PROBLEM SETS

SPRING 2023: LINEAR ALGEBRA (MAT-1001)

GAURAV BHATNAGAR

Problem solving in Linear Algebra

To become good in Linear Algebra, you have to do many problems.
(1) For the quizzes, mark corrections or redo the problems as required.
(2) Do all the problems from the book, as and when we complete the relevant
chapters.
(3) In addition, try the problems I give in class. (I have included them here.)
(4) Problems marked RB are provided by Professor Bhatia.
(5) These problems are to be tried on your own for as long as you possibly can,
and then discussed with friends. Learning mathematics is an activity best
done as a team. Hearing the words repeatedly, even if in your own voice,
helps in absorbing the material.
(6) After doing the problem, spend some time in writing your solution neatly
in a manner that your argument is clear. If you see your solution after a
year, will you understand what you have done? Even if you find out the
solution from someone else, write on your own.
(7) You may show your solutions to the TAs and TFs and ask them informally
whether it is correct.
(8) In DS sessions, plan to present a problem if called upon (groups can present
too).
(9) Some of these problems will show up in the midterms/final. It may be
impossible to do them there if you have not tried them earlier.

1. Problem Set 1
Problem 1 (Herstein p. 8).
(a.) If A is a subset of B and B is a subset of C, prove that A is a subset of C.
(b.) If B ⊂ A, prove that A ∪ B = A. Prove the converse. (The converse of “If P
then Q’ is ‘If Q then P .)
(c.) If B ⊂ A, prove that for any set C both B ∪ C ⊂ A ∪ C and B ∩ C ⊂ A ∩ C.
Remark 1. Herstein refers to Topics in Algebra by I. N. Herstein,
Problem 2 (Herstein p. 8). Prove: A ∪ (B ∩ C) = (A ∪ B) ∩ (A ∪ C).
1
2 G. BHATNAGAR

Problem 3 (Herstein p. 8). For a subset A of the universal set S, we denote by

Ac the complement of A in S (that is, Ac = S − A.) For any two subsets A and B
of S, prove the following known as de Morgan Laws.
(a.) (A ∩ B)c = Ac ∪ B c .
(b.) (A ∪ B)c = Ac ∩ B c .
Problem 4. This problem contains more details of a proof we did in class.
(a.) Prove that for any polynomial p, and any complex number α, there is a
polynomial q such that p(x) = (x − α)q(x) + r, where r is a complex number.
(Hint: Consider
p(x) = an xn + · · · + a0 = (x − a)(bn−1 xn−1 + · · · + b0 ) + r
and show you can calculate bk and r in terms of the given coefficients ak and
a.)
(b.) Prove that if p(α) = 0 then p(x) = (x − α)q(x) for some polynomial q(x).
(That is, r = 0.)
(c.) Prove the converse of the statement above.
(d.) Argue that if p is a polynomial of degree n, then q is of degree n − 1.
After trying out this problem yourself, you should read the relevant theorem in
Chapter 3 of the book.
Problem 5. The following are vector spaces V over the given field F. State the
relevant definitions of addition of vectors + and scalar multiplication · in each of
these. Verify the axioms of vector spaces hold for all of these (do at least five, and
then go on as long as you don’t fall asleep due to boredom).
(a.) V = C, F = R.
(b.) V = R, F = R.
(c.) V = Pn (z), where Pn (z) is the set of polynomials with coefficients in C. F = C.
(d.) V = Q[x], the set of polynomials with coefficients in Q. F = Q. If the field
F = R, is it still a vector space?
(e.) V = Fn over a field F.
(f.) A sequence a is a function a : N ∪ {0} → C. We represent a(i) by ai , so denote
the sequence as
∞
a = ai i=0 .
V is the set of all sequences in C. F = C.
(g.) V = C[[x]], the set of formal power series with coefficients in C. F = C.
(h.) V = C 1 [0, 1], the set of all functions on [0, 1] that are continuous, have a
derivative on (0, 1) and the derivative is continuous. (You may use any prop-
erties of continuous functions and derivatives but define all the terms just to
remind yourself what they mean, and state the theorems used as is from some
calculus book.)
Definition 1 (Binary operation). Let A be a set. A binary operation ∗ on A is a
function A × A → A.
 PROBLEM SETS SPRING 2023: LINEAR ALGEBRA (MAT-1001) 3

Problem 6 (RB). Let ∗ be the binary operation on N = {1, 2, 3, . . . , } defined as

a ∗ b = ab .

(a) Is this operation commutative?

(b) Is it associative?
Problem 7 (RB). Let V be a vector space over a field F; x ∈ V and a ∈ F. Show
that if ax = 0, then a = 0 or x = 0. (To show a statement of the form P or Q,
one way to organize your argument is to assume P is false, and show Q is true.)
Problem 8. Prove the following statements which follow immediately from the
definition of the vector space. After trying them yourself, look over the proofs in
the book and compare with your proofs. In the following V is a vector space over
the field F.
(a.) Prove that in a vector space, the additive identity is unique.
(b.) Prove that the additive inverse of v ∈ V is unique.
(c.) Prove that: 0 · v = 0 for all v ∈ V . Here the 0 on the left is in F and the one
on the right is in the vector space V . For example, in R4 : 0 · (1, 2, 3, −1) =
(0, 0, 0, 0).
(d.) Prove that a · 0 = 0 for all a ∈ F.
(e.) Prove that (−1) · v = −v for all v ∈ V . Here the −1 is the additive inverse of
F (satisfying the famous field axioms) and −v the additive inverse of v in the
vector space V .
(f.) (Bonus) When you next go home, explain to that irritating kid in your family
why
(negative) × (negative) = (positive).
Problem 9. The following problems are about arithmetic on Zn . This consists of
the set
Zn := {0, 1, 2, . . . , n − 1}.
Both + and multiplication × can be defined on this set. The only proviso is that
if you get an answer bigger or equal to n, then divide the number by n and take
the remainder. The remainder is always one of 0, 1, 2, . . . , n − 1. We say that we
are doing arithmetic “modulo n” or “mod n”.
(a) Write down multiplication tables for + and × mod n for n = 2, 3, 4, 5.
(b) Using the multiplication table, verify that F = Zn is a field for n = 2, 3, 5.
Show it is not a field for n = 4.
(c) (Optional). Prove that Zp is a field if p is a prime. You may need to use the
following fact: If p | ab, then p | a or p | b. Here p | a means p divides a.
(d) Let F be Z3 and let Fn [x] denote the set of polynomials of degree less than or
equal to n with coefficients in F. Use Sage to generate the elements of Fn [x],
for n = 1, 2, 3, . . . .. How many elements are there in Fn [x]?
4 G. BHATNAGAR

2. Problem Set 2
Exercise 1 (Really exercise). Go for a long walk, alone, lasting at-least one hour,
and mentally go over all the definitions and theorems done in linear algebra so
far. You may make and carry a crib sheet; you may talk to yourself during this
time—but otherwise stay silent. No music on headphones allowed and if possible,
leave your phone behind. This needs to be repeated at least once every week.
Problem 10. Let S = {u1 , u2 , . . . , um } be a set of vectors in a Vector Space V
over F. Show that X m 
< S >:= ai ui : ai ∈ F
i=1
is a subspace of V . (This is what we defined to be < S >, the linear span of S.
We have done this, but you should do it on your own once.)
Problem 11. Let U1 and U2 be subspaces of a vector space V over a field F.
(a.) Show that U1 ∩ U2 is a subspace of V .
(b.) Show that U1 ∪ U2 may not be a subspace.
Problem 12. Let S be a set of vectors in a vector space V and let W be defined
as the intersection of all subspaces of V which contain S. Show that < S >= W ;
that is (in symbols):
\
< S >= U.
S⊂U
U subspace of V

