Pith. Sign in

REVIEW

Automatized analytic continuation of Mellin-Barnes integrals

PaperPDF from arXiv

data files

ReviewThe objections and verdict, as JSON Paper breakdownThe authors' claims, conditions and measurements, as JSON Signed receiptProof this review has not been altered
X Bluesky LinkedIn Reddit Hacker News
M. Czakon

High Energy Physics - Phenomenology · Submitted 16 Nov 2005 · 18 references

Pith's verdict: The MB package paper: a genuinely useful automation tool, honest about its own failure modes, with the contour optimizer as the one soft spot worth watching.

T0 · AI review

How this review was made: An AI referee (selfhosted/deepseek-v4.1-flash) read the paper on October 29, 2026 and answered a fixed list of questions about it. T0 is the starting level every paper gets. The levels describe how deeply Pith has checked the paper, never the authors.

  1. T0 AI review An AI read the paper and answered a fixed list of questions.
  2. T1 Audit Automatic checks that the references and files the paper relies on exist and match.
  3. T2 Compute The paper's code or calculations re-run with the same results.
  4. T3 Formal statement The claims written out in a formal math language a computer can read.
  5. T4 Proof Those claims proved, and the proof checked by computer.

How Pith checks papers →

· 2 major objections · 8 minor · Ask about this paper · The compiled record · Read the full review →
Challenge this review Comment (0) Read on arXiv ↗
Cite ▾
arXiv IDhep-ph/0511200 (version 2)
Pith IDD424RPWC, Pith's permanent record of this paper and its review
CitationM. Czakon. "Automatized analytic continuation of Mellin-Barnes integrals." arXiv:hep-ph/0511200 (2005). https://pith.science/paper/D424RPWC
Cite this reviewBibTeX for Pith's review of this paper
Claim paper

Why Pith says this

Highlight any text to comment on it.

Referee report

How Pith checks papers → · Full audit of this paper →

The paper

The paper claims

A Mathematica package automates the analytic continuation of Mellin-Barnes integrals in a chosen parameter, and shows where the same integrals stop being numerically tractable.

The reading

This is a program paper: it presents MB, a Mathematica package that automates the step in Mellin-Barnes (MB) calculations where an integral must be continued in a parameter — usually the dimensional-regularization parameter $epsilon$ — from a region where every gamma function has positive argument down to a target value where some of them do not. The claim is that this continuation can be done by machine: the package finds a contour assignment, tracks which poles cross which contours as the parameter moves, and returns the resulting residue integrals in a form ready for expansion or numerical evaluation. It matters because that continuation step was the part of the MB method that resisted automation, so a working package turns a case-by-case craft into a routine call and lowers the entry cost of higher-order Feynman diagram calculations. The paper is candid that the second half of the promise — numerical evaluation — holds cleanly only in the Euclidean regime and for massless graphs at physical kinematics; for massive graphs in the physical region the author exhibits integrands that are not even absolutely integrable and argues that full automatization there may not be achievable.

What carries the argument

The load-bearing mechanism is the recursion sketched in Fig. 1: for each integral on the current list, find every gamma-function argument $F(z,a,epsilon)$ that reaches a non-positive integer as $epsilon$ runs from its starting value to the target, take the residue at the corresponding $z$ with the sign fixed by the coefficient of that variable in $F$, and add the resulting lower-dimensional integral to the next round, stopping when the target is reached. Two supporting pieces do real work: MBoptimizedRules, which searches for contour placements that avoid as many residues as possible and guarantees that no contour starts or ends on a pole, and the Barnes1/Barnes2 routines, which regularize a would-be lemma application by shifting parameters by $epsilon$ so the contour separates the poles and then continuing back to zero. The numerical side rests on writing the multidimensional integrals out as Fortran programs for CUBA, with a contour-shifting utility that moves each contour toward the middle of the gap between neighbouring poles to reduce peaking.

Extended reading notes

Core claim

On the author's own terms the contribution is a working implementation of a known algorithm plus a set of tools built around it. Following the prescription of Ref. [6], the package fixes contours parallel to the imaginary axis and performs the analytic continuation by accounting for poles that cross those contours as the parameter is taken from $epsilon_I$ to $epsilon_0$; one modification is imposed, namely that no two contours may be crossed simultaneously, which the author argues is always achievable by infinitesimal shifts of one contour. The output is a nested list of residue integrals, each carrying its own contour, that can be merged by linearity, filtered for terms that vanish at a given order, and expanded in $epsilon$. Around this core the paper adds a contour optimizer that tries to place contours so that fewer residues occur, an automatic application of the two Barnes lemmas by shifting parameters by $epsilon$ and continuing back to zero, and a numerical pipeline that writes Fortran, calls CUBA and CERNlib, and returns values with quoted errors. The second, negative claim is that numerical evaluation of MB integrals is reliable in the Euclidean region and for massless graphs in the physical region, but not in general for massive graphs there: the factor $(-s)^{-z}$ at $s>0$ contributes $e^{bpi}$, which can cancel the exponential damping of the gamma functions, and the author gives a leading-pole integral whose integrand is not absolutely integrable and whose oscillation frequency, $log(s/4)$, encodes the threshold.

Load-bearing premise

The recursion is assumed to terminate and to return a complete set of residue contributions, and contours are assumed to be arrangeable so that no two are ever crossed at the same time; the paper argues the second point and demonstrates the first on worked examples rather than proving it.

Editorial extensions

If this is right

  • Extracting the singularity structure of a multiloop Feynman integral in $epsilon$ becomes a single package call rather than a per-graph contour analysis.
  • Contour optimization directly reduces the number of residue integrals, and therefore the number of terms that must be expanded and evaluated.
  • Barnes' lemmas become usable automatically after continuation, so some integrations can be done exactly instead of numerically.
  • Numerical MB evaluation provides an independent cross-check of analytic results; the paper reports agreement with earlier analytic work on planar QED boxes and with sector decomposition, and permille-level agreement on the tennis court integral.
  • The paper fixes the boundary of the method's applicability: Euclidean kinematics and massless physical-region graphs are reliable, massive physical-region graphs are not.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The ban on simultaneous contour crossings looks like a bookkeeping convenience rather than a mathematical necessity; a treatment that allowed simultaneous crossings through a multivariate residue or a contour rotation would remove the need for infinitesimal shifts and might shrink the residue set further.
  • Because the troublesome oscillation frequency $log(s/4)$ is itself the threshold logarithm, one could imagine reading threshold expansions directly off that frequency instead of integrating numerically, which would turn the paper's worst numerical case into an analytic one.
  • Since the package emits structured MBint objects with explicit contours, the same output could be piped into a nested-sum reduction or a differential-equation solver, making the continuation step a front end for analytic rather than numeric evaluation.
  • The permille-level agreement reported for the tennis court integral suggests that a Pantis-type tail treatment can make even poorly convergent massive integrals usable as cross-checks, at least in low dimension.
Original abstract

I describe a package written in MATHEMATICA that automatizes typical operations performed during evaluation of Feynman graphs with Mellin-Barnes (MB) techniques. The main procedure allows to analytically continue a MB integral in a given parameter without any intervention from the user and thus to resolve the singularity structure in this parameter. The package can also perform numerical integrations at specified kinematic points, as long as the integrands have satisfactory convergence properties. I demonstrate that, at least in the case of massive graphs in the physical region, the convergence may turn out to be poor, making naive numerical integration of MB integrals unusable. I present possible solutions to this problem, but argue that full automatization in such cases may not be achievable.

The compiled record

The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.

Compiled by classifier-1.13-20260917 via OpenRouter, question bank abstract-p1-20260923, 2026-09-23: 1 request, 4,289 input tokens, $0.0002 reported.

What the abstract itself states, as probabilities: claims first 0.05 · states limitation 0.97 · human participants 0.02 · controlled comparison 0.02 · randomized assignment 0.01 · states quantitative result 0.04 · compares with prior results 0.17 · mentions code or data availability 0.24.

The abstract, sentence by sentence

arXiv listing abstract v2. Results and the main claim are underlined; .

I describe a package written in MATHEMATICA that automatizes typical operations performed during evaluation of Feynman graphs with Mellin-Barnes (MB) techniques. The main procedure allows to analytically continue a MB integral in a given parameter without any intervention from the user and thus to resolve the singularity structure in this parameter. The package can also perform numerical integrations at specified kinematic points, as long as the integrands have satisfactory convergence properties. I demonstrate that, at least in the case of massive graphs in the physical region, the convergence may turn out to be poor, making naive numerical integration of MB integrals unusable. I present possible solutions to this problem, but argue that full automatization in such cases may not be achievable.

background contribution method result interpretation main claim

What the classifier reports for each sentence. The number is the probability it gave the label.
#sentencerolethe authors' own workstates the main conclusionstrength
s0 I describe a package written in MATHEMATICA that automatizes typical operations performed during evaluation of Feynman graphs with Mellin-Barnes (MB) techniques. contribution 0.97 0.92 0.08 asserted 1.00
s1 The main procedure allows to analytically continue a MB integral in a given parameter without any intervention from the user and thus to resolve the singularity structure in this parameter. method 0.61 0.82 0.28 asserted 0.98
s2 The package can also perform numerical integrations at specified kinematic points, as long as the integrands have satisfactory convergence properties. contribution 0.62 0.71 0.05 hedged possibility 0.67
s3 I demonstrate that, at least in the case of massive graphs in the physical region, the convergence may turn out to be poor, making naive numerical integration of MB integrals unusable. result 0.58 0.86 0.17 hedged possibility 0.60
s4 I present possible solutions to this problem, but argue that full automatization in such cases may not be achievable. limitation 0.99 0.80 0.44 hedged possibility 0.96

What kind of paper the classifier reports this is: method or tool 1.00 · evidence from simulation or computation 0.87.

What was examined, and what was not

sources read
  • arXiv listing abstract v2
not examined
  • full text
  • figures and tables
  • references and citing sentences
  • relations to other papers
levels run
  • P abstract
not run
  • S sections
  • T body sentences
  • C citations
  • R cross-paper relations

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. Every number in the measurements table is the authors', checked by code. Record as JSON

Figures
Figure 1. Analytic continuation algorithm. be formalized as in
Figure 2. The B0(s,ms,ms) function, with s = p 2 and ms = m2 . 3.2 Low level routines The routines described in the previous section form the interface. It might happen that the user would like to use the low level routines, which actually perform the calculation. MBresidues[integrand, limit, {fixedVarRules, intVarRules}, options] Performs a single step in the recursive analytic continuation algorithm, i.e. it finds all the residues for a given integral, but does not proceed with the analytic continuation of the resulting integrals. All the arguments and options are the same as in MBcontinue,…
Figure 3. First planar QED box master integral, B1. The wavy lines are massless, whereas the continuous are massive and on-shell
Figure 4. Second planar QED box master integral, B2. The notation is the same as in
Figure 5. The B5l3md2 integral. The notation is the same as in
Figure 6. A regularized version of the B5l3md2 integral. A finite result can be derived from the general representation for B2, when a6 → 0. An interesting example is the B5l3md2 integral,
Figure 7. Massless on-shell non-planar double box integral, NP
Figure 8. Massless on-shell triple box integral, T. obtain at the symmetric point, s = −1, t = −1, u = −1 NP = Γ(3 + 2ǫ)
Figure 9. Real part of the integral Eq. (14) at s = 2. oscillation, log (s/4), encodes the threshold. Further examples seem to confirm that this is a general property. Fortunately, this integral can be evaluated using standard techniques for infinite range oscillatory integrands. With the Pantis’ method [17] Z∞ −∞ db e−iωbf(b) ≃ Z∞ −b0 db e−iωbf(b) + 1 iω e iωb0 f(−b0), (13) setting s = 2 and b0 = 40 the value of the integral in Eq. (11) is ≃ −3.17−0.09i, to be compared to the exact result, which is −π. One would be tempted to assume that the slowly convergent oscillatory be￾haviour can
Figure 10. Tennis court integral, I (3)b 4 , containing a factor of (p + r) 2 . evaluation and the value of every integral is given within MBval[value, error, probability, part] objects, where value, error and probability are given by CUBA, and part is the number of the integral. This provides a primitive means of improving the calculation by tuning only specific integrals, since the integration parameters can be easily changed in the Fortran programs. · Verbose: by default the progress of the integration is printed to the screen. This can be switched off by setting this option to False. I
Figure 11. Regularization of the integrals in the Barnes lemmas by shifting the poles, in such a way that the contour separates the poles. Out[1]:= (ms^z1*(-s)^(-ep - z1)*Gamma[1 - ep - z1]^2*Gamma[-z1]* Gamma[ep + z1])/Gamma[2 - 2*ep - 2*z1] This is, however, a rare situation. Most of the time, the Barnes’ lemmas are applicable to integrals after analytic continuation and expansion. In this case, the procedure generates integrals with contours parallel to the imaginary axis and the contour might not separate the poles of the Γ functions. In such cases one uses various corollaries to the…

Figures reproduced from arXiv:hep-ph/0511200 by the authors.

References
How later papers used it

150 later papers cite it

Evidence: substantial. 146 of the citing papers share no author with it, and 105 of those carry a confident label from Pith's citation classifier.

Pith read 25 of the 221 citing sentences, choosing only those whose citation markers point at this paper alone, and wrote the paraphrases from them. The summary, the counts, the evidence level, the links and the quotes come from code; the summary counts the papers listed below. A later paper is listed under a label when Pith's citation classifier gives one of its sentences that label at probability 0.8 or more (0.95 for support), Pith's reading finds that the citing paper itself did it, and a second check confirms the paraphrase against the sentence. This section lists uses only. It does not list criticism or disagreement, so their absence here says nothing about how the paper has been received. Papers that share an author with this one are listed apart, under "By the same authors". A quote is the citing authors' own sentence, shown when their paper's licence is CC BY, CC BY-SA or CC0. JSON

Uses its method or tool 17 of 108 papers with this label · 2 by the same authors

By the same authors

How the literature uses it

149 arXiv papers cite this in 221 sentences

Each citing sentence is taken from the citing paper's LaTeX source and typed by Pith's classifier. A label appears at probability 0.8 or more (0.95 for support), a contest at 0.7 or more; the quoted words are the citing authors'. How each label was checked · JSON

Papers citing it · 54 sentences without a confident label

Uses its method or tool 108 papers
Compares with it 2 papers
  • For a final accuracy of the order of $10^-6$ the method is as fast as $$ [ref] and $$ [ref] , while for an accuracy of $10^-11$ and higher, turns out to be more than 10 times slower than other two packages; see Tab. [ref].

    New prospects for the numerical calculation of Mellin-Barnes integrals in Minkowskian kinematics · Bottom-Top MBDE approach - optimal steepest descent contours · 0.99

  • An excellent agreement has been obtained when comparing the evaluation of the pentagon from the analytic computation of these integrals, derived with , with the direct numerical evaluation of this Feynman integral with the help of [ref] (as well as with the numerical evaluation of each of the eleven MB integrals separately using [ref] ).

    Multiple Mellin-Barnes integrals with straight contours · Conclusions · 0.86

Cites it as background 21 papers

Discussion

Disagree with the review, or know something it missed? Leave a comment, highlight any passage above to comment on it, or challenge this review formally.