Carleson operators on doubling metric measure spaces

4 Proof of Finitary Carleson

To prove Proposition 2.0.1, we already fixed in Chapter 2 measurable functions σ1,σ2,Q and Borel sets F,G. We have also defined S to be the smallest integer such that the ranges of σ1 and σ2 are contained in [S,S] and F and G are contained in the ball B(o,14DS).

The proof of the next lemma is done in Section 4.1, following the construction of dyadic cubes in [ .

Lemma 4.0.1 grid existence
#

There exists a grid structure (D,c,s).

The next lemma, which we prove in Section 4.2, should be compared with the construction in [ .

Lemma 4.0.2 tile structure

For a given grid structure (D,c,s), there exists a tile structure (P,I,Ω,Q,c,s).

Choose a grid structure (D,c,s) with Lemma 4.0.1 and a tile structure for this grid structure (P,I,Ω,Q,c,s) with Lemma 4.0.2. Applying Proposition 2.0.2, we obtain a Borel set G in X with 2μ(G)μ(G) such that for all Borel functions f:XC with |f|1F we have 2.0.22.

Lemma 4.0.3 tile sum operator

We have for all xGG

pPTpf(x)=s=σ1(x)σ2(x)Ks(x,y)f(y)e(Q(x)(y)Q(x)(x))dμ(y).
4.0.1

Proof

We use this to prove Proposition 2.0.1.

Proof of Proposition 2.0.1

4.1 Proof of Grid Existence Lemma

We begin with the construction of the centers of the dyadic cubes.

Lemma 4.1.1 counting balls
#

Let SkS. Consider YX such that for any yY, we have

yB(o,4DSDk),
4.1.1

furthermore, for any yY with yy, we have

B(y,Dk)B(y,Dk)=.
4.1.2

Then the cardinality of Y is bounded by

|Y|23a+200Sa3.
4.1.3

Proof

For each SkS, let Yk be a set of maximal cardinality in X such that Y=Yk satisfies the properties 4.1.1 and 4.1.2 and such that oYk. By the upper bound of Lemma 4.1.1, such a set exists.

For each SkS, choose an enumeration of the points in the finite set Yk and thus a total order < on Yk.

Lemma 4.1.2 cover big ball
#

For each SkS, the ball B(o,4DSDk) is contained in the union of the balls B(y,2Dk) with yYk.

Proof

Define the set

C:={(y,k):SkS,yYk}
4.1.8

We totally order the set C lexicographically by setting (y,k)<(y,k) if k<k or both k=k and y<y. In what follows, we define recursively in the sense of this order a function

(I1,I2,I3):CP(X)×P(X)×P(X).
4.1.9

Assume the sets Ij(y,k) have already been defined for j=1,2,3 if k<k and if k=k and y<y.

If k=S, define for j{1,2} the set Ij(y,k) to be B(y,jDS). If S<k, define for j{1,2} and yYk the set Ij(y,k) to be

{I3(y,k1):yYk1B(y,jDk)}.
4.1.10

Define for SkS and yYk

I3(y,k):=I1(y,k)[I2(y,k)[Xk{I3(y,k):yYk,y<y})]]
4.1.11

with

Xk:={I1(y,k):yYk}.
4.1.12

Lemma 4.1.3 basic grid structure

For each SkS and 1j3 the following holds.

If j2 and for some xX and y1,y2Yk we have

xIj(y1,k)Ij(y2,k),
4.1.13

then y1=y2.

If j1, then

B(o,4DS2Dk)yYkIj(y,k).
4.1.14

We have for each yYk,

B(y,12Dk)I3(y,k)B(y,4Dk).
4.1.15

Proof
Lemma 4.1.4 cover by cubes
#

Let SlkS and yYk. We have

I3(y,k)yYlI3(y,l).
4.1.16

Proof
Lemma 4.1.5 dyadic property
#

Let SlkS and yYk and yYl with I3(y,l)I3(y,k). Then

I3(y,l)I3(y,k).
4.1.17

Proof

For SkkS and yYk, yYk write (y,k|y,k) if I3(y,k)I3(y,k) and

infxXI3(y,k)ρ(y,x)<6Dk.
4.1.19

Lemma 4.1.6 transitive boundary
#

Assume Sk<k<kS and yYk, yYk, yYk. Assume there is xX such that

xI3(y,k)I3(y,k)I3(y,k).
4.1.20

If (y,k|y,k), the also (y,k|y,k) and (y,k|y,k)

Proof
Lemma 4.1.7 small boundary
#

Let K=24a+1. For each S+KkS and yYk we have

zYkK:(z,kK|y,k)μ(I3(z,kK))12μ(I3(y,k)).
4.1.24

Proof
Lemma 4.1.8 smaller boundary
#

Let K=24a+1 and let n0 be an integer. Then for each S+nKkS we have

yYknK:(y,knK|y,k)μ(I3(y,knK))2nμ(I3(y,k)).
4.1.31

Proof
Lemma 4.1.9 boundary measure
#

For each SkS and yYk and 0<t<1 with tDkDS we have

μ({xI3(y,k) : ρ(x,XI3(y,k))tDk})2tκμ(I3(y,k)).
4.1.35

Proof

Let D~ be the set of all I3(y,k) with k[S,S] and yYk. Define

s(I3(y,k)):=k
4.1.37

c(I3(y,k)):=y.
4.1.38

We define D to be the set of all ID~ such that II3(o,S).

Proof of Lemma 4.0.1

4.2 Proof of Tile Structure Lemma

Choose a grid structure (D,c,s) with Lemma 4.0.1 Let ID. Suppose that

ZQ(X)
4.2.1

is such that for any ϑ,θZ with ϑθ we have

BI(ϑ,0.3)BI(θ,0.3)Q(X)=.
4.2.2

Since Q(X) is finite, there exists a set Z satisfying both 4.2.1 and 4.2.2 of maximal cardinality among all such sets. We pick for each ID such a set Z(I).

Lemma 4.2.1 frequency ball cover
#

For each ID, we have

Q(X)zZ(I)BI(z,0.7).
4.2.3

Proof

We define

P={(I,z) : ID,zZ(I)},
I((I,z))=IandQ((I,z))=z.

We further set

s(p)=s(I(p)),c(p)=c(I(p)).

Then 2.0.18, 2.0.19 hold by definition.

It remains to construct the map Ω, and verify properties 2.0.13, 2.0.14 and 2.0.15. We first construct an auxiliary map Ω1. For each ID, we pick an enumeration of the finite set Z(I)

Z(I)={z1,,zM}.

We define Ω1:PP(Θ) as below. Set

Ω1((I,z1))=BI(z1,0.7)zZ(I){z1}BI(z,0.3)

and then define iteratively

Ω1((I,zk))=BI(zk,0.7)zZ(I){zk}BI(z,0.3)i=1k1Ω1((I,zi)).
4.2.4

Lemma 4.2.2 disjoint frequency cubes
#

For each ID, and p1,p2P(I), if

Ω1(p1)Ω1(p2),

then p1=p2.

Proof

For each ID, it holds that

zZ(I)BI(z,0.7)pP(I)Ω1(p).
4.2.5

For every pP, it holds that

Bp(Q(p),0.3)Ω1(p)Bp(Q(p),0.7).
4.2.6

Proof

Now we are ready to define the function Ω. We define for all pP(I0)

Ω(p)=Ω1(p).
4.2.7

For all other cubes ID, II0, there exists, by 2.0.8 and 2.0.9, JD with IJ and IJ. On such I we define Ω by recursion. We can pick an inclusion minimal JD among the finitely many cubes such that IJ and IJ. This J is unique: Suppose that J is another inclusion minimal cube with IJ and IJ. Without loss of generality, we have that s(J)s(J). By 2.0.8, it follows that JJ. Since J is minimal with respect to inclusion, it follows that J=J. Then we define

Ω(p)=zZ(J)Ω1(p)Ω((J,z))Bp(Q(p),0.2).
4.2.8

We now verify that (P,I,Ω,Q,c,s) forms a tile structure.

Proof of Lemma 4.0.2