WEBVTT

00:00:00.040 --> 00:00:02.460
The following content is
provided under a Creative

00:00:02.460 --> 00:00:03.870
Commons license.

00:00:03.870 --> 00:00:06.320
Your support will help
MIT OpenCourseWare

00:00:06.320 --> 00:00:10.560
continue to offer high quality
educational resources for free.

00:00:10.560 --> 00:00:13.300
To make a donation or
view additional materials

00:00:13.300 --> 00:00:17.116
from hundreds of MIT courses,
visit MIT OpenCourseWare

00:00:17.116 --> 00:00:17.740
at ocw.mit.edu.

00:00:32.145 --> 00:00:32.920
PROFESSOR: Hi.

00:00:32.920 --> 00:00:34.870
Today we, in a
manner of speaking,

00:00:34.870 --> 00:00:39.150
start the second part of
this part of our course,

00:00:39.150 --> 00:00:42.080
that what we have done
up to now is, hopefully,

00:00:42.080 --> 00:00:46.270
to have presented the concept
of partial derivatives

00:00:46.270 --> 00:00:47.620
with all the trimmings.

00:00:47.620 --> 00:00:50.250
And now what we would like
to do is devote the remainder

00:00:50.250 --> 00:00:52.510
of the course to
what we could call

00:00:52.510 --> 00:00:56.850
selected topics, in which
the partial derivative plays

00:00:56.850 --> 00:01:00.580
a fundamental underlying thread.

00:01:00.580 --> 00:01:02.720
In other words, a
situation in which

00:01:02.720 --> 00:01:06.730
we will now pick specific
areas of calculus,

00:01:06.730 --> 00:01:09.370
and show how what we've learned
about partial derivatives

00:01:09.370 --> 00:01:11.920
can be applied to
these particular areas.

00:01:11.920 --> 00:01:15.590
Now, the first natural area,
which would suggest itself,

00:01:15.590 --> 00:01:18.050
it seems to me,
would be the concept

00:01:18.050 --> 00:01:21.480
of integrating a function
of several variables.

00:01:21.480 --> 00:01:23.590
Since, after we
study derivatives,

00:01:23.590 --> 00:01:26.790
we next associated
the definite integral

00:01:26.790 --> 00:01:29.780
with an antiderivative,
it might seem

00:01:29.780 --> 00:01:32.280
that a very natural
inquiry now is

00:01:32.280 --> 00:01:35.900
to discuss the concept
of infinite multiple sums

00:01:35.900 --> 00:01:39.090
as being the analog of
the definite integral

00:01:39.090 --> 00:01:41.720
in our study of calculus
of a single variable.

00:01:41.720 --> 00:01:44.370
Now, because these
topics that we're picking

00:01:44.370 --> 00:01:48.180
have great application and have
been studied in great depth,

00:01:48.180 --> 00:01:52.030
we by necessity will
sort of skim the surface

00:01:52.030 --> 00:01:55.550
of many of these topics, so
that we can present as many

00:01:55.550 --> 00:01:59.130
of the key points as
we can in the time

00:01:59.130 --> 00:02:01.410
allotted for the course.

00:02:01.410 --> 00:02:03.600
We will make the
lectures, hopefully,

00:02:03.600 --> 00:02:07.460
as short as possible,
as pungent as possible,

00:02:07.460 --> 00:02:10.770
and leave most of the details
to the learning exercises,

00:02:10.770 --> 00:02:13.130
and to the supplementary notes.

00:02:13.130 --> 00:02:16.310
At any rate, what our lecture
today is called is "Double

00:02:16.310 --> 00:02:20.260
Sums, " or perhaps a better
word would be multiple sums.

00:02:20.260 --> 00:02:22.030
But we will
concentrate on the case

00:02:22.030 --> 00:02:24.370
of double sums for
the usual reason

00:02:24.370 --> 00:02:26.840
that namely, in the case of
two independent variables,

00:02:26.840 --> 00:02:30.240
we have convenient pictures
still left at our convenience.

00:02:30.240 --> 00:02:33.930
Now the idea is, let's
consider a region

00:02:33.930 --> 00:02:38.230
R. Say, the square,
the unit square

00:02:38.230 --> 00:02:42.010
on the coordinate axes with
vertices (0, 0), (1, 0),

00:02:42.010 --> 00:02:44.400
(1, 1), and (0, 1).

00:02:44.400 --> 00:02:46.590
Now, what we're
going to do now is

00:02:46.590 --> 00:02:50.470
imagine that R is a thin plate.

00:02:50.470 --> 00:02:54.220
And what we would like to do is
to find the mass of this plate

00:02:54.220 --> 00:02:59.410
R, assuming that the plate
has a variable density.

00:02:59.410 --> 00:03:00.980
In other words,
what we want to do

00:03:00.980 --> 00:03:05.320
is find the mass of the plate R,
if its density rho at the point

00:03:05.320 --> 00:03:08.505
x comma y in R is
given by rho of x,

00:03:08.505 --> 00:03:11.306
y equals x squared
plus y squared.

00:03:11.306 --> 00:03:12.680
You see, by the
way, the reason I

00:03:12.680 --> 00:03:15.920
picked this particular
density distribution is,

00:03:15.920 --> 00:03:19.900
you may notice that x squared
plus y squared, quite simply,

00:03:19.900 --> 00:03:22.510
is the square of the
distance of the point

00:03:22.510 --> 00:03:25.550
x comma y from the origin.

00:03:25.550 --> 00:03:27.670
And this will make
it very easy later

00:03:27.670 --> 00:03:31.300
for me to find where
lowest densities occur,

00:03:31.300 --> 00:03:33.000
where greatest densities occur.

00:03:33.000 --> 00:03:36.310
Whereas this may not seem
like a very applied problem.

00:03:36.310 --> 00:03:38.650
Notice that, if you want the
geometric picture of what's

00:03:38.650 --> 00:03:42.390
going on here, the idea
here is that the density

00:03:42.390 --> 00:03:46.970
is varying radially as we
move away from the origin.

00:03:46.970 --> 00:03:49.420
In other words, the
density is the square

00:03:49.420 --> 00:03:52.350
of the distance of the
points from the origin, which

00:03:52.350 --> 00:03:55.640
means that the density is
constant on any circle centered

00:03:55.640 --> 00:03:57.700
at the origin.

