Lecture 35

IIT Kanpur July 201
20 followers
Sep 26, 2024
Hello students, in this video we are going to talk about Arjun function expansion. So,
let me start with some preliminaries and then we will see what exactly I would like to do
here. So, first of all see thing is in R n let us say that let E 1, E 2, E n. So, this be the basis
of R n, basis of R n, is it ok? And then if this is the case then every let us say x in R n, for
every x in R n. So, there exists, there is a unique choice, there is an unique choice,
unique choice of constants, of constants, ok? What is the constant? Alpha 1, alpha 2, alpha n,
for which, for which x equals to summation i equals to 1 to n alpha i E i, is it ok? I can
write it like this. Now, you see let us say that if such a representation is already given or let
us say if we given any, you know, eigenvalues, given a basis and if we want to find out what
the i, a i, alpha i's are, ok? How do you do it? See here what we are using is this. Now,
since E i's are orthonormal, ok, for all i between 1 and n, what you can do is this,
you can actually see what is the, you know, action of x on E i, is it ok? And what is that?
That will be nothing but the summation alpha i E i acting at E i, ok, i equals to 1 to n and that
we can use the linearity property of the dot product and we can write it like this, right?
So, this is nothing but summation of alpha i E i acting at, maybe I can write this thing as E g,
that will be better. So, basically I want to find alpha j that is. So, E, ok, this is fine and then
E j, is it ok? Now, if we do that, so i equals to 1 to n, ok? So, I am just taking any j between 1
to n, I am just writing it like this, ok? So, if I do that, then you see E i E j, essentially what
is happening is this, I should write right here, E i E j, you see for i equals to j it is 1,
otherwise it is going to be 0. So, that will actually give me the coefficient alpha j,
ok? And this holds for every j between 1 and n, right? So, hence the unique representation,
the unique representation given, you know, orthonormal basis, hence the unique representation
given an orthonormal basis, orthonormal, sorry, orthonormal basis, basis is x can be written as
summation, you see, the coefficients is nothing but x acting at E j, ok? And times E i, i equals
to 1 to n, clear? So, that is your unique representation and that is how you write any
element x in terms of its, you know, normal orthonormal basis that is, yeah. Now, you see,
we want to generalize this idea, is it ok? So, basically let us say given orthogonal set of
functions in some interval alpha beta, yes? With respect to some weight, let us say R x,
yes? Can we represent an arbitrary function? So, the question now is this question,
question is this. Given, ok, an ortho, so here it is orthonormal basis, right? So,
let us just start with the orthogonal basis. Of course, we can make it normal. So, let us start
with the orthogonal basis now, ok? So, orthogonal basis, basis and what is it? Phi n, ok? Phi n and
n in, n equals to 0 to infinity. So, the orthogonal basis given like this, ok? Orthogonal basis, yes,
in the interval alpha beta, in the interval alpha beta, is it ok? With respect to the weight,
with respect to some weight, ok? With respect to the weight R x, weight R x, is it ok?
Now, other thing is this, can we represent, can we represent another function, ok? Let us say f
of x, nice enough. Of course, you cannot expect that you cannot, you can represent any function
like that, but nice enough, let us say, yes? We do not know what nice enough means right now,
but let us just, I mean, you know it is a nice function, exponential, sine, cosine, that sort of
function, can we do that? That is the question. As an infinite series, as an infinite series,
infinite series, ok? Involving phi n's, involving phi n of x, is it ok? So, essentially f of x should
look like this, you see, summation n equals to 0 to infinity C n phi n of x, is it ok?
See, if you remember, why are we writing this as infinite series? Because the orthonormal sets
which we are getting from the Sturm-Liouville problems, we talked about it, right? So,
if you have a regular Sturm-Liouville boundary value problem, yes, you are going to get,
let us say, that for a distinct eigenvalue, right? So, you are going to get, you know,
eigenfunctions and those eigenfunctions are orthonormal to each other, right? Yes, that
we know, yes? And the thing is, this is like an infinite series, right? So, basically you see,
there is a way of actually producing an orthonormal basis. Now, the thing is,
the question is this, can we use that idea to represent any function, is it ok?
