ARO Typography Center

MKU Linear Algebra - Uni 4 : SMTJC61

MKUL4

Contents

Chapter 4 Theory of Matrices

4.1 Introduction

We have already seen that an \(m \times n\) matrix \(A\) is an array of \(mn\) numbers \(a_{ij}\) where \(1 \leq i \leq m\), \(1 \leq j \leq n\) arranged in \(m\) rows and \(n\) columns as follows

\[ A = \begin {pmatrix} a_{11} & a_{12} & \cdots & a_{1n} \\ a_{21} & a_{22} & \cdots & a_{2n} \\ \vdots & \vdots & \cdots & \vdots \\ a_{m1} & a_{m2} & \cdots & a_{mn} \\ \end {pmatrix} \]

We shall denote this matrix by the symbol \(a_{ij}\). If \(m=n\), \(A\) is called a square matrix of order \(n\).

4.2 Elementary Transformations

    Definition 4.2.1. Let \(A\) be an \(m \times n\) matrix over a field \(F\). An elementary row (column) operation on \(A\) is of any one of the following three types.

      1. The interchange of any two rows (columns).

      2. Multiplication of a row (column) by a non-zero element \(c\) in \(F\).

      3. Addition of any multiple of one row (column) with any other row (column).

    Example 4.2.2. Let \(A=\left (\begin {array}{rr}1 & 2 \\ 2 & 1 \\ 3 & -1\end {array}\right )\), \(A_{1}=\left (\begin {array}{rr}3 & -1 \\ 2 & 1 \\ 1 & 2\end {array}\right )\), \(A_{2}=\left (\begin {array}{rr}2 & 2 \\ 4 & 1 \\ 6 & -1\end {array}\right )\),
    \(A_{3}=\left (\begin {array}{rr}1 & 2 \\ 5 & 7 \\ 3 & -1\end {array}\right ) .\)
    \(A_{1}\) is obtained from \(A\) by interchanging the first and third rows.
    \(A_{2}\) is obtained from \(A\) by multiplying the first column of \(A\) by 2.
    \(A_{3}\) is obtained from \(A\) by adding to the second row the multiple by 3 of the first row.

    Note 4.2.3. We may use the following notations for elementary transformations.

      (i) Interchange of \(i^{\text {th }}\) and \(j^{\text {th }}\) rows will be denoted by \(R_{i} \leftrightarrow R_{j}\).

      (ii) Multiplication of \(i^{\text {th }}\) row by a non-zero element \(c \in F\) will be denoted by \(R_{i} \rightarrow c R_{i}\).

      (iii) Addition of \(k\) times the \(j^{\text {th }}\) row to the \(i^{\text {th }}\) row will be denoted by \(R_{i} \rightarrow R_{i}+k R_{j}\).

    The corresponding column operations will be denoted by writing \(C\) in the place of \(R\).

    Definition 4.2.4. An \(m \times n\) matrix \(B\) is said to be row equivalent (column equivalent) to an \(m \times n\) matrix \(A\) if \(B\) can be obtained from \(A\) by a finite succession of elementary row operations (column operations).

    \(A\) and \(B\) are said to be equivalent if \(B\) can be obtained from \(A\) by a finite succession of elementary row or column operations.

    If \(A\) and \(B\) are equivalent. We write \(A \sim B\).

    Definition 4.2.5. A matrix obtained form the identity matrix by applying a single elementary row or column operation is called an elementary matrix.

    Example 4.2.6. The following \(\left (\begin {array}{lll}0 & 1 & 0 \\ 1 & 0 & 0 \\ 0 & 0 & 1\end {array}\right )\), \(\left (\begin {array}{lll}4 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 1\end {array}\right )\), \(\left (\begin {array}{lll}1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 2 & 1\end {array}\right )\) are elementary matrices obtained from the identity matrix \(\left (\begin {array}{lll}1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 1\end {array}\right )\) by applying the elementary operations \(R_{1} \leftrightarrow R_{2}\), \(R_{1} \rightarrow 4 R_{1}\), \(R_{3} \rightarrow R_{3} +2 R_2 \) respectively.

    Theorem 4.2.7. Any elementary matrix is non-singular.

    Proof : The determinant of the identity matrix of any order is 1 . Hence the determinant of an elementary matrix obtained by interchanging any two rows is \(-1\). The determinant of an elementary matrix obtained by multiplying any row by \(k \neq 0\) is \(k\). The determinant of an elementary matrix obtained by adding a multiple of one row with another row is 1. Hence any elementary matrix is non-singular.

    Theorem 4.2.8. Let \(A\) be an \(m \times n\) matrix and \(B\) be an \(n \times p\) matrix. Then every elementary row (column) operation of the product \(A B\) can be obtained by subjecting the matrix \(A\) (matrix \(B\) ) to the same elementary row (column) operation.

    Proof : Let \(R_{1}, R_{2}, \ldots , R_{m}\) denote the rows of the matrix \(A\) and \(C_{1}, C_{2}, \ldots , C_{p}\) denote the columns of \(B\). By the definition of matrix multiplication

    \[ A B=\left (\begin {array}{cccc} R_{1} C_{1} & R_{1} C_{2} & \ldots \ldots & R_{1} C_{p} \\ R_{2} C_{1} & R_{2} C_{2} & \ldots \ldots & R_{2} C_{p} \\ \cdot & \cdot & \ldots \ldots & \cdot \\ \cdot & \cdot & \ldots \ldots & \cdot \\ \cdot & \cdot & \ldots \ldots & \cdot \\ R_{m} C_{1} & R_{m} C_{2} & \ldots \ldots & R_{m} C_{p} \end {array}\right ) \]

    It is obvious from the above representation of \(A B\) that if we apply any elementary row operation on \(A\) the matrix \(A B\) is also subjected to the same elementary row operation. Also if we apply any elementary column operation on \(B\) the matrix \(A B\) is also subjected to the same elementary column operation.

    Theorem 4.2.9. Each elementary row operation on an \(m \times n\) matrix \(A\) is equivalent to pre-multiplying the matrix \(A\) by the corresponding elementary \(m \times m\) matrix.

    Proof : Since \(A\) is an \(m \times n \) matrix we can write \(A=I A\) where \(I\) is the identity matrix of order \(m\). By Theorem 4.2.8 an elementary row operation on \(IA\) is equivalent to the same row operation on \(I\). But an elementary row operation on \(I\) gives an elementary matrix. Hence by pre-multiplying \(A\) by the corresponding elementary matrix we get the required row operation on \(A\).

    Note 4.2.10. Similarly each elementary column operation of an \(m \times n\) matrix \(A\) is equivalent to post-multiplying the matrix \(A\) by the corresponding elementary \(n \times n\) matrix.

    Corollary 4.2.11. If two \(m \times n\) matrices \(A\) and \(B\) are row equivalent then \(A=P B\) where \(P\) is a non-singular \(m \times m\) matrix.

    Proof : Since \(A\) is row equivalent to \(B\), \(A\) can be obtained from \(B\) by applying successive elementary row operations.
    Hence \(A=E_{1} E_{2} \ldots E_{n} B\) where each \(E_{i}\) is an elementary matrix. Since each \(E_{i}\) is non-singular, \(A=P B\) where \(P=E_{1} E_{2} \ldots E_{n}\) and \(P\) is non-singular.

    Corollary 4.2.12. If two matrices \(A\) and \(B\) are column equivalent then \(A=B Q\) where \(Q\) is a non-singular matrix.

    Corollary 4.2.13. If two \(m \times n\) matrices \(A\) and \(B\) are equivalent then \(A=P B Q\) where \(P\) is a non-singular \(m \times m\) matrix and \(Q\) is a non- singular \(n \times n\) matrix.

    Corollary 4.2.14. The inverse of an elementary matrix is again an elementary matrix.

    Proof : Let \(E\) be an elementary matrix obtained from \(I\) by applying some elementary operations. If we apply the reverse operation on \(E\), then \(E\) is carried back to \(I\). Let \(E^{*}\) be the elementary matrix corresponding to the reverse operation.
    Then \(E^{*} E=E E^{*}=I\). Hence \(E^{*}=E^{-1}\).
    Hence \(E^{-1}\) is also an elementary matrix.

Canonical form of a matrix

We now use elementary tow and column operations to reduce any matrix to a simple form, called the canonical form of a matrix.

    Theorem 4.2.15. By successive applications of elementary row and column operations, any non-zero \(m \times n\) matrix \(A\) can be reduced to a diagonal matrix \(D\) in Which the diagonal entries are either 0 or 1 and all the 1’s proceeding all the zeros on the diagonal.
    In other Words, any non-zero \(m \times n\) matrix is equivalent to a matrix of the form \(\left [\begin {array}{cc}I_{r} & 0 \\ 0 & 0\end {array}\right ]\) where \(I_{r}\) is the \(r \times r\) identity matrix and \(0\) is the zero matrix.

    Proof : We shall prove the theorem by induction on the number of rows of \(A\). Suppose \(A\) has just one row.
    Let \(A=\left (a_{11} a_{12} \ldots a_{1 n}\right )\).
    Since \(A \neq 0\), by interchanging columns, if necessary, we can bring a non-zero entry \(c\) to the position \(a_{11}\).
    Multiplying \(A\) by \(c^{-1}\) we get 1 as the first entry.
    Other entries in \(A\) can be made zero by adding - suitable multiples of 1 . Thus the result is true when \(m=1\).
    Now, suppose that the result is true for any non-zero matrix with \(m-1\) rows.
    Let \(A\) be a non-zero \(m \times n\) matrix. By permuting rows and columns we can bring some non-zero entry \(c\) to the position \(a_{11}\).
    Multiplying the first row by \(c^{-1}\) we get 1 as the first entry.
    All other entries in the first column can be made zero by adding suitable multiples of the first row to each other row.
    Similarly all the other entries in the first row can be made zero:
    This reduces \(A\) to a matrix of the form

    \[ B=\left (\begin {array}{ll} I_{1} & O \\ 0 & C \end {array}\right ) \]

    where \(C \) is an \((m-1) \times (n-1)\) matrix.
    Now by induction hypothesis \(C\) can be reduced to the desired form by elementary row and column operations.
    Hence \(A\) is equivalent to a matrix of the required form.

    Corollary 4.2.16. If \(A\) is an \(m \times n\) matrix there exist non-singular square matrices \(P\) and \(Q\) of orders \(m\) and \(n\) respectively such that \(P A Q=\left (\begin {array}{ll}I_r & 0 \\ 0 & 0\end {array}\right )\).
    The result follows from Corollary 4.2.13 of Theorem 4.2.9.

    Corollary 4.2.17. Any non-singular square matrix \(A\) of order \(n\) is equivalent to the identity matrix.

    Proof : By Corollary 4.2.16, \(P A Q=\left (\begin {array}{ll}I_{r} & O \\ O & 0\end {array}\right )\).
    Since \(P, A, Q\) are all non-singular \(\left (\begin {array}{ll}I_{r} & O \\ O & O\end {array}\right )\) is non-singular. This is possible if and only if \(\left (\begin {array}{cc}I_{r} & O \\ O & O\end {array}\right )=I_{n}\).

    Corollary 4.2.18. Any non-singular matrix \(A\) can be expressed as a product of elementary matrices.

    Proof : By Corollary 4.2.17, \(P A Q=I_{n}\). Hence \(A=P^{-1} Q^{-1}\). Further by Corollary 4.2.14 of Theorem 4.2.9, \(P^{-1}\) and \(Q^{-1}\) are products of elementary matrices.
    Hence \(A\) is a product of elementary matrices.

    Note 4.2.19. The inverse of a non-singular matrix \(A\) can be computed by using elementary transformations. Let \(A\) be a non-singular matrix of order \(n\). Then \(A A^{-1}=\) \(A^{-1} A=I\). Now, the non-singular matrix \(A^{-1}\) can be expressed as the product of elementary matrices.
    Let \(A^{-1}=E_{1} E_{2} \ldots E_{n}\).
    Then \(I=A^{-1} A=E_{1} E_{2} \ldots E_{n} A\).
    Thus every non-singular matrix \(A\) can be reduced to \(I\) by pre-multiplying \(A\) by elementary matrices.
    Hence \(A\) can be reduced to the identity matrix by applying successive elementary row operations.
    Now, \(A=I A\). Reduce the matrix \(A\) in the left hand side to \(I\) by applying successive elementary row operations and apply the same elementary row operations to the factor \(I\) in the right hand side. Then we get \(I=B A\) so that \(B=A^{-1}\).

Solved Problems

    Problem 4.2.20. Reduce the matrix \(A=\left (\begin {array}{rrr}1 & 2 & -1 \\ 1 & 1 & 2 \\ 2 & 4 & -2\end {array}\right )\) to the canonical form.

Solution :

\begin{align*} A & =\begin{pNiceArray}{ccc}1 & 2 & -1 \\ 1 & 1 & 2 \\ 2 & 4 & -2\end {pNiceArray}\\ &\sim \begin{pNiceArray}[last-col]{ccc} 1 & 2 & -1 \\ 0 & -1 & 3 & R_{2} \rightarrow R_{2}-R_{1}\\ 0 & 0 & 0 & R_{3} \rightarrow R_{3}-2 R_{1} \end {pNiceArray}\\ &\sim \begin{pNiceArray}[last-col]{ccc}1 & 0 & 0 \\ 0 & -1 & 3 & C_{2} \rightarrow C_{2}-2 C_{1}\\ 0 & 0 & 0 & C_{3} \rightarrow C_{3}+C_{1} \end {pNiceArray}\\ &\sim \begin{pNiceArray}[last-col]{ccc}1 & 0 & 0 \\ 0 & -1 & 0 \\ 0 & 0 & 0 & C_{3} \rightarrow C_{3}+3 C_{2} \end {pNiceArray}\\ &\sim \begin{pNiceArray}[last-col]{ccc} 1 & 0 & 0 \\ 0 & 1 & 0 & R_{2} \rightarrow -R_{2}\\ 0 & 0 & 0 \end {pNiceArray} \end{align*}

    Problem 4.2.21. Find the inverse of the matrix \(A=\left (\begin {array}{rrr}1 & 0 & 2 \\ 3 & 1 & -1 \\ -2 & 1 & 3\end {array}\right )\).

