• Nie Znaleziono Wyników

Morphic words, Beatty sequences and integer images of the Fibonacci language

N/A
N/A
Protected

Academic year: 2021

Share "Morphic words, Beatty sequences and integer images of the Fibonacci language"

Copied!
12
0
0

Pełen tekst

(1)

Delft University of Technology

Morphic words, Beatty sequences and integer images of the Fibonacci language

Dekking, Michel

DOI

10.1016/j.tcs.2019.12.036

Publication date

2020

Document Version

Final published version

Published in

Theoretical Computer Science

Citation (APA)

Dekking, M. (2020). Morphic words, Beatty sequences and integer images of the Fibonacci language.

Theoretical Computer Science, 809, 407-417. https://doi.org/10.1016/j.tcs.2019.12.036

Important note

To cite this publication, please use the final published version (if applicable).

Please check the document version above.

Copyright

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons. Takedown policy

Please contact us and provide details if you believe this document breaches copyrights. We will remove access to the work immediately and investigate your claim.

This work is downloaded from Delft University of Technology.

(2)

Contents lists available atScienceDirect

Theoretical

Computer

Science

www.elsevier.com/locate/tcs

Morphic

words,

Beatty

sequences

and

integer

images

of

the

Fibonacci

language

Michel Dekking

DelftUniversityofTechnology,FacultyEEMCS,P.O.Box5031,2600GADelft,theNetherlands

a

r

t

i

c

l

e

i

n

f

o

a

b

s

t

r

a

c

t

Articlehistory:

Received29September2019

Receivedinrevisedform18December2019 Accepted29December2019

Availableonline7January2020 CommunicatedbyR.Giancarlo Keywords:

Morphicword HD0L-system

IteratedBeattysequence Frobeniusproblem Goldenmeanlanguage

Morphic words are letter-to-letter images of fixed points x of morphisms on finite alphabets. Thereare situationswhere theseletter-to-lettermapsdo not occur naturally, but have to be replaced by amorphism. We call this adecoration ofx. Theoretically, decorationsofmorphicwordsareagainmorphicwords,butinseveralproblemstheidea ofdecoratingthefixedpointofamorphismisuseful.Wepresenttwoofsuchproblems. The first considersthe so called A A sequences, where

α

isa quadraticirrational, A is the Beatty sequence defined by A(n)= 

α

n,and A A is the sequence (A(A(n))). The secondexample considershomomorphic embeddingsofthe Fibonaccilanguageintothe integers,whichturnsouttoleadtogeneralizedBeattysequenceswithtermsoftheform V(n)=p

α

n+qn+r,wherep,q andr areintegers.

©2020ElsevierB.V.Allrightsreserved.

1. Introduction

Thegoalofthispaperistoshowthat ifonesuspectsaninfinitewordona finitealphabettobea morphicword,i.e., theletter-to-letterimage ofafixedpoint ofamorphism,thenthewaytoachieve thisisnottotrytodothisdirectly,but indirectly.Bythelatterwe meanthatone replacesthe searchforafixed pointandaletter-to-lettermap bya searchfor afixedpointandamoregeneralobject:amorphism.Toemphasizethisprinciple,wecallthismorphismadecoration,and theinfinitewordwillthenbeadecoration ofafixedpoint.Itiswell-knownthattheclassofdecorationsoffixedpointsof morphisms isequaltothe classofmorphic words,see,e.g.,Corollary 7.7.5inthemonographby AlloucheandShallit[2]. Theirproof, although algorithmic, issomewhat indirect.We willbe using a‘natural’ algorithm togo fromthe decorated fixedpointtoamorphicword,given,e.g.,in[16].WedescribethisalgorithmintheproofofCorollary9.

Weillustratetheusefulnessofthis‘decorationprinciple’bygivingtwoexamples:iteratedBeattysequencesinSection3

andintegerimagesoftheFibonaccilanguage inSection 4.Inthat sectionwe solvethe Frobeniusproblemfor homomor-phicembeddingsofthe Fibonaccilanguage intheset ofintegers,which meansthat we givea precisedescription ofthe complementofthisembedding.

Althoughthe two examples are seemingly unrelated,they are connected by theappearance ofgeneralized Beatty se-quences,whichwedefineinSection2.

Inthe appendixwe give adifferentproof thatthe differencesequence oftheiteratedBeatty sequence A A definedby A A

(

n

)

= 

n



n

2



isa morphic word.Thisleads toa morphic wordonan alphabetofsize4.We conjecturethat thisis thesmallestsizepossible,whichisequivalenttotheconjecturethatthedifferencesequenceofA A isnotafixedpointofa morphism.

E-mailaddress:F.M.Dekking@math.tudelft.nl. https://doi.org/10.1016/j.tcs.2019.12.036

(3)

Forsomegeneralresultsforaspecialclassofdecorationsoffixedpointsofmorphismssee[15].In[15] thedecorations are socalledmarkedmorphisms,whichinsomesense aretheopposite ofthedecorationsthatone willencounterinthe presentpaper.Wementionalsothatdecorations ofmorphisms arecloselyconnectedtoHD0L-systems.See[19] forsome recentresultsontheseinthecontextofBeattysequences,whichinsomesensearealsooppositetoourresults.

2. GeneralizedBeattysequences

Let

α

beanirrationalnumberlargerthan1,then A definedby A

(

n

)

= 

n

α



forn

1 isknownastheBeattysequenceof

α

.Here,

·

denotesthefloorfunction.Following[3] wecallanysequence V oftheform V

(

n

)

=

p A

(

n

)

+

qn

+

r for n

1

where p

,

q

,

r areintegers,ageneralizedBeattysequence,forshortaGBS.

