Fall Semester Faculty of Aeronautics and
Review of Matrices MAT 201E
2023/2024 Astronautics
A matrix A consists of a rectangular array of numbers, or elements, arranged in
𝑚 rows and 𝑛 columns—that is,
𝑎11 𝑎12 ⋯ 𝑎1𝑛
𝑎21 𝑎22 ⋯ 𝑎2𝑛
𝐀=( ⋮ ⋮ ⋱ ⋮ )
𝑎𝑚1 𝑎𝑚2 ⋯ 𝑎𝑚𝑛
A is an 𝑚 × 𝑛 matrix.
NOTE: If A is an 𝑛 × 1 matrix, then it is called a column vector or simply a vector.
The element in the 𝑖th row and 𝑗th column is designated by 𝑎𝑖𝑗 . Hence, the first
subscript identifies its row and the second its column.
Each 𝐀 = (𝑎𝑖𝑗 ) matrix has:
• the transpose of 𝐀, denoted as 𝐀T , in which the columns and rows are
interchanged such that,
𝐀T = (𝑎𝑗𝑖 )
• ̅ , in which 𝑎𝑖𝑗 is replaced with its complex
the conjugate of 𝐀, denoted as 𝐀
conjugates 𝑎̅𝑖𝑗
𝐀̅ = (𝑎̅𝑖𝑗 )
• the transpose of the conjugate matrix (𝐀̅T ) which is called the adjoint of 𝐀,
and denoted as 𝐀 .∗
Properties of Matrices
1. Equality. Two 𝑚 × 𝑛 matrices 𝐀 and 𝐁 are equal, if corresponding elements are
equal—that is, if 𝑎𝑖𝑗 = 𝑏𝑖𝑗 for each 𝑖 and 𝑗.
2. Zero. The symbol 0 will be used to denote the matrix (or vector) each of whose
elements is zero.
3. Addition & subtraction. Two 𝑚 × 𝑛 matrices 𝐀 and 𝐁 can be added or
subtracted by adding/subtracting the corresponding elements as follows,
𝐀 ± 𝐁 = (𝑎𝑖𝑗 ) ± (𝑏𝑖𝑗 ) = (𝑎𝑖𝑗 ± 𝑏𝑖𝑗 )
𝐁 ± 𝐀 = (𝑏𝑖𝑗 ) ± (𝑎𝑖𝑗 ) = (𝑎𝑖𝑗 ± 𝑏𝑖𝑗 )
4. Multiplication by a Number. The product of a matrix 𝐀 by a number 𝛾 is
defined as follows,
𝛾𝐀 = 𝛾(𝑎𝑖𝑗 ) = (𝛾𝑎𝑖𝑗 )
The distributive laws,
𝛾(𝐀 + 𝐁) = 𝛾𝐀 + 𝛾𝐁 ; (𝛾 + 𝜑)𝐀 = 𝛾𝐀 + 𝜑𝐀 ; −𝐀 = (−1)𝐀
1
Fall Semester Faculty of Aeronautics and
Review of Matrices MAT 201E
2023/2024 Astronautics
5. Multiplication of matrices. The product of 𝐀 and 𝐁 matrices is defined
whenever the number of columns in the first factor and number of rows in the
second factor are the same. If 𝐀 and 𝐁 are 𝑚 × 𝑛 and 𝑛 × 𝑟 matrices, respectively,
then the product 𝐂 = 𝐀𝐁 is an 𝑚 × 𝑟 matrix.
𝑛
𝑐𝑖𝑗 = ∑ 𝑎𝑖𝑘 𝑏𝑘𝑗
𝑘=1
Matrix multiplication satisfies the associative law,
(𝐀𝐁)𝐂 = 𝐀(𝐁𝐂)
and the distributive law,
𝐀(𝐁 + 𝐂) = 𝐀𝐁 + 𝐀𝐂
However, in general, matrix multiplication is not commutative, such that
𝐀𝐁 ≠ 𝐁𝐀
for most cases.