Solution :

\begin{align*} \begin{pNiceArray}{ccc} 1 & 0 & 2 \\ 3 & 1 & -1 \\ -2 & 1 & 3\end {pNiceArray} & = A \begin{pNiceArray}{ccc} 1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 1\end {pNiceArray} \\ \Rightarrow \begin{pNiceArray}{ccc} 1 & 0 & 2 \\ 0 & 1 & -7 \\ 0 & 1 & 7\end {pNiceArray}&=A \begin{pNiceArray}[last-col]{ccc} 1 & 0 & 0 \\ -3 & 1 & 0 & R_{2} \rightarrow R_{2}-3 R_{1} \\ 2 & 0 & 1 & R_{3} \rightarrow R_{3}+2 R_{1} \end {pNiceArray} \\ \Rightarrow \begin{pNiceArray}{ccc} 1 & 0 & 2 \\ 0 & 1 & -7 \\ 0 & 0 & 14 \end {pNiceArray}&=A \begin{pNiceArray}[last-col]{ccc} 1 & 0 & 0 \\ -3 & 1 & 0 \\ 5 & -1 & 1 & R_{3} \rightarrow R_{3}-R_{2} \end {pNiceArray} \\ \Rightarrow \begin{pNiceArray}{ccc}1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 1\end {pNiceArray} & =A \begin{pNiceArray}[last-col]{ccc} \frac {2}{7} & \frac {1}{7} & -\frac {1}{7} & R_{1} \rightarrow R_{1}-\frac {1}{7} R_{3} \\ -\frac {1}{2} & \frac {1}{2} & \frac {1}{2} & R_{2} \rightarrow R_{2}+\frac {1}{2} R_{3} \\ \frac {5}{14} & -\frac {1}{14} & \frac {1}{14}& R_{3} \rightarrow \frac {1}{14} R_{3} \end {pNiceArray}\\ \Rightarrow A^{-1}&=\begin{pNiceArray}{ccc} \frac {2}{7} & \frac {1}{7} & -\frac {1}{7} \\ -\frac {1}{2} & \frac {1}{2} & \frac {1}{2} \\ \frac {5}{14} & -\frac {1}{14} & \frac {1}{14}\end {pNiceArray} \end{align*}

    Definition 4.2.22. Let \(A\) and \(B\) be two square matrices of order \(n \). \(B\) is said to be similar to \(A\) if there exists a \(n \times n\) non-singular matrix \(P\) such that \(B=P^{-1} A P\).

Solved Problems

    Problem 4.2.23. Similarity of matrices is an equivalence relation in the set of all \(n \times n\) matrices.

Solution : Let \(S\) be the set of all \(n \times n\) matrices.
Let \(A \in S\).
Since \(A=I^{-1} A I\) and \(I\) is non-singular, \(A\) is similar to \(A\).
Hence similarity of matrices is reflexive.
Now, let \(A, B \in S\) and let \(A\) be similar to \(B\) matrix.

\[ A=P^{-1} B P\]

where \(P \in S\) is a non-singular. Now,

\begin{align*} P^{-1} B P=A & \Rightarrow P P^{-1} B P P^{-1}=P A P^{-1} \\ & \Rightarrow B=P A P^{-1} \\ & \Rightarrow B=\left (P^{-1}\right )^{-1} A\left (P^{-1}\right ) \end{align*} Since \(P\) is non-singular \(P^{-1} \in S\) is also non-singular.
\(B\) is similar to \(A\). Hence similarity of matrices is symmetric.
Now, let \(A, B, C \in S\).
Let \(A\) be similar to \(B\) to \(B\) be similar to \(C\). Hence there exist non-singular matrices \(P, Q \in S\) such that

\[ A=P^{-1} B P \text { and } B=Q^{-1} C Q \text {. } \]

Now,

\begin{align*} A& =P^{-1} B P\\ &=P^{-1}\left (Q^{-1} C Q\right ) P \\ &=\left (P^{-1} Q^{-1}\right ) C Q P \\ &=(Q P)^{-1} C(Q P) . \end{align*} Since \(P, Q \in S\) are non-singular, \(Q P \in S\) is also non-singular.
Hence \(A\) is similar to \(C\). \(\therefore \) Similarity of matrices is transitive.
Hence similarity of matrices is an equivalence relation.

    Problem 4.2.24. If \(A\) and \(B\) are similar matrices show that their determinants are same.

Solution : Let \(A\) and \(B\) be two similar matrices.
\(\therefore \) There exists a non-singular matrix \(P\) such that \(B=P^{-1} A P\). Now,

\begin{align*} |B| &=\left |P^{-1} A P\right | \\ &=\left |P^{-1} \| A\right ||P| \\ &=|A| \quad \left (\text { since }\left |P^{-1}\right |=\frac {1}{|P|}\right ) \end{align*} Hence the result.

4.3 Rank of a Matrix

We now proceed to introduce the concept of the rank of a matrix.

    Definition 4.3.1. Let \(A=\left (a_{i j}\right )\) be an \(m \times n\) matrix. The rows \(R_{i}=\left (a_{i 1}, a_{i 2}, \ldots \ldots \ldots , \ldots a_{i n}\right )\) of \(A\) can be thought of as elements of \(F^{A}\). The subspace of \(F^{n}\) generated by the \(m\) rows of \(A\) is called the row space of \(A\).

    Similarly, the subspace of \(F^{m}\) generated by the \(n\) columns of \(A\) is called the column space of \(A\).

    The dimension of the row space (column space) of \(A\) is called the row rank (column rank) of \(A\).

    Theorem 4.3.2. Any two row equivalent matrices have the same row space and have the same row rank.

    Proof : Let \(A\) be an \(m \times n\) matrix.
    It is enough if we prove that the row space of \(A\) is not altered by any elementary row operation.
    Obviously the row space of \(A\) is not altered by an elementary row operation of the type \(R_{i} \leftrightarrow R_{j}\).
    Now, consider the elementary row operation \(R_{i} \rightarrow c R_{i}\) where \(c \in F-\{0\} .\)
    Since \(L\left (\left \{R_{1}, R_{2}, \ldots \ldots , R_{i}, \ldots \ldots \ldots , R_{n}\right \}\right )=\) \(L\left (\left \{R_{1}, R_{2}, \ldots \ldots \ldots , c R_{i}, \ldots \ldots , R_{n}\right \}\right )\) the row space of \(A\) is not altered by this type of elementary row operation.
    Similarly we can easily prove that the row space of \(A\) is not altered by an elementary row operation of the type \(R_{i} \rightarrow R_{i}+c R_{i}\).
    Hence row equivalent matrices have the same row space and hence the same row rank.

Similarly we can prove the following theorem.

    Theorem 4.3.3. Any two column equivalent matrices have the same column rank.

    Theorem 4.3.4. The row rank and the column rank of any matrix are equal.

    Proof : Let \(A=\left (a_{i j}\right )\) be an \(m \times n\) matrix.
    Let \(R_{1}, R_{2}, \ldots \ldots \ldots , R_{m}\) denote the rows of \(A\).
    Hence \(R_{i}=\left (a_{i 1}, a_{i 2}, \ldots \ldots \ldots , a_{i n}\right )\).
    Suppose the row rank of \(A\) is \(r\).
    Then the dimension of the row space is \(r\).
    Let \(v_{1}=\left (b_{11}, \ldots , b_{1 n}\right ), v_{2}=\left (b_{21}, \ldots , b_{2 n}\right ), \ldots \) \(\ldots , v_{r}=\left (b_{r 1}, \ldots \ldots \ldots , b_{r n}\right )\) be a basis for the row space of \(A\).
    Then each row is a linear combination of the vectors \(v_{1}, v_{2}, \ldots \ldots \ldots , v_{r}\). Let,

    \begin{align*} R_{1} & =k_{11} v_{1}+k_{12} v_{2}+\cdots \cdots +k_{1 r} v_{r}\\ R_{2} & =k_{21} v_{1}+k_{22} v_{2}+\cdots \cdots +k_{2 r} v_{r}\\ R_{m}=& k_{m 1} v_{1}+k_{m 2} v_{2}+\cdots \cdots +k_{m r} v_{r} && \text { where } k_{i j} \in F \end{align*} Equating the \(i^{\text {th }}\) component of each of the above equations, we get

    \begin{align*} a_{1 i}=& k_{11} b_{1 i}+k_{12} b_{2 i}+\cdots \cdots +k_{1 r} b_{r i} \\ a_{2 i}=& k_{21} b_{1 i}+k_{22} b_{2 i}+\cdots \cdots +k_{2 r} b_{r i} \\ \cdots & \cdots \cdots \cdots \cdots \\ a_{m i}=& k_{m 1} b_{1 i}+k_{m 2} b_{2 i}+ \cdots \cdots + k_{m r} b_{r i} \end{align*} Hence

    \[ \left (\begin {array}{c} a_{1 i} \\ \cdot \\ \cdot \\ a_{m i} \end {array}\right )=b_{1 i}\left (\begin {array}{c} k_{11} \\ \cdot \\ \cdot \\ k_{m 1} \end {array}\right )+b_{2 i}\left (\begin {array}{c} k_{12} \\ \cdot \\ \cdot \\ k_{m 2} \end {array}\right )+\cdots +b_{r i}\left (\begin {array}{c} k_{1 r} \\ \cdot \\ \cdot \\ k_{m r} \end {array}\right ) \]

    Thus each column of \(A\) is a linear combination of vectors.
    Hence the dimension of the column space \(\leq r\).
    \(\therefore \quad \) Column rank of \(A \leq r=\) row rank of \(A\).
    Similarly, row rank of \(A \leq \) column rank of \(A\).
    Hence the row rank and the column rank of \(A\) are equal.

    Definition 4.3.5. Definition. The rank of a matrix \(A\) is the common value of its row and column rank.

    Note 4.3.6. Since the row rank and the column rank of matrix are unaltered by elementary row and column operations, equivalent matrices have the same rank In particular if a matrix \(A\) is reduced to its canonical form \(\left (\begin {array}{ll}I_{r} & O \\ O & O\end {array}\right )\), then rank of \(A=r\).

    Thus to find the rank of a matrix \(A\), we reduce to the canonical form and find the number of non-zero entries in the diagonal.

    Note that in the canonical form of the matrix there exists an \(r \times r\) sub-matrix, namely, \(I_{r}\), whose determinant is not zero.

    Further every \((r+1) \times (r+1)\) sub-matrix contains a row of zeros and hence its determinant is zero.

    Also under any elementary row or column operation the value of a determinant is either unaltered or multiplied by a non-zero constant.

    Hence the matrix \(A\) is also such that

      (i) there exists an \(r \times r\) sub-matrix whose determinant is nonzero.

      (ii) the determinant of every \((r+1) \times (r+1)\) sub-matrix is zero.

    Hence one can also define the rank of a matrix A to be \(r\) if A satisfies (i) and (ii).

    Note 4.3.7. Any non-singular matrix of order \(n\) is equivalent to the identity matrix and hence its rank is \(n\).

    Note 4.3.8. The rank of a matrix is not altered on multiplication by non-singular matrices, since pre multiplication by a non-singular matrix is equivalent to applying elementary row operations and post-multiplication by a non- singular matrix is equivalent to applying elementary column operations.

Solved Problems

    Problem 4.3.9. Find the rank of the matrix \(A=\left (\begin {array}{llll} 4 & 2 & 1 & 3 \\ 6 & 3 & 4 & 7 \\ 2 & 1 & 0 & 7 \end {array}\right ) \)

Solution :

\begin{align*} A &=\begin{pNiceArray}{cccc} 4 & 2 & 1 & 3 \\ 6 & 3 & 4 & 7 \\ 2 & 1 & 0 & 7 \end {pNiceArray} \\ & \sim \begin{pNiceArray}[last-col]{cccc} 1 & 2 & 4 & 3 & C_{1} \leftrightarrow C_{3} \\ 4 & 3 & 6 & 7 \\ 0 & 1 & 2 & 7 \end {pNiceArray} \\ & \sim \begin{pNiceArray}[last-col]{cccc} 1 & 0 & 0 & 0 & C_{1} \rightarrow C_2 - 2 C_1 \\ 4 & -5 & -10 & -5 & C_{3} \rightarrow C_{3}-4 C_{1} \\ 0 & 1 & 2 & 7 & C_{4} \rightarrow C_{4}-3 C_{1} \end {pNiceArray} \\ & \sim \begin{pNiceArray}[last-col]{cccc} 1 & 0 & 0 & 0 \\ 0 & -5 & -10 & -5 & R_{2} \rightarrow R_{2}-4 R_{1}\\ 0 & 1 & 2 & 7 \end {pNiceArray} \\ & \sim \begin{pNiceArray}[last-col]{cccc} 1 & 0 & 0 & 0 & C_{3} \rightarrow C_{3}-2 C_{2} \\ 0 & -5 & 0 & 0 & C_4 \to C_4 - C_2 \\ 0 & 1 & 0 & 6 \end {pNiceArray} \\ & \sim \begin{pNiceArray}[last-col]{cccc} 1 & 0 & 0 & 0 \\ 0 & -5 & 0 & 0 \\ 0 & 0 & 0 & 6 & R_{3} \rightarrow R_{3}+\frac {1}{5} R_{2} \end {pNiceArray} \\ & \sim \begin{pNiceArray}[last-col]{cccc} 1 & 0 & 0 & 0 \\ 0 & -5 & 0 & 0 & C_{2} \leftrightarrow C_{3} \\ 0 & 0 & 6 & 0 \end {pNiceArray} \\ & \sim \begin{pNiceArray}[last-col]{cccc} 1 & 0 & 0 & 0 \\ 0 & 1 & 0 & 0 & R_{2} \rightarrow -\frac {1}{5} R_{2} \\ 0 & 0 & 1 & 0 & R_{3} \rightarrow \frac {1}{6} R_{3} \end {pNiceArray} \end{align*}

\[\text { Rank of } A=3.\]

    Problem 4.3.10. Find the rank of the matrix \(A=\left (\begin {array}{llll} 1 & 1 & 1 & 1 \\ 4 & 1 & 0 & 2 \\ 0 & 3 & 4 & 2 \end {array}\right )\) by examining the determinant minors.

