WEBVTT

00:00:00.000 --> 00:00:07.330
JOEL LEWIS: Hi.

00:00:07.330 --> 00:00:08.830
Welcome back to recitation.

00:00:08.830 --> 00:00:11.277
We've been talking about Taylor
series and different sorts

00:00:11.277 --> 00:00:12.860
of manipulations you
can do with them,

00:00:12.860 --> 00:00:14.757
and different examples
of Taylor series.

00:00:14.757 --> 00:00:16.340
So I have an example
here that I don't

00:00:16.340 --> 00:00:17.620
think you've seen in lecture.

00:00:17.620 --> 00:00:21.480
So this is the Taylor
series 1 plus 2x plus 3 x

00:00:21.480 --> 00:00:26.510
squared plus 4 x cubed plus
5 x to the fourth, and so on.

00:00:26.510 --> 00:00:29.680
I'm going to tell you that this
is a Taylor series for a fairly

00:00:29.680 --> 00:00:31.287
nice function.

00:00:31.287 --> 00:00:33.370
And what I'd like you to
do, is try and figure out

00:00:33.370 --> 00:00:34.700
what that function is.

00:00:34.700 --> 00:00:37.830
Now, I'm kind of sending
you off onto this task

00:00:37.830 --> 00:00:39.400
without giving
you much guidance,

00:00:39.400 --> 00:00:41.400
so let me give you a
little bit of a hint, which

00:00:41.400 --> 00:00:43.879
is that the thing
to do here, is you

00:00:43.879 --> 00:00:45.920
have a bunch of different
tools that you know how

00:00:45.920 --> 00:00:47.590
to manipulate Taylor series by.

00:00:47.590 --> 00:00:50.880
So you know how to, you know,
do calculus on Taylor series,

00:00:50.880 --> 00:00:52.520
you can take derivatives
and integrals,

00:00:52.520 --> 00:00:55.640
you know how to add and
multiply and perform

00:00:55.640 --> 00:00:57.040
these sorts of manipulations.

00:00:57.040 --> 00:00:58.900
So you might think
of a manipulation

00:00:58.900 --> 00:01:01.100
you could perform on this
Taylor series that'll

00:01:01.100 --> 00:01:03.800
make it a simpler expression,
or something you're already

00:01:03.800 --> 00:01:05.440
familiar with, or so on.

00:01:05.440 --> 00:01:10.390
So let me give you some
time to work on that.

00:01:10.390 --> 00:01:11.710
Think about it for a while.

00:01:11.710 --> 00:01:13.970
Try and come up with the
expression for this Taylor

00:01:13.970 --> 00:01:14.640
series.

00:01:14.640 --> 00:01:16.900
And come back, and we
can work on it together.

00:01:26.009 --> 00:01:28.300
So hopefully you had some
luck working on this problem,

00:01:28.300 --> 00:01:31.880
and figured out what the
right manipulation to use is.

00:01:31.880 --> 00:01:35.330
I didn't-- I kind of tossed
it at you without a whole lot

00:01:35.330 --> 00:01:37.830
of guidance, and I don't think
you've done a lot of examples

00:01:37.830 --> 00:01:38.880
like this before.

00:01:38.880 --> 00:01:42.400
So it's a little tricky.

00:01:42.400 --> 00:01:44.750
So what I suggested was
that you think about things

00:01:44.750 --> 00:01:51.930
that you know how to do that
could be used to simplify this.

00:01:51.930 --> 00:01:56.270
So looking at it, it's
just one Taylor series.

00:01:56.270 --> 00:01:59.430
So it's not clear that sort
of Taylor series arithmetic

00:01:59.430 --> 00:02:00.740
is going to help you very much.

00:02:00.740 --> 00:02:03.370
It's not obviously
a substitution

00:02:03.370 --> 00:02:06.657
of some value into some
other Taylor series

00:02:06.657 --> 00:02:07.740
that you're familiar with.

00:02:07.740 --> 00:02:12.220
It's not obviously, say,
a sum of two Taylor series

00:02:12.220 --> 00:02:15.150
that you already know very well.

00:02:15.150 --> 00:02:19.340
So those things don't,
aren't immediately,

00:02:19.340 --> 00:02:22.320
it's not clear where
to go with them.

00:02:22.320 --> 00:02:26.180
One thing that does
pop out-- well, OK.

00:02:26.180 --> 00:02:29.184
So let's talk about some of the
other tools that I mentioned.

00:02:29.184 --> 00:02:30.600
We have calculus
on Taylor series.

00:02:30.600 --> 00:02:32.650
So we have differentiation.

00:02:32.650 --> 00:02:35.350
And if you take-- so we see
here that the coefficients are

00:02:35.350 --> 00:02:37.410
sort of a linear polynomial.

00:02:37.410 --> 00:02:39.550
If you take a derivative,
what happens is,

00:02:39.550 --> 00:02:44.090
well, this becomes
2 plus 6x plus 12x

00:02:44.090 --> 00:02:46.080
squared plus 20x cubed.

00:02:46.080 --> 00:02:48.740
And those coefficients,
2, 6, 12, 20, those

00:02:48.740 --> 00:02:51.730
are given by a
quadratic polynomial.

00:02:51.730 --> 00:02:55.360
So that makes our life kind
of more complicated, somehow.

00:02:55.360 --> 00:03:00.450
But if we look at this, we see
that integrating this power

00:03:00.450 --> 00:03:02.110
series is really easy to do.

00:03:02.110 --> 00:03:05.014
This power series has a
really nice antiderivative.

00:03:05.014 --> 00:03:06.555
So what I'm going
to do, is I'm going

00:03:06.555 --> 00:03:11.980
to call this power series
by the name f of x.

00:03:11.980 --> 00:03:13.930
And then what I'd
like you to notice,

00:03:13.930 --> 00:03:18.730
is that the
antiderivative of f of x

00:03:18.730 --> 00:03:21.480
dx-- well, we can integrate
power series termwise.

00:03:21.480 --> 00:03:24.119
And so what we get is--
well, so all right.

00:03:24.119 --> 00:03:25.660
So I'm going to do--
I'm going to put

00:03:25.660 --> 00:03:27.700
the constant of
integration first.

00:03:27.700 --> 00:03:33.360
So the antiderivative of
this is c plus-- well, 1,

00:03:33.360 --> 00:03:36.660
you take its integral
and you get x.

00:03:36.660 --> 00:03:39.655
And 2x, you take its
integral, and you just

00:03:39.655 --> 00:03:41.550
get plus x squared.

00:03:41.550 --> 00:03:44.370
And 3 x squared, you
take its integral

00:03:44.370 --> 00:03:47.150
and you get plus x
cubed, and plus x

00:03:47.150 --> 00:03:51.640
to the fourth from the next
one, and plus x to the fifth,

00:03:51.640 --> 00:03:53.240
and so on.

00:03:53.240 --> 00:03:53.860
So, OK.

00:03:53.860 --> 00:03:57.260
So I put the c here, right?

00:03:57.260 --> 00:04:02.280
So any power series of this
form is an antiderivative

00:04:02.280 --> 00:04:04.020
of the power series
that we started with.

00:04:04.020 --> 00:04:08.444
Any power series of this
form has the derivative equal

00:04:08.444 --> 00:04:09.860
to thing that we're
interested in.

00:04:09.860 --> 00:04:12.550
And since we really
care about what f is,

00:04:12.550 --> 00:04:14.060
and not what its
antiderivative is,

00:04:14.060 --> 00:04:17.620
we can choose c to be
any convenient value.

00:04:17.620 --> 00:04:21.030
So I'm gonna, in
particular, look

00:04:21.030 --> 00:04:22.950
at the power series
where c is equal to 1.

00:04:22.950 --> 00:04:24.000
And why am I going
to make that choice?

00:04:24.000 --> 00:04:25.625
Well, because we've
seen a power series

00:04:25.625 --> 00:04:28.040
that looks very much like
this before, with the 1 there.

00:04:28.040 --> 00:04:38.730
So we know 1 plus x plus
x squared plus x cubed

00:04:38.730 --> 00:04:41.900
plus x to the fourth and so on.

00:04:41.900 --> 00:04:47.930
We know that this is
equal to 1 over 1 minus x.

00:04:47.930 --> 00:04:48.700
OK?

00:04:48.700 --> 00:04:50.770
So since we know that
this is the case,

00:04:50.770 --> 00:04:53.269
that means that our
power series, f of x,

00:04:53.269 --> 00:04:54.560
is just the derivative of this.

00:04:54.560 --> 00:04:56.240
That's what we just showed here.

00:04:56.240 --> 00:05:06.170
So, f of x is equal
to d over dx of 1

00:05:06.170 --> 00:05:15.350
over-- whoops that should
be 1 over 1 minus x.

00:05:15.350 --> 00:05:15.850
OK?

00:05:15.850 --> 00:05:17.224
And notice that,
you know, if I'd

00:05:17.224 --> 00:05:19.390
chosen a different
choice of constant here,

00:05:19.390 --> 00:05:21.930
it would be killed off
by this differentiation.

00:05:21.930 --> 00:05:24.340
So it really was irrelevant.

00:05:24.340 --> 00:05:26.620
So d over dx of
1 over 1 minus x.

00:05:28.695 --> 00:05:29.986
Yes, of d over dx of 1 minus x.

00:05:29.986 --> 00:05:30.250
And, OK.

00:05:30.250 --> 00:05:32.210
Well, this is an easy
derivative to compute.

00:05:32.210 --> 00:05:34.640
This is 1 minus
x to the minus 1,

00:05:34.640 --> 00:05:36.950
so you do a little
chain rule thing,

00:05:36.950 --> 00:05:41.000
and I think what you get
out is that this is 1 over 1

00:05:41.000 --> 00:05:43.720
minus x squared.

00:05:46.300 --> 00:05:51.760
So this gives me a nice formula
for this function f of x.

00:05:51.760 --> 00:05:54.420
If you wanted to check,
one thing you could do,

00:05:54.420 --> 00:05:58.080
is you could set about
computing a few terms, the power

00:05:58.080 --> 00:06:01.592
series for this function,
for 1 over 1 minus x squared,

00:06:01.592 --> 00:06:03.550
1 over 1 minus x quantity
squared I should say.

00:06:05.959 --> 00:06:08.500
So you could do that either by
using your derivative formula,

00:06:08.500 --> 00:06:10.004
which is easy enough to do.

00:06:10.004 --> 00:06:11.420
Another thing you
could do, is you

00:06:11.420 --> 00:06:15.430
could realize that this is 1
over 1 minus x times 1 over 1

00:06:15.430 --> 00:06:17.360
minus x, so you
could try multiplying

00:06:17.360 --> 00:06:20.610
that polynomial by itself and
that power series by itself,

00:06:20.610 --> 00:06:25.140
and see if it's easy to get
back this formula that we had.

00:06:25.140 --> 00:06:26.680
But any of those
ways is a good way

00:06:26.680 --> 00:06:28.650
to double check
that this is really

00:06:28.650 --> 00:06:32.450
true, that this function, that
this power series here, f of x,

00:06:32.450 --> 00:06:34.350
has this functional form.

00:06:34.350 --> 00:06:36.400
Now of course, I
haven't said anything

00:06:36.400 --> 00:06:39.090
about the radius of
convergence, but the thing

00:06:39.090 --> 00:06:43.800
to remember when you're doing
calculus on power series,

00:06:43.800 --> 00:06:47.150
is that when you take a
derivative or an antiderivative

00:06:47.150 --> 00:06:48.630
of a power series,
you don't change

00:06:48.630 --> 00:06:49.790
the radius of convergence.

00:06:49.790 --> 00:06:52.290
Sometimes you can fiddle with
what happens at the endpoints,

00:06:52.290 --> 00:06:56.310
but the radius of
convergence stays the same.

00:06:56.310 --> 00:06:58.340
So this is going
to be true whenever

00:06:58.340 --> 00:07:00.680
x is between negative 1 and 1.

00:07:00.680 --> 00:07:03.160
And in fact, this series
diverges at this endpoint.

00:07:03.160 --> 00:07:05.030
That's pretty easy to check.

00:07:05.030 --> 00:07:06.700
So there we go.

00:07:06.700 --> 00:07:09.300
So this is the functional
form for this power series.

00:07:09.300 --> 00:07:11.850
It's valid whenever x is
between negative 1 and 1.

00:07:11.850 --> 00:07:16.220
That's its range of convergence.

00:07:16.220 --> 00:07:19.270
And so here we have a cute
little trick for figuring out

00:07:19.270 --> 00:07:20.642
some forms of power series.

00:07:20.642 --> 00:07:22.850
And of course if you were
interested, so like I said,

00:07:22.850 --> 00:07:25.410
now you had a-- you could look
at the derivative of this,

00:07:25.410 --> 00:07:27.990
which I mentioned had some
coefficients that were given

00:07:27.990 --> 00:07:29.510
by some quadratic polynomial.

00:07:29.510 --> 00:07:32.519
Now that you have a functional
form, you could figure out,

00:07:32.519 --> 00:07:33.935
you know, "Oh what,
you know, what

00:07:33.935 --> 00:07:37.640
is the function
whose power series

00:07:37.640 --> 00:07:40.090
has that quadratic
polynomial as coefficients?"

00:07:40.090 --> 00:07:42.240
And you can do a whole
bunch of other stuff

00:07:42.240 --> 00:07:45.150
by similarly taking derivatives
of other power series

00:07:45.150 --> 00:07:47.430
that you're familiar
with, or integrals.

00:07:47.430 --> 00:07:49.236
So I'll stop there.