So, let us understand this, what happens. So, without, you know, going into, I mean,
proper mathematics, let us just quickly understand that how should we formally proceed, ok? So,
how to formally proceed? To formally proceed without convergence, without the question of
convergence. What does it mean? See, in the earlier case, ok, when we were in R n, it was
like a finite number of points, you see, this series, this series which we are talking about,
ok, summation alpha i E i, this is like from i to 1, i from 1 to n, right? So, it is a finite series.
In case of finite series, you really do not have to worry about convergence because it is like a
infinite kind of cases, then we have to talk about convergence, yes? We do not know whether
this actually converges or not. So, what we are going to do is basically, without the question
of convergence, let us just assume that all of this converges, they are nice, everything is fine
and can we find the coefficient C n, that is just the question, ok? So, what we do is this, see,
basically such that f x must look like summation n equals to 0 to infinity C n phi n x, can we do
that? Yes, let us just understand that, ok. So, what we do is let us just call this as 1,
1. So, we multiply, we multiply 1 with R of x times phi m of x, is it ok? We just multiply it
and integrate between alpha to beta and integrate between alpha to beta, is it ok? Yes, what do we
get? You see, if you multiply between, integrate between alpha to beta R of x phi m of x f of x,
right, dx, that is what we are going to get. I am multiplying it by R of x phi m of x, clear?
So, that will be written by alpha to beta summation n equals to 0 to infinity C n R of
x phi n of x phi m of x dx, is it ok? See, similar sort of thing what we are doing here,
here we are multiplying by R j, sorry, E j, right, that is my basis kind, the orthonormal basis. Now,
here I am multiplying with my orthonormal basis element phi n, but there is a weight involved,
right, R of x. So, basically we are incorporating that R of x also here, right. So, once we do that,
you see, the thing is here there is a problem. If I want to take this integral inside this
summation, there is a problem, but the thing is we are just assuming that we can do that,
let us just assume that, yes. So, once you do that, then what happens is,
you see, from here what can you write? This is summation n equals to 0 to infinity, yes,
and this integral is from alpha to beta, clear, R of x phi m of x phi n of x dx, yes,
and that is nothing but alpha to beta R of x phi m of x f of x dx, is it ok, yes.
And by the way, what about this? This is nothing but, you see, this is a, let us say,
this is an orthonormal basis, right, phi n's are orthonormal. So, basically for I n not equals to
m, we have basically, we have the integral alpha to beta, it is orthonormal with respect to the
weight R, right. So, basically R phi m phi n, the integral of that is going to be 0 for all n not
equals to m, yes. So, the only part which is remaining is m equals to m. So, in that case,
it is just the R of x norm of phi m square, yes, between alpha to, ok, one second, sorry,
sorry, it is not R of, sorry, I made a small mistake somewhere, right, that C m is, I did
not write here, I am really sorry about it, let me just write it, this C m should be there,
no, C m, C m, let us just write it C m, ok. Now, you see, the thing is, this is the only
thing which is remaining is C m times norm phi m square, is it ok. And then this part,
what we have is, this is alpha to beta R of x phi m of x phi n of x dx, is it ok, yes.
So, therefore, what we have, C m is nothing but alpha to beta R of x phi m of x phi n
of x dx by norm of phi m square, ok. Phi m is orthonormal, sorry, orthogonal, right.
So, phi m square is of course, not 0, yes, and if it is orthogonal, so basically that
will be 1 in that case, if you start with that, ok. So, in that case C m will be just
a to b alpha to beta R phi m phi n dx, ok. So, see, the thing is, therefore, see, if
this operation, the first operation, the integral and the sum, if we can, you know, interchange
that, so basically when can we do it. So, basically if the series, the series, what
is the series? Summation n equals to 0 to infinity C n phi n of x, ok, if it converges
uniformly, if it converges uniformly in alpha beta, in alpha beta, then, then the above
procedure is justified, procedure is justified. I hope it is clear to you, very, very simple,
very, very simple, ok. Now, the thing is this, see, the coefficient C n which you are going
to get, this C m which you are getting, those coefficients, if, see, if, let us say, all
of this is possible, so thus it is convergent, see, convergent uniformly and everything is
fine. Then, what we are going to do is, we are going to call this thing as the Fourier
coefficient, Fourier coefficient of the function f of f of x, is it ok, right.
