Radiative pion capture in 2H, 3He and 3H

The $\pi^- + {^2{\rm H}} \rightarrow \gamma + n + n$, $\pi^- + {^3{\rm He}} \rightarrow \gamma + {^3{\rm H}}$, $\pi^- + {^3{\rm He}} \rightarrow \gamma + n + d$, $\pi^- + {^3{\rm He}} \rightarrow \gamma + n + n + p$ and $\pi^- + {^3{\rm H}} \rightarrow \gamma + n + n + n$ capture reactions are studied with the AV18 two-nucleon potential and the Urbana~IX three-nucleon potential. We provide for the first time realistic predictions for the differential and total capture rates for all these processes, treating consistently the initial and final nuclear states. Our results are based on the single nucleon Kroll-Ruderman-type transition operator and concentrate on the full treatment of the nuclear final state interactions. They are compared with older theoretical predictions and experimental data.


I. INTRODUCTION
Studies of radiative pion capture were initiated in 1951 by Panofsky, Aamodt and Hadley [1], who measured the ratio of mesonic (π − + p → n + π 0 ) to radiative (π − + p → n + γ) capture of stopped negative pions in hydrogen. Measurements of capture reactions in the early 1950s played an important role in fixing fundamental properties of the pions and their interactions with the nucleons. Later, experiments performed at the Lawrence Radiation Laboratory in Berkeley delivered photon spectra from radiative pion capture on different nuclei. Many such measurements with improved resolution were conducted in the 1970s at the Swiss Institute for Nuclear Research (SIN) (later the Paul Scherrer Institute (PSI)) by a collaboration including the Universities of Lausanne, Munich and Zurich. All the early experimental and theoretical work prior to January 1976 was summarized in Ref. [2]. Since we restrict ourselves to reactions with two and three nucleons in the present paper, we only refer the reader to Refs. [3][4][5][6][7][8][9][10][11][12] for later studies of systems with A > 3.
The studies of the π − + 2 H → γ + n + n reaction concentrated on the extraction of the 1 S 0 neutron-neutron scattering length, a nn . This reaction produces three detectable particles in a final state and the interaction of the photon with two emerging neutrons is so weak that the final state interaction (FSI) is absolutely dominated by the neutron-neutron (nn) scattering. As early as in 1951 Watson and Stuart [13] showed with quite simple dynamics that the corresponding photon spectrum very strongly depends on the properties of the nn interaction. Since then many theoretical efforts [14][15][16][17][18][19][20][21] combined with more and more precise measurements [22][23][24][25][26][27][28][29][30][31][32], have contributed decisively to our present day knowledge about a nn . Detailed information about this rich field can be found in the review byŠlaus, Akaishi, and Tanaka [33] and in the more recent review by Gårdestig [34], where also complementary efforts [35,36] to determine a nn from the 2 H(n, np)n reaction were reported.
In order to extract a nn from the π − + 2 H → γ + n + n process, various theoretical frameworks were employed. The approach formulated and applied by Gibbs, Gibson, and Stephenson [16,17,37] used a nonrelativistic one-body transition operator containing relativistic corrections and a nn scattering wave function generated from the Reid soft-core potential [38]. The wave function was obtained in coordinate space, starting from the asymptotic region and then integrating towards smaller inter-nucleon distances r. For r ≤ 1.4 fm a fifth degree polynomial with appropriate boundary conditions was used to represent the wave function. In order to minimize the error in the a nn extraction, the calculations had to be restricted to small relative nn energies, that is to the nn FSI peak region. While the first analysis of the experiment conducted at the Clinton P. Anderson Meson Physics Facility at Los Alamos (LAMPF) reported in Ref. [31] used theoretical cross sections derived with nonrelativistic phase space factors, the results published later in Ref. [32] were obtained with the corresponding relativistic formulas.
The second theoretical approach treated the nn rescatterings by means of Muskhelishvili-Omnès dispersion relations [39,40]. De Téramond and collaborators [18,19,41] considered various dynamical ingredients, like pion rescattering terms, off-shell effects, the impact of higher partial waves and studied their importance for the extraction of a nn . This theoretical framework was employed in the analysis of two experiments performed at SIN [27][28][29][30]. It is quite remarkable that the analyses of the SIN and LAMPF experiments led to equivalent results for a nn , with very small theoretical errors of 0.3 fm in a nn , even though they were based on different theories. Namely, the final result from the SIN experiment [30] was a nn = (−18.7 ± 0.6) fm, representing a weighted mean of two data sets with systematic and theoretical errors added in quadrature. The corresponding result from the LAMPF experiment reported in Ref. [32] read a nn = (−18.63 ± 0.27 (experiment) ± 0.30 (theory)) fm.
