{"id":"c93b9a3d-3ebe-4d8b-b104-aa30206cb226","arxiv_id":"1908.03208","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper defines and exactly computes threshold parameters il(p) and cn(p) governing circle-rootedness and interlacing with roots of unity, and constructs families where il/cn grows without bound.","lead":"This paper gives exact formulas for two thresholds attached to symmetric polynomials: one that forces roots onto the unit circle, and one that forces them to interlace the roots of unity. It classifies the second threshold by a polyhedral fan and shows that the two thresholds can be arbitrarily far apart as the degree grows.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 7.8 uses an incorrect root set for Q_n^2, so the proof that il/cn is unbounded (Theorem 7.7) is invalid as written.","rationale":"The reader identifies Proposition 6.6 as the weakest assumption, but that issue appears repairable: once S_ω(p_β) and S_ω(x^n+1) have a common real interlace C, sign preservation shows C is also a common interlace of S_ω(p_β)+tS_ω(x^n+1) and S_ω(x^n+1) for all t≥0, so the 'for every α≥β' version follows from the proof's construction. The more serious problem is Proposition 7.8. The root sets asserted there are not just mislabelled; they are inconsistent with the definition of Q_n and with x^n+1, and the claimed interlacing already fails at n=4. Proposition 7.8 is the only proof offered for Theorem 7.7, so the paper's claim that i l/c n is unbounded is unproven as written. This does not overturn the rest of the paper: the Interlace Formula, the FOIC, the double-root discriminant formula, and the small-darga computations may all stand. Thus the reader's CONDITIONAL verdict is appropriate, though the specific repair needed is in §7 rather than mainly in §6. The concrete computational check would settle whether the unboundedness theorem can be salvaged with a corrected proof or whether it fails.","tokens_in":46577,"tokens_out":26972,"duration_ms":272958,"concrete_test":"Recompute c n(P_n) directly from Theorem 6.9 for n=4, 8, 12: compute Disc(p_α/gcd(p,x^n+1)) symbolically for n=4 and with high-precision root finding for n=8,12, and compare with i l(P_n) from the Interlace Formula. Also verify the interlacing claim by listing the arguments of the roots of Q_n^2 and x^n+1 for n=8; the sets should not alternate if the root-set statement in Proposition 7.8 is wrong. If c n(P_n)≤1 and i l(P_n) grows as claimed, Theorem 7.7 survives with a repaired proof; if c n(P_n) exceeds 1 on some subsequence, or if the corrected i l bound fails, the unboundedness result is in jeopardy.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The unboundedness result depends on Proposition 7.8, whose proof contains a concrete false statement. With n=4m, θ=π/(2n), ω=e^{iθ}, F_k=(x−ω^{2k})(x−ω^{−2k}), and Q_n=∏_{0≤j<m}F_{4j+1}, the paper says that the roots of x^n+1 are ω^2,ω^4,…,ω^{2n} and that those of Q_n^2 are ω^2,ω^6,…,ω^{2n−2}, all double. But (ω^{2r})^n=(−1)^r, so the roots of x^n+1 are the ω^{2r} with r odd; and F_{4j+1} has roots at exponents ±2(4j+1), not the listed even exponents. For n=4, Q_4^2=(x^2−√2x+1)^2 has roots only at e^{±iπ/4}, so Q_4^2 does not alternate with all roots of x^4+1; the asserted interlacing, and hence the claim c n(P_4)=1 via Corollary 6.8, is unsupported. Direct computation from p_α(x)=α(x^4+1)−2√2x^3+4x^2−2√2x shows the largest double-root parameter of p_α/gcd(p,x^4+1) is 2√2−2<1, not 1. Since the lower bound b e(P_n)=Ω(n^{1−ε}) is the engine of Theorem 7.7, the headline unboundedness result is not established by this argument; a corrected interlacing argument or a different upper bound for c n(P_n) is needed.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces two threshold invariants for a trim self-inversive polynomial p of darga n: the interlace number il(p), the least parameter such that p_α = α(x^n+1)+p strictly angle-interlaces the nth roots of unity, and the circle number cn(p), the least parameter such that p_α is circle-rooted for all larger α. The main results are the Interlace Formula (Theorems 4.2 and 4.3), expressing il(p) as half the maximum of −p on the relevant roots of unity; a discriminant characterization of cn(p) (Theorems 6.7 and 6.9); a polyhedral-fan classification of palindromic polynomials by their interlacing certificates (Section 5); and a study of exact polynomials where il = cn (Section 7). The paper also claims that the quotient il/cn is unbounded as the degree grows (Theorem 7.7) and includes small-degree computations plus several families of examples.","tokens_in":46905,"tokens_out":10288,"duration_ms":105832,"significance":"If