Meromorphic solutions of higher order Briot-Bouquet differential equations
A. Eremenko, L. W. Liao, T. W. Ng

TL;DR
This paper proves that meromorphic solutions with poles of higher order Briot-Bouquet differential equations are elliptic functions, extending the understanding of their solution structure.
Contribution
It establishes that all meromorphic solutions with poles for these equations are elliptic functions, including degenerate cases, providing a comprehensive classification.
Findings
Meromorphic solutions with poles are elliptic functions.
Degenerate elliptic functions are also solutions.
The result generalizes previous classifications of such differential equations.
Abstract
For differential equations where is a polynomial, we prove that all meromorphic solutions having at least one pole are elliptic functions, possibly degenerate.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
\checkfont
msam10
\newromanexpr\HessHess
Meromorphic solutions of higher order Briot–Bouquet
differential equations
ALEXANDRE E EREMENKO
Department of Mathematics Supported by NSF grants DMS-0555279 and DMS-0244547.
Purdue University
West Lafayette IN
47907 USA
e-mail: [email protected]
LIANGWEN LIAO
Department of Mathematics Partially supported by the grant of the education department of Jiangsu Province, China 07KJB110069.
Purdue University
Nanjing University
Nanjing
210093
China
e-mail: [email protected]
TUEN WAI NG
Department of Mathematics Partially supported by RGC grant HKU 7020/03P, and NSF grant DMS-0244547.
The University of Hong Kong
Pokfulam
Hong Kong
e-mail: [email protected]
(2009)
Abstract
For differential equations where is a polynomial, we prove that all meromorphic solutions having at least one pole are elliptic functions, possibly degenerate.
††volume: 121
1 Introduction
According to a theorem of Weierstrass, meromorphic functions in the complex plane that satisfy an algebraic addition theorem
[TABLE]
are elliptic functions, possibly degenerate [17, 1].
More precisely, let us denote by the class of meromorphic functions in that consists of doubly periodic functions, rational functions and functions of the form where is rational and . Then each function satisfies an identity of the form (1), and conversely, every meromorphic function111A “meromorphic function” in this paper means a function meromorphic in the complex plane, unless some other domain is specified. See [17, 20] for discussion of the equation (1) in more general classes of functions. that satisfies such an identity belongs to .
One way to prove this result is to differentiate (1) with respect to and then set . Then we obtain a Briot–Bouquet differential equation
[TABLE]
The fact that every meromorphic solution of such an equation belongs to was known to Abel and Liouville, but probably it was stated for the first time in the work of Briot and Bouquet [5, 6].
Here we consider meromorphic solutions of higher order Briot–Bouquet equations
[TABLE]
Picard [18] proved that for , all meromorphic solutions belong to the class . This work was one of the first applications of the famous Picard’s theorems on omitted values.
In the end of 1970-s Hille [12, 13, 14, 15] considered meromorphic solutions of (2) for arbitrary . The result of Picard was already forgotten, and Hille stated it as a conjecture. Then Bank and Kaufman [4] gave another proof of Picard’s theorem.
These investigations were continued in [8]. To state the main results from [8] we assume without loss of generality that the polynomial in (2) is irreducible. Let denote the compact Riemann surface defined by the equation
[TABLE]
Then every meromorphic solution of (2) defines a holomorphic map . According to another theorem of Picard, a Riemann surface which admits a non-constant holomorphic map from has to be of genus [math] or , ([19], see also [2]). The following theorems were proved in [8]:
Theorem A. If is of genus , then every meromorphic solution of is an elliptic function.
Theorem B. If is odd, then every meromorphic solution of having at least one pole, belongs to the class .
The main result of the present paper is the extension of Theorem B to the case of even .
Theorem 1. If is a meromorphic solution of an equation and has at least one pole, then .
This can be restated in the following way. Let be a meromorphic function in the plane which is not entire and does not belong to . Then and are algebraically independent.
It is easy to see that for every function of class and every natural integer there exists an equation of the form (2) which satisfies.
It is not true that all meromorphic solutions of higher order Briot–Bouquet equations belong to , a simple counterexample is . We don’t know whether non-linear irreducible counterexamples exist.
