Professional Documents
Culture Documents
net/publication/318444307
CITATIONS READS
0 15,301
1 author:
Ayman Badawi
American University of Sharjah,
114 PUBLICATIONS 3,023 CITATIONS
SEE PROFILE
All content following this page was uploaded by Ayman Badawi on 17 August 2017.
several variables. They are used system of linear equations (Cramer rule) and in
to define the characteristic finding the inverse of a matrix (Adjoint method).
polynomial of a matrix that is an
essential tool in eigenvalue
problems in linear algebra Introduction
Elementary An elementary matrix is a matrix
matrix which differs from the identity Graphs are very useful ways of presenting infor-
matrix (In) by one single mation about social networks. However, when
elementary row operation there are many actors and/or many kinds of rela-
Equivalent Two matrices are equivalent if tions, they can become so visually complicated
matrices each is obtained from the other that it is very difficult to see patterns. It is also
by applying a sequence of row possible to represent information about social net-
operations works in the form of matrices. Representing the
Identity In Let n 2 be a positive integer. information in this way also allows the application
matrix Then B = In is the square matrix, of mathematical and computer tools to summarize
n n, where bij = 1 if i = j and and find patterns. Social network analysts use
bij = 0 if i 6¼ j. If A is an n m matrices in a number of different ways. So, under-
matrix, then AIm = A and In standing a few basic things about matrices from
A = A. mathematics is necessary. For example, the sim-
Invertible A square matrix A, n n, is said plest and most common matrix is binary. That is, if
(nonsingular) to be invertible or nonsingular if a tie is present, a one is entered in a cell; if there is
matrix there exists a matrix n n no tie, a zero is entered. This kind of a matrix is the
denoted by A1 such that starting point for almost all network analysis and
AA1 = A1 A = In is called an “adjacency matrix” because it repre-
Matrix n m A matrix is a block consisting of sents who is next to or adjacent to whom in the
n row and m column. An entry in “social space” mapped by the relations that we
a matrix B located in the i th row have measured. The following is an example of a
and j th column of B is denoted binary matrix:
by bij 2 3
Symmetric A square matrix A, n n, is said 0 1 1
matrix and to be symmetric if AT = A, and it 40 0 15
skew- is called a skew symmetric if 0 1 0
symmetric AT = A
matrix Matrices and linear algebra are surely inseparable
Transpose of The transpose of a matrix A is subjects, and they are important “concepts”
a matrix denoted by AT such that aTij ¼ aji needed in many aspects of real life science. The
subject of linear algebra can be partially explained
Definition by the meaning of the two terms comprising the
title. We can understand “linear” to mean anything
In this entry, we describe all basic matrix opera- that is “straight” or “flat”. For example, in the
tions: Addition, subtraction, and multiplication. xy-plane we are accustomed to describing straight
We show the importance of matrices in studying lines as the set of solutions to an equation of the
system of linear equations (augmented matrix and form y ¼ mx þ b , where the slope m and the
row operations). We show different methods used y-intercept b are constants that together describe
in calculating determinant of a square matrix. We the line. Living in three dimensions, with coordi-
show the importance of determinant in solving nates described by triples (x, y, z), they can be
described as the set of solutions to equations of
the form ax þ by þ cz ¼ d, where a, b, c, d are
Matrix Algebra, Basics of 3
constants that together determine the plane. While The material presented here can be found in
we might describe planes as “flat”, lines in three every textbook on basic linear algebra. Since there
dimensions might be described as “straight”. are so many textbooks on basic linear algebra, and
From a multivariate calculus course, we recall we cannot list all of them, we refer to a few books
that lines are sets of points described by equations here. For example, Axler (1997), Bernstein
such as x = 3t 4, y = 7t + 2, z = 9t, where t is a (2005), Beezer (2012), Blyth and Robertson
parameter that can take on any value. (2002), Kaw (2011), Lang (1986), Lay (2003),
Another view of this notion of “flatness” is to Robbiano (2011), and Shores (2007.
recognize that the sets of points just described are
solutions to equations of a relatively simple form.
These equations involve addition and multiplica-
System of Linear Equations
tion only. Here are some examples of typical
equations:
Matrices play an important role in solving a sys-
tem of linear equations as we will see later on in
2x þ 3y 4z ¼ 134
this section. Let R be the set of all real numbers
x1 þ 5x2 x3 þ x4 þ x5 ¼ 0
and C be the set of all complex numbers. Then
9a 2b þ 7c þ 2d ¼ 7
Rn ¼ fða1 , a2 , . . . , an Þ=a1 , a2 , . . . , an Rg and
Cn ¼ fða1 , a2 , . . . , an Þ=a1 , a2 , . . . , an Cg . An
What we will not see in a linear algebra course are
element of Rn(C n) is called a point.
equations like:
A system of linear equations is a collection of
m equations with n variable x1, x2, x3, . . ., xn of the
xy þ 5yz ¼ 13x1 þ x32 =x4 x3 x4 x25 ¼ 0
form
cos ðabÞ þ logðc dÞ ¼ 2
Example 1 F ¼ fð2, 1Þg is the set of all solu- Let A be an m n matrix (i.e., A has m rows
tions to the system 2x1 þ x2 ¼ 5, x1 þ 2x2 ¼ 4 and n columns). Then each of the following is
(i.e., x1 ¼ 2, x2 ¼ 1 , and hence, the solution is called a row operation on A.
unique).
