0% found this document useful (0 votes)
17 views7 pages

Lecture Note-2

Linear algebra

Uploaded by

Bogda Balasi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
17 views7 pages

Lecture Note-2

Linear algebra

Uploaded by

Bogda Balasi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

SECTION 7.

C Positive Operators and Isometries 225

7.C Positive Operators and Isometries


Positive Operators

7.31 Definition positive operator


An operator T 2 L.V / is called positive if T is self-adjoint and
hT v; vi  0
for all v 2 V.

If V is a complex vector space, then the requirement that T is self-adjoint


can be dropped from the definition above (by 7.15).

7.32 Example positive operators

(a) If U is a subspace of V, then the orthogonal projection PU is a positive


operator, as you should verify.

(b) If T 2 L.V / is self-adjoint and b; c 2 R are such that b 2 < 4c, then
T 2 C bT C cI is a positive operator, as shown by the proof of 7.26.

7.33 Definition square root


An operator R is called a square root of an operator T if R2 D T.

7.34 Example If T 2 L.F3 / is defined by T .z1 ; z2 ; z3 / D .z3 ; 0; 0/,


then the operator R 2 L.F3 / defined by R.z1 ; z2 ; z3 / D .z2 ; z3 ; 0/ is a
square root of T.

The characterizations of the positive The positive operators correspond


operators in the next result correspond to the numbers Œ0; 1/, so better
to characterizations of the nonnegative terminology would use the term
numbers among C. Specifically, a com- nonnegative instead of positive.
plex number z is nonnegative if and However, operator theorists consis-
only if it has a nonnegative square root, tently call these the positive opera-
tors, so we will follow that custom.
corresponding to condition (c). Also,
z is nonnegative if and only if it has a real square root, corresponding to
condition (d). Finally, z is nonnegative if and only if there exists a complex
number w such that z D ww, N corresponding to condition (e).
226 CHAPTER 7 Operators on Inner Product Spaces

7.35 Characterization of positive operators


Let T 2 L.V /. Then the following are equivalent:

(a) T is positive;

(b) T is self-adjoint and all the eigenvalues of T are nonnegative;

(c) T has a positive square root;

(d) T has a self-adjoint square root;

(e) there exists an operator R 2 L.V / such that T D R R.

Proof We will prove that (a) ) (b) ) (c) ) (d) ) (e) ) (a).
First suppose (a) holds, so that T is positive. Obviously T is self-adjoint
(by the definition of a positive operator). To prove the other condition in (b),
suppose  is an eigenvalue of T. Let v be an eigenvector of T corresponding
to . Then
0  hT v; vi D hv; vi D hv; vi:
Thus  is a nonnegative number. Hence (b) holds.
Now suppose (b) holds, so that T is self-adjoint and all the eigenvalues
of T are nonnegative. By the Spectral Theorem (7.24 and 7.29), there is
an orthonormal basis e1 ; : : : ; en of V consisting of eigenvectors of T. Let
1 ; : : : ; n be the eigenvalues of T corresponding to e1 ; : : : ; en ,; thus each
j is a nonnegative number. Let R be the linear map from V to V such that
q
Rej D j ej
for j D 1; : : : ; n (see 3.5). Then R is a positive operator, as you should verify.
Furthermore, R2 ej D j ej D T ej for each j , which implies that R2 D T.
Thus R is a positive square root of T. Hence (c) holds.
Clearly (c) implies (d) (because, by definition, every positive operator is
self-adjoint).
Now suppose (d) holds, meaning that there exists a self-adjoint operator
R on V such that T D R2 . Then T D R R (because R D R). Hence (e)
holds.
Finally, suppose (e) holds. Let R 2 L.V / be such that T D R R. Then
T  D .R R/ D R .R / D R R D T. Hence T is self-adjoint. To
complete the proof that (a) holds, note that
hT v; vi D hR Rv; vi D hRv; Rvi  0
for every v 2 V. Thus T is positive.
SECTION 7.C Positive Operators and Isometries 227

Each nonnegative number has a Some mathematicians also use the


unique nonnegative square root. The term positive semidefinite opera-
next result shows that positive operators tor, which means the same as posi-
enjoy a similar property. tive operator.

7.36 Each positive operator has only one positive square root
Every positive operator on V has a unique positive square root.

Proof Suppose T 2 L.V / is positive. A positive operator can have in-


Suppose v 2 V is an eigenvector of T. finitely many square roots (al-
Thus there exists   0 such that T v D though only one of them can be
v. positive). For example, the identity
Let R be a positive square operator on V has infinitely many
p root of T. square roots if dim V > 1.
We will prove that Rv D v. This
will imply that the behavior of R on the eigenvectors of T is uniquely deter-
mined. Because there is a basis of V consisting of eigenvectors of T (by the
Spectral Theorem), this will p imply that R is uniquely determined.
To prove that Rv D v, note that the Spectral Theorem asserts that
there is an orthonormal basis e1 ; : : : ; en of V consisting of eigenvectors of R.
Because R is a positive operator, all its eigenvalues are nonnegative. p Thus
there exist nonnegative numbers 1 ; : : : ; n such that Rej D j ej for
j D 1; : : : ; n.
Because e1 ; : : : ; en is a basis of V, we can write
v D a1 e1 C    C an en
for some numbers a1 ; : : : ; an 2 F. Thus
p p
Rv D a1 1 e1 C    C an n en
and hence
R2 v D a1 1 e1 C    C an n en :
But R2 D T, and T v D v. Thus the equation above implies
a1 e1 C    C an en D a1 1 e1 C    C an n en :
The equation above implies that aj .  j / D 0 for j D 1; : : : ; n. Hence
X
vD aj ej ;
fj W j Dg
and thus X p p
Rv D aj ej D v;
fj W j Dg
as desired.
228 CHAPTER 7 Operators on Inner Product Spaces

Isometries
Operators that preserve norms are sufficiently important to deserve a name:

7.37 Definition isometry

 An operator S 2 L.V / is called an isometry if


kS vk D kvk
for all v 2 V.

 In other words, an operator is an isometry if it preserves norms.

The Greek word isos means equal; For example, I is an isometry


the Greek word metron means whenever  2 F satisfies jj D 1. We
measure. Thus isometry literally will see soon that if F D C, then the
means equal measure. next example includes all isometries.

7.38 Example Suppose 1 ; : : : ; n are scalars with absolute value 1 and


S 2 L.V / satisfies Sej D j ej for some orthonormal basis e1 ; : : : ; en of V.
Show that S is an isometry.

Solution Suppose v 2 V. Then

7.39 v D hv; e1 ie1 C    C hv; en ien

and

7.40 kvk2 D jhv; e1 ij2 C    C jhv; en ij2 ;

where we have used 6.30. Applying S to both sides of 7.39 gives

S v D hv; e1 iSe1 C    C hv; en iSen


D 1 hv; e1 ie1 C    C n hv; en ien :

The last equation, along with the equation jj j D 1, shows that

7.41 kS vk2 D jhv; e1 ij2 C    C jhv; en ij2 :

Comparing 7.40 and 7.41 shows that kvk D kS vk. In other words, S is an
isometry.
SECTION 7.C Positive Operators and Isometries 229

The next result provides several con- An isometry on a real inner product
ditions that are equivalent to being an space is often called an orthogonal
isometry. The equivalence of (a) and (b) operator. An isometry on a com-
shows that an operator is an isometry if plex inner product space is often
and only if it preserves inner products. called a unitary operator. We use
The equivalence of (a) and (c) [or (d)] the term isometry so that our re-
sults can apply to both real and
shows that an operator is an isometry
complex inner product spaces.
if and only if the list of columns of its
matrix with respect to every [or some] basis is orthonormal. Exercise 10
implies that in the previous sentence we can replace “columns” with “rows”.

7.42 Characterization of isometries


Suppose S 2 L.V /. Then the following are equivalent:

(a) S is an isometry;

(b) hS u; Svi D hu; vi for all u; v 2 V ;

(c) Se1 ; : : : ; Sen is orthonormal for every orthonormal list of vectors


e1 ; : : : ; en in V ;

(d) there exists an orthonormal basis e1 ; : : : ; en of V such that


Se1 ; : : : ; Sen is orthonormal;

(e) S S D I ;

(f) SS D I ;

(g) S  is an isometry;

(h) S is invertible and S 1 D S .

Proof First suppose (a) holds, so S is an isometry. Exercises 19 and 20 in


Section 6.A show that inner products can be computed from norms. Because
S preserves norms, this implies that S preserves inner products, and hence
(b) holds. More precisely, if V is a real inner product space, then for every
u; v 2 V we have
hS u; Svi D .kS u C S vk2  kS u  S vk2 /=4
D .kS.u C v/k2  kS.u  v/k2 /=4
D .ku C vk2  ku  vk2 /=4
D hu; vi;
230 CHAPTER 7 Operators on Inner Product Spaces

where the first equality comes from Exercise 19 in Section 6.A, the second
equality comes from the linearity of S , the third equality holds because S is an
isometry, and the last equality again comes from Exercise 19 in Section 6.A.
If V is a complex inner product space, then use Exercise 20 in Section 6.A
instead of Exercise 19 to obtain the same conclusion. In either case, we see
that (b) holds.
Now suppose (b) holds, so S preserves inner products. Suppose that
e1 ; : : : ; en is an orthonormal list of vectors in V. Then we see that the list
Se1 ; : : : ; Sen is orthonormal because hSej ; Sek i D hej ; ek i. Thus (c) holds.
Clearly (c) implies (d).
Now suppose (d) holds. Let e1 ; : : : ; en be an orthonormal basis of V such
that Se1 ; : : : ; Sen is orthonormal. Thus

hS  Sej ; ek i D hej ; ek i

for j; k D 1; : : : ; n [because the term on the left equals hSej ; Sek i and
.Se1 ; : : : ; Sen / is orthonormal]. All vectors u; v 2 V can be written as
linear combinations of e1 ; : : : ; en , and thus the equation above implies that
hS  Su; vi D hu; vi. Hence S  S D I ; in other words, (e) holds.
Now suppose (e) holds, so that S  S D I. In general, an operator S need
not commute with S . However, S  S D I if and only if S S  D I ; this is a
special case of Exercise 10 in Section 3.D. Thus S S  D I, showing that (f)
holds.
Now suppose (f) holds, so S S  D I. If v 2 V, then

kS  vk2 D hS  v; S  vi D hS S  v; vi D hv; vi D kvk2 :

Thus S  is an isometry, showing that (g) holds.


Now suppose (g) holds, so S  is an isometry. We know that (a) ) (e) and
(a) ) (f) because we have shown (a) ) (b) ) (c) ) (d) ) (e) ) (f). Using
the implications (a) ) (e) and (a) ) (f) but with S replaced with S  [and
using the equation .S  / D S], we conclude that S S  D I and S  S D I.
Thus S is invertible and S 1 D S  ; in other words, (h) holds.
Now suppose (h) holds, so S is invertible and S 1 D S . Thus S  S D I.
If v 2 V, then

kSvk2 D hSv; Svi D hS  S v; vi D hv; vi D kvk2 :

Thus S is an isometry, showing that (a) holds.


We have shown (a) ) (b) ) (c) ) (d) ) (e) ) (f) ) (g) ) (h) ) (a),
completing the proof.
SECTION 7.C Positive Operators and Isometries 231

The previous result shows that every isometry is normal [see (a), (e), and
(f) of 7.42]. Thus characterizations of normal operators can be used to give
descriptions of isometries. We do this in the next result in the complex case
and in Chapter 9 in the real case (see 9.36).

7.43 Description of isometries when F D C


Suppose V is a complex inner product space and S 2 L.V /. Then the
following are equivalent:

(a) S is an isometry.

(b) There is an orthonormal basis of V consisting of eigenvectors of S


whose corresponding eigenvalues all have absolute value 1.

Proof We have already shown (see Example 7.38) that (b) implies (a).
To prove the other direction, suppose (a) holds, so S is an isometry. By the
Complex Spectral Theorem (7.24), there is an orthonormal basis e1 ; : : : ; en
of V consisting of eigenvectors of S . For j 2 f1; : : : ; ng, let j be the
eigenvalue corresponding to ej . Then

jj j D kj ej k D kSej k D kej k D 1:

Thus each eigenvalue of S has absolute value 1, completing the proof.

EXERCISES 7.C

1 Prove or give a counterexample: If T 2 L.V / is self-adjoint and there


exists an orthonormal basis e1 ; : : : ; en of V such that hT ej ; ej i  0 for
each j , then T is a positive operator.

2 Suppose T is a positive operator on V. Suppose v; w 2 V are such that

Tv D w and T w D v:

Prove that v D w.

3 Suppose T is a positive operator on V and U is a subspace of V invariant


under T. Prove that T jU 2 L.U / is a positive operator on U.

4 Suppose T 2 L.V; W /. Prove that T  T is a positive operator on V and


T T  is a positive operator on W.

You might also like