Further progress in the theoretical treatment of radiative pion capture reactions and the inverse process, pion photoproduction, was made in the framework of chiral effective field theory. In particular neutral pion photoproduction from a nucleon was studied by Bernard and collaborators [42] in the framework of heavy baryon chiral perturbation theory (HBCHPT). The same authors calculated also the one-loop corrections to the Kroll-Ruderman low-energy theorems for charged pion photoproduction at threshold in Ref. [43]. Within the same framework Fearing et al. [44] evaluated the transition amplitude for the photoproduction process away from threshold and obtained expressions for the s-and p-wave multipoles. They made connection with the radiative capture reaction at the cross section level via the detailed balance equation.
Several years later, Gårdestig and Phillips applied HBCHPT to the π − + 2 H → γ + n + n reaction [20,21,45] and for the first time used a consistent transition operator (with oneand two-body contributions) as well as the deuteron and nn scattering states. Namely, they worked in coordinate space and, starting from the asymptotic state, calculated the nn wave function solving the Schrödinger equation, which contained the lowest order chiral potential. For distances r smaller than a few fermis, a solution for the spherical well potential was chosen to account for the unknown short-distance physics. First calculations of Gårdestig and Phillips were carried out at next-to-next-to-leading order of chiral expansion and made it possible to extract a nn with a precision at the 0.2 fm level.
Further investigations [21,45] at higher order of chiral expansion revealed important relations between the short-distance physics in a number of reactions on light nuclei, since the same axial isovector two-body contact term was found to contribute in the radiative pion capture on the deuteron, pion production in nucleon-nucleon (NN) scattering, triton β decay, proton-proton fusion, neutrino-deuteron scattering, muon capture on the deuteron, nucleon-deuteron scattering and the p + 3 He → 4 He + e + + ν e (hep) process. The correlation found by Gårdestig and Phillips had direct impact on the accuracy with which a nn was estimated, and allowed them to reduce the theoretical error to approximately 0.05 fm.
Investigation of the radiative (and non-radiative) negative pion capture in the threenucleon (3N) bound states also started in the early 1950s. Even if this work prior to January 1976 is described in Ref. [2], we mention here a pioneering contribution by Messiah [46], who formulated a theoretical framework to deal with pion capture (not only radiative), discussed various dynamical aspects and gave predictions for all the six reaction channels for the 3 He nucleus, which were however based on very crude approximations in the closure formulas.
Further theoretical efforts to describe radiative capture concentrated on the π − + 3 He → γ + 3 H process. Some authors used very simple parametrizations of the 3 He and 3 H wave functions and the Kroll-Ruderman [47] form of the transition operator to calculate directly the pertinent nuclear matrix elements (see for example Ref. [48]). In other papers the so-called "elementary particle treatment of nuclei" was adopted [49][50][51][52]. Since the early measurements [53,54] yielded no absolute capture rates but rather relative probabilities of various processes, the theories tried to reproduce measured branching ratios. In particular the so-called Panofsky ratio, that is the ratio of the probabilities of the charge exchange π − + 3 He → π 0 + 3 H and radiative capture π − + 3 He → γ + 3 H reactions was studied in many papers, as exemplified by Refs. [55,56]. An important paper by Truöl et al. [57] not only brought new experimental data on the photon spectrum and several branching ratios but also corrections to the earlier theoretical predictions published in Refs. [48,52,58]. Using a formula from Ref. [59] and making a connection between the Gamow-Teller matrix element in triton β-decay and the corresponding matrix elements in non-breakup pion radiative capture on 3 He, the authors of Ref. [57] also provided a result for the π − + 3 He → γ + 3 H capture rate.
The photon spectrum and the branching ratios from Ref. [57] were analyzed by Phillips and Roig [55]. Radiative breakup rates were calculated in the impulse approximation and FSI among the three nucleons was treated in the Amado model [60,61], by solving the Faddeev equations with a simple separable s-wave NN potential. The 3N bound states were not calculated consistently but had an analytical form, which allowed the authors to regulate strengths of the principal S-state, S ′ -state and D-state components. Despite these simplifications, Phillips and Roig could describe the shapes of the experimental photon spectrum and the branching ratios given in Ref. [57]. They also predicted the decisive role of FSI and the dominant contribution of the π − + 3 He → γ + n + d channel in the radiative breakup of 3 He.
