WorksPhysics and Astronomy
Pion fragmentation functions from a quark-jet model in a functional approach
Abstract
The elementary fragmentation function that describes the process is predicted applying crossing and charge symmetry to the cut diagram of the pion valence quark distribution function. This elementary probability distribution defines the ladder-kernel of a quark jet fragmentation equation, which is solved self-consistently to obtain the full pion fragmentation function. The hadronization into a pion employs the complete Poincaré invariant Bethe-Salpeter wave function, though the overwhelming contribution to the fragmentation function is due the leading Bethe-Salpeter amplitude. Compared to a Nambu–Jona-Lasinio model prediction, the fragmentation function we obtain is enhanced in the range but otherwise in good qualitative agreement. The full pion fragmentation function is overall greater than the elementary fragmentation function below .
Introduction
Many high-energy scattering processes can be treated perturbatively in Quantum Chromodynamics (QCD) within the framework of factorization theorems. The latter allow for a separation of the perturbatively calculable part of a given cross section from the nonperturbative matrix elements at a given energy scale, where the latter are frequently codified in parton distribution functions. They also are fundamental tools in understanding the internal structure of protons and other hadrons.
Likewise, quark jet fragmentation functions describe how the energy and momentum are distributed amongst highly energetic hadrons that form a jet with nearly parallel longitudinal momenta and negligible transverse momenta [1]. Such jets are produced in high-energy collisions where reaction and decay processes produce gluon and quark partons that fragment into a cascade of hadrons once they escape the interaction region. Besides being of fundamental interest in their own right, their knowledge is essential for the reconstruction of the transversity quark distribution functions [2, 3]. At the heart of quark jet fragmentations functions are elementary fragmentation functions of quarks and gluons [4], which describe the probability for such a parton to fragment into a single hadron with a given fraction of its light-front momentum.
In here, we obtain the elementary fragmentation function from the corresponding cut diagram [5, 6, 7] using a functional approach to QCD modeling. In doing so, we solve the Dyson-Schwinger equation (DSE) for a light quark, while the solution of the Bethe-Salpeter equation (BSE) in the pseudoscalar channel yields the pion wave function. The fragmentation function for a physical light-front momentum fraction and the parton distribution function of a quark in a pion for unphysical obey a Drell–Levy–Yan (DLY) relation [8, 9, 10] due to charge conjugation and crossing symmetry. However, in a single step fragmentation process, such as , momentum and isospin sum rules cannot be satisfied [7] and one must consider the possibility that the fragmenting quark produces a cascade of mesons. The infinite fragmentation tower can in principle involve any hadrons, though we here limit ourselves to the hadronization into a cascade of pions described by jet functions, much alike the original quark jet-model of Fields and Feynman [1]. Our approach satisfies momentum and isospin sum rules and represents a systematic extension of a Nambu–Jona-Lasinio (NJL) model calculation [7] where the Poincaré invariant Bethe-Salpeter amplitude is not point-like but possesses a spatial extension and the quark masses are momentum dependent.
In Section II we present the theoretical framework, namely the DSE and BSE that describe the quark and the bound state wave function of the pion, respectively. The elementary fragmentation function we obtain is discussed in Section III and our prediction is compared to that of a NJL model [7]. We emphasize that while the latter is about 10–12% suppressed compared with the fragmentation function obtained in the DSE-BSE approach, the qualitative behavior of both functions is very similar. This is chiefly due to the fact that the fragmentation function is dominated by the contribution of the leading Bethe-Salpeter amplitude, which mirrors the NJL calculation where the wave function is a single, point-like Lorentz covariant. We then discuss the quark-jet fragmentation equations from which we obtain full fragmentation function that describes the infinite tower of quark hadronization into pions of different isospin. We wrap up with concluding remarks in Section IV.
From quark to pion: gap and bound-state equations
Dressed quark propagators
In a nonperturbative approach to the quark fragmentation function, the fully dressed quark propagator is a crucial ingredient. We obtain this propagator from the corresponding DSE of a dressed quark,
whose solutions are Euclidean Schwinger functions [11, 12]. In Eq. (1), is the bare current-quark mass, and are the vertex and wave-function renormalization constants at the renormalization scale , respectively. The integral describes the self-energy of the quark and involves the dressed-quark propagator , the dressed-gluon propagator with and the quark-gluon vertex, [13, 14, 15, 16], where are SU(3) color matrices. The regularization scale can be taken to infinity. For a given flavor, the general covariant form of the dressed propagator is,
where is the wave renormalization function and is the mass function of the quark, and the functions and can be related to them. In a subtractive renormalization scheme one imposes the conditions,
where is the renormalized current-quark mass related to the bare mass by,
and is the mass-renormalization constant in the QCD Lagrangian.
The leading rainbow truncation of the DSE (1) and the ladder truncation of the corresponding BSE kernel has proven to be a robust approximation that preserves the axialvector Ward-Green-Takahashi identity and allows for the description of light ground-state mesons in the isospin-nonzero pseudoscalar and vector channels. In essence, the fully dressed quark-gluon vertex is reduced to the perturbative vertex, . Therefore, the DSE kernel becomes [17],
This truncation is characterized by an Abelianized Ward identity and, as in QED, implies [11] due to the omission of the three-gluon interaction in . The additional factor in Eq. (6) ensures multiplicative renormalizability of the DSE and consequently is a renormalization-point invariant quantity [18].
The free-gluon propagator is transverse in Landau gauge,
and the dressing function is composed of two pieces,
where we deliberately absorb a factor of the gluon propagator (7). The first term in Eq. (8) dominates in the infrared domain, , and is suppressed at large momenta, while the second term describes a bounded and monotonically decreasing continuation of the perturbative coupling that dominates at large momenta: . In here, we use the infrared-finite dressing model of Ref. [19] which reads,
where is the anomalous mass dimension and is the active flavor number, , , , , and . In solving the DSE (1), we determine the renormalization constants and with the conditions and [20].
Pion wave function
The pions produced in the fragmentation of a quark are described by a Poincaré-covariant Bethe-Salpeter amplitude (BSA), . The latter is the solution of the homogeneous BSE for pseudoscalar mesons,
where is the relative quark-antiquark momentum, is the on-shell pion momentum, , and is the fully amputated scattering kernel. The Bethe-Salpeter wave function of the pion, is defined by attaching the quark propagators to the BSA,
with the momentum-partition parameters and the quark and antiquark momenta and . The ladder truncation compatible with the rainbow truncation of the DSE in Eq. (6) reads,
and is the gluon momentum exchanged infinitely many times between the quark and antiquark.
The BSA is composed of Dirac matrices and the relative and total momenta consistent with the quantum numbers . Its most general Poincaré-covariant form is given by [21],
in which , , , are Lorentz-invariant scalar amplitudes. In the calculation of the elementary fragmentation function (16), we employ the BSE solution of Ref. [20] for the pion.
Fragmentation functions
Elementary quark fragmentation function
Fragmentation functions describe the hadronization of highly energetic partons with nearly longitudinal momenta, though the transverse momentum dependence can be included [3]. The original parton model of fragmentation [1] is nowadays understood as the process of very energetic quarks and gluons escaping the initial collision region and then fragmenting into jets of hadrons.
The elementary fragmentation function of the process for physical light-front momentum fraction can be related to the parton distribution function for unphysical using crossing and charge-conjugation symmetry,
which is the DLY relation [8, 9, 10], and is the fraction of light-front momentum transferred from a valence quark to a pion. This elementary fragmentation function represents a probability density and can be formulated with the cut diagram depicted in Figure 1 whose evaluation leads to the expression [7]:

Figure 1. Cut diagram of the fragmentation function in Eq. (16). Shaded circle with outgoing/incoming double-solid line denote the pion in the fragmentation process, solid lines are quark propagators and the solid dots represent the between the two quarks with momentum .
In Eq. (16), , and are the incoming and outgoing four-momenta of the quark, is the fully dressed quark propagator.
and are the BSA (14) and the charge-conjugate BSA of the pion with total momentum , respectively. is an isospin factor. The light-front momenta are defined as , and , where is a light-like four-vector: .
In order to calculate the fragmentation function (16) in light-front variables, a suitable algebraic representation of the BSA and quark propagators is indispensable1, as the dressing functions and scalar amplitudes are only available as numerical functions of their momentum squared and angles between the relative and total momentum of the meson [20, 22, 23, 24, 25]. As discussed in more detail in appendix A, the generalized Nakanishi integral representation we use for the BSA and the complex-conjugate pole parametrization of the quark propagators depend on squared momenta and scalar products of momenta which we define in the following.
We follow Ref. [26] and define light-front variables in Euclidean metric:
where . In the rest frame of the pion, , and therefore . Evaluating the trace in Eq. (16) leads to scalar products which in light-front coordinates read,
while the four-momentum integral is:
As discussed in Refs. [2, 27], Eq. (16) refers to the frame where the produced meson has no transverse momentum, , whereas the fragmenting quark is characterized by non-zero . In order to interpret this fragmentation function as a distribution of the pion in the quark, the quark has to be boosted to the frame where . This can be achieved with a Lorentz transformation and it follows that one may simply substitute,
Note that we distinguish between both frames with the subscripts and , that is but .
We recall that the quarks described by a running mass functions (2) are confined and therefore there is no solution of the mass-shell equation . On the other hand, , provides a working definition for a Euclidean-quark mass, and therefore a realistic estimate of the quark’s active constituent mass. We here use: [17]. With this, the second Dirac function in Eq. (16), using Eqs. (18), (19), and (21) and , becomes,
The first Dirac function in Eq. (16) leads to after integration over which modifies the second Dirac function to,
Integrating over , and become:
With this, the integral over can be calculated numerically and normalized according to,
which satisfies isospin and momentum sum rules. We have used the isospin notation and in Eqs. (16) and (26).
The elementary fragmentation function is compared to that obtained with a NJL model [7] in Figure 2. Both predictions are qualitatively in good agreement, though the fragmentation function obtained in the present work is clearly enhanced up to and slightly softer for larger . In order to understand the similarity of both predictions, we plot the contributions of and (14) to in Figure 3. Clearly, the fragmentation function is dominated by the leading amplitude, not unlike the single covariant though point-like pion wave function employed in the NJL calculation [7].
Figure 2. The elementary fragmentation function derived from the expression in Eq. (16) compared with a prediction in the NJL model [7]. The red-shaded error band stems from the systematic error in fitting the generalized Nakanishi integral representation (A7), as discussed in Ref. [20].
Figure 3. and contributions (14) to the fragmentation function. The contributions of and are not plotted, as they are negligible.
Since we use a gluon-model interaction in Eq. (8), the renormalization scale of the quark propagators is not that of a given scheme in perturbation theory. Following Ref. [7, 28], we also compute in the present approach the pion’s PDF. We then match the first moment of the valence quark extracted from a Drell-Yan analysis, [29, 30], by using next-to-leading order Dokshitzer-Gribov-Lipatov-Altarelli-Parisi (DGLAP) evolution [31] including strange- and charm-quark mass thresholds from the experimental scale, , to our model scale. For the latter we find .
Quark-jet functions
There is no reason to assume the quark fragments into a single pion described by the amplitude in Eq. (16). On the contrary, one must consider the possibility that the fragmenting quark produces a cascade of mesons. In the quark-jet model formulated by Field and Feynman [1], the meson observed in a semi-inclusive process is one amongst many others which form a jet of hadrons. The probability for finding a pion with light-front momentum fraction in the jet is described by the full fragmentation or jet function , obtained via an iterative resummation of all possible fragmentations.
With the convenient isospin decomposition [7],
one can distinguish between favored, unfavored and neutral fragmentation functions:
The unfavored functions denote those fragmentation functions where the hadronizing quark or antiquark is
not a valence quark of the pion. The functions and of Eq. (27) are obtained by solving the set of Volterra integral equations [7],
which are normalized as and . In the limit the full fragmentation function Eq. (27) reduces to the elementary one,
since in this case the quark gives all its momentum to the initial pion, leaving no momentum left for a jet. The solution to Eqs. (31) and (32) is plotted in Figure 4 for the pion fragmentation function along with the elementary fragmentation function. We observe that is overall enhanced and greater than elementary fragmentation function in the range .
Figure 4. The elementary fragmentation functions and [normalized as in Eq. (26)] and the pion fragmentation function at the scale , where error bands are as in Figure 2.
The moments of the fragmentation functions are defined as,
and at the model scale we find:
After leading order DGLAP evolution of the fragmentation functions [32] including thresholds, their moments at a scale GeV are:
The moments in Eqs. (36a) and (36b) describe the momentum fractions of the , and pairs generated in the evolution equations from the original -quark. The corresponding pion fragmentation functions evolved to the scale GeV are drawn as functions of in Figure 5. Notably, as seen from comparison of Figure 4 and Figure 5, the fragmentation function is strongly enhanced at small , while the hump structure persists even though the original maximum is considerably lower. In the evolution we assume that the pion is made of and quarks only at the model scale. It has been advocated [33] that a singlet jet equation in QCD should involve gluon contributions to the cascade because of gluon and mixing, which implies that contributes to momentum conservation. Doing so would obviously lead to an enhancement of in Figure 5, that is if we assumed some model for that fragmentation function at our model scale .
Figure 5. The fragmentation functions evolved from the model scale to and the evolution-generated gluon, , and quark fragmentation functions. In the latter, the momentum fractions are of , and pairs generated by the initial -quark in the DGLAP evolution [32].
Conclusive Remarks
We present the first calculation of the elementary fragmentation function in the DSE-BSE framework employing a cut-diagram representation of the fragmentation process and a convenient algebraic representations of the pion’s BSA and quark propagators. The latter allows us to analytically integrate over the light-front variables and while the transverse momentum can be integrated numerically. The fragmentation function we obtain is in qualitative agreement with an earlier calculation in the NJL model, but is about 10–12% enhanced for momentum fractions .
This elementary fragmentation function feeds into the kernel of a quark jet fragmentation equation, which is solved to obtain the full pion fragmentation function. The resulting functions and , amongst others, describe the probability that a quark escaping the collision region produces a positively or negatively charged pion, respectively, thereby transferring a fraction of its light-front momentum .
The present approach is readily applicable to other light and heavy meson fragmentation functions and calculations are underway. As a light quark can fragment in a cascade of mesons (and baryons) other than pions, this leads to a set of coupled quark jet fragmentation functions that must be solved self-consistently [34]. Future improvements must also consider gluon contributions to the isospin-singlet fragmentation function at the initial hadronic scale, as one expects gluon and mixing. The fragmentation of a gluon poses a more daunting challenge, as the topology of their hadronization involves higher-order diagrams and a robust ansatz for the quark-gluon vertex. Nevertheless, their calculation will reveal unknown aspects of hadronization and shed light on the underlying confinement mechanism.
Acknowledgments
We are indebted to Ian Cloët for very helpful discussions. B. E. acknowledges support by the São Paulo Research Foundation (FAPESP), grant no. 2023/00195-8, and the National Council for Scientific and Technological Development (CNPq), grant no. 409032/2023-9, and R. C. S is supported by a postdoctoral CAPES-PIPD fellowship. B. E. and R. C. S are members of the Brazilian network project INCT-Física Nuclear e Aplicações, grant no. 464898/2014-5.
Algebraic representation of propagators and Bethe-Salpeter amplitudes
Since we work with light-front variables, a suitable algebraic representation for the propagators and bound-state amplitudes that enter the fragmentation function (16) is required. A sum of complex conjugate mass poles [35, 36] provides an adequate reproduction of the numerical solution of the scalar and vector dressing functions of the quark propagator ,
where are complex-valued light-quark mass scales and are complex coefficients fitted to the numerical values of and . Their values are found in Table 2 of Ref. [20].
The Lorentz-invariant amplitudes of the BSA (14) are expanded in terms of Chebyshev moments, ,
which results in a faster convergence in solving the BSE. We consider Chebyshev polynomials of the second kind, where is an angle between and . The eigenvalue problem for the vector that describes the pseudoscalar meson is solved by means of Arnoldi factorization implemented with the ARPACK library [37]. For details we refer, for instance, to Refs. [38, 39].
One can separate the BSA in even and odd components,
where are even under and for flavorless pseudoscalar mesons, as they are eigenstates of the charge-conjugation operator defined as,
In here, we assume isospin symmetry with . The constraint that the covariant basis (14) must satisfy with for pseudoscalar mesons is the reason that is even.
The four scalar amplitudes are parametrized with a generalized Nakanishi integral representation [40, 41],
where , and the spectral density is given by,
and are Gegenbauer polynomials of order . The parameters , , and for the pion are listed in Table IV of Ref. [24]. The amplitudes are fitted to the sum of the 0th and 2nd Chebyshev moments.
References
References
- [1]R. D. Field and R. P. Feynman, Phys. Rev. D 15, 2590-2616 (1977) doi:10.1103/PhysRevD.15.2590DOI
- [2]V. Barone, A. Drago and P. G. Ratcliffe, Phys. Rept. 359 (2002), 1-168 doi:10.1016/S0370-1573(01)00051-5DOI
- [3]H. H. Matevosyan, W. Bentz, I. C. Cloët and A. W. Thomas, Phys. Rev. D 85, 014021 (2012) doi:10.1103/PhysRevD.85.014021DOI
- [4]A. Metz and A. Vossen, Prog. Part. Nucl. Phys. 91 (2016), 36-202 doi:10.1016/j.ppnp.2016.08.003DOI
- [5]R. L. Jaffe, Nucl. Phys. B 229 (1983), 205-230 doi:10.1016/0550-3213(83)90361-9DOI
- [6]W. Bentz, T. Hama, T. Matsuki and K. Yazaki, Nucl. Phys. A 651 (1999), 143-173 doi:10.1016/S0375-9474(99)00130-XDOI
- [7]T. Ito, W. Bentz, I. C. Cloët, A. W. Thomas and K. Yazaki, Phys. Rev. D 80, 074008 (2009) doi:10.1103/PhysRevD.80.074008DOI
- [8]S. D. Drell, D. J. Levy and T. M. Yan, Phys. Rev. 187, 2159-2171 (1969) doi:10.1103/PhysRev.187.2159DOI
- [9]S. D. Drell, D. J. Levy and T. M. Yan, Phys. Rev. D 1, 1035-1068 (1970) doi:10.1103/PhysRevD.1.1035DOI
- [10]S. D. Drell, D. J. Levy and T. M. Yan, Phys. Rev. D 1, 1617-1639 (1970) doi:10.1103/PhysRevD.1.1617DOI
- [11]A. Bashir, L. Chang, I. C. Cloët, B. El-Bennich, Y. X. Liu, C. D. Roberts and P. C. Tandy, Commun. Theor. Phys. 58, 79-134 (2012) doi:10.1088/0253-6102/58/1/16DOI
- [12]I. C. Cloët and C. D. Roberts, Prog. Part. Nucl. Phys. 77, 1-69 (2014) doi:10.1016/j.ppnp.2014.02.001DOI
- [13]L. Albino, A. Bashir, L. X. G. Guerrero, B. E. Bennich and E. Rojas, Phys. Rev. D 100, no.5, 054028 (2019) doi:10.1103/PhysRevD.100.054028DOI
- [14]L. Albino, A. Bashir, B. El-Bennich, E. Rojas, F. E. Serna and R. C. da Silveira, JHEP 11, 196 (2021) doi:10.1007/JHEP11(2021)196DOI
- [15]B. El-Bennich, F. E. Serna, R. C. da Silveira, L. A. F. Rangel, A. Bashir and E. Rojas, Rev. Mex. Fis. Suppl. 3, no.3, 0308092 (2022) doi:10.31349/Sup lRevMexFis.3.0308092
- [16]J. R. Lessa, F. E. Serna, B. El-Bennich, A. Bashir and O. Oliveira, Phys. Rev. D 107, no.7, 074017 (2023) doi:10.1103/PhysRevD.107.074017DOI
- [17]F. E. Serna, C. Chen and B. El-Bennich, Phys. Rev. D 99, no.9, 094027 (2019) doi:10.1103/PhysRevD.99.094027DOI
- [18]J. C. R. Bloch, Phys. Rev. D 66, 034032 (2002) doi:10.1103/PhysRevD.66.034032DOI
- [19]S. x. Qin, L. Chang, Y. x. Liu, C. D. Roberts and D. J. Wilson, Phys. Rev. C 84, 042202 (2011) doi:10.1103/PhysRevC.84.042202DOI