F ¼ fða1 , 2a1 þ 3Þ=a1 Rg is the set of all 1. Swap the locations of two rows.
solutions to the system 2x1 þ x2 ¼ 3, 4x1 þ 2. Multiply each entry of a single row by a non-
2x2 ¼ 6 (i.e., the system has infinitely many zero quantity.
solutions). 3. Multiply each entry of one row by some quan-
The system x1 þ 2x2 ¼ 0, 2x1 þ 4x2 ¼ 1 has tity and add these values to the entries in the
no solutions. same columns of a second row. Leave the first
A system of linear equations is called consis- row the same after this operation but replace
tent if it has a solution, and it is called inconsistent the second row by the new values.
if it has no solutions. A system of linear equations
as in (1) is called homogeneous if b1 ¼ b2 ¼ We will use the following notations to describe
¼ bm ¼ 0. the row operations stated above:
Theorem 2 Every homogeneous system of linear 1. Ri $ Rj : Swap the location of rows i and j.
equations is consistent (i.e., (0, 0,. . .,0) is always a 2. aRi: Multiply row i by the nonzero scalar a.
solution of such system). If a system of linear 3. aRi þ Rj ! Rj : Multiply row i by the scalar a
equations is consistent and it has more variables and add to row j, so that row j will change but
than equations, then the system has infinitely no change in row i.
many solutions. In particular, if a homogenous
system has more variables than equations, then it Two matrices, A, B, of the same size, say m n,
has infinitely many solutions. are said to be row-equivalent if and only if A is
Two systems of linear equations are equivalent obtained from B by applying a sequence of row
if their solution sets are equal. operations on B.
The following type of matrices is needed in
Theorem 3 If we apply one or two or all of the order to achieve our main goal and solve a system
following equation operations to a system of lin- of linear equations using augmented matrices.
ear equations as many times as we want, then the A matrix A, m n, is called in reduced row-
original system and the transformed system are echelon form if it meets all of the following
equivalent. conditions:
1. Swap the locations of two equations in the list 1. If there is a row where every entry is zero, then
of equations. this row lies below any other row that contains
2. Multiply each term of an equation by a nonzero a nonzero entry.
quantity. 2. The leftmost nonzero entry of a row is equal to 1.
3. Multiply each term of one equation by some 3. The leftmost nonzero entry of a row is the only
quantity and add these terms to a second equa- nonzero entry in its column.
tion, on both sides of the equality. Leave the 4. Consider any two different leftmost nonzero
first equation the same after this operation but entries, one located in row i, column j and the
replace the second equation by the new one. other located in row s, column t. If s > i, then
t > j.
In light of Theorem 3, one can view each
equation of a system of linear equations as a row A matrix A, m n, is said to be in row-echelon
of a matrix and each equation operation in Theo- form if and only if A satisfies conditions (1), (2),
rem 3 as a row operation on a matrix. Hence, we and (4) as above.
have the following well-known row operations.
Matrix Algebra, Basics of 5
2 3
0 1 3 0 0 independent variable, we understand that xk can
Example 2 The matrix A ¼ 4 0 0 0 1 0 5 take any real (complex) value.
0 0 0 0 1 Consider the following system:
is in reduced row-echelon form.
2 3
1 1 1 1 x1 þ 2x2 x3 þ 2x4 ¼ 1
A matrix A ¼ 4 0 0 1 4 5 is in row- 2x1 4x2 þ 3x3 4x4 ¼ 2
0 0 0 1 x1 2x2 þ x3 2x4 ¼ 1
echelon form but not in reduced row-
echelon form. Whose augmented matrix is equivalent to the
0 1 matrix
The matrix A ¼ is a neither in row-
1 0 2 3
echelon form nor in reduced row-echelon form. 1 2 0 2 j5
E ¼ 40 0 1 0 j4 5
Theorem 4 Let A be a matrix, m n. Then A is 0 0 0 0 j0
row-equivalent to a unique matrix, m n, in
reduced row-echelon form. in reduced row-echelon form, which corresponds
Consider the system in (1). The augmented to the system:
matrix of the system is
x1 þ 2x2 þ 2x4 ¼ 5
2 3 x3 ¼ 4
a11 a12 ... a1n jb1
6 a21 a22 ... a2n jb2 7 0¼0
6 7
4⋮ ⋮ ... ⋮ ⋮5
am1 am2 ... amn j bm No equation of this system has a form zero =
nonzero. Therefore, the original system is consis-
where a1i, a2i, . . ., ami are the coefficients of the tent. The dependent (leading) variables are x1 and
variable xi in the system (1). In view of Theorem x3. The independent (free) variables are x2 and x4.
3, we have the following result. Thus, x2, x4 can take any real (complex) values.
We write the leading variables in terms of the
Theorem 5 Let A be the augmented matrix of a dependent variables. Thus, we have:
given system of linear equations and let E be the
reduced row-echelon form of A. Then the solution x1 ¼ 5 2x2 2x4
set of the given system is equal to the solution set x3 ¼ 4
x2 , x4 RðCÞ
of the system that has E as its augmented matrix.
Hence, the solution set of the original system is
Theorem 6 Let A be the augmented matrix of a
given system of linear equations and let E be the
F ¼ fð5 2x2 2x4 , x2 , 4, x4 Þ=x2 , x4 RðCÞg:
reduced row-echelon form of A. Then the given
system is consistent if and only if none of the
equations that correspond to the matrix E has a
form zero = nonzero. Matrix Operations
Suppose A is the augmented matrix of a con-
sistent system of linear equations and let E be the Let Mnm be the set of all n m matrices with
reduced row-echelon form of A. Suppose j is the entries from R(C) for some positive integers n,m.
index of a column of B that contains the leading If A Mnm, then aij denotes the entry in the
1 for some row. Then the variable xj is said to be matrix A that is located in the i th row and the j th
dependent or leading. A variable that is not depen- column of A. If A Mnm, then we say that
dent is called independent or free. If xk is an size(A) = n m.
6 Matrix Algebra, Basics of
Theorem 7 (Addition, subtraction, and multipli- 3. The second row of D is Dr2 ¼ 3½ 1 2 3
cation by a scalar) Let A, B be two matrices and þ 0½ 0 5 2 þ 6 ½ 3 7 4 .
a R(C). Then A + B and A B is defined if and
only if size(A) = size(B). Furthermore, suppose Note that BA is undefined by Theorem 8. In
that size(A) = size(B), A + B = C, A B = D, fact, it is possible that for some matrices A, B that
and aA = F. Then cij = aij + bij, dij = aij bij, and AB and BA are defined but AB 6¼ BA.
fij = aaij.
Theorem 9 Let i be a positive integer and A be a
Example 3 matrix. Then Ai ¼ A A A (i times)
is defined if and only if A is a square matrix, i.e.,
3 4 1 2 0 1 A Mnn for some positive integer n.
Let A ¼ ,B ¼ , and
1 0 2 4 2 1
1 4 2 Theorem 10
a ¼ 2: Then A þ B ¼ C ¼ ,
3 2 3
5 4 0 1. Let a, b RðCÞ, A, B Mnm and let C Mmk.
AB¼D¼ , and
5 2 1 Then ðaA þ bBÞC ¼ aAC þ bBC.
6 8 2 2. Let a, b RðCÞ, A, B Mnm and let C Mkn.
2A ¼
2 0 4 Then ðaA þ bBÞC ¼ aAC þ bBC.
3. Let A Mnm , B Mmk , and C Mki . Then
For a matrix A Mnm , let Ari denote the i th ABC ¼ ðABÞC ¼ AðBCÞ.
row of A and let Aci denote the i th column of A.
Theorem 11 (Matrix-form of a system of linear
Theorem 8 (Matrix multiplication) Let A be an equations) Consider the system in (1).
2 3 2 3
m n matrix and B be a n k matrix. Then the a11 a12 a1n x1
matrix multiplication AB is defined if and only if 6 a21 a22 a2n 7 6 x2 7
C¼6 4⋮
7, X ¼ 6 7, and
n = n. Furthermore, suppose that n = n and let ⋮ ⋮ ⋮5 4⋮5
AB = D. Then the following statements hold: a a amn xn
2 m13 m2
1. Size(D) = m k and d i, j ¼ ai, 1 b1, j þ ai, 2 b2, j b1
þ þ ai, n bn, j [dot product]. 6 b2 7
B¼6 7
4 ⋮ 5. Then (in view of Theorem 8(2)), the
2. Dci ¼ b1i Ac1 þ b2i Ac2 þ þ bni Acn [linear
combination of the columns of A]. bm
3. Dri ¼ ai1 Br1 þ ai2 Br2 þ þ ain Brn [linear matrix-form of the system in (1) is CX = B, where
combination of the rows of B]. C is called the coefficient matrix of the system,
X is called the variables-column of the system, and
1 2 5 B is called the constant column of the system.
