• Non ci sono risultati.

Zeros of regular functions of quaternionic and octonionic variable: a division lemma and the camshaft effect

N/A
N/A
Protected

Academic year: 2021

Condividi "Zeros of regular functions of quaternionic and octonionic variable: a division lemma and the camshaft effect"

Copied!
16
0
0

Testo completo

(1)

arXiv:0904.2667v2 [math.CV] 5 Nov 2009

o tonioni variable: a division lemma and the

amshaft ee t R. Ghiloni,A. Perotti

Department of Mathemati s University of Trento ViaSommarive,14 I38123 PovoTrento ITALY

perottis ien e.unitn.it November5, 2009

Abstra t

Westudyindetailthezerosetofasli eregularfun tionofa quater-nioni oro tonioni variable. Bymeansofadivisionlemmafor onvergent powerseries,wendtheexa trelationexisting betweenthezerosoftwo o tonioni regular fun tions and those of their produ t. In the ase of o tonioni polynomials,wegetastrongformofthefundamentaltheorem ofalgebra. Weprovethatthesumofthemultipli itiesofzerosequalsthe degreeofthepolynomialandobtainafa torizationinlinearpolynomials. Keywords: Quaternions, O tonions, Fun tions of ahyper omplex variable, Fundamentaltheoremofalgebra

Math.Subj.Class: 30C15,30G35,32A30

1 Introdu tion

Let

H

denotetheskeweldofquaternionsover

R

. Re ently,GentiliandStruppa [13,15℄introdu edthenotionofsli eregularity (orCullen regularity)for fun -tionsofaquaternioni variable. Thisnotiondonot oin idewith the lassi al oneof Fueter regularity(we referthereader to [30℄ and[19℄for thetheory of Fueter regular fun tions). In fa t, the set of sli e regular fun tions on a ball

B

R

enteredin the originof

H

oin ides withthat of allpowerseries

P

i

w

i

a

i

that onverges in

B

R

. In parti ular, all the standardpolynomials

P

n

i=0

w

i

a

i

withrightquaternioni oe ients,whi h failtobeFueter regular,dene sli e

WorkpartiallysupportedbyMIUR(PRINProje tProprietàgeometri hedellevarietà realie omplesse")andGNSAGAofINdAM

(2)

regular fun tions on the whole spa e

H

. The lass of sli e regular fun tions ontainsalsothe radiallyholomorphi fun tions introdu ed byFueter ( f. [19, Ÿ11℄). Su h fun tions x every omplex plane generated by the reals and by aunit imaginaryquaternion, while the sli e regular fun tions do not need to fullthis property. A tually,radiallyholomorphi fun tions orrespondto the veryspe ial lassof those sli eregularfun tions, whi h anbeexpanded into powerserieswith real oe ients. Thenotionofsli eregularitywasextended tofun tionsofano tonioni variablein[14,12℄andtofun tionswithvaluesin aCliordalgebrain[17, 4℄.

Inthispaper,westudyin detailthezerosetofasli eregularfun tion. For simpli ity, in thesequel,weuse theterminologyregular fun tion in pla e of sli eregularfun tion. Weobtainby dierent te hniques some properties al-readyknown( f.[24,14,15,11,16,18℄)andweextendotherstotheo tonioni ase. We nd the exa trelation existing between the zerosof twoo tonioni regular fun tions and those of their produ t. In the ase of o tonioni poly-nomials,weobtainastrongformof thefundamental theoremof algebra. The main tool we use is a division lemma, whi h generalizes to o tonioni power seriesaresultprovedbyBe k[1℄ forquaternioni polynomialsand by Serdio [27℄foro tonioni polynomials. Inthesimpler aseofregularpolynomialsand onvergentpowerserieswithreal(i.e. entral) oe ients,thepropertiesofthe zeroset havebeenstudied alsobyotherauthors( f. [5℄).

We des ribe in more details the stru ture of the paper. In Se tion 2, we re allsomebasi denitions and provethedivision lemma (Lemma 1). Given a regular fun tion

f

and an o tonion

α

, we an asso iate to

f

a remainder, whi h is a linear or onstant regular polynomial. This remainder des ribes ompletelythe interse tion ofthe zeroset

V (f )

of

f

withthe onjuga y lass of

α

. We an then obtaineasily a stru ture theorem for

V (f )

(Theorem 6). This result was proved for quaternioni polynomials by Pogorui and Shapiro [24℄, for quaternioni regular fun tions byGentili and Stoppato in [11℄ and it wasextendedto o tonioni regularfun tions in [18℄. Thedivision lemma and the on eptofnormal series asso iatedtoaregularfun tion (seeSe tion2for pre isedenitions)suggestthedenitionofthemultipli ityofazeroofaregular fun tion. Ourdenitionisequivalentonquaternioni polynomialswiththeone givenin[2℄andin[16℄.

InSe tion3,wedes ribethezerosoftheprodu t

f ∗g

oftworegularfun tions intermsofthezerosof

f

and

g

. Weprove(Lemma7)thatthenormalseriesof aregularfun tionismultipli ative,aresultthatisnontrivialintheo tonioni ase. An immediate onsequen eis that the onjuga y lassesof the zerosof

f ∗ g

areexa tlythose ontainingzerosof

f

or

g

. Thepre iserelationbetween

su hzerosisstatedinTheorem9. We allthisrelationthe amshaftee t. In theasso iative ase,itredu estoknownresults( f.[11℄forquaternioni regular fun tionsand[22,29℄forpolynomials).

Se tion4 ontainssomeappli ationsofthepre edingresultstoregular poly-nomials. The fundamental theorem of algebra was proved by Niven [23℄ for standardquaternioni polynomials. Itwasextended to otherpolynomialsover

(3)

theorem valid for a lass of real algebras in ludingthe omplex numbers, the quaternionsandtheo tonions. Seealso[31℄,[25℄and[18℄ forotherproofs.

In this ontext, our aim is a strong form of the fundamental theorem, in whi h aformulafor thesumof themultipli ities of zerosisa hieved. Gordon andMotzkin[20℄ proved,forpolynomialsona(asso iative)division ring,that thenumberof onjuga y lasses ontainingzerosof

f

annot be greaterthan thedegree

n

of

f

. This estimatewasimprovedonthequaternions by Pogorui and Shapiro [24℄: if

f

has

m

spheri al zeros and

l

nonspheri al zeros, then

2m + l ≤ n

. Gentili and Struppa [16℄ showed that, using the right denition

of multipli ity, the number of zeros of

f

equals the degreeof the polynomial. Wegeneralize this strong form to theo tonions (Theorem 11), givingaproof inspiredbythesimpleargumentusedin[24℄. Fromthisresultandthedivision lemma,wegetafa torizationlemmaforregularpolynomials(Lemma12)and somesu ient onditionsforthenitenessof

V (f )

.

Thedenitions andresultsarestatedovero tonions only, buttheyremain validoverquaternionsaswell. Sin e

H

anbeidentiedwitharealsubalgebra oftheo tonions,the orrespondingstatementsforquaternions anbeobtained dire tlyfromtheiro tonioni versionbyspe ializing oe ientsandvariables.

2 Division of regular series and multipli ity of

zeros

Let

O

bethenonasso iativedivision algebraofo tonions (also alled Cayley numbers). Wereferto[26,22,6℄forthemainpropertiesofthealgebras

H

and

O

. Were allthat

O

anbeobtainedfrom

H

bytheCayleyDi ksonpro ess. Any

x ∈ O

an be written as

x = x

1

+ x

2

k

, where

x

1

, x

2

∈ H

and

k

is a xed

imaginary unit of

O

. The addition on

O

is dened omponentwise and the produ tisdenedby

xy = (x

1

+ x

2

k)(y

1

+ y

2

k) = x

1

y

1

− ¯

y

2

x

2

+ (x

2

y

¯

1

+ y

2

x

1

)k.

Let

{1, i, j, ij}

denote arealbasis of

H

. Then abasis ofthe real algebra

O

is

formedby

{1, i, j, ij, k, ik, jk, (ij)k}

.

Let

B

R

betheopenballof

O

enteredin theoriginwith(possiblyinnite) radius

R

. Let

f : B

R

−→ O

and

g : B

R

−→ O

be regular fun tions with seriesexpansions

f (w) =

P

i

w

i

a

i

and

g(w) =

P

j

w

j

b

j

. Theusual produ tof polynomials,where

w

is onsideredtobea ommutingvariable( f.forexample [22℄ and [9, 10℄), an be extended to power series, and hen e to sli e regular fun tions( f. [11,16℄). Re allthat

f ∗ g

is theregular fun tiondened on

B

R

bysetting

(f ∗ g)(w) :=

P

k

w

k

P

i+j=k

a

i

b

j

.

We denote by

f

the regular fun tion dened on

B

R

by

f (w) :=

P

i

w

i

a

i

. If

(4)

(f ∗ g)(α) = f (α)g(α)

forevery

g

andevery

α ∈ O

andwewillwrite

f g

inpla e

of

f ∗ g

. Moreover,if

f

isreal then

(f ∗ g) ∗ h = (f g) ∗ h = f (g ∗ h)

for every

g, h

,and

f ∗ g

isequalto

f g

.

Wedenoteby

N (f )

thenormalseries (or symmetrizedseries)of

f

dened by

N (f ) := f ∗ f = f ∗ f.

Weremark that

N (f )

isregularandrealon

B

R

.

Let

α ∈ O

. Wedenotethetra e

tr(α) = α + α

of

α

alsoby

t

α

,thesquared

normof

α

by

n

α

anddenetherealpolynomial

α

, alled hara teristi poly-nomialof

α

, by

α

(w) := N (w − α) = (w − α) ∗ (w − α) = w

2

− w · t

α

+ n

α

.

It is wellknown ( f. [27℄) that

α

is onjugate to an o tonion

β

if and only if

α

= ∆

β

. Indi ate by

S

α

the onjuga y lass of

α

:

S

α

:= {β ∈ O | β =

Re(α) + |Im(α)| I, I ∈ S}

, where

S = {I ∈ O | I

2

= −1}

is the sphere of imaginaryunits. Notethat

S

α

redu estothepoint

{α}

when

α

isrealanditis asixdimensionalspherewhen

α

isnonreal.

Wedenote by

V (f )

thezero set of

f

. If

α ∈ V (f )

is real, we all

α

areal zero of

f

. If

α

is nonreal and

S

α

⊂ V (f )

, we all

α

a spheri al zero of

f

. Otherwise,we all

α ∈ V (f )

anisolatedzeroof

f

. Thisterminologyisjustied byTheorem 6below.

Letuspresentourdivisionlemma.

Lemma1. Let

f : B

R

−→ O

regularandlet

α ∈ B

R

. Thefollowingstatements hold.

(1)

There exist, and are unique, a regular fun tion

g : B

R

−→ O

and an o tonion

r ∈ O

su hthat

f (w) = (w − α) ∗ g(w) + r

.

(2)

Let

α ∈ O

be nonreal. There exist, and are unique, a regular fun tion

h : B

R

−→ O

ando tonions

a, b ∈ O

su hthat

f (w) = (∆

α

h)(w)+w a+b

.

Proof. (1)Letusprovetheexisten eof

g

andof

r

. First,observethat,ifsu h

g

and

r

exist,then

r

mustbeequalto

f (α)

. Infa t,it iseasyto see that the valueofthefun tion

(w − α) ∗ g(w)

in

α

iszero. Assume that

f =

P

i

w

i

a

i

has positive onvergen eradius

R

. Re allthat

R =



lim sup

n→+∞

n

p|a

n

|



−1

.

Let

α ∈ B

R

. If

α = 0

,thenwe anwrite

f = w



P

+∞

i=0

w

i

a

i+1



+ a

0

.

Suppose

α 6= 0

. A formal series omputation imposes to dene the oe ients of the

powerseries

g =

P

n

w

n

b

n

asfollows:

b

n

:= α

−1−n

f (α) −

n

X

j=0

α

j

a

j

= α

−1−n

+∞

X

j=n+1

α

j

a

j

(5)

forea h

n ∈ N

. Letusshowthattheseries

g

hasradiusof onvergen eatleast

R

. Sin e

|α| < R

,

lim sup

n→+∞

p|a

n

n

| < |α|

−1

. Thisimpliesthat,foreveryxed

ρ

with

|α| < ρ < R

,there existsaninteger

n

ρ

su hthattheinequality

|a

j

| ≤ ρ

−j

issatisedforevery

j > n

ρ

. Let

x ∈ B

R

andlet

ρ

be hoseninsu hawaythat

max{|α|, |x|} < ρ < R

. Then,forevery

n ≥ n

ρ

,itholds:

|x

n

b

n

| ≤ |x|

n

|b

n

| ≤ |x|

n

|α|

−1−n

+∞

X

j=n+1

|α|

j

|a

j

| ≤ |x|

n

|α|

−1−n

+∞

X

j=n+1

|α|

j

ρ

−j

= |x|

n

|α|

−1−n

 |α|

ρ



n+1

1

1 − |α|/ρ

=

 |x|

ρ



n

1

ρ − |α|

.

Therefore, the power series

P

n

w

n

b

n

onverges at every

x ∈ B

R

. It remains to prove the uniqueness of

g

and of

r

. We have just seen that

r

must be equal to

f (α)

so the o tonion

r

is uniquely dened by

f

and

α

. If

f (w) =

(w − α) ∗ g(w) + r = (w − α) ∗ g

(w) + r

,then

(w − α) ∗ (g(w) − g

(w)) = 0

and thiseasilyimpliesthat

g = g

.

(2)Let

α

benonreal. Usingtherstparttwi e,weget

s ∈ O

andaregular fun tion

h

on

B

R

su hthat

f (w) = (w − α) ∗ g(w) + r = (w − α) ∗ ((w − ¯

α) ∗ h(w) + s) + r

= (∆

α

h)(w) + ws − αs + r.

To proveuniqueness, assume that

f (w) = ∆

α

h + wa + b = ∆

α

h

+ wa

+ b

.

Then

f (α) = αa + b = αa

+ b

,

f (¯

α) = ¯

αa + b = ¯

αa

+ b

, from whi h weget

(α − ¯

α)(a − a

) = 0

andthen

a = a

,

b = b

. Finally,

α

h = ∆

α

h

gives

h = h

on

B

R

\ S

α

. Byadensityargument,

h

and

h

must oin ideeverywhere. Remark 1. As abyprodu t ofthe proofof thepre eding theorem, weobtain that,if

f

isaregularo tonioni polynomialofdegree

n > 0

,then

g

isaregular o tonioni polynomialofdegree

n − 1

.

Denition1. Lettheregularfun tions

g, h

andtheo tonions

r, a, b

beasabove. If

α

is real, we all

r

the remainderof

f

withrespe t to

α

andthe fun tion

g

the quotientof

f

w.r.t.

α

. If

α

isnonreal,we all remainderof

f

withrespe t to

α

theo tonioni polynomialdenedby

w a + b

and quotientof

f

w.r.t.

α

the fun tion

h

. Notethat theydependonly onthe onjuga y lass

S

α

of

α

. In any ase,wewill denote theremainder by

r

α

(f )

.

If

f

isreal,thenthe oe ientsofthequotientfun tionandoftheremainder arereal. Thisfa tisa onsequen eoftheuniquenessofquotientandremainder:

(w − α) g + r = f = ¯

f = (w − α) ¯

g + ¯

r

(

α

real)

or

α

h + w a + b = f = ¯

f = ∆

α

h + w ¯

¯

a + ¯b

(

α

nonreal)

imply

g = ¯

g

,

r = ¯

r

,

h = ¯

h

,

a = ¯

a

and

b = ¯b

. Asarstappli ation ofLemma1,weobtain:

(6)

Corollary2. Let

r

α

(f )

betheremainder of

f

withrespe tto

α

. Thefollowing statementshold.

(1)

Let

α

be real. Then

α

is a zero of

f

if and only if

r

α

(f ) = 0

; that is,

w − α | f

.

(2)

Let

α

benonreal andlet

r

α

(f )(w) = w a + b

. Then wehave:

(i) α

isaspheri al zeroof

f

if andonlyif

a = 0 = b

;thatis,

α

| f

.

(ii) α

isanisolatedzeroof

f

if andonlyif

a 6= 0

and

α = −ba

−1

. In parti ular, either

S

α

⊂ V (f )

or

S

α

∩ V (f )

is empty or

S

α

∩ V (f )

onsists ofasinglepoint. Thelattersituation o urs ifandonly if

a 6= 0

and

−ba

−1

∈ S

α

.

Proof. By using Lemma 1, we an follow the same lines of proof as in the polynomial ase( f. [27,Ÿ3℄).

If

α ∈ R

,thestatementistrivial,sin e

f (α) = r

α

(f )

.

Now assumethat

α

is nonreal. If

r

α

(f ) = 0

,i.e.

a = b = 0

, then

f (β) =

α

(β) · h(β) = 0

for every

β ∈ S

α

. If

a 6= 0

and

α = −ba

−1

, then

f (α) =

α

(α) · h(α) = 0

. If

S

α

ontainstwozeros

β

and

γ

of

f

,thentheequality

α

(β) · h(β) + β a + b = f (β) = 0 = f (γ) = ∆

α

(γ) · h(γ) + γ a + b

gives

β a = γ a

, sin e

α

vanishes on

β, γ ∈ S

α

. If

β 6= γ

, then it must be

a = b = 0

.

If

f

isreal,then

r

α

(f ) = w a + b

with

a, b

real. Therefore

−ba

−1

∈ R

and

f

hasnoisolated(nonreal)zeros. Thismeansthat thezerosof

f

are allrealor spheri aland

V (f )

istheunionofthespheres

S

α

with

α ∈ V (f )

( f.[5℄).

If

f

diersfrom areal

g

bya onstantnonrealo tonion,thenwehavethe opposite situation:

f

an have only isolatedzeros. This property was proved forregularpolynomialsover

H

in[22, Ÿ16.19℄.

Proposition 3. Let

f (w) = g(w) + a

0

be aregular fun tion on

B

R

. Assume that

g

isreal and

a

0

∈ O

isnonreal. Then, if

V (f )

isnotempty, itselements areisolatedzeros of

f

ontainedin the omplex planegeneratedby

1

and

a

0

. Proof. Let

α ∈ V (f )

. Sin e

a

0

= −g(α)

is nonreal,also

α

must benonreal. From thedivision lemma (Lemma 1) applied to

g

, weget:

g(w) = ∆

α

h(w) +

wa + b

, with real

a, b

. Then

r

α

(f ) = wa + b + a

0

and

α

annotbe spheri al.

Otherwise, it would be

a

0

= −b

. Therefore,

α = −(b + a

0

)a

−1

is anisolated zeroof

f

belongingtothe omplexplanegeneratedby

1

and

a

0

.

We still get the nonexisten e of spheri al zeros under a weaker ondition on

f

.

Proposition 4. Let

f (w) = g(w) + wa

1

+ a

0

be a regular fun tion on

B

R

. Assumethat

g

isreal andeither

a

0

or

a

1

isnonreal. Then

f

has no spheri al zeros.

(7)

Proof. We pro eed as before. Assume that

V (f ) 6= ∅

. Let

α ∈ V (f )

. Then

r

α

(f ) = w(a+a

1

)+b+a

0

,with

a, b

real. Then

α

annotbespheri al.Otherwise,

itwouldbe

a

0

= −b

and

a

1

= −a

.

If

α

is real and

f (w) = (w − α) g(w) + r

forsome regular fun tion

g

and

r ∈ O

,then

r

α

(f ) = r

,

r

α

( ¯

f ) = ¯

r

and

N (f ) = (w − α)

2

N (g) + (w − α)(g ∗ ¯

r + r ∗ ¯

g) + n

r

.

(1)

Itfollowsthat

r

α

(N (f )) = n

r

. If

α

isnonrealand

f

hasremainder

r

α

(f )(w) =

w a + b

,thenwehavethat

r

α

( ¯

f )(w) = w ¯

a + ¯b

andhen e

r

α

(N (f )) = r

α

(f ∗ ¯

f )(w) = w(a¯b + b¯

a + t

α

n

a

) + n

b

− n

α

n

a

.

Thelatterequalityfollowsimmediatelyfromthefollowingfa t

(wa + b) ∗ (w¯

a + ¯b) =

w

2

n

a

+ w(a¯b + b¯

a) + n

b

=

=

(∆

α

+ wt

α

− n

α

)n

a

+ w(a¯b + b¯

a) + n

b

.

Corollary5. Given

α ∈ O

,the followingstatements areequivalent:

(1) S

α

∩ V (f )

isnonempty.

(2) S

α

∩ V (N (f ))

isnonempty.

(3) ∆

α

| N (f )

.

Inparti ular, wehave that

V (N (f )) =

[

α∈V (f )

S

α

.

Proof. If

α

isreal and

r

α

(f ) = r

,then

r

α

(N (f )) = n

r

= 0

ifandonlyif

r = 0

. If this is the ase, Eq. (1) implies that

α

= (w − α)

2

divides

N (f )

. If

α

is nonreal,then

α

| N (f )

ifandonlyif

r

α

(N (f )) = 0

;thatis,

a¯b + b¯

a + t

α

n

a

= 0,

n

b

− n

α

n

a

= 0.

(2)

If

a = b = 0

, then

S

α

⊆ V (f )

and equations (2) are trivial. If

a 6= 0

and

β = −ba

−1

∈ S

α

∩ V (f )

,then

t

α

n

a

= t

β

n

a

= βa¯

a + a¯

a ¯

β = −b¯

a − a¯b

. Moreover,

n

α

n

a

= n

β

n

a

= n

b

andequations(2)aresatised.

Conversely,supposethatequations(2)aresatised. If

a = 0

,then

b = 0

and hen e

S

α

⊂ V (f )

. Let

a 6= 0

. Dene theo tonion

β := −ba

−1

. Pro eeding as above,weobtainthat

t

α

n

a

= t

β

n

a

and

n

α

n

a

= n

β

n

a

or,equivalently,

t

α

= t

β

and

n

α

= n

β

. Itfollowsthat

β ∈ S

α

∩ V (f )

. Sin e

N (f )

isreal,itszerosareall realorspheri al,so

V (N (f ))

isequalto

S

α∈V (f )

S

α

.

Fromthepre eding orollaries,weimmediatelygetanewproofofthe follow-ingresult. Itwasprovedforquaternioni polynomialsbyPogoruiandShapiro [24℄, for quaternioni regular fun tions byGentili and Stoppato in [11℄ and it wasextendedtoo tonioni regularfun tionsin [18℄.

(8)

Theorem6. Let

f : B

R

−→ O

bearegularfun tion,whi hdoesnotvanishon the whole

B

R

. For ea h

I ∈ S

, denote by

C

I

the omplexplane of

O

generated

by

1

and

I

; that is,

C

I

:= {x + yI ∈ O | x, y ∈ R}

. The following statements

hold.

(1)

Forall

I ∈ S

,

C

I

S

α∈V (f )

S

α

is losedanddis retein

C

I

∩ B

R

.

(2)

For ea h

β ∈

S

α∈V (f )

S

α

, either

S

β

⊂ V (f )

or

S

β

∩ V (f )

onsists of a singlepoint.

Proof. Sin e

N (f )

is regular on

B

R

, and then holomorphi on

C

I

∩ B

R

, its zeroset

C

I

S

α∈V (f )

S

α

= C

I

∩ V (N (f ))

is losed anddis rete in

C

I

∩ B

R

. These ond statementfollowsfromCorollary2.

Denition2. Given anonnegativeinteger

s

and ano tonion

α

in

V (f )

,we say that

α

is a zeroof

f

of multipli ity

s

if

s

α

| N (f )

and

s+1

α

∤ N (f )

. We

willdenote theinteger

s

, alled multipli ityof

α

,by

m

f

(α)

.

Inthe aseof

α

arealelement,this onditionis equivalentto

(w − α)

s

| f

and

(w − α)

s+1

∤ f

. If

α

isaspheri alzero,then

α

divides

f

and

f

¯

. Therefore

m

f

(α)

is at least 2. If

m

f

(α) = 1

,

α

is alled asimple zero of

f

. The reader observesthatthemultipli ityof

α

dependsonlyonthe onjuga y lass

S

α

of

α

. Remark 2. Thepre eding denition isequivalenton quaternioni polynomials withtheonegivenin[2℄andin[16℄.

3 Zeros of produ ts: the  amshaft ee t

Let

f, g : B

R

−→ O

beregularfun tions. Theaimofthisse tionistodes ribe

the zeros of

f ∗ g

in terms of the zeros of

f

and of

g

. The following result is nontrivial andveryimportant in the o tonioni setting. It wasprovedby Serdio([28,Theorem10℄)in theparti ular asein whi h

g

isa onstant. Lemma7. Itholds:

N (f ∗ g) = N (f ) N (g).

Proof. Let

g

bexed. Wemustprovethat thetworealquadrati forms

Q

1

(f ) := (f ∗ g) ∗ (¯

g ∗ ¯

f )

and

Q

2

(f ) := (f ∗ ¯

f )(g ∗ ¯

g)

oin ide.Bypolarization,theequalityof

Q

1

and

Q

2

isequivalenttotheequality ofthe

R

bilinearsymmetri forms

B

1

(f

1

, f

2

) := (f

1

∗ g) ∗ (¯

g ∗ ¯

f

2

) + (f

2

∗ g) ∗ (¯

g ∗ ¯

f

1

),

B

2

(f

1

, f

2

) := (f

1

∗ ¯

f

2

)N (g) + (f

2

∗ ¯

f

1

)N (g).

By linearity, it is su ient to prove that

B

1

and

B

2

oin ide for

f

1

= w

i

a

,

f

2

= w

j

b

, with

a, b ∈ O

. Sin e

B

α

(w

i

a, w

j

b) = w

i+j

∗ B

α

(a, b)

(

α = 1, 2

), it

(9)

totheequality

Q

1

(a) = Q

2

(a)

forevery

a ∈ O

. Nowwe anx

a ∈ O

andapply thesameargumenttothequadrati forms

Q

1

(g) := (a ∗ g) ∗ (¯

g ∗ ¯

a)

and

Q

2

(g) := |a|

2

(g ∗ ¯

g).

We get that these two forms oin ide if and only if

Q

1

(b) = Q

2

(b)

for every

b ∈ O

. Thereforeitsu estoprovetheequality

(a b)(¯b ¯

a) = |a|

2

|b|

2

forevery onstant

a, b ∈ O

.

Ineveryalternativealgebrathesubalgebragenerated bytwoelementsis asso- iative. Therefore

(a b)(¯b ¯

a) = a(b¯b)¯

a = |a|

2

|b|

2

. This ompletestheproof. Corollary8. Itholds:

[

α∈V (f ∗g)

S

α

=

[

α∈V (f )∪V (g)

S

α

or, equivalently, given any

α ∈ O

,

V (f ∗ g) ∩ S

α

is nonempty if and only if

(V (f ) ∪ V (g))∩S

α

isnonempty. Inparti ular, thezerosetof

f ∗ g

is ontained

inthe unionof spheres

S

α∈V (f )∪V (g)

S

α

.

Proof. By ombiningCorollary5andLemma 7,weget:

S

α∈V (f ∗g)

S

α

= V (N (f ∗ g)) = V (N (f )) ∪ V (N (g)) =

=



S

α∈V (f )

S

α





S

α∈V (g)

S

α



=

S

α∈V (f )∪V (g)

S

α

.

Sin e

V (f ∗ g) ⊂ V (N (f ∗ g))

, theproofis omplete.

Nowwegiveapre isedes riptionof thezerosoftheprodu t

f ∗ g

. Thanks to Corollary8, itis su ient to analyze separately thespheres

S

α

ontaining zerosof

f

orof

g

.

Firstly,we onsider thesigni ant ase: thenonrealzeros.

Let

α ∈ O

be nonreal. Supposethat

r

α

(f )(w) = wa + b

and

r

α

(g)(w) =

wc + d

forsome

a, b, c, d ∈ O

. Thenwehavethat

r

α

(f ∗ g)(w) = w(ad + bc + t

α

ac) + bd − n

α

ac = (r

α

(f ) ∗ r

α

(g))(w) − ∆

α

ac.

(3)

Thelatterassertionisanimmediate onsequen eofthefollowingfa t

(wa + b) ∗ (wc + d) =

w

2

(ac) + w(ad + bc) + bd =

=

(∆

α

+ wt

α

− n

α

)(ac) + w(ad + bc) + bd

ThankstoEq.(3),weobtain:

(10)

(1)

Suppose

f

has anisolatedzero

α

in

S

α

of multipli ity

s

and

g

isnowhere

zero on

S

α

. Then

f ∗ g

has an isolated zero

α

in

S

α

of multipli ity

s

,

ad + (¯

αa)c 6= 0

anditholds:

α

= ((αa)d + n

α

ac) (ad + (¯

αa)c)

−1

.

Inparti ular, if

(αa)d = α(ad)

and

n

α

ac = α ((¯

αa)c) (

for example,if

a

is

realor if

a, b, c, d

belong toanasso iative subalgebraof

O )

,then

α

= α

.

(2)

Suppose

f

is nowhere zero on

S

α

and

g

has an isolated zero

β

in

S

α

of multipli ity

t

. Then

f ∗ g

has an isolated zero

β

in

S

α

of multipli ity

t

,

bc + a( ¯

βc) 6= 0

anditholds:

β

= (b(βc) + n

α

ac) bc + a( ¯

βc)



−1

.

Inparti ular, if

b(βc) = (bβ)c

and

a( ¯

βc) = (a ¯

β)c (

for example, if

c

isreal

orif

a, b, c, d

belongstoanasso iativesubalgebraof

O )

,thenwehavethat

β

= (b + a ¯

β )β(b + a ¯

β )

−1

.

(3)

Suppose

f

has an isolated zero

α

in

S

α

of multipli ity

s

and

g

has an isolatedzero

β

in

S

α

ofmultipli ity

t

. Then,if

a(βc) = (¯

αa)c

,then

α

isa spheri alzeroof

f ∗ g

ofmultipli ity

s + t

. Otherwise,

f ∗ g

hasanisolated zero

γ

in

S

α

ofmultipli ity

s + t

su hthat

γ = (−(αa)(βc) + n

α

ac) (−a(βc) + (¯

αa)c)

−1

.

(4)

Let

α

beaspheri alzeroof

f

orof

g

,andlet

s, t

betherespe tive multipli -itiesof

α (

possiblyzero

)

. Then

α

isaspheri al zeroof

f ∗ g

ofmultipli ity

s + t

.

Proof. In thefollowing,we willuse thefa t that, in anyalternativequadrati algebra,thetra eisasso iativeand ommutative( f.e.g.[6,Ÿ9.1.2℄).

First,weprovethetheoremwithoutthepart on erningmultipli ities.

(1)Inthis ase

a 6= 0

and

αa + b = 0

( f.Corollary2(2)(ii)). If

c = 0

,then

d 6= 0

(otherwise

g

wouldvanish onthewhole sphere

S

α

). Then

ad + (¯

αa)c =

ad 6= 0

andEq.(3)givestheisolatedzero

α

= −(bd)(ad)

−1

= ((αa)d)(ad)

−1

of

f ∗ g

. If

c 6= 0

,then

ad + (¯

αa)c

annotvanish. Infa t,ifitwere

ad + (¯

αa)c = 0

,

then we would have that

α = −((ad)c

¯

−1

)a

−1

. But then

tr(α) = tr(−dc

−1

)

and

|α| = | − dc

−1

|

. This would mean that

β = −dc

−1

belongsto

S

α

∩ V (g)

, ontradi tingourassumptions.

Corollary8tellsthat

S

α

must ontainazeroof

f ∗ g

. FromCorollary2(2)(ii) andEq.(3),thiszeromustbe

α

= (−bd + n

α

ac)(ad + bc + t

α

ac)

−1

.

Note that the o tonion

ad + bc + t

α

ac

does not vanish, sin e it is equal to

ad − (αa)c + (t

α

a)c = ad − (αa)c + (αa)c + (¯

αa)c = ad + (¯

αa)c

. We on lude

that

α

= ((αa)d + n

(11)

(2)This se ond ase issimilarto therstone. Now

c 6= 0

and

βc + d = 0

.

If

a = 0

,then

b 6= 0

and

bc + a( ¯

βc) = bc 6= 0

. FromEq.(3)wegettheisolated

zero

β

= (b(βc))(bc)

−1

.

If

a

doesnotvanish,thenitstill holds

bc + a( ¯

βc) 6= 0

. Infa t,ifthiswere nottrue,wewouldhavethat

β

is onjugateto

−ba

−1

and therefore

−ba

−1

∈ S

α

∩ V (f )

. ByusingCorollary8,Corollary2(2)(ii),

βc = −d

and

t

α

= t

β

,weobtaintheuniquezero

β

= (−bd + n

α

ac)(ad + bc + t

α

ac)

−1

= (b(βc) + n

α

ac)(bc + a( ¯

βc))

−1

of

f ∗ g

in

S

α

.

(3) In this ase,

a, c 6= 0

,

b = −αa

,

d = −βc

. From Eq. (3), we get the remainder:

r

α

(f ∗ g)(w) = w(−a(βc) − (αa)c + t

α

ac) + (αa)(βc) − n

α

ac

= w(−a(βc) + (¯

αa)c) + (αa)(βc) − n

α

ac.

If

a(βc) = (¯

αa)c

,then

r

α

(f ∗ g)

is onstant. Ontheotherhand,byCorollary8,

r

α

(f ∗ g)

vanishesonsomepointof

S

α

andhen eitisnull. Itfollowsthat

f ∗ g

has

α

asaspheri alzero. Instead,if

a(βc) 6= (¯

αa)c

,then

f ∗ g

hastheisolated zero

γ = (−(αa)(βc) + n

α

ac) (−a(βc) + (¯

αa)c)

−1

.

(4) If the real polynomial

α

divides

f

or

g

, then it divides

f ∗ g

. The on lusionfollowsfromCorollary2(2)(i).

Now we onsider multipli ities. If

α

is a zero of

f

of multipli ity

s

and

β ∈ S

α

isa zeroof

g

ofmultipli ity

t

, then

N (f ) = ∆

s

α

h

1

and

N (g) = ∆

t

α

h

2

, where

h

1

and

h

2

real regular fun tions su h that

r

α

(h

1

) = wa + b 6= 0

and

r

α

(h

2

) = wc + d 6= 0

. FromLemma 7, itfollowsthat

N (f ∗ g) = ∆

s+t

α

(h

1

h

2

)

. Wehaveto provethat

s+t+1

α

∤ N (f ∗ g)

. Byadensityargument,wesee that

this is equivalent to show that

α

∤ h

1

h

2

, i.e. that theremainder

r

α

(h

1

h

2

)

is notzero. FromEq.(3),

r

α

(h

1

h

2

)

iszeroifandonlyif

ad + bc + t

α

ac = bd − n

α

ac = 0.

(4)

Sin e

a, b, c, d

are real, Eq. (4) impliesthat

a(d + αc)(d + ¯

αc) = 0

. But this annothappensin e

α

isnonrealand

a 6= 0

.

Example 1. Foreverynonreal

α ∈ O

,

α

¯

is onjugatedto

α

. Let

a

besu hthat

¯

α = aαa

−1

. If

f (w) = wa−αa

,

g(w) = w−α

,then

f ∗g = w

2

a−w(aα+αa)+αaα

and

V (f ) = V (g) = {α}

while

V (f ∗ g) = S

α

(seeTheorem9(3)).

Remark 3. In the o tonioni ase, the amshaft ee t appears even if oneof the fun tions is onstant: the zeros of

f

and of

f ∗ c

(

c ∈ O

a onstant) an bedierent( f.Serdio[27℄when

f

isapolynomial). Forexample,let

f (w) =

wi − j

,

g(w) = k

. Then

f ∗ g = w(ik) − jk

has

−ij

as unique zero, while

V (f ) = {ij}

.

(12)

Proposition 10. Let

α

be a real o tonion. Suppose that

α

is a zero of

f

of multipli ity

s

and azeroof

g

of multipli ity

t ( s

and

t

possibly zero

)

. Then

α

isazeroof

f ∗ g

ofmultipli ity

s + t

.

Proof. If the real polynomial

w − α

divides

f

or

g

, then it divides

f ∗ g

. In parti ular,

r

α

(f ∗ g) = 0

. Moreover, if

f = (w − α)

s

h

1

and

g = (w − α)

t

h

2

with

h

1

(α) 6= 0

and

h

2

(α) 6= 0

, then

f ∗ g = (w − α)

s+t

(h

1

∗ h

2

)

. It remains toprovethat

(w − α)

s+t+1

∤ f ∗ g

. Bydensity, this isequivalenttoshowthat

w − α ∤ h

1

∗ h

2

, i.e.

(h

1

∗ h

2

)(α) 6= 0

. But

(h

1

∗ h

2

)(α) = 0

would imply

that

N (h

1

∗ h

2

)(α) = N (h

1

)(α) · N (h

2

)(α) = 0

. On the other hand, sin e

α

is real,

N (h

1

)(α) = |h

1

(α)|

2

6= 0

and

N (h

2

)(α) = |h

2

(α)|

2

6= 0

, whi h is a ontradi tion. 4 Appli ations

Webeginwithano tonioni versionofthefundamentaltheoremofalgebra. Theorem11 (Fundamental theorem of algebra). If

f

isa regularo tonioni polynomial of degree

n > 0

,then its zero set

V (f )

is nonempty and the ar-dinality

k

of the distin t onjuga y lasses of elements of

V (f )

is less than or equal to

n

. More pre isely, if

α

1

, . . . , α

r

are the distin t real zeros of

f

,

α

r+1

, . . . , α

r+i

are the distin t isolated zeros of

f

and

α

r+i+1

, . . . , α

r+i+s

are

pairwisenon onjugatespheri al zerosof

f

su hthat

S

s

j=1

S

α

r+i+j

isthe setof allspheri al zerosof

f

,then

k = r + i + s

andthe following equality holds:

k

X

j=1

m

f

j

) = n.

Inparti ular, wehave that

r + i + 2s ≤ n

.

Proof. Thenormalpolynomial

N (f )

isarealpolynomialofdegree

2n

. Let

I ∈ S

.

Thentheset

V

I

(N (f )) := {z ∈ C

I

| N (f )(z) = 0} = C

I

S

α∈V (f )

S

α

is non emptyand ontainsat most

2n

elements. Corollary5tellsthat

V (f ) ∩ S

α

6= ∅

forevery

α

su hthat

V

I

(N (f )) ∩ S

α

6= ∅

. Therefore,

V (f )

isnonempty.

If

f = ∆

α

h + wa + b

forsome regular fun tion

h

and

a, b ∈ O

, then

f =

¯

α

¯

h+w¯

a+¯b

. Thisimpliesthatanisolatedzero

α = −ba

−1

of

f

,with

m

f

(α) = s

,

orrespondstoanisolatedzero

α = −¯b¯

ˆ

a

−1

∈ S

α

of

f

¯

withthesamemultipli ity. From Theorem 9(3), we get that

α

is a spheri al zero of

N (f ) = f ∗ ¯

f

of multipli ity

2s

. The same property holds for spheri al and real zeros of

f

. Then,given

α

1

, . . . , α

k

as inthestatementofthetheorem,itfollowsthat

2

k

X

j=1

m

f

j

) =

k

X

j=1

m

N

(f )

j

) = 2n.

Sin e the multipli ity of a spheri al zero is at least

2

, it follows immediately

(13)

andofTheorem11gives:

Corollary12(Fa torizationlemma). Let

f

bearegularo tonioni polynomial of degree

n > 0

. Let

α

1

∈ V (f )

. Then thereexist

α

2

, . . . , α

n

, c ∈ O

with

c 6= 0

su hthat

f

fa tors asfollows:

f (w) = (w − α

1

) ∗ f

1

(w),

where

f

k

(w) : = (w − α

k+1

) ∗ f

k+1

(w)

for

k = 1, 2, . . . , n − 2

and

f

n−1

(w) : = (w − α

n

) ∗ c.

Moreover, forevery

k = 1, . . . , n − 1

,

α

k+1

isazeroof

f

k

anditis onjugateto azero

α

k+1

of

f

.

Remark 4. Theorem9tellshowto obtainthezeros

α

k+1

∈ V (f )

from theset

i

}

i=2,...,n

.

A standardappli ation ofthefundamental theorem of algebrain the om-plexeldistheuniquenessofmoni polynomialswithpres ribedzeros. Onthe quaternionsandontheo tonions,anewphenomenonappears. The multipli i-ty of an isolated zero of the sumof two regular fun tions an bestri tly less thanthoseofthefun tions. This fa timpliesthat twodierentmoni polyno-mials an havethesame zeroswith thesame multipli ities. Forexample, the polynomials

f (w) = (w − i) ∗ (w − i)

and

g(w) = (w − i) ∗ (w − j)

have the uniqueisolatedzero

w = i

withmultipli ity2,whilethedieren e

f − g

hasa simplezero

i

. Thisfa t annothappenwhenallthezerosarerealorspheri al orsimple,sin ein this ase

m

f+g

(α) ≥ min{m

f

(α), m

g

(α)}.

On the quaternions, the existen e of a unique moni regular polynomial withassignedpairwisenon onjugatezeroswasprovedbyBe k[1℄(seealso[3℄, [2℄ and [16℄). On the o tonions, it wasre ently proved by Serdio [28℄. The existen eofaninnitenumbersofmoni regularquaternioni polynomialswith pres ribedmultipleisolatedzeroswasprovedbyBe k[1℄.

An immediate onsequen e of the fundamental theorem and Propositions 3, 4 is the following result, whi h generalizes the one proved by Lam in the quaternioni ase( f.[22, Ÿ16.19℄):

Corollary 13. Let

f (w) =

P

n

i=0

w

i

a

i

be a regular o tonioni polynomial of degree

n > 0

. Thefollowing statementshold.

(1)

If the oe ients

a

1

, . . . , a

n

are real and

a

0

is nonreal, then

f

has only isolatedzeros.

(2)

If

a

2

, . . . , a

n

arereal andeither

a

0

or

a

1

isnonreal,then

f

has no

spher-i al zeros.

Inparti ular, theset

V (f )

ontainsexa tly

n

elements, ountedwiththeir mul-tipli ities.

(14)

onsider the polynomial

f (w) = w

2

+ wi + j

. Sin e

N (f ) = w

4

+ w

2

+ 1

,

S

α∈V (f )

S

α

is equalto

S

β

∪ S

γ

, where

β :=

1

2

+ i

3

2

and

γ := −

1

2

+ i

3

2

. The orrespondingremaindersare

r

β

(f ) = w(1 + i) − (1 − j)

and

r

γ

(f ) = w(−1 + i) − (1 − j),

whi hgivethefollowingtwosimpleisolatedzerosof

f

:

α

1

=

1

2

(1−i−j−ij) ∈ S

β

and

α

2

=

1

2

(−1 − i + j − ij) ∈ S

γ

. Referen es

[1℄ B. Be k, Sur les équations polynomiales dans les quaternions. Enseign. Math. (2)25(1979),no.3-4,193201.

[2℄ U. Bray, G. Whaples, Polynomials with oe ients from a division ring. Canad.J.Math.35(1983),no.3,509515.

[3℄ L.Cerlien o,M.Mureddu,Anoteonpolynomialequationsinquaternions. Rend. Sem.Fa .S i.Univ.Cagliari51(1981),no.1,9599.

[4℄ F. Colombo,I. Sabadini,D.C. Struppa,Sli e monogeni fun tions. Israel J.Math.,171(2009),385403.

[5℄ B. Datta, S. Nag, Zero-sets of quaternioni ando tonioni analyti fun -tions with entral oe ients. Bull. London Math. So . 19 (1987), no. 4, 329336.

[6℄ H.-D.Ebbinghaus,H. Hermes,F.Hirzebru h, M.Koe her,K.Mainzer,J. Neukir h, A. Prestel, R.Remmert, Numbers. Springer-Verlag,NewYork, 1991.

[7℄ S. Eilenberg, I.Niven, The fundamental theorem of algebra for quater-nions.Bull.Amer. Math.So .50,(1944),246248.

[8℄ S. Eilenberg, N. Steenrod, Foundations of algebrai topology. Prin eton UniversityPress,Prin eton,NewJersey,1952.

[9℄ I.Gelfand,V.Retakh,R.Wilson,Quadrati -linearalgebrasasso iatedwith de ompositions of non ommutative polynomials anddierential p olynomi-als.Sele taMath. (N.S.)7,493523(2001).

[10℄ I.Gelfand, S. Gelfand, V. Retakh, R.L.Wilson, Fa torizations of polyno-mials overnon ommutativealgebrasandsu ientsetsofedges indire ted graphs.Lett.Math.Phys.74(2005),no.2,153167.

[11℄ G. Gentili, C. Stoppato, Zeros of regular fun tions and polynomials of a quaternioni variable. Mi higanMath. J.,56(2008),665667.

(15)

forregularfun tionsofahyper omplexvariable.InHyper omplexanalysis, eds. I. Sabadini, M. Shapiro and F. Sommen, Trends Math., Birkhäuser, Basel,2009,pp.165186.

[13℄ G. Gentili, D. Struppa, A new approa h toCullen-regular fun tions of a quaternioni variable. C.R.A ad.S i.Paris,342(2006),741744.

[14℄ G.Gentili, D.Struppa,Regularfun tionsonthespa eof Cayley numbers. ToappearinRo kyMountainJ.Math.

[15℄ G.Gentili,D.Struppa,Anewtheoryofregularfun tionsofaquaternioni variable. Adv.Math. 216(2007),279301.

[16℄ G. Gentili, D. Struppa, On the multipli ity of zeroes of polynomials with quaternioni oe ients.MilanJ.Math.,76(2008),1525.

[17℄ G. Gentili, D. Struppa,Regularfun tionson a CliordAlgebra. Complex Var.Ellipti Equ. 53(2008),475483.

[18℄ G.Gentili,D. Struppa,F.Vla i, Thefundamental theoremof algebrafor Hamilton andCayley numbers.Math. Z.,259(2008),895-902.

[19℄ K.Gürlebe k,K.HabethaandW. Sprössig,Holomorphi fun tionsin the planeand

n

-dimensionalspa e.Translatedfrom the2006Germanoriginal Funktionentheoriein EbeneundRaum, BirkhäuserVerlag,Basel,2008. [20℄ B.Gordon,T.S. Motzkin,Onthe zerosof polynomialsoverdivision rings.

Trans.AMS1965,116,218226.

[21℄ Y. Jou, The fundamentaltheorem of algebra for Cayley numbers. A ad. Sini aS ien e Re ord3,(1950),2933.

[22℄ T.Y. Lam, A rst ourse in non ommutativerings. Springer-Verlag, New York,1991.

[23℄ I.Niven,Equationsinquaternions.Amer.Math.Monthly48,(1941),654 661.

[24℄ A. Pogorui, M.V. Shapiro, On the stru tureof the set of zeros of quater-nioni polynomials.ComplexVariables49,(2004)no.6,379389.

[25℄ H. Rodríguez-Ordóñez,A note onthe fundamental theorem of algebra for the o tonions.Expo.Math. 25(2007),no.4,355361.

[26℄ R.D.S hafer,Anintrodu tiontononasso iativealgebras.A ademi Press, NewYork-London1966.

[27℄ R.Serdio,Ono tonioni polynomials.Adv.Appl.CliordAlg.17(2007), 245258.

(16)

Alg.published online,(2008),DOI 10.1007/s00006-008-0140-5.

[29℄ R. Serdio,L.Siu,Zeros of quaternionpolynomials.Appl. Math.Lett.14 (2001),no.2,237239.

[30℄ A.Sudbery,Quaternioni analysis.Mat.Pro .Camb.Phil.So .85(1979), 199225.

[31℄ N.Topuridze,Onthe rootsofpolynomialsoverdivisionalgebras.Georgian Math. J.10(2003),no.4,745762.

Riferimenti

Documenti correlati

I use writing reform as a cover term for reforms of writing systems on three levels: script reforms, which change the script used for a language (e.g. from the Latin to the Cyril-

blockchain; Bitcoin; cryptocurrencies; sovereignty; central banking; fiat money; permissionless; public economic law... 1 Revolutionary by design: the invention of the

Dapprima, avvalendosi dei sondaggi sull’uso del tempo, mostra come alcuni recenti mutamenti relativi ai modelli lavorativi (diffusione di attività meno routinarie e

Consecutive 42 patients have been operated upon in our Department. The patients ranging 4 to 44 years of age were affected by various degrees of pectus excavatum. :

We calculate the critical radiation intensity needed for supermassive star formation by direct collapse and examine its dependence on the H + 2 level population.. Under the

On 19 December 2018 the UN General Assembly approved the Global Compact for Safe, Orderly and Regular Migration (GCM), with 152 votes in favor, five against (Czech