Solution :

\begin{align*} \left |\begin{array}{lll} 1 & 1 & 1 \\ 4 & 1 & 0 \\ 0 & 3 & 4 \end {array}\right | & =0=\left |\begin{array}{lll} 1 & 1 & 1 \\ 1 & 0 & 2 \\ 3 & 4 & 2 \end {array}\right | \\ \left |\begin{array}{lll} 1 & 1 & 1 \\ 4 & 1 & 2 \\ 0 & 3 & 2 \end {array}\right |& =0=\left |\begin{array}{lll} 1 & 1 & 1 \\ 4 & 0 & 2 \\ 0 & 4 & 2 \end {array}\right | \end{align*} Every \(3 \times 3\) submatrix of \(A\) has determinant zero. Also,

\begin{align*} \left |\begin{array}{ll} 1 & 1 \\ 4 & 1 \end {array}\right | & =-3 \neq 0 \\ \text { Rank of } A& =2 \end{align*}

4.4 Simultaneous Linear Equations

In this section we shall apply the theory of matrices developed in the preceding sections to study the existence of simultaneous linear equations.

Matrix form of a set of linear equations

Consider a system of \(m\) linear equations in \(n\) unknowns \(x_{1}, x_{2}, \ldots , x_{n}\) given by

\begin{align*} a_{11}x_1 + a_{12}x_2 + \cdots + a_{1n}x_n & = b_1 \\ a_{21}x_1 + a_{22}x_2 + \cdots + a_{2n}x_n & = b_2 \\ \cdots \cdots \cdots \cdots & = b_1 \\ a_{m1}x_1 + a_{m2}x_2 + \cdots + a_{mn}x_n & = b_m \\ \end{align*} Using the concept of matrix multiplication and equality of matrices this system can be written as \(A X=B\) where,

\[ A = \begin {pmatrix} a_{11} & a_{12} & \cdots & a_{1n} \\ a_{21} & a_{22} & \cdots & a_{2n} \\ \vdots & \vdots & \ddots & \vdots \\ a_{m1} & a_{m2} & \cdots & a_{mn} \\ \end {pmatrix} , \quad X = \begin {pmatrix} x_1 \\ x_2 \\ \vdots \\ x_n \end {pmatrix} , \quad B= \begin {pmatrix} b_1 \\ b_2 \\ \vdots \\ b_m \end {pmatrix} \]

Then \(m \times n \) matrix \(A\) is called the coefficient matrix.

    Definition 4.4.1. A set of values of \(x_1, x_2, \ldots , x_n\) which satisfy the above system of equations is called a solution of the system. The system of equations is said to be consistent if it has at least one solution. Otherwise the system is said to be inconsistent.

    The \(m \times (n+1)\) matrix given by

    \[A = \begin {pmatrix} a_{11} & a_{12} & \cdots & a_{1n} & b_1 \\ a_{21} & a_{22} & \cdots & a_{2n} & b_2 \\ \vdots & \vdots & \ddots & \vdots & \vdots \\ a_{m1} & a_{m2} & \cdots & a_{mn}& b_m \end {pmatrix}\]

Thus the augmented matrix \((A,B)\) is obtained by annexing to \(A \) the column matrix \(B\), which becomes the \((n+1)^{th}\) column in \((A,B)\).

    Note 4.4.2. Since every column in \(A\) appears in \((A, B)\) the column space of the matrix \(A\) is a subspace of the column space of the matrix \((A, B)\). Hence the rank of \(A \leq \) rank of \((A, B)\).

    Theorem 4.4.3. The system of linear equations \(A X=B \) is consistent if and only if rank of \(A=\) rank of \((A, B)\).

    Proof : Let the system be consistent.
    Let \(u_{1}, u_{2}, \ldots \ldots , u_{n}\) be a solution of the system.
    Then \(B=u_{1} C_{1}+u_{2} C_{2}+\cdots \cdots +u_{n} C_{n}\) where \(C_{1}, C_{2}, \cdots , C_{n}\) denote the columns of \(A\).
    Hence the column space of the augmented matrix \((A, B)\), namely \(\left \langle C_{1}, C_{2}, \ldots \ldots , C_{n}, B\right \rangle \) is the same as the column space \(\left \langle C_{1}, C_{2}, \ldots \ldots , C_{n}\right \rangle \) of \(A\).
    Hence the rank of \(A=\operatorname {rank}\) of \((A, B)\).
    Conversely let rank of \(A=\operatorname {rank}\) of \((A, B)\).
    Then the column rank of \(A=\) column rank of \((A, B)\).

    \[\operatorname {dim}\left \langle C_{1}, C_{2}, \ldots , C_{n}\right \rangle =\operatorname {dim}\left \langle C_{1}, C_{2}, \ldots , C_{n}, B\right \rangle . \]

    But \(\left \langle C_{1}, C_{2}, \ldots \ldots , C_{n}\right \rangle \) is a subspace of \(\left \langle C_{1}, C_{2}, \ldots \ldots , C_{n}, B\right \rangle \).
    \(B\) is a linear combination of \(C_{1}, C_{2}, \ldots \ldots , C_{n}\).
    If \(B=u_{1} C_{1}+\ldots +u_{n} C_{n}\) then \(u_{1}, u_{2}, \ldots \ldots , u_{n}\) is a solution of the system. This completes the proof.

Solved Problems

    Problem 4.4.4. Show that the equations \(x+y+z=6 , x+2 y+3 z=14 , \ x+4 y+7 z=30 \) are consistent and solve them.

Solution : The given system of equations can be put in the matrix form

\begin{align*} A X & =\left (\begin{array}{lll} 1 & 1 & 1 \\ 1 & 2 & 3 \\ 1 & 4 & 7 \end {array}\right )\left (\begin{array}{l} x \\ y \\ z \end {array}\right )=\left (\begin{array}{c} 6 \\ 14 \\ 30 \end {array}\right )=B \end{align*} The augmented matrix is given by

\begin{align*} (A, B) &=\left (\begin{array}{cccc} 1 & 1 & 1 & 6 \\ 1 & 2 & 3 & 14 \\ 1 & 4 & 7 & 30 \end {array}\right ) \\ & \sim \begin{pNiceArray}[last-col]{cccc} 1 & 1 & 1 & 6 \\ 0 & 1 & 2 & 8 & R_{2} \rightarrow R_{2}-R_{1} \\ 0 & 3 & 6 & 24 & R_{3} \rightarrow R_{3}-R_{1} \end {pNiceArray}\\ & \sim \begin{pNiceArray}[last-col]{cccc} 1 & 1 & 1 & 6 \\ 0 & 1 & 2 & 8 \\ 0 & 0 & 0 & 0 & R_{3} \rightarrow R_{3}-3 R_{2} \end {pNiceArray} \end{align*} Hence rank of \(A=\operatorname {rank}\) of \((A, B)=2\).
Hence the given system is consistent.
Also the given system of equations reduces to

\begin{align*} \left (\begin{array}{lll} 1 & 1 & 1 \\ 0 & 1 & 2 \\ 0 & 0 & 0 \end {array}\right ) \left (\begin{array}{l} x \\ y \\ z \end {array}\right ) & =\left (\begin{array}{l} 6 \\ 8 \\ 0 \end {array}\right ) \\ \therefore \quad x+y+z & =6 \\ y+2 z & =8 \end{align*} Putting \(z=c\) we obtain the general solution of the system as \(x=c-2, y=8-2 c, z=c\).

    Problem 4.4.5. Verify whether the following system of equations is consistent. If it is consistent, find the solution. \(x-4 y-3 z =-16 , \ 4 x-y+6 z =16, 2 x+7 y+12 z =48 , \ 5 x-5 y+3 z =0\)

Solution : The matrix form of the system is given by

\begin{align*} \left (\begin{array}{rrr} 1 & -4 & -3 \\ 4 & -1 & 6 \\ 2 & 7 & 12 \\ 5 & -5 & 3 \end {array}\right )\left (\begin{array}{l} x \\ y \\ z \end {array}\right ) & =\left (\begin{array}{r} -16 \\ 16 \\ 48 \\ 0 \end {array}\right ) \end{align*} The augmented matrix is given by

\begin{align*} (A, B)& =\left (\begin{array}{rrrr} 1 & -4 & -3 & -16 \\ 4 & -1 & 6 & 16 \\ 2 & 7 & 12 & 48 \\ 5 & -5 & 3 & 0 \end {array}\right )\\ & \sim \begin{pNiceArray}[last-col]{cccc} 1 & -4 & -3 & -16 \\ 0 & 15 & 18 & 80 & R_{2} \rightarrow R_{2}-4 R_{1} \\ 0 & 15 & 18 & 80 & R_{3} \rightarrow R_{3}-2 R_{1} \\ 0 & 15 & 18 & 80 & R_{4} \rightarrow R_{4}-5 R_{1} \end {pNiceArray} \\ & \sim \begin{pNiceArray}[last-col]{cccc} 1 & -4 & -3 & -16 \\ 0 & 15 & 18 & 80 \\ 0 & 0 & 0 & 0 & R_{3} \rightarrow R_{3}-R_{2} \\ 0 & 0 & 0 & 0 & R_{4} \rightarrow R_{4}-R_{2} \end {pNiceArray} \end{align*} Rank of \(A=\operatorname {Rank}\) of \((A, B)=2\) and hence the system is consistent. Also the system of equations reduces to

\[ \left (\begin {array}{ccc} 1 & -4 & -3 \\ 0 & 15 & 18 \\ 0 & 0 & 0 \\ 0 & 0 & 0 \end {array}\right )\left (\begin {array}{l} x \\ y \\ z \end {array}\right )=\left (\begin {array}{r} -16 \\ 80 \\ 0 \\ 0 \end {array}\right )\]

\(x-4 y-3 z=-16 \) and \(15 y+18 z=80 .\)
Putting \(z=c\) we obtain the general solution of the systems as

\[x=-(9 c / 5)+(16 / 3),\ y=-(6 c / 5)+(16 / 3) , z=c .\]

    Problem 4.4.6. For what values of \(\eta \) the equations \(x+y+z =1, \ x+2 y+4 z =\eta , \ x+4 y+10 z =\eta ^{2}\) are consistent?

Solution : The matrix form of the system is given by

\begin{align*} \left (\begin{array}{ccc} 1 & 1 & 1 \\ 1 & 2 & 4 \\ 1 & 4 & 10 \end {array}\right )\left (\begin{array}{l} x \\ y \\ z \end {array}\right ) & =\left (\begin{array}{c} 1 \\ \eta \\ \eta ^{2} \end {array}\right ) \end{align*} The augmented matrix is given by

\begin{align*} (A, B) &=\left (\begin{array}{cccc} 1 & 1 & 1 & 1 \\ 1 & 2 & 4 & \eta \\ 1 & 4 & 10 & \eta ^{2} \end {array}\right ) \\ & \sim \begin{pNiceArray}[last-col]{cccc} 1 & 1 & 1 & 1 \\ 0 & 1 & 3 & \eta -1 & R_{2} \rightarrow R_{2}-R_{1} \\ 0 & 3 & 9 & \eta ^{2}-1 & R_3 \to R_3 - R_1 \end {pNiceArray}\\ & \sim \begin{pNiceArray}[last-col]{cccc} 1 & 1 & 1 & 1 \\ 0 & 1 & 3 & \eta -1 \\ 0 & 0 & 0 & \eta ^{2}-3 \eta +2 & R_3 \to R_3 - 3 R_2 \end {pNiceArray} \end{align*} The given system is consistent if and only if \(\eta ^{2}-3 \eta +2=0\) \(\therefore \quad \eta =2\) or \(1\).

    Problem 4.4.7. Show that the system of equations \(x+2 y+z =11\), \(4 x+6 y+5 z =8 \), \(2 x+2 y+3 z =19\) is inconsistent.

Solution : The matrix form of the system is given by

\[ \left (\begin {array}{lll} 1 & 2 & 1 \\ 4 & 6 & 5 \\ 2 & 2 & 3 \end {array}\right )\left (\begin {array}{l} x \\ y \\ z \end {array}\right )=\left (\begin {array}{c} 11 \\ 8 \\ 19 \end {array}\right ) \]

\(\therefore \) The augmented matrix is given by

\begin{align*} (A, B) & =\left (\begin{array}{cccc} 1 & 2 & 1 & 11 \\ 4 & 6 & 5 & 8 \\ 2 & 2 & 3 & 19 \end {array}\right ) \\ & \sim \begin{pNiceArray}[last-col]{cccc} 1 & 2 & 1 & 11 \\ 0 & -2 & 1 & -36 &R_{2} \rightarrow R_{2}-4 R_{1} \\ 0 & -2 & 1 & -3&R_{3} \rightarrow R_{3}-2 R_{1} \end {pNiceArray}\\ & \sim \begin{pNiceArray}[last-col]{cccc} 1 & 2 & 1 & 11 \\ 0 & -2 & 1 & -36 \\ 0 & 0 & 0 & 33 & R_{3} \rightarrow -R_{3}-R_{2} \end {pNiceArray} \end{align*} Rank of \(A=2 \) and rank of \((A, B)=3 \). The given system is inconsistent.

