]>PLB135398135398S0370-2693(20)30202-110.1016/j.physletb.2020.135398The AuthorsPhenomenologyFig. 1Matrix element 〈ψ(q)O(−q−p)ψ¯(p)〉 of a bilinear quark operator in momentum space. The black box denotes the operator, and solid lines denote the external quarks.Fig. 1Table 1Numerical cancellation of the 1/ε poles in the O(a3) coefficient of the form factor F1V for Nc = 3, nf = 0, and μ¯=μ.Table 1O(ε−2)O(ε−1)O(1)

(0.08 ± 1.2)⋅10−4(0.68 ± 11.9)⋅10−4286.17 ± 0.12

Bilinear quark operators in the RI/SMOM scheme at three loopsBernd A.Kniehla⁎kniehl@desy.deOleg L.Veretinboleg.veretin@desy.deaII. Institut für Theoretische Physik, Universität Hamburg, Luruper Chaussee 149, 22761 Hamburg, GermanyII. Institut für Theoretische PhysikUniversität HamburgLuruper Chaussee 149Hamburg22761GermanyII. Institut für Theoretische Physik, Universität Hamburg, Luruper Chaussee 149, 22761 Hamburg, GermanybInstitut für Theoretische Physik, Universität Regensburg, Universitätsstraße 31, 93040 Regensburg, GermanyInstitut für Theoretische PhysikUniversität RegensburgUniversitätsstraße 31Regensburg93040GermanyInstitut für Theoretische Physik, Universität Regensburg, Universitätsstraße 31, 93040 Regensburg, Germany⁎Corresponding author.Editor: B. GrinsteinAbstractWe consider the renormalization of the matrix elements of the bilinear quark operators ψ¯ψ, ψ¯γμψ, and ψ¯σμνψ at next-to-next-to-next-to-leading order in QCD perturbation theory at the symmetric subtraction point. This allows us to obtain conversion factors between the MS‾ scheme and the regularization invariant symmetric momentum subtraction (RI/SMOM) scheme. The obtained results can be used to reduce the errors in determinations of quark masses from lattice QCD simulations. The results are given in Landau gauge.KeywordsLattice QCDBilinear quark operatorsMS‾ schemeRegularization invariant symmetric MOM schemeThree-loop approximation1IntroductionThe lattice formulation of quantum chromodynamics (QCD) provides a possibility to estimate long-distance operator matrix elements from first principles using Monte Carlo methods. Many important physical observables can be related to matrix elements of bilinear quark operators of the form Oμ…ν=ψ¯Γμ…νψ, where Γμ…ν is some Dirac structure that can contain covariant derivatives.We start from the following expression in Minkowski space:(1)∫dxdye−iq⋅x−ip⋅y〈ψξ,i(x)Oμ…ν(0)ψ¯ζ,j(y)〉=δijSξξ′(−q)Λξ′ζ′(p,q)Sζ′ζ(p), where ξ,ζ are spinor indices, i,j are color indices in the fundamental representation, S(q) is the quark propagator, and Λ(p,q) is the amputated Green's function, which is shown schematically in Fig. 1.The renormalization of the matrix elements on the lattice is done in some appropriate renormalization scheme. One of the popular schemes is the regularization independent momentum subtraction (RI/MOM) scheme or its variant, the RI′/MOM scheme [1], where the subtraction is done at the momentum configuration p2=q2=−μ2, (p+q)2=0. It has been realized, however, that such a prescription may suffer from a strong sensitivity to infrared (IR) effects. For example, the pseudoscalar current receives contributions from the pseoduscalar-meson pole at (p+q)2=0 and is sensitive to condensate effects of order O(ΛQCD2/μ2). To avoid such problems, the regularization independent symmetric MOM (RI/SMOM) scheme has