(2) EFRAIMIDIS AND VUKOTI'C. 1503. the small subclasses of the normalized univalent functions 𝑆 such as the Hurwitz and Noshiro–Warschawski classes and also for the closed convex hulls of convex functions and starlike functions, two classes that also contain non-univalent functions. • Thanks to our Lemma 2.1, each result formulated as an inequality that holds for all 𝜆 ∈ ℂ can also be enunciated in an equivalent way as a single new inequality for the coeﬃcients, which could be of some independent interest. • Theorem 3.1 reﬂects a new phenomenon for the generalized functional Φ(𝑓 ) = 𝜆𝑎𝑚 𝑎𝑛 − 𝑎𝑚+𝑛−1 : the sharp bounds obtained diﬀer in an essential way in the case 𝑚 ≠ 𝑛 from the case 𝑚 = 𝑛. • We generalize the results proved by Brown and Tsao [2] and by Ma [17] for starlike functions and also answer a question from [17] on the smallest positive 𝜆 for which Ma's estimates hold. • We show that the generalized Zalcman conjecture is asymptotically true for every complex value of 𝜆 and is also equivalent to other related statements which may provide further insight into the problem. • We improve upon the observation that the Zalcman conjecture implies the Bieberbach conjecture by showing that this implication passes through three related but weaker conjectures than Zalcman's which may be of independent interest.. 2. P R E L I M I NA R I E S. The following simple but very useful lemma for complex numbers will allow us to rephrase several statements (to be proved later) in a diﬀerent language. In this way, inﬁnitely many conditions can typically be replaced by just a single one of diﬀerent type. Lemma 2.1. Let 𝑎, 𝑏 ∈ ℂ be arbitrary and let 𝐶, 𝑀 > 0. Then |𝑎 + 𝜆𝑏| ≤ 𝑀 max{𝐶, |𝜆|},. for all 𝜆 ∈ ℂ. (2.1). if and only if |𝑎| + |𝑏|𝐶 ≤ 𝑀𝐶.. (2.2). Assuming that 𝑎, 𝑏 ≠ 0, equality holds in (2.1) for some 𝜆 ≠ 0 if and only if it holds in (2.2) and also |𝜆| = 𝐶 and arg 𝜆 = arg 𝑎 − arg 𝑏 (taking the values of the argument function modulo 2𝜋). Proof. If (2.1) holds, we can choose 𝜆 with |𝜆| = 𝐶 and arg 𝜆 = arg 𝑎 − arg 𝑏 to get |𝑎| + |𝑏|𝐶 ≤ 𝑀𝐶. Conversely, assuming that |𝑎| + |𝑏|𝐶 ≤ 𝑀𝐶, the triangle inequality yields { } { } { } |𝜆| |𝜆| |𝜆| |𝜆| | | | | | | | | |𝑎 + 𝜆𝑏| ≤ |𝑎| + | ||𝑏|𝐶 ≤ max 1, | | |𝑎| + max 1, | | |𝑏|𝐶 ≤ 𝑀𝐶 max 1, | | = 𝑀 max{𝐶, |𝜆|}. |𝐶 | |𝐶 | |𝐶 | |𝐶 | By inspecting the chains of inequalities in (2.1) and (2.2), it is quite direct to see that equality for 𝜆 ≠ 0 is possible in either case only when |𝜆| = 𝐶 and arg 𝑎 = arg(𝜆𝑏) as claimed. □ On Livingston-type inequalities. The class of analytic functions 𝑔 such that Re𝑔 > 0 in 𝔻, normalized so that 𝑔(0) = 1, is considered frequently in connection with univalent functions. The classical Carathéodory lemma [5, Chap. 2] states that the Taylor coeﬃcients 𝑝𝑛 of any such function 𝑔 must satisfy the sharp inequality |𝑝𝑛 | ≤ 2, 𝑛 ≥ 1. Another inequality: |𝑝𝑛 − 𝑝𝑘 𝑝𝑛−𝑘 | ≤ 2 for the functions in was proved by Livingston [15, Lemma 1]. The following generalization was obtained by the ﬁrst author in [6] and will be crucial in this paper. ∑ 𝑛 Theorem A. If 𝑔 ∈ , 𝑔(𝑧) = 1 + ∞ 𝑛=1 𝑝𝑛 𝑧 and 1 ≤ 𝑘 ≤ 𝑛 − 1 then |𝑝𝑛 − 𝑤𝑝𝑘 𝑝𝑛−𝑘 | ≤ 2 max{1, |1 − 2𝑤|},. for all 𝑤 ∈ ℂ,. (2.3). and the inequality is sharp. We can now deduce an inequality which may be of independent interest. It can be seen as a generalization of the well-known estimate: | | |𝑝2 − 1 𝑝2 | + 1 |𝑝1 |2 ≤ 2, | 1| 2 | | 2 which can be found in [19, p. 166] and can also be deduced from the classical Schwarz–Pick lemma..