Problem 13 (RB). Let u and v be the vectors in R2 given by

" # " #
1 3
u= ,v = .
2 1
(a) Express the vectors
" # " # " # " #
1 0 2 −1
, , ,
0 1 4 1
as linear combinations of u and v.
(b) Is there any vector in R2 that cannot be expressed as a linear combination of
u and v?
Problem 14 (RB). Are the vectors
       
1 1 1 1
1 −1  1  −1
  ,   ,   , and  
       
1  1  −1 −1
1 −1 −1 1
in R4 linearly independent.
 PROBLEM SETS SPRING 2023: LINEAR ALGEBRA (MAT-1001) 5

Problem 15 (RB). Suppose u, v and w are linearly independent vectors in a

vector space V .
(a) Is it necessary that u + v, v + w, and w + u are linearly independent?
(b) Is it possible that the span of u and v is equal to the span of v and w?
Problem 16 (RB). Do the following by hand, and then check your work using
Sage.
(a) Solve the system of equations
x+ y+ z =1
x + 2y + 2z = 0
x + 2y + 3z = 0.
(b) Replace the three numbers (1, 0, 0) on the right-hand side above by (0, 0, 1)
and solve the system.
(c) Do the same with the numbers (0, 0, 0) instead.
(d) In Sage, reduce the relevant matrices to row-reduced echelon form and then
solve.

3. Problem Set 3
Note. For problems involving matrices, solve the problems by hand, and also
using Sage.
Problem 17. For all the vector spaces mentioned in Problem 5, Problem Set 1,
find an elementary basis (if you can).
Problem 18. For each of the following augmented matrices, write down the cor-
responding system of equations and solve them (if possible). Use the symbols x,
y, z, u as required.
(a)
 
1 0 0 0 1
0 2 0 0 2 
M1 = 
 

0 0 3 0 3 
0 0 0 −1 −4
(b)
 
1 0 0 0 1
0 2 0 0 2 
M2 = 
 

0 0 3 0 3 
0 0 0 0 −1
Problem 19.
6 G. BHATNAGAR

(a) Solve the system of equations represented by this augmented matrix using back
substitution. (Solve the last equation, then use solution in the previous one,
and so on.) Use x, y and z for variable names.
 
1 0 2 0
M3 = 0 −2 3 2 
 
0 0 1 3
(b) Use row operations to reduce the matrix M3 to the following form, and then
solve the corresponding systems of equations.
 
1 0 0 ∗
M3 = 0 −2 0 ∗ 
 
0 0 1 3
Here ∗ are ‘wildcards’. They stand for ‘some numbers’.
Problem 20. The following questions are all related.
(a) Solve the following system of equations.
x+y =2
y−z =3
x + z = −1
(b) Write the vector v (below) as a linear combination of vectors in the set S :=
{v1 , v2 , v3 }. (The underlying field is R.) Here
 
2
v =  3 ;
 
−1
     
1 1 0
v1 = 0 , v2 = 1 , v3 = −1 .
     
1 0 1
(c) Find a vector u which is not a linear combination of vectors in S.
(d) For v2 and v3 above, show that if av2 + bv3 = 0, then a = 0 and b = 0.
(e) For v1 , v2 and v3 above, find a1 , a2 , a3 , not all 0, such that
a1 v1 + a2 v2 + a3 v3 = 0.
Problem 21. Let V be a finite dimensional vector space, with dim V = n. Then
prove the following.
(a.) If U ⊂ V is a subspace of V , then dim U < dimV .
(b.) Let S = {u1 , u2 , . . . , un }. If < S >= V , then S is a basis (that is, S is linearly
independent).
 PROBLEM SETS SPRING 2023: LINEAR ALGEBRA (MAT-1001) 7

(c.) Let S = {u1 , u2 , . . . , un }. If S is linearly independent, then S is a basis (that

is, < S >= V ).
Problem 22. Here we use the notation < S > for the linear span of S. Let S and
T be subsets of a vector space V . Then prove the following.
(a) S ⊂ T implies < S >⊂< T >.
(b) < S ∪ T >=< S > + < T >.
(c) << S >>=< S >. (Here, on the left, its the subspace generated by < S >.)
Problem 23. Let V be a vector space of dimension n.
(a.) Show that any n+1 elements of the vector space have to be linearly dependent.
(b.) No set of n − 1 elements of the vector space can span V .
Problem 24. Let S = {u1 , u2 , . . . , um } be a set of linearly independent vectors in
a vector space V over a field F. Let
T = {w1 , w2 , . . . , wn }
such that ui ∈< T >, for i = 1, 2, . . . , m.
(a.) (Steinitz Exchange Lemma) Show that there exists a j (1 ≤ j ≤ n) such that
S ∪ {wj } \ {u1 } = {wj , u2 , u3 , . . . , um }
is linearly independent.
(b.) (Another way to state the first miracle of Linear Algebra) Show that
m≤n
.
(c.) What does this imply when < S >= V ?

4. Problem Set 4
Problem 25.
(a.) Do any 10 problems from any linear algebra book solving systems of linear
equations using Sage. Use matrix operations to do them, and also find out
commands to solve the equations directly. For each of the solutions, if there
are infinite number of solutions, find the dimension of the solution space.
(b.) For each matrix A considered above, consider the transformation T : Rm →
Rn defined by the matrix, i.e., T : X 7→ AX. For each matrix A, identify m
and n. Find the dimension of the null space of the matrix (i.e., the dimension
of the null space of T and the dimension of the range space of T . (The
dimension of the range space is called the column rank of the matrix.)
Problem 26. Let T : V → W be a linear transformation, where V is a finite
dimensional vector space.
(a.) Show that if dim V > dim W , then T is not injective.
(b.) If dim V < dim W , then T is not surjective.
(c.) Write the contrapositive of (a).
8 G. BHATNAGAR

(d.) Write the contrapositive of (b).

(e.) Argue that if T is bijective, then dim V = dim W .
Problem 27. Consider the map T : R2 → R2 defined as T (x, y) = (x − y, y).
(a.) Show that it is well defined.
(b.) Show that it is linear.
(c.) Show that it is onto.
(d.) Find the rank and nullity of T .
(e.) Find m(T ) the matrix of T (in the elementary basis).
(f.) Find the matrix of T where the domain uses the basis S = {(1, 1), (1, −1)}
and the range uses the basis {(1, 0), (0, 1)}. (First verify pictorially that S is
indeed a basis of R2 .
(g.) Find the matrix where the elementary basis E = {e1 , e2 } is mapped to S =
{w1 , w2 }, where w1 = e1 − e2 and w2 = e2 . (First verify pictorially that S is
indeed a basis of R2 .)
Problem 28. Consider the map T : R3 → R3 defined as
T (x, y, x) = (2x − 3y + z, x + 2y − z, x − 3y + z + 2).
Is T a linear map? Explain your answer geometrically.
Problem 29. Consider the set
N = {(x1 , x2 , x3 , x4 , x5 ) ∈ R5 : x1 = 5x2 , x3 = x4 = x5 }.
(a.) What is the dimension of N ?
(b.) Let T : R5 → R2 . Can T have N as a null space?
(c.) Define a T : R5 → R3 such that null(T ) = N . (Use a matrix.)
Problem 30. Suppose there is a linear transformation T : V → V such that
both the rank T and nullity T are finite; that is null T and range T are both finite
dimensional subspaces of V . Prove that V must be finite dimensional.
Problem 31. Give an example of a function T : R2 → R such that T (au) = aT u
for all a ∈ R and u ∈ R2 , but T is not linear.
Problem 32. Let T : V → F be a linear transformation. Show that if u ∈ V is
not in the null space of T (i.e. T u ̸= 0), then
V = null T ⊕ < {u} > .