So, now let us look at a problem and then we will go from there, ok. So, first start,
let us doing, start doing a problem. So, exercise example 1 is there. So, you see, we know that
the set, this set 1 cosine n pi x by L sin n pi x by L and L positive n is greater than
equal to 1, ok. This is orthogonal, this is orthogonal with respect to, what is the weight,
r x equals to 1, r x equals to 1, ok. First of all, how do you show this is orthogonal?
Of course, you can take the integral in equation, that is going to be 0 for n not equals to
m or it is going to be not 0 for n equals to m, ok, that of course, you can do.
The other way of doing it is, basically you see, y double prime equals to y, you remember
that problem we did, y at the point, with the periodic solution, this thing, with the
periodic data, yes. And we have seen that these are the eigenvalues, right, yes, sorry,
this is the eigenfunctions. And we know from that theorem which we did in the earlier class
also that those eigenfunctions going to be orthogonal. So, you have this set which is
orthogonal, you can do that, ok. So, this is orthogonal in minus L to L. So, we did
it for minus pi to pi, but for minus L to L, this is what you are going to deal with,
ok. Now, of course, then what happens is minus
L to L, minus L to L that is, ok, cosine square n pi x by L d x, this is nothing, but 2 L
and L, ok. This is for n equals to 0 and this is for n greater than equal to 1, yeah. And
minus L to L sin square n pi x by L d x, this is L for n greater than equal to 1. What about
n equals to 0? It is basically 0. So, you do not worry about it, ok. So, in that case
you see by going by this earlier idea, what happens is this f of x. So, given any f, yeah,
so see we define something like this. So, the Fourier series, the Fourier series, Fourier
series of f of x, f of x is defined to be, is defined to be f of x, ok, it will look
like this, a naught by 2 plus summation n equals to 1 to infinity a n cosine n pi x
by L plus b n sin n pi x by L, is it ok? We can write it like this, where let me just
write it what is a naught, a n, a n is 1 by L. What is, why 1 by L? Because you remember
this, this part is there phi m square. What is phi m square? They are like sines and cosines,
right, ok. So, a n is basically that cosine part. So, basically that 1 by L thing is there,
you see phi n square, ok. So, phi n square cosine square. So, it is, let me write it,
1 by L f x cosine n pi x by L d x, ok, n greater than equal 0, is it ok?
Let me just once check if everything is fine what I wrote.
a n is 1 by L, yes this is fine, minus L to L. And let us write b n. So, if you can see,
if we do the exact same procedure, b n, this, this b n's will be in respect to, you know,
sin, ok. So, in that case it will be again 1 by L sin and this is minus L to L and it will
be just a multiplication. So, let me write it. So, it is sin of n pi x by L d x, ok, b n. So,
that is there. Now, the thing is this, see the thing is, what we will do is this. So, this is
fine. So, if you do it like this, you can just write down the Fourier series of f of x. So,
by the way, this Fourier series in particular is called a trigonometric Fourier series, is it ok?
Now, the thing is this. You see, as I told you already that all of this is valid provided the
Fourier series converges, ok. So, how are you going to show that this Fourier series actually
converges, ok? And so, for this, you know, seeing this theory to work, what we are going to do is,
we are going to assume. So, the convergence of Fourier series, let us just write it like this.
We are going to talk about the convergence of the Fourier series, Fourier series, ok. So,
see the thing is, first of all, we are going to start with some phi n's. So, let phi n,
n equals to 0, 1, 2 and f of x be, are piecewise continuous, piecewise continuous in alpha beta.
I know, I hope you guys all know what piecewise continuous in alpha beta means.
It basically means that there are finite number of points, ok, but they may not be continuous,
except that everywhere they are going to be continuous. So, that is what we call
all the fields piecewise continuous, ok. Now, you see, let and we will call, define
S capital N of x, this we are going to define it as the sum of summation n equals to 0 to capital
N, the first n plus 1 term of C n phi n of x, is it ok, yes, and we donate it by capital S n of x.