been suggested in Ref. [2]. The subtraction in this scheme is performed at the symmetric Euclidean point(2)p2=q2=(p+q)2=−μ2,p⋅q=μ22. In Eq. (2), none of the four-momenta is exceptional any more, which provides much better IR behavior for the scheme.To confront lattice simulations with phenomenological analyses, it is necessary to convert the matrix elements to the MS‾ renormalization scheme, which is usually adopted in continuum perturbation theory. The conversion factors for field strengths and masses are known in the RI/MOM scheme through the three-loop order [3,4]. The corresponding matching calculations in the RI/SMOM scheme are more involved. The one-loop results are given in Refs. [2,5]. At the two-loop level, calculations have been done in Refs. [6–8] for the quark currents. The n=2 and n=3 twist-two operators have been considered at the two-loop level in Refs. [9,10]. The two-loop singlet axial-vector current has been considered in Ref. [11]. Recently, in Ref. [12], the conversion of the strong-coupling constant has been evaluated at the three-loop order.The goal of the present work is to evaluate the matching factors between the MS‾ and RI/SMOM schemes for the bilinear quark operators in the three-loop approximation. This paper is organized as follows. In Section 2, we give the underlying definitions and discuss the main steps of our evaluation procedure. In Section 3, we present our results. In Section 4, we conclude with a summary.2EvaluationIn this paper, we consider scalar, vector, and tensor non-singlet bilinear quark operators in QCD,(3)JS=ψ¯ψ,JμV=ψ¯γμψ,JμνT=ψ¯σμνψ, where σμν=i2[γμ,γν], taken at the Euclidean symmetric kinematic point of Eq. (2).We adopt the tensor decomposition used in Ref. [5] and write the amputated Green's functions ΛS,V,T in terms of scalar form factors FjS,V,T=FjS,V,T(p,q) and the relevant tensor structures built from the four-momenta p,q and Dirac γ matrices,(4)ΛS(p,q)=Γ0F1S+1μ2Γ2,pqF2S,(5)ΛμV(p,q)=γμF1V+1μ2p̸pμF2V+1μ2q̸pμF3V+1μ2p̸qμF4V+1μ2q̸qμF5V+1μ2Γ3,μpqF6V,(6)ΛμνT(p,q)=Γ2,μνF1T+1μ2(pμqν−pνqμ)F2T+1μ2(Γ2,μppν−Γ2,νppμ)F3T+1μ2(Γ2,μpqν−Γ2,νpqμ)F4T+1μ2(Γ2,μqpν−Γ2,νqpμ)F5T+1μ2(Γ2,μqqν−Γ2,νqqμ)F6T+1μ4(pμqν−pνqμ)Γ2,pqF7T+1μ2Γ4,μνpqF8T, where Γn denote antisymmetric products of γ matrices with the normalization factor 1/n! included, i.e.(7)Γ0=I,Γ2,ν1ν2=12!(γν1γν2−γν2γν1),etc. In Eqs. (4)–(6), we have also used the shorthand notation pαΓ…α…=Γ…p… for the contraction of a four-momentum with a tensor.The evaluation of the above matrix elements is organized as usual, in two steps: the reduction to master integrals and their evaluation. After the projection and the evaluation of the color and Dirac traces, we first reduce the large number of Feynman integrals with the help of integration-by-parts (IBP) relations [13] to a small set of master integrals. This is done with the help of the computer package FIRE [14]. Besides the IBP relations, we have additional relations arising from the symmetric kinematics of Eq. (2). With these new relations, we can further reduce the number of master integrals. Finally, we can express all the amplitudes in terms of 2 one-loop, 8 two-loop, and 60 three-loop master integrals.Generally, an amplitude can be written as a sum of N master integrals Mj,(8)A=∑j=1Ncj(d)Mj, where the coefficients cj(d) are rational functions of the space-time dimension d and are determined in the course of the reduction procedure.In our case, we have the situation where the master integrals Mj have at most 1/ε3 poles in the ε expansion, while the coefficients cj(d) can have poles up to 1/ε5. This means that some of the three-loop master integrals have to be expanded up to O(ε5) to extract the finite contributions! If the master integrals are evaluated numerically with only restricted accuracy (see the discussion below), this is expected to lead to a significant loss of accuracy in the final renormalized results.Notice, however, that the choice of the master integrals Mj in Eq. (8) is not unique. While the number N of master integrals remains invariant, any N linearly independent integrals can be chosen as a basis. This freedom can be exploited to improve the behavior of the coefficients cj(d) in the limit d→4. In particular, if all coefficients cj(d) have no poles as ε→0, then the set of master integrals Mj represents an ϵ-finite basis [15,16]. Such a basis can always be constructed as long as no restrictions on the choice of master integrals Mj are imposed. In our evaluation, we choose master integrals with positive indices and work with a restricted set of master integrals, namely, the master integrals that are present in the original expressions in Landau gauge before the reduction procedure. With such conditions, we construct a basis that has at most 1/ε poles in the coefficients cj(d).We now turn to the evaluation of the master integrals. In the kinematics of Eq. (2), we have single-scale integrals. The integrals through two loops have been considered previously, in Ref. [17] using the parametric-integral representation and, more systematically, in Ref. [18] using the method of differential equations. We can briefly summarize our knowledge about the analytic structure of the master integrals in the kinematics of Eq. (2) as follows. They can be expressed in terms of harmonic polylogarithms [19] taken at the special point eiπ/6 on the unit circle in the complex plane. At three loops, we have polylogarithms through weight six. The basis of relevant constants up to weight six has been constructed in Ref. [20] using differential equations. These master integrals have been considered recently in Ref. [12].In our evaluation, we compute the master integrals numerically instead. For this purpose, we use the method of sector decomposition [21,22], which is based on the analytic resolution of singularities and the successive numerical integration of the parametric integrals by Monte Carlo methods. For this purpose, we use the implementation of the program package FIESTA [23]. At the three-loop level, we have four-fold to nine-fold parametric integrals resulting usually in several hundreds of so-called sector integrals, which are then evaluated numerically using the program library CUBA [24]. With a typical sample of 108 function calls, we achieve a relative accuracy of order 10−6 for individual master integrals. However, due to large cancellations between different terms in sums like the one in Eq. (8), the resulting relative accuracy is expected to be worse.3Results and discussionIn this section, we present our results for the individual form factors defined in Eqs. (4)–(6). We perform our evaluations in Landau gauge and for the kinematics in Eq. (2). After the renormalization of the matrix elements, the 1/εj poles cancel. Since the master integrals are known numerically with restricted accuracy, such cancellations are only approximate. We demonstrate the cancellation of the ε poles for the vector form factor F1V in Table 1. We observe from Table 1 that the coefficients of the 1/εj poles are suppressed relative to the O(1) term by 4 to 5 orders of magnitude. On the other hand, the absolute error grows from 10−4 for the O(ε−2) term to 10−1 for the O(1) one. We compare our results at the one-loop level with Refs. [2,5] and at the two-loop level with Ref. [8]. We find agreement with these papers, however, with one caveat: if we understand the form factors Σ(j)S,V,T in Refs. [2,8] as being evaluated for the matrix element 〈ψ(q)O(−q−p)ψ¯(p)〉 instead of 〈ψ(p)O(−q−p)ψ¯(q)〉, then we find FjS,V,T=−Σ(j)S,V,T through two loops.We now provide all the form factors in Landau gauge, renormalized in the MS‾ scheme. Our results for the general SU(N) group are listed in the ancillary file, and those for the SU(3) group, with CF=4/3 and CA=3, are presented here:(9)F1S=1+a(0.6455188559544156)+a2(48.48858821140752−6.346872802912313nf)+a3(2396.16(7)−417.872(2)nf+8.6453(1)nf2),(10)F2S=a(1.0417365505286484)+a2(11.166805203854107−0.462994022457177nf)+a3(340.47(8)−65.674(2)nf+0.87117(2)nf2),(11)F1V=1+a(−1.6249301161380183)+a2(−6.1248320773321865−0.2362586437086281nf)+a3(286.17(12)−47.915(2)nf+2.37045(6)nf2),(12)F2V=a(−1.4720824544982587)+a2(−18.797490826183573+1.3299518483852129nf)+a3(−658.77(12)+120.165(2)nf−2.13077(4)nf2),(13)F3V=a(−1.7777777777777777)+a2(−44.38058543042832+2.8641975308641974nf)+a3(−1748.85(6)+276.890(3)nf−6.0501(3)nf2),(14)F4V=F3V,(15)F5V=F2V,(16)F6V=a(2.083473101057297)+a2(39.7873695861749−3.009461145971651nf)+a3(1501.31(4)−269.099(2)nf+6.37227(3)nf2),(17)F1T=1+a(−0.06232529034504568)+a2(−17.00995398724325+1.6001145201364706nf)+a3(−482.80(7)+57.814(2)nf−0.36852(5)nf2),(18)F2T=a(−3.125209651585945)+a2(−76.20220922856834+5.555928269486126nf)+a3(−3002.31(16)+488.628(7)nf−11.87338(5)nf2),(19)F3T=2F4T,(20)F4T=a(0.15284766163975952)+a2(0.9097900360378999−0.16983073515528835nf)+a3(−16.54(9)−6.074(2)nf+0.23416(3)nf2),(21)F5T=F4T,(22)F6T=2F4T,(23)F7T=4F4T,(24)F8T=−F2T, where a=αs/(4π) and the 't Hooft mass μ¯ is taken to be equal to the subtraction point μ in the SMOM scheme. We retain the error intervals only for the three-loop contributions, while the one- and two-loop coefficients can be obtained from Ref. [8] in analytic form.From the above results, we also obtain the conversion factors between the MS‾ and RI/SMOM schemes for the quark field and mass, Cq and Cm, namely,(25)ψRMS‾=CqRI/SMOMψRRI/SMOM,mRMS‾=CmRI/SMOMmRRI/SMOM. The result for CqRI/SMOM is known, since it was shown in Ref. [2] that CqRI/SMOM=CqRI′/MOM as a result of a Ward–Takahashi identity. The result for CmRI/SMOM at the three-loop order is new and reads(26)CqRI/SMOM=1+a2(−25.46420605097376+2.333333333333333nf)+a3(−1489.980500+246.442650nf−6.460905nf2),(27)CmRI/SMOM=1+a(−0.6455188559544155533)+a2(−22.607687567041+4.0135394695790nf)+a3(−860.28(7)+164.742(2)nf−2.18440(10)nf2).If we use, for an estimation, αs/π∼0.1, which approximately corresponds to μ=2 GeV, then we obtain for the mass conversion factor, with nf=3,(28)CmRI/SMOM=1−0.0161−0.0066−0.0060+…. Thus, the three-loop contribution is of the same sign and size as the two-loop one.4ConclusionIn this paper, we have established a framework for the evaluation of the bilinear quark operators of QCD in the MOM scheme at the symmetric kinematic point. The results have been obtained for the scalar, vector, and tensor currents with general tensor and Dirac structures in the three-loop approximation. The conversion factors between the MS‾ and RI/SMOM schemes have been given to three loops as well. The three-loop corrections appear to be comparable in size with the two-loop ones.Note added in proofAfter submission, a preprint [25] has appeared in which our Eq. (27) is confirmed and represented in analytic form.Declaration of Competing InterestThe authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.AcknowledgementsThis work was supported in part by DFG Research Unit FOR 2926 through Grant No. 409651613. O.L.V. is grateful to V. Braun and M. Göckeler for fruitful discussions and to the University of Hamburg for the warm hospitality.References[1]G.MartinelliC.PittoriC.T.SachrajdaM.TestaA.VladikasA general method for non-perturbative renormalization of lattice operatorsNucl. Phys. B44519958110810.1016/0550-3213(95)00126-DarXiv:hep-lat/9411010G. Martinelli, C. Pittori, C. T. Sachrajda, M. Testa, A. Vladikas, A General method for nonperturbative renormalization of lattice operators, Nucl. Phys. B445 (1995) 81–108. arXiv:hep-lat/9411010, doi:10.1016/0550-3213(95)00126-D.[2]C.SturmY.AokiN.H.ChristT.IzubuchiC.T.C.SachrajdaA.SoniRenormalization of quark bilinear operators in a momentum-subtraction scheme with a nonexceptional subtraction pointPhys. Rev. D80200901450110.1103/PhysRevD.80.014501arXiv:0901.2599 [hep-ph]C. Sturm, Y. Aoki, N. H. Christ, T. Izubuchi, C. T. C. Sachrajda, A. Soni, Renormalization of quark bilinear operators in a momentum-subtraction scheme with a nonexceptional subtraction point, Phys. Rev. D80 (2009) 014501. arXiv:0901.2599, doi:10.1103/PhysRevD.80.014501.[3]K.G.ChetyrkinA.RéteyRenormalization and running of quark mass and field in the regularization invariant and MS‾ schemes at three and four loopsNucl. Phys. B583200033410.1016/S0550-3213(00)00331-XarXiv:hep-ph/9910332K. G. Chetyrkin, A. Retey, Renormalization and running of quark mass and field in the regularization invariant and MS-bar schemes at three loops and four loops, Nucl. Phys. B583 (2000) 3–34. arXiv:hep-ph/9910332, doi:10.1016/S0550-3213(00)00331-X.[4]J.A.GraceyThree loop anomalous dimension of non-singlet quark currents in the RI' schemeNucl. Phys. B662200324727810.1016/S0550-3213(03)00335-3arXiv:hep-ph/0304113J. A. Gracey, Three loop anomalous dimension of nonsinglet quark currents in the RI-prime scheme, Nucl. Phys. B662 (2003) 247–278. arXiv:hep-ph/0304113, doi:10.1016/S0550-3213(03)00335-3.[5]J.A.GraceyRI'/SMOM scheme amplitudes for deep inelastic scattering operators at one loop in QCDPhys. Rev. D83201105402410.1103/PhysRevD.83.054024arXiv:1009.3895 [hep-ph]J. A. Gracey, RI'/SMOM scheme amplitudes for deep inelastic scattering operators at one loop in QCD, Phys. Rev. D83 (2011) 054024. arXiv:1009.3895, doi:10.1103/PhysRevD.83.054024.[6]L.G.AlmeidaC.SturmTwo-loop matching factors for light quark masses and three-loop mass anomalous dimensions in the regularization invariant symmetric momentum-subtraction schemesPhys. Rev. D82201005401710.1103/PhysRevD.82.054017arXiv:1004.4613 [hep-ph]L. G. Almeida, C. Sturm, Two-loop matching factors for light quark masses and three-loop mass anomalous dimensions in the RI/SMOM schemes, Phys. Rev. D82 (2010) 054017. arXiv:1004.4613, doi:10.1103/PhysRevD.82.054017.[7]M.GorbahnS.JägerPrecise MS‾ light-quark masses from lattice QCD in the regularization invariant symmetric momentum-subtraction schemePhys. Rev. D82201011400110.1103/PhysRevD.82.114001arXiv:1004.3997 [hep-ph]M. Gorbahn, S. Jager, Precise MS-bar light-quark masses from lattice QCD in the RI/SMOM scheme, Phys. Rev. D82 (2010) 114001. arXiv:1004.3997, doi:10.1103/PhysRevD.82.114001.[8]J.A.GraceyRI'/SMOM scheme amplitudes for quark currents at two loopsEur. Phys. J. C712011156710.1140/epjc/s10052-011-1567-8arXiv:1101.5266 [hep-ph]J. A. Gracey, RI'/SMOM scheme amplitudes for quark currents at two loops, Eur. Phys. J. C71 (2011) 1567. arXiv:1101.5266, doi:10.1140/epjc/s10052-011-1567-8.[9]J.A.GraceyTwo loop renormalization of the n=2 Wilson operator in the RI'/SMOM schemeJ. High Energy Phys.03201110910.1007/JHEP03(2011)109arXiv:1103.2055 [hep-ph]J. A. Gracey, Two loop renormalization of the n = 2 Wilson operator in the RI'/SMOM scheme, JHEP 03 (2011) 109. arXiv:1103.2055, doi:10.1007/JHEP03(2011)109.[10]J.A.GraceyAmplitudes for the n=3 moment of the Wilson operator at two loops in the RI/'SMOM schemePhys. Rev. D84201101600210.1103/PhysRevD.84.016002arXiv:1105.2138 [hep-ph]J. A. Gracey, Amplitudes for the n = 3 moment of the Wilson operator at two loops in the RI/'SMOM scheme, Phys. Rev. D84 (2011) 016002. arXiv:1105.2138, doi:10.1103/PhysRevD.84.016002.[11]J.A.GraceySymmetric point flavour singlet axial vector current renormalization at two loopsarXiv:2001.11282 [hep-ph]2020J. A. Gracey, Symmetric point flavour singlet axial vector current renormalization at two loops (2020). arXiv:2001.11282.[12]A.BednyakovA.PikelnerFour-loop QCD MOM beta functions from the three-loop vertices at the symmetric pointarXiv:2002.02875 [hep-ph]2020A. Bednyakov, A. Pikelner, Four-loop QCD MOM beta functions from the three-loop vertices at the symmetric point (2020). arXiv:2002.02875.[13]K.G.ChetyrkinF.V.TkachovIntegration by parts: The algorithm to calculate β-functions in 4 loopsNucl. Phys. B192198115920410.1016/0550-3213(81)90199-1K. G. Chetyrkin, F. V. Tkachov, Integration by Parts: The Algorithm to Calculate beta Functions in 4 Loops, Nucl. Phys. B192 (1981) 159–204. doi:10.1016/0550-3213(81)90199-1.[14]A.V.SmirnovFIRE5: A C++ implementation of Feynman Integral REductionComput. Phys. Commun.189201518219110.1016/j.cpc.2014.11.024arXiv:1408.2372 [hep-ph]A. V. Smirnov, FIRE5: a C++ implementation of Feynman Integral REduction, Comput. Phys. Commun. 189 (2015) 182–191. arXiv:1408.2372, doi:10.1016/j.cpc.2014.11.024.[15]K.G.ChetyrkinM.FaisstC.SturmM.Tentyukovϵ-finite basis of master integrals for the integration-by-parts methodNucl. Phys. B742200620822910.1016/j.nuclphysb.2006.02.030arXiv:hep-ph/0601165K. G. Chetyrkin, M. Faisst, C. Sturm, M. Tentyukov, epsilon-finite basis of master integrals for the integration-by-parts method, Nucl. Phys. B742 (2006) 208–229. arXiv:hep-ph/0601165, doi:10.1016/j.nuclphysb.2006.02.030.[16]M.FaisstP.MaierhöferC.SturmStandard and ϵ-finite master integrals for the ρ-parameterNucl. Phys. B766200724626810.1016/j.nuclphysb.2006.12.014arXiv:hep-ph/0611244M. Faisst, P. Maierhoefer, C. Sturm, Standard and epsilon-finite Master Integrals for the rho-Parameter, Nucl. Phys. B766 (2007) 246–268. arXiv:hep-ph/0611244, doi:10.1016/j.nuclphysb.2006.12.014.[17]N.I.UssyukinaA.I.DavydychevNew results for two-loop off-shell three-point diagramsPhys. Lett. B332199415916710.1016/0370-2693(94)90874-5arXiv:hep-ph/9402223N. I. Usyukina, A. I. Davydychev, New results for two loop off-shell three point diagrams, Phys. Lett. B332 (1994) 159–167. arXiv:hep-ph/9402223, doi:10.1016/0370-2693(94)90874-5.[18]T.G.BirthwrightE.W.N.GloverP.MarquardMaster integrals for massless two-loop vertex diagrams with three offshell legsJ. High Energy Phys.09200404210.1088/1126-6708/2004/09/042arXiv:hep-ph/0407343T. G. Birthwright, E. W. N. Glover, P. Marquard, Master integrals for massless two-loop vertex diagrams with three offshell legs, JHEP 09 (2004) 042. arXiv:hep-ph/0407343, doi:10.1088/1126-6708/2004/09/042.[19]E.RemiddiJ.A.M.VermaserenHarmonic polylogarithmsInt. J. Mod. Phys. A15200072575410.1142/S0217751X00000367arXiv:hep-ph/9905237E. Remiddi, J. A. M. Vermaseren, Harmonic polylogarithms, Int. J. Mod. Phys. A15 (2000) 725–754. arXiv:hep-ph/9905237, doi:10.1142/S0217751X00000367.[20]B.A.KniehlA.F.PikelnerO.L.VeretinThree-loop massive tadpoles and polylogarithms through weight sixJ. High Energy Phys.08201702410.1007/JHEP08(2017)024arXiv:1705.05136 [hep-ph]B. A. Kniehl, A. F. Pikelner, O. L. Veretin, Three-loop massive tadpoles and polylogarithms through weight six, JHEP 08 (2017) 024. arXiv:1705.05136, doi:10.1007/JHEP08(2017)024.[21]T.BinothG.HeinrichAn automatized algorithm to compute infrared divergent multi-loop integralsNucl. Phys. B585200074175910.1016/S0550-3213(00)00429-6arXiv:hep-ph/0004013T. Binoth, G. Heinrich, An automatized algorithm to compute infrared divergent multiloop integrals, Nucl. Phys. B585 (2000) 741–759. arXiv:hep-ph/0004013, doi:10.1016/S0550-3213(00)00429-6.[22]T.BinothG.HeinrichNumerical evaluation of multi-loop integrals by sector decompositionNucl. Phys. B680200437538810.1016/j.nuclphysb.2003.12.023arXiv:hep-ph/0305234T. Binoth, G. Heinrich, Numerical evaluation of multiloop integrals by sector decomposition, Nucl. Phys. B680 (2004) 375–388. arXiv:hep-ph/0305234, doi:10.1016/j.nuclphysb.2003.12.023.[23]A.V.SmirnovFIESTA4: Optimized Feynman integral calculations with GPU supportComput. Phys. Commun.204201618919910.1016/j.cpc.2016.03.013arXiv:1511.03614 [hep-ph]A. V. Smirnov, FIESTA4: Optimized Feynman integral calculations with GPU support, Comput. Phys. Commun. 204 (2016) 189–199. arXiv:1511.03614, doi:10.1016/j.cpc.2016.03.013.[24]T.HahnCuba—a library for multidimensional numerical integrationComput. Phys. Commun.1682005789510.1016/j.cpc.2005.01.010arXiv:hep-ph/0404043T. Hahn, CUBA: A Library for multidimensional numerical integration, Comput. Phys. Commun. 168 (2005) 78–95. arXiv:hep-ph/0404043, doi:10.1016/j.cpc.2005.01.010.[25]A.BednyakovA.PikelnerQuark masses: NNLO bridge from RI/SMOM to MS‾ schemearXiv:2002.12758 [hep-ph]2020A. Bednyakov, A. Pikelner, Quark masses: NNLO bridge from RI/SMOM to MS‾ scheme, arXiv:2002.12758 [hep-ph], 2020.