If S isasequence,wedenoteitssequenceoffirstorderdifferencesas



S,i.e.,



S isdefinedby



S

(

n

)

=

S

(

n

+

1

)

S

(

n

),

for n

=

1

,

2

. . .

HowdoesonerecognizeGBS’s?Ingeneralthisisnoteasy,butthereisausefulcharacterizationforquadraticirrational numbers

α

,whichhavethepropertythat

α

∈ (

0

,

1

)

andtheiralgebraicconjugate

α

∈ (

/

0

,

1

)

.TheseareknownastheSturm numbers.Ingeneral,thesequencesoffirstdifferences

:=



(

n

+

1

)

α

 − 

n

α





=





A

(

n

)



are calledSturmiansequences. The characterization is derived from the followingkey result, which isalso proved in the monographs[2] and[17].

Proposition1.([10],[1]) Let

α

beaSturmnumber.Thenthereexistsamorphism

σ

α onthealphabet

{

0

,

1

}

,suchthat

σ

α

(

)

=

cα. Inthefollowingwe willconsiderthevariantsof

σ

α onvariousotheralphabetsthan

{

0

,

1

}

,butwillnotindicatethisin thenotation.Asnotedin[3],thefollowinglemmafollowsdirectlyfromProposition1byrealisingthat

V

=

p A

+

q Id

+

r

V

(

n

+

1

)

V

(

n

)

=

p

(

A

(

n

+

1

)

A

(

n

))

+

q

=

p cα

(

n

)

+

q

.

Lemma2. (AlloucheandDekking[3]) Let

α

beaSturmnumber.LetV

= (

V

(

n

))

n≥1bethegeneralizedBeattysequencedefined

byV

(

n

)

=

p

(



n

α

)

+

qn

+

r,andlet



V bethesequenceofitsfirstdifferences.Then



V isthefixedpointof

σ

α onthealphabet

{

q

,

p

+

q

}

.

3. IteratedBeattysequences

RecallthataBeattysequenceisasequenceA

= (

A

(

n

))

n≥1,withA

(

n

)

= 

n

α



forn

1,where

α

isapositiverealnumber.

WhatBeattyobservedisthatwhenB

= (

B

(

n

))

n≥1isthesequencedefinedby B

(

n

)

= 

n

β



,with

α

and

β

satisfying

1

α

+

1

β

=

1

,

(1)

then A andB arecomplementary sequences,thatis,thesets

{

A

(

n

)

:

n

1

}

and

{

B

(

n

)

:

n

1

}

aredisjointandtheirunionis thesetofpositiveintegers.Inparticularif

α

=

ϕ

=

1+2√5 isthegoldenmean,thisgivesthatthesequences

(

n

ϕ

)

n≥1 and

(



n

ϕ

2

)

n≥1arecomplementary.

In thispaperwe lookatsequences asfunctionsfrom

N

to

N

.Inthiswaycompositions Z

=

X Y oftwosequences X andY aredefinedasthesequencegivenby Z

(

n

)

=

X

(

Y

(

n

))

forn

∈ N

.

AwellknownresultonthecompositionofBeattysequencesinthegoldenmeancaseisthefollowing.

Theorem 3. (Carlitz-Scoville-Hoggatt [7]) Let U

= (

U

(

n

))

n≥1 be a composition of the sequences A

= (

n

ϕ

)

n≥1 and B

=

(



n

ϕ

2

)

n≥1,containingi occurrencesofA andj occurrencesofB,thenforalln

1

U

(

n

)

=

Fi+2 jA

(

n

)

+

Fi+2 j−1n

− λ

U

,

whereFkaretheFibonaccinumbers(F0

=

0,F1

=

1,Fn+2

=

Fn+1

+

Fn)and

λ

Uisaconstant.

Thismeansthatanycompositionof A and B canbewrittenasanintegerlinearcombinationp A

+

qId

+

r, whereId is definedbyId

(

n

)

=

n.Thelength2compositionsinTheorem3give

(4)

A A

=

B

1

=

A

+

Id

1

,

A B

=

A

+

B

=

2 A

+

Id

,

B A

=

A

+

B

1

=

2 A

+

Id

1

,

B B

=

A

+

2B

=

3 A

+

2Id

.

AresultasTheorem3doesnotholdforallquadraticirrationals.Ifwetake,forexample,

α

=

2,i.e.,we considerthe Beattysequencegivenby A

(

n

)

= 

n

2



,thenthecomplementaryBeatty sequenceB isgivenby B

(

n

)

= 

n

(

2

+

2

)



.Itis provedin[8] (seealso[14])thatforn

1

A B

(

n

)

= 

2



n

(

2

+

2

)

 =

A

(

n

)

+

B

(

n

)

=

2 A

(

n

)

+

2n

.

However,noexpressionforA A isgiven.1 Infact,onecaneasilyprovethattheredonotexistintegers p

,

q andr suchthat A A

=

p A

+

qId

+

r.ThisfollowsfromLemma2inSection2,sincethefirstorderdifferencesequenceof A A takesmorethan 2values.Still,expressionsforA A areknowninvolvingthesequence



2

{

n

2

}

,seeTheorem1in[13],andsee[5].

Whydoes thegolden mean always yieldGBS’s forthe difference sequences ofthe compositionsof A and B,butthe silvermeandoesnot?OurTheorem5clarifiesthesituation.

From nowon we focuson theiterated Beatty sequence A A givenby A A

(

n

)

= 

n

α



α



.It hasbeenstudied by many authors. See, among others, [7], [8], [13], [4], [5]. The main effort in these papers has been to express A A as a linear combinationof A,Id andtheconstantfunction.

HereisanimportantbasicresultontheiteratesoftheBeattysequence A

(

n

)

= 

n

α



foralgebraic

α

ofdegreed.Inthe following,

{

n

α

}

=

n

α

− 

n

α



isthefractionalpartofn

α

.

Theorem4. (Fraenkel[13]) Letd

1 anda0

,

...,

ad

,

n

,

K

,

L

,

M

∈ Z

.Supposethatadxd

+

ad−1xd−1

+ · · · +

a1x

+

a0

=

0 hasareal

nonzeroroot

α

.LetA

(

n

)

= 

n

α



.Then A



M

+

Ln

+



id=02Ai

(

K a

i+2A

(

n

))



= (

L

K a1

)

A

(

n

)

K a0n

+

D,whereD isboundedinn,namely,

D

= 

M

α

+ (

L

+

K a0

α

−1

)

{

n

α

}

− θ

α



,where

θ

=



di=−12

(

K ai+2A

(

n

)

α

i

Ai

(

K ai+2A

(

n

)))

.

Weareinterestedonlyinthecased

=

2.Let

(

x

α

)(

x

α

)

betheminimalpolynomialofaquadraticirrational

α

.

Theorem5.Let

α

>

1 beaquadraticirrationalwithminimalpolynomialin

Z

[

x

]

.LetA

(

n

)

= 

n

α



.ThesequenceA A isageneralized Beattysequenceifandonlyif

|

α

|

<

1.

Proof. Ifonesubstitutes K

=

1,L

=

M

=

0,d

=

2 anda2

=

1 inTheorem4,oneobtains

A A

(

n

)

= −

a1A

(

n

)

a0n

+

D

(

n

),

where

(

x

α

)(

x

α

)

=

x2

+

a 1x

+

a0,and D

(

n

)

=



a 0

α

{

n

α

}



.

Thetheoremnowfollows,since

αα

=

a0,andsincethesequence

(

{

n

α

})

isequidistributedover

[

0

,

1

]

.

Hereweusedthat

θ

=

0 ford

2,asindicatedbyFraenkelintheNotesonpage642of[13].

2

Example6.Let

α

=

1

+

2, with corresponding A

(

n

)

= 

n

(

1

+

2

)



. As in the proof of Theorem 5 one computes that A

(

A

(

n

))

=

2 A

(

n

)

+

n

1.Thenumber

α

2

=

2

1 is aSturmnumber.The correspondingSturmiansequence isfixed point ofthe morphism

σ

α−2 given by 0

01

,

1

010,asfollowsfroma computation by continuedfractions ([10],[2]).

Since A

(

n

)

= 

n

(

α

2

)



+

2n,



A isfixed pointofthemorphism givenby2

23

,

1

232.Since A A

(

n

+

1

)

A A

(

n

)

=

2

(

A

(

n

+

1

)

A

(

n

))

+

1,



A A isfixedpointofthemorphismgivenby5

57

,

7

575.Sointhisparticularcase



A A is puremorphic.

2

Whatisthestructureof A A if

α

>

1 and

|

α

|

>

1?

WedeterminethisfortheFraenkelfamily,alsoknownasthemetallicmeans,whicharethepositivesolutionstox2

+ (

t

2

)

x

=

t,wherethenaturalnumbert istheparameter.Fort

=

1 oneobtainsthegoldenmean,fort

=

2 thesilvermean

2.

Theorem7.Let

α

=



2

t

+

t2

+

4



/

2,fort

=

2

,

3

,

. . .

,andletA

(

n

)

= 

n

α



forn

1.Then



A A isamorphicword.Infact,



A A

isadecoration

δ

ofafixedpointofamorphism

τ

,bothdefinedonthealphabet

{

1

,

2

. . . ,

t

+

1

}

.Fort

=

2 andt

=

3 themorphisms

τ

and

δ

aregivenrespectivelyby

τ

(

1

)

=

12

,

τ

(

2

)

=

131

,

τ

(

3

)

=

121

, δ(

1

)

=

13

, δ(

2

)

=

222

, δ(

3

)

=

132

,

τ

(

1

)

=

123

,

τ

(

2

)

=

124

,

τ

(

3

)

=

1141

,

τ

(

4

)

=

1241

, δ(

1

)

=

113

, δ(

2

)

=

122

, δ(

3

)

=

2122

, δ(

4

)

=

1222

.

(5)

Fort

4 themorphism

τ

isgiven2by

τ

(

1

)

=

1

...

[

t

1

]

t

,

τ

(

2

)

=

1

...

[

t

1

]

[

t

+

1

]

,andforj

=

3

,

...,

t

1

τ

(

j

)

=

1

...

[

t

j

] [

t

j

+

1

] [

t

j

+

1

] [

t

j

+

2

] . . . [

t

2

] [

t

+

1

],

τ

(

t

)

=

112

. . .

[

t

2

] [

t

+

1

]

1

,

τ

(

t

+

1

)

=

1223

. . .

[

t

2

] [

t

+

1

]

1

.

For t

4 the morphism

δ

is given by

δ(

1

)

=

1t−13

,

δ(

2

)

=

1t−222

,

δ(

j

)

=

1tj21j−22 for j

=

3

,

...,

t

1, and

δ(

t

)

=

21t−222

,

δ(

t

+

1

)

=

121t−322.

IntheproofofthistheoremweneedthecombinatorialLemma8.Weknowthat



A isfixedpointofthemorphism

σ

onthealphabet

{

1

,

2

}

givenby

σ

(

1

)

=

1t−12

,

σ

(

2

)

=

1t−121

,

(2)

ascanbefoundinCrispetal. [10],orAlloucheandShallit[2].Hereoneusesthat

α

hasaverysimplecontinuedfraction expansion:

α

= [

1

;

t

,

t

,

t

,

. . .

]

.

Lemma8.Lett

2 beaninteger.Fort

=

2,definethethreewordsu1

=

121

,

v

=

2112,andw

=

1212.

