IFIC/15-71 January2016 Loop-tree duality and quantum field theory in four dimensions 6 1 0 2 Germán F. R. Sborlini∗a,b n a aInstitutodeFísicaCorpuscular,UniversitatdeValència–ConsejoSuperiordeInvestigaciones J Científicas,ParcCientífic,E-46980Paterna,Valencia,Spain. 8 1 bDepartamentodeFísicaandIFIBA,FCEyN,UniversidaddeBuenosAires,(1428)Pabellón1 CiudadUniversitaria,CapitalFederal,Argentina. ] h p E-mail: [email protected] - p e h Loop-tree duality allows to express virtual contributions in terms of phase-space integrals, thus [ leadingtoadirectcomparisonwithrealradiationterms. Inthistalk, wereviewthebasisofthe 1 method and describe its application to regularize Feynman integrals. Performing an integrand- v levelcombinationofrealandvirtualterms, weobtainfinitecontributionsthatcanbecomputed 4 3 infour-dimensions. Moreover,thismethodprovidesanaturalphysicalinterpretationofinfrared 6 singularities,theiroriginandthewaythattheycancelinthecompletecomputation. 4 0 . 1 0 6 1 : v i X r a 12thInternationalSymposiumonRadiativeCorrections(Radcor2015)andLoopFestXIV(Radiative CorrectionsfortheLHCandFutureColliders) 15-19June,2015 UCLADepartmentofPhysics&Astronomy,LosAngeles,USA ∗Speaker. (cid:13)c Copyrightownedbytheauthor(s)underthetermsoftheCreativeCommons Attribution-NonCommercial-NoDerivatives4.0InternationalLicense(CCBY-NC-ND4.0). http://pos.sissa.it/ LTDandquantumfieldtheoryin4D GermánF.R.Sborlini 1. Introduction The knowledge of higher-order corrections is crucial for the discovery and characterization of potential new-physics signals. From the theoretical point of view, this is a very challenging task due to the increasing computational complexity of multi-loop multi-leg processes. In fact, the presence of singularities (or ill-defined expressions) forces the application of regularization methodstoobtainfiniteresults,suchasdimensionalregularization(DREG)[1,2,3,4]. Thereare differentkindofsingularities; • ultraviolet(UV),associatedwiththehigh-energybehaviorofthetheory; • infrared(IR),relatedwiththepresenceofdegenerateconfigurationsinthelow-energylimit; • andthresholdsingularities,whichmayappearwhenvirtualparticlesareproducedon-shell; as well as other spurious or non-physical divergences. In the context of DREG, threshold singu- larities do not introduce ε-poles because they are integrable singularities; however, it is crucial to introduce a proper prescription to deal with them. On the other hand, renormalization success- fullycures UVdivergences byapplyinga well-knownand systematicprocedure. For theIRcase, it is also possible to achieve a finite result when IR-safe observables are considered. Only under that assumption, Kinoshita-Lee-Nauenberg (KLN) [5] theorem guarantees the cancellation of IR singularities among all the possible degenerate configurations associated with the same final state observable. Inparticular,thisinvolvestakingintoaccountrealandvirtualcorrections,whichdiffer inthenumberofexternalparticles. In this article, we focus in the treatment of IR singularities through a new method, which is based in the loop-tree duality (LTD) theorem [6, 7, 8]. In the standard approach, which is based inthesubtractionmethod[9,10,11,12,13,14,15,16,17],theIR-singularbehaviorofscattering amplitudesisexploitedtowritelocalcounter-termswhichcouldbeintegratedanalytically. Thein- tegrated counter-terms