• 検索結果がありません。

Multiplication Operators between Lipschitz-Type Spaces on a Tree

N/A
N/A
Protected

Academic year: 2022

シェア "Multiplication Operators between Lipschitz-Type Spaces on a Tree"

Copied!
37
0
0

読み込み中.... (全文を見る)

全文

(1)

Volume 2011, Article ID 472495,36pages doi:10.1155/2011/472495

Research Article

Multiplication Operators between Lipschitz-Type Spaces on a Tree

Robert F. Allen,

1

Flavia Colonna,

2

and Glenn R. Easley

3

1Department of Mathematics, University of Wisconsin-La Crosse, La Crosse, WI 54601, USA

2Department of Mathematical Sciences, George Mason University, Fairfax, VA 22030, USA

3System Planning Corporation, Arlington, VA 22209, USA

Correspondence should be addressed to Flavia Colonna,fcolonna@gmu.edu Received 4 December 2010; Accepted 16 March 2011

Academic Editor: Ingo Witt

Copyrightq2011 Robert F. Allen et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Let L be the space of complex-valued functionsf on the set of vertices T of an infinite tree rooted atosuch that the difference of the values off at neighboring vertices remains bounded throughout the tree, and letLwbe the set of functionsf∈ Lsuch that|fv−fv|O|v|−1, where|v|is the distance betweenoandvandvis the neighbor ofvclosest too. In this paper, we characterize the bounded and the compact multiplication operators betweenLandLwand provide operator norm and essential norm estimates. Furthermore, we characterize the bounded and compact multiplication operators betweenLw and the spaceL of bounded functions on T and determine their operator norm and their essential norm. We establish that there are no isometries among the multiplication operators between these spaces.

1. Introduction

LetXandYbe complex Banach spaces of functions defined on a setΩ. For a complex-valued functionψdefined onΩ, the multiplication operator with symbolψfromXtoYis defined as

Mψfψf, ∀f∈ X. 1.1

A fundamental objective in the study of the operators with symbol is to tie the properties of the operator to the function theoretic properties of the symbol.

(2)

WhenΩis taken to be the open unit disk in the complex plane, an important space of functions to study is the Bloch space, defined as the setBof analytic functionsf : → for which

βf sup

z∈

1− |z|2fz<∞. 1.2

The Bloch space can also be described as the set consisting of the Lipschitz functions between metric spaces from endowed with the Poincar´e distanceρto endowed with the Euclidean distance, a fact that was proved by the second author in1 see also2. In fact, f∈ Bif and only if there existβ >0 such that for allz, w

fzfwβρz, w, βf sup

z /w

fzfw

ρz, w . 1.3

More recently, considerable research has been carried out in the field of operator theory when the setΩis taken to be a discrete structure, such as a discrete group or a graph. In this paper, we consider the case whenΩis taken to be an infinite tree.

By a treeT we mean a locally finite, connected, and simplyconnected graph, which, as a set, we identify with the collection of its vertices. Two verticesuandvare called neighbors if there is an edge connecting them, and we use the notationuv. A vertex is called terminal if it has a unique neighbor. A path is a finite or infinite sequence of verticesv0, v1, . . .such thatvkvk 1andvk−1/vk 1, for allk.

Given a treeT rooted atoand a vertexuT, a vertexvis called a descendant ofuifu lies in the unique path fromotov. The vertexuis then called an ancestor ofv. Given a vertex v /o, we denote byv the unique neighbor which is an ancestor ofv. ForvT, the setSv

consisting ofvand all its descendants is called the sector determined byv.

Define the length of a finite pathuu0, u1, . . . , vun withukuk 1fork0, . . . , n to be the numbernof edges connectingutov. The distance,du, v, between verticesuandv is the length of the path connectingutov. The treeTis a metric space under the distanced.

Fixingoas the root of the tree, we define the length of a vertexvby|v|do, v. By a function on a tree we mean a complex-valued function on the set of its vertices.

In this paper, the tree will be assumed to be rooted at a vertexoand without terminal verticesand hence infinite.

Infinite trees are discrete structures which exhibit significant geometric and potential theoretic characteristics that are present in the Poincar´e disk . For instance, they have a boundary, which is defined as the set of equivalence classes of paths which differ by finitely many vertices. The union of the boundary with the tree yields a compact space. A useful resource for the potential theory on trees illustrating the commonalities with the disk is3.

In4it was shown that, if the tree has the property that all its vertices have the same number of neighbors, then there is a natural embedding of the tree in the unit disk such that the edges of the tree are arcs of geodesics in with the same hyperbolic length and the set of cluster points of the vertices is the entire unit circle.

In5, the last two authors defined the Lipschitz spaceLon a treeTas the set consisting of the functionsf :T which are Lipschitz with respect to the distancedonT and the Euclidean distance on. For this reason, the Lipschitz spaceLcan be viewed as a discrete

(3)

analogue of the Bloch spaceB. It was also shown that the Lipschitz functions onT are pre- cisely the functions for which

Df

sup

v∈TDfv<∞, 1.4

whereDfv |fv−fv|andTT\ {o}. Under the norm f

Lfo Df

, 1.5

Lis a Banach space containing the spaceLof the bounded functions onT. Furthermore, for fL,fL≤2f.

