Table Of Content

8 Generalized parton distributions and the structure of the nucleon 0 0 2 Sigfrido Boffi and Barbara Pasquini n Dipartimento di Fisica Nucleare e Teorica, Università degli Studi di Pavia, Pavia, Italy a Istituto Nazionale di Fisica Nucleare, Sezione di Pavia, Pavia J 8 2 ] h Summary. — Generalized parton distributions have been introduced in recent p years as a suitable theoretical tool to study the structure of the nucleon. Unifying - theconceptsofpartondistributionsandhadronicformfactors,theyprovideacom- p e prehensive framework for describing the quark and gluon structure of the nucleon. h Inthisreview theirformal properties and modeling arediscussed, summarizing the [ most recent developments in the phenomenological description of these functions. Thestatus of available data is also presented. 2 v PACS 12.38.-t – Quantumchromodynamics. 5 PACS 12.39.-x – Phenomenological quark models. 2 PACS 13.60.-r – Photon and charged-lepton interactions with hadrons. 6 2 . 1 1 7 0 1. – Introduction : v i Thecompositenatureoftheprotonwasmademanifestbythediscoveryofitsanoma- X lous magnetic moment [1] and confirmedby the observationof the electromagnetic form r a factors by Hofstadter and coworkers[2]. Nowadaysit is firmly believed that the internal dynamicsofhadronssuchasthe protonandneutron(collectivelyindicatedasnucleons), is determined by the strong interactions between quarks exchanginggluons, as governed by quantum chromodynamics (QCD). However, a detailed description of the nucleon structure is still missing because QCD can only be solved in the perturbative regime of short distance phenomena probed in hard collisions, whereas the soft part of the interaction corresponding to the long-distance behaviour requires a nonperturbative and/or numerical treatment like, e.g., in lattice simulations. In order to probe the internal structure of the nucleon large energy (ν) and momentum (q) transfers are necessary as in the so-called deep inelastic scattering (DIS) and ultimatelyintheBjorkenregime,i.e. whenbothν andQ2 = q2 =q2 ν2 becomevery − − largewithfixedx =Q2/2p q(orx =Q2/2M ν foranucleonwithmassM atrest). B B N N · Undertheseconditionsscalingoccurs,i.e. structurefunctionsparametrizingtheinclusive DIS cross section become independent of Q2 at fixed x . In the parton model [3] this B phenomenon is interpreted as the incoherent elastic scattering off the partons with the Bjorken x being just the fractional (light-cone, longitudinal) momentum of the struck B 1 2 SIGFRIDOBOFFIand BARBARAPASQUINI parton. At the parton level one may distinguish three kinds of quark distributions, i.e. thequarkdensityq(x ),thehelicitydistribution∆q(x ),andthetransversity∆ q(x ). B B T B The first two are well-knownquantities: q(x ) is the probability of finding a quark with B afractionx ofthelongitudinalmomentumoftheparent(fast-moving)nucleon,regard- B less of its spin orientation; ∆q(x ) gives the net helicity of a quark in a longitudinally B polarizednucleon,i.e. itisthe numberdensityofquarkswithpositivehelicity minusthe number density of quarks with negative helicity, assuming the parent nucleon to have positive helicity; in a transversely polarized nucleon, the transversity ∆ q(x ) is the T B number density of quarks with polarization parallel to that of the nucleon minus the number density of quarks with antiparallel