betaFeedback

WorksMathematics

On the periodicity of meromorphic functions of the form ff(k)ff^{(k)}

Abstract

For a non-entire meromorphic function ff and an integer kk, we will show that whenever ff(k)ff^{(k)} is periodic so is ff. This result gives an affirmative answer to a significant class of the so called generalized Yang's conjecture.

Introduction and main results

We recall Yang’s conjecture, whose precise statement is as follows: given a holomorphic function ff, if ff(k)f f^{(k)} (k∈N+k \in\mathbb{N}^{+}) is periodic, is ff necessarily periodic? Recently, this question has attracted extensive attention from researchers. Wang and Hu [8] proved this conjecture is true when k=1k=1. However, the case k≥2k \ge2 remains open as of now. Later, several scholars have started to study its differential or difference variants, as well as the case where ff is meromorphic, see [1, 2, 3, 4, 5, 6, 7]. These research lead to the so called generalized Yang’s conjecture, which asks whether ff is periodic whenever ff is meromorphic and fnf(k)f^n f^{(k)} (n,k∈N+n,k \in\mathbb{N}^{+}) is periodic. Some investigations were mentioned in [7]. And main results from [7] studied this question for fnf(k)f^n f^{(k)} when ff is non-holomorphic meromorphic. Specifically,

Theorem A ([7]). Let nn be a positive integer and kk be an even integer. Assume ff is transcendental meromorphic with at least one pole. If fnf(k)f^n f^{(k)} is TT-periodic, then ff is a periodic function with period pTpT, where p∈{1,2,⋯ ,n+1}p \in\{1,2,\cdots,n+1\}.

Theorem B ([7]). Let m,n≥2m,n \ge2 be integers and kk be an odd integer with k>m(n−1)k > m(n-1). Assume ff is transcendental meromorphic such that the minimum multiplicity of all its poles is mm. If fnf(k)f^n f^{(k)} is TT-periodic, then ff is a periodic function with period pTpT, where p∈{1,2,⋯ ,n+1}p \in\{1,2,\cdots,n+1\}.

Through observation, they provided an affirmative answer to fnf(k)f^n f^{(k)} for even kk and for meromorphic ff having at least one pole. However, in the odd case, they imposed restrictions on nn and the multiplicities of the poles. In the following, we consider n=1n=1 and allow poles of any positive multiplicity. For k=1k=1, the conclusion is already known for transcendental meromorphic functions, including holomorphic functions, see [6]. The main result of the present paper treats arbitrary positive odd kk. And we do not impose conditions on the growth of ff.

Throughout this paper, a meromorphic function means a function meromorphic on C\mathbb{C}, unless some other domain is specified. A cc-period of ff is a number c≠0c \ne0 satisfying

f(z+c)=f(z).f(z+c)=f(z).

which need not be a smallest period.

Theorem 1.1. Let kk be an odd integer. Assume that ff is meromorphic with at least one pole. If ff(k)ff^{(k)} is cc-periodic, then ff is a periodic function with period cc or 2c2c.

Together with Theorem A, we immediately deduce the following conclusion.

Theorem 1.2. Assume that ff is meromorphic with at least one pole. If ff(k)ff^{(k)} is cc-periodic, then ff is a periodic function with period cc or 2c2c.

Two cases in Theorem 1.2 do occur. To illustrate, when f(z)=1/sin⁡(πz/c)f(z)=1/\sin(\pi z/c), ff(k)ff^{(k)} has period cc, whereas ff has period 2c2c rather than cc; when f(z)=cot⁡(πz/c)f(z)=\cot(\pi z/c), both ff(k)ff^{(k)} and ff have period cc.

Preliminaries

This section presents several lemmas that will be needed. For convenience, we define some notations as follows. For z0∈Cz_0 \in\mathbb{C} and r>0r>0, write

D(a,r)={z∈C:∣z−a∣<r},D∗(a,r)={z∈C:0<∣z−a∣<r}.D(a,r)=\{z\in\mathbb{C}:|z-a|<r\},\qquad D^*(a,r)=\{z\in\mathbb{C}:0<|z-a|<r\}.

For a meromorphic function uu, define

Qk[u]=12∑l=0k−1(−1)lu(l)u(k−1−l).Q_k[u]=\frac{1}{2}\sum_{l=0}^{k-1}(-1)^l u^{(l)}u^{(k-1-l)}.

It is not hard to check that

(Qk[u])′=uu(k),Qk[−u]=Qk[u].(Q_k[u])'=uu^{(k)},\qquad Q_k[-u]=Q_k[u].

A family of holomorphic functions on a domain is called normal if every sequence has a subsequence that converges uniformly on compact subsets to a holomorphic function, or tends uniformly to infinity on compact subsets. And a family is locally bounded if it has a common bound on each compact subset of its domain. We refer to [9] for more details.

Lemma 2.1 ([9]). Let F\mathcal{F} be a family of holomorphic functions on a domain D⊂CD\subset\mathbb{C}. If F\mathcal{F} is not normal on DD, then there exist a sequence {fn}⊂F\{f_n\}\subset\mathcal{F}, a sequence {zn}\{z_n\} with lim⁡n→∞zn=z0∈D\lim_{n\to\infty}z_n=z_0\in D, and a sequence of positive number {εn}\{\varepsilon_n\} with lim⁡n→∞εn=0\lim_{n\to\infty}\varepsilon_n=0 such that

gn(ξ)=fn(zn+εnξ)g_n(\xi)=f_n(z_n+\varepsilon_n\xi)

converges uniformly to a nonconstant holomorphic function g(ξ)g(\xi) on any compact subset of C\mathbb{C}.

Lemma 2.2. Let D⊂CD \subset\mathbb{C} be a domain and F\mathcal{F} be a family of zero-free holomorphic functions on DD. If

{ff(k):f∈F}\{f f^{(k)}: f \in\mathcal{F}\}

is locally bounded on DD, then F\mathcal{F} is normal on DD.

Proof. On the contrary, suppose that F\mathcal{F} is not normal. Applying Lemma 2.1, we can get a sequence {fn}⊂F\{f_n\} \subset\mathcal{F} satisfying

gn(ξ)=fn(zn+εnξ)⟹g(ξ)g_n(\xi)=f_n(z_n+\varepsilon_n\xi)\Longrightarrow g(\xi)

on compact subsets of C\mathbb{C}, where lim⁡n→∞zn=z0∈D\lim_{n\to\infty}z_n=z_0\in D, lim⁡n→∞εn=0\lim_{n\to\infty}\varepsilon_n=0 and gg is a nonconstant holomorphic function.

We claim that gg has no zeros. Suppose that ξ0\xi_0 is a zero of gg. Choose r>0r>0 such that gg has no zeros on ∣ξ−ξ0∣=r|\xi-\xi_0|=r. Then, for nn sufficiently large, we have

max⁡∣ξ−ξ0∣=r∣gn(ξ)−g(ξ)∣<min⁡∣ξ−ξ0∣=r∣g(ξ)∣.\max_{|\xi-\xi_0|=r}|g_n(\xi)-g(\xi)|<\min_{|\xi-\xi_0|=r}|g(\xi)|.

Rouché’s theorem implies that gng_n and gg have the same number of zeros inside the circle ∣ξ−ξ0∣=r|\xi-\xi_0|=r counting multiplicities. This contradicts that every gng_n has no zeros. Hence gg is zero-free.

Cauchy’s integral formula shows that all derivatives of each order converge uniformly on compact subsets, i.e., gn(k)(ξ)⟹g(k)(ξ)g_n^{(k)}(\xi)\Longrightarrow g^{(k)}(\xi). Take a compact set K⊂DK\subset D. Since {ff(k)}\{f f^{(k)}\} is locally bounded, a direct computation gives

gn(ξ)gn(k)(ξ)=εnk(fnfn(k))(zn+εnξ)⟹0g_n(\xi)g_n^{(k)}(\xi)=\varepsilon_n^k\left(f_nf_n^{(k)}\right)(z_n+\varepsilon_n\xi)\Longrightarrow0

on KK. Furthermore, it yields

gg(k)≡0gg^{(k)}\equiv0

on C\mathbb{C}. Since gg has no zeros, g(k)≡0g^{(k)}\equiv0, which means gg is a polynomial of degree at most k−1k-1. However, a nonconstant polynomial must have zeros. Therefore gg is constant, a contradiction. The assumption is not valid, i.e., F\mathcal{F} is normal on DD. □

Lemma 2.3. Assume that ff is a meromorphic function with a pole at z0z_0 and ff(k)f f^{(k)} is cc-periodic. Denote

fj(z)=f(z+jc),j∈N.f_j(z)=f(z+jc),\qquad j\in\mathbb{N}.

Then there exists r>0r>0 such that the family {fj}\{f_j\} is holomorphic, zero-free and locally bounded on D∗(z0,r)D^*(z_0,r). Moreover, for any ζ∈D∗(z0,r)\zeta\in D^*(z_0,r) and any l∈Nl\in\mathbb{N}, we have

sup⁡j∈N∣fj(l)(ζ)∣<∞.\sup_{j\in\mathbb{N}}|f_j^{(l)}(\zeta)|<\infty.

Proof. Set F=ff(k)F=f f^{(k)}. Suppose that the pole z0z_0 of ff has order mm. Then, in a neighborhood of z0z_0, we have

f(z)=Aτ−m+O(τ−m+1),f(z)=A\tau^{-m}+O(\tau^{-m+1}),

where A≠0A\ne0, m∈Nm\in\mathbb{N}, and τ=z−z0\tau=z-z_0. Differentiation gives

f(k)(z)=(−1)k(m+k−1)!(m−1)!Aτ−m−k+O(τ−m−k+1).f^{(k)}(z)=(-1)^k\frac{(m+k-1)!}{(m-1)!}A\tau^{-m-k}+O(\tau^{-m-k+1}).

Therefore,

F(z)=(−1)k(m+k−1)!(m−1)!A2τ−2m+k+O(τ−2m−k+1).(1)F(z)=(-1)^k\frac{(m+k-1)!}{(m-1)!}A^2\tau^{-2m+k}+O(\tau^{-2m-k+1}). \tag*{(1)}

It follows that z0z_0 is also a pole of FF, of order 2m+k2m+k. Then, for any j∈Nj \in\mathbb{N}, a simple calculation yields

fjfj(k)(z)=F(z+jc)=F(z),(2)f_j f_j^{(k)}(z)=F(z+jc)=F(z), \tag*{(2)}

which indicates z0z_0 is also a pole of fjf_j. Let its order be mjm_j and its leading coefficient of Laurent expansion of fjf_j be AjA_j. Differentiating the Laurent expansion of fjf_j kk times, substituting the expansions fjf_j and fj(k)f_j^{(k)} into (2), and then comparing the order and the leading coefficient with (1) in turn, we obtain

mj=m,Aj=±A.m_j=m,\qquad A_j=\pm A.

Take rr so that FF is holomorphic and nonzero on D∗(z0,r)D^*(z_0,r). It yields that fjf_j is holomorphic and zero-free on D∗(z0,r)D^*(z_0,r) as well. In addition, all fjfj(k)f_jf_j^{(k)} are locally bounded on D∗(z0,r)D^*(z_0,r) due to (2). Employing Lemma 2.2, we can see {fj}\{f_j\} is normal on D∗(z0,r)D^*(z_0,r). Furthermore, we claim {fj}\{f_j\} is local boundedness on D∗(z0,r)D^*(z_0,r). If not, then there exists a subsequence {fjn}\{f_{j_n}\} tending uniformly to infinity on a compact subset KK of D∗(z0,r)D^*(z_0,r). Denote

Vj(z)=(z−z0)mfj(z),V_j(z)=(z-z_0)^m f_j(z),

which is a holomorphic and zero-free function on D(z0,r)D(z_0,r), and

Vj(z0)=Aj,∣Vj(z0)∣=∣A∣.V_j(z_0)=A_j,\qquad|V_j(z_0)|=|A|.

In particular, 1/Vj1/V_j is also holomorphic on D(z0,r)D(z_0,r). Fix r~\widetilde r with {z:∣z−z0∣=r~}⊂K\{z:|z-z_0|=\widetilde r\}\subset K. Then

max⁡∣z−z0∣=r~∣1fjn(z)∣⟶0.\max_{|z-z_0|=\widetilde r}\left|\frac{1}{f_{j_n}(z)}\right|\longrightarrow0.

Applying maximum modulus principle, we deduce that

1∣A∣=∣1Vjn(z0)∣≤max⁡∣z−z0∣=r~∣1Vjn(z)∣=r−mmax⁡∣z−z0∣=r~∣1fjn(z)∣⟶0,n⟶∞,\frac{1}{|A|}=\left|\frac{1}{V_{j_n}(z_0)}\right|\leq\max_{|z-z_0|=\widetilde r}\left|\frac{1}{V_{j_n}(z)}\right|=r^{-m}\max_{|z-z_0|=\widetilde r}\left|\frac{1}{f_{j_n}(z)}\right|\longrightarrow0,\qquad n\longrightarrow\infty,

which is impossible. Thus, normality implies local boundedness.

For any ζ∈D∗(z0,r)\zeta\in D^*(z_0,r), and choose ρ>0\rho>0 such that D(ζ,ρ)‾⊂D∗(z0,r)\overline{D(\zeta,\rho)}\subset D^*(z_0,r). According to the local boundedness, there exists a constant M>0M>0 for which

∣fj(z)∣≤M,z∈D(ζ,ρ)‾.|f_j(z)|\leq M,\qquad z\in\overline{D(\zeta,\rho)}.

It follows Cauchy’s integral formula that

∣fj(l)(ζ)∣≤l!Mρl.(3)\left|f_j^{(l)}(\zeta)\right|\leq\frac{l!M}{\rho^l}. \tag*{(3)}

Thus, for any l∈Nl\in\mathbb{N}, we have

sup⁡j∈N∣fj(l)(ζ)∣<∞.\sup_{j\in\mathbb{N}}\left|f_j^{(l)}(\zeta)\right|<\infty.

□\square

Lemma 2.4. Let kk be odd. Assume that ff and gg are meromorphic functions with the same order m≥1m\geq1 and the same leading Laurent coefficient α≠0\alpha\neq0 at pole z0z_0. If

ff(k)=gg(k),Qk[f]=Qk[g],(4)f f^{(k)}=g g^{(k)},\qquad Q_k[f]=Q_k[g], \tag*{(4)}

then f=gf=g in a neighborhood of z0z_0.

Proof. Suppose, on the contrary, that h=g−fh=g-f is not identically zero in a neighborhood of z0z_0. Since the leading terms of Laurent expansions ff and gg agree, there are an integer n>−mn>-m and a constant b≠0b\ne0 such that

f(z)=aτ−m+O(τ−m+1),f(z)=a\tau^{-m}+O(\tau^{-m+1}),
h(z)=bτn+O(τn+1).h(z)=b\tau^n+O(\tau^{n+1}).

Combining the first identity in (4) with g=f+hg=f+h, we have

0=fh(k)+hf(k)+hh(k).(5)0=fh^{(k)}+hf^{(k)}+hh^{(k)}. \tag*{(5)}

A straightforward computation yields

fh(k)=−(m+k−1)!(m−1)!abτn−m−k+O(τn−m−k+1),fh^{(k)}=-\frac{(m+k-1)!}{(m-1)!}ab\tau^{n-m-k}+O(\tau^{n-m-k+1}),
hf(k)=n(n−1)⋯(n−k+1)abτn−m−k+O(τn−m−k+1),hf^{(k)}=n(n-1)\cdots(n-k+1)ab\tau^{n-m-k}+O(\tau^{n-m-k+1}),
hh(k)=O(τ2n−k).hh^{(k)}=O(\tau^{2n-k}).

Comparing the lowest-order terms in (5) yield n=m+k−1≥kn=m+k-1\ge k. Further calculation leads to

Qk[g]−Qk[f]=Qk[f+h]−Qk[f]=12∑l=0k−1(−1)l(f(l)h(k−1−l)+h(l)f(k−1−l))+Qk[h]=∑l=0k−1(−1)lf(l)h(k−1−l)+Qk[h]=∑l=0k−1(−1)l(−1)l(m+l−1)!(m−1)!n!(n−k+1+l)!ab+O(τ)=abn!(m−1)!∑l=0k−11m+l+O(τ).\begin{aligned} Q_k[g]-Q_k[f]&=Q_k[f+h]-Q_k[f]\\ &=\frac{1}{2}\sum_{l=0}^{k-1}(-1)^l\left(f^{(l)}h^{(k-1-l)}+h^{(l)}f^{(k-1-l)}\right)+Q_k[h]\\ &=\sum_{l=0}^{k-1}(-1)^l f^{(l)}h^{(k-1-l)}+Q_k[h]\\ &=\sum_{l=0}^{k-1}(-1)^l(-1)^l\frac{(m+l-1)!}{(m-1)!}\frac{n!}{(n-k+1+l)!}ab+O(\tau)\\ &=ab\frac{n!}{(m-1)!}\sum_{l=0}^{k-1}\frac{1}{m+l}+O(\tau). \end{aligned}

In terms of the second identity in (4), we conclude that

0=abn!(m−1)!∑l=0k−11m+l,0=ab\frac{n!}{(m-1)!}\sum_{l=0}^{k-1}\frac{1}{m+l},

which is impossible. Hence hh must vanish identically near z0z_0. This proves f=gf=g.

Proof of Theorem 1.1

Proof. Assume that ff has a pole z0z_0 of order mm. In a neighborhood of z0z_0, we have

f(z)=aτ−m+O(τ−m+1),f(z)=a\tau^{-m}+O(\tau^{-m+1}),

where a≠0a\ne0 and τ=z−z0\tau=z-z_0. Lemma 2.2 shows that f(z+c)f(z+c) has the same order at z0z_0 and its leading coefficient is a1=±aa_1=\pm a. Set

κ=a1a∈{1,−1},g(z)=κf(z+c).\kappa=\frac{a_1}{a}\in\{1,-1\},\qquad g(z)=\kappa f(z+c).

Then gg and ff have the same leading coefficient aa at z0z_{0} and

gg(k)=ff(k).gg^{(k)} = ff^{(k)}.

The periodicity of F=ff(k)F = ff^{(k)} leads to

ddz(Qk[f](z+c)−Qk[f](z))=F(z+c)−F(z)=0.\frac{d}{dz}\left(Q_{k}[f](z+c)-Q_{k}[f](z)\right)=F(z+c)-F(z)=0.

Hence, we obtain

Qk[f](z+c)−Qk[f](z)=C0,C0∈C.(6)Q_{k}[f](z+c)-Q_{k}[f](z)=C_{0},\qquad C_{0}\in\mathbb{C}. \tag*{(6)}

Set

fj(z)=f(z+jc),j∈N.f_{j}(z)=f(z+jc),\qquad j\in\mathbb{N}.

Employing Lemma 2.3, there exists r>0r>0 such that for ζ∈D∗(z0,r)\zeta\in D^{*}(z_{0},r),

∣Qk[fj](ζ)∣≤12∑l=0k−1∣fj(l)(ζ)∣∣fj(k−1−l)(ζ)∣≤M22ρk−1∑l=0k−1l!(k−1−l)!.\left|Q_{k}[f_{j}](\zeta)\right|\leq\frac{1}{2}\sum_{l=0}^{k-1}\left|f_{j}^{(l)}(\zeta)\right|\left|f_{j}^{(k-1-l)}(\zeta)\right|\leq\frac{M^{2}}{2\rho^{k-1}}\sum_{l=0}^{k-1}l!(k-1-l)!.

where M,ρM,\rho satisfies

∣fj(z)∣≤M,z∈D(ζ,ρ)⊂D∗(z0,r).\left|f_{j}(z)\right|\leq M,\qquad z\in D(\zeta,\rho)\subset D^{*}(z_{0},r).

On the other hand, (6) gives

Qk[fj](ζ)=Qk[f](ζ+jc)=Qk[f](ζ)+jC0.Q_{k}[f_{j}](\zeta)=Q_{k}[f](\zeta+jc)=Q_{k}[f](\zeta)+jC_{0}.

If C0≠0C_{0}\ne0, then

∣Qk[f](ζ)+jC0∣≥j∣C0∣−∣Qk[f](ζ)∣⟶∞,j⟶∞,\left|Q_{k}[f](\zeta)+jC_{0}\right|\geq j\left|C_{0}\right|-\left|Q_{k}[f](\zeta)\right|\longrightarrow\infty,\qquad j\longrightarrow\infty,

a contradiction. Therefore C0=0C_{0}=0, which means

Qk[f]=Qk[f](z+c)=Qk[g](z).Q_{k}[f]=Q_{k}[f](z+c)=Q_{k}[g](z).

Lemma 2.4 shows that g=fg=f near z0z_{0}, i.e. κf(z+c)=f(z)\kappa f(z+c)=f(z). If κ=1\kappa=1, then cc is a period of ff. If κ=−1\kappa=-1, then 2c2c is a period of ff. This completes the proof. □

School of Mathematical Sciences, Suzhou University of Science and Technology, Suzhou, P. R. China

Email address: [email protected]

School of Mathematics, Nanjing University, Nanjing, P. R. China

Email address: [email protected]

School of Mathematics, Nanjing University, Nanjing, P. R. China

Email address: [email protected]

School of Mathematics, Nanjing University, Nanjing, P. R. China

Email address: [email protected]

References

  1. [1]Z. Latreuch and A. Zemirni. On a theorem of A. and C. rényi and a conjecture of C. C. Yang concerning periodicity of entire functions. Analysis Mathematica, 48:111–125, 2022.DOI
  2. [2]P. Li, W.R. Lv, and C.C. Yang. Entire solutions of certain types of nonlinear differential equations. Houston Journal of Mathematics, 45:431–437, 2019.
  3. [3]K. Liu, Y.M. Wei, and P.Y. Yu. Generalized Yang’s conjecture on the periodicity of entire functions. Bulletin of the Korean Mathematical Society, 57(5):1259–1267, 2020.DOI
  4. [4]K. Liu and P.Y. Yu. A note on the periodicity of entire function. Bulletin of the Australian Mathematical Society, 100(2):290–296, 2019.DOI
  5. [5]X.L. Liu and R. Korhonen. On the periodicity of transcendental entire functions. Bulletin of the Australian Mathematical Society, 101(3):453–465, 2020.DOI
  6. [6]X.L. Liu, R. Korhonen, and K. Liu. Variations on a conjecture of C. C. Yang concerning periodicity. Computational Methods and Function Theory, 22:27–33, 2022.DOI
  7. [7]S.X. Mei, W.Q. Shen, J. Wang, and X. Yao. Finiteness property and the periodicity of meromorphic functions. Analysis Mathematica, 51(1):269–277, 2025.
  8. [8]Q. Wang and P.C. Hu. On zeros and periodic entire functions. Acta Mathematica Scientia, 38A(2):209–214, 2018.
  9. [9]L. Zalcmann. Normal families: new perspectives. Bulletin of the American Mathematical Society, 35(3):215–230, 1998.

Paper details

vlnc:W_cezmbvnx
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