(3) EFRAIMIDIS AND VUKOTI'C. 1504. Proposition 2.2. If 𝑔 ∈ , 𝑔(𝑧) = 1 +. ∑∞. 𝑛 𝑛=1 𝑝𝑛 𝑧. and 1 ≤ 𝑘 ≤ 𝑛 − 1 then. | | |𝑝 − 1 𝑝 𝑝 | + 1 |𝑝 𝑝 | ≤ 2. | 𝑛 2 𝑘 𝑛−𝑘 | 2 𝑘 𝑛−𝑘 | |. (2.4). The inequality is sharp. Proof. Rewriting inequality (2.3) from Theorem A in the form |2𝑝𝑛 − 𝑝𝑘 𝑝𝑛−𝑘 + (1 − 2𝑤)𝑝𝑘 𝑝𝑛−𝑘 | ≤ 4 max{1, |1 − 2𝑤|}, □. the statement follows by Lemma 2.1.. We note that the inequality stated in Proposition 2.2 also appeared (with a diﬀerent proof) in Campschroer's thesis [3, §1.4] which contains various interesting ideas on extremal problems. The combined use of Theorem A, Lemma 2.1, and Proposition 2.2 will be the key to a number of results throughout this paper.. 3. SHARP EST I MAT E S FOR SO M E SP ECIAL CLAS S ES. In this section we will obtain various estimates on the generalized Zalcman functional Φ(𝑓 ) = 𝜆𝑎𝑚 𝑎𝑛 − 𝑎𝑚+𝑛−1 with complex values 𝜆. We do this for four diﬀerent classes of functions which are either subclasses of 𝑆 or closed convex hulls of important subclasses of 𝑆 (which also contain non-univalent functions). All estimates are sharp and each one of them is also formulated in an equivalent way The Hurwitz class. The name Hurwitz class is often used to denote the set of all functions 𝑓 of the form 𝑓 (𝑧) = 𝑧 + 𝑎2 𝑧2 + 𝑎3 𝑧3 + … , ∑ analytic in 𝔻 and with the property that ∞ 𝑛=2 𝑛|𝑎𝑛 | ≤ 1. Obviously, the 𝑛-th coeﬃcient of a function in𝑛 is subject to the estimate |𝑎𝑛 | ≤ 1∕𝑛 for each 𝑛. The simplest example of a function in is the polynomial 𝑃𝑛 (𝑧) = 𝑧 + 𝑧𝑛 , 𝑛 ≥ 2. It is a wellknown exercise that ⊂ 𝑆. The reader is referred to [8] for further properties of . For the functions in this class we obtain a much smaller bound on the Zalcman functional than for the entire class 𝑆. We stress the diﬀerence between items (a) and (b) of the theorem below: the estimates on the functional Φ(𝑓 ) = 𝜆𝑎𝑚 𝑎𝑛 − 𝑎𝑚+𝑛−1 diﬀer in an essential way in the cases 𝑚 = 𝑛 and 𝑚 ≠ 𝑛, with the presence of an extra factor of four in the denominator in the latter case. Theorem 3.1. (a) If 𝑓 ∈ and 𝑛 ≥ 2 then the following inequality holds for the coeﬃcients of 𝑓 : 𝑛2 |𝑎2𝑛 | + (2𝑛 − 1)|𝑎2𝑛−1 | ≤ 1.. (3.1). This single inequality is equivalent to |𝜆 𝑎2𝑛. { − 𝑎2𝑛−1 | ≤ max. |𝜆| 1 , 2 𝑛 2𝑛 − 1. } ,. for all 𝜆 ∈ ℂ.. (3.2). Equality holds if and only if 𝛼 ⎧ 2𝑛−1 , ⎪𝑧 + 2𝑛 − 1 𝑧 𝑓 (𝑧) = ⎨ 𝛼 ⎪𝑧 + 𝑧𝑛 , 𝑛 ⎩. for |𝜆| ≤. 𝑛2 , 2𝑛−1. for |𝜆| ≥. 𝑛2 , 2𝑛−1. where 𝛼 is a complex number of modulus one. (b) If 𝑓 ∈ , then for any two distinct values 𝑚, 𝑛 ≥ 2 we have 4𝑚𝑛|𝑎𝑚 𝑎𝑛 | + (𝑚 + 𝑛 − 1)|𝑎𝑚+𝑛−1 | ≤ 1.. (3.3).

(4) EFRAIMIDIS AND VUKOTI'C. 1505. The last inequality is equivalent to { |𝜆 𝑎𝑚 𝑎𝑛 − 𝑎𝑚+𝑛−1 | ≤ max. |𝜆| 1 , 4𝑚𝑛 𝑚 + 𝑛 − 1. } ,. for all 𝜆 ∈ ℂ.. (3.4). Equality holds if and only if 𝛼 ⎧𝑧 + 𝑧𝑚+𝑛−1 , for |𝜆| ≤ 𝑚+𝑛−1 ⎪ 𝑓 (𝑧) = ⎨ 𝛽 𝑛 ⎪𝑧 + 𝛼 𝑧𝑚 + for |𝜆| ≥ 𝑧 , ⎩ 2𝑚 2𝑛. 4𝑚𝑛 , 𝑚+𝑛−1 4𝑚𝑛 , 𝑚+𝑛−1. where 𝛼 and 𝛽 are complex numbers such that |𝛼| = |𝛽| = 1. Proof. (a) By the deﬁnition of we have that 𝑛|𝑎𝑛 | ≤ 1 and therefore 𝑛2 |𝑎𝑛 |2 + (2𝑛 − 1)|𝑎2𝑛−1 | ≤ 𝑛|𝑎𝑛 | + (2𝑛 − 1)|𝑎2𝑛−1 | ≤ 1. Taking 𝑀=. 1 , 𝑛2. 𝐶=. 𝑛2 2𝑛 − 1. in Lemma 2.1, the above inequality is equivalent to (3.2). Obviously, equality is only possible when 𝑛|𝑎𝑛 | = 1 or 𝑛|𝑎𝑛 | = 0. The ﬁrst case implies that 𝑎2𝑛−1 = 0 and all remaining coeﬃcients are zero. The second yields that (2𝑛 − 1)|𝑎2𝑛−1 | = 1 and all remaining coeﬃcients are zero, which easily leads to the desired conclusion. (b) The proof is slightly more involved in the case 𝑚 ≠ 𝑛. Set 𝑥 = 𝑚|𝑎𝑚 | and 𝑦 = 𝑛|𝑎𝑛 |. Clearly 𝑥, 𝑦 ≥ 0 and by the deﬁnition of they satisfy 𝑥 + 𝑦 ≤ 1. This and (𝑥 − 𝑦)2 ≥ 0 imply 4𝑥𝑦 ≤ (𝑥 + 𝑦)2 ≤ 𝑥 + 𝑦. It follows readily from the deﬁnition of that 4𝑚𝑛|𝑎𝑚 𝑎𝑛 | + (𝑚 + 𝑛 − 1)|𝑎𝑚+𝑛−1 | ≤ 1. Using Lemma 2.1 with 𝑀=. 1 , 4𝑚𝑛. 𝐶=. 4𝑚𝑛 𝑚+𝑛−1. we see that this is equivalent to (3.4). Equality holds in (b) if and only if either 𝑚|𝑎𝑚 | = 𝑛|𝑎𝑛 | = 0 or 𝑚|𝑎𝑚 | = 𝑛|𝑎𝑛 | = 1∕2, which again easily leads to the claim on extremal functions. □ The Noshiro–Warschawski class. We now consider the functions in the normalized class } { = 𝑓 ∈ (𝔻) ∶ Re𝑓 ′ (𝑧) > 0, 𝑓 (0) = 0, 𝑓 ′ (0) = 1 . 1 − 𝑧 whose derivative is 𝑓 ′ (𝑧) = (1 + 𝑧)∕(1 − 𝑧), a mapping of 𝔻 1−𝑧 onto the right half-plane. The branch of the logarithm is chosen so that log 1 = 0. Note that ⊂ 𝑆 by the basic Noshiro–Warschawski lemma [5, Theorem 2.16]. MacGregor [18] showed that for 𝑓 in we have |𝑎𝑛 | ≤ 2∕𝑛. On the other hand, contains the Hurwitz class . This can be seen as follows. If 𝑓 is a function in other than the identity, then 𝑓 ′ (0) = 1 and, when 𝑧 ≠ 0, we have the strict inequality A typical example of a function in is 𝑓 (𝑧) = 2 log. Re 𝑓 ′ (𝑧) = 1 +. ∞ ∑ 𝑛=2. ∞ ∞ ∑ ∑ { } 𝑛 Re 𝑎𝑛 𝑧𝑛−1 ≥ 1 − 𝑛|𝑎𝑛 ||𝑧|𝑛−1 > 1 − 𝑛|𝑎𝑛 | ≥ 0. 𝑛=2. 𝑛=2. It should not be too surprising to have larger upper bounds for the generalized Zalcman functional among the functions in than for those in . This is indeed the case, as our next result shows..

(5) EFRAIMIDIS AND VUKOTI'C. 1506. Theorem 3.2. Let 𝑓 ∈ and let 𝑚, 𝑛 ≥ 2. Then the following inequality holds for the coeﬃcients of 𝑓 : | | 𝑚𝑛|𝑎𝑚 𝑎𝑛 | 2 𝑚𝑛 | | | 2(𝑚 + 𝑛 − 1) 𝑎𝑚 𝑎𝑛 − 𝑎𝑚+𝑛−1 | + 2(𝑚 + 𝑛 − 1) ≤ 𝑚 + 𝑛 − 1 . | | This is equivalent to 2 |𝜆𝑎𝑚 𝑎𝑛 − 𝑎𝑚+𝑛−1 | ≤ max 𝑚+𝑛−1. {. } | 𝑚 + 𝑛 − 1 || | 1, |1 − 2𝜆 𝑚𝑛 || |. for all 𝜆 ∈ ℂ.. Equality holds in both inequalities for the function 𝑓 (𝑧) = 2 log. 1 −𝑧 1−𝑧. 𝑓 (𝑧) =. 1 + 𝜁 𝑚+𝑛−2 𝑑𝜁 1 − 𝜁 𝑚+𝑛−2. (3.5). | | when |1 − 2𝜆 𝑚+𝑛−1 | ≥ 1 and for 𝑚𝑛 | | ∫[0,𝑧]. | | (meaning integration over the segment from 0 to 𝑧) when |1 − 2𝜆 𝑚+𝑛−1 | < 1. 𝑚𝑛 | | ∑∞ ∑ 𝑛 ′ 𝑛 Proof. Let 𝑓 ∈ , 𝑓 (𝑧) = 𝑧 + ∞ 𝑛=2 𝑎𝑛 𝑧 in 𝔻. Then 𝑔 = 𝑓 ∈ and, writing 𝑔(𝑧) = 1 + 𝑛=1 𝑝𝑛 𝑧 , the coeﬃcients of 𝑓 and 𝑔 are related by 𝑝𝑛−1 = 𝑛𝑎𝑛 . The desired inequalities now follow from Theorem A and Proposition 2.2. The function given by (3.5) has coeﬃcients 2∕𝑛 and yields equality in the cases indicated. For the remaining case, when | 𝑚+𝑛−1 | |1 − 2𝜆 𝑚𝑛 | < 1, we ﬁnd that the function | | 𝑓 ′ (𝑧) =. 1 + 𝑧𝑚+𝑛−2 1 − 𝑧𝑚+𝑛−2. belongs to . Since 𝑓 (0) = 0, it follows that 𝑓 ∈ . Clearly, 𝑓 (𝑧) = 𝑧 +. ∞ ∑. 2 𝑧𝑘(𝑚+𝑛−2)+1 , 𝑘(𝑚 + 𝑛 − 2) + 1 𝑘=1 □. and it is easily checked that equality is attained for this function.. We observe that one can write down an explicit formula for the extremal function written above as a primitive but there is really no need for this. The closed convex hull of convex functions. Denote by 𝐶 the class of convex functions in 𝑆. A typical example is the half𝑧 . It is well known that the coeﬃcient estimate can be improved a great deal for the functions in 𝐶: by plane function 𝓁(𝑧) = 1−𝑧 a theorem of Loewner, they must satisfy |𝑎𝑛 | ≤ 1, with equality only for the function 𝓁 and its rotations (see [5, Corollary on p. 45]). Denote by co(𝐶) the convex hull of 𝐶 and by co(𝐶) its closure in the topology of uniform convergence on compact subsets of 𝔻. Note that this larger class no longer consists exclusively of univalent functions. A well-known result of Marx and Strohhäcker [19, p. 45] implies that { ( ) } co(𝐶) = 𝑓 ∈ 𝐻(𝔻) ∶ Re 𝑓 (𝑧)∕𝑧 > 1∕2, 𝑓 (0) = 𝑓 ′ (0) − 1 = 0 . Thus, a connection with the class is readily established by the formula { 𝑓 (𝑧) − 1, 2 𝑔(𝑧) = 𝑧 1,. for 𝑧 ≠ 0, for 𝑧 = 0,. that is, 𝑓 ∈ co(𝐶) if and only if 𝑔 ∈ . For real parameters 𝜆 and in the case when 𝑚 = 𝑛, the inequality in the following theorem appeared in [7] for 0 ≤ 𝜆 ≤ 2 and in [14] for 𝜆 ≥ 2. Here we give a complete answer for all complex 𝜆 and all 𝑚, 𝑛 ≥ 2..

