betaFeedback

WorksPhysics and Astronomy

Pion and kaon electromagnetic form factors in a flavor-dressed contact interaction: the role of Nf-screening on dynamical chiral symmetry breaking

Abstract

We compute the electromagnetic form factors of the pion and the kaon in a symmetry-preserving contact interaction whose effective coupling depends on the number of active light flavors Nf , so that dynamical chiral symmetry breaking is screened as Nf grows toward the chiral transition. The calculation uses the rainbow-ladder truncation of the Schwinger-Dyson and Bethe-Salpeter equations with proper-time regularization, and it includes the dressed quark-photon vertex obtained from its own inhomogeneous Bethe-Salpeter equation in the same truncation. At Nf = 2 the model reproduces the established contact interaction values of the meson masses, decay constants and condensates, and the charge radii rpi = 0.447 fm and rK = 0.417 fm, which undershoot experiment by about one third, as is known for this class of interactions. We therefore restrict every comparison to Q2 less then or equal to 0.6 GeV2, confront it with the NA7 data and with one-loop chiral perturbation theory, and focus on the response to Nf . Both radii grow monotonically with Nf , and the growth survives every variation of the screening function and of the infrared confinement scale that we test. The growth is carried by the quark core rather than by the vector-meson dressing: between Nf = 2 and 6 the bare-vertex contribution to <r2pi> rises from 0.097 to 0.226 fm2, while the dressing contribution moves from 0.102 to 0.127 fm2 even though the rho pole drops from 0.93 to 0.70 GeV. The finite strange mass slows the kaon response only moderately.

Introduction

The pion occupies a singular position in QCD. As the lightest hadron and the pseudo-Goldstone boson of dynamical chiral symmetry breaking (DCSB), it is both the clearest signal that chiral symmetry is spontaneously broken in the QCD vacuum and the simplest bound state available for studying quark confinement and strong-interaction dynamics. These two roles are not independent. The pion is light precisely because chiral symmetry is broken, and its internal structure is shaped by the same nonperturbative mechanisms that generate the bulk of visible mass in the universe [1, 2]. Any quantitative model of the pion must therefore capture both faces simultaneously. Quantum chromodynamics admits two qualitatively distinct regimes. At short distances the running coupling becomes small, and the theory is perturbative [3, 4]. At large distances the coupling grows, quarks and gluons are confined within color-singlet hadrons, and perturbation theory is useless [5]. The properties of the pion, including its mass, decay constant, and electromagnetic structure, are determined entirely in this nonperturbative sector. Accessing them requires either lattice QCD or a continuum approach to the Schwinger-Dyson equations (SDEs) of the theory.

Within the SDE framework, the rainbow-ladder (RL) truncation provides the minimal consistent scheme that preserves the axial-vector Ward-Takahashi identity (axWTI) and therefore correctly encodes Goldstone boson dynamics [6, 7]. In this setting, the pion emerges as a simultaneous solution of the quark gap equation and the homogeneous Bethe-Salpeter equation (BSE). Early studies using a dressed gluon propagator in the Landau gauge, the Maris-Tandy model, produced pion and kaon form factors in good agreement with available data [8, 9]. A simplified but analytically tractable alternative is the contact interaction (CI) model, in which the gluon-mediated interaction between quarks is replaced by a momentum-independent coupling. The CI model, regularized in a symmetry-preserving way via the Schwinger proper-time scheme, describes pion static properties well but yields form factors that are too hard: the charge radii come out too small, and the form factors approach nonzero constants at large spacelike momentum transfer, a direct consequence of the momentum-independent Bethe-Salpeter amplitude [10, 11, 12, 13]. It is worth stating plainly where this model sits. A momentum-independent gluon exchange between quarks is, at the level of the Lagrangian, a four-fermion contact term, so the framework used here belongs to the Nambu-Jona-Lasinio (NJL) family [14, 15]. Two features separate it from the textbook NJL model. The first is the proper-time regularization with a second, infrared cutoff, which removes the real-axis quark production threshold and thereby supplies a confinement mechanism that NJL lacks. The second is the explicit dependence of the coupling on the number of active flavors. Neither modification changes the fact that the rainbow-ladder truncation1 of the Schwinger-Dyson and Bethe-Salpeter system is, for a contact interaction, the same resummation as the Hartree plus random-phase approximation of the NJL model at leading order in 1/Nc1/N_c. Meson observables in that framework have a long history, and the pion electromagnetic and scalar form factors in particular have been computed in the extended NJL model with vector and axial-vector degrees of freedom [16]; the connection between such models and chiral effective Lagrangians is reviewed in Ref. [17].

A separate but related question concerns how these properties change as the number of active light quark flavors NfN_f varies. QCD with NfN_f massless quarks in the fundamental representation of SU(3)SU(3) loses asymptotic freedom above NfAF=16.5N_f^{\mathrm{AF}} = 16.5 [4]. Below this value, an infrared fixed point of the renormalization group emerges in a window Nfc≲Nf<NfAFN_f^c \lesssim N_f < N_f^{\mathrm{AF}}, commonly called the conformal window [18, 19, 20]. For Nf<NfcN_f < N_f^c, the infrared fixed point shifts to strong coupling, chiral symmetry breaks dynamically, and quarks confine. At Nf=NfcN_f = N_f^c, a chiral phase transition occurs. The precise value of NfcN_f^c is not yet settled: lattice simulations place it in the range 8≲Nfc≲128 \lesssim N_f^c \lesssim12 [20, 21, 22], while continuum SDE studies and effective models suggest Nfc≈8N_f^c \approx8 [23, 24, 25, 26]. A recent self-consistent SDE analysis including the gluon propagator finds Nfc=6.81N_f^c = 6.81 [27]. The dependence of quark dynamics on NfN_f has been investigated in the context of chiral symmetry breaking, confinement, and the finite-temperature QCD phase diagram using the CI model with a flavor-dependent effective coupling [25, 26, 28]. In this framework, the effective coupling carries an explicit factor of 1−(Nf−2)/Nc\sqrt{1 - (N_f - 2)/N_c}, which suppresses quark-gluon interactions as NfN_f grows and drives the system toward chiral symmetry restoration. Previous work within this model established the NfN_f- and NcN_c-dependence of the dynamical quark mass, the quark-antiquark condensate [25, 26, 29], and the leptonic decay constants of the pion and kaon [30]. What has not been addressed is how the electromagnetic structure of the pion and the kaon responds to the screening of the effective coupling as NfN_f increases.

This is the question we take up here. Working within the rainbow-ladder truncation of the SDE-BSE system and employing the symmetry-preserving Schwinger proper-time regularization scheme, we compute the electromagnetic form factors of the charged pion and kaon for varying NfN_f at fixed Nc=3N_c = 3. The calculation proceeds in the generalized impulse approximation following Refs. [8, 10, 12], and the photon couples to the quarks through the dressed quark-photon vertex, which in this truncation obeys its own inhomogeneous Bethe-Salpeter equation and carries the ρ\rho-meson pole. The kaon provides a complementary window, because its strange constituent responds to the screening differently from the light one.

The comparison with experiment is confined to the region where a contact interaction can be read at all.

The low-momentum spacelike pion form factor has been mapped by pion-electron elastic scattering, most precisely by the CERN NA7 experiment [31, 32], with earlier Fermilab measurements by Dally et al. [33]; the kaon charge radius was measured in the same way [34, 35]. We add one-loop chiral perturbation theory (χ\chiPT) [36, 37, 38] as a model-independent benchmark of the low-energy behavior. The momentum transfers reached at Jefferson Lab and targeted by the Electron-Ion Collider [39, 40, 41, 42] lie outside that region, and we make no statement about them.

Two results anticipate the rest of the paper. The model reproduces the established static properties at Nf=2N_f = 2 but undershoots the measured charge radii by about one third; this is a known feature of the contact interaction [12, 13] [43], and it means that the absolute radii are not predictions to be taken at face value. What the model does control is the response to NfN_f. We find that both radii grow monotonically as the coupling is screened, that the growth is robust under every variation of the two least constrained ingredients of the model, and that it is driven by the quark core of the bound state rather than by the vector-meson dressing of the photon vertex.

The rest of this paper is organized as follows. § 2 presents the flavor-dressed contact interaction model, the gap equation, the physical basis of the NfN_f-dependent confinement scale and the sensitivity of the results to the model assumptions. § 3 develops the Bethe-Salpeter formalism, the static observables, the form factors and the dressed quark-photon vertex. § 4 presents the numerical results, and § 5 offers the conclusions. Appendix A collects the expressions for the triangle diagram with unequal masses.

