Single spin asymmetry for $p^{uparrow}p to pi X$ in perturbative QCD

a r X i v :h e p -p h /9503290v 2 3 O c t 1995DFTT 48/94INFNCA-TH-94-27hep-ph/9503290Revised version -July 1995SINGLE SPIN ASYMMETRY FOR p ↑p →πX IN PERTURBATIVE QCD M.Anselmino 1,M.Boglione 1,F.Murgia 21Dipartimento di Fisica Teorica,Universit`a di Torino and INFN,Sezione di Torino,Via P.Giuria 1,10125Torino,Italy 2INFN,Sezione di Cagliari,Via A.Negri 18,09127Cagliari,Italy Abstract:Within the QCD-improved parton model and assuming the factorization theorem to hold in the helicity basis and for higher twist contributions,we show how non zero single spin asymmetries in hadron-hadron high energy and moderately large p T inclusive processes can be obtained,even in massless perturbative QCD,provided the quark intrinsic motion is taken into account.A simple model is constructed whichreproduces the main features of the data on the single spin asymmetry observed in inclusive pion production in p p collisions.1IntroductionSingle spin asymmetries in high energy and moderately large p T inclusive hadronic processes have recently received much attention,both experimentally[1]-[4]and the-oretically[5]-[13].Whereas they are expected to vanish at leading twist in massless perturbative QCD,higher twist effects might still be important in the kinematical region of the available data and may give origin to non zero values.Among the attempted explanations,quark-gluon correlations[5,8,13]and transverse k⊥effects in the quark distribution[6,7]or fragmentation functions[11]have been considered.We analyse here the large single spin asymmetries observed in the collision of transversely polarized protons offunpolarized protons,with the production of pions with p T values up to2GeV/c,p↑p→πX.(1) Several experimental results are available for such processes[2]-[4]and show clear patterns in the dependence of the spin asymmetrydσ↑−dσ↓A N=s,where p L is the pion longitudinal momentum in the p p c.m.√system,andWe know that the above hard scattering scheme works well for unpolarized pro-cesses and indeed it has been tested in many experiments.It has also been general-ized to the polarized case[18],so that it may be applied to the description of several processes involving polarized hadrons[19];however,the existing spin data do not allow yet a definite test of its validity.In Section2we adapt the formalism of Refs.[9]and[18]to the case of the single spin asymmetry(2).In order not to obtain a zero result higher twist effects have to be introduced;this can be done at different stages and several suggestions or attempts have been proposed in the literature[6]-[11],[20,21].Single spin effects in the elementary reactions alone are bound to be proportional toαs m q/√√√d3pπ∼1z(5)ρa/p↑λa,λ′af a/p↑(x a)f b/p(x b)ˆMλc,λd;λa,λbˆM∗λ′c,λd;λ′a,λbDλc,λ′cπ/c(z),where f a/p↑(x a)is the number density of partons a with momentum fraction x a inside the polarized proton[similarly for f b/p(x b)]andρa/p↑λa,λ′a(x a)is the helicity densitymatrix of parton a inside the polarized proton p↑.TheˆMλc ,λd;λa,λb’s are the helicityamplitudes for the elementary process ab→cd;if one wishes to consider higher order(inαs)contributions also elementary processes involving more partons should be included.Dλc,λ′cπ/c(z)is the product of fragmentation amplitudesDλc,λ′cπ/c= X,λX DλX;λc D∗λX;λ′c(6) where the X,λX stays for a spin sum and phase space integration of the undetected particles,considered as a system X.The usual unpolarized fragmentation function Dπ/c(z),i.e the density number of pions resulting from the fragmentation of an unpolarized parton c and carrying a fraction z of its momentum is given byDπ/c(z)=11Equation(5)holds as an equality if the elementary amplitudes are normalized so that (1/4) λa,λb,λc,λd|ˆMλa,λb,λc,λd|2=(1/π)(dˆσ/dˆt)final particle C with respect to the axis of the jet generated by parton c does not imply any more λc =λ′c and allows a non zero value of the asymmetrydσp ↑p →π,k ⊥X −dσp ↓p →π,k ⊥X .(9)The above asymmetry (9)is related to the so called Collins [11,24,25]or sheared jet [26]effect;it requires the measurement of the azimuthal angle φof the outgoing hadron around the jet axis,but,apart from a small sin φdependence,it is a leading twist effect and it depends on some non perturbative quark fragmentation analysing power.When integrating over the azimuthal angle the effect might not entirely disappear because of some φdependence in the elementary parton interaction.This idea was exploited in Ref.[11]where,essentially,the parton c is produced in the forward direction and the final hadron p T is due to its transverse k ⊥inside the jet.One cannot expect such a model to work at large p T .Another possible k ⊥effect,suggested by Sivers [6,7],may originate in the dis-tribution functions.To see how this comes out from the general scheme we rewrite Eq.(5)taking into account the parton intrinsic momentum in the number density f a/p :E πdσp ↑p →πX 2 a,b,c,d λa ,λ′a ;λb ;λc ,λ′c ;λd d 2k ⊥a dx a dxb1d 3p π−E πdσp ↓p →πX4 a,b,c,d λa ,λb ;λc ,λdd 2k ⊥a dx a dx b 1whereˆf a,λa /p↑(↓)(x a,k⊥a)is the number density of partons a with helicityλa,momen-tum fraction x a and intrinsic transverse momentum k⊥a in a transversely polarized proton[with spin↑or↓according to Eq.(3)or(4)].Eq.(13)follows from Eq.(12)by rotational invariance and explicitely shows that∆N f a/p↑(x,k⊥)=0when k⊥=0.This new quantity can be regarded as a single spin asymmetry or analysing power for the p↑→a+X process;if we define the polarized number densities in terms of distribution amplitudes asˆfa,λa/p↑(x a,k⊥a)= X p,λX p|G a/pλX p,λa;↑(x a,k⊥a)|2(14) then we have,in the helicity basis,∆N f a/p↑(x a,k⊥a)= X p,λX p λa2Im G a/pλX p,λa;+(x a,k⊥a)G a/p∗λX p,λa;−(x a,k⊥a)≡2I a/p+−(x a,k⊥a).(15) Eq.(15)simply follows from Eqs.(12)and(14)via Eqs.(3)and(4)and showsthe non diagonal nature,in the helicity indices,of I a/p+−(x a,k⊥a).Collins[9]has argued that a non zero value of I a/p+−(x,k⊥)is forbidden by the timereversal invariance of QCD;his argument is based on the analysis of the non diagonal matrix elements of the leading twist quark operator¯ψγ+ψ,whose diagonal matrixelements are related to the distribution functions f q/p(x,k⊥).For such operator his argument is correct and indeed time reversal invariance forces the matrix elements between proton states with different helicities to be zero.In a more physical languagethis amounts to say that single spin asymmetries for the process p↑→qX are forbidden by parity and time reversal invariance,which is true.However,as we said,initial state interactions(like soft gluon exchanges)between the incoming protons must certainly occur,and,at least in cases in which their neglect gives a zero result, one should consider them and relate the distribution functions to the inclusive cross-section for the process p1p2→qX;single(transverse)spin asymmetries are then certainly allowed(as the problem we are studying here confirms)via time reversal invariant scalar quantities likeεµνρσpµ1pν2qρsσ1.These soft gluon and initial state interactions which correlate partons from dif-ferent hadrons and allow a non zero value of I a/p+−can only survive at higher twist;ithas been shown[18]that at leading twist-2the proof of the factorization theorem for the unpolarized case–with the cancellation of all soft gluon contributions–holds for the polarized case as well.Eq.(11),as it will be shown in the next Section,onlygives higher twist contributions;assuming a non zero value of I a/p+−in the factorizedstructure of Eq.(11)amounts to assume the validity of the factorization theorem beyond leading twist.In the operator language this approach has been advocated by Qiu and Sterman [8]who use generalized factorization theorems valid at higher twist and relate nonzero single spin asymmetries in p p collisions to the expectation value of a higher twist operator,a twist-3parton distribution,which explicitly involves correlations between the two protons and combines quarkfields with a gluonicfield strength. However,they still consider only collinear partonic configurations so that,in order to obtain non zero results,they have to take into account the contributions of higher order elementary interactions.Our function I a/p+−(x a,k⊥a)introduced in Eqs.(15)and(12)can be consideredas a new phenomenological quantity which takes into account the non perturbative long distance physics,including initial state interactions,and plays,for single spin asymmetries in p p collisions,the same role plaid by the distributions functions f a/p(x a)in unpolarized processes.This function is zero in the absence of parton intrinsic motion,but,for k⊥=0,allows non zero single spin asymmetries even taking into account only lowest order perturbative QCD interactions among theconstituents.Measurements of the asymmetries supply information on I a/p+−,likemeasurements of unpolarized cross-sections supply information on f a/p.In the next Section we introduce a simple parametrization of I a/p+−,based on ourknowledge of the distribution functions,and show that it can reproduce with good accuracy the data on the single spin asymmetryA N=dσp↑p→πX−dσp↓p→πX2dσunp.(16)3A simple phenomenological modelInserting Eqs.(11)and(15),with the proper kinematical factors[23],into Eq.(16)yieldsA N= a,b,c,d dx a dx b d2k⊥a I a/p+−(x a,k⊥a)f b/p(x b)[dˆσ/dˆt(k⊥a)]Dπ/c(z)/zI a/p +−which does not originate from the k⊥a dependence is taken to be of the simpleformN a xαa a(1−x a)βa,(18) so that we approximately haved2k⊥a I a/p+−(x a,k⊥a)dˆσdˆt (+k⊥a)−dˆσMN a xαa a(1−x a)βa dˆσdˆt(−k0⊥a) ,(19)where M is a hadronic mass scale,M≃1GeV/c.A numerical estimate of k0⊥a/M can be found in Ref.[27]and can be accurately reproduced by the expressionk0⊥aAs the experimental data [2]cover a p T range between 0.7and 2.0GeV/c we have computed A N at a fixed value p T =1.5GeV/c .Our asymmetry decreases with increasing p T and increases at smaller p T ;however,we do not expect our approach to be valid at p T smaller than,say,1GeV/c .We will further comment on the p T range of our computation in the conclusions.Notice that the above values (21)are very reasonable indeed;actually,apart from an overall normalization constant,they might even have been approximately guessed.The exponents αu,d and βu,d are not far from the very na¨ıve values one canobtain by assuming,as somehow suggested by Eqs.(15)and (14),that I a/p +−(x a )∼2D π+/c +D π−/c .(22)Eq.(17)show that the relation (22)also holds for N 0and D 0,so thatA 0N =N ++N −A +N D −1+D −of polarized protons offunpolarized ones.Such region of p T is a delicate one;even if we assume,somewhat optimistically,that the hard scattering scheme is suitable for the description of these processes,we expect that,due to the moderate p T values, higher twist contributions may still be sizeable and important.Indeed,the leading twist contribution to the single spin asymmetries is zero and the large experimental data should be explained via non leading terms.In evaluating them one has to assume that the factorization theorem still holds at higher twist level;moreover,one can only take into account a few out of the many higher twist contributions and corrections to the simple hard scattering for-mulae.Here,we have considered the role of the parton intrinsic k⊥in the initial polarized proton[6,7],while neglecting other possible effects due,e.g.,to transverse k⊥in the fragmentation process.In this our model can only be regarded as a phe-nomenological approach to the description of the otherwise mysterious large single spin asymmetries.Further application of the same model should test its validity.We have shown how single spin asymmetries can be different from zero and originate from lowest order perturbative QCD interactions,provided some non per-turbative and intrinsic k⊥effects are properly taken into account.In the helicity basis,suitable for the use of the factorization theorem,this non perturbative longdistance information is contained in the function I a/p+−(x,k⊥),which has a simpleparton model interpretation in the transverse spin basis,Eqs.(12)-(15).Detailed data on single spin asymmetries in p p inclusive interactions would yield information on I a/p+−(x,k⊥),similar to the information gathered on the unpolarizedstructure functions;a knowledge of I a/p+−from some process could then be used topredict spin effects in other processes.Here we have shown how a simple and realisticparametrization of I a/p+−can easily explain the data on single spin asymmetries forthe process p↑p→πX at large energy and p T≃2GeV/c.Our model clearly exhibits the observed increase with x F,atfixed p T values,of the magnitude of the asymmetries.A more detailed application of our approach to other processes,like ¯p↑p→πX,p↑p→γX andπp↑→πX is in progress[31]:this should help in assessing the relevance and importance of our estimates.AcknowledgementsWe would like to thank J.C.Collins for useful correspondenceReferences[1]D.L.Adams et al.,Phys.Lett.B261(1991)201[2]D.L.Adams et al.,Phys.Lett.B264(1991)462[3]D.L.Adams et al.,Phys.Lett.B276(1992)531[4]D.L.Adams et al.,Z.Phys.C56(1992)181[5]M.G.Ryskin,Yad.Fiz.48(1988)1114[Sov.J.Nucl.Phys.48(1988)708][6]D.Sivers,Phys.Rev.D41(1990)83[7]D.Sivers,Phys.Rev.D41(1991)261[8]J.Qiu,G.Sterman,Phys.Rev.Lett.67(1991)2264[9]J.Collins,Nucl.Phys.B396(1993)161[10]C.Boros,Liang Zuo-tang and Meng Ta-chung,Phys.Rev.Lett.70(1993)1751[11]X.Artru,J.Czyzewski and H.Yabuki,preprint LYCEN/9423and TPJU12/94,May1994,hep-ph/9405426[12]M.Anselmino,M.Boglione and F.Murgia,Proceedings of the XI International Sym-posium on High Energy Spin Physics,Bloomington,Indiana,September1994,hep-ph/9410382[13]A.V.Efremov,V.M.Korotkiyan and O.V.Teryaev,Phys.Lett.B348(1995)577[14]S.Nurushev,Proceedings of the XI International Symposium on High Energy SpinPhysics,Bloomington,Indiana,September1994[15]J.Antille et al.,Phys.Lett.B94(1980)523[16]B.E.Bonner et al.,Phys.Rev.D41(1990)13[17]S.Saroffet al.,Phys.Rev.Lett.64(1990)995[18]J.C.Collins,Nucl.Phys.B394(1993)169[19]J.C.Collins,S.H.Heppelmann and dinsky,Nucl.Phys.B420(1994)565[20]J.Szwed,Phys.Lett.B105(1981)403[21]A.V.Efremov and O.V.Teryaev,Yad.Fiz.36(1982)242[Sov.J.Nucl.Phys.36(1982)140][22]G.Kane,J.Pumplin and W.Repko,Phys.Rev.Lett.41(1978)1698[23]D.Sivers,R.Blankenbecler and S.Brodsky,Phys.Rep.23C(1976)1[24]A.Kotzinian,preprint hep-ph/9412283[25]R.D.Tangerman and P.J.Mulders,preprint NIKHEF95-001,hep-ph/9501202[26]X.Artru and J.C.Collins,preprint PSU/TH/158,hep-ph/9504220[27]J.D.Jackson,G.G.Ross and R.G.Roberts,Phys.Lett.B226(1989)159[28]A.D.Martin,W.J.Stirling and R.G.Roberts,Phys.Rev.D50(1994)6734[29]R.D.Field,Applications of perturbative QCD,Redwood City,Addison Wesley(1989)[30]R.Cutler,D.Sivers,Phys.Rev.D17(1978)196[31]M.Anselmino,M.Boglione and F.Murgia,in preparationFigure captionFig.1Fit of the data on A N[2],with the parameters given in Eq.(21);the upper, middle and lower sets of data and curves refer respectively toπ+,π0andπ−.This figure "fig1-1.png" is available in "png" format from: /ps/hep-ph/9503290v2。