• Nie Znaleziono Wyników

(1)Distin t zeros of L-fun tions by E

N/A
N/A
Protected

Academic year: 2021

Share "(1)Distin t zeros of L-fun tions by E"

Copied!
11
0
0

Pełen tekst

(1)

Distin t zeros of L-fun tions

by

E. Bombieri (Prin eton, N.J.)and A. Perelli (Genova)

1. Introdu tion. Let L

1

(s) and L

2

(s) be two \independent" L-fun -

tions, where themeaning of \independent" willbe lari edlater on. Sin e

theL-fun tionsaredeterminedbytheirzeros,wemayexpe tthatL

1

(s)and

L

2

(s)have few ommon zeros. Thisproblemappearsto bevery diÆ ultat

present,thereforewemayasktheeasierquestionofgettingafairquantityof

distin tzeros of su h fun tions. Inthispaperwe showthat, under suitable

onditions,L

1

(s) and L

2

(s) have a positiveproportionofdistin tzeros.

We state our results in the moderately general setting of Bombieri{

Hejhal's paper [1℄, whi h also provides thebasi ingredients of the present

paper. Moreover, we willwork outour maintool,Theorem 2below,in the

ase of several L-fun tions. Hen e, for a given integer N 2, we onsider

N fun tions L

1

(s);:::;L

N

(s) satisfyingthefollowingbasi hypothesis.

HypothesisB.(I)Ea hfun tionL

j

(s)hasanEulerprodu toftheform

L

j (s)=

Y

p d

Y

i=1 (1

ip p

s

) 1

withj

ip jp



for some xed 0<1=2 andi=1;:::;d.

(II) For every ">0 we have

X

px d

X

i=1 j

ip j

2

x 1+"

:

(III) The fun tions L

j

(s) have an analyti ontinuation to C as mero-

morphi fun tions of nite order with a nite number of poles, all on the

line =1,and satisfy a fun tional equationof the form

(s)="(1 s);

where(s)=Q s

Q

m

i=1 (

i s+

i

), Q>0, 

i

>0,Re

i

0 andj"j=1.

1991 Mathemati sSubje tClassi ation: Primary11M41.

(2)

(IV ) The oeÆ ients a

j

(p) of the Diri hlet series

L

j (s)=

1

X

n=1 a

j (n)n

s

satisfy

X

px a

j (p)a

k (p)

p

jk n

j

loglogx+

jk +O



1

logx



for ertain onstants n

j

>0.

Weexpli itlyremarkthatallthedatainvolvedinHypothesisB on ern-

inga fun tionL

j

(s) maydependon j. We also remarkthatthe onditions

of Hypothesis B may be somewhat relaxed (see Selberg [10℄) in order to

dedu e ourresultsbelow.

Wereferto Se tion3of[1℄ forathoroughdis ussionofHypothesisB,of

itsstandard onsequen esand ofseveralexamplesoffun tionssatisfyingit.

HerewepointoutonlythatB(II) impliesthatboththeDiri hletseriesand

theEulerprodu tof L

j

(s) onverge absolutelyfor >1,B(I) ensuresthat

L

j

(s) 6=0 for >1 and B(III) gives riseto thefamiliar notions of riti al

strip, riti allineand trivialand non-trivialzeros. Moreover, writing



j

= m

X

i=1



i

;

N

j

(t)=jf%:L

j

(%)=0; 0Re%1 and 0Im%tgj

and

S

j (t)=

1

 argL

j

(1=2+it);

forsuÆ ientlylargetwe have

(1) N

j (t)=



j



tlogt+

j t+

0

j +S

j

(t)+O(1=t)

with ertain onstants

j and

0

j .

Condition B(IV), introdu ed bySelberg [10℄, plays a spe ialrole, sin e

it provides a form of \near-orthogonality" of the fun tions L

j

(s); the \in-

dependen e" alluded to at the beginning of the se tion omes from this

\near-orthogonality". For instan e, B(IV) implies that L

1

(s);:::;L

N (s)

arelinearlyindependentoverC; seeBombieri{Hejhal[1℄and Ka zorowski{

Perelli[7℄forfurther resultsinthisdire tion.

We expe t that the fun tions L

j

(s) satisfy the Generalized Riemann

Hypothesis. Asa substitute of itinour arguments, we willinstead assume

thefollowingdensityestimate. Let

N (;T)=jf%:L (%)=0; Re% and jIm%jTgj:

(3)

Hypothesis D.Thereexists 0<a<1 su h that

N

j

(;T)T

1 a( 1=2)

logT

uniformly for 1=2 and j=1;:::;N.

The mainpoint in introdu ingHypothesis D is that, unlike the Gener-

alizedRiemann Hypothesis,it an beveri edin manyinteresting ases. In

fa t, it has been proved by Selberg [9℄ for the Riemann zeta fun tion, by

Fujii[5℄ for Diri hlet L-fun tions, and byLuo[8℄ in themore diÆ ult ase

of L-fun tionsatta hed to ertainmodularforms.

In order to state our main result, we de ne the ounting fun tion

D(T;L

1

;L

2

) of the distin t non-trivial zeros, ounted with multipli ity,of

two fun tions L

1

(s) and L

2 (s) as

D(T;L

1

;L

2 )=

X

0Re%1

0Im%T

max(m

1

(%) m

2 (%);0);

where%runsoverthezerosofL

1 (s)L

2

(s)andis ountedwithoutmultipli -

ity. We also de ne

D(T)=D(T;L

1

;L

2

)+D(T;L

2

;L

1 )=

X

0Re%1

0Im%T jm

1

(%) m

2 (%)j;

withthe same onvention about%.

Our mainresult is

Theorem 1.Let L

1

(s) and L

2

(s) satisfy Hypotheses Band D and sup-

pose that 

1

=

2

. Then

D(T;L

1

;L

2

)TlogT:

Clearly,thesame lower boundholdsforD(T;L

2

;L

1

) andD(T)too.

The rst resultof thistypehasbeenobtainedbyFujii[6℄in the aseof

twoprimitiveDiri hletL-fun tions,bymeansofSelberg'smomentsmethod.

The problem of ounting strongly distin t zeros, i.e., zeros pla ed at dif-

ferent points, appears to be more diÆ ult, and the best result is due to

Conrey{Ghosh{Gonek [3℄, [4℄. They deal withthis problem, inthe ase of

two primitive Diri hlet L-fun tions, by onsidering the more diÆ ultques-

tion of getting simple zeros of L(s;

1 )L(s;

2

), and show that there are

 T 6=11

su h zeros up to T. Moreover, if the Riemann Hypothesis is as-

sumedforone ofthetwofun tions,thenapositiveproportionofsu hzeros

is obtained. However, the te hniques in[3℄ and [4℄ do not extend to over

the ase of moregeneral L-fun tions, su hasGL

2

L-fun tions.

Let us all oprime two fun tions in Selberg's lass S (see [10℄) ea h

(4)

su h that there are no ommon fa tors of su h fa torizations. Assuming

Selberg's Conje tures 1.1 and 1.2 in [10℄, we see that B(IV) holds for o-

prime fun tions. Hen e, assuming Hypothesis D for every fun tion in S,

we may regardthelowerboundinTheorem 1,inthe ase of oprimefun -

tions, as a onsequen e of Selberg's onje tures. Another onsequen e of

Selberg's onje turesisthat S hasuniquefa torization (see Conrey{Ghosh

[2℄). Weremarkherethat thelatter onsequen eof Selberg's onje turesis

easilyimplied by a very weak form of theformer. Pre isely,assumingthat

two oprimefun tions inS have D(T) 1 forsuÆ iently large T, we get

theunique fa torization in S. In fa t, the assumption impliesthat two o-

primefun tions arene essarily distin t,andthis learlyimpliesthe unique

fa torization.

Theorem 1appears to bethe limitofour method,althoughmu h more

is expe ted to hold. Forinstan e, if L

1

(s) and L

2

(s) are distin t primitive

fun tions, we expe t that almost all zeros of L

1

(s) and L

2

(s) are distin t,

i.e.,

D(T)



1 +

2



TlogT;

inwhi h ase almostall zeros area tually stronglydistin t,oreven that

D(T)=N

1

(T)+N

2

(T)+O(1);

i.e., L

1

(s) and L

2

(s) have O(1) ommon non-trivialzeros.

