Chapter 3: Linear Maps

Linear Algebra Done Right, by Sheldon Axler

A: The Vector Space of Linear Maps

Problem 1
Suppose b, c ∈ R. Define T : R3 → R2 by

T (x, y, z) = (2x − 4y + 3z + b, 6x + cxyz).

Show that T is linear if and only if b = c = 0.

Proof. (⇐) Suppose b = c = 0. Then

T (x, y, z) = (2x − 4y + 3z, 6x).

Let (x1 , y1 , z1 ), (x2 , y2 , z2 ) ∈ R3 . Then

T ((x1 , y1 , z1 ) + (x2 , y2 , z2 )) = T (x1 + x2 , y1 + y2 , z1 + z2 )

= (2(x1 + x2 ) − 4(y1 + y2 ) + 3(z1 + z2 ), 6(x1 + x2 ))
= (2x1 + 2x2 − 4y1 − 4y2 + 3z1 + 3z2 , 6x1 + 6x2 )
= (2x1 − 4y1 + 3z1 , 6x1 ) + (2x2 − 4y2 + 3z2 , 6x2 )
= T (x1 , y1 , z1 ) + T (x2 , y2 , z2 ).

Now, for λ ∈ F and (x, y, z) ∈ R3 , we have

T (λ(x, y, z)) = T (λx, λy, λz)

= (2(λx) − 4(λy) + 3(λz), 6(λx))
= (λ(2x − 4y + 3z), λ(6x))
= λ(2x − 4y + 3z, 6x)
= λT (x, y, z),

and thus T is a linear map.

(⇒) Supose T is a linear map. Then

T (x1 + x2 , y1 + y2 , z1 + z2 ) = T (x1 , y1 , z1 ) + T (x2 , y2 , z2 ) (†)

for all (x1 , y1 , z1 ), (x2 , y2 , z2 ) ∈ R3 . In particular, by applying the definition of

T and comparing first coordinates of both sides of (†), we have

2(x1 + x2 ) − 4(y1 + y2 ) + 3(z1 + z2 ) + b =

(2x1 − 4y1 + 3z1 + b) + (2x2 − 4y2 + 3z2 + b),

1
and after simplifying, we have b = 2b, and hence b = 0. Now by applying the
definition of T and comparing second coordinates of both sides of (†), we have

6(x1 + x2 ) + c(x1 + x2 )(y1 + y2 )(z1 + z2 ) = 6x1 + c(x1 y1 z1 ) + 6x2 + c(x2 y2 z2 )

= 6(x1 + x2 ) + c(x1 y1 z1 + x2 y2 z2 ),

which implies

c(x1 + x2 )(y1 + y2 )(z1 + z2 ) = c(x1 y1 z1 + x2 y2 z2 ).

Now suppose c ≠ 0. Then choosing (x1 , y1 , z1 ) = (x2 , y2 , z2 ) = (1, 1, 1), the

equation above implies 8 = 2, a contradiction. Thus c = 0, completing the
proof.

Problem 3
Suppose T ∈ L(Fn , Fm ). Show that there exist scalars Aj,k ∈ F for
j = 1, . . . , m and k = 1, . . . , n such that

T (x1 , . . . , xn ) = (A1,1 x1 + · · · + A1,n xn , . . . , Am,1 x1 + · · · + Am,n xn )

for every (x1 , . . . , xn ) ∈ Fn .

Proof. Given x ∈ Fn , we may write

x = x1 e1 + · · · + xn en ,

where e1 , . . . , en is the standard basis of Fn . Since T is linear, we have

T x = T (x1 e1 + · · · + xn en ) = x1 T e1 + · · · + xn T en .

Now for each T ek ∈ Fm , where k = 1, . . . , n, there exist A1,k , . . . , Am,k ∈ F such

that

T ek = A1,k e1 + · · · + Am,k em


= A1,k , . . . , Am,k

and thus

xk T ek = A1,k xk , . . . , Am,k xk .
Therefore, we have
n
X

Tx = A1,k xk , . . . , Am,k xk
k=1
 
n
X n
X
= A1,k xk , . . . , Am,k xk  ,
k=1 k=1

and thus there exist scalars Aj,k ∈ F for j = 1, . . . , m and k = 1, . . . , n of the

desired form.

2
 Problem 5
Prove that L(V, W ) is a vector space.

Proof. We check each property in turn.

Commutative: Given S, T ∈ L(V, W ) and v ∈ V , we have

(T + S)(v) = T v + Sv = Sv + T v = (S + T )(v)

and so addition is commutative.

Associative: Given R, S, T ∈ L(V, W ) and v ∈ V , we have

((R + S) + T )(v) = (R + S)(v) + T v = Rv + Sv + T v

= R + (S + T )(v) = (R + (S + T ))(v)

and so addition is associative. And given a, b ∈ F, we have

((ab)T )(v) = (ab)(T v) = a(b(T v)) = (a(bT ))(v)

and so scalar multiplication is associative as well.

Additive identity: Let 0 ∈ L(V, W ) denote the zero map, let T ∈ L(V, W ),
and let v ∈ V . Then

(T + 0)(v) = T v + 0v = T v + 0 = T v

and so the zero map is the additive identity.

Additive inverse: Let T ∈ L(V, W ) and define (−T ) ∈ L(V, W ) by (−T )v =
−T v. Then
(T + (−T ))(v) = T v + (−T )v = T v − T v = 0,
and so (−T ) is the additive inverse for each T ∈ L(V, W ).
Multiplicative identity: Let T ∈ L(V, W ). Then

(1T )(v) = 1(T v) = T v

and so the multiplicative identity of F is the multiplicative identity of scalar

multiplication.
Distributive properties: Let S, T ∈ L(V, W ), a, b ∈ F, and v ∈ V . Then

(a(S + T ))(v) = a((S + T )(v)) = a(Sv + T v) = a(Sv) + a(T v)

= (aS)(v) + (aT )(v)

and

((a + b)T )(v) = (a + b)(T v) = a(T v) + b(T v) = (aT )(v) + (bT )(v)

and so the distributive properties hold.

Since all properties of a vector space hold, we see L(V, W ) is in fact a vector
space, as desired.

3
 Problem 7
Show that every linear map from a 1-dimensional vector space to itself is
multiplication by some scalar. More precisely, prove that if dim V = 1
and T ∈ L(V, V ), then there exists λ ∈ F such that T v = λv for all
v ∈V.

Proof. Since dim V = 1, a basis of V consists of a single vector. So let w ∈ V

be such a basis. Then there exists α ∈ F such that v = αw and λ ∈ F such that
T w = λw. It follows
T v = T (αw) = αT w = αλw = λ(αw) = λv,
as desired.

Problem 9
Give an example of a function φ : C → C such that

φ(w + z) = φ(w) + φ(z)

for all w, z ∈ C but φ is not linear. (Here C is thought of as a complex

vector space.)

Proof. Define
φ:C→C
x + yi 7→ x − yi.
Then for x1 + y1 i, x2 + y2 i ∈ C, it follows
φ((x1 + y1 i) + (x2 + y2 i)) = φ((x1 + x2 ) + (y1 + y2 )i)
= (x1 + x2 ) − (y1 + y2 )i
= (x1 − y1 )i + (x2 − y2 )i
= φ(x1 + y1 i) + φ(x2 + y2 i)
and so φ satisfies the additivity requirement. However, we have
φ(i · i) = φ(−1) = −1
and
i · φ(i) = i(−i) = 1
and hence φ fails the homogeneity requirement of a linear map.
Problem 11
Suppose V is finite-dimensional. Prove that every linear map on a
subspace of V can be extended to a linear map on V . In other words,
show that if U is a subspace of V and S ∈ L(U, W ), then there exists
T ∈ L(V, W ) such that T u = Su for all u ∈ U .

4
Proof. Suppose U is a subspace of V and S ∈ L(U, W ). Let v1 , . . . , um be a
basis of U and let v1 , . . . , vm , vm+1 , . . . , vn be an extension
Pn of this basis to V .
For any z ∈ V , there exist a1 , . . . , an ∈ F such that z = k=1 ak vk , and so we
define

T :V →W
n
X m
X n
X
ak vk 7→ ak Svk + ak vk .
k=1 k=1 k=m+1

Since every v ∈ V has a unique representation as a linear combination of elements

of our basis, the map is well-defined. We first show T is a linear map. So suppose
z1 , z2 ∈ V . Then there exist a1 , . . . an ∈ F and b1 , . . . , bn ∈ F such that

z1 = a1 v1 + · · · + an vn and z2 = b1 v1 + · · · + bn vn .

It follows
 
n
X n
X
T (z1 + z2 ) = T  ak vk + bk v k 
k=1 k=1
 
n
X
=T (ak + bk )vk 
k=1
m
X n
X
= (ak + bk )Svk + (ak + bk )vk
k=1 k=m+1
   
m
X n
X m
X n
X
= ak Svk + ak vk  +  bk Svk + bk v k 
k=1 k=m+1 k=1 k=m+1
   
n
X n
X
=T ak v k  + T  bk vk 
k=1 k=1

= T z1 + T z2

and so T is additive. To see that T is homogeneous, let λ ∈ F and z ∈ V , so

5
 Pn
that we may write z = k=1ak vk for some a1 , . . . , an ∈ F. We have
 
Xn
T (λz) = T λ a k vk 
k=1
 
n
X
=T (λak )vk 
k=1
 
m
X n
X
=S (λak )vk  + (λak )vk
k=1 k=m+1
 
m
X n
X
= λS  ak vk  + λ ak vk
k=1 k=m+1
   
m
X n
X
= λ S  ak vk  + λak vk 
 
k=1 k=m+1
 
n
X
= λT  ak vk 
k=1

= λT z

and so T is homogeneous as well hence T ∈ L(V, W ). Lastly, to see that T |U = S,

Pm
let u ∈ U . Then there exist a1 , . . . , am ∈ F such that u = k=1 ak vk , and hence
 
m
X
Tu = T  ak vk 
k=1
 
m
X
=S ak vk 
k=1

= Su,

and so indeed T agrees with S on U , completing the proof.

Problem 13
Suppose v1 , . . . , vm is a linearly dependent list of vectors in V . Suppose
also that W ̸= {0}. Prove that there exist w1 , . . . , wm ∈ W such that no
T ∈ L(V, W ) satisfies T vk = wk for each k = 1, . . . , m.

Proof. Since v1 , . . . , vm is linearly dependent, one of them may be written as a

linear combination of the others. Without loss of generality, suppose this is vm .

6
Then there exist a1 , . . . , am−1 ∈ F such that

vm = a1 v1 + · · · + am−1 vm−1 .

Since W ̸= {0}, there exists some nonzero z ∈ W . Define w1 , . . . , wm ∈ W by

(
z if k = m
wk =
0 otherwise.

Now suppose there exists T ∈ L(V, W ) such that T vk = wk for k = 1, . . . , m. It

follows

T (0) = T (vm − a1 v1 − · · · − am−1 vm−1 )

= T vm − a1 T v1 − · · · − am−1 T vm−1
= z.

But z ̸= 0, and thus T (0) ̸= 0, a contradiction, since linear maps take 0 to 0.

Therefore, no such linear map can exist.

B: Null Spaces and Ranges

Problem 1
Give an example of a linear map T such that dim null T = 3 and
dim range T = 2.

Proof. Define the map

T : R5 → R5
(x1 , x2 , x3 , x4 , x5 ) 7→ (0, 0, 0, x4 , x5 ).

First we show T is a linear map. Suppose x, y ∈ R5 . Then

T (x + y) = T ((x1 , x2 , x3 , x4 , x5 ) + (y1 , y2 , y3 , y4 , y5 ))
= T (x1 + y1 , x2 + y2 , x3 + y3 , x4 + y4 , x5 + y5 )
= (0, 0, 0, x4 + y5 , x5 + y5 )
= (0, 0, 0, x4 , x5 ) + (0, 0, 0, y4 , y5 )
= T (x) + T (y).

Next let λ ∈ R. Then

T (λx) = T (λx1 , λx2 , λx3 , λx4 , λx5 )

= (0, 0, 0, λx4 , λx5 )
= λ(0, 0, 0, x4 , x5 )
= λT (x),

7
and so T is in fact a linear map. Now notice that

null T = {(x1 , x2 , x3 , 0, 0) ∈ R5 | x1 , x2 , x3 ∈ R}.

This space clearly has as a basis e1 , e2 , e3 ∈ R5 and hence has dimension 3. Now,
by the Fundamental Theorem of Linear Maps, we have

dim R5 = 3 + dim range T

and hence dim range T = 2, as desired.

Problem 3
Suppose v1 , . . . , vm is a list of vectors in V . Define T ∈ L(Fm , V ) by

T (z1 , . . . , zm ) = z1 v1 + · · · + zm vm .

(a) What property of T corresponds to v1 , . . . , vm spanning V ?

(b) What property of T corresponds to v1 , . . . , vm being linearly inde-

pendent?

Proof. (a) We claim surjectivity of T corresponds to v1 , . . . , vm spanning V .

To see this, suppose T is surjective, and let w ∈ V . Then there exists
z ∈ Fm such that T z = w. This yields

z1 v1 + · · · + zm vm = w,

and hence every w ∈ V can be expressed as a linear combination of

v1 , . . . , vn . That is, span(v1 , . . . , vn ) = V .
(b) We claim injectivity of T corresponds to v1 , . . . , vm being linearly inde-
pendent. To see this, suppose T is injective, and let a1 , . . . , an ∈ F such
that
a1 v1 + · · · + an vn = 0.
Then
T (a) = T (a1 , . . . , an ) = a1 v1 + · · · + an vn = 0
which is true iff a = 0 since T is injective. That is, a1 = · · · = an = 0 and
hence v1 , . . . , vn is linearly independent.

Problem 5
Give an example of a linear map T : R4 → R4 such that

range T = null T.

8
Proof. Define

T : R4 → R4
(x1 , x2 , x3 , x4 ) 7→ (x3 , x4 , 0, 0).

Clearly T is a linear map, and we have

null T = {(x1 , x2 , x3 , x4 ) | x3 = x4 = 0 ∈ R} = R2 × {0}2

and
range T = {(x, y, 0, 0) | x, y ∈ R} = R2 × {0}2 .
Hence range T = null T , as desired.

Problem 7
Suppose V and W are finite-dimensional with 2 ≤ dim V ≤ dim W . Show
that {T ∈ L(V, W ) | T is not injective} is not a subspace of L(V, W ).

Proof. Let Z = {T ∈ L(V, W ) | T is not injective}, let v1 , . . . , vm be a basis of

V , where m ≥ 2, and let w1 , . . . , wn be a basis of W , where n ≥ m. We define
T ∈ L(V, W ) by its behavior on the basis

0 if k = 1


T vk := w2 if k = 2
 1 wk otherwise


2

so that clearly T is not injective since T v1 = 0 = T (0), and hence T ∈ Z.

Similarly, define S ∈ L(V, W ) by its behavior on the basis

w1 if k = 1


Svk := 0 if k = 2
 1 wk otherwise


2

and note that S is not injective either since Sv2 = 0 = S(0), and hence S ∈ Z.
However, notice

(S + T )(vk ) = wk for k = 1, . . . , n

and hence null(S + T ) = {0} since it takes the basis of V to the basis of W , so
that S + T is in fact injective. Therefore S + T ̸∈ Z, and Z is not closed under
addition. Thus Z is not a subspace of L(V, W ).

Problem 9
Suppose T ∈ L(V, W ) is injective and v1 , . . . , vn is linearly independent
in V . Prove that T v1 , . . . , T vn is linearly independent in W .

9
Proof. Suppose a1 , . . . , an ∈ F are such that
a1 T v1 + · · · + an T vn = 0.
Since T is a linear map, it follows
T (a1 v1 + · · · + an vn ) = 0.
But since null T = {0} (by virtue of T being a linear map), this implies a1 v1 +
· · · + an vn = 0. And since v1 , . . . , vn are linearly independent, we must have
a1 = · · · = an = 0, which in turn implies T v1 , . . . , T vn is indeed linearly
independent in W .
Problem 11
Suppose S1 , . . . , Sn are injective linear maps such that S1 S2 . . . Sn makes
sense. Prove that S1 S2 . . . Sn is injective.

Proof. For n ∈ Z≥2 , let P (n) be the statement: S1 , . . . , Sn are injective linear
maps such that S1 S2 . . . Sn makes sense, and the product S1 S2 . . . Sn is injective.
We induct on n.
Base case: Suppose n = 2, and assume S1 ∈ L(V0 , V1 ) and S2 ∈ L(V1 , V2 ), so
that the product S1 S2 is defined, and assume that both S1 and S2 are injective.
Suppose v1 , v2 ∈ V0 are such that v1 = ̸ v2 , and let w1 = S2 v1 and w2 = S2 v.
Since S2 is injective, w1 ̸= w2 . And since S1 is injective, this in turn implies
that S1 (w1 ) ̸= S1 (w2 ). In other words, S1 (S2 (v1 )) ̸= S1 (S2 (v2 )), so that S1 S2 is
injective as well, and hence P (2) is true.
Inductive step: Suppose P (k) is true for some k ∈ Z+ , and consider the
product (S1 S2 . . . Sk )Sk+1 . The term in parentheses is injective by hypothesis,
and the product of this term with Sk+1 is injective by our base case. Thus
P (k + 1) is true.
By the principle of mathematical induction, the statement P (n) is true for
all n ∈ Z≥2 , as was to be shown.

Problem 13
Suppose T is a linear map from F4 to F2 such that

null T = {(x1 , x2 , x3 , x4 ) ∈ F4 | x1 = 5x2 and x3 = 7x4 }.

Prove that T is surjective.

Proof. We claim the list

(5, 1, 0, 0), (0, 0, 7, 1)
is a basis of null T . This implies
dim range T = dim F4 − dim null T
=4−2
= 2,

10
and hence T is surjective (since the only 2-dimensional subspace of F2 is the
space itself). So let’s prove our claim that this list is a basis.
Clearly the list is linearly independent. To see that it spans null T , suppose
x = (x1 , x2 , x3 , x4 ) ∈ null T , so that x1 = 5x2 and x3 = 7x4 . We may write
       
x1 5x2 5 0
x2   x2  1 0
 =
x3  7x4  = x2 0 + x4 7 ,
    

x4 x4 0 1

and indeed x is in the span of our list, so that our list is in fact a basis, completing
the proof.

Problem 15
Prove that there does not exist a linear map from F5 to F2 whose null
space equals

{(x1 , x2 , x3 , x4 , x5 ) ∈ F5 | x1 = 3x2 and x3 = x4 = x5 }.

Proof. Suppose such a T ∈ L(F5 , F2 ) did exist. We claim

(3, 1, 0, 0, 0), (0, 0, 1, 1, 1)

is a basis of null T . This implies

dim range T = dim F5 − dim null T

=5−2
= 3,

which is absurd, since the codomain of T has dimension 2. Hence such a T

cannot exist. So, let’s prove our claim that this list is a basis.
Clearly (3, 1, 0, 0, 0), (0, 0, 1, 1, 1) is linearly independent. To see that it spans
null T , suppose x = (x1 , . . . , x5 ) ∈ null T , so that x1 = 3x2 and x3 = x4 = x5 .
We may write
       
x1 3x2 3 0
x2   x2  1 0
       
x3  =  x3  = x2 0 + x3 1 ,
       
x4   x3  0 1
x5 x3 0 1

and indeed x is in the span of our list, so that our list is in fact a basis, completing
the proof.

Problem 17
Suppose V and W are both finite-dimensional. Prove that there exists
an injective linear map from V to W if and only if dim V ≤ dim W .

11
Proof. (⇒) Suppose T ∈ L(V, W ) is injective. If dim V > dim W , Thereom 3.23
tells us that no map from V to W would be injective, a contradiction, and so we
must have dim V ≤ dim W .
(⇐) Suppose dim V ≤ dim W . Then the inclusion map ι : V → W is both a
linear map and injective.

Problem 19
Suppose V and W are finite-dimensional and that U is a subspace of V .
Prove that there exists T ∈ L(V, W ) such that null T = U if and only if
dim U ≥ dim V − dim W .

Proof. (⇐) Suppose dim U ≥ dim V − dim W . Since U is a subspace of V , there

exists a subspace U ′ of V such that

V = U ⊕ U ′.

Let u1 , . . . um be a basis for U , let u′1 , . . . , u′n be a basis for U ′ , and let w1 , . . . , wp
be a basis for W . By hypothesis, we have

m ≥ (m + n) − p,

which implies p ≥ n. Thus we may define a linear map T ∈ L(V, W ) by its values
on the basis of V = U ⊕ U ′ by taking T uk = 0 for k = 1, . . . m and T u′j = wj
for j = 1, . . . , n (since p ≥ n, there is a wj for each u′j ). The map is linear by
Theorem 3.5, and its null space is U by construction.
(⇒) Suppose U is a subspace of V , T ∈ L(V, W ), and null T = U . Then,
since range T is a subspace of W , we have dim range T ≤ dim W . Combining
this inequality with the Fundamental Theorem of Linear Maps yields

dim null T = dim V − dim range T

≥ dim V − dim W.

Since dim null T = dim U , we have the desired inequality.

Problem 21
Suppose V is finite-dimensional and T ∈ L(V, W ). Prove that T is
surjective if and only if there exists S ∈ L(W, V ) such that T S is the
identity map on W .

Proof. (⇒) Suppose T ∈ L(V, W ) is surjective, so that W is necessarily finite-

dimensional as well. Let v1 , . . . , vm be a basis of V and let n = dim W , where
m ≥ n by surjectivity of T . Note that

T v1 , . . . , T vm

12
span W . Thus we may reduce this list to a basis by removing some elements
(possibly none, if n = m). Suppose this reduced list were T vi1 , . . . , T vin for some
i1 , . . . , in ∈ {1, . . . , m}. We define S ∈ L(W, V ) by its behavior on this basis

S(T vik ) := vik for k = 1, . . . , n.

Suppose w ∈ W . Then there exist a1 , . . . , an ∈ F such that

w = a1 T vi1 + · · · + an T vin

and thus

T S(w) = T S (a1 T vi1 + · · · + an T vin )

= T S (a1 T vi1 + · · · + an T vin )


= T a1 S(T vi1 ) + · · · + an S(T vin )
= T (a1 vi1 + · · · + an vin )
= a1 T vi1 + · · · + an T vin
= w,

and so T S is the identity map on W .

(⇐) Suppose there exists S ∈ L(W, V ) such that T S ∈ L(W, W ) is the
identity map, and suppose by way of contradiction that T is not surjective, so
that dim range T S < dim W . By the Fundamental Theorem of Linear Maps,
this implies

dim W = dim null T S + dim range T S

< dim null T S + dim W

and hence dim null T S > 0, a contradiction, since the identity map can only
have trivial null space. Thus T is surjective, as desired.

Problem 23
Suppose U and V are finite-dimensional vector spaces and S ∈ L(V, W )
and T ∈ L(U, V ). Prove that

dim range ST ≤ min{dim range S, dim range T }.

Proof. We will show that both dim range ST ≤ dim range S and dim range ST ≤
dim range T , since this implies the desired inequality.
We first show that dim range ST ≤ dim range S. Suppose w ∈ range ST .
Then there exists u ∈ U such that ST (u) = w. But this implies that w ∈
range S as well, since T u ∈ S −1 (w). Thus range ST ⊆ range S, which implies
dim range ST ≤ dim range S.
We now show that dim range ST ≤ dim range T . Note that if v ∈ null T , so
that T v = 0, then ST (v) = 0 (since linear maps take zero to zero). Thus we

13
have null T ⊆ null ST , which implies dim null T ≤ dim null ST . Combining this
inequality with the Fundamental Theorem of Linear Maps applied to T yields

dim U ≤ dim null ST + dim range T. (1)

Similarly, we have

dim U = dim null ST + dim range ST. (2)

Combining (1) and (2) yields

dim null ST + dim range ST ≤ dim null ST + dim range T

and hence dim range ST ≤ dim range T , completing the proof.

Problem 25
Suppose V is finite-dimensional and T1 , T2 ∈ L(V, W ). Prove that
range T1 ⊆ range T2 if and only if there exists S ∈ L(V, V ) such that
T1 = T2 S.

Proof. (⇐) Suppose there exists S ∈ L(V, V ) such that T1 = T2 S, and let
w ∈ range T1 . Then there exists v ∈ V such that T1 v = w, and hence T2 S(v) = w.
But then w ∈ range T2 as well, and hence range T1 ⊆ range T2 .
(⇒) Suppose range T1 ⊆ range T2 , and let v1 , . . . , vn be a basis of V . Let
wk = T vk for k = 1, . . . , n. Then there exist u1 , . . . , un ∈ V such that T2 uk = wk
for k = 1, . . . , n (since wk ∈ range T1 implies wk ∈ range T2 ). Define S ∈ L(V, V )
by its behavior on the basis

Svk := uk for k = 1, . . . , n.

It follows that T2 S(vk ) = T2 uk = wk = T1 vk . Since T2 S and T1 are equal on the

basis, they are equal as linear maps, as was to be shown.

Problem 27
Suppose p ∈ P(R). Prove that there exists a polynomial q ∈ P(R) such
that 5q ′′ + 3q ′ = p.

Proof. Suppose deg p = n, and consider the linear map

D : Pn+1 (R) → Pn (R)

q 7→ 5q ′′ + 3q ′ .

If we can show D is surjective, we’re done, since this implies that there exists
some q ∈ Pn+1 (R) such that Dq = 5q ′′ + 3q ′ = p. To that end, suppose
r ∈ null D. Then we must have r′′ = 0 and r′ = 0, which is true if and only if r

14
is constant. Thus any α ∈ R× is a basis of null D, and so dim null D = 1. By
the Fundamental Theorem of Linear Maps, we have

dim range D = dim Pn+1 (R) − dim null D,

and hence

dim range D = (n + 2) − 1 = n + 1.

Since the only subspace of Pn (R) with dimension n + 1 is the space itself, D is
surjective, as desired.

Problem 29
Suppose φ ∈ L(V, F). Suppose u ∈ V is not in null φ. Prove that

V = null φ ⊕ {au | a ∈ F}.

Proof. First note that since u ∈ V − null φ, there exists some nonzero φ(u) ∈
range φ and hence dim range φ ≥ 1. But since range φ ⊆ F, and dim F = 1, we
must have dim range φ = 1. Thus, letting n = dim V , it follows

dim null φ = dim V − dim range φ

= n − 1.

Let v1 , . . . , vn−1 be a basis for null φ. We claim v1 , . . . , vn−1 , u is an extension

of this basis to a basis of V , which would then imply V = null φ ⊕ {au | a ∈ F},
as desired.
To show v1 , . . . , vn−1 , u is a basis of V , it suffices to show linearly independence
(since it has length n = dim V ). So suppose a1 , . . . , an ∈ F are such that

a1 v1 + · · · + an−1 vn−1 + an u = 0.

We may write
an u = −a1 v1 − · · · − an−1 vn−1 ,
which implies an u ∈ null φ. By hypothesis, u ̸∈ null φ, and thus we must
have an = 0. But now each of the a1 , . . . , an−1 must be 0 as well (since
v1 , . . . , vn−1 form a basis of null φ and thus are linearly independent). Therefore,
v1 , . . . , vn−1 , u is indeed linearly independent, proving our claim.

Problem 31
Give an example of two linear maps T1 and T2 from R5 to R2 that have
the same null space but are such that T1 is not a scalar multiple of T2 .

15
Proof. Let e1 , . . . , e5 be the standard basis of R5 . We define T1 , T2 ∈ L(R5 , R2 )
by their behavior on the basis (using the standard basis for R2 as well)

T1 e1 := e2
T1 e2 := e1
T1 ek := 0 for k = 3, 4, 5

and

T2 e1 := e1
T2 e2 := e2
T2 ek := 0 for k = 3, 4, 5.

Clearly null T1 = null T2 . We claim T2 is not a scalar multiple of T1 . To see this,

suppose not. Then there exists α ∈ R such that T1 = αT2 . In particular, this
implies T1 e1 = αT2 e1 . But this is absurd, since T1 e1 = e2 and T2 e1 = e1 , and of
course e1 , e2 is linearly independent. Thus no such α can exist, and T1 , T2 are
as desired.

C: Matrices
Problem 1
Suppose V and W are finite-dimensional and T ∈ L(V, W ). Show that
with respect to each choice of bases of V and W , the matrix of T has at
least dim range T nonzero entries.

Proof. Let v1 , . . . , vn be a basis of V , let w1 , . . . , wm be a basis of W , let

r = dim range T , and let s = dim null T . Then there are s basis vectors of V
which map to zero and r basis vectors of V with nontrivial representation as linear
combinations of w1 , . . . , wm . That is, suppose T vk ̸= 0, where k ∈ {1, . . . , n}.
Then there exist a1 , . . . , am ∈ F, not all zero, such that

T vk = a1 w1 + · · · + am wm .

The coefficients form column k of M(T ), and there are r such vectors in the
basis of V . Hence there are r columns of M(T ) with at least one nonzero entry,
as was to be shown.

Problem 3
Suppose V and W are finite-dimensional and T ∈ L(V, W ). Prove that
there exist a basis of V and a basis of W such that with respect to these
bases, all entries of M(T ) are 0 except the entries in row j, column j,
equal 1 for 1 ≤ j ≤ dim range T .

16
Proof. Let R be the subspace of V such that
V = R ⊕ null T,
let r1 , . . . , rm be a basis of R (where m = dim range T ), and let v1 , . . . , vn be
a basis of null T (where n = dim null T ). Then r1 , . . . , rm , v1 , . . . , vn is a basis
of V . It follows that T r1 , . . . , T rm is a basis of range T , and hence there is an
extension of this list to a basis of W . Suppose T r1 , . . . , T rm , w1 , . . . , wp is such
an extension (where p = dim W − m). Then, for j = 1, . . . m, we have
   
X m p
X
T rj =  δi,j · T rt  +  0 · wk  ,
i=1 k=1

where δi,j is the Kronecker delta function. Thus, column j of M(T ) is has an
entry of 1 in row j and 0’s elsewhere, where j ranges over 1 to m = dim range T .
Since T v1 = · · · = T vn = 0, the remaining columns of M(T ) are all zero. Thus
M(T ) has the desired form.

Problem 5
Suppose w1 , . . . , wn is a basis of W and V is finite-dimensional. Suppose
T ∈ L(V, W ). Prove that there exists a basis v1 , . . . , vm of V such that all
the entries in the first row of M(T ) (with respect to the bases v1 , . . . , vm
and w1 , . . . , wn ) are 0 except for possibly a 1 in the first row, first column.

Proof. First note that if range T ⊆ span(w2 , . . . , wn ), the first row of M(T ) will
be all zeros regardless of choice of basis for V .
So suppose range T ̸⊆ span(w2 , . . . , wn ) and let u1 ∈ V be such that T u1 ̸∈
span(w2 , . . . , wn ). There exist a1 , . . . , an ∈ F such that
T u1 = a1 w1 + · · · + an wn ,
and notice a1 ̸= 0 since T u1 ̸∈ span(w2 , . . . , wn ). Hence we may define
1
z1 := u1 .
a1
It follows
a2 an
T z1 = w1 + w2 + · · · + wn . (3)
a1 a1
Now extend z1 to a basis z1 , . . . , zm of V . Then for k = 2, . . . , m, there exist
A1,k , . . . , An,k ∈ F such that
T zk = A1,k w1 + · · · + An,k wn ,
and notice
T (zk − A1,k z1 ) = T zk − A1,k T z1
 

a2 an
= A1,k w1 + · · · + An,k wn − A1,k w1 + w2 + · · · + wn
a1 a1

a2
an
= A2,k − A1,k w2 + · · · + An,k − A1,k wn . (4)
a1 a1

17
Now we define a new list in V by
(
z1 if k = 1
vk :=
zk − A1,k z1 otherwise

for k = 1, . . . , m. We claim v1 , . . . , vm is a basis. To see this, it suffices to

prove the list is linearly independent, since its length equals dim V . So suppose
b1 , . . . , bm ∈ F are such that

b1 v1 + · · · + bm vm = 0.

By definition of the vk , it follows

b1 z1 + b2 (z2 − A1,k z1 ) + · · · + bm (zm − A1,k z1 ) = 0.

But since z1 , . . . , zm is a basis of V , the expression on the LHS above is simply a

linear combination of vectors in a basis. Thus we must have b1 = · · · = bm = 0,
and indeed v1 , . . . , vm are linearly independent, as claimed.
Finally, notice (3) tells us the first column of M(T, vk , wk ) is all 0’s except a
1 in the first entry, and (4) tells us the remaining columns have a 0 in the first
entry. Thus M(T, vk , wk ) has the desired form, completing the proof.

Problem 7
Suppose S, T ∈ L(V, W ). Prove that M(S + T ) = M(S) + M(T ).

Proof. Let v1 , . . . , vm be a basis of V and let w1 , . . . , wn be a basis of W . Also,

let A = M(S) and B = M(T ) be the matrices of these linear transformations
with respect to these bases. It follows

(S + T )vk = Svk + T vk



= A1,k w1 + · · · + An,k wn + B1,k w1 + · · · + Bn,k wn
= (A1,k + B1,k )w1 + · · · + (An,k + Bn,k )wn .

Hence M(S+T )j,k = Aj,k +Bj,k , and indeed we have M(S+T ) = M(S)+M(T ),
as desired.

Problem 9
 
c1
.
Suppose A is an m-by-n matrix and c =  .. 

 is an n-by-1 matrix.
cn
Prove that
Ac = c1 A·,1 + · · · + cn A·,n .

18
Proof. By definition, it follows
  
A1,1 A1,2 . . . A1,n c1
 A2,1 A2,2 . . . A2,n   c2 
  
Ac =  .. .. ..  . 
..   
. . . .   .. 


Am,1 Am,2 . . . Am,n cn
 
A1,1 c1 + A1,2 c2 + · · · + A1,n cn
 A2,1 c1 + A2,2 c2 + · · · + A2,n cn 
 
=  .. 
.

 
Am,1 c1 + Am,2 c2 + · · · + Am,n cn
     
A1,1 A1,2 A1,n
 A2,1   A2,2   A2,n 
     
= c1 
 .. 
 + c 2 . 
  + · · · + c n . 
 
 .   ..   .. 
Am,1 Am,2 Am,n
= c1 A·,1 + · · · + cn A·,n ,
as desired.

Problem 11
Suppose a = (a1 , . . . , an ) is a 1-by-n matrix and C is an n-by-p matrix.
Prove that
aC = a1 C1,· + · · · + an Cn,· .

Proof. By definition, it follows

 
C1,1 C1,2 ... C1,p
C2,1 C2,2 ... C2,p 
 

aC = (a1 , . . . , an )  .. .. .. .. 

 . . . . 

Cn,1 Cn,2 ... Cn,p
 
n
X n
X n
X
= ak Ck,1 , ak Ck,2 , . . . , ak Ck,p 
k=1 k=1 k=1
n
X

= ak Ck,1 , . . . , ak Ck,p
k=1
Xn


= ak Ck,1 , . . . , Ck,p
k=1
Xn
= ak Ck,· ,
k=1

as desired.

19
 Problem 13
Prove that the distributive property holds for matrix addition and matrix
multiplication. In other words, suppose A, B, C, D, E, and F are matrices
whose sizes are such that A(B + C) and (D + E)F make sense. Prove that
AB + AC and DF + EF both make sense and that A(B + C) = AB + AC
and (D + E)F = DF + EF .

Proof. First note that if A(B + C) makes sense, then the number of columns of
A must equal the number of rows of B + C. But the sum of two matrices is only
defined if their dimensions are equal, and hence the number of rows of both B
and C must equal the number of columns of A. Thus AB + AC makes sense.
So suppose A ∈ Fm,n and B, C ∈ Fn,p . It follows
n
X


A(B + C) j,k = Aj,r (B + C)r,k
r=1
n
X
= Aj,r (Br,k + Cr,k )
r=1
Xn


= Aj,r Br,k + Aj,r Cr,k
r=1
n
X n
X
= Aj,r Br,k + Aj,r Cr,k
r=1 r=1
= (AB)j,k + (AC)j,k ,

proving the first distributive property.

Now note that if (D + E)F makes sense, then the number of columns of
D + E must equal the number of rows of F . Hence the number of columns of
both D and E must equal the number of rows of F , and thus DF + EF makes
sense as well. So suppose D, E ∈ Fm,n and F ∈ Fn,p . It follows
n
X


(D + E)F j,k
= (D + E)j,r Fr,k
r=1
n
X
= (Dj,r + Ej,r )Fr,k
r=1
n
X
= Dj,r Fr,k + Ej,r Fr,k
r=1
Xn n
X
= Dj,r Fr,k + Ej,r Fr,k
r=1 r=1
= (DF )j,k + (EF )j,k ,

proving the second distributive property.

20
 Problem 15
Suppose A is an n-by-n matrix and 1 ≤ j, k ≤ n. show that the entry in
row j, column k, of A3 (which is defined to mean AAA) is
n X
X n
Aj,p Ap,r Ar,k .
p=1 r=1

Proof. For 1 ≤ p, k ≤ n, we have

n
X
(A2 )p,k = Ap,r Ar,k .
r=1

Thus, for 1 ≤ j, k ≤ n, it follows

n
X
(A3 )j,k = Aj,p (A2 )p,k
p=1
n
X n
X
= Aj,p Ap,r Ar,k
p=1 r=1
Xn Xn
= Aj,p Ap,r Ar,k ,
p=1 r=1

as desired.

D: Invertibility and Isomorphic Vector Spaces

Problem 1
Suppose T ∈ L(U, V ) and S ∈ L(V, W ) are both invertible linear maps.
Prove that ST ∈ L(U, W ) is invertible and that (ST )−1 = T −1 S −1 .

Proof. For all u ∈ U , we have

(T −1 S −1 ST )(u) = T −1 (S −1 (S(T (u))))
= T −1 (I(T (u))
= T −1 (T (u))
=v
−1 −1
and hence T S is a left inverse of ST . Similarly, for all w ∈ W , we have
(ST T −1 S −1 )(w) = S(T (T −1 (S −1 (w))))
= S(I(S −1 (w)))
= S(S −1 (w))
=w

21
and hence T −1 S −1 is a right inverse of ST . Therefore, ST is invertible, as
desired.

Problem 3
Suppose V is finite-dimensional, U is a subspace of V , and S ∈ L(U, V ).
Prove there exists an invertible operator T ∈ L(V ) such that T u = Su
for every u ∈ U if and only if S is injective.

Proof. (⇐) Suppose S is injective, and let W be the subspace of V such that
V = U ⊕ W . Let u1 , . . . , um be a basis of U and let w1 , . . . , wn be a basis of W ,
so that u1 , . . . , um , w1 , . . . , wn is a basis of V . Define T ∈ L(V ) by its behavior
on this basis of V

T uk := Suk
T wj := wj

for k = 1, . . . , m and j = 1, . . . , n. Since S is injective, so too is T . And since V

is finite-dimensional, this implies that T is invertible, as desired.
(⇒) Suppose there exists an invertible operator T ∈ L(V ) such that T u = Su
for every u ∈ U . Since T is invertible, it is also injective. And since T is injective,
so to is S = T |U , completing the proof.

Problem 5
Suppose V is finite-dimensional and T1 , T2 ∈ L(V, W ). Prove that
range T1 = range T2 if and only if there exists an invertible operator
S ∈ L(V ) such that T1 = T2 S.

Proof. (⇒) Suppose range T1 = range T2 := R, so that null T1 = null T2 := N as

well. Let Q be the unique subspace of V such that

V = N ⊕ Q,

and let u1 , . . . , um be a basis of N and v1 , . . . , vn be a basis of Q. We claim

there exists a unique qk ∈ Q such that T2 qk = T1 vk for k = 1, . . . , n. To see this,
suppose qk , qk′ ∈ Q are such that T2 qk = T2 qk′ = T1 vk . Then T2 (qk − qk′ ) = 0,
and hence qk − qk′ ∈ N . But since N ∩ Q = {0}, this implies qk − qk′ = 0 and
thus qk = qk′ . And so the choice of qk is indeed unique. We now define S ∈ L(V )
by its behavior on the basis

Suk = uk for k = 1, . . . , m
Svj = qj for j = 1, . . . , n.

Let v ∈ V , so that there exist a1 , . . . , am , b1 , . . . , bn ∈ F such that

v = a1 u1 + · · · + am um + b1 v1 + · · · + bn vn .

22
It follows

(T2 S)(v) = T2 (S(a1 u1 + · · · + am um + b1 v1 + · · · + bn vn ))

= T2 (a1 Su1 + · · · + am Sum + b1 Sv1 + · · · + bn Svn )
= T2 (a1 u1 + · · · + am um + b1 q1 + · · · + bn qn )
= a1 T2 u1 + · · · + am T2 um + b1 T2 q1 + · · · + bn T2 qn
= b1 T1 v1 + · · · + bn T1 vn .

Similarly, we have

T1 v = T1 (a1 u1 + · · · + am um + b1 v1 + · · · + bn vn )
= a1 T1 u1 + · · · + am T1 um + b1 T1 v1 + · · · + bn T1 vn
= b1 T1 v1 + · · · + bn T1 vn ,

and so indeed T1 = T2 S. To see that S is invertible, it suffices to prove it is

injective. So let v ∈ V be as before, and suppose Sv = 0. It follows

Sv = S(a1 u1 + · · · + am um + b1 v1 + · · · + bn vn )
= (a1 u1 + · · · + am um ) + (b1 Sv1 + · · · + bn Svn )
= 0.

By the proof of Theorem 3.22, Sv1 , . . . , Svn is a basis of R, and thus the list
u1 , . . . , um , Sv1 , . . . , Svn is a basis of V , and each of the a’s and b’s must be zero.
Therefore S is indeed injective, completing the proof in this direction.
(⇐) Suppose there exists an invertible operator S ∈ L(V ) such that T1 = T2 S.
If w ∈ range T1 , then there exists v ∈ V such that T1 v = w, and hence (T2 S)(v) =
T2 (S(v)) = w, so that w ∈ range T2 and we have range T1 ⊆ range T2 . Conversely,
suppose w′ ∈ range T2 , so that there exists v ′ ∈ V such that T2 v ′ = w′ . Then,
since T2 = T1 S −1 , we have (T1 S −1 )(v ′ ) = T1 (S −1 (v ′ )) = w′ , so that w′ ∈
range T1 . Thus range T2 ⊆ range T1 , and we have shown range T1 = range T2 , as
desired.

Problem 7
Suppose V and W are finite-dimensional. Let v ∈ V . Let

E = {T ∈ L(V, W ) | T v = 0}.

(a) Show that E is a subspace of L(V, W ).

(b) Suppose v ̸= 0. What is dim E?

Proof. (a) First note that the zero map is clearly an element of E, and hence
E contains the additive identity of L(V, W ). Now suppose T1 , T2 ∈ E.
Then

(T1 + T2 )(v) = T1 v + T2 v = 0

23
 and hence T1 + T2 ∈ E, so that E is closed under addition. Finally, suppose
T ∈ E and λ ∈ F. Then

(λT )(v) = λT v = λ0 = 0,

and so E is closed under scalar multiplication as well. Thus E is indeed a

subspace of L(V, W ).
(b) Suppose v ̸= 0, and let dim V = m and dim W = n. Extend v to a basis
v, v2 , . . . , vm of V , and endow W with any basis. Let E denote the subspace
of Fm,n of matrices whose first column is all zero.
We claim T ∈ E if and only if M(T ) ∈ E, so that M : E → E is an
isomorphism. Clearly if T ∈ E (so that T v = 0), then M(T )·,1 is all zero,
and hence T ∈ E. Conversely, suppose M(T ) ∈ E. It follows

M(T v) = M(T )M(v)

  
0 A1,2 ... A1,n 1
0 A2,2 ... A2,n  0
  
= .. .. .. ..  .
 
. . . .   .. 
0 Am,2 ... Am,n 0
 
0
0
 
=
 ..  ,

.
0

and thus we must have T v = 0 so that T ∈ E, proving our claim. So

indeed E ∼
= E.
Now note that E has as a basis the set of all matrices with a single 1 in
a column besides the first, and zeros everywhere else. There are mn − n
such matrices, and hence dim E = mn − n. Thus we have dim E = mn − n
as well, as desired.

Problem 9
Suppose V is finite-dimensional and S, T ∈ L(V ). Prove that ST is
invertible if and only if both S and T are invertible.

Proof. (⇐) Suppose S and T are both invertible. Then by Problem 1, ST is

invertible.
(⇒) Suppose ST is invertible. We will show T is injective and S is surjective.
Since V is finite-dimensional, this implies that both S and T are invertible. So
suppose v1 , v2 ∈ V are such that T v1 = T v2 . Then (ST )(v1 ) = (ST )(v2 ), and
since ST is invertible (and hence injective), we must have v1 = v2 , so that T
is injective. Next, suppose v ∈ V . Since T −1 is surjective, there exists w ∈ V

24
such that T −1 w = v. And since ST is surjective, there exists p ∈ V such that
(ST )(p) = w. It follows that (ST T −1 )(p) = T −1 (w), and hence Sp = v. Thus S
is surjective, completing the proof.

Problem 11
Suppose V is finite-dimensional and S, T, U ∈ L(V ) and ST U = I. Show
that T is invertible and that T −1 = U S.

Proof. Notice ST U is invertible since ST U = I and I is invertible. By Problem

9, we have that (ST )U is invertible if and only if ST and U are invertible. By
a second application of the result, ST is invertible if and only if S and T are
invertible. Thus S, T, and U are all invertible. To see that T −1 = U S, notice
U S = (T −1 T )U S
= T −1 (S −1 S)T U S
= T −1 S −1 (ST U )S
= T −1 S −1 S
= T −1 ,
as desired.

Problem 13
Suppose V is a finite-dimensional vector space and R, S, T ∈ L(V ) are
such that RST is surjective. Prove that S is injective.

Proof. Since V is finite-dimensional and RST is surjective, RST is also invertible.

By Problem 9, we have that (RS)T is invertible if and only if RS and T are
invertible. By a second application of the result, RS is invertible if and only if
R and S are invertible. Thus R, S, and T are all invertible, and hence injective.
In particular, S is injective, as desired.

Problem 15
Prove that every linear map from Fn,1 to Fm,1 is given by a matrix
multiplication. In other words, prove that if T ∈ L(Fn,1 , Fm,1 ), then
there exists an m-by-n matrix A such that T x = Ax for every x ∈ Fn,1 .

Proof. Endow both Fn,1 and Fm,1 with the standard basis, and let T ∈
L(Fn,1 , Fm,1 ). Let A = M(T ) with respect to these bases, and note that if
x ∈ Fn,1 , then M(x) = x (and similarly if y ∈ Fm,1 , then M(y) = y). Hence
T x = M(T x)
= M(T )M(x)
= Ax,
as desired.

25
 Problem 16
Suppose V is finite-dimensional and T ∈ L(V ). Prove that T is a scalar
multiple of the identity if and only if ST = T S for every S ∈ L(V ).

Proof. (⇒) Suppose T = λI for some λ ∈ F, and let S ∈ L(V ) be arbitrary. For
any v ∈ V , we have ST v = S(λI)v = λSv and T Sv = (λI)Sv = λSv, and hence
ST = T S. Since S was arbitrary, we have the desired result.
(⇐) Suppose ST = T S for every S ∈ L(V ), and let v ∈ V be arbitrary.
Consider the list v, T v. We claim it is linearly dependent. To see this, suppose
not. Then v, T v can be extended to a basis v, T v, u1 , . . . , un of V . Define
S ∈ L(V ) by
S(αv + βT v + γ1 u1 + · · · + γn un ) = βv,
where α, β, γ1 , . . . , γn are the unique coefficients of our basis for the given input
vector. In particular, notice S(T v) = v and Sv = 0. Thus ST v = T Sv implies
v = T (0) = 0, contradicting our assumption that v, T V is linearly independent.
So v, T v must be linearly dependent, and so for for all v ∈ V there exists λv ∈ F
such that T v = λv v (where λ0 can be any nonzero element of F, since T 0 = 0).
We claim λv is independent of the choice of v for v ∈ V − {0}, hence T v = λv
for all v ∈ V (including v = 0) and some λ ∈ F, and thus T = λI.
So suppose w, z ∈ V − {0} are arbitrary. We want to show λw = λz . If w and
z are linearly dependent, then there exists α ∈ F such that w = αz. It follows

λw w = T w
= T (αz)
= αT z
= αλz z
= λz (αz)
= λz w.

Since w ̸= 0, this implies λw = λz . Next suppose w and z are linearly independent.

Then we have

λw+z (w + z) = T (w + z)
= Tw + Tz
= λw w + λz z,

and hence

(λw+z − λw )w + (λw+z − λz )z = 0.

Since w and z are assumed to be linearly independent, we have λw+z = λw and

λw+z = λz , and hence again we have λw = λz , completing the proof.

26
 Problem 17
Suppose V is finite-dimensional and E is a subspace of L(V ) such that
ST ∈ E and T S ∈ E for all S ∈ L(V ) and all T ∈ E. Prove that E = {0}
or E = L(V ).

Proof. If E = {0}, we’re done. So suppose E ̸= {0}. If dim V = n, then

L(V ) ∼= Fn,n , and so there exists an isomorphic subspace E := M(E) ⊆ Fn,n
with the property that AB ∈ E and BA ∈ E for all A ∈ Fn,n and all B ∈ E. It
suffices to show E = Fn,n .
Define E i,j to be the matrix which is 1 in row i and column j and 0 everywhere
else, and let A ∈ Fn,n be nonzero. Then there exists some 1 ≤ j, k ≤ n such
that Aj,k ̸= 0. Notice for 1 ≤ i, j, r, s ≤ n, we have E i,j A ∈ E, and hence
E i,j AE r,s ∈ E. This product has the form

E i,j AE k,ℓ = Aj,k · E i,ℓ .

In other words, E i,j AE k,ℓ takes Aj,k and puts it in the ith row and ℓth column
of a matrix which is 0 everywhere else. Since E is closed under addition, this
implies

E 1,j AE k,1 + E 2,j AE k,2 + · · · + E n,j AE k,n = Aj,k · I ∈ E.

But since E is closed under scalar multiplication, and Aj,k ̸= 0, we have

!
1
· Aj,k · I = I ∈ E.
Aj,k

Since E contains I, by our characterization of E it must also contain every

element of Fn,n . Thus E = Fn,n , and since E ∼
= E, we must have E = L(V ), as
desired.

Problem 19
Suppose T ∈ L(P(R)) is such that T is injective and deg T p ≤ deg p for
every nonzero polynomial p ∈ P(R).
(a) Prove that T is surjective.

(b) Prove that deg T p = deg p for every nonzero p ∈ P(R).

Proof. (a) Let q ∈ P(R), and suppose deg q = n. Let Tn = T |Pn (R) , so
that Tn is the restriction of T to a linear operator on Pn (R). Since T
is injective, so is Tn . And since Tn is an injective linear operator over a
finite-dimensional vector space, Tn is surjective as well. Thus there exists
r ∈ Pn (R) such that Tn r = q, and so we have T r = q as well. Therefore T
is surjective.

27
 (b) We induct on the degree of the restriction maps Tn ∈ L(Pn (R)), each of
which is bijective by (a). Let P (k) be the statement: deg Tk p = k for every
nonzero p ∈ Pk (R).
Base case: Suppose p ∈ P0 (R) is nonzero. Since T0 is a bijective , T0 p = 0
iff p = 0 (the zero polynomial), which is the only polynomial with degree
< 0. Since p is nonzero by hypothesis, we must have deg T0 p = 0. Hence
P (0) is true.
Inductive step: Let n ∈ Z+ , and suppose P (k) is true for all 0 ≤ k < n.
Let p ∈ Pn (R) be nonzero. If deg Tn p < n, then for some k < n there
exists q ∈ Pk (R) and Tk ∈ P(R) such that Tk q = p (since Tk is surjective).
Hence T q = T p, a contradiction since deg p = ̸ deg q and T is injective.
Thus we must have deg Tn p = n, and P (n) is true.
By the principle of mathematical induction, P (k) is true for all k ∈ Z≥0 .
Hence deg T p = deg p for all nonzero p ∈ P(R), since T p = Tk p for
k = deg p.

E: Products and Quotients of Vector Spaces

Problem 1
Suppose T is a function from V to W . The graph of T is the subset of
V × W defined by

graph of T = {(v, T v) ∈ V × W | v ∈ V }.

Prove that T is a linear map if and only if the graph of T is a subspace

of V × W .

Proof. Define G := {(v, T v) ∈ V × W | v ∈ V }.

(⇒) Suppose T is a linear map. Since T is linear, T 0 = 0, and hence (0, 0) ∈ G,
so that G contains the additive identity. Next, let (v1 , T v1 ), (v2 , T v2 ) ∈ G. Then

(v1 , T v1 ) + (v2 , T v2 ) = (v1 + v2 , T v1 + T v2 ) = (v1 + v2 , T (v1 + v2 )) ∈ G,

and G is closed under addition. Lastly, let λ ∈ F and (v, T v) ∈ G. It follows

λ(v, T v) = (λv, λT v) = (λv, T (λv)) ∈ G,

and G is closed under scalar multiplication. Thus G is a subspace of V × W .

(⇐) Suppose G is a subspace of V × W , and let (v1 , T v1 ), (v2 , T v2 ) ∈ G.
Since G is closed under addition, it follows

(v1 + v2 , T v1 + T v2 ) ∈ G,

and hence we must have T v1 + T v2 = T (v1 + v2 ), so that T is additive. And since

G is closed under scalar multiplication, for λ ∈ F and (v, T v) ∈ G, it follows

(λv, λT v) ∈ G,

28
and hence we must have λT v = T (λv), so that T is homogeneous. Therefore, T
is a linear map, as desired.

Problem 3
Give an example of a vector space V and subspaces U1 , U2 of V such that
U1 × U2 is isomorphic to U1 + U2 but U1 + U2 is not a direct sum.

Proof. Define the following two subspaces of P(R)

U1 := P(R)
U2 := R,

so that U1 ∩ U2 = R and the sum U1 + U2 = P(R) is not direct. Endow P(R)

and R with their standard bases, and define φ by its behavior on the basis of
U1 × U2

φ : U1 × U2 → U1 + U2
 
X k , 0 7→ X k+1
(0, 1) 7→ 1.

We claim φ is an isomorphism. To see that φ is injective, suppose

(a0 + a1 X + · · · + am X m , α) , (b0 + b1 X + · · · + bn X n , β) ∈ U1 × U2

and
(a0 + a1 X + · · · + am X m , α) ̸= (b0 + b1 X + · · · + bn X n , β) .
We have

φ (a0 + a1 X + · · · + am X m , α) = α + a0 X + a1 X 2 + · · · + am X m+1 (5)

and

φ (b0 + b1 X + · · · + bn X n , β) = β + b0 X + b1 X 2 + · · · + bn X n+1 . (6)

Since α ̸= β, this implies (5) does not equal (6) and hence φ is injective. To see
that φ is surjective, suppose c0 + c1 X + · · · + cp X p ∈ U1 + U2 . Then
 
φ c1 + c2 X + · · · + cp X p−1 , c0 = c0 + c1 X + · · · + cp X p

and φ is indeed surjective.

Since φ an injective and surjective linear map, it is an isomorphism. Thus
U1 × U2 ∼
= U1 + U2 , as was to be shown.
Problem 5
Suppose W1 , . . . , Wm are vector spaces. Prove that L(V, W1 × · · · × Wm )
and L(V, W1 ) × · · · × L(V, Wm ) are isomorphic vector spaces.

29
Proof. Define the projection map πk for k = 1, . . . , m by
πk : W1 × · · · × Wm → Wk
(w1 , . . . , wm ) 7→ wk .
Clearly πk is linear. Now define
φ : L(V, W1 × · · · × Wm ) → L(V, W1 ) × · · · × L(V, Wm )
T 7→ (π1 T, . . . , πm T ).
To see that φ is linear, let T1 , T2 ∈ L(V, W1 × · · · × Wm ). It follows
φ(T1 + T2 ) = (π1 (T1 + T2 ), . . . , πm (T1 + T2 ))
= (π1 T1 + π1 T2 , . . . , πm T1 + πm T2 )
= (π1 T1 , . . . , πm T1 ) + (π1 T2 , . . . , πm T2 )
= φ(T1 ) + φ(T2 ),
and hence φ is additive. Now for λ ∈ F and T ∈ L(V, W1 × · · · × Wm ), we have
φ(λT ) = (π1 (λT ), . . . , πm (λT ))
= (λ(π1 T ), . . . , λ(πm T ))
= λ(π1 T, . . . , πm T ),
and thus φ is homogenous. Therefore, φ is linear.
We now show φ is an isomorphism. To see that it is injective, suppose
T ∈ L(V, W1 × · · · × Wm ) and φ(T ) = 0. Then
(π1 T, . . . , πm T ) = (0, . . . , 0)
which is true iff T is the zero map. Thus φ is injective. To see that φ is surjective,
suppose (S1 , . . . , Sm ) ∈ L(V, W1 ) × · · · × L(V, Wm ). Define
S : V → W1 × · · · × Wm
v 7→ (S1 v, . . . , Sm v),
so that φk S = Sk for k = 1, . . . , m. Then
φ(S) = (π1 S, . . . , πm S)
= (S1 , . . . , Sn )
and S is indeed surjective. Therefore, φ is an isomorphism, and we have
L(V, W1 × · · · × Wm ) ∼
= L(V, W1 ) × · · · × L(V, Wm ),
as desired.

Problem 7
Suppose v, x are vectors in V and U, W are subspaces of V such that
v + U = x + W . Prove that U = W .

30
Proof. First note that since v + 0 = v ∈ v + U , there exists w0 ∈ W such that
v = x + w0 , and hence v − x = w0 ∈ W . Similarly, there exists u0 ∈ U such that
x − v = u0 ∈ U .
Suppose u ∈ U . Then there exists w ∈ W such that v + u = x + w, and hence

u = (x − v) + w = −w0 + w ∈ W,

and we have U ⊆ W . Conversely, suppose w′ ∈ W . Then there exists u′ ∈ U

such that x + w′ = v + u′ , and hence

w′ = (v − x) + u′ = −u0 + u′ ∈ U,

and we have W ⊆ U . Therefore U = W , as desired.

Problem 8
Prove that a nonempty subset A of V is an affine subset of V if and only
if λv + (1 − λ)w ∈ A for all v, w ∈ A and all λ ∈ F.

Proof. (⇒) Suppose A ⊆ V is an affine subset of V . Then there exists x ∈ V

and a subspace U ⊆ V such that A = x + U . Suppose v, w ∈ A. Then there
exist u1 , u2 ∈ U such that v = x + u1 and w = x + u2 . Thus, for all λ ∈ F, we
have

λv + (1 − λ)w = λ(x + u1 ) + (1 − λ)(x + u2 )

= x + λu1 + (1 − λ)u2 .

Since λu1 + (1 − λ)u2 ∈ U , this implies v + (1 − λ)w ∈ x + U = A, as desired.

(⇐) Suppose λv + (1 − λ)w ∈ A for all v, w ∈ A and all λ ∈ F. Choose a ∈ A
and define
U := −a + A.
We claim U is a subspace of V . Clearly 0 ∈ U since a ∈ A. Let x ∈ U , so that
x = −a + a0 for some a0 ∈ A, and let λ ∈ F. It follows

λa0 + (1 − λ)a ∈ A ⇒ −λa + λa0 + a ∈ A ⇒ λ(−a + a0 ) ∈ −a + A = U,

and thus λx = λ(−a + a0 ) ∈ U , and U is closed under scalar multiplication. Now

let x, y ∈ U . Then there exist a1 , a2 ∈ A such that x = −a + a1 and y = −a + a2 .
Notice
 
1 1 1 1
a1 + 1 − a2 = a1 + a2 ∈ A,
2 2 2 2

and hence
1 1
−a + a1 + a2 ∈ U.
2 2

31
It follows

x + y = −2a + a1 + a2
 
1 1
= 2 −a + a1 + a2 ∈ U,
2 2
using the fact that U has already been shown to be closed under scalar multipli-
cation. Thus U is also closed under addition, and so U is a subspace of V . Now,
since A = a + U , we have that A is indeed an affine subset of V , as desired.

Problem 9
Suppose A1 and A2 are affine subsets of V . Prove that the intersection
A1 ∩ A2 is either an affine subset of V or the empty set.

Proof. If A1 ∩ A2 = ∅, we’re done, so suppose A1 ∩ A2 is nonempty and let

v ∈ A1 ∩ A2 . Then we may write

A1 = v + U1 and A2 = v + U2

for some subspaces U1 , U2 ⊆ V .

We claim A1 ∩ A2 = v + (U1 ∩ U2 ), which is an affine subset of V . To see
this, suppose x ∈ v + (U1 ∩ U2 ). Then there exists u ∈ U1 ∩ U2 such that
x = v + u. Since u ∈ U1 , we have x ∈ v + U1 = A1 . And since u ∈ U2 , we have
x ∈ v + U2 = A2 . Thus x ∈ A1 ∩ A2 and v + (U1 ∩ U2 ) ⊆ A1 ∩ A2 . Conversely,
suppose y ∈ A1 ∩ A2 . Then there exist u1 ∈ U1 and u2 ∈ U2 such that y = v + u1
and y = v + u2 . But this implies u1 = u2 , and hence u1 = u2 ∈ U1 ∩ U2 , thus
y ∈ v + (U1 ∩ U2 ). Therefore A1 ∩ A2 ⊆ v + (U1 ∩ U2 ), and hence we have
A1 ∩ A2 = v + (U1 ∩ U2 ), as claimed.

Problem 11
Suppose v1 , . . . , vm ∈ V . Let

A = {λ1 v1 + · · · + λm vm | λ1 , . . . , λm ∈ F and λ1 + · · · + λm = 1}.

(a) Prove that A is an affine subset of V .

(b) Prove that every affine subset of V that contains v1 , . . . , vm also

contains A.
(c) Prove that A = v + U for some v ∈ V and some subspace U of V
with dim U ≤ m − 1.

Proof. (a) Let v, w ∈ A, so that there exist α1 , . . . , αm ∈ F and β1 , . . . , βm ∈ F

such that

v = α1 v1 + · · · + αm vm
w = β1 v1 + · · · + βm vm ,

32
 P P
where αk = 1 and βk = 1. Given λ ∈ F, it follows
m
X m
X
λv + (1 − λ)w = λ αk vk + (1 − λ) βk vk
k=1 k=1
m
X  
= λαk + (1 − λ)βk vk .
k=1

But notice
m
X  
λαk + (1 − λ)βk = λ + (1 − λ) = 1,
k=1

and hence λv + (1 − λ)w ∈ A by the way we defined A. By Problem 8,

this implies that A is an affine subset of V , as was to be shown.
(b) We induct on m.
Base case: When m = 1, the statement is trivially true, since A = {v1 },
and hence any affine subset of V that contains v1 of course contains A.
Inductive step: Let k ∈ Z+ , and suppose the statement is true for m = k.
Suppose A′ is an affine subset of V that contains P v1 , . . . , vk+1 , and let
x ∈ A. Then there exist λ1 , . . . , λk+1 ∈ F such that j λj = 1 and

x = λ1 v1 + · · · + λk+1 vk+1 .

Now, if λk+1 = 1, then x = vk+1 ∈ A′ . Otherwise, we have

λ1 λk
+ ··· + = 1,
1 − λk+1 1 − λk+1

and hence by our inductive hypothesis, this implies

λ1 λk
v1 + · · · + v k ∈ A′ .
1 − λk+1 1 − λk+1

By Problem 8, we know
 
λ1 λk
(1 − λk+1 ) v1 + · · · + vk + λk+1 vk+1 ∈ A′ .
1 − λk+1 1 − λk+1

But after simplifying, this tells us

λ1 v1 + · · · + λk+1 vk+1 = x ∈ A′ .

Hence A ⊆ A′ , and the statement is true for m = k + 1.

By the principal of mathematical induction, the statement is true for all
m ∈ Z+ . Thus any affine subset of V that contains v1 , . . . , vm also contains
A, as was to be shown.

33
 (c) Define U := span(v2P − v1 , . . . , vm − v1 ). Let x ∈ A, so that there exist
λ1 , . . . , λm ∈ F with k λk = 1 such that

x = λ1 v1 + · · · + λm vm .

Notice
 
m
X
v1 + λ2 (v2 − v1 ) + · · · + λm (vm − v1 ) = 1 − λk  v1 + λ2 v2 + · · · + λm vm
k=2

= λ1 v1 + · · · + λm vm
= x,

and hence x ∈ v1 + U , so that A ⊆ v1 + U . Next suppose y ∈ v1 + U , so

that there exist α1 , . . . , αm−1 ∈ F such that

y = v1 + α1 (v2 − v1 ) + · · · + αm−1 (vm − v1 ).

Expanding the RHS yields

 
m−1
X
y = 1 − αk  v1 + α1 v2 + · · · + αm−1 vm .
k=1

But since  
m−1
X m−1
X
1 − αk  + αk = 1,
k=1 k=1

this implies y ∈ A, and hence v1 + U ⊆ A. Therefore A = v1 + U , and

since dim U ≤ m − 1, we have the desired result.

Problem 13
Suppose U is a subspace of V and v1 + U, . . . , vm + U is a basis of V /U
and u1 , . . . , un is a basis of U . Prove that v1 , . . . , vm , u1 , . . . , un is a basis
of V .

Proof. Since

dim V = dim V /U + dim U

= m + n,

it suffices to show v1 , . . . , vm , u1 , . . . , un spans V . Suppose v ∈ V . Then there

exist α1 , . . . , αm ∈ F such that

v + U = α1 (v1 + U ) + · · · + αm (vm + U ).

But then
v + U = (α1 v1 + · · · + αm vm ) + U

34
and hence
v − (α1 v1 + · · · + αm vm ) ∈ U.
Thus there exist β1 , . . . , βn ∈ U such that

v − (α1 v1 + · · · + αm vm ) = β1 u1 + · · · + βn un ,

and we have
v = α1 v1 + · · · + αm vm + β1 u1 + · · · + βn un ,
so that indeed v1 , . . . , vm , u1 , . . . , un spans V .

Problem 15
Suppose φ ∈ L(V, F) and φ ̸= 0. Prove that dim V /(null φ) = 1.

Proof. Since φ ̸= 0, we must have dim range φ = 1, so that range φ = F. Since

V /(null φ) ∼
= range φ, this implies V /(null φ) ∼
= F, and hence dim V /(null φ) = 1,
as desired.

Problem 17
Suppose U is a subspace of V such that V /U is finite-dimensional. Prove
that there exists a subspace W of V such that dim W = dim V /U and
V = U ⊕ W.

Proof. Suppose dim V /U = n, and let v1 + U, . . . , vn + U be a basis of V /U .

Define W := span(v1 , . . . , vn ). We claim v1 , . . . , vn must be linearly independent,
so that v1 , . . . , vn is a basis of W . To see this, suppose α1 , . . . , αn ∈ F are such
that
α1 v1 + · · · + αn vn = 0.
Then
(α1 v1 + · · · + αn vn ) + U = α1 (v1 + U ) + · · · + αn (vn + U ),
and hence we must have α1 = · · · = αn = 0. Thus v1 , . . . , vn is indeed linearly
independent, as claimed.
We now claim V = U ⊕ W . To see that V = U + W , suppose v ∈ V . Then
there exist β1 , . . . , βn ∈ F such that

v + U = β1 (v1 + U ) + · · · + βn (vn + U ).

It follows
n
X
v− βk vk ∈ U,
k=1

and hence    
n
X n
X
v = v − βk vk  +  β k vk  .
k=1 k=1

35
Since first term in parentheses is in U and the second term in parentheses is in
W , we have v ∈ U + W , and hence V ⊆ U + W . Clearly U + W ⊆ V , since U
and W are each subspaces of V , and hence V = U + W . To see that the sum
is direct, suppose w ∈ U ∩ W . Since w ∈ W , there exist λ1 , . . . , λn such that
w = λ1 v1 + · · · + λn vn , and hence

w + U = (λ1 v1 + · · · + λn vn ) + U
= λ1 (v1 + U ) + · · · + λn (vn + U ).

Since w ∈ U , we have w + U = 0 + U . Thus λ1 = · · · = λn = 0, which implies

w = 0. Since U ∩ W = {0}, the sum is indeed direct. Thus V = U ⊕ W , with
dim W = n = dim V /U , as desired.

Problem 19
Find a correct statement analogous to 3.78 that is applicable to finite
sets, with unions analogous to sums of subspaces and disjoint unions
analogous to direct sums.

Theorem. Suppose |V | < ∞ and U1 , . . . , Un ⊆ V . Then U1 , . . . , Un are pairwise

disjoint if and only if

|U1 ∪ · · · ∪ Un | = |U1 | + · · · + |Un |.

Proof. We induct on n.
Base case: Let n = 2. Since |U1 ∪ U2 | = |U1 | + |U2 | − |U1 ∩ U2 |, we have
U1 ∩ U2 = ∅ iff |U1 ∪ U2 | = |U1 | + |U2 |.
Inductive hypothesis: Let k ∈ Z≥2 , and suppose the statement is true for
n = k. Let Uk+1 ⊆ V . Then

|U1 ∪ · · · ∪ Uk+1 | = |U1 ∪ · · · ∪ Uk | + |Uk+1 |

iff Uk+1 ∩ (U1 ∪ · · · ∪ Uk ) = ∅ by our base case. Combining this with our inductive
hypothesis, we have

|U1 ∪ · · · ∪ Uk+1 | = |U1 | + · · · + |Uk | + |Uk+1 |

iff U1 , . . . , Uk+1 are pairwise disjoint, and the statement is true for n = k + 1.
By the principal of mathematical induction, the statement is true for all
n ∈ Z≥2 .

F: Duality
Problem 1
Explain why every linear functional is either surjective or the zero map.

36
Proof. Since dim F = 1, the only subspaces of F are F itself and {0}. Let V be
a vector space (not necessarily finite-dimensional) and suppose φ ∈ V ′ . Since
range φ is a subspace of F, it must be either F itself (in which case φ is surjective)
or {0} (in which case φ is the zero map).

Problem 3
Suppose V is finite-dimensional and U is a subspace of V such that
U ̸= V . Prove that there exists φ ∈ V ′ such that φ(u) = 0 for every
u ∈ U but φ ̸= 0.

Proof. Suppose dim U = m and dim V = n for some m, n ∈ Z+ such that m < n.
Let u1 , . . . , um be a basis of U . Expand this to a basis u1 , . . . , um , um+1 , . . . , un
of V , and let φ1 , . . . , φn be the corresponding dual basis of V ′ . For any u ∈ U ,
there exist α1 , . . . , αm such that u = α1 u1 + · · · + αm um . Now notice

φm+1 (u) = φm+1 (α1 u1 + · · · + αm um )

= α1 φm+1 (u1 ) + · · · + αm φm+1 (um )
= 0,

but φm+1 (um+1 ) = 1. Thus φm+1 (u) = 0 for every u ∈ U but φm+1 ̸= 0, as
desired.

Problem 5
Suppose V1 , . . . , Vm are vector spaces. Prove that (V1 × · · · × Vm )′ and
V1′ × · · · × Vm′ are isomorphic vector spaces.

Proof. For i = 1, . . . , m, let

ξi : V i → V 1 × · · · × V m
vi 7→ (0, . . . , vi , . . . , 0).

Now define

T : (V1 × · · · × Vm )′ → V1′ × · · · × Vm′

φ 7→ (φ ◦ ξ1 , . . . , φ ◦ ξm ) .

We claim T is an isomorphism. We must show three things: (1) that T is a

linear map; (2) that T is injective; and (3) that T is surjective.
To see that T is a linear map, first suppose φ1 , φ2 ∈ (V1 × · · · × Vm )′ . It
follows


T (φ1 + φ2 ) = (φ1 + φ2 ) ◦ ξ1 , . . . , (φ1 + φ2 ) ◦ ξm
= (φ1 ◦ ξ1 + φ2 ◦ ξ1 , . . . , φ1 ◦ ξm + φ2 ◦ ξm )
= (φ1 ◦ ξ1 , . . . , φ1 ◦ ξm ) + (φ2 ◦ ξ1 , . . . , φ2 ◦ ξm )
= T (φ1 ) + T (φ2 ),

37
thus T is additive. To see that it is also homogeneous, suppose λ ∈ F and
φ ∈ (V1 × · · · × Vm )′ . We have


T (λφ) = (λφ) ◦ ξ1 , . . . , (λφ) ◦ ξm


= λ(φ ◦ ξ1 ), . . . , λ(φ ◦ ξm )
= λ (φ ◦ ξ1 , . . . , φ ◦ ξm )
= λT (φ),

and thus T is homogeneous as well and therefore it is a linear map.

To see that T is injective, suppose φ, ψ ∈ (V1 × · · · × Vm )′ but φ ̸= ψ.
Then there exists some (v1 , . . . , vm ) ∈ V1 × · · · × Vm such that φ(v1 , . . . , vm ) ̸=
ψ(v1 , . . . , vm ). Since φ and ψ are linear, this means that there exists some index
k ∈ {1, . . . , m} such that φ(0, . . . , vk , . . . , 0) ̸= ψ(0, . . . , vk , . . . , 0). But then
φ ◦ ξk ̸= ψ ◦ ξk , and hence T (φ) ̸= T (ψ), so that T is injective.
To see that T is surjective, suppose (φ1 , . . . , φm ) ∈ V1′ × · · · × Vm′ and define

θ : V1 × · · · × Vm → F
m
X
(v1 , . . . , vm ) 7→ φk (vk ).
k=1

We claim T (θ) = (φ1 , . . . , φm ). To see this, let k ∈ {1, . . . , m}. We will show
that the map in the k-th component of T (θ) is equal to φk . Given vk ∈ Vk , we
have

T (θ)k (vk ) = (θ ◦ ξk )(vk )

= θ(ξk (vk ))
= θ(0, . . . , vk , . . . , 0)
= φ1 (0) + · · · + φk (vk ) + · · · + φm (0)
= φk (vk ),

as desired. Thus T (θ) = (φ1 , . . . , φm ), and T is indeed surjective. Since T is

both injective and surjective, it’s an isomorphism.

Problem 7
Suppose m is a positive integer. Show that the dual basis of the basis
(j)
1, . . . , xm of Pm (R) is φ0 , φ1 , . . . , φm , where φj (p) = p j!(0) . Here p(j)
denotes the j th derivative of p, with the understanding that the 0th
derivative of p is p.

Proof. For j = 0, . . . , m, we have by direct computation of the j-th derivative

(
 
k 1 if j = k
φj x =
0 otherwise,

38
so that φ0 , φ1 , . . . , φm is indeed the dual basis of 1, . . . , xm . Note the uniqueness
of the dual basis follows by uniqueness of a linear map (including the linear
functionals in the dual basis) whose values on a basis are specified.

Problem 9
Suppose v1 , . . . , vn is a basis of V and φ1 , . . . , φn is the corresponding
dual basis of V ′ . Suppose ψ ∈ V ′ . Prove that

ψ = ψ(v1 )φ1 + · · · + ψ(vn )φn .

Proof. Let α1 , . . . , αn ∈ F be such that

ψ = α1 φ1 + · · · + αn φn .

For k = 1, . . . , n, we have

ψ(vk ) = α1 φ1 (vk ) + · · · + αk φk (vk ) + · · · + αn φn (vk )

= α1 · 0 + · · · + αk · 1 + · · · + αn · 0
= αk .

Thus we have
ψ = ψ(v1 )φ1 + · · · + ψ(vn )φn ,
as desired

Problem 11
Suppose A is an m-by-n matrix with A ̸= 0. Prove that the rank of A is
1 if and only if there exist (c1 , . . . , cm ) ∈ Fm and (d1 , . . . , dn ) ∈ Fn such
that Aj,k = cj dk for every j = 1, . . . , m and every k = 1, . . . , n.

Proof. (⇒) Suppose the rank of A is 1. By the assumption that A ̸= 0, there

exists a nonzero entry Ai.,j for some i ∈ {1, . . . , m} and j ∈ {1, . . . , n}. Thus
span{A·,1 , . . . , A·,n } = span{A·,j }, and hence there exist α1 , . . . , αn ∈ F such
that A·,c = αc A·,j for c = 1, . . . , n. Expanding out each out these columns, we
have

Ar,c = αc Ar,j (7)

for r = 1, . . . , m. Similarly for the rows, we have span{A1,· , . . . , Am,· } =

span{Ai,· }, and hence there exist β1 , . . . , βm ∈ F such that Ar′ ,· = βr Ai,· for
r′ = 1, . . . , m. Expanding out each of these rows, we have

Ar′ ,c′ = βr′ Ai,c′ (8)

for c′ = 1, . . . , n. Now by replacing the Ar,j term in (7) according to (8), we

have Ar,j = βr Ai,j , and hence (7) may be rewritten

Ar,c = αc βr Ai,j ,

39
and the result follows by defining cr = βr Ai,j and dc = αc for r = 1, . . . , m and
c = 1, . . . , n.
(⇐) Suppose there exist (c1 , . . . , cm ) ∈ Fm and (d1 , . . . , dn ) ∈ Fn such that
Aj,k = cj dk for every j = 1, . . . , m and every k = 1, . . . , n. Then each of the
columns is a scalar multiple of (d1 , . . . , dn )t ∈ Fn,1 and the column rank is 1.
Since the rank of a matrix equals its column rank, the rank of A is 1 as well.

Problem 13
Define T : R3 → R2 by T (x, y, z) = (4x + 5y + 6z, 7x + 8y + 9z). Suppose
φ1 , φ2 denotes the dual basis of the standard basis of R2 and ψ1 , ψ2 , ψ3
denotes the dual basis of the standard basis of R3 .
(a) Describe the linear functionals T ′ (φ1 ) and T ′ (φ2 ).
(b) Write T ′ (φ1 ) and T ′ (φ2 ) as a linear combination of ψ1 , ψ2 , ψ3 .

Proof. (a) Endowing R3 and R2 with their respective standard basis, we have

(T ′ (φ1 ))(x, y, z) = (φ1 ◦ T )(x, y, z)

= φ1 (T (x, y, z))
= φ1 (4x + 5y + 6z, 7x + 8y + 9z)
= 4x + 5y + 6z

and similarly

(T ′ (φ2 ))(x, y, z) = φ2 (4x + 5y + 6z, 7x + 8y + 9z)

= 7x + 8y + 9z.

(b) Notice

(4ψ1 + 5ψ2 + 6ψ3 )(x, y, z) = 4ψ1 (x, y, z) + 5ψ2 (x, y, z) + 6ψ3 (x, y, z)
= 4x + 5y + 6z
= T ′ (φ1 )(x, y, z)

and

(7ψ1 + 8ψ2 + 9ψ3 )(x, y, z) = 7ψ1 (x, y, z) + 8ψ2 (x, y, z) + 9ψ3 (x, y, z)
= 7x + 8y + 9z
= T ′ (φ2 )(x, y, z),

as desired.

Problem 15
Suppose W is finite-dimensional and T ∈ L(V, W ). Prove that T ′ = 0 if
and only if T = 0.

40
Proof. (⇒) Suppose T ′ = 0. Let φ ∈ W ′ and v ∈ V be arbitrary. We have

0 = (T ′ (φ))(v) = φ(T v).

Since φ is arbitrary, we must have T v = 0. But now since v is arbitrary, this

implies T = 0 as well.
(⇐) Suppose T = 0. Again let φ ∈ W ′ and v ∈ V be arbitrary. We have

(T ′ (φ))(v) = φ(T v) = φ(0) = 0,

and hence T ′ = 0 as well.

Problem 17
Suppose U ⊆ V . Explain why U 0 = {φ ∈ V ′ | U ⊆ null φ}.

Proof. It suffices to show that, for arbitrary φ ∈ V ′ , we have U ⊆ null φ if and

only if φ(u) = 0 for all u ∈ U . So suppose U ⊆ null φ. Then for all u ∈ U , we
have φ(u) = 0 (simply by definition of null φ). Conversely, suppose φ(u) = 0
for all u ∈ U . Then if u′ ∈ U , we must have u′ ∈ null φ. That is, U ⊆ null φ,
completing the proof.

Problem 19
Suppose V is finite-dimensional and U is a subspace of V . Show that
U = V if and only if U 0 = {0}.

Proof. (⇒) Suppose U = V . Then

U 0 = {φ ∈ V ′ | U ⊆ null φ}
= {φ ∈ V ′ | V ⊆ null φ}
= {0},

since only the zero functional can have all of V in its null space.
(⇐) Suppose U 0 = {0}. It follows

dim V = dim U + dim U 0

= dim U + 0
= dim U.

Since the only subspace of V with dimension dim V is V itself, we have U = V ,

as desired.

Problem 20
Suppose U and W are subsets of V with U ⊆ W . Prove that W 0 ⊆ U 0 .

41
Proof. Suppose φ ∈ W 0 . Then φ(w) = 0 for all w ∈ W . If φ ̸∈ U 0 , then there
exists some u ∈ U such that φ(u) ̸= 0. But since U ⊆ W , u ∈ W . This is absurd,
hence we must have φ ∈ U 0 . Thus W 0 ⊆ U 0 , as desired.

Problem 21
Suppose V is finite-dimensional and U and W are subspaces of V with
W 0 ⊆ U 0 . Prove that U ⊆ W .

Proof. Suppose not. Then there exists a nonzero vector u ∈ U such that u ̸∈ W .
There exists some basis of U containing u. Define φ ∈ V ′ such that, for any
vector v in this basis, we have
(
1 if v = u
φ(v) =
0 otherwise.

By construction, φ ∈ W 0 , and hence φ ∈ U 0 . But this implies φ(u) = 0, a

contradiction.

Problem 22
Suppose U, W are subspaces of V . Show that (U + W )0 = U 0 ∩ W 0 .

Proof. Since U ⊆ U + W and W ⊆ U + W , Problem 20 tells us that (U + W )0 ⊆

U 0 and (U + W )0 ⊆ W 0 . Thus (U + W )0 ⊆ U 0 ∩ W 0 . Conversely, suppose
φ ∈ U 0 ∩ W 0 . Let x ∈ U + W . Then there exist u ∈ U and w ∈ W such that
x = U + W . Then

φ(x) = φ(u + w)
= φ(u) + φ(w)
= 0,

where the second equality follows since φ ∈ U 0 and φ ∈ W 0 by assumption.

Hence φ ∈ (U + W )0 and we have U 0 + W 0 ⊆ (U + W )0 . Thus (U + W )0 =
U 0 ∩ W 0 , as desired.

Problem 23
Suppose V is finite-dimensional and U and W are subspaces of V . Prove
that (U ∩ W )0 = U 0 + W 0 .

Proof. Since U ∩W ⊆ U and U ∩W ⊆ W , Problem 20 tells us that U 0 ⊆ (U ∩W )0

and W 0 ⊆ (U ∩ W )0 . Thus U 0 + W 0 ⊆ (U ∩ W )0 . Now, notice (using Problem

42
22 to deduce the second equality)
dim(U 0 + W 0 ) = dim(U 0 ) + dim(W 0 ) − dim(U 0 ∩ W 0 )
= dim(U 0 ) + dim(W 0 ) − dim((U + W )0 )
= (dim V − dim U ) + (dim V − dim W ) − [dim V − dim(U + W )]
= dim V − dim U − dim W + dim(U + W )
= dim V − [dim U + dim W − dim(U + W )]
= dim V − dim(U ∩ W )
= dim((U ∩ W )0 ).
Hence we must have U 0 + W 0 = (U ∩ W )0 , as desired.

Problem 25
Suppose V is finite-dimensional and U is a subspace of V . Show that

U = {v ∈ V | φ(v) = 0 for every φ ∈ U 0 }.

Proof. Let A = {v ∈ V | φ(v) = 0 for every φ ∈ U 0 }. Suppose u ∈ U . Then

φ(u) = 0 for all φ ∈ U 0 , and hence u ∈ A, showing U ⊆ A.
Conversely, suppose v ∈ A but v ̸∈ U . Since 0 ∈ U , we must have v ̸= 0.
Thus there exists a basis u1 , . . . , um , v, v1 , . . . , vn of V such that u1 , . . . , um is a
basis of U . Let ψ1 , . . . , ψm , φ, φ1 , . . . , φn be the dual basis of V ′ , and consider
for a moment the functional φ. Clearly we have both φ ∈ U 0 and φ(v) = 1 by
construction, but this is a contradiction, since we assumed v ∈ A. Thus A ⊆ U ,
and we conclude U = A, as was to be shown.

Problem 27
Suppose T ∈ L(P5 (R), P5 (R)) and null T ′ = span(φ), where φ is the
linear functional on P5 (R) defined by φ(p) = p(8). Prove that range T =
{p ∈ P5 (R) | p(8) = 0}.

Proof. By Theorem 3.107, we know null T ′ = (range T )0 , and hence (range T )0 =

{αφ | α ∈ R}. It follows by Problem 25
range T = {p ∈ P5 (R) | ψ(p) = 0 for all ψ ∈ (range T )0 }
= {p ∈ P5 (R) | (αφ)(p) = 0 for all α ∈ R}
= {p ∈ P5 (R) | φ(p) = 0}
= {p ∈ P5 (R) | p(8) = 0},
as desired.

Problem 29
Suppose V and W are finite-dimensional, T ∈ L(V, W ), and there exists
φ ∈ V ′ such that range T ′ = span(φ). Prove that null T = null φ.

43
Proof. By Theorem 3.107, we know range T ′ = (null T )0 , and hence (null T )0 =
{αφ | α ∈ R}. It follows by Problem 25

null T = {v ∈ V | ψ(v) = 0 for all ψ ∈ (null T )0 }

= {v ∈ V | αφ(v) = 0 for all α ∈ F}
= {v ∈ V | φ(v) = 0}
= null φ,

as desired.

Problem 31
Suppose V is finite-dimensional and φ1 , . . . , φn is a basis of V ′ . Show
that there exists a basis of V whose dual basis is φ1 , . . . , φn .

Proof. To prove this, we will first show V ∼ = V ′′ . We will then take an existing
basis of V , map it to its dual basis in V ′′ , and then use the inverse of the
′

isomorphism to take this basis of V ′′ to a basis in V . This basis of V will have

the known basis of V ′ as its dual.
So, for any v ∈ V , define Ev ∈ V ′′ by Ev (φ) = φ(v). We claim the map
ˆ· : V → V ′′ given by v̂ = Ev is an isomorphism. To do so, it suffices to show it to
be both linear and injective, since dim(V ′′ ) = dim((V ′ )′ ) = dim(V ′ ) = dim(V ).
We first show ˆ· is linear. So suppose u, v ∈ V . Then for any φ ∈ V ′ , we have

(u[
+ v)(φ) = Eu+v (φ)
= φ(u + v)
= φ(u) + φ(v)
= Eu (φ) + Ev (φ)
= û(φ) + v̂(φ)

so that ˆ· is indeed linear. Next we show it to be homogeneous. So suppose λ ∈ F,

and again let v ∈ V . Then for any φ ∈ V ′ , we have

(λv)(φ)
c = Eλv (φ)
= φ(λv)
= λφ(v)
= λEv (φ)
= λv̂,

so that ˆ· is homogenous as well. Being both linear and homogenous, it is a linear

map.
Next we show ˆ· is injective. So suppose v̂ = 0 for some v ∈ V . We want to
show v = 0. Let v1 , . . . , vn be a basis of V . Then there exist α1 , . . . , αn ∈ F

44
such that v = α1 v1 + · · · + αn vn . Then, for all φ ∈ V ′ , we have

v̂ = 0 =⇒ (α1 v1 + · · · + αn vn )∧ = 0
=⇒ α1 vb1 + · · · + αn vc
n =0
=⇒ (α1 vb1 + · · · + αn vc
n )(φ) = 0
=⇒ α1 vb1 (φ) + · · · + αn vc
n (φ) = 0
=⇒ α1 φ(v1 ) + · · · + αn φn (vn ) = 0.

Since this last equation holds for all φ ∈ V ′ , it holds in particular for each
element of the dual basis φ1 , . . . , φn . That is, for k = 1, . . . , n, we have

α1 φk (v1 ) + · · · + αk φk (vk ) + · · · + αn φk (vn ) = 0 =⇒ αk = 0,

and therefore v = 0 · v1 + · · · + 0 · vn = 0, as desired. Thus ˆ· is indeed an

isomorphism.
We now prove the main result. Suppose φ1 , . . . , φn is a basis of V ′ , and let
Φ1 , . . . , Φn be the dual basis in V ′′ . For each Φk , let vk be the inverse of Φk
under the isomorphism ˆ·. Since the inverse of an isomorphism is an isomorphism,
and isomorphisms take bases to bases, v1 , . . . , vn is a basis of V . Let us now
check that its dual basis is φ1 , . . . , φn . For j, k = 1, . . . , n, we have

φj (vk ) = vbk (φj )

= Φk (φj )
(
1 if k = j
=
0 otherwise,

so indeed there exists a basis of V whose dual basis is φ1 , . . . , φn , as was to be

shown.

Problem 32
Suppose T ∈ L(V ) and u1 , . . . , un and v1 , . . . , vn are bases of V . Prove
that the following are equivalent:
(a) T is invertible.

(b) The columns of M(T ) are linearly independent in Fn,1 .

(c) The columns of M(T ) span Fn,1 .
(d) The rows of M(T ) are linearly independent in F1,n .

(e) The rows of M(T ) span F 1,n .

Here M(T ) means M(T, (u1 , . . . , un ), (v1 , . . . , vn )).

Proof. We prove the following: (a) ⇐⇒ (b) ⇐⇒ (c) ⇐⇒ (e) ⇐⇒ (d).

(a) ⇐⇒ (b). Suppose T is invertible. That is, for any w ∈ V , there exists a

45
unique x ∈ V such that w = T x. It follows

M(w) = M(T x)
= M(T )M(x)
= M(x)1 M(T )·,1 + · · · + M(x)n M(T )·,n .

That is, every vector in F1,n can be exhibited as a unique linear combination
of the columns of M(T ). This is true if and only if the columns of M(T ) are
linearly independent.
(b) ⇐⇒ (c). Suppose the columns of M(T ) are linearly independent in Fn,1 .
Since they form a linearly independent list of length dim(Fn,1 ), they are a basis.
But this is true if and only if they span Fn,1 as well.
(c) ⇐⇒ (e). Suppose the columns of M(T ) span Fn,1 , so that the column
rank is n. Since the row rank equals the column rank, so too must the rows of
M(T ) span F1,n .
(e) ⇐⇒ (d). Suppose the rows of M(T ) span F1,n . Since they form a
spanning list of length dim(F1,n ), they are a basis. But this is true if and only if
they are linearly independent in F1,n as well.

Problem 33
Suppose m and n are positive integers. Prove that the function that
takes A to At is a linear map from F m,n to Fn,m . Furthermore, prove
that this linear map is invertible.

Proof. We first show taking the transpose is linear. So suppose A, B ∈ Fm,n and
let j = 1, . . . , n and k = 1, . . . , m. It follows

(A + B)tj,k = (A + B)k,j
= Ak,j + Bk,j
= Atj,k + Bj,k
t
,

so that taking the transpose is additive. Next, let λ ∈ F. It follows

(λA)tj,k = (λA)k,j
= λAk,j
= λAtj,k ,

so that taking the transpose is homogenous. Since it is both additive and

homogeneous, it is a linear map. To see that taking the transpose is invertible,
note that (At )t = A, so that the inverse of the transpose is the transpose
itself.

46
 Problem 34
The double dual space of V , denoted V ′′ , is defined to be the dual
space of V ′ . In other words, V ′′ = (V ′ )′ . Define Λ : V → V ′′ by

(Λv)(φ) = φ(v)

for v ∈ V and φ ∈ V ′ .
(a) Show that Λ is a linear map from V to V ′′ .
(b) Show that if T ∈ L(V ), then T ′′ ◦ Λ = Λ ◦ T , where T ′′ = (T ′ )′ .

(c) Show that if V is finite-dimensional, then Λ is an isomorphism from

V onto V ′′ .

Proof. We proved (a) and (c) in Problem 31 (where we defined ˆ· in precisely the
same way as Λ). So it only remains to prove (b). So suppose v ∈ V and φ ∈ V ′
are arbitrary. Evaluating T ′′ ◦ Λ, notice

((T ′′ ◦ Λ)(v))(φ) = (T ′′ (Λv))(φ)

= (Λv)(T ′ φ)
= (T ′ φ)(v)
= φ(T v),

where the second and fourth equalities follow by definition of the dual map, and
the third equality follows by definition of Λ. And evaluating Λ ◦ T , we have

((Λ ◦ T )(v))(φ) = (Λ(T v))(φ)

= φ(T v),

so that the two expressions evaluate to the same thing. Since the choice of both
v and φ was arbitrary, we have T ′′ ◦ Λ = Λ ◦ T , as desired.

Problem 35
Show that (P(R))′ and R∞ are isomorphic.

Proof. For any sequence α = (α0 , α1 , . . . ) ∈ R∞ , let φα be the unique linear

functional in (P(R))′ such that φα (X k ) = αk for all k ∈ Z+ (note that since
the list 1, X, X 2 , . . . is a basis of P(R), this description of φα is sufficient). We
claim

Φ : R∞ → (P(R))′
α 7→ φα

is an isomorphism. There are three things to show: that Φ is a linear map, that
it’s injective, and that it’s surjective.

47
 We first show Φ is linear. Suppose α, β ∈ R∞ . For any k ∈ Z+ , it follows
(Φ(α + β))(X k ) = φα+β (X k )
= (α + β)k
= α k + βk
= φα (X k ) + φβ (X k )
= (Φ(α))(X k ) + (Φ(β))(X k ),
so that Φ is additive. Next suppose λ ∈ R. Then we have
Φ(λα)(X k ) = φλα (X k )
= (λα)k
= λαk
= λΦ(α),
so that Φ is homogenous. Being both additive and homogeneous, Φ is indeed
linear.
Next, to see that Φ is injective, suppose Φ(α) = 0 for some α ∈ R∞ . Then
φα (X k ) = αk = 0 for all k ∈ Z+ , and hence α = 0. Thus Φ is injective.
Lastly, to see that Φ is surjective, suppose φ ∈ (P(R))′ . Define αk = φ(X k )
for all k ∈ Z+ and let α = (α0 , α1 , . . . ). By construction, we have (Φ(α))(X k ) =
αk for all k ∈ Z+ , and hence Φ(α) = φα . Thus Φ is surjective.
Since Φ is linear, injective, and surjective, it’s an isomorphism, as desired.
Problem 37
Suppose U is a subspace of V . Let π : V → V /U be the usual quotient
map. Thus π ′ ∈ L((V /U )′ , V ′ ).
(a) Show that π ′ is injective.

(b) Show that range π ′ = U 0 .

(c) Conclude that π ′ is an isomorphism from (V /U )′ onto U 0 .

Proof. (a) Let φ ∈ (V /U )′ , and suppose π ′ (φ) = 0. Then (φ ◦ π)(v) =

φ(v + U ) = 0 for all v ∈ V . This is true only if φ = 0, and hence π ′ is
indeed injective.
(b) First, suppose φ ∈ range π ′ . Then there exists ψ ∈ (V /U )′ such that
π ′ (ψ) = φ. So for all u ∈ U , we have
φ(u) = (π ′ (ψ))(u)
= ψ(π(u))
= ψ(u + U )
= ψ(0 + U )
= 0,

48
 and thus φ ∈ U 0 , showing range π ′ ⊆ U 0 . Conversely, suppose φ ∈ U 0 , so
that φ(u) = 0 for all u ∈ U . Define ψ ∈ (V /U )′ by ψ(v + U ) = φ(v) for
all v ∈ V . Then (π ′ (ψ))(v) = ψ(π(v)) = ψ(v + U ) = φ(v), and so indeed
φ ∈ range π ′ , showing U 0 ⊆ range π ′ . Therefore, we have range π ′ = U 0 ,
as desired.

(c) Notice that (b) may be interpreted as saying π ′ : (V /U )′ → U 0 is surjective.

Since π ′ was shown to be injective in (a), we conclude π ′ is an isomorphism
from (V /U )′ onto U 0 , as desired.

49