Lecture 12: Diagonalization
A square matrix D is called diagonal if all but diagonal entries are zero:
2 3
a1 0 ¢¢¢ 0
6 0 a2 ¢¢¢ 0 7
D=6 4¢ ¢ ¢ ¢ ¢ ¢
7 : (1)
¢ ¢ ¢ ¢ ¢ ¢5
0 0 ¢ ¢ ¢ an n£n
Diagonal matrices are the simplest matrices that are basically equivalent to vectors in Rn :
Obviously, D has eigenvalues a1 ; a2 ; :::; an ; and for each i = 1; 2; :::; n;
2 32 3 2 3
a1 ¢ ¢ ¢ 0 ¢ ¢ ¢ 0 0 0
.
6 . ... . . .
.. .
.. 7 6 7 6.7
.
6. . 7 6 .. 7 6 .. 7
6 76 7 6 7
D~ei = 6 0 ¢ ¢ ¢ ai ¢ ¢ ¢ 0 7 617 = 6ai 7 = ai~ei :
6. .. .. . . . 7 6 .7 6 . 7
4 .. . . . .. 5 4 .. 5 4 .. 5
0 ¢ ¢ ¢ 0 ¢ ¢ ¢ an 0 0
We thus conclude that for any diagonal matrix, eigenvalues are all diagonal entries, and ~ei
is an eigenvector associated with ai ; the ith diagonal entry, i.e.,
~ei is an eigenvector associated with eigenvalue ai ; for i = 1; 2; :::; n:
Due to the simplicity of diagonal matrices, one likes to know whether any matrix can be
similar to a diagonal matrix. Diagonalization is a process of …nding a diagonal matrix that
is similar to a given non-diagonal matrix.
De…nition 12.1. An n £ n matrix A is called diagonalizable if A is similar to a diagonal
matrix D:
Example 12.1. Consider
· ¸ · ¸ · ¸
7 2 5 0 1 1
A= ; D= ; P = :
¡4 1 0 3 ¡1 ¡2
(a) Verify A = P DP ¡1
(b) Find Dk and Ak
(c) Find eigenvalues and eigenvectors for A:
Solution: (a) It su¢ces to show that AP = P D and that P is invertible. Direct
calculations lead to
det P = ¡1 6= 0 =) P is invertible
Again, direct computation leads to
· ¸ · ¸
5 3 5 3
AP = ; PD = :
¡5 ¡6 ¡5 ¡6
1
Therefore AP = P D:
(b) ·
¸· ¸ · 2 ¸ · k ¸
25 0 5 0 5 0 k 5 0
D = = ; D = :
0 3 0 3 0 32 0 3k
¡ ¢
A2 = P DP ¡1 P DP ¡1 = P DP ¡1 P DP ¡1 = P D2 P ¡1
Ak = P Dk P ¡1
(c) Eigenvalues of A = Eigenvalues of D; which are ¸1 = 5; ¸2 = 3: For D; we know that
· ¸
1
~e1 = is an eigenvectors for ¸1 = 5 : D~e1 = 5~e1 (2)
0
· ¸
0
~e2 = is an eigenvectors for ¸2 = 3 : D~e2 = 3~e2 : (3)
1
Since AP = P D; from (2) and (3), respectively, we see
AP~e1 = P D~e1 = P (5~e1 ) = 5P~e1 ;
AP~e2 = P D~e2 = P (3~e2 ) = 3P~e2 :
Therefore, · ¸· ¸ · ¸
1 1 1 1
P~e1 = =
¡1 ¡2 0 ¡1
is an eigenvector associated with eigenvalue ¸ = 5 for matrix A; and
· ¸· ¸ · ¸
1 1 0 1
P ~e2 = =
¡1 ¡2 1 ¡2
is an eigenvector of A associated with eigenvalue ¸= 3:
From this example, we observation that if A is diagonalizable and A is similar to a
diagonal matrix D (as in (1)) through an invertible matrix P;
AP = P D:
Then
P~ei is an eigenvector associated with ai ; for i = 1; 2; :::; n:
This generalization can be easily veri…ed in the manner analogous to Example 12.1. More-
over, these n eigenvectors P~e1 ; P ~e2 ; :::; P~en are linear independent. To see independence, we
consider the linear system
Xn
¸i P~ei = ~0:
i=1
2
Suppose the linear system admits a solution (¸1 ; :::; ¸n ). Then
à n !
X
P ¸i~ei = ~0:
i=1
Since P is invertible, the above equation leads to
X
n
¸i~ei = ~0:
i=1
Since has ~e1 ; ~e2 ; :::; ~en are linear independent, it follows that the last linear system has only
the trivial solution ¸i = 0 for all i = 1; 2; :::; n: Therefore, P~e1 ; P~e2 ; :::; P~en are linear inde-
pendent. This observation leads to
Theorem 12.1. (Diagonalization).A n £ n matrix A is diagonalizable i¤ A has n lin-
early independent eigenvectors. Furthermore, suppose that A has n linearly independent
eigenvectors f~v1 ; ~v2 ; :::; ~vn g. Set
2 3
a1 0 ¢ ¢ ¢ 0
6 0 a2 ¢ ¢ ¢ 0 7
P = [~v1 ; ~v2 ; :::; ~vn ] ; D = 6 7
4¢ ¢ ¢ ¢ ¢ ¢ ¢ ¢ ¢ ¢ ¢ ¢5
0 0 ¢ ¢ ¢ an
where ai is the eigenvalue associated with ~vi ; i.e., A~vi = ai~vi :Then, P is invertible and
P ¡1 AP = D:
Proof. We have demonstrated that if A is diagonalizable, then A has n linearly independent
eigenvectors. We next show that the reverse is also true: if A has n linearly independent
eigenvectors, then A must be diagonalizable. To this end, we only need to verify that
AP = P D with P and D described above, since P is obviously invertible due to independence
of eigenvectors. The proof of AP = P D is straightforward by calculation as follows:
AP = A [~v1 ; ~v2 ; :::; ~vn ]
= [A~v1 ; A~v2 ; :::; A~vn ]
= [a1~v1 ; a2~v2 ; :::; an~vn ] ;
P D = P [a1~e1 ; a2~e2 ; :::; an~en ]
= [a1 P~e1 ; a2 P~e2 ; :::; an P~en ] :
Now, 2 3 2 3
1 0
607 617
P~e1 = [~v1 ; ~v2 ; :::; ~vn ] 6 7 6 7
4¢ ¢ ¢5 = ~v1 ; P~e2 = [~v1 ; ~v2 ; :::; ~vn ] 4¢ ¢ ¢5 = ~v2 ; :::
0 0
3
This shows AP = P D.
Example 12.2. Diagonalize
2 3
1 3 3
A = 4¡3 ¡5 ¡35 :
3 3 1
Solution: Diagonalization means …nding a diagonal matrix D and an invertible matrix
P such that AP = P D: We shall follow Theorem 12.1 step by step.
Step 1. Find all eigenvalues. From computations
2 3
1¡¸ 3 3
det (A ¡ ¸I) = det 4 ¡3 ¡5 ¡ ¸ ¡3 5
3 3 1¡¸
= ¡¸3 ¡ 3¸2 + 4
¡ ¢
= ¡¸3 + ¸2 ¡ 4¸2 + 4
¡ ¢ ¡ ¢
= ¡¸3 + ¸2 + ¡4¸2 + 4
¡ ¢
= ¡¸2 (¸¡ 1) ¡ 4 ¸2 ¡ 1
£ ¤
= ¡ (¸¡ 1) ¸2 + 4 (¸+ 1)
= ¡ (¸¡ 1) (¸+ 2)2 :
we see that A has eigenvalues ¸1 = 1; ¸2 = ¡2 (this is a double root).
Step 2. Find all eigenvalues – …nd a basis for each eigenspace Null (A ¡ ¸Ii ).
For ¸1 = 1;
2 3 2 3
1 3 3 1 0 0
A ¡ ¸1 I = 4¡3 ¡5 ¡35 ¡ 40 1 05
3 3 1 0 0 1
2 3 2 3
0 3 3 3 3 0
R !R
= 4¡3 ¡6 ¡35 1! 2 4¡3 ¡6 ¡35
3 3 0 0 3 3
2 3 2 3
3 3 0 1 0 ¡1
R2 +R1 !R2
! 40 ¡3 ¡35 ! 40 1 1 5 :
0 3 3 0 0 0
Therefore, linear system
(A ¡ I) ~x = 0
reduces to
x1 ¡ x3 = 0
x2 + x3 = 0;
4
where x3 is the free variable. Choose x3 = 1; we obtain an eigenvector
2 3
1
~x = 4¡15 :
1
For ¸2 = ¡2;
2 3 2 3
1 3 3 1 0 0
A ¡ ¸2 I = 4¡3 ¡5 ¡35 ¡ (¡2) 40 1 05
3 3 1 0 0 1
2 3 2 3
3 3 3 1 1 1