00:03:57.700 --> 00:04:00.740
Again, the key building
block that we're going to use

00:04:00.740 --> 00:04:03.320
is again analogous to
what we used in calculus

00:04:03.320 --> 00:04:04.690
of a single variable.

00:04:04.690 --> 00:04:06.960
For example, in calculus
of a single variable,

00:04:06.960 --> 00:04:09.970
when we said that distance
was equal to rate times time,

00:04:09.970 --> 00:04:12.930
it was assumed that
the rate was constant.

00:04:12.930 --> 00:04:15.170
In a similar way,
notice that if we

00:04:15.170 --> 00:04:18.010
assumed that the
density were constant--

00:04:18.010 --> 00:04:20.839
if the density were
constant, then the mass

00:04:20.839 --> 00:04:24.930
would just be the
density times the area.

00:04:24.930 --> 00:04:27.420
Again, most of us are used
to thinking of density

00:04:27.420 --> 00:04:29.440
in terms of three
dimensions, that the mass is

00:04:29.440 --> 00:04:31.610
the density times the volume.

00:04:31.610 --> 00:04:34.130
I prefer to pick
a thin plate here,

00:04:34.130 --> 00:04:36.560
simply to utilize
a simple diagram.

00:04:36.560 --> 00:04:38.240
But I did not want
to get involved

00:04:38.240 --> 00:04:40.670
with three-dimensional diagrams.

00:04:40.670 --> 00:04:43.180
Another drawback,
perhaps, to the system

00:04:43.180 --> 00:04:45.420
is that you may be
wondering why we're not

00:04:45.420 --> 00:04:48.715
using the counterpart of
what area was in calculus

00:04:48.715 --> 00:04:49.590
of a single variable.

00:04:49.590 --> 00:04:54.170
Why aren't we talking about
the volume in two dimensions?

00:04:54.170 --> 00:04:57.060
Why couldn't we somehow
visualize a volume problem

00:04:57.060 --> 00:04:57.560
here?

00:04:57.560 --> 00:04:59.950
Again, if that's what
you'd like to do,

00:04:59.950 --> 00:05:02.780
notice that we could
graph the density

00:05:02.780 --> 00:05:06.490
as a function of the point
x comma y in the region R.

00:05:06.490 --> 00:05:09.030
Plot the density
in the direction

00:05:09.030 --> 00:05:11.970
of the positive
z-axis, in which case

00:05:11.970 --> 00:05:14.620
the problem that we would
then be trying to solve

00:05:14.620 --> 00:05:17.490
would be to find the
volume of a solid

00:05:17.490 --> 00:05:21.000
that you get by taking the
parallelepiped, whose base is

00:05:21.000 --> 00:05:25.450
the region R, and intersect
that with the surface z

00:05:25.450 --> 00:05:27.510
equals x squared plus y squared.

00:05:27.510 --> 00:05:31.090
Again, we will leave these
details for the exercises.

00:05:31.090 --> 00:05:33.240
But the surface z
equals x squared

00:05:33.240 --> 00:05:36.340
plus y squared, as you look
in along the z-axis here,

00:05:36.340 --> 00:05:38.610
is a parabolic bowl opening up.

00:05:38.610 --> 00:05:40.470
In other words, z
equals x squared

00:05:40.470 --> 00:05:44.300
plus y squared, if you take z
equals a positive constant--

00:05:44.300 --> 00:05:47.300
in other words, slice this
parallel to the xy-plane--

00:05:47.300 --> 00:05:49.430
if z is a positive
constant, x squared

00:05:49.430 --> 00:05:53.140
plus y squared equals a
positive constant is a circle.

00:05:53.140 --> 00:05:55.740
In any rate, just
to summarize that,

00:05:55.740 --> 00:05:57.490
the geometric equivalent
to our problem

00:05:57.490 --> 00:06:01.200
is find the volume
of a solid S if S

00:06:01.200 --> 00:06:03.700
is the parallelepiped
with the base R

00:06:03.700 --> 00:06:07.050
where R is just as given,
and the top-- meaning

00:06:07.050 --> 00:06:09.340
it's intersected
with the surface z

00:06:09.340 --> 00:06:11.640
equals x squared plus y squared.

00:06:11.640 --> 00:06:13.660
And again, let me
mention as an aside

00:06:13.660 --> 00:06:17.970
that the same analogy appears in
calculus of a single variable.

00:06:17.970 --> 00:06:22.200
Namely, if I write down the
definite integral from 0 to 1

00:06:22.200 --> 00:06:26.100
x squared dx, I may view
this, well, in many ways,

00:06:26.100 --> 00:06:27.660
but two of the major
ways in which I

00:06:27.660 --> 00:06:29.618
may view this, the major
one that we emphasized

00:06:29.618 --> 00:06:31.780
in our course was as an area.

00:06:31.780 --> 00:06:34.570
I can also view this as a mass.

00:06:34.570 --> 00:06:39.860
More specifically, let me take
the region R as shown here.

00:06:39.860 --> 00:06:42.470
Then, you see,
one interpretation

00:06:42.470 --> 00:06:45.130
of the integral from
0 to 1 x squared dx

00:06:45.130 --> 00:06:46.940
is that it's the
area of the region

00:06:46.940 --> 00:06:50.320
R. Another
interpretation is imagine

00:06:50.320 --> 00:06:55.300
that you have a very thin rod,
geometrically of uniform shape.

00:06:55.300 --> 00:06:57.050
In other words, it's
essentially a segment

00:06:57.050 --> 00:06:59.110
of the x-axis from 0 to 1.

00:06:59.110 --> 00:07:01.590
And suppose that the
density of this rod

00:07:01.590 --> 00:07:05.860
is a function of position x,
say rho of x is x squared.

00:07:05.860 --> 00:07:08.470
Then notice, in a manner
analogous to what we were just

00:07:08.470 --> 00:07:11.280
talking about, the
mass of this rod

00:07:11.280 --> 00:07:16.010
would be found by integrating
x squared from 0 to 1.

00:07:16.010 --> 00:07:18.780
By the way, notice again
a very important thing

00:07:18.780 --> 00:07:22.620
that comes up here, and that
is that these two diagrams may

00:07:22.620 --> 00:07:24.670
be used for the same problem.

