1 7A

1 Proof. We need to define T ∗ such that ⟨v, T ∗ w⟩ = z1 x2 + z2 x3 + ... + zn−1 xn = ⟨T v, w⟩. Let v
= (z1 , ..., zn ) and w = (x1 , ..., xn ) be in Fn . Then define T ∗ w = (x2 , x3 , ..., xn , 0). Observe that
⟨v, T ∗ w⟩ = z1 x2 + z2 x3 + ... + zn−1 xn , as desired. ■

2 Proof. Let T ∈ L(V, W ). Suppose T = 0. Then ⟨T v, w⟩ = ⟨0, w⟩ = 0 for every v in V and w in W.

Then by the definition of adjoint, ⟨v, T ∗ w⟩ = 0. In general, V will not be trivial, which means this
implies that T ∗ w = 0 for all w in W. Thus, T ∗ = 0. It is obvious that the converse is true as well.

Suppose T ∗ = 0. Then T ∗ T v = T ∗ (T v) = 0 for every v in V.

Suppose T ∗ T = 0. Then either T ∗ or T must be the zero map. In either case, T T ∗ = 0, as desired.

Finally, suppose T T ∗ = 0. Then either T or T ∗ must be the zero map. As we proved in the first
paragraph, one being 0 implies the other is 0 as well. Therefore, both T and T ∗ are 0, specifically T
in this instance. We have shown that T = 0 =⇒ T ∗ = 0 =⇒ T ∗ T = 0 =⇒ T T ∗ = 0 =⇒ T = 0.
Thus, we have completed the proof, and shown that these are all equivalent. ■

3 Proof. Let T ∈ L(V ) and λ ∈ F.

( =⇒ ) Suppose λ is an eigenvalue of T. Then Tv = λv for some v in V. Then

⟨T v, w⟩ = ⟨λv, w⟩ = λ⟨v, w⟩ = ⟨v, λw⟩ = ⟨v, T ∗ w⟩

Therefore, T ∗ w = λw, which shows that λ is an eigenvalue of T ∗ .

( ⇐= ) Similarly in the converse direction, we have that

⟨v, T ∗ ⟩ = ⟨v, λ⟩ = λ⟨v, w⟩ = λ⟨v, w⟩ = ⟨λv, w⟩ = ⟨T v, w⟩

Thus, Tv = λv, as was to be shown. ■

4 Proof. Let T ∈ L(V ) and let U be a subspace of V.

( =⇒ ) Suppose U is T-invariant. Then ⟨T u, w⟩ = 0 for all u in U and w in U ⊥ , since we know that

Tu is in U. This tells us that ⟨u, T ∗ w⟩ = 0, which means T ∗ w is orthogonal to every u in U. In other
words, T ∗ w ∈ U ⊥ , meaning U ⊥ is invariant under T ∗ .

( ⇐= ) Suppose U ⊥ is invariant under T ∗ . Then for every w in U ⊥ and u in U, ⟨u, T ∗ w⟩ = 0 = ⟨T u, w⟩.

Thus, Tu is orthogonal to every w in U ⊥ , meaning Tu is in U. thus, U is T-invariant, completing the
proof. ■

6 Proof. ■

1
 7 Proof. (a) ( =⇒ ) Suppose T is injective. Then null T = (rangeT ∗ )⊥ = {0}. Then range T ∗ = V,
which means T ∗ is surjective.

( ⇐= ) Suppose T ∗ is surjective. Then range T ∗ = (nullT )⊥ = V . Then null T = {0}, and so T is

injective, as desired.

(b) ( =⇒ ) Suppose T is surjective. Then range T (nullT ∗ )⊥ = W . That is, null T ∗ = 0. Therefore,
T ∗ is injective.

( ⇐= ) Suppose T ∗ is injective. Then null T ∗ = (rangeT )⊥ = 0. Then range T = W. Thus, T is

surjective. ■

12 Proof. (a) In solving for T ∗ q = T ∗ (b0 + b1 x + b2 x2 ) = c0 + c1 x + c2 x2 , we find that both c0 and c2

are 0, and (a0 + a1 + a2 )c1 = (b0 + b1 + b2 )a1 . If (a0 + a1 + a2 ) = 0, then a1 must be 0 by this
equation (the b’s are a random polynomial in this case). This implies that Tp = 0, so p is in null T.
Assuming p is nonzero, we find that T ∗ q = 0. Otherwise, assuming (a0 + a1 + a2 ) is nonzero, we have
that c1 = (ba0 0+b1 +b2 )a1 ∗ 2
+a1 +a2 , so T (b0 + b1 x + b2 x ) =
(b0 +b1 +b2 )a1
a0 +a1 +a2 x ̸= b1 x in general. Therefore, T is not
self-adjoint.
(b) This is not a contradiction because 1,x,x2 is not an orthonormal basis, meaning the matrix of the
adjoint of T is not the conjugate transpose. Therefore, this is not a contradiction. ■

13 Proof. Let T ∈ L(V ) be invertible. (a) ( =⇒ ) Suppose T is self-adjoint. Then T = T ∗ . Since T is

invertible, we know that T ∗ is invertible and that (T ∗ )−1 = (T −1 )∗ . However, since T is self-adjoint,
we can substitute T into the previous equation to obtain T −1 = (T −1 )∗ . Therefore, T −1‘ is self-adjoint.

( ⇐= ) Suppose T −1 is self-adjoint. Once again, we know that ((T −1 )∗ )−1 = ((T −1 )−1 )∗ = T ∗ . By our
hyptothesis, the first expression equals (T −1 )−1 = T . Thus, we have T = T ∗ , and our proof is complete.

(b) ( =⇒ ) Suppose T is normal. Then T T ∗ = T ∗ T . Then

T −1 (T −1 )∗ = T −1 (T ∗ )−1 = (T ∗ T )−1 = (T T ∗ )−1 = (T ∗ )−1 T −1 = (T −1 )∗ T −1

where the third equality is given by our inital hypothesis that T is normal. Therefore, T −1 is normal.
( ⇐= ) Suppose T −1 is normal. Then T T ∗ = (T −1 )−1 ((T ∗ )−1 )−1 = ((T ∗ )−1 T −1 )−1 = ((T −1 )∗ T −1 )−1 =
(T −1 (T −1 )∗ )−1 = ((T ∗ )−1 )−1 (T −1 )−1 = T ∗ T . Therefore, T is normal. ■
0 1



16 Proof. Consider two normal operators S and T on R2 such that M (S) = −1 01
0 and M (T ) = 1 0 .


These operators are both normal, however, M (S + T ) = 00 20 which is not normal. Therefore, the set
of normal operators on V is not closed under addition, so it is not a subspace of L(V ). ■

17 Proof. ( =⇒ )Suppose there is a U such that P = PU . Let v1 , v2 be elements of V. Then v1 = u1 + w1

and v2 = u2 + w2 such that u1 , u2 ∈ U and w1 , w2 ∈ U ⊥ . Then ⟨P v1 , v2 ⟩ = ⟨P (u1 + w1 ), v2 ⟩ =
⟨P u1 , v2 ⟩ + ⟨P w1 , v2 ⟩ = ⟨u1 , u2 + w2 ⟩ = ⟨u1 , u2 ⟩ + ⟨u1 , w2 ⟩ = ⟨u1 , u2 ⟩ = ⟨v1 , P v2 ⟩. Thus, P is self-
adjoint.

2
 ( ⇐= )Suppose P is self-adjoint. Then null P is orthogonal to range P. In 6C exercise 9, we showed that
this implies there exists a subspace U of V, namely range P, such that P = PU , and we are done. ■

20 Proof. Suppose the assumptions from the problem are true. Then v is eigenvector of T with a corre-
sponding eigenvalue of 3, and w is an eigenvector of T with a corresponding eigenvalue of 4. Since T
is normal, we can infer that v and w are orthogonal. With this in mind, ∥T (v + w)∥2 = ∥T v + T w∥2 =
∥3v + 4w∥2 = ⟨3v + 4w, 3v + 4w⟩ = ⟨3v, 3v⟩ + ⟨3v, 4w⟩ + ⟨4w, 3v⟩ + ⟨4w, 4w⟩ = 9∥v∥2 + 0 + 0 + 16∥w∥2 =
36 + 64 = 100. Thus, ∥T (v + w)∥2 = 100, so ∥T (v + w)∥ = 10. ■

23 Proof. First we will address null T = null T k . First we have the case where k = 2. We claim that since
T is normal, null T ∗ T = null T 2 . Let v be in null T ∗ T . Then T ∗ T v = T ∗ (T v) = 0. Then Tv must be
in null T as well. Therefore, T(Tv) = T 2 v = 0, so v is in null T 2 . Reverse the order of steps to see
inclusion in the other direction as well. Since null T = null T ∗ T , we have that null T = null T 2 .

Now assume the statement holds for any positive integer less than or equal to k. Then null T = null
T 2 = ... = null T k . Let v be in null T k+1 . Then T k+1 v = 0. Then T k (T v) = 0, which means Tv is in
null T k . By our induction hyptothesis, Tv must be in null T. Thus, T(Tv) = T 2 v = 0, which means v
is in null T 2 , and by extension that v is in null T. Inclusion in the other direction is trivial. Thus, we
conclue that null T = null T k .

Now we will look at the claim range T = range T k . We know range T ∗ T = range T ∗ . First we will show
that range T = range T 2 . Let v be in range T = range T ∗ = range T ∗ T . Then v = T ∗ T w = T T ∗ w
for some w in V. Then obviously, T ∗ w is in range T ∗ which means T ∗ w = T w1 for some w1 in V.
Therefore, v = T T w1 = T 2 w1 , so v is in range T 2 . Inclusion in the other direction is trivial. Thus,
range T = range T 2 .

Now let v be in range T. Then v = T w1 = for some w1 in V. Thus, v = T 2 w2 = T (T w2 ) for some w2

in V. Then we have v = T (T 2 w3 ), and so on. By repeated applications of this process, it is clear that
v is in range T k for any positive integer k. Also, if v is in range T k , then v = T k w = T (T k−1 w) for
some w in V, and so v is in range T. Thus, range T = range T k . ■

range T. Also, range T = (null T ∗ )⊥ = (null T)⊥ . Since

L
26 Proof. Since T is normal, V = null T
T(1,1,1) = (2,2,2) ̸= 0, (1,1,1) must be orthogonal to (z1 , z2 , z3 ). Assuming our inner product space is
defined with the Euclidean inner product, we have ⟨(z1 , z2 , z3 ), (1, 1, 1)⟩ = z1 + z2 + z3 = 0, completing
the proof. ■

28 Proof. Since T = D2 we have that T = DD = (-D)(-D) = D∗ D∗ = (DD)∗ = T ∗ . Thus T is self-

adjoint. ■