{"id":"64e40905-aa73-442f-bbca-90e33cfd3089","arxiv_id":"2505.09390","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A path integral over a real Gaussian free field with a Hankel-contour zero mode reproduces the imaginary DOZZ three-point function without a neutrality condition.","lead":"A new recipe for the imaginary version of Liouville quantum field theory is proposed, based on a random surface field plus an integration contour borrowed from the Gamma function. The paper gives exact results on a circle and numerical evidence on a sphere that the theory matches the conjectured imaginary DOZZ structure constants.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The equality with the imaginary DOZZ constant hinges on an unproved power-law tail for the sphere moment generating function G^hat C_{β,α}(µ); if that tail is wrong, the U-contour integral in Eq. (15) can diverge or acquire extra terms.","rationale":"The reader's weakest assumption is the tail of the sphere moment generating function, and that is also the point I would defend as load-bearing. The clean circle computation provides the template, but the sphere case is genuinely conjectural: the moments are not Selberg integrals of the same form, and no integral representation for G^hat C exists. The numerical comparison is evidence but not a proof, and the paper itself flags the under-sampling near poles. I do not see an internal inconsistency in the Lagrangian setup; the issue is missing support for the exact equality with the imaginary DOZZ constant. A controlled tail estimate, or a rare-event numerics check of the tail, would settle whether Eq. (17) is a theorem or a conjecture. Nothing in my read changes the reader's CONDITIONAL verdict.","tokens_in":17411,"tokens_out":17363,"duration_ms":198564,"concrete_test":"At the parameter point of Fig. 9 (β=0.44, α1=-3.94, α2=-3.58, α3=-2.4), compute G^hat C_{β,α}(µ) with an importance-sampling estimator that biases samples toward large |Re M|, for µ from 10^4 to 10^8, and compare the Nmodes=2^11 and 2^14 cutoffs. Fit log|G(µ)| versus log µ and check for oscillations in arg G(µ). Then recompute the contour integral (15) with the fitted tail and compare to Eq. (25). If the fitted exponent is not compatible with the pole positions of C_ImDOZZ, or if the recomputed integral deviates from C_ImDOZZ by more than the relative deviation reported in Fig. 9, the conjecture underlying Eq. (17) is falsified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Equation (17) is the central claim: the U-contour integral (15) equals the imaginary DOZZ structure constant. The integral only has the asserted value if G^hat C_{β,α}(µ) decays as a power law C µ^{-λ} as µ→+∞ on the positive real axis, with no oscillatory or logarithmic corrections, for all charges in the allowed range. The paper does not have this input. The text says exactly: 'We conjecture that the G^hat C_{β,α}(µ) behaves similarly to the circle case', and the general-surface section says 'We postulate that E[e^{-µMΣ,g}]∼µ^{-λ}'. The circle case is proven via exact Selberg moments and a one-dimensional integral representation; neither ingredient is available on the sphere, where the moments are Dotsenko-Fateev integrals and no integral representation of the Laplace transform is given. The numerical check (Fig. 9) is carried out at moderate values of µ and, as the authors themselves note, loses accuracy near poles of the imaginary DOZZ because the tail is under-sampled. Unless a controlled estimate of this tail is supplied, the vertical branches of U are not known to converge, the contour deformation in Eq. (16) may miss boundary terms, and the claimed agreement with the unique analytic continuation is not established.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a probabilistic construction of imaginary Liouville field theory based on a real (non-compactified) Gaussian free field, with the zero mode integrated over a U-shaped Hankel-type contour in the complex plane. For the circle, the authors rigorously compute the one-point function by combining known Selberg-integral moments of imaginary Gaussian multiplicative chaos with the exact power-law tail of its Laplace transform, obtaining the closed expression in Eq. (8). For the sphere, they define the three-point function as the U-contour integral in Eq. (15) involving the Laplace transform of the sphere imaginary GMC, rewrite it as Eq. (16), and conjecture that this equals the imaginary DOZZ structure constant (Eq. (17)). The equality is supported by numerical Monte Carlo simulations shown in Fig. 9 and by the observation that at integer values of the parameter s the expressions reduce to Dotsenko-Fateev/Coulomb gas integrals. The manuscript also sketches a definition for general surfaces under a postulated power-law decay of the Laplace transform.","tokens_in":17774,"tokens_out":2641,"duration_ms":29204,"significance":"If the central claim (17) were established rigorously, this would be an important step: it would provide the first explicit Lagrangian path-integral construction reproducing the imaginary DOZZ structure constants without a neutrality constraint, with direct relevance to c ≤ 1 conformal field theories, Potts models, and O(n) loop models. The circle one-point function is a rigorous and valuable exact result, obtained from Selberg integrals and a clean asymptotic analysis. The general idea of using a complex zero-mode contour to continue the Coulomb gas integrals to real s is attractive and well motivated by prior work on analytic continuation of Liouville theory. However, the sphere three-point claim rests on a conjectured tail behavior and on numerical evidence of limited scope, so the paper is best viewed as a compelling proposal with a rigorous core on the circle and an open analytical gap on the sphere.","major_comments":[{"comment":"The numerical support in Fig. 9 is not sufficient for the full parameter range of Eq. (17). The figure uses a single parameter set beta = 0.44, alpha1 = -3.94, alpha2 = -3.58, a fixed cutoff N_modes = 211, and no reported error bars or convergence tests in N_modes and sample number. The authors state in the supplementary material that the simulations lose accuracy near a pole of the imaginary DOZZ function, precisely where the tail of the GMC distribution determines the result. Since the claimed equality is supposed to hold for all allowed alpha_i > Q, the numerical evidence covers only a small slice of the domain. At a minimum, the paper should show error estimates, convergence with cutoff, and tests at several non-integer values of s away from integer points and from poles.","section":"Eqs. (15)–(17) and the paragraph after Eq. (15)"},{"comment":"Part of the agreement with the imaginary DOZZ constant is by construction: for integer s, Eq. (16) reduces exactly to the Dotsenko-Fateev/Coulomb gas integrals, and the imaginary DOZZ constant is defined as the unique analytic continuation of those integrals (the authors cite Schomerus, Kostov-Petkova, Zamolodchikov, Dotsenko for this). Therefore equality at integer s is not an independent confirmation of Eq. (17). The genuinely independent content of the claim is the extension to non-integer s, and the numerical evidence for that extension is the limited set shown in Fig. 9. The paper should state this distinction explicitly and focus the numerical and analytical effort on the non-integer regime.","section":"Eq. (16) and the paragraph beginning 'When s in N'"},{"comment":"The proposed definition for a general surface (Sigma,g) is based entirely on the postulated power-law decay E[e^{-mu M} ] ~ mu^{-lambda}, with no proof or numerical test except on the circle and sphere. The manuscript presents this as a definition for LFT_i_beta, but the convergence of the U-contour integral and the validity of the Seiberg-type bounds depend critically on the value of lambda and on the absence of oscillatory or logarithmic corrections. Since the sphere case, which is the paper's main result, already depends on an unproved conjecture, the general-surface proposal should be clearly labeled as conditional on this postulate and not as a construction of the theory.","section":"General-surface section and Seiberg bounds"}],"minor_comments":[{"comment":"The text refers to 'the asymptotic behavior (24)' and says it is checked in the supplementary material, but Eq. (24) appears only in the supplementary material and is not labeled in the main text; the cross-reference should be corrected.","section":"Main text after Eq. (23)"},{"comment":"Some axis labels in Fig. 5 are garbled, e.g. 'G S_{beta,epsilon}(e^{i beta/2 c})]' appears to be missing a 'Re' or 'Im' prefix; please fix the LaTeX and ensure both panels have clear axis labels.","section":"Fig. 5 caption"},{"comment":"The statement 'the accuracy of the numerical results diminishes in the vicinity of the pole s = -4/beta^2' is important and should appear in the main text as a limitation of the numerical method, not only in the supplementary material.","section":"Supplementary material, sentence after Eq. (45)"},{"comment":"The notation 's in -N*' is nonstandard; write 's in -N' or 's = -n with n in N' for clarity.","section":"Main text, paragraph before Eq. (11)"},{"comment":"The reference '[Not]' is informal; if the Girsanov/Cameron-Martin formula is used, cite a standard textbook or the original source instead of a parenthetical remark.","section":"Reference list"}],"recommendation":"major_revision","confidential_remarks":"The paper is a well-written and interesting proposal, and the circle result is rigorous and valuable. The major obstacle is the unproved tail behavior of the sphere Laplace transform, which is load-bearing for the central equality with the imaginary DOZZ constant. I would be willing to reconsider after the authors either supply a rigorous tail estimate (even for a restricted parameter range) or substantially reframe the central claim as a conjecture supported by more extensive numerics with error bars. I did not see any indication of inappropriate citation practices or scope mismatch; the topic is appropriate for hep-th."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, this is a genuinely novel construction: a real non-compactified GFF plus a complex zero mode integrated over a Hankel contour, which sidesteps the neutrality condition that blocks earlier compactified imaginary Liouville theories. Second, the central sphere claim is conditional: the equality to the imaginary DOZZ constant depends on a power-law tail for the Laplace transform of the sphere imaginary GMC, which the paper itself states as a conjecture. I think the reader's assessment is right.\n\nWhat is actually new: the circle one-point function (Eq. 8) is derived exactly from Selberg moments and the moment generating function; that part is solid. The reduction of Eq. (16) to Dotsenko-Fateev integrals for integer s is exact, and the fact that the U contour reproduces those residues is a nice structural feature. The numerical test on the sphere (Fig. 9) is a genuine check for non-integer s, across many orders of magnitude, even if error bars are absent.\n\nSoft spots. The load-bearing assumption is the conjectured behavior G^C_hat_{beta,alpha}(mu) ~ mu^{-lambda} as mu -> +infinity. If there are log corrections or oscillatory terms, the vertical branches of U may not converge, or boundary terms in the contour deformation could appear. The paper is honest about this -- it says \"conjecture\" and \"postulate\" -- but it means Eq. (17) is not established. The integer-s agreement is partly by construction: for integer s, the contour reduces to the Coulomb gas integrals that define the integer values of the imaginary DOZZ constant. The non-integer comparison is the meaningful check, and it looks good at moderate mu, but near the poles the simulations lose accuracy precisely because the tail is under-sampled. The general-surface section is explicitly heuristic, and no code is shipped. None of this is disqualifying, but it is exactly where a referee should push.\n\nThe citation pattern is appropriate; the relevant prior work (CILT, RS15, LRV15) is cited, and the novelty claim is modest and credible. This paper deserves a serious referee: the circle result and the construction are worth publishing even if the sphere conjecture ends up requiring revision. I would send it to peer review and ask for a proof or much stronger numerical support of the sphere tail, plus error bars and code.","headline":"A genuinely new probabilistic proposal for imaginary Liouville theory with an exact circle result and a striking numerical match on the sphere; the sphere three-point claim rests on an unproved tail conjecture, so the paper is conditional rather than established.","tokens_in":18214,"tokens_out":2405,"would_cite":true,"duration_ms":24690,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81T40","60G60"],"pacs":["11.25.Hf"],"model":"deepseek-v4-flash","headline":"Imaginary Liouville field theory, defined by a U-shaped zero-mode contour, reproduces the imaginary DOZZ structure constants without a neutrality constraint.","keywords":["imaginary Liouville field theory","Gaussian free field","Gaussian multiplicative chaos","imaginary DOZZ structure constants","Coulomb gas integrals","Hankel contour","conformal bootstrap","Potts and O(n) loop models"],"falsifier":"For fixed $\\beta\\in(0,\\sqrt{2})$ and charges $\\alpha_i>Q$, estimate the sphere tail $E[e^{-\\mu M^{\\hat C}_{\\beta,\\alpha}}]$ for large $\\mu$; if $\\mu^{4/\\beta^2}E[e^{-\\mu M^{\\hat C}_{\\beta,\\alpha}}]$ does not approach a nonzero constant, the contour integral (15) will diverge or acquire extra terms and $C_\\alpha=C^{\\mathrm{ImDOZZ}}_\\alpha$ will fail. The same test near a pole of $C^{\\mathrm{ImDOZZ}}$, where the numerical method already loses accuracy, is a direct place to look.","tokens_in":17240,"feed_emoji":"🎲","tokens_out":14316,"duration_ms":123787,"temperature":0.7,"pith_summary":"The paper proposes a probabilistic construction of imaginary Liouville field theory ($\\mathrm{LFT}_{i\\beta}$), the theory with imaginary coupling whose correlation functions were previously available only in a compactified version with a strict neutrality condition. It splits the field into a real Gaussian free field plus a complex constant $c$, and integrates the constant along a U-shaped Hankel-type contour $U$ instead of the real line. The central claim is that this non-compactified path integral reproduces the imaginary DOZZ structure constants, with no neutrality constraint, and that the theory is the first explicit Lagrangian field theory to do so. This matters because these structure constants are the amplitudes behind three-point connectivities in geometric statistical models such as Potts and $O(n)$ loop models, and because a direct Lagrangian definition for central charge $c_{\\mathrm{charge}}\\le 1$ has been missing. The one-point function on the circle is obtained exactly, and the three-point function on the sphere is matched numerically to the imaginary DOZZ constant.","feed_headline":"Imaginary Liouville path integral yields the imaginary DOZZ constants","feed_subtitle":"Choosing the U-shaped contour removes the neutrality condition and yields the amplitudes behind geometric statistical models.","key_machinery":"The central object is the moment generating function $G_{\\beta,\\alpha}(\\mu)=E[e^{-\\mu M_{\\beta,\\alpha}}]$ of the imaginary Gaussian multiplicative chaos $M_{\\beta,\\alpha}$, the limit of $\\int e^{i\\beta X_\\epsilon}\\,\\epsilon^{-\\beta^2/2} dv$ with charge insertions, together with the U-shaped Hankel contour for the zero mode $c$. On the circle the Laplace transform is evaluated through Selberg integrals as $G^S_\\beta(\\mu)=\\int_0^\\infty e^{-\\mu t^{\\beta^2/4}-t}dt$, whose algebraic tail $\\mu^{-4/\\beta^2}$ selects $U$ as the contour on which the one-point integral converges, in direct analogy with the inverse Gamma function. On the sphere the same mechanism is conjectured to hold: sufficient decay of the integrand in $C_\\alpha=\\int_U e^{i(\\bar\\alpha-2Q)c}G^{\\hat C}_{\\beta,\\alpha}(\\mu e^{i\\beta c})dc$ makes the three-point function finite, and the shift equations together with duality $\\beta\\to -4/\\beta$ identify the result with $C^{\\mathrm{ImDOZZ}}_\\alpha$. The U contour is the load-bearing choice: it replaces the neutrality constraint of the compactified theory by a prescription valid for real $s=(2Q-\\sum_i\\alpha_i)/\\beta$.","core_discovery":"The discovery is a contour prescription that makes the imaginary Liouville path integral well defined. Writing $\\phi = X + c$ with $X$ a real Gaussian free field and $c$ a complex constant, the paper integrates $c$ along $U$, made of the lines $[0,-i\\infty)$, $[0,2\\pi/\\beta]$ and $[2\\pi/\\beta,2\\pi/\\beta-i\\infty)$ on the sphere, with $4\\pi/\\beta$ replacing $2\\pi/\\beta$ on the circle. On the circle this yields the exact one-point formula $\\langle V_\\alpha\\rangle = \\frac{4\\pi}{\\beta} e^{-i\\pi s} \\mu_B^s \\frac{\\Gamma(1+s\\beta^2/4)}{(\\Gamma(1+\\beta^2/4))^s \\Gamma(1+s)}$, built on the rigorous tail $G^S_\\beta(\\mu)\\sim C \\mu^{-4/\\beta^2}$ of the Laplace transform of imaginary Gaussian multiplicative chaos. On the sphere the same contour gives $C_\\alpha = \\int_U e^{i(\\bar\\alpha-2Q)c} G^{\\hat C}_{\\beta,\\alpha}(\\mu e^{i\\beta c}) dc$, and the paper claims this equals the imaginary DOZZ structure constant $C^{\\mathrm{ImDOZZ}}_\\alpha$, with numerical simulations in close agreement. The construction is presented as the first explicit Lagrangian theory reproducing those constants without a neutrality condition, and as a real-$s$ extension of the Coulomb-gas Dotsenko-Fateev integrals.","pith_inferences":["A proof of the conjectured power-law tail on the sphere, perhaps by extending the Selberg-type moment computations used on the circle, would turn the numerical match into a theorem; the same exact moment method that produces the circle tail is a plausible route.","The numerical breakdown near poles suggests rare-event tails dominate the Laplace transform there, so importance sampling or a direct tail-exponent computation at the pole would sharpen the test without a full proof.","Because the paper leaves four-point factorization open, a natural next test is to compute a four-point ratio from the U-contour and compare with the known non-compact bootstrap solution, which would reveal whether the U-contour theory is the same conformal field theory or a new one.","Rational $\\beta^2$ or the $\\beta\\to\\sqrt{2}$ boundary could anchor the construction to exact minimal-model or $c=1$ answers, providing cross-checks outside the parameter range currently simulated."],"forward_implications":["The three-point functions of a Lagrangian field theory at central charge $c_{\\mathrm{charge}}\\le 1$ become one-dimensional contour integrals over imaginary Gaussian multiplicative chaos, with no neutrality condition.","For integer $s$, the U-contour definition reduces to the Dotsenko-Fateev Coulomb-gas integrals, so the theory is the real-$s$ interpolation of the compactified imaginary Liouville theory.","The circle one-point function obeys the expected shift relations and the duality $\\beta\\to -4/\\beta$, and the same zero-mode integral with the contour $i\\mathbb{R}$ gives the analytic continuation of real Liouville theory.","On general closed surfaces, the same prescription defines correlation functions under the Seiberg-type bounds $\\alpha_j>Q$ and $s<\\beta\\lambda$, with a pole at $s=\\beta\\lambda$ fixed by the Laplace-transform tail.","If the equality holds, the paper supplies a Lagrangian counterpart to the bootstrap solutions used in Potts and $O(n)$ loop models."],"supporting_citations":[{"why":"Defines compactified imaginary Liouville theory, proves the exponential moments that make the generating functions well defined, and gives the normalization used in the imaginary DOZZ formula.","marker":"[GKRa]"},{"why":"Establishes existence and the nontrivial regime of complex Gaussian multiplicative chaos for beta in (0, sqrt 2), the probabilistic input for the imaginary chaos.","marker":"[LRV15]"},{"why":"Supplies the Selberg integral evaluation used to compute the moments of the imaginary chaos and hence the exact circle Laplace transform.","marker":"[FW08]"},{"why":"Motivates the U-shaped Lefschetz-thimble contour as the analytic-continuation prescription for imaginary Liouville theory.","marker":"[HMW11]"},{"why":"Shows U-shaped integrals perform the lattice Liouville analytic continuation as b goes to i beta, the direct precedent for the contour.","marker":"[CSU23]"},{"why":"Introduces the imaginary DOZZ three-point function as the continuation consistent with beta to -4/beta, the target of the identity (17).","marker":"[Zam05]"},{"why":"Provides the c <= 1 bootstrap solution whose structure constants are the imaginary DOZZ constants and whose four-point factorization is the paper's open question.","marker":"[RS15]"},{"why":"Derives the unique analytic continuation of Dotsenko-Fateev integrals with three electric charges, used to argue the U-contour theory must coincide with the imaginary DOZZ constant.","marker":"[Dot16]"}],"fun_headline_variants":["U-shaped contour makes imaginary Liouville theory well-defined","Imaginary Liouville without neutrality via new contour","New contour yields imaginary DOZZ constants from path integral","Contour trick gives imaginary DOZZ, no neutrality needed"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The sphere claim stands on the unproven assumption that the average $E[e^{-\\mu M}]$ decays as a pure power of $\\mu$ for large $\\mu$, just as it does on the circle; the paper's own numerical agreement also degrades near the poles of the imaginary DOZZ constant.","fun_headline_variants_meta":{"raw":{"variants":["U-shaped contour makes imaginary Liouville theory well-defined","Imaginary Liouville without neutrality via new contour","New contour yields imaginary DOZZ constants from path integral","Contour trick gives imaginary DOZZ, no neutrality needed"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00024,"raw_usage":{"total_tokens":1511,"prompt_tokens":933,"completion_tokens":578,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":549,"completion_tokens_details":{"reasoning_tokens":516}},"tokens_in":549,"tokens_out":578,"duration_ms":6500,"temperature":1.0,"reasoning_tokens":516,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T21:33:06.713424+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For fixed $\\beta\\in(0,\\sqrt{2})$ and charges $\\alpha_i>Q$, estimate the sphere tail $E[e^{-\\mu M^{\\hat C}_{\\beta,\\alpha}}]$ for large $\\mu$; if $\\mu^{4/\\beta^2}E[e^{-\\mu M^{\\hat C}_{\\beta,\\alpha}}]$ does not approach a nonzero constant, the contour integral (15) will diverge or acquire extra terms and $C_\\alpha=C^{\\mathrm{ImDOZZ}}_\\alpha$ will fail. The same test near a pole of $C^{\\mathrm{ImDOZZ}}$, where the numerical method already loses accuracy, is a direct place to look.","supporting_citations":[],"review_version":1}