WEBVTT

00:00:05.247 --> 00:00:07.420
PROFESSOR: Welcome
to this recitation

00:00:07.420 --> 00:00:09.390
on matrix exponential.

00:00:09.390 --> 00:00:13.520
So here, we're given matrix
A with entries 6, 5, 1, 2.

00:00:13.520 --> 00:00:16.290
And we're asked to compute
the matrix exponential,

00:00:16.290 --> 00:00:19.530
exponential A*t, and to use
it to solve the initial value

00:00:19.530 --> 00:00:23.570
problem u prime of
t equals A*u(t),

00:00:23.570 --> 00:00:25.900
where here u are
basically vectors,

00:00:25.900 --> 00:00:29.120
with initial condition,
u of 0 equals [4, 1].

00:00:29.120 --> 00:00:31.216
So why don't you
pause the video, work

00:00:31.216 --> 00:00:32.049
through the problem?

00:00:32.049 --> 00:00:33.007
And I'll be right back.

00:00:44.030 --> 00:00:46.930
Welcome back.

00:00:46.930 --> 00:00:51.180
So first, to go ahead and
compute the matrix exponential,

00:00:51.180 --> 00:00:54.960
we need to identify the
eigenvalues of matrix

00:00:54.960 --> 00:00:57.990
A and its eigenvectors.

00:00:57.990 --> 00:01:02.872
So this is a matrix--
I'll just rewrite here--

00:01:02.872 --> 00:01:04.720
that we saw before.

00:01:04.720 --> 00:01:14.020
And its eigenvalues are again,
solution of 6 minus lambda, 5,

00:01:14.020 --> 00:01:21.400
1, 2 minus lambda, equals
to 0, which gives us 6...,

00:01:21.400 --> 00:01:27.830
2 minus lambda
minus 5 equals to 0.

00:01:27.830 --> 00:01:31.570
Lambda square minus 8*lambda.

00:01:31.570 --> 00:01:35.610
Then we have a 12 minus 5.

00:01:35.610 --> 00:01:41.410
So you can verify that the two
eigenvalues would be 1 and 7.

00:01:49.980 --> 00:01:51.980
Lambda_1 equals to 1.

00:01:51.980 --> 00:01:54.530
And lambda_1 equals to 7.

00:01:54.530 --> 00:01:59.090
So now, we need to seek
the eigenvectors associated

00:01:59.090 --> 00:02:00.680
to each one of the eigenvalues.

00:02:00.680 --> 00:02:03.070
So the idea here is
to basically move

00:02:03.070 --> 00:02:05.610
toward a diagonalization
of the matrix A.

00:02:05.610 --> 00:02:08.100
So let's seek the eigenvectors.

00:02:08.100 --> 00:02:12.110
And here, I'm just going
to give them to you,

00:02:12.110 --> 00:02:16.004
and you can verify
the calculation.

00:02:16.004 --> 00:02:17.420
And this calculation
was performed

00:02:17.420 --> 00:02:19.200
in a previous recitation.

00:02:19.200 --> 00:02:21.080
So the eigenvectors.

00:02:21.080 --> 00:02:30.510
v_1 associated to the eigenvalue
lambda_1 was, for example, 1,

00:02:30.510 --> 00:02:32.650
minus 1.

00:02:32.650 --> 00:02:38.080
And the other one
that we found--

00:02:38.080 --> 00:02:47.200
again, this is one form of
the eigenvector-- was 5 and 1.

00:02:49.912 --> 00:02:53.660
So these are from the notes
of a previous recitation.

00:02:53.660 --> 00:02:56.930
So you can verify that these
are the two eigenvectors.

00:02:56.930 --> 00:03:03.430
And from this point, then we
can rewrite this solution,

00:03:03.430 --> 00:03:04.780
if you recall.

00:03:04.780 --> 00:03:07.420
I'm just going to
go through the steps

00:03:07.420 --> 00:03:10.622
toward getting to the definition
of the exponential matrix.

00:03:10.622 --> 00:03:12.080
So here, if we
didn't know anything

00:03:12.080 --> 00:03:13.663
about the exponential
matrix, we would

00:03:13.663 --> 00:03:20.340
be able to write the
solution as c_1 exponential t

00:03:20.340 --> 00:03:33.670
v_1 plus c_2 exponential 7t v_2,
which basically gives us here,

00:03:33.670 --> 00:03:40.390
if I write it in this form,
for example, an exponential t,

00:03:40.390 --> 00:03:47.990
minus exponential t and an
exponential 5t multiplied

00:03:47.990 --> 00:03:53.530
by the entry of this vector,
an exponential 7t here,

00:03:53.530 --> 00:03:57.010
multiplying [c_1, c_2].

00:03:57.010 --> 00:04:03.340
So this is where the idea of the
matrix exponential comes from.

00:04:03.340 --> 00:04:10.320
We're basically introducing
the matrix phi of t

00:04:10.320 --> 00:04:14.480
for which we can write u equals
phi of t multiplied by this

00:04:14.480 --> 00:04:18.579
[c_1, c_2], general constants.

00:04:18.579 --> 00:04:24.240
So phi of t would then
be equal to this matrix.

00:04:24.240 --> 00:04:29.410
But what we want is to be
able to solve an initial value

00:04:29.410 --> 00:04:38.590
problem for which
e of A of 0 applied

00:04:38.590 --> 00:04:42.700
to our initial conditions
would give us back

00:04:42.700 --> 00:04:46.260
our initial condition.

00:04:46.260 --> 00:04:50.840
So we're seeking for a form for
this exponential matrix that

00:04:50.840 --> 00:04:53.260
would allow us to do this.

00:04:53.260 --> 00:04:59.880
So the way that we define the
matrix exponential give us

00:04:59.880 --> 00:05:03.809
exponential A*t-- now, I
won't go into the proof,

00:05:03.809 --> 00:05:05.600
but we're just going
to check it together--

00:05:05.600 --> 00:05:08.930
multiplied by phi of 0 minus 1.

00:05:08.930 --> 00:05:14.030
So let's check that if we
use this form of the matrix

00:05:14.030 --> 00:05:15.590
exponential, we would have e.

00:05:18.920 --> 00:05:26.740
We will have that at
0 applied to u(0).

00:05:26.740 --> 00:05:37.210
We have phi(0), phi(0)^(-1)
applied to u(0).

00:05:37.210 --> 00:05:38.850
This is a matrix
with its inverse,

00:05:38.850 --> 00:05:40.780
which gives us the identity.

00:05:40.780 --> 00:05:45.820
And so basically, this
gives us back u of 0.

00:05:48.847 --> 00:05:50.930
I mean you don't need to
do that when you're asked

00:05:50.930 --> 00:05:52.480
to find the matrix exponential.

00:05:52.480 --> 00:05:55.460
But just to remember
where it's coming from,

00:05:55.460 --> 00:05:58.900
you write down your
system in matrix form.

00:05:58.900 --> 00:06:00.940
You identify the
matrix phi of t.

00:06:00.940 --> 00:06:06.080
And then you recall why you want
the matrix exponential to have

00:06:06.080 --> 00:06:08.490
this form, basically to be
able to solve initial value

00:06:08.490 --> 00:06:14.620
problems for which the value
u of t is projected to u of 0

00:06:14.620 --> 00:06:19.500
when we take t equals 0
for the matrix exponential.

00:06:19.500 --> 00:06:22.070
So now let's go
back to our problem.

00:06:22.070 --> 00:06:24.020
So let's compute this
matrix exponential.

00:06:24.020 --> 00:06:26.300
We have phi of t.

00:06:26.300 --> 00:06:31.730
So now from this formula, we
know that we need phi of 0.

00:06:31.730 --> 00:06:35.270
So that give us, basically,
exponential of 0,

00:06:35.270 --> 00:06:42.080
5, minus 1, and 1.

00:06:42.080 --> 00:06:44.390
We need to find its inverse.

00:06:44.390 --> 00:06:52.810
So recall that the inverse
of a two-by-two matrix

00:06:52.810 --> 00:07:01.200
is basically just the
determinant, minus b,

00:07:01.200 --> 00:07:04.860
minus c, and reversing
the diagonal entries.

00:07:04.860 --> 00:07:10.850
So we can just apply this
to get our phi of 0 minus 1.

00:07:10.850 --> 00:07:14.930
So here, our
determinant is basically

00:07:14.930 --> 00:07:18.800
1 plus 5, which is 1 over 6.

00:07:18.800 --> 00:07:25.620
And then the entries are
simply 1, 1, minus 5, and 1.

00:07:25.620 --> 00:07:29.230
So now, we're just left with the
multiplication of two matrices

00:07:29.230 --> 00:07:31.442
to get our matrix exponential.

00:07:31.442 --> 00:07:42.500
So our matrix exponential
would give us this one sixth.

00:07:42.500 --> 00:07:46.150
And we now have to
multiply the entries.

00:07:46.150 --> 00:07:47.850
So I'm not going to
rewrite everything.

00:07:47.850 --> 00:07:50.280
I'm just going to
use this space here.

00:07:50.280 --> 00:07:54.030
So we have exponential
t multiplying

00:07:54.030 --> 00:07:57.755
1, plus 5 exponential 7t.

00:08:04.280 --> 00:08:07.520
Then, we have exponential t
dot minus 5 for this entry.

00:08:10.570 --> 00:08:14.050
5 exponential t
multiplying our 1.

00:08:20.225 --> 00:08:23.820
7t, thank you.

00:08:23.820 --> 00:08:26.190
Then for the second
entry, we basically

00:08:26.190 --> 00:08:36.070
have minus exponential
t 1 exponential 7t 1

00:08:36.070 --> 00:08:44.520
minus exponential t minus
5 and exponential 7t 1.

00:08:49.700 --> 00:08:53.306
So we're done with the
matrix exponential.

00:08:53.306 --> 00:08:56.790
So now we were asked to solve
for the initial value problem

00:08:56.790 --> 00:09:00.570
with initial condition 4 and 1.

00:09:00.570 --> 00:09:01.830
So how do we go about that?

00:09:01.830 --> 00:09:05.170
Well, recall that
I just reminded you

00:09:05.170 --> 00:09:07.650
what did we want to use
this matrix exponential for.

00:09:07.650 --> 00:09:11.920
And what we wanted it for is
to be able to basically project

00:09:11.920 --> 00:09:17.200
an initial condition into a
solution u of t, t times later.

00:09:17.200 --> 00:09:21.600
And we constructed this matrix
to be able to basically give us

00:09:21.600 --> 00:09:23.890
this solution by just
multiplying the matrix

00:09:23.890 --> 00:09:25.550
by the initial value vector.

00:09:25.550 --> 00:09:28.450
So basically, to find the
solution of this initial value

00:09:28.450 --> 00:09:31.780
problem, we simply need
to multiply this matrix

00:09:31.780 --> 00:09:35.210
by the initial vector
that we were given.

00:09:35.210 --> 00:09:39.210
And I'm just going to
write it here to not

00:09:39.210 --> 00:09:41.340
have to rewrite everything.

00:09:41.340 --> 00:09:44.090
And it was 4 and 1.

00:09:44.090 --> 00:09:48.980
And this is u of 0.

00:09:48.980 --> 00:09:51.218
So let me just do
a dash here just so

00:09:51.218 --> 00:09:52.840
that we can do the computation.

00:09:52.840 --> 00:09:57.770
And we would end up
with a solution--

00:09:57.770 --> 00:10:01.510
I'm going to keep it
in matrix form for now.

00:10:01.510 --> 00:10:08.960
So we end up with 4 exponential
t minus 5 exponential t,

00:10:08.960 --> 00:10:11.500
so minus 1 exponential t.

00:10:11.500 --> 00:10:13.950
And we have a one sixth.

00:10:13.950 --> 00:10:24.350
Here, 5 exponential 7t, so
we have 20, plus 5, so 25,

00:10:24.350 --> 00:10:26.630
exponential 7t.

00:10:29.260 --> 00:10:34.070
Then for the second entry
of the vector solution,

00:10:34.070 --> 00:10:40.700
we have minus exponential here
minus 4 that we add to a 5,

00:10:40.700 --> 00:10:45.540
and here, a 7 multiplied
by 4 that we add to a 1.

00:10:45.540 --> 00:10:52.310
So we have basically
plus 5 exponential 7t.

00:10:52.310 --> 00:10:56.370
And that basically gives us one
way of writing this solution.

00:10:56.370 --> 00:11:07.843
And we can split this down,
if we will, into two vectors,

00:11:07.843 --> 00:11:15.940
plus t; minus 1, 1;
exponential 7t; [25, 5].

00:11:15.940 --> 00:11:18.610
And this form is as valid.

00:11:18.610 --> 00:11:21.720
Yes, thank you.

00:11:21.720 --> 00:11:24.060
So that ends the
laborious calculations.

00:11:24.060 --> 00:11:25.680
But basically,
the key point here

00:11:25.680 --> 00:11:28.340
was just to remember where is
the matrix exponential coming

00:11:28.340 --> 00:11:33.270
from, basically, from the
eigenvalues and eigenvectors

00:11:33.270 --> 00:11:36.850
of the original matrix
present in the system,

00:11:36.850 --> 00:11:39.340
and where is the
definition coming from,

00:11:39.340 --> 00:11:43.250
why do we define it as
phi of t phi minus 1 of 0,

00:11:43.250 --> 00:11:49.910
and how to use it then
to give the solution

00:11:49.910 --> 00:11:51.290
to an initial value problem.

00:11:51.290 --> 00:11:53.700
So that ends this recitation.