Estimates for some square functions of littlewood-paley type
Abstract
Rubio de Francia, José L.
Full text
Pub
.
Ma
.
UAB
Vol
.
27 Ns
2
Juny
1983
ESTIMATES
FOR
SOME
SQUAREFUNCTIONS
1
.
In oduc ion
One
o
he
classical esul s
o
he
Li lewood-
Paley heo ys a es
ha
he
Lp-no m
o a
unc ion
is
equi alen
o
he
Lp-no m
o
he
quad a ic
mean
o
i s
pa ial
sums
co esponding
o
he
dyadic
in e als,
which
o m
a
decomposi ion
o
1R
n
.
The
p ecise
s a e-
men
can
be
seen
in
101, Ch
.
IV
.
Mo e
ecen ly,o he
ypes
o
pa i ions
ha e
been
conside ed
and,
in
pa icu-
la ,
ha
ob ained om
a
ixed
in e al
and
i s
ans-
la es,
i
.e .,
in
heone-dimensional
case
:
A (x)
_
~
'
1
.
1
1
(u)e
27
ixIdj
1
2
1/2
OF
LITTLEWOOD-PALEY
TYPE
José
L
.
Rubio
de
F ancia
Fo
he
ope a o
A
,
one
does
no
ob ain
equi-
alence
o
no ms
(excep
o
L
2
),
bu
only
he
inequali y
I~
A
.1I
p
<_
Cp
11
II
p
which,
mo eo e ,
is
only
,
alid
in
he
ange
2
<_
p
<
-
.
A
ske ch
o
p oo
o
his
esul
appea s
in [1],
whe e
i is
shown
o be
a
basic
ing e-
dien
o
A
.
Có doba'sapp oach o
he
es ima es
o
sphe ical
summa ion
mul iplie s
.
A
mo ede ailed
p oo
is
gi en
in
[2]
.
This
pape
g ew
ou
o
con e sa ions
on
his sub-
jec
wi h
A
.
Có doba,
o
whom
I
am indeb ed
o
sha ing
his
knowledge
wi h
me
.
The
pu pose
is
wo- old
.
Fi s ,
we
gi e
in
sec ion
2
ano he
p oo
o
he
esul
jus
men ioned
and
some
sligh
gene alisa ions
.
This
new
a
simple
app oach
is
based
on
he
unexpec ed
poin wiseinequali y
:
G (x)
2
1
Cons
.
M(I 1
2
)(x)
(whe e
G
is a
smoo h
e sion
o
A),
om
which,
weigh ed
Lp
inequali ies
o
he
ope a o
A
a e
also
ob ained
almos
immedia ely
.
Secondly,
we explo e
some
o
he
analogueso
A
in
o de
o
ge
a
deepe
unde s anding
o
wha
is
eally
going
on wi h hesequad a icope a o s
.
In
pa icula ,
some
con inuous
analogues
o
A
conside ed
in
sec ion
4
lead
o
a
s iking
esul
on
poin wise
con e gen e
o
a e ages
o
ball
mul iplie s
.
Fu he
gene alisa ion
is
gained
in
sec ion
5,
whe e
we
deal
wi h
A
and i s
smoo h
e sion G
in
locallycompac
Abelian
g oups,
a
con ex which equi es
ye
ano he di e en
p oo
o
he
Lp-inequali ies
.
The
no a ion
used
is
ai lys anda d
.
We
deno e
by
M
=
M
1
he
Ha dy-Li lewoodmaximal
ope a o
in
and,
mo egene ally,we de ine
M
g
(x)
=
M(I l
q
)(x)
1/q
=
sup(IQI-1
J
I (Y)Igdy)1/q
XEQ
Q
IR
n
.
The
classes
o
weigh s
conside ed
a e
:
A
p
( he
usual
Muckenhoup 's
class)
and
Ap
,
whichconsis s
o
all
w(x)
>
0
such ha
sup
(III
-1
w)(III-1
1
w
l-p
)
p-1
2
.
Quad a i
c
Ope a o s
o
Disc
e e
Type
whe e
1
1
q
5
-
and
Q
s ands
o
an a bi a y
cube
in
whe e
I
is
an
a bi a ybounded
n-dimensional
in e al,
and
he
usual
modi ica ion
is
conside ed
o
p =
1
.
Weigh s
in
A
p
co espond
o
p oduc s
o
ope a o s
which
a e
bounded
wi h espec
o
wEA
P
,
such
as
he
s ong
maximal unc ion
o
he
double
Hilbe ans o m
(see
[5j)
.