are combined with the virtual contributions, whilst their unintegrated form is used to perform a local regularization of the real terms. Contrary to this, the main idea behind LTDistorewritevirtualamplitudesasrealradiationterms,andperformanintegrandlevelcombi- nation with the real contribution. In this way, dual amplitudes originated from the virtual part act directlyascounter-termsforthereal-emissionamplitudes. Theoutlineofthisarticleisthefollowing. InSection2wereviewthebasicideasbehindLTD, introducing suitable notation. Then, in Section 3, we study the divergent structure of a triangle Feynman integral. In particular, we prove that IR poles are originated in a compact region of the integration space. After that, we explain the implementation of a fully inclusive cross-section computationatnext-to-leadingorder(NLO)inSection4. Wefocusthediscussioninthereal-virtual momenta mapping, which plays a crucial role to achieve the IR cancellation. Some details about thenumericalimplementationaregiven,emphasizingthepossibilityofobtainingfour-dimensional representations at integrand level. Finally, in Section 5 we present the conclusions and discuss aboutpossibleextensionsoftheseideas. 2 LTDandquantumfieldtheoryin4D GermánF.R.Sborlini 2. BasicsofLTD TherealandvirtualcontributionssharethesamekindofIRsingularities,inspiteofinvolving a different number of final-state particles. In LTD loop integrals are related with phase-space in- tegrals of tree-level objects called dual integrals [6]. In any relativistict, local and unitary theory, this property is directly translated to scattering amplitudes, and virtual amplitudes become expre- sible in terms of dual amplitudes. In this article, we analyze one-loop corrections, but these ideas arenaturallyextendedtohigher-loops[6,18,19,20,21]. Let’sconsideranN-legscalarone-loop integral, whose dual representation is given by the sum of N dual integrals. Each dual integral is associatedwithapossibleone-cut,sowehave (cid:90) L(1)(p ,...,p ) = − ∑ ∏ δ˜(q)G (q;q ), (2.1) 1 N i D i j i∈α1j∈α1,j(cid:54)=i (cid:96) wherethecutconditionisobtainedthrough δ˜(q)≡2πıθ(q )δ(q2−m2), (2.2) i i,0 i i whichforcesq tobeon-shell,and i 1 G (q;q )= i,j∈α ={1,2,...N}, (2.3) D i j q2−m2−ı0η·k 1 j j ji are dual propagators, with k = q −q. The four-momenta of the external legs are p, which ji j i i are taken as outgoing, and we use (cid:96) as the loop momentum. Also, we denote the internal line momentaasq =(q ,q),whereq istheenergycomponentwhilstq referstothespacialpart. i,µ i,0 i i,0 i Ifk = p +...+p representsasumofexternalmomenta(whichfulfillsk =0duetomomentum i 1 i N conservation),thenwehaveq =(cid:96)+k andk =k −k. i i ji j i The distinctive feature of LTD is the introduction of the modified ı0 prescription. The idea ofusingtheresiduetheoremandcutdiagramswasalreadyappliedintheFeynman’streetheorem (FTT) [22, 23], which establishes that loop amplitudes are reconstructed as the sum over all pos- sible multiple-cuts. In fact, these m-cuts are simply obtained by replacing propagators with the associated Dirac’s delta given in Eq. (2.2); no other modification is required to recover the exact loopintegral. Ontheotherhand, LTDonlymakesuseofsinglecutsbymodifyingtheFeynman’s prescriptionandintroducingthearbitraryfuture-likevectorη. Noticethattheprescriptiondepends onthesignoftheproductη·k , i.e. differentcutsmighthaveadifferentprescription. Thispoint ij iscrucialtoexactlyrecoverthediscontinuitystructureofvirtualamplitudes. Inotherterms,since FTT and LTD are equivalent, the multiple-cut information is codified into the dual propagators throughtheirmodifiedı0prescription. Finally,let’smakeacommentabouttherelationamongloopandphase-space(PS)measures. In the context of DREG, virtual contributions involve the integral over (cid:96) without any other con- straint. On the other hand, real radiation terms are obtained after the integration over the extra particle’s PS. Since real particles are involved, they are subjected to physical requirements, i.e. theymustfulfillmomentumconservationandthecorrespondingon-shellrelation. Thelastcondi- tionisimplementedthroughtheintroductionoftheDirac’sdeltagiveninEq.(2.2). Inconsequence, 3 LTDandquantumfieldtheoryin4D GermánF.R.Sborlini we can start from the virtual contribution and use LTD at one-loop; the dual integration measure becomes (cid:90) (cid:90) dd(cid:96) δ˜(q)=−ıµ4−d δ˜(q) , (2.4) (cid:96) i (2π)d i which resembles a real PS measure in d-dimensions. So, LTD converts the usual d-dimensional loop measure into a (d−1)-dimensional integration. The associated integration domain becomes the forward on-shell hyperboloid associated with the solution of the equation G (q)−1 =(q2− F i i m2+ı0) = 0. In this way, dual contributions are expressed in the same form as the real part, i allowing a direct combination at integrand level. We will explain the consequences of this fact in thefollowingsections. 3. Formalexample: disentanglingIRsingularitiesinloops LTDofferstheattractivepossibilityofstudyingthedivergentstructureofvirtualcontributions (loop integrals, in particular) and identifying the regions responsible of the appearance of these singularities. Let’sanalyzethesimplestIR-divergentFeynmanintegral,i.e. ascalartriangleinthe time-like (TL) region. The process is represented by the kinematical configuration p → p +p , 3 1 2 with p2=0=p2and p2=s >0. UsingthenotationintroducedinSection2,wehaveq =(cid:96)+p , 1 2 3 12 1 1 q =(cid:96)+p and q =(cid:96). Then, we parametrize q in the center-of-mass frame, choosing p (p ) 2 12 3 i 1 2 along the positive (negative) z-axis. Using the dimensionless variables (ξ ,v), the kinematical i,0 i invariantsofthissystemaregivenby 2q ·p /s =ξ v , 2q ·p /s =ξ (1−v), (3.1) i 1 12 i,0 i i 2 12 i,0 i andwedefinethed-dimensionalintegrationmeasuresas (cid:18)s (cid:19)−ε d[ξ ]=c 12 ξ−2εdξ , d[v]=(v(1−v))−εdv , (3.2) i,0 Γ µ2 i,0 i,0 i i i i withc theusualloopvolumefactorind dimensions. Here,ξ isrelatedtotheenergycomponent Γ i,0 of q, whilst v is an angular variable. Since there are three internal lines, the application of LTD i i leadsto c (cid:18)−s −ı0(cid:19)−ε 3 L(1)(p ,p ,−p ) = Γ 12 = ∑I , (3.3) 1 2 3 s ε2 µ2 i 12 i=1 withthedualintegrals 1 (cid:90) ∞ (cid:90) 1 I = d[ξ ] d[v ]ξ−1(v (1−v ))−1 , (3.4) 1 s 1,0 1 1,0 1 1 12 0 0 1 (cid:90) ∞ (cid:90) 1 (1−v )−1 2 I = d[ξ ] d[v ] , (3.5) 2 2,0 2 s 1−ξ +ı0 12 0 0 2,0 1 (cid:90) ∞ (cid:90) 1 v−1 I = − d[ξ ] [v ] 3 . (3.6) 3 3,0 3 s 1+ξ 12 0 0 3,0 ItisworthappreciatingthatthedualprescriptionbecomesrelevantforI ,becausethedenominator 2 vanishesinsidetheintegrationdomain. Thisisrelatedwiththepresenceofathresholdsingularity. 