Lecture 25: Proof of Weyl's criterion

IIT Kanpur July 201
20 followers
Sep 6, 2024
[Music]
welcome back uh in the last lecture we
are trying to prove the Wes Criterion
for EQ distributed sequence so we are
halfway through our proof on one side
that was uh
what we wanted to prove that uh
if limit n goes to Infinity 1 / n n = 1
to n e^ 2 pi i k
JN this is equal to zero for
all zero nonzero
integer then
the sequence JN is equ
distributed so in order to do that what
essentially what we have observed to so
some something is equ distributed what
we need to so need to
prove 1 / n summ n = 1 to n and Kai of
ab for a sub interval AB in 01 this is
this goes to 0 to 1 Kai of ab T DT this
is what we want to prove and what we
have achieved we have achieved that for
every one periodic
continuous
function
f
we
proved limit n goes to Infinity 1 / n
summation / n from 1 to n f of JN this
is equal to 0 to 1 ft DT now from The
Continuous function we would like to go
to the indicator function of Z One Step
function of so pictorially let us look
like
this I
have a b so indicator function of 01 is
a Ab is this and then zero here and
outside a zero here and one here that is
what the indicator function of this now
what I can do is that depending on my
a where is the position of a I can find
because a is strictly bigger than zero
therefore I can find some Epsilon where
this a minus
Epsilon this is a minus Epsilon and B +
Epsilon this is going to lie inside the
uh interval
01 so now I will take a continuous
function which is
this which takes the value at uh uh
value one in the interval AB and it
drops us that as we have seen it in
earlier occasion this uh area of
this this is of course is this two
triangle the year triangle the area is
going to be Epsilon plus Epsilon which
is 2
Epsilon that's the maximum what I will
get so now so
Define so we can and
find Epsilon
uh positive such that a minus Epsilon is
strictly positive and B+ Epsilon is
strictly less than
one that I can always get now if I take
from or rather this this Epsilon not so
any Epsilon less than Epsilon KN we have
a minus Epsilon and B + Epsilon this is
in
01 so now
Define f Epsilon
plus this is equal to of X is equal to 1
if x belongs to AB
and uh
0o if x uh does not belongs to a minus
Epsilon and B + Epsilon outside this
this is zero that is what I am
considering and this is a continuous
function and
continuous linear
on the interval B uh a minus Epsilon to
a
union uh B2 b +
Epsilon so as you can see that if I take
the integral 0 to
1 of this continuous function
tdt then this is going to be
a minus Epsilon to a + a to B
+ B2 b +
Epsilon and uh here in this this is f
Epsilon + f Epsilon + f Epsilon plus so
here this is equal to B minus a plus
this one would be less than definitely
one because it is linear so the the
entire thing is less than
Epsilon and this one is also so this is
equal to b - A + 2
Epsilon
similarly in a bigger picture my this is
01 this is my
ab so
now what I will do is so this is my a
minus Epsilon and then this is a +
Epsilon this is B minus Epsilon
so I will take one over here and then I
will drop
it
linear let us say this is B- 3 by 2
Epsilon then this is a + uh 3 Epsilon by
2 and it will take the value one in
between
and after this this is
zero so this F Epsilon minus of 1 T this
is equal to 1 on a + 3 Epsilon by 2 b -
3 Epsilon by 2
Z if uh T does not belongs to a minus a
+ Epsilon to B minus
Epsilon and
linear then clearly in this case
integral 0 to 1 F Epsilon minus of T DT
this is equal to B minus of
a minus of 2
Epsilon so by the construction what we
can see is that uh F Epsilon minus of T
this is is always less or equal to
indicator function of ab of T because uh
uh on a + 3 Epsilon by 2 to B minus 3
Epsilon by 2 it is taking the same value
now uh indicator function of ab is goes
up to takes the value one up to AB now
you are chopping it off which is less
than one so this is true for all T which
again is less or equal to F Epsilon plus
of T because there you are putting the
ear outside a small to small here there
so this is always true hence what do we
get from this so
therefore F Epsilon minus of this ZN is
less equal to Kai of ab
of
Zan Which is less or equal to F Epsilon
plus of JN therefore the summation / n
from 1 to n is less or equal to
summation over 1 to n is less or equal
to summation 1 to
n therefore if I divide 1 by n 1 by n 1
by n remember that the both F Epsilon
