Lecture 39: Fejer kernel on R

IIT Kanpur July 201
20 followers
1 Video View·Sep 6, 2024
[Music]
welcome back uh in the last lecture we
uh asked ourself that what is going to
be the analog of the fair kernel in
R so
by following
what information we have from from the
circle case for your series case
Define f of R for some positive R this
is equal
to R
sin pi x r
Square ided by
P
XR Square
If X is not equal to0 and is equal to R
if x isal to0 this is the analog
of the fair Kel and we are just copying
it what we have there in the series only
with the n part now we have for our
positive real
number okay
now why Fair kernel was
important because it has some nice
property and through that
property we were able to get back our
function by knowing the for year
coefficient and that is through the con
what we have learned there if f is a 2
pi periodic function and if we assumed
it to be
continuous then F convolution of KN
there we denoted the fair kernel this
converges to F
uniformly in the four year series case
so what properties of the fair kernel
there worked one thing is that they are
non- negative if the integral over the
period of 2 pi then that was going to
give us
1 and then away from
zero the integral of the fair
kernel which as n goes to Infinity this
uh goes to zero which means this is the
picture what we had done is uh for 2 pi
periodic function so
now this is some 2 n +
1 this is a positive function non
negative function and as n goes large
and large this portion the area is going
to contribute very
little that's what made Wonder to get
back F from F convolution of KN which is
in the form of 1 - mod Z by n f of n e^
i NX now let us see that similar kind of
property whether this holds here for our
Fair kernel or not unless and until that
happens then there is no reason to call
this
as the fair kernel for R in
R okay so the first thing first
clearly f r of X is greater or equal to
zero now we would like
to see what is the integral
of
DX this is equal
to uh
R * - Infinity to
Infinity sin pi r
x² ided by P R
x²
DX now make a a change of variable you
take
uh P RX to be
to be y therefore
this is uh 1 /
Pi - Infinity to
Infinity sin y by y whole Square
Dy and we know that I mean what is uh uh
the value of this that is pi so this
which is equal to 1 so this in to
compute this integral this is well known
and we must have encountered it earlier
otherwise one hand waving argument is
that if you define let us say i t this
is equal to minus infinity to Infinity
sin t y
by uh y whole squar Dy and then you take
I prime t
then this is going to give us basically
now assuming everything fine formally we
can by doing hand over so this is going
to be
sin 2T X by X
DX so if we make
uh a change of variable I mean so then
at uh you will get that it is equal to
Pi
a plus some
constant so therefore uh this is equal
to Pi a equal to T = to 1 you will get
that this is pi okay so so what we are
essentially this integral is a known
integral which takes the value pi
now third property what we need to see
is that if I
am going away from zero then if I take
my R to be large enough then the
integral of f r x should vanish that
means
is limit R goes to Infinity integral mod
y greater than Delta f r of y d y is
equal to0
true because that is one of the most
fundamental property what we have used
in the fair kernel business in the four
year
series
so let's uh compute this let
Delta greater than
0o so now this is limit R goes to
Infinity so let us uh do this uh mod of
X greater than Delta this is 1 by
R
uh then
p² sin squ PK
RX /
x² DX that is what the fair canel
is
now this one is lesser equal to 1 / < s
r mod x greater than
Delta uh sin s pix is always bounded by
1 1 so I have DX by
x² now away from the origin DX by x²
this is a finite number integral Delta
to Infinity 1 / x² is
uh uh - 1 /x from Delta to Infinity so
this is nothing but 1 / Delta and Delta
is positive so this is some constant
times
uh 1 / R into 1
/
Delta so this goes to zero as R goes to
Infinity therefore this property of the
fair kernel is also
true so
hence f is a continuous function f is
if continuous moderate
Decay what we have seen is that uh f is
UN will be uniformly continuous then F
convolution of f r this converges to F
uniformly the exactly if we follow the
uh same method just to recall what it is
so F convolution of f r of x minus of FX
so this is uh I we can write that this
is integral of f of x minus of y f r of
y d
y and minus
FX into one 1 in place of one I can
write that this is f this is the
technique what we used in the for year
series case so this is lesser equal to
minus infinity to Infinity mod of f ofx
- of y - F
ofx and then f r of Y
Dy now as f is continuous therefore for
Epsilon positive it is uniformly
continuous so Epsilon positive there
exist a Delta greater than Z
such that mod of f of x- of Yus of f
ofx is uh less than Epsilon
whenever mod Y is less than
Delta so therefore this integral I can
break it it into mod Y less than Delta
plus mod y greater greater or equal
to Delta now on the first I one use as