And now, you see, we consider, we consider the difference, we consider the difference
S n of x minus f of x, this difference for large values of N, large values of capital N, is it ok.
Now, you see, if, if given epsilon greater than 0, given epsilon greater than 0,
there exists a integer N epsilon, which is positive, such that mod S n of x minus f of x
can be made less than epsilon, then the Fourier series, the Fourier series converges uniformly,
uniformly, uniformly to f of x for all x in alpha beta, is it ok.
So, the idea is this, see, essentially what we are saying is this, we are, we were approximating
f, yes, and how are we doing, doing it, we are going to use f of phi x, in such a way that we
are basically taking the finite sums and after that we are actually saying that for large N of N,
what happens is this, the difference between S n and f, it can be made arbitrarily small,
right, and if that holds for all x, that is important, then we say it converges uniformly,
is it ok. And of course, if n depends on, so let me make a small remark here.
So, you guys already know this, but I, I still wrote it in case of,
since there should not be any confusion here about the definition. So, what is the remark,
see N, the capital N, here depends only on epsilon, depends only on epsilon, only
on epsilon and not on x, clear, and not on x, not on x. So, if it depends on x also,
then we basically say this is a point wise convergent, right.
So, now let us look at a definition, let me just write down the definition and then I will
try and explain what are you, what I am trying to do here. So, let us write down a definition.
So, we are going to start with some function phi n of x. So, let phi n of x,
n greater than equals 0, of course natural, yes, and phi of x are piecewise continuous,
piecewise continuous in alpha beta. We say, we say the sequence phi n of x,
this converges in mean, this is very important, converges in mean, this is the,
see there are different kind of convergences, this convergent is convergent in mean to phi of x,
to phi of x with respect to, with respect to the weight function R of x, the weight function,
weight function R of x. See, here I am incorporating R of x,
if you want you can just take R of x to be 1, it is not a problem. In alpha beta,
if this is the meaning, see what we are saying is this, the norm of phi n minus phi, is it okay,
that if you take the square of that norm and then you take the limit n tends to infinity,
that is, let me just write down what is the definition, that is given by limit n tends
to infinity, what is this norm, this is nothing but the integral alpha to beta R of x phi n of x
minus phi of x square dx, this is going to be 0. Yes, we talked about this sort of norms and all,
right, yes, if you remember in the first lecture itself we talked about it, so we are using that
norm. So, this norm is defined by integral R phi square, yes, so if this goes to 0,
then we say that the, you know, the Fourier series converges to f of x essentially, okay.
So, therefore, the Fourier series, the Fourier series converges in mean, in mean to f of x,
f of x provided, provided limit n tends to infinity, okay, and alpha to beta R of x
S n of x minus, okay, you do not have to put a mod here, it is basically the same,
square is there here, minus f of x square dx, square dx, this should go to 0, is it okay.
Let me call this converges and the converges in mean, okay. So, now see, is it okay, now you see
the thing is we will see a small thing that what happens, okay, if there is a possibility of
representing f in terms of a different series, is it okay. So, basically let f of x, okay,
f of x is represented, it is represented by summation n equals to 0 to infinity d n phi n
of x, okay, where phi n are not necessarily, are not necessarily Fourier series coefficients, okay,
Fourier coefficients, coefficients of f x, is it okay, you understand what I am trying to say,
right. See, the thing is it may happen that there is a different representation which is basically
not a Fourier representation and we can write it like this, okay. So, now, and we want to see what
happens if we do something like this, okay. So, now, you know, define, let us see what happens.
Another finite sum p n, what is the sum? It depends on x of course, and then d naught d 1
d n, okay, just to specify this particular thing that it depends, the d n is a finite sum
with respect to the coefficients d 1, d 2, d n, right, and that is given by summation n equals
to 0 to infinity d n times phi n of x, okay. Now, and let us say we define capital E, sorry,
E capital N, okay, that quantity is nothing but the norm of T n minus f, okay, we want to see what
is the difference, T n minus f, the norm, if we calculate it, let us just call that T n, I want to
calculate what that is, okay. Now, you see, since phi n's are orthogonal,
this way is given to us, right, orthogonal, then what is E n square? Let us just calculate what is
E n square. E n square is nothing but summation T A capital N minus f square, is it okay,
and that is given by, by definition what is it? It is alpha to beta r of x and then
summation d n phi n of x, n equals to 0 to infinity minus f of x square d x, is it okay?