00:07:24.670 --> 00:07:27.800
You see, in many cases
when one has a density

00:07:27.800 --> 00:07:30.530
distribution like
this, one visualizes

00:07:30.530 --> 00:07:33.610
the density being plotted.

00:07:33.610 --> 00:07:36.560
In other words, the density
rho of x equals x squared

00:07:36.560 --> 00:07:39.400
is plotted as the curve y
equals rho of x, in this case,

00:07:39.400 --> 00:07:40.930
y equals x squared.

00:07:40.930 --> 00:07:43.800
And we then think
of a way of being

00:07:43.800 --> 00:07:47.370
able to visualize what the
density is doing here in terms

00:07:47.370 --> 00:07:49.130
of the shape of the curve here.

00:07:49.130 --> 00:07:52.250
That's exactly what
we were doing earlier

00:07:52.250 --> 00:07:54.440
in the lecture, when we
said that we can visualize

00:07:54.440 --> 00:07:59.860
the density rho of x, y as a
function z equals rho of x, y.

00:07:59.860 --> 00:08:02.300
The analogy being that with
two independent variables,

00:08:02.300 --> 00:08:06.530
we wouldn't have a curve, but
rather we would have a surface.

00:08:06.530 --> 00:08:10.200
By the way, again, notice
how this emphasizes the fact

00:08:10.200 --> 00:08:15.200
that the definite integral is
in reality an infinite sum.

00:08:15.200 --> 00:08:17.490
That it is convenient
in many cases

00:08:17.490 --> 00:08:21.870
to view this limit of a sum in
terms of an area under a curve.

00:08:21.870 --> 00:08:25.600
But that the definite integral
itself exists for more problems

00:08:25.600 --> 00:08:27.380
than just for computing areas.

00:08:27.380 --> 00:08:30.100
Now at any rate, all we're
trying to point out here

00:08:30.100 --> 00:08:32.059
is that the same
analogy will exist

00:08:32.059 --> 00:08:34.970
when we go to functions
of several variables.

00:08:34.970 --> 00:08:38.340
If, in fact, we now return
to the problem stated

00:08:38.340 --> 00:08:41.169
at the beginning of the
lecture, we take our region

00:08:41.169 --> 00:08:43.990
R. Remember what this is, now.

00:08:43.990 --> 00:08:48.340
It's this square with
vertices at these points.

00:08:48.340 --> 00:08:52.560
We know that the density is
given at any point x comma y

00:08:52.560 --> 00:08:54.150
by x squared plus y squared.

00:08:54.150 --> 00:08:56.670
And what we'd like to
do now is find the mass

00:08:56.670 --> 00:08:58.360
of this particular plate.

00:08:58.360 --> 00:09:00.880
Now, just as in calculus
of a single variable,

00:09:00.880 --> 00:09:03.410
the argument that we
use now is nothing more

00:09:03.410 --> 00:09:06.240
than if we took a small
enough region here,

00:09:06.240 --> 00:09:08.910
we could assume that
the variable density was

00:09:08.910 --> 00:09:10.095
essentially constant.

00:09:10.095 --> 00:09:13.160
In other words, this is how one
makes physical approximations.

00:09:13.160 --> 00:09:15.930
We break this thing
up into small pieces.

00:09:15.930 --> 00:09:19.250
Somehow or other find the
mass of each small piece,

00:09:19.250 --> 00:09:20.790
and add those all up.

00:09:20.790 --> 00:09:23.470
The key point being, if
the pieces are small enough

00:09:23.470 --> 00:09:26.530
for an approximation, we
can say that the density

00:09:26.530 --> 00:09:28.170
is essentially constant.

00:09:28.170 --> 00:09:30.260
By the way, in the
same way that we

00:09:30.260 --> 00:09:34.390
had to learn the sigma notation
back in part one of our course,

00:09:34.390 --> 00:09:37.080
we will now have to
somehow or other learn

00:09:37.080 --> 00:09:40.190
to master the notation
for double sums.

00:09:40.190 --> 00:09:45.040
The idea, again, being this: if
I partition this into n equal

00:09:45.040 --> 00:09:48.890
parts, and this segment
into m equal parts,

00:09:48.890 --> 00:09:53.160
and if I call a typical
x segment, delta x sub i,

00:09:53.160 --> 00:09:56.100
and a typical y
segment, delta y_j,

00:09:56.100 --> 00:10:00.210
then to match up in coordinate
fashion what the mass of this

00:10:00.210 --> 00:10:03.477
little piece would be, I
would call that the (i,

00:10:03.477 --> 00:10:07.330
j)-th mass corresponding
to the i-th x partition,

00:10:07.330 --> 00:10:09.980
and the j.th y partition.

00:10:09.980 --> 00:10:12.500
And the idea is, I'm going
to add all of these up.

00:10:12.500 --> 00:10:16.280
That would motivate the
notation called the double sum.

00:10:16.280 --> 00:10:18.590
Namely, the way you
read this is you

00:10:18.590 --> 00:10:22.590
say, let me first pick
a fixed value of j,

00:10:22.590 --> 00:10:27.460
and for that fixed value of j,
I will sum up all these pieces

00:10:27.460 --> 00:10:29.380
as i goes from 1 to n.

00:10:29.380 --> 00:10:31.530
What that means
pictorially is this.

00:10:31.530 --> 00:10:33.660
I will pick a fixed value of j.

00:10:33.660 --> 00:10:36.380
Well, since j controls
the y-coordinate,

00:10:36.380 --> 00:10:39.500
a fixed value of j fixes
your horizontal strip.

00:10:39.500 --> 00:10:41.000
And what you're
saying is I will now

00:10:41.000 --> 00:10:44.210
add up all the masses
along that strip.

00:10:44.210 --> 00:10:46.430
Then what you say
is, and then as I

00:10:46.430 --> 00:10:51.780
do that, let me perform that for
each j as j goes from 1 to m.

00:10:51.780 --> 00:10:53.370
In other words,
what this would say

00:10:53.370 --> 00:10:58.010
is, add up all of the
delta m's along each row,

00:10:58.010 --> 00:11:02.250
and then take the sum of
all the possible values

00:11:02.250 --> 00:11:03.430
that the rows can take on.

00:11:03.430 --> 00:11:05.860
In other words, I'll add up
these pieces, these pieces,

00:11:05.860 --> 00:11:08.990
these pieces, these pieces,
then add those sums together.