the continuity of f uniform continuity
pull out the Epsilon we left with the
integral of FR which value is one now in
this case as R goes to Infinity we know
that this is nothing but
I I will pull
out now the second part so this is less
than
Epsilon plus the second part I will pull
out 2 * supremum over mod of f of x such
that X belongs to R because this is this
is a bounded function and we are left
with
and just by the third property this goes
to zero as R goes to Infinity
exactly the same proof works if for this
Fair
kernel if our F has moderate Decay and
continuous okay so there you know the
fair L there it was looking as F
convolution of FN in the series
case is equal to summation over - n to
n 1 - mod J by n f hat of n e the power
2i I
NX so this is uh uh what was the fair
kernel convolution was looking like so
natural to see if we are talking about
the analog then whether this is going to
be true here in thisa case or
not so for that let's uh
do try to see that what will be the
analog the sum would be integral minus r
to R then 1 - mod Z to the by R then
this is f hat of Z and e to the power 2
pi i z x
DJ we must look that this is f
convolution of f in our
case okay
so
we compute this this
integral - 2 pi I
JX
DX uh the there are two x symbol so we
will take this out we'll put that
y j
y d y and e the power 2 pi I
JX
DJ now our functions are nice so we can
interchange the integral so this is
going to be minus infinity to Infinity F
of Y then integral minus r to r 1 - mod
Z by r
e to the power 2 pi
i x z xus of Y then this is DJ integral
then we have left with Dy so now
essentially what we need to if we can
show that this is f
if this is equal to f r of x minus of Y
then we can write this as F convolution
of capital FR at
X okay so that essentially asks us to
compute minus of R to r 1 - mod Z by R E
power 2 pi
ijx
DJ this obviously mod is there we want
to get rid of that we will get it is
1us J by r e to the power 2 pi I
JX
DJ and here this is minus of R 2
0 1 + J by r e the power 2 by I JX
DJ now
just compute this integral then this is
the first one would be e the^ 2 pi i
j 2 pi i
x z divided by 2 pi i x which is
evaluated at 0 to R so now this we can
write that this is r R minus of 1 at
zero this is going to give minus of one
if I compute with this then obviously we
are going to do the integration by parts
in the next so now in the integration by
parts the Zero part will get killed
therefore what we get here
minus
r e the power 2 pi I
RX ided
by 2 pi
IX and then 1 / R is there 1 / R and
then plus then the integral over 0 to
R e^ 2 pi I JX by 2 pi
iix into derivative of this is DJ
so this we can write this as if we do
the integral again so
this and then the plus the other one the
other one is going to be exactly the
same thing which uh would give 1 - e ^ -
2 Pi
ixr / by 2 pi
IX and then minus
uh e to the
power
uh so this quantity is e the^
minus this is uh minus of r e the power
- 2i I RX
by 2 pi IX and a factor of R is also
there plus the integral minus r to Z 0 e
to the power 2
pi
i j x / 2 pi
I
x² then this is oh 2 pi i
x then this is
DJ
now as you can see this this and this
gets cancelled so I have here 1 minus 1
by 2 pi IX here 1 by 2 pi IX this gets
cancel Then This this term and then this
term they also get cancell so what we
need to evaluate only this so now this
is going to be e the power 2 pi I JX / 2
pi I
x² and uh then at uh this is from 0 to R
and then this one is
uh e to the
power this is
uh yeah no this is plus e the^ 2 pi
iix Z / by 2 pi IX
Square then Min - R to
r0 so if we compute this then this is
going to give us e to the power so let's
write it in the next page this is going
to be equal to E power 2 piun
ix+ e^ - 2 pi I
RX minus of 2 - 1 - 1 is 2 and then this
is R * 2 < I
x² so this is going to be of course this
is uh this is
2
cos 2
pi
r
x - 2 by -
4r >
x² now this is equal
to 1 - cos 2x is 2 sin s x so now this
is minus of 4 sin s piun
RX divided by - 4 R PX Square this is
nothing but R * sin s <
RX divided
by P
x² If X is not equal to Z because we are
dividing it by X we are assuming that X
is not equal to Z now if you uh take
that F if x is equal to zero then you
are left with
X = to0 then - R to r 1 - mod Z by
R and
uh d z this obviously you will get to
get R so now what you have got
minus R2
r 1 - mod Z Z by r f hat of z e to the
power 2 pi I
JX
DJ this is nothing but so we have a
concrete
analogy with the for year
series okay so now successfully we have
done uh this um the fair kernel so
actually what we have seen I mean we can
expect uh
if
inversion when uh we have
uh a fact we are assuming a f to have
moderate Decay so it's natural to
uh uh think about what uh are the a
large class of function in which we can
do our forer analysis in a very uh
successful way uh
so in the next lecture what
we will do is that we will compute the
for year transform of
some um nice function and uh then by
looking at the behavior we will uh
kind of will
be we decide that uh uh what kind of
space we are looking for thank you
[Music]
[Music]
[Music]
