• Nem Talált Eredményt

A note on the location of the zeros of polynomials defined by linear recursions.

N/A
N/A
Protected

Academic year: 2022

Ossza meg "A note on the location of the zeros of polynomials defined by linear recursions."

Copied!
6
0
0

Teljes szövegt

(1)

A NOTE ON THE LOCATION OF ZEROS OF POLYNOMIALS DEFINED BY LINEAR RECURSIONS

Ferenc Mátyás (Eger, Hungary)

Abstract. In this paper it is proved that some earlier results on the location of zeros of polynomials defined by special linear recursions can be improved if the Brauer’s theorem is applied instead of the Gershgorin’s theorem.

AMS Classification Number:11B39, 12D10

1. Introduction

Letn≥2an integer and define the polynomialsGn(x)by the recursive formula (1) Gn(x) =P(x)Gn−1(x) +Q(x)Gn−2(x),

where the polynomials P(x), Q(x), G0(x) and G1(x) are fixed polynomials from C[x] and at most G0(x) is the zeropolynomial. If it is needed then we use the notation

(2) Gn(P(x), Q(x), G0(x), G1(x))

instead ofGn(x). Thus, for example the wellknown Fibonacci (Fn(x))and Cheby- shev(Un(x))polynomials of the second kind can be obtained as

Fn(x) =Gn(x,1,0,1) and Un(x) =Gn(2x,−1,0,1), respectively.

Recently, we have delt with the location of zeros of polynomials defined by (1), where the polynomialsP(x), Q(x), G0(x) and G1(x) are special ones (see [3], [4], [5]). If the explicit values of the zeros of polynomialsGn(x)are unknown then one can try to determine such a subset of C that contains the zeros of Gn(x) for all n≥1. For example, P. E. Ricci [7] proved that if a complex numberzis a zero of the polynomialGn(x,1,1, x+ 1)for somen≥1then|z|<2. In [3] we investigated

Research was sponsored by the Hungarian OTKA Foundation No. T 032898.

(2)

the location of zeros of polynomialsGn(x,1, c, x+e)ifc 6= 0, and proved that if z∈Cis a zero of these polynomials for somen≥1then

(3) |z| ≤max (|e|+|c|,2).

Similar result was obtained in [5] and for special recursions of orderk≥2 in [4].

To give the location of the zeros of the abovementioned polynomials we applied the wellknown Gershgorin’s theorem. But, some papers written by J. Gilewicz and E. Leopold ([1], [2]) suggest that it would be better to apply the Brauer’s theorem, since the results are sharper ones. First, see these theorems.

LetA= (aij)be a quadratic matrix of ordern≥2andaij∈C. For1≤i≤n let

(4) Gi =



ω∈C : |ω−aii| ≤ Xn

t=1 t6=i

|ait|



 and for1≤i < j ≤n

(5) Bij =



ω∈C : |ω−aii| · |ω−ajj| ≤

 Xn

t=1 t6=i

|ait|



 Xn

t=1 t6=j

|ajt|





.

Gershgorin’s theorem.All the eigenvalues ofAare contained in the set

G= [n i=1

Gi.

Brauer’s theorem.All the eigenvalues ofAare contained in the set

B= [

1i<jn

Bij

(see [6]).

The purpose of this paper is to obtain a general theorem for the location of the zeros of polynomials defined by (1). Applying the Brauer’s theorem we improve the result given in (3).

(3)

2. Results

First we need the following lemma.

Lemma.For everyn≥1

Gn(P(x), Q(x), G0(x), G1(x)) = det(An), whereAn is the following tridiagonal Jacobi matrix of ordern:

An=







G1(x) ip

Q(x)G0(x) 0 0 . . . 0 0

ip

Q(x) P(x) ip

Q(x) 0 . . . 0 0

0 ip

Q(x) P(x) ip

Q(x) . . . 0 0

... ... ... ... ... ... ...

0 0 0 0 . . . ip

Q(x) P(x)





 .

Proof.The statement of the Lemma can be obtained by induction on n.

Theorem.Forn≥2all the zeros of the polynomials Gn(P(x), Q(x), G0(x), G1(x)) are located in the sets defined by

(6) n

z∈C : |G1(z)| ≤p

Q(z)G0(z)o

∪n

z∈C : |P(z)| ≤2p Q(z)o or

(7) {z∈C: |G1(z)P(z)| ≤2|Q(z)G0(z)|} ∪n

z∈C: |P(z)| ≤2p Q(z)o

.

Proof.It is known that the eigenvalues λ1, λ2, . . . , λn ofAn are the roots of the equationdet(λIn−An) = 0, which can be rewritten as

λn+an−1(x)λn1+an−2(x)λn2+. . .+a1(x)λ+ det(An) = 0,

whereIn is the unit matrix of ordernand the coefficients ai(x)ofλi-s depend on x. Thus forn≥2, by our Lemma, a complex numberzis a zero of the polynomial

Gn(P(x), Q(x), G0(x), G1(x))

(4)

iff 0 is an eigenvalue of the tridiagonal matrix An. Applying the Gershgorin’s theorem and (4) we get that

|G1(z)| ≤p

Q(z)G0(z) or |P(z)| ≤2p Q(z), while according to the Brauer’s theorem and (5)

|G1(z)P(z)| ≤2|Q(z)G0(z)| or |P(z)| ≤2p Q(z).

These prove the theorem.

We note that forn≥1

Gn(P(x), Q(x),0, G1(x)) =G1(x)·Gn(P(x), Q(x),0,1) =

G1(x)·Gn1(P(x), Q(x),1, P(x)),

thus ifzis a zero of the polynomialGn(P(x), Q(x),0, G1(x))then eitherG1(z) = 0 or z is a zero of the polynomial Gn1(P(x), Q(x),1, P(x)). In the latter case, by our theorem,z satisfies the inequality

|P(z)| ≤2p Q(z),

which matches with a direct consequence of Theorem 1 in [5].

3. Application

In the following part of this paper we shall apply our theorem to give the location of zeros of polynomialsGn(x,1, c, x+e), wherec, e∈C andc6= 0, since a large class of polynomials Gn(x) can be traced back to this form (see [5]). We have already mentioned that the result (3) can be obtained by the Gershgorin’s theorem, thus we demonstrate that the Brauer’s theorem (or (7)) gives in general a better estimation for the location of the zeros.

Forn≥2, according to (7), the zerosz ofGn(x,1, c, x+e)belong to the set (8) {z∈C : |z+e| · |z| ≤2|c|} ∪ {z∈C: |z| ≤2},

while by (6), they belong to the set

(9) {z∈C : |z+e| ≤ |c|} ∪ {z∈C: |z| ≤2},

from which (3) immediately follows. It can be seen that the zero ofG1(x) =x+e also belongs to the sets (8) and (9), further if a complex numberzsatisfies (8) then

(5)

z also satisfies (9). Therefore, the set defined by (8) can be a narrower one than the set defined by (9).

Let|e|+|c| ≤2. In this case the sets (8) and (9) are equals. Thus (3) cannot be improved, that is,|z| ≤2.

Let|e|+|c|>2. Applying the mappingC−→Cdefined by

z=|z|(cosϕ+isinϕ)7→z =|z|(cos(ϕ−arg(−e)) +isin(ϕ−arg(−e))), the sets (8) and (9) are transformed into the sets

(10) {z ∈C: |z− |e|| · |z| ≤2|c|} ∪ {z ∈C : |z| ≤2} and

(11) {z∈C:|z− |e|| ≤ |c|} ∪ {z ∈C:|z| ≤2},

respectively. Without loss of generality it is sufficient to deal with only (10) and (11) since we want to estimate|z| =|z|. Let z = x+iy, where x, y ∈R. Then (10) and (11) can be rewritten as

(12) (x− |e|)2+y2

x2+y2

≤4|c|2 or x2+y2≤4 and

(x− |e|)2+y2≤ |c|2 or x2+y2≤4, respectively. Investigating the graph of the implicit function

(x− |e|)2+y2

x2+y2

−4|c|2= 0, one can calculate that the graph always intersects the axisxin (13) x1=|e| −p

|e|2+ 8|c|

2 and x2= |e|+p

|e|2+ 8|c|

2 ,

while in the case|e|2≥8|c|the points (14) x3=|e| −p

|e|2−8|c|

2 and x4= |e|+p

|e|2−8|c| 2

are also intersecting points, and the inequalities 0<−x1< x3≤x4< x2

hold. Further, ifz =x+iysatisfies (12) and|e2| ≤8|c|then (15) |z|=|z| ≤x2<max(|e|+|c|,2) =|e|+|c|,

(6)

while in the case|e|2>8|c|

(16) |z|=|z| ≤max(2, x3) or x4≤ |z|=|z| ≤x2<|e|+|c|,

wherex1, x2, x3and x4 are defined by (13) and (14). It can be seen that (15) and (16) realy improve (3). (The numerical calculations are omitted in (13)–(16).)

References

[1] Gilewicz, J. & Leopold, E.,Location of the zeros of polynomials satisfying three-terms recurrence relation with complex coefficients,Integral Transforms and Special Functions,2(1994), 267–278.

[2] Gilewicz, J. & Leopold, E.,Zeros of polynomials and recurrence relation with periodic coefficients, Journal of Computational and Applied Math., 107 (1999), 241–255.

[3] Mátyás, F.,Bound for the zeros of Fibonacci type polynomials,Acta Acad.

Paed. Agriensis Sectio Math.,25(1998), 17–23.

[4] Mátyás, F., On a bound of the zeros of polynomials defined by special linear recurrences of orderk, Rivista di Mat. Univ. Parma, 6/1(1998), 173–180.

[5] Mátyás, F., On the location of the zeros of polynomials defined by linear recursions,Publ. Math. Debrecen,55/3–4(1999), 453–464.

[6] Parodi, M., La localisation des valeurs caractéristiques des matrices et ses applications, Gauthier Villars, Paris, 1959.

[7] Ricci, P. E., Generalized Lucas polynomials and Fibonacci polynomials, Rivista di Mat. Univ. Parma,4/5(1995), 137–146.

Ferenc Mátyás

Institute of Mathematics and Informatics Eszterházy Károly College

Leányka str. 4.

H-3300 Eger, Hungary e-mail: matyas@ektf.hu

Hivatkozások

KAPCSOLÓDÓ DOKUMENTUMOK

These sets (circles) are called to be Gershgorin-circles. Using these notations we formulate the following well-known theorem.. Now we are able to formulate our main result.. .)

In conclusion, it can be stated that stem cells that are able to rescue the vast majority of injured motoneurons destined to die following transplantation into the injured spinal

It is proved that three is the upper bound for the number of isolated zeros of Abelian integrals, and there exists some α, β and γ such that the Abelian integrals could have three

and Yang C.-C., Some further results on the zeros and growths of entire solutions of second order linear differential equations, Kodai Math.. and Yang C.-C., On the zeros

The theoretical possibility that a polynomial g ∈ E may not have all its n zeros on the unit circle can indeed occur in the case where n is odd.. This is illustrated by our

It is intriguing that this reasoning works fine for any even n, and so does the one that was used to prove Theorem C, but somehow both lack the sophistication needed to settle the

te specifies how to give the input parameters of the program. It equals one, if the poles and the zeros of the function ekeS) as well as the polynomials of

We mention first of all, that for a single Jordan curve Theorem 1.2 can be easily deduced from [1, Theorem 4.1.1] by taking the balayage of the normalized zero counting measure ν n