δ > 0 we take, we can find an irrational number x ∈ (c − δ, c + δ), and

|f (x) − f (c)| = |0 − q22 | > . Thus f (x) is not continuous at any rational point.

limx→c f (x) exists, because if limx→c f (x) = l, we can define f (c) = l, and make f

continuous at c.

1

Let > 0, and let δ > 0 be chosen so that 0 < δ < . Let x ∈ R, 0 < |x − pq | < δ.

2q

Then

|f (x) − 0| = |0 − 0| < if x 6∈ Q

m p

If x ∈ Q, let x = , x 6= , then

n q

p m p qm − pn qm − pn

x − = − = < ⇒ <

q n q qn 2q n 2

2 qm − pn m p

Now |f (x) − 0| = | n22 − 0| ≤ < because |qm − pn| ≥ 1 as 6= .

n n n n q

Hence we have shown that 0 < |x − pq | < δ with δ < ⇒ |f (x) − 0| < , i.e.

2q

limx→ pq f (x) = 0, showing that f (x) has a removable discontinuity at all rational

points.

Question 1(b) If f (x) is differentiable three times in (a − h, a + h), prove that there is a

point c in the interval such that

f (a + h) + f (a − h) h2

− f 0 (a) = f 000 (c)

2h 6

Solution. In order to prove this we prove the following particular case.

If f (x) is thrice differentiable and f (a) = 0, f (a − h) = 0, f (a + h) = 0, f 0 (a) = 0 then

there exists a point c ∈ (a − h, a + h) such that f 000 (c) = 0.

1. f (a − h) = f (a) = 0 and f 0 (x) exists, therefore Rolle’s theorem tells us that there

exists ξ1 ∈ (a − h, a) such that f 0 (ξ1 ) = 0.

2. f (a + h) = f (a) = 0 and f 0 (x) exists, therefore Rolle’s theorem tells us that there

exists ξ2 ∈ (a, a + h) such that f 0 (ξ2 ) = 0.

3. f 0 (ξ1 ) = f 0 (a) = 0 and f 00 (x) exists in [ξ1 , a], Rolle’s theorem gives a point η1 ∈ (ξ1 , a)

such that f 00 (η1 ) = 0.

4. f 0 (a) = f 0 (ξ2 ) = 0 and f 00 (x) exists in [a, ξ2 ], Rolle’s theorem gives a point η2 ∈ (a, ξ2 )

such that f 00 (η2 ) = 0.

5. Finally f 00 (η1 ) = f 00 (η2 ) = 0 and f 000 (x) exists in [η1 , η2 ], then Rolle’s theorem gives a

point c ∈ (η1 , η2 ) such that f 000 (c) = 0. Clearly, c ∈ (η1 , η2 ) ⊆ (a − h, a + h).

2

Now to prove the main result, let

Q(x) = f (a) + (x − a)f 0 (a) + B(x − a)2 + A(x − a)3

where A, B are so determined that Q(a + h) = f (a + h), Q(a − h) = f (a − h). Thus

f (a + h) − Q(a + h) = 0 ⇒ f (a + h) − f (a) − hf 0 (a) − h2 B − h3 A = 0 (1)

f (a − h) − Q(a − h) = 0 ⇒ f (a − h) − f (a) + hf 0 (a) − h2 B + h3 A = 0 (2)

Subtracting (2) from (1) we get

f (a + h) − f (a − h) − 2hf 0 (a) − 2h3 A = 0 (3)

We now take φ(x) = f (x) − Q(x), then φ(a) = 0, φ0 (a) = 0, φ(a − h) = 0, φ(a + h) = 0,

so by the result proved above, we get a point c ∈ (a − h, a + h) such that φ000 (c) = 0. But

000

φ000 (c) = f 000 (c) − Q000 (c) = f 000 (c) − 6A ⇒ A = f 6(c) . Substituting in (3) we have

h3 000

f (a + h) − f (a − h) − 2hf 0 (a) = f (c)

3

or

f (a + h) + f (a − h) h2

− f 0 (a) = f 000 (c)

2h 6

as required.

Note: In such questions the general tendency is to take φ(t) = f (a + th) − f (a − th), 0 ≤

t ≤ 1 and apply Taylor’s theorem to φ(t) in [0, 1]. We get

1 00

φ(1) = φ(0) + φ0 (0) + φ (θ), 0 < θ < 1

2!

φ(1) = f (a + h) − f (a − h)

φ(0) = 0

φ0 (0) = h(f 0 (a) − (−h)f 0 (a) = 2hf 0 (a)

φ00 (0) = h2 f 00 (a) − h2 f 00 (a) = 0

φ000 (θ) = h3 f 000 (a + θh) − (−h3 )f 000 (a − θh) = h3 [f 000 (a + θh) + f 000 (a − θh)]

Thus

f (a + h) + f (a − h) h2

− f 0 (a) = [f 000 (a + θh) + f 000 (a − θh)]

2h 6

i.e. we get two points in (a − h, a + h) and not one, so such a solution does not work in our

case.

Question 1(c) The arc of the parabola y 2 = 4ax from the vertex to one extremity of the

latus rectum is√revolved about the corresponding chord. Prove that the volume of the spindle

2 5 3

so formed is πa .

75

Solution.

3

Arc OL joining (0, 0) to (a, 2a) is rotated

about the chord OL. Let P be the point L

(x, y), M the foot of the the perpendicular on (a, 2a)

|y − 2x| P

OL i.e. y = 2x. Therefore P M = √ .

5 M

y 2 − 4xy + 4x2 O

OM 2 = OP 2 − P M 2 = x2 + y 2 −

5

4y 2 + 4xy + x2

=

5 √

2y + x 4 ax + x y 2 = 4ax

OM = √ = √

5 5

The required volume is V , given by

Z a

V = π(P M )2 d(OM )

0

√

y 2 − 4xy + 4x2 4ax − 4ax · 4x + 4x2

(P M )2 = =

√

5 5

a √ √

2 √x + 1 2 a+ x

d(OM ) = √ dx = √ dx

5 5x

Z a √ √

4 √ 2 a + x

⇒V = π x a − 2 ax + x √ dx

0 5 5x

Z a

4π √ √ √ 3

= √ x2 + ax − 2x ax + 2a ax − 4ax + 2 ax 2 dx

5 5 0

a

4π x3 x2

3 2 3

= √ − 3a + 2a 2 x 2

5 5 3 2 3 0

√ 3 3 3

√

4 5π a 3a 4a 2 5 3

= − + = πa

25 3 2 3 75

as required.

∂ 2u ∂u ∂u

v2 +v = cos 2u

∂x ∂y ∂x ∂y

Solution. Differentiate x cos u + y sin u = 1 with respect to x to get

∂u ∂u

cos u − x sin u + y cos u =0

∂x ∂x

or

∂u ∂u ∂u

cos u − (x sin u − y cos u) = cos u − v =0 ⇒ v = cos u (1)

∂x ∂x ∂x

4

Differentiate x cos u + y sin u = 1 with respect to y to get

∂u ∂u

−x sin u + sin u + y cos u =0

∂y ∂y

or

∂u ∂u ∂u

sin u − (x sin u − y cos u) = sin u − v =0 ⇒ v = sin u (2)

∂y ∂y ∂y

Differentiate (1) with respect to y to get

∂u ∂ 2u ∂u ∂v

− sin u =v +

∂y ∂x ∂y ∂x ∂y

But

∂v ∂u ∂u ∂u

= x cos u − cos u + y sin u = − cos u

∂y ∂y ∂y ∂y

as x cos u + y sin u = 1. Therefore

∂u ∂ 2u ∂u ∂u ∂u

− sin u =v + − cos u

∂y ∂x ∂y ∂x ∂y ∂x

Multiplying by v, and substituting (1), (2), we get

∂ 2u ∂u ∂u

− sin u(sin u) = v 2 +v − cos u(cos u)

∂x ∂y ∂x ∂y

or

∂ 2u ∂u ∂u

v2 +v = cos2 u − sin2 u = cos 2u

∂x ∂y ∂x ∂y

as required.

√ 1 √ 1 2

1

π2 Γ 3= 3 Γ

6 3

π

Z 1 Z

Γ(m)Γ(n) m−1 n−1

2

= B(m, n) = x (1 − x) dx = 2 sin2m−1 θ cos2n−1 θ dθ

Γ(m + n) 0 0

on putting x = sin2 θ.

1

1. We take n = 2

to get

π

Γ( 12 )Γ(m)

Z

2

=2 sin2m−1 θ dθ

Γ(m + 21 ) 0

5

2. We take m = n to get

π π

π

[Γ(n)]2

Z Z Z

2

2n−1 1 2

2n−1 1

=2 (sin θ cos θ) dθ = (sin 2θ) dθ = (sin φ)2n−1 dφ

Γ(2n) 0 22n−2 0 22n−1 0

by putting 2θ = φ. But

π

Z π Z

2

2n−1

(sin φ) dφ = 2 (sin φ)2n−1 dφ

0 0

so π

[Γ(n)]2

Z

1 2

= 2n−2 (sin θ)2n−1 dθ

Γ(2n) 2 0

=

Γ(n + 12 ) Γ(2n)

or √

1 Γ( 12 )Γ(2n) π

Γ(n)Γ n + = 2n−1

= 2n−1 Γ(2n)

2 2 2

1

Put n = 3

to get

√ 1 2

2 2

1

√

1 π 23 Γ 3 1 1 Γ

3

Γ 3

Γ = =⇒ Γ = π 23

3 Γ( 56 ) 3 Γ( 56 )

π 2

1

2π 5

1

Now Γ(n)Γ(1 − n) = , so Γ 3

Γ 3

= √ and Γ 6

Γ 6

= 2π. Thus we get

sin nπ 3

2π

Γ( 61 )

2 √

√

1 1 3

Γ = π 23

3 2π

so 2

√ 1 √ 1 1

3 Γ = π 23 Γ

3 6

as required.

Question 2(c) Find a point within a triangle such that the sum of its distances from the

angular points may be a minimum.

Solution.

6

Let the point be P = (x, y), and the A

vertices of the triangle ABC be (xi , yi ), i =

1,

P2, 3.p Then u = P A + P B + P C =

3 2 2

r1

i=1 (x −

pxi ) + (y − yi ) = r1 + r2 + r3 , a c

where ri = (x − xi )2 + (y − yi )2 , i = 1, 2, 3. P γ

For extreme values α

∂u

3

x − xi β

X r2 r3

= =0

∂x ri

i=1 B b C

3

∂u X y − yi

= =0

∂y i=1

ri

Thus

x − x1 x − x2 x − x3 y − y1 y − y2 y − y3

+ =− , + =−

r1 r2 r3 r1 r2 r3

Squaring and adding, we get

2

2+ (x − x1 )(x − x2 ) + (y − y1 )(y − y2 ) = 1

r1 r2

or

1 1

(x − x1 )(x − x2 ) + (y − y1 )(y − y2 ) = −

r1 r2 2

Now by the cosine formula, r12 + r22 − 2r1 r2 cos α = a2 or

Thus

1

cos α = (2x2 − 2xx1 − 2xx2 + 2x1 x2 + 2y 2 − 2yy1 − 2yy2 + 2y1 y2 )

2r1 r2

1 1

= (x − x1 )(x − x2 ) + (y − y1 )(y − y2 ) = −

r1 r2 2

Thus α = 2π 3

. By symmetry, β = γ = 2π 3

, i.e. the sides of the triangle subtend the same

angle at P .

Note that this solution is inadmissible when any angle of the triangle is > 2π3

. In this case,

the partial derivatives of u do not vanish anywhere inside the triangle, hence the minimum

must lie at one of the vertices. A simple check shows that the minimum is at the vertex

whose angle is > 2π3

.

7

3 3

∂ 2u X ∂ 2 ri X ∂ x − xi

= =

∂x2 i=1

∂x2 i=1

∂x ri

3

X 1 x − xi x − xi

= −

i=1

r i ri2 ri

3 3

(x − xi )2 (y − yi )2

X

X 1

= 1− 2

= 3

>0

i=1

ri ri i=1

r i

∂2u

Similarly ∂y 2

> 0.

3 3 3 3

∂ 2u X ∂ x − xi X x − xi ∂ri X x − xi y − y i X (x − xi )(y − yi )

= =− 2

=− 2

=−

∂x ∂y i=1

∂y ri i=1

ri ∂y i=1

ri ri i=1

ri3

2 2

∂ 2u ∂ 2u ∂ u

Now we need to check that 2 2

− > 0. To make the calculations easier,

∂x ∂y ∂x ∂y

let x − xi = ri cos αi , y − yi = ri sin αi . Then

2 3 3 3 2

∂ 2u ∂ 2u ∂ 2u sin2 αi cos2 αi

X X X

sin αi cos αi

− = −

∂x2 ∂y 2 ∂x ∂y i=1

ri i=1

ri i=1

ri

1

= [sin2 α1 cos2 α2 + cos2 α1 sin2 α2 − 2 sin α1 cos α2 cos α1 sin α2 ] + . . .

r1 r2

1 1 1

= [sin α1 cos α2 − cos α1 sin α2 ]2 + [. . .]2 + [. . .]2 > 0

r1 r2 r1 r3 r2 r3

Thus u = P A + P B + P C is at a minimum when P is the point at which the sides subtend

the same angle.

Note: This is called the Steiner’s problem, and the point P is called the Fermat Point or

the Fermat-Torricelli Point of the triangle.

Paper II

Question 3(a) Find the extreme values of f (x1 , x2 , x3 ) = x21 +x22 +x23 subject to 3i,j=1 aij xi xj =

P

1, where aij = aji .

Solution. Let F (x1 , x2 , x3 ) = x21 + x22 + x23 − λ 3i,j=1 aij xi xj − 1 . For extreme values,

P

3

∂F X

= 2xi − 2λ aij xj = 0, 1 ≤ i ≤ 3 (∗)

∂xi j=1

8

3 3

X ∂F X

This implies 0 = xi = 2f − 2λ aij xi xj = 2f − 2λ ⇒ f = λ at the extreme values.

i=1

∂xi i,j=1

Now the equations (*) are a set of linear equations, which can be written as 2(I−λA)x =

0, where I is a 3 × 3 identity matrix,

P3 and A = (aij ) is a 3 × 3 symmetric matrix, so has

real eigenvalues. Since x 6= 0 as i,j=1 aij xi xj = 1, I − λA must be singular ⇒ λ 6= 0, so

A − λ−1 I is singular. Thus λ−1 is an eigenvalue of A.

Hence the extreme values of f are λ−1 , where λ is an eigenvalue of A = (aij ).