The proof of Theorem 1 is based on Bombieri{Hejhal's [1℄ variant of

Selberg's[9℄moments method,whi hleadsinamore dire twayto thedis-

tributionfun tionforthelogL

j

(1=2+it)(see TheoremBof[1℄). Although

we ould followa variant more inthe spiritof Selberg [9℄ and Fujii[6℄, we

will prove Theorem 1 by means of a short intervals analog of the above

mentionedTheoremB,whi hwe believeto beof interestinitself.

Let M 10, writeh=M=logT and

V

j (t)=

logL

j

(1=2+i(t+h)) logL

j

(1=2+it)

(2n

j

logM) 1=2

;

and let

T

denote theasso iatedprobabilitymeasureon C N

,de ned by

(2) 

T ()=

1

T

jft2[T;2T℄:(V

1

(t);:::;V

N

(t))2gj

for every open set  C N

. Moreover, let e

kzk 2

denote the gaussian

measure on C N

and let d!be theeu lideandensityon C N

.

Theorem 2. Let L

1

(s);:::;L

N

(s) satisfy Hypotheses B and D and let

M =M(T)!1withM (logT)=loglogT asT !1. Then,asT !1,

 tends to the gaussian measure withasso iated density e

kzk 2

d!.

(5)

We remark thatwe an easily get a slight variant of Theorem 2,where

h=M=logt and M =M(t)! 1with M log 1 "

t ast!1. Therefore,

ifwe separatetheV

j

(t)into theirrealand imaginaryparts,Theorem2 an

beexpressed bysayingthat thefun tions

log

L

j 1

2

+i t+ M

logt



log

L

j 1

2 +it



(2n

j

logM) 1=2

; j=1;:::;N;

and

argL

j 1

2

+i t+ M

logt



argL

j 1

2 +it



(2n

j

logM) 1=2

; j=1;:::;N;

be ome distributed, in the limit of large t, like independent random vari-

ables, ea h having gaussian density exp( u 2

)du, provided M ! 1 with

M log 1 "

tas t!1.

A knowledgments. The se ond named author wishes to thank the

InstituteforAdvan edStudy forits hospitalityand forprovidingex ellent

working onditions.

2. Basi lemmas. Inthisse tionwefollowtheargumentsinSe tion5

of Bombieri{Hejhal[1℄. For>1and j=1;:::;N we write

logL

j (s)=

1

X

n=1

j (n)

1 (n)n

s

; 

1 (n)=



0; n=1;

(n)=logn; n2;

and denote by u(x) a real positive C 1

fun tion with ompa t support in

[1;e℄ andbyu(s)e its Mellin transform. Wealso write

v(x)= 1

\

x

u(t)dt

and assumethat u isnormalized sothatv(0)=1. We refer to Lemma1 of

[1℄ and theremarkfollowingitforrelevant propertiesof eu(s).

By (5.4) of [1℄ we havethe approximateformula

logL

j

(1=2+it)= 1

X

n=1

j (n)

1 (n)

n 1=2+it

v(e

(logn)=logX

) (3)

+ X

% 1

\

1=2 1

% s e

u(1+(% s)logX)d+O(1);

where jtjis suÆ ientlylarge and nottheordinate of azero of L

j

(s),where

2  X  t 2

and where % runs over zeros of L

j

(s) with 0  Re%  1. We

write(3)as

logL

j

(1=2+it)=D

j

(1=2+it;X)+R

j

(1=2+it;X)

whereD (1=2+it;X) is theDiri hletserieson theright handsideof (3).

(6)

FromLemma 3 of[1℄ we immediatelygetour rstbasi lemma.

Lemma 1. Assume Hypotheses B and D, and let 2 X  T a=2

and T

suÆ iently large. Then for j=1;:::;N we have

2T

\

T jR

j

(1=2+it;X)jdtT logT

logX :

Our se ond basi lemma is a short intervals analog of Lemma 6 of [1℄,

i.e., themixedmomentsofthedi eren es oftheD

j

(1=2+it;X). Sin e the

proofof Lemma2belowfollowsthatof Lemma6of [1℄,we willonlysket h

it. ForsuÆ ientlylargeM,writeh=M=logT and



j

(t)=D

j

(1=2+i(t+h);X) D

j

(1=2+it;X):

Moreover, let k

j

 0 and l

j

 0, j = 1;:::;N, be integers and let us

abbreviatek=(k

1

;:::;k

N ),K

j

=k

1

+:::+k

j

,K =K

N

and similarlyfor

l;L

j

and L. We also writek!= Q

N

j=1 k

j

!.