Yes, we can write it like this. Now, you see, we can just break it up, yes,
let us just break it up and write it. So, this is nothing but summation n equals to 0 to infinity,
okay, d n square, d n square alpha to beta r of x phi n square, phi n square and d x,
the first term, right. What about this last term? Let us just write down the last term now, plus
alpha to beta r of x f square of x d x, the last term and then the middle term,
two heavy term, minus two times summation n equals to 0 to infinity d n integral alpha to
beta r of x phi n of x f of x d x, okay. Let me just check once if everything is fine or not,
yes, this is okay, right, this, this seems everything is fine, okay.
Now, you see what exactly is this term? This is nothing but phi n square, right. So,
this is nothing but summation, I am not writing 0 to infinity, it is always, sorry,
what am I doing? Sorry, sorry, this is not 0 to infinity, right, this is 0 to n,
sorry, I am doing it with e n, so this is 0 to n, that is why I can change the limit, sorry, yes,
this is my mistake, very sorry about it, 0 to n, yes, okay. So, 0 to n I am not writing,
okay, so that is given because it is a finite sum, I can take the integral inside, so it is
not a problem, yes, okay. So, this is d n square, d n square and phi n square, right, that is just
this thing, definition minus 2 times summation d n C n norm of phi n square, okay, plus f square,
is it okay? So, I will just write down everything in terms of norm, so that is given by
and of course, the integral is from 0 to n, okay, let me just write it 0 to n, 0 to n, okay.
So, if this is true, you see I can write this as summation 0 to n, this is nothing but norm of
phi n square d n minus C n square, okay, d n minus C n square minus sum n equals 0 to n
phi n square C n square plus norm phi n f square, yes, I hope everything is fine,
let me just once check, yes, this is fine, yes, 2 times this and plus f square, I think it is okay,
yeah. So, you see if we can write it like this, now the thing is this, you can see that the
quantity E n, E capital N, when is this least, if d n equals to C n, okay.
So, therefore, capital E n, this quantity is least, so basically the norm of capital E n
is least, least if C n equals to d n for n equals to 0, 1, 2, right, yeah and therefore,
therefore, see the thing is what we proved is the following, let me just write it here.
So, we proved this theorem, theorem, we proved the theorem, let me just write down the theorem
first. So, for a given non-negative integer n, so given n in N, okay, the best approximation,
the best approximation, okay, in mean, very important in mean to a function f of x,
f, okay, x by an expression, what sort of expression, 1, by an expression of the form
summation d n phi n x, okay, is obtained, is obtained when the coefficients, when the coefficients
d n are the Fourier series, are the Fourier series of f of x, of f of x.
I hope everything is clear here, yeah, see what we did, this part is this, f x we wrote it as
summation d n phi n, d n is not necessarily the Eigen, sorry, not necessarily the Fourier
coefficient of f of x, okay and what we did is this, we calculated E n square which is nothing
but the finite sum with respect to those coefficients d n minus f square, okay. So,
in the mean, they will converge in the mean, okay, so you see we have to talk about this particular
term, yeah, that is what we are doing here. Now, you see E n square is nothing but summation d n
minus f square, right, which we can write it like this and now you see this is a positive term,
yeah, so when is E n least, E n is least if d n is equals to c n essentially, okay, so that is what,
so what we proved is this, that basically if you want to write f of x in terms of Fourier series,
looks like this, n equals to 0 to infinity d n f phi n x, then the best, you know, the best
candidate for the coefficients has to be the Fourier coefficients, is it okay, right.
Now, you see the thing is, let us say that here, let us just call this thing as 2,
let us just call this thing as 2. Now, into, into if d n equals to c n,
okay, it is a Fourier coefficient, they are Fourier coefficients, so n equals to
0, 1, capital N to obtain, to obtain, okay, one second, sorry, this I should write it as a finite,
we did it for 0 to N, capital N, right, for any capital N, we can do it, okay.