9

UPSC Civil Services Main 1983 - Mathematics

Calculus

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

f (x) = lim

n→∞ 1 + x2n

in the interval 0 ≤ x ≤ π2 . Show that the function does not vanish anywhere in this interval

and explain why is so although f (0), f ( π2 ) differ in sign.

log 3 − sin 1

f (x) = log(2 + x), 0 ≤ x < 1, f (1) =

2

1 log(2+x)

For x > 1, x2n

→ 0 and x2n

→ 0 as n → ∞, therefore

log(2+x)

x2n

− sin x

f (x) = lim 1 = − sin x

n→∞

x2n

+1

Thus

log(2 + x), 0 ≤ x < 1

f (x) = log 3−sin

2

1)

, x=1

− sin x, 1 < x ≤ π2

Clearly f (0) = log 2 > 0, f ( π2 ) = − sin π2 = −1 < 0, and f (x) 6= 0 for any x ∈ [0, π2 ].

The reason why this can happen is that f (x) is not continuous in the interval [0, π2 ], so

the intermediate value theorem does not apply. The function is discontinuous at x = 1 as

limx→1+ f (x) = − sin 1, limx→1− f (x) = log 3 and both of these are different from f (1).

1

Question 1(b) Prove that there is a point a ∈ (−h, h) such that

Z h

2h3 00

f (x) dx = h[f (h) + f (−h)] − f (a)

−h 3

Solution. Integrating by parts, we get

Z h ih Z h Z h

0

f (x) dx = xf (x) − xf (x) dx = h[f (h) + f (−h)] − xf 0 (x) dx

−h −h −h −h

R0

Now put x = −t in −h

xf 0 (x) dx to get

Z 0 Z 0 Z h

0 0

xf (x) dx = (−t)f (−t)(−dt) = − tf 0 (−t) dt

−h h 0

Thus Z h Z h

x f 0 (x) − f 0 (−x) dx

f (x) dx = h[f (h) + f (−h)] −

−h 0

0 0

f (x) − f (−x) f (x) + f 00 (−x)

00

f 0 (x) − f 0 (−x)

We now note that lim = lim = 2f 00 (0), showing that

x→0 x x→0 1 x

is Riemann integrable in [0, h]. Since x2 keeps the same sign, the first mean value theorem

of integral calculus gives us

Z h 0 0

2 f (x) − f (−x) f 0 (ξ) − f 0 (−ξ) h 2

Z

x dx = x dx

0 x ξ 0

Z h

f 0 (ξ) − f 0 (−ξ) 2h3

f (x) dx = h[f (h) + f (−h)] −

−h 2ξ 3

We now apply Lagrange’s mean value theorem to the function f 0 (x) in [−ξ, ξ] and get

f 0 (ξ) − f 0 (−ξ)

= f 00 (a) for some a ∈ (−ξ, ξ)

2ξ

Thus

h

2h3 00

Z

f (x) dx = h[f (h) + f (−h)] − f (a)

−h 3

Note that 0 < ξ < h ⇒ a ∈ [−h, h].

Corollary: If f 000 (x) exists in (−h, h) prove that

f (h) − f (−h) f 0 (h) + f 0 (−h) h3 000

=h − f (a)

2 2 3

for some a ∈ (−h, h).

Proof: Use the above result for f 0 (x) instead of f (x). The left hand side now becomes

f (h) − f (−h), and dividing by two, we get the above equation.

2

Question 1(c) Find the least perimeter of an isoceles triangle in which a circle of radius r

can be inscribed.

Solution.

midpoint of BC. E is the point of contact α

of AC and the circle or radius r inscribed

OE

in ABC. Therefore = tan α, or AE =

AE F E

OE cot α = r cot α. Similarly, we find out r r

that OA = r csc α, and AD = AO + OD = O

r(1 + csc α), consequently BD = AD tan α =

r(1 + csc α) tan α. Since 4ABC is isoceles,

BD = DC, DC = CE and AE = AF . B D C

The perimeter

P = BD + DC + CE + AE + BF + AF

= 4BD + 2AE = 4r(1 + csc α) tan α + 2r cot α

= 4r tan α + 4r sec α + 2r cot α

dP

= 4r sec2 α + 4r sec α tan α − 2r csc2 α = 0

dα

2 2 sin α 1

⇒0 = + −

cos2 α cos2 α sin2 α

⇒ 0 = 2 sin α + 2 sin α − cos2 α

2 3

⇒ 0 = 2 sin3 α + 3 sin2 α − 1

⇒ 0 = (sin α + 1)2 (2 sin α − 1)

Thus dP

dα

= 0 ⇒ sin α = 21 , since sin α = −1 is not possible as 2α is the angle of a triangle.

Thus α = π6 .

d2 P

= 8r sec2 α tan α + 4r(sec α tan2 α + sec3 α) − 4r csc α(− csc α cot α) > 0

dα2

for α = π6 . Therefore we have a minimum when α = π6 . The required perimeter is 4r √13 +

√ √

4r √23 + 2r 3 = 6r 3.

1

2 , prove without using the method of integration

that x+y

f (x) + f (y) = f

1 + xy

3

x+y

Solution. Let u = f (x) + f (y), v = 1+xy

. Then

∂v (1 − xy)(1) − (x + y)(−y) 1 + y2

= =

∂x (1 − xy)2 (1 − xy)2

∂v (1 − xy)(1) − (x + y)(−x) 1 + x2

= =

∂y (1 − xy)2 (1 − xy)2

∂(u, v) 12 1

1+x 1+y 2

= 1+y2 1+x = 0

2

∂(x, y) (1−xy)2 (1−xy) 2

x+y

Therefore u, v are functionally dependent. Let u = φ(v) ⇒ f (x) + f (y) = φ .

1 − xy

Substituting y = x0+gives us f (x) + f (0) = φ(x) or φ(x) = f (x) as f (0) = 0. Hence

y

f (x) + f (y) = f , which was to be proved.

1 − xy

Question 2(b) Show that under the transformation u = x2 − y 2 , v = 2xy the equation

∂ 2H 2

2∂ H ∂H ∂H

y2 2

− x 2

=x −y

∂x ∂y ∂x ∂y

∂ ∂ ∂H

becomes u −v = 0.

∂v ∂u ∂v

Solution.

∂H ∂H ∂u ∂H ∂v ∂H ∂H

= + = 2x + 2y

∂x ∂u ∂x ∂v 2∂x ∂u ∂v 2

∂ 2H 2

∂ 2H

∂H ∂ H ∂ H ∂ H

= 2 + 2x 2x + 2y + 2y 2x + (2y)

∂x2 ∂u ∂u2 ∂u ∂v ∂v ∂u ∂v 2

∂H ∂ 2H ∂ 2H ∂ 2H

= 2 + 4x2 2 + 8xy + 4y 2 2 (1)

∂u ∂u ∂u ∂v ∂v

∂H ∂H ∂u ∂H ∂v ∂H ∂H

= + = −2y + 2x

∂y ∂u ∂y ∂v ∂y ∂u ∂v

2

2 2

2

∂ 2H

∂ H ∂H ∂ H ∂ H ∂ H

= −2 − 2y (−2y) + 2x + 2x (−2y) + (2x)

∂y 2 ∂u ∂u2 ∂u ∂v ∂v ∂u ∂v 2

∂H ∂ 2H ∂ 2H ∂ 2H

= −2 + 4y 2 2 − 8xy + 4x2 2 (2)

∂u ∂u ∂u ∂v ∂v

∂H ∂H ∂H ∂H ∂H ∂H

x −y = 2x2 + 2xy + 2y 2 − 2xy

∂x ∂y ∂u ∂v ∂u ∂v

∂H

= 2(x2 + y 2 ) (3)

∂u

4

Using (1), (2), (3) we get

2 2

2∂ H 2∂ H ∂H ∂H

0 = y −x − x −y

∂x2 ∂y 2 ∂x ∂y

2 2 2

2 ∂H 2 2∂ H 3 ∂ H 4∂ H

= 2y + 4x y + 8xy + 4y

∂u ∂u2 ∂u ∂v ∂v 2

2 2

∂H ∂ H ∂ H ∂ 2H ∂H

+2x2 − 4x2 y 2 2 + 8x3 y − 4x4 2 − 2(x2 + y 2 )

∂u ∂u ∂u ∂v ∂v ∂u

2 2

∂ H ∂ H

= 8xy(x2 + y 2 ) + 4(y 4 − x4 ) 2

∂u ∂v ∂v

2 2

∂ H ∂ H

⇒0 = (y 2 − x2 ) 2 + 2xy

∂v ∂u ∂v

∂ 2H ∂ 2H

⇒0 = −u 2 + v

∂v ∂u

∂v

∂ ∂ ∂H

⇒0 = u −v

∂v ∂u ∂v

as required.

Question 2(c) Prove that the volume of the solid generated by the revolution of the tractrix

t

x = a cos t + a log tan , y = a sin t

2

about the asymptote is equal to half the volume of a sphere of radius a.

Solution. The x-axis is the asymptote, as t → 0 ⇒ x → ∞, y → 0. The volume required

is twice the volume in the first quadrant.

Z π

2 dx

V = 2 πy 2 dt

0 dt

Z π

2 a

= 2π a2 sin2 t(−a sin t + ) dt

0 sin t

Z π

2

3

= 2πa (sin t − sin3 t) dt

0

Z π

2

3

= 2πa sin t cos2 t dt

0

π

cos3 t 2 2πa3

3

= 2πa − =

3 0 3

3

1 4πa 1

= = × Volume of a sphere of radius a

2 3 2

5

Paper II

Question 3(a) Obtain a set of sufficient conditions such that for a function f (x, y)

∂ 2f ∂ 2f

=

∂x ∂y ∂y ∂x

Solution. Euler’s Theorem: If the partial derivatives fxy and fyx are continuous at (a, b)

then fxy (a, b) = fyx (a, b).

Proof: Let Ψ(h, k) = f (a + h, b + k) − f (a + h, b) − f (a, b + k) + f (a, b) where (a + h, b +

k), (a + h, b), (a, b + k) all belong to a neighborhood N of (a, b) — we can take for N an open

disc with center (a, b) in which fxy and fyx exist.

Let G(x) = f (x, b + k) − f (x, b) for x ∈ Ih where Ih = [a, a + h] or [a + h, a] according

as h > 0 or h < 0. Clearly G0 (x) = fx (x, b + k) − fx (x, b) for x ∈ Ih . We apply Lagrange’s

Mean Value Theorem to G(x) and obtain:

G(a+h)−G(a) = Ψ(h, k) = hG0 (a+θh) = h[fx (a+θh, b+k)−fx (a+θh, b)] (∗)

Now we consider F (t) = fx (a + θh, t) for t ∈ Ik where Ik = [b, b + k] or [b + k, b] according

as k > 0 or k < 0. We apply Lagrange’s Mean Value Theorem to F (t) and obtain:

f (a

∂y x

+ θh, t), so

∂ 2f

F (b + k) − F (b) = k (a + θh, b + θ1 k)

∂y ∂x

∂ 2f

Ψ(h, k) = hk (a + θh, b + θ1 k)

∂y ∂x

Since fyx is continuous at (a, b), we get

Ψ(h, k) ∂ 2f ∂ 2f

lim = lim (a + θh, b + θ1 k) = (a, b)

h→0,k→0 hk h→0,k→0 ∂y ∂x ∂y ∂x

Now instead of G(x), we start with H(y) = f (a+h, y)−f (a, y) for y ∈ Ik , and proceeding

exactly as above, we get

Ψ(h, k) ∂ 2f

lim = (a, b)

h→0,k→0 hk ∂x ∂y

∂ 2f ∂ 2f

Hence (a, b) = (a, b), which completes the proof.

∂x ∂y ∂y ∂x

6

Question 3(b) Find the maximum and minimum values of x2 + y 2 + z 2 subject to the

conditions x + y + z = 1, xyz + 1 = 0.

Lagrange’s undetermined multipliers. The extreme values are given by

∂F

= 2x + λ1 + λ2 yz = 0

∂x

∂F

= 2y + λ1 + λ2 xz = 0

∂y

∂F

= 2z + λ1 + λ2 xy = 0

∂z

Subtracting the first two, 2(x − y) + λ2 z(y − x) = 0 ⇒ x = y or λ2 = z2 (Note that

x 6= 0, y 6= 0, z 6= 0 because xyz + 1 = 0). Similarly from the other pairs of equations, we

get y = z or λ2 = x2 , and x = z or λ2 = y2 .

Since xyz = −1, x + y + z = 1, it follows that x, y, z cannot be all positive or all negative,

moreover two must be positive and one negative. In particular, x = y = z is not possible.

Suppose x 6= y, then λ2 = z2 . Now ∂F ∂x

= 0 ⇒ 2x + λ1 + 2y = 0.

Substituting the values of λ1 , λ2 in ∂F

∂z

= 0, we get

2xy

2z − 2(x + y) + =0

z

But x + y = 1 − z, xy = − z1 , so we get 2z − 2(1 − z) − z22 = 0 ⇒ 2z 3 − z 2 − 1 = 0 ⇒

(z − 1)(2z 2 + z + 1) = 0. But 2z 2 + z + 1 has no real roots, so the only real root is z = 1.

Thus we get λ2 = 2, z = 1, λ1 = 0(∵ x + y + z = 1 ⇒ x + y = 0 ⇒ λ1 = 0), x = ±1, y =

∓1(∵ x + y = 0, xy = 1 ⇒ x2 = 1 ⇒ x = ±1, y = ∓1). Hence the stationary values are

(1, −1, 1), (−1, 1, 1) and f (x, y, z) = 3 at these points.

On taking y 6= z we shall get (1, 1, −1), (1, −1, 1) as stationary points (by symmetry),

and for x 6= z, we get (1, 1, −1), (−1, 1, 1). f (x, y, z) is always equal to 3, hence we cannot

say whether it is a maximum or minimum without checking d2 F .

Considering the point (−1, 1, 1),

0 ⇒ dx − dy − dz = 0 ⇒ dx = 0, dz = −dy.

Thus d2 F = 8(dz)2 > 0, so f has a minimum at all these stationary points.

Note: The question can be treated as that of one variable as y, z can be eliminated, but

the calculation becomes quite messy.

7

UPSC Civil Services Main 1984 - Mathematics

Calculus

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

(

x2 sin2 x1 , x 6= 0

f (x) =

0, x=0

is continuous at 0.

π

2. tan x is not continuous at x = 2

Solution.

√

1. Given > 0, let δ =√ , then |x| ≤ δ =⇒ |x2 sin2 x1 | ≤ |x2 | < , because | sin2 x1 | ≤ 1.

Thus |x − 0| < δ = ⇒ |f (x) − f (0)| < , showing that f (x) is continuous at x = 0.

1

2. Since limx→ π2 sin x = 1, given = 2

there exists δ1 > 0 such that 0 < |x − π2 | < δ1 ⇒

| sin x − 1| < 21 ⇒ sin x > 12 .

Since limx→ π2 cos x = 0, given any real number G > 0 there exists δ2 > 0 such that

0 < |x − π2 | < δ2 ⇒ | cos x| < 2G

1

.

Let δ = min(δ1 , δ2 ). Then δ > 0 and 0 < |x − π2 | < δ ⇒ | tan x| > 12 · 2G = G. This

shows that tan x is not bounded in any neighborhood of π2 , therefore limx→ π2 tan x does

not exist, so tan x is not continuous at x = π2 .

Note that if limx→a f (x) = l, then f (x) is bounded in a neighborhood of a, because given

> 0, there exists δ > 0 such that 0 < |x − a| < δ ⇒ l − < f (x) < l + .

1

Question 1(b) Find the volume of the torus generated by revolving a disc of radius r about

a line at a distance a > r from the center of the circle.

Solution. Let the line be the x-axis, and let the circle have center (0, a).√ The circle’s

equation is x2 + (y − a)2 = r2√

. The upper semicircle is given by f1 (x) = a + r2 − x2 , and

the lower one by f2 (x) = a − r2 − x2 , and the desired volume is given by

Z r

V = π (f12 (x) − f22 (x)) dx

Z−rr

= π (f1 (x) − f2 (x))(f1 (x) + f2 (x)) dx

−r

Z r √

= π (2 r2 − x2 )(2a) dx

−r

Z r√

= 8aπ r2 − x2 dx

0

Put x = r sin θ

Z π

2

= 8aπ r cos θ · r cos θ dθ

0

Z π

2

2

= 8ar π cos2 θ dθ

0

1 π

= 8ar2 π = 2π 2 ar2

22

We could get the same result by applying Pappus’ Theorem — the volume of a solid of

revolution generated by rotating a plane figure about an external axis is equal to the product

of the area of the figure and the distance traveled by its geometric centroid during revolution.

Thus V = πr2 · 2aπ = 2π 2 ar2 .

f (0, y) = 0. Examine whether (i) f (x, y) is continuous, and (ii) limx→0 f (x, y) for y 6= 0 and

limy→0 f (x, y) for x 6= 0 exist.

Solution. (i) Given > 0, let δ1 = δ2 = 2 . Then |x| < δ1 , |y| < δ2 ⇒ |f (x, y) − f (0, 0)| =

|(x + y) sin( x1 + y1 )| ≤ |x| + |y| < , as | sin( x1 + y1 )| ≤ 1. Thus f (x, y) is continuous at (0, 0).

(ii) Since |x sin( x1 + y1 )| ≤ |x|, it follows that limx→0 x sin( x1 + y1 ) = 0. Thus if limx→0 f (x, y)

exists for y 6= 0, then limx→0 y sin( x1 + y1 ) should also exist, for y 6= 0, because y sin( x1 + y1 ) =

(x + y) sin( x1 + y1 ) − x sin( x1 + y1 ).

Let g(x) = y sin( x1 + y1 ). Suppose limx→0 g(x) = l for y = π2 . Then given > 0, < π2 ,

there exists δ > 0 such that 0 < |x| < δ ⇒ |g(x) − l| < 2 . Let x1 = 2nπ 1 1

, x2 = (2n+1)π , n large

2

so that |x1 | < δ, |x2 | < δ. Now

≤ |g(x1 ) − l| + |g(x2 ) − l| <

2 π 2

g(x1 ) = sin(2nπ + ) =

π 2 π

2 π 2

g(x2 ) = sin((2n + 1)π + ) = −

π 2 π

4

⇒ |g(x1 ) − g(x2 )| = >

π

Thus we have a contradiction. Hence limx→0 y sin( x1 + y1 ) does not exist for y 6= 0, so

limx→0 f (x, y) for y 6= 0 does not exist. Similarly it can be seen that limy→0 f (x, y) for x 6= 0

does not exist, by symmetry.

ZZ

Question 2(a) Evaluate xy dx dy over the area given by the boundary y = 0(0 ≤ x ≤

3); y = (x − 3)2 (2 ≤ x ≤ 3); y = 1(1 ≤ x ≤ 2); y = x(0 ≤ x ≤ 1).

Solution.

y = (x − 3)2

parts

y=x

1. The triangle 4 bounded by

y = 0, x = 1, y = x.

2. The square bounded by

x = 1, x = 2, y = 0, y = 1. 4 R

3. The region R bounded by O x=1

x = 2, y = 0, y = (x − 3)2 , (2 ≤ x ≤ 3).

1 x 1

y 2 ix 1 1 3

ZZ Z Z Z Z

1

I1 = xy dx dy = xy dy dx = x dx = x dx =

4 x=0 y=0 0 2 0 2 0 8

Z 2 Z 1

x2 i 2 y 2 i 1 3 1

ZZ

3

I2 = xy dx dy = xy dy dx = · = · =

x=1 y=0 2 1 2 0 2 2 4

Z 3 Z (x−3)2 Z 3

y 2 i(x−3)2

ZZ

I3 = xy dx dy = yx dy dy = x dx

R x=2 0 x=2 2 0

1 3 1 0 4 1 h u6 u5 i0

Z Z

4 1 3 1 13

= x(x − 3) dx = u (u + 3) du = +3 = − =

2 2 2 −1 2 6 5 −1 2 5 6 60

1 3 13 131

Thus the given integral = I1 + I2 + I3 = + + =

8 4 60 120

3

Question 2(b) If B(p, q) is the Beta function, show that

pB(p, q) = (q − 1)B(p + 1, q − 1)

where p, q are real, p > 0, q > 1. Hence or otherwise find B(p, n) where n is an integer > 0.

Z 1

Solution. By definition, B(p, q) = xp−1 (1 − x)q−1 dx. Integrating by parts, we get

0

p

1 1

xp

Z

x

B(p, q) = (1 − x)q−1 + (q − 1)(1 − x)q−2 dx

p 0 0 p

Since p > 0, (q − 1) > 0, we get

Z 1

pB(p, q) = (q − 1) xp+1−1 (1 − x)(q−1)−1 dx = (q − 1)B(p + 1, q − 1)

0

n−1

B(p, n) = B(p + 1, n − 1)

p

(n − 1)(n − 2)

= B(p + 2, n − 2)

p(p + 1)

(n − 1)!

= B(p + n − 1, 1)

p(p + 1) . . . (p + n − 1)

Z 1

(n − 1)!

= xp+n−2 (1 − x)1−1 dx

p(p + 1) . . . (p + n − 2) 0

(n − 1)! Γ(p)Γ(n)

= =

p(p + 1) . . . (p + n − 1) Γ(p + n)

x+y ∂(u, v)

Question 2(c) If u = and v = tan−1 x + tan−1 y, find . Are u and v func-

1 − xy ∂(x, y)

tionally related? If so, find the relationship.

Solution.

∂u (1 − xy) · 1 − (x + y)(−y) 1 + y2

= =

∂x (1 − xy)2 (1 − xy)2

∂u (1 − xy) · 1 − (x + y)(−x) 1 + x2

= =

∂y (1 − xy)2 (1 − xy)2

∂v 1

=

∂x 1 + x2

∂v 1

=

∂y 1 + y2

∂u ∂u

∂(u, v) ∂x ∂y ∂u ∂v ∂u ∂v

= ∂x ∂y − ∂y ∂x = 0

∂v ∂v =

∂(x, y)

∂x ∂y

4

This shows that u, v are functionally related. Let x = tan θ, y = tan φ. Then v = θ + φ.

tan θ + tan φ

u= = tan(θ + φ) = tan v.

1 − tan θ tan φ

Alternatively, usingv = tan−1 x + tan−1 y we write

tan v − tan(tan−1 y) tan v − y

x = tan(v − tan−1 y) = −1

=

1 + tan v tan(tan y) 1 + y tan v

tan v−y

1+y tan v

+y (tan v)(1 + y 2 )

u = tan v−y = = tan v

1 − y 1+y tan v

1 + y2

as before.

Paper II

Question 3(a) Show that the maximum and minimum values of the function u = x2 +y 2 +xy

where ax2 + by 2 = ab, a > b > 0 are given by 4(u − a)(u − b) = ab.

Solution. Let F (x, y) = x2 +y 2 +xy +λ(ax2 +by 2 −ab) where λ is Lagrange’s undetermined

multiplier. The extreme values are obtained from

∂F ∂F

= 2x + y + 2λax = 0, = 2y + x + 2λby = 0

∂x ∂y

∂F ∂F

0 = x +y = 2(x2 + y 2 + xy) + 2λ(ax2 + by 2 ) = 2u + 2λab

∂x ∂y

u

Thus λ = − ab . Consequently 2x + y − 2ux b

= 0 ⇒ 2x(b − u) + yb = 0, and 2y + x − 2uy a

=

0 ⇒ 2y(a − u) + ax = 0.

Since ax2 + by 2 = ab > 0, (x, y) 6= (0, 0), so the coefficient matrix of the above linear

2(b − u) b

equations must be singular i.e. = 0 or 4(a − u)(b − u) − ab = 0.

a 2(a − u)

Thus the maximum and minimum √ values are √ given by 4(a − u)(b − u) − ab = 0.

Note: We can substitute x = b cos t, y = a sin t to make u a function of one variable,

and proceed accordingly.

(

x2 tan−1 xy − y 2 tan−1 xy , x 6= 0, y 6= 0

f (x, y) =

0, x=y=0

Solution.

1. The function is continuous at (0, 0), because |f (x, y) − f (0, 0)| = 0 or ≤ π2 (x2 + y 2 ).

5

2. fxy (0, 0) 6= fyx (0, 0). For this we calculate the following:

−1 y 2 1 y 1 1

fx (x, y) = 2x tan +x y 2 − 2 − y2 2

x 1+ 2 x

x 1 + xy2 y

y x2 y y3 y

= 2x tan−1 − 2 2

− 2 2

= 2x tan−1 − y

x x +y x +y x

(b)

f (h, k) − f (0, k) h2 tan−1 k

h

− k 2 tan−1 h

k

fx (0, k) = lim = lim

h→0 h h→0 h

h

tan−1

Now limh→0 h tan−1 k

h

= 0, limh→0 h

k

= 1

k

— if θ = tan−1 hk , hk = tan θ, then

h

tan−1 θ

h

= k tan

k

θ

, and tanθ θ → 1 as θ → 0.

Thus fx (0, k) = −k, in particular fx (0, 0) = 0.

(c) By symmetry, fy (x, y) = x − 2y tan−1 x

y

is x 6= 0, y 6= 0, and fy (h, 0) = h.

fx (0, k) − fx (0, 0) −k − 0

fyx (0, 0) = lim = lim = −1

k→0 k k→0 k

fy (h, 0) − fy (0, 0) h−0

fxy (0, 0) = lim = lim =1

h→0 h h→0 h

3. The function is differentiable at (0, 0) as both fx (x, y) and fy (x, y) are continuous at

(0, 0). Note that limx→0,y→0 fx (x, y) = limx→0,y→0 fy (x, y) = 0 and fx (0, 0) = 0 =

fy (0, 0).

6

UPSC Civil Services Main 1985 - Mathematics

Calculus

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) If (

xy

x2 +y 2

, (x, y) 6= (0, 0)

f (x, y) =

0, (x, y) = (0, 0)

Show that both the partial derivatives fx , fy exist at (0, 0), but the function is not continuous

there.

Question 1(b) If for all values of the parameter λ and some constant n, F (λx, λy) =

∂F ∂F

λn F (x, y) identically, where F is assumed to be differentiable, prove that x +y =

∂x ∂y

nF (x, y).

Question 1(c) Prove the relation between the beta and gamma functions

Γ(m)Γ(n)

β(m, n) =

Γ(m + n)

1

Question 2(a) If a function f defined on [a, b] is continuous on [a, b] and differentiable on

(a, b), and f (a) = f (b), then prove that there exists at least one real number c, a < c < b

such that f 0 (c) = 0.

and attains its maximum and minimum. Since f (a) = f (b) and we assume that f is not a

constant function, it follows that there exists a real number c such that f (c) = M , maximum

of f (without loss of generality) and a < c < b. Then the right hand derivative of f at c is

f (c + h) − f (c)

lim+ and clearly it is ≤ 0, because f (c + h) − f (c) ≤ 0 and h > 0. However

h→0 h

f (c + h) − f (c)

the left hand derivative of f at c is lim− ≥ 0 because f (c + h) − f (c) ≤ 0 and

h→0 h

0

h < 0. Since f (c) exists, the right hand derivative of f at c equals the left hand derivative

of f at c, which is possible only when both are zero, i.e. f 0 (c) = 0.

x2 x3 x4

log(1 + x) = x − + − + ...

2 3 4

Hence find the value of log(1 + x + x2 + x3 + x4 ).

Solution. Since log(1 + x) possesses continuous derivatives of all orders for every value of

x for which 1 + x > 0 i.e. x > −1, we have

x2 00 xn−1 (n−1)

f (x) = log(1 + x) = f (0) + xf 0 (0) + f (0) + . . . + f (0) + Rn

2! (n − 1)!

where

xn (n) xn (n − 1)!

Rn = f (θx) = (−1)n−1 , 0<θ<1

n! n! (1 + θx)n

(−1)n−1 (n − 1)!

as f (n) (x) for f (x) = log(1 + x) is .

(x + 1)n

x 1

Let 0 < x ≤ 1, so that 0 < < 1, then |Rn | < n

and therefore Rn → 0 as n → ∞.

1 + θx

Let −1 < x ≤ 0, we consider the expansion

x2 00 xn−1 (n−1)

f (x) = log(1 + x) = f (0) + xf 0 (0) + f (0) + . . . + f (0) + Rn

2! (n − 1)!

where

n−1

xn

n−1 (n) n n 1−θ 1

Rn = (1 − θ) f (θx) = (−1) x , 0<θ<1

(n − 1)! 1 + θx 1 + θx

2

— this is Cauchy’s form of the remainder — note that we consider Cauchy’s form of the

remainder as Lagrange’s form of the remainder does not enable us to prove that Rn → 0 as

n → ∞. when −1 < x < 0.

1−θ 1 1

Since 0 < θ < 1 and |x| < 1, 0 < < 1. Also, < . Now since xn → 0

1 + θx 1 + θx 1 − |x|

as n → ∞, we get Rn → 0 as n → ∞.