Gi en
mCL~(IR
n
),
we
deno e
by
Tm
he
mul i-
plie ope a o
:
(T
m
)~
=
.m,
which
is well
de ined
in
L2 (IR
n
) .
When
T
m
can
be
ex ended
o a
boundedope a o
in
Lp,
we
shall
deno eagain
his
ex ension
by
T
m
.
In
pa icula ,
i
m
is
a
Schwa z unc ion,
mG
J(IR
n
),
we
know
ha
Tm
can
be de ined
in Lp
o
all
1
1
p
The
quad a ic
ope a o
o
be
conside ed
he e
is
(1)
G (x)
=
{y
IT
(x)I
2
}
1/
2
ke2
n
whe e
m(k+
.)
means
he
ansla ion
o
ou
mul iplie
:
m(k+
.)(E)
=
m(k+j)
.
Finally,
i
Q
s ands
o
he
uni
cube
in
IR
n
,
Q
=
[-1,
1 ]
n
,
we
de ine
he
unc ions
:
q
j
(x)
=
I2)QI
-1
X
2jQ
(x)
=
2
- i
n
q
o
(2
-
)x)
Theo emA
:
p
ecis
ely
:
j-0
(2)
G (x)
1
cj{qj*1 1
2
(x)}
112
<
C
M
2
(x)
ho1ds
o
e e y
EL
1
+L
.
,
whe e
he
cons an sc
j
depend
only
on
m
and
c
.
-
C
<
-
.
As
a
consequence
,
G
is
j-0
a
bounded
op
e a o
in
LP(IR
(3)
G (x)P
w
(x)
dx
_<
CP
(w)
1 (x)1P
w
(x)
dx
o
all
weA
p/
2
,
2
<__
p<
hen
he
poin wis
e
majo
isa ion
2<
p
an
d
mo e
P oo
:
Le
ge
A
IR
n
)
be
he
in e seFou ie
ans o m
o
m
.
Fo
each
ini e
sequence
a =
(a k
)
o uni
R
2
-
no m,
we
de ine
G
x
(x)
_
Z
ak
Tm(k+
.)
(x)
k
k
e
-21 ik
.y
g
(y)
(x-y)
dy
=
J
g
(y)
h
a
(y),
(x-y)
dy
whe e
h
;k
(y)
is
pe iodic
:
h
x
(y+k)
=
h
x
(Y)
(ks
ZZ
n
)
and
has
uni no m
in L
2
(Q)
.
Now,
we
de ine
:
co =
sup
{Ig(x)I
:
x
.Q}
c
j
=_
23n
sup
{Ig(x)I
:
xs(2jQ)<2j
-I
Q)}
k
so
ha
2c
j
<
m
(because
g
GA,
and
IGa
(x)I
< Z
cj
2-jn
J
.
Iha(y)
(x-y)I
dY
j=0
2 ]
Q
1
c
j
{2
- j
n
j=0
1/2
JJ
I (x-y)I
2
dy}
Q
(j=1,2,
.
.
.)
Since
G
(x)
=
sup
IG
a
(x)I,
he
i s inequali y
in
a
(2) is
p o ed,
and
he
second
one
ollows
because
qJ
.
*
1
M
o
e e y
unc ion
.
Since
M
2
is
a
bounded
ope a o
in
L
(IR
n
)
and
in
LP(w)
when
p
>
2
and
w
s
Ap
/2,
only
he
case
p =
2
o
(3)
emains
o be
p o ed
.
Bu
his
is a
conse-
quence
o
he
i s es ima e
in
(2)
oge he
wi h
he
obse a ion ha
J
(
1
Q
1
-I
XQ
)
*
(x)
w
(x)
dx
:5
C
I
(x)
1
w
(x)
dx
o
e e ycube
Q
and
e e y
weigh
w
s
A
I
.
The
p e ious
heo em
is
a
-
smoo h e sion
o
he
Li lewood-Paley
ype
inequali ies
we
ac uallywish
o
ob ain,
which
a e
conce ned
wi h
he
pa ial
sum
ope a o s
SI,
whe e
I
is
an
a bi a y
n-dimensional
in e al
and
(S
I
)
,
.
= XI
.
Now,
he e
is a
s anda d
unca ion
a gu-
men
o
ob ain
he
s ong
esul om
i s
smoo h e sion,
which
is
based
on
he
inequali y
(4)
¡SI
.