In Ref. [62], the same authors made calculations for the π − + 3 H → γ + n + n + n capture reaction, later confronted with experimental results obtained by Bistirlich et al. [63] and (in an improved experiment) by Miller et al. [64]. This reaction allows one to study the 3N system in a pure total isospin T = 3/2 state and to search for resonant or even bound three-neutron states. None were found in the two above-mentioned LAMPF experiments and the smooth shape of the experimental photon spectrum was in satisfactory agreement with the theoretical predictions by Phillips and Roig. In particular FSI raised the theoretical spectrum obtained under a plane wave impulse approximation. Since later all pion beam facilities were shut down and the more recent theoretical work on radiative pion capture was focused on the single-nucleon and two-nucleon (2N) sector, no calculations for the 3N system with modern realistic nuclear forces have been performed.
Recently we have established a theoretical framework for the A ≤ 3 muon capture reactions [65,66]. Important building blocks of this framework were cross-checked with the results from Ref. [67], obtained using the hyperspherical harmonics formalism [68]. It has then become very natural to adapt our momentum space techniques for corresponding radiative capture reactions. Thus we provide, for the first time, predictions with consistent treatment of the initial and final nuclear states calculated from realistic 2N and 3N forces for the differential and total capture rates of the π − + 2 H → γ + n + n, π − + 3 He → γ + 3 H, π − + 3 He → γ + n + d, π − + 3 He → γ + n + n + p, and π − + 3 H → γ + n + n + n reactions.
The paper is organized in the following way. In Sec. II we introduce the single nucleon transition operator. In the following sections, we show selected results, concentrating on the photon spectra and total capture rates. In Sec. III, we start with the π − + 2 H → γ + n + n reaction and demonstrate that our framework possesses the same sensitivity to a nn as the older [30,32] and more recent, chiral calculations [20]. Our results concerning π − + 3 He → γ + 3 H and breakup reactions with trinucleons are described in further sections, where we compare predictions obtained with different treatment of FSI. All our results for the total capture rates are shown together and compared to earlier theoretical predictions in Sec. VII. Finally, Sec. VIII contains our summary and outlook.

II. THE TRANSITION OPERATOR
The radiative capture process is treated in the same way as muon capture from the 1s atomic orbit. Namely, the initial state | i comprises the K-shell pion wave function | ψ and the initial nucleus state | Ψ i P i m i with the three-momentum P i and the spin projection m i : In the final state | f the photon occurs, described by the state | γ p γ ǫ with the threemomentum p γ and the polarization vector ǫ perpendicular to p γ , accompanied by the final nuclear state | Ψ f P f m f with the total three-momentum P f and the set of spin projections m f : We assume that the transition from the initial to final state is governed by the one-body Kroll-Rudermann operator j KR [47] and start with the simple nonrelativistic form from Ref. [2]. It is given as where e, g A , g V , f π and σ are the elementary charge, axial-vector and vector coupling constants, the pion decay constant and the nucleon spin operator, respectively. In the isospin formalism, also the isospin lowering operator, To make a simple connection to our work on muon capture [65,66] we replace Eq. (2.3) by where j A is the single nucleon axial current from Refs. [65,69]. Its matrix elements in momentum space of one nucleon read where M ≈ 939 MeV is the nucleon mass and the detailed information about the nucleon form factors g A 1 and g A 2 is provided in Ref. [69]. In practice it is always possible to setp γ = −ẑ, which means that for a real photon only two nuclear matrix elements of the transverse components of j A (here represented in the spherical notation) need to be calculated The transversality condition implies also that the term in Eq. (2.5) proportional to g A 2 does not contribute.
The form of the transition operator employed in this article definitely leaves room for improvement. Radiative pion capture by a single nucleon and the inverse, photoproduction, process was studied in heavy baryon chiral perturbation theory in Refs. [42][43][44] and corrections to the Kroll-Ruderman low-energy theorem were calculated. One can also expect many-nucleon, most importantly 2N, contributions to the capture process. They were derived more recently by Gårdestig [20,21,45].