4 LTDandquantumfieldtheoryin4D GermánF.R.Sborlini Explicitly, two internal lines become simultaneously on-shell when ξ =1 and the diagram can 2,0 be split into two physical tree-level terms. These contributions resemble those obtained after the application of Cutkowsky’s rules. In fact, we can remove the imaginary part of I by adding the 2 Cutkowskycontribution,andthisremovesthepurelyimaginaryε-polesallowingtoobtainafour- dimensionalrepresentationofL(1)(p ,p ,−p )[7,24]. 1 2 3 Once we have written the dual contributions, we analyze the structure of their integrands to identifytheoriginofloopsingularities. Theintegrationdomainisdefinedbythepossitive-energy solutions of G (q)−1 = 0 in the loop-momentum space, which are geometrically described by F i intersecting light-cones (LCs) when considering massless propagators. In Fig. 1, these domains are shown for each dual contribution; we distinguish between forward (q > 0) and backward i,0 (q <0)LCs. TheintersectionsamongLCsproducedifferentkindsofsingularities,asdiscussed i,0 in Refs. [20, 21]. For the massless triangle, the intersection among the three LCs originates a soft singularity, which is associated with double ε-poles. On the other hand, forward-backward intersections lead to collinear singularities. For instance, the intersection among the forward LC of I and the backward LC of I originates single ε-poles related with those introduced in the 1 3 collinearlimit p (cid:107)q . Analogously,whentheforwardLCofI intersectsthebackwardLCforI , 1 1 2 1 we obtain those singularities associated with the collinear region q (cid:107) p . While forward-forward 1 2 intersections cancel among dual contributions, the integrable threshold-singularity is associated withtheintersectionoftheforwardLCforI andthebackwardLCforI . Inthatregion,q2=0=q2 2 3 2 3 impliesthattwointernallinesareon-shelland,moreover,theyhavepositiveenergy;thus,thedual diagramfactorizesintotheproductoftwotree-levelamplitudes. dcuaalncellation p1 -k3 collinear ξ0 -k1 soft collinearp threshold 2 -k2 ξ z Figure1: LocationofthresholdandIRsingularitiesinthe(ξ ,ξ )space. Forward-forward(FF)singular- 0 z itiescancelamongdualcontributionswhilstforward-backward(FB)intersectionsleadtoIRandthreshold singularities. Inparticular,thecollinearsingularitiesarespreadalongtheFBintersections,andthesoftone isassociatedwiththeintersectionofthethreelight-cones. To conclude this section, let’s use the graphical information available in Fig. 1 to identify the regions that contribute to the IR singular structure of the triangle. We define the following soft/collinearintegrals; (s) I = I (ξ ≤w), (3.7) 1 1 1,0 5 LTDandquantumfieldtheoryin4D GermánF.R.Sborlini (c) I = I (w≤ξ ≤1;v ≤1/2), (3.8) 1 1 1,0 1 (c) I = I (ξ ≤1+w;v ≥1/2), (3.9) 2 2 2,0 2 whereweintroducedanarbitrarycutw>0todealwiththethresholdregion. Performinganexplicit computation,wefind c (cid:18)−s −ı0(cid:19)−ε IIR = I(s)+I(c)+I(c)= Γ 12 1 1 2 s µ2 12 (cid:20) 1 π2 (cid:18) 1(cid:19) (cid:21) × +log(2)log(w)− −2Li − +ıπlog(2) +O(ε), (3.10) ε2 3 2 w whichimpliesthatL(1)(p ,p ,−p )=IIR+O(ε0),i.e. alltheIRsingularstructureofthetriangle 1 2 3 is due to a compact region in the integration domain. This is a crucial property to achieve a local cancellationofIRsingularitieswhenaddingtherealcorrections[7,8,24]. 