Lecture 38: Inversion formula and convolution

IIT Kanpur July 201
20 followers
Sep 6, 2024
[Music]
welcome back uh in the previous lectures
we have started doing forer analysis on
real line
r and we have assumed our function to
have a moderate decay in the sense that
f of x modulus is less equal to some
constant times by 1 +
x² now what we have seen is that F hat
is defined through e^ 2 pi I JX so this
is just to write the notation F hat at Z
this is equal to - Infinity to
Infinity f of x e^ - 2 pi i x z DX this
is for J and then what we have seen is
that F hat of Z is less or equal to some
constant that means f is a bounded
function and uh f
reman leema says that fat of
Z goes to zero as mod
Z tends to infinity and we are assuming
F to be continuous then actually fat is
also continuous in other words fat
belongs to c0 of R set of all continuous
function which banishes at Infinity
okay
so today uh we would like to address the
question that what is the resemblance of
this forer transform visis the forer
series suppose for example in particular
if we take F to be a
function uh which is supported on minus
pi to Pi then what can
what is the kind of the relationship
between the forer transform and the
forer series so can we recover some of
the result by using the for year series
techniques now towards that end let's
have a toy example so let's take our
F let FBA a continuous function
with compact
support uh that is and and support
of f is contained in some minus n to n
in some interval
that means uh f is a continuous function
if you have minus n to n so outside this
interval this is going
to
vanish zero and here it's a continuous
function so this is uh outside certain
finite interval this is
zero
and so let's
assume F hat of
Z is less or equal to some constant by 1
+ mod z s that means both F being a
continuous function obviously has
moderate
Decay and on top of of that we are
assuming that F hat also has moderate
Decay so now
for any L greater than
0
Define a
n of L this is equal to 1 by 2 L
integral
- L to l f of x e^ - 2 piun I
NX ided 2 L
DX
so now if uh remember that if we have a
2L periodic function so this is going to
be the foure coefficient of that two 2L
periodic function
uh so now we have if L
is greater than n then this I can
represent in terms of the 2 pi periodic
2 L periodic
function and hence we can get that a
NL this is equal to F hat of n by 2
L so as uh we know that
uh 1X 2 L
Factor
so here as you can see because we are
assuming that F hat has a moderate Decay
so summation over mod of f hat of n by 2
L this is
finite therefore the for year series is
an absolute summable series and it
converges to a continuous function at
every point that is what we have learned
from our foure series
course
so then we can write F ofx this is equal
to summation over a n
l e to the power 2 pi i n x by 2
L which is equal
to 1 by
2 summation n we can write actually
summation n from we can take it from
minus infinity to
infinity and uh F hat of n by 2 l e to
the power 2 pi i n x by 2 L
we have already assumed that f is a
continuous function uh and so now let us
try to look at what is the meaning of n
to n f of x
DX now this uh that means I am taking
minus n to n the meaning of this is that
we will take a partition
this is 1 / n so this is actually all
this partition is 2 N /
n
so and so take the partition
XJ minus n isal to X is less than
X1 so which is going up to xn which is
equal to n this is what the partition we
have
taken so therefore this
reman integral can be approximated by
remen Sum so what is the reman sum remen
sum would be 1 /
n summation over J from 0 to n f of
minus of n + 2 N J by n
each 1 is 2 N by n so this I can always
rewrite this as 1 / n summation over K
is = - Capital NN to
NN F of K by
n so this is going to approximate as n
goes to Infinity this is going to
approximate the integral a integral
minus n to
n so therefore that's what we will write
minus n to n f of
x DX this is equal to
limit n goes to Infinity 1 /
[Music]
N I can write that this is
NN F K by n
so
therefore minus infinity to Infinity F
ofx DX this is equal to to limit n goes
to Infinity so this we can write that n
goes to
Infinity 1 by n summation K is = -
Infinity to Infinity F of K by n this is
true for any continuous function
if F continuous then this we can write
for every continuous function so
therefore if I take my f in particular
to be minus infinity to Infinity F hat