Our first predictions, however, are based on the single nucleon transition operator and focus on other dynamical ingredients. Like for muon capture, we want to concentrate on FSI in the nuclear sector. To this end we calculate 2N and 3N scattering states using the AV18 NN potential [70] and the Urbana IX 3N force [71]. To the best of our knowledge, we provide, for the first time, consistent predictions for the total capture rates of the π − + 2 H → γ +n+n, π − + 3 He → γ+ 3 H, π − + 3 He → γ+n+d, π − + 3 He → γ+n+n+p, and π − + 3 H → γ+n+n+n reactions, obtained with realistic 2N and 3N forces.
Many elements of our calculations for all the listed reactions are essentially the same as performed for the corresponding muon capture reactions in Refs. [65,66]. In particular the formulas concerning kinematics can be directly used, if the muon mass is replaced by the negative pion mass. The radiative pion capture rates for the totally unpolarized reactions are also easily obtained from the corresponding expressions for the muon capture rates.
Our description of nuclear initial and final states is based on the nonrelativistic potentials and dynamical equations. It should then be used with the nonrelativistic kinematics. It is then mandatory to verify if the nonrelativistic approximations in the kinematics of the nuclear sector (the photon is of course treated relativistically) is justified. Clearly the pion is heavier than the muon, so pion absorption brings more energy to the nuclear system. Thus the comparisons of various results computed from the nonrelativistic and relativistic nuclear kinematics performed in Refs. [65,66] for muon capture had to be repeated with a new mass of the absorbed particle.
Starting from the energy and momentum conservation, we obtain first the maximal relativistic and nonrelativistic photon energies: and In Fig. 1, we demonstrate additionally that the kinematically allowed regions in the E γ − E n plane calculated relativistically and nonrelativistically essentially overlap, which means that the nonrelativistic kinematics can be safely used.
Taking the form of the transition matrix element into account, introducing the bosonic factors from the pion and photon fields and evaluating the phase space factor in terms of the relative nn momentum, p = 1 2 (p 1 − p 2 ) (p 1 and p 2 are the two individual neutron FIG. 1. The kinematically allowed region in the E γ − E n plane calculated relativistically (solid curve) and nonrelativistically (dashed curve) for the π − + 2 H → γ + n + n capture process. The lines practically overlap. momenta), we arrive at the following expression for the total unpolarized capture rate where the factor (M ′ d α ) 3 π stems from the K-shell atomic wave function, is the fine structure constant and p ≡ |p| = p(E γ ). We use in our calculations f π ≡ √ 2F π = 0.932 M π [2].
We can further simplify Eq. (3.3), since for the unpolarized case only the relative angle between p and p γ matters. Therefore we setp γ = −ẑ and choose the azimuthal angle of the relative momentum φ p = 0, which leads to: We generate the nuclear matrix elements N ±1 (m 1 , m 2 , m d ) in momentum space [65]. They contain plane wave as well as rescattering contributions. Although the three-dimensional formalism of Ref. [72] could be applied also to radiative pion capture in 2 H, in this article we discuss results obtained solely with a standard partial wave decomposition (PWD). In the calculations, we exclusively used the AV18 NN potential [70]. However, based on our experience from the muon capture process [65], we expect that predictions calculated with other realistic NN potentials would not differ significantly.
The calculations have been performed including all partial wave states with the total angular momentum j ≤ 4. In order to achieve fully converged results, 60 E γ points and 50 θ p points are used.
From Eq. (3.4), one can easily extract the differential capture rate dΓ nn /dE γ . This quantity is shown in the left panel of Fig. 2 for the plane wave part of N ±1 (m 1 , m 2 , m d ) ((PW) dashed line) and for the full N ±1 (m 1 , m 2 , m d ) ((Full) solid line). When the full result for N ±1 (m 1 , m 2 , m d ) is taken, a very narrow peak in the vicinity of E max,nn γ emerges.
Here, the relative energy of the two-neutron system is very small, which explains strong rescattering effects in dΓ nn /dE γ . Another form of the differential capture rate, dΓ nn /dp, is displayed in the right panel of Fig. 2, now as a function of the magnitude of the relative nn momentum. The transition between dΓ nn /dE γ and dΓ nn /dp reads Despite the fact that the shapes of the differential rates, shown in Fig. 2 for the plane wave and full dynamics are quite different, the corresponding integrated results for the total capture rate are rather similar. We obtain Γ nn = 0.318 ×10 15 1/s (PW) and Γ nn = 0.328 ×10 15 1/s (Full). Results of earlier calculations are displayed in Table I and discussed in Sec. VII.
