Finding the Eigenvalues and Eigenvectors of a Matrix
Finding the eigenvalues and eigenvectors of a matrix are
fundamental operations in linear algebra. The Mathcad functions
eigenvals and eigenvecs perform these operations. You can also
use eigenvecs to determine whether a matrix can be diagonalized.
Eigenvalues and Eigenvectors
A number c is an eigenvalue of a matrix A if there is a non-zero
vector v, called an eigenvector for c, so that the following equation
is true:
A v = c v
Here's an example:
9 2 5
2
A 2 5 2
5
1
2
2
2
1
v 1
1
Since
5
A v 5
5
is equal to
5
5 v 5
5
v is an eigenvector for A, corresponding to eigenvalue c = 5.
Geometrically, this means that multiplying A times v stretches v by
a factor of 5.
07feb_Eigenvalues.xmcd
7/29/2011
The function eigenvals returns a vector containing the eigenvalues of A.
1
eigenvals ( A) 5
4
The function eigenvecs returns a matrix whose columns are the
corresponding eigenvectors of A.
U eigenvecs ( A)
0.267 0.577 0.196
U 0.535 0.577 0.784
0.802 0.577 0.588
Note that the second column of U is a scalar multiple of the
eigenvector v above.
To return a single column of U, first type
ORIGIN 1
Doing so starts the numbering of the columns at 1. Next, type U
followed by [Ctrl] 6 to insert the column operator.
U
Finally, type the number of the column in the placeholder to the
right of U. For example, the third column of U is
0.196
3
U 0.784
0.588
07feb_Eigenvalues.xmcd
7/29/2011
You can verify that this is the eigenvector corresponding to 4, the
third entry of the vector of eigenvalues.
0.784
0.784
3
3
A U 3.138 is equal to 4 U 3.138
2.353
2.353
Diagonalizing a Matrix
A square matrix is called a diagonal matrix if its only non-zero
entries are on the main diagonal. For example:
1 0 0
0 5 0
0 0 4
The eigenvalues of a diagonal matrix are its diagonal entries. The
corresponding eigenvectors are the standard basis vectors of R3.
If a matrix A is not diagonal, it is often necessary to diagonalize it to find an invertible matrix U and a diagonal matrix D that satisfy
the matrix equation
A U = D
If you can find the matrices U and D, then A is diagonalizable.
How can you tell if an n-by-n matrix A is diagonalizable? The
answer is that A is diagonalizable if it has n linearly independent
eigenvectors. You can determine whether this is true by evaluating
eigenvecs(A) symbolically, provided that the entries of A are
written as common fractions, such as
9
, rather than as decimals,
2
such as 4.5. If the result is a square matrix, then A is
diagonalizable.
Here's how to test whether a matrix is diagonalizable:
07feb_Eigenvalues.xmcd
7/29/2011
Diagonalizability Test
1. Type
U eigenvecs ( A)
2. Press [Ctrl] [.], followed by [Enter], to evaluate U with the
symbolic equal sign.
If U is a square matrix - it has the same number of columns as
rows - then A is diagonalizable. U is invertible and satisfies the
matrix equation above, where D is the diagonal matrix whose
diagonal entries are the eigenvalues of A.
If U has fewer columns than rows, A is not diagonalizable. Since U
is not a square matrix, it cannot be invertible.
Note: You must use the symbolic equal sign for this test, and the
entries of A must be symbolic expressions. Evaluating U with the
numerical equal sign always returns a square matrix, even when A is
not diagonalizable.
Let's apply the test to the matrix A in the example above:
U eigenvecs ( A)
1
3
U 2
1
3
3
1
Since U is a square matrix, A is diagonalizable. The diagonal
matrix D is
1 0 0
1
U A U 0 5 0
0 0 4
The result is a diagonal matrix whose diagonal entries are the
eigenvalues of A.
Note: When you evaluate U symbolically, the columns are displayed
in a different order than when you evaluate it numerically. However,
this does not affect the test.
07feb_Eigenvalues.xmcd
7/29/2011
Not all matrices are diagonalizable. Here's an example.
2 0 1
B 1 2 2
0 0 3
If you apply the diagonalizability test to B, the result is
U eigenvecs ( B)
0 1
U 1 3
0 1
Since U is not a square matrix, B is not diagonalizable.
What this means is that B does not have three linearly independent
eigenvectors, so its eigenvectors do not span all of R3.
07feb_Eigenvalues.xmcd
7/29/2011
07feb_Eigenvalues.xmcd
7/29/2011
