Radiative decay of the ∆∗(1700) M. D¨oring∗ Departamento de F´ısica Te´orica and IFIC, Universidad de Valencia–CSIC, 46100 Burjassot (Valencia), Spain Electromagnetic properties provide information about the structure of strongly interacting sys- tems and allow for independent tests of hadronic models. The radiative decay of the ∆∗(1700) is studied,which appears dynamically generated in a coupled channelapproach from therescattering ofthe(3/2+)decupletofbaryonswiththe(0−)octetofpseudoscalarmesons. Theradiativedecayis predictedfromthewell-knowncouplingsofthephotontothemesonsandhadronswhichconstitute thisresonance in the dynamical picture. PACSnumbers: 24.10.Eq,25.20.Lj,11.30.Rd 7 0 0 2 I. INTRODUCTION n a J The unitary extensions of chiral perturbation theory (UχPT) have brought new light to the meson-baryon in- 4 teraction, showing that some well-known resonances qualify as being dynamically generated. In this picture the 2 Bethe-Salpeter resummation of elementary interactions, derived from chiral Lagrangians, guarantees unitarity and leads at the same time to genuine non-perturbative phenomena suchas poles of the scattering amplitude in the com- 1 plex plane ofthe invariantscatteringenergy√s, whichcan be identified with resonances. Coupled channel dynamics v 0 plays an essential role in this scheme, with the chiral Lagrangians providing the corresponding transitions of the 7 multiplets; even physically closed channels contribute as intermediate virtual states. 0 After earlierstudies inthis directionexplainingthe Λ(1405)andthe N∗(1535)as meson-baryon(MB) quasibound 1 states [1–6] from the interaction of the meson octet of the pion (M) with the baryon octet of the nucleon (B), new 0 efforts have been undertaken [7, 8] to investigate the low lying 3/2− baryonic resonances which decay in s-wave into 7 0− mesons (M) and 3/2+ baryons (B∗) of the decuplet. The leading interaction of these hadrons, given by the 0 isovector term from Ref. [9], is unitarized by the use of the Bethe-Salpeter equation (BSE) in the on-shell reduction / h schemefrom[8]whichallowsforafactorizationofverticesandtheintermediateloopfunction,thus reducingthe BSE t - to analgebraicmatrixequationincoupledchannels. The unitarizedamplitude developspoles indifferentisospinand cl strangenesschannels in the complex plane of √s, which have been identified with resonancesfromthe PDB [10] such u as the Λ∗(1520), Σ∗(1670), ∆∗(1700), etc. [8]. n In isospin 3/2,strangeness 0, which are the quantum numbers of the ∆∗(1700),the coupled channels in s-waveare v: givenby∆(1232)π,Σ∗(1385)K,and∆(1232)η. The∆∗(1700),togetherwithaseriesofotherproductionmechanisms, i hasbeenincludedinRef. [11]inthestudyoftheγp π0ηpandγp π0K0Σ+ photoproductionreactions,currently X → → measured at ELSA/Bonn. In the detailed work of [11] the ∆∗(1700), together with its strong couplings to ∆(1232)η r and Σ∗(1385)K, turned out to provide the dominant contribution. The branching ratios into these two channels are a predicted from the scheme of dynamical generation and differ from a simple SU(3) extrapolation of the ∆π channel by up to a factor of 30 [12]. The predictions for both reactions are in good agreement with preliminary data [13]. Recently, new measurements at low photon energies have been published [14] which also agree well with [11]. This has motivated another study [12] of altogether nine additional pion- and photon-induced reactions. From considerations of quantum numbers and the experimentally established s-wave dominance of the Σ∗ production close to threshold, the ∆∗(1700) channel is expected to play a major, in some reactions dominant, role. Indeed, good global agreement has been found for the studied reactions that span nearly two orders of magnitude in their respective cross sections. Thus, evidence from quite different experiments has been accumulated that the strong ∆∗(1700) Σ∗K, ∆η cou- → plings,predictedbythecoupledchannelmodel,arerealistic. Thisgivessupporttotheschemeofdynamicalgeneration of this resonance. However, in all the photon-induced reactions from [11, 12] the initial γp ∆∗(1700) transition → has been taken from the experimental [10] helicity amplitudes A and A [15]: In Ref. [16] the electromagnetic 1/2 3/2 form factors G′, G′, and G′, which appear in the scalar and vector part of the γp ∆∗(1700)transition, have been 1 2 3 → Electronicaddress: [email protected] ∗ 2 expressed in terms of the experimentally known A and A [10]; this provides the transition on which we rely in 1/2 3/2 all the photoproduction reactions via ∆∗(1700) in [11, 12]. Such a semi-phenomenologicalansatz is well justified: the photon coupling and the width of the ∆∗(1700)is taken from phenomenology, whereas the strong decays of the ∆∗(1700) into hadronic channels are