Mathematics for Data Science 2

Professor Sarang Sane

Department of Mathematics
Indian Institute of Technology Madras
Lecture 06
Determinants (Part 3)
Hello, and welcome to the maths 2 component of the online BSc course on data science. Today’s
video is a continuation of the determinants topic that we have been studying. So, this is part three
of that same topic.

(Refer Slide Time: 00:32)

So, let us recall quickly what we did in the previous so far in this topic. So, we have defined for a
1 × 1 matrix, that the determinant of the matrix a is just the number a. And then we have seen
that for an n n matrix, the determinant is defined inductively. So, we did the definition explicitly
for 2 × 2 and 3 × 3.

And then I showed by for 4 Object 8

4 how I can use a 3 Object 10
3 definition and then you we said that, in
general, we use the minors or co-factors, corresponding to the first row in order to define the
determinant. So, that means if so the idea is if you know how to determine to define the
determinant for (n-1) Object 12
(n-1) matrix, then you can use this procedure to get it for the n Object 14
n
matrix.

So, just to recall, the (i, j) th minor is the determinant of the submatrix formed by deleting the i th
row and jth columns, so you have n rows and n columns, you delete the i th row and the jth column.
And then you are left with an (n -1) × (n – 1) matrix. And then you take the determinant of that.
And the (i, j)th cofactor is minus (-1)i+j times the (i, j)th minor. And just to recall again, this was
the final thing that we did in the previous video, we have seen that the determinant of a is (-1) 1+j
a1j M1j.

So, this was given by what I called expansion along the first row and then you can take this (-
1)1+j, inside along with M1j and then you can replace it by C 1j that is why we define the cofactors
where we can get rid of the science coming here. So, this is what we have studied so far. So, why
did we study the (i, j)th minor or the (i, j)th cofactor, because it is not occurred so far. So, that is
one of the things we will see in this video and we will also see some other properties of the
determinant.

(Refer Slide Time: 02:51)

So, why did we define the (i, j)th minor or the (i, j)th cofactor? So, the reason is because we had a
formula for the determinant in fact the definition in terms of expanding for any row or any sorry
with respect to the first row. And now what we are going to see is that you can actually do this
with any row or any column.

And there is nothing very special about the first row. So, this is called expansion along any row
or column. So, here is what happens when we do expansion along the i th row. So, in that case we
can write the expression for the determinant as the sum of the (i, j) th entry of a times the (i, j) th
minor multiplied by (-1)i+j.
So, remember, when i was 1 that was the definition. So, when i was 1, we had (-1) 1+j a1j M1j. So,
now what we are saying is we can replace that 1 by i and the formula works equally well. So, the
expansion along the first row was the definition that we are going to use and what we are saying
is, it does not matter whether you use 1 there or any other row, you can expand along any other
row.

And the fact is, you can even use a column to do the same thing. So, you can expand along any
column. So, here, notice that the j was what was varying and the i was fixed. So, this is
expansion along the ith row and this is the expansion. So, now the j is fixed and the i is varying.
Here i is varying and j is fixed.

So, that is expansion along the j th column. So, maybe let us do an explicit expression, let us write
down one explicit expression here. So, suppose I have a 3 × 3 matrix, and I want to expand along
the second row. So, what is this term telling me? So, this is expansion along second row.

So, this is saying what you have to do is, you have to take (-1) to the power. So, now i is 2 and
now j varies, so minus (-1)2+1 times a21 times the (2, 1)th cofactor sorry the (2, 1)th minor plus (-
1)2+2, a22 times the (2,2)th minor (-1)2+3 a23 times the (2,3)th minor. So that is the expression.

So, if you write it out, concretely, this is -a 21 times determinant of so you have to drop the second

a12 a13
row and first column. So, you get det [ a31 a33 ] plus a22 times
Object 22
minus a23 times det

. So, I hope it is clear what we mean by expanding along the i th row this is an example
Object 24

of how to do this, you can do this for an n Object 26

n matrix also. It is correct.

The proof is slightly involved I mean, it is not hard, but it involves manipulation of a lot of
expressions. So, if you like your algebra, I will suggest you go ahead. And let me do maybe a
similar example for expansion along the second row sorry second column. So, this is expansion
along second column.

So, what this is saying is that this is(-1)1+2 a12 times determinant of plus (-1)2+2 a22 times
Object 28

similar determinant, I will encourage you to write this down plus (-1) 2+3 sorry this was what the
second column, so my bad, this is a 32 times the determinant of the corresponding matrix you
obtain namely so that is a minor.

So, this is what we mean by expansion along any row or column. And it is a fact that the
determinant can be computed in any of these ways. And it remains the same, so there is no
nothing special about the first row of the matrix. That is how we define it. So, now, that we have
seen that the determinant can be computed in terms of any row or column meaning by expanding
along any row or column, we see that in particular, there is nothing special about the first row.

