DESY 16-020 ISSN 0418-9833 January 2016 6 Counting master integrals: Integration by parts vs. 1 0 functional equations 2 n Bernd A. Kniehl∗, Oleg V. Tarasov1 a J II. Institute fu¨r Theoretische Physik, Universita¨t Hamburg, Luruper Chaussee 149, 22761 0 Hamburg, Germany 3 ] h t - p e Abstract h [ We illustrate the usefulness offunctionalequations inestablishingrelationships 1 betweenmasterintegralsundertheintegration-by-partsreductionprocedureby v 5 considering a certain two-loop propagator-type diagram as an example. 1 1 Keywords: Two-loop sunset diagram, Recurrence relations, Functional 0 0 equations, Multiloop calculations . 2 PACS: 02.30.Gp, 02.30.Lt, 11.15.Bt, 12.38.Bx 0 6 1 An adequate theoretical interpretation of the increasingly precise data col- : v lected by the experiments at the CERN Large Hadron Collider and elsewhere i X necessitates advanced technologies for the calculation of radiative corrections, r a which typically depend on several different mass scales. Feynman diagrams involving quantum loops may be reduced to so-called master integrals via ded- icated algorithms, such as integration by parts (IBP) [1, 2]. The evaluation of the master integrals often turns out to be a bottleneck of the entire theoretical ∗Correspondingauthor Email addresses: [email protected] (BerndA.Kniehl),[email protected] (OlegV. Tarasov) 1On leave of absence from Joint Institute for Nuclear Research, 141980 Dubna (Moscow Region),Russia. Preprint submitted toJournal of LATEX Templates February 2, 2016 analysis, the more if many different mass scales are involved. Any method to reduce the number of master integrals of a given set of Feynman diagrams is, therefore, highly welcome. Recently, relationships between master integrals of the two-loop sunset diagram were found in Refs. [3, 4]. In the present paper, a new relationship of this type will be presented, which is found using functional equations [5]. Thederivationoffunctionalequationsforintegralswithtwoandmoreloops is much more complicated than in the one-loop case. In this following, we consider two-loop propagator-type integrals. Two-loop integrals differ by the number of internal lines. According to the algorithm of Refs. [5, 6], functional equations may be obtained from recurrence relations connecting two-loop inte- grals with different numbers of lines. The most complicated integrals in such a functionalequationmaybeeliminatedbyanappropriatechoiceoffour-momenta and masses. 2 1 3 4 Figure1: Two-loopdiagramcorrespondingtointegralV(d) . 1111 Let us consider the following two-loop propagator-type integral with four internal lines: 1 V(d) (m2,m2,m2,m2;q2)= ν1ν2ν3ν4 1 2 3 4 iπd/2 2 (cid:0)ddk (cid:1)ddk × 1 2 ,(1) Z Z [(k1−k2)2−m21]ν1[k22−m22]ν2[(k1−q)2−m23]ν3[(k2−q)2−m24]ν4 where d is the space-time dimension. The Feynman diagram corresponding to this integralis showninFig.1. With the aidofgeneralizedrecurrencerelations giveninRef.