The Flavor-Dressed Contact Interaction Model

Schwinger-Dyson Equation and the Dressed Quark Propagator

The starting point is the SDE for the dressed-quark propagator [44, 45]. In Euclidean space and Landau gauge, the full propagator Sf(p)S_f(p) of a quark with flavor ff can be written as

Sf−1(p)=Sf,0−1(p)+Σf(p),(1)S_f^{-1}(p) = S_{f,0}^{-1}(p) + \Sigma_f(p), \tag*{(1)}

where Sf,0(p)=(ip̸+mf)−1S_{f,0}(p) = (i\not{p} + m_f)^{-1} is the bare propagator, mfm_f is the current quark mass, and the self-energy reads

Σf(p)=∫d4q(2π)4g2Dμν(p−q)λa2γμSf(q)λa2Γν(p,q).(2)\Sigma_f(p) = \int\frac{d^4q}{(2\pi)^4} g^2 D_{\mu\nu}(p-q) \frac{\lambda^a}{2}\gamma_\mu S_f(q) \frac{\lambda^a}{2}\Gamma_\nu(p,q). \tag*{(2)}

Here g2Dμνg^2D_{\mu\nu} is the dressed gluon propagator, Γν\Gamma_\nu the dressed quark-gluon vertex, and λa\lambda^a the Gell-Mann matraces, which satisfy

∑a=1Nc2−1λa2λa2=12(Nc−1Nc)I≡CFI,(3)\sum_{a=1}^{N_c^2-1} \frac{\lambda^a}{2}\frac{\lambda^a}{2} = \frac{1}{2}\left(N_c-\frac{1}{N_c}\right) I \equiv C_F I , \tag*{(3)}

with II the identity matrix in color space.

The flavor-dependent contact interaction

The flavor-dressed contact interaction (FCI) model replaces the momentum-dependent gluon exchange with a momentum-independent effective coupling whose strength depends explicitly on NfN_f [25, 26]. In the Landau gauge and within the infrared region where gluons acquire a dynamically generated mass mgm_g [10, 46, 47], the model is defined by

g2Dμν(p−q)=δμναeff(Nf),(4)g^2D_{\mu\nu}(p-q)=\delta_{\mu\nu}\alpha_{\mathrm{eff}}(N_f), \tag*{(4)}

where

αeff(Nf)=4παirmg2[1−Nf−2Nfc]p,p=12.(5)\alpha_{\mathrm{eff}}(N_f)=\frac{4\pi\alpha_{\mathrm{ir}}}{m_g^2}\left[1-\frac{N_f-2}{\mathcal{N}_f^c}\right]^p,\qquad p=\frac{1}{2}. \tag*{(5)}

Here Nfc\mathcal{N}_f^c is not the critical flavor number itself but a shifted version of it, Nfc=Nfc+η\mathcal{N}_f^c=N_f^c+\eta. The parameter αir=0.93π\alpha_{\mathrm{ir}}=0.93\pi sets the infrared-enhanced interaction strength, and mg=0.8 GeVm_g=0.8\ \mathrm{GeV} is the gluon mass scale [48]. The NfN_f-dependent factor encodes the progressive weakening of the effective coupling as the number of light quark flavors increases. The shift η\eta, in the range 1.8–2.3, is adjusted to place the onset of chiral symmetry restoration at Nfc≈8N_f^c\approx8, consistent with lattice and continuum QCD studies [21, 23, 24]. The factor (Nf−2)(N_f-2) in Eq. (5) anchors the model at Nf=2N_f=2, where the standard CI model is recovered; the physical motivation for this parametrization is discussed in Ref. [25]. The exponent pp is written explicitly because Sec. II E varies it.

When this model is inserted into the SDE, the quark wave function renormalization reduces to unity and the dressed propagator takes the simple form Sf(p)=(ip̸+Mf)−1S_f(p)=(i\not{p}+M_f)^{-1}, with MfM_f a momentum-independent dynamical mass. This momentum independence is the defining feature of the CI class of models and the source of their analytical tractability [10, 11, 49]. All the color and coupling factors enter every equation below through a single combination,

G(Nf)≡CFαeff(Nf)4π2=4αir3πmg2[1−Nf−2Nfc]p,(6)\mathcal{G}(N_f)\equiv\frac{C_F\alpha_{\mathrm{eff}}(N_f)}{4\pi^2}=\frac{4\alpha_{\mathrm{ir}}}{3\pi m_g^2}\left[1-\frac{N_f-2}{\mathcal{N}_f^c}\right]^p, \tag*{(6)}

where the second form holds for Nc=3N_c=3. At Nf=2N_f=2 one has G=1.9375 GeV−2\mathcal{G}=1.9375\ \mathrm{GeV}^{-2}.

Gap equation and confinement

Inserting the FCI model into Eq. (1) and performing the Dirac traces in the rainbow-ladder truncation, where the dressed vertex is replaced by its bare counterpart Γν=γν\Gamma_\nu=\gamma_\nu, the dynamical quark mass satisfies

Mf=mf+MfG(Nf)∫0∞ds ss+Mf2.(7)M_f=m_f+M_f\mathcal{G}(N_f)\int_0^\infty ds\,\frac{s}{s+M_f^2}. \tag*{(7)}

The integral is ultraviolet divergent and must be regulated. The Schwinger proper-time scheme [44] introduces two cutoff scales, an ultraviolet regulator τuv=Λuv−1\tau_{\mathrm{uv}}=\Lambda_{\mathrm{uv}}^{-1} and an infrared regulator τir=Λir−1\tau_{\mathrm{ir}}=\Lambda_{\mathrm{ir}}^{-1},

1s+Mf2→∫τuv2τir2dτ e−τ(s+Mf2)=e−τuv2(s+Mf2)−e−τir2(s+Mf2)s+Mf2.(8)\frac{1}{s+M_f^2}\rightarrow\int_{\tau_{\mathrm{uv}}^2}^{\tau_{\mathrm{ir}}^2}d\tau\,e^{-\tau(s+M_f^2)} =\frac{e^{-\tau_{\mathrm{uv}}^2(s+M_f^2)}-e^{-\tau_{\mathrm{ir}}^2(s+M_f^2)}}{s+M_f^2}. \tag*{(8)}

The role of each regulator is distinct. The ultraviolet cutoff τuv\tau_{\mathrm{uv}} removes quadratic and logarithmic divergences and sets the scale for all dimensioned quantities in the model; it cannot be eliminated since the CI model is non-renormalizable. The infrared cutoff τir\tau_{\mathrm{ir}}, on the other hand, implements quark confinement: its nonzero value removes the real-axis quark production threshold from the propagator, so that the quark never reaches its mass shell and therefore cannot appear in the physical spectrum [12, 49, 50, 51], (see also a recent comparative study on confinement order parameters in [52]).

After the integration over ss, the gap equation reduces to

Mf=mf+MfG(Nf)C(Mf2),(9)M_f=m_f+M_f\mathcal{G}(N_f)\mathcal{C}(M_f^2), \tag*{(9)}

where, suppressing the dependence on the regulators,

C(z)=z[Γ(−1,zτuv2)−Γ(−1,zτir2)],\mathcal{C}(z)=z\left[\Gamma(-1,z\tau_{\mathrm{uv}}^2)-\Gamma(-1,z\tau_{\mathrm{ir}}^2)\right],
C‾1(z)=−dC(z)dz=Γ(0,zτuv2)−Γ(0,zτir2),(10)\overline{\mathcal{C}}_1(z)=-\frac{d\mathcal{C}(z)}{dz}=\Gamma(0,z\tau_{\mathrm{uv}}^2)-\Gamma(0,z\tau_{\mathrm{ir}}^2), \tag*{(10)}
C‾2(z)=−12dC‾1(z)dz=e−zτuv2−e−zτir22z,\overline{\mathcal{C}}_2(z)=-\frac{1}{2}\frac{d\overline{\mathcal{C}}_1(z)}{dz}=\frac{e^{-z\tau_{\mathrm{uv}}^2}-e^{-z\tau_{\mathrm{ir}}^2}}{2z},

and C1(z)=zC‾1(z)\mathcal{C}_1(z)=z\overline{\mathcal{C}}_1(z), with Γ(a,y)=∫y∞ta−1e−t dt\Gamma(a,y)=\int_y^\infty t^{a-1}e^{-t}\,dt the upper incomplete Gamma function. Every regularized integral in this paper reduces to one of these three functions: C\mathcal{C} for one propagator, C‾1\overline{\mathcal{C}}_1 for two and C‾2\overline{\mathcal{C}}_2 for three, each evaluated at the appropriate Feynman-parametrized argument.

