{"id":"817907d2-8718-487d-9a3e-0c37aade4e1e","arxiv_id":"2504.20284","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"In any degenerating one-parameter family of rational maps, either all periodic multipliers stay uniformly bounded or almost all of them blow up at a power rate, and degenerating cubic families always contain a short repelling cycle with blowing-up multiplier.","lead":"This paper proves a sharp dichotomy for one-parameter families of rational maps that degenerate at a parameter value: either the multipliers of all periodic orbits stay bounded, or the multipliers of nearly all periodic orbits blow up like a positive power of the parameter.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem A's exponent and (1−ε)d^n count depend on the contraction rate in Lemma 1.3 and the λ=0 equivalence, both imported without proof from unpublished [FavRL24].","rationale":"The reader's weakest_assumption correctly identifies the imported non-Archimedean results as the most load-bearing unverified input. My re-reading of §2.1 confirms that Theorem A(2) depends pointwise on the contraction rate in Lemma 1.3, and Case (1) depends on Theorem 1.1 plus [FavRL24, Théorème 4.4]. I found no internal inconsistency in the complex-side argument strong enough to replace this concern: the Galois-orbit argument in §2.1 supplies the uniformity needed to pass from component-wise non-Archimedean norms to pointwise complex multiplier lower bounds, and the 'reducing A+' step is a routine constant adjustment for each fixed n. The broken cross-reference and the missing figure are presentation issues, not correctness risks. Therefore the reader's CONDITIONAL verdict is appropriate, with the condition being independent confirmation of the imported non-Archimedean results.","tokens_in":20005,"tokens_out":12696,"duration_ms":131135,"concrete_test":"Independently verify the two imported results in [FavRL24]: (a) Theorem 1.1, that over C((t)) λ(f)=0 holds iff f is affine Bernoulli iff f has no repelling rigid periodic cycle; (b) Lemma 1.3, that the inverse branches of f^n contract B(x0,τ) into B(x_{-n}, L e^{-nλ(f)/2}) for a full-measure set of orbits in the natural extension, with the stated rate and not a merely suboptimal rate e^{-n(λ(f)/2−δ)}. Concretely, re-derive the proof of Theorem 1.2 from [FavRL24, Theorem B] and check the sharpness of the exponential rate; then recompute §2.1 replacing e^{-nλ(f)/2} by e^{-n(λ(f)/2−δ)} and confirm whether Theorem A(2) still holds with λ=λ(f)/2−δ and density 1−ε. If the weaker rate is all that follows, Theorem A(2) must be restated accordingly.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem A in §2.1 splits at λ(f_na)=0 versus >0. In the zero case it invokes [FavRL24, Théorème 4.4] for the uniform multiplier bound; that invocation rests on the equivalence in Theorem 1.1 (λ=0 iff no repelling rigid cycle). In the positive case, Theorem 1.2 is the engine, and its proof is entirely built on Lemma 1.3, imported from [FavRL24]: almost every point in the natural extension admits inverse branches of f^n contracting B(x0,τ) into B(x_{-n}, L e^{-nλ(f)/2}). That rate e^{-nλ(f)/2} is exactly what makes the complex multipliers on good components have pole order at least λ(f)/2, which is then weakened to the arbitrary exponent λ<λ(f)/2 in Theorem A(2). The full-measure union over L,τ is what yields the proportion 1−ε of rigid periodic points with large multiplier, hence the count (1−ε)d^n. If the contraction rate in [FavRL24] were actually e^{-n(λ(f)/2−δ)} for some δ>0, or the full-measure statement failed on a set of positive measure, then the positive-Lyapunov half of Theorem A would only give the weaker exponent λ=λ(f)/2−δ and a degraded count. Since [FavRL24] is an unpublished preprint and the present paper does not reproduce the proof, this is the most load-bearing unverified step.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies one-parameter holomorphic degenerating families \\{f_t\\} of rational maps of degree d≥2 on the Riemann sphere. The main result, Theorem A, asserts a dichotomy: either all multipliers of periodic points satisfy a uniform bound |(f_t^n)'(z)|^{1/n}≤C for |t|≤1/2, or there exists λ>0 such that for every ε>0 and every n, at least (1−ε)d^n of the period-n points have max{1,|(f_t^n)'(z)|}^{1/n} ≥ C(ε,n)|t|^{-λ}. The proof passes to the non-Archimedean limit f_na over C((t)) and uses its Lyapunov exponent λ(f_na), together with a quantitative equidistribution of rigid repelling cycles (Theorem 1.2). The paper also proves Theorem B (existence of repelling rigid cycles of period ≤2 for polynomials and ≤3 for cubic rational maps, with fixed-point statements for quadratic and cubic-polynomial cases) and Corollary C (properness and birationality of certain multiplier maps on moduli spaces). The non-Archimedean results for polynomials (Theorem 3.1) and cubic maps (Theorem 3.2) are proved within the paper, and several known results are used and contextualized.","tokens_in":20314,"tokens_out":15246,"duration_ms":140842,"significance":"If correct, Theorem A is a substantial advance: it upgrades the positive-proportion blow-up in [FavRL24] to a (1−ε) proportion and gives a quantitative exponent, showing that multiplier blow-up is typical in degenerating families. The proof strategy via the non-Archimedean limit and the Lyapunov exponent is elegant and well-motivated, and the paper is clearly written. The self-contained portions — in particular the counting proof for polynomial cycles (Theorem 3.1) and the case analysis for cubic maps (Theorem 3.2) — are valuable. However, the central quantitative engine (Theorem 1.2) rests on Lemma 1.3, which is imported from the unpublished preprint [FavRL24] without proof, and the zero-Lyapunov alternative of Theorem A also relies on [FavRL24, Théorème 4.4] and Theorem 1.1 of that preprint. In addition, the proof of Theorem 1.2 contains a shift-index error in the mixing argument. These issues must be addressed before the main theorem can be fully verified.","major_comments":[{"comment":"The proof of Theorem 1.2 and hence of the blow-up alternative in Theorem A(2) is entirely dependent on Lemma 1.3, which is stated without proof and attributed to the unpublished preprint [FavRL24]. The exact contraction rate e^{-nλ(f)/2} and the full-measure union over L,τ in Lemma 1.3 are what force the exponent λ<λ(fna)/2 and the proportion (1−ε)d^n in Theorem A(2); any weakening of the rate (e.g., e^{-n(λ/2−δ)}) or a failure of the full-measure property would degrade both the exponent and the count. Because [FavRL24] is not part of the present manuscript, the central quantitative claim is not self-contained and cannot be checked from the text as it stands.","section":"§1.3, Lemma 1.3"},{"comment":"The first alternative of Theorem A is obtained by invoking [FavRL24, Théorème 4.4], which itself relies on the equivalence in Theorem 1.1 (λ=0 iff affine Bernoulli iff no repelling rigid cycle). These results are also imported from the unpublished preprint [FavRL24] and are not proved or stated in sufficient detail in the present paper. The dichotomy is therefore conditional on external results; the author should either include proofs of the relevant statements or state Theorem A with an explicit dependence on [FavRL24].","section":"§2.1, first paragraph"},{"comment":"The mixing argument uses the set σ^n(\\hat B(x)) ∩ \\hat B(x) ∩ E_{L,τ}, but Lemma 1.3 provides an inverse branch from y_0 to y_{-n}, and the final ball B(y_{-n}, L e^{-nλ/2}) must be contained in B(x,r). This requires y_{-n}∈B(x,r), i.e., \\hat y ∈ σ^{-n}(\\hat B(x)). The set σ^n(\\hat B(x)) only gives y_n∈B(x,r), which does not control y_{-n}. As written, the covering argument does not go through. If the intended set is σ^{-n}(\\hat B(x)), the proof should be corrected; this is load-bearing because it produces the counting estimate (1−ε)d^n in Theorem 1.2.","section":"§1.3, proof of Theorem 1.2"}],"minor_comments":[{"comment":"The theorem states 'rigid periodic points z of period n', but the proof counts periodic points of period dividing n. The difference is O(d^{n/2}) and does not affect the asymptotic statement, but the wording should be aligned.","section":"§1.3, Theorem 1.2"},{"comment":"The sentence 'Reducing A_+ if necessary, we may suppose this is true over |t|≤1/2' is potentially confusing; since the constant is allowed to depend on n and ε, it may help to spell out that one decreases C to handle the range r_n≤|t|≤1/2.","section":"§2.1, final paragraph"},{"comment":"Typos such as 'holomomorphic' (§1.4), 'We dont know' (Introduction), and 'mulitpliers' (Introduction) should be corrected.","section":"Throughout"},{"comment":"The proof of Theorem 3.2 is long and contains many cases; the reference to Figure 1 is helpful but the figure itself is not described in the caption, making it hard to follow. A more detailed caption or a short overview of the cases would improve readability.","section":"§3.2"}],"recommendation":"major_revision","confidential_remarks":"The manuscript relies heavily on the author's own unpublished preprint [FavRL24] for the two most load-bearing statements (Lemma 1.3 and the zero-Lyapunov bound). This makes independent verification difficult; the editor may wish to request a copy or assurance of publication. The shift-index error in the proof of Theorem 1.2 is easy to fix but shows the need for a careful revision of §1.3."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, this is a real improvement: Theorem A turns the earlier positive-proportion blow-up result into a dichotomy where in the non-bounded case almost all period-n points — (1−ε)d^n of them — have multipliers blowing up like |t|^{-λ}. Second, the proof leans heavily on the author's unpublished preprint [FavRL24]: the contraction lemma (Lemma 1.3) and the λ=0 equivalence (Theorem 1.1) are both imported without proof, and the zero-Lyapunov case uses a theorem from that preprint. If those tools are solid, the paper is likely correct; if not, Theorem A's exponent and count degrade.\n\nWhat is genuinely new: Theorem A(2) with the explicit (1−ε) proportion and power rate; Theorem 3.2, a first classification step for cubic rational maps over C((t)) without potential good reduction, yielding a repelling rigid cycle of period at most 3; and Corollary C(2), properness of the period-1,2,3 multiplier map on M3. The architecture is coherent: reduce to the non-Archimedean limit, use the Lyapunov dichotomy, and bridge back to complex multipliers via Lemma 1.7. The polynomial counting argument in §3.1 is clean and self-contained.\n\nSoft spots, in proportion. The central quantitative engine, Theorem 1.2, is built entirely on Lemma 1.3, which is stated as 'proved in [FavRL24]' and not reproduced. The stress-test concern is legitimate: if the contraction rate is e^{-n(λ/2−δ)} rather than e^{-nλ/2}, or if the full-measure set is smaller, the exponent and the (1−ε) count in Theorem A(2) would not follow. The same holds for the λ=0 equivalence in Theorem 1.1. Since [FavRL24] is unpublished, a referee cannot verify this without access to the preprint; this is the main barrier to accepting Theorem A at face value. The proof of Theorem 3.2 is a long case analysis with a missing figure; it looks plausible but needs a non-Archimedean dynamics expert to certify. Minor issues: a broken cross-reference (§??) and a figure header without the figure.\n\nMy own verdict: the central argument holds up conditional on the imported results; no circularity that I can see. The paper deserves a serious referee. The recommendation: send to peer review, and have the referee require either a proof of Lemma 1.3 and Theorem 1.1 in an appendix or a precise pointer to a publicly available version of [FavRL24], and specifically check the cubic classification.","headline":"Strong dichotomy result upgrading Favre–Rivera-Letelier from positive proportion to (1−ε) proportion, but the proof leans on an unpublished preprint and the cubic case needs expert checking.","tokens_in":20876,"tokens_out":3156,"would_cite":true,"duration_ms":29760,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["37F10","37F45","37P05"],"pacs":[],"model":"deepseek-v4-flash","headline":"For any degenerating family of rational maps, either all periodic multipliers are uniformly bounded or, for every ε>0, at least (1−ε)d^n of the period-n points have multipliers growing like |t|^{−λ}.","keywords":["rational maps","multipliers","periodic cycles","degenerating families","non-Archimedean dynamics","Berkovich space","Lyapunov exponent","moduli space"],"falsifier":"Compute the multipliers of all period-1, period-2, and period-3 cycles in an explicit degenerating cubic rational family and check whether at least one has a pole at t=0; if none does, Theorem B fails. For Theorem A, take an explicit degenerating family whose non-Archimedean limit has positive Lyapunov exponent, count the period-n solutions of f_t^n(z)=z for small |t|, and verify that the proportion with max{1,|df_t^n(z)|}^{1/n} ≥ C|t|^{−λ} tends to 1; finding a proportion bounded away from 1 for every λ>0 would refute the dichotomy.","tokens_in":19775,"feed_emoji":"♾️","tokens_out":8732,"duration_ms":86990,"temperature":0.7,"pith_summary":"This paper proves a dichotomy for one-parameter holomorphic families of rational maps that degenerate at t=0. Along such a family, either the multipliers of all periodic cycles are bounded by a constant independent of the period and the parameter, or there is a rate λ>0 such that for every ε>0 and every n, at least (1−ε)d^n of the period-n points have multiplier at least C(ε,n)|t|^{−λ} as t→0. In other words, if multipliers blow up at all, they blow up for almost every periodic orbit, and the blow-up is at least polynomial in 1/|t|. The proof passes through the non-Archimedean limit of the family over the field of Laurent series, where the dichotomy becomes a statement about the Lyapunov exponent of the limit map: zero exponent gives bounded multipliers, positive exponent forces many repelling cycles with large non-Archimedean multipliers. The paper also shows that small periods suffice to detect degeneracy: in polynomial families a period-1 or period-2 multiplier must blow up, and in cubic rational families a period-3 multiplier must blow up.","feed_headline":"Most periodic multipliers blow up, or none do, as maps degenerate","feed_subtitle":"A dichotomy proved through a non-Archimedean limit shows intermediate multiplier growth is impossible.","key_machinery":"The load-bearing object is the non-Archimedean limit f_na, a rational map of degree d over the complete valued field C((t)), viewed on the Berkovich projective line. Its equilibrium measure carries a Lyapunov exponent λ(f_na)=∫log|df_na| dμ, and the dichotomy is carried by the equivalence that λ(f_na)=0 iff f_na is affine Bernoulli iff f_na has no repelling rigid periodic cycle. The second ingredient is a contraction lemma in the natural extension: almost every point admits analytic inverse branches of f_na^n shrinking at rate $Le^{{−nλ(f_na)/2}}$, which lets the proof count rigid periodic points with large multiplier. A Puiseux-series comparison lemma then converts the non-Archimedean growth into the complex multiplier lower bound |t|^{−λ}, giving the precise rate of blow-up.","core_discovery":"The central discovery is Theorem A: a degenerating family of rational maps of degree d≥2 over the unit disk falls into one of two exclusive cases. Either there is a constant C such that |df_t^n(z)|^{1/n}≤C for all |t|≤1/2, all n, and all period-n points z; or there is λ>0 such that for every ε>0 and every n, at least (1−ε)d^n of the period-n points satisfy max{1,|df_t^n(z)|}^{1/n}≥C(ε,n)|t|^{−λ}. This upgrades an earlier result that only a positive proportion of multipliers blow up in the second case. The proof identifies the rate λ with any number below half the non-Archimedean Lyapunov exponent λ(f_na) of the limit map f_na over the field C((t)). When λ(f_na)>0, a contraction lemma for inverse branches in the natural extension yields at least (1−ε)d^n rigid periodic points of period n with non-Archimedean multiplier at least A $e^{{nλ(f_na)/2}}$; a Puiseux-series comparison transfers this into the complex lower bound |t|^{−λ}. When λ(f_na)=0, the limit is affine Bernoulli and all complex multipliers stay bounded. This mechanism makes blow-up generic rather than exceptional.","pith_inferences":["The dichotomy suggests a general principle for degenerating rational maps: the parameter-space boundary is governed by large multipliers, so compactness statements for moduli spaces should follow from controlling small-period multipliers.","One might expect a quantitative refinement: the optimal blow-up exponent should be exactly λ(f_na)/2, with the proportion (1−ε)d^n replaced by an error term depending on the mixing rate of the non-Archimedean equilibrium measure.","If the paper's conjecture on rigid repelling cycles of bounded period holds in every degree, then the multiplier map on moduli space would be proper in all degrees, giving a uniform finite-type description of degeneracy.","The affine Bernoulli exceptional case, where all multipliers stay bounded, suggests a dictionary between non-Archimedean entropy and the asymptotic growth of complex multiplier spectra that could be tested numerically on families such as z^d + t z^{−d}."],"forward_implications":["In any degenerating family, either no multiplier grows at all or a fraction tending to 1 of period-n multipliers grows at least polynomially in 1/|t|, so intermediate behavior with only a positive proportion blowing up is impossible.","For polynomial families of any degree, degeneracy forces a period-1 or period-2 multiplier to have a pole at t=0; for cubic rational families, period at most 3 suffices.","Consequently, the multiplier spectrum map on the relevant moduli spaces is proper and birational: the boundary at infinity is detected by small-period multipliers.","In stable polynomial families whose non-Archimedean limit has positive Lyapunov exponent and no recurrent critical points, the blow-up rate |t|^{−λ} holds uniformly for all large n with a constant independent of n.","The rate of blow-up is governed by the non-Archimedean Lyapunov exponent: any λ below λ(f_na)/2 can be chosen, linking the complex multiplier growth to the expansion rate of the limit map."],"supporting_citations":[{"why":"Supplies the contraction lemma for inverse branches in the natural extension, the equivalence between zero Lyapunov exponent and absence of repelling rigid cycles, and the earlier positive-proportion blow-up result that Theorem A refines.","marker":"[FavRL24]"},{"why":"Proves the non-negativity of the non-Archimedean Lyapunov exponent, which starts the λ=0 versus λ>0 dichotomy.","marker":"[Ok15]"},{"why":"Provides the Berkovich projective line, equilibrium measures, and the theorem on local degrees used in the cubic periodic-cycle analysis.","marker":"[BR10]"},{"why":"Supplies background on Berkovich dynamics, type II points, and Rivera-Letelier's theorem on fixed type II points used in Section 3.","marker":"[Ben19]"},{"why":"Gives the ergodic theory of rational maps over ultrametric fields, including the mixing of the equilibrium measure used in the counting argument.","marker":"[FavRL10]"},{"why":"Relates potential good reduction of the non-Archimedean limit to non-degeneracy of the complex family through local degree at the Gauss point.","marker":"[Ki15]"},{"why":"Shows the quadratic moduli space embeds properly via fixed-point multipliers, used in Theorem B(1) and the properness consequences.","marker":"[Mi93]"},{"why":"Controls the intersections of periodic-point curve components with t=0 when there are no recurrent critical points, used in the uniform polynomial version.","marker":"[Tr14]"}],"fun_headline_variants":["Multiplier blow-up dichotomy for degenerating rational maps","Almost all multipliers blow up, or none do, in degenerating maps","Non-Archimedean proof yields multiplier blow-up dichotomy","Degenerating rational maps: all multipliers bounded or almost all blow up"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The dichotomy rests on the imported contraction lemma that almost every point in the natural extension of the non-Archimedean limit admits inverse branches shrinking at rate $Le^{{−nλ(f_na)/2}}$, together with the equivalence between zero Lyapunov exponent, having no repelling rigid cycle, and having all complex multipliers bounded; should either estimate fail in the stated strength, the rate |t|^{−λ} and the (1−ε)d^n count would not follow from the proof.","fun_headline_variants_meta":{"raw":{"variants":["Multiplier blow-up dichotomy for degenerating rational maps","Almost all multipliers blow up, or none do, in degenerating maps","Non-Archimedean proof yields multiplier blow-up dichotomy","Degenerating rational maps: all multipliers bounded or almost all blow up"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000502,"raw_usage":{"total_tokens":2393,"prompt_tokens":826,"completion_tokens":1567,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":442,"completion_tokens_details":{"reasoning_tokens":1494}},"tokens_in":442,"tokens_out":1567,"duration_ms":12505,"temperature":1.0,"reasoning_tokens":1494,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T05:34:07.637346+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the multipliers of all period-1, period-2, and period-3 cycles in an explicit degenerating cubic rational family and check whether at least one has a pole at t=0; if none does, Theorem B fails. For Theorem A, take an explicit degenerating family whose non-Archimedean limit has positive Lyapunov exponent, count the period-n solutions of f_t^n(z)=z for small |t|, and verify that the proportion with max{1,|df_t^n(z)|}^{1/n} ≥ C|t|^{−λ} tends to 1; finding a proportion bounded away from 1 for every λ>0 would refute the dichotomy.","supporting_citations":[],"review_version":1}