arXiv reader

On the difference of the two initial logarithmic coefficients for Bazilevic class of univalent functions

M. Obradovic and N. Tuneski

arXiv
2504.09928v1 · PDF · 14 April 2025
Categories
math.CV
Licence
CC BY 4.0 The site shows the text.

Abstract

In this paper we give sharp bounds of the difference of the moduli of the second and the first logarithmic coefficient for Bazilevič class of univalent functions.

Items 2

Theorem

Theorem 1

054B

Let f\in\mathcal{B}_{1}(\alpha) for some \alpha>0. Then

-\frac{1}{\sqrt{(\alpha+1)^{2}+1}}\leq |\gamma_{2}|-|\gamma_{1}|\leq\frac{1}{\alpha+2}.

The right estimation is sharp for all \alpha>0, while the left is sharp for \alpha_1\le\alpha\le\alpha_2, where \alpha_1= \frac{1}{2} \left(\sqrt{6}-2\right)=0.2247\ldots and \alpha_2 = \sqrt2-1=0.4142\ldots are the positive roots of the equation

2 \alpha ^4+8 \alpha ^3+5 \alpha ^2-6 \alpha +1=0.
Proof of Theorem 1

Let f\in\mathcal{B}_{1}(\alpha), \alpha>0, and let

F(f,\alpha;z)=\left(\frac{f(z)}{z}\right)^{\alpha-1}f'(z).

Then, by the definition of the class \mathcal{B}_{1}(\alpha), the function

\begin{split} G(f,\alpha;z) &=\frac{1-F(f,\alpha;z)}{1+F(f,\alpha;z)}\\ &=-\frac{\alpha+1}{2}a_{2}z- \left[\frac{\alpha+2}{2}a_{3}-\frac{\alpha+3}{4}a_{2}^{2}\right]z^{2}+\cdots \\ &=:c_{1}z+c_{2}z^{2}+\cdots \end{split}

is a Schwartz function, and by using the inequalities |c_{1}|\leq1 and |c_{2}|\leq 1-|c_{1}|^{2}, we receive

\frac{1}{2}(1+\alpha)|a_{2}|\leq1

and

\left|\frac{\alpha+2}{2}a_{3}-\frac{\alpha+3}{4}a_{2}^{2}\right|\leq1-\frac{1}{4}(1+\alpha)^{2}|a_{2}|^{2}.

From these inequalities we obtain |a_{2}|\leq\frac{2}{1+\alpha} and

\left|a_{3}-\frac{\alpha+3}{2(\alpha+2)}a_{2}^{2}\right| \leq\frac{2}{\alpha+2}-\frac{(\alpha+1)^{2}}{2(\alpha+2)}|a_{2}|^{2}.

For the upper bounds (right estimates) of |\gamma _{2}|-|\gamma_{1}|, using (?) and (Proof), for \alpha>0 we have:

\begin{split} |\gamma _{2}|-|\gamma_{1}|&=\frac{1}{2}\left|a_3-\frac{1}{2}a_{2}^{2}\right|-\frac{1}{2}|a_{2}|\\ &=\frac{1}{2}\left|\left( a_{3}-\frac{1}{2}a_{2}^{2}-\frac{1}{2(\alpha+2)}a_{2}^{2} \right) + \frac{1}{2(\alpha+2)}a_{2}^{2}\right|-\frac{1}{2}|a_{2}|\\ &\leq\frac12\left|a_{3}-\frac{\alpha+3}{2(\alpha+2)}a_{2}^{2}\right|+\frac{1}{4(\alpha+2)}|a_{2}|^{2} -\frac{1}{2}|a_{2}|\\ &\leq \frac{1}{\alpha+2}-\frac{(\alpha+1)^{2}}{4(\alpha+2)}|a_{2}|^{2}+\frac{1}{4(\alpha+2)}|a_{2}|^{2} -\frac{1}{2}|a_{2}|\\ &=\frac{1}{\alpha+2}-\frac{\alpha}{4}|a_{2}|^{2}-\frac{1}{2}|a_{2}|\\ &\leq \frac{1}{\alpha+2}. \end{split}

As for sharpness of this estimate, let first consider relation

\left( \frac{f(z)}{z}\right)^{\alpha-1}f'(z)=h(z),\quad \alpha>0, h(0)=1.