4. Physicalexample: γ∗→qq¯ atNLO Let’s combine real and virtual contributions through the application of LTD. The first imple- mentationofthisapproachwaspresentedinRef. [7],whereweobtainedtheNLOcorrectionstoa generic1→2decayinthecontextofascalartheory. Here,wewillbrieflydescribetheprocedure forcomputingNLOQCDcorrectionstoγ∗→qq¯. AmoredetailedpresentationisavailableinRef. [24]. In first place, we compute the renormalized one-loop correction to γ∗(p )→q(p )+q¯(p ), 3 1 2 i.e. σ(1)= 1 (cid:90) dΦ (cid:104)2Re(cid:104)M(0)|M(1)(cid:105)−(cid:0)∆Z (p )+∆Z (p )(cid:1)|M(0)|2(cid:105) , (4.1) V 2s 1→2 2 1 2 2 12 that also includes the self-energy corrections to all the external particles; this is a crucial point to achieve a fully local cancellation of IR singularities. Besides that, let’s mention that the presence of a non-trivial structure in the numerator leads to UV divergent terms that must be absorbed into renormalization counter-terms. Using LTD and expressing the internal momenta qµ in terms of i (ξ ,v),weobtain i,0 i (1) (1) (1) (1) σ =σ +σ +σ , (4.2) V V,1 V,2 V,3 (1) whereσ aretherenormalizeddualcross-sections. V,i On the other hand, the real correction corresponds to the process γ∗(p )→q(p(cid:48))+q¯(p(cid:48))+ 3 1 2 g(p(cid:48)),andisgivenby r (cid:18)s (cid:19)−ε (cid:90) 1 (cid:90) 1−y(cid:48) σ(1) = σ(0)c g2C 12 dy(cid:48) 1rdy(cid:48) (y(cid:48) y(cid:48) y(cid:48) )−ε R (cid:101)Γ S F µ2 1r 2r 1r 2r 12 0 0 (cid:20) (cid:18) y(cid:48) (cid:19) (cid:18)y(cid:48) y(cid:48) (cid:19)(cid:21) × 4 12 −ε +2(1−ε) 2r + 1r , (4.3) y(cid:48) y(cid:48) y(cid:48) y(cid:48) 1r 2r 1r 2r wherey(cid:48) s =2p(cid:48)·p(cid:48),σ(0) denotestheBornlevelcross-sectionandc isthePSvolumefactorin ij 12 i j (cid:101)Γ d-dimensions. Then, we separate the real-radiation PS into two regions which only contains one collinearconfiguration. Explicitly,weusetheidentity 1=θ(y(cid:48) −y(cid:48) )+θ(y(cid:48) −y(cid:48) ), (4.4) 2r 1r 1r 2r 6 LTDandquantumfieldtheoryin4D GermánF.R.Sborlini tosplitthethree-bodyPSintotwodisjointregions,R ={y(cid:48) <y(cid:48) }andR ={y(cid:48) <y(cid:48) }. Imple- 1 1r 2r 2 2r 1r mentingthisseparationatintegrandlevelinEq.(4.3),weobtain σ(1)=σ(1)(y(cid:48) <y(cid:48) ) , i,j={1,2}, (4.5) R,i R ir jr (1) (1) (1) which fulfill σ =σ +σ . The following step is the introduction of a momentum mapping R R,1 R,2 that allows to generate 1→3 on-shell kinematics using the variables (ξ ,v). For instance, we i,0 i define p(cid:48)µ =qµ , p(cid:48)µ =−qµ+α pµ = pµ−qµ+α pµ , r 1 1 3 1 2 1 1 1 2 q2 p(cid:48)µ =(1−α )pµ , α = 3 , (4.6) 2 1 2 1 2q ·p 3 2 andexpressy(cid:48) intermsof(ξ ,v )accordingto ij 1,0 1 v ξ (1−v )(1−ξ )ξ y(cid:48) = 1 1,0 , y(cid:48) = 1 1,0 1,0 , y(cid:48) =1−ξ . (4.7) 1r 1−(1−v )ξ 2r 1−(1−v )ξ 12 1,0 1 1,0 1 1,0 ThismappingisspeciallysuitedforR becauseitproperlydescribesthelimity(cid:48) →0,correspond- 1 1r ingtothecollinearconfigurationq (cid:107) p . Ananalogousmomentumtransformationisobtainedfor 1 1 (1) R . Then,weusethecorrespondingmappingtorewriteσ intermsof(ξ ,v). Since 2 R,i i,0 i (cid:18) (cid:19) 1−2v θ(y(cid:48) −y(cid:48) )≡R (ξ ,v ) = θ(1−2v )θ 1 −ξ , (4.8) 2r 1r 1 1,0 1 1 1−v 1,0 1 √ (cid:18) (cid:19) 1− 1−v θ(y(cid:48) −y(cid:48) )≡R (ξ ,v ) = θ 2 −ξ , (4.9) 1r 2r 2 2,0 2 v 2,0 2 the regions R and R are mapped into subsets inside the integration domain of the dual cross- 1 2 sections σ(1) and σ(1), respectively. So, we define σ˜(1) =σ(1)(R) with i={1,2}. Notice that it V,1 V,2 V,i V,i i (1) is not required to deal with σ because it does not contribute to the IR-singular structure of the V,3 (1) (1) virtual part. In other words, only the integrands of σ and σ are needed to locally regularize V,1 V,2 thedivergentbehavioroftherealcontributionintheregionsR andR ,respectively. 