WEBVTT

00:00:07.165 --> 00:00:07.790
JOEL LEWIS: Hi.

00:00:07.790 --> 00:00:09.364
Welcome back to recitation.

00:00:09.364 --> 00:00:11.280
You've been learning in
lecture about matrices

00:00:11.280 --> 00:00:13.820
and their various
applications, and one of them

00:00:13.820 --> 00:00:16.490
is to solving systems
of linear equations.

00:00:16.490 --> 00:00:19.690
So I have here a system of
three linear equations for you.

00:00:19.690 --> 00:00:25.203
2x plus c*z equals 4, x
minus y plus 2z equals pi,

00:00:25.203 --> 00:00:29.730
and x minus 2y plus
2z equals minus 12.

00:00:29.730 --> 00:00:31.790
So what I'd like you
to do is the following.

00:00:31.790 --> 00:00:35.080
Find the value of c--
or all values of c--

00:00:35.080 --> 00:00:38.100
for which, first of all,
there's a unique solution

00:00:38.100 --> 00:00:40.010
to this system.

00:00:40.010 --> 00:00:43.140
Second of all, for which the
corresponding homogeneous

00:00:43.140 --> 00:00:44.790
system has a unique solution.

00:00:44.790 --> 00:00:47.250
So remember that the
corresponding homogeneous

00:00:47.250 --> 00:00:50.330
system is the system where you
just replace these constants

00:00:50.330 --> 00:00:51.620
on the right by 0.

00:00:51.620 --> 00:00:53.310
So it's a very
similar-looking system.

00:00:53.310 --> 00:00:55.130
The left-hand sides
are all the same,

00:00:55.130 --> 00:00:57.775
but the right-hand sides
are replaced with 0.

00:00:57.775 --> 00:01:00.275
So you want to find the value
of c for which this system has

00:01:00.275 --> 00:01:02.070
a unique solution,
the value of c

00:01:02.070 --> 00:01:04.520
for which the corresponding
homogeneous system has

00:01:04.520 --> 00:01:10.300
a unique solution, and
also the values of c

00:01:10.300 --> 00:01:13.830
for which the corresponding
homogeneous system has

00:01:13.830 --> 00:01:15.310
infinitely many solutions.

00:01:15.310 --> 00:01:18.390
Note that I'm not asking you to
solve this system of equations,

00:01:18.390 --> 00:01:21.790
although you're welcome
to do so if you like.

00:01:21.790 --> 00:01:23.720
Although, of course,
whether you can or not

00:01:23.720 --> 00:01:25.850
might depend on the value of c.

00:01:25.850 --> 00:01:30.210
So why don't you pause the
video, take a little while

00:01:30.210 --> 00:01:33.179
to work out the solutions
to these three questions,

00:01:33.179 --> 00:01:34.970
come back, and we can
work it out together.

00:01:43.530 --> 00:01:46.820
So hopefully you have some luck
working out these problems.

00:01:46.820 --> 00:01:48.860
Let's start working
through them together.

00:01:48.860 --> 00:01:52.590
So I'm actually going
to take parts a and b

00:01:52.590 --> 00:01:53.682
together at the same time.

00:01:53.682 --> 00:01:55.140
And the reason that
I'm going to do

00:01:55.140 --> 00:01:58.230
that is that one
thing you've learned

00:01:58.230 --> 00:02:04.140
is that a system has
a unique solution for,

00:02:04.140 --> 00:02:08.470
on the right-hand side-- sorry--
a system has a unique solution,

00:02:08.470 --> 00:02:11.580
like this, a square
system of linear equations

00:02:11.580 --> 00:02:13.670
has a unique
solution if and only

00:02:13.670 --> 00:02:15.490
if it has a unique
solution regardless

00:02:15.490 --> 00:02:17.070
of what the right-hand side is.

00:02:17.070 --> 00:02:20.320
So in particular, the answer
to a and the answer to b

00:02:20.320 --> 00:02:21.870
are exactly the same.

00:02:21.870 --> 00:02:25.760
So values of c for which this
system has a unique solution

00:02:25.760 --> 00:02:29.480
are exactly the same as values
of c for which the homogeneous

00:02:29.480 --> 00:02:30.940
system has a unique solution.

00:02:30.940 --> 00:02:33.750
Now the solutions will
be different, of course.

00:02:33.750 --> 00:02:38.400
But the value of
c-- or the values

00:02:38.400 --> 00:02:40.660
of c-- that make it
solvable uniquely,

00:02:40.660 --> 00:02:43.660
make it solvable uniquely
for all right-hand sides.

00:02:43.660 --> 00:02:46.210
And so which values
of c are those?