polarization. Information on the last distribution is missing on the experimental side because ∆ q(x ) decouples from inclusive T B DISandthereforecannotbemeasuredinsuchatraditionalsourceofinformation. Other processes,suchaspolarizedDrell-Yandileptonproduction,arebettersuitedforaccessing transversity. Experiments show scaling violation with a Q2 dependence due to the contribution of gluons taking part to the scattering process as active particles or being radiated by the initial and scattered quarks as described by QCD with the Dokshitzer-Gribov-Lipatov- Altarelli-Parisi (DGLAP) evolution equations [4]. Thus, QCD fits to DIS data also determine the gluon distribution. A large amount of data over the years have provided us with a fair description of quark and gluon distributions (see, e.g., [5]). More generally, in the Bjorken regime one probes space-time correlations along the light-cone. WhereasinclusiveDISinvolvesdiagonalmatrixelementsofcertainoperators, thus allowing a probability interpretation in terms of distributions, a full knowledge of thecorrelationscanonlybeachievedbyconsideringalsothenondiagonalmatrixelements of the same operators. This is possible in exclusive processes under suitable conditions whereonecanfactorizeshort-andlong-distancecontributionstothereactionmechanism. These nondiagonal matrix elements can be parametrized in terms of generalized parton distributions (GPDs). GPDs havebeen introduced in the pastin differentcontexts (see, e.g., [6, 7]), but have raised a large interest in the hadron community only when their importancewasstressedinstudiesofdeeplyvirtualComptonscattering(DVCS)[8,9,10] and hard meson production [11] in connection with the possibility of factorizing their contribution[12] andgaininginformationonthe spinstructure ofthe nucleon[8]. Being relatedtonondiagonalmatrixelements,GPDsdonotrepresentanylongeraprobability, butrathertheinterferencebetweenamplitudesdescribingdifferentpartonconfigurations of the nucleon so that they give access to momentum correlations of partons in the nucleon. The finite momentum transfer to the proton makes a second space-time structure of the process possible. Whereas in inclusive DIS the partonic subprocess is the scattering of a photon on a quark or antiquark, in the case of GPDs the virtual photon can also annihilate on a quark-antiquark pair with transverse separation of order 1/Q in the proton target. The momentum transfer can also have a transverse component. This provides a powerful tool to study hadron structure in three dimensions, because in addition to the information on the (longitudinal) behaviour in momentum space along the direction in which the nucleon is moving (as in the case of ordinary parton distributions), they also give insights on how partons are spatially distributed in the transverse plane [13] (as in the case of elastic form factors). The GPDs can also be viewed as the generating functions for the form factors of the twist-two operators governing the interaction mechanisms of hard processes in the GENERALIZEDPARTONDISTRIBUTIONSANDTHESTRUCTUREOFTHENUCLEON 3 deepinelasticregime. Thesegeneralizedformfactorsdonotcoupledirectlytoanyknown fundamentalinteractions,butcanbestudiedindirectlylookingatmomentsoftheGPDs. The most peculiar example are the form factors of the energy momentum tensor, which allowtoaccessthe