When NfN_f differs from the reference value Nf=2N_f=2, the infrared cutoff is adjusted to maintain the connection between DCSB and confinement through the rescaled confinement scale

τ~ir=τirMu(Nf=2)Mu(Nf),(11)\widetilde{\tau}_{\mathrm{ir}}=\tau_{\mathrm{ir}}\frac{M_u(N_f=2)}{M_u(N_f)}, \tag*{(11)}

For each NfN_f we iterate the light-quark gap equation and Eq. (11) until τ~ir\widetilde{\tau}_{\mathrm{ir}} and Mu(Nf)M_u(N_f) are mutually consistent, and use the resulting τ~ir\widetilde{\tau}_{\mathrm{ir}} for both flavors. The light quark sets the scale because it carries the chiral order parameter; the strange mass contains a large explicit piece and would tie the confinement scale to the current mass rather than to DCSB. The converged infrared scale Λ~ir=1/τ~ir\widetilde{\Lambda}_{\rm ir}=1/\widetilde{\tau}_{\rm ir} is 0.240, 0.204, 0.152 and 0.071 GeV at Nf=2N_f=2, 4, 6 and 8. The iteration converges quickly, and a single update starting from τir\tau_{\rm ir} already reproduces the converged masses to better than 10−310^{-3} GeV. The light-quark condensate follows directly from the gap equation. The regularized trace of the propagator gives

−⟨qˉq⟩f=Nc4π2MfC(Mf2)=Nc4π2Mf−mfG(Nf),(12)-\langle\bar{q}q\rangle_f=\frac{N_c}{4\pi^2}M_f\mathcal{C}(M_f^2)=\frac{N_c}{4\pi^2}\frac{M_f-m_f}{G(N_f)}, \tag*{(12)}

which is the chiral condensate in the chiral limit. The structure encoded in Eqs. (9)–(12) has a transparent physical content. As NfN_f grows, G(Nf)G(N_f) falls, the dynamical part of the mass shrinks, and in the chiral limit the gap equation eventually admits only the trivial solution. Simultaneously, Eq. (11) drives τ~ir\widetilde{\tau}_{\rm ir} upward, the infrared cutoff Λ~ir\widetilde{\Lambda}_{\rm ir} falls from 0.240 GeV at Nf=2N_f=2 to 0.071 GeV at Nf=8N_f=8, and the region of the propagator that confinement removes shrinks with it.

Why the confinement scale follows the chiral order parameter along NfN_f

Confinement and DCSB are distinct phenomena. At zero temperature and density they are governed by different scales, and there are settings in which they separate: at large baryon chemical potential a confined but chirally symmetric phase has been proposed [53], and in two dimensions confinement is linear and unconditional. Eq. (11) is therefore a modeling choice and needs a justification that does not rest on finite-temperature thermodynamics, where the two transitions happen to coincide in the crossover region.

The justification we offer is specific to the direction in theory space that this paper explores. The variable is NfN_f at T=0T=0 and μ=0\mu=0, and the endpoint is the lower edge of the conformal window [18, 19, 20]. Inside the window the infrared dynamics is controlled by a fixed point of the renormalization group, so the theory is scale invariant at long distances. A scale-invariant theory has neither a string tension nor a dynamically generated mass. Entering the conformal window from below therefore removes confinement and DCSB together, and it does so for a reason that is not thermal: both require a scale, and the fixed point removes it. Equation (11) is the simplest prescription that builds this endpoint into a model with a single dynamical scale. It does not claim that the two phenomena are the same, nor that they vanish together anywhere else in the phase diagram.

The prescription can still bias the results, and the test is to remove it. Section IIE repeats the calculation with τir\tau_{\rm ir} held at its Nf=2N_f=2 value, which decouples confinement from the chiral order parameter completely. The answer is quantitative and is given in Table I: the pion radius at Nf=6N_f=6 changes from 0.594 to 0.532 fm, about 10%, and

VariantNfN_fMuM_uMsM_sfπf_\pifKf_Krπr_\pirKr_K
baseline40.3120.5000.0970.1050.4950.459
60.2330.4580.0870.1010.5940.545
p=0.25p=0.2540.3400.5160.1000.1060.4690.436
60.3030.4950.0970.1050.5030.466
p=1p=140.2510.4670.0900.1020.5660.520
60.0790.390————
η=1.8\eta=1.840.3100.4990.0970.1050.4960.460
60.2290.4560.0870.1000.6000.551
η=2.3\eta=2.340.3140.5010.0980.1050.4930.457
60.2390.4610.0880.1010.5850.537
fixed τir\tau_{\rm ir}40.3090.4990.0960.1050.4810.446
60.2240.4580.0810.0990.5320.487

Table 1. Sensitivity of the main observables to the model assumptions, at Nf=4N_f=4 and 6. Masses and decay constants in GeV, radii (dressed at vertex) in fm. At Nf=2N_f=2 all variants coincide with the first row of Table IV. For p=1p=1 at Nf=6N_f=6 the pion is no longer bound.

neither the monotonic growth with NfN_f nor the ordering rK<rπr_K<r_\pi is affected.

Sensitivity to the screening function and to the rescaled confinement scale

Equations (5) and (11) are the two least constrained ingredients of the model. Neither is derived from QCD. The square-root screening factor is the simplest function that is unity at Nf=2N_f=2, decreases monotonically, and vanishes at Nf=Nfc+2N_f=N_f^c+2; the rescaling of τir\tau_{\rm ir} was discussed above. We recalculate the entire calculation with three families of alternatives: the exponent p∈{0.25,0.5,1}p\in\{0.25,0.5,1\} in Eq. (5), which keeps the endpoint fixed while changing how fast it is approached; the shift η\eta at the two ends of its interval, 1.8 and 2.3; and a fixed τir\tau_{\rm ir}. The results are collected in Table I.

Three conclusions follow from the table. The ordering in NfN_f is monotonic in every variant, for every observable, and rK<rπr_K<r_\pi holds throughout. The rate of growth is not universal: between Nf=2N_f=2 and 6 the pion radius grows by 13% for p=0.25p=0.25 and by 33% for p=0.5p=0.5, while for p=1p=1 the pion unbinds before Nf=6N_f=6. This is expected, since pp controls how fast the transition is approached, and it means that the numbers quoted at a given NfN_f inherit the uncertainty of the screening function. Changes in η\eta move the radii by less than 2%, and fixing τir\tau_{\rm ir} reduces the growth without reversing it. No qualitative statement of Sec. IV depends on the baseline choices.

Model parameters

The model parameters are fixed at the reference point Nf=2N_f=2, Nc=3N_c=3 by reproducing the properties of the ground-state π\pi and ρ\rho mesons following Ref. [12], and the kaon following Ref. [54]. The values are collected in Table II.

mu=dm_{u=d}msm_sΛir\Lambda_{\rm ir}Λuv\Lambda_{\rm uv}αir\alpha_{\rm ir}mgm_gNfcN_f^c
0.0070.1700.2400.9050.93π0.93\pi0.8009.98

Table II. Parameters of the FCI model. All dimensioned quantities are in GeV. The value of NfcN_f^c corresponds to η=1.98\eta= 1.98 in Nfc=Nfc+ηN_f^c = N_f^c+\eta with Nfc=8N_f^c=8.

mπm_\pimKm_Kfπf_\pifKf_Kκπ1/3\kappa_\pi^{1/3}κK1/3\kappa_K^{1/3}
FCI0.1400.4990.1010.1060.2430.252
Exp.0.1400.4940.0920.110––

Table III. Static observables at Nf=2N_f=2, Nc=3N_c=3, in GeV. The decay constants are given in the convention fπexp≃0.092f_\pi^{\rm exp}\simeq0.092 GeV. The experimental values are those of the Particle Data Group [55].

At these parameter values, the solution of the gap and Bethe-Salpeter equations yields the dynamical masses and meson observables listed in Table III. They coincide with the published contact interaction values obtained with the same parameters [12, 13, 54], which we use as a check of our numerical implementation, and they agree with experiment at the level expected of the truncation.

Bethe-Salpeter equation and electromagnetic form factors

Pseudoscalar meson bound states

The pion emerges from the SDE-BSE coupled system as a simultaneous solution of two equations: the gap equation already presented in Sec. II, and the homogeneous BSE for the quark-antiquark bound state. The BSE determines the meson Bethe-Salpeter amplitude (BSA) ΓH(k;P)\Gamma_H(k;P) through

[ΓH(k;P)]tu=∫d4q(2π)4Ktu;rs(k,q;P)[Sf(q+)ΓH(q;P)Sg(q−)]sr(13)[\Gamma_H(k;P)]_{tu}=\int\frac{d^4q}{(2\pi)^4}K_{tu;rs}(k,q;P)[S_f(q_+)\Gamma_H(q;P)S_g(q_-)]_{sr} \tag*{(13)}

where q+=q+ϑPq_+=q+\vartheta P and q−=q−(1−ϑ)Pq_-=q-(1-\vartheta)P, with ϑ∈[0,1]\vartheta\in[0,1] a momentum-sharing parameter that leaves physical observables unchanged [6]. The indices rr, ss, tt, uu collect color, flavor, and spinor degrees of freedom. The symbol H=fgˉH=f\bar{g} denotes the flavor content of the meson and KK is the quark-antiquark scattering kernel.

