and
x1 = 3x2 + x4
so
v = (3x2 + x4 , x2 , x4 , x4 ) = x2 (3, 1, 0, 0) + x4 (1, 0, 1, 1).
This shows that V span ((3, 1, 0, 0), (1, 0, 1, 1)). On the other hand, since
both of those vectors are in V , we in fact have
V = span ((3, 1, 0, 0), (1, 0, 1, 1)) .
It remains to show these vectors are independent: suppose
(0, 0, 0, 0) = 1 (3, 1, 0, 0) + 2 (1, 0, 1, 1) = (31 + 2 , 1 , 2 , 2 ).
Looking at e.g. the second and third coordinates, we see 1 = 2 = 0, as
required.
The dimension is two, since we found a basis with two elements.
Note: there are many possible bases; solving for (say) x1 and x4 in terms of the
others in the above solution would yield a different, but equally valid, basis.
(b) For 1 i m and 1 j n, let Eij denote the matrix whose (i, j)th entry is
1 and all other entries are 0. Well show the set
{Eij }1im
1jn
m X
n
X
i=1
aij Eij
j=1
ij Eij = st ,
i,j
a1 = a3 = 0.
Page 2
Solution: This becomes easy if we use the characterization of span that says that
for any set {x1 , . . . , xk }, the span span(x1 , . . . , xk ) is the smallest subspace containing
all the xi s.
Clearly v1 , . . . , vn span(w, v1 , . . . , vn ), so
span(v1 , . . . , vn ) span(w, v1 , . . . , vn ).
Conversely, we have v1 . . . , vn span(v1 , . . . , vn ) and w span(v1 , . . . , vn ) by assumption, so
span(w, v1 , . . . , vn ) span(v1 , . . . , vn ).
Thus span(w, v1 , . . . , vn ) = span(v1 , . . . , vn ) as required.
[Alternatively, since w span(v1 , . . . , vn ), there are scalars a1 , . . . , an such that
w = a1 v1 + + an vn .
Then for any
w0 = 0 w + 1 v1 + + n vn span(w, v1 , . . . , vn ),
we can write
w0 = 0 (a1 v1 + + an vn ) + 1 v1 + + n vn
= (0 a1 + 1 )v1 + + (0 an + n )vn
span(v1 , . . . , vn )
so span(w, v1 , . . . , vn ) span(v1 , . . . , vn ). ]
3. (8 points) (Axler 2.B.8) Suppose U and W are subspaces of a vector space V , and that
{u1 , . . . , un } and {w1 , . . . , wm } are bases of U and W respectively. Show that
{u1 , . . . , un , w1 , . . . , wm }
is a basis of V if and only if V = U W .
Solution: First, suppose that V = U W . To show that {u1 , . . . , un , w1 , . . . , wm }
is a spanning set for V , we note that
U = span(u1 , . . . , un )
and
Page 3
W = span(w1 , . . . , wm )
by assumption, since {u1 , . . . , un } and {w1 , . . . , wm } are bases of U and W respectively. Then
span(u1 , . . . , un , w1 , . . . wm ) = span(u1 , . . . , un ) + span(w1 , . . . , wm ) = U + W = V.
To show that {u1 , . . . , un , w1 , . . . , wm } is linearly independent, suppose
0 = 1 u1 + + n un + 01 w1 + + 0m wm
for some scalars 1 , . . . , n , 01 , . . . , 0n F. Let
v = 1 u1 + + n un U.
Observe that
v = 1 u1 + + n un = (01 w1 + + 0m wm ) W
so v U W . By assumption, the sum U + W is direct, so U W = {0}, i.e. v = 0.
Therefore
0 = 1 u1 + + n un = (01 w1 + + 0m wm ).
Since the ui s are linearly independent, we must have 1 = = n = 0, and since
the wi s are independent, we get 01 = = 0n = 0.
[Alternative proof of independence: We use the dim-sum formula: since V = U + W
is direct by assumption, we obtain
dim(V ) = dim(U W ) = dim(U ) + dim(W ) = m + n.