of
z e to the power 2 pi iix z d z
you see uh we are assuming F hat to have
moderate Decay so we can Define this
integral uh as limit n goes
to Infinity minus n to n of this
function so this by our previous
observation in terms of this if we are
if we are taking a f of Z e^ 2 pi i x z
d z then this uh tells us that this is
limit n goes to
Infinity 1 / n summation over K from
minus infinity to Infinity F hat of K by
n and E the power 2 pi i z k by
n and
uh this what we have seen is that that
this is going
to be nothing but what we get is that
this is f ofx in our previous
observation so here this is essentially
what we are trying to do is that we are
getting some sort of the forer
inversion uh if we are assuming F and F
at they are nicely behaved moderate
Decay through by converting the problem
into the four year series and then we
are getting it back now one of the
fundamental
Tool uh while dealing with the foror
series either in the circle
or for finite abelan
group we uh have exploited the
convolution of two function a
lot so it's uh now assume
both f and
g uh
have moderate
Decay and continuous let's
take then
Define f convolution of g at X
this is equal to Min - Infinity to
Infinity f of x - y g of y d
y look at that this integral exist
because uh both f and g they have
moderate Decay so now in particular they
are bounded so if you push the modulus
inside then you can pull out the Bound
for G or F then mod of G is integrable
so there is absolutely no problem in
defining this integral now this
also I by the change of variable with
the REM integration what we have seen I
can write that this is G of xus of Y Dy
that in a sense is also you can write it
as G convolution of F at X
now both f and g they have moderate
Decay now one would like to ask this
whether this function f convolution g
also has moderate Decay or not so now in
order to see that we need to get F
convolution of g at X so this is less or
equal to integral minus infinity to
Infinity mod of f of x - of Y mod of G
of
Y
Dy now I break this integral into two
parts this integral is mod
Y is less or equal to mod x by
2 mod f of x - of Y
g
y Dy plus integral mod y greater than
mod x by 2
mod gy now look at for the first
integral let's name them 1 + 2 now for
one now mod one y as you can see that
mod Y is less or equal to mod x by 2 now
mod of x - of Y this is greater or equal
to mod x - mod Y now minus of mod Y is
greater or equal to minus of mod x by 2
so this is greater or equal to mod x -
mod x by 2 which is equal to mod x by
2 therefore mod of f of xus of Y this is
going to be less or equal
to as F has moderate Decay so what I
will what we are going to get by
a mod of x - of y² now mod of x- of Y is
greater or equal to
mod x by 2 therefore this is less equal
to a by 1 + mod x² by 4 so this
definitely is less or equal to some
constant ided 1 + mod
x² so there and so the mod of I is less
or equal to
some constant b 1 + mod x² to minus
infinity to Infinity mod of G of Y Dy
which is a finite number and which is
independent of X therefore the first
part we can see that it gives the
moderate Decay this is what we are
getting for the first
part now let's look at the second
part
now
for the second
part mod of Y is greater than mod x by 2
this is already given so now I
will use the moderate decay of G so this
is less or equal to some constant by 1 +
y² and now this is less or equal to
again B by 1 + mod x² therefore
2 is less or equal to B by 1 + mod x²
into integral minus infinity to Infinity
mod of f of x - of y d y by making a
change of variable we know that this is
equal to 1 + mod x² integral minus
infinity to Infinity mod of f of Y
Dy which is again independent of X
therefore what we have got is that the
second is also giving us the moderate
Decay therefore so in conclusion we have
F convolution of g at X modulus is less
or equal to some constant by 1 + mod
x² now one would like to uh ask uh that
what about if I have a convolution so
which has a moderate Decay so I can talk
about the forer transform of that
convolution and what's going to be the
forer transform of this F convolution G
in this case in the case of the circle
what we have seen is that f convolution
g hat at n is f of n into G of n the
similar result holds true in the setting
of finite Aban group now in this case
also this will be
standard argument F convolution of G hat
at Z this is minus infinity to Infinity