6. Multiplication of vectors.
—There are several ways of forming a product of two vectors 𝐱 and 𝐲, each with
𝑛 components. One is a direct extension to 𝑛 dimensions of the familiar dot
product from calculus; we denote it by 𝐱 𝐓 𝐲 and write
𝑛
𝐓
𝐱 𝐲 = ∑ 𝑥𝑖 𝑦𝑖 (1)
𝑖=1
the results is a (complex) number, and it follows directly that
𝐱 𝐓 𝐲 = 𝐲 𝐓 𝐱, 𝐱 𝐓 (𝐲 + 𝐳) = 𝐱 𝐓 𝐲 + 𝐱 𝐓 𝐳, (𝜶𝐱 𝐓 )𝐲 = 𝜶(𝐱 𝐓 𝐲) = 𝐱 𝐓 (𝜶𝐲)
—There is another vector product tat is also defined for any two vectors having
the same number of components. This product, denoted by (𝐱, 𝐲), is called the
scalar or inner product, and is defined by
𝑛
(𝐱, 𝐲) = ∑ 𝑥𝑖 𝑦̅𝑖 (2)
𝑖=1
—The scalar product is also a (complex) number, and by comparing the Eqs. (1)
and (2) we see that
(𝐱, 𝐲) = 𝐱 𝐓 𝐲̅
2
Fall Semester Faculty of Aeronautics and
Review of Matrices MAT 201E
2023/2024 Astronautics
If all the elements of 𝐲 is real then the two products (1) and (2) are identical.
From the Eq. (2) it follows that
̅̅̅̅̅̅̅
(𝐱, 𝐲) = (𝐲, 𝐱), (𝐱, 𝐲 + 𝐳) = (𝐱, 𝐲) + (𝐱, 𝐳),
(𝜶𝐱, 𝐲) = α(𝐱, 𝐲), (𝐱, 𝛂𝐲) = α
̅(𝐱, 𝐲)
Note that even if the vector 𝐱 has elements with nonzero imaginary parts, the
scalar product of 𝐱 with itself yields a nonnegative real number,
𝑛 𝑛
(𝐱, 𝐱) = ∑ 𝑥𝑖 𝑥̅𝑖 = ∑|𝑥𝑖 |2
𝑖=1 𝑖=1
The nonnegative quantity (𝐱, 𝐱)1/2, often denoted by ‖𝐱‖, is called the length or
magnitude, of 𝐱. If (𝐱, 𝐲) = 0, then the two vectors 𝐱 and 𝐲 are said to be
orthogonal. For example, the unit vectors 𝐢, 𝐣, 𝐤 of three-dimensional vector
geometry form an orthogonal set. On the other hand, if some of the elements of 𝐱
are not real, then the product
𝑛
(𝐱 T , 𝐱) = ∑ 𝑥𝑖 2
𝑖=1
may not be a real number.
Example: Let
𝑖 2−𝑖
𝐱 = [ −2 ], 𝐲=[ 𝑖 ]
1+𝑖 3
Then
𝐱 𝐓 𝐲 = (i)(2 − i) + (−2)(i) + (1 + i)(3) = 4 + 3i
(𝐱, 𝐲) = (i)(2 + i) + (−2)(−i) + (1 + i)(3) = 2 + 7i
𝐱 𝐓 𝐱 = (i)2 + (−2)2 + (1 + i)2 = 3 + 2i
(𝐱, 𝐱) = (𝑖)(−𝑖) + (−2)(−2) + (1 + 𝑖)(1 − 𝑖) = 7
7. Identity. The multiplicative identity, or simply the identity matrix 𝐈 is given by
1 0 ⋯ 0
0 1 ⋯ 0
𝐈=( )
⋮ ⋮ ⋱ ⋮
0 0 ⋯ 1
From the definition of matrix multiplication, we have
𝐀𝐈 = 𝐈𝐀 = 𝐀
for any square matrix 𝐀.
3
Fall Semester Faculty of Aeronautics and
Review of Matrices MAT 201E
2023/2024 Astronautics
8. Inverse. The square matrix 𝐀 is said to be nonsingular or invertible if there is
another matrix 𝐁 such that 𝐀𝐁 = 𝐈 and 𝐁𝐀 = 𝐈, where 𝐈 is the identity. 𝐁 is called
multiplicative inverse, or simply the inverse of 𝐀, and we write 𝐁 = 𝐀−𝟏 Hence,
𝐀𝐀−𝟏 = 𝐀−𝟏 𝐀 = 𝐈
Matrices that do not have an inverse is called singular or noninvertible.