4.5 Characteristic Equation And Cayley Hamilton Theorem

    Definition 4.5.1. An expression of the form \(A_{0}+A_{1} x+A_{2} x^{2}+\cdots +A_{n} x^{n}\) where \(A_{0}, A_{1}, \ldots , A_{n}\) are square matrices of the same order and \(A_{n} \neq 0\) is called a matrix polynomial of degree \(n\).

    Example 4.5.2. If \(\left (\begin {array}{ll} 1 & 2 \\ 0 & 3 \end {array}\right )+\left (\begin {array}{ll} 1 & 1 \\ 2 & 1 \end {array}\right ) x+ \left (\begin {array}{ll} 2 & 0 \\ 3 & 1 \end {array}\right ) x^{2}\) is a matrix polynomial of degree 2 and it is simply the matrix \(\left (\begin {array}{cc}1+x+2 x^{2} & 2+x \\ 2 x+3 x^{2} & 3+x+x^{2}\end {array}\right )\).

    Definition 4.5.3. Let \(A\) be any square matrix of order \(n\) and let \(I\) be the identity matrix of order \(n\). Then the matrix polynomial given by \(A-x I\) is called the characteristic matrix of \(A\).

    The determinant \(|A-x I|\) which is an ordinary polynomial in \(x\) of degree \(n\) is called the characteristic polynomial of \(A\).

    The equation \(|A-x I|=0\) is called the characteristic equation of \(A\).

    Example 4.5.4. Find the characteristic equation of the matrix \(A=\left (\begin {array}{ll}1 & 2 \\ 3 & 4\end {array}\right )\).

Solution : Then the characteristic matrix of \(A\) is \(A-x l\) given by

\begin{align*} A-x l &=\left (\begin{array}{ll} 1 & 2 \\ 3 & 4 \end {array}\right )-x\left (\begin{array}{ll} 1 & 0 \\ 0 & 1 \end {array}\right ) \\ &=\left (\begin{array}{cc} 1-x & 2 \\ 3 & 4-x \end {array}\right ) \end{align*} The characteristic polynomial of \(A\) is

\begin{align*} |A-xI | & = \begin{vmatrix} 1-x & 2 \\ 3 & 4-x \end {vmatrix}\\ &=(1-x)(4-x)-6 \\ &=x^{2}-5 x-2 \end{align*} \(\therefore \) The characteristic equation of \(A\) is \(|A-x I|=0\). \(\therefore x^{2}-5 x-2=0\) is the characteristic equation of \(A\),

    Example 4.5.5. Find the characteristic equation of the matrix \(A=\left (\begin {array}{lll}1 & 0 & 2 \\ 0 & 1 & 2 \\ 1 & 2 & 0\end {array}\right )\).

Solution : The characteristic matrix of \(A\) is \(A-x I\) given by

\[ A-x I=\left (\begin {array}{ccc} 1-x & 0 & 2 \\ 0 & 1-x & 2 \\ 1 & 2 & -x \end {array}\right ) \]

The characteristic polynomial of \(A\) is

\begin{align*} |A-x I| &=\left |\begin{array}{ccc} 1-x & 0 & 2 \\ 0 & 1-x & 2 \\ 1 & 2 & -x \end {array}\right | \\ &=(1-x)[(1-x)(-x)-4]-2(1-x)] \\ &=-x(1-x)^{2}-4(1-x)-2+2 x \\ &=-x^{3}+2 x^{2}-x-4+4 x-2+2 x \\ &=-x^{3}+2 x^{2}+5 x-6 \end{align*} The characteristic equation of \(A\) is

\begin{align*} -x^{3}+2 x^{2}+5 x-6&=0 \\ \text { (i.e) } x^{3}-2 x^{2}-5 x+6&=0 \end{align*}

    Theorem 4.5.6 (Cayley Hamilton Theorem)). Any square matrix \(A\) satisfies its characteristic equation. (i.e.) if \(a_{0}+a_{1} x+a_{2} x^{2}+\cdots +a_{n} x^{n}\) is the characteristic polynomial of degree \(n\) of \(A\) then

    \[ a_{0} I+a_{1} A+a_{2} A^{2}+\cdots +a_{n} A^{n}=0 \]

    Proof : Let \(A\) be a square matrix of order \(n\).

    \begin{equation} \label {p726eq1} \text { Let }|A-x l|=a_{0}+a_{1} x+a_{2} x^{2}+\cdots +a_{n} x^{n} \end{equation}

    be the characteristic polynomial of \(A\).

    Now, \(\operatorname {adj}(A-x I)\) is a matrix polynomial of degree \(n-1\) since each entry of the matrix \(adj(A-x I)\) is a cofactor of \(A-x I\) and hence is a polynomial of degree \(\leq n-1\).

    \begin{equation} \text { Let } \operatorname {adj}(A-x I)=B_{0}+B_{1} x+B_{2} x^{2}+\cdots +B_{n-1} x^{n-1} \label {p726eq2} \end{equation}

    Now,

    \begin{align*} (A-x I) \operatorname {adj}(A-x I)&=|A-x I| I . \quad (\because \quad (\operatorname {adj} A) A=A(\operatorname {adj} A)=|A| I) \\ \therefore \quad (A-x I)\left (B_{0}+B_{1} x+\cdots +B_{n-1} x^{n-1}\right ) &=\left (a_{0}+a_{1} x+\cdots +a_{n} x^{n}\right ) I \text { using } \eqref {p726eq1} \text { and }\eqref {p726eq2} . \end{align*} Equating the coefficients of the corresponding powers of \(x\) we get

    \begin{align*} A B_{0}=& a_{0} I \\ A B_{1}-B_{0}=& a_{1} I \\ A B_{2}-B_{1}=& a_{2} I \\ \ldots & \ldots . . . . \\ \ldots & \ldots . \\ AB_{n-1}-B_{n-2}=& a_{n-1} I \\ -B_{n-1}=& a_{n} I \end{align*} Pre-multiplying the above equations by \(I, A, A^{2}, \ldots \) \(A^{n}\) respectively and adding we get

    \[ a_{0} I+a_{1} A+a_{2} A^{2}+\cdots +a_{n} A^{n}=0 \]

    Note 4.5.7. The inverse of a non-singular matrix can be calculated by using the Cayley Hamilton theorem as follows.

    Let \(a_{0}+a_{1} x+a_{2} x^{2}+\cdots +a_{n} x^{n}\) be the characteristic polynomial of \(A\)

    Then by Theorem 4.5.6, we have

    \begin{equation} a_{0} I+a_{1} A+a_{2} A^{2}+\cdots +a_{n} A^{n}=0 \label {p726eq3} \end{equation}

    Since \(|A-xI|=a_{0}+a_{1} x+a_{2} x^{2}+\cdots +a_{n} x^{n}\) we get \(a_{0}=|A|\) (by putting \(x=0\) ).

    \begin{align*} \therefore \quad a_{0} & \neq 0 (\because A \text { is a non singular matrix. }) \\ \therefore \quad I & =-\frac {1}{a_{0}}\left [a_{1} A+a_{2} A^{2}+\cdots +a_{n} A^{n}\right ] \text { by \eqref {p726eq3} }\\ {A^{-1}&=-\frac {1}{a_{0}}\left [a_{1} I+a_{2} A+\cdots +a_{n} A^{n-1}\right ] \end{align*}

Solved Problems

    Problem 4.5.8. Find the characteristic equation of the matrix \(A=\begin {pmatrix} 8 & -6 & 2 \\ -6 &7 & -4 \\ 2 & -4 & 3 \end {pmatrix}\).

Solution : The characteristic equation of the matrix \(A\) is given by \(|A- \lambda I|=0\).

\begin{align*} \begin{vmatrix}8-\lambda &-6&2\\ -6&7-\lambda &-4\\ 2&-4&3-\lambda \end {vmatrix} &=0 \\ \left (8-\lambda \right ) \begin{vmatrix}7-\lambda &-4\\ -4&3-\lambda \end {vmatrix}-\left (-6\right ) \begin{vmatrix}-6&-4\\ 2&3-\lambda \end {vmatrix}+2\cdot \begin{vmatrix}-6&7-\lambda \\ 2&-4\end {vmatrix} & =0 \\ \left (8-\lambda \right )[\left (7-\lambda \right )\left (3-\lambda \right )-\left (-4\right )\left (-4\right )] -\left (-6\right )[\left (-6\right )\left (3-\lambda \right )-\left (-4\right )\cdot 2] & \\ \quad +2[ \left (-6\right )\left (-4\right )-\left (7-\lambda \right )\cdot 2] &= 0 \\ \left (8-\lambda \right )\left (\lambda ^2-10\lambda +5\right )-\left (-6\right )\left (6\lambda -10\right )+2\left (2\lambda +10\right ) &= 0 \\ -\lambda ^3+18\lambda ^2-85\lambda +40+36\lambda -60+4\lambda +20 &= 0 \\ -\lambda ^3+18\lambda ^2-45\lambda &= 0 \\ \lambda ^3-18\lambda ^2+45\lambda &= 0 \end{align*} ie., \(\lambda ^3-18\lambda ^2+45\lambda = 0\), which represents the characteristic equation of \(A\).

    Problem 4.5.9. Show that the non-singular matrix \(A = \begin {pmatrix} 1& 2 \\ 3 & 1 \end {pmatrix} \) satisfies the equation \(A^2 - 2A -5I=0\). Hence evaluate \(A^{-1}\).

Solution : The characteristic polynomial of \(A\) is

\begin{align*} |A- \lambda I | & = \begin{vmatrix} 1-x & 2 \\ 3 & 1-x \end {vmatrix} \\ & =\left (1-\lambda \right )\left (1-\lambda \right )-2\cdot 3 \\ &=\lambda ^2-2\lambda -5 \end{align*} By Cayley-Hamiltonian theorem

\begin{align*} A^2 -2A -5I&=0\\ I & = \dfrac {1}{5} (A^2 -2A) \\ A^{-1} & = \dfrac {1}{5} (A -2I)\\ & = \dfrac {1}{5} \left [\begin{pmatrix} 1& 2 \\ 3 & 1 \end {pmatrix} - 2 \begin{pmatrix} 1& 0 \\ 0 & 1 \end {pmatrix} \right ] \\ A^{-1}& = \dfrac {1}{5} \begin{pmatrix} -1& 2 \\ 3 & -1 \end {pmatrix} \end{align*}

    Problem 4.5.10. Show that the matrix \(A= \begin {pmatrix} 2 & -3 & 1 \\ 3 & 1 & 3 \\ -5 & 2 & -4 \end {pmatrix} \) satisfies the equation \(A(A-1) (A+2I)=0\).

Solution : The characteristic polynomial of \(A\) is

\begin{align*} |A-\lambda I|&=0\\ \begin{vmatrix}2-\lambda &-3&1\\ 3&1-\lambda &3\\ -5&2&-4-\lambda \end {vmatrix}&=0 \\ \left (2-\lambda \right ) \begin{vmatrix}1-\lambda &3\\ 2&-4-\lambda \end {vmatrix}-\left (-3\right ) \begin{vmatrix}3&3\\ -5&-4-\lambda \end {vmatrix}+1\cdot \begin{vmatrix}3&1-\lambda \\ -5&2\end {vmatrix}&=0 \\ \left (2-\lambda \right )\left (\lambda ^2+3\lambda -10\right )-\left (-3\right )\left (-3\lambda +3\right )+1\cdot \left (-5\lambda +11\right )&=0\\ -\lambda ^3-\lambda ^2+16\lambda -20-9\lambda +9-5\lambda +11&=0 \\ -\lambda ^3-\lambda ^2+2\lambda &=0 \end{align*} By Cayley-Hamilton Theorem

\begin{align*} -A^3 -A^2 +2A&=0\\ A^3 +A^2 -2A&=0\\ A(A^2 +A-2I)&=0\\ A(A+2I)(A-I)&=0 \end{align*}

    Problem 4.5.11. Using Cayley-Hamilton theorem find the inverse of the matrix \(A= \begin {pmatrix} 7 & 2 & -2 \\ -6 & -1 & 2 \\ 6 & 2 & -1 \end {pmatrix}\).

Solution : Let \(A =\left [\begin {array}{rrr} 7 & 2 & -2 \\ -6 & -1 & 2 \\ 6 & 2 & -1 \end {array}\right ] \).

The characteristic polynomial of \(A=|A-xI|\)

\begin{align*} |A-xI|=&\left |\begin{array}{ccc} 7-x & 2 & -2 \\ -6 & -1-x & 2 \\ 6 & 2 & -1-x \end {array}\right | \\ &=(7-x)\left [(1+x)^{2}-4\right ]-2[6(1+x)-12] -2[-12+6(1+x)]\\ &=(7-x)\left (x^{2}+2 x-3\right )-12(x-1)-12(x-1)\\ &= 7 x^{2}+14 x-21-x^{3}-2 x^{2}+3 x-12 x+12 -12 x+12 \\ &=-x^{3}+5 x^{2}-7 x+3 \end{align*} By Cayley-Hamilton theorem

\begin{align*} -A^{3}+5 A^{2}-7 A+3 I_{3}&={0} \\ A^{3}-5 A^{2}+7 A-3 I_{3}&=0 \\ 3 I_{3}&=A^{3}-5 A^{2}+7 A \\ I_{3}&=\frac {1}{3}\left (A^{3}-5 A^{2}+7 A\right ) \end{align*} Pre (or post) multiplying by \(A^{-1}\) on both sides we get

\begin{equation} \label {p728eq1} A^{-1}=\frac {1}{3}\left [A^{2}-5 A+7 I_{3}\right ] \end{equation}

Now,

\begin{align*} A^{2}&=\left (\begin{array}{rrr} 7 & 2 & -2 \\ -6 & -1 & 2 \\ 6 & 2 & -1 \end {array}\right )\left (\begin{array}{rrr} 7 & 2 & -2 \\ -6 & -1 & 2 \\ 6 & 2 & -1 \end {array}\right ) \\ & =\begin{pmatrix}7\cdot \:7+2\left (-6\right )+\left (-2\right )\cdot \:6&7\cdot \:2+2\left (-1\right )+\left (-2\right )\cdot \:2&7\left (-2\right )+2\cdot \:2+\left (-2\right )\left (-1\right )\\ \left (-6\right )\cdot \:7+\left (-1\right )\left (-6\right )+2\cdot \:6&\left (-6\right )\cdot \:2+\left (-1\right )\left (-1\right )+2\cdot \:2&\left (-6\right )\left (-2\right )+\left (-1\right )\cdot \:2+2\left (-1\right )\\ 6\cdot \:7+2\left (-6\right )+\left (-1\right )\cdot \:6&6\cdot \:2+2\left (-1\right )+\left (-1\right )\cdot \:2&6\left (-2\right )+2\cdot \:2+\left (-1\right )\left (-1\right )\end {pmatrix} \\ &=\left (\begin{array}{rrr} 25 & 8 & -8 \\ -24 & -7 & 8 \\ 24 & 8 & -7 \end {array}\right ) \end{align*} From (4.4)

\begin{align*} A^{-1}&=\frac {1}{3}\left [\left (\begin{array}{rrr} 25 & 8 & -8 \\ -24 & -7 & 8 \\ 24 & 8 & -7 \end {array}\right ) -\left [\begin{array}{rrr} 35 & 10 & -10 \\ -30 & -5 & 10 \\ 30 & 10 & -5 \end {array}\right ] +\left [\begin{array}{rrr} 7 & 0 & 0 \\ 0 & 7 & 0 \\ 0 & 0 & 7 \end {array}\right ]\right ] \\ A^{-1}&=\frac {1}{3}\left (\begin{array}{rrr} -3 & -2 & 2 \\ 6 & 5 & -2 \\ -6 & -2 & 5 \end {array}\right ) \end{align*}

    Problem 4.5.12. Find the inverse of the matrix \(\begin {pmatrix} 3 & 3 & 4 \\ 2 & -3 & 4 \\ 0 & -1 & 1 \end {pmatrix}\) using Cayley-Hamilton theorem.

Solution : The characteristic polynomial of \(A\) is

\begin{align*} |A-x I| &=\begin{vmatrix} 3-x & 3 & 4 \\ 2 & -3-x & 4 \\ 0 & -1 & 1-x \end {vmatrix}\\ &= \left (3-x\right )\left (x^2+2x+1\right )-3\cdot \:2\left (-x+1\right )+4\left (-2\right )\\ &= -x^3+x^2+5x+6x+3-6-8\\ &=-x^{3}+x^{2}+11 x-11 \end{align*} By Cayley-Hamilton theorem

\begin{align*} A^{3}+A^{2}+11 A-11 I_{3}&=0 . \\ \therefore \quad A^{3}-A^{2}-11 A+11 I_{3}&=0 . \\ 11 I_{3}&=-\left (A^{3}-A^{2}-11 A\right ) . \\ \therefore \quad I_{3}&=-\frac {1}{11}\left [A^{3}-A^{2}-11 A\right ] . \end{align*} Pre (post) multiplying by \(A^{-1}\) on both sides we get

\begin{align*} A^{-1} & =-\frac {1}{11}\left [A^{2}-A-11 I_{3}\right ] \\ & =-\frac {1}{11} \left [ \begin{pmatrix} 15 & -4 & 28 \\ 0 & 11 & 0 \\ -2 & 2 & -3 \end {pmatrix} - \begin{pmatrix} 3 & 3 & 4 \\ 2 & -3 & 4 \\ 0 & -1 & 1 \end {pmatrix} -11 \begin{pmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 1 \end {pmatrix} \right ] \\ &=\begin{pmatrix} -\frac {1}{11} & \frac {7}{11} & -\frac {24}{11} \\ \frac {2}{11} & -\frac {3}{11} & \frac {4}{11} \\ \frac {2}{11} & -\frac {3}{11} & \frac {15}{11} \end {pmatrix} \end{align*}

    Problem 4.5.13. Verify Cayley Hamilton’s theorem for the matrix \(A =\begin {pmatrix} 1&2 \\ 4 & 3 \end {pmatrix} \).

Solution : The characteristic equation of \(A\) is

\begin{align*} |A-\lambda l|&=0 . \\ \therefore \left |\begin{array}{cc} 1-\lambda & 2 \\ 4 & 3-\lambda \end {array}\right |&=0 \\ \therefore \quad (1-\lambda )(3-\lambda )-8 &=0\\ \lambda ^{2}-4 \lambda -5 &=0 \end{align*} By Cayley Hamilton’s theorem \(A\) satisfies its characteristic equation. We have

\begin{align*} A^2 - 4A - 5I &=0 \\ A^2 &= \begin{pmatrix} 1&2 \\ 4 & 3 \end {pmatrix}\begin{pmatrix} 1&2 \\ 4 & 3 \end {pmatrix} = \begin{pmatrix} 9&8\\16&17 \end {pmatrix}\\ 4A&= 4 \begin{pmatrix} 1&2 \\ 4 & 3 \end {pmatrix} = \begin{pmatrix} 4 &8 \\ 16& 12 \end {pmatrix} \\ 5I & = 5 \begin{pmatrix} 1&0 \\0&1 \end {pmatrix} =\begin{pmatrix} 5&0 \\0&5 \end {pmatrix}\\ A^2 - 4A - 5I & = \begin{pmatrix} 9&8\\16&17 \end {pmatrix} - \begin{pmatrix} 4 &8 \\ 16& 12 \end {pmatrix} - \begin{pmatrix} 5&0 \\0&5 \end {pmatrix} \\ & = \begin{pmatrix} 0&0 \\0 &0 \end {pmatrix} =0 \end{align*} Thus Cayley-Hamilton theorem is verified.

    Problem 4.5.14. Using Cayley Hamilton’s theorem for the matrix \(A= \begin {pmatrix} 1 & 0 & -2 \\ 2 & 2 & 4 \\ 0 & 0 & 2 \end {pmatrix} \). Find (i) \(A^{-1}\) and (ii) \(A^4\).

Solution : The characteristic equation of \(A\) is

\begin{align*} |A-\lambda I | &=0 \\ \begin{vmatrix} 1 -\lambda & 0 & -2 \\ 2 & 2 -\lambda & 4 \\ 0 & 0 & 2 - \lambda \end {vmatrix}&=0\\ (2 -\lambda ) (2 - \lambda ) (1-\lambda )&=0 \\ \lambda ^{3}- 5 \lambda ^{2}+8 \lambda -4 &=0 \end{align*} By Cayley-Hamilton’s theorem

\begin{align} A^{3}-5 A^{2}+8 A-4 I&=0 \label {p729eq1} \\ 4I &= A^3 -5A^2 +8A \notag \end{align} (i) To find \(A^{-1}\) pre multiplying by \(A^{-1}\) we get

\begin{align} 4 A^{-1} &=A^{-1} A^{3}-5 A^{-1} A^{2}+8 A^{-1} A\notag \\ &=A^{2}-5 A+8 I \notag \\ \therefore A^{-1} &=\frac {1}{4}\left [A^{2}-5 A+8 I\right ] \label {p729eq2} \end{align} Now,

\begin{align*} A^{2} &=\left (\begin{array}{rrr} 1 & 0 & -2 \\ 2 & 2 & 4 \\ 0 & 0 & 2 \end {array}\right )\left (\begin{array}{rrr} 1 & 0 & -2 \\ 2 & 2 & 4 \\ 0 & 0 & 2 \end {array}\right ) \\ &=\left (\begin{array}{rrr} 1 & 0 & -6 \\ 6 & 4 & 12 \\ 0 & 0 & 4 \end {array}\right ) \end{align*} From (4.6)

\begin{align*} A^{-1}&= \frac {1}{4}\left [\left (\begin{array}{rrr} 1 & 0 & -6 \\ 6 & 4 & 12 \\ 0 & 0 & 4 \end {array}\right ) -\left (\begin{array}{rrr} 5 & 0 & -10 \\ 10 & 10 & 20 \\ 0 & 0 & 10 \end {array}\right )+\left (\begin{array}{lll} 8 & 0 & 0 \\ 0 & 8 & 0 \\ 0 & 0 & 8 \end {array}\right )\right ]\\ &=\frac {1}{4}\left (\begin{array}{rrr} 4 & 0 & 4 \\ 4 & 2 & -8 \\ 0 & 0 & 2 \end {array}\right ) \\ A^{-1}&=\left (\begin{array}{rrr} 1 & 0 & 1 \\ 1 & \frac {1}{2} & -2 \\ 0 & 0 & -\frac {1}{2} \end {array}\right ) \end{align*} (ii) To find \(A^{4}\). From (4.5)

\begin{align*} A^{3} &=5 A^{2}-8 A+4 I \\ \therefore A^{4}&=5 A^{3}-8 A^{2}+4 A \\ &=5\left [5 A^{2}-8 A+4 I\right ]-8 A^{2}+4 A \text { by } \eqref {p729eq1} \\ &=17 A^{2}-36 A+20 I\\ & = 17 \left (\begin{array}{rrr} 1 & 0 & -6 \\ 6 & 4 & 12 \\ 0 & 0 & 4 \end {array}\right ) - 36 \begin{pmatrix} 1 & 0 & -2 \\ 2 & 2 & 4 \\ 0 & 0 & 2 \end {pmatrix} +20 \begin{pmatrix} 1 & 0 & 0 \\0& 1 & 0 \\0& 0 & 1 \end {pmatrix} \\ &=\left (\begin{array}{rrr} 17 & 0 & -102 \\ 102 & 68 & 204 \\ 0 & 0 & 68 \end {array}\right ) - \left (\begin{array}{rrr} 36 & 0 & -72 \\ 72 & 72 & 144 \\ 0 & 0 & 72 \end {array}\right ) +\left (\begin{array}{rrr} 20 & 0 & 0 \\ 0 & 20 & 0 \\ 0 & 0 & 20 \end {array}\right ) \\ A^{4}&=\left (\begin{array}{rrr} 1 & 0 & -30 \\ 30 & 16 & 60 \\ 0 & 0 & 16 \end {array}\right ) \end{align*}

4.6 Eigen Values And Eigen Vectors

    Definition 4.6.1. Let \(A\) be an \(n \times n\) matrix. A number \(\lambda \) is called an eigen value of \(A\) if there exists a nonzero vector \(X=\left (\begin {array}{c}x_{1} \\ x_{2} \\ \vdots \\ x_{n}\end {array}\right )\) such that \(A X=\lambda X\) and \(X\) is called an eigen vector corresponding to the eigen value \(\lambda \).

    Remark 4.6.2. If \(X\) is an eigen vector corresponding to the eigen value \(\lambda \) of \(A\), then \(\alpha X\) where \(\alpha \) is any non-zero number, is also an eigen vector corresponding to \(\lambda \).

    Remark 4.6.3. Let \(X\) be an eigen vector corresponding to the eigen value \(\lambda \) of \(A\). Then \(A X=\lambda X\) so that \((A-\lambda I) X=0\). Thus \(X\) is a non-trivial solution of the system of homogeneous linear equations \((A-\lambda I) X=0\). Hence \(|A-\lambda I|=0\), which is the characteristic polynomial of \(A \) . Let \(|A-\lambda I|=a_{0} \lambda ^{n}+a_{1} \lambda ^{n-1}+\cdots +a_{n}\). The roots of this polynomial give the eigen values of \(A\). Hence eigen values are also called characteristic roots.

4.6.1 Properties of Eigen Values

    Property 4.6.4. Let \(X\) be an eigen vector corresponding to the eigen values \(\lambda _{1}\) and \(\lambda _{2}\). Then \(\lambda _{1}=\lambda _{2}\).

    Proof : By definition \(X \neq 0, A X=\lambda _{1} X\) and \(A X=\lambda _{2} X\).

    \begin{align*} \lambda _{1} X & =\lambda _{2} X \\ \left (\lambda _{1}-\lambda _{2}\right ) X&=0 \end{align*} Since \(X \neq 0, \quad \lambda _{1}=\lambda _{2}\).

    Property 4.6.5. Let \(A\) be a square matrix. Then
    (i) the sum of the eigen values of \(A\) is equal to the sum of the diagonal elements (trace) of \(A\).
    (ii) Product of eigen values of \(A\) is \(|A|\).

    Proof : (i) Let \(A=\left (\begin {array}{cccc}a_{11} & a_{12} & \cdots & a_{1 n} \\ a_{21} & a_{22} & \cdots & a_{2 n} \\ \vdots & \vdots & \vdots & \vdots \\ a_{n 1} & a_{n 2} & \cdots & a_{n n}\end {array}\right )\). The eigenvalues of \(A\) are the roots of the characteristic equation

    \begin{align} |A-\lambda I| & =\left |\begin{array}{cccc} a_{11}-\lambda & a_{12} & \cdots & a_{1 n} \\ a_{21} & a_{22}-\lambda & \cdots & a_{2 n} \\ \vdots & \vdots & \vdots & \vdots \\ a_{n 1} & a_{n 2} & \cdots & a_{n n}-\lambda \end {array}\right |=0 \label {p731eq1} \end{align} Let,

    \begin{align} |A-\lambda I|& =a_{0} \lambda ^{n}+a_{1} \lambda ^{n-1}+\cdots +a_{n} \ldots \label {p731eq2} \end{align} From (4.7) and (4.8) we get

    \begin{equation} a_{0}=(-1)^{n} ; a_{1}=(-1)^{n-1}\left (a_{11}+a_{22}+\cdots +a_{n n}\right ) \label {p731eq3} \end{equation}

    Also by putting \(\lambda =0\) in (4.8), we get \(a_{n}=|A|\). Now let \(\lambda _{1}, \lambda _{2}, \ldots , \lambda _{n}\) be the eigen values of \(A\). \(\therefore \lambda _{1}, \lambda _{2}, \ldots , \lambda _{n}\) are the roots of (4.8).

    \begin{align*} \therefore \quad \lambda _{1}+\lambda _{2}+\cdots +\lambda _{n} & =-\frac {a_{1}}{a_{0}} \\ &=a_{11}+a_{22}+\cdots +a_{n n}(\text { using } \eqref {p731eq3} ) \\ \text {Sum of the eigen values} & =\text { trace.of } A \end{align*} (ii) Product of the eigen values = product of the roots

    \begin{align*} \text {Product of the eigen values} &=\lambda _{1} \lambda _{2} \cdots \lambda _{n} \\ &=(-1)^{n} \frac {a_{n}}{a_{0}}=\frac {(-1)^{n} a_{n}}{(-1)^{n}} \\ &=a_{n} \\ &=|A| . \end{align*}

    Property 4.6.6. The eigen values of \(A\) and its transpose \(A^{T}\) are the same.

    Proof : It is enough if we prove that \(A\) and \(A^{T}\) have the same characteristic polynomial. Since for any square matrix \(M,|M|=\left |M^{T}\right |\) we have,

    \begin{align*} |A-\lambda I|=\left |(A-\lambda I)^{T}\right | &=\left |A^{T}-(\lambda I)^{T}\right | \\ &=\left |A^{T}-\lambda I\right | . \end{align*} Hence the result.

    Property 4.6.7. If \(\lambda \) is an eigen value of a non singular matrix \(A\) then \(\frac {1}{\lambda }\) is an eigen value of \(A^{-1}\).

    Proof : Let \(X\) be an eigen vector corresponding to \(\lambda \).
    Then \(A X=\lambda X .\) Since \(A\) is non singular \(A^{-1}\) exists.

    \begin{align*} \therefore \quad A^{-1}(A X) &=A^{-1}(\lambda X) \\ I X &=\lambda A^{-1} X\\ \therefore \quad A^{-1} X &=\left (\frac {1}{\lambda }\right ) X \end{align*} \(\therefore \quad \frac {1}{\lambda }\) is an eigen value of \(A^{-1}\)

    Corollary 4.6.8. If \(\lambda _{1}, \lambda _{2}, \ldots , \lambda _{n}\) are the eigen values of a non singular matrix \(A\) then \(\frac {1}{\lambda _{1}}, \frac {1}{\lambda _{2}}, \ldots , \frac {1}{\lambda _{n}} \) are the eigen values of \(A^{-1}\).

    Property 4.6.9. If \(\lambda \) is an eigen value of \(A\) then \(k \lambda \) is an eigen value of \(k A\) where \(k\) is a scalar.

    Proof : Let \(X\) be an eigen vector corresponding to \(\lambda \). Then

    \begin{equation} A X=\lambda X \label {prop5eq1} \end{equation}

    Now,

    \begin{align*} (k A) X= & k(A X)\\ &=k(\lambda X) \quad (\text { by } \eqref {prop5eq1}) \\ &=(k \lambda ) X . \end{align*} \(\therefore \quad k \lambda \) is an eigen value of \(k A\).

    Property 4.6.10. If \(\lambda \) is an eigen value of \(A\) then \(\lambda ^{k}\) is an eigen value of \(A^{k}\) where \(k\) is any positive integer.

    Proof : Let \(X\) be an eigen vector corresponding to \(\lambda \). Then

    \begin{equation} A X=\lambda X \label {prop6eq1} \end{equation}

    Now,

    \begin{align*} A^{2} X &=(A A) X=A(A X) \\ &=A(\lambda X) \quad (\text { by } \eqref {prop6eq1} ) \\ &=\lambda (A X) \\ &=\lambda (\lambda X) \quad \text { (by } \eqref {prop6eq1} ) \\ &=\lambda ^{2} X . \end{align*} \(\lambda ^{2} \) is an eigen value of \(A^{2}\).
    Proceeding like this we can prove that \(\lambda ^{k}\) is an eigen value of \(A^{k}\) for any positive integer.

    Corollary 4.6.11. If \(\lambda _{1}, \lambda _{2}, \ldots , \lambda _{n}\) are eigen values of \(A\) then \(\lambda _{1}^{k}, \lambda _{2}^{k}, \ldots , \lambda _{n}^{k}\) are eigen values of \(A^{k}\) for any positive integer \(k\).

    Property 4.6.12. Eigen vectors corresponding to distinct eigen values of a matrix are linearly independent.

    Proof : Let \(\lambda _{1}, \lambda _{2}, \ldots , \lambda _{k}\) be distinct eigen values of a matrix and let \(X_{i}\) be the eigen vector corresponding to \(\lambda _{i}\). Hence

    \begin{equation} \label {prop7eq1} A X_{i}=\lambda _{i} X_{i} \quad (i=1,2, \ldots , k) \end{equation}

    Now, suppose \(X_{1}, X_{2}, \ldots , X_{k}\) are linearly dependent. Then there exist real numbers \(\alpha _{1}, \alpha _{2}, \ldots , \alpha _{k}\), not all zero, such that \(\alpha _{1} X_{1}+\alpha _{2} X_{2}+\cdots +\alpha _{k} X_{k}=0\). Among all such relations, we choose one of shortest length, say \(j\).

    By rearranging the vectors \(X_{1}, X_{2}, \ldots , X_{k}\) we may assume that

    \begin{equation} \label {prop7eq2} \alpha _{1} X_{1}+\alpha _{2} X_{2}+\cdots +\alpha _{j} X_{j}=0 \end{equation}

    \begin{align} A\left (\alpha _{1} X_{1}\right )+A\left (\alpha _{2} X_{2}\right )+\cdots +A\left (\alpha _{j} X_{j}\right ) & =0 \notag \\ \alpha _{1}\left (A X_{1}\right )+\alpha _{2}\left (A X_{2}\right )+\cdots +\alpha _{j}\left (A X_{j}\right )&=0\notag \\ \alpha _{1} \lambda _{1} X_{1}+\alpha _{2} \lambda _{2} X_{2}+\cdots +\alpha _{j} \lambda _{j} X_{j} & =0 \label {prop7eq3} \end{align} Multiplying (4.13) by \(\lambda _{1}\) and subtracting from (4.14), we get

    \begin{align} \alpha _{2}\left (\lambda _{1}-\lambda _{2}\right ) X_{2}+& \alpha _{3}\left (\lambda _{1}-\lambda _{3}\right ) X_{3}+\cdots +\alpha _{j}\left (\lambda _{1}-\lambda _{j}\right ) X_{j}=0 \label {prop7eq4} \end{align} and since \(\lambda _{1}, \lambda _{2}, \ldots , \lambda _{j}\) are distinct and \(\alpha _{2}, \ldots , \alpha _{j}\) are non-zero we have

    \[ \alpha _{i}\left (\lambda _{1}-\lambda _{i}\right ) \neq 0 ; \quad i=2,3, \ldots , j . \]

    Thus (4.15) gives a relation whose length is \(j-1\), giving a contradiction. Hence \(X_{1}, X_{2}, \ldots , X_{k}\) are linearly independent.

    Property 4.6.13. The characteristic roots of a Hermitian matrix are all real.

    Proof : Let \(A\) be a Hermitian matrix. Hence

    \begin{equation} A=\overline {A}^{T} \label {prop8eq1} \end{equation}

    Let \(\lambda \) be a characteristic root of \(A\) and let \(X\) be a characteristic vector corresponding to \(\lambda \).

    \begin{equation} A X=\lambda X \label {prop8eq2} \end{equation}

    Now,

    \begin{align*} A X &= \lambda X \\ \Rightarrow \overline {X}^{T} A X &= \lambda \overline {X}^{T} X \\ \Rightarrow \left (\overline {X}^{T} A X\right )^{T}&= \lambda \overline {X}^{T} X \quad \left (\text { since } X^{T} A X\right . \text { is a $1 \times 1$ matrix)}\\ \Rightarrow X^{T} A^{T}\left (\overline {X}^{T}\right )^{T}&= \lambda \overline {X}^{T} X \\ \Rightarrow X^{T} A^{T} \overline {X}&= \lambda \overline {X}^{T} X \\ \Rightarrow \overline {X^TA^T} \overline {X} &= \overline {\lambda X^T X } \\ \Rightarrow \overline {X}^{T} \overline {A}^{T} X&= \overline {\lambda X^{T} X} \\ \Rightarrow \overline {X}^{T} A X&= \overline {\lambda } X^{T} \overline {X} \quad \text { (using \eqref {prop8eq1} ) } \\ \Rightarrow \overline {X}^{T} \lambda X&= \overline {\lambda } X^{T} \overline {X} \quad \text { (using \eqref {prop8eq2} ) } \end{align*}

    \begin{equation} \Rightarrow \lambda \left (\overline {X}^{T} X\right )= \overline {\lambda }\left (X^{T} \overline {X}\right ) \label {prop8eq3} \end{equation}

    Now,

    \begin{align*} \overline {X}^{T} X &=X^{T} \overline {X}=\overline {x_{1}} x_{1}+\overline {x_{2}} x_{2}+\ldots \ldots +\overline {x_{n}} x_{n} \\ &=\left |x_{1}\right |^{2}+\left |x_{2}\right |^{2}+\ldots \ldots +\left |x_{n}\right |^{2} \\ &\neq 0 \end{align*} From (4.18) we get \(\lambda =\overline {\lambda }\). Hence \(\lambda \) is real.

    Corollary 4.6.14. The characteristic roots of a real symmetric matrix are real.

    Proof : We know that any real symmetric matrix is Hermitian. Hence the result follows from the above property.

    Property 4.6.15. The characteristic roots of a skew Hermitian matrix are either purely imaginary or zero.

    Proof : Let \(A\) be a skew Hermitian matrix and \(\lambda \) be a characteristic root of \(A\).

    \begin{align*} \therefore |A-\lambda I| & =0 \\ &\therefore |i A-i \lambda I|& =0 \end{align*} \(\therefore i \lambda \) is a characteristic root of \(i A .\)

    Since \(A\) is skew Hermitian \(i A\) is Hermitian. \(\therefore \, i \lambda \) is real. Hence \(\lambda \) is purely imaginary or zero.

    Corollary 4.6.16. The characteristic roots of a real skew symmetric matrix are either purely imaginary or zero.

    Proof : We know that any real skew symmetric matrix is skew Hermitian. Hence the result follows from the above property.

    Property 4.6.17. Let \(\lambda \) be a characteristic root of an unitary matrix \(A\). Then \(|\lambda |=1\) (i.e) the characteristic roots of a unitary matrix are all the unit modulus.

    Proof : Let \(\lambda \) be a characteristic root of an unitary matrix \(A\) and \(X\) be a characteristic vector corresponding to \(\lambda \),

    \begin{equation} AX = \lambda X \label {prop10eq1} \end{equation}

    Taking conjugate and transpose in (4.19) we get \((\overline {A X})^{T}=(\overline {\lambda X})^{T}\).

    \begin{equation} \therefore \overline {X}^{T} \overline {A}^{T}=\overline {\lambda } \overline {X}^{T} \label {prop10eq2} \end{equation}

    Multiplying (4.19) and (4.20), we get

    \begin{align*} \left (\overline {X^{T} A^{T}}\right )(A X)&=\left (\overline {\lambda } \overline {X}^{T}\right )(\lambda X) \\ \therefore \overline {X}^{T}\left (\overline {A}^{T} A\right ) X &=\overline {\lambda } \lambda \left (\overline {X}^{T} X\right ) \end{align*} Now, since \(A\) is an unitary matrix \(\overline {A}^{T} A=I\).
    Hence \(\overline {X}^{T} X=(\overline {\lambda } \lambda ) \overline {X}^{T} X\).
    Since \(X\) is non-zero vector \(\overline {X}^{T}\) is also non-zero vector and \(\overline {X}^{T} X=\left |x_{1}\right |^{2}+\left |x_{2}\right |^{2}+\ldots \ldots +\left |x_{n}\right |^{2} \neq 0\) we get \(\lambda \overline {\lambda }=1\).
    Hence \(|\lambda |^{2}=1\). Hence \(|\lambda |=1\).

    Corollary 4.6.18. Let \(\lambda \) be a characteristic root of an orthogonal matrix \(A\). Then \(|\lambda |=1\).

Since any orthogonal matrix-is unitary the result follows from Property 10.

    Property 4.6.19. Zero is an eigen value of \(A\) if and only if \(A\) is a singular matrix.

    Proof : The eigen values of \(A\) are the roots of the characteristic equation \(|A-\lambda I|=0 .\) Now, 0 is an eigen value of \(A \Leftrightarrow |A-0 I|=0\)

    \begin{align*} \Leftrightarrow |A| =0 \\ \end{align*} \(\Leftrightarrow A\) is a singular matrix

    Property 4.6.20. If \(A\) and \(B\) are two square matrices of the same order then \(A B\) and \(B A\) have the same eigen values.

    Proof : Let \(\lambda \) be an eigen value of \(A B\) and \(X\) be an eigen vector corresponding to \(\lambda \).

    \begin{align*} (A B) X & =\lambda X \\ B(A B) X& =B(\lambda X)=\lambda (B X) \\ (B A)(B X)&=\lambda (B X) \\ (B A) Y&=\lambda Y \text { where } Y=B X . \end{align*} Hence \(\lambda \) is an eigen value of \(B A\).
    Also \(B X\) is the corresponding eigen vector.

