betaFeedback

WorksMathematics

Improved criteria for geometric properties of confluent hypergeometric functions

Abstract

In this paper, we establish sufficient conditions for several geometric properties of confluent hypergeometric functions. Using differential subordination techniques, we derive parameter conditions ensuring lower bounds for the real part of the confluent hypergeometric function and sectorial subordination of its derivative. We also obtain criteria for strong convexity and strong starlikeness of prescribed order. In addition, we investigate the close-to-convexity and starlikeness of the confluent hypergeometric functions. Particular cases yield corresponding results for generalized Laguerre polynomials. The obtained criteria are compared with existing sufficient conditions, and their implications for the geometric theory of special functions are discussed.

Introduction

The confluent hypergeometric function of the first kind, also called Kummer’s function, is defined by

Φ(a;c;z)=1F1(a;c;z)=∑k=0∞(a)k(c)kzkk!(1)\Phi(a;c;z)={}_1F_1(a;c;z)=\sum_{k=0}^{\infty}\frac{(a)_k}{(c)_k}\frac{z^k}{k!} \tag*{(1)}

where a,c∈Ca,c \in\mathbb{C}, c≠0,−1,−2,…c \ne0,-1,-2,\ldots, and (a)n(a)_n denotes the Pochhammer symbol. The series converges for every z∈Cz \in\mathbb{C}. The function Φ(a;c;z)\Phi(a;c;z) satisfies Kummer’s differential equation

zΦ′′(z)+(c−z)Φ′(z)−aΦ(z)=0(2)z\Phi^{\prime\prime}(z)+(c-z)\Phi^{\prime}(z)-a\Phi(z)=0 \tag*{(2)}

Several classical special functions admit representations in terms of Φ\Phi, including Laguerre polynomials, Whittaker functions, and modified Bessel functions.

The geometric properties of confluent hypergeometric functions have been investigated using differential subordination and coefficient techniques. Merkes and Scott [12] studied starlikeness properties of hypergeometric functions, while Miller and Mocanu [14] derived sufficient conditions for confluent hypergeometric functions to have positive real part, starlike, and convex in Δ\Delta. Ponnusamy and Vuorinen [15] obtained parameter conditions for univalence, convexity, and close-to-convexity. Bohra and Ravichandran [3] investigated strong starlikeness and strong convexity. Confluent hypergeometric functions also arise in problems from mathematical physics, engineering, and applied sciences; see [1, 2, 9, 10, 11, 19]. Further results on their geometric properties and related hypergeometric functions can be found in [5, 15, 18].

For analytic functions ff and gg in Δ\Delta, we say that ff is subordinate to gg, written as f(z)≺g(z)f(z) \prec g(z), z∈Δz \in\Delta, if there exists a Schwarz function ω\omega (analytic in Δ\Delta, ω(0)=0\omega(0) = 0, ∣ω(z)∣<1\lvert\omega(z)\rvert< 1) such that f(z)=g(ω(z))f(z) = g(\omega(z)). If gg is univalent, then f(z)≺g(z)f(z) \prec g(z), z∈Δz \in\Delta is equivalent to f(0)=g(0)f(0) = g(0) and f(Δ)⊂g(Δ)f(\Delta) \subset g(\Delta).

Let A\mathcal{A} denote the class of analytic functions ff in Δ\Delta normalized by f(0)=0f(0) = 0 and f′(0)=1f'(0) = 1. The subclass of univalent functions in A\mathcal{A} is denoted by S\mathcal{S}. A function f∈Af \in\mathcal{A} is called starlike if Re⁡(zf′(z)/f(z))>0\operatorname{Re}(zf'(z)/f(z)) > 0, z∈Δz \in\Delta. The class of such functions is denoted by S∗\mathcal{S}^{*}. Similarly, f∈Af \in\mathcal{A} is called convex if Re⁡(1+zf′′(z)/f′(z))>0\operatorname{Re}(1 + zf''(z)/f'(z)) > 0, z∈Δz \in\Delta. We denote the class of convex functions by K\mathcal{K}. By Alexander’s theorem f∈K⇔zf′(z)∈S∗f \in\mathcal{K} \Leftrightarrow zf'(z) \in\mathcal{S}^{*}. A function f∈Af \in\mathcal{A} is called close-to-convex if there exist a convex function g∈Ag \in\mathcal{A} and a real number θ\theta such that Re⁡(eiθf′(z)/g′(z))>0\operatorname{Re}(e^{i\theta}f'(z)/g'(z)) > 0. The class of close-to-convex functions is denoted by C\mathcal{C}. Every close-to-convex function is univalent, but the converse is not generally true. Further details concerning these classes can be found in [6, 8].

For 0<β≤10 < \beta\le1, a function f∈Af \in\mathcal{A} is said to be strongly starlike of order β\beta if

∣arg⁡(zf′(z)f(z))∣<π2β,z∈Δ,\left|\arg\left(\frac{zf'(z)}{f(z)}\right)\right| < \frac{\pi}{2}\beta,\qquad z \in\Delta,

Equivalently,

zf′(z)f(z)≺(1+z1−z)β,z∈Δ,\frac{zf'(z)}{f(z)} \prec\left(\frac{1+z}{1-z}\right)^{\beta},\qquad z \in\Delta,

where the principal branch is understood. The class of strongly starlike functions of order β\beta, denoted by SS∗(β)\mathcal{SS}^{*}(\beta), was introduced independently by Stankiewicz [17] and Brannan and Kirvan [4]. In particular, SS∗(1)=S∗\mathcal{SS}^{*}(1) = \mathcal{S}^{*}. A function f∈Af \in\mathcal{A} is said to be strongly convex of order β\beta, where 0<β≤10 < \beta\le1, if zf′(z)zf'(z) is strongly starlike of order β\beta. Equivalently,

∣arg⁡(1+zf′′(z)f′(z))∣<πβ2,z∈Δ.\left|\arg\left(1+\frac{zf''(z)}{f'(z)}\right)\right| < \frac{\pi\beta}{2},\qquad z \in\Delta.

The corresponding class is denoted by SK(β)\mathcal{SK}(\beta). In particular, SK(1)=K\mathcal{SK}(1) = \mathcal{K}.

The main objective of this article is to establish sufficient conditions for several geometric properties of confluent hypergeometric functions. We first establish lower bounds for the real part of Φ(a;c;z)\Phi(a;c;z), subordination of its derivative, and strong convexity and strong starlikeness of prescribed order. We then investigate the close-to-convexity condition for Φ(a;c;zm)\Phi(a;c;z^{m}) and starlikeness of zΦ(a;c;z)z\Phi(a;c;z) through monotonicity properties of its coefficients. The results are compared with earlier results, and applications to related special functions are developed.

We close this section with lemmas that will be used to prove several theorems.

Lemma 1.1 ([14]). Let Λ⊂C\Lambda\subset\mathbb{C} and let a function Ψ:C3×Δ→C\Psi: \mathbb{C}^{3} \times\Delta\to\mathbb{C} satisfy the condition Ψ(is,t,u+iv;z)∉Λ\Psi(is,t,u+iv;z) \notin\Lambda, for z∈Δz \in\Delta and for real ss, tt, uu, vv satisfying t≤−(1+s2)/2t \le-(1+s^{2})/2 and t+u≤0t+u \le0. If pp is analytic in Δ\Delta, with p(0)=1p(0) = 1 and Ψ(p(z),zp′(z),z2p′′(z);z)∈Λ\Psi(p(z),zp'(z),z^{2}p''(z);z) \in\Lambda, z∈Δz \in\Delta, then Re⁡(p(z))>0\operatorname{Re}(p(z)) > 0 in Δ\Delta.

It may be noted that if we only have Ψ:C2×Δ→C\Psi: \mathbb{C}^{2} \times\Delta\to\mathbb{C}, then the condition Ψ(is,t;z)∉Λ\Psi(is,t;z) \notin\Lambda of Lemma 1.1 simplifies to Ψ(is,t;z)∉Λ\Psi(is,t;z) \notin\Lambda, for z∈Δz \in\Delta and for real ss, tt with t≤−(1+s2)/2t \le-(1+s^{2})/2.

Lemma 1.2 ([7]). Let f(z)=z+∑n=2∞Anznf(z)=z+\sum_{n=2}^{\infty} A_{n}z^{n}, with A1=1A_{1}=1. If An≥0A_{n}\ge0, {nAn}n≥1\{nA_{n}\}_{n\ge1} and {nAn−(n+1)An+1}n≥1\{nA_{n}-(n+1)A_{n+1}\}_{n\ge1} both are non-increasing, then f∈S∗f\in\mathcal{S}^{*}.

Main Results

We first establish the following extension of [15 Theorem 1.9].

Theorem 2.1 If aa, cc, β\beta are real such that a≠0a\ne0 and

c≥{1+1+(a−β(a+1))21−2β,0≤β<12,1+∣a∣β1−β+∣1−aβ1−β∣,β≥12, β≠1.c \ge\begin{cases} 1+\sqrt{1+\dfrac{(a-\beta(a+1))^{2}}{1-2\beta}}, & 0\le\beta<\dfrac{1}{2},\\ 1+\dfrac{|a|\beta}{1-\beta}+\left|1-\dfrac{a\beta}{1-\beta}\right|, & \beta\ge\dfrac{1}{2},\ \beta\ne1. \end{cases}

Then Re⁡Φ(a;c;z)>β\operatorname{Re}\Phi(a;c;z)>\beta for all z∈Δz\in\Delta.

Proof. Let p:Δ→Cp:\Delta\to\mathbb{C} be given by

p(z)=Φ(a;c;z)−β1−β,(β≠1, z∈Δ).(3)p(z)=\frac{\Phi(a;c;z)-\beta}{1-\beta},\qquad(\beta\ne1,\ z\in\Delta). \tag*{(3)}

Then pp is analytic in Δ\Delta and p(0)=1p(0)=1. As Φ(a;c;z)\Phi(a;c;z) satisfies the differential equation (2), we get

z2p′′(z)+(c−z)zp′(z)−az(p(z)+β1−β)=0.(4)z^{2}p^{\prime\prime}(z)+(c-z)zp^{\prime}(z)-az\left(p(z)+\frac{\beta}{1-\beta}\right)=0. \tag*{(4)}

Let

Ψ(w1,w2,w3;z)=w3+(c−z)w2−az(w1+β1−β).\Psi(w_{1},w_{2},w_{3};z)=w_{3}+(c-z)w_{2}-az\left(w_{1}+\frac{\beta}{1-\beta}\right).

then (4) can be written as Ψ(p(z),zp′(z),z2p′′(z);z)∈Λ\Psi(p(z),zp^{\prime}(z),z^{2}p^{\prime\prime}(z);z)\in\Lambda, where Λ={0}\Lambda=\{0\}. By Lemma 1.1, Re⁡p(z)>0\operatorname{Re}p(z)>0 if Ψ(is,t,u+iv;z)∉Λ\Psi(is,t,u+iv;z)\notin\Lambda, where ss, tt, uu, v∈Rv\in\mathbb{R}, t≤−(1+s2)/2t\le-(1+s^{2})/2, and t+u≤0t+u\le0 for z∈Δz\in\Delta. We prove a stronger statement that Re⁡Ψ(is,t,u+iv;z)<0\operatorname{Re}\Psi(is,t,u+iv;z)<0, which eventually leads to the required condition. Considering ss, tt, uu, v∈Rv\in\mathbb{R}, z=x+iy∈Δz=x+iy\in\Delta, t≤−(1+s2)/2t\le-(1+s^{2})/2 and t+u≤0t+u\le0, we obtain

Re⁡(Ψ(is,t,u+iv;z))=u+t+(c−x−1)t+ays−axβ1−β≤−12[(c−x−1)s2−2ays+(c−x−1+2axβ1−β)]=:Qs(x,y,β).\begin{aligned} \operatorname{Re}(\Psi(is,t,u+iv;z)) &=u+t+(c-x-1)t+ays-\frac{ax\beta}{1-\beta}\\ &\le-\frac{1}{2}\left[(c-x-1)s^{2}-2ays+\left(c-x-1+\frac{2ax\beta}{1-\beta}\right)\right]\\ &=:Q_{s}(x,y,\beta). \end{aligned}

Since the coefficient of s2s^{2} in Qs(x,y,β)Q_{s}(x,y,\beta) is negative, Qs(x,y,β)<0Q_{s}(x,y,\beta)<0 for all s∈Rs\in\mathbb{R} if its discriminant is strictly negative, that is,

(2ay)2<4(c−x−1)(c−x−1+2axβ1−β).(2ay)^{2}<4(c-x-1)\left(c-x-1+\frac{2ax\beta}{1-\beta}\right).

Since z∈Δz \in\Delta, and so it is sufficient to verify the inequality Rx(a,c,β)>0R_x(a,c,\beta)>0, where

Rx(a,c,β)=(a2+1−2aβ1−β)x2−2(c−1)(1−aβ1−β)x+(c−1)2−a2.R_x(a,c,\beta)=\left(a^2+1-\frac{2a\beta}{1-\beta}\right)x^2-2(c-1)\left(1-\frac{a\beta}{1-\beta}\right)x+(c-1)^2-a^2.

The case 0≤β<1/20\leq\beta<1/2 was established in [15 Theorem 1.9]. We now consider the case for β=1/2\beta=1/2. If β=1/2\beta=1/2, then

Rx(a,c,1/2)=(a−1)2x2−2(c−1)(1−a)x+(c−1)2−a2=((a−1)x+(c−1))2−a2≥0\begin{aligned} R_x(a,c,1/2)&=(a-1)^2x^2-2(c-1)(1-a)x+(c-1)^2-a^2\\ &=((a-1)x+(c-1))^2-a^2\geq0 \end{aligned}

if and only if c≥1+∣a∣+∣a−1∣c\geq1+|a|+|a-1| for a≠0a\neq0, c≥1c\geq1.

Moreover, for β∈(1/2,1)\beta\in(1/2,1), we set t=β/(1−β)t=\beta/(1-\beta), B=1−atB=1-at, then ∣t∣>1|t|>1 and

Rx(a,c,β)=(c−1−Bx)2−a2(1+(t2−1)x2).R_x(a,c,\beta)=(c-1-Bx)^2-a^2(1+(t^2-1)x^2).

Therefore,

Rx(a,c,β)≥0⟺∣c−1−Bx∣≥∣a∣1+(t2−1)x2.R_x(a,c,\beta)\geq0\Longleftrightarrow|c-1-Bx|\geq|a|\sqrt{1+(t^2-1)x^2}.

Hence Rx(a,c,β)≥0R_x(a,c,\beta)\geq0 for all x∈(−1,1)x\in(-1,1) if and only if either

c−1≥sup⁡∣x∣<1(Bx+∣a∣1+(t2−1)x2)c-1\geq\sup_{|x|<1}\left(Bx+|a|\sqrt{1+(t^2-1)x^2}\right)

or

c−1≤inf⁡∣x∣<1(Bx−∣a∣1+(t2−1)x2).c-1\leq\inf_{|x|<1}\left(Bx-|a|\sqrt{1+(t^2-1)x^2}\right).

Now define f(x)=Bx+∣a∣1+(t2−1)x2f(x)=Bx+|a|\sqrt{1+(t^2-1)x^2}. Since

f′′(x)=∣a∣t2−1(1+(t2−1)x2)3/2>0,f''(x)=|a|\frac{t^2-1}{(1+(t^2-1)x^2)^{3/2}}>0,

the function ff is convex on (−1,1)(-1,1). Hence sup⁡∣x∣<1f(x)=max⁡{f(1),f(−1)}\sup_{|x|<1}f(x)=\max\{f(1),f(-1)\}. But f(1)=B+∣a∣tf(1)=B+|a|t and f(−1)=−B+∣a∣tf(-1)=-B+|a|t, hence

sup⁡∣x∣<1f(x):=sup⁡∣x∣<1(Bx+∣a∣1+(t2−1)x2)=∣a∣t+∣B∣.\sup_{|x|<1}f(x):=\sup_{|x|<1}\left(Bx+|a|\sqrt{1+(t^2-1)x^2}\right)=|a|t+|B|.

Similarly,

inf⁡∣x∣<1(Bx−∣a∣1+(t2−1)x2)=−(∣B∣+∣a∣t).\inf_{|x|<1}\left(Bx-|a|\sqrt{1+(t^2-1)x^2}\right)=-(|B|+|a|t).

Thus, the sufficient condition for Rx(a,c,β)≥0R_x(a,c,\beta)\geq0,

∣c−1∣≥∣a∣t+∣B∣=∣∣a∣β1−β∣+∣1−aβ1−β∣.|c-1|\geq|a|t+|B|=\left|\frac{|a|\beta}{1-\beta}\right|+\left|1-\frac{a\beta}{1-\beta}\right|.

Since c>1c>1, the required sufficient condition is

c≥1+∣a∣β1−β+∣1−aβ1−β∣.c\geq1+\frac{|a|\beta}{1-\beta}+\left|1-\frac{a\beta}{1-\beta}\right|.

This completes the proof.

Now, using the first derivative identity of the confluent hypergeometric function

Φ′(a;c;z)=acΦ(a+1;c+1;z)(5)\Phi'(a;c;z) = \frac{a}{c}\Phi(a+1;c+1;z) \tag*{(5)}

and restricting the parameter β\beta to β∈[0,1)\beta\in[0,1), we obtain the following extension to the result [15 Theorem 1.14].

Corollary 2.2 If aa, cc, β\beta are real such that a≠−1a \ne-1, β∈[0,1)\beta\in[0,1), and

c≥{1+(a+1−β(a+2))21−2β,0≤β<12,∣a+1∣β1−β+∣1−(a+1)β1−β∣,12≤β<1.c \ge \begin{cases} \sqrt{1+\dfrac{(a+1-\beta(a+2))^2}{1-2\beta}}, & 0 \le\beta< \dfrac{1}{2},\\ \dfrac{|a+1|\beta}{1-\beta}+\left|1-\dfrac{(a+1)\beta}{1-\beta}\right|, & \dfrac{1}{2} \le\beta< 1. \end{cases}

Then the function ca(Φ(a;c;z)−1)\dfrac{c}{a}(\Phi(a;c;z)-1) is univalent in Δ\Delta.

The generalized Laguerre polynomials Ln(α)(z)L_n^{(\alpha)}(z) are related with the confluent Hypergeometric function by the relation

Φ(−n;α+1;z):=Mn,α(z)=n!(α+1)nLn(α)(z)\Phi(-n;\alpha+1;z) := M_{n,\alpha}(z) = \frac{n!}{(\alpha+1)_n}L_n^{(\alpha)}(z)

(α>−1,n∈N∪{0})(\alpha>-1,n\in\mathbb{N}\cup\{0\}).

Hence, taking a=−na=-n and c=α+1c=\alpha+1 in Theorem 2.1, we obtain the following improved result of the normalized Laguerre polynomial Mn,α(z)M_{n,\alpha}(z) [16 Theorem 1].

Corollary 2.3 Let α\alpha, β∈R\beta\in\mathbb{R}, n∈Nn \in\mathbb{N}. If

α≥{1+(n−β(n−1))21−2β,0≤β<12,1+2nβ1−β,β≥12, β≠1.\alpha\ge \begin{cases} \sqrt{1+\dfrac{(n-\beta(n-1))^2}{1-2\beta}}, & 0 \le\beta< \dfrac{1}{2},\\ 1+\dfrac{2n\beta}{1-\beta}, & \beta\ge\dfrac{1}{2},\ \beta\ne1. \end{cases}

Then Re⁡(Mn,α(z))>β\operatorname{Re}(M_{n,\alpha}(z))>\beta for all z∈Δz\in\Delta.

Our next result gives conditions on aa and cc for which (c/a)Φ′(a;c;z)≺((1+z)/(1−z))1/2(c/a)\Phi'(a;c;z)\prec((1+z)/(1-z))^{1/2}, or equivalently, the function (c/a)Φ′(a;c;z)(c/a)\Phi'(a;c;z) lies in the sector ∣arg⁡z∣<πβ/4|\arg z|<\pi\beta/4, β∈(0,1]\beta\in(0,1].

Theorem 2.4 Let aa, cc, β∈R\beta\in\mathbb{R}, a≠0a\ne0, β∈(0,1]\beta\in(0,1]. If

c≥1+(a+1β)2,β>0,c\ge\sqrt{1+\left(\frac{a+1}{\beta}\right)^2},\qquad\beta>0,

then caΦ′(a;c;z)≺(1+z1−z)β\dfrac{c}{a}\Phi'(a;c;z)\prec\left(\dfrac{1+z}{1-z}\right)^\beta.

Proof. Let

p(z)=exp⁡[1βLog⁡(caΦ′(a;c;z))],p(z)=\exp\left[\frac{1}{\beta}\operatorname{Log}\left(\frac{c}{a}\Phi'(a;c;z)\right)\right],

where Log⁡\operatorname{Log} denotes the analytic logarithm in the right half-plane. Then pp is analytic in Δ\Delta, p(0)=1p(0)=1, and

caΦ′(a;c;z)=p(z)β.\frac{c}{a}\Phi'(a;c;z)=p(z)^\beta.

Substituting this identity into the differential equation (2), and multiplying by zz, yields

βp(z)z2p′′(z)+β(β−1)(zp′(z))2+β(c+1−z)p(z)(zp′(z))−(a+1)zp(z)2=0.(6)\beta p(z)z^{2}p''(z)+\beta(\beta-1)\left(zp'(z)\right)^{2}+\beta(c+1-z)p(z)\left(zp'(z)\right)-(a+1)zp(z)^{2}=0. \tag*{(6)}

This equation (6) can be written as Ψ(p(z),zp′(z),z2p′′(z);z)∈Λ\Psi(p(z),zp'(z),z^{2}p''(z);z)\in\Lambda, where Λ={0}\Lambda=\{0\}. By Lemma 1.1, it is sufficient to show that Ψ(is,t,u+iv;z)∉Λ\Psi(is,t,u+iv;z)\notin\Lambda, where s,t,u,v∈Rs,t,u,v\in\mathbb{R}, t+u≤0t+u\leq0 and t≤−(1+s2)/2t\leq-(1+s^{2})/2 for z=x+iy∈Δz=x+iy\in\Delta. Considering s,t,u,v∈Rs,t,u,v\in\mathbb{R}, z∈Δz\in\Delta, t+u≤0t+u\leq0 and t≤−(1+s2)/2t\leq-(1+s^{2})/2, we obtain

Re⁡Ψ(is,t,u+iv;x+iy)=−βsv+β(β−1)t2+βsty+(a+1)xs2,(7)\operatorname{Re}\Psi(is,t,u+iv;x+iy)=-\beta sv+\beta(\beta-1)t^{2}+\beta sty+(a+1)xs^{2}, \tag*{(7)}

and

Im⁡Ψ(is,t,u+iv;x+iy)=βsu+βst(c+1−x)+(a+1)ys2.(8)\operatorname{Im}\Psi(is,t,u+iv;x+iy)=\beta su+\beta st(c+1-x)+(a+1)ys^{2}. \tag*{(8)}

We now consider three cases depending upon s=0s=0, s>0s>0, s<0s<0.

Case (i). If s=0s=0, then from the equation (7), we have Re⁡Ψ(is,t,u+iv;z)=−β(1−β)t2<0\operatorname{Re}\Psi(is,t,u+iv;z)=-\beta(1-\beta)t^{2}<0, thus Re⁡Ψ(is,t,u+iv;z)∉Λ\operatorname{Re}\Psi(is,t,u+iv;z)\notin\Lambda.

Case (ii). If s>0s>0, then using u+t≤0u+t\leq0 and t≤−(1+s2)/2t\leq-(1+s^{2})/2, we get from the equation (8),

Im⁡Ψ(is,t,u+iv;x+iy)=s(−β2(c−x)s2+(a+1)ys−β2(c−x)).(9)\operatorname{Im}\Psi(is,t,u+iv;x+iy)=s\left(-\frac{\beta}{2}(c-x)s^{2}+(a+1)ys-\frac{\beta}{2}(c-x)\right). \tag*{(9)}

Let

Ss(x,y,β)=−β2(c−x)s2+(a+1)ys−β2(c−x)(c>1, ∣z∣<1, β∈[0,1)).S_s(x,y,\beta)=-\frac{\beta}{2}(c-x)s^{2}+(a+1)ys-\frac{\beta}{2}(c-x)\qquad(c>1,\ |z|<1,\ \beta\in[0,1)).

By hypothesis c−x>0c-x>0, hence Ss(x,y,β)S_s(x,y,\beta) is a quadratic polynomial in ss with negative leading coefficient. Therefore, Ss(x,y,β)<0S_s(x,y,\beta)<0 for all s∈Rs\in\mathbb{R} if and only if its discriminant is negative. Hence, the discriminant DD of Ss(x,y,β)S_s(x,y,\beta),

D=(a+1)2y2−β2(c−x)2<(a+1)2(1−x2)−β2(c−x)2.\begin{aligned} D&=(a+1)^{2}y^{2}-\beta^{2}(c-x)^{2}\\ &<(a+1)^{2}(1-x^{2})-\beta^{2}(c-x)^{2}. \end{aligned}

Thus we need

1−x2(c−x)2≤β2(a+1)2for all −1<x<1.\frac{1-x^{2}}{(c-x)^{2}}\leq\frac{\beta^{2}}{(a+1)^{2}}\qquad\text{for all }-1<x<1.