- [20]F. E. Serna, R. C. da Silveira, J. J. Cobos-Martínez, B. El-Bennich and E. Rojas, Eur. Phys. J. C 80, no.10, 955 (2020) doi:10.1140/epjc/s10052-020-08517-3DOI
- [21]C. H. Llewellyn-Smith, Annals Phys. 53, 521-558 (1969) doi:10.1016/0003-4916(69)90035-9DOI
- [22]F. E. Serna, R. C. da Silveira and B. El-Bennich, Phys. Rev. D 106, no.9, L091504 (2022) doi:10.1103/PhysRevD.106.L091504DOI
- [23]F. E. Serna and B. El-Bennich, PoS CHARM2020, 047 (2021) doi:10.22323/1.385.0047DOI
- [24]R. C. da Silveira, F. E. Serna and B. El-Bennich, Phys. Rev. D 107, no.3, 034021 (2023) doi:10.1103/PhysRevD.107.034021DOI
- [25]F. E. Serna, B. El-Bennich and G. Krein, Phys. Rev. D 110, no.11, 114033 (2024) doi:10.1103/PhysRevD.110.114033DOI
- [26]G. Eichmann, E. Ferreira and A. Stadler, Phys. Rev. D 105, no.3, 034009 (2022) doi:10.1103/PhysRevD.105.034009DOI
- [27]J. C. Collins and D. E. Soper, Nucl. Phys. B 194, 445-492 (1982)
- [28]C. Shi and I. C. Cloët, Phys. Rev. Lett. 122, no.8, 082301 (2019) doi:10.1103/PhysRevLett.122.082301DOI
- [29]P. J. Sutton, A. D. Martin, R. G. Roberts and W. J. Stirling, Phys. Rev. D 45, 2349-2359 (1992) doi:10.1103/PhysRevD.45.2349
- [30]M. Glück, E. Reya and I. Schienbein, Eur. Phys. J. C 10, 313-317 (1999) doi:10.1007/s100529900124DOI
- [31]M. Botje, Comput. Phys. Commun. 182, 490-532 (2011) doi:10.1016/j.cpc.2010.10.020DOI
- [32]M. Hirai and S. Kumano, Comput. Phys. Commun. 183 (2012), 1002-1013 doi:10.1016/j.cpc.2011.12.022DOI
- [33]H. Y. Xing, Z. Q. Yao, B. L. Li, D. Binosi, Z. F. Cui and C. D. Roberts, Eur. Phys. J. C 84 (2024) no.1, 82 doi:10.1140/epjc/s10052-024-12403-7DOI
- [34]A. Casey, H. H. Matevosyan and A. W. Thomas, Phys. Rev. D 85 (2012), 114049 doi:10.1103/PhysRevD.85.114049DOI
- [35]M. Bhagwat, M. A. Pichowsky and P. C. Tandy, Phys. Rev. D 67, 054019 (2003) doi:10.1103/PhysRevD.67.054019DOI
- [36]B. El-Bennich, G. Krein, E. Rojas and F. E. Serna, Few Body Syst. 57, no.10, 955-963 (2016) doi:10.1007/s00601-016-1133-x
- [37]R. B. Lehoucq, D. C. Sorensen, C. Yang, ARPACK Users’ Guide: Solution of Large-Scale Eigenvalue Problems with Implicitly Restarted Arnoldi Methods (Society for Industrial & Applied Mathematics, 1998).DOI
- [38]M. Blank and A. Krassnigg, Comput. Phys. Commun. 182, 1391-1401 (2011) doi:10.1016/j.cpc.2011.03.003DOI
- [39]E. Rojas, B. El-Bennich and J. P. B. C. de Melo, Phys. Rev. D 90, 074025 (2014) doi:10.1103/PhysRevD.90.074025DOI
- [40]N. Nakanishi, Phys. Rev. 138 (1965), B1182-B1192 doi:10.1103/PhysRev.138.B1182DOI
- [41]N. Nakanishi, Phys. Rev. 139 (1965), B1401-B1406 doi:10.1103/PhysRev.139.B1401DOI