The little Lipschitz space is defined as L0

f ∈ L: lim

|v| → ∞Dfv 0

1.6

and was proven to be a separable closed subspace ofL. We state the following results that will be useful in the present paper.

Lemma 1.1see5, Lemma 3.4. aIff ∈ LandvT, then fvfo |v|Df

. 1.7

In particular, iffL1, then|fv| ≤ |v|for eachvT. bIff∈ L0, then

|v| → ∞lim fv

|v| 0. 1.8

Lemma 1.2see5, Proposition 2.4. Let{fn}be a sequence of functions inL0 converging to 0 pointwise inT such that{fnL}is bounded. Thenfn0 weakly inL0.

In6, we introduced the weighted Lipschitz space on a treeTas the setLwof the func- tionsf :T such that

sup

v∈T|v|Dfv<∞. 1.9

The interest in this space is due to its connection to the bounded multiplication operators onL. Specifically, it was shown in 5 that the bounded multiplication operators onLare precisely those operatorsMψ whose symbolψ is a bounded function in Lw. The spaceLw was shown to be a Banach space under the norm

f

wfo sup

v∈T|v|Dfv. 1.10

(4)

The little weighted Lipschitz space was defined as

Lw,0

f∈ Lw: lim

|v| → ∞|v|Dfv 0

1.11

and was shown to be a closed separable subspace ofLw.

In this paper, we will make repeated use of the following results proved in6.

Lemma 1.3see6, Propositions 2.1 and 2.6. aIff∈ Lw, andvT, then fv≤

1 log|v| f

w. 1.12

bIff∈ Lw,0, then

|v| → ∞lim fv

log|v|0. 1.13

Lemma 1.4see6, Proposition 2.7. Let{fn}be a sequence of functions inLw,0converging to 0 pointwise inT such that{fnw}is bounded. Thenfn0 weakly inLw,0.

In this paper, we consider the multiplication operators betweenLandLw, as well as betweenLw andL. The multiplication operators betweenLandL were studied by the last two authors in7.

1.1. Organization of the Paper

In Sections2and3, we study the multiplication operators betweenLwandL. We characterize the bounded and the compact operators and give estimates on their operator norm and their essential norm. We also prove that no isometric multiplication operators exist between the respective spaces.

InSection 4, we characterize the bounded operators and the compact operators from Lw toLand determine their operator norm and their essential norm. As was the case in Sections2and3, we show that no isometries exist amongst such operators. In addition, we characterize the multiplication operators that are bounded from below.

Finally, in Section 5, we characterize the bounded and the compact multiplication operators fromLtoLw. We also determine their operator norm and their essential norm.

As with all the other cases, we show that there are no isometries amongst such operators.

2. Multiplication Operators from L

w

to L

We begin the section with the study of the bounded multiplication operatorsMψ :Lw → L andMψ :Lw,0 → L0.

(5)

2.1. Boundedness and Operator Norm Estimates

Letψbe a function on the treeT. Define τψ sup

v∈TDψvlog1 |v|, σψ sup

v∈T

ψv

|v| 1.

2.1

In the following theorem, we give a boundedness criterion in terms of the quantitiesτψ and σψ.

Theorem 2.1. For a functionψonT, the following statements are equivalent:

aMψ :Lw → Lis bounded.

bMψ :Lw,0 → L0is bounded.

cτψ andσψare finite.

Furthermore, under these conditions, we have max

τψ, σψ

≤ Mψτψ σψ. 2.2

Proof. a⇒cAssumeMψ :Lw → Lis bounded. ApplyingMψ to the constant function 1, we haveψ ∈ L, so that, byLemma 1.1, we haveσψ <∞. Next, consider the functionf onT defined byfv log1 |v|. Thenfo 0; forv /o, a straightforward calculation shows that

|v|Dfv |v|

log1 |v|−log|v| ≤1 2.3

and lim|v| → ∞|v|Dfv 1. Thus,fw 1 and soMψfL ≤ Mψ. Therefore, forvT, noting that

D

ψf v Dψvfv ψ

v Dfv, 2.4

one has

DψvfvD

ψf v ψ

v Dfv

Mψf

L σψ|v|Dfv≤Mψ σψ. 2.5 Henceτψ <∞.

c⇒aAssumeτψandσψ are finite. Then, byLemma 1.3, forf∈ LwandvT, we have

D

ψf v≤Dψvfv ψ

v Dfv

Dψv

1 log|v| f

w |v|σψDfv

τψfw σψfwfo .

