Analysis of the triplet production by the circularly polarized photon at high energies * G. I. Gakh, M. I. Konchatnij, I. S. Levandovsky, and N. P. Merenkov Kharkov Institute of Physics and Technology 61108, Akademicheskaya, 1, Kharkov, Ukraine 3 1 The possibility in principle of the determining high energy photon circular polar- 0 2 ization by the measurement of the created electron polarization in the process of n triplet photoproduction γ +e− e+e− +e− is investigated. The respective event a → J number which depend on polarization states of photon and created electron does not 5 2 decrease with the growth of the photon energy, and this circumstance can ensure the ] h high efficiency in such kind of experiments. We study different double and single p - p distributions of the created electron (or positron), which allow to probe the photon e h circular polarization and to measure its magnitude (the Stock’s parameter ξ ), using 2 [ the technique of the Sudakov’s variables. Some experimental setups with different 1 v 0 rules for event selection are studied and corresponding numerical estimations are 1 0 presented. 6 . 1 0 3 1 1. INTRODUCTION : v i X It is well known that process of the triplet production r a γ(k)+e−(p) e−(k )+e+(k )+e−(p ) (1) 1 2 1 → by the high-energy photons on the atomic electrons can be used to measure the photon linear polarization degree [1–3]. This possibility arises due to azimuthal asymmetry of the corresponding cross-section, i.e., due to its dependence on the angle between the plane in which the photon is polarized, and the plane (k,p ) where the recoil electron 3-momentum 1 lies. The detailed description of the different differential distributions, such as the depen- dence on the momentum value, on the polar angle and minimal recorded momentum of the * Electronic address: [email protected] 2 recoil electron, dependence on the invariant mass of the created electron-positron pair, on the positron energy and others, has been investigated in Ref. [4]. This single-spin effect is the basis for theoretical background of polarimeters where the different angular and energy distributions are used [5]. The exact expressions for differential and partly integrated cross sections of the process (1) is very cumbersome and exist in the complete form only for unpolarized case [6]. At high collision energy only two (from eight) diagrams contribute with leading accuracy (neglecting terms of the order of m2/s, s = 2(kp), m is the electron mass) and the corresponding expressions are essentially simplified. These diagrams (the so-called Borselino diagrams [7]) are shown in Fig.1. Nevertheless, at the boundaries of the final particle phase space the non- leading terms can be reinforced, and in Ref. [8] some of such effects had been investigated for the case of linearly polarized photons. As regards the photon circular polarization, it can be probed by at least double-spin effects. In the region of small and intermediate photon energies the circular polarization can be measured using double-spin correlation in the Compton scattering. For example, in Ref. [9] the corresponding possibility was considered for the Compton cross-section asymmetry in the scattering of photon on polarized electrons. In principle, one can also measure the polarization of the recoil electron. The double-spin effects may be used to create polarized electron beams using the laser photons [10]. At high energies of the photon beams the use of the Compton scattering is not effective because the Compton cross-section decreases very fast with the growth of the photon energy. If the photon energy is large, the cross-section of the electron-positron pair production, which does not decrease with the growth of the energy, has become larger than the Compton scattering one. To estimate the respective energy one can use the asymptotic formulas for the total cross-sections [11] 2πr2 28αr2 σ 0 lnx, σ 0 lnx, (2) C pair ≈ x ≈ 9 s 1 x = , α = , m2 137 where r = α/m is the classical radius of electron. In the rest system of the initial electron 0 (s = 2ωm)thephotonenergyω hastobelargerthanforabout80MeV.Thus, tomeasurethe circularpolarizationofthephotonswiththeenergiesmorethan100MeV itisadvantageously to use the process (1) rather than the Compton scattering. 