• 検索結果がありません。

ON THE ZEROS AND CRITICAL POINTS OF A RATIONAL MAP

N/A
N/A
Protected

Academic year: 2022

シェア "ON THE ZEROS AND CRITICAL POINTS OF A RATIONAL MAP"

Copied!
4
0
0

読み込み中.... (全文を見る)

全文

(1)

IJMMS 28:4 (2001) 243–246 PII. S0161171201011589 http://ijmms.hindawi.com

© Hindawi Publishing Corp.

ON THE ZEROS AND CRITICAL POINTS OF A RATIONAL MAP

XAVIER BUFF (Received 23 January 2001)

Abstract.Letf:P1P1be a rational map of degreed. It is well known thatf hasd zeros and 2d2 critical points counted with multiplicities. In this note, we explain how those zeros and those critical points are related.

2000 Mathematics Subject Classification. 30C15.

In this note,f:P1P1is a rational map. We denote by i}iI the set of zeros off, and byj}jJthe set of critical points off which are not zeros off(the sets I and J are finite). Moreover, we denote byni the multiplicity ofαi as a zero off and bymjthe multiplicity ofωjas a critical point off. The local degree off atαi

isniand the local degree off atωj isdj=mj+1. In particular, whenωjand f (ωj), the pointωjis a zero offof ordermj.

Our goal is to understand the relations that exist between the pointsαi and the pointsωj.

Proposition1. Given a finite collection of distinct pointsαiP1with multiplicities niandωjP1with multiplicitiesmj, there exists a rational mapfvanishing exactly at the pointsαi with multiplicitiesni and having extra critical points exactly at the pointsωjwith multiplicitiesmjif and only if

(i)

(ni+1)

mj=2,and (ii) for anyksuch thatαkC,

res

ωj∈C z−ωj

mj

αi∈C z−αi

ni+1dz, αk

=0. (1)

We will give a geometric interpretation of (ii) in the case whereαkis a simple zero off: working in a coordinate whereαk= ∞, the barycentre of the remaining zeros weighted with their multiplicities is equal to the barycentre of the critical points off weighted with their multiplicities (seeProposition 3below).

Proof. The proof is elementary. It is based on the observation that the 1-forms d(1/f )and

φ=

ωj∈C z−ωj

mj

αi∈C z−αi

ni+1dz (2)

are proportional. The differential equationd(1/f )=φhas a rational solution if and only ifφis exact, if and only if the residues ofφat all finite poles are equal to zero.

(2)

244 XAVIER BUFF

Lemma2. Letfbe a rational map. Denote byαiits zeros and bynitheir multiplici- ties. Denote byωjthe critical points offwhich are not multiple zeros offand bymj

their multiplicities. The zeros of the1-formd(1/f )are exactly the pointsωjwith order mjand its poles are exactly the pointsαiwith orderni+1.

Proof. A singularity of the 1-formd(1/f )= −df /f2is necessarily a zero or a pole off, a zero off, or(whereφis defined by analytic continuation). Considering the Laurent series off at each of those points, one immediately gets the result.

Now assume that there exists a rational map f with the required properties.

Lemma 2shows that the 1-formsφand d(1/f ) have the same poles and the same zeros inC, with the same multiplicities. Hence, their ratio is a rational function which does not vanish inC. Thus,φ and d(1/f )are proportional. In particular, φhas a singularity atif and only ifd(1/f )has a singularity atand the singularity is of the same kind for both 1-forms. Since the number of poles minus the number of zeros of any nonzero 1-form onP1is equal to 2 (the Euler characteristic ofP1), we see that (ni+1)

mj=2 which is precisely condition (i) inProposition 1. Besides, since φis exact, it follows that the residues at all the polesαkvanish and condition (ii) is satisfied.

Conversely, the 1-formφhas poles of orderni+1 at the pointsαiCand zeros of ordermjat the pointsωjC. Condition (ii) implies thatφis exact, that is, there exists a rational mapg:P1P1such thatφ=dg. Since the number of poles ofφin P1minus the number of zeros ofφinP1is equal to 2, condition (i) implies that when

is neither a pointαinor a pointωj, it is a regular point ofφ, when∞ =αi0, it is a pole ofφof orderni0, and when∞ =ωj0, it is a zero ofφof ordermj0. Finally, φ=d(1/f ), withf=1/g, andLemma 2shows that the rational mapf=1/gvanishes exactly at the pointsαiwith multiplicitiesniand has extra critical points exactly at the pointsωjwith multiplicitiesmj.

We will now give a geometric interpretation of (ii) whenαkis a simple zero off.

We first work in a coordinate whereis neither one of the pointsαinor a pointωj. Define

R(z)=

j

z−ωj

mj

ik

z−αi

ni+1. (3)

Then,

res

j

z−ωj

mj

i

z−αini+1dz, αk

=res R(z)

z−αk2dz, αk

=R αk

. (4)

SinceR(αk)≠0, this residue vanishes if and only if R

αk

R

αk =

j

mj

αk−ωj

ik

ni+1

αk−αi=0. (5)

Letdbe the number of zeros counted with multiplicities, that is,d=

ini. The total number of critical points is 2d2=

jmj+

i(ni1)(the critical points off are

(3)

ON THE ZEROS AND CRITICAL POINTS OF A RATIONAL MAP 245 the pointsωjand the multiple zeros off). Then, (5) can be rewritten as

1 2d2

j

mj

αk−ωj+

i≠k

ni1 αk−αi

= 1 d−1

i≠k

ni

αk−αi

. (6)

This last equality can be interpreted in the following way.

Proposition3. Assume thatf is a rational map having a simple zero at∞. Then, the barycentre of the remaining zeros weighted with their multiplicities is equal to the barycentre of the critical points off weighted with their multiplicities.

Remark4. One can prove this proposition directly. We may writef=P /Q, where

P=

d−1

k=0

akzk, Q= d k=0

bkzk, (7)

are co-prime polynomials with deg(Q)=deg(P )+1=d. Without loss of generality, we may assume that the barycentre of the zeros offis equal to 0. In other words, we may assume thatP is a centered polynomial, that is,ad2=0. A simple calculation shows that

PQ−QP=

2d2 k=0

ckzk (8)

is a polynomial of degree 2d2 and thatc2k−1=0. Therefore, the barycentre of the zeros ofPQ−QP, that is, the barycentre of the critical points off, is equal to 0.

Apply this geometric interpretation in order to re-prove two known results. The first corollary is related to the Sendov conjecture (cf. [1] and more particularly Section 4).

This conjecture asserts that if a polynomialPhas all its roots in the closed unit disk, then, for each zeroαithere exists a critical pointω(possibly a multiple zero) such thati−ω| ≤1.

Corollary5. LetP:CCbe a polynomial. Assume the zeros ofPare all contained in the closed unit disk andα0∈S1is a zero ofP. Then, the closed disk of diameter[0, α0] contains at least one critical point off.

Proof. Denote bydthe degree ofP. Ifα0is a multiple zero ofP, then the result is trivial. Thus, assumeα0 is a simple zero of P. We work in the coordinateZ = α0/(α0−z). The rational mapf:ZP (α0−α0/Z)has a simple zero atZ= ∞and the remaining zeros are contained in the half-plane{Z∈P1| (Z)≥1/2}. Thus the barycentreβof those zeros satisfies(β)≥1/2. Moreover,f has a critical point of multiplicitydatZ=0. Thus, the barycentre of thedremaining critical points is 2β.

Since(2β)≥1, we see thatf has at least one critical pointωcontained in the half plane{Z∈P1| (Z)≥1}. Then,α0−α0is a critical point ofP contained in the closed disk of diameter[0, α0].

The second corollary has been proved by Videnskii [2]. Our result provides an al- ternate proof.

(4)

246 XAVIER BUFF

Corollary6. Assume thatf :P1P1is a rational map andP1 is a closed disk or a closed half-plane containing all the zeros off. Then,contains at least one critical point off.

Proof. Without loss of generality, we may assume that the zeros are simple and that at least one zero, sayα0, is on the boundary of∆. In a coordinate whereα0= ∞,

∆is a closed half-plane. The barycentre of the remaining zeros is contained in this half-plane. Consequently, the barycentre of the critical points is contained in∆. Thus,

∆contains at least one critical point.

Videnskii also proved that this result is optimal in the sense that there exist rational maps of arbitrary degrees with simple zeros contained in a disk∆but only one critical point in∆.

References

[1] M. Marden,Conjectures on the critical points of a polynomial, Amer. Math. Monthly90 (1983), no. 4, 267–276.MR 84e:30007. Zbl 535.30010.

[2] I. Videnskii,On the zeros of the derivative of a rational function and invariant subspaces for the backward shift operator on the Bergman space, in preparation.

Xavier Buff: Université Paul Sabatier, Laboratoire Emile Picard, 118, Route de Narbonne,31062Toulouse Cedex, France

E-mail address:[email protected]

参照

関連したドキュメント

The following theorem will be proved: For any C 3 unimodal map of an interval with a nonflat critical point there exists an interval around the critical value such that the first

The equality in (2.6) holds if and only if a and b are linearly dependent, or the vector c is a linear combination of a and b, and (a, c)(b, c) = 0, but (a, c) and (b, c) are

Bandelt [1], it was proved in [7] that the tolerance lattice TolA of an algebra A with a majority term is 0-modular and pseudocomplemented.. As TolA is an algebraic lattice, the

We introduce a modified process on the graph ~ L 2 alt with the same structure as the original oriented site percolation problem except in that if any two sites are wetted at time

In this work we study an exact boundary control problem for the standard wave equation on a domain with moving boundary which has a single fixed hole.. The boundary of such domains

As we shall see, these two 3-parameter noncompact groups are rudiments of the 3-parameter groups of relativistic symmetry of the axially symmetric Fins- lerian spaces with the

The main result is Theo- rem 1 which shows that under certain circumstances a critical group of a directed graph is the quotient of a critical group of its directed line graph..

Using a modified Routh array, ˘Siljak has provided a finite algebraic procedure to count the number of positive (or negative) zeros, with their multiplicity (see also the more