The axWTI imposes a tight constraint on the BSE kernel [56]. In the chiral limit it reads

−PμΓ5μ(k;P)=S−1(k+)γ5+γ5S−1(k−)(14)-P_\mu\Gamma_{5\mu}(k;P)=S^{-1}(k_+)\gamma_5+\gamma_5S^{-1}(k_-) \tag*{(14)}

with k±=k±P/2k_\pm=k\pm P/2. Any truncation scheme that preserves this identity guarantees that the pion is a massless Nambu-Goldstone boson in the chiral limit [7, 56].

In the rainbow-ladder truncation combined with the FCI model, the scattering kernel is

K(p,q;P)=−g2Dμν(p−q)[λa2γμ]⊗[λa2γν](15)K(p,q;P)=-g^2D_{\mu\nu}(p-q)\left[\frac{\lambda^a}{2}\gamma_\mu\right]\otimes\left[\frac{\lambda^a}{2}\gamma_\nu\right] \tag*{(15)}

and the homogeneous BSE for the ground-state pseudoscalar meson becomes

ΓH(P)=−CFαeff(Nf)∫d4q(2π)4γμSf(q+P)ΓH(P)Sg(q)γμ(16)\Gamma_H(P)=-C_F\alpha_{\rm eff}(N_f)\int\frac{d^4q}{(2\pi)^4}\gamma_\mu S_f(q+P)\Gamma_H(P)S_g(q)\gamma_\mu \tag*{(16)}

Within the FCI model the BSA cannot depend on the relative momentum between the constituents, and the most general decomposition consistent with Poincaré covariance and the quantum numbers of the meson is [10, 12, 54]

ΓH(P)=iγ5EH(P)+12MRγ5γ⋅PFH(P)(17)\Gamma_H(P)=i\gamma_5E_H(P)+\frac{1}{2M_R}\gamma_5\gamma\cdot P F_H(P) \tag*{(17)}

where MR=MfMg/(Mf+Mg)M_R=M_fM_g/(M_f+M_g) is the reduced mass of the constituents, so that 2MR=M2M_R=M for equal masses. The pseudovector component FHF_H is not a correction but an intrinsic feature of any Poincaré-covariant treatment: it materially affects the form factor and cannot be discarded [10, 56]. Substituting Eq. (17) into Eq. (16), introducing a Feynman parameter α\alpha and regularizing as in Sec. II, the BSE reduces to the 2×22\times2 eigenvalue problem

(EHFH)=G(Nf)(KEEHKEFHKFEHKFFH)(EHFH)(18)\begin{pmatrix}E_H\\F_H\end{pmatrix} =\mathcal{G}(N_f) \begin{pmatrix} \mathcal{K}_{EE}^H & \mathcal{K}_{EF}^H\\ \mathcal{K}_{FE}^H & \mathcal{K}_{FF}^H \end{pmatrix} \begin{pmatrix}E_H\\F_H \tag*{(18)}\end{pmatrix}

which has a nontrivial solution at P2=−mH2P^2=-m_H^2. With ω=ω(α,P2)=αMf2+(1−α)Mg2+α(1−α)P2\omega=\omega(\alpha,P^2)=\alpha M_f^2+(1-\alpha)M_g^2+\alpha(1-\alpha)P^2, the kernels are

KEEH=∫01dα{C(ω)+[MfMg−α(1−α)P2−ω]C‾1(ω)}\mathcal{K}_{EE}^H=\int_0^1d\alpha\left\{\mathcal{C}(\omega)+[M_fM_g-\alpha(1-\alpha)P^2-\omega]\overline{\mathcal{C}}_1(\omega)\right\}
KEFH=P22MfMg∫01dα[αMf2+MfMg+(1−α)Mg2]C‾1(ω)\mathcal{K}_{EF}^H=\frac{P^2}{2M_fM_g}\int_0^1d\alpha\left[\alpha M_f^2+M_fM_g+(1-\alpha)M_g^2\right]\overline{\mathcal{C}}_1(\omega)
KFEH=MR∫01dα[αMf+(1−α)Mg]C‾1(ω)\mathcal{K}_{FE}^H=M_R\int_0^1d\alpha\left[\alpha M_f+(1-\alpha)M_g\right]\overline{\mathcal{C}}_1(\omega)
KFFH=−(Mf+Mg)22MfMgKFEH(19)\mathcal{K}_{FF}^H=-\frac{(M_f+M_g)^2}{2M_fM_g}\mathcal{K}_{FE}^H \tag*{(19)}

For Mf=MgM_f=M_g these reduce to the kernels of Ref. [12]. The expressions printed in the previous version of this manuscript for KFE\mathcal{K}_{FE} and KFF\mathcal{K}_{FF} were dimensionally inconsistent; the ones above were derived anew and checked numerically against the published values.

The last line of Eq. (19) is where the axWTI enters the regularization. Preserving Eq. (14) after regularization requires [10, 12]

0=∫01dα[C(ω)+C1(ω)],(20)0 = \int_0^1 d\alpha\left[C(\omega) + C_1(\omega)\right], \tag*{(20)}

which the proper-time integrals do not satisfy identically. Following Refs. [10, 12, 54], it is imposed as a constraint on the regularized integrals; applied to the FFFF channel it produces the relation between KFF\mathcal{K}_{FF} and KFE\mathcal{K}_{FE} quoted above. With this constraint the pion is massless in the chiral limit, because at P2=0P^2 = 0 the eigenvalue condition coincides with the chiral-limit gap equation.

Static observables

Once the amplitudes EHE_H and FHF_H are determined, the BSA must be normalized. The canonical normalization condition in the rainbow-ladder truncation is [6, 12]

Pμ=Nctr⁡D∫d4q(2π)4ΓH(−P)∂Sf(q+P)∂PμΓH(P)Sg(q),(21)P_\mu= N_c \operatorname{tr}_{D} \int\frac{d^4q}{(2\pi)^4} \Gamma_H(-P) \frac{\partial S_f(q+P)}{\partial P_\mu} \Gamma_H(P) S_g(q), \tag*{(21)}

evaluated as the derivative of the regularized two-point function; in the chiral limit it reduces to 1=(Nc/4π2)Cˉ1(M2)Eπ(Eπ−2Fπ)1 = (N_c/4\pi^2)\bar{C}_1(M^2) E_\pi(E_\pi- 2F_\pi), the result of Ref. [12]. The leptonic decay constant and the in-meson condensate are then

fH=Nc4π21MR[EHKFEH+FHKFFH],(22)f_H = \frac{N_c}{4\pi^2}\frac{1}{M_R}\left[E_H \mathcal{K}_{FE}^H + F_H \mathcal{K}_{FF}^H\right], \tag*{(22)}
κH=fHρH,ρH=Nc4π2[EHKEEH+FHKEFH],(23)\kappa_H = f_H \rho_H,\qquad\rho_H = \frac{N_c}{4\pi^2}\left[E_H \mathcal{K}_{EE}^H + F_H \mathcal{K}_{EF}^H\right], \tag*{(23)}

with all kernels at P2=−mH2P^2 = -m_H^2 [6, 57]. The decay constant and the in-meson condensate are tied by the Gell-Mann–Oakes–Renner-type identity

fHmH2=(mf+mg)ρH,(24)f_H m_H^2 = (m_f + m_g)\rho_H, \tag*{(24)}

which a symmetry-preserving truncation must satisfy for any current masses [56]. Since both sides depend on NfN_f through different combinations of the same quantities, it is a nontrivial check of the numerics along the whole flavor trajectory; we return to it in § 4 F.

Electromagnetic form factors

The elastic form factor of the meson H=fgˉH = f\bar{g} is defined through the matrix element of the electromagnetic current between one-meson states. In the generalized impulse approximation it is the sum of the two diagrams in which the photon couples to the quark or to the antiquark [8, 10, 45],

FHem(Q2)=efFHf(Q2)+egFHgˉ(Q2),(25)F_H^{\mathrm{em}}(Q^2) = e_f F_H^f(Q^2) + e_g F_H^{\bar{g}}(Q^2), \tag*{(25)}

with eu=2/3e_u = 2/3, edˉ=esˉ=1/3e_{\bar{d}} = e_{\bar{s}} = 1/3, and

2KμFHf(Q2)=2Nc∫d4t(2π)4tr⁡D[iΓH(−p2)Sf(t+p2)×iΓμγ(Q)Sf(t+p1)iΓH(p1)Sg(t)],(26)2K_\mu F_H^f(Q^2) = 2N_c \int\frac{d^4t}{(2\pi)^4} \operatorname{tr}_{D}\left[i\Gamma_H(-p_2)S_f(t+p_2) \times i\Gamma_\mu^\gamma(Q)S_f(t+p_1)i\Gamma_H(p_1)S_g(t)\right], \tag*{(26)}

where p1=K−Q/2p_1 = K - Q/2 and p2=K+Q/2p_2 = K + Q/2 are the momenta of the initial and final meson, QQ is the spacelike photon momentum, KK is the average meson momentum, and Γμγ\Gamma_\mu^\gamma is the quark-photon vertex discussed below. The antiquark contribution FHgˉF_H^{\bar{g}} follows from Eq. (26) with the roles of ff and gg exchanged.