3 k1 −k2 k k −k2 + k1 p p1 p p1 Fig.1. Borselino diagrams which give the nondecreasing contribution in the cross section at high energies and small momentum transferred. The above estimate of the pair production cross-section is made taking into account only the Borselino diagrams. The events, described by these diagrams, have very specific kinematics in the rest system of the initial electron, namely: the recoil electron has small 3- momentum (of the order of m) whereas the created electron-positron pair carries out all the photon energy and moves along the photon momentum direction. In the reaction c.m.s the scattered(recoil)electronhassmall(oftheorderofm)perpendicularmomentumtransferand very small (of the order of m3/s) longitudinal one. Just such kind of the events contribute to the nondecreasing cross-section. The contribution of the rest diagrams, describing the direct capture of the photon by the initial electron and exchange effects due to the identity of the final electrons, decreases at least as m/ω. Fig.2. Spectral energy distribution of the Crab Nebula renormalized to the total phase interval. The synchrotron radiation (region up to 103MeV) and due to inverse Compton scattering (region upper 103MeV) are presented. To clarify the experimental points, statistical and systematic errors, solid and dotted curves see Refs. [12, 13]. The very high-energy photon component can be contained in the cosmic rays. For exam- 4 ple, in Fig.2 we show the spectral energy distribution of the Crab Nebula [12]. In this figure the high-energy photons in the region of 102 103MeV are synchrotron ones and in the − region 103 105MeV they arise due to possibility of the inverse Compton scattering on an − ultra-energetic electrons. The analysis of their polarization is very important to understand the remarkable features of the cosmologically distant gamma ray bursts. Thereareafewpossibilities tomeasurethephotoncircular polarizationintheprocess (1). It is possible: i)to use longitudinally polarized electrons and measure the asymmetry of the cross-section at two opposite directions of the polarization, ii)to measure the polarization of the recoil electrons, iii)to measure the polarizationof the created electrons or positrons. The double-spincorrelationeffectsinfirsttwocasesdecreasewiththegrowthofthephotonenergy and, therefore, are not effective for the measurement of the photon circular polarization at thehighenergies. So, inthispaperwe concentrateonthethirdexperimental setup whichcan be realized in the scattering of the photons on unpolarized atomic electrons or the electron beam. Our study in some aspects is close to the approach developed in Ref. [14], where the process (1) with circularly polarized photons has been suggested to create high energy polarized electrons, and in the last section we discuss the corresponding similarities and differences in more details. 