TA class
Pu-Zhao, Kow
2.11.2016
Throughout this lecture note, we assume that the linear spaces are nite-dimensional (thus a vector
space), and of over a eld
#2
F=C
, unless stated.
Revised homework
Suppose
of
1.
T L(V )
C.
Prove that
is an eigenvalue of
if and only if
is an eigenvalue
Although it is similar, you still have to write out the converse. Actually, note that
and
2.
and
T .
(T ) = T ,
In real case, the statement of the problem says that
general,
() =
you can prove the converse rapidly.
and
they do not share the same eigenvectors.
share same eigenvalue, but in
For example, consider the real matrix
A=
1
0
2
3
,
which has eigenvalues and corresponding eigenvectors as follows:
rst eigenvalue:
1 = 3and
its correspoding eigenvector is
v1 = (1 1)T
second eigenvalue:
2 = 1and
its correspoding eigenvector is
v2 = (1 0)T .
But, note that its transpost (i.e. adjoint) is the real matrix
A=
1
2
0
3
,
which has eigenvalues and corresponding eigenvectors as follows:
rst eigenvalue:
second eigenvalue:
Note that
1
1 =
and
1 = 3and
2 = 1and
2,
2 =
its correspoding eigenvector is
v1 = (1 0)T
its correspoding eigenvector is
v2 = (1 1)T .
but
null(A
1 I) = span{v1 } span{
1 I) null(AT
v1 } = {0}
null(A
2 I) = span{v2 } span{
2 I) null(AT
v2 } = {0},
which shows that they do not share the same eigenvalues (not even one!).
3.
Beware of the notation. Note that
ker(T ) = N (T ) = null(T ),
dim(ker(T )) = nullity(T ) = dim(N (T )) = dim(null(T )).
1
TA class - 2.11.2016
4.
T v = v ,
Again, if you want to use the relation
T,
5.
Linear Algebra
with corresponding eigenvector
not arbitrary in V .
v .
you must state that is an eigenvalue of
In general, null(T
I) 6= V ,
which means that
is
This result is quite important. I will prove it:
Proof.
uV,
Suppose that
is an eigenvalue of
T,
with corresponding eigenvector
v.
Given any
we have
hT u u, vi = hT u, vi hu, vi = hu, T vi hu, vi = hu, T vi = hu, 0i = 0.
This shows that T u u null(T I)
. Since u V is arbitrary, this also shows that
range(T I) null(T I)
. Since is an eigenvalue, thus null(T I) 6= {0}, and
thus null(T I)
6= V , which means that range(T I) 6= V , shows that T I is
not surjective, which means that is an eigenvalue.
#2
Exercise 7B (due 2.11.2016)
Suppose that
and
is a self-adjoint operator on a nite-dimensional inner product space
are the only eigenvalues of
T.
Prove that
and that
T 2 5T + 6I = 0.
=E(2,T )
=E(3,T )
}|
{ z
}|
{
Proof. By Exercise
#4 we shall do later, rst we note that V = null(T 2I) null(T 3I),
null(T 2I)
= null(T 3I). Given any v, w V , then
z
and
h(T 2 5T + 6I)v, wi = hT 2 v, wi 5hT v, wi + 6hv, wi
= hT v, T wi 5hT v, wi + 6hv, wi
Since
is self adjoint (hence normal)
= v w hv, wi 5v hv, wi + 6hv, wi
= v w 5v + 6 hv, wi,
v , w {2, 3},
If v and w are not in
v and w are in the same
eigenspace, thus v = w , and clearly v w 5v + 6 = (v 2)(v 3) = 0. (How about the
general case v = e1 + e2 , where e1 E(2, T ) and general w ? Think about it.)
where
which are eigenvalues of
the same eigenspace, then
#4
hv, wi = 0,
(This is an important fact!) Suppose
and
w,
respectively.
which completes our proof. If
F=C
and
T L(V ).
Prove that
all pairs of eigenvectors corresponding to distinct eigenvalues of
is normal if and only if
are orthogonal and
V = E(1 , T ) E(m , T ),
where
Proof.
1 , , m
denote the distinct eigenvalues of
T.
This is an immediately consequence of (Complex) Spectral Theorem, try to relate with
Proposition 7.24(c).
#6
(1)
(This is an important fact!) Prove that a normal operator on a complex inner product space is
self-adjoint if and only if all its eigenvalues are real.
Proof.
v.
Suppose that
is self-adjoint. Let
be an eigenvalue of T , with corresponding eigenvector
Note that
hv, vi = hT v, vi = hv, T vi = hv, vi.
