Math 2135: Linear Algebra

Assignment 1

Solutions

√
1 3
1. Show that 2 + 2 i ∈ C is a cube root of −1. Find two more cube roots of −1 in C.
We compute:
Ç √ å3 Ç √ å2 Ç √ å
1 3 1 3 1 3
+ i = + i + i
2 2 2 2 2 2
Å ã2 Å ã Ç √ å Ç √ å2 ! Ç √ å Ç √ åÇ √ å
1 1 3 3 1 3 1 3 3 1 3
= +2 i + i + i = + i− + i
2 2 2 2 2 2 4 2 4 2 2
Ç √ åÇ √ å Å ã2 Ç √ å Å ã Ç √ å Å ã Ç √ å2
1 3 1 3 1 3 1 3 1 3
= − + i + i =− + i −+ i + i
2 2 2 2 2 2 2 2 2 2
1 3
= − + 0 − = −1
4 4
√
1 3
Exactly the same calculations, with −i in place of i, show that 2 − 2 i is a cube root of
−1. And of course −1 is a (real) cube root of −1.

2. (a) Does there exist a λ ∈ C such that λ(2, 3i, 4 + 5i) = (6, 7i, 8 − 9i)? Why or why
not?
No. Note that the left-hand side is (2λ, 3iλ, (4 + 5i)λ). For this to be equal to the
right-hand side, we must have 2λ = 6, which forces λ = 3. However, for this λ, we find
that 3iλ = 9i 6= 7i. So there is no equality.
(b) Does there exist a λ ∈ C such that λ(2, 1 + i) = (1 − i, 1)? Why or why not?
Yes. Indeed, (1 + i)(1 − i) = 12 − i2 = 1 − (−1) = 2. So λ = 1/(1 + i) works. What is
this λ? It is
1 1−i 1−i 1 1
= = = − i.
1+i (1 + i)(1 − i) 2 2 2

3. Suppose that F = R or C and that V is a vector space over F. Show that, if α ∈ F

and v ∈ V , and if αv = 0, then either α = 0 or v = 0 (or both).
Equivalently, we are asked to show that if α 6= 0, then v = 0. But if α 6= 0, then there is a
number α−1 ∈ F such that αα−1 = 1, in which case

0 = α−1 0 = α−1 αv = 1v = v.

4. The first three of the following six sets are subsets of the vector space R3 , and
the last three are subsets of the vector space RR . Three of these six subsets are

1
 actually vector subspaces, and the other three are not. For those that are, simply
state that they are indeed vector spaces: you do not need to prove why. For those
that are not, state that they are not vector spaces, and give a reason: explain
one of the axioms for vector space that fails. (There might be more than one!)

(a) The set of triples (x, y, z) ∈ R3 such that x + y = z.

This is a vector subspace of R3 . Basically this is because the equation “x + y = z” is
linear.
(b) The set of triples (x, y, z) ∈ R3 such that xy = z.
This is not a vector subspace of R3 . For example, the vector v = (1, 1, 1) is in this
subset, but v + v = (2, 2, 2) is not, so it is not closed under +.
(c) The set of triples (x, y, z) ∈ R3 such that x, y, z are nonnegative.
This is not a vector subspace of R3 . For example, the vector v = (1, 1, 1) is in this
subset, but −v = (−1, −1, −1) is not, so it is not closed under scalar multiplication. (It
is closed under +, on the other hand.)
(d) The set of all smooth functions f : R → R.
This is a vector subspace of RR , basically because sums of products of smooth functions
are again smooth. (A function is smooth when it has a continuous derivative, which has
a continuous derivative, which has a continuous derivative, ad infinitum.)
(e) The set of all functions f : R → R which solve the differential equation
f 00 (x) = f (x), where f 00 denotes the second derivative of f .
This is a vector subspace of RR , basically because the equation “f 00 = f ” is linear.
Indeed, 0 solves f 00 = f (because the derivative of the zero function is again zero); if
f1 , f2 both solve f 00 = f , then so does f1 + f2 (because derivative of a sum is sum of
derivatives); and if α ∈ R then (αf )0 = αf 0 , and so multiplying a solution to f 00 = f by
α produces another solution.
(f) The set of functions f : R → R such that f (0) is an integer.
This is not a vector subspace of RR . It does pass some tests: it contains the 0 function
and is closed under + and −. However, it is not closed under scalar multiplication.
Indeed: the constant function f (x) ≡ 1 is in this set, but multiplying this by the number
α = 21 produces the constant function f (x) ≡ 21 , which is not in this set.

5. Is R naturally a vector space over C? Is C naturally a vector space over R?

V = C is naturally a vector space over F = R. First, C comes equipped with an addition
law which satisfies all the requirements for addition in a vector space. To be a vector space
over R, the other datum C needs is a way of taking a “scalar” α ∈ R and a “vector” v ∈ C
and multiply them to produce a new “vector” αv ∈ C. Since the real numbers are naturally
a subset of the complex numbers, we can use the complex multiplication for this: just read
“αv” as that product of complex numbers. This multiplication law does indeed satisfy the
necessary associativity and distributivity rules.
V = R is not naturally a vector space over FC. It does come with an addition law satisfying
all requirements. However, in order to be a vector space, we would need a way to multiply a
“scalar” α ∈ C and a “vector” v ∈ R and produce a “vector” αv ∈ R. Let’s try to find such a
multiplication law; by trying and failing, we will amass evidence that it might be impossible.
(We will not produce a proof that it is impossible, however.) Well, we could remember that R
is a subset of C and so evaluate αv as a complex multiplication. But the result will typically

2
 be a complex number which is not real. We could take just its real part. But that will not
be suitably associative.
Actually proving that V = R is not naturally a vector space over F = C is hard because it
requires saying more carefully what the word “naturally” means. Let’s agree that in order to
be natural, the vector addition should be the addition in R. Let’s also agree that “naturality”
requires that the multiplication law αv, where α ∈ C and v ∈ R, have the following property:
if α happens to be real, then αv is means the usual multiplication of real numbers. With
these agreements, we can prove
√ that R is not naturally a vector space over C. Indeed, suppose
that it were, and take α = −1 and v = 1. Then, by supposition, there would be some real
number r = αv. But this is also equal to βv where β = r ∈ R ⊂ C and still v = 1. So
(α − β)v = 0. By exercise
√ 2, this forces α − β = 0 or v = 0 or both. Well, v 6= 0, so it forces
α − β = 0. But α = −1 whereas β is real, and so we have a contradiction.

6. Let V := R>0 denote the set of positive real numbers. Let’s define, for v, w ∈ V ,
their “V -sum” to be v +V w := vw, where the subscript on the left-hand side
the subscript reminds that it is a new notion of “sum” in the set V , and on the
right-hand side we mean the product of positive numbers. Also, for λ ∈ R and
v ∈ V , let’s define their “product” λ ·V v := v λ , where the right-hand side means
the exponential of real numbers. Is this V , with these notions of addition and
scalar multiplication, a vector space over R? Explain.
This V , with these notions of addition and scalar multiplication, is a vector space over R!
There is a sneaky reason for this. Given v ∈ V , consider the real number r = log v ∈
R1 ; conversely, given r ∈ R1 , consider the positive real number v = exp v ∈ V = R>0 .
(Mathematicians typically write “exp” for the natural, i.e. base e, exponential function, and
“log” for the natural, i.e. base e, logarithm. But any base will work so long as they are the
same: if you want to think of log as the base-10 logarithm, then just decide that exp(r) := 10r .)
These two functions exp : r ↔ v : log provide a perfect matching between R1 and V .
Moreover, they relate the usual addition and scalar multiplication in R1 to the funny versions
for V .
There is also a non-sneaky reason: we can just check the axioms directly. First, we need to
?
check whether +V is commutative and associative: If u, v, w ∈ V , do we have v +V w = w +V v
?
and (u +V v) +V w = u +V (v +V w)? Yes, because these questions are equivalent to asking,
? ?
for the usual multiplication of real numbers, whether vw = wv and (uv)w = u(vw), and the
answers to both questions are yes. Second, we need to check whether there is an element
0V ∈ V such that, for every v ∈ V , 0V +V v = v? Yes: take 0V := 1, and use that for
usual multiplication, 1v = v. And is there negation? I.e. given v ∈ V , is there −V v such
that v +V (−V v) = 0V ? Yes: take −V v := v −1 . Third, we need to check some properties of
?
multiplication. We need to check whether, given α, β ∈ R and v ∈ V , is α·V (β·V v) = (αβ)·V v?
?
Yes, because this unpacks to the question of whether (v α )β = v αβ . We need to check whether,
? ?
given v ∈ V , is 1 ·V v = v? Yes, because this unpacks to the question of whether v 1 = v. And
?
last, we need to check some distributivity laws, which unpack to whether (vw)α = v α wα and
?α
whether v α+β v v β .

7. Let V = R t {+∞, −∞}. In other words, we take the set of real numbers, and
add two new elements to the set, named “+∞” and “−∞.” Equip V with an

3
 “addition” law which is the usual addition in R, extended by

v + (+∞) = (+∞) + v = ∞, v + (−∞) = (−∞) + v = −∞,

(+∞) + (+∞) = (+∞), (+∞) + (−∞) = (−∞) + (+∞) = 0, (−∞) + (−∞) = −∞

Also, define a “scalar multiplication” law which is the usual multiplication in R,

and, if α ∈ R, then
 
−∞,

 α < 0, +∞,

 α < 0,
α(+∞) = 0, α = 0, α(−∞) = 0, α = 0,
 

+∞, α > 0, −∞,

α > 0.

Is this V , with these notions of addition and scalar multiplication, a vector space
over R? Explain.
This V , with these notions of addition and scalar multiplication, is not a vector space over
R. Almost every axiom fails. For example, associativity of + fails: ((+∞)+(+∞))+(−∞) =
(+∞) + (−∞) = 0, but (+∞) + ((+∞) + (−∞)) = (+∞) + 0 = +∞.

8. Let V be a vector space.

(a) Prove that the intersection of two vector subspaces of V is always another
vector subspace of V .
Let U1 , U2 ⊂ V be vector subspaces. To show that U := U1 ∩ U2 is a vector subspace of
V , it suffices to show: (i) 0 ∈ U ; (ii) if u, v ∈ U then u + v ∈ U ; (iii) if u ∈ U and α ∈ R,
then αu ∈ U .
To show (i), note that 0 ∈ U1 and 0 ∈ U2 and hence 0 ∈ U1 ∩ U2 = U .
To show (ii), note that, if u, v ∈ U , then u, v ∈ U1 and so u + v ∈ U1 , but also u, v ∈ U2
so u + v ∈ U2 , and hence u + v ∈ U1 ∩ U2 = U .
To show (iii), note that, if u ∈ U , then u ∈ U1 , and so αu ∈ U1 for every α ∈ R. But
also u ∈ U2 , and so αu ∈ U2 . Since αu is in both U1 and U2 , it is in U = U1 ∩ U2 .
(b) Prove that the union of two vector subspaces of V is another vector subspace
of V only if one of the two original subspaces contains the other one.
Let U1 , U2 ⊂ V be vector subspaces. We are asked to show: the only way for U1 ∪ U2 to
be a vector subspace is if U1 ⊂ U2 or U2 ⊂ U1 (or both). Equivalently, we are asked to
show that if U1 6⊂ U2 and U2 6⊂ U1 , then U1 ∪ U2 is not a vector subspace.
Well, if U1 6⊂ U2 , then we can choose an element u1 ∈ U1 which is not also in U2 .
Similarly, if U2 6⊂ U1 , then we can choose an element u2 ∈ U2 which is not also in U1 .
So let’s make such choices. Then these u1 and u2 are both in U1 ∪ U2 . We will show,
however, that their sum u1 + u2 is not in U1 ∪ U2 . This will prove that U1 ∪ U2 is not
closed under +, and so is not a vector subspace.
First, we claim that u1 + u2 is not in U1 . If it were, then since also u1 ∈ U1 and since
U1 is a vector subspace, we would have u2 = (u1 + u2 ) − u1 ∈ U1 . But when we selected
u2 , we explicitly selected it to not be in U1 . Second, we claim that u1 + u2 is not in
U2 . If it were, then since also u2 ∈ U2 and since U2 is a vector subspace, we would have
u1 = (u1 + u2 ) − u2 ∈ U2 . But when we selected u1 , we explicitly selected it to not be in
U2 . So both statements “u1 + u2 ∈ U1 ” and “u1 + u2 ∈ U2 ” are false. This means that
u1 + u2 is also not in the union U1 ∪ U2 , which is what we wanted to show.