predictions from the unitary coupled channel model; the strengths of these strong transitions are responsible for the good agreement with experimentfoundin[11,12]. Itis,however,straightforwardtoimproveatthispoint,andthisistheaimofthisstudy. Electromagnetic properties provide additional information about the structure of strongly interacting systems and allow for an independent test of hadronic models, in this case the hypothesis that the ∆∗(1700) is dynamically generated. Avirtue ofthe presentmodelis thatonecanmakepredictionsforthe radiativedecay,orequivalently,the inverse process of photoproduction; the components of the ∆∗(1700) in the meson-baryonbase are all what matters, together with the well-known coupling of the photon to these constituents. A similar study has been carried out for theradiativedecayoftheΛ∗(1520)inRef. [17]thathasbeendescribedrecentlyasadynamicallygeneratedresonance [18]. For the Λ∗(1520) Σ0γ decay, where the dominant channels add up, indeed good agreement with experiment → has been found. The present study is carried out along the lines of [17] but several modifications will be necessary: the (πN) channel in d-wave plays an important role and is implemented in the coupled channel scheme. Second, a fully gauge invariant phototransition amplitude for the s-wave channels is derived that includes also couplings of the photon to the B∗ baryons. ∗ II. THE MODEL FOR THE RADIATIVE ∆ (1700) DECAY In Sec. IIA the coupled channel model from Ref. [8] is revised and extended to the inclusion of the πN channel in d-wave, (πN) . In Sec. IIB the model for the phototransition amplitude ∆∗(1700) γN is derived: The photon d → interacts with the dynamically generated resonance via a one-loop intermediate state that is given by all coupled channels which constitute the resonance. A. The (πN) channel in the unitary coupled channel approach d IntheunitarizedmodelfromRef. [8]the∆∗(1700)resonanceappearsasaquasi-boundstateinthecoupledchannels ∆π, Σ∗K, and ∆η. The attraction in these channels leads to the formation of a pole in the D channel which has 33 been identified with the ∆∗(1700)resonance. In Ref. [8] the chiralinteraction from [9] is adapted in a nonrelativistic reductionprovidingisovectorMB∗ MB∗ transitionswhereM(B∗)standsforthe octetof0− pseudoscalarmesons (3/2+ decuplet baryons). This inter→action is unitarized by the use of the Bethe-Salpeter equation (BSE) T =(1 VG)−1V (1) − which turns out to be a matrix equation in coupled channels, factorized to an algebraic equation according to the on-shell reduction scheme of [8, 19]. In the recent work of [17] this scheme is compared in detail to other possible treatments of the BSE. The function G is a diagonal matrix with the MB∗ loop functions G of the channels i MB ∗ which are regularized in dimensional regularization with one subtraction constant α. As it will appear in a different context in the phototransition, the function G has been re-derived with the result MB ∗ d4p 2M 1 G = i MB∗ (2π)4 (p+q)2 M2 p2 m2 Z − − 1 ddℓ 1 E = 2M lim dx − d→4 (2π)d (ℓ2 +xM2+(x 1)(xq2 m2))2 Z Z0 E − − 2M  m2 M2 m2+s M2 Q(√s)  = α+log + − log + f (√s) (2) (4π)2 µ2 2s m2 √s 1 (cid:20) (cid:21) where m (M) is the meson (decuplet baryon) mass, q2 s is the invariant scattering energy, and ≡ 2 α(µ)=γ log(4π) 2, (3) − ǫ − − withthe Eucledianintegrationoverℓ andǫ=4 d. Thec.m. energy√sinEq. (2)andallthefollowingexpressions E − of this study has to be taken at the physical sheet, i.e., √s √s+iǫ. The only exception is the pole search in the → 3 second Riemann sheet discussedat the end of this subsection. Values of the regularizationscale of µ=700 MeV and α= 2 are natural, as argued in [8]. The c.m. momentum function Q and f in Eq. (2) are given by 1 − (s (M +m)2)(s (M m)2) Q(√s) = − − − , 2√s p f (√s) = log s (M2 m2)+2√sQ(√s) +log s+(M2 m2)+2√sQ(√s) 1 − − − log(cid:0) s+(M2 m2)+2√sQ(√s(cid:1)) lo(cid:0)g s (M2 m2)+2√sQ(√(cid:1)s) . (4) − − − − − − − The loop function from Eq. ((cid:0)2) has a real part, which implie(cid:1)s a ma(cid:0)jor difference of the present app(cid:1)roach compared to unitarizations with the K-matrix. The real parts of the G , together with the attractive kernel V in the isospin i 3/2,strangeness 0 channel, provide enough strength for the formation of a pole in the complex plane of the invariant scattering energy √s which is identified with the ∆∗(1700). However, additional channels will also couple to the dynamically generated resonance, changing in general its position and branching ratios, as these new channels can rescatter as well. In this study, the (πN) channel is d included in the analysis, because this is the lightest channel that can couple to the ∆∗(1700)and precise information of the πN πN transition in the D channel exists from