Fort

3,definethet

1 wordsuj

=

1tj21jforj

=

1

,

...,

t

1,andthetwowordsv

=

21t2

,

w

=

121t−12.

Let

σ

bethemorphismin (2),thenfort

=

2,onehas

σ

(

u1

)

=

u1v

,

σ

(

v

)

=

u1wu1

,

σ

(

w

)

=

u1vu1.

Fort

=

3 onehas

σ

(

u1

)

=

u1u2v

,

σ

(

u2

)

=

u1u2w

,

σ

(

v

)

=

u1u1wu1

,

σ

(

w

)

=

u1u2wu1.

Fort

4 onehas

σ

(

u1

)

=

u1

...

ut−1v

,

σ

(

u2

)

=

u1

...

ut−1w,andforj

=

3

,

...,

t

1 onehas

σ

(

uj

)

=

u1

...

utjutj+1utj+1utj+2

. . .

ut−2w

,

σ

(

v

)

=

u1u1u2

. . .

ut−2wu1

,

σ

(

w

)

=

u1u2u2u3

. . .

ut−2wu1

.

Proof. First we take t

=

2.Then

σ

is given by

σ

(

1

)

=

12

,

σ

(

2

)

=

121. One easily verifies the statement ofthe lemma:

σ

(

u1

)

=

1212112

=

u1v

,

σ

(

v

)

=

1211212121

=

u1wu1

,

σ

(

w

)

=

1212112121

=

u1vu1.

Thecaset

=

3 followsfromananalogouscomputation.

Next,thecaset

4.Wefirstmentionfourrelations,directlyimpliedbythedefinitions,whichwillbeusedintheproof: v

=

21

σ

(

1

),

w

=

12

σ

(

1

),

w

=

ut−12

,

u1

=

σ

(

2

).

Wealsouserepeatedly

σ

(

1j

)

=

u1

. . .

uj−11tj2 for j

=

2

, . . .

t

1

,

whichcanbeprovedbyinduction:

σ

(

1j+1

)

=

u

1

. . .

uj−11tj2

σ

(

1

)

=

u1

. . .

uj−11tj21t−12

=

u1

. . .

uj1tj−12.

Wethenhave

σ

(

u1

)

=

σ

(

1t−121

)

=

u1

. . .

ut−212

σ

(

2

)

σ

(

1

)

=

u1

. . .

ut−2121t−121

σ

(

1

)

=

u1

. . .

ut−121

σ

(

1

)

=

u1

...

ut−1v

,

σ

(

u2

)

=

σ

(

1t−2211

)

=

u1

. . .

ut−31121t−1211t−212

σ

(

1

)

=

u1

. . .

ut−112

σ

(

1

)

=

u1

...

ut−1w

.

Now foruj,with3

j

t

1:(interpretingu1

. . .

u0 asanemptyprefixinthecase j

=

t

1;sointhatcasetheoutcome

is

σ

(

ut−1

)

=

u1u2u2u3

. . .

ut−2w (ift

4).)

σ

(

uj

)

=

σ

(

1tj

)

σ

(

21j

)

=

u1

. . .

utj−11j2 1t−121

σ

(

1j

)

=

u1

. . .

utj−1utj1j−121 1t−12

σ

(

1j−1

)

=

u1

. . .

utj−1utjutj+11j−12

σ

(

1j−1

)

=

u1

. . .

utj−1utjutj+11j−12 1t−12

σ

(

1j−2

)

=

u1

. . .

utj−1utjutj+1utj+11j−22

σ

(

1j−2

)

=

u1

. . .

utj−1utjutj+1utj+11j−22 1t−12

σ

(

1j−3

)

=

u1

. . .

utj−1utjutj+1utj+1utj+21j−32

σ

(

1j−3

)

= · · ·

(6)

=

u1

. . .

utj−1utjutj+1utj+1

. . .

ut−212

σ

(

1

)

=

u1

. . .

utjutj+1utj+1utj+2

. . .

ut−2w

.

Forv and w onederives:

σ

(

v

)

=

σ

(

2

)

σ

(

1t−1

)

σ

(

12

)

=

u1u1

. . .

ut−212

σ

(

1

)

σ

(

2

)

=

u1u1u2

. . .

ut−2w u1

σ

(

w

)

=

σ

(

ut−12

)

=

u1u2u2u3

. . .

ut−2w

σ

(

2

)

=

u1u2u2u3

. . .

ut−2w u1

.

2

ProofofTheorem7. Inview of thecomplexity of theproof we first give theproof forthe caset

=

3,i.e., the case

α

=

(

13

1

)/

2,thebronzemean.

We then have toshow that



A A is a decoration

δ

ofa fixed point ofa morphism

τ

, both defined onthe alphabet

{

1

,

2

,

3

,

4

}

,where

τ

isgivenby

τ

(

1

)

=

123

,

τ

(

2

)

=

124

,

τ

(

3

)

=

1141

,

τ

(

4

)

=

1241

,

andthedecoration

δ

isgivenby

δ(

1

)

=

113

,

δ(

2

)

=

122

,

δ(

3

)

=

2122

,

δ(

4

)

=

1222

.

ThewordsfromLemma8areinthiscase

u1

=

1121

,

u2

=

1211

,

v

=

21112

,

w

=

12112

,

andtheirimagesunder

σ

are

σ

(

u1

)

=

u1u2v

,

σ

(

u2

)

=

u1u2w

,

σ

(

v

)

=

u1u1wu1

σ

(

w

)

=

u1u2wu1

.

Thecodingu1

1

,

u2

2

,

v

3

,

w

4 transforms

σ

workingon

{

u1

,

u2

,

v

,

w

}

into

τ

.

LetL bethemapthatassignstoanyworditslength,so,e.g.,L

(

u1

)

=

4

,

L

(

v

)

=

5.

CLAIM:1)Theword



A canbewrittenas



A

=

x1x2

. . .

whereeachxi isanelementfrom

{

u1

,

u2

,

v

,

w

}

.

2)Thewordr

:=

L

(

x1

)

L

(

x2

)

. . .

isfixedpointofthemorphism

σ

4,5 givenby4

445

,

5

4454.

Proof of part1) of the claim: we know that



A is the unique fixed point of the morphism

σ

=

σ

1,2 givenby 1

112

,

2

1121.Since1121

=

u1 isaprefixof



A,theword

σ

n

(

u1

)

isalsoaprefix of



A foralln

1.SowithLemma8

thisprovestheCLAIM,part1). Part2)ofthe claimthenfollowsfrom L

(

u1

)

=

L

(

u2

)

=

4

,

L

(

v

)

=

L

(

w

)

=

5,whichinduces

themorphism

σ

4,5fortheinfinitewordr oflengths.

Howdoweobtain



A A from



A?SinceA

(N)

=

A A

(N)

A B

(N)

,adisjointunion,oneobtains A A fromA byremoving theintegers A B

(

n

)

,which,ofcourse,haveindexB

(

n

)

inthesequence A.The differencesequence



B ofthissequenceis the unique fixed point of the morphism

σ

4,5,since

β

=

α

+

3. It follows then from the CLAIMthat the integers A B

(

n

)

occuratpositionswhichcorrespondtothethird letterinthewordxi.Hereit isthethird letter, becausethefirsttermof

thesequence

(

A

(

B

(

n

))

=

5

,

10

,

15

,

22

,

. . .

occursatposition4inthesequence

(

A

(

n

))

=

1

,

2

,

3

,

5

,

. . .

.Removalofthe A B

(

n

)

is then performed by adding the third and the fourth letterin the xi. This operation turns u1

=

1121 into

δ(

1

)

=

113,

u2

=

1211 into

δ(

2

)

=

122, v

=

21112 into

δ(

3

)

=

2122, and w

=

12112 into

δ(

4

)

=

1222. The conclusion is that this

decoration

δ

turnstheuniquefixedpointof

τ

into



A A.Thisendstheproofforthecaset

=

3.

Forgeneralt,thecodingu1

1

,

...,

ut−1

t

1

,

v

t

,

w

t

+

1 transforms

σ

workingon

{

u1

,

...,

ut−1

,

v

,

w

}

into

τ

.

Ananalogousclaimasforthet

=

3 caseholds,andnowthemap L satisfies L

(

u1

)

=

L

(

u2

)

= ... =

L

(

ut−1

)

=

t

+

1

,

L

(

v

)

=

L

(

w

)

=

t

+

2

,

whichinduces themorphism

σ

t+1,t+2 fortheinfinitewordr oflengths.One continuesinthesameway,usingnow that

β

=

α

+

t. Thistime, the integers A B

(

n

)

occur at positions in A with correspond tothe tth letter inthe words xi from

{

u1

,

...,

ut−1

,

v

,

w

}

.Hereitisthetth letter,becausethefirsttermofthesequence

(

A

(

B

(

n

))

occursatpositionB

(

1

)

=

t

+

1

inthesequence

(

A

(

n

))

.HereB

(

1

)

= β

= 

α

+

t



=

t

+

1,sinceasimplecomputationshowsthat1

<

α

<

2 forallt. Removal of the A B

(

n

)

is then performed by adding the tth and the

(

t

+

1

)

th letter in the x

i. This operation turns

u1

=

1t−121 into

δ(

1

)

=

1t−13,u2

=

1t−2211 into

δ(

2

)

=

1t−222 anduj

=

1tj21j into

δ(

j

)

=

1tj21j−22,for j

=

3

,

...,

t

1.

Moreover,thetwowordsv

=

21t2

,

w

=

121t−12 areturnedinto

δ(

t

)

=

21t−222,respectively

δ(

t

+

1

)

=

121t−322. Theconclusionisthatthisdecoration

δ

mapsthefixedpointof

τ

tothefirstdifferences



A A.

2

Corollary9.Hereisawaytowrite



A A

=

11312221222

. . .

asamorphicwordforthecaset

=

3,i.e.,

α

= (

13

1

)/

2,thebronze mean.Let

θ

on

{

1

,

. . . ,

6

}

bethemorphismgivenby

(7)

θ

:

1

123

,

2

164

,

3

5145

,

4

1645

,

5

123

,

6

164

.

Lettheletter-to-lettermorphism

λ

begivenby

λ

:

1

1

,

2

1

,

4

2

,

5

2

,

6

2

,

3

3

.

Then



A A

= λ(θ

(

1

))

.

Proof. This corollary isderived fromTheorem 7 byusing the‘natural’ algorithm given, forexample,by Honkala in[16], Lemma4.Honkala’srequirementof‘cyclicity’inthatlemmaisnotnecessary.

Tomakethispapermoreself-containedwegiveadescriptionofthis‘natural’algorithm.

Let x be a fixed point of a morphism

τ

on an alphabet A, and

δ

:

A

B a decoration. Let d

(

a

)

:= |δ(

a

)

|

, so that

δ(

a

)

= δ

1

(

a

)

. . . δ

d(a)

(

a

)

foreacha

A.The‘natural’ algorithmconsistsofreplacingeach lettera inx by d

(

a

)

copiesofthe

lettera,denotedasC1

(

a

),

. . . ,

Cd(a)

(

a

)

.Thelettertolettermap

λ

onthealphabetAC

:= {

Cj

(

a

)

:

a

A

,

j

=

1

,

. . . ,

d

(

a

)

}

isthen

definedas

λ(

Cj

(

a

))

= δ

j

(

a

)

.Themorphism

τ

inducesalargenumberofmorphisms

θ

onAC,byfirstmappingforeacha the

wordC

(

a

)

:=

C1

(

a

),

. . .

Cd(a)

(

a

)

totheconcatenationofwords

(

a

)

:=

C

(

τ

1

(

a

))

. . .

C

(

τ

t(a)

(

a

))

,when

τ

(

a

)

=

τ

1

(

a

)

. . .

τ

t(a)

(

a

)

,

andthensplitting

(

a

)

intod

(

a

)

words,defining

θ (

Ci

(

a

))

astheith wordinthissplitting.Thesplittingshouldbedonein

suchawaythataprimitivemorphism

θ

results.

IntheproofofCorollary9thealphabetAC hasapriorid

(

1

)

+

d

(

2

)

+

d

(

3

)

+

d

(

4

)

=

14 letters.Thesituationisspecialhere,

since d

(

a

)

=

t

(

a

)

fora

=

1

,

2

,

3

,

4.This suggeststo define

θ

by splittingthe

(

a

)

intothewords C

(

τ

1

(

a

))

till C

(

τ

t(a)

(

a

))

.

Afterprojectingletterswiththesame

θ

-imageandthesame

λ

-imageonasingleletter,thenumberoflettersreducesto6, andoneobtainsthemorphism

θ

inthecorollary.

2

Fraenkel’sTheorem4withthe‘defect’function D

=

D

(

n

)

suggeststhat the



A A sequencescantakemanyvalues.This isnotthecase.

Proposition10.Foranyirrational

α

largerthan1 thesequence



A A

=



(

n

+

1

)

α



α



− 

n

α



α





takesvaluesinanalphabetof sizetwo,threeorfour.

Proof. Weillustratetheproofwiththecase1

<

α

<

2.Then s

:= 

A isaSturmianwordtakingvaluesd

=

1 ord

=

2.So



A A

(

n

)

=

A

(

A

(

n

+

1

))

A

(

A

(

n

))

=

A

(

A

(

n

)

+

d

)

A

(

A

(

n

)),

where d

=

1 or 2

.

We puti

:=

A

(

n

)

.Incased

=

1, A

(

A

(

n

)

+

d

)

A

(

A

(

n

))

=

A

(

i

+

1

)

A

(

i

)

=

1 or2.Incased

=

2, A

(

A

(

n

)

+

d

)

A

(

A

(

n

))

=

A

(

i

+

2

)

A

(

i

)

=

A

(

i

+

2

)

A

(

i

+

1

)

+

A

(

i

+

1

)

A

(

i

)

.SoeitherA

(

A

(

n

)

+

d

)

A

(

A

(

n

))

=

2 or3,orA

(

A

(

n

)

+

d

)

A

(

A

(

n

))

=

3 or4,respectivelyif11,12and21arethesubwordsoflength2ofs,orif12,21and22arethesubwordsoflength2ofs. Whatwefoundisthat



A A takesvaluesin

{

1

,

2

,

3

}

if1

<

α

<

3

/

2,and



A A takesvaluesin

{

1

,

2

,

3

,

4

}

if3

/

2

<

α

<

2.In somecases



A A maytakeonly2values,forexample,if

α

isthegoldenmean.

Theproofforothervaluesof

α

issimilar,exploitingbalancednessoftheSturmianwords

= 

A.

2

Example.Take

α

=

11

/

2

=

1

.

658

. . .

.Then A A

(

n

)

=

1

,

4

,

6

,

9

,

13

,

14

,

18

,

21

,

23

. . .

,so



A A takesthe fourvalues 1

,

2

,

3 and4.

Remark.Oncemore,let

α

=

2.Thedifferencesx2,k

:= (

A B

)

k

− (

B A

)

k,wherek

1,arethe‘commutator’functions.They

are extensivelystudied in[9]. Theyare all similarto x2,1,whichis equalto x2,1

=

A B

B A

=

2Id

A A.One can derive

fromthisthatallcommutatorfunctionsaremorphicwords.

4. EmbeddingsoftheFibonaccilanguageintotheintegers

Let

L

bealanguage,i.e.,asub-semigroupofthefreesemigroupgeneratedbyafinitealphabetundertheconcatenation operation.Ahomomorphismof

L

intothenaturalnumbersisamapS

:

L

→ N

satisfying

S

(

v w

)

=

S

(

v

)

+

S

(

w

),

for all v

,

w

L

.

The classical Frobenius problemasks whetherthe complementof S

(L)

inthe naturalnumbers will be infinite orfinite, andinthelattercasethevalueofthelargestelementinthiscomplement.IntheclassicalFrobeniusproblem

L

isthefull language consistingofallwords overafinitealphabet.We willsolvethisproblemwhen

L

=

L

F i.e.,thesetofall words

occurringinxF,wherexFistheFibonacciword,theinfinitewordfixedbythemorphism0

01

,

1

0.

Recallthat

ϕ

= (

1

+

5

)/

2.ThekeyingredientinthissectionisthelowerWythoffsequence

(



n

ϕ

)

n≥1

=

1

,

3

,

4

,

6

,

8

,

9

,

(8)

Theorem11.([12]) LetS

:

L

F

→ N

beahomomorphism.Definea

=

S

(

0

),

b

=

S

(

1

)

.ThenS

(L

F)istheunionofthetwogeneralized

Beattysequences



(

a

b

)



n

ϕ



+ (

2b

a

)

n



and



(

a

b

)



n

ϕ



+ (

2b

a

)

n

+

a

b



.

WhatremainstobedoneistodeterminethecomplementofthesetS

(L

F)in

N

.Weshallshowthatthecorresponding

infinitewordisalways amorphicword,byrepresentingitasadecorationofafixedpointofamorphism.Itappearsthat thisisamatterofcomplicatedbookkeeping,especiallywhenthetwovaluesS

(

0

)

andS

(

1

)

aresmall.

Thereare threemorphisms f

,

g andh thatplay animportantrole inthissection, whereitisconvenientto lookata andb bothasintegersandasabstractletters.Themorphismsaregivenby

f

:



a

ab b

a

,

g

:



a

baa b

ba

,

h

:



a

aab b

ab

.

Lemma12.LetxFbetheFibonaccisequenceonthealphabet

{

a

,

b

}

,fixedpointof f .ThenthefixedpointxGofg isthesequencebxF,

andthefixedpointxHofh isaxF.

Proof. BerstelandSééboldprovein[6] (Theorem3.1)thatforanymorphicSturmianwordcα onthealphabet

{

a

,

b

}

both acα andbcα areagainmorphicwords.Theirproofisconstructive,andtheygivethemorphisms g andh forcα

=

xFintheir

Example1.

2

HereisaresultthatgivesanideaoftheproofingeneralforthecaseS

(

0

)

>

S

(

1

)

.

Theorem13.LetS

:

L

F

→ N

beahomomorphismdeterminedbya

=

S

(

0

),

b

=

S

(

1

)

.Supposethat

a

+

2

<

2b

+

1

<

2a

1

.

Thenthefirstdifferencesofthecomplement

N

\

S

(

L

F)ofS

(

L

F)isthewordobtainedbydecoratingthefixedpointxHofthemorphism

h bythemorphism

δ

givenby

δ(

a

)

=

1b−22 1ab−22

,

δ(

b

)

=

12ba−22 1ab−22

.

Proof. ThesequenceoffirstdifferencesofageneralizedBeattysequence



p



n

ϕ



+

qn

+

r

)

isthefixedpointoftheFibonacci morphism f on the alphabet

{

2p

+

q

,

p

+

q

}

.This followsdirectly formLemma2, see alsoLemma8 in[3]. Sothe two generalizedBeattysequencesG1

:=



(

a

b

)



n

ϕ



+(

2b

a

)

n



andG2,givenbyG2

(

n

)

=

G1

(

n

)

+

a

b inTheorem11havethe

propertythat



G1

= 

G2isthefixedpointxFoftheFibonaccimorphismonthealphabetwithsymbols2

(

a

b

)

+

2b

a

=

a

anda

b

+

2b

a

=

b.

Weillustratetheproofbyfirstconsideringthecasea

=

8

,

b

=

5.Inthiscasewehave

G1

=

5

,

13

,

18

,

26

,

34

,

39

,

47

,

52

,

60

, . . . ,

G2

=

G1

+

3

=

8

,

16

,

21

,

29

,

37

,

42

,

50

,

55

,

63

, . . . ...

Partitionthepositiveintegers

N

intoadjacentsetsVi

,

i

=

1

,

2

,

. . .

definedby

Vi

= {

G2

(

i

1

)

+

1

, . . . ,

G2

(

i

)

}.

Here we put G2

(

0

)

=

0. As a consequence, Card

(

Vi

)

=

8 if xH

(

i

)

=

a and Card

(

Vi

)

=

5 if xH

(

i

)

=

b, where xH

=

xH

(

1

)

xH

(

2

)

· · · =

aabaab

. . .

is the fixed point of h. The reason that the directive sequence is xH instead of xF is that

thelast elementofeach Vi isequaltoG2

(

i

)

fori

=

1

,

2

,

. . .

.

V1 V2 V3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

     

Inthetableabove,theintegersinG1

(N)

aremarkedwith



,thoseinG2

(N)

with



,andthoseinthecomplementwitha

.Byconstruction,all theVi withcardinality8havethesamepattern

 

fortheirmembers.Alsoall Vi with

cardinality5havethesamepattern

 

.Notethatthelasttwosymbolsare



,forbothsize5andsize8Vi’s,and

theirfirstsymbolsare

forboth.Thisimpliesthatifwegluethepatternstogether,thentheinfinitesequenceofdifferences ofthepositionsof

intheinfinitepatternyieldsfirstdifferencesofthesequenceofelementsin

N

\ (

G1

(N)

G2

(N))

.For

Vi ofsize8thesedifferences(includingthe‘jumpover’lastvalue2)aregivenby1,1,1,2,1,2,andforVi ofsize5by2,1,2.It

followsthatthefirstdifferencesareobtainedbydecoratingthefixedpointxHbythemorphism

δ

givenby

δ

:

a

111212

,

b

212

.

Forthe generalcase one considers sets Vi ofconsecutive integers ofsize a or sizeb, wherethe order isagaindictated

(9)

positioneda

b placesbeforetheend.ItfollowsagainthatovertheVi’sthefirstdifferencesofthecomplementsetendin

2(the‘jumpover’value),areprecededbya

b

2 1’s,whichisprecededbya2.Thefirstdifferencesstartwithanumber of1’s,whichis

(

a

2

)

1

− (

a

b

2

)

1

=

b

2 forthe Vi’soflengtha,and

(

b

2

)

1

− (

a

b

2

)

1

=

2b

a

2

fortheVi’soflengthb.Thisyieldsthedecoration

δ

statedinthetheorem.

2

WenowgiveanexampleofthedifficultiesoneencounterswhenS

(

0

)

orS

(

1

)

are(relatively)small.

Theorem14.LetS

:

L

F

→ N

bethehomomorphismdeterminedbya

=

S

(

0

)

=

3

,

b

=

S

(

1

)

=

1.Thenthesequenceoffirstdifferences

ofthecomplement

N

\

S

(

L

F

)

ofS

(

L

F

)

isthewordobtainedbydecoratingthefixedpointxHofh by

δ

: {

a

,

b

}

→ {

7

,

11

}

givenby

δ(

a

)

=

7

,

11,and

δ(

b

)

=

11.

Proof. AccordingtoTheorem11,S

(L

F

)

istheunionofthetwosets G1

(N)

andG2

(N)

givenby

G1

(

N)

= {

2

n

ϕ

 −

n

,

n

1

} =

1

,

4

,

5

,

8

,

11

,

12

. . . ,

G2

(

N)

= {

2

n

ϕ

 −

n

+

2

,

n

1

} =

3

,

6

,

7

,

10

,

13

,

14

. . . .

Thefirstdifferences



G1

= 

G2aretheFibonacciwordonthealphabet

{

3

,

1

}

.Imitatingtheproofoftheprevioustheorem,

weobtainthefollowingtable,inducedbythemorphismh givenby1

331

,

3

31.OnehasCard

(

Vi

)

=

a

=

3 ifxH

(

i

)

=

a

andCard

(

Vi

)

=

b

=

1 ifxH

(

i

)

=

b,wherexH

=

xH

(

1

)

xH

(

2

)

· · · =

aabaababaa

. . .

isthefixedpointofh.

V1 V2 V3 V4 V5 V6 V7 V8 V9 V10

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

                    

Thereareatleasttwothingswrongwiththis:

[E1] The Vi’soflength3donotallhavethesamepattern,

[E2] Therearepatternsthatdonotcontaina

.

To counter theseproblems,wego fromtheletters a

=

3

,

b

=

1 tothewordsh

(

3

),

h

(

1

)

, yieldinga partition with Wi’sof

length7and4.Thetableweobtainis

W1 W2 W3 W4

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

                    

Problem[E1]iscausedbythefactthatVi’soflength3havedifferentpatternsdependingonwhethertheyarefollowedby

a Vioflength1oroflength3.Problem[E1]isnowsolvedwiththeWi’s,since33canonlyoccurasaprefixofh

(

1

)

=

331,

and31canonlyoccurasasuffixofeitherh

(

1

)

orh

(

3

)

.

However,[E2]isnotyetsolved,since W3doesnotcontaina

.Thewaytotacklethisistopasstothesquareofh,i.e.,

taketheWi’soflength18and11correspondingtoh2

(

1

)

=

33133131 andh2

(

3

)

=

33131.

It isobviousfromthecorresponding patterns,that thedifferencesofthecomplement

N

\

S

(L

F)are givenby the

dec-oration W1

7

,

11

,

W3

11 oftheWi’s.Butsinceh2

(

x

H

)

=

xH,thisisthesameasdecoratingthelettersa

7

,

11,and

b

11 inxH.

2

Remark15.Theorem25in[3] statesthatthethreesequences

(

2



n

ϕ



n

,

n

1

)

= (

1

,

4

,

5

,

8

,

11

,

12

. . . ),

(

2



n

ϕ



n

+

2n

1

)

,andz

:= (

4



n

ϕ



+

3n

+

2

,

n

0

)

= (

2

,

9

,

20

,

27

,

. . .)

formacomplementarytriple.FromLemma2appliedwiththeSturm number

α

=

ϕ

1 onededucesthat



z

=

7x11,7 theFibonaccisequenceonthealphabet

{

11

,

7

}

,precededbytheletter7

(seealsoLemma8in[3],whichstatesthatifV

(

n

)

=

p

(



n

ϕ

)

+

qn

+

r then



V

=

x2p+q,p+q).

Ontheother hand,we haveTheorem14,tellingusthat



z

= δ(

xH

)

,where

δ

isthedecorationa

7

,

11,andb

11.

Applyingthe‘natural’algorithmto

δ(

xH

)

,weobtainthat

δ(

xH

)

isthemorphicwordobtainedbyapplyingtheletter-to-letter

map

λ(

a

)

=

11

,

λ(

b

)

=

7 tothefixedpointxGofthemorphism g.Thus

δ(

xH

)

= λ(

xG

)

= λ(

bxF

)

= λ(

b

)λ(

xF

)

=

7

λ(

xF

)

=

7x11,7

.

Conclusion:Theorem14isessentiallyequaltoTheorem25in[3],buthasacompletelydifferentproof.

2

WeletC betheincreasingsequenceofintegersinthecomplementofS

(

L

F

)

,soC

(N)

= N \

S

(

L

F

)

.

Theorem16.LetS

:

L

F

→ N

beahomomorphism.Thenthesequence



C offirstdifferencesofthecomplement

N

\

S

(

L

F)ofS

(

L

F)

Cytaty

Powiązane dokumenty

To study various properties of continuous subadditive processes we gen- eralize some old definitions including those of central, singular and global exponents (Definition 2.4) and

Key words and phrases : evolution problem, stable family of operators, stable approx- imations of the evolution operator, fundamental solution, Cauchy problem, uniformly correct

It should be stressed here that the possibility of almost sure approximation of A by the multiples of orthog- onal projections in L 2 (µ) depends heavily on the properties of

[4] —, Pad´e approximation for infinite words generated by certain substitutions, and Hankel determinants, in: Number Theory and Its Applications, K. W e n, Some properties of

A family of elliptic Q-curves defined over biquadratic fields and their

This is a Ramsey-type question (see [3], [4]) for integers: when is it true that every partition is such that at least one of the parts has a certain property.. Our aim in this note

The factorization has been completed, but the factor U is exactly singular, and division by zero will occur if it is used to solve a system of equations... The leading dimensions

Zamiast oblicza´ c go od razu wprost z definicji, poszukuj¸ ac najwi¸ ekszego nieze- rowego minora, mo˙zemy po prostu przekszta lca´ c macierz bez zmiany rz¸ edu do takiej postaci,