Since
v 6= 0,
hence
= ,
i.e.
(2)
R.
Conversely, suppose that all eigenalues of a normal operator
Theorem, there exists an orthonormal basis
{ei }ni=1
is real. Using (Complex) Spectral
with eigenvalue
i .
Using Proposition 7.21,
we know that
=0,
h(T T )ei , ej i =
Pu-Zhao Kow
since
i R
z }| {
(i i ) hei , ej i = 0
2
for all
i, j = 1, , n.
(3)
L16057015
TA class - 2.11.2016
Linear Algebra
T T = 0,
This concludes that
#4
T = T .
i.e.
Exercise 7C (due 2.11.2016)
Suppose
on
T L(V, W ).
T T
Prove that
is a positive operator on
and
TT
is a positive operator
W.
Proof.
Given any
v V,
thus
hT T v, vi = hT v, T vi 0,
which shows that
T T
is positive. Similarly, given any
(4)
w W,
thus
hT T w, wi = hT w, T wi 0,
which shows that
#8
Suppose
TT
T L(V ).
For
is positive.
u, v V ,
(5)
dene
hu, viT
by
hu, viT = hT u, vi.
Prove that
h, iT
V if and
h, i).
is an inner product on
(with respect to the original inner product
Proof.
Suppose that
h, iT
(6)
only if
is an invertible positive operator
is an inner product. Thus we have
hv, viT = hT v, vi 0,
which shows that
any
v V,
is positive. We now shows that
(7)
ker(T ) = {0}.
Let
x ker(T ),
thus, given
we have
hx, viT = hT x, vi = h0, vi = 0,
and by arbitrariness of
x = 0,
which shows that
and the assumption that
h, iT
(8)
is an inner product, we conclude that
is injective, and hence (bijective) invertible.
The converse is just straight-forward verication of denition of inner product.
Positivity. hv, viT = hT v, vi 0, since T is positive.
Deniteness. Suppose that hv, viT = hT v, vi = 0. Since
contradiction will appeared. (Hint: use the properties
T is invertible,
ker(T ) = {0}).
thus
v = 0,
or else
Additivity. hu + w, viT = hT (u + w), vi = hT u + T w, vi = hT u, vi + hT w, vi = hu, viT + hw, viT .
Homogeneity. hu, viT = hT (u), vi = hT u, vi = hT u, vi = hu, viT .
Conjugate_Symmetry. hu, viT = hT u, vi = hv, T ui = hu, viT .
#9
Prove or disprove: the identity operator on
Proof.
Consider
F = C.
d
c
1d2
c
generally which are not self-adjoint.
Suppose
S L(V ).
has innitely many self-adjoint square roots.
The general form of square roots of identity are
#10
F2
and
d
1d2
c
c
d
,
(9)
Prove that the following are equivalent:
(a) S is an isometry;
(b) hS u, S vi = hu, vi for all u, v V ;
(c) S e1 , , S em is an orthonormal list for every orthonormal list of vectors e1 , , em in V ;
(d) S e1 , , S en is an orthonormal basis for some orthonormal basis e1 , , en of V .
Pu-Zhao Kow
L16057015
TA class - 2.11.2016
Proof.
Linear Algebra
(a)(b): These implications are trivial using Proposition 7.42.
(b)(c) and (c)(d): These implications are trivial.
(d)(b): Since
{e1 , , en }
is an orthonormal basis of
v=
n
X
cvi ei
V,
thus given any
v V,
we can write
ci F.
for some
(10)
i=1
Thus
Sv =
n
X
cvi S .
(11)
i=1
Note that, by orthonomality of
S e1 , , S en ,
we have
n
n
n
X
X
X
hS u, S vi = h
cui S ei ,
cvi S ei i =
cui cvi .
i=1
Also, by orthonomality of
e1 , , e n ,
i=1
(12)
i=1
we have
n
n
n
X
X
X
hu, vi = h
cui ei ,
cvi ei i =
cui cvi ,
i=1
i=1
and the result immediately follows from (12) and (13).
#14
Let
(13)
i=1
be the second derivative operator in Exercise 21 in Section 7A. Show that
is a positive
operator.
Proof.
Recall
V =
span{1, cos x,
, cos nx, sin x, , sin nx}.
Note that
is span by an or-
thogonal basis (I assume that you all done in last homework, need NOT to to show). Also note
that
T cos mx = m2 cos mx
thus its positivity is clear.
Pu-Zhao Kow
T sin mx = m2 sin mx,
(14)
L16057015