With the bare vertex Γμγ=γμ\Gamma_\mu^\gamma= \gamma_\mu, the trace is reduced in two steps. Numerator terms proportional to the spectator denominator t2+Mg2t^2 + M_g^2 are cancelled first, leaving a two-point integral over the struck-quark propagators; the remainder is a genuine three-point integral. Both are then Feynman-parametrized and regularized with the functions of Eq. (10). The result has the structure

FHf(Q2)∣bare=EH2TEEf+EHFHTEFf+FH2TFFf,(27)\left.F_H^f(Q^2)\right|_{\mathrm{bare}} = E_H^2 T_{EE}^f + E_HF_H T_{EF}^f + F_H^2 T_{FF}^f, \tag*{(27)}

with the functions TXYf(Q2)T_{XY}^f(Q^2) given in Appendix A. For equal masses they reproduce the expressions of Refs. [10, 12], and for unequal masses the EEEE term coincides with that of Ref. [13]. In the pion the charge computed with the canonical normalization of Eq. (21) is Fπem(0)=1.001F_\pi^{\mathrm{em}}(0) = 1.001, which measures the consistency of the regularization between the normalization and the triangle. For the kaon the canonical normalization computed from the light leg and from the strange leg differ by 4%, a known consequence of regularizing loops with unequal masses; we therefore normalize each flavor contribution so that FHf(0)=FHgˉ(0)=1F_H^f(0) = F_H^{\bar{g}}(0) = 1, which enforces charge conservation exactly.

The dressed quark-photon vertex

In the rainbow-ladder truncation the quark-photon vertex is not an independent object. It obeys an inhomogeneous Bethe-Salpeter equation with the same kernel as Eq. (15), and for a contact interaction that equation can be solved in closed form [12]. The solution is

Γμγ(Q)=γμTPT(Q2)+γμL,γμT=γμ−γ⋅Q QμQ2,(28)\Gamma_\mu^\gamma(Q) = \gamma_\mu^T P_T(Q^2) + \gamma_\mu^L,\qquad\gamma_\mu^T = \gamma_\mu- \frac{\gamma\cdot Q\,Q_\mu}{Q^2}, \tag*{(28)}

with γμL=γμ−γμT\gamma_\mu^L = \gamma_\mu- \gamma_\mu^T undressed, as the vector Ward-Takahashi identity requires, and

PTf(Q2)=11+Kγf(Q2),P_T^f(Q^2) = \frac{1}{1 + K_\gamma^f(Q^2)},
Kγf(Q2)=G(Nf)∫01dα α(1−α)Q2×Cˉ1(Mf2+α(1−α)Q2).(29)K_\gamma^f(Q^2) = \mathcal{G}(N_f)\int_0^1 d\alpha\,\alpha(1-\alpha)Q^2 \times\bar{C}_1\left(M_f^2 + \alpha(1-\alpha)Q^2\right). \tag*{(29)}

Because K⋅Q=0K \cdot Q = 0 for elastic scattering, only the transverse part contributes to Eq. (26) after contraction with KμK_\mu, and the dressed form factor is simply

FHf(Q2)=PTf(Q2)FHf(Q2)∣bare.(30)F_H^f(Q^2)=P_T^f(Q^2)F_H^f(Q^2)\big|_{\mathrm{bare}}. \tag*{(30)}

The factor satisfies PTf(0)=1P_T^f(0)=1, so the charge is untouched, and PTf→1P_T^f\to1 at large Q2Q^2. In between it develops a pole at the timelike point Q2=−mρ2Q^2=-m_\rho^2 where 1+Kfγf=01+Kf_\gamma^f=0, which is the ρ\rho-meson bound-state equation of the model. The vertex is screened by the same coupling G(Nf)\mathcal{G}(N_f) as everything else, so the ρ\rho pole moves with NfN_f; we find mρ=0.929m_\rho=0.929 GeV at Nf=2N_f=2, in agreement with Ref. [12], decreasing to 0.846, 0.704 and 0.394 GeV at Nf=4N_f=4, 6 and 8. All form factors and radii below are computed with the dressed vertex unless stated otherwise.

Results and Discussion

Dynamical quark masses and static observables

We solve Eqs. (9) and (11) self-consistently for the dynamical masses MuM_u and MsM_s as a function of NfN_f, at fixed Nc=3N_c=3, and then the BSE for both mesons. The results are collected in Table IV.

NfN_fMuM_uMsM_smπm_\pimKm_Kfπf_\pifKf_Kκπ1/3\kappa_\pi^{1/3}κK1/3\kappa_K^{1/3}rπbr_\pi^brπdr_\pi^drKbr_K^brKdr_K^dmρm_\rho
20.3670.5330.1400.4990.1010.1060.2430.2520.3120.4470.2930.4170.929
40.3120.5000.1410.5000.0970.1050.2380.2500.3660.4950.3410.4590.846
60.2330.4580.1460.5040.0870.1010.2260.2440.4750.5940.4390.5450.704
80.1080.4030.1660.5050.0550.0800.1810.2100.9661.0520.9991.0650.394

Table IV. Dependence on NfN_f at Nc=3N_c = 3. Masses and decay constants in GeV, radii in fm. The labels b and d denote the bare and the dressed quark-photon vertex, and mρm_\rho is the pole of Eq. (29). The infrared cutoff follows Eq. (11), solved self-consistently.

As NfN_f increases from 2, the coupling G(Nf)\mathcal{G}(N_f) decreases monotonically according to Eq. (6), the strength of DCSB is reduced, and both dynamical masses fall. The light mass falls much faster in absolute and relative terms, from 0.367 to 0.108 GeV between Nf=2N_f=2 and 8, while the strange mass falls from 0.533 to 0.403 GeV. The reason is that MsM_s contains a sizable explicit piece: ms=0.17m_s=0.17 GeV is unaffected by the screening and acts as a floor below which MsM_s cannot fall. The decay constants and the in-meson condensates follow the same trend. The meson masses first remain nearly constant and then rise near the end of the trajectory, as the Goldstone mechanism weakens and the system approaches the symmetric phase, in a pattern analogous to thermal screening [26].

Pion form factor at low momentum transfer

Figure 1 shows the pion form factor for Nf=2,4,6,8N_f=2,4,6,8 in the range 0≤Q2≤0.60\leq Q^2\leq0.6 GeV2^2, together with the NA7 data [32] and the one-loop χ\chiPT prediction. For χ\chiPT we use the SU(3)SU(3) expression of Refs. [37, 38],

Fπ+(t)=1+2Hππ(t)+HKK(t),FK+(t)=1+Hππ(t)+2HKK(t),(31)\begin{aligned} F_{\pi^+}(t)&=1+2H_{\pi\pi}(t)+H_{KK}(t),\\ F_{K^+}(t)&=1+H_{\pi\pi}(t)+2H_{KK}(t), \tag*{(31)} \end{aligned}

with t=−Q2t=-Q^2 and

HPP(t)=(t−4MP2)JˉP(t)12F2+tF2[2L9r3−ln⁡(MP2/μ2)+1/3192π2],(32)H_{PP}(t)=\frac{(t-4M_P^2)\bar{J}_P(t)}{12F^2} +\frac{t}{F^2}\left[\frac{2L_9^r}{3}-\frac{\ln(M_P^2/\mu^2)+1/3}{192\pi^2}\right], \tag*{(32)}

where JˉP\bar{J}_P is the standard one-loop function [36], F=92.2F=92.2 MeV, μ=0.77\mu=0.77 GeV and L9r(μ)=(6.9±0.7)×10−3L_9^r(\mu)=(6.9\pm0.7)\times10^{-3} [37]. Equation (32) reproduces the one-loop radii

⟨rπ+2⟩=12L9rF2−2ln⁡(Mπ2/μ2)+ln⁡(MK2/μ2)+332π2F2.(33)\langle r_{\pi^+}^2\rangle=\frac{12L_9^r}{F^2}-\frac{2\ln(M_\pi^2/\mu^2)+\ln(M_K^2/\mu^2)+3}{32\pi^2F^2}. \tag*{(33)}

and the corresponding kaon expression, which give rπ=0.669r_\pi=0.669 fm and rK=0.641r_K=0.641 fm. The band reflects the uncertainty in L9rL_9^r and is drawn only up to Q2=0.2Q^2=0.2 GeV2^2, beyond which one-loop χ\chiPT is not reliable.

