Professional Documents
Culture Documents
RowReduce
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
Submit
Our goal now is understand these solutions of , not just as computional algorithms, but in terms of vector
spaces associated with the coefficient matrix . All this is in preparation for developing solution techniques that
more sophisticated problems require. In fact, you've met some of these sophisticated problems already when
solving systems of linear differential equations. There are four important vector subspaces associated with each
coefficient matrix , each going under a variety of names differing from one text to another, and from
whichever concept is being emphasized by that text! The first vector subspace is
of spanned by the columns of .
Theorem 1: a matrix equation is consistent, i.e., has a solution, if and only if
Discussion of a model example illustrates well some features of all four vector subspaces.
Model Example 1:
in other words,
of . So we have to choose a linearly
independent subset of the columns of that still
spans . Look first at : while
Claim 1. are linearly independent: for if
, then
which in column form can be written as
where are standard basis vectors in .
Thus is a basis for . But Claim 2. for each in
is obtained from by row operations on
which 'mess up' the columns of .
But then
On the other hand, there is a invertible
matrix such that . Thus by matrix
column multiplication,
Hence is a basis for .
Soln Prob 2: the augmented equation associated Thus
with the equation is
But then Consequently,
so are free variables. Set .
Then In particular, is the span of two linearly
independent vectors in , hence a twodimensional
vector subspace of .
This last solution generalizes to any matrix :
1. Since , contains .
which shows that belongs to .
2. Fix arbitrary vectors in . Then
Thus is a subspace of for any
matrix .
which shows that belongs to .
Later when we get to eigenvectors we'll see why it's particularly important to know the dimension and eigenbasis
basis for the Null Space of a matrix. The Null Space enters in the nonhomogeneous case also.
Problem 3: find all solutions of the matrix Thus
equation when is Model Example 1 and
Solution: the augmented equation associated
with the equation is
Consequently, with
But then
,
,
In the case of square matrices the equivalent conditions can be sharpened:
is invertible ,
,
,
.
A close examination of the solutions to Problems 1 and 2 reveals some important results which we then collect in
a basic theorem:
Definition: the rank of a matrix is the dimension of the column
space of .
Fundamental Theorem of Linear Algebra: Part I. for an
matrix
Superposition Principle: the Null Space of an matrix is a
subspace of .
For convenience, it's worth collecting together basic interpretations of concepts developed in this section.