So, into if d n equals to c n, what do we obtain? We obtain that summation s n minus f square,
okay, this is nothing but norm of f square minus summation n equals to 0 to infinity
phi n square c n square, right, that is what we are going to get, see this T is will be
s in that case, because the coefficient d is will be e, d is basically e i's, okay, sorry,
d and c are basically same, d n's and c n's are basically same, so this is, this d n is basically
s n, right, okay. Now, so what is the norm of s n minus f square? So, that is nothing but this one
and this one and this is gone, because d n equals to c n, is it okay, that is why, that is what I
wrote it here, okay. So, now therefore, therefore what we can say is this norm of d n minus f
square, okay, that is nothing but summation 0 to capital N norm of phi n square d n minus c n
square, right, that is what we actually showed and then s capital N minus f square, we can write it
like this, yes, the last two terms is in terms of s n minus f, we can write it like this, okay.
Now, if we do that, see, since this is a positive term, all of these are positive,
okay, so we can say 0 will be dominated by norm of s n minus f, which is again dominated by
norm of T n minus f, right, this is a positive term, we can throw it away, so s n minus f square
will be dominated by T n minus f square and hence this inequality holds, okay.
Now, you see, if therefore, if summation n equals to 0 to infinity, now it is infinity, yes,
d n phi n x, if this converges uniformly, converges, sorry, not uniformly, in the mean,
in the mean to f of x, in the mean to f of x, okay, then the Fourier series converges, right,
then the Fourier series, because this is with respect to the Fourier series, right,
the Fourier series converges in mean, converges in mean to f of x, to f of x, okay,
but then let us just call this thing as 3, but then 1 to 3 implies,
implies that limit n tends to infinity, okay, summation 0 to capital N phi n square
d n minus c n square, this equals to 0, right, yes, see the Fourier series converges to f of x
in mean, right, so basically this is 0, yes, and of course, this also is given to be,
it converges, so that is 0, so basically we have limit n tends to infinity,
this particular term is going to 0, is it okay, yeah, and this is, and when is this possible,
see this is a positive term, this is a positive term, I am taking a finite sum,
and then I am taking the limit, I am telling that that is 0, it only happens if d n equals to c n,
therefore the above is only possible, above is possible, possible only if d n equals to c n,
okay, for n equals to 0, 1, 2, it goes on like this, is it okay, yes, and what we proved now
is essentially this thing, so let me write it here, what we proved is this theorem,
theorem 2, what is this theorem, so if a series of the form, series of the form
summation 0 to infinity d n phi n of x converges in mean, converges in mean to f of x, is it okay,
if it converges in mean to f of x, then the coefficients, then the coefficients,
coefficients d n must be the Fourier coefficients, coefficients of f of x, okay,
see we just saw, right, that if, I mean, if you have something like this, that d n converges to
f of x in mean, then it has to be the Fourier coefficient, d n has to be c n, so basically
that is what we showed, that if the series converges to f of x in mean, then it has to be
the Fourier coefficient, so the series converges, is basically the Fourier series, that is what,
okay, right, so now also let us see, also, yeah, so this is proved, this is proved,
now let us look at a different thing here, see the thing is from this inequality,
from this inequality, I do not know, I wrote it, yes, here, let us just call this star,
also from star, from star what do we have, we have 0 is s n plus 1 minus f is dominated by
s n minus f, right, see, see star, so basically if you are taking n plus 1, there is another term,
so I, we are basically have a decreasing sequence, right, non-increasing sequence that is,
okay, so thus, thus, you know, this sequence norm of s n minus f, okay, and n equals to 0,
1, this, this sequence is non-increasing and bounded, is non-increasing and bounded,
okay, is non-increasing and bounded, okay, below by 0, of course, below by 0, by 0, okay,
what does it mean, it means that it must converge, right, and if it converges to 0, then of course,
the Fourier series f x converges to, the Fourier series of f converges to f of x in mean, right,
see, of course, right, if s n minus f, this, this, this sequence, that basically converges,
it converges to 0, though, so basically in that case, we just say that the Fourier
series converges in mean, that is there, okay. So, also, also we will have this, see,
summation n equals to 0 to infinity phi n square, c n square is dominated by norm of f square,
is it okay, yeah, that is true. Now, because of this, right, see,
see, because of this, this is always dominated by, this always dominates 0, so basically
phi n square minus c n square is dominated by f of, f square, right, okay.