the partial wave analysis (PWA) of Ref. [20]. The (ρN) 33 s → channel has been found important [10, 21], but for the radiative decay its influence is expected to be moderate as discussed below. In order to include the (πN) channel in the coupled channel model, one has to determine the (πN) (MB∗) d d s → transitions, where MB∗ stands for the channels ∆π, Σ∗K, and ∆η from [8]. There is no experimental information on these transitions. From the theoretical side, there is no information either due to the large number of low energy constants in the d-wave to s-wave transition. Thus, the coupling strengths have to be introduced as free parameters, called β. For the inclusion of d-wave potentials, we follow the lines of Ref. [18, 22] where it has been shown that the d-wave transitions can be factorized on-shell in the same way as the s-wave transitions; as a consequence, the meson-baryond-waveloop function is the same as the s-wave loop function from Eq. (2). We also allow for a d-wave to d-wave transition of the (πN) channel. d With the channel ordering i=1 4 for ∆π, Σ∗K, ∆η, (πN) the interaction kernel is given by d ··· √5 2 k0+k′0 2 k0+k′0 0 Q2 rβ −4fπ2 −4fπ2 πN (πN)d→∆π V = (cid:0) (cid:1) −1 (cid:0)k0+k′0(cid:1) √32 k0+k′0 Q2 rβ  (5) −4fπ2 −4fπ2 πN (πN)d→Σ∗K  0 Q2 rβ   (cid:0) (cid:1) (cid:0) (cid:1) πN (πN)d→∆η   Q4 r2β   πN (πN)d→(πN)d    where we have multiplied some elements with r = 1/(4 932 1700) MeV−3 = 1.7 10−8 MeV−3 in order to obtain · · · dimensionless transition strengths β of the order of one. In Eq. (5) one can recognize the chiral isovector transitions with (k0+k′0) from Ref. [8] where k0 = (s M2+m2)/(2√s) is the meson energy and f = 93 MeV. In Eq. (5), π − Q is the on-shell c.m. momentum of the πN system and the β are the s-wave to d-wave transition strengths. See πN also Refs. [18, 22] where the analytic form of the transitions V (i=1, ,4) has been derived. i4 ··· Although a natural value for the subtraction constants is given by α= 2 [8], it is a common procedure [5, 18] to − absorb higher order effects in small variations around this value. Thus, as these higher order effects are undoubtly present,we allowfor variationsof the α ofthe four channels. Togetherwith the transitionstrengths β, the set offree parameters is fitted to the single-energy-bin solution of the PWA of Ref. [20]. Note that there is a conversion factor according to M Q M Q T˜ (√s)= i i j j T (√s) (6) ij ij −s4π√ss4π√s in order to express the solution T of the Bethe-Salpeter equation (1) in terms of the dimensionless amplitude T˜(√s) from [20]. In Eq. (6), M (Q ) is the baryon mass (c.m. momentum) of channel i; in the present case, i=j =4. i i Forenergiesabove1.7GeVatheoreticalerrorof0.08hasbeenaddedtotheerrorbarsfrom[20]becauseadditional channelssuchasρN starttoopenandonecannotexpectgoodagreementmuchbeyondthepositionofthe∆∗(1700). The resulting amplitudes of four different fits are plotted in Fig. 1 with the parameter values displayed in Tab. I. In fit 1, all parameters have been left free. The values for the β are small: For √s M , they lead to values in ∆ ∼ ∗ the kernel (5) aroundone order of magnitude smaller than the chiralinteractions between channels 1 to 3; the size of β correspondstoavaluesometimeseventwoordersofmagnitudesmaller. Theseparametervaluesreflect (πN)d→(πN)d the fact that the ∆∗(1700) couples only weakly to πN and that the πN interaction in D is weak in general. 33 4 0.3 Re D Re D 33 33 0.2 IFmit D133 0.2 IFmit D333 Fit 2 Fit 3’ 0.1 0.1 D33 0 D33 0 -0.1 -0.1 -0.2 1200 1400 1600 1800 2000 1200 1400 1600 1800 2000 s1/2 [MeV] s1/2 [MeV] FIG. 1: Fit results: Fit 1: All parameters free. Fit 2: without πN πN d-wave transition kernel. Fit 3: As fit 2, but all subtraction constants for the s-wave loops chosen to be equal. Fit 3’→: As fit 3 but different minimum in χ2. The error bars show the single-energy-bin solution T˜ from Ref. [20]. 44 TABLEI:Parameter valuesof thefitsto(πN)d: theαi aresubtraction constants, theβi are(πN)d (B∗M)s and (πN)d → → (πN) transition strengths. Notethe sign changes of theβ between fit 3 and fit 3’. d i α α α α β β β β ∆π Σ∗K ∆η (πN)d (πN)d→∆π (πN)d→Σ∗K (πN)d→∆η (πN)d→(πN)d Fit 1 1.96 1.28 0.87 1.00 2.19 1.15 0.07 63.5 − − − − − − Fit 2 1.25 1.29 0.66 1.96 3.30 0.37 0.31 0 − − − − − Fit 3 1.21 1.21 1.21 1.00 3.25 0.85 0.54 0 − − − − − − Fit 3’ 1.22 1.22 1.22 1.00 3.19 0.92 0.42 0 − − − − − Thus, the (πN) (πN) transition strength can be set to zero which is done for the fits 2, 3, and 3’. One can d d → even choose all subtraction constants of the s-wave channels to be equal, which is done in fit 3, and still obtain a sufficiently good result as Fig. 1 shows. However, for fit three, there is another minimum in χ2, almost as good as the best one found. This fit is called