Basic Linear Algebra

Prof. Inder K. Rana

Department of Mathematics
Indian Institute of Technology- Bombay

Lecture – 10
Solvability of a Linear System, Linear Span, Basis I

(Refer Slide Time: 00:33)

Let us just recall, what we have done in the previous lecture, we have started looking at what is
called the rank of a matrix, so we defined for the matrix 𝐴 which is 𝑚 × 𝑛, look at its row echelon
form particularly you can also look at the reduced row echelon form and the rank of the matrix is
defined as the number of nonzero rows in that matrix or the number of pivots, both are same.

So, we had mention that if you take row echelon form of a matrix that may not be unique but the
number 𝑟 which is number of nonzero rows, its same in all row echelon forms, okay and the first
nonzero entry was called pivot, so number of pivots is also fixed, once you have row echelon form,
so the rank is defined to be the number of nonzero rows in the row echelon form, now reduced row
echelon form.

(Refer Slide Time: 01:52)

So and we stated the theorem of solving a system of equations in terms of this new terminology
rank, so it says the following, first of all, the solution set is non-empty that means the system is
consistent if and only if the rank of 𝐴 is equal to rank of the augmented matrix, the rank; 2 ranks
are same that means the pivot, it should not happened that there is no pivot in a row in A but there
is a pivot in the augmented matrix, right.

So that will give you an inconsistent equation 0 = nonzero thing, so existence is if only if the rank
of 𝐴 is same as the rank of augmented matrix that means, the number of nonzero rows in 𝐴 and
the number of nonzero rows in augmented matrix should be same, okay, once you have them in
row echelon form. So, one is a unique solution possible, 𝑛 is the number of variables, 𝑚 is some
m cross n, n is number of variables.

So, if there are n pivots exactly, right that means each variable will get a unique value, so the
system has unique solution if and only if the rank of 𝐴 = rank of [𝐴 | 𝑏] = 𝑛 that is consistency and
then there is a unique solution and non-unique means, actually there are infinite number of
solutions.

The number of the pivots is strictly less than the number of variables, okay, so then it says there
are infinite number of solution and we said that the difference 𝑛 – 𝑟, the rank, right those many
variables will get infinite values, their arbitrary, they are free to give any values, right, so you can
choose the remaining.

If it is less than n, there are infinite solutions and although infinite solutions can be obtained by
backward substitution, right, putting the non-pivotal variables arbitrary values and calculating the
pivotal variables in terms of rows arbitrary values, right.
(Refer Slide Time: 04:29)

So, let us write a; so the I think proof we have already gone through, the existence and; so let us
keep the proof, right.
(Refer Slide Time: 04:34)
(Refer Slide Time: 04:39)

Uniqueness also we gave gone through, so let us; so I am just summarising now, if the rank of 𝐴
is 𝑟, rank of [𝐴 | 𝑏] is 𝑟 + 1 that is one more column is there in augmented matrix, so possibility
is rank of 𝐴 = 𝑟 but the rank of the augmented matrix is 𝑟 + 1, then there is no solution, it is an
inconsistent solution, the solution set, right, the set of all solutions is an empty set. If both are 𝑟
and 𝑟 is strictly less than n, the number of variables, then there are infinitely many solutions
possible, right.

And if both are equal to n, the rank of 𝐴 is same as rank of augmented matrix, which is equal to 𝑛,
then the solution set will be a single point that means there is a unique solution, right, this cannot
happen, 𝑟 cannot be strictly bigger than 𝑛, right, okay the rank of the matrix cannot be bigger than
number of rows or number of columns, so this possibility does not arise, so these are the only 3
possibilities which can arise.
(Refer Slide Time: 05:44)
And as a consequence of this, there is a some simple application which we have already discussed,
a square matrix, 𝑛 × 𝑛 is invertible if only if 𝐴𝑋 = 𝑂 has only trivial solution, right, there should
not be any nontrivial solution for this, is it clear because 𝐴𝑋 = 𝑂, 0 is always a solution, right
and if 𝐴 is invertible that means only one solution possible, for 0 should be the only one, is it clear
to everybody.

𝐴𝑋 = 0, always has 0 as the solution, if we put 𝑋 = 0 that is the solution, right and when
invertible that means there is only one solution possible, right invertible means that was also
echelon to the row echelon form is identity matrix, right that was also same as saying the rank = 𝑛
and we also showed that this is same as saying the product, 𝐴 is a product of elementary matrices,
right, 𝐴 is invertible if and only if it is a product of elementary matrices, we showed that.

The row echelon form should be identity matrix, right, invertibility is same as saying that rank =
𝑛, number of nonzero rows must be same as a number of variables and that is same as a row, the
number of rows or number of columns in the square matrix and product of elementary matrices,
so there are various ways of saying when a matrix is invertible, right all are same but there is
different ways of saying the same thing, okay.
(Refer Slide Time: 07:57)
So, let us look at that the previous one said that system 𝐴𝑋 = 0 seems to be an important 𝐴 is to
consider, so such a system is called a homogenous system, so system of linear equations were the
right hand side that vector is a 𝑏 is 0 is called homogenous system, right, so given below, we want
to relate it with the; so, given any linear system; 𝐴𝑋 = 𝑏, if you put 𝑏 = 0, will get a
homogeneous system, right.