J
12
)1~2
II
p
s
Cp
~ j
~2)1~2
ii
p
J
i
which
holds
o
a bi a y
in e als
I
j
and
unc ions
s
Lp
(IR
n
)
(see
S ein
[lo])
.
The
in e als
{I
j }
a e
said
o be
almos cong u
-
en
(wi h
cons an
C ?
1)
i
sjp
R i
(I
j)
<__
C
ij
i
(I
J
)
(i
=
1,2,
.
.
.,
n)
whe e
i (
"
)
s ands
o
he
side
leng h
o an
in e al
along
he
x
i
-di ec ion
.
Theo em
B
:
I
2s p
<
w
and
m s Ap l2
,
hen,
o
e e y
sequence
{I
j
}
o
disjoin almos
cong uen
in e
-
als
in
IR
n,
he
inequali y
(5)
II
(1
1S
I
.
1
2
)
112
11
c
(w)
II
II
J
Lp
(W)
p
Lp
(w)
ho1ds o
all
eLP(w)
.
P oo
:
Suppose i s
ha
all
I
j
a e
bounded
.
Since
e e y hing
is
in a ian
unde
dila ions
in
each
coo dina e,
we
can
assume
ha
Qi
(I
j)
:S
C
(1
:i
i
5
n)
o
e e y
in e al
I j
in
ou
sequence
.
Then,
each
in e al
con ains
a
leas
one
la ice
poin
:
k
j
s
Ii
n
zZ
n
.
Take
a
Schwa z unc ion
m
such ha m(1)
=
1
when
s
I
= [-C,
C]',
so
ha
SIJ
=
S
IJ
(T
m(-kJ+
.)
)
(
s
Lp)
and
an
applica ion
o inequali y
(4)
oge he
wi h
Theo em
A
yields
:
II
(E
¡si
.
82
)
1/2
II
p
<
Cp
II
G
11
p
>>
cp
II
11
p
This
p o es
he
case
w(x)
__
1
.
Fo
a
gene al
w
s
AP
/2
he
a gumen
is
exac ly
he
same,
bu
we
mus
use he
weigh ed e siono
(4),
namely,
ha
suchinequali y
holds
no
only
in
LP(IR
n),
bu
also
in
LP(w)
i
w
s
AP
and
1
<
p
<
°°
.
(This
ac ually
a
a he
s aigh o wa d
consequence
o
a
gene al
heo em
o
Ma cinkiewicz
and
Zygmund
[6]
oge he
wi h
he
boundedness
in
LP(w)
o
he
mul ipleHilbe
ans o m
;
see
also
Ku z
[5])
.
Finally,
i
may
be
he
case
ha ,
o
some
i
=
1,2,
.
.
.,
n,
we
ha e
,
Zi(I
J)
_
o
all
Ij
.
The
necessa y
modi ica ions
o
deal
wi h
his case
a e
a he
i ial,
a e a e
le
o
he
eade
.
Gi en
a
sequence
o
disjoin
consecu i e
in e als
in
IR,
i
hey
all
ha e
he
lame
leng h
we
ha e
jus
p o ed
ha
inequali y
(5)
holds
ue, while
in
he
case
o
leng hs
inc easing
a an
exponen ial
a e,
he
same
inequali y
is
ob ained
(and
no
only
o
pi2,
bu
o
all
1
<
p
<
-) by
classical
Li lewood-Paley heo y
.
l
seems
na u al
o
.expec
he
same
kind
o
esul
when
he
leng hs
inc ease
a bi a ily
.
A
pa ialposi i e
answe
is
con ained
in
ou
nex
esul
which,
o
he
sake
o
simplici y,will
be
s a ed
in
i s
one-dimensional
e -
sion
.
Theo emC
:
Le
{aej}j
s
~Z
be
an
odd
sequence
o eal
num
-
be s
(i .e
.
a-3
=
-a
j
),
and
assume
ha
i s
posi i e
pa
is co
n ex
-
and
slow1y
inc easing,
i-e
.
j-1
j
2
,~-1 ,~+1
a2
~
j
:1
C
a
.
(j
1
1)
Then,
he
quad a icexp ession
a
.
{
~
1
?(C)
e
2uix~
d1
12
}
112
-~
a
.,1-1
de ines
a
bou
nded
ope
a o
in
LP(IR
),
2
1
p
<
~
.
GN
(x)
=
{
1J
e
-2T il-y
g
(
y
)
(x-y)
dy
12
dj}l/2
.
M<N
IR
n
When
sL
2
,
he
inne
in eg al
is
absolu elycon e gen ,
and,
i we
ix
xs
IR
n
and
N>
0,
he e
exis s
a
unc-
ion
a(J)
(depending
on
bo h
x
and
N)
o
uni
no m
in
L
2
(IR
n
)
such
ha
GN
(x)
11 ¡<N
J
a(I)
e-2T il-y
g(y)
I
n
=
jh(y)
g(y)
(x-y)
dy
whe e
h
s L
2
and
jjh11
2 =
1
.
Since
K
=
(gl
2
s L
1 ,
we
áppiy Scha z
inequali y
o
ob ain
G
N
(x)
i
{1
K(y)
j (x-Y)I
2
dy}
1/2
and
pa
(i)
is
p o ed
by
le ing
N
- -
.
Now,
unde
he
hypo hesis
o
(ii),
jK(x)j
is
domina ed
by
a
dec easing,
adial,
in eg able unc ion,
and
hus
:
(K
*
¡ `2)1/2
<
C
M(I l
2
)
l/
2
=C
M
2
In
pa icula ,
he
maximalope a o
:
(x-y)
dy d1
8-
U0
{
J
I
T
m(u+d
"
)
-
m
(u)
12
du}
1/2
IR
n
is
bounded
in
LPOR
n
),
2 <
p
<-_
and
o
weak ype
(2,2)
.
By
a
s anda d
echnique,
h
epoin wise
con e gence
esul
will
be
p o ed
o
e e y
s
Lp,
2
<
p
<
-,
i
we
es a-
blish
his
esul
o
e e y
Schwa z unc ion
.
Bu , i
eJ(IR
n
)
:
{
JI
T
M(u+d
.)
(x)
-
m(u)
)()¿)I
2
du}l/2
<
<-_
{J{JIM(u+d~)
-
m(u)I
I (&)I
dj}
2
du}
1/2
<
J
I (j)I{
JIM(U+61)
-
m(u)I
2
du}
1/2
d1
Since
m
s L
2
,
he
inne
in eg al
in
he
las
exp ession
is
bounded
independen ly
o
d,
1,
and
i
anisheswhen
d
->
0,
so
ha
an
applica ion
o
Lebe gue's
dominá ed
.
con e gence
heo em
ends
he
p oo
.
The
ollowingpa icula case
o
Theo emD
is
wo h
men ioning
:
Le
m
=
X
B,
whe e
B
is
he
uni
ball
in
IR
n
.
Then
II(x)1
=
Ixi-n/2
IJn/2
(2-nIx1)I
5
C(1+Ix1)-n/2-1/2
and
bo hpa s
o
he
heo em
can
be applied
.
I
we
w i e
S
E
o
he
pa ial
sum
ope a o
co esponding
o
he
mul iplie
XE,
hen
he
i s
pa
shows
ha i
makes
sense
o
de ine
he
ope a o
:
-->
{1
nISu+B
l2
du}
1/2
I
o
e e y
s
Lp(1R
n
),
2
<__
p
<__
-,
e en
hough,
by
Fe e -
man's
heo em
([3]),
each
one
o
he
ope a o s
Su+B
can
be
de ined
only
in
L
2
.
Mo eo e ,
he
second
pa
gi es
:
Co olla y
3
:
l
s
Lp(
.?n)
,
2 : p <
-,
hen
S
*
(x)
=
sup
{J
¡S (u+B)
(x)12
du}1/2
C
9
2
(x)
0< <w
IR n
and
Zim
J
15
(u+B)
(x)
-
(x)I
2
du =
0
a
.e
.
->M
(10)
lim
S
V
(x)
=
(x)
a
.e
.
Iu1<1
Gi en
an
open
ball
V
in
]R
n
con aining
he
o igin,
i
is
no
known
whe he
is
ue
o
e e y
s
L
2
.
Wha
Co olla y
3
shows
is
ha
a
ce aina e age
o
he
s a emen s
(10)
o all
balls
con-
aining
he
o igin
is
ue,
bu
his
is
only
a
poo
subs i u e
which
is
a
away om
(10)
i sel
.
5
:
The
Case
o
Locall
Compac
G
oups
An
essen ial
pa
o
he
esul s
in
sec ions
2
and
4
can
be
o mula ed
in
he
con ex
o
locally
compac Abelian
(
.c
.a
.)
g oups,
p o iding
some so
o
uni ied
e sion
o
he
disc e e
and
con inuous
cases
.
Since he e
is
no
Ha dy-Li lewood
ope a o
in
con ex ,
we
only
ob ain
he
Lp
inequali ies,
poin wise
es ima es,
and
a
di e en ,
sligh ly
app oach
mus be ollowed
.
Le
X
be
a .c
.a
.
g oup
wi h
dual
g oup
X
A
gene alelemen
o
X
( esp
.
X
)
will
be deno ed
by
x,
y,
e c
.
( esp
.,
a,
e c
.),
and
ds,
d1
will
be
he
Haa measu es
on
X
and
X
,
whichwe
shallassume
o
be
sui ably
no malised
wi h espec
o
each
o he
.
The
le e s
H
and
P
will
be
used
o
ep esen
closed
sub-
g oups
o
X
and
X
,
hei
Haa
measu es
being
m
H
and
m
,
espec i ely
.
The
ope a o
o be
conside ed
he e
is
his
gene al
bu no
he
longe
G
(x)
I(ag)
-
(x)I
2
dm,
(a)}
1/2
When
X
=X=
IR
n
,
we
can
ake
=
2z
n
,
ob aining
he
ope a o
de ined
in
(1),
o
=
X
=
IR
n, in
which
case
we
ge he
ope a o
G
o
(9)
(wi h
g
=
m in
bo h
cases)
.
Theo em
E
:
Le
g
s L
1 (
X)
be a
posi i e
unc ion
such
ha
(11)
{
Ij(C,)12
dm
(a)}1/2
-
M
<
Then
he,
ope a o
l
:
de d
ed
a
bo e
bound
ea
Lp
(
X)
when
2
< p
<
II
G ll
p
5Mil
11
p
P oo
:
We
de ine
he
closed
subg oup
o
X
1
H
=
=
{x
s
X
:
<
x,a>
=
1
o
all
as
}
so ha
=
(
X/H)
.
The
Haa measu e
m
in
X
/H
is
de ined
as
he
dual
o
m
.
Then,
a
uniqueHaa measu e
mH
can
be
ixed
in
H
so
ha ,
de ining
a>g(y)
(x-y)
dyl2
dm
(a)}
1/2
and
mo e
p ecisely
:
(
s
Lp
;
2 <
p
5
-)
(z)
_
1
(x+y)
din
H
(y)
(
s
L~
(
X)
)
H
(whe e
z
= (x +H) s
X/H),
he
ollowing
iden i y
always
holds
(12)
J
(x)
dx
=
1
T(z)
dm(x)
X
X/H
A
se ies
o
simplelemmas
will
be
needed
in
he
p oo
:
Lemma
1
:
I
g s
L
1(
X)
is
posi i e,
hen,
o
e e y
sX
J
19-(CL-I)1Z
dm,,
(a)
<_
1
9(a)I
Z
dm
(a)
=MZ
P oo
:
Since
(T)
^
(a)
=
(a)
o
e e y
a e
,
Plan-
che el's
heo em
implies
1
9
12
din,
(a)
J
ji(a)j2
dm,
(a) =
1
g(x)
2
dm(x)
X/H
Bu
I (x)1
1
j j
'
(
x)
o
e e y
,
and
in
ou
pa icu-
la
case,
we
ha e
1(1g)
'
(x)j
:S
g'(S¿),
whic
hconclude
s h
e
p oo
o
he
lemma
.
In
ou
nex
esul ,
we
shall
deno e
by
B
a
Banach
space
(wi h
dual
B*)
and
by
LB
(
X)
he
Bochne -
Lebesgue
space
consis ing
o
all
(s ongly)
measu able
B- alued
unc ions
F(x)
de ined
on
X
and
such ha
JI
F(x)
II
B
is
in eg able
.
Lemma
2
:
Le
K
(x) be
a
measu able
B*- alued
unc ion
such ha
(13)
1
I
K(x)-b1
dx
5
CII
bII
B
bEB
Then,
he
o mula
(14)
T
F(x)
=
1
K(x-y)-
F(y) dy
de ines
a bounded
ope a o om
L
1(
X)
o
L
1
(
X)
mi h
no m
5
C
.
P oo
:
I
su ices
o
e i y
ha
II
TF
II
1
1
C
II
F
II
L
1
o
all
simple
in eg able
unc ions
F,
since
hese
B
a e
dense
in LB
.
Fo
each
such unc ion
F, (13)
implies
ha
TF(x)
is
well
de ined,
and
JITF(X)j
dx
5
1
1
IK(x-y)
"
F(y)I
dy dx
=
=
J
{J
¡x (x)
-F(y)
j
dx}
dy
<_
c
111F(y)
11
Bdy
.
In
he
nex
lemma,
we ake
as
ou
Banach
space
B
=B
*
=
L
2
(
0
)
=L
2
(
0
;
m
),
whe e
o
is
a
ixed
compac subse
o
.
Func ions
F s
L
2
(X)
a e
hen
iso-
me ically
iden i iedwi h unc ions
(x,
a) =
F(x)
(a)
in
he
p oduc
space
L
2
(X
x
o
) .
Lemma
3
:
Le
U
:
L
2
, LB be
de ined
by U (x,
a)
(ag)
*
(x)
.
Then
U
is
a
bounded
ope a o
wi hno m
5
5
M,
and
i s
adjoin
T
= U
:
LB
, L 2 is
gi en
by
(14)
whe e
he
ke nel
is
:
X(x)
=
k(x,
a)
= g(-x)
<x,
a>
.
P oo
:
The
i s asse ion ollows omPlanche el's
heo em
and
Lemma
1 :
~I
U
(x)
11
B
dx
x
J
1
^
?(j)1
2
d1
dm
(a)
x
0
M
2
1
^
I?(J)j
2
d1
=
M2
J
I (x)I2
dx
The
ke nel
o
U
*
is
ob ained
by
a
simple
compu a ion
which
is
le
o
he
eade
( he
ac ha
we
ake
a
com-
pac
subse
0
c
makes
all he
in eg als
absolu ely
con e gen )
.
We
a e
now
in
a
posi ion
o
comple e
he
p o
.o
o
Theo em
E
.
Le
G
o
deno e
he
ope a o
de ined
as G
a e eplacing
by o
.
I i is
p o ed
ha
II
Go
II
p
=
MII
II
p
(2 < p
<
-)
o
an
a bi a y compac
subse
o
o
,
hen,
le ing
o
inc ease
o
,
we
ge he
desi ed
esul
o G
.
Bu
IGo
(X)1
=
=
IIU
(x)II
B,
so
ha
he
inequali ies
o be
ob ained
a e
(15)
IIU II
LP
~MIl ll
p
(2<p5~)
B
Fo
p
=
2,
his
was
p o ed
in
Lemma
3
.
Fo
p
=
-,
(15)
is
equi alen ,
by
duali y,
o
11T
FII,1
5
MI¡
FIIT1
"
LB
and
his
can be
p o ed
by
e i ying
ha
he
ke nel
K(x)
de ined
in
Lemma
3
sa is ies
(14)
wi h
cons an
C =
M
.
In
ac ,
gi en
b
=
b(n)
s
B
=
L 2
(
0
):
IK(-x)
.bl
=
I
g(x)
<x,
a>b(a)
dm
(a)I
0
=
Ih(x)I
g(x)
wi h
h s
L
2
(
X
/H)
and
II
h
II
2
=
II
b
II
B,
so
ha
we
can
use
he
iden i y
(12) o
ob ain
IK(x)-bl
dx
=
Ih(x)I
g(x)
dm(z)
<
X
X
/H
~~
h
11
2
11
8
11
2
b
11
BM
Finally,
he
case
2<
p
<
-
o
(15)
ollows
by
in e pola-
ion,
and he
p oo
is
comple ed
.
When
=
X
,
(11)
simplymeans
ha
g
e L
2
(X)
.
Mo eo e ,
in
his
case,
he
assump ion
g i
0
is
no
neces-
sa y,
and
we
ha e
a
s a emen
comple ely
analogous
o
Theo em
D(i)
.
On
he
o he
hand,
i
is
disc e e
and
g
has
compac
suppo , hen
(11) is
i ially e i ied
.
The
unca ion
a gumen
used
in
Theo emB
can
also
be
applied
o
he
smoo h
G- unc ionconside ed
he e
:
Gi en
m
s
L
w
(
X),
we
say
ha
i
is
an
Lp-mul iplie
i
he
ope a o
Tm
de ined
(and
bounded)
in
L
2
(
X)
by
(T
m
)
^_
1m
admi s
a
bounded
ex ension
o
Lp(
X)
.
Theo em
F
:
Le
me
Lw
(
X)
be
-
a
compac ly
suppo ed
Lp-
mul iplie ,
wi h
2<p
<
g
oup
o,
X
,
hen,
he
ope a o
ás ITM(a+
:)
(x)I
2
}
1/2
is
a
disc e e
sub-