On the location of ratios of zeros of special trinomials

2025-04-27 0 0 371.27KB 17 页 10玖币
侵权投诉
On the location of ratios of zeros of special trinomials
A. S. Bamunobaa,1, I. Ndikubwayoa,1,
aMakerere University, Department of Mathematics, Kampala, 256, Uganda
Abstract
Given coprime integers k, ℓ with k> ℓ 1 and arbitrary complex polynomials A(z),B(z) with deg(A(z)B(z)) 1,
we consider the polynomial sequence {Pn(z)}satisfying a three-term recurrence Pn(z)+B(z)Pn(z)+A(z)Pnk(z)=0
subject to the initial conditions P0(z)=1, P1(z)=··· =P1k(z)=0 and fully characterize the real algebraic curve Γ
on which the zeros of the polynomials in {Pn(z)}lie. In addition, we show that, for any (randomly chosen) nZ1
and zero z0of Pn(z) with A(z0),0, at-least two of the distinct zeros of the trinomial D(t;z0) :=A(z0)tk+B(z0)t+1
have a ratio that lies on the real line and /or on the unit circle centred at the origin. This reveals a previously unknown
geometric property exhibited by the zeros of trinomials of the form tk+at+1 where aC{0}is such that akR.
Keywords: trinomial, null-collinearity, q-discriminant, real-rootedness, reciprocal polynomials
2010 MSC: Primary 12D10, 30C15, Secondary 26C10, 30C10
1. Basic notions and main result
Recursively defined polynomials have always been a subject of interest since they can be used to explain phe-
nomena frequently occurring in mathematics, statistics, physics and engineering. For example, (recursive) poly-
nomial sequences arise in physics and approximation theory as the solutions of certain ordinary dierential equations.
In particular, the Hermite polynomials are used in quantum and statistical mechanics to describe solutions of the
Schr¨
odinger equation for a harmonic oscillator while Laguerre polynomials are used to describe the eigenfunctions
for the Schr¨
odinger operator associated with the hydrogen atom, [1]. In combinatorics, these recursive polynomials
can be shown to represent certain combinatorial objects like graphs and therefore can be used to come up with new
identities and generating functions, the validity of which can be proved using the combinatorial interpretation.
Some of the above mentioned polynomial sequences are examples of sequences of orthogonal polynomials satisfy-
ing three-term recursive formulae. These three-term recursion formulae are famous since they provide a necessary
condition for polynomial sequences to be orthogonal, a highly sought for property in functional representations and
Corresponding author
Email addresses: alex.bamunoba@mak.ac.ug (A. S. Bamunoba), innocent.ndikubwayo@mak.ac.ug (I. Ndikubwayo)
1This research was funded by the NORHED-II project “Mathematics for Sustainable Development (MATH4SD), 2021-2026” at Makerere
University in collaboration with the University of Dar es Salaam and University of Bergen in Norway.
Preprint submitted to Quaestiones Mathematicae November 8, 2024
arXiv:2210.06403v3 [math.CV] 7 Nov 2024
numerical computations. Typically, orthogonal polynomials are used as basis functions in which to expand other
more complicated functions to be used in either interpolation, approximation or numerical quadrature. In addition,
interesting properties such as real-rootedness of the generated polynomials can be deduced from the orthogonality
property, for details [2]. Now, it is almost a tradition in mathematics that, whenever one encounters a family of poly-
nomials (preferably generated by recursions), it is of interest to study properties such as orthogonality, unimodularity,
real-rootedness, stability or interlacing roots among others, see [3] for further reading.
In recent years however, several authors for example [4, 5, 6, 7, 8, 9] have interested themselves in the problem of the
location of zeros of polynomials in polynomial sequences, especially those that are generated by linear recurrences.
This is partly because, knowing the zeros of some polynomials in the polynomial sequence may give information
concerning the zeros of other polynomials in the same sequence. In all of these studies, the authors either search for
