betaFeedback

WorksMathematics

Uniform bounds for planar polynomial limit cycles

Abstract

For every degree, we prove that the number of isolated periodic orbits of a real planar polynomial vector field is bounded by a finite constant depending only on that degree. This establishes the uniform boundedness assertion in the second part of Hilbert's sixteenth problem. The proof uses separation of asymptotic expansions on nested complex domains and a finite-dimensional counting argument.

Introduction

Let V=(P,Q)V=(P,Q) be a real polynomial vector field on R2\mathbb{R}^2. A periodic orbit is the image of a nonconstant periodic solution of

x˙=P(x,y),y˙=Q(x,y).\dot{x}=P(x,y),\qquad\dot{y}=Q(x,y).

It is a limit cycle if some open neighborhood of that image contains no other periodic orbit. We count geometric images once, without multiplicity.

Theorem 1.1 (Uniform boundedness). For each integer d≥1d \ge1, there is a finite nonnegative integer B(d)B(d) such that every real planar polynomial vector field of degree at most dd has at most B(d)B(d) limit cycles in the whole plane.

There is no restriction on the coefficients, the locations of the cycles, or their stability. Theorem 1.1 gives a positive resolution of the uniform boundedness assertion in the second part of Hilbert’s sixteenth problem. It does not furnish an effective formula for B(d)B(d).

Historical context

Hilbert’s sixteenth problem asks, in its differential-equation part, for the maximum number and arrangement of limit cycles of planar polynomial systems [8] [Problem 16, p. 465]. Two finiteness assertions must be distinguished. Individual finiteness concerns one fixed vector field; uniform boundedness asks for a bound depending only on its degree [9] [pp. 301–304]. Even finiteness for every member of a family does not by itself bound the numbers occurring as its coefficients vary.

Dulac claimed a proof of individual finiteness in 1923 [4]; after a gap in that argument was identified [11], Écalle and Ilyashenko developed proofs using complex and asymptotic analysis of return maps [5, 12, 13]. Yeung identifies a coefficient-closure obstruction in a leading-term argument of Ilyashenko’s 1991 monograph [22]. This concerns that proof mechanism; it is not a counterexample to individual finiteness.

The geometric approach to uniformity studies the limiting sets of periodic orbits and bounds the number of nearby cycles in a parameter family. A foundational local result is Bautin’s sharp bound of three for cycles bifurcating from a nondegenerate weak focus or center of a quadratic vector field, under perturbations within the quadratic family [1] [§3]. Roussarie develops the study of limiting periodic sets through graphics and desingularization [20]. Ilyashenko and Yakovenko prove finite cyclicity for elementary polycycles in generic finite-parameter families [10]; Kaloshin subsequently gives an explicit cyclicity bound in terms of the number of parameters [15]. The elementary hypothesis requires each singularity of the polycycle to have at least one nonzero eigenvalue. It and the genericity assumptions restrict these results. Another major partial advance concerns the infinitesimal problem: Binyamini, Novikov, and Yakovenko give explicit bounds for zeros of Abelian integrals and for cycles arising from nonsingular Hamiltonian ovals in the stated nonconservative perturbations [3]. These results illustrate both the power of local transition analysis and the extra work required at unrestricted degenerations.

There is a parallel development through definability and quasianalytic asymptotics. Kaiser, Rolin, and Speissegger construct an o-minimal expansion containing the transition maps at nonresonant hyperbolic singularities. For small analytic unfoldings that preserve the singular points and their linear parts, they obtain uniform bounds on isolated return-map fixed points near polycycles with such singularities [14]. Galal, Kaiser, and Speissegger construct Ilyashenko algebras with injective transserial asymptotics and a corresponding Hardy field [7]. They prove closure under field operations and differentiation for germs in one variable. The multivariable closure needed for a parameter argument is a separate issue. The present proof establishes separation, differentiated implicit closure, and projection bounds for its own passage class. Those properties are proved in Sections 3–5 and 11; definability of arbitrary passage maps is not an assumption.