2.6

(6)

Since|ψo| ≤σψ, we obtain Mψf

Lψofo τψf

w σψf

wfo

τψ σψ f

w ψoσψ fo

τψ σψ f

w,

2.7

proving the boundedness ofMψ :Lw → Land the upper estimate.

b⇒cSupposeMψ :Lw,0 → L0is bounded. The finiteness ofσψfollows again from the fact thatψ Mψ1 ∈ L0and fromLemma 1.1. To prove thatτψ < ∞, let 0< α < 1 and, forvT, definefαv log1 |v|α. Thenfαo 0 and|v|Dfav → 0 as|v| → ∞; so fα ∈ Lw,0. Since for 0 < α < 1, the functionxxxαis increasing forx ≥1, the function Dfαvis increasing inα, andDfαv≤DfvforvT, wherefv log1 |v|, forvT. Thus,fαw≤ fw1. Therefore, forvT, we have

Dψvfαv≤D

ψfα v ψ

v Dfαv

Mψfα σψ|v|Dfαv≤Mψ σψ. 2.8 Lettingα → 1, we obtain

Dψvlog1 |v|≤Mψ σψ. 2.9

Henceτψ <∞.

c⇒bAssumeσψandτψare finite, and letf∈ Lw,0. Then, byLemma 1.3, forvT, we have

D

ψf v≤Dψvfv ψ

v Dfv

Dψvlog1 |v| fv log1 |v|

ψv

|v| |v|Dfv

τψ

fv

log1 |v| σψ|v|Dfv−→0

2.10

as|v| → ∞. Thus,ψf ∈ L0. The boundedness ofMψ and the estimateMψfLτψ σψ can be shown as in the proof ofc⇒a.

Finally we show that, under boundedness assumptions onMψ,Mψ ≥max{τψ, σψ}.

ForvT, letfv 1/|v| 1χv, whereχvdenotes the characteristic function of{v}. Then fvw1 and

ψfv

L ψv

|v| 1. 2.11

Furthermore, lettingfo 1/2χo, we see thatfow1 andψfoL |ψo|. Therefore, we deduce thatMψσψ.

(7)

Next, fixvTand forwT, define

gvw

⎧⎨

log1 |w| if|w|<|v|,

log1 |v| if|w| ≥ |v|. 2.12

Then,gv ∈ Lwand

|v| → ∞lim gv

w lim

|v| → ∞|v|

log1 |v|−log|v|

1. 2.13

Observe that, forwT, we have

D

ψgv w

⎧⎨

ψwlog1 |w|−ψwlog|w| if|w|<|v|,

Dψwlog1 |v| if|w| ≥ |v|. 2.14

Hence

sup

w∈TD

ψgv w≥ sup

|w|≥|v|Dψwlog1 |v|≥Dψvlog1 |v|. 2.15

Definefvgv/gvw. Thenfvw1 and MψMψfv

L D

ψgv gv

w

Dψvlog1 |v|

gv

w

. 2.16

Taking the limit as|v| → ∞, we obtainMψτψ. Therefore,Mψ ≥max{τψ, σψ}.

2.2. Isometries

In this section, we show there are no isometric multiplication operatorsMψ from the spaces LworLw,0to the spacesLorL0, respectively.

AssumeMψ :Lw → Lis an isometry. ThenψL Mψ1L 1. On the other hand,

|ψo| 1/2MψχoL 1/2χow 1. Thus supv∈TDψv ψL− |ψo| 0, which implies thatψis a constant of modulus 1. Yet, forvT, lettingfv 1/|v| 1χv, we see that

1fv

wMψfv

L 1

|v| 1, 2.17

which yields a contradiction. Therefore, we obtain the following result.

Theorem 2.2. There are no isometriesMψ fromLwtoLorLw,0toL0, respectively.

(8)

2.3. Compactness and Essential Norm Estimates

In this section, we characterize the compact multiplication operators. As with many classical spaces, the characterization of the compact operators is a “little-oh” condition corresponding the the “big-oh” condition for boundedness. We first collect some useful results about com- pact operators fromLworLw,0toL.

Lemma 2.3. A bounded multiplication operatorMψ fromLwtoLis compact if and only if for every bounded sequence{fn}inLwconverging to 0 pointwise, the sequence{ψfnL} → 0 asn → ∞.

Proof. AssumeMψ is compact, and let {fn}be a bounded sequence in Lw converging to 0 pointwise. Without loss of generality, we may assume fnw ≤ 1 for alln. Then the sequence{Mψfn}{ψfn}has a subsequence{ψfnk}which converges in theL-norm to some functionf∈ L. Clearlyψofnko → ψofo, and by partaofLemma 1.1, forvT, we have

ψvfnkv−fvψofnko−fo |v|D

ψfnkf

≤1 |v|ψfnkfL. 2.18

Thus,ψfnkfpointwise onT. Sincefn → 0 pointwise, it follows thatfmust be identically 0, which implies thatψfnkL → 0. With 0 being the only limit point of{ψfn}inL, it follows thatψfnL → 0 asn → ∞.

Conversely, assume every bounded sequence {fn}inLw converging to 0 pointwise has the property thatψfnL → 0 asn → ∞. Let{gn}be a sequence inLwwithgnw≤1 for alln. Then|gno| ≤ 1 for alln, and by partaofLemma 1.2, forvT, we obtain

gnv≤

1 log|v| gn

w≤1 log|v|. 2.19

Thus,{gn}is uniformly bounded on finite subsets ofT. So some subsequence{gnk}converges pointwise to some functiong. FixvTandε >0. Then forksufficiently large, we have

gvgnkv< ε

2|v|, gnk

vg

v < ε

2|v|. 2.20

We deduce

|v|Dgv≤ |v|gvgnkv gnk

vg

v |v|Dgnkv

≤ |v|gvgnkv |v|gnk

vg

v |v|Dgnkv

< ε |v|Dgnkv≤ε 1,

2.21

for allksufficiently large. Sog ∈ Lw. The sequence defined byfk gnkg is bounded in Lwand converges to 0 pointwise. Thus by hypothesis, we obtainψfkL → 0 ask → ∞. It follows thatMψgnk ψgnkψgin theL-norm, thus proving the compactness ofMψ.