So, given a system of linear equations 𝐴𝑋 = 𝑏, if you put 𝑏 = 0 that is called the homogeneous
system associated with the given system of linear equations, right we are putting, so what is the
advantage of doing that and what is the use of it; so, let us; the observation is the solutions of the
general system are related with the solutions of the corresponding homogeneous system, in what
way, we will describe that.

So, first of all let us is note that the homogeneous system always have a solution, right, 𝐴𝑋 = 0,
so 0 is always the solution of that okay, so let us define all solution of the homogeneous system
has a set, so let us call that as null space given a matrix 𝐴, you are given a matrix m cross n, so
look at all possible, the vectors 𝑋, so I will want to get 𝐴𝑋 = 0, so what will be the order of the
vector 𝑋, you have to multiply 𝐴 with 𝑋, 𝐴 is 𝑚 × 𝑛.

So, vector 𝑥 has to be 𝑛 × 1, right so you are looking at all vectors 𝑋 ∈ ℝ𝑛 such that 𝐴𝑋 = 0, so
look at that side, so collection of all vectors which are killed by 𝐴, you can think it that way that
is called the null space of the matrix 𝐴, right and that is same as a solution space of the
homogeneous system 𝐴𝑋 = 0, right in terms of set theory, we call it as null space denoted by n
of 𝐴, right that is a solution space of the homogeneous system 𝐴𝑋 = 0 for a given matrix 𝐴.

So, for every matrix we are associating is space with it, a set with it, the null space which is the
solution set of the homogeneous system, okay. What is the advantage of that? So, the advantage is
look at, we want to solve the system 𝐴𝑋 = 𝑏, right, the theorem says if the solution set of 𝐴𝑋 =
𝑏 is same as; you find at least one solution of the system 𝐴𝑋 = 𝑏 and add to it all solutions of
homogeneous systems, you get all solutions of the system.

So, finding a solution for the non-homogeneous case or the general system is find at least one, find
all solutions of the homogeneous case, take 1 from there and add it to it, we will get all, one by
one pick and add, right, so the solution of the; solution space of 𝐴𝑋 = 𝑏 is same as 𝑋0 + 𝑋, where
𝑋 is in the null space that means 𝐴𝑋 = 0 and what is 𝑋0; some particular solution, if I were able
to find at least one solution of the system 𝐴𝑋 = 𝑏, add to it a solution of the homogeneous system,
you get a solution of the general system, right.

So, what it says is that the solution of the homogeneous system is important thing to consider using
that we can describe all solutions of the system 𝐴𝑋 = 𝑏 provided we can find the null space, the
solution space of homogeneous system plus one particular solution, right.
(Refer Slide Time: 12:21)
So, we will concentrate on this, so questions are; how to list all solutions of a homogeneous
system? So, the problem; we are reducing the problem to the homogeneous case in a sense and
what is the best economical way of telling what are all the solutions, you can say this is the
solutions, right but can I give a method of generating all the solutions of the homogeneous system,
right what is the most economical way?

So, let us observe something that is 𝑋1 , 𝑋2 are solutions of the homogeneous system, 𝐴𝑋 = 0,
then 𝛼 𝑋1 + 𝛽 𝑋2 also is a solution, if 𝑋1 and 𝑋2 are the solutions of the homogeneous system that
means 𝐴𝑋1 = 0, 𝐴𝑋2 = 0, then the claim is 𝛼 𝑋1 + 𝛽 𝑋2, where 𝛼 and 𝛽 are any scalars, is also
a solution of the homogeneous system because we multiply by 𝐴, then what is a product?

A multiplied by 𝛼 𝑋1 + 𝛽 𝑋2 that is same as 𝛼 𝐴𝑋1 + 𝛽 𝐴𝑋2 , why the distributive property of

multiplication of matrices, right but 𝐴𝑋1 = 0, 𝐴𝑋2 = 0, so, 𝛼 𝑋1 + 𝛽 𝑋2 of that that again also is
a solution of it, right.
(Refer Slide Time: 14:05)
You say that a subset 𝑉 of ℝ𝑛 , right is called a vector space if it has that property may be, take any
2 elements 𝑣, 𝑤 in 𝑉, take 2 any scalars 𝑎, 𝑏 and take the linear combination 𝑎𝑣 + 𝑏𝑤, that is
called a linear combination that also should belong to V, right.

So, if the set 𝑉 has this property, then we say it is a vector space.

So, null space is a subset of ℝ^𝑛 and it is a vector space.

(Refer Slide Time: 18:50)

(Refer Slide Time: 21:24)

So, visualising; so what is 𝐿, so that is {𝛼 𝑋0 | 𝛼 ∈ ℝ}, let us try to draw a picture of it, so what is
the picture; 𝑋0 is the vector, right, so I can draw that vector, okay, let us draw it from here, so this
is a vector, okay, so that is my vector 𝑋0, what is 𝛼 𝑋0, what I am doing? I am taking a fix vector
and only stretching it, right, so when I stretch it, its length will change only, right, so this will give
me, right, the vector could be anywhere here from here to here or here to here or here to depending
on whether it is positive or negative.

So, what is this; this is a line through the origin right in ℝ2 , is that okay, right, so this is a line
through the origin that is a vector space, okay.