SOME REMARKS ON THE ZEROS OF TRIBONACCI POLYNOMIALS

In this paper, the zeros of Tribonacci polynomials are studied. The bound containing the zeros of Tribonacci polynomials has been numerically examined with comparisons. On the other hand, a new algorithm is given so that it can be used in other boundary problems.


Introduction
The features of polynomials have played an important role in many scientific areas such as control theory, signal processing, crytography and mathematical biology.If we have an accurate estimate of the region containing all the zeros of a polynomial, then the amount of work needed to find exact zeros can be considerably reduced in comparison with using the classical methods, and for this reason there is always a need for better and better estimates for the region containing all the zeros of a polynomial.So, there have been a number of theorems on computations for the roots of polynomials.We consider the generalization of well-known Fibonacci polynomials called R−Bonacci polynomials.The polynomial R n (x) is defined by the following recursive equation in [6] for any integer n and r ≥ 2 : where is the r−nomial coefficient.Recently, some results have been given on the zeros of R−Bonacci polynomials R n (x) when r = 2, 3 in [5].M. X. He, P. E. Ricci and D. Simon found interesting curves composing of the zeros of R−Bonacci polynomials in [12].(If we set r = 2, 3, then the classical Fibonacci polynomials and Tribonacci polynomials are obtained respectively.)Explicit forms have been given for the roots Fibonacci polynomials in [7].Since these forms have been used in many practical areas eg.Graph Energy, Stable Polynomials, Hand Printed Characters in [15], [16], [18] and [19], it is important to find the zeros of these polynomials.On the other hand, Moore and Prodinger examined the asymptotic behaviour of the maximal roots of Fibonacci-type polynomials in [2] and [9], respectively.Furthermore, the absolute values of complex zeros of Fibonacci-like polynomials have been investigated by Matyas in [8].
Unlike Fibonacci polynomials, explicit forms for the zeros of Tribonacci polynomials have not been found yet.Instead, zero attractors of these polynomials for r = 3 are determined by W. Goh, M. X.He and P. E.
Ricci in [11].The symmetric polynomials of the zeros of Tribonacci polynomials are found by M. X. He, D. Simon and P. E. Ricci in [3].These results are generalized for the zeros of R−Bonacci polynomials and their derivatives in [17].
The aim of this paper is to determine the desired region on the complex plane containing the zeros of Tribonacci polynomials.To support this region, we present numerical results for these polynomials to compare the results with the known regions.Afterwards, we develop a new algorithm to use in numerical calculations.

Bounds For Polynomials
There have been a number of theorems on computations for the roots of polynomials.These studies date back to the work of Cauchy [1].Let be a complex polynomial of degree n.All the zeros of the polynomial P (z) lie in the region where Recently, Diaz-Barrero have improved this region by describing two annuli containing all the zeros of a polynomial where the inner and outer radii are stated in terms of the well-known Fibonacci numbers, respectively [13].
Theorem 2.1.[13] Let Then, all its zeros lie in the ring shaped region where and For another annulus, they considered Lucas numbers which are defined by L n = L n−1 +L n−2 for n ≥ 0 with the initial conditions L 0 = 2, L 1 = 1 (see [5] and [10] for the basic properties of these number sequences).
Then its all zeros of P (z) lie in the annulus where and . (2.8)

Tribonacci polynomials
For r = 3 in (1.1), R−Bonacci polynomials are named Tribonacci polynomials, defined by the recurrence relation with T 0 (x) = 0, T 1 (x) = 1 and T 2 (x) = x 2 .Despite its complicate form, the coefficients of these polynomials have many interesting properties.It is known that Tribonacci zeros constitute 3-stars as shown below in Figure 1.
As seen in the above Figure 1, the zeros of these polynomial being invariant with rotation 2π/3 are separated into 3 sets.So, it may be interesting to find a smallest disc that contains the zeros of these polynomials since the explicit expressions for the zeros of these polynomials are not known.In order to examine a new disc containing the zeros of a given polynomial in (3.1), we have used the following identity where F n is the n−th Fibonacci number in [14].Recall that Fibonacci numbers are defined by F n = F n−1 + F n−2 for n ≥ 2 with the initial conditions F 0 = 0, F 1 = 1.This identity (3.2) can be found in [5].Now, we consider the following theorem.
Theorem 3.1.[14] Let P (z) = n k=0 a k z k (a k = 0, 0 ≤ k ≤ n) be a non-consant complex polynomial.Then all its zeros lie in the annulus where Now we give the following theorem.
Theorem 3.2.Let T n (x) be a Tribonacci polynomial with n ≡ k(mod 3).Then all its zeros lie where r 1 = F3 Fn+2−1 Proof.The result is obviously obtained when the r-nomial coefficients are written instead It will be immediately noticed that the areas containing the zeros of Tribonacci polynomials will be a disc or an annulus.After reaching this result for the zeros of these polynomials, it is numerically examined by comparing with the existing boundaries.When comparisons are made, one of the oldest boundaries and the latest bounds are selected.Below is a table giving inner, outer radius and the area for the zeros of these polynomials.And this shows that this bound is unusable for these polynomials.On the other hand, it is clear that our boundary( 3 ) gives better results in terms of both r 1 , r 2 and area, when compared to the bound of Cauchy, one of the oldest boundaries and the bound of Dalal and Govil( 2 ) which have been recently obtained.
Although not included in the table, these calculations have been numerically verified for the first 20 Tribonacci polynomials.Hence, the disk or annulus obtained by our boundary and containing the zeros of each Tribonacci polynomial will remain the smallest.This is the desired ideal situation.

Application
In this section, we develop an algorithm.At the beginning of the algorithm, all Tribonacci polynomials are generated and then the coefficients are taken to be used at any boundary.The algorithm makes it easy for us to make this computation for the Tribonacci Polynomial as many as we would like to.Furthermore, this algorithm can be developed and adapted for other boundary value problems.And, thanks to this algorithm, testing of the usability of any boundary can be done without needing long and complicated operations.

Figure 1 .
Figure 1.The Zeros of T 25 (x) ≤ |z| ≤ 140.30− 61839.30.03 ≤ |z| ≤ 31.15−3048.36|z|≤ 51 − 8171.28 0.03 ≤ |z| ≤ 30.02 − 2831.2Above, the upper bounds containing the zeros of the first ten Tribonacci polynomials are shown numerically.(Sincethezero of T 0 (x) , T 2 (x) is only 0 and T 1 (x) is a fixed polynomial, they are not included in this table.)Atfirstglance, it seems that Diaz-Barrero's boundary(1) is the most useful among others.But, it is clearly observed that actual area is gradually growing except for first few Tribonacci polynomials.

Table 1 .
The Algorithm