1 2 Once the real and dual cross-sections are expressed using the same sets of variables, we pro- ceed to combine them at integrand level. At this point, we realize that the isolation of the IR singularitiesoftheloopintegralsintoacompactregionoftheparameterspacebecomesessential, becausetherealcontributionhasafiniteintegrationdomain. Weconcludethat σ(1)=σ˜(1)+σ(1) , i={1,2} , (4.10) i V,i R,i arefiniteintegralsinthelimitε→0. Moreover,sincetheNLOcorrectionstothetotalcross-section are finite by virtue of KLN theorem, the sum of the dual contributions which are not used in the (1) definitionofσ ,i.e. i (cid:16) (cid:17) (cid:16) (cid:17) σ¯(1)= σ(1)−σ˜(1) + σ(1)−σ˜(1) +σ(1), (4.11) V V,1 V,1 V,2 V,2 V,3 isalsofinitewhenε =0. Themostimportantpointisthatwecanexplicitlyfindfour-dimensional (1) representations for Eqs. (4.10) and (4.11). For σ it is straightforward to prove that the limit i 7 LTDandquantumfieldtheoryin4D GermánF.R.Sborlini ε →0 leads to a regular integrand. In the case of σ¯(1), there are some subtleties since it involves V threedifferentsetsofvariables. Ifwejustimplementashiftintheenergycomponentandunifythe angular variables, i.e. (ξ +a,v)=(ξ ,v), taking the limit ε →0 would lead to missing finite i,0 i i 0 parts. This is due to a mismatch in the collinear limits of the different dual integrands. To cure thisbehavior,wemustusethesamecoordinatesystemtodescribethethreeinternalmomentaqµ. i Once we perform this change of variables, we can consider ε →0 at integrand level and recover thesameresultobtainedinDREG. Finally,ifweaddthefour-dimensionalrepresentationsofEqs. (4.10)and(4.11),weobtain α σ(1)=σ(1)+σ(1)+σ¯(1)=3C S σ(0) , (4.12) 1 2 V F 4π which agrees with the well-known total cross-section at NLO in α . We would like to emphasize S that, following the procedure sketched here, the result shown in Eq. (4.12) is obtained through a purelyfour-dimensionalimplementation. 5. Conclusionsandoutlook The loop-tree duality (LTD) theorem is a theoretical tool which allows to decompose loop- integrals in terms of dual contributions. These dual contributions are build using single cuts by invokingtoasuitablemodificationintheı0prescription. Inthisarticle, wefirstlyappliedLTDto analyzetheIRsingularstructureofthemasslesstriangleintegral. Thisledustotheconclusionthat softandcollinearpolesareoriginatedinacompactregionoftheintegrationdomain. After that, we center in the description of the NLO QCD corrections to the processγ∗→qq¯. Weshowedaproceduretocombinerealandvirtualcontributions,andimplementthecomputation considering the limit ε →0 at integrand level. This a distinctive aspect of our approach, because it is not trivial to find an integral representation which is compatible with commuting the limit ε →0andtheintegral. Inotherwords,weknowthattheadditionofrealandvirtualcontributions shouldleadtofiniteresults; i.e. usingDREG,weintegrateandafterthatwetakethelimitε →0, because all ε-poles cancel. Our claim is stronger, because we found an algorithm that