totalandorbitalangularmomentumofthenucleoncarriedbyquarks and gluons. Inthe lasttenyearsGPDs havebeeninvestigatedingreatdetailfromthe theoretical point of view, and many review papers already have summarized the progressing status of their understanding and modeling [14, 15, 16, 17, 18, 19]. Experimentally, measuring GPDs in exclusive processes is a big challenge requiring high luminosity andresolution. Thereforeonly veryrecently first dedicated experiments have been planned and performed. This review is an attempt to provide a general overview on the most recent developments in the phenomenological description of the nucleon GPDs. After a summary of properties of the GPDs in sect. 2 and their physical content in sect. 3, the main lines of researchin modeling GPDs for the nucleon are presented in sect. 4. Results obtained for quantities describing the nucleon structure are discussed in sect. 5, and the status of experimental investigation is presented in sect. 6. Conclusions are drawn in the final section. 2. – Definition and properties of generalized parton distributions . 21. Generalized form factors. – In quantum field theory one can construct currents, i.e. Dirac tensors as bilinear combinations of Γ matrices (Γ = 1,γ ,γµ,γµγ ,σµν) and 5 5 Dirac fields, ψ¯Γψ. Depending on the selected Γ matrix and its Lorentz properties one can define scalar, pseudoscalar, vector, axial and tensor currents. Matrixelementsoftheabovecurrentsonnucleonstateswithinitial(final)momentum pµ (p′µ) and covariantly normalized as p′ p =2p0(2π)3δ(p′ p), are usually expressed h | i − intermsofformfactors. Forexample,withψ ψ (z)forquarksofflavourq,thematrix q → elements of quark (electroweak) vector and axial currents are decomposed as 1 (2.1) p′ ψ¯ (0)γµψ (0)p = u¯(p′) Fq(t)γµ+i Fq(t)σµν∆ u(p), h | q q | i 1 2M 2 ν (cid:20) N (cid:21) 1 (2.2) p′ ψ¯ (0)γµγ ψ (0)p = u¯(p′) gq(t)γµγ + gq(t)∆µγ u(p), h | q 5 q | i A 5 2M P 5 (cid:20) N (cid:21) where u(p) is the Dirac spinor normalized as u¯(p)u(p) = 2M , and ∆µ = p′µ pµ and N t = ∆2. For each separate flavour q the vector current involves the contribu−tions Fq 1 and Fq respectively to the Dirac and Pauli electromagnetic form factors of proton and 2 neutron, Fp,n and Fp,n, while the weak axial current involves the contributions gq and 1 2 A gq respectively to the axial and induced pseudoscalar form factors. P Restricting oneself to up and down quarks only and referringto their contribution in the proton, the form factors Fq are related to the physical Fp,n as 1,2 1,2 (2.3) Fu =2Fp +Fn , Fd =Fp +2Fn . 1,2 1,2 1,2 1,2 1,2 1,2 For the axial vector form factor one uses the isospin decomposition (2.4) gu = 1g + 1g0, gd = 1g + 1g0, A 2 A 2 A A −2 A 2 A 4 SIGFRIDOBOFFIand BARBARAPASQUINI where g (g0) is the isovector (isoscalar) axial form factor with g (0) = 1.267 and, A A A from quark models, g0(t) = 3g (t). Relations similar to (2.4) hold for the induced A 5 A pseudoscalar form factor containing an important pion pole contribution through the partial conservation of the axial current. Theabovecurrentscanbeconsideredasparticularcasesofmoregeneraloperatorsin QCD.Inthedeepinelasticregimeofhardprocesses,whereanoperatorproductexpansion is performed in order to overcome the problem of light-cone singularities arising as a consequence of the explored short distances, the major contribution comes from the so- called twist-two tensor operators (see, e.g., [20]). Formally, the twist τ is defined as the dimension d in mass units minus the Lorentz spin s of the operator,τ =d s. A spin-s − tensor transforms as an irreducible representation of the Lorentz group. The maximal spin for a given number of Lorentz indices is achieved when they are all symmetrized. The irreducibility implies that the reduction to lower-spin tensors is not possible: as a consequence, the contraction of any pair of indices with the metric tensor gives zero. Thus, the Lorentz structure has to be traceless. InQCD, there are six towersof twist-two operatorsforming totally symmetric repre- sentations of the Lorentz group [14, 18, 19]: ↔ ↔ (2.5a) µµ1···µn−1 = ψ¯ γ(µi µ1 i µn−1)ψ , Oq q D ··· D q ↔ ↔ (2.5b) ˜µµ1···µn−1 = ψ¯ γ(µγ i µ1 i µn−1)ψ , Oq q 5 D ··· D q (2.5c) µνµ1···µn−1 = ψ¯ σµ(νi↔µ1 i↔µn−1)ψ , OqT q D ··· D q ↔ ↔ (2.5d) µµ1···µn−1ν = F(µαi µ1 i µn−1F ν) , Og D ··· D α ↔ ↔ (2.5e) ˜µµ1···µn−1ν = iF(µαi µ1 i µn−1F˜ ν) , Og − D ··· D α (2.5f) µµ1···µn−1ναβ = F(µαi↔µ1 i↔µn−1Fν)β , OgT D ··· D where F˜αβ = 1ǫαβγδF is the dualfieldstrengthtensorwith ǫ =1. In Eqs.(2.5)all 2 γδ 0123 indices within ( ) are symmetrized and traceless, and ··· ↔ → ← → → ← ← (2.6) Dµ≡ 12 Dµ −Dµ , Dµ=∂µ −igtaAaµ(z), Dµ=∂µ +igtaAaµ(z) (cid:16) (cid:17) are covariant derivatives expressed in terms of the gluon vector potential Aa(z). The µ label a = 1,...,8 is the octet colour label, and the t are (one half of) the Gell-Mann a matrices for the triplet representation of SU(3), with (2.7) [t ,t ]=if t . a b abc c The gluon field tensor is (2.8) Fa =∂ Aa ∂ Aa +gf AbAc, µν µ ν − ν µ abc µ ν where g is a constant representing the coupling strength between ψ and Aa. In the q µ following when not necessary the colour index will be omitted. Like the form factors of the electromagnetic current, additional information about nucleonstructurecanbefoundinthe(generalized)formfactorsofthetwist-twooperators GENERALIZEDPARTONDISTRIBUTIONSANDTHESTRUCTUREOFTHENUCLEON 5 when the matrix elements are taken between states of unequal momenta. Using Lorentz symmetry and parity and time reversalinvariance, one can write down all possible form factors of the spin-n operator µµ1···µn−1 in Eq. (2.5a) [14, 18] q O (2.9) p′ ψ¯ γ(µi↔µ1 i↔µn−1)ψ p q q h | D ··· D | i n−1 =u¯(p′) γ(µ∆µ1 ∆µiPµi+1 Pµn−1)Aq (t)  ··· ··· n,i i=0 eXven  n−1 ∆ α iσα(µ∆µ1 ∆µiPµi+1 Pµn−1)Bq (t) − 2M ··· ··· n,i N i=0 eXven 1 + ∆(µ∆µ1 ∆µn−1)Cq(t)Mod(n+1,2) u(p), M ··· n N (cid:21) where P = 1(p+p′) is the averagenucleon momentum, and Mod(n+1,2) is 1 for even 2 n and 0 for odd n. Thus Cq(t) is present only when n is even, and in general there are n n+1 form factors [21, 22]. Similar decompositions in terms of generalized form factors are possible also for nucleon matrix elements of the other twist-two operators in Eqs. (2.5). The problem of counting the number of allowedform factors has been addressedin Ref. [21] making use of a method basedonpartialwaveformalismandcrossingsymmetry,i.e. by considering the number of independent amplitudes in the crossedchannel pp¯ 0 corresponding to h |O| i proton-antiprotoncreation from the twist-two source. In the case of the vector operator one finds exactlyn+1formfactors,asquotedabove. Thesituationis morecomplicated in the case of the axial vector and tensor operators [22, 23]. As has been observed in Ref. [17], there is no Cq(t)-like generalized form factor n present for the axial vector, and the parametrizationreads [22] (2.10) p′ ψ¯ (0)γ(µγ i µ1 i µn−1)ψ (0) p q 5 q h | D ··· D | i n−1 =u¯(p′) γ(µγ ∆µ1 ∆µiPµi+1 Pµn−1)A˜q (t) 5 ··· ··· n,i eXiv=e0nn 1 +γ ∆(µ∆µ1 ∆µiPµi+1 Pµn−1)B˜q (t) u(p), 52M ··· ··· n,i N (cid:27) for a total of 2[n−1]+2 independent form factors. 2 There is a total of 2[n−1]+n+2 independent form factors for the parametrization 2 of the tensor operator µνµ1···µn−1 in Eq. (2.5c) [22]: OqT (2.11) p′ ψ¯ (0)iσµ(νi µ1 i µn−1)ψ (0) p q q h | D ··· D | i n−1 = A S u¯(p′) iσµν∆µ1 ∆µiPµi+1 Pµn−1A (t)  Tn,i [µν](νµ1...) i=0 ··· ··· eXven  6 SIGFRIDOBOFFIand BARBARAPASQUINI n−1 P[µ∆ν] + ∆µ1 ∆µiPµi+1 Pµn−1A˜ (t) M2 ··· ··· Tn,i i=0 N eXven n−1 γ[µ∆ν] + ∆µ1 ∆µiPµi+1 Pµn−1B (t) Tn,i 2M ··· ··· N i=0 eXven n−1γ[µPν] + ∆µ1 ∆µiPµi+1 Pµn−1B˜ (t) u(p), M ··· ··· Tn,i  N i=0 Xodd   where one has first to symmetrize and then to antisymmetrize as indicated [24]. Analogous form factor decompositions are possible for the gluon operators [17]. . 22. Generalized parton distributions. – In hard scattering processes, where hadrons andpartonsmovefastinthezˆ-direction,itisnaturaltoworkwithlight-conecoordinates in terms oftwo light-likefour-vectorsn =(1,0,0,1)/√2 andn =(1,0,0, 1)/√2, i.e. + − − (2.12) vµ v+nµ +v−nµ +vµ, ≡ + − ⊥ where v+ =v n =(v0+v3)/√2, v− =v n =(v0 v3)/√2 and v =(0,v ,0). The − + ⊥ ⊥ · · − relevantmomentaarethenthelight-coneplus-momentap+ andtherelevantΓstructures become, e.g., n/ γ n =γ+. − − ≡ · Light-cone bilocal operators arising in hard scattering processes can be expanded in terms of the above currents making use of the relation (2.13) (P+)n dxxn−1 dz− eixP+z− ψ¯ ( 1z)γ+ψ (1z) Z Z 2π h q −2 q 2 iz+=0,z⊥=0 d n−1 ↔ = i ψ¯ ( 1z)γ+ψ (1z) =ψ¯ (0)γ+(i∂+)n−1ψ (0) dz− q −2 q 2 q q (cid:12)z=0 (cid:16) (cid:17) h i(cid:12) (cid:12) and its analogs for operators involving other(cid:12)Γ matrices than γ+. In general, with the covariant derivative instead of ∂ on the right-hand side, a Wilson link operator D W[ 1z,1z] appears between the operators at positions 1z and 1z, where −2 2 −2 2 b (2.14) W[a,b]= exp ig dz−A+(z−n ) , − P − Za ! and denotespathorderingbetweena andb. Inthe light-conegauge,A+ =0,the Wil- P son link reduces to unity, as it will be assumed in the following. However, the condition A+ = 0 does not remove all gauge freedom because z−-independent gauge transforma- tions are still possible with consequences on the correlation functions in hard processes sensitive to transverse momenta of partons [25, 26, 27, 28, 29]. Technical details needed in the calculation of the gauge link corresponding to a given partonic subprocess and a calculational scheme are provided in Refs. [30, 31]. Partondistributions are just defined in terms of matrix elements oflight-cone bilocal operators between proton states of equal momenta [32]. In general, with initial (final) GENERALIZEDPARTONDISTRIBUTIONSANDTHESTRUCTUREOFTHENUCLEON 7 Fig. 1. – (a) Kinematic variables in the symmetric frame of reference; (b) parametrization of the GPD Hq(x,ξ,t) in terms of momentum fractions. momentump(p′)andhelicityλ (λ′ )onedefinesasetofgeneralizedquarkdistributions N N for a hadron with spin 1 that have been classified in Refs. [33, 34]: 2 (2.15) dz− eixP+z− p′,λ′ ψ¯ ( 1z)γ+ψ (1z) p,λ Z 4π h N| q −2 q 2 | Ni(cid:12)z+=0,z⊥=0 = 1 u¯(p′,λ′ ) Hqγ++Eq iσ+α∆α u(p(cid:12)(cid:12),λ ), 2P+ N 2M N (cid:20) N (cid:21) (2.16) dz− eixP+z− p′,λ′ ψ¯ ( 1z)γ+γ ψ (1z) p,λ Z 4π h N| q −2 5 q 2 | Ni(cid:12)z+=0,z⊥=0 = 1 u¯(p′,λ′ ) H˜qγ+γ +E˜q γ5∆+ u(p,λ(cid:12)(cid:12) ), 2P+ N 5 2M N (cid:20) N (cid:21) (2.17) dz− eixP+z− p′,λ′ ψ¯ ( 1z)iσ+iψ (1z) p,λ Z 4π h N| q −2 q 2 | Ni(cid:12)z+=0,z⊥=0 1 P+∆i ∆+(cid:12)Pi = u¯(p′,λ′ ) Hq iσ+i+H˜q − (cid:12) 2P+ N T T M2 (cid:20) N γ+∆i ∆+γi γ+Pi P+γi +Eq − +E˜q − u(p,λ ). T 2M T M N N N (cid:21) BecauseofLorentzinvariancethe eightGPDs Hq, Eq,H˜q,E˜q,Hq,H˜q,Eq,E˜q can T T T T onlydependonthreekinematicalvariables(Fig.1),i.e. the(average)quarklongitudinal momentum fraction x = k+/P+, the invariant momentum square t and the skewness parameter ξ given by p+ p′+ ∆+ (2.18) ξ = − = . p++p′+ −2P+ In addition, as in the case of parton distributions, there is an implicit scale dependence in the definitionof GPDs correspondingto the renormalizationscale µ2, i.e. the scale at which the QCD operators in Eqs. (2.15)-(2.17)are understood to be renormalized. The eight GPDs are all required to be real valued as a consequence of time reversal invariance, with support in the interval x,ξ [ 1,1]. ∈ − Similarly for gluons one has [34] (2.19) 1 dz− eixP+z− p′,λ′ F+i( 1z)F +(1z) p,λ P+ Z 2π h N| −2 i 2 | Ni(cid:12)z+=0,z⊥=0 (cid:12) (cid:12) 8 SIGFRIDOBOFFIand BARBARAPASQUINI 1 iσ+α∆ = u¯(p′,λ′ ) Hgγ++Eg α u(p,λ ), 2P+ N 2M N (cid:20) N (cid:21) (2.20) i dz− eixP+z− p′,λ′ F+i( 1z)F˜+(1z) p,λ −P+ Z 2π h N| −2 i 2 | Ni(cid:12)z+=0,z⊥=0 = 1 u¯(p′,λ′ ) H˜gγ+γ +E˜g γ5∆+ u(p,λ ), (cid:12)(cid:12) 2P+ N 5 2M N (cid:20) N (cid:21) (2.21) 1 dz− eixP+z− p′,λ′ SF+i( 1z)F+j(1z) p,λ −P+ Z 2π h N| −2 2 | Ni(cid:12)z+=0,z⊥=0 1 P+∆j ∆+Pj P+(cid:12)∆i ∆+Pi =S − u¯(p′,λ′ ) Hg iσ+i+H˜g (cid:12) − 2P+ 2M P+ N T T M2 N (cid:20) N γ+∆i ∆+γi γ+Pi P+γi +Eg − +E˜g − u(p,λ ), T 2M T M N N N (cid:21) where asummationoveri=1,2is implied andSdenotes symmetrizationini andj and subtraction of the trace. Quarkand gluonoperatorscan be usefully rearrangedin orderto make explicit their actiononthepartonhelicity,asindicatedinTab.I. Thecorrespondingmatrixelements, (2.22a) Aq = dz− eixP+z− p′,λ′ q (z) p,λ , λ′Nµ′,λNµ Z 2π h N|Oµ′,µ | Ni(cid:12)z+=0,z⊥=0 (2.22b) Ag = 1 dz− eixP+z− p′,λ′ g (z) p,(cid:12)(cid:12)λ , λ′Nµ′,λNµ P+ Z 2π h N|Oµ′,µ | Ni(cid:12)z+=0,z⊥=0 (cid:12) are similar to helicity amplitudes [17]. In particular they share the(cid:12) same property from Table I.–Quark and gluon operators, Oµq′µ and Oµg′µ respectively, for leading-twist GPDs and the parton helicity transitions they describe. It is implied that the first field is taken at −1z 2 carrying helicity µ′ and the second at 1z carrying helicity µ. 2 x≤−ξ x<|ξ| x≥ξ Oq µ′ µ µ′ µ µ′ µ µ′µ 14ψ¯qγ+(1+γ5)ψq − − − + + + 14ψ¯qγ+(1−γ5)ψq + + + − − − −41iψ¯q(σ+1−iσ+2)ψq + − + + − + 14iψ¯q(σ+1+iσ+2)ψq − + − − + − Og µ′µ 12[F+iFi+−iF+iF˜i+] − − − + + + 21[F+iFi++iF+iF˜i+] + + + − − − 1[F+1F1+−F+2F2+−iF+1F2+−iF+2F1+] + − + + − + 2 1[F+1F1+−F+2F2++iF+1F2++iF+2F1+] − + − − + − 2 GENERALIZEDPARTONDISTRIBUTIONSANDTHESTRUCTUREOFTHENUCLEON 9 parity invariance, (2.23) A−λ′N−µ′,−λN−µ =(−1)λ′N−µ′−λN+µ Aλ′Nµ′,λNµ ∗. h i Using light-cone helicity spinors explicit calculation [17] gives the matrix elements H +H˜ ξ2 E+E˜ (2.24a) A = 1 ξ2 , ++,++ − 2 − 1 ξ2 2 ! − p H H˜ ξ2 E E˜ (2.24b) A = 1 ξ2 − − , −+,−+ − 2 − 1 ξ2 2 ! − p √t t E ξE˜ (2.24c) A = eiϕ 0− − , ++,−+ − 2M 2 N √t t E+ξE˜ (2.24d) A = eiϕ 0− , −+,++ 2M 2 N for both quarks and gluons, where ϕ is the azimuthal angle of the vector D = p′/(1 ξ) p/(1+ξ), i.e. eiϕ = (D1 +iD2)/D . Therefore, the four GPDs H, E, H˜, E˜ ar−e − | | parton helicity conserving and chiral even. The distributions H and E are sometimes referredtoasunpolarizedandH˜ andE˜ aspolarizedbecauseH andE correspondtothe sum over parton helicities, and H˜ and E˜ to the difference. In the case of parton-helicity flip, for quarks one obtains √t t Eq +E˜q (2.25a) Aq =eiϕ 0− H˜q +(1 ξ) T T , ++,+− 2MN T − 2 ! √t t Eq E˜q (2.25b) Aq =eiϕ 0− H˜q +(1+ξ) T − T , −+,−− 2M T 2 N ! t t ξ2 ξ (2.25c) Aq = 1 ξ2 Hq + 0− H˜q Eq + E˜q , ++,−− − T 4M2 T − 1 ξ2 T 1 ξ2 T (cid:18) N − − (cid:19) p t t (2.25d) Aq = e2iϕ 1 ξ2 0− H˜q, −+,+− − − 4M2 T N p where 4ξ2M2 (2.26) t = N − 0 1 ξ2 − is the minimal value for t at given ξ. Analogous expressions hold for gluons with − an additional global factor eiϕ 1 ξ2√t t/(2M ) on the right-hand side. Other 0 N − − helicity combinationsaregivenby parity invariance. Therefore,the four GPDs H , H˜ , p T T E , E˜ are parton helicity flipping and chiral odd. T T By inverting the set of Eqs. (2.24), (2.25) the different GPDs can be separately ex- tracted from the amplitudes Aλ′Nµ′,λNµ. 10 SIGFRIDOBOFFIand BARBARAPASQUINI . 23. Forward limit. – In the forward case, p = p′, both ∆ and ξ are zero. As ξ 0, → alsox x ,wheretheBjorkenvariablex isthefractionofthelongitudinalmomentum B B carried→by the active parton. In this case the functions Hq, H˜q and Hq reduce to the T usual DIS parton distribution functions, i.e. q(x ), x >0, (2.27) Hq(x ,0,0)= B B B q¯( x ), x <0, B B (cid:26) − − ∆q(x ), x >0, (2.28) H˜q(x ,0,0)= B B B ∆q¯( x ), x <0, B B (cid:26) − ∆ q(x ), x >0, (2.29) Hq(x ,0,0)= T B B T B ∆ q¯( x ), x <0, T B B (cid:26) − − whereq(x ),(q¯(x )),∆q(x )(∆q¯(x ))and∆ q(x )(∆ q¯(x ))arequark(antiquark) B B B B T B T B density, helicity and transversity distributions, respectively: (2.30) q(x )= dz− eixBp+z− pψ¯(0)γ+ψ(z)p , B 4π h | | i Z (cid:12)z+=z⊥=0 (cid:12) (2.31) ∆q(xB)= d4zπ− eixBp+z−hpSk|ψ¯(0)γ+γ5ψ(z(cid:12)(cid:12))|pSki , Z (cid:12)z+=z⊥=0 (2.32) ∆Tq(xB)= dz− eixBp+z− pS⊥ ψ¯(0)γ+γ1γ5ψ(z)pS(cid:12)(cid:12)(cid:12)⊥ , 4π h | | i Z (cid:12)z+=z⊥=0 (cid:12) (cid:12) (cid:12) with S (S ) being the longitudinal (transverse) nucleon-spin projection. k ⊥ NocorrespondingrelationsexistforthefunctionsEq,E˜q,Eq andH˜q,becauseinthe T T forward limit they decouple in their defining equations. However, they do not vanish. In particular, Eq(x,0,0) carries important information about the quark orbital angular momentum (see. Eq.(2.67)). In contrast,E˜q vanishes identically being an odd function T of ξ by time reversalsymmetry [34]. For gluons one has (2.33) Hg(x ,0,0)=x g(x ), H˜g(x ,0,0)=x ∆g(x ), x >0. B B B B B B B All the gluon helicity-flip matrix elements go to zero in the forward case corresponding to the collinear limit (t = t in Eqs. (2.25)) and therefore for a spin-1 target decouple 0 2 from any observable for collinear scattering. . 24. Polynomiality. – Introducing a light-like vector nµ, which is conjugate to Pµ in the sense that P n = 1, and contracting both sides of Eq. (2.9) with n n n , · µ µ1··· µn−1 one obtains (2.34) n n n p′ µµ1···µn−1 p µ µ1··· µn−1h |Oq | i σµαn i∆ =u¯(p′)n/u(p)Hq(ξ,t)+u¯(p′) µ α u(p)Eq(ξ,t), n 2M n N