And we see that the expression that we are getting, there is some kind of symmetry about them.
But of course, with some negative signs thrown in and so on. So, keeping that in mind, let us,
now look at some properties of the determinant. So, we already saw in the first video, that the
determinant of a product is the product of determinants and we could do various operations on
matrices and the determinant either remain the same or change sign. So, we need all of those 2 Object 30

2 and 3 Object 32
3 matrices. So, let us see now what happens for the general case. So, I will make these
statements but I will not really prove too many of these.

(Refer Slide Time: 10:05)

So, the first property that we want to look at is the product. So, the same thing that was earlier
that we saw earlier still holds, namely that a determinant of a product is the product of
determinants. So, what we mean by that is determinant of (AB) is determinant of A times
determinant of B.
So, we explicitly check this, when they were of size 2 × 2 and I said you can try to check this for
3 Object 36
3. If you did, you must have seen that the expression starts becoming rather complicated, it is
not difficult, but you have to be careful with your algebra. And so you can imagine that for 3 Object 38

3, sorry, beyond 3 Object 40

3, if you work it out in terms of the explicit expressions, it going to be
slightly hard.

So, I will not, we will not really get into the proof of the statement. But we have seen it for 2 Object 42
2,
and hopefully we have taken for 3 Object 44
3. So, this is believable. So, let us use this instead and
derive some interesting formulae. So, the first thing that we can do is we can ask, if you take the
power of a matrix of a square matrix of course, namely you do A 2 A3 A4, An in general, then what
happens to the determinant.

And what this identity tells you is that you can take this power out, so determinant of (A n) is the
det(A)n. The second thing that you get out of this identity is if you take the determinant of the
inverse, then that is just the reciprocal of the determinant of A. So, I can take this more fancy
language as det(A)-1.

So, the proof of this is exactly the same as we did for the 2 Object 46
2 and 3 Object 48
3 case, I will suggest that
you go back and look at it. Another very useful formula is that when you get determinant of P
inverse AP, so you have a matrix P, you take its inverse. So, A and P are both of size square
matrices of size n by n, and P has an inverse.

So, take determinant of P-1AP, then you can use this formula and what we derive just above to
get that the determinant is the same as the determinant of A. And then another thing to note very
easy fact. Remember that we saw that the product of two matrices may not be even defined, if
you do not, if you take the opposite order and even if it is, they may be of completely different
size and even if they are of the same size, AB and BA may not be the same.

But the determinant must be the same because the determinant is the product of determinants.
This is of course remember if A and B are of square matrices of size n Object 50
n and the final identity
maybe one can write down. So, determinant of ATA. So, maybe before I even do this, I would
ask, what is determinant of A transpose? So, determinant of A transpose. So, now given that, we
can expand along any row or column, so for A transpose instead of expanding along the first
row, we expand along the first column.
So expand along the first column, so I say expand along first column. So, I strongly suggest you
do this and if you do that, the expression you are going to get will allow you to conclude that this
is determinant of A. So, I will leave this to you so along with so I will have I should use this
through this plus what I will call induction.

Namely, assume that this formula is correct for (n – 1) Object 52

(n – 1) matrices, that the transpose of a
matrix has the same determinant, then expand along the first column and in the minors,
remember there are determinants of (n – 1) Object 54
(n – 1) matrices. So, think of those as transpose of
something, and you will see that you can equate these two expressions. So, once we know this
determinant of A transpose is determinant of A, so determinant of ATA is det(A)2, this might not
be very seem very interesting right now, but maybe at some later point, you might find
expressions of this type coming up.

(Refer Slide Time: 16:07)

So, this long list of interesting observations, let us go on to another property. So, again, we have
check this property for 2 Object 56
2 and 3 × 3 matrices or other I have checked it for 2 × 2 and I hope
you have solved 3 × 3. So, switching two rows or columns changes the sign of the determinant.
So, what do we mean by switching two rows or columns, let me recall on that was, so if A is like
this, this is the ith row, this is the jth row.
 ai 1 ai 2−−−a¿
So, this is [ ]
a j 1 a j 2−−−a jn
. So, we will switch these two rows to get this new matrix A . So,

~
now it is ith row will be aj1, aj2, ajn this is ith row of A and the jth row will be ai1, ai2 up to ain this is
~
the jth row. So, we have switched these two column rows, and everything else in A is exactly the
~
same as it was for A, all other entries are exactly the same. So, then determinant of A is minus
determinant of A, how do we get this? So, the again the idea is to expand along the i th row and
then use induction. So, expand along ith row and use induction. So, if you do that, you will see a
minus sign coming up.

(Refer Slide Time: 18:30)

Let us, move on to the next property that we want to discuss, I should have said in the previous
slide that similarly, if you interchange columns you get the same result. So, adding multiples of a
row to another row leaves the determinant unchanged. So, what does this mean? So, here is A,
~
which was like in the previous slide and what is A . So, let us say I add t times the jth row to the ith

ai 1 +t a j 1 a i2 +t a j 2−−−a ℑ +t a jn
row. So, then here I get the [ aj 1 a j2 −−−a jn ] and the jth row remains the same.

So, the only thing that has changed is the is the i th row and what happens in that case, so the
~
claim is the determinant of A is the same as the determinant of A. So, again, one has to prove
this. So, the proof is so one way of proving this is to expand along the i th row and then you will
get two different expressions one as this determinant away and the other is determinant of a
matrix, which has two rows which are the same and then one will have to prove that is 0. So, that
is how you get this.

So, one can do this also for columns. So, for columns, let us say if I want to add the k th column to
the lth, sorry t times lth column to the kth column, then we will get we want a 1k + ta11, a2k + ta21 and
so on ank + t anl this is the new kth column and the same thing will have to work, the determinant
~
of ~
A is determinant of A. So, adding multiples of a columns to another column leaves the
determinant unchanged.

(Refer Slide Time: 21:19)

So, then another property. So, again, we have seen this for 2 2 and I leave it to you to check for
3 3, if you multiply a particular row by a constant t, then the determinant gets multiplied by t.
So, the same thing holds for columns, if you multiply a column by t, then the determinant gets
multiplied by t. So, what is this thing, so you have your matrix A and let us say in Object 80
you
multiply the ith row by t.

So, we have [t ai 1 t ai 2−−−t a¿ ] everything else is the same. So, then what you do is you expand
using the ith row so, if you expand using ith row you get t times ai1, maybe for that there is a minus
~
sign which I should have included, anyways sign I forget. So, determinant of A , so expansion

n
along the ith row gives us ∑ (−1)i+ j . So, maybe let me do this proof so that you get an idea of
j=1

how to prove such things.

So, you can take t out and so you get ¿ . So, now so you have to be very careful here because
when you write down the minus sign with respect to a particular matrix, so let me put this tilde
~
here. So, this is the (i,j)th of the matrix A but if you delete ith row and whatever column then the
remaining matrix remember it may have the same matrix. So, Object 92
for this particular row i is the
same as Mij.

So, this is just t and now, this is exactly the expression for the determinant of A. So, we have
Object 94

actually proved this. So, what is upshot? The upshot is that if you multiply a particular row by t
then in the determinant it comes out, the same thing holds for columns. So, I will suggest that
you check this so there you will have to expand along columns. So, I will also put in a warning
here because this is something that often students make mistakes.

If you multiply the entire matrix by a constant, remember that we have studied multiplying
matrices by scalars. So, scalar multiplication of matrices that means you multiply each entry of
that matrix by that scalar. So, if you do that, and this is an n × n matrix, so, just to keep track of
that this is an n n matrix, then this is like multiplying not just one particular row or one
particular column by t, but every row or every column by t.

So, you will pick up a t in the determinant from each row. So, what you will get this t n det( A).
What is this n? This n is exactly the size of A, the number of rows or number of columns in A.
So, t n det ( A), you will not get tdet ( A). So, be careful of this property, it says you multiply a row
by a constant t, then that t comes out in the determinant. If you multiply the entire matrix by t,
then t n comes out of the determinant. This is something students often make mistakes, so be
careful with this.

(Refer Slide Time: 25:56)

So finally, let me end with some computational tips. So, if you have a zero row or a zero column
in a matrix, then the determinant is 0. How do you get this? We expand along that particular row
or column. The determinant of a matrix in which one row or column is a linear combination of
the other rows, respectively columns is 0. Why is that? Because you can remember, we can add
multiples of a row to another row and the determinant remains unchanged.

So, you keep adding the correct expressions of the other rows to that particular row and you
make it 0. So, once you make it 0, remember, the determinant remained unchanged and then you
have a zero row or column. And that is how you get the determinant to be 0. So, the second
property we will be using later on. So, we will come back to this if you did not understand this.

The third one is what we just saw scalar multiplication of a row by a constant t multiplies the
determinant by t. And the fourth is a general sort of tip, while computing the determinant, you
can choose to compute it using whatever row or column is most convenient. For example, as we
saw in the first thing here, we used that row or column which is 0.
So, this is a general statement or it may be of use so what have we done in this video, we have
seen how to compute determinants. First of all, we define determinants for square matrices and
then we saw that you can compute them using expansion along any row or column in terms of
minors and then we use that to look at some other properties. So, most importantly, determinant
of a product is product of determinants. So, keep this in mind.

And then we also saw that if you do some particular operations, namely adding two rows or
adding a multiple of one row to another, or multiplying a particular row by a constant, or
swapping rows, we saw how the determinant behaves. So, that is it for now. In the next video,
we will see how to use determinants to compute solutions to systems of linear equations.