5. Problem Set 5
Remarks. Some parts in these questions require definition of matrix multiplication
which we will do this week. You should attempt these parts on Sage as well as by
hand.
Problem 33. Suppose I = {u1 , u2 , . . . , um }} is a linearly independent set in a
vector space U , and T : U → V is injective (or one-one). Show that
{T u1 , T u2 , . . . , T um }
 PROBLEM SETS SPRING 2023: LINEAR ALGEBRA (MAT-1001) 9

is linearly independent.
Problem 34. Let U , V and W be vector spaces and S : U → V and T : V → W
be linear transformations. Show the following:
Nullity(T S) ≤ Nullity T + Nullity S.
Consider matrices corresponding to the transformations S and T w.r.t the standard
basis for all of U , V and W . Interpret this result in terms of the matrices obtained.
Problem 35 (RB). Given a vector x = (x1 , . . . , xn ), let
x1 + · · · + xn
x̄ :=
n
n n
be the “mean” of x. Let A : R → R be the map
Ax = (x̄, x̄, . . . , x̄).
Is this a linear map? Is so, what is the matrix corresponding to A (in the standard
basis). What is the null space of A. (The null space of A is the space of solutions
of Ax = 0.)
Problem 36. Let V be a vector space and §V = {u1 , u2 , . . . , un } be a basis of V .
Let Ui =< {ui }, for i = 1, 2, . . . , n. Then show that
V = U1 ⊕ U2 ⊕ · · · ⊕ Un .
Give an explcit isomorphism and show its an isomorphism.
Problem 37. Let T1 and T2 be linear transformations, and a ∈ F. Let some
basis be fixed, Let m(T ) denote the matrix of T in wrt that basis for all linear
transformations T in this problem. Prove that:
(i.) m(T1 + T2 ) = m(T1 ) + m(T2 ), and;
(ii.) m(aT1 ) = am(T1 ).
Given that every matrix is the matrix of a linear transformation, this implies that
m × n matrices with entries in F form a vector place over F.
Problem 38. A square matrix A is called nilpotent if Ak = 0 (the zero matrix)
for some k > 0.
(a) (RB) Make an n × n matrix A such that A, A2 , . . . , An−1 are not zero, but
An = 0.
(b) Prove that
1
= 1 + x + x2 + · · · ,
1−x
for suitable values of x.
(c) Show that if A is nilpotent, then I + A is invertible.
Problem 39 (RB). Let e1 , e2 , . . . , e5 be the standard basis in R5 and let S :
R5 → R5 be the linear map defined by
S(e1 ) = e2 , S(e2 ) = e3 , S(e3 ) = e4 , S(e4 ) = e5 , S(e5 ) = e1 .
10 G. BHATNAGAR

(a) Write the matrix of S in the standard basis.

(b) Calculate the powers S 2 , S 3 , S 4 , . . . .
Problem 40 (RB). Let e1 , e2 , . . . , e5 be the standard basis in R5 and let T :
R5 → R5 be the operator defined by
T (e1 ) = e2 , T (e2 ) = e3 , T (e3 ) = e4 , T (e4 ) = e5 , T (e5 ) = 0.
(a) Write the matrix of T in the standard basis.
(b) Calculate the powers T 2 , T 3 , T 4 , . . . .
(c) Construct an operator A on R5 such that A2 = A and rank(A) = 3.
(d) Calculate AT and T A.

6. Problem Set 6
Problem 41. Let v ∈ V , U =< v >. Define S : U → V such that m(S) = m(v)
and show it is well-defined and a linear transformation. Using this and the theorem
in Lecture 11, show that for T : V → W a linear transformation, and v ∈ V , we
must have
M (T v) = M (T )M (v).
Problem 42. Find two distinct examples of matrices that commute and one which
doesn’t.
Problem 43. All the following problems are about a Hilbert Hotel. In each case
you have to give a bijection. These problems are taken from my QRMT class.
(a.) The Hilbert Hotel has an infinite number of rooms and they are all full. A
new guest comes. But the hotel manager manages to find a room for the new
guest. How?
(b.) Again, the Hilbert Hotel is full and a long bus, carrying a 1000 guests arrives.
The tour leader sees the sign that the hotel is full, and asks the manager if he
can suggest another nearby hotel.
The manager says: “No problem. We have space for 1000 more guests”
How does the manager accommodate 1000 more guests?
(c.) The Hilbert Hotel is full, and this time a really very, very long bus comes.
The bus contains an infinite number of guests. Say the guests are numbered
1, 2, 3, . . . .
Again, the manager says: “No problem.” How does the manager find rooms
for the new guests this time?
(d.) For a change, the Hilbert Hotel is empty. All the guests have left, and the
manager is looking forward to some free time. As you can imagine, looking
after an infinite number of guests can get tiring. So the manager heaves a
sigh of relief as a bus containing all his guests leaves for the airport!
But the relief is short-lived. An infinite number of buses (numbered 1, 2, 3,
. . . ) come in and each of them contain an infinite number of guests (named
1, 2, 3, . . . ).
 PROBLEM SETS SPRING 2023: LINEAR ALGEBRA (MAT-1001) 11

How does the manager accommodate all of the guests?

Problem 44. Let T : V → W be an invertible linear transformation. Show that
T −1 is also a linear transformation. You can prove this directly
Problem 45. Let T : U → V be an invertible linear transformation. Then its
inverse is unique; that is, if R and S are two inverses, then R = S.
Problem 46. Let A and B be n × n square matrices such that AB = In , where In
denotes the n × n identity matrix. Show that BA = In too. (Prove this directly by
using row reduction. Also use properties of linear transformations to give another
proof.)

7. Problem Set 7: Matrix Problems

These problems are taken from “An introduction to Linear Algebra”, by Krishna-
murthy, Mainra and Arora. These are to be done using Sage. All these problems
can be directly solved in Sage and also solved after row reduction. Do both ap-
proaches so you can check your work. Implement all these in your Sage worksheet
and bring to class. It will help with the quiz.
Problem 47. Consider the following systems of equations. For each of these do
the following.
(a.) Write the system in the form AX = B.
(b.) Solve the system of equations by first writing the system in the form A′ X = B ′ ,
where A′ is a row-reduced echelon matrix.
(c.) Find linearly independent vectors that generate the row space of A. State
dimension of row space.
(d.) Solve the system of equations AX = 0. (This is null(T ) for the associated
linear transformation T : X 7→ AX.)
(e.) Find a basis of the null space.
(f.) Find a maximal set of linearly independent vectors for the column space of
A. State the dimension of the column space.
(g.) Find the rank of the matrix A. (Is there is a method in Sage to find the rank?)
(h.) Find a maximal set of columns that are linearly independent. Delete some
entries to obtain a square matrix with linearly independent rows and columns.
Say they make a matrix A∗ . Find the inverse of the matrix A∗ by two methods:
(i.) By row-reduction of a suitable augmented matrix.
(ii.) By directly finding the inverse in Sage.



 u − v + 2w + 3x = 1

2u + 2v + 2x = 1
(7.1)

4u + v − w − x = 1

u + 2v − w − x = 1

12 G. BHATNAGAR



 u + 2v + 4w + x = 4

2u − w + 3x = 4
(7.2)

 u − 2v − w =0

3u + v − w − 5x = 7



 2u
 + w−x+ y =2
(7.3) u + w−x+ y =1

12u + 2v + 8w + 2y = 12



2u + 2v − w − x = 0

(7.4) 2u + 4v + 2w + 4x = 4

3u + 6v + 3w + 6x = 6




 u −w =1

2u + v + w =2



(7.5) v−w =3

u+ v+w =4





 2v − w =0


 u + 2w =1

2u + v + 2w =1



(7.6) v − 2w =1

u+v =1





 u − v + 4w =1


 2u + v + w + x = 2

3u − v + w − x = 2
(7.7)

 u + 2v − w + x = 1

6u + 2v + w + x = 5




 u + 3v − 3w + 2x = 1

4u + v − 2w + x = 1
(7.8)

 6u + 5v + 10w + 3x = 15

u + 2v + 3w + x = 6

 PROBLEM SETS SPRING 2023: LINEAR ALGEBRA (MAT-1001) 13

8. Problem Set 8
Problem 48. The following will help you figure out the arithmetic of permuta-
tions.
(a.) Express the following permutation as a product of disjoint cycles as well as a
product of transpositions; in addition, find the inverse.
!
1 2 3 4 5 6
5 3 2 6 4 1
(b.) Express the following product of permutations as a product of disjoint cycles.
(1, 2, 3, 4, 5)(6, 7)(1, 3, 5, 7)(1, 6, 3)
(c.) Write the m-cycle (1, 2, 3, . . . , m) as a product of transpositions of the form
(1, 2), (1, 3), (1, 4), . . . .
(d.) Write the m-cycle (1, 2, 3, . . . , m) as a product of transpositions of the form
(i, i + 1).
(e.) Prove that Sn is generated by the cycles:
(1, 2, 3, . . . , n − 1) and (n − 1, n),
that is any permutation can be written as a product of these two permutations.
(f.) Show that transpositions of the form (i, i + 1), for i = 1, 2, . . . , n − 1, generate
Sn ; that is, any permutation can be written as a product of transpositions of
this form.
Problem 49. (a.) Find all the inversions of (i, j)
(b.) Show that sgn(σ) = sgn(σ −1 ).
(c.) Show that sgn(στ ) = sgn(σ) sgn(τ ).
(d.) Show that |Sn | = n!.
Problem 50. Let Pσ be a permutation matrix corresponding to the permutation
σ; that is, it is obtained by permuting the columns of the n × n identity matrix.
(a.) Show that |Pσ | = sgn(σ)
(b.) Show that P T = P −1 .
Problem 51. List all the 24 elements of S4 and give their pictorial representation.
Find the inverse of each element and find the sign of each element. (You do a few
by hand, and then use Sage. Verify all entries mentally once you have listed using
Sage.)
Problem 52. Let B be a matrix obtained from A (an n × n matrix) by replacing
Row i (denoted Ri ) of A by Ri + cRj , for some other row index j. Show that
det B = det A.
Problem 53. Let A be an n × n matrix.
(a.) Make a matrix E such that EA gives the matrix obtained by interchanging
the ith row with the jth row of A.
14 G. BHATNAGAR

(b.) Make a matrix E such that EA gives the matrix obtained by performing the
row operation Ri ← cRi .
(c.) Make a matrix E such that EA gives the matrix obtained by performing the
row operation Rj ← Rj + cRi .
Problem 54 (RB). Let A be the matrix
 
1 2 3 4 ... 10
11 12 . . . . . . . . . . . 20 
 
. . .
. . .

A= . . ........... .  .
 .. .. .. 
. . ........... . 
91 . . . . . . . . . . . . . . . 100

(a) Calculate det A for the matrix. You may use Sage to make experiments. How-
ever, since RB has given this problem, it means it can be guessed without
computer experiments.
(b) Find the rank of the matrix A.
Problem 55. This problem is about a number theoretic matrix. Let Gn be the
n × n matrix with (i, j)th entry given by the GCD (greatest common divisor) of i
and j, denoted gcd(i, j).
(a) Write down what are G1 , G2 , G3 .
(b) Compute the determinants of G1 , G2 , G3 by hand.
(c) Use Sage to calculate det(Gn ) for several small values of n.
(d) Go to OEIS.org to guess what the value of det(Gn ).
(e) Prove your guess.
Problem 56 (Vandermonde determinant). The Vandermonde determinant ∆(x)
is the determinant of the n × n matrix
∆(x) := (xn−j
i )1≤i,j≤n .
Here x represents the variables (x1 , x2 , . . . , xn ). Show that

xn−1
1 x1n−2 . . . . . . x1 1
xn−1
2 x2n−2 . . . . . . . . x2 1
. .. . ..
∆(x) = .. . . . . . . . . ..
Y
. . = (xi − xj ).
.. .. . ..
. . . . . . . . ..
i<j
. . .
n−1 n−2
xn xn . . . . . . . . xn 1
The product on the right is over all 1 ≤ i, j ≤ n such that i < j. For example,
when n = 3, the product is
(x1 − x2 )(x1 − x3 )(x2 − x3 ).
 PROBLEM SETS SPRING 2023: LINEAR ALGEBRA (MAT-1001) 15

This determinant shows up in mathematics (and physics) in several places. The

objective of this problem is to prove this determinant. (We saw this determinant
when defining even permutations.) Prove the result by following these steps.
(a.) Before proving, verify the determinant for n = 2 and n = 3 by hand, and then
n = 3, 4 by Sage. Verification using symbols is required.
(b.) Prove that if p(z) is a polynomial and a is a complex number such that
p(a) = 0, then (z − a) is a factor of p(z).
(c.) Show that the right hand side divides the determinant. So
Y
∆(x) = c(x) (xi − xj ),
i<j

for some polynomial c(x) of x1 , x2 , . . . , xn .

(d.) By comparing the total degree of the polynomial on both sides, show that
c(x) is a constant c and does not depend on any xi .
(e.) Find the leading term on the left hand side, that is, the coefficient of
xn−1
1 xn−2
2 · · · xn−1 .
Argue that c = 1.
This completes the proof.

9. Problem Set 9
Problem 57. List the three types of elementary matrices. Find inverses of each
type.
Problem 58. Show that the inverse of a lower-triangular matrix is also lower-
triangular.
Problem 59. Show that sin x and cos x are linearly independent over R.
Problem 60 (RB). The vector space considered here is of continuous functions
on R.
(a) Show that if a ̸= b, then the functions f1 (x) = eax and f2 (x) = ebx are linearly
independent.
(b) Show that if a1 , a2 , . . . , an are distinct real numbers, then the functions
fi (x) = eai x , i = 1, 2, . . . , n
are linearly independent. (Hint. You may need to use one of the problems
above.)
(c) Consider the differential equation y ′′ − 3y ′ + 2y = 0. If you assume that y = eax
is a solution, what values must a take? Find two linearly independent solutions
of this differential equation.
Problem 61. Suppose U is invariant under T and dim U = 1. Then prove that
we must have a λ ∈ F and u such that T u = λu.
16 G. BHATNAGAR

Problem
62 (See Problem 74). Let M be the (n + 1) × (n + 1) matrix with entries

M = pij where
 
i
pij = .
j
We take nk = 0 if k > n. Find the inverse of M .

Problem 63 (RB). Make an n × n matrix A such that A, A2 , . . . , An−1 are not

zero, but An = 0.
Problem 64. A square matrix A is called nilpotent if Ak = 0 (the zero matrix)
for some k > 0.
(a) Let V = Fn [x] be a vector space of all polynomials of degree less than equal
d
to n. Let D be the linear operator D = dx , which sends a polynomial to its
derivative. Find a matrix of this operator in a suitable basis.
(b) Show that the matrix in part (a) is nilpotent.
(c) Prove that
1
= 1 + x + x2 + · · · ,
1−x
for suitable values of x.
(d) Show that if A is nilpotent, then I − A is invertible.
(e) How will you define eA ? Does A have to be a square matrix to define eA ?
Calculate eA for a matrix A where A3 = 0 but A, A2 are not 0. (You may use
a matrix you found earlier.)

10. Problem Set 10

Problem 65 (RB). Find the eigenvalues and eigenvectors of the following matri-
ces. Do it by hand and then check your work using Sage.
" # " # " # " # " #
1 1 −1 1 1 −1 1 1 1 1
, , , ,
1 1 1 −1 −1 1 1 0 0 1

Problem 66 (RB). Let α be any real number. What is the relation between the
eigenvalues of A and A + αI.?
Problem 67 (RB). If the eigenvalues of A are λ1 , λ2 , . . . , λn , what are the
eigenvalues of
B = 2I + 7A + 4A2 .
Problem 68 (Inverses). Use elementary matrices to prove the following. Let A,
B be square matrices such that BA = I, where I is a suitable identity matrix.
Then AB = I. Thus “left-inverses” are also “right-inverses”.
 PROBLEM SETS SPRING 2023: LINEAR ALGEBRA (MAT-1001) 17

Problem 69 (The matrix J of all 1s). Consider the matrix

 
1 1 1 1 1
1 1 1 1 1
 
J = 1 1 1 1 1
 
 
1 1 1 1 1
1 1 1 1 1
(a) Solve the matrix equation
JX = 5X,
T
where X = (x1 , x2 , x3 , x4 , x5 ) . Let v1 be a solution.
(b) Find 4 linearly independent solutions (say v2 , v3 , v4 , v5 ) of
JX = 0
where 0 on the right is the column vector with all 0s.
(c) Let P be the matrix formed with vi as column vectors. Compute P −1 .
(d) Compute P −1 J, JP and P −1 JP .
(e) Compute det(J). Explain.
(f) Generalize the above for an n × n matrix Jn with all 1’s.
Problem 70 (RB). Given a vector x = (x1 , . . . , xn ), let
x1 + · · · + xn
x̄ :=
n
n n
be the “mean” of x. Let A : R → R be the map
Ax = (x̄, x̄, . . . , x̄).
(a) Is this a linear map? If so, what is the matrix corresponding to A (in the
standard basis).
(b) What is the kernel of A?
(c) Find eigenvalues and the corresponding eigenvectors of A.
Problem 71 (Binet’s formula for the Fibonacci numbers). The Fibonacci numbers
are generated recursively by F0 := 0, F1 := 1 and, for n ≥ 2:
Fn = Fn−1 + Fn−2 .
The first few Fibonacci numbers are
0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, . . . .
In this problem we find the following formula for the Fibonacci numbers.
 √ √ 
1  1 + 5 n  1 − 5 n
(10.1) Fn = √ − .
5 2 2
(a) Use the recurrence relation for Fn and use Sage to generate 100 values of Fn
and Fn /Fn−1 for n > 2 (decimal expansions). Use (10.1) to generate the first
few values and check that it is correct.
18 G. BHATNAGAR

(b) Show that " # " #" #

Fn 0 1 Fn−1
=
Fn+1 1 1 Fn
and iterate to show that
" # " #n " #
Fn 0 1 F0
= .
Fn+1 1 1 F1
(c) Use eigenvalues and eigenvectors to compute F n , where
" #
0 1
F = .
1 1
(d) Use the above to prove (10.1). Check values from Sage.
(e) Show that
$ √ % √
1  1 + 5 n 1 1  1 + 5 n
Fn = √ + ≈√ .
5 2 2 5 2
Use this to show Fn /Fn−1 converges and find the value where it converges.
Note the role of the ‘largest eigenvalue’ in the solution.

11. Problem Set 11

Problem 72. Recall that a linear transformation T : V → W where V and W
are finite dimensional vector corresponds to a matrix M (T ). This matrix depends
on bases B1 and B2 for V and W respectively.
(a) Let A be the matrix given by
" #
3 4
A= .
4 −3
Find a formula for the corresponding linear transformation T : R2 → R2
for which M (T ) = A where both the bases B1 and B2 are taken to be the
elementary basis {e1 , e2 }. Verify your answer.
(b) Give intuitive reason(s) why the following set B must be a basis of R2 , where
" # " #
2 1
B= , .
1 −2
(c) For the transformation in T in part (a), find M (T ) using the basis B (for both
copies of R). (Carefully state the theorems used from the lecture notes.)
Problem 73. The following problem is about the matrix
" #
cos θ − sin θ
Rθ = .
sin θ cos θ
 PROBLEM SETS SPRING 2023: LINEAR ALGEBRA (MAT-1001) 19

(a.) Let C1 and C2 be columns of Rθ . Show that they are linearly independent.
This shows that the columns form a basis for R2 .
(b.) Draw a picture of the columns which shows intuitively these form a basis.
(c.) Find Rθ−1 .
(d.) Let P be the change of basis from the elementary basis to Rθ (in the notation
of Lecture 20). Given coordinates (x, y) of a point in R2 , find the coordinates
of a point in the new basis Rθ .
(e.) Interpret now Rθ as a linear transformation. Describe its action on e1 and e2 ,
the vectors "which
# form the elementary basis. How would you describe Rθ X
x
where X = ?
y
(f.) Now we consider eigenvalues. Show that the characteristic polynomial of Rθ
is given by
p(λ) = λ2 − 2λ cos θ + 1.
(g.) To find roots of p(λ) (over C) , recall that
eiθ = cos θ + i sin θ.
and thus:
eiθ + e−iθ eiθ − e−iθ
cos θ = and sin θ = .
2 2i
Show that the roots over C are
λ = e±iθ .
From here, find the values of θ when Rθ has eigenvalues in R.
(h.) Now we interpret Rθ : C → C by taking (x, y) ∈ R2 to be the same as
x + iy ∈ C. Argue that Rθ corresponds to
z 7→ zeiθ .
What is the inverse of this transformation? Does this explain your answer in
part (c.)?
(i.) Finally, note that z will be an eigenvector if
Rθ z = λz = e±iθ z.
Does this explain the values of θ when you get eigenvalues in part (g.)?
Problem 74 (Binomial Coefficients). Let V = Fn [x] made up of polynomials of
degree ≤ n, so of the form
p(x) = a0 + a1 x + · · · + an xn , for ai ∈ F.
(a) Let p0 (x), p1 (x), . . . , pn (x) be such that:
(i) deg pk (x) = k, for k = 0, 1, . . . , n.
(ii) They are monic, that is, the coefficient of the highest degree term is 1.
Show that {pk (x) : k = 0, 1, . . . , n} is a linearly independent set in V .
20 G. BHATNAGAR

(b) Show that {xk : k = 0, 1, . . . , n} and {(1 + x)k : k = 0, 1, . . . , n} are bases of

V.
(c) For m = 0, 1, . . . , n, write the identity
m  
m
X m k
(1 + x) = x
k=0
k
in matrix form as follows:
   
1 1
 (1 + x)  x
   
   2
 (1 + x)2  x 
..  = M  ..  .
   

 .  .
 ..  .
.
. .
 
 
n
(1 + x) xn
Write M when n = 5. Interpret it as a change of basis transformation. Explain
why M has to be invertible.

(d) Let M be the (n + 1) × (n + 1) matrix with entries M = pij where
 
i
pij = .
j
We take nk = 0 if k > n. Find the inverse of M . (Hint: try few values in


Sage and then prove the pattern.)
Problem 75 (Fourier expansions). I guess everyone is tired of linear algebra. Let
us take a break and do some calculus. Recall that
eiz = cos z + i sin z
and assume integration formulas apply whether the variable is real or complex.
You may use sage (or google to find an online integrating service) to compute the
integrals if you wish, but they are quite easy.
This problem guides you through what is called a Fourier expansion of a function.
A lot of real analysis was discovered to justify the missing steps in this and similar
calculations.
(a) Show that
Z π (
1 1 n = 0,
einx dx =
2π −π 0 n ̸= 0.
(b) From (a), obtain
(
Z π
2π m = n,
einx e−imx dx =
−π 0 m ̸= n.
 PROBLEM SETS SPRING 2023: LINEAR ALGEBRA (MAT-1001) 21

(c) Suppose a function can be written in the form

∞
X
(11.1) f (x) = an einx .
n=−∞

Then show that Z π

1
am = f (x)e−imx dx,
2π −π

assuming you can treat infinite series like finite sums, that is, integral of a sum
of terms is equal to the sum of integrals.
(d) Now we take a specific function. Let g : [−π, π] → R be defined as
(
x if 0 ≤ x ≤ π,
g(x) = |x| =
−x if − π ≤ x ≤ 0.

Sketch the graph of g.

(e) For this function g, assume it can be written in the form (11.1). Then show
that 
π/2 n = 0,
an = −1 + (−1)n
 n ̸= 0.
πn2
(f) Basel problem (Euler): Assuming the expansion of the form (11.1) that
you obtain equals the function g (so, in particular, the series converges), show
that
X 1 π2
2
= ;
n≥1
n 8
n odd

and,
∞
X 1 π2
2
= .
n=1
n 6

The last of these is called the Basel problem.

(g) Plot y = g(x) and graphs of the partial expansions obtained by truncating
(11.1) to a finite sum of terms:
N
X
gN (x) = an einx ,
n=−N

(with an as calculated above) for N = 1, 2, 3, 4, 5 on the same graph. You may

use desmos.com or sage.
22 G. BHATNAGAR

12. Problem Set 12

Problem 76 (Moved from earlier). Consider the vector space V = C over the
field R. For u, v ∈ C, we define f : C × C → C as f (u, v) := uv.
(i) Is f linear in the first slot?
(ii) Is it true that f (u, u) ≥ 0 for all u ∈ C?
(iii) Is it true that f (u, u) = 0 if and only if u = 0?
(iv) Is it true that f (u, v) = f (v, u)?
(v) Is f a norm on V ? Explain.
Problem 77. Let V = R3 be the 3-dimensional space of vectors. for vectors
u = a1 i + b1 j + c1 k and v = a2 i + b2 j + c2 k
we define the scalar product as
u · v := a1 a2 + b1 b2 + c1 c2 .
√
The length of a vector is given by |v| := v.v. Two vectors are called orthogonal
if u · v = 0.
(i) Find six 3 × 3 matrices, whose entries are all 0 and 1, and whose columns
are linearly independent and mutually orthogonal and have length 1. (Hint.
Which set with 6 = 3! objects do you know?)
(ii) Explain by invoking determinants why the columns are linearly independent.
(iii) Let M be an n × n matrix whose columns are mutually orthogonal and have
length 1. Here the definition of dot product, orthogonal vectors and length
is extended in an obvious way. Then prove that
(a) M T M = I, where M T is the transpose of M and I the n × n identity
matrix.
(b) The columns of M are linearly independent.
Problem 78 (Legendre Polynomials). Define an inner product on the space of
continuous and bounded functions defined on the interval [−1, 1] by
Z 1
⟨f, g⟩ := f (x)g(x)dx.
−1

(a.) Show that this does indeed define an inner product.

(b.) Apply the Gram-Schmidt process as applied to the set
{1, x, x2 , x3 , . . . , xn }.
Report the results uptil n = 3.
(c.) Write a program in Sage for the same.
(d.) Verify that the polynomials obtained are indeed orthogonal.
Problem 79 (RB). For v1 , v2 ∈ R, let v = (v1 , v2 ) ∈ R2 . Let
∥v∥1 := |v1 | + |v2 |
 PROBLEM SETS SPRING 2023: LINEAR ALGEBRA (MAT-1001) 23

and
∥v∥∞ := max(|v1 |, |v2 |).
(a) Show that ∥v∥1 and ∥v∥∞ are norms on R2 .
(b) Show that for any vector v in R2 , we have
√
∥v∥ ≤ ∥v∥1 ≤ 2 ∥v∥
and
√
∥v∥∞ ≤ ∥v∥ ≤ 2 ∥v∥∞ .
(c) The “unit ball” for any norm ∥ · ∥ on R2 is the set
D = {v ∈ R2 : ∥v∥ ≤ 1}.
Draw a picture of the unit balls for the norms ∥v∥, ∥v∥1 and ∥v∥∞ .
(d) Do the above for R3 .
(e) Is the function
f (v) = min(|v1 |, |v2 |)
2
a norm on R .
Problem 80 (RB). Let u, v be any two points in Rn . Show that
4⟨u, v⟩ = ∥u + v∥2 − ∥u − v∥2 .
Problem 81. Prove or disprove the following statements.
(i) The product of two orthogonal matrices is orthogonal.
(ii) The sum of two orthogonal matrices is orthogonal.
(iii) Every orthogonal matrix is invertible and its inverse is an orthogonal matrix.
(iv) If U is a 2 × 2 orthogonal matrix, then there exists a 2 × 2 orthogonal matrix
V such that V 2 = U .
Problem 82. Prove the following inequalities.
(a.) Suppose a1 , a2 , . . . are positive numbers. Show that:
 a + a + · · · + a  2 a2 + a2 + · · · + a2
1 2 n
≤ 1 2 n
.
n n
Try out the inequality for n = 2 and 3 for some choice of ai .
(b.) Suppose a1 , a2 , . . . , an are positive numbers. Then show that
1 1 1
(a1 + a2 + · · · + an ) + + ··· + ≥ n2 .
a1 a2 an
Try out the inequality for n = 2 and 3 for some choice of ai .
Problem 83 (Roots of unity). This problem is based on some central ideas of
mathematics which you will encounter a million times if you choose a scientific
career. If you follow the steps, you can discover them on your own.
24 G. BHATNAGAR

