Mathematics for Electrical Engineers (EE 143):
Eigenvalues and Eigenvectors
Krishnan C.M.C
Assistant Professor, E&E,
NITK Surathkal
Contents
Eigenvectors and Eigenvalues
The Characteristic Equation
Diagonalization
Complex Eigenvalues
EE143 Dept. of E & E, NITK Surathkal
2
Contents
Eigenvectors and Eigenvalues
The Characteristic Equation
Diagonalization
Complex Eigenvalues
EE143 Dept. of E & E, NITK Surathkal
3
Eigenvector
DEF An eigenvector of an 𝑛 × 𝑛 matrix 𝐴 is any non zero vector 𝑥 such that
𝐴𝑥 = 𝜆𝑥 for some scalar 𝜆.The scalar 𝜆 is called an eigenvalue of 𝐴 if
there is a non trivial solution 𝑥 of 𝐴𝑥 = 𝜆𝑥; such an 𝑥 is called as an
eigenvector corresponding to 𝜆.
Ex-1 1 6 6 3
Let 𝐴 = ,𝑥 = ,𝑦 = Check if the two given vectors are
5 2 −5 −2
eigenvectors of the matrix
−24 6
𝐴𝑥 = = −4 = −4𝑥 Eigenvector or 𝐴
20 −5
−9 3
𝐴𝑦 = ≠𝜆 Not an eigenvector or 𝐴
11 −2
Ex-2 Is 𝜆 = 7 an eigenvalue of 𝐴 in Ex-1?
Check if 𝐴𝑥 = 𝜆𝑥 has a non trivial solution
1 6 1 6 1 0 −6 6
𝑥 = 7𝑥 ⇒ −7 𝑥=0 ⇒ 𝑥=0
5 2 5 2 0 1 5 −5
EE143 Dept. of E & E, NITK Surathkal
4
Eigenvector
Ex-2 Is 𝜆 = 7 an eigenvalue of 𝐴 in Ex-1?
Check if 𝐴𝑥 = 𝜆𝑥 has a non trivial solution
1 6 1 6 1 0 −6 6
𝑥 = 7𝑥 ⇒ −7 𝑥=0 ⇒ 𝑥=0
5 2 5 2 0 1 5 −5
−6 6 0 1 −1 0 1
⇒ ~ ⇒ 𝑥 = 𝑥2
5 −5 0 0 0 0 1
Eigen space = eigenvectors + zero vector
𝑑2
1
1
𝑑1
Taken from Lay et.al. and reproduced here
EE143 Dept. of E & E, NITK Surathkal
5
Eigen Space
DEF • The eigenspace of an 𝑛 × 𝑛 matrix 𝐴 for a given eigenvalue 𝜆 is the null
space of the matrix (𝐴 − 𝜆𝐼)
• The eigen space consists of the eigenvectors and the zero vector
• The eigen space is a subspace of ℝ𝑛
• The product 𝐴𝑥 = 𝜆𝑥 will belong to the eigen space of 𝐴 for the given 𝜆
Ex-3 4 −1 6
Let 𝐴 = 2 1 6 and one of the eigenvalues is 2. Find a basis for the
2 −1 8
corresponding eigenspace
4−2 −1 6 0
Nul 𝐴 − 2𝐼3 ⇒ solve 2 1−2 6 ⋮ 0
2 −1 8 − 2 0
2 −1 6 0 2 −1 6 0 0.5 −3
2 −1 6 ⋮ 0 ~ 0 0 0 ⋮ 0 ⇒ 𝑥 = 1 𝑥2 + 0 𝑥3
2 −1 6 0 0 0 0 0 0 1
EE143 Dept. of E & E, NITK Surathkal Basis set 𝑢, 𝑣
6
Eigen Space
4 −1 6
Ex-3
Let 𝐴 = 2 1 6 and one of the eigenvalues is 2. Find a basis for the
2 −1 8
corresponding eigenspace
0.5 −3
Basis of the eigen space 1 , 0
0 1
Taken from Lay et.al. and reproduced here
EE143 Dept. of E & E, NITK Surathkal
7
Eigen Space
Theorem The eigenvalues of a triangular matrix are its diagonal entries
Theorem If 𝑣1 , 𝑣2 , … , 𝑣𝑟 are eigenvectors corresponding to distinct
eigenvalues 𝜆1 , 𝜆2 , … , 𝜆𝑟 of an 𝑛 × 𝑛 of a matrix 𝐴, then the set
𝑣1 , 𝑣2 , … , 𝑣𝑟 is linearly independent
Proof In the indexed set, let 𝑣𝑝+1 be a linear combination of the preceding linearly
independent vectors (with 𝑝 < 𝑟 ).