The total capture rate Γ nn can be also evaluated using other variables where E γ is the only physical solution of the nonrelativistic equation and depends on the magnitude of the detected neutron momentum, p 1 , as well as on the angle between the detected neutron and photon momentum, θ γ1 . Note that the maximal neutron energy E max The building block of Eq. (3.6) is the differential capture rate d 5 Γ nn / (dp γ dp 1 dE 1 ) = 1 2 As early as in 1951 Watson and Stuart [13] showed with quite simple dynamics that the corresponding photon spectrum is very sensitive to the properties of the low-energy nn interaction. Since then many calculations [14-21, 32, 41], summarized in Ref. [34], have also demonstrated that this capture process can be used to study the properties of low-energy nn scattering. Also results of our calculations (see Figs. [3][4] show that the photon spectrum has two salient peaks: the nn FSI peak around E n ≡ E 1 = 2 MeV and the so-called QFS peak arising from the quasi-free π − p process around E n = 9 MeV. From Fig. 3, it is clear that heights of the peaks increase with increasing θ γ1 . What is more important, for fixed θ γ1 small variations of the 1 S 0 nn interaction lead to quite visible changes in the FSI peak. We study this effect, performing additional calculations with the altered nn AV18 potential whose 1 S 0 matrix elements are multiplied by the factor 1.01 and 0.99. This leads to the following changes of the a nn values: for the stronger version of the potential we get a nn = −21.8 fm, while for the weakened force a nn = −16.5 fm. (The original value is a nn = −18.8 fm.) The primary and modified neutron spectra are displayed in Fig. 4 and we see that the FSI peak becomes higher if the absolute value of a nn grows. Note that in all these calculations the nn AV18 potential was used without electromagnetic contributions.
Actually, in order to minimize systematic uncertainties in the a nn extraction, the very shape of the neutron time-of-flight spectrum in the area corresponding to the FSI peak is considered [30,32]. The nonrelativistic relation between the time-of-flight variable t 1 and the energy of the neutron E n reads where s is the flight path to the neutron detector. A simple step then leads to the neutron time-of-flight spectrum demonstrated in Fig. 5 (for s= 2.55 m) for the same three values of a nn : d 5 Γ nn / (dp γ dp 1 dt 1 ) = 1 2 (3.10) Equation (3.10) can be compared with the relativistic formula from Eq. (3) in Ref. [20]. We do not intend here to work on the extraction of a nn with our present theory. Before moving to calculations with 3N systems we wanted to make sure that our framework possesses the same important features as the calculations used in the old and more recent analyses of the π − + 2 H → γ + n + n process [14][15][16][17][18][19][20]32].
In this case, we deal with two-body kinematics and we can compare the photon energy calculated nonrelativistically and using relativistic equations. The relativistic result, based on FIG. 2. The differential capture rate dΓ nn /dE γ as a function of the photon energy E γ (a) and the differential capture rate dΓ nn /dp as a function of the magnitude of the relative nn momentum p (b) for the π − + 2 H → γ + n + n process, calculated with the AV18 potential [70] and using the transition operator from Eq. (2.4). The dashed curves show the plane wave results and the solid curves are used for the full results. FIG. 3. The differential capture rate d 5 Γ nn /(dp γ dp 1 dE 1 ) calculated with the AV18 potential [70] and using the transition operator from Eq. (2.4) as a function of the neutron energy E n ≡ E 1 for three different angles between the emitted photon and neutron momentum θ γ1 : 179 • (solid line), 175 • (dashed line) and 171 • (dotted line). All the results correspond to a nn = -18.8 fm.
In the nonrelativistic case, we start with   FIG. 4. The differential capture rate d 5 Γ nn /(dp γ dp 1 dE 1 ) as a function of the neutron energy FIG. 5. The differential capture rate d 5 Γ nn /(dp γ dp 1 dt 1 ) (neutron time-of-flight spectrum for a flight path of 2.55 m) obtained from the results shown in Fig. 4. The results corresponding to different a nn values, represented by the same lines as in Fig. 4, are normalized at the (left) QFS peak.
For this reaction we calculate only the total capture rate where the factor (2M ′ function and M ′ 3 He = M 3 He Mπ M 3 He +Mπ . For the unpolarized reaction the angular integration over the photon momenta leads to the 4π factor and in all the considered cases we setp γ = −ẑ. The phase space factor ρ is Like for muon capture, the additional factor R accounts for the finite volume of the 3 He charge and we use the same value R = 0.98 as in Ref. [67]. (The corresponding correction in the deuteron case has been neglected.) The nuclear matrix elements involve the initial 3 He and final 3 H states: and are obtained employing our standard PWD techniques [65,73].
Our results for this process are obtained for two cases. When we generate the 3 He and 3 H wave functions using the AV18 NN potential only, we get Γ3 H = 2.059 ×10 15 1/s. For the wave functions calculated with the AV18 NN potential augmented by the Urbana IX 3N force, the rate is slightly reduced to Γ3 H = 2.013 ×10 15 1/s. These predictions are compared with the results of earlier calculations in Table II and discussed in Sec. VII.
The kinematics of the π − + 3 He → γ + n + d and π − + 3 He → γ + n + n + p reactions is treated exactly in the same way as in muon capture on 3 He [65], so we can immediately evaluate the maximal photon energies for the two breakup channels as  Fig. 6. For both cases, lines obtained with the relativistic and nonrelativistic kinematics fully overlap except for very small photon energies. The same is also true for the three-body breakup, as demonstrated in Fig. 7 for the allowed region in the E γ − E p plane.
In this case the minimal proton kinetic energy is greater than zero for E γ > E 2sol γ (see the inset in Fig. 7) and the values of E 2sol γ based on the relativistic kinematics, and nonrelativistic one, yield very similar numerical values: 126.318 MeV and 126.314 MeV, respectively. All these results clearly show that the nonrelativistic kinematics can be safely used also for the breakup channels.
Using standard steps we obtain the formulas for the total capture rates. In the case of the two-body breakup it reads: where we used the relative neutron-deuteron momentum given in terms of the final neutron (p n ) and deuteron (p d ) momenta, to evaluate Γ nd . For the π − + 3 He → γ + n + n + p reaction we obtain in a similar way: Here the integral is expressed in terms of the Jacobi relative momenta p and q, that is obtained from the proton momentum (p p ) and the momenta of the two neutrons (p 1 and p 2 ). In Eq. (5.9) p max is a function of E γ and q ≡ |q| = q(E γ , p ) [66]. Note that we used the same geometrical arguments as before to simplify the angular integrations in Eqs. (5.7) and (5.9). The crucial matrix elements are calculated in momentum space, as outlined in Ref. [65], within the numerical framework developed in Refs. [73,74]. Also in Ref. [73] the detailed definitions of various 3N dynamics can be found.
Since the convergence of our PWD-based results with respect to the total subsystem angular momentum j and the total 3N angular momentum J was discussed in Ref. [65], we can start the discussion of our predictions with Fig. 8, where, for the π − + 3 He → γ + n + d reaction, we compare results of calculations employing various 3N dynamics: symmetrized plane wave approximation obtained with the AV18 NN potential, consistent calculations of the initial and final nuclear states with the AV18 interaction only and calculations based on the Hamiltonian containing additionally the Urbana IX 3N force. The results are qualitatively quite similar to the ones obtained for muon capture. The three differential capture rates dΓ nd /dE γ rise very slowly with the photon energy and form a single maximum close to the maximal photon energy. This maximum is higher and broader for the plane wave case. Effects introduced by FSI are very important and, in the maximum, reduce the full dΓ nd /dE γ to about 1/2 of the plane wave prediction. The inclusion of the 3N force lowers the peak further by about 14 % .
The FSI effects are even stronger for the π − + 3 He → γ +n+n+p reaction, as displayed in Fig. 9 for the differential rate dΓ nnp /dE γ . The reduction factor is already about 4.5 for the case without 3N force. The 3N force has roughly the same effect as in the two-body breakup case. In Ref. [65] we suggested that this might be a consequence of the overprediction of the A = 3 radii when 3N interaction is neglected.
Since the values of dΓ nnp /dE γ are much smaller (by a factor of 5) than dΓ nd /dE γ , at least in the peak area, the picture shown in Fig. 10 for the total breakup capture rate, dΓ br /dE γ = dΓ nd /dE γ + dΓ nnp /dE γ , is similar to the one for dΓ nd /dE γ . Finally, in Fig. 11, we show the contributions from the two-body and three-body breakup channels calculated with our full dynamics, that is including the 3N force. This figure clearly demonstrates that the breakup is dominated by the two-body channel. The corresponding predictions for the total Γ nd and Γ nnp capture rates are presented in Table II together with earlier theoretical predictions and experimental information about the relative probability of the breakup and non-breakup radiative capture. These results will be discussed in Sec. VII.
The kinematically allowed region in the E γ − E n plane for the three-body breakup of 3 H is shown in Fig. 12. As for the π − + 3 He → γ + n + n + p capture process, we show the border lines based on the relativistic and nonrelativistic kinematics and evaluate correspondingly the maximal photon energy relativistically E max,nnn  The kinematically allowed region in the E γ − E p plane calculated relativistically (solid curve) and nonrelativistically (dashed curve) for the π − + 3 He → γ + n + n + p process. The inset focuses on the highest photon energy region. The lines practically overlap except for very small photon energies.
In the calculations of the three-neutron continuum only the nn version of the AV18 potential [70] appears. The nuclear Hamiltonian contains the same Urbana IX 3N force [71]. The formula for the total Γ nnn capture rate,  FIG. 8. The differential capture rate dΓ nd /dE γ for the π − + 3 He → γ + n + d process as a function of the photon energy, calculated with the single nucleon transition operator and with different treatment of 3N dynamics: taking the symmetrized plane wave approximation (dotted line), calculating the initial and final 3N states without (dashed curve) and with 3N force (solid line). The calculations are based on the AV18 NN potential [70] and the Urbana IX 3N force [71]. FIG. 9. The same as in Fig. 8 for the differential capture rate dΓ nnp /dE γ in the case of the π − + 3 He → γ + n + n + p process.
The values of the differential capture rates dΓ nnn /dE γ are smaller (see Fig. 13) than the dΓ nnp /dE γ results for the three-body breakup of 3 He. The maximum is still broader for the plane wave case but FSI now raises the results by a factor of 1.8, playing a crucial role also for this process. The inclusion of the 3N force leads to a reduction of the peak's height by about 15 %. Also the total rates for pion capture on 3 H are displayed in Table II and described in the next section.

VII. COMPARISON WITH EARLIER THEORETICAL PREDICTIONS
A comparison with earlier theoretical predictions and experimental data is not simple for several reasons. Experimental work never actually aimed at obtaining total radiative capture rates. Very often measurements covered different reaction channels and tried to FIG. 11. Two-body (dotted curve) and three-body breakup contributions (dashed curve) to the differential breakup radiative pion capture rate (solid curve) as a function of the photon energy, calculated with the single nucleon transition operator and with full 3N dynamics (including 3N force in the initial and final nuclear states). As before, the calculations are based on the AV18 NN potential [70] and the Urbana IX 3N force [71].
gather information about their relative probabilities. The famous Panofsky ratio for 3 He, studied in Refs. [55,56], is a very good example. Unfortunately, such ratios involve rates, which are not calculated by us in the present paper. Also we do not have experimental (unnormalized) photon energy spectra for the reactions of interest at our disposal. Many calculations for the π − + 2 H → γ + n + n concentrated on the extraction of the nn scattering length and their authors did not provide results for the total radiative capture Γ nn . This is because already Gibbs, Gibson and Stephenson pointed out in Ref. [17] that the total radiative rate is clearly insensitive to uncertainties of the low-energy nn scattering parameters. Namely, they observed that variation of a nn from −15 to −20 fm or the effective range r nn from 2.6 to 3.0 fm changed Γ nn by less than 1 %. Thus predictions for the Γ nn capture rate can be found in only few theoretical papers [17,75,76].
Additional problems are caused by the fact that some predictions originate from combined theoretical and experimental evidence. Finally, earlier calculations are often corrected in FIG. 13. The same as in Fig. 8 for the differential capture rate dΓ nnn /dE γ in the case of the π − + 3 H → γ + n + n + n process.
subsequent publications. Nonetheless, we have tried to collect the available information and we show it in Tables I and II. In Table I the total radiative capture rates for the π − + 2 H → γ + n + n reaction are displayed. As already mentioned, for this observable our plane wave and full results are rather similar despite the fact that final state rescattering is very important and substantially affects the differential rates in Figs. 2 and 3. We notice also that our full predictions agree with the earlier theoretical results, except for Ref. [17]. Table II contains rates for radiative capture in the trinucleons. We consider several capture channels, starting from the only non-breakup process: π − + 3 He → γ + 3 H. Here the values of the rates are much higher than for the π − + 2 H → γ + n + n reaction. The capture rate is raised by approximately 3.5 %, when the 3N bound states are calculated consistently not only with the 2N but also with the 3N potentials.
The breakup of 3 He is clearly dominated by the π − + 3 He → γ + n + d reaction, since the rate for the π − + 3 He → γ + n + n + p process is three times smaller. For both breakup reactions FSI effects based on the 2N forces are very strong and reduce the rates significantly. The inclusion of the 3N force leads to a further reduction, which amounts to 9 % (two-body breakup) and to 7 % (three-body breakup). Our best results (obtained with the 3N force) for the ratio of the total breakup rate to the non-breakup rate agree both with the experimental data from Ref. [57] and with the theoretical prediction in Ref. [55]. That means that the non-breakup and breakup channels in radiative pion capture in 3 He are equally important.
Comparing the total rates for the π − + 3 H → γ + n+ n+ n reaction one might draw a false conclusion that FSI and 3N force effects are very small. From Fig. 13 it is, however, clear that the agreement between the plane wave and full results for the total rates is rather accidental, since the differential rates are quite different. Contrary to the three-body breakup of 3 He, FSI effects enhance the plane wave result. The 3N force reduces the rate by approximately 10 %. Our result for the total radiative capture rate is by about 70 % larger than the prediction from Ref. [62].
In view of the fact that the theoretical results obtained before were based on quite different approaches, in many cases the agreement with earlier theoretical predictions is satisfactory. In particular, we obtained similar shapes of the photon energy spectra for all studied reactions. Our predictions about the role of the final state interactions based on the realistic semi-phenomenological nuclear forces are fully converged with respect to the number of partial wave states. That means that, in contrast to Phillips and Roig [55,64], we are ready to calculate not only capture rates but also any polarization observables. Our much more advanced model confirms qualitatively the S-wave based results for capture rates from Refs. [55,64]. Thus it will be very interesting to compare results of modern calculations performed with improved transition operator and consistent nuclear forces.  [76] (1976) 0.375 (based on pion photoproduction data) Ref. [76] (1976) 0.383 (based on soft-pion limit) Ref. [17] (1977) 0.420 ± 0.05

VIII. SUMMARY AND CONCLUSIONS
Recent theoretical work by Gårdestig and Phillips [21,45] shows that radiative pion capture is not only interesting by itself but also correlated with a variety of other processes when studied within chiral effective field theory. This is very important because the other reactions (like muon capture) or neutrino induced processes are much harder to measure  [62] (1975) 0.07 and the information from the radiative capture on light nuclei could significantly improve our understanding of the other processes. Thus a uniform framework for the calculations of various electromagnetic and weak reactions on the single nucleon, deuteron, 3 He, 3 H and other light nuclei should be formulated and applied. This framework would comprise consistent two-nucleon and more-nucleon forces as well as transition operators ("currents") with one-body and many-body parts. Results of fully converged calculations should be ultimately compared with precise experimental data, to yield a broad and complete picture of these reactions in few-nucleon systems.
In the present paper, we studied the π − + 2 H → γ + n + n, π − + 3 He → γ + 3 H, π − + 3 He → γ + n + d, π − + 3 He → γ + n + n + p and π − + 3 H → γ + n + n + n reactions using traditional nuclear forces (the AV18 NN potential and the Urbana IX 3N force) and a simple single-nucleon transition operator. These calculations, like our studies of muon capture [65,66] or very recent investigations of some neutrino induced reactions [77], are ready to be systematically improved to encompass more complicated dynamical input. Many aspects of the performed calculations, like the role of the relativistic kinematics, the efficient methods of partial wave decomposition or the convergence of our results with respect to the number of partial wave states, have been already established and the predictions presented here can serve as an important benchmark.
Our calculations already provide first realistic predictions for the differential dΓ3 H /dE γ , dΓ nd /dE γ , dΓ nnp /dE γ and dΓ nnn /dE γ capture rates as well as for the corresponding total radiative capture rates Γ3 H , Γ nd , Γ nnp and Γ nnn . The formalism used in the present paper will be in the future extended to study other pion capture reactions, including non-radiative and double radiative pion capture.