(a) Solve the equations in complex numbers: z = 1, z 2 = 1, 1 + z + z 2 = 0,

z 3 = 1, and z 4 = 1. For these use high-school algebra (the quadratic formula
is allowed, if needed). You cannot use the approach outlined in the next part.
Plot the solutions on the complex plane by hand.
(b) We consider the equation z n = 1. Write 1 in polar form as reiθ and obtain the
solutions
2πik
wk = e n , where k = 0, 1, 2, . . . , n − 1.
Verify that these are indeed the solutions you obtained in part (a). For n =
1, 2, 3, 4, 5, 8, 10, use Sage to plot them. What do you observe?
2πik
(c) Let w = wk = e n for some k = 1, . . . , n − 1. Prove that
1 + w + w2 + · · · + wn−1 = 0.
Illustrate wk , wk2 , . . . , on your graph for some small values of n and k. Say
n = 8 and k = 2, 3, 4, 5.
2πi
(d) Now let w = e n and wk be as above. We want to guess a function y = f (x)
as a linear combination
c0 ew0 x + c1 ew1 x + c2 ew2 x + · · · + cn−1 ewn−1 x .
We assume f is infinitely differentiable at 0. Let f (n) (0) denote the nth deriv-
ative of f evaluated at 0 (here f (0) (x) = f (x)). Suppose further we know the
values of f (k) (0) = yk , for all k = 0, 1, . . . , n − 1. Then show that the following
system of equations must hold.
c0 + c1 + c2 + · · · + cn−1 = y0
c0 + c1 w + c2 w2 + · · · + cn−1 wn−1 = y1
c0 + c1 w2 + c2 w4 + · · · + cn−1 w2(n−1) = y2
..
.
2
c0 + c1 wn−1 + c2 w2(n−1) + · · · + cn−1 w(n−1) = yn−1

2πi
(e) Again let w = e n . Let F be the n × n matrix (it helps to have rows and
columns indexed from 0 to n − 1, like it is done in Sage):
 
1 1 1 ... 1
1 w w2 . . . wn−1 
 

n−1  2 4 2(n−1) 

F = wij i,j=0 = 1 w w . . . w
 .. .. .. .. .. 
 
. . . . . 
2
1 wn−1 w2(n−1) . . . w(n−1)
That above system of equations can be written F C = Y , where
C = (c0 , . . . , cn−1 )T and Y = (y0 , . . . , yn−1 )T .
 PROBLEM SETS SPRING 2023: LINEAR ALGEBRA (MAT-1001) 25

Now the question. Show that

F F ∗ = nI,
where F ∗ is the conjugate transpose of F .
This shows that F is invertible; and that √1n F is unitary with orthonormal columns.
Remarks. The matrix F is called a Fourier matrix and is used a lot in digital
signal processing. The sequences (ck ) and (yk )’s are called signals in digital signal
processing, and they are encoded as above using F . The nth roots of unity form
a group with n elements.
Problem 84 (The orthogonal complement of a space). Let V be a finite dimen-
sional inner product space and U be a subset of V . The orthogonal complement
of U is defined as
U ⊥ := {v ∈ V : ⟨u, v⟩ = 0 for all u ∈ U }.
(a.) Show that U ⊥ is a subspace of V .
(b.) Show that {0}⊥ = V and V ⊥ = {0}.
(c.) If U is a subspace of V , then V = U ⊕ U ⊥ .
(d.) If U is a subspace of V , then U = (U ⊥ )⊥ .
Problem 85 (The projection operator). Let U be a subspace of V , a finite di-
mensional inner product space. By the above, any vector v ∈ V can be written in
a unique way as v = u + u⊥ , where u⊥ ∈ U ⊥ is some vector. Define PU : V → V
by v 7→ u.
(a.) Show that PU is a linear transformation.
(b.) Show that PU2 = PU .
(c.) What is the range and null space of PU ?
(d.) Let v ∈ V . Show that for every u ∈ U ,
∥v − PU v∥ ≤ ∥v − u∥.
The equality happens when PU v = u. This shows that PU can be used to
solve minimization problems.
13. Problem Set 13
Problem 86. Let A be an m × n real matrix of rank n. Let P = A(AT A)−1 AT .
Is P an orthogonal projection?
Problem 87. Let T be defined as X 7→ AX, where
" #
2 1+i
A=
1−i 3
Find its eigenvalues and verify they are real.
Problem 88. Let T : V → V be a linear operator, where V is finite dimensional.
Use the following to show that if ⟨T v, v⟩ = 0 for all v ∈ V , then T is identically 0.
26 G. BHATNAGAR

(a)

1
⟨T u, w⟩ = ⟨T (u + w), u + w⟩ − ⟨T (u − w), u − w⟩
4 
+ i⟨T (u + iw), u + iw⟩ − i⟨T (u − iw), u − iw⟩

(b) Show if ⟨T v, v⟩ = 0 for all v ∈ V then ⟨T u, w⟩ = 0 for all u, w ∈ V . From here

show that T u = 0 for all u ∈ V .
Problem 89. Suppose A = A∗ and all it’s eigenvalues are non-negative. Show
that A is positive.
Problem 90. The trace of an n × n matrix A (denoted tr) is the sum of the
diagonal entries of A. Show the following.
(a.) tr αA = α tr A
(b.) tr A + B = tr A + tr B
(c.) tr At = tr A
(d.) tr A∗ = tr AP P
(e.) tr(A∗ A) = k j |akj |2
(f.) If A∗ A = I then tr A∗ A = n.
(g.) tr AB = tr BA
Along with the determinant, the trace is an important invariant of an operator. It
stays the same for similar matrices.
Problem 91. Let T : V → V be a linear operator
(a.) Show that T is positive if and only if it has a self adjoint square root.
(b.) Show that if T is positive if and only if there is an operator such that T = S ∗ S.
Problem 92. Let S : V → V be an isometry, that is, ∥Sv∥ = ∥v∥ for all v ∈ V .
(a.) Show that Rθ defined in Problem 73 is an isometry.
(b.) Show that an isometry is invertible.
(c.) Show that S is an isometry if and only if S ∗ S = I.
Problem 93. We extend the analysis of Lecture 20. Let T : V → W be a linear
transformation. Let dim V = n and dim W = m. Let

B = {v1 , . . . , vn }
C = {w1 , . . . , wm }

be bases of V and W respectively. Let

 
A = aij = m(T )B-C

be the matrix of the linear transformation from B to C (see Lecture 8). The
matrix A is an m × n matrix. Let B ′ and C ′ be two other bases and let P and Q
 PROBLEM SETS SPRING 2023: LINEAR ALGEBRA (MAT-1001) 27

be such that
BP = B ′ = v1′ , . . . , vn′ ;
 

CQ = C ′ = w1′ , . . . , wm ′
 
.
Thus P and Q are the respective change of basis matrices. By following the steps
below, we find the matrix of T with respect to B ′ and C ′ . Let this matrix be
denoted A′ , that is,
A′ = a′ij = m(T )B ′ -C ′ .
 

(a.) Write the dimensions of A, A′ , P and Q.

(b.) Show that
Xm Xn 
T vj ′ = aik Pkj wi .
i=1 k=1
(c.) On the other hand, show that
Xm X
m 
′ ′
T vj = Qik akj wi .
i=1 k=1
′
(d.) Show that AP = QA , or
A′ = Q−1 AP.
This gives us the matrix of the linear transformations with changed bases.
(e.) Obtain the theorem of Lecture 20 as a special case of the above.
Note. The following problems from Professor Bhatia’s course will help you get
used to some notation and definitions. Draw pictures and try small examples by
hand to see what is going on. RB only considered real matrices/vector spaces in
his course. If you take LAMA with him, then do these questions for sure before
he begins.
Problem 94 (RB). Let S be a subset of R2 and let u be any point of R2 . We say
a point v in S is a best approximant to u from S if
∥u − v∥ ≤ ∥u − w∥ for all w ∈ S.
(This says that no point of S is closer to u than v is.) Let S be the subset of R2
given by
S = {(a, a) : a ∈ R}.
(i) Draw a picture of S.
(ii) For each of the following points find the best approximant from S.
" # " # " # " #
2 1 1 1
u= ; u= ; u= ; u= .
1 −1 0 1
(iii) Let u = (u1 , u2 ) be any vector in R2 . Find a best approximant to u from S.
(iv) Can any point u in R2 have two best approximants from S?
28 G. BHATNAGAR

(v) Replace S by the set

S ′ = {(a, −a) : a ∈ R}
and repeat the above for this set.
(vi) Can you think of a subset S and a point u which has more than one best
approximant from S?
Problem 95 (RB). A subset S of Rn is called convex if whenever v and w belong
to S, then so does (1 − t)v + tw, for every t in [0, 1].
(i) What does a convex subset of R look like?
(ii) Find a convex subset of R whose complement S ′ is also convex.
(iii) Draw pictures of a few convex sets in R2 .
(iv) Show that the unit ball in R2 for any norm ∥ · ∥ is convex.
(v) Is the union of any two convex sets convex? Is the intersection of any two
convex sets convex?
Problem 96 (RB). A 2 × 2 matrix is called “norm-preserving” is ∥Ax∥ = ∥x∥ for
every x ∈ R2 . Which of the following matrices are norm preserving.
" # " # " # " #
1 0 0 1 1 0 1 1
A = I, B = , C= , D= , E= ,
0 −1 1 0 0 0 0 0
" # " √ √ # " √ √ #
1 1 1/ 2 −1/ 2 1/ 2 1/ 2
F = , G= √ √ , H= √ √ .
1 1 1/ 2 1/ 2 1/ 2 1/ 2

Problem 97 (RB). A 2 × 2 matrix is called an “isometry” if for all x, y ∈ R2 ,

∥Ax − Ay∥ = ∥x − y∥.
Is the following statement correct: A is an isometry if and only if it is norm-
preserving.
Problem 98 (RB). A 2 × 2 matrix is said to be “angle-preserving” if ⟨Ax, Ay⟩ =
⟨x, y⟩. Is the following statement correct: A is angle-preserving if and only if it is
an isometry.
A question from the Final of 2022.
Problem 99 (Graph Theory). For this question, it is best if you perform your own
experiments and see what is happening. Resist the temptation to google what is
known. I want you to do lots of computations and make conjectures. Proofs
are not required, but if some observations are trivially provable by quoting some
theorem then do so. It’s a test of your mathematical creativity. Enjoy!
A graph G consists of a pair of sets (V, E), where V (the set of vertices) and E
(the set of edges) such that each e ∈ E is a 2-subset {v1 , v2 } of V . If an edge
e = {v1 , v2 }, we say e is an edge between the vertices v1 and v2 . This is also called
an undirected, simple, graph. The word simple refers to no loops (i.e., vertices
 PROBLEM SETS SPRING 2023: LINEAR ALGEBRA (MAT-1001) 29

joined to itself) or parallel edges (i.e., two edges between a pair of vertices). We
can represent a graph by using the adjacency matrix. The rows and columns
are indexed by the vertices and we place a 1 if there is an edge between vi and vj .
Here is an example of a graph on 5 vertices.
4 2

1
5
3

The corresponding matrix is

 
0 0 1 1 0
0 0 0 1 1
 
1 0 0 1 0
 
 
1 1 1 0 0
0 1 0 0 0
A graph is connected if any two vertices are connected by a path of edges. For
example, {1, 4}, {4, 2}, {2, 5} is a path of edges between 1 and 5. A cycle or
circuit is a closed path (the triangle made by vertices 1, 3, 4 is a cycle).
(a) Draw a minimum of 10 graphs on sufficiently large number of vertices and
write down their matrices. For each of these matrices, find the transpose,
eigenvalues, trace, determinants, sum of the squares of the eigenvalues. Any
observations? Is the matrix of a graph symmetric?
(b) A complete graph on n vertices is one where every pair of vertices is con-
nected. For n = 1, 2, 3, 4, 5 draw the complete graphs and input their matrices.
Write down a function in Sage to create a matrix of the complete graph on
n vertices. Find eigenvalues and eigenvectors and state any conjectures you
make from these.
(c) A tree is a connected graph with no loops or cycles. A tree is ‘minimally
connected’, that is, if you remove any edge it becomes disconnected. If you
add an edge, it will contain a cycle; in this sense it is ‘maximal’ with no cycles.
4 2

1
3
5

Make several trees and write their corresponding matrices. What is the
maximum number of 1’s in any row? Calculate the powers of the matrices
30 G. BHATNAGAR

for several examples. Can you write the powers of the matrices without any
calculation, just by looking at the tree?
(d) A Directed graph or a network has edges i → j (so each edge has a direction).
How will you represent a directed graph as a matrix?
(e) A weighted graph is a (possibly directed) graph with wieghts on edges. You
can represent this as a matrix by putting weight wij on the edge i → j and 0
in case there is no edge. Write down a few weighted graphs and examine how
various matrix operations work.
Remarks. Graph theory is a large area of mathematics (and computer science).
Graphs may be implemented using matrices and there is a large body of literature
on “spectral graph theory” where linear algebra methods are used to study graphs.
This problem is just a chance for you to discover stuff for yourself. You can learn
more by following your curiosity along with google and Chat GPT.
A wish for the future. For many of you, this may be the last math course you
take; your chosen career may not require much interaction with mathematics. But
I hope you have got a taste of the joy there is in problem solving. Here is a quote
from a mathematician who won the Steele prize for his work (google him). He spoke
about why he is a mathematician. But his remarks are applicable more widely.
Chinese poems by a decadent Japanese monk of the fifteenth cen-
tury greatly impressed me; they could be interpreted to suggest the
purpose of one’s life could be found only in an enjoyable pastime
(or more precisely, a way to kill time), lasting until one’s demise.
From that time on, I tried in earnest to find such a way to kill time.
∼ Haruzo Hida on why he does math; response to the Steele Prize (No-
tices, April 2019)
Whatever the problems you choose to solve in real life, if you enjoy solving math
problems, nothing can prevent you from making some time to do so as an enjoyable
pastime. All the best.
Ashoka University, Sonipat, Haryana 131029, India
Email address: [email protected]