(6) EFRAIMIDIS AND VUKOTI'C. 1507. Theorem 3.3. Let 𝑓 be in co(𝐶) and 𝑚, 𝑛 ≥ 2. Then |𝑎𝑚 𝑎𝑛 − 𝑎𝑚+𝑛−1 | + |𝑎𝑚 𝑎𝑛 | ≤ 1. This is equivalent to the following statement: |𝜆𝑎𝑚 𝑎𝑛 − 𝑎𝑚+𝑛−1 | ≤ max{1, |1 − 𝜆|},. for all 𝜆 ∈ ℂ.. Equality holds in both inequalities for the function given by 𝑧 1−𝑧. (3.6). 𝑧 1 − 𝑧𝑚+𝑛−2. (3.7). 𝑓 (𝑧) = when |1 − 𝜆| ≥ 1 and for 𝑓 (𝑧) = when |1 − 𝜆| < 1. Proof. The function 𝑔 given by 𝑔(𝑧) = 2. ∞ ∑ 𝑓 (𝑧) 𝑝𝑛 𝑧𝑛 , −1=1+ 𝑧 𝑛=1. 𝑧 ≠ 0,. 𝑔(0) = 1,. belongs to and the coeﬃcients of the functions 𝑓 and 𝑔 are related by 𝑝𝑛−1 = 2𝑎𝑛 . Theorem A yields the desired inequality in 𝜆 and the equivalent formulation as a single inequality follows by Lemma 2.1. The function given by (3.6) clearly yields equality in the cases indicated. For the remaining case, when |1 − 𝜆| < 1, we consider the function 𝑔(𝑧) =. 1 + 𝑧𝑚+𝑛−2 , 1 − 𝑧𝑚+𝑛−2. which belongs to . Let 𝑓 be the function in co(𝐶) for which 𝑔(𝑧) = 2𝑓 (𝑧)∕𝑧 − 1. We see that ∞. 𝑓 (𝑧) =. ∑ 𝑧 = 𝑧𝑘(𝑚+𝑛−2)+1 𝑚+𝑛−2 1−𝑧 𝑘=0 □. and equality is attained for this function.. A further generalization. More generally, one can consider the class 𝐶(𝛼) of analytic functions in 𝔻 of the form 𝑓 (𝑧) = 𝑧 + 𝑎2 𝑧2 + 𝑎3 𝑧3 + ⋯ which satisfy ( ) 𝑧𝑓 ′′ Re 1 + ′ > 𝛼. 𝑓 This class was introduced by Robertson in [21, Section 3], cf. also [9, §2.3]. Of course, 𝐶(0) = 𝐶, the class of convex functions. Since these classes become smaller as 𝛼 increases, all functions in 𝐶(𝛼) are univalent and convex whenever 0 ≤ 𝛼 < 1. When −1∕2 ≤ 𝛼 < 0 these functions are known to be univalent and convex in one direction [23]. The function given by ⎧ 1 − (1 − 𝑧)2𝛼−1 , for 𝛼 ≠ 1∕2, ⎪ 2𝛼 − 1 𝑓𝛼 (𝑧) = ⎨ ⎪log 1 , for 𝛼 = 1∕2, ⎩ 1−𝑧 is often extremal in this class. Its coeﬃcients are easily computed: 𝑛. 𝐴𝑛 =. Γ(𝑛 + 1 − 2𝛼) 1 ∏ (𝑘 − 2𝛼). = 𝑛! Γ(2 − 2𝛼) 𝑛! 𝑘=2.

(7) EFRAIMIDIS AND VUKOTI'C. 1508. It is known that |𝑎𝑛 | ≤ 𝐴𝑛 for functions in 𝐶(𝛼) (see [21] for 0 ≤ 𝛼 < 1 and [22] for starlike functions of any order 𝛼 < 1, which directly implies the inequality considered here through Alexander's Theorem). Arguments similar to those used earlier allow us to recover, without much eﬀort, a recent theorem from [14]; thus, we omit several details below. Theorem 3.4. Let 𝛼 < 1, 𝑓 ∈ co(𝐶(𝛼)), 𝑚, 𝑛 ≥ 2, and let 𝐴𝑛 be as above. Then |𝑎 𝑎 𝑎𝑚+𝑛−1 || |𝑎𝑚 𝑎𝑛 | | 𝑚 𝑛 − ≤ 1. | |+ | 𝐴𝑚 𝐴𝑛 𝐴𝑚+𝑛−1 | 𝐴𝑚 𝐴𝑛 | | This is equivalent to the following statement: { } |𝜆𝑎𝑚 𝑎𝑛 − 𝑎𝑚+𝑛−1 | ≤ max 𝐴𝑚+𝑛−1 , |𝜆𝐴𝑚 𝐴𝑛 − 𝐴𝑚+𝑛−1 | ,. for all 𝜆 ∈ ℂ.. Equality holds in both inequalities above for the function given by 𝑓 = 𝑓𝛼 in the case when |𝜆𝐴𝑚 𝐴𝑛 − 𝐴𝑚+𝑛−1 | ≥ 𝐴𝑚+𝑛−1 and for the function 𝑓 (𝑧) =. 𝑚+𝑛−2 ( 2𝜋𝑘𝑖 ) ∑ − 2𝜋𝑘𝑖 1 𝑒 𝑚+𝑛−2 𝑓𝛼 𝑒 𝑚+𝑛−2 𝑧 𝑚 + 𝑛 − 2 𝑘=1. in the case when |𝜆𝐴𝑚 𝐴𝑛 − 𝐴𝑚+𝑛−1 | < 𝐴𝑚+𝑛−1 . Proof. Theorem 4 in [1] provides the following Herglotz-type representation: there exists a probability measure 𝜇 on 𝕋 such that 𝑓 ′ (𝑧) =. 𝑑𝜇(𝜆) ∫𝕋 (1 − 𝜆𝑧)2−2𝛼. for every 𝑓 ∈ co(𝐶(𝛼)). From here the coeﬃcients of such 𝑓 relate with those of a function in by 𝑎𝑛 =. 𝐴𝑛 𝑝 . 2 𝑛−1 □. The desired result now follows from Theorem A and Proposition 2.2 as before.. The closed convex hull of starlike functions. A set 𝐸 is said to be starlike with respect to the origin if for every 𝑧 ∈ 𝐸 the entire segment [0, 𝑧] is contained in 𝐸. A function 𝑓 is said to be starlike if it is a univalent function of the disk onto a domain starlike with respect to the origin. The usual notation for the subclass of 𝑆 consisting of all starlike functions is 𝑆 ∗ . Obviously, 𝐶 ⊂ 𝑆 ∗ ⊂ 𝑆. Brown and Tsao [2, Theorem 2] showed that the Zalcman conjecture is true for starlike functions and Ma [17, Theorem 2.3] generalized their result further to show that |𝜆𝑎𝑚 𝑎𝑛 − 𝑎𝑚+𝑛−1 | ≤ 𝜆 𝑚𝑛 − 𝑚 − 𝑛 + 1 whenever 𝜆 ∈ ℝ and 𝜆 ≥ 𝜆0 = 2(𝑚+𝑛−1) . The following result generalizes his result to the case of complex parameters and at the 𝑚𝑛 same time answers in the aﬃrmative his question posed in [17] as to whether 𝜆0 is the smallest positive number for which the above bound remains true. Theorem 3.5. Let 𝑓 ∈ co(𝑆 ∗ ) and let 𝑚, 𝑛 ≥ 2. Then | 𝑎𝑚 𝑎𝑛 𝑎𝑚+𝑛−1 | |𝑎𝑚 𝑎𝑛 | | | | 𝑚𝑛 − 𝑚 + 𝑛 − 1 | + 𝑚𝑛 ≤ 1. | | This statement is equivalent to |𝜆𝑎𝑚 𝑎𝑛 − 𝑎𝑚+𝑛−1 | ≤ (𝑚 + 𝑛 − 1) max. { } | | 𝑚𝑛 1, ||1 − 𝜆|| , 𝑚+𝑛−1 | |. for all 𝜆 ∈ ℂ..

(8) EFRAIMIDIS AND VUKOTI'C. 1509. In both cases, equality holds for the function given by 𝑧 (1 − 𝑧)2. (3.8). 𝑧𝑚+𝑛−1 𝑧 + (𝑚 + 𝑛 − 2) 𝑚+𝑛−2 1−𝑧 (1 − 𝑧𝑚+𝑛−2 )2. (3.9). 𝑓 (𝑧) = | when |1 − |. | 𝑚𝑛 𝜆| 𝑚+𝑛−1 |. ≥ 1 and for 𝑓 (𝑧) =. | when |1 − |. | 𝑚𝑛 𝜆| 𝑚+𝑛−1 |. < 1.. Proof. By Alexander's theorem [5, §2.5] we know that every starlike function 𝑓 is of the form 𝑓 (𝑧) = 𝑧ℎ′ (𝑧) for some ℎ in 𝐶. Such a relation is preserved upon taking convex combinations and uniform limits on compact subsets of the disk, hence we obtain the same conclusion for every function 𝑓 in co(𝑆 ∗ ) and some corresponding ℎ in co(𝐶). Next, taking into account the connection between the classes co(𝐶) and we readily get that for every 𝑓 in co(𝑆 ∗ ) there is a function 𝑔 in such that ( )′ ) 𝑧 + 𝑧𝑔(𝑧) 𝑧( 𝑓 (𝑧) = 𝑧ℎ′ (𝑧) = 𝑧 = 1 + 𝑔(𝑧) + 𝑧𝑔 ′ (𝑧) . 2 2 ∑ 𝑛𝑝𝑛−1 𝑛 Writing 𝑔(𝑧) = 1 + ∞ , 𝑛 ≥ 2. Now, Theorem A yields the ﬁrst and Proposition 2.2 𝑛=1 𝑝𝑛 𝑧 we can easily deduce that 𝑎𝑛 = 2 the second of the two inequalities. We note that the Koebe function (3.8) clearly satisﬁes the equality in the cases indicated. For the remaining case, when | | 𝑚𝑛 |1 − 𝑚+𝑛−1 𝜆| < 1, we consider the function | | 𝑔(𝑧) =. 1 + 𝑧𝑚+𝑛−2 , 1 − 𝑧𝑚+𝑛−2. which belongs to . Hence, in view of the above computation, the function 𝑓 = 2𝑧 (1 + 𝑔 + 𝑧𝑔 ′ ) belongs to co(𝑆 ∗ ) and has the form (3.9). We now compute 𝑓 (𝑧) = 𝑧 +. ∞ ∑ (. ) 𝑘(𝑚 + 𝑛 − 2) + 1 𝑧𝑘(𝑚+𝑛−2)+1 .. 𝑘=1. Clearly, equality is attained for this function in both inequalities.. □. The class co(𝑆 ∗ ) is obviously strictly larger than 𝑆 ∗ and it turns out that, in the simplest case 𝜆 = 1, the above Theorem 3.5 yields the sharp bound |𝑎𝑚 𝑎𝑛 − 𝑎𝑚+𝑛−1 | ≤ max{𝑚 + 𝑛 − 1, (𝑚 − 1)(𝑛 − 1)}, which is diﬀerent from (𝑚 − 1)(𝑛 − 1) when either 𝑚 = 2 or 𝑚 = 𝑛 = 3; this is explained in [17]. In particular, when 𝑚 = 𝑛 ∈ {2, 3} we have the estimate |𝑎2𝑛 − 𝑎2𝑛−1 | ≤ 2𝑛 − 1. In this case, 2𝑛 − 1 > (𝑛 − 1)2 , the general estimate in the Zalcman conjecture (also conﬁrmed by Brown and Tsao for starlike functions). However, there is no contradiction since the class co(𝑆 ∗ ) also contains non-univalent functions.. 4. S OME GEN E R A L CON SI D E R ATIO NS. An asymptotic version of the Zalcman conjecture. Let 𝑓 ∈ 𝑆 and let 𝑀∞ (𝑟, 𝑓 ) = max|𝑧|=𝑟 |𝑓 (𝑧)|. Recall that the Hayman index of 𝑓 is the number 𝛼 = lim(1 − 𝑟)2 𝑀∞ (𝑟, 𝑓 ). 𝑟→1. It is well known [5, p. 157] that 0 ≤ 𝛼 ≤ 1. Moreover, Hayman's regularity theorem [5, Theorem 5.6] asserts that for each 𝑓 in 𝑆 its 𝑛-th Taylor coeﬃcient 𝑎𝑛 satisﬁes lim𝑛→∞ |𝑎𝑛 ∕𝑛| = 𝛼..

(9) EFRAIMIDIS AND VUKOTI'C. 1510. Even though the Zalcman conjecture continues to be an open problem, we now show that its asymptotic version is true and we give it in a precise quantitative form. Theorem 4.1. Let 𝑓 (𝑧) = 𝑧 + 𝑎2 𝑧2 + ⋯ be in 𝑆, with Hayman index 𝛼, and let 𝜆 ∈ ℂ. Then lim. 𝑚,𝑛→∞. |𝜆𝑎𝑚 𝑎𝑛 − 𝑎𝑚+𝑛−1 | = 𝛼2. |𝜆𝑚𝑛 − 𝑚 − 𝑛 + 1|. (4.1). Also, if we deﬁne 𝐵𝑚,𝑛 (𝜆) = sup𝑓 ∈𝑆 |𝜆𝑎𝑚 𝑎𝑛 − 𝑎𝑚+𝑛−1 |, then lim. 𝑚,𝑛→∞. 𝐵𝑚,𝑛 (𝜆) |𝜆𝑚𝑛 − 𝑚 − 𝑛 + 1|. = 1.. In both limits, we understand that (𝑚, 𝑛) → (∞, ∞) unconditionally in ℕ2 (meaning that 𝑚 + 𝑛 → ∞). Proof. Applying the triangle inequality we get |𝜆𝑎𝑚 𝑎𝑛 − 𝑎𝑚+𝑛−1 | |𝑎𝑚+𝑛−1 | |𝑎 𝑎 | |𝜆|𝑚𝑛 𝑚+𝑛−1 ≤ 𝑚 𝑛 + , |𝜆𝑚𝑛 − 𝑚 − 𝑛 + 1| 𝑚𝑛 |𝜆𝑚𝑛 − 𝑚 − 𝑛 + 1| 𝑚 + 𝑛 − 1 |𝜆𝑚𝑛 − 𝑚 − 𝑛 + 1| where the right-hand side converges to 𝛼 2 in view of Hayman's regularity theorem. Analogously, we can use the triangle inequality to get a lower bound converging to 𝛼 2 . Hence (4.1) follows. The Koebe function clearly shows that 𝐵𝑚,𝑛 (𝜆) ≥ |𝜆 𝑚𝑛 − 𝑚 − 𝑛 + 1|. Using the customary notation 𝐴𝑛 = sup𝑓 ∈𝑆 |𝑎𝑛 |, we have 1≤. 𝐵𝑚,𝑛 (𝜆) |𝜆 𝑚𝑛 − 𝑚 − 𝑛 + 1|. ≤. |𝜆|𝐴𝑚 𝐴𝑛 + 𝐴𝑚+𝑛−1 → 1, |𝜆 𝑚𝑛 − 𝑚 − 𝑛 + 1|. when (𝑚, 𝑛) → (∞, ∞).. □. As is usual [5, Chap. 2], by a rotation of a function 𝑓 in 𝑆 we mean the function 𝑓𝑐 (𝑧) = 𝑐𝑓 (𝑐𝑧), |𝑐| = 1, which is again in 𝑆. Note that the rotations of the Koebe function give equality in Zalcman's conjecture. ) ( Corollary 4.2. If 𝑓 ∈ 𝑆 is not a rotation of the Koebe function, then for every 𝛿 ∈ 0, 1 − 𝛼 2 there exist 𝑚0 and 𝑛0 in ℕ (which depend on 𝑓 ) such that |𝜆𝑎𝑚 𝑎𝑛 − 𝑎𝑚+𝑛−1 | ≤ (1 − 𝛿)|𝜆𝑚𝑛 − 𝑚 − 𝑛 + 1|, for all 𝑚 ≥ 𝑚0 , 𝑛 ≥ 𝑛0 . Some equivalent reformulations of the Zalcman conjecture. For the sake of simplicity, we treat only the original conjecture: |𝑎2𝑛 − 𝑎2𝑛−1 | ≤ (𝑛 − 1)2 . We ﬁrst recall that, if assumed true for all 𝑛, it easily implies the Bieberbach conjecture (now de Branges' theorem). Since the proof of this implication for one value of 𝑛 uses the validity of the conjecture for another 𝑛, in order to avoid this discussion in the sequel, we shall simply take for granted the Bieberbach conjecture for odd integers: |𝑎2𝑛−1 | ≤ 2𝑛 − 1. With this in mind, the Zalcman conjecture can be reformulated in several ways. ∑ 𝑘 Theorem 4.3. Let 𝑓 ∈ 𝑆 be ﬁxed, 𝑓 (𝑧) = 𝑧 + ∞ 𝑘=2 𝑎𝑘 𝑧 , and let 𝑛 ≥ 2 be arbitrary. Then the following statements are equivalent: (a) The Zalcman conjecture holds: |𝑎2𝑛 − 𝑎2𝑛−1 | ≤ (𝑛 − 1)2 = 𝑛2 − (2𝑛 − 1);. (b) |𝑎2𝑛 − 𝑡𝑎2𝑛−1 | ≤ 𝑛2 − 𝑡(2𝑛 − 1) for all 𝑡 ∈ [0, 1];. (c) |𝑎2𝑛 − 𝑎2𝑛−1 | + 𝑟|𝑎2𝑛−1 | ≤ (𝑛 − 1)2 + 𝑟(2𝑛 − 1) for all 𝑟 > 0;. (d) |𝑎2𝑛 − 𝑤𝑎2𝑛−1 | ≤ (𝑛 − 1)2 + |𝑤 − 1|(2𝑛 − 1) for all 𝑤 ∈ ℂ.. Proof. We will show that (b) ⇒ (a) ⇒ (c) ⇒ (d) ⇒ (b). Of course, other schemes of proof are also possible. (𝑏) ⇒ (𝑎) . This implication is trivial. (𝑎) ⇒ (𝑐) . Suppose that (a) holds. In view of the inequality |𝑎2𝑛−1 | ≤ 2𝑛 − 1, we deduce directly from (a) that |𝑎2𝑛 − 𝑎2𝑛−1 | + 𝑟|𝑎2𝑛−1 | ≤ (𝑛 − 1)2 + 𝑟(2𝑛 − 1) for all 𝑟 > 0..

(10) EFRAIMIDIS AND VUKOTI'C. 1511. (𝑐) ⇒ (𝑑) . Suppose that |𝑎2𝑛 − 𝑎2𝑛−1 | + 𝑟|𝑎2𝑛−1 | ≤ (𝑛 − 1)2 + 𝑟(2𝑛 − 1) holds for all 𝑟 > 0 (hence, by taking limits, also for 𝑟 = 0). Let 𝑤 be arbitrary. If 𝑤 = 1 then (d) follows from the assumption for 𝑟 = 0. For every other value of 𝑤 there is a positive 𝑟 such that |𝑤 − 1| = 𝑟 and we get |𝑎2𝑛 − 𝑤𝑎2𝑛−1 | = |𝑎2𝑛 − 𝑎2𝑛−1 + (1 − 𝑤)𝑎2𝑛−1 | ≤ |𝑎2𝑛 − 𝑎2𝑛−1 | + 𝑟|𝑎2𝑛−1 | ≤ (𝑛 − 1)2 + 𝑟(2𝑛 − 1) = (𝑛 − 1)2 + |𝑤 − 1|(2𝑛 − 1), and (d) is proved. (𝑑) ⇒ (𝑏) . This follows readily by taking 𝑤 = 𝑡 ∈ [0, 1].. □. Several remarks are in order to show that Theorem 4.3 may shed some new light on the problem. • In view of Theorem 4.3, proving the Zalcman conjecture amounts to proving any of the equivalent statements while disproving it would amount to ﬁnding one single example of a function which does not satisfy one of the inequalities (b), (c), or (d) for one single value of 𝑡, 𝑟, or 𝑤 respectively. • Statement (b) in the theorem had already been veriﬁed for the typically real functions and follows from [2, Theorem 1]. • The fact that Bieberbach's conjecture is true means that (d) holds for 𝑤 = 0. If Zalcman's conjecture were to be true, we would have many more new inequalities such as, for example, |𝑎2𝑛 − 2𝑎2𝑛−1 | ≤ 𝑛2 , obtained by taking 𝑤 = 2 in (d). • We also note that the validity of Bieberbach's conjecture readily implies that (d) is true for any 𝑤 = −𝑀, where 𝑀 is real and positive; indeed: |𝑎2𝑛 + 𝑀𝑎2𝑛−1 | ≤ 𝑛2 + 𝑀(2𝑛 − 1) = (𝑛 − 1)2 + (𝑀 + 1)(2𝑛 − 1). However, we do not know whether (d) is true in general for any other value of 𝑤 except for those in (−∞, 0]. So there appears to be a signiﬁcant gap between Bieberbach and Zalcman. Three related but weaker conjectures. At this point it seems natural to formulate three closely related conjectures. They could be of interest since they are all weaker than Zalcman's but each of them also implies the Bieberbach conjecture. In relation to condition (b) of our preceding theorem, for a given value 𝑡 in [0, 1] we will denote by (𝐵𝑡 ) the following statement: |𝑎2𝑛 − 𝑡𝑎2𝑛−1 | ≤ 𝑛2 − 𝑡(2𝑛 − 1), for all 𝑓 ∈ 𝑆 with 𝑓 (𝑧) = 𝑧 + as follows.. ∑∞. 𝑘 𝑘=2 𝑎𝑘 𝑧 , and all 𝑛. (Bt ). ≥ 2. Thus, we can formulate the ﬁrst weak version of the Zalcman conjecture. Conjecture 1. There exists 𝑡 ∈ (0, 1] such that (𝐵𝑡 ) holds. It is not clear in any obvious way that this statement is true. However, (𝐵0 ) is precisely the Bieberbach conjecture and we know it is true. Thus, the set of all 𝑡 ∈ [0, 1] for which (𝐵𝑡 ) holds is non-empty. It is easy to see that this set is closed as the deﬁning condition contains a non-strict inequality. It is also convex; indeed, if (𝐵𝑠 ) and (𝐵𝑡 ) hold and 𝛼, 𝛽 ∈ [0, 1] with 𝛼 + 𝛽 = 1 then clearly |𝑎2𝑛 − (𝛼𝑠 + 𝛽𝑡)𝑎2𝑛−1 | ≤ 𝛼|𝑎2𝑛 − 𝑠𝑎2𝑛−1 | + 𝛽|𝑎2𝑛 − 𝑡𝑎2𝑛−1 | ≤ 𝛼(𝑛2 − 𝑠(2𝑛 − 1)) + 𝛽(𝑛2 − 𝑡(2𝑛 − 1)) = 𝑛2 − (𝛼𝑠 + 𝛽𝑡)(2𝑛 − 1),.

(11) EFRAIMIDIS AND VUKOTI'C. 1512. hence (𝐵𝛼𝑠+𝛽𝑡 ) is also true. Thus, it seems natural to consider the quantity 𝑇 = sup{𝑡 ∈ [0, 1] ∶ (𝐵𝑡 ) is true }. With this notation, the Zalcman conjecture claims that 𝑇 = 1, while the weak Zalcman conjecture only claims that 𝑇 > 0. Now consider the situation when condition (c) in Theorem 4.3 holds only for some 𝑟 > 0. So for a ﬁxed 𝑟 > 0 we can consider the statement (𝐶𝑟 ): |𝑎2𝑛 − 𝑎2𝑛−1 | + 𝑟|𝑎2𝑛−1 | ≤ (𝑛 − 1)2 + 𝑟(2𝑛 − 1), for all 𝑓 ∈ 𝑆 with 𝑓 (𝑧) = 𝑧 + conjecture.. ∑∞. 𝑘 𝑘=2 𝑎𝑘 𝑧 ,. (Cr ). and all 𝑛 ≥ 2. This clearly gives rise to the second weak version of the Zalcman. Conjecture 2. There exists 𝑟 ∈ [0, 1] such that (𝐶𝑟 ) holds. It also makes sense to consider a weaker version of condition (d) in Theorem 4.3. For a ﬁxed 𝑟, say 𝑟 ∈ [0, 1], consider |𝑎2𝑛 − 𝑤𝑎2𝑛−1 | ≤ (𝑛 − 1)2 + |𝑤 − 1|(2𝑛 − 1), for all 𝑓 ∈ 𝑆 with 𝑓 (𝑧) = 𝑧 +. ∑∞. 𝑘 𝑘=2 𝑎𝑘 𝑧 ,. for all 𝑤 with |𝑤 − 1| = 𝑟,. (Dr ). and all 𝑛 ≥ 2. Thus, we have the third weak version of the Zalcman conjecture.. Conjecture 3. There exists 𝑟 ∈ [0, 1] such that (𝐷𝑟 ) holds. The following relationship exists between the conjectures mentioned. Theorem 4.4. Assume only a weaker statement than the Bieberbach conjecture, for example, Littlewood's theorem [5, Theorem 2.8]: |𝑎𝑛 | < 𝑒𝑛, for all 𝑛 ≥ 2. Under these assumptions we have: (a) The Zalcman conjecture implies Conjecture 3. (b) Conjecture 3 implies Conjecture 2. (c) Conjecture 2 implies Conjecture 1 (with 𝑡 = 1 − 𝑟). (d) Conjecture 1 implies the Bieberbach conjecture. (e) All weak conjectures: Conjecture 1, Conjecture 2, and Conjecture 3 are asymptotically true. For example, if 𝑓 is a function in 𝑆 with Hayman index 𝛼, and 𝑡 ∈ [0, 1] then lim. 𝑛→∞. |𝑎2𝑛 − 𝑡𝑎2𝑛−1 | 𝑛2 − 𝑡(2𝑛 − 1). = 𝛼2.. Proof. (a) This implication is trivial. (b) If Conjecture 3 is true, then for the corresponding value of 𝑟 we have |𝑎2𝑛 − 𝑤𝑎2𝑛−1 | ≤ (𝑛 − 1)2 + 𝑟(2𝑛 − 1) for all 𝑤 on the circle {𝑤 ∶ |𝑤 − 1| = 𝑟}. If 𝑎𝑛 − 𝑎2𝑛−1 ≠ 0 and 𝑎2𝑛−1 ≠ 0 we can choose a (unique) 𝑤 on this circle with ) ( arg(1 − 𝑤) = arg 𝑎2𝑛 − 𝑎2𝑛−1 − arg 𝑎2𝑛−1 so as to obtain |𝑎2𝑛 − 𝑤𝑎2𝑛−1 | = |𝑎2𝑛 − 𝑎2𝑛−1 + (1 − 𝑤)𝑎2𝑛−1 | = |𝑎2𝑛 − 𝑎2𝑛−1 | + 𝑟|𝑎2𝑛−1 |, and (𝐶𝑟 ) follows. If any of the values 𝑎2𝑛 − 𝑎2𝑛−1 , 𝑎2𝑛−1 is zero, the statement also holds trivially. (c) Assume that Conjecture 2 is true. For the corresponding 𝑟 ∈ [0, 1], consider 𝑡 = 1 − 𝑟 ∈ [0, 1]. Then by the triangle inequality |𝑎2𝑛 − 𝑡𝑎2𝑛−1 | ≤ |𝑎2𝑛 − 𝑎2𝑛−1 | + 𝑟|𝑎2𝑛−1 | ≤ (𝑛 − 1)2 + 𝑟(2𝑛 − 1) ≤ 𝑛2 − 𝑡(2𝑛 − 1), which proves that Conjecture 1 is true. (d) To show that Conjecture 1 implies the Bieberbach inequality, we follow the idea of Brown and Tsao from [2]. Begin with a weaker bound for the 𝑛-th coeﬃcient, say |𝑎𝑛 | ≤ 𝐶𝑛, for some 𝐶 > 1, and then improve on it using condition (𝐵𝑡 ). As was mentioned, we can start oﬀ from Littlewood's theorem and 𝐶 = 𝑒. Note that 𝑡≤1<. 𝑛2 , 2𝑛 − 1. for all 𝑛 ≥ 2..

(12) EFRAIMIDIS AND VUKOTI'C. 1513. Hence |𝑎𝑛 |2 ≤ |𝑎2𝑛 − 𝑡𝑎2𝑛−1 | + 𝑡|𝑎2𝑛−1 ≤ 𝑛2 − 𝑡(2𝑛 − 1) + 𝐶𝑡(2𝑛 − 1) = 𝑛2 + 𝑡(𝐶 − 1)(2𝑛 − 1) ≤ 𝐶𝑛2 . √ −𝑘 Hence, |𝑎𝑛 | ≤ 𝐶 𝑛. Iterating this procedure, we obtain |𝑎𝑛 | ≤ 𝐶 2 𝑛 for all positive integers 𝑘, which yields |𝑎𝑛 | ≤ 𝑛. (e) The proof is quite similar to that of Theorem 4.1 so we omit it.. □. ACKNOW LEDGEMENTS The authors would like to thank Professor Dmitry V. Yakubovich for some useful comments and, in particular, for suggesting that a statement like Lemma 2.1 should be true. The authors were partially supported by MTM2015-65792-P, MINECO/FEDER-EU and MTM2015-69323-REDT, MINECO, Spain. O RC ID Dragan Vukotić. http://orcid.org/0000-0002-8617-628X. REFERENCES [1] L. Brickman et al., Convex hulls and extreme points of families of starlike and convex mappings, Trans. Amer. Math. Soc. 185 (1973), 413–428. [2] J. E. Brown and A. Tsao, On the Zalcman conjecture for starlike and typically real functions, Math. Z. 191 (1986), no. 3, 467–474. [3] J. T. P. Campschroer, Inverse coeﬃcients and symmetrization of univalent functions, Doctoral Thesis, Katholieke Universiteit Nijmegen, 1984. [4] L. de Branges, A proof of the Bieberbach conjecture, Acta Math. 154 (1985), no. 1-2, 137–152. [5] P. L. Duren, Univalent functions, Springer-Verlag, New York, 1983. [6] I. Efraimidis, A generalization of Livingston's coeﬃcient inequalities for functions with positive real part, J. Math. Anal. Appl. 435 (2016), no. 1, 369–379. [7] I. Efraimidis and D. Vukotić, On the generalized Zalcman functional for some classes of univalent functions, available at http://arxiv.org/abs/ 1403.5240 (unpublished). [8] A. W. Goodman, Univalent functions and nonanalytic curves, Proc. Amer. Math. Soc. 8 (1957), 598–601. [9] I. Graham and G. Kohr, Geometric function theory in one and higher dimensions, Monographs and Textbooks in Pure and Applied Mathematics, vol. 255, Marcel Dekker, Inc., New York, 2003. [10] W. K. Hayman, Multivalent functions, second edition, Cambridge University Press, Cambridge, 1994. [11] S. Krushkal, Univalent functions and holomorphic motions, J. Anal. Math. 66 (1995), 253–275. [12] S. Krushkal, Proof of the Zalcman conjecture for initial coeﬃcients, Georgian Math. J. 17 (2010), no. 4, 663–681. (Erratum in Georgian Math. J. 19 (2012), no. 4, 777.) [13] L. Li and S. Ponnusamy, On the generalized Zalcman functional 𝜆𝑎2𝑛 − 𝑎2𝑛−1 in the close-to-convex family, Proc. Amer. Math. Soc. 145 (2017), no. 2, 833–846. [14] L. Li, S. Ponnusamy, and J. Qiao, Generalized Zalcman conjecture for convex functions of order 𝛼, Acta Math. Hungar. 150 (2016), no. 1, 234–246. [15] A. E. Livingston, The coeﬃcients of multivalent close-to-convex functions, Proc. Amer. Math. Soc. 21 (1969), 545–552. [16] W. Ma, The Zalcman conjecture for close-to-convex functions, Proc. Amer. Math. Soc. 104 (1988), no. 3, 741–744. [17] W. Ma, Generalized Zalcman conjecture for starlike and typically real functions, J. Math. Anal. Appl. 234 (1999), no. 1, 328–339. [18] T. H. MacGregor, Functions whose derivative has a positive real part, Trans. Amer. Math. Soc. 104 (1962), 532–537. [19] Ch. Pommerenke, Univalent functions, Vandenhoeck & Ruprecht, Göttingen, 1975. [20] V. Ravichandran and S. Verma, Generalized Zalcman conjecture for some classes of analytic functions, J. Math. Anal. Appl. 450 (2017), 592–605. [21] M. S. Robertson, On the theory of univalent functions, Ann. of Math. (2) 37 (1936), no. 2, 374–408. [22] T. J. Suﬀridge, Some special classes of conformal mappings, Handbook of Complex Analysis: Geometric Function Theory, Vol. 2, Elsevier, Amsterdam, 2005, pp. 309–338. [23] T. Umezawa, Analytic functions convex in one direction, J. Math. Soc. Japan 4 (1952), 194–202.. How to cite this article: Efraimidis I, Vukotić D. Applications of Livingston-type inequalities to the generalized Zalcman functional. Mathematische Nachrichten. 2018;291:1502–1513. https://doi.org/10.1002/mana.201700022.

(13)