In the process of proving of Theorem 1 we will establish an estimate of the degrees of possible meromorphic solutions in terms of the polynomials . Here by degree of a function of class we mean the degree of a rational function , or the degree of in , or the number of poles in the fundamental parallelogram of an elliptic function . Thus our result permits in principle the determination of all meromorphic solutions having at least one pole of a given equation (2).
Our method of proof is based on the so-called “finiteness property” of certain autonomous differential equations: there are only finitely many formal Laurent series with a pole at zero that satisfy these equations. The idea seems to occur for the first time in [12, p. 274] but the argument given there contains a mistake. This mistake was corrected in [8]. Later the same method was applied in [7] and [10] to study meromorphic solutions of other differential equations.
2 Preliminaries
We will use the following refined version of Wiman–Valiron theory which is due to Bergweiler, Rippon and Stallard.
Let be a meromorphic function and a component of the set which contains no poles (so is unbounded). Set
[TABLE]
and
[TABLE]
This derivative exists for all except possibly a discrete set. According to a theorem of Fuchs [11],
[TABLE]
unless the singularity of at is a pole. For every we choose a point with the properties .
Theorem C. For every , there exists a set of finite logarithmic measure, such that for , the disk
[TABLE]
is contained in and we have
[TABLE]
When is entire, this is a classical theorem of Wiman. Wiman’s proof used power series, so it cannot be extended to the situation when is not entire. A more flexible proof, not using power series is due to Macintyre [16]; it applies, for example to functions analytic and unbounded in . The final result stated above was recently established in [3].
3 Proof of Theorem 1
In what follows, we always assume that the polynomial in (2) is irreducible.
To state a result of [8] which we will need, we introduce the following notation. Let be the field of meromorphic functions on . The elements of can be represented as rational functions whose denominators are co-prime with . In particular, and in (3) are elements of . For and a point , we denote by the order of at the point . Thus if then is the multiplicity of the zero of , if then is the multiplicity of the pole, and at all other points .
Let be the set of poles of . For we set
Theorem D. *Suppose that an irreducible equation has a transcendental meromorphic solution . Let be the holomorphic map defined by . Then:
a) Every pole of belongs to .
b) For every , the number is either or , where is a positive integer.
c) If for some , then the equation has infinitely many solutions, and all these solutions are poles of order of .
d) If for some , then the equation has no solutions.*
Picard’s theorem on omitted values implies that can happen for at most two points . For the convenience of the reader we include a proof of Theorem D in the Appendix.
The numbers can be easily determined from the Newton polygon of . Thus Theorem D gives several effective necessary conditions for the equation (2) to have meromorphic or entire solutions.
Remark. The proof of Theorem D in [8] uses Theorem C which was stated in [8] but not proved. One can also give an alternative proof of Theorem D, using Nevanlinna theory instead of Theorem C, by the arguments similar to those in [9].
Lemma 1. Suppose that is a meromorphic solution of . If for some then has order one, normal type.
Proof. In view of Theorem A and Theorem D, d), we conclude that the genus of is zero. Therefore, we can find in which has a single simple pole at . Then is an entire function by Theorem D, d). As has a simple pole at , the element is a local parameter at , and in a neighborhood of we have
[TABLE]
where as , and the dots stand for the terms of degree smaller than . Substituting and and differentiating the first equation times we obtain for a differential equation of the form
[TABLE]
where the dots stand for the terms of degree smaller than . Now we use a standard argument of Wiman–Valiron theory. Applying Theorem C to the entire function , with and , we compare the asymptotic relations (5) and (6) to conclude that where is a constant. This implies , which means that is of order , normal type. So is also of order , normal type, because and satisfy a polynomial relation of the form , where is a polynomial with constant coefficients.
Lemma 2. Suppose that is a meromorphic solution of . If for two different points and in , then is a rational function of , where .
Proof. As in the previous lemma, the genus of is zero. Let be a function in with a single simple pole at and a single simple zero at . Then is an entire function of order , normal type (by Lemma 1) omitting [math] and (by Theorem D, d). So for some . Since is a generator of , by Lüroth’s theorem, both and are rational functions of and the lemma follows.
Lemma 3. Suppose that is even, the Riemann surface is of genus zero, is a non-constant meromorphic solution of , and for at most one point . Then the Abelian differential is exact, that is for some .
Proof. It is sufficient to show that under the assumptions of Lemma 3, the integral of over every closed path in is zero. As is of genus zero, we only have to consider residues of . By Theorem D, a), all poles of our differential belong to the set .
Consider first a point with . By Theorem D, c), we have a meromorphic solution with a pole of order at zero, such that the corresponding function has the property . In a neighborhood of we have a Puiseaux expansion
[TABLE]
with some positive integer . We substitute and obtain
[TABLE]
where is the residue of at . Now we notice that for even ,
[TABLE]
Using this, we integrate (7) over a small circle around [math] in the -plane, described times anticlockwise. We obtain that , so .
Now we consider a point with . By the assumptions of the lemma, there is at most one such point. Then the residue of at is zero because the sum of all residues of a differential on a compact Riemann surface is zero. This proves the lemma.
Using (8) and Lemma 3, if the assumptions of Lemma 3 are satisfied, we can rewrite our differential equation
[TABLE]
as
[TABLE]
where is an integral of the exact differential , and is a constant that depends on the particular solution . We have the relation .
Lemma 4. For a given differential equation of the form , there are only finitely many formal Laurent series with a pole at zero that satisfy the equation.
Proof. By making a linear change of the independent variable, we may assume that
[TABLE]
Then
[TABLE]
Now we substitute a Laurent series with undetermined coefficients
[TABLE]
to the equation (9), which is a consequence of (10). With even we have:
[TABLE]
and
[TABLE]
In the last formula, the symbol stands for a finite sum of products of the coefficients of the series (11) which contain no coefficients with . Substituting to (9) and comparing the coefficients at we obtain
[TABLE]
This equation has finitely many non-zero roots . We have
[TABLE]
Further we obtain
[TABLE]
Substituting here the value of from (12), we see that the coefficient at is different from zero, because
[TABLE]
Thus is uniquely determined from (13). The situation is analogous for all coefficients with . These coefficients are uniquely determined from the equation (9) once is chosen.
Now we consider the coefficients with . We have
[TABLE]
Again we substitute the value of from (12) and conclude that the coefficient at equals
[TABLE]
This coefficient is zero for a single value of , namely . Thus cannot be determined from the equation (9), but once and are chosen, the rest of the coefficients of the series (11) are determined uniquely.
To determine we invoke the equation (10):
[TABLE]
where the dots stand for the terms of lower degrees. We have
[TABLE]
Substituting this to our equation (14) we write the condition that the constant terms in both sides of (14) are equal. This condition is a polynomial equation in (it is linear with respect to ) and the coefficient at in this equation equals
[TABLE]
This expression is not zero because each term of the sum is positive. Thus is determined uniquely, and this completes the proof of the lemma.
Remark. It follows from this proof that the only meromorphic solutions of the differential equations
[TABLE]
are exponential polynomials when and functions where and is an appropriate constant.
The rest of the proof of Theorem 1 is a repetition of the argument from [8].
By Theorems A and B, we may assume that is of genus zero, and is even. In view of Lemmas 2 and 3, it is enough to consider the case that the differential is exact. Then every solution of (2) also satisfies (10) with some constant .
Assume that is a transcendental meromorphic solution of (10), having at least one pole. By Theorem D, d), c), has infinitely many poles The functions satisfy the assumptions of Lemma 4, therefore some of them are equal. We conclude that is a periodic function. By making a linear change of the independent variable we may assume that the smallest period is .
Consider the strip .
Case 1. has infinitely many poles in . Applying Lemma 4 again, we conclude that has a period in , so is doubly periodic.
Case 2. is bounded in for some . Since is -periodic, we have where is meromorphic in . As is bounded in some neighborhoods of [math] and , we conclude that is rational.
Case 3. has finitely many poles in and is unbounded in for every . As is -periodic, we write where is meromorphic in . Now has finitely many poles and is unbounded either in a neighborhood of [math] or in a neighborhood of . Suppose that it is unbounded in a neighborhood of . Then the set , where is large enough has an unbounded component containing no poles of . On this component , the function satisfies a differential equation
[TABLE]
where is some constant and or is one of the numbers from Theorem D. Applying Theorem C in as we did in the proof of Lemma 1, we obtain that and that has a pole at infinity. Similar argument works for the singularity at [math], so is rational, and this completes the proof.
4 Appendix
Proof of Theorem D. Statement a) is a special case of [9, Th. 10], but we give a simple independent proof using Theorem C. Proving it by contradiction, suppose that has a pole at a point such that . Let be a disk of radius centered at , and a component of containing . We assume that the disk is so small that contains no other poles of , except the pole at . Let be a meromorphic solution of our equation (2) and consider the map given by . The image of this map is dense in and the point is evidently omitted by . Let be a component of the preimage Consider the meromorphic function . It is holomorphic and unbounded in , and for . We conclude that is unbounded. Now we apply Theorem C to in .
Set and let be defined as in (4). For any , we choose a point with and . By Theorem C, we have
[TABLE]
where outside a set of finite logarithmic measure.
From Lemma 6.10 of [3], we have for every ,
[TABLE]
as outside a set of finite logarithmic measure.
Differentiating the equation we obtain
[TABLE]
where is a polynomial. On the other hand, from the Puiseaux expansion at the point we obtain
[TABLE]
where is a constant and . Combining (17) and (18) we obtain
[TABLE]
Inserting to this asymptotic relation and using (15) and (16) we obtain a contradiction which proves a).
Consider now a point . From the Puiseaux expansion we obtain
[TABLE]
If has a preimage under the map , then this preimage is a pole of . If this pole is of order we have and as . Substituting to (19) we conclude that . Thus if has at least one preimage under then with a positive integer , and every preimage of is a pole of order of . This implies d).
Now suppose that a point has only finitely many preimages. Let be a neighborhood of infinity, and a component of the preimage . We may assume that is so small that does not contain other poles of except . Let be a component of the preimage . If is bounded then is a ramified covering of a finite degree, and takes the value somewhere in . As we assume that is transcendental but has only finitely many preimages, there should exist an unbounded component . Choosing a smaller if necessary, we achieve that this unbounded component contains no -preimages of . Then is a holomorphic function in , , and is unbounded in . Applying Theorem C to the function in we obtain the asymptotic relation (5). Putting in this relation, taking (16) into account, and comparing with (19) we conclude that in (19). This implies c). Thus in any case or , which proves b).
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[1] N. Akhiezer, Elements of the theory of elliptic functions. Transl. Math. Monogr., 79 . AMS, Providence, RI, 1990.
- 2[2] A.F. Beardon and T.W. Ng, Parametrizations of algebraic curves. Ann. Acad. Sci. Fenn. Math. 31 (2006), no. 2, 541-554.
- 3[3] W. Bergweiler, P. Rippon and G. Stallard, Dynamics of meromorphic functions with direct or logarithmic singularities, Proc. London Math. Soc. (to appear, see also ar Xiv:0704.2712).
- 4[4] S. Bank and R. Kaufman, On Briot-Bouquet differential equations and a question of Einar Hille. Math. Z. 177 (1981), no. 4, 549–559.
- 5[5] Ch. Briot et J. Bouquet, Théorie des fonctions doublement périodiques et, en particulier, des fonctions elliptiques; Paris, Mallet-Bachelier, 1859.
- 6[6] Ch. Briot et J. Bouquet, Intégration des équations différentielles au moyen de fonctions elliptiques, J. École Polytechnique, 21 (1856) 199–254.
- 7[7] Y. M. Chiang and R. Halburd, On the meromorphic solutions of an equation of Hayman, J. Math. Anal. Appl. 281 (2003) 663–667.
- 8[8] A. Eremenko, Meromorphic solutions of equations of Briot–Bouquet type, Teor. Funktsii, Funk. Anal. i Prilozh., 38 (1982) 48–56. English translation: Amer. Math. Soc. Transl. (2) 133 (1986) 15–23.