F convolution of G of x e ^ - 2 pi I JX
DX which is equal to minus infinity to
infinity and then I will open up the F
convolution of G this is f of x minus of
Y uh G of
Y Dy e to^ - 2i I
JX DX now both f and g both have
moderate decay
modulus of f in times the modulus of G
they are the finite so integral is
finite therefore I can interchange the
order of the integration so the moment
we change so we have got Infinity so
this is g y minus infinity to Infinity F
of xus of Y here I will put - 2i I
J xus of Y
DX now extra factor I am writing is e^ 2
pi i j y so I need to cancel
that and then this is
dy by making a change of variable for
the inner integral we have
f ofx e the^ - 2i I
JX DX then e the^ - 2i i j y
Dy this is equal
to minus infinity to Infinity G of y e
to the power - 2i i j y Dy and hit it
with F hat of Z because the inner
integral is going to give us F hat of Z
so this is nothing but F hat of J G hat
of Z
so that's also
true
now we have both f and g they are uh
continuous function therefore it exactly
following in the same fashion it is uh
one can so that f convolution g is also
continuous exactly the way we have
proved in the case of the
circle or 2 pi periodic function
so what you can do is that F convolution
of G of x minus f convolution g of Y
this is equal to minus infinity to
Infinity F of
xus uh let's say
s - F of y- of
S and then G of s d s the
mod now what one does is that if x is
close
to y then what we need to show that
this one is small so this is less equal
to minus infinity to Infinity mod of f
of x - of s - F of Yus of s mod GS
DS now if F given to us if it is
uniformly
continuous then we can use that if x
less X is close to Y then f of x - s
minus of f of y - s will be less than
Epsilon so what we know that by
continuity for Epsilon positive there
exist a Delta greater than 0o such that
mod of F of
x- of s- of f of Yus of s is less than
Epsilon
whenever mod of xus of Y is less than
Delta but as you can see that here the
integral uh here for every s i can my
Delta s Delta can depend on S so if in
instead if we have the uniform
continuity then
definitely we could pull out the
smallness of the difference outside the
integral then we will be left with minus
infinity to Infinity mode of
gsds so
this uh says that what we need is the
uniform continuity now we have only with
we're assuming that f is a continuous
function and with a moderate Decay now
if F has modate Decay you can easily see
that f of
x Which is less or equal to a by some
constant x² and this goes to zero as X
goes to mod x goes to
Infinity therefore now this function is
a continuous function although
and around Infinity this is what is
happening is that it will
be it goes to Z
zero that means F if continuous then
this is c0 of
R now
F so now here what you can say is that
mod of f
ofx is that for Epsilon positive if
there exist a r greater than Z this is
less than
Epsilon
whenever mod of X is greater or equal to
R now outside this outside minus r to R
we have 1 Epsilon it is less than
Epsilon so we don't need to bother about
the tail part of the function that is
itself is giving the uniform
continuity and uh now what we are left
with we are left with minus r to R and
this is a compact set so
f f
is
continuous
hence if I look at f as a map from minus
r to R then
then this
restriction will be uniformly
continuous so f
is uniformly continuous
therefore I can pull out the smallness
of f from from this
integral without having any problem and
then thus we get F convolution of G of x
minus F convolution of G of Y is less
than Epsilon * minus infinity to
Infinity mod of G of s
DS
whenever less than Delta so we can we
can choose our Epsilon because the
instead of Epsilon if we want to work
Epsilon
by minus infinity to Infinity mod GS DS
then this will give the smallness anyway
this is a finite integral minus uh uh
infinity to Infinity mode gsds is a
finite number and I can choose my
Epsilon to be small so that this
is so this this also says that with
assuming the continuity of F and then f
convolution g if both have moderate
Decay then f convolution g is also a
continuous
function okay so now to draw the
analogy uh further so there what we had
seen is that the fair
kernel was extremely important for us to
study for year
series and it's natural
to ask what is the equivalence of fair
Kel here thank you we'll discuss this in
the next lecture
[Music]
[Music]
[Music]