00:11:08.990 --> 00:11:10.770
Notice, of course,
I could have written

00:11:10.770 --> 00:11:12.700
this in the reverse order.

00:11:19.480 --> 00:11:20.910
Namely, this would say what?

00:11:20.910 --> 00:11:25.130
First hold i constant, and
for a fixed i, sum these as j

00:11:25.130 --> 00:11:26.540
goes from 1 to m.

00:11:26.540 --> 00:11:31.300
In other words, for a fixed
i, sum these as j goes from 1

00:11:31.300 --> 00:11:34.820
to m, and then sum
over all the j's.

00:11:34.820 --> 00:11:36.520
Over all the i's.

00:11:36.520 --> 00:11:38.920
At any rate, the
learning exercises

00:11:38.920 --> 00:11:41.790
will take care of making sure
that you learn to manipulate

00:11:41.790 --> 00:11:43.520
this particular notation.

00:11:43.520 --> 00:11:45.560
The theory works as follows.

00:11:45.560 --> 00:11:49.650
What we do is, assuming that the
density function is continuous,

00:11:49.650 --> 00:11:53.300
at each particular little
rectangle that we have here we,

00:11:53.300 --> 00:11:56.400
we pick the point which
has the smallest density,

00:11:56.400 --> 00:11:59.330
we picked the point which
has the largest density.

00:11:59.330 --> 00:12:04.070
We then imagine that we had
replaced this little segment

00:12:04.070 --> 00:12:07.780
by the same area, but
with a material which

00:12:07.780 --> 00:12:12.380
has a constant smallest
density, so that the true mass

00:12:12.380 --> 00:12:16.200
must be greater than
the mass of that piece.

00:12:16.200 --> 00:12:19.050
And then we assume that this
little piece was replaced

00:12:19.050 --> 00:12:20.930
by a new piece with
the same shape,

00:12:20.930 --> 00:12:23.490
but whose density was
constantly equal to the greatest

00:12:23.490 --> 00:12:24.610
density on this.

00:12:24.610 --> 00:12:26.940
So that this particular
mass must be less

00:12:26.940 --> 00:12:29.210
than the mass of the new piece.

00:12:29.210 --> 00:12:34.480
And in that way, we squeeze the
true mass between two extremes.

00:12:34.480 --> 00:12:37.740
Again, just as we did in
calculus of a single variable.

00:12:37.740 --> 00:12:40.320
And again, to keep the
notation going the same way,

00:12:40.320 --> 00:12:46.650
I let delta m sub i,j lower bar
denote the mass of the piece

00:12:46.650 --> 00:12:51.100
that I get by replacing the
density by the constant density

00:12:51.100 --> 00:12:55.960
equal to the smallest density,
rho sub i,j lower bar.

00:12:55.960 --> 00:12:58.510
Treating that as a
constant, multiplying that

00:12:58.510 --> 00:13:03.090
by the cross sectional area,
delta x_i times delta y_j.

00:13:03.090 --> 00:13:08.160
I prefer to call that delta a
sub i,j so we don't prejudice

00:13:08.160 --> 00:13:10.730
either the order in which
we write these factors,

00:13:10.730 --> 00:13:14.400
or the particular coordinate
system that one wants to use.

00:13:14.400 --> 00:13:16.240
But because I think
you're more familiar

00:13:16.240 --> 00:13:18.900
with Cartesian coordinates, I
will continue to write this.

00:13:18.900 --> 00:13:24.650
But the idea is, I compute
the lowest, the smallest

00:13:24.650 --> 00:13:27.230
mass in other words,
and underestimate.

00:13:27.230 --> 00:13:30.630
In a similar way, I
compute and overestimate.

00:13:30.630 --> 00:13:33.535
I then know that the
true mass of my (i,

00:13:33.535 --> 00:13:37.120
j)-th piece is caught
between these two,

00:13:37.120 --> 00:13:40.820
therefore my total mass, which
is the double sum of these over

00:13:40.820 --> 00:13:44.320
i and j, must be because
must be, in turn,

00:13:44.320 --> 00:13:47.370
caught between these
two double sums.

00:13:47.370 --> 00:13:49.420
In other words,
what I have now done

00:13:49.420 --> 00:13:52.750
is caught the true
mass between a lower

00:13:52.750 --> 00:13:55.450
bound and an upper bound.

00:13:55.450 --> 00:13:58.800
And again, to write this
thing more symbolically

00:13:58.800 --> 00:14:01.850
so we can look at it here, what
I'm saying is, on the one hand,

00:14:01.850 --> 00:14:05.370
my mass must be
at least as great

00:14:05.370 --> 00:14:07.460
as the value of this double sum.

00:14:07.460 --> 00:14:09.330
On the other hand,
it can be no greater

00:14:09.330 --> 00:14:11.630
than the value of
this double sum.

00:14:11.630 --> 00:14:13.220
Somehow or other,
the same as I did

00:14:13.220 --> 00:14:15.650
i calculus of a single
variable, I would now

00:14:15.650 --> 00:14:19.830
like to put the squeeze on
this as the sizes of the pieces

00:14:19.830 --> 00:14:23.400
delta x_i and delta
y_j are allowed to go

00:14:23.400 --> 00:14:25.560
get arbitrarily
close to 0 in size.

00:14:25.560 --> 00:14:29.280
What I hope happens
is that in that limit,

00:14:29.280 --> 00:14:32.970
this difference goes to 0 so
that M is caught between two

00:14:32.970 --> 00:14:35.770
equal things, and hence
the true value of M

00:14:35.770 --> 00:14:38.740
must be that common value.

00:14:38.740 --> 00:14:41.190
You see, in a manner of
speaking, theoretically,

00:14:41.190 --> 00:14:43.530
nothing new is happening
that didn't already

00:14:43.530 --> 00:14:45.730
happen in calculus
of a single variable,

00:14:45.730 --> 00:14:50.010
but from a computational point
of view, limits of single sums

00:14:50.010 --> 00:14:53.150
have now been replaced
by limits of double sums,

00:14:53.150 --> 00:14:56.790
so that computationally, there
is a degree of difficulty

00:14:56.790 --> 00:14:59.830
more present in
our present study

00:14:59.830 --> 00:15:02.810
than there was in our study of
calculus of a single variable.