Now using the fact that f (n) (0) = (−1)n−1 (n − 1)!, we get the required result

x2 x3 x4

log(1 + x) = x − + − + ...

2 3 4

1 − x5

log(1 + x + x2 + x3 + x4 ) = log( ) = log(1 − x5 ) − log(1 − x)

1−x

x10 x15 x2 x3 x4

= −(x5 + + + . . .) + (x + + + + . . .)

2 3 2 3 4

x2 x3 x4 1 x6 x7 x8 x9 x10 1

= x+ + + + x5 −1 + + + + + − 1 + ...

2 3 4 5 6 7 8 9 2 5

which is the required expansion.

Question 2(c) Find the volume generated by revolving y 2 = 4ax about the latus rectum.

Solution.

the perpendicular on the latus rectum x = a. P M

Then P M = a − x. SM = y, where S is

the focus (a, 0). Because of symmetry we S (a, 0)

can confine ourselves to the first quadrant, x=a

0 ≤ y ≤ 2a.

y 2 = 4ax

Z Z 2a

2

V = 2 π(P M ) d(SM ) = 2π (a2 − 2ax + x2 ) dy

0

2a

y2 y4

Z

= 2π + a2 − dy

0 2 16a2

2a

y3 y5 8a3 32a5

2 3

= 2π a y − + = 2π 2a − +

6 80a2 0 6 80a2

30 − 20 + 6 32πa3

= 2πa3 =

15 15

3

Paper II

Question 3(a) Show that for the function

( 2

x y2

x2 +y 2

, (x, y) 6= (0, 0)

f (x, y) =

0, x = 0, y = 0

1. fx is not differentable at (0, 0).

2. fyx is not continuous at (0, 0).

3. fxy (0, 0) = fyx (0, 0).

Solution.

1. For (x, y) 6= (0, 0),

(x2 + y 2 )2xy 2 − x2 y 2 (2x) 2xy 4

fx (x, y) = = 2

(x2 + y 2 )2 (x + y 2 )2

2.

h2 y 2

f (h, y) − f (0, y) h2 +y 2

−0 hy 2

fx (0, y) = lim = lim = lim =0

h→0 h h→0 h h→0 h2 + y 2

3.

f (x + h, 0) − f (x, 0)

fx (x, 0) = lim =0

h→0 h

In particular, fx (0, 0) = 0.

fx (0,k)−fx (0,0)

4. fyx (0, 0) = limk→0 k

=0

f (x,k)−f (x,0) x2 k

5. fy (x, 0) = limk→0 k

= limk→0 x2 +k2

= 0 when x 6= 0.

f (0,k)−f (0,0)

6. fy (0, 0) = limk→0 k

= 0.

fy (h,0)−fy (0,0)

7. fxy (0, 0) = limh→0 h

= 0.

Thus we have shown that fxy (0, 0) = fyx (0, 0) = 0.

8. For (x, y) 6= (0, 0),

∂ ∂ 2xy 4

fyx (x, y) = fx (x, y) =

∂y ∂y (x2 + y 2 )2

8xy 3 (x2 + y 2 )2 − 2xy 4 · 2(x2 + y 2 ) · 2y

=

(x2 + y 2 )4

3

8xy 3 (x2 + y 2 ) − 8xy 5 8x3 y 3

xy

= = 2 =8 2

(x2 + y 2 )3 (x + y 2 )3 x + y2

Now from question 1(a) above, it follows that fyx is not continuous at (0, 0).

4

9. If fx were differentiable at (0, 0), we would have

√

fx (h, k) = fx (0, 0) + hfxx (0, 0) + kfyx (0, 0) + h2 + k 2 φ(h, k)

2hk 4 1

φ(h, k) = √

(h2 + k 2 )2 h2 + k 2

2m4

Now we can see that limh→0,k→0 φ(h, k) does not exist — if k = mh, φ(h, mh) = 5 ,

(1+m2 ) 2

so limh→0 φ(h, mh) is different for different values of m (say 0, 1).

5

UPSC Civil Services Main 1986 - Mathematics

Calculus

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

( 1

e1− x2 sin x1 , x 6= 0

f (x) =

0, x=0

1 x

Z

lim (x − 3y)f (y) dy

x→0 x2 0

Solution.

f (x) − f (0)

1. By definition, f 0 (0) = lim , if this limit exists.

x→0 x

f (x) − f (0) 1 1 1

lim = lim e1− x2 sin

x→0 x x→0 x x

1 − 12 1

= e lim 2 e x lim x sin

x→0 x x→0 x

1 1

= e lim te−t · 0 Letting t = 2

, also lim x sin = 0

t→∞ x x→0 x

= 0 ∵ lim te−t = 0

t→∞

1

2. Define

Z x

F (x) = (x − 3y)f (y) dy

0

Z x Z x

= x f (y) dy − 3 yf (y) dy

0 0

Z x Z x

0

F (x) = f (y) dy + xf (x) − 3xf (x) = f (y) dy − 2xf (x)

0 0

F 00 (x) = f (x) − 2f (x) − 2xf 0 (x) = −f (x) − 2xf 0 (x)

Note that F 00 (x) exists because f 0 (x) exists.

F (x)

The required limit is lim 2 . Since lim F (x) = 0, lim x2 = 0 and both are differen-

x→0 x x→0 x→0

tiable, we can apply L’Hospital’s rule to get

F (x) F 0 (x)

lim = lim

x→0 x2 x→0 2x

F (x) F 00 (x) −f (x) − 2xf 0 (x) f (0)

lim 2

= lim = lim =−

x→0 x x→0 2 x→0 2 2

since f 0 (x) exists and is continuous, so limx→0 xf 0 (x) = 0.

Thus Z x

1 f (0)

lim 2 (x − 3y)f (y) dy = −

x→0 x 0 2

Question 1(b) Use Rolle’s theorem to establish that under suitable conditions (to be stated)

f (a) f (b) f (a) f 0 (ξ)

g(a) g(b) = (b − a) g(a) g 0 (ξ) , a < ξ < b

nbn−1 (a − b) < an − bn < nan−1 (a − b)

where a > b and n > 1.

Solution. Let f (x), g(x) be continuous in the closed interval [a, b] and differentiable in the

open interval (a, b). Let

f (a) f (x) x − a f (a) f (b)

F (x) = −

g(a) g(x) b − a g(a) g(b)

x−a

= [f (a)g(x) − g(a)f (x)] − [f (a)g(b) − g(a)f (b)]

b−a

Thus F (x) is

2

1. continuous in the closed interval [a, b].

2. differentiable in the open interval (a, b)

3. F (a) = F (b) = 0.

Thus F (x) satisfies the requirements of Rolle’s theorem, consequently, there exists ξ, a <

ξ < b such that F 0 (ξ) = 0. But

1

F 0 (ξ) = [f (a)g 0 (ξ) − g(a)f 0 (ξ)] − [f (a)g(b) − g(a)f (b)]

b−a

f (a) f 0 (ξ)

− 1 f (a) f (b) = 0

= 0

g(a) g (ξ) b − a g(a) g(b)

therefore

f (a) f (b) 0

= (b − a) f (a) f 0 (ξ) , a < ξ < b

g(a) g(b) g(a) g (ξ)

Let f (x) = xn , g(x) = 1, then the above result implies

n n n

b a b nξ n−1

1 1 = (a − b) 1

, b < ξ < a

0

or an − bn = (a − b)nξ n−1 . Now since bn−1 < ξ n−1 < an−1 , we get

nbn−1 (a − b) < an − bn < nan−1 (a − b), as required.

(ax3 + by 3 )n ∂ 2u ∂ 2u ∂ 2u

Question 1(c) If u = + xf ( xy ), find the value of x2 2 + 2xy + y2 2 .

3n(3n − 1) ∂x ∂x ∂y ∂y

Solution. We use the result proved in 2006, question 2(b): If f (x, y) is a homogeneous

function of degree n possessing continuous partial derivatives of degree 2,

∂ 2f ∂ 2f 2

2∂ f

x2 + 2xy + y = n(n − 1)f

∂x2 ∂x ∂y ∂y 2

(ax3 + by 3 )n

Let v = , then v is homogeneous of degree 3n, so

3n(3n − 1)

2

2∂ v ∂ 2v 2

2∂ v (ax3 + by 3 )n

x 2

+ 2xy +y 2

= 3n(3n − 1) = (ax3 + by 3 )n

∂x ∂x ∂y ∂y 3n(3n − 1)

Let w = xf ( xy ), then w is homogeneous of degree 1, so

∂ 2w ∂ 2w 2

2∂ w

x2 + 2xy + y = 1(1 − 1)w = 0

∂x2 ∂x ∂y ∂y 2

Now u = v + w, so adding the above equations we have

∂ 2u ∂ 2u 2

2∂ u

x2 + 2xy + y = (ax3 + by 3 )n

∂x2 ∂x ∂y ∂y 2

3

Question 2(a) 1. Without evaluating the involved integrals, show that

1

Z x Z

t dt x dt

+ =0

1 1 + t2 1 t(1 + t2 )

Z a+T

2. If f (x) is periodic of period T , show that f (t) dt is independent of a.

a

Solution.

du t 1 u2 u

1. Let t = u1 , so that dt = − 2

and 2

= 2

= , implying that

u 1+t u 1+u 1 + u2

Z x Z 1 Z 1

t dt x u du x du

2

= 2

− 2

= − 2

1 1+t 1 1+u u 1 u(1 + u )

Thus 1

Z x Z

t dt x dt

+ =0

1 1 + t2 1 t(1 + t2 )

Z a+T

2. Define F (a) = f (t) dt. We shall prove that F (a) = F (b) for any b. In particular

Ra T

F (a) = F (0) = 0

f (t) dt which is independent of a.

Z b+T Z a+T

F (b) − F (a) = f (t) dt − f (t) dt

b a

Z b+T Z a+T Z b

= f (t) dt − f (t) dt − f (t) dt

b b a

Z b+T Z b Z b

= f (t) dt + f (t) dt − f (t) dt

b a+T a

Z b+T Z b

= f (t) dt − f (t) dt

a+T a

Z b Z b

= f (u + T ) du − f (t) dt ∵u+T =t

a a

Z b Z b

= f (u) du − f (t) dt = 0 ∵ f (u + T ) = f (u)

a a

4

Question 2(b) Find the volume of the solid generated by revolving one arc of x = a(t −

sin t), y = a(1 − cos t) about its base.

Solution. One arc is given by 0 ≤ t ≤ 2π and the base is the x-axis. Thus

Z 2π

dx

V = π y 2 dt

dt

Z0 2π Z 2π

2 2 3 t 3

= π a (1 − cos t) a(1 − cos t) dt = πa 2 sin2 dt

0 0 2

Z π

3 t

= 8πa 2 sin6 θ dθ (θ = )

0 2

Z π

2

= 32πa3 sin6 θ dθ (sin(π − θ) = sin θ, so double the integral and half the limit)

0

5·3·1 π

= 32πa3 = 5π 2 a3

6·4·2 2

Z 2 Z x − 21

Question 2(c) Evaluate (x−y)2 +2(x+y)+1 dx dy by using the transformation

0 0

x = u(1 + v), y = v(1 + u). Assume u, v are positive in the region concerned.

Solution.

1. The Jacobian of the transformation is

∂(x, y) 1 + v u

= =1+u+v >0

∂(u, v) v 1 + u

because u ≥ 0, v ≥ 0.

2. The region of integration in the xy-plane is the triangle bounded by the lines y = 0, y =

x, x = 2.

3. The region of integration in the uv-plane lies in the first quadrant as u ≥ 0, v ≥ 0.

Clearly

x=y ⇔ u=v

x = 2 ⇔ u(1 + v) = 2

y = 0 ⇔ v = 0 (∵ 1 + u > 0)

(1, 1)

v-axis

u(1 + v) = 2

5

The curve u(1 + v) = 2 meets the u-axis at (2, 0) and the line u = v at (1, 1). The

region of integration is bounded by v = 0, v = u, u(1 + v) = 2 and therefor consists

of the triangle with vertices (0, 0), (1, 0), (1, 1) and the portion bounded by u = 1, v =

0, u(1 + v) = 2, in which u varies from 1 to 2, and v varies from 0 to 2−u u

.

4.

− 12 − 21

(x − y)2 + 2(x + y) + 1 = (u − v)2 + 2(2uv + u + v) + 1

1

= (u2 + v 2 + 2uv + 2u + 2v + 1)− 2 = (1 + u + v)−1

Thus

Z 2 Z x − 21

2

I = (x − y) + 2(x + y) + 1 dx dy

0 0

2

Z 1 Z u Z 2 Z

u

−1

= dv du + dv du

0 0 1 0

Z 2

1 2

= + du − 1

2 1 u

1

= 2 log 2 −

2

Paper II

Question 3(a) Find the maximum and minimum values of f (x, y) = 7x2 + 8xy + y 2 where

x, y are constrained by the relation x2 + y 2 = 1.

Solution. Let F (x, y) = 7x2 +8xy +y 2 +λ(x2 +y 2 −1), where λ is Lagrange’s undetermined

multiplier. For extreme values

∂F ∂F

= 14x + 8y + 2λx = 0, = 8x + 2y + 2λy = 0

∂x ∂y

follows that

7 + λ 4

= (7 + λ)(1 + λ) − 16 = λ2 + 8λ − 9 = 0 ⇒ λ = −9, 1

4 1 + λ

∂F ∂F

x +y = 14x2 + 8xy + 2λx2 + 8xy + 2y 2 + 2λy 2 = 2f + 2λ(x2 + y 2 ) = 0

∂x ∂y

6

Since x2 + y 2 = 1, we get f = −λ ⇒ at stationary points f = 9, −1. Thus the maximum

value of f is 9, minimum value is −1.

Check: We have found the maximum and minimum values without finding the stationary

points.

1 2

λ = −9 ⇒ x = 2y, x2 + y 2 = 1 ⇒ 5y 2 = 1 ⇒ y = ± √ , x = ± √

5 5

1 2

λ = 1 ⇒ y = −2x, x2 + y 2 = 1 ⇒ 5x2 = 1 ⇒ x = ± √ , y = ∓ √

5 5

5

+ 165

+ 15 = 9. In case (2) f (x, y) = 57 − 16

5

+ 54 = −1, confirming the

above.

Note: The question could also be done by substituting x = cos t, y = sin t, and then

f (x, y) = 7 cos2 t+8 cos t sin t+sin2 t, which is now a function of one variable. Differentiating

and letting the derivative be 0, we get −14 cos t sin t + 8(cos2 t − sin2 t) + 2 sin t cos t = 0.

Let z = tan t, then 2z 2 + 3z − 2 = 0 ⇒ z = −2, 21 ⇒ (x, y) = (± √15 , ∓ √25 ), (± √25 , ± √15 ) ⇒

f (x, y) = 9, −1.

7

UPSC Civil Services Main 1987 - Mathematics

Calculus

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

1 9 1 9

Question 1(a) If x1 = x+ and for n > 0, xn+1 = xn + , find the value of

2 x 2 xn

limn→∞ xn (x > 0 is assumed).

x + x9 9

Now x1 = ≥ x · = 3, because the arithmetic mean is always ≥ the geometric

2 x

mean of two positive numbers. Similarly xn ≥ 3 for all n ≥ 1, thus the sequence {xn } is

bounded below. Moreover

1 9 1

xn+1 − xn = xn + − xn = (9 − x2n ) ≤ 0

2 2xn 2xn

as x2n ≥ 9. Thus {xn } is a monotonically decreasing sequence. Let l be the greatest lower

bound of {xn } — then

1 9 1 9 1 9

l = lim xn = lim xn−1 + = lim xn−1 + = l+

n→∞ n→∞ 2 xn−1 2 n→∞ limn→∞ xn−1 2 l

Thus 2l2 = l2 + 9 ⇒ l = 3 (l = −3 is not admissible as all elements of the sequence are ≥ 3).

Thus limn→∞ xn = 3.

1

Question 1(b) 1. If x = −a, h = 2a, f (x) = x 3 , find θ from the mean value theorem:

f (x + h) = f (x) + hf 0 (x + θh).

functional relationship between u, v, w and find the relationship, if any.

Solution.

1

1. Substituting the given values, we get f (a) = f (−a) + 2af 0 (−a + 2aθ). Now f (x) =

1 2

x 3 ⇒ f 0 (x) = 13 x− 3 , so

1 12a 2

a 3 = (−a) 3 + (−a + 2aθ)− 3

3

1 2a − 2 2

= −a 3 + a 3 (2θ − 1)− 3

3

2 2

⇒2 = (2θ − 1)− 3

3

2

⇒3 = (2θ − 1)− 3

⇒ 27 = (2θ − 1)−2

1

⇒ ±√ = 2θ − 1

27

So θ = 21 (1 ± 1

√

3 3

).

2. To establish that u, v, w have a functional relation, we have to show that the Jacobian

∂(u, v, w)

= 0. Clearly

∂(x, y, z)

1 1 −1 1 1 0

∂(u, v, w)

=1 −1 1 =1

−1 0 = 0

∂(x, y, z) 2x 2(y − z) −2(y − z) 2x 2(y − z) 0

Now u2 = x2 + (y − z)2 + 2x(y − z), v 2 = x2 + (y − z)2 − 2x(y − z), thus u2 + v 2 = 2w,

which is the desired functional relationship.

x n1 +1 + y n1 +1 12

Question 1(c) If u = csc−1 , show that

x+y

∂ 2u ∂ 2u 2

2∂ u 1

x2 2

+ 2xy + y 2

= 2 tan u(2n + sec2 u)

∂x ∂x ∂y ∂y 4n

1

Solution. Let v = csc u. Then v is a homogeneous function of degree 2n

, because

(ax) n1 +1 + (ay) n1 +1 12 1

x n1 +1 + y n1 +1 12 1

v(ax, ay) = = a 2n = a 2n v(x, y)

a(x + y) x+y

2

Applying Euler’s theorem to v,

∂v ∂v 1

x +y = v

∂x ∂y 2n

∂u ∂u 1 ∂v ∂u

−x csc u cot u − y csc u cot u = csc u ∵ v = csc u, = − csc u cot u

∂x ∂y 2n ∂x ∂x

∂u ∂u 1

x +y = − tan u (1)

∂x ∂y 2n

2 2

∂u ∂ u ∂ u 1 ∂u

+x 2 +y = − sec2 u Differentiating (1) w.r.t. x (2)

∂x ∂x ∂x ∂y 2n ∂x

2

∂ u ∂u ∂ 2u 1 ∂u

x + +y 2 = − sec2 u Differentiating (1) w.r.t. y (3)

∂x ∂y ∂y ∂y 2n ∂y

2 2

∂ 2u

∂ u ∂ u 1 ∂u ∂u

x2 2 + 2xy + y2 2 = (−1 − 2

sec u) x +y (2) × x + (3) × y

∂x ∂x ∂y ∂y 2n ∂x ∂y

1 1

= − (2n + sec2 u)(− tan u) From (1)

2n 2n

1

= tan u(2n + sec2 u)

4n2

π

∞

tx−1

Z Z

2

2x−1 2y−1 1

sin θ cos θ dθ = dt, x, y > 0

0 2 0 (1 + t)x+y

1 dx π

< 1 <

2 0 (4 − x2 − x3 ) 2 6

Solution.

Z 1

1. By definition B(y, x) = z y−1 (1 − z)x−1 dz.

0

2

Put z = cos θ, dz = −2 cos θ sin θ dθ, to get

Z 0

B(y, x) = cos2y−2 θ sin2x−2 θ (−2 cos θ sin θ) dθ

π

2

Z π

2

= 2 sin2x−1 θ cos2y−1 θ dθ

0

3

1

Now, put z = t+1

in B(y, x),

Z 1

B(y, x) = z y−1 (1 − z)x−1 dz

0

Z 0

1 y−1 1 x−1 −dt

= 1−

t+1 t+1 (1 + t)2

Z∞∞

tx−1

= dt

0 (1 + t)x+y

Z π

1 ∞ tx−1

Z

2

2x−1 2y−1

sin θ cos θ dθ = dt

0 2 0 (1 + t)x+y

as required.

1 1

4 − x2 + x3 < 4 ⇒ < 1 .

2 (4 − x2 − x3 ) 2

1 1

On the other hand, 4 − x2 + x3 > 4 − x2 for x > 0, so (4 − x2 + x3 )− 2 < (4 − x2 )− 2 .

Hence

Z 1 Z 1 Z 1

1 dx dx

dx < 1 < √

0 2 4 − x2

2 3

0 (4 − x − x ) 2 0

Z 1 1

1 dx −1 x π

< 1 < sin =

2 2 3

0 (4 − x − x ) 2 2 0 6

x3

Question 2(b) Find the volume of the solid generated by revolving the curve y 2 = ,a >

2a − x

0 about its asymptote x = 2a.

Solution. Consider a thin vertical strip of thickness dx, and rotate it about the asymptote

4

r

x3

— its volume is 2πrh dx = 2π(2a − x)2 dx. Thus

2a − x

s

Z 2a

x3

V = 4π (2a − x) dx

0 2a − x

√

Z 2a

3

= 4π 2a − x x 2 dx

0

Put x = 2a sin2 θ ⇒ dx = 4a sin θ cos θ dθ

Z π√

2 3

= 4π 2a cos θ(2a) 2 sin3 θ 4a sin θ cos θ dθ

0

Z π

2

3

= 64πa sin4 θ cos2 θ dθ

0

3·1·1 π

= 64πa3 = 2π 2 a3

6·4·2 2

ZZ

1 1 2

Question 2(c) Evaluate x 2 y 3 (1 − x − y) 3 dx dy where D is the domain bounded by the

D

lines x = 0, y = 0, x + y = 1.

Solution. We convert this to a Dirichlet integral using the standard transformation. Put

x + y = u, y = uv, so that dx dy = u du dv, (see 1989, question 3(a) for example).

Z 1Z 1

1 1 1 1 2

I = u 2 (1 − v) 2 u 3 v 3 (1 − u) 3 u du dv

Z0 1 Z0 1

11 2 1 1

= u 6 (1 − u) 3 v 3 (1 − v) 2 du dv

0 0

Γ( 17

6

)Γ( 53 ) Γ( 34 )Γ( 32 ) 2

3

Γ( 23 ) 31 Γ( 13 )Γ( 32 ) 16 2 1

= = = Γ Γ

Γ( 276

) Γ( 17 6

) 753

222

Γ( 32 ) 945 3 3

But Γ( 23 )Γ( 13 ) = π

sin π3

= 2π

√

3

. Thus I = 32π

√ .

945 3

Paper II

continuous only at x = 12 .

5

Question 4(a) Find the maximum and minimum value of f (x, y) = xy subject to the con-

dition that x2 + y 2 + xy = a2 .

multiplier. For extreme values,

∂F ∂F y x

= y − 2λx − λy = 0, = x − 2λy − λx = 0 ⇒ λ = =

∂x ∂y 2x + y 2y + x

x2 + y 2 + xy = a2 we get

a2

1. x = y ⇒ 3x2 = a2 ⇒ x = y = ± √a3 ⇒ f (x, y) = 3

f (x, y) = −a2 .

2

Thus the required maximum value is a3 and the required minimum is −a2 .

Note: In this problem there was no need to check the nature of the critical points, as the

maximum amd minimum values occur at these points. If it were required, it could be done

as follows.

d2 F = −2λ(dx)2 − 2λ(dy)2 + 2(1 − λ)dx dy

2x+y

Now x2 + y 2 + xy = a2 ⇒ 2x dx + x dy + y dx + 2y dy = 0, or dy = − x+2y dx. Thus

2

2 2 2x + y 2x + y

d F = −2λ(dx) − 2λ (dx)2 − 2(1 − λ)(dx)2

x + 2y x + 2y

Case 2: x = y = ± √a3 , λ = 31 .

2 4 2

d2 F = − (dx)2 − (dx)2 − (dx)2 < 0

3 3 3

so we have a local maximum at x = y = ± √a3 .

6

UPSC Civil Services Main 1988 - Mathematics

Calculus

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

(n) n (n−2) n (n−4) nπ

f (0) − f (0) + f (0) − . . . = sin

2 4 2

Solution. Clearly sin x = f (x) cos x. Using Leibnitz’s formula for the derivative of the

product of two functions, we get

n

dn sin x X n (n−r)

= f (x) cos(r) (x)

dxn r=0

r

dr rπ dr rπ dr

Now cos x = cos(x + 2

) and sin x = sin(x + 2

). Therefore, cos x = 0 when

dxr r

dxr n

dxr

d n d

x = 0, n odd, and r

cos x = (−1) 2 when x = 0, n even. Also, sin x = sin nπ

2

when

dx dxn

x = 0.

Thus

nπ (n) n (n−2) n (n−4)

sin = f (0) − f (0) + f (0) − . . .

2 2 4

.

multiplier. For stationary values,

∂F ∂F ∂F

= 2x + λ = 0, = 2y + λ = 0, = 2z + λ = 0

∂x ∂y ∂z

1

Thus λ = −2x = −2y = −2z ⇒ x = y = z. From x + y + z = k we have x = y = z = k3 .

Now

∂ 2F ∂ 2F ∂ 2F ∂ 2F ∂ 2F ∂ 2F

d2 F = (dx)2

+ (dy)2

+ (dz)2

+ dx dy + dy dz + dz dx

∂x2 ∂y 2 ∂z 2 ∂x ∂y ∂y ∂z ∂z ∂x

= 2(dx)2 + 2(dy)2 + 2(dz)2

Now dx + dy + dz = 0 from x + y + z = k, so substituting dz = −dx − dy, we get

d2 F = 2(dx)2 + 2(dy)2 + 2(dx + dy)2 = 4(dx)2 + 4(dy)2 + 4dx dy

dy 2 3 2

= 4 dx + + (dy) > 0

2 4

k k2 2 2 k2

Thus x2 +y 2 +z 2 is minimum when x = y = z = 3

and the minimum value of 9

+ k9 + k9 = 3

.

x3 − xy 2 − 2xy + 2x − y = 0

and show that they cut the curve again in points which lie on the line 3x − y = 0.

Solution. Since the coefficient of the highest degree term of y in the equation, i.e.y 2 is −x,

it follows that x = 0 is an asymptote parallel to the y axis.

There is no asymptote parallel to the x-axis, as the coefficient of x3 , the highest degree

term of x peresent in the equation, is a constant.

If y = mx + c is an asymptote, then m is given by Φ3 (m) = 1 − m2 = 0 ⇒ m = ±1. c

Φ2 (m)

is given by − 0 provided it is not indeterminate, where Φr are homogeneous terms of

Φ3 (m)

degree r present in the equation. Thus c = − −2m−2m

= −1. Thus the asymptotes of the type

y = mx + c are y = x − 1, y + x = −1.

The joint equation of asymptotes is given by x(y − x + 1)(y + x + 1) = 0 or x(y 2 − x2 +

2y + 1) = xy 2 − x3 + 2yx + x = 0, or P3 = x3 − xy 2 − 2yx − x = 0.

Thus the point of intersection of the curve and the asymptotes lie on the line 3x − y = 0

as the equation of the given curve is P3 + 3x − y = 0.

Note: The asymptotes parallel to the y-axis can also be found as follows:

Let x = my + d be an asymptote. Dividing the equation by y 3 and letting y → ∞, we

get m3 − m = 0, as xy → m when y → ∞. Thus m = 0, ±1.

If m = 0, then x = d, so d3 − dy 2 − 2dy + 2d − y = 0. Dividing by y 2 and letting y → ∞

we get d = 0, so x = 0 is an asymptote.

Similarly we can find d = 1 when m = ±1. This will determine all three asymptotes.

Question 2(a) Find the center of gravity of one loop of the lemniscate of Bernoulli r2 =

a2 cos 2θ.

Solution.

2

Because of symmetry, y = 0. The area of

the lemniscate is

Z π Z π π4

1 4 2 4 sin 2θ a2

r dθ = a2 cos 2θ dθ = a2 =

2 − π4 0 2 0 2

π

Now θ= 4

Z π

2 4

3

θ = − π4

x = r cos θ dθ

3a2 − π4

Z π

4a 4 3

= (1 − 2 sin2 θ) 2 cos θ dθ

3 0

√ cos t dt

Let 2 sin θ = sin t ⇒ cos θ dθ = √

2

Z π Z π √

4a 2

2 3 4a 2

4 4a 3 · 1 · π π 2a

= √ (1 − sin t) 2 cos t dt = √ cos t dt = √ =

3 2 0 3 2 0 3 24·2·2 8

√

the centroid is π 82a , 0 .