From (Proof) we receive

[f^\alpha(z)]' = \alpha z^{\alpha-1}h(z),

and after some calculations,

f(z)=z \left( \frac{\alpha}{z^{\alpha}}\int_{0}^{z}t^{\alpha-1}h(t)dt\right)^{\frac{1}{\alpha}}.

Choosing h(z)=\frac{1+z^{2}}{1-z^{2}}=1+2z^2+\cdots, after equating the coefficients, we receive a_2=0 and a_3=\frac{2}{\alpha+2}, i.e.,

f(z)=z+\frac{2}{\alpha+2}z^{3}+\cdots .

This gives that

|\gamma _{2}|-|\gamma_{1}|=\frac{1}{2}|a_{3}|=\frac{1}{\alpha+2}.

Thus, the right hand estimation is sharp for every \alpha>0.

Now, for the lower bound we should prove

\frac{1}{2}\left|a_3-\frac{1}{2}a_{2}^{2}\right|-\frac{1}{2}|a_{2}|\geq-\frac{1}{\sqrt{(\alpha +1)^{2}+1}},

i.e.,

\left|a_3-\frac{1}{2}a_{2}^{2}\right|\geq|a_{2}|-\frac{2}{\sqrt{(\alpha +1)^{2}+1}}.

If 0\leq|a_{2}|<\frac{2}{\sqrt{(\alpha +1)^{2}+1)}}, then the previous inequality obviously holds.

Further, let study the case when |a_{2}|\geq \frac{2}{\sqrt{(\alpha +1)^{2}+1)}}. Using (?) and (Proof) we have

\begin{split} \left|a_3-\frac{1}{2}a_{2}^{2}\right|&= \left|\left(a_{3}-\frac{\alpha+3}{2(\alpha+2)}a_{2}^{2}\right)+ \frac{1}{2(\alpha+2)}a_{2}^{2}\right|\\ &\geq\frac{1}{2(\alpha+2)}|a_{2}|^{2}-\left|a_{3}-\frac{\alpha+3}{2(\alpha+2)}a_{2}^{2}\right|\\ &\geq\frac{1}{2(\alpha+2)}|a_{2}|^{2}-\frac{2}{\alpha+2}+\frac{(\alpha+1)^{2}}{2(\alpha+2)}|a_{2}|^{2}\\ &=\frac{(\alpha+1)^{2}+1}{2(\alpha+2)}|a_{2}|^{2}-\frac{2}{\alpha+2}. \end{split}

So, it is enough to prove that

\frac{(\alpha+1)^{2}+1}{2(\alpha+2)}|a_{2}|^{2}-\frac{2}{\alpha+2}\geq |a_{2}|- \frac{2}{\sqrt{(\alpha +1)^{2}+1}},

which is equivalent to

\frac{(\alpha+1)^{2}+1}{2(\alpha+2)}\left(|a_{2}|^{2}- \frac{4}{(\alpha +1)^{2}+1}\right) -\left(|a_{2}|- \frac{2}{\sqrt{(\alpha +1)^{2}+1}}\right)\geq0 ,

i.e., to

\left(|a_{2}|- \frac{2}{\sqrt{(\alpha +1)^{2}+1}}\right)\left[\frac{(\alpha+1)^{2}+1}{2(\alpha+2)} \left(|a_{2}|+\frac{2}{\sqrt{(\alpha +1)^{2}+1}}\right)-1\right]\geq 0.

The last inequality indeed holds since by assumption |a_{2}|\geq \frac{2}{\sqrt{(\alpha +1)^{2}+1}} and

\begin{split} \frac{(\alpha+1)^{2}+1}{2(\alpha+2)}\left(|a_{2}|+\frac{2}{\sqrt{(\alpha +1)^{2}+1}}\right)-1 &\geq \frac{(\alpha+1)^{2}+1}{2(\alpha+2)}\frac{4}{\sqrt{(\alpha +1)^{2}+1}}-1 \\ &=\frac{2\sqrt{(\alpha +1)^{2}+1}}{\alpha+2}-1\\ &>\frac{2(\alpha+1)}{\alpha+2}-1=\frac{\alpha}{\alpha+2}>0. \end{split}

As for the sharpness of the lower (left) estimate, equality sign in (Proof) is obtained if |a_{2}|=\frac{2}{\sqrt{(\alpha +1)^{2}+1}}=:b_{2}, i.e., if a_{2}=b_{2}e^{i\theta}, and a_3=\frac{1}{2}a_{2}^{2}=\frac{1}{2}b_{2}e^{i2\theta}. Now, the function

f(z)=z+b_{2}e^{i\theta}z^{2}+\frac{1}{2}b_{2}^{2}e^{2i\theta}z^{3}

has such coefficients a_2 and a_3, and further

\left( \frac{f(z)}{z}\right)^{\alpha-1}f'(z) = 1+(\alpha +1)b_{2}e^{i\theta}z+\frac{\alpha(\alpha+2)}{2}b^2_{2}e^{i2\theta}z^{2} =: h_2(z),

It remains to check that under the conditions of the theorem, i.e., for \alpha_1\le \alpha\le\alpha_2, {\operatorname{Re}\,}\{h_2(z)\}>0,\, z\in{\mathbb D}. In that direction, we study

{\operatorname{Re}\,}\{h_2(e^{i\phi})\}=1+(\alpha +1)b_{2}\cos(\theta+\phi)+\frac{\alpha(\alpha+2)}{2}b_{2}^{2}\cos(2(\theta+\phi)),

-\pi\le \phi \le\pi, which with notation \cos(\theta+\phi)=t, -1\leq t\leq1, becomes

{\operatorname{Re}\,}\{h_2(e^{i\phi})\} = \alpha(\alpha+2)b_{2}^{2}t^{2}+(\alpha+1)b_{2}t+1-\frac{\alpha(\alpha+2)}{2}=: \psi(t),

and we need to show that \psi(t)\ge0 for all \alpha_0\le\alpha\le\alpha_2 and -1\le t\le 1. The vertex of the concave upwards quadratic function \psi lies on the left half-plane, so on [-1,1] it attains its minimal value at t_* = \max\{-1, t_1\},

t_1 = -\frac{\left((\alpha +1)^2+1\right) (\alpha +1)}{4 \alpha (\alpha +2) \sqrt{(\alpha +1)^2+1}} .

Now, t_1\ge -1 is equivalent to t_1^2 \le 1, i.e., to

\frac{(\alpha +1)^2 \left((\alpha +1)^2+1\right)}{16 \alpha ^2 (\alpha +2)^2} \le 1,

and further

2 + 6 \alpha - 57 \alpha ^2 - 60 \alpha ^3 - 15 \alpha ^4 \le 0.

The last, for \alpha>0, is equivalent to \alpha\ge \alpha_3 = \sqrt{\frac{1}{30} \left(\sqrt{129}+33\right)}-1 = 0.21597\ldots.

Thus, for \alpha\ge \alpha_3, the minimal value of \psi on the interval [-1,1] is

\psi(t_*) = -\frac{(\alpha +1)^2}{4 \alpha (\alpha +2)}-\frac{1}{2} \alpha (\alpha +2)+1

which is greater or equal to zero if, and only, if \alpha_3<\alpha_1\le\alpha\le\alpha_2.

The case when \alpha\le \alpha_3 leads to the minimal value of \psi on the interval [-1,1]:

\psi(-1) = -\frac{2 (\alpha +1)}{\sqrt{(\alpha +1)^2+1}}-\frac{1}{2} \alpha (\alpha +2)+\frac{4 \alpha (\alpha +2)}{(\alpha +1)^2+1}+1,

which is negative and does not bring sharpness.

Source and licence

The text on this page is from arXiv:2504.09928v1 by M. Obradovic and N. Tuneski, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.

The record of each item is in analysis/objects/2504/2504.09928.jsonl.