Study of Monotonicity of Trinomial Arcs M(p, k, r, n) when 1 <α<+

Similar documents
MATH 5510 Mathematical Models of Financial Derivatives. Topic 1 Risk neutral pricing principles under single-period securities models

Non replication of options

Essays on Some Combinatorial Optimization Problems with Interval Data

Prize offered for the solution of a dynamic blocking problem

Equivalence between Semimartingales and Itô Processes

arxiv: v1 [math.oc] 23 Dec 2010

4: SINGLE-PERIOD MARKET MODELS

3.2 No-arbitrage theory and risk neutral probability measure

Asymptotic Notation. Instructor: Laszlo Babai June 14, 2002

Application of an Interval Backward Finite Difference Method for Solving the One-Dimensional Heat Conduction Problem

3 Arbitrage pricing theory in discrete time.

CS364A: Algorithmic Game Theory Lecture #3: Myerson s Lemma

CHOICE THEORY, UTILITY FUNCTIONS AND RISK AVERSION

Comparing Allocations under Asymmetric Information: Coase Theorem Revisited

More On λ κ closed sets in generalized topological spaces

Euler Savary s Formula On Complex Plane C

Interpolation. 1 What is interpolation? 2 Why are we interested in this?

THE NUMBER OF UNARY CLONES CONTAINING THE PERMUTATIONS ON AN INFINITE SET

Palindromic Permutations and Generalized Smarandache Palindromic Permutations

On the Number of Permutations Avoiding a Given Pattern

TWO-PERIODIC TERNARY RECURRENCES AND THEIR BINET-FORMULA 1. INTRODUCTION

The Value of Information in Central-Place Foraging. Research Report

MAT25 LECTURE 10 NOTES. = a b. > 0, there exists N N such that if n N, then a n a < ɛ

OPTIMAL BLUFFING FREQUENCIES

arxiv: v1 [math.co] 31 Mar 2009

A note on the number of (k, l)-sum-free sets

A No-Arbitrage Theorem for Uncertain Stock Model

BINOMIAL TRANSFORMS OF QUADRAPELL SEQUENCES AND QUADRAPELL MATRIX SEQUENCES

Applied Mathematics Letters

Problem 1: Random variables, common distributions and the monopoly price

Richardson Extrapolation Techniques for the Pricing of American-style Options

Brouwer, A.E.; Koolen, J.H.

GUESSING MODELS IMPLY THE SINGULAR CARDINAL HYPOTHESIS arxiv: v1 [math.lo] 25 Mar 2019

An Application of Ramsey Theorem to Stopping Games

Yao s Minimax Principle

Some Bounds for the Singular Values of Matrices

Computational Independence

CHARACTERIZATION OF CLOSED CONVEX SUBSETS OF R n

Introduction to Probability Theory and Stochastic Processes for Finance Lecture Notes

A class of coherent risk measures based on one-sided moments

On the Lower Arbitrage Bound of American Contingent Claims

Permutation Factorizations and Prime Parking Functions

Lecture 5: Iterative Combinatorial Auctions

Chapter 5 Finite Difference Methods. Math6911 W07, HM Zhu

Haiyang Feng College of Management and Economics, Tianjin University, Tianjin , CHINA

4 Martingales in Discrete-Time

Browder-type theorems for direct sums of operators

On the h-vector of a Lattice Path Matroid

TEST 1 SOLUTIONS MATH 1002

On Existence of Equilibria. Bayesian Allocation-Mechanisms

Tug of War Game. William Gasarch and Nick Sovich and Paul Zimand. October 6, Abstract

Probability. An intro for calculus students P= Figure 1: A normal integral

SNELL S LAW AND UNIFORM REFRACTION. Contents

A relation on 132-avoiding permutation patterns

PURITY IN IDEAL LATTICES. Abstract.

Notes on the symmetric group

Quadrant marked mesh patterns in 123-avoiding permutations

DENSITY OF PERIODIC GEODESICS IN THE UNIT TANGENT BUNDLE OF A COMPACT HYPERBOLIC SURFACE

Square-Root Measurement for Ternary Coherent State Signal

Part 3: Trust-region methods for unconstrained optimization. Nick Gould (RAL)

Lecture Quantitative Finance Spring Term 2015

Laurence Boxer and Ismet KARACA

MTH6154 Financial Mathematics I Interest Rates and Present Value Analysis

Martingale Pricing Theory in Discrete-Time and Discrete-Space Models

1 Consumption and saving under uncertainty

Mixed Strategies. Samuel Alizon and Daniel Cownden February 4, 2009

Best-Reply Sets. Jonathan Weinstein Washington University in St. Louis. This version: May 2015

Realizability of n-vertex Graphs with Prescribed Vertex Connectivity, Edge Connectivity, Minimum Degree, and Maximum Degree

Effects of Wealth and Its Distribution on the Moral Hazard Problem

Optimal retention for a stop-loss reinsurance with incomplete information

Lecture 8: Asset pricing

Method of Characteristics

Alain Hertz 1 and Sacha Varone 2. Introduction A NOTE ON TREE REALIZATIONS OF MATRICES. RAIRO Operations Research Will be set by the publisher

Forecast Horizons for Production Planning with Stochastic Demand

CONSTRUCTION OF CODES BY LATTICE VALUED FUZZY SETS. 1. Introduction. Novi Sad J. Math. Vol. 35, No. 2, 2005,

B. Online Appendix. where ɛ may be arbitrarily chosen to satisfy 0 < ɛ < s 1 and s 1 is defined in (B1). This can be rewritten as

Hints on Some of the Exercises

Characterization of the Optimum

DRAFT. 1 exercise in state (S, t), π(s, t) = 0 do not exercise in state (S, t) Review of the Risk Neutral Stock Dynamics

Applied Mathematical Sciences, Vol. 8, 2014, no. 1, 1-12 HIKARI Ltd,

SHORT-TERM RELATIVE ARBITRAGE IN VOLATILITY-STABILIZED MARKETS

Lecture 4: Divide and Conquer

The ruin probabilities of a multidimensional perturbed risk model

The Limiting Distribution for the Number of Symbol Comparisons Used by QuickSort is Nondegenerate (Extended Abstract)

Sequential Investment, Hold-up, and Strategic Delay

In Discrete Time a Local Martingale is a Martingale under an Equivalent Probability Measure

Pricing Dynamic Solvency Insurance and Investment Fund Protection

1 Rational Expectations Equilibrium

Lecture 7: Bayesian approach to MAB - Gittins index

On a Manufacturing Capacity Problem in High-Tech Industry

Laurence Boxer and Ismet KARACA

A Preference Foundation for Fehr and Schmidt s Model. of Inequity Aversion 1

LECTURE 2: MULTIPERIOD MODELS AND TREES

Andreas Wagener University of Vienna. Abstract

On Forchheimer s Model of Dominant Firm Price Leadership

MANAGEMENT SCIENCE doi /mnsc ec pp. ec1 ec14

Supplemental Materials for What is the Optimal Trading Frequency in Financial Markets? Not for Publication. October 21, 2016

Best response cycles in perfect information games

No-arbitrage theorem for multi-factor uncertain stock model with floating interest rate

Handout 4: Deterministic Systems and the Shortest Path Problem

Transcription:

International Journal of Algebra, Vol. 1, 2007, no. 10, 477-485 Study of Monotonicity of Trinomial Arcs M(p, k, r, n) when 1 <α<+ Kaoutar Lamrini Uahabi F.A.R. Blvd., 49, Apartment N. 9 Nador 62000, Morocco Mohammed Zaoui Dept. of Mathematics, Faculty of Sciences, Mohamed first University P.O. Box 524, Oujda 60000, Morocco Abstract We are interested in this paper in the family of curves M(p, k, r, n), solutions of equation z n = αz k +(1 α), where z is a complex variable, n, k, p and r are nonzero integers and α is a real number greater than 1. Expressing each of these curves in polar coordinates (ρ, θ), we will prove that ρ (θ) is an increasing function. Mathematics Subject Classification: 03F65, 12D10, 14H45, 26C10, 30C15, 65H05 Keywords: curve, derivable function, feasible angles, increasing function, monotonicity, trinomial arc, trinomial equation 1 Introduction In [6], Fell studied the trinomial equation λz n + (1 λ) z k 1 = 0, (1) where z = ρe iθ is a complex variable, n and k are two integers such that 1 k n 1 and λ is a real number. She established especially a large description of the trajectories of roots of the trinomial equation (1), called trinomial arcs. These arcs are continuous curves, corresponding to a number λ which is restricted whether to [0, 1],orto[1, + [, or also to ], 0]. Each of these arcs can be expressed in polar coordinates (ρ, θ) by a function ρ (θ). The problem of monotonicity of the function ρ (θ) for the different families

478 K. Lamrini Uahabi and M. Zaoui of trinomial curves is pointed out in [6]. This problem is never solved. The descriptive results of Fell [6] gave us the information about the form and the localization of the trinomial arcs. However, these families of curves are not well-defined, in order to be studied. The original intent of this paper was to study the behavior of one of these different families. In fact, we consider in our work the equivalent trinomial equation z n = αz k + (1 α), (2) where z = ρe iθ is a complex variable, n is an integer larger than 1, k =1, 2,..., n 1 and α is a real number. We restrict our attention in this paper to a family of trinomial curves, denoted by M(p, k, r, n), solutions of equation (2) with 1 <α<+ and p, k, r and n are four integers satisfying some conditions. The rest of this paper is organized as follows. Section 2 contains some preliminary materiel concerning the expression of the trinomial equation (2) in polar coordinates. In section 3, we establish the description and the definition of the arcs M(p, k, r, n). Finally, we study in section 4 the monotonicity of the function ρ (θ) for these trinomial curves and we arrive at the main result stating that ρ (θ) is an increasing function for the arcs M(p, k, r, n). 2 Basic definitions and formulas The two trinomial equations (1) and (2) are equivalent. In fact, in order to pass from (2) to (1), we put α =1 1/λ. Indeed, substituting into equation (1) the expression given for z n by equation (2) yields [1 λ (1 α)] ( z k 1 ) =0. So, λ (1 α) = 1 or z k = 1. Because z is a complex number, we obtain α =1 1/λ. By this relation, we can say that the case 1 α<+ of equation (2) corresponds to the case <λ<0 of equation (1). Remark 2.1 The case α =1is one particular case for the trinomial equation (2). In fact, when α =1, this equation becomes z k [ z n k 1 ] =0. Then, the n roots of equation (2) are exactly the (n k) th roots of unity, which are simple roots and 0, a root of multiplicity k. In the rest of this paper, we will study the trinomial equation (2) with 1 < α<+. Putting z = ρe iθ in this equation and separating real and imaginary parts, we obtain ρ n sin nθ = αρ k sin kθ and ρ n cos nθ = αρ k cos kθ +(1 α). Then, if θ lπ/n, where l is an integer, one has ρ n k = α sin kθ / sin nθ. (3)

Study of monotonicity 479 On the other side, divide equation (2) by z n and take the imaginary part. When θ lπ/(n k) such that l is an integer, we get ρ k = (1 1/α) sin nθ / sin(n k)θ. (4) Lastly, we obtain the following equation for the trajectories of roots of (2): ρ n k sin nθ ρ n sin(n k)θ = sin kθ. (5) There is one basic remark : as 1 <α<+ and ρ>0, from (3) and (4), we are interested in those angles θ for which : sign (sin nθ) = sign (sin kθ) = sign (sin(n k)θ). (6) Definition 2.2 An angle θ which fulfills (6) will be called a (n, k)-feasible angle for the trinomial equation (2) with 1 <α<+. Remark 2.3 Because the upper and lower half-planes are symmetrical, we restrict our study of the trinomial arcs to the upper half-plane. 3 Trinomial arcs M(p, k, r, n) By Remark 2.1, when α =1,thenroots of equation (2) are the (n k) th roots of unity, which are simple roots and 0, a root of multiplicity k. In [6], Fell asserts that when α moves from 1 to +, the trajectories of these n roots are continuous arcs, such that some of these arcs are situated inside the unit disk and the others are outside the disk. In fact, inside the unit disk, she considers in [6] the following sets: A = {k th roots of unity}, B = {(n k) th roots of 1} and C = {n th roots of unity}. Let γ be in C and β be the unique nearest neighbor of γ in A B. We are interested in the case β A, β/ B. In [6], Fell tells us that in this situation, as α moves from + to 1, there is a trajectory starting at β such that ρ moves from 1 to 0, this trajectory is tangent to the ray θ = arg(β). Fell has showed also that, in this case, the feasible angles θ belong to intervals of length less than or equal to π/n and bounded by arg (γ) and arg (β), such that γ is an n th root of 1 and β is an k th root of unity. In this paper, we will consider and restrict our attention to the trinomial arcs such that the feasible angles θ belong to an interval of the form [(2r + 1)π/n, 2πp/k], where p and r are nonzero integers verifying some conditions. In fact, γ such that arg (γ) =(2r +1)π/n is an n th root of 1 and β such that arg (β) =2πp/k is an k th root of unity. These trinomial curves will be denoted by M(p, k, r, n). Remark that, when n = 2 and n = 3, the trajectories of roots of equation (2) with 1 <α<+ are linear. Moreover, when n = 4, the roots of (2) don t move in trajectories of the form M(p, k, r, n). We define so the family of trinomial arcs M(p, k, r, n) as follows :

480 K. Lamrini Uahabi and M. Zaoui Definition 3.1 If n is an integer greater than or equal to 5, som(p, k, r, n) is the set of roots of equation (2) with α>1 and the feasible angles belong to the interval [(2r +1)π/n, 2πp/k], where p and r are nonzero integers verifying r p and k is an integer such that pn/(r +1)<k<2pn/(2r +1). The trinomial curves M(p, k, r, n) are illustrated in the figure below. These arcs exist in view of the next lemma. Lemma 3.2 If n is an integer greater than or equal to 5, k is an integer such that 1 k n 1 and α>1, so in the trinomial equation (2) with pn/(r +1)<k<2pn/(2r +1), where p and r are nonzero integers verifying r p, any angle of the interval [(2r +1)π/n, 2πp/k] is feasible. Proof. We assume that the integer k satisfy pn/(r+1) < k < 2pn/(2r+1). First, notice that because 0 <k<n, the two nonzero integers p and r verify the condition 2p/(2r +1) 1, i. e. r p. Now, let be (2r +1)π/n < θ<2πp/k. We have (2r +1)π < nθ < 2πpn/k. Since pn/(r +1) <k,we obtain that 2πpn/k < 2(r +1)π and that sin nθ < 0. On the other side, we have (2r +1)πk/n < kθ < 2πp. From the inequality r p stems that (2p 1)n/(2r +1)< pn/(r +1)<k. So, (2p 1)π <(2r +1)πk/n and so sin kθ < 0. Finally, we have (2r + 1)π(1 k/n) < (n k)θ < 2πp(n/k 1). Because k<2pn/(2r + 1), then [2(r p)+1]π<(2r +1)π(1 k/n) and since pn/(r + 1) < k, so2πp(n/k 1) < [2(r p)+2]π. It follows that sin(n k)θ <0. Thus, the conditions (6) are fulfilled and the proof is achieved. Trinomial arcs M(p, k, r, n) inside the upper half unit disk Remark 3.3 From the proof of Lemma 3.2, we get sin nθ < 0, sin kθ < 0 and sin(n k)θ <0 for any angle θ in the interval ](2r +1)π/n, 2πp/k[.

Study of monotonicity 481 4 Behavior of the trinomial arcs M(p, k, r, n) In this section, our objective is to prove that for each trinomial arc M(p, k, r, n), ρ(θ) is a monotonic function, i. e. dρ/dθ is never zero. Thus, we begin by proving that the derivative dρ/dθ exists and it is well-defined. Proposition 4.1 ρ(θ) is a derivable function for each arc M(p, k, r, n). Proof. Consider a trinomial arc M(p, k, r, n). By equation (4), we have ρ k (θ) =(1 1/α) sin nθ/ sin(n k)θ. From Remark 3.3, the feasible angles θ are such that sin nθ < 0 and sin(n k)θ <0. Define the function g (θ) = (1 1/α) sin nθ/ sin(n k)θ on the interval ](2r + 1)π/n, 2πp/k[. So, as α > 1, the denominator of g (θ) is never 0. It follows that g (θ) is well-defined. In addition, g (θ) is derivable and positive. Hence, the function ρ(θ) =[g (θ)] 1/k is derivable. Therefore, its derivative dρ/dθ exists and it is well-defined. Now, Proposition 4.1 allows us to estimate dρ/dθ. In fact, differentiating both sides of equation (5) with respect to θ, we get [ (n k) ρ n k 1 sin nθ nρ n 1 sin(n k)θ ] dρ/dθ = k cos kθ +(n k) ρ n cos(n k)θ nρ n k cos nθ. Supposing that dρ/dθ = 0 and considering equation (5) of trajectories of roots, we will consider ρ n and ρ n k as solutions of the next system: { k cos kθ +(n k) ρ n cos(n k)θ nρ n k cos nθ = 0 ρ n k sin nθ ρ n sin(n k)θ sin kθ = 0. This system is equivalent to the following system : { R (θ) ρ n k = N 1 (θ) R (θ) ρ n = N 2 (θ), (7) where R(θ) = (n k) sin kθ k cos nθ sin(n k)θ N 1 (θ) = n sin kθ cos(n k)θ k sin nθ N 2 (θ) = (n k) sin kθ cos nθ k sin(n k)θ. Remark that N 1 (θ) and N 2 (θ) can be expressed as follows : N 1 (θ) =(n k) sin nθ n sin(n k)θ cos kθ and N 2 (θ) =(n k) sin nθ cos kθ n sin(n k)θ. Then, the difference of the two equalities of (7) leads to the relation: R(θ) [ρ n k ρ n ]=S(θ) [1 cos kθ], (8) with S(θ) = (n k) sin nθ + n sin(n k)θ.

482 K. Lamrini Uahabi and M. Zaoui Remark 4.2 By Remark 3.3, we can deduce, for the arcs M(p, k, r, n), that S(θ) < 0 for any feasible angle θ in the interval ](2r +1)π/n, 2πp/k[. Let us recall that our main interest is to prove that dρ/dθ is never zero, for each trinomial arc M(p, k, r, n). So, in what follows, the question is to show that the hypothesis dρ/dθ = 0 considered above is not possible for this family of arcs. For that, we will need the two following propositions and we will use the remark below. Remark 4.3 For any feasible angle θ in the interval ](2r +1)π/n, 2πp/k[, the sign of cos nθ is determined as follows : Case i): pn/(r + 1) < k < 4pn/(4r + 3). We have cos nθ = 0 on ](2r +1)π/n, 2πp/k[ if and only if θ =(4r +3)π/2n. Moreover, we have cos nθ < 0 for θ<(4r +3)π/2n and cos nθ > 0 for θ>(4r +3)π/2n. Case ii): 4pn/(4r +3) k < 2pn/(2r +1). In this situation, we have cos nθ < 0 for any feasible angle θ in the interval ](2r +1)π/n, 2πp/k[. Proposition 4.4 Let M(p, k, r, n) be a trinomial arc. For any integer k such that pn/(r +1)<k<2pn/(2r +1), we have N 2 (θ) > 0 for any feasible angle θ in the interval ](2r +1)π/n, 2πp/k[. Proof. Let θ be a feasible angle in ](2r +1)π/n, 2πp/k[. Recall that N 2 (θ) =(n k) sin kθ cos nθ k sin(n k)θ. From Remark 3.3, we get sin kθ < 0 and sin(n k)θ <0. Moreover, according to Remark 4.3, two cases are possible. So, in Case i), we distinguish the two following subcases : Subcase 1): θ belongs to ](2r + 1)π/n, (4r + 3)π/2n]. In this situation, we have cos nθ 0. So, it follows that N 2 (θ) > 0. Subcase 2): θ belongs to ](4r +3)π/2n, 2πp/k[. Since N 2 (θ) can be expressed as N 2 (θ) = n sin kθ cos nθ k sin nθ cos kθ, we will consider the function L 2 (θ) = N 2 (θ)/ cos nθ cos kθ = n tan kθ k tan nθ. According to Remark 4.3, in this subcase, we have cos nθ > 0. On the other side, we have (2r +3/2)πk/n < kθ < 2πp. Because the two integers p and r verify r p, we obtain that (2p 1/2)/(2r +3/2) <p/(r +1) <k/nand that (2p 1/2)π <(2r + 3/2)πk/n. So, we deduce that (2p 1/2)π < kθ < 2πp and that cos kθ > 0. Hence, the sign of N 2 (θ) is exactly the sign of L 2 (θ). For that, we study this last function on the interval ](4r + 3)π/2n, 2πp/k[. L 2 (θ) is derivable with L 2 (θ) =nk[tan2 kθ tan 2 nθ ]. As tan nθ < 0 and tan kθ < 0, the roots of the equation L 2 (θ) = 0 are those of the equation tan nθ = tan kθ. The unique root of this equation is of the form θ = lπ/(n k) where l is an integer. However, lπ/(n k) ](4r +3)π/2n, 2πp/k[ if and only if (2r +3/2)(1 k/n) <l<2p(n/k 1). Because k<4pn/(4r + 3), so 2(r p) +3/2 < (2r +3/2)(1 k/n) and because pn/(r +1) <k,so 2p(n/k 1) < 2(r p + 1). The integer l verify then 2(r p) +3/2 <l<

Study of monotonicity 483 2(r p) + 2, which is not possible. We deduce so that L 2 (θ) 0 for any θ in ](4r +3)π/2n, 2πp/k[. In addition, L 2 (θ) goesto+ as θ tends on the right to (4r +3)π/2n and L 2 (2πp/k) > 0. So, it follows that L 2 (θ) > 0 and that N 2 (θ) > 0 for any θ in ](4r +3)π/2n, 2πp/k[. On the other hand, Remark 4.3 tells us that in Case ii), we have cos nθ < 0 for any θ in ](2r +1)π/n, 2πp/k[. Then, N 2 (θ) > 0 for any θ in this interval. Consequently, for any integer k such that pn/(r + 1) < k < 2pn/(2r + 1), we proved that N 2 (θ) > 0 for any feasible angle θ in ](2r +1)π/n, 2πp/k[. Proposition 4.5 Let M(p, k, r, n) be a trinomial arc. For any integer k such that pn/(r +1)< k < 2pn/(2r +1), we have the following results : -Ifpn/(r+1) <k<4pn/(4r+3), so there exists θ 0 in ](2r +1)π/n, 2πp/k[ such that R (θ 0 )=0. Moreover, R(θ) < 0 for θ<θ 0 and R(θ) > 0 for θ>θ 0. -If4pn/(4r +3) k<2pn/(2r +1),soR(θ) < 0 for any angle θ in the interval ](2r +1)π/n, 2πp/k[. Proof. Let θ be a feasible angle in ](2r + 1)π/n, 2πp/k[. Let us recall that R(θ) =(n k) sin kθ k cos nθ sin(n k)θ. We know by Remark 3.3 that sin kθ < 0 and sin(n k)θ <0. Then, in view of Remark 4.3, we distinguish in Case i) the two subcases: Subcase 1): θ belongs to ](2r +1)π/n, (4r +3)π/2n]. Because cos nθ 0 in this subcase, we deduce that R(θ) < 0. Subcase 2): θ belongs to ](4r + 3)π/2n, 2πp/k[. Observing that R (θ) can be expressed as R(θ) =(n k) sin nθ cos (n k) θ n sin(n k)θ cos nθ, we will define K (θ) = R(θ)/ cos nθ cos(n k)θ =(n k) tan nθ n tan(n k)θ. In this situation, we have cos nθ > 0. In addition, we have (2r +3/2)π(1 k/n) < (n k)θ <2πp(n/k 1). Since k<4pn/(4r + 3), so [2(r p)+3/2] π< (2r +3/2)π(1 k/n) and as pn/(r +1) <k,so2πp(n/k 1) < [2(r p)+2]π. It follows that [2(r p)+3/2] π < (n k)θ < [2(r p) +2]π and that cos(n k)θ >0. Thus, R(θ) has the same sign as K (θ). Because tan nθ < 0 and tan(n k)θ <0, the zeros of K (θ) =n(n k) [tan 2 nθ tan 2 (n k)θ] are those of the equation tan nθ = tan(n k)θ. The unique solution of this last equation is of the form θ = lπ/k with l is an integer. However, lπ/k ](4r +3)π/2n, 2πp/k[ if and only if (2r +3/2)k/n < l < 2p. As pn/(r + 1)< k,so(2r + 3/2)p/(r + 1)< (2r + 3/2)k/n and because the two integers p and r verify r p, we get (2p 1/2) < (2r +3/2)p/(r + 1). We obtain so that (2p 1/2) <l<2p, which is impossible as l is an integer. Then, we conclude that K (θ) is never zero. We have also that K(θ) goesto as θ tends on the right to (4r +3)π/2n and K(2πp/k) > 0. Thus, there exists θ 0 in the interval ](4r +3)π/2n, 2πp/k[ such that R(θ 0 ) = 0. Moreover, we deduce that R(θ) < 0 for θ<θ 0 and that R(θ) > 0 for θ>θ 0. Consequently, in Case i), i. e. if pn/(r + 1) < k < 4pn/(4r + 3), there exists θ 0 in the interval ](2r +1)π/n, 2πp/k[ such that R(θ 0 ) = 0. Moreover,

484 K. Lamrini Uahabi and M. Zaoui we obtain that R(θ) < 0 for θ<θ 0 and that R(θ) > 0 for θ>θ 0. Finally, from Remark 4.3, we have in Case ii) that cos nθ < 0 for any θ in ](2r +1)π/n, 2πp/k[. So, R(θ) < 0 for any angle θ in this interval. Now, the two propositions above allow us to state the next main result for the trinomial curves M(p, k, r, n). Theorem 4.6 For the trinomial curves M(p, k, r, n), the function ρ(θ) is monotonic on the interval [(2r +1)π/n, 2πp/k]. Proof. Assume that θ is a feasible angle in the interval ](2r + 1)π/n, 2πp/k[. Proposition 4.5 and Remark 4.3 tell us that in Case i), there exists θ 0 in this interval such that R(θ 0 ) = 0. So, we can distinguish the two following subcases: Subcase 1): θ belongs to ](2r +1)π/n, θ 0 ]. From Propositions 4.4 and 4.5, we have respectively that N 2 (θ) > 0 and R(θ) 0. Then, the equation R(θ) ρ n = N 2 (θ) given by (7) is not possible, because ρ>0. Subcase 2): θ belongs to ]θ 0, 2πp/k[. On the one hand, we have R(θ) > 0 by Proposition 4.5. On the other hand, we have S(θ) < 0 by Remark 4.2. Then, the relation R(θ) [ρ n k ρ n ]=S(θ) [1 cos kθ] given by (8) implies that ρ n k ρ n < 0, which is impossible because 0 <ρ<1. Hence, in Case i), the hypothesis dρ/dθ = 0 is false. On the other side, Proposition 4.5 tells us that in Case ii), we have R(θ) < 0 for any θ in the interval ](2r +1)π/n, 2πp/k[. Since N 2 (θ) > 0 by Proposition 4.4, then, the fact that ρ>0 contradicts the equation R(θ) ρ n = N 2 (θ) given by (7). Consequently, in the two cases i) and ii), the hypothesis dρ/dθ = 0 is not possible. Thus, we deduce that dρ/dθ is never 0, i.e. ρ (θ) is monotonic on the interval ](2r + 1)π/n, 2πp/k[. Thus, we achieve the proof. Remark 4.7 When k = pn/(r +1), we have 2πp/k =2(r +1)π/n. By using the notations of Fell [6], we can tell that 2(r +1)π/n = arg(γ), such that γ belongs to C = {n th roots of unity} and that 2πp/k = arg(β), such that β belongs to A = {k th roots of unity}. According to Fell [6], because arg(β) = arg(γ), the particular case k = pn/(r +1) corresponds to a linear trinomial arc inside the unit disk. Remark 4.8 When k = 2pn/(2r +1), we have (2r +1)π/n = 2πp/k. Then, in this case, the trinomial arc M(p, k, r, n) is such that the bounds of the interval of feasible angles are identical. So, the particular case k =2pn/(2r+1) corresponds to a linear trinomial arc inside the unit disk. Lastly, Theorem 4.6 allows us to prove the following main result. Theorem 4.9 Let M(p, k, r, n) be a trinomial arc. For any integer k such that pn/(r +1) < k < 2pn/(2r +1), ρ(θ) is an increasing function on the interval [(2r +1)π/n, 2πp/k].

Study of monotonicity 485 Proof. Let M(p, k, r, n) be a trinomial arc. From Theorem 4.6, the function ρ(θ) is monotonic on the interval [(2r + 1)π/n, 2πp/k]. Let us estimate ρ(θ) at the bound (2r +1)π/n. So, put θ =(2r +1)π/n in the equation ρ k (θ) = (1 1/α) sin nθ/ sin(n k)θ given by (4). Because (2p 1)π < (2r +1)πk/n < 2πp, we get sin [(2r +1)πk/n] 0. Moreover, as α>1, we obtain that ρ [(2r +1)π/n] = 0. Therefore, since ρ (θ) changes between 0 and 1, it follows that ρ(θ) is an increasing function on the interval [(2r +1)π/n, 2πp/k]. References [1] J. Dieudonné, Sur quelques points de la théorie des polynômes, Bull. Sci. Math. (2), 58 (1934), 273-296. [2] K. Dilcher, J. Nulton and K.B. Stolarsky, The zeros of certain family of trinomials, Glasgow Math. J., 34, N. 1 (1992), 55-74. [3] D.Z. Djokovic, Cyclic polygons, roots of polynomials with decreasing nonnegative coefficients, and eigenvalues of stochastic matrices, Linear Algebra Appl., 142 (1990), 173-193. [4] S. Dubuc and A. Malik, Convex hull of powers of a complex number, trinomial equations and the Farey sequence, Num. Algorithms, 2 (1992), 1-32. [5] S. Dubuc and M. Zaoui, Sur la quasi-convexité des arcs trinomiaux, Rendiconti del Circolo Matematico di Palermo, Serie II, Tomo XLV (1996), 493-514. [6] H. Fell, The geometry of zeros of trinomial equations, Rendiconti del Circolo Matematico di Palermo, Serie II, Tomo XXIX (1980), 303-336. [7] F. Pécastaings and J. Sevin, Chemins vers la géométrie, Edition Vuibert, Paris, 1981. [8] K. Toda, On the location of the roots of linear combinations of some polynomials, J. Sci. Hiroschima Univ., 7 (1937), 144-150. Received: May 7, 2007