Deep inelastic scattering and factorization in the ’t Hooft Model 9 0 0 2 n a J Jorge Mondejara and Antonio Pinedab 0 2 a Department of Physics, University of Alberta, Edmonton, Alberta, Canada T6G 2G7 ] h p b Grup de F´ısica Te`orica and IFAE, Universitat Aut`onoma de Barcelona, E-08193 - p Bellaterra, Barcelona, Spain e h [ 1 v 3 1 1 Abstract 3 . 1 We study in detail deep inelastic scattering in the ’t Hooft model. We are able to an- 0 alytically check current conservation and to obtain analytic expressions for the matrix 9 0 elements with relative precision (1/Q2) for 1 x β2/Q2. This allows us to compute v: the electron-meson differential cOross section an−d its≫moments with 1/Q2 precision. For i X the former we find maximal violations of quark-hadron duality, as it is expected for a r large Nc analysis. For the latter we find violations of the operator product expansion at a next-to-leading order in the 1/Q2 expansion. PACS numbers: 12.38.Aw, 12.39.St, 11.10.Kk, 11.15.Pg 1 Contents 1 Introduction 2 2 QCD in the large N limit 5 1+1 c 2.1 QCD in the light front . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 1+1 2.1.1 Light-cone coordinates . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.1.2 The QCD lagrangian in the light-cone frame . . . . . . . . . . . . . 7 2.2 The ’t Hooft model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.3 The ’t Hooft equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 2.3.1 Semiclassical solution of the ’t Hooft equation . . . . . . . . . . . . 12 2.3.2 The boundary-layer approximation . . . . . . . . . . . . . . . . . . 13 2.4 Transition matrix elements . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 2.4.1 Neutral currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 2.4.2 Charged meson: Non-equal mass case . . . . . . . . . . . . . . . . . 18 2.4.3 Chargeless meson: Equal mass case . . . . . . . . . . . . . . . . . . 19 3 DIS in the ’t Hooft model 20 3.1 Hadronic computation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 3.1.1 Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 3.1.2 Scattering cross section . . . . . . . . . . . . . . . . . . . . . . . . . 24 3.1.3 Matrix elements in the DIS limit . . . . . . . . . . . . . . . . . . . 26 3.1.4 The hadronic tensor . . . . . . . . . . . . . . . . . . . . . . . . . . 30 3.1.5 The forward Compton scattering amplitude . . . . . . . . . . . . . 32 3.1.6 Moments at next-to-leading order . . . . . . . . . . . . . . . . . . . 33 3.1.7 Determination of TNLO . . . . . . . . . . . . . . . . . . . . . . . . . 35 3.2 Perturbative factorization . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 3.2.1 Calculation of TOPE . . . . . . . . . . . . . . . . . . . . . . . . . . 39 3.2.2 Moments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 4 Conclusions 46 A Semileptonic B decays in the ’t Hooft model 49 1 Introduction At its birth Quantum Chromodynamics (QCD) looked like a rather peculiar theory. It is constructed in terms of quarks and gluons, whereas all that one observes experimentally are hadrons, very specific combinations of those “elementary” degrees of freedom. Indeed, when the idea of quarks and gluons was first proposed [1], they were considered a mere fictitious tool to try to describe the hadron phenomenology. Nowadays, no one doubts their actual existence, as they leave their footprint in Deep Inelastic Scattering (DIS) experimentswithhadrons,orintheratioR = σe+e−→hadrons/σe+e−→µ+µ−,forexample. Nor 2 does anyone doubt that QCD is the correct theory to explain their dynamics. However, thirty-six years after QCD was vindicated as the theory of strong interactions [2], we still lack a satisfactory analytic description of the hadrons in terms of the degrees of freedom and parameters that appear in its lagrangian. The difficulty resides in the fact that, leaving aside symmetry considerations (or how symmetries are realized), the only quantitative and analytic computational scheme to check the dynamics of QCD from first principles consists in weak-coupling computa- tions. In principle, those are limited to the computation of Green functions in the Deep Euclidean limit. The connection with experiment, however, requires the treatment of non-perturbative effects as well, and to relate those computations done in the Euclidean domain to the physical cut. Non-perturbative effects are taken into account through perturbative factorization techniques. The idea behind this approach is to try to separate the non-perturbative effects from the perturbative ones, dividing our calculations into two pieces: one which we can calculate perturbatively, and another which we leave unevaluated and determine through comparison with experiment or lattice calculations, for example. Essentially, all these factorization techniques are inspired on Wilson’s Operator Product Expansion (OPE)[3]. The basis of the OPE is the application of the following relation in the deep Euclidean region1, i eiqxdxT(A(x)B(0)) q→∞ CAB(q,µ)O (µ), (1) −→ n n Z n X whereA,B aresomelocaloperators,O arelocaloperatorswithincreasingdimensionality n in n and the right quantum numbers to reproduce the left-hand side, CAB are distribu- n tions, and µ is the renormalization scale. The coefficients CAB encode the physics beyond n the scale µ, and the local operators encode the physics below this scale. In QCD the coefficients are calculated using perturbation theory, and the operators are assumed to hold all the non-perturbative physics. This is not completely accurate, however, as per- turbative effects can enter the matrix elements of the operators between the initial and final states, and non-perturbative effects make their way as well into the coefficients (for example, in the form of small-size instantons)[4]. However, this is generally disregarded, and the OPE is used as a series of perturbative coefficients times some matrix elements to be determined experimentally or otherwise. The series is understood to be asymptotic: at some point in the expansion, non-perturbative effects are expected to cause it to break down [5]. Actually, the validity of the OPE is only established in perturbation theory [6]. 1It should be mentioned that the primary definition of the OPE is without the time-ordering. The time-ordering introduces some ambiguities in the definition of the left-hand side of the equation (and consequently on the right-hand side). This is due to the fact that local terms in time are not fully determined (and we should also specify, in principle, in which frame we consider the time evolution). As amatterofprinciple,onemaytrytofixthembyaskingthecorrelatortohavethedesiredtransformation properties under the symmetries of the system. In practice, we will consider the imaginary part of the correlator and obtain the complete result through dispersion relations. This guarantees the desired analytic properties for the correlator. 3 There is no mathematical proof that Eq. (1) can indeed reproduce well the (unknown) exact solution ofQCDforprocesses which involve non-perturbative effects, even accepting its asymptotic nature. The validity of the OPE in these cases is just an assumption. The connection of the OPE with the physical cut can be performed through dispersion relations. This is a well defined procedure, and the sum rules obtained with it are as good as the OPE is. But the OPE, as stated above, is also used to directly compute quantities on the physical cut. If we knew the exact solution to QCD in the Deep Euclidean region (in a finite region) we could safely perform the analytic continuation from there to the physical cut, but as all we have at best are truncated expansions, this procedure can be a source of uncertainties, usually called quark-hadron duality violations (see [7] for a generaldiscussion). Westressthatquark-hadrondualityviolationsareusuallydisregarded without agoodtheoreticalbasis. Typically theyareonlydiscussed, sometimes, inanalysis ofthevacuumpolarization, andeven morescarcely inother processes like DISorB decays like B X γ, see for instance [8,9,10,7,11,12]. Note that in these processes, perturbative s → factorizationtechniques, ortheassociatedeffectivefieldtheorieslikesoft-collineareffective theory, simply neglect duality-violationeffects completely. These effects canbeeasily seen in the large N limit and quantified in the ’t Hooft model (two dimensional QCD in the c large N limit [13]). We do so here for the case of DIS (see [14] for the case of B decays). c The practical version of the OPE (perturbative coefficients times non-perturbative op- erators [5]) is at the basis of computations at large Euclidean momentum of (the moments in) DIS and the vacuum polarization tensor, which so far have been thought to be among the more solid predictions of QCD, since they are not affected by quark-hadron duality problems. Therefore, the importance of setting the OPE