The figure makes the limits of the model visible, and we state them directly. At Nf=2N_f=2 the contact interaction gives rπ=0.447r_\pi=0.447 fm against the experimental 0.659±0.0040.659\pm0.004 fm [55], an undershoot of about one third, and the curve lies above the data wherever they exist; against the 22 NA7 points of Figure 1, with statistical errors only, the χ2\chi^2 per point is 52 with the dressed vertex and 105 with the bare one. One-loop χ\chiPT, whose only free constant L9rL_9^r was itself fixed from pion data, gives a χ2\chi^2 per point of 2.0 below Q2=0.1Q^2=0.1 GeV2^2 and falls below the data above it, where higher orders are needed. The undershoot is the hardness of the contact interaction form factor [12, 13] [43], and it already shows at the level of the slope at the origin. The model therefore does not predict the absolute pion radius, and we do not present it as doing so.

Figure 1. Pion electromagnetic form factor, with the dressed quark-photon vertex, for Nf=2,4,6,8N_f=2,4,6,8 at Nc=3N_c=3, compared with 22 of the 45 NA7 data points [32], in the intervals 0.015≤Q2≤0.0350.015\leq Q^2\leq0.035 GeV2^2 and 0.083≤Q2≤0.1440.083\leq Q^2\leq0.144 GeV2^2 (statistical errors, converted from the published ∣Fπ∣2|F_\pi|^2), and with one-loop χ\chiPT (band, L9r=(6.9±0.7)×10−3L_9^r=(6.9\pm0.7)\times10^{-3}), shown only where the latter applies.

What the figure does show reliably is the response to NfN_f. The form factor softens monotonically as the coupling is screened: at Q2=0.3Q^2=0.3 GeV2^2 it falls from 0.812 at Nf=2N_f=2 to 0.779, 0.716 and 0.510 at Nf=4,6N_f=4,6 and 8. The radius grows accordingly (Table IV). This softening has a simple reading in the model: as MuM_u decreases, the bound state becomes less compact, and the scale over which the photon resolves its charge distribution moves to smaller Q2Q^2. That the softening brings the Nf=6N_f = 6 curve closer to the data is a coincidence of no physical significance, since the physical world has Nf=2+1N_f = 2+1 light flavors.

One class of measurements is deliberately absent. The e+e−→π+π−e^+e^- \to\pi^+\pi^- and τ\tau-decay data [58] determine the pion form factor in the timelike region, where it is complex and dominated by the ρ\rho resonance. The quantity computed here is the spacelike form factor, a real function. A dispersion relation connects the two, but it requires the two-pion cut, which a contact interaction does not have, so we do not attempt the continuation.

Kaon form factor and the flavor asymmetry

Figure 2 shows the kaon form factor. The pattern is the same as for the pion, with two differences. The kaon is more compact at every Nf≤6N_f \le6, rK/rπ=0.93r_K/r_\pi= 0.93 at Nf=2N_f = 2, in line with experiment and with χ\chiPT. And its response to the screening is slower: between Nf=2N_f = 2 and 6, at Q2=0.3 GeV2Q^2 = 0.3\ \mathrm{GeV}^2, the pion form factor drops by 11.7% and the kaon form factor by 9.4%. The strange current mass, which keeps MsM_s from falling as fast as MuM_u, is responsible. The effect is real but moderate. In terms of radii the two mesons grow by 33% and 31% over the same interval, because the kaon still contains a light quark, whose dynamical mass is screened exactly as in the pion.

Figure 2. Kaon electromagnetic form factor, with the dressed quark-photon vertex, for Nf=2,4,6,8N_f = 2, 4, 6, 8 at Nc=3N_c = 3. The hatched band is the NA7 monopole fit with ⟨rK2⟩=0.34±0.05 fm2\langle r_K^2\rangle= 0.34 \pm0.05\ \mathrm{fm}^2 [35], drawn over the measured range; the orange band is one-loop χ\chiPT.

The ratio FKem/FπemF_K^{\rm em}/F_\pi^{\rm em} is larger than unity at every spacelike Q2Q^2 we consider, and it grows with NfN_f: at Q2=0.3 GeV2Q^2 = 0.3\ \mathrm{GeV}^2 it equals 1.025, 1.034, 1.052 and 1.075 for Nf=2N_f = 2, 4, 6 and 8. This corrects a statement of the previous version, which claimed that the ratio approaches unity as NfN_f grows; the calculation does not support it in the range where the model can be used. The radii, in contrast, do converge near the end of the trajectory, rK≃rπ≃1.06r_K \simeq r_\pi\simeq1.06 fm at Nf=8N_f = 8, where the pion is barely bound and both mesons are dominated by the soft light quark. We regard that point as the least reliable of the trajectory and draw no quantitative conclusion from it.

What drives the growth of the radius

Figure 3 compares the form factors computed with the bare and with the dressed quark-photon vertex at Nf=2N_f = 2 and 6, and Figure 4 shows both radii along the full trajectory. The dressing enlarges the radius substantially. At Nf=2N_f = 2 it raises rπr_\pi from 0.312 to 0.447 fm and rKr_K from 0.293 to 0.417 fm, in agreement with Ref. [12]. It does not change the sign of any trend, and it does not change the ordering rK<rπr_K < r_\pi.

Figure 3. Pion (left) and kaon (right) form factors with the bare (dashed) and the dressed (solid) quark-photon vertex, at Nf=2N_f = 2 and Nf=6N_f = 6. The NA7 pion data of Figure 1 are shown for reference.

Figure 4. Charge radii of the pion and kaon as functions of NfN_f, with the bare and with the dressed quark-photon vertex. The diamonds mark the experimental values [55].

The separation into the two pieces is the result we consider most informative. Since the dressing multiplies the bare form factor, the mean-square radius splits additively, ⟨r2⟩=⟨r2⟩bare+6dKγ/dQ2∣0\langle r^2\rangle= \langle r^2\rangle_{\rm bare} + 6dK_\gamma/dQ^2|_0. For the pion, between Nf=2N_f = 2 and 6, the bare contribution rises from 0.097 to 0.226 fm2^2, while the dressing contribution moves from 0.102 to 0.127 fm2^2. For the kaon the corresponding numbers are 0.086 →\to 0.193 and 0.088 →\to 0.105 fm2^2. The vector-meson pole drops by about a quarter over the same interval, so one might have expected the dressing to dominate the growth. It does not. The growth of the electromagnetic size under screening is carried by the quark core of the bound state, which responds directly to the falling dynamical mass, and only secondarily by the ρ\rho-like dressing of the photon.

Comparison of the radii with other approaches

Table V places the Nf=2N_f = 2 radii in context. Our values coincide with the earlier contact interaction calculations with the same parameters, and all of them fall short of experiment and of χ\chiPT by a similar fraction. The ratio rK/rπr_K/r_\pi is closer to experiment than either radius separately: the model gives 0.93, χ\chiPT 0.96, and the PDG averages 0.85±0.050.85 \pm0.05. The difference of mean-square radii, ⟨rπ2⟩−⟨rK2⟩=0.025 fm2\langle r_\pi^2\rangle- \langle r_K^2\rangle= 0.025\ {\rm fm}^2, is smaller than the NA7 value 0.10±0.045 fm20.10 \pm0.045\ {\rm fm}^2 [35].

rπr_\pirKr_K
This work, bare vertex0.3120.293
This work, dressed vertex0.4470.417
CI, Ref. [12]0.45–
CI, Ref. [13]0.450.42
One-loop χ\chiPT, Eq. (31)0.669 ± 0.0290.641 ± 0.030
NA7 [32, 35]0.663 ± 0.0060.58 ± 0.04
Dally et al. [34]–0.53 ± 0.05
PDG [55]0.659 ± 0.0040.560 ± 0.031

Table 5. Charge radii (fm) at Nf=2N_f = 2.

The Gell-Mann–Oakes–Renner identity along the flavor trajectory

Equation (24) has no reason to survive the screening automatically, because fHf_H, mHm_H and ρH\rho_H each evolve with NfN_f through different combinations of EHE_H, FHF_H and the kernels. We find that it holds to better than one part in 10410^4 for both mesons at every NfN_f in Table IV, including Nf=8N_f = 8, where the pion is barely bound. This confirms that the screened coupling, the rescaled infrared cutoff and the axWTI constraint of Eq. (20) are mutually consistent, and that the trends reported above are not artifacts of the numerical implementation. A discussion of the NfN_f dependence of the decay constants within this model is given in Ref. [30].

Conclusions

We have studied the electromagnetic form factors of the charged pion and kaon within a symmetry-preserving, flavor-dressed contact interaction model of quarks, in which the effective coupling depends on the number of active light flavors and is progressively screened as $N_f grows toward the chiral transition. The photon couples to the quarks through the dressed quark-photon vertex obtained in the same truncation, and the calculation has been confronted with the NA7 data and with one-loop chiral perturbation theory in the only momentum range, Q2≤0.6 GeV2Q^2 \le0.6\ \mathrm{GeV}^2, where a contact interaction can be read. At Nf=2N_f = 2 the model reproduces the established contact interaction values of the masses, decay constants, condensates and radii, and it inherits their main limitation: the charge radii, 0.447 fm for the pion and 0.417 fm for the kaon, fall short of experiment by about one third. We do not regard these absolute values as predictions.