We state here the basi estimate we will repeatedly use in the proof of

Lemma2. ForX3 we have

(4) X

p a

j (p)a

k (p)

p

v(e

(logp) =logX

) 2

je

ihlogp

1j 2

jk 2n

j log

+



h

2 logX



+O(1)

uniformly for h  1=loglogX, where log +

x = max (logx;0). In fa t,

je

ihlogp

1j = 4sin 2

((h=2)logp) and hen e (4) follows from B(IV) by

partialsummation(see also (3.8) of [1℄).

Lemma 2. Assume Hypothesis B and let X  T

1=(K+L+1 )

and M 

(logT)=loglogX. Write



j (t)=

1

X

n=1 b

j (n)

n 1=2+it

; b

j

(n)=

j (n)

1 (n)v(e

(logn) =logX

)(e

ihlogn

1):

Then

2T

\

T N

Y

j=1 (

j (t))

k

j

(

j (t))

l

j

dt =Æ

k;l k!T

N

Y

j=1



2n

j log

+



M

2 logT

logX



k

j

+O



T



log +



M

2 logT

logX



(K+L 1)=2



:

Proof. Wemay learlyassumethatK+L1. Fornotationalsimpli -

ity,weabbreviate 

j

=

j

(t). Sin e 

j

issupported atprime powers only,

we splitit as



j

= 0

+ 00

(7)

where 0

j

rangesover primesp and  00

j

over primepowersp r

,r2. Then,

a ordingly,we get

(5)

N

Y

j=1 (

j )

k

j

(

j )

l

j

= N

Y

j=1 (

0

j )

k

j

(

0

j )

l

j

+R (t);

where, asintheproof of Lemma6 of[1℄,

(6)

2T

\

T

jR (t)jdt 2T

\

T j

00

j1 jj

0

j2 j

K+L 1

dt+ 2T

\

T j

00

j3 j

K+L

dt

fora suitable hoi e of j

1

;j

2 and j

3 .

Sin e e

ihlogn

11, by(5.14) of [1℄ we have

(7)

2T

\

T j

00

j j

2(K+L)

dtT

forj=1;:::;N,providedX T

1=(K+L+1 )

.

By Montgomery{Vaughan'smean-valuetheoremforDiri hletpolynomi-

als (see,e.g.,Lemma 4 of[1℄) wehave

2T

\

T j

0

j j

2(K+L)

dt=T X

jB 0

(n)j 2

n

+O



X

jB 0

(n)j 2



;

where

B 0

(n)=

X

p1:::p

K+L

=n b

j (p

1 ):::b

j (p

K+L ):

Sin e

j

(p)=a

j

(p)and 

1

(p)=1,from (5.16) of[1℄ and (4)we get

X

jB 0

(n)j 2

n

(K+L)!



X

p jb

j (p)j

2

p



K+L





log +



M

2 logT

logX



K+L

:

Moreover, from (5.17) of [1℄ we obtain

X

jB 0

(n)j 2

X

(1+")(K+L)

;

and hen e

(8)

2T

\

T j

0

j j

2(K+L)

dtT



log +



M

2 logT

logX



K+L

providedX T

1=(K+L+1 )

.

From (6){(8)and Holder'sinequalitywe get

(9)

2T

\

T

jR (t)jdtT



log +



M

2 logT

logX



(K+L 1)=2

1=(K+L+1 )

(8)

In order to treat themainprodu t on theright handsideof (5)we use

again Lemma4 of[1℄. We abbreviaten=(n

1

;:::;n

K ),

b(n;k)= N

Y

j=1 K

j

Y

r=K

j 1 +1

b

j (n

r

) and B(n;k)= X

n

1 :::n

K

=n

b(n;k);

and asin(5.18) of [1℄ we have

2T

\

T N

Y

j=1 (

0

j )

k

j

(

0

j )

l

j

;dt =T X

B(n;k)B(n;l)

n (10)

+O



X

jB(n;k)j 2



1=2



X

jB(n;l)j 2



1=2



;

where the sums are restri ted to n of type n = p

1 :::p

K

for k and n =

q

1 :::q

L

for l; here p and q denote prime numbers. By a variant of the

argument leading to (8)we seethat

(11)



X

jB(n;k)j 2



1=2



X

jB(n;l)j 2



1=2

T



log +



M

2 logT

logX



(K+L 1 )=2

providedX T

1=(K+L+1 )

.