and the factorization methods used in quantum field theories, especially in QCD, on solid theoretical ground can hardly be overemphasized. The OPE has been only partially checked in models, for instance in the ’t Hooft model. This theory is superrenormalizable and asymptotically free, so it is a nice ground on which to test the OPE2. This was done at the lowest order in the OPE in Refs. [15,16] for the vacuum polarization and for DIS off a meson with nice agreement between the results of the model and the OPE expectations. In Ref. [17] the OPE was numerically checked in this model at next-to-leading order (NLO) in the 1/Q2 expansion, with logarithmic accuracy, for the vacuum polarization. In Ref. [18] the main results for DIS at NLO were presented. In particular a violation of the OPE was found at NLO in the 1/Q2 expansion. In this paper the details of that computation are presented. The paper is organized as follows. In section 2 we review the ’t Hooft model. We will present the model, the semiclassical approximation to its solution [16,19], and the transition matrix elements for a vector current (in two dimensions one can also obtain from them the matrix elements for the axial-vector current). Insection3westudyDISinthe’tHooftmodel. Wecalculatethefull,non-perturbative expression of the forward Compton scattering amplitude in terms of the ’t Hooft wave functions and energies. As we mentioned, we observe maximal duality violations in the 2In the ’t Hooft model there are no marginal operators. Therefore, the coupling constant has dimen- sions and does not run; no renormalons should then arise. 4 physical cut when comparedwith the expression obtainedfromperturbative factorization. Analytic expressions for the matrix elements with 1/Q2 precision for 1 x β2/Q2 are − ≫ also given. We then compute the forward Compton tensor and expand it in the Deep Euclidean domain with 1/Q2 precision. This result is compared to what we would obtain with theOPE. Onewould expect aperfect agreement at this order. However, surprisingly, we find that our expansion contains, besides the expected local matrix elements, some non-local ones at O(1/Q2), which cannot be part of the OPE. These non-local matrix elements arise from the constructive interference between two (non-analytic) oscillating terms. In section 4 we present our conclusions. In appendix A we present corrections to some formulas of Ref. [14], where we studied duality violations in the context of semileptonic B decays in the ’t Hooft model with 1/m2 precision. Nevertheless, the main conclusion Q of that paper remains unchanged. Namely, one observes no duality violations in the moments with 1/m2 precision. Q 2 QCD in the large N limit 1+1 c The framework formed by QCD in two dimensions in the large N limit is usually called c the ’t Hooft model [13]. This model exhibits confinement: there are no free quarks, and the only states with finite mass are mesons (the mass of baryons grows with N ), c which are composed of exactly one quark and one antiquark, with an infinite ladder of gluons exchanged between them and an infinite series of “rainbow” radiative corrections to their propagators (in the large N limit only planar diagrams with no internal quark c loops contribute, and in an appropriate gauge gluons don’t interact with each other). In two space-time dimensions, the Dirac structure of the lagrangian becomes trivial and gluons can be integrated out, leaving us just with quark fields with no spinor structure. This allows us to solve the meson spectrum, which consists of an infinite tower of in- finitely narrow resonances, due to the large N limit, and features Regge behavior for c large excitations. All of this makes the ’t Hooft model an attractive framework where one can exactly “solve” QCD, and test computational techniques employed in the real world against exact results. In section 2.1 we will first present the appropriate coordinates and quantization frame to treat QCD in 1+1 dimensions, and in section 2.2 we will consider its large N limit, c the ’t Hooft model; in section 2.4 we will present the transition matrix elements for a vector current at leading order in 1/N (from which, in two dimensions, one can obtain c the matrix elements for an axial-vector current). 