Example 4 Let A¼ and
3 0 6
2 3 Theorem 12 Let CX = B be the matrix-form of a
1 2 3
B¼ 0 4 5 5
2 . Then AB is defined by given system of linear equations with m equations
3 7 4 and n variables (hence, size(C) = m n, size(X) =
Theorem 8. Let AB = D. Then n 1, and size(B) = m 1). Then the given
system is consistent if and only if there are some
1. Size(D) = 2 3 and d23 ¼ ð 3Þð3Þ þ ð0Þð2Þ real (complex) numbers, r1, r2,. . ., rn, such that
þ ð6Þð4Þ. B ¼ r 1 Cc1 þ r 2 Cc2 þ þ r n Ccn is a linear com-
bination of the columns of C.
1
2. The second column of D is Dc2 ¼ 2
3
Example 5 Consider the system:
2 5
þ5 þ7 .
0 6
Matrix Algebra, Basics of 7
2 3
3þ1 4 5 2 1 2 1
detðAÞ ¼ ð1Þ 3 det ¼ 69 6 2 1 2 67
3 2 Example 10 Let A ¼ 6 4 4
7. We
2 5 15
Let A Mnn . If aij = 0 whenever i > j, then 2 1 2 3
A is called an upper triangular matrix. If aij = 0 when- use Theorem 10 in order to calculate det(A).
ever i < j, then A is called a lower triangular matrix.
If aij = 0 whenever i 6¼ j, then A is called a AR1 þ R2 ! R2 , 2R1 þ R3 ! R3 , R1 þ R4 ! R4
diagonal matrix. If A is upper triangular or lower
triangular or diagonal, then A is said to be a 2 3
2 1 2 1
triangular matrix. 60 2 4 7 7
B¼6
40
7
0 1 1 5
Theorem 24 Let A Mnn be a triangular matrix. 0 0 0 2
Then detðAÞ ¼ a11 a22 : :an1n1 ann .
Since B is obtained from A by applying row
Theorem 25 operation number three exactly three times on
A and row operation number three has no effect
1. Let A Mnn be an invertible upper triangular on det(A), we conclude that det(A) = det(B) by
matrix. Then A1 is an upper triangular matrix. Theorem 28(3). Hence, det(A) = det(B) = 8 by
2. Let A Mnn be an invertible lower triangular Theorem 25.
matrix. Then A1 is a lower triangular matrix.
3. Let A Mnn be an invertible diagonal matrix. Example 11 Let A M44 and suppose that
Then A1 is a diagonal matrix. 2 3
2 2 2 2
6 0 1 1 4 7
Theorem 26 Let A Mnn . If one of the follow- A R1 $ R2 B 2R3 C ¼ 6 7
4 0 0 2 3 5. By
ing statements hold, then det(A) = 0. 0 0 0 4
1. Two rows or two columns of A are identical. Theorem 25, det(C) = 16. By Theorem 28(1),
2. One row of A is a multiple of another row of A, 16 = det(C) = 2 det(B), and hence det(B) = 8. By
or one column of A is a multiple of another Theorem 28(2), 8 = det(B) = det(A), and hence
column of A. det(A) = 8.
3. One row or one column of A is entirely zeros.
Theorem 29 (Characterizing invertible matrices
Theorem 27 (The effect of row operations on in terms of determinant) Let A Mnn . Then A is
determinant) Let A Mnn . Then, invertible (nonsingular) if and only if det(A) 6¼ 0.
1. Suppose that we applied row operation number
one exactly once on A : A Ri $ Rj B. Then Theorem 30 (Characterizing consistent and
det(B) = det(A). inconsistent systems of linear equations in terms
2. Suppose that we applied row operation number of determinant) Let CX = B be the matrix-form of
two exactly once on A : A aRi B . a system of linear equations with n equations and
3. Suppose that we applied row operation number n variables (i.e., size (C) = n n). Then:
three exactly once on A : A aRi þ Rj ! Rj B .
Then det(B) = det(A). 1. The system has a unique solution if and only if
det(C) 6¼ 0.
Theorem 28 (Most used method to find a deter- 2. Assume that the given system is consistent.
minant) Let A Mnn . It is always recommended Then the system has infinitely many solutions
that we apply row operations on A in order to if and only if det(C) = 0.
transform A to a triangular matrix, then we use
Theorem 28 and Theorem 25 to calculate det(A).
10 Matrix Algebra, Basics of
Lang S (1986) Introduction to linear algebra. Springer, Robbiano L (2011) Linear algebra for everyone. Springer,
Berlin Berlin
Lay DC (2003) Linear algebra and its applications, 3rd edn. Shores TS (2007) Applied linear algebra and matrix anal-
Pearson, Toronto ysis. Springer, Berlin