1 Introduction and main results

Let D be a domain in \({\mathbb {C}}\) and \({\mathcal {F}}\) be a family of meromorphic functions in D. A family \({\mathcal {F}}\) is said to be normal in D, in the sense of Montel, if each sequence \(\left\{ f_n \right\} \) has a subsequence \(\left\{ f_{n_k}\right\} \) that converges spherically locally uniformly in D to a meromorphic function or to the constant \(\infty \).

The following well-known normal conjecture was proposed by Hayman in 1967.

Theorem A

[1] Let \(n \ge 5\) be a positive integer, \({\mathcal {F}}\) be a family of meromorphic functions in D and let \(a(\ne 0), b\) be two finite complex numbers. If, for any function \(f \in {\mathcal {F}}, f'(z)-af^{n}(z)\ne b\) in D, then \({\mathcal {F}}\) is normal in D.

The conjecture was proved by Li [2], Li [3], Langley [4] (for \(n \ge 5\)), Pang [5] (for \(n = 4\)), Chen and Fang [6], Bergweiler and Eremenko [7] (for \(n = 3\)).

Up to now, this result has undergone various extensions and improvements (see [8,9,10,11,12]).

The following generalization of Theorem A was proved by Yang et al. [13].

Theorem B

Let \(n \ge 4\) be a positive integer, \({\mathcal {F}}\) be a family of meromorphic functions in D and let \(a(z)(\not \equiv 0), b(z)\) be two holomorphic functions in D. If, for any function \(f \in {\mathcal {F}}\), (1)\(f(z) \ne \infty \) when \(a(z)=0\), (2)\(f'(z)-a(z)f^{n}(z)\ne b(z)\) in D, then \({\mathcal {F}}\) is normal in D.

It is natural to ask what can be said if \(f'(z)-a(z)f^{n}(z)= b(z)\) has solutions in Theorem B.

In this article, we study this problem and prove the following result.

Theorem 1.1

Let \(n \ge 4\) be a positive integer, \({\mathcal {F}}\) be a family of meromorphic functions in D and let \(a(z)(\not \equiv 0), b(z)\) be two holomorphic functions in D. If, for any function \(f \in {\mathcal {F}}\), (1)\(f(z) \ne \infty \) when \(a(z)=0\), (2) \(f'(z)-a(z)f^{n}(z)-b(z)\) has at most one zero in D, then \({\mathcal {F}}\) is normal in D.

The following examples show that both \(n \ge 4\) and the conditions (1), (2) in Theorem 1.1 are necessary and sharp.

Example 1.1

Let \({\Delta }=\{z:| z| <1\}\) and \(a(z)=-2z^{3}, b(z)=0\). Let \({\mathcal {F}}=\{f_j(z)\}\), where

$$\begin{aligned} f_j(z)=\frac{1}{z^{2}-\frac{1}{j}},z\in {\Delta },j=1,2\ldots . \end{aligned}$$

Then \(f'_{j}(z)-a(z)f^{3}_{j}(z)-b(z)= \frac{\frac{2}{j}z}{(z^{2}-\frac{1}{j})^{3}}\), which has a zero in \(\Delta \). Clearly,\(f \ne \infty \) when \(a(z)=0\), however \({\mathcal {F}}\) is not normal at 0.

Example 1.2

Let \({\Delta }=\{z:| z| <1\}, a(z)=-z, b(z)=0\). Let \({\mathcal {F}}=\{f_j(z)\}\), where

$$\begin{aligned} f_j(z)=\frac{1}{j+z},z\in \Delta ,j=1,2\ldots . \end{aligned}$$

Then \(f'_{j}(z)-a(z)f^{4}_{j}(z)-b(z)= \frac{-z^{2}+(1-2j)z-j^{2}}{(j+z)^{4}}\), which has exactly two distinct zeros in \(\Delta \). Clearly,\(f \ne \infty \) when \(a(z)=0\), however \({\mathcal {F}}\) is not normal at 0.