minus and F Epsilon plus they continuous
function so this is going to
be 0 to this
is this of course is going to be 0 to
1 F Epsilon
minus of T DT this is going
to goes to 0 to 1 F Epsilon + T DT now
this we have seen that this is B minus
of a + 2 Epsilon this is B minus of a -
2 Epsilon therefore for every
Epsilon uh for large
n we are going to get that this so now
this is going to be sandwiched by this
so when this goes
to something like this will converge let
us say it converge to some limit Alpha
then this Alpha is lesser equal
to b - Epsilon - 2 Epsilon and B minus
Epsilon + 2 Epsilon and this is true for
every very arbitrarily small Epsilon
therefore thus
limit n goes to
Infinity
true
for all
arbitrary
small Epsilon
and that's what exactly what we wanted
to prove and this shows that this will
imply
that JN is equ
distributed okay so now we would like
to prove the converse of while Criterion
so the converse is it says the converse
of
while what does it say
if
for all nonzero integers
K uh limit n goes to Infinity
1 by uh
if no
no so
if
Zen is equid
distributed
then limit n goes to Infinity 1 by n
kjn this is equal to zero for all
nonzero integer
K okay so now this
statement
uh Jen is e distributed means what this
we have seen that this implies that for
for
every sub
interval
a
01 what limit n goes to
Infinity 1 / n
in how many GI are there in the interval
AB this is the cardinality and then this
is equal to B minus of a that means this
is integral 0 to 1 indicator function of
ab this is given to us now what we will
prove that this
happens now
if if we can prove if we can replace the
indicator
function uh by a continuous function
then obviously in particular we will get
this this is going to be a continuous
function therefore so now we had been
given with let's say a continuous
function this is in fact the argument
goes for
uh every REM and integrable function uh
now uh if I have a continuous function
and if I want to look at the integral of
that then look at the reman prescription
how we are doing we are doing we are
taking the uh lower sum and the upper
sum and where it converges then that
that is what
the integral is so that
means if I take f
l to be for certain thing if I am
dividing it uh by this F L of of X this
is some M star and K of i
k f of x of x k let's say from 1 to n
and Le this is M Small n f where
MF
K or rather MK let's just put it that's
much more
simpler MK
is infimum or rather in this case
minimum because we are doing uh The
Continuous function f of x so that X
belongs to
ik now for this F now if I look at some
M now this F
L of X this is less or equal to f of
x now what one can do is that this uh
then one of similarly if I take the U
opposum
is K = 1 to n I divide AB into the N
equal parts with the 1 by n and uh then
this is
MK F uh indicator function
i
k now if I take the this integral then
this is integral of FL is a lower
sum this integral of f is upper
sum and now the FL L if I replace this
this is true for every sub interval AB
therefore if I replace it in the 01 in
interval uh 0
to
uh this
Criterion this is true for so therefore
limit n goes to Infinity 1 / n summ Over
N from 1 to
n uh f l of JN this is going to converge
into the reman sum this is going to n
goes to Infinity this is going to get me
the integral of this and the integral of
this is nothing but the REM integral of
this now f l of JN is lesser equal
to F of Z is less or equal to F U of
Zen uh hence the sum which is lesser if
we take out the limit here this is
lesser equal to 1 by n summation / n = 1
to n f of Z which is 1 by n summation
Over N to1 to n f u of Z
now this side converges to integral
of 0 to 1 ft
DT and this converges to 0 to 1 ft
DT so therefore this the middle one has
no other choice but to converse to 0 to
1 integral 0 to 1 ft DT
so therefore in particular
particular if we
choose our F to be e to the^ 2 pi i
k uh
T then by the previous observation we
have got summation over 1 by n n = 1 to
n e^ 2 pi i k
JN this is going to converse to 0 to 1
e^ 2 pi i k t
DT now K is a non-zero quantity so once
K is a nonzero quantity this we have
seen many times that this is equal to
zero so this is uh
one application of the foure series what
we have seen is the this is called the
while e distribution theorem that that's
what the Ws Criterion is known as and
which has many many
applications thank you
[Music]
[Music]
[Music]
