Connectivity of the Julia set for the Chebyshev-Halley family on degree n polynomials
Abstract
We study the Chebyshev-Halley family of root finding algorithms from the point of view of holomorphic dynamics. Numerical experiments show that the speed of convergence to the roots may be slower when the basins of attraction are not simply connected. In this paper we provide a criterion which guarantees the simple connectivity of the basins of attraction of the roots. We use the criterion for the Chebyshev-Halley methods applied to the degree n polynomials z n + c, obtaining a characterization of the parameters for which all Fatou components are simply connected and, therefore, the Julia set is connected. We also study how increasing n affects the dynamics.
Full text
Connec i i y o he Julia se o he Chebyshe -Halley
amily on deg ee npolynomials
∗B. Campos, †J. Canela and ∗P. Vindel
camp[email p o ec ed], [email p o ec ed], [email p o ec ed]
Abs ac
We s udy he Chebyshe -Halley amily o oo inding algo i hms om he poin o
iew o holomo phic dynamics. Nume ical expe imen s show ha he speed o con e gence
o he oo s may be slowe when he basins o a ac ion a e no simply connec ed. In
his pape we p o ide a c i e ion which gua an ees he simple connec i i y o he basins o
a ac ion o he oo s. We use he c i e ion o he Chebyshe -Halley me hods applied o
he deg ee npolynomials zn+c, ob aining a cha ac e iza ion o he pa ame e s o which
all Fa ou componen s a e simply connec ed and, he e o e, he Julia se is connec ed. We
also s udy how inc easing na ec s he dynamics.
1 In oduc ion
Mos o he p oblems aced by scien is s and enginee s in ol e equa ions ha do no ha e a
known analy ical solu ion. Nume ical me hods a e a good op ion o ackle and sol e eal wo ld
p oblems. In pa icula , i e a i e me hods a e used o ind app oxima ions o he solu ions o
(z) = 0.
The bes -known oo - inding algo i hm is New on’s me hod, which has o de o con e gence
2. Many nume ical me hods o o de h ee o mo e a e de i ed om New on’s scheme: Cheby-
she me hod, also known as supe -New on me hod (see [15], o example), Halley’s me hod and
supe -Halley me hod. A mo e de ailed s udy o he cons uc ion and e olu ion o hese nume -
ical me hods can be seen in [13]. These me hods belong o a amily o nume ical algo i hms
called he Chebyshe -Halley amily, which is gi en by
xn+1 =xn−1 + 1
2
L (xn)
1−αL (xn) (xn)
0(xn),(1)
whe e
L (xn) = (xn) 00 (xn)
( 0(xn))2
∗Ins i u o de Ma em´a icas y Aplicaciones de Cas ell´on (IMAC), Uni e si a Jaume I. Spain
†Uni e si ´e Pa is-Es Ma ne-la-Vall´ee. F ance
1
and α∈C. Wi hin his amily, Chebyshe me hod is ob ained o α= 0, Halley’s me hod is
ob ained o α=1
2and supe -Halley me hod is ob ained o α= 1. Mo eo e , as α ends o ∞
hese algo i hms con e ge o New on’s me hod.
Failu es in in e media e calcula ions made by a compu e a e e y di icul o de ec . Con-
sequen ly, one o he common aims o nume ical analysis is o selec obus algo i hms, ha is,
algo i hms wi h a good nume ical s abili y in a wide ange o si ua ions. The e o e, i is usual
o conside i e a i e me hods wi h high o de o con e gence. Howe e , he adii o con e -
gence which ensu e ha he solu ion o he me hod is co ec may dec ease when we inc ease
he o de o he me hod. One way o add ess his issue is o s udy he nume ical me hod om
a dynamical poin o iew, i.e., o conside he i e a i e me hod as a disc e e dynamical sys em
and o s udy i s s abili y. This is a line o wo k ha has p o en o be especially ui ul in
ecen yea s (see, o example, he pape s [1], [2], [3], [6], [7], [9], [14]).
In his pape we s udy he Chebyshe -Halley amily om a dynamical poin o iew in o de
o ind which membe s o he amily ha e be e s abili y. To ca y ou his dynamical s udy,
he oo - inding algo i hm is applied o a polynomial P. By doing so, we ob ain a a ional map
Q:b
C→b
C, whe e b
Cdeno es de Riemann sphe e, whose dynamics desc ibes how beha es he
me hod when applied o he polynomial P. Indeed, he poin s which con e ge o he oo s o
Pwhen applying o he nume ical me hod a e exac ly hose which con e ge o he oo s o P
when i e a ing he map Q.
We shall gi e a sho in oduc ion o he concep s used in holomo phic dynamics. A mo e
de ailed desc ip ion o he opic can be ound in [4] and [17]. We conside he disc e e dynamical
sys em gi en by he i e a es o a a ional map Q:b
C→b
C. A poin z0∈b
Cis called ixed i
Q(z0) = z0and pe iodic o pe iod p > 1 i Qp(z0) = z0and Qk(z0)6=z0 o k < p. In he la e
case, we say ha hz0i={z0, Q(z0),··· , Qp−1(z0)}is a p-cycle. A poin is p epe iodic i i is
e en ually mapped unde i e a ion o Qon o a pe iodic poin . The mul iplie o a ixed poin
z0is gi en by λ=Q0(z0). Analogously, he mul iplie o a p-cycle hz0iis gi en by (Qp)0(z0).
We say ha a ixed poin o a cycle is a ac ing i |λ|<1 (supe a ac ing i λ= 0), epelling
i |λ|>1, and indi e en i |λ|= 1. In he la e case λ=e2πiθ, whe e θ∈[0,1). We say
ha an indi e en poin o cycle is a ionally indi e en o pa abolic i θ∈Q. Any a ac ing
o pa abolic poin z0has a basin o a ac ion, an open se o poin s which con e ge unde
i e a ion o Q o z0, ela ed o i . We deno e i by
A(z0) = {z∈b
C:Qk(z)→z0when k→∞}.
The basin o a ac ion o an a ac ing (o pa abolic) p-cycle can be de ined analogously using
ha all elemen s o hz0ia e a ac ing (o pa abolic) ixed poin s o Qp.
The dynamics o a ional map Qspli s he Riemann sphe e in o wo o ally in a ian sub-
se s. The Fa ou se ,F(Q), consis s o he z∈b
Csuch ha he amily o i e a es o Q,
{Q(z), Q2(z), . . . , Qk(z), . . .}, is no mal, o equi alen ly equicon inuous, in some open neigh-
bou hood Uo z. The Fa ou se is open and co esponds o he se o poin s wi h s able
dynamics. I s complemen , he Julia se J(Q), is closed and co esponds o he se o poin s
which p esen chao ic beha iou . The connec ed componen s o he Fa ou se , called Fa ou
componen s, a e mapped among hemsel es unde i e a ion. All Fa ou componen s o a a-
ional map a e ei he pe iodic o p epe iodic ([19]). By he Classi ica ion Theo em (see e.g.
[17]), all pe iodic Fa ou componen s a e ei he basins o a ac ion o a ac ing o pa abolic
cycles, o simply connec ed o a ion domains (Siegel disks) o doubly connec ed o a ion do-
mains (He man ings). Mo eo e , all pe iodic Fa ou componen s a e ela ed o a c i ical poin ,
ha is, a poin z∈b
Csuch ha Q0(z) = 0. Indeed, he basin o a ac ion o an a ac ing o
2
a pa abolic cycle con ains, a leas , a c i ical poin . Also, he o bi o , a leas , a c i ical poin
accumula es on he bounda y o a Siegel disk o a He man ing.
When a oo - inding algo i hm is s udied om he poin o iew o holomo phic dynamics,
i is usual o apply he me hod o low deg ee polynomials (see o ins ance [11], [6], [7], [9]).
The eason is ha when he deg ee o he polynomial inc eases, he numbe o c i ical poin s
o he a ional map ob ained also inc eases. This is a se ious d awback when analysing he
pa ame e spaces o he me hods applied o high deg ee polynomials, since he o bi s o he
c i ical poin s a e c ucial o es ablish he exis ence o basins o a ac ion which do no come
om he oo s.
In he pape [5], we conside he Chebyshe -Halley amily o nume ical me hods applied o
he deg ee npolynomials zn+c, whe e z∈Cand c∈C {0}. As a as we know, his is he
i s ime ha a s udy o a amily o oo - inding algo i hms applied o deg ee npolynomials
is conside ed om he poin o iew o dynamical sys ems and he co esponding pa ame e
spaces a e p o ided. In Figu e 1 we show he pa ame e spaces o his amily o di e en alues
o n. Despi e o he inc ease in he numbe o c i ical poin s, we use he symme ies o he
dynamical sys em ob ained o jus i y ha i is enough o ollow he o bi o a single c i ical
poin o de e mine he exis ence o basins o a ac ion o he han he ones p o ided by he
oo s. This p ope y allows us o ob ain pic u es o he pa ame e spaces o polynomials o his
amily. We use hese nume ical s udies oge he wi h some heo e ical esul s o p o ide a i s
analysis o how he se s o pa ame e s wi h good dynamical beha iou a y as ninc eases. Fo
ixed n, we show ha he e exis s a unique pa ame e α o which he Chebyshe -Halley me hod
has o de o con e gence 4. Howe e , he dynamics o such poin s may a y when we inc ease
n. Fo ins ance, o α= 1 (supe -Halley me hod) he algo i hm has o de o con e gence 4 o
n= 2, bu i p esen s bad dynamical beha iou o big n(see Figu es 9, 10 and 11).
The goal o his pape is o con inue he esea ch began in [5]. We ocus on he s udy
o pa hological dynamical beha iou which appea s bo h in he dynamical and he pa ame e
planes. Fi s , we s udy he dynamical condi ions which lead o non-simply connec ed basins o
a ac ion o he oo s. I has been nume ically obse ed ha he basins o a ac ion may ha e
holes, which seem o lead o slowe speed o con e gence. In Sec ion 3 we p o ide a dynamical
condi ion (P oposi ion 3.1) which can be applied o any holomo phic amily o a ional maps
wi h a single ee c i ical poin (modulo symme ies). Using his condi ion, we p o e ha he
basins o a ac ions o he oo s o he Chebyshe -Halley amily a e mul iply connec ed i
and only i he immedia e basin o a ac ion o he oo z= 1 con ains ano he c i ical poin
c6= 1, and no p eimage o z= 1 o he han himsel . This cha ac e iza ion is used o s udy he
connec i i y o hei Julia se s. In Theo em 3.9 we p o e ha he Julia se o he Chebyshe -
Halley ope a o is disconnec ed i and only i he p e ious condi ion holds. These esul s allow
us o loca e he alues o he pa ame e o which he Julia se is disconnec ed, and he e o e,
he nume ical me hods a e mo e uns able.
A e wa ds, we s udy he pa ame e s which p esen bad beha iou when d awing he pa-
ame e planes. Pa ame e planes o he Chebyshe -Halley amily applied o zn−1 a e shown
in Figu e 1, o se e al alues o n. The d awings a e ob ained as ollows. Fo each pa ame e
αwe i e a e a c i ical poin up o 150 imes. I he o bi o he c i ical poin con e ges o a
oo ( ha is, he i e a e wsa is ies |w−ξ|<10−4 o some n h- oo o he uni y ξ) in less
han 150 i e a es, we plo he co esponding poin using a scale om ed ( as con e gence) o
g een o pu ple and o g ey (slow con e gence). I a e 150 i e a es he c i ical o bi has no
con e ged o a oo , we plo he poin in black. See Sec ion 5.1 o a mo e de ailed explana ion
on how he images o pa ame e planes a e p oduced. We obse e in ed pa ame e s o which
3
−1 0 1 2 3 4
−2
−1
0
1
2
(a) n=2
−1 0 1 2 3 4
−2
−1
0
1
2
(b) n=3
−1 0 1 2 3 4
−2
−1
0
1
2
(c) n=5
−1 0 1 2 3 4
−2
−1
0
1
2
(d) n=10
−1 0 1 2 3 4
−2
−1
0
1
2
(e) n=25
−1 0 1 2 3 4
−2
−1
0
1
2
( ) n=100
Figu e 1: Pa ame e spaces o he Chebyshe -Halley amily applied o zn−1.
4
he co esponding c i ical poin s con e ge as o he oo s. We can also obse e he Ca se ,
he se o pa ame e s o which he c i ical poin s do no belong o he basins o a ac ion o
he oo s. In all he igu es we can dis inguish wo big disks called he head an he body o
he Ca . These se s co espond o pa ame e s o which s ange ixed poin s a e a ac ing (see
P oposi ion 2.3 and P oposi ion 2.4). The Ca se consis s o he head and he body (wi h hei
deco a ions) and a necklace-like s uc u e which su ounds he head, ha we call he Colla .
Fo nsmall, we obse e he Colla colou ed in yellow. Howe e , o n∈ {25,100} he e appea
some black disks which do no co espond o s able beha iou , as is he case o he head and
he body. In his pape we analyse a ound which bi u ca ion pa ame e s hese egions appea
(see Sec ion 4 and 5.1), and s udy he associa ed ope a o s om a nume ical poin o iew
(Sec ion 5.2).
The pape is s uc u ed as ollows. In Sec ion 2 we ecall he main p ope ies o he ope a o s
ob ained applying he Chebyshe -Halley me hods o he polynomials zn+cand analyse he
s abili y o he s ange ixed poin s. In Sec ion 3 we p o ide a dynamical condi ion o he
simple connec i i y o he basins o a ac ion o he oo s and he connec i i y o he Julia
se . In Sec ion 4 we s udy di e en bi u ca ion pa ame e s wi hin he Colla o he Ca se .
Finally, in Sec ion 5 we s udy he pa ame e and he dynamical planes o he amily om a
nume ical poin o iew.
2 Dynamical s udy on deg ee npolynomials
We s udy is he Chebyshe -Halley amily o nume ical me hods applied on he polynomials
zn+c. Resul s o his amily applied on polynomials o deg ee wo can be seen in [11], [10],
[12].
In [5] we show ha he ope a o ob ained o (z) = zn−1 is conjuga e o he one ob ained
when applying he Chebyshe -Halley me hod o (z) = zn+c,c∈C−{0}. The e o e, i is
enough o s udy he Chebyshe -Halley me hods o zn−1 o unde s and hem o all zn+c,
whe e c∈C−{0}.
We deno e by On,α(z) he ixed poin ope a o ob ained o zn−1.By subs i u ing (z) =
zn−1 in (1) we ha e:
On,α(z) = z−(zn−1)((−1+2α+n−2αn) + (1 −2α−3n+ 2αn)zn)
2nzn−1(α(n−1)(zn−1) −nzn)=
=(1 −2α)(n−1) + (2 −4α−4n+ 6αn −2αn2)zn+ (n−1)(1 −2α−2n+ 2αn)z2n
2nzn−1(α(1 −n)+(−α−n+αn)zn).(2)
Excep o degene a e cases s udied in Sec ion 4, he deg ee o his ope a o is 2n. The e o e,
i has 2n+ 1 ixed poin s and 4n−2 c i ical poin s.
The ixed poin s a e ob ained by sol ing On,α(z) = z. On one hand, we ob ain he n h - oo s
o he uni y, co esponding o he ze os o he polynomial zn−1, which a e supe a ac ing
ixed poin s. The o he n+ 1 ixed poin s a e z=∞and he nsolu ions o he equa ion
−1+2α+n−2αn + (1 −2α−3n+ 2αn)zn= 0.(3)
These poin s a e called s ange ixed poin s, because hey do no ma ch wi h he solu ions o
he polynomial. The c i ical poin s o he ope a o On,α a e he solu ions o O0
n,α(z) = 0, whe e
O0
n,α(z) = (zn−1)2(n−1)(α(1 −2α)(n−1)2+ (1 −2n−2α+ 2αn) (−α−n+αn)zn)
2nzn(α(n−1) + (α+n−αn)zn)2.(4)
5
The n h- oo s o he uni y a e double c i ical poin s and, hence, a e supe a ac ing ixed poin s
o local deg ee 3. The poin z= 0 is a c i ical poin o mul iplici y n−2 since i is mapped
wi h deg ee n−1 o z=∞. This asse ion ollows om he e m 1/zn−1on Equa ion (2). The
emaining nc i ical poin s a e gi en by
cn,α,ξ =cξ=ξα(n−1)2(2α−1)
n(2n−1) −α(4n−1)(n−1) + 2α2(n−1)21/n
,(5)
whe e ξdeno es an n h- oo o he uni y, ı.e. ξn= 1. The exis ence o any s able beha iou o
he dynamical sys em o he han he basins o a ac ion o he ze os o zn−1 is con olled by
he o bi s o hese n ee c i ical poin s.
The o de o con e gence o he oo s o all membe s he Chebyshe -Halley amily is a
leas 3. The nex esul , which co esponds o [5, P oposi ion 6.3], s a es ha he e is a single
pa ame e o which he o de o con e gence inc eases o 4. In his case, he n h- oo s o he
uni y a e c i ical poin s o mul iplici y h ee and a e supe a ac ing ixed poin s o local deg ee
4.
P oposi ion 2.1. Fo n≥2, he ope a o On,α has o de o con e gence 4 i and only i
α=2n−1
3n−3.
The ollowing lemma, which co esponds o [5, Lemma 6.2], s a es ha he dynamics o he
maps On,α is symme ic wi h espec o he n h- oo o he uni y. I ollows om he lemma
ha he n ee c i ical o bi s a e symme ic and, he e o e, i is enough o con ol one o hem.
Lemma 2.2. Le n∈Nand le ξbe an n h- oo o he uni y, i.e. ξn= 1. Then Iξ(z) = ξz
conjuga es On,α(z)wi h i sel , i.e.
Iξ◦On,α(z) = On,α ◦Iξ(z).
The s abili y o he ixed poin z=∞was s udied in [5], whe e we p o ed he ollowing
p oposi ion.
P oposi ion 2.3. The ixed poin z=∞sa is ies he ollowing s a emen s.
1. I α−1−4n+5n2
2(n−1)(2n−1)<n
2(2n−1), hen z=∞is an a ac o . In pa icula , i is a supe a -
ac ing ixed poin i i s mul iplie is equal o 0, i.e. α=n
n−1.
2. I α−1−4n+5n2
2(n−1)(2n−1)=n
2(2n−1), hen z=∞is an indi e en ixed poin .
3. I α−1−4n+5n2
2(n−1)(2n−1)>n
2(2n−1), hen z=∞is a epelling ixed poin .
We inish his sec ion s udying he s abili y o he o he s ange ixed poin s, which a e
gi en by he solu ions o (3).
P oposi ion 2.4. The ns ange ixed poin s a e gi en by
z∗
ξ=ξ1−2α−n+ 2αn
1−2α−3n+ 2αn1/n
,(6)
whe e ξdeno es an n h- oo o he uni y, and sa is y he ollowing s a emen s.
6
1. I α−2n−1
n−1<1
2, hen hey a e a ac ing ixed poin s. In pa icula , hey a e supe a -
ac ing ixed poin s i α=2n−1
n−1.
2. I α−2n−1
n−1=1
2, hen hey a e indi e en ixed poin s.
3. I α−2n−1
n−1>1
2, hen hey a e epelling ixed poin s.
P oo . The de i a i e o he ope a o is gi en in Equa ion (4). Subs i u ing zn=1−2α−n+2αn
(1−2α−3n+2αn)
in (4), we ha e ha
O0
n,α(z∗
ξ) = −2−2α(n−1) + 4n
n−1.
W i ing α=a+ib and de eloping he equa ion O0
n,α(z∗
ξ)= 1 we ob ain
a−2n−1
n−12
+b2=1
4,
which is he equa ion o he ci cle cen ed a 2n−1
n−1wi h adius 1/2. The s ange ixed poin s a e
a ac ing inside his ci cle, since O0
n,α(z∗
ξ)<1.They a e supe a ac ing i O0
n,α(z∗
ξ) = 0, which
is sa is ied o α=2n−1
n−1. Ou side his ci cle hey sa is y O0
n,α(z∗)>1, so hey a e epelling.
Finally, on he ci cle hey sa is y O0
n,α(z∗
ξ)= 1, so hey a e indi e en ixed poin s.
P oposi ion 2.3 explains he e olu ion o he loca ion and size o he head o he Ca se .
In pa icula , i shows ha he adius o he co esponding disk dec eases o 1
4as n→ ∞. On
he o he hand, in P oposi ion 2.4 we show ha he body o he Ca se is a disk o adius 1
2,
which does no depend on n.
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
(a) n= 3
0 0.5 1 1.5 2
−1
−0.5
0
0.5
1
(b) n= 10
Figu e 2: Dynamical planes o On,α o α=2n−1
n−1.
In Figu e 2 we show he dynamical planes o On,α o n∈ {3,10}and α=2n−1
n−1, which
co espond o he supe a ac ing case in P oposi ion 2.4. We obse e wi h a scaling om ed
7
o g een o pu ple and o g ey he poin s which con e ge o he he oo s in 75 i e a es, ha
is, he i e a e wsa is ies |w−ξ|<10−4 o some n h- oo o he uni y ξ. We obse e in black
he poin s which ha e no con e ged o he oo s a e 75 i e a es. The basins o a ac ion o
he s ange ixed poin s appea as black inge s coming om in ini y. No ice ha he s ange
ixed poin s, and he co esponding basins o a ac ion, a e symme ically loca ed. We e e
o Sec ion 5.2 o a mo e de ailed explana ion on how he images a e p oduced.
3 Simple connec i i y o Fa ou componen s
Nume ical expe imen s show ha he speed o con e gence o he oo s may be slowe when
he basins o a ac ion a e no simply connec ed, ha is, he e a e connec ed componen s o
he basins o a ac ion which ha e holes. In his sec ion we gi e a dynamical condi ion which
cha ac e izes whe he he immedia e basins o a ac ion o he ope a o s On,α a e simply
connec ed. We also p o e ha he ope a o s On,α can no ha e He man ings, ha is, doubly
connec ed o a ion domains. The exis ence o such domains would lead o a posi i e measu e
se o ini ial condi ions no con e ging o he oo s. Fo comple eness, we end his sec ion
showing how he p e ious esul s lead o a dynamical cha ac e iza ion o he connec i i y o
he co esponding Julia se s.
3.1 Simple connec i i y o immedia e basins o a ac ion o supe -
a ac ing ixed poin s
In his subsec ion we gi e a cha ac e iza ion o he simple connec i i y o he immedia e basin
o a ac ion o a suppe a ac ing ixed poin in he case ha i can con ain a mos one c i ical
poin besides he suppe a ac ing ixed poin i sel . This esul is s a ed o a gene al a ional
map . Fo simplici y, we assume ha i has a suppe a ac ing ixed poin a z= 0. We
deno e by A (0) i s basin o a ac ion and by A∗
(0) i s immedia e basin o a ac ion, ha is,
he connec ed componen o A (0) which con ains z= 0.
P oposi ion 3.1. Le :b
C→b
Cbe a a ional map and le z= 0 be a supe a ac ing ixed
poin o . Assume ha A (0) con ains a mos one c i ical poin o he han z= 0. Then,
exac ly one o he ollowing s a emen s holds.
1. The se A∗
(0) con ains no c i ical poin o he han z= 0. Then, A∗
(0) is simply con-
nec ed.
2. The se A∗
(0) con ains a c i ical poin c6= 0 and a p eimage z0o z= 0,z06= 0. Then,
A∗
(0) is simply connec ed.
3. The se A∗
(0) con ains a c i ical poin c6= 0 and no p eimage o z= 0 o he han z= 0
i sel . Then, A∗
(0) is mul iply connec ed.
The p oo o P oposi ion 3.1 uses he Riemann-Hu wi z o mula (see [4]). Gi en a p ope
map :U→V, his o mula ela es he connec i i ies o Uand V. The connec i i y o an
open se U⊂b
Cis gi en by he numbe o connec ed componen s o ∂U. A map is p ope i
he p eimages o compac se s a e compac se s. Gi en a a ional map , an open se Vand a
connec ed componen Uo −1(V), he map :U→Vis p ope .
8
Theo em 3.2 (Riemann-Hu wi z Fo mula).Le Uand Vbe wo connec ed domains o b
Co
ini e connec i i y mUand mVand le :U→Vbe a deg ee kp ope map b anched o e c
c i ical poin s coun ed wi h mul iplici y. Then
mU−2 = k(mV−2) + c.
P oo o P oposi ion 3.1. We conside he B¨o che coo dina e o he supe a ac ing ixed
poin z= 0 (see [17, Theo em 9.1]). The B¨o che coo dina e is a con o mal map φ:U→D ,
whe e Uis a simply connec ed neighbou hood o z= 0, 0 < ≤1, and D is he open disk
o cen e 0 and adius . This map conjuga es wi h z→zd, whe e dis he local deg ee
o he supea a ac ing ixed poin z= 0. Mo e speci ically, o all zwi h |z|< we ha e
φ◦ ◦φ−(z) = zn. In pa icula , φ(0) = 0.
We now assume ha Uis he maximal domain o de ini ion o he B¨o che coo dina e
and, hence, is maximal. I A∗
(0) does no con ain any ex a c i ical poin , hen = 1 and
U=A∗
(0) (see [17, Theo em 9.3]). In pa icula , A∗
(0) is simply connec ed. On he o he
hand, i A∗
(0) con ains an ex a c i ical poin c6= 0, hen < 1, U⊆A∗
(0), and c∈∂U (see
[17, Theo em 9.3]). Le V= (U). Then, γ0:= ∂V is a simple closed cu e (i is he p eimage
unde φo he ci cle o adius d). Deno e by γ1 he connec ed componen o −1(γ0) which
con ains ∂U. Then, γ1is he union o wo simple closed cu es which in e sec a he c i ical
poin c, say γ1
1, and γ2
1. Fu he mo e, ei he ∂U =γ1
1o ∂U =γ1
1∪γ2
1=γ1(see Figu e 3).
0γ0
γ2
1
γ1
1
z0
γ2
1
γ1
1
0γ0
U
VV
U
Figu e 3: The wo possibili ies o he con igu a ion o γ1=γ1
1∪γ2
1.
Assume ha ∂U =γ1
1and deno e by U0 he connec ed componen o b
C γ2
1which does no
con ain z= 0. Since (γ1
1) = (γ2
1) = γ0, i ollows ha (U0) = (U) = V. In pa icula U0
con ains a p eimage z0o 0. Mo eo e , U∪U0⊂A∗
(0) and we can ake a simply connec ed
neighbou hood U ⊂ A∗
(0) o U∪U0. Deno e by W he connec ed componen o −1(U) ha
con ains 0. Since does no ha e any c i ical poin in A∗
(0) o he han cand 0, i ollows
om he Riemann-Hu wi z o mula (Theo em 3.2) ha |W:W → U has deg ee d+ 1 and W
is simply connec ed. No ice ha he deg ee o |Wis a leas d+ 1 since z= 0 is a p eimage
o mul iplici y do i sel and Wcon ains an ex a p eimage z0o 0. Repea ing he p ocess,
we ob ain a sequence o simply connec ed domains U=U0⊂U1⊂ ··· ⊂ Un⊂ ···, whe e
Unis a connec ed componen o −1(Un−1) and Un−1⊂Un. Mo eo e , A∗
(0) = S∞
n=0 Un, om
which we conclude ha A∗
(0) is a simply connec ed se con aining a p eimage z0o z= 0 and
a c i ical poin c6= 0.
Now assume ha ∂U =γ1
1∪γ2
1=γ1. Then, he connec ed componen s W1and W2o b
C γ1
1
and b
C γ2
1, espec i ely, which do no con ain z= 0 a e mapped unde on o open se s which
con ain b
C V. In pa icula , bo h W1and W2con ain pa o he Julia se J( ) and, he e o e,
9
(1 −n)α(1 −2n+ 2α(n−1))(n−1)
n+α(1 −n)−(2n−1 + α(n−1) (n−2))2
=
=nq(α(n−1) −1)2+ 2 (n−1)2
,
which implies
(1 + 2α(n−1) −2n) (n−1)2n2= 0.
The oo o he las equa ion is
α=2n−1
2n−2.
Howe e , his is no a alid solu ion since we a e conside ing he case α6=2n−1
2n−2.
Finally, we conside he alue α=n
n−1disca ded in Equa ion (8). In his case, he deg ee
o he ope a o emains 2n:
On, n
n−1(z) = (n+ 1) + 2(n2−1)zn−(n−1) z2n
2n2zn−1.
Then, he only alues o which he ope a o degene a es o a a ional unc ion o lowe
deg ee a e α=1
2and α=2n−1
2n−2.
In he ollowing we analyse he dynamical beha iou o he nume ical me hods co esponding
o he bi u ca ion pa ame e s ob ained abo e.
Lemma 4.2. Fo α=1
2 he s ange ixed poin s a e z= 0 and z=∞. Mo eo e , he only
c i ical poin s a e he oo s o he polynomial, which ha e mul iplici y 2
P oo . The a ional unc ion o α=1
2is gi en by
On, 1
2(z) = (n+ 1)z+ (n−1)zn+1
(n−1) + (n+ 1)zn.(9)
The ixed poin s o On, 1
2(z) a e he oo s, z= 0 ( he ns ange ixed poin s collide a 0) and
z=∞. The de i a i e is gi en by
O0
n, 1
2
(z) = (n2−1) (zn−1)2
((n−1) + (n+ 1)zn)2.(10)
The only c i ical poin s, which a e he ze os o O0
n, 1
2
(z), a e he oo s o he polynomial zn−1
and ha e mul iplici y 2.
Lemma 4.3. Fo α=2n−1
2n−2, he oo s o he polynomial a e c i ical poin s o mul iplici y 2.
Mo eo e , he poin s {0,∞} o m a cycle o pe iod 2, which is supe a ac ing i n > 2.
P oo . The ope a o o α=2n−1
2n−2is
On, 2n−1
2n−2(z) = 1 + (2n−1)zn
(2n−1 + zn)zn−1.(11)
16
The ixed poin s a e he oo s o he polynomial and he n h- oo s o −1. The de i a i e is
gi en by
O0
n, 2n−1
2n−2
(z) = −(2n2−3n+ 1) (zn−1)2
zn(2n−1 + zn)2.(12)
The oo s o he polynomial a e c i ical poin s wi h mul iplici y 2. Since he nume a o o
On, 2n−1
2n−2(z) does no anish a z= 0 and z= 0 is a oo o mul iplici y n−1 o he denomina o ,
we conclude ha z= 0 is mapped o z=∞wi h deg ee n−1. The e o e, z= 0 is a c i ical
poin i n−1>1, ha is, n > 2. Simila ly, z=∞is mapped o z= 0 wi h deg ee n−1 and
is a c i ical poin i n > 2. In pa icula {0,∞} is a pe iod wo cycle. Since a pe iodic cycle
is supe a ac ing when any o he poin s o he cycle is c i ical, we conclude ha {0,∞} is a
supe a ac ing cycle i n > 2.
No ice ha he poin z=∞is a ixed poin o all α6=2n−1
2n−2bu i becomes a pe iod 2
supe a ac ing pe iodic poin i n > 2 and α=2n−1
2n−2(see Lemma 4.3). This may lead o e y
uns able dynamics (compa e Figu e 7).
We now look o bi u ca ion pa ame e s o which he ee c i ical poin s coincide wi h z= 0
o a e mapped on o i . No ice ha z= 0 is a p eimage o he ixed poin z=∞. In ha case,
he ee c i ical poin s a e p epe iodic and, he e o e, he e is no o he s able beha iou han
he basins o a ac ion o he oo s.
Lemma 4.4. The ee c i ical poin s collapse wi h z= 0 i and only i α= 0.
P oo . In he exp ession o he c i ical poin s (Equa ion (5)) we obse e ha c i ical poin s
become equal o 0 o α= 0 and α=1
2. Ne e heless, α=1
2is a degene a e case; in his case,
as we ha e seen in Lemma 4.2, z= 0 is no a c i ical poin . So, he only alue o which ee
c i ical poin s collapse wi h z= 0 is α= 0.
Now, we calcula e he alues o he pa ame e such ha c i ical poin s a e p eimages o
z= 0. As be o e, o such pa ame e s he c i ical poin s a e p epe iodic since z= 0 maps o
he ixed poin z=∞.
Lemma 4.5. C i ical poin s a e p eimages o z= 0 i and only i α=1±√2(1−n)
n−1.
P oo . Fo α6={n
n−1,2n−1
2n−2}, he ee c i ical poin s a e gi en by
cn,α,ξ =ξ (n−1)2(−1+2α)α
n(2n−1) −(4n−1) (n−1) α+ 2 (n−1)2α2!1/n
,(13)
whe e ξdeno es an n h- oo o he uni y. We disca d he pa ame e α=n
n−1since he c i ical
poin s coincide wi h z=∞(see P oposi ion 2.3). We also disca d he pa ame e 2n−1
2n−2since he
deg ee o On,α dec eases and he e a e no ee c i ical poin s (see Lemma 4.1 and Lemma 4.3).
P eimages o z= 0 a e ob ained om On,α (z) = 0, ha is
(1 −2α)(n−1) + (2 −4α−4n+ 6αn −2αn2)zn+ (n−1)(1 −2α−2n+ 2αn)z2n= 0.
The solu ions sa is y
zn=−1+2n+α(n−1) (n−2) ±nq2 (n−1) + (1 + α(1 −n))2
(n−1) (1 −2n+ 2α(n−1)) .
17
The solu ions coincide wi h c i ical poin s i
−1+2n+α(n−1) (n−2) ±nq2 (n−1) + (1 + α(1 −n))2
(n−1) (1 −2n+ 2α(n−1)) =
=(n−1)2(2α−1) α
n(2n−1) −(4n−1) (n−1) α+ 2 (n−1)2α2.
Ope a ing and simpli ying we ob ain
(2α−1) (1 + 2α(n−1) −2n)2(n−1)2α2(n−1)2−2α(n−1) + 2n−1= 0.
The solu ions o he las equa ion a e
α=1
2, α =2n−1
2 (n−1), α =1±p2 (1 −n)
n−1.
The pa ame e s α=1
2and α=2n−1
2n−2a e degene a ed cases wi h no ee c i ical poin s (see
Lemma 4.1).
We conclude ha c i ical poin s a e he p eimages o z= 0 i and only i α=1±√2(1−n)
n−1.
No ice ha he alues o αin oduced in Lemma 4.5 coincide, in he pa ame e space, wi h
wo symme ic poin s o he Colla o he Ca se , whe e he main ami ica ions appea (see
Figu e 6).
5 Nume ical s udies and conclusions
The goal o his sec ion is o pe o m a nume ical s udy on he Chebyshe -Halley amily.
We i s analyse he pa ame e planes nea he bi u ca ion pa ame e s desc ibed in Sec ion 4.
A e wa ds we s udy he dynamical planes o he amily. We ocus on he dynamics o maps
wi h a disconnec ed Julia se and he e olu ion when ng ows o he dynamics o some ele an
membe s o he Chebyshe -Halley amily.
5.1 Pa ame e planes
The d awings o pa ame e planes o he ope a o s On,α along he pape a e done using a
p og am w i en in C which wo ks as ollows. We ake a g id o poin s (1500 ×1000 poin s
o Figu e 1, 1200 ×2000 poin s o Figu e 6, and 1500 ×1500 poin s o Figu e 7). Then,
we associa e a pa ame e α∈C o each poin o he g id. The ange o he eal and he
imagina y pa o he pa ame e s αis indica ed in he ho izon al and e ical axes o he
images, espec i ely. Fo ixed αwe compu e one o he n ee c i ical poin s and i e a e i up
o 150 imes. A each i e a ion we e i y i he i e a ed poin whas con e ged o any o he
n h- oo s o he uni y (we e i y i |w−ξ|<10−4, o any n h- oo o he uni y ξ). I he
c i ical o bi con e ges o a oo o he uni y, we colou he co esponding poin wi h an scaling
om ed ( as con e gence) o yellow, g een, blue, pu ple and o g ey (slow con e gence). I
a e 150 i e a ions he o bi o he c i ical poin has no con e ged o a oo , we plo he poin
in black. In Figu e 1 we show he pa ame e planes o he ope a o On,α o di e en alues o
n.
18
−0.5 0 0.5
−1
−0.5
0
0.5
1
(a) n= 10
−0.5 0 0.5
−1
−0.5
0
0.5
1
(b) n= 25
−0.5 0 0.5
−1
−0.5
0
0.5
1
(c) n= 100
Figu e 6: Zooms in on he pa ame e planes nea he main bi u ca ion poin s o he Colla .
The aim o his subsec ion is o analyse he pa ame e planes o he amily nea he bi-
u ca ion pa ame e s ound in Sec ion 4. In Figu e 6 we show zooms in on he pa ame e
planes o se e al n. The anges o αinclude he bi u ca ion pa ame e s α= 0 (Lemma 4.4),
α= 1/2 (Lemma 4.1) and α±=1±√2(1−n)
n−1(Lemma 4.5). The pa ame e α= 1/2 (Halley’s
me hod) appea s in he igu es as a ip o an an enna which joins i wi h α= 0 (Chebyshe ’s
me hod). We obse e ha he bi u ca ion s uc u e a α= 1/2 is a he simple, which could
indica e s abili y o he amily o pa ame e s nea Halley’s alue. On he o he hand, a mo e
complex bi u ca ion s uc u e appea s a Chebyshe ’s pa ame e (α= 0). Recall ha o his
pa ame e all c i ical poin s collide a z= 0, which is a p eimage o he ixed poin z=∞.
Indeed, o n= 25 we obse e a black disk o pa ame e s o which he o bi o he c i ical
poin has no con e ged o a oo a e 150 i e a es. This black disk does no co espond o
any s able beha iou , bu o pa ame e s o which he c i ical o bi would equi e mo e i e a es
o con e ge o a oo . Indeed, his disk o pa ame e s dec eases when we inc ease he numbe
o i e a ions. A simila si ua ion occu s nea he pa ame e s α±, which a e ma ked wi h whi e
poin s in Figu e 6. These pa ame e s co espond o ope a o s o which he c i ical poin s a e
p eimages o z= 0. Se e al o hese black egions no co esponding o s able beha iou appea
when d awing he pa ame e planes o big n, being he bigge ones a ound α±. Nume ical
expe imen s seem o indica e ha he bigges black egions no ela ed o s able beha iou ap-
pea a ound pa ame e s o which he ee c i ical poin s a e e en ually mapped unde i e a ion
o he co esponding ope a o on o z= 0. In Sec ion 5.2 we show he dynamics o On,α o all
he bi u ca ion pa ame e s discussed in his pa ag aph in o de o analyse i hey ha e o be
a oided o i hey p esen good dynamical beha iou .
In Figu e 7 we show zooms in on he pa ame e plane nea he pa ame e α=2n−1
2n−2. I
n > 2, his bi u ca ion pa ame e co esponds o a lowe deg ee ope a o o which {0,∞} is
a supe a ac ing cycle. The e o e, his is a pa ame e o be a oided. In Figu e 7 we indica e
i wi h a small whi e poin . We obse e how, o n= 10 and n= 25, his pa ame e is he
19
1 1.02 1.04 1.06 1.08 1.1
−4
−2
0
2
4
·10−2
(a) n= 10
0.96 0.98 1 1.02 1.04
−4
−2
0
2
4
·10−2
(b) n= 25
Figu e 7: Zooms in on he pa ame e planes nea he degene acy pa ame e α=2n−1
2n−2.
cen e o a cascade o bi u ca ions. Simila ly o wha happens in he pa ame e plane nea he
o he bi u ca ion pa ame e s, his cascade o bi u ca ions may lead o a black egion which does
no co espond o s able beha iou (see Figu e 7 (b)). This cascade o bi u ca ions akes place
in a a he small egion which is adjacen o he hype bolic componen o pa ame e s which
con ains α=2n−1
3n−3.Fo ixed n, his pa ame e co esponds o he single ope a o On,α wi h
o de o con e gence 4 o he oo s (see P oposi ion 2.1). Mo eo e , he pa ame e s wi hin his
hype bolic componen ha e connec ed Julia se (see Co olla y 3.11), which a p io i makes o
hem desi able pa ame e s. Howe e , he hype bolic componen seems o become smalle and
o mo e sligh ly o he le when we inc ease n(see Figu e 1). In pa icula , he pa ame e
α= 1, which has o de o con e gence 4 o he oo s o n= 2, alls in o he cascade o
bi u ca ions o n= 25 (see Figu e 7). In Sec ion 5.2 we show how he dynamics o α= 1
e ol es when we inc ease nand i app oaches and en e s he cascade o bi u ca ions.
5.2 Dynamical planes
In his sec ion we p o ide nume ical d awings o dynamical planes o he ope a o s On,α. The
d awings o dynamical planes a e done using a p og am w i en in C which wo ks as ollows.
We ake a g id o 1500 ×1500 and we associa e a poin z∈C o each poin o he g id. The
ange o he eal and he imagina y pa o he poin s zis indica ed in he ho izon al and
e ical axes o he images, espec i ely. Then, we i e a e he poin zup o 75 imes. A each
i e a ion we e i y i he i e a ed poin whas con e ged o any o he n h- oo s o he uni y
(we e i y i |w−ξ|<10−4, o any n h- oo o he uni y ξ). I i con e ges o any oo o he
uni y, we colou he co esponding poin in he g id wi h an scaling om ed ( as con e gence)
o yellow, g een, blue, pu ple and o g ey (slow con e gence). I a e 75 i e a ions he o bi
has no con e ged o a oo , we plo he poin in black.
The goal o his sec ion is o analyse he dynamical beha iou o he ope a o s On,α o se -
e al pa ame e s in o de o ha e a be e idea o which pa ame e s p o ide a good dynamical
beha iou . We i s in es iga e nume ically o which pa ame e s he Julia se is disconnec ed
20
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
(a) n= 3, α = 0.2+1.4i
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
(b) n= 3, α = 2i
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
(c) n= 25, α = 0.2+1.4i
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
(d) n= 25, α = 2i
Figu e 8: Dynamical planes o se e al On,α(z) wi h a disconnec ed Julia se .
and we analyse hei dynamics. Be o e doing so, we ecall he concep o hype bolic componen
o pa ame e s. A hype bolic componen is a connec ed se o pa ame e s o which all c i ical
o bi s accumula e on a ac ing o supe a ac ing cycles. The Julia se is s able wi hin hype -
bolic componen s, i.e. he Julia se is connec ed o one pa ame e o he componen i and only
i i is connec ed o all pa ame e s o he componen . Hype bolic componen s o pa ame e s
o which he ee c i ical poin s belong o he basins o a ac ion o he oo s appea in ed in
Figu e 1. In Co olla y 3.11 we ha e p o en ha he ope a o s co esponding o he hype bolic
componen which con ains α=2n−1
3n−3ha e connec ed Julia se . Nume ical simula ions seem
o indica e ha he o he ed hype bolic componen s bounded by he Colla o he Ca se
co espond o pa ame e s o which he ee c i ical poin s belong o he basins o a ac ion o
he oo s bu no o hei immedia e basins o a ac ion and, hence, co espond o ope a o s
wi h a connec ed Julia se (see Theo em 3.9). The only hype bolic componen co esponding
o ope a o s wi h disconnec ed Julia se seems o be he ed unbounded componen on he
complemen o he Colla (see Figu e 1). In Figu e 8, Figu e 9 ( ), and Figu e 11 ( ) we show
21
he dynamical planes o se e al pa ame e s wi hin his hype bolic componen o n= 3 and
n= 25. We ake he pa ame e α= 0.2 + 1.4i, which is close o he Colla , and he pa ame e s
α= 2iand α= 4i, which a e a he away om he Colla . Fo α= 0.2 + 1.4i he holes in
he basins o a ac ion a e ela i ely big and a e easy o obse e. Fo he pa ame e s α= 2i
and α= 4i hese holes a e much mo e di icul o see since hey become smalle . Howe e , he
con e gence o he oo s wi hin his las pa ame e s is slowe . Indeed, o n= 3 we obse e
how he basins o a ac ions o he oo s ha e a mo e in ense ed onali y o α= 0.2 + 1.4i,
which indica es as con e gence. Fo n= 25 we can obse e how he basins o a ac ion o
he oo s acqui e a blueish onali y o α= 2iand, pa icula ly, o α= 4i, which indica es
slowe con e gence o he oo s. We may conclude ha , wi hin his hype bolic componen ,
pa ame e s close o he Colla ha e bigge holes on he basins o a ac ion o he oo s, bu
he con e gence o he oo s is as e .
In Figu e 9, Figu e 10 and Figu e 11 we show he dynamical planes o he ope a o On,α o
se e al o he mos ele an pa ame e s s udied along he pape o n= 3, n= 10 and n= 25,
espec i ely. The pa ame e s analysed a e he ollowing. Sub igu e (a) co esponds o α=2n−1
3n−3,
which p o ides he me hod wi h o de o con e gence 4 o he oo s (see P oposi ion 2.1).
Sub igu es (b), (c), and (d) co espond o α= 1/2 (Halley’s me hod), α= 0 (Chebyshe ’s
me hod) and α= 1 (Supe -Halley me hod), espec i ely. Sub igu e (e) co esponds o α=
1+√2(1−n)
n−1, which is a bi u ca ion pa ame e o which he ee c i ical poin s a e mapped o
z= 0 (see Lemma 4.5). Finally, sub igu e ( ) co esponds o α= 4iwhich is a pa ame e in he
unbounded componen o he complemen o he Colla o which he Julia se is disconnec ed.
A p io i, he pa ame e ha should p o ide a be e dynamical beha iou is α=2n−1
3n−3
(sub igu e (a)), since he co esponding ope a o s ha e o de o con e gence 4 o he oo s.
We obse e ha , nea he oo s, i is he pa ame e wi h a as es a e o con e gence (i has
he mos in ense ed nea he oo s). Howe e , he global dynamics a e complex. Indeed, o
n= 10 and n= 25 we obse e how he Julia se becomes complica ed and he e appea black
disks o ini ial condi ions which do no con e ge o he oo s a e 75 i e a es. We wan o
poin ou ha i we inc ease he numbe o i e a es we can make disappea his disk o n= 10.
Ne e heless, o n= 25 he disk dec eases e y slowly when we inc ease he numbe o i e a es.
This is a kind o pa hological beha iou which is no ela ed o s able beha iou di e en om
he oo s and can lead o big se s o ini ial condi ions which do no con e ge o he oo s.
The pa ame e s α= 0 (Chebyshe ’s me hod, sub igu e (c)), α=1+√2(1−n)
n−1(sub igu e (e))
and α= 4i(sub igu e ( )) p esen a simila dynamical beha iou . F om hem, he one which
p esen s a la ge black disk a ound z= 0 is α= 0. This indica es ha Chebyshe ’s me hod
may no be an ou s anding oo inding algo i hm o apply o he amily zn+cwhen nis
big. On he o he hand, he pa ame e α=1+√2(1−n)
n−1, which is a bi u ca ion pa ame e in he
Colla , p esen s a be e dynamical beha iou han he pa ame e α= 4i, which is a pa ame e
co esponding o an ope a o wi h a disconnec ed Julia se . Indeed, he speed o con e gence
o he oo s o he ope a o s co esponding o Chebyshe ’s me hod and α=1+√2(1−n)
n−1is much
as e han he one o α= 4i. We can conclude ha he black egions which appea a ound
bi u ca ion pa ame e s in he pa ame e plane (see Figu e 1 and Figu e 6) a e no necessa ily
ela ed o pa icula ly bad dynamical beha iou and hey a e p e e able han he pa ame e s
co esponding o disconnec ed Julia se .
The pa ame e α= 1 is pa icula ly in e es ing. Fo n= 2 he co esponding ope a o
has o de o con e gence 4 o he oo s. I s ill p esen s e y good dynamical beha iou o
22
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
(a) α= 5/6
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
(b) α= 0.5, Halley’s me hod
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
(c) α= 0, Chebyshe ’s me hod
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
(d) α= 1, Supe -Halley me hod
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
(e) α= (1 + i√4)/2
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
( ) α= 4i
Figu e 9: Dynamical planes o On,α(z) o n= 3 and di e en alues o α.
23
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
(a) α= 19/27
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
(b) α= 0.5, Halley’s me hod
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
(c) α= 0, Chebyshe ’s me hod
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
(d) α= 1, Supe -Halley me hod
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
(e) α= (1 + i√18)/9
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
( ) α= 4i
Figu e 10: Dynamical planes o On,α(z) o n= 10 and di e en alues o α.
24
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
(a) α= 49/72
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
(b) α= 0.5, Halley’s me hod
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
(c) α= 0, Chebyshe ’s me hod
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
(d) α= 1, Supe -Halley me hod
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
(e) α= (1 + i√48)/24
−3−2−1 0 1 2 3
−3
−2
−1
0
1
2
3
( ) α= 4i
Figu e 11: Dynamical planes o On,α(z) o n= 25 and di e en alues o α.
25