    Property 4.6.21. If \(P\) and \(A\) are \(n \times n\) matrices and \(P\) is a nonsingular matrix then \(A\) and \(P^{-1} A P\) have the same eigen values.

    Proof : Let \(B=P^{-1} A P\).
    To prove \(A\) and \(B\) have same eigen values, it is enough to prove that the characteristic polynomials of \(A\) and \(B\) are the same. Now

    \begin{align*} |B-\lambda I| &=\left |P^{-1} A P-\lambda I\right | \\ &=\left |P^{-1} A P-P^{-1}(\lambda I) P\right | \\ &=\left |P^{-1}(A-\lambda I) P\right | \\ &=\left |P^{-1}\right ||A-\lambda I||P| \\ &=\left |P^{-1}\right ||P||A-\lambda I| \\ &=\left |P^{-1} P \| A-\lambda I\right | \\ &=|I||A-\lambda I| \\ &=|A-I \lambda | . \end{align*} The characteristic equations of \(A\) and \(P^{-1} A P\) are the same.

    Property 4.6.22. If \(\lambda \) is a characteristic root of \(A\) then \(f(\lambda )\) is a characteristic root of the matrix \(f(A)\) where \(f(x)\) is any polynomial.

    Proof : Let \(f(x)=a_{0} x^{n}+a_{1} x^{n-1}+\ldots + a_{n-1} x+a_{n}\) where \(a_{0} \neq 0\) and \(a_{1}, a_{2}, \ldots , a_{n}\) are all real numbers.

    \[ f(A)=a_{0} A^{n}+a_{1} A^{n-1}+\ldots +a_{n-1} A+a_{n} I . \]

    Since \(\lambda \) is a characteristic root of \(A\), \(\lambda ^{n}\) is a characteristic root of \(A^{n}\) for any positive integer \(n\) (By Property 6)

    \begin{align*} A^n X & = \lambda ^n X\\ A^{n-1}X & = \lambda ^{n-1}X\\ \cdots & \cdots \cdots \\ \cdots & \cdots \cdots \\ AX & = \lambda X\\ a_0 A^n X & = a_0 \lambda ^ n X \\ a_1 A^{n-1}X & = a_1 \lambda ^{n-1}X \\ \cdots & \cdots \cdots \\ \cdots & \cdots \cdots \\ a_{n-1} A X & =a_{n-1} \lambda X \end{align*} Adding the above equations we have

    \begin{align*} a_{0} A^{n} X+a_{1} A^{n-1} X+\ldots +a_{n-1} A X& =a_{0} \lambda ^{n} X+a_{1} \lambda ^{n-1} X+\ldots +a_{n-1} \lambda X\\ \left (a_{0} A^{n}+a_{1} A^{n-1}+\ldots +a_{n-1} A\right ) X&=\left (a_{0} \lambda ^{n}+a_{1} \lambda ^{n-1}+\ldots +a_{n-1} \lambda \right ) X \\ \left (a_{0} A^{n}+a_{1} A^{n-1}+\ldots +a_{n-1} A+a_{n} I\right ) X & =\left (a_{0} \lambda ^{n}+a_{1} \lambda ^{n-1}+\ldots + a_{n-1} \lambda +a_{n}\right ) X\\ f(A) X&=f(\lambda ) X \end{align*} Hence \(f(\lambda )\) is a characteristic root of \(f(A)\).

Solved Problems

    Problem 4.6.23. If \(X_{1}, X_{2}\) are eigen vectors corresponding to an eigen value \(\lambda \) then \(a X_{1}+b X_{2}(a, b\) non-zero scalars) is also an eigen vector corresponding to \(\lambda \).

Solution : Since \(X_{1}\) and \(X_{2}\) are given vectors corresponding to \(\lambda \), we have

\[ A X_{1}=\lambda X_{1} \text { and } A X_{2}=\lambda X_{2}. \]

Hence \(A\left (a X_{1}\right )=\lambda \left (a X_{1}\right )\) and \(A\left (b X_{2}\right )=\lambda \left (b X_{2}\right )\)

\[ A\left (a X_{1}+b X_{2}\right )=\lambda \left (a X_{1}+b X_{2}\right ) \text {. } \]

\(a X_{1}+b X_{2}\) is an eigen vector corresponding to \(\lambda \).

    Problem 4.6.24. If the eigen values of \(A =\begin {pmatrix} 3 & 10 & 5 \\ -2 & -3 & -4 \\ 3 & 5 & 7 \end {pmatrix} \) are 2,2,3 find the eigen values of \(A^{-1}\) and \(A^{2}\).

Solution : Since 0 is not an eigen value of \(A\), \(A\) is non singular matrix and hence \(A^{-1}\) exists.

Eigen values of \(A^{-1}\) are \(,\dfrac {1}{2},\dfrac {1}{2},\dfrac {1}{3}\) and eigen values of \(A^2\) are \(2^2, 2^2, 3^2 \).

    Problem 4.6.25. Find the eigen values of \(A^{5}\) when \(A= \begin {pmatrix} 3 & 0 & 0 \\ 5 & 4 & 0 \\ 3 & 6 & 1 \end {pmatrix}\).

Solution : The characteristic equation of \(A\) is obviously

\[(3-\lambda )(4-\lambda )(1-\lambda )=0.\]

Hence the eigen values of \(A\) are \(3,4,1\).
The eigen values of \(A^{5}\) are \(3^{5}, 4^{5}, 1^{5}\).

    Problem 4.6.26. Find the sum and product of the eigen values of the matrix \(\left (\begin {array}{lll}3 & -4 & 4 \\ 1 & -2 & 4 \\ 1 & -1 & 3\end {array}\right ) \) without actually finding the eigen values.

Solution : Let \(A=\left (\begin {array}{lll}3 & -4 & 4 \\ 1 & -2 & 4 \\ 1 & -1 & 3\end {array}\right )\)
Sum of the eigen values \(=\) trace of \(A=3+(-2)+3=4\).
Product of the eigen values \(=|A| .\) Now,

\begin{align*} |A| &=\left |\begin{array}{rrr}3 & -4 & 4 \\ 1 & -2 & 4 \\ 1 & -1 & 3\end {array}\right | \\ &=3(-6+4)+4(3-4)+4(-1+2) \\ &=-6-4+4=-6 \end{align*} Product of the eigen values \(=-6\).

    Problem 4.6.27. Find the characteristic roots of the matrix \(\begin {pmatrix} \cos \theta & -\sin \theta \\ -\sin \theta & \cos \theta \end {pmatrix} \)

Solution : Let \(A=\left (\begin {array}{rr}\cos \theta & -\sin \theta \\ -\sin \theta & \cos \theta \end {array}\right )\)
The characteristic equation of \(A\) is given by \(|A-\lambda I |=0\).

\begin{align*} \begin{vmatrix} \cos \theta -\lambda & -\sin \theta \\ -\sin \theta & \cos \theta -\lambda \end {vmatrix}&=0 \\ (\cos \theta -\lambda )^{2}-\sin ^{2} \theta &=0\\ (\cos \theta -\lambda -\sin \theta )(\cos \theta -\lambda +\sin \theta )&=0\\ [\lambda -(\cos \theta -\sin \theta )][\lambda -(\cos \theta +\sin \theta )]&=0 \end{align*} The two characteristic roots, (the two eigen values) of the matrix are \((\cos \theta -\sin \theta )\) and \((\cos \theta +\sin \theta )\).

    Problem 4.6.28. Find the characteristic roots of the matrix \(A=\left (\begin {array}{rr} \cos \theta & -\sin \theta \\ -\sin \theta & -\cos \theta \end {array}\right ) \).

Solution : The characteristic equation of \(A\) is given by \(|A-\lambda I|=0\).

\begin{align*} \begin{vmatrix} \cos \theta -\lambda & -\sin \theta \\ -\sin \theta & -\cos \theta -\lambda \end {vmatrix} & =0 \\ -\left (\cos ^{2} \theta -\lambda ^{2}\right )-\sin ^{2} \theta &=0 \\ \therefore \quad \lambda ^{2}-\left (\cos ^{2} \theta +\sin ^{2} \theta \right ) &=0 \\ \therefore \quad \lambda ^{2}-1&=0 . \end{align*} The characteristic roots are 1 and \(-1\).

    Problem 4.6.29. Find the sum and product of the eigen values of the matrix \(A=\left (\begin {array}{ll}a_{11} & a_{12} \\ a_{21} & a_{22}\end {array}\right )\) without finding the roots of the characteristic equation.

Solution : Sum of the eigen values of

\[ A=\text { trace of } A=a_{11}+a_{22} \]

Product of the eigen values of

\[ A= |A|=a_{11} a_{22}-a_{12} a_{21} . \]

    Problem 4.6.30. Verify the statement that the sum of the elements in the diagonal of a matrix is the sum of the eigen values of the matrix \(A= \begin {pmatrix} -2 & 2 & -3 \\ 2 & 1 & -6 \\ -1 & -2 & 0 \end {pmatrix} \)

Solution : The characteristic equation of \(A\) is \(|A- \lambda I|=0\).

\begin{align*} \begin{vmatrix}-2-\lambda &2&-3\\ 2&1-\lambda &-6\\ -1&-2&-\lambda \end {vmatrix} &=0\\ \left (-2-\lambda \right ) \begin{vmatrix}1-\lambda &-6\\ -2&-\lambda \end {vmatrix}-2\cdot \begin{vmatrix}2&-6\\ -1&-\lambda \end {vmatrix}-3\cdot \begin{vmatrix}2&1-\lambda \\ -1&-2\end {vmatrix} &=0\\ \left (-2-\lambda \right )\left (-\lambda +\lambda ^2-12\right )-2\left (-2\lambda -6\right )-3\left (-\lambda -3\right ) &=0\\ -\lambda ^3-\lambda ^2+14\lambda +24+4\lambda +12+3\lambda +9 &=0\\ -\lambda ^3-\lambda ^2+21\lambda +45 &=0 \\ \lambda ^3+\lambda ^2-21\lambda -45 &=0 \end{align*} This is a cubic equation in \(\lambda \) and hence it has 3 roots and the three roots are the three eigen values of the matrix.
The sum of the eigen values = \(- \left (\dfrac {\text {coefficient of } \lambda ^2 }{\text {coefficient of } \lambda ^3} \right )=-1\)
The sum of the elements on the diagonal of the matrix \(A = -2+1+0=-1 \). Hence the result.

    Problem 4.6.31. The product of two eigen values of the matrix \(A = \begin {pmatrix} 6 & -2 & 2 \\ -2 & 3 & -1 \\ 2 & -1&3 \end {pmatrix} \) is 16. Find the third eigen value. What is the sum of the eigen values of \(A\)?

Solution : Let \(\lambda _{1}, \lambda _{2}, \lambda _{3}\) be the eigen values of \(A\).
Given, product of 2 eigen values (say) \(\lambda _{1}, \lambda _{2}\) is \(16.\)

\[ \therefore \lambda _{1} \lambda _{2}=16 \]

We know that the product of the eigen values is \(|A|\).

\begin{align*} \lambda _{1} \lambda _{2} \lambda _{3} & =\left |\begin{array}{rrr}6 & -2 & 2 \\ -2 & 3 & -1 \\ 2 & -1 & 3\end {array}\right |\\ 16 \lambda _{3} &=6(9-1)+2(-6+2)+2(2-6) \\ &=48-8-8 \\ &=32 \\ \therefore \lambda _{3} &=2 \end{align*} \(\therefore \) The third eigen value is 2.
Also we know that the sum of the eigen values of

\[ A=\operatorname {trace} \text { of } A=6+3+3=12 \]

    Problem 4.6.32. The product of two eigen values of the matrix \(A=\left (\begin {array}{rrr}2 & 2 & -7 \\ 2 & 1 & 2 \\ 0 & 1 & -3\end {array}\right )\) is \(-12\). Find the eigen values of \(A\).

Solution : Let \(\lambda _{1}, \lambda _{2}, \lambda _{1}\) be the eigen values of \(A\).
Given product of 2 eigen values, say, \(\lambda _{1}\) and \(\lambda _{1}\) is \(-12\).

\begin{equation} \lambda _{1} \lambda _{2}=-12 \label {prob10eq1} \end{equation}

We know that the product of the eigen values is \(|A|\) We know that the product of the eigen values is \(|A|\).

\begin{align} \therefore \quad \lambda _{1} \lambda _{2} \lambda _{3} & =\left |\begin{array}{rrr} 2 & 2 & -7 \\ 2 & 1 & 2 \\ 0 & 1 & -3 \end {array}\right | \notag \\ 12 \lambda _{3}& =-12 \notag \\ \therefore \quad \lambda _{3} & =1 \label {prob10eq2} \end{align} Also we know sum of the eigen values \(=\) Trace of \(A\).

\begin{align} \lambda _{1}+\lambda _{2}+\lambda _{3}& =2+1-3=0\notag \\ \lambda _{1}+\lambda _{2} & =-1 \quad \text { (using \eqref {prob10eq2} ) } \label {prob10eq3} \end{align} Using (4.23) in (4.21) we get

\begin{align*} \lambda _{1}\left (-1-\lambda _{1}\right )&=-12\\ \lambda _{1}^{2}+\lambda _{1}-12 &=0 \\ \left (\lambda _{1}+4\right )\left (\lambda _{1}-3\right )&=0 \\ \therefore \lambda _{1}&=3 \text { or }-4 \end{align*} Putting \(\lambda _{1}=3\) in (4.21) we get \(\lambda _{2}=-4\). Or putting \(\lambda _{1}=-4\) in (4.21) we get \(\lambda _{2}=3\).
Thus the three eigen values are \(3,-4,1\).

    Problem 4.6.33. Find the sum of the squares of the eigen values of \(A=\left (\begin {array}{lll}3 & 1 & 4 \\ 0 & 2 & 6 \\ 0 & 0 & 5\end {array}\right )\)

Solution : Let \(\lambda _{1}, \lambda _{2}, \lambda _{3}\) be the eigen values of \(A\).
We know that \(\lambda _{1}^{2}, \lambda _{2}^{2}, \lambda _{3}^{2}\) are the eigen values of \(A^{2}\).

\begin{align*} A^{2} &=\left (\begin{array}{lll} 3 & 1 & 4 \\ 0 & 2 & 6 \\ 0 & 0 & 5 \end {array}\right ) \cdot \left (\begin{array}{lll} 3 & 1 & 4 \\ 0 & 2 & 6 \\ 0 & 0 & 5 \end {array}\right ) \\ &=\left (\begin{array}{lll} 9 & 5 & 38 \\ 0 & 4 & 42 \\ 0 & 0 & 25 \end {array}\right ) \end{align*} Sum of the eigen values of \(A^{2}=\) Trace of \(A^{2}=9+4+25\)

\[i.e., \lambda _{1}^{2}+\lambda _{2}^{2}+\lambda _{3}^{2}=38 \]

Sum of the squares of the eigen values of \(A=38\).

    Problem 4.6.34. Find the eigen values and eigen vectors of the matrix \(A=\left (\begin {array}{lll} 1 & 1 & 3 \\ 1 & 5 & 1 \\ 3 & 1 & 1 \end {array}\right )\).

Solution : The characteristic equation of \(A\) is

\begin{align*} |A-\lambda I| & =0 \\ \therefore \left |\begin{array}{ccc} 1-\lambda & 1 & 3 \\ 1 & 5-\lambda & 1 \\ 3 & 1 & 1-\lambda \end {array}\right | & =0 \\ \therefore (1-\lambda )[(5-\lambda )(1-\lambda )-1]-[(1-\lambda )-3]+3[1-3(5-\lambda )] & =0 \\ (1-\lambda )\left (\lambda ^{2}-6 \lambda +4\right )+(\lambda +2)+3(3 \lambda -14) & =0 \\ \lambda ^{2}-6 \lambda +4-\lambda ^{3}+6 \lambda ^{2}-4 \lambda +\lambda +2+9 \lambda -42 & =0 \\ -\lambda ^{3}+7 \lambda ^{2}-36& =0 \\ \lambda ^{3}-7 \lambda ^{2}+36&=0 \\ (\lambda +2)\left (\lambda ^{2}-9 \lambda +18\right )&=0 \\ (\lambda +2)(\lambda -6)(\lambda -3)&=0 \end{align*} \(\lambda =-2,3,6\) are the three eigen values.

Case (i) Eigen vector corresponding to \(\lambda =-2\).
Let \(X=\left (\begin {array}{l} x_{1} \\ x_{2} \\ x_{3} \end {array}\right ) \) be an eigen vector corresponding to \(\lambda =-2\)
Hence \(A X=-2 X\).

\begin{align*} \left (\begin{array}{lll} 1 & 1 & 3 \\ 1 & 5 & 1 \\ 3 & 1 & 1 \end {array}\right )\left (\begin{array}{l} x_{1} \\ x_{2} \\ x_{3} \end {array}\right )&=\left (\begin{array}{l} -2 x_{1} \\ -2 x_{2} \\ -2 x_{3} \end {array}\right ) \\ \therefore x_{1}+x_{2}+3 x_{3}&=-2 x_{1} \\ x_{1}+5 x_{2}+x_{3}&=-2 x_{2} \\ 3 x_{1}+x_{2}+x_{3}&=-2 x_{3} \end{align*}

\begin{align} 3 x_{1}+x_{2}+3 x_{3}& =0 \label {prob12eq1} \\ x_{1}+7 x_{2}+x_{3}&=0 \label {prob12eq2}\\ 3 x_{1}+x_{2}+3 x_{3}&=0 \label {prob12eq3} \end{align} Clearly this system of three equations reduces to two equations only. From (4.24) and (4.25) we get

\[ \because \quad x_{1}=-2 k ; x_{2}=0 ; x_{3}=2 k. \]

\(\therefore \) It has only one independent solution and can be obtained by giving any value to \(k\) say \(k=1\).
\(\therefore \quad (-2,0,2)\) is an eigen vector corresponding to \(\lambda =-2\).

Case (ii) Eigen vector corresponding to \(\lambda =3\).
Then \(AX = 3 X\)

\begin{align*} -2 x_{1}+x_{2}+3 x_{3}&= 0 \\ x_{1}+2 x_{2}+x_{3} &=0 \\ 3 x_{1}+x_{2}-2 x_{3}&= 0 \end{align*} Taking the first 2 equations we get

\begin{align*} \frac {x_{1}}{-5} & =\frac {x_{2}}{5}=\frac {x_{3}}{-5}=k \text { (say). } \\ x_{1}&=-k ; x_{2}=k ; x_{3}=-k . \end{align*} Taking \(k=1\) (say) \((-1,1,-1)\) is an eigen vector corresponding to \(\lambda =3\).

Case (iii) Eigen vector corresponding to \(\lambda =6\),
We have \(A X=6 X\).

\begin{align*} -5 x_{1}+x_{2}+3 x_{3} &=0 \\ x_{1}-x_{2}+x_{3} &=0 \\ 3 x_{1}+x_{2}-5 x_{3} &=0 \end{align*} Taking the first two equations we get

\[ \frac {x_{1}}{4}=\frac {x_{2}}{8}=\frac {x_{3}}{4}=k \]

\(\therefore \quad x_{1}=k ; x_{2}=2 k ; x_{3}=k\). It satisfies the third equation also.
Taking \(k=1\) (say) \((1,2,1)\) is an eigen vector corresponding to \(\lambda =6\).
Therefore \((-2,0,2), (-1,1,-1)\) and \((1,2,1)\) are the eigen vectors of the eigen values \(-2,3,6\) respectively.

    Problem 4.6.35. Find the eigen values and eigen vectors of the matrix \(A= \begin {pmatrix} 6 & -2 & 2 \\ -2 & 3 & -1 \\ 2 & -1 & 3 \end {pmatrix}\).

Solution : The characteristic equation of \(A\) is \(|A-\lambda I|=0\).

\begin{align*} \left |\begin{array}{ccc} 6-\lambda & -2 & 2 \\ -2 & 3-\lambda & -1 \\ 2 & -1 & 3-\lambda \end {array}\right | & =0 . \\ \therefore \quad (6-\lambda )\left [(3-\lambda )^{2}-1\right ]+2[(2 \lambda -6)+2] +2(2-6+2 \lambda )&=0\\ \therefore \quad (6-\lambda )\left (8+\lambda ^{2}-6 \lambda \right )+4 \lambda -8+4 \lambda -8&=0 \\ 48+6 \lambda ^{2}-36 \lambda -8 \lambda -\lambda ^{3}+6 \lambda ^{2}+8 \lambda -16&=0 \\ -\lambda ^{3}+12 \lambda ^{2}-36 \lambda +32&=0\\ \lambda ^{3}-12 \lambda ^{2}+36 \lambda -32&=0 \\ (\lambda -2)(\lambda -2)(\lambda -8)&=0 \end{align*} The eigen values are \(2,2,8\). We now find the eigen vectors.
Case (i) Eigen vector corresponding to \(\lambda =2\).
The eigen vector \(X=\left (\begin {array}{l}x_{1} \\ x_{2} \\ x_{3}\end {array}\right )\) is got from \(A X=2 X\)

\begin{align*} 6 x_{1}-2 x_{2}+2 x_{3}&=2 x_{1} \\ -2 x_{1}+3 x_{2}-x_{3}&=2 x_{2} \\ 2 x_{1}-x_{2}+3 x_{3}&=2 x_{3} \\ 4 x_{1}-2 x_{2}+2 x_{3}&=0 \\ -2 x_{1}+x_{2}-x_{3}&=0\\ 2 x_{1}-x_{2}+x_{3}&=0 \end{align*} The above three equations are equivalent to the single equation \(2 x_{1}-x_{2}+x_{3}=0\).
The independent eigen vectors can be obtained by giving arbitrary values to any two of the unknowns \(x_{1}, x_{2}, x_{3}\). Let \(x_{1}=1 ; x_{2}=2\) we get \(x_{3}=0\).
Let \(x_{1}=3 ; x_{2}=4 \) we get \(x_{3}=-2\).
\(\therefore \quad \) Two independent vectors corresponding to \(\lambda = 2\) are \((1,2,0)\) and \((3,4,-2)\).
Case (ii) Eigen vector corresponding to \(\lambda =8\).
The eigen vector \(X= \begin {pmatrix} x_{1} \\ x_{2} \\ x_{3} \end {pmatrix}\) is got from \(X=8 X\).

\begin{align} \therefore \quad -2 x_{1}-2 x_{2}+2 x_{3}&=0 \label {prob13eq1} \\ -2 x_{1}-5 x_{2}-x_{3}&=0\label {prob13eq2} \\ 2 x_{1}-x_{2}-5 x_{3}&=0\label {prob13eq3} \end{align} From (4.27) and (4.28) we get

\begin{align*} \frac {x_{1}}{12} &=\frac {x_{2}}{-6} &=\frac {x_{3}}{6}=k \text { (say). } \\ \therefore \quad x_{1} &=2 k ; x_{2}=-k ; x_{3}=k . \end{align*} Giving \(k=1\) we get an eigen vector corresponding to 8 as \((2,-1,1)\).
Therefore \((1,2,0), (3,4,-2)\) and \((2,-1,1)\) are the eigen vectors of the eigen values \(2,2,8\) respectively.

    Problem 4.6.36. Find the eigen values and eigen vectors of the matrix \(A=\left (\begin {array}{rrr} 2 & -2 & 2 \\ 1 & 1 & 1 \\ 1 & 3 & -1 \end {array}\right ) \)

Solution : The characteristic equation of \(A\) is \(|A-\lambda l|=0\).

\begin{align*} \left |\begin{array}{ccc}2-\lambda & -2 & 2 \\ 1 & 1-\lambda & 1 \\ 1 & 3 & -1-\lambda \end {array}\right |&=0 \\ (2-\lambda )[-(1-\lambda )(1+\lambda )-3] +2[-(1+\lambda )-1]+2[3-(1-\lambda )]&=0\\ (2-\lambda )\left (\lambda ^{2}-4\right )-2(2+\lambda )+2(2+\lambda )&=0\\ 2 \lambda ^{2}-8-\lambda ^{3}+4 \lambda -4-2 \lambda +4+2 \lambda &=0\\ -\lambda ^{3}+2 \lambda ^{2}+4 \lambda -8&=0\\ \lambda ^{3}-2 \lambda ^{2}-4 \lambda +8&=0 \\ (\lambda -2)\left (\lambda ^{2}-4\right )&=0\\ (\lambda -2)(\lambda -2)(\lambda +2)&=0 \end{align*} \(\lambda =2,2,-2 \) are the three eigen values.
Case (i) Eigen vector corresponding to \(\lambda =2\).
Let \(X=\left (x_{1}, x_{2}, x_{3}\right )\) be an eigen vector corresponding to \(\lambda =2, X\) is got from \(A X=2 X\).

\[\left (\begin {array}{rrr} 2 & -2 & 2 \\ 1 & 1 & 1 \\ 1 & 3 & -1 \end {array}\right )\left (\begin {array}{l} x_{1} \\ x_{2} \\ x_{3} \end {array}\right )=\left (\begin {array}{l} 2 x_{1} \\ 2 x_{2} \\ 2 x_{3} \end {array}\right ) \]

The eigen vector corresponding to \(\lambda =2\) is given by the equations

\begin{align*} 2 x_{1}-2 x_{2}+2 x_{3} &=2 x_{1} \\ x_{1}+x_{2}+x_{3} &=2 x_{2} \\ x_{1}+3 x_{2}-x_{3} &=2 x_{3} \end{align*}

\begin{align} \text { (i.e.) }-x_{2}+x_{3} &=0\label {prob14eq1} \\ x_{1}-x_{2}+x_{3} &=0 \label {prob14eq2}\\ x_{1}+3 x_{2}-3 x_{3} &=0 \label {prob14eq3} \end{align} Taking (4.30) and (4.31) we get

\begin{align*} \frac {x_{1}}{0} &=\frac {x_{2}}{1}=\frac {x_{3}}{1}=k \text { (say)}.\\ \therefore \quad x_{1}&=0 ; x_{2}=k ; x_{3}=k \end{align*} Taking \(k=1\), we get \((0,1,1)\) as an eigen vector corresponding to \(\lambda =2\).
Case (ii) Eigen vector corresponding to \(\lambda =-2\).
Corresponding to \(\lambda =-2\) we have \(A X=-2 X\).

\begin{align*} 2 x_{1}-2 x_{2}+2 x_{3} &=-2 x_{1} \\ x_{1}+x_{2}+x_{3} &=-2 x_{2} \\ x_{1}+3 x_{2}-x_{3} &=-2 x_{3} \end{align*}

\begin{align} 2 x_{1}-x_{2}+x_{3} &=0 \label {prob14eq4} \\ x_{1}+3 x_{2}+x_{3} &=0 \label {prob14eq5}\\ x_{1}+3 x_{2}+x_{3} &=0 \label {prob14eq6} \end{align} Taking (4.33) and (4.34) we get,

\begin{align*} \frac {x_{1}}{-4}=\frac {x_{2}}{-1}=\frac {x_{3}}{7}&=k \text { (say). } \\ x_{1}&=-4 k ; x_{2}=-k ; x_{3}=7 k. \end{align*} Taking \(k=1\) we get \((-4,-1,7)\) as an eigen vector corresponding to the eigen value \(\lambda =-2\).

Post a Comment

0 Comments

/* MCQ Temp */