{"id":"edf9e64b-fd79-45b5-84bc-2d02e69e48f8","arxiv_id":"2504.16766","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A new R-bar operation yields locally finite Feynman integrals in higher dimensions and reduces epsilon-expanded master integrals to a minimal basis, shown on one-, two-, and three-loop examples.","lead":"This paper introduces the R-bar operation, a way to remove ultraviolet divergences from Feynman integrals without bringing back infrared or collinear divergences. The authors use it to build finite integrals that reduce the epsilon-expanded master integrals to a smaller basis, which can speed up high-order particle physics calculations.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"R-bar IR/CO-safety proof in Sec. III.C is the load-bearing step; the assertion that counterterm regions inherit from G is not fully proven, and a failure there would invalidate the reduction method.","rationale":"The paper's main claim is conditional on the R-bar operation producing IR/CO-safe UV counterterms. The reader's weakest-assumption identification matches the most load-bearing step: Sec. III.C's proof that G_r inherits IR/CO finiteness from G is only sketched. I considered alternative concerns, such as the lack of a completeness proof for the claimed 'minimal basis' and the reliance of the double-pentagon reduction on a single numerical point, but these are secondary to the core mechanism. If the R-bar construction is unsound, the locally finite integrals are not actually finite and all resulting relations among master integrals are unreliable; if it is sound, the method has a plausible basis. The paper includes one worked analytical example (Appendix A) and a numerical cross-check in the three-loop case, which provide some support, but the governing lemma is not proven in full generality. No internal inconsistency or clear counterexample is evident, so the concern is a serious gap rather than a demonstrated failure. The appropriate verdict therefore remains CONDITIONAL, with no adjustment from the reader's assessment.","tokens_in":16830,"tokens_out":6759,"duration_ms":70876,"concrete_test":"Test IR/CO finiteness of the individual counterterms and of G + G^UVCT for a nontrivial two-loop seed, e.g. the double-box integral I^(6-2ε)(1,1,1,1,1,1,2,0,0) used in Eq. (29). Concretely: introduce the η-regulator of Eq. (7) into every massless propagator of G and of each G_r; use asy2 to compute the η→0 region expansion of G + G^UVCT; verify that no η^{-κ} ln^i η terms survive and that every G_r is finite in all 13 soft/collinear regions listed in Sec. IV.B. A weaker but still useful check is to integrate the Appendix A sunrise sum at a numerical phase-space point with pySecDec: if the result depends on a UV cutoff, the R-bar subtraction is incomplete.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central mechanism is the claim in Sec. III.C that the R-bar counterterms G_r, defined by truncating the large-momentum expansion (15) at order wD in Eq. (16), are IR/CO finite whenever the original integrand G is. The proof hinges on three unproved assertions: (i) after the mass substitution (13), only IR/CO regions of the small combinations S_i can appear in G_r; (ii) any such IR/CO region R is also an IR/CO region of G; and (iii) each term g_k retained in (16) inherits the power suppression N = O(λ^a) of G. The argument that each homogeneous numerator piece φ_k independently satisfies φ_k = O(λ^a) assumes no cancellation between terms of different k after the denominator expansion, and the statement that the negative powers of λ introduced by the denominator of G_r are the same as those of G is asserted rather than derived. If the truncation or the mass shift (13) changes the scaling of a denominator factor in some soft or collinear region, a divergence can survive in G_r, making the 'locally finite' integral actually divergent and invalidating every relation of the form (2). Appendix A checks only the two-loop sunrise; it does not exercise the double-box or double-pentagon seeds used for the main reductions. This is not evidence that the proof is wrong, but it is the point where the central assertion is least secure.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a systematic method to reduce the epsilon-expanded master integrals of a Feynman integral family to a smaller set of expansion coefficients. The strategy is to construct locally finite integrals by raising the spacetime dimension (which suppresses infrared and collinear divergences) and then subtracting ultraviolet divergences with a new operation called the R-bar operation, designed to preserve IR and CO safety. These finite integrals are expressed in terms of 4-dimensional master integrals using integration-by-parts identities and dimension recurrence relations, yielding constraints of the form sum_i c_i M_i = O(epsilon^0). Expanding in epsilon gives linear relations among master-integral expansion coefficients. The method is illustrated on a one-loop pentagon, a two-loop double box, a two-loop double pentagon, and a three-loop example, with the number of independent expansion coefficients substantially reduced. An open-source Mathematica package, FFI, is provided. The central technical claim is the R-bar construction in Section III.C, which asserts that the UV counterterms are IR and CO finite whenever the original integrand is.","tokens_in":17164,"tokens_out":6094,"duration_ms":58589,"significance":"If the central construction is valid, the paper offers a practical and general route to generating finite/evanescent relations among multiloop Feynman integrals, which could meaningfully reduce the cost of high-order perturbative computations. The explicit reductions in the double-box and three-loop examples, together with the numerical cross-check of the three-loop relations, demonstrate that the method works in nontrivial settings. The open-source package FFI is a useful deliverable. However, the proof of the key R-bar property is only sketched, and a technical issue with the dimension-recurrence formula for the five-point example is not addressed; these points prevent the paper from being fully convincing in its current form.","major_comments":[{"comment":"The proof that the counterterm G_r is IR and CO finite rests on two unproved assertions. First, the statement that 'IR and CO regions of G_r must inherit from those of G' is asserted without argument; the mass substitution (13) modifies propagators in a way that could in principle create new soft or collinear scaling regions in G_r that are not regions of G. Second, the claim that each homogeneous component φ_k independently satisfies φ_k = O(λ^a) 'because there cannot be cancellation between φ_k's' is not justified: the power-counting behavior applies to the full numerator N in region R, but the expansion (15) mixes φ_k with denominator factors, so individual terms g_k need not share the same scaling. The authors should either give a complete proof of these two points or verify the IR/CO finiteness of the counterterms numerically for every seed integral used in the main examples; Appendix A only checks the two-loop sunrise, which is too simple to test the double-box and double-pentagon cases where the subtraction is more intricate. This gap is load-bearing because all subsequent constraints assume the constructed integrals are actually finite.","section":"III.C, Eqs. (13)-(17)"},{"comment":"The dimension-recurrence relation (22) contains the Gram determinant V(p_1,...,p_E) in the denominator. For the double-pentagon example of Section IV.C, E=5 and the external momenta satisfy p_1+...+p_5=0 with four-dimensional kinematics, so V(p_1,...,p_5) vanishes identically. The manuscript does not explain how the dimension shift from D=6-2ε to D=4-2ε is implemented for this family. If Eq. (22) is used directly, the denominator is zero; if a different procedure is used, such as taking a limit or using an alternative recurrence, it should be spelled out. Since the double pentagon is one of the principal examples supporting the method, this technical point needs to be resolved or explicitly addressed.","section":"III.D, Eq. (22), and IV.C"},{"comment":"The paper claims that the methods reduce the epsilon-expanded master integrals to a 'minimal basis', but it does not prove that the finite integrals generated by scanning a limited set of dimensions, ranks, and dot orders are sufficient to obtain the true minimal basis. The statement in Section IV.C that 'there are 108 MIs for the double pentagon' and that the divergent parts reduce to '127 MCs' describes the result of a particular finite scan, not a proof of completeness. Furthermore, the double-pentagon analysis is performed at a single numerical kinematic point; unless it is shown that the rank of the constraint system is independent of the point (or that the chosen point is generic), the reported MC counts could be special. The authors should either prove completeness of the generated constraints for the examples or soften the claims of minimality to 'minimal among the constraints generated by this procedure'.","section":"IV.C and VI"}],"minor_comments":[{"comment":"The text contains the typo 'cen deduce' instead of 'can deduce' in the sentence following Eq. (24).","section":"IV.A, after Eq. (24)"},{"comment":"The phrase 'cij's can be choses to satisfy' contains a grammatical error; it should read 'can be chosen to satisfy'.","section":"III.C, paragraph after Eq. (13)"},{"comment":"The text reads 'FFI alse provides methods', which should be 'FFI also provides methods'.","section":"V, second paragraph"},{"comment":"The affiliation line contains the typo 'Shanghai Jiao Tong Univeristy', which should be 'University'.","section":"Author affiliations"},{"comment":"The notation 'η−κ lniη' is printed without a superscript on the logarithm; this should be typeset as 'η^{-κ} ln^i η' or similar to be unambiguous.","section":"II, after Eq. (7)"}],"recommendation":"major_revision","confidential_remarks":"The paper presents a promising technique with a useful package and nontrivial examples, but the R-bar proof gap and the unresolved Gram-determinant issue for the double pentagon are serious enough to require a major revision. The authors should be encouraged to either provide a rigorous proof or add direct numerical verification of IR/CO finiteness of the counterterms for the multi-loop seeds. The minimality claim should also be moderated unless a completeness argument is supplied. There are no concerns about novelty or citation practices."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This paper is worth a serious read. The R-bar operation, a UV subtraction that preserves IR/CO safety, is a genuinely new twist on the standard R-operation. The strategy of building high-dimensional finite integrals and using them to constrain epsilon-expanded master integrals is clever, and the examples make it concrete: they reproduce the known one-loop pentagon relation, reduce the two-loop double box and double pentagon, and give a three-loop case with numerical cross-checks. The FFI package is open-source, which is real evidence of reproducibility.\n\nThe central logic holds up: higher dimensions suppress IR/CO regions, R-bar removes UV, and finite combinations impose linear constraints. The explicit double-box equations (28)-(30) are consistent, and I found no circular reasoning. Reliance on Blade and DRR is tool use, not circularity.\n\nThe soft spots are where the proof is sketched. Section III.C argues that the counterterm G_r inherits IR/CO finiteness from G, but the argument rests on three assertions that are stated rather than fully derived: that only IR/CO regions of the small combinations can appear after the mass substitution, that those regions are also regions of G, and that each homogeneous numerator piece independently has the required power suppression. Appendix A checks only the two-loop sunrise, not the double-box or double-pentagon seeds used in the main reductions. This is not evidence of a flaw, but it is the load-bearing step and needs a tighter proof. The double-pentagon reduction is also checked at a single numerical point, and the 'minimal basis' claim is not a formal lower bound; it is minimal with respect to the constraints found. These are addressable, not fatal.\n\nIf I were refereeing, I would ask for a fuller proof of the inheritance lemma, more numerical checks, and a precise statement of what 'minimal' means. The paper deserves peer review; it is a serious contribution to multiloop reduction methods. I would cite it, and I would bring it to the reading group.","headline":"A genuinely new method for reducing epsilon-expanded master integrals via a UV-subtraction scheme that preserves IR/CO safety; the main proof sketch needs tightening, but the work deserves a serious referee.","tokens_in":17663,"tokens_out":2928,"would_cite":true,"duration_ms":24458,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81Q30","81T18","81T15"],"pacs":[],"model":"deepseek-v4-flash","headline":"$\\bar{R}$-operation removes ultraviolet divergences from Feynman integrals while preserving their infrared and collinear safety, and reduces the $\\epsilon$-expanded master integrals to a minimal basis.","keywords":["Feynman integrals","master integrals","epsilon expansion","ultraviolet subtraction","infrared divergence","collinear divergence","dimensional recurrence relations","integration by parts"],"falsifier":"Compute the full $\\bar{R}$-subtracted integrand $G+G^{UVCT}$ for the two-loop double box at a generic kinematic point, introducing a small regulator $\\eta$ into every propagator; if its expansion as $\\eta\\to 0$ contains any $1/\\eta$ or $\\ln\\eta$ term, the subtraction has reintroduced an infrared or collinear divergence and the central claim of local finiteness is false.","tokens_in":16603,"feed_emoji":"🧮","tokens_out":10794,"duration_ms":95389,"temperature":0.7,"pith_summary":"Feynman integrals are normally reduced to master integrals in general dimension $D=4-2\\epsilon$, but those master integrals are not independent when only the first few terms of the $\\epsilon$-expansion are needed. This paper establishes a systematic way to construct locally finite integrals, finite in every infrared, collinear, and ultraviolet region, by raising the spacetime dimension and then applying a modified ultraviolet subtraction, the $\\bar{R}$-operation, whose counterterms do not reintroduce soft or collinear divergences. Expressed back in four dimensions through integration-by-parts identities and dimensional recurrence relations, these finite integrals impose constraints that cut the $\\epsilon$-expansion coefficients of master integrals to a much smaller set of master coefficients. If correct, this substantially lowers the cost of high-order perturbative computations in quantum field theory, since only the reduced set of coefficients needs to be evaluated.","feed_headline":"R-bar subtraction cuts multiloop integrals to a minimal basis","feed_subtitle":"It removes ultraviolet poles without creating new soft or collinear divergences, so fewer expansion coefficients are needed.","key_machinery":"The load-bearing object is the $\\bar{R}$-operation: the standard $R$-operation subtraction formula $G^{UVCT} = \\sum (-1)^k G_{r_1\\cdots r_k}$, but with each counterterm $G_r$ built after the variable substitution $l_i\\cdot l_j = \\hat{l}_i\\cdot \\hat{l}_j + c_{ij}m^2$, $p_i\\cdot l_j = p_i\\cdot \\hat{l}_j$, which puts a mass $m^2$ into the propagators of the large loop momenta while leaving the numerator alone. The counterterm is then the part of the large-momentum expansion (15) down to the superficial degree of divergence $wD$. Because the denominators of $G_r$ inherit their infrared and collinear scaling from the original integrand, and each term in the expansion inherits one homogeneous piece of the numerator, the counterterm cannot create new soft or collinear poles. Around this object, the method uses the strategy of regions with a small mass regulator to certify infrared and collinear finiteness, and Tarasov's dimensional recurrence relations to bring high-dimensional finite integrals back to $4-2\\epsilon$.","core_discovery":"Starting from any Feynman integral in a given family, one can first choose a sufficiently high spacetime dimension so that all infrared and collinear regions are power-counting finite. The paper's central proposal is the $\\bar{R}$-operation, a variant of the $R$-operation in which each ultraviolet counterterm is constructed by expanding the integrand in the large loop momenta after shifting the squared perpendicular components of those momenta by a mass term, while leaving the numerator untouched. This mass substitution is what prevents the counterterms from developing new infrared or collinear singularities. The resulting locally finite integrals, once reduced to master integrals by IBP and converted to $D=4-2\\epsilon$ by dimensional recurrence relations, yield relations of the form $\\sum_i c_i M_i = O(\\epsilon^0)$ or $O(\\epsilon)$; expanding in $\\epsilon$ gives linear constraints that reduce the divergent and low-order finite parts of the master integrals to a minimal set of master coefficients. The paper demonstrates the reduction on the one-loop pentagon, the two-loop double box and double pentagon, and a three-loop example, and packages the procedure in an automated tool.","pith_inferences":["If the $\\bar{R}$-operation preserves infrared and collinear safety for all families, the number of master coefficients at each order in $\\epsilon$ becomes a well-defined invariant of an integral family; computing this number across a set of standard topologies would provide a clean stress test of the method.","The appearance of vacuum-like ultraviolet-counterterm families among the master coefficients hints that the divergent structure of $\\epsilon$-expanded integrals may be controlled by a small universal set of vacuum integrals, though the paper's examples do not prove this.","The same mass-shift idea could be transplanted to local subtraction schemes for physical cross sections, where one also needs to remove ultraviolet poles without disturbing soft and collinear safety; the authors do not explore that application.","A natural extension would be to apply the reduction to a family with known analytic results, such as a non-planar two-loop box, and compare the predicted master coefficients with direct integration; this would test the completeness of the generated constraints."],"forward_implications":["For a given integral family, the divergent parts of all master integrals can be expressed through a small set of master coefficients, many of which are simpler subsector or vacuum integrals, so only those coefficients need to be evaluated.","The one-loop pentagon relation $E_0 = \\sum_i a_i D_{0i}+O(\\epsilon)$ is recovered from the method, confirming that it generalizes a known simplification to higher loops.","In the two-loop double box, the 32 expansion coefficients of the eight master integrals through order $\\epsilon^{-1}$ are cut by half; in the two-loop double pentagon, the divergent parts of 108 master integrals reduce to 127 master coefficients drawn from subsectors.","The method works at three loops in the tested family, where the divergent parts of all nine master integrals reduce to ten master coefficients, and the automated package makes the reduction turnkey once a family is defined."],"supporting_citations":[{"why":"Establishes that the space of Feynman integrals in a family is finite-dimensional, so a master-integral basis exists.","marker":"[1]"},{"why":"Supplies the one-loop pentagon relation $E_0=\\sum_i a_iD_{0i}+O(\\epsilon)$ that motivates and benchmarks the reduction method.","marker":"[26]"},{"why":"Provides the dimensional recurrence relations used to convert high-dimensional finite integrals back to $D=4-2\\epsilon$.","marker":"[44]"},{"why":"Describes an earlier method for constructing finite Feynman integrals by special numerators, which the paper compares with and extends.","marker":"[45]"},{"why":"Gives another construction of finite integrals from Feynman polytopes, used as an alternative route in the four-dimensional approach.","marker":"[46]"},{"why":"Supplies the R-operation subtraction formula whose modified form is the $\\bar{R}$-operation.","marker":"[47–49]"},{"why":"Provides the strategy of regions used throughout to identify and power-count all singular integration regions.","marker":"[50]"},{"why":"Automatic expansion-by-regions analysis used to certify which integrals are infrared and collinear finite.","marker":"[54]"}],"fun_headline_variants":["R-bar subtraction yields minimal basis for multiloop integrals","R-bar operation tames UV poles, preserves IR safety, fewer coefficients","Automated R-bar: subtract UV, keep IR, shrink epsilon basis","R-bar method cuts Feynman master integrals to minimal basis"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The method rests on the assumption that putting a mass-like shift into the large loop momenta of each ultraviolet counterterm cannot create new infrared or collinear singular regions, so every subtraction term stays finite wherever the original integrand was safe.","fun_headline_variants_meta":{"raw":{"variants":["R-bar subtraction yields minimal basis for multiloop integrals","R-bar operation tames UV poles, preserves IR safety, fewer coefficients","Automated R-bar: subtract UV, keep IR, shrink epsilon basis","R-bar method cuts Feynman master integrals to minimal basis"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000587,"raw_usage":{"total_tokens":2729,"prompt_tokens":890,"completion_tokens":1839,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":506,"completion_tokens_details":{"reasoning_tokens":1774}},"tokens_in":506,"tokens_out":1839,"duration_ms":11321,"temperature":1.0,"reasoning_tokens":1774,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T10:56:37.982944+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the full $\\bar{R}$-subtracted integrand $G+G^{UVCT}$ for the two-loop double box at a generic kinematic point, introducing a small regulator $\\eta$ into every propagator; if its expansion as $\\eta\\to 0$ contains any $1/\\eta$ or $\\ln\\eta$ term, the subtraction has reintroduced an infrared or collinear divergence and the central claim of local finiteness is false.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes that the space of Feynman integrals in a family is finite-dimensional, so a master-integral basis exists."}],"review_version":1}