{"id":"6e513111-1f28-4b41-bf32-dfa5856a619a","arxiv_id":"1908.06026","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Every closed subset of the plane containing 0 and at least one other point is realized as the singular-value set of a locally univalent entire function made from nested exponentials.","lead":"A new class of entire functions built from nested exponentials is shown to contain locally univalent functions whose singular-value set is any prescribed closed set containing zero and another point. The paper supplies a new constructive proof of a 1954 theorem of Heins and a new tool for building entire functions with controlled dynamics.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The written proof of Theorem 1 falsely asserts a common inverse-branch neighbourhood for every point outside the orbit; restricting the argument to the complement of the orbit closure repairs it and preserves the theorem.","rationale":"The central claim of Theorem 1 depends on showing that F0 has no singular values outside V. That is exactly the role of the inverse-branch argument in Section 3. The paper's choice of Ω as the complement of the orbit is too broad: for points in V that are accumulation points of the orbit, no common inverse-branch neighbourhood can exist because the finite singular sets of the approximants accumulate there. This is an internal overstatement rather than a disagreement with consensus, and it is load-bearing because it threatens the proof of S(F0) ⊆ V. The repair is straightforward: since V is closed, points outside V have positive distance from the orbit, so the common neighbourhood exists there; with Ω replaced by the complement of V, Lemma 7 and the uniform convergence estimates apply unchanged. I therefore agree with the reader's conditional verdict and recommend no change to it. I also note a minor typographical issue in equation (5): the intended definition appears to be b_n + z/(b_0...b_{n-1}) rather than (b_n + z)/(b_0...b_{n-1}); the printed form makes f_{k,n}(0) = b_k and f'_{0,n}(0) = 1 fail, though the inverse-branch formulas later in Section 3 confirm the corrected normalization. This does not alter the main concern.","tokens_in":12842,"tokens_out":33581,"duration_ms":309877,"concrete_test":"Specialize to V = {0} ∪ {1/n : n ≥ 1} and choose a dense sequence in V with a_0 = 0. Let ζ = 0. Every neighbourhood of ζ contains some orbit point, so no fixed U_ζ avoids the singular values of all f_{0,n}, disproving the paper's assertion for this ζ. To test the repair, rewrite the proof with Ω equal to the complement of V and verify that Lemma 7 and the inequalities (15)–(16) are used only for points outside V, where a fixed U_ζ exists because V is closed. If they apply verbatim, the stated inclusion S(F0) ⊆ V is restored without changing the construction.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Section 3, after defining Ω as the complement of the orbit {E(0,k)(0): k ≥ 0}, the paper states that every ζ ∈ Ω has a neighbourhood U_ζ on which all inverse branches of every approximant f_{0,n} are defined for all n > 0. This is false when ζ is an accumulation point of the orbit but is not itself an orbit point: every neighbourhood of ζ contains some orbit point, and once n exceeds the index of that point, f_{0,n} has a singular value inside that neighbourhood. Thus no fixed U_ζ can avoid the excluded finite sets for all n. The subsequent convergence argument (Lemma 7 and the estimates leading to (15)–(16)) therefore only works on the complement of the orbit closure, not on all of Ω. This is load-bearing because the stated overclaim would imply S(F0) is contained in the orbit itself, contradicting the already-established inclusion V ⊆ S(F0). The gap is repairable: since V is closed, for ζ outside V there is positive distance to every orbit point, so a common U_ζ exists there; replacing Ω by the complement of V makes the proof go through unchanged. As written, however, the proof contains an unsupported and false intermediate claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a class E of entire functions F0 for which there exist entire Fn and nonzero constants lambda_n satisfying Fn = lambda_{n+1} exp(F_{n+1}) for all n. The main result (Theorem 1) states that for every closed set U containing 0 and at least one further point, and for every compact K and epsilon>0, there is a locally univalent f in E with S(f)=U and ||f-E_1||_K<epsilon. The proof first constructs, for any prescribed nonzero lambda_n, a uniform limit F0 of explicit compositions of exponentials, and then chooses the lambda_n so that the points E_{(0,n)}(0) form a dense subset of U. A separate inverse-branch argument is used to show that no other singular values appear. From Theorem 1 the author derives Corollary 2 (Heins' theorem that every closed set is the set of singular values of a locally univalent entire function), further closure properties of E, and two examples, one asserting the existence of functions in E with nonempty Fatou set and one asserting the existence of functions with empty Fatou set.","tokens_in":13122,"tokens_out":19768,"duration_ms":167708,"significance":"If the proof is repaired, this is a valuable and original contribution. The construction is explicit and self-contained, avoiding uniformization and quasiconformal folding; the functions are obtained as uniform limits of concrete approximants, and the paper shows that despite the known instability of the singular-value set under limits, the particular limiting process can be used to prescribe S(f) exactly. This gives a new proof of a classical theorem of Heins. The paper also correctly identifies useful structural properties of the class E, such as closure under composition and dense embedding into the space of non-vanishing entire functions, and the proof is based on standard tools (Picard, Hurwitz, Rouche) with no fitted parameters. The advertised dynamical consequences (Corollaries 8 and 9 and the Fatou-set examples) broaden the interest of the paper, provided Example 3 is fully justified.","major_comments":[{"comment":"The assertion after the definition of Omega, that every zeta in Omega has a neighbourhood U_zeta on which all inverse branches of f_{0,n} are well defined for all n>0, is false. If zeta is an accumulation point of the orbit {E_{(0,k)}(0)} that is not itself an orbit point, then zeta lies in Omega, but every neighbourhood of zeta contains some E_{(0,ell)}(0), and for all n>ell this point is a singular value of f_{0,n}; hence no inverse branch of f_{0,n} can be defined on the whole neighbourhood. This false claim is load-bearing as written, because Lemma 7 and the estimates leading to (15) and (16) are applied to all of Omega, and the proof concludes S(F0) subset of the orbit closure. The gap is repairable: replace Omega by C\\V. Since V is closed, every zeta outside V has positive distance to V and therefore to every orbit point, so a fixed U_zeta avoiding all excluded finite sets exists; the remaining argument then proves S(F0) subset of V, which is exactly what is needed. The repair is local, but as it stands the proof contains an unsupported intermediate claim.","section":"Section 3, paragraph after defining Omega"},{"comment":"The statement 'hence the Fatou set of f is empty' is not justified. From |f'(1)|>1 one only obtains that 1 is a repelling fixed point. To conclude that F(f)=emptyset one must rule out attracting, parabolic, and Siegel periodic Fatou components as well as wandering and Baker domains. The finiteness of S(f) and the postsingular set {0,1} make this plausible and probably standard, but the argument is not written and no citation is given. Please either supply a short proof or cite a theorem that, for an entire function with finite postsingular set all of whose periodic points are repelling, the Fatou set is empty.","section":"Section 4, Example 3"}],"minor_comments":[{"comment":"The phrase 'and its is dense' should read 'and it is dense'.","section":"Abstract"},{"comment":"The sentence 'Let (r_n) be an increasing sequence of integers' should say 'real numbers' or 'positive numbers', since the radii of the disks need not be integers.","section":"Section 2, proof of Theorem 3"},{"comment":"The word 'Pioncaré' is a typo for 'Poincaré'.","section":"Section 2, alternative proof of Proposition 6"},{"comment":"Please spell out the choice of epsilon in the Rouché argument and the Cauchy estimate for the derivative; the sentence 'using Cauchy estimates we obtain |f'(1)|>1' is too terse for the claimed quantitative conclusion.","section":"Section 4, Example 3"},{"comment":"Reference [5] contains the garbled phrase 'Konvergenzwert st'; the intended German term should be corrected.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The two flagged issues are local and repairable: replacing Omega by C\\V in Section 3 fixes the main proof, and Example 3 needs a short justification or a precise citation. I do not see a fundamental obstruction to the central theorem, but the manuscript should not be accepted in its current form because the main proof contains a false intermediate claim and one advertised consequence is unsupported."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The main thing you should know: this paper is real, not a curiosity. It introduces a class E of entire functions (those expressible as F0 with F_n = λ_{n+1}e^{F_{n+1}}), shows E is dense in the non-vanishing entire functions, and proves that every closed set containing 0 and at least one other point is the singular-value set of some locally univalent function in E. That implies Heins's theorem as a corollary, but with a genuinely different proof mechanism: explicit towers of exponentials whose inverse branches can be tracked.\n\nThe best part is the proof of Theorem 1. The author constructs F0 as a uniform limit of f_{0,n}, where each f_{0,n} is in the Speiser class with singular values exactly the first n points of a chosen orbit. The inverse-branch argument in Lemma 7 is clever: it forces the winding indices to stabilize, and the estimates (15)–(16) show the inverse branches converge. The claim S(F0) = V follows cleanly once you have convergence of all inverse branches outside V. This is a real advance over Heins's uniformization construction, and the recovery of Heins is not a trivial consequence.\n\nThe soft spot is in Section 3, as you flagged. The paper defines Ω as the complement of the orbit and asserts every ζ ∈ Ω has a fixed neighbourhood U_ζ on which all inverse branches of every f_{0,n} are defined for all n. That is false when ζ is an accumulation point of the orbit: any neighbourhood of ζ contains orbit points that are singular values of f_{0,n} for all large n. The fix is easy: replace Ω with the complement of V (the orbit closure). For ζ outside V, positive distance to the orbit gives the needed neighbourhood, and that is all the proof needs to get S(F0) ⊆ V. So the theorem survives, but the written proof contains an overstatement that must be corrected.\n\nThe second issue is minor: Example 3 claims the Fatou set is empty from finite postsingular set plus |f'(1)|>1. That doesn't follow without an argument (or a citation), and the text just asserts it. It does not affect the main theorem, but a referee should ask for a justification.\n\nNet: this deserves a serious referee. The construction is new, the main theorem is correct in substance, and the gaps are repairable. I would send it to review and ask for a revision that fixes the neighbourhood statement and fleshes out Example 3.","headline":"A new construction class for entire functions with prescribed singular values; the central theorem is sound modulo a repairable overstatement in Section 3 and a minor gap in Example 3.","tokens_in":13615,"tokens_out":3489,"would_cite":true,"duration_ms":29078,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["30D20","30D05","37F10"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that every closed set containing 0 and one more point is the singular-value set of a locally univalent entire function in a new explicit class, and that this class is dense in the non-vanishing entire functions.","keywords":["entire functions","singular values","asymptotic values","locally univalent functions","class E","iterated exponentials","Eremenko-Lyubich class","Fatou set"],"falsifier":"Take $U=\\{0,1\\}$ and choose the constants $\\lambda_n$ so that the iterated points $E_{(0,n)}(0)$ accumulate at $1$; then inspect the inverse branches $g_{I_n}$ of the explicit approximants $f_{0,n}$ near $1$. If no sequence of these branches converges uniformly on a fixed neighbourhood to a branch of the limiting $F_0$, then $1$ is not an asymptotic value and $S(F_0)$ misses $1$, contradicting Theorem 1 for this choice of $U$.","tokens_in":12662,"feed_emoji":"🔁","tokens_out":22296,"duration_ms":199739,"temperature":0.7,"pith_summary":"This paper introduces a class $\\mathcal{E}$ of entire functions (complex-differentiable maps defined on all of $\\mathbb{C}$) generated by infinite backward compositions of rescaled exponentials, meaning $F_0\\in\\mathcal{E}$ when $F_n(z)=\\lambda_{n+1}e^{F_{n+1}(z)}$ for a sequence of nonzero constants $\\lambda_n$. Its main theorem states that every closed set $U\\subset\\mathbb{C}$ containing $0$ and at least one other point is the singular-value set $S(f)$ of a locally univalent entire function $f\\in\\mathcal{E}$, and that $f$ can be chosen uniformly close to the exponential map on any compact set. Because adding a constant translates the singular-value set, the theorem recovers the classical result that every closed subset of the plane occurs as the singular-value set of a locally univalent entire function. The point is methodological: the functions are explicit uniform limits of approximants with known singular values, so the construction tracks a quantity that is normally unstable under uniform convergence.","feed_headline":"Iterated exponentials realize every closed singular-value set","feed_subtitle":"An explicit class of approximating functions keeps singular values stable through uniform limits.","key_machinery":"The mechanism has two parts. First, the class $\\mathcal{E}$ is generated by the identities $F_n(z)=\\lambda_{n+1}e^{F_{n+1}(z)}$, giving the telescoping relation $F_0=E_{(0,n)}\\circ F_n$; this makes every orbit point $E_{(0,k)}(0)$ an asymptotic value of $F_0$. Second, the explicit approximants $f_{k,n}$ and their inverse branches $g_{I_n}(z)=(L_{I_n}(z)-b_n)b_0\\cdots b_{n-1}$, where $L_{I_n}$ is a composition of logarithmic branches $L^\\lambda_k(z)=\\log z-\\log\\lambda+2k\\pi i$, control which asymptotic values survive the uniform limit. The load-bearing tool is Lemma 7: if $g_{I_n}$ converges uniformly on a neighbourhood of a point $\\zeta$ outside the orbit, then the index tuple eventually has the form $(k_1,\\ldots,k_j,m_{j+1},\\ldots,m_n)$. This forces the limiting inverse branches to be exactly the ones anchored by the first $j$ choices, so a point outside the dense orbit cannot be an asymptotic value of $F_0$.","core_discovery":"The central claim is Theorem 1: given any closed $U\\subset\\mathbb{C}$ containing $0$ and one further point, any compact set $K$, and any $\\varepsilon>0$, there is a locally univalent $f\\in\\mathcal{E}$ with $S(f)=U$ and $\\|f-E_1\\|_K<\\varepsilon$, where $E_1(z)=e^z$. The construction first chooses constants $\\lambda_n$ so that the orbit points $E_{(0,n)}(0)$, with $E_{(0,n)}=E_{\\lambda_1}\\circ\\cdots\\circ E_{\\lambda_n}$ and $E_\\lambda(z)=\\lambda e^z$, are dense in $U$. The approximants are $f_{0,n}(z)=E_{(0,n)}((b_n+z)/(b_0\\cdots b_{n-1}))$, where $b_0=1$ and $b_n=2\\pi i m_n+\\log b_{n-1}-\\log\\lambda_n$ for rapidly increasing integers $m_n$. They converge uniformly on compact sets to a nonconstant locally univalent $F_0\\in\\mathcal{E}$, and each orbit point $E_{(0,k)}(0)$ is an asymptotic value of $F_0$. The proof that no other points are asymptotic values uses inverse branches of $f_{0,n}$: a stabilization lemma forces every convergent sequence of inverse branches to agree eventually with a single branch of $F_0$, so no accidental asymptotic values arise. Hence $S(F_0)$ equals the closure of the dense orbit, namely $U$.","pith_inferences":["The explicit approximants suggest a testable extension: prescribe, along with $S(f)=U$, finitely many local data such as $f(0)$ and $f'(0)$; the paper already constructs functions in $\\mathcal{E}$ with $f(1)=1$ and $f'(1)=\\lambda$, so a systematic interpolation theorem is plausible.","The branch-stabilization mechanism is not obviously tied to exponentials; replacing $E_\\lambda$ by another one-parameter family of covering maps could yield analogous classes with controlled singular values if the constants can force inverse branches to stabilize.","One could study quantitative versions of the theorem, such as how the rate of approximation to $e^z$ on growing compact sets depends on the rapidly increasing integers $m_n$ and on the geometry of $U$; the paper leaves this trade-off unexplored."],"forward_implications":["Every closed set $V\\subset\\mathbb{C}$ is the singular-value set of some locally univalent entire function: the one-point cases are covered by $e^z+c$, the empty case by the identity map, and all other cases by Theorem 1 followed by a constant shift.","For any closed $U$ containing $0$, the exponential functions $E_\\lambda(z)=\\lambda e^z$ lie in the closure of the class $\\{f:S(f)=U\\}$, so prescribed singular-value sets can be realized while staying uniformly close to exponentials.","The class $\\mathcal{E}$ is dense in the space of non-vanishing entire functions, with the zero function also in the closure, giving many explicit uniform limits with controlled singular values.","Functions in $\\mathcal{E}$ can have empty Fatou set with finite singular values $\\{0,1\\}$, so the construction produces examples with completely repelling postsingular dynamics, and it also produces functions with nonempty Fatou set.","The class $\\mathcal{E}$ is closed under composition and under composition with arbitrary entire functions, so it forms a compositional reservoir of explicit maps whose singular values remain controlled, intersecting both the Speiser class and the Eremenko-Lyubich class."],"supporting_citations":[{"why":"the classical result that every closed set is the singular-value set of a locally univalent entire function, which this paper reproves by a uniform-limit construction.","marker":"[6]"},{"why":"defines the Eremenko-Lyubich class of functions with bounded singular values, one of the classes whose intersection with the new class $\\mathcal{E}$ is established.","marker":"[4]"},{"why":"the prior construction with controlled singular values via quasiconformal folding, the main comparison point for the paper's explicit limit method.","marker":"[2]"},{"why":"supplies the functional-equation and local-linearization facts used to exhibit nonconstant functions in $\\mathcal{E}$, including the special solutions of $f(\\mu z)=\\lambda e^{f(z)}$.","marker":"[8]"}],"fun_headline_variants":["Iterated exponentials yield every closed singular-value set","Entire functions from exponentials achieve any prescribed singular set","Exponential towers can prescribe singular values arbitrarily","New class of entire functions realizes all closed singular sets"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof relies on every point excluded from the singular set having a fixed neighbourhood on which all inverse functions of the approximating maps are defined with uniform bounds, including at points where the chosen iterated points accumulate.","fun_headline_variants_meta":{"raw":{"variants":["Iterated exponentials yield every closed singular-value set","Entire functions from exponentials achieve any prescribed singular set","Exponential towers can prescribe singular values arbitrarily","New class of entire functions realizes all closed singular sets"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001288,"raw_usage":{"total_tokens":5344,"prompt_tokens":1109,"completion_tokens":4235,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":725,"completion_tokens_details":{"reasoning_tokens":4173}},"tokens_in":725,"tokens_out":4235,"duration_ms":29173,"temperature":1.0,"reasoning_tokens":4173,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:03:07.084339+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take $U=\\{0,1\\}$ and choose the constants $\\lambda_n$ so that the iterated points $E_{(0,n)}(0)$ accumulate at $1$; then inspect the inverse branches $g_{I_n}$ of the explicit approximants $f_{0,n}$ near $1$. If no sequence of these branches converges uniformly on a fixed neighbourhood to a branch of the limiting $F_0$, then $1$ is not an asymptotic value and $S(F_0)$ misses $1$, contradicting Theorem 1 for this choice of $U$.","supporting_citations":[{"cited_title":"Heins: The set of asymptotic values of an entire function , Tolfte Skandinaviska Matematik- erkongressen (Lund, Sweden, 1953), Proceedings of the Scan dinavian Math","cited_arxiv_id":null,"evidence_quote":"the classical result that every closed set is the singular-value set of a locally univalent entire function, which this paper reproves by a uniform-limit construction."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"defines the Eremenko-Lyubich class of functions with bounded singular values, one of the classes whose intersection with the new class $\\mathcal{E}$ is established."},{"cited_title":"Bishop: Constructing entire functions by quasiconformal folding Acta Math., 214 (2015), 160","cited_arxiv_id":null,"evidence_quote":"the prior construction with controlled singular values via quasiconformal folding, the main comparison point for the paper's explicit limit method."},{"cited_title":"Milnor: Dynamics in one complex variable , 3rd ed","cited_arxiv_id":null,"evidence_quote":"supplies the functional-equation and local-linearization facts used to exhibit nonconstant functions in $\\mathcal{E}$, including the special solutions of $f(\\mu z)=\\lambda e^{f(z)}$."}],"review_version":1}