2.1 QCD in the light front 1+1 The QCD Lagrangian is given by 1 = Ga Ga,µν + ψ¯ (iγµD m +iǫ)ψ , (2) L1+1 −4 µν i µ − i i i X 5 where D = ∂ +igA and the index i labels the flavour. µ µ µ One usually works with Minkowskian coordinates, and quantizes the fields in the equal-time frame, where fields are defined at x0 = constant. However, in some cases a different set of coordinates and a different quantization frame prove to be more useful. In the so-called light-cone coordinates, the Dirac structure of the lagrangian becomes trivial in two dimensions, and once everything is expressed in these coordinates, it is natural to choose a quantization frame in which fields are defined at a constant value of one of the light-cone coordinates, and not x0. This is the light-cone quantization frame [20]. This quantization frame may be convenient when dealing with nearly massless particles. In four dimensions this line of research has been pursued by many groups, see [21] for a review. In two dimensions it can be seen that it is a natural framework on which to solve QCD in the large N limit. 1+1 c 2.1.1 Light-cone coordinates Let us define a basis in 1+ 1 dimensions with the two following light-like vectors (with the metric g+− = g−+ = 2 and zero elsewhere), nµ = (1,1), nµ = (1, 1). (3) − + − Light-cone coordinates are defined like x+ n x = x0 +x1 , x− n x = x0 x1 , (4) + − ≡ · ≡ · − (cid:0) (cid:1) (cid:0) (cid:1) which implies that 1 1 x0 x+ +x− , x1 x+ x− , (5) ≡ 2 ≡ 2 − and (cid:0) (cid:1) (cid:0) (cid:1) ∂ ∂ ∂ ∂ ∂ ∂ ∂− = 2 = + = ∂ +∂ p−,∂+ = 2 = = ∂ ∂ p+, (6) ∂x+ ∂x0 ∂x1 0 1 ∼ ∂x− ∂x0 −∂x1 0− 1 ∼ P+x− P−x+ P x = + , (7) · 2 2 1 dDx = dx+dx−dD−2x . (8) ⊥ 2 For the Dirac algebra it is useful to define the corresping light-cone matrices n/ = γ+, n/ = γ−. (9) + − To have explicit expressions, it is useful to work with an explicit representation of the Dirac algebra. We will use the following Weyl-like representation for the Dirac algebra 0 i 0 i γ0 = − γ1 = , (10) i 0 i 0 (cid:18) (cid:19) (cid:18) (cid:19) 6 so that the corresponding light-cone matrices are given by 0 1 0 0 γ− = γ0 γ1 = 2i γ+ = γ0 +γ1 = 2i . (11) − − 0 0 1 0 (cid:18) (cid:19) (cid:18) (cid:19) We can define as well the following projection operators (γ = γ0γ1), 5 1+γ γ0γ+ 1 1 1 0 Λ 5 = = n/ n/ = γ−γ+ = , (12) + ≡ 2 2 4 − + 4 0 0 (cid:18) (cid:19) 1 γ γ0γ− 1 1 0 0 Λ − 5 = = n/ n/ = γ+γ− = . − ≡ 2 2 4 + − 4 0 1 (cid:18) (cid:19) If we write ψ ψ = + , (13) ψ − (cid:18) (cid:19) then the projection operators separate the two components of the field, ψ 0 Λ ψ = + , Λ ψ = . (14) + 0 − ψ − (cid:18) (cid:19) (cid:18) (cid:19) 2.1.2 The QCD lagrangian in the light-cone frame Once the light-cone coordinates have been defined, the next step is choosing a certain gauge, the light-cone gauge. In light-cone coordinates, the gluonic field is represented by the components Aa,+ n Aa, Aa,− n Aa. (15) + − ≡ · ≡ · The light-cone gauge consists in fixing Aa,+(x) = 0; the reason for this choice will become evident in the next lines. In this gauge the QCD lagrangian in two dimensions can be written like 1 = ∂+Aa,− 2+ ψ† iD−ψ +ψ† i∂+ψ m ψ† ( iψ )+( iψ )†ψ , L1+1 8 i,+ i,+ i,− i,−− i i,+ − i,− − i,− i,+ (cid:0) (cid:1) Xi (cid:16) (1(cid:17)6) where i is the flavour index. Now, quantizing in the light-cone frame consists in defining the fields in this lagrangian at x+ = constant. The coordinate x+ plays therefore the role of time, the role of the energy being played by the conjugated variable P−. The other variables, P+ (and P in four dimensions) are kinematical. For instance, the P+ ⊥ H component of an hadron H behaves in a “free”-particle way, P+ = p+, (17) H i i X where the sum extends over all the partonic components of the bound state. This allows one to define the variable “x ”, which measures the fraction of P+ momentum carried by i H a given parton, p+ x = i . (18) i P+ H 7 In this quantization frame the field ψ is not dynamical (it doesn’t evolve with “time”, − and is therefore a constraint) and can be integrated out, m i ψ = i ψ . (19) i,− i∂+ i,+ In our gauge, gluons, represented by the component A−, are non-dynamical and can be integrated out as well3. After removing the constraints, the resulting Lagrangian can be written like m2 iǫ = ψ† i∂−ψ +i i − dy−ψ† (x−,x+)ǫ(x− y−)ψ (y−,x+) L i,+ i,+ 4 i,+ − i,+ i i Z X X g2 + dy−ψ† taψ (x−,x+) x− y− ψ† taψ (y−,x+) , (20) 4 i,+ i,+ | − | j,+ j,+ ij Z X where we have defined 1 , x < 0, − ǫ(x) = 0 , x = 0, (21)  1 , x > 0.  