the main theorems are correct, the framework is a genuinely useful organizing device: the Interlace Formula reduces interlacing with roots of unity to a finite Fourier transform, the discriminant formula gives an explicit algebraic algorithm for the circle number, and the fan of interlace certificates provides a clean geometric classification. The interlace formula and the small-darga computations are carefully argued and appear reproducible, with no fitted parameters. The discriminant characterization and the exact-polynomial criteria are attractive and likely to be cited. However, the paper's headline unboundedness claim depends on Proposition 7.8, whose proof contains a concrete false statement about the roots of x^n+1 and Q_n^2; as written, Theorem 7.7 is not established. In addition, Proposition 6.6 is stated in a stronger form than its proof supports and is contradicted by a degree-4 example, and Theorem 6.13 has sign errors in the displayed candidate values. These issues require substantive repair before the paper can be recommended for publication.","major_comments":[{"comment":"The proof of Proposition 7.8 contains an incorrect description of the roots involved. With n=4m, θ=π/(2n) and ω=e^{iθ}, one has (ω^{2r})^n = e^{irπ} = (−1)^r, so the roots of x^n+1 are the ω^{2r} with r odd, not the full sequence ω^2, ω^4, ..., ω^{2n}. Moreover, F_{4j+1} has roots at x = ω^{±2(4j+1)}, whose exponents are congruent to 2 and 4m−2−8j modulo n, not the listed even exponents ω^2, ω^6, ..., ω^{2n−2}. For n=4, Q_4^2 has only the two roots e^{±iπ/4}, each double, so P_4 + x^4 + 1 does not angle-interlace x^4+1; consequently the appeal to Corollary 6.8 to obtain cn(P_4)=1 is unsupported. Because the lower bound on il(P_n) is the engine of Theorem 7.7, the unboundedness result is not established by this argument and needs a corrected construction or a genuinely different upper bound for cn(P_n).","section":"§7, Proposition 7.8"},{"comment":"The proposition is stated in a stronger form than the proof establishes. The proof uses Proposition 2.4 and root correspondence to show that p_α is circle-rooted for all α≥β is equivalent to S_ω(p_β) being real rooted and having a common interlace with S_ω(x^n+1), i.e. to a common angle interlace between p_β and x^n+1 at the single parameter β. The printed statement instead requires p_α and x^n+1 to have a common angle interlace for every α≥β. That stronger assertion is false for the degree-4 example p(x)=2x^2 from Section 8.3: at β=cn(p)=1, p_β=(x^2+1)^2 has double roots ±i, which cannot have a common angle interlace with x^4+1. The statement should be weakened to the β-level condition actually used, and the dependence of Corollary 6.8 and the circle-number computation on this assertion should be re-examined.","section":"§6, Proposition 6.6"},{"comment":"The displayed formulas for r1 and r2 contain sign errors. As printed, r1 = p(−1)/2 and r2 = p(−1)/2 (n even) or p′(−1)/n (n odd), but the proof identifies the candidate parameters as −p(1)/2, −p(−1)/2, and −p′(−1)/n, respectively. The computations in Section 8.3 use the negative forms (for example r1=−b−1 and r2=b−1 for darga 4), so the theorem as stated cannot be applied directly. The statement must be corrected; otherwise the algorithmic content of the circle-number computation is misstated.","section":"§6, Theorem 6.13"}],"minor_comments":[{"comment":"The displayed formula i l(B_n) = 2^{n−1} cos^n(π/n) − 1 appears to have a sign error: for n=3 it gives −1/2, whereas direct computation from the Interlace Formula gives 3/2. The proof text also omits the constant contribution p(ω) = (1+ω)^n − 2 when applying the Interlace Formula.","section":"§9.5, Eq. (28)"},{"comment":"Reference [29] lists arXiv:1908.00839 in the citation line but ends with the identifier arXiv:1902.04231; please reconcile the announced and actual identifiers.","section":"References"},{"comment":"There are several small typos and spacing artifacts, including 'wich' for 'which' in the introduction and 'Erhart' for 'Ehrhart' in Problem 9; a careful copyedit would improve readability.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The unboundedness of il/cn is a headline claim of the paper, and the current proof of Proposition 7.8 is invalid as written. I would ask for a substantive revision that either repairs the construction or supplies a different family of polynomials, and that reconciles Proposition 6.6 and Theorem 6.13 with their proofs, before a further round of review."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the interlace-number machinery is a real contribution, and most of the paper is carefully done, but the proof of the headline unboundedness result (Theorem 7.7 via Proposition 7.8) contains a concrete false statement, so that result is not established as written.\n\nWhat is new and good: the interlace formula (Theorems 4.2/4.3) turns a coefficient check into a finite Fourier maximization, and the fan of interlace certs gives a clean geometric classification of palindromic polynomials by their certs. The discriminant characterization of the circle number (Theorem 6.9) is useful and computationally practical, and the small-darga analysis in Section 8 is careful and reproducible. Known results (Lakatos-Losonczi, Kwon, Chen, Lee-Yang) are recovered as bounds inside the framework rather than patched on. This is an honest paper, and the authors clearly know the literature.\n\nWhere it gets soft. The reader's report is accurate. Proposition 6.6 claims an equivalence for every alpha >= beta but the proof only gives the statement at beta alone; the printed stronger form is false for the darga-4 polynomials in Section 8.3. Theorem 6.13 has an apparent sign/argument typo, and Example 5.3 contains sign errors; these are minor but need fixing. The loading-bearing problem is Proposition 7.8. In the proof, the roots of x^n+1 are listed as omega^2, omega^4, ..., omega^{2n}, but with n=4m and omega=e^{i pi/(2n)}, the correct roots are the omega^{2r} with r odd. The listed roots of Q_n^2—omega^2, omega^6, ..., omega^{2n-2}, all double—do not match the definition F_{4j+1}, whose roots are at exponents plus or minus 2(4j+1). For n=4, Q_4^2 has roots only at e^{±i pi/4}, which do not alternate with all roots of x^4+1. The claim cn(P_4)=1 via Corollary 6.8 is therefore unsupported; direct computation gives the double-root parameter as 2*sqrt(2)-2 < 1. Since Proposition 7.8 is the engine of Theorem 7.7, the unboundedness of il/cn is not proven by this argument.\n\nWho this is for: anyone working on circle-rooted polynomials, root location criteria, or real-rooted/interlacing methods. Despite the flaw in the unboundedness claim, the framework is valuable: il(p), cn(p), exactness, and the fan give a unifying language that deserves serious consideration. The fix may be local—a corrected interlacing argument for the P_n family, or a different upper bound for cn(P_n)—but the paper should not appear with Proposition 7.8 as it stands.\n\nRecommendation: send to peer review, but require major revision. The core ideas are worth refereeing carefully; the errors are specific and repairable, not signs of a hollow paper.","headline":"The interlace-number framework is genuinely new and mostly careful, but the headline unboundedness result rests on a concrete root-set error in Prop. 7.8 and needs major repair before this is publishable.","tokens_in":47475,"tokens_out":2027,"would_cite":false,"duration_ms":23288,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["12D10","26C10","30C15","11C08","11L03","14P10","52B99"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves two sharp thresholds—one Fourier, one discriminant—control when a self-inversive polynomial's roots land on the unit circle and interlace the roots of unity.","keywords":["self-inversive polynomial","palindromic polynomial","polynomial roots","interlacing","roots of unity","polyhedral fan","discriminant","finite Fourier transform"],"falsifier":"Compute $p_\\alpha$ for the degree-4 palindromic polynomial $p(x)=\\sqrt{2}(x^3+x)+2x^2$ at a value $\\beta$ equal to its circle number, or just above it, and check directly whether $p_\\beta$ together with $x^4+1$ admit a common angle interlacing set; the proposition as printed predicts the interlacing holds for every $\\alpha\\ge\\beta$, so finding one $\\alpha$ in $(\\operatorname{cn}(p),\\operatorname{il}(p))$ whose roots fail to share a common angle interlace with $x^4+1$ refutes the universal form of Proposition 6.6.","tokens_in":46329,"feed_emoji":"⭕","tokens_out":8569,"duration_ms":87331,"temperature":0.7,"pith_summary":"This paper asks when a self-inversive polynomial, one whose coefficients are symmetric under reversal so its roots come in reciprocal-conjugate pairs, has all roots on the unit circle, and when those roots fall one in each angular sector between consecutive roots of unity. The authors attach to each trimmed such polynomial $p$ a one-parameter family $p_\\alpha(x)=\\alpha(x^n+1)+p(x)$, with $\\alpha$ dragging the roots toward the circle, and define two sharp thresholds: the interlace number $\\operatorname{il}(p)$ and the circle number $\\operatorname{cn}(p)$. The main claim is that $\\operatorname{il}(p)$ is exactly half the largest negative value of $p$ evaluated at the $n$th roots of unity, so one finite Fourier transform of the coefficients settles interlacing, and that $\\operatorname{cn}(p)$ is the largest real root of a discriminant built from $p_\\alpha$ divided by its common factor with $x^n+1$. The two thresholds always satisfy $\\operatorname{cn}(p)\\le\\operatorname{il}(p)$, and the paper classifies the exact polynomials where equality holds, while showing the ratio can grow without bound as the degree grows.","feed_headline":"Interlacing roots of unity is now a one-line Fourier check","feed_subtitle":"A second threshold, read from a discriminant, controls when all roots land on the unit circle.","key_machinery":"The central object is the dragged family $p_\\alpha(x)=\\alpha(x^n+1)+p(x)$, whose roots move continuously toward the unit circle as $\\alpha\\to\\infty$. The argument is carried by three devices: the discrete Fourier transform of the coefficient vector, since evaluating $p$ at $\\omega\\in U_n$ is exactly that transform and Theorem 3.7 converts the sign of those evaluations into angle-interlacing with the roots of unity; the Möbius root correspondence, which maps the unit circle minus a point to the real line and turns circle rootedness of a self-inversive polynomial into real rootedness of an associated real polynomial, so that interlacing theorems on the line apply; and the discriminant of the normalized family, whose largest real root locates the first double root and hence the circle number. The interlace certs, the roots of unity attaining the maximum in the interlace formula, index the cones of a complete polyhedral fan in the space of trim palindromic polynomials.","core_discovery":"For a trim self-inversive polynomial $p$ of darga $n$ (darga is the sum of the smallest and largest indices of nonzero coefficients), the paper defines $p_\\alpha=\\alpha(x^n+1)+p(x)$ and proves two exact threshold formulas. The interlace number, the least $\\alpha$ beyond which $p_\\alpha$ strictly angle-interlaces the $n$th roots of unity, equals $\\tfrac12\\max\\{-p(\\omega):\\omega\\in U_n\\}$ (Theorem 4.2); because a polynomial that angle-interlaces a set of $n$ circle points is automatically circle-rooted, this gives a cheap sufficient condition. The circle number, the least $\\alpha$ beyond which $p_\\alpha$ is circle-rooted for every larger parameter, equals the largest real root of $\\operatorname{Disc}\\bigl(p_\\alpha(x)/\\gcd(p(x),x^n+1)\\bigr)$ (Theorem 6.9). Interlacing always implies circle rootedness, so $\\operatorname{cn}(p)\\le\\operatorname{il}(p)$, and the paper calls $p$ exact when equality holds; exact polynomials have an interlace cert, a root of unity attaining the maximum, that is a double root at the threshold. The same toolkit reinterprets known coefficient criteria as upper bounds on $\\operatorname{il}$, organizes real palindromic polynomials of fixed darga into a complete polyhedral fan according to which root of unity witnesses the interlace number, and shows the quotient $\\operatorname{il}(p)/\\operatorname{cn}(p)$ can be arbitrarily large in high degree.","pith_inferences":["A testable extension is that the same minimax Fourier formula might certify interlacing for matrix-valued polynomials or for roots constrained to a finite union of circles, since the threshold is read from finitely many evaluations.","The unbounded quotient suggests that for large degree the practical route to circle-rootedness certificates should go through interlacing-certified subfamilies rather than through coefficient-only conditions.","The fan-of-certs classification suggests a probabilistic corollary: for random palindromic polynomials, the cone containing the polynomial, and hence which root of unity is the cert, could be studied from the geometry of the interlace simplex.","The exact-polynomial condition, equality of the two thresholds with a double root at the threshold, may provide an algebraic certificate of tight interlacing analogous to real-rooted interlacing pairs."],"forward_implications":["For any trim self-inversive polynomial with integer coefficients, the interlace number is an algebraic integer, and verifying interlacing with roots of unity reduces to checking $n$ Fourier evaluations rather than factoring or root-finding.","Known circle-rootedness criteria in the literature become upper bounds on $\\operatorname{il}(p)$, so any family satisfying such a bound is automatically circle-rooted and interlaces $U_n$.","The circle number is semi-algebraic in the coefficients and computable from a single discriminant; the parity simplification in Theorem 6.13 cuts the determinant size roughly in half.","Polynomials whose interlace cert is $1$ or $-1$ are exact, and exactness is decidable by checking whether the threshold polynomial has a double root.","In high degree the ratio $\\operatorname{il}/\\operatorname{cn}$ is unbounded, so no coefficient-only condition can approximate the circle number uniformly."],"supporting_citations":[{"why":"Supplies the Fourier-evaluation idea behind Theorem 3.7, from which the interlace formula follows.","marker":"[14]"},{"why":"Provides the interlacing background and the common-interlace criterion used in Propositions 2.3 and 2.4 for the circle-number characterization.","marker":"[16]"},{"why":"Explains the Möbius transformation that carries roots on the circle to real roots, the dictionary used in Section 6.","marker":"[13]"},{"why":"Quoted for the root correspondence between circle roots and real roots under the Cayley map.","marker":"[39]"},{"why":"Provides the discriminant, subresultant, and semi-algebraic machinery behind Theorem 6.9 and Corollary 6.10.","marker":"[3]"},{"why":"Gives the self-inversive coefficient bound reinterpreted as an upper bound on the interlace number.","marker":"[25]"},{"why":"Gives the median-based bound for palindromic polynomials used as another upper bound on the interlace number.","marker":"[21]"}],"fun_headline_variants":["A single parameter drags roots to the unit circle","Two thresholds: interlacing and circle-rootedness","Self-inversive polynomials organized into a complete fan","When does interlacing match circle-rootedness?","Dragging roots: a parametric family with exact bounds"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that circle-rootedness of $p_\\alpha$ for all $\\alpha\\ge\\beta$ can be tested by a single common interlace between $p_\\beta$ and $x^n+1$ via the Möbius map and Proposition 2.4; the printed version of this equivalence in Proposition 6.6 makes a stronger for-every-$\\alpha$ statement that the proof only establishes at $\\beta$ and that already fails for degree-4 examples in Section 8.3.","fun_headline_variants_meta":{"raw":{"variants":["A single parameter drags roots to the unit circle","Two thresholds: interlacing and circle-rootedness","Self-inversive polynomials organized into a complete fan","When does interlacing match circle-rootedness?","Dragging roots: a parametric family with exact bounds"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000347,"raw_usage":{"total_tokens":2000,"prompt_tokens":1148,"completion_tokens":852,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":764,"completion_tokens_details":{"reasoning_tokens":776}},"tokens_in":764,"tokens_out":852,"duration_ms":8154,"temperature":1.0,"reasoning_tokens":776,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:31:08.457938+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute $p_\\alpha$ for the degree-4 palindromic polynomial $p(x)=\\sqrt{2}(x^3+x)+2x^2$ at a value $\\beta$ equal to its circle number, or just above it, and check directly whether $p_\\beta$ together with $x^4+1$ admit a common angle interlacing set; the proposition as printed predicts the interlacing holds for every $\\alpha\\ge\\beta$, so finding one $\\alpha$ in $(\\operatorname{cn}(p),\\operatorname{il}(p))$ whose roots fail to share a common angle interlace with $x^4+1$ refutes the universal form of Proposition 6.6.","supporting_citations":[{"cited_title":"Zeros and irreducibility of polynomi- als with gcd powers as coefficients","cited_arxiv_id":null,"evidence_quote":"Supplies the Fourier-evaluation idea behind Theorem 3.7, from which the interlace formula follows."},{"cited_title":"Roots on a Circle","cited_arxiv_id":null,"evidence_quote":"Explains the Möbius transformation that carries roots on the circle to real roots, the dictionary used in Section 6."},{"cited_title":"How to count the number of zeros that a polynomial has on the unit circle?","cited_arxiv_id":"1902.04231","evidence_quote":"Quoted for the root correspondence between circle roots and real roots under the Cayley map."},{"cited_title":"Algorithms in real algebraic geometry","cited_arxiv_id":null,"evidence_quote":"Provides the discriminant, subresultant, and semi-algebraic machinery behind Theorem 6.9 and Corollary 6.10."},{"cited_title":"Self-inversive polynomials whose zeros are on the unit circle","cited_arxiv_id":null,"evidence_quote":"Gives the self-inversive coefficient bound reinterpreted as an upper bound on the interlace number."},{"cited_title":"Reciprocal polynomials with all zeros on the unit cir- cle","cited_arxiv_id":null,"evidence_quote":"Gives the median-based bound for palindromic polynomials used as another upper bound on the interlace number."}],"review_version":1}