Consider a function

f(x)=1−x2(c−x)2,−1<x<1,f(x)=\frac{1-x^{2}}{(c-x)^{2}},\qquad-1<x<1,

which attains its maximum at x=1/cx=1/c. Thus max⁡−1<x<1f(x)=f(1/c)=1/(c2−1)\max_{-1<x<1}f(x)=f(1/c)=1/(c^{2}-1). Hence a necessary and sufficient condition for Im⁡Ψ(is,t,u+iv;x+iy)<0\operatorname{Im}\Psi(is,t,u+iv;x+iy)<0 is

1c2−1≤β2(a+1)2,\frac{1}{c^{2}-1}\leq\frac{\beta^{2}}{(a+1)^{2}},

that is

c≥1+(a+1β)2,β>0.c\geq\sqrt{1+\left(\frac{a+1}{\beta}\right)^{2}},\qquad\beta>0.

Case (iii). If s<0s<0, then by using (9), with conditions t≤−(1+s2)/2t\leq-(1+s^{2})/2 and u+t≤0u+t\leq0 gives

Im⁡Ψ(is,t,u+iv;x+iy)=s(−β2(c−x)s2+(a+1)ys−β2(c−x)).\operatorname{Im}\Psi(is,t,u+iv;x+iy)=s\left(-\frac{\beta}{2}(c-x)s^{2}+(a+1)ys-\frac{\beta}{2}(c-x)\right).

Since s<0s<0, Im⁡Ψ>0\operatorname{Im}\Psi>0 provided Ss(x,y,β)<0S_s(x,y,\beta)<0, which reduces to case (ii)(ii) only. Thus, in all the cases, Ψ(is,t,u+iv;x+iy)∉Λ\Psi(is,t,u+iv;x+iy)\notin\Lambda. This implies that

caΦ′(a;c;z)<(1+z1−z)β,z∈Δ.\frac{c}{a}\Phi'(a;c;z)<\left(\frac{1+z}{1-z}\right)^{\beta},\qquad z\in\Delta.

This completes the proof.

Theorem 2.5 Let a,c,β∈Ra,c,\beta\in\mathbb{R}, a≠0a\ne0. Suppose that

α=sin⁡(πβ2),γ=cos⁡(πβ2),β0=4πtan⁡−1(12),η1(a,β)=1+∣a∣α+14α2,\alpha=\sin\left(\frac{\pi\beta}{2}\right),\qquad\gamma=\cos\left(\frac{\pi\beta}{2}\right),\qquad\beta_0=\frac{4}{\pi}\tan^{-1}\left(\frac{1}{2}\right),\qquad\eta_1(a,\beta)=1+\frac{|a|}{\alpha}+\frac{1}{4\alpha^2},
η2(a,β)=1+α2a2+γ∣a∣α+14α2,andη3(a,β)=1+(1+γα)∣a∣.\eta_2(a,\beta)=1+\alpha^2a^2+\frac{\gamma|a|}{\alpha}+\frac{1}{4\alpha^2},\qquad\text{and}\qquad\eta_3(a,\beta)=1+\left(1+\frac{\gamma}{\alpha}\right)|a|.

If β∈(0,β0]\beta\in(0,\beta_0] and

c≥{η2(a,β),if ∣a∣≤1α(1+γ),η1(a,β),if 1α(1+γ)≤∣a∣≤12α2,η3(a,β),if ∣a∣>12α2,c\ge \begin{cases} \eta_2(a,\beta),&\text{if }|a|\le\frac{1}{\alpha(1+\gamma)},\\ \eta_1(a,\beta),&\text{if }\frac{1}{\alpha(1+\gamma)}\le|a|\le\frac{1}{2\alpha^2},\\ \eta_3(a,\beta),&\text{if }|a|>\frac{1}{2\alpha^2}, \end{cases}

then Φ(a;c;z)\Phi(a;c;z) is strongly convex of order β\beta. Moreover if β∈(β0,1)\beta\in(\beta_0,1) and

c≥{η2(a,β),0<∣a∣≤12α2,η3(a,β),∣a∣>12α2,c\ge \begin{cases} \eta_2(a,\beta),&0<|a|\le\frac{1}{2\alpha^2},\\ \eta_3(a,\beta),&|a|>\frac{1}{2\alpha^2}, \end{cases}

then Φ(a;c;z)\Phi(a;c;z) is strongly convex of order β\beta.

Proof. Let

q(z)=exp⁡[1βLog⁡(1+zΦ′′(z)Φ′(z))],q(z)=\exp\left[\frac{1}{\beta}\operatorname{Log}\left(1+\frac{z\Phi''(z)}{\Phi'(z)}\right)\right],

where Log⁡\operatorname{Log} denotes the analytic logarithm in the right half-plane. Then qq is analytic in Δ\Delta, q(0)=1q(0)=1, and

1+zΦ′′(z)Φ′(z)=qβ(z)(10)1+\frac{z\Phi''(z)}{\Phi'(z)}=q^\beta(z) \tag*{(10)}

Assume that z≠0z\ne0. Dividing (2) by Φ′(z)\Phi'(z) and using (10), we get

Φ(z)Φ′(z)=aΦ′(z)=qβ(z)+c−1−z.(11)\frac{\Phi(z)}{\Phi'(z)}=\frac{a}{\Phi'(z)}=q^\beta(z)+c-1-z. \tag*{(11)}

Differentiate (11) with respect to zz, we get

1−Φ(z)Φ′(z)⋅Φ′′(z)Φ′(z)=βqβ−1(z)q′(z)−1a.(12)1-\frac{\Phi(z)}{\Phi'(z)}\cdot\frac{\Phi''(z)}{\Phi'(z)}=\frac{\beta q^{\beta-1}(z)q'(z)-1}{a}. \tag*{(12)}

Further, using (10) and (11) in (12), we obtain

zq′(z)+1βqβ+1(z)+c−z−2βq(z)+1−c−azβq1−β(z)=0.(13)zq'(z)+\frac{1}{\beta}q^{\beta+1}(z)+\frac{c-z-2}{\beta}q(z)+\frac{1-c-az}{\beta}q^{1-\beta}(z)=0. \tag*{(13)}

This equation (13) can be written as Ψ(w1,w2;z)∈Λ\Psi(w_1,w_2;z)\in\Lambda, where Λ={0}\Lambda=\{0\}. By Lemma 1.1, Re⁡p(z)>0\operatorname{Re}p(z)>0 if Ψ(is,t;z)∉Λ\Psi(is,t;z)\notin\Lambda, where s,t∈Rs,t\in\mathbb{R} and t≤−(1+s2)/2t\leq-(1+s^2)/2, for z=x+iy∈Δz=x+iy\in\Delta. We prove the stronger statement Re⁡Ψ(is,t;z)<0\operatorname{Re}\Psi(is,t;z)<0, which eventually leads to the required condition. Considering s,t∈Rs,t\in\mathbb{R}, z∈Δz\in\Delta and t≤−(1+s2)/2t\leq-(1+s^2)/2, we obtain

Ψ(is,t;z)=t+1β(is)β+1+c−z−2β(is)+1−c−azβ(is)1−β.(14)\Psi(is,t;z)=t+\frac{1}{\beta}(is)^{\beta+1}+\frac{c-z-2}{\beta}(is)+\frac{1-c-az}{\beta}(is)^{1-\beta}. \tag*{(14)}

We consider three cases depending upon s=0s=0, s>0s>0 and s<0s<0.

Case (i). If s=0s=0, then Re⁡Ψ(is,t;z)=t<−(1+s2)/2<0\operatorname{Re}\Psi(is,t;z)=t<-(1+s^2)/2<0.

Case (ii). If s>0s>0, then using the principal branch,

(is)β+1=sβ+1ei(β+1)π/2,(is)1−β=s1−βei(1−β)π/2.(is)^{\beta+1}=s^{\beta+1}e^{i(\beta+1)\pi/2},\qquad(is)^{1-\beta}=s^{1-\beta}e^{i(1-\beta)\pi/2}.

For s>0s>0 gives is=(1+i)s/2\sqrt{is}=(1+i)\sqrt{s/2}, which is the principal square root of isis. Hence, the above power formulas are consistent with the principal branch. Now

Re⁡(1β(is)β+1)=−sβ+1βsin⁡βπ2,Re⁡(c−z−2β(is))=ysβ,\operatorname{Re}\left(\frac{1}{\beta}(is)^{\beta+1}\right)=-\frac{s^{\beta+1}}{\beta}\sin\frac{\beta\pi}{2},\qquad\operatorname{Re}\left(\frac{c-z-2}{\beta}(is)\right)=\frac{ys}{\beta},

and

Re⁡(1−c−azβ(is)1−β)=s1−ββ[(1−c−ax)sin⁡βπ2+aycos⁡βπ2].\operatorname{Re}\left(\frac{1-c-az}{\beta}(is)^{1-\beta}\right)=\frac{s^{1-\beta}}{\beta}\left[(1-c-ax)\sin\frac{\beta\pi}{2}+ay\cos\frac{\beta\pi}{2}\right].

Therefore

Re⁡Ψ(is,t;z)=t−sβ+1βsin⁡βπ2+ysβ+s1−ββ[(1−c−ax)sin⁡βπ2+aycos⁡βπ2]≤−1+s22−sβ+1βsin⁡βπ2+ysβ+s1−ββ[(1−c−ax)sin⁡βπ2+aycos⁡βπ2]=:Q(s).\begin{aligned} \operatorname{Re}\Psi(is,t;z) &=t-\frac{s^{\beta+1}}{\beta}\sin\frac{\beta\pi}{2}+\frac{ys}{\beta} +\frac{s^{1-\beta}}{\beta}\left[(1-c-ax)\sin\frac{\beta\pi}{2}+ay\cos\frac{\beta\pi}{2}\right]\\ &\leq-\frac{1+s^2}{2}-\frac{s^{\beta+1}}{\beta}\sin\frac{\beta\pi}{2}+\frac{ys}{\beta} +\frac{s^{1-\beta}}{\beta}\left[(1-c-ax)\sin\frac{\beta\pi}{2}+ay\cos\frac{\beta\pi}{2}\right]\\ &=:Q(s). \end{aligned}

Now we need to show Q(s)<0Q(s)<0 for all s>0s>0. Let α=sin⁡(βπ/2)\alpha=\sin(\beta\pi/2) and γ=cos⁡(βπ/2)\gamma=\cos(\beta\pi/2). Then α>0\alpha>0, and

Q(s)=−1+s22−αβsβ+1+ysβ+s1−ββ[(1−c−ax)α+ayγ].(15)Q(s)=-\frac{1+s^2}{2}-\frac{\alpha}{\beta}s^{\beta+1}+\frac{ys}{\beta}+\frac{s^{1-\beta}}{\beta}\left[(1-c-ax)\alpha+ay\gamma\right]. \tag*{(15)}

Since ∣x∣,∣y∣<1|x|,|y|<1, we have ys<sys<s, and by the Cauchy–Schwarz inequality, we have

−axα+ayγ≤∣a∣α2+γ2x2+y2<∣a∣.-ax\alpha+ay\gamma\leq|a|\sqrt{\alpha^2+\gamma^2}\sqrt{x^2+y^2}<|a|.

Also

sβ−αs2β=−α(sβ−12α)2+14α<14α.s^\beta-\alpha s^{2\beta}=-\alpha\left(s^\beta-\frac{1}{2\alpha}\right)^2+\frac{1}{4\alpha}<\frac{1}{4\alpha}.

Hence

Q(s)<−1+s22−αβsβ+1+sβ+s1−ββ((1−c)α+∣a∣)=−1+s22+s1−ββ[sβ−αs2β+(1−c)α+∣a∣]<−1+s22+s1−ββ[(1−c)α+∣a∣+14α]≤0,if (1−c)α+∣a∣+14α≤0.\begin{aligned} Q(s) &< -\frac{1+s^2}{2}-\frac{\alpha}{\beta}s^{\beta+1}+\frac{s}{\beta}+\frac{s^{1-\beta}}{\beta}\left((1-c)\alpha+|a|\right)\\ &= -\frac{1+s^2}{2}+\frac{s^{1-\beta}}{\beta}\left[s^\beta-\alpha s^{2\beta}+(1-c)\alpha+|a|\right]\\ &< -\frac{1+s^2}{2}+\frac{s^{1-\beta}}{\beta}\left[(1-c)\alpha+|a|+\frac{1}{4\alpha}\right]\\ &\leq0,\qquad\text{if }(1-c)\alpha+|a|+\frac{1}{4\alpha}\leq0. \end{aligned}

Therefore, a sufficient condition for Q(s)<0Q(s)<0 is

c≥1+∣a∣α+14α2=:η1(a,β)(16)c\geq1+\frac{|a|}{\alpha}+\frac{1}{4\alpha^2}=:\eta_1(a,\beta) \tag*{(16)}

Now using a different inequality, we obtain another lower bound for cc. Using (15), we have

Q(s)=−1+s22+s1−ββ(−αs2β+ysβ+(1−c−αx)α+ayγ):=−1+s22+s1−ββR(sβ),\begin{aligned} Q(s)&=-\frac{1+s^2}{2}+\frac{s^{1-\beta}}{\beta}\left(-\alpha s^{2\beta}+ys^\beta+(1-c-\alpha x)\alpha+ay\gamma\right)\\ &:=-\frac{1+s^2}{2}+\frac{s^{1-\beta}}{\beta}R(s^\beta), \end{aligned}

where R(u)=−αu2+yu+(1−c−αx)α+ayγR(u)=-\alpha u^2+yu+(1-c-\alpha x)\alpha+ay\gamma. Now the discriminant DD of R(u)R(u)

D=y2+4α((1−c−αx)α+ayγ)<1−x2+4α2(1−c−αx)+4αγ∣a∣=:−S(x),\begin{aligned} D&=y^2+4\alpha\left((1-c-\alpha x)\alpha+ay\gamma\right)\\ &<1-x^2+4\alpha^2(1-c-\alpha x)+4\alpha\gamma|a|=: -S(x), \end{aligned}

where

S(x)=x2+4α2ax+4α2c−4α2−4αγ∣a∣−1=(x+2α2a)2−4α4a2+4α2c−4α2−4αγ∣a∣−1.\begin{aligned} S(x)&=x^2+4\alpha^2ax+4\alpha^2c-4\alpha^2-4\alpha\gamma|a|-1\\ &=(x+2\alpha^2a)^2-4\alpha^4a^2+4\alpha^2c-4\alpha^2-4\alpha\gamma|a|-1. \end{aligned}

Since S(x)S(x) is a quadratic polynomial in xx with positive leading coefficient, its minimum over (−1,1)(-1,1) is attained either at the vertex or at the boundary points x=±1x=\pm1. The vertex of S(x)S(x) is x0=−2α2ax_0=-2\alpha^2a. Thus, we consider two cases.

Let ∣x0∣≤1|x_0|\leq1, that is, ∣a∣≤1/2α2|a|\leq1/2\alpha^2. Then the vertex lies in [−1,1][-1,1], hence

min⁡S(x)=S(x0)=−4α4a2+4α2c−4α2−4αγ∣a∣−1.\min S(x)=S(x_0)=-4\alpha^4a^2+4\alpha^2c-4\alpha^2-4\alpha\gamma|a|-1.

Therefore S(x)≥0S(x)\geq0 for all x∈(−1,1)x\in(-1,1) if and only if

c≥α2a2+1+γα∣a∣+14α2.c\geq\alpha^2a^2+1+\frac{\gamma}{\alpha}|a|+\frac{1}{4\alpha^2}.

Moreover, let ∣x0∣>1|x_0|>1, that is, ∣a∣>1/2α2|a|>1/2\alpha^2. Then the vertex lies outside (−1,1)(-1,1), hence

inf⁡x∈(−1,1)S(x)=1−4α2∣a∣+4α2c−4α2−4αγ∣a∣−1.\inf_{x\in(-1,1)}S(x)=1-4\alpha^2|a|+4\alpha^2c-4\alpha^2-4\alpha\gamma|a|-1.

Therefore, S(x)≥0S(x)\geq0 for all x∈(−1,1)x\in(-1,1) if and only if

c≥1+∣a∣+γα∣a∣.c\geq1+|a|+\frac{\gamma}{\alpha}|a|.

Combining these two cases, the best condition on cc is

c≥{1+α2a2+γ∣a∣α+14α2≔η2(a,β),if ∣a∣≤12α2,1+(1+γα)∣a∣≔η3(a,β),if ∣a∣>12α2.c \ge\begin{cases} 1+\alpha^{2}a^{2}+\frac{\gamma|a|}{\alpha}+\frac{1}{4\alpha^{2}} \coloneqq\eta_{2}(a,\beta), & \text{if } |a|\le\frac{1}{2\alpha^{2}},\\ 1+\left(1+\frac{\gamma}{\alpha}\right)|a| \coloneqq\eta_{3}(a,\beta), & \text{if } |a|>\frac{1}{2\alpha^{2}}. \end{cases}

We now compare different lower bounds of cc. If β∈(0,β0]\beta\in(0,\beta_{0}], then 1α(1+γ)≤12α2\frac{1}{\alpha(1+\gamma)}\le\frac{1}{2\alpha^{2}}. Using 1−γ=α2/(1+γ)1-\gamma=\alpha^{2}/(1+\gamma), we have

η2−η1=α2∣a∣2+γ−1α∣a∣=∣a∣α(α∣a∣−11+γ).\eta_{2}-\eta_{1}=\alpha^{2}|a|^{2}+\frac{\gamma-1}{\alpha}|a|=|a|\alpha\left(\alpha|a|-\frac{1}{1+\gamma}\right).

Hence

η2≤η1⟺∣a∣≤1α(1+γ)andη2>η1⟺∣a∣>1α(1+γ).\eta_{2}\le\eta_{1}\Longleftrightarrow|a|\le\frac{1}{\alpha(1+\gamma)} \qquad\text{and}\qquad \eta_{2}>\eta_{1}\Longleftrightarrow|a|>\frac{1}{\alpha(1+\gamma)}.

Therefore

c≥{η2(a,β),if ∣a∣≤1α(1+γ),η1(a,β),if 1α(1+γ)≤∣a∣≤12α2,η3(a,β),if ∣a∣>12α2.c\ge\begin{cases} \eta_{2}(a,\beta), & \text{if } |a|\le\frac{1}{\alpha(1+\gamma)},\\ \eta_{1}(a,\beta), & \text{if } \frac{1}{\alpha(1+\gamma)}\le|a|\le\frac{1}{2\alpha^{2}},\\ \eta_{3}(a,\beta), & \text{if } |a|>\frac{1}{2\alpha^{2}}. \end{cases}

Moreover, if β∈(β0,1]\beta\in(\beta_{0},1], then 12α2<1α(1+γ)\frac{1}{2\alpha^{2}}<\frac{1}{\alpha(1+\gamma)}. Hence

c≥{η2(a,β),0<∣a∣≤12α2,η3(a,β),∣a∣>12α2.c\ge\begin{cases} \eta_{2}(a,\beta), & 0<|a|\le\frac{1}{2\alpha^{2}},\\ \eta_{3}(a,\beta), & |a|>\frac{1}{2\alpha^{2}}. \end{cases}

Case (iii). If s<0s<0. In this case is=(1−i)m/2\sqrt{is}=(1-i)\sqrt{m/2}, where m=−s>0m=-s>0. Again from (14), we get

Re⁡Ψ(is,t;z)=t−mβ+1βsin⁡βπ2+ymβ+m1−ββ[(1−c−αx)sin⁡βπ2+αycos⁡βπ2]≤−1+m22−mβ+1βsin⁡βπ2+ymβ+m1−ββ[(1−c−αx)sin⁡βπ2+αycos⁡βπ2].\begin{aligned} \operatorname{Re}\Psi(is,t;z)=t-\frac{m^{\beta+1}}{\beta}\sin\frac{\beta\pi}{2}+\frac{ym}{\beta}+\frac{m^{1-\beta}}{\beta}\left[(1-c-\alpha x)\sin\frac{\beta\pi}{2}+\alpha y\cos\frac{\beta\pi}{2}\right] \\ &\le-\frac{1+m^{2}}{2}-\frac{m^{\beta+1}}{\beta}\sin\frac{\beta\pi}{2}+\frac{ym}{\beta}+\frac{m^{1-\beta}}{\beta}\left[(1-c-\alpha x)\sin\frac{\beta\pi}{2}+\alpha y\cos\frac{\beta\pi}{2}\right]. \end{aligned}

The last inequality is the same as the one in case (ii) and therefore holds. Combining all the cases and using Lemma 1.1, we conclude that Φ(a;c;z)\Phi(a;c;z) is strongly convex of order β\beta. □

Using the derivative identity (5), we get

1+Φ′′(a−1;c−1;z)Φ′(a−1;c−1;z)=z(Φ′(a;c;z))′zΦ(a;c;z).1+\frac{\Phi''(a-1;c-1;z)}{\Phi'(a-1;c-1;z)}=\frac{z\left(\Phi'(a;c;z)\right)'}{z\Phi(a;c;z)}.

Hence, we obtain the following result of strong starlikeness.

Corollary 2.6 Let a,c,β∈Ra,c,\beta\in\mathbb{R}, a≠1a \ne1. Suppose that

α=sin⁡(πβ2),γ=cos⁡(πβ2),β0=4πtan⁡−1(12),η~1(a,β)=1+∣a−1∣α+14α2,\alpha= \sin\left(\frac{\pi\beta}{2}\right),\quad\gamma= \cos\left(\frac{\pi\beta}{2}\right),\quad\beta_0 = \frac{4}{\pi}\tan^{-1}\left(\frac{1}{2}\right),\quad\widetilde{\eta}_1(a,\beta) = 1+\frac{|a-1|}{\alpha}+\frac{1}{4\alpha^2},
η~2(a,β)=1+α2(a−1)2+γ∣a−1∣α+14α2,andη~3(a,β)=1+(1+γα)∣a−1∣.\widetilde{\eta}_2(a,\beta) = 1+\alpha^2(a-1)^2+\frac{\gamma|a-1|}{\alpha}+\frac{1}{4\alpha^2},\quad\text{and}\quad\widetilde{\eta}_3(a,\beta) = 1+\left(1+\frac{\gamma}{\alpha}\right)|a-1|.

If β∈(0,β0]\beta\in(0,\beta_0] and

c≥{1+η~2(a,β),∣a−1∣≤1α(1+γ),1+η~1(a,β),1α(1+γ)≤∣a−1∣≤12α2,1+η~3(a,β),∣a−1∣>12α2,c \ge\begin{cases} 1+\widetilde{\eta}_2(a,\beta), & |a-1|\le\dfrac{1}{\alpha(1+\gamma)},\\ 1+\widetilde{\eta}_1(a,\beta), & \dfrac{1}{\alpha(1+\gamma)}\le|a-1|\le\dfrac{1}{2\alpha^2},\\ 1+\widetilde{\eta}_3(a,\beta), & |a-1|>\dfrac{1}{2\alpha^2}, \end{cases}

then Φ(a;c;z)\Phi(a;c;z) is strongly starlike of order β\beta. Moreover if β∈(β0,1)\beta\in(\beta_0,1) and

c≥{1+η~2(a,β),0<∣a−1∣≤12α2,1+η~3(a,β),∣a−1∣>12α2,c \ge\begin{cases} 1+\widetilde{\eta}_2(a,\beta), & 0<|a-1|\le\dfrac{1}{2\alpha^2},\\ 1+\widetilde{\eta}_3(a,\beta), & |a-1|>\dfrac{1}{2\alpha^2}, \end{cases}

then Φ(a;c;z)\Phi(a;c;z) is strongly starlike of order β\beta.

Setting β=1/2\beta=1/2 in Theorem 2.5 and Corollary 2.6 yields the following improvement in [3 Theorem 1.1, Corollary 1.2].

Corollary 2.7 Let a,c∈Ra,c\in\mathbb{R}, a≠0a\ne0. Suppose that

c≥{32+∣a∣+a22,0<∣a∣≤2(2−1),32+2∣a∣,2(2−1)<∣a∣≤1,1+2∣a∣,∣a∣>1.c \ge\begin{cases} \dfrac{3}{2}+|a|+\dfrac{a^2}{2}, & 0<|a|\le2(\sqrt{2}-1),\\ \dfrac{3}{2}+\sqrt{2}|a|, & 2(\sqrt{2}-1)<|a|\le1,\\ 1+2|a|, & |a|>1. \end{cases}

Then Φ(a;c;z)\Phi(a;c;z) is strongly convex of order 1/21/2 in the unit disk Δ\Delta.

Corollary 2.8 Let a,c∈Ra,c\in\mathbb{R}, a≠1a\ne1. Suppose that

c≥{52+∣a−1∣+(a−1)22,0<∣a−1∣≤2(2−1),32+2∣a−1∣,2(2−1)<∣a−1∣≤1,2+2∣a−1∣,∣a−1∣>1.c \ge\begin{cases} \dfrac{5}{2}+|a-1|+\dfrac{(a-1)^2}{2}, & 0<|a-1|\le2(\sqrt{2}-1),\\ \dfrac{3}{2}+\sqrt{2}|a-1|, & 2(\sqrt{2}-1)<|a-1|\le1,\\ 2+2|a-1|, & |a-1|>1. \end{cases}

Then Φ(a;c;z)\Phi(a;c;z) is strongly starlike of order 1/21/2 in the unit disk Δ\Delta.

To prove the next result, we need the following Lemma.

Lemma 2.9 Let m∈Nm\in\mathbb{N} and let f(z)=z+∑n=1∞amn+1zmn+1∈Af(z)=z+\sum_{n=1}^{\infty}a_{mn+1}z^{mn+1}\in\mathcal{A}. Suppose that the coefficients are real and satisfy either

1≥(m+1)am+1≥(2m+1)a2m+1≥⋯≥0,1\ge(m+1)a_{m+1}\ge(2m+1)a_{2m+1}\ge\cdots\ge0,

or

1≤(m+1)am+1≤(2m+1)a2m+1≤⋯ ,lim⁡n→∞(mn+1)amn+1≤2.1\le(m+1)a_{m+1}\le(2m+1)a_{2m+1}\le\cdots,\qquad\lim_{n\to\infty}(mn+1)a_{mn+1}\le2.

Then ff is close-to-convex with respect to gm(z)=∫0z(1−tm)−1 dtg_m(z)=\int_{0}^{z}(1-t^m)^{-1}\,\mathrm{d}t in Δ\Delta for m≤2m\le2, and in ∣z∣<Rm:=(m−1)−1/m|z|<R_m:=(m-1)^{-1/m} for m≥3m\ge3.

Proof. Clearly gm∈Ag_m\in\mathcal{A} and

1+zgm′′(z)gm′(z)=1+(m−1)zm1−zm.1+\frac{zg_m''(z)}{g_m'(z)}=\frac{1+(m-1)z^m}{1-z^m}.

Thus g1g_1 and g2g_2 are convex in Δ\Delta. Now suppose that m≥3m\ge3. Since m>2m>2 and Re⁡(zm)≥−∣z∣m\operatorname{Re}(z^m)\ge-|z|^m, then

Re⁡(1+zgm′′(z)gm′(z))≥1−(m−2)∣z∣m−(m−1)∣z∣2m∣1−zm∣2=(1+∣z∣m)(1−(m−1)∣z∣m)∣1−zm∣2>0\begin{aligned} \operatorname{Re}\left(1+\frac{zg_m''(z)}{g_m'(z)}\right)\ge\frac{1-(m-2)|z|^m-(m-1)|z|^{2m}}{|1-z^m|^2} \\ &=\frac{(1+|z|^m)(1-(m-1)|z|^m)}{|1-z^m|^2}>0 \end{aligned}

whenever ∣z∣<(m−1)−1/m=Rm|z|<(m-1)^{-1/m}=R_m. Hence gmg_m is convex in ∣z∣<Rm|z|<R_m.

In either case (i) or (ii), the sequence {bn}\{b_n\} is monotone and satisfies ∑n=1∞∣bn−bn−1∣=∣lim⁡n→∞bn−b0∣≤1\sum_{n=1}^{\infty}|b_n-b_{n-1}|=|\lim_{n\to\infty}b_n-b_0|\le1. Hence, for every z∈Δz\in\Delta,

Re⁡{f′(z)gm′(z)}=Re⁡{(1−zm)f′(z)}\operatorname{Re}\left\{\frac{f'(z)}{g_m'(z)}\right\}=\operatorname{Re}\{(1-z^m)f'(z)\}
≥1−∑n=1∞∣bn−bn−1∣∣z∣mn>0.\ge1-\sum_{n=1}^{\infty}|b_n-b_{n-1}||z|^{mn}>0.

Therefore, ff is close-to-convex with respect to gmg_m in the respective disks.

We now prove a sufficient condition for the close-to-convexity of zΦ(a;c;zm)z\Phi(a;c;z^m).

Theorem 2.10. Let m∈Nm\in\mathbb{N}, a>0a>0, and c≥(m+1)ac\ge(m+1)a. Then the function fm(z)=zΦ(a;c;zm)f_m(z)=z\Phi(a;c;z^m) is close-to-convex in Δ\Delta with respect to gm(z)=∫0z(1−tm)−1 dtg_m(z)=\int_{0}^{z}(1-t^m)^{-1}\,\mathrm{d}t for m≤2m\le2, and close-to-convex in ∣z∣<Rm:=(m−1)−1/m|z|<R_m:=(m-1)^{-1/m} for m≥3m\ge3.

Proof. Let

fm(z)=zΦ(a;c;zm)=z+∑n=1∞Anzmn+1,f_m(z)=z\Phi(a;c;z^m)=z+\sum_{n=1}^{\infty}A_nz^{mn+1},

where An=(a)n/(c)nn!A_n=(a)_n/(c)_n n! and set bn=(mn+1)Anb_n=(mn+1)A_n, where A0=1A_0=1. Then b0=1b_0=1, b1≤1b_1\le1, and

bnbn+1−1=(mn+1)(c+n)(n+1)−(mn+m+1)(a+n)(mn+1)(c+n)(n+1).\frac{b_n}{b_{n+1}}-1=\frac{(mn+1)(c+n)(n+1)-(mn+m+1)(a+n)}{(mn+1)(c+n)(n+1)}.

For n≥0n\ge0, the hypothesis c≥(m+1)ac\ge(m+1)a gives

Dn=(mn+1)(c+n)(n+1)−(mn+m+1)(a+n)≥a[(m+1)(mn+1)(n+1)−(mn+m+1)]+n[(mn+1)(n+1)−(mn+m+1)]=a[m(m+1)n2+(m2+m+1)n]+n(mn2+n−m)≥0.\begin{aligned} D_n&=(mn+1)(c+n)(n+1)-(mn+m+1)(a+n)\\ &\ge a[(m+1)(mn+1)(n+1)-(mn+m+1)]\\ &\quad+n[(mn+1)(n+1)-(mn+m+1)]\\ &=a[m(m+1)n^2+(m^2+m+1)n]+n(mn^2+n-m)\ge0. \end{aligned}

This implies that bn≥bn+1b_n\ge b_{n+1}, n≥1n\ge1. Hence 1=b0≥b1≥b2≥⋯>01=b_0\ge b_1\ge b_2\ge\cdots>0. Now using Lemma 2.9, fmf_m is close-to-convex with respect to gmg_m in respective discs.

Remark 2.1. (i) Taking m=1m=1 in Theorem 2.10, we conclude that zΦ(a;c;z)z\Phi(a;c;z) is close-to-convex with respect to −log⁡(1−z)-\log(1-z) in Δ\Delta, provided that a>0a>0 and c≥2ac\ge2a. For 0<a<1/30<a<1/3, this condition is best than the corresponding sufficient condition c≥(1+a)/2c\ge(1+a)/2 obtained in [15 Theorem 1.3].

(ii) Taking m=2m=2 in Theorem 2.10, we conclude that zΦ(a;c;z2)z\Phi(a;c;z^2) is close-to-convex with respect to (1/2)log⁡[(1+z)/(1−z)](1/2)\log[(1+z)/(1-z)] in Δ\Delta, provided that a>0a>0 and c≥3ac\ge3a. For 0<a<2/70<a<2/7, this condition is best than the corresponding sufficient condition c≥2(1+a)/3c\ge2(1+a)/3 obtained in [15 Theorem 1.5].

(iii) Taking m=3m=3 in Theorem 2.10, we conclude that zΦ(a;c;z3)z\Phi(a;c;z^3) is close-to-convex with respect to

g3(z)=−13log⁡(1−z)+16log⁡(1+z+z2)+13(arctan⁡2z+13−π6)g_3(z)=-\frac{1}{3}\log(1-z)+\frac{1}{6}\log(1+z+z^2)+\frac{1}{\sqrt{3}}\left(\arctan\frac{2z+1}{\sqrt{3}}-\frac{\pi}{6}\right)

in ∣z∣<2−1/3≈0.7937|z|<2^{-1/3}\approx0.7937, provided that a>0a>0 and c≥4ac\ge4a.

(iv) Taking m=4m=4 in Theorem 2.10, we conclude that zΦ(a;c;z4)z\Phi(a;c;z^4) is close-to-convex with respect to

g4(z)=14log⁡1+z1−z+12arctan⁡zg_4(z)=\frac{1}{4}\log\frac{1+z}{1-z}+\frac{1}{2}\arctan z

in ∣z∣<3−1/4≈0.7598|z|<3^{-1/4}\approx0.7598, provided that a>0a>0 and c≥5ac\ge5a.

Finally, using Lemma 1.2 and Theorem 2.10, we prove the following improved version of [15 Theorem 1.8].

Theorem 2.11 Let a>0a>0, and define

C1(a)=4a−1+10a2+2a+12andC2(a)=9a−9+3(11a2+42a+43)12.C_1(a)=\frac{4a-1+\sqrt{10a^2+2a+1}}{2} \qquad\text{and}\qquad C_2(a)=\frac{9a-9+\sqrt{3(11a^2+42a+43)}}{12}.

If c≥max⁡{C1(a),C2(a)}c\ge\max\{C_1(a),C_2(a)\}, then f(z)=zΦ(a;c;z)f(z)=z\Phi(a;c;z) is starlike and close-to-convex with respect to −log⁡(1−z)-\log(1-z) in Δ\Delta.

Proof. Let f(z)=zΦ(a;c;z)=z+∑n=2∞Anznf(z)=z\Phi(a;c;z)=z+\sum_{n=2}^{\infty}A_nz^n, where

A1=1,An=(a)n−1(c)n−1(n−1)!,n≥2.A_1=1,\qquad A_n=\frac{(a)_{n-1}}{(c)_{n-1}(n-1)!},\qquad n\ge2.

Since a>0a>0 and c≥C1(a)>2a>0c\ge C_1(a)>2a>0, we have An>0A_n>0 for every n≥1n\ge1. By Theorem 2.10, the sequence {nAn}n≥1\{nA_n\}_{n\ge1} is non-increasing. Hence, ff is close-to-convex with respect to −log⁡(1−z)-\log(1-z) in Δ\Delta.

To establish starlikeness, define Bn=nAn−(n+1)An+1B_n=nA_n-(n+1)A_{n+1}, n≥1n\ge1. We shall show that {Bn}n≥1\{B_n\}_{n\ge1} is non-increasing. First note that

B1−B2=2c2+(2−8a)c+3a2−5a2c(c+1)≥0B_1-B_2=\frac{2c^2+(2-8a)c+3a^2-5a}{2c(c+1)}\ge0

for all c≥C1(a)c\ge C_1(a), where C1(a)C_1(a) is the larger root of the quadratic polynomial in the numerator. Next, for n≥2n\ge2, we obtain

Bn−Bn+1=P(n)n(n+1)(c+n−1)(c+n)An,(17)B_n-B_{n+1}=\frac{P(n)}{n(n+1)(c+n-1)(c+n)}A_n, \tag*{(17)}

where PP is a polynomial of degree five. Writing n=y+2n=y+2, we obtain

P(y+2)=y5+p4y4+p3y3+p2y2+p1y+p0,P(y+2)=y^5+p_4y^4+p_3y^3+p_2y^2+p_1y+p_0,

with

p4=2(c+4),p3=c2+15c−2a+22,p2=7c2+39c−2ac−14a+23,p1=a2+16c2+42c−12ac−31a+4,p0=2(6c2+9c−9ac+2a2−12a−2).\begin{aligned} p_4&=2(c+4), & p_3&=c^2+15c-2a+22, & p_2&=7c^2+39c-2ac-14a+23,\\ p_1&=a^2+16c^2+42c-12ac-31a+4, & p_0&=2(6c^2+9c-9ac+2a^2-12a-2). \end{aligned}

We now verify that pj≥0p_j\geq0 for j=0,1,…,4j=0,1,\ldots,4. Clearly, p4>0p_4>0. Moreover, p1p_1, p2p_2, and p3p_3 are increasing functions of cc for c≥2ac\geq2a. Evaluating them at c=2ac=2a, we find

p3>p3∣c=2a,p1≥p1∣c=2a>0,andp2≥p2∣c=2a>0.p_3>p_3\big|_{c=2a},\qquad p_1\geq p_1\big|_{c=2a}>0,\qquad\text{and}\qquad p_2\geq p_2\big|_{c=2a}>0.

Hence p1,p2,p3>0p_1,p_2,p_3>0 whenever c≥2ac\geq2a.

It remains to consider p0p_0. Write p0=2Q(c)p_0=2Q(c), where Q(c)=6c2+(9−9a)c+2a2−12a−2Q(c)=6c^2+(9-9a)c+2a^2-12a-2. Since QQ is strictly increasing on [2a,∞)[2a,\infty) and C2(a)C_2(a) is the larger root of Q(c)=0Q(c)=0, the condition c≥max⁡{C1(a),C2(a)}c\geq\max\{C_1(a),C_2(a)\} implies Q(c)≥0Q(c)\geq0. Therefore p0≥0p_0\geq0. It follows that P(y+2)≥0P(y+2)\geq0, y≥0y\geq0.

In particular, P(n)≥0P(n)\geq0 for every integer n≥2n\geq2. This implies Bn−Bn+1≥0B_n-B_{n+1}\geq0 (n≥2n\geq2). Now combining this inequality with B1≥B2B_1\geq B_2, we conclude that {Bn}n≥1\{B_n\}_{n\geq1} is non-increasing. Therefore Lemma 1.2 yields f(z)=zΦ(a;c;z)∈S∗f(z)=z\Phi(a;c;z)\in S^*. This completes the proof. □

Concluding remarks. The close connections between confluent hypergeometric functions and several classical special functions, including generalized Laguerre polynomials, Whittaker functions, Bessel functions, and Hermite polynomials, provide opportunities to extend the results obtained in this paper. Through suitable parameter specializations, the established sufficient conditions may be used to investigate the geometric properties of these related special functions.

Furthermore, the present study opens avenues for future research, particularly in determining sharper parameter bounds and the exact radii of starlikeness, convexity, and close-to-convexity for the associated functions. Such investigations may lead to refined geometric criteria and a deeper understanding of the relationships between confluent hypergeometric functions and their special cases.

Funding Declaration. The authors declare that no financial support was received for this manuscript.

Clinical Trial Declaration. This manuscript does not involve any clinical trials.

References

  1. [1]Arzikulov O Z, Khasanov A, Ergashev T G. Confluent hypergeometric functions and their application to solving the Dirichlet problem for the Helmholtz equation with three singular coefficients. Vestn. Samar. Gos. Tekhn. Univ., Ser. Fiz.-Mat. Nauki, 2025, 29(3): 407–429
  2. [2]Berisha N Sh, Berisha F M, Fejzullahu B Xh. Some Fourier transforms involving confluent hypergeometric functions. Int Trans Spec Funct, 2024, 35(5-6), 316–329DOI
  3. [3]Bohra N, Ravichandran V. On confluent hypergeometric functions and generalized Bessel functions. Anal Math, 2017, 43(4), 533–545DOI
  4. [4]Brannan D A, Kirwan W E. On some classes of bounded univalent functions. J London Math Soc, 1969, 2, 431–443DOI
  5. [5]Carlson B C, Shaffer D B. Starlike and prestarlike hypergeometric functions. J Math Anal Anal, 1984, 15, 737–745DOI
  6. [6]Duren P L. Univalent Functions. Springer-Verlag, 1983.
  7. [7]Fejér L. Untersuchungen über Potenzreihen mit mehrfach monotoner Koeffizientenfolge. Acta Litt Sci Szeged, 1937, 8, 89–115
  8. [8]Goodman A W. Univalent functions. Mariner Publishing Co., Tampa, FL 1983
  9. [9]Gil A, Ruiz-Antolín D, Segura J, Temme N M. Computation of the confluent hypergeometric function U(a,b,x) and its derivative for positive arguments. Numer Algorithms, 2023, 94, 669–679
  10. [10]Hasanov A, Yuldashev T K. Analytic Continuation Formula for a Confluent Hypergeometric Function E₁ of Three Arguments. Lobachevskii J Math, 2025, 46, 5934–5941DOI
  11. [11]Lin J, Xu X. Explicit evaluation of the Stokes matrices for certain quantum confluent hypergeometric equations. J Geom Phys, 2015, 209, 105364
  12. [12]Merkes E P, Scott W T. Starlike hypergeometric functions. Proc Amer Math Soc, 1961, 12, 885–888DOI
  13. [14]Miller S S, Mocanu P T. Univalence of Gaussian and Confluent Hypergeometric Functions. Proc Am Math Soc, 1990, 110(2), 332–342DOI
  14. [15]Ponnusamy S, Vuorinen M. Univalence and Convexity Properties for Confluent Hypergeometric Functions. Complex Var Elliptic Eqn, 1998, 36, 73–97DOI
  15. [16]Prajapat J K, Dash P P, Sheshma A, Raina R K. Certain characterization properties of the Laguerre polynomials. J Anal, 2024, 32, 3139–3154DOI
  16. [17]Stankiewicz J. On a family of starlike functions. Ann Univ Mariae Curie-Sklodowska Sect A, 1968, 22/24, 175–181
  17. [18]Silverman H. Starlike and convexity properties for hypergeometric functions. J Math Anal Appl, 1983, 172, 574–581
  18. [19]Tudela-Pi M, Colombaro I. Passive two-plateau relaxation from Tricomi confluent hypergeometric kernels. Nonlinear Dyn, 2026, 114, 901

Paper details

vlnc:W_p8b92fgm
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents