Introduction
The LHCb Collaboration observed hidden-charm pentaquark candidates in the J / ψ p J/\psi p J / ψ p invariant mass distribution of Λ b 0 → J / ψ p K − \Lambda_b^0 \to J/\psi pK^- Λ b 0 → J / ψ p K − decays [1 ] . With an enlarged data sample, three narrow states, P c ( 4312 ) + P_c(4312)^+ P c ( 4312 ) + , P c ( 4440 ) + P_c(4440)^+ P c ( 4440 ) + and P c ( 4457 ) + P_c(4457)^+ P c ( 4457 ) + , were resolved [2 ] , and a candidate with strangeness was later observed in the J / ψ Λ J/\psi\Lambda J / ψ Λ system [3 ] . The three nonstrange states lie 9, 22 and 5 MeV below the D ˉ Σ c \bar{D}\Sigma_c D ˉ Σ c and D ˉ ∗ Σ c \bar{D}^{*}\Sigma_c D ˉ ∗ Σ c thresholds. This proximity suggests that they are hadronic molecules composed of an anticharmed meson and a charmed baryon [4 , 5 , 6 , 7 , 8 , 9 ] , although kinematical explanations of part of the signals have also been proposed [10 , 11 , 12 , 13 , 14 , 15 ] . Heavy-quark spin symmetry (HQSS) relates the D ˉ Σ c \bar{D}\Sigma_c D ˉ Σ c , D ˉ Σ c ∗ \bar{D}\Sigma_c^* D ˉ Σ c ∗ , D ˉ ∗ Σ c \bar{D}^{*}\Sigma_c D ˉ ∗ Σ c and D ˉ ∗ Σ c ∗ \bar{D}^{*}\Sigma_c^* D ˉ ∗ Σ c ∗ systems and predicts seven S S S -wave states with isospin 1 2 \frac{1}{2} 2 1 [8 , 9 , 16 ] . The spin and parity of the observed states have not yet been determined experimentally. Models with contact or vector-exchange interactions assign J P = 1 2 − J^P=\frac{1}{2}^{-} J P = 2 1 − to P c ( 4440 ) P_c(4440) P c ( 4440 ) and 3 2 − \frac{3}{2}^{-} 2 3 − to P c ( 4457 ) P_c(4457) P c ( 4457 ) [9 ] , whereas analyses that include one-pion exchange and the observed J / ψ p J/\psi p J / ψ p spectrum prefer the opposite ordering [17 , 18 , 19 ] .
Electromagnetic properties provide direct information on the internal structure of hadrons. The magnetic moments of the P c P_c P c states have been calculated in the constituent quark model for molecular and compact configurations [20 , 21 , 22 , 23 ] , in a molecular picture with coupled channels and D D D -wave components [24 ] , and in chiral perturbation theory [25 ] . The magnetic dipole and higher multipole moments have also been studied with light-cone QCD sum rules for molecular and compact interpolating currents [26 , 27 , 28 , 29 , 30 ] , and the strange partners have been investigated in Refs. [31 , 32 ] . The radiative transition P c ( 4457 ) → P c ( 4312 ) γ P_c(4457) \to P_c(4312)\gamma P c ( 4457 ) → P c ( 4312 ) γ has been studied in Ref. [33 ] . To the best of our knowledge, however, the momentum dependence of the electromagnetic form factors of the nonstrange P c P_c P c states at their pole positions has not been studied so far.
In this work, we calculate the electromagnetic form factors of the seven states generated in the coupled-channel model with the HQSS vector-exchange kernel of Ref. [9 ] . We follow the method developed by Sekihara, Hyodo and Jido for the Λ ( 1405 ) \Lambda(1405) Λ ( 1405 ) [34 , 35 ] , in which the form factor of a resonance is defined by the residue at the double pole of the photon-coupled scattering amplitude. The same definition is used in the on-shell formalism of Ref. [36 ] . In this approach, no ansatz for the wave function is required. Compared with the Λ ( 1405 ) \Lambda(1405) Λ ( 1405 ) , the present system has three distinct features. First, the P c + P_c^+ P c + is charged, so that the photon vertex generated by gauging the energy-dependent kernel contributes to the current. This vertex provides the energy-dependence term of the compositeness sum rule [37 , 38 , 39 ] in the charge normalization. Second, the coupled channels are related by HQSS, and the magnetic operators connect different channels. For this channel space, we construct the photon–four-hadron vertex, which consists of the gauged kernel and of the lowest magnetic and electric contact operators allowed by HQSS. Third, decay channels open slightly below the poles. The photon loops of these channels have a triangle singularity at spacelike momentum transfer, which requires a specification of the analytic continuation of the residue in Q 2 Q^2 Q 2 .
This paper is organized as follows. In Sec. II, we define the form factors and introduce the coupled-channel model and the photon couplings. In Sec. III, we present the form factors and radii and discuss their relation to the zero-range limit. The magnetic moments and the two spin orderings are discussed in Sec. IV. The sensitivity of the results to the observed masses and widths is examined in § 1.4 , and the open channels are treated in Sec. VI. Section VII is devoted to a summary. Radii are given in f m 2 \mathrm{fm}^2 fm 2 with ℏ c = 197.327 M e V f m \hbar c = 197.327\ \mathrm{MeV\,fm} ℏ c = 197.327 MeV fm .
Throughout this paper, we use the metric ( + , − , − , − ) (+,-,-,-) ( + , − , − , − ) , natural units ℏ = 1 = c \hbar= 1 = c ℏ = 1 = c , e = 4 π α e = \sqrt{4\pi\alpha} e = 4 π α , α = 1 / 137 \alpha= 1/137 α = 1/137 and L i n t = − e J e m μ A μ \mathcal{L}_{\mathrm{int}} = -eJ_{\mathrm{em}}^\mu A_\mu L int = − e J em μ A μ . The virtual photon is incoming in the process P c ( P ) + γ ∗ ( q ) → P c ( P ′ ) P_c(P) + \gamma^*(q) \to P_c(P') P c ( P ) + γ ∗ ( q ) → P c ( P ′ ) , with q = P ′ − P q = P' - P q = P ′ − P and Q 2 = − q 2 Q^2 = -q^2 Q 2 = − q 2 . Following Refs. [34 , 35 ] , we define the form factors of a resonance with the pole position z R = M R − i Γ R / 2 z_R = M_R-i\Gamma_R/2 z R = M R − i Γ R /2 by the matrix element of the current between the resonance states,
⟨ z R ( P ′ ) ( − ) ∣ J μ ( x ) ∣ z R ( P ) ( + ) ⟩ = e i ( P ′ − P ) ⋅ x F μ ( Q 2 ) , (1) \langle z_R(P')^{(-)}|J^\mu(x)|z_R(P)^{(+)}\rangle= e^{i(P'-P)\cdot x}F^\mu(Q^2), \tag*{(1)} ⟨ z R ( P ′ ) ( − ) ∣ J μ ( x ) ∣ z R ( P ) ( + ) ⟩ = e i ( P ′ − P ) ⋅ x F μ ( Q 2 ) , (1)
which is complex in general. In the Breit frame, where P = − q / 2 P=-q/2 P = − q /2 and P ′ = + q / 2 P'=+q/2 P ′ = + q /2 , the charge and magnetic dipole form factors are defined with nonrelativistic spin wave functions as
⟨ J 0 ⟩ M ′ M = G E ( Q 2 ) δ M ′ M , \langle J^0\rangle_{M'M}=G_E(Q^2)\delta_{M'M}, ⟨ J 0 ⟩ M ′ M = G E ( Q 2 ) δ M ′ M ,
⟨ J ⟩ M ′ M = i 2 M p G M ( Q 2 ) J ( J × q ) M ′ M , (2) \langle\mathbf{J}\rangle_{M'M}=\frac{i}{2M_p}\frac{G_M(Q^2)}{J}(\mathbf{J}\times\mathbf{q})_{M'M}, \tag*{(2)} ⟨ J ⟩ M ′ M = 2 M p i J G M ( Q 2 ) ( J × q ) M ′ M , (2)
where J \mathbf{J} J are the spin-J J J matrices, M p M_p M p is the proton mass and the current is given in units of e e e . Both form factors are dimensionless and normalized as G E ( 0 ) = 1 G_E(0)=1 G E ( 0 ) = 1 and G M ( 0 ) = μ / μ N G_M(0)=\mu/\mu_N G M ( 0 ) = μ / μ N with μ N = e / ( 2 M p ) \mu_N=e/(2M_p) μ N = e / ( 2 M p ) . They coincide with the covariant form factors up to corrections of relative order Q 2 / ( 4 M R 2 ) Q^2/(4M_R^2) Q 2 / ( 4 M R 2 ) , which amount to 1.3 % 1.3\% 1.3% at Q 2 = 1 G e V 2 Q^2=1\ \mathrm{GeV}^2 Q 2 = 1 GeV 2 . The electric quadrupole and magnetic octupole form factors of the states with J ≥ 3 2 J\geq\frac{3}{2} J ≥ 2 3 require D D D -wave components or the higher multipoles of the D ˉ ∗ \bar{D}^* D ˉ ∗ and Σ c ∗ \Sigma_c^* Σ c ∗ , and are therefore not included in the present S S S -wave model. Nonvanishing higher multipole moments of the spin-3 2 \frac{3}{2} 2 3 states have been obtained with light-cone sum rules [29 ] .
In addition to the electromagnetic current, we consider three probe currents: the baryon-number current (F B F_B F B ), the current that counts the anticharm quarks (F c ˉ F_{\bar{c}} F c ˉ ) and the charm current J C μ = J N c μ − J N ˉ c μ ( F C ) J_C^\mu=J_{N_c}^\mu-J_{\bar{N}_c}^\mu(F_C) J C μ = J N c μ − J N ˉ c μ ( F C ) . For pointlike constituents, F B F_B F B traces the position of the baryon and F c ˉ F_{\bar{c}} F c ˉ that of the meson. The mean squared radius associated with a probe is ⟨ r 2 ⟩ = − 6 d F / d Q 2 \langle r^2\rangle=-6\,dF/dQ^2 ⟨ r 2 ⟩ = − 6 d F / d Q 2 at Q 2 = 0 Q^2=0 Q 2 = 0 , divided by G M ( 0 ) G_M(0) G M ( 0 ) in the case of the magnetic radius.
We denote by T l ′ l ( s ) T_{l'l}(\sqrt{s}) T l ′ l ( s ) the S S S -wave meson–baryon amplitude for M l B l → M l ′ B l ′ M_lB_l\to M_{l'}B_{l'} M l B l → M l ′ B l ′ . Near a pole on the appropriate Riemann sheet, T l ′ l ≃ g l ′ g l / ( s − z R ) T_{l'l}\simeq g_{l'}g_l/(\sqrt{s}-z_R) T l ′ l ≃ g l ′ g l / ( s − z R ) , where the residues define the effective couplings g l g_l g l of the state to the channels [40 ] . In the photon-coupled amplitude T γ , l ′ l μ ( s ′ , s ; Q 2 ) T^\mu_{\gamma,l'l}(\sqrt{s'},\sqrt{s};Q^2) T γ , l ′ l μ ( s ′ , s ; Q 2 ) for M l B l + γ ∗ → M l ′ B l ′ M_lB_l+\gamma^*\to M_{l'}B_{l'} M l B l + γ ∗ → M l ′ B l ′ , the resonance appears as a double pole, and the matrix element (1) is given by its coefficient [34 , 35 , 36 ] ,
F μ ( Q 2 ) = ( s ′ − z R ) T γ , l ′ l μ ( s ′ , s ; Q 2 ) ( s − z R ) T l ′ l ( s ) ∣ s , s ′ → z R . (3) F^\mu(Q^2)=\left.\frac{(\sqrt{s'}-z_R)T^\mu_{\gamma,l'l}(\sqrt{s'},\sqrt{s};Q^2)(\sqrt{s}-z_R)}{T_{l'l}(\sqrt{s})}\right|_{\sqrt{s},\sqrt{s'}\to z_R}. \tag*{(3)} F μ ( Q 2 ) = T l ′ l ( s ) ( s ′ − z R ) T γ , l ′ l μ ( s ′ , s ; Q 2 ) ( s − z R ) s , s ′ → z R . (3)
The nonresonant background and all terms with at most a single pole do not contribute to Eq. (3) . The amplitudes are normalized as i M γ = − i e ϵ μ T γ μ i\mathcal{M}_\gamma=-ie\epsilon_\mu T^\mu_\gamma i M γ = − i e ϵ μ T γ μ and T = − M T=-\mathcal{M} T = − M . With our sign convention for L i n t \mathcal{L}_{\mathrm{int}} L int , the amplitude T γ T_\gamma T γ differs in overall sign from that of Ref. [35 ] , whereas the form factors are identical.
Coupled-channel model
We employ the isospin-1 2 \frac{1}{2} 2 1 coupled channels of Ref. [9 ] ,
1 2 − : { η c N , J / ψ N , D ˉ Λ c , D ˉ Σ c , D ˉ ∗ Λ c , D ˉ ∗ Σ c , D ˉ ∗ Σ c ∗ } , \frac{1}{2}^{-}:\{\eta_cN,\ J/\psi N,\ \bar{D}\Lambda_c,\ \bar{D}\Sigma_c,\ \bar{D}^{*}\Lambda_c,\ \bar{D}^{*}\Sigma_c,\ \bar{D}^{*}\Sigma_c^{*}\}, 2 1 − : { η c N , J / ψ N , D ˉ Λ c , D ˉ Σ c , D ˉ ∗ Λ c , D ˉ ∗ Σ c , D ˉ ∗ Σ c ∗ } ,
3 2 − : { J / ψ N , D ˉ ∗ Λ c , D ˉ ∗ Σ c , D ˉ ∗ Σ c ∗ } , \frac{3}{2}^{-}:\{J/\psi N,\ \bar{D}^{*}\Lambda_c,\ \bar{D}^{*}\Sigma_c,\ \bar{D}^{*}\Sigma_c^{*}\}, 2 3 − : { J / ψ N , D ˉ ∗ Λ c , D ˉ ∗ Σ c , D ˉ ∗ Σ c ∗ } ,
5 2 − : { D ˉ ∗ Σ c ∗ } , (4) \frac{5}{2}^{-}:\{\bar{D}^{*}\Sigma_c^{*}\}, \tag*{(4)} 2 5 − : { D ˉ ∗ Σ c ∗ } , (4)
which are labeled by l l l , with the meson mass m l m_l m l and the baryon mass M l M_l M l . The isospin states are ∣ D ˉ ( ∗ ) Σ c ( ∗ ) ⟩ = 2 / 3 ∣ D ˉ − ( ∗ ) Σ c ( ∗ ) + ⟩ − 1 / 3 ∣ D ˉ ( ∗ ) 0 Σ c ( ∗ ) + ⟩ |\bar{D}^{(*)}\Sigma_c^{(*)}\rangle=\sqrt{2/3}\,|\bar{D}^{(*)}_{-}\Sigma_c^{(*)+}\rangle-\sqrt{1/3}\,|\bar{D}^{(*)0}\Sigma_c^{(*)+}\rangle ∣ D ˉ ( ∗ ) Σ c ( ∗ ) ⟩ = 2/3 ∣ D ˉ − ( ∗ ) Σ c ( ∗ ) + ⟩ − 1/3 ∣ D ˉ ( ∗ ) 0 Σ c ( ∗ ) + ⟩ and ∣ D ˉ ( ∗ ) Λ c ⟩ = ∣ D ˉ ( ∗ ) 0 Λ c + ⟩ |\bar{D}^{(*)}\Lambda_c\rangle=|\bar{D}^{(*)0}\Lambda_c^+\rangle ∣ D ˉ ( ∗ ) Λ c ⟩ = ∣ D ˉ ( ∗ ) 0 Λ c + ⟩ . We use the isospin-averaged masses of Ref. [41 ] (in MeV: η c \eta_c η c 2984.1, J / ψ J/\psi J / ψ 3096.9, D ˉ \bar{D} D ˉ 1867.25, D ˉ ∗ \bar{D}^* D ˉ ∗ 2008.56, N N N 938.92, Λ c \Lambda_c Λ c 2286.46, Σ c \Sigma_c Σ c 2453.46, Σ c ∗ \Sigma_c^* Σ c ∗ 2518.1).
In the heavy-quark limit, the interaction depends only on the light degrees of freedom. The most general S S S -wave kernel V l ′ l ( J ) V_{l'l}(J) V l ′ l ( J ) consistent with HQSS is given in Eqs. (3) –(5) of Ref. [9 ] in terms of seven couplings, μ 1 \mu_1 μ 1 , μ 2 \mu_2 μ 2 , μ 3 \mu_3 μ 3 , μ 12 \mu_{12} μ 12 , μ 13 \mu_{13} μ 13 and μ 23 \mu_{23} μ 23 for light-quark spin 1 2 \frac{1}{2} 2 1 and λ 2 \lambda_2 λ 2 for light-quark spin 3 2 \frac{3}{2} 2 3 [16 , 42 ] . These couplings are determined by the extended local hidden gauge approach [4 , 5 , 16 ] , in which the interaction arises from the exchange of ρ \rho ρ and ω \omega ω mesons between the light quarks and from D ∗ D^* D ∗ exchange in the transitions to the hidden-charm channels.
Following Eq. (6) of Ref. [9 ] , we use
μ 1 = μ 23 = 0 , μ 2 = − μ 3 = k l 0 + k l ′ 0 4 f 2 , λ 2 = r μ 3 , μ 13 = − μ 12 , μ 12 = − 6 κ m ρ 2 p D ∗ , l ′ l 2 − m D ∗ 2 k l ′ 0 + k l 0 4 f 2 , (5) \begin{aligned}
\mu_1 &= \mu_{23}=0,\qquad\mu_2=-\mu_3=\frac{k_l^{0}+k_{l'}^{0}}{4f^2},\\
\lambda_2 &= r\mu_3,\qquad\mu_{13}=-\mu_{12},\\
\mu_{12} &= -\sqrt{6}\,\kappa\,\frac{m_\rho^2}{p_{D^*,l'l}^2-m_{D^*}^2}\,\frac{k_{l'}^{0}+k_l^{0}}{4f^2}, \tag*{(5)}
\end{aligned} μ 1 λ 2 μ 12 = μ 23 = 0 , μ 2 = − μ 3 = 4 f 2 k l 0 + k l ′ 0 , = r μ 3 , μ 13 = − μ 12 , = − 6 κ p D ∗ , l ′ l 2 − m D ∗ 2 m ρ 2 4 f 2 k l ′ 0 + k l 0 , (5)
where two additional parameters r r r and κ \kappa κ have been introduced, f = 93 M e V f=93\,\mathrm{MeV} f = 93 MeV , m ρ = 775.5 M e V m_\rho=775.5\,\mathrm{MeV} m ρ = 775.5 MeV , k l 0 = ( s + m l 2 − M l 2 ) / ( 2 s ) k_l^0=(s+m_l^2-M_l^2)/(2\sqrt{s}) k l 0 = ( s + m l 2 − M l 2 ) / ( 2 s ) is the meson energy and p D ∗ , l ′ l 2 = m l ′ 2 + m l 2 − 2 k l ′ 0 k l 0 p_{D^*,l'l}^2=m_{l'}^2+m_l^2-2k_{l'}^0k_l^0 p D ∗ , l ′ l 2 = m l ′ 2 + m l 2 − 2 k l ′ 0 k l 0 [5 ] . The hidden gauge approach corresponds to r = κ = 1 r=\kappa=1 r = κ = 1 , whereas HQSS alone does not fix these parameters. Since the magnetic operators of Sec. II C connect different channels, the relative phases of the channel states affect the form factors. We therefore evaluate the kernel and all photon vertices in a single phase convention, defined by the multiplet fields H ˉ = − D ˉ + σ ⋅ D ˉ ∗ \bar{H}=-\bar{D}+\boldsymbol{\sigma}\cdot\bar{\boldsymbol{D}}^* H ˉ = − D ˉ + σ ⋅ D ˉ ∗ , S i = 1 3 σ i Σ c + Σ c ∗ i S^i=\frac{1}{\sqrt{3}}\sigma^i\Sigma_c+\Sigma_c^{*i} S i = 3 1 σ i Σ c + Σ c ∗ i and Ψ = − η c + σ ⋅ ψ \Psi=-\eta_c+\boldsymbol{\sigma}\cdot\boldsymbol{\psi} Ψ = − η c + σ ⋅ ψ [43 ] . This convention corresponds to the Condon–Shortley coupling scheme, in which the c ˉ \bar{c} c ˉ is coupled first in the mesons, the c c c first in the baryons and the meson first in the two-hadron state. In this convention, we reproduce the kernel matrices of Ref. [9 ] , including the relative sign μ 13 = − μ 12 \mu_{13}=-\mu_{12} μ 13 = − μ 12 that enters the interference terms of G M G_M G M .
Using the on-shell factorization of the kernel, the Bethe–Salpeter equation reduces to the algebraic form [44 , 45 ]
T ( s ) = [ 1 − V ( s ) G ( s ) ] − 1 V ( s ) , G = diag ( G l ) , (6) T(\sqrt{s})=\left[1-V(\sqrt{s})G(\sqrt{s})\right]^{-1}V(\sqrt{s}),\qquad G=\operatorname{diag}(G_l), \tag*{(6)} T ( s ) = [ 1 − V ( s ) G ( s ) ] − 1 V ( s ) , G = diag ( G l ) , (6)
where G l G_l G l is the meson–baryon loop function in dimensional regularization [45 ] ,
G l ( s ) = i ∫ d 4 q ( 2 π ) 4 2 M l [ ( P − q ) 2 − M l 2 ] [ q 2 − m l 2 ] = 2 M l 16 π 2 [ a l + 2 + ∫ 0 1 d x ln x M l 2 + ( 1 − x ) m l 2 − x ( 1 − x ) s μ r e g 2 ] . (7) \begin{aligned}
G_l(\sqrt{s})&=i\int\frac{\mathrm{d}^4q}{(2\pi)^4}\frac{2M_l}{\left[(P-q)^2-M_l^2\right]\left[q^2-m_l^2\right]}\\
&=\frac{2M_l}{16\pi^2}\left[a_l+2+\int_0^1\mathrm{d}x\ln\frac{xM_l^2+(1-x)m_l^2-x(1-x)s}{\mu_{\mathrm{reg}}^2}\right]. \tag*{(7)}
\end{aligned} G l ( s ) = i ∫ ( 2 π ) 4 d 4 q [ ( P − q ) 2 − M l 2 ] [ q 2 − m l 2 ] 2 M l = 16 π 2 2 M l [ a l + 2 + ∫ 0 1 d x ln μ reg 2 x M l 2 + ( 1 − x ) m l 2 − x ( 1 − x ) s ] . (7)
Here, a l a_l a l is the subtraction constant at the regularization scale μ r e g = 1 G e V \mu_{\mathrm{reg}}=1\,\mathrm{GeV} μ reg = 1 GeV , and s → s + i ϵ s\to s+i\epsilon s → s + i ϵ . The factor 2 M l 2M_l 2 M l of the baryon propagator corresponds to the normalization u ˉ u = 1 \bar{u}u=1 u ˉ u = 1 [44 ] . In this convention, V V V and T T T have dimension E − 1 E^{-1} E − 1 , G G G has dimension E E E , and the couplings g l g_l g l are dimensionless. A negative V V V corresponds to attraction. The loop function on the second Riemann sheet of channel l l l is given by
G l I I ( s ) = G l I ( s ) + i M l q l 2 π s , Im q l > 0 , (8) G_l^{\mathrm{II}}(\sqrt{s})=G_l^{\mathrm{I}}(\sqrt{s})+i\frac{M_lq_l}{2\pi\sqrt{s}},\qquad\operatorname{Im}q_l>0, \tag*{(8)} G l II ( s ) = G l I ( s ) + i 2 π s M l q l , Im q l > 0 , (8)
where q l q_l q l is the center-of-mass momentum on the first sheet. We refer to a channel as open if its threshold lies below M R M_R M R , and we search for the poles on the Riemann sheet on which all open channels are taken on their second sheets.
In this work we consider two parameter sets as follows. Set I corresponds to the model of Ref. [9 ] , with r = κ = 1 r=\kappa=1 r = κ = 1 and a common subtraction constant a l = − 2.09 a_l=-2.09 a l = − 2.09 . In this set, the two D ˉ ∗ Σ c \bar{D}^*\Sigma_c D ˉ ∗ Σ c states lie within 0.5 MeV of each other and deviate by 5 and 12 MeV from the observed masses. In set II , the parameters are fitted to the observed masses and widths reported by the LHCb Collaboration [2 ] . We identify these values with the pole positions. A fit of a a a , r r r and κ \kappa κ reproduces the widths, but the D ˉ Σ c \bar{D}\Sigma_c D ˉ Σ c state then lies 5.7 MeV below the observed mass of P c ( 4312 ) P_c(4312) P c ( 4312 ) , whose lower uncertainty is 0.9 MeV. We therefore allow the subtraction constant of the loops with a D ˉ \bar{D} D ˉ meson to differ from that of the loops with a D ˉ ∗ \bar{D}^* D ˉ ∗ meson, a D ˉ = a + δ a_{\bar{D}}=a+\delta a D ˉ = a + δ , which corresponds to an effect of order Λ Q C D / m c \Lambda_{\mathrm{QCD}}/m_c Λ QCD / m c . The three masses determine a a a , δ \delta δ and r r r , and κ \kappa κ is fixed by a least-squares fit to the three widths, with the statistical and systematic uncertainties added in quadrature. We obtain
a = − 2.227 , δ = + 0.052 , a=-2.227,\qquad\delta=+0.052, a = − 2.227 , δ = + 0.052 ,
r = 0.554 , κ = 0.7 2 − 0.14 + 0.10 . (9) r=0.554,\qquad\kappa=0.72^{+0.10}_{-0.14}. \tag*{(9)} r = 0.554 , κ = 0.7 2 − 0.14 + 0.10 . (9)
with χ min 2 = 1.6 \chi_{\min}^2=1.6 χ m i n 2 = 1.6 . The uncertainty of κ \kappa κ is given by the interval χ 2 ≤ χ min 2 + 1 \chi^2\leq\chi_{\min}^2+1 χ 2 ≤ χ m i n 2 + 1 , in which the other parameters are redetermined for each value of κ \kappa κ . The value r < 1 r<1 r < 1 reduces the attraction for light-quark spin 3 2 \frac{3}{2} 2 3 , so that the 1 2 − D ˉ ∗ Σ c \frac{1}{2}^{-}\bar{D}^*\Sigma_c 2 1 − D ˉ ∗ Σ c state lies below the 3 2 − \frac{3}{2}^{-} 2 3 − state.
The pole positions obtained in the two parameter sets are listed in Table I. The poles of set I agree with those of Ref. [9 ] within 0.2 MeV in the mass and 0.1 MeV in the width, except for the D ˉ Σ c \bar{D}\Sigma_c D ˉ Σ c state. For this state, the pole position reported in Ref. [9 ] , 4306.4 − 7.6 i 4306.4-7.6i 4306.4 − 7.6 i MeV, is reproduced if the weakly coupled D ˉ ∗ Λ c \bar{D}^*\Lambda_c D ˉ ∗ Λ c channel is taken on its first Riemann sheet. On the sheet specified above, we find the pole at 4306.6 − 7.0 i 4306.6-7.0i 4306.6 − 7.0 i MeV. In set II, the masses of the three observed states are reproduced by construction, and their widths are 10.1, 23.8 and 3.0 MeV, compared with the experimental values 9. 8 − 2.5 + 4.6 9.8^{+4.6}_{-2.5} 9. 8 − 2.5 + 4.6 , 20. 6 − 11.9 + 10.0 20.6^{+10.0}_{-11.9} 20. 6 − 11.9 + 10.0 and 6. 4 − 2.0 + 6.0 6.4^{+6.0}_{-2.0} 6. 4 − 2.0 + 6.0 MeV. The model also predicts four partner states, a D ˉ Σ c ∗ \bar{D}\Sigma_c^* D ˉ Σ c ∗ state at 4378 MeV and three D ˉ ∗ Σ c ∗ \bar{D}^*\Sigma_c^* D ˉ ∗ Σ c ∗ states at 4505, 4517 and 4525 MeV. Since the constituents are treated as stable particles and the width of the Σ c ∗ \Sigma_c^* Σ c ∗ is comparable with or larger than the binding energies of these four states, their pole positions should be regarded as estimates (see § 1.4 ).
We adopt the spin assignment of Ref. [9 ] , 1 2 − \frac{1}{2}^{-} 2 1 − for P c ( 4440 ) P_c(4440) P c ( 4440 ) and 3 2 − \frac{3}{2}^{-} 2 3 − for P c ( 4457 ) P_c(4457) P c ( 4457 ) , which we refer to as ordering A. Within the present kernel, this assignment is favored by the widths. Since the J / ψ N J/\psi N J / ψ N channel couples to the D ˉ ∗ Σ c \bar{D}^*\Sigma_c D ˉ ∗ Σ c channel with the weights 5 6 \frac{5}{6} 6 5 for J = 1 2 J=\frac{1}{2} J = 2 1 and 1 3 \frac{1}{3} 3 1 for J = 3 2 J=\frac{3}{2} J = 2 3 , the 1 2 − \frac{1}{2}^{-} 2 1 − state is broader than the 3 2 − \frac{3}{2}^{-} 2 3 − state. For the opposite ordering B, the observed masses require r = 3.5 r=3.5 r = 3.5 –4.4, the resulting widths show the reverse of the observed pattern, and the D ˉ Σ c \bar{D}\Sigma_c D ˉ Σ c state lies 15 MeV below P c ( 4312 ) P_c(4312) P c ( 4312 ) mass. This argument, however, does not apply to models with one-pion exchange, in which the tensor force modifies the relation between spin and width [17 , 18 , 19 ] . We therefore present the magnetic moments for both orderings in Sec. IV.
Table 1. Pole positions of the seven states in the two parameter sets (masses and widths in MeV), compositeness X m a i n X_{\mathrm{main}} X main of the dominant channel and energy-dependence term Z Z Z of Eq. (10) in set II, and observed masses and widths [2 ] . The states are labeled by their dominant channel. In set I, X m a i n = 0.99 X_{\mathrm{main}}=0.99 X main = 0.99 –1.01 1.01 1.01 and Re Z = 0.006 \operatorname{Re}Z=0.006 Re Z = 0.006 –0.013 0.013 0.013 .
Table 2. One-body moments and transition moments in μ N \mu_N μ N in the constituent quark model, for the charge states and for the I = 1 2 I=\frac{1}{2} I = 2 1 average. μ q q \mu_{qq} μ q q is the moment of the light diquark and μ q \mu_q μ q that of the light quark of the meson. For Σ c ( ∗ ) → Λ c \Sigma_c^{(*)}\to\Lambda_c Σ c ( ∗ ) → Λ c the last column is the overlap with ∣ D ˉ ( ∗ ) 0 Λ c ⟩ \lvert\bar{D}^{(*)0}\Lambda_c\rangle ∣ D ˉ ( ∗ ) 0 Λ c ⟩ . The lattice values are those of Table III of Ref. [46 ] for the linear and the quadratic extrapolation in m π 2 m_\pi^2 m π 2 , with μ Σ c + = 1 2 ( μ Σ c + + μ Σ c 0 ) \mu_{\Sigma_c^{+}}=\frac{1}{2}(\mu_{\Sigma_c^{+}}+\mu_{\Sigma_c^{0}}) μ Σ c + = 2 1 ( μ Σ c + + μ Σ c 0 ) in the average.
The compositeness of channel l l l and the energy-dependence term are defined as [37 , 38 , 39 , 47 ]
X l = − g l 2 d G l d s ∣ z R , ∑ l X l + Z = 1 , X_l=-g_l^2\left.\frac{\mathrm{d}G_l}{\mathrm{d}\sqrt{s}}\right|_{z_R},\qquad\sum_l X_l+Z=1, X l = − g l 2 d s d G l z R , l ∑ X l + Z = 1 ,
Z = − ∑ l ′ , l g l ′ G l ′ d V l ′ l d s ∣ z R G l g l . (10) Z=-\sum_{l',l}g_{l'}G_{l'}\left.\frac{\mathrm{d}V_{l'l}}{\mathrm{d}\sqrt{s}}\right|_{z_R}G_lg_l. \tag*{(10)} Z = − l ′ , l ∑ g l ′ G l ′ d s d V l ′ l z R G l g l . (10)
For a resonance, the X l X_l X l are complex and cannot be interpreted as probabilities. All states are dominated by a single channel, with ∣ X m a i n − 1 ∣ < 0.03 \lvert X_{\mathrm{main}}-1\rvert<0.03 ∣ X main − 1 ∣ < 0.03 and ∣ X l ∣ < 0.02 \lvert X_l\rvert<0.02 ∣ X l ∣ < 0.02 for the other channels.
Photon couplings
The photon couples both to the constituents and to their interaction. Each coupling enters through the Feynman rule − i e ϵ μ Γ μ -ie\epsilon_\mu\Gamma^\mu − i e ϵ μ Γ μ , where the vertex function Γ μ \Gamma^\mu Γ μ is the matrix element of the current in units of e e e . The electromagnetic current is thus the sum of one-body currents and a photon–four-hadron vertex.
One-body currents
The charge couplings are given by the convection currents Q M ( k + k ′ ) μ Q_M(k+k')^\mu Q M ( k + k ′ ) μ of the mesons, which are diagonal in the polarization of the D ˉ ∗ \bar{D}^{*} D ˉ ∗ , and Q B ( p + p ′ ) μ / ( 2 M B ) Q_B(p+p')^\mu/(2M_B) Q B ( p + p ′ ) μ / ( 2 M B ) of the baryons in the nonrelativistic reduction [35 ] . The electric charges of the meson and the baryon are ( Q M , Q B ) = ( − 2 3 , 5 3 ) (Q_M,Q_B)=(-\frac{2}{3},\frac{5}{3}) ( Q M , Q B ) = ( − 3 2 , 3 5 ) in the D ˉ ( ∗ ) Σ c \bar{D}^{(*)}\Sigma_c D ˉ ( ∗ ) Σ c channels, obtained as the isospin average over the two charge states, and ( 0 , 1 ) (0,1) ( 0 , 1 ) in the D ˉ ( ∗ ) Λ c \bar{D}^{(*)}\Lambda_c D ˉ ( ∗ ) Λ c and hidden-charm channels. In all channels, the baryon-number and anticharm probes have the charges ( 0 , 1 ) (0,1) ( 0 , 1 ) and ( 1 , 0 ) (1,0) ( 1 , 0 ) , respectively, while the charm probe has ( − 1 , 1 ) (-1,1) ( − 1 , 1 ) in the open-charm channels and ( 0 , 0 ) (0,0) ( 0 , 0 ) in the hidden-charm channels.
The magnetic moments and the M1 transition moments of the hadrons are taken from the constituent quark model, which is consistent with the HQSS structure of the fields H ˉ \bar{H} H ˉ , S i S^i S i and Ψ \Psi Ψ . We use μ u = 1.862 \mu_u=1.862 μ u = 1.862 , μ d = − 0.931 \mu_d=-0.931 μ d = − 0.931 and μ c = 0.417 μ N \mu_c=0.417\mu_N μ c = 0.417 μ N (m u , d = 336 m_{u,d}=336 m u , d = 336 MeV, m c = 1.50 m_c=1.50 m c = 1.50 GeV) together with the experimental nucleon moments. The resulting values are listed in Table II, together with the lattice QCD results of Ref. [46 ] for the Σ c \Sigma_c Σ c . The magnetic current of a hadron is given by J m a g ( q ) = i ( μ o p / e ) α × q J_{\mathrm{mag}}(q)=i(\mu_{\mathrm{op}}/e)\boldsymbol{\alpha}\times\mathbf{q} J mag ( q ) = i ( μ op / e ) α × q . In the channel basis, the one-body magnetic operator is reduced by means of the Wigner–Eckart theorem,
⟨ l ′ ; J M ′ ∣ μ ^ ( 1 ) ∣ l ; J M ⟩ = M l ′ l ( 1 ) ( J ) ⟨ J M ′ ∣ J ∣ J M ⟩ J . (11) \langle l';JM'|\hat{\mu}^{(1)}|l;JM\rangle=\mathcal{M}_{l'l}^{(1)}(J)\frac{\langle JM'|\mathcal{J}|JM\rangle}{J}. \tag*{(11)} ⟨ l ′ ; J M ′ ∣ μ ^ ( 1 ) ∣ l ; J M ⟩ = M l ′ l ( 1 ) ( J ) J ⟨ J M ′ ∣ J ∣ J M ⟩ . (11)
The matrix M ( 1 ) = M M + M B \mathcal{M}^{(1)}=\mathcal{M}^{M}+\mathcal{M}^{B} M ( 1 ) = M M + M B consists of a meson-line and a baryon-line contribution (see Appendix C). Its diagonal elements are the additive moments of the two-hadron states, and its off-diagonal elements describe transitions between channels with a common spectator. With the isospin weights 2 3 \frac{2}{3} 3 2 and 1 3 \frac{1}{3} 3 1 , the light-quark contribution to the moment of the D ˉ ∗ \bar{D}^{*} D ˉ ∗ in the D ˉ ( ∗ ) Σ c ( ∗ ) \bar{D}^{(*)}\Sigma_c^{(*)} D ˉ ( ∗ ) Σ c ( ∗ ) channels is 1 3 ( μ u + 2 μ d ) \frac{1}{3}(\mu_u+2\mu_d) 3 1 ( μ u + 2 μ d ) . This contribution vanishes for μ u = − 2 μ d \mu_u=-2\mu_d μ u = − 2 μ d , which yields ⟨ μ D ˉ ∗ ⟩ = − μ c \langle\mu_{\bar{D}^{*}}\rangle=-\mu_c ⟨ μ D ˉ ∗ ⟩ = − μ c .
Photon–four-hadron vertex
The coupling of the photon to the interaction is described by the photon–four-hadron vertex
Γ l ′ l μ ( P , P ′ ) = Γ min , l ′ l μ + Γ M , l ′ l μ + Γ E , l ′ l μ (12) \Gamma^\mu_{l'l}(P,P')=\Gamma^\mu_{\min,l'l}+\Gamma^\mu_{M,l'l}+\Gamma^\mu_{E,l'l} \tag*{(12)} Γ l ′ l μ ( P , P ′ ) = Γ m i n , l ′ l μ + Γ M , l ′ l μ + Γ E , l ′ l μ (12)
where P P P and P ′ P' P ′ are the total four-momenta before and after the absorption of the photon. The first term is determined by the kernel through current conservation. The magnetic and electric contact terms, Γ t r μ = Γ M μ + Γ E μ \Gamma^\mu_{\mathrm{tr}}=\Gamma^\mu_M+\Gamma^\mu_E Γ tr μ = Γ M μ + Γ E μ , are transverse and contain couplings that are not fixed by the model. To the best of our knowledge, this vertex has not been derived previously for the D ˉ ( ∗ ) Σ c ( ∗ ) \bar{D}^{(*)}\Sigma_c^{(*)} D ˉ ( ∗ ) Σ c ( ∗ ) coupled-channel system, and its construction is one of the new elements of the present work.
Gauged kernel. An energy-dependent contact interaction corresponds to a Lagrangian with time derivatives. Gauging these derivatives generates the part Γ min μ \Gamma^\mu_{\min} Γ m i n μ of the vertex, which is constrained by current conservation, q μ Γ l ′ l μ = − Q H [ V l ′ l ( s ′ ) − V l ′ l ( s ) ] q_\mu\Gamma^\mu_{l'l}=-Q_H[V_{l'l}(\sqrt{s'})-V_{l'l}(\sqrt{s})] q μ Γ l ′ l μ = − Q H [ V l ′ l ( s ′ ) − V l ′ l ( s )] , where Q H Q_H Q H is the charge of the state [48 , 49 ] . For an arbitrary energy dependence of the kernel, a solution is given by
Γ min , l ′ l μ ( P , P ′ ) = − Q H ( P + P ′ ) μ s ′ − s [ V l ′ l ( s ′ ) − V l ′ l ( s ) ] → s ′ → s − Q H ( P + P ′ ) μ 2 s ∂ V l ′ l ∂ s (13) \Gamma^\mu_{\min,l'l}(P,P')=-Q_H\frac{(P+P')^\mu}{s'-s}\left[V_{l'l}(\sqrt{s'})-V_{l'l}(\sqrt{s})\right]
\xrightarrow{s'\to s}-Q_H\frac{(P+P')^\mu}{2\sqrt{s}}\frac{\partial V_{l'l}}{\partial\sqrt{s}} \tag*{(13)} Γ m i n , l ′ l μ ( P , P ′ ) = − Q H s ′ − s ( P + P ′ ) μ [ V l ′ l ( s ′ ) − V l ′ l ( s ) ] s ′ → s − Q H 2 s ( P + P ′ ) μ ∂ s ∂ V l ′ l (13)
which is unique up to transverse terms. For the kernel of Eq. (5) , the derivative is obtained from
∂ k l 0 ∂ s = ω l s , ∂ p D ∗ , l ′ l 2 ∂ s = − 2 ( k l 0 ω l + ω l ′ k l ′ 0 ) s (14) \frac{\partial k_l^0}{\partial\sqrt{s}}=\frac{\omega_l}{\sqrt{s}},\qquad
\frac{\partial p_{D^*,l'l}^2}{\partial\sqrt{s}}=-\frac{2(k_l^0\omega_l+\omega_{l'}k_{l'}^0)}{\sqrt{s}} \tag*{(14)} ∂ s ∂ k l 0 = s ω l , ∂ s ∂ p D ∗ , l ′ l 2 = − s 2 ( k l 0 ω l + ω l ′ k l ′ 0 ) (14)
where ω l = ( s + M l 2 − m l 2 ) / ( 2 s ) \omega_l=(s+M_l^2-m_l^2)/(2\sqrt{s}) ω l = ( s + M l 2 − m l 2 ) / ( 2 s ) is the baryon energy. The derivative of the D ∗ D^* D ∗ -exchange factor in μ 12 \mu_{12} μ 12 must be included to satisfy the Ward identity of Sec. II D. For the Weinberg–Tomozawa kernel, Eq. (13) reduces to the vertex of Ref. [35]. This vertex vanishes for a neutral state but contributes for the charged P c + P_c^+ P c + . In the Breit frame, it contributes only to G E G_E G E .
Magnetic contact vertex. Gauge invariance does not determine the transverse part of the vertex. Transverse vertices are constructed from the field strength F μ ν F_{\mu\nu} F μν , and each of them introduces a new coupling. We consider the operators that contain one magnetic field and no further derivative, δ L M = e B ⋅ X \delta\mathcal{L}_M=e\,\boldsymbol{B}\cdot\mathcal{X} δ L M = e B ⋅ X , where X \mathcal{X} X is a two-hadron operator that transforms as a vector in the light-quark spins.
In the present model space, parity, HQSS, time-reversal invariance, Hermiticity and the isospin structure of the photon allow 34 such operators (see Appendix B). For an I = 1 2 I=\frac{1}{2} I = 2 1 state, they contribute through ten combinations c ˉ s \bar{c}_s c ˉ s of their couplings, X = ∑ s c ˉ s X ( s ) \mathcal{X}=\sum_s\bar{c}_s\mathcal{X}^{(s)} X = ∑ s c ˉ s X ( s ) . In the D ˉ ( ∗ ) Σ c ( ∗ ) \bar{D}^{(*)}\Sigma_c^{(*)} D ˉ ( ∗ ) Σ c ( ∗ ) channels, these operators act on the spin σ q \sigma_q σ q of the light quark in the meson and on the spin S q q S_{qq} S q q of the light diquark in the baryon, X ( a ) = 2 σ q \mathcal{X}^{(a)}=2\sigma_q X ( a ) = 2 σ q , X ( b ) = 2 S q q \mathcal{X}^{(b)}=2S_{qq} X ( b ) = 2 S q q and X ( c ) j = − 2 σ q i Q q q i j \mathcal{X}^{(c)j}=-2\sigma_q^iQ_{qq}^{ij} X ( c ) j = − 2 σ q i Q q q ij , where Q q q i j Q_{qq}^{ij} Q q q ij is the quadrupole operator of the diquark, see Eq. (46) . For an incoming photon, B → i q × ϵ \boldsymbol{B}\to i\boldsymbol{q}\times\boldsymbol{\epsilon} B → i q × ϵ , and the vertex takes the form
Γ M , l ′ l μ = 2 m l ′ m l i ϵ μ ν ρ σ v ν q ρ X l ′ l , σ (15) \Gamma^\mu_{M,l'l}=2\sqrt{m_{l'}m_l}\,i\epsilon^{\mu\nu\rho\sigma}v_\nu q_\rho\mathcal{X}_{l'l,\sigma} \tag*{(15)} Γ M , l ′ l μ = 2 m l ′ m l i ϵ μν ρ σ v ν q ρ X l ′ l , σ (15)
where v μ = ( P + P ′ ) μ / [ ( P + P ′ ) 2 ] 1 / 2 v^\mu=(P+P')^\mu/[(P+P')^2]^{1/2} v μ = ( P + P ′ ) μ / [( P + P ′ ) 2 ] 1/2 , ϵ 0123 = + 1 \epsilon^{0123}=+1 ϵ 0123 = + 1 , and X μ \mathcal{X}^\mu X μ is the spin four-vector, which reduces to ( 0 , X ) (0,\boldsymbol{\mathcal{X}}) ( 0 , X ) in the rest frame of the meson–baryon pair. The factor 2 m l ′ m l 2\sqrt{m_{l'}m_l} 2 m l ′ m l converts the nonrelativistic normalization of the meson fields in the operators to the normalization of Eq. (7) . The vertex is transverse, q μ Γ M μ = 0 q_\mu\Gamma^\mu_M=0 q μ Γ M μ = 0 , and in the Breit frame at s = s ′ s=s' s = s ′ its components are Γ M 0 = 0 \Gamma^0_M=0 Γ M 0 = 0 and Γ M = 2 m l ′ m l i X × q \boldsymbol{\Gamma}_M=2\sqrt{m_{l'}m_l}\,i\boldsymbol{\mathcal{X}}\times\boldsymbol{q} Γ M = 2 m l ′ m l i X × q . Between states of total spin J J J , the Wigner–Eckart theorem yields
⟨ l ′ ; J M ′ ∣ Γ M ∣ l ; J M ⟩ = 2 m l ′ m l i A l ′ l ( J ) ( J × q ) M ′ M , A l ′ l ( J ) = ∑ s c ˉ s Ξ l ′ l ( s ) ( J ) (16) \langle l';JM'|\Gamma_M|l;JM\rangle
=2\sqrt{m_{l'}m_l}\,i\mathcal{A}^{(J)}_{l'l}(\boldsymbol{\mathcal{J}}\times\boldsymbol{q})_{M'M},
\qquad
\mathcal{A}^{(J)}_{l'l}=\sum_s\bar{c}_s\Xi^{(s)}_{l'l}(J) \tag*{(16)} ⟨ l ′ ; J M ′ ∣ Γ M ∣ l ; J M ⟩ = 2 m l ′ m l i A l ′ l ( J ) ( J × q ) M ′ M , A l ′ l ( J ) = s ∑ c ˉ s Ξ l ′ l ( s ) ( J ) (16)
where the spin factors Ξ l ′ l ( s ) ( J ) = ⟨ l ′ ; J J ∣ X z ( s ) ∣ l ; J J ⟩ / J \Xi^{(s)}_{l'l}(J)=\langle l';JJ|\mathcal{X}^{(s)}_z|l;JJ\rangle/J Ξ l ′ l ( s ) ( J ) = ⟨ l ′ ; J J ∣ X z ( s ) ∣ l ; J J ⟩ / J are listed in Table VII .
TABLE VII. Spin factors Ξ l ′ l ( s ) ( J ) \Xi^{(s)}_{l'l}(J) Ξ l ′ l ( s ) ( J ) of the lowest-derivative magnetic operators, Eq. (16) . Bases: ( D ˉ Σ c , D ˉ ∗ Σ c , D ˉ ∗ Σ c ∗ ) (\bar{D}\Sigma_c,\bar{D}^{*}\Sigma_c,\bar{D}^{*}\Sigma_c^{*}) ( D ˉ Σ c , D ˉ ∗ Σ c , D ˉ ∗ Σ c ∗ ) for J = 1 2 J=\frac{1}{2} J = 2 1 and ( D ˉ Σ c ∗ , D ˉ ∗ Σ c , D ˉ ∗ Σ c ∗ ) (\bar{D}\Sigma_c^{*},\bar{D}^{*}\Sigma_c,\bar{D}^{*}\Sigma_c^{*}) ( D ˉ Σ c ∗ , D ˉ ∗ Σ c , D ˉ ∗ Σ c ∗ ) for J = 3 2 J=\frac{3}{2} J = 2 3 . For Ξ ( d ) \Xi^{(d)} Ξ ( d ) the basis is ( D ˉ Λ c , D ˉ ∗ Λ c ) (\bar{D}\Lambda_c,\bar{D}^{*}\Lambda_c) ( D ˉ Λ c , D ˉ ∗ Λ c ) for J = 1 2 J=\frac{1}{2} J = 2 1 and ( D ˉ ∗ Λ c ) (\bar{D}^{*}\Lambda_c) ( D ˉ ∗ Λ c ) for J = 3 2 J=\frac{3}{2} J = 2 3 . The rows (columns) of Ξ ( f ) \Xi^{(f)} Ξ ( f ) and Ξ ( f ′ ) \Xi^{(f')} Ξ ( f ′ ) are the Σ c ( ∗ ) \Sigma_c^{(*)} Σ c ( ∗ ) (Λ c \Lambda_c Λ c ) channels. The rows of Ξ ( h ) \Xi^{(h)} Ξ ( h ) , Ξ ( h ′ ) \Xi^{(h')} Ξ ( h ′ ) and Ξ ( k ) \Xi^{(k)} Ξ ( k ) are the hidden-charm channels, ( η c N , J / ψ N ) (\eta_cN,J/\psi N) ( η c N , J / ψ N ) for J = 1 2 J=\frac{1}{2} J = 2 1 and ( J / ψ N ) (J/\psi N) ( J / ψ N ) for J = 3 2 J=\frac{3}{2} J = 2 3 , and their columns the Σ c ( ∗ ) \Sigma_c^{(*)} Σ c ( ∗ ) or Λ c \Lambda_c Λ c channels. Ξ ( n ) \Xi^{(n)} Ξ ( n ) is diagonal in the hidden-charm channels. The matrices of the transition operators are symmetric in the channel indices.
Table VII.
Electric contact vertex. Parity forbids operators that contain an electric field without a further derivative. The lowest electric operators are therefore
δ L E = e d ( ∇ ⋅ E ) X (17) \delta\mathcal{L}_E=e\,d\,(\boldsymbol{\nabla}\cdot\boldsymbol{E})\mathcal{X} \tag*{(17)} δ L E = e d ( ∇ ⋅ E ) X (17)
where X \mathcal{X} X is a two-hadron operator that is a scalar in the light-quark spins. Since ∇ ⋅ E = v ν ∂ μ F μ ν \boldsymbol{\nabla}\cdot\boldsymbol{E}=v_\nu\partial_\mu F^{\mu\nu} ∇ ⋅ E = v ν ∂ μ F μν in the rest frame of the pair, the corresponding vertex is
Γ E , l ′ l μ = 2 m l ′ m l E l ′ l [ q 2 v μ − q μ ( q ⋅ v ) ] (18) \Gamma^\mu_{E,l'l}=2\sqrt{m_{l'}m_l}\,\mathcal{E}_{l'l}\left[q^2v^\mu-q^\mu(q\cdot v)\right] \tag*{(18)} Γ E , l ′ l μ = 2 m l ′ m l E l ′ l [ q 2 v μ − q μ ( q ⋅ v ) ] (18)
where E \mathcal{E} E is a real symmetric matrix of couplings with dimension E − 5 E^{-5} E − 5 . This vertex is transverse, and in the Breit frame at s = s ′ s=s' s = s ′ its only nonvanishing component is Γ E 0 = − Q 2 2 m l ′ m l E l ′ l \Gamma^0_E=-Q^2\,2\sqrt{m_{l'}m_l}\,\mathcal{E}_{l'l} Γ E 0 = − Q 2 2 m l ′ m l E l ′ l . It therefore changes the slope of G E G_E G E without changing the charge. In the D ˉ ( ∗ ) Σ c ( ∗ ) \bar{D}^{(*)}\Sigma_c^{(*)} D ˉ ( ∗ ) Σ c ( ∗ ) channels, the scalar operators are 1 1 1 and σ q ⋅ S q q \boldsymbol{\sigma}_q\cdot\boldsymbol{S}_{qq} σ q ⋅ S q q , the latter with the eigenvalues − 2 -2 − 2 and + 1 +1 + 1 for light-quark spin 1 2 \frac{1}{2} 2 1 and 3 2 \frac{3}{2} 2 3 , respectively. Consequently, the matrix E \mathcal{E} E has the same HQSS structure as the kernel, with one coupling for each light-quark spin. Including the other pairs of channel types, there are 22 such operators, which contribute to an I = 1 2 I=\frac{1}{2} I = 2 1 state through seven combinations (see Appendix B).
The magnetic contact operators are analogous to the four-nucleon–one-photon counterterms in the pionless effective field theory of the nucleon–nucleon interaction [50 ] . In the present case, however, they are constrained by HQSS and connect channels with different heavy-meson and heavy-baryon spins.
Contact couplings
The contact couplings are neither fixed by the kernel nor constrained by experimental data. We set their central values to zero for the following three reasons. First, neither their magnitudes nor their signs are known, and the value zero favors neither sign and lies at the center of the variation considered below. Second, with this choice the central form factors are determined solely by the one-body currents and the gauged kernel, which are fixed by the strong-interaction model, the charges and the constituent moments. Third, only an order-of-magnitude estimate of these couplings is available. The anomalous π ρ γ \pi\rho\gamma π ρ γ and π ω γ \pi\omega\gamma π ω γ exchange currents yield magnetic couplings of about 10 G e V − 4 10\ \mathrm{GeV}^{-4} 10 GeV − 4 in magnitude with undetermined signs, which change the magnetic moments by at most 0.07 μ N 0.07\ \mu_N 0.07 μ N (see Appendix B). This change is smaller than the variation adopted in this work.
The range of the magnetic couplings is estimated by dimensional analysis, ∣ c ˉ ∣ ∼ ∣ V ∣ / Λ 2 |\bar{c}| \sim|V|/\Lambda^2 ∣ c ˉ ∣ ∼ ∣ V ∣/ Λ 2 . With the diagonal kernel of the dominant channels, ∣ V ∣ = 16 |V| = 16 ∣ V ∣ = 16 to 29 G e V − 2 29\ \mathrm{GeV}^{-2} 29 GeV − 2 in the normalization of the operators, and Λ = 0.7 \Lambda= 0.7 Λ = 0.7 to 1 G e V 1\ \mathrm{GeV} 1 GeV , this estimate gives values between 15 15 15 and 60 G e V − 4 60\ \mathrm{GeV}^{-4} 60 GeV − 4 . We therefore adopt the bounds
∣ c ˉ s ∣ ≤ C s , C s = { 30 G e V − 4 s = a , b , c , d , f , f ′ , 6 G e V − 4 s = h , h ′ , k , n , (19) |\bar{c}_s| \le C_s,\qquad C_s =
\begin{cases}
30\ \mathrm{GeV}^{-4} & s=a,b,c,d,f,f',\\
6\ \mathrm{GeV}^{-4} & s=h,h',k,n,
\end{cases} \tag*{(19)} ∣ c ˉ s ∣ ≤ C s , C s = { 30 GeV − 4 6 GeV − 4 s = a , b , c , d , f , f ′ , s = h , h ′ , k , n , (19)
where the first six combinations act between the D ˉ ( ∗ ) Σ c ( ∗ ) \bar{D}^{(*)}\Sigma_c^{(*)} D ˉ ( ∗ ) Σ c ( ∗ ) and D ˉ ( ∗ ) Λ c \bar{D}^{(*)}\Lambda_c D ˉ ( ∗ ) Λ c channels and the last four involve the hidden-charm channels. For the transitions to the hidden-charm channels, the size is reduced by the factor m ρ 2 / ∣ p D ∗ 2 − m D ∗ 2 ∣ ≃ 0.2 m_\rho^2/|p_{D^*}^2-m_{D^*}^2| \simeq0.2 m ρ 2 /∣ p D ∗ 2 − m D ∗ 2 ∣ ≃ 0.2 of the D ∗ D^* D ∗ exchange, and the same bound is used for the diagonal combination n ˉ \bar{n} n ˉ of these channels. Since the contact terms are linear in the couplings,
G M ( 0 ) = G M ( 0 ) ∣ c ˉ = 0 + ∑ s K s c ˉ s G_M(0) = G_M(0)|_{\bar{c}=0}+\sum_s K_s\bar{c}_s G M ( 0 ) = G M ( 0 ) ∣ c ˉ = 0 + s ∑ K s c ˉ s
with complex coefficients K s K_s K s . The maximal shifts of the real part are
δ b o x = ∑ s ∣ Re K s ∣ C s , δ e l l = [ ∑ s ( Re K s ) 2 C s 2 ] 1 / 2 (20) \delta_{\mathrm{box}}=\sum_s|\operatorname{Re}K_s|C_s,\qquad
\delta_{\mathrm{ell}}=\left[\sum_s(\operatorname{Re}K_s)^2C_s^2\right]^{1/2} \tag*{(20)} δ box = s ∑ ∣ Re K s ∣ C s , δ ell = [ s ∑ ( Re K s ) 2 C s 2 ] 1/2 (20)
for independently bounded couplings and for couplings inside the ellipsoid ∑ s ( c ˉ s / C s ) 2 ≤ 1 \sum_s(\bar{c}_s/C_s)^2\le1 ∑ s ( c ˉ s / C s ) 2 ≤ 1 , respectively. The ranges of the magnetic moments in Sec. IV are obtained with δ b o x \delta_{\mathrm{box}} δ box .
The contribution of the contact terms is ψ T A ~ ψ \psi^\mathrm{T}\tilde{A}\psi ψ T A ~ ψ with ψ l = g l G l ( z R ) \psi_l=g_lG_l(z_R) ψ l = g l G l ( z R ) and A ~ l ′ l = 2 m l ′ m l A l ′ l \tilde{A}_{l'l}=2\sqrt{m_{l'}m_l}A_{l'l} A ~ l ′ l = 2 m l ′ m l A l ′ l , where G l ( z R ) G_l(z_R) G l ( z R ) depends on the subtraction constants. A shift G → G + C G\to G+C G → G + C by a constant diagonal matrix C C C , compensated in the kernel by V − 1 → V − 1 + C V^{-1}\to V^{-1}+C V − 1 → V − 1 + C , leaves T T T , the couplings g l g_l g l , the one-body loops and Z Z Z unchanged. The contact term, however, remains unchanged only if
A ~ → ( 1 + V C ) − 1 A ~ ( 1 + C V ) − 1 at s = z R . (21) \tilde{A}\to(1+VC)^{-1}\tilde{A}(1+CV)^{-1}\quad\text{at }\sqrt{s}=z_R. \tag*{(21)} A ~ → ( 1 + V C ) − 1 A ~ ( 1 + C V ) − 1 at s = z R . (21)
Hence, the couplings c ˉ s \bar{c}_s c ˉ s and the bounds (19) refer to the subtraction constants of Eq. (7) , and only the complete current is independent of the regularization scheme.
The electric couplings are not determined by the model either, and they are not included in the form factors. Inserted between the amplitudes ψ l \psi_l ψ l , the vertex (18) shifts the charge radius by
δ ⟨ r 2 ⟩ E = 12 ∑ l ′ , l ψ l ′ m l ′ m l E l ′ l ψ l , (22) \delta\langle r^2\rangle_E=12\sum_{l',l}\psi_{l'}\sqrt{m_{l'}m_l}\,\mathcal{E}_{l'l}\psi_l, \tag*{(22)} δ ⟨ r 2 ⟩ E = 12 l ′ , l ∑ ψ l ′ m l ′ m l E l ′ l ψ l , (22)
whose size is estimated in Sec. III C.
Two further properties support the choice of zero central values. First, the magnetic moments and the charge radii are linear in the contact couplings. Their values at c ˉ s = 0 \bar{c}_s=0 c ˉ s = 0 and E = 0 \mathcal{E}=0 E = 0 are therefore the expectation values for any distribution of the couplings that is symmetric under a change of their sign. Second, the electric and magnetic contact vertices enter the amplitude in the same way, so that under a change of the subtraction constants the electric couplings transform as A ~ \tilde{A} A ~ does in Eq. (21) . This transformation is linear and homogeneous, and zero is the only value of the couplings that is the same for all subtraction constants related in this way. The central values are therefore independent of the subtraction constants within this class, whereas a nonzero central value would refer to one particular choice.
Form factors at the pole
Attaching the photon to the scattering series yields three classes of diagrams with a double pole, as shown in Figure 1 . Diagrams in which the photon couples to an external leg have only a single pole and therefore do not contribute to Eq. (3) [34 , 49 ] . In channel space, the photon-coupled amplitude reads
Figure 1. The three classes of diagrams of the photon-coupled amplitude T γ μ T^\mu_\gamma T γ μ with a double pole of the state: photon attached to the meson line (T γ 1 T_{\gamma1} T γ 1 ), to the baryon line (T γ 2 T_{\gamma2} T γ 2 ) and to the four-hadron vertex (T γ 3 T_{\gamma3} T γ 3 ). The shaded ellipses are the amplitudes T ( s ) T(\sqrt{s}) T ( s ) (left) and T ( s ′ ) T(\sqrt{s'}) T ( s ′ ) (right), dashed (solid) lines are mesons (baryons), and the wavy line is the incoming virtual photon or another probe current. P P P and P ′ = P + q P'=P+q P ′ = P + q are the total momenta before and after the absorption of q q q , with s = P 2 s=P^2 s = P 2 and s ′ = P ′ 2 s'=P'^2 s ′ = P ′2 . The loops of T γ 1 , 2 T_{\gamma1,2} T γ 1 , 2 are matrices in channel space, because the magnetic transitions D ˉ ∗ → D ˉ γ \bar{D}^{*}\to\bar{D}\gamma D ˉ ∗ → D ˉ γ , Σ c ∗ → Σ c γ \Sigma_c^{*}\to\Sigma_c\gamma Σ c ∗ → Σ c γ and Σ c ( ∗ ) → Λ c γ \Sigma_c^{(*)}\to\Lambda_c\gamma Σ c ( ∗ ) → Λ c γ connect different channels. The vertex of T γ 3 T_{\gamma3} T γ 3 is the sum of the gauged kernel Γ min μ \Gamma^\mu_{\min} Γ m i n μ and the transverse contact terms Γ t r μ \Gamma^\mu_{\mathrm{tr}} Γ tr μ .
T γ μ ( s ′ , s ) = T ( s ′ ) Π μ ( P , P ′ ) T ( s ) , T^\mu_\gamma(\sqrt{s'},\sqrt{s})=T(\sqrt{s'})\Pi^\mu(P,P')T(\sqrt{s}), T γ μ ( s ′ , s ) = T ( s ′ ) Π μ ( P , P ′ ) T ( s ) ,
Π l ′ l μ = D M ; l ′ l μ + D B ; l ′ l μ + G l ′ ( s ′ ) Γ l ′ l μ G l ( s ) , (23) \Pi^\mu_{l'l}=D^\mu_{M;l'l}+D^\mu_{B;l'l}+G_{l'}(\sqrt{s'})\Gamma^\mu_{l'l}G_l(\sqrt{s}), \tag*{(23)} Π l ′ l μ = D M ; l ′ l μ + D B ; l ′ l μ + G l ′ ( s ′ ) Γ l ′ l μ G l ( s ) , (23)
where D M μ D^\mu_M D M μ and D B μ D^\mu_B D B μ denote the triangle loops with the photon attached to the meson and baryon lines, respectively [35 ] , and Γ μ \Gamma^\mu Γ μ is the vertex of Eq. (12) . Inserting the pole terms of T T T and comparing with Eq. (3) , we obtain
⟨ P c ( P ′ ) ∣ J μ ∣ P c ( P ) ⟩ = ∑ l ′ , l g l ′ Π l ′ l μ ( P , P ′ ) g l ∣ s = s ′ = z R , (24) \langle P_c(P')|J^\mu|P_c(P)\rangle
=\left.\sum_{l',l}g_{l'}\Pi^\mu_{l'l}(P,P')g_l\right|_{\sqrt{s}=\sqrt{s'}=z_R}, \tag*{(24)} ⟨ P c ( P ′ ) ∣ J μ ∣ P c ( P )⟩ = l ′ , l ∑ g l ′ Π l ′ l μ ( P , P ′ ) g l s = s ′ = z R , (24)
where the loops are continued to the Riemann sheet of the pole. Note that the couplings enter without complex conjugation. At the pole, the Breit frame is defined by P = ( E , − q / 2 ) P=(E,-q/2) P = ( E , − q /2 ) and P ′ = ( E , q / 2 ) P'=(E,q/2) P ′ = ( E , q /2 ) with E = E= E = z R 2 + Q 2 / 4 \sqrt{z_R^2+Q^2/4} z R 2 + Q 2 /4 . The form factor F P F_P F P of a probe with the charges ( Q M , l P , Q B , l P ) (Q^P_{M,l},Q^P_{B,l}) ( Q M , l P , Q B , l P ) and the total charge Q H P Q^P_H Q H P , together with the magnetic form factor, is given by
F P = ∑ l g l 2 [ Q M , l P E l M + Q B , l P E l B ] + Q H P E z R Z , F_P=\sum_l g_l^2\left[Q^P_{M,l}E^M_l+Q^P_{B,l}E^B_l\right]+Q^P_H\frac{E}{z_R}Z, F P = l ∑ g l 2 [ Q M , l P E l M + Q B , l P E l B ] + Q H P z R E Z ,
G M = ∑ l ′ , l g l ′ [ M l ′ l M M l ′ l M + M l ′ l B M l ′ l B ] g l G_M=\sum_{l',l}g_{l'}\left[\mathcal{M}^M_{l'l}\mathcal{M}^M_{l'l}+\mathcal{M}^B_{l'l}\mathcal{M}^B_{l'l}\right]g_l G M = l ′ , l ∑ g l ′ [ M l ′ l M M l ′ l M + M l ′ l B M l ′ l B ] g l
{#eq:25}
+2M_p\mu_N\sum_{l',l}\psi_{l'},2\sqrt{m_{l'}m_l}\mathcal{I}^{(J)}_{l'l},\psi_l.
where E l M , B ( Q 2 ) E^{M,B}_l(Q^2) E l M , B ( Q 2 ) and M l ′ l M , B ( Q 2 ) \mathcal{M}^{M,B}_{l'l}(Q^2) M l ′ l M , B ( Q 2 ) are the electric and magnetic loop functions of Appendix A, and G E = F P G_E=F_P G E = F P for the electric charges. The matrices M M , B \mathcal{M}^{M,B} M M , B are given in units of μ N \mu_N μ N . The term proportional to Z Z Z represents the contribution of the gauged kernel (13) . In the nonrelativistic limit, the meson and the baryon are located at the fractions
ξ M = M l m l + M l , ξ B = m l m l + M l (25) \xi_M=\frac{M_l}{m_l+M_l},\qquad\xi_B=\frac{m_l}{m_l+M_l} \tag*{(25)} ξ M = m l + M l M l , ξ B = m l + M l m l (25)
of the relative distance from the center of mass. The one-body loops satisfy the Ward identity q μ D M , B ; l l μ = Q M , B ; l [ G l ( s ) − G l ( s ′ ) ] q_\mu D^{\mu}_{M,B;ll}=Q_{M,B;l}[G_l(\sqrt{s})-G_l(\sqrt{s'})] q μ D M , B ; l l μ = Q M , B ; l [ G l ( s ) − G l ( s ′ )] on every Riemann sheet. Combining this identity with q μ Γ min μ q_\mu\Gamma^\mu_{\min} q μ Γ m i n μ and the scattering equation, we obtain
q μ T γ μ ( s ′ , s ) = − Q H [ T ( s ′ ) G ( s ′ ) V ( s ) − V ( s ′ ) G ( s ) T ( s ) ] (26) q_\mu T^\mu_\gamma(\sqrt{s'},\sqrt{s})=-Q_H\left[T(\sqrt{s'})G(\sqrt{s'})V(\sqrt{s})-V(\sqrt{s'})G(\sqrt{s})T(\sqrt{s})\right] \tag*{(26)} q μ T γ μ ( s ′ , s ) = − Q H [ T ( s ′ ) G ( s ′ ) V ( s ) − V ( s ′ ) G ( s ) T ( s ) ] (26)
which contains only single poles. The coefficient of the double pole is therefore conserved. At Q 2 = 0 Q^2=0 Q 2 = 0 , where E l M ( 0 ) = E l B ( 0 ) = − d G l / d s E^M_l(0)=E^B_l(0)=-dG_l/d\sqrt{s} E l M ( 0 ) = E l B ( 0 ) = − d G l / d s , this leads to
G E ( 0 ) = Q H [ ∑ l X l + Z ] = Q H . (27) G_E(0)=Q_H\left[\sum_l X_l+Z\right]=Q_H. \tag*{(27)} G E ( 0 ) = Q H [ l ∑ X l + Z ] = Q H . (27)
The charge normalization is thus equivalent to the compositeness sum rule (10) , in which the gauged kernel provides the term Z Z Z . Numerically, Eq. (27) is satisfied to better than 10 − 5 10^{-5} 1 0 − 5 at all poles. This result extends the relation between the normalization of the form factor and the compositeness established in Refs. [35 , 39 ] to a charged state. While this identity fixes the charge, it imposes no constraint on the contact currents of Sec. II C 3 or on the Q 2 Q^2 Q 2 dependence of the form factors.
The residue is evaluated on the Riemann sheet of the pole. For an open channel, the photon loops on the second sheet are obtained by adding the discontinuity across the unitarity cut (see Appendix A). In addition, these loops have a triangle singularity at spacelike momentum transfer. We define the form factor as the branch of the residue that is analytic in Q 2 Q^2 Q 2 and contains the point Q 2 = 0 Q^2=0 Q 2 = 0 (see Sec. VI).
In Figure 2 , we show the form factors of the three observed states in set II for Q 2 ≤ 1 G e V 2 Q^2\leq1\,\mathrm{GeV}^2 Q 2 ≤ 1 GeV 2 . At Q 2 = 1 G e V 2 Q^2=1\,\mathrm{GeV}^2 Q 2 = 1 GeV 2 , the electric form factors are 0.71 + 0.09 i 0.71+0.09i 0.71 + 0.09 i , 0.83 + 0.06 i 0.83+0.06i 0.83 + 0.06 i and 0.56 + 0.06 i 0.56+0.06i 0.56 + 0.06 i for P c ( 4312 ) P_c(4312) P c ( 4312 ) , P c ( 4440 ) P_c(4440) P c ( 4440 ) and P c ( 4457 ) P_c(4457) P c ( 4457 ) , respectively. The form factor of the most weakly bound state, P c ( 4457 ) P_c(4457) P c ( 4457 ) , decreases most rapidly. The imaginary parts of G E G_E G E vanish at Q 2 = 0 Q^2=0 Q 2 = 0 and remain below 0.10.
Figure 2. Electric (top) and magnetic (bottom) form factors of the three observed states at the poles of set II for pointlike constituents. Thick solid (dashed) lines are real (imaginary) parts. Thin gray lines are the zero-range form of Eq. (29) . Dotted lines are the real part of G E G_E G E with quark-level vector-meson dominance for the constituents (Sec. III C). The insets show the region of low Q 2 Q^2 Q 2 . The bands in the lower panels are Re G M ± δ b o x \operatorname{Re}G_M \pm\delta_{\mathrm{box}} Re G M ± δ box , where δ b o x \delta_{\mathrm{box}} δ box is the maximal shift from the magnetic contact couplings, Eq. (20) , around Re G M \operatorname{Re}G_M Re G M .
The Q 2 Q^2 Q 2 dependence is characteristic of a weakly bound two-body system. We denote by m m m and M M M the meson and baryon masses of the dominant channel, by μ r e d = m M / ( m + M ) \mu_{\mathrm{red}}=mM/(m+M) μ red = m M / ( m + M ) its reduced mass, and by
B c = m + M − z R = B + i 2 Γ R , γ = 2 μ r e d B c (28) B_c=m+M-z_R=B+\frac{i}{2}\Gamma_R,\qquad\gamma=\sqrt{2\mu_{\mathrm{red}}B_c} \tag*{(28)} B c = m + M − z R = B + 2 i Γ R , γ = 2 μ red B c (28)
the complex binding energy and the binding momentum. For a zero-range wave function ψ ∝ e − γ x / x \psi\propto e^{-\gamma x}/x ψ ∝ e − γ x / x , the form factor of a constituent located at the fraction ξ \xi ξ of the relative distance x x x from the center of mass is
F Z R ( Q ; ξ ) = 2 γ ξ Q arctan ξ Q 2 γ . (29) F_{\mathrm{ZR}}(Q;\xi)=\frac{2\gamma}{\xi Q}\arctan\frac{\xi Q}{2\gamma}. \tag*{(29)} F ZR ( Q ; ξ ) = ξ Q 2 γ arctan 2 γ ξ Q . (29)
The thin lines in Figure 2 show 5 3 F Z R ( Q ; ξ B ) − 2 3 F Z R ( Q ; ξ M ) \frac{5}{3}F_{\mathrm{ZR}}(Q;\xi_B)-\frac{2}{3}F_{\mathrm{ZR}}(Q;\xi_M) 3 5 F ZR ( Q ; ξ B ) − 3 2 F ZR ( Q ; ξ M ) for G E G_E G E and the corresponding combination of the baryon-line and meson-line contributions to G M ( 0 ) G_M(0) G M ( 0 ) for G M G_M G M . These curves contain no parameter other than the pole position. The full loop results agree with them within 0.02 for G E G_E G E and within 0.05 for G M G_M G M up to Q 2 = 1 G e V 2 Q^2=1\,\mathrm{GeV}^2 Q 2 = 1 GeV 2 . For pointlike constituents, the form factors are therefore essentially determined by the complex binding energy, the masses and charges of the constituents and the static moments.
Mean squared distance and the zero-range relation
The mean squared distance between the constituents of the dominant channel can be obtained from either the baryon-number or the anticharm radius,
⟨ x 2 ⟩ = ⟨ r 2 ⟩ B ξ B 2 = ⟨ r 2 ⟩ c ˉ ξ M 2 . (30) \langle x^2\rangle=\frac{\langle r^2\rangle_B}{\xi_B^2}=\frac{\langle r^2\rangle_{\bar c}}{\xi_M^2}. \tag*{(30)} ⟨ x 2 ⟩ = ξ B 2 ⟨ r 2 ⟩ B = ξ M 2 ⟨ r 2 ⟩ c ˉ . (30)
The two determinations agree within 0.5% for all states. The values of ⟨ x 2 ⟩ \langle x^2\rangle ⟨ x 2 ⟩ and the radii are listed in Table III.
For a zero-range wave function, one obtains
⟨ x 2 ⟩ Z R = 1 2 γ 2 = 1 4 μ r e d B c . (31) \langle x^2\rangle_{\mathrm{ZR}}=\frac{1}{2\gamma^2}=\frac{1}{4\mu_{\mathrm{red}}B_c}. \tag*{(31)} ⟨ x 2 ⟩ ZR = 2 γ 2 1 = 4 μ red B c 1 . (31)
The computed values exceed Eq. (31) by 3 to 11% in modulus, whereas the phases agree within 4° (see Figure 3 ). This excess can be understood as follows. The asymptotic normalization of the wave function of a shallow state is determined by the residue at the pole, which depends on the effective range r e r_e r e of the amplitude [51 ] . If the mean squared distance is dominated by the tail of the wave function, one finds
Figure 3. Left: modulus of the ratio of ⟨ x 2 ⟩ \langle x^2\rangle ⟨ x 2 ⟩ to the zero-range value 1 / ( 4 μ r e d B c ) 1/(4\mu_{\mathrm{red}}B_c) 1/ ( 4 μ red B c ) for the seven states in the two parameter sets, as a function of the modulus of the binding momentum. The solid line is the effective-range factor of Eq. (33)} for real γ \gamma γ with r e = 0.095 f m r_e=0.095\,\mathrm{fm} r e = 0.095 fm , the dotted line the same for r e = 1 / m ρ r_e=1/m_\rho r e = 1/ m ρ , and the dashed line the single-channel D ˉ ∗ Σ c ∗ \bar{D}^{*}\Sigma_c^{*} D ˉ ∗ Σ c ∗ bound state with varied interaction strength. Middle: phase of ⟨ x 2 ⟩ \langle x^2\rangle ⟨ x 2 ⟩ compared with the phase of 1 / B c 1/B_c 1/ B c . Right: the complex ratio w = 4 μ r e d B c ( 1 − γ r e ) ⟨ x 2 ⟩ w=4\mu_{\mathrm{red}}B_c(1-\gamma r_e)\langle x^2\rangle w = 4 μ red B c ( 1 − γ r e ) ⟨ x 2 ⟩ with the complex effective range of the full amplitude (large symbols) and with the real range of the decoupled channel (small gray symbols).
⟨ x 2 ⟩ ≃ 1 4 μ r e d B c 1 1 − γ r e . (32) \langle x^2\rangle\simeq\frac{1}{4\mu_{\mathrm{red}}B_c}\frac{1}{1-\gamma r_e}. \tag*{(32)} ⟨ x 2 ⟩ ≃ 4 μ red B c 1 1 − γ r e 1 . (32)
The effective range is defined through the expansion k cot δ l = − 1 / a 0 + 1 2 r e k 2 k\cot\delta_l=-1/a_0+\frac{1}{2}r_e k^2 k cot δ l = − 1/ a 0 + 2 1 r e k 2 of the amplitude f l = − M l T l l / ( 4 π s ) = 1 / ( k cot δ l − i k ) f_l=-M_lT_{ll}/(4\pi\sqrt{s})=1/(k\cot\delta_l-ik) f l = − M l T l l / ( 4 π s ) = 1/ ( k cot δ l − ik ) of the dominant channel. In the coupled-channel system, r e r_e r e is complex because the lower channels are open. For the full amplitude, we obtain Re r e = 0.081 \operatorname{Re}r_e=0.081 Re r e = 0.081 –0.121 f m 0.121\,\mathrm{fm} 0.121 fm and ∣ Im r e ∣ ≤ 0.017 f m |\operatorname{Im}r_e|\leq0.017\,\mathrm{fm} ∣ Im r e ∣ ≤ 0.017 fm at the fourteen poles. When the other channels are switched off, the effective range becomes real, r e = 0.091 r_e=0.091 r e = 0.091 –0.098 f m 0.098\,\mathrm{fm} 0.098 fm . The expansion up to order k 2 r e k^2r_e k 2 r e reproduces the binding momenta of the poles within 0.6 % 0.6\% 0.6% . To test Eq. (32) as a complex relation, we define the ratio w = 4 μ r e d B c ( 1 − γ r e ) ⟨ x 2 ⟩ w=4\mu_{\mathrm{red}}B_c(1-\gamma r_e)\langle x^2\rangle w = 4 μ red B c ( 1 − γ r e ) ⟨ x 2 ⟩ , which equals unity if Eq. (32) holds. With the effective range of the full amplitude, we find ∣ w − 1 ∣ ≤ 0.018 |w-1|\leq0.018 ∣ w − 1∣ ≤ 0.018 for the fourteen cases of Table III, which cover the binding momenta ∣ γ ∣ = 0.07 |\gamma|=0.07 ∣ γ ∣ = 0.07 –0.23 G e V 0.23\,\mathrm{GeV} 0.23 GeV (see Figure 3 ). With the real effective range of the decoupled channel, ∣ w − 1 ∣ ≤ 0.027 |w-1|\leq0.027 ∣ w − 1∣ ≤ 0.027 . At the level of 2 % 2\% 2% , the loop calculation with pointlike vertices is therefore essentially determined by the binding energy, the width and an effective range of 0.1 f m 0.1\,\mathrm{fm} 0.1 fm . The other channels contribute mainly through the width in B c B_c B c and through the complex effective range. Equation (32) is the extension of the shallow-bound-state relation to a complex binding energy. As shown by the dashed line in Figure 3 , a single-channel bound state with varied interaction strength follows the same factor.
Table 3. Complex mean squared distance ⟨ x 2 ⟩ \langle x^2\rangle ⟨ x 2 ⟩ between the constituents of the dominant channel, Eq. (30) , and its zero-range value 1 / ( 4 μ r e d B c ) 1/(4\mu_{\mathrm{red}}B_c) 1/ ( 4 μ red B c ) in the two parameter sets, and mean squared radii of the charge, of the magnetization and of the charm current in set II. All values are in f m 2 \mathrm{fm}^2 fm 2 , for pointlike constituents and without contact currents. The charm radius is not normalized, because F C ( 0 ) = 0 F_C(0)=0 F C ( 0 ) = 0 .
The effective range of the model is determined by the energy dependence of the kernel and by the regular part of the loop function. It is unrelated to the range 1 / m ρ = 0.25 f m 1/m_\rho=0.25\,\mathrm{fm} 1/ m ρ = 0.25 fm of the vector-meson exchange, which is not resolved by the present kernel. The values in Table III should therefore be understood as results for the effective range of this model. The sensitivity to r e r_e r e can be illustrated by two examples. For r e = 1 / m ρ r_e=1/m_\rho r e = 1/ m ρ , the modulus of the factor in Eq. (32) is 1.22 1.22 1.22 , 1.40 1.40 1.40 and 1.15 1.15 1.15 for P c ( 4312 ) P_c(4312) P c ( 4312 ) , P c ( 4440 ) P_c(4440) P c ( 4440 ) and P c ( 4457 ) P_c(4457) P c ( 4457 ) in set II, compared with the computed ratios 1.07 1.07 1.07 , 1.10 1.10 1.10 and 1.04 1.04 1.04 . For the deuteron, with γ = 0.2316 f m − 1 \gamma=0.2316\,\mathrm{fm}^{-1} γ = 0.2316 fm − 1 and the triplet effective range 1.753 f m 1.753\,\mathrm{fm} 1.753 fm , the factor is 1.68 1.68 1.68 , while the experimental structure radius of 1.975 f m 1.975\,\mathrm{fm} 1.975 fm [52 ] exceeds the zero-range value by a factor of 1.67 1.67 1.67 in the mean squared radius. These examples, however, do not provide a bound. For γ = u + i v \gamma=u+iv γ = u + i v and real r e r_e r e , the factor exceeds unity in modulus only for 0 < r e < 2 u / ∣ γ ∣ 2 0<r_e<2u/|\gamma|^2 0 < r e < 2 u /∣ γ ∣ 2 , whereas a negative or complex effective range can reduce the modulus.
Since the radii of resonances are complex, their interpretation is not unique, and several interpretations have been discussed in the literature [53 , 54 , 55 , 56 ] . Here we do not adopt a particular interpretation and present both the real and the imaginary parts. In Eq. (31) , the modulus and the phase have a simple meaning, ⟨ x 2 ⟩ ≃ 1 / ( 4 μ r e d ∣ B c ∣ ) \langle x^{2}\rangle\simeq1/(4\mu_{\mathrm{red}}|B_{c}|) ⟨ x 2 ⟩ ≃ 1/ ( 4 μ red ∣ B c ∣ ) and arg ⟨ x 2 ⟩ ≃ − arctan ( Γ R / 2 B ) \arg\langle x^{2}\rangle\simeq-\arctan(\Gamma_{R}/2B) arg ⟨ x 2 ⟩ ≃ − arctan ( Γ R /2 B ) . For Γ R ≪ 2 B \Gamma_{R} \ll2B Γ R ≪ 2 B , the imaginary part is small, and the real part can be interpreted as a mean squared distance. For the states considered here, however, Γ R / 2 B = 0.3 \Gamma_{R}/2B=0.3 Γ R /2 B = 0.3 to 1.8 1.8 1.8 , except for the 3 2 − D ˉ ∗ Σ c \frac{3}{2}^{-}\bar{D}^{*}\Sigma_{c} 2 3 − D ˉ ∗ Σ c state in set I. The modulus defines a characteristic length, ∣ ⟨ x 2 ⟩ ∣ 1 / 2 = 0.98 |\langle x^{2}\rangle|^{1/2}=0.98 ∣ ⟨ x 2 ⟩ ∣ 1/2 = 0.98 , 0.63 and 1.36 fm for the three observed states in set II.
Charge radii
For pointlike constituents, the charge radius is given by the charge-weighted sum of the radii of the two constituents,
⟨ r 2 ⟩ E ≃ [ Q M ξ M 2 + Q B ξ B 2 ] ⟨ x 2 ⟩ , (33) \langle r^{2}\rangle_{E}\simeq\left[Q_{M}\xi_{M}^{2}+Q_{B}\xi_{B}^{2}\right]\langle x^{2}\rangle, \tag*{(33)} ⟨ r 2 ⟩ E ≃ [ Q M ξ M 2 + Q B ξ B 2 ] ⟨ x 2 ⟩ , (33)
with Q M = − 2 3 Q_{M}=-\frac{2}{3} Q M = − 3 2 and Q B = 5 3 Q_{B}=\frac{5}{3} Q B = 3 5 . The factor in brackets is 0.096 for D ˉ Σ c \bar{D}\Sigma_{c} D ˉ Σ c and 0.136 for D ˉ ∗ Σ c \bar{D}^{*}\Sigma_{c} D ˉ ∗ Σ c . Since the meson and the baryon carry charges of opposite sign and the heavier baryon is located closer to the center of mass, the two terms largely cancel. The values in Table III satisfy Eq. (33) within 1%. The charm radius ⟨ r 2 ⟩ C \langle r^{2}\rangle_{C} ⟨ r 2 ⟩ C is negative because the anticharm quark resides in the lighter constituent. Two further contributions to the slope of G E G_{E} G E are not included in these values.
The first contribution arises from the electric contact currents of Eq. (17) . Their size can only be estimated by dimensional analysis, ∣ E ∣ ∼ ∣ V ∣ / Λ 3 |\mathcal{E}|\sim|V|/\Lambda^{3} ∣ E ∣ ∼ ∣ V ∣/ Λ 3 , which gives values between 15 and 90 GeV− 5 ^{-5} − 5 . Taking into account only the diagonal entry of the dominant channel, Eq. (22) gives ∣ δ ⟨ r 2 ⟩ E ∣ = 0.014 f m 2 |\delta\langle r^{2}\rangle_{E}|=0.014\,\mathrm{fm}^{2} ∣ δ ⟨ r 2 ⟩ E ∣ = 0.014 fm 2 for E = 30 G e V − 5 \mathcal{E}=30\,\mathrm{GeV}^{-5} E = 30 GeV − 5 and 0.04 f m 2 0.04\,\mathrm{fm}^{2} 0.04 fm 2 for 90 G e V − 5 90\,\mathrm{GeV}^{-5} 90 GeV − 5 for each of the three observed states in set II. These shifts correspond to 16, 28 and 5% of the moduli of the values in Table III for 30 G e V − 5 30\,\mathrm{GeV}^{-5} 30 GeV − 5 , and to 47, 83 and 16% for 90 G e V − 5 90\,\mathrm{GeV}^{-5} 90 GeV − 5 . The effect is largest for P c ( 4440 ) P_{c}(4440) P c ( 4440 ) , whose molecular charge radius is small owing to the cancellation in Eq. (33) . The distance ⟨ x 2 ⟩ \langle x^{2}\rangle ⟨ x 2 ⟩ , which is defined through the baryon-number and anticharm probes, is not affected by these terms.
The second contribution arises from the finite size of the constituents, which adds their charge-weighted radii, including the radius of the neutral D ˉ ( ∗ ) 0 \bar{D}^{(*)0} D ˉ ( ∗ ) 0 . We estimate this effect by vector-meson dominance at the quark level, which respects the quark content of the constituents: the light quarks couple to the photon through the ρ \rho ρ and ω \omega ω poles, and the charm quarks through the J / ψ J/\psi J / ψ pole. Since the charm contributions of the c c c and the c ˉ \bar{c} c ˉ cancel at Q 2 = 0 Q^{2}=0 Q 2 = 0 and the net charge of the light quarks of the P c + P_{c}^{+} P c + is + 1 +1 + 1 in every channel, the correction is the same for all states, δ ⟨ r 2 ⟩ E = 6 / m ρ 2 = 0.39 f m 2 \delta\langle r^{2}\rangle_{E}=6/m_{\rho}^{2}=0.39\,\mathrm{fm}^{2} δ ⟨ r 2 ⟩ E = 6/ m ρ 2 = 0.39 fm 2 . An explicit evaluation with these form factors in all loops gives ⟨ r 2 ⟩ E = 0.47 − 0.04 i \langle r^{2}\rangle_{E}=0.47-0.04i ⟨ r 2 ⟩ E = 0.47 − 0.04 i , 0.43 − 0.02 i 0.43-0.02i 0.43 − 0.02 i and 0.63 − 0.08 i f m 2 0.63-0.08i\,\mathrm{fm}^{2} 0.63 − 0.08 i fm 2 for the three observed states in set II. At Q 2 = 0.5 G e V 2 Q^{2}=0.5\,\mathrm{GeV}^{2} Q 2 = 0.5 GeV 2 , the electric form factors are reduced to 0.48, 0.51 and 0.40 from 0.83, 0.91 and 0.69 (see Figure 2 ). The same charge form factor is obtained if only the transverse part of each vertex is dressed, so that the Ward identities remain satisfied. Since lattice QCD calculations find charge radii of the charmed baryons smaller than that of the proton [46 ] , this estimate may overestimate the effect. Nevertheless, it shows that the finite size of the constituents can exceed the molecular contribution to the charge radius, 0.05 to 0.24 fm2 ^{2} 2 , and that it dominates the Q 2 Q^{2} Q 2 dependence of G E G_{E} G E above about 0.3 G e V 2 0.3\,\mathrm{GeV}^{2} 0.3 GeV 2 . The distance ⟨ x 2 ⟩ \langle x^{2}\rangle ⟨ x 2 ⟩ between the centers of the constituents is not affected.
Magnetic moments
Additive moments and loop results
The magnetic moment μ = G M ( 0 ) μ N \mu=G_{M}(0)\,\mu_{N} μ = G M ( 0 ) μ N is complex at a resonance pole, and our statements on signs and ranges refer to its real part. For an S S S -wave meson-baryon state, the additive moment paid is the diagonal element of M ( 1 ) \mathcal{M}^{(1)} M ( 1 ) in Eq. (11) . With the isospin-averaged moments of Table II, it equals μ Σ c \mu_{\Sigma_{c}} μ Σ c for the D ˉ Σ c \bar{D}\Sigma_{c} D ˉ Σ c state, 2 3 μ D ˉ ∗ − 1 3 μ Σ c = − 0.85 μ N \frac{2}{3}\mu_{\bar{D}^{*}}-\frac{1}{3}\mu_{\Sigma_{c}}=-0.85\,\mu_{N} 3 2 μ D ˉ ∗ − 3 1 μ Σ c = − 0.85 μ N for the 1 2 − D ˉ ∗ Σ c \frac{1}{2}^{-}\bar{D}^{*}\Sigma_{c} 2 1 − D ˉ ∗ Σ c state and μ D ˉ ∗ + μ Σ c = 1.31 μ N \mu_{\bar{D}^{*}}+\mu_{\Sigma_{c}}=1.31\,\mu_{N} μ D ˉ ∗ + μ Σ c = 1.31 μ N for the 3 2 − \frac{3}{2}^{-} 2 3 − state (second column of Table IV ), as in the quark-model calculations of Refs. [20 , 24 ] .
Table IV. Magnetic moments in μ N \mu_N μ N . Second column: additive moment of the dominant channel. Third and fourth columns: G M ( 0 ) G_M(0) G M ( 0 ) in the two parameter sets with the constituent quark moments and without contact terms. All other columns refer to the real part in set II. Fifth and sixth columns: light-quark part of the Σ c ( ∗ ) \Sigma_c^{(*)} Σ c ( ∗ ) moments scaled to the lattice results of Ref. [46 ] for the linear and the quadratic extrapolation. Seventh and eighth columns: maximal shifts δ e l l \delta_{\rm ell} δ ell and δ b o x \delta_{\rm box} δ box from the contact couplings, Eq. (20) . Ninth column: excess of the loop normalization over the additive value. Last column: range of Re μ \operatorname{Re}\mu Re μ over these variations (see text).
The full results in the third and fourth columns of Table IV differ from the additive values by 1 to 6% in set I and by 1 to 11% in set II. Two effects contribute to the results. First, the ratio of the diagonal magnetic and electric loops of the dominant channel, which equals unity in the nonrelativistic limit, ranges from 1.02 to 1.07 at the fourteen poles in dimensional regularization. With a three-momentum cutoff that reproduces the subtraction constant at threshold, it is 1.006 to 1.024 for the observed states, so that this effect depends on loop momenta of the order of the hadron masses. Second, the transition moments and the subdominant channels contribute − 0.06 -0.06 − 0.06 to 0 in set I and − 0.07 -0.07 − 0.07 to + 0.17 μ N +0.17\,\mu_{N} + 0.17 μ N in set II, where r ≠ 1 r\ne1 r = 1 couples the D ˉ ( ∗ ) Σ c ( ∗ ) \bar{D}^{(*)}\Sigma_{c}^{(*)} D ˉ ( ∗ ) Σ c ( ∗ ) channels directly. The imaginary parts of G M ( 0 ) G_{M}(0) G M ( 0 ) are − 0.01 -0.01 − 0.01 , − 0.05 -0.05 − 0.05 and − 0.03 -0.03 − 0.03 for the observed states in set II and below 0.22 in modulus otherwise. The additive moments therefore dominate also in the present approach.
The variations of the inputs exceed these loop corrections. We consider four sources as follows.
Constituent moments. In the I = 1 2 I=\frac{1}{2} I = 2 1 average, the lattice results for the Σ c \Sigma_{c} Σ c [46 ] (Table II) lie 36% below the quark-model value for the linear extrapolation in m π 2 m_{\pi}^{2} m π 2 and 13% below for the quadratic one. Scaling the light-quark parts of all Σ c ( ∗ ) \Sigma_{c}^{(*)} Σ c ( ∗ ) moments and transition moments by 0.67 0.67 0.67 and 0.88 0.88 0.88 , with the Σ c \Sigma_c Σ c and the Σ c ∗ \Sigma_c^{*} Σ c ∗ related by HQSS, gives the fifth and sixth columns of Table IV . The meson transition moments extracted from the radiative widths of D ∗ + → D + γ D^{*+}\to D^{+}\gamma D ∗+ → D + γ and J / ψ → η c γ J/\psi\to\eta_c\gamma J / ψ → η c γ are 14 and 20% below the constituent-quark values (Appendix C). A 30% reduction of the meson-line contribution changes μ \mu μ by at most 0.13 μ N 0.13\mu_N 0.13 μ N .
Contact couplings and loop normalization. The maximal shifts δ e l l \delta_{\rm ell} δ ell and δ b o x \delta_{\rm box} δ box for the bounds of Eq. (19) are listed in Table IV for set II. In set I they are 0.10 0.10 0.10 to 0.16 μ N 0.16\mu_N 0.16 μ N and 0.15 0.15 0.15 to 0.30 μ N 0.30\mu_N 0.30 μ N . Both scale linearly with the bounds, and the exchange-current estimate of Appendix B gives at most 0.07 μ N 0.07\mu_N 0.07 μ N . The excess of the diagonal loops over the additive value, 0.07 0.07 0.07 to 0.16 μ N 0.16\mu_N 0.16 μ N in set II, is included as a one-sided variation. It corresponds to a shift of a diagonal contact coupling by 8 to 17 G e V − 4 \mathrm{GeV}^{-4} GeV − 4 for the observed states, within the bounds.
Regularization. In a refit, the common subtraction constant a a a was shifted by ± 0.2 \pm0.2 ± 0.2 and the kernel was multiplied by a factor λ V \lambda_V λ V , with λ V \lambda_V λ V , δ \delta δ and r r r determined from the masses and κ \kappa κ from the widths. This gives λ V = 0.57 \lambda_V=0.57 λ V = 0.57 for a = − 2.43 a=-2.43 a = − 2.43 and λ V = 3.27 \lambda_V=3.27 λ V = 3.27 for a = − 2.03 a=-2.03 a = − 2.03 (χ 2 = 1.6 \chi^2=1.6 χ 2 = 1.6 and 1.9). Without contact terms, the moments of the observed states change by at most 0.024 μ N 0.024\mu_N 0.024 μ N and ∣ 4 μ r e d B c ⟨ x 2 ⟩ ∣ |4\mu_{\rm red}B_c\langle x^2\rangle| ∣4 μ red B c ⟨ x 2 ⟩ ∣ by at most 0.016. At fixed pole positions, these quantities are thus insensitive to the subtraction constant, although the kernel strength changes by a factor of 5.7. The contact terms behave differently, because the amplitudes ψ l \psi_l ψ l of the dominant channels change by factors of 1.0 to 1.6 and 0.4 to 0.6, respectively. For the fixed bounds of Eq. (19) , δ e l l \delta_{\rm ell} δ ell of the observed states changes from 0.16 0.16 0.16 –0.27 μ N 0.27\mu_N 0.27 μ N to 0.36 0.36 0.36 –0.59 μ N 0.59\mu_N 0.59 μ N and 0.05 0.05 0.05 –0.10 μ N 0.10\mu_N 0.10 μ N . For bounds scaled with the kernel strength, as required by the estimate [ c ~ ∼ ∣ V ∣ / λ 2 ] [ \tilde{c} \sim|V|/\lambda^2] [ c ~ ∼ ∣ V ∣/ λ 2 ] , it is 0.20 0.20 0.20 –0.34 μ N 0.34\mu_N 0.34 μ N and 0.18 0.18 0.18 –0.31 μ N 0.31\mu_N 0.31 μ N . The shifts δ e l l \delta_{\rm ell} δ ell and δ b o x \delta_{\rm box} δ box thus represent the dimensional estimate within a given regularization scheme, consistent with Eq. (21) .
Since the constituent moments and the loop normalization shift the moments in one direction only, we do not add the variations in quadrature. The last column of Table IV gives the range of Re μ \operatorname{Re}\mu Re μ , from the smallest modulus obtained with the input variations, reduced by δ b o x \delta_{\rm box} δ box and by the loop excess, to the largest modulus, increased by δ b o x \delta_{\rm box} δ box . With δ e l l \delta_{\rm ell} δ ell , the ranges of the observed states become + 0.8 +0.8 + 0.8 to + 1.9 +1.9 + 1.9 , − 1.2 -1.2 − 1.2 to − 0.3 -0.3 − 0.3 and + 0.3 +0.3 + 0.3 to + 1.7 μ N +1.7\mu_N + 1.7 μ N . These ranges represent model variations rather than statistical confidence intervals. All real parts retain their signs. For the two D ˉ ∗ Σ c \bar{D}^{*}\Sigma_c D ˉ ∗ Σ c states, however, the ranges reach zero if the bounds are increased by factors of 1.7 for P c ( 4440 ) P_c(4440) P c ( 4440 ) and 1.4 for P c ( 4457 ) P_c(4457) P c ( 4457 ) . The signs are therefore determined only for contact couplings below about 40 G e V − 4 40\ \mathrm{GeV}^{-4} 40 GeV − 4 in the present scheme, a value within the range of the dimensional estimate.
Spin assignment and comparison with other calculations
The additive moment of a D ˉ ∗ Σ c \bar{D}^{*}\Sigma_c D ˉ ∗ Σ c state is negative for J = 1 2 J=\frac{1}{2} J = 2 1 if μ D ˉ ∗ < 0 \mu_{\bar{D}^{*}}<0 μ D ˉ ∗ < 0 and μ Σ c > 0 \mu_{\Sigma_c}>0 μ Σ c > 0 , and positive for J = 3 2 J=\frac{3}{2} J = 2 3 if, in addition, μ Σ c > ∣ μ D ˉ ∗ ∣ \mu_{\Sigma_c}>|\mu_{\bar{D}^{*}}| μ Σ c > ∣ μ D ˉ ∗ ∣ . Both conditions hold for all inputs considered here, with μ Σ c \mu_{\Sigma_c} μ Σ c between 1.10 1.10 1.10 and 1.72 μ N 1.72\mu_N 1.72 μ N and ∣ μ D ˉ ∗ ∣ |\mu_{\bar{D}^{*}}| ∣ μ D ˉ ∗ ∣ between 0.29 0.29 0.29 and 0.42 μ N 0.42\mu_N 0.42 μ N , so that the signs follow from the constituent moments and the one-body current. For contact couplings within the bounds of Eq. (19) , the two spin assignments give real parts of opposite sign, and the sign of the moment would determine the spin of a D ˉ ∗ Σ c \bar{D}^{*}\Sigma_c D ˉ ∗ Σ c molecule.
Table V compares the two orderings. Since the kernel of Sec. II B cannot be fitted to the data for ordering B, we place the two poles at the observed masses by shifting the subtraction constant of the D ˉ ∗ Σ c \bar{D}^{*}\Sigma_c D ˉ ∗ Σ c channel in set I by − 0.092 -0.092 − 0.092 for J = 3 2 J=\frac{3}{2} J = 2 3 and by + 0.046 +0.046 + 0.046 for J = 1 2 J=\frac{1}{2} J = 2 1 . This breaks HQSS and relies only to evaluate the loop at the required pole positions. The resulting moments, + 1.35 − 0.01 i +1.35-0.01i + 1.35 − 0.01 i and − 0.89 − 0.07 i μ N -0.89-0.07i\mu_N − 0.89 − 0.07 i μ N , differ in their real parts from those of ordering A for the same spin by less than 0.1 μ N 0.1\mu_N 0.1 μ N , so that the moments of the two states are interchanged. The ranges for ordering B were computed at its own pole positions.
Table 5. Real parts of the magnetic moments of the three observed states in μ N \mu_N μ N for the two spin orderings of the D ˉ ∗ Σ c \bar{D}^{*}\Sigma_c D ˉ ∗ Σ c states, compared with other calculations for molecular configurations. This work: central value and range as in Table IV . Reference [20 ] : quark model. Reference [24 ] : molecular picture with a single channel, with coupled channels, and with D-wave components. Reference [25 ] : chiral perturbation theory. Reference [27 ] : light-cone QCD sum rules with a molecular current.
The quark-model values of Ref. [20 ] for molecular configurations are additive moments and agree with the second column of Table IV within 0.05 μ N 0.05\mu_N 0.05 μ N . The closest previous calculation of the static moments, Ref. [24 ] , treats the states as coupled-channel molecules with D-wave components and constituent quark moments. Its values differ from our central values by at most 0.14 μ N 0.14\mu_N 0.14 μ N , and the D-wave components change the moments by less than 0.03 μ N 0.03\mu_N 0.03 μ N . Chiral perturbation theory gives 1.53 μ N 1.53\mu_N 1.53 μ N for the D ˉ Σ c \bar{D}\Sigma_c D ˉ Σ c state [25 ] . All these values lie within our ranges. In contrast, the light-cone sum rules of Ref. [27 ] with a molecular interpolating current and ordering B yield larger positive moments for both states, for the 1 2 − \frac{1}{2}^{-} 2 1 − state opposite in sign to the additive value. Compact configurations give still different moments [20 , 21 , 27 , 30 ] . Within the molecular picture, the sign of the moment of either D ˉ ∗ Σ c \bar{D}^{*}\Sigma_c D ˉ ∗ Σ c state would therefore constrain the spin assignment and discriminate among the calculations listed in Table V. Its uncertainty, the real parts cover 0.4 0.4 0.4 –2.2 2.2 2.2 , 0.3 0.3 0.3 –0.5 0.5 0.5 and 0.1 0.1 0.1 –2.5 f m 2 2.5\ \mathrm{fm}^{2} 2.5 fm 2 . Only for P c ( 4440 ) P_c(4440) P c ( 4440 ) is ⟨ x 2 ⟩ \langle x^2\rangle ⟨ x 2 ⟩ thus determined within a factor of two. The magnetic moments change by at most 0.06 μ N 0.06\mu_N 0.06 μ N over the mass ranges.
The uncertainty of κ \kappa κ has a smaller effect. At the ends of its interval, with the other parameters redetermined from the masses, the widths of the observed states are 6.6 6.6 6.6 , 16.7 16.7 16.7 and 2.4 2.4 2.4 MeV and 13.4 13.4 13.4 , 30.2 30.2 30.2 and 3.6 3.6 3.6 MeV. The moduli of ⟨ x 2 ⟩ \langle x^2\rangle ⟨ x 2 ⟩ then change by 8 8 8 , 6 6 6 and 2 % 2\% 2% , and the real parts of the magnetic moments by less than 0.01 μ N 0.01\mu_N 0.01 μ N . Since the spin does not enter Eq. (32) , the ranges in Table VI apply to both spin orderings of the D ˉ ∗ Σ c \bar{D}^{*}\Sigma_c D ˉ ∗ Σ c states.
Table VI. Mean squared distance and charge radius of the three observed states in fm2 ^2 2 at the poles of set II and their ranges. B B B is the binding energy in MeV at the central mass. “Mass”: the mass is varied within its experimental uncertainty at the width of set II. “Mass and width”: the mass and the width are varied within the experimental uncertainties. For P c ( 4457 ) P_c(4457) P c ( 4457 ) the width of set II lies below the experimental interval, so that the two ranges are not nested.
The calculation employs isospin-averaged masses, whereas the thresholds of the two charge states differ by 6.1 6.1 6.1 MeV for D ˉ Σ c \bar{D}\Sigma_c D ˉ Σ c and 4.7 4.7 4.7 MeV for D ˉ ∗ Σ c \bar{D}^{*}\Sigma_c D ˉ ∗ Σ c , which is comparable with the binding energies of P c ( 4312 ) P_c(4312) P c ( 4312 ) and P c ( 4457 ) P_c(4457) P c ( 4457 ) . Isospin breaking in P c ( 4457 ) P_c(4457) P c ( 4457 ) has been proposed as a test of its molecular nature [57 ] . Instead of a calculation in the charge basis, we estimate the effect in the zero-range limit, with the pole held fixed and the tails of the two charge components determined by their physical thresholds. The real part of ⟨ x 2 ⟩ \langle x^2\rangle ⟨ x 2 ⟩ then changes by − 6 -6 − 6 , − 2 -2 − 2 and + 7 % +7\% + 7% for the three observed states. The charge radius is more sensitive, because the neutral-meson component has the lower threshold and hence the longer tail, and its charges do not cancel. The real part of ⟨ r 2 ⟩ E \langle r^2\rangle_E ⟨ r 2 ⟩ E increases by 8 8 8 , 2 2 2 and 30 % 30\% 30% .
Finally, the constituents are treated as stable, whereas the isospin-weighted widths are Γ Σ c = 2.0 \Gamma_{\Sigma_c}=2.0 Γ Σ c = 2.0 MeV and Γ Σ c ∗ = 15.6 \Gamma_{\Sigma_c^*}=15.6 Γ Σ c ∗ = 15.6 MeV [41 ] . A constant constituent width Γ c o n \Gamma_{\mathrm{con}} Γ con shifts the threshold and, to leading order, the pole by − i Γ c o n / 2 -i\Gamma_{\mathrm{con}}/2 − i Γ con /2 , so that B c B_c B c and Eq. (32) remain unchanged and the width of the state becomes Γ R + Γ c o n \Gamma_R+\Gamma_{\mathrm{con}} Γ R + Γ con . Corrections from the energy dependence of the self-energy and from the three-body channels D ˉ ( ∗ ) Λ c π \bar{D}^{(*)}\Lambda_c\pi D ˉ ( ∗ ) Λ c π are not included. Their size is indicated by Γ c o n / ( 2 ∣ B c ∣ ) \Gamma_{\mathrm{con}}/(2|B_c|) Γ con / ( 2∣ B c ∣ ) , which is 0.10 0.10 0.10 , 0.04 0.04 0.04 and 0.20 0.20 0.20 for the observed states and 0.3 0.3 0.3 to 3.8 3.8 3.8 for the four states with a Σ c ∗ \Sigma_c^* Σ c ∗ . For the latter, the pole positions and radii should therefore be regarded as reference values for stable constituents.
Sensitivity on the observed masses and widths
The observed masses have uncertainties of − 0.9 + 6.8 ^{+6.8}_{-0.9} − 0.9 + 6.8 , − 4.9 + 4.3 ^{+4.3}_{-4.9} − 4.9 + 4.3 and − 1.8 + 4.1 ^{+4.1}_{-1.8} − 1.8 + 4.1 MeV, which are not small compared with the binding energies of 8.8 8.8 8.8 , 21.7 21.7 21.7 and 4.7 4.7 4.7 MeV. Since ⟨ x 2 ⟩ ∝ 1 / B c \langle x^2\rangle\propto1/B_c ⟨ x 2 ⟩ ∝ 1/ B c , they dominate the variation of the radii. When each pole is shifted to the ends of the observed mass range by varying the subtraction constant of the dominant channel, the ratio of the computed ⟨ x 2 ⟩ \langle x^2\rangle ⟨ x 2 ⟩ to 1 / ( 4 μ r e d B c ) 1/(4\mu_{\mathrm{red}}B_c) 1/ ( 4 μ red B c ) remains between 1.02 1.02 1.02 and 1.16 1.16 1.16 in modulus for binding energies between 0.6 0.6 0.6 and 26.6 26.6 26.6 MeV, in agreement with Eq. (32) . The ranges in Table VI are therefore obtained from the zero-range relation multiplied by the computed ratio at the central pole. With the widths of set II, ∣ ⟨ x 2 ⟩ ∣ |\langle x^2\rangle| ∣ ⟨ x 2 ⟩ ∣ changes by − 7 -7 − 7 to + 88 % +88\% + 88% , − 15 -15 − 15 to + 17 % +17\% + 17% and − 26 -26 − 26 to + 205 % +205\% + 205% for the three states. At fixed width, the real part of Eq. (31) is maximal, 1 / ( 4 μ r e d Γ R ) 1/(4\mu_{\mathrm{red}}\Gamma_R) 1/ ( 4 μ red Γ R ) , at B = Γ R / 2 B=\Gamma_R/2 B = Γ R /2 and vanishes at threshold, where the imaginary part approaches − 1 / ( 2 μ r e d Γ R ) -1/(2\mu_{\mathrm{red}}\Gamma_R) − 1/ ( 2 μ red Γ R ) . If the width is also varied within
Open channels and the triangle singularity
The photon loops of a channel l l l that is open at the energy s \sqrt{s} s are singular at the spacelike momentum transfer [35 ]
Q d i v 2 ( s ) = λ ( s , m s t 2 , m s p 2 ) m s p 2 = 4 q l 2 ( s ) s m s p 2 , (34) Q_{\mathrm{div}}^2(s)=\frac{\lambda(s,m_{\mathrm{st}}^2,m_{\mathrm{sp}}^2)}{m_{\mathrm{sp}}^2}=\frac{4q_l^2(s)s}{m_{\mathrm{sp}}^2}, \tag*{(34)} Q div 2 ( s ) = m sp 2 λ ( s , m st 2 , m sp 2 ) = m sp 2 4 q l 2 ( s ) s , (34)
where m s t m_{\mathrm{st}} m st is the mass of the constituent struck by the photon, m s p m_{\mathrm{sp}} m sp that of the spectator, and λ \lambda λ is the Källén function. This singularity is the leading Landau singularity of the triangle graph [58 , 59 , 60 ] . It corresponds to a process in which the state decays into the channel, the photon reverses the momentum of the struck constituent in the rest frame of the spectator, and the two constituents recombine with all three internal lines on shell. For a closed channel, the same expression gives the anomalous threshold of a bound-state form factor at timelike momentum transfer [61 ] .
At the pole, s = s ′ = z R 2 s=s'=z_R^2 s = s ′ = z R 2 , the triangle point lies at a complex momentum transfer below the real axis, for example at Q d i v 2 ( z R ) = ( 0.67 − 0.20 i ) G e V 2 Q_{\mathrm{div}}^2(z_R)=(0.67-0.20i)\,\mathrm{GeV}^2 Q div 2 ( z R ) = ( 0.67 − 0.20 i ) GeV 2 for the baryon loop of the D ˉ ∗ Λ c \bar{D}^{*}\Lambda_c D ˉ ∗ Λ c channel of P c ( 4312 ) P_c(4312) P c ( 4312 ) in set II. Consequently, the double-pole residue (3) is an analytic function of Q 2 Q^2 Q 2 in a neighborhood of the real axis. We define the form factor F a n F^{\mathrm{an}} F an as the branch of the residue that contains Q 2 = 0 Q^2=0 Q 2 = 0 . Following Ref. [35 ] , it is obtained by the continuation in the energy at Q 2 → 0 Q^2\to0 Q 2 → 0 , followed by the continuation in Q 2 Q^2 Q 2 along the real axis at fixed z R z_R z R . This branch satisfies the Ward identities and is normalized to the charge, and its Taylor coefficients at Q 2 = 0 Q^2=0 Q 2 = 0 give the radii and the moments. The triangle point is a logarithmic branch point of the residue. The continuation of F a n F^{\mathrm{an}} F an around this point leads to the adjacent branch, on which the diagonal electric loops Ξ l a n \Xi_l^{\mathrm{an}} Ξ l an are replaced by Ξ l a n + Δ l \Xi_l^{\mathrm{an}}+\Delta_l Ξ l an + Δ l with
Δ l ( Q 2 ) = M l [ 2 ( z R 2 + m s t 2 − m s p 2 ) + Q 2 ] 2 Q ( 4 z R 2 + Q 2 ) (35) \Delta_l(Q^2)=\frac{M_l\left[2(z_R^2+m_{\mathrm{st}}^2-m_{\mathrm{sp}}^2)+Q^2\right]}{2Q\left(4z_R^2+Q^2\right)} \tag*{(35)} Δ l ( Q 2 ) = 2 Q ( 4 z R 2 + Q 2 ) M l [ 2 ( z R 2 + m st 2 − m sp 2 ) + Q 2 ] (35)
(see Appendix A). For the magnetic loops with the photon on the meson and the baryon line, the corresponding differences are M l m l m l ′ / ( 2 Q E ) M_l\sqrt{m_lm_l'}/(2QE) M l m l m l ′ / ( 2 QE ) and M l M l ′ / ( 2 Q E ) M_lM_l'/(2QE) M l M l ′ / ( 2 QE ) . These analytic expressions agree with the difference of the two numerical continuations to machine precision (see Figure 4 ). Although the adjacent branch also satisfies the Ward identities, it is not normalized, since Δ l \Delta_l Δ l grows as 1 / Q 1/Q 1/ Q for Q 2 → 0 Q^2\to0 Q 2 → 0 . The requirement of regularity at Q 2 = 0 Q^2=0 Q 2 = 0 therefore selects F a n F^{\mathrm{an}} F an .
Figure 4. Left: contribution g l 2 F l B g_l^2 F_l^{\mathrm{B}} g l 2 F l B of the D ˉ ∗ Λ c \bar{D}^{*}\Lambda_c D ˉ ∗ Λ c channel to the baryon-number form factor of P c ( 4312 ) P_c(4312) P c ( 4312 ) in set II. Solid lines: the branch that is analytic in Q 2 Q^2 Q 2 and contains Q 2 = 0 Q^2=0 Q 2 = 0 . Dashed lines: the adjacent branch, shown above the triangle point Q d i v 2 ( M R 2 ) Q_{\mathrm{div}}^2(M_R^2) Q div 2 ( M R 2 ) . Right: difference of the two branches, g l 2 Δ l g_l^2\Delta_l g l 2 Δ l . Symbols are the difference of the two numerical continuations, lines the closed form (35) . The shaded area is the triangle interval (37) .
The relation between the two branches and the amplitude at real energies is as follows. For a given real Q 2 > 0 Q^2>0 Q 2 > 0 , each loop of an open channel has a logarithmic branch point at the real energy E t r ( Q 2 ) E_{\mathrm{tr}}(Q^2) E tr ( Q 2 ) defined by Q d i v 2 ( E t r 2 ) = Q 2 Q_{\mathrm{div}}^2(E_{\mathrm{tr}}^2)=Q^2 Q div 2 ( E tr 2 ) = Q 2 . The pole can be reached from real energies on either side of this point. The continuation in the energy at fixed Q 2 Q^2 Q 2 yields F l a n F_l^{\mathrm{an}} F l an if it starts at s > E t r \sqrt{s}>E_{\mathrm{tr}} s > E tr , and F l a n + Δ l F_l^{\mathrm{an}}+\Delta_l F l an + Δ l if it starts at s < E t r \sqrt{s}<E_{\mathrm{tr}} s < E tr . For Q 2 < Q d i v 2 ( M R 2 ) Q^2<Q_{\mathrm{div}}^2(M_R^2) Q 2 < Q div 2 ( M R 2 ) , the point E t r E_{\mathrm{tr}} E tr lies below M R M_R M R , whereas for larger Q 2 Q^2 Q 2 it lies above M R M_R M R . The value of the residue that is connected to the amplitude at the resonance energy by a continuation at fixed Q 2 Q^2 Q 2 is therefore
F c o n ( Q 2 ) = F a n ( Q 2 ) + ∑ l θ ( Q 2 − Q d i v 2 ( M R 2 ) ) Δ F ( Q 2 ) , (36) F^{\mathrm{con}}(Q^2)=F^{\mathrm{an}}(Q^2)+\sum_l\theta\left(Q^2-Q_{\mathrm{div}}^2(M_R^2)\right)\Delta F(Q^2), \tag*{(36)} F con ( Q 2 ) = F an ( Q 2 ) + l ∑ θ ( Q 2 − Q div 2 ( M R 2 ) ) Δ F ( Q 2 ) , (36)
where the sum runs over the loops of the open channels, each with its own triangle point and with the difference Δ F \Delta F Δ F that follows from Eq. (35) and the couplings and charges of Eq. (25). Below the lowest triangle point, F c o n = F a n F^{\mathrm{con}}=F^{\mathrm{an}} F con = F an , so that the charge, the radii and the moments coincide. Since F c o n F^{\mathrm{con}} F con is a piecewise selection of branches and not an analytic function of Q 2 Q^2 Q 2 , we do not use it to define the form factor. Instead, it specifies which branch of the residue multiplies the double pole of the amplitude at real energies close to M R M_R M R . If the triangle point lies within the half width of the state,
Q d i v 2 ( M R − 1 2 Γ R ) 2 < Q 2 < Q d i v 2 ( ( M R + 1 2 Γ R ) 2 ) , (37) Q_{\mathrm{div}}^2\left(M_R-\frac{1}{2}\Gamma_R\right)^2<Q^2<Q_{\mathrm{div}}^2\left(\left(M_R+\frac{1}{2}\Gamma_R\right)^2\right), \tag*{(37)} Q div 2 ( M R − 2 1 Γ R ) 2 < Q 2 < Q div 2 ( ( M R + 2 1 Γ R ) 2 ) , (37)
the branch point E t r ( Q 2 ) E_{\mathrm{tr}}(Q^2) E tr ( Q 2 ) lies inside the resonance peak, and the amplitude on either side of it is connected to a different branch.
Among the three observed states, only P c ( 4312 ) P_c(4312) P c ( 4312 ) has a triangle point below 1 G e V 2 1\,\mathrm{GeV}^2 1 GeV 2 . In set II, its D ˉ ∗ Σ c \bar{D}^{*}\Sigma_c D ˉ ∗ Σ c channel is open by 16.9 MeV, with χ l = − 0.002 − 0.002 i \chi_l=-0.002-0.002i χ l = − 0.002 − 0.002 i . The triangle points are located at Q d i v 2 ( M R 2 ) = 0.51 G e V 2 Q_{\mathrm{div}}^2(M_R^2)=0.51\,\mathrm{GeV}^2 Q div 2 ( M R 2 ) = 0.51 GeV 2 for the meson loop and 0.67 G e V 2 0.67\,\mathrm{GeV}^2 0.67 GeV 2 for the baryon loop, and the corresponding intervals (37) are 0.36 < Q 2 < 0.67 G e V 2 0.36<Q^2<0.67\,\mathrm{GeV}^2 0.36 < Q 2 < 0.67 GeV 2 and 0.47 < Q 2 < 0.87 G e V 2 0.47<Q^2<0.87\,\mathrm{GeV}^2 0.47 < Q 2 < 0.87 GeV 2 . The two branches differ by at most 0.009 in G E G_E G E , which corresponds to 1.2%, and by 0.009 in G M G_M G M . In set I, the channel is open by 11.6 MeV with ∣ X l ∣ = 0.010 \lvert X_l\rvert=0.010 ∣ X l ∣ = 0.010 , and the difference reaches at most 0.035. For the other states, the lowest triangle points of the diagonal loops are located at 1.3, 1.6, 2.0, 2.6 and 3.4 GeV2 ^2 2 in set II. The lowest triangle point of a magnetic transition loop between two open channels lies at 2.1 GeV2 ^2 2 . The difference between the two branches is thus smaller than the variation of the form factors with the parameter set, the structure of the constituents and the contact terms.
The on-shell formalism of Ref. [36 ] defines the form factor of a resonance by the same double-pole residue. The continuation of the triangle function to the second sheets given in Eq. (24) of Ref. [36 ] agrees with ours at real energies (see Appendix A). In that work, however, the form factor is obtained by a continuation in the energies at fixed real Q 2 Q^2 Q 2 . When started at s = M R \sqrt{s}=M_R s = M R , this continuation yields F c o n F^{\mathrm{con}} F con , which coincides with F a n F^{\mathrm{an}} F an below the triangle point but requires the specification given here above it. The two branches exist for the form factor of any resonance with an open decay channel.
In the Λ ( 1405 ) \Lambda(1405) Λ ( 1405 ) study of Ref. [35 ] , the π Σ \pi\Sigma π Σ channel is open, and for the higher pole Eq. (34) gives 0.17 GeV2 ^2 2 for the loop with the photon on the pion. This singularity produces the structure near 0.2 GeV2 ^2 2 in the π + \pi^+ π + component of the electric form factor shown in that reference. Since the π \pi π and the Σ \Sigma Σ have zero hypercharge, their charges are pure isovector charges. For an isoscalar state with isospin-averaged masses, the π + Σ − \pi^+\Sigma^- π + Σ − and π − Σ + \pi^-\Sigma^+ π − Σ + components therefore cancel at every Q 2 Q^2 Q 2 and on both branches. The baryon-number, strangeness and magnetic currents couple only to the Σ \Sigma Σ . With the light pion as spectator, the corresponding triangle point lies at 12.7 GeV2 ^2 2 , far outside the range of Q 2 Q^2 Q 2 considered there. In P c ( 4312 ) P_c(4312) P c ( 4312 ) , the D ˉ ∗ 0 Λ c + \bar{D}^{*0}\Lambda_c^+ D ˉ ∗ 0 Λ c + channel has the single charge state D ˉ ∗ 0 Λ c + \bar{D}^{*0}\Lambda_c^+ D ˉ ∗ 0 Λ c + , so that no such cancellation occurs.
Finally, we note that this singularity is of the same type as the triangle singularities proposed as the origin of the P c P_c P c signals [10 , 11 , 12 , 13 , 14 ] , but it appears in a different graph and in a different kinematic variable. Since it presupposes the existence of the pole, it does not bear on the question of whether the observed peaks are of kinematical origin.
Summary
In this work, we have studied the electromagnetic form factors of the hidden-charm pentaquarks as D ˉ ( ∗ ) Σ c ( ∗ ) \bar{D}^{(*)}\Sigma_c^{(*)} D ˉ ( ∗ ) Σ c ( ∗ ) molecules in a coupled-channel unitary approach with a vector-meson exchange kernel that respects HQSS. The form factors are obtained from the residues at the double pole of the photon-coupled amplitude. We have constructed the photon–four-hadron vertex of the HQSS channel space. For a charged state, the vertex of the gauged energy-dependent kernel is required, and the charge normalization then coincides with the compositeness sum rule. The one-body and contact magnetic operators connect the channels related by HQSS. Our results are obtained for stable, pointlike constituents and isospin-averaged masses.
We have found that the complex mean squared distance between the constituents satisfies ⟨ x 2 ⟩ = [ 4 μ r e d B c ( 1 − γ r e ) ] − 1 \langle x^2\rangle=[4\mu_{\mathrm{red}}B_c(1-\gamma r_e)]^{-1} ⟨ x 2 ⟩ = [ 4 μ red B c ( 1 − γ r e ) ] − 1 within 2%, where B c B_c B c is the complex binding energy and r e r_e r e the complex effective range of the dominant channel, with Re r e = 0.08 \operatorname{Re}r_e=0.08 Re r e = 0.08 –0.12 fm. With the parameters fitted to the three observed masses and widths, we obtain ⟨ x 2 ⟩ = 0.85 − 0.47 i \langle x^2\rangle=0.85-0.47i ⟨ x 2 ⟩ = 0.85 − 0.47 i , 0.35 − 0.18 i 0.35-0.18i 0.35 − 0.18 i and 1.76 − 0.57 i f m 2 1.76-0.57i\ \mathrm{fm}^2 1.76 − 0.57 i fm 2 for P c ( 4312 ) P_c(4312) P c ( 4312 ) , P c ( 4440 ) P_c(4440) P c ( 4440 ) and P c ( 4457 ) P_c(4457) P c ( 4457 ) , respectively. The uncertainties of these masses change the modulus of ⟨ x 2 ⟩ \langle x^2\rangle ⟨ x 2 ⟩ by − 7 -7 − 7 to + 88 % +88\% + 88% , − 15 -15 − 15 to + 17 % +17\% + 17% and − 26 -26 − 26 to + 205 % +205\% + 205% . The molecular contribution to the charge radius is small, 0.05 to 0.24 fm2 ^2 2 in the real part, because the meson and the baryon carry charges of opposite sign. Electric contact currents of the size suggested by dimensional analysis change this contribution by 0.01 to 0.04 f m 2 0.04\,\mathrm{fm}^{2} 0.04 fm 2 , while the finite size of the constituents adds 0.39 f m 2 0.39\,\mathrm{fm}^{2} 0.39 fm 2 in a vector-meson dominance estimate. The real parts of the magnetic moments agree with the additive moments of the constituents within 1–11%. Using constituent quark moments, we obtain Re μ = 1.74 \operatorname{Re}\mu= 1.74 Re μ = 1.74 , − 0.95 -0.95 − 0.95 and 1.26 μ N 1.26\,\mu_N 1.26 μ N , with imaginary parts below 0.06 μ N 0.06\,\mu_N 0.06 μ N . Taking into account the lattice results for the Σ c \Sigma_c Σ c moments, contact couplings within the bounds from dimensional analysis and the loop normalization, we obtain the ranges 0.8 to 2.0, − 1.3 -1.3 − 1.3 to − 0.2 -0.2 − 0.2 and 0.2 to 1.8 μ N \mu_N μ N . These results correspond to the spin assignment 1 2 − \frac{1}{2}^{-} 2 1 − for P c ( 4440 ) P_c(4440) P c ( 4440 ) and 3 2 − \frac{3}{2}^{-} 2 3 − for P c ( 4457 ) P_c(4457) P c ( 4457 ) . For the opposite ordering, the moments of the two states are interchanged. The opposite signs of the moments of the two D ˉ ∗ Σ c \bar{D}^{*}\Sigma_c D ˉ ∗ Σ c states persist for the constituent moments considered here and for contact couplings up to 1.4 times the adopted bounds.
We have also shown that the photon loops of an open channel have a triangle singularity at spacelike momentum transfer. At the pole, this singularity becomes a branch point at complex Q 2 Q^2 Q 2 , and the form factor is defined as the branch of the residue that is analytic at Q 2 = 0 Q^2=0 Q 2 = 0 . Beyond the triangle point associated with the resonance energy, the value connected to the amplitude at real energies lies on the adjacent branch, which differs by a term given in closed form. The charge, the radii and the moments are the same on both branches. For the observed states and Q 2 ≤ 1 G e V 2 Q^2\leq1\,\mathrm{GeV}^{2} Q 2 ≤ 1 GeV 2 , this effect concerns only P c ( 4312 ) P_c(4312) P c ( 4312 ) , for which the two branches differ by 1% in set II and 4% in set I.
The quantities at the poles are complex and are not direct observables, and the present model contains only the molecular component. The largest variation of the magnetic moments arises from the moments of the Σ ( ∗ ) \Sigma^{(*)} Σ ( ∗ ) and D ˉ ∗ \bar{D}^{*} D ˉ ∗ , so that lattice QCD results at the physical pion mass would reduce the present ranges. A determination of the effective range of the D ˉ ( ∗ ) Σ c \bar{D}^{(*)}\Sigma_c D ˉ ( ∗ ) Σ c amplitudes from line shapes would test the value of 0.1 fm obtained with the present kernel. The inclusion of one-pion exchange with D D D waves, a calculation in the charge basis and the effects of the widths of the constituents are left for future work. The partner states, including a 5 2 − D ˉ ∗ Σ c ∗ \frac{5}{2}^{-}\bar{D}^{*}\Sigma_c^{*} 2 5 − D ˉ ∗ Σ c ∗ state 2 MeV below its threshold in set II, are predictions of the model for stable constituents.
Data availability
There are no publicly available research data or software supporting this manuscript. Requests for further information or data should be sent to the authors.
Acknowledgments
This research project is supported by National Research Council of Thailand (NRCT): (Contract No. N41A680287). DS is supported by Thailand NSRF via PMU-B [grant number B39G680009]. DS has also received funding support from the Fundamental Fund of Khon Kaen University. Generative AI tools (Claude, Anthropic, and ChatGPT, OpenAI) were used to assist with the drafting and grammatical correction of the manuscript text and with numerical implementation. The authors directed this assistance, verified all scientific content and equations against their own calculations, and take full responsibility for the paper.
Photon loops and their continuation
The triangle loops D M μ D_M^\mu D M μ and D B μ D_B^\mu D B μ in Eq. (23) are evaluated with Feynman parameters, as in Appendix B of Ref. [35 ] . With ∬ = ∫ 0 1 d x ∫ 0 1 − x d y \iint=\int_0^1 dx\int_0^{1-x}dy ∬ = ∫ 0 1 d x ∫ 0 1 − x d y and z = 1 − x − y z=1-x-y z = 1 − x − y , the loop functions in Eq. (25) at s = s ′ = z R \sqrt{s}=\sqrt{s'}=z_R s = s ′ = z R are given by
F l M = 2 M l E 16 π 2 ∬ 2 z D l l M , M l ′ l M = 2 M l ′ 2 m l m l ′ 16 π 2 ∬ 1 D l ′ l M , \mathcal{F}^{M}_{l}=\frac{2M_lE}{16\pi^2}\iint\frac{2z}{D^{M}_{ll}},\qquad\mathcal{M}^{M}_{l'l}=\frac{2M_{l'}2\sqrt{m_lm_{l'}}}{16\pi^2}\iint\frac{1}{D^{M}_{l'l}}, F l M = 16 π 2 2 M l E ∬ D l l M 2 z , M l ′ l M = 16 π 2 2 M l ′ 2 m l m l ′ ∬ D l ′ l M 1 ,
F l B = 2 M l E 16 π 2 ∬ 2 x D l l B , M l ′ l B = 2 M l ′ 2 M l ′ 16 π 2 ∬ 1 D l ′ l B , (38) \mathcal{F}^{B}_{l}=\frac{2M_lE}{16\pi^2}\iint\frac{2x}{D^{B}_{ll}},\qquad\mathcal{M}^{B}_{l'l}=\frac{2M_{l'}2M_l'}{16\pi^2}\iint\frac{1}{D^{B}_{l'l}}, \tag*{(38)} F l B = 16 π 2 2 M l E ∬ D l l B 2 x , M l ′ l B = 16 π 2 2 M l ′ 2 M l ′ ∬ D l ′ l B 1 , (38)
with the denominators
D l ′ l M = x m l 2 + y m l ′ 2 + z M l 2 + x y Q 2 − z ( 1 − z ) z R 2 , D^{M}_{l'l}=xm_l^2+ym_{l'}^2+zM_l^2+xyQ^2-z(1-z)z_R^2, D l ′ l M = x m l 2 + y m l ′ 2 + z M l 2 + x y Q 2 − z ( 1 − z ) z R 2 ,
D l ′ l B = x m l 2 + y M l ′ 2 + z M l 2 + y z Q 2 − x ( 1 − x ) z R 2 . (39) D^{B}_{l'l}=xm_l^2+yM_{l'}^2+zM_l^2+yzQ^2-x(1-x)z_R^2. \tag*{(39)} D l ′ l B = x m l 2 + y M l ′ 2 + z M l 2 + y z Q 2 − x ( 1 − x ) z R 2 . (39)
The integrals over the real simplex define the loop functions on the first sheet in s s s and s ′ s' s ′ . For an open channel, the second sheet is reached by adding the discontinuity across the unitarity cut, which follows from the Cutkosky rule [62 , 63 ] . If both channels of a loop are open,
F I I , I I = F I , I + disc s F I ∣ + i ε + disc s ′ F I ∣ − i ε , (40) F^{\mathrm{II,II}}=F^{\mathrm{I,I}}+\operatorname{disc}_{s}F^{\mathrm{I}}|_{+i\varepsilon}+\operatorname{disc}_{s'}F^{\mathrm{I}}|_{-i\varepsilon}, \tag*{(40)} F II , II = F I , I + disc s F I ∣ + i ε + disc s ′ F I ∣ − i ε , (40)
where the uncut propagator carries the Feynman prescription in the first discontinuity and the conjugate prescription in the second. For real energies above both thresholds, Eq. (40) reduces to F I I , I I = F − 2 i Im F F^{\mathrm{II,II}}=F-2i\,\operatorname{Im}F F II , II = F − 2 i Im F , which is the continuation of the triangle function given in Eq. (24) of Ref. [36 ] .
In the discontinuity in s s s , the two cut lines are on shell, and the uncut propagator is linear in the cosine of the angle between the on-shell constituent and the photon in the rest frame of P P P , α ≠ β cos θ \alpha\ne\beta\cos\theta α = β cos θ . The angular integration yields the logarithm L = ln w L=\ln w L = ln w with w = ( α + β ) / ( α − β ) w=(\alpha+\beta)/(\alpha-\beta) w = ( α + β ) / ( α − β ) . For the scalar triangle S = ∬ 1 / D S=\iint1/D S = ∬ 1/ D , where D D D is one of the denominators in Eq. (39) , we obtain
disc s S = − i q l s 2 π β L , (41) \operatorname{disc}_{s}S=-\frac{iq_l}{\sqrt{s}}\frac{2\pi}{\beta}L, \tag*{(41)} disc s S = − s i q l β 2 π L , (41)
where q l q_l q l denotes the momentum on the sheet of the pole, with Re q l > 0 \operatorname{Re}q_l>0 Re q l > 0 for an open channel. For a diagonal loop at s = s ′ s=s' s = s ′ , one has β = 2 q l Q E / s \beta=2q_lQE/\sqrt{s} β = 2 q l QE / s , α = − Q 2 ω s p / s \alpha=-Q^2\omega_{\mathrm{sp}}/\sqrt{s} α = − Q 2 ω sp / s and
w = 1 − z ∗ 1 + z ∗ , z ∗ = Q ω s p 2 q l E , (42) w=\frac{1-z^*}{1+z^*},\qquad z^*=\frac{Q\omega_{\mathrm{sp}}}{2q_lE}, \tag*{(42)} w = 1 + z ∗ 1 − z ∗ , z ∗ = 2 q l E Q ω sp , (42)
where ω s p = ( s + m s p 2 − m s t 2 ) / ( 2 s ) \omega_{\mathrm{sp}}=(s+m_{\mathrm{sp}}^{2}-m_{\mathrm{st}}^{2})/(2\sqrt{s}) ω sp = ( s + m sp 2 − m st 2 ) / ( 2 s ) is the energy of the spectator. The zero of w w w at z ⋆ = 1 z^{\star}=1 z ⋆ = 1 corresponds to the triangle point (34) . Since Im z ⋆ > 0 \operatorname{Im}z^{\star}>0 Im z ⋆ > 0 at the pole, w w w does not cross the cut of the principal value Ln \operatorname{Ln} Ln for real Q 2 > 0 Q^{2}>0 Q 2 > 0 . For the continuation in the energy at Q 2 → 0 Q^{2}\to0 Q 2 → 0 followed by the continuation in Q 2 Q^{2} Q 2 , which defines F a n a F_{\mathrm{an}}^{\phantom{a}} F an a , the logarithms of the terms cut in s s s and in s ′ s^{\prime} s ′ are
L s = Ln w − 2 π i , L s ′ = Ln w (43) L_s=\operatorname{Ln}w-2\pi i,\qquad L_{s^{\prime}}=\operatorname{Ln}w \tag*{(43)} L s = Ln w − 2 π i , L s ′ = Ln w (43)
for all Q 2 Q^{2} Q 2 . For the continuation in the energy at fixed Q 2 Q^{2} Q 2 from s = M R \sqrt{s}=M_R s = M R , they coincide with Eq. (43) below the triangle point, whereas L s = L s ′ = Ln w L_s=L_{s^{\prime}}=\operatorname{Ln}w L s = L s ′ = Ln w above it. The two branches therefore differ by δ L = 2 π i \delta L=2\pi i δ L = 2 π i in the term cut in s s s . In Eq. (41) , the momentum q l q_l q l then cancels, and δ S = 2 π 2 / ( Q E ) \delta S=2\pi^{2}/(Q E) δ S = 2 π 2 / ( QE ) . The same steps applied to the charge vertices give Eq. (35) for the electric loops, while the prefactors in Eq. (38) give the differences M l m l m l ′ / ( 2 Q E ) M_l\sqrt{m_lm_l^{\prime}}/(2QE) M l m l m l ′ / ( 2 QE ) and M l M l ′ / ( 2 Q E ) M_lM_l^{\prime}/(2QE) M l M l ′ / ( 2 QE ) for the magnetic loops. For a transition loop between two open channels, in which the mass of the struck constituent changes from m s t m_{\mathrm{st}} m st to m s t ′ m_{\mathrm{st}}^{\prime} m st ′ , the triangle point is located at Q d i v 2 = [ ( λ s t 1 / 2 + λ s t ′ 1 / 2 ) 2 − ( m s t 2 − m s t ′ 2 ) ] / ( 4 m s p 2 ) Q_{\mathrm{div}}^{2}=[(\lambda_{\mathrm{st}}^{1/2}+\lambda_{\mathrm{st}^{\prime}}^{1/2})^{2}-(m_{\mathrm{st}}^{2}-m_{\mathrm{st}}^{\prime2})]/(4m_{\mathrm{sp}}^{2}) Q div 2 = [( λ st 1/2 + λ st ′ 1/2 ) 2 − ( m st 2 − m st ′2 )] / ( 4 m sp 2 ) with λ s t = λ ( s , m s t 2 , m s p 2 ) \lambda_{\mathrm{st}}=\lambda(s,m_{\mathrm{st}}^{2},m_{\mathrm{sp}}^{2}) λ st = λ ( s , m st 2 , m sp 2 ) and λ s t ′ = λ ( s , m s t ′ 2 , m s p 2 ) \lambda_{\mathrm{st}^{\prime}}=\lambda(s,m_{\mathrm{st}}^{\prime2},m_{\mathrm{sp}}^{2}) λ st ′ = λ ( s , m st ′2 , m sp 2 ) .
The magnetic operators of lowest order in derivatives contain one magnetic field B B B and no further derivative. Owing to HQSS, their spin structures act only on the light-quark spins, i.e., on the first index of H ˉ \bar H H ˉ , the Cartesian index of S i S^i S i and the spinor index of N N N . The photon couples as an isoscalar plus the third component of an isovector. Since the model space contains three types of channels, D ˉ ( ∗ ) Σ c ( ∗ ) \bar D^{(*)}\Sigma_c^{(*)} D ˉ ( ∗ ) Σ c ( ∗ ) , D ˉ ( ∗ ) Λ c \bar D^{(*)}\Lambda_c D ˉ ( ∗ ) Λ c and the hidden-charm channels, there are six pairs of channel types. In terms of the light-spin bilinears O H ˉ 0 = ⟨ H ˉ † H ˉ ⟩ \mathcal{O}_{\bar H}^{0}=\langle\bar H^\dagger\bar H\rangle O H ˉ 0 = ⟨ H ˉ † H ˉ ⟩ , O H ˉ j = ⟨ H ˉ † σ j H ˉ ⟩ \mathcal{O}_{\bar H}^{j}=\langle\bar H^\dagger\sigma^j\bar H\rangle O H ˉ j = ⟨ H ˉ † σ j H ˉ ⟩ , O S 0 = S i † S i \mathcal{O}_{S}^{0}=S^{i\dagger}S^i O S 0 = S i † S i , O S j = i ϵ i j k S i † S k \mathcal{O}_{S}^{j}=i\epsilon^{ijk}S^{i\dagger}S^k O S j = i ϵ ij k S i † S k and O S i j = S i † S j + S j † S i − 2 3 δ i j S k † S k \mathcal{O}_{S}^{ij}=S^{i\dagger}S^j+S^{j\dagger}S^i-\frac{2}{3}\delta^{ij}S^{k\dagger}S^k O S ij = S i † S j + S j † S i − 3 2 δ ij S k † S k , and of six isospin structures labeled by t t t for the H ˉ † H ˉ S † S \bar H^\dagger\bar H S^\dagger S H ˉ † H ˉ S † S operators, the operators read
L γ ( 1 ) = e B j ∑ t ≠ 5 { d t ( a ) [ O H ˉ j ⊗ O S 0 ] t + d t ( b ) [ O H ˉ 0 ⊗ O S j ] t + d t ( c ) [ O H ˉ i ⊗ O S i j ] t } + e d 5 ( e ) ϵ i j k B i [ O H ˉ j ⊗ O S k ] 5 , L γ ( 2 ) = e B j { d 0 ( d ) ⟨ H ˉ † σ j H ˉ ⟩ + d 1 ( d ) ⟨ H ˉ † σ j τ 3 H ˉ ⟩ } Λ c † Λ c , L γ ( 3 ) = e B i ∑ t = 1 3 f t R t i + i e ϵ i j k B i ∑ t = 1 3 f t R t j k + h . c . , L γ ( 4 ) = e B j ∑ t = 1 3 { h t N † σ i σ j H ˉ Ψ † S i + h t ′ N † σ j σ i H ˉ Ψ † S i } N t + h . c . , L γ ( 5 ) = e B j N † σ j ( k 0 + k 1 τ 3 ) H ˉ Ψ † Λ c + h . c . , L γ ( 6 ) = e B j N † σ j ( n 0 + n 1 τ 3 ) N ⟨ Ψ † Ψ ⟩ , (44) \begin{aligned}
\mathcal{L}_{\gamma}^{(1)}&=eB^j\sum_{t\ne5}\left\{d_t^{(a)}[\mathcal{O}_{\bar H}^{j}\otimes\mathcal{O}_{S}^{0}]_t+d_t^{(b)}[\mathcal{O}_{\bar H}^{0}\otimes\mathcal{O}_{S}^{j}]_t+d_t^{(c)}[\mathcal{O}_{\bar H}^{i}\otimes\mathcal{O}_{S}^{ij}]_t\right\}+ed_5^{(e)}\epsilon^{ijk}B^i[\mathcal{O}_{\bar H}^{j}\otimes\mathcal{O}_{S}^{k}]_5,\\
\mathcal{L}_{\gamma}^{(2)}&=eB^j\left\{d_0^{(d)}\langle\bar H^\dagger\sigma^j\bar H\rangle+d_1^{(d)}\langle\bar H^\dagger\sigma^j\tau_3\bar H\rangle\right\}\Lambda_c^\dagger\Lambda_c,\\
\mathcal{L}_{\gamma}^{(3)}&=eB^i\sum_{t=1}^{3}f_t\mathcal{R}_t^i+ie\epsilon^{ijk}B^i\sum_{t=1}^{3}f_t\mathcal{R}_t^{jk}+\mathrm{h.c.},\\
\mathcal{L}_{\gamma}^{(4)}&=eB^j\sum_{t=1}^{3}\left\{h_tN^\dagger\sigma^i\sigma^j\bar H\Psi^\dagger S^i+h_t^{\prime}N^\dagger\sigma^j\sigma^i\bar H\Psi^\dagger S^i\right\}_{\mathcal{N}_t}+\mathrm{h.c.},\\
\mathcal{L}_{\gamma}^{(5)}&=eB^jN^\dagger\sigma^j(k_0+k_1\tau_3)\bar H\Psi^\dagger\Lambda_c+\mathrm{h.c.},\\
\mathcal{L}_{\gamma}^{(6)}&=eB^jN^\dagger\sigma^j(n_0+n_1\tau_3)N\langle\Psi^\dagger\Psi\rangle, \tag*{(44)}
\end{aligned} L γ ( 1 ) L γ ( 2 ) L γ ( 3 ) L γ ( 4 ) L γ ( 5 ) L γ ( 6 ) = e B j t = 5 ∑ { d t ( a ) [ O H ˉ j ⊗ O S 0 ] t + d t ( b ) [ O H ˉ 0 ⊗ O S j ] t + d t ( c ) [ O H ˉ i ⊗ O S ij ] t } + e d 5 ( e ) ϵ ij k B i [ O H ˉ j ⊗ O S k ] 5 , = e B j { d 0 ( d ) ⟨ H ˉ † σ j H ˉ ⟩ + d 1 ( d ) ⟨ H ˉ † σ j τ 3 H ˉ ⟩ } Λ c † Λ c , = e B i t = 1 ∑ 3 f t R t i + i e ϵ ij k B i t = 1 ∑ 3 f t R t j k + h.c. , = e B j t = 1 ∑ 3 { h t N † σ i σ j H ˉ Ψ † S i + h t ′ N † σ j σ i H ˉ Ψ † S i } N t + h.c. , = e B j N † σ j ( k 0 + k 1 τ 3 ) H ˉ Ψ † Λ c + h.c. , = e B j N † σ j ( n 0 + n 1 τ 3 ) N ⟨ Ψ † Ψ ⟩ , (44)
with
R 1 , 2 , 3 i = { S i , a † ⟨ H ˉ † τ a H ˉ ⟩ , S i , 3 † ⟨ H ˉ † H ˉ ⟩ , i ϵ 3 a b S i , a † ⟨ H ˉ † τ b H ˉ ⟩ } Λ c , \mathcal{R}_{1,2,3}^{i}=\left\{S^{i,a\dagger}\langle\bar H^\dagger\tau^a\bar H\rangle,\ S^{i,3\dagger}\langle\bar H^\dagger\bar H\rangle,\ i\epsilon^{3ab}S^{i,a\dagger}\langle\bar H^\dagger\tau^b\bar H\rangle\right\}_{\Lambda_c}, R 1 , 2 , 3 i = { S i , a † ⟨ H ˉ † τ a H ˉ ⟩ , S i , 3 † ⟨ H ˉ † H ˉ ⟩ , i ϵ 3 ab S i , a † ⟨ H ˉ † τ b H ˉ ⟩ } Λ c ,
N 1 , 2 , 3 = { ( N † τ a H ˉ ) S a , ( N † H ˉ ) S 3 , i ϵ 3 a b ( N † τ a H ˉ ) S b } (45) \mathcal{N}_{1,2,3}=\left\{(N^\dagger\tau^a\bar H)S^a,\ (N^\dagger\bar H)S^3,\ i\epsilon^{3ab}(N^\dagger\tau^a\bar H)S^b\right\} \tag*{(45)} N 1 , 2 , 3 = { ( N † τ a H ˉ ) S a , ( N † H ˉ ) S 3 , i ϵ 3 ab ( N † τ a H ˉ ) S b } (45)
where R t j k \mathcal{R}_t^{jk} R t j k is defined as R t i \mathcal{R}_t^i R t i with σ j \sigma^j σ j inserted in the H ˉ \bar H H ˉ trace, and all couplings are real with dimension E − 4 E^{-4} E − 4 . The subscript N t \mathcal{N}_t N t denotes the isospin contraction. The six isospin structures are ⟨ H ˉ † H ˉ ⟩ S † S \langle\bar H^\dagger\bar H\rangle S^\dagger S ⟨ H ˉ † H ˉ ⟩ S † S , V a T a V^a\mathcal{T}^a V a T a , V 3 S † S V^3S^{\dagger}S V 3 S † S , ⟨ H ˉ † H ˉ ⟩ I S 3 \langle\bar H^\dagger\bar H\rangle I^3_S ⟨ H ˉ † H ˉ ⟩ I S 3 , ϵ 3 a b V a T b \epsilon^{3ab}V^a\mathcal{T}^b ϵ 3 ab V a T b and V a Q S a 3 V^aQ_S^{a3} V a Q S a 3 for t = 1 , … , 6 t=1,\ldots,6 t = 1 , … , 6 , with V a = ⟨ H ˉ † τ a H ˉ ⟩ V^a=\langle\bar H^\dagger\tau^a\bar H\rangle V a = ⟨ H ˉ † τ a H ˉ ⟩ , the isospin operator T a \mathcal{T}^a T a of the Σ c ( ∗ ) \Sigma_c^{(*)} Σ c ( ∗ ) and Q S a b = 3 2 δ a b − { T a , T b } Q_S^{ab}=\frac{3}{2}\delta^{ab}-\{\mathcal{T}^a,\mathcal{T}^b\} Q S ab = 2 3 δ ab − { T a , T b } . Time-reversal invariance combines the structures d d d , a a a , b b b and c c c with the five isospin structures that are even under time reversal, and the fourth spin structure with t = 5 t=5 t = 5 . This yields 16 operators for the D ˉ ( ∗ ) Σ c ( ∗ ) \bar D^{(*)}\Sigma_c^{(*)} D ˉ ( ∗ ) Σ c ( ∗ ) channels. Including the operators for the other pairs of channel types, the total number is 16 + 2 + 6 + 6 + 2 + 2 = 34 16+2+6+6+2+2=34 16 + 2 + 6 + 6 + 2 + 2 = 34 . Two-body operators that act on the heavy-quark spins are suppressed by Λ Q C D / m c \Lambda_{\mathrm{QCD}}/m_c Λ QCD / m c and are not included.
All operators have the form e c B j X j e\,c\,B^j\mathcal{X}^j e c B j X j , where c c c is a coupling and X j \mathcal{X}^j X j is a field bilinear that transforms as a vector in the light-quark spins. For an incoming photon, A μ = ϵ μ e − i q ⋅ x A^\mu=\epsilon^\mu e^{-iq\cdot x} A μ = ϵ μ e − i q ⋅ x with ϵ 0 = 0 \epsilon^0=0 ϵ 0 = 0 , one has B = i q × ϵ B=i\mathbf{q}\times\boldsymbol{\epsilon} B = i q × ϵ , and L i n t = − e J μ A μ = e J ⋅ ϵ \mathcal{L}_{\mathrm{int}}=-eJ^\mu A_\mu=e\mathbf{J}\cdot\boldsymbol{\epsilon} L int = − e J μ A μ = e J ⋅ ϵ gives the current J = i c X × q \mathbf{J}=i c\mathcal{X}\times\mathbf{q} J = i c X × q . This is the vertex (15) in the rest frame of the pair. Between the channel states of Sec. IIB, the bilinears of L γ ( 1 ) \mathcal{L}_{\gamma}^{(1)} L γ ( 1 ) and L γ ( 2 ) \mathcal{L}_{\gamma}^{(2)} L γ ( 2 ) reduce to spin operators of the light quark in the meson and of the light diquark,
X ( a ) = X ( d ) = 2 σ q , X ( b ) = 2 S q q , \mathcal{X}^{(a)}=\mathcal{X}^{(d)}=2\boldsymbol{\sigma}_q,\qquad\mathcal{X}^{(b)}=2\mathbf{S}_{qq}, X ( a ) = X ( d ) = 2 σ q , X ( b ) = 2 S q q ,
X ( c ) j = − 2 σ q i Q q q i j (46) \mathcal{X}^{(c)j}=-2\sigma_q^iQ_{qq}^{ij} \tag*{(46)} X ( c ) j = − 2 σ q i Q q q ij (46)
Q q q i j = S q q i S q q j + S q q j S q q i − 4 3 δ i j . Q_{qq}^{ij}=S_{qq}^{i}S_{qq}^{j}+S_{qq}^{j}S_{qq}^{i}-\frac{4}{3}\delta^{ij}. Q q q ij = S q q i S q q j + S q q j S q q i − 3 4 δ ij .
up to isospin factors, which are absorbed into the combinations (B4) below. The factor 2 arises from the trace over the spin of the c ˉ \bar{c} c ˉ . In the Condon–Shortley coupling scheme of Sec. II B, Eq. (46) reproduces the spin factors to Ξ ( d ) \Xi^{(d)} Ξ ( d ) to Ξ ( d ) \Xi^{(d)} Ξ ( d ) of tbl:VII , including their signs. In the D ˉ ( ∗ ) Σ c ( ∗ ) \bar{D}^{(*)}\Sigma_c^{(*)} D ˉ ( ∗ ) Σ c ( ∗ ) channels, the magnetic vertex in the Breit frame is therefore
Γ M = 4 m l ′ m l i [ d ˉ ( a ) σ q + d ˉ ( b ) S q q − d ˉ ( c ) σ q ⋅ Q q q ] l l ′ × q \Gamma_M = 4\sqrt{m_l'm_l}\,i\left[\bar{d}^{(a)}\sigma_q+\bar{d}^{(b)}S_{qq}-\bar{d}^{(c)}\sigma_q\cdot Q_{qq}\right]^{l'}_{l}\times q Γ M = 4 m l ′ m l i [ d ˉ ( a ) σ q + d ˉ ( b ) S q q − d ˉ ( c ) σ q ⋅ Q q q ] l l ′ × q
with ( σ q ⋅ Q q q ) j = σ q i Q q q i j (\sigma_q\cdot Q_{qq})^{j}=\sigma_q^{i}Q_{qq}^{ij} ( σ q ⋅ Q q q ) j = σ q i Q q q ij . The operators of L γ ( 3 ) \mathcal{L}_{\gamma}^{(3)} L γ ( 3 ) to L γ ( 6 ) \mathcal{L}_{\gamma}^{(6)} L γ ( 6 ) change the spin of the diquark from one to zero or connect the open-charm and hidden-charm channels. Their spin factors were computed with the same fields and are listed in tbl:VII . In the I = 1 2 I=\frac{1}{2} I = 2 1 , I 3 = + 1 2 I_3=+\frac{1}{2} I 3 = + 2 1 state, the couplings enter through the combinations
d ˉ ( s ) = d 1 ( s ) − 2 d 2 ( s ) − 1 3 d 3 ( s ) + 2 3 d 4 ( s ) + 20 9 d 6 ( s ) , d ˉ ( d ) = d 0 ( d ) + d 1 ( d ) , f ˉ ( ′ ) = − 3 f 1 ( e ) − 1 3 f 2 ( e ) + 2 3 f 3 ( e ) , n ˉ = n 0 + n 1 , (47) \begin{aligned}
\bar{d}^{(s)}&=d_1^{(s)}-2d_2^{(s)}-\frac{1}{3}d_3^{(s)}+\frac{2}{3}d_4^{(s)}+\frac{20}{9}d_6^{(s)},\\
\bar{d}^{(d)}&=d_0^{(d)}+d_1^{(d)},\\
\bar{f}^{(\prime)}&=-\sqrt{3}f_1^{(e)}-\frac{1}{\sqrt{3}}f_2^{(e)}+\frac{2}{\sqrt{3}}f_3^{(e)},\\
\bar{n}&=n_0+n_1, \tag*{(47)}
\end{aligned} d ˉ ( s ) d ˉ ( d ) f ˉ ( ′ ) n ˉ = d 1 ( s ) − 2 d 2 ( s ) − 3 1 d 3 ( s ) + 3 2 d 4 ( s ) + 9 20 d 6 ( s ) , = d 0 ( d ) + d 1 ( d ) , = − 3 f 1 ( e ) − 3 1 f 2 ( e ) + 3 2 f 3 ( e ) , = n 0 + n 1 , (47)
where ( s = a , b , c ) (s=a,b,c) ( s = a , b , c ) and analogously h ˉ ( ′ ) \bar{h}^{(\prime)} h ˉ ( ′ ) and k ˉ = k 0 + k 1 \bar{k}=k_0+k_1 k ˉ = k 0 + k 1 . Since the structure t = 5 t=5 t = 5 has no diagonal matrix element in an isospin eigenstate, d 5 ( e ) d_5^{(e)} d 5 ( e ) does not contribute. The combinations d ˉ ( a , b , c , d ) \bar{d}^{(a,b,c,d)} d ˉ ( a , b , c , d ) , f ˉ \bar{f} f ˉ , f ˉ ′ \bar{f}^{\prime} f ˉ ′ , h ˉ \bar{h} h ˉ , h ˉ ′ \bar{h}^{\prime} h ˉ ′ , k ˉ \bar{k} k ˉ and n ˉ \bar{n} n ˉ are the ten couplings c ˉ s \bar{c}_s c ˉ s of Sec. II C 3. The spin factors are listed in tbl:VII . For J = 5 2 J=\frac{5}{2} J = 2 5 , Ξ ( a ) = Ξ ( b ) = 4 5 \Xi^{(a)}=\Xi^{(b)}=\frac{4}{5} Ξ ( a ) = Ξ ( b ) = 5 4 and Ξ ( c ) = − 8 15 \Xi^{(c)}=-\frac{8}{15} Ξ ( c ) = − 15 8 .
The size of the couplings can be estimated within the vector-exchange model. Since the photon couples to the exchanged ρ \rho ρ only through the isospin structure t = 5 t=5 t = 5 and the leading vector-meson couplings are spin independent, the leading remaining mechanism is the anomalous π V γ \pi V\gamma π V γ exchange current. In this mechanism, one hadron emits a pion, the photon converts it into a ρ \rho ρ or ω \omega ω meson, and the second hadron absorbs the vector meson. We take the π V γ \pi V\gamma π V γ couplings from the experimental radiative widths of the ω \omega ω and the ρ \rho ρ , the pion couplings from Γ ( D ∗ + ) \Gamma(D^{*+}) Γ ( D ∗+ ) and Γ ( Σ c + + → Λ c + π + ) \Gamma(\Sigma_c^{++}\to\Lambda_c^+\pi^+) Γ ( Σ c ++ → Λ c + π + ) and the vector coupling from the normalization of the kernel. The combinations (B4) are then 9 to 12 G e V − 4 \mathrm{GeV}^{-4} GeV − 4 in modulus for d ˉ ( a ) \bar{d}^{(a)} d ˉ ( a ) , d ˉ ( d ) \bar{d}^{(d)} d ˉ ( d ) and f ˉ \bar{f} f ˉ and below 1 G e V − 4 \mathrm{GeV}^{-4} GeV − 4 for d ˉ ( b ) \bar{d}^{(b)} d ˉ ( b ) . Couplings of this size change the magnetic moments by at most 0.07 μ N \mu_N μ N . Since the exchange current has the range of the pion and the relative signs of the couplings were not determined, this provides only an order-of-magnitude estimate.
The electric operators of Eq. (17) contain a two-hadron operator X \mathcal{X} X that is a Hermitian scalar in the light-quark spins and even under time reversal. For the D ˉ ( ∗ ) Σ c ( ∗ ) \bar{D}^{(*)}\Sigma_c^{(*)} D ˉ ( ∗ ) Σ c ( ∗ ) channels, the scalar spin structures are O H ˉ 0 O S 0 \mathcal{O}_{\bar{H}}^{0}\mathcal{O}_{S}^{0} O H ˉ 0 O S 0 and O H ˉ j O S j \mathcal{O}_{\bar{H}}^{j}\mathcal{O}_{S}^{j} O H ˉ j O S j , which are equivalent to one coupling for each light-quark spin, as in the case of the kernel. Combined with the five even isospin structures, they give 10 operators. With one scalar spin structure for each of the other pairs of channel types, the total number is 10 + 2 + 3 + 3 + 2 + 2 = 22 10+2+3+3+2+2=22 10 + 2 + 3 + 3 + 2 + 2 = 22 . In an I = 1 2 I=\frac{1}{2} I = 2 1 state, each scalar spin structure of each pair of channel types contributes through one combination, which yields the seven couplings of the matrix E \mathcal{E} E . Between the channel states, the two scalar structures reduce to 2 and 2 σ q ⋅ S q q 2\sigma_q\cdot S_{qq} 2 σ q ⋅ S q q . With ∇ ⋅ E = v ν ∂ μ F μ ν \nabla\cdot E=v_\nu\partial_\mu F^{\mu\nu} ∇ ⋅ E = v ν ∂ μ F μν and ∂ μ → − i q μ \partial_\mu\to-iq_\mu ∂ μ → − i q μ for the incoming photon, we have ∂ μ F μ ν → − q 2 ϵ ν + q ν ( q ⋅ ϵ ) \partial_\mu F^{\mu\nu}\to-q^2\epsilon^\nu+q^\nu(q\cdot\epsilon) ∂ μ F μν → − q 2 ϵ ν + q ν ( q ⋅ ϵ ) , and L i n t = − e J μ A μ \mathcal{L}_{\mathrm{int}}=-eJ^\mu A_\mu L int = − e J μ A μ gives J μ = d [ q 2 v μ − q μ ( q ⋅ v ) ] X J^\mu=d\left[q^2v^\mu-q^\mu(q\cdot v)\right]\mathcal{X} J μ = d [ q 2 v μ − q μ ( q ⋅ v ) ] X , which is Eq. (18) . In the Breit frame at s = s ′ s=s' s = s ′ , the charge density is − d Q 2 X -dQ^2\mathcal{X} − d Q 2 X , and ⟨ r 2 ⟩ = − 6 d F / d Q 2 \langle r^2\rangle=-6\,dF/dQ^2 ⟨ r 2 ⟩ = − 6 d F / d Q 2 leads to Eq. (22) .
One-body magnetic matrix elements
tbl:VIII lists the reduced matrix elements (11) of the one-body magnetic operator between the channel states, separated into the meson-line and baryon-line contributions. They are computed with the spin wave functions of Sec. II B and the moments of tbl:II . For two spins j 1 j_1 j 1 and j 2 j_2 j 2 coupled to J J J , ⟨ j 1 ⟩ = [ J ( J + 1 ) + j 1 ( j 1 + 1 ) − j 2 ( j 2 + 1 ) ] / [ 2 J ( J + 1 ) ] J \langle j_1\rangle=[J(J+1)+j_1(j_1+1)-j_2(j_2+1)]/[2J(J+1)]\mathcal{J} ⟨ j 1 ⟩ = [ J ( J + 1 ) + j 1 ( j 1 + 1 ) − j 2 ( j 2 + 1 )] / [ 2 J ( J + 1 )] J , which yields the additive moments on the diagonal. In the Ξ c ( ∗ ) \Xi_c^{(*)} Ξ c ( ∗ ) channels, the meson moments are averaged with the isospin weights 2 3 \frac{2}{3} 3 2 and 1 3 \frac{1}{3} 3 1 , while in the Λ c \Lambda_c Λ c channels the meson is a D ˉ ( ∗ ) 0 \bar{D}^{(*)0} D ˉ ( ∗ ) 0 . In the hidden-charm channels, the baryon line contributes μ p \mu_p μ p for η c N \eta_cN η c N with J = 1 2 J=\frac{1}{2} J = 2 1 and for J / ψ N J/\psi N J / ψ N with J = 3 2 J=\frac{3}{2} J = 2 3 , and − μ p / 3 -\mu_p/3 − μ p /3 for J / ψ N J/\psi N J / ψ N with J = 1 2 J=\frac{1}{2} J = 2 1 , whereas the meson contribution is 0.482 μ N \mu_N μ N to the transition between η c N \eta_cN η c N and J / ψ N J/\psi N J / ψ N .
The transition moments of the mesons can be compared with experimental data. With Γ ( V → P γ ) = ( α / 3 ) ( k 3 / M p 2 ) ∣ μ V P / μ N ∣ 2 \Gamma(V\to P\gamma)=(\alpha/3)(k^3/M_p^2)|\mu_{VP}/\mu_N|^2 Γ ( V → P γ ) = ( α /3 ) ( k 3 / M p 2 ) ∣ μ V P / μ N ∣ 2 , where k k k is the photon momentum, the experimental widths [41 ] yield ∣ μ ( D ∗ + → D + γ ) ∣ = 0.44 ± 0.06 μ N |\mu(D^{*+}\to D^+\gamma)|=0.44\pm0.06\,\mu_N ∣ μ ( D ∗+ → D + γ ) ∣ = 0.44 ± 0.06 μ N and ∣ μ ( J / ψ → η c γ ) ∣ = 0.67 ± 0.03 μ N |\mu(J/\psi\to\eta_c\gamma)|=0.67\pm0.03\,\mu_N ∣ μ ( J / ψ → η c γ ) ∣ = 0.67 ± 0.03 μ N , which are 14 and 20% below the constituent-quark values 0.51 and 0.83 μ N \mu_N μ N , respectively.
TABLE VIII. Isospin-averaged one-body magnetic matrix elements M l ′ l ( 1 ) ( J ) \mathcal{M}^{(1)}_{l'l}(J) M l ′ l ( 1 ) ( J ) in μ N \mu_N μ N between the open-charm channel states of the P c + P_c^{+} P c + . Each entry gives the meson-line part M M \mathcal{M}^{\mathrm{M}} M M / the baryon-line part M B \mathcal{M}^{\mathrm{B}} M B .
Table VIII.