WEBVTT

00:00:05.712 --> 00:00:07.090
BEN HARRIS: Hi.

00:00:07.090 --> 00:00:10.250
Today we're going to do a
problem about similar matrices.

00:00:10.250 --> 00:00:12.440
The problem is right here.

00:00:12.440 --> 00:00:15.870
So the question asks, which
of the following statements

00:00:15.870 --> 00:00:19.670
are true, and it asks
you to explain why.

00:00:19.670 --> 00:00:24.400
The first statement is: if A
and B are similar matrices, then

00:00:24.400 --> 00:00:28.760
2 A cubed plus A minus 3
times the identity is similar

00:00:28.760 --> 00:00:34.350
to 2 times B cubed plus B
minus 3 times the identity.

00:00:34.350 --> 00:00:36.970
The second question
asks: if A and B

00:00:36.970 --> 00:00:41.470
are 3 by 3 matrices with
eigenvalues 1, 0, and -1,

00:00:41.470 --> 00:00:43.780
then they're similar.

00:00:43.780 --> 00:00:49.070
And the third part asks you
whether these two J matrices

00:00:49.070 --> 00:00:50.460
are similar.

00:00:50.460 --> 00:00:51.160
OK.

00:00:51.160 --> 00:00:55.700
I'll give you a moment to hit
pause and try it on your own.

00:00:55.700 --> 00:00:58.475
I'll be back in just a moment,
and we can do it together.

00:01:09.300 --> 00:01:10.520
And we're back.

00:01:15.560 --> 00:01:22.280
Let's start with part
A. Part A is true.

00:01:25.130 --> 00:01:26.620
Why?

00:01:26.620 --> 00:01:31.490
Well, what does it mean
for A and B to be similar?

00:01:31.490 --> 00:01:37.240
Right, we know there's
some matrix M, such

00:01:37.240 --> 00:01:40.850
that-- that's what
"st" means-- when

00:01:40.850 --> 00:01:46.320
I multiply A on the left by
M and the right by M inverse,

00:01:46.320 --> 00:01:48.380
I get B.

00:01:48.380 --> 00:01:49.560
OK.

00:01:49.560 --> 00:01:55.710
So let's take that same
matrix M and multiply it

00:01:55.710 --> 00:02:01.840
on the left and the right of
2 A cubed plus A minus 3 times

00:02:01.840 --> 00:02:03.075
the identity.

00:02:05.830 --> 00:02:06.630
OK.

00:02:06.630 --> 00:02:08.720
What do we get here?

00:02:08.720 --> 00:02:14.190
Well, the point is what M
times A cubed times M inverse,

00:02:14.190 --> 00:02:19.515
we can just write that
as three M A M inverses.

00:02:24.190 --> 00:02:32.910
Similarly, we have an M*A
M inverse and an M times

00:02:32.910 --> 00:02:35.980
the identity times M inverse.

00:02:35.980 --> 00:02:38.270
Good.

00:02:38.270 --> 00:02:41.280
Remember that M times
A times M inverse is B.

00:02:41.280 --> 00:02:50.250
So we just get 2 B cubed--
three B's-- plus B minus-- well,

00:02:50.250 --> 00:02:53.390
M times the identity is just
M. So we just get the identity

00:02:53.390 --> 00:02:57.872
back, and we have 3
times the identity.

00:02:57.872 --> 00:02:58.760
Good.

00:02:58.760 --> 00:03:02.140
And this is a general remark,
that if you have matrices

00:03:02.140 --> 00:03:05.600
A and B that are similar,
then any polynomials

00:03:05.600 --> 00:03:09.080
in these matrices A
and B will be similar.

00:03:09.080 --> 00:03:13.090
It's the exact
same justification.

00:03:13.090 --> 00:03:15.360
OK.

00:03:15.360 --> 00:03:19.820
Now let's go on to
part B. So now A and B

00:03:19.820 --> 00:03:24.160
are 3 by 3 similar matrices
with the same eigenvalues.

00:03:24.160 --> 00:03:28.690
And their eigenvalues
are distinct.

00:03:28.690 --> 00:03:31.085
So it turns out that
b is true as well.

00:03:35.830 --> 00:03:38.940
And why is that?

00:03:38.940 --> 00:03:43.020
A matrix with distinct
eigenvalues is diagonalizable.

00:03:43.020 --> 00:03:55.600
So we can write A as S lambda
S inverse, where lambda is just

00:03:55.600 --> 00:03:57.982
this eigenvalue matrix.

00:04:00.550 --> 00:04:08.300
We can also write B as T
lambda T inverse, where lambda

00:04:08.300 --> 00:04:10.250
is the same in both
cases because they

00:04:10.250 --> 00:04:12.420
have the same eigenvalues.

00:04:12.420 --> 00:04:13.310
Good.

00:04:13.310 --> 00:04:18.420
Now I'll let you-- so
before we check it,

00:04:18.420 --> 00:04:20.560
let's just say the point.

00:04:20.560 --> 00:04:24.370
The point is if two matrices
are similar to the same matrix,

00:04:24.370 --> 00:04:26.950
then they're similar
to each other.

00:04:26.950 --> 00:04:30.100
Similarity is a
transitive relation.

00:04:30.100 --> 00:04:38.880
And I'll just let you check
that you can take T S inverse

00:04:38.880 --> 00:04:43.430
A times T S inverse
inverse, and you'll

00:04:43.430 --> 00:04:52.630
get B. This follows directly
from these two relations.

00:04:52.630 --> 00:04:53.130
Good.

00:04:53.130 --> 00:04:58.185
Now let's take on part
C. Part C is false.

00:05:01.240 --> 00:05:04.690
Let's come back over here and
look at these two matrices,

00:05:04.690 --> 00:05:06.310
J_1 and J_2.

00:05:06.310 --> 00:05:07.850
The first thing
you should see is

00:05:07.850 --> 00:05:12.810
that these two are
Jordan blocks-- sorry,

00:05:12.810 --> 00:05:17.840
not Jordan blocks, they're
matrices in Jordan normal form.

00:05:17.840 --> 00:05:20.200
They're different matrices
in Jordan normal form,

00:05:20.200 --> 00:05:23.350
so they will not be similar.

00:05:23.350 --> 00:05:27.060
But let's actually see why.

00:05:27.060 --> 00:05:30.450
Let's look at-- remember,
one of the things

00:05:30.450 --> 00:05:34.790
that similarity preserves are
eigenvectors and eigenvalues.

00:05:34.790 --> 00:05:39.710
So let's look at the eigenspace
with eigenvalue minus one

00:05:39.710 --> 00:05:41.710
with these two matrices.

00:05:41.710 --> 00:05:49.300
So J_1 plus the identity--
let's look at the nullspace

00:05:49.300 --> 00:05:50.720
of this matrix.

00:05:50.720 --> 00:05:58.620
So this is just
0's on the diagonal

00:05:58.620 --> 00:06:01.000
and 1's right
above the diagonal.

00:06:01.000 --> 00:06:05.530
And J_2 plus the identity.

00:06:05.530 --> 00:06:09.790
this is just 0, 1, 0, 0.

00:06:12.350 --> 00:06:20.430
So the point is that the
nullspace of this matrix

00:06:20.430 --> 00:06:22.050
is just one-dimensional.

00:06:22.050 --> 00:06:24.595
So there's only one
independent eigenvector

00:06:24.595 --> 00:06:27.570
of J_1 with eigenvalue minus 1.

00:06:27.570 --> 00:06:31.900
Whereas the nullspace of this
matrix is two-dimensional.

00:06:36.550 --> 00:06:38.970
There are two
independent eigenvectors

00:06:38.970 --> 00:06:41.130
with eigenvalue minus 1.

00:06:41.130 --> 00:06:49.390
So the dimension-- the nullspace
of J_1 plus the identity,

00:06:49.390 --> 00:06:54.650
this is 1, and this is 2.

00:06:54.650 --> 00:06:58.880
So they cannot
possibly be similar.

00:07:02.700 --> 00:07:03.200
Good.

00:07:03.200 --> 00:07:04.820
So that completes the problem.

00:07:04.820 --> 00:07:08.030
It's a nice exercise to
do this more generally.

00:07:08.030 --> 00:07:13.790
And you can use these techniques
not just looking at the number

00:07:13.790 --> 00:07:20.170
of independent eigenvectors and
the nullspace of your J minus

00:07:20.170 --> 00:07:23.340
lambda*I matrix, but also
powers of J minus lambda I

00:07:23.340 --> 00:07:24.710
and their nullspaces.

00:07:24.710 --> 00:07:28.690
You can use this to
show that any two

00:07:28.690 --> 00:07:31.470
matrices in Jordan normal
form that are different

00:07:31.470 --> 00:07:33.500
are not similar.

00:07:33.500 --> 00:07:34.770
This same method works.

00:07:34.770 --> 00:07:36.960
And that's a nice
exercise if you

00:07:36.960 --> 00:07:41.340
want to go a little further
with similar matrices.

00:07:41.340 --> 00:07:45.530
Let's just recap the
properties we saw here.

00:07:45.530 --> 00:07:48.180
We saw that if we had
two similar matrices,

00:07:48.180 --> 00:07:51.840
then any polynomials in
those matrices were similar.

00:07:51.840 --> 00:07:56.550
And we saw that if we
have two matrices that

00:07:56.550 --> 00:08:01.300
have the same distinct
eigenvalues, then

00:08:01.300 --> 00:08:02.030
they're similar.

00:08:02.030 --> 00:08:04.450
And we saw that,
in a special case,

00:08:04.450 --> 00:08:07.390
we saw that two matrices
in Jordan normal form

00:08:07.390 --> 00:08:10.360
that are different
are not similar.

00:08:10.360 --> 00:08:11.910
Thanks.