allows to directly combine real and virtual terms before integration, based in the dual decomposition of virtual amplitudes. As explained in Section 4, the representation that we obtained is not only four-dimensionalbutalsocompatiblewiththecommutativityofthelimitε →0. Andtheessential componentofthistechniqueisthereal-virtualmapping,whichallowstogeneratetherealradiation kinematics from the Born level invariants, plus the spatial component of the loop momentum. In thisway,singularitiesofthedualandrealtermsaremappedintothesamepointsintheintegration domain,leadingtoafullylocalregularization. In conclusion, this approach constitutes an alternative to the traditional subtraction method, with the appealing possibility of increasing the computational efficiency [24, 25]. Moreover, the four-dimensional representations that we found can be obtained using purely algebraic methods, which also might shed light into the mathematical structures behind scattering amplitudes and cross-sections. 8 LTDandquantumfieldtheoryin4D GermánF.R.Sborlini Acknowledgments This work has been supported by the Research Executive Agency (REA) under the Grant AgreementNo. PITN-GA-2010-264564(LHCPhenoNet),byCONICETArgentina,bytheSpanish Government and ERDF funds from the European Commission (Grants No. FPA2014-53631-C2- 1-P,FPA2011-23778)andbyGeneralitatValencianaunderGrantNo. PROMETEOII/2013/007. References [1] C.G.BolliniandJ.J.Giambiagi,NuovoCim.B12(1972)20. [2] G.’tHooftandM.J.G.Veltman,Nucl.Phys.B44(1972)189. [3] G.M.CicutaandE.Montaldi,Lett.NuovoCim.4(1972)329. [4] J.F.Ashmore,Lett.NuovoCim.4(1972)289. [5] T.Kinoshita,J.Math.Phys.3(1962)650. [6] S.Catani,T.Gleisberg,F.Krauss,G.RodrigoandJ.C.Winter,JHEP0809(2008)065 [arXiv:0804.3170[hep-ph]]. [7] R.J.Hernández-Pinto,G.F.R.SborliniandG.Rodrigo,arXiv:1506.04617[hep-ph]. [8] G.F.R.Sborlini,R.Hernandez-PintoandG.Rodrigo,arXiv:1510.01079[hep-ph]. [9] S.CataniandM.H.Seymour,Nucl.Phys.B485(1997)291[Erratum-ibid.B510(1998)503] [hep-ph/9605323]. [10] S.CataniandM.H.Seymour,Phys.Lett.B378(1996)287[hep-ph/9602277]. [11] S.Frixione,Z.KunsztandA.Signer,Nucl.Phys.B467(1996)399[hep-ph/9512328]. [12] A.Gehrmann-DeRidder,T.GehrmannandE.W.N.Glover,JHEP0509(2005)056 [hep-ph/0505111]. [13] S.CataniandM.Grazzini,Phys.Rev.Lett.98(2007)222002[hep-ph/0703012]. [14] V.DelDuca,C.Duhr,G.Somogyi,F.TramontanoandZ.Trócsányi,JHEP1504(2015)036 [arXiv:1501.07226[hep-ph]]. [15] M.Czakon,Phys.Lett.B693(2010)259[arXiv:1005.0274[hep-ph]]. [16] R.Boughezal,C.Focke,X.LiuandF.Petriello,arXiv:1504.02131[hep-ph]. [17] J.Gaunt,M.Stahlhofen,F.J.TackmannandJ.R.Walsh,JHEP1509(2015)058[arXiv:1505.04794 [hep-ph]]. [18] I.Bierenbaum,S.Catani,P.DraggiotisandG.Rodrigo,JHEP1010(2010)073[arXiv:1007.0194 [hep-ph]]. [19] I.Bierenbaum,S.Buchta,P.Draggiotis,I.MalamosandG.Rodrigo,JHEP1303(2013)025 [arXiv:1211.5048[hep-ph]]. [20] S.Buchta,G.Chachamis,P.Draggiotis,I.MalamosandG.Rodrigo,JHEP1411(2014)014 [arXiv:1405.7850[hep-ph]]. [21] S.Buchta,PhD.thesis,UniversitatdeValència,2015.arXiv:1509.07167[hep-ph]. [22] R.P.Feynman,ActaPhys.Polon.24(1963)697. 9 LTDandquantumfieldtheoryin4D GermánF.R.Sborlini [23] R.P.Feynman,In*Brown,L.M.(ed.): SelectedpapersofRichardFeynman*867-887 [24] G.F.R.Sborlini,R.Hernández-PintoandG.Rodrigo,inpreparation.IFIC/15-73. [25] S.Buchta,G.Chachamis,P.DraggiotisandG.Rodrigo,arXiv:1510.00187[hep-ph]. 10