The response to NfN_f is where the model has something to say. Both radii grow monotonically as the coupling is screened, from 0.447 to 0.594 fm for the pion and from 0.417 to 0.545 fm for the kaon between Nf=2N_f = 2 and 6. The growth survives every variation of the screening function and of the infrared confinement scale that we tested, including one in which confinement is decoupled from the chiral order parameter altogether, although its rate depends on the shape of the screening function. The kaon responds more slowly than the pion because the strange current mass keeps MsM_s from falling as fast as MuM_u, but the difference is moderate, since the kaon still carries a light quark. The Gell-Mann–Oakes–Renner identity holds along the whole trajectory.

The most informative outcome is the decomposition of the radius. The growth of the electromagnetic size under screening comes mainly from the quark core of the bound state, whose contribution to the pion mean-square radius more than doubles between Nf=2N_f = 2 and 6, and only secondarily from the vector-meson dressing of the photon, whose contribution grows by a quarter even though the ρ\rho pole drops by about the same fraction. Within this framework, then, the size of the lightest pseudoscalars tracks the strength of DCSB through the quark mass function itself. Testing whether this statement survives a momentum-dependent interaction, where the absolute radii are realistic, is the natural continuation of this work.

Acknowledgments

A. Raya acknowledges support from SECIHTI (Mexico) under grant CBF-2026-465 and from UMSNH-CIC under grant 18371. A. Ahmad thanks the Institute of Physics, Gomal University for support.

Triangle diagram with unequal masses

For a meson H=fgˉH = f\bar{g} with P2=−mH2P^2 = -m_H^2 and the photon attached to the quark ff, the bare-vertex form factor of Eq. (27) is, up to a normalization fixed by FHf(0)=1F_H^f(0) = 1,

TXYf(Q2)=bXY∫01dα Cˉ1(ωb)+2∫01du∫01dβ{aXY[Cˉ1(Δ)−ΔCˉ2(Δ)]+rXYCˉ2(Δ)},(34)T_{XY}^f(Q^2) = b_{XY}\int_0^1 d\alpha\,\bar{\mathcal{C}}_1(\omega_b) + 2\int_0^1 du\int_0^1 d\beta\left\{a_{XY}\left[\bar{\mathcal{C}}_1(\Delta)-\Delta\bar{\mathcal{C}}_2(\Delta)\right]+r_{XY}\bar{\mathcal{C}}_2(\Delta)\right\}, \tag*{(34)}

with ωb=Mf2+α(1−α)Q2\omega_b = M_f^2 + \alpha(1-\alpha)Q^2 and

Δ=uMf2+(1−u)Mg2−u(1−u)mH2+u2β(1−β)Q2.(35)\Delta= uM_f^2+(1-u)M_g^2-u(1-u)m_H^2+u^2\beta(1-\beta)Q^2. \tag*{(35)}

Writing S=Mf+MgS = M_f + M_g, D=Mf−MgD = M_f - M_g and k=S2/(4Mf2Mg2)k = S^2/(4M_f^2M_g^2), the coefficients are

bEE=1,aEE=0,b_{EE}=1,\qquad a_{EE}=0,
rEE=u(D2+mH2)+2MgD,r_{EE}=u\left(D^2+m_H^2\right)+2M_gD,
bEF=−SMf,aEF=−SD2MfMg,b_{EF}=-\frac{S}{M_f},\qquad a_{EF}=-\frac{SD}{2M_fM_g},
rEF=S2MfMg{Q2u(Du+Mg)−4DmH2u(1−u)−2Mg(SD+mH2)},r_{EF}=\frac{S}{2M_fM_g}\left\{Q^2u(Du+M_g)-4D m_H^2u(1-u)-2M_g\left(SD+m_H^2\right)\right\},
bFF=k(Q2+mH2),aFF=−k[u(Q2+3mH2)−mH2],b_{FF}=k\left(Q^2+m_H^2\right),\qquad a_{FF}=-k\left[u\left(Q^2+3m_H^2\right)-m_H^2\right],
rFF=k{Q4u3β3(1−β)+Q2[mH2u2(2u−1)−MgSu]r_{FF}=k\left\{Q^4u^3\beta^3(1-\beta)+Q^2\left[m_H^2u^2(2u-1)-M_gSu\right]\right.
+mH2[Mf2u−2MfMgu+2MfMg−3Mg2u+2Mg2]+mH4u(2u−1)2}.(36)\left.+m_H^2\left[M_f^2u-2M_fM_gu+2M_fM_g-3M_g^2u+2M_g^2\right]+m_H^4u(2u-1)^2\right\}. \tag*{(36)}

The antiquark contribution follows by exchanging MfM_f and MgM_g. For Mf=MgM_f=M_g these expressions reduce to those of Refs. [10, 12], and the EEEE coefficient coincides with that of Ref. [13] for unequal masses. We have verified numerically that Eqs. (34)–(36) reproduce the triangle computed directly from the Dirac trace to machine precision.