(9)

By an analogous argument, we obtain the corresponding compactness criterion forMψ

fromLw,0toL0.

Lemma 2.4. A bounded multiplication operatorMψ fromLw,0 toL0 is compact if and only if for every bounded sequence{fn}in Lw,0 converging to 0 pointwise, the sequence {ψfnL} → 0 as n → ∞.

The following result is a variant ofLemma 1.3a, which will be needed to prove a char-

acterization of the compact multiplication operators from Lw to L and from Lw,0 to L0

Theorem 2.6.

Lemma 2.5. Forf∈ Lwand vT

fvfo 2 log1 |v|sw

f , 2.22

whereswf supw∈T|w|Dfw.

Proof. FixvTand argue by induction onn|v|. Forn0, inequality2.22is obvious. So assume|v|n >0 and|fu| ≤ |fo| 2 log1 |u|swffor all verticesusuch that|u|< n.

Then

fvfvf

v f v

≤ 1

|v|sw

f fo 2 log|v|sw f

fo 1

|v| 2 log|v|

sw

f .

2.23

Next, observe that 1/|v| 1≤log|v| 1/|v|, so 1

|v| ≤ 2

|v| 1 ≤2 log

|v| 1

|v|

. 2.24

Hence

1

|v| 2 log|v| ≤2 log|v| 1. 2.25

Inequality2.22now follows immediately from2.23and2.25.

Theorem 2.6. LetMψbe a bounded multiplication operator fromLwtoL(or equivalently fromLw,0

toL0). Then the following statements are equivalent:

aMψ :Lw → Lis compact.

bMψ :Lw,0 → L0is compact.

clim|v| → ∞|ψv|/|v| 1 0 and lim|v| → ∞Dψvlog|v|0.

(10)

Proof. We first provea⇒c. AssumeMψ : Lw → Lis compact. It suffices to show that, for any sequence{vn}inT such that 2≤ |vn| → ∞, we have limn→ ∞|ψvn|/|vn| 1 0 and limn→ ∞Dψvnlog|vn| 0. Let{vn} be such a sequence, and for eachn, define fn 1/|vn| 1χvn. Thenfno 0,fn → 0 pointwise asn → ∞, and fnw 1. By Lemma 2.3, it follows thatψfnL → 0 asn → ∞. Furthermore

ψfn

Lsup

v∈T

ψvfnv−ψvfnvψvnfnvn ψvn

|vn| 1. 2.26

Thus limn→ ∞|ψvn|/|vn| 1 0.

Next, for eachnandvT, define

gnv

⎧⎪

⎪⎪

⎪⎪

⎪⎩

0 if|v|<

|vn|, 2 log|v| −log|vn| if

|vn| ≤ |v|<|vn| −1, log|vn| if|v| ≥ |vn| −1.

2.27

ThenDgnv 0 if|v| ≤

|vn| or|v| > |vn| −1. In addition, if

|vn| < |v| ≤ |vn| −1, then

|v|Dgnv<4. Indeed, there are two possibilities. Either

|vn| ≤ |v| −1, in which case

|v|Dgnv 2|v|

log|v| −log|v| −1 ≤ 2|v|

|v| −1 ≤3, 2.28

or|v| −1<

|vn|<|v|, in which case

|v|Dgnv |v|

2 log|v| −log|vn|

|vn| 1

log

|vn| 12

|vn|

≤ 2

|vn| 1 |vn| ≤2

1 1

√2

<4.

2.29

Thus{gnw}is bounded, and{gn}converges to 0 pointwise. ByLemma 2.3, it follows that ψgnL → 0 asn → ∞. Moreover

ψgn

LψvngnvnψvngnvnDψvnlog|vn|. 2.30 Therefore limn→ ∞Dψvnlog|vn|0.

To prove the implication c⇒a, suppose lim|v| → ∞Dψvlog|v| 0 and lim|v| → ∞|ψv|/|v| 1 0. Clearly, if ψ is identically 0, then Mψ is compact. So assume Mψ :Lw → Lis bounded withψnot identically 0. ByLemma 2.3, it suffices to show that if

(11)

{fn}is bounded inLwconverging to 0 pointwise, thenψfnL → 0 asn → ∞. Let{fn}be such a sequence,ssupn∈fnw, and fixε >0. Note that

|v| → ∞lim Dψvlog1 |v| lim

|v| → ∞Dψvlog|v|log1 |v|

log|v| 0. 2.31

Thus there exists anMsuch that fno< ε

3sψ

L

, Dψvlog1 |v|< ε

6s, ψv

|v| 1 < ε

3s, 2.32

for|v| ≥M. UsingLemma 2.5, for|v|> M, we have

D

ψfn v≤Dψvfnv Dfn

v ψ v

Dψvfno 2 log|v| 1 fn

w fn

w

ψv

|v|

ψ

Lfno 2Dψvlog|v| 1 ψv

|v|

fn

w

< ε.

2.33

On the other hand, on the setBM{v∈T :|v| ≤M},{fn}converges to 0 uniformly, and thusDfndoes as well. Moreover

D

ψfn v≤Dψvfnv ψ

v Dfnv

ψ

Lfnv max

|w|≤MψwDfnv−→0, 2.34

uniformly on BM. Therefore Dψfn → 0 uniformly on T. Furthermore, the sequence {ψfno}converges to 0 asn → ∞. HenceψfnL → 0 asn → ∞, proving thatMψ is compact.

Finally, note that the functionsfnandgndefined in the proof ofa⇒care inLw,0. So the equivalence ofbandcis proved analogously.

Recall the essential norm of a bounded operatorSbetween Banach spacesXandYis defined as

Seinf

S−K:K is compact fromXtoY

. 2.35

(12)

Forψa function onT, define the quantities

A

ψ lim

n→ ∞sup

|v|≥n

ψv

|v| 1, B

ψ lim

n→ ∞sup

|v|≥nDψvlog|v|.

2.36

Theorem 2.7. LetMψ be a bounded multiplication operator fromLwtoL. Then Mψ

e≥max A

ψ , B

ψ . 2.37

Proof. For eachn, definefn 1/n 1χn, whereχndenotes the characteristic function of the set {v ∈ T : |v| n}. Thenfn ∈ Lw,0,fnw 1, andfn → 0 pointwise. Thus, by Lemma 1.4,{fn}converges to 0 weakly inLw,0. LetKbe the set of compact operators from Lw,0 toL0, and letK ∈ K. ThenK is completely continuous 8, and soKfnL → 0 as n → ∞. Thus

MψK ≥lim sup

n→ ∞

MψK fn

L≥lim sup

n→ ∞

Mψfn

L. 2.38

Now note that

Mψfn

Lsup

|v|n

ψv

n 1 . 2.39

Hence

Mψ

e≥infMψK:K∈ K

≥lim sup

n→ ∞

Mψfn

L

lim

n→ ∞sup

|v|≥n

ψv

|v| 1 A

ψ .

2.40

We will now show thatMψe. This estimate is clearly true if 0. So assume {vn}is a sequence inT such that 2≤ |vn| → ∞asn → ∞and

nlim→ ∞Dψvnlog|vn|B

ψ . 2.41

(13)

Forn andvT, define

hnv

⎧⎪

⎪⎩

log|v| 12

log|vn| if 0≤ |v|<|vn|, log|vn| if |v| ≥ |vn|.

2.42

Thenhno 0,hnvn hnvn log|vn|, and

|v|Dhnv

⎧⎪

⎪⎩

|v|

log|vn|log

|v| 1

|v|

log|v||v| 1 if 1≤ |v|<|vn|,

0 if|v| ≥ |vn|. 2.43

The supremum of|v|Dhnvis attained at the vertices of length|vn| −1 and is given by snsup

v∈T|v|Dhnv |vn| −1log |vn|

|vn| −1

log|vn| −1|vn|

log|vn| . 2.44

Since|vn| −1log|vn|/|vn| −1≤1, we have log 2 2

log|vn| ≤ hnwsn≤ log|vn| −1|vn|

log|vn| <2. 2.45

By lettinggn hn/hnw, we havegn ∈ Lw,0,gnw 1, andgn → 0 pointwise. By Lemma 1.4, the sequence{gn}converges to 0 weakly inLw,0. ThusKgnL → 0 asn → ∞.

Therefore,

MψK≥lim sup

n→ ∞

MψK gn

L≥lim sup

n→ ∞

ψgnL. 2.46

For eachn, we havegnvn gnvn log|vn|/sn. So D

ψgn vn 1

snDψvnlog|vn|. 2.47

Since limn→ ∞sn1, we have Mψ

e≥infMψK:K∈ K

≥lim sup

n→ ∞ sup

v∈TD

ψgn v

≥ lim

n→ ∞

1

snDψvnlog|vn| B

ψ .

2.48

Therefore,Mψe≥max{Aψ, Bψ}.

(14)

We now derive an upper estimate on the essential norm.

Theorem 2.8. LetMψ be a bounded multiplication operator fromLwtoL. Then

Mψ

eA

ψ B

ψ . 2.49

Proof. Forn, define the operatorKnonLwby

Knf v

⎧⎨

fv if |v| ≤n,

fvn if |v|> n, 2.50

wheref ∈ Lwand vn is the ancestor of vof length n. Forf ∈ Lw,Knfo fo, and Knf ∈ Lw,0. LetBn {v ∈ T : |v| ≤ n}, and note that Knf attains finitely many values, whose number does not exceed the cardinality ofBn. Let{gk}be a sequence inLwsuch that gkw≤1 for eachk. Thenasupk∈|gko| ≤1, and|Kngko| ≤a. Furthermore, by part aofLemma 1.3, for eachvTand for eachk, we have|Kngkv| ≤1 logn. Thus, some subsequence of{Kngk}k∈must converge to a functiong onT attaining constant values on the sectors determined by the vertices of lengthn. It follows that this subsequence converges toginLwas well, proving thatKnis a compact operator onLw. SinceMψ is bounded as an operator fromLwtoL, it follows thatMψKn:Lw → Lis compact for alln.

Define the operatorJn IKn, whereI denotes the identity operator onLw. Then Jnfo 0, and forvT, we have

|v|D

Jnf v |v|Jnf v−

Jnf v ≤ |v|Dfv≤f

w. 2.51

By partaofLemma 1.3, we see that

Jnf v≤

1 log|v| f

w. 2.52

(15)

Using2.51and2.52, we obtain MψMψKn f

LψJnfL sup

|v|>n

ψv

Jnf v−ψ

v Jnf v

≤sup

|v|>n

Jnf vDψv ψ v D

Jnf v

sup

|v|>n

Jnf vDψv ψv

|v| |v|D

Jnf v

≤sup

|v|≥n

1 log|v| Dψv ψv

|v| 1 f

w

≤sup

|v|≥n

log|v|Dψv1 log|v|

log|v|

ψv

|v| 1 fw

sup

|v|≥nlog|v|Dψv1 logn logn sup

|v|≥n

ψv

|v| 1 f

w.

2.53

Since

Mψ

e≤lim sup

n→ ∞

MψMψKnlim sup

n→ ∞ sup

fw1

MψMψKn f

L, 2.54

taking the limit asn → ∞, we obtain Mψ

eB

ψ A

ψ , 2.55

as desired.

3. Multiplication Operators from L to L

w

We begin this section with a boundedness criterion for the multiplication operators fromMψ : L → LwandMψ :L0 → Lw,0.

3.1. Boundedness and Operator Norm Estimates Letψbe a function on the treeT. Define the quantities

θψ sup

v∈T|v|2Dψv, ωψ sup

v∈T|v| 1ψv.

3.1

(16)

Theorem 3.1. For a functionψonT, the following statements are equivalent:

aMψ :L → Lwis bounded.

bMψ :L0 → Lw,0is bounded.

cθψ andωψare finite.

Furthermore, under the above conditions, one has

max θψ, ωψ

Mψθψ ωψ. 3.2

Proof. a⇒cAssumeMψ is bounded fromLtoLw. The functionfo 1/2χo ∈ Land foL1. Thus

ψoψfow≤ Mψ. 3.3

Next, fixvT. Thenχv ∈ LandχvL1; so

|v| 1ψvψχv

wMψ. 3.4

Taking the supremum over allvT, from3.3and3.4we see thatωψ is finite and

ωψMψ. 3.5

WithvT, we now define

fvw

⎧⎨

|w| if |w|<|v|,

|v| if |w| ≥ |v|. 3.6

Thenfv∈ L,fvo 0 andfvL1. By the boundedness ofMψ we obtain MψMψfv

w

≥ sup

1≤|w|≤|v||w|ψw|w| −ψ

w |w| −1

≥ sup

1≤|w|≤|v||w|2Dψw− sup

1≤|w|≤|v||w|ψ w .

3.7

(17)

Therefore,

|v|2Dψv≤ sup

1≤|w|≤|v||w|2DψwMψ ωψ. 3.8

Taking the supremum over allvT, we obtainθψ <∞. From this and3.5, we deduce the lower estimate

Mψ≥max θψ, ωψ

. 3.9

c⇒aAssumeθψandωψare finite. Then,ψ∈ Lw, and byLemma 1.1, forf ∈ Lwith fL 1 andvT, we have

|v|D

ψf v≤ |v|Dψvfv |v|ψ

v Dfv

≤ |v|Dψvfo |v|2DψvDf ωψDf

≤ |v|Dψvfo

θψ ωψ Df

.

3.10

Thus,ψf∈ Lw. Note that|fo| Df1 and ψ

wψo sup

v∈T|v|Dψv≤ωψ sup

v∈T|v|2Dψv ωψ θψ. 3.11 From this, we have

ψf

wψ

wfo

θψ ωψ Df

θψ ωψ, 3.12

proving the boundedness ofMψ :L → Lwand the upper estimate

Mψθψ ωψ. 3.13 b⇒cThe proof is the same as fora⇒c; since forvT, the functionsχvandfv

used there belong toL0.

c⇒bAssumeθψ andωψare finite, and letf∈ L0. Then, byLemma 1.1, forvT, we have

|v|D

ψf v≤ |v|Dψvfv |v|ψ

v Dfv

≤ |v|2Dψvfv

|v| |v|ψ

v Dfv

θψfv

|v| ωψDfv−→0

3.14

as |v| → ∞. Thus, ψf ∈ Lw,0. The proof of the boundedness ofMψ is similar to that in c⇒a.

(18)

3.2. Isometries

In this section, we show there are no isometric multiplication operatorsMψfrom the spaceL toLwor fromL0toLw,0.

SupposeMψ :L → Lwis an isometry. ThenψwMψ1w1. On the other hand, ψo 1

2ψχo

w 1 2χo

L1. 3.15

Thus supv∈T|v|Dψv ψw− |ψo|0, which implies thatψis a constant of modulus 1.

Now observe that, forvT, we have 1χv

LMψχ

w |v| 1ψv|v| 1, 3.16 which is a contradiction. Sinceχv ∈ L0for allvT, ifMψ :L0 → Lw,0is an isometry, then the above argument yields again a contradiction. Thus, we proved the following result.

Theorem 3.2. There are no isometriesMψ fromLtoLwor fromL0toLw,0.

3.3. Compactness and Essential Norm

We now characterize the compact multiplication operators, but first we give a useful com- pactness criterion for multiplication operators fromLtoLwor fromL0toLw,0.

Lemma 3.3. A bounded multiplication operatorMψ fromLtoLw(L0 toLw,0) is compact if and only if for every bounded sequence{fn}inL(L0) converging to 0 pointwise, the sequenceψfnw converges to 0 asn → ∞.

Proof. SupposeMψis compact fromLtoLwand{fn}is a bounded sequence inLconverging to 0 pointwise. Without loss of generality, we may assumefnL ≤1 for alln. SinceMψ is compact, the sequence{ψfn}has a subsequence{ψfnk}that converges in theLw-norm to some functionf∈ Lw.

ByLemma 1.3, forvTwe have ψvfnkv−fv≤

1 log|v| ψfnkf

w. 3.17

Thus, ψfnkf pointwise on T. Furthermore, since |ψofnko−fo| ≤ ψfnkfw, ψ0fnk0 → f0ask → ∞. Thusψfnkfpointwise onT. Since by assumption,fn → 0 pointwise, it follows thatf is identically 0, and thusψfnkw → 0. Since 0 is the only limit point inLwof the sequence{ψfn}, we deduce thatψfnw → 0 asn → ∞.

Conversely, suppose that every bounded sequence {fn} in L that converges to 0 pointwise has the property thatψfnw → 0 asn → ∞. Let{gn}be a sequence inLsuch that gnL ≤1 for alln. Then|gno| ≤ 1, and by partaofLemma 1.1, forvTwe have

|gnv| ≤ |v|. So{gn}is uniformly bounded on finite subsets ofT. Thus there is a subsequence {gnk}, which converges pointwise to some functiong.

(19)

Fixε >0 andvT. Then|gnkv−gv|< ε/2 as well as|gnkvgv|< ε/2 fork sufficiently large. Therefore, for allksufficiently large, we have

Dgvgvgnkv gnk

vg

v Dgnkv< ε Dgnkv. 3.18 Thusg ∈ L. The sequencefnk gnkg is bounded inLand converges to 0 pointwise. So ψfnkw → 0 ask → ∞. Thusψgnkψgin theLw-norm. Therefore,Mψis compact.

The proof for the case ofMψ :L0 → Lw,0is similar.

Theorem 3.4. LetMψ be a bounded multiplication operator fromLtoLw(or equivalently fromL0

toLw,0). Then the following are equivalent:

aMψ :L → Lwis compact.

bMψ :L0 → Lw,0is compact.

clim|v| → ∞|v|2Dψv 0 and lim|v| → ∞|v| 1|ψv|0.

Proof. a⇒c Suppose Mψ : L → Lw is compact. We need to show that if {vn} is a sequence in T such that 2 ≤ |vn|increasing unboundedly, then limn→ ∞|vn|2Dψvn 0 and limn→ ∞|vn| 1|ψvn| 0. Let {vn} be such a sequence, and for n define fn |vn| 1/|vnvn. Clearly fn → 0 pointwise, and fnL ≤ 3/2. UsingLemma 3.3, we see that

ψfn

w−→0 asn−→ ∞. 3.19

On the other hand, sincefno 0 for alln, we have ψfn

wsup

v∈T|v|D

ψfn v≥ |vn|

|vn| 1

|vn|

ψvn |vn| 1ψvn. 3.20

Hence limn→ ∞|vn| 1|ψvn|0.

Next, forn, define

gnv

⎧⎪

⎪⎪

⎪⎪

⎪⎨

⎪⎪

⎪⎪

⎪⎪

0 if|v|<

|vn| 2

, 2|v| − |vn| 2 if

|vn| 2

≤ |v|<|vn|,

|vn| if|v| ≥ |vn|,

3.21

wherexdenotes the largest integer less than or equal tox. Then gn → 0 pointwise, and gnL 2. Sincegnvn gnvn |vn|, we have

ψgn

w≥ |vn|ψvngnvnψ vn gn

vn |vn|2Dψvn. 3.22

ByLemma 3.3we obtain limn→ ∞|vn|2Dψvn≤limn→ ∞ψgnw0.

c⇒aSuppose lim|v| → ∞|v|2Dψv 0 and lim|v| → ∞|v| 1|ψv|0. Assumeψis not identically zero, otherwiseMψ is trivially compact. ByLemma 3.3, to prove that Mψ is

(20)

compact, it suffices to show that if{fn}is a bounded sequence inLconverging to 0 pointwise, thenψfnw → 0 asn → ∞. Let{fn}be such a bounded sequence,ssupn∈fnL, and fix ε >0. There existsM such that|v| 1|ψv|< ε/2sand|v|2Dψv< ε/2sfor|v| ≥M.

ForvTand byLemma 1.1, we have

|v|D

ψfn v≤ |v|ψvDfnv |v|Dψvfn v

≤ |v|ψvDfnv |v|Dψvfno |v|Dfn

≤|v| 1ψvDfnv |v|2Dψvfno Dfn

|v| 1ψvDfnv |v|2Dψvfn

L.

3.23

Sincefn → 0 uniformly on{v∈T :|v| ≤M}asn → ∞, so doesDfn. So, on the set{v∈T :

|v| ≤M},|v|Dψfnv → 0 asn → ∞. On the other hand, on{v∈T :|v| ≥M}, we have

|v|D

ψfn v≤|v| 1ψvDfnv |v|2Dψvfn

L< ε. 3.24 So|v|Dψfnv → 0 asn → ∞. Sincefn → 0 pointwise,ψofno → 0 asn → ∞. Thus ψfnw → 0 asn → ∞. The compactness ofMψ follows at once fromLemma 3.3.

The proof of the equivalence ofbandcis analogous.

Forψa function onT, define A

ψ lim

n→ ∞sup

|v|≥n|v|ψv, B

ψ lim

n→ ∞sup

|v|≥n|v|2Dψv.

3.25

Theorem 3.5. LetMψ be a bounded multiplication operator fromLtoLw. Then Mψ

e≥max

A ψ ,1

2B

ψ . 3.26

Proof. Fixk, and for eachn, consider the sets

En,k{v∈T :n≤ |v| ≤kn, |v|even},

On,k{v∈T :n≤ |v| ≤kn, |v|odd}. 3.27 Define the functionsfn,k χEn,k andgn,k χOn,k. Thenfn,k, gn,k ∈ L0,fn,kL gn,kL 1, andfn,kandgn,kapproach 0 pointwise asn → ∞. ByLemma 1.2, the sequences{fn,k}and {gn,k}approach 0 weakly inL0 asn → ∞. LetK0 be the set of compact operators fromL0

toLw,0, and note that every operator inK0is completely continuous. Thus, ifK ∈ K0, then Kfn,kw → 0 andKgn,kw → 0, asn → ∞.

(21)

Therefore, ifK∈ K0, then

MψK≥lim sup

n→ ∞

MψK fn,k

w

≥lim sup

n→ ∞

Mψfn,k

w

≥lim sup

n→ ∞ sup

v∈En,k

|v| 1ψv.

3.28

Similarly,

MψK≥lim sup

n→ ∞ sup

v∈On,k

|v| 1ψv. 3.29

Therefore, combining3.28and3.29, we obtain Mψ

einfMψK:K∈ K0

≥lim sup

n→ ∞ sup

kn≥|v|≥n|v| 1ψv

≥lim sup

n→ ∞ sup

kn≥|v|≥n|v|ψv.

3.30

Lettingk → ∞, we obtainMψe≥ Aψ.

Next, we wish to show thatMψe ≥1/2Bψ. The result is clearly true ifBψ 0.

So assume there exists a sequence{vn}inTsuch that 2<|vn| → ∞asn → ∞and

nlim→ ∞|vn|2Dψvn B

ψ . 3.31

Forn, define

hnv

⎧⎪

⎪⎪

⎪⎨

⎪⎪

⎪⎪

0 if vo,

|v| 12

|vn| if 1≤ |v|<|vn|,

|vn| if |v| ≥ |vn|.

3.32

Clearly,hno 0, hnvn hnvn |vn|, and

Dhnv

⎧⎪

⎪⎪

⎪⎪

⎪⎪

⎪⎪

⎪⎩ 4

|vn| if |v|1, 2|v| 1

|vn| if 1<|v|<|vn|, 0 if |v| ≥ |vn|.

3.33

参照

関連したドキュメント

proved that on any bounded symmetric domain (Hermitian symmetric space of non-compact type), for any compactly supported smooth functions f and g , the product of the Toeplitz

“rough” kernels. For further details, we refer the reader to [21]. Here we note one particular application.. Here we consider two important results: the multiplier theorems

From the delayed cosine and sine type matrix function on the fractal set R αn (0 &lt; α ≤ 1) corresponding to second order inhomogeneous delay differential equations with

A bounded linear operator T ∈ L(X ) on a Banach space X is said to satisfy Browder’s theorem if two important spectra, originating from Fredholm theory, the Browder spectrum and

Komal and Shally Gupta, Multiplication operators between Orlicz spaces, Integral Equations and Operator Theory, Vol.. Lorentz, Some new function

We note that in the case m = 1, the class K 1,n (D) properly contains the classical Kato class K n (D) introduced in [1] as the natural class of singular functions which replaces the

In the section we investigate the connection between DF-valued holomorphic mappings of uniformly bounded type on DF-spaces and the linear topological invariants (LB ∞ ) and (DN ).

Noor, “On analytic functions related to certain family of integral operators,” Journal of Inequalities in Pure and Applied Mathematics, vol.. Goel, “Functions starlike and convex