𝑣𝑝+1 = 𝑐1 𝑣1 + 𝑐2 𝑣2 + ⋯ + 𝑐𝑝 𝑣𝑝 → (𝟏)
Multiply (𝟏) by the matrix 𝐴 ⇒ 𝜆𝑝+1 𝑣𝑝+1 = 𝑐1 𝜆1 𝑣1 + 𝑐2 𝜆2 𝑣2 + ⋯ + 𝑐𝑝 𝜆𝑝 𝑣𝑝 → (𝟐)
𝟐 − 𝝀𝒑+𝟏 (𝟏) ⇒
0 = 𝑐1 𝜆1 − 𝜆𝑝+1 𝑣1 + 𝑐2 𝜆2 − 𝜆𝑝+1 𝑣2 + ⋯ + 𝑐𝑝 𝜆𝑝 − 𝜆𝑝+1 𝑣𝑝 → (𝟑)
• Since 𝑣1 , 𝑣2 , … , 𝑣𝑝 is linearly independent, the weights 𝑐𝑖 𝜆𝑖 − 𝜆𝑝+1 , 1 ≤
𝑖 ≤ 𝑝 are all zero. But 𝜆𝑖 ≠ 𝜆𝑝+1 .
• ⇒ 𝑐𝑖 are all zeros.
• Then by 𝟏 , 𝑣𝑝+1 = 0 which is not possible since it is an eigenvector
• Thus the set 𝑣1 , 𝑣2 , … , 𝑣𝑝+1 𝑝 < 𝑟 must be linearly independent
EE143 Dept. of E & E, NITK Surathkal
8
Contents
Eigenvectors and Eigenvalues
The Characteristic Equation
Diagonalization
Complex Eigenvalues
EE143 Dept. of E & E, NITK Surathkal
9
Characteristic Equation
2 3
Ex-4 Find the eigenvalues of the matrix 𝐴 =
3 −6
Find 𝜆 such that 𝐴 − 𝜆𝐼 𝑥 = 0 has a non-trivial solution
⇒ 𝐴 − 𝜆𝐼 Should not be invertible
⇒ det 𝐴 − 𝜆𝐼 = 0 ⇒
2−𝜆 3
=0 ⇒ 𝜆2 + 4𝜆 − 21 = 0
3 −6 − 𝜆
⇒ 𝜆 = 3, −7
DEF 𝜆2 + 4𝜆 − 21 = 0 is the characteristic equation of 𝐴
𝜆 = 3, −7 are the characteristic roots or eigenvalues of 𝐴
𝑝 𝜆 = 𝜆2 + 4𝜆 − 21 is the characteristic polynomial of 𝐴
Theorem The scalar 𝜆 is an eigenvalue of 𝐴 IFF it satisfies the characteristic
equation det 𝐴 − 𝜆𝐼 = 0
EE143 Dept. of E & E, NITK Surathkal
10
Characteristic Equation
0.95 0.03
Ex-5 Application to a dynamical system: Let 𝐴 = . Analyze the long-
0.05 0.97
term behavior of a system described by an iterative equation, 𝑥𝑘+1 =
0.6
𝐴𝑥𝑘 (𝑘 = 0,1,2, … ) with 𝑥0 =
0.4
Method 1: Direct iterative method: find 𝑥1 = 𝐴𝑥0 , then𝑥2 = 𝐴𝑥1 etc.
Method 2: Using eigenvalues and eigenvectors to get a closed form solution
Step 1: Find the eigenvalues of 𝐴
0.95 − 𝜆 0.03
𝑑𝑒𝑡 = 0 ⇒ 𝜆2 − 1.92𝜆 + 0.92 = 0 ⇒ 𝜆 = 1.0, 0.92
0.05 0.97 − 𝜆
Step 2: Find the eigenvectors corresponding to 𝜆 = 1.0, & 𝜆 = 0.92
𝜆=1
−0.05 0.03 0 −0.05 0.03 0 3/5 3
⋮ ~ ⋮ ⇒ 𝑣1 = 𝑥2 ⇒ 𝑣1 =
0.05 −0.03 0 0 0 0 1 5
EE143 Dept. of E & E, NITK Surathkal
11
Characteristic Equation
0.95 0.03
Ex-5 Application to a dynamical system: Let 𝐴 = . Analyze the long-
0.05 0.97
Contnd… term behavior of a system is described by an iterative equation, 𝑥
𝑘+1 =
0.6
𝐴𝑥𝑘 (𝑘 = 0,1,2, … ) with 𝑥0 =
0.4
Step 2: Find the eigenvectors corresponding to 𝜆 = 1.0, & 𝜆 = 0.92
𝜆 = 0.92
0.03 0.03 0 0.03 0.03 0 −1 1
⋮ ~ ⋮ ⇒ 𝑣2 = 𝑥2 ⇒ 𝑣2 =
0.05 0.05 0 0 0 0 1 −1
0.6 3 1
Step 3: Represent 𝑥0 = as a linear combination of 𝑣1 = &𝑣2 =
0.4 5 −1
2
Note: it will be always possible since 𝑣1 , 𝑣2 span ℝ Why?
𝑐1 𝑐1 −1
𝑐1 𝑣1 + 𝑐2 𝑣2 = 𝑥0 ⇒ [𝑣1 𝑣2 ] 𝑐 = 𝑥0 ⇒ 𝑐 = 𝑣1 𝑣2 −1 𝑥0 = 3 1 0.6
2 2 5 −1 0.4
𝑐1 0.125
⇒ 𝑐 =
2 0.225
EE143 Dept. of E & E, NITK Surathkal
12
Characteristic Equation
0.95 0.03
Ex-5 Application to a dynamical system: Let 𝐴 = . Analyze the long-
0.05 0.97
Contnd… term behavior of a system is described by an iterative equation, 𝑥
𝑘+1 =
0.6
𝐴𝑥𝑘 (𝑘 = 0,1,2, … ) with 𝑥0 =
0.4
0.6 3 1
Step 3: Represent 𝑥0 = as a linear combination of 𝑣1 = &𝑣2 =
0.4 5 −1
2
Note: it will be always possible since 𝑣1 , 𝑣2 span ℝ
𝑐1 𝑐1 −1
𝑐1 𝑣1 + 𝑐2 𝑣2 = 𝑥0 ⇒ [𝑣1 𝑣2 ] 𝑐 = 𝑥0 ⇒ 𝑐 = 𝑣1 𝑣2 −1 𝑥0 = 3 1 0.6
2 2 5 −1 0.4
𝑐1 0.125
⇒ 𝑐 =
2 0.225
Step 4: Finding a general formula for 𝑥𝑘
We have 𝑥1 = 𝐴𝑥0 = 𝐴 𝑐1 𝑣1 + 𝑐2 𝑣2 = 𝑐1 𝐴𝑣1 + 𝑐2 𝐴𝑣2 = 𝑐1 𝜆1 𝑣1 + 𝑐2 𝜆2 𝑣2
Then 𝑥2 = 𝐴𝑥1 = 𝐴 𝑐1 𝜆1 𝑣1 + 𝑐2 𝜆2 𝑣2 = 𝑐1 𝜆12 𝑣1 + 𝑐2 𝜆22 𝑣2
Similarly 𝑥𝑘 = 𝑐1 𝜆1𝑘 𝑣1 + 𝑐2 𝜆𝑘2 𝑣2 𝟑 𝒌 𝟏
= 𝟎. 𝟏𝟐𝟓 + 𝟎. 𝟐𝟐𝟓 𝟎. 𝟗𝟐
𝟓 −𝟏
EE143 Dept. of E & E, NITK Surathkal
13
Characteristic Equation
Theorem
Invertible matrix theorem continued: An 𝑛 × 𝑛 matrix 𝐴 is invertible IFF
The number 0 is not an eigenvalue of 𝐴.
Proof (i) Let us consider that 0 is an eigenvalue of 𝐴
Corresponding eigenvectors are the non trivial solutions of 𝐴 − 0𝐼 𝑥 = 0
i.e., 𝐴 𝑥 = 0
If 𝑥 is a non trivial solution, not all 𝑛 columns of 𝐴 are pivot columns
Thus 𝐴 is not invertible IF 0 is an eigenvalue of 𝐴
(ii) Now let us check when 𝐴 is invertible
Left multiply the equation 𝐴𝑥 = 𝜆𝑥 by 𝐴−1
⇒ 𝑥 = 𝜆𝐴−1 𝑥
Now since 𝑥 ≠ 0 as it is an eigenvector, 𝜆 ≠ 0
Thus 𝐴 is invertible IIF 0 is not an eigenvalue of 𝐴
EE143 Dept. of E & E, NITK Surathkal
14
Similarity transformation and eigenvalues
DEF If 𝐴 & 𝐵 are 𝑛 × 𝑛 matrices, then 𝐴 is similar to 𝐵 if there is an invertible
matrix 𝑃 such that P −1 𝐴𝑃 = 𝐵. This is known as a similarity transformation.
Theorem If two 𝑛 × 𝑛 matrices 𝐴 & 𝐵 are similar, then they have the same
characteristic polynomial and hence the same eigenvalues.
Proof det 𝐵 − 𝜆𝐼 = det P −1 𝐴𝑃 − 𝜆𝐼
= det P × det P −1 𝐴𝑃 − 𝜆𝐼 × det 𝑃−1
= det 𝑃 P −1 𝐴𝑃 − 𝜆𝐼 𝑃−1
= det 𝑃P −1 𝐴𝑃𝑃−1 − 𝑃𝜆𝐼𝑃−1
= det 𝐴 − 𝜆𝐼
Thus they have the same characteristic polynomials
Warning:
1. If two matrices have the same eigenvalues, that do not mean that they
are similar
2. If 𝐴~𝐵 they may not have the same eigenvalues.
EE143 Dept. of E & E, NITK Surathkal
15
Examples
Ex-6 Find the eigenvalues of the following matrices using determinants
4 0 −1
0 4 −1 4,3,3
1 0 2
3 1 1
0 5 0 5, 6.41, 3.58
−2 0 7
Ex-7 Find the eigenvalues of the following matrices by inspection
5,5,2,3 3,1,3,1,0
EE143 Dept. of E & E, NITK Surathkal
16
Examples
Ex-8 Explain why det 𝐴 is the product of the eigenvalues of 𝐴
Hint: eigenvalues are the solution of det 𝐴 − 𝜆𝐼 = 0
Ex-9 Let 𝜆 be an eigenvalue of 𝐴 , an invertible matrix. Then show that 𝜆−1 is an
eigenvalue of 𝐴−1.
We have 𝐴𝑥 = 𝜆 𝑥 Left multiply both sides by 𝐴−1 , We have 𝑥 = 𝜆 𝐴−1 𝑥
i.e.,𝜆−1 𝑥 = 𝐴−1 𝑥 Hence 𝜆−1 is an eigenvalue of 𝐴−1
Ex-10 If 𝐴2 is the zero matrix, then show that the eigenvalues of 𝐴 are all zeros.
We have 𝐴𝑥 = 𝜆 𝑥 Left multiply both sides by𝐴, We have A2 𝑥 = 𝜆 𝐴𝑥
But 𝐴2 is zero matrix and the eigenvector 𝑥 ≠ 0 . Thus 𝜆 = 0
Ex-11 Consider an 𝑛 × 𝑛 matrix 𝐴 such that the row sums equals 𝑠. Prove
that 𝑠 is an eigenvalue.
We can see that 𝑛 × 1 column vector with all values 1 (or equal) will
be an eigenvector
EE143 Dept. of E & E, NITK Surathkal
17
Examples
Ex-8 Find the value of ℎ such that the eigen space corresponding to 𝜆 = 4 is 2
dimensional for the matrix
That means, 𝐴 − 4𝐼 𝑥 = 0 should have
Dimension of eigen space =2
two free variables
0 2 3 3 0 2 3 3
𝐴 − 4𝐼 = 0 −2 ℎ 3 ~ 0 0 ℎ + 3 6
0 0 0 14 0 0 0 14
0 0 0 −2 0 0 0 −2
In this example, if ℎ + 3 = 0 then the system will have two free
variables (𝑥1 and 𝑥3 ) 0 1 3/2 0 0 1
1 −1.5 0
𝐴 − 4𝐼~ 0 0 0 ⇒ 𝑥 = 𝑥3 +𝑥1
0 1 0
0 0 0
0 0 0 0 0 0
EE143 Dept. of E & E, NITK Surathkal Dim (eigen space) =2
18
Contents
Eigenvectors and Eigenvalues
The Characteristic Equation
Diagonalization
Complex Eigenvalues
EE143 Dept. of E & E, NITK Surathkal
19
Diagonalization
Ex-12 7 2
Let 𝐴 = , find a formula for 𝐴𝑘 , given 𝐴 = 𝑃𝐷𝑃−1 , where 𝑃 =
−4 1
1 1 5 0
and 𝐷 =
−1 −2 0 3
𝑘 2 × 5𝑘 − 3𝑘 5𝑘 − 3𝑘
𝐴 =
2 × 3𝑘 − 5𝑘 2 × 3𝑘 − 5𝑘
Theorem 1. An 𝑛 × 𝑛 matrix 𝐴 is diagonalizable if and only if 𝐴 has 𝑛 linearly
independent eigenvectors
2. In fact, 𝐴 = 𝑃𝐷𝑃−1 , with 𝐷 a diagonal matrix, if and only if the
columns of 𝑃 are 𝑛 linearly independent eigenvectors of 𝐴.
3. In this case, the diagonal entries of 𝐷 are the 𝑛 eigenvalues of 𝐴
1 3 3
Ex-13 Diagonalize the following matrix 𝐴 = −3 −5 −3 if possible.
3 3 1
EE143 Dept. of E & E, NITK Surathkal
20
Diagonalization 1 3 3
𝐴 = −3 −5 −3
3 3 1
Ex-13 Step-1: Find the eigenvalues
1−𝜆 3 3
det −3 −5 − 𝜆 −3 = 0
3 3 1−𝜆
1−𝜆 −5 − 𝜆 1 − 𝜆 + 9 − 3 −3 + 3𝜆 + 9 + 3 −9 + 15 + 3𝜆 = 0
1 − 𝜆 𝜆2 + 4𝜆 + 4 = 0 ⇒ 𝝀 = 1, −2, −2
Step-2: Find three linearly independent eigenvectors
we have the coefficient matrix for the homogeneous equation as
For 𝜆 = 1 For 𝜆 = −2
0 3 3 1 0 −1 3 3 3 1 1 1
−3 −6 −3 ~ 0 1 1 −3 −3 −3 ~ 0 0 0
3 3 0 0 0 0 3 3 3 0 0 0
1 Eigenspace
Eigenspace
−1 −1
for𝜆 = −2
⇒ 𝑥 = 𝑥3 −1 for𝜆 = 1 ⇒ 𝑥 = 𝑥2
Dim? 1 + 𝑥3 0
1 Dim? 1-D 2-D 0 1
𝑣1 𝑣2 𝑣3
1+2=3
EE143 Dept. of E & E, NITK Surathkal
21
Diagonalization
Step-3: Form the matrix P and D (with the same order of eigenvalues and
vectors)
1 −1 −1 1 𝜆1 0 0
𝑃 = −1 1 0 𝐷 = 0 −2 𝜆2 0
1 0 1 0 0 −2
𝜆3
𝑣1 𝑣2 𝑣3
Please verify that 𝑃𝐷 = 𝐴𝑃 as a homework
Ex-14 2 4 3
(H.W.) Diagonalize the following matrix 𝐴 = −4 −6 −3 if possible.
3 3 1
Note: You will see that there are repeated roots and each eigenspace is 1-D. Hence you wont
be able to generate a matrix P
Theorem An 𝑛 × 𝑛 matrix 𝐴 with 𝑛 distinct eigenvalues is diagonalizable.
Proof: ref. to previous theorem which says that n distinct eigenvalues
generate n linearly independent eigenvectors.
Is the converse true? i.e., if 𝐴 is diagonalizable, then 𝐴 has 𝑛 distinct
eigenvalues? NO: Ref Ex:13
EE143 Dept. of E & E, NITK Surathkal
22
Diagonalization
Theorem
Let 𝐴 be an 𝑛 × 𝑛 matrix with 𝑝 ≤ 𝑛 distinct eigenvalues 𝜆1 , 𝜆2 , … , 𝜆𝑝 .
1. For 1 ≤ 𝑘 ≤ 𝑝, the dimension of eigen space for 𝜆𝑘 is less than or
equal to the multiplicity of the eigenvalue 𝜆𝑘 .
2. The matrix A is diagonalizable iff the sum of the dimensions of the
eigenspaces equal 𝑛, and this happens iff:
i. The characteristic polynomial factors completely into linear
factor, or,
ii. The dimension of the eigen space for each 𝜆𝑘 equals the
multiplicity of 𝜆𝑘 .
3. If A is diagonalizable and 𝐵𝑘 is a basis for the eigenspace
corresponding to 𝜆𝑘 for each 𝑘 , then the total collection of vectors in
the sets 𝐵1 , 𝐵2 , … , 𝐵𝑝 forms an eigenvector basis for ℝ𝑛 .
Ex-14 𝐴 is a 4 × 4 matrix with three distinct eigenvalues. One eigenspace is 1-dim,
and one of the other eigenspaces is 2-dim. Is it possible that 𝐴 is not
diagonalizable? Justify.
If A is not diagonalizable means 3rd eigenspace is 0-d. This is impossible
as eigenvector can not be a zero vector (not even for 𝜆 = 0).
EE143 Dept. of E & E, NITK Surathkal
23
Diagonalization
Ex-16 Let the matrix be diagonalized and is given below. Find its eigenvalues and
basis set for each eigen space
1 −1 0
𝜆1 = 3 , 𝐵1 = −1 𝜆2 = 2 , 𝐵2 = 1 𝜆3 = 3 , 𝐵3 = −1
0 −1 1
1 0 −1
𝜆1,3 = 3 , 𝐵1,3 = −1 , −1 𝜆2 = 2 , 𝐵2 = 1
0 1 −1
Ex-17 Show that if a matrix is invertible and diagonalizable, so will be its inverse
𝐴 = 𝑃𝐷𝑃−1 Then 𝐴−1 = 𝑃𝐷𝑃−1 = 𝑃𝐷−1 𝑃−1
−1
We have 𝐷−1 diagonal with entries as the reciprocal of the eigenvalues of 𝐴
Thus 𝐴−1 is diagonalizable Eigenvalues of 𝐴−1
EE143 Dept. of E & E, NITK Surathkal
24
Diagonalization
Ex-18 Find a 2 × 2 matrix which is invertible but not diagonalizable
𝛼 𝛽
𝐴= with 𝛼 and 𝛽 two non-zero scalars
0 𝛼
Here eigenvalue is 𝛼
1 Which is 1 dimensional Hence not diagonalizable
Eigen space is 𝑥1
0
Ex-19 Find a 2 × 2 matrix which is not invertible but diagonalizable
𝛼 𝛽
𝐴= with 𝛼 and 𝛽 two non-zero scalars
0 0
Here eigenvalues are 𝛼 & 0 Since 0 is an eigenvalue, A is not invertible.
1 1
The eigenspace for 𝜆 = 𝛼 is 𝑥1 The eigenspace for 𝜆 = 0 is 𝑥2 −𝛽/𝛼
0
Since 1 + 1 = 2 , 𝑃 can be generated and hence 𝐴 is diagonalizable
EE143 Dept. of E & E, NITK Surathkal
25
Contents
Eigenvectors and Eigenvalues
The Characteristic Equation
Diagonalization
Complex Eigenvalues
EE143 Dept. of E & E, NITK Surathkal
26
Complex Eigenvalues
Ex-20 0 −1
Let 𝐴 = .