Example 1.3

Let \({\Delta }=\{z:| z| <1\}, a(z)=-z, b(z)=0\). Let \({\mathcal {F}}=\{f_j(z)\}\), where

$$\begin{aligned} f_j(z)=\frac{1}{jz},z\in \Delta ,j=2,3\ldots . \end{aligned}$$

Then \(f'_{j}(z)-a(z)f^{4}_{j}(z)-b(z)= \frac{1-j^{3}z}{j^{4}z^{3}}\), which has a zero in \(\Delta \). But \({\mathcal {F}}\) is not normal at 0.

The next example shows that Theorem 1.1 is not valid if a(z) is a meromorphic function in D.

Example 1.4

Let \({\Delta }=\{z:| z| <1\}, a(z)=z^{-4}, b(z)=0\). Let \({\mathcal {F}}=\{f_j(z)\}\), where

$$\begin{aligned} f_j(z)=jz,z\in \Delta ,j=2,3\ldots . \end{aligned}$$

Then \(f'_{j}(z)-a(z)f^{4}_{j}(z)-b(z)= j-j^4\), which has no zero in \(\Delta \). But \({\mathcal {F}}\) is not normal at 0.

Remark 1.1

Clearly Theorem 1.1 improves and extends Theorem B.

2 Some lemmas

Lemma 2.1

[14] Let k be a positive integer and let \({\mathcal {F}}\) be a family of functions meromorphic in the unit disc \(\Delta \), all of whose zeros have multiplicity at least k. If \({\mathcal {F}}\) is not normal in any neighbourhood of \(z_{0}\in \Delta \), then for each \(\alpha , 0\le \alpha <k\) there exist a sequence of complex numbers \(z_n, z_n\rightarrow z_0, z_0\in \Delta \), a sequence of positive numbers \(\rho _n\rightarrow 0\), and a sequence of functions \(f_{n}\in {\mathcal {F}}\) such that \( g_n (\xi ) = \rho _n^{ - \alpha } f_n (z_n + \rho _n \xi ) \rightarrow g(\xi ) \) spherically uniformly on compact subsets of \({\mathbb {C}}\), where g is a non-constant meromorphic function, all of whose zeros have multiplicity at least k. Moreover, \(g(\xi )\) has order at most 2.

We remark that one can take \(-k < \alpha \le 0 \) in the above lemma if all poles of each \(f_{n}\in {\mathcal {F}}\) have multiplicity at least k (see [15]).

Lemma 2.2

[16, Theorem 4, p. 381] Let f(z) be a transcendental meromorphic function with finite order, all of whose zeros are of multiplicity at least 2, and let \(P(z)(\not \equiv 0)\) be a polynomial, then \(f'(z)-P(z)\) has infinitely many zeros.

Lemma 2.3

[17, Lemma 6] Let k be a positive integer, f(z) be a transcendental meromorphic function with finite order, all of whose zeros have multiplicity at least \(k+1\), then \(f^{(k)}-1\) has infinitely many zeros.

Lemma 2.4

[18, Lemma 5] Let k be a positive integer, f(z) be a nonconstant rational function, all of whose zeros have multiplicity at least \(k+2\), and all of whose poles have multiplicity at least 3, then \(f^{(k)}(z)-1\) has at least two distinct zeros.

Lemma 2.5

[12, Lemma 3] Let \(n\ge 4\) be a positive integer, \(a\ne 0\) be a finite complex number and let f(z) be a nonconstant meromorphic function, then \(f'(z)-af^{n}(z)\) has at least two distinct zeros.

Lemma 2.6

[19, Lemma 7(iv), p. 261] Let kl be two positive integers, Q(z) be a rational function, all of whose zeros have multiplicity at least \(k+2\) and all of whose poles have multiplicity at least 2 with the possible exception of \(z=0\), then \(Q^{(k)}(z)= z^l\) has a solution in \({\mathbb {C}}\).

Lemma 2.7

[20, Lemma 11] Let km be two positive integers, f(z) be a rational function. If \(f(z)\ne 0\) for \(z\in {\mathbb {C}}\), and \(f^{(k)}(z)\ne z^{m}\) for \(z\ne z_{0}\), where \(z_{0}\in {\mathbb {C}}\), then f(z) is a constant.

Lemma 2.8

Let \(n\ge 4\) be a positive integer, \({\mathcal {F}}=\{f_j\}\) be a family of meromorphic functions in a domain D, and let \(a_j(z),b_j(z)\) be two sequences of analytic functions in D such that \(a_j(z)\longrightarrow a(z)\ne 0, b_j(z)\longrightarrow b(z)\). If \(f'_j(z)-a_j(z)f_{j}^{n}(z)-b_{j}(z)\) has at most one zero, then \({\mathcal {F}}\) is normal in D.

Proof

Suppose that \({\mathcal {F}}\) is not normal at \(z_{0}\in D\). By Lemma 2.1, there exists \(z_{j}\rightarrow z_{0}, \rho _{j}\rightarrow 0^{+}\), and \(f_{j}\in {\mathcal {F}}\) such that

$$\begin{aligned} g_{j}(\xi )={\rho _{j}^{\frac{1}{n-1}}}{f_{j}(z_{j}+\rho _{j}\xi )}\longrightarrow g(\xi ) \end{aligned}$$

locally uniformly on compact subsets of \({\mathbb {C}}\), where \(g(\xi )\) is a non-constant meromorphic function in \({\mathbb {C}}\).

For each \(\xi \in {\mathbb {C}}/\{g^{-1}(\infty )\}\), we have \( g_{j}'(\xi )-a_{j}(z_{j}+\rho _{j}\xi )g_{j}^{n}(\xi )-\rho _{j}^{\frac{n}{n-1}}b_{j}(z_{j}+\rho _{j}\xi ) \longrightarrow g'(\xi )-a(z_{0})g^{n}(\xi ).\)

Obviously, \(g'(\xi )-a(z_{0})g^{n}(\xi )\not \equiv 0\).

Suppose that \(g'(\xi )-a(z_{0})g^{n}(\xi )\equiv 0\), then \(g(\xi )\) must be an entire function. Hence, since \(n\ge 3\) and \(a(z_{0})\ne 0\), we have

$$\begin{aligned} nT(r,g)=T(r,g^{n})=T(r,\frac{g'}{a(z_{0})})=T(r,g')+log^{+}\frac{1}{|a(z_{0})|} \le T(r,g)+S(r,g). \end{aligned}$$

It follows that we have \(T(r,g)=S(r,g)\). But this is impossible since \(g(\xi )\) is a non-constant meromorphic function.

Claim: \(g'(\xi )-a(z_{0})g^{n}(\xi )\) has at most one zero.

Suppose this is not the case, and \(g'(\xi )-a(z_{0})g^{n}(\xi )\) has two distinct zeros \(\xi _{1}\), and \(\xi _{2}\). We choose a positive number \(\delta \) small enough such that \(D_{1}\cap D_{2}=\emptyset \) and \(g'(\xi )-a(z_{0})g^{n}(\xi )\) has no other zeros in \(D_{1}\cup D_{2}\) except for \(\xi _{1}\) and \(\xi _{2}\), where \(D_{1}=\{\xi :|\xi -\xi _{1}|<\delta \}\) and \(D_{2}=\{\xi :|\xi -\xi _{2}|<\delta \}\).

By Hurwitz’s theorem, for sufficiently large j, there exist points \(\xi _{1,j}\rightarrow \xi _{1}\) and \(\xi _{2,j}\rightarrow \xi _{2}\) such that

$$\begin{aligned} f_{j}'(\xi _{1,j})-a_{j}(z_{j}+\rho _{j}\xi _{1,j})f_{j}^{n}(z_{j}+\rho _{j}\xi _{1,j})-b_{j}(z_{j}+\rho _{j}\xi _{1,j})=0 \end{aligned}$$

and

$$\begin{aligned} f_{j}'(\xi _{2,j})-a_{j}(z_{j}+\rho _{j}\xi _{2,j})f_{j}^{n}(z_{j}+\rho _{j}\xi _{2,j})-b_{j}(z_{j}+\rho _{j}\xi _{2,j})=0 \end{aligned}$$

Since \(f'_j(z)-a_j(z)f_{j}^{n}(z)-b_{j}(z)\) has at most one zero in D, then

$$\begin{aligned} z_{j}+\rho _{j}\xi _{1,j}=z_{j}+\rho _{j}\xi _{2,j}, \end{aligned}$$

this is

$$\begin{aligned} \xi _{1,j}=\xi _{2,j}=\frac{z_{0}-z_{j}}{\rho _{j}}, \end{aligned}$$

which contradicts the fact \(D_{1}\cap D_{2}=\emptyset \). The claim is proved.

From Lemma 2.5, we have \(g'(\xi )-a(z_{0})g^{n}(\xi )\) has at least two distinct zeros, this contradicts Claim which says that \(g'(\xi )-a(z_{0})g^{n}(\xi )\) has at most one zero. Therefore \({\mathcal {F}}\) is normal in D. \(\square \)

3 Proof of theorems

Proof of Theorem 1.1

Suppose that \({\mathcal {F}}\) is not normal at \(z_{0}\). From Lemma 2.8, we obtain \(a(z_{0})= 0\). Without loss of generality, we assume that \(z_0 =0\) and \(a(z)=z^m\phi (z)\), where \(m\in {\mathbb {N}}\), and \(\phi \) is holomorphic with \(\phi (0)\ne 0\). We can say \(\phi \ne 0\) in \(\Delta (0, \delta _{0})=\{z:|z|<\delta _{0}\}\) with the normalization \(\phi (0) = 1\).

Since \(f(z)\ne \infty \) when \(a(z)=0\) and \(f'(z)-a(z)f^{n}(z)-b(z)\) has at most one zero in \(\Delta \), then \(\frac{f'(z)}{a(z)f^{n}(z)}-\frac{b(z)}{a(z)f^{n}(z)}-1\) has at most one zero in \(\Delta \). Consider the family as follows

$$\begin{aligned} {\mathcal {G}}=\{g(z)=\frac{1}{a(z)f^{n-1}(z)},f\in {\mathcal {F}}\}, \end{aligned}$$
(3.1)

where all zeros and poles of g(z) have multiplicity at least \(n-1(\ge 3)\), except possibly the pole at 0, which has order at least m. Hence \({\mathcal {G}}\) is not normal at \(z_0=0\) in \(\Delta \). Then, by Lemma 2.1, there exists \(z_{j} \longrightarrow 0,g_{j} \in {\mathcal {G}}\) and \(\rho _{j} \longrightarrow 0^{+}\) such that

$$\begin{aligned} G_{j}(\xi )=\frac{g_{j}(z_{j}+\rho _{j}\xi )}{\rho _{j}}\longrightarrow G(\xi ) \end{aligned}$$
(3.2)

locally uniformly on compact subsets of \({\mathbb {C}}\), where \(G(\xi )\) is a non-constant meromorphic functions in \({\mathbb {C}}\), and all zeros and poles of \(G(\xi )\) have multiplicity at least \(n-1(\ge 3)\), except possibly a pole, which has order at least m.

Next we consider two cases according to whether the sequence \({\frac{z_{n}}{\rho _{n}}}\) is bounded or unbounded.

Case  1. First assume that the sequence \({\frac{z_{n}}{\rho _{n}}}\) is unbounded. Then, there exists a subsequence, which we continue to call \({\frac{z_{n}}{\rho _{n}}}\), such that \({\frac{z_{n}}{\rho _{n}}}\rightarrow \infty \).

By simple calculation, from (3.1) and (3.2), we have

$$\begin{aligned} G'_{j}(\xi )= & {} \left[ \frac{g_{j}(z_{j}+\rho _{j}\xi )}{\rho _{j}}\right] '=\left[ \frac{1}{\rho _{j}a(z_{j}+\rho _{j}\xi )f^{n-1}_{j}(z_{j}+\rho _{j}\xi )}\right] '\\= & {} -\frac{a'(z_{j}+\rho _{j}\xi )}{a^{2}(z_{j}+\rho _{j}\xi )f^{n-1}_{j}(z_{j}+\rho _{j}\xi )}-\frac{(n-1)f_{j}'(z_{j}+\rho _{j}\xi )}{a(z_{j}+\rho _{j}\xi )f^{n}_{j}(z_{j}+\rho _{j}\xi )}\\= & {} -\frac{a'(z_{j}+\rho _{j}\xi )}{a^{2}(z_{j}+\rho _{j}\xi )f^{n-1}_{j}(z_{j}+\rho _{j}\xi )}-\frac{(n-1)b(z_{j}+\rho _{j}\xi )}{a(z_{j}+\rho _{j}\xi )f^{n}_{j}(z_{j}+\rho _{j}\xi )}\\&-(n-1)\left[ \frac{f_{j}'(z_{j}+\rho _{j}\xi )}{a(z_{j}+\rho _{j}\xi )f^{n}_{j}(z_{j}+\rho _{j}\xi )}-\frac{b(z_{j}+\rho _{j}\xi )}{a(z_{j}+\rho _{j}\xi )f^{n}_{j}(z_{j}+\rho _{j}\xi )}\right] . \end{aligned}$$

Noting that

$$\begin{aligned}&\frac{a'(z_{j}+\rho _{j}\xi )}{a^{2}(z_{j}+\rho _{j}\xi )f^{n-1}_{j}(z_{j}+\rho _{j}\xi )}\\&\quad =\frac{m(z_{j}+\rho _{j}\xi )^{m-1}\rho _{j}\phi (z_{j}+\rho _{j}\xi )+(z_{j}+\rho _{j}\xi )^{m}\rho _{j}\phi '(z_{j}+\rho _{j}\xi )}{(z_{j}+\rho _{j}\xi )^{2m}\phi ^{2}(z_{j}+\rho _{j}\xi )}\\&\qquad g_{j}(z_{j}+\rho _{j}\xi )(z_{j}+\rho _{j}\xi )^{m}\phi (z_{j}+\rho _{j}\xi )\\&\quad =\frac{m\rho _{j}}{z_{j}+\rho _{j}\xi }g_{j}(z_{j}+\rho _{j}\xi )+\frac{\rho _{j}\phi '(z_{j}+\rho _{j}\xi )}{\phi (z_{j}+\rho _{j}\xi )}g_{j}(z_{j}+\rho _{j}\xi )\rightarrow 0,\hbox { and }\\&\left[ \frac{b(z_{j}+\rho _{j}\xi )}{a(z_{j}+\rho _{j}\xi )f^{n}_{j}(z_{j}+\rho _{j}\xi )}\right] ^{n-1}\\&\quad =\rho ^{n}_{j}a(z_{j}+\rho _{j}\xi )b^{n-1}(z_{j}+\rho _{j}\xi )\left[ \frac{1}{\rho _{j}a(z_{j}+\rho _{j}\xi )f^{n-1}_{j}(z_{j}+\rho _{j}\xi )}\right] ^{n}\\&\quad =\rho ^{n}_{j}a(z_{j}+\rho _{j}\xi )b^{n-1}(z_{j}+\rho _{j}\xi )G^{n}_{j}(\xi )\rightarrow 0 \end{aligned}$$

uniformly on compact subsets of \({\mathbb {C}}\) disjoint from the poles of G. we deduce that \(-(n-1)\left[ \frac{f_{j}'(z_{j}+\rho _{j}\xi )}{a(z_{j}+\rho _{j}\xi )f^{n}_{j}(z_{j}+\rho _{j}\xi )}-\frac{b(z_{j}+\rho _{j}\xi )}{a(z_{j}+\rho _{j}\xi )f^{n}_{j}(z_{j}+\rho _{j}\xi )}\right] \rightarrow G'(\xi )\)

uniformly on compact subsets of \({\mathbb {C}}\) disjoint from the poles of G.

If \(G'(\xi )\equiv -(n-1)=1-n\), then \(G(\xi )=(1-n)\xi +A\), where A is a constant, which contradicts the fact that all zeros of \(G(\xi )\) have multiplicity at least \(n-1\). Hence, we have \(G'(\xi )\not \equiv -(n-1)\).

Using an argument similar to Claim in Lemma 2.8, we can obtain \(G'(\xi )+(n-1)\) has at most one zero.

On the other hand, by Lemmas 2.3 and 2.4, we have \(G'(\xi )+(n-1)\) has at least two distinct zeros. Hence \(G(\xi )\) is a constant, a contradiction.

Case 2. Now we consider the case that \({\frac{z_{n}}{\rho _{n}}}\) is bounded. Then, there is a subsequence, which we continue to call \({\frac{z_{n}}{\rho _{n}}}\), such that \({\frac{z_{n}}{\rho _{n}}}\rightarrow \alpha \in {\mathbb {C}}\).

It follows from (3.2) that

$$\begin{aligned} \frac{g_{j}(\rho _{j}\xi )}{\rho _{j}}=\frac{g_{j}(z_{j}+\rho _{j}(\xi -\frac{z_{j}}{\rho _{j}}))}{\rho _{j}}= G_{j}(\xi -\frac{z_{j}}{\rho _{j}})\rightarrow G(\xi -\alpha )={\widehat{G}}(\xi ) \end{aligned}$$
(3.3)

spherically uniformly on compact subsets of \({\mathbb {C}}\). Clearly, all zeros and all poles of \({\widehat{G}}(\xi )\) have multiplicity at least \(n-1\), and \(\xi = 0\) is a pole of \({\widehat{G}}(\xi )\) with multiplicity at least m.

Set

$$\begin{aligned} H_{j}(\xi )=\frac{1}{\rho ^{m+1}_{j}f^{n-1}_{j}(\rho _{j}\xi )}. \end{aligned}$$

It follows from (3.2) and (3.3) that

$$\begin{aligned} H_{j}(\xi )=a(\rho _{j}\xi )\frac{g_{j}(z_{j}+\rho _{j}(\xi -\frac{z_{j}}{\rho _{j}}))}{\rho _{j}}\rightarrow H(\xi )=\xi ^{m}G(\xi -\alpha )=\xi ^{m}{\widehat{G}}(\xi ) \end{aligned}$$
(3.4)

spherically locally uniformly on \({\mathbb {C}}\backslash \{0\}\), or on \({\mathbb {C}}\) if \({G}(-\alpha )\ne \infty \). Obviously, all zeros and all poles of \(H(\xi )\) have multiplicity at least \(n-1\), and \(H(0)\ne 0\) since \(f(\xi )\ne \infty \) when \(a(\xi )=0\).

Noting that

$$\begin{aligned}&H'_{j}(\xi )+(n-1)\frac{b(\rho _{j}\xi )}{\rho ^{m}_{j}f^{n}_{j}(\rho _{j}\xi )}=\left[ \frac{1}{\rho ^{m+1}_{j}f^{n-1}_{j}(\rho _{j}\xi )}\right] '+(n-1)\frac{b(\rho _{j}\xi )}{\rho ^{m}_{j}f^{n}_{j}(\rho _{j}\xi )} \\&\quad =-(n-1)\xi ^{m}\phi (\rho _{j}\xi )\left[ \frac{f'_{j}(\rho _{j}\xi )}{a(\rho _{j}\xi )f^{n}_{j}(\rho _{j}\xi )}-\frac{b(\rho _{j}\xi )}{a(\rho _{j}\xi )f^{n}_{j}(\rho _{j}\xi )}\right] . \end{aligned}$$

On the other hand,

$$\begin{aligned}&\left[ \frac{b(\rho _{j}\xi )}{\rho ^{m}_{j}f^{n}_{j}(\rho _{j}\xi )}\right] ^{n-1}=\frac{\rho ^{n}_{j}b^{n-1}(\rho _{j}\xi )a^{n}(\rho _{j}\xi )}{\rho ^{m(n-1)}_{j}}\frac{1}{\rho ^{n}_{j}a^{n}(\rho _{j}\xi )f^{n(n-1)}_{j}(\rho _{j}\xi )} \\&\quad =\frac{\rho ^{n}_{j}b^{n-1}(\rho _{j}\xi )a^{n}(\rho _{j}\xi )}{\rho ^{m(n-1)}_{j}}G^{n}_{j}(\rho _{j}\xi )=\frac{\rho ^{n}_{j}b^{n-1}(\rho _{j}\xi )a^{n}(\rho _{j}\xi )}{\rho ^{m(n-1)}_{j}}G^{n}_{j}(z_{j}+\rho _{j}(\xi -\frac{z_{j}}{\rho _{j}})) \\&\quad =\frac{\rho ^{n(m+1)}_{j}b^{n-1}(\rho _{j}\xi )\xi ^{nm}\phi ^{n}(\rho _{j}\xi )}{\rho ^{m(n-1)}_{j}}G^{n}_{j}(z_{j}+\rho _{j}(\xi -\frac{z_{j}}{\rho _{j}})) \\&\quad =\rho ^{m+n}_{j}b^{n-1}(\rho _{j}\xi )\xi ^{nm}\phi ^{n}(\rho _{j}\xi )G^{n}_{j}(z_{j}+\rho _{j}(\xi -\frac{z_{j}}{\rho _{j}}))\rightarrow 0 \end{aligned}$$

uniformly on compact subsets of \({\mathbb {C}}\) disjoint from the poles of G. Hence

$$\begin{aligned} -(n-1)\xi ^{m}\phi (\rho _{j}\xi )\left[ \frac{f'_{j}(\rho _{j}\xi )}{a(\rho _{j}\xi )f^{n}_{j}(\rho _{j}\xi )}-\frac{b(\rho _{j}\xi )}{a(\rho _{j}\xi )f^{n}_{j}(\rho _{j}\xi )}\right] \rightarrow H'(\xi ) \end{aligned}$$

uniformly on compact subsets of \({\mathbb {C}}\) disjoint from the poles of G.

If \(H'(\xi )\equiv -(n-1)\xi ^{m}\), then \(H(\xi )=\frac{(1-n)\xi ^{m+1}}{m+1}+B\), where B is a constant. Thus, \(H(\xi )\) has at least one zero in \({\mathbb {C}}\). Let \(\xi _{0}\) be a zero of \(H(\xi )\), then by the fact that \(\xi _{0}\) has multiplicity at least \(n-1(\ge 3)\), we get \(H'(\xi _{0})=0\) and hence \(\xi _{0} = 0\) by \(H'(\xi ) \equiv -(n-1)\xi ^{m}\). It follows that \(H(\xi )=c~\xi ^{l}, c\in {\mathbb {C}}\) and \(l\in {\mathbb {N}}\). Thus, \(\frac{(1-n)\xi ^{m+1}}{m+1}+B=c~\xi ^{l}\). By comparing the degrees and coefficients, we see that \(B=0\). Thus, \(H(\xi )=\frac{(1-n)\xi ^{m+1}}{m+1}\). It follows that \(H(0)=0\), which contradicts that \(H(0)\ne 0\).

Using an argument similar to Claim in Lemma 2.8, we can obtain \(H'(\xi )+(n-1)\xi ^{m}\) has at most one zero.

Next we consider two subcases according to whether \(H'(\xi )+(n-1)\xi ^{m}\) has zero or not.

Case 2.1 \(H'(\xi )\ne (n-1)\xi ^{m}\). It now follows from Lemmas 2.2 and 2.6 that Case 2.1 cannot occur.

Case 2.2 \(H'(\xi )= (n-1)\xi ^{m}\). Then \(H'(\xi )\ne (n-1)\xi ^{m}\) for \(\xi \ne \xi _{0}\), where \(\xi _{0}\in {\mathbb {C}}\).

Case 2.2.1 \(\xi _{0}=0\). Thus \(H'(0)=0\). It follows that \(H(0)=0\), which contradicts that \(H(0)\ne 0\).

Case 2.2.2 \(\xi _{0}\ne 0\).

If \(H(\xi _{0})=0\), then \(H'(\xi _{0})=0\) since all zeros of \(H(\xi )\) have multiplicity at least \(n-1\). Thus \(\xi _{0}=0\), which contradicts that \(\xi _{0}\ne 0\).

If \(H(\xi )\ne 0, \xi \in {\mathbb {C}}\). Combined with the fact that \(H'(\xi )\ne (n-1)\xi ^{m}\) for \(\xi \ne \xi _{0}\), where \(\xi _{0}\in {\mathbb {C}}\). Hence by Lemma 2.7\(H(\xi )\) is a constant, a contradiction.

If \(H(\xi )=0\), and \(H(\xi _{0})\ne 0, \xi \in {\mathbb {C}}\backslash \{\xi _{0}\}\). Combined with the fact that \(H'(\xi )\ne (n-1)\xi ^{m}\) for \(\xi \in {\mathbb {C}}\backslash \{\xi _{0}\}\). It now follows from Lemmas 2.2 and 2.6 that this case cannot occur. Hence we show that \({\mathcal {G}}\) is normal at \( z_{0}=0\).

Next, we show that \({\mathcal {F}}\) is normal at \(z_0=0\). Since \({\mathcal {G}}\) is normal at \(z_0=0\), let \(g_j\longrightarrow g\) in a neighborhood of 0, then there exist \({\Delta _\delta } = \left\{ {z:\left| z \right| < \delta } \right\} \) and a subsequence of \(\{g_j\}\) such that \(\{g_j\}\) converges uniformly to a meromorphic function or \(\infty \). Noting \(g(0)=\infty \), we can find a \(\varepsilon \) with \(0<\varepsilon < \delta \) and \(M>0\) such that \(|g(z)|>M, z\in \Delta _\varepsilon \). So, for sufficiently large j, we get \(|g_j(z)|\ge \frac{M}{2}\), hence \(|a(z)f^{n-1}_{j}(z)|\le \frac{2}{M}\). Therefore \(f_j(z)\ne \infty \) for sufficiently large j and \(z\in \Delta _\varepsilon \).

Hence \({f_j(z)}\) is analytic in \(\Delta _\varepsilon \). Choosing \(\varepsilon \) small enough that \(|a(z)|\ge \frac{|z|^{m}}{M}\), it follows that, for sufficiently large j, we have

$$\begin{aligned} |f_j(z)|=\left| {\frac{1}{a(z)g_{j}(z)}}\right| ^{\frac{1}{n-1}} \le {\left( {\frac{2^{m+1}}{\varepsilon ^{m} }} \right) ^\frac{1}{n-1}},\quad \left| z \right| = \frac{\varepsilon }{2}. \end{aligned}$$

By the Maximum Principle and Montel’s theorem, \({\mathcal {F}}\) is normal at \(z_0=0\). The complete proof of Theorem 1.1 is given. \(\square \)