15.0pt 31pt 588pt     6  Linear part of a Function Increment

MATH1060
Introductory Linear Algebra

Vladimir V. Kisil

Abstract

This is lecture notes for the course MATH1060 Introductory Linear Algebra at School of Mathematics of University of Leeds. They are based on the notes of Dr Reg B. J. T. Allenby used in the previous years. However all misprints, omissions, and errors are only my responsibility. Please let me know if you find any.
The notes are available also for download in PDF and PostScript formats.
The suggested textbooks are [1,2]

Booklist:

[1]
R. B. J. T. Allenby. Linear Algebra. Edward Arnold, 1995.
[2]
H. Anton and C. Rorres. Elementary Linear Algebra: applications version. Wiley, sixth edition, 1991.

Useful Information

Notations

To avoid possible confusion, it is usual to insert commas between the components of a vector or a 1×n matrix, thus: [5,2], [1,11,1,111].

1  General Systems of Linear Equations

1.1  Introduction

The subject of Linear Algebra is based on the study of systems of simultaneous linear equations. As far as we are concerned there is one basic technique - that of reducing a matrix to echelon form. I will not assume any prior knowledge of such reduction nor even of matrices. The subject has ramifications in many areas of science, engineering etc. We shall look at only a minuscule part of it. We shall have no time for "real" applications.
Why the name Algebra , by the way?
Very often, in dealing with "real life" problems we find it easier-or even necessary! -to simplify the problem so that it becomes mathematically tractable. This often means "linearising" it so that it reduces to a system of (simultaneous) linear equations.
We shall deal with specific systems of linear equations in a moment.

1.2  The different possibilities

There are essentially three different possibilities which can arise when we solve systems of linear equations. These may be illustrated by the following example each part of which involves two equations in two unknowns. Since there are only two unknowns it is more convenient to label them x and y − instead of x1 and x2.
Example 1
  1. The first case illustrated by the system:

    2x+5y
    =
    3
    (α)
    3x−2y
    =
    14
    (β)
    To solve this system eliminate x from equation (β) by replacing (β) by 2· (β)−3·(α), that is
    (6x−4y)−(6x+15y)=28−9.

    2x+5y
    =
    3
    (α)
    0x−19y
    =
    19
    (γ)
    So the given pair of equations are changed to (α)-(γ). Equation (γ) shows that y=−1 and then (α) shows that x=4. This system is consistent and has the unique solution.
  2. The second case:

    6.8x+10.2y
    =
    2.72
    (α)
    7.8x+11.7y
    =
    3.11
    (β)
    Replace here (β) by 6.8·(β)−7.8·(α) gives the following system:
    6.8x+10.2y
    =
    2.72
    (α)
    0x+0y
    =
    0.068
    (γ)
    Then (γ) shows that there can be no such x and y, i.e. no solution. This means that system is inconsistent.
  3. The third case:

    6.8x+10.2y
    =
    2.72
    (α)
    7.8x+11.7y
    =
    3.12
    (β)
    Replace here (β) again by 6.8·(β)−7.8·(α) gives the following system:
    6.8x+10.2y
    =
    2.72
    (α)
    0x+0y
    =
    0
    (γ)
    Here (γ) imposes no restriction on possible values of x and y, so the given system of equations reduces to the single equation (α). Taking y to be any real number you wish, say y=c, then (α) determines a corresponding value of x, namely
    x= 2.72−10.2·c

    6.8
    .
    Thus the given system is consistent and has infinitely many solutions.
This three cases could be seen geometrically if we do three drawings:
Figure 1: Three cases of linear systems considered in Example 1.1.
(i) Lines are in generic position and intersect in one point-the unique solution.
(ii) Lines are parallel and distinct-there is no solution.
(iii) Lines coinside-all points are solutions.
When there are more than two equations we try to eliminate even more unknowns. A typical example would proceed as follows:
Example 2 (i) We get, successfully, the solution of the following system:
3x+y−z
=
2
(α)
x+y+z
=
2
(β)
x+2y+3z
=
5
(γ)
      
3x+y−z
=
2
(α)
0x−2y−4z
=
−4
(δ)=(α)−3(β)
0x+y+2z
=
3
(ε)=(γ)−(β)