There are various ways to compute 𝐀−𝟏 from 𝐀, assuming that 𝐀 is nonsingular.
One way involves the use of determinants. Associated with each element 𝑎𝑖𝑗 of a
given matrix is the minor 𝑀𝑖𝑗 , which is the determinant of the matrix obtained by
deleting the 𝑖th row and 𝑗th column of the original matrix (i.e. deleting the row
and column containing 𝑎𝑖𝑗 ). Also, associated with the element 𝑎𝑖𝑗 is the cofactor
𝐶𝑖𝑗 defined by the equation
𝐶𝑖𝑗 = (−1)𝑖+𝑗 𝑀𝑖𝑗
If 𝐁 = 𝐀−𝟏 , then it can be shown that the general element 𝑏𝑖𝑗 is given by
𝐶𝑖𝑗
𝑏𝑖𝑗 =
det 𝐀
This is not an efficient way of calculate 𝐀−𝟏 . However, this clearly indicates that if
det 𝐀 = 𝟎, then 𝐀 is singular. And, if det 𝐀 ≠ 𝟎, 𝐀 is nonsingular.
A better way to calculate 𝐀−𝟏 is by the use elementary row operations which are
I. Interchange of two rows.
II. Multiplication of a row by a nonzero scalar.
III. Addition of any multiple of one row to another row.
This method is referred to as row reduction or Gaussian elimination. Any
nonsingular matrix 𝐀 cane be transformed to identity 𝐈 by a systematic sequence
of operations. It is possible to show that if the same sequence of operations is
then performed on 𝐈, it is transformed to 𝐀−𝟏 . It is most efficient to perform the
sequence of operations on both matrices at the same time by forming the
augmented matrix 𝐀|𝐈.
Example: Find the inverse of
1 −1 −1
𝐀 = (3 −1 2 )
2 2 3
Solution:
—Form the augmented matrix 𝐀|𝐈:
1 −1 −1 | 1 0 0
𝐀 = (3 −1 2 | 0 1 0)
2 2 3 | 0 0 1
4
Fall Semester Faculty of Aeronautics and
Review of Matrices MAT 201E
2023/2024 Astronautics
—Obtain zeros in the off-diagonal positions in the first column
Add (-3) times the first row to the second row,
Add (-2) times the first row to the third row.
1 −1 −1 | 1 0 0
(0 2 5 | −3 1 0)
0 4 5 | −2 0 1
—Obtain a 1 in the diagonal position
1
Multiply the second row by (2)
1 −1 −1 | 1 0 0
(0 1 5/2 | −3/2 1/2 0)
0 4 5 | −2 0 1
—Obtain zeros in the off-diagonal position in the second column
Add second row to the first row
Add (-4) times the second row to the third row
1 0 3/2 | −1/2 1/2 0
(0 1 5/2 | −3/2 1/2 0)
0 0 −5 | 4 −2 1
—Obtain a 1 in the diagonal position in the third column
1
Multiply the third column by (− 5)
1 0 3/2 | −1/2 1/2 0
(0 1 5/2 | −3/2 1/2 0 )
0 0 1 | −4/5 2/5 −1/5
—Obtain zeros for off-diagonal positions in the third column
3
Add (− 2) times the third row to the first row
5
Add (− 2) times the third row to the second row
1 0 0 | 7/10 −1/10 3/10
(0 1 0 | 1/2 −1/2 1/2 )
0 0 1 | −4/5 2/5 −1/5
The last matrix is 𝐈|𝐀−𝟏, a fact that can be verified with direct multiplication with
the original matrix 𝐀.
