Exercise 5.62:
-
For you may assume that .
You may assume that all eigenvalues of these matrices are integers (they were constructed this way for ease of computation, but this will not be true in general). For each of these matrices :
- i.
Find the characteristic polynomial and minimal polynomial.
- ii.
For each eigenvalue, find a basis for every generalized eigenspace .
- iii.
Find the JNF.
- iv.
Find a Jordan basis.
Solution.[5.62] . We compute its characteristic polynomial as follows:
So there are three different eigenvalues. Therefore there is only one possibility for the minimal polynomial (which is a factor of , is monic, and also has the same roots as ):
Next we determine all of the generalized eigenspaces. First we find the generalized eigenspaces of index 1 for each eigenvalue:
These are all 1 dimensional. One could check that are also 1-dimensional, by computing . But alternatively, the existence of the Jordan normal form of a matrix shows that any collection of generalized eigenvectors for different eigenvalues must be linearly independent. Therefore the dimensions of , , and must together add up to less than or equal to 3. But that means they are each of dimension 1.
So a Jordan basis is given by 3 distinct Jordan chains of length 1 (i.e. eigenvectors), and we can take:
-
,
-
,
-
.
There together form a Jordan basis. Now forming the transition matrix and performing the final matrix multiplication:
-
,
-
,
-
.
This is the Jordan normal form of . Notice that had we chose to write our basis in a different order, then the Jordan blocks of our final matrix would have been in a different order.
. We find the characteristic polynomial of as follows.
So there are two possible polynomials which are simultaneously (i) a factor of , (ii) monic, and (iii) have the same roots as . So is one of these two:
-
,
-
.
First we test the one of smallest degree to see whether or not . We compute:
Therefore the minimal polynomial of .
To find the generalized eigenspaces for each eigenvalue, let’s start with the generalized eigenspaces of index 1:
Since these dimensions add up to 3, there is a basis of eigenvectors, and therefore the generalized eigenspaces of higher index are no bigger than the usual eigenspaces. In other words, and for all .
So the Jordan normal form of is a diagonal matrix, , and a Jordan basis is made up of three Jordan chains of length 1, which we could take as follows:
-
,
-
,
-
.
So we can now form the transition matrix, and perform the final verification:
-
,
-
,
-
.
. JNF is .
. JNF is .
. JNF is .
Exercise 5.63:
Assume a matrix has the following characteristic polynomial. Find all possible JNF’s for (up to reordering of the Jordan blocks).
- i.
,
- ii.
.
Solution.[5.63] (i) Since the characteristic polynomial does not change under a similarity transformation, the Jordan norm form of must have diagonal entries consisting of two 1’s and two ’s. The only choice is whether, for each eigenvalue, there is a single Jordan block of size 2, or two blocks of size 1. So there are four possibilities:
-
,
-
,
-
,
-
.
These may also be written explicitly as matrices as follows:
-
,
-
,
-
,
-
.
It is equally correct to have written the Jordan blocks in a different order.
[5.63] (ii) Since the characteristic polynomial doesn’t change under similarity, the JNF must have three 1’s on the diagonal, and one . So the only possible choices for Jordan normal forms as:
-
,
-
,
-
.
So, up to rearranging the Jordan blocks, there are only three possible JNF’s. If we were to write these as matrices, they would look as follows:
-
,
-
,
-
.
Exercise 5.64:
Let be the differentiation map.
- i.
Choose a basis of , and write down the matrix .
- ii.
What is the minimal polynomial of ?
- iii.
Find the JNF of .
Solution.[5.64] First we will write down the matrix of in the standard basis of . To do this, we compute for each basis vector:
-
,
-
,
-
,
-
.
Then we find the coordinate matrices of these vectors, and the matrix of is constructed by taking those coordinate matrices as its columns:
The characteristic polynomial of this matrix is , and so the choices for the minimal polynomial are:
-
,
-
,
-
,
-
.
To see which it is, we need to evaluate each of these polynomials are the matrix . We evalute as follows:
So the minimal polynomial of is . Unlike matrices, for matrices, to find the Jordan normal form it is not enough to consider the characteristic and minimal polynomials. For matrices of size 4 and higher, it is possible to have matrices with different JNF’s (even up to reordering of blocks), but with the same characteristic and minimal polynomials. So to find the JNF we need to consider the generalized eigenspaces.
Since the only eigenvalue is , we could consider the generalized eigenspaces for that eigenvalue. In fact, all we will need is the dimension of the first one:
So . By Theorem 5.47 the number of Jordan blocks is 1, so the JNF must be .
Exercise 5.65:
Assume are similar matrices, and let be an eigenvalue (by Exercise 3.55, and have the same eigenvalues). Prove that is the same for as it is for .
Solution.[5.65] Assume that for some invertible matrix . Firstly we need to understand how similarity changes the kernel of a matrix. If is a square matrix of size , then the following statements are equivalent to each other:
-
,
-
,
-
.
Since is a bijective linear map, this proves that the dimension of is the same as the dimension of . In particular,
Therefore, the generalized eigenspaces of index have equal dimensions for similar matrices and .
Exercise 5.66:
Consider the function defined by
For example, .
- •
Choose a basis of , and find the matrix .
- •
Find the JNF of that matrix
- •
Find a Jordan basis for ; this should form a basis of .
Solution.[5.66] Let , which is the easiest basis of . To find the matrix of , we need to compute where transforms the elements of the basis:
-
,
-
,
-
.
So the columns of the matrix are the coordinates of the vectors with respect to the basis . Therefore
The characteristic polynomial of this matrix is , since it is upper-triangular. This shows that the only eigenvalue is 1. Since it is a matrix, to find the JNF, it is enough to compute the minimal polynomial of (recall that this technique doesn’t work for matrices of size 4 or larger). The possibilities for the minimal polynomial are:
-
,
-
,
-
.
To check which one is correct, we evaluate the matrix at each of these possibilities, and the one of smallest degree which evaluates to zero is the minimal polynomial. We compute:
Therefore the minimal polynomial is , and hence the JNF is .
To find a Jordan basis, we know we need to find a single Jordan chain of length 3. To do this, we need to find an element . According to our matrix multiplications above, we can deduce that
-
,
-
.
So we define
-
,
-
,
-
.
Then the sequence forms a Jordan chain of length 3. In particular, it forms a Jordan basis.
As a final check, we could construct the transition matrix, and make the following verification:
-
,
-
,
-
.
Exercise 5.67:
A student is asked to prove Theorem 5.9(i). He writes the following:
[Start of student’s work]
Take two monic polynomials of minimal degree such that , and assume . Since and are both of the same degree , and are monic, the polynomial is monic of degree . But notice that
So contradicts the minimality of . Therefore, our assumption must have been false. This proves , and in other words, the polynomial is unique.
[End of student’s work]
This student has made a logical mistake. What is it, and how could it be fixed?
Exercise 5.68:
A student is asked to prove the Cayley-Hamilton theorem (which is not an easy thing to do, and is omitted from this module). He writes the following:
[Start of student’s work]
Substitute with in the characteristic polynomial. Then
Therefore, for any square matrix , as required.
[End of student’s work]
Identify the student’s mistake. You do not have to give a correct proof.
Exercise 5.69:
Prove that a matrix is diagonalizable if and only if
where for .
Solution.[5.69] There are two statements that need to be proved, because it says “if and only if”. For either direction, we can define distinct eigenvalues . By putting into Jordan normal form, we see that the characteristic polynomial of must be
Here is called the multiplicity of the eigenvalue .
Assume is diagonalizable, so there is an invertible matrix such that , where ; it may be the case that each occurs multiple times in the diagonal matrix . The minimal polynomials of and are the same, since they are similar matrices. Now I claim that is the minimal polynomial. Since the minimal polynomial must share the same roots as the characteristic polynomial, this is the smallest possibility that might be the minimal polynomial. We just need to check that it sends the matrix to the zero matrix. Indeed, is the product of diagonal matrices, and the resulting matrix is the zero matrix. This proves that is the minimal polynomial, as required.
Next, assume that is not diagonalizable. Then we can put is Jordan normal form, , and since it is not diagonalizable, there must exist at least one Jordan block of size . To prove that is not the minimal polynomial, it is enough to check that is not the zero matrix; in other words there is a vector such that . Since we have assumed is not diagonalizable, we can choose a , for some eigenvalue . Since we are free to reorder the eigenvalues, there is no loss of generality in assuming , the last eigenvalue. Then
Since we have assumed are the distinct eigenvalues, for any , we use that is an eigenvector to see that:
In particular, this proves . Therefore is not the minimal polynomial of , and hence it is also not the minimal polynomial of . This finishes the proof.
Exercise 5.70 (Bonus):
Let , and consider the set of all matrices which are similar to . This is called the orbit of under the conjugation action. [ Aside: The words ‘‘orbit’’, ‘‘conjugation’’, and ‘‘action’’ will all be defined in MATH321.]
How many different orbits are there, among nilpotent complex matrices? Recall the definition of “nilpotent” from Exercise 3.62.
Learning objectives for Chapter 5:
Pass Level: You should be able to…
-
Verify the Cayley-Hamilton theorem for specific matrices (e.g. Exercise 5.4).
-
For a given (mostly factored) polynomial, produce a list of all possible monic factors which share the same roots (e.g. Exercise 5.14).
-
Given a matrix and its factored characteristic polynomial, find its minimal polynomial (e.g. Exercise 5.16).
-
Be able to express any generalized eigenspace of a matrix as the kernel of another matrix (e.g. Definition 5.18).
-
Given the generalized eigenspaces of a matrix, deduce its JNF using Theorem 5.47 (e.g. Exercise 5.49).
-
Correctly answer, with full justification, at least 50% of the true / false questions relevant to this Chapter.
First class level: You should be able to…
-
Explain, in your own words, the main ideas used in the proofs of Theorem 5.11, Theorem 5.22, and Theorem 5.26
-
Identify, in your own words, the most important concepts and results of this Chapter.
-
Write a complete solution, without referring to any notes, to at least 80% of the exercises in this Chapter, and in particular the proof questions.
-
Correctly answer, with full justification, all of the true / false questions relevant to this Chapter.