Thus

ZZ

3

Question 2(b) Evaluate x 2 y 2 (1 − x2 − y 2 ) dx dy over the positive quadrant of the circle

x2 + y 2 = 1.

dX dY

Solution. Substitute x2 = X, y 2 = Y ⇒ dx dy = 1 1 and the region of integration is

4X 2 Y 2

transformed to X ≥ 0, Y ≥ 0, X + Y ≤ 1. The required integral becomes

ZZ ZZ

3 dX dY 1 1 1

I= X Y (1 − X − Y )

4

1 1 = X 4 Y 2 (1 − X − Y ) dX dY

X≥0,Y ≥0 4X 2 Y 2 4 X≥0,Y ≥0

X+Y ≤1 X+Y ≤1

∂(X, Y ) 1 − v −u

= =u

∂(u, v) v u

Y

Also, X ≥ 0, Y ≥ 0, X + Y ≤ 1 ⇒ 0 ≤ u ≤ 1, 0 ≤ v ≤ 1(∵ v = X+Y

).

1 1 1 1

Z Z

1 1 1

I = u 4 (1 − v) 4 u 2 v 2 (1 − u)u du dv

4 0 0

1 1 7

Z Z 1

1 1

= u 4 (1 − u) du v 2 (1 − v) 4 dv

4 0 0

1 11 3 5 1 Γ( 114

)Γ(2) Γ( 32 )Γ( 54 ) 1 Γ( 32 )Γ( 54 )

= B( 4 , 2)B( 2 , 4 ) = =

4 4 Γ( 19 4

) Γ( 114

) 4 Γ( 19 4

)

3

Question 2(c) Show that the volume of the solid obtained by revolving the curve (a−x)y 2 =

π 2 a3

a2 x about its asymptote is .

2

Solution. The asymptote is x = a, because a − x is the coefficient of the highest degree

term in y present in the equation of the curve. Thus

Z ∞

V = π(a − x)2 dy

−∞

Z ∞ Z ∞

ay 2 2 6 dy

= 2π a− 2 2

dy = 2πa

0 a +y 0 (a + y 2 )2

2

Z π Z π

6

2 a sec2 θ dθ 3

2

2 31 π π 2 a3

V = 2πa 4 2 2

= 2πa cos θ dθ = 2πa =

0 a (1 + tan θ) 0 22 2

Paper II

x2 y 2 32

ZZ

1− − 2 dx dy

a2 b

x2 y 2

over the area of the ellipse + 2 = 1.

a2 b

Solution.

ZZ Let x = aX, y = bY , so dx dy = ab dX dY . The given integral becomes

3

ab (1 − X − Y 2 ) 2 dX dY over the circle X 2 + Y 2 ≤ 1. Changing to polar coordinates,

2

π

Z

2

Z 1

3

I = 4ab (1 − r2 ) 2 r dr dθ

0 0

Z 1

3

= 2πab (1 − r2 ) 2 r dr

0

Put r2 = t, 2r dr = dt

Z 1

3 2 5 1

i 2πab

= πab (1 − t) dt = −πab (1 − t)

2 2 =

0 5 0 5

4

Question 4(a) Show that a local extreme value of f given by

f (x) = xk1 + . . . + xkn , x = (x1 , . . . , xn )

subject to the condition x1 + . . . + xn = a is ak n1−k .

Solution. Let F (x) = xk1 +. . .+xkn −λ(x1 +. . .+xn −a) where λ is Lagrange’s undetermined

multiplier. For extreme values, for all i

∂F λ

= kxk−1

i − λ = 0 ⇒ xk−1

i =

∂xi k

λ a

Thus xk−1

1 = . . . = xk−1

n = , or x1 = . . . = xn . Using x1 + . . . + xn = a we get xi = , so

n

k n

X a k nak

the extreme value is = k = ak n1−k .

i=0

n n

Note: To decide the nature of the extreme value, we consider

d2 F = k(k − 1)xk−2 2 2

1 [(dx1 ) + . . . + (dxn ) ]

minimum. If 0 < k < 1, d2 F < 0, and we have a maximum. This is not required for this

question.

Question 4(b) The function f : R2 → R is given by

( 2

x y

x2 +y 2

, (x, y) 6= (0, 0)

f (x, y) =

0, (x, y) = (0, 0)

Prove that at (0, 0) f is continuous and possesses all directional derivatives, but is not dif-

ferentiable.

Solution. Let > 0 and δ = , then

x2 y r2 cos2 θ r sin θ p

|f (x, y) − f (0, 0)| = 2 = ≤ r = x2 + y 2

x + y2 r2

Thus for any (x, y) belonging to the open disc with center (0, 0) and radius δ = i.e. for

x2 + y 2 < 2 , we get |f (x, y) − f (0, 0)| < , so f (x, y) is continuous at (0, 0).

Let v = (cos θ, sin θ), then by definition the directional derivative at (a, b) in the direction

of v is given by

f (a + t cos θ, b + t sin θ) − f (a, b)

lim

t→0 t

when this limit exists, and it is denoted by fv (a, b).

Here (a, b) = (0, 0), so

f (t cos θ, t sin θ) t2 cos2 θ t sin θ

fv (0, 0) = lim = lim = cos2 θ sin θ

t→0 t t→0 t3

Thus all directional derivatives exist.

However for f to be differentiable, all directional derivatives should be equal — this is

clearly not the case here, as the directional derivative for θ = 0 is 0, but for θ = π4 it is 2√1 2 .

5

UPSC Civil Services Main 1989 - Mathematics

Calculus

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

h2 00

f (a + h) = f (a) + hf 0 (a) + f (a + θh)

2!

where θ lies between 0 and 1, and prove that limh→0 θ = 13 .

Solution. Let

(a + h − x)2

φ(x) = f (a + h) − f (x) − (a + h − x)f 0 (x) − A

2!

where A is so determined that φ(a + h) = φ(a). Clearly φ(a + h) = 0, and φ(a) = f (a + h) −

2 2

f (a) − hf 0 (a) − h2! A, so φ(a) = 0 ⇒ h2! A = f (a + h) − f (a) − hf 0 (a).

Now φ satisfies the requirements of Rolle’s theorem in [a, a + h], therefore there exists

θ ∈ (0, 1) such that φ0 (a + θh) = 0, note that a + θh ∈ (a, a + h). But

h2 00

f (a + h) = f (a) + hf 0 (a) + f (a + θh)

2!

as required.

Now consider the equality

h2 00

f (a + h) − f (a) − hf 0 (a) − 2

f (a) h2 f 00 (a + θh) − f 00 (a)

=

h3 2 h3

1

Taking the limit as h → 0 on both sides, we get

2

f (a + h) − f (a) − hf 0 (a) − h2 f 00 (a)

LHS = lim

h→0 h3

f (a + h) − f (a) − hf 00 (a)

0 0

= lim L’Hospital’s rule

h→0 3h2

f 00 (a + h) − f 00 (a)

= lim L’Hospital’s rule

h→0 6h

1 000

= f (a)

6

f 00 (a + θh) − f 00 (a) θ f 00 (a + θh) − f 00 (a)

RHS = lim = lim

h→0 2h h→0 2 θh

θ 000

= f (a)

2

Thus θ = 31 as h → 0, provided f 000 (a) 6= 0.

The question above is a particular case of the following:

Taylor’s Theorem: Let f (n−1) , the (n − 1)-th derivative of a real valued function be

continuous in the closed interval [a, a + h], and let f (n) exist in the open interval (a, a + h),

then there exists a real number θ, 0 < θ < 1, such that

hn−1 (n−1)

f (a + h) = f (a) + hf 0 (a) + . . . + f (a) + Rn

(n − 1)!

where

hn (1 − θ)n−p (n)

Rn = f (a + θh), 0 < θ < 1

(n − 1)!p

Proof: The condition that f (n−1) (x) is continuous in [a, a + h] ⇒ f, f 0 , f 00 , . . . , f (n−2) are

continuous in [a, a + h]. Let

(a + h − x)n−1 (n−1)

φ(x) = f (x) + (a + h − x)f 0 (x) + . . . + f (x) + A(a + h − x)p

(n − 1)!

where A is a constant to be determined so that φ(a + h) = φ(a) i.e.

0 (hn−1 (n−1)

f (a + h) = f (a) + hf (a) + . . . + f (a) + Ahp

(n − 1)!

Now (i) φ(x) is continuous in [a, a + h] (ii) φ(x) is differentiable in (a, a + h) and (iii)

φ(a + h) = φ(a). Thus all requirements of Rolle’s theorem are satisfied,so there exists a real

number θ, 0 < θ < 1 such that φ0 (a + θh) = 0 (note that any real number c ∈ (a, a + h) can

be written as c = a + c−a

h

h, and since a < c < a + h, we get 0 < θ = c−a

h

< 1). But

(a + h − x)n−1 (n)

φ0 (x) = f (x) − pA(a + h − x)p−1

(n − 1)!

2

so

(h(1 − θ))n−1 (n) hn−p (1 − θ)n−p (n)

φ0 (a+θh) = f (a+θh)−pAhp−1 (1−θ)p−1 = 0 ⇒ A = f (a+θh)

(n − 1)! (n − 1)!p

Thus

hn−1 (n−1) hn (1 − θ)n−p (n)

f (a + h) = f (a) + hf 0 (a) + . . . + f (a) + f (a + θh)

(n − 1)! (n − 1)!p

If we put n = p we get

hn−1 (n−1) hn

f (a + h) = f (a) + hf 0 (a) + . . . + f (a) + f (n) (a + θh)

(n − 1)! n!

h2 00

f (a + h) = f (a) + hf 0 (a) + f (a + θh)

2!

Now we prove the second part: If f (n+1) (x) is continuous from the right at x = a, then

1

θ → n+1 as h → 0 provided f (n+1) (a) 6= 0.

In the given question, n = 2, and f (3) (x) is continuous in [a, a + h] so θ → 13 .

f (n+1) (x) is continuous from the right at x = a ⇒ there exists k, 0 < k < h, such that

f (n+1) (x) exists in [a, a + k]. Thus by the above theorem:

f (a + k) = f (a) + kf (a) + . . . + f (a) + f (a + θk)

(n − 1)! n!

n n+1

(k (n) k

f (a + k) = f (a) + kf 0 (a) + . . . + f (a) + f (n+1) (a + φk)

n! (n + 1)!

where 0 < θ < 1, 0 < φ < 1. Subtracting the first from the second,

k

f (n) (a + θk) − f (n) (a) = f (n+1) (a + φk)

n+1

Now we use Lagrange’s mean value theorem for f (n) (x) in [a, a + θk] and obtain

k

θkf (n+1) (a + θθ1 k) = f (n+1) (a + φk), 0 < θ1 < 1

n+1

1

Taking limit as h → 0 i.e. k → 0, we get limh→0 θ = n+1 as f (n+1) (a+θθ1 k) and f (n+1) (a+φk)

both tend to f (n+1) (a) as k → 0, and f (n+1) (a) 6= 0.

Question 1(b) Prove that the√volume of a right circular cylinder of greatest volume which

can be inscribed in a sphere is 33 times that of a sphere.

3

Solution. Let the radius of the sphere be a, and the radius and height of the cylinder be r

and h respectively. Then r2 + ( h2 )2 = a2 . Letting r = a cos θ, h = 2a sin θ, the volume of the

cylinder is 2πa3 cos2 θ sin θ.

dV

For extreme values, = 2πa3 [cos3 θ − 2 cos θ sin2 θ] = 0. Now cos θ = 0 ⇒ θ = π2 ⇒

dθ

r = 0, which is not admissible. So cos2 θ − 2 sin2 θ = 0 ⇒ tan θ = √12 as 0 < θ < π2 . Thus

q

sin θ = √13 , cos θ = 23 .

d2 V 3

q

2

= 2πa [−3 cos θ sin θ + 2 sin θ − 4 cos θ sin θ] < 0 for sin θ = 3 , cos θ = 23 . Thus

3 2 2 √1

dθ

V is maximum when tan θ = √12 .

√

V = 2πa3 23 · √1 = 4 πa3 · 3

√ 3 3 3

3

Thus V is 3

times the volume of the sphere.

∂ 2z

Question 1(c) Show that at the point of the surface xx y y z z = c where x = y = z, =

∂x ∂y

−[x log ex]−1 .

∂z ∂z 1 + log x log ex

(1 + log z) + (1 + log x) = 0 ⇒ =− =− (∗)

∂x ∂x 1 + log z log ez

∂z log ey

Similarly, =− .

∂y log ez

Differentiating (*) w.r.t. y,

∂ 2z 1 ∂z ∂z ∂ 2z 1 ∂z ∂z log ex log ey

(1 + log z) + =0⇒ =− =−

∂x ∂y z ∂y ∂x ∂x ∂y z log ez ∂y ∂x z(log ez)3

∂ 2z log ex log ex

Letting x = y = z, we get =− = −[x log ex]−1 as required.

∂x ∂y x(log ex)3

Question 2(a) Find the surface of the solid generated by revolving x = a cos3 t, y = a sin3 t

about the x-axis.

4

Solution. We confine ourselves to the first quadrant because of symmetry. The curve is

2 2 2

x3 + y 3 = a3 .

Z π

2 ds

S = 2 2πy dt

0 dt

Z π r

2

3 dx 2 dy 2

= 4π a sin t + dt

0 dt dt

Z π

2 1

sin3 t (−3a cos2 t sin t)2 + (3a sin2 t cos t)2 2 dt

= 4πa

0

Z π

2

= 4πa 3a sin4 t cos t dt

0

5

π2

2 sin t 12 2

= 12πa = πa

5 0 5

Question 2(b) If for a curve x sin θ + y cos θ = f 0 (θ) and x cos θ − y sin θ = f 00 (θ) then show

that S = f (θ) + f 00 (θ) + C.

y = f 0 (θ) cos θ − f 00 (θ) sin θ

dx

= f 0 (θ) cos θ + f 00 (θ) sin θ − f 00 (θ) sin θ + f 000 (θ) cos θ

dθ

= f 0 (θ) + f 000 (θ) cos θ

dy

= −f 0 (θ) sin θ + f 00 (θ) cos θ − f 00 (θ) cos θ − f 000 (θ) sin θ

dθ

= − f 0 (θ) + f 000 (θ) sin θ

Z r 2 2 Z

dx dy 0

f (θ) + f 000 (θ) dθ

S = + dθ =

dθ dθ

00

= f (θ) + f (θ) + C

as required.

ZZZ

1

Question 2(c) Evaluate (1 − z) 2 dx dy dz over the interior of the tetrahedron with

faces x = 0, y = 0, z = 0, x + y + z = 1.

5

Solution.

Z 1 Z 1−z Z 1−y−z

1

I = (1 − z) 2 dx dy dz

z=0 y=0 x=0

Z1 Z1−z

1

= (1 − z) 2 (1 − y − z) dy dz

z=0 y=0

1

y2

Z h i1−z

1

= (1 − z) 2 y − − zy dz

z=0 2 0

Z 1

1

h (1 − z)2 i

= (1 − z) 1 − z −

2 − z(1 − z) dz

z=0 2

Z 1

3

h (1 − z) i

= (1 − z) 2 1 − − z dz

z=0 2

Z 1 1

1 5 1 −2 7 1

= (1 − z) 2 dz = (1 − z) 2 =

z=0 2 2 7 0 7

Paper II

ZZ

1 1

(1 − x − y)3 x 2 y 2 dx dy

taken over the interior of the triangle whose vertices are the origin and the points (0, 1) and

π

(1, 0) is 480 .

∂(x, y) 1 − v −u

= u. u varies from 0 to 1, and v = y = y also

x = u(1 − v), and = u x+y

∂(u, v) v u

varies from 0 to 1.

Z 1 Z 1

1 1 1 1

I = (1 − u)3 u 2 (1 − v) 2 u 2 v 2 u du dv

Z0 1 Z0 1

3 3

= (1 − u)4−1 u3−1 (1 − v) 2 −1 v 2 −1 du dv

0 0

Γ(4)Γ(3) Γ( 23 )Γ( 32 )

=

Γ(7) Γ(3)

3! π π

= =

6! 4 480

as Γ( 12 ).

6

Question 3(b) Obtain the largest and the least values of 2(x + y + z) − xyz on the closed

ball x2 + y 2 + z 2 ≤ 9.

undetermined multiplier. For extreme values:

∂F ∂F ∂F

= 2 − yz + 2λx = 0, = 2 − xz + 2λy = 0, = 2 − yx + 2λz = 0

∂x ∂y ∂z

Subtracting the first two, z(x − y) + 2λ(x − y) = 0 ⇒ x = y or z = −2λ. Similarly from the

other pairs, we get y = z or x = −2λ and x = z or y = −2λ.

We now explicitly find extreme vaues to get the greatest and least values of f (x, y, z) =

2(x + y + z) − xyz

√

1. x = y = z: This gives us x2 + x2 + x2 = 9, so x = ± 3.

√ √ √ √

(a) x = y = z = 3: f (x, y, z) = 6 3 − 3 3 = 3 3.

√ √ √ √

(b) x = y = z = − 3: f (x, y, z) = −6 3 + 3 3 = −3 3.

2λ2 − 1

2 − 4λ2 + 2λz = 0 ⇒ z = . Thus

λ

2λ2 − 1 2

= 9 − 8λ2

λ

⇒ 4λ4 − 4λ2 + 1 = 9λ2 − 8λ4

⇒ 12λ4 − 13λ2 + 1 = 0

1

⇒ λ2 = 1,

12

λ = 1 ⇒ x = y = −2, z 2 = 1

λ = −1 ⇒ x = y = 2, z 2 = 1

1 1 25

λ = √ ⇒ x = y = − √ , z2 =

2 3 3 3

1 1 25

λ = − √ ⇒ x = y = √ , z2 =

2 3 3 3

We examine each of these 8 cases:

(b) x = y = −2, z = −1: f (x, y, z) = −10 + 4 = −6.

(c) x = y = 2, z = 1: f (x, y, z) = 10 − 4 = 6.

(d) x = y = 2, z = −1: f (x, y, z) = 6 + 4 = 10.

(e) x = y = − √13 , z = √5 :

3

f (x, y, z) = √6

3

− 5

√

3 3

= 13

√ .

3 3

7

(f) x = y = − √13 , z = − √53 : f (x, y, z) = − √143 + 5

√

3 3

= − 337

√ .

3

(g) x = y = √1 , z = √5 : f (x, y, z) = 14

√ − 5

√ = 37

√ .

3 3 3 3 3 3 3

√ = − 313

√ .

3 3 3 3

We need not consider the other possibilities like y = z, z = −2λ as the situation is symmetric

and no new values will result. The greatest value of f (x, y, z) is 10, and the least value is

-10.

We don’t consider d2 F , as it is not needed. We use the fact the all the maximum and

minimum values occur at the extreme values. Calculation of d2 F is a very lengthy process.

8

UPSC Civil Services Main 1990 - Mathematics

Calculus

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

Question 1(a) If a function f (x) of the real variable x has the first five derivatives 0 at a

given value x = a, show that it has a maximum or a minimum at x = a according as the

sixth derivative is negative or positive. What happens if only the first four derivatives are 0,

but not the fifth?

Solution. We shall take up the general case, which would cover both the cases —

Let f (n) (a) exist and f (n) (a) 6= 0. Let f (r) (a) = 0, 1 ≤ r ≤ n − 1. Then f (a) is not an

extreme value when n is odd, and if n is even, then f has a maximum at x = a if f (n) (a) < 0

and a minimum if f (n) (a) > 0.

Thus the above result is proved when n = 6. When n = 5, f (x) has neither maximum

nor minimum at x = a.

Proof: The existence of f (n) (a) imlies that f 0 (x), f 00 (x), . . . , f (n−1) (x) all exist in a certain

neighborhood of a, say (a − δ1 , a + δ1 ).

Case (1): If f (n) (a) > 0, then f (n−1) (x) is increasing at a i.e. there exists 0 < δ < δ1

such that f (n−1) (x) < f (n−1) (a) = 0 for a − δ < x < a, and f (n−1) (x) > f (n−1) (a) = 0 for

a < x < a + δ.

Case (2): If f (n) (a) < 0, then f (n−1) (x) is decreasing at a i.e. there exists 0 < δ < δ1

such that f (n−1) (x) > f (n−1) (a) = 0 for a − δ < x < a, and f (n−1) (x) < f (n−1) (a) = 0 for

a < x < a + δ.

Let h be such that |h| < δ, then by Taylor’s theorem

hn−1 (n−1)

f (a + h) = f (a) + hf 0 (a) + . . . + f (a + θh), 0 < θ < 1

(n − 1)!

hn−1 (n−1)

Thus f (a + h) − f (a) = f (a + θh) as f (r) (a) = 0, 1 ≤ r ≤ n − 2. Since

(n − 1)!

a + θh ∈ (a − δ, a + δ), we have the following conclusions:

1

1. n even and f (n) (a) > 0: If h < 0, then hn−1 < 0 and f (n−1) (a + θh) < 0 (by Case

(1) above, as a − δ < a + θh < a. If h > 0, then hn−1 > 0 and f (n−1) (a + θh) > 0 (as

a < a + θh < a + δ.

hn−1 (n−1)

In either case f (a + θh) > 0 for all h with |h| < δ, i.e. f (a + h) > f (a)

(n − 1)!

for |h| < δ, thus f (x) has a minimum at x = a.

hn−1 (n−1)

2. n even and f (n) (a) < 0: Using Case (2), we get f (a + θh) < 0 for all h

(n − 1)!

with |h| < δ, i.e. f (a + h) < f (a) for |h| < δ, thus f (x) has a maximum at x = a.

3. n odd and f (n) (a) > 0: If h < 0, then hn−1 > 0 and f (n−1) (a + θh) < 0, so f (a +

h) < f (a) for h < 0, |h| < δ. If h > 0, then hn−1 > 0 and f (n−1) (a + θh) > 0, so

f (a + h) > f (a) for 0 < h < δ. Thus f has neither minimum nor maximum at x = a.

The case for n odd and f (n) (a) < 0 is similar — h < 0 ⇒ f (a + h) > f (a), h > 0 ⇒

f (a + h) < f (a), so there is no extreme value at x = a.

Solution.

f 0 (x) = 2(x − 2)(x2 + 2bx + c)(x + 3)3 + (x − 2)2 (2x + 2b)(x + 3)3

+ 3(x − 2)2 (x2 + 2bx + c)(x + 3)2

f 0 (−1) = 2(−3)(1 − 2b + c)8 + 72(−2 + 2b) + 108(1 − 2b + c)

= −48 + 96b − 48c − 144 + 144b + 108 − 216b + 108c = −84 + 24b + 60c

Question 2(b) Assuming that the above condition is satisfied, examine the nature of f (x)

at its three critical points.

Solution.

2

1. x = 2.

(x − 2)2 (x2 + 2bx + c)(x + 3)3 = (x − 2)2 g(x)

f (x) =

f 0 (x) = 2(x − 2)g(x) + (x − 2)2 g 0 (x)

f 00 (x) = 2g(x) + 4(x − 2)g 0 (x) + (x − 2)2 g 00 (x)

f 00 (2) = 2g(2) = 250(4 + 4b + c)

= 250(4 + 14 − 10c + c) = 250 · 9(2 − c)

There is a maximum at x = 2 if 2 < c, and a minimum at x = 2 if 2 > c. The test

fails when c = 2. Considering f 000 (x) when x = 2, c = 2, 2b = −3,

f 000 (x) = 2g 0 (x) + 4g 0 (x) + 4(x − 2)g 00 (x) + (x − 2)2 g 000 (x) + 2(x − 2)g 00 (x)

f 000 (2) = 6 (4 + 2b)(2 + 3)3 + 3(4 + 4b + c)(2 + 3)2

= 6 125 + 75(4 − 6 + 2) 6= 0

so there is no maximum or minimum at x = 2 when c = 2.

2. x = −3

(x + 3)3 (x − 2)2 (x2 + 2bx + c) = (x + 3)3 h(x)

f (x) =

f 0 (x) = 3(x + 3)2 h(x) + (x + 3)3 h0 (x)

f 00 (x) = 6(x + 3)h(x) + 3(x + 3)2 h0 (x) + 3(x + 3)2 h0 (x) + (x + 3)3 h00 (x)

f 00 (−3) = 0

f 000 (x) = 6h(x) + 12(x + 3)h0 (x) + (x + 3)2 [some polynomial in x]

f 000 (−3) = 6h(−3) = 150(9 − 6b + c) = 150(9 − 21 + 15c + c) = 600(4c − 3)

So if c 6= 34 , f (x) does not have a maximum or minimum at x = −3. If c = 34 , then

so f (4) (−3) = 18 2(−5)(9 − 6b + c) + 25(−6 + 2b) = 18(−240 + 110b − 10c) < 0. Hence

if c = 43 , b = 13

8

, f (x) has a maximum at x = −3.

3. x = −1. Let y = x + 3. Then we need to consider the value y = 2.

f (y) = y 3 (y − 5)2 (y 2 + (2b − 6)y + 9 − 6b + c)

= y 3 (y 2 − 10y + 25)(y 2 + (1 − 5c)y − 12 + 16c)

= y 7 + (1 − 5c − 10)y 6 + (−12 + 16c + 25 − 10 + 50c)y 5

+(120 − 160c + 25 − 125c)y 4 + (400c − 300)y 3

= y 7 − (9 + 5c)y 6 + (3 + 66c)y 5 + (145 − 285c)y 4 + (400c − 300)y 3

f 00 (y) = 42y 5 − 30(9 + 5c)y 4 + 20(3 + 66c)y 3 + 12(145 − 285c)y 2 + 6(400c − 300)y

f 00 (2) = 1344 − 480(9 + 5c) + 160(3 + 66c) + 48(145 − 285c) + 12(400c − 300)

= 864 − 720c

3

So if c > 65 , then f 00 (−1) < 0 and f (x) has a maximum at x = −1. If c < 56 , f (x) has

a minimum at x = −1. If c = 56 , then this test fails, we consider higher derivatives at

y = 2 with c = 65 :

when c = 65 .

Question 2(c) Show that f (x) has at least three points of inflection irrespective of the

condition 2b + 5c = 7.

f (2) = 0, there exists a real number b between −3 and 2 such that f 0 (b) = 0. Now f 0 (2) =

f 0 (b) = f 0 (−3) = 0, thus we get a1 ∈ (−3, b), a2 ∈ (b, 2) where f 00 (a1 ) = 0, f 00 (a2 ) = 0. Thus

there are least 3 points where f 00 (x) = 0.

Note: This question is not complete: A point of inflection is a point where the curvature

of a curve changes sign (the curve goes from convex to concave or vice versa). Any point a

is a point of inflection if f 00 (a) = 0 and the lowest order non-zero derivative at a is of odd

order, or equivalently, f 00 (a + ) and f 00 (a − ) have opposite signs in the neighborhood of a.

Thus f 00 (x) = 0 is a necessary condition for a point to be an inflection point, but is not

sufficient. f 00 (x) is a 5th degree curve. Take two of the roots and equate them, and two other

roots and equate them too. We can thus get two equations in b, c, which can be solved to get

their values. Now f 00 (x) = A(x − B)(x − C)2 (x − D)2 , where B, C, D are distinct. This curve

has only one point of inflection, B. For example, the similar curve g(x) = 35x3 − 21x5 + 5x7

has only one inflection point, although g 00 (x) has 3 roots.

n

X n (n−i)

f (x)g (i) (x)

i=0

i

n

where f (m) denotes the m-th derivative of f (x) and i

are the binomial coefficients.

1

f (1) (x)g (0) (x) + 11 f (0) (x)g (1) (x) where f (0) (x) = f (x).

0

Induction hypothesis: Assume that the result is true for n = m.

m

(m)

X m (m−i)

(f g) = f (x)g (i) (x)

i=0

i

4

Differentiating both sides,

m

(m+1)

X m

(f g) (x) = f (m−i+1) (x)g (i) (x) + f (m−i) (x)g (i+1) (x)

i=0

i

m h

m (m+1) (0)

X m m i (m−i+1)

= f (x)g (x) + + f (x)g (i) (x)

0 i=1

i i − 1

m (0)

+ f (x)g (m+1) (x)

m

m

m + 1 (m+1) (0)

X m + 1 (m−i+1)

= f (x)g (x) + f (x)g (i) (x)

0 i=1

i

m + 1 (0)

+ f (x)g (m+1) (x)

m+1

m+1

X m + 1

= f (m+1−i) (x)g (i) (x)

i=0

i

We used m0 = 1 = m+1

m m+1

m m

m+1

0

, m

= 1 = m+1

, i

+ i−1

= i

. The result is now

established for n = m + 1, and hence by induction for all n.

Question 3(b) Show that the m-th derivative gm (x) of g(x) = tan−1 x satisfies

(−1)m−1 (m − 1)!φm (x)

gm (x) =

(1 + x2 )m

where φm (x) is a polynomial of dergree m given by φ1 (x) = 1, φ2 (x) = 2x and the recursion

φm+1 (x) = 2xφm (x) − (1 + x2 )φm−1 (x).

1

Solution. g1 (x) = or (1 + x2 )g1 (x) = 1.

1 + x2

Using Leibnitz’s theorem proved above and differentiating the above equation m times,

we get

m 2 m m

(1 + x )gm+1 (x) + 2xgm (x) + 2gm−1 (x) = 0

0 1 2

or

(1 + x2 )gm+1 (x) + 2mxgm (x) + m(m − 1)gm−1 (x) = 0

which is the required relation.