The proof below treats coefficient variation and degeneration together. It does not use individual finiteness as a premise. This is also why cycles that move arbitrarily far from the origin and nonhyperbolic isolated cycles must be included in the last step of the argument.

Proof strategy

Choose a short transverse interval through a point of a periodic orbit, with coordinate rr. Following nearby trajectories once around the orbit defines a real analytic return map Π(r)\Pi(r). The nearby limit cycles correspond to isolated zeros of the displacement Π(r)−r\Pi(r)-r. A simple zero has Π′(r)≠1\Pi'(r) \ne1; such a cycle is hyperbolic. The difficulty is to bound these zeros while the coefficients vary and the passages making up the return map degenerate.

A local saddle already exhibits the scales that occur. For x˙=x\dot{x}=x, y˙=−λy\dot{y}=-\lambda y, with λ>0\lambda>0, the passage from (r,1)(r,1) to (1,rλ)(1,r^\lambda), 0<r<10<r<1, has flight time L=log⁡(1/r)L=\log(1/r) and contraction exponent W=λLW=\lambda L. When r→0r\to0 and λ→0\lambda\to0, the first clock can diverge while the second stays bounded or diverges at a different rate. The later passage models treat such differing scales together with their smaller correction terms.

We represent a return by equations matching the endpoints of finitely many local passages. This formulation has a predecessor in Ilyashenko and Yakovenko’s cyclic systems for elementary polycycles [10]. Here the cuts are allowed to move along the orbit, giving continuous families of representations. Section 11 shows that, within each regular chart of a fixed matching system, representations of different hyperbolic cycles lie in different connected components of the fiber. It therefore remains to bound those components uniformly as the field coefficients and preparation parameters vary.

The counting argument reduces this uniform bound to a different assertion: each fixed finite auxiliary system has only finitely many isolated solutions when all its variables, including those later used as parameters, are allowed to vary. We call this absolute isolated-zero finiteness. It must hold for the differential and algebraic systems created by the counting argument, including their graph relations and multiplier equations, with the original open domains retained. The analytic and geometric parts of the proof establish these hypotheses for the matching systems.

Here is the organization of the proof.

  1. Sections 2–5 prove the abstract absolute-zero criterion in Theorem 5.7. Along an alleged escaping sequence, a finite saturated flag records the ordered large scales and their exact logarithmic relations. Theorem 3.2 compares two lateral trees of asymptotic coefficients and identifies the first nonzero term. If both trees vanish, the function vanishes exactly. A differentiated implicit calculus and an induction on ordinary variables then turn this information into a local chart through each sufficiently far solution. An independent large coordinate survives in a zero chart, contradicting isolation. The differentiated implicit calculus

is stated in Theorem 4.4.

  1. Sections 6–8 verify the packet and retestability hypotheses of that criterion for the analytic passage maps used later (Theorems 6.1, 7.1, and 8.1). The models include additive transfers, regular or stable transfers with one large clock, and transfers with two clocks that may be comparable or widely separated. Their arguments are treated independently before endpoint relations are imposed. In particular, exponentially small terms and the differences between lateral determinations remain part of the packet data.

  1. Sections 9 and 10 give a finite geometric description for fields of bounded degree. Strong one-variable subanalytic preparation is used on the scalar fields, before passage maps are introduced. It supplies finitely many boxes and annuli (Theorem 9.7), which reduce to the preceding analytic models (Theorem 10.1). A crossing argument for periodic Jordan curves bounds the length of the resulting passage words. This finite description depends on the degree; it is fixed before any sequence is chosen for an absolute-zero test.

  1. Section 11 constructs the matching equations with their exact graph relations and verifies absolute finiteness for the required closure (Theorem 11.4). The independent counting principle is Theorem 11.2. Within each regular chart, representations of different hyperbolic cycles lie in different connected components of the fiber. A suitable small signed rotation produces at least as many hyperbolic cycles as any prescribed finite collection of isolated cycles (Lemma 11.8). Spatial rescaling then gives the assertion throughout the plane.

Figure 1 records where the analytic and geometric parts meet in the counting argument. The analytic flags and charts in an absolute-zero contradiction may depend on the chosen sequence. They are not used as a finite covering of coefficient space. The finite geometric description and the projection-counting argument supply the uniformity.

Main theorem interfaces: Abstract isolated-zero criterion; Sections 2–5; Verified passage packets; Sections 6–8; Finite geometric templates; reductions to transfer models; Sections 9–10; terminal reductions; Matching systems: absolute isolated-zero finiteness for every finite system in the required differential and algebraic closure; Theorem 11.4; Projection count for regular fibers; Theorem 11.2; Uniform hyperbolic-cycle bound in a fixed square; rotation and rescaling; A bound depending only on degree, throughout the plane; Theorem 1.1

Figure 1. Main theorem interfaces. Matching combines the finite geometric description with the packet criterion, retaining the exact graph ties and open domains. The independent projection theorem converts absolute finiteness for the full required closure into uniform component bounds. The sequence-dependent flags and internal factorizations used in the absolute-zero test do not enlarge the finite geometric alphabet. The packet input to Sections 9–10 concerns the terminal reductions, not the preparation of scalar fields.

For a first reading, the theorem statements cited above and the final proof in Section 11 display the chain of implications. Definition 11.1 specifies the precise closure required at the counting interface. The preceding sections establish its analytic and geometric inputs.

Analytic ingredients

Three features of the argument deserve emphasis.

A packet records an actual analytic function together with two trees of successive asymptotic expansions and the domain, transition, and error estimates relating them. The two signs refer to lateral determinations on either side of the real path; higher-band raw realizations may be smooth interpolations of holomorphic determinations.

First, the separation theorem uses a tree of expansions on successive complex bands. It does not assume a simultaneous convergent series in all small exponentials. The hypotheses record retained arguments, permissible error costs, differentiated transition estimates, and growth on the full fringes of the domains. The proof upgrades arbitrarily high fixed exponential decay to exact vanishing by weighted Cauchy corrections and maximum principles. The domain and error conditions are indispensable: real asymptotic flatness alone gives no such conclusion.

Second, regular ordinary implicit changes must be compatible with dependent scales that can rotate rapidly under tiny complex perturbations. The construction uses high real Taylor jets and interpolation into small comparison tubes, with a fixed finite coherence order at each stage. It does not require one ordinary complex neighborhood valid for all derivative orders. The later center-lifting argument additionally keeps the power exponents controlling maps, inverses, and neighborhoods

fixed while the accuracy of auxiliary records is increased. This quantitative property permits exact correction of approximate chart hits.

Third, the projection argument is formulated for an analytic class closed under the particular derivatives, graph variables, and multiplier equations used in its proof. Proper barriers and a fixed generic tilt reduce the number of parameters. The resulting absolute-to-uniform implication is independent of the particular planar passage models and can be applied to other classes for which the same hypotheses are established.

Notation and conventions

The technical constructions are local at a limiting ordinary analytic germ. An asymptotic assertion is made for a fixed finite request: a finite collection of expressions, derivative orders, Taylor accuracies, and transition estimates. Constants, sufficiently far starting points, and smaller neighborhoods may depend on that request. Whenever an exponent must be independent of a later accuracy choice, this is stated explicitly.

The notation used throughout the analytic part is summarized below; see Definitions 2.1 and 3.1. Local clock and state variables are introduced separately in each passage model.

NotationMeaning
Z1≫⋯≫Zm≫1Z_1 \gg\cdots\gg Z_m \gg1Ordered scales of a finite flag
fi=log⁡Zi,hi=e−Zif_i = \log Z_i,\quad h_i = e^{-Z_i}Logarithmic scales and small exponentials
SSTuple of independent, or free, flag scales
t,tct, t_cOrdinary inputs and their interior limiting germ point
bib_iBounded ordinary backgrounds in exact log relations
σ(i)\sigma(i)Leading later index in a dependent log relation
s=u+ivs = u + ivComplex parameter of a selected flag path
Λi=fi′\Lambda_i = f_i'Logarithmic rate along the real path
Vi=efi/ΛiV_i = e^{f_i}/\Lambda_iPrecision scale at level ii
Sαj,±S_{\alpha}^{j,\pm}Raw packet in band jj at prefix α\alpha
hν=exp⁡(−∑iνiZi)h^\nu= \exp(-\sum_i \nu_i Z_i)Exponential shift with lexicographic order

Table 1.

Lexicographic order starts at the largest scale. A positive exponent vector has positive first nonzero coordinate; later coordinates need not be nonnegative. A vanishing coefficient means a vanishing ordinary analytic germ. Complex logarithms and powers always use the specified lifted determinations. A dependent scale is evaluated at its current background values, including after an implicit change of ordinary variables.

All differentiations of primitive packets are in their independent arguments. Equations tying those arguments are imposed only afterward; differentiation of the resulting physical maps uses the chain rule on the tied graph. This distinction is used repeatedly in the terminal reductions and the matching equations.

Finite logarithmic flags and angular paths

An unbounded sequence of equation arguments can involve several widely separated scales. This section records those scales and their exact logarithmic relations in a finite flag, then constructs complex paths through the sequence anchors. Along these paths, the largest scale approaches a vertical complex direction before the smaller scales do. We construct nested contours on which their real parts remain positive and quantitatively separated. These are the domains for the separation argument in Section 3. The final two lemmas control anchor-dependent growth and simultaneous avoidance of finitely many lattices of poles.

Sequence conventions and exact logarithmic relations

All constructions in this section concern a selected sequence of real points. Passing to a subsequence is allowed. For positive quantities, X≫YX \gg Y means X/Y→+∞X/Y \to+\infty. A comparison involving finitely many quantities, and comparisons of a finite list at every fixed iterated exponential height, can be decided after a diagonal subsequence. None of the resulting numbers of scales or heights is asserted to be uniform over the original sequences.

The ordinary variables have a finite limit tct_c in the interior of an analytic argument neighborhood. Bounded ratios that a formula needs are included among these arguments. Neighborhoods can be shrunk for each finite request. The endpoint inequalities defining the actual problem can nevertheless approach their boundary. An ordinary coefficient equal to zero below means the zero analytic germ at tct_c, and not merely a function vanishing on the selected sequence.

Definition 2.1 (Logarithmic flag). A logarithmic flag consists of positive scales

Z1≫Z2≫⋯≫Zm≫1,fi=log⁡Zi,hi=e−Zi.Z_1 \gg Z_2 \gg\cdots\gg Z_m \gg1,\qquad f_i = \log Z_i,\qquad h_i = e^{-Z_i}.

each declared either free or dependent. The tuple of free scales is denoted by SS. A dependent scale has the exact relation

fi=∑j>iaijZj+bi,σ(i)=min⁡{j:aij≠0},ai,σ(i)>0.f_i = \sum_{j>i} a_{ij}Z_j + b_i,\qquad\sigma(i) = \min\{j : a_{ij} \ne0\},\qquad a_{i,\sigma(i)} > 0.

The coefficients aija_{ij} are fixed real numbers, and the backgrounds bib_i are bounded ordinary coordinates, or their current values on an ordinary chart. The free scales are independent arguments; dependent scales are always evaluated by (2.1).

The saturation condition below will ensure that bounded changes of the backgrounds preserve the ordered hierarchy. Two scales belong to the same height block if each is eventually bounded by a finite iterated exponential of the other. A flag is saturated if, in every block containing a dependent node, all dependent nodes precede all free nodes and

log⁡Zfirst free=o(log⁡Zlast dependent).\log Z_{\text{first free}} = o(\log Z_{\text{last dependent}}).

An all-free block has no additional condition.

The relation defining height blocks is an equivalence relation: a composition of finitely many iterated exponentials is another finite iterate, after increasing the iterate to absorb fixed constants. Its classes are consecutive in the ordering of the scales. A dependent node and its leading child σ(i)\sigma(i) are in the same block, because

fi=ai,σ(i)Zσ(i)(1+o(1)).f_i = a_{i,\sigma(i)}Z_{\sigma(i)}(1+o(1)).

In particular a dependency chain ends at a free node in the same block.

We use lifted logarithms and powers in complex continuations. Thus fif_i continues as the expression in (2.1), even if Zi=efiZ_i=e^{f_i} winds in the plane. An exponential monomial is

hν=exp⁡(−∑i=1mνiZi).h^\nu= \exp\left(-\sum_{i=1}^{m}\nu_i Z_i\right).

Its order is the sign of the first nonzero entry of ν\nu, with the largest scale first. This convention does not refer to the numerical sign of the monomial. Known signs and ordinary units can be kept as separate factors. Later expansions will be trees indexed by successive prefixes of ν\nu; no simultaneous convergence of all these monomials is implicit in this notation.

Inserting logarithms expresses fixed real powers of large inputs as exponential monomials times bounded nonvanishing ordinary analytic factors. A later insertion of −log⁡∣t∣-\log|t|, when an ordinary coordinate tt tends to zero, must also preserve the large coordinates already free at that stage. The next lemma supplies these operations.

Lemma 2.2 (Finite saturation and preservation of free coordinates). A finite collection of large real inputs admits a flag after a fixed linear change of coordinates and the addition of bounded ordinary coordinates. One can insert the logarithm of any free scale, retaining the old scale values, by making that scale dependent and adding at most one positive free residual or one bounded ordinary residual. The free residual is O(log⁡S)O(\log S) if the scale being converted is SS.

A flag admits a finite saturation by such insertions. After saturation, promoting one ordinary coordinate t→0t \to0 of known sign, by inserting −log⁡∣t∣-\log|t|, admits a finite further saturation in which every old free coordinate stays free. The number of ordinary coordinates does not increase in this promotion.

Proof. For the initial linear reduction, choose a largest absolute value among a basis of the finite span of the inputs. After a subsequence, the ratios of the other elements to it have finite limits. Subtract these limiting multiples. In the remaining span repeat the operation, changing the sign of an unbounded residual when necessary. At each step the dimension of the remaining span drops. A bounded residual is retained as an ordinary coordinate, even if its selected values vanish; every unbounded residual chosen as a new scale is strictly smaller than its predecessor. This gives a finite ordered scale basis with constant coefficients.

Apply the same elimination to log⁡S\log S against the existing scale basis, beginning with the largest scale. A scale much larger than log⁡S\log S has zero coefficient. After subtracting its limiting multiple at each comparable scale, the remainder is either bounded or, after changing sign, a new positive scale separated from every old scale. All the subtracted terms and the remainder are O(log⁡S)O(\log S). Hence only one new residual is needed. Since log⁡S=o(S)\log S=o(S), its expression uses only scales below SS; it is therefore a relation of the form Equation (2.1). Its first nonzero coefficient is positive, since log⁡S→+∞\log S \to+\infty. The old scale values have not been changed.

Process the blocks containing dependent nodes from largest to smallest. On entering a block fix its then smallest scale TT. Convert every free scale RR in that block for which log⁡R≠o(log⁡T)\log R \ne o(\log T), and repeat this test only on any new residual still in the block. Every continuing chain satisfies

Rq+1≤Cqlog⁡Rq.R_{q+1} \le C_q \log R_q.

The initial node of a chain is bounded by exp⁡N(T)\exp^N(T) for a fixed NN, by the definition of the block. Finitely many applications of Equation (2.3) give a residual O(T)O(T) and then one O(log⁡T)O(\log T). The logarithm of the latter is o(log⁡T)o(\log T), so it is no longer converted. Fixed constants can be absorbed by an extra logarithmic step. A residual that has entered a lower block has already left the chain under consideration. There are finitely many initial chains and each step has at most one child; thus the procedure is finite. Residuals entering lower blocks are free until that block is processed.

Every converted node in the block has, after the comparison subsequence, log⁡R≥clog⁡T\log R \ge c\log T. Every free node left at the end has log⁡F=o(log⁡T)\log F=o(\log T). The original dependent nodes also have logarithms at least log⁡T\log T. Consequently all dependent nodes precede all remaining free nodes, and Equation (2.2) holds. All-free blocks need no processing.

For a later promotion first insert −log⁡∣t∣-\log|t| by the same linear elimination. In each old block with dependent nodes keep the old cut

L=log⁡Zold last dependent.L = \log Z_{\mathrm{old\ last\ dependent}}.

Only the new residual chain is tested for conversion, with criterion log⁡R≠o(L)\log R \ne o(L). For every old free node FF, log⁡F=o(L)\log F=o(L). For each newly converted node RR, the comparison subsequence gives log⁡R≥cL\log R\ge cL; therefore

log⁡Flog⁡R⟶0,RF⟶+∞.\frac{\log F}{\log R}\longrightarrow0,\qquad\frac{R}{F}\longrightarrow+\infty.

In particular a new residual below an old free node is never converted in this block. Thus no old free coordinate is consumed, and the new last-dependent cut still dominates the logarithms of all old free nodes. The residual-chain argument proves termination here as well. If the chain enters an old all-free block, leaving its residual free introduces no dependent node into that block. Inserting new scales cannot join two distinct old blocks: transitivity of finite-height comparability would already have put their old members in one block. The coordinate tt has been consumed; at most the final step of its single chain produces a bounded residual. This proves the assertion about the number of ordinary coordinates. The inverse expression for tt has the form

t=sgn⁡(t)exp⁡(−∑iνiZi−B),t=\operatorname{sgn}(t)\exp\left(-\sum_i \nu_i Z_i-B\right),

with BB bounded and with a positive first nonzero large coefficient when such a coefficient occurs.

The flag is now fixed. The next assertion concerns a different operation: we hold its free inputs fixed and vary its bounded backgrounds, always reevaluating dependent scales by their exact relations. This stability will be needed when ordinary variables are changed implicitly.

Lemma 2.3 (Stability under bounded backgrounds). On a saturated output flag of Lemma 2.2, bounded variations of the backgrounds at fixed free coordinates preserve the ordered scale hierarchy, uniformly on each fixed bounded admissible real background set. More precisely, let

H=∑i=1mciZi+B,H=\sum_{i=1}^{m}c_iZ_i+B,

where the cic_i are fixed real numbers, their vector is nonzero, and BB is uniformly bounded on that background set. If kk is the first index with ck≠0c_k\ne0, then

HckZk⟶1\frac{H}{c_kZ_k}\longrightarrow1

uniformly under those variations; the denominator uses the current, reevaluated value of ZkZ_k. Thus the leading index and sign are preserved. Dependent logarithms may enter such a comparison after substitution of their exact triangular formulas. No sign assertion is made when the resulting large part vanishes, or for an expression containing an uninserted free logarithm.

Proof. Process the saturated output flag’s height blocks from smallest to largest. Each evaluated lower block is bounded above, uniformly on the fixed bounded real background set, by a fixed finite exponential iterate of its largest free member. Indeed it contains only finitely many triangular relations, whose coefficients are fixed and whose backgrounds are bounded. Every free member of a higher block dominates every such fixed iterate. Thus lower-block values remain negligible compared with every free member of the block being processed.

The free members of the current block are unchanged and ordered. If the block has dependent members, let dd be its last one and put k=σ(d)k=\sigma(d). The leading child lies in the same block and, by saturation, is free. Its value is therefore unchanged. All later scales have already been ordered, so the triangular relation gives

fd=adkZk(1+o(1))f_d=a_{dk}Z_k(1+o(1))

uniformly on the background set. The logarithm of the first free member is o(fd)o(f_d) on the original sequence. It remains so uniformly, since that logarithm and ZkZ_k are fixed. This proves the last-dependent/first-free boundary.

Now proceed upwards through the dependent members. The difference of two consecutive dependent log relations is a fixed affine combination of later scales and a bounded remainder. Its large part is nonzero, since otherwise their logarithmic gap on the original sequence would be bounded. Its first nonzero coefficient is positive on that sequence. The later scales are already ordered uniformly by the induction, so this coefficient still dominates. This proves all remaining gaps and hence the full hierarchy uniformly.

Finally divide HH by ckZkc_kZ_k, evaluated at the current backgrounds. Every later-scale ratio tends uniformly to zero, as does the bounded remainder divided by ZkZ_k. This proves the precise assertion about affine comparisons. A combination involving dependent logarithms first reduces to this form by exact substitution. If all large coefficients cancel, their cancellation is an identity of fixed coefficient vectors, leaving the current bounded remainder; it supplies no additional sign conclusion.

Paths through far anchors

We keep a saturated flag fixed and vary its free coordinates along a complex path, with parameter s=u+ivs=u+iv. The path must pass through the chosen anchor without depending on the ordinary backgrounds, and it must preserve scale separation on every required fixed backward interval. Its angular purpose comes from the identity

ℜZi=eℜficos⁡(ℑfi).\Re Z_i=e^{\Re f_i}\cos(\Im f_i).

As ∣v∣|v| grows, the larger scales must approach the directions ℑfi=±π/2\Im f_i=\pm\pi/2 before the smaller ones. With Λi=fi′\Lambda_i=f_i' on the real ray, the path estimates below give ℑfi(u+iv)∼vΛi(u)\Im f_i(u+iv)\sim v\Lambda_i(u): the level-ii direction is therefore reached at distance about π/(2Λi)\pi/(2\Lambda_i) from that ray. We first construct the paths and estimate these logarithmic rates; the next subsection chooses the precise contours and their margins.

Write polylog XX for a bound C(1+log⁡X)NC(1+\log X)^N, where C,NC,N may change between occurrences and may depend on a fixed derivative order. In every use the number of derivatives and the other requests are finite. Thresholds, constants, and ordinary neighborhoods may depend on this finite request.

References

  1. [1]N. N. Bautin. On the number of limit cycles appearing with variation of the coefficients from an equilibrium state of the type of a focus or a center. Matematicheskii Sbornik (N.S.), 30(72)(1):181–196, 1952. URL https://www.mathnet.ru/eng/sm5421. In Russian.
  2. [2]Edward Bierstone and Pierre D. Milman. Semianalytic and subanalytic sets. Publications Mathématiques de l’IHÉS, 67:5–42, 1988. doi: 10.1007/BF02699126. URL https://www.numdam.org/item/10.1007/BF02699126.pdf.DOI
  3. [3]Gal Binyamini, Dmitry Novikov, and Sergei Yakovenko. On the number of zeros of Abelian integrals: A constructive solution of the infinitesimal Hilbert sixteenth problem. Inventiones Mathematicae, 181(2):227–289, 2010. doi: 10.1007/s00222-010-0244-0.DOI
  4. [4]Henri Dulac. Sur les cycles limites. Bulletin de la Société Mathématique de France, 51:45–188, 1923. doi: 10.24033/bsmf.1031. URL https://numdam.org/articles/10.24033/bsmf.1031/.DOI
  5. [5]Jean Écalle. Introduction aux fonctions analysables et prévue constructive de la conjecture de Dulac. Actualités Mathématiques. Hermann, Paris, 1992. URL https://www.imo.universite-paris-saclay.fr/~jean.ecalle/ecadulac.html. The linked author-hosted page reproduces the preface.
  6. [6]A. M. Gabrielov. Projections of semi-analytic sets. Functional Analysis and Its Applications, 2(4):282–291, 1968. doi: 10.1007/BF01075680. URL https://www.math.purdue.edu/~gabrea/ProjectionsOfSemianalyticSets.pdf.DOI
  7. [7]Zeinab Galal, Tobias Kaiser, and Patrick Speissegger. Ilyashenko algebras based on transserial asymptotic expansions. Advances in Mathematics, 367:107095, 2020. doi: 10.1016/j.aim.2020.107095.DOI
  8. [8]David Hilbert. Mathematical problems. Bulletin of the American Mathematical Society, 8 (10):437–479, 1902. doi: 10.1090/S0002-9904-1902-00923-3. URL https://people.reed.edu/~davidp/341/resources/hilbert.pdf. Translated by Mary Winston Newson.DOI
  9. [9]Yu. Ilyashenko. Centennial history of Hilbert’s 16th problem. Bulletin of the American Mathematical Society, 39(3):301–354, 2002. doi: 10.1090/S0273-0979-02-00946-1.DOI
  10. [10]Yu. Ilyashenko and Sergei Yakovenko. Finite cyclicity of elementary polycycles in generic families. In Concerning the Hilbert 16th Problem, volume 165 of American Mathematical Society Translations, Series 2, pages 21–95. American Mathematical Society, Providence, RI, 1995. doi: 10.1090/trans2/165/02.DOI
  11. [11]Yu. S. Ilyashenko. Dulac’s memoir “on limit cycles” and related problems of the local theory of differential equations. Russian Mathematical Surveys, 40(6):1–49, 1985. doi: 10.1070/RM1985v040n06ABEH003701. URL https://www.mathnet.ru/eng/rm2786.DOI
  12. [12]Yu. S. Ilyashenko. Finiteness theorems for limit cycles. Russian Mathematical Surveys, 45(2):129–203, 1990. doi: 10.1070/RM1990v045n02ABEH002335. URL https://www.mathnet.ru/eng/rm4718.DOI
  13. [13]Yu. S. Il’yashenko. Finiteness Theorems for Limit Cycles, volume 94 of Translations of Mathematical Monographs. American Mathematical Society, Providence, RI, 1991. doi: 10.1090/mmono/094. Translated from the Russian by H. H. McFaden.DOI
  14. [14]Tobias Kaiser, Jean-Philippe Rolin, and Patrick Speissegger. Transition maps at non-resonant hyperbolic singularities are o-minimal. Journal für die reine und angewandte Mathematik, 636:1–45, 2009. doi: 10.1515/CRELLE.2009.081.DOI
  15. [15]Vadim Kaloshin. The existential Hilbert 16-th problem and an estimate for cyclicity of elementary polycycles. Inventiones Mathematicae, 151(3):451–512, 2003. doi: 10.1007/s00222-002-0244-9.DOI
  16. [16]A. G. Khovanskii. Real analytic varieties with the finiteness property and complex Abelian integrals. Functional Analysis and Its Applications, 18(2):119–127, 1984. doi: 10.1007/BF01077822. URL https://www.math.utoronto.ca/askold/1984-FaA-2-English.pdf.DOI
  17. [17]A. G. Khovanskii. Cycles of dynamical systems on the plane and Rolle’s theorem. Siberian Mathematical Journal, 25(3):502–506, 1984. URL https://www.math.utoronto.ca/askold/1984-Sib_mat-zj-3English.pdf.DOI
  18. [18]Jean-Marie Lion and Jean-Philippe Rolin. Théorème de préparation pour les fonctions logarithmico-exponentielles. Annales de l’Institut Fourier, 47(3):859–884, 1997. doi: 10.5802/aif.1583. URL https://www.numdam.org/item/10.5802/aif.1583.pdf.DOI
  19. [19]Andre Opris. On preparation theorems for ℝ_an,exp-definable functions. Journal of Logic & Analysis, 15(1):1–50, 2023. doi: 10.4115/jla.2023.15.1. URL https://math.hawaii.edu/home/jla/450-1521-1-PB.pdf.DOI
  20. [20]Robert Roussarie. Bifurcation of Planar Vector Fields and Hilbert’s Sixteenth Problem, volume 164 of Progress in Mathematics. Birkhäuser, Basel, 1998. doi: 10.1007/978-3-0348-8798-4.DOI
  21. [21]Arthur Sard. The measure of the critical values of differentiable maps. Bulletin of the American Mathematical Society, 48(12):883–890, 1942. doi: 10.1090/S0002-9904-1942-07811-6. URL https://webhomes.maths.ed.ac.uk/~vlranick/papers/sard.pdf.DOI
  22. [22]Melvin Yeung. Dulac’s theorem revisited. Qualitative Theory of Dynamical Systems, 24:57, 2025. doi: 10.1007/s12346-025-01220-2. Article 57; earlier preprint arXiv:2402.12506.

Paper details

vlnc:W_wttu4qk9
Version
v1 · latest
Published
License
LicenseRef-OpenAI-public
Usage

views · 22

downloads · 4

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

unreleased internal OpenAI model

View statement
Looking for related papers…
Contents