2. FOUR-RANK COMPTON TENSOR In the approximation considered here, the matrix element squared of the process (1) is defined by a contraction of two second-rank Lorentz tensors V and B , and the differential µν µν cross-section of this process can be written in the form e6 dσ = V B dΦ, (3) µν µν 2(2π)5sq4 d3k d3k d3p 1 2 1 dΦ = δ(k +p p k k ) , 1 1 2 2E 2E 2ε − − − 1 2 1 where q = k k k = p p, E (E ) is the energy of the created electron (positron) and 1 2 1 1 2 − − − ε is the energy of the recoil electron with 4-momentum p . The tensor B is defined by the 1 1 µν electron current j µ ∗ B = j j , j = u¯(p )γ u(p), (4) µν µ ν µ 1 µ and in the case of polarized initial electron 1 B = Tr(pˆ +m)γ (pˆ+m)(1 γ Wˆ )γ , µν 1 µ 5 ν 2 − 5 where W is its polarization 4-vector. Taking the trace over the spinor indices we have µ B = q2g +2(pp ) 2im(µνqW), (5) µν µν 1 µν − where the following notation is used (ab) = a b +a b , (µνqW) = ǫ q W , ǫ = 1. µν µ ν ν µ µνλρ λ ρ 1230 If the initial electron is unpolarized and one want to measure the recoil electron polarization, it needs to substitute p ⇆ p , µ ⇆ ν, W W 1 1 → in the right hand side of Eq.(5) which results simply to change W W , where W is the 1 1 → polarization 4-vector of the recoil electron. Forevents witharbitrarilypolarizedphotonbeam, thetensorV inEq.(3)canbewritten µν in terms of its Stock’s parameters ξ (i = 1,2,3) and the four-rank Compton tensor T i µνλρ (such that its contractions with q ,q and k ,k equal to zero and which will be defined µ ν λ ρ below) as follows 1 V = [e e +e e ]+ξ [e e e e ]+ (6) µν 1λ 1ρ 2λ 2ρ 3 1λ 1ρ 2λ 2ρ 2 − ξ [e e(cid:0) +e e ] iξ [e e e e ] T , 1 1λ 2ρ 2λ 1ρ 2 1λ 2ρ 2λ 1ρ µνλρ − − (cid:1) where the mutually orthogonal space-like 4-vectors e and e , relative to which the photon 1 2 polarization properties are defined, have to satisfy the following relations e2 = e2 = 1, (e k) = (e k) = (e e ) = 0. 1 2 − 1 2 1 2 The first term inside the parentheses in r.h.s. of Eq.(6) is in charge of the events with unpolarized photon, the second and third ones are responsible for the events with linear photon polarization and the last one – for the events with the circular polarization. The parameters ξ and ξ , which define the linear polarization degree of the photon, depend 1 3 on the choice of the 4-vectors e and e , whereas parameter ξ does not depend. Because 1 2 2 we want to investigate the events with circular photon polarization, we can choose these 4-vectors by the most convenient way, namely χ k χ k (λkk k ) 1 2λ 2 1λ 1 2 e = − , e = (7) 1λ 2λ N N 6 with the short notation N = 2χχ χ m2(χ2 +χ2), χ = (kk ), χ = (k k ). 1 2 − 1 2 1,2 1,2 1 2 The 4-vector e appears kindly in the expression for the four-rank Compton tensor T , 1 µνλρ see Eq.(8) below. The polarization properties of a real photon are defined by two orthogonal 3-vectors n1 andn2.Each ofthese two vectors areorthogonalalso to 3-vector ofthephotonmomentum k, and the 4-vectors e and e are their covariant generalizations. It follows from the definition 1 2 of e and e that they have both time and space components. Adding to them 4-vector k 1 2 with appropriate factors (it is, in fact, the gauge transformation), one can eliminate the time and longitudinal (along the vector k) components. Thus, in arbitrary Lorentz system with Z-axis directed along the vector k and 3-momentum lying in the (ZX) plane, where k = (ω,0,0,ω), k = (E ,k ,0,k ), k = (E ,k ,k ,k ), 1 1 1x 1z 2 2 2x 2y 2z the corresponding transformation has the form E k E k k k (0,n1) = e1λ 1 2z − 2 1z kλ, (0,n2) = e1λ 1x 2y kλ, − N − N where n1 = (nx,ny,0), n2 = (ny, nx,0), − ω[(E k )k (E k )k ] ω(E k )k 1 1z 2x 2 2z 1x 1 1z 2y n = − − − , n = − , x y N N N2 = ω2 [(E k )k (E k )k ]2 +(E k )2k2 . 1 − 1z 2x − 2 − 2z 1x 1 − 1z 2y (cid:8) (cid:9) At such transformation no observables are changed due to the gauge invariance, that mani- fests itself by means of the above mention restrictions on the tensor T µνλρ k T = k T = 0. λ µνλρ ρ µνλρ That is why the description of all polarization phenomena in process (1) by means of the 4-vectors e1 and e2 is completely equivalent to the description in terms of 3-vectors n1 and n2. The evident advantage of the covariant description is independence from the Lorentz system. 7 For events in which the created electron polarization states in the process (1) must be determined, the Borselino diagrams lead to following expression for the tensor T µνλρ 1 T = Tr (kˆ +m)(1 γ Sˆ)Qˆ (kˆ m)Qˆ , (8) µνλρ 1 5 λµ 2 νρ 2 − − (cid:8) N γ kˆγ γ kˆγ(cid:9) Qˆ = e γ + λ µ µ λ , λµ 1λ µ χ χ 2χ − 2χ 1 2 1 2 where S is the electron spin 4-vector, with properties S2 = 1, (Sk ) = 0. 1 − Let us divide T into two parts: the firs part depends on 4-vector S and the second µνλρ one does not (0) (S) T = T +T . µνλρ µνλρ µνλρ Then we can write (0) T = T +T , (9) µνλρ (µν)(λρ) [µν][λρ] (S) T = im T +T , µνλρ (µν)[λρ] [µν](λρ) (cid:2) (cid:3) where we use the index notation (αβ) ([αβ]) to emphasize the symmetry (antisymmetry) under permutation of indices α and β. These symmetry properties (9) and form (5) for the tensor B allow to discuss some features of the process (1) with high energy polarized µν photon on the quality level. As we noted in the Introduction, the cross section of the process (1) (when all particles are unpolarized) does not decrease with the growth of the photon energy. Such behavior is caused by the terms proportional to s2 in the contraction T B which enters differential µνλρ µν cross section (3). On the other hand, only symmetrical component 2(pp ) in Eq.(5) can 1 µν ensure appearance of these terms. This simple observation suggest us that the non decreas- ing spin correlations in the differential cross section in considered case are connected only with symmetrical, relative (µ ⇆ ν) permutation, tensors T and T . The first one (µν)(λρ) (µν)[λρ] describes single-spin correlations which depend on Stock’s parameters ξ and ξ caused by 1 3 the photon linear polarization [1]. The second one can contribute on condition that the polarization of the created electron (or positron) is measured, or in other words, it describes double-spin correlation which depends on Stock’s parameter ξ that is the degree of the pho- 2 ton circular polarization. In further we will concentrate just on this double-spin correlation that can be used to measure parameter ξ . 2 The antisymmetric, under (µ ⇆ ν) permutation, tensors T and T have not a [µν][λρ] [µν](λρ) largephysical sense intheconsidered problembecause theycandescribe thespincorrelations 8 in the differential cross section which decrease with the energy growth at least as s−1. For the full description in the such approximation there is not enough to consider only the Borselino diagramsandonehave toaccount foralltherest sixones. Butthesetensors areconnected by thecrosssymmetry withthecorresponding tensors inannihilationchannel which aresuitable for description of the subprocess e++e− γ+γ∗, which is important in different radiative → return measurements [15] and where there are no contribution of any other diagrams. That is why we give here all the corresponding expressions in very compact form 2 N2 T = g χ +χ 2g q2e e (10) (µν)(λρ) µν 1 2 λρ 1λ 1ρ χ χ − χ χ − 1 2 1 2 n h i (cid:0) ˆ (cid:1) 2χ χ (1+P )g g 2(k k ) k k + 1 2 λρ µρ νλ 1 2 λρ µ ν − 1+Pˆ +Pˆ +Pˆ Pˆ g k (χ k +χ k )+N(k k )e + λρ µν λρ µν νρ µ 2 1λ 1 2λ 1µ 2µ 1λ − (k e ) (kk ) (k e ) (kk ) (cid:0) (cid:1) (cid:2)N 1 1 λρ 2 µν 2 1 λρ 1 µν (cid:3) χ − χ − 1 1 h i g (χ +χ )(k k) 2(m2 +χ)k k , λρ 1 2 12 µν µ ν − o (cid:2) (cid:3) where k = k +k and Pˆ is operator of the (α ⇄ β) permutation. 12 1 2 αβ 2 (χ2k χ2k )k [k k ] T = (1 Pˆ ) (χ2 +χ2)g g + 1 2µ − 2 1µ ν 1 2 λρ + (11) [µν][λρ] χ χ − µν 1 2 µλ νρ χ χ 1 2 1 2 n h i N (1 Pˆ Pˆ +Pˆ Pˆ ) g e (χ2k χ2k )+ − µν − λρ µν λρ χ χ νλ 1ρ 2 1µ − 1 2µ 1 2 h χ2 (χ χ )(m2 +χ) χ2 (χ χ )(m2 +χ) 1 − 1 − 2 g k k + 2 − 2 − 1 g k k , µρ ν 1λ µρ ν 2λ χ χ 1 2 io where we use notation [ab] = a b a b . αβ α β β α − The spin-dependent parts of T read µνλρ (µνqk) T = 2(µνqS)h h + (χ χ )(kS)g [µν](λρ) − λ ρ χ2χ 2 − 1 λρ− 1 2 (cid:2) (kS) χ χ (Sh) h (µνρq)+h (µνλq) , (12) 1 2 λρ λ ρ − χ 1 (cid:3) (cid:2) (cid:3) where 4-vector h is defined as k k 2 1 h = , χ − χ 2 1 and (k k ) +(k k ) 1 1 1 1 µν 2 2 µν T = + (k k ) + (µν)[λρ] χ χ − χ2 χ2 1 2 µν 1 2 1 2 h (cid:16) (cid:17) 9 q2(χ χ )2 +χ (χ2 χ2) + 1 − 2 1 1 − 2 (λρkS)+ 2χ2χ2 1 2 i (λρkk ) 1 1 12 + (Sh) +(Sk ) g + µν 2 µν χ χ − χ 2 1 2 h (cid:16) (cid:17) i (k S) (kS) k k 2 2ν 1ν (µλρk) + χ − χ χ − χ 2 1 1 2 n h(cid:16) (cid:17)(cid:16) (cid:17) (χ χ ) 1 − 2 (q2 +2χ +2χ )S +(µ ⇆ ν) . (13) 2χ χ2 1 2 ν 1 2 i o 3. DIFFERENTIAL CROSS SECTION When calculating the non-decreasing (with the energy growth) contribution to the unpo- larized part of the cross section one ought to account for the terms proportional to s2 in the contraction T (e e +e e )B , which arise due to scalar products (k p), (k p), and µνλρ 1λ 1ρ 2λ 2ρ µν 1 2 (kp). To do this it is enough to use approximation B = 4p p (see Ref. [3]). Then we have µν µ ν 4m2 T (e e +e e )B = 16 (k p)(k p)+ µνλρ 1λ 1ρ 2λ 2ρ µν 1 2 − χ χ 1 2 h q2 2m2 q2 2m2 (k p)2 +(k p)2 . (14) 1 χ χ − χ2 2 χ χ − χ2 1 2 2 1 2 1 (cid:16) (cid:17) (cid:16) (cid:17)i The main contribution to the differential cross section, within the chosen accuracy, gives the region of small momentum transferred q2 m2. In this case it is useful to introduce | | ∼ the so-called Sudakov’s variables [16] which are suitable for the calculation at high energies and small momentum transferred. These variables, in fact, define the expansion of the final state 4-momenta on the longitudinal and transversal components relative to the 4-momenta of the initial particles. For the process (1) we have (see also [14]) ′ ′ k2 = αp +βk +k⊥, q = αqp +βqk +q⊥, m2 p′ = p k, s = 2(kp), p′2 = 0, − s (k⊥p) = (k⊥k) = (q⊥p) = (q⊥k) = 0, s s d4k2 = dαdβd2k⊥, d4q = dαqdβqd2q⊥, (15) 2 2 where the 4-vectors k⊥ and q⊥ are the space-like ones, so k⊥2 = k2, q⊥ = q2, and k,q − − are two-dimensional Euclidean vectors. 10 The phase space of the final particles with the over-all δ function (see Eq.(3)) can be − written as s2 dΦ = 4 dαdβd2k⊥dαqdβqd2q⊥δ(k22 −m2)δ(k12 −m2)δ(p21 −m2). (16) By using the conservation laws we derive k2 = sαβ k2, k2 = s(1 β)(α+α ) (k+q)2, 2 − 1 − − q − m2 +k2 p2 = sβ +m2 q2, sα = , sβ = q2, 1 q − β q m2 +k2 m2 +(k+q)2 sα = , q − β − 1 β − and after integration over α,α ,β by help of three δ-functions the phase space reduces to q q very simple expression 1 dΦ = dβd2kd2q, (17) 4sβ(1 β) − The variable β is the photon energy fraction that is carried out by the positron β = E /ω 2 (the created electron energy E = (1 β)ω). In terms of the Sudakov’s variables, the 1 − independent invariants are expressed as follows m2 +(k+q)2 m2 +k2 m2(m2 +k2)2 χ = , χ = , q2 = q2 . (18) 1 2(1 β) 2 2β − − s2β2(1 β)2 − − Further we will consider two possible experimental situations: i)when both the scattered (recoil) and created electron are recorded, ii)only created electron is recorded. In the first case we suggest that events with q2 < q2 are not detected, where the minimal selected | | | 0| momentum transfer squared q2 is of the order of m2. In the second case all events with | 0| q2 q2 are included, where q2 is the minimal possible value of q2 , which is defined | | ≥ | min| | min| | | bythesecondtermintheexpressionfor q2 inEq.(18). Itisjustthelongitudinalmomentum − transfer squared. These two event selections give very different values for the differential cross section. If q2 is of the order of m2 one can neglect everywhere with q2 . Such procedure leads to the | | min cross section that depends on q , but does not depend on the collision energy (invariant s). 0 On the other hand, when values of q2 for selected events begin from q2 , the integration | | min over d2q leads to logarithmic rise of the corresponding cross section when the collision energy increases. This leading logarithmic contribution can be derived by the equivalent photon method [17] but our goal is to calculate also the next-to-leading (constant) one.