Matrix functions. Sometimes vectors or matrices consisted of elements that
are functions of a real variable 𝑡, such as
5
Fall Semester Faculty of Aeronautics and
Review of Matrices MAT 201E
2023/2024 Astronautics
𝑥1 (𝑡) 𝑎11 (𝑡) … 𝑎1𝑛 (𝑡)
𝐱(𝑡) = ( ⋮ ), 𝐀(𝑡) = ( ⋮ ⋮ )
𝑥𝑛 (𝑡) 𝑎𝑛1 (𝑡) … 𝑎𝑛𝑛 (𝑡)
The matrix 𝐀(𝑡) is said to be continuous at 𝑡 = 𝑡0 or on an interval 𝛼 < 𝑡 < 𝛽 if
each element of 𝐀 is continuous function at the given point or on the given interval.
Similarly, 𝐀(𝑡) is said to be differentiable if each of its elements is differentiable,
and its derivative 𝑑𝐀⁄𝑑𝑡 is defined by
𝒅𝐀 𝑑𝑎𝑖𝑗
=( );
𝒅𝒕 𝑑𝑡
that is, each element of 𝑑𝐀⁄𝑑𝑡 is the derivative of the corresponding element of 𝐀
In the same way, the integral of a matrix functions is defined as
𝑏 𝑏
∫ 𝐀(𝑡)𝑑𝑡 = (∫ 𝑎𝑖𝑗 (𝑡)𝑑𝑡)
𝑎 𝑎
Eigenvalues, Eigenvectors, Linear Independence and Hermitian
Matrices
𝐀 = (aij ) is an 𝑛 × 𝑛 square matrix, and 𝐱 and 𝐲 are 𝑛 × 1 vectors
𝐀𝐱 = 𝐲
and this equation can be viewed as a linear transformation that transforms a given
vector 𝐱 to a new vector 𝐲. Vectors that are transformed into multiples of
themselves are important in many applications (e.g. finding the principal axes of
stress or strain in an elastic body, and finding the modes of free vibration in a
conservative system with a finite number of degrees of freedom). To find such
vectors, we set 𝐲 = λ𝐱, where λ is a scalar proportionality factor, and seek
solutions of the equation
𝐀𝐱 = λ𝐱 (3)
or
(𝐀 − λ𝐈)𝐱 = 0 (4)
The latter has nonzero solutions if and only if λ is chosen so that
det(𝐀 − λ𝐈) = 0. (5)
Values of λ that satisfy the above Eq. (5) is called eigenvalues of the matrix 𝐀, and
the nonzero solutions of Eqs. (3) or (4) that are obtained by using such a value of
λ are called eigenvectors corresponding to that eigenvalue.
6
Fall Semester Faculty of Aeronautics and
Review of Matrices MAT 201E
2023/2024 Astronautics
Example: Find the eigenvalues and eigenvectors of the matrix
3 −1
𝐀=( )
4 −2
The eigenvalues λ and eigenvectors 𝐱 satisfy the equation (𝐀 − λ𝐈)𝐱 = 0, or
3−λ −1 𝑥1 0
( ) (𝑥 ) = ( )
4 −2 − λ 2 0
The eigenvalues are the roots of the equation
3−λ −1
det(𝐀 − λ𝐈) = | | = λ2 − λ − 2 = 0
4 −2 − λ
Thus, eigenvalues are λ1 = 2 and λ2 = −1.
—To find the eigenvectors, we replace λ by each of the eigenvalues. For λ = 2 we
have
1 −1 𝑥1 0
( )( ) = ( )
4 −4 𝑥2 0
Hence, each row of this vector equation leads to the condition 𝑥1 − 𝑥2 = 0, so 𝑥1
and 𝑥2 are equal but their values is not determined. In other words, if 𝑥1 = 𝑐, then
𝑥2 = 𝑐 too, and the eigenvector 𝐱 (1) is
1
𝐱 (1) = 𝑐 ( ) , 𝑐≠0
1
Usually, we will drop the arbitrary constant 𝑐 when finding the eigenvectors; thus
we write
1
𝐱 (1) = ( )
1
We say that 𝐱 (1) is the eigenvector corresponding to the eigenvalue λ1 = 2. Any
nonzero multiple of 𝐱 (1) is also an eigenvector.
