Math 25a Homework 8 Solutions

Ivan Corwin and Alison Milller.

1 Alison’s problems
(1) We know that the set M2×2 (R) of 2 × 2 real matrices forms a vector space over R (with the
usual definitions of addition and scalar multiplications for matrices).
(a) Show that the subspace W consisting of symmetric matrices is a subspace of M2×2 (R).
(Note: W = {A ∈ M2×2 (R) : A = AT }.)
(b) Find a basis for W . What is the dimension of W?
(c) Find a basis for M2×2 (R) which contains a basis for W .

Solution. (a). Let w1 , w2 be any symmetric matrices in M2×2 (R). We need to show that w1 + w2
and cw1 are symmetric. Write out
   
x1 y1 x2 y2
w1 = , w2 = .
y 1 z1 y 2 z2

Then    
cx1 cy1 x1 + x2 y1 + y2
cw1 = , w1 + w2 =
cy1 cz1 y 1 + y 2 z1 + z2
is also symmetric. So W is a subspace.
(b) We claim that      
1 0 0 1 0 0
, ,
0 0 1 0 0 1
is a basis for W Linear independence: for a1 , a2 , a3 ∈ R, the linear combination
       
1 0 0 1 0 0 a1 a2
a1 + a2 + a3 =
0 0 1 0 0 1 a2 a3

is only 0 if all its entries are 0, that is,if a1 = 

a2 = a3 = 0.
a1 a2
Spanning: Any symmetric matrix can be represented as a linear combination by
a2 a3
       
a1 a2 1 0 0 1 0 0
= a1 + a2 + a3 .
a2 a3 0 0 1 0 0 1

So W has a basis of three elements, and has dimension 3.

(c) Extend the basis from (b) to the list
       
1 0 0 1 0 0 0 1
, , ,
0 0 1 0 0 1 0 0

1
By the linear dependence lemma, if these vectors are linearly independent, one of them must lie
in the span of the previous vectors. It can’t be one of
 the first three, because they are linearly
0 1
independent. Neither can it be the last one, because is asymmetric and the span of the
0 0
first three contains only symmetric matrices. Hence the four matrices are linearly independent.
Because dim M2×2 (R) = 4, they are a basis.

(2) Given a finite dimensional vector space V over a field F . We aim to show that V ∗∗ ∼ = V . Recall
the handy bracket notation we discussed in class (Monday 21st November). We discussed the map
h, i : V ∗ × V → F with (f, v) 7→ hf, vi = f (v).
(a) Given any nonzero vector v ∈ V , show that there is a f ∈ V ∗ such that f (v) = 0.
(b) Define a map φ : V → V ∗∗ by φ(v) = h·, vi. (That is, for f ∈ V ∗ , φ(v)(f ) = hf, vi = f (v).)
Show that this map is injective (you might need part (a)). Then show the map is an isomorphism
by counting dimensions (carefully).
Remark: Once you write out the (short) proof, it might seem that we haven’t really done
anything in part (b). In fact, the isomorphism we constructed is canonical. If we have time, we’ll
see precisely what this means in a future HW. For now, roughly, a canonical isomorphism means
that it is independent of any choices. Contrast this with previous work. We (now) know that in
fact V ∗ ∼ = V . This is because we constructed a basis for V ∗ from a basis of V — but note this
isomorphism depended on the choices made.

Solution. (a) The set {v} is linearly independent: extend it to a basis {v1 = v, v2 , v3 , . . . , vn }. By
the last problem set, there is a functional f 1 ∈ V ∗ such that f1 (vi ) = 1 if i = 1 and 0 otherwise.
So f 1 (v) = f 1 (v1 ) = 1 6= 0.
(b) We note that our map φ is linear because hf, vi is linear in v. Now we show that its kernel is
trivial. Consider any nonzero v. Then by (a), there exists an f such that hf, vi = 6 0, so φ(v)(f ) 6= 0.
This means that φ(v) is not the zero map for any nonzero v, and Ker φ must contain only the zero
vector. Hence φ is injective.
Now we count dimensions. From the last problem set we know that dim V ∗ = dim V , so also
dim V ∗∗ = dim V ∗ = dim V . Also, dim V = dim Ker φ + dim Im φ = dim Im φ by injectivity. So
dim Ker φ is a subspace of V with the same dimension as V : that must be all of V , so φ is surjective
as well as injective, hence an isomorphism.

(3) Recall from class (Monday 21st November), that we defined the annihilator of a subset of a
vector space. Given a vector space V and subset S (note: S does not have to be a subspace of V !),
then the annihilator S 0 := {f ∈ V ∗ : hf, si ≡ 0, ∀s ∈ S}.
(a) If W is an m-dimensional subspace of an n-dimensional vector space V , then show that W 0
is an (n − m)-dimensional subspace of V ∗ .
(b) Show that if W is a subspace of a finite dimensional vector space V , then W 00 = (W 0 )0 = W .
(c) Show that if S is any subset of a finite-dimensional vector space V , then S 00 is the same as
the subspace of V spanned by vectors in S.
(d) If S and T are subspaces of a finite-dimensional vector space V and if S ⊂ T , then T 0 ⊂ S 0 .
(e) If W and U are subspaces of a finite-dimensional vector space V , then (W ∩ U )0 = W 0 + U 0
and (W + U )0 = W 0 ∩ U 0 .

2
Solution. (a) Let v1 , . . . , vm be a basis of W , and extend it to a basis v1 , . . . , vn of V . Consider the
map φ : V ∗ → Fm given by φ(f ) = (f (v1 ), f (v2 ), . . . , f (vm )). The kernel of φ is exactly W 0 : for any
f ∈ W 0 , all the entries of φ(f ) are zero because v1 , v2 , . . . , vm ∈ Ker f . Conversely, if φ(f ) = 0, for
any v ∈ V , write v = a1 v1 + a2 v2 + · · · + an vn . Then f (v) = a1 f (v1 ) + a2 f (v2 ) + · · · + an f (vn ) = 0.
So W 0 = Ker φ is a subspace (we’ve shown before that the kernel of any linear map is a subspace.)
We now need to check its dimension.
Our map φ is surjective by problem set 9: given any (a1 , a2 , . . . , am ) ∈ Fm we can find f ∈ V ∗
such that f (vi ) = ai for i = 1, . . . , m, and f (vi ) is whatever we want for i > m. Then this f satisfies
φ(f ) = (a1 , a2 , . . . , am ): because our a’s could be anything, the map is surjective. So dim Im φ =
dim Fm = m. Also n = dim V = dim Ker φ + dim Im φ = dim W 0 + m, so dim W 0 = n − m.
(b) Counting dimensions again! First we prove that W ⊂ W 00 . This is straight from the
definition: for w ∈ W and any f ∈ W 0 , hw, f i = hf, wi = 0, so w ∈ W 00 (note that we are
identifying w with its canonical image, as given by problem 2, in V ∗∗ . This means that W ⊂ W 00 .
On the other hand, if dim W = m, dim W 0 = n − m, and dim W 00 = dim V ∗ − dim W 0 =
n − (n − m) = m = dim W . So the two spaces have the same dimension, and one is contained in
the other. They must be equal.
(c) We show that S 0 = (Span S)0 . The inclusion (Span S)0 ⊂ S 0 is clear, because if f ∈
(Span S)0 , for any v ∈ S, v ∈ Span S also, so f (v) = 0, and f ∈ S 0 as well. For the other one, let
f ∈ S 0 , and take any v ∈ Span S. By definition, v is a linear combination v = a1 v1 +a2 v2 +· · ·+an vn
of vectors v1 , . . . , vn of S. By linearity, f (v) = a1 f (v1 ) + a2 f (v2 ) + · · · + an f (vn ) = 0 because f
annihilates each vi ∈ S. So f ∈ (Span S)0 also, S 0 ⊂ (Span S)0 , and the two sets are equal.
Now we get to apply (b). Taking annihilators of S 0 = (Span S)0 , S 00 = (Span S)00 = Span S,
which is what we wanted.
(d) Because S is a subset of T , anything in V ∗ that sends all elements of T to 0 also sends all
elements of S to 0. So T 0 ⊂ S 0 .
(e) We first prove the easier half: (W + U )0 = W 0 ∩ U 0 . This is double inclusion. We have
U, W ⊂ W + U , so by (d), (W + U )0 ⊂ W 0 , U 0 , so (W + U )0 is also a subset of the intersection
W 0 ∩ U 0 . That proves one inclusion, (W + U )0 ⊂ W 0 ∩ U 0 . For the reverse inclusion, suppose that
f ∈ W 0 ∩ U 0 . Then for all u ∈ U , f (u) = 0, and for all w ∈ W , f (w) = 0. By linearity, then, for
all u + w ∈ U + W , f (u + w) = f (u) + f (w) = 0, so f ∈ (W + U )0 . So W 0 ∩ U 0 ⊂ (W + U )0 . By
double inclusion, W 0 ∩ U 0 = (W + U )0 .
It’s possible, but a bit messy, to do the second part with bases or counting dimensions. Instead,
we’ll use a trick. In the result we’ve just proved, W 0 ∩ U 0 = (W + U )0 , we replace W with W 0
and U with U 0 . This gives us (W 0 + U 0 )0 = W 00 ∩ U 00 = W ∩ U . Take annihilators of both sides:
(W 0 + U 0 )00 = (W ∩ U )0 . By (b), this is just W 0 + U 0 = (W ∩ U )0 .

(4) (a) Read Chapter 4 of Axler

(b) Problem 5 page 73 of Axler. (I can think of two different proofs - can you? Note: just
submit one for grading.)

Solution. (a) The solution is left as an exercise to the reader.

(b) By Axler 4.10, if λ is a root of f , so is λ̄. This means that the non-real roots of f come in
complex conjugate pairs. So if f had only non-real roots, it would have an even number of complex
roots. By the Fundamental Theorem of Algebra, the number of complex roots of f equals the

3
degree of f , so f would have even degree. Conversely, if f has odd degree, it must instead have
some real root.