1 (−1)1−1 (1 − 1)!φ1 (x)

Now g1 (x) = = as φ1 (x) = 1.

1 + x2 (1 + x2 )1

5

−2x (−1)2−1 (2 − 1)!φ2 (x)

g2 (x) = = as φ2 (x) = 2x.

(1 + x2 )2 (1 + x2 )2

Assume that the result is true for all n ≤ m. From the relation proved above:

(1 + x2 )gm+1 (x) + 2mx + m(m − 1) =0

(1 + x2 )m (1 + x2 )m−1

or

(−1)m (m − 1)!φm (x) (−1)m−1 (m − 2)!φm−1 (x)

gm+1 (x) = 2mx + m(m − 1)

(1 + x2 )m+1 (1 + x2 )m

m

(−1) m! h i

= 2 m+1

2xφm (x) − (1 + x2 )φm−1 (x)

(1 + x )

(−1)m m!

= φm+1 (x)

(1 + x2 )m+1

as φm+1 (x) = 2xφm (x) − (1 + x2 )φm−1 (x). This completes the proof.

Question 3(c) Find φm (x) for x ≤ 6. Can you find a general formula?

Solution.

φ1 (x) = 1

φ2 (x) = 2x

φ3 (x) = 2x · 2x − (1 + x2 ) = 3x2 − 1

φ4 (x) = 2x(3x2 − 1) − (1 + x2 )(2x) = 4x3 − 4x

φ5 (x) = 2x(4x3 − 4x) − (1 + x2 )(3x2 − 1) = 5x4 − 10x2 + 1

φ6 (x) = 2x(5x4 − 10x2 + 1) − (1 + x2 )(4x3 − 4x)

= 6x5 − 20x3 + 6x

1 0

φ1 (x) = x

0

2 1

φ2 (x) = x

1

3 2 3 0

φ3 (x) = x − x

2 0

4 3 4 1

φ4 (x) = x − x

3 1

6

5 4 5 2 5 0

φ5 (x) = x − x + x

4 2 0

6 5 6 3 6 1

φ6 (x) = x − x + x

5 3 1

n n−1 n n−3 n

φn (x) = x − x + xn−5 − . . .

n−1 n−3 n−5

2r+1≤n 2r+1≤n

X

r n n−2r−1

X

r n

= (−1) x = (−1) xn−2r−1

r=0

n − 2r − 1 r=0

2r + 1

n n n

n − 2! + 4! − 6! + ...

3 5 7

n n

dn x −1

X n dn−r ex X n

(e tan x) = gr (x) n−r = gr (x)ex

dxn r=0

r dx r=0

r

Thus n n

dn x

−1

X n X n

(e tan x) = g r (0) = (−1)r−1 (r − 1)!φr (0)

dxn x=0 r=0

r r=1

r

as g(0) = 0 and gr (x) was computed above in terms of φr (x). Now φ2r (0) = 0, and φ2r+1 =

(−1)r . Thus

2r+1≤n

dn x

X n

−1

(e tan x) = (2r)!(−1)r

dxn x=0 r=0

2r + 1

or

dn x

n n n n

(e tan−1 x) = 0! − 2! + 4! − 6! + ...

dxn x=0 1 3 5 7

as required.

Paper II

Z π

2

Question 4(a) Discuss the convergence of the integral log(sin x) dx and if convergent,

0

evaluate it.

7

Solution.π The integrand log sin x has a discontinuity only at x = 0. We consider the

Z

2

integral − log(sin x) dx so that the integrand is positive.

0

Let g(x) = x−m , 0 < m < 1, then

lim = lim+ = lim =0

x→0+ x−m x→0 sin x −mx−m−1 x→0+ m sin x

as L’Hospital’s rule applies.

Z π Z π

2 2

But g(x) dx is convergent as 0 < m < 1, therefore − log(sin x) dx converges. Now

0 0

Z π Z π Z π Z π

2 2 2 2

log(sin 2x) dx = log 2 dx + log(cos x) dx + log(sin x) dx

0 0 0 0

Z π Z π

2 2

But log(cos x) dx = log(sin x) dx, so

0 0

Z π Z π

2 π 2

2 log(sin x) dx = − log 2 + log(sin 2x) dx

0 2 0

By putting 2x = t. π

Z Z π

2 1

log(sin 2x) dx = log(sin t) dt

0 2 0

Now substituting t = π − y, we have

π

Z π Z 0 Z

2

log(sin t) dt = log(sin y) (−dy) = log(sin y) dy

π π

2 2

0

Z π Z π

2 2

∴ log(sin 2x) dx = log(sin y) dy

0 0

Z π

2 π

⇒ log(sin x) dx = − log 2

0 2

2

Question 4(b) Find the point on the parabola

√ y = 2x, z = 0 which is nearest to the plane

z = x + 2y + 8. Show that this distance is 6.

Solution. Let the point (x, y, 0) by on the parabola. Then the distance d from the plane

z = x + 2y + 8 is given by

|x + 2y − 0 + 8| |x + 2y + 8|

d= √ = √

1+4+1 6

8

⇒ 6d2 = x2 + 4xy + 4y 2 + 16x + 32y + 64.

Put 2x = y 2 , and let F = 6d2 to get

y4 2

F = 6d = + 2y 3 + 4y 2 + 8y 2 + 32y + 64

4

We need to minimize F . The critical points are given by

dF

= y 3 + 6y 2 + 24y + 32 = 0

dy

⇒ (y + 2)(y 2 + 4y + 16) = 0

d2 F

= 3y 2 + 12y + 24 = 12 − 24 + 24 > 0

dy 2

at y = −2. Hence y = −2 is a minimum. √

When y = −2, x = 2, and d = |2−4+8|

√

6

= 6 as required.

ZZZ

1

(a2 b2 c2 − b2 c2 x2 − c2 a2 y 2 − a2 b2 z 2 ) 2 dx dy dz

x2 y 2 z 2 π 2 a2 b2 c2

over the volume bounded by 2 + 2 + 2 = 1 is equal to .

a b c 4

Solution. Let

ZZZ

1

I = 2

(a2 b2 c2 − b2 c2 x2 − c2 a2 y 2 − a2 b2 z 2 ) 2 dx dy dz

x2 2

{(x,y,z) | + y2 + z2 ≤1}

a2 b c

x2 y 2 z 2 21

ZZZ

= 8abc 1 − 2 − 2 − 2 dx dy dz

{(x,y,z) | x2 2 2

+ y2 + z2 ≤1,x≥0,y≥0,z≥0} a b c

a2 b c

ZZZ

2 2 2 2 2 2 12

I = 8a b c 2 2 2

(1 − X − Y − Z ) dX dY dZ

X +Y +Z ≤1

X≥0,Y ≥0,Z≥0

du dv dw

Let X 2 = u, Y 2 = v, Z 2 = w, so that dX = √ , dY

2 u

= √ , dZ

2 v

= √

2 w

and

ZZZ

1 1 1 1

2 2 2

I=a b c (1 − u − v − w) 2 u− 2 v − 2 w− 2 du dv dw

u+v+w≤1

u≥0,v≥0,w≥0

9

We now convert this to a Dirichlet integral — let u + v + w = α, v + w = αβ, w = αβγ,

∂(u, v, w)

so that = −α2 β and

∂(α, β, γ)

Z 1Z 1Z 1

1 1 1 1

I = abc 2 2 2

(1 − α) 2 (α − αβ)− 2 (αβ − αβγ)− 2 (αβγ)− 2 α2 β dα dβ dγ

Z0 1 Z0 1 Z0 1

1 1 1 1 1

= a2 b2 c2 (1 − α) 2 α 2 (1 − β)− 2 (1 − γ)− 2 γ − 2 dα dβ dγ

0 0 0

3 3

2 2 2 Γ( 2 )Γ( 2 ) Γ( 21 )Γ(1) Γ( 21 )Γ( 21 )

= abc

Γ(3) Γ( 32 ) Γ(1)

1 1 4

2 2 2 2 Γ( 2 ) π 2 a2 b2 c2

= abc =

2 4

√

as required, as Γ( 12 ) = π.

10

UPSC Civil Services Main 1991 - Mathematics

Calculus

Sunder Lal

Retired Professor of Mathematics

Panjab University

Chandigarh

1. The curve is symmetrical about both the coordinate axes. Enough to consider when

x ≥ 0, y ≥ 0.

2. (0, 0) lies on the curve.

3. The coefficient of the highest power of x is y 2 − b2 . Thus y = ±b are two asymptotes

parallel to the x-axis. The coefficient of the highest power of y is x2 − a2 , thus x = ±a

are two asymptotes parallel to the y-axis. For other asymptotes of the form y =

mx + c, we know that limx→∞ xy = m. We divide by x4 and let x → ∞ to obtain

a2 y 2 b2 a2 b2

lim 1− 2 − − 4 = m2 = 0 so the only asymptotes of the form y = mx+c

x→∞ x x2 x2 x

have m = 0, and we have already found those. Thus the curve has only 4 asymptotes.

a2 b 2 a2 b2 b2 x 2 bx

4. y 2 − b2 = 2 2

⇒ y 2

= 2 2

+ b 2

= 2 2

⇒ y = ±√ . Since we need

x −a x −a x −a x 2 − a2

bx

to trace only in the first quadrant, we let y = √ .

x 2 − a2

dy b bx2 −a2 b

5. = √ − 3 = 3 , showing that the curve has no critical

dx x 2 − a2 (x2 − a2 ) 2 (x2 − a2 ) 2

points.

6. If x2 − a2 < 0 , then (x2 − a2 )y 2 = b2 x2 ⇒ x = 0, y = 0. Hence x2 − a2 > 0, so the

curve lies beyond the line x = ±a. Similarly it lies beyond the lines y = ±b. (0, 0) is

an isolated point.

1

d2 y 5

7. 2

= 32 a2 b(x2 − a2 )− 2 2x > 0 for x > a, so the curve is downwards convex.

dx

A sketch is shown here:

x = −a x=a

y=b

y = −b

Question 1(b) Find the cubic curve which has the same asymptotes as the curve

x3 − 6x2 y + 11xy 2 − 6y 3 + x + y + 1 = 0

and which passes through the points (0, 0), (1, 0) and (0, 1).

same terms of degree 3 and 2 as the given curve, it has the same asymptotes. Now since

it must pass through (0, 0), c = 0. Since it passes through (1, 0), 1 + a = 0 ⇒ a = −1,

and since it passes through (0, 1), −6 + b = 0 ⇒ b = 6. Hence the required curve is

x3 − 6x2 y + 11xy 2 − 6y 3 − x + 6y = 0.

Question 1(c) Show that the function f (x, y) = y 2 + x2 y + x4 has (0, 0) as the only critical

point and f (x, y) has a minimum at that point.

∂f ∂f

Solution. The critical points are given by = 2xy + 4x3 = 0 ⇒ x = 0, y = −2x2 , =

∂x ∂y

x2

2y + x2 = 0 ⇒ y = − .

2 2

When x = 0, y = − x2 ⇒ y = 0.

2

y = −2x2 , y = − x2 ⇒ x = 0, y = 0. Hence (0, 0) is the only critical point.

2

∂ 2f 2 ∂ 2f ∂ 2f

Now = 2y + 12x = 0 at (0, 0). = 2 and = 2x = 0 at (0, 0). Thus

∂x2 ∂y 2 ∂x ∂y

∂ 2 f ∂ 2 f ∂ 2 f 2

− = 0 at (0, 0)

∂x2 ∂y 2 ∂x ∂y

so we cannot say whether it is maximum or minimum. However f (x, y) = (x2 + y2 )2 + 34 y 2 ≥ 0

for all (x, y) ∈ R2 , so f (x, y) has a minimum at (0, 0).

Question 1(d) Find the percentage error in the volume of a right circular cone when an

error of 1% is made in measuring the height and an error of 0.5% is made in measuring the

base radius.

Solution.

π 2

V = r h

3

π

log V = log + 2 log r + log h

3

dV dr dh

= 2 +

V r h

dr dh dV

Now = 0.5% = 0.005, = 1% = 0.01. Thus = 0.02 = 2%. Hence the error in

r h V

measuring the volume is 2%.

ZZ

Question 2(a) Evaluate F (x + y)xm−1 y n−1 dx dy with F (u) = (1 − u)l−1 where D is

D

the interior of the triangle formed by x = 0, y = 0, x + y = 1 and l, m, n are all positive.

∂(x, y) v u

Put u = x + y, uv = x ⇒ y = u − uv, so that = = −u.

∂(u, v) 1 − v −u

Z 1Z 1

I = F (u)(uv)m−1 un−1 (1 − v)n−1 u du dv

Z0 1 Z0 1

= F (u)um+n−1 v m−1 (1 − v)n−1 du dv

0 0

Γ(m)Γ(n) 1

Z

= F (u)um+n−1 du

Γ(m + n) 0

Γ(m)Γ(n) 1

Z

= (1 − u)l−1 um+n−1 du

Γ(m + n) 0

Γ(m)Γ(n) Γ(l)Γ(m + n) Γ(m)Γ(n)Γ(l)

= =

Γ(m + n) Γ(m + n + l) Γ(m + n + l)

3

x2 y 2 z 2

Question 2(b) Find the center of gravity of the solid bounded by the ellipsoid 2 + 2 + 2 =

a b c

1 and lying in the positive octant.

ZZZ

Solution. Let S be the positive octant of the ellipsoid. Let M = dx dy dz.

ZZZ ZZZ Z Z ZS

1 1 1

Then x = x dx dy dz, y = y dx dy dz, z = z dx dy dz.

M S M S M S

Now put x = aX, y = bY, z = cZ so that

ZZZ

π

M = abc dX dY dZ = abc

X≥0,Y ≥0,Z≥0,

2 2 2

6

X +Y +Z ≤1

3

.

6a2 bc

ZZZ

x= X dX dY dZ

πabc X≥0,Y ≥0,Z≥0,

2 2 2

X +Y +Z ≤1

Put X 2 = u, Y 2 = v, Z 2 = w. Then

ZZZ

6a 1 1

−1 1 −1 1 −1

x = u≥0,v≥0,w≥0, u 2 u 2 v 2 w 2 du dv dw

π D: 2 2 2

u+v+w≤1

ZZZ ZZZ

3a 1 1 3a 1 1

= u0 v − 2 w− 2 du dv dw = u1−1 v 2 −1 w 2 −1 du dv dw

4π D 4π D

1 1

3a Γ(1)Γ(