fit 3’. As Tab. I and Fig. 1 show, one obtains an almost identical amplitude with a set of β’s with opposite sign as compared to fit 3 (see explanation below). For the differentsolutions 1 to 3’,the coupling strengths ofthe resonanceto the different channelscan be obtained by expanding the amplitude around the resonance position in a Laurent series (see also Sec. IIE). The residues give the coupling strengths which are uniquely determined up to a globalsign which we fix by demanding the realpart of the couplingto∆π tobe positive. InTab. II the resultingcouplingsaredisplayed. The valuesinbracketsquotethe values of the original model from Ref. [8] without the inclusion of (πN) . Compared to these values, all couplings d TABLE II: Position s1p/o2le and couplings of the ∆∗(1700). The values in brackets show the original results from [8] without the inclusion of the (πN) -channel. The PDB [10] quotes a value of s1/2 =(1620 1680) i(160 240) MeV and couplings d pole − − − corresponding to g =1.57 0.3, g =0.94 0.2. Notethe sign change of g between fit 3 and fit 3’. | ∆π| ± | (πN)d| ± (πN)d s1/2 [MeV] g g g g g g g pole ∆π | ∆π| Σ∗K | Σ∗K| ∆η | ∆η| (πN)d (1827 i108) (0.5+i0.8) (1.0) (3.3+i0.7) (3.4) (1.7 i1.4) (2.2) () − − Fit 1 1707 i160 1.09 i0.92 1.4 3.57+i1.91 4.0 1.98 i1.68 2.6 0.84 i0.05 − − − − − − Fit 2 1692 i166 0.62 i1.03 1.2 3.44+i2.28 4.1 1.89 i1.78 2.6 0.77+i0.00 − − − − − Fit 3 1697 i214 0.68 i1.07 1.3 3.01+i1.95 3.6 2.27 i1.89 3.0 0.89+i0.15 − − − − − Fit 3’ 1698 i216 0.68 i1.07 1.3 3.02+i1.95 3.6 2.27 i1.89 3.0 +0.92 i0.12 − − − − − 5 increase slightly in strength and have different phases. The largest change is observed for g which has increased ∆π | | by40%forfit1;thenewvalueiswellinsidetherangequotedbythePDB[10]thatcorrespondsto g =1.57 0.3. ∆π | | ± Also, the d-wave coupling to πN coincides well with g =0.94 0.2 [10]. The main properties of the resonance | (πN)d| ± are conserved; in particular, the absolute values g do not change much. This is relevant with respect to previous | | studies [11, 12] where the model from [8] has been used for the couplings of the ∆∗(1700) to ∆η and Σ∗K. Some explicitcorrectionsto the resultsfrom[11,12]aregiven,basedonthe values fromfit 1whichis the preferred one as discussed below. The cross sections of the pion induced processes from [12] with the Σ∗K final state will change by a factor ( 0.84 i0.05/0.94)2(4.0/3.4)2 =1.1, i.e., they stay practically the same. In the pion induced |− − | reaction with the ∆η final state the factor is again 1.1. Anticipating the result for fit 1 from Tab. IV, that the radiativecouplingis closeto the experimentalone (whichis the one usedin[11,12]), the crosssectionofthe reaction γp K0π+Λ increases by a factor of (4.0/3.4)2 = 1.4 which improves the agreement with data, see Fig. 5 of [12] and→also the recent measurements in Ref. [23]. For γp π0K0Σ+, which is also studied in [11, 12], and where the → photoproductionviathe∆∗(1700)dominates[11],thefactorisalsocloseto1.4. Forthecrosssectionoftheγp π0ηp reaction, the transition via the ∆∗(1700)is only one of the reactions, and one has to re-evaluate the coherent→sum of all processes of the model of [11], with the new values from Tab. II. The numerical results for this reaction [11] have been updated in Fig. 5 of [12]. With the values of fit 1 from Tab. II, the cross section stays practically the same as in Fig. 5 of [12] (10 % decrease). Interestingly, the sign of the coupling to (πN) is reversed in fit 3’ compared to fit 3. This behavior has been d noted before for the parameter values in Tab. I. The reason for the difference between fit 3 and 3’ is that the coupling of the πN channel to the ∆∗(1700) is small: Although (πN) is contained to all orders in the rescattering d scheme, higher orders are smaller. Then, the (by far) dominant order in the fit to (πN) (πN) is g2 and the d → d (πN)d relative sign to the other couplings g , g , and g is difficult to fix. As the influence of the (πN) channel in ∆π Σ K ∆η d ∗ the rescattering scheme is moderate, the main differences to the original model are caused by other parameters: the important parameters, which have fine-tuned the original model from [8], are the subtraction constants α , α , ∆π Σ K ∗ and α . Their variation brings the pole down from s1/2 =1827 i108 MeV [8] to the values quoted in Tab. II. ∆η pole − In Sec. III results for the radiative decay for all four fits are given, in order to obtain an idea of the systematic theoretical uncertainties. The fit 1 is preferred, though, because in the reduction to less free parameters, as it is the case for the fits 2, 3, and 3’, the remaining free parameters have to absorb effects such as direct (πN) (πN) d d → transitions; results might become distorted, despite the fact that the fits appear to be good in Fig. 1. The larger spaceoffreeparametersalsohelpsto fixthe ambiguitiesfoundinfit3and3’: Forfit1and2,noalternativesolutions with a reversed sign for g have been found, and, thus, the sign is fixed. (πN)d I have also performed a search for poles in the second Riemann sheet. For the different fits, the pole positions are given in Tab. II. The positions are in agreement with the values given by the PDB [10]. Note that one of the virtues of the coupled channel analysis is that a separationof backgroundand resonancepart of the amplitude is not necessary; thus, the position of the pole does not suffer from ambiguities from this separation process required by other analyses. Nevertheless, some theoretical uncertainties are present in the model from the omission of other channels such as ρN in s-waveor even ρN, ∆π and KΣ in d-wave,allof which havebeen reportedin the PDB [10]. Although the ρN channel is closed at the position of the ∆∗(1700), it contributes through the real part of the ρN loop function in the rescattering scheme, and through the finite width of the ρ even to the imaginary part. However,inthecalculationoftheradiativedecay,whichistheaimofthisstudy,nolargecontributionsareexpected from these heavy channels. In the study of the radiative decay of the Λ∗(1520) [17] we have seen that contributions to the radiative decay width from the heavy channels are systematically suppressed, as also discussed in Sec. III. B. The phototransition amplitude The only known radiative decay of the ∆∗(1700), which is also the one of relevance in Refs. [11, 12], is into γN and we concentrate on this channel. The coupled channel model for the ∆∗(1700) has the virtue that the radiative decay can be calculated in a parameter-free and well-known way through the coupling of the photon to the particles which constitute the resonance. The dominant photon couplings to the coupled channels ∆π, Σ∗K, ∆η, and (πN) d are displayed in Fig. 2 . We have chosen here the I = +1/2 or charge C = +1 state of the ∆∗(1700). The set of 3 diagrams (1) to (9) is gauge invariant and finite as shown in Sec. IIC. There is also a loop diagramwhere the photon couples to the ∆+ in an intermediate ∆+η state. This contribution isdoublysuppressed: First,thediagramswithγB∗B∗ couplingsaremuchsmallerthantheotherones(seeSec. IIC), and, second, the ∆η state is heavy which renders this contribution even smaller. Apart from the photon coupling to the ∆π and Σ∗K channels, Fig. 2 shows also the coupling to the πN d-wave channelin diagram(10). Note thatthere is no Kroll-Rudermanterm (whichhas πN in s-wavein the γBMB-vertex) 6 γ (k) π+ (p) π+ π0(q-p) γ q( (p-k) (1) π (2) π (3) π (p-k) ∆+(p) proton ∆0 (q-p) ∆ p ∆0 ∆ proton ∆ q ( γ(k) γ π− π− π− γ (4) π (5) π (6) π p ∆++ ∆ p ∆++ ∆ p ∆++ ∆ γ γ K+ K+ K0 γ (7) π (8) π (9) π p Σ∗0 ∆ p Σ∗0 ∆ p Σ∗+ ∆ γ γ π+ (10) π p n ∆ FIG.2: Mechanismsforthe∆∗(1700)decayins-andd-waveloops. Theshadedcirclesrepresentthe∆∗(1700)inthetransition B∗M ∆π or π+n ∆π. The diagrams in the left column are referred to as ”meson pole loops”. Diagrams (2), (5), and → → (8) have a γBMB∗ transition and are referred to as ”Kroll-Ruderman loops” (diagrams (3), (6), (9): ”baryon pole loops”). Diagram (10) shows theprocess with a d-wave coupling of the(πN)d intermediate state to the∆∗(1700). as the vertex on the right hand side of the πN loop is in d-wave, and the term vanishes in the integration over the loop momentum. The photon coupling to the d-wave loop is evaluated in Sec. IID. As foradditionalphotoncouplings,ase.g. tothe externalproton,to verticesofthe rescattering,orto intermediate meson-baryon loops, apart from the ones considered, these processes are present [24] in general but negligible as discussed at the end of Sec. IIC. The MBB∗ vertices appearingin Fig. 2 are providedby the Lagrangianfrom Ref. [25], with the relevantparts for the present reaction given by = T AµB+BA Tµ µ µ L C (cid:0) 1,···,3 (cid:1) 1,···,3 = ǫ Tade u Ab,µ Bc+ ǫabc Be Ad T uµ C abc µ d e c b,µ ade  a,b,c,d,e a,b,c,d,e X X   (7) where B is the standard 3 3 matrix of the 1/2+ baryon fields, T is the field of the 3/2+ decuplet baryons with the definitionsandphaseconve×ntionsof[25],andAµ isproportionaltotheaxialcurrent[25]. Inthepresentsituation,Aµ is expandedupto one mesonfield. Fromminimalcoupling ofthe photon,the γBMB∗ Kroll-Rudermancontactterm is straightforward deduced from Eq. (7). The constant is fitted to phenomenologically known branching ratios. C Explicit expressions for the required Feynman rules can be found, e.g., in [11, 12, 17]. 7 Let us start with the evaluation of diagram (1) in Fig. 2. The amplitude is given by f∗ 1 ( it) = ∆πN eS†ǫ T − π+∆0 − m 3 µ ν ∆π→∆π r d4p 2M 1 1 ( i) (p k)µ(2p k)ν (8) × − (2π)4 (q p)2 M2+iǫ p2 m2+iǫ (p k)2 m2+iǫ − − Z − − − − − where m(M) is the pion (∆) mass, f∗ m /(√2f ) = 2.13 is the ∆πN coupling strength, e2 = 4π/137 is the electric charge, S† is the spin∆1/π2N ≡ 3/π2 transiπtioCn∆o→pNeπrator which we approximate by S† = (0,S†), and ǫ is the polarization of thµe photon which i→n Coulomb gauge is given by ǫ = (0,ǫ). The shadedµ circle in diagramν ν (1) represents the T-matrix element T of the unitary coupled channel scheme in which the ∆∗(1700) appears ∆π→∆π dynamically generated. In Sec. IIE it will be matched to the results from Sec. IIA. Note the simplified structure of the ∆-propagatorin Eq. (8). This is the same simplification as made in the model of dynamical generation[8] of the ∆∗(1700), see Eq. (2). In order to ensure gauge invariance, the loop in Eq. (8) can be evaluated using a calculation technique from Refs. [26, 27]. In Sec. IIC we will compare this scheme to a straightforwardcalculation of the loops of Fig. 2. The general structure of the loop function, or phototransition amplitude, is given by Tµν =agµν +bqµqν +cqµkν +dkµqν +ekµkν. (9) with the momenta q and k as defined in diagram(1). The terms with c and e do not contribute once contractedwith ǫν accordingto Eq. (8) and using the transversalityof the photon, ǫk =0. The terms with b and d do not contribute as ǫq =0 in the c.m. frame where q =0 and using the fact that ǫ0 =0. Thus, the only term that will not vanish in | | Eq. (9) is (agµν). It can be shown that the sets of diagrams (1) to (6) and (7) to (9) of Fig. 2 are gauge invariant. Contracting Tµν with the photon momentum kν and using the Ward identity k Tµν 0, leads to the condition a+dkq = 0. Note ν ≡ that diagram (2) of Fig. 2 contributes only to the term with a, whereas diagram (1) contributes both to a and d. However,evaluating d (and from this, a through the condition a+dkq=0) has the advantage that the loop integral isfinitewhereasbothdiagram(1)and(2)arelogarithmicallydivergent. UsingFeynmanparametersandkeepingonly the terms proportional to kµqν , the second line of Eq. (8) becomes 1 1−x 4Mkµqν x(z 1) dkµqν = dx dz − , (10) − (4π2) x[(x 1)q2+z(q2 M2)+M2]+(1 x)m2 Z Z − − e − 0 0 where we have written the product 2qk =2q0k0 =q2 M2 in the c.m. system where q = 0 and M is the mass of the external baryon, in this case a proton. Note that −k0 =e k = 1/(2√s)(s M2) whe|re| √s q0 wheich we will use | | − e ≡ several times in the following. From Eq. (10) we calculate the term a through the condition a+dkq =0, GI a= dkq g.i. ≡ − 2M M2 M2+m2 k√s (M2 m2)2 2M2M2 m2M4 M2 = e − + − − e − e log (4π)2 2M2 2M4s m2 (cid:20) e (cid:2) e (cid:3) M2 M2+m2 2M2 s M2 M2 m2+2k√s + e − e − Q(M )f (M )+ − e − Q(√s)f (√s) 4M3k√s e 1 e 4ks 1 (cid:0) e (cid:1)(cid:0) (cid:1) m2 M2+M2+m2 2M Q(M ) M2+M2+m2+2M Q(M ) + Li − e − e e +Li − e e e 2k√s 2 2m2 2 2m2 (cid:20) (cid:18) (cid:19) (cid:18) (cid:19) M2+s+m2 2√sQ(√s) M2+s+m2+2√sQ(√s) Li − − Li − (11) − 2 2m2 − 2 2m2 (cid:18) (cid:19) (cid:18) (cid:19)(cid:21)(cid:21) which is gauge invariant by construction. In Eq. (11), Li is the dilogarithm and Q and f are given in Eq. (4). 2 1 Having calculated a from d, the loop function GI corresponds to the meson pole diagram (1) from Fig. 2 plus the g.i. Kroll-Ruderman term from diagram (2). Furthermore,thephotoncanalsocoupledirectlytothebaryonasdisplayedindiagram(3). Onlythenon-magnetic partoftheγB∗B∗ coupling(convectionterm),givenin[16],is considered. This termhasthesamestructureandsign as the γMM vertex. For the magnetic part, see the discussion at the end of Sec. IIC. Diagram (3) also contributes 8 to the term dkµqν in Eq. (9). The contribution to d, let it be dII, leads to an extra modification of the term a, GII aII = dIIkq g.i. ≡ − 2M M2+M2 m2 k√s(M2 m2)2 sM2M2 M2 = e − − − e log (4π)2 2M2 − 2M4s m2 (cid:20) e e m2 M2 2M2 s sM2 M2 m2+s − e − − e Q(M )f (M ) − Q(√s)f (√s) − 4M3k√s e 1 e − 4ks 1 (cid:0) (cid:1)(cid:0) e (cid:1) M2 m2+M2+M2 2M Q(M ) m2+M2+M2+2M Q(M ) + Li − e − e e +Li − e e e 2k√s 2 2M2 2 2M2 (cid:20) (cid:18) (cid:19) (cid:18) (cid:19) m2+s+M2 2√sQ(√s) m2+s+M2+2√sQ(√s) Li − − Li − . (12) − 2 2M2 − 2 2M2 (cid:18) (cid:19) (cid:18) (cid:19)(cid:21)(cid:21) In order to determine the effective coupling of the photon to the ∆∗(1700) we construct isospin amplitudes from the diagrams given in Fig. 2. The isospin states of ∆π and Σ∗K in (I =3/2,I =1/2) are given by 3 2 1 8 ∆π,I =3/2,I =1/2 = ∆++π− + ∆+π0 + ∆0π+ , 3 | i 5| i 15| i 15| i r r r 1 2 Σ∗K,I =3/2,I =1/2 = Σ∗+K0 + Σ∗0K+ (13) 3 | i 3| i 3| i r r with the phase convention π+ = 1,1 . | i −| i With the loop functions from Eqs. (11) and (12) and standard Feynman rules [11, 12, 17] we can calculate the isospin amplitudes for the sum of all ∆π-loops and the Σ∗K-loops according to Eq. (13) with the result √10 f∗ ( it ǫ)(I=3/2,I3=1/2) = eS† ǫ ∆πN GI +GII T , − · γp→∆π→∆π 3 · mπ g.i. g.i. |m=mπ,M=M∆,Me=MN ∆π→∆π (−it·ǫ)(γIp=→3Σ/2∗,KI3→=1∆/π2) = 3√12eS†·ǫCΣ∗f→πN(cid:0)K¯ GIg.i.+G(cid:1)IgI.i. |m=mK,M=MΣ∗,Me=MN TΣ∗K→∆π (14) (cid:0) (cid:1) where we have indicated which masses m, M, M have to be used in the loop functions. In Eq. (14), e 6(D+F) CΣ∗→NK¯ =1.508≃ 5 . (15) ThestrengthCΣ∗→NK¯ fortheΣ∗ decayintothephysicallyclosedchannelNK¯ hasbeendeterminedfromfromaSU(6) quark model [28] in the same way as in Refs. [11, 12]: the SU(6) spin-flavor symmetry connects the πNN coupling strength to the πN∆ strength, and then SU(3) symmetry is used to connect the πN∆ transition with K¯NΣ∗. The use of SU(6) symmetry allows to express CΣ∗→NK¯ in terms of D and F. In the decuplet, the SU(3) symmetry is broken. This can be taken into account phenomenologically by allowing for different in the Lagrangian (7). For the open channels of the Σ∗ decay modes one obtains = 1.64 and Σ →Σπ C C ∗ C(1Σ5∗)→iΛsπcl=os1e.t7o1tfhroemsefivtatliunegst(ocotmhepapraerttioal decay w=idft∗hs in=to2t.h1e3s)e.channels. The constant CΣ∗→NK¯ =1.508 from Eq. C∆→Nπ ∆πN C. Gauge Invariance The construction of the gauge invariant amplitude in the last section can be compared to a straightforwardcalcu- lation of the diagrams (1) to (9) in Fig. 2. In this section we show that both ways give identical results; at the end of this section we discuss further issues related to gauge invariance. In Fig. 2 there are three types of loops: the Kroll-Ruderman structure, the meson pole term, and the baryon pole term. All loops are logarithmically divergent and we calculate in dimensional regularization for the sake of conservationof gauge invariance. The Kroll-Rudermannloop function is identical to the commonmeson-baryonloop function from Eq. (2), G =G . (16) γBMB MB ∗ ∗ 9 TABLE III: Coefficients A for the diagrams (1) to (9) from Fig. 2 with the amplitude given in Eq. (20). The lower row i shows how theinfinities 2/ǫ (ǫ=4 d, see Eq. (3) for ǫ 0) scale for each diagram. Oncemultiplied with thecorresponding Clebsch-Gordan coefficients (CG) a−ccording to Eq. (13),→the sum over theinfinities cancels, Σ (CG) r 2 =0. i i i`ǫ´ (1) (2) (3) (4) (5) (6) (7) (8) (9) A 1G 1G 2G G G 2G 1G 1G 2G i q3 γMM q3 γBMB∗ −q3 γB∗B∗ γMM γBMB∗ γB∗B∗ q3 γMM q3 γBMB∗ q3 γB∗B∗ r 2 1 2 1 2 1 2 √3 2 √3 2 √3 2 1 2 1 2 1 2 i`ǫ´ −2`ǫ´ `ǫ´ √2`ǫ´ − 2 `ǫ´ `ǫ´ − `ǫ´ −2`ǫ´ `ǫ´ −√2`ǫ´ With the momenta assigned as in diagram (1) of Fig. 2, we define the meson pole loop function by d4p 2M 1 1 ( i) (p k)µ(2p k)ν − (2π)4 (q p)2 M2+iǫ p2 m2+iǫ (p k)2 m2+iǫ − − Z − − − − − gµν G γMM → 1 1−x gµνM 2/ǫ γ+log(4π)+logµ2 gµν2M = − dx dz log x[(x 1)q2+2zqk+M2]+(1 x)m2 (4π)2 − (4π)2 − − (cid:0) (cid:1) Z Z 0 0 (cid:0) (cid:1) 2M = gµν R(m,M) (17) (4π)2 with 1 m2 (M2 m2)2+sM2 M2 R(m,M) = 1 α log − e log 2 − − µ2 − 4sM2 m2 (cid:18) (cid:19) e M2 M2+m2 s M2+m2 + e − Q(M )f (M ) − Q √s)f (√s) 4M k√s e 1 e − 4ks ( 1 e m2 M2+M2+m2 2M Q(M ) M2+M2+m2+2M Q(M ) + Li − e − e e +Li − e e e 2k√s 2 2m2 2 2m2 (cid:20) (cid:18) (cid:19) (cid:18) (cid:19) M2+s+m2 2√sQ(√s) M2+s+m2+2√sQ(√s) Li − − Li − (18) − 2 2m2 − 2 2m2 (cid:18) (cid:19) (cid:18) (cid:19)(cid:21) and Q, f from Eq. (4). The arrow in Eq. (17) indicates that we only keep the terms proportional to gµν because 1 all other possible structures from Eq. (9) do not contribute as commented following Eq. (9). Similarly, and with the assignment of momenta as in diagram (3) of Fig. 2, the loop function where the photon couples directly to the baryon, is given by 2M gµν G =gµν R(M,m). (19) γB∗B∗ (4π)2 Note that the convection part of the γB∗B∗ coupling (the non-magnetic part) is of the same structure and sign as the γMM coupling [16]. The next step is to express the amplitudes of the diagrams in Fig. 2 in terms of these three loop functions. Using standard Feynman rules [11, 12, 17], we obtain for the diagrams (1) to (9) ( it ǫ) =A eg S† ǫT (20) − · (i) i B∗MB · B∗M→∆π (wΣh∗eKre).gBT∗hMeBco=effifc∆∗ieπnNts/mAπfoorr tghBe∗MdiBagr=amCsΣ∗(→i)NfoK¯r/i(2=fπ1)tofro9marEeqg.ive(1n5i)ndTeapbe.ndIIinI.g on whether (B∗M) = (∆π) or i The sum over all diagrams results in √10 f∗ ( it ǫ)(I=3/2,I3=1/2) = eS† ǫ ∆πN (G +G +G ) T , − · γp→∆π→∆π 3 · mπ γMM γBMB∗ γB∗B∗ |m=mπ,M=M∆,Me=MN ∆π→∆π (−it·ǫ)(γIp=→3Σ/2∗,KI3→=1∆/π2) = 3√12eS†·ǫCΣ∗f→πNK¯ (GγMM +GγBMB∗ +GγB∗B∗)|m=mK,M=MΣ∗,Me=MN TΣ∗K→∆π. (21) 10 The infinities of the nine diagrams, i.e., the terms with 2/ǫ from the Eqs. (16, 17, 19) scale as shown in the lower row of Tab. III. In order to construct the isospin states according to Eq. (13) each infinity is multiplied with the corresponding Clebsch-Gordan coefficient; as a result, the sums over all infinities cancel for the ∆π loops and also for the Σ∗K loops. Note that the phase convention π+ = 1,1 is crucial at this point. The Ward identity is | i −| i working. In other words, gauge invariance renders the phototransition amplitude finite and leads to a parameter-free expression. Comparing Eqs. (21) and (14), obviously G +G +G =GI +GII (22) γMM γBMB∗ γB∗B∗ g.i. g.i. which can also be seen by comparing the explicit expressions given in Eqs. (11, 12, 16, 17, 19). In other words, the scheme from Eq. (9), which allows for the construction of a gauge invariant amplitude through the condition a+dkq =0, leads to the same result as a straightforwardcalculation of the amplitude, in which the infinities cancel systematically. However,this is only the case if allcontributionsto dare takeninto account,inthe presentcase from the meson pole term plus the baryon pole term (GI and GII ). g.i. g.i. In the rest of this section, further issues of gauge invariance are discussed such as a comparisonto cut-off schemes, the role of magnetic couplings, and gauge invariance in the context of the rescattering scheme. In several recent studies [11, 17] the occurring photon loops have been regularized with a cut-off. As we have now a gauge invariant, parameter-free scheme at hand, we would like to compare both methods numerically. As a first test, the scheme has been implemented in the calculation of the radiative decay width of the Λ∗(1520) from Ref. [17]. This means a gauge invariant evaluation of the s-wave loops from Fig. 2 of [17] formed by πΣ∗ and KΞ∗, plus additional diagrams with γΣ∗Σ∗ and γΞ∗Ξ∗ couplings in analogy to the diagrams in Fig. 2. In practice, the re-calculation only requires the replacement of the terms G + 2 G˜ from Eq. (41) of [17] by GI +GII or i 3 i g.i. g.i. GγMM +GγBMB +GγB B from Eq. (22). The final resu(cid:16)lt from Ref.(cid:17)[17] for the radiative decay(cid:0)Λ∗(1520) (cid:1)γΣ0 ∗ ∗ ∗ → changes from Γ = 60 keV [17] to Γ = 61 keV. Thus, the approximations made in [17] and the violation of gauge invariance are well under control. Note that the additional diagram with a γΣ∗Σ∗ coupling cancels for the second radiative decay studied in [17], Λ∗(1520) γΛ, in the same way as the πΣ∗ meson pole and Kroll-Rudermanterms. → This is a consequence of gauge invariance but can be also seen directly by noting that the γMM interaction and the convectionterm of the γB∗B∗ interaction have the same structure and sign [16]. Thus, the Λ∗(1520) γΛ radiative → width stays as small as already found in [17]. In the cut-off scheme, the meson pole loop is defined as d4q q2 (q k)2/k2 1 M 1 G˜(cut) = i − · | | i , i Z (2π)4 (q−k)2−m2i +iǫ q2−m2i +iǫ Ei(q) P0−q0−Ei(q)+iǫ Λ 1 dq q2 q2(1 x2) 1 1 M i = dx − −Z (2π)2 Z 2ωiωi′ k+ωi+ωi′ k−ωi−ωi′+iǫ Ei(q) 0 −1 1 1 (ω +ω′)2+(ω +ω′) E (q) √s +kω′ . (23) × √s ω E (q)+iǫ √s k ω′ E (q)+iǫ i i i i i − i − i− i − − i− i h (cid:0) (cid:1) i For consistency, the notation is as in [17]; x is the cosinus of the angle between q and k with k the momentum of the realphoton (k k); m is the meson mass, P0 √s, and ω , ω′ are the energies of the mesons at momentum q | |≡ i ≡ i i and q k, respectively; E the energy of the baryon. Note that Eq. (23) is slightly different from the corresponding i − expression in [17], called G˜ there, as one kinematical approximation made in [17], the angle average over the term 1 x2, is not sufficient in the present case, because the momenta of the decay products are higher. One obtains G˜ fro−m [17] by substituting (1 x2) 1 in Eq. (23). Note that G˜(cut) corresponds to 2/3 of G˜. − → The meson pole term G˜(cut) is accompaniedby the correspondingKroll-Rudermanterm with cut-off, called G(cut), where explicit expressions can be found, e.g., in Eq. (22) of [17]. The cut-off can be determined by requiring the real part of the Kroll-Ruderman loop function to be equal in both dimensional regularization and cut-off scheme, at the energy of the resonance. For ∆π in the loop, this leads to a cut-off of Λ=881 MeV. In Fig. 3 the cut-off loops for ∆π (G˜(cut) +G(cut)), are shown as the dotted line. The gauge invariant function GI from Eq. (11) is plotted with the solid line. The imaginary parts of both results are identical as expected, but, g.i. moreinterestingly,atthe energiesofthe ∆∗ ofaround1700MeV,alsothe realpartscoincideclosely. Thedashedline shows the gauge invariant function GII from Eq. (12). This contribution comes from the baryon pole diagrams. As g.i. there are two baryon propagators,the diagramshould be smaller which is indeed the case as Fig. 3 shows. However, results can be affected noticeable and one should include this term in general. From the comparison in Fig. 3 we see that the cut-off scheme as it has been used in [17] (Kroll-Ruderman plus meson pole term) indeed takes into account the dominant contributions. The baryon pole term, which has not been