Review () Linear Algebra I (-nyc-/) Winter 

Review exercises
Question . — In this problem you are given the matrix A and its reduced echelon form R:
   
 1 −2 2 −1 1 −2 0 3 
   
A = −1 2 −1 −1 ; R = 0 0 1 −2 .
   
2 −4 1 4 0 0 0 0
T is the linear transformation defined by T (x) = Ax.
a. What is the domain of T ? What is the codomain of T ?
b. Find a basis of the kernel of T .
c. Find a basis of the range of T .
d. Give a normal equation of range of T .
e. Find a vector of length 2 which points in the direction opposite to T (e1 ).
f. Find matrices M and E such that M linearly independent columns, E has a pivot position in every row, and A = ME.
! !
3 −1
Question . — Let `1 be the line Span{v1 }, and let `2 be the line Span{v2 }, where v1 = and v2 = .
4 1
a. Find a vertical shear which maps `1 onto `2 .
b. Find a horizontal shear which maps `1 onto `2 .
c. Find a rotation which maps `2 onto `1 .
d. Find a reflection which maps `1 onto `2 .
e. Find a linear transformation which maps the line v1 + Span{v2 } to e1 .
   
Note: If a2 + b2 = 1 then ba −b
a is a rotation matrix and a b is a(n orthogonal) reflection matrix.
b −a
     
 1  1  2 
−1 1 −1
      Let U = Span{u, v} and let V be the set of all vectors in R4 which are
Question . — Let u =  , v =   and x =  .
 1  1  3  orthogonal to U .
     
−1 1 2
a. Check that u is orthogonal to v.
b. Find linearly independent vectors u0 , v0 which span V .
c. Express x as a linear combination of u, v, u0 and v0 (where u0 and v0 are from part b).
d. Find the orthogonal projection of x onto U .
e. Find the orthogonal reflection of x through U .
f. Give a normal equation of U .
Question . — a. Let A be the standard matrix of the linear transformation T : R3 → R3 which rotates points about the x2 -axis
by 34 π. Find A, A4 and A23 .
   
 1 −3 2 3 1 −3 0 −1 Find a basis of ker(T ), and a
   
b. Let T : R4 → R3 be defined by T (x) = Bx, where B = −1 3 1 3 ∼ 0 0 1 2 .
    normal vector n to ran(T ).
1 −3 5 9 0 0 0 0
c. Let T : R7 → R7 be the transformation which reflects vectors in the hyperplane defined by x1 + x2 + x3 + x4 + x5 + x6 + x7 = 1.
Explain briefly why T is not a linear transformation.
d. Give four distinct linear transformations R2 → R2 , each of which is equal to its inverse.
Question . — a. Find the standard matrix of the linear transformation T : R2 → R2 which first rotates points by 23 π about
the origin and then performs a vertical shear which maps e1 to e1 − 2e2 and leaves e2 unchanged.
b. Let P : R3 → R3 be the linear transformation which projects vectors orthogonally onto the plane defined by 2x1 −5x2 +x3 = 0.
Give linearly independent spanning vectors of the kernel of P , and give linearly independent spanning vectors of the range of
P . Note that you do not need to find the standard matrix of P in order to answer this part.
c. Let ! !
2 2
u= and v= .
−3 1
Find the standard matrix of the linear transformation S : R2 → R2 such that for all x ∈ R2 , S(x) ∈ Span{u} and S(x) − x ∈ Span{v}.
Find the standard matrix of the linear transformation T : R2 → R2 such that for all x ∈ R2 , T (x) ∈ Span{u} and T (x) − x is
orthogonal to v.
d. Let R : R3 → R3 be the linear transformation which reflects vectors orthogonally through the plane defined by x1 + 2x2 +
3x3 = 0. Find the standard matrix A of R, and explain briefly why the equation A = R is complete nonsense.
Review () Linear Algebra I (-nyc-/) Winter 

Question . — Let ` be the line with parametric representation p + tv, where

   
1 2
   
p = 3 and v = 4 ,
   
2 1
and let S : R3 → R2 be the linear transformation defined by
  !
x1  2x1 − 3x2 + 5x3
 
S x2  = .
  −x1 + 2x2 − 4x3
x3

a. Write the standard matrix of S.

b. Give a parametric vector representation of S(`) = { S(x) : x ∈ ` }, and draw a picture of S(`).
c. Find a line in R3 whose image under S is a single point.
d. Is there a plane in R3 whose image under S is a single point? Explain.
e. Give implicit and explicit descriptions of the plane in R3 whose image under S is the line q + Span{w}, where
! !
2 1
q= and w= .
−1 1
Question . — Let
 
 1 1 3 1 6 
 2 −1 0 1 −1

A =  .
−3 2 1 −2 1 
 
4 1 6 1 3
a. Find bases for, and the dimensions of, Col(A), Nul(A), Col(AT ) and Nul(AT ).
b. Do the columns of A span R4 ?
c. Find a matrix B with linearly independent rows such that
Col(A) = Nul(B).
d. Find a matrix C with linearly independent rows such that
Col(AT ) = Nul(C).
Question . — Let T : M2×2 → M2×2 be the linear transformation defined by
!
1 2
T (X) = AX − XA, where A= .
−1 1
Find bases of the kernel and range of T .
Question . — Let
 
1 1
 
A = 1 2 .
 
1 3

a. Find all left inverses of A (i.e., find all matrices Y such that Y A = I.)
b. Give a particular left inverse of A of the form !
5 a b
.
c 7 d
c. Suppose that Y is a left inverse of A and C is a 3 × 3 matrix. Explain why Y C need not be a solution of the equation AX = C.
Under what conditions is Y C a solution of the equation AX = C?
Question . — For each of the following, give specific items as described, or prove that there are no such items. Justify all
assertions carefully.
a. A 3 × 3 matrix A with no zero entries, such that AT , A and AT A = I.
b. A 3 × 3 matrix B and a vector b ∈ R3 such that the equation BT Bx = BT b has no solution.
c. A 3 × 3 matrix C such that C T , C and Cx − x is orthogonal to the columns of C for all x ∈ R3 .
d. Four different 2 × 2 matrices X which satisfy the equation X 2 = 4(I − X).
Review () Linear Algebra I (-nyc-/) Winter 

Question . — a. Given  

1 0 −1
 
A = 1 0 0  ,
 
0 2 0
express A and A−1 as products of elementary matrices. Be sure to write each elementary matrix explicitly.
b. Let M and N be 27 × 27 matrices such that N , M − N −1 and MN − I are invertible, where I denotes the 27 × 27 identity
matrix. Simplify
3M(M − N −1 )−1 − (2I + MN )(MN − I)−1
as much as possible.
Question . — In this question A is an n × m matrix, B is an m × n matrix, and AB is its own inverse.
a. What are the possible values of m?
b. Compute (AB + I)2 and (AB + I)3 and simplify the results completely. Give a formula for (AB + I)k which involves no
matrix products besides AB.
c. Suppose that m = n. What is the inverse of: i. A? ii. B? iii. BA?
Give your answers using only products involving A and B.
Question . — Let
! !  
3 5 1 −1 0 7 3
 
A= , B= and C = 4 5 .
2 3 0 1 −2  
0 1

a. Evaluate A − 2(BC)T .
b. Solve the equation A(Y + 2I2 )T = BC for Y .
Question . — a. Let T : R3 → R3 be the linear transformation that projects points orthogonally onto the x1 x2 -plane and
then rotates points about the x3 -axis by 13 π. Find the standard matrix A of T , and find A5 .
b. Show that if A and B are invertible matrices and (AB)2 = A2 B2 , then A and B commute.
c. Find 2 × 2 matrices A and B such that (AB)2 = A2 B2 , but A and B do not commute.
d. Let V = Span{v1 , v2 }, where
   
 2   1 
   
v1 = −1 and v2 =  0  ,
   
0 −1
and let R be the standard matrix of the linear transformation which reflects points orthogonally in V . Find a matrix Q such that
Q−1 RQ is a diagonal matrix, and write this diagonal matrix.
Question . — A matrix X is called a weak generalized inverse of A if AXA = A.
a. For which values of α, if any, is
  !
α α 
1 1 1
 
α α  a weak generalized inverse of ?
  1 1 1
α α
b. Suppose that A is not a zero matrix, and that X and Y are weak generalized inverses of A. Under what conditions on s and
t is sX + tY a weak generalized inverse of A?
In Parts c and d below you are given that A is an m × n matrix and that X is a weak generalized inverse of A. So that AXA = A,
even though A is not necessarily invertible and may not even be square.
c. True or false: If y is any vector in Rn then (I − XA)y is a solution of Ax = 0. Justify your answer.
d. True or false: If Ax = b is consistent then Xb is a solution of this equation. Justify your answer.
e. Prove that every matrix has a weak generalized inverse.
Review () Linear Algebra I (-nyc-/) Winter 
(solutions)

Solutions to the review exercises

Solution to Question . — A. The pivot columns of A are a1 and a3 ; also,
a2 = −2a1 and a4 = 3a1 − 2a3 , (∗)
i.e.,
2a1 + a2 = 0 and − 3a1 + 2a3 + a4 = 0. (†)
a. Since A is a 3 × 4 matrix, T : R4 → R3 ; i.e., the domain of T is R4 and the codomain of T is R3 .
b. ker(T ) = { x ∈ R4 : T (x) = 0 } = Span{u, v}, where
   
2 −3
1  0 
   
u =   and v =  
0  2 
   
0 1
are obtained from (†). One sees that {u, v} is linearly independent by looking at the second and fourth entries of u and v.
c. ran(T ) = { Ax : x ∈ R4 } = Span{a1 , a2 , a3 , a4 } = Span{a1 , a3 }, by definition and by (∗), and {a1 , a3 } is linearly independent.
d. Since ran(T ) = Span{a1 , a2 }, and
! !  
 T 1 −1 2 1 0 −1 1
 
a1 a3 = ∼ , the vector w = 3
2 −1 1 0 1 −3  
1
T the range of T is defined by wT x = 0, i.e., x1 + 3x2 + x3 = 0.
spans the general solution of p A x = 0, and therefore
√ √
e. Since kT (e1 )k = ka1 k = 1 + (−1) + 2 = 6, the vector − 31 6 a1 has length 2 and points in the direction opposite to T (e1 ).
2 2 2
f. From A and its reduced echelon form, it is immediate that
  !
 1 2 
  1 −2 0 3
A = −1 −1 .
  0 0 1 −2
2 1

Solution to Question . — a. Since

! ! ! !
1 0 3 3 −1
= =t
α 1 4 3α + 4 1
for some real number t if, and only if, 3α + 7 = 0, or α = − 73 , it follows that
 
 1 0
 
 7 
− 3 1
is the standard matrix of a vertical shear which maps `1 onto `2 .
b. Since ! ! ! !
1 β 3 3 + 4β −1
= =t
0 1 4 4 1
for some real number t if, and only if, 7 + 4β = 0, or β = − 74 , it follows that
 
1 − 7 
 4
 
0 1
is the standard matrix of a horizontal shear which maps `1 onto `2 .
c. Since ! ! ! !
a −b −1 −a − b 3
= =t
b a 1 −b + a 4
1 √
for some real number t if, and only if, −4(a + b) = 3(a − b), or b = −7a. Now 1 = a2 + b2 = 50a2 , hence a = ± 10 2. Thus (for
example),
!
1 √ 1 7
10 2 −7 1

is the standard matrix of a rotation which maps `2 onto `1 .

Review () Linear Algebra I (-nyc-/) Winter 
(solutions)

d. Since ! ! ! !
a b 3 3a + 4b −1
= =t
b −a 4 −4a + 3b 1
1 √
for some real number t if, and only if, −a + 7b = 0, or a = 7b. Now, since 1 = a2 + b2 = 50b2 , it follows that b = ± 10 2. Thus (for
example), !
1 √ 7 1
10 2 1 −7

is the standard matrix of a(n orthogonal) reflection which transforms `1 into `2 .

e. If T (x) = Ax and T (v1 + Span{v2 }) = Av1 + Span{Av2 } is {e1 }, then Av2 = 0, which implies that
!
α α
A= ,
β β
for some α, β ∈ R, and that Av1 = e1 , which gives 7α = 1 and 7β = 0. Therefore,
!
1 1 1
A= 7
0 0
is the standard matrix of a linear transformation which maps v1 + Span{v2 } to e1 .
Solution to Question . — a. A direct calculation gives uT v = 1 − 1 + 1 − 1 = 0.
b. Reducing ! ! !
 T 1 −1 1 −1 1 −1 1 −1 1 0 1 0
uv = ∼ ∼ ,
1 1 1 1 0 2 0 2 0 1 0 1
gives V = Span{u0 , v0 }, where
   
 1   0 
 0   1 
   
u0 =   and v0 =   .
−1  0 
   
0 −1
0 0
c. Note that u and v are mutually orthogonal, and that
uT u = vT v = 4, u0 T u0 = v0 T v0 = 2, uT x = 4, vT x = 6, u0 T x = −1 and v0 T x = −3.
Therefore, x = u + 23 v − 12 u0 − 23 v0 .
 
5
0 3 1
1
d. The orthogonal projection of x onto U is x = u + 2 v = 2  .
5
1
     
5  2   3 
1 −1  2 
     
e. The orthogonal reflection of x in U is 2x0 − x =   −   =  .
5  3   2 
     
1 2 −1
 
−1
−3
 
f. The orthogonal projection of x onto V is x00 = − 12 u0 − 32 v0 = 12  .
 1 
 
3
g. From part b, U is the solution of N T x = 0, where
 
 1 0 
   0  1 
N = u0 v0 =   .
−1 0 
0 −1
√ √ √ √
Solution to Question . — a. As T maps e1 { − 12 2e1 − 12 2e3 , e2 { e2 and e3 { 21 2e1 − 21 2e3 , it follows that
 √ √     √ √ 
− 2 0 2  −1 0 0  − 2 0 − 2
     
A = 12  √ 0 2 0  ,
√ A4 =  0 1 0  and A23 = 12  √0 2 0  ,
√
     
− 2 0 − 2 0 0 −1 2 0 − 2
The matrix A4 is deduced from the fact that T 4 is a rotation by π about the x2 -axis, and the matrix A23 is deduced from the fact
that T 8 = idR3 and T preserves length, so A23 = A−1 = AT .
Review () Linear Algebra I (-nyc-/) Winter 
(solutions)

b. Inspecting the reduced echelon form of B gives

   


 3  1  
1  0 
 
    
ker(T ) = Span 
   ,    .

 
0 
 
−2 
 


 0     
1 
A normal vector to the range of T is obtained by reducing
! !  
 T 1 −1 1 1 0 2 −2
 
b1 b3 = ∼ , so if n = −1 ,
2 1 5 0 1 1  
1
then ran(T ) is the solution of nT y = 0.
c. The solution of T (x) = x is the given hyperplane, which does not contain 0, so T (0) , 0. Therefore, T is not linear.
d. The identity mapping of R2 , reflection in the origin, and reflection in any line containing the origin, are all examples of
linear transformations R2 → R2 which are equal to their inverses. Four simple examples are idR2 , − idR2 (i.e., reflection in the
origin), reflection in Span{e1 } and reflection in Span{e2 }. (Recall it is required to give linear transformations, not matrices.)
√  √  √
Solution to Question . — a. The linear transformation T maps e1 { − 21 e1 + 12 3e2 { − 12 e1 + 1 + 12 3 e2 and e2 { − 12 3e1 −
 
1 1√ 1 √
2 e2 { − 2 3e1 + − 2 + 3 e2 , so
 √ 
h i  − 1 − 21 3 
2
T =   .
1 + 1 √3 − 1 + √3
2 2
b. It is immediate from the geometry of orthogonal projections that
     

  2 
 
 5 0 

 
 
   

ker(P ) = Span 
 
−5 

 , and ran(P ) = Span 
 
 2 
 , 
1 
.

 
 1    
 0 5   

where the spanning vectors of the range of P are any two linearly independent vectors in R3 which are orthogonal to the
spanning vector of the kernel (i.e., the given normal to the plane).
c. Note that S(x) = x + vt belongs to Span{u}, so it is orthogonal to
!
3 wT x
w= , and thus wT (x + vt) = 0, or t=− T .
2 w v
Therefore, !
wT x vwT
S(x) = x − v T = I2 − T x.
w v w v
Hence, standard matrix of S is ! ! !
vwT 1 0 1 6 4 1 2 −4
I2 − = − = .
wT v 0 1 8 3 2 8 −3 6
Note that T (x) = ut and T (x) − x = ut − x is orthogonal to v, so that
vT x
vT (ut − x) = 0, or t= .
vT u
Therefore, !
vT x uvT uT v 4 2
T (x) = u T = T x, and so =
v u v u vT u −6 −3
is the standard matrix of T .  
d. The orthogonal reflection of x in the plane containing 0 with normal n = e1 +2e2 +3e3 is R(x) = x−2 projn x = I3 − 17 nT n x
(since nT n = 14), so the standard matrix of R is
     
1 0 0 1 2 3  6 −2 −3
     
A = 0 1 0 − 17 2 4 6 = 17 −2 3 −6 .
     
0 0 1 3 6 9 −3 −6 −2
The equation A = R is complete nonsense because A and R are not even the same kind of thing. (A is a matrix and R is a linear
transformation.)
Review () Linear Algebra I (-nyc-/) Winter 
(solutions)

Solution to Question . — a. The standard matrix of S is

!
2 −3 5
A= .
−1 2 −4

b. S(`) = { Ap + tAv : t ∈ R }, where

! 1 2
 !
    2 −3 5   3 −3
Ap Av = A p v = 3 4 = .
−1 2 −4   −3 2
2 1

  n o x2
S(`) = 3 + Span −3
−3 2

x1

Av

Ap

c. Reducing A gives
! !
2 −3 5 1 0 −2
∼ ,
−1 2 −4 0 1 −3

so ker(S) = Span{u}, where

 
2
 
u = 3 ,
 
1

is a line in R3 whose image under S is a single point.

d. The image of a flat x + V under S is S(x) + S(V ), which is a single point if, and only if S(V ) = {0}, i.e., V is a subset of the
line ker(S). Therefore, there is no plane in R3 whose image under S is a single point.
e. Let $ denote the plane in R3 whose image under S is ` 0 = q + Span{w}. Then $ contains x if, and only if, mT (Ax − q) = 0,
i.e., nT x = mT q, where m = e1 − e2 is a normal vector to ` 0 and
  !  3 
 2 −1
1  
 
n = AT m = −3 2  = −5
  −1  
5 −4 9

is a normal vector to $. Therefore nT x = mT q, or 3x1 − 5x2 + 9x3 = 3, is an implicit equation defining $. Solving this equation,
one obtains the explicit description e1 + Span{y, z} of $, where
   
5 −3
   
y = 3 and z =  0  .
   
0 1

 
Solution to Question . — First reduce A I4 so as to bring A into reduced echelon form.
 
1 0 1 0 −1 0 2 1 0
 
  0 1 2 0 3 1 1 1
0
A I4 ∼  2 2 2 
0 0 0 1
 4 1
2 − 52 − 23 0
 
0 0 0 0 0 −1 −6 −3 1
Review () Linear Algebra I (-nyc-/) Winter 
(solutions)

 
a. A basis of Col(A) is formed by the pivot columns of A, A = {a1 , a2 , a4 }, and dim Col(A) = rank(A) = 3. A basis of Nul(A)
is obtained as usual from the reduced echelon form of A,
   

 −1  1  
 
   

 

 
−2 −3

 
    
N =  
 1  ,  0   ,

 
 
 
 
 

     
 0  −4

 

 
 0   1 
 
 
and dim Nul(A) = nullity(A) = 2. The transposes of the non-zero rows of the reduced echelon form of A,
     

  1  0 0 
     

 


  0  1 0
     

 
     
 

R = ,
 1  , 2 , 0

      


  0  0 1
     

      

 
−1 3 4
 
 
form a basis of Col(AT ) = Row A, and so dim Col(AT ) = 3. The transposes of the rows of the bookkeeping matrix to the right
of the zero rows of the reduced echelon form of A,
 
−1
 

   

   

−6
 
B = −3
  ,

   


   

 1 
 
 
form a basis of Nul(AT ) = RowNul(A), and dim Nul(AT ) = 1.
b. A has 3 pivot columns, so the columns of A do not span R4 .
c. Col(A) is the null space of
 
B = −1 −6 −3 1 ,
   
since BA = 0, so Col(A) ⊂ Nul(B). But dim Col(A) = dim Nul(B) = 3, so Col(A) = Nul(B).
d. Col(AT ) is the null space of
 
−1 −2 1 0 0
C =   ,

1 −3 0 −4 1
T    
since CAT = (AC T ) = 0, so Col(AT ) ⊂ Nul(C). But dim Col(AT ) = dim Nul(C) = 3, so Col(AT ) = Nul(C).

Solution to Question . — Recall that E = {E11 , E21 , E12 , E22 } is a basis of M2×2 , where Eij has 1 in row i and column j, and 0
in every other position. Since
( ! ! ! !)
0 −2 2 0 1 0 0 2
T (E ) = , , , ,
−1 0 0 −2 0 −1 1 0
The matrix M of T relative to E is
   
 0 2 1 0 1 0 0 −1
−1 0 0 1 0 1 1
0 
 2
 ∼ ,
−2 0 0 2 0 0 0 0 
   
0 −2 −1 0 0 0 0 0
from which it follows that ( ! !)
n o 0 −2 2 0
T (E11 ), T (E21 ) = ,
−1 0 0 −2
is a basis of the range of T , and
( ! !)
n o 1 0 0 1
E11 + E22 , − 12 E21 + E12 = ,
0 1 − 12 0
is a basis of the kernel of T .
Review () Linear Algebra I (-nyc-/) Winter 
(solutions)

Note. — A basis of the kernel of T could be found by noting that {T (E11 ), T (E21 )} is linearly independent, so the range of T has
dimension > 2 and hence the kernel of T has dimension 6 4 − 2 = 2. Since I, A are linearly independent matrices in the kernel of
T , they must form a basis of the kernel of T .

 . —
Solution to Question a. The equation Y A = I2 is equivalent to AT Y T = I2 . Solving the equation AZ = I2 by reducing
the matrix AT I2 so as to bring AT into reduced echelon form,
! ! !
1 1 1 1 0 1 1 1 1 0 1 0 −1 2 −1
∼ ∼ ,
1 2 3 0 1 0 1 2 −1 1 0 1 2 −1 1

yields
   
 2 −1  1   
   
Z = −1 1  + −2 t1 t2 ,
   
0 0 1
and so ! !
2 −1 0 t  
Y = ZT = + 1 1 −2 1
−1 1 0 t2
is an explicit description of all left inverses of A.
b. The matrix ! ! !
2 −1 0 3   5 −7 3
+ 1 −2 1 =
−1 1 0 −3 −4 7 −3
is a left inverse of A as required.
c. Since the columns of A do not span R3 , the equation AX = C may have no solution, in which case Y C (although it is certainly
defined) cannot be a solution of the equation AX = C. However, if AX = C has a solution X , then X = I2 X = Y AX = Y C.
Thus, Y C is a solution of the equation AX = C if, and only if, this equation has a solution.

Solution to Question . — a. The standard matrix R of the orthogonal reflection in the line spanned by n = e1 + e2 + e3 is
given by
 
   1 −2 −2
 
R = I − 2n(nT n)−1 nT = r1 r2 r3 = 13 −2 1 −2 .
 
−2 −2 1
The columns of R are mutually orthogonal T
 unit vectors,
 so R R = I3 . Now permuting the columns of R will yield a matrix A for
which AT A = I3 , and in particular A = 13 r2 r3 r1 satisfies AT A = I3 , yet AT , A.
b. Let B be any 3 × 3 matrix, b ∈ R3 , and let b0 be the orthogonal projection of b onto Col(B). Then b0 − b is orthogonal to
the columns of B and b0 = Bt for some t ∈ R3 , so BT (Bt − b) = 0, or BT Bt = BT b. This shows that the equation BT Bx = BT b is
consistent for all B ∈ M3×3 and all b ∈ R3 . Therefore, there is no such example.
c. If Cx − x is orthogonal to the columns of C for all x ∈ R3 then
   
columnj C T (C − I) = C T (C − I) ej = C T (Cej − ej ) = 0,

for j = 1, 2, 3, so C T (C − I) = 0, i.e., C T C − C T = 0, where 0 denotes the zero 3 × 3 matrix. Hence, C T = C T C, which implies that
T T
C = (C T ) = (C T C) = C T C = C T .
Therefore, there is no such example.
d. The equation X 2 = 4(I − X) is equivalent to the equation (X + 2I)2 = 8I. Now
! !
1 0 −1 0
Z1 = I, Z2 = −I, Z3 = and Z4 =
0 −1 0 1
√
are four solutions of Z 2 = I, and so the matrices (X = 2 2Z − 2I)
√ ! √ !
± 2−1 √ 0 ± 2 − 1 √ 0
2 , 2
0 ± 2−1 0 ∓ 2−1

are four solutions of the equation in question.

Review () Linear Algebra I (-nyc-/) Winter 
(solutions)

Solution to Question . — a. Since

     
1 0 −1 1 0 −1 1 0 0
     
A ∼ 0 0 1  ∼ 0 2 0  ∼ 0 2 0 ∼ I,
     
0 2 0 0 0 1 0 0 1
1
via the sequence R2 ← R2 − R1 , R2 ↔ R3 , R1 ← R1 + R3 , R2 ← 2 R2 of elementary row operations, it follows that
    
1 0 0 1 0 1 1 0 0  1 0 0
       
 
A−1 = 0 12 0 0 1 0 0 0 1 −1 1 0 ,
       
0 0 1 0 0 1 0 1 0 0 0 1
and     
1 0 0 1 0 0 1 0 −1 1 0 0
    
A = 1 1 0 0 0 1 0 1 0  0 2 0 .
    
0 0 1 0 1 0 0 0 1 0 0 1
b. First notice that
3M(M − N −1 )−1 = 3MN N −1 (M − N −1 )−1
 −1
= 3MN (M − N −1 )N
= 3MN (MN − I)−1 ;

so that,

3M(M−N −1 )−1 − (2I + MN )(MN − I)−1

= 3MN (MN − I)−1 − (2I + MN )(MN − I)−1
= (2MN − 2I)(MN − I)−1
= 2(MN − I)(MN − I)−1
= 2I.

Solution to Question . — a. If AB is invertible then B has a left inverse, which implies that m > n.
b. If AB is its own inverse, i.e., (AB)2 = I; so
(AB + I)2 = (AB)2 + 2AB + I = 2(AB + I),
and
(AB + I)3 = (AB + I)2 (AB + I) = 2(AB + I)2 = 22 (AB + I).
In general, one has (AB + I)k = 2k−1 (AB + I).
c. If m = n and AB is its own inverse, i.e., ABAB = I, then A−1 = BAB, B−1 = ABA and BABA = I, so (BA)−1 = BA.

Solution to Question . — a. Since

! 7 
3 !
1 −1 0   3 −2
BC = 4 5 = ,
0 1 −2   4 3
0 1
one has ! ! !
T 3 5 3 4 −3 −3
A − 2(BC)= −2 = .
2 3 −2 3 6 −3
T
b. A is invertible, so the equation A(Y + 2I2 )T = BC is equivalent to (Y + 2I2 )T = A−1 BC, or Y = (A−1 BC) − 2I2 . Now,
! ! !
−1 −3 5 3 −2 11 21
A BC = = ,
2 −3 4 3 −6 −13
and therefore,
! ! !
11 −6 1 0 9 −6
Y= −2 = .
21 −13 0 1 21 −15
Review () Linear Algebra I (-nyc-/) Winter 
(solutions)

Solution to Question . — a. Since

√
T (e1 ) = cos( 13 π)e1 + sin( 13 π)e2 = 21 e1 + 12 3 e2 ,
√
T (e2 ) = − sin( 31 π)e1 + cos( 13 π)e2 = − 12 3 e1 + 21 e2 and
T (e3 ) = 0,
one has  
 21 1√   
 √ − 2 3 0  R 1
h i  1  0
0 =  3
1 π
T = A =  2 3  ,
0
 2 
  0
0 0 0
where R 1 π is the matrix of a rotation about the origin by 13 π in R2 . It follows that
3
 
   1 1√
3 0 
R 5 π 0  2√ 2 

A =  3
5  = − 1 3
  2 1
2 0  ,
0 0  

0 0 0
Since R 5 π = R−1
1 = RT1 .
3 3π 3π
b. If A and B are invertible matrices and (AB)2 = A2 B2 then
AB = A−1 A2 B2 B−1 = A−1 (AB)2 B−1 = A−1 ABABB−1 = BA,
which shows that A and B commute.
c. Let ! !
1 0 0 0
A= and B= ;
0 0 1 0
then AB = 0 and BA = B , 0, so A2 B2 = 0 and (AB)2 = 0 (since AB is a factor of each). Thus, A2 B2 = (AB)2 , but A and B do not
commute.
d. Observe that  
1
 
n = 2
 
1
 
is a normal vector to V . Since {v1 , v2 } is linearly independent, and n is orthogonal to both v1 and v2 , it follows that Q = v1 v2 n
has linearly independent columns, and is therefore invertible. Notice that
Qe1 = v1 , Qe2 = v2 and Qe3 = n,
also
R(v1 ) = v1 , R(v2 ) = v2 and R(n) = −n
and finally
Q−1 v1 = e1 , Q−1 v2 = e2 and Q−1 n = e3 .
Together, these equations imply that
 
  1 0 0 
 
Q−1 RQ = e1 e2 −e3 = 0 1 0  ,
 
0 0 −1
as required.
Solution to Question . — a. A direct calculation gives
! α α  ! !
1 1 1   1 1 1 1 1 1
α α  = 6α ,
1 1 1   1 1 1 1 1 1
α α
from which it follows that  
α α  !
  1 1 1
α α  is a weak generalized inverse of
  1 1 1
α α
if, and only if, α = 61 .
Review () Linear Algebra I (-nyc-/) Winter 
(solutions)

b. If X and Y are weak generalized inverses of A, then

A(sX + tY )A = sAXA + tAYA = sA + tA = (s + t)A.
Since A is not a zero matrix, this is equal to A if, and only if, s + t = 1. Therefore, sX + tY is a weak generalized inverse of A if,
and only if, s + t = 1.
c. Since A(I − XA) = A − AXA = A − A = 0 (the m × n zero matrix), any vector of the form (I − XA)y, where y ∈ Rn , is a solution
of Ax = 0. Therefore, the statement in question is true.
d. If Ap = b then AXb = AXAp = Ap = b, so Xb is a solution of Ax = b. Therefore, the statement in question is true.
e. First observe that the zero n × m matrix is a weak generalized inverse of the zero m × n matrix: 0m×n 0n×m 0m×n = 0m×n . If A
is not a zero matrix, then A has r pivot columns, where 1 6 r 6 min{m, n}, and A = ME, where M is an m × r matrix with r pivot
columns and E is an r × n matrix with r pivot columns. (For example, take the columns of M to be the pivot columns of A, in
order, and take the rows of E to be the non-zero rows of the reduced echelon form of A, in order.) Since E has r pivot columns
there is an n × r matrix Y such that EY = Ir , and since M has r pivot columns there is an r × m matrix Z such that ZM = Ir . Let
X = Y Z; then
AXA = (ME)Y Z(ME) = M(EY )(ZM)E = MIr Ir E = ME = A,
which shows that X is a weak generalized inverse of A.