In view of (5), (9), (10) and (11), to omplete theproof of Lemma 2 it

suÆ es to showthat

X

B(n;k)B(n;l)

n

k;l k!

N

Y

j=1



2n

j log

+



M

2 logT

logX



k

j

(12)

+O



log +



M

2 logT

logX



(K+L 1 )=2



:

IfK 6=Lthere isnothingtoprove,sin eB(n;k)B(n;l)=0forevery n; we

an therefore assume K =L1 and pro eed by indu tionasinLemma 6

of [1℄.

If K =1, (12)follows immediately from (4). Supposenow that K 2.

Arguing again as inLemma 6 of [1℄ and using (3.8) of [1℄, we see that the

ontribution to the left hand side of (12) oming from n's whi h are not

square-free is





log +



M

2 logT

logX



(K+L 1 )=2

:

In order to deal with the remaining part of the sum on the left hand side

(9)

take into a ount thefa tor e

ihlogn

1 in ourde nition of theb

j

(n). In

thiswaywesee that(12) holdsforanyK 1,and Lemma 2is proved.

We remarkthat we an easily obtain a version of Lemma 2 with h re-

pla edbyM=logt, providedanadditional errorterm

O



T(loglogX) K+L

MlogX

log 2

T



isaddedinthestatementofLemma2. Weleaveitsveri ationtothereader.

3. Proof of theorems. Theproofof Theorem2 follows loselythatof

TheoremBof [1℄. LetM !1 asT !1 and hoose

logX=

logT

(logM) 1=4

;

sothat

log +



M

2 logT

logX



logM;

logX

logT

=(logM) 1=4

; X=T o(1)

:

Moreover, let

U

j

(t)=(2n

j

logM) 1=2



j (t)

and e

T

be theasso iatedprobabilitymeasure on C N

,de ned asin(2).

Then, assuming that M  (logT)=loglogT and arguing exa tly as in

theproofof TheoremBof [1℄,from Lemma2 we see thate

T

onverges, as

T ! 1, to the gaussian measure e

kzk 2

. Also, from Lemma 1 we easily

dedu e that

1

T 2T

\

T jV

j

(t) U

j

(t)jdt(logM) 1=4

;

andhen e

T

onvergestothesamegaussianmeasure, ompletingtheproof.

The proofofTheorem1 isby ontradi tion. LetT



bea sequen ealong

whi h

D



:=D(2T



;L

1

;L

2

) D(T



;L

1

;L

2

)=o(T

 logT

 ):

We set

(13) M



=min



logT



loglogT



; r

T

 logT



1+D





:

Then M



! 1 and M



 (logT



)=loglogT



, sothat Theorem2 isappli-

able to L

1 ,L

2

and thesequen e T

 ,M

 .

Write

h



=M



=logT



;



N (t;h

 )=(N

1 (t+h



) N

1

(t)) (N

2 (t+h



) N

2 (t));

 (t;h )=(S (t+h ) S (t)) (S (t+h ) S (t))

(10)

and observe that (1)and 

1

=

2 imply

(14) 

N (t;h

 )=

S (t;h

 )+O



M



logT





uniformlyfort2[T



;2T



℄.

Forj=1;2and t2[T



;2T



℄we have

(15) ImV

j (t)=



(2n

j logM

 )

1=2 (S

j (t+h

 ) S

j (t)):

Thusfrom (14) and (15)we see thatift2[T



;2T



℄is su h that

(16) ImV

2

(t)<0 and ImV

1

(t)>1;

then



N (t;h

 )=

1

 (2n

1 logM

 )

1=2

ImV

1 (t)

1

 (2n

2 logM

 )

1=2

ImV

2

(t)+O(M



=logT

 )

 1

 (2n

1 logM

 )

1=2

+O(M



=logT

 ):

Denote byE



thesetof t2[T



;2T



℄forwhi h(16) holds.

