EigenValuesEigenvectors

download EigenValuesEigenvectors

of 14

Transcript of EigenValuesEigenvectors

  • 8/4/2019 EigenValuesEigenvectors

    1/14

    Primer on Eigenvalues andEigenvectors

    Copyright Brian G. Higgins (2003)

    Introduction

    In this notebook we discuss the solution of the following matrix problem

    (1)Ax = l x

    where A is a nn matrix, x is a 1n column vector, and l is a scalar constant. Matrix problems of this form occur

    frequently in the solution of differential equations, and have important physical significance in applications in the

    physical sciences and engineering.

    We explore various analytically methods commonly used to solve the algebraic eigenvalue problem when n is small.

    We also discuss important properties of the eigenvalues and eigenvectors for these systems, and show howMathematica

    can be used to perform these calculations.

    The contents of this notebook are the essential building blocks for studying more advanced topics in linear algebra,

    numerical methods, statistics and differential equations. Thus all chemical engineering students should master this

    material early in their academic careers.

    Eigenvalue/Eigenvector Calculations

    In this section we illustrate how to compute eigenvalues and eigenvectors for a nn matrix A. We also discuss various

    properties of the eigenvalues and eigenvectors. We start with a general definition

    Definition 1

    Let A be a nn matrix. A scalar l is called an eigenvalue ofA if there exists a nonzero 1n column vector x that is a

    solution to the following system of linear equations:

    (2)A x= l x

    The solution vector x is called an eigenvector, and the scalar l is called its associated eigenvalue.

    EigenValuesEigenvectors.nb 1

  • 8/4/2019 EigenValuesEigenvectors

    2/14

    Example 1: Eigenvectors of a Matrix A

    Suppose we have the matrix Agiven by

    A=

    i

    k

    jjjjjjj

    1 -1 0

    0 1 1

    0 0 -2

    y

    {

    zzzzzzz

    then

    x=

    i

    k

    jjjjjjj1

    3

    -9

    y

    {

    zzzzzzzis an eigenvector since

    A x=

    i

    k

    jjjjjjj1 -1 0

    0 1 1

    0 0 -2

    y

    {

    zzzzzzz i

    k

    jjjjjjj1

    3

    -9

    y

    {

    zzzzzzz =i

    k

    jjjjjjj-2

    -6

    18

    y

    {

    zzzzzzz = H-2Li

    k

    jjjjjjj1

    3

    -9

    y

    {

    zzzzzzz

    with l = -2. Thus the associated eigenvalue ofx is l = -2

    Property 1

    The eigenvalues and eigenvectors of a matrix A are determined by solving the following homogeneous system of linear

    equations:

    HA11 - lL x1 +A12 x2 + + A1n xn = 0 A21 x1 + HA22 - lL x2 + + A2n xn = 0

    An1 x1 +An2 x2 + + HAnn - lL xn = 0

    where x= 8x1 , x2 , , xn

  • 8/4/2019 EigenValuesEigenvectors

    3/14

    Example 2: Calculation of eigenvalues and eigenvectors of a matrix A: Distinct Eigenvalues

    Suppose we are given the following matrix:

    A=

    i

    k

    jjjjjjj

    1 2 1

    6 -1 0

    -1 -2 -1

    y

    {

    zzzzzzz

    Determine the eigenvalues li and the associated eigenvectors xi ofA.

    First we determine the characteristic equation for the eigenvalues

    DetAi

    k

    jjjjjjj1 - l 2 1

    6 -1 - l 0

    -1 -2 -1 - l

    y

    {

    zzzzzzzE = 0

    Expanding the determinant gives

    PHlL = H1 - lL8H-1 - lLH-1 - lL< - 286H-1 - lL< + 18-12 + H-1 - lL< = 0Multiplying out gives

    PHlL = 12 l - l2 - l3 = 0

    Thus the roots ofPHlL = 0 are then

    PHlL = l Hl + 4LHl - 3L = 0

    Hence the eigenvalues are:

    l1 = 0, l2 = -4, l3 = 3

    To find the eigenvectors xi we need to solve the following system of homogeneous equations for each li :

    Axi = li xi

    Consider first l1 = 0. The equations we must solve are

    i

    k

    jjjjjjj1 2 1

    6 -1 0

    -1 -2 -1

    y

    {

    zzzzzzzi

    k

    jjjjjjjx11

    x12

    x13

    y

    {

    zzzzzzz = 0

    Thus we obtain

    x11 + 2 x12 + x13 = 0, 6 x11 - x12 = 0, -x11 - 2 x12 - x13 = 0

    We can solve this system by Gauss elimination to obtain

    x11 = -1

    13x13 , x12 = -

    613

    x13

    Hence the eigenvector is

    EigenValuesEigenvectors.nb 3

  • 8/4/2019 EigenValuesEigenvectors

    4/14

    x1 =

    i

    k

    jjjjjjjjj

    - 113

    a

    - 613

    a

    a

    y

    {

    zzzzzzzzz, where a is arbitrary

    A suitable eigenvector is

    x1 =i

    k

    jjjjjjj1

    6

    -13

    y

    {

    zzzzzzz

    Consider next l2 = -4. The equations we must solve are

    i

    k

    jjjjjjj5 2 1

    6 3 0

    -1 -2 3

    y

    {

    zzzzzzzi

    k

    jjjjjjjx21

    x22

    x23

    y

    {

    zzzzzzz = 0

    We obtain the following system of equations

    5 x21 + 2 x22 + x23 = 0, 6 x21 + 3 x22 = 0, -x21 - 2 x22 + 3 x23 - 0

    Solving this set gives

    x22 = 2 x23 , x21 = -x23

    Thus the eigenvector is

    x2 =

    i

    k

    jjjjjjj-a

    2 a

    a

    y

    {

    zzzzzzz where a is arbitrary.

    A suitable vector is then

    x2 =

    i

    k

    jjjjjjj-1

    2

    1

    y

    {

    zzzzzzz

    A similar calculation for l3 = 3 gives

    i

    k

    jjjjjjj-2 2 1

    6 -4 0

    -1 -2 -4

    y

    {

    zzzzzzzi

    k

    jjjjjjjx31

    x32

    x33

    y

    {

    zzzzzzz = 0

    The eigenvector from this calculation is

    x3 =

    i

    k

    jjjjjjj2

    3

    -2

    y

    {

    zzzzzzz

    To summarize: In this problem we found 3 distinct eigenvalues, and 3 associated eigenvectors

    EigenValuesEigenvectors.nb 4

  • 8/4/2019 EigenValuesEigenvectors

    5/14

    Example 2: Mathematica Solution

    Mathematica has several built-in functions for determining the eigenvalues and eigenvectors of a square matrix. To

    calculate the eigenvalues the function is Eigenvalues; to find the eigenvectors, the function is Eigenvectors; and to

    solve for both the eigenvalues and eigenvectors one can use Eigensystem. We illustrate this function using the matrix

    A.

    A=

    i

    k

    jjjjjjj1 2 1

    6 -1 0

    -1 -2 -1

    y

    {

    zzzzzzz;

    8Eigenvalues@AD, Eigenvectors@AD