4
9. Let V be a vector space.
(a) Is the addition of vector subspaces associative? In other words, given three
vector subspaces U1 , U2 , U3 ⊂ V , is
?
U1 + (U2 + U3 ) = (U1 + U2 ) + U3

Yes. Indeed, the left-hand side unpacks to the set of all vectors in V which can be written
as u1 + (u2 + u3 ) for some elements u1 ∈ U1 , u2 ∈ U2 , and u3 ∈ U3 . The right-hand side
unpacks to the set of all vectors in V which can be written as (u1 + u2 ) + u3 for some
elements u1 ∈ U1 , u2 ∈ U2 , and u3 ∈ U3 . The associativity of vector addition implies
that these two sets are the same
(b) Is the addition of vector subspaces commutative? In other words, given two
vector subspaces U1 , U2 ⊂ V , is
?
U1 + U2 = U2 + U1

Yes. After unpacking, the commutativity of vector addition implies that these two sets
are the same.
(c) Is there a vector subspace “O” ⊂ V such that U + O = U for every vector
subspace U ⊂ V ?
Yes. The subspace O := {0} works. Indeed, for this U + O is the set of vectors of the
form u + o where u ∈ U and o ∈ O. But the only choice for o is o = 0, and so U + O is
the set of vectors of the form u + 0 where u ∈ U . But this is precisely the set U .
10. Let U1 ⊂ RR be the set of functions f : R → R such that f (x) = 0 if x < 0. Let
U2 ⊂ RR be the set of functions f : R → R such that f (x) = 0 if x ≥ 0.
(a) Show that U1 and U2 are vector subspaces of RR .
For each U = U1 or U2 , it suffices to show: (i) 0 ∈ U ; (ii) if f, g ∈ U then f + g ∈ U ;
(iii) if f ∈ U and α ∈ R, then αf ∈ U .
Well, the zero function 0 certainly vanishes on all negative numbers, and so it is in U1 .
It also vanishes on all nonnegative numbers, so it is in U2 . This confirms (i) for both U1
and U2 .
Suppose that f, g ∈ U1 , i.e. if x < 0, then f (x) = g(x) = 0. Then, continuing to
assume that x < 0, we find that (f + g)(x) = f (x) + g(x) = 0 + 0 = 0. This shows
that f + g ∈ U1 , establishing (ii) for U1 . Essentially the same proof works for U2 :
suppose that f, g ∈ U2 , i.e. that if x ≥ 0 then f (x) = g(x) = 0; thus if x ≥ 0 then
(f + g)(x) = f (x) + g(x) = 0 + 0 = 0; thus f + g ∈ U2 .
Property (iii) is proved similarly. Suppose that f ∈ U1 and that α ∈ R and that x < 0.
Then (αf )(x) = α f (x) = α 0 = 0. Since this was true for all x < 0, we find that αf ∈ U1 .
Suppose that f ∈ U2 and that α ∈ R and that x ≥ 0. Then (αf )(x) = α f (x) = α 0 = 0.
Since this was true for all x ≥ 0, we find that αf ∈ U1
(b) Show that U1 + U2 = RR .
We must show that, for any function f : R → R, there exists functions f1 ∈ U1 and
f2 ∈ U2 such that f = f1 + f2 . Set
( (
f (x), x ≥ 0, 0, x ≥ 0,
f1 (x) := f2 (x) :=
0, x < 0, f (x), x < 0.

5
 Then we find that indeed f1 does satisfy the requirements to be in U1 (i.e. it is a function
which vanishes on negative numbers), and f2 does satisfy the requirements to be in U2
(i.e. it is a function which vanishes on nonnegative numbers). We can add these functions
by considering the two cases:
(
f (x) + 0, x ≥ 0,
(f1 + f2 )(x) = f1 (x) + f2 (x) =
0 + f (x), x < 0.

But this is just f (x) in all cases, and so f1 + f2 = f .

(c) Show that the sum is direct.
It suffices to show that U1 ∩ U2 = {0}. In other words, we claim that the only function
which is in both U1 and U2 is the zero function. Well, if f ∈ U1 ∩ U2 , then certainly
f ∈ U1 and so f (x) = 0 for all negative x. On the other hand, if f ∈ U1 ∩ U2 , then
certainly f ∈ U2 and so f (x) = 0 for all nonnegative x. But every real x is either
negative or nonnegative, and so we see that if f ∈ U1 ∩ U2 , then f (x) = 0 for all x. And
so f = 0, which is what we wanted to prove.