In order to geta lower boundforjE



j,we onsidertheset

=f(z

1

;z

2 )2C

2

:Imz

1

>1 and Imz

2

<0g;

sothat

(17) jE

 j=T





T

():

From Theorem2 we obtain

(18) lim

!1



T

()=

\

e

kzk 2

d!1:

From (15), (17) and (18) we see that jE

 j T



, and hen e we dedu e

theexisten e of T



=h



values t

r 2[T



;2T



℄, withjt

r t

s jh



if r6=s,

su h that



N (t

r

;h

 )

1

 (2n

1 logM

 )

1=2

+O(M



=logT

 ):

Therefore

(19) D



 X

r



N (t

r

;h

 )

p

logM



M

 T

 logT

 :

Nowre all that

(13) M



=min



logT



loglogT

; r

T

 logT



1+D



:

(11)

If in(13)we have M



=(logT



)=loglogT



wemust also have

D



T



(loglogT

 )

2

logT



;

while (19) gives D



 T



(loglogT

 )

3=2

, a ontradi tion. The other al-

ternative in (13) gives M



= p

(T

 logT



)=(1+D



) , whi h substituted in

(19) shows that D



 T

 logT



; this ontradi ts our assumption D



=

o(T

 logT

 ).

The proof of Theorem1is omplete.

Referen es

[1℄ E.BombieriandD.A.Hejhal,Onthedistributionofzerosoflinear ombinations

of Eulerprodu ts,DukeMath.J.80(1995),821{862.

[2℄ J.B.ConreyandA.Ghosh,OntheSelberg lassofDiri hletseries:smalldegrees,

ibid.72(1993),673{693.

[3℄ J.B.Conrey,A.Ghoshand S.M. Gonek,Simplezerosof thezeta fun tion of

aquadrati number eld,I,Invent.Math.86(1986),563{576.

[4℄ |,|,|,Simplezerosofthezetafun tionofaquadrati number eld,II,in:Ana-

lyti NumberTheoryandDioph.Probl.,A.C.Adolphsonetal.(eds.),Birkhauser,

1987,87{114.

[5℄ A.Fujii,Onthe zerosofDiri hlet'sL-fun tions. I,Trans.Amer.Math. So .196

(1974),225{235.

[6℄ |,On thezerosof Diri hlet'sL-fun tions.V,A taArith.28(1976),395{403.

[7℄ J.Ka zorowskiandA. Perelli,Fun tionalindependen e of thesingularitiesof

a lassof Diri hletseries,Amer.J.Math.,toappear.

[8℄ W. Luo, Zeros of He ke L-fun tions asso iated with usp forms, A ta Arith.71

(1995),139{158.

[9℄ A. Selberg, Contributions to the theory of the Riemann zeta-fun tion, Ar hiv

Math. Naturvid.48(1946),89{155; Colle tedPapers,Vol. I,Springer,1989, 214{

280.

[10℄ |,Oldandnew onje tures andresultsabouta lass ofDiri hletseries,in:Pro .

Amal Conf. Analyti Number Theory, E. Bombieri et al. (eds.), Universita di

Salerno,1992,367{385;Colle tedPapers,Vol.II,Springer,1991,47{63.

S hoolofMathemati s DipartimentodiMatemati a

InstituteforAdvan edStudy ViaDode aneso35

Prin eton,NewJersey08540 16146 Genova

U.S.A. Italy

E-mail:ebmath.ias.edu E-mail:perellidima.unige.it

Re eivedon24.3.1997 (3156)

Cytaty

Powiązane dokumenty

to shout: 'I don't understand anything! What's all about this black and white thing? We gotta stop racism in any way!! What's this with blacks and whites! I'm gonna make a rule in

It seems quite difficult to obtain more theoretical results on the convergence of the hybrid procedure in the general case... It was introduced in [6, 7] and applied to some well

In Section 3 we apply estimates for linear forms in (p-adic) logarithms of algebraic numbers.. This is derived from

In this paper, we use a zero density estimate for L-functions and Brun’s sieve to obtain an asymptotic formula for N (H, p) which holds uniformly, for “almost all” primes p, in

The object of this paper is to evaluate an integral involving an Hermite polynomial, a generalized hypergeometric series and Fox’s H-function and utilize it to evaluate a

Augment a given network graph with an additional nullor (whose nullator branch and norator branch, respectively, connect the nodes κ 1 and κ 2 with the reference node), and then

the range of space shrouded by this shadow depends on the shape of the illuminated object, on the direction of light incidence and, to a lesser extent, on the distance from the

in this article the issue of architecture as fun or having fun with architecture will be discussed on the basis of some concrete architectural examples, namely the hundertwasserhaus