a criterion for real-rootedness of the polynomial sequence (for applications in numerical analysis and combinatorics)
or explicit determination of the limiting curve on which the zeros of the polynomials lie, see for example [4, 5, 7, 8].
In this setting, one considers arbitrary non-zero complex polynomials A(z),B(z) with deg(A(z)B(z)) 1, two coprime
integers k, ℓ with k> ℓ 1 and the polynomial sequence {Pn(z)}satisfying a three-term recurrence
Pn(z)+B(z)Pn(z)+A(z)Pnk(z)=0,(1)
subject to the initial conditions P0(z)=1, P1(z)=··· =P1k(z)=0. If we take a solution of the form Pn(z)=(t(z))n
where nZ0and t(z) is a nonzero complex rational function, then substituting it in the recurrence (1) yields
0=(t(z))n=B(z)(t(z))n+A(z)(t(z))nk=((t(z))k+B(z)(t(z))k+A(z))(t(z))nk.
Since t(z).0, it follows that, (t(z))k+B(z)(t(z))k+A(z)=0 except probably for finitely many zC. The equation
tk+B(z)tk+A(z)=0 (with the variable zof tdropped) is called the characteristic equation of recurrence in (1),
while the polynomial (t;z) :=tk+B(z)tk+A(z) is called the characteristic polynomial of recurrence in (1). In [5,
Lemma 1], the ordinary generating function of the polynomial sequence {Pn(z)}generated by (1) is shown to be
G(t;z) :=
X
n=0
Pn(z)tn=1
1+B(z)t+A(z)tk.
We shall denote the denominator of G(t;z) by D(t;z) :=A(z)tk+B(z)t+1, the reciprocal polynomial of (t;z).
Remark 1.1. We consider k and to be coprime since the case for non-coprime integers, does not provide any new
information about the distribution of the zeros of non-zero polynomials in its sequence. To see this, we take r,sZ1
with r =d, s =kd, gcd(k, ℓ)=1and k > ℓ 1. The ordinary generating function GR(t;z)for the polynomial
sequence {Rm(z)}generated by the three-term recurrence Rn(z)+B(z)Rnr(z)+A(z)Rns(z)=0, subject to the initial
conditions R0(z)=1, R1(z)=··· =R1s(z)=0is related to G(t;z)as follows:
GR(t;z)=
X
m=0
Rm(z)tm=1
1+B(z)tr+A(z)ts=1
1+B(z)wd+A(z)wkd =
X
n=0
Pn(z)tnd =G(td;z).
2
Therefore, Rm(z)=Pn(z)if m =nd and Rm(z)=0if d m, i.e., the sequence {Rm(z)}is “essentially” {Pn(z)}.
Of all the results in the literature regarding the location of (the algebraic curve containing) the zeros of {Pn(z)}, the
most general and crucial to us is [4, Theorem 1.1] due to B¨
ogvad, et. al. To state it, we need the following notation:
For a non-empty subset Xof nonzero complex polynomials, we set ZX, (or ZA(z)if X={A(z)}) to be the union of the
set of zeros of all the polynomials in X, i.e., ZX:={zC:f(z)=0 for some f∈ X} =[
f(z)∈X Zf(z).
Theorem 1.2 ([4, Theorem 1.1]).If {Pn(z)}is the polynomial sequence generated by (1) and z0∈ Z{Pn(z)}− ZA(z),
then z0lies on the real algebraic curve Γ:=(zC: Im (1)kBk(z)
A(z)!=0).
Remark 1.3. .
1. Theorem 1.2 generalizes the specific cases k =2,3and 4with =1proved in [7, Theorems 1, 3, 5] respectively.
2. The zeros of Pn(z)become dense in H:= Γ nzC: 0 Re (1)kBk(z)
A(z)!<kk
(k)koas n tends to infinity
for the cases (k, ℓ)∈ {(2,1),(3,1),(4,1)}, as proved in [7, Theorems 1, 3, 5]. For k 5and =1, Tran
established the density of zeros of Pn(z)in Hfor suciently large n, see [8, Theorem 1] for details.
3. The case ℓ > 1is of a dierent flavour, (see Theorem 1.4)and will be established in Subsection 2.2.
Theorem 1.2 gives a partial characterisation of the real algebraic curve on which the zeros of the generated polynomials
lie. This motivated the authors to search for the full characterisation of the curve Γ, as stated in Theorem 1.4.
Theorem 1.4. If {Pn(z)}is the polynomial sequence generated by (1) with ℓ > 1and z0∈ Z{Pn(z)}− ZA(z), then z0lies
on the curve
Γ1:=
nzΓ: Re (1)kBk(z)
A(z)0o,where Z{B(z)}− ZA(z),
nzΓ: Re (1)kBk(z)
A(z)<0o,where Z{B(z)}⊂ ZA(z)
Another problem that we deal with in the present paper concerns the location of zeros of trinomials. In general, the
problem of the number and location of zeros of an arbitrary trinomial atk+bt+c(where a,b,cC{0}and k, ℓ Z
such that k3 and k> ℓ 1) has a long history of study starting with J. Lambert in 1758, followed by L. Euler
in 1777 and many others to the present, see [10, 11] and the references therein. Of all the results in the literature
about this problem, we are interested in those that take the form of bounds on the magnitude of the zeros or of sectors
in the complex plane containing the zeros. These stem from J. Egerv`
ary’s observation that, the zeros of trinomials
can be interpreted as the equilibrium points of a force field created by unit masses that are located at the vertices of
two regular concentric polygons centered at the origin in the complex plane. Utilising the symmetry and continuity
properties of this force field, the roots of a trinomial can be separated according to their argument and moduli, see [12]
for details. From these results, A. Melman in 2012 obtained finer results involving smaller annular sectors containing
the zeros of a trinomial that take into account the magnitude of the coecients. For further details, see [10].
3
Most of the approaches in the above studies could be characterised as being algebraic and/or geometric. However,
most recently, in 2019, D. Belki´
c employed some analytic tools that primarily focus upon derivations of the analytical
formulae for all the roots of trinomials through series developments using the Bell polynomials and the Fox-Wright
function. These are based on the fact, that all the roots of the general nth degree trinomial admit certain convenient
representations in terms of the Lambert and Euler series for the asymmetric and symmetric cases of the trinomial
equation, respectively. As an application of their work, these analytical solutions are numerically illustrated in the
genome multiplicity corrections for survival of synchronous cell populations after irradiation, see [11] for details.
Despite the above eorts, the problem of locating the zeros of a general trinomial is still unresolved. In the current
paper, we take a dierent approach and instead study the location of ratios of zeros of trinomials. In particular, we
choose an arbitrary nZ1, then a z0∈ ZPn(z)− ZA(z)and study the location of ratios of the zeros of the trinomil
D(t;z0)=A(z0)tk+B(z0)t+1. We obtain the following results, namely: Theorem 1.5 and Corollary 1.6.
Theorem 1.5 (Main result).If {Pn(z)}is the polynomial sequence generated by (1) and z0∈ Z{Pn(z)}− ZA(z), then
there exists at-least two zeros of D(t;z0)=A(z0)tk+B(z0)t+1whose ratio is real and /or has modulus 1.
Corollary 1.6. If {Pn(z)}is the polynomial sequence generated by (1) and z0∈ Z{Pn(z)}− ZA(z), then D(t;z0)has
at-least two equimodular zeros and /or exactly three “null-collinear”2zeros for k odd, or two null-collinear zeros for
k even.
The remaining sections of the paper are devoted to proving Theorem 1.4 and Theorem 1.5 in reverse order.
2. Proofs
2.1. Some lemmata and important results
Let C:={zC:|z|=1}and h:XC,w7→ (1 wk)k
(1 w)(wwk)kwhere Xis the domain of h.
Lemma 2.1. If z X(R∪ C), then h(z)R.
Proof. The domain of his X=C(µµk∪ {0}) where µndenotes the set of nth complex zeros of unity. Now,
there are two cases to consider, namely:
(i) if zXR, then zR, hence h(z)Ras his a quotient of two polynomials with real coecients, i.e., hR(w).
(ii) if zX∩ C, then h(z)Rby [5, Lemma 2].
2Null-collinear means two points (for keven) or three (for kodd) points lie on a line via the origin Owith at-least one point on either side of O.
4
摘要:

OnthelocationofratiosofzerosofspecialtrinomialsA.S.Bamunobaa,1,I.Ndikubwayoa,1,∗aMakerereUniversity,DepartmentofMathematics,Kampala,256,UgandaAbstractGivencoprimeintegersk,ℓwithk>ℓ⩾1andarbitrarycomplexpolynomialsA(z),B(z)withdeg(A(z)B(z))⩾1,weconsiderthepolynomialsequence{Pn(z)}satisfyingathree-term...

展开>> 收起<<
On the location of ratios of zeros of special trinomials.pdf

共17页,预览4页

还剩页未读, 继续阅读

声明:本站为文档C2C交易模式,即用户上传的文档直接被用户下载,本站只是中间服务平台,本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私,请立即通知玖贝云文库,我们立即给予删除!
分类:图书资源 价格:10玖币 属性:17 页 大小:371.27KB 格式:PDF 时间:2025-04-27

开通VIP享超值会员特权

  • 多端同步记录
  • 高速下载文档
  • 免费文档工具
  • 分享文档赚钱
  • 每日登录抽奖
  • 优质衍生服务
/ 17
客服
关注