[7],theintegralV(d) witharbitraryintegersν maybereduced ν1ν2ν3ν4 j 2 to the integral V(d) , four two-loop integrals with three lines of the type 1111 J(d) (m2,m2,m2;q2) ν1ν2ν3 1 2 3 1 ddk ddk = 1 2 , (2) iπd/2 2 Z Z [k12−m21]ν1[(k1−k2)2−m22]ν2[(k2−q)2−m23]ν3 (cid:0) (cid:1) the product of a one-loop propagator-type integral, 1 ddk I(d)(m2,m2;q2)= 1 , (3) 2 1 2 iπd/2 Z (k2−m2)[(k −q)2−m2] 1 1 1 2 with masses m2 and m2, times a one-loop vacuum-type integral, 2 3 1 ddk T (m2)= 1 , (4) a 1 iπd/2 Z k2−m2 1 1 and products of one-loop vacuum-type integrals with different masses. In Ref. [7], severalgeneralized recurrence relations for the integral V(d) ν1ν2ν3ν4 were presented. One of these relations, namely the one in Eq. (55) therein, reads: (d−3)u u ∆ +(d−3)u u ∆ +(d−4)∆ ∆ m2V(d) = 624 246 134 314 134 246 134 246 4 1112 2∆ ∆ 134 246 u ∆ −u ∆ ×V(d) + 624 134 314 246m2J(d)(m2,m2,m2;q2) 1111 ∆ ∆ 3 112 1 2 3 134 246 u ∆ −u ∆ + 624 134 134 246m2J(d)(m2,m2,m2;q2) ∆ ∆ 1 211 1 2 3 134 246 2m2(q2−m2) + 2 2 J(d)(m2,m2,m2,q2) ∆ 121 1 2 3 246 u (3d−8) − 624 J(d)(m2,m2,m2;q2) 2∆ 111 1 2 3 246 (d−2)u + 624J (m2,m2,m2;0) 2∆ 111 1 3 4 246 (d−2) + u T(d)(m2)+u T(d)(m2) I(d)(m2,m2;q2). (5) 2∆134 h 314 1 3 134 1 1 i 2 2 4 whereu =m2−m2−m2 and∆ =−u (u +u )−u u . According ijk i j k ijk ijk jik kij jik kij to the algorithm of Ref. [5] to obtain functional equation, one has to eliminate from Eq. (5) the integrals V(d) and V(d) by an appropriate choice of four- 1111 1112 momentum and masses. For m = 0, the left-hand side of Eq. (5) vanishes, so 4 that we obtain an expression for the integral V(d) in terms of integrals with 1111 3 lesser numbers of lines, namely, 2m2(q2+u ) V(d) = 1 312 J (m2,m2,m2;q2) 1111 (d−2)(m2−m2)(q2−m2) 211 1 2 3 3 1 2 2m2(q2+u ) − 3 123 J (m2,m2,m2;q2) (d−2)(m2−m2)(q2−m2) 112 1 2 3 3 1 2 4m2 + 2 J (m2,m2,m2;q2) (d−2)(q2−m2) 121 1 2 3 2 (3d−8) − J (m2,m2,m2;q2) (d−2)(q2−m2) 111 1 2 3 2 1 + J (m2,m2,0;0) q2−m2 111 1 3 2 1 + T(d)(m2)−T(d)(m2) I(d)(m2,0;q2). (6) m2−m2 h 1 1 1 3 i 2 2 1 3 After multiplying Eq. (6) with the factor q2 −m2 and then setting q2 = m2, 2 2 the contribution proportional to the integral V(d) drops out, and obtain the 1111 following relationship: 0 = 2m2J(d)(m2,m2,m2;m2)+4m2J(d)(m2,m2,m2;m2) 1 211 1 2 3 2 2 121 1 2 3 2 +2m2J(d)(m2,m2,m2;m2)−(3d−8)J(d)(m2,m2,m2;m2) 3 112 1 2 3 2 111 1 2 3 2 +(d−2)J(d)(m2,m2,0;0). (7) 111 1 3 Equation (7) connects two-loop propagator-type integrals with different kine- matics. The analytic expression for the integral J(d)(m2,m2,m2;0) in terms 111 1 2 3 of the Gauss hypergeometric function F presented in Ref. [8] is considerably 2 1 simplerthanthe analyticexpressionsforthe integralsJ(d) andJ(d) withexter- 111 211 nal momentum square being different from zero. It is interesting to notice that the Cayley–Menger determinant D = [q2−(m +m +m )2][q2−(m −m +m )2] 123 1 2 3 1 2 3 ×[q2−(m +m −m )2][q2−(m −m −m )2] (8) 1 2 3 1 2 3 for this kinematics is different from zero: D | =(m2−m2)2[(m2−m2)2+8m2(2m2−m2−m2)]. (9) 123 q2=m22 1 3 1 3 2 2 1 3 Thus,Eq.(7)isaclearillustrationthatthenumberofnontrivialbasisintegrals, aspredictedby IBP,mayre reducednotonlyif D =0 orone massis zeroas 123 4 wasobservedinRef.[7],butalsoforothervaluesoffour-momentummomentum squareandmasses. Onepossibleinterpretationisthatthetotalnumberofbasis integralsarisingfromthe IBPreductionofthe integralJ(d) (m2,m2,m2;m2) ν1ν2ν3 1 2 3 2 with arbitrary integer powers of propagators remains the same, but that one nontrivial integral may be replaced by simpler one. For the particular case when m = 0 and m2 = m2 = m2, the reduction 1 2 3 of the number of basis integrals was observed in Ref. [9]. For this kinematics, Eq. (7) yields 3d−8 d−2 J(d)(m2,0,m2;m2)= J(d)(m2,0,m2;m2)− J(d)(m2,0,0;0), (10) 211 6m2 111 6m2 111 so that, instead of two nontrivial integrals, only one nontrivial basis integral, J(d)(m2,0,m2;m2), remains. 111 Puttingm =0inEq.(7),werecoverEq.(9)inRef.[3],whichwasobtained 3 there as a special case via differential reduction [10, 11, 12, 13, 14, 15, 16, 17, 18,19,20, 21, 22]. Another generalizationof Eq.(9) inRef. [3] wasobtained in Ref. [4] using IBP in connection with an effective propagator mass [23]. We would like mention that Eq. (7) connects integrals of different mass as- signments. Such integrals may arise from rather different Feynman diagrams. Relationships of this type may be very useful, e.g., for proving the gauge inde- pendence of radiative corrections to physical observables. In conclusion, functional equations [5, 6] provide a powerfultool for disclos- inghiddenrelationshipsbetweenwhatappeartobemasterintegralsuponstan- dard applications of the IBP reduction procedure [1, 2]. Similar relationships havepreviouslybeenrevealedusingdifferentialreduction[3]andanonstandard variantoftheIBPreductionprocedureimplementedwithpropagatormassesto be integrated over [4]. Acknowledgments ThisworkwassupportedbytheGermanResearchFoundationDFGthrough the Collaborative Research Center SFB 676 Particles, Strings and the Early Universe: the Structure of Matter and Space-Time. 5 References [1] F. Tkachov, A theorem on analytical calculability of 4-loop renormalization group functions, Phys.Lett. B100 (1981) 65–68. doi:10.1016/0370-2693(81)90288-4. [2] K. Chetyrkin, F. Tkachov, Integration by parts: The algorithm to calculate β-functions in 4 loops, Nucl.Phys. B192 (1981) 159–204. doi:10.1016/0550-3213(81)90199-1. [3] M. Yu. Kalmykov, B. A. Kniehl, Counting master integrals: Integra- tion by parts vs. differential reduction, Phys. Lett. B702 (2011) 268–271. arXiv:1105.5319, doi:10.1016/j.physletb.2011.06.094. [4] B. A. Kniehl, A. V. Kotikov, Counting master integrals: Integration-by- parts procedure with effective mass, Phys. Lett. B712 (2012) 233–234. arXiv:1202.2242, doi:10.1016/j.physletb.2012.04.071. [5] O. Tarasov, New relationships between Feynman inte- grals, Phys.Lett. B670 (2008) 67–72. arXiv:0809.3028, doi:10.1016/j.physletb.2008.10.021. [6] O. Tarasov, Functional equations for Feynman in- tegrals, Phys.Part.Nucl.Lett. 8 (2011) 419–427. doi:10.1134/S1547477111050219. [7] O. Tarasov, Generalized recurrence relations for two-loop propaga- tor integrals with arbitrary masses, Nucl.Phys. B502 (1997) 455–482. arXiv:hep-ph/9703319, doi:10.1016/S0550-3213(97)00376-3. [8] A. Davydychev, J. Tausk, Two-loop self-energy diagrams with different masses and the momentum expansion, Nucl.Phys. B397 (1993) 123–142. doi:10.1016/0550-3213(93)90338-P. [9] A. Davydychev, M. Kalmykov, New results for the ǫ-expansion of certain one-,two-andthree-loopFeynmandiagrams,Nucl.Phys.B605(2001)266– 318. arXiv:hep-th/0012189, doi:10.1016/S0550-3213(01)00095-5. 6 [10] M. Yu. Kalmykov, Gauss hypergeometric function: reduc- tion, ǫ-expansion for integer/half-integer parameters and Feyn- man diagrams, JHEP 04 (2006) 056. arXiv:hep-th/0602028, doi:10.1088/1126-6708/2006/04/056. [11] M.Kalmykov,V.V.Bytev,B.A.Kniehl,B.F.L.Ward,S.A.Yost,Feyn- manDiagrams,DifferentialReduction,andHypergeometricFunctions,PoS ACAT08 (2008) 125. arXiv:0901.4716. [12] V. V. Bytev, M. Kalmykov, B. A. Kniehl, B. F. L. Ward, S. A. Yost, Differential Reduction Algorithms for Hypergeometric Functions Applied to Feynman Diagram Calculation, arXiv:0902.1352. [13] V. V. Bytev, M. Yu. Kalmykov, B. A. Kniehl, Differential reduction of generalized hypergeometric functions from Feynman diagrams: One- variable case, Nucl. Phys. B836 (2010) 129–170. arXiv:0904.0214, doi:10.1016/j.nuclphysb.2010.03.025. [14] M. Yu. Kalmykov, B. A. Kniehl, All-order ǫ expansions of hypergeo- metric functions of one variable, Phys. Part. Nucl. 41 (2010) 942–945. arXiv:1003.1965, doi:10.1134/S1063779610060250. [15] M. Yu. Kalmykov, B. A. Kniehl, “Sixth root of unity” and Feyn- man diagrams: hypergeometric function approach point of view, Nucl. Phys. B (Proc. Suppl.) 205-206 (2010) 129–134. arXiv:1007.2373, doi:10.1016/j.nuclphysbps.2010.08.031. [16] S.A.Yost,V.V.Bytev,M.Yu.Kalmykov,B.A.Kniehl,B.F.L.Ward,Dif- ferential Reduction Techniques for the Evaluation of Feynman Diagrams, PoS ICHEP2010 (2010) 135. arXiv:1101.2348. [17] V. V. Bytev, M. Yu. Kalmykov, B. A. Kniehl, HYPERDIRE, HYPERge- ometric functions DIfferential REduction: MATHEMATICA-based pack- ages for differential reduction of generalized hypergeometric functions 7 pFp−1, F1, F2, F3, F4, Comput. Phys. Commun. 184 (2013) 2332–2342. arXiv:1105.3565, doi:10.1016/j.cpc.2013.05.009. [18] S.A.Yost,V.V.Bytev,M.Yu.Kalmykov,B.A.Kniehl,B.F.L.Ward,The Epsilon Expansion of Feynman Diagrams via Hypergeometric Functions and Differential Reduction, arXiv:1110.0210. [19] M. Yu. Kalmykov, B. A. Kniehl, Mellin–Barnes representations of Feyn- man diagrams, linear systems of differential equations, and polyno- mial solutions, Phys. Lett. B714 (2012) 103–109. arXiv:1205.1697, doi:10.1016/j.physletb.2012.06.045. [20] V. V. Bytev, M. Kalmykov, B. A. Kniehl, When epsilon-expansion of hy- pergeometric functions is expressible in terms of multiple polylogarithms: the two-variables examples, PoS LL2012 (2012) 029. arXiv:1212.4719. [21] V. V. Bytev, M. Yu. Kalmykov, S.-O. Moch, HYPERgeometric func- tions DIfferential REduction (HYPERDIRE): MATHEMATICA based packages for differential reduction of generalized hypergeometric func- tions: F and F Horn-type hypergeometric functions of three vari- D S ables, Comput. Phys.Commun. 185 (2014)3041–3058. arXiv:1312.5777, doi:10.1016/j.cpc.2014.07.014. [22] V.V.Bytev,B.A.Kniehl,HYPERDIREHYPERgeometricfunctions DIf- ferential REduction: Mathematica-based packages for the differential re- duction of generalized hypergeometric functions: Horn-type hypergeomet- ricfunctionsoftwovariables,Comput.Phys.Commun.189(2014)128–154. arXiv:1309.2806, doi:10.1016/j.cpc.2014.11.022. [23] B. A. Kniehl, A. V. Kotikov, Calculating four-loop tadpoles with one non-zero mass, Phys.Lett. B638 (2006) 531–537. arXiv:hep-ph/0508238, doi:10.1016/j.physletb.2006.04.057. 8