The set {u1 , . . . , un , w1 , . . . , wm } is a spanning set for V whose length is the same as
dim(V ), hence it is automatically independent, i.e. a basis. ]
Conversely, suppose {u1 , . . . , un , w1 , . . . , wm } is a basis of V . Then since it is a
spanning set,
V = span(u1 , . . . , un , w1 , . . . , wm ) = span(u1 , . . . , un ) + span(w1 , . . . , wm )
= U + W,
since by definition, the ui s span U and the wi s span W . To show that the sum is
direct, we argue in reverse: suppose v U W . Then there are scalars 1 , . . . , n
such that
v = 1 u1 + + n un
and scalars 01 , . . . , 0m such that
v = 01 w1 + + 0m wm .
Page 4
denote the highest degree among the polynomials appearing in the list. Then any
linear combination of the fi s has degree less than or equal to N , i.e.
span(f1 , . . . , fn ) F[x]N .
In particular, the polynomial xN +1
/ span(f1 , . . . , fn ).
Thus for every finite set of polynomials {f1 , . . . , fn } , there is another polynomial g
that is not in the span of the fi s. Hence there cannot be a finite spanning set for all
of F[t].
5. (5 points) Let V be a vector space over the finite field Fp with p elements, where p is
a prime number. Suppose dim(V ) = n. What is the cardinality of V , i.e. how many
elements does V have? Hint: fix a basis of V and use coordinates.
Solution: Let = {v1 , . . . , vn } be a basis for V . Then the coordinate map T : V
(Fp )n that sends v to [v] is an isomorphism, and in particular is a bijection. So we
need to count the number of elements of
(Fp )n = {(1 , . . . , n ) | i Fp }.
Page 5
Solution:
(a) Let {x1 , . . . , xn } X, and suppose
0 = 1 fx1 + + n fxn
for some scalars 1 , . . . n F. For each i, we evaluate the previous expression
at xi :
0 = 1 fx1 (xi ) + + n fxn (xi )
= 1 0 + + i1 0 + i fxi (xi ) + i+1 0 + + n 0
= i 1 = i .
Hence all the i s must be zero, and so {fx1 , . . . , fxn } is linearly independent.
(b) We claim that V is finite-dimensional if and only if X is a finite set.
First, suppose V is finite dimensional. Then every linearly independent list of
vectors has length dim(V ). By part (a), this implies that there cannot be
more than dim(V ) elements in X; in particular, X is a finite set.
Conversely, suppose X = {x1 , . . . , xn } is a finite set and let V . Define
ai = (xi ) F
and consider
g = a1 fx1 + + an fxn V.
Page 6
8. (10 bonus points) * This is a bonus question, and is quite challenging! Suppose F is an
infinite field, and that V is a finite-dimensional vector space over F. Show that V is not
the union of finitely many proper subspaces. (compare with Question 6 on Assigment 1)
Page 7
n
[
Ui
i=1
for some proper subspaces U1 , . . . , Un . Without loss of generality, we may suppose that
Ui 6 Uj for all i 6= j; if one subspace were contained in another, we can throw away the
smaller one without changing the union.
The goal will be to show that
()
U1 =
n
[
(Ui U1 );
i=2
if we can do this, then we are done for the following reasons. Each Ui U1 is a proper
subspace of U1 since U1 6 Ui . On the other hand since U1 is a proper subspace of V ,
we have dim(U1 ) < dim(V ), and our induction step assumes that the statement is true
for all vector spaces of smaller dimension than V . Therefore () is a contradiction, and
were done.
Now lets show () holds. Let u U1 and choose w V such that w
/ U1 . Let F be
any scalar, and consider the vector
w + u V.
S
Since V = Ui , the vector w + u is in one of the Ui s. However it cannot be in U1 ,
because w
/ U1 . Therefore, for every scalar ,
w + u
n
[
Ui
i=2
Page 8
n
[
Ui ,
i=2
and so
U1 =
n
[
(Ui U1 )
i=2
as required.
Page 9