Now, you see, the sequence, if I, now, since the sequence, the sequence c n, what is c n,
c n, we will define it to be summation n equals to 0 to capital N, phi n square,
c n square, this sequence and n equals to 0, 1, 2, yes, is non-decreasing and bounded,
is non-decreasing and bounded, above by, bounded above by norm f square,
okay, it must converge, it must converge, clear, okay.
And therefore, what we have is summation, once again, I did a mistake again here,
it should be up till n, because all of those, the earlier calculations were up till n,
so then, it must follow that 0 to infinity, norm of phi n square, c n square must be dominated
by norm f square, yes, this is non-decreasing and this holds for all n. So, basically, I can write
it as, update the, sum is from 0 to infinity, right, we can do that. Now, you see, therefore,
what we can say is, therefore, the Fourier series, the Fourier series of f of x converges in mean to
f of x, converges in mean to f of x, yes, if and only if this is true, right, if and only if this
and this is true, right, summation f square is equals to summation n equals to 0 to infinity
phi n square, c n square, yes, I hope this is clear, this converges if and only if norm f
square has to be equals to this, yes, okay. So, now, this inequality what we wrote here,
this is very important, this inequality is called the Parseval's inequality.
So, this is called, how do I put it, maybe I can put it in a red, this inequality is called
Parseval's inequality, is it okay, yes, and this particular inequality, this inequality,
you see, if phi n's are basically, you know, orthonormal, so let me put it this way, small note,
if phi n's are orthonormal, orthonormal, then this thing implies that summation phi n square
0 to infinity is always dominated by f square, okay, and this is called Bessel's inequality,
Bessel's inequality, yes, I hope this is clear now, okay.
So, let us put everything in summary and I will put it in red, so that it becomes
much clearer what we did up till now, okay. So, let me put it as this theorem,
theorem. So, this is about Fourier, this thing, series. So, let we start with phi n of x,
okay, phi n of x and this is for n equals to 0, 1, this thing, okay, be an orthonormal set or
orthonormal set, is it okay. And let C n, C n be the Fourier coefficient, Fourier coefficient
of f of x, of f of x given by f of x is summation C n phi n of x, is it okay.
So, this is between n equals to 0 to infinity, yes, okay. Then the following holds,
then the following holds, following holds, what is that? So, it is A, the series, this series
summation 0 to infinity C n square, okay, converges, converges, this is what we showed,
right, you see, it is dominated by norm f square, so basically this series converges C n square,
and what does it mean? It means that limit n tends to infinity C n equals to limit n tends to
infinity integral alpha to beta R of x phi n of x f of x d x, this is going to be 0, okay.
See the thing is here that norm of phi n square is not there because phi n is taken to be
orthonormal, okay, right. Now B, the second thing is the Bessel's inequality holds,
Bessel's inequality, what we wrote this last page, this inequality holds, okay, Bessel's inequality
holds, holds. And C, what is the C which you have? You have that the Fourier, the
Fourier series of f of x, of f of x, okay, converges in mean, converges in mean,
mean to f of x, is it okay, if and only if the Bessel's inequality holds,
if and only if Bessel's inequality holds, is it okay, yes.
So, this is the idea of Fourier series. So, basically what we have is this, let me conclude
what we had did and why, where is it coming from. See, we know that if you have a regular
or a periodic sum level boundary value problem, then basically corresponding to the distinct
eigenvalues, we are going to get a, you know series, so basically the infinite series of
orthogonal functions, right, which we are basically, we call it an eigenfunction.
And the thing is, the thing is since they are orthogonal, okay, that set we can start with
and we can actually formulate a Fourier series out of it, yes, and then we can represent any
f with respect to those Fourier series according to whatever theory we just developed, okay.
So, that is the idea. So, basically what happens is if you want to produce, you know, different,
different Fourier series, where are you going to get the supply of f of x, those supplies are
actually, we can get it using the sum level boundary value problems, okay. So, with this,
I am going to end the video.