1 0
Apply 4 × to get the
Do you recognize this matrix? Rotation by 90°
original vector back
Does it have any eigenvector in ℝ2 ? None
Eigenvalues: 𝜆2 + 1 = 0 ⇒ 𝜆 = ±𝑖
Eigenvectors:
𝜆=𝑖
Check: 𝐴𝑣1 = 0 −1 𝑖
−𝑖 −1 0 −𝑖 −1 0 𝑖 1 0 1
⋮ ~ ⋮ ⇒ 𝑣1 = 𝑐 −1 𝑖
1 −𝑖 0 0 0 0 1 =
𝑖 =𝑖 1
⇒ 𝑐 is in ℂ
𝜆 = −𝑖
𝑖 −1 0 𝑖 −1 0 −𝑖
⋮ ~ ⋮ ⇒ 𝑣2 = 𝑐
1 𝑖 0 0 0 0 1 Also 𝑣1 &𝑣2 are in ℂ2
Also notice, 𝜆1 = 𝜆∗2 And𝑣1 = 𝑣2 ∗
i.e., eigenvalues and the corresponding eigenvectors come in complex conjugate pairs
∗
IMP: . represents a conjugation operation (different from Lay et.al. notations)
EE143 Dept. of E & E, NITK Surathkal
27
Complex Eigenvalues
Also notice, 𝜆1 = 𝜆∗2 And𝑣1 = 𝑣2 ∗
i.e., eigenvalues and the corresponding eigenvectors come in complex conjugate pairs
In general, for an 𝑛 × 𝑛 real valued matrix 𝐴 with complex eigenvalue
𝜆1 and eigenvector 𝑣1 we have
𝐴𝑣1 = 𝜆1 𝑣1
Taking conjugate on both sided
𝐴 𝑣1 ∗ = 𝜆1 ∗ 𝑣1 ∗
∗ ∗
Thus 𝜆1 is an eigenvalue and the corresponding eigenvector is 𝑣1
Ex-21 0.5 −0.6
Let 𝐴 = . Find the eigenvalues and basis for each eigenspace.
0.75 1.1
Characteristic eqn: 𝜆2 − 1.6𝜆 + 1 = 0 Ch. roots:𝜆 = 0.8 ± 0.6𝑖
Eigenvectors: −0.3 + 0.6𝑖 −0.6 0
⋮ −2 − 4𝑖
𝜆 = 0.8 − 0.6𝑖 ⇒ ⇒𝑥=𝑐
0.75 0.3 + 0.6𝑖 0 5
EE143 Dept. of E & E, NITK Surathkal
28
Complex Eigenvalues
Ex-21 0.5 −0.6
Let 𝐴 = . Find the eigenvalues and basis for each eigenspace.
0.75 1.1
Characteristic eqn: 𝜆2 − 1.6𝜆 + 1 = 0 Ch. roots:𝜆 = 0.8 ± 0.6𝑖
Eigenvectors: −0.3 + 0.6𝑖 −0.6 0
⋮ −2 − 4𝑖
𝜆1 = 0.8 − 0.6𝑖 ⇒ ⇒ 𝑣1 = 𝑐
0.75 0.3 + 0.6𝑖 0 5
Simillarly for: −2 + 4𝑖
𝜆2 = 0.8 + 0.6𝑖 ⇒ 𝑣2 = 𝑐
5 Taken from Lay
Lets now see how 𝐴 acts on some 𝑥 in ℝ2 et.al. and
reproduced here
2
Let 𝑥0 =
0
1.0
Then 𝑥1 = 𝐴𝑥0 =
1.5
−0.4
𝑥2 = 𝐴𝑥1 =
2.4
⋮
• Matrix with complex eigenvalues will results in rotations
• How to quantify them? Lets take a simple example in ℝ2
EE143 Dept. of E & E, NITK Surathkal
29
Complex Eigenvalues
Ex-22 𝑎 −𝑏
Let 𝐶 = . Find the eigenvalues.
𝑏 𝑎
Ch. roots:𝜆 = 𝑎 ± 𝑏𝑖 if:𝑟 = 𝜆 = 𝑎2 + 𝑏2
𝑟
𝑏
𝑟 0 cos 𝜑 − sin 𝜑
Then :𝐶 =
0 𝑟 sin 𝜑 cos 𝜑 𝜑 𝑎
scaling rotation
Ex-23 0.5 −0.6 −2 − 4𝑖
Let us now take Ex-21 𝐴 = , 𝜆1 = 0.8 − 0.6𝑖, 𝑣1 = .
0.75 1.1 5
−2 −4 1 0 4
Lets take 𝑃 = 𝑅𝑒 𝑣1 𝐼𝑚 𝑣1 = 𝑃 −1 =
5 0 20 −5 −2
0.8 −0.6 𝑎 −𝑏
𝑃−1 𝐴𝑃 = Compare with 𝐶 =
0.6 0.8 𝑏 𝑎
𝑎 = 0.8, 𝑏 = 0.6 𝜑 = 36.86° 𝑟 = 𝜆 = 1
EE143 Dept. of E & E, NITK Surathkal
30
Complex Eigenvalues
Ex-23 0.5 −0.6 −2 − 4𝑖
Let us now take Ex-21 𝐴 = , 𝜆1 = 0.8 − 0.6𝑖, 𝑣1 = .
0.75 1.1 5
−2 −4 1 0 4
Lets take 𝑃 = 𝑅𝑒 𝑣1 𝐼𝑚 𝑣1 = 𝑃 −1
=
5 0 20 −5 −2
0.8 −0.6 𝑎 −𝑏
𝑃−1 𝐴𝑃 = Compare with 𝐶 =
0.6 0.8 𝑏 𝑎
𝑎 = 0.8, 𝑏 = 0.6 𝜑 = 36.86° 𝑟= 𝜆 =1
−2 −4 0.8 −0.6 1 0 4
𝐴 = 𝑃𝐶𝑃−1 =
5 0 0.6 0.8 20 −5 −2
Any 𝑥 in ℝ2
Change of Rotation Change of
coordinate by36.86° coordinate
back to 𝑥 = 𝑢 = 𝑃−1 𝑥
𝑃𝑢
EE143 Dept. of E & E, NITK Surathkal
31
Complex Eigenvalues
Ex-24 Let 𝐴 be a real valued 𝑛 × 𝑛 matrix, and let 𝑥 be a vector in ℂ𝑛 . Show that
Re 𝐴𝑥 = 𝐴 Re 𝑥 and Im 𝐴𝑥 = 𝐴 Im 𝑥
Proof 𝐴𝑥 = 𝐴 Re 𝑥 + 𝑖 Im 𝑥 = 𝐴Re 𝑥 + 𝑖 A Im 𝑥
Now Re 𝐴𝑥 = Re 𝐴Re 𝑥 + 𝑖 A Im 𝑥 = 𝐴Re 𝑥
And Im 𝐴𝑥 = Im 𝐴Re 𝑥 + 𝑖 A Im 𝑥 = 𝐴Im 𝑥
Ex-25 Let 𝐴 be a real valued 2 × 2 matrix, with a complex eigenvalue 𝜆 = 𝑎 −
𝑏𝑖 (𝑏 ≠ 0) and 𝑣 an associated eigenvector in ℂ2 .
1. Show that ARe 𝑣 = 𝑎 Re 𝑣 + 𝑏 Im 𝑣 & A Im 𝑣 = −𝑏 Re 𝑣 + 𝑎 Im 𝑣
𝑎 −𝑏
2. Verify that if 𝑃 = 𝑅𝑒 𝑣 𝐼𝑚 𝑣 & 𝐶 = , then 𝐴𝑃 = 𝑃𝐶
𝑏 𝑎
Proof-1 We have A𝑣 = 𝜆𝑣 = 𝑎 − 𝑏𝑖 𝑣 = (𝑎 − 𝑏𝑖)(𝑅𝑒 𝑣 + 𝐼𝑚 𝑣 𝑖)
Proof-2 is now direct = 𝑎𝑅𝑒 𝑣 + 𝑏𝐼𝑚 𝑣 + 𝑖 −𝑏 𝑅𝑒 𝑣 + 𝑎𝐼𝑚 𝑣
multiplication 𝐼𝑚 𝐴𝑣 = 𝐴𝐼𝑚 𝑣
𝑅𝑒 𝐴𝑣 = 𝐴𝑅𝑒 𝑣
Theorem Let 𝐴 be a 2 × 2 matrix with a complex eigenvalue 𝜆 = 𝑎 − 𝑏𝑖 𝑏 ≠ 0 and
𝑣 an associated eigenvector in ℂ2 . Then
𝑎 −𝑏
𝐴 = 𝑃𝐶𝑃−1 where 𝑃 = 𝑅𝑒 𝑣1 𝐼𝑚 𝑣1 and 𝐶 =
𝑏 𝑎
EE143 Dept. of E & E, NITK Surathkal
32
Questions?
EE143 Dept. of E & E, NITK Surathkal
33
