• No results found

How to find the eigenvalues and eigenvectors of a symmetric 2x2 matrix.pdf

N/A
N/A
Protected

Academic year: 2021

Share "How to find the eigenvalues and eigenvectors of a symmetric 2x2 matrix.pdf"

Copied!
5
0
0

Loading.... (view fulltext now)

Full text

(1)

Version of 25 June 2012 Corrections and Suggestions to Vince_Cronin@baylor.edu

How to find the eigenvalues and eigenvectors of a symmetric 2x2 matrix

Introduction

We will leave the theoretical development of eigensystems for you to read in textbooks on linear algebra or tensor mathematics, or from reliable sources on the web such as those listed in the references section at the end of this document. Here, we accept that for any given stress or strain tensor, a coordinate system can be identified in which all of the shear stresses or shear strains have zero value, and the only non-zero values are along the diagonal of the tensor matrix. The eigenvectors are parallel to those special coordinate axes, and the eigenvalues are the values along the diagonal. Another way of characterizing them is that the eigenvectors are along the principal directions of the stress or strain ellipsoids, and the eigenvalues are the magnitudes of the principal stresses or strains.

Let’s call the square matrix we are analyzing matrix M.

M = d e f g ⎡ ⎣ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥

We want to find all possible values for a variable we will call λ that satisfy the following

characteristic equation:

det(M – λI) = 0, or using another common symbol for determinant,

MλI = 0,

Matrix I in the preceding equations is the identity matrix

(I = 1 0 0 1 ⎡ ⎣ ⎢ ⎤ ⎦ ⎥ if M is a 2x2 matrix) and variable λ is an eigenvalue. Then

(M – λI) = d e f g ⎡ ⎣ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥− λ 0 0 λ ⎡ ⎣ ⎢ ⎤ ⎦ ⎥ = (d−λ) e f (g−λ) ⎡ ⎣ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ The characteristic equation is

(d−λ) e

f (g−λ) = 0 This determinant can be unpacked as the equation

((d – λ)(g – λ)) – (f e) = 0 or, simplifying,

(2)

The result is a quadratic equation. You might remember how to solve a quadratic equation from deep in your childhood. In case you don’t, the general solution for the quadratic equation

(a x2) + (b x) + c = 0 is

x= −b± b

2− 4ac 2a

This expression, which is known as the quadratic function, yields two answers for x that satisfy the quadratic equation. In our case, the equivalent terms in the quadratic function are

a = 1 b = −(d + g),

noting that [−(d + g)]2 = (d + g)2, and

c = ((d g) − (f e))

Consequently, the eigenvalues are given by the quadratic function

λ = (d+ g) ± (d + g)2− 4(dg − fe) 2

or, somewhat simplified,

λ=(d+ g) ± 4 ∗e∗ f

(

)

+ d − g

(

)

2

2 .

We now have two values of λ that satisfy our quadratic equation, and these are the two eigenvalues of our 2x2 matrix. We will refer to the larger eigenvalue as λ1, and the smaller

eigenvalue is λ2.

Now we need to find the eigenvectors that correspond to λ1 and λ2, respectively. Returning to

our example using matrix M, we have the following equation to solve to find the eigenvector associated with λ1 0 0 ⎡ ⎣ ⎢ ⎤ ⎦ ⎥ = dfλ1 geλ 1 ⎡ ⎣ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ x1 y1 ⎡ ⎣ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥, which can be restated as

((d – λ1) x1) + (e y1) = 0 and

(f x1) + ((g – λ1) y1) = 0.

The best we can do with these two equations and their two unknown values (x1 and y1) is to

determine how one of the unknowns relates to the other. In other words, we can determine the slope of the vector. If we arbitrarily choose x1 = 1, we can rearrange either or both of the

previous equations to determine the value of y1 when x1 = 1:

y1= − d −

(

λ1

)

e = λ1− d

(

)

e or

(3)

y1= − f g−λ1

(

)

An eigenvector that corresponds with eigenvalue λ1 has the following coordinates:

1,

(

λ1− d

)

e ⎧ ⎨ ⎩ ⎫ ⎬ ⎭ The length or norm of that vector is

12+

(

λ1− d

)

e ⎛ ⎝⎜ ⎞ ⎠⎟ 2

Recalling that a unit vector is a vector whose length is 1, we can find the unit vector associated with any vector of arbitrary length by dividing each component of the initial vector by that vector’s length or norm. The unit eigenvector that corresponds to eigenvalue λ1 is

1 12+

(

λ1− d

)

e ⎛ ⎝⎜ ⎞ ⎠⎟ 2, λ1− d

(

)

e 12+

(

λ1− d

)

e ⎛ ⎝⎜ ⎞ ⎠⎟ 2 ⎧ ⎨ ⎪ ⎪ ⎩ ⎪ ⎪ ⎫ ⎬ ⎪ ⎪ ⎭ ⎪ ⎪

We repeat the process to find the coordinates of the unit eigenvector that corresponds to eigenvalue λ1.

Worked example.

Given the matrix M below, calculate the eigenvalues and the corresponding unit eigenvectors.

M = 6 3 3 4 ⎡ ⎣ ⎢ ⎤ ⎦ ⎥ Solution.

The eigenvalues are

(6+ 4) + 4 ∗ 3∗ 3

(

)

+ 6 − 4

(

)

2 2 ⎛ ⎝ ⎜ ⎜ ⎞ ⎠ ⎟ ⎟ = 8.16228 and (6+ 4) − 4 ∗ 3∗ 3

(

)

+ 6 − 4

(

)

2 2 ⎛ ⎝ ⎜ ⎜ ⎞ ⎠ ⎟ ⎟ = 1.83772

We refer to the larger eigenvalue (8.16228) as λ1, and the smaller (1.83772) is λ2. An

eigenvector associated with λ1 is

8.16228− 6

(

)

(4)

The unit eigenvector associated with λ1 is found by dividing the individual components of the

previous vector by the vector’s length

1 12+

(

8.16228− 6

)

3 ⎛ ⎝⎜ ⎞ ⎠⎟ 2 , 8.16228− 6

(

)

3 12+

(

8.16228− 6

)

3 ⎛ ⎝⎜ ⎞ ⎠⎟ 2 ⎧ ⎨ ⎪ ⎪ ⎩ ⎪ ⎪ ⎫ ⎬ ⎪ ⎪ ⎭ ⎪ ⎪ = {0.811242, 0.58471}

An eigenvector associated with λ2 is

1,

(

1.83772− 6

)

3 ⎧ ⎨ ⎩ ⎫ ⎬ ⎭ and the corresponding unit eigenvector associated with λ2 is

1 12+

(

1.83772− 6

)

3 ⎛ ⎝⎜ ⎞ ⎠⎟ 2 , 1.83772− 6

(

)

3 12+

(

1.83772− 6

)

3 ⎛ ⎝⎜ ⎞ ⎠⎟ 2 ⎧ ⎨ ⎪ ⎪ ⎩ ⎪ ⎪ ⎫ ⎬ ⎪ ⎪ ⎭ ⎪ ⎪ = {0.58471, -0.811242}

In the specific case in which we have a 2-D Lagrangian strain tensor

εij = ε11 ε12 ε21 ε22 ⎡ ⎣ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ the eigenvalues are given by

λ = (ε11+ε22)± 4ε

(

12ε21

)

+

(

ε11−ε22

)

2

2 .

The larger eigenvalue is λ1, and the smaller eigenvalue is λ2. An eigenvector corresponding to

λ1 is 1,λ1−ε11 ε12 ⎧ ⎨ ⎩ ⎫ ⎬ ⎭, and an eigenvector associated with λ2 is

1,λ2−ε11 ε12 ⎧ ⎨ ⎩ ⎫ ⎬ ⎭.

The unit eigenvectors can then be determined by dividing each of the components of these vectors by their length or norm. The unit eigenvector associated with eigenvalue λ1 is

(5)

1 1+

(

λ1−ε11

)

ε12 ⎛ ⎝⎜ ⎞ ⎠⎟ 2 , λ1−ε11

(

)

ε12 1+

(

λ1−ε11

)

ε12 ⎛ ⎝⎜ ⎞ ⎠⎟ 2 ⎧ ⎨ ⎪ ⎪ ⎩ ⎪ ⎪ ⎫ ⎬ ⎪ ⎪ ⎭ ⎪ ⎪

The unit eigenvector associated with eigenvalue λ2 is

1 1+

(

λ2−ε11

)

ε12 ⎛ ⎝⎜ ⎞ ⎠⎟ 2 , λ2−ε11

(

)

ε12 1+

(

λ2−ε11

)

ε12 ⎛ ⎝⎜ ⎞ ⎠⎟ 2 ⎧ ⎨ ⎪ ⎪ ⎩ ⎪ ⎪ ⎫ ⎬ ⎪ ⎪ ⎭ ⎪ ⎪ . References

Ferguson, J., 1994, Introduction to linear algebra in geology: London, Chapman & Hall, 203 p., ISBN 0-412-49350-0.

Online resources

Eigenvalue calculator, accessed 25 June 2012 via

http://www.wolframalpha.com/entities/calculators/eigenvalue_calculator/gv/9h/bi/

Weisstein, E.W., Eigenvalue: MathWorld—A Wolfram web resource, accessed 25 June 2012 via http://mathworld.wolfram.com/Eigenvalue.html

Weisstein, E.W., Eigenvector: MathWorld—A Wolfram web resource, accessed 25 June 2012 via http://mathworld.wolfram.com/Eigenvector.html

Weisstein, E.W., Quadratic equation: MathWorld—A Wolfram web resource, accessed 25 June 2012 via http://mathworld.wolfram.com/QuadraticEquation.html

Eigenvalue calculator, accessed 25 June 2012 via

References

Related documents

The revenue of the Group increased by 23.0% to £65.16 million (2014 revenue: £52.96 million), which was mainly contributed by growth in the mobile phone prepaid airtime reload and

Now we pass 3 dates to a method we have created, before we declare how many variables we use declaring the method (we can put them between parenthesis), after we recall the

Goodman Matthew ENGR Materials Science and Engineering 02-1 2021 A VACANT ENGR Materials Science and Engineering 02-2 2020 Al-Qadi Imad L ENGR Civil and Environmental Engineering

Hvis Regjeringens publikumskostnad øker så mye at Opprørernes forventede nytte av krig blir større enn det største kravet Regjeringen med høye krigskostnader aksepterer (a R >

• how they and others choose and adapt features and functions of spoken language for specific purposes in different contexts. • some common influences on spoken language and

In this updated policy, Hulu.com includes a link to Adobe’s Flash cookie manager, discloses that it uses Flash cookies, but then downplays their potential for tracking: “Local

Where the Client terminates Services at will although the Project is continuing, the Consultant is entitled to 5% of the difference between the Fees payable under clause 14.26

Objectives To explore the association between depressive symptoms and recent head- related trauma (diagnosed concussion, subconcussive impacts) in semiprofessional male