CAFPE/70-06 LU TP 06-03 UG-FT/200-06 hep-ph/0601197 January 2006 6 0 0 2 n a The B Kaon Parameter in the Chiral Limit J K 4 2 1 v Johan Bijnensa), Elvira G´amizb) and Joaquim Pradesc) 7 9 1 a) Department of Theoretical Physics, Lund University 1 S¨olvegatan 14A, S-22362 Lund, Sweden. 0 6 0 b) Department of Physics & Astronomy, University of Glasgow / h Glasgow G12 8QQ, United Kingdom. p - p c) Centro Andaluz de F´ısica de las Part´ıculas Elementales (CAFPE) and Departamento e h de F´ısica Te´orica y del Cosmos, Universidad de Granada : v Campus de Fuente Nueva, E-18002 Granada, Spain. i X r a Abstract We introduce four-point functions in the hadronic ladder resummation approach to large N QCD Green functions. We determine the relevant one to calculate the B c K kaon parameter in the chiral limit. This four-point function contains both the large momenta QCD OPE and the small momenta ChPT at NLO limits, analytically. We get Bˆχ = 0.38±0.15. We also give the ChPT result at NLO for the relevant four- K point function to calculate B outside the chiral limit, while the leading QCD OPE K is the same as the chiral limit one. 1 Introduction Indirect kaon CP-violation in the Standard Model (SM) is proportional to the matrix element hK0|K0i = −iC C(ν)hK0| d4yQ (y)|K0i ∆S=2 ∆S=2 Z 16 ≡ −iC Bˆ F2m2 (1) ∆S=2 3 K K K with Q (x) ≡ 4Lµ(x)L (x); 2L (x) ≡ [sγ (1−γ )d](x). (2) ∆S=2 µ µ µ 5 and C(ν) a Wilson coefficient. C(ν) and C are known in perturbative QCD to next- ∆S=2 to-leading order (NLO). A review can be found in [1]. The kaon bag parameter, Bˆ , defined in (1) is an important input for the unitarity K triangle analysis and its calculation has been addressed many times in the past. There have been four main QCD-based techniques used to calculate it: QCD-Hadronic Duality [2, 3], three-point function QCD Sum Rules [4], lattice QCD and the 1/N (N = number c c of colors) expansion. Recent reviews of the unitarity triangle, where the relevant references for the inputs can be found, are [5]. For recent advances using lattice QCD see [6, 7, 8]. Here, we present a determination of the Bˆ parameter at NLO in the 1/N expansion. K c That the 1/N expansion introduced in [9, 10] would be useful in this regard was first sug- c gested by Bardeen, Buras and G´erard [11] and reviewed by Bardeen in [12, 13]. There one can find most of the references to previous work and applications of this non-perturbative technique. Work directly related to this work can be found in [14, 15] where a NLO in 1/N calcu- c lation of Bˆ within and outside the chiral limit was presented. There, the relevant spectral K function is calculated at very low energies using Chiral Perturbation Theory (ChPT), at intermediate energies with the extended Nambu-Jona-Lasinio (ENJL) model [16] and at very large energies with the operator product expansion (OPE). The method, including how to deal with the short-distance scheme dependence, has been discussed extensively in [15]. The improvement in the present work is basically in the intermediate energies and the numerical quality of the matching to short distances. Another calculation of Bˆ in the chiral limit at NLO in the 1/N expansion is in [17]. K c There, the relevant spectral function is saturated by the pion pole and the first rho meson resonance –minimal hadronic approximation (MHA). In [18], the same technique as in [17] was used but including also the effects of dimension eight operators in the OPE of the ∆S = 2 Green’s function and adding the first scalar meson resonance to the relevant spectralfunction. Thisworks differs fromthatintwo aspects. Weuseour X-bosonmethod which allows for aneasy identification of the precise scheme used in the hadronic picture by only using currents and densities directly in four dimensions rather than using dimensional regularization throughout. This is a difference in formalism, not in physical content. On the other hand, we use a much more elaborate way to include hadronic states allowing 1 the use of more constraints [19] and thus more resonances than the work of [17, 18]. The hadronic model developed in [19] also includes current quark masses effects and will be ˆ used to determine B in the real case [20]. K The present manuscript is organized as follows. In Section 2 we collect a few perturba- tive formulas at NLO in QCD for completeness. Section 3 explains in detail how we derive the value of Bˆ from a large N Green’s function with two densities and two currents. K c This can be found essentially in [15] but we add here an extra QCD calculation of the short-distance constraint analogous to the equivalent terms in our work on Q and Q [21]. 7 8 QCD is still not solved even at leading order in the 1/N expansion. Therefore, at present, c any NLO in 1/N approach to weak matrix elements needs a large N hadronic model. c c We present in Section 4 our hadronic model based on a severe approximation to a ladder resummation of QCD as explained in [19]. This model is used at intermediate distances and is the main uncertainty in our approach. Then we discuss the chiral limit results in Section 5. The extension beyond the chiral limit needs more work both at intermediate distances but we include some comments already in Section 6. At short distances, the leading dimension six OPE operator is known and agrees with the chiral one. The last section contains our conclusions and a comparison with other approaches. Some preliminary results were already presented in [22] and [23]. 2 Some perturbative QCD formulas This section is a compilation of the perturbative QCD coefficients at NLO from [1] needed in our analysis of Bˆ and derived in [24]. K The coefficients C(ν) and C are known in perturbative QCD at next-to-leading ∆S=2 order (NLO) in a ≡ α /π in two schemes [1, 24], the ’t Hooft-Veltman (HV) scheme (MS S subtraction and non-anti-commuting γ in D 6= 4) and in the Naive Dimensional Regu- 5 larization (NDR) scheme (MS subtraction and anti-commuting γ in D 6= 4). C col- 5 ∆S=2 lects known functions of the integrated out heavy particle masses and Cabibbo-Kobayashi- Maskawa matrix elements, it can be found in [1, 24]. Its actual form is very important for CKM analysis but not needed here. The Wilson coefficient C(ν) is γ β γ C(ν) = 1+a(ν) 2 − 2 1 [α (ν)]γ1/β1. (3) "β1 β12 #! s γ is the one-loop ∆S = 2 anomalous dimension 1 3 1 γ = 1− = 1. (4) 1 2 N (cid:18) c(cid:19) γ is the two-loop ∆S = 2 anomalous dimension [24] 2 1 1 4 57 N2 17 γNDR = − 1− 17+ (3−n )+ c −1 = − , 2 32 (cid:18) Nc(cid:19)" 3 f Nc 9 !# 48 1 1 γHV = γNDR − 1− β . (5) 2 2 2 N 1 (cid:18) c(cid:19) 2 β and β are the first two coefficients of the QCD beta function 1 2 da(ν) ν = β a(ν)k+1, k dν k=1 X 1 9 β = − [11N −2n ] = − , 1 c f 6 2 1 n β = − 34N2 −13n N +3 f = −8. (6) 2 24 c f c N (cid:20) c(cid:21) The explicit numbers are for N = n = 3. c f 3 X-boson Method and Known Constraints The X-boson method was explained in detail in [15]. It takes the idea of [11] of reducing the four-quark operator in (2) to products of currents and follows through the full scheme and scale dependence. In [15] it was explicitly showed how short-distance scale and scheme dependences can be taken into account analytically in the 1/N expansion. Here, we only c sketch the procedure introducing the notation. The effective action Γ , ∆S=2 Γ ≡ −C C(ν) d4yQ (y)+h.c., (7) ∆S=2 ∆S=2 ∆S=2 Z contains all the short-distance physics of the SM. We replace it by the exchange of a colorless heavy ∆S = 2 X-boson with couplings Γ ≡ 2g (µ ,···) d4yXµ(y)L (y) +h.c.. (8) LD ∆S=2 C µ Z The coupling g (µ ,···) is obtained [15] with an analytical short-distance matching us- ∆S=2 C ing perturbative QCD.1 Afterwards we only need to identify the current L in the hadronic µ picture, not the four-quark operator Q . ∆S=2 The matching leads to g2 (µ ,···) M ∆S=2 C ≡ C C(ν) 1+a γ log X +∆r . (9) M2 ∆S=2 1 ν X (cid:20) (cid:18) (cid:18) (cid:19) (cid:19)(cid:21) The one-loop finite term ∆r is scheme dependent 11 1 11 7 1 ∆rNDR = − 1− = − ; ∆rHV = − 1− (10) 8 N 12 8 N (cid:18) c(cid:19) (cid:18) c(cid:19) and makes the coupling |g | scheme independent to order a2. This coupling is also ∆S=2 independent of the scale ν to the same order. Notice that there is no dependence on the 1At energies small compared to the W-boson mass but where perturbative QCD is still valid. 3 cut-off scale µ –this feature is general of four-point functions which are product of con- C served currents [15]. This procedure thusincludes thestandard leading andnext-to-leading resummation to all orders of the large logs in [α log(M /ν)]n and α [α log(M /ν)]n in- S W S S W cluding the short-distance scheme dependence. Inorder toget atthematrix-element (1), we calculatea two-point Greenfunctioninthe presence of the weak effective action, with pseudo-scalar densities carrying kaon quantum numbers. Afterreducing thekaontwo-quarkdensities, thetwo-pointfunction(11)provides the matrix element in (1) via standard LSZ reduction. The two-point function is evaluated using the matching with the X-boson effective action. We thus want to calculate the two-point function [14, 15] Π∆S=2(q2) = i d4eiq·xh0|T P†0(0)PK0(x)eiΓLD |0i (11) K Z (cid:16) (cid:17) which we need to order g2 . After reducing, we obtain ∆S=2 16 hK0(q)|eiΓ∆S=2|K0(q)i = hK0(q)|eiΓLD|K0(q)i ≡ −iC Bˆ q2F2 ∆S=2 3 K K d4p g2 ig = X ∆S=2 µν Πµν(p2 ,q2) (12) (2π)4 2 p2 −M2 X Z X X where q2 is the external momentum carried by the kaons. The basic object is the reduced four-point function Πµν(p2 ,q2) ≡ i24hK0(q)| d4x d4ye−ipX·(x−y)T (Lµ(x)Lν(y))|K0(q)i. (13) X Z Z This can be obtained from the four-point Green’s function with two kaon pseudo-scalar densities and two currents L . µ At large N , the reduced four-point function (13) factorizes into two disconnected two- c pointfunctionsatallordersinquarkmassesandexternalmomentumq2. Thisdisconnected part is g Πµν (p2 ,q2) = (2π)4δ(4)(p )8q2F2 , (14) µν disconn. X X K which leads to the well-known large-N prediction2 c 3 BˆNc = . (15) K 4 At next-to-leading order in the 1/N expansion, one has c 3 g2 1 ∞ Bˆ = ∆S=2 1− dQ2F[Q2] (16) K 4MX2 C∆S=2 " 16π2FK2 Z0 # with Q2 the X-boson momentum in Euclidean space and 1 Q2 g Πµν (Q2,q2) F[Q2] ≡ − lim dΩ µν conn. . (17) 8π2 q2→m2KZ Q 1+(Q2/MX2) q2 2In the strict large Nc limit we also have C(ν)=1. 4 The next point is the calculation of (17). There are two energy regimes where we know how to calculate F(Q2) within QCD, namely, at very large Q2 and at very small Q2. In the first regime Q2 >> 1GeV2, with q2 kept small and we can use the operator product expansion in QCD. One gets ∞ C(i) (ν,Q2)hK0(q)|Q(i) |K0(q)i g Πµν (Q2,q2) = 2n+2 2n+2 , (18) µν conn. Q2n n=2i=1 XX where Q(i) are local ∆S = 2 operators of dimension 2n+2. In particular, 2n+2 Q = 4 d4xLµ(x)L (x) (19) 6 µ Z and Q C (ν,Q2) = −8π2γ a 1+a (β −γ ) log +F +O(a2) (20) 6 1 1 1 1 ν (cid:20) (cid:20) (cid:18) (cid:19) (cid:21) (cid:21) with3 γ ∆r 1 119 F = 2 +(β −γ ) − = (21) 1 1 1 γ1 " γ1 2# 16 where the term with F was not known before. The finite term F is order N andtherefore 1 1 c this a2 term is of the same order in N as the leading term. In fact, at the same order in c N , there is an infinite series in powers of a. c The above can be used to take the limit M → ∞ explicitly via X 3 µ Bˆ = C(ν) 1+a γ log +∆r × K 1 4 ν (cid:18) (cid:18) (cid:18) (cid:19) (cid:19)(cid:19) 1 µ2 ∞ 1− dQ2F[Q2]MX→∞ + dQ2F[Q2]MX→∞ " 16π2FK2 Z0 Zµ2 D≥8 ! µ2 +O +O(a2) . (22) M2 ! # X Where F[Q2] is obtained inserting in (17) the result in (18) minus the dimension six D≥8 term. Forthe list anda discussion ofthedimension eight operatorssee [18,25]. In[18]thereis a calculation in the factorizable limit of the contribution of the dimension eight operators. Numerically, the finite term of order a2 competes with that contribution when Q2 is around (1 ∼ 2) GeV2. The second energy regime where we can calculate F[Q2] model independently is for Q2 → 0, where the effective quantum field theory of QCD is chiral perturbation theory. In ChPT the result is known up to order p4 both in the chiral limit [15, 17] and outside the chiral limit [22]. The result in the chiral limit is 12 Fχ[Q2] = 3+A Q2 +A Q4 +··· , A = − (2L +5L +L +L ) , (23) 4 6 4 F2 1 2 3 9 0 3The leading terms differs from the one in [17] only in terms subleading in 1/Nc. 5 with F the chiral limit of the pion decay constant F = 92.4 MeV. The next term, A , 0 π 6 can be easily calculated but contains some of the unknown C constants from the p6 ChPT i Lagrangian. We still need to describe the intermediate energy region for which we use the large N c hadronic model described in the next section. We have used that model to predict the series in (23) minus 3+A Q2, which is known. This is equivalent to predict the relevant 4 C and higher couplings combinations. i 4 A Ladder Resummation Large N Model c All the results and information we have presented so far on Bˆ are model independent. In K particular, we have seen in the previous section that there are two energy regimes which can be calculated within QCD. In this section, we describe a large N hadronic model that c provides the full Πµν (Q2,q2) which, apart from other QCD information, contains these conn. two QCD regimes analytically. The large N hadronic model we use was introduced in [19]. It can be thought of as c QCD in the rainbow or ladder-resummation approximation. The basic objects are vertex functions with one, two, three, ... two-quark currents or density sources attached to them, referred to as one-point, two-point, three-point, ... vertex functions. These correspond to the two-particle irreducible diagrams in large N QCD. These vertex functions are glued c , , , ... Figure 1: One-point, two-point, three-point, ··· vertex functions. The crosses can be vector or axial-vector currents and scalar or pseudo-scalar densities. into infinite geometrical series with couplings g for vector or axial-vector sources and g V S for scalar or pseudo-scalar sources, what can be seen as a very crude approximation to the two-particle reducible part. In this way one can construct full n-point Green’s functions in the presence of current quark masses –see for instance, how to get full two-point functions in Figure 2. ... Figure 2: Infinite geometrical series which gives full two-point functions at large N . The c black vertices thatglue the vertex functions together areeither g forvector or axial-vector V sources or g for scalar and pseudo-scalar sources. S 6 The basic vertex functions in Figure 1 have to be polynomials in momenta and quark masses to keep the large N structure. As explained in [19], only the first nonet of hadronic c states per channel, i.e., the pseudo-scalar pseudo-Goldstone bosons, the first vectors, the first axial-vectors and the first scalars can be easily generated within this framework. The coefficients of the vertex functions are free constants of order N , many of which can be c fixed by imposing chiral Ward identities on the full Green’s functions. Chiral perturbation theory at order p4 and the operator product expansion in QCD help to determine many more of these free coefficients in the vertex functions. A consequence of the formalism described here is that the vertex functions obey Ward identities with a constituent quark mass [19]. The full two-point Green’s functions obtained from the resummation in Figure 2 agree withtheonesoflargeN QCDwhen onelimitsthehadroniccontent tobejustonehadronic c state per channel –our model does not produce any less or more constraints than large N c QCD and all parameters can be fixed in terms of resonance masses [15] in agreement with other groups using large N and one state per channel. Introducing two or more hadronic c states per channel systematically is difficult as explained in [19]. We leave for future work the investigation into how to carry it out. Most low-energy hadronic effective actions used for large N phenomenology are in the c approximation of keeping the resonances below some hadronic scale and always in the chiral limit. In many cases less than the four states included here are used. The procedure described above of obtaining full Green’s functions can be done in the presence of current quark masses. In fact, in [15] two-point functions were calculated outside the chiral limit and all the new parameters that appear up to order m2 can be q determined except one; namely, the second derivative of the quark condensate with respect to quark masses. Some predictions of the model we are discussing involving coupling constants and masses of vectors and axial-vectors in the presence of masses are f2 M2 = f2 M2 , Vij Vij Vkl Vkl 1 f2 M4 −f2 M4 = − hqqi (m +m −m −m ) Vij Vij Vkl Vkl 2 χ i j k l f2 M2 +f2 = f2 M2 +f2 , Aij Aij ij Akl Akl kl 1 f2 M4 −f2 M4 = hqqi (m +m −m −m ) , (24) Aij Aij Akl Akl 2 χ i j k l where i,j,k,l are indices for the up, down and strange quark flavors. Three-point functions, as shown in Figure 3, were calculated in the chiral limit in [19]. We have now all the needed ones in the study of Bˆ also outside the chiral limit and they K will be presented elsewhere [26]. Some three-point functions have also been calculated in other large N approaches and in the chiral limit [27, 28, 29, 30] –for instance, PVV, PVA, c PPV and PSP three-point functions, where P stands for pseudo-scalar, V for vector, A for axial-vector and S for scalar sources. They agree fully with the ones we get in our model when restricted to just one hadronic state per channel. Four-point functions are constructed analogously with the two-topologies dictated by large N , one with two three-point vertex functions and one with a four-point vertex c 7 Figure 3: Infinite geometrical series which gives full three-point functions at large N . The c crosses that glue the vertex functions are either g for vector or axial-vector sources or g V S for scalar and pseudo-scalar sources. function. The final four-point functions fulfill the correct chiral Ward identities –including current quark masses and we impose short-distance QCD constraints as well as the ChPT results at NLO to determine the free parameters allowed by the chiral Ward identities. 5 Chiral Limit Results We now calculate F[Q2] of Eq. (17) in the chiral limit using the model of the previous section. After integrating over the four-dimensional Euclidean solid angle Ω and doing Q the limit q2 → 0 as in (17), we get α α α β Fχ[Q2] = δ Q2 +δ + V + A + S + V 1 2 Q2 +M2 Q2 +M2 Q2 +M2 (Q2 +M2)2 V A S V β γ γ A V A + + + . (25) (Q2 +M2)2 (Q2 +M2)3 (Q2 +M2)3 A V A We have used here the two-point vertex functions to second order in momenta, higher orders will make it impossible to satisfy the Weinberg sum rules. The three- and four- point vertex functions have been expanded to fourth order in the external momenta. The short-distance constraints used are the Weinberg sum rules and the fact that the vector form factor should vanish as 1/Q2 for large Q2. As expected from the arguments in [19], we find clashes between short-distance constraints for three-point functions and those of the needed four-point function here. I.e., with just a finite number of hadronic states per channel not all short-distance constraints for n-point Green are compatible in general. Of course, one can always impose the latter which are compatible with the subset of the short- distance constraints corresponding to the momenta structure of the three-point functions 8 p6 input p4 input F 87.7 MeV 81.1 MeV 0 103L 0.43 0.38 1 103L 0.73 1.59 2 103L −2.3 −2.91 3 103L 0.97 1.46 5 103L 5.93 6.9 9 103L −4.4 −5.5 10 M 805 MeV 690 MeV V M 1.16 GeV 895 MeV A M 1.41 GeV 1.06 GeV S A −12.7 GeV−2 −23 GeV−2 4 Table 1: The low energy inputs used together with some derived quantities for the two sets of input. entering in the four-point functions but not with all three-point functions short-distance constraints. That is what we have done for calculating Bˆ . K The short-distance constraints on F[Q2] discussed in Sect. 3 and Ref. [17, 18] require δ = δ = 0 in (25). Imposing these we obtain the expression for F[Q2] given in the 1 2 appendix. The explicit calculation reveals that diagrams with the exchange of vector and axial- vector states produce not only single poles but also double and triple poles. These also arise from diagrams with only one of these propagators due to the factor 1/q2 present in (17).4 The function Fχ[Q2] in (17) reproduces the large N -pole structure found in [17, 18] c but including the first hadronic state in all the spin zero and one channels. The relevant free parameters are the masses M , M and M , the pion decay constant V A S in the chiral limit, F , and three combinations of constants appearing in the three- and 0 four-point vertex functions, Aχ . 1,2,3 The low energy inputs we use are F and the O(p4) couplings L determined from the 0 i ChPT fits to data. We use two different fits, namely, fit 10 of [31] for L and L fit 1,2,3 9,10 from [32], both with the full p6 fits and the ones using only p4 expressions. The input values for these two cases p6 and p4 are given together with some derived quantities in Table 1. The values of the masses and the slope A follow from the L and F used as 4 i 0 input using the relations of [19] and (23). One could also use the physical masses and then use the L only to get at the slope. The case with p6 input is within 30% of the physical i masses. For the scalars, which mass to use is still an open question, but with the mounting evidence that the first one are not present in large N , [33] and references therein, a mass c of around 1.4 GeV seems fine. 4Ref. [17, 18] use a different underlying four-point function. The same pole structure does show up there as well as is required by chiral symmetry. 9