Once we have the Lagrangian we can construct the Hamiltonian,  m2 iǫ P− = i i − dx−dy−ψ† (x−,x+)ǫ(x− y−)ψ (y−,x+) − 4 i,+ − i,+ i Z X g2 dx−dy−ψ† taψ (x−,x+) x− y− ψ† taψ (y−,x+) . (22) − 4 i,+ i,+ | − | j,+ j,+ ij Z X The representation of the quarks in terms of free fields in the light-cone quantization frame is (note that this assumes that P2 0) ≥ ∞ dp+ ψ (x) = a(p)e−ipx +b†(p)eipx , (23) + 2(2π) Z0 (cid:0) (cid:1) and the anticommuting relations are a(p),a†(q) = b(p),b†(q) = 2(2π)δ(p+ q+), { } { } − a(p),b†(q) = b(p),a†(q) = 0. (24) { } { } The free propagator in the light-cone quantization frame looks like d2k k+γ− + m2i γ+ +m Pf vac T ψf(x)ψ¯f(0) vac = e−ik·xi 2 k+ 2 i , (25) i ≡ h | i i | i (2π)2 k+k− m2 +iǫ (cid:16) (cid:17) Z − i 3Oneshouldnotforgetthatthereisanotherconstraint,theGausslaw,thatrestrictstheHilbertspace of physical states to those which are singlet under gauge transformations. See for instance [22], where one can also find a quantization in the path integral formulation. 8 Figure 1: The infinite series of “rainbow” radiative corrections to the quark propagator. where the f stands for free. The renormalized propagator is given by the infinite sum of one-particle-irreducible diagrams. In the light-cone frame, and in the large N limit, c these diagrams are limited to the rainbow-like diagrams shown in Figure 1, since no gluon lines can cross each other and there is no gluon self-interaction (the gluon lines in that diagram are not truly propagating in our gauge; strictly speaking, all gluon lines should begin and end at the same point in time). However, only the first diagram of this kind contributes, k−p γ+ 1 = iβ2 , (26) 2 k+ p where we have defined β2 g2N /(2π). Adding a gluon line on top of this diagram c ≡ produces a vanishing integral, which kills all the other “rainbow” diagrams. The infinite sum yields + + + ... d2k k+γ− + m2i γ+ +m P (x) = e−ik·xi 2 k+ 2 i . (27) ≡ i (2π)2 k+k− m2 +β2 +iǫ Z − i Recall that this expression is gauge-dependent. Should we have chosen to quantize in the equal-time frame, the expression of the renormalized propagator would be (again in the large N limit) c d2k k+γ− + k+k−+β2γ+ +m P (x)eq.time = e−ik·xi 2 k+ 2 i . (28) i (2π)2 k+k− m2 +β2 +iǫ Z − i The difference between the two propagators is (in momentum space) γ+ 1 P (p)eq.time P (p)light−cone = i , (29) i − i 2 p+ 9 which illustrates the fact that the imaginary part of the propagator is independent of the quantization frame chosen. Note as well that this term is local in “time”, proportional to δ(x+). Such terms would jeopardize the expected covariance of the Green function. Let us note that if, for instance, we consider the OPE of such Green functions at leading order in 1/Q2, we get something proportional to γµq /q2 in the equal-time quantization µ frame but γ−/q− in the light-cone quantization frame. This is not a problem by itself, since the propagator is not a physical quantity. 2.2 The ’t Hooft model By solving the eigenstate equation (taking into account the constraints, and using n to schematically label the quantum numbers of the bound state) P− n = P− n , (30) | i n | i one obtains the basis of states over which the Hilbert space of physical states can be spanned. Here we will focus on the meson sector of the Hilbert space and we will generi- cally label the state as ij;n , where i labels the flavour of the valence quark, j labels the | i flavour of the valence antiquark and n labels the excitation of the bound state. The solution to Eq. (30) in the large N limit gives us the spectrum in the ’t Hooft c model. In this limit the sectors with fixed number of quarks and antiquarks are conserved and consequently the number of mesons; in particular, the sector with only one meson is stable inthelargeN limit. Therefore, the boundstatecan berepresented in thefollowing c way 1 Pn+ dp+ p+ ij;n (0) = φij a† (p)b† (P p) vac , (31) | i √N 2(2π) n P+ i,α j,α n − | i c Z0 (cid:18) n (cid:19) where α is the color index, φij is apwave function representing the bound state, and the n state is normalized as (0)hij;m|i′j′;ni(0) = (2π)2Pn(0)+δmnδii′δjj′δ(Pm(0)+ −Pn(0)+). (32) The superscript (0) stands for the large N limit, and P(0) the eigenvalue of ij;n (0) (we c n | i (0) do not explicitely display the flavour content of P except in cases where it can produce n confusion). φij can also be understood in terms of the gauge invariant “null-plane” matrix element n 1 φinj(x) = N 2(2π) dy−e2iy−Pn+xhvac|ψ¯j,+(0,0)φ(0,y−)ψi,+(y−,0)|i,j;ni(0) , (33) c Z where Φ(x−,py−) is a Wilson line, − Φ(x−,y−) = P[e(igRyx− dz−A+(z−))] . (34) 10