3x+y−z
=
2
(α)
0x+y+2z
=
2
(ζ)= (δ)

2
0x+y+2z
=
3
(ε)
      
3x+y−z
=
2
(α)
0x+y+2z
=
2
(ζ)
0x+0y+0z
=
1
(η)=(ε)−(ζ)
The last equation η shows that the given system of equations has no solutions. Could you imagine it graphically in a space?
(ii) Let consider the same system but change only the very last number:
3x+y−z
=
2
(α)
x+y+z
=
2
(β)
x+2y+3z
=
4
(γ)
      
3x+y−z
=
2
(α)
0x−2y−4z
=
−4
(δ)=(α)−3(β)
0x+y+2z
=
2
(ε) = (γ)−(β)

3x+y−z
=
2
(α)
0x+y+2z
=
2
(ζ)= (δ)

2
0x+y+2z
=
2
(ε)
      
3x+y−z
=
2
(α)
0x+y+2z
=
2
(ζ)
0x+0y+0z
=
0
(η)=(ε)−(ζ)
This time equation (η) places no restrictions whatsoever on x,y and z and so can be ignored. Equation (ζ) tells us that if we take z to have the "arbitrary" value c, say, then y must take the value 2 − 2z = 2 − 2c and then (β) tells us that x must take the value 2 − y − z = 2 − (2 − 2c) − c = c. That is, the general solution of the given system of equations is: x = c, y = 2−2c, z = c, with c being any real number. So solutions include, for example, (x,y,z) = (1,0,1), (3,−4,3), (−7π,2−14π,−7π),…. The method of determining y from z and then x from y and z is the method of back substitution.
The general system of m linear equations in n "unknowns" takes the form:
a11x1
+
a12x2
+
+
a1j xj
+
+
a1nxn
=
b1
a21x1
+
a22x2
+
+
a2j xj
+
+
a2nxn
=
b2
:
:
:
:
:
:
:
ai1x1
+
ai2x2
+
+
aij xj
+
+
ainxn
=
bi
:
:
:
:
:
:
:
am1x1
+
am2x2
+
+
amj xj
+
+
amnxn
=
bm
(L)
Remark 3
  1. The aij and bi are given real numbers and the PROBLEM is to find all n-tuples (c1,c2,…,cj,…,cn ) of real numbers such that when the c1, c2, ..., cn are substituted for the x1, x2, ..., xn, each of the equalities in (L) is satisfied. Each such n-tuple is called a solution of (the system) (L).
  2. If b1 = b2 = … = bn = 0 we say that the system (L) is homogeneous.
  3. Notice the useful double suffix notation in which the symbol aij denotes the coefficient of xj in the i-th equation.
  4. In this module the aij and the bj will always be real numbers.
  5. All the equations are linear. That is, in each term aij xj, each xj occurs to the power exactly 1. (E.g.: no √{xj} nor products such as xj 2 xk are allowed.)
  6. It is not assumed that the number of equations is equal to the number of "unknowns".

1.3  Introduction of Matrices

Consider the system of equations
2u + 5v
=
3
3u−2v
=
14
(cf. Example 1.1.1). We easily obtain the answer u = 4, v = −1 (cf. x = 4, y = −1 in Example 1.1.1).
This shows that it is not important which letters are used for the unknowns.
The important facts are what values the m×n coefficients aij and the m coefficients bj have. Thus we can abbreviate the equations of Example 1.1.1 to the arrays  
2
5
3
3
−2
14
  and  
2
5
3
0
−19
−19
 and those in Example 1.1.2 to the arrays  
6.8
10.2
2.72
7.8
11.7
3.11
 and  
6.8
10.2
2.72
0
0
0.068
 correspondingly.
Any such (rectangular) array (usually enclosed in brackets instead of a box) is called a matrix. More formally we give the following definition:
Definition 4 An array A of m×n numbers arranged in m rows and n columns is called an m by n matrix (written "m×n matrix").
Example 5 (
π
−[99/23]
8.1
0
e2
−700
) is a 2×3 matrix with a1,2= −[99/23] and a2,1=0. What are a2,3 and a3,2?

1.4  Linear Equations and Matrices

With the system of equations (L) as given above we associate two matrices:


File translated from TEX by TTH, version 3.60.
On 16 Feb 2005, 14:17.