00:02:46.210 --> 00:02:48.590
Well, those are the
values of c for which

00:02:48.590 --> 00:02:51.957
the coefficient matrix on the
left-hand side is invertible.

00:02:51.957 --> 00:02:54.040
So if the coefficient
matrix on the left-hand side

00:02:54.040 --> 00:02:56.620
is invertible, then we
can solve this system

00:02:56.620 --> 00:02:58.040
and we get a unique solution.

00:02:58.040 --> 00:03:01.040
If it's not invertible,
then either we

00:03:01.040 --> 00:03:04.170
can't solve this system--
like, there are no solutions--

00:03:04.170 --> 00:03:06.640
or we can solve this system,
but there are infinitely

00:03:06.640 --> 00:03:07.780
many solutions.

00:03:07.780 --> 00:03:10.700
So in both questions
a and b, we're

00:03:10.700 --> 00:03:14.680
asking for the value of c for
which the coefficient matrix

00:03:14.680 --> 00:03:16.450
of the left-hand
side is invertible,

00:03:16.450 --> 00:03:19.270
and that will be when we
have a unique solution.

00:03:19.270 --> 00:03:22.780
So how do we know when
a matrix is invertible?

00:03:22.780 --> 00:03:25.620
Well, let's write down what
the matrix is first of all.

00:03:25.620 --> 00:03:29.410
So this matrix M that we're
after is equal to the matrix

00:03:29.410 --> 00:03:43.290
2, 0, c; 1, minus
1, 2; 1, minus 2, 2.

00:03:43.290 --> 00:03:46.850
So this is the coefficient
matrix M of that system,

00:03:46.850 --> 00:03:49.650
and we want to know for which
values of c is it invertible.

00:03:49.650 --> 00:03:52.531
Well, when is a
matrix invertible?

00:03:52.531 --> 00:03:54.780
A matrix is invertible--
square matrix is invertible--

00:03:54.780 --> 00:03:58.310
precisely when it has
non-zero determinant.

00:03:58.310 --> 00:04:01.610
So we just need to look at the
determinant of this matrix.

00:04:01.610 --> 00:04:05.070
So you've learned how to compute
determinants of matrices,

00:04:05.070 --> 00:04:05.740
I think.

00:04:05.740 --> 00:04:10.830
So let's, in this case,
we have the det M.

00:04:10.830 --> 00:04:13.380
So it's a sum or difference
of six different terms,

00:04:13.380 --> 00:04:15.070
and you could get
it, for example,

00:04:15.070 --> 00:04:17.781
by the Laplace expansion
if you wanted to.

00:04:17.781 --> 00:04:20.030
So I'm just going to write
out what the six terms are.

00:04:20.030 --> 00:04:28.900
So it's 2 times minus 1 times
2, plus 0 times 2 times 1,

00:04:28.900 --> 00:04:38.910
plus c times 1 times minus
2, minus c times minus 1

00:04:38.910 --> 00:04:49.990
times 1, minus 2 times minus 2
times 2, minus 0 times 1 times

00:04:49.990 --> 00:04:50.900
2.

00:04:50.900 --> 00:04:53.530
So this is the determinant
of this matrix.

00:04:53.530 --> 00:04:56.517
You can get it either
just by remembering

00:04:56.517 --> 00:04:58.475
which terms are which
and which get a plus sign

00:04:58.475 --> 00:05:01.430
and which get a minus sign, or
by doing the Laplace expansion,

00:05:01.430 --> 00:05:05.600
or by whatever other tricks
you might happen to know.

00:05:05.600 --> 00:05:09.100
So now we need to know whether
or not this determinant is 0.

00:05:09.100 --> 00:05:11.622
So let's work out what this is.

00:05:11.622 --> 00:05:13.330
So this is-- let me
start simplifying it.

00:05:13.330 --> 00:05:19.060
So this is minus 4
plus 0 minus 2c--

00:05:19.060 --> 00:05:22.560
this is minus minus
c, so plus c--

00:05:22.560 --> 00:05:29.520
this is minus minus 8, so plus
8, which is equal to 4 minus c.

00:05:29.520 --> 00:05:32.390
So the determinant--
right, two of those terms

00:05:32.390 --> 00:05:35.810
are 0, and so I just
get to leave them out.

00:05:35.810 --> 00:05:39.900
So the determinant of
this matrix is 4 minus c.

00:05:39.900 --> 00:05:41.750
And what we're
interested in is when

00:05:41.750 --> 00:05:44.820
this determinant is non-zero.

00:05:44.820 --> 00:05:56.970
So in particular, for c
not equal to 0-- sorry,

00:05:56.970 --> 00:06:03.090
for c not equal to 4-- when c
is not 4, the determinant of M

00:06:03.090 --> 00:06:04.510
is not 0.

00:06:04.510 --> 00:06:12.930
So when c is not 4,
determinant of M is not 0,

00:06:12.930 --> 00:06:18.160
so both systems-- both
the original system

00:06:18.160 --> 00:06:21.210
and the corresponding
homogeneous system--

00:06:21.210 --> 00:06:33.170
have a unique solution.

00:06:33.170 --> 00:06:38.090
So when c is not 4-- so
for most values of c--

00:06:38.090 --> 00:06:41.140
the determinant is
not 0, and the system

00:06:41.140 --> 00:06:42.530
has a unique solution.

00:06:42.530 --> 00:06:45.150
So when c is equal
to 4, what happens?

00:06:45.150 --> 00:06:48.900
Well, when c is equal to 4,
we're in the bottom case.

00:06:48.900 --> 00:06:51.980
We're in the case where
the homogeneous system has

00:06:51.980 --> 00:06:53.960
infinitely many solutions.

00:06:53.960 --> 00:06:54.470
OK?

00:06:54.470 --> 00:06:58.640
So let me write that over here.

00:06:58.640 --> 00:07:04.260
When c equals 4-- I'm
going to abbreviate

00:07:04.260 --> 00:07:12.920
again-- the homogeneous
system has-- I'm

00:07:12.920 --> 00:07:17.120
going to use this symbol-- this
sort of sideways eight symbol

00:07:17.120 --> 00:07:19.230
means infinity, so
I'm going to use it

00:07:19.230 --> 00:07:26.660
for infinitely many solutions.

00:07:26.660 --> 00:07:28.976
So when c is 4, the
homogeneous system

00:07:28.976 --> 00:07:30.225
has infinitely many solutions.

00:07:30.225 --> 00:07:33.240
And you might be
curious-- well, so let

00:07:33.240 --> 00:07:35.240
me say one more
thing about that.

00:07:35.240 --> 00:07:39.640
We know when the coefficient
matrix isn't invertible

00:07:39.640 --> 00:07:42.960
that the system either has zero
or infinitely many solutions.

00:07:42.960 --> 00:07:45.160
But the homogeneous system
always has a solution.

00:07:45.160 --> 00:07:47.680
It always has the solution
where everything is all 0.

00:07:47.680 --> 00:07:48.180
Right?

00:07:48.180 --> 00:07:51.060
So that's why we know that
it's infinitely many here.

00:07:51.060 --> 00:07:54.100
And one thing you might ask
is can you find any others?

00:07:54.100 --> 00:08:00.280
Can you find any solutions
that aren't just [0, 0, 0]?

00:08:00.280 --> 00:08:01.690
And the answer is yes.

00:08:01.690 --> 00:08:04.312
So this is now going beyond
when I asked you to do,

00:08:04.312 --> 00:08:06.645
but I think it's, you know,
an interesting thing to see.

00:08:12.670 --> 00:08:17.740
So if you wanted to find another
solution, what do you know?

00:08:17.740 --> 00:08:21.840
Well, let's go back to
the equations that we had.

00:08:21.840 --> 00:08:24.460
So when we're dealing
with a homogeneous system,

00:08:24.460 --> 00:08:25.850
the right-hand sides are 0.

00:08:25.850 --> 00:08:28.058
So I'm just going to cross
out these right-hand sides

00:08:28.058 --> 00:08:30.280
and replace them with 0
so we don't get confused.

00:08:30.280 --> 00:08:34.950
So this is 0, 0, and 0.

00:08:34.950 --> 00:08:39.242
So we're dealing with this
system: 2x plus c*z equals 0,

00:08:39.242 --> 00:08:45.140
x minus y plus 2z equals 0, and
x minus 2y plus 2z equals 0.

00:08:45.140 --> 00:08:48.550
OK, so if you want
a solution [x, y, z]

00:08:48.550 --> 00:08:50.380
to this system,
what do you know?

00:08:50.380 --> 00:08:53.970
Well, from the second equation,
you know that the vector

00:08:53.970 --> 00:08:59.650
[x, y, z] is orthogonal to
the vector 1, minus 1, 2.

00:08:59.650 --> 00:09:00.540
How do you know that?

00:09:00.540 --> 00:09:03.890
Because this left-hand
side, x minus y plus 2z,

00:09:03.890 --> 00:09:10.350
is equal to [x, y, z]
dot 1, minus 1, 2.

00:09:10.350 --> 00:09:12.380
And similarly from
the third equation,

00:09:12.380 --> 00:09:14.270
you know that the
vector [x, y,  z]

00:09:14.270 --> 00:09:18.310
is orthogonal to the
vector 1, minus 2, 2,

00:09:18.310 --> 00:09:22.930
because this left-hand side
is equal to [x, y, z] dot 1,

00:09:22.930 --> 00:09:25.160
minus 2, 2.

00:09:25.160 --> 00:09:26.040
Yeah?

00:09:26.040 --> 00:09:27.700
And that's equal to 0.

00:09:27.700 --> 00:09:30.210
So from the second
and third equations,

00:09:30.210 --> 00:09:32.370
you know that you're
looking for a vector that's

00:09:32.370 --> 00:09:36.290
orthogonal to both
x-- or sorry-- both 1,

00:09:36.290 --> 00:09:40.039
minus 1, 2, and 1, minus 2, 2.

00:09:40.039 --> 00:09:41.580
How do you get a
vector perpendicular

00:09:41.580 --> 00:09:42.840
to two known vectors?

00:09:42.840 --> 00:09:44.510
Well, you just take
their cross product.

00:09:44.510 --> 00:09:46.110
So let's go back over here.

00:09:46.110 --> 00:10:04.750
So to find one, you
take a cross product

00:10:04.750 --> 00:10:07.380
of two rows of the
coefficient matrix.

00:10:07.380 --> 00:10:10.310
So in this case, for example,
we can take these rows, 1,

00:10:10.310 --> 00:10:13.210
minus 1, 2; and 1, minus 2, 2.

00:10:13.210 --> 00:10:18.000
So, for example, the
vector 1, minus 1,

00:10:18.000 --> 00:10:24.549
2-- OK-- cross the
vector 1, minus 2, 2.

00:10:24.549 --> 00:10:26.215
Now I've kind of run
out of board space,

00:10:26.215 --> 00:10:28.631
so I'm not going to work out
precisely what this vector is

00:10:28.631 --> 00:10:29.260
for you.

00:10:29.260 --> 00:10:32.090
But if you like, you
can certainly check.

00:10:32.090 --> 00:10:34.017
You can compute this
cross product out

00:10:34.017 --> 00:10:35.850
with our nice formula
for the cross product.

00:10:35.850 --> 00:10:37.830
It will give you some
vector, and then you

00:10:37.830 --> 00:10:40.220
can check that that
vector is indeed

00:10:40.220 --> 00:10:42.640
a solution of the
homogeneous system.

00:10:42.640 --> 00:10:45.117
So that will give
us a second solution

00:10:45.117 --> 00:10:46.200
of the homogeneous system.

00:10:46.200 --> 00:10:50.140
Nontrivial we say, because
it's not just the 0 solution.

00:10:50.140 --> 00:10:54.090
So to quickly recap, we had
a system of linear equations.

00:10:54.090 --> 00:10:57.380
I've now crossed out what the
original right-hand side was.

00:10:57.380 --> 00:10:59.100
We had a system of
linear equations,

00:10:59.100 --> 00:11:03.540
and we were looking for a choice
of c for which that system had

00:11:03.540 --> 00:11:06.700
a unique solution and for which
the corresponding homogeneous

00:11:06.700 --> 00:11:08.340
system had a unique solution.

00:11:08.340 --> 00:11:11.490
And the values of c that
make that work are precisely

00:11:11.490 --> 00:11:15.470
the values of c such that
the coefficient matrix

00:11:15.470 --> 00:11:17.030
has a non-zero determinant.

00:11:17.030 --> 00:11:18.975
So that's true for
both parts a and b.

00:11:18.975 --> 00:11:22.920
And for part c, when we were
looking for what values of c

00:11:22.920 --> 00:11:27.550
give the homogeneous system
infinitely many solutions,

00:11:27.550 --> 00:11:30.150
the answer is any
other value of c.

00:11:30.150 --> 00:11:34.140
Any value of c for which
the coefficient matrix does

00:11:34.140 --> 00:11:37.635
have 0 determinant will give
you infinitely many solutions

00:11:37.635 --> 00:11:40.530
in the homogeneous case,
and in non-homogeneous cases

00:11:40.530 --> 00:11:44.630
will either give you 0 solutions
or infinitely many solutions.

00:11:44.630 --> 00:11:46.390
And then we also at
the end, we briefly

00:11:46.390 --> 00:11:53.080
discussed one way to
find nontrivial solutions

00:11:53.080 --> 00:11:55.965
in the homogeneous case
when there are infinitely

00:11:55.965 --> 00:11:57.180
many solutions.

00:11:57.180 --> 00:11:58.954
So I'll end there.