Instanton modifications of the bound state singularity in the Schwinger Model Tomasz Radoz˙ycki Department of Mathematical Methods in Physics, Warsaw University, Hoz˙a 74, 00-682 Warsaw, Poland∗ We consider the quark-antiquark Green’s function in the Schwinger Model with instanton con- tributions taken into account. Thanks to the fact that this function may analytically be found, we draw out singular terms, which arise due to the formation of the bound state in the theory — the massive Schwinger boson. The principal term has a pole character. The residue in this pole containscontributionsfrom variousinstantonsectors: 0,±1,±2. Itisshown,thatthenonzeroones change the factorizability property. The formula for the residue is compared to the Bethe-Salpeter wave function found as a field amplitude. Next, it is demonstrated, that apart from polar part, there appears in the Green’s function also the weak branch point singularity of the logarithmic and dilogarithmic nature. These results are not in variance with the universally adopted S-matrix 7 factorization. 0 0 PACSnumbers: 11.10.Kk,11.10.St,11.55.-m 2 n I. INTRODUCTION lated [3, 4]. This equation is, however, extremely com- a J plicated. Firstly, it is a multidimensional integral equa- tionandthesearealwaysmorechallengingthandifferen- 9 The Schwinger Model (SM) [1], describing a massless 2 fermion field in interaction with a gauge boson in two tial ones. Secondly, being an equation for the unknown space-time dimensions, has become a very fruitful sys- bound state wave function, it requires, as an input, the 3 knowledge of the nonperturbative propagatorsfor ingre- tem in field theory. Above all it possesses many of the v dient particles and the interaction kernel between them. features, one is convinced to find in the theory of strong 2 Naturally no such quantities are known in realistic field 2 interactions. Therefore, thanks to its relative simplic- theories. Fromtheverybeginningoneisthendoomedto 2 ity, SM is expected to be a perfect testing layout for extremely strong approximations. As to the calculations 1 the various nonperturbative aspects of QCD. They in- 0 volve confinement, topological sectors, instantons, con- in field theoretical models, there are only few examples 7 (for instance the so-called Wick-Cutkosky Model [5]), densates. Other interesting and nontrivial features one 0 whereitispossibletofindsolutionsbuteveninthatcase should mention, are the existence of anomaly and gauge / oneisforcedtosimplifytheequationeitherbyneglecting h bosonnonzeromassgenerationwithouttheneedofintro- t ducing an auxiliary Higgs field. Naturally some of these relative time dependence or simplifying the interaction - kernel together with one-particle propagators. p properties are relatedwith one another being in fact dif- e ferent manifestations of the same physical phenomena. The otherpossibilityininvestigatingboundstates,in- h Another aspect, that should be appreciated in SM, stead of solving Bethe-Salpeter equation, is to consider : v is the exact solvability of the equations, which allows four-point Green’s function in the appropriate channel, Xi one to perform many calculations without unpleasant andfind the residue ofthe pole correspondingto a given approximations and additional assumptions or simplifi- bound particle [6, 7]. This residue is constructed just r a cations. One example is the quark propagator(the term fromtheBethe-Salpeterwavefunctions. However,know- ‘quark’ means here the fundamental fermion, due to its ing the full four-point function is usually as unlikely as confinement property),which has alreadybeen found by solving exactly the Bethe-Salpeter equation itself. One Schwinger,butalsootherfunctionsstandwideopen. For should then all the more appreciate the SM, for which instance, in our previous work [2], we were able to ob- the exact form of the Green’s function in question, was tain the explicit formulae for the whole set of four-point found [2, 8]. This is the main reason, why we decided Green’s functions. to look for the bound state singularity in this model, to analyze the nature of this singularity, and to derive the A very troublesome and long-standing question in formulaforthe Bethe-Salpeterfunction. Thecalculation Quantum Field Theory is the consistent description of performed primarily in section III shows the complete bound states. The problem becomes particularly severe agreement with general considerations [6, 7, 9]. when such a state has a relativistic character and re- tardation effects in the interactions among constituent Acertainnewaspectappearshowever,ifonetakesinto particles cannot be neglected. Over half a century account that the SM is a topologically nontrivial the- ago, the equation for the bound state wave function ory [10, 11, 12, 13, 14, 15]. As it is well known, fermion — the so called Bethe-Salpeter equation — was formu- Green’s functions gain contributions from nonzero in- stanton sectors. This is connected with the existence in the theory of the (infinite) set of topological vacua, the superposition of which constitutes the true ground ∗Electronicaddress: [email protected] state. The expectation values for field operators in this 2 newstate,becomenowsumsofmatrixelementstakenbe- residue of the factorizable part of the preceding section. tween various topological vacua. If a given operator has non-vanishing off-diagonal elements, then they have to enter into the calculation,apartfromthe main, diagonal II. BASIC DEFINITIONS AND NOTATION contributions. Distinct topological vacua correspond to different, and topologically inequivalent, configurations The Lagrangian of the Schwinger Model, describing a of the gauge field, and the corresponding transition be- fermion(quark)fieldΨininteractionwithagaugeboson tween them is described by the instanton field. There Aµ, has the following form: is a proportionality between topological index of a vac- uum and the chiral charge and, therefore, off-diagonal (x) = Ψ(x)(iγµ∂µ gγµAµ(x))Ψ(x) L − elements may appear only for the chirality variant oper- 1 λ Fµν(x)F (x) (∂ Aµ(x))2 , (1) ators. To that category belong the products of fermion µν µ − 4 − 2 fields and, consequently, we may expect the appearance where λ fixes the gauge and g is the coupling constant. of certain new terms in the Green’s functions involving The space-time has one time and one space dimension. quarks. The explicit calculations for the two- and four- In this world, the most simply is to choose the quark point functions has been done in [16], where these sup- field in the two-component form. In this case, the 2 2 plementary terms were explicitly given. × Diracgammamatricesmayhavetherepresentationgiven Now the question arises if, and how, these contribu- below: tions modify the behaviour of the two-fermion Green’s function close to the bound state singularity. To this 0 1 0 1 γ0 = , γ1 = − , (2) questionwedevotesectionIVA(oneinstantoncontribu- (cid:18)1 0 (cid:19) (cid:18)1 0(cid:19) tions) and section IVB (two instantons contributions). Our results indicate two modifications. Firstly, we see that the residue in the polar term acquires new ingredi- γ5 =γ0γ1 = 1 0 . ents: theBethe-Salpeterwavefunctionismodified—and (cid:18)0 1(cid:19) − thisisquiteobvious,sinceitisdefinedthroughtheprod- For the metric tensor we choose the convention: uctofthetwoquarkfields,whichissensitivetoinstanton background—andtheothertermappears,whichmakes g00 = g11 =1, the residue nonfactorizable. Yet, it may still be written − as a sum of several factorizable terms. Secondly, apart and the antisymmetric symbol εµν is defined by: fromthepole,wefindterms,whichshowanothertypeof singularity—branchpointsingularityofthelogarithmic ε01 = ε10 =1, ε00 =ε11 =0. (3) − and dilogarithmic nature. These contributions are small All the Green’s functions may be obtained by differenti- in comparison with the principal term, and in the usual atingoverexternalcurrentsofthe generatingfunctional, bound state pole derivations [6, 7], are neglected by the which is, as usually, defined by the Feynman path inte- proof construction itself. In the SM, we happily dispose gral: the full Green’s function without being obliged to make any addition assumptions or simplifications, which usu- allyremainbeyondourcontrol,andthoseterms survive. Z[η,η,J]= Ψ Ψ Aei d2x[L+ηΨ+Ψη+JµAµ] , (4) Z D D D R On the other hand, the calculated instanton con- tributions disappear for transition amplitudes between Aswehavealreadywritteninthe Introduction,this the- asymptotic states and to that extent our results are ory is topologically nontrivial. The true vacuum, which in agreement with the analyticity properties of the S- is usually called the θ-vacuum, becomes a superposition matrix [6, 17, 18, 19, 20, 21, 22]. For the confined parti- of the infinite set of degenerate vacua, bearing different cles one cannot require the cluster decomposition on the topological indices n [23, 24]: level of the Green’s function. The main argument refer- ∞ ring to the possible separation of certain subgroups of θ >= einθ n> . (5) particles into the distant space-time regions, may be ap- | | n=X−∞ plied only for hadrons, but neither for individual quarks nor even qq structures, if they do not bear the complete Inthisnewvacuum,thegeneratingfunctional(4),which set of quantum numbers of a meson. It is then not sur- constitutes the Feynman form of < θ θ > , breaks η,η,J | prising, that the quark Green’s function reveals much into pieces: richer analytical structure, than the hadronic S-matrix ∞ does. Z[η,η,J]= eikθZ(k)[η,η,J]. (6) Finally, in section V we calculate directly the Bethe- k=X−∞ Salpeter wave function as a qq field amplitude between one-particlestateofmomentumP andthevacuumstate, Inthisformula—contraryto(5)—thesummationruns and find that it corresponds to that obtained from the over instanton number, i.e. k corresponds to the change 3 of the topologicalindices in the two vacua in the expres- this sumisnowlimited onlyto the lowvaluesofk. Con- sion: < θ θ > . Henceforth we will use the com- trary to the expression exp i d2x ηΨ+Ψη , which η,η,J | monsymbol 0>,for denoting the true vacuum. Ineach contains arbitrary high powe(cid:8)rsRof Ψ(cid:2)and Ψ an(cid:3)(cid:9)d, there- of Z(k)’s the|functional integration with respect to the fore, contributes to all sectors in (6), the four-fermion gaugefield Aµ, is restrictedto a specific class ofconfigu- function gets terms with at most k = 2. This is due to rations bearing the Pontryaginindex k. the fact, thatfour fermionfields canch±angethe chirality NowtheqqGreen’sfunctionmayberepresentedasthe of the state by (at most) 4, and the topological vacuum appropriate functional derivative of Z: index is a half of the chiral charge [23]. In the following sections we investigate in detail all possible terms. G (x ,x ;x ,x )= (7) ab,cd 1 2 3 4 =<0T(Ψ (x )Ψ (x )Ψ (x )Ψ (x ))0>= a 1 b 2 c 3 d 4 | | δ4Z III. QUARK-ANTIQUARK BOUND STATE =N , δη (x )δη (x )δη (x )δη (x )(cid:12) WITHOUT TOPOLOGICAL CONTRIBUTIONS a 1 b 2 c 3 d 4 (cid:12)η,η,J=0 (cid:12) (cid:12) where N is a constant ensuring that < 00 > = 1. In this section we will concentrate on the 0-instanton η,η,J=0 | In the following, we are only interested in singularities term of the function G (x ,x ;x ,x ). It is a great ab,cd 1 2 3 4 that may be associated with bound states, and hence, merit of the Schwinger Model, that this function may throughoutthis work,byGwe understandonly the con- be found exactly, in an explicit and analytic way. The nected part of this function. details of our calculation have already been given in [2] Due to the fact, that Z expands in the sum of Z(k)’s, andwewillnotelaborateonthemhere,butinsteadrecall the same refers also to the Green’s function. However, only the final result: G(0) (x ,x ;x ,x )= ab;cd 1 2 3 4 1 = S (x x )S (x x )+[S(x x )γ5] [S(x x )γ5] exp ig2[β(x x ) β(x x ) ac 1 3 bd 2 4 1 3 ac 2 4 bd 1 2 1 4 2{ − − − − } { − − − 1 β(x x )+β(x x )] + S (x x )S (x x ) [S(x x )γ5] [S(x x )γ5] 2 3 3 4 ac 1 3 bd 2 4 1 3 ac 2 4 bd − − − } 2{ − − − − − } x x ×exp{−ig2[β(x1−x2)−β(x1−x4)−β(x2−x3)+β(x3−x4))]}−(cid:26) c3 ↔ d4 (cid:27) , (8) ↔ where the index (0) refers to the instantonnumber. The mustlater getcontributionsfromthe nontrivialtopolog- functionβ(z)isdefinedthroughthetwo-dimensionalmo- ical configurations [16]. mentum integral: Now we will concentrate on the structure of the qq d2p 1 eipz function (8) and try to reveal, whether it has a pole in β(z)= − , (9) the t-channel. To this goal we introduce new space-time Z (2π)2(p2 g2(cid:0)/π+iǫ)((cid:1)p2+iǫ) − variables: X, Y, x and y, defined in the following way: and may be expressed in terms of MacDonald or Hankel function, depending on whether z is space- or time-like, 1 but the above representationis more suitable for us. We X = (x +x ), x=x x , 1 3 1 3 recallthatin2Dthecouplingconstantg isdimensionfull 2 − and the quantity µ2 = g2/π plays a role of the mass 1 Y = (x +x ), y =x x . (11) 2 4 2 4 squared of the so called Schwinger boson – a qq bound 2 − state we are interested in. The object S(x) in (8) is the full quark propagator, Thanks to the translational invariance, which manifests which has the known form: itself through the dependence of G on x x ’s only, we i j − S(x)= (x)exp ig2β(x) , (10) have three independent variables: Z = Y X, x and S0 − − (cid:2) (cid:3) y. From the four terms present in (8), the two explic- with (x) = 1 6x being the free propagator. The itly written are of interest for us, since the two other, S0 −2πx2−iε term‘full’usedabove,meanshere‘fullinthe0-instanton obtained by the antisymmetrization, contribute to the sector’, i.e. it is given in the form of the original singularities not in t, but eventually in the u-channel. Schwinger works [1]. However S – bilinear in Ψ’s — First of the two interesting terms, denoted by I, takes 4 the form: eventual singularity is cancelled by the sine func- tions present in the numerator in (16), either by 1 G(0)I(Z;x,y)= S(x) S(y)+[S(x)γ5] [S(y)γ5] that explicitly written or by the one hidden in the 2{ ⊗ ⊗ } formula for I . If both parameters are nonzero, on exp ig2[β( Z +(x y)/2) β( Z +(x+y)/2)) 1 2 × { − − − − can easily obtain q2 = α+β p2 =0. Thus, there β( Z (x+y)/2))+β( Z (x y)/2))] , (12) β − − − − − − } may only arise a contri(cid:16)butio(cid:17)n to the singularity at where we made use of the evenness of β: β(x) =β( x). q2 = 0. It cannot, however, be worse than that in − For brevity we omitted the spinor indices, and used the I . 1 tensorproductnotation,whichshouldnotcauseanycon- 2. α(p2 µ2)=0andβ(p q)2 =0. Ifα=0,wehave fusion. Thepolecorrespondingtotheboundstateshould − − q =p and, ingeneral,one mightgeta contribution befoundinthecomplexplaneofP2,whereP isthetwo- to the pole at q2 = µ2 (actually it would become momentum canonically conjugated to Z. The matrical a branchpoint), but it does nothappen because of partdependsonlyontherelativecoordinatesxandyand the sine functions becoming zero for q p = 0. If entersintotheresidueofthispole,butdoesnotinfluence − β = 0, we obtain p = 0, which stays in contradic- its position. Let us then perform the Fourier transform tion with the assumption p2 = µ2. For α,β = 0, over Z of the exponent containing β functions. Making we find α2p2 = β2(p q)2, which is incompa6tible use of (9) we my write the quantity in question as: − with p2 =µ2 and (p q)2 =0. − FI(P,x,y)= d2ZeiPZe−4ig2 (2dπ2)p2e−ipZI1(p,x,y) , 3. αp2 = 0 and β((p q)2 µ2) = 0. This case is − − Z R almost identical with the previous one and need (13) not be separately considered. where I (p,x,y) = sin(px/2)sin(py/2). Expanding the ex- 1 (p2−µ2+iǫ)(p2+iǫ) 4. α(p2 µ2)=0andβ[(p q)2 µ2]=0. Ifoneofthe ponent in series, we can execute the Z integration ob- − − − parameters is zero, then we find a contradiction of taining: αp+β(p q)=0,witheitherp2 =µ2 or(p q)2 = FI(P,x,y)= ∞ (−4ig2)nI (P,x,y)+(2π)2δ(2)(P), wµ2e.haWvehaα−t2pr2em=aβin2s(pis tqo)2asasnudmaeftαer,βca6=nce0l.l−inIgftshoe, n nX=1 n! arisingfactorsµ2onb−othsides,wegetα=β. This (14) leads to q = 2p and a singularity emerges now at The In’s satisfy the following recurrent relation: q2 =4µ2. This is the two-particle singularity. In (q,x,y)= (15) Summarizing,weobservethatinI2 thepoleatµ2 disap- pears and a new singularity arises at (2µ)2. Now we are d2p sin(px/2)sin(py/2) = I (q p,x,y). in a position to carry out a proof by induction: the as- Z (2π)2(p2−µ2+iǫ)(p2+iǫ) n−1 − sumedsingularitiesundertheintegralin(15)are: p2 =0, It is obvious that I1(p,x,y) has a pole at p2 = µ2. The p2 =µ2,(p−q)2 =0and(p−q)2 =((n−1)µ)2. Wehave other singularity is at p2 = 0. But what about terms of again four possibilities, the analysis of which is identical higher n? Do they alter the behaviour at p2 = µ2? Let as above, except the last point, where we now get the equation α2 = (n 1)2β2. This leads to q = np and us firstconstructI2(p,x,y). Accordingto (15) ithasthe q2 =n2µ2. Hence,−the conclusion is, that the singularity form: corresponding to n intermediate particles (q2 =n2µ2) is I2 (q,x,y)= (16) present only in In. This means, that the pole at µ2 may be simply picked up from the first term of the expan- d2p sin(px/2)sin(py/2) = I (q p,x,y). sion (14). Z (2π)2(p2 µ2+iǫ)(p2+iǫ) 1 − − Now we need to apply the same to the second expres- Thepossiblesingularitiesinq2,exhibitedinI ,cancome sion in the Green’s function (8). Again we take out the 2 exponent and perform the Fourier transform, the result- from the singularities of the integrand functions. There are four of them: obliviously p2 = 0, p2 = µ2, and the ing expression being called FII(P,x,y). Happily both othertwooriginatingfromI . Theseare: (p q)2 =0and exponents differ only by the sign, so we are able to im- 1 (p q)2 = µ2. It is not possible to satisfy −all these four mediately write down the final result. The total contri- − bution (i.e. I+II) to the bound state pole in G(0) is: or even three conditions simultaneously, so the singular points may appearonly asa resultofthe mergingoftwo G(0)I(P;x,y)= 4iπ S(x)γ5 S(y)γ5 singularities. In accordance with Landau procedure [6, b.s. − ⊗ sin((cid:2)Px/2)si(cid:3)n(P(cid:2)y/2) (cid:3) 25,26],weintroducetwoparameters(αandβ)satisfying , (17) α,β 0 and α2 +β2 = 0. With the assumption αp+ × P2 µ2+iǫ ≥ 6 − β(p q)=0, the following possibilities occur: We see the perfect factorization of the residue, and the − quantity: 1. αp2 =0andβ(q p)2 =0. Ifoneoftheparameters equalszero,then−we getp=0 or p q =0 andthe Φ (x)= 2√πS(x)γ5sin(Px/2), (18) P − − 5 plays a role of the Bethe-Salpeter amplitude [3, 4, 7] or dom, corresponding to the transition amplitude in ques- the wave function. The minus sign is chosen for later tion, is then zero and the tunnelling phenomenon disap- convenience. Note, that this quantity is found explicitly pears. This does not mean, however, that the notion of and exactly, without any simplification referring neither θ-vacuum is useless, since topologicalvacua (contrary to to the relative time x0 nor to the structure on the in- the θ-vacuum) violate cluster property [30]. We are the teraction kernel. It is an essential point, since even in fated to use (5) and must put up with modification in the simplified model theories, one usually is limited to the fermion Green’s functions the case of the so called ‘equal time wave function’ (i.e. The contributions to the four-point function were al- defined only on the hypersurface x0 = 0 in the center ready found in our previous work [16]. Below we will of mass frame), which neglects retardation effects and concentrate on the eventual bound state pole in the for- consequently leads to the nonrelativistic bound state, or mulae. ladder approximation for the interaction kernel. On the otherhand,thefunction(18)isafullyrelativisticobject. A. 1-instanton sector IV. TOPOLOGICAL CONTRIBUTIONS TO THE QUARK-ANTIQUARK BOUND STATE By virtue on the Atiyah-Singer index theorem [24, 27, Asithasbeenmentioned,thevacuumintheSchwinger 28, 29, 31], the massless Dirac operator in the external Model is nontrivial: it is a superposition (5) of vacua fieldbearingnonzeroinstantonnumberkhaszeromodes. correspondingto varioustopologicalnumbers. As a con- In the present section we consider the case of k = 1, ± sequence, all chirality changing operators— and to that for which there is only one such mode. Its detailed form classthe productoffermionfields belongs—whichhave may be found in the literature [10, 13, 32]) and we will nonvanishingmatrix elements betweendifferent topolog- not recall it here. This mode gives now its contribution ical vacua, acquire contributions from nonzero instanton while calculating the Feynman path integral (4). Since sectors. We recallhere, thatif the theorycontainsmass- it corresponds to the zero eigenvalue, it disappears from less fermions (at least one) — and this is just the case the quadratic part of the Lagrangian and remains only of the Schwinger Model — the amplitude of the spon- in source terms. Due to the Grassman character of the taneous tunnelling between different topological vacua quark fields, of the four differentiations in (7), two must vanishes[23,27]. Mathematicallyitisexpressedthrough be saturated by sources multiplying zero modes. The zero value of the euclidean Dirac operator determinant: other two are performed in a common way. The result det[i∂ eA],becauseofthe appearanceofzeroeigenval- forthefour-pointfunction,withbothk=+1andk= 1 uesw6he−nA6µbearstheinstantonindex[24,27,28,29,31]. instanton numbers taken into account, obtained in [1−6], The functional integral over fermionic degrees of free- is: ig G(1)(x1,x2;x3,x4)= −8π3/2eγE(cosθ−iγ5sinθ)⊗S0(x2−x4) (11⊗11−γ5⊗γ5) (cid:2) eig2[β(x1−x3)−β(x2−x4)−β(x1−x4)+β(x1−x2)+β(x2−x3)−β(x3−x4)]+(11 11+γ5 γ5) (19) × ⊗ ⊗ eig2[β(x1−x3)−β(x2−x4)+β(x1−x4)−β(x1−x2)−β(x2−x3)+β(x3−x4)] + antisymmetrization. × (cid:3) The antisymmetrization runs here both over ‘initial’, again contributes to the pole in the t-channel, the other i.e. x and x and ‘final’ i.e. x and x (accompanied twoonlyintheu-channelandwillbe omitted. Introduc- 1 2 3 4 by the appropriate spinor indices change) variables in ing center of mass and relative variables (11), similarly G(x ,x ;x ,x ). Thus(19)getsthreemoreterms,which as it was done in section III, we obtain for the first term 1 2 3 4 are not explicitly written here for brevity. One of them in (19): (with the simultaneous substitution 1 2 and 3 4) ↔ ↔ ig G(1)Ia(Z;x,y)= eγE (cosθ iγ5sinθ)eig2β(x) (y) (11 11 γ5 γ5) (20) −8π3/2 − ⊗S ⊗ − ⊗ (cid:2) (cid:3) exp ig2[ β( Z+(x+y)/2)+β( Z+(x y)/2))+β( Z (x+y)/2)) β( Z (x y)/2))] , × { − − − − − − − − − − } 6 The last exponent in the above, per analogiam to (13), the exponent has only the sign invertedand need not be may now be written in the form: elaboratedon. Their joint effect is then simply to cancel 11 11 in (19) and to double γ5 γ5. Their contribution e4g2 (2dπ2)p2e−ipZJ1(p,x,y) , (21) to⊗the bound state is therefore:⊗ R where J1(p,x,y) = (cpo2s(−pµx2/+2)iǫsi)n(p(p2y+/i2ǫ)). Now we might in- G(b1.s).I(P;x,y)= iµceoγEs((cid:2)Peixg/2β2()xs)ien−(iPγ5yθ/γ25)(cid:3)⊗(cid:2)S(y)γ5(cid:3) troduce J ’s and consider the recurrent relation similar n , (22) to (15), but it is unnecessary. Due to the presence of × P2 µ2+iǫ − thesinefunctioninthenumerator,wecandirectlyapply theanalysisofsectionIIIandcometotheresultthatthe where,insteadofcosθ iγ5sinθ,wehavewrittene−iγ5θ. − contributiontotheboundstatecomesonlyfromthefirst The antisymmetrized term, which is left so far, has an termoftheseriesfortheexponent(21). Astothesecond analogous form, but with x and y swopped, and tensor term (in our nomenclature it would be called Ib) in (19) products inverted. Our final result in this sector is: G(1)(P;x,y)= iµeγE eig2β(x)e−iγ5θγ5 S(y)γ5 b.s. P2 µ2+iǫ ⊗ − (cid:8)(cid:2) (cid:3) (cid:2) (cid:3) cos(Px/2)sin(Py/2)+ S(x)γ5 eig2β(y)e−iγ5θγ5 sin(Px/2)cos(Py/2) . (23) × ⊗ (cid:2) (cid:3) (cid:2) (cid:3) (cid:9) Thankstothequarkfieldbeingmassless,whichresultsin As we shall see in the following section, this is actually thechiralinvarianceoftheinitialLagrangian,thedepen- the case, but it contributes also other terms that modify dence on the parameter θ may be gauged away, so one the factorization and even change the character of the might simply put θ = 0. Comparing the two obtained singularity at P2 =µ2. equations (17) and (23) we see that introducing a new Bethe-Salpeter amplitude in the form: Φ (x)= 2√πS(x)γ5sin(Px/2) (24) P − µ + eγEeig2β(x)e−iγ5θγ5cos(Px/2), 2√π the residue factorization in the pole will be restored i.e. one will have: B. 2-instanton sector Φ (x) Φ (y) P P G (P;x,y) ⊗ (25) b.s. ∼ P2 µ2+iǫ − In the 2-instanton sector there are two zero modes of if the 2-instanton sector contributes to the bound state thetheDiracoperator[10,13,27,28,32]. Nowallofthe a term (with the appropriate constant coefficient): differentiationsin(7)aresaturatedwithsourcesstanding with zero modes. The 2-instanton contribution to the [eig2β(x)e−iγ5θγ5] [eig2β(y)e−iγ5θγ5] Green’s function, which was evaluated in [16], has the ⊗ cos(Px/2)cos(Py/2). form: × g4 G(2)(x1,x2;x3,x4)=−256π4e4γE e2iθ 11−γ5 ⊗ 11−γ5 (x02+x12) (26) (cid:2) (cid:0) (cid:1) (cid:0) (cid:1) ×(−x04+x14)+e−2iθ 11+γ5 ⊗ 11+γ5 (−x02+x12)(x04+x14) exp ig2 β(x1−x4) +β(x x )+β(x (cid:0)x )+(cid:1)β(x(cid:0) x )+(cid:1) β(x x )+β(x (cid:3)x ) (cid:8)+(cid:2)antisymmetrization. 2 3 1 2 3 4 1 3 2 4 − − − − − (cid:3)(cid:9) Using again the variables Z, x and y, performing the required antisymmetrization, and replacing β functions with integral formula (9), we obtain, with the additional use of the identity e−iθγ5(11 γ5)=e∓iθ(11 γ5): ± ± 7 µ4 1 G(2)(Z;x,y)= e4γE (eig2β(x)e−iγ5θ) (eig2β(y)e−iγ5θ) (Z2 (x y)2)γ5 γ5 −64π2 ⊗ × − 4 − ⊗ (cid:2) (cid:3) (cid:2) 1 1 + (Z2 (x y)2) (11 γ5) (11+γ5)+(11+γ5) (11 γ5) (27) 2 − 4 − − ⊗ ⊗ − (cid:0) (cid:1) 1 d2p 1 e−ipZcos(px/2)cos(py/2) −2εαβZα(x−y)β(11⊗γ5+γ5⊗11) exp 4πiµ2Z (2π)2 −(p2+iǫ)(p2 µ2+iǫ) , (cid:3) (cid:2) − (cid:3) with εαβ being the antisymmetric symbol defined in (3). This is exactly what we needed for (25). This is not, Now we will show that the first term (i.e. Z2 times however,thewholestory. Thesimilarcontributioncomes γ5 γ5) gives just the contribution needed to restore from the second matrical term in (27), which also con- the⊗factorization (25). tains Z2. Since we have already saturated (25) this sup- One could suspect that there might be some cancel- plementary part violates the factorization. It has the lations, among various expressions in the formula (27). form: Below we shall argue that the cancellation that would iµ2 lead to complete disappearance of any term, is not pos- − e2γE(eig2β(x)e−iθγ5 eig2β(y)e−iθγ5) 8π ⊗ sible due to their independent matrical structure. Only cancellationsamongaspecificmatrixelements—butnot [(11 γ5) (11+γ5)+(11+γ5) (11 γ5)] × − ⊗ ⊗ − matrices as a whole — are admissible. Consider namely cos(Px/2)cos(Py/2) . (31) the expression: × P2 µ2+iǫ − aγ5 γ5 + b[(1 γ5) (1+γ5)+(1+γ5) (1 γ5)] Due to the above mentioned lack of cancellations among ⊗ − ⊗ ⊗ − + c(11 γ5+γ5 11) (28) independent matrix structures, this term must survive. ⊗ ⊗ Moreover it cannot cancel with the (x y)2 term with andmultiplyitfirstlyby(11+γ5) (11+γ5)andsecondly − ⊗ the same matrical coefficient since, contrary to (31), the by(11 γ5) (11 γ5). Weobtain(a+2c)(11+γ5) (11+ γ5)+b−0an⊗d(a−2c)(11 γ5) (11 γ5)+b 0respec⊗tively. latter disappears for x = y. Besides, it does not con- · − − ⊗ − · tribute to the bound state pole at all. This can be On the other hand if we take the double trace (i.e over easily argued as follows. After rearrangements identi- firstmatrixandoversecondmatrixinthetensorproduct calto (29)one getsthe additional1/Z2 comingfromthe indices), we get a 0+8 b+0 c. This shows that the · · · logarithm in the exponent. This can be neutralized, if three tensors listed in (28) are really independent. we act with the two-dimensionald’Alambert operator in Letusobservenow,thatthequantityinthelastexpo- variable P. When this is done, we obtain a first order nent of (27), after rearranging the terms under the inte- pole at P2 = µ2. Now the question is: given a func- gral(the integrandfunction willbe calledbelowM) and tion f satisfying: ∂2f(P) 1 . Does f also dis- partially executing it, may be rewritten in the following P ∼ P2−µ2 play a pole? The answer is negative. It may easily be way: found that f(P) Li (P2/µ2) and dilogarithm has a 2 d2p limit π2/6 when i∼ts argument goes to unity. Therefore 4πiµ2 M(Z,p,x,y)= 2γ ln( µ2Z2/4) Z (2π)2 − E− − it may be neglected in comparison with the polar term. One should emphasize, however, that although diloga- d2p 1 4πi e−ipZ (29) rithmic function has a well defined limit, it is not an an- − Z (2π)2 p2 µ2+iǫ − alyticfunction inthis point. Itisa branchpointandthe d2p 1 cos(px/2)cos(py/2) function has a cut on the half line [µ2, [. The presence +4πiµ2 e−ipZ − , ∞ Z (2π)2 (p2+iǫ)(p2 µ2+iǫ) ofcos(Px/2)cos(Py/2)does not changethe conclusions, − since cosine is an analytic function, and may only make If we expand, as before, the exponent of the last two things better. On the other hand for x = y = 0 cosines terms and take the Fourier transform over Z, it be- can be replaced with unities and a branch point arises. comes obvious that both of them contribute to the pole Such terms are usually lost in the common approach, at P2 = µ2. The further analysis is again identical to where one fishes out only the polar contribution of the that of Sec. III, except for the difference, that now there bound state. is additional cancellation between the integrals in (29). ForthetermproportionaltoεαβZ (x y) ,similarar- Finally all the factors combine together in (27) and we α − β gumentationis applicable, with a difference that we now obtain the expected term: have to do with logarithmic function, which again may −iµ2e2γE(eig2β(x)e−iθγ5γ5) (eig2β(y)e−iθγ5γ5) beneglectedifcomparedtothepolarterm. Hence,aswe 4π ⊗ see, due to the topological background,not only nonfac- cos(Px/2)cos(Py/2) torization appears, but the character of the singularity . (30) × P2 µ2+iǫ in P2 =µ2 is slightly changed. − 8 Theseresultsdonot,however,contradicttheS-matrix we now have to do with an object billinear in quark pole factorization [17, 18, 19, 20, 21]. The S-matrix de- fields). We now pass to the CM and relative coordi- scribesthetransitionamplitudesamongphysicalasymp- nates: x = X + x/2 and x = X x/2, and intro- 1 2 − totic states. In the Schwinger Model these are the Fock duce a wave function by separating the uniform motion states of Schwingerbosons. Consequently we are limited of the bound particle as a whole, from the internal mo- onlytothecurrent-currentGreen’sfunctionsandthisre- tion: Φ (x)=eiPXχ(x ,x ). Finally we obtain: P 1 2 quire putting x,y 0, and taking the trace over spinor indices with γµ γ→ν matrices. Since the trace of an odd ΦP(x)= 2√πS(x)γ5sin(Px/2) number of γ’s is⊗zero, it is easy to observe that the only −+ µ eγEeig2β(x)e−iθγ5γ5cos(Px/2).(37) contributiontothepolecomesnowfrom(17)andfactor- 2√π ization reappears [10]. This is what should be expected since the vector current does not couple to instantons. This amplitude wave function entirely agrees with that Summarizing,thatpartofthefullqq Green’sfunction, found in (24) from the factorizable part of the bound in which one is limited to the polar terms, has the form: state residue. 1 iG (P;x,y)= [Φ (x) Φ (y) (32) b.s. P2 µ2+iǫ P ⊗ P VI. SUMMARY − +Ψ(−)(x) Ψ(+)(y)+Ψ(+)(x) Ψ(−)(y)], P ⊗ P P ⊗ P In conclusion we would like to recapitulate our main with results. We have found analytically the polar contribu- µ tiontotheqq Green’sfunction. Inthetopologicallytriv- Ψ(±)(x)= eγEeig2β(x)e−iθγ5(11 γ5), (33) P 2√2π ± ialcase,theresidueturnsouttobe factorizable,whichis commonlytakenforgranted. Outofthisweobtainedthe and Φ (x) given by (24). formulafortheBethe-Salpeterwavefunction. Thesitua- P tionturnedouttobedifferent,ifoneconsidersoneortwo instanton background. We have obtained expression for V. WAVE FUNCTION AS A FIELD the residue in this case and found, that the BS function AMPLITUDE has been modified by the θ-vacuum structure. The full polar term may now be written as a sum of several fac- The question arises, how the obtained results harmo- torizableterms. Oneseesthat,apartfromthe formation nizewiththeBethe-Salpeterwavefunctiondefinedasan oftheboundstate(Schwingerboson),thereappearsalso amplitude of fields. Fortunately in the Schwinger Model another way for a pair qq, not bearing meson quantum wehavealltools,whichallowustofinditexplicitlyeven numbers, to travel over the space. They may jump from inthenontrivialtopologicalsectors. Letusthenconsider one instanton to the other. This phenomenon is possible the amplitude in question: in there are at least two instantons. Moreover if one takes into account also these compo- χP(x1,x2)=<0T(Ψ(x1)Ψ(x2)P > . (34) nents, that are usually neglected for P2 µ2 as being | | → smallin comparisonwiththatrevealinga pole,one finds Thecalculationofthisquantityturnsouttobeverysim- another type of singularity — a branch point connected ple,ifonealreadyknowstheexpressionforthefullquark with the appearance of logarithmic and dilogarithmic propagator in the instanton background[16]: functions. They arise because of long-range correlations in the condensed vacuum. ie Sfull(x)=S(x)+ 4π3/2 cosθ−iγ5sinθ eγE+ie2β(x) . Alltheseinstantoncontributionsvanish,ifweconsider (cid:0) (cid:1) (35) theS-matrixelements,i.e. onlytransitionamplitudesfor the‘physical’asymptoticstates. These,intheSchwinger The state P > is the one Schwinger boson state of two- momentum| P, and (34) is expressible,thanks to the for- Model, are Fock states of one or more bosons. The four- point(quark)function is thenreplacedwith the current- mulaeoftheLSZformalism,throughthevertexfunction. current function, for which it is shown, that simple pole It is known that, thanks to the two gauge symmetries structureaswellasfactorizationofthe residuereappear. (U(1) and U (1)), the vertex function may in turn en- A TheanalyticalstructureoftheGreen’sfunctionmaythen tirely be constructedfromthe quark propagator. This is bemoreintricatethanthatoftheS-matrix,especiallyfor a miracle of the Schwinger Model. One then gets: topologically nontrivial theory. χ (x ,x )= i√π e−iPx1γ5S(x x ) P 1 2 1 2 − +e−(cid:2)iPx2S(x x )γ5 . (36) 1− 2 Acknowledgments (cid:3) If we now substitute for the quark propagator the ex- pression S , we obtain the formula for the amplitude I would like to thank to ProfessorsJ´ozef Namys lowski full χ inallinstantonsectorsinquestion: 0, 1(recallthat and Krzysztof Meissner for elucidating discussions. P ± 9 [1] J. Schwinger, in Theoretical Physics, Trieste Lectures Berkeley, UCRL-11961 Report (unpublished). 1962 (I.A.E.A. Vienna 1963), p. 89; Phys. Rev. 128, [18] W. Zimmermann, N. Cim. 13, 503(1959). 2425(1962). [19] D.I. Olive, Phys. Rev. 13, B745(1964). [2] T. Radoz˙ycki and J. Namysl owski, Phys. Rev. D59, [20] J.C. Taylor, Phys. Rev.(142, 1236(1966). 065010(1999). [21] D.S. Kershaw, Phys.Rev. D5, 1976(1972). [3] E.E. Salpeter and H.A. Bethe, Phys. Rev. 84, [22] D.Iagolnitzer,TheSmatrix,North-Holland,Amsterdam 1232(1951). 1978. [4] M. Gell-Mann and F. Low, Phys. Rev. 84, 350(1951). [23] R. Rajaraman, An Introductions to Solitons and Instan- [5] G.C. Wick, Phys. Rev. 96, 1124(1954); R.E. Cutkosky, tonsinQuantumFieldTheory,North-Holland,NewYork Phys.Rev.96, 1135(1954); 1982. [6] R.J. Eden, Proc. Roy.Soc. A 217, 390(1953). [24] K.Huang,Quarks,LeptonsandGaugeFields,WorldSci- [7] S.Mandelstam, Proc. Roy.Soc. A 233, 248(1955). entific 1982. [8] L. S.Brown, N.Cim. 29, 617(1963). [25] L.D. Landau, Nucl. Phys. 13, 181(1959). [9] E.H. Wichmann and J.H. Crichton, Phys. Rev. 132, [26] R.J.Eden,P.V.Landshoff,D.J.Olive,J.C.Polkinghorne, 2788(1963). The Analytic S-Matrix, Cambridge University Press, [10] C. Adam, Z.Phys. C63, 169(1994). Cambridge 1966. [11] A.V.Smilga, Phys. Rev. D49, 5480(1994). [27] G. ’t Hooft, Phys. Rev. D14, 3432(1976). [12] C. Gattringer, hep-th/9503137. [28] M. Atiyahand I.Singer, Ann.Math. 87, 484(1968). [13] G. Maiella and F. Schaposnik, Nucl. Phys. B132, [29] J. Kiskis, Phys. Rev. D15, 2329(1977). 357(1978). [30] C.G. Callan, R.F. Dashen and D.J. Gross, Phys. Lett. [14] K.D.RotheandJ.A.Swieca,Ann.Phys.117,382(1979). 63B, 334(1976). [15] G. McCartor, Int. J. Mod. Phys. A12, 1091(1997). [31] M.M. Ansourian, Phys. Lett. 70B, 301(1977). [16] T. Radoz˙ycki, Phys. Rev. D60, 105027(1999). [32] C. Jayewardena, Helv.Phys. Acta61, 636(1988). [17] J. Crichton, Ph.D. Thesis, University of California,