References

  1. [1]C. D. Roberts, D. G. Richards, T. Horn, and L. Chang, Insights into the emergence of mass from studies of pion and kaon structure, Prog. Part. Nucl. Phys. 120, 103883 (2021), arXiv:2102.01579.
  2. [2]T. Horn and C. D. Roberts, The pion: an enigma within the standard model, J. Phys. G 43, 073001 (2016), arXiv:1602.04016.DOI
  3. [3]D. J. Gross and F. Wilczek, Ultraviolet Behavior of Nonabelian Gauge Theories, Phys. Rev. Lett. 30, 1343 (1973).DOI
  4. [4]H. D. Politzer, Reliable Perturbative Results for Strong Interactions?, Phys. Rev. Lett. 30, 1346 (1973).DOI
  5. [5]K. G. Wilson, Confinement of Quarks, Phys. Rev. D 10, 2445 (1974).DOI
  6. [6]P. Maris and C. D. Roberts, Pi- and K meson Bethe-Salpeter amplitudes, Phys. Rev. C 56, 3369 (1997), arXiv:nucl-th/9708029.DOI
  7. [7]H. J. Munczek, Dynamical chiral symmetry breaking, goldstone’s theorem and the consistency of the schwinger-dyson and bethe-salpeter equations, Phys. Rev. D 52, 4736 (1995).
  8. [8]P. Maris and P. C. Tandy, The π, k+, and k0 electromagnetic form factors, Phys. Rev. C 62, 055204 (2000), arXiv:nucl-th/00050015.DOI
  9. [9]P. Maris and P. C. Tandy, The quark photon vertex and the pion charge radius, Phys. Rev. C 61, 045202 (2000), arXiv:nucl-th/9905056.
  10. [10]L. X. Gutierrez-Guerrero, A. Bashir, I. C. Cloet, and C. D. Roberts, Pion form factor from a contact interaction, Phys. Rev. C81, 065202 (2010), arXiv:1002.1968 [nucl-th].
  11. [11]H. L. L. Roberts, C. D. Roberts, A. Bashir, L. X. Gutierrez-Guerrero, and P. C. Tandy, Abelian anomaly and neutral pion production, Phys. Rev. C82, 065202 (2010), arXiv:1009.0067 [nucl-th].DOI
  12. [12]H. L. L. Roberts, A. Bashir, L. X. Gutierrez-Guerrero, C. D. Roberts, and D. J. Wilson, pi- and rho-mesons, and their diquark partners, from a contact interaction, Phys. Rev. C83, 065206 (2011), arXiv:1102.4376 [nucl-th].DOI
  13. [13]R. J. Hernández-Pinto, L. X. Gutiérrez-Guerrero, A. Bashir, M. A. Bedolla, and I. M. Higuera-Angulo, Electromagnetic form factors and charge radii of pseudoscalar and scalar mesons: A comprehensive contact interaction analysis, Phys. Rev. D 107, 054002 (2023), arXiv:2301.11881 [hep-ph].
  14. [14]Y. Nambu and G. Jona-Lasinio, Dynamical Model of Elementary Particles Based on an Analogy with Superconductivity. I, Phys. Rev. 122, 345 (1961).
  15. [15]Y. Nambu and G. Jona-Lasinio, Dynamical Model of Elementary Particles Based on an Analogy with Superconductivity. II, Phys. Rev. 124, 246 (1961).
  16. [16]V. Bernard, A. H. Blin, B. Hiller, Y. P. Ivanov, A. A. Osipov, and U.-G. Meißner, Pion observables in the extended njl model with vector and axial-vector mesons, Annals of Physics 249, 499 (1996).DOI
  17. [17]J. Bijnens, Chiral lagrangians and nambu-jona-lasinio like models, Physics Reports 265, 370 (1996).
  18. [18]W. E. Caswell, Asymptotic Behavior of Nonabelian Gauge Theories to Two Loop Order, Phys. Rev. Lett. 33, 244 (1974).DOI
  19. [19]T. Banks and A. Zaks, On the Phase Structure of Vector-Like Gauge Theories with Massless Fermions, Nucl. Phys. B 196, 189 (1982).DOI
  20. [20]T. Appelquist, G. T. Fleming, and E. T. Neil, Lattice Study of Conformal Behavior in SU(3) Yang-Mills Theories, Phys. Rev. D 79, 076010 (2009), arXiv:0901.3766 [hep-ph].DOI
  21. [21]T. Appelquist et al. (LSD), Lattice simulations with eight flavors of domain wall fermions in SU(3) gauge theory, Phys. Rev. D 90, 114502 (2014), arXiv:1405.4752 [hep-lat].DOI
  22. [22]A. Hasenfratz and D. Schaich, Nonperturbative p function of twelve-flavor SU(3) gauge theory, JHEP 02, 132, arXiv:1610.10004 [hep-lat].DOI
  23. [23]A. Bashir, A. Raya, and J. Rodriguez-Quintero, Qcd: restoration of chiral symmetry and deconfinement for large n f, Physical Review D 88, 054003 (2013).DOI
  24. [24]M. Hopfer, C. S. Fischer, and R. Alkofer, Running coupling in the conformal window of large-Nf QCD, JHEP 11, 035, arXiv:1405.7031 [hep-ph].DOI
  25. [25]A. Ahmad, A. Bashir, M. A. Bedolla, and J. J. Cobos-Martínez, Flavor, temperature and magnetic field dependence of the QCD phase diagram: magnetic catalysis and its inverse, J. Phys. G 48, 075002 (2021), arXiv:2008.03847 [hep-ph].DOI
  26. [26]A. Ahmad and A. Murad, Color-flavor dependence of the Nambu-Jona-Lasinio model and QCD phase diagram, Chin. Phys. C 46, 083109 (2022), arXiv:2201.09980 [hep-ph].DOI
  27. [27]Y.-h. Chen et al., Flavor dependence of chiral symmetry breaking and the conformal window, arXiv:2602.14489 (2026), arXiv:2602.14489.DOI
  28. [28]A. Ahmad and A. Farooq, Schwinger Pair Production in QCD from Flavor-Dependent Contact Interaction Model of Quarks, Braz. J. Phys. 54, 212 (2024), arXiv:2302.13265 [hep-ph].arxiv.org/abs/2302.13265
  29. [29]A. Ahmad, Qcd phase diagram for large nf : analysis from contact interaction effective potential, Rev.Mex. Fís. 72, 030801–030814 (2026).arxiv.org/abs/2503.16903
  30. [30]A. Ahmad, π- and K-Mesons Properties for Large Nf, Int. J. Theor. Phys. 65, 88 (2026), arXiv:2401.11186 [hep-ph].
  31. [31]S. R. Amendolia et al., A Measurement of the Pion Charge Radius, Phys. Lett. B 146, 116 (1984).
  32. [32]S. R. Amendolia et al. (NA7), A measurement of the space-like pion electromagnetic form factor, Nucl. Phys. B 277, 168 (1986).DOI
  33. [33]E. B. Dally, J. M. Hauptman, J. Kubic, D. H. Stork, A. B. Watson, Z. Guzik, T. S. Nigmanov, V. D. Riabtsoy, E. N. Tsyganov, A. S. Vodopianov, A. Beretvas, A. Grigorian, J. C. Tompkins, T. E. Toohig, A. A. Wehmann, J. A. Poirier, C. A. Rey, J. T. Volk, P. D. Rapp, and P. F. Shepard, Elastic-scattering measurement of the negative-pion radius, Phys. Rev. Lett. 48, 375 (1982).DOI
  34. [34]E. B. Dally et al., DIRECT MEASUREMENT OF THE NEGATIVE KAON FORM-FACTOR, Phys. Rev. Lett. 45, 232 (1980).DOI
  35. [35]S. R. Amendolia et al., A Measurement of the Kaon Charge Radius, Phys. Lett. B 178, 435 (1986).DOI
  36. [36]J. Gasser and H. Leutwyler, Chiral Perturbation Theory to One Loop, Annals Phys. 158, 142 (1984).DOI
  37. [37]J. Gasser and H. Leutwyler, Chiral Perturbation Theory: Expansions in the Mass of the Strange Quark, Nucl. Phys. B 250, 465 (1985).DOI
  38. [38]J. Gasser and H. Leutwyler, Low-Energy Expansion of Meson Form-Factors, Nucl. Phys. B 250, 517 (1985).DOI
  39. [39]G. M. Huber et al. (Jefferson Lab F(π)), Charged pion form factor between q² = 0.60 and 2.45 GeV², Phys. Rev. C 78, 045203 (2008), arXiv:0809.3052.
  40. [40]T. Horn et al. (Jefferson Lab F(π-2)), Determination of the charged pion form factor at q² = 1.60 and 2.45 GeV², Phys. Rev. Lett. 97, 192001 (2006), arXiv:nucl-ex/0607005.DOI
  41. [41]A. C. Aguilar et al., Pion and Kaon Structure at the Electron-Ion Collider, Eur. Phys. J. A 55, 190 (2019), arXiv:1907.08218 [nucl-ex].DOI
  42. [42]R. Abdul Khalek et al., Science requirements and detector concepts for the electron-ion collider: EIC yellow report, Nucl. Phys. A 1026, 122447 (2022), arXiv:2103.05419. [43] C. Chen, L. Chang, C. D. Roberts, S. M. Schmidt, S. Wan, and D. J. Wilson, Features and flaws of a contact interaction treatment of the kaon, Phys. Rev. C 87, 045207 (2013), arXiv:1212.2212 [nucl-th].DOI
  43. [44]J. S. Schwinger, On gauge invariance and vacuum polarization, Phys. Rev. 82, 664 (1951).
  44. [45]C. D. Roberts and A. G. Williams, Dyson-schwinger equations and their application to hadronic physics, Prog. Part. Nucl. Phys. 33, 477 (1994).
  45. [46]J. M. Cornwall, Dynamical Mass Generation in Continuum QCD, Phys. Rev. D 26, 1453 (1982).DOI
  46. [47]A. C. Aguilar, D. Binosi, and J. Papavassiliou, The Gluon Mass Generation Mechanism: A Concise Primer, Front. Phys. (Beijing) 11, 111203 (2016), arXiv:1511.08361 [hep-ph].arxiv.org/abs/1511.08361
  47. [48]P. Boucaud, J. P. Leroy, A. L. Yaouanc, J. Micheli, O. Pene, and J. Rodriguez-Quintero, The Infrared Behaviour of the Pure Yang-Mills Green Functions, Few Body Syst. 53, 387 (2012), arXiv:1109.1936 [hep-ph].
  48. [49]H. L. L. Roberts, L. Chang, I. C. Cloet, and C. D. Roberts, Masses of ground and excited-state hadrons, Few Body Syst. 51, 1 (2011), arXiv:1101.4244 [nucl-th].DOI
  49. [50]D. Ebert, T. Feldmann, and H. Reinhardt, Extended NJL model for light and heavy mesons without q - anti-q thresholds, Phys. Lett. B388, 154 (1996), arXiv:hep-ph/9608223 [hep-ph].DOI
  50. [51]C. D. Roberts, M. S. Bhagwat, A. Holl, and S. V. Wright, Aspects of hadron physics, Eur. Phys. J. ST 140, 53 (2007), arXiv:0802.0217 [nucl-th].DOI
  51. [52]A. Ahmad and M. Jamil, Confinement order parameters in QCD: a view from Schwinger-Dyson equations and related low energy QCD models, Chin. Phys. C 50, 113102 (2026).DOI
  52. [53]L. McLerran and R. D. Pisarski, Phases of cold, dense quarks at large N(c), Nucl. Phys. A 796, 83 (2007), arXiv:0706.2191 [hep-ph].DOI
  53. [54]C. Chen, L. Chang, C. D. Roberts, S. Wan, and D. J. Wilson, Spectrum of hadrons with strangeness, Few Body Syst. 53, 293 (2012), arXiv:1204.2553 [nucl-th].DOI
  54. [55]S. Navas et al. (Particle Data Group), Review of particle physics, Phys. Rev. D 110, 030001 (2024).
  55. [56]P. Maris, C. D. Roberts, and P. C. Tandy, Pion mass and decay constant, Phys. Lett. B 420, 267 (1998), arXiv:nucl-th/9707003.DOI
  56. [57]S. J. Brodsky, C. D. Roberts, R. Shrock, and P. C. Tandy, Confinement contains condensates, Phys. Rev. C 85, 065202 (2012), arXiv:1202.2376 [nucl-th].DOI
  57. [58]L. M. Barkov et al. (OLYA, CMD), Electromagnetic Pion Form-Factor in the Timelike Region, Nucl. Phys. B 256, 365 (1985).DOI

Paper details

vlnc:W_mw74t6j6
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents