Department of Mathematics, IIT Madras

MA-1102 Series & Matrices

Assignment-3-Sol Matrix Operations
1. Show that given any n ∈ N there exist matrices A, B ∈ Rn×n such that AB 6= BA.
Let A = [e2 e1 e3 e4 · · · en ] and B = [v u u · · · u], where e1 , . . . , en are standard basis vectors of Rn×1
and u = (1, 1, 1, . . . , 1)t , v = (0, 0, 0, . . . , 0)t .
 
1 1 0
2. Let A = 0 1 2 . Compute An .
0 0 1
 
1 n n(n − 1)
We show that An = 0 1 2n  for n ∈ N by induction.
0 0 1
The basis case n = 1 is obvious. Suppose An is as given. Now,
    
1 1 0 1 n n(n − 1) 1 n + 1 (n + 1)n
An+1 = AAn = 0 1 2 0 1 2n  = 0 1 2(n + 1)  .
0 0 1 0 0 1 0 0 1
Notice that taking n = 0 in the matrix An , we see that A0 = I.

3. Let A ∈ Fm×n ; B ∈ Fn×k . Let A1 , . . . , Am be the rows of A and let B1 , . . . , Bk be the columns of B. Show
that
(a) A1 B, . . . , AM B are the rows of AB. (b) AB1 , . . . , ABk are the columns of AB.
(a) The jth entry in Ai B is Ai · Bj , which is the (i, j)th entry in AB.
(b) The ith entry in ABi is Ai · Bj , which is the (i, j)th entry in AB.
 
I A
4. Let A ∈ Fn×n ; I be the identity matrix of order n. Find the inverse of the 2n × 2n matrix .
0 I
 −1  
I A I −A
= .
0 I 0 I
5. If A is a hermitian (symmetric) invertible matrix, then show that A−1 is hermitian (symmetric).
A∗ = A. Then (A−1 )∗ = (A∗ )−1 = A−1 . So, A−1 is hermitian.
6. If A is a lower (upper) triangular invertible matrix, then A−1 is lower (upper) triangular.
Suppose A is a lower triangular matrix. Let D be the diagonal matrix whose diagonal entries are exactly the
diagonal entries of A in the correct order. Since A is invertible, D is also invertible. Then write A = D(I + N ).
Here, N is a lower triangular matrix with all diagonal entries as 0. Then verify that A−1 = (I − N )D−1 . Also,
verify that this is a lower triangular matrix.
7. Let x, y ∈ F1×n (or in Fn×1 ); α ∈ F. Prove the following:
(a) kx + yk2 + kx − yk2 = 2(kxk2 + kyk2 . (Parallelogram Law)
(b) |hx, yi| ≤ kxk kyk. (Cauchy-Schwartz inequality)
(c) kx + yk = kxk + kyk. (Triangle inequality)
(d) If x ⊥ y, then kx + yk2 = kxk2 + kyk2 . (Pythagoras’ Law)
(a) Expand the norms using inner product.
(b) A proof of Cauchy-Schwartz inequality goes as follows:
hx,yi hy,xi
If y = 0, then the inequality clearly holds. Else, hy, yi =
6 0. Write α = hy,yi . Then α = hy,yi and αhx, yi =
2 2
|α| kyk . Then


0 ≤ hx − αy, x − αyi = hx, xi − αhx, yi + α αhy, yi − hy, xi
|hx, yi|2
= kxk2 − αhx, yi = kxk2 − |α|2 kyk2 = kxk2 − kyk2 .
kyk4
 (c) The triangle inequality can be proved using Cauchy-Schwartz, as in the following:

kx + yk2 = hx + y, x + yi = kxk2 + kyk2 + hx, yi + hy, xi ≤ kxk2 + kyk2 + 2kxk kyk.

(d) hx + y, x + yi = hx, xi + hx, yi + hy, xi + hy, yi = kxk2 + kyk2 .

8. Show that each orthogonal 2 × 2 matrix is either a reflection or a rotation.

If A = [aij ] is an orthogonal matrix of order 2, then At A = I implies

a211 + a221 = 1 = a212 + a222 , a11 a12 + a21 a22 = 0.

Thus, there exist α, β such that a11 = cos α, a21 = sin α, a12 = cos β, a22 = sin β and
cos(α − β) = 0. It then follows that A is in one of the following forms:
   
cos θ − sin θ cos θ sin θ
O1 := , O2 := .
sin θ cos θ sin θ − cos θ
−−−→
Let (a, b) be the vector in the plane that starts at the origin and ends at the point (a, b). Writing the point (a, b)
as a column vector [a b]t , we see that the matrix product O1 [a b]t is the end-point of the vector obtained by
−−−→
rotating the vector (a, b) by an angle θ. Similarly, O2 [a b]t gives a point obtained by reflecting (a, b) along a
straight line that makes an angle θ/2 with the x-axis. Thus, O1 is said to be a rotation by an angle θ and O2 is
called a reflection by an angle θ/2 along the x-axis.

9. Determine linear independence of {(1, 2, 2, 1), (1, 3, 2, 1), (4, 1, 2, 2), (5, 2, 4, 3)} in C1×4 .
(5, 2, 4, 3) = 2(1, 2, 2, 1) − 1(1, 3, 2, 1) + 1(4, 1, 2, 2). So, the set is lineraly dependent.

10. Let u, v, w ∈ Fn×1 . Show that {u, v, w} is linearly independent iff {u+v, v+w, w+u} is linearly independent.
a(u + v) + b(v + w) + c(w + u) = 0 ⇒ (a + c)u + (a + b)v + (b + c)w = 0
⇒ a + c = 0, a + b = 0, b + c = 0 ⇒ a = 0, b = 0, c = 0.
Hence {u + v, v + w, w + u} is linearly independent.
Conversely, αu + βv + γw = 0 ⇒ α+β−γ 2 (u + v) + β+γ−α
2 (v + w) + α+γ−β
2 (w + u) = 0
⇒ α + β − γ = 0, β + γ − α = 0, α + γ − β = 0 ⇒ α = β = γ = 0.
So, {u, v, w} is linearly independent.

11. Find a basis for the subspace {(a, b, c) : 2a + 3b − 4c = 0} of R1×4 .

As a subspace of R1×4 , {(a, b, c) : 2a + 3b − 4c = 0} = {(a, b, 2a+3b
4 ) : a, b ∈ R}.
The vectors (1, 0, 1/2) and (0, 1, 3/4) are in the subspace.
(a, b, 2a+3b
4 ) = a(1, 0, 1/2) + b(0, 1, 3/4). So, these two vectors span the subspace.
Now, a(1, 0, 1/2) + b(0, 1, 3/4) = (0, 0, 0) ⇒ a = 0, b = 0, 2a+3b 4 = 0. So, the vectors are linearly indepen-
dent. Hence a basis for the subspace is {(1, 0, 1/2), (0, 1, 3/4)}.

12. Let A ∈ R3×3 satisfy A(a, b, c)t = (a + b, 2a − b − c, a + b + c)t . Determine A and also its rank and nullity.
Ae1 = A(1, 0, 0)t = (1, 2, 1)t , Ae2 = A(0, t t t t
 1, 0) = (1, −1, 1) , Ae3 = A(0, 0, 1) = (0, −1, 1) .
1 1 0
So, A = [Ae1 Ae2 Ae3 ] = 2 −1 −1 .

1 1 1
Now, aAe1 + bAe2 + cAe3 = 0 ⇒ a + b = 0, 2a − b − c = 0, a + b + c = 0 ⇒ a = b = c = 0.
Thus, the three columns of A are linearly independent. So, rank(A) = 3 and null(A) = 3 − 3 = 0.

13. Determine a basis of the subspace U = {(a, b, c, d, e) : a = c = e, b + d = 0} of R1×5 .

U = {(a, b, a, −b, a) : a, b ∈ R}. We claim that {(1, 0, 1, 0, 1), (0, 1, 0, −1, 0)} is a basis of U.
First, these two vectors are in U.
Second, (a, b, a, −b, a) = a(1, 0, 1, 0, 1) + b(0, 1, 0, −1, 0). Thus, the set spans U.
Third, if a(1, 0, 1, 0, 1) + b(0, 1, 0, −1, 0) = 0, then a = b = 0. So, the set is linearly independent.

14. Let A ∈ Fm×n have rank r. Give reasons for the following:

(a) rank(A) ≤ min{m, n}.

 (b) If n > m, then there exist x, y ∈ Fn×1 such that x 6= y and Ax = Ay.
(c) If n < m, then there exists y ∈ Fm×1 such that for no x ∈ Fn×1 , Ax = y.
(d) If n = m, then as a map, A is one-one iff A is onto.

(a) rank(A) is the maximum number of linearly independent rows in A. So, r ≤ m. Also, rank(A) is the
maximum number of linearly independent columns in A. So, r ≤ n. Therefore, r ≤ min{m, n}.
(b) Suppose n > m. Then rank(A) ≤ m < n and rank(A) + null(A) = n implies that null(A) > 0. That is,
N (A) is at least one dimensional. So, there exists a nonzero vector v ∈ Fn×1 such that Av = 0. But A(0) = 0.
Thus we take x = 0 and y as this v so that Ax = Ay but x 6= y. In other words, A as a map is not one-one.
(c) Suppose n < m. Then rank(A) ≤ n < m. So, R(A) is a proper subspace of Fm×1 . That is, as a map, A is
not onto. Then the conclusion follows.
(d) Suppose n = m. Now, A is one-one iff N (A) = {0} iff null(A) = 0 iff rank(A) = n = m iff
R(A) = Fm×1 iff A is onto.
 
1 −3 2 −2 11
15. Convert  3 8 −7 5 8  into its row echelon form and row reduced echelon form using the algo-
5 2 −3 1 7
rithms. Then determine its rank and nullity.
     
1 −3 2 −2 11 1 −3 2 −2 11 1 0 −5/17 −1/17 0
(a)  3 8 −7 5 8  −→  0 17 −13 11 −25  −→  0 1 −13/17 11/17 0 
5 2 −3 1 7 0 0 0 0 23 0 0 0 0 1
rank = 3, nullity = 2

16. The vectors u1 = (1, 1, 0), u2 = (0, 1, 1), u3 = (1, 0, 1) form a basis for F3 . Apply Gram-Schmidt Orthogo-
nalization.
v1 = (1, 1, 0).
hu2 , v1 i (0, 1, 1) · (1, 1, 0) 1  1 1 
v2 = u2 − v1 = (0, 1, 1) − (1, 1, 0) = (0, 1, 1) − (1, 1, 0) = − , , 1 .
hv1 , v1 i (1, 1, 0) · (1, 1, 0) 2 2 2
hu3 , v1 i hu3 , v2 i
v3 = u3 − v1 − v2
hv1 , v1 i hv2 , v2 i
 1 1  1 1 
= (1, 0, 1) − (1, 0, 1) · (1, 1, 0)(1, 1, 0) − (1, 0, 1) · − , , 1 − , ,1
2 2 2 2
1 1  1 1   2 2 2
= (1, 0, 1) − (1, 1, 0) − − , ,1 = − , ,− .
n 2 3
2 2
o 3 3 3
The set (1, 1, 0), − 12 , 21 , 1 , − 23 , 32 , − 23 is orthogonal.
√ √ √ √ √
17. Let A ∈ R3×3 have the first two columns as (1/ 3, 1/ 3, 1/ 3, )t and (1/ 2, 0, −1/ 2)t . Determine the
third column of A so that A is an orthogonal matrix.
Notice that the first two columns of A are already orthonormal, and orthogonal to each other. You can start with
the third as (0, 0, 1)t and use Gram-Schmidt process. Alternatively, let the third column be (a, b, c)t . Then the
first two are orthogonal to the third implies a + b + c = 0, a − c = 0. This gives (a, t t
2 2 2 2
√ b, c) = (a, −2a, a) . Now,
the third column
√ √norm√1 implies that 1 = a + 4a + a = 6a ⇒ a = ±1/ 6. Thus the third column of
has
A is ±(1/ 6, −2/ 6, 1/ 6)t .