—Now, for λ = −1 we have,
4 −1 𝑥1 0
( ) (𝑥 ) = ( )
4 −1 2 0
Again, we obtain a single condition on 𝑥1 and 𝑥2 . 4𝑥1 − 𝑥2 = 0. Therefore, the
eigenvector corresponding to the eigenvalue λ2 = −1 is
1
𝐱 (2) = ( )
4
or any nonzero multiple of this vector.
7
Fall Semester Faculty of Aeronautics and
Review of Matrices MAT 201E
2023/2024 Astronautics
Now, recall that Eq. (5) generates a polynomial equation of degree 𝑛 in λ, so there
are 𝑛 number of eigenvalues λ1 , λ2 , … λn , some of which may be repeated. If a given
eigenvalue repeats 𝑚 times as a root of Eq. (5), that eigenvalue is said to have
algebraic multiplicity 𝑚.
A set of 𝑘 vectors 𝐱 (1) , 𝐱 (2) , … , 𝐱 (𝑘) is said to be linearly independent if there
exists a set of (complex) numbers 𝑐1 , 𝑐2 , … , 𝑐𝑘 , at least one of which is nonzero,
such that
𝑐1 𝐱 (1) + 𝑐2 𝐱 (2) + ⋯ +𝑐𝑘 𝐱 (𝑘) = 0. (6)
In other words, 𝐱 (1) , 𝐱 (2) , … , 𝐱 (𝑘) are linearly independent if there is a linear
relation among them. On the other hand, if the only set 𝑐1 , 𝑐2 , … , 𝑐𝑘 for which the
Eq. (6) is satisfied is 𝑐1 = 𝑐2 = ⋯ = 𝑐𝑘 = 0, then 𝐱 (1) , 𝐱 (2) , … , 𝐱 (𝑘) are said to be
linearly independent.
Consider now a set of 𝑛 vectors 𝐱 (1) , 𝐱 (2) , … , 𝐱 (𝑛) , each of which has 𝑛 components.
(𝑗)
Let 𝑥𝑖𝑗 = 𝑥𝑖 be the 𝑖th component of the vector 𝐱 (𝑗) , and let 𝐗 = 𝑥𝑖𝑗 . Then Eq. (6)
can be written as
(1) (𝑛)
x1 𝑐1 + ⋯ + x1 𝑐𝑛 𝑥11 𝑐1 + ⋯ + 𝑥1𝑛 𝑐𝑛
( ⋮ ⋮ )=( ⋮ ⋮ ) = 𝐗𝐜 = 0 (7)
(1)
x𝑛 𝑐1 + ⋯
(𝑛)
+ x𝑛 𝑐𝑛 𝑥𝑛1 𝑐1 + ⋯ + 𝑥𝑛𝑛 𝑐𝑛
If det 𝐗 ≠ 0, then the only solution of Eq. (7) is 𝐜 = 0, but if det 𝐗 = 0, there are
nonzero solutions. Thus, the set of vectors 𝐱 (1) , 𝐱 (2) , … , 𝐱 (𝑛) is linearly independent
if and only if det 𝐗 ≠ 0.
An important class of matrices, called self-adjoint or Hermitian matrices, are
those for which 𝐀∗ = 𝐀; that is 𝑎̅𝑖𝑗 = 𝑎𝑖𝑗 . Hermitian matrices have a subclass
named real symmetric matrices—that is, matrices that have real elements and for
which 𝐀𝐓 = 𝐀. The eigenvalues and eigenvectors of Hermitian matrices have the
following properties which are always useful,
I. All eigenvalues are real.
II. There always exists a full set of 𝑛 linearly independent eigenvectors,
regardless of the algebraic multiplicities of the eigenvalues.
III. If 𝐱 (1) and 𝐱 (2) are eigenvectors that correspond to different eigenvalues,
then (𝐱 (1) , 𝐱 (2) ) = 0. Thus, if all eigenvalues are simple, then the associated
eigenvectors form an orthogonal set of vectors.
IV. Corresponding to an eigenvalue of algebraic multiplicity 𝑚, it is possible to
choose 𝑚 eigenvectors that are mutually orthogonal. Thus, the full set of 𝑛
eigenvectors can always be chosen to be orthogonal as well as
independent.