00:15:02.810 --> 00:15:04.920
The computation
becomes more difficult,

00:15:04.920 --> 00:15:06.600
the theory stays the same.

00:15:06.600 --> 00:15:09.850
Now before we start getting into
the idea of what limits mean,

00:15:09.850 --> 00:15:11.800
let me just summarize
this for you

00:15:11.800 --> 00:15:14.450
in terms of a more
concrete interpretation.

00:15:14.450 --> 00:15:19.450
Let me again take that
square R with its vertices,

00:15:19.450 --> 00:15:23.540
the same one we dealt with,
(0, 0), (1, 0), (1, 1), (0, 1).

00:15:23.540 --> 00:15:26.610
And let me now take as
a special case the case

00:15:26.610 --> 00:15:31.310
in which I divide both the
x region and the y region

00:15:31.310 --> 00:15:36.440
into two equal parts, so that
my points of partition, you see,

00:15:36.440 --> 00:15:38.710
look like this.

00:15:38.710 --> 00:15:42.550
Now, you see, the advantage
of picking my density

00:15:42.550 --> 00:15:45.500
to be x squared
plus y squared is

00:15:45.500 --> 00:15:47.230
that since the
density is radial,

00:15:47.230 --> 00:15:50.460
what this means is that
in each of my partition

00:15:50.460 --> 00:15:54.320
rectangles-- in this case, each
of my rectangles is a square.

00:15:54.320 --> 00:15:56.760
You see, keep in mind
that first of all, I

00:15:56.760 --> 00:16:00.290
don't have to divide these two
dimensions into the same number

00:16:00.290 --> 00:16:01.410
of equal parts.

00:16:01.410 --> 00:16:04.890
Secondly, they don't even
have to be equal parts.

00:16:04.890 --> 00:16:08.052
But I've, just for symmetry
here, elected to do this.

00:16:08.052 --> 00:16:10.510
Just so we get an idea of what's
happening computationally,

00:16:10.510 --> 00:16:12.010
and the more difficult
cases will be

00:16:12.010 --> 00:16:13.720
taken care of in the exercises.

00:16:13.720 --> 00:16:15.080
The idea now is what?

00:16:15.080 --> 00:16:19.380
That for each of these
particular squares, the lowest

00:16:19.380 --> 00:16:22.380
density occurs at the
point which makes up

00:16:22.380 --> 00:16:24.600
the lower left hand
corner, because that

00:16:24.600 --> 00:16:28.120
will be the closest point
to the origin in each

00:16:28.120 --> 00:16:29.150
of these squares.

00:16:29.150 --> 00:16:33.540
And the furthest point from the
origin in each of these squares

00:16:33.540 --> 00:16:36.720
would be the point in the
upper right hand corner.

00:16:36.720 --> 00:16:40.090
So you see in this
particular case,

00:16:40.090 --> 00:16:41.600
let me do both of these at once.

00:16:41.600 --> 00:16:45.680
To find the smallest density
in each of these regions,

00:16:45.680 --> 00:16:47.370
let me have the
smallest one written

00:16:47.370 --> 00:16:50.960
in the black chalk on the
lower part of this diagonal.

00:16:50.960 --> 00:16:54.470
You see the point (0, 0)
corresponds to a density of 0.

00:16:54.470 --> 00:16:57.750
So the lowest possible
density is 0 in this block.

00:16:57.750 --> 00:17:00.440
The point 1/2 comma 0
is the closest point

00:17:00.440 --> 00:17:02.780
to the origin in
this little square.

00:17:02.780 --> 00:17:05.400
So x squared plus y
squared in that case

00:17:05.400 --> 00:17:08.780
is simply 1/2 squared plus 0
squared, which is a quarter.

00:17:08.780 --> 00:17:12.630
The closest point
here is 1/2 comma 1/2,

00:17:12.630 --> 00:17:15.150
so if we square the
coordinates and add,

00:17:15.150 --> 00:17:17.760
we get 1/4 plus
1/4, which is 1/2.

00:17:17.760 --> 00:17:21.200
And in a similar way, the
lowest density in this block

00:17:21.200 --> 00:17:23.829
is also 1/4.

00:17:23.829 --> 00:17:26.710
Now again, using
the same argument,

00:17:26.710 --> 00:17:29.940
the point in the
first block which

00:17:29.940 --> 00:17:33.260
is further from the
origin is 1/2 comma 1/2.

00:17:33.260 --> 00:17:36.290
That density corresponds to 1/2.

00:17:36.290 --> 00:17:38.530
In other words, 1/2
squared plus 1/2 squared.

00:17:38.530 --> 00:17:40.880
So now putting in the biggest
densities in each block,

00:17:40.880 --> 00:17:42.260
and carrying out
the computations

00:17:42.260 --> 00:17:43.410
in a very trivial way.

00:17:43.410 --> 00:17:47.420
We see that the maximum
densities are 1/2 here, 5/4

00:17:47.420 --> 00:17:52.020
here, 2 here, and 5/4 here.

00:17:52.020 --> 00:17:55.040
The idea is going to be
now that what we can now do

00:17:55.040 --> 00:17:57.060
is we know that the area
of each of these pieces

00:17:57.060 --> 00:17:59.770
is 1/2 by 1/2,
which is a quarter.

00:17:59.770 --> 00:18:02.490
I will now assume
two different plates.

00:18:02.490 --> 00:18:05.500
One will be the plate
which has the same shape

00:18:05.500 --> 00:18:07.880
as this, but in
this quadrant here,

00:18:07.880 --> 00:18:10.990
has a constant density of
0; in this quadrant here,

00:18:10.990 --> 00:18:14.130
has a constant density of
1/4; in this quadrant here,

00:18:14.130 --> 00:18:15.850
a constant density of 1/2;

00:18:15.850 --> 00:18:19.050
and in this quadrant, a
constant density of 1/4.

00:18:19.050 --> 00:18:21.310
I can now compute
that mass, and I

00:18:21.310 --> 00:18:24.960
know that mass must be less than
the mass that I'm looking for,

00:18:24.960 --> 00:18:27.700
because I have put the
lowest possible density

00:18:27.700 --> 00:18:28.830
in each quadrant.

00:18:28.830 --> 00:18:31.340
Correspondingly,
if I now compute

00:18:31.340 --> 00:18:34.020
the mass of another plate
which has the same shape,

00:18:34.020 --> 00:18:37.970
but in which each quadrant has
the density of the greatest

00:18:37.970 --> 00:18:39.980
value of the density
in each quadrant,

00:18:39.980 --> 00:18:42.930
that must give me an
over-approximation.

00:18:42.930 --> 00:18:46.100
And so my mass is caught
between these two.

00:18:46.100 --> 00:18:48.750
And just to let you get an idea
of what this notation means,

00:18:48.750 --> 00:18:53.500
noticed the delta a sub
i,j in this case is 1/4.

00:18:53.500 --> 00:18:56.810
For i equals 1, 2, and j equals
1 and 2-- in other words,

00:18:56.810 --> 00:19:03.060
our four elements of area are a
sub 1, 1, a sub 1, 2, a sub 2,

00:19:03.060 --> 00:19:07.330
1, a sub 2, 2, where if you
want to do these in order,

00:19:07.330 --> 00:19:11.330
just let i be 1 and 2
here, j be 1 and 2 here.

00:19:11.330 --> 00:19:17.010
For example, the maximum
density in the first row,

00:19:17.010 --> 00:19:19.700
second column-- well, I'm
reading these the wrong way

00:19:19.700 --> 00:19:20.200
here.

00:19:20.200 --> 00:19:21.290
The (i, j)-th.

00:19:21.290 --> 00:19:25.860
First row, second column,
up this way, is 5/4.

00:19:25.860 --> 00:19:31.330
The minimum density in that
same quadrant is 1/4, et cetera.

00:19:31.330 --> 00:19:33.300
At any rate, what I'm
saying is if I now

00:19:33.300 --> 00:19:35.840
take the fact that
for constant density,

00:19:35.840 --> 00:19:39.510
the mass is the
density times the area,

00:19:39.510 --> 00:19:42.390
a lower approximation
to my mass is

00:19:42.390 --> 00:19:45.520
obtained by adding up all
of my constant densities

00:19:45.520 --> 00:19:47.680
and multiplying by the area.

00:19:47.680 --> 00:19:51.540
In other words, 0 times 1/4 plus
1/4 times 1/4 plus 1/2 times

00:19:51.540 --> 00:19:53.830
1/4 plus 1/4 times 1/4.

00:19:53.830 --> 00:19:56.320
That sum comes out to be 1/4.

00:19:56.320 --> 00:19:59.120
Correspondingly,
the upper sum is

00:19:59.120 --> 00:20:01.770
what I get by taking each
of the constant densities

00:20:01.770 --> 00:20:05.860
and multiplying it by
the constant area, 1/4.

00:20:05.860 --> 00:20:08.810
That leads to a sum of 5/4.

00:20:08.810 --> 00:20:10.840
And therefore, no
matter what else I

00:20:10.840 --> 00:20:13.400
know about the mass
of my region R,

00:20:13.400 --> 00:20:19.150
I know that it must be more
than 1/4, but less than 5/4.

00:20:19.150 --> 00:20:22.710
And the idea is that by
putting on the squeeze,

00:20:22.710 --> 00:20:24.820
by making more and
more subdivisions,

00:20:24.820 --> 00:20:29.720
I can hopefully find the
exact value of M sub R.

00:20:29.720 --> 00:20:33.420
But finding this exact
value involves a very, very

00:20:33.420 --> 00:20:35.430
difficult computation.

00:20:35.430 --> 00:20:39.030
Namely, I divide
this up into m times

00:20:39.030 --> 00:20:41.880
n little elements of area.

00:20:41.880 --> 00:20:43.535
I pick a point in each region.

00:20:46.470 --> 00:20:50.290
The density at a particular
point that I've picked out

00:20:50.290 --> 00:20:53.250
in that region, which I'll
call c_(i, j) comma d_(i,

00:20:53.250 --> 00:20:58.240
j) is c_(i, j) squared
plus d_(i, j) squared.

00:20:58.240 --> 00:21:03.660
This would give me the
mass of a particular piece,

00:21:03.660 --> 00:21:05.650
and I add these up
as i goes from 1

00:21:05.650 --> 00:21:08.010
to n, as j goes from
1 to n, and take

00:21:08.010 --> 00:21:11.620
the limit as the maximum delta
X sub i and the maximum delta y

00:21:11.620 --> 00:21:13.740
sub j approach zero.

00:21:13.740 --> 00:21:17.340
And this, in most cases
is very, very difficult.

00:21:17.340 --> 00:21:19.680
And in fact, even in this
relatively simple case,

00:21:19.680 --> 00:21:21.780
what do I mean by
relatively simple case?

00:21:21.780 --> 00:21:23.620
In terms of a
region, what could be

00:21:23.620 --> 00:21:27.170
more straightforward than
our little square, R,

00:21:27.170 --> 00:21:27.760
that we chose?

00:21:27.760 --> 00:21:30.790
In general, we'll have
more complicated regions.

00:21:30.790 --> 00:21:34.860
Let me summarize today's lesson,
essentially, by showing you,

00:21:34.860 --> 00:21:37.780
in particular, what it
is that we're talking

00:21:37.780 --> 00:21:39.700
about in terms of double sums.

00:21:39.700 --> 00:21:41.570
Namely, what we're
going to do is

00:21:41.570 --> 00:21:45.670
we'll assume that R is a
reasonable subset of two

00:21:45.670 --> 00:21:46.860
dimensional space.

00:21:46.860 --> 00:21:50.180
And by reasonable, I mean
that it's not actually

00:21:50.180 --> 00:21:53.320
a kinky type of thing that
leads to pathological problems.

00:21:53.320 --> 00:21:56.740
The technical words are things
like R is simply connected

00:21:56.740 --> 00:21:57.630
and things like this.

00:21:57.630 --> 00:21:59.470
We'll talk about
that in more detail.

00:21:59.470 --> 00:22:01.330
What I'm saying
is let me imagine

00:22:01.330 --> 00:22:06.780
that R is a fairly well
defined region in the xy-plane.

00:22:06.780 --> 00:22:08.620
Let me assume that
I have a mapping

00:22:08.620 --> 00:22:13.490
f that carries R into
E. What does that mean?

00:22:13.490 --> 00:22:17.580
f is a real-valued function
of two independent variables.

00:22:17.580 --> 00:22:20.600
Namely, R is a
two-dimensional space.

00:22:20.600 --> 00:22:24.590
So a typical element in
R is a two-tuple x comma

00:22:24.590 --> 00:22:27.350
y, and what we're saying
that f maps x comma

00:22:27.350 --> 00:22:29.450
y into some number.

00:22:29.450 --> 00:22:32.760
Remember E, E^1, if you
want to call it that,

00:22:32.760 --> 00:22:34.740
is the set of real numbers.

00:22:34.740 --> 00:22:38.280
So the idea is let's-- what
we can say now is this.

00:22:38.280 --> 00:22:41.060
Let's take this region
R. Let's partition it

00:22:41.060 --> 00:22:46.520
into a mesh of little
rectangles here.

00:22:46.520 --> 00:22:50.390
Let's, in the (i, j)-th
rectangle-- namely,

00:22:50.390 --> 00:22:54.190
the rectangle whose dimensions
are delta x_i by delta y_j--

00:22:54.190 --> 00:22:58.370
let's pick a particular
point, and to identify it,

00:22:58.370 --> 00:23:01.900
we'll call it c sub
i, j comma d sub i, j.

00:23:01.900 --> 00:23:05.260
Let's compute f at that point.

00:23:05.260 --> 00:23:09.290
Then we'll compute
f times this area,

00:23:09.290 --> 00:23:12.790
and we'll sum this thing
up as i goes from 1 to n,

00:23:12.790 --> 00:23:15.310
and j goes from
1 to m, and we're

00:23:15.310 --> 00:23:19.480
going to take a limit as this
partition gets as fine as we

00:23:19.480 --> 00:23:21.060
wish, meaning the
size of the pieces

00:23:21.060 --> 00:23:24.210
get arbitrarily close
to zero in area.

00:23:24.210 --> 00:23:26.600
Both in the x and
in the y direction.

00:23:26.600 --> 00:23:29.820
Notice, by the way,
that because this is not

00:23:29.820 --> 00:23:33.790
a rectangular region, pieces
of rectangles are left out.

00:23:33.790 --> 00:23:36.120
And that from a
computational point of view,

00:23:36.120 --> 00:23:38.330
as you make this refinement
greater and greater,

00:23:38.330 --> 00:23:41.530
this is why the theory
is so voluminous on

00:23:41.530 --> 00:23:43.310
this particular type of topic.

00:23:43.310 --> 00:23:45.370
You must show that
you're squeezing out all

00:23:45.370 --> 00:23:46.670
the error and things like this.

00:23:46.670 --> 00:23:49.760
But conceptually, all that
we're really doing here

00:23:49.760 --> 00:23:50.940
is the following.

00:23:50.940 --> 00:23:53.690
What we do is we pick
a particular point

00:23:53.690 --> 00:23:57.280
in a particular one
of these rectangles.

00:23:57.280 --> 00:24:00.192
We evaluate f at that
particular point,

00:24:00.192 --> 00:24:01.650
which exists because
we're assuming

00:24:01.650 --> 00:24:03.214
that f is defined on this.

00:24:03.214 --> 00:24:05.130
And if you're having
trouble visualizing this,

00:24:05.130 --> 00:24:07.510
f has nothing to do
with this boundary.

00:24:07.510 --> 00:24:11.360
F is a function which maps
each point into a number,

00:24:11.360 --> 00:24:15.240
so graphically, you can think
of f as being a surface here.

00:24:15.240 --> 00:24:19.260
Namely if we visualize
f of x comma y

00:24:19.260 --> 00:24:22.410
as being mapped with
respect to the z-axis,

00:24:22.410 --> 00:24:24.750
we have a surface
coming out here.

00:24:24.750 --> 00:24:26.830
And in a sense, what
we're really trying to do

00:24:26.830 --> 00:24:30.860
is to find an approximation for
the volume of the region cut

00:24:30.860 --> 00:24:33.000
off by this cylinder.

00:24:33.000 --> 00:24:37.825
And whose top is the particular
surface z equals f of x, y.

00:24:37.825 --> 00:24:39.700
But forgetting about
that for the time being,

00:24:39.700 --> 00:24:42.090
mechanically, all
we do is we form

00:24:42.090 --> 00:24:44.200
this particular double sum.

00:24:44.200 --> 00:24:47.160
We form this double sum, and now
I've put the brackets in here

00:24:47.160 --> 00:24:50.000
to emphasize that what
this says is you first pick

00:24:50.000 --> 00:24:56.290
a fixed value of j, and sum
this over all i from 1 to n.

00:24:56.290 --> 00:24:59.240
And when you're all through with
that, the i no longer appears.

00:24:59.240 --> 00:25:02.340
You have something
that depends only on j.

00:25:02.340 --> 00:25:03.290
And you now do what?

00:25:03.290 --> 00:25:06.290
Sum this result up
over all values of j

00:25:06.290 --> 00:25:07.970
as j goes from 1 to m.

00:25:07.970 --> 00:25:11.460
And you then compute the
limit as the maximum delta x_i

00:25:11.460 --> 00:25:14.960
and maximum delta
y_j both approach 0.

00:25:14.960 --> 00:25:17.830
Now first of all, this
limit may not exist.

00:25:17.830 --> 00:25:20.297
Secondly, the limit
may exist, but it

00:25:20.297 --> 00:25:22.630
depends on the order in which
you're adding these terms.

00:25:22.630 --> 00:25:24.850
If after all, we saw
that we could change

00:25:24.850 --> 00:25:27.080
the sum of a simple
infinite series

00:25:27.080 --> 00:25:28.880
by changing the
order of the terms,

00:25:28.880 --> 00:25:30.850
certainly it shouldn't
be surprising

00:25:30.850 --> 00:25:34.160
that an infinite double
sum isn't affected--

00:25:34.160 --> 00:25:35.910
or is affected, it
shouldn't be surprising

00:25:35.910 --> 00:25:38.650
that it is affected if we
change the order of the terms.

00:25:38.650 --> 00:25:41.560
So we have one of two
ways of doing this,

00:25:41.560 --> 00:25:42.985
in terms of being systematized.

00:25:42.985 --> 00:25:45.920
Do we first sum in terms
of the x's, and then

00:25:45.920 --> 00:25:48.110
add up the y increments?

00:25:48.110 --> 00:25:49.350
Or the other way around?

00:25:49.350 --> 00:25:51.840
The key result is this, though.

00:25:51.840 --> 00:25:56.020
It's that this particular
limit will exist, and will

00:25:56.020 --> 00:25:59.180
be independent of what order
you add these things in,

00:25:59.180 --> 00:26:02.360
provided only that
the function f is

00:26:02.360 --> 00:26:04.780
at least piecewise continuous.

00:26:04.780 --> 00:26:07.070
In other words, that f
is a continuous function.

00:26:07.070 --> 00:26:08.650
And if it's not
continuous, it has

00:26:08.650 --> 00:26:12.020
breaks of only a finite
number of places.

00:26:12.020 --> 00:26:16.520
And the idea is that not
only will this limit exist,

00:26:16.520 --> 00:26:20.050
but in the manner analogous to
that definite integral, when

00:26:20.050 --> 00:26:23.460
this limit exists,
we denote it by what?

00:26:23.460 --> 00:26:25.830
We replace this by x and y.

00:26:25.830 --> 00:26:30.100
This gets replaced by dx
and dy, and the double sum

00:26:30.100 --> 00:26:33.070
gets replaced by
a double integral,

00:26:33.070 --> 00:26:38.360
and what we now do is indicate
what the region R is over here.

00:26:38.360 --> 00:26:40.480
In other words, in
terms of a summary,

00:26:40.480 --> 00:26:45.630
whenever I write
down this expression,

00:26:45.630 --> 00:26:48.639
it's an abbreviation for
a limit of a double sum.

00:26:48.639 --> 00:26:50.180
And what limit of
a double sum is it?

00:26:50.180 --> 00:26:55.300
It's the particular limit of
this double sum, if it exists,

00:26:55.300 --> 00:26:59.260
and that limit will exist if
f is piecewise continuous.

00:26:59.260 --> 00:27:01.560
Now, the key point that
I want to bring out here,

00:27:01.560 --> 00:27:05.240
and in fact, this will lead into
the lecture of next time, also,

00:27:05.240 --> 00:27:08.450
is that if I had never heard
of partial derivatives,

00:27:08.450 --> 00:27:11.150
it makes sense to talk
about this kind of a sum.

00:27:11.150 --> 00:27:14.080
In other words, note
that forming this limit

00:27:14.080 --> 00:27:17.550
does not require any knowledge
of partial derivatives.

00:27:17.550 --> 00:27:22.250
In the same way that finding
infinite sums in calculus one

00:27:22.250 --> 00:27:26.700
could be done independently of
any knowledge of a derivative,

00:27:26.700 --> 00:27:30.460
this particular concept,
this particular limits

00:27:30.460 --> 00:27:34.660
makes sense even if we've never
heard of partial derivatives.

00:27:34.660 --> 00:27:37.690
And what does this
double sum represent?

00:27:37.690 --> 00:27:40.100
Just by way of a
quick summary again,

00:27:40.100 --> 00:27:43.310
there are two very specific
physical interpretations

00:27:43.310 --> 00:27:44.540
we can give this.

00:27:44.540 --> 00:27:47.350
One is that this
double sum represents

00:27:47.350 --> 00:27:50.660
the density of the plate
whose shape is the region

00:27:50.660 --> 00:27:56.320
R-- the mass of the plate
whose shape is the region R,

00:27:56.320 --> 00:27:59.340
and whose density at
the point x comma y in R

00:27:59.340 --> 00:28:01.540
is given by f of x, y.

00:28:01.540 --> 00:28:05.070
In other words, this part
here controls the density,

00:28:05.070 --> 00:28:08.270
and this part here controls
the shape of the region.

00:28:08.270 --> 00:28:12.000
A second interpretation is that
this limit, when it exists,

00:28:12.000 --> 00:28:14.620
represents the volume
of the right cylinder.

00:28:14.620 --> 00:28:17.130
That means a
cylinder is obtained

00:28:17.130 --> 00:28:20.200
by tracing over the
region R with a line

00:28:20.200 --> 00:28:22.000
perpendicular to the xy-plane.

00:28:22.000 --> 00:28:26.990
That particular cylinder,
whose top is the surface z

00:28:26.990 --> 00:28:30.230
equals f of x, y.

00:28:30.230 --> 00:28:32.770
Here are, then, the
two interpretations

00:28:32.770 --> 00:28:33.800
that we give this.

00:28:33.800 --> 00:28:35.900
One is it can be a mass.

00:28:35.900 --> 00:28:38.400
The other is it can be a volume.

00:28:38.400 --> 00:28:41.880
The key point is that to
evaluate these double sums,

00:28:41.880 --> 00:28:43.470
these limits of
double sums, or if you

00:28:43.470 --> 00:28:45.940
want to call them
infinite double sums,

00:28:45.940 --> 00:28:49.880
there is no necessity for
knowing partial derivatives.

00:28:49.880 --> 00:28:53.450
However, one might expect that
in the same way that there

00:28:53.450 --> 00:28:56.160
was a connection
between single infinite

00:28:56.160 --> 00:29:00.590
sums and ordinary
antiderivatives,

00:29:00.590 --> 00:29:04.740
there should be a connection
between double or multiple sums

00:29:04.740 --> 00:29:08.610
and antiderivatives of partial--

00:29:08.610 --> 00:29:11.630
Well, the antiderivative
involving partial derivatives.

00:29:11.630 --> 00:29:13.810
It turns out that this
is indeed the case.

00:29:13.810 --> 00:29:16.870
We will talk about this
in more detail next time,

00:29:16.870 --> 00:29:20.310
but between now and next time,
what I would like you to do

00:29:20.310 --> 00:29:23.630
is to drill particularly
hard on these exercises,

00:29:23.630 --> 00:29:26.500
and make sure that you
become familiar and feel

00:29:26.500 --> 00:29:29.800
at home with the notation
that's used in denoting

00:29:29.800 --> 00:29:31.680
double and other multiple sums.

00:29:31.680 --> 00:29:34.260
At any rate then, until
next time, good bye.

00:29:38.340 --> 00:29:40.730
Funding for the
publication of this video

00:29:40.730 --> 00:29:45.580
was provided by the Gabriella
and Paul Rosenbaum Foundation.

00:29:45.580 --> 00:29:49.760
Help OCW continue to provide
free and open access to MIT

00:29:49.760 --> 00:29:54.178
courses by making a donation
at ocw.mit.edu/donate.