= 4¡3 ¡3 ¡35 ! 40 0 05 :
3 3 3 0 0 0
The corresponding linear system is
x1 + x2 + x3 = 0
It follows that x2 and x3 are free variables. As we did before, we need to select (x2 ; x3 )
to be (1; 0) and (0; 1) : Choose x2 = 1; x3 = 0 =) x1 = ¡x2 ¡ x3 = ¡1; choose x2 = 0;
x3 = 1 =) x1 = ¡x2 ¡ x3 = ¡1: We thus got two independent eigenvectors for ¸2 = ¡2 :
2 3 2 3 2 3 2 3
x1 ¡1 x1 ¡1
~v2 = x2 = 1 ; ~v3 = x2 = 0 5 :
4 5 4 5 4 5 4
x3 0 x3 1
Step 3. Assemble orderly D and P as follows: there are several choices to pair D and P:
2 3 2 3
1 0 0 1 ¡1 ¡1
Choice#1 : D = 40 ¡2 0 5 ; P = [~v1 ; ~v2 ; ~v3 ] = 4¡1 1 05
0 0 ¡2 1 0 1
2 3 2 3
1 0 0 1 ¡1 ¡1
Choice#2 : D = 40 ¡2 0 5 ; P = [~v1 ; ~v3 ; ~v2 ] = 4¡1 0 15
0 0 ¡2 1 1 0
2 3 2 3
¡2 0 0 ¡1 ¡1 1
Choice#3 : D = 4 0 ¡2 05 ; P = [~v2 ; ~v3 ; ~v1 ] = 4 1 0 ¡15 :
0 0 1 0 1 1
Remark. Not every matrix is diagonalizable. For instance,
· ¸
1 1
A= ; det (A ¡ ¸I) = (¸¡ 1)2 :
0 1
The only eigenvalue is ¸ = 1; it has the multiplicity m = 2: From
· ¸
0 1
A¡I = ;
0 0
5
we see that (A ¡ I) has only one pivot. Thus r (A ¡ I) = 1: By Dimension Theorem, we
have
dim Null (A ¡ I) = 2 ¡ r (A) = 2 ¡ 1 = 1:
In other words, the basis consists of Null (A ¡ I) consists of only one vector. For instance,
one may choose · ¸
1
0
as a basis. According to the discussion above, if A is diagonalizable, i.e., AP = P D for a
diagonal matrix · ¸
a 0
D=
0 b
then
P~e1 is an eigenvector of A associated with a;
P~e2 is an eigenvector of A associated with b:
Furthermore, since P is invertible,
P~e1 and P~e2 are linearly independent (why?).
Since we have already demonstrated that there is no more than two linearly independent
eigenvectors, it is impossible to diagonalize A.
In general, if ¸0 is an eigenvalue of multiplicity m (i.e., the characteristic polynomial
det (A ¡ ¸I) = (¸¡ ¸0 )m Q (¸) ; Q (¸0 ) 6= 0 );
then
dim (Null (A ¡ ¸I)) · m:
Theorem 12.2. Let A be an n £ n matrix with distinct real eigenvalues ¸1 ; ¸2 ; :::; ¸p
with multiplicity m1 ; m2 ; :::; mp ; respectively. Then,
1. ni = dim (N ull (A ¡ ¸i I)) · mi and m1 + m2 + ::: + mp · n:
2. A is diagonalizable i¤ ni = mi for all i = 1; 2; :::; p; and
m1 + m2 + ::: + mp = n:
In this case, we have P ¡1 AP = D; where P and D can be obtained as follows. Let Bi
be a basis of Null (A ¡ ¸i I) for each i: Then
2 3
¸1 Im1 ::: 0
P = [B1 ; :::; Bp] ; D = 4 ::: ::: ::: 5 ; Imi = (mi £ mi ) identity
0 ::: ¸pImp
i.e., the …rst m1 columns of P form B1 ; the eigenvectors for ¸1 ; the next m2 columns
of P are B2 ; then B3 ; etc. The last mp columns of P are from Bp; the …rst m1 diagonal
entries of D are ¸1 ; the next m2 diagonal entries of D are ¸2 , and so on.
6
3. In particular, if A has n distinct real eigenvalues, then A is diagonalizable.
4. Any symmetric matrix is diagonalizable.
Note that, as we saw before, there are multiple choices for assembling P: For instance, if
A is 5 £ 5; and
n A hasotwo eigenvalues ¸1 = 2; ¸2 = 3 with a basis f~a1 ; ~a2 g for Null (A ¡ 2I)
and a basis ~b1 ; ~b2 ; ~b3 for Null (A ¡ 3I) ; respectively, then, we have several choices to select
pairs of (P; D) :
h i · ¸
~ ~ ~ 2I2 0
choice#1 : P = ~a1 ; ~a2 ; b1 ; b2 ; b3 ; D =
0 3I3
2· ¸ 3 2 3
2 0 2 0 0 0 0
6 0 2 0 7 6 7
6 2 37 60 2 0 0 07
=66 3 0 0 7 6
7 = 60 0 3 0 07
7
4 0 40 3 055 40 0 0 3 05
0 0 3 0 0 0 0 3
2 3
2 0 0 0 0
h i 60 3 0 0 07
6 7
choice#2 : P = ~a1 ; ~b1 ; ~b3 ;~a2 ; ~b2 ; D = 66 0 0 3 0 0 7:
7
40 0 0 2 05
0 0 0 0 3
Example 12.3. Diagonalize A
2 3
5 0 0 0
60 5 0 07
A=6
41
7:
4 ¡3 0 5
¡1 ¡2 0 ¡3
Solution: We need to …nd all eigenvalues and eigenvectors. Since A is an upper triangle
matrix, we see that eigenvalues are ¸1 = 5; m1 = 2 , ¸2 = ¡3; m2 = 2:
For ¸1 = 5;
2 3 2 3
0 0 0 0 0 0 0 0
60 0 0 07 6 0 0
3 !R4 60 07
A ¡ 5I = 6 7 R4 +R
! 7
41 4 ¡8 0 5 41 4 ¡8 0 5
¡1 ¡2 0 ¡8 0 2 ¡8 ¡8
2 3 2 3
0 0 0 0 0 0 0 0
60 0 0 07 60 0 0 07
!6 7 6
41 4 ¡8 0 5 ! 41 0 8 16 5
7
0 1 ¡4 ¡4 0 1 ¡4 ¡4
7
Therefore, x3 and x4 are free variable, and
x1 = ¡8x3 ¡ 16x4
x2 = 4x3 + 4x4 :
Choose (x3 ; x4 ) = (1; 0) =) x1 = ¡8; x2 = 4; Choose (x3 ; x4 ) = (0; 1) =) x1 = ¡16; x2 = 4:
We obtain two independent eigenvectors
2 3 2 3
¡8 ¡16
647 6 4 7
6 7;6 7
4 1 5 4 0 5 (for ¸1 = 5):
0 1
For ¸2 = ¡3;
2 3 2 3
8 0 0 0 8 0 0 0
60 8 070 6 8 0 07
3 !R4 60
A ¡ (¡3) I = 6 7 R4 +R
! 7
41 4 050 41 4 0 05
¡1 ¡2 0 0 0 2 0 0
2 3 2 3
R3 ¡ 2R4 ! R3 8 0 0 0 0 0 0 0
R2 ¡ 4R4 ! R2 60 0 0 07 R1 ¡8R3 !R1 60 0 0 07
! 6 7 ! 6 7:
41 0 0 05 41 0 0 05
0 2 0 0 0 2 0 0
Hence
x1 = 0; x2 = 0:
Choose (x3 ; x4 ) = (1; 0) and (x3 ; x4 ) = (0; 1) ; respectively, we have eigenvectors
2 3 2 3
0 0
607 607
6 7 and 6 7 (for ¸2 = ¡3).
415 405
0 1
Assemble pairs (P; D) : There are several choice. For instance,
2 3 2 3
¡8 ¡16 0 0 5 0 0 0
64 4 0 07 6 07
P =6 7 ; D = 60 5 0 7
41 0 1 05 40 0 ¡3 0 5
0 1 0 1 0 0 0 ¡3
or 2 3 2 3
¡8 0 ¡16 0 5 0 0 0
64 0 4 07 6 7
P =6 7 ; D = 60 ¡3 0 0 7 :
41 1 0 05 40 0 5 0 5
0 0 1 1 0 0 0 ¡3
8
² Homework 12.
² 1. Suppose that A is similar to
2 3
¡1 0 0 0
60 2 0 07
D=6
40
7:
0 0 05
0 0 0 ¡1
(a) Find the characteristic polynomial and eigenvalues of A:
(b) Let P ¡1 AP = D; where
2 3
2 1 0 5
6 0 3 0 27
P =64¡1 0 1 25 :
7
0 1 0 4
Find a basis for each eigenspace of A:
2. Diagonalize the following matrices.
2 3
2 2 ¡1
(a) A = 4 1 3 ¡15 ; ¸= 1; 5:
¡1 ¡2 2
2 3
2 0 0 0
64 ¡3 0 07
(b) B = 641 2 1 05
7
0 1 0 1
3. Show that if A is diagonalizable, so is A2 :
4. Suppose that n £ n matrices A and B have exact the same linearly independent
eigenvectors that span Rn . Show that (a) both A and B can be simultaneously
diagonalized (i.e., there are the same P such that P ¡1 AP and P ¡1 BP are diagonal
matrices), and (b) AB is also diagonalizable.
5. for each statement, determine and explain true or false.
(a) A is diagonalizable if A = P ¡1 DP for some diagonal matrix D:
(b) If A is diagonalizable, then A is invertible.
(c) If AP = P D with D diagonal matrix, then nonzero columns of P are eigen-
vectors of A:
(d) Let A be a symmetric matrix and ¸ be an